notaz.gp2x.de / mupen64plus-pandora.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
rice: optimize IsTriangleVisible
[mupen64plus-pandora.git] / source / gles2rice / src / RenderBase.cpp
CommitLineData
292f9317 1/*
2Copyright (C) 2003 Rice1964
3
4This program is free software; you can redistribute it and/or
5modify it under the terms of the GNU General Public License
6as published by the Free Software Foundation; either version 2
7of the License, or (at your option) any later version.
8
9This program is distributed in the hope that it will be useful,
10but WITHOUT ANY WARRANTY; without even the implied warranty of
11MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
12GNU General Public License for more details.
13
14You should have received a copy of the GNU General Public License
15along with this program; if not, write to the Free Software
16Foundation, 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
30extern 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
42inline 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
106inline void RSP_Vtx_Clipping(int i) {}
107#endif
108
109/*
110 * Global variables
111 */
112ALIGN(16,RSP_Options gRSP)
113ALIGN(16,RDP_Options gRDP)
114
115static ALIGN(16,XVECTOR4 g_normal)
116//static int norms[3];
117
118ALIGN(16,XVECTOR4 g_vtxNonTransformed[MAX_VERTS])
119ALIGN(16,XVECTOR4 g_vecProjected[MAX_VERTS])
120ALIGN(16,XVECTOR4 g_vtxTransformed[MAX_VERTS])
121
122float g_vtxProjected5[1000][5];
123float g_vtxProjected5Clipped[2000][5];
124
125//uint32 g_dwVtxFlags[MAX_VERTS]; // Z_POS Z_NEG etc
126VECTOR2 g_fVtxTxtCoords[MAX_VERTS];
127uint32 g_dwVtxDifColor[MAX_VERTS];
128uint32 g_clipFlag[MAX_VERTS];
129uint32 g_clipFlag2[MAX_VERTS];
130RenderTexture g_textures[MAX_TEXTURES];
131float g_fFogCoord[MAX_VERTS];
132
133EXTERNAL_VERTEX g_vtxForExternal[MAX_VERTS];
134
135TLITVERTEX g_vtxBuffer[1000];
136TLITVERTEX g_clippedVtxBuffer[2000];
137uint8 g_oglVtxColors[1000][4];
138int g_clippedVtxCount=0;
139TLITVERTEX g_texRectTVtx[4];
140unsigned short g_vtxIndex[1000];
141unsigned int g_minIndex, g_maxIndex;
142
143float gRSPfFogMin;
144float gRSPfFogMax;
145float gRSPfFogDivider;
146
147uint32 gRSPnumLights;
148Light gRSPlights[16];
149
150ALIGN(16,Matrix gRSPworldProjectTransported)
151ALIGN(16,Matrix gRSPworldProject)
152ALIGN(16,Matrix gRSPmodelViewTop)
153ALIGN(16,Matrix gRSPmodelViewTopTranspose)
154ALIGN(16,Matrix dkrMatrixTransposed)
155
156N64Light gRSPn64lights[16];
157
158
159void (*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);
167n.y = (g_normal.x * matWorld.m01) + (g_normal.y * matWorld.m11) + (g_normal.z * matWorld.m21);
168n.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#define Vec3TransformNormal(vec, m) \
246 VECTOR3 temp; \
247 temp.x = (vec.x * m._11) + (vec.y * m._21) + (vec.z * m._31); \
248 temp.y = (vec.x * m._12) + (vec.y * m._22) + (vec.z * m._32); \
249 temp.z = (vec.x * m._13) + (vec.y * m._23) + (vec.z * m._33); \
250 float norm = sqrt(temp.x*temp.x+temp.y*temp.y+temp.z*temp.z); \
251 if (norm == 0.0) { vec.x = 0.0; vec.y = 0.0; vec.z = 0.0;} else \
252 { vec.x = temp.x/norm; vec.y = temp.y/norm; vec.z = temp.z/norm; }
253#endif
254
255
256#if !defined(__GNUC__) && !defined(NO_ASM)
257__declspec( naked ) void __fastcall SSEVec3Transform(int i)
258{
259 __asm
260 {
261 shl ecx,4; // ecx = i
262
263 movaps xmm1, DWORD PTR g_vtxNonTransformed [ecx]; // xmm1 as original vector
264
265 movaps xmm4, DWORD PTR gRSPworldProjectTransported; // row1
266 movaps xmm5, DWORD PTR gRSPworldProjectTransported[0x10]; // row2
267 movaps xmm6, DWORD PTR gRSPworldProjectTransported[0x20]; // row3
268 movaps xmm7, DWORD PTR gRSPworldProjectTransported[0x30]; // row4
269
270 mulps xmm4, xmm1; // row 1
271 mulps xmm5, xmm1; // row 2
272 mulps xmm6, xmm1; // row 3
273 mulps xmm7, xmm1; // row 4
274
275 movhlps xmm0, xmm4; // xmm4 high to xmm0 low
276 movlhps xmm0, xmm5; // xmm5 low to xmm0 high
277
278 addps xmm4, xmm0; // result of add are in xmm4 low
279 addps xmm5, xmm0; // result of add are in xmm5 high
280
281 shufps xmm0, xmm4, 0x44; // move xmm4 low DWORDs to xmm0 high
282 shufps xmm4, xmm5, 0xe4; // move xmm5 high DWORS to xmm4
283 movhlps xmm5, xmm0; // xmm4, xmm5 are mirrored
284
285 shufps xmm4, xmm4, 0x08; // move xmm4's 3rd uint32 to its 2nd uint32
286 shufps xmm5, xmm5, 0x0d; // move xmm5's 4th uint32 to its 2nd uint32,
287 // and move its 2nd uint32 to its 1st uint32
288
289 addps xmm4, xmm5; // results are in 1st and 2nd uint32
290
291
292 movhlps xmm0, xmm6; // xmm6 high to xmm0 low
293 movlhps xmm0, xmm7; // xmm7 low to xmm0 high
294
295 addps xmm6, xmm0; // result of add are in xmm6 low
296 addps xmm7, xmm0; // result of add are in xmm7 high
297
298 shufps xmm0, xmm6, 0x44; // move xmm6 low DWORDs to xmm0 high
299 shufps xmm6, xmm7, 0xe4; // move xmm7 high DWORS to xmm6
300 movhlps xmm7, xmm0; // xmm6, xmm7 are mirrored
301
302 shufps xmm6, xmm6, 0x08; // move xmm6's 3rd uint32 to its 2nd uint32
303 shufps xmm7, xmm7, 0x0d; // move xmm7's 4th uint32 to its 2nd uint32,
304 // and move its 2nd uint32 to its 1st uint32
305
306 addps xmm6, xmm7; // results are in 1st and 2nd uint32
307
308 movlhps xmm4, xmm6; // final result is in xmm4
309 movaps DWORD PTR g_vtxTransformed [ecx], xmm4;
310
311 movaps xmm0,xmm4;
312 shufps xmm0,xmm0,0xff;
313 divps xmm4,xmm0;
314 rcpps xmm0,xmm0;
315 movhlps xmm0,xmm4;
316 shufps xmm0,xmm0,0xe8;
317 movlhps xmm4,xmm0;
318
319 movaps DWORD PTR g_vecProjected [ecx], xmm4;
320
321 emms;
322 ret;
323 }
324}
325
326// Only used by DKR
327__declspec( naked ) void __fastcall SSEVec3TransformDKR(XVECTOR4 &pOut, const XVECTOR4 &pV)
328{
329 __asm
330 {
331 movaps xmm1, DWORD PTR [edx]; // xmm1 as original vector
332
333 movaps xmm4, DWORD PTR dkrMatrixTransposed; // row1
334 movaps xmm5, DWORD PTR dkrMatrixTransposed[0x10]; // row2
335 movaps xmm6, DWORD PTR dkrMatrixTransposed[0x20]; // row3
336 movaps xmm7, DWORD PTR dkrMatrixTransposed[0x30]; // row4
337
338 mulps xmm4, xmm1; // row 1
339 mulps xmm5, xmm1; // row 2
340 mulps xmm6, xmm1; // row 3
341 mulps xmm7, xmm1; // row 4
342
343 movhlps xmm0, xmm4; // xmm4 high to xmm0 low
344 movlhps xmm0, xmm5; // xmm5 low to xmm0 high
345
346 addps xmm4, xmm0; // result of add are in xmm4 low
347 addps xmm5, xmm0; // result of add are in xmm5 high
348
349 shufps xmm0, xmm4, 0x44; // move xmm4 low DWORDs to xmm0 high
350 shufps xmm4, xmm5, 0xe4; // move xmm5 high DWORS to xmm4
351 movhlps xmm5, xmm0; // xmm4, xmm5 are mirrored
352
353 shufps xmm4, xmm4, 0x08; // move xmm4's 3rd uint32 to its 2nd uint32
354 shufps xmm5, xmm5, 0x0d; // move xmm5's 4th uint32 to its 2nd uint32,
355 // and move its 2nd uint32 to its 1st uint32
356
357 addps xmm4, xmm5; // results are in 1st and 2nd uint32
358
359
360 movhlps xmm0, xmm6; // xmm6 high to xmm0 low
361 movlhps xmm0, xmm7; // xmm7 low to xmm0 high
362
363 addps xmm6, xmm0; // result of add are in xmm6 low
364 addps xmm7, xmm0; // result of add are in xmm7 high
365
366 shufps xmm0, xmm6, 0x44; // move xmm6 low DWORDs to xmm0 high
367 shufps xmm6, xmm7, 0xe4; // move xmm7 high DWORS to xmm6
368 movhlps xmm7, xmm0; // xmm6, xmm7 are mirrored
369
370 shufps xmm6, xmm6, 0x08; // move xmm6's 3rd uint32 to its 2nd uint32
371 shufps xmm7, xmm7, 0x0d; // move xmm7's 4th uint32 to its 2nd uint32,
372 // and move its 2nd uint32 to its 1st uint32
373
374 addps xmm6, xmm7; // results are in 1st and 2nd uint32
375
376 movlhps xmm4, xmm6; // final result is in xmm4
377 movaps DWORD PTR [ecx], xmm4;
378
379 emms;
380 ret;
381 }
382}
383#elif defined(__GNUC__) && defined(__x86_64__) && !defined(NO_ASM)
384void SSEVec3Transform(int i)
385{
386 asm volatile(" shl $4, %0 \n"
387 " movslq %k0, %q0 \n"
388 " movaps (%1,%q0), %%xmm1 \n"
389 " movaps 0(%2), %%xmm4 \n"
390 " movaps 16(%2), %%xmm5 \n"
391 " movaps 32(%2), %%xmm6 \n"
392 " movaps 48(%2), %%xmm7 \n"
393 " mulps %%xmm1, %%xmm4 \n"
394 " mulps %%xmm1, %%xmm5 \n"
395 " mulps %%xmm1, %%xmm6 \n"
396 " mulps %%xmm1, %%xmm7 \n"
397 " movhlps %%xmm4, %%xmm0 \n"
398 " movlhps %%xmm5, %%xmm0 \n"
399 " addps %%xmm0, %%xmm4 \n"
400 " addps %%xmm0, %%xmm5 \n"
401 " shufps $0x44, %%xmm4, %%xmm0 \n"
402 " shufps $0xe4, %%xmm5, %%xmm4 \n"
403 " movhlps %%xmm0, %%xmm5 \n"
404 " shufps $0x08, %%xmm4, %%xmm4 \n"
405 " shufps $0x0d, %%xmm5, %%xmm5 \n"
406 " addps %%xmm5, %%xmm4 \n"
407 " movhlps %%xmm6, %%xmm0 \n"
408 " movlhps %%xmm7, %%xmm0 \n"
409 " addps %%xmm0, %%xmm6 \n"
410 " addps %%xmm0, %%xmm7 \n"
411 " shufps $0x44, %%xmm6, %%xmm0 \n"
412 " shufps $0xe4, %%xmm7, %%xmm6 \n"
413 " movhlps %%xmm0, %%xmm7 \n"
414 " shufps $0x08, %%xmm6, %%xmm6 \n"
415 " shufps $0x0d, %%xmm7, %%xmm7 \n"
416 " addps %%xmm7, %%xmm6 \n"
417 " movlhps %%xmm6, %%xmm4 \n"
418 " movaps %%xmm4, (%3,%q0) \n"
419 " movaps %%xmm4, %%xmm0 \n"
420 " shufps $0xff, %%xmm0, %%xmm0 \n"
421 " divps %%xmm0, %%xmm4 \n"
422 " rcpps %%xmm0, %%xmm0 \n"
423 " movhlps %%xmm4, %%xmm0 \n"
424 " shufps $0xe8, %%xmm0, %%xmm0 \n"
425 " movlhps %%xmm0, %%xmm4 \n"
426 " movaps %%xmm4, (%4,%q0) \n"
427 : "+r"(i)
428 : "r"(g_vtxNonTransformed), "r"(&gRSPworldProjectTransported.m[0][0]), "r"(g_vtxTransformed), "r"(g_vecProjected)
429 : "memory", "%xmm0", "%xmm1", "%xmm4", "%xmm5", "%xmm6", "%xmm7"
430 );
431}
432#elif !defined(NO_ASM) // 32-bit GCC assumed
433void SSEVec3Transform(int i)
434{
435 asm volatile(" shl $4, %0 \n"
436 " movaps (%1,%0), %%xmm1 \n"
437 " movaps 0(%2), %%xmm4 \n"
438 " movaps 16(%2), %%xmm5 \n"
439 " movaps 32(%2), %%xmm6 \n"
440 " movaps 48(%2), %%xmm7 \n"
441 " mulps %%xmm1, %%xmm4 \n"
442 " mulps %%xmm1, %%xmm5 \n"
443 " mulps %%xmm1, %%xmm6 \n"
444 " mulps %%xmm1, %%xmm7 \n"
445 " movhlps %%xmm4, %%xmm0 \n"
446 " movlhps %%xmm5, %%xmm0 \n"
447 " addps %%xmm0, %%xmm4 \n"
448 " addps %%xmm0, %%xmm5 \n"
449 " shufps $0x44, %%xmm4, %%xmm0 \n"
450 " shufps $0xe4, %%xmm5, %%xmm4 \n"
451 " movhlps %%xmm0, %%xmm5 \n"
452 " shufps $0x08, %%xmm4, %%xmm4 \n"
453 " shufps $0x0d, %%xmm5, %%xmm5 \n"
454 " addps %%xmm5, %%xmm4 \n"
455 " movhlps %%xmm6, %%xmm0 \n"
456 " movlhps %%xmm7, %%xmm0 \n"
457 " addps %%xmm0, %%xmm6 \n"
458 " addps %%xmm0, %%xmm7 \n"
459 " shufps $0x44, %%xmm6, %%xmm0 \n"
460 " shufps $0xe4, %%xmm7, %%xmm6 \n"
461 " movhlps %%xmm0, %%xmm7 \n"
462 " shufps $0x08, %%xmm6, %%xmm6 \n"
463 " shufps $0x0d, %%xmm7, %%xmm7 \n"
464 " addps %%xmm7, %%xmm6 \n"
465 " movlhps %%xmm6, %%xmm4 \n"
466 " movaps %%xmm4, (%3,%0) \n"
467 " movaps %%xmm4, %%xmm0 \n"
468 " shufps $0xff, %%xmm0, %%xmm0 \n"
469 " divps %%xmm0, %%xmm4 \n"
470 " rcpps %%xmm0, %%xmm0 \n"
471 " movhlps %%xmm4, %%xmm0 \n"
472 " shufps $0xe8, %%xmm0, %%xmm0 \n"
473 " movlhps %%xmm0, %%xmm4 \n"
474 " movaps %%xmm4, (%4,%0) \n"
475 : "+r"(i)
476 : "r"(g_vtxNonTransformed), "r"(&gRSPworldProjectTransported.m[0][0]), "r"(g_vtxTransformed), "r"(g_vecProjected)
477 : "memory", "%xmm0", "%xmm1", "%xmm4", "%xmm5", "%xmm6", "%xmm7"
478 );
479}
480#endif
481float real255 = 255.0f;
482float real128 = 128.0f;
483
484#if !defined(__GNUC__) && !defined(NO_ASM)
485__declspec( naked ) void __fastcall SSEVec3TransformNormal()
486{
487 __asm
488 {
489 mov DWORD PTR [g_normal][12], 0;
490
491 movaps xmm4, DWORD PTR gRSPmodelViewTopTranspose; // row1
492 movaps xmm5, DWORD PTR gRSPmodelViewTopTranspose[0x10]; // row2
493 movaps xmm1, DWORD PTR [g_normal]; // xmm1 as the normal vector
494 movaps xmm6, DWORD PTR gRSPmodelViewTopTranspose[0x20]; // row3
495
496 mulps xmm4, xmm1; // row 1
497 mulps xmm5, xmm1; // row 2
498 mulps xmm6, xmm1; // row 3
499
500 movhlps xmm0, xmm4; // xmm4 high to xmm0 low
501 movlhps xmm0, xmm5; // xmm5 low to xmm0 high
502
503 addps xmm4, xmm0; // result of add are in xmm4 low
504 addps xmm5, xmm0; // result of add are in xmm5 high
505
506 shufps xmm0, xmm4, 0x44; // move xmm4 low DWORDs to xmm0 high
507 shufps xmm4, xmm5, 0xe4; // move xmm5 high DWORS to xmm4
508 movhlps xmm5, xmm0; // xmm4, xmm5 are mirrored
509
510 shufps xmm4, xmm4, 0x08; // move xmm4's 3rd uint32 to its 2nd uint32
511 shufps xmm5, xmm5, 0x0d; // move xmm5's 4th uint32 to its 2nd uint32,
512
513 addps xmm4, xmm5; // results are in 1st and 2nd uint32
514
515 movaps xmm1,xmm4;
516 mulps xmm1,xmm1; //square
517 movlhps xmm7, xmm1;
518 shufps xmm7, xmm7,0x03;
519 addss xmm7, xmm1;
520
521 movhlps xmm0, xmm6; // xmm6 high to xmm0 low
522 addps xmm6, xmm0; // result of add are in xmm6 low
523
524 movlhps xmm0, xmm6;
525 shufps xmm0, xmm0, 0x03;
526 addss xmm0, xmm6; // result of add is at xmm0's 1st uint32
527
528 movlhps xmm4, xmm0;
529
530 mulss xmm0,xmm0;
531 addss xmm7,xmm0; // xmm7 1st uint32 is the sum of squares
532
533#ifdef DEBUGGER
534 movaps DWORD PTR [g_normal], xmm4;
535 movss DWORD PTR [g_normal][12], xmm7;
536#endif
537 xorps xmm0,xmm0;
538 ucomiss xmm0,xmm7;
539 jz l2
540
541 rsqrtss xmm7,xmm7;
542 shufps xmm7,xmm7,0;
543#ifdef DEBUGGER
544 movss DWORD PTR [g_normal][12], xmm7;
545#endif
546 mulps xmm4,xmm7;
547
548 movaps DWORD PTR [g_normal], xmm4; // Normalized
549 mov DWORD PTR [g_normal][12], 0;
550
551 emms;
552 ret;
553l2:
554 movss DWORD PTR [g_normal], xmm0;
555 movss DWORD PTR [g_normal][12], xmm0;
556 emms;
557 ret;
558 }
559}
560#elif defined(__GNUC__) && !defined(NO_ASM) // this code should compile for both 64-bit and 32-bit architectures
561void SSEVec3TransformNormal(void)
562{
563 asm volatile(" movl $0, 12(%0) \n"
564 " movaps (%1), %%xmm4 \n"
565 " movaps 16(%1), %%xmm5 \n"
566 " movaps (%0), %%xmm1 \n"
567 " movaps 32(%1), %%xmm6 \n"
568 " mulps %%xmm1, %%xmm4 \n"
569 " mulps %%xmm1, %%xmm5 \n"
570 " mulps %%xmm1, %%xmm6 \n"
571 " movhlps %%xmm4, %%xmm0 \n"
572 " movlhps %%xmm5, %%xmm0 \n"
573 " addps %%xmm0, %%xmm4 \n"
574 " addps %%xmm0, %%xmm5 \n"
575 " shufps $0x44, %%xmm4, %%xmm0 \n"
576 " shufps $0xe4, %%xmm5, %%xmm4 \n"
577 " movhlps %%xmm0, %%xmm5 \n"
578 " shufps $0x08, %%xmm4, %%xmm4 \n"
579 " shufps $0x0d, %%xmm5, %%xmm5 \n"
580 " addps %%xmm5, %%xmm4 \n"
581 " movaps %%xmm4, %%xmm1 \n"
582 " mulps %%xmm1, %%xmm1 \n"
583 " movlhps %%xmm1, %%xmm7 \n"
584 " shufps $0x03, %%xmm7, %%xmm7 \n"
585 " addss %%xmm1, %%xmm7 \n"
586 " movhlps %%xmm6, %%xmm0 \n"
587 " addps %%xmm0, %%xmm6 \n"
588 " movlhps %%xmm6, %%xmm0 \n"
589 " shufps $0x03, %%xmm0, %%xmm0 \n"
590 " addss %%xmm6, %%xmm0 \n"
591 " movlhps %%xmm0, %%xmm4 \n"
592 " mulss %%xmm0, %%xmm0 \n"
593 " addss %%xmm0, %%xmm7 \n"
594#ifdef DEBUGGER
595 " movaps %%xmm4, (%0) \n"
596 " movss %%xmm7, 12(%0) \n"
597#endif
598 " xorps %%xmm0, %%xmm0 \n"
599 " ucomiss %%xmm7, %%xmm0 \n"
600 " jz 0f \n"
601 " rsqrtss %%xmm7, %%xmm7 \n"
602 " shufps $0x00, %%xmm7, %%xmm7 \n"
603#ifdef DEBUGGER
604 " movss %%xmm7, 12(%0) \n"
605#endif
606 " mulps %%xmm7, %%xmm4 \n"
607 " movaps %%xmm4, (%0) \n"
608 " movl $0, 12(%0) \n"
609 " jmp 1f \n"
610 "0: \n"
611 " movss %%xmm0, (%0) \n"
612 " movss %%xmm0, 12(%0) \n"
613 "1: \n"
614 :
615 : "r"(&g_normal.x), "r"(&gRSPmodelViewTopTranspose.m[0][0])
616 : "memory", "cc", "%xmm0", "%xmm1", "%xmm4", "%xmm5", "%xmm6", "%xmm7"
617 );
618}
619#endif
620
621void NormalizeNormalVec()
622{
623 float w = 1/sqrtf(g_normal.x*g_normal.x + g_normal.y*g_normal.y + g_normal.z*g_normal.z);
624 g_normal.x *= w;
625 g_normal.y *= w;
626 g_normal.z *= w;
627}
628
629
630void InitRenderBase()
631{
632#if !defined(NO_ASM)
633 if( status.isSSEEnabled && !g_curRomInfo.bPrimaryDepthHack && options.enableHackForGames != HACK_FOR_NASCAR)
634 {
635 ProcessVertexData = ProcessVertexDataSSE;
636 }
637 else
0764a475 638#elif defined(__ARM_NEON__)
639 if( !g_curRomInfo.bPrimaryDepthHack && options.enableHackForGames != HACK_FOR_NASCAR && options.enableHackForGames != HACK_FOR_ZELDA_MM && !options.bWinFrameMode)
640 {
641 ProcessVertexData = ProcessVertexDataNEON;
642 }
643 else
292f9317 644#endif
645 {
646 ProcessVertexData = ProcessVertexDataNoSSE;
647 }
648
649 gRSPfFogMin = gRSPfFogMax = 0.0f;
650 windowSetting.fMultX = windowSetting.fMultY = 2.0f;
651 windowSetting.vpLeftW = windowSetting.vpTopW = 0;
652 windowSetting.vpRightW = windowSetting.vpWidthW = 640;
653 windowSetting.vpBottomW = windowSetting.vpHeightW = 480;
654 gRSP.maxZ = 0;
655 gRSP.nVPLeftN = gRSP.nVPTopN = 0;
656 gRSP.nVPRightN = 640;
657 gRSP.nVPBottomN = 640;
658 gRSP.nVPWidthN = 640;
659 gRSP.nVPHeightN = 640;
660 gRDP.scissor.left=gRDP.scissor.top=0;
661 gRDP.scissor.right=gRDP.scissor.bottom=640;
662
663 gRSP.bLightingEnable = gRSP.bTextureGen = false;
664 gRSP.curTile=gRSPnumLights=gRSP.ambientLightColor=gRSP.ambientLightIndex= 0;
665 gRSP.fAmbientLightR=gRSP.fAmbientLightG=gRSP.fAmbientLightB=0;
666 gRSP.projectionMtxTop = gRSP.modelViewMtxTop = 0;
667 gRDP.fogColor = gRDP.primitiveColor = gRDP.envColor = gRDP.primitiveDepth = gRDP.primLODMin = gRDP.primLODFrac = gRDP.LODFrac = 0;
668 gRDP.fPrimitiveDepth = 0;
669 gRSP.numVertices = 0;
670 gRSP.maxVertexID = 0;
671 gRSP.bCullFront=false;
672 gRSP.bCullBack=true;
673 gRSP.bFogEnabled=gRDP.bFogEnableInBlender=false;
674 gRSP.bZBufferEnabled=true;
675 gRSP.shadeMode=SHADE_SMOOTH;
676 gRDP.keyR=gRDP.keyG=gRDP.keyB=gRDP.keyA=gRDP.keyRGB=gRDP.keyRGBA = 0;
677 gRDP.fKeyA = 0;
678 gRSP.DKRCMatrixIndex = gRSP.dwDKRVtxAddr = gRSP.dwDKRMatrixAddr = 0;
679 gRSP.DKRBillBoard = false;
680
681 gRSP.fTexScaleX = 1/32.0f;
682 gRSP.fTexScaleY = 1/32.0f;
683 gRSP.bTextureEnabled = FALSE;
684
685 gRSP.clip_ratio_left = 0;
686 gRSP.clip_ratio_top = 0;
687 gRSP.clip_ratio_right = 640;
688 gRSP.clip_ratio_bottom = 480;
689 gRSP.clip_ratio_negx = 1;
690 gRSP.clip_ratio_negy = 1;
691 gRSP.clip_ratio_posx = 1;
692 gRSP.clip_ratio_posy = 1;
693 gRSP.real_clip_scissor_left = 0;
694 gRSP.real_clip_scissor_top = 0;
695 gRSP.real_clip_scissor_right = 640;
696 gRSP.real_clip_scissor_bottom = 480;
697 windowSetting.clipping.left = 0;
698 windowSetting.clipping.top = 0;
699 windowSetting.clipping.right = 640;
700 windowSetting.clipping.bottom = 480;
701 windowSetting.clipping.width = 640;
702 windowSetting.clipping.height = 480;
703 windowSetting.clipping.needToClip = false;
704 gRSP.real_clip_ratio_negx = 1;
705 gRSP.real_clip_ratio_negy = 1;
706 gRSP.real_clip_ratio_posx = 1;
707 gRSP.real_clip_ratio_posy = 1;
708
709 gRSP.DKRCMatrixIndex=0;
710 gRSP.DKRVtxCount=0;
711 gRSP.DKRBillBoard = false;
712 gRSP.dwDKRVtxAddr=0;
713 gRSP.dwDKRMatrixAddr=0;
714
715
716 gRDP.geometryMode = 0;
717 gRDP.otherModeL = 0;
718 gRDP.otherModeH = 0;
719 gRDP.fillColor = 0xFFFFFFFF;
720 gRDP.originalFillColor =0;
721
722 gRSP.ucode = 1;
723 gRSP.vertexMult = 10;
724 gRSP.bNearClip = false;
725 gRSP.bRejectVtx = false;
726
727 gRDP.texturesAreReloaded = false;
728 gRDP.textureIsChanged = false;
729 gRDP.colorsAreReloaded = false;
730
731 memset(&gRDP.otherMode,0,sizeof(RDP_OtherMode));
732 memset(&gRDP.tiles,0,sizeof(Tile)*8);
733
734 for( int i=0; i<MAX_VERTS; i++ )
735 {
736 g_clipFlag[i] = 0;
737 g_vtxNonTransformed[i].w = 1;
738 }
739
740 memset(gRSPn64lights, 0, sizeof(N64Light)*16);
741}
742
743void SetFogMinMax(float fMin, float fMax, float fMul, float fOffset)
744{
745 if( fMin > fMax )
746 {
747 float temp = fMin;
748 fMin = fMax;
749 fMax = temp;
750 }
751
752 {
753 gRSPfFogMin = max(0,fMin/500-1);
754 gRSPfFogMax = fMax/500-1;
755 }
756
757 gRSPfFogDivider = 255/(gRSPfFogMax-gRSPfFogMin);
758 CRender::g_pRender->SetFogMinMax(fMin, fMax);
759}
760
761void InitVertexColors()
762{
763}
764
765void InitVertexTextureConstants()
766{
767 float scaleX;
768 float scaleY;
769
770 RenderTexture &tex0 = g_textures[gRSP.curTile];
771 //CTexture *surf = tex0.m_pCTexture;
772 Tile &tile0 = gRDP.tiles[gRSP.curTile];
773
774 scaleX = gRSP.fTexScaleX;
775 scaleY = gRSP.fTexScaleY;
776
777 gRSP.tex0scaleX = scaleX * tile0.fShiftScaleS/tex0.m_fTexWidth;
778 gRSP.tex0scaleY = scaleY * tile0.fShiftScaleT/tex0.m_fTexHeight;
779
780 gRSP.tex0OffsetX = tile0.fhilite_sl/tex0.m_fTexWidth;
781 gRSP.tex0OffsetY = tile0.fhilite_tl/tex0.m_fTexHeight;
782
783 if( CRender::g_pRender->IsTexel1Enable() )
784 {
785 RenderTexture &tex1 = g_textures[(gRSP.curTile+1)&7];
786 //CTexture *surf = tex1.m_pCTexture;
787 Tile &tile1 = gRDP.tiles[(gRSP.curTile+1)&7];
788
789 gRSP.tex1scaleX = scaleX * tile1.fShiftScaleS/tex1.m_fTexWidth;
790 gRSP.tex1scaleY = scaleY * tile1.fShiftScaleT/tex1.m_fTexHeight;
791
792 gRSP.tex1OffsetX = tile1.fhilite_sl/tex1.m_fTexWidth;
793 gRSP.tex1OffsetY = tile1.fhilite_tl/tex1.m_fTexHeight;
794 }
795
796 gRSP.texGenXRatio = tile0.fShiftScaleS;
797 gRSP.texGenYRatio = gRSP.fTexScaleX/gRSP.fTexScaleY*tex0.m_fTexWidth/tex0.m_fTexHeight*tile0.fShiftScaleT;
798}
799
800void TexGen(float &s, float &t)
801{
802 if (gRDP.geometryMode & G_TEXTURE_GEN_LINEAR)
803 {
804 s = acosf(g_normal.x) / 3.14159f;
805 t = acosf(g_normal.y) / 3.14159f;
806 }
807 else
808 {
809 s = 0.5f * ( 1.0f + g_normal.x);
810 t = 0.5f * ( 1.0f - g_normal.y);
811 }
812}
813
814void ComputeLOD(bool openGL)
815{
816 TLITVERTEX &v0 = g_vtxBuffer[0];
817 TLITVERTEX &v1 = g_vtxBuffer[1];
818 RenderTexture &tex0 = g_textures[gRSP.curTile];
819
820 float d,dt;
821 if( openGL )
822 {
823 float x = g_vtxProjected5[0][0] / g_vtxProjected5[0][4] - g_vtxProjected5[1][0] / g_vtxProjected5[1][4];
824 float y = g_vtxProjected5[0][1] / g_vtxProjected5[0][4] - g_vtxProjected5[1][1] / g_vtxProjected5[1][4];
825
826 x = windowSetting.vpWidthW*x/windowSetting.fMultX/2;
827 y = windowSetting.vpHeightW*y/windowSetting.fMultY/2;
828 d = sqrtf(x*x+y*y);
829 }
830 else
831 {
832 float x = (v0.x - v1.x)/ windowSetting.fMultX;
833 float y = (v0.y - v1.y)/ windowSetting.fMultY;
834 d = sqrtf(x*x+y*y);
835 }
836
837 float s0 = v0.tcord[0].u * tex0.m_fTexWidth;
838 float t0 = v0.tcord[0].v * tex0.m_fTexHeight;
839 float s1 = v1.tcord[0].u * tex0.m_fTexWidth;
840 float t1 = v1.tcord[0].v * tex0.m_fTexHeight;
841
842 dt = sqrtf((s0-s1)*(s0-s1)+(t0-t1)*(t0-t1));
843
844 float lod = dt/d;
845 float frac = log10f(lod)/log10f(2.0f);
846 //DEBUGGER_IF_DUMP(pauseAtNext,{DebuggerAppendMsg("LOD frac = %f", frac);});
847 frac = (lod / powf(2.0f,floorf(frac)));
848 frac = frac - floorf(frac);
849 //DEBUGGER_IF_DUMP(pauseAtNext,{DebuggerAppendMsg("LOD = %f, frac = %f", lod, frac);});
850 gRDP.LODFrac = (uint32)(frac*255);
851 CRender::g_pRender->SetCombinerAndBlender();
852}
853
854bool bHalfTxtScale=false;
855extern uint32 lastSetTile;
856
d6e5b275 857#ifndef __ARM_NEON__
858static void multiply_subtract2(float *d, const float *m1, const float *m2, const float *s)
859{
860 int i;
861 for (i = 0; i < 2; i++)
862 d[i] = m1[i] * m2[i] - s[i];
863}
864#else
865extern "C" void multiply_subtract2(float *d, const float *m1, const float *m2, const float *s);
866#endif
867
292f9317 868void InitVertex(uint32 dwV, uint32 vtxIndex, bool bTexture, bool openGL)
869{
870 VTX_DUMP(TRACE2("Init vertex (%d) to vtx buf[%d]:", dwV, vtxIndex));
871
872 TLITVERTEX &v = g_vtxBuffer[vtxIndex];
873 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));
874 if( openGL )
875 {
876 g_vtxProjected5[vtxIndex][0] = g_vtxTransformed[dwV].x;
877 g_vtxProjected5[vtxIndex][1] = g_vtxTransformed[dwV].y;
878 g_vtxProjected5[vtxIndex][2] = g_vtxTransformed[dwV].z;
879 g_vtxProjected5[vtxIndex][3] = g_vtxTransformed[dwV].w;
880 g_vtxProjected5[vtxIndex][4] = g_vecProjected[dwV].z;
881
882 if( g_vtxTransformed[dwV].w < 0 )
883 g_vtxProjected5[vtxIndex][4] = 0;
884
885 g_vtxIndex[vtxIndex] = vtxIndex;
886 }
887
888 if( !openGL || options.bOGLVertexClipper == TRUE )
889 {
890 v.x = g_vecProjected[dwV].x*gRSP.vtxXMul+gRSP.vtxXAdd;
891 v.y = g_vecProjected[dwV].y*gRSP.vtxYMul+gRSP.vtxYAdd;
892 v.z = (g_vecProjected[dwV].z + 1.0f) * 0.5f; // DirectX minZ=0, maxZ=1
893 //v.z = g_vecProjected[dwV].z; // DirectX minZ=0, maxZ=1
894 v.rhw = g_vecProjected[dwV].w;
895 VTX_DUMP(TRACE4(" Proj : x=%f, y=%f, z=%f, rhw=%f", v.x,v.y,v.z,v.rhw));
896
897 if( gRSP.bProcessSpecularColor )
898 {
899 v.dcSpecular = CRender::g_pRender->PostProcessSpecularColor();
900 if( gRSP.bFogEnabled )
901 {
902 v.dcSpecular &= 0x00FFFFFF;
903 uint32 fogFct = 0xFF-(uint8)((g_fFogCoord[dwV]-gRSPfFogMin)*gRSPfFogDivider);
904 v.dcSpecular |= (fogFct<<24);
905 }
906 }
907 else if( gRSP.bFogEnabled )
908 {
909 uint32 fogFct = 0xFF-(uint8)((g_fFogCoord[dwV]-gRSPfFogMin)*gRSPfFogDivider);
910 v.dcSpecular = (fogFct<<24);
911 }
912 }
913 VTX_DUMP(TRACE2(" (U,V): %f, %f", g_fVtxTxtCoords[dwV].x,g_fVtxTxtCoords[dwV].y));
914
915 v.dcDiffuse = g_dwVtxDifColor[dwV];
916 if( gRDP.otherMode.key_en )
917 {
918 v.dcDiffuse &= 0x00FFFFFF;
919 v.dcDiffuse |= (gRDP.keyA<<24);
920 }
921 else if( gRDP.otherMode.aa_en && gRDP.otherMode.clr_on_cvg==0 )
922 {
923 v.dcDiffuse |= 0xFF000000;
924 }
925
926 if( gRSP.bProcessDiffuseColor )
927 {
928 v.dcDiffuse = CRender::g_pRender->PostProcessDiffuseColor(v.dcDiffuse);
929 }
930 if( options.bWinFrameMode )
931 {
932 v.dcDiffuse = g_dwVtxDifColor[dwV];
933 }
934
935 if( openGL )
936 {
937 g_oglVtxColors[vtxIndex][0] = v.r;
938 g_oglVtxColors[vtxIndex][1] = v.g;
939 g_oglVtxColors[vtxIndex][2] = v.b;
940 g_oglVtxColors[vtxIndex][3] = v.a;
941 }
942
943 if( bTexture )
944 {
945 // If the vert is already lit, then there is no normal (and hence we can't generate tex coord)
946 // Only scale if not generated automatically
947 if (gRSP.bTextureGen && gRSP.bLightingEnable)
948 {
949 // Correction for texGen result
950 float u0,u1,v0,v1;
951 RenderTexture &tex0 = g_textures[gRSP.curTile];
952 u0 = g_fVtxTxtCoords[dwV].x * 32 * 1024 * gRSP.fTexScaleX / tex0.m_fTexWidth;
953 v0 = g_fVtxTxtCoords[dwV].y * 32 * 1024 * gRSP.fTexScaleY / tex0.m_fTexHeight;
954 u0 *= (gRDP.tiles[gRSP.curTile].fShiftScaleS);
955 v0 *= (gRDP.tiles[gRSP.curTile].fShiftScaleT);
956
957 if( CRender::g_pRender->IsTexel1Enable() )
958 {
959 RenderTexture &tex1 = g_textures[(gRSP.curTile+1)&7];
960 u1 = g_fVtxTxtCoords[dwV].x * 32 * 1024 * gRSP.fTexScaleX / tex1.m_fTexWidth;
961 v1 = g_fVtxTxtCoords[dwV].y * 32 * 1024 * gRSP.fTexScaleY / tex1.m_fTexHeight;
962 u1 *= gRDP.tiles[(gRSP.curTile+1)&7].fShiftScaleS;
963 v1 *= gRDP.tiles[(gRSP.curTile+1)&7].fShiftScaleT;
964 CRender::g_pRender->SetVertexTextureUVCoord(v, u0, v0, u1, v1);
965 }
966 else
967 {
968 CRender::g_pRender->SetVertexTextureUVCoord(v, u0, v0);
969 }
970 }
971 else
972 {
d6e5b275 973 TexCord tex0;
974 multiply_subtract2(&tex0.u, &g_fVtxTxtCoords[dwV].x, &gRSP.tex0scaleX, &gRSP.tex0OffsetX);
292f9317 975
976 if( CRender::g_pRender->IsTexel1Enable() )
977 {
d6e5b275 978 TexCord tex1;
979 multiply_subtract2(&tex1.u, &g_fVtxTxtCoords[dwV].x, &gRSP.tex1scaleX, &gRSP.tex1OffsetX);
292f9317 980
d6e5b275 981 CRender::g_pRender->SetVertexTextureUVCoord(v, tex0, tex1);
982 VTX_DUMP(TRACE2(" (tex0): %f, %f", tex0.u,tex0.v));
983 VTX_DUMP(TRACE2(" (tex1): %f, %f", tex1.u,tex1.v));
292f9317 984 }
985 else
986 {
d6e5b275 987 CRender::g_pRender->SetVertexTextureUVCoord(v, tex0);
988 VTX_DUMP(TRACE2(" (tex0): %f, %f", tex0.u,tex0.v));
292f9317 989 }
990 }
991
992 // Check for txt scale hack
993 if( !bHalfTxtScale && g_curRomInfo.bTextureScaleHack &&
994 (gRDP.tiles[lastSetTile].dwSize == TXT_SIZE_32b || gRDP.tiles[lastSetTile].dwSize == TXT_SIZE_4b ) )
995 {
996 int width = ((gRDP.tiles[lastSetTile].sh-gRDP.tiles[lastSetTile].sl+1)<<1);
997 int height = ((gRDP.tiles[lastSetTile].th-gRDP.tiles[lastSetTile].tl+1)<<1);
998 if( g_fVtxTxtCoords[dwV].x*gRSP.fTexScaleX == width || g_fVtxTxtCoords[dwV].y*gRSP.fTexScaleY == height )
999 {
1000 bHalfTxtScale=true;
1001 }
1002 }
1003 }
1004
1005 if( g_curRomInfo.bEnableTxtLOD && vtxIndex == 1 && gRDP.otherMode.text_lod )
1006 {
1007 if( CRender::g_pRender->IsTexel1Enable() && CRender::g_pRender->m_pColorCombiner->m_pDecodedMux->isUsed(MUX_LODFRAC) )
1008 {
1009 ComputeLOD(openGL);
1010 }
1011 else
1012 {
1013 gRDP.LODFrac = 0;
1014 }
1015 }
1016
1017 VTX_DUMP(TRACE2(" DIF(%08X), SPE(%08X)", v.dcDiffuse, v.dcSpecular));
1018 VTX_DUMP(TRACE0(""));
1019}
1020
1021uint32 LightVert(XVECTOR4 & norm, int vidx)
1022{
1023 float fCosT;
1024
1025 // Do ambient
1026 register float r = gRSP.fAmbientLightR;
1027 register float g = gRSP.fAmbientLightG;
1028 register float b = gRSP.fAmbientLightB;
1029
1030 if( options.enableHackForGames != HACK_FOR_ZELDA_MM )
1031 {
1032 for (register unsigned int l=0; l < gRSPnumLights; l++)
1033 {
1034 fCosT = norm.x*gRSPlights[l].x + norm.y*gRSPlights[l].y + norm.z*gRSPlights[l].z;
1035
1036 if (fCosT > 0 )
1037 {
1038 r += gRSPlights[l].fr * fCosT;
1039 g += gRSPlights[l].fg * fCosT;
1040 b += gRSPlights[l].fb * fCosT;
1041 }
1042 }
1043 }
1044 else
1045 {
1046 XVECTOR4 v;
1047 bool transformed = false;
1048
1049 for (register unsigned int l=0; l < gRSPnumLights; l++)
1050 {
1051 if( gRSPlights[l].range == 0 )
1052 {
1053 // Regular directional light
1054 fCosT = norm.x*gRSPlights[l].x + norm.y*gRSPlights[l].y + norm.z*gRSPlights[l].z;
1055
1056 if (fCosT > 0 )
1057 {
1058 r += gRSPlights[l].fr * fCosT;
1059 g += gRSPlights[l].fg * fCosT;
1060 b += gRSPlights[l].fb * fCosT;
1061 }
1062 }
1063 else //if( (gRSPlights[l].col&0x00FFFFFF) != 0x00FFFFFF )
1064 {
1065 // Point light
1066 if( !transformed )
1067 {
1068 Vec3Transform(&v, (XVECTOR3*)&g_vtxNonTransformed[vidx], &gRSPmodelViewTop); // Convert to w=1
1069 transformed = true;
1070 }
1071
1072 XVECTOR3 dir(gRSPlights[l].x - v.x, gRSPlights[l].y - v.y, gRSPlights[l].z - v.z);
1073 //XVECTOR3 dir(v.x-gRSPlights[l].x, v.y-gRSPlights[l].y, v.z-gRSPlights[l].z);
1074 float d2 = sqrtf(dir.x*dir.x+dir.y*dir.y+dir.z*dir.z);
1075 dir.x /= d2;
1076 dir.y /= d2;
1077 dir.z /= d2;
1078
1079 fCosT = norm.x*dir.x + norm.y*dir.y + norm.z*dir.z;
1080
1081 if (fCosT > 0 )
1082 {
1083 //float f = d2/gRSPlights[l].range*50;
1084 float f = d2/15000*50;
1085 f = 1 - min(f,1);
1086 fCosT *= f*f;
1087
1088 r += gRSPlights[l].fr * fCosT;
1089 g += gRSPlights[l].fg * fCosT;
1090 b += gRSPlights[l].fb * fCosT;
1091 }
1092 }
1093 }
1094 }
1095
1096 if (r > 255) r = 255;
1097 if (g > 255) g = 255;
1098 if (b > 255) b = 255;
1099 return ((0xff000000)|(((uint32)r)<<16)|(((uint32)g)<<8)|((uint32)b));
1100}
1101
1102uint32 LightVertNew(XVECTOR4 & norm)
1103{
1104 float fCosT;
1105
1106 // Do ambient
1107 register float r = gRSP.fAmbientLightR;
1108 register float g = gRSP.fAmbientLightG;
1109 register float b = gRSP.fAmbientLightB;
1110
1111
1112 for (register unsigned int l=0; l < gRSPnumLights; l++)
1113 {
1114 fCosT = norm.x*gRSPlights[l].tx + norm.y*gRSPlights[l].ty + norm.z*gRSPlights[l].tz;
1115
1116 if (fCosT > 0 )
1117 {
1118 r += gRSPlights[l].fr * fCosT;
1119 g += gRSPlights[l].fg * fCosT;
1120 b += gRSPlights[l].fb * fCosT;
1121 }
1122 }
1123
1124 if (r > 255) r = 255;
1125 if (g > 255) g = 255;
1126 if (b > 255) b = 255;
1127 return ((0xff000000)|(((uint32)r)<<16)|(((uint32)g)<<8)|((uint32)b));
1128}
1129
1130
1131float zero = 0.0f;
1132float onef = 1.0f;
1133float fcosT;
1134
1135#if !defined(__GNUC__) && !defined(NO_ASM)
1136__declspec( naked ) uint32 __fastcall SSELightVert()
1137{
1138 __asm
1139 {
1140 movaps xmm3, DWORD PTR gRSP; // loading Ambient colors, xmm3 is the result color
1141 movaps xmm4, DWORD PTR [g_normal]; // xmm4 is the normal
1142
1143 mov ecx, 0;
1144loopback:
1145 cmp ecx, DWORD PTR gRSPnumLights;
1146 jae breakout;
1147 mov eax,ecx;
1148 imul eax,0x44;
1149 movups xmm5, DWORD PTR gRSPlights[eax]; // Light Dir
1150 movups xmm1, DWORD PTR gRSPlights[0x14][eax]; // Light color
1151 mulps xmm5, xmm4; // Lightdir * normals
1152
1153 movhlps xmm0,xmm5;
1154 addps xmm0,xmm5;
1155 shufps xmm5,xmm0,0x01;
1156 addps xmm0,xmm5;
1157
1158 comiss xmm0,zero;
1159 jc endloop
1160
1161 shufps xmm0,xmm0,0; // fcosT
1162 mulps xmm1,xmm0;
1163 addps xmm3,xmm1;
1164endloop:
1165 inc ecx;
1166 jmp loopback;
1167breakout:
1168
1169 movss xmm0,DWORD PTR real255;
1170 shufps xmm0,xmm0,0;
1171 minps xmm0,xmm3;
1172
1173 // Without using a memory
1174 cvtss2si eax,xmm0; // move the 1st uint32 to eax
1175 shl eax,10h;
1176 or eax,0FF000000h;
1177 shufps xmm0,xmm0,0E5h; // move the 2nd uint32 to the 1st uint32
1178 cvtss2si ecx,xmm0; // move the 1st uint32 to ecx
1179 shl ecx,8;
1180 or eax,ecx;
1181 shufps xmm0,xmm0,0E6h; // Move the 3rd uint32 to the 1st uint32
1182 cvtss2si ecx,xmm0;
1183 or eax,ecx;
1184
1185 ret;
1186 }
1187}
1188#elif defined(__GNUC__) && defined(__x86_64__) && !defined(NO_ASM)
1189uint32 SSELightVert(void)
1190{
1191 uint32 rval;
1192 float f255 = 255.0, fZero = 0.0;
1193
1194 asm volatile(" movaps %1, %%xmm3 \n" // xmm3 == gRSP.fAmbientLight{RGBA}
1195 " movaps %2, %%xmm4 \n" // xmm4 == g_normal.{xyz}
1196 " xor %%rcx, %%rcx \n"
1197 "0: \n"
1198 " cmpl %3, %%ecx \n"
1199 " jae 2f \n"
1200 " mov %%rcx, %%rax \n"
1201 " imul $0x44, %%rax, %%rax \n"
1202 " movups (%4,%%rax,), %%xmm5 \n" // xmm5 == gRSPlights[l].{xyzr}
1203 " movups 20(%4,%%rax,), %%xmm1 \n" // xmm1 == gRSPlights[l].{frfgfbfa}
1204 " mulps %%xmm4, %%xmm5 \n"
1205 " movhlps %%xmm5, %%xmm0 \n"
1206 " addps %%xmm5, %%xmm0 \n"
1207 " shufps $0x01, %%xmm0, %%xmm5 \n"
1208 " addps %%xmm5, %%xmm0 \n"
1209 " comiss %6, %%xmm0 \n"
1210 " jc 1f \n"
1211 " shufps $0x00, %%xmm0, %%xmm0 \n"
1212 " mulps %%xmm0, %%xmm1 \n"
1213 " addps %%xmm1, %%xmm3 \n"
1214 "1: \n"
1215 " inc %%rcx \n"
1216 " jmp 0b \n"
1217 "2: \n"
1218 " movss %5, %%xmm0 \n"
1219 " shufps $0x00, %%xmm0, %%xmm0 \n"
1220 " minps %%xmm3, %%xmm0 \n"
1221 " cvtss2si %%xmm0, %%eax \n"
1222 " shll $0x10, %%eax \n"
1223 " orl $0xff000000, %%eax \n"
1224 " shufps $0xe5, %%xmm0, %%xmm0 \n"
1225 " cvtss2si %%xmm0, %%ecx \n"
1226 " shll $8, %%ecx \n"
1227 " orl %%ecx, %%eax \n"
1228 " shufps $0xe6, %%xmm0, %%xmm0 \n"
1229 " cvtss2si %%xmm0, %%ecx \n"
1230 " orl %%ecx, %%eax \n"
1231 : "=&a"(rval)
1232 : "m"(gRSP), "m"(g_normal), "m"(gRSPnumLights), "r"(gRSPlights), "m"(f255), "m"(fZero)
1233 : "%rcx", "memory", "cc", "%xmm0", "%xmm1", "%xmm3", "%xmm4", "%xmm5"
1234 );
1235 return rval;
1236}
1237#elif !defined(NO_ASM) // 32-bit GCC assumed
1238uint32 SSELightVert(void)
1239{
1240 uint32 rval;
1241 float f255 = 255.0, fZero = 0.0;
1242
1243 asm volatile(" movaps %1, %%xmm3 \n"
1244 " movaps %2, %%xmm4 \n"
1245 " xor %%ecx, %%ecx \n"
1246 "0: \n"
1247 " cmpl %3, %%ecx \n"
1248 " jae 2f \n"
1249 " mov %%ecx, %%eax \n"
1250 " imul $0x44, %%eax, %%eax \n"
1251 " movups (%4,%%eax,), %%xmm5 \n"
1252 " movups 20(%4,%%eax,), %%xmm1 \n"
1253 " mulps %%xmm4, %%xmm5 \n"
1254 " movhlps %%xmm5, %%xmm0 \n"
1255 " addps %%xmm5, %%xmm0 \n"
1256 " shufps $0x01, %%xmm0, %%xmm5 \n"
1257 " addps %%xmm5, %%xmm0 \n"
1258 " comiss %6, %%xmm0 \n"
1259 " jc 1f \n"
1260 " shufps $0x00, %%xmm0, %%xmm0 \n"
1261 " mulps %%xmm0, %%xmm1 \n"
1262 " addps %%xmm1, %%xmm3 \n"
1263 "1: \n"
1264 " inc %%ecx \n"
1265 " jmp 0b \n"
1266 "2: \n"
1267 " movss %5, %%xmm0 \n"
1268 " shufps $0x00, %%xmm0, %%xmm0 \n"
1269 " minps %%xmm3, %%xmm0 \n"
1270 " cvtss2si %%xmm0, %%eax \n"
1271 " shll $0x10, %%eax \n"
1272 " orl $0xff000000, %%eax \n"
1273 " shufps $0xe5, %%xmm0, %%xmm0 \n"
1274 " cvtss2si %%xmm0, %%ecx \n"
1275 " shll $8, %%ecx \n"
1276 " orl %%ecx, %%eax \n"
1277 " shufps $0xe6, %%xmm0, %%xmm0 \n"
1278 " cvtss2si %%xmm0, %%ecx \n"
1279 " orl %%ecx, %%eax \n"
1280 : "=&a"(rval)
1281 : "m"(gRSP), "m"(g_normal), "m"(gRSPnumLights), "r"(gRSPlights), "m"(f255), "m"(fZero)
1282 : "%rcx", "memory", "cc", "%xmm0", "%xmm1", "%xmm3", "%xmm4", "%xmm5"
1283 );
1284 return rval;
1285}
1286#endif
1287
1288inline void ReplaceAlphaWithFogFactor(int i)
1289{
1290 if( gRDP.geometryMode & G_FOG )
1291 {
1292 // Use fog factor to replace vertex alpha
1293 if( g_vecProjected[i].z > 1 )
1294 *(((uint8*)&(g_dwVtxDifColor[i]))+3) = 0xFF;
1295 if( g_vecProjected[i].z < 0 )
1296 *(((uint8*)&(g_dwVtxDifColor[i]))+3) = 0;
1297 else
1298 *(((uint8*)&(g_dwVtxDifColor[i]))+3) = (uint8)(g_vecProjected[i].z*255);
1299 }
1300}
1301
1302
1303// Bits
1304// +-+-+-
1305// xxyyzz
1306#define Z_NEG 0x01
1307#define Z_POS 0x02
1308#define Y_NEG 0x04
1309#define Y_POS 0x08
1310#define X_NEG 0x10
1311#define X_POS 0x20
1312
1313// Assumes dwAddr has already been checked!
1314// Don't inline - it's too big with the transform macros
1315
1316#if !defined(NO_ASM)
1317void ProcessVertexDataSSE(uint32 dwAddr, uint32 dwV0, uint32 dwNum)
1318{
1319 UpdateCombinedMatrix();
1320
1321 // This function is called upon SPvertex
1322 // - do vertex matrix transform
1323 // - do vertex lighting
1324 // - do texture cooridinate transform if needed
1325 // - calculate normal vector
1326
1327 // Output: - g_vecProjected[i] -> transformed vertex x,y,z
1328 // - g_vecProjected[i].w -> saved vertex 1/w
1329 // - g_dwVtxFlags[i] -> flags
1330 // - g_dwVtxDifColor[i] -> vertex color
1331 // - g_fVtxTxtCoords[i] -> vertex texture cooridinates
1332
1333 FiddledVtx * pVtxBase = (FiddledVtx*)(g_pRDRAMu8 + dwAddr);
1334 g_pVtxBase = pVtxBase;
1335
1336 for (uint32 i = dwV0; i < dwV0 + dwNum; i++)
1337 {
1338 SP_Timing(RSP_GBI0_Vtx);
1339
1340 FiddledVtx & vert = pVtxBase[i - dwV0];
1341
1342 g_vtxNonTransformed[i].x = (float)vert.x;
1343 g_vtxNonTransformed[i].y = (float)vert.y;
1344 g_vtxNonTransformed[i].z = (float)vert.z;
1345
1346 SSEVec3Transform(i);
1347
1348 if( gRSP.bFogEnabled )
1349 {
1350 g_fFogCoord[i] = g_vecProjected[i].z;
1351 if( g_vecProjected[i].w < 0 || g_vecProjected[i].z < 0 || g_fFogCoord[i] < gRSPfFogMin )
1352 g_fFogCoord[i] = gRSPfFogMin;
1353 }
1354
1355 ReplaceAlphaWithFogFactor(i);
1356
1357
1358 VTX_DUMP(
1359 {
1360 uint32 *dat = (uint32*)(&vert);
1361 DebuggerAppendMsg("vtx %d: %08X %08X %08X %08X", i, dat[0],dat[1],dat[2],dat[3]);
1362 DebuggerAppendMsg(" : %f, %f, %f, %f",
1363 g_vtxTransformed[i].x,g_vtxTransformed[i].y,g_vtxTransformed[i].z,g_vtxTransformed[i].w);
1364 DebuggerAppendMsg(" : %f, %f, %f, %f",
1365 g_vecProjected[i].x,g_vecProjected[i].y,g_vecProjected[i].z,g_vecProjected[i].w);
1366 });
1367
1368 RSP_Vtx_Clipping(i);
1369
1370 if( gRSP.bLightingEnable )
1371 {
1372 g_normal.x = (float)vert.norma.nx;
1373 g_normal.y = (float)vert.norma.ny;
1374 g_normal.z = (float)vert.norma.nz;
1375
1376 SSEVec3TransformNormal();
1377 if( options.enableHackForGames != HACK_FOR_ZELDA_MM )
1378 g_dwVtxDifColor[i] = SSELightVert();
1379 else
1380 g_dwVtxDifColor[i] = LightVert(g_normal, i);
1381 *(((uint8*)&(g_dwVtxDifColor[i]))+3) = vert.rgba.a; // still use alpha from the vertex
1382 }
1383 else
1384 {
1385 if( (gRDP.geometryMode & G_SHADE) == 0 && gRSP.ucode < 5 ) //Shade is disabled
1386 {
1387 //FLAT shade
1388 g_dwVtxDifColor[i] = gRDP.primitiveColor;
1389 }
1390 else
1391 {
1392 register IColor &color = *(IColor*)&g_dwVtxDifColor[i];
1393 color.b = vert.rgba.r;
1394 color.g = vert.rgba.g;
1395 color.r = vert.rgba.b;
1396 color.a = vert.rgba.a;
1397 }
1398 }
1399
1400 if( options.bWinFrameMode )
1401 {
1402 g_dwVtxDifColor[i] = COLOR_RGBA(vert.rgba.r, vert.rgba.g, vert.rgba.b, vert.rgba.a);
1403 }
1404
1405 // Update texture coords n.b. need to divide tu/tv by bogus scale on addition to buffer
1406
1407 // If the vert is already lit, then there is no normal (and hence we
1408 // can't generate tex coord)
1409 if (gRSP.bTextureGen && gRSP.bLightingEnable )
1410 {
1411 TexGen(g_fVtxTxtCoords[i].x, g_fVtxTxtCoords[i].y);
1412 }
1413 else
1414 {
1415 g_fVtxTxtCoords[i].x = (float)vert.tu;
1416 g_fVtxTxtCoords[i].y = (float)vert.tv;
1417 }
1418 }
1419
1420 VTX_DUMP(TRACE2("Setting Vertexes: %d - %d\n", dwV0, dwV0+dwNum-1));
1421 DEBUGGER_PAUSE_AND_DUMP(NEXT_VERTEX_CMD,{TRACE0("Paused at Vertex Cmd");});
1422}
1423#endif
1424
1425void ProcessVertexDataNoSSE(uint32 dwAddr, uint32 dwV0, uint32 dwNum)
1426{
1427
1428 UpdateCombinedMatrix();
1429
1430 // This function is called upon SPvertex
1431 // - do vertex matrix transform
1432 // - do vertex lighting
1433 // - do texture cooridinate transform if needed
1434 // - calculate normal vector
1435
1436 // Output: - g_vecProjected[i] -> transformed vertex x,y,z
1437 // - g_vecProjected[i].w -> saved vertex 1/w
1438 // - g_dwVtxFlags[i] -> flags
1439 // - g_dwVtxDifColor[i] -> vertex color
1440 // - g_fVtxTxtCoords[i] -> vertex texture cooridinates
1441
1442 FiddledVtx * pVtxBase = (FiddledVtx*)(g_pRDRAMu8 + dwAddr);
1443 g_pVtxBase = pVtxBase;
1444
1445 for (uint32 i = dwV0; i < dwV0 + dwNum; i++)
1446 {
1447 SP_Timing(RSP_GBI0_Vtx);
1448
1449 FiddledVtx & vert = pVtxBase[i - dwV0];
1450
1451 g_vtxNonTransformed[i].x = (float)vert.x;
1452 g_vtxNonTransformed[i].y = (float)vert.y;
1453 g_vtxNonTransformed[i].z = (float)vert.z;
1454
1455 Vec3Transform(&g_vtxTransformed[i], (XVECTOR3*)&g_vtxNonTransformed[i], &gRSPworldProject); // Convert to w=1
1456
1457 g_vecProjected[i].w = 1.0f / g_vtxTransformed[i].w;
1458 g_vecProjected[i].x = g_vtxTransformed[i].x * g_vecProjected[i].w;
1459 g_vecProjected[i].y = g_vtxTransformed[i].y * g_vecProjected[i].w;
1460 if ((g_curRomInfo.bPrimaryDepthHack || options.enableHackForGames == HACK_FOR_NASCAR ) && gRDP.otherMode.depth_source )
1461 {
1462 g_vecProjected[i].z = gRDP.fPrimitiveDepth;
1463 g_vtxTransformed[i].z = gRDP.fPrimitiveDepth*g_vtxTransformed[i].w;
1464 }
1465 else
1466 {
1467 g_vecProjected[i].z = g_vtxTransformed[i].z * g_vecProjected[i].w;
1468 }
1469
1470 if( gRSP.bFogEnabled )
1471 {
1472 g_fFogCoord[i] = g_vecProjected[i].z;
1473 if( g_vecProjected[i].w < 0 || g_vecProjected[i].z < 0 || g_fFogCoord[i] < gRSPfFogMin )
1474 g_fFogCoord[i] = gRSPfFogMin;
1475 }
1476
1477 VTX_DUMP(
1478 {
1479 uint32 *dat = (uint32*)(&vert);
1480 DebuggerAppendMsg("vtx %d: %08X %08X %08X %08X", i, dat[0],dat[1],dat[2],dat[3]);
1481 DebuggerAppendMsg(" : %f, %f, %f, %f",
1482 g_vtxTransformed[i].x,g_vtxTransformed[i].y,g_vtxTransformed[i].z,g_vtxTransformed[i].w);
1483 DebuggerAppendMsg(" : %f, %f, %f, %f",
1484 g_vecProjected[i].x,g_vecProjected[i].y,g_vecProjected[i].z,g_vecProjected[i].w);
1485 });
1486
1487 RSP_Vtx_Clipping(i);
1488
1489 if( gRSP.bLightingEnable )
1490 {
1491 g_normal.x = (float)vert.norma.nx;
1492 g_normal.y = (float)vert.norma.ny;
1493 g_normal.z = (float)vert.norma.nz;
1494
1495 Vec3TransformNormal(g_normal, gRSPmodelViewTop);
1496 g_dwVtxDifColor[i] = LightVert(g_normal, i);
1497 *(((uint8*)&(g_dwVtxDifColor[i]))+3) = vert.rgba.a; // still use alpha from the vertex
1498 }
1499 else
1500 {
1501 if( (gRDP.geometryMode & G_SHADE) == 0 && gRSP.ucode < 5 ) //Shade is disabled
1502 {
1503 //FLAT shade
1504 g_dwVtxDifColor[i] = gRDP.primitiveColor;
1505 }
1506 else
1507 {
1508 register IColor &color = *(IColor*)&g_dwVtxDifColor[i];
1509 color.b = vert.rgba.r;
1510 color.g = vert.rgba.g;
1511 color.r = vert.rgba.b;
1512 color.a = vert.rgba.a;
1513 }
1514 }
1515
1516 if( options.bWinFrameMode )
1517 {
1518 g_dwVtxDifColor[i] = COLOR_RGBA(vert.rgba.r, vert.rgba.g, vert.rgba.b, vert.rgba.a);
1519 }
1520
1521 ReplaceAlphaWithFogFactor(i);
1522
1523 // Update texture coords n.b. need to divide tu/tv by bogus scale on addition to buffer
1524
1525 // If the vert is already lit, then there is no normal (and hence we
1526 // can't generate tex coord)
1527 if (gRSP.bTextureGen && gRSP.bLightingEnable )
1528 {
1529 TexGen(g_fVtxTxtCoords[i].x, g_fVtxTxtCoords[i].y);
1530 }
1531 else
1532 {
1533 g_fVtxTxtCoords[i].x = (float)vert.tu;
1534 g_fVtxTxtCoords[i].y = (float)vert.tv;
1535 }
1536 }
1537
1538 VTX_DUMP(TRACE2("Setting Vertexes: %d - %d\n", dwV0, dwV0+dwNum-1));
1539 DEBUGGER_PAUSE_AND_DUMP(NEXT_VERTEX_CMD,{TRACE0("Paused at Vertex Cmd");});
1540}
1541
61b9f2df 1542/* NEON code */
1543
1544#include "RenderBase_neon.h"
1545
0764a475 1546extern "C" void pv_neon(XVECTOR4 *g_vtxTransformed, XVECTOR4 *g_vecProjected,
1547 uint32 *g_dwVtxDifColor, VECTOR2 *g_fVtxTxtCoords,
1548 float *g_fFogCoord, uint32 *g_clipFlag2,
61b9f2df 1549 uint32 dwNum, int neon_state,
1550 const FiddledVtx *vtx,
0764a475 1551 const Light *gRSPlights, const float *fRSPAmbientLightRGBA,
1552 const XMATRIX *gRSPworldProject, const XMATRIX *gRSPmodelViewTop,
61b9f2df 1553 uint32 gRSPnumLights, float gRSPfFogMin,
1554 uint32 primitiveColor, uint32 primitiveColor_);
1555
3db2a2f9 1556extern "C" int tv_direction(const XVECTOR4 *v0, const XVECTOR4 *v1, const XVECTOR4 *v2);
1557
0764a475 1558void ProcessVertexDataNEON(uint32 dwAddr, uint32 dwV0, uint32 dwNum)
1559{
1560 if (gRSP.bTextureGen && gRSP.bLightingEnable) {
1561 ProcessVertexDataNoSSE(dwAddr, dwV0,dwNum);
1562 return;
1563 }
1564
1565 // assumtions:
1566 // - g_clipFlag is not used at all
61b9f2df 1567 // - g_fFogCoord is not used at all
0764a475 1568 // - g_vtxNonTransformed is not used after ProcessVertexData*() returns
1569 // - g_normal - same
1570
0764a475 1571 int neon_state = 0;
1572 if ( gRSP.bLightingEnable )
1573 neon_state |= PV_NEON_ENABLE_LIGHT;
1574 if ( (gRDP.geometryMode & G_SHADE) || gRSP.ucode >= 5 )
1575 neon_state |= PV_NEON_ENABLE_SHADE;
1576 if ( gRSP.bFogEnabled )
1577 neon_state |= PV_NEON_ENABLE_FOG;
1578 if ( gRDP.geometryMode & G_FOG )
1579 neon_state |= PV_NEON_FOG_ALPHA;
1580
1581 uint32 i;
1582
1583 UpdateCombinedMatrix();
1584
1585 // This function is called upon SPvertex
1586 // - do vertex matrix transform
1587 // - do vertex lighting
1588 // - do texture cooridinate transform if needed
1589 // - calculate normal vector
1590
1591 // Output: - g_vecProjected[i] -> transformed vertex x,y,z
1592 // - g_vecProjected[i].w -> saved vertex 1/w
1593 // - g_vtxTransformed[i]
1594 // - g_dwVtxDifColor[i] -> vertex color
1595 // - g_fVtxTxtCoords[i] -> vertex texture cooridinates
61b9f2df 1596 // - g_fFogCoord[i] -> unused
0764a475 1597 // - g_clipFlag2[i]
1598
1599 const FiddledVtx * pVtxBase = (const FiddledVtx*)(g_pRDRAMu8 + dwAddr);
1600 g_pVtxBase = (FiddledVtx *)pVtxBase;
1601
61b9f2df 1602 gRSPmodelViewTop._14 = gRSPmodelViewTop._24 =
1603 gRSPmodelViewTop._34 = 0;
1604
0764a475 1605 // SP_Timing(RSP_GBI0_Vtx);
1606 status.SPCycleCount += Timing_RSP_GBI0_Vtx * dwNum;
1607
61b9f2df 1608#if 1
1609 i = dwV0;
1610 pv_neon(&g_vtxTransformed[i], &g_vecProjected[i],
1611 &g_dwVtxDifColor[i], &g_fVtxTxtCoords[i],
1612 &g_fFogCoord[i], &g_clipFlag2[i],
1613 dwNum, neon_state, &pVtxBase[i - dwV0],
1614 gRSPlights, gRSP.fAmbientColors,
1615 &gRSPworldProject, &gRSPmodelViewTop,
1616 gRSPnumLights, gRSPfFogMin,
1617 gRDP.primitiveColor, gRDP.primitiveColor);
1618#else
0764a475 1619 for (i = dwV0; i < dwV0 + dwNum; i++)
1620 {
1621 const FiddledVtx & vert = pVtxBase[i - dwV0];
1622 XVECTOR3 vtx_raw; // was g_vtxNonTransformed
1623
1624 vtx_raw.x = (float)vert.x;
1625 vtx_raw.y = (float)vert.y;
1626 vtx_raw.z = (float)vert.z;
1627
1628 Vec3Transform(&g_vtxTransformed[i], &vtx_raw, &gRSPworldProject); // Convert to w=1
1629
1630 g_vecProjected[i].w = 1.0f / g_vtxTransformed[i].w;
1631 g_vecProjected[i].x = g_vtxTransformed[i].x * g_vecProjected[i].w;
1632 g_vecProjected[i].y = g_vtxTransformed[i].y * g_vecProjected[i].w;
1633 g_vecProjected[i].z = g_vtxTransformed[i].z * g_vecProjected[i].w;
1634
0764a475 1635 // RSP_Vtx_Clipping(i);
1636 g_clipFlag2[i] = 0;
1637 if( g_vecProjected[i].w > 0 )
1638 {
1639 if( g_vecProjected[i].x > 1 ) g_clipFlag2[i] |= X_CLIP_MAX;
1640 if( g_vecProjected[i].x < -1 ) g_clipFlag2[i] |= X_CLIP_MIN;
1641 if( g_vecProjected[i].y > 1 ) g_clipFlag2[i] |= Y_CLIP_MAX;
1642 if( g_vecProjected[i].y < -1 ) g_clipFlag2[i] |= Y_CLIP_MIN;
1643 }
1644
1645 if( neon_state & PV_NEON_ENABLE_LIGHT )
1646 {
1647 XVECTOR3 normal; // was g_normal
1648 float r, g, b;
1649
1650 normal.x = (float)vert.norma.nx;
1651 normal.y = (float)vert.norma.ny;
1652 normal.z = (float)vert.norma.nz;
1653
1654 Vec3TransformNormal(normal, gRSPmodelViewTop);
1655
1656 r = gRSP.fAmbientLightR;
1657 g = gRSP.fAmbientLightG;
1658 b = gRSP.fAmbientLightB;
1659
1660 for (unsigned int l=0; l < gRSPnumLights; l++)
1661 {
1662 float fCosT = normal.x * gRSPlights[l].x + normal.y * gRSPlights[l].y + normal.z * gRSPlights[l].z;
1663
1664 if (fCosT > 0 )
1665 {
1666 r += gRSPlights[l].fr * fCosT;
1667 g += gRSPlights[l].fg * fCosT;
1668 b += gRSPlights[l].fb * fCosT;
1669 }
1670 }
1671 if (r > 255) r = 255;
1672 if (g > 255) g = 255;
1673 if (b > 255) b = 255;
1674 g_dwVtxDifColor[i] = ((vert.rgba.a<<24)|(((uint32)r)<<16)|(((uint32)g)<<8)|((uint32)b));
1675 }
1676 else if( neon_state & PV_NEON_ENABLE_SHADE )
1677 {
1678 IColor &color = *(IColor*)&g_dwVtxDifColor[i];
1679 color.b = vert.rgba.r;
1680 color.g = vert.rgba.g;
1681 color.r = vert.rgba.b;
1682 color.a = vert.rgba.a;
1683 }
61b9f2df 1684 else
1685 g_dwVtxDifColor[i] = gRDP.primitiveColor; // FLAT shade
0764a475 1686
1687 // ReplaceAlphaWithFogFactor(i);
1688 if( neon_state & PV_NEON_FOG_ALPHA )
1689 {
1690 // Use fog factor to replace vertex alpha
1691 if( g_vecProjected[i].z > 1 )
1692 *(((uint8*)&(g_dwVtxDifColor[i]))+3) = 0xFF;
61b9f2df 1693 // missing 'else' in original code??
1694 else if( g_vecProjected[i].z < 0 )
0764a475 1695 *(((uint8*)&(g_dwVtxDifColor[i]))+3) = 0;
1696 else
1697 *(((uint8*)&(g_dwVtxDifColor[i]))+3) = (uint8)(g_vecProjected[i].z*255);
1698 }
1699
1700 g_fVtxTxtCoords[i].x = (float)vert.tu;
1701 g_fVtxTxtCoords[i].y = (float)vert.tv;
61b9f2df 1702 }
1703#endif
0764a475 1704}
1705
292f9317 1706bool PrepareTriangle(uint32 dwV0, uint32 dwV1, uint32 dwV2)
1707{
1708 if( status.isVertexShaderEnabled || status.bUseHW_T_L )
1709 {
1710 g_vtxIndex[gRSP.numVertices++] = dwV0;
1711 g_vtxIndex[gRSP.numVertices++] = dwV1;
1712 g_vtxIndex[gRSP.numVertices++] = dwV2;
1713 status.dwNumTrisRendered++;
1714 gRSP.maxVertexID = max(gRSP.maxVertexID,max(dwV0,max(dwV1,dwV2)));
1715 }
1716 else
1717 {
1718 SP_Timing(SP_Each_Triangle);
1719
1720 bool textureFlag = (CRender::g_pRender->IsTextureEnabled() || gRSP.ucode == 6 );
1721 bool openGL = CDeviceBuilder::m_deviceGeneralType == OGL_DEVICE;
1722
1723 InitVertex(dwV0, gRSP.numVertices, textureFlag, openGL);
1724 InitVertex(dwV1, gRSP.numVertices+1, textureFlag, openGL);
1725 InitVertex(dwV2, gRSP.numVertices+2, textureFlag, openGL);
1726
1727 gRSP.numVertices += 3;
1728 status.dwNumTrisRendered++;
1729 }
1730
1731 return true;
1732}
1733
1734
1735
1736// Returns TRUE if it thinks the triangle is visible
1737// Returns FALSE if it is clipped
1738bool IsTriangleVisible(uint32 dwV0, uint32 dwV1, uint32 dwV2)
1739{
1740 //return true; //fix me
1741
1742 if( status.isVertexShaderEnabled || status.bUseHW_T_L ) return true; // We won't have access to transformed vertex data
1743
1744 DEBUGGER_ONLY_IF( (!debuggerEnableTestTris || !debuggerEnableCullFace), {return TRUE;});
1745
1746#ifdef DEBUGGER
1747 // Check vertices are valid!
1748 if (dwV0 >= MAX_VERTS || dwV1 >= MAX_VERTS || dwV2 >= MAX_VERTS)
1749 return false;
1750#endif
1751
1752 // Here we AND all the flags. If any of the bits is set for all
1753 // 3 vertices, it means that all three x, y or z lie outside of
1754 // the current viewing volume.
1755 // Currently disabled - still seems a bit dodgy
1756 if ((gRSP.bCullFront || gRSP.bCullBack) && gRDP.otherMode.zmode != 3)
1757 {
1758 XVECTOR4 & v0 = g_vecProjected[dwV0];
1759 XVECTOR4 & v1 = g_vecProjected[dwV1];
1760 XVECTOR4 & v2 = g_vecProjected[dwV2];
1761
1762 // Only try to clip if the tri is onscreen. For some reason, this
1763 // method doesnt' work well when the z value is outside of screenspace
1764 //if (v0.z < 1 && v1.z < 1 && v2.z < 1)
1765 {
3db2a2f9 1766#ifndef __ARM_NEON__
292f9317 1767 float V1 = v2.x - v0.x;
1768 float V2 = v2.y - v0.y;
1769
1770 float W1 = v2.x - v1.x;
1771 float W2 = v2.y - v1.y;
1772
1773 float fDirection = (V1 * W2) - (V2 * W1);
1774 fDirection = fDirection * v1.w * v2.w * v0.w;
1775 //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;
3db2a2f9 1776#else
1777 // really returns float, but we only need sign
1778 int fDirection = tv_direction(&v0, &v1, &v2);
1779#endif
292f9317 1780
1781 if (fDirection < 0 && gRSP.bCullBack)
1782 {
1783 status.dwNumTrisClipped++;
1784 return false;
1785 }
1786 else if (fDirection > 0 && gRSP.bCullFront)
1787 {
1788 status.dwNumTrisClipped++;
1789 return false;
1790 }
1791 }
1792 }
1793
1794#ifdef ENABLE_CLIP_TRI
1795 //if( gRSP.bRejectVtx && (g_clipFlag[dwV0]|g_clipFlag[dwV1]|g_clipFlag[dwV2]) )
1796 // return;
1797 if( g_clipFlag2[dwV0]&g_clipFlag2[dwV1]&g_clipFlag2[dwV2] )
1798 {
1799 //DebuggerAppendMsg("Clipped");
1800 return false;
1801 }
1802#endif
1803
1804 return true;
1805}
1806
1807
1808void SetPrimitiveColor(uint32 dwCol, uint32 LODMin, uint32 LODFrac)
1809{
1810 gRDP.colorsAreReloaded = true;
1811 gRDP.primitiveColor = dwCol;
1812 gRDP.primLODMin = LODMin;
1813 gRDP.primLODFrac = LODFrac;
1814 if( gRDP.primLODFrac < gRDP.primLODMin )
1815 {
1816 gRDP.primLODFrac = gRDP.primLODMin;
1817 }
1818
1819 gRDP.fvPrimitiveColor[0] = ((dwCol>>16)&0xFF)/255.0f; //r
1820 gRDP.fvPrimitiveColor[1] = ((dwCol>>8)&0xFF)/255.0f; //g
1821 gRDP.fvPrimitiveColor[2] = ((dwCol)&0xFF)/255.0f; //b
1822 gRDP.fvPrimitiveColor[3] = ((dwCol>>24)&0xFF)/255.0f; //a
1823}
1824
1825void SetPrimitiveDepth(uint32 z, uint32 dwDZ)
1826{
1827 gRDP.primitiveDepth = z & 0x7FFF;
1828 gRDP.fPrimitiveDepth = (float)(gRDP.primitiveDepth)/(float)0x8000;
1829
1830 //gRDP.fPrimitiveDepth = gRDP.fPrimitiveDepth*2-1;
1831 /*
1832 z=0xFFFF -> 1 the farest
1833 z=0 -> -1 the nearest
1834 */
1835
1836 //how to use dwDZ?
1837
1838#ifdef DEBUGGER
1839 if( (pauseAtNext && (eventToPause == NEXT_VERTEX_CMD || eventToPause == NEXT_FLUSH_TRI )) )//&& logTriangles )
1840 {
1841 DebuggerAppendMsg("Set prim Depth: %f, (%08X, %08X)", gRDP.fPrimitiveDepth, z, dwDZ);
1842 }
1843#endif
1844}
1845
1846void SetVertexXYZ(uint32 vertex, float x, float y, float z)
1847{
1848 g_vecProjected[vertex].x = x;
1849 g_vecProjected[vertex].y = y;
1850 g_vecProjected[vertex].z = z;
1851
1852 g_vtxTransformed[vertex].x = x*g_vtxTransformed[vertex].w;
1853 g_vtxTransformed[vertex].y = y*g_vtxTransformed[vertex].w;
1854 g_vtxTransformed[vertex].z = z*g_vtxTransformed[vertex].w;
1855}
1856
1857void ModifyVertexInfo(uint32 where, uint32 vertex, uint32 val)
1858{
1859 switch (where)
1860 {
1861 case RSP_MV_WORD_OFFSET_POINT_RGBA: // Modify RGBA
1862 {
1863 uint32 r = (val>>24)&0xFF;
1864 uint32 g = (val>>16)&0xFF;
1865 uint32 b = (val>>8)&0xFF;
1866 uint32 a = val&0xFF;
1867 g_dwVtxDifColor[vertex] = COLOR_RGBA(r, g, b, a);
1868 LOG_UCODE("Modify vert %d color, 0x%08x", vertex, g_dwVtxDifColor[vertex]);
1869 }
1870 break;
1871 case RSP_MV_WORD_OFFSET_POINT_XYSCREEN: // Modify X,Y
1872 {
1873 uint16 nX = (uint16)(val>>16);
1874 short x = *((short*)&nX);
1875 x /= 4;
1876
1877 uint16 nY = (uint16)(val&0xFFFF);
1878 short y = *((short*)&nY);
1879 y /= 4;
1880
1881 // Should do viewport transform.
1882
1883
1884 x -= windowSetting.uViWidth/2;
1885 y = windowSetting.uViHeight/2-y;
1886
1887 if( options.bEnableHacks && ((*g_GraphicsInfo.VI_X_SCALE_REG)&0xF) != 0 )
1888 {
1889 // Tarzan
1890 // I don't know why Tarzan is different
1891 SetVertexXYZ(vertex, x/windowSetting.fViWidth, y/windowSetting.fViHeight, g_vecProjected[vertex].z);
1892 }
1893 else
1894 {
1895 // Toy Story 2 and other games
1896 SetVertexXYZ(vertex, x*2/windowSetting.fViWidth, y*2/windowSetting.fViHeight, g_vecProjected[vertex].z);
1897 }
1898
1899 LOG_UCODE("Modify vert %d: x=%d, y=%d", vertex, x, y);
1900 VTX_DUMP(TRACE3("Modify vert %d: (%d,%d)", vertex, x, y));
1901 }
1902 break;
1903 case RSP_MV_WORD_OFFSET_POINT_ZSCREEN: // Modify C
1904 {
1905 int z = val>>16;
1906
1907 SetVertexXYZ(vertex, g_vecProjected[vertex].x, g_vecProjected[vertex].y, (((float)z/0x03FF)+0.5f)/2.0f );
1908 LOG_UCODE("Modify vert %d: z=%d", vertex, z);
1909 VTX_DUMP(TRACE2("Modify vert %d: z=%d", vertex, z));
1910 }
1911 break;
1912 case RSP_MV_WORD_OFFSET_POINT_ST: // Texture
1913 {
1914 short tu = short(val>>16);
1915 short tv = short(val & 0xFFFF);
1916 float ftu = tu / 32.0f;
1917 float ftv = tv / 32.0f;
1918 LOG_UCODE(" Setting vertex %d tu/tv to %f, %f", vertex, (float)tu, (float)tv);
1919 CRender::g_pRender->SetVtxTextureCoord(vertex, ftu/gRSP.fTexScaleX, ftv/gRSP.fTexScaleY);
1920 }
1921 break;
1922 }
1923 DEBUGGER_PAUSE_AND_DUMP(NEXT_VERTEX_CMD,{TRACE0("Paused at ModVertex Cmd");});
1924}
1925
1926void ProcessVertexDataDKR(uint32 dwAddr, uint32 dwV0, uint32 dwNum)
1927{
1928 UpdateCombinedMatrix();
1929
1930 long long pVtxBase = (long long) (g_pRDRAMu8 + dwAddr);
1931 g_pVtxBase = (FiddledVtx*)pVtxBase;
1932
1933 Matrix &matWorldProject = gRSP.DKRMatrixes[gRSP.DKRCMatrixIndex];
1934
1935 int nOff;
1936
1937 bool addbase=false;
1938 if ((!gRSP.DKRBillBoard) || (gRSP.DKRCMatrixIndex != 2) )
1939 addbase = false;
1940 else
1941 addbase = true;
1942
1943 if( addbase && gRSP.DKRVtxCount == 0 && dwNum > 1 )
1944 {
1945 gRSP.DKRVtxCount++;
1946 }
1947
1948 LOG_UCODE(" ProcessVertexDataDKR, CMatrix = %d, Add base=%s", gRSP.DKRCMatrixIndex, gRSP.DKRBillBoard?"true":"false");
1949 VTX_DUMP(TRACE2("DKR Setting Vertexes\nCMatrix = %d, Add base=%s", gRSP.DKRCMatrixIndex, gRSP.DKRBillBoard?"true":"false"));
1950
1951 nOff = 0;
1952 uint32 end = dwV0 + dwNum;
1953 for (uint32 i = dwV0; i < end; i++)
1954 {
1955 XVECTOR3 w;
1956
1957 g_vtxNonTransformed[i].x = (float)*(short*)((pVtxBase+nOff + 0) ^ 2);
1958 g_vtxNonTransformed[i].y = (float)*(short*)((pVtxBase+nOff + 2) ^ 2);
1959 g_vtxNonTransformed[i].z = (float)*(short*)((pVtxBase+nOff + 4) ^ 2);
1960
1961 //if( status.isSSEEnabled )
1962 // SSEVec3TransformDKR(g_vtxTransformed[i], g_vtxNonTransformed[i]);
1963 //else
1964 Vec3Transform(&g_vtxTransformed[i], (XVECTOR3*)&g_vtxNonTransformed[i], &matWorldProject); // Convert to w=1
1965
1966 if( gRSP.DKRVtxCount == 0 && dwNum==1 )
1967 {
1968 gRSP.DKRBaseVec.x = g_vtxTransformed[i].x;
1969 gRSP.DKRBaseVec.y = g_vtxTransformed[i].y;
1970 gRSP.DKRBaseVec.z = g_vtxTransformed[i].z;
1971 gRSP.DKRBaseVec.w = g_vtxTransformed[i].w;
1972 }
1973 else if( addbase )
1974 {
1975 g_vtxTransformed[i].x += gRSP.DKRBaseVec.x;
1976 g_vtxTransformed[i].y += gRSP.DKRBaseVec.y;
1977 g_vtxTransformed[i].z += gRSP.DKRBaseVec.z;
1978 g_vtxTransformed[i].w = gRSP.DKRBaseVec.w;
1979 }
1980
1981 g_vecProjected[i].w = 1.0f / g_vtxTransformed[i].w;
1982 g_vecProjected[i].x = g_vtxTransformed[i].x * g_vecProjected[i].w;
1983 g_vecProjected[i].y = g_vtxTransformed[i].y * g_vecProjected[i].w;
1984 g_vecProjected[i].z = g_vtxTransformed[i].z * g_vecProjected[i].w;
1985
1986 gRSP.DKRVtxCount++;
1987
1988 VTX_DUMP(TRACE5("vtx %d: %f, %f, %f, %f", i,
1989 g_vtxTransformed[i].x,g_vtxTransformed[i].y,g_vtxTransformed[i].z,g_vtxTransformed[i].w));
1990
1991 if( gRSP.bFogEnabled )
1992 {
1993 g_fFogCoord[i] = g_vecProjected[i].z;
1994 if( g_vecProjected[i].w < 0 || g_vecProjected[i].z < 0 || g_fFogCoord[i] < gRSPfFogMin )
1995 g_fFogCoord[i] = gRSPfFogMin;
1996 }
1997
1998 RSP_Vtx_Clipping(i);
1999
2000 short wA = *(short*)((pVtxBase+nOff + 6) ^ 2);
2001 short wB = *(short*)((pVtxBase+nOff + 8) ^ 2);
2002
2003 s8 r = (s8)(wA >> 8);
2004 s8 g = (s8)(wA);
2005 s8 b = (s8)(wB >> 8);
2006 s8 a = (s8)(wB);
2007
2008 if (gRSP.bLightingEnable)
2009 {
2010 g_normal.x = (char)r; //norma.nx;
2011 g_normal.y = (char)g; //norma.ny;
2012 g_normal.z = (char)b; //norma.nz;
2013
2014 Vec3TransformNormal(g_normal, matWorldProject)
2015#if !defined(NO_ASM)
2016 if( status.isSSEEnabled )
2017 g_dwVtxDifColor[i] = SSELightVert();
2018 else
2019#endif
2020 g_dwVtxDifColor[i] = LightVert(g_normal, i);
2021 }
2022 else
2023 {
2024 int nR, nG, nB, nA;
2025
2026 nR = r;
2027 nG = g;
2028 nB = b;
2029 nA = a;
2030 // Assign true vert colour after lighting/fogging
2031 g_dwVtxDifColor[i] = COLOR_RGBA(nR, nG, nB, nA);
2032 }
2033
2034 ReplaceAlphaWithFogFactor(i);
2035
2036 g_fVtxTxtCoords[i].x = g_fVtxTxtCoords[i].y = 1;
2037
2038 nOff += 10;
2039 }
2040
2041
2042 DEBUGGER_PAUSE_AND_DUMP(NEXT_VERTEX_CMD,{DebuggerAppendMsg("Paused at DKR Vertex Cmd, v0=%d, vn=%d, addr=%08X", dwV0, dwNum, dwAddr);});
2043}
2044
2045
2046extern uint32 dwPDCIAddr;
2047void ProcessVertexDataPD(uint32 dwAddr, uint32 dwV0, uint32 dwNum)
2048{
2049 UpdateCombinedMatrix();
2050
2051 N64VtxPD * pVtxBase = (N64VtxPD*)(g_pRDRAMu8 + dwAddr);
2052 g_pVtxBase = (FiddledVtx*)pVtxBase; // Fix me
2053
2054 for (uint32 i = dwV0; i < dwV0 + dwNum; i++)
2055 {
2056 N64VtxPD &vert = pVtxBase[i - dwV0];
2057
2058 g_vtxNonTransformed[i].x = (float)vert.x;
2059 g_vtxNonTransformed[i].y = (float)vert.y;
2060 g_vtxNonTransformed[i].z = (float)vert.z;
2061
2062#if !defined(NO_ASM)
2063 if( status.isSSEEnabled )
2064 SSEVec3Transform(i);
2065 else
2066#endif
2067 {
2068 Vec3Transform(&g_vtxTransformed[i], (XVECTOR3*)&g_vtxNonTransformed[i], &gRSPworldProject); // Convert to w=1
2069 g_vecProjected[i].w = 1.0f / g_vtxTransformed[i].w;
2070 g_vecProjected[i].x = g_vtxTransformed[i].x * g_vecProjected[i].w;
2071 g_vecProjected[i].y = g_vtxTransformed[i].y * g_vecProjected[i].w;
2072 g_vecProjected[i].z = g_vtxTransformed[i].z * g_vecProjected[i].w;
2073 }
2074
2075 g_fFogCoord[i] = g_vecProjected[i].z;
2076 if( g_vecProjected[i].w < 0 || g_vecProjected[i].z < 0 || g_fFogCoord[i] < gRSPfFogMin )
2077 g_fFogCoord[i] = gRSPfFogMin;
2078
2079 RSP_Vtx_Clipping(i);
2080
2081 uint8 *addr = g_pRDRAMu8+dwPDCIAddr+ (vert.cidx&0xFF);
2082 uint32 a = addr[0];
2083 uint32 r = addr[3];
2084 uint32 g = addr[2];
2085 uint32 b = addr[1];
2086
2087 if( gRSP.bLightingEnable )
2088 {
2089 g_normal.x = (char)r;
2090 g_normal.y = (char)g;
2091 g_normal.z = (char)b;
2092#if !defined(NO_ASM)
2093 if( status.isSSEEnabled )
2094 {
2095 SSEVec3TransformNormal();
2096 g_dwVtxDifColor[i] = SSELightVert();
2097 }
2098 else
2099#endif
2100 {
2101 Vec3TransformNormal(g_normal, gRSPmodelViewTop);
2102 g_dwVtxDifColor[i] = LightVert(g_normal, i);
2103 }
2104 *(((uint8*)&(g_dwVtxDifColor[i]))+3) = (uint8)a; // still use alpha from the vertex
2105 }
2106 else
2107 {
2108 if( (gRDP.geometryMode & G_SHADE) == 0 && gRSP.ucode < 5 ) //Shade is disabled
2109 {
2110 g_dwVtxDifColor[i] = gRDP.primitiveColor;
2111 }
2112 else //FLAT shade
2113 {
2114 g_dwVtxDifColor[i] = COLOR_RGBA(r, g, b, a);
2115 }
2116 }
2117
2118 if( options.bWinFrameMode )
2119 {
2120 g_dwVtxDifColor[i] = COLOR_RGBA(r, g, b, a);
2121 }
2122
2123 ReplaceAlphaWithFogFactor(i);
2124
2125 VECTOR2 & t = g_fVtxTxtCoords[i];
2126 if (gRSP.bTextureGen && gRSP.bLightingEnable )
2127 {
2128 // Not sure if we should transform the normal here
2129 //Matrix & matWV = gRSP.projectionMtxs[gRSP.projectionMtxTop];
2130 //Vec3TransformNormal(g_normal, matWV);
2131
2132 TexGen(g_fVtxTxtCoords[i].x, g_fVtxTxtCoords[i].y);
2133 }
2134 else
2135 {
2136 t.x = vert.s;
2137 t.y = vert.t;
2138 }
2139
2140
2141 VTX_DUMP(
2142 {
2143 DebuggerAppendMsg("vtx %d: %d %d %d", i, vert.x,vert.y,vert.z);
2144 DebuggerAppendMsg(" : %f, %f, %f, %f",
2145 g_vtxTransformed[i].x,g_vtxTransformed[i].y,g_vtxTransformed[i].z,g_vtxTransformed[i].w);
2146 DebuggerAppendMsg(" : %X, %X, %X, %X", r,g,b,a);
2147 DebuggerAppendMsg(" : u=%f, v=%f", t.x, t.y);
2148 });
2149 }
2150
2151 VTX_DUMP(TRACE2("Setting Vertexes: %d - %d\n", dwV0, dwV0+dwNum-1));
2152 DEBUGGER_PAUSE_AND_DUMP(NEXT_VERTEX_CMD,{TRACE0("Paused at Vertex Cmd");});
2153}
2154
2155extern uint32 dwConkerVtxZAddr;
2156void ProcessVertexDataConker(uint32 dwAddr, uint32 dwV0, uint32 dwNum)
2157{
2158 UpdateCombinedMatrix();
2159
2160 FiddledVtx * pVtxBase = (FiddledVtx*)(g_pRDRAMu8 + dwAddr);
2161 g_pVtxBase = pVtxBase;
2162 //short *vertexColoraddr = (short*)(g_pRDRAMu8+dwConkerVtxZAddr);
2163
2164 for (uint32 i = dwV0; i < dwV0 + dwNum; i++)
2165 {
2166 SP_Timing(RSP_GBI0_Vtx);
2167
2168 FiddledVtx & vert = pVtxBase[i - dwV0];
2169
2170 g_vtxNonTransformed[i].x = (float)vert.x;
2171 g_vtxNonTransformed[i].y = (float)vert.y;
2172 g_vtxNonTransformed[i].z = (float)vert.z;
2173
2174#if !defined(NO_ASM)
2175 if( status.isSSEEnabled )
2176 SSEVec3Transform(i);
2177 else
2178#endif
2179 {
2180 Vec3Transform(&g_vtxTransformed[i], (XVECTOR3*)&g_vtxNonTransformed[i], &gRSPworldProject); // Convert to w=1
2181 g_vecProjected[i].w = 1.0f / g_vtxTransformed[i].w;
2182 g_vecProjected[i].x = g_vtxTransformed[i].x * g_vecProjected[i].w;
2183 g_vecProjected[i].y = g_vtxTransformed[i].y * g_vecProjected[i].w;
2184 g_vecProjected[i].z = g_vtxTransformed[i].z * g_vecProjected[i].w;
2185 }
2186
2187 g_fFogCoord[i] = g_vecProjected[i].z;
2188 if( g_vecProjected[i].w < 0 || g_vecProjected[i].z < 0 || g_fFogCoord[i] < gRSPfFogMin )
2189 g_fFogCoord[i] = gRSPfFogMin;
2190
2191 VTX_DUMP(
2192 {
2193 uint32 *dat = (uint32*)(&vert);
2194 DebuggerAppendMsg("vtx %d: %08X %08X %08X %08X", i, dat[0],dat[1],dat[2],dat[3]);
2195 DebuggerAppendMsg(" : %f, %f, %f, %f",
2196 g_vtxTransformed[i].x,g_vtxTransformed[i].y,g_vtxTransformed[i].z,g_vtxTransformed[i].w);
2197 DebuggerAppendMsg(" : %f, %f, %f, %f",
2198 g_vecProjected[i].x,g_vecProjected[i].y,g_vecProjected[i].z,g_vecProjected[i].w);
2199 });
2200
2201 RSP_Vtx_Clipping(i);
2202
2203 if( gRSP.bLightingEnable )
2204 {
2205 {
2206 uint32 r= ((gRSP.ambientLightColor>>16)&0xFF);
2207 uint32 g= ((gRSP.ambientLightColor>> 8)&0xFF);
2208 uint32 b= ((gRSP.ambientLightColor )&0xFF);
2209 for( uint32 k=1; k<=gRSPnumLights; k++)
2210 {
2211 r += gRSPlights[k].r;
2212 g += gRSPlights[k].g;
2213 b += gRSPlights[k].b;
2214 }
2215 if( r>255 ) r=255;
2216 if( g>255 ) g=255;
2217 if( b>255 ) b=255;
2218 r *= vert.rgba.r ;
2219 g *= vert.rgba.g ;
2220 b *= vert.rgba.b ;
2221 r >>= 8;
2222 g >>= 8;
2223 b >>= 8;
2224 g_dwVtxDifColor[i] = 0xFF000000;
2225 g_dwVtxDifColor[i] |= (r<<16);
2226 g_dwVtxDifColor[i] |= (g<< 8);
2227 g_dwVtxDifColor[i] |= (b );
2228 }
2229
2230 *(((uint8*)&(g_dwVtxDifColor[i]))+3) = vert.rgba.a; // still use alpha from the vertex
2231 }
2232 else
2233 {
2234 if( (gRDP.geometryMode & G_SHADE) == 0 && gRSP.ucode < 5 ) //Shade is disabled
2235 {
2236 g_dwVtxDifColor[i] = gRDP.primitiveColor;
2237 }
2238 else //FLAT shade
2239 {
2240 g_dwVtxDifColor[i] = COLOR_RGBA(vert.rgba.r, vert.rgba.g, vert.rgba.b, vert.rgba.a);
2241 }
2242 }
2243
2244 if( options.bWinFrameMode )
2245 {
2246 //g_vecProjected[i].z = 0;
2247 g_dwVtxDifColor[i] = COLOR_RGBA(vert.rgba.r, vert.rgba.g, vert.rgba.b, vert.rgba.a);
2248 }
2249
2250 ReplaceAlphaWithFogFactor(i);
2251
2252 // Update texture coords n.b. need to divide tu/tv by bogus scale on addition to buffer
2253 //VECTOR2 & t = g_fVtxTxtCoords[i];
2254
2255 // If the vert is already lit, then there is no normal (and hence we
2256 // can't generate tex coord)
2257 if (gRSP.bTextureGen && gRSP.bLightingEnable )
2258 {
2259 g_normal.x = (float)*(char*)(g_pRDRAMu8+ (((i<<1)+0)^3)+dwConkerVtxZAddr);
2260 g_normal.y = (float)*(char*)(g_pRDRAMu8+ (((i<<1)+1)^3)+dwConkerVtxZAddr);
2261 g_normal.z = (float)*(char*)(g_pRDRAMu8+ (((i<<1)+2)^3)+dwConkerVtxZAddr);
2262 Vec3TransformNormal(g_normal, gRSPmodelViewTop);
2263 TexGen(g_fVtxTxtCoords[i].x, g_fVtxTxtCoords[i].y);
2264 }
2265 else
2266 {
2267 g_fVtxTxtCoords[i].x = (float)vert.tu;
2268 g_fVtxTxtCoords[i].y = (float)vert.tv;
2269 }
2270 }
2271
2272 VTX_DUMP(TRACE2("Setting Vertexes: %d - %d\n", dwV0, dwV0+dwNum-1));
2273 DEBUGGER_PAUSE_AND_DUMP(NEXT_VERTEX_CMD,{DebuggerAppendMsg("Paused at Vertex Cmd");});
2274}
2275
2276
2277typedef struct{
2278 short y;
2279 short x;
2280 short flag;
2281 short z;
2282} RS_Vtx_XYZ;
2283
2284typedef union {
2285 struct {
2286 uint8 a;
2287 uint8 b;
2288 uint8 g;
2289 uint8 r;
2290 };
2291 struct {
2292 char na; //a
2293 char nz; //b
2294 char ny; //g
2295 char nx; //r
2296 };
2297} RS_Vtx_Color;
2298
2299
2300void ProcessVertexData_Rogue_Squadron(uint32 dwXYZAddr, uint32 dwColorAddr, uint32 dwXYZCmd, uint32 dwColorCmd)
2301{
2302 UpdateCombinedMatrix();
2303
2304 uint32 dwV0 = 0;
2305 uint32 dwNum = (dwXYZCmd&0xFF00)>>10;
2306
2307 RS_Vtx_XYZ * pVtxXYZBase = (RS_Vtx_XYZ*)(g_pRDRAMu8 + dwXYZAddr);
2308 RS_Vtx_Color * pVtxColorBase = (RS_Vtx_Color*)(g_pRDRAMu8 + dwColorAddr);
2309
2310 uint32 i;
2311 for (i = dwV0; i < dwV0 + dwNum; i++)
2312 {
2313 RS_Vtx_XYZ & vertxyz = pVtxXYZBase[i - dwV0];
2314 RS_Vtx_Color & vertcolors = pVtxColorBase[i - dwV0];
2315
2316 g_vtxNonTransformed[i].x = (float)vertxyz.x;
2317 g_vtxNonTransformed[i].y = (float)vertxyz.y;
2318 g_vtxNonTransformed[i].z = (float)vertxyz.z;
2319
2320#if !defined(NO_ASM)
2321 if( status.isSSEEnabled )
2322 SSEVec3Transform(i);
2323 else
2324#endif
2325 {
2326 Vec3Transform(&g_vtxTransformed[i], (XVECTOR3*)&g_vtxNonTransformed[i], &gRSPworldProject); // Convert to w=1
2327 g_vecProjected[i].w = 1.0f / g_vtxTransformed[i].w;
2328 g_vecProjected[i].x = g_vtxTransformed[i].x * g_vecProjected[i].w;
2329 g_vecProjected[i].y = g_vtxTransformed[i].y * g_vecProjected[i].w;
2330 g_vecProjected[i].z = g_vtxTransformed[i].z * g_vecProjected[i].w;
2331 }
2332
2333 VTX_DUMP(
2334 {
2335 DebuggerAppendMsg(" : %f, %f, %f, %f",
2336 g_vtxTransformed[i].x,g_vtxTransformed[i].y,g_vtxTransformed[i].z,g_vtxTransformed[i].w);
2337 DebuggerAppendMsg(" : %f, %f, %f, %f",
2338 g_vecProjected[i].x,g_vecProjected[i].y,g_vecProjected[i].z,g_vecProjected[i].w);
2339 });
2340
2341 g_fFogCoord[i] = g_vecProjected[i].z;
2342 if( g_vecProjected[i].w < 0 || g_vecProjected[i].z < 0 || g_fFogCoord[i] < gRSPfFogMin )
2343 g_fFogCoord[i] = gRSPfFogMin;
2344
2345 RSP_Vtx_Clipping(i);
2346
2347 if( gRSP.bLightingEnable )
2348 {
2349 g_normal.x = (float)vertcolors.nx;
2350 g_normal.y = (float)vertcolors.ny;
2351 g_normal.z = (float)vertcolors.nz;
2352
2353#if !defined(NO_ASM)
2354 if( status.isSSEEnabled )
2355 {
2356 SSEVec3TransformNormal();
2357 g_dwVtxDifColor[i] = SSELightVert();
2358 }
2359 else
2360#endif
2361 {
2362 Vec3TransformNormal(g_normal, gRSPmodelViewTop);
2363 g_dwVtxDifColor[i] = LightVert(g_normal, i);
2364 }
2365 *(((uint8*)&(g_dwVtxDifColor[i]))+3) = vertcolors.a; // still use alpha from the vertex
2366 }
2367 else
2368 {
2369 if( (gRDP.geometryMode & G_SHADE) == 0 && gRSP.ucode < 5 ) //Shade is disabled
2370 {
2371 g_dwVtxDifColor[i] = gRDP.primitiveColor;
2372 }
2373 else //FLAT shade
2374 {
2375 g_dwVtxDifColor[i] = COLOR_RGBA(vertcolors.r, vertcolors.g, vertcolors.b, vertcolors.a);
2376 }
2377 }
2378
2379 if( options.bWinFrameMode )
2380 {
2381 g_dwVtxDifColor[i] = COLOR_RGBA(vertcolors.r, vertcolors.g, vertcolors.b, vertcolors.a);
2382 }
2383
2384 ReplaceAlphaWithFogFactor(i);
2385
2386 /*
2387 // Update texture coords n.b. need to divide tu/tv by bogus scale on addition to buffer
2388 VECTOR2 & t = g_fVtxTxtCoords[i];
2389
2390 // If the vert is already lit, then there is no normal (and hence we
2391 // can't generate tex coord)
2392 if (gRSP.bTextureGen && gRSP.bLightingEnable && g_textures[gRSP.curTile].m_bTextureEnable )
2393 {
2394 TexGen(g_fVtxTxtCoords[i].x, g_fVtxTxtCoords[i].y);
2395 }
2396 else
2397 {
2398 t.x = (float)vert.tu;
2399 t.y = (float)vert.tv;
2400 }
2401 */
2402 }
2403
2404 VTX_DUMP(TRACE2("Setting Vertexes: %d - %d\n", dwV0, dwV0+dwNum-1));
2405 DEBUGGER_PAUSE_AND_DUMP(NEXT_VERTEX_CMD,{TRACE0("Paused at Vertex Cmd");});
2406}
2407
2408void SetLightCol(uint32 dwLight, uint32 dwCol)
2409{
2410 gRSPlights[dwLight].r = (uint8)((dwCol >> 24)&0xFF);
2411 gRSPlights[dwLight].g = (uint8)((dwCol >> 16)&0xFF);
2412 gRSPlights[dwLight].b = (uint8)((dwCol >> 8)&0xFF);
2413 gRSPlights[dwLight].a = 255; // Ignore light alpha
2414 gRSPlights[dwLight].fr = (float)gRSPlights[dwLight].r;
2415 gRSPlights[dwLight].fg = (float)gRSPlights[dwLight].g;
2416 gRSPlights[dwLight].fb = (float)gRSPlights[dwLight].b;
2417 gRSPlights[dwLight].fa = 255; // Ignore light alpha
2418
2419 //TRACE1("Set light %d color", dwLight);
2420 LIGHT_DUMP(TRACE2("Set Light %d color: %08X", dwLight, dwCol));
2421}
2422
2423void SetLightDirection(uint32 dwLight, float x, float y, float z, float range)
2424{
2425 //gRSP.bLightIsUpdated = true;
2426
2427 //gRSPlights[dwLight].ox = x;
2428 //gRSPlights[dwLight].oy = y;
2429 //gRSPlights[dwLight].oz = z;
2430
2431 register float w = range == 0 ? (float)sqrt(x*x+y*y+z*z) : 1;
2432
2433 gRSPlights[dwLight].x = x/w;
2434 gRSPlights[dwLight].y = y/w;
2435 gRSPlights[dwLight].z = z/w;
2436 gRSPlights[dwLight].range = range;
2437 DEBUGGER_PAUSE_AND_DUMP(NEXT_SET_LIGHT,TRACE5("Set Light %d dir: %.4f, %.4f, %.4f, %.4f", dwLight, x, y, z, range));
2438}
2439
2440static float maxS0, maxT0;
2441static float maxS1, maxT1;
2442static bool validS0, validT0;
2443static bool validS1, validT1;
2444
2445void LogTextureCoords(float fTex0S, float fTex0T, float fTex1S, float fTex1T)
2446{
2447 if( validS0 )
2448 {
2449 if( fTex0S<0 || fTex0S>maxS0 ) validS0 = false;
2450 }
2451 if( validT0 )
2452 {
2453 if( fTex0T<0 || fTex0T>maxT0 ) validT0 = false;
2454 }
2455 if( validS1 )
2456 {
2457 if( fTex1S<0 || fTex1S>maxS1 ) validS1 = false;
2458 }
2459 if( validT1 )
2460 {
2461 if( fTex1T<0 || fTex1T>maxT1 ) validT1 = false;
2462 }
2463}
2464
2465bool CheckTextureCoords(int tex)
2466{
2467 if( tex==0 )
2468 {
2469 return validS0&&validT0;
2470 }
2471 else
2472 {
2473 return validS1&&validT1;
2474 }
2475}
2476
2477void ResetTextureCoordsLog(float maxs0, float maxt0, float maxs1, float maxt1)
2478{
2479 maxS0 = maxs0;
2480 maxT0 = maxt0;
2481 maxS1 = maxs1;
2482 maxT1 = maxt1;
2483 validS0 = validT0 = true;
2484 validS1 = validT1 = true;
2485}
2486
2487void ForceMainTextureIndex(int dwTile)
2488{
2489 if( dwTile == 1 && !(CRender::g_pRender->IsTexel0Enable()) && CRender::g_pRender->IsTexel1Enable() )
2490 {
2491 // Hack
2492 gRSP.curTile = 0;
2493 }
2494 else
2495 {
2496 gRSP.curTile = dwTile;
2497 }
2498}
2499
2500float HackZ2(float z)
2501{
2502 z = (z+9)/10;
2503 return z;
2504}
2505
2506float HackZ(float z)
2507{
2508 return HackZ2(z);
2509
2510 if( z < 0.1 && z >= 0 )
2511 z = (.1f+z)/2;
2512 else if( z < 0 )
2513 //return (10+z)/100;
2514 z = (expf(z)/20);
2515 return z;
2516}
2517
2518void HackZ(std::vector<XVECTOR3>& points)
2519{
2520 int size = points.size();
2521 for( int i=0; i<size; i++)
2522 {
2523 XVECTOR3 &v = points[i];
2524 v.z = (float)HackZ(v.z);
2525 }
2526}
2527
2528void HackZAll()
2529{
2530 if( CDeviceBuilder::m_deviceGeneralType == DIRECTX_DEVICE )
2531 {
2532 for( uint32 i=0; i<gRSP.numVertices; i++)
2533 {
2534 g_vtxBuffer[i].z = HackZ(g_vtxBuffer[i].z);
2535 }
2536 }
2537 else
2538 {
2539 for( uint32 i=0; i<gRSP.numVertices; i++)
2540 {
2541 float w = g_vtxProjected5[i][3];
2542 g_vtxProjected5[i][2] = HackZ(g_vtxProjected5[i][2]/w)*w;
2543 }
2544 }
2545}
2546
2547
2548extern XMATRIX reverseXY;
2549extern XMATRIX reverseY;
2550
2551void UpdateCombinedMatrix()
2552{
2553 if( gRSP.bMatrixIsUpdated )
2554 {
2555 gRSPworldProject = gRSP.modelviewMtxs[gRSP.modelViewMtxTop] * gRSP.projectionMtxs[gRSP.projectionMtxTop];
2556 gRSP.bMatrixIsUpdated = false;
2557 gRSP.bCombinedMatrixIsUpdated = true;
2558 }
2559
2560 if( gRSP.bCombinedMatrixIsUpdated )
2561 {
2562 if( options.enableHackForGames == HACK_REVERSE_XY_COOR )
2563 {
2564 gRSPworldProject = gRSPworldProject * reverseXY;
2565 }
2566 if( options.enableHackForGames == HACK_REVERSE_Y_COOR )
2567 {
2568 gRSPworldProject = gRSPworldProject * reverseY;
2569 }
2570#if !defined(NO_ASM)
2571 if( status.isSSEEnabled )
2572 {
2573 MatrixTranspose(&gRSPworldProjectTransported, &gRSPworldProject);
2574 }
2575#endif
2576 gRSP.bCombinedMatrixIsUpdated = false;
2577 }
2578
2579 //if( gRSP.bWorldMatrixIsUpdated || gRSP.bLightIsUpdated )
2580 //{
2581 // // Update lights with transported world matrix
2582 // for( unsigned int l=0; l<gRSPnumLights; l++)
2583 // {
2584 // Vec3TransformCoord(&gRSPlights[l].td, &gRSPlights[l].od, &gRSPmodelViewTopTranspose);
2585 // Vec3Normalize(&gRSPlights[l].td,&gRSPlights[l].td);
2586 // }
2587
2588 // gRSP.bWorldMatrixIsUpdated = false;
2589 // gRSP.bLightIsUpdated = false;
2590 //}
2591}
2592
Unnamed repository; edit this file 'description' to name the repository.