?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
Merged identical VertexManager code from DX9/DX11/OGL plugins into VideoCommon. Still...
[dolphin.git] / Source / Core / VideoCommon / Src / VertexLoader.cpp
blobfba141d4cc569e3518d5ed93cbe665108f8e9ed7
1 // Copyright (C) 2003 Dolphin Project.
2
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.
6
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.
11
12 // A copy of the GPL 2.0 should have been included with the program.
13 // If not, see http://www.gnu.org/licenses/
14
15 // Official SVN repository and contact information can be found at
16 // http://code.google.com/p/dolphin-emu/
17
18 #include <assert.h>
19
20 #include "Common.h"
21 #include "VideoCommon.h"
22 #include "VideoConfig.h"
23 #include "Profiler.h"
24 #include "MemoryUtil.h"
25 #include "StringUtil.h"
26 #include "x64Emitter.h"
27 #include "ABI.h"
28 #include "PixelEngine.h"
29
30 #include "LookUpTables.h"
31 #include "Statistics.h"
32 #include "VertexLoaderManager.h"
33 #include "VertexLoader.h"
34 #include "BPMemory.h"
35 #include "DataReader.h"
36 #include "VertexManagerBase.h"
37
38 #include "VertexLoader_Position.h"
39 #include "VertexLoader_Normal.h"
40 #include "VertexLoader_Color.h"
41 #include "VertexLoader_TextCoord.h"
42
43 //BBox
44 #include "XFMemory.h"
45 extern float GC_ALIGNED16(g_fProjectionMatrix[16]);
46
47 #define USE_JIT
48
49 #define COMPILED_CODE_SIZE 4096
50
51 NativeVertexFormat *g_nativeVertexFmt;
52
53 #ifndef _WIN32
54 #undef inline
55 #define inline
56 #endif
57
58 // Matrix components are first in GC format but later in PC format - we need to store it temporarily
59 // when decoding each vertex.
60 static u8 s_curposmtx;
61 static u8 s_curtexmtx[8];
62 static int s_texmtxwrite = 0;
63 static int s_texmtxread = 0;
64
65 static int loop_counter;
66
67 // Vertex loaders read these. Although the scale ones should be baked into the shader.
68 int tcIndex;
69 int colIndex;
70 TVtxAttr* pVtxAttr;
71 int colElements[2];
72 float posScale;
73 float tcScale[8];
74
75 static const float fractionTable[32] = {
76 1.0f / (1U << 0), 1.0f / (1U << 1), 1.0f / (1U << 2), 1.0f / (1U << 3),
77 1.0f / (1U << 4), 1.0f / (1U << 5), 1.0f / (1U << 6), 1.0f / (1U << 7),
78 1.0f / (1U << 8), 1.0f / (1U << 9), 1.0f / (1U << 10), 1.0f / (1U << 11),
79 1.0f / (1U << 12), 1.0f / (1U << 13), 1.0f / (1U << 14), 1.0f / (1U << 15),
80 1.0f / (1U << 16), 1.0f / (1U << 17), 1.0f / (1U << 18), 1.0f / (1U << 19),
81 1.0f / (1U << 20), 1.0f / (1U << 21), 1.0f / (1U << 22), 1.0f / (1U << 23),
82 1.0f / (1U << 24), 1.0f / (1U << 25), 1.0f / (1U << 26), 1.0f / (1U << 27),
83 1.0f / (1U << 28), 1.0f / (1U << 29), 1.0f / (1U << 30), 1.0f / (1U << 31),
84 };
85
86 using namespace Gen;
87
88 void LOADERDECL PosMtx_ReadDirect_UByte()
89 {
90 s_curposmtx = DataReadU8() & 0x3f;
91 PRIM_LOG("posmtx: %d, ", s_curposmtx);
92 }
93
94 void LOADERDECL PosMtx_Write()
95 {
96 *VertexManager::s_pCurBufferPointer++ = s_curposmtx;
97 *VertexManager::s_pCurBufferPointer++ = 0;
98 *VertexManager::s_pCurBufferPointer++ = 0;
99 *VertexManager::s_pCurBufferPointer++ = 0;
100 }
101
102 void LOADERDECL UpdateBoundingBox()
103 {
104 if (!PixelEngine::bbox_active)
105 return;
106
107 // Truly evil hack, reading backwards from the write pointer. If we were writing to write-only
108 // memory like we might have been with a D3D vertex buffer, this would have been a bad idea.
109 float *data = (float *)(VertexManager::s_pCurBufferPointer - 12);
110 // We must transform the just loaded point by the current world and projection matrix - in software.
111 // Then convert to screen space and update the bounding box.
112 float p[3] = {data[0], data[1], data[2]};
113
114 const float *world_matrix = (float*)xfmem + MatrixIndexA.PosNormalMtxIdx * 4;
115 const float *proj_matrix = &g_fProjectionMatrix[0];
116
117 float t[3];
118 t[0] = p[0] * world_matrix[0] + p[1] * world_matrix[1] + p[2] * world_matrix[2] + world_matrix[3];
119 t[1] = p[0] * world_matrix[4] + p[1] * world_matrix[5] + p[2] * world_matrix[6] + world_matrix[7];
120 t[2] = p[0] * world_matrix[8] + p[1] * world_matrix[9] + p[2] * world_matrix[10] + world_matrix[11];
121
122 float o[4];
123 o[2] = t[0] * proj_matrix[8] + t[1] * proj_matrix[9] + t[2] * proj_matrix[10] + proj_matrix[11];
124 // Depth culling
125 if (o[2] < 0.0) {
126 // No pixels are likely to be drawn - don't update bounding box.
127 return;
128 }
129 o[0] = t[0] * proj_matrix[0] + t[1] * proj_matrix[1] + t[2] * proj_matrix[2] + proj_matrix[3];
130 o[1] = t[0] * proj_matrix[4] + t[1] * proj_matrix[5] + t[2] * proj_matrix[6] + proj_matrix[7];
131 o[3] = t[0] * proj_matrix[12] + t[1] * proj_matrix[13] + t[2] * proj_matrix[14] + proj_matrix[15];
132
133 o[0] /= o[3];
134 o[1] /= o[3];
135
136 // should possibly adjust for viewport?
137 o[0] = (o[0] + 1.0f) * 320.0f;
138 o[1] = (o[1] + 1.0f) * 240.0f;
139
140 if (o[0] < PixelEngine::bbox[0]) PixelEngine::bbox[0] = (u16)std::max(0.0f, o[0]);
141 if (o[0] > PixelEngine::bbox[1]) PixelEngine::bbox[1] = (u16)std::min(640.0f, o[0]);
142 if (o[1] < PixelEngine::bbox[2]) PixelEngine::bbox[2] = (u16)std::max(0.0f, o[1]);
143 if (o[1] > PixelEngine::bbox[3]) PixelEngine::bbox[3] = (u16)std::min(480.0f, o[1]);
144 /*
145 if (GetAsyncKeyState(VK_LSHIFT)) {
146 ERROR_LOG(VIDEO, "XForm: %f %f %f to %f %f", p[0], p[1], p[2], o[0], o[1]);
147 ERROR_LOG(VIDEO, "%i %i %i %i", g_VideoInitialize.pBBox[0], g_VideoInitialize.pBBox[1], g_VideoInitialize.pBBox[2], g_VideoInitialize.pBBox[3]);
148 }*/
149 }
150
151 void LOADERDECL TexMtx_ReadDirect_UByte()
152 {
153 s_curtexmtx[s_texmtxread] = DataReadU8() & 0x3f;
154 PRIM_LOG("texmtx%d: %d, ", s_texmtxread, s_curtexmtx[s_texmtxread]);
155 s_texmtxread++;
156 }
157
158 void LOADERDECL TexMtx_Write_Float()
159 {
160 *(float*)VertexManager::s_pCurBufferPointer = (float)s_curtexmtx[s_texmtxwrite++];
161 VertexManager::s_pCurBufferPointer += 4;
162 }
163
164 void LOADERDECL TexMtx_Write_Float2()
165 {
166 ((float*)VertexManager::s_pCurBufferPointer)[0] = 0;
167 ((float*)VertexManager::s_pCurBufferPointer)[1] = (float)s_curtexmtx[s_texmtxwrite++];
168 VertexManager::s_pCurBufferPointer += 8;
169 }
170
171 void LOADERDECL TexMtx_Write_Float4()
172 {
173 ((float*)VertexManager::s_pCurBufferPointer)[0] = 0;
174 ((float*)VertexManager::s_pCurBufferPointer)[1] = 0;
175 ((float*)VertexManager::s_pCurBufferPointer)[2] = s_curtexmtx[s_texmtxwrite++];
176 ((float*)VertexManager::s_pCurBufferPointer)[3] = 0; // Just to fill out with 0.
177 VertexManager::s_pCurBufferPointer += 16;
178 }
179
180 VertexLoader::VertexLoader(const TVtxDesc &vtx_desc, const VAT &vtx_attr)
181 {
182 m_compiledCode = NULL;
183 m_numLoadedVertices = 0;
184 m_VertexSize = 0;
185 m_numPipelineStages = 0;
186 m_NativeFmt = NativeVertexFormat::Create();
187 loop_counter = 0;
188 VertexLoader_Normal::Init();
189 VertexLoader_Position::Init();
190 VertexLoader_TextCoord::Init();
191
192 m_VtxDesc = vtx_desc;
193 SetVAT(vtx_attr.g0.Hex, vtx_attr.g1.Hex, vtx_attr.g2.Hex);
194
195 AllocCodeSpace(COMPILED_CODE_SIZE);
196 CompileVertexTranslator();
197 WriteProtect();
198 }
199
200 VertexLoader::~VertexLoader()
201 {
202 FreeCodeSpace();
203 delete m_NativeFmt;
204 }
205
206 void VertexLoader::CompileVertexTranslator()
207 {
208 m_VertexSize = 0;
209 const TVtxAttr &vtx_attr = m_VtxAttr;
210
211 #ifdef USE_JIT
212 if (m_compiledCode)
213 PanicAlert("trying to recompile a vtx translator");
214
215 m_compiledCode = GetCodePtr();
216 ABI_EmitPrologue(4);
217
218 // Start loop here
219 const u8 *loop_start = GetCodePtr();
220
221 // Reset component counters if present in vertex format only.
222 if (m_VtxDesc.Tex0Coord || m_VtxDesc.Tex1Coord || m_VtxDesc.Tex2Coord || m_VtxDesc.Tex3Coord ||
223 m_VtxDesc.Tex4Coord || m_VtxDesc.Tex5Coord || m_VtxDesc.Tex6Coord || m_VtxDesc.Tex7Coord) {
224 WriteSetVariable(32, &tcIndex, Imm32(0));
225 }
226 if (m_VtxDesc.Color0 || m_VtxDesc.Color1) {
227 WriteSetVariable(32, &colIndex, Imm32(0));
228 }
229 if (m_VtxDesc.Tex0MatIdx || m_VtxDesc.Tex1MatIdx || m_VtxDesc.Tex2MatIdx || m_VtxDesc.Tex3MatIdx ||
230 m_VtxDesc.Tex4MatIdx || m_VtxDesc.Tex5MatIdx || m_VtxDesc.Tex6MatIdx || m_VtxDesc.Tex7MatIdx) {
231 WriteSetVariable(32, &s_texmtxwrite, Imm32(0));
232 WriteSetVariable(32, &s_texmtxread, Imm32(0));
233 }
234 #endif
235
236 // Colors
237 const int col[2] = {m_VtxDesc.Color0, m_VtxDesc.Color1};
238 // TextureCoord
239 // Since m_VtxDesc.Text7Coord is broken across a 32 bit word boundary, retrieve its value manually.
240 // If we didn't do this, the vertex format would be read as one bit offset from where it should be, making
241 // 01 become 00, and 10/11 become 01
242 const int tc[8] = {
243 m_VtxDesc.Tex0Coord, m_VtxDesc.Tex1Coord, m_VtxDesc.Tex2Coord, m_VtxDesc.Tex3Coord,
244 m_VtxDesc.Tex4Coord, m_VtxDesc.Tex5Coord, m_VtxDesc.Tex6Coord, (m_VtxDesc.Hex >> 31) & 3
245 };
246
247 // Reset pipeline
248 m_numPipelineStages = 0;
249
250 // It's a bit ugly that we poke inside m_NativeFmt in this function. Planning to fix this.
251 m_NativeFmt->m_components = 0;
252
253 // Position in pc vertex format.
254 int nat_offset = 0;
255 PortableVertexDeclaration vtx_decl;
256 memset(&vtx_decl, 0, sizeof(vtx_decl));
257 for (int i = 0; i < 8; i++) {
258 vtx_decl.texcoord_offset[i] = -1;
259 }
260
261 // m_VBVertexStride for texmtx and posmtx is computed later when writing.
262
263 // Position Matrix Index
264 if (m_VtxDesc.PosMatIdx) {
265 WriteCall(PosMtx_ReadDirect_UByte);
266 m_NativeFmt->m_components |= VB_HAS_POSMTXIDX;
267 m_VertexSize += 1;
268 }
269
270 if (m_VtxDesc.Tex0MatIdx) {m_VertexSize += 1; m_NativeFmt->m_components |= VB_HAS_TEXMTXIDX0; WriteCall(TexMtx_ReadDirect_UByte); }
271 if (m_VtxDesc.Tex1MatIdx) {m_VertexSize += 1; m_NativeFmt->m_components |= VB_HAS_TEXMTXIDX1; WriteCall(TexMtx_ReadDirect_UByte); }
272 if (m_VtxDesc.Tex2MatIdx) {m_VertexSize += 1; m_NativeFmt->m_components |= VB_HAS_TEXMTXIDX2; WriteCall(TexMtx_ReadDirect_UByte); }
273 if (m_VtxDesc.Tex3MatIdx) {m_VertexSize += 1; m_NativeFmt->m_components |= VB_HAS_TEXMTXIDX3; WriteCall(TexMtx_ReadDirect_UByte); }
274 if (m_VtxDesc.Tex4MatIdx) {m_VertexSize += 1; m_NativeFmt->m_components |= VB_HAS_TEXMTXIDX4; WriteCall(TexMtx_ReadDirect_UByte); }
275 if (m_VtxDesc.Tex5MatIdx) {m_VertexSize += 1; m_NativeFmt->m_components |= VB_HAS_TEXMTXIDX5; WriteCall(TexMtx_ReadDirect_UByte); }
276 if (m_VtxDesc.Tex6MatIdx) {m_VertexSize += 1; m_NativeFmt->m_components |= VB_HAS_TEXMTXIDX6; WriteCall(TexMtx_ReadDirect_UByte); }
277 if (m_VtxDesc.Tex7MatIdx) {m_VertexSize += 1; m_NativeFmt->m_components |= VB_HAS_TEXMTXIDX7; WriteCall(TexMtx_ReadDirect_UByte); }
278
279 // Write vertex position loader
280 WriteCall(VertexLoader_Position::GetFunction(m_VtxDesc.Position, m_VtxAttr.PosFormat, m_VtxAttr.PosElements));
281 m_VertexSize += VertexLoader_Position::GetSize(m_VtxDesc.Position, m_VtxAttr.PosFormat, m_VtxAttr.PosElements);
282 nat_offset += 12;
283
284 // OK, so we just got a point. Let's go back and read it for the bounding box.
285
286 #ifdef BBOX_SUPPORT
287 WriteCall(UpdateBoundingBox);
288 #endif
289
290 // Normals
291 vtx_decl.num_normals = 0;
292 if (m_VtxDesc.Normal != NOT_PRESENT) {
293 m_VertexSize += VertexLoader_Normal::GetSize(m_VtxDesc.Normal, m_VtxAttr.NormalFormat, m_VtxAttr.NormalElements, m_VtxAttr.NormalIndex3);
294 TPipelineFunction pFunc = VertexLoader_Normal::GetFunction(m_VtxDesc.Normal, m_VtxAttr.NormalFormat, m_VtxAttr.NormalElements, m_VtxAttr.NormalIndex3, g_Config.bAllowSignedBytes);
295 if (pFunc == 0)
296 {
297 char temp[256];
298 sprintf(temp,"%i %i %i %i", m_VtxDesc.Normal, m_VtxAttr.NormalFormat, m_VtxAttr.NormalElements, m_VtxAttr.NormalIndex3);
299 g_VideoInitialize.pSysMessage("VertexLoader_Normal::GetFunction returned zero!");
300 }
301 WriteCall(pFunc);
302
303 vtx_decl.num_normals = vtx_attr.NormalElements ? 3 : 1;
304 vtx_decl.normal_offset[0] = -1;
305 vtx_decl.normal_offset[1] = -1;
306 vtx_decl.normal_offset[2] = -1;
307 switch (vtx_attr.NormalFormat) {
308 case FORMAT_UBYTE:
309 case FORMAT_BYTE:
310 {
311 vtx_decl.normal_gl_type = VAR_BYTE;
312 int native_size = 4;
313 if (vtx_attr.NormalFormat == FORMAT_BYTE && !g_Config.bAllowSignedBytes)
314 {
315 vtx_decl.normal_gl_type = VAR_SHORT;
316 native_size = 8;
317 }
318 vtx_decl.normal_gl_size = 4;
319 vtx_decl.normal_offset[0] = nat_offset;
320 nat_offset += native_size;
321 if (vtx_attr.NormalElements) {
322 vtx_decl.normal_offset[1] = nat_offset;
323 nat_offset += native_size;
324 vtx_decl.normal_offset[2] = nat_offset;
325 nat_offset += native_size;
326 }
327 break;
328 }
329 case FORMAT_USHORT:
330 case FORMAT_SHORT:
331 vtx_decl.normal_gl_type = VAR_SHORT;
332 vtx_decl.normal_gl_size = 4;
333 vtx_decl.normal_offset[0] = nat_offset;
334 nat_offset += 8;
335 if (vtx_attr.NormalElements) {
336 vtx_decl.normal_offset[1] = nat_offset;
337 nat_offset += 8;
338 vtx_decl.normal_offset[2] = nat_offset;
339 nat_offset += 8;
340 }
341 break;
342 case FORMAT_FLOAT:
343 vtx_decl.normal_gl_type = VAR_FLOAT;
344 vtx_decl.normal_gl_size = 3;
345 vtx_decl.normal_offset[0] = nat_offset;
346 nat_offset += 12;
347 if (vtx_attr.NormalElements) {
348 vtx_decl.normal_offset[1] = nat_offset;
349 nat_offset += 12;
350 vtx_decl.normal_offset[2] = nat_offset;
351 nat_offset += 12;
352 }
353 break;
354 default: _assert_(0); break;
355 }
356
357 int numNormals = (m_VtxAttr.NormalElements == 1) ? NRM_THREE : NRM_ONE;
358 m_NativeFmt->m_components |= VB_HAS_NRM0;
359
360 if (numNormals == NRM_THREE)
361 m_NativeFmt->m_components |= VB_HAS_NRM1 | VB_HAS_NRM2;
362 }
363
364 vtx_decl.color_gl_type = VAR_UNSIGNED_BYTE;
365 vtx_decl.color_offset[0] = -1;
366 vtx_decl.color_offset[1] = -1;
367 for (int i = 0; i < 2; i++) {
368 m_NativeFmt->m_components |= VB_HAS_COL0 << i;
369 switch (col[i])
370 {
371 case NOT_PRESENT:
372 m_NativeFmt->m_components &= ~(VB_HAS_COL0 << i);
373 vtx_decl.color_offset[i] = -1;
374 break;
375 case DIRECT:
376 switch (m_VtxAttr.color[i].Comp)
377 {
378 case FORMAT_16B_565: m_VertexSize += 2; WriteCall(Color_ReadDirect_16b_565); break;
379 case FORMAT_24B_888: m_VertexSize += 3; WriteCall(Color_ReadDirect_24b_888); break;
380 case FORMAT_32B_888x: m_VertexSize += 4; WriteCall(Color_ReadDirect_32b_888x); break;
381 case FORMAT_16B_4444: m_VertexSize += 2; WriteCall(Color_ReadDirect_16b_4444); break;
382 case FORMAT_24B_6666: m_VertexSize += 3; WriteCall(Color_ReadDirect_24b_6666); break;
383 case FORMAT_32B_8888: m_VertexSize += 4; WriteCall(Color_ReadDirect_32b_8888); break;
384 default: _assert_(0); break;
385 }
386 break;
387 case INDEX8:
388 m_VertexSize += 1;
389 switch (m_VtxAttr.color[i].Comp)
390 {
391 case FORMAT_16B_565: WriteCall(Color_ReadIndex8_16b_565); break;
392 case FORMAT_24B_888: WriteCall(Color_ReadIndex8_24b_888); break;
393 case FORMAT_32B_888x: WriteCall(Color_ReadIndex8_32b_888x); break;
394 case FORMAT_16B_4444: WriteCall(Color_ReadIndex8_16b_4444); break;
395 case FORMAT_24B_6666: WriteCall(Color_ReadIndex8_24b_6666); break;
396 case FORMAT_32B_8888: WriteCall(Color_ReadIndex8_32b_8888); break;
397 default: _assert_(0); break;
398 }
399 break;
400 case INDEX16:
401 m_VertexSize += 2;
402 switch (m_VtxAttr.color[i].Comp)
403 {
404 case FORMAT_16B_565: WriteCall(Color_ReadIndex16_16b_565); break;
405 case FORMAT_24B_888: WriteCall(Color_ReadIndex16_24b_888); break;
406 case FORMAT_32B_888x: WriteCall(Color_ReadIndex16_32b_888x); break;
407 case FORMAT_16B_4444: WriteCall(Color_ReadIndex16_16b_4444); break;
408 case FORMAT_24B_6666: WriteCall(Color_ReadIndex16_24b_6666); break;
409 case FORMAT_32B_8888: WriteCall(Color_ReadIndex16_32b_8888); break;
410 default: _assert_(0); break;
411 }
412 break;
413 }
414 // Common for the three bottom cases
415 if (col[i] != NOT_PRESENT) {
416 vtx_decl.color_offset[i] = nat_offset;
417 nat_offset += 4;
418 }
419 }
420
421 // Texture matrix indices (remove if corresponding texture coordinate isn't enabled)
422 for (int i = 0; i < 8; i++) {
423 vtx_decl.texcoord_offset[i] = -1;
424 const int format = m_VtxAttr.texCoord[i].Format;
425 const int elements = m_VtxAttr.texCoord[i].Elements;
426
427 if (tc[i] == NOT_PRESENT) {
428 m_NativeFmt->m_components &= ~(VB_HAS_UV0 << i);
429 } else {
430 _assert_msg_(VIDEO, DIRECT <= tc[i] && tc[i] <= INDEX16, "Invalid texture coordinates!\n(tc[i] = %d)", tc[i]);
431 _assert_msg_(VIDEO, FORMAT_UBYTE <= format && format <= FORMAT_FLOAT, "Invalid texture coordinates format!\n(format = %d)", format);
432 _assert_msg_(VIDEO, 0 <= elements && elements <= 1, "Invalid number of texture coordinates elemnts!\n(elements = %d)", elements);
433
434 m_NativeFmt->m_components |= VB_HAS_UV0 << i;
435 WriteCall(VertexLoader_TextCoord::GetFunction(tc[i], format, elements));
436 m_VertexSize += VertexLoader_TextCoord::GetSize(tc[i], format, elements);
437 }
438
439 if (m_NativeFmt->m_components & (VB_HAS_TEXMTXIDX0 << i)) {
440 if (tc[i] != NOT_PRESENT) {
441 // if texmtx is included, texcoord will always be 3 floats, z will be the texmtx index
442 vtx_decl.texcoord_offset[i] = nat_offset;
443 vtx_decl.texcoord_gl_type[i] = VAR_FLOAT;
444 vtx_decl.texcoord_size[i] = 3;
445 nat_offset += 12;
446 WriteCall(m_VtxAttr.texCoord[i].Elements ? TexMtx_Write_Float : TexMtx_Write_Float2);
447 }
448 else {
449 m_NativeFmt->m_components |= VB_HAS_UV0 << i; // have to include since using now
450 vtx_decl.texcoord_offset[i] = nat_offset;
451 vtx_decl.texcoord_gl_type[i] = VAR_FLOAT;
452 vtx_decl.texcoord_size[i] = 4;
453 nat_offset += 16; // still include the texture coordinate, but this time as 6 + 2 bytes
454 WriteCall(TexMtx_Write_Float4);
455 }
456 }
457 else {
458 if (tc[i] != NOT_PRESENT) {
459 vtx_decl.texcoord_offset[i] = nat_offset;
460 vtx_decl.texcoord_gl_type[i] = VAR_FLOAT;
461 vtx_decl.texcoord_size[i] = vtx_attr.texCoord[i].Elements ? 2 : 1;
462 nat_offset += 4 * (vtx_attr.texCoord[i].Elements ? 2 : 1);
463 }
464 }
465
466 if (tc[i] == NOT_PRESENT) {
467 // if there's more tex coords later, have to write a dummy call
468 int j = i + 1;
469 for (; j < 8; ++j) {
470 if (tc[j] != NOT_PRESENT) {
471 WriteCall(VertexLoader_TextCoord::GetDummyFunction()); // important to get indices right!
472 break;
473 }
474 }
475 // tricky!
476 if (j == 8 && !((m_NativeFmt->m_components & VB_HAS_TEXMTXIDXALL) & (VB_HAS_TEXMTXIDXALL << (i + 1)))) {
477 // no more tex coords and tex matrices, so exit loop
478 break;
479 }
480 }
481 }
482
483 if (m_VtxDesc.PosMatIdx) {
484 WriteCall(PosMtx_Write);
485 vtx_decl.posmtx_offset = nat_offset;
486 nat_offset += 4;
487 } else {
488 vtx_decl.posmtx_offset = -1;
489 }
490
491 native_stride = nat_offset;
492 vtx_decl.stride = native_stride;
493
494 #ifdef USE_JIT
495 // End loop here
496 #ifdef _M_X64
497 MOV(64, R(RAX), Imm64((u64)&loop_counter));
498 SUB(32, MatR(RAX), Imm8(1));
499 #else
500 SUB(32, M(&loop_counter), Imm8(1));
501 #endif
502
503 J_CC(CC_NZ, loop_start, true);
504 ABI_EmitEpilogue(4);
505 #endif
506 m_NativeFmt->Initialize(vtx_decl);
507 }
508
509 void VertexLoader::WriteCall(TPipelineFunction func)
510 {
511 #ifdef USE_JIT
512 #ifdef _M_X64
513 MOV(64, R(RAX), Imm64((u64)func));
514 CALLptr(R(RAX));
515 #else
516 CALL((void*)func);
517 #endif
518 #else
519 m_PipelineStages[m_numPipelineStages++] = func;
520 #endif
521 }
522
523 void VertexLoader::WriteGetVariable(int bits, OpArg dest, void *address)
524 {
525 #ifdef USE_JIT
526 #ifdef _M_X64
527 MOV(64, R(RAX), Imm64((u64)address));
528 MOV(bits, dest, MatR(RAX));
529 #else
530 MOV(bits, dest, M(address));
531 #endif
532 #endif
533 }
534
535 void VertexLoader::WriteSetVariable(int bits, void *address, OpArg value)
536 {
537 #ifdef USE_JIT
538 #ifdef _M_X64
539 MOV(64, R(RAX), Imm64((u64)address));
540 MOV(bits, MatR(RAX), value);
541 #else
542 MOV(bits, M(address), value);
543 #endif
544 #endif
545 }
546
547 void VertexLoader::RunVertices(int vtx_attr_group, int primitive, int count)
548 {
549 DVSTARTPROFILE();
550
551 m_numLoadedVertices += count;
552
553 // Flush if our vertex format is different from the currently set.
554 if (g_nativeVertexFmt != NULL && g_nativeVertexFmt != m_NativeFmt)
555 {
556 // We really must flush here. It's possible that the native representations
557 // of the two vtx formats are the same, but we have no way to easily check that
558 // now.
559 VertexManager::Flush();
560 // Also move the Set() here?
561 }
562 g_nativeVertexFmt = m_NativeFmt;
563
564 if (bpmem.genMode.cullmode == 3 && primitive < 5)
565 {
566 // if cull mode is none, ignore triangles and quads
567 DataSkip(count * m_VertexSize);
568 return;
569 }
570
571 m_NativeFmt->EnableComponents(m_NativeFmt->m_components);
572
573 // Load position and texcoord scale factors.
574 m_VtxAttr.PosFrac = g_VtxAttr[vtx_attr_group].g0.PosFrac;
575 m_VtxAttr.texCoord[0].Frac = g_VtxAttr[vtx_attr_group].g0.Tex0Frac;
576 m_VtxAttr.texCoord[1].Frac = g_VtxAttr[vtx_attr_group].g1.Tex1Frac;
577 m_VtxAttr.texCoord[2].Frac = g_VtxAttr[vtx_attr_group].g1.Tex2Frac;
578 m_VtxAttr.texCoord[3].Frac = g_VtxAttr[vtx_attr_group].g1.Tex3Frac;
579 m_VtxAttr.texCoord[4].Frac = g_VtxAttr[vtx_attr_group].g2.Tex4Frac;
580 m_VtxAttr.texCoord[5].Frac = g_VtxAttr[vtx_attr_group].g2.Tex5Frac;
581 m_VtxAttr.texCoord[6].Frac = g_VtxAttr[vtx_attr_group].g2.Tex6Frac;
582 m_VtxAttr.texCoord[7].Frac = g_VtxAttr[vtx_attr_group].g2.Tex7Frac;
583
584 pVtxAttr = &m_VtxAttr;
585 posScale = fractionTable[m_VtxAttr.PosFrac];
586 if (m_NativeFmt->m_components & VB_HAS_UVALL)
587 for (int i = 0; i < 8; i++)
588 tcScale[i] = fractionTable[m_VtxAttr.texCoord[i].Frac];
589 for (int i = 0; i < 2; i++)
590 colElements[i] = m_VtxAttr.color[i].Elements;
591
592 // if strips or fans, make sure all vertices can fit in buffer, otherwise flush
593 int granularity = 1;
594 switch (primitive) {
595 case 3: // strip .. hm, weird
596 case 4: // fan
597 if (VertexManager::GetRemainingSize() < 3 * native_stride)
598 VertexManager::Flush();
599 break;
600 case 6: // line strip
601 if (VertexManager::GetRemainingSize() < 2 * native_stride)
602 VertexManager::Flush();
603 break;
604 case 0: granularity = 4; break; // quads
605 case 2: granularity = 3; break; // tris
606 case 5: granularity = 2; break; // lines
607 }
608
609 int startv = 0, extraverts = 0;
610 int v = 0;
611
612 //int remainingVerts2 = VertexManager::GetRemainingVertices(primitive);
613 while (v < count)
614 {
615 int remainingVerts = VertexManager::GetRemainingSize() / native_stride;
616 //if (remainingVerts2 - v + startv < remainingVerts)
617 //remainingVerts = remainingVerts2 - v + startv;
618 if (remainingVerts < granularity) {
619 INCSTAT(stats.thisFrame.numBufferSplits);
620 // This buffer full - break current primitive and flush, to switch to the next buffer.
621 u8* plastptr = VertexManager::s_pCurBufferPointer;
622 if (v - startv > 0)
623 VertexManager::AddVertices(primitive, v - startv + extraverts);
624 VertexManager::Flush();
625 //remainingVerts2 = VertexManager::GetRemainingVertices(primitive);
626 // Why does this need to be so complicated?
627 switch (primitive) {
628 case 3: // triangle strip, copy last two vertices
629 // a little trick since we have to keep track of signs
630 if (v & 1) {
631 memcpy_gc(VertexManager::s_pCurBufferPointer, plastptr-2*native_stride, native_stride);
632 memcpy_gc(VertexManager::s_pCurBufferPointer+native_stride, plastptr-native_stride*2, 2*native_stride);
633 VertexManager::s_pCurBufferPointer += native_stride*3;
634 extraverts = 3;
635 }
636 else {
637 memcpy_gc(VertexManager::s_pCurBufferPointer, plastptr-native_stride*2, native_stride*2);
638 VertexManager::s_pCurBufferPointer += native_stride*2;
639 extraverts = 2;
640 }
641 break;
642 case 4: // tri fan, copy first and last vert
643 memcpy_gc(VertexManager::s_pCurBufferPointer, plastptr-native_stride*(v-startv+extraverts), native_stride);
644 VertexManager::s_pCurBufferPointer += native_stride;
645 memcpy_gc(VertexManager::s_pCurBufferPointer, plastptr-native_stride, native_stride);
646 VertexManager::s_pCurBufferPointer += native_stride;
647 extraverts = 2;
648 break;
649 case 6: // line strip
650 memcpy_gc(VertexManager::s_pCurBufferPointer, plastptr-native_stride, native_stride);
651 VertexManager::s_pCurBufferPointer += native_stride;
652 extraverts = 1;
653 break;
654 default:
655 extraverts = 0;
656 break;
657 }
658 startv = v;
659 }
660 int remainingPrims = remainingVerts / granularity;
661 remainingVerts = remainingPrims * granularity;
662 if (count - v < remainingVerts)
663 remainingVerts = count - v;
664
665 #ifdef USE_JIT
666 if (remainingVerts > 0) {
667 loop_counter = remainingVerts;
668 ((void (*)())(void*)m_compiledCode)();
669 }
670 #else
671 for (int s = 0; s < remainingVerts; s++)
672 {
673 tcIndex = 0;
674 colIndex = 0;
675 s_texmtxwrite = s_texmtxread = 0;
676 for (int i = 0; i < m_numPipelineStages; i++)
677 m_PipelineStages[i]();
678 PRIM_LOG("\n");
679 }
680 #endif
681 v += remainingVerts;
682 }
683
684 if (startv < count)
685 VertexManager::AddVertices(primitive, count - startv + extraverts);
686 }
687
688
689
690
691 void VertexLoader::RunCompiledVertices(int vtx_attr_group, int primitive, int count, u8* Data)
692 {
693 DVSTARTPROFILE();
694
695 m_numLoadedVertices += count;
696
697 // Flush if our vertex format is different from the currently set.
698 if (g_nativeVertexFmt != NULL && g_nativeVertexFmt != m_NativeFmt)
699 {
700 // We really must flush here. It's possible that the native representations
701 // of the two vtx formats are the same, but we have no way to easily check that
702 // now.
703 VertexManager::Flush();
704 // Also move the Set() here?
705 }
706 g_nativeVertexFmt = m_NativeFmt;
707
708 if (bpmem.genMode.cullmode == 3 && primitive < 5)
709 {
710 // if cull mode is none, ignore triangles and quads
711 DataSkip(count * m_VertexSize);
712 return;
713 }
714
715 m_NativeFmt->EnableComponents(m_NativeFmt->m_components);
716
717 // Load position and texcoord scale factors.
718 m_VtxAttr.PosFrac = g_VtxAttr[vtx_attr_group].g0.PosFrac;
719 m_VtxAttr.texCoord[0].Frac = g_VtxAttr[vtx_attr_group].g0.Tex0Frac;
720 m_VtxAttr.texCoord[1].Frac = g_VtxAttr[vtx_attr_group].g1.Tex1Frac;
721 m_VtxAttr.texCoord[2].Frac = g_VtxAttr[vtx_attr_group].g1.Tex2Frac;
722 m_VtxAttr.texCoord[3].Frac = g_VtxAttr[vtx_attr_group].g1.Tex3Frac;
723 m_VtxAttr.texCoord[4].Frac = g_VtxAttr[vtx_attr_group].g2.Tex4Frac;
724 m_VtxAttr.texCoord[5].Frac = g_VtxAttr[vtx_attr_group].g2.Tex5Frac;
725 m_VtxAttr.texCoord[6].Frac = g_VtxAttr[vtx_attr_group].g2.Tex6Frac;
726 m_VtxAttr.texCoord[7].Frac = g_VtxAttr[vtx_attr_group].g2.Tex7Frac;
727
728 pVtxAttr = &m_VtxAttr;
729 posScale = fractionTable[m_VtxAttr.PosFrac];
730 if (m_NativeFmt->m_components & VB_HAS_UVALL)
731 for (int i = 0; i < 8; i++)
732 tcScale[i] = fractionTable[m_VtxAttr.texCoord[i].Frac];
733 for (int i = 0; i < 2; i++)
734 colElements[i] = m_VtxAttr.color[i].Elements;
735
736 if(VertexManager::GetRemainingSize() < native_stride * count)
737 VertexManager::Flush();
738 memcpy_gc(VertexManager::s_pCurBufferPointer, Data, native_stride * count);
739 VertexManager::s_pCurBufferPointer += native_stride * count;
740 DataSkip(count * m_VertexSize);
741 VertexManager::AddVertices(primitive, count);
742 }
743
744
745
746 void VertexLoader::SetVAT(u32 _group0, u32 _group1, u32 _group2)
747 {
748 VAT vat;
749 vat.g0.Hex = _group0;
750 vat.g1.Hex = _group1;
751 vat.g2.Hex = _group2;
752
753 m_VtxAttr.PosElements = vat.g0.PosElements;
754 m_VtxAttr.PosFormat = vat.g0.PosFormat;
755 m_VtxAttr.PosFrac = vat.g0.PosFrac;
756 m_VtxAttr.NormalElements = vat.g0.NormalElements;
757 m_VtxAttr.NormalFormat = vat.g0.NormalFormat;
758 m_VtxAttr.color[0].Elements = vat.g0.Color0Elements;
759 m_VtxAttr.color[0].Comp = vat.g0.Color0Comp;
760 m_VtxAttr.color[1].Elements = vat.g0.Color1Elements;
761 m_VtxAttr.color[1].Comp = vat.g0.Color1Comp;
762 m_VtxAttr.texCoord[0].Elements = vat.g0.Tex0CoordElements;
763 m_VtxAttr.texCoord[0].Format = vat.g0.Tex0CoordFormat;
764 m_VtxAttr.texCoord[0].Frac = vat.g0.Tex0Frac;
765 m_VtxAttr.ByteDequant = vat.g0.ByteDequant;
766 m_VtxAttr.NormalIndex3 = vat.g0.NormalIndex3;
767
768 m_VtxAttr.texCoord[1].Elements = vat.g1.Tex1CoordElements;
769 m_VtxAttr.texCoord[1].Format = vat.g1.Tex1CoordFormat;
770 m_VtxAttr.texCoord[1].Frac = vat.g1.Tex1Frac;
771 m_VtxAttr.texCoord[2].Elements = vat.g1.Tex2CoordElements;
772 m_VtxAttr.texCoord[2].Format = vat.g1.Tex2CoordFormat;
773 m_VtxAttr.texCoord[2].Frac = vat.g1.Tex2Frac;
774 m_VtxAttr.texCoord[3].Elements = vat.g1.Tex3CoordElements;
775 m_VtxAttr.texCoord[3].Format = vat.g1.Tex3CoordFormat;
776 m_VtxAttr.texCoord[3].Frac = vat.g1.Tex3Frac;
777 m_VtxAttr.texCoord[4].Elements = vat.g1.Tex4CoordElements;
778 m_VtxAttr.texCoord[4].Format = vat.g1.Tex4CoordFormat;
779
780 m_VtxAttr.texCoord[4].Frac = vat.g2.Tex4Frac;
781 m_VtxAttr.texCoord[5].Elements = vat.g2.Tex5CoordElements;
782 m_VtxAttr.texCoord[5].Format = vat.g2.Tex5CoordFormat;
783 m_VtxAttr.texCoord[5].Frac = vat.g2.Tex5Frac;
784 m_VtxAttr.texCoord[6].Elements = vat.g2.Tex6CoordElements;
785 m_VtxAttr.texCoord[6].Format = vat.g2.Tex6CoordFormat;
786 m_VtxAttr.texCoord[6].Frac = vat.g2.Tex6Frac;
787 m_VtxAttr.texCoord[7].Elements = vat.g2.Tex7CoordElements;
788 m_VtxAttr.texCoord[7].Format = vat.g2.Tex7CoordFormat;
789 m_VtxAttr.texCoord[7].Frac = vat.g2.Tex7Frac;
790 };
791
792 void VertexLoader::AppendToString(std::string *dest) const
793 {
794 dest->reserve(250);
795 static const char *posMode[4] = {
796 "Inv",
797 "Dir",
798 "I8",
799 "I16",
800 };
801 static const char *posFormats[5] = {
802 "u8", "s8", "u16", "s16", "flt",
803 };
804 static const char *colorFormat[8] = {
805 "565",
806 "888",
807 "888x",
808 "4444",
809 "6666",
810 "8888",
811 "Inv",
812 "Inv",
813 };
814
815 dest->append(StringFromFormat("%ib skin: %i P: %i %s-%s ",
816 m_VertexSize, m_VtxDesc.PosMatIdx,
817 m_VtxAttr.PosElements ? 3 : 2, posMode[m_VtxDesc.Position], posFormats[m_VtxAttr.PosFormat]));
818 if (m_VtxDesc.Normal) {
819 dest->append(StringFromFormat("Nrm: %i %s-%s ",
820 m_VtxAttr.NormalElements, posMode[m_VtxDesc.Normal], posFormats[m_VtxAttr.NormalFormat]));
821 }
822 int color_mode[2] = {m_VtxDesc.Color0, m_VtxDesc.Color1};
823 for (int i = 0; i < 2; i++)
824 {
825 if (color_mode[i])
826 {
827 dest->append(StringFromFormat("C%i: %i %s-%s ", i, m_VtxAttr.color[i].Elements, posMode[color_mode[i]], colorFormat[m_VtxAttr.color[i].Comp]));
828 }
829 }
830 int tex_mode[8] = {
831 m_VtxDesc.Tex0Coord, m_VtxDesc.Tex1Coord, m_VtxDesc.Tex2Coord, m_VtxDesc.Tex3Coord,
832 m_VtxDesc.Tex4Coord, m_VtxDesc.Tex5Coord, m_VtxDesc.Tex6Coord, m_VtxDesc.Tex7Coord
833 };
834 for (int i = 0; i < 8; i++)
835 {
836 if (tex_mode[i])
837 {
838 dest->append(StringFromFormat("T%i: %i %s-%s ",
839 i, m_VtxAttr.texCoord[i].Elements, posMode[tex_mode[i]], posFormats[m_VtxAttr.texCoord[i].Format]));
840 }
841 }
842 dest->append(StringFromFormat(" - %i v\n", m_numLoadedVertices));
843 }
Dolphin, the Gamecube and Wii emulator