d07c171f |
1 | /* |
2 | Copyright (C) 2003 Rice1964 |
3 | |
4 | This program is free software; you can redistribute it and/or |
5 | modify it under the terms of the GNU General Public License |
6 | as published by the Free Software Foundation; either version 2 |
7 | of the License, or (at your option) any later version. |
8 | |
9 | This program is distributed in the hope that it will be useful, |
10 | but WITHOUT ANY WARRANTY; without even the implied warranty of |
11 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the |
12 | GNU General Public License for more details. |
13 | |
14 | You should have received a copy of the GNU General Public License |
15 | along with this program; if not, write to the Free Software |
16 | Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA. |
17 | |
18 | */ |
19 | |
20 | #include <cmath> |
21 | #include <vector> |
22 | |
23 | #include "osal_preproc.h" |
24 | #include "float.h" |
25 | #include "DeviceBuilder.h" |
26 | #include "VertexShaderConstantDef.h" |
27 | #include "Render.h" |
28 | #include "Timing.h" |
29 | |
30 | extern FiddledVtx * g_pVtxBase; |
31 | |
32 | #define ENABLE_CLIP_TRI |
33 | #define X_CLIP_MAX 0x1 |
34 | #define X_CLIP_MIN 0x2 |
35 | #define Y_CLIP_MAX 0x4 |
36 | #define Y_CLIP_MIN 0x8 |
37 | #define Z_CLIP_MAX 0x10 |
38 | #define Z_CLIP_MIN 0x20 |
39 | |
40 | #ifdef ENABLE_CLIP_TRI |
41 | |
42 | inline void RSP_Vtx_Clipping(int i) |
43 | { |
44 | g_clipFlag[i] = 0; |
45 | g_clipFlag2[i] = 0; |
46 | if( g_vecProjected[i].w > 0 ) |
47 | { |
48 | /* |
49 | if( gRSP.bRejectVtx ) |
50 | { |
51 | if( g_vecProjected[i].x > 1 ) |
52 | { |
53 | g_clipFlag2[i] |= X_CLIP_MAX; |
54 | if( g_vecProjected[i].x > gRSP.real_clip_ratio_posx ) |
55 | g_clipFlag[i] |= X_CLIP_MAX; |
56 | } |
57 | |
58 | if( g_vecProjected[i].x < -1 ) |
59 | { |
60 | g_clipFlag2[i] |= X_CLIP_MIN; |
61 | if( g_vecProjected[i].x < gRSP.real_clip_ratio_negx ) |
62 | g_clipFlag[i] |= X_CLIP_MIN; |
63 | } |
64 | |
65 | if( g_vecProjected[i].y > 1 ) |
66 | { |
67 | g_clipFlag2[i] |= Y_CLIP_MAX; |
68 | if( g_vecProjected[i].y > gRSP.real_clip_ratio_posy ) |
69 | g_clipFlag[i] |= Y_CLIP_MAX; |
70 | } |
71 | |
72 | if( g_vecProjected[i].y < -1 ) |
73 | { |
74 | g_clipFlag2[i] |= Y_CLIP_MIN; |
75 | if( g_vecProjected[i].y < gRSP.real_clip_ratio_negy ) |
76 | g_clipFlag[i] |= Y_CLIP_MIN; |
77 | } |
78 | |
79 | //if( g_vecProjected[i].z > 1.0f ) |
80 | //{ |
81 | // g_clipFlag2[i] |= Z_CLIP_MAX; |
82 | // g_clipFlag[i] |= Z_CLIP_MAX; |
83 | //} |
84 | |
85 | //if( gRSP.bNearClip && g_vecProjected[i].z < -1.0f ) |
86 | //{ |
87 | // g_clipFlag2[i] |= Z_CLIP_MIN; |
88 | // g_clipFlag[i] |= Z_CLIP_MIN; |
89 | //} |
90 | } |
91 | else |
92 | */ |
93 | { |
94 | if( g_vecProjected[i].x > 1 ) g_clipFlag2[i] |= X_CLIP_MAX; |
95 | if( g_vecProjected[i].x < -1 ) g_clipFlag2[i] |= X_CLIP_MIN; |
96 | if( g_vecProjected[i].y > 1 ) g_clipFlag2[i] |= Y_CLIP_MAX; |
97 | if( g_vecProjected[i].y < -1 ) g_clipFlag2[i] |= Y_CLIP_MIN; |
98 | //if( g_vecProjected[i].z > 1.0f ) g_clipFlag2[i] |= Z_CLIP_MAX; |
99 | //if( gRSP.bNearClip && g_vecProjected[i].z < -1.0f ) g_clipFlag2[i] |= Z_CLIP_MIN; |
100 | } |
101 | |
102 | } |
103 | } |
104 | |
105 | #else |
106 | inline void RSP_Vtx_Clipping(int i) {} |
107 | #endif |
108 | |
109 | /* |
110 | * Global variables |
111 | */ |
112 | ALIGN(16,RSP_Options gRSP) |
113 | ALIGN(16,RDP_Options gRDP) |
114 | |
115 | static ALIGN(16,XVECTOR4 g_normal) |
116 | //static int norms[3]; |
117 | |
118 | ALIGN(16,XVECTOR4 g_vtxNonTransformed[MAX_VERTS]) |
119 | ALIGN(16,XVECTOR4 g_vecProjected[MAX_VERTS]) |
120 | ALIGN(16,XVECTOR4 g_vtxTransformed[MAX_VERTS]) |
121 | |
122 | float g_vtxProjected5[1000][5]; |
123 | float g_vtxProjected5Clipped[2000][5]; |
124 | |
125 | //uint32 g_dwVtxFlags[MAX_VERTS]; // Z_POS Z_NEG etc |
126 | VECTOR2 g_fVtxTxtCoords[MAX_VERTS]; |
127 | uint32 g_dwVtxDifColor[MAX_VERTS]; |
128 | uint32 g_clipFlag[MAX_VERTS]; |
129 | uint32 g_clipFlag2[MAX_VERTS]; |
130 | RenderTexture g_textures[MAX_TEXTURES]; |
131 | float g_fFogCoord[MAX_VERTS]; |
132 | |
133 | EXTERNAL_VERTEX g_vtxForExternal[MAX_VERTS]; |
134 | |
135 | TLITVERTEX g_vtxBuffer[1000]; |
136 | TLITVERTEX g_clippedVtxBuffer[2000]; |
137 | uint8 g_oglVtxColors[1000][4]; |
138 | int g_clippedVtxCount=0; |
139 | TLITVERTEX g_texRectTVtx[4]; |
140 | unsigned short g_vtxIndex[1000]; |
141 | unsigned int g_minIndex, g_maxIndex; |
142 | |
143 | float gRSPfFogMin; |
144 | float gRSPfFogMax; |
145 | float gRSPfFogDivider; |
146 | |
147 | uint32 gRSPnumLights; |
148 | Light gRSPlights[16]; |
149 | |
150 | ALIGN(16,Matrix gRSPworldProjectTransported) |
151 | ALIGN(16,Matrix gRSPworldProject) |
152 | ALIGN(16,Matrix gRSPmodelViewTop) |
153 | ALIGN(16,Matrix gRSPmodelViewTopTranspose) |
154 | ALIGN(16,Matrix dkrMatrixTransposed) |
155 | |
156 | N64Light gRSPn64lights[16]; |
157 | |
158 | |
159 | void (*ProcessVertexData)(uint32 dwAddr, uint32 dwV0, uint32 dwNum)=NULL; |
160 | |
161 | /* |
162 | * |
163 | */ |
164 | |
165 | |
166 | /*n.x = (g_normal.x * matWorld.m00) + (g_normal.y * matWorld.m10) + (g_normal.z * matWorld.m20); |
167 | n.y = (g_normal.x * matWorld.m01) + (g_normal.y * matWorld.m11) + (g_normal.z * matWorld.m21); |
168 | n.z = (g_normal.x * matWorld.m02) + (g_normal.y * matWorld.m12) + (g_normal.z * matWorld.m22);*/ |
169 | |
170 | // Multiply (x,y,z,0) by matrix m, then normalize |
171 | #if defined(__INTEL_COMPILER) && !defined(NO_ASM) |
172 | #define Vec3TransformNormal(vec, m) __asm \ |
173 | { \ |
174 | __asm fld dword ptr [vec + 0] \ |
175 | __asm fmul dword ptr [m + 0] \ /* x m00*/ |
176 | __asm fld dword ptr [vec + 0] \ |
177 | __asm fmul dword ptr [m + 4] \ /* x m01 x m00*/ |
178 | __asm fld dword ptr [vec + 0] \ |
179 | __asm fmul dword ptr [m + 8] \ /* x m02 x m01 x m00*/ |
180 | \ |
181 | __asm fld dword ptr [vec + 4] \ |
182 | __asm fmul dword ptr [m + 16] \ /* y m10 x m02 x m01 x m00*/ |
183 | __asm fld dword ptr [vec + 4] \ |
184 | __asm fmul dword ptr [m + 20] \ /* y m11 y m10 x m02 x m01 x m00*/ |
185 | __asm fld dword ptr [vec + 4] \ |
186 | __asm fmul dword ptr [m + 24] \ /* y m12 y m11 y m10 x m02 x m01 x m00*/ |
187 | \ |
188 | __asm fxch st(2) \ /* y m10 y m11 y m12 x m02 x m01 x m00*/ |
189 | __asm faddp st(5), st(0) \ /* y m11 y m12 x m02 x m01 (x m00 + y m10)*/ |
190 | __asm faddp st(3), st(0) \ /* y m12 x m02 (x m01 + ym11) (x m00 + y m10)*/ |
191 | __asm faddp st(1), st(0) \ /* (x m02 + y m12) (x m01 + ym11) (x m00 + y m10)*/ |
192 | \ |
193 | __asm fld dword ptr [vec + 8] \ |
194 | __asm fmul dword ptr [m + 32] \ /* z m20 (x m02 + y m12) (x m01 + ym11) (x m00 + y m10)*/ |
195 | __asm fld dword ptr [vec + 8] \ |
196 | __asm fmul dword ptr [m + 36] \ /* z m21 z m20 (x m02 + y m12) (x m01 + ym11) (x m00 + y m10)*/ |
197 | __asm fld dword ptr [vec + 8] \ |
198 | __asm fmul dword ptr [m + 40] \ /* z m22 z m21 z m20 (x m02 + y m12) (x m01 + ym11) (x m00 + y m10)*/ |
199 | \ |
200 | __asm fxch st(2) \ /* z m20 z m21 z m22 (x m02 + y m12) (x m01 + ym11) (x m00 + y m10)*/ |
201 | __asm faddp st(5), st(0) \ /* z m21 z m22 (x m02 + y m12) (x m01 + ym11) (x m00 + y m10 + z m20)*/ |
202 | __asm faddp st(3), st(0) \ /* z m22 (x m02 + y m12) (x m01 + ym11 + z m21) (x m00 + y m10 + z m20)*/ |
203 | __asm faddp st(1), st(0) \ /* (x m02 + y m12 + z m 22) (x m01 + ym11 + z m21) (x m00 + y m10 + z m20)*/ |
204 | \ |
205 | __asm fxch st(2) \ /* (x m00 + y m10 + z m20) (x m01 + ym11 + z m21) (x m02 + y m12 + z m 22) */ |
206 | \ |
207 | __asm fld1 \ /* 1 x y z */ |
208 | __asm fld st(1) \ /* x 1 x y z */ |
209 | __asm fmul st(0),st(0) \ /* xx 1 x y z */ |
210 | __asm fld st(3) \ /* y xx 1 x y z */ |
211 | __asm fmul st(0),st(0) \ /* yy xx 1 x y z */ |
212 | __asm fld st(5) \ /* z yy xx 1 x y z */ |
213 | __asm fmul st(0),st(0) \ /* zz yy xx 1 x y z */ |
214 | \ |
215 | __asm fxch st(2) \ /* xx yy zz 1 x y z */ |
216 | \ |
217 | __asm faddp st(1),st(0) \ /* (xx+yy) zz 1 x y z */ |
218 | __asm faddp st(1),st(0) \ /* (xx+yy+zz) 1 x y z */ |
219 | \ |
220 | __asm ftst \ /* Compare ST to 0 */ |
221 | __asm fstsw ax \ /* Store FPU status word in a */ |
222 | __asm sahf \ /* Transfer ax to flags register */ |
223 | __asm jz l2 \ /* Skip if length is zero */ |
224 | \ |
225 | __asm fsqrt \ /* l 1 x y z */ |
226 | \ |
227 | __asm fdivp st(1),st(0) \ /* (1/l) x y z */ |
228 | \ |
229 | __asm fmul st(3),st(0) \ /* f x y fz */ |
230 | __asm fmul st(2),st(0) \ /* f x fy fz */ |
231 | __asm fmulp st(1),st(0) \ /* fx fy fz */ |
232 | \ |
233 | __asm fstp dword ptr [vec + 0] \ /* fy fz*/ |
234 | __asm fstp dword ptr [vec + 4] \ /* fz */ |
235 | __asm fstp dword ptr [vec + 8] \ /* done */ |
236 | __asm jmp l3 \ |
237 | __asm l2: \ |
238 | __asm mov dword ptr [vec + 0], 0 \ |
239 | __asm mov dword ptr [vec + 4], 0 \ |
240 | __asm mov dword ptr [vec + 8], 0 \ |
241 | __asm l3: \ |
242 | } \ |
243 | |
244 | //#else // use C code in other cases, this is probably faster anyway |
245 | #elif defined(__ARM_NEON__0) |
246 | #define Vec3TransformNormal(vec, mtx) \ |
247 | { \ |
248 | asm volatile ( \ |
249 | "vld1.32 {d0}, [%0] \n\t" \ |
250 | "flds s2, [%0, #8] \n\t" \ |
251 | "vld1.32 {d18, d19}, [%1] \n\t" \ |
252 | "vld1.32 {d20, d21}, [%1, #16] \n\t" \ |
253 | "vld1.32 {d22, d23}, [%1, #32] \n\t" \ |
254 | \ |
255 | "vmul.f32 q2, q9, d0[0] \n\t" \ |
256 | "vmla.f32 q2, q10, d0[1] \n\t" \ |
257 | "vmla.f32 q2, q11, d1[0] \n\t" \ |
258 | \ |
259 | "vmul.f32 d0, d4, d4 \n\t" \ |
260 | "vpadd.f32 d0, d0, d0 \n\t" \ |
261 | "vmla.f32 d0, d5, d5 \n\t" \ |
262 | \ |
263 | "vmov.f32 d1, d0 \n\t" \ |
264 | "vrsqrte.f32 d0, d0 \n\t" \ |
265 | "vmul.f32 d2, d0, d1 \n\t" \ |
266 | "vrsqrts.f32 d3, d2, d0 \n\t" \ |
267 | "vmul.f32 d0, d0, d3 \n\t" \ |
268 | "vmul.f32 d2, d0, d1 \n\t" \ |
269 | "vrsqrts.f32 d3, d2, d0 \n\t" \ |
270 | "vmul.f32 d0, d0, d3 \n\t" \ |
271 | \ |
272 | "vmul.f32 q2, q2, d0[0] \n\t" \ |
273 | \ |
274 | "vst1.32 {d4}, [%0] \n\t" \ |
275 | "fsts s10, [%0, #8] \n\t" \ |
276 | : :"r"(vec), "r"(&mtx._11) \ |
277 | : "d0","d1","d2","d3","d18","d19","d20","d21","d22", "d23", "memory" \ |
278 | ); \ |
279 | } |
280 | #else |
281 | #define Vec3TransformNormal(vec, m) \ |
282 | VECTOR3 temp; \ |
283 | temp.x = (vec.x * m._11) + (vec.y * m._21) + (vec.z * m._31); \ |
284 | temp.y = (vec.x * m._12) + (vec.y * m._22) + (vec.z * m._32); \ |
285 | temp.z = (vec.x * m._13) + (vec.y * m._23) + (vec.z * m._33); \ |
286 | float norm = sqrtf(temp.x*temp.x+temp.y*temp.y+temp.z*temp.z); \ |
287 | if (norm == 0.0) { vec.x = 0.0; vec.y = 0.0; vec.z = 0.0;} else \ |
288 | { float rep=1/norm;vec.x = temp.x*rep; vec.y = temp.y*rep; vec.z = temp.z*rep; } |
289 | #endif |
290 | |
291 | |
292 | #if !defined(__GNUC__) && !defined(NO_ASM) |
293 | __declspec( naked ) void __fastcall SSEVec3Transform(int i) |
294 | { |
295 | __asm |
296 | { |
297 | shl ecx,4; // ecx = i |
298 | |
299 | movaps xmm1, DWORD PTR g_vtxNonTransformed [ecx]; // xmm1 as original vector |
300 | |
301 | movaps xmm4, DWORD PTR gRSPworldProjectTransported; // row1 |
302 | movaps xmm5, DWORD PTR gRSPworldProjectTransported[0x10]; // row2 |
303 | movaps xmm6, DWORD PTR gRSPworldProjectTransported[0x20]; // row3 |
304 | movaps xmm7, DWORD PTR gRSPworldProjectTransported[0x30]; // row4 |
305 | |
306 | mulps xmm4, xmm1; // row 1 |
307 | mulps xmm5, xmm1; // row 2 |
308 | mulps xmm6, xmm1; // row 3 |
309 | mulps xmm7, xmm1; // row 4 |
310 | |
311 | movhlps xmm0, xmm4; // xmm4 high to xmm0 low |
312 | movlhps xmm0, xmm5; // xmm5 low to xmm0 high |
313 | |
314 | addps xmm4, xmm0; // result of add are in xmm4 low |
315 | addps xmm5, xmm0; // result of add are in xmm5 high |
316 | |
317 | shufps xmm0, xmm4, 0x44; // move xmm4 low DWORDs to xmm0 high |
318 | shufps xmm4, xmm5, 0xe4; // move xmm5 high DWORS to xmm4 |
319 | movhlps xmm5, xmm0; // xmm4, xmm5 are mirrored |
320 | |
321 | shufps xmm4, xmm4, 0x08; // move xmm4's 3rd uint32 to its 2nd uint32 |
322 | shufps xmm5, xmm5, 0x0d; // move xmm5's 4th uint32 to its 2nd uint32, |
323 | // and move its 2nd uint32 to its 1st uint32 |
324 | |
325 | addps xmm4, xmm5; // results are in 1st and 2nd uint32 |
326 | |
327 | |
328 | movhlps xmm0, xmm6; // xmm6 high to xmm0 low |
329 | movlhps xmm0, xmm7; // xmm7 low to xmm0 high |
330 | |
331 | addps xmm6, xmm0; // result of add are in xmm6 low |
332 | addps xmm7, xmm0; // result of add are in xmm7 high |
333 | |
334 | shufps xmm0, xmm6, 0x44; // move xmm6 low DWORDs to xmm0 high |
335 | shufps xmm6, xmm7, 0xe4; // move xmm7 high DWORS to xmm6 |
336 | movhlps xmm7, xmm0; // xmm6, xmm7 are mirrored |
337 | |
338 | shufps xmm6, xmm6, 0x08; // move xmm6's 3rd uint32 to its 2nd uint32 |
339 | shufps xmm7, xmm7, 0x0d; // move xmm7's 4th uint32 to its 2nd uint32, |
340 | // and move its 2nd uint32 to its 1st uint32 |
341 | |
342 | addps xmm6, xmm7; // results are in 1st and 2nd uint32 |
343 | |
344 | movlhps xmm4, xmm6; // final result is in xmm4 |
345 | movaps DWORD PTR g_vtxTransformed [ecx], xmm4; |
346 | |
347 | movaps xmm0,xmm4; |
348 | shufps xmm0,xmm0,0xff; |
349 | divps xmm4,xmm0; |
350 | rcpps xmm0,xmm0; |
351 | movhlps xmm0,xmm4; |
352 | shufps xmm0,xmm0,0xe8; |
353 | movlhps xmm4,xmm0; |
354 | |
355 | movaps DWORD PTR g_vecProjected [ecx], xmm4; |
356 | |
357 | emms; |
358 | ret; |
359 | } |
360 | } |
361 | |
362 | // Only used by DKR |
363 | __declspec( naked ) void __fastcall SSEVec3TransformDKR(XVECTOR4 &pOut, const XVECTOR4 &pV) |
364 | { |
365 | __asm |
366 | { |
367 | movaps xmm1, DWORD PTR [edx]; // xmm1 as original vector |
368 | |
369 | movaps xmm4, DWORD PTR dkrMatrixTransposed; // row1 |
370 | movaps xmm5, DWORD PTR dkrMatrixTransposed[0x10]; // row2 |
371 | movaps xmm6, DWORD PTR dkrMatrixTransposed[0x20]; // row3 |
372 | movaps xmm7, DWORD PTR dkrMatrixTransposed[0x30]; // row4 |
373 | |
374 | mulps xmm4, xmm1; // row 1 |
375 | mulps xmm5, xmm1; // row 2 |
376 | mulps xmm6, xmm1; // row 3 |
377 | mulps xmm7, xmm1; // row 4 |
378 | |
379 | movhlps xmm0, xmm4; // xmm4 high to xmm0 low |
380 | movlhps xmm0, xmm5; // xmm5 low to xmm0 high |
381 | |
382 | addps xmm4, xmm0; // result of add are in xmm4 low |
383 | addps xmm5, xmm0; // result of add are in xmm5 high |
384 | |
385 | shufps xmm0, xmm4, 0x44; // move xmm4 low DWORDs to xmm0 high |
386 | shufps xmm4, xmm5, 0xe4; // move xmm5 high DWORS to xmm4 |
387 | movhlps xmm5, xmm0; // xmm4, xmm5 are mirrored |
388 | |
389 | shufps xmm4, xmm4, 0x08; // move xmm4's 3rd uint32 to its 2nd uint32 |
390 | shufps xmm5, xmm5, 0x0d; // move xmm5's 4th uint32 to its 2nd uint32, |
391 | // and move its 2nd uint32 to its 1st uint32 |
392 | |
393 | addps xmm4, xmm5; // results are in 1st and 2nd uint32 |
394 | |
395 | |
396 | movhlps xmm0, xmm6; // xmm6 high to xmm0 low |
397 | movlhps xmm0, xmm7; // xmm7 low to xmm0 high |
398 | |
399 | addps xmm6, xmm0; // result of add are in xmm6 low |
400 | addps xmm7, xmm0; // result of add are in xmm7 high |
401 | |
402 | shufps xmm0, xmm6, 0x44; // move xmm6 low DWORDs to xmm0 high |
403 | shufps xmm6, xmm7, 0xe4; // move xmm7 high DWORS to xmm6 |
404 | movhlps xmm7, xmm0; // xmm6, xmm7 are mirrored |
405 | |
406 | shufps xmm6, xmm6, 0x08; // move xmm6's 3rd uint32 to its 2nd uint32 |
407 | shufps xmm7, xmm7, 0x0d; // move xmm7's 4th uint32 to its 2nd uint32, |
408 | // and move its 2nd uint32 to its 1st uint32 |
409 | |
410 | addps xmm6, xmm7; // results are in 1st and 2nd uint32 |
411 | |
412 | movlhps xmm4, xmm6; // final result is in xmm4 |
413 | movaps DWORD PTR [ecx], xmm4; |
414 | |
415 | emms; |
416 | ret; |
417 | } |
418 | } |
419 | #elif defined(__GNUC__) && defined(__x86_64__) && !defined(NO_ASM) |
420 | void SSEVec3Transform(int i) |
421 | { |
422 | asm volatile(" shl $4, %0 \n" |
423 | " movslq %k0, %q0 \n" |
424 | " movaps (%1,%q0), %%xmm1 \n" |
425 | " movaps 0(%2), %%xmm4 \n" |
426 | " movaps 16(%2), %%xmm5 \n" |
427 | " movaps 32(%2), %%xmm6 \n" |
428 | " movaps 48(%2), %%xmm7 \n" |
429 | " mulps %%xmm1, %%xmm4 \n" |
430 | " mulps %%xmm1, %%xmm5 \n" |
431 | " mulps %%xmm1, %%xmm6 \n" |
432 | " mulps %%xmm1, %%xmm7 \n" |
433 | " movhlps %%xmm4, %%xmm0 \n" |
434 | " movlhps %%xmm5, %%xmm0 \n" |
435 | " addps %%xmm0, %%xmm4 \n" |
436 | " addps %%xmm0, %%xmm5 \n" |
437 | " shufps $0x44, %%xmm4, %%xmm0 \n" |
438 | " shufps $0xe4, %%xmm5, %%xmm4 \n" |
439 | " movhlps %%xmm0, %%xmm5 \n" |
440 | " shufps $0x08, %%xmm4, %%xmm4 \n" |
441 | " shufps $0x0d, %%xmm5, %%xmm5 \n" |
442 | " addps %%xmm5, %%xmm4 \n" |
443 | " movhlps %%xmm6, %%xmm0 \n" |
444 | " movlhps %%xmm7, %%xmm0 \n" |
445 | " addps %%xmm0, %%xmm6 \n" |
446 | " addps %%xmm0, %%xmm7 \n" |
447 | " shufps $0x44, %%xmm6, %%xmm0 \n" |
448 | " shufps $0xe4, %%xmm7, %%xmm6 \n" |
449 | " movhlps %%xmm0, %%xmm7 \n" |
450 | " shufps $0x08, %%xmm6, %%xmm6 \n" |
451 | " shufps $0x0d, %%xmm7, %%xmm7 \n" |
452 | " addps %%xmm7, %%xmm6 \n" |
453 | " movlhps %%xmm6, %%xmm4 \n" |
454 | " movaps %%xmm4, (%3,%q0) \n" |
455 | " movaps %%xmm4, %%xmm0 \n" |
456 | " shufps $0xff, %%xmm0, %%xmm0 \n" |
457 | " divps %%xmm0, %%xmm4 \n" |
458 | " rcpps %%xmm0, %%xmm0 \n" |
459 | " movhlps %%xmm4, %%xmm0 \n" |
460 | " shufps $0xe8, %%xmm0, %%xmm0 \n" |
461 | " movlhps %%xmm0, %%xmm4 \n" |
462 | " movaps %%xmm4, (%4,%q0) \n" |
463 | : "+r"(i) |
464 | : "r"(g_vtxNonTransformed), "r"(&gRSPworldProjectTransported.m[0][0]), "r"(g_vtxTransformed), "r"(g_vecProjected) |
465 | : "memory", "%xmm0", "%xmm1", "%xmm4", "%xmm5", "%xmm6", "%xmm7" |
466 | ); |
467 | } |
468 | #elif !defined(NO_ASM) // 32-bit GCC assumed |
469 | void SSEVec3Transform(int i) |
470 | { |
471 | asm volatile(" shl $4, %0 \n" |
472 | " movaps (%1,%0), %%xmm1 \n" |
473 | " movaps 0(%2), %%xmm4 \n" |
474 | " movaps 16(%2), %%xmm5 \n" |
475 | " movaps 32(%2), %%xmm6 \n" |
476 | " movaps 48(%2), %%xmm7 \n" |
477 | " mulps %%xmm1, %%xmm4 \n" |
478 | " mulps %%xmm1, %%xmm5 \n" |
479 | " mulps %%xmm1, %%xmm6 \n" |
480 | " mulps %%xmm1, %%xmm7 \n" |
481 | " movhlps %%xmm4, %%xmm0 \n" |
482 | " movlhps %%xmm5, %%xmm0 \n" |
483 | " addps %%xmm0, %%xmm4 \n" |
484 | " addps %%xmm0, %%xmm5 \n" |
485 | " shufps $0x44, %%xmm4, %%xmm0 \n" |
486 | " shufps $0xe4, %%xmm5, %%xmm4 \n" |
487 | " movhlps %%xmm0, %%xmm5 \n" |
488 | " shufps $0x08, %%xmm4, %%xmm4 \n" |
489 | " shufps $0x0d, %%xmm5, %%xmm5 \n" |
490 | " addps %%xmm5, %%xmm4 \n" |
491 | " movhlps %%xmm6, %%xmm0 \n" |
492 | " movlhps %%xmm7, %%xmm0 \n" |
493 | " addps %%xmm0, %%xmm6 \n" |
494 | " addps %%xmm0, %%xmm7 \n" |
495 | " shufps $0x44, %%xmm6, %%xmm0 \n" |
496 | " shufps $0xe4, %%xmm7, %%xmm6 \n" |
497 | " movhlps %%xmm0, %%xmm7 \n" |
498 | " shufps $0x08, %%xmm6, %%xmm6 \n" |
499 | " shufps $0x0d, %%xmm7, %%xmm7 \n" |
500 | " addps %%xmm7, %%xmm6 \n" |
501 | " movlhps %%xmm6, %%xmm4 \n" |
502 | " movaps %%xmm4, (%3,%0) \n" |
503 | " movaps %%xmm4, %%xmm0 \n" |
504 | " shufps $0xff, %%xmm0, %%xmm0 \n" |
505 | " divps %%xmm0, %%xmm4 \n" |
506 | " rcpps %%xmm0, %%xmm0 \n" |
507 | " movhlps %%xmm4, %%xmm0 \n" |
508 | " shufps $0xe8, %%xmm0, %%xmm0 \n" |
509 | " movlhps %%xmm0, %%xmm4 \n" |
510 | " movaps %%xmm4, (%4,%0) \n" |
511 | : "+r"(i) |
512 | : "r"(g_vtxNonTransformed), "r"(&gRSPworldProjectTransported.m[0][0]), "r"(g_vtxTransformed), "r"(g_vecProjected) |
513 | : "memory", "%xmm0", "%xmm1", "%xmm4", "%xmm5", "%xmm6", "%xmm7" |
514 | ); |
515 | } |
516 | #endif |
517 | float real255 = 255.0f; |
518 | float real128 = 128.0f; |
519 | |
520 | #if !defined(__GNUC__) && !defined(NO_ASM) |
521 | __declspec( naked ) void __fastcall SSEVec3TransformNormal() |
522 | { |
523 | __asm |
524 | { |
525 | mov DWORD PTR [g_normal][12], 0; |
526 | |
527 | movaps xmm4, DWORD PTR gRSPmodelViewTopTranspose; // row1 |
528 | movaps xmm5, DWORD PTR gRSPmodelViewTopTranspose[0x10]; // row2 |
529 | movaps xmm1, DWORD PTR [g_normal]; // xmm1 as the normal vector |
530 | movaps xmm6, DWORD PTR gRSPmodelViewTopTranspose[0x20]; // row3 |
531 | |
532 | mulps xmm4, xmm1; // row 1 |
533 | mulps xmm5, xmm1; // row 2 |
534 | mulps xmm6, xmm1; // row 3 |
535 | |
536 | movhlps xmm0, xmm4; // xmm4 high to xmm0 low |
537 | movlhps xmm0, xmm5; // xmm5 low to xmm0 high |
538 | |
539 | addps xmm4, xmm0; // result of add are in xmm4 low |
540 | addps xmm5, xmm0; // result of add are in xmm5 high |
541 | |
542 | shufps xmm0, xmm4, 0x44; // move xmm4 low DWORDs to xmm0 high |
543 | shufps xmm4, xmm5, 0xe4; // move xmm5 high DWORS to xmm4 |
544 | movhlps xmm5, xmm0; // xmm4, xmm5 are mirrored |
545 | |
546 | shufps xmm4, xmm4, 0x08; // move xmm4's 3rd uint32 to its 2nd uint32 |
547 | shufps xmm5, xmm5, 0x0d; // move xmm5's 4th uint32 to its 2nd uint32, |
548 | |
549 | addps xmm4, xmm5; // results are in 1st and 2nd uint32 |
550 | |
551 | movaps xmm1,xmm4; |
552 | mulps xmm1,xmm1; //square |
553 | movlhps xmm7, xmm1; |
554 | shufps xmm7, xmm7,0x03; |
555 | addss xmm7, xmm1; |
556 | |
557 | movhlps xmm0, xmm6; // xmm6 high to xmm0 low |
558 | addps xmm6, xmm0; // result of add are in xmm6 low |
559 | |
560 | movlhps xmm0, xmm6; |
561 | shufps xmm0, xmm0, 0x03; |
562 | addss xmm0, xmm6; // result of add is at xmm0's 1st uint32 |
563 | |
564 | movlhps xmm4, xmm0; |
565 | |
566 | mulss xmm0,xmm0; |
567 | addss xmm7,xmm0; // xmm7 1st uint32 is the sum of squares |
568 | |
569 | #ifdef DEBUGGER |
570 | movaps DWORD PTR [g_normal], xmm4; |
571 | movss DWORD PTR [g_normal][12], xmm7; |
572 | #endif |
573 | xorps xmm0,xmm0; |
574 | ucomiss xmm0,xmm7; |
575 | jz l2 |
576 | |
577 | rsqrtss xmm7,xmm7; |
578 | shufps xmm7,xmm7,0; |
579 | #ifdef DEBUGGER |
580 | movss DWORD PTR [g_normal][12], xmm7; |
581 | #endif |
582 | mulps xmm4,xmm7; |
583 | |
584 | movaps DWORD PTR [g_normal], xmm4; // Normalized |
585 | mov DWORD PTR [g_normal][12], 0; |
586 | |
587 | emms; |
588 | ret; |
589 | l2: |
590 | movss DWORD PTR [g_normal], xmm0; |
591 | movss DWORD PTR [g_normal][12], xmm0; |
592 | emms; |
593 | ret; |
594 | } |
595 | } |
596 | #elif defined(__GNUC__) && !defined(NO_ASM) // this code should compile for both 64-bit and 32-bit architectures |
597 | void SSEVec3TransformNormal(void) |
598 | { |
599 | asm volatile(" movl $0, 12(%0) \n" |
600 | " movaps (%1), %%xmm4 \n" |
601 | " movaps 16(%1), %%xmm5 \n" |
602 | " movaps (%0), %%xmm1 \n" |
603 | " movaps 32(%1), %%xmm6 \n" |
604 | " mulps %%xmm1, %%xmm4 \n" |
605 | " mulps %%xmm1, %%xmm5 \n" |
606 | " mulps %%xmm1, %%xmm6 \n" |
607 | " movhlps %%xmm4, %%xmm0 \n" |
608 | " movlhps %%xmm5, %%xmm0 \n" |
609 | " addps %%xmm0, %%xmm4 \n" |
610 | " addps %%xmm0, %%xmm5 \n" |
611 | " shufps $0x44, %%xmm4, %%xmm0 \n" |
612 | " shufps $0xe4, %%xmm5, %%xmm4 \n" |
613 | " movhlps %%xmm0, %%xmm5 \n" |
614 | " shufps $0x08, %%xmm4, %%xmm4 \n" |
615 | " shufps $0x0d, %%xmm5, %%xmm5 \n" |
616 | " addps %%xmm5, %%xmm4 \n" |
617 | " movaps %%xmm4, %%xmm1 \n" |
618 | " mulps %%xmm1, %%xmm1 \n" |
619 | " movlhps %%xmm1, %%xmm7 \n" |
620 | " shufps $0x03, %%xmm7, %%xmm7 \n" |
621 | " addss %%xmm1, %%xmm7 \n" |
622 | " movhlps %%xmm6, %%xmm0 \n" |
623 | " addps %%xmm0, %%xmm6 \n" |
624 | " movlhps %%xmm6, %%xmm0 \n" |
625 | " shufps $0x03, %%xmm0, %%xmm0 \n" |
626 | " addss %%xmm6, %%xmm0 \n" |
627 | " movlhps %%xmm0, %%xmm4 \n" |
628 | " mulss %%xmm0, %%xmm0 \n" |
629 | " addss %%xmm0, %%xmm7 \n" |
630 | #ifdef DEBUGGER |
631 | " movaps %%xmm4, (%0) \n" |
632 | " movss %%xmm7, 12(%0) \n" |
633 | #endif |
634 | " xorps %%xmm0, %%xmm0 \n" |
635 | " ucomiss %%xmm7, %%xmm0 \n" |
636 | " jz 0f \n" |
637 | " rsqrtss %%xmm7, %%xmm7 \n" |
638 | " shufps $0x00, %%xmm7, %%xmm7 \n" |
639 | #ifdef DEBUGGER |
640 | " movss %%xmm7, 12(%0) \n" |
641 | #endif |
642 | " mulps %%xmm7, %%xmm4 \n" |
643 | " movaps %%xmm4, (%0) \n" |
644 | " movl $0, 12(%0) \n" |
645 | " jmp 1f \n" |
646 | "0: \n" |
647 | " movss %%xmm0, (%0) \n" |
648 | " movss %%xmm0, 12(%0) \n" |
649 | "1: \n" |
650 | : |
651 | : "r"(&g_normal.x), "r"(&gRSPmodelViewTopTranspose.m[0][0]) |
652 | : "memory", "cc", "%xmm0", "%xmm1", "%xmm4", "%xmm5", "%xmm6", "%xmm7" |
653 | ); |
654 | } |
655 | #endif |
656 | |
657 | void NormalizeNormalVec() |
658 | #ifdef __ARM_NEON__0 |
659 | { |
660 | asm volatile ( |
661 | "vld1.32 {d4}, [%0] \n\t" //d4={x,y} |
662 | "flds s10, [%0, #8] \n\t" //d5[0] = z |
663 | // "sub %0, %0, #8 \n\t" //d5[0] = z |
664 | "vmul.f32 d0, d4, d4 \n\t" //d0= d4*d4 |
665 | "vpadd.f32 d0, d0, d0 \n\t" //d0 = d[0] + d[1] |
666 | "vmla.f32 d0, d5, d5 \n\t" //d0 = d0 + d5*d5 |
667 | |
668 | "vmov.f32 d1, d0 \n\t" //d1 = d0 |
669 | "vrsqrte.f32 d0, d0 \n\t" //d0 = ~ 1.0 / sqrt(d0) |
670 | "vmul.f32 d2, d0, d1 \n\t" //d2 = d0 * d1 |
671 | "vrsqrts.f32 d3, d2, d0 \n\t" //d3 = (3 - d0 * d2) / 2 |
672 | "vmul.f32 d0, d0, d3 \n\t" //d0 = d0 * d3 |
673 | "vmul.f32 d2, d0, d1 \n\t" //d2 = d0 * d1 |
674 | "vrsqrts.f32 d3, d2, d0 \n\t" //d3 = (3 - d0 * d3) / 2 |
675 | "vmul.f32 d0, d0, d3 \n\t" //d0 = d0 * d4 |
676 | |
677 | "vmul.f32 q2, q2, d0[0] \n\t" //d0= d2*d4 |
678 | "vst1.32 {d4}, [%0] \n\t" //d2={x0,y0}, d3={z0, w0} |
679 | "fsts s10, [%0, #8] \n\t" //d2={x0,y0}, d3={z0, w0} |
680 | |
681 | ::"r"(g_normal) |
682 | : "d0", "d1", "d2", "d3", "d4", "d5", "memory" |
683 | ); |
684 | } |
685 | #else |
686 | { |
687 | float w = 1/sqrtf(g_normal.x*g_normal.x + g_normal.y*g_normal.y + g_normal.z*g_normal.z); |
688 | g_normal.x *= w; |
689 | g_normal.y *= w; |
690 | g_normal.z *= w; |
691 | } |
692 | #endif |
693 | |
694 | void InitRenderBase() |
695 | { |
696 | #if !defined(NO_ASM) |
697 | if( status.isSSEEnabled && !g_curRomInfo.bPrimaryDepthHack && options.enableHackForGames != HACK_FOR_NASCAR) |
698 | { |
699 | ProcessVertexData = ProcessVertexDataSSE; |
700 | } |
701 | else |
f18a2655 |
702 | #elif defined(__ARM_NEON__) |
703 | if( !g_curRomInfo.bPrimaryDepthHack && options.enableHackForGames != HACK_FOR_NASCAR && options.enableHackForGames != HACK_FOR_ZELDA_MM && !options.bWinFrameMode) |
704 | { |
705 | ProcessVertexData = ProcessVertexDataNEON; |
706 | } |
707 | else |
d07c171f |
708 | #endif |
709 | { |
710 | ProcessVertexData = ProcessVertexDataNoSSE; |
711 | } |
712 | |
713 | gRSPfFogMin = gRSPfFogMax = 0.0f; |
714 | windowSetting.fMultX = windowSetting.fMultY = 2.0f; |
715 | windowSetting.vpLeftW = windowSetting.vpTopW = 0; |
716 | windowSetting.vpRightW = windowSetting.vpWidthW = 640; |
717 | windowSetting.vpBottomW = windowSetting.vpHeightW = 480; |
718 | gRSP.maxZ = 0; |
719 | gRSP.nVPLeftN = gRSP.nVPTopN = 0; |
720 | gRSP.nVPRightN = 640; |
721 | gRSP.nVPBottomN = 640; |
722 | gRSP.nVPWidthN = 640; |
723 | gRSP.nVPHeightN = 640; |
724 | gRDP.scissor.left=gRDP.scissor.top=0; |
725 | gRDP.scissor.right=gRDP.scissor.bottom=640; |
726 | |
727 | gRSP.bLightingEnable = gRSP.bTextureGen = false; |
728 | gRSP.curTile=gRSPnumLights=gRSP.ambientLightColor=gRSP.ambientLightIndex= 0; |
729 | gRSP.fAmbientLightR=gRSP.fAmbientLightG=gRSP.fAmbientLightB=0; |
730 | gRSP.projectionMtxTop = gRSP.modelViewMtxTop = 0; |
731 | gRDP.fogColor = gRDP.primitiveColor = gRDP.envColor = gRDP.primitiveDepth = gRDP.primLODMin = gRDP.primLODFrac = gRDP.LODFrac = 0; |
732 | gRDP.fPrimitiveDepth = 0; |
733 | gRSP.numVertices = 0; |
734 | gRSP.maxVertexID = 0; |
735 | gRSP.bCullFront=false; |
736 | gRSP.bCullBack=true; |
737 | gRSP.bFogEnabled=gRDP.bFogEnableInBlender=false; |
738 | gRSP.bZBufferEnabled=true; |
739 | gRSP.shadeMode=SHADE_SMOOTH; |
740 | gRDP.keyR=gRDP.keyG=gRDP.keyB=gRDP.keyA=gRDP.keyRGB=gRDP.keyRGBA = 0; |
741 | gRDP.fKeyA = 0; |
742 | gRSP.DKRCMatrixIndex = gRSP.dwDKRVtxAddr = gRSP.dwDKRMatrixAddr = 0; |
743 | gRSP.DKRBillBoard = false; |
744 | |
745 | gRSP.fTexScaleX = 1/32.0f; |
746 | gRSP.fTexScaleY = 1/32.0f; |
747 | gRSP.bTextureEnabled = FALSE; |
748 | |
749 | gRSP.clip_ratio_left = 0; |
750 | gRSP.clip_ratio_top = 0; |
751 | gRSP.clip_ratio_right = 640; |
752 | gRSP.clip_ratio_bottom = 480; |
753 | gRSP.clip_ratio_negx = 1; |
754 | gRSP.clip_ratio_negy = 1; |
755 | gRSP.clip_ratio_posx = 1; |
756 | gRSP.clip_ratio_posy = 1; |
757 | gRSP.real_clip_scissor_left = 0; |
758 | gRSP.real_clip_scissor_top = 0; |
759 | gRSP.real_clip_scissor_right = 640; |
760 | gRSP.real_clip_scissor_bottom = 480; |
761 | windowSetting.clipping.left = 0; |
762 | windowSetting.clipping.top = 0; |
763 | windowSetting.clipping.right = 640; |
764 | windowSetting.clipping.bottom = 480; |
765 | windowSetting.clipping.width = 640; |
766 | windowSetting.clipping.height = 480; |
767 | windowSetting.clipping.needToClip = false; |
768 | gRSP.real_clip_ratio_negx = 1; |
769 | gRSP.real_clip_ratio_negy = 1; |
770 | gRSP.real_clip_ratio_posx = 1; |
771 | gRSP.real_clip_ratio_posy = 1; |
772 | |
773 | gRSP.DKRCMatrixIndex=0; |
774 | gRSP.DKRVtxCount=0; |
775 | gRSP.DKRBillBoard = false; |
776 | gRSP.dwDKRVtxAddr=0; |
777 | gRSP.dwDKRMatrixAddr=0; |
778 | |
779 | |
780 | gRDP.geometryMode = 0; |
781 | gRDP.otherModeL = 0; |
782 | gRDP.otherModeH = 0; |
783 | gRDP.fillColor = 0xFFFFFFFF; |
784 | gRDP.originalFillColor =0; |
785 | |
786 | gRSP.ucode = 1; |
787 | gRSP.vertexMult = 10; |
788 | gRSP.bNearClip = false; |
789 | gRSP.bRejectVtx = false; |
790 | |
791 | gRDP.texturesAreReloaded = false; |
792 | gRDP.textureIsChanged = false; |
793 | gRDP.colorsAreReloaded = false; |
794 | |
795 | memset(&gRDP.otherMode,0,sizeof(RDP_OtherMode)); |
796 | memset(&gRDP.tiles,0,sizeof(Tile)*8); |
797 | |
798 | for( int i=0; i<MAX_VERTS; i++ ) |
799 | { |
800 | g_clipFlag[i] = 0; |
801 | g_vtxNonTransformed[i].w = 1; |
802 | } |
803 | |
804 | memset(gRSPn64lights, 0, sizeof(N64Light)*16); |
805 | } |
806 | |
807 | void SetFogMinMax(float fMin, float fMax, float fMul, float fOffset) |
808 | { |
809 | if( fMin > fMax ) |
810 | { |
811 | float temp = fMin; |
812 | fMin = fMax; |
813 | fMax = temp; |
814 | } |
815 | |
816 | { |
817 | gRSPfFogMin = max(0,fMin/500-1); |
818 | gRSPfFogMax = fMax/500-1; |
819 | } |
820 | |
821 | gRSPfFogDivider = 255/(gRSPfFogMax-gRSPfFogMin); |
822 | CRender::g_pRender->SetFogMinMax(fMin, fMax); |
823 | } |
824 | |
825 | void InitVertexColors() |
826 | { |
827 | } |
828 | |
829 | void InitVertexTextureConstants() |
830 | { |
831 | float scaleX; |
832 | float scaleY; |
833 | |
834 | RenderTexture &tex0 = g_textures[gRSP.curTile]; |
835 | //CTexture *surf = tex0.m_pCTexture; |
836 | Tile &tile0 = gRDP.tiles[gRSP.curTile]; |
837 | |
838 | scaleX = gRSP.fTexScaleX; |
839 | scaleY = gRSP.fTexScaleY; |
840 | |
841 | gRSP.tex0scaleX = scaleX * tile0.fShiftScaleS/tex0.m_fTexWidth; |
842 | gRSP.tex0scaleY = scaleY * tile0.fShiftScaleT/tex0.m_fTexHeight; |
843 | |
844 | gRSP.tex0OffsetX = tile0.fhilite_sl/tex0.m_fTexWidth; |
845 | gRSP.tex0OffsetY = tile0.fhilite_tl/tex0.m_fTexHeight; |
846 | |
847 | if( CRender::g_pRender->IsTexel1Enable() ) |
848 | { |
849 | RenderTexture &tex1 = g_textures[(gRSP.curTile+1)&7]; |
850 | //CTexture *surf = tex1.m_pCTexture; |
851 | Tile &tile1 = gRDP.tiles[(gRSP.curTile+1)&7]; |
852 | |
853 | gRSP.tex1scaleX = scaleX * tile1.fShiftScaleS/tex1.m_fTexWidth; |
854 | gRSP.tex1scaleY = scaleY * tile1.fShiftScaleT/tex1.m_fTexHeight; |
855 | |
856 | gRSP.tex1OffsetX = tile1.fhilite_sl/tex1.m_fTexWidth; |
857 | gRSP.tex1OffsetY = tile1.fhilite_tl/tex1.m_fTexHeight; |
858 | } |
859 | |
860 | gRSP.texGenXRatio = tile0.fShiftScaleS; |
861 | gRSP.texGenYRatio = gRSP.fTexScaleX/gRSP.fTexScaleY*tex0.m_fTexWidth/tex0.m_fTexHeight*tile0.fShiftScaleT; |
862 | } |
863 | |
864 | void TexGen(float &s, float &t) |
865 | { |
866 | if (gRDP.geometryMode & G_TEXTURE_GEN_LINEAR) |
867 | { |
868 | s = acosf(g_normal.x) / 3.14159f; |
869 | t = acosf(g_normal.y) / 3.14159f; |
870 | } |
871 | else |
872 | { |
873 | s = 0.5f * ( 1.0f + g_normal.x); |
874 | t = 0.5f * ( 1.0f - g_normal.y); |
875 | } |
876 | } |
877 | |
878 | void ComputeLOD(bool openGL) |
879 | { |
880 | TLITVERTEX &v0 = g_vtxBuffer[0]; |
881 | TLITVERTEX &v1 = g_vtxBuffer[1]; |
882 | RenderTexture &tex0 = g_textures[gRSP.curTile]; |
883 | |
884 | float d,dt; |
885 | if( openGL ) |
886 | { |
887 | float x = g_vtxProjected5[0][0] / g_vtxProjected5[0][4] - g_vtxProjected5[1][0] / g_vtxProjected5[1][4]; |
888 | float y = g_vtxProjected5[0][1] / g_vtxProjected5[0][4] - g_vtxProjected5[1][1] / g_vtxProjected5[1][4]; |
889 | |
890 | x = windowSetting.vpWidthW*x/windowSetting.fMultX/2; |
891 | y = windowSetting.vpHeightW*y/windowSetting.fMultY/2; |
892 | d = sqrtf(x*x+y*y); |
893 | } |
894 | else |
895 | { |
896 | float x = (v0.x - v1.x)/ windowSetting.fMultX; |
897 | float y = (v0.y - v1.y)/ windowSetting.fMultY; |
898 | d = sqrtf(x*x+y*y); |
899 | } |
900 | |
901 | float s0 = v0.tcord[0].u * tex0.m_fTexWidth; |
902 | float t0 = v0.tcord[0].v * tex0.m_fTexHeight; |
903 | float s1 = v1.tcord[0].u * tex0.m_fTexWidth; |
904 | float t1 = v1.tcord[0].v * tex0.m_fTexHeight; |
905 | |
906 | dt = sqrtf((s0-s1)*(s0-s1)+(t0-t1)*(t0-t1)); |
907 | |
908 | float lod = dt/d; |
909 | float frac = log10f(lod)/log10f(2.0f); |
910 | //DEBUGGER_IF_DUMP(pauseAtNext,{DebuggerAppendMsg("LOD frac = %f", frac);}); |
911 | frac = (lod / powf(2.0f,floorf(frac))); |
912 | frac = frac - floorf(frac); |
913 | //DEBUGGER_IF_DUMP(pauseAtNext,{DebuggerAppendMsg("LOD = %f, frac = %f", lod, frac);}); |
914 | gRDP.LODFrac = (uint32)(frac*255); |
915 | CRender::g_pRender->SetCombinerAndBlender(); |
916 | } |
917 | |
918 | bool bHalfTxtScale=false; |
919 | extern uint32 lastSetTile; |
48d77f73 |
920 | #define noinline __attribute__((noinline)) |
921 | |
922 | static noinline void InitVertex_scale_hack_check(uint32 dwV) |
923 | { |
924 | // Check for txt scale hack |
925 | if( gRDP.tiles[lastSetTile].dwSize == TXT_SIZE_32b || gRDP.tiles[lastSetTile].dwSize == TXT_SIZE_4b ) |
926 | { |
927 | int width = ((gRDP.tiles[lastSetTile].sh-gRDP.tiles[lastSetTile].sl+1)<<1); |
928 | int height = ((gRDP.tiles[lastSetTile].th-gRDP.tiles[lastSetTile].tl+1)<<1); |
929 | if( g_fVtxTxtCoords[dwV].x*gRSP.fTexScaleX == width || g_fVtxTxtCoords[dwV].y*gRSP.fTexScaleY == height ) |
930 | { |
931 | bHalfTxtScale=true; |
932 | } |
933 | } |
934 | } |
935 | |
936 | static noinline void InitVertex_notopengl_or_clipper_adjust(TLITVERTEX &v, uint32 dwV) |
937 | { |
938 | v.x = g_vecProjected[dwV].x*gRSP.vtxXMul+gRSP.vtxXAdd; |
939 | v.y = g_vecProjected[dwV].y*gRSP.vtxYMul+gRSP.vtxYAdd; |
940 | v.z = (g_vecProjected[dwV].z + 1.0f) * 0.5f; // DirectX minZ=0, maxZ=1 |
941 | //v.z = g_vecProjected[dwV].z; // DirectX minZ=0, maxZ=1 |
942 | v.rhw = g_vecProjected[dwV].w; |
943 | VTX_DUMP(TRACE4(" Proj : x=%f, y=%f, z=%f, rhw=%f", v.x,v.y,v.z,v.rhw)); |
944 | |
945 | if( gRSP.bProcessSpecularColor ) |
946 | { |
947 | v.dcSpecular = CRender::g_pRender->PostProcessSpecularColor(); |
948 | if( gRSP.bFogEnabled ) |
949 | { |
950 | v.dcSpecular &= 0x00FFFFFF; |
951 | uint32 fogFct = 0xFF-(uint8)((g_fFogCoord[dwV]-gRSPfFogMin)*gRSPfFogDivider); |
952 | v.dcSpecular |= (fogFct<<24); |
953 | } |
954 | } |
955 | else if( gRSP.bFogEnabled ) |
956 | { |
957 | uint32 fogFct = 0xFF-(uint8)((g_fFogCoord[dwV]-gRSPfFogMin)*gRSPfFogDivider); |
958 | v.dcSpecular = (fogFct<<24); |
959 | } |
960 | } |
961 | |
962 | static noinline void InitVertex_texgen_correct(TLITVERTEX &v, uint32 dwV) |
963 | { |
964 | // Correction for texGen result |
965 | float u0,u1,v0,v1; |
966 | RenderTexture &tex0 = g_textures[gRSP.curTile]; |
967 | u0 = g_fVtxTxtCoords[dwV].x * 32 * 1024 * gRSP.fTexScaleX / tex0.m_fTexWidth; |
968 | v0 = g_fVtxTxtCoords[dwV].y * 32 * 1024 * gRSP.fTexScaleY / tex0.m_fTexHeight; |
969 | u0 *= (gRDP.tiles[gRSP.curTile].fShiftScaleS); |
970 | v0 *= (gRDP.tiles[gRSP.curTile].fShiftScaleT); |
971 | |
972 | if( CRender::g_pRender->IsTexel1Enable() ) |
973 | { |
974 | RenderTexture &tex1 = g_textures[(gRSP.curTile+1)&7]; |
975 | u1 = g_fVtxTxtCoords[dwV].x * 32 * 1024 * gRSP.fTexScaleX / tex1.m_fTexWidth; |
976 | v1 = g_fVtxTxtCoords[dwV].y * 32 * 1024 * gRSP.fTexScaleY / tex1.m_fTexHeight; |
977 | u1 *= gRDP.tiles[(gRSP.curTile+1)&7].fShiftScaleS; |
978 | v1 *= gRDP.tiles[(gRSP.curTile+1)&7].fShiftScaleT; |
979 | CRender::g_pRender->SetVertexTextureUVCoord(v, u0, v0, u1, v1); |
980 | } |
981 | else |
982 | { |
983 | CRender::g_pRender->SetVertexTextureUVCoord(v, u0, v0); |
984 | } |
985 | } |
986 | |
987 | #ifndef __ARM_NEON__ |
988 | static void multiply_subtract2(float *d, const float *m1, const float *m2, const float *s) |
989 | { |
990 | int i; |
991 | for (i = 0; i < 2; i++) |
992 | d[i] = m1[i] * m2[i] - s[i]; |
993 | } |
994 | #else |
995 | extern "C" void multiply_subtract2(float *d, const float *m1, const float *m2, const float *s); |
996 | #endif |
d07c171f |
997 | |
998 | void InitVertex(uint32 dwV, uint32 vtxIndex, bool bTexture, bool openGL) |
999 | { |
1000 | VTX_DUMP(TRACE2("Init vertex (%d) to vtx buf[%d]:", dwV, vtxIndex)); |
1001 | |
48d77f73 |
1002 | #ifdef __linux__ |
1003 | openGL = 1; // what else there is? |
1004 | #endif |
d07c171f |
1005 | TLITVERTEX &v = g_vtxBuffer[vtxIndex]; |
1006 | VTX_DUMP(TRACE4(" Trans: x=%f, y=%f, z=%f, w=%f", g_vtxTransformed[dwV].x,g_vtxTransformed[dwV].y,g_vtxTransformed[dwV].z,g_vtxTransformed[dwV].w)); |
1007 | if( openGL ) |
1008 | { |
1009 | g_vtxProjected5[vtxIndex][0] = g_vtxTransformed[dwV].x; |
1010 | g_vtxProjected5[vtxIndex][1] = g_vtxTransformed[dwV].y; |
1011 | g_vtxProjected5[vtxIndex][2] = g_vtxTransformed[dwV].z; |
1012 | g_vtxProjected5[vtxIndex][3] = g_vtxTransformed[dwV].w; |
1013 | g_vtxProjected5[vtxIndex][4] = g_vecProjected[dwV].z; |
1014 | |
48d77f73 |
1015 | if( *(int *)&g_vtxTransformed[dwV].w < 0 ) |
d07c171f |
1016 | g_vtxProjected5[vtxIndex][4] = 0; |
1017 | |
1018 | g_vtxIndex[vtxIndex] = vtxIndex; |
1019 | } |
1020 | |
48d77f73 |
1021 | if( __builtin_expect(!openGL || options.bOGLVertexClipper == TRUE, 0) ) |
d07c171f |
1022 | { |
48d77f73 |
1023 | InitVertex_notopengl_or_clipper_adjust(v, dwV); |
d07c171f |
1024 | } |
1025 | VTX_DUMP(TRACE2(" (U,V): %f, %f", g_fVtxTxtCoords[dwV].x,g_fVtxTxtCoords[dwV].y)); |
1026 | |
1027 | v.dcDiffuse = g_dwVtxDifColor[dwV]; |
1028 | if( gRDP.otherMode.key_en ) |
1029 | { |
1030 | v.dcDiffuse &= 0x00FFFFFF; |
1031 | v.dcDiffuse |= (gRDP.keyA<<24); |
1032 | } |
1033 | else if( gRDP.otherMode.aa_en && gRDP.otherMode.clr_on_cvg==0 ) |
1034 | { |
1035 | v.dcDiffuse |= 0xFF000000; |
1036 | } |
1037 | |
1038 | if( gRSP.bProcessDiffuseColor ) |
1039 | { |
1040 | v.dcDiffuse = CRender::g_pRender->PostProcessDiffuseColor(v.dcDiffuse); |
1041 | } |
1042 | if( options.bWinFrameMode ) |
1043 | { |
1044 | v.dcDiffuse = g_dwVtxDifColor[dwV]; |
1045 | } |
1046 | |
1047 | if( openGL ) |
1048 | { |
1049 | g_oglVtxColors[vtxIndex][0] = v.r; |
1050 | g_oglVtxColors[vtxIndex][1] = v.g; |
1051 | g_oglVtxColors[vtxIndex][2] = v.b; |
1052 | g_oglVtxColors[vtxIndex][3] = v.a; |
1053 | } |
1054 | |
1055 | if( bTexture ) |
1056 | { |
1057 | // If the vert is already lit, then there is no normal (and hence we can't generate tex coord) |
1058 | // Only scale if not generated automatically |
48d77f73 |
1059 | if ( __builtin_expect(gRSP.bTextureGen && gRSP.bLightingEnable, 0) ) |
d07c171f |
1060 | { |
48d77f73 |
1061 | InitVertex_texgen_correct(v, dwV); |
d07c171f |
1062 | } |
1063 | else |
1064 | { |
48d77f73 |
1065 | TexCord tex0; |
1066 | multiply_subtract2(&tex0.u, &g_fVtxTxtCoords[dwV].x, &gRSP.tex0scaleX, &gRSP.tex0OffsetX); |
d07c171f |
1067 | |
1068 | if( CRender::g_pRender->IsTexel1Enable() ) |
1069 | { |
48d77f73 |
1070 | TexCord tex1; |
1071 | multiply_subtract2(&tex1.u, &g_fVtxTxtCoords[dwV].x, &gRSP.tex1scaleX, &gRSP.tex1OffsetX); |
d07c171f |
1072 | |
48d77f73 |
1073 | CRender::g_pRender->SetVertexTextureUVCoord(v, tex0, tex1); |
1074 | VTX_DUMP(TRACE2(" (tex0): %f, %f", tex0.u,tex0.v)); |
1075 | VTX_DUMP(TRACE2(" (tex1): %f, %f", tex1.u,tex1.v)); |
d07c171f |
1076 | } |
1077 | else |
1078 | { |
48d77f73 |
1079 | CRender::g_pRender->SetVertexTextureUVCoord(v, tex0); |
1080 | VTX_DUMP(TRACE2(" (tex0): %f, %f", tex0.u,tex0.v)); |
d07c171f |
1081 | } |
1082 | } |
1083 | |
48d77f73 |
1084 | if( __builtin_expect(g_curRomInfo.bTextureScaleHack && !bHalfTxtScale, 0) ) |
1085 | InitVertex_scale_hack_check(dwV); |
d07c171f |
1086 | } |
1087 | |
1088 | VTX_DUMP(TRACE2(" DIF(%08X), SPE(%08X)", v.dcDiffuse, v.dcSpecular)); |
1089 | VTX_DUMP(TRACE0("")); |
1090 | } |
1091 | |
1092 | uint32 LightVert(XVECTOR4 & norm, int vidx) |
1093 | { |
1094 | float fCosT; |
1095 | |
1096 | // Do ambient |
1097 | register float r = gRSP.fAmbientLightR; |
1098 | register float g = gRSP.fAmbientLightG; |
1099 | register float b = gRSP.fAmbientLightB; |
1100 | |
1101 | if( options.enableHackForGames != HACK_FOR_ZELDA_MM ) |
1102 | { |
1103 | for (register unsigned int l=0; l < gRSPnumLights; l++) |
1104 | { |
1105 | fCosT = norm.x*gRSPlights[l].x + norm.y*gRSPlights[l].y + norm.z*gRSPlights[l].z; |
1106 | |
1107 | if (fCosT > 0 ) |
1108 | { |
1109 | r += gRSPlights[l].fr * fCosT; |
1110 | g += gRSPlights[l].fg * fCosT; |
1111 | b += gRSPlights[l].fb * fCosT; |
1112 | } |
1113 | } |
1114 | } |
1115 | else |
1116 | { |
1117 | XVECTOR4 v; |
1118 | bool transformed = false; |
1119 | |
1120 | for (register unsigned int l=0; l < gRSPnumLights; l++) |
1121 | { |
1122 | if( gRSPlights[l].range == 0 ) |
1123 | { |
1124 | // Regular directional light |
1125 | fCosT = norm.x*gRSPlights[l].x + norm.y*gRSPlights[l].y + norm.z*gRSPlights[l].z; |
1126 | |
1127 | if (fCosT > 0 ) |
1128 | { |
1129 | r += gRSPlights[l].fr * fCosT; |
1130 | g += gRSPlights[l].fg * fCosT; |
1131 | b += gRSPlights[l].fb * fCosT; |
1132 | } |
1133 | } |
1134 | else //if( (gRSPlights[l].col&0x00FFFFFF) != 0x00FFFFFF ) |
1135 | { |
1136 | // Point light |
1137 | if( !transformed ) |
1138 | { |
1139 | Vec3Transform(&v, (XVECTOR3*)&g_vtxNonTransformed[vidx], &gRSPmodelViewTop); // Convert to w=1 |
1140 | transformed = true; |
1141 | } |
1142 | |
1143 | XVECTOR3 dir(gRSPlights[l].x - v.x, gRSPlights[l].y - v.y, gRSPlights[l].z - v.z); |
1144 | //XVECTOR3 dir(v.x-gRSPlights[l].x, v.y-gRSPlights[l].y, v.z-gRSPlights[l].z); |
1145 | float d2 = sqrtf(dir.x*dir.x+dir.y*dir.y+dir.z*dir.z); |
1146 | dir.x /= d2; |
1147 | dir.y /= d2; |
1148 | dir.z /= d2; |
1149 | |
1150 | fCosT = norm.x*dir.x + norm.y*dir.y + norm.z*dir.z; |
1151 | |
1152 | if (fCosT > 0 ) |
1153 | { |
1154 | //float f = d2/gRSPlights[l].range*50; |
1155 | float f = d2/15000*50; |
1156 | f = 1 - min(f,1); |
1157 | fCosT *= f*f; |
1158 | |
1159 | r += gRSPlights[l].fr * fCosT; |
1160 | g += gRSPlights[l].fg * fCosT; |
1161 | b += gRSPlights[l].fb * fCosT; |
1162 | } |
1163 | } |
1164 | } |
1165 | } |
1166 | |
1167 | if (r > 255) r = 255; |
1168 | if (g > 255) g = 255; |
1169 | if (b > 255) b = 255; |
1170 | return ((0xff000000)|(((uint32)r)<<16)|(((uint32)g)<<8)|((uint32)b)); |
1171 | } |
1172 | |
1173 | uint32 LightVertNew(XVECTOR4 & norm) |
1174 | { |
1175 | float fCosT; |
1176 | |
1177 | // Do ambient |
1178 | register float r = gRSP.fAmbientLightR; |
1179 | register float g = gRSP.fAmbientLightG; |
1180 | register float b = gRSP.fAmbientLightB; |
1181 | |
1182 | |
1183 | for (register unsigned int l=0; l < gRSPnumLights; l++) |
1184 | { |
1185 | fCosT = norm.x*gRSPlights[l].tx + norm.y*gRSPlights[l].ty + norm.z*gRSPlights[l].tz; |
1186 | |
1187 | if (fCosT > 0 ) |
1188 | { |
1189 | r += gRSPlights[l].fr * fCosT; |
1190 | g += gRSPlights[l].fg * fCosT; |
1191 | b += gRSPlights[l].fb * fCosT; |
1192 | } |
1193 | } |
1194 | |
1195 | if (r > 255) r = 255; |
1196 | if (g > 255) g = 255; |
1197 | if (b > 255) b = 255; |
1198 | return ((0xff000000)|(((uint32)r)<<16)|(((uint32)g)<<8)|((uint32)b)); |
1199 | } |
1200 | |
1201 | |
1202 | float zero = 0.0f; |
1203 | float onef = 1.0f; |
1204 | float fcosT; |
1205 | |
1206 | #if !defined(__GNUC__) && !defined(NO_ASM) |
1207 | __declspec( naked ) uint32 __fastcall SSELightVert() |
1208 | { |
1209 | __asm |
1210 | { |
1211 | movaps xmm3, DWORD PTR gRSP; // loading Ambient colors, xmm3 is the result color |
1212 | movaps xmm4, DWORD PTR [g_normal]; // xmm4 is the normal |
1213 | |
1214 | mov ecx, 0; |
1215 | loopback: |
1216 | cmp ecx, DWORD PTR gRSPnumLights; |
1217 | jae breakout; |
1218 | mov eax,ecx; |
1219 | imul eax,0x44; |
1220 | movups xmm5, DWORD PTR gRSPlights[eax]; // Light Dir |
1221 | movups xmm1, DWORD PTR gRSPlights[0x14][eax]; // Light color |
1222 | mulps xmm5, xmm4; // Lightdir * normals |
1223 | |
1224 | movhlps xmm0,xmm5; |
1225 | addps xmm0,xmm5; |
1226 | shufps xmm5,xmm0,0x01; |
1227 | addps xmm0,xmm5; |
1228 | |
1229 | comiss xmm0,zero; |
1230 | jc endloop |
1231 | |
1232 | shufps xmm0,xmm0,0; // fcosT |
1233 | mulps xmm1,xmm0; |
1234 | addps xmm3,xmm1; |
1235 | endloop: |
1236 | inc ecx; |
1237 | jmp loopback; |
1238 | breakout: |
1239 | |
1240 | movss xmm0,DWORD PTR real255; |
1241 | shufps xmm0,xmm0,0; |
1242 | minps xmm0,xmm3; |
1243 | |
1244 | // Without using a memory |
1245 | cvtss2si eax,xmm0; // move the 1st uint32 to eax |
1246 | shl eax,10h; |
1247 | or eax,0FF000000h; |
1248 | shufps xmm0,xmm0,0E5h; // move the 2nd uint32 to the 1st uint32 |
1249 | cvtss2si ecx,xmm0; // move the 1st uint32 to ecx |
1250 | shl ecx,8; |
1251 | or eax,ecx; |
1252 | shufps xmm0,xmm0,0E6h; // Move the 3rd uint32 to the 1st uint32 |
1253 | cvtss2si ecx,xmm0; |
1254 | or eax,ecx; |
1255 | |
1256 | ret; |
1257 | } |
1258 | } |
1259 | #elif defined(__GNUC__) && defined(__x86_64__) && !defined(NO_ASM) |
1260 | uint32 SSELightVert(void) |
1261 | { |
1262 | uint32 rval; |
1263 | float f255 = 255.0, fZero = 0.0; |
1264 | |
1265 | asm volatile(" movaps %1, %%xmm3 \n" // xmm3 == gRSP.fAmbientLight{RGBA} |
1266 | " movaps %2, %%xmm4 \n" // xmm4 == g_normal.{xyz} |
1267 | " xor %%rcx, %%rcx \n" |
1268 | "0: \n" |
1269 | " cmpl %3, %%ecx \n" |
1270 | " jae 2f \n" |
1271 | " mov %%rcx, %%rax \n" |
1272 | " imul $0x44, %%rax, %%rax \n" |
1273 | " movups (%4,%%rax,), %%xmm5 \n" // xmm5 == gRSPlights[l].{xyzr} |
1274 | " movups 20(%4,%%rax,), %%xmm1 \n" // xmm1 == gRSPlights[l].{frfgfbfa} |
1275 | " mulps %%xmm4, %%xmm5 \n" |
1276 | " movhlps %%xmm5, %%xmm0 \n" |
1277 | " addps %%xmm5, %%xmm0 \n" |
1278 | " shufps $0x01, %%xmm0, %%xmm5 \n" |
1279 | " addps %%xmm5, %%xmm0 \n" |
1280 | " comiss %6, %%xmm0 \n" |
1281 | " jc 1f \n" |
1282 | " shufps $0x00, %%xmm0, %%xmm0 \n" |
1283 | " mulps %%xmm0, %%xmm1 \n" |
1284 | " addps %%xmm1, %%xmm3 \n" |
1285 | "1: \n" |
1286 | " inc %%rcx \n" |
1287 | " jmp 0b \n" |
1288 | "2: \n" |
1289 | " movss %5, %%xmm0 \n" |
1290 | " shufps $0x00, %%xmm0, %%xmm0 \n" |
1291 | " minps %%xmm3, %%xmm0 \n" |
1292 | " cvtss2si %%xmm0, %%eax \n" |
1293 | " shll $0x10, %%eax \n" |
1294 | " orl $0xff000000, %%eax \n" |
1295 | " shufps $0xe5, %%xmm0, %%xmm0 \n" |
1296 | " cvtss2si %%xmm0, %%ecx \n" |
1297 | " shll $8, %%ecx \n" |
1298 | " orl %%ecx, %%eax \n" |
1299 | " shufps $0xe6, %%xmm0, %%xmm0 \n" |
1300 | " cvtss2si %%xmm0, %%ecx \n" |
1301 | " orl %%ecx, %%eax \n" |
1302 | : "=&a"(rval) |
1303 | : "m"(gRSP), "m"(g_normal), "m"(gRSPnumLights), "r"(gRSPlights), "m"(f255), "m"(fZero) |
1304 | : "%rcx", "memory", "cc", "%xmm0", "%xmm1", "%xmm3", "%xmm4", "%xmm5" |
1305 | ); |
1306 | return rval; |
1307 | } |
1308 | #elif !defined(NO_ASM) // 32-bit GCC assumed |
1309 | uint32 SSELightVert(void) |
1310 | { |
1311 | uint32 rval; |
1312 | float f255 = 255.0, fZero = 0.0; |
1313 | |
1314 | asm volatile(" movaps %1, %%xmm3 \n" |
1315 | " movaps %2, %%xmm4 \n" |
1316 | " xor %%ecx, %%ecx \n" |
1317 | "0: \n" |
1318 | " cmpl %3, %%ecx \n" |
1319 | " jae 2f \n" |
1320 | " mov %%ecx, %%eax \n" |
1321 | " imul $0x44, %%eax, %%eax \n" |
1322 | " movups (%4,%%eax,), %%xmm5 \n" |
1323 | " movups 20(%4,%%eax,), %%xmm1 \n" |
1324 | " mulps %%xmm4, %%xmm5 \n" |
1325 | " movhlps %%xmm5, %%xmm0 \n" |
1326 | " addps %%xmm5, %%xmm0 \n" |
1327 | " shufps $0x01, %%xmm0, %%xmm5 \n" |
1328 | " addps %%xmm5, %%xmm0 \n" |
1329 | " comiss %6, %%xmm0 \n" |
1330 | " jc 1f \n" |
1331 | " shufps $0x00, %%xmm0, %%xmm0 \n" |
1332 | " mulps %%xmm0, %%xmm1 \n" |
1333 | " addps %%xmm1, %%xmm3 \n" |
1334 | "1: \n" |
1335 | " inc %%ecx \n" |
1336 | " jmp 0b \n" |
1337 | "2: \n" |
1338 | " movss %5, %%xmm0 \n" |
1339 | " shufps $0x00, %%xmm0, %%xmm0 \n" |
1340 | " minps %%xmm3, %%xmm0 \n" |
1341 | " cvtss2si %%xmm0, %%eax \n" |
1342 | " shll $0x10, %%eax \n" |
1343 | " orl $0xff000000, %%eax \n" |
1344 | " shufps $0xe5, %%xmm0, %%xmm0 \n" |
1345 | " cvtss2si %%xmm0, %%ecx \n" |
1346 | " shll $8, %%ecx \n" |
1347 | " orl %%ecx, %%eax \n" |
1348 | " shufps $0xe6, %%xmm0, %%xmm0 \n" |
1349 | " cvtss2si %%xmm0, %%ecx \n" |
1350 | " orl %%ecx, %%eax \n" |
1351 | : "=&a"(rval) |
1352 | : "m"(gRSP), "m"(g_normal), "m"(gRSPnumLights), "r"(gRSPlights), "m"(f255), "m"(fZero) |
1353 | : "%rcx", "memory", "cc", "%xmm0", "%xmm1", "%xmm3", "%xmm4", "%xmm5" |
1354 | ); |
1355 | return rval; |
1356 | } |
1357 | #endif |
1358 | |
1359 | inline void ReplaceAlphaWithFogFactor(int i) |
1360 | { |
1361 | if( gRDP.geometryMode & G_FOG ) |
1362 | { |
1363 | // Use fog factor to replace vertex alpha |
1364 | if( g_vecProjected[i].z > 1 ) |
1365 | *(((uint8*)&(g_dwVtxDifColor[i]))+3) = 0xFF; |
1366 | if( g_vecProjected[i].z < 0 ) |
1367 | *(((uint8*)&(g_dwVtxDifColor[i]))+3) = 0; |
1368 | else |
1369 | *(((uint8*)&(g_dwVtxDifColor[i]))+3) = (uint8)(g_vecProjected[i].z*255); |
1370 | } |
1371 | } |
1372 | |
1373 | |
1374 | // Bits |
1375 | // +-+-+- |
1376 | // xxyyzz |
1377 | #define Z_NEG 0x01 |
1378 | #define Z_POS 0x02 |
1379 | #define Y_NEG 0x04 |
1380 | #define Y_POS 0x08 |
1381 | #define X_NEG 0x10 |
1382 | #define X_POS 0x20 |
1383 | |
1384 | // Assumes dwAddr has already been checked! |
1385 | // Don't inline - it's too big with the transform macros |
1386 | |
1387 | #if !defined(NO_ASM) |
1388 | void ProcessVertexDataSSE(uint32 dwAddr, uint32 dwV0, uint32 dwNum) |
1389 | { |
1390 | UpdateCombinedMatrix(); |
1391 | |
1392 | // This function is called upon SPvertex |
1393 | // - do vertex matrix transform |
1394 | // - do vertex lighting |
1395 | // - do texture cooridinate transform if needed |
1396 | // - calculate normal vector |
1397 | |
1398 | // Output: - g_vecProjected[i] -> transformed vertex x,y,z |
1399 | // - g_vecProjected[i].w -> saved vertex 1/w |
1400 | // - g_dwVtxFlags[i] -> flags |
1401 | // - g_dwVtxDifColor[i] -> vertex color |
1402 | // - g_fVtxTxtCoords[i] -> vertex texture cooridinates |
1403 | |
1404 | FiddledVtx * pVtxBase = (FiddledVtx*)(g_pRDRAMu8 + dwAddr); |
1405 | g_pVtxBase = pVtxBase; |
1406 | |
1407 | for (uint32 i = dwV0; i < dwV0 + dwNum; i++) |
1408 | { |
1409 | SP_Timing(RSP_GBI0_Vtx); |
1410 | |
1411 | FiddledVtx & vert = pVtxBase[i - dwV0]; |
1412 | |
1413 | g_vtxNonTransformed[i].x = (float)vert.x; |
1414 | g_vtxNonTransformed[i].y = (float)vert.y; |
1415 | g_vtxNonTransformed[i].z = (float)vert.z; |
1416 | |
1417 | SSEVec3Transform(i); |
1418 | |
1419 | if( gRSP.bFogEnabled ) |
1420 | { |
1421 | g_fFogCoord[i] = g_vecProjected[i].z; |
1422 | if( g_vecProjected[i].w < 0 || g_vecProjected[i].z < 0 || g_fFogCoord[i] < gRSPfFogMin ) |
1423 | g_fFogCoord[i] = gRSPfFogMin; |
1424 | } |
1425 | |
1426 | ReplaceAlphaWithFogFactor(i); |
1427 | |
1428 | |
1429 | VTX_DUMP( |
1430 | { |
1431 | uint32 *dat = (uint32*)(&vert); |
1432 | DebuggerAppendMsg("vtx %d: %08X %08X %08X %08X", i, dat[0],dat[1],dat[2],dat[3]); |
1433 | DebuggerAppendMsg(" : %f, %f, %f, %f", |
1434 | g_vtxTransformed[i].x,g_vtxTransformed[i].y,g_vtxTransformed[i].z,g_vtxTransformed[i].w); |
1435 | DebuggerAppendMsg(" : %f, %f, %f, %f", |
1436 | g_vecProjected[i].x,g_vecProjected[i].y,g_vecProjected[i].z,g_vecProjected[i].w); |
1437 | }); |
1438 | |
1439 | RSP_Vtx_Clipping(i); |
1440 | |
1441 | if( gRSP.bLightingEnable ) |
1442 | { |
1443 | g_normal.x = (float)vert.norma.nx; |
1444 | g_normal.y = (float)vert.norma.ny; |
1445 | g_normal.z = (float)vert.norma.nz; |
1446 | |
1447 | SSEVec3TransformNormal(); |
1448 | if( options.enableHackForGames != HACK_FOR_ZELDA_MM ) |
1449 | g_dwVtxDifColor[i] = SSELightVert(); |
1450 | else |
1451 | g_dwVtxDifColor[i] = LightVert(g_normal, i); |
1452 | *(((uint8*)&(g_dwVtxDifColor[i]))+3) = vert.rgba.a; // still use alpha from the vertex |
1453 | } |
1454 | else |
1455 | { |
1456 | if( (gRDP.geometryMode & G_SHADE) == 0 && gRSP.ucode < 5 ) //Shade is disabled |
1457 | { |
1458 | //FLAT shade |
1459 | g_dwVtxDifColor[i] = gRDP.primitiveColor; |
1460 | } |
1461 | else |
1462 | { |
1463 | register IColor &color = *(IColor*)&g_dwVtxDifColor[i]; |
1464 | color.b = vert.rgba.r; |
1465 | color.g = vert.rgba.g; |
1466 | color.r = vert.rgba.b; |
1467 | color.a = vert.rgba.a; |
1468 | } |
1469 | } |
1470 | |
1471 | if( options.bWinFrameMode ) |
1472 | { |
1473 | g_dwVtxDifColor[i] = COLOR_RGBA(vert.rgba.r, vert.rgba.g, vert.rgba.b, vert.rgba.a); |
1474 | } |
1475 | |
1476 | // Update texture coords n.b. need to divide tu/tv by bogus scale on addition to buffer |
1477 | |
1478 | // If the vert is already lit, then there is no normal (and hence we |
1479 | // can't generate tex coord) |
1480 | if (gRSP.bTextureGen && gRSP.bLightingEnable ) |
1481 | { |
1482 | TexGen(g_fVtxTxtCoords[i].x, g_fVtxTxtCoords[i].y); |
1483 | } |
1484 | else |
1485 | { |
1486 | g_fVtxTxtCoords[i].x = (float)vert.tu; |
1487 | g_fVtxTxtCoords[i].y = (float)vert.tv; |
1488 | } |
1489 | } |
1490 | |
1491 | VTX_DUMP(TRACE2("Setting Vertexes: %d - %d\n", dwV0, dwV0+dwNum-1)); |
1492 | DEBUGGER_PAUSE_AND_DUMP(NEXT_VERTEX_CMD,{TRACE0("Paused at Vertex Cmd");}); |
1493 | } |
1494 | #endif |
1495 | |
1496 | void ProcessVertexDataNoSSE(uint32 dwAddr, uint32 dwV0, uint32 dwNum) |
1497 | { |
1498 | |
1499 | UpdateCombinedMatrix(); |
1500 | |
1501 | // This function is called upon SPvertex |
1502 | // - do vertex matrix transform |
1503 | // - do vertex lighting |
1504 | // - do texture cooridinate transform if needed |
1505 | // - calculate normal vector |
1506 | |
1507 | // Output: - g_vecProjected[i] -> transformed vertex x,y,z |
1508 | // - g_vecProjected[i].w -> saved vertex 1/w |
1509 | // - g_dwVtxFlags[i] -> flags |
1510 | // - g_dwVtxDifColor[i] -> vertex color |
1511 | // - g_fVtxTxtCoords[i] -> vertex texture cooridinates |
1512 | |
1513 | FiddledVtx * pVtxBase = (FiddledVtx*)(g_pRDRAMu8 + dwAddr); |
1514 | g_pVtxBase = pVtxBase; |
1515 | |
1516 | for (uint32 i = dwV0; i < dwV0 + dwNum; i++) |
1517 | { |
1518 | SP_Timing(RSP_GBI0_Vtx); |
1519 | |
1520 | FiddledVtx & vert = pVtxBase[i - dwV0]; |
1521 | |
1522 | g_vtxNonTransformed[i].x = (float)vert.x; |
1523 | g_vtxNonTransformed[i].y = (float)vert.y; |
1524 | g_vtxNonTransformed[i].z = (float)vert.z; |
1525 | |
1526 | Vec3Transform(&g_vtxTransformed[i], (XVECTOR3*)&g_vtxNonTransformed[i], &gRSPworldProject); // Convert to w=1 |
1527 | |
1528 | g_vecProjected[i].w = 1.0f / g_vtxTransformed[i].w; |
1529 | g_vecProjected[i].x = g_vtxTransformed[i].x * g_vecProjected[i].w; |
1530 | g_vecProjected[i].y = g_vtxTransformed[i].y * g_vecProjected[i].w; |
1531 | if ((g_curRomInfo.bPrimaryDepthHack || options.enableHackForGames == HACK_FOR_NASCAR ) && gRDP.otherMode.depth_source ) |
1532 | { |
1533 | g_vecProjected[i].z = gRDP.fPrimitiveDepth; |
1534 | g_vtxTransformed[i].z = gRDP.fPrimitiveDepth*g_vtxTransformed[i].w; |
1535 | } |
1536 | else |
1537 | { |
1538 | g_vecProjected[i].z = g_vtxTransformed[i].z * g_vecProjected[i].w; |
1539 | } |
1540 | |
1541 | if( gRSP.bFogEnabled ) |
1542 | { |
1543 | g_fFogCoord[i] = g_vecProjected[i].z; |
1544 | if( g_vecProjected[i].w < 0 || g_vecProjected[i].z < 0 || g_fFogCoord[i] < gRSPfFogMin ) |
1545 | g_fFogCoord[i] = gRSPfFogMin; |
1546 | } |
1547 | |
1548 | VTX_DUMP( |
1549 | { |
1550 | uint32 *dat = (uint32*)(&vert); |
1551 | DebuggerAppendMsg("vtx %d: %08X %08X %08X %08X", i, dat[0],dat[1],dat[2],dat[3]); |
1552 | DebuggerAppendMsg(" : %f, %f, %f, %f", |
1553 | g_vtxTransformed[i].x,g_vtxTransformed[i].y,g_vtxTransformed[i].z,g_vtxTransformed[i].w); |
1554 | DebuggerAppendMsg(" : %f, %f, %f, %f", |
1555 | g_vecProjected[i].x,g_vecProjected[i].y,g_vecProjected[i].z,g_vecProjected[i].w); |
1556 | }); |
1557 | |
1558 | RSP_Vtx_Clipping(i); |
1559 | |
1560 | if( gRSP.bLightingEnable ) |
1561 | { |
1562 | g_normal.x = (float)vert.norma.nx; |
1563 | g_normal.y = (float)vert.norma.ny; |
1564 | g_normal.z = (float)vert.norma.nz; |
1565 | |
1566 | Vec3TransformNormal(g_normal, gRSPmodelViewTop); |
1567 | g_dwVtxDifColor[i] = LightVert(g_normal, i); |
1568 | *(((uint8*)&(g_dwVtxDifColor[i]))+3) = vert.rgba.a; // still use alpha from the vertex |
1569 | } |
1570 | else |
1571 | { |
1572 | if( (gRDP.geometryMode & G_SHADE) == 0 && gRSP.ucode < 5 ) //Shade is disabled |
1573 | { |
1574 | //FLAT shade |
1575 | g_dwVtxDifColor[i] = gRDP.primitiveColor; |
1576 | } |
1577 | else |
1578 | { |
1579 | register IColor &color = *(IColor*)&g_dwVtxDifColor[i]; |
1580 | color.b = vert.rgba.r; |
1581 | color.g = vert.rgba.g; |
1582 | color.r = vert.rgba.b; |
1583 | color.a = vert.rgba.a; |
1584 | } |
1585 | } |
1586 | |
1587 | if( options.bWinFrameMode ) |
1588 | { |
1589 | g_dwVtxDifColor[i] = COLOR_RGBA(vert.rgba.r, vert.rgba.g, vert.rgba.b, vert.rgba.a); |
1590 | } |
1591 | |
1592 | ReplaceAlphaWithFogFactor(i); |
1593 | |
1594 | // Update texture coords n.b. need to divide tu/tv by bogus scale on addition to buffer |
1595 | |
1596 | // If the vert is already lit, then there is no normal (and hence we |
1597 | // can't generate tex coord) |
1598 | if (gRSP.bTextureGen && gRSP.bLightingEnable ) |
1599 | { |
1600 | TexGen(g_fVtxTxtCoords[i].x, g_fVtxTxtCoords[i].y); |
1601 | } |
1602 | else |
1603 | { |
1604 | g_fVtxTxtCoords[i].x = (float)vert.tu; |
1605 | g_fVtxTxtCoords[i].y = (float)vert.tv; |
1606 | } |
1607 | } |
1608 | |
1609 | VTX_DUMP(TRACE2("Setting Vertexes: %d - %d\n", dwV0, dwV0+dwNum-1)); |
1610 | DEBUGGER_PAUSE_AND_DUMP(NEXT_VERTEX_CMD,{TRACE0("Paused at Vertex Cmd");}); |
1611 | } |
1612 | |
48d77f73 |
1613 | /* NEON code */ |
1614 | |
1615 | #include "RenderBase_neon.h" |
1616 | |
f18a2655 |
1617 | extern "C" void pv_neon(XVECTOR4 *g_vtxTransformed, XVECTOR4 *g_vecProjected, |
1618 | uint32 *g_dwVtxDifColor, VECTOR2 *g_fVtxTxtCoords, |
1619 | float *g_fFogCoord, uint32 *g_clipFlag2, |
48d77f73 |
1620 | uint32 dwNum, int neon_state, |
1621 | const FiddledVtx *vtx, |
f18a2655 |
1622 | const Light *gRSPlights, const float *fRSPAmbientLightRGBA, |
1623 | const XMATRIX *gRSPworldProject, const XMATRIX *gRSPmodelViewTop, |
48d77f73 |
1624 | uint32 gRSPnumLights, float gRSPfFogMin, |
1625 | uint32 primitiveColor, uint32 primitiveColor_); |
1626 | |
1627 | extern "C" int tv_direction(const XVECTOR4 *v0, const XVECTOR4 *v1, const XVECTOR4 *v2); |
f18a2655 |
1628 | |
1629 | void ProcessVertexDataNEON(uint32 dwAddr, uint32 dwV0, uint32 dwNum) |
1630 | { |
1631 | if (gRSP.bTextureGen && gRSP.bLightingEnable) { |
1632 | ProcessVertexDataNoSSE(dwAddr, dwV0,dwNum); |
1633 | return; |
1634 | } |
1635 | |
1636 | // assumtions: |
1637 | // - g_clipFlag is not used at all |
48d77f73 |
1638 | // - g_fFogCoord is not used at all |
f18a2655 |
1639 | // - g_vtxNonTransformed is not used after ProcessVertexData*() returns |
1640 | // - g_normal - same |
1641 | |
f18a2655 |
1642 | int neon_state = 0; |
1643 | if ( gRSP.bLightingEnable ) |
1644 | neon_state |= PV_NEON_ENABLE_LIGHT; |
1645 | if ( (gRDP.geometryMode & G_SHADE) || gRSP.ucode >= 5 ) |
1646 | neon_state |= PV_NEON_ENABLE_SHADE; |
1647 | if ( gRSP.bFogEnabled ) |
1648 | neon_state |= PV_NEON_ENABLE_FOG; |
1649 | if ( gRDP.geometryMode & G_FOG ) |
1650 | neon_state |= PV_NEON_FOG_ALPHA; |
1651 | |
1652 | uint32 i; |
1653 | |
1654 | UpdateCombinedMatrix(); |
1655 | |
1656 | // This function is called upon SPvertex |
1657 | // - do vertex matrix transform |
1658 | // - do vertex lighting |
1659 | // - do texture cooridinate transform if needed |
1660 | // - calculate normal vector |
1661 | |
1662 | // Output: - g_vecProjected[i] -> transformed vertex x,y,z |
1663 | // - g_vecProjected[i].w -> saved vertex 1/w |
1664 | // - g_vtxTransformed[i] |
1665 | // - g_dwVtxDifColor[i] -> vertex color |
1666 | // - g_fVtxTxtCoords[i] -> vertex texture cooridinates |
48d77f73 |
1667 | // - g_fFogCoord[i] -> unused |
f18a2655 |
1668 | // - g_clipFlag2[i] |
1669 | |
1670 | const FiddledVtx * pVtxBase = (const FiddledVtx*)(g_pRDRAMu8 + dwAddr); |
1671 | g_pVtxBase = (FiddledVtx *)pVtxBase; |
1672 | |
48d77f73 |
1673 | gRSPmodelViewTop._14 = gRSPmodelViewTop._24 = |
1674 | gRSPmodelViewTop._34 = 0; |
1675 | |
f18a2655 |
1676 | // SP_Timing(RSP_GBI0_Vtx); |
1677 | status.SPCycleCount += Timing_RSP_GBI0_Vtx * dwNum; |
1678 | |
48d77f73 |
1679 | |
1680 | #if 1 |
1681 | i = dwV0; |
1682 | pv_neon(&g_vtxTransformed[i], &g_vecProjected[i], |
1683 | &g_dwVtxDifColor[i], &g_fVtxTxtCoords[i], |
1684 | &g_fFogCoord[i], &g_clipFlag2[i], |
1685 | dwNum, neon_state, &pVtxBase[i - dwV0], |
1686 | gRSPlights, gRSP.fAmbientColors, |
1687 | &gRSPworldProject, &gRSPmodelViewTop, |
1688 | gRSPnumLights, gRSPfFogMin, |
1689 | gRDP.primitiveColor, gRDP.primitiveColor); |
1690 | #else |
f18a2655 |
1691 | |
1692 | for (i = dwV0; i < dwV0 + dwNum; i++) |
1693 | { |
1694 | const FiddledVtx & vert = pVtxBase[i - dwV0]; |
1695 | XVECTOR3 vtx_raw; // was g_vtxNonTransformed |
1696 | |
1697 | vtx_raw.x = (float)vert.x; |
1698 | vtx_raw.y = (float)vert.y; |
1699 | vtx_raw.z = (float)vert.z; |
1700 | |
1701 | Vec3Transform(&g_vtxTransformed[i], &vtx_raw, &gRSPworldProject); // Convert to w=1 |
1702 | |
1703 | g_vecProjected[i].w = 1.0f / g_vtxTransformed[i].w; |
1704 | g_vecProjected[i].x = g_vtxTransformed[i].x * g_vecProjected[i].w; |
1705 | g_vecProjected[i].y = g_vtxTransformed[i].y * g_vecProjected[i].w; |
1706 | g_vecProjected[i].z = g_vtxTransformed[i].z * g_vecProjected[i].w; |
1707 | |
f18a2655 |
1708 | // RSP_Vtx_Clipping(i); |
1709 | g_clipFlag2[i] = 0; |
1710 | if( g_vecProjected[i].w > 0 ) |
1711 | { |
1712 | if( g_vecProjected[i].x > 1 ) g_clipFlag2[i] |= X_CLIP_MAX; |
1713 | if( g_vecProjected[i].x < -1 ) g_clipFlag2[i] |= X_CLIP_MIN; |
1714 | if( g_vecProjected[i].y > 1 ) g_clipFlag2[i] |= Y_CLIP_MAX; |
1715 | if( g_vecProjected[i].y < -1 ) g_clipFlag2[i] |= Y_CLIP_MIN; |
1716 | } |
1717 | |
1718 | if( neon_state & PV_NEON_ENABLE_LIGHT ) |
1719 | { |
1720 | XVECTOR3 normal; // was g_normal |
1721 | float r, g, b; |
1722 | |
1723 | normal.x = (float)vert.norma.nx; |
1724 | normal.y = (float)vert.norma.ny; |
1725 | normal.z = (float)vert.norma.nz; |
1726 | |
1727 | Vec3TransformNormal(normal, gRSPmodelViewTop); |
1728 | |
1729 | r = gRSP.fAmbientLightR; |
1730 | g = gRSP.fAmbientLightG; |
1731 | b = gRSP.fAmbientLightB; |
1732 | |
1733 | for (unsigned int l=0; l < gRSPnumLights; l++) |
1734 | { |
1735 | float fCosT = normal.x * gRSPlights[l].x + normal.y * gRSPlights[l].y + normal.z * gRSPlights[l].z; |
1736 | |
1737 | if (fCosT > 0 ) |
1738 | { |
1739 | r += gRSPlights[l].fr * fCosT; |
1740 | g += gRSPlights[l].fg * fCosT; |
1741 | b += gRSPlights[l].fb * fCosT; |
1742 | } |
1743 | } |
1744 | if (r > 255) r = 255; |
1745 | if (g > 255) g = 255; |
1746 | if (b > 255) b = 255; |
1747 | g_dwVtxDifColor[i] = ((vert.rgba.a<<24)|(((uint32)r)<<16)|(((uint32)g)<<8)|((uint32)b)); |
1748 | } |
1749 | else if( neon_state & PV_NEON_ENABLE_SHADE ) |
1750 | { |
1751 | IColor &color = *(IColor*)&g_dwVtxDifColor[i]; |
1752 | color.b = vert.rgba.r; |
1753 | color.g = vert.rgba.g; |
1754 | color.r = vert.rgba.b; |
1755 | color.a = vert.rgba.a; |
1756 | } |
48d77f73 |
1757 | else |
1758 | g_dwVtxDifColor[i] = gRDP.primitiveColor; // FLAT shade |
f18a2655 |
1759 | |
1760 | // ReplaceAlphaWithFogFactor(i); |
1761 | if( neon_state & PV_NEON_FOG_ALPHA ) |
1762 | { |
1763 | // Use fog factor to replace vertex alpha |
1764 | if( g_vecProjected[i].z > 1 ) |
1765 | *(((uint8*)&(g_dwVtxDifColor[i]))+3) = 0xFF; |
48d77f73 |
1766 | // missing 'else' in original code?? |
1767 | else if( g_vecProjected[i].z < 0 ) |
f18a2655 |
1768 | *(((uint8*)&(g_dwVtxDifColor[i]))+3) = 0; |
1769 | else |
1770 | *(((uint8*)&(g_dwVtxDifColor[i]))+3) = (uint8)(g_vecProjected[i].z*255); |
1771 | } |
1772 | |
1773 | g_fVtxTxtCoords[i].x = (float)vert.tu; |
1774 | g_fVtxTxtCoords[i].y = (float)vert.tv; |
1775 | } |
48d77f73 |
1776 | #endif |
f18a2655 |
1777 | } |
1778 | |
1779 | bool PrepareTriangle(uint32 dwV0, uint32 dwV1, uint32 dwV2) |
d07c171f |
1780 | { |
1781 | if( status.isVertexShaderEnabled || status.bUseHW_T_L ) |
1782 | { |
1783 | g_vtxIndex[gRSP.numVertices++] = dwV0; |
1784 | g_vtxIndex[gRSP.numVertices++] = dwV1; |
1785 | g_vtxIndex[gRSP.numVertices++] = dwV2; |
1786 | status.dwNumTrisRendered++; |
1787 | gRSP.maxVertexID = max(gRSP.maxVertexID,max(dwV0,max(dwV1,dwV2))); |
1788 | } |
1789 | else |
1790 | { |
1791 | SP_Timing(SP_Each_Triangle); |
1792 | |
1793 | bool textureFlag = (CRender::g_pRender->IsTextureEnabled() || gRSP.ucode == 6 ); |
1794 | bool openGL = CDeviceBuilder::m_deviceGeneralType == OGL_DEVICE; |
1795 | |
1796 | InitVertex(dwV0, gRSP.numVertices, textureFlag, openGL); |
1797 | InitVertex(dwV1, gRSP.numVertices+1, textureFlag, openGL); |
1798 | InitVertex(dwV2, gRSP.numVertices+2, textureFlag, openGL); |
1799 | |
48d77f73 |
1800 | if( __builtin_expect(gRSP.numVertices == 0 && g_curRomInfo.bEnableTxtLOD && gRDP.otherMode.text_lod, 0) ) |
1801 | { |
1802 | if( CRender::g_pRender->IsTexel1Enable() && CRender::g_pRender->m_pColorCombiner->m_pDecodedMux->isUsed(MUX_LODFRAC) ) |
1803 | { |
1804 | ComputeLOD(openGL); |
1805 | } |
1806 | else |
1807 | { |
1808 | gRDP.LODFrac = 0; |
1809 | } |
1810 | } |
1811 | |
d07c171f |
1812 | gRSP.numVertices += 3; |
1813 | status.dwNumTrisRendered++; |
1814 | } |
1815 | |
1816 | return true; |
1817 | } |
1818 | |
1819 | |
1820 | |
1821 | // Returns TRUE if it thinks the triangle is visible |
1822 | // Returns FALSE if it is clipped |
1823 | bool IsTriangleVisible(uint32 dwV0, uint32 dwV1, uint32 dwV2) |
1824 | { |
1825 | //return true; //fix me |
1826 | |
1827 | if( status.isVertexShaderEnabled || status.bUseHW_T_L ) return true; // We won't have access to transformed vertex data |
1828 | |
1829 | DEBUGGER_ONLY_IF( (!debuggerEnableTestTris || !debuggerEnableCullFace), {return TRUE;}); |
1830 | |
1831 | #ifdef DEBUGGER |
1832 | // Check vertices are valid! |
1833 | if (dwV0 >= MAX_VERTS || dwV1 >= MAX_VERTS || dwV2 >= MAX_VERTS) |
1834 | return false; |
1835 | #endif |
1836 | |
1837 | // Here we AND all the flags. If any of the bits is set for all |
1838 | // 3 vertices, it means that all three x, y or z lie outside of |
1839 | // the current viewing volume. |
1840 | // Currently disabled - still seems a bit dodgy |
1841 | if ((gRSP.bCullFront || gRSP.bCullBack) && gRDP.otherMode.zmode != 3) |
1842 | { |
1843 | XVECTOR4 & v0 = g_vecProjected[dwV0]; |
1844 | XVECTOR4 & v1 = g_vecProjected[dwV1]; |
1845 | XVECTOR4 & v2 = g_vecProjected[dwV2]; |
1846 | |
1847 | // Only try to clip if the tri is onscreen. For some reason, this |
1848 | // method doesnt' work well when the z value is outside of screenspace |
1849 | //if (v0.z < 1 && v1.z < 1 && v2.z < 1) |
1850 | { |
48d77f73 |
1851 | #ifndef __ARM_NEON__ |
d07c171f |
1852 | float V1 = v2.x - v0.x; |
1853 | float V2 = v2.y - v0.y; |
1854 | |
1855 | float W1 = v2.x - v1.x; |
1856 | float W2 = v2.y - v1.y; |
1857 | |
1858 | float fDirection = (V1 * W2) - (V2 * W1); |
1859 | fDirection = fDirection * v1.w * v2.w * v0.w; |
1860 | //float fDirection = v0.x*v1.y-v1.x*v0.y+v1.x*v2.y-v2.x*v1.y+v2.x*v0.y-v0.x*v2.y; |
48d77f73 |
1861 | #else |
1862 | // really returns float, but we only need sign |
1863 | int fDirection = tv_direction(&v0, &v1, &v2); |
1864 | #endif |
d07c171f |
1865 | |
1866 | if (fDirection < 0 && gRSP.bCullBack) |
1867 | { |
1868 | status.dwNumTrisClipped++; |
1869 | return false; |
1870 | } |
1871 | else if (fDirection > 0 && gRSP.bCullFront) |
1872 | { |
1873 | status.dwNumTrisClipped++; |
1874 | return false; |
1875 | } |
1876 | } |
1877 | } |
1878 | |
1879 | #ifdef ENABLE_CLIP_TRI |
1880 | //if( gRSP.bRejectVtx && (g_clipFlag[dwV0]|g_clipFlag[dwV1]|g_clipFlag[dwV2]) ) |
1881 | // return; |
1882 | if( g_clipFlag2[dwV0]&g_clipFlag2[dwV1]&g_clipFlag2[dwV2] ) |
1883 | { |
1884 | //DebuggerAppendMsg("Clipped"); |
1885 | return false; |
1886 | } |
1887 | #endif |
1888 | |
1889 | return true; |
1890 | } |
1891 | |
1892 | |
1893 | void SetPrimitiveColor(uint32 dwCol, uint32 LODMin, uint32 LODFrac) |
1894 | { |
1895 | gRDP.colorsAreReloaded = true; |
1896 | gRDP.primitiveColor = dwCol; |
1897 | gRDP.primLODMin = LODMin; |
1898 | gRDP.primLODFrac = LODFrac; |
1899 | if( gRDP.primLODFrac < gRDP.primLODMin ) |
1900 | { |
1901 | gRDP.primLODFrac = gRDP.primLODMin; |
1902 | } |
1903 | |
1904 | gRDP.fvPrimitiveColor[0] = ((dwCol>>16)&0xFF)/255.0f; //r |
1905 | gRDP.fvPrimitiveColor[1] = ((dwCol>>8)&0xFF)/255.0f; //g |
1906 | gRDP.fvPrimitiveColor[2] = ((dwCol)&0xFF)/255.0f; //b |
1907 | gRDP.fvPrimitiveColor[3] = ((dwCol>>24)&0xFF)/255.0f; //a |
1908 | } |
1909 | |
1910 | void SetPrimitiveDepth(uint32 z, uint32 dwDZ) |
1911 | { |
1912 | gRDP.primitiveDepth = z & 0x7FFF; |
1913 | gRDP.fPrimitiveDepth = (float)(gRDP.primitiveDepth)/(float)0x8000; |
1914 | |
1915 | //gRDP.fPrimitiveDepth = gRDP.fPrimitiveDepth*2-1; |
1916 | /* |
1917 | z=0xFFFF -> 1 the farest |
1918 | z=0 -> -1 the nearest |
1919 | */ |
1920 | |
1921 | //how to use dwDZ? |
1922 | |
1923 | #ifdef DEBUGGER |
1924 | if( (pauseAtNext && (eventToPause == NEXT_VERTEX_CMD || eventToPause == NEXT_FLUSH_TRI )) )//&& logTriangles ) |
1925 | { |
1926 | DebuggerAppendMsg("Set prim Depth: %f, (%08X, %08X)", gRDP.fPrimitiveDepth, z, dwDZ); |
1927 | } |
1928 | #endif |
1929 | } |
1930 | |
1931 | void SetVertexXYZ(uint32 vertex, float x, float y, float z) |
1932 | { |
1933 | g_vecProjected[vertex].x = x; |
1934 | g_vecProjected[vertex].y = y; |
1935 | g_vecProjected[vertex].z = z; |
1936 | |
1937 | g_vtxTransformed[vertex].x = x*g_vtxTransformed[vertex].w; |
1938 | g_vtxTransformed[vertex].y = y*g_vtxTransformed[vertex].w; |
1939 | g_vtxTransformed[vertex].z = z*g_vtxTransformed[vertex].w; |
1940 | } |
1941 | |
1942 | void ModifyVertexInfo(uint32 where, uint32 vertex, uint32 val) |
1943 | { |
1944 | switch (where) |
1945 | { |
1946 | case RSP_MV_WORD_OFFSET_POINT_RGBA: // Modify RGBA |
1947 | { |
1948 | uint32 r = (val>>24)&0xFF; |
1949 | uint32 g = (val>>16)&0xFF; |
1950 | uint32 b = (val>>8)&0xFF; |
1951 | uint32 a = val&0xFF; |
1952 | g_dwVtxDifColor[vertex] = COLOR_RGBA(r, g, b, a); |
1953 | LOG_UCODE("Modify vert %d color, 0x%08x", vertex, g_dwVtxDifColor[vertex]); |
1954 | } |
1955 | break; |
1956 | case RSP_MV_WORD_OFFSET_POINT_XYSCREEN: // Modify X,Y |
1957 | { |
1958 | uint16 nX = (uint16)(val>>16); |
1959 | short x = *((short*)&nX); |
1960 | x /= 4; |
1961 | |
1962 | uint16 nY = (uint16)(val&0xFFFF); |
1963 | short y = *((short*)&nY); |
1964 | y /= 4; |
1965 | |
1966 | // Should do viewport transform. |
1967 | |
1968 | |
1969 | x -= windowSetting.uViWidth/2; |
1970 | y = windowSetting.uViHeight/2-y; |
1971 | |
1972 | if( options.bEnableHacks && ((*g_GraphicsInfo.VI_X_SCALE_REG)&0xF) != 0 ) |
1973 | { |
1974 | // Tarzan |
1975 | // I don't know why Tarzan is different |
1976 | SetVertexXYZ(vertex, x/windowSetting.fViWidth, y/windowSetting.fViHeight, g_vecProjected[vertex].z); |
1977 | } |
1978 | else |
1979 | { |
1980 | // Toy Story 2 and other games |
1981 | SetVertexXYZ(vertex, x*2/windowSetting.fViWidth, y*2/windowSetting.fViHeight, g_vecProjected[vertex].z); |
1982 | } |
1983 | |
1984 | LOG_UCODE("Modify vert %d: x=%d, y=%d", vertex, x, y); |
1985 | VTX_DUMP(TRACE3("Modify vert %d: (%d,%d)", vertex, x, y)); |
1986 | } |
1987 | break; |
1988 | case RSP_MV_WORD_OFFSET_POINT_ZSCREEN: // Modify C |
1989 | { |
1990 | int z = val>>16; |
1991 | |
1992 | SetVertexXYZ(vertex, g_vecProjected[vertex].x, g_vecProjected[vertex].y, (((float)z/0x03FF)+0.5f)/2.0f ); |
1993 | LOG_UCODE("Modify vert %d: z=%d", vertex, z); |
1994 | VTX_DUMP(TRACE2("Modify vert %d: z=%d", vertex, z)); |
1995 | } |
1996 | break; |
1997 | case RSP_MV_WORD_OFFSET_POINT_ST: // Texture |
1998 | { |
1999 | short tu = short(val>>16); |
2000 | short tv = short(val & 0xFFFF); |
2001 | float ftu = tu / 32.0f; |
2002 | float ftv = tv / 32.0f; |
2003 | LOG_UCODE(" Setting vertex %d tu/tv to %f, %f", vertex, (float)tu, (float)tv); |
2004 | CRender::g_pRender->SetVtxTextureCoord(vertex, ftu/gRSP.fTexScaleX, ftv/gRSP.fTexScaleY); |
2005 | } |
2006 | break; |
2007 | } |
2008 | DEBUGGER_PAUSE_AND_DUMP(NEXT_VERTEX_CMD,{TRACE0("Paused at ModVertex Cmd");}); |
2009 | } |
2010 | |
2011 | void ProcessVertexDataDKR(uint32 dwAddr, uint32 dwV0, uint32 dwNum) |
2012 | { |
2013 | UpdateCombinedMatrix(); |
2014 | |
2015 | long long pVtxBase = (long long) (g_pRDRAMu8 + dwAddr); |
2016 | g_pVtxBase = (FiddledVtx*)pVtxBase; |
2017 | |
2018 | Matrix &matWorldProject = gRSP.DKRMatrixes[gRSP.DKRCMatrixIndex]; |
2019 | |
2020 | int nOff; |
2021 | |
2022 | bool addbase=false; |
2023 | if ((!gRSP.DKRBillBoard) || (gRSP.DKRCMatrixIndex != 2) ) |
2024 | addbase = false; |
2025 | else |
2026 | addbase = true; |
2027 | |
2028 | if( addbase && gRSP.DKRVtxCount == 0 && dwNum > 1 ) |
2029 | { |
2030 | gRSP.DKRVtxCount++; |
2031 | } |
2032 | |
2033 | LOG_UCODE(" ProcessVertexDataDKR, CMatrix = %d, Add base=%s", gRSP.DKRCMatrixIndex, gRSP.DKRBillBoard?"true":"false"); |
2034 | VTX_DUMP(TRACE2("DKR Setting Vertexes\nCMatrix = %d, Add base=%s", gRSP.DKRCMatrixIndex, gRSP.DKRBillBoard?"true":"false")); |
2035 | |
2036 | nOff = 0; |
2037 | uint32 end = dwV0 + dwNum; |
2038 | for (uint32 i = dwV0; i < end; i++) |
2039 | { |
2040 | XVECTOR3 w; |
2041 | |
2042 | g_vtxNonTransformed[i].x = (float)*(short*)((pVtxBase+nOff + 0) ^ 2); |
2043 | g_vtxNonTransformed[i].y = (float)*(short*)((pVtxBase+nOff + 2) ^ 2); |
2044 | g_vtxNonTransformed[i].z = (float)*(short*)((pVtxBase+nOff + 4) ^ 2); |
2045 | |
2046 | //if( status.isSSEEnabled ) |
2047 | // SSEVec3TransformDKR(g_vtxTransformed[i], g_vtxNonTransformed[i]); |
2048 | //else |
2049 | Vec3Transform(&g_vtxTransformed[i], (XVECTOR3*)&g_vtxNonTransformed[i], &matWorldProject); // Convert to w=1 |
2050 | |
2051 | if( gRSP.DKRVtxCount == 0 && dwNum==1 ) |
2052 | { |
2053 | gRSP.DKRBaseVec.x = g_vtxTransformed[i].x; |
2054 | gRSP.DKRBaseVec.y = g_vtxTransformed[i].y; |
2055 | gRSP.DKRBaseVec.z = g_vtxTransformed[i].z; |
2056 | gRSP.DKRBaseVec.w = g_vtxTransformed[i].w; |
2057 | } |
2058 | else if( addbase ) |
2059 | { |
2060 | g_vtxTransformed[i].x += gRSP.DKRBaseVec.x; |
2061 | g_vtxTransformed[i].y += gRSP.DKRBaseVec.y; |
2062 | g_vtxTransformed[i].z += gRSP.DKRBaseVec.z; |
2063 | g_vtxTransformed[i].w = gRSP.DKRBaseVec.w; |
2064 | } |
2065 | |
2066 | g_vecProjected[i].w = 1.0f / g_vtxTransformed[i].w; |
2067 | g_vecProjected[i].x = g_vtxTransformed[i].x * g_vecProjected[i].w; |
2068 | g_vecProjected[i].y = g_vtxTransformed[i].y * g_vecProjected[i].w; |
2069 | g_vecProjected[i].z = g_vtxTransformed[i].z * g_vecProjected[i].w; |
2070 | |
2071 | gRSP.DKRVtxCount++; |
2072 | |
2073 | VTX_DUMP(TRACE5("vtx %d: %f, %f, %f, %f", i, |
2074 | g_vtxTransformed[i].x,g_vtxTransformed[i].y,g_vtxTransformed[i].z,g_vtxTransformed[i].w)); |
2075 | |
2076 | if( gRSP.bFogEnabled ) |
2077 | { |
2078 | g_fFogCoord[i] = g_vecProjected[i].z; |
2079 | if( g_vecProjected[i].w < 0 || g_vecProjected[i].z < 0 || g_fFogCoord[i] < gRSPfFogMin ) |
2080 | g_fFogCoord[i] = gRSPfFogMin; |
2081 | } |
2082 | |
2083 | RSP_Vtx_Clipping(i); |
2084 | |
2085 | short wA = *(short*)((pVtxBase+nOff + 6) ^ 2); |
2086 | short wB = *(short*)((pVtxBase+nOff + 8) ^ 2); |
2087 | |
2088 | s8 r = (s8)(wA >> 8); |
2089 | s8 g = (s8)(wA); |
2090 | s8 b = (s8)(wB >> 8); |
2091 | s8 a = (s8)(wB); |
2092 | |
2093 | if (gRSP.bLightingEnable) |
2094 | { |
2095 | g_normal.x = (char)r; //norma.nx; |
2096 | g_normal.y = (char)g; //norma.ny; |
2097 | g_normal.z = (char)b; //norma.nz; |
2098 | |
2099 | Vec3TransformNormal(g_normal, matWorldProject) |
2100 | #if !defined(NO_ASM) |
2101 | if( status.isSSEEnabled ) |
2102 | g_dwVtxDifColor[i] = SSELightVert(); |
2103 | else |
2104 | #endif |
2105 | g_dwVtxDifColor[i] = LightVert(g_normal, i); |
2106 | } |
2107 | else |
2108 | { |
2109 | int nR, nG, nB, nA; |
2110 | |
2111 | nR = r; |
2112 | nG = g; |
2113 | nB = b; |
2114 | nA = a; |
2115 | // Assign true vert colour after lighting/fogging |
2116 | g_dwVtxDifColor[i] = COLOR_RGBA(nR, nG, nB, nA); |
2117 | } |
2118 | |
2119 | ReplaceAlphaWithFogFactor(i); |
2120 | |
2121 | g_fVtxTxtCoords[i].x = g_fVtxTxtCoords[i].y = 1; |
2122 | |
2123 | nOff += 10; |
2124 | } |
2125 | |
2126 | |
2127 | DEBUGGER_PAUSE_AND_DUMP(NEXT_VERTEX_CMD,{DebuggerAppendMsg("Paused at DKR Vertex Cmd, v0=%d, vn=%d, addr=%08X", dwV0, dwNum, dwAddr);}); |
2128 | } |
2129 | |
2130 | |
2131 | extern uint32 dwPDCIAddr; |
2132 | void ProcessVertexDataPD(uint32 dwAddr, uint32 dwV0, uint32 dwNum) |
2133 | { |
2134 | UpdateCombinedMatrix(); |
2135 | |
2136 | N64VtxPD * pVtxBase = (N64VtxPD*)(g_pRDRAMu8 + dwAddr); |
2137 | g_pVtxBase = (FiddledVtx*)pVtxBase; // Fix me |
2138 | |
2139 | for (uint32 i = dwV0; i < dwV0 + dwNum; i++) |
2140 | { |
2141 | N64VtxPD &vert = pVtxBase[i - dwV0]; |
2142 | |
2143 | g_vtxNonTransformed[i].x = (float)vert.x; |
2144 | g_vtxNonTransformed[i].y = (float)vert.y; |
2145 | g_vtxNonTransformed[i].z = (float)vert.z; |
2146 | |
2147 | #if !defined(NO_ASM) |
2148 | if( status.isSSEEnabled ) |
2149 | SSEVec3Transform(i); |
2150 | else |
2151 | #endif |
2152 | { |
2153 | Vec3Transform(&g_vtxTransformed[i], (XVECTOR3*)&g_vtxNonTransformed[i], &gRSPworldProject); // Convert to w=1 |
2154 | g_vecProjected[i].w = 1.0f / g_vtxTransformed[i].w; |
2155 | g_vecProjected[i].x = g_vtxTransformed[i].x * g_vecProjected[i].w; |
2156 | g_vecProjected[i].y = g_vtxTransformed[i].y * g_vecProjected[i].w; |
2157 | g_vecProjected[i].z = g_vtxTransformed[i].z * g_vecProjected[i].w; |
2158 | } |
2159 | |
2160 | g_fFogCoord[i] = g_vecProjected[i].z; |
2161 | if( g_vecProjected[i].w < 0 || g_vecProjected[i].z < 0 || g_fFogCoord[i] < gRSPfFogMin ) |
2162 | g_fFogCoord[i] = gRSPfFogMin; |
2163 | |
2164 | RSP_Vtx_Clipping(i); |
2165 | |
2166 | uint8 *addr = g_pRDRAMu8+dwPDCIAddr+ (vert.cidx&0xFF); |
2167 | uint32 a = addr[0]; |
2168 | uint32 r = addr[3]; |
2169 | uint32 g = addr[2]; |
2170 | uint32 b = addr[1]; |
2171 | |
2172 | if( gRSP.bLightingEnable ) |
2173 | { |
2174 | g_normal.x = (char)r; |
2175 | g_normal.y = (char)g; |
2176 | g_normal.z = (char)b; |
2177 | #if !defined(NO_ASM) |
2178 | if( status.isSSEEnabled ) |
2179 | { |
2180 | SSEVec3TransformNormal(); |
2181 | g_dwVtxDifColor[i] = SSELightVert(); |
2182 | } |
2183 | else |
2184 | #endif |
2185 | { |
2186 | Vec3TransformNormal(g_normal, gRSPmodelViewTop); |
2187 | g_dwVtxDifColor[i] = LightVert(g_normal, i); |
2188 | } |
2189 | *(((uint8*)&(g_dwVtxDifColor[i]))+3) = (uint8)a; // still use alpha from the vertex |
2190 | } |
2191 | else |
2192 | { |
2193 | if( (gRDP.geometryMode & G_SHADE) == 0 && gRSP.ucode < 5 ) //Shade is disabled |
2194 | { |
2195 | g_dwVtxDifColor[i] = gRDP.primitiveColor; |
2196 | } |
2197 | else //FLAT shade |
2198 | { |
2199 | g_dwVtxDifColor[i] = COLOR_RGBA(r, g, b, a); |
2200 | } |
2201 | } |
2202 | |
2203 | if( options.bWinFrameMode ) |
2204 | { |
2205 | g_dwVtxDifColor[i] = COLOR_RGBA(r, g, b, a); |
2206 | } |
2207 | |
2208 | ReplaceAlphaWithFogFactor(i); |
2209 | |
2210 | VECTOR2 & t = g_fVtxTxtCoords[i]; |
2211 | if (gRSP.bTextureGen && gRSP.bLightingEnable ) |
2212 | { |
2213 | // Not sure if we should transform the normal here |
2214 | //Matrix & matWV = gRSP.projectionMtxs[gRSP.projectionMtxTop]; |
2215 | //Vec3TransformNormal(g_normal, matWV); |
2216 | |
2217 | TexGen(g_fVtxTxtCoords[i].x, g_fVtxTxtCoords[i].y); |
2218 | } |
2219 | else |
2220 | { |
2221 | t.x = vert.s; |
2222 | t.y = vert.t; |
2223 | } |
2224 | |
2225 | |
2226 | VTX_DUMP( |
2227 | { |
2228 | DebuggerAppendMsg("vtx %d: %d %d %d", i, vert.x,vert.y,vert.z); |
2229 | DebuggerAppendMsg(" : %f, %f, %f, %f", |
2230 | g_vtxTransformed[i].x,g_vtxTransformed[i].y,g_vtxTransformed[i].z,g_vtxTransformed[i].w); |
2231 | DebuggerAppendMsg(" : %X, %X, %X, %X", r,g,b,a); |
2232 | DebuggerAppendMsg(" : u=%f, v=%f", t.x, t.y); |
2233 | }); |
2234 | } |
2235 | |
2236 | VTX_DUMP(TRACE2("Setting Vertexes: %d - %d\n", dwV0, dwV0+dwNum-1)); |
2237 | DEBUGGER_PAUSE_AND_DUMP(NEXT_VERTEX_CMD,{TRACE0("Paused at Vertex Cmd");}); |
2238 | } |
2239 | |
2240 | extern uint32 dwConkerVtxZAddr; |
2241 | void ProcessVertexDataConker(uint32 dwAddr, uint32 dwV0, uint32 dwNum) |
2242 | { |
2243 | UpdateCombinedMatrix(); |
2244 | |
2245 | FiddledVtx * pVtxBase = (FiddledVtx*)(g_pRDRAMu8 + dwAddr); |
2246 | g_pVtxBase = pVtxBase; |
2247 | //short *vertexColoraddr = (short*)(g_pRDRAMu8+dwConkerVtxZAddr); |
2248 | |
2249 | for (uint32 i = dwV0; i < dwV0 + dwNum; i++) |
2250 | { |
2251 | SP_Timing(RSP_GBI0_Vtx); |
2252 | |
2253 | FiddledVtx & vert = pVtxBase[i - dwV0]; |
2254 | |
2255 | g_vtxNonTransformed[i].x = (float)vert.x; |
2256 | g_vtxNonTransformed[i].y = (float)vert.y; |
2257 | g_vtxNonTransformed[i].z = (float)vert.z; |
2258 | |
2259 | #if !defined(NO_ASM) |
2260 | if( status.isSSEEnabled ) |
2261 | SSEVec3Transform(i); |
2262 | else |
2263 | #endif |
2264 | { |
2265 | Vec3Transform(&g_vtxTransformed[i], (XVECTOR3*)&g_vtxNonTransformed[i], &gRSPworldProject); // Convert to w=1 |
2266 | g_vecProjected[i].w = 1.0f / g_vtxTransformed[i].w; |
2267 | g_vecProjected[i].x = g_vtxTransformed[i].x * g_vecProjected[i].w; |
2268 | g_vecProjected[i].y = g_vtxTransformed[i].y * g_vecProjected[i].w; |
2269 | g_vecProjected[i].z = g_vtxTransformed[i].z * g_vecProjected[i].w; |
2270 | } |
2271 | |
2272 | g_fFogCoord[i] = g_vecProjected[i].z; |
2273 | if( g_vecProjected[i].w < 0 || g_vecProjected[i].z < 0 || g_fFogCoord[i] < gRSPfFogMin ) |
2274 | g_fFogCoord[i] = gRSPfFogMin; |
2275 | |
2276 | VTX_DUMP( |
2277 | { |
2278 | uint32 *dat = (uint32*)(&vert); |
2279 | DebuggerAppendMsg("vtx %d: %08X %08X %08X %08X", i, dat[0],dat[1],dat[2],dat[3]); |
2280 | DebuggerAppendMsg(" : %f, %f, %f, %f", |
2281 | g_vtxTransformed[i].x,g_vtxTransformed[i].y,g_vtxTransformed[i].z,g_vtxTransformed[i].w); |
2282 | DebuggerAppendMsg(" : %f, %f, %f, %f", |
2283 | g_vecProjected[i].x,g_vecProjected[i].y,g_vecProjected[i].z,g_vecProjected[i].w); |
2284 | }); |
2285 | |
2286 | RSP_Vtx_Clipping(i); |
2287 | |
2288 | if( gRSP.bLightingEnable ) |
2289 | { |
2290 | { |
2291 | uint32 r= ((gRSP.ambientLightColor>>16)&0xFF); |
2292 | uint32 g= ((gRSP.ambientLightColor>> 8)&0xFF); |
2293 | uint32 b= ((gRSP.ambientLightColor )&0xFF); |
2294 | for( uint32 k=1; k<=gRSPnumLights; k++) |
2295 | { |
2296 | r += gRSPlights[k].r; |
2297 | g += gRSPlights[k].g; |
2298 | b += gRSPlights[k].b; |
2299 | } |
2300 | if( r>255 ) r=255; |
2301 | if( g>255 ) g=255; |
2302 | if( b>255 ) b=255; |
2303 | r *= vert.rgba.r ; |
2304 | g *= vert.rgba.g ; |
2305 | b *= vert.rgba.b ; |
2306 | r >>= 8; |
2307 | g >>= 8; |
2308 | b >>= 8; |
2309 | g_dwVtxDifColor[i] = 0xFF000000; |
2310 | g_dwVtxDifColor[i] |= (r<<16); |
2311 | g_dwVtxDifColor[i] |= (g<< 8); |
2312 | g_dwVtxDifColor[i] |= (b ); |
2313 | } |
2314 | |
2315 | *(((uint8*)&(g_dwVtxDifColor[i]))+3) = vert.rgba.a; // still use alpha from the vertex |
2316 | } |
2317 | else |
2318 | { |
2319 | if( (gRDP.geometryMode & G_SHADE) == 0 && gRSP.ucode < 5 ) //Shade is disabled |
2320 | { |
2321 | g_dwVtxDifColor[i] = gRDP.primitiveColor; |
2322 | } |
2323 | else //FLAT shade |
2324 | { |
2325 | g_dwVtxDifColor[i] = COLOR_RGBA(vert.rgba.r, vert.rgba.g, vert.rgba.b, vert.rgba.a); |
2326 | } |
2327 | } |
2328 | |
2329 | if( options.bWinFrameMode ) |
2330 | { |
2331 | //g_vecProjected[i].z = 0; |
2332 | g_dwVtxDifColor[i] = COLOR_RGBA(vert.rgba.r, vert.rgba.g, vert.rgba.b, vert.rgba.a); |
2333 | } |
2334 | |
2335 | ReplaceAlphaWithFogFactor(i); |
2336 | |
2337 | // Update texture coords n.b. need to divide tu/tv by bogus scale on addition to buffer |
2338 | //VECTOR2 & t = g_fVtxTxtCoords[i]; |
2339 | |
2340 | // If the vert is already lit, then there is no normal (and hence we |
2341 | // can't generate tex coord) |
2342 | if (gRSP.bTextureGen && gRSP.bLightingEnable ) |
2343 | { |
2344 | g_normal.x = (float)*(char*)(g_pRDRAMu8+ (((i<<1)+0)^3)+dwConkerVtxZAddr); |
2345 | g_normal.y = (float)*(char*)(g_pRDRAMu8+ (((i<<1)+1)^3)+dwConkerVtxZAddr); |
2346 | g_normal.z = (float)*(char*)(g_pRDRAMu8+ (((i<<1)+2)^3)+dwConkerVtxZAddr); |
2347 | Vec3TransformNormal(g_normal, gRSPmodelViewTop); |
2348 | TexGen(g_fVtxTxtCoords[i].x, g_fVtxTxtCoords[i].y); |
2349 | } |
2350 | else |
2351 | { |
2352 | g_fVtxTxtCoords[i].x = (float)vert.tu; |
2353 | g_fVtxTxtCoords[i].y = (float)vert.tv; |
2354 | } |
2355 | } |
2356 | |
2357 | VTX_DUMP(TRACE2("Setting Vertexes: %d - %d\n", dwV0, dwV0+dwNum-1)); |
2358 | DEBUGGER_PAUSE_AND_DUMP(NEXT_VERTEX_CMD,{DebuggerAppendMsg("Paused at Vertex Cmd");}); |
2359 | } |
2360 | |
2361 | |
2362 | typedef struct{ |
2363 | short y; |
2364 | short x; |
2365 | short flag; |
2366 | short z; |
2367 | } RS_Vtx_XYZ; |
2368 | |
2369 | typedef union { |
2370 | struct { |
2371 | uint8 a; |
2372 | uint8 b; |
2373 | uint8 g; |
2374 | uint8 r; |
2375 | }; |
2376 | struct { |
2377 | char na; //a |
2378 | char nz; //b |
2379 | char ny; //g |
2380 | char nx; //r |
2381 | }; |
2382 | } RS_Vtx_Color; |
2383 | |
2384 | |
2385 | void ProcessVertexData_Rogue_Squadron(uint32 dwXYZAddr, uint32 dwColorAddr, uint32 dwXYZCmd, uint32 dwColorCmd) |
2386 | { |
2387 | UpdateCombinedMatrix(); |
2388 | |
2389 | uint32 dwV0 = 0; |
2390 | uint32 dwNum = (dwXYZCmd&0xFF00)>>10; |
2391 | |
2392 | RS_Vtx_XYZ * pVtxXYZBase = (RS_Vtx_XYZ*)(g_pRDRAMu8 + dwXYZAddr); |
2393 | RS_Vtx_Color * pVtxColorBase = (RS_Vtx_Color*)(g_pRDRAMu8 + dwColorAddr); |
2394 | |
2395 | uint32 i; |
2396 | for (i = dwV0; i < dwV0 + dwNum; i++) |
2397 | { |
2398 | RS_Vtx_XYZ & vertxyz = pVtxXYZBase[i - dwV0]; |
2399 | RS_Vtx_Color & vertcolors = pVtxColorBase[i - dwV0]; |
2400 | |
2401 | g_vtxNonTransformed[i].x = (float)vertxyz.x; |
2402 | g_vtxNonTransformed[i].y = (float)vertxyz.y; |
2403 | g_vtxNonTransformed[i].z = (float)vertxyz.z; |
2404 | |
2405 | #if !defined(NO_ASM) |
2406 | if( status.isSSEEnabled ) |
2407 | SSEVec3Transform(i); |
2408 | else |
2409 | #endif |
2410 | { |
2411 | Vec3Transform(&g_vtxTransformed[i], (XVECTOR3*)&g_vtxNonTransformed[i], &gRSPworldProject); // Convert to w=1 |
2412 | g_vecProjected[i].w = 1.0f / g_vtxTransformed[i].w; |
2413 | g_vecProjected[i].x = g_vtxTransformed[i].x * g_vecProjected[i].w; |
2414 | g_vecProjected[i].y = g_vtxTransformed[i].y * g_vecProjected[i].w; |
2415 | g_vecProjected[i].z = g_vtxTransformed[i].z * g_vecProjected[i].w; |
2416 | } |
2417 | |
2418 | VTX_DUMP( |
2419 | { |
2420 | DebuggerAppendMsg(" : %f, %f, %f, %f", |
2421 | g_vtxTransformed[i].x,g_vtxTransformed[i].y,g_vtxTransformed[i].z,g_vtxTransformed[i].w); |
2422 | DebuggerAppendMsg(" : %f, %f, %f, %f", |
2423 | g_vecProjected[i].x,g_vecProjected[i].y,g_vecProjected[i].z,g_vecProjected[i].w); |
2424 | }); |
2425 | |
2426 | g_fFogCoord[i] = g_vecProjected[i].z; |
2427 | if( g_vecProjected[i].w < 0 || g_vecProjected[i].z < 0 || g_fFogCoord[i] < gRSPfFogMin ) |
2428 | g_fFogCoord[i] = gRSPfFogMin; |
2429 | |
2430 | RSP_Vtx_Clipping(i); |
2431 | |
2432 | if( gRSP.bLightingEnable ) |
2433 | { |
2434 | g_normal.x = (float)vertcolors.nx; |
2435 | g_normal.y = (float)vertcolors.ny; |
2436 | g_normal.z = (float)vertcolors.nz; |
2437 | |
2438 | #if !defined(NO_ASM) |
2439 | if( status.isSSEEnabled ) |
2440 | { |
2441 | SSEVec3TransformNormal(); |
2442 | g_dwVtxDifColor[i] = SSELightVert(); |
2443 | } |
2444 | else |
2445 | #endif |
2446 | { |
2447 | Vec3TransformNormal(g_normal, gRSPmodelViewTop); |
2448 | g_dwVtxDifColor[i] = LightVert(g_normal, i); |
2449 | } |
2450 | *(((uint8*)&(g_dwVtxDifColor[i]))+3) = vertcolors.a; // still use alpha from the vertex |
2451 | } |
2452 | else |
2453 | { |
2454 | if( (gRDP.geometryMode & G_SHADE) == 0 && gRSP.ucode < 5 ) //Shade is disabled |
2455 | { |
2456 | g_dwVtxDifColor[i] = gRDP.primitiveColor; |
2457 | } |
2458 | else //FLAT shade |
2459 | { |
2460 | g_dwVtxDifColor[i] = COLOR_RGBA(vertcolors.r, vertcolors.g, vertcolors.b, vertcolors.a); |
2461 | } |
2462 | } |
2463 | |
2464 | if( options.bWinFrameMode ) |
2465 | { |
2466 | g_dwVtxDifColor[i] = COLOR_RGBA(vertcolors.r, vertcolors.g, vertcolors.b, vertcolors.a); |
2467 | } |
2468 | |
2469 | ReplaceAlphaWithFogFactor(i); |
2470 | |
2471 | /* |
2472 | // Update texture coords n.b. need to divide tu/tv by bogus scale on addition to buffer |
2473 | VECTOR2 & t = g_fVtxTxtCoords[i]; |
2474 | |
2475 | // If the vert is already lit, then there is no normal (and hence we |
2476 | // can't generate tex coord) |
2477 | if (gRSP.bTextureGen && gRSP.bLightingEnable && g_textures[gRSP.curTile].m_bTextureEnable ) |
2478 | { |
2479 | TexGen(g_fVtxTxtCoords[i].x, g_fVtxTxtCoords[i].y); |
2480 | } |
2481 | else |
2482 | { |
2483 | t.x = (float)vert.tu; |
2484 | t.y = (float)vert.tv; |
2485 | } |
2486 | */ |
2487 | } |
2488 | |
2489 | VTX_DUMP(TRACE2("Setting Vertexes: %d - %d\n", dwV0, dwV0+dwNum-1)); |
2490 | DEBUGGER_PAUSE_AND_DUMP(NEXT_VERTEX_CMD,{TRACE0("Paused at Vertex Cmd");}); |
2491 | } |
2492 | |
2493 | void SetLightCol(uint32 dwLight, uint32 dwCol) |
2494 | { |
2495 | gRSPlights[dwLight].r = (uint8)((dwCol >> 24)&0xFF); |
2496 | gRSPlights[dwLight].g = (uint8)((dwCol >> 16)&0xFF); |
2497 | gRSPlights[dwLight].b = (uint8)((dwCol >> 8)&0xFF); |
2498 | gRSPlights[dwLight].a = 255; // Ignore light alpha |
2499 | gRSPlights[dwLight].fr = (float)gRSPlights[dwLight].r; |
2500 | gRSPlights[dwLight].fg = (float)gRSPlights[dwLight].g; |
2501 | gRSPlights[dwLight].fb = (float)gRSPlights[dwLight].b; |
2502 | gRSPlights[dwLight].fa = 255; // Ignore light alpha |
2503 | |
2504 | //TRACE1("Set light %d color", dwLight); |
2505 | LIGHT_DUMP(TRACE2("Set Light %d color: %08X", dwLight, dwCol)); |
2506 | } |
2507 | |
2508 | void SetLightDirection(uint32 dwLight, float x, float y, float z, float range) |
2509 | { |
2510 | //gRSP.bLightIsUpdated = true; |
2511 | |
2512 | //gRSPlights[dwLight].ox = x; |
2513 | //gRSPlights[dwLight].oy = y; |
2514 | //gRSPlights[dwLight].oz = z; |
2515 | |
2516 | register float w = range == 0 ? (float)sqrt(x*x+y*y+z*z) : 1; |
2517 | |
2518 | gRSPlights[dwLight].x = x/w; |
2519 | gRSPlights[dwLight].y = y/w; |
2520 | gRSPlights[dwLight].z = z/w; |
2521 | gRSPlights[dwLight].range = range; |
2522 | DEBUGGER_PAUSE_AND_DUMP(NEXT_SET_LIGHT,TRACE5("Set Light %d dir: %.4f, %.4f, %.4f, %.4f", dwLight, x, y, z, range)); |
2523 | } |
2524 | |
2525 | static float maxS0, maxT0; |
2526 | static float maxS1, maxT1; |
2527 | static bool validS0, validT0; |
2528 | static bool validS1, validT1; |
2529 | |
2530 | void LogTextureCoords(float fTex0S, float fTex0T, float fTex1S, float fTex1T) |
2531 | { |
2532 | if( validS0 ) |
2533 | { |
2534 | if( fTex0S<0 || fTex0S>maxS0 ) validS0 = false; |
2535 | } |
2536 | if( validT0 ) |
2537 | { |
2538 | if( fTex0T<0 || fTex0T>maxT0 ) validT0 = false; |
2539 | } |
2540 | if( validS1 ) |
2541 | { |
2542 | if( fTex1S<0 || fTex1S>maxS1 ) validS1 = false; |
2543 | } |
2544 | if( validT1 ) |
2545 | { |
2546 | if( fTex1T<0 || fTex1T>maxT1 ) validT1 = false; |
2547 | } |
2548 | } |
2549 | |
2550 | bool CheckTextureCoords(int tex) |
2551 | { |
2552 | if( tex==0 ) |
2553 | { |
2554 | return validS0&&validT0; |
2555 | } |
2556 | else |
2557 | { |
2558 | return validS1&&validT1; |
2559 | } |
2560 | } |
2561 | |
2562 | void ResetTextureCoordsLog(float maxs0, float maxt0, float maxs1, float maxt1) |
2563 | { |
2564 | maxS0 = maxs0; |
2565 | maxT0 = maxt0; |
2566 | maxS1 = maxs1; |
2567 | maxT1 = maxt1; |
2568 | validS0 = validT0 = true; |
2569 | validS1 = validT1 = true; |
2570 | } |
2571 | |
2572 | void ForceMainTextureIndex(int dwTile) |
2573 | { |
2574 | if( dwTile == 1 && !(CRender::g_pRender->IsTexel0Enable()) && CRender::g_pRender->IsTexel1Enable() ) |
2575 | { |
2576 | // Hack |
2577 | gRSP.curTile = 0; |
2578 | } |
2579 | else |
2580 | { |
2581 | gRSP.curTile = dwTile; |
2582 | } |
2583 | } |
2584 | |
2585 | float HackZ2(float z) |
2586 | { |
2587 | z = (z+9)/10; |
2588 | return z; |
2589 | } |
2590 | |
2591 | float HackZ(float z) |
2592 | { |
2593 | return HackZ2(z); |
2594 | |
2595 | if( z < 0.1 && z >= 0 ) |
2596 | z = (.1f+z)/2; |
2597 | else if( z < 0 ) |
2598 | //return (10+z)/100; |
2599 | z = (expf(z)/20); |
2600 | return z; |
2601 | } |
2602 | |
2603 | void HackZ(std::vector<XVECTOR3>& points) |
2604 | { |
2605 | int size = points.size(); |
2606 | for( int i=0; i<size; i++) |
2607 | { |
2608 | XVECTOR3 &v = points[i]; |
2609 | v.z = (float)HackZ(v.z); |
2610 | } |
2611 | } |
2612 | |
2613 | void HackZAll() |
2614 | { |
2615 | if( CDeviceBuilder::m_deviceGeneralType == DIRECTX_DEVICE ) |
2616 | { |
2617 | for( uint32 i=0; i<gRSP.numVertices; i++) |
2618 | { |
2619 | g_vtxBuffer[i].z = HackZ(g_vtxBuffer[i].z); |
2620 | } |
2621 | } |
2622 | else |
2623 | { |
2624 | for( uint32 i=0; i<gRSP.numVertices; i++) |
2625 | { |
2626 | float w = g_vtxProjected5[i][3]; |
2627 | g_vtxProjected5[i][2] = HackZ(g_vtxProjected5[i][2]/w)*w; |
2628 | } |
2629 | } |
2630 | } |
2631 | |
2632 | |
2633 | extern XMATRIX reverseXY; |
2634 | extern XMATRIX reverseY; |
2635 | |
2636 | void UpdateCombinedMatrix() |
2637 | { |
2638 | if( gRSP.bMatrixIsUpdated ) |
2639 | { |
2640 | gRSPworldProject = gRSP.modelviewMtxs[gRSP.modelViewMtxTop] * gRSP.projectionMtxs[gRSP.projectionMtxTop]; |
2641 | gRSP.bMatrixIsUpdated = false; |
2642 | gRSP.bCombinedMatrixIsUpdated = true; |
2643 | } |
2644 | |
2645 | if( gRSP.bCombinedMatrixIsUpdated ) |
2646 | { |
2647 | if( options.enableHackForGames == HACK_REVERSE_XY_COOR ) |
2648 | { |
2649 | gRSPworldProject = gRSPworldProject * reverseXY; |
2650 | } |
2651 | if( options.enableHackForGames == HACK_REVERSE_Y_COOR ) |
2652 | { |
2653 | gRSPworldProject = gRSPworldProject * reverseY; |
2654 | } |
2655 | #if !defined(NO_ASM) |
2656 | if( status.isSSEEnabled ) |
2657 | { |
2658 | MatrixTranspose(&gRSPworldProjectTransported, &gRSPworldProject); |
2659 | } |
2660 | #endif |
2661 | gRSP.bCombinedMatrixIsUpdated = false; |
2662 | } |
2663 | |
2664 | //if( gRSP.bWorldMatrixIsUpdated || gRSP.bLightIsUpdated ) |
2665 | //{ |
2666 | // // Update lights with transported world matrix |
2667 | // for( unsigned int l=0; l<gRSPnumLights; l++) |
2668 | // { |
2669 | // Vec3TransformCoord(&gRSPlights[l].td, &gRSPlights[l].od, &gRSPmodelViewTopTranspose); |
2670 | // Vec3Normalize(&gRSPlights[l].td,&gRSPlights[l].td); |
2671 | // } |
2672 | |
2673 | // gRSP.bWorldMatrixIsUpdated = false; |
2674 | // gRSP.bLightIsUpdated = false; |
2675 | //} |
2676 | } |
2677 | |