2 Copyright (C) 2003 Rice1964
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.
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.
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.
23 #include "osal_preproc.h"
25 #include "DeviceBuilder.h"
26 #include "VertexShaderConstantDef.h"
30 extern FiddledVtx * g_pVtxBase;
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
40 #ifdef ENABLE_CLIP_TRI
42 inline void RSP_Vtx_Clipping(int i)
46 if( g_vecProjected[i].w > 0 )
51 if( g_vecProjected[i].x > 1 )
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;
58 if( g_vecProjected[i].x < -1 )
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;
65 if( g_vecProjected[i].y > 1 )
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;
72 if( g_vecProjected[i].y < -1 )
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;
79 //if( g_vecProjected[i].z > 1.0f )
81 // g_clipFlag2[i] |= Z_CLIP_MAX;
82 // g_clipFlag[i] |= Z_CLIP_MAX;
85 //if( gRSP.bNearClip && g_vecProjected[i].z < -1.0f )
87 // g_clipFlag2[i] |= Z_CLIP_MIN;
88 // g_clipFlag[i] |= Z_CLIP_MIN;
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;
106 inline void RSP_Vtx_Clipping(int i) {}
112 ALIGN(16,RSP_Options gRSP)
113 ALIGN(16,RDP_Options gRDP)
115 static ALIGN(16,XVECTOR4 g_normal)
116 //static int norms[3];
118 ALIGN(16,XVECTOR4 g_vtxNonTransformed[MAX_VERTS])
119 ALIGN(16,XVECTOR4 g_vecProjected[MAX_VERTS])
120 ALIGN(16,XVECTOR4 g_vtxTransformed[MAX_VERTS])
122 float g_vtxProjected5[1000][5];
123 float g_vtxProjected5Clipped[2000][5];
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];
133 EXTERNAL_VERTEX g_vtxForExternal[MAX_VERTS];
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;
145 float gRSPfFogDivider;
147 uint32 gRSPnumLights;
148 Light gRSPlights[16];
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)
156 N64Light gRSPn64lights[16];
159 void (*ProcessVertexData)(uint32 dwAddr, uint32 dwV0, uint32 dwNum)=NULL;
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);*/
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 \
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*/
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*/
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)*/
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)*/
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)*/
205 __asm fxch st(2) \ /* (x m00 + y m10 + z m20) (x m01 + ym11 + z m21) (x m02 + y m12 + z m 22) */
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 */
215 __asm fxch st(2) \ /* xx yy zz 1 x y z */
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 */
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 */
225 __asm fsqrt \ /* l 1 x y z */
227 __asm fdivp st(1),st(0) \ /* (1/l) x y z */
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 */
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 */
238 __asm mov dword ptr [vec + 0], 0 \
239 __asm mov dword ptr [vec + 4], 0 \
240 __asm mov dword ptr [vec + 8], 0 \
244 //#else // use C code in other cases, this is probably faster anyway
245 #elif defined(__ARM_NEON__0)
246 #define Vec3TransformNormal(vec, mtx) \
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" \
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" \
259 "vmul.f32 d0, d4, d4 \n\t" \
260 "vpadd.f32 d0, d0, d0 \n\t" \
261 "vmla.f32 d0, d5, d5 \n\t" \
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" \
272 "vmul.f32 q2, q2, d0[0] \n\t" \
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" \
281 #define Vec3TransformNormal(vec, m) \
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; }
292 #if !defined(__GNUC__) && !defined(NO_ASM)
293 __declspec( naked ) void __fastcall SSEVec3Transform(int i)
297 shl ecx,4; // ecx = i
299 movaps xmm1, DWORD PTR g_vtxNonTransformed [ecx]; // xmm1 as original vector
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
306 mulps xmm4, xmm1; // row 1
307 mulps xmm5, xmm1; // row 2
308 mulps xmm6, xmm1; // row 3
309 mulps xmm7, xmm1; // row 4
311 movhlps xmm0, xmm4; // xmm4 high to xmm0 low
312 movlhps xmm0, xmm5; // xmm5 low to xmm0 high
314 addps xmm4, xmm0; // result of add are in xmm4 low
315 addps xmm5, xmm0; // result of add are in xmm5 high
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
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
325 addps xmm4, xmm5; // results are in 1st and 2nd uint32
328 movhlps xmm0, xmm6; // xmm6 high to xmm0 low
329 movlhps xmm0, xmm7; // xmm7 low to xmm0 high
331 addps xmm6, xmm0; // result of add are in xmm6 low
332 addps xmm7, xmm0; // result of add are in xmm7 high
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
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
342 addps xmm6, xmm7; // results are in 1st and 2nd uint32
344 movlhps xmm4, xmm6; // final result is in xmm4
345 movaps DWORD PTR g_vtxTransformed [ecx], xmm4;
348 shufps xmm0,xmm0,0xff;
352 shufps xmm0,xmm0,0xe8;
355 movaps DWORD PTR g_vecProjected [ecx], xmm4;
363 __declspec( naked ) void __fastcall SSEVec3TransformDKR(XVECTOR4 &pOut, const XVECTOR4 &pV)
367 movaps xmm1, DWORD PTR [edx]; // xmm1 as original vector
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
374 mulps xmm4, xmm1; // row 1
375 mulps xmm5, xmm1; // row 2
376 mulps xmm6, xmm1; // row 3
377 mulps xmm7, xmm1; // row 4
379 movhlps xmm0, xmm4; // xmm4 high to xmm0 low
380 movlhps xmm0, xmm5; // xmm5 low to xmm0 high
382 addps xmm4, xmm0; // result of add are in xmm4 low
383 addps xmm5, xmm0; // result of add are in xmm5 high
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
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
393 addps xmm4, xmm5; // results are in 1st and 2nd uint32
396 movhlps xmm0, xmm6; // xmm6 high to xmm0 low
397 movlhps xmm0, xmm7; // xmm7 low to xmm0 high
399 addps xmm6, xmm0; // result of add are in xmm6 low
400 addps xmm7, xmm0; // result of add are in xmm7 high
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
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
410 addps xmm6, xmm7; // results are in 1st and 2nd uint32
412 movlhps xmm4, xmm6; // final result is in xmm4
413 movaps DWORD PTR [ecx], xmm4;
419 #elif defined(__GNUC__) && defined(__x86_64__) && !defined(NO_ASM)
420 void SSEVec3Transform(int i)
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"
464 : "r"(g_vtxNonTransformed), "r"(&gRSPworldProjectTransported.m[0][0]), "r"(g_vtxTransformed), "r"(g_vecProjected)
465 : "memory", "%xmm0", "%xmm1", "%xmm4", "%xmm5", "%xmm6", "%xmm7"
468 #elif !defined(NO_ASM) // 32-bit GCC assumed
469 void SSEVec3Transform(int i)
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"
512 : "r"(g_vtxNonTransformed), "r"(&gRSPworldProjectTransported.m[0][0]), "r"(g_vtxTransformed), "r"(g_vecProjected)
513 : "memory", "%xmm0", "%xmm1", "%xmm4", "%xmm5", "%xmm6", "%xmm7"
517 float real255 = 255.0f;
518 float real128 = 128.0f;
520 #if !defined(__GNUC__) && !defined(NO_ASM)
521 __declspec( naked ) void __fastcall SSEVec3TransformNormal()
525 mov DWORD PTR [g_normal][12], 0;
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
532 mulps xmm4, xmm1; // row 1
533 mulps xmm5, xmm1; // row 2
534 mulps xmm6, xmm1; // row 3
536 movhlps xmm0, xmm4; // xmm4 high to xmm0 low
537 movlhps xmm0, xmm5; // xmm5 low to xmm0 high
539 addps xmm4, xmm0; // result of add are in xmm4 low
540 addps xmm5, xmm0; // result of add are in xmm5 high
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
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,
549 addps xmm4, xmm5; // results are in 1st and 2nd uint32
552 mulps xmm1,xmm1; //square
554 shufps xmm7, xmm7,0x03;
557 movhlps xmm0, xmm6; // xmm6 high to xmm0 low
558 addps xmm6, xmm0; // result of add are in xmm6 low
561 shufps xmm0, xmm0, 0x03;
562 addss xmm0, xmm6; // result of add is at xmm0's 1st uint32
567 addss xmm7,xmm0; // xmm7 1st uint32 is the sum of squares
570 movaps DWORD PTR [g_normal], xmm4;
571 movss DWORD PTR [g_normal][12], xmm7;
580 movss DWORD PTR [g_normal][12], xmm7;
584 movaps DWORD PTR [g_normal], xmm4; // Normalized
585 mov DWORD PTR [g_normal][12], 0;
590 movss DWORD PTR [g_normal], xmm0;
591 movss DWORD PTR [g_normal][12], xmm0;
596 #elif defined(__GNUC__) && !defined(NO_ASM) // this code should compile for both 64-bit and 32-bit architectures
597 void SSEVec3TransformNormal(void)
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"
631 " movaps %%xmm4, (%0) \n"
632 " movss %%xmm7, 12(%0) \n"
634 " xorps %%xmm0, %%xmm0 \n"
635 " ucomiss %%xmm7, %%xmm0 \n"
637 " rsqrtss %%xmm7, %%xmm7 \n"
638 " shufps $0x00, %%xmm7, %%xmm7 \n"
640 " movss %%xmm7, 12(%0) \n"
642 " mulps %%xmm7, %%xmm4 \n"
643 " movaps %%xmm4, (%0) \n"
644 " movl $0, 12(%0) \n"
647 " movss %%xmm0, (%0) \n"
648 " movss %%xmm0, 12(%0) \n"
651 : "r"(&g_normal.x), "r"(&gRSPmodelViewTopTranspose.m[0][0])
652 : "memory", "cc", "%xmm0", "%xmm1", "%xmm4", "%xmm5", "%xmm6", "%xmm7"
657 void NormalizeNormalVec()
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
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
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}
682 : "d0", "d1", "d2", "d3", "d4", "d5", "memory"
687 float w = 1/sqrtf(g_normal.x*g_normal.x + g_normal.y*g_normal.y + g_normal.z*g_normal.z);
694 void InitRenderBase()
697 if( status.isSSEEnabled && !g_curRomInfo.bPrimaryDepthHack && options.enableHackForGames != HACK_FOR_NASCAR)
699 ProcessVertexData = ProcessVertexDataSSE;
702 #elif defined(__ARM_NEON__)
703 if( !g_curRomInfo.bPrimaryDepthHack && options.enableHackForGames != HACK_FOR_NASCAR && options.enableHackForGames != HACK_FOR_ZELDA_MM && !options.bWinFrameMode)
705 ProcessVertexData = ProcessVertexDataNEON;
710 ProcessVertexData = ProcessVertexDataNoSSE;
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;
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;
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;
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;
742 gRSP.DKRCMatrixIndex = gRSP.dwDKRVtxAddr = gRSP.dwDKRMatrixAddr = 0;
743 gRSP.DKRBillBoard = false;
745 gRSP.fTexScaleX = 1/32.0f;
746 gRSP.fTexScaleY = 1/32.0f;
747 gRSP.bTextureEnabled = FALSE;
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;
773 gRSP.DKRCMatrixIndex=0;
775 gRSP.DKRBillBoard = false;
777 gRSP.dwDKRMatrixAddr=0;
780 gRDP.geometryMode = 0;
783 gRDP.fillColor = 0xFFFFFFFF;
784 gRDP.originalFillColor =0;
787 gRSP.vertexMult = 10;
788 gRSP.bNearClip = false;
789 gRSP.bRejectVtx = false;
791 gRDP.texturesAreReloaded = false;
792 gRDP.textureIsChanged = false;
793 gRDP.colorsAreReloaded = false;
795 memset(&gRDP.otherMode,0,sizeof(RDP_OtherMode));
796 memset(&gRDP.tiles,0,sizeof(Tile)*8);
798 for( int i=0; i<MAX_VERTS; i++ )
801 g_vtxNonTransformed[i].w = 1;
804 memset(gRSPn64lights, 0, sizeof(N64Light)*16);
807 void SetFogMinMax(float fMin, float fMax, float fMul, float fOffset)
817 gRSPfFogMin = max(0,fMin/500-1);
818 gRSPfFogMax = fMax/500-1;
821 gRSPfFogDivider = 255/(gRSPfFogMax-gRSPfFogMin);
822 CRender::g_pRender->SetFogMinMax(fMin, fMax);
825 void InitVertexColors()
829 void InitVertexTextureConstants()
834 RenderTexture &tex0 = g_textures[gRSP.curTile];
835 //CTexture *surf = tex0.m_pCTexture;
836 Tile &tile0 = gRDP.tiles[gRSP.curTile];
838 scaleX = gRSP.fTexScaleX;
839 scaleY = gRSP.fTexScaleY;
841 gRSP.tex0scaleX = scaleX * tile0.fShiftScaleS/tex0.m_fTexWidth;
842 gRSP.tex0scaleY = scaleY * tile0.fShiftScaleT/tex0.m_fTexHeight;
844 gRSP.tex0OffsetX = tile0.fhilite_sl/tex0.m_fTexWidth;
845 gRSP.tex0OffsetY = tile0.fhilite_tl/tex0.m_fTexHeight;
847 if( CRender::g_pRender->IsTexel1Enable() )
849 RenderTexture &tex1 = g_textures[(gRSP.curTile+1)&7];
850 //CTexture *surf = tex1.m_pCTexture;
851 Tile &tile1 = gRDP.tiles[(gRSP.curTile+1)&7];
853 gRSP.tex1scaleX = scaleX * tile1.fShiftScaleS/tex1.m_fTexWidth;
854 gRSP.tex1scaleY = scaleY * tile1.fShiftScaleT/tex1.m_fTexHeight;
856 gRSP.tex1OffsetX = tile1.fhilite_sl/tex1.m_fTexWidth;
857 gRSP.tex1OffsetY = tile1.fhilite_tl/tex1.m_fTexHeight;
860 gRSP.texGenXRatio = tile0.fShiftScaleS;
861 gRSP.texGenYRatio = gRSP.fTexScaleX/gRSP.fTexScaleY*tex0.m_fTexWidth/tex0.m_fTexHeight*tile0.fShiftScaleT;
864 void TexGen(float &s, float &t)
866 if (gRDP.geometryMode & G_TEXTURE_GEN_LINEAR)
868 s = acosf(g_normal.x) / 3.14159f;
869 t = acosf(g_normal.y) / 3.14159f;
873 s = 0.5f * ( 1.0f + g_normal.x);
874 t = 0.5f * ( 1.0f - g_normal.y);
878 void ComputeLOD(bool openGL)
880 TLITVERTEX &v0 = g_vtxBuffer[0];
881 TLITVERTEX &v1 = g_vtxBuffer[1];
882 RenderTexture &tex0 = g_textures[gRSP.curTile];
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];
890 x = windowSetting.vpWidthW*x/windowSetting.fMultX/2;
891 y = windowSetting.vpHeightW*y/windowSetting.fMultY/2;
896 float x = (v0.x - v1.x)/ windowSetting.fMultX;
897 float y = (v0.y - v1.y)/ windowSetting.fMultY;
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;
906 dt = sqrtf((s0-s1)*(s0-s1)+(t0-t1)*(t0-t1));
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();
918 bool bHalfTxtScale=false;
919 extern uint32 lastSetTile;
921 void InitVertex(uint32 dwV, uint32 vtxIndex, bool bTexture, bool openGL)
923 VTX_DUMP(TRACE2("Init vertex (%d) to vtx buf[%d]:", dwV, vtxIndex));
925 TLITVERTEX &v = g_vtxBuffer[vtxIndex];
926 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));
929 g_vtxProjected5[vtxIndex][0] = g_vtxTransformed[dwV].x;
930 g_vtxProjected5[vtxIndex][1] = g_vtxTransformed[dwV].y;
931 g_vtxProjected5[vtxIndex][2] = g_vtxTransformed[dwV].z;
932 g_vtxProjected5[vtxIndex][3] = g_vtxTransformed[dwV].w;
933 g_vtxProjected5[vtxIndex][4] = g_vecProjected[dwV].z;
935 if( g_vtxTransformed[dwV].w < 0 )
936 g_vtxProjected5[vtxIndex][4] = 0;
938 g_vtxIndex[vtxIndex] = vtxIndex;
941 if( !openGL || options.bOGLVertexClipper == TRUE )
943 v.x = g_vecProjected[dwV].x*gRSP.vtxXMul+gRSP.vtxXAdd;
944 v.y = g_vecProjected[dwV].y*gRSP.vtxYMul+gRSP.vtxYAdd;
945 v.z = (g_vecProjected[dwV].z + 1.0f) * 0.5f; // DirectX minZ=0, maxZ=1
946 //v.z = g_vecProjected[dwV].z; // DirectX minZ=0, maxZ=1
947 v.rhw = g_vecProjected[dwV].w;
948 VTX_DUMP(TRACE4(" Proj : x=%f, y=%f, z=%f, rhw=%f", v.x,v.y,v.z,v.rhw));
950 if( gRSP.bProcessSpecularColor )
952 v.dcSpecular = CRender::g_pRender->PostProcessSpecularColor();
953 if( gRSP.bFogEnabled )
955 v.dcSpecular &= 0x00FFFFFF;
956 uint32 fogFct = 0xFF-(uint8)((g_fFogCoord[dwV]-gRSPfFogMin)*gRSPfFogDivider);
957 v.dcSpecular |= (fogFct<<24);
960 else if( gRSP.bFogEnabled )
962 uint32 fogFct = 0xFF-(uint8)((g_fFogCoord[dwV]-gRSPfFogMin)*gRSPfFogDivider);
963 v.dcSpecular = (fogFct<<24);
966 VTX_DUMP(TRACE2(" (U,V): %f, %f", g_fVtxTxtCoords[dwV].x,g_fVtxTxtCoords[dwV].y));
968 v.dcDiffuse = g_dwVtxDifColor[dwV];
969 if( gRDP.otherMode.key_en )
971 v.dcDiffuse &= 0x00FFFFFF;
972 v.dcDiffuse |= (gRDP.keyA<<24);
974 else if( gRDP.otherMode.aa_en && gRDP.otherMode.clr_on_cvg==0 )
976 v.dcDiffuse |= 0xFF000000;
979 if( gRSP.bProcessDiffuseColor )
981 v.dcDiffuse = CRender::g_pRender->PostProcessDiffuseColor(v.dcDiffuse);
983 if( options.bWinFrameMode )
985 v.dcDiffuse = g_dwVtxDifColor[dwV];
990 g_oglVtxColors[vtxIndex][0] = v.r;
991 g_oglVtxColors[vtxIndex][1] = v.g;
992 g_oglVtxColors[vtxIndex][2] = v.b;
993 g_oglVtxColors[vtxIndex][3] = v.a;
998 // If the vert is already lit, then there is no normal (and hence we can't generate tex coord)
999 // Only scale if not generated automatically
1000 if (gRSP.bTextureGen && gRSP.bLightingEnable)
1002 // Correction for texGen result
1004 RenderTexture &tex0 = g_textures[gRSP.curTile];
1005 u0 = g_fVtxTxtCoords[dwV].x * 32 * 1024 * gRSP.fTexScaleX / tex0.m_fTexWidth;
1006 v0 = g_fVtxTxtCoords[dwV].y * 32 * 1024 * gRSP.fTexScaleY / tex0.m_fTexHeight;
1007 u0 *= (gRDP.tiles[gRSP.curTile].fShiftScaleS);
1008 v0 *= (gRDP.tiles[gRSP.curTile].fShiftScaleT);
1010 if( CRender::g_pRender->IsTexel1Enable() )
1012 RenderTexture &tex1 = g_textures[(gRSP.curTile+1)&7];
1013 u1 = g_fVtxTxtCoords[dwV].x * 32 * 1024 * gRSP.fTexScaleX / tex1.m_fTexWidth;
1014 v1 = g_fVtxTxtCoords[dwV].y * 32 * 1024 * gRSP.fTexScaleY / tex1.m_fTexHeight;
1015 u1 *= gRDP.tiles[(gRSP.curTile+1)&7].fShiftScaleS;
1016 v1 *= gRDP.tiles[(gRSP.curTile+1)&7].fShiftScaleT;
1017 CRender::g_pRender->SetVertexTextureUVCoord(v, u0, v0, u1, v1);
1021 CRender::g_pRender->SetVertexTextureUVCoord(v, u0, v0);
1026 float tex0u = g_fVtxTxtCoords[dwV].x *gRSP.tex0scaleX - gRSP.tex0OffsetX ;
1027 float tex0v = g_fVtxTxtCoords[dwV].y *gRSP.tex0scaleY - gRSP.tex0OffsetY ;
1029 if( CRender::g_pRender->IsTexel1Enable() )
1031 float tex1u = g_fVtxTxtCoords[dwV].x *gRSP.tex1scaleX - gRSP.tex1OffsetX ;
1032 float tex1v = g_fVtxTxtCoords[dwV].y *gRSP.tex1scaleY - gRSP.tex1OffsetY ;
1034 CRender::g_pRender->SetVertexTextureUVCoord(v, tex0u, tex0v, tex1u, tex1v);
1035 VTX_DUMP(TRACE2(" (tex0): %f, %f", tex0u,tex0v));
1036 VTX_DUMP(TRACE2(" (tex1): %f, %f", tex1u,tex1v));
1040 CRender::g_pRender->SetVertexTextureUVCoord(v, tex0u, tex0v);
1041 VTX_DUMP(TRACE2(" (tex0): %f, %f", tex0u,tex0v));
1045 // Check for txt scale hack
1046 if( !bHalfTxtScale && g_curRomInfo.bTextureScaleHack &&
1047 (gRDP.tiles[lastSetTile].dwSize == TXT_SIZE_32b || gRDP.tiles[lastSetTile].dwSize == TXT_SIZE_4b ) )
1049 int width = ((gRDP.tiles[lastSetTile].sh-gRDP.tiles[lastSetTile].sl+1)<<1);
1050 int height = ((gRDP.tiles[lastSetTile].th-gRDP.tiles[lastSetTile].tl+1)<<1);
1051 if( g_fVtxTxtCoords[dwV].x*gRSP.fTexScaleX == width || g_fVtxTxtCoords[dwV].y*gRSP.fTexScaleY == height )
1058 if( g_curRomInfo.bEnableTxtLOD && vtxIndex == 1 && gRDP.otherMode.text_lod )
1060 if( CRender::g_pRender->IsTexel1Enable() && CRender::g_pRender->m_pColorCombiner->m_pDecodedMux->isUsed(MUX_LODFRAC) )
1070 VTX_DUMP(TRACE2(" DIF(%08X), SPE(%08X)", v.dcDiffuse, v.dcSpecular));
1071 VTX_DUMP(TRACE0(""));
1074 uint32 LightVert(XVECTOR4 & norm, int vidx)
1079 register float r = gRSP.fAmbientLightR;
1080 register float g = gRSP.fAmbientLightG;
1081 register float b = gRSP.fAmbientLightB;
1083 if( options.enableHackForGames != HACK_FOR_ZELDA_MM )
1085 for (register unsigned int l=0; l < gRSPnumLights; l++)
1087 fCosT = norm.x*gRSPlights[l].x + norm.y*gRSPlights[l].y + norm.z*gRSPlights[l].z;
1091 r += gRSPlights[l].fr * fCosT;
1092 g += gRSPlights[l].fg * fCosT;
1093 b += gRSPlights[l].fb * fCosT;
1100 bool transformed = false;
1102 for (register unsigned int l=0; l < gRSPnumLights; l++)
1104 if( gRSPlights[l].range == 0 )
1106 // Regular directional light
1107 fCosT = norm.x*gRSPlights[l].x + norm.y*gRSPlights[l].y + norm.z*gRSPlights[l].z;
1111 r += gRSPlights[l].fr * fCosT;
1112 g += gRSPlights[l].fg * fCosT;
1113 b += gRSPlights[l].fb * fCosT;
1116 else //if( (gRSPlights[l].col&0x00FFFFFF) != 0x00FFFFFF )
1121 Vec3Transform(&v, (XVECTOR3*)&g_vtxNonTransformed[vidx], &gRSPmodelViewTop); // Convert to w=1
1125 XVECTOR3 dir(gRSPlights[l].x - v.x, gRSPlights[l].y - v.y, gRSPlights[l].z - v.z);
1126 //XVECTOR3 dir(v.x-gRSPlights[l].x, v.y-gRSPlights[l].y, v.z-gRSPlights[l].z);
1127 float d2 = sqrtf(dir.x*dir.x+dir.y*dir.y+dir.z*dir.z);
1132 fCosT = norm.x*dir.x + norm.y*dir.y + norm.z*dir.z;
1136 //float f = d2/gRSPlights[l].range*50;
1137 float f = d2/15000*50;
1141 r += gRSPlights[l].fr * fCosT;
1142 g += gRSPlights[l].fg * fCosT;
1143 b += gRSPlights[l].fb * fCosT;
1149 if (r > 255) r = 255;
1150 if (g > 255) g = 255;
1151 if (b > 255) b = 255;
1152 return ((0xff000000)|(((uint32)r)<<16)|(((uint32)g)<<8)|((uint32)b));
1155 uint32 LightVertNew(XVECTOR4 & norm)
1160 register float r = gRSP.fAmbientLightR;
1161 register float g = gRSP.fAmbientLightG;
1162 register float b = gRSP.fAmbientLightB;
1165 for (register unsigned int l=0; l < gRSPnumLights; l++)
1167 fCosT = norm.x*gRSPlights[l].tx + norm.y*gRSPlights[l].ty + norm.z*gRSPlights[l].tz;
1171 r += gRSPlights[l].fr * fCosT;
1172 g += gRSPlights[l].fg * fCosT;
1173 b += gRSPlights[l].fb * fCosT;
1177 if (r > 255) r = 255;
1178 if (g > 255) g = 255;
1179 if (b > 255) b = 255;
1180 return ((0xff000000)|(((uint32)r)<<16)|(((uint32)g)<<8)|((uint32)b));
1188 #if !defined(__GNUC__) && !defined(NO_ASM)
1189 __declspec( naked ) uint32 __fastcall SSELightVert()
1193 movaps xmm3, DWORD PTR gRSP; // loading Ambient colors, xmm3 is the result color
1194 movaps xmm4, DWORD PTR [g_normal]; // xmm4 is the normal
1198 cmp ecx, DWORD PTR gRSPnumLights;
1202 movups xmm5, DWORD PTR gRSPlights[eax]; // Light Dir
1203 movups xmm1, DWORD PTR gRSPlights[0x14][eax]; // Light color
1204 mulps xmm5, xmm4; // Lightdir * normals
1208 shufps xmm5,xmm0,0x01;
1214 shufps xmm0,xmm0,0; // fcosT
1222 movss xmm0,DWORD PTR real255;
1226 // Without using a memory
1227 cvtss2si eax,xmm0; // move the 1st uint32 to eax
1230 shufps xmm0,xmm0,0E5h; // move the 2nd uint32 to the 1st uint32
1231 cvtss2si ecx,xmm0; // move the 1st uint32 to ecx
1234 shufps xmm0,xmm0,0E6h; // Move the 3rd uint32 to the 1st uint32
1241 #elif defined(__GNUC__) && defined(__x86_64__) && !defined(NO_ASM)
1242 uint32 SSELightVert(void)
1245 float f255 = 255.0, fZero = 0.0;
1247 asm volatile(" movaps %1, %%xmm3 \n" // xmm3 == gRSP.fAmbientLight{RGBA}
1248 " movaps %2, %%xmm4 \n" // xmm4 == g_normal.{xyz}
1249 " xor %%rcx, %%rcx \n"
1251 " cmpl %3, %%ecx \n"
1253 " mov %%rcx, %%rax \n"
1254 " imul $0x44, %%rax, %%rax \n"
1255 " movups (%4,%%rax,), %%xmm5 \n" // xmm5 == gRSPlights[l].{xyzr}
1256 " movups 20(%4,%%rax,), %%xmm1 \n" // xmm1 == gRSPlights[l].{frfgfbfa}
1257 " mulps %%xmm4, %%xmm5 \n"
1258 " movhlps %%xmm5, %%xmm0 \n"
1259 " addps %%xmm5, %%xmm0 \n"
1260 " shufps $0x01, %%xmm0, %%xmm5 \n"
1261 " addps %%xmm5, %%xmm0 \n"
1262 " comiss %6, %%xmm0 \n"
1264 " shufps $0x00, %%xmm0, %%xmm0 \n"
1265 " mulps %%xmm0, %%xmm1 \n"
1266 " addps %%xmm1, %%xmm3 \n"
1271 " movss %5, %%xmm0 \n"
1272 " shufps $0x00, %%xmm0, %%xmm0 \n"
1273 " minps %%xmm3, %%xmm0 \n"
1274 " cvtss2si %%xmm0, %%eax \n"
1275 " shll $0x10, %%eax \n"
1276 " orl $0xff000000, %%eax \n"
1277 " shufps $0xe5, %%xmm0, %%xmm0 \n"
1278 " cvtss2si %%xmm0, %%ecx \n"
1279 " shll $8, %%ecx \n"
1280 " orl %%ecx, %%eax \n"
1281 " shufps $0xe6, %%xmm0, %%xmm0 \n"
1282 " cvtss2si %%xmm0, %%ecx \n"
1283 " orl %%ecx, %%eax \n"
1285 : "m"(gRSP), "m"(g_normal), "m"(gRSPnumLights), "r"(gRSPlights), "m"(f255), "m"(fZero)
1286 : "%rcx", "memory", "cc", "%xmm0", "%xmm1", "%xmm3", "%xmm4", "%xmm5"
1290 #elif !defined(NO_ASM) // 32-bit GCC assumed
1291 uint32 SSELightVert(void)
1294 float f255 = 255.0, fZero = 0.0;
1296 asm volatile(" movaps %1, %%xmm3 \n"
1297 " movaps %2, %%xmm4 \n"
1298 " xor %%ecx, %%ecx \n"
1300 " cmpl %3, %%ecx \n"
1302 " mov %%ecx, %%eax \n"
1303 " imul $0x44, %%eax, %%eax \n"
1304 " movups (%4,%%eax,), %%xmm5 \n"
1305 " movups 20(%4,%%eax,), %%xmm1 \n"
1306 " mulps %%xmm4, %%xmm5 \n"
1307 " movhlps %%xmm5, %%xmm0 \n"
1308 " addps %%xmm5, %%xmm0 \n"
1309 " shufps $0x01, %%xmm0, %%xmm5 \n"
1310 " addps %%xmm5, %%xmm0 \n"
1311 " comiss %6, %%xmm0 \n"
1313 " shufps $0x00, %%xmm0, %%xmm0 \n"
1314 " mulps %%xmm0, %%xmm1 \n"
1315 " addps %%xmm1, %%xmm3 \n"
1320 " movss %5, %%xmm0 \n"
1321 " shufps $0x00, %%xmm0, %%xmm0 \n"
1322 " minps %%xmm3, %%xmm0 \n"
1323 " cvtss2si %%xmm0, %%eax \n"
1324 " shll $0x10, %%eax \n"
1325 " orl $0xff000000, %%eax \n"
1326 " shufps $0xe5, %%xmm0, %%xmm0 \n"
1327 " cvtss2si %%xmm0, %%ecx \n"
1328 " shll $8, %%ecx \n"
1329 " orl %%ecx, %%eax \n"
1330 " shufps $0xe6, %%xmm0, %%xmm0 \n"
1331 " cvtss2si %%xmm0, %%ecx \n"
1332 " orl %%ecx, %%eax \n"
1334 : "m"(gRSP), "m"(g_normal), "m"(gRSPnumLights), "r"(gRSPlights), "m"(f255), "m"(fZero)
1335 : "%rcx", "memory", "cc", "%xmm0", "%xmm1", "%xmm3", "%xmm4", "%xmm5"
1341 inline void ReplaceAlphaWithFogFactor(int i)
1343 if( gRDP.geometryMode & G_FOG )
1345 // Use fog factor to replace vertex alpha
1346 if( g_vecProjected[i].z > 1 )
1347 *(((uint8*)&(g_dwVtxDifColor[i]))+3) = 0xFF;
1348 if( g_vecProjected[i].z < 0 )
1349 *(((uint8*)&(g_dwVtxDifColor[i]))+3) = 0;
1351 *(((uint8*)&(g_dwVtxDifColor[i]))+3) = (uint8)(g_vecProjected[i].z*255);
1366 // Assumes dwAddr has already been checked!
1367 // Don't inline - it's too big with the transform macros
1369 #if !defined(NO_ASM)
1370 void ProcessVertexDataSSE(uint32 dwAddr, uint32 dwV0, uint32 dwNum)
1372 UpdateCombinedMatrix();
1374 // This function is called upon SPvertex
1375 // - do vertex matrix transform
1376 // - do vertex lighting
1377 // - do texture cooridinate transform if needed
1378 // - calculate normal vector
1380 // Output: - g_vecProjected[i] -> transformed vertex x,y,z
1381 // - g_vecProjected[i].w -> saved vertex 1/w
1382 // - g_dwVtxFlags[i] -> flags
1383 // - g_dwVtxDifColor[i] -> vertex color
1384 // - g_fVtxTxtCoords[i] -> vertex texture cooridinates
1386 FiddledVtx * pVtxBase = (FiddledVtx*)(g_pRDRAMu8 + dwAddr);
1387 g_pVtxBase = pVtxBase;
1389 for (uint32 i = dwV0; i < dwV0 + dwNum; i++)
1391 SP_Timing(RSP_GBI0_Vtx);
1393 FiddledVtx & vert = pVtxBase[i - dwV0];
1395 g_vtxNonTransformed[i].x = (float)vert.x;
1396 g_vtxNonTransformed[i].y = (float)vert.y;
1397 g_vtxNonTransformed[i].z = (float)vert.z;
1399 SSEVec3Transform(i);
1401 if( gRSP.bFogEnabled )
1403 g_fFogCoord[i] = g_vecProjected[i].z;
1404 if( g_vecProjected[i].w < 0 || g_vecProjected[i].z < 0 || g_fFogCoord[i] < gRSPfFogMin )
1405 g_fFogCoord[i] = gRSPfFogMin;
1408 ReplaceAlphaWithFogFactor(i);
1413 uint32 *dat = (uint32*)(&vert);
1414 DebuggerAppendMsg("vtx %d: %08X %08X %08X %08X", i, dat[0],dat[1],dat[2],dat[3]);
1415 DebuggerAppendMsg(" : %f, %f, %f, %f",
1416 g_vtxTransformed[i].x,g_vtxTransformed[i].y,g_vtxTransformed[i].z,g_vtxTransformed[i].w);
1417 DebuggerAppendMsg(" : %f, %f, %f, %f",
1418 g_vecProjected[i].x,g_vecProjected[i].y,g_vecProjected[i].z,g_vecProjected[i].w);
1421 RSP_Vtx_Clipping(i);
1423 if( gRSP.bLightingEnable )
1425 g_normal.x = (float)vert.norma.nx;
1426 g_normal.y = (float)vert.norma.ny;
1427 g_normal.z = (float)vert.norma.nz;
1429 SSEVec3TransformNormal();
1430 if( options.enableHackForGames != HACK_FOR_ZELDA_MM )
1431 g_dwVtxDifColor[i] = SSELightVert();
1433 g_dwVtxDifColor[i] = LightVert(g_normal, i);
1434 *(((uint8*)&(g_dwVtxDifColor[i]))+3) = vert.rgba.a; // still use alpha from the vertex
1438 if( (gRDP.geometryMode & G_SHADE) == 0 && gRSP.ucode < 5 ) //Shade is disabled
1441 g_dwVtxDifColor[i] = gRDP.primitiveColor;
1445 register IColor &color = *(IColor*)&g_dwVtxDifColor[i];
1446 color.b = vert.rgba.r;
1447 color.g = vert.rgba.g;
1448 color.r = vert.rgba.b;
1449 color.a = vert.rgba.a;
1453 if( options.bWinFrameMode )
1455 g_dwVtxDifColor[i] = COLOR_RGBA(vert.rgba.r, vert.rgba.g, vert.rgba.b, vert.rgba.a);
1458 // Update texture coords n.b. need to divide tu/tv by bogus scale on addition to buffer
1460 // If the vert is already lit, then there is no normal (and hence we
1461 // can't generate tex coord)
1462 if (gRSP.bTextureGen && gRSP.bLightingEnable )
1464 TexGen(g_fVtxTxtCoords[i].x, g_fVtxTxtCoords[i].y);
1468 g_fVtxTxtCoords[i].x = (float)vert.tu;
1469 g_fVtxTxtCoords[i].y = (float)vert.tv;
1473 VTX_DUMP(TRACE2("Setting Vertexes: %d - %d\n", dwV0, dwV0+dwNum-1));
1474 DEBUGGER_PAUSE_AND_DUMP(NEXT_VERTEX_CMD,{TRACE0("Paused at Vertex Cmd");});
1478 void ProcessVertexDataNoSSE(uint32 dwAddr, uint32 dwV0, uint32 dwNum)
1481 UpdateCombinedMatrix();
1483 // This function is called upon SPvertex
1484 // - do vertex matrix transform
1485 // - do vertex lighting
1486 // - do texture cooridinate transform if needed
1487 // - calculate normal vector
1489 // Output: - g_vecProjected[i] -> transformed vertex x,y,z
1490 // - g_vecProjected[i].w -> saved vertex 1/w
1491 // - g_dwVtxFlags[i] -> flags
1492 // - g_dwVtxDifColor[i] -> vertex color
1493 // - g_fVtxTxtCoords[i] -> vertex texture cooridinates
1495 FiddledVtx * pVtxBase = (FiddledVtx*)(g_pRDRAMu8 + dwAddr);
1496 g_pVtxBase = pVtxBase;
1498 for (uint32 i = dwV0; i < dwV0 + dwNum; i++)
1500 SP_Timing(RSP_GBI0_Vtx);
1502 FiddledVtx & vert = pVtxBase[i - dwV0];
1504 g_vtxNonTransformed[i].x = (float)vert.x;
1505 g_vtxNonTransformed[i].y = (float)vert.y;
1506 g_vtxNonTransformed[i].z = (float)vert.z;
1508 Vec3Transform(&g_vtxTransformed[i], (XVECTOR3*)&g_vtxNonTransformed[i], &gRSPworldProject); // Convert to w=1
1510 g_vecProjected[i].w = 1.0f / g_vtxTransformed[i].w;
1511 g_vecProjected[i].x = g_vtxTransformed[i].x * g_vecProjected[i].w;
1512 g_vecProjected[i].y = g_vtxTransformed[i].y * g_vecProjected[i].w;
1513 if ((g_curRomInfo.bPrimaryDepthHack || options.enableHackForGames == HACK_FOR_NASCAR ) && gRDP.otherMode.depth_source )
1515 g_vecProjected[i].z = gRDP.fPrimitiveDepth;
1516 g_vtxTransformed[i].z = gRDP.fPrimitiveDepth*g_vtxTransformed[i].w;
1520 g_vecProjected[i].z = g_vtxTransformed[i].z * g_vecProjected[i].w;
1523 if( gRSP.bFogEnabled )
1525 g_fFogCoord[i] = g_vecProjected[i].z;
1526 if( g_vecProjected[i].w < 0 || g_vecProjected[i].z < 0 || g_fFogCoord[i] < gRSPfFogMin )
1527 g_fFogCoord[i] = gRSPfFogMin;
1532 uint32 *dat = (uint32*)(&vert);
1533 DebuggerAppendMsg("vtx %d: %08X %08X %08X %08X", i, dat[0],dat[1],dat[2],dat[3]);
1534 DebuggerAppendMsg(" : %f, %f, %f, %f",
1535 g_vtxTransformed[i].x,g_vtxTransformed[i].y,g_vtxTransformed[i].z,g_vtxTransformed[i].w);
1536 DebuggerAppendMsg(" : %f, %f, %f, %f",
1537 g_vecProjected[i].x,g_vecProjected[i].y,g_vecProjected[i].z,g_vecProjected[i].w);
1540 RSP_Vtx_Clipping(i);
1542 if( gRSP.bLightingEnable )
1544 g_normal.x = (float)vert.norma.nx;
1545 g_normal.y = (float)vert.norma.ny;
1546 g_normal.z = (float)vert.norma.nz;
1548 Vec3TransformNormal(g_normal, gRSPmodelViewTop);
1549 g_dwVtxDifColor[i] = LightVert(g_normal, i);
1550 *(((uint8*)&(g_dwVtxDifColor[i]))+3) = vert.rgba.a; // still use alpha from the vertex
1554 if( (gRDP.geometryMode & G_SHADE) == 0 && gRSP.ucode < 5 ) //Shade is disabled
1557 g_dwVtxDifColor[i] = gRDP.primitiveColor;
1561 register IColor &color = *(IColor*)&g_dwVtxDifColor[i];
1562 color.b = vert.rgba.r;
1563 color.g = vert.rgba.g;
1564 color.r = vert.rgba.b;
1565 color.a = vert.rgba.a;
1569 if( options.bWinFrameMode )
1571 g_dwVtxDifColor[i] = COLOR_RGBA(vert.rgba.r, vert.rgba.g, vert.rgba.b, vert.rgba.a);
1574 ReplaceAlphaWithFogFactor(i);
1576 // Update texture coords n.b. need to divide tu/tv by bogus scale on addition to buffer
1578 // If the vert is already lit, then there is no normal (and hence we
1579 // can't generate tex coord)
1580 if (gRSP.bTextureGen && gRSP.bLightingEnable )
1582 TexGen(g_fVtxTxtCoords[i].x, g_fVtxTxtCoords[i].y);
1586 g_fVtxTxtCoords[i].x = (float)vert.tu;
1587 g_fVtxTxtCoords[i].y = (float)vert.tv;
1591 VTX_DUMP(TRACE2("Setting Vertexes: %d - %d\n", dwV0, dwV0+dwNum-1));
1592 DEBUGGER_PAUSE_AND_DUMP(NEXT_VERTEX_CMD,{TRACE0("Paused at Vertex Cmd");});
1595 extern "C" void pv_neon(XVECTOR4 *g_vtxTransformed, XVECTOR4 *g_vecProjected,
1596 uint32 *g_dwVtxDifColor, VECTOR2 *g_fVtxTxtCoords,
1597 float *g_fFogCoord, uint32 *g_clipFlag2,
1598 uint32 dwNum, const FiddledVtx *vtx,
1599 const Light *gRSPlights, const float *fRSPAmbientLightRGBA,
1600 const XMATRIX *gRSPworldProject, const XMATRIX *gRSPmodelViewTop,
1601 uint32 gRSPnumLights, float gRSPfFogMin);
1603 void ProcessVertexDataNEON(uint32 dwAddr, uint32 dwV0, uint32 dwNum)
1605 if (gRSP.bTextureGen && gRSP.bLightingEnable) {
1606 ProcessVertexDataNoSSE(dwAddr, dwV0,dwNum);
1611 // - g_clipFlag is not used at all
1612 // - g_vtxNonTransformed is not used after ProcessVertexData*() returns
1613 // - g_normal - same
1615 #define PV_NEON_ENABLE_LIGHT (1 << 0)
1616 #define PV_NEON_ENABLE_SHADE (1 << 1)
1617 #define PV_NEON_ENABLE_FOG (1 << 2)
1618 #define PV_NEON_FOG_ALPHA (1 << 3)
1621 if ( gRSP.bLightingEnable )
1622 neon_state |= PV_NEON_ENABLE_LIGHT;
1623 if ( (gRDP.geometryMode & G_SHADE) || gRSP.ucode >= 5 )
1624 neon_state |= PV_NEON_ENABLE_SHADE;
1625 if ( gRSP.bFogEnabled )
1626 neon_state |= PV_NEON_ENABLE_FOG;
1627 if ( gRDP.geometryMode & G_FOG )
1628 neon_state |= PV_NEON_FOG_ALPHA;
1632 UpdateCombinedMatrix();
1634 // This function is called upon SPvertex
1635 // - do vertex matrix transform
1636 // - do vertex lighting
1637 // - do texture cooridinate transform if needed
1638 // - calculate normal vector
1640 // Output: - g_vecProjected[i] -> transformed vertex x,y,z
1641 // - g_vecProjected[i].w -> saved vertex 1/w
1642 // - g_vtxTransformed[i]
1643 // - g_dwVtxDifColor[i] -> vertex color
1644 // - g_fVtxTxtCoords[i] -> vertex texture cooridinates
1648 const FiddledVtx * pVtxBase = (const FiddledVtx*)(g_pRDRAMu8 + dwAddr);
1649 g_pVtxBase = (FiddledVtx *)pVtxBase;
1651 // SP_Timing(RSP_GBI0_Vtx);
1652 status.SPCycleCount += Timing_RSP_GBI0_Vtx * dwNum;
1654 if (!(neon_state & (PV_NEON_ENABLE_LIGHT | PV_NEON_ENABLE_SHADE))) {
1655 for (i = dwV0; i < dwV0 + dwNum; i++)
1656 g_dwVtxDifColor[i] = gRDP.primitiveColor; // FLAT shade
1659 for (i = dwV0; i < dwV0 + dwNum; i++)
1661 const FiddledVtx & vert = pVtxBase[i - dwV0];
1662 XVECTOR3 vtx_raw; // was g_vtxNonTransformed
1664 vtx_raw.x = (float)vert.x;
1665 vtx_raw.y = (float)vert.y;
1666 vtx_raw.z = (float)vert.z;
1668 Vec3Transform(&g_vtxTransformed[i], &vtx_raw, &gRSPworldProject); // Convert to w=1
1670 g_vecProjected[i].w = 1.0f / g_vtxTransformed[i].w;
1671 g_vecProjected[i].x = g_vtxTransformed[i].x * g_vecProjected[i].w;
1672 g_vecProjected[i].y = g_vtxTransformed[i].y * g_vecProjected[i].w;
1673 g_vecProjected[i].z = g_vtxTransformed[i].z * g_vecProjected[i].w;
1675 if( neon_state & PV_NEON_ENABLE_FOG )
1677 g_fFogCoord[i] = g_vecProjected[i].z;
1678 if( g_vecProjected[i].w < 0 || g_vecProjected[i].z < 0 || g_fFogCoord[i] < gRSPfFogMin )
1679 g_fFogCoord[i] = gRSPfFogMin;
1682 // RSP_Vtx_Clipping(i);
1684 if( g_vecProjected[i].w > 0 )
1686 if( g_vecProjected[i].x > 1 ) g_clipFlag2[i] |= X_CLIP_MAX;
1687 if( g_vecProjected[i].x < -1 ) g_clipFlag2[i] |= X_CLIP_MIN;
1688 if( g_vecProjected[i].y > 1 ) g_clipFlag2[i] |= Y_CLIP_MAX;
1689 if( g_vecProjected[i].y < -1 ) g_clipFlag2[i] |= Y_CLIP_MIN;
1692 if( neon_state & PV_NEON_ENABLE_LIGHT )
1694 XVECTOR3 normal; // was g_normal
1697 normal.x = (float)vert.norma.nx;
1698 normal.y = (float)vert.norma.ny;
1699 normal.z = (float)vert.norma.nz;
1701 Vec3TransformNormal(normal, gRSPmodelViewTop);
1703 r = gRSP.fAmbientLightR;
1704 g = gRSP.fAmbientLightG;
1705 b = gRSP.fAmbientLightB;
1707 for (unsigned int l=0; l < gRSPnumLights; l++)
1709 float fCosT = normal.x * gRSPlights[l].x + normal.y * gRSPlights[l].y + normal.z * gRSPlights[l].z;
1713 r += gRSPlights[l].fr * fCosT;
1714 g += gRSPlights[l].fg * fCosT;
1715 b += gRSPlights[l].fb * fCosT;
1718 if (r > 255) r = 255;
1719 if (g > 255) g = 255;
1720 if (b > 255) b = 255;
1721 g_dwVtxDifColor[i] = ((vert.rgba.a<<24)|(((uint32)r)<<16)|(((uint32)g)<<8)|((uint32)b));
1723 else if( neon_state & PV_NEON_ENABLE_SHADE )
1725 IColor &color = *(IColor*)&g_dwVtxDifColor[i];
1726 color.b = vert.rgba.r;
1727 color.g = vert.rgba.g;
1728 color.r = vert.rgba.b;
1729 color.a = vert.rgba.a;
1732 // ReplaceAlphaWithFogFactor(i);
1733 if( neon_state & PV_NEON_FOG_ALPHA )
1735 // Use fog factor to replace vertex alpha
1736 if( g_vecProjected[i].z > 1 )
1737 *(((uint8*)&(g_dwVtxDifColor[i]))+3) = 0xFF;
1738 if( g_vecProjected[i].z < 0 )
1739 *(((uint8*)&(g_dwVtxDifColor[i]))+3) = 0;
1741 *(((uint8*)&(g_dwVtxDifColor[i]))+3) = (uint8)(g_vecProjected[i].z*255);
1744 g_fVtxTxtCoords[i].x = (float)vert.tu;
1745 g_fVtxTxtCoords[i].y = (float)vert.tv;
1749 bool PrepareTriangle(uint32 dwV0, uint32 dwV1, uint32 dwV2)
1751 if( status.isVertexShaderEnabled || status.bUseHW_T_L )
1753 g_vtxIndex[gRSP.numVertices++] = dwV0;
1754 g_vtxIndex[gRSP.numVertices++] = dwV1;
1755 g_vtxIndex[gRSP.numVertices++] = dwV2;
1756 status.dwNumTrisRendered++;
1757 gRSP.maxVertexID = max(gRSP.maxVertexID,max(dwV0,max(dwV1,dwV2)));
1761 SP_Timing(SP_Each_Triangle);
1763 bool textureFlag = (CRender::g_pRender->IsTextureEnabled() || gRSP.ucode == 6 );
1764 bool openGL = CDeviceBuilder::m_deviceGeneralType == OGL_DEVICE;
1766 InitVertex(dwV0, gRSP.numVertices, textureFlag, openGL);
1767 InitVertex(dwV1, gRSP.numVertices+1, textureFlag, openGL);
1768 InitVertex(dwV2, gRSP.numVertices+2, textureFlag, openGL);
1770 gRSP.numVertices += 3;
1771 status.dwNumTrisRendered++;
1779 // Returns TRUE if it thinks the triangle is visible
1780 // Returns FALSE if it is clipped
1781 bool IsTriangleVisible(uint32 dwV0, uint32 dwV1, uint32 dwV2)
1783 //return true; //fix me
1785 if( status.isVertexShaderEnabled || status.bUseHW_T_L ) return true; // We won't have access to transformed vertex data
1787 DEBUGGER_ONLY_IF( (!debuggerEnableTestTris || !debuggerEnableCullFace), {return TRUE;});
1790 // Check vertices are valid!
1791 if (dwV0 >= MAX_VERTS || dwV1 >= MAX_VERTS || dwV2 >= MAX_VERTS)
1795 // Here we AND all the flags. If any of the bits is set for all
1796 // 3 vertices, it means that all three x, y or z lie outside of
1797 // the current viewing volume.
1798 // Currently disabled - still seems a bit dodgy
1799 if ((gRSP.bCullFront || gRSP.bCullBack) && gRDP.otherMode.zmode != 3)
1801 XVECTOR4 & v0 = g_vecProjected[dwV0];
1802 XVECTOR4 & v1 = g_vecProjected[dwV1];
1803 XVECTOR4 & v2 = g_vecProjected[dwV2];
1805 // Only try to clip if the tri is onscreen. For some reason, this
1806 // method doesnt' work well when the z value is outside of screenspace
1807 //if (v0.z < 1 && v1.z < 1 && v2.z < 1)
1809 float V1 = v2.x - v0.x;
1810 float V2 = v2.y - v0.y;
1812 float W1 = v2.x - v1.x;
1813 float W2 = v2.y - v1.y;
1815 float fDirection = (V1 * W2) - (V2 * W1);
1816 fDirection = fDirection * v1.w * v2.w * v0.w;
1817 //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;
1819 if (fDirection < 0 && gRSP.bCullBack)
1821 status.dwNumTrisClipped++;
1824 else if (fDirection > 0 && gRSP.bCullFront)
1826 status.dwNumTrisClipped++;
1832 #ifdef ENABLE_CLIP_TRI
1833 //if( gRSP.bRejectVtx && (g_clipFlag[dwV0]|g_clipFlag[dwV1]|g_clipFlag[dwV2]) )
1835 if( g_clipFlag2[dwV0]&g_clipFlag2[dwV1]&g_clipFlag2[dwV2] )
1837 //DebuggerAppendMsg("Clipped");
1846 void SetPrimitiveColor(uint32 dwCol, uint32 LODMin, uint32 LODFrac)
1848 gRDP.colorsAreReloaded = true;
1849 gRDP.primitiveColor = dwCol;
1850 gRDP.primLODMin = LODMin;
1851 gRDP.primLODFrac = LODFrac;
1852 if( gRDP.primLODFrac < gRDP.primLODMin )
1854 gRDP.primLODFrac = gRDP.primLODMin;
1857 gRDP.fvPrimitiveColor[0] = ((dwCol>>16)&0xFF)/255.0f; //r
1858 gRDP.fvPrimitiveColor[1] = ((dwCol>>8)&0xFF)/255.0f; //g
1859 gRDP.fvPrimitiveColor[2] = ((dwCol)&0xFF)/255.0f; //b
1860 gRDP.fvPrimitiveColor[3] = ((dwCol>>24)&0xFF)/255.0f; //a
1863 void SetPrimitiveDepth(uint32 z, uint32 dwDZ)
1865 gRDP.primitiveDepth = z & 0x7FFF;
1866 gRDP.fPrimitiveDepth = (float)(gRDP.primitiveDepth)/(float)0x8000;
1868 //gRDP.fPrimitiveDepth = gRDP.fPrimitiveDepth*2-1;
1870 z=0xFFFF -> 1 the farest
1871 z=0 -> -1 the nearest
1877 if( (pauseAtNext && (eventToPause == NEXT_VERTEX_CMD || eventToPause == NEXT_FLUSH_TRI )) )//&& logTriangles )
1879 DebuggerAppendMsg("Set prim Depth: %f, (%08X, %08X)", gRDP.fPrimitiveDepth, z, dwDZ);
1884 void SetVertexXYZ(uint32 vertex, float x, float y, float z)
1886 g_vecProjected[vertex].x = x;
1887 g_vecProjected[vertex].y = y;
1888 g_vecProjected[vertex].z = z;
1890 g_vtxTransformed[vertex].x = x*g_vtxTransformed[vertex].w;
1891 g_vtxTransformed[vertex].y = y*g_vtxTransformed[vertex].w;
1892 g_vtxTransformed[vertex].z = z*g_vtxTransformed[vertex].w;
1895 void ModifyVertexInfo(uint32 where, uint32 vertex, uint32 val)
1899 case RSP_MV_WORD_OFFSET_POINT_RGBA: // Modify RGBA
1901 uint32 r = (val>>24)&0xFF;
1902 uint32 g = (val>>16)&0xFF;
1903 uint32 b = (val>>8)&0xFF;
1904 uint32 a = val&0xFF;
1905 g_dwVtxDifColor[vertex] = COLOR_RGBA(r, g, b, a);
1906 LOG_UCODE("Modify vert %d color, 0x%08x", vertex, g_dwVtxDifColor[vertex]);
1909 case RSP_MV_WORD_OFFSET_POINT_XYSCREEN: // Modify X,Y
1911 uint16 nX = (uint16)(val>>16);
1912 short x = *((short*)&nX);
1915 uint16 nY = (uint16)(val&0xFFFF);
1916 short y = *((short*)&nY);
1919 // Should do viewport transform.
1922 x -= windowSetting.uViWidth/2;
1923 y = windowSetting.uViHeight/2-y;
1925 if( options.bEnableHacks && ((*g_GraphicsInfo.VI_X_SCALE_REG)&0xF) != 0 )
1928 // I don't know why Tarzan is different
1929 SetVertexXYZ(vertex, x/windowSetting.fViWidth, y/windowSetting.fViHeight, g_vecProjected[vertex].z);
1933 // Toy Story 2 and other games
1934 SetVertexXYZ(vertex, x*2/windowSetting.fViWidth, y*2/windowSetting.fViHeight, g_vecProjected[vertex].z);
1937 LOG_UCODE("Modify vert %d: x=%d, y=%d", vertex, x, y);
1938 VTX_DUMP(TRACE3("Modify vert %d: (%d,%d)", vertex, x, y));
1941 case RSP_MV_WORD_OFFSET_POINT_ZSCREEN: // Modify C
1945 SetVertexXYZ(vertex, g_vecProjected[vertex].x, g_vecProjected[vertex].y, (((float)z/0x03FF)+0.5f)/2.0f );
1946 LOG_UCODE("Modify vert %d: z=%d", vertex, z);
1947 VTX_DUMP(TRACE2("Modify vert %d: z=%d", vertex, z));
1950 case RSP_MV_WORD_OFFSET_POINT_ST: // Texture
1952 short tu = short(val>>16);
1953 short tv = short(val & 0xFFFF);
1954 float ftu = tu / 32.0f;
1955 float ftv = tv / 32.0f;
1956 LOG_UCODE(" Setting vertex %d tu/tv to %f, %f", vertex, (float)tu, (float)tv);
1957 CRender::g_pRender->SetVtxTextureCoord(vertex, ftu/gRSP.fTexScaleX, ftv/gRSP.fTexScaleY);
1961 DEBUGGER_PAUSE_AND_DUMP(NEXT_VERTEX_CMD,{TRACE0("Paused at ModVertex Cmd");});
1964 void ProcessVertexDataDKR(uint32 dwAddr, uint32 dwV0, uint32 dwNum)
1966 UpdateCombinedMatrix();
1968 long long pVtxBase = (long long) (g_pRDRAMu8 + dwAddr);
1969 g_pVtxBase = (FiddledVtx*)pVtxBase;
1971 Matrix &matWorldProject = gRSP.DKRMatrixes[gRSP.DKRCMatrixIndex];
1976 if ((!gRSP.DKRBillBoard) || (gRSP.DKRCMatrixIndex != 2) )
1981 if( addbase && gRSP.DKRVtxCount == 0 && dwNum > 1 )
1986 LOG_UCODE(" ProcessVertexDataDKR, CMatrix = %d, Add base=%s", gRSP.DKRCMatrixIndex, gRSP.DKRBillBoard?"true":"false");
1987 VTX_DUMP(TRACE2("DKR Setting Vertexes\nCMatrix = %d, Add base=%s", gRSP.DKRCMatrixIndex, gRSP.DKRBillBoard?"true":"false"));
1990 uint32 end = dwV0 + dwNum;
1991 for (uint32 i = dwV0; i < end; i++)
1995 g_vtxNonTransformed[i].x = (float)*(short*)((pVtxBase+nOff + 0) ^ 2);
1996 g_vtxNonTransformed[i].y = (float)*(short*)((pVtxBase+nOff + 2) ^ 2);
1997 g_vtxNonTransformed[i].z = (float)*(short*)((pVtxBase+nOff + 4) ^ 2);
1999 //if( status.isSSEEnabled )
2000 // SSEVec3TransformDKR(g_vtxTransformed[i], g_vtxNonTransformed[i]);
2002 Vec3Transform(&g_vtxTransformed[i], (XVECTOR3*)&g_vtxNonTransformed[i], &matWorldProject); // Convert to w=1
2004 if( gRSP.DKRVtxCount == 0 && dwNum==1 )
2006 gRSP.DKRBaseVec.x = g_vtxTransformed[i].x;
2007 gRSP.DKRBaseVec.y = g_vtxTransformed[i].y;
2008 gRSP.DKRBaseVec.z = g_vtxTransformed[i].z;
2009 gRSP.DKRBaseVec.w = g_vtxTransformed[i].w;
2013 g_vtxTransformed[i].x += gRSP.DKRBaseVec.x;
2014 g_vtxTransformed[i].y += gRSP.DKRBaseVec.y;
2015 g_vtxTransformed[i].z += gRSP.DKRBaseVec.z;
2016 g_vtxTransformed[i].w = gRSP.DKRBaseVec.w;
2019 g_vecProjected[i].w = 1.0f / g_vtxTransformed[i].w;
2020 g_vecProjected[i].x = g_vtxTransformed[i].x * g_vecProjected[i].w;
2021 g_vecProjected[i].y = g_vtxTransformed[i].y * g_vecProjected[i].w;
2022 g_vecProjected[i].z = g_vtxTransformed[i].z * g_vecProjected[i].w;
2026 VTX_DUMP(TRACE5("vtx %d: %f, %f, %f, %f", i,
2027 g_vtxTransformed[i].x,g_vtxTransformed[i].y,g_vtxTransformed[i].z,g_vtxTransformed[i].w));
2029 if( gRSP.bFogEnabled )
2031 g_fFogCoord[i] = g_vecProjected[i].z;
2032 if( g_vecProjected[i].w < 0 || g_vecProjected[i].z < 0 || g_fFogCoord[i] < gRSPfFogMin )
2033 g_fFogCoord[i] = gRSPfFogMin;
2036 RSP_Vtx_Clipping(i);
2038 short wA = *(short*)((pVtxBase+nOff + 6) ^ 2);
2039 short wB = *(short*)((pVtxBase+nOff + 8) ^ 2);
2041 s8 r = (s8)(wA >> 8);
2043 s8 b = (s8)(wB >> 8);
2046 if (gRSP.bLightingEnable)
2048 g_normal.x = (char)r; //norma.nx;
2049 g_normal.y = (char)g; //norma.ny;
2050 g_normal.z = (char)b; //norma.nz;
2052 Vec3TransformNormal(g_normal, matWorldProject)
2053 #if !defined(NO_ASM)
2054 if( status.isSSEEnabled )
2055 g_dwVtxDifColor[i] = SSELightVert();
2058 g_dwVtxDifColor[i] = LightVert(g_normal, i);
2068 // Assign true vert colour after lighting/fogging
2069 g_dwVtxDifColor[i] = COLOR_RGBA(nR, nG, nB, nA);
2072 ReplaceAlphaWithFogFactor(i);
2074 g_fVtxTxtCoords[i].x = g_fVtxTxtCoords[i].y = 1;
2080 DEBUGGER_PAUSE_AND_DUMP(NEXT_VERTEX_CMD,{DebuggerAppendMsg("Paused at DKR Vertex Cmd, v0=%d, vn=%d, addr=%08X", dwV0, dwNum, dwAddr);});
2084 extern uint32 dwPDCIAddr;
2085 void ProcessVertexDataPD(uint32 dwAddr, uint32 dwV0, uint32 dwNum)
2087 UpdateCombinedMatrix();
2089 N64VtxPD * pVtxBase = (N64VtxPD*)(g_pRDRAMu8 + dwAddr);
2090 g_pVtxBase = (FiddledVtx*)pVtxBase; // Fix me
2092 for (uint32 i = dwV0; i < dwV0 + dwNum; i++)
2094 N64VtxPD &vert = pVtxBase[i - dwV0];
2096 g_vtxNonTransformed[i].x = (float)vert.x;
2097 g_vtxNonTransformed[i].y = (float)vert.y;
2098 g_vtxNonTransformed[i].z = (float)vert.z;
2100 #if !defined(NO_ASM)
2101 if( status.isSSEEnabled )
2102 SSEVec3Transform(i);
2106 Vec3Transform(&g_vtxTransformed[i], (XVECTOR3*)&g_vtxNonTransformed[i], &gRSPworldProject); // Convert to w=1
2107 g_vecProjected[i].w = 1.0f / g_vtxTransformed[i].w;
2108 g_vecProjected[i].x = g_vtxTransformed[i].x * g_vecProjected[i].w;
2109 g_vecProjected[i].y = g_vtxTransformed[i].y * g_vecProjected[i].w;
2110 g_vecProjected[i].z = g_vtxTransformed[i].z * g_vecProjected[i].w;
2113 g_fFogCoord[i] = g_vecProjected[i].z;
2114 if( g_vecProjected[i].w < 0 || g_vecProjected[i].z < 0 || g_fFogCoord[i] < gRSPfFogMin )
2115 g_fFogCoord[i] = gRSPfFogMin;
2117 RSP_Vtx_Clipping(i);
2119 uint8 *addr = g_pRDRAMu8+dwPDCIAddr+ (vert.cidx&0xFF);
2125 if( gRSP.bLightingEnable )
2127 g_normal.x = (char)r;
2128 g_normal.y = (char)g;
2129 g_normal.z = (char)b;
2130 #if !defined(NO_ASM)
2131 if( status.isSSEEnabled )
2133 SSEVec3TransformNormal();
2134 g_dwVtxDifColor[i] = SSELightVert();
2139 Vec3TransformNormal(g_normal, gRSPmodelViewTop);
2140 g_dwVtxDifColor[i] = LightVert(g_normal, i);
2142 *(((uint8*)&(g_dwVtxDifColor[i]))+3) = (uint8)a; // still use alpha from the vertex
2146 if( (gRDP.geometryMode & G_SHADE) == 0 && gRSP.ucode < 5 ) //Shade is disabled
2148 g_dwVtxDifColor[i] = gRDP.primitiveColor;
2152 g_dwVtxDifColor[i] = COLOR_RGBA(r, g, b, a);
2156 if( options.bWinFrameMode )
2158 g_dwVtxDifColor[i] = COLOR_RGBA(r, g, b, a);
2161 ReplaceAlphaWithFogFactor(i);
2163 VECTOR2 & t = g_fVtxTxtCoords[i];
2164 if (gRSP.bTextureGen && gRSP.bLightingEnable )
2166 // Not sure if we should transform the normal here
2167 //Matrix & matWV = gRSP.projectionMtxs[gRSP.projectionMtxTop];
2168 //Vec3TransformNormal(g_normal, matWV);
2170 TexGen(g_fVtxTxtCoords[i].x, g_fVtxTxtCoords[i].y);
2181 DebuggerAppendMsg("vtx %d: %d %d %d", i, vert.x,vert.y,vert.z);
2182 DebuggerAppendMsg(" : %f, %f, %f, %f",
2183 g_vtxTransformed[i].x,g_vtxTransformed[i].y,g_vtxTransformed[i].z,g_vtxTransformed[i].w);
2184 DebuggerAppendMsg(" : %X, %X, %X, %X", r,g,b,a);
2185 DebuggerAppendMsg(" : u=%f, v=%f", t.x, t.y);
2189 VTX_DUMP(TRACE2("Setting Vertexes: %d - %d\n", dwV0, dwV0+dwNum-1));
2190 DEBUGGER_PAUSE_AND_DUMP(NEXT_VERTEX_CMD,{TRACE0("Paused at Vertex Cmd");});
2193 extern uint32 dwConkerVtxZAddr;
2194 void ProcessVertexDataConker(uint32 dwAddr, uint32 dwV0, uint32 dwNum)
2196 UpdateCombinedMatrix();
2198 FiddledVtx * pVtxBase = (FiddledVtx*)(g_pRDRAMu8 + dwAddr);
2199 g_pVtxBase = pVtxBase;
2200 //short *vertexColoraddr = (short*)(g_pRDRAMu8+dwConkerVtxZAddr);
2202 for (uint32 i = dwV0; i < dwV0 + dwNum; i++)
2204 SP_Timing(RSP_GBI0_Vtx);
2206 FiddledVtx & vert = pVtxBase[i - dwV0];
2208 g_vtxNonTransformed[i].x = (float)vert.x;
2209 g_vtxNonTransformed[i].y = (float)vert.y;
2210 g_vtxNonTransformed[i].z = (float)vert.z;
2212 #if !defined(NO_ASM)
2213 if( status.isSSEEnabled )
2214 SSEVec3Transform(i);
2218 Vec3Transform(&g_vtxTransformed[i], (XVECTOR3*)&g_vtxNonTransformed[i], &gRSPworldProject); // Convert to w=1
2219 g_vecProjected[i].w = 1.0f / g_vtxTransformed[i].w;
2220 g_vecProjected[i].x = g_vtxTransformed[i].x * g_vecProjected[i].w;
2221 g_vecProjected[i].y = g_vtxTransformed[i].y * g_vecProjected[i].w;
2222 g_vecProjected[i].z = g_vtxTransformed[i].z * g_vecProjected[i].w;
2225 g_fFogCoord[i] = g_vecProjected[i].z;
2226 if( g_vecProjected[i].w < 0 || g_vecProjected[i].z < 0 || g_fFogCoord[i] < gRSPfFogMin )
2227 g_fFogCoord[i] = gRSPfFogMin;
2231 uint32 *dat = (uint32*)(&vert);
2232 DebuggerAppendMsg("vtx %d: %08X %08X %08X %08X", i, dat[0],dat[1],dat[2],dat[3]);
2233 DebuggerAppendMsg(" : %f, %f, %f, %f",
2234 g_vtxTransformed[i].x,g_vtxTransformed[i].y,g_vtxTransformed[i].z,g_vtxTransformed[i].w);
2235 DebuggerAppendMsg(" : %f, %f, %f, %f",
2236 g_vecProjected[i].x,g_vecProjected[i].y,g_vecProjected[i].z,g_vecProjected[i].w);
2239 RSP_Vtx_Clipping(i);
2241 if( gRSP.bLightingEnable )
2244 uint32 r= ((gRSP.ambientLightColor>>16)&0xFF);
2245 uint32 g= ((gRSP.ambientLightColor>> 8)&0xFF);
2246 uint32 b= ((gRSP.ambientLightColor )&0xFF);
2247 for( uint32 k=1; k<=gRSPnumLights; k++)
2249 r += gRSPlights[k].r;
2250 g += gRSPlights[k].g;
2251 b += gRSPlights[k].b;
2262 g_dwVtxDifColor[i] = 0xFF000000;
2263 g_dwVtxDifColor[i] |= (r<<16);
2264 g_dwVtxDifColor[i] |= (g<< 8);
2265 g_dwVtxDifColor[i] |= (b );
2268 *(((uint8*)&(g_dwVtxDifColor[i]))+3) = vert.rgba.a; // still use alpha from the vertex
2272 if( (gRDP.geometryMode & G_SHADE) == 0 && gRSP.ucode < 5 ) //Shade is disabled
2274 g_dwVtxDifColor[i] = gRDP.primitiveColor;
2278 g_dwVtxDifColor[i] = COLOR_RGBA(vert.rgba.r, vert.rgba.g, vert.rgba.b, vert.rgba.a);
2282 if( options.bWinFrameMode )
2284 //g_vecProjected[i].z = 0;
2285 g_dwVtxDifColor[i] = COLOR_RGBA(vert.rgba.r, vert.rgba.g, vert.rgba.b, vert.rgba.a);
2288 ReplaceAlphaWithFogFactor(i);
2290 // Update texture coords n.b. need to divide tu/tv by bogus scale on addition to buffer
2291 //VECTOR2 & t = g_fVtxTxtCoords[i];
2293 // If the vert is already lit, then there is no normal (and hence we
2294 // can't generate tex coord)
2295 if (gRSP.bTextureGen && gRSP.bLightingEnable )
2297 g_normal.x = (float)*(char*)(g_pRDRAMu8+ (((i<<1)+0)^3)+dwConkerVtxZAddr);
2298 g_normal.y = (float)*(char*)(g_pRDRAMu8+ (((i<<1)+1)^3)+dwConkerVtxZAddr);
2299 g_normal.z = (float)*(char*)(g_pRDRAMu8+ (((i<<1)+2)^3)+dwConkerVtxZAddr);
2300 Vec3TransformNormal(g_normal, gRSPmodelViewTop);
2301 TexGen(g_fVtxTxtCoords[i].x, g_fVtxTxtCoords[i].y);
2305 g_fVtxTxtCoords[i].x = (float)vert.tu;
2306 g_fVtxTxtCoords[i].y = (float)vert.tv;
2310 VTX_DUMP(TRACE2("Setting Vertexes: %d - %d\n", dwV0, dwV0+dwNum-1));
2311 DEBUGGER_PAUSE_AND_DUMP(NEXT_VERTEX_CMD,{DebuggerAppendMsg("Paused at Vertex Cmd");});
2338 void ProcessVertexData_Rogue_Squadron(uint32 dwXYZAddr, uint32 dwColorAddr, uint32 dwXYZCmd, uint32 dwColorCmd)
2340 UpdateCombinedMatrix();
2343 uint32 dwNum = (dwXYZCmd&0xFF00)>>10;
2345 RS_Vtx_XYZ * pVtxXYZBase = (RS_Vtx_XYZ*)(g_pRDRAMu8 + dwXYZAddr);
2346 RS_Vtx_Color * pVtxColorBase = (RS_Vtx_Color*)(g_pRDRAMu8 + dwColorAddr);
2349 for (i = dwV0; i < dwV0 + dwNum; i++)
2351 RS_Vtx_XYZ & vertxyz = pVtxXYZBase[i - dwV0];
2352 RS_Vtx_Color & vertcolors = pVtxColorBase[i - dwV0];
2354 g_vtxNonTransformed[i].x = (float)vertxyz.x;
2355 g_vtxNonTransformed[i].y = (float)vertxyz.y;
2356 g_vtxNonTransformed[i].z = (float)vertxyz.z;
2358 #if !defined(NO_ASM)
2359 if( status.isSSEEnabled )
2360 SSEVec3Transform(i);
2364 Vec3Transform(&g_vtxTransformed[i], (XVECTOR3*)&g_vtxNonTransformed[i], &gRSPworldProject); // Convert to w=1
2365 g_vecProjected[i].w = 1.0f / g_vtxTransformed[i].w;
2366 g_vecProjected[i].x = g_vtxTransformed[i].x * g_vecProjected[i].w;
2367 g_vecProjected[i].y = g_vtxTransformed[i].y * g_vecProjected[i].w;
2368 g_vecProjected[i].z = g_vtxTransformed[i].z * g_vecProjected[i].w;
2373 DebuggerAppendMsg(" : %f, %f, %f, %f",
2374 g_vtxTransformed[i].x,g_vtxTransformed[i].y,g_vtxTransformed[i].z,g_vtxTransformed[i].w);
2375 DebuggerAppendMsg(" : %f, %f, %f, %f",
2376 g_vecProjected[i].x,g_vecProjected[i].y,g_vecProjected[i].z,g_vecProjected[i].w);
2379 g_fFogCoord[i] = g_vecProjected[i].z;
2380 if( g_vecProjected[i].w < 0 || g_vecProjected[i].z < 0 || g_fFogCoord[i] < gRSPfFogMin )
2381 g_fFogCoord[i] = gRSPfFogMin;
2383 RSP_Vtx_Clipping(i);
2385 if( gRSP.bLightingEnable )
2387 g_normal.x = (float)vertcolors.nx;
2388 g_normal.y = (float)vertcolors.ny;
2389 g_normal.z = (float)vertcolors.nz;
2391 #if !defined(NO_ASM)
2392 if( status.isSSEEnabled )
2394 SSEVec3TransformNormal();
2395 g_dwVtxDifColor[i] = SSELightVert();
2400 Vec3TransformNormal(g_normal, gRSPmodelViewTop);
2401 g_dwVtxDifColor[i] = LightVert(g_normal, i);
2403 *(((uint8*)&(g_dwVtxDifColor[i]))+3) = vertcolors.a; // still use alpha from the vertex
2407 if( (gRDP.geometryMode & G_SHADE) == 0 && gRSP.ucode < 5 ) //Shade is disabled
2409 g_dwVtxDifColor[i] = gRDP.primitiveColor;
2413 g_dwVtxDifColor[i] = COLOR_RGBA(vertcolors.r, vertcolors.g, vertcolors.b, vertcolors.a);
2417 if( options.bWinFrameMode )
2419 g_dwVtxDifColor[i] = COLOR_RGBA(vertcolors.r, vertcolors.g, vertcolors.b, vertcolors.a);
2422 ReplaceAlphaWithFogFactor(i);
2425 // Update texture coords n.b. need to divide tu/tv by bogus scale on addition to buffer
2426 VECTOR2 & t = g_fVtxTxtCoords[i];
2428 // If the vert is already lit, then there is no normal (and hence we
2429 // can't generate tex coord)
2430 if (gRSP.bTextureGen && gRSP.bLightingEnable && g_textures[gRSP.curTile].m_bTextureEnable )
2432 TexGen(g_fVtxTxtCoords[i].x, g_fVtxTxtCoords[i].y);
2436 t.x = (float)vert.tu;
2437 t.y = (float)vert.tv;
2442 VTX_DUMP(TRACE2("Setting Vertexes: %d - %d\n", dwV0, dwV0+dwNum-1));
2443 DEBUGGER_PAUSE_AND_DUMP(NEXT_VERTEX_CMD,{TRACE0("Paused at Vertex Cmd");});
2446 void SetLightCol(uint32 dwLight, uint32 dwCol)
2448 gRSPlights[dwLight].r = (uint8)((dwCol >> 24)&0xFF);
2449 gRSPlights[dwLight].g = (uint8)((dwCol >> 16)&0xFF);
2450 gRSPlights[dwLight].b = (uint8)((dwCol >> 8)&0xFF);
2451 gRSPlights[dwLight].a = 255; // Ignore light alpha
2452 gRSPlights[dwLight].fr = (float)gRSPlights[dwLight].r;
2453 gRSPlights[dwLight].fg = (float)gRSPlights[dwLight].g;
2454 gRSPlights[dwLight].fb = (float)gRSPlights[dwLight].b;
2455 gRSPlights[dwLight].fa = 255; // Ignore light alpha
2457 //TRACE1("Set light %d color", dwLight);
2458 LIGHT_DUMP(TRACE2("Set Light %d color: %08X", dwLight, dwCol));
2461 void SetLightDirection(uint32 dwLight, float x, float y, float z, float range)
2463 //gRSP.bLightIsUpdated = true;
2465 //gRSPlights[dwLight].ox = x;
2466 //gRSPlights[dwLight].oy = y;
2467 //gRSPlights[dwLight].oz = z;
2469 register float w = range == 0 ? (float)sqrt(x*x+y*y+z*z) : 1;
2471 gRSPlights[dwLight].x = x/w;
2472 gRSPlights[dwLight].y = y/w;
2473 gRSPlights[dwLight].z = z/w;
2474 gRSPlights[dwLight].range = range;
2475 DEBUGGER_PAUSE_AND_DUMP(NEXT_SET_LIGHT,TRACE5("Set Light %d dir: %.4f, %.4f, %.4f, %.4f", dwLight, x, y, z, range));
2478 static float maxS0, maxT0;
2479 static float maxS1, maxT1;
2480 static bool validS0, validT0;
2481 static bool validS1, validT1;
2483 void LogTextureCoords(float fTex0S, float fTex0T, float fTex1S, float fTex1T)
2487 if( fTex0S<0 || fTex0S>maxS0 ) validS0 = false;
2491 if( fTex0T<0 || fTex0T>maxT0 ) validT0 = false;
2495 if( fTex1S<0 || fTex1S>maxS1 ) validS1 = false;
2499 if( fTex1T<0 || fTex1T>maxT1 ) validT1 = false;
2503 bool CheckTextureCoords(int tex)
2507 return validS0&&validT0;
2511 return validS1&&validT1;
2515 void ResetTextureCoordsLog(float maxs0, float maxt0, float maxs1, float maxt1)
2521 validS0 = validT0 = true;
2522 validS1 = validT1 = true;
2525 void ForceMainTextureIndex(int dwTile)
2527 if( dwTile == 1 && !(CRender::g_pRender->IsTexel0Enable()) && CRender::g_pRender->IsTexel1Enable() )
2534 gRSP.curTile = dwTile;
2538 float HackZ2(float z)
2544 float HackZ(float z)
2548 if( z < 0.1 && z >= 0 )
2551 //return (10+z)/100;
2556 void HackZ(std::vector<XVECTOR3>& points)
2558 int size = points.size();
2559 for( int i=0; i<size; i++)
2561 XVECTOR3 &v = points[i];
2562 v.z = (float)HackZ(v.z);
2568 if( CDeviceBuilder::m_deviceGeneralType == DIRECTX_DEVICE )
2570 for( uint32 i=0; i<gRSP.numVertices; i++)
2572 g_vtxBuffer[i].z = HackZ(g_vtxBuffer[i].z);
2577 for( uint32 i=0; i<gRSP.numVertices; i++)
2579 float w = g_vtxProjected5[i][3];
2580 g_vtxProjected5[i][2] = HackZ(g_vtxProjected5[i][2]/w)*w;
2586 extern XMATRIX reverseXY;
2587 extern XMATRIX reverseY;
2589 void UpdateCombinedMatrix()
2591 if( gRSP.bMatrixIsUpdated )
2593 gRSPworldProject = gRSP.modelviewMtxs[gRSP.modelViewMtxTop] * gRSP.projectionMtxs[gRSP.projectionMtxTop];
2594 gRSP.bMatrixIsUpdated = false;
2595 gRSP.bCombinedMatrixIsUpdated = true;
2598 if( gRSP.bCombinedMatrixIsUpdated )
2600 if( options.enableHackForGames == HACK_REVERSE_XY_COOR )
2602 gRSPworldProject = gRSPworldProject * reverseXY;
2604 if( options.enableHackForGames == HACK_REVERSE_Y_COOR )
2606 gRSPworldProject = gRSPworldProject * reverseY;
2608 #if !defined(NO_ASM)
2609 if( status.isSSEEnabled )
2611 MatrixTranspose(&gRSPworldProjectTransported, &gRSPworldProject);
2614 gRSP.bCombinedMatrixIsUpdated = false;
2617 //if( gRSP.bWorldMatrixIsUpdated || gRSP.bLightIsUpdated )
2619 // // Update lights with transported world matrix
2620 // for( unsigned int l=0; l<gRSPnumLights; l++)
2622 // Vec3TransformCoord(&gRSPlights[l].td, &gRSPlights[l].od, &gRSPmodelViewTopTranspose);
2623 // Vec3Normalize(&gRSPlights[l].td,&gRSPlights[l].td);
2626 // gRSP.bWorldMatrixIsUpdated = false;
2627 // gRSP.bLightIsUpdated = false;