1
// Copyright (c) 2012- PPSSPP Project.
3
// This program is free software: you can redistribute it and/or modify
4
// it under the terms of the GNU General Public License as published by
5
// the Free Software Foundation, version 2.0 or later versions.
7
// This program is distributed in the hope that it will be useful,
8
// but WITHOUT ANY WARRANTY; without even the implied warranty of
9
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
10
// GNU General Public License 2.0 for more details.
12
// A copy of the GPL 2.0 should have been included with the program.
13
// If not, see http://www.gnu.org/licenses/
15
// Official git repository and contact information can be found at
16
// https://github.com/hrydgard/ppsspp and http://www.ppsspp.org/.
20
#include "Common/ChunkFile.h"
21
#include "Common/GraphicsContext.h"
22
#include "base/NativeApp.h"
23
#include "base/logging.h"
24
#include "profiler/profiler.h"
25
#include "Core/Debugger/Breakpoints.h"
26
#include "Core/MemMapHelpers.h"
27
#include "Core/MIPS/MIPS.h"
28
#include "Core/Host.h"
29
#include "Core/Config.h"
30
#include "Core/Reporting.h"
31
#include "Core/System.h"
33
#include "helper/dx_state.h"
35
#include "GPU/GPUState.h"
36
#include "GPU/ge_constants.h"
37
#include "GPU/GeDisasm.h"
39
#include "GPU/Common/FramebufferCommon.h"
40
#include "GPU/Directx9/helper/global.h"
41
#include "GPU/Directx9/ShaderManagerDX9.h"
42
#include "GPU/Directx9/GPU_DX9.h"
43
#include "GPU/Directx9/FramebufferDX9.h"
44
#include "GPU/Directx9/DrawEngineDX9.h"
45
#include "GPU/Directx9/TextureCacheDX9.h"
47
#include "Core/HLE/sceKernelThread.h"
48
#include "Core/HLE/sceKernelInterrupt.h"
49
#include "Core/HLE/sceGe.h"
55
FLAG_FLUSHBEFOREONCHANGE = 2,
56
FLAG_EXECUTE = 4, // needs to actually be executed. unused for now.
57
FLAG_EXECUTEONCHANGE = 8,
58
FLAG_ANY_EXECUTE = 4 | 8,
63
struct CommandTableEntry {
66
GPU_DX9::CmdFunc func;
69
static const CommandTableEntry commandTable[] = {
70
// Changes that dirty the framebuffer
71
{GE_CMD_FRAMEBUFPTR, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, &GPU_DX9::Execute_FramebufType},
72
{GE_CMD_FRAMEBUFWIDTH, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, &GPU_DX9::Execute_FramebufType},
73
{GE_CMD_FRAMEBUFPIXFORMAT, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, &GPU_DX9::Execute_FramebufType},
74
{GE_CMD_ZBUFPTR, FLAG_FLUSHBEFOREONCHANGE},
75
{GE_CMD_ZBUFWIDTH, FLAG_FLUSHBEFOREONCHANGE},
77
// Changes that dirty uniforms
78
{GE_CMD_FOGCOLOR, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE},
79
{GE_CMD_FOG1, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE},
80
{GE_CMD_FOG2, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE},
82
// Should these maybe flush?
83
{GE_CMD_MINZ, FLAG_FLUSHBEFOREONCHANGE},
84
{GE_CMD_MAXZ, FLAG_FLUSHBEFOREONCHANGE},
86
// Changes that dirty texture scaling.
87
{ GE_CMD_TEXMAPMODE, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, &GPU_DX9::Execute_TexMapMode },
88
{ GE_CMD_TEXSCALEU, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, &GPU_DX9::Execute_TexScaleU },
89
{ GE_CMD_TEXSCALEV, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, &GPU_DX9::Execute_TexScaleV },
90
{ GE_CMD_TEXOFFSETU, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, &GPU_DX9::Execute_TexOffsetU },
91
{ GE_CMD_TEXOFFSETV, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, &GPU_DX9::Execute_TexOffsetV },
93
// Changes that dirty the current texture. Really should be possible to avoid executing these if we compile
94
// by adding some more flags.
95
{GE_CMD_TEXSIZE0, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTE, &GPU_DX9::Execute_TexSize0},
96
{GE_CMD_TEXSIZE1, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, &GPU_DX9::Execute_TexSizeN},
97
{GE_CMD_TEXSIZE2, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, &GPU_DX9::Execute_TexSizeN},
98
{GE_CMD_TEXSIZE3, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, &GPU_DX9::Execute_TexSizeN},
99
{GE_CMD_TEXSIZE4, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, &GPU_DX9::Execute_TexSizeN},
100
{GE_CMD_TEXSIZE5, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, &GPU_DX9::Execute_TexSizeN},
101
{GE_CMD_TEXSIZE6, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, &GPU_DX9::Execute_TexSizeN},
102
{GE_CMD_TEXSIZE7, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, &GPU_DX9::Execute_TexSizeN},
103
{GE_CMD_TEXFORMAT, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, &GPU_DX9::Execute_TexFormat},
104
{GE_CMD_TEXLEVEL, FLAG_EXECUTE, &GPU_DX9::Execute_TexLevel},
105
{GE_CMD_TEXADDR0, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, &GPU_DX9::Execute_TexAddr0},
106
{GE_CMD_TEXADDR1, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, &GPU_DX9::Execute_TexAddrN},
107
{GE_CMD_TEXADDR2, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, &GPU_DX9::Execute_TexAddrN},
108
{GE_CMD_TEXADDR3, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, &GPU_DX9::Execute_TexAddrN},
109
{GE_CMD_TEXADDR4, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, &GPU_DX9::Execute_TexAddrN},
110
{GE_CMD_TEXADDR5, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, &GPU_DX9::Execute_TexAddrN},
111
{GE_CMD_TEXADDR6, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, &GPU_DX9::Execute_TexAddrN},
112
{GE_CMD_TEXADDR7, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, &GPU_DX9::Execute_TexAddrN},
113
{GE_CMD_TEXBUFWIDTH0, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, &GPU_DX9::Execute_TexBufw0},
114
{GE_CMD_TEXBUFWIDTH1, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, &GPU_DX9::Execute_TexBufwN},
115
{GE_CMD_TEXBUFWIDTH2, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, &GPU_DX9::Execute_TexBufwN},
116
{GE_CMD_TEXBUFWIDTH3, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, &GPU_DX9::Execute_TexBufwN},
117
{GE_CMD_TEXBUFWIDTH4, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, &GPU_DX9::Execute_TexBufwN},
118
{GE_CMD_TEXBUFWIDTH5, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, &GPU_DX9::Execute_TexBufwN},
119
{GE_CMD_TEXBUFWIDTH6, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, &GPU_DX9::Execute_TexBufwN},
120
{GE_CMD_TEXBUFWIDTH7, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, &GPU_DX9::Execute_TexBufwN},
121
{GE_CMD_CLUTADDR, FLAG_FLUSHBEFOREONCHANGE},
122
{GE_CMD_CLUTADDRUPPER, FLAG_FLUSHBEFOREONCHANGE},
123
{GE_CMD_CLUTFORMAT, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, &GPU_DX9::Execute_ClutFormat},
125
// These affect the fragment shader so need flushing.
126
{GE_CMD_CLEARMODE, FLAG_FLUSHBEFOREONCHANGE},
127
{GE_CMD_TEXTUREMAPENABLE, FLAG_FLUSHBEFOREONCHANGE},
128
{GE_CMD_FOGENABLE, FLAG_FLUSHBEFOREONCHANGE},
129
{GE_CMD_TEXMODE, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, &GPU_DX9::Execute_TexParamType},
130
{GE_CMD_TEXSHADELS, FLAG_FLUSHBEFOREONCHANGE},
131
{GE_CMD_SHADEMODE, FLAG_FLUSHBEFOREONCHANGE},
132
{GE_CMD_TEXFUNC, FLAG_FLUSHBEFOREONCHANGE},
133
{GE_CMD_COLORTEST, FLAG_FLUSHBEFOREONCHANGE},
134
{GE_CMD_ALPHATESTENABLE, FLAG_FLUSHBEFOREONCHANGE},
135
{GE_CMD_COLORTESTENABLE, FLAG_FLUSHBEFOREONCHANGE},
136
{GE_CMD_COLORTESTMASK, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, &GPU_DX9::Execute_ColorTestMask},
138
// These change the vertex shader so need flushing.
139
{GE_CMD_REVERSENORMAL, FLAG_FLUSHBEFOREONCHANGE},
140
{GE_CMD_LIGHTINGENABLE, FLAG_FLUSHBEFOREONCHANGE},
141
{GE_CMD_LIGHTENABLE0, FLAG_FLUSHBEFOREONCHANGE},
142
{GE_CMD_LIGHTENABLE1, FLAG_FLUSHBEFOREONCHANGE},
143
{GE_CMD_LIGHTENABLE2, FLAG_FLUSHBEFOREONCHANGE},
144
{GE_CMD_LIGHTENABLE3, FLAG_FLUSHBEFOREONCHANGE},
145
{GE_CMD_LIGHTTYPE0, FLAG_FLUSHBEFOREONCHANGE},
146
{GE_CMD_LIGHTTYPE1, FLAG_FLUSHBEFOREONCHANGE},
147
{GE_CMD_LIGHTTYPE2, FLAG_FLUSHBEFOREONCHANGE},
148
{GE_CMD_LIGHTTYPE3, FLAG_FLUSHBEFOREONCHANGE},
149
{GE_CMD_MATERIALUPDATE, FLAG_FLUSHBEFOREONCHANGE},
151
// This changes both shaders so need flushing.
152
{GE_CMD_LIGHTMODE, FLAG_FLUSHBEFOREONCHANGE},
153
{GE_CMD_TEXFILTER, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, &GPU_DX9::Execute_TexParamType},
154
{GE_CMD_TEXWRAP, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, &GPU_DX9::Execute_TexParamType},
157
{GE_CMD_ALPHATEST, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, &GPU_DX9::Execute_AlphaTest},
158
{GE_CMD_COLORREF, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, &GPU_DX9::Execute_ColorRef},
159
{GE_CMD_TEXENVCOLOR, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, &GPU_DX9::Execute_TexEnvColor},
161
// Simple render state changes. Handled in StateMapping.cpp.
162
{GE_CMD_OFFSETX, FLAG_FLUSHBEFOREONCHANGE},
163
{GE_CMD_OFFSETY, FLAG_FLUSHBEFOREONCHANGE},
164
{GE_CMD_CULL, FLAG_FLUSHBEFOREONCHANGE},
165
{GE_CMD_CULLFACEENABLE, FLAG_FLUSHBEFOREONCHANGE},
166
{GE_CMD_DITHERENABLE, FLAG_FLUSHBEFOREONCHANGE},
167
{GE_CMD_STENCILOP, FLAG_FLUSHBEFOREONCHANGE},
168
{GE_CMD_STENCILTEST, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, &GPU_DX9::Execute_StencilTest},
169
{GE_CMD_STENCILTESTENABLE, FLAG_FLUSHBEFOREONCHANGE},
170
{GE_CMD_ALPHABLENDENABLE, FLAG_FLUSHBEFOREONCHANGE},
171
{GE_CMD_BLENDMODE, FLAG_FLUSHBEFOREONCHANGE},
172
{GE_CMD_BLENDFIXEDA, FLAG_FLUSHBEFOREONCHANGE},
173
{GE_CMD_BLENDFIXEDB, FLAG_FLUSHBEFOREONCHANGE},
174
{GE_CMD_MASKRGB, FLAG_FLUSHBEFOREONCHANGE},
175
{GE_CMD_MASKALPHA, FLAG_FLUSHBEFOREONCHANGE},
176
{GE_CMD_ZTEST, FLAG_FLUSHBEFOREONCHANGE},
177
{GE_CMD_ZTESTENABLE, FLAG_FLUSHBEFOREONCHANGE},
178
{GE_CMD_ZWRITEDISABLE, FLAG_FLUSHBEFOREONCHANGE},
180
// These can't be emulated in D3D (except a few special cases)
181
{GE_CMD_LOGICOP, FLAG_FLUSHBEFOREONCHANGE},
182
{GE_CMD_LOGICOPENABLE, FLAG_FLUSHBEFOREONCHANGE},
184
// Can probably ignore this one as we don't support AA lines.
185
{GE_CMD_ANTIALIASENABLE, FLAG_FLUSHBEFOREONCHANGE},
187
// Morph weights. TODO: Remove precomputation?
188
{GE_CMD_MORPHWEIGHT0, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE},
189
{GE_CMD_MORPHWEIGHT1, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE},
190
{GE_CMD_MORPHWEIGHT2, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE},
191
{GE_CMD_MORPHWEIGHT3, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE},
192
{GE_CMD_MORPHWEIGHT4, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE},
193
{GE_CMD_MORPHWEIGHT5, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE},
194
{GE_CMD_MORPHWEIGHT6, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE},
195
{GE_CMD_MORPHWEIGHT7, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE},
197
// Control spline/bezier patches. Don't really require flushing as such, but meh.
198
{GE_CMD_PATCHDIVISION, FLAG_FLUSHBEFOREONCHANGE},
199
{GE_CMD_PATCHPRIMITIVE, FLAG_FLUSHBEFOREONCHANGE},
200
{GE_CMD_PATCHFACING, FLAG_FLUSHBEFOREONCHANGE},
201
{GE_CMD_PATCHCULLENABLE, FLAG_FLUSHBEFOREONCHANGE},
204
{GE_CMD_VIEWPORTXSCALE, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, &GPU_DX9::Execute_ViewportType},
205
{GE_CMD_VIEWPORTYSCALE, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, &GPU_DX9::Execute_ViewportType},
206
{GE_CMD_VIEWPORTZSCALE, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, &GPU_DX9::Execute_ViewportType},
207
{GE_CMD_VIEWPORTXCENTER, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, &GPU_DX9::Execute_ViewportType},
208
{GE_CMD_VIEWPORTYCENTER, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, &GPU_DX9::Execute_ViewportType},
209
{GE_CMD_VIEWPORTZCENTER, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, &GPU_DX9::Execute_ViewportType},
210
{GE_CMD_CLIPENABLE, FLAG_FLUSHBEFOREONCHANGE},
213
{GE_CMD_REGION1, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, &GPU_DX9::Execute_Region},
214
{GE_CMD_REGION2, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, &GPU_DX9::Execute_Region},
217
{GE_CMD_SCISSOR1, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, &GPU_DX9::Execute_Scissor},
218
{GE_CMD_SCISSOR2, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, &GPU_DX9::Execute_Scissor},
220
// These dirty various vertex shader uniforms. Could embed information about that in this table and call dirtyuniform directly, hm...
221
{GE_CMD_AMBIENTCOLOR, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE},
222
{GE_CMD_AMBIENTALPHA, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE},
223
{GE_CMD_MATERIALDIFFUSE, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE},
224
{GE_CMD_MATERIALEMISSIVE, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE},
225
{GE_CMD_MATERIALAMBIENT, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE},
226
{GE_CMD_MATERIALALPHA, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE},
227
{GE_CMD_MATERIALSPECULAR, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE},
228
{GE_CMD_MATERIALSPECULARCOEF, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE},
230
// These precompute a value. not sure if worth it. Also dirty uniforms, which could be table-ized to avoid execute.
231
{GE_CMD_LX0, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, &GPU_DX9::Execute_Light0Param},
232
{GE_CMD_LY0, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, &GPU_DX9::Execute_Light0Param},
233
{GE_CMD_LZ0, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, &GPU_DX9::Execute_Light0Param},
234
{GE_CMD_LX1, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, &GPU_DX9::Execute_Light1Param},
235
{GE_CMD_LY1, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, &GPU_DX9::Execute_Light1Param},
236
{GE_CMD_LZ1, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, &GPU_DX9::Execute_Light1Param},
237
{GE_CMD_LX2, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, &GPU_DX9::Execute_Light2Param},
238
{GE_CMD_LY2, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, &GPU_DX9::Execute_Light2Param},
239
{GE_CMD_LZ2, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, &GPU_DX9::Execute_Light2Param},
240
{GE_CMD_LX3, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, &GPU_DX9::Execute_Light3Param},
241
{GE_CMD_LY3, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, &GPU_DX9::Execute_Light3Param},
242
{GE_CMD_LZ3, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, &GPU_DX9::Execute_Light3Param},
244
{GE_CMD_LDX0, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, &GPU_DX9::Execute_Light0Param},
245
{GE_CMD_LDY0, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, &GPU_DX9::Execute_Light0Param},
246
{GE_CMD_LDZ0, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, &GPU_DX9::Execute_Light0Param},
247
{GE_CMD_LDX1, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, &GPU_DX9::Execute_Light1Param},
248
{GE_CMD_LDY1, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, &GPU_DX9::Execute_Light1Param},
249
{GE_CMD_LDZ1, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, &GPU_DX9::Execute_Light1Param},
250
{GE_CMD_LDX2, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, &GPU_DX9::Execute_Light2Param},
251
{GE_CMD_LDY2, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, &GPU_DX9::Execute_Light2Param},
252
{GE_CMD_LDZ2, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, &GPU_DX9::Execute_Light2Param},
253
{GE_CMD_LDX3, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, &GPU_DX9::Execute_Light3Param},
254
{GE_CMD_LDY3, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, &GPU_DX9::Execute_Light3Param},
255
{GE_CMD_LDZ3, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, &GPU_DX9::Execute_Light3Param},
257
{GE_CMD_LKA0, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, &GPU_DX9::Execute_Light0Param},
258
{GE_CMD_LKB0, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, &GPU_DX9::Execute_Light0Param},
259
{GE_CMD_LKC0, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, &GPU_DX9::Execute_Light0Param},
260
{GE_CMD_LKA1, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, &GPU_DX9::Execute_Light1Param},
261
{GE_CMD_LKB1, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, &GPU_DX9::Execute_Light1Param},
262
{GE_CMD_LKC1, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, &GPU_DX9::Execute_Light1Param},
263
{GE_CMD_LKA2, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, &GPU_DX9::Execute_Light2Param},
264
{GE_CMD_LKB2, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, &GPU_DX9::Execute_Light2Param},
265
{GE_CMD_LKC2, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, &GPU_DX9::Execute_Light2Param},
266
{GE_CMD_LKA3, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, &GPU_DX9::Execute_Light3Param},
267
{GE_CMD_LKB3, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, &GPU_DX9::Execute_Light3Param},
268
{GE_CMD_LKC3, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, &GPU_DX9::Execute_Light3Param},
270
{GE_CMD_LKS0, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, &GPU_DX9::Execute_Light0Param},
271
{GE_CMD_LKS1, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, &GPU_DX9::Execute_Light1Param},
272
{GE_CMD_LKS2, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, &GPU_DX9::Execute_Light2Param},
273
{GE_CMD_LKS3, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, &GPU_DX9::Execute_Light3Param},
275
{GE_CMD_LKO0, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, &GPU_DX9::Execute_Light0Param},
276
{GE_CMD_LKO1, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, &GPU_DX9::Execute_Light1Param},
277
{GE_CMD_LKO2, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, &GPU_DX9::Execute_Light2Param},
278
{GE_CMD_LKO3, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, &GPU_DX9::Execute_Light3Param},
280
{GE_CMD_LAC0, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, &GPU_DX9::Execute_Light0Param},
281
{GE_CMD_LDC0, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, &GPU_DX9::Execute_Light0Param},
282
{GE_CMD_LSC0, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, &GPU_DX9::Execute_Light0Param},
283
{GE_CMD_LAC1, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, &GPU_DX9::Execute_Light1Param},
284
{GE_CMD_LDC1, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, &GPU_DX9::Execute_Light1Param},
285
{GE_CMD_LSC1, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, &GPU_DX9::Execute_Light1Param},
286
{GE_CMD_LAC2, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, &GPU_DX9::Execute_Light2Param},
287
{GE_CMD_LDC2, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, &GPU_DX9::Execute_Light2Param},
288
{GE_CMD_LSC2, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, &GPU_DX9::Execute_Light2Param},
289
{GE_CMD_LAC3, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, &GPU_DX9::Execute_Light3Param},
290
{GE_CMD_LDC3, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, &GPU_DX9::Execute_Light3Param},
291
{GE_CMD_LSC3, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, &GPU_DX9::Execute_Light3Param},
294
{GE_CMD_TEXFLUSH, 0},
295
{GE_CMD_TEXLODSLOPE, 0},
298
// These are just nop or part of other later commands.
301
{GE_CMD_TRANSFERSRC, 0},
302
{GE_CMD_TRANSFERSRCW, 0},
303
{GE_CMD_TRANSFERDST, 0},
304
{GE_CMD_TRANSFERDSTW, 0},
305
{GE_CMD_TRANSFERSRCPOS, 0},
306
{GE_CMD_TRANSFERDSTPOS, 0},
307
{GE_CMD_TRANSFERSIZE, 0},
309
// From Common. No flushing but definitely need execute.
310
{GE_CMD_OFFSETADDR, FLAG_EXECUTE},
311
{GE_CMD_ORIGIN, FLAG_EXECUTE | FLAG_READS_PC}, // Really?
312
{GE_CMD_PRIM, FLAG_EXECUTE},
313
{GE_CMD_JUMP, FLAG_EXECUTE | FLAG_READS_PC | FLAG_WRITES_PC},
314
{GE_CMD_CALL, FLAG_EXECUTE | FLAG_READS_PC | FLAG_WRITES_PC},
315
{GE_CMD_RET, FLAG_EXECUTE | FLAG_READS_PC | FLAG_WRITES_PC},
316
{GE_CMD_END, FLAG_FLUSHBEFORE | FLAG_EXECUTE | FLAG_READS_PC | FLAG_WRITES_PC}, // Flush?
317
{GE_CMD_VADDR, FLAG_EXECUTE, &GPU_DX9::Execute_Vaddr},
318
{GE_CMD_IADDR, FLAG_EXECUTE, &GPU_DX9::Execute_Iaddr},
319
{GE_CMD_BJUMP, FLAG_EXECUTE | FLAG_READS_PC | FLAG_WRITES_PC}, // EXECUTE
320
{GE_CMD_BOUNDINGBOX, FLAG_EXECUTE, &GPU_DX9::Execute_BoundingBox}, // + FLUSHBEFORE when we implement... or not, do we need to?
322
// Changing the vertex type requires us to flush.
323
{GE_CMD_VERTEXTYPE, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, &GPU_DX9::Execute_VertexType},
325
{GE_CMD_BEZIER, FLAG_FLUSHBEFORE | FLAG_EXECUTE, &GPU_DX9::Execute_Bezier},
326
{GE_CMD_SPLINE, FLAG_FLUSHBEFORE | FLAG_EXECUTE, &GPU_DX9::Execute_Spline},
328
// These two are actually processed in CMD_END.
329
{GE_CMD_SIGNAL, FLAG_FLUSHBEFORE},
330
{GE_CMD_FINISH, FLAG_FLUSHBEFORE},
332
// Changes that trigger data copies. Only flushing on change for LOADCLUT must be a bit of a hack...
333
{GE_CMD_LOADCLUT, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTE, &GPU_DX9::Execute_LoadClut},
334
{GE_CMD_TRANSFERSTART, FLAG_FLUSHBEFORE | FLAG_EXECUTE | FLAG_READS_PC},
336
// We don't use the dither table.
342
// These handle their own flushing.
343
{GE_CMD_WORLDMATRIXNUMBER, FLAG_EXECUTE | FLAG_READS_PC | FLAG_WRITES_PC, &GPU_DX9::Execute_WorldMtxNum},
344
{GE_CMD_WORLDMATRIXDATA, FLAG_EXECUTE, &GPU_DX9::Execute_WorldMtxData},
345
{GE_CMD_VIEWMATRIXNUMBER, FLAG_EXECUTE | FLAG_READS_PC | FLAG_WRITES_PC, &GPU_DX9::Execute_ViewMtxNum},
346
{GE_CMD_VIEWMATRIXDATA, FLAG_EXECUTE, &GPU_DX9::Execute_ViewMtxData},
347
{GE_CMD_PROJMATRIXNUMBER, FLAG_EXECUTE | FLAG_READS_PC | FLAG_WRITES_PC, &GPU_DX9::Execute_ProjMtxNum},
348
{GE_CMD_PROJMATRIXDATA, FLAG_EXECUTE, &GPU_DX9::Execute_ProjMtxData},
349
{GE_CMD_TGENMATRIXNUMBER, FLAG_EXECUTE | FLAG_READS_PC | FLAG_WRITES_PC, &GPU_DX9::Execute_TgenMtxNum},
350
{GE_CMD_TGENMATRIXDATA, FLAG_EXECUTE, &GPU_DX9::Execute_TgenMtxData},
351
{GE_CMD_BONEMATRIXNUMBER, FLAG_EXECUTE | FLAG_READS_PC | FLAG_WRITES_PC, &GPU_DX9::Execute_BoneMtxNum},
352
{GE_CMD_BONEMATRIXDATA, FLAG_EXECUTE, &GPU_DX9::Execute_BoneMtxData},
354
// Vertex Screen/Texture/Color
355
{ GE_CMD_VSCX, FLAG_EXECUTE },
356
{ GE_CMD_VSCY, FLAG_EXECUTE },
357
{ GE_CMD_VSCZ, FLAG_EXECUTE },
358
{ GE_CMD_VTCS, FLAG_EXECUTE },
359
{ GE_CMD_VTCT, FLAG_EXECUTE },
360
{ GE_CMD_VTCQ, FLAG_EXECUTE },
361
{ GE_CMD_VCV, FLAG_EXECUTE },
362
{ GE_CMD_VAP, FLAG_EXECUTE },
363
{ GE_CMD_VFC, FLAG_EXECUTE },
364
{ GE_CMD_VSCV, FLAG_EXECUTE },
366
// "Missing" commands (gaps in the sequence)
367
{GE_CMD_UNKNOWN_03, FLAG_EXECUTE},
368
{GE_CMD_UNKNOWN_0D, FLAG_EXECUTE},
369
{GE_CMD_UNKNOWN_11, FLAG_EXECUTE},
370
{GE_CMD_UNKNOWN_29, FLAG_EXECUTE},
371
{GE_CMD_UNKNOWN_34, FLAG_EXECUTE},
372
{GE_CMD_UNKNOWN_35, FLAG_EXECUTE},
373
{GE_CMD_UNKNOWN_39, FLAG_EXECUTE},
374
{GE_CMD_UNKNOWN_4E, FLAG_EXECUTE},
375
{GE_CMD_UNKNOWN_4F, FLAG_EXECUTE},
376
{GE_CMD_UNKNOWN_52, FLAG_EXECUTE},
377
{GE_CMD_UNKNOWN_59, FLAG_EXECUTE},
378
{GE_CMD_UNKNOWN_5A, FLAG_EXECUTE},
379
{GE_CMD_UNKNOWN_B6, FLAG_EXECUTE},
380
{GE_CMD_UNKNOWN_B7, FLAG_EXECUTE},
381
{GE_CMD_UNKNOWN_D1, FLAG_EXECUTE},
382
{GE_CMD_UNKNOWN_ED, FLAG_EXECUTE},
383
{GE_CMD_UNKNOWN_EF, FLAG_EXECUTE},
384
{GE_CMD_UNKNOWN_FA, FLAG_EXECUTE},
385
{GE_CMD_UNKNOWN_FB, FLAG_EXECUTE},
386
{GE_CMD_UNKNOWN_FC, FLAG_EXECUTE},
387
{GE_CMD_UNKNOWN_FD, FLAG_EXECUTE},
388
{GE_CMD_UNKNOWN_FE, FLAG_EXECUTE},
389
// Appears to be debugging related or something? Hit a lot in GoW.
390
{ GE_CMD_UNKNOWN_FF, 0 },
393
GPU_DX9::CommandInfo GPU_DX9::cmdInfo_[256];
395
GPU_DX9::GPU_DX9(GraphicsContext *gfxCtx)
396
: resized_(false), gfxCtx_(gfxCtx) {
397
lastVsync_ = g_Config.bVSync ? 1 : 0;
398
dxstate.SetVSyncInterval(g_Config.bVSync);
400
shaderManager_ = new ShaderManagerDX9();
401
drawEngine_.SetShaderManager(shaderManager_);
402
drawEngine_.SetTextureCache(&textureCache_);
403
drawEngine_.SetFramebufferManager(&framebufferManager_);
404
framebufferManager_.Init();
405
framebufferManager_.SetTextureCache(&textureCache_);
406
framebufferManager_.SetShaderManager(shaderManager_);
407
framebufferManager_.SetTransformDrawEngine(&drawEngine_);
408
textureCache_.SetFramebufferManager(&framebufferManager_);
409
textureCache_.SetDepalShaderCache(&depalShaderCache_);
410
textureCache_.SetShaderManager(shaderManager_);
412
// Sanity check gstate
413
if ((int *)&gstate.transferstart - (int *)&gstate != 0xEA) {
414
ERROR_LOG(G3D, "gstate has drifted out of sync!");
417
// Sanity check cmdInfo_ table - no dupes please
418
std::set<u8> dupeCheck;
419
memset(cmdInfo_, 0, sizeof(cmdInfo_));
420
for (size_t i = 0; i < ARRAY_SIZE(commandTable); i++) {
421
const u8 cmd = commandTable[i].cmd;
422
if (dupeCheck.find(cmd) != dupeCheck.end()) {
423
ERROR_LOG(G3D, "Command table Dupe: %02x (%i)", (int)cmd, (int)cmd);
425
dupeCheck.insert(cmd);
427
cmdInfo_[cmd].flags |= commandTable[i].flags;
428
cmdInfo_[cmd].func = commandTable[i].func;
429
if (!cmdInfo_[cmd].func) {
430
cmdInfo_[cmd].func = &GPU_DX9::Execute_Generic;
433
// Find commands missing from the table.
434
for (int i = 0; i < 0xEF; i++) {
435
if (dupeCheck.find((u8)i) == dupeCheck.end()) {
436
ERROR_LOG(G3D, "Command missing from table: %02x (%i)", i, i);
440
// No need to flush before the tex scale/offset commands if we are baking
441
// the tex scale/offset into the vertices anyway.
444
BuildReportingInfo();
446
// Some of our defaults are different from hw defaults, let's assert them.
447
// We restore each frame anyway, but here is convenient for tests.
449
textureCache_.NotifyConfigChanged();
452
void GPU_DX9::UpdateCmdInfo() {
453
if (g_Config.bPrescaleUV) {
454
cmdInfo_[GE_CMD_TEXSCALEU].flags &= ~FLAG_FLUSHBEFOREONCHANGE;
455
cmdInfo_[GE_CMD_TEXSCALEV].flags &= ~FLAG_FLUSHBEFOREONCHANGE;
456
cmdInfo_[GE_CMD_TEXOFFSETU].flags &= ~FLAG_FLUSHBEFOREONCHANGE;
457
cmdInfo_[GE_CMD_TEXOFFSETV].flags &= ~FLAG_FLUSHBEFOREONCHANGE;
459
cmdInfo_[GE_CMD_TEXSCALEU].flags |= FLAG_FLUSHBEFOREONCHANGE;
460
cmdInfo_[GE_CMD_TEXSCALEV].flags |= FLAG_FLUSHBEFOREONCHANGE;
461
cmdInfo_[GE_CMD_TEXOFFSETU].flags |= FLAG_FLUSHBEFOREONCHANGE;
462
cmdInfo_[GE_CMD_TEXOFFSETV].flags |= FLAG_FLUSHBEFOREONCHANGE;
465
if (g_Config.bSoftwareSkinning) {
466
cmdInfo_[GE_CMD_VERTEXTYPE].flags &= ~FLAG_FLUSHBEFOREONCHANGE;
467
cmdInfo_[GE_CMD_VERTEXTYPE].func = &GPU_DX9::Execute_VertexTypeSkinning;
469
cmdInfo_[GE_CMD_VERTEXTYPE].flags |= FLAG_FLUSHBEFOREONCHANGE;
470
cmdInfo_[GE_CMD_VERTEXTYPE].func = &GPU_DX9::Execute_VertexType;
476
void GPU_DX9::CheckGPUFeatures() {
479
features |= GPU_SUPPORTS_BLEND_MINMAX;
480
features |= GPU_SUPPORTS_TEXTURE_LOD_CONTROL;
481
features |= GPU_PREFER_CPU_DOWNLOAD;
482
features |= GPU_SUPPORTS_ACCURATE_DEPTH;
485
ZeroMemory(&caps, sizeof(caps));
488
result = pD3DdeviceEx->GetDeviceCaps(&caps);
490
result = pD3Ddevice->GetDeviceCaps(&caps);
492
if (FAILED(result)) {
493
WARN_LOG_REPORT(G3D, "Direct3D9: Failed to get the device caps!");
495
if ((caps.RasterCaps & D3DPRASTERCAPS_ANISOTROPY) != 0 && caps.MaxAnisotropy > 1)
496
features |= GPU_SUPPORTS_ANISOTROPY;
497
if ((caps.TextureCaps & (D3DPTEXTURECAPS_NONPOW2CONDITIONAL | D3DPTEXTURECAPS_POW2)) == 0)
498
features |= GPU_SUPPORTS_OES_TEXTURE_NPOT;
501
if (!g_Config.bHighQualityDepth) {
502
features |= GPU_SCALE_DEPTH_FROM_24BIT_TO_16BIT;
503
} else if (PSP_CoreParameter().compat.flags().PixelDepthRounding) {
504
// Assume we always have a 24-bit depth buffer.
505
features |= GPU_SCALE_DEPTH_FROM_24BIT_TO_16BIT;
506
} else if (PSP_CoreParameter().compat.flags().VertexDepthRounding) {
507
features |= GPU_ROUND_DEPTH_TO_16BIT;
510
gstate_c.featureFlags = features;
513
GPU_DX9::~GPU_DX9() {
514
framebufferManager_.DestroyAllFBOs(true);
515
shaderManager_->ClearCache(true);
516
delete shaderManager_;
519
// Needs to be called on GPU thread, not reporting thread.
520
void GPU_DX9::BuildReportingInfo() {
521
Thin3DContext *thin3d = gfxCtx_->CreateThin3DContext();
523
reportingPrimaryInfo_ = thin3d->GetInfoString(T3DInfo::VENDORSTRING);
524
reportingFullInfo_ = reportingPrimaryInfo_ + " - " + System_GetProperty(SYSPROP_GPUDRIVER_VERSION) + " - " + thin3d->GetInfoString(T3DInfo::SHADELANGVERSION);
529
void GPU_DX9::DeviceLost() {
530
// Simply drop all caches and textures.
531
// FBOs appear to survive? Or no?
532
shaderManager_->ClearCache(false);
533
textureCache_.Clear(false);
534
framebufferManager_.DeviceLost();
537
void GPU_DX9::DeviceRestore() {
541
void GPU_DX9::InitClear() {
542
ScheduleEvent(GPU_EVENT_INIT_CLEAR);
545
void GPU_DX9::InitClearInternal() {
546
bool useNonBufferedRendering = g_Config.iRenderingMode == FB_NON_BUFFERED_MODE;
547
if (useNonBufferedRendering) {
548
dxstate.depthWrite.set(true);
549
dxstate.colorMask.set(true, true, true, true);
550
pD3Ddevice->Clear(0, NULL, D3DCLEAR_STENCIL|D3DCLEAR_TARGET|D3DCLEAR_ZBUFFER, D3DCOLOR_XRGB(0, 0, 0), 1.f, 0);
554
void GPU_DX9::DumpNextFrame() {
555
dumpNextFrame_ = true;
558
void GPU_DX9::BeginFrame() {
559
ScheduleEvent(GPU_EVENT_BEGIN_FRAME);
562
void GPU_DX9::ReapplyGfxStateInternal() {
564
GPUCommon::ReapplyGfxStateInternal();
567
void GPU_DX9::BeginFrameInternal() {
570
drawEngine_.Resized();
571
textureCache_.NotifyConfigChanged();
575
// Turn off vsync when unthrottled
576
int desiredVSyncInterval = g_Config.bVSync ? 1 : 0;
577
if ((PSP_CoreParameter().unthrottle) || (PSP_CoreParameter().fpsLimit == 1))
578
desiredVSyncInterval = 0;
579
if (desiredVSyncInterval != lastVsync_) {
580
dxstate.SetVSyncInterval(desiredVSyncInterval);
581
lastVsync_ = desiredVSyncInterval;
584
textureCache_.StartFrame();
585
drawEngine_.DecimateTrackedVertexArrays();
586
depalShaderCache_.Decimate();
587
// fragmentTestCache_.Decimate();
589
if (dumpNextFrame_) {
590
NOTICE_LOG(G3D, "DUMPING THIS FRAME");
591
dumpThisFrame_ = true;
592
dumpNextFrame_ = false;
593
} else if (dumpThisFrame_) {
594
dumpThisFrame_ = false;
596
shaderManager_->DirtyShader();
598
framebufferManager_.BeginFrame();
601
void GPU_DX9::SetDisplayFramebuffer(u32 framebuf, u32 stride, GEBufferFormat format) {
602
host->GPUNotifyDisplay(framebuf, stride, format);
603
framebufferManager_.SetDisplayFramebuffer(framebuf, stride, format);
606
bool GPU_DX9::FramebufferDirty() {
607
// FIXME: Workaround for displaylists sometimes hanging unprocessed. Not yet sure of the cause.
608
if (ThreadEnabled()) {
609
// FIXME: Workaround for displaylists sometimes hanging unprocessed. Not yet sure of the cause.
610
ScheduleEvent(GPU_EVENT_PROCESS_QUEUE);
611
// Allow it to process fully before deciding if it's dirty.
614
VirtualFramebuffer *vfb = framebufferManager_.GetDisplayVFB();
616
bool dirty = vfb->dirtyAfterDisplay;
617
vfb->dirtyAfterDisplay = false;
622
bool GPU_DX9::FramebufferReallyDirty() {
623
// FIXME: Workaround for displaylists sometimes hanging unprocessed. Not yet sure of the cause.
624
if (ThreadEnabled()) {
625
// FIXME: Workaround for displaylists sometimes hanging unprocessed. Not yet sure of the cause.
626
ScheduleEvent(GPU_EVENT_PROCESS_QUEUE);
627
// Allow it to process fully before deciding if it's dirty.
631
VirtualFramebuffer *vfb = framebufferManager_.GetDisplayVFB();
633
bool dirty = vfb->reallyDirtyAfterDisplay;
634
vfb->reallyDirtyAfterDisplay = false;
640
void GPU_DX9::CopyDisplayToOutput() {
641
ScheduleEvent(GPU_EVENT_COPY_DISPLAY_TO_OUTPUT);
644
void GPU_DX9::CopyDisplayToOutputInternal() {
645
dxstate.depthWrite.set(true);
646
dxstate.colorMask.set(true, true, true, true);
650
framebufferManager_.CopyDisplayToOutput();
651
framebufferManager_.EndFrame();
653
// shaderManager_->EndFrame();
654
shaderManager_->DirtyLastShader();
656
gstate_c.textureChanged = TEXCHANGE_UPDATED;
659
// Maybe should write this in ASM...
660
void GPU_DX9::FastRunLoop(DisplayList &list) {
661
PROFILE_THIS_SCOPE("gpuloop");
662
const CommandInfo *cmdInfo = cmdInfo_;
663
for (; downcount > 0; --downcount) {
664
// We know that display list PCs have the upper nibble == 0 - no need to mask the pointer
665
const u32 op = *(const u32 *)(Memory::base + list.pc);
666
const u32 cmd = op >> 24;
667
const CommandInfo info = cmdInfo[cmd];
668
const u8 cmdFlags = info.flags; // If we stashed the cmdFlags in the top bits of the cmdmem, we could get away with one table lookup instead of two
669
const u32 diff = op ^ gstate.cmdmem[cmd];
670
// Inlined CheckFlushOp here to get rid of the dumpThisFrame_ check.
671
if ((cmdFlags & FLAG_FLUSHBEFORE) || (diff && (cmdFlags & FLAG_FLUSHBEFOREONCHANGE))) {
674
gstate.cmdmem[cmd] = op; // TODO: no need to write if diff==0...
675
if ((cmdFlags & FLAG_EXECUTE) || (diff && (cmdFlags & FLAG_EXECUTEONCHANGE))) {
676
(this->*info.func)(op, diff);
682
void GPU_DX9::FinishDeferred() {
683
// This finishes reading any vertex data that is pending.
684
drawEngine_.FinishDeferred();
687
void GPU_DX9::ProcessEvent(GPUEvent ev) {
689
case GPU_EVENT_INIT_CLEAR:
693
case GPU_EVENT_BEGIN_FRAME:
694
BeginFrameInternal();
697
case GPU_EVENT_COPY_DISPLAY_TO_OUTPUT:
698
CopyDisplayToOutputInternal();
701
case GPU_EVENT_INVALIDATE_CACHE:
702
InvalidateCacheInternal(ev.invalidate_cache.addr, ev.invalidate_cache.size, ev.invalidate_cache.type);
705
case GPU_EVENT_FB_MEMCPY:
706
PerformMemoryCopyInternal(ev.fb_memcpy.dst, ev.fb_memcpy.src, ev.fb_memcpy.size);
709
case GPU_EVENT_FB_MEMSET:
710
PerformMemorySetInternal(ev.fb_memset.dst, ev.fb_memset.v, ev.fb_memset.size);
713
case GPU_EVENT_FB_STENCIL_UPLOAD:
714
PerformStencilUploadInternal(ev.fb_stencil_upload.dst, ev.fb_stencil_upload.size);
718
GPUCommon::ProcessEvent(ev);
722
inline void GPU_DX9::CheckFlushOp(int cmd, u32 diff) {
723
const u8 cmdFlags = cmdInfo_[cmd].flags;
724
if ((cmdFlags & FLAG_FLUSHBEFORE) || (diff && (cmdFlags & FLAG_FLUSHBEFOREONCHANGE))) {
725
if (dumpThisFrame_) {
726
NOTICE_LOG(G3D, "================ FLUSH ================");
732
void GPU_DX9::PreExecuteOp(u32 op, u32 diff) {
733
CheckFlushOp(op >> 24, diff);
736
void GPU_DX9::ExecuteOp(u32 op, u32 diff) {
737
const u8 cmd = op >> 24;
738
const CommandInfo info = cmdInfo_[cmd];
739
const u8 cmdFlags = info.flags;
740
if ((cmdFlags & FLAG_EXECUTE) || (diff && (cmdFlags & FLAG_EXECUTEONCHANGE))) {
741
(this->*info.func)(op, diff);
745
void GPU_DX9::Execute_Vaddr(u32 op, u32 diff) {
746
gstate_c.vertexAddr = gstate_c.getRelativeAddress(op & 0x00FFFFFF);
749
void GPU_DX9::Execute_Iaddr(u32 op, u32 diff) {
750
gstate_c.indexAddr = gstate_c.getRelativeAddress(op & 0x00FFFFFF);
753
void GPU_DX9::Execute_VertexType(u32 op, u32 diff) {
754
if (diff & (GE_VTYPE_TC_MASK | GE_VTYPE_THROUGH_MASK)) {
755
shaderManager_->DirtyUniform(DIRTY_UVSCALEOFFSET);
759
void GPU_DX9::Execute_VertexTypeSkinning(u32 op, u32 diff) {
760
// Don't flush when weight count changes, unless morph is enabled.
761
if ((diff & ~GE_VTYPE_WEIGHTCOUNT_MASK) || (op & GE_VTYPE_MORPHCOUNT_MASK) != 0) {
763
gstate.vertType ^= diff;
765
gstate.vertType ^= diff;
766
if (diff & (GE_VTYPE_TC_MASK | GE_VTYPE_THROUGH_MASK))
767
shaderManager_->DirtyUniform(DIRTY_UVSCALEOFFSET);
768
// In this case, we may be doing weights and morphs.
769
// Update any bone matrix uniforms so it uses them correctly.
770
if ((op & GE_VTYPE_MORPHCOUNT_MASK) != 0) {
771
shaderManager_->DirtyUniform(gstate_c.deferredVertTypeDirty);
772
gstate_c.deferredVertTypeDirty = 0;
777
void GPU_DX9::Execute_Prim(u32 op, u32 diff) {
778
// This drives all drawing. All other state we just buffer up, then we apply it only
779
// when it's time to draw. As most PSP games set state redundantly ALL THE TIME, this is a huge optimization.
781
u32 data = op & 0xFFFFFF;
782
u32 count = data & 0xFFFF;
783
// Upper bits are ignored.
784
GEPrimitiveType prim = static_cast<GEPrimitiveType>((data >> 16) & 7);
789
// Discard AA lines as we can't do anything that makes sense with these anyway. The SW plugin might, though.
791
if (gstate.isAntiAliasEnabled()) {
792
// Discard AA lines in DOA
793
if (prim == GE_PRIM_LINE_STRIP)
795
// Discard AA lines in Summon Night 5
796
if ((prim == GE_PRIM_LINES) && gstate.isSkinningEnabled())
800
// This also make skipping drawing very effective.
801
framebufferManager_.SetRenderFrameBuffer(gstate_c.framebufChanged, gstate_c.skipDrawReason);
802
if (gstate_c.skipDrawReason & (SKIPDRAW_SKIPFRAME | SKIPDRAW_NON_DISPLAYED_FB)) {
803
drawEngine_.SetupVertexDecoder(gstate.vertType);
804
// Rough estimate, not sure what's correct.
805
int vertexCost = drawEngine_.EstimatePerVertexCost();
806
cyclesExecuted += vertexCost * count;
810
u32 vertexAddr = gstate_c.vertexAddr;
811
if (!Memory::IsValidAddress(vertexAddr)) {
812
ERROR_LOG_REPORT(G3D, "Bad vertex address %08x!", vertexAddr);
816
void *verts = Memory::GetPointerUnchecked(vertexAddr);
818
u32 vertexType = gstate.vertType;
819
if ((vertexType & GE_VTYPE_IDX_MASK) != GE_VTYPE_IDX_NONE) {
820
u32 indexAddr = gstate_c.indexAddr;
821
if (!Memory::IsValidAddress(indexAddr)) {
822
ERROR_LOG_REPORT(G3D, "Bad index address %08x!", indexAddr);
825
inds = Memory::GetPointerUnchecked(indexAddr);
828
#ifndef MOBILE_DEVICE
829
if (prim > GE_PRIM_RECTANGLES) {
830
ERROR_LOG_REPORT_ONCE(reportPrim, G3D, "Unexpected prim type: %d", prim);
835
drawEngine_.SubmitPrim(verts, inds, prim, count, vertexType, &bytesRead);
837
int vertexCost = drawEngine_.EstimatePerVertexCost() * count;
838
gpuStats.vertexGPUCycles += vertexCost;
839
cyclesExecuted += vertexCost;
841
// After drawing, we advance the vertexAddr (when non indexed) or indexAddr (when indexed).
842
// Some games rely on this, they don't bother reloading VADDR and IADDR.
843
// The VADDR/IADDR registers are NOT updated.
844
AdvanceVerts(vertexType, count, bytesRead);
847
void GPU_DX9::Execute_Bezier(u32 op, u32 diff) {
848
// This also make skipping drawing very effective.
849
framebufferManager_.SetRenderFrameBuffer(gstate_c.framebufChanged, gstate_c.skipDrawReason);
850
if (gstate_c.skipDrawReason & (SKIPDRAW_SKIPFRAME | SKIPDRAW_NON_DISPLAYED_FB)) {
851
// TODO: Should this eat some cycles? Probably yes. Not sure if important.
855
if (!Memory::IsValidAddress(gstate_c.vertexAddr)) {
856
ERROR_LOG_REPORT(G3D, "Bad vertex address %08x!", gstate_c.vertexAddr);
860
void *control_points = Memory::GetPointerUnchecked(gstate_c.vertexAddr);
861
void *indices = NULL;
862
if ((gstate.vertType & GE_VTYPE_IDX_MASK) != GE_VTYPE_IDX_NONE) {
863
if (!Memory::IsValidAddress(gstate_c.indexAddr)) {
864
ERROR_LOG_REPORT(G3D, "Bad index address %08x!", gstate_c.indexAddr);
867
indices = Memory::GetPointerUnchecked(gstate_c.indexAddr);
870
if (gstate.vertType & GE_VTYPE_MORPHCOUNT_MASK) {
871
DEBUG_LOG_REPORT(G3D, "Bezier + morph: %i", (gstate.vertType & GE_VTYPE_MORPHCOUNT_MASK) >> GE_VTYPE_MORPHCOUNT_SHIFT);
873
if (vertTypeIsSkinningEnabled(gstate.vertType)) {
874
DEBUG_LOG_REPORT(G3D, "Bezier + skinning: %i", vertTypeGetNumBoneWeights(gstate.vertType));
877
GEPatchPrimType patchPrim = gstate.getPatchPrimitiveType();
878
int bz_ucount = op & 0xFF;
879
int bz_vcount = (op >> 8) & 0xFF;
880
bool computeNormals = gstate.isLightingEnabled();
881
bool patchFacing = gstate.patchfacing & 1;
883
drawEngine_.SubmitBezier(control_points, indices, gstate.getPatchDivisionU(), gstate.getPatchDivisionV(), bz_ucount, bz_vcount, patchPrim, computeNormals, patchFacing, gstate.vertType, &bytesRead);
885
// After drawing, we advance pointers - see SubmitPrim which does the same.
886
int count = bz_ucount * bz_vcount;
887
AdvanceVerts(gstate.vertType, count, bytesRead);
890
void GPU_DX9::Execute_Spline(u32 op, u32 diff) {
891
// This also make skipping drawing very effective.
892
framebufferManager_.SetRenderFrameBuffer(gstate_c.framebufChanged, gstate_c.skipDrawReason);
893
if (gstate_c.skipDrawReason & (SKIPDRAW_SKIPFRAME | SKIPDRAW_NON_DISPLAYED_FB)) {
894
// TODO: Should this eat some cycles? Probably yes. Not sure if important.
898
if (!Memory::IsValidAddress(gstate_c.vertexAddr)) {
899
ERROR_LOG_REPORT(G3D, "Bad vertex address %08x!", gstate_c.vertexAddr);
903
void *control_points = Memory::GetPointerUnchecked(gstate_c.vertexAddr);
904
void *indices = NULL;
905
if ((gstate.vertType & GE_VTYPE_IDX_MASK) != GE_VTYPE_IDX_NONE) {
906
if (!Memory::IsValidAddress(gstate_c.indexAddr)) {
907
ERROR_LOG_REPORT(G3D, "Bad index address %08x!", gstate_c.indexAddr);
910
indices = Memory::GetPointerUnchecked(gstate_c.indexAddr);
913
if (gstate.vertType & GE_VTYPE_MORPHCOUNT_MASK) {
914
DEBUG_LOG_REPORT(G3D, "Spline + morph: %i", (gstate.vertType & GE_VTYPE_MORPHCOUNT_MASK) >> GE_VTYPE_MORPHCOUNT_SHIFT);
916
if (vertTypeIsSkinningEnabled(gstate.vertType)) {
917
DEBUG_LOG_REPORT(G3D, "Spline + skinning: %i", vertTypeGetNumBoneWeights(gstate.vertType));
920
int sp_ucount = op & 0xFF;
921
int sp_vcount = (op >> 8) & 0xFF;
922
int sp_utype = (op >> 16) & 0x3;
923
int sp_vtype = (op >> 18) & 0x3;
924
GEPatchPrimType patchPrim = gstate.getPatchPrimitiveType();
925
bool computeNormals = gstate.isLightingEnabled();
926
bool patchFacing = gstate.patchfacing & 1;
927
u32 vertType = gstate.vertType;
929
drawEngine_.SubmitSpline(control_points, indices, gstate.getPatchDivisionU(), gstate.getPatchDivisionV(), sp_ucount, sp_vcount, sp_utype, sp_vtype, patchPrim, computeNormals, patchFacing, vertType, &bytesRead);
931
// After drawing, we advance pointers - see SubmitPrim which does the same.
932
int count = sp_ucount * sp_vcount;
933
AdvanceVerts(gstate.vertType, count, bytesRead);
936
void GPU_DX9::Execute_ViewportType(u32 op, u32 diff) {
937
gstate_c.framebufChanged = true;
938
gstate_c.textureChanged |= TEXCHANGE_PARAMSONLY;
940
case GE_CMD_VIEWPORTZSCALE:
941
case GE_CMD_VIEWPORTZCENTER:
942
shaderManager_->DirtyUniform(DIRTY_PROJMATRIX | DIRTY_DEPTHRANGE);
947
void GPU_DX9::Execute_BoundingBox(u32 op, u32 diff) {
948
// Just resetting, nothing to bound.
949
const u32 data = op & 0x00FFFFFF;
951
// TODO: Should this set the bboxResult? Let's set it true for now.
952
currentList->bboxResult = true;
955
if (((data & 7) == 0) && data <= 64) { // Sanity check
956
void *control_points = Memory::GetPointer(gstate_c.vertexAddr);
957
if (gstate.vertType & GE_VTYPE_IDX_MASK) {
958
ERROR_LOG_REPORT_ONCE(boundingbox, G3D, "Indexed bounding box data not supported.");
959
// Data seems invalid. Let's assume the box test passed.
960
currentList->bboxResult = true;
964
// Test if the bounding box is within the drawing region.
965
if (control_points) {
966
currentList->bboxResult = drawEngine_.TestBoundingBox(control_points, data, gstate.vertType);
969
ERROR_LOG_REPORT_ONCE(boundingbox, G3D, "Bad bounding box data: %06x", data);
970
// Data seems invalid. Let's assume the box test passed.
971
currentList->bboxResult = true;
975
void GPU_DX9::Execute_Region(u32 op, u32 diff) {
976
gstate_c.framebufChanged = true;
977
gstate_c.textureChanged |= TEXCHANGE_PARAMSONLY;
980
void GPU_DX9::Execute_Scissor(u32 op, u32 diff) {
981
gstate_c.framebufChanged = true;
982
gstate_c.textureChanged |= TEXCHANGE_PARAMSONLY;
985
void GPU_DX9::Execute_FramebufType(u32 op, u32 diff) {
986
gstate_c.framebufChanged = true;
987
gstate_c.textureChanged |= TEXCHANGE_PARAMSONLY;
990
void GPU_DX9::Execute_TexScaleU(u32 op, u32 diff) {
991
gstate_c.uv.uScale = getFloat24(op);
992
shaderManager_->DirtyUniform(DIRTY_UVSCALEOFFSET);
995
void GPU_DX9::Execute_TexScaleV(u32 op, u32 diff) {
996
gstate_c.uv.vScale = getFloat24(op);
997
shaderManager_->DirtyUniform(DIRTY_UVSCALEOFFSET);
1000
void GPU_DX9::Execute_TexOffsetU(u32 op, u32 diff) {
1001
gstate_c.uv.uOff = getFloat24(op);
1002
shaderManager_->DirtyUniform(DIRTY_UVSCALEOFFSET);
1005
void GPU_DX9::Execute_TexOffsetV(u32 op, u32 diff) {
1006
gstate_c.uv.vOff = getFloat24(op);
1007
shaderManager_->DirtyUniform(DIRTY_UVSCALEOFFSET);
1010
void GPU_DX9::Execute_TexAddr0(u32 op, u32 diff) {
1011
gstate_c.textureChanged = TEXCHANGE_UPDATED;
1012
shaderManager_->DirtyUniform(DIRTY_UVSCALEOFFSET);
1015
void GPU_DX9::Execute_TexAddrN(u32 op, u32 diff) {
1016
gstate_c.textureChanged |= TEXCHANGE_PARAMSONLY;
1019
void GPU_DX9::Execute_TexBufw0(u32 op, u32 diff) {
1020
gstate_c.textureChanged = TEXCHANGE_UPDATED;
1023
void GPU_DX9::Execute_TexBufwN(u32 op, u32 diff) {
1024
gstate_c.textureChanged |= TEXCHANGE_PARAMSONLY;
1027
void GPU_DX9::Execute_TexSize0(u32 op, u32 diff) {
1028
// Render to texture may have overridden the width/height.
1029
// Don't reset it unless the size is different / the texture has changed.
1030
if (diff || gstate_c.textureChanged != TEXCHANGE_UNCHANGED) {
1031
gstate_c.curTextureWidth = gstate.getTextureWidth(0);
1032
gstate_c.curTextureHeight = gstate.getTextureHeight(0);
1033
shaderManager_->DirtyUniform(DIRTY_UVSCALEOFFSET);
1034
// We will need to reset the texture now.
1035
gstate_c.textureChanged |= TEXCHANGE_PARAMSONLY;
1039
void GPU_DX9::Execute_TexSizeN(u32 op, u32 diff) {
1040
gstate_c.textureChanged |= TEXCHANGE_PARAMSONLY;
1043
void GPU_DX9::Execute_TexFormat(u32 op, u32 diff) {
1044
gstate_c.textureChanged = TEXCHANGE_UPDATED;
1047
void GPU_DX9::Execute_TexMapMode(u32 op, u32 diff) {
1048
shaderManager_->DirtyUniform(DIRTY_UVSCALEOFFSET);
1051
void GPU_DX9::Execute_TexParamType(u32 op, u32 diff) {
1052
gstate_c.textureChanged |= TEXCHANGE_PARAMSONLY;
1055
void GPU_DX9::Execute_TexEnvColor(u32 op, u32 diff) {
1056
shaderManager_->DirtyUniform(DIRTY_TEXENV);
1059
void GPU_DX9::Execute_TexLevel(u32 op, u32 diff) {
1060
// I had hoped that this would let us avoid excessively flushing in Gran Turismo, but not so,
1061
// as the game switches rapidly between modes 0 and 1.
1063
if (gstate.getTexLevelMode() == GE_TEXLEVEL_MODE_CONST) {
1064
gstate.texlevel ^= diff;
1066
gstate.texlevel ^= diff;
1069
gstate_c.textureChanged |= TEXCHANGE_PARAMSONLY;
1072
void GPU_DX9::Execute_LoadClut(u32 op, u32 diff) {
1073
gstate_c.textureChanged |= TEXCHANGE_PARAMSONLY;
1074
textureCache_.LoadClut(gstate.getClutAddress(), gstate.getClutLoadBytes());
1075
// This could be used to "dirty" textures with clut.
1078
void GPU_DX9::Execute_ClutFormat(u32 op, u32 diff) {
1079
gstate_c.textureChanged |= TEXCHANGE_PARAMSONLY;
1080
// This could be used to "dirty" textures with clut.
1083
void GPU_DX9::Execute_Ambient(u32 op, u32 diff) {
1084
shaderManager_->DirtyUniform(DIRTY_AMBIENT);
1087
void GPU_DX9::Execute_MaterialDiffuse(u32 op, u32 diff) {
1088
shaderManager_->DirtyUniform(DIRTY_MATDIFFUSE);
1091
void GPU_DX9::Execute_MaterialEmissive(u32 op, u32 diff) {
1092
shaderManager_->DirtyUniform(DIRTY_MATEMISSIVE);
1095
void GPU_DX9::Execute_MaterialAmbient(u32 op, u32 diff) {
1096
shaderManager_->DirtyUniform(DIRTY_MATAMBIENTALPHA);
1099
void GPU_DX9::Execute_MaterialSpecular(u32 op, u32 diff) {
1100
shaderManager_->DirtyUniform(DIRTY_MATSPECULAR);
1103
void GPU_DX9::Execute_Light0Param(u32 op, u32 diff) {
1104
shaderManager_->DirtyUniform(DIRTY_LIGHT0);
1107
void GPU_DX9::Execute_Light1Param(u32 op, u32 diff) {
1108
shaderManager_->DirtyUniform(DIRTY_LIGHT1);
1111
void GPU_DX9::Execute_Light2Param(u32 op, u32 diff) {
1112
shaderManager_->DirtyUniform(DIRTY_LIGHT2);
1115
void GPU_DX9::Execute_Light3Param(u32 op, u32 diff) {
1116
shaderManager_->DirtyUniform(DIRTY_LIGHT3);
1119
void GPU_DX9::Execute_FogColor(u32 op, u32 diff) {
1120
shaderManager_->DirtyUniform(DIRTY_FOGCOLOR);
1123
void GPU_DX9::Execute_FogCoef(u32 op, u32 diff) {
1124
shaderManager_->DirtyUniform(DIRTY_FOGCOEF);
1127
void GPU_DX9::Execute_ColorTestMask(u32 op, u32 diff) {
1128
shaderManager_->DirtyUniform(DIRTY_ALPHACOLORMASK);
1131
void GPU_DX9::Execute_AlphaTest(u32 op, u32 diff) {
1132
shaderManager_->DirtyUniform(DIRTY_ALPHACOLORREF);
1133
shaderManager_->DirtyUniform(DIRTY_ALPHACOLORMASK);
1136
void GPU_DX9::Execute_StencilTest(u32 op, u32 diff) {
1137
shaderManager_->DirtyUniform(DIRTY_STENCILREPLACEVALUE);
1140
void GPU_DX9::Execute_ColorRef(u32 op, u32 diff) {
1141
shaderManager_->DirtyUniform(DIRTY_ALPHACOLORREF);
1144
void GPU_DX9::Execute_WorldMtxNum(u32 op, u32 diff) {
1145
// This is almost always followed by GE_CMD_WORLDMATRIXDATA.
1146
const u32_le *src = (const u32_le *)Memory::GetPointerUnchecked(currentList->pc + 4);
1147
u32 *dst = (u32 *)(gstate.worldMatrix + (op & 0xF));
1148
const int end = 12 - (op & 0xF);
1151
while ((src[i] >> 24) == GE_CMD_WORLDMATRIXDATA) {
1152
const u32 newVal = src[i] << 8;
1153
if (dst[i] != newVal) {
1156
shaderManager_->DirtyUniform(DIRTY_WORLDMATRIX);
1163
const int count = i;
1164
gstate.worldmtxnum = (GE_CMD_WORLDMATRIXNUMBER << 24) | ((op + count) & 0xF);
1166
// Skip over the loaded data, it's done now.
1167
UpdatePC(currentList->pc, currentList->pc + count * 4);
1168
currentList->pc += count * 4;
1171
void GPU_DX9::Execute_WorldMtxData(u32 op, u32 diff) {
1172
// Note: it's uncommon to get here now, see above.
1173
int num = gstate.worldmtxnum & 0xF;
1174
u32 newVal = op << 8;
1175
if (num < 12 && newVal != ((const u32 *)gstate.worldMatrix)[num]) {
1177
((u32 *)gstate.worldMatrix)[num] = newVal;
1178
shaderManager_->DirtyUniform(DIRTY_WORLDMATRIX);
1181
gstate.worldmtxnum = (GE_CMD_WORLDMATRIXNUMBER << 24) | (num & 0xF);
1184
void GPU_DX9::Execute_ViewMtxNum(u32 op, u32 diff) {
1185
// This is almost always followed by GE_CMD_VIEWMATRIXDATA.
1186
const u32_le *src = (const u32_le *)Memory::GetPointerUnchecked(currentList->pc + 4);
1187
u32 *dst = (u32 *)(gstate.viewMatrix + (op & 0xF));
1188
const int end = 12 - (op & 0xF);
1191
while ((src[i] >> 24) == GE_CMD_VIEWMATRIXDATA) {
1192
const u32 newVal = src[i] << 8;
1193
if (dst[i] != newVal) {
1196
shaderManager_->DirtyUniform(DIRTY_VIEWMATRIX);
1203
const int count = i;
1204
gstate.viewmtxnum = (GE_CMD_VIEWMATRIXNUMBER << 24) | ((op + count) & 0xF);
1206
// Skip over the loaded data, it's done now.
1207
UpdatePC(currentList->pc, currentList->pc + count * 4);
1208
currentList->pc += count * 4;
1211
void GPU_DX9::Execute_ViewMtxData(u32 op, u32 diff) {
1212
// Note: it's uncommon to get here now, see above.
1213
int num = gstate.viewmtxnum & 0xF;
1214
u32 newVal = op << 8;
1215
if (num < 12 && newVal != ((const u32 *)gstate.viewMatrix)[num]) {
1217
((u32 *)gstate.viewMatrix)[num] = newVal;
1218
shaderManager_->DirtyUniform(DIRTY_VIEWMATRIX);
1221
gstate.viewmtxnum = (GE_CMD_VIEWMATRIXNUMBER << 24) | (num & 0xF);
1224
void GPU_DX9::Execute_ProjMtxNum(u32 op, u32 diff) {
1225
// This is almost always followed by GE_CMD_PROJMATRIXDATA.
1226
const u32_le *src = (const u32_le *)Memory::GetPointerUnchecked(currentList->pc + 4);
1227
u32 *dst = (u32 *)(gstate.projMatrix + (op & 0xF));
1228
const int end = 16 - (op & 0xF);
1231
while ((src[i] >> 24) == GE_CMD_PROJMATRIXDATA) {
1232
const u32 newVal = src[i] << 8;
1233
if (dst[i] != newVal) {
1236
shaderManager_->DirtyUniform(DIRTY_PROJMATRIX);
1243
const int count = i;
1244
gstate.projmtxnum = (GE_CMD_PROJMATRIXNUMBER << 24) | ((op + count) & 0xF);
1246
// Skip over the loaded data, it's done now.
1247
UpdatePC(currentList->pc, currentList->pc + count * 4);
1248
currentList->pc += count * 4;
1251
void GPU_DX9::Execute_ProjMtxData(u32 op, u32 diff) {
1252
// Note: it's uncommon to get here now, see above.
1253
int num = gstate.projmtxnum & 0xF;
1254
u32 newVal = op << 8;
1255
if (newVal != ((const u32 *)gstate.projMatrix)[num]) {
1257
((u32 *)gstate.projMatrix)[num] = newVal;
1258
shaderManager_->DirtyUniform(DIRTY_PROJMATRIX);
1261
gstate.projmtxnum = (GE_CMD_PROJMATRIXNUMBER << 24) | (num & 0xF);
1264
void GPU_DX9::Execute_TgenMtxNum(u32 op, u32 diff) {
1265
// This is almost always followed by GE_CMD_TGENMATRIXDATA.
1266
const u32_le *src = (const u32_le *)Memory::GetPointerUnchecked(currentList->pc + 4);
1267
u32 *dst = (u32 *)(gstate.tgenMatrix + (op & 0xF));
1268
const int end = 12 - (op & 0xF);
1271
while ((src[i] >> 24) == GE_CMD_TGENMATRIXDATA) {
1272
const u32 newVal = src[i] << 8;
1273
if (dst[i] != newVal) {
1276
shaderManager_->DirtyUniform(DIRTY_TEXMATRIX);
1283
const int count = i;
1284
gstate.texmtxnum = (GE_CMD_TGENMATRIXNUMBER << 24) | ((op + count) & 0xF);
1286
// Skip over the loaded data, it's done now.
1287
UpdatePC(currentList->pc, currentList->pc + count * 4);
1288
currentList->pc += count * 4;
1291
void GPU_DX9::Execute_TgenMtxData(u32 op, u32 diff) {
1292
// Note: it's uncommon to get here now, see above.
1293
int num = gstate.texmtxnum & 0xF;
1294
u32 newVal = op << 8;
1295
if (num < 12 && newVal != ((const u32 *)gstate.tgenMatrix)[num]) {
1297
((u32 *)gstate.tgenMatrix)[num] = newVal;
1298
shaderManager_->DirtyUniform(DIRTY_TEXMATRIX);
1301
gstate.texmtxnum = (GE_CMD_TGENMATRIXNUMBER << 24) | (num & 0xF);
1304
void GPU_DX9::Execute_BoneMtxNum(u32 op, u32 diff) {
1305
// This is almost always followed by GE_CMD_BONEMATRIXDATA.
1306
const u32_le *src = (const u32_le *)Memory::GetPointerUnchecked(currentList->pc + 4);
1307
u32 *dst = (u32 *)(gstate.boneMatrix + (op & 0x7F));
1308
const int end = 12 * 8 - (op & 0x7F);
1311
// If we can't use software skinning, we have to flush and dirty.
1312
if (!g_Config.bSoftwareSkinning || (gstate.vertType & GE_VTYPE_MORPHCOUNT_MASK) != 0) {
1313
while ((src[i] >> 24) == GE_CMD_BONEMATRIXDATA) {
1314
const u32 newVal = src[i] << 8;
1315
if (dst[i] != newVal) {
1324
const int numPlusCount = (op & 0x7F) + i;
1325
for (int num = op & 0x7F; num < numPlusCount; num += 12) {
1326
shaderManager_->DirtyUniform(DIRTY_BONEMATRIX0 << (num / 12));
1329
while ((src[i] >> 24) == GE_CMD_BONEMATRIXDATA) {
1330
dst[i] = src[i] << 8;
1336
const int numPlusCount = (op & 0x7F) + i;
1337
for (int num = op & 0x7F; num < numPlusCount; num += 12) {
1338
gstate_c.deferredVertTypeDirty |= DIRTY_BONEMATRIX0 << (num / 12);
1342
const int count = i;
1343
gstate.boneMatrixNumber = (GE_CMD_BONEMATRIXNUMBER << 24) | ((op + count) & 0x7F);
1345
// Skip over the loaded data, it's done now.
1346
UpdatePC(currentList->pc, currentList->pc + count * 4);
1347
currentList->pc += count * 4;
1350
void GPU_DX9::Execute_BoneMtxData(u32 op, u32 diff) {
1351
// Note: it's uncommon to get here now, see above.
1352
int num = gstate.boneMatrixNumber & 0x7F;
1353
u32 newVal = op << 8;
1354
if (num < 96 && newVal != ((const u32 *)gstate.boneMatrix)[num]) {
1355
// Bone matrices should NOT flush when software skinning is enabled!
1356
if (!g_Config.bSoftwareSkinning || (gstate.vertType & GE_VTYPE_MORPHCOUNT_MASK) != 0) {
1358
shaderManager_->DirtyUniform(DIRTY_BONEMATRIX0 << (num / 12));
1360
gstate_c.deferredVertTypeDirty |= DIRTY_BONEMATRIX0 << (num / 12);
1362
((u32 *)gstate.boneMatrix)[num] = newVal;
1365
gstate.boneMatrixNumber = (GE_CMD_BONEMATRIXNUMBER << 24) | (num & 0x7F);
1368
void GPU_DX9::Execute_Generic(u32 op, u32 diff) {
1370
u32 data = op & 0xFFFFFF;
1372
// Handle control and drawing commands here directly. The others we delegate.
1378
Execute_Vaddr(op, diff);
1382
Execute_Iaddr(op, diff);
1386
Execute_Prim(op, diff);
1389
// The arrow and other rotary items in Puzbob are bezier patches, strangely enough.
1391
Execute_Bezier(op, diff);
1395
Execute_Spline(op, diff);
1398
case GE_CMD_BOUNDINGBOX:
1399
Execute_BoundingBox(op, diff);
1402
case GE_CMD_REGION1:
1403
case GE_CMD_REGION2:
1404
Execute_Region(op, diff);
1407
case GE_CMD_CLIPENABLE:
1408
//we always clip, this is opengl
1411
case GE_CMD_CULLFACEENABLE:
1415
case GE_CMD_TEXTUREMAPENABLE:
1417
gstate_c.textureChanged = TEXCHANGE_UPDATED;
1420
case GE_CMD_LIGHTINGENABLE:
1423
case GE_CMD_FOGCOLOR:
1425
shaderManager_->DirtyUniform(DIRTY_FOGCOLOR);
1430
shaderManager_->DirtyUniform(DIRTY_FOGCOEF);
1435
shaderManager_->DirtyUniform(DIRTY_FOGCOEF);
1438
case GE_CMD_FOGENABLE:
1441
case GE_CMD_DITHERENABLE:
1444
case GE_CMD_OFFSETX:
1447
case GE_CMD_OFFSETY:
1450
case GE_CMD_TEXSCALEU: Execute_TexScaleU(op, diff); break;
1451
case GE_CMD_TEXSCALEV: Execute_TexScaleV(op, diff); break;
1452
case GE_CMD_TEXOFFSETU: Execute_TexOffsetU(op, diff); break;
1453
case GE_CMD_TEXOFFSETV: Execute_TexOffsetV(op, diff); break;
1455
case GE_CMD_SCISSOR1:
1456
case GE_CMD_SCISSOR2:
1457
Execute_Scissor(op, diff);
1465
case GE_CMD_FRAMEBUFPTR:
1466
case GE_CMD_FRAMEBUFWIDTH:
1467
case GE_CMD_FRAMEBUFPIXFORMAT:
1468
Execute_FramebufType(op, diff);
1471
case GE_CMD_TEXADDR0:
1472
Execute_TexAddr0(op, diff);
1475
case GE_CMD_TEXADDR1:
1476
case GE_CMD_TEXADDR2:
1477
case GE_CMD_TEXADDR3:
1478
case GE_CMD_TEXADDR4:
1479
case GE_CMD_TEXADDR5:
1480
case GE_CMD_TEXADDR6:
1481
case GE_CMD_TEXADDR7:
1482
Execute_TexAddrN(op, diff);
1485
case GE_CMD_TEXBUFWIDTH0:
1486
Execute_TexAddr0(op, diff);
1489
case GE_CMD_TEXBUFWIDTH1:
1490
case GE_CMD_TEXBUFWIDTH2:
1491
case GE_CMD_TEXBUFWIDTH3:
1492
case GE_CMD_TEXBUFWIDTH4:
1493
case GE_CMD_TEXBUFWIDTH5:
1494
case GE_CMD_TEXBUFWIDTH6:
1495
case GE_CMD_TEXBUFWIDTH7:
1496
Execute_TexBufwN(op, diff);
1499
case GE_CMD_CLUTFORMAT:
1500
Execute_ClutFormat(op, diff);
1503
case GE_CMD_CLUTADDR:
1504
case GE_CMD_CLUTADDRUPPER:
1505
// Hm, LOADCLUT actually changes the CLUT so no need to dirty here.
1508
case GE_CMD_LOADCLUT:
1509
Execute_LoadClut(op, diff);
1512
case GE_CMD_TEXMAPMODE:
1513
Execute_TexMapMode(op, diff);
1516
case GE_CMD_TEXSHADELS:
1519
case GE_CMD_TRANSFERSRC:
1520
case GE_CMD_TRANSFERSRCW:
1521
case GE_CMD_TRANSFERDST:
1522
case GE_CMD_TRANSFERDSTW:
1523
case GE_CMD_TRANSFERSRCPOS:
1524
case GE_CMD_TRANSFERDSTPOS:
1527
case GE_CMD_TRANSFERSIZE:
1530
case GE_CMD_TRANSFERSTART: // Orphis calls this TRXKICK
1532
// TODO: Here we should check if the transfer overlaps a framebuffer or any textures,
1533
// and take appropriate action. This is a block transfer between RAM and VRAM, or vice versa.
1534
// Can we skip this entirely on SkipDraw? It skips some things internally.
1535
DoBlockTransfer(gstate_c.skipDrawReason);
1537
// Fixes Gran Turismo's funky text issue, since it overwrites the current texture.
1538
gstate_c.textureChanged = TEXCHANGE_UPDATED;
1542
case GE_CMD_TEXSIZE0:
1543
Execute_TexSize0(op, diff);
1546
case GE_CMD_TEXSIZE1:
1547
case GE_CMD_TEXSIZE2:
1548
case GE_CMD_TEXSIZE3:
1549
case GE_CMD_TEXSIZE4:
1550
case GE_CMD_TEXSIZE5:
1551
case GE_CMD_TEXSIZE6:
1552
case GE_CMD_TEXSIZE7:
1553
Execute_TexSizeN(op, diff);
1556
case GE_CMD_ZBUFPTR:
1557
case GE_CMD_ZBUFWIDTH:
1560
case GE_CMD_AMBIENTCOLOR:
1561
case GE_CMD_AMBIENTALPHA:
1563
shaderManager_->DirtyUniform(DIRTY_AMBIENT);
1566
case GE_CMD_MATERIALDIFFUSE:
1568
shaderManager_->DirtyUniform(DIRTY_MATDIFFUSE);
1571
case GE_CMD_MATERIALEMISSIVE:
1573
shaderManager_->DirtyUniform(DIRTY_MATEMISSIVE);
1576
case GE_CMD_MATERIALAMBIENT:
1577
case GE_CMD_MATERIALALPHA:
1579
shaderManager_->DirtyUniform(DIRTY_MATAMBIENTALPHA);
1582
case GE_CMD_MATERIALSPECULAR:
1583
case GE_CMD_MATERIALSPECULARCOEF:
1585
shaderManager_->DirtyUniform(DIRTY_MATSPECULAR);
1588
case GE_CMD_LIGHTTYPE0:
1589
case GE_CMD_LIGHTTYPE1:
1590
case GE_CMD_LIGHTTYPE2:
1591
case GE_CMD_LIGHTTYPE3:
1593
case GE_CMD_LX0:case GE_CMD_LY0:case GE_CMD_LZ0:
1594
case GE_CMD_LDX0:case GE_CMD_LDY0:case GE_CMD_LDZ0:
1595
case GE_CMD_LKA0:case GE_CMD_LKB0:case GE_CMD_LKC0:
1596
case GE_CMD_LKS0: // spot coef ("conv")
1597
case GE_CMD_LKO0: // light angle ("cutoff")
1601
shaderManager_->DirtyUniform(DIRTY_LIGHT0);
1604
case GE_CMD_LX1:case GE_CMD_LY1:case GE_CMD_LZ1:
1605
case GE_CMD_LDX1:case GE_CMD_LDY1:case GE_CMD_LDZ1:
1606
case GE_CMD_LKA1:case GE_CMD_LKB1:case GE_CMD_LKC1:
1612
shaderManager_->DirtyUniform(DIRTY_LIGHT1);
1614
case GE_CMD_LX2:case GE_CMD_LY2:case GE_CMD_LZ2:
1615
case GE_CMD_LDX2:case GE_CMD_LDY2:case GE_CMD_LDZ2:
1616
case GE_CMD_LKA2:case GE_CMD_LKB2:case GE_CMD_LKC2:
1622
shaderManager_->DirtyUniform(DIRTY_LIGHT2);
1624
case GE_CMD_LX3:case GE_CMD_LY3:case GE_CMD_LZ3:
1625
case GE_CMD_LDX3:case GE_CMD_LDY3:case GE_CMD_LDZ3:
1626
case GE_CMD_LKA3:case GE_CMD_LKB3:case GE_CMD_LKC3:
1632
shaderManager_->DirtyUniform(DIRTY_LIGHT3);
1635
case GE_CMD_VIEWPORTXSCALE:
1636
case GE_CMD_VIEWPORTYSCALE:
1637
case GE_CMD_VIEWPORTXCENTER:
1638
case GE_CMD_VIEWPORTYCENTER:
1639
case GE_CMD_VIEWPORTZSCALE:
1640
case GE_CMD_VIEWPORTZCENTER:
1641
Execute_ViewportType(op, diff);
1644
case GE_CMD_LIGHTENABLE0:
1645
case GE_CMD_LIGHTENABLE1:
1646
case GE_CMD_LIGHTENABLE2:
1647
case GE_CMD_LIGHTENABLE3:
1650
case GE_CMD_SHADEMODE:
1653
case GE_CMD_PATCHDIVISION:
1654
case GE_CMD_PATCHPRIMITIVE:
1655
case GE_CMD_PATCHFACING:
1659
case GE_CMD_MATERIALUPDATE:
1662
//////////////////////////////////////////////////////////////////
1664
//////////////////////////////////////////////////////////////////
1665
case GE_CMD_CLEARMODE:
1668
//////////////////////////////////////////////////////////////////
1670
//////////////////////////////////////////////////////////////////
1671
case GE_CMD_ALPHABLENDENABLE:
1672
case GE_CMD_BLENDMODE:
1673
case GE_CMD_BLENDFIXEDA:
1674
case GE_CMD_BLENDFIXEDB:
1677
case GE_CMD_ALPHATESTENABLE:
1678
case GE_CMD_COLORTESTENABLE:
1679
// They are done in the fragment shader.
1682
case GE_CMD_COLORTEST:
1683
case GE_CMD_COLORTESTMASK:
1685
shaderManager_->DirtyUniform(DIRTY_ALPHACOLORMASK);
1688
case GE_CMD_ALPHATEST:
1689
if (((data >> 16) & 0xFF) != 0xFF && (data & 7) > 1)
1690
WARN_LOG_REPORT_ONCE(alphatestmask, G3D, "Unsupported alphatest mask: %02x", (data >> 16) & 0xFF);
1691
// Intentional fallthrough.
1692
case GE_CMD_COLORREF:
1694
shaderManager_->DirtyUniform(DIRTY_ALPHACOLORREF);
1697
case GE_CMD_TEXENVCOLOR:
1699
shaderManager_->DirtyUniform(DIRTY_TEXENV);
1702
case GE_CMD_TEXFUNC:
1703
case GE_CMD_TEXFLUSH:
1706
case GE_CMD_TEXFORMAT:
1707
Execute_TexFormat(op, diff);
1710
case GE_CMD_TEXMODE:
1711
case GE_CMD_TEXFILTER:
1712
case GE_CMD_TEXWRAP:
1713
Execute_TexParamType(op, diff);
1716
//////////////////////////////////////////////////////////////////
1717
// Z/STENCIL TESTING
1718
//////////////////////////////////////////////////////////////////
1720
case GE_CMD_STENCILTESTENABLE:
1721
case GE_CMD_ZTESTENABLE:
1723
case GE_CMD_ZWRITEDISABLE:
1726
case GE_CMD_MORPHWEIGHT0:
1727
case GE_CMD_MORPHWEIGHT1:
1728
case GE_CMD_MORPHWEIGHT2:
1729
case GE_CMD_MORPHWEIGHT3:
1730
case GE_CMD_MORPHWEIGHT4:
1731
case GE_CMD_MORPHWEIGHT5:
1732
case GE_CMD_MORPHWEIGHT6:
1733
case GE_CMD_MORPHWEIGHT7:
1734
gstate_c.morphWeights[cmd - GE_CMD_MORPHWEIGHT0] = getFloat24(data);
1743
case GE_CMD_LOGICOPENABLE:
1744
case GE_CMD_LOGICOP:
1747
case GE_CMD_ANTIALIASENABLE:
1749
WARN_LOG_REPORT_ONCE(antiAlias, G3D, "Unsupported antialias enabled: %06x", data);
1752
case GE_CMD_TEXLODSLOPE:
1754
WARN_LOG_REPORT_ONCE(texLodSlope, G3D, "Unsupported texture lod slope: %06x", data);
1757
case GE_CMD_TEXLEVEL:
1758
Execute_TexLevel(op, diff);
1763
WARN_LOG_REPORT_ONCE(vscx, G3D, "Unsupported Vertex Screen Coordinate X : %06x", data);
1768
WARN_LOG_REPORT_ONCE(vscy, G3D, "Unsupported Vertex Screen Coordinate Y : %06x", data);
1773
WARN_LOG_REPORT_ONCE(vscz, G3D, "Unsupported Vertex Screen Coordinate Z : %06x", data);
1778
WARN_LOG_REPORT_ONCE(vtcs, G3D, "Unsupported Vertex Texture Coordinate S : %06x", data);
1783
WARN_LOG_REPORT_ONCE(vtct, G3D, "Unsupported Vertex Texture Coordinate T : %06x", data);
1788
WARN_LOG_REPORT_ONCE(vtcq, G3D, "Unsupported Vertex Texture Coordinate Q : %06x", data);
1793
WARN_LOG_REPORT_ONCE(vcv, G3D, "Unsupported Vertex Color Value : %06x", data);
1798
WARN_LOG_REPORT_ONCE(vap, G3D, "Unsupported Vertex Alpha and Primitive : %06x", data);
1803
WARN_LOG_REPORT_ONCE(vfc, G3D, "Unsupported Vertex Fog Coefficient : %06x", data);
1808
WARN_LOG_REPORT_ONCE(vscv, G3D, "Unsupported Vertex Secondary Color Value : %06x", data);
1812
case GE_CMD_UNKNOWN_03:
1813
case GE_CMD_UNKNOWN_0D:
1814
case GE_CMD_UNKNOWN_11:
1815
case GE_CMD_UNKNOWN_29:
1816
case GE_CMD_UNKNOWN_34:
1817
case GE_CMD_UNKNOWN_35:
1818
case GE_CMD_UNKNOWN_39:
1819
case GE_CMD_UNKNOWN_4E:
1820
case GE_CMD_UNKNOWN_4F:
1821
case GE_CMD_UNKNOWN_52:
1822
case GE_CMD_UNKNOWN_59:
1823
case GE_CMD_UNKNOWN_5A:
1824
case GE_CMD_UNKNOWN_B6:
1825
case GE_CMD_UNKNOWN_B7:
1826
case GE_CMD_UNKNOWN_D1:
1827
case GE_CMD_UNKNOWN_ED:
1828
case GE_CMD_UNKNOWN_EF:
1829
case GE_CMD_UNKNOWN_FA:
1830
case GE_CMD_UNKNOWN_FB:
1831
case GE_CMD_UNKNOWN_FC:
1832
case GE_CMD_UNKNOWN_FD:
1833
case GE_CMD_UNKNOWN_FE:
1835
WARN_LOG_REPORT_ONCE(unknowncmd, G3D, "Unknown GE command : %08x ", op);
1837
case GE_CMD_UNKNOWN_FF:
1838
// This is hit in quite a few games, supposedly it is a no-op.
1839
// Might be used for debugging or something?
1843
GPUCommon::ExecuteOp(op, diff);
1848
void GPU_DX9::FastLoadBoneMatrix(u32 target) {
1849
const int num = gstate.boneMatrixNumber & 0x7F;
1850
const int mtxNum = num / 12;
1851
uint32_t uniformsToDirty = DIRTY_BONEMATRIX0 << mtxNum;
1852
if ((num - 12 * mtxNum) != 0) {
1853
uniformsToDirty |= DIRTY_BONEMATRIX0 << ((mtxNum + 1) & 7);
1856
if (!g_Config.bSoftwareSkinning || (gstate.vertType & GE_VTYPE_MORPHCOUNT_MASK) != 0) {
1858
shaderManager_->DirtyUniform(uniformsToDirty);
1860
gstate_c.deferredVertTypeDirty |= uniformsToDirty;
1862
gstate.FastLoadBoneMatrix(target);
1865
void GPU_DX9::GetStats(char *buffer, size_t bufsize) {
1866
float vertexAverageCycles = gpuStats.numVertsSubmitted > 0 ? (float)gpuStats.vertexGPUCycles / (float)gpuStats.numVertsSubmitted : 0.0f;
1867
snprintf(buffer, bufsize - 1,
1868
"DL processing time: %0.2f ms\n"
1869
"Draw calls: %i, flushes %i\n"
1870
"Cached Draw calls: %i\n"
1871
"Num Tracked Vertex Arrays: %i\n"
1872
"GPU cycles executed: %d (%f per vertex)\n"
1873
"Commands per call level: %i %i %i %i\n"
1874
"Vertices submitted: %i\n"
1875
"Cached, Uncached Vertices Drawn: %i, %i\n"
1877
"Textures active: %i, decoded: %i invalidated: %i\n"
1878
"Vertex, Fragment shaders loaded: %i, %i\n",
1879
gpuStats.msProcessingDisplayLists * 1000.0f,
1880
gpuStats.numDrawCalls,
1881
gpuStats.numFlushes,
1882
gpuStats.numCachedDrawCalls,
1883
gpuStats.numTrackedVertexArrays,
1884
gpuStats.vertexGPUCycles + gpuStats.otherGPUCycles,
1885
vertexAverageCycles,
1886
gpuStats.gpuCommandsAtCallLevel[0], gpuStats.gpuCommandsAtCallLevel[1], gpuStats.gpuCommandsAtCallLevel[2], gpuStats.gpuCommandsAtCallLevel[3],
1887
gpuStats.numVertsSubmitted,
1888
gpuStats.numCachedVertsDrawn,
1889
gpuStats.numUncachedVertsDrawn,
1890
(int)framebufferManager_.NumVFBs(),
1891
(int)textureCache_.NumLoadedTextures(),
1892
gpuStats.numTexturesDecoded,
1893
gpuStats.numTextureInvalidations,
1894
shaderManager_->NumVertexShaders(),
1895
shaderManager_->NumFragmentShaders()
1899
void GPU_DX9::DoBlockTransfer(u32 skipDrawReason) {
1900
// TODO: This is used a lot to copy data around between render targets and textures,
1901
// and also to quickly load textures from RAM to VRAM. So we should do checks like the following:
1902
// * Does dstBasePtr point to an existing texture? If so maybe reload it immediately.
1904
// * Does srcBasePtr point to a render target, and dstBasePtr to a texture? If so
1905
// either copy between rt and texture or reassign the texture to point to the render target
1909
u32 srcBasePtr = gstate.getTransferSrcAddress();
1910
u32 srcStride = gstate.getTransferSrcStride();
1912
u32 dstBasePtr = gstate.getTransferDstAddress();
1913
u32 dstStride = gstate.getTransferDstStride();
1915
int srcX = gstate.getTransferSrcX();
1916
int srcY = gstate.getTransferSrcY();
1918
int dstX = gstate.getTransferDstX();
1919
int dstY = gstate.getTransferDstY();
1921
int width = gstate.getTransferWidth();
1922
int height = gstate.getTransferHeight();
1924
int bpp = gstate.getTransferBpp();
1926
DEBUG_LOG(G3D, "Block transfer: %08x/%x -> %08x/%x, %ix%ix%i (%i,%i)->(%i,%i)", srcBasePtr, srcStride, dstBasePtr, dstStride, width, height, bpp, srcX, srcY, dstX, dstY);
1928
if (!Memory::IsValidAddress(srcBasePtr)) {
1929
ERROR_LOG_REPORT(G3D, "BlockTransfer: Bad source transfer address %08x!", srcBasePtr);
1933
if (!Memory::IsValidAddress(dstBasePtr)) {
1934
ERROR_LOG_REPORT(G3D, "BlockTransfer: Bad destination transfer address %08x!", dstBasePtr);
1938
// Check that the last address of both source and dest are valid addresses
1940
u32 srcLastAddr = srcBasePtr + ((height - 1 + srcY) * srcStride + (srcX + width - 1)) * bpp;
1941
u32 dstLastAddr = dstBasePtr + ((height - 1 + dstY) * dstStride + (dstX + width - 1)) * bpp;
1943
if (!Memory::IsValidAddress(srcLastAddr)) {
1944
ERROR_LOG_REPORT(G3D, "Bottom-right corner of source of block transfer is at an invalid address: %08x", srcLastAddr);
1947
if (!Memory::IsValidAddress(dstLastAddr)) {
1948
ERROR_LOG_REPORT(G3D, "Bottom-right corner of destination of block transfer is at an invalid address: %08x", srcLastAddr);
1952
// Tell the framebuffer manager to take action if possible. If it does the entire thing, let's just return.
1953
if (!framebufferManager_.NotifyBlockTransferBefore(dstBasePtr, dstStride, dstX, dstY, srcBasePtr, srcStride, srcX, srcY, width, height, bpp, skipDrawReason)) {
1954
// Do the copy! (Hm, if we detect a drawn video frame (see below) then we could maybe skip this?)
1955
// Can use GetPointerUnchecked because we checked the addresses above. We could also avoid them
1956
// entirely by walking a couple of pointers...
1957
if (srcStride == dstStride && (u32)width == srcStride) {
1958
// Common case in God of War, let's do it all in one chunk.
1959
u32 srcLineStartAddr = srcBasePtr + (srcY * srcStride + srcX) * bpp;
1960
u32 dstLineStartAddr = dstBasePtr + (dstY * dstStride + dstX) * bpp;
1961
const u8 *src = Memory::GetPointerUnchecked(srcLineStartAddr);
1962
u8 *dst = Memory::GetPointerUnchecked(dstLineStartAddr);
1963
memcpy(dst, src, width * height * bpp);
1965
for (int y = 0; y < height; y++) {
1966
u32 srcLineStartAddr = srcBasePtr + ((y + srcY) * srcStride + srcX) * bpp;
1967
u32 dstLineStartAddr = dstBasePtr + ((y + dstY) * dstStride + dstX) * bpp;
1969
const u8 *src = Memory::GetPointerUnchecked(srcLineStartAddr);
1970
u8 *dst = Memory::GetPointerUnchecked(dstLineStartAddr);
1971
memcpy(dst, src, width * bpp);
1975
textureCache_.Invalidate(dstBasePtr + (dstY * dstStride + dstX) * bpp, height * dstStride * bpp, GPU_INVALIDATE_HINT);
1976
framebufferManager_.NotifyBlockTransferAfter(dstBasePtr, dstStride, dstX, dstY, srcBasePtr, srcStride, srcX, srcY, width, height, bpp, skipDrawReason);
1979
CBreakPoints::ExecMemCheck(srcBasePtr + (srcY * srcStride + srcX) * bpp, false, height * srcStride * bpp, currentMIPS->pc);
1980
CBreakPoints::ExecMemCheck(dstBasePtr + (srcY * dstStride + srcX) * bpp, true, height * dstStride * bpp, currentMIPS->pc);
1982
// TODO: Correct timing appears to be 1.9, but erring a bit low since some of our other timing is inaccurate.
1983
cyclesExecuted += ((height * width * bpp) * 16) / 10;
1986
void GPU_DX9::InvalidateCache(u32 addr, int size, GPUInvalidationType type) {
1987
GPUEvent ev(GPU_EVENT_INVALIDATE_CACHE);
1988
ev.invalidate_cache.addr = addr;
1989
ev.invalidate_cache.size = size;
1990
ev.invalidate_cache.type = type;
1994
void GPU_DX9::InvalidateCacheInternal(u32 addr, int size, GPUInvalidationType type) {
1996
textureCache_.Invalidate(addr, size, type);
1998
textureCache_.InvalidateAll(type);
2000
if (type != GPU_INVALIDATE_ALL && framebufferManager_.MayIntersectFramebuffer(addr)) {
2001
// If we're doing block transfers, we shouldn't need this, and it'll only confuse us.
2002
// Vempire invalidates (with writeback) after drawing, but before blitting.
2003
if (!g_Config.bBlockTransferGPU || type == GPU_INVALIDATE_SAFE) {
2004
framebufferManager_.UpdateFromMemory(addr, size, type == GPU_INVALIDATE_SAFE);
2009
void GPU_DX9::NotifyVideoUpload(u32 addr, int size, int width, int format) {
2010
if (Memory::IsVRAMAddress(addr)) {
2011
framebufferManager_.NotifyVideoUpload(addr, size, width, (GEBufferFormat)format);
2013
textureCache_.NotifyVideoUpload(addr, size, width, (GEBufferFormat)format);
2014
InvalidateCache(addr, size, GPU_INVALIDATE_SAFE);
2017
void GPU_DX9::PerformMemoryCopyInternal(u32 dest, u32 src, int size) {
2018
if (!framebufferManager_.NotifyFramebufferCopy(src, dest, size, false, gstate_c.skipDrawReason)) {
2019
// We use a little hack for Download/Upload using a VRAM mirror.
2020
// Since they're identical we don't need to copy.
2021
if (!Memory::IsVRAMAddress(dest) || (dest ^ 0x00400000) != src) {
2022
Memory::Memcpy(dest, src, size);
2025
InvalidateCache(dest, size, GPU_INVALIDATE_HINT);
2028
void GPU_DX9::PerformMemorySetInternal(u32 dest, u8 v, int size) {
2029
if (!framebufferManager_.NotifyFramebufferCopy(dest, dest, size, true, gstate_c.skipDrawReason)) {
2030
InvalidateCache(dest, size, GPU_INVALIDATE_HINT);
2034
void GPU_DX9::PerformStencilUploadInternal(u32 dest, int size) {
2035
framebufferManager_.NotifyStencilUpload(dest, size);
2038
bool GPU_DX9::PerformMemoryCopy(u32 dest, u32 src, int size) {
2039
// Track stray copies of a framebuffer in RAM. MotoGP does this.
2040
if (framebufferManager_.MayIntersectFramebuffer(src) || framebufferManager_.MayIntersectFramebuffer(dest)) {
2041
if (IsOnSeparateCPUThread()) {
2042
GPUEvent ev(GPU_EVENT_FB_MEMCPY);
2043
ev.fb_memcpy.dst = dest;
2044
ev.fb_memcpy.src = src;
2045
ev.fb_memcpy.size = size;
2048
// This is a memcpy, so we need to wait for it to complete.
2051
PerformMemoryCopyInternal(dest, src, size);
2056
InvalidateCache(dest, size, GPU_INVALIDATE_HINT);
2060
bool GPU_DX9::PerformMemorySet(u32 dest, u8 v, int size) {
2061
// This may indicate a memset, usually to 0, of a framebuffer.
2062
if (framebufferManager_.MayIntersectFramebuffer(dest)) {
2063
Memory::Memset(dest, v, size);
2065
if (IsOnSeparateCPUThread()) {
2066
GPUEvent ev(GPU_EVENT_FB_MEMSET);
2067
ev.fb_memset.dst = dest;
2069
ev.fb_memset.size = size;
2072
// We don't need to wait for the framebuffer to be updated.
2074
PerformMemorySetInternal(dest, v, size);
2079
// Or perhaps a texture, let's invalidate.
2080
InvalidateCache(dest, size, GPU_INVALIDATE_HINT);
2084
bool GPU_DX9::PerformMemoryDownload(u32 dest, int size) {
2085
// Cheat a bit to force a download of the framebuffer.
2086
// VRAM + 0x00400000 is simply a VRAM mirror.
2087
if (Memory::IsVRAMAddress(dest)) {
2088
return PerformMemoryCopy(dest ^ 0x00400000, dest, size);
2093
bool GPU_DX9::PerformMemoryUpload(u32 dest, int size) {
2094
// Cheat a bit to force an upload of the framebuffer.
2095
// VRAM + 0x00400000 is simply a VRAM mirror.
2096
if (Memory::IsVRAMAddress(dest)) {
2097
return PerformMemoryCopy(dest, dest ^ 0x00400000, size);
2102
bool GPU_DX9::PerformStencilUpload(u32 dest, int size) {
2103
if (framebufferManager_.MayIntersectFramebuffer(dest)) {
2104
if (IsOnSeparateCPUThread()) {
2105
GPUEvent ev(GPU_EVENT_FB_STENCIL_UPLOAD);
2106
ev.fb_stencil_upload.dst = dest;
2107
ev.fb_stencil_upload.size = size;
2110
PerformStencilUploadInternal(dest, size);
2117
void GPU_DX9::ClearCacheNextFrame() {
2118
textureCache_.ClearNextFrame();
2121
void GPU_DX9::Resized() {
2123
framebufferManager_.Resized();
2126
void GPU_DX9::ClearShaderCache() {
2127
shaderManager_->ClearCache(true);
2130
std::vector<FramebufferInfo> GPU_DX9::GetFramebufferList() {
2131
return framebufferManager_.GetFramebufferList();
2134
void GPU_DX9::DoState(PointerWrap &p) {
2135
GPUCommon::DoState(p);
2137
// TODO: Some of these things may not be necessary.
2138
// None of these are necessary when saving.
2139
if (p.mode == p.MODE_READ) {
2140
textureCache_.Clear(true);
2141
drawEngine_.ClearTrackedVertexArrays();
2143
gstate_c.textureChanged = TEXCHANGE_UPDATED;
2144
framebufferManager_.DestroyAllFBOs(true);
2145
shaderManager_->ClearCache(true);
2149
bool GPU_DX9::GetCurrentFramebuffer(GPUDebugBuffer &buffer, int maxRes) {
2150
return framebufferManager_.GetCurrentFramebuffer(buffer, maxRes);
2153
bool GPU_DX9::GetCurrentDepthbuffer(GPUDebugBuffer &buffer) {
2154
return framebufferManager_.GetCurrentDepthbuffer(buffer);
2157
bool GPU_DX9::GetCurrentStencilbuffer(GPUDebugBuffer &buffer) {
2158
return framebufferManager_.GetCurrentStencilbuffer(buffer);
2161
bool GPU_DX9::GetCurrentTexture(GPUDebugBuffer &buffer, int level) {
2162
if (!gstate.isTextureMapEnabled()) {
2166
textureCache_.SetTexture(true);
2167
textureCache_.ApplyTexture();
2168
int w = gstate.getTextureWidth(level);
2169
int h = gstate.getTextureHeight(level);
2171
LPDIRECT3DBASETEXTURE9 baseTex;
2172
LPDIRECT3DTEXTURE9 tex;
2173
LPDIRECT3DSURFACE9 offscreen = nullptr;
2176
bool success = false;
2177
hr = pD3Ddevice->GetTexture(0, &baseTex);
2178
if (SUCCEEDED(hr) && baseTex != NULL) {
2179
hr = baseTex->QueryInterface(IID_IDirect3DTexture9, (void **)&tex);
2180
if (SUCCEEDED(hr)) {
2181
D3DSURFACE_DESC desc;
2182
D3DLOCKED_RECT locked;
2183
tex->GetLevelDesc(level, &desc);
2184
RECT rect = {0, 0, (LONG)desc.Width, (LONG)desc.Height};
2185
hr = tex->LockRect(level, &locked, &rect, D3DLOCK_READONLY);
2187
// If it fails, this means it's a render-to-texture, so we have to get creative.
2189
LPDIRECT3DSURFACE9 renderTarget = nullptr;
2190
hr = tex->GetSurfaceLevel(level, &renderTarget);
2191
if (renderTarget && SUCCEEDED(hr)) {
2192
hr = pD3Ddevice->CreateOffscreenPlainSurface(desc.Width, desc.Height, desc.Format, D3DPOOL_SYSTEMMEM, &offscreen, NULL);
2193
if (SUCCEEDED(hr)) {
2194
hr = pD3Ddevice->GetRenderTargetData(renderTarget, offscreen);
2195
if (SUCCEEDED(hr)) {
2196
hr = offscreen->LockRect(&locked, &rect, D3DLOCK_READONLY);
2199
renderTarget->Release();
2203
if (SUCCEEDED(hr)) {
2204
GPUDebugBufferFormat fmt;
2206
switch (desc.Format) {
2207
case D3DFMT_A1R5G5B5:
2208
fmt = gstate_c.bgraTexture ? GPU_DBG_FORMAT_5551 : GPU_DBG_FORMAT_5551_BGRA;
2211
case D3DFMT_A4R4G4B4:
2212
fmt = gstate_c.bgraTexture ? GPU_DBG_FORMAT_4444 : GPU_DBG_FORMAT_4444_BGRA;
2216
fmt = gstate_c.bgraTexture ? GPU_DBG_FORMAT_565 : GPU_DBG_FORMAT_565_BGRA;
2219
case D3DFMT_A8R8G8B8:
2220
fmt = gstate_c.bgraTexture ? GPU_DBG_FORMAT_8888 : GPU_DBG_FORMAT_8888_BGRA;
2224
fmt = GPU_DBG_FORMAT_INVALID;
2228
if (fmt != GPU_DBG_FORMAT_INVALID) {
2229
buffer.Allocate(locked.Pitch / pixelSize, desc.Height, fmt, false);
2230
memcpy(buffer.GetData(), locked.pBits, locked.Pitch * desc.Height);
2236
offscreen->UnlockRect();
2237
offscreen->Release();
2239
tex->UnlockRect(level);
2250
bool GPU_DX9::GetCurrentClut(GPUDebugBuffer &buffer) {
2251
return textureCache_.GetCurrentClutBuffer(buffer);
2254
bool GPU_DX9::GetDisplayFramebuffer(GPUDebugBuffer &buffer) {
2255
return FramebufferManagerDX9::GetDisplayFramebuffer(buffer);
2258
bool GPU_DX9::GetCurrentSimpleVertices(int count, std::vector<GPUDebugVertex> &vertices, std::vector<u16> &indices) {
2259
return drawEngine_.GetCurrentSimpleVertices(count, vertices, indices);
2262
std::vector<std::string> GPU_DX9::DebugGetShaderIDs(DebugShaderType type) {
2263
if (type == SHADER_TYPE_VERTEXLOADER) {
2264
return drawEngine_.DebugGetVertexLoaderIDs();
2266
return shaderManager_->DebugGetShaderIDs(type);
2270
std::string GPU_DX9::DebugGetShaderString(std::string id, DebugShaderType type, DebugShaderStringType stringType) {
2271
if (type == SHADER_TYPE_VERTEXLOADER) {
2272
return drawEngine_.DebugGetVertexLoaderString(id, stringType);
2274
return shaderManager_->DebugGetShaderString(id, type, stringType);
added
removed
GPU/Directx9/GPU_DX9.cpp
