2 /* FF is big and ugly so feel free to write lines as long as you like.
5 * Let me make that clearer:
6 * Aieeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeee ! !! !!!
10 #include "basetexture9.h"
11 #include "vertexdeclaration9.h"
12 #include "vertexshader9.h"
13 #include "pixelshader9.h"
15 #include "nine_defines.h"
16 #include "nine_helpers.h"
17 #include "nine_pipe.h"
18 #include "nine_dump.h"
20 #include "pipe/p_context.h"
21 #include "tgsi/tgsi_ureg.h"
22 #include "tgsi/tgsi_dump.h"
23 #include "util/u_box.h"
24 #include "util/u_hash_table.h"
25 #include "util/u_upload_mgr.h"
27 #define DBG_CHANNEL DBG_FF
29 #define NINE_FF_NUM_VS_CONST 196
30 #define NINE_FF_NUM_PS_CONST 24
41 uint32_t position_t
: 1;
42 uint32_t lighting
: 1;
43 uint32_t darkness
: 1; /* lighting enabled but no active lights */
44 uint32_t localviewer
: 1;
45 uint32_t vertexpointsize
: 1;
46 uint32_t pointscale
: 1;
47 uint32_t vertexblend
: 3;
48 uint32_t vertexblend_indexed
: 1;
49 uint32_t vertextween
: 1;
50 uint32_t mtl_diffuse
: 2; /* 0 = material, 1 = color1, 2 = color2 */
51 uint32_t mtl_ambient
: 2;
52 uint32_t mtl_specular
: 2;
53 uint32_t mtl_emissive
: 2;
54 uint32_t fog_mode
: 2;
55 uint32_t fog_range
: 1;
56 uint32_t color0in_one
: 1;
57 uint32_t color1in_zero
: 1;
58 uint32_t has_normal
: 1;
60 uint32_t normalizenormals
: 1;
63 uint32_t tc_dim_input
: 16; /* 8 * 2 bits */
65 uint32_t tc_dim_output
: 24; /* 8 * 3 bits */
67 uint32_t tc_gen
: 24; /* 8 * 3 bits */
73 uint64_t value64
[3]; /* don't forget to resize VertexShader9.ff_key */
78 /* Texture stage state:
80 * COLOROP D3DTOP 5 bit
81 * ALPHAOP D3DTOP 5 bit
82 * COLORARG0 D3DTA 3 bit
83 * COLORARG1 D3DTA 3 bit
84 * COLORARG2 D3DTA 3 bit
85 * ALPHAARG0 D3DTA 3 bit
86 * ALPHAARG1 D3DTA 3 bit
87 * ALPHAARG2 D3DTA 3 bit
88 * RESULTARG D3DTA 1 bit (CURRENT:0 or TEMP:1)
89 * TEXCOORDINDEX 0 - 7 3 bit
90 * ===========================
100 uint32_t colorarg0
: 3;
101 uint32_t colorarg1
: 3;
102 uint32_t colorarg2
: 3;
103 uint32_t alphaarg0
: 3;
104 uint32_t alphaarg1
: 3;
105 uint32_t alphaarg2
: 3;
106 uint32_t resultarg
: 1; /* CURRENT:0 or TEMP:1 */
107 uint32_t textarget
: 2; /* 1D/2D/3D/CUBE */
109 /* that's 32 bit exactly */
111 uint32_t projected
: 16;
112 uint32_t fog
: 1; /* for vFog coming from VS */
113 uint32_t fog_mode
: 2;
114 uint32_t fog_source
: 1; /* 0: Z, 1: W */
115 uint32_t specular
: 1;
116 uint32_t pad1
: 11; /* 9 32-bit words with this */
117 uint8_t colorarg_b4
[3];
118 uint8_t colorarg_b5
[3];
119 uint8_t alphaarg_b4
[3]; /* 11 32-bit words plus a byte */
122 uint64_t value64
[6]; /* don't forget to resize PixelShader9.ff_key */
123 uint32_t value32
[12];
127 static unsigned nine_ff_vs_key_hash(void *key
)
129 struct nine_ff_vs_key
*vs
= key
;
131 uint32_t hash
= vs
->value32
[0];
132 for (i
= 1; i
< ARRAY_SIZE(vs
->value32
); ++i
)
133 hash
^= vs
->value32
[i
];
136 static int nine_ff_vs_key_comp(void *key1
, void *key2
)
138 struct nine_ff_vs_key
*a
= (struct nine_ff_vs_key
*)key1
;
139 struct nine_ff_vs_key
*b
= (struct nine_ff_vs_key
*)key2
;
141 return memcmp(a
->value64
, b
->value64
, sizeof(a
->value64
));
143 static unsigned nine_ff_ps_key_hash(void *key
)
145 struct nine_ff_ps_key
*ps
= key
;
147 uint32_t hash
= ps
->value32
[0];
148 for (i
= 1; i
< ARRAY_SIZE(ps
->value32
); ++i
)
149 hash
^= ps
->value32
[i
];
152 static int nine_ff_ps_key_comp(void *key1
, void *key2
)
154 struct nine_ff_ps_key
*a
= (struct nine_ff_ps_key
*)key1
;
155 struct nine_ff_ps_key
*b
= (struct nine_ff_ps_key
*)key2
;
157 return memcmp(a
->value64
, b
->value64
, sizeof(a
->value64
));
159 static unsigned nine_ff_fvf_key_hash(void *key
)
161 return *(DWORD
*)key
;
163 static int nine_ff_fvf_key_comp(void *key1
, void *key2
)
165 return *(DWORD
*)key1
!= *(DWORD
*)key2
;
168 static void nine_ff_prune_vs(struct NineDevice9
*);
169 static void nine_ff_prune_ps(struct NineDevice9
*);
171 static void nine_ureg_tgsi_dump(struct ureg_program
*ureg
, boolean override
)
173 if (debug_get_bool_option("NINE_FF_DUMP", FALSE
) || override
) {
174 const struct tgsi_token
*toks
= ureg_get_tokens(ureg
, NULL
);
176 ureg_free_tokens(toks
);
180 #define _X(r) ureg_scalar(ureg_src(r), TGSI_SWIZZLE_X)
181 #define _Y(r) ureg_scalar(ureg_src(r), TGSI_SWIZZLE_Y)
182 #define _Z(r) ureg_scalar(ureg_src(r), TGSI_SWIZZLE_Z)
183 #define _W(r) ureg_scalar(ureg_src(r), TGSI_SWIZZLE_W)
185 #define _XXXX(r) ureg_scalar(r, TGSI_SWIZZLE_X)
186 #define _YYYY(r) ureg_scalar(r, TGSI_SWIZZLE_Y)
187 #define _ZZZZ(r) ureg_scalar(r, TGSI_SWIZZLE_Z)
188 #define _WWWW(r) ureg_scalar(r, TGSI_SWIZZLE_W)
192 /* AL should contain base address of lights table. */
193 #define LIGHT_CONST(i) \
194 ureg_src_indirect(ureg_DECL_constant(ureg, i), _X(AL))
196 #define MATERIAL_CONST(i) \
197 ureg_DECL_constant(ureg, 19 + (i))
199 #define _CONST(n) ureg_DECL_constant(ureg, n)
201 /* VS FF constants layout:
203 * CONST[ 0.. 3] D3DTS_WORLD * D3DTS_VIEW * D3DTS_PROJECTION
204 * CONST[ 4.. 7] D3DTS_WORLD * D3DTS_VIEW
205 * CONST[ 8..11] D3DTS_PROJECTION
206 * CONST[12..15] D3DTS_VIEW^(-1)
207 * CONST[16..18] Normal matrix
209 * CONST[19].xyz MATERIAL.Emissive + Material.Ambient * RS.Ambient
210 * CONST[20] MATERIAL.Diffuse
211 * CONST[21] MATERIAL.Ambient
212 * CONST[22] MATERIAL.Specular
213 * CONST[23].x___ MATERIAL.Power
214 * CONST[24] MATERIAL.Emissive
215 * CONST[25] RS.Ambient
217 * CONST[26].x___ RS.PointSizeMin
218 * CONST[26]._y__ RS.PointSizeMax
219 * CONST[26].__z_ RS.PointSize
220 * CONST[26].___w RS.PointScaleA
221 * CONST[27].x___ RS.PointScaleB
222 * CONST[27]._y__ RS.PointScaleC
224 * CONST[28].x___ RS.FogEnd
225 * CONST[28]._y__ 1.0f / (RS.FogEnd - RS.FogStart)
226 * CONST[28].__z_ RS.FogDensity
228 * CONST[30].x___ TWEENFACTOR
230 * CONST[32].x___ LIGHT[0].Type
231 * CONST[32]._yzw LIGHT[0].Attenuation0,1,2
232 * CONST[33] LIGHT[0].Diffuse
233 * CONST[34] LIGHT[0].Specular
234 * CONST[35] LIGHT[0].Ambient
235 * CONST[36].xyz_ LIGHT[0].Position
236 * CONST[36].___w LIGHT[0].Range
237 * CONST[37].xyz_ LIGHT[0].Direction
238 * CONST[37].___w LIGHT[0].Falloff
239 * CONST[38].x___ cos(LIGHT[0].Theta / 2)
240 * CONST[38]._y__ cos(LIGHT[0].Phi / 2)
241 * CONST[38].__z_ 1.0f / (cos(LIGHT[0].Theta / 2) - cos(Light[0].Phi / 2))
242 * CONST[39].xyz_ LIGHT[0].HalfVector (for directional lights)
243 * CONST[39].___w 1 if this is the last active light, 0 if not
251 * NOTE: no lighting code is generated if there are no active lights
253 * CONST[100].x___ Viewport 2/width
254 * CONST[100]._y__ Viewport 2/height
255 * CONST[100].__z_ Viewport 1/(zmax - zmin)
256 * CONST[100].___w Viewport width
257 * CONST[101].x___ Viewport x0
258 * CONST[101]._y__ Viewport y0
259 * CONST[101].__z_ Viewport z0
261 * CONST[128..131] D3DTS_TEXTURE0
262 * CONST[132..135] D3DTS_TEXTURE1
263 * CONST[136..139] D3DTS_TEXTURE2
264 * CONST[140..143] D3DTS_TEXTURE3
265 * CONST[144..147] D3DTS_TEXTURE4
266 * CONST[148..151] D3DTS_TEXTURE5
267 * CONST[152..155] D3DTS_TEXTURE6
268 * CONST[156..159] D3DTS_TEXTURE7
270 * CONST[160] D3DTS_WORLDMATRIX[0] * D3DTS_VIEW
271 * CONST[164] D3DTS_WORLDMATRIX[1] * D3DTS_VIEW
273 * CONST[192] D3DTS_WORLDMATRIX[8] * D3DTS_VIEW
277 struct ureg_program
*ureg
;
278 const struct nine_ff_vs_key
*key
;
280 uint16_t input
[PIPE_MAX_ATTRIBS
];
283 struct ureg_src aVtx
;
284 struct ureg_src aNrm
;
285 struct ureg_src aCol
[2];
286 struct ureg_src aTex
[8];
287 struct ureg_src aPsz
;
288 struct ureg_src aInd
;
289 struct ureg_src aWgt
;
291 struct ureg_src aVtx1
; /* tweening */
292 struct ureg_src aNrm1
;
294 struct ureg_src mtlA
;
295 struct ureg_src mtlD
;
296 struct ureg_src mtlS
;
297 struct ureg_src mtlE
;
300 static inline unsigned
301 get_texcoord_sn(struct pipe_screen
*screen
)
303 if (screen
->get_param(screen
, PIPE_CAP_TGSI_TEXCOORD
))
304 return TGSI_SEMANTIC_TEXCOORD
;
305 return TGSI_SEMANTIC_GENERIC
;
308 static inline struct ureg_src
309 build_vs_add_input(struct vs_build_ctx
*vs
, uint16_t ndecl
)
311 const unsigned i
= vs
->num_inputs
++;
312 assert(i
< PIPE_MAX_ATTRIBS
);
313 vs
->input
[i
] = ndecl
;
314 return ureg_DECL_vs_input(vs
->ureg
, i
);
317 /* NOTE: dst may alias src */
319 ureg_normalize3(struct ureg_program
*ureg
,
320 struct ureg_dst dst
, struct ureg_src src
)
322 struct ureg_dst tmp
= ureg_DECL_temporary(ureg
);
323 struct ureg_dst tmp_x
= ureg_writemask(tmp
, TGSI_WRITEMASK_X
);
325 ureg_DP3(ureg
, tmp_x
, src
, src
);
326 ureg_RSQ(ureg
, tmp_x
, _X(tmp
));
327 ureg_MUL(ureg
, dst
, src
, _X(tmp
));
328 ureg_release_temporary(ureg
, tmp
);
332 nine_ff_build_vs(struct NineDevice9
*device
, struct vs_build_ctx
*vs
)
334 const struct nine_ff_vs_key
*key
= vs
->key
;
335 struct ureg_program
*ureg
= ureg_create(PIPE_SHADER_VERTEX
);
336 struct ureg_dst oPos
, oCol
[2], oPsz
, oFog
;
339 unsigned label
[32], l
= 0;
340 boolean need_aNrm
= key
->lighting
|| key
->passthrough
& (1 << NINE_DECLUSAGE_NORMAL
);
341 boolean has_aNrm
= need_aNrm
&& key
->has_normal
;
342 boolean need_aVtx
= key
->lighting
|| key
->fog_mode
|| key
->pointscale
|| key
->ucp
;
343 const unsigned texcoord_sn
= get_texcoord_sn(device
->screen
);
347 /* Check which inputs we should transform. */
348 for (i
= 0; i
< 8 * 3; i
+= 3) {
349 switch ((key
->tc_gen
>> i
) & 0x7) {
350 case NINED3DTSS_TCI_CAMERASPACENORMAL
:
353 case NINED3DTSS_TCI_CAMERASPACEPOSITION
:
356 case NINED3DTSS_TCI_CAMERASPACEREFLECTIONVECTOR
:
357 need_aVtx
= need_aNrm
= TRUE
;
359 case NINED3DTSS_TCI_SPHEREMAP
:
360 need_aVtx
= need_aNrm
= TRUE
;
367 /* Declare and record used inputs (needed for linkage with vertex format):
368 * (texture coordinates handled later)
370 vs
->aVtx
= build_vs_add_input(vs
,
371 key
->position_t
? NINE_DECLUSAGE_POSITIONT
: NINE_DECLUSAGE_POSITION
);
373 vs
->aNrm
= ureg_imm1f(ureg
, 0.0f
);
375 vs
->aNrm
= build_vs_add_input(vs
, NINE_DECLUSAGE_NORMAL
);
377 vs
->aCol
[0] = ureg_imm1f(ureg
, 1.0f
);
378 vs
->aCol
[1] = ureg_imm1f(ureg
, 0.0f
);
380 if (key
->lighting
|| key
->darkness
) {
381 const unsigned mask
= key
->mtl_diffuse
| key
->mtl_specular
|
382 key
->mtl_ambient
| key
->mtl_emissive
;
383 if ((mask
& 0x1) && !key
->color0in_one
)
384 vs
->aCol
[0] = build_vs_add_input(vs
, NINE_DECLUSAGE_i(COLOR
, 0));
385 if ((mask
& 0x2) && !key
->color1in_zero
)
386 vs
->aCol
[1] = build_vs_add_input(vs
, NINE_DECLUSAGE_i(COLOR
, 1));
388 vs
->mtlD
= MATERIAL_CONST(1);
389 vs
->mtlA
= MATERIAL_CONST(2);
390 vs
->mtlS
= MATERIAL_CONST(3);
391 vs
->mtlE
= MATERIAL_CONST(5);
392 if (key
->mtl_diffuse
== 1) vs
->mtlD
= vs
->aCol
[0]; else
393 if (key
->mtl_diffuse
== 2) vs
->mtlD
= vs
->aCol
[1];
394 if (key
->mtl_ambient
== 1) vs
->mtlA
= vs
->aCol
[0]; else
395 if (key
->mtl_ambient
== 2) vs
->mtlA
= vs
->aCol
[1];
396 if (key
->mtl_specular
== 1) vs
->mtlS
= vs
->aCol
[0]; else
397 if (key
->mtl_specular
== 2) vs
->mtlS
= vs
->aCol
[1];
398 if (key
->mtl_emissive
== 1) vs
->mtlE
= vs
->aCol
[0]; else
399 if (key
->mtl_emissive
== 2) vs
->mtlE
= vs
->aCol
[1];
401 if (!key
->color0in_one
) vs
->aCol
[0] = build_vs_add_input(vs
, NINE_DECLUSAGE_i(COLOR
, 0));
402 if (!key
->color1in_zero
) vs
->aCol
[1] = build_vs_add_input(vs
, NINE_DECLUSAGE_i(COLOR
, 1));
405 if (key
->vertexpointsize
)
406 vs
->aPsz
= build_vs_add_input(vs
, NINE_DECLUSAGE_PSIZE
);
408 if (key
->vertexblend_indexed
|| key
->passthrough
& (1 << NINE_DECLUSAGE_BLENDINDICES
))
409 vs
->aInd
= build_vs_add_input(vs
, NINE_DECLUSAGE_BLENDINDICES
);
410 if (key
->vertexblend
|| key
->passthrough
& (1 << NINE_DECLUSAGE_BLENDWEIGHT
))
411 vs
->aWgt
= build_vs_add_input(vs
, NINE_DECLUSAGE_BLENDWEIGHT
);
412 if (key
->vertextween
) {
413 vs
->aVtx1
= build_vs_add_input(vs
, NINE_DECLUSAGE_i(POSITION
,1));
414 vs
->aNrm1
= build_vs_add_input(vs
, NINE_DECLUSAGE_i(NORMAL
,1));
419 oPos
= ureg_DECL_output(ureg
, TGSI_SEMANTIC_POSITION
, 0); /* HPOS */
420 oCol
[0] = ureg_saturate(ureg_DECL_output(ureg
, TGSI_SEMANTIC_COLOR
, 0));
421 oCol
[1] = ureg_saturate(ureg_DECL_output(ureg
, TGSI_SEMANTIC_COLOR
, 1));
422 if (key
->fog
|| key
->passthrough
& (1 << NINE_DECLUSAGE_FOG
)) {
423 oFog
= ureg_DECL_output(ureg
, TGSI_SEMANTIC_FOG
, 0);
424 oFog
= ureg_writemask(oFog
, TGSI_WRITEMASK_X
);
427 if (key
->vertexpointsize
|| key
->pointscale
) {
428 oPsz
= ureg_DECL_output_masked(ureg
, TGSI_SEMANTIC_PSIZE
, 0,
429 TGSI_WRITEMASK_X
, 0, 1);
430 oPsz
= ureg_writemask(oPsz
, TGSI_WRITEMASK_X
);
433 if (key
->lighting
|| key
->vertexblend
)
434 AR
= ureg_DECL_address(ureg
);
436 /* === Vertex transformation / vertex blending:
439 if (key
->position_t
) {
440 if (device
->driver_caps
.window_space_position_support
) {
441 ureg_MOV(ureg
, oPos
, vs
->aVtx
);
443 struct ureg_dst tmp
= ureg_DECL_temporary(ureg
);
444 /* vs->aVtx contains the coordinates buffer wise.
445 * later in the pipeline, clipping, viewport and division
446 * by w (rhw = 1/w) are going to be applied, so do the reverse
447 * of these transformations (except clipping) to have the good
448 * position at the end.*/
449 ureg_MOV(ureg
, tmp
, vs
->aVtx
);
450 /* X from [X_min, X_min + width] to [-1, 1], same for Y. Z to [0, 1] */
451 ureg_ADD(ureg
, ureg_writemask(tmp
, TGSI_WRITEMASK_XYZ
), ureg_src(tmp
), ureg_negate(_CONST(101)));
452 ureg_MUL(ureg
, ureg_writemask(tmp
, TGSI_WRITEMASK_XYZ
), ureg_src(tmp
), _CONST(100));
453 ureg_ADD(ureg
, ureg_writemask(tmp
, TGSI_WRITEMASK_XY
), ureg_src(tmp
), ureg_imm1f(ureg
, -1.0f
));
454 /* Y needs to be reversed */
455 ureg_MOV(ureg
, ureg_writemask(tmp
, TGSI_WRITEMASK_Y
), ureg_negate(ureg_src(tmp
)));
457 ureg_RCP(ureg
, ureg_writemask(tmp
, TGSI_WRITEMASK_W
), _W(tmp
));
458 /* multiply X, Y, Z by w */
459 ureg_MUL(ureg
, ureg_writemask(tmp
, TGSI_WRITEMASK_XYZ
), ureg_src(tmp
), _W(tmp
));
460 ureg_MOV(ureg
, oPos
, ureg_src(tmp
));
461 ureg_release_temporary(ureg
, tmp
);
463 } else if (key
->vertexblend
) {
464 struct ureg_dst tmp
= ureg_DECL_temporary(ureg
);
465 struct ureg_dst tmp2
= ureg_DECL_temporary(ureg
);
466 struct ureg_dst aVtx_dst
= ureg_DECL_temporary(ureg
);
467 struct ureg_dst aNrm_dst
= ureg_DECL_temporary(ureg
);
468 struct ureg_dst sum_blendweights
= ureg_DECL_temporary(ureg
);
469 struct ureg_src cWM
[4];
471 for (i
= 160; i
<= 195; ++i
)
472 ureg_DECL_constant(ureg
, i
);
474 /* translate world matrix index to constant file index */
475 if (key
->vertexblend_indexed
) {
476 ureg_MAD(ureg
, tmp
, vs
->aInd
, ureg_imm1f(ureg
, 4.0f
), ureg_imm1f(ureg
, 160.0f
));
477 ureg_ARL(ureg
, AR
, ureg_src(tmp
));
480 ureg_MOV(ureg
, aVtx_dst
, ureg_imm4f(ureg
, 0.0f
, 0.0f
, 0.0f
, 0.0f
));
481 ureg_MOV(ureg
, aNrm_dst
, ureg_imm4f(ureg
, 0.0f
, 0.0f
, 0.0f
, 0.0f
));
482 ureg_MOV(ureg
, sum_blendweights
, ureg_imm4f(ureg
, 1.0f
, 1.0f
, 1.0f
, 1.0f
));
484 for (i
= 0; i
< key
->vertexblend
; ++i
) {
485 for (c
= 0; c
< 4; ++c
) {
486 cWM
[c
] = ureg_src_dimension(ureg_src_register(TGSI_FILE_CONSTANT
, (160 + i
* 4) * !key
->vertexblend_indexed
+ c
), 0);
487 if (key
->vertexblend_indexed
)
488 cWM
[c
] = ureg_src_indirect(cWM
[c
], ureg_scalar(ureg_src(AR
), i
));
491 /* multiply by WORLD(index) */
492 ureg_MUL(ureg
, tmp
, _XXXX(vs
->aVtx
), cWM
[0]);
493 ureg_MAD(ureg
, tmp
, _YYYY(vs
->aVtx
), cWM
[1], ureg_src(tmp
));
494 ureg_MAD(ureg
, tmp
, _ZZZZ(vs
->aVtx
), cWM
[2], ureg_src(tmp
));
495 ureg_MAD(ureg
, tmp
, _WWWW(vs
->aVtx
), cWM
[3], ureg_src(tmp
));
498 /* Note: the spec says the transpose of the inverse of the
499 * WorldView matrices should be used, but all tests show
501 * Only case unknown: D3DVBF_0WEIGHTS */
502 ureg_MUL(ureg
, tmp2
, _XXXX(vs
->aNrm
), cWM
[0]);
503 ureg_MAD(ureg
, tmp2
, _YYYY(vs
->aNrm
), cWM
[1], ureg_src(tmp2
));
504 ureg_MAD(ureg
, tmp2
, _ZZZZ(vs
->aNrm
), cWM
[2], ureg_src(tmp2
));
507 if (i
< (key
->vertexblend
- 1)) {
508 /* accumulate weighted position value */
509 ureg_MAD(ureg
, aVtx_dst
, ureg_src(tmp
), ureg_scalar(vs
->aWgt
, i
), ureg_src(aVtx_dst
));
511 ureg_MAD(ureg
, aNrm_dst
, ureg_src(tmp2
), ureg_scalar(vs
->aWgt
, i
), ureg_src(aNrm_dst
));
512 /* subtract weighted position value for last value */
513 ureg_ADD(ureg
, sum_blendweights
, ureg_src(sum_blendweights
), ureg_negate(ureg_scalar(vs
->aWgt
, i
)));
517 /* the last weighted position is always 1 - sum_of_previous_weights */
518 ureg_MAD(ureg
, aVtx_dst
, ureg_src(tmp
), ureg_scalar(ureg_src(sum_blendweights
), key
->vertexblend
- 1), ureg_src(aVtx_dst
));
520 ureg_MAD(ureg
, aNrm_dst
, ureg_src(tmp2
), ureg_scalar(ureg_src(sum_blendweights
), key
->vertexblend
- 1), ureg_src(aNrm_dst
));
522 /* multiply by VIEW_PROJ */
523 ureg_MUL(ureg
, tmp
, _X(aVtx_dst
), _CONST(8));
524 ureg_MAD(ureg
, tmp
, _Y(aVtx_dst
), _CONST(9), ureg_src(tmp
));
525 ureg_MAD(ureg
, tmp
, _Z(aVtx_dst
), _CONST(10), ureg_src(tmp
));
526 ureg_MAD(ureg
, oPos
, _W(aVtx_dst
), _CONST(11), ureg_src(tmp
));
529 vs
->aVtx
= ureg_src(aVtx_dst
);
531 ureg_release_temporary(ureg
, tmp
);
532 ureg_release_temporary(ureg
, tmp2
);
533 ureg_release_temporary(ureg
, sum_blendweights
);
535 ureg_release_temporary(ureg
, aVtx_dst
);
538 if (key
->normalizenormals
)
539 ureg_normalize3(ureg
, aNrm_dst
, ureg_src(aNrm_dst
));
540 vs
->aNrm
= ureg_src(aNrm_dst
);
542 ureg_release_temporary(ureg
, aNrm_dst
);
544 struct ureg_dst tmp
= ureg_DECL_temporary(ureg
);
546 if (key
->vertextween
) {
547 struct ureg_dst aVtx_dst
= ureg_DECL_temporary(ureg
);
548 ureg_LRP(ureg
, aVtx_dst
, _XXXX(_CONST(30)), vs
->aVtx1
, vs
->aVtx
);
549 vs
->aVtx
= ureg_src(aVtx_dst
);
551 struct ureg_dst aNrm_dst
= ureg_DECL_temporary(ureg
);
552 ureg_LRP(ureg
, aNrm_dst
, _XXXX(_CONST(30)), vs
->aNrm1
, vs
->aNrm
);
553 vs
->aNrm
= ureg_src(aNrm_dst
);
557 /* position = vertex * WORLD_VIEW_PROJ */
558 ureg_MUL(ureg
, tmp
, _XXXX(vs
->aVtx
), _CONST(0));
559 ureg_MAD(ureg
, tmp
, _YYYY(vs
->aVtx
), _CONST(1), ureg_src(tmp
));
560 ureg_MAD(ureg
, tmp
, _ZZZZ(vs
->aVtx
), _CONST(2), ureg_src(tmp
));
561 ureg_MAD(ureg
, oPos
, _WWWW(vs
->aVtx
), _CONST(3), ureg_src(tmp
));
562 ureg_release_temporary(ureg
, tmp
);
565 struct ureg_dst aVtx_dst
= ureg_writemask(ureg_DECL_temporary(ureg
), TGSI_WRITEMASK_XYZ
);
566 ureg_MUL(ureg
, aVtx_dst
, _XXXX(vs
->aVtx
), _CONST(4));
567 ureg_MAD(ureg
, aVtx_dst
, _YYYY(vs
->aVtx
), _CONST(5), ureg_src(aVtx_dst
));
568 ureg_MAD(ureg
, aVtx_dst
, _ZZZZ(vs
->aVtx
), _CONST(6), ureg_src(aVtx_dst
));
569 ureg_MAD(ureg
, aVtx_dst
, _WWWW(vs
->aVtx
), _CONST(7), ureg_src(aVtx_dst
));
570 vs
->aVtx
= ureg_src(aVtx_dst
);
573 struct ureg_dst aNrm_dst
= ureg_writemask(ureg_DECL_temporary(ureg
), TGSI_WRITEMASK_XYZ
);
574 ureg_MUL(ureg
, aNrm_dst
, _XXXX(vs
->aNrm
), _CONST(16));
575 ureg_MAD(ureg
, aNrm_dst
, _YYYY(vs
->aNrm
), _CONST(17), ureg_src(aNrm_dst
));
576 ureg_MAD(ureg
, aNrm_dst
, _ZZZZ(vs
->aNrm
), _CONST(18), ureg_src(aNrm_dst
));
577 if (key
->normalizenormals
)
578 ureg_normalize3(ureg
, aNrm_dst
, ureg_src(aNrm_dst
));
579 vs
->aNrm
= ureg_src(aNrm_dst
);
583 /* === Process point size:
585 if (key
->vertexpointsize
|| key
->pointscale
) {
586 struct ureg_dst tmp
= ureg_DECL_temporary(ureg
);
587 struct ureg_dst tmp_x
= ureg_writemask(tmp
, TGSI_WRITEMASK_X
);
588 struct ureg_dst tmp_y
= ureg_writemask(tmp
, TGSI_WRITEMASK_Y
);
589 struct ureg_dst tmp_z
= ureg_writemask(tmp
, TGSI_WRITEMASK_Z
);
590 if (key
->vertexpointsize
) {
591 struct ureg_src cPsz1
= ureg_DECL_constant(ureg
, 26);
592 ureg_MAX(ureg
, tmp_z
, _XXXX(vs
->aPsz
), _XXXX(cPsz1
));
593 ureg_MIN(ureg
, tmp_z
, _Z(tmp
), _YYYY(cPsz1
));
595 struct ureg_src cPsz1
= ureg_DECL_constant(ureg
, 26);
596 ureg_MOV(ureg
, tmp_z
, _ZZZZ(cPsz1
));
599 if (key
->pointscale
) {
600 struct ureg_src cPsz1
= ureg_DECL_constant(ureg
, 26);
601 struct ureg_src cPsz2
= ureg_DECL_constant(ureg
, 27);
603 ureg_DP3(ureg
, tmp_x
, vs
->aVtx
, vs
->aVtx
);
604 ureg_RSQ(ureg
, tmp_y
, _X(tmp
));
605 ureg_MUL(ureg
, tmp_y
, _Y(tmp
), _X(tmp
));
606 ureg_CMP(ureg
, tmp_y
, ureg_negate(_Y(tmp
)), _Y(tmp
), ureg_imm1f(ureg
, 0.0f
));
607 ureg_MAD(ureg
, tmp_x
, _Y(tmp
), _YYYY(cPsz2
), _XXXX(cPsz2
));
608 ureg_MAD(ureg
, tmp_x
, _Y(tmp
), _X(tmp
), _WWWW(cPsz1
));
609 ureg_RSQ(ureg
, tmp_x
, _X(tmp
));
610 ureg_MUL(ureg
, tmp_x
, _X(tmp
), _Z(tmp
));
611 ureg_MUL(ureg
, tmp_x
, _X(tmp
), _WWWW(_CONST(100)));
612 ureg_MAX(ureg
, tmp_x
, _X(tmp
), _XXXX(cPsz1
));
613 ureg_MIN(ureg
, tmp_z
, _X(tmp
), _YYYY(cPsz1
));
616 ureg_MOV(ureg
, oPsz
, _Z(tmp
));
617 ureg_release_temporary(ureg
, tmp
);
620 for (i
= 0; i
< 8; ++i
) {
621 struct ureg_dst tmp
, tmp_x
, tmp2
;
622 struct ureg_dst oTex
, input_coord
, transformed
, t
, aVtx_normed
;
623 unsigned c
, writemask
;
624 const unsigned tci
= (key
->tc_gen
>> (i
* 3)) & 0x7;
625 const unsigned idx
= (key
->tc_idx
>> (i
* 3)) & 0x7;
626 unsigned dim_input
= 1 + ((key
->tc_dim_input
>> (i
* 2)) & 0x3);
627 const unsigned dim_output
= (key
->tc_dim_output
>> (i
* 3)) & 0x7;
629 /* No texture output of index s */
630 if (tci
== NINED3DTSS_TCI_DISABLE
)
632 oTex
= ureg_DECL_output(ureg
, texcoord_sn
, i
);
633 tmp
= ureg_DECL_temporary(ureg
);
634 tmp_x
= ureg_writemask(tmp
, TGSI_WRITEMASK_X
);
635 input_coord
= ureg_DECL_temporary(ureg
);
636 transformed
= ureg_DECL_temporary(ureg
);
638 /* Get the coordinate */
640 case NINED3DTSS_TCI_PASSTHRU
:
641 /* NINED3DTSS_TCI_PASSTHRU => Use texcoord coming from index idx *
642 * Else the idx is used only to determine wrapping mode. */
643 vs
->aTex
[idx
] = build_vs_add_input(vs
, NINE_DECLUSAGE_i(TEXCOORD
,idx
));
644 ureg_MOV(ureg
, input_coord
, vs
->aTex
[idx
]);
646 case NINED3DTSS_TCI_CAMERASPACENORMAL
:
647 ureg_MOV(ureg
, ureg_writemask(input_coord
, TGSI_WRITEMASK_XYZ
), vs
->aNrm
);
648 ureg_MOV(ureg
, ureg_writemask(input_coord
, TGSI_WRITEMASK_W
), ureg_imm1f(ureg
, 1.0f
));
651 case NINED3DTSS_TCI_CAMERASPACEPOSITION
:
652 ureg_MOV(ureg
, ureg_writemask(input_coord
, TGSI_WRITEMASK_XYZ
), vs
->aVtx
);
653 ureg_MOV(ureg
, ureg_writemask(input_coord
, TGSI_WRITEMASK_W
), ureg_imm1f(ureg
, 1.0f
));
656 case NINED3DTSS_TCI_CAMERASPACEREFLECTIONVECTOR
:
657 tmp
.WriteMask
= TGSI_WRITEMASK_XYZ
;
658 aVtx_normed
= ureg_DECL_temporary(ureg
);
659 ureg_normalize3(ureg
, aVtx_normed
, vs
->aVtx
);
660 ureg_DP3(ureg
, tmp_x
, ureg_src(aVtx_normed
), vs
->aNrm
);
661 ureg_MUL(ureg
, tmp
, vs
->aNrm
, _X(tmp
));
662 ureg_ADD(ureg
, tmp
, ureg_src(tmp
), ureg_src(tmp
));
663 ureg_ADD(ureg
, ureg_writemask(input_coord
, TGSI_WRITEMASK_XYZ
), ureg_src(aVtx_normed
), ureg_negate(ureg_src(tmp
)));
664 ureg_MOV(ureg
, ureg_writemask(input_coord
, TGSI_WRITEMASK_W
), ureg_imm1f(ureg
, 1.0f
));
665 ureg_release_temporary(ureg
, aVtx_normed
);
667 tmp
.WriteMask
= TGSI_WRITEMASK_XYZW
;
669 case NINED3DTSS_TCI_SPHEREMAP
:
670 /* Implement the formula of GL_SPHERE_MAP */
671 tmp
.WriteMask
= TGSI_WRITEMASK_XYZ
;
672 aVtx_normed
= ureg_DECL_temporary(ureg
);
673 tmp2
= ureg_DECL_temporary(ureg
);
674 ureg_normalize3(ureg
, aVtx_normed
, vs
->aVtx
);
675 ureg_DP3(ureg
, tmp_x
, ureg_src(aVtx_normed
), vs
->aNrm
);
676 ureg_MUL(ureg
, tmp
, vs
->aNrm
, _X(tmp
));
677 ureg_ADD(ureg
, tmp
, ureg_src(tmp
), ureg_src(tmp
));
678 ureg_ADD(ureg
, tmp
, ureg_src(aVtx_normed
), ureg_negate(ureg_src(tmp
)));
679 /* now tmp = normed(Vtx) - 2 dot3(normed(Vtx), Nrm) Nrm */
680 ureg_MOV(ureg
, ureg_writemask(tmp2
, TGSI_WRITEMASK_XYZ
), ureg_src(tmp
));
681 ureg_MUL(ureg
, tmp2
, ureg_src(tmp2
), ureg_src(tmp2
));
682 ureg_DP3(ureg
, ureg_writemask(tmp2
, TGSI_WRITEMASK_X
), ureg_src(tmp2
), ureg_src(tmp2
));
683 ureg_RSQ(ureg
, ureg_writemask(tmp2
, TGSI_WRITEMASK_X
), ureg_src(tmp2
));
684 ureg_MUL(ureg
, ureg_writemask(tmp2
, TGSI_WRITEMASK_X
), ureg_src(tmp2
), ureg_imm1f(ureg
, 0.5f
));
685 /* tmp2 = 0.5 / sqrt(tmp.x^2 + tmp.y^2 + (tmp.z+1)^2)
686 * TODO: z coordinates are a bit different gl vs d3d, should the formula be adapted ? */
687 ureg_MUL(ureg
, tmp
, ureg_src(tmp
), _X(tmp2
));
688 ureg_ADD(ureg
, ureg_writemask(input_coord
, TGSI_WRITEMASK_XY
), ureg_src(tmp
), ureg_imm1f(ureg
, 0.5f
));
689 ureg_MOV(ureg
, ureg_writemask(input_coord
, TGSI_WRITEMASK_ZW
), ureg_imm4f(ureg
, 0.0f
, 0.0f
, 0.0f
, 1.0f
));
690 ureg_release_temporary(ureg
, aVtx_normed
);
691 ureg_release_temporary(ureg
, tmp2
);
693 tmp
.WriteMask
= TGSI_WRITEMASK_XYZW
;
700 /* Apply the transformation */
701 /* dim_output == 0 => do not transform the components.
702 * XYZRHW also disables transformation */
703 if (!dim_output
|| key
->position_t
) {
704 ureg_release_temporary(ureg
, transformed
);
705 transformed
= input_coord
;
706 writemask
= TGSI_WRITEMASK_XYZW
;
708 for (c
= 0; c
< dim_output
; c
++) {
709 t
= ureg_writemask(transformed
, 1 << c
);
711 /* dim_input = 1 2 3: -> we add trailing 1 to input*/
712 case 1: ureg_MAD(ureg
, t
, _X(input_coord
), _XXXX(_CONST(128 + i
* 4 + c
)), _YYYY(_CONST(128 + i
* 4 + c
)));
714 case 2: ureg_DP2(ureg
, t
, ureg_src(input_coord
), _CONST(128 + i
* 4 + c
));
715 ureg_ADD(ureg
, t
, ureg_src(transformed
), _ZZZZ(_CONST(128 + i
* 4 + c
)));
717 case 3: ureg_DP3(ureg
, t
, ureg_src(input_coord
), _CONST(128 + i
* 4 + c
));
718 ureg_ADD(ureg
, t
, ureg_src(transformed
), _WWWW(_CONST(128 + i
* 4 + c
)));
720 case 4: ureg_DP4(ureg
, t
, ureg_src(input_coord
), _CONST(128 + i
* 4 + c
)); break;
725 writemask
= (1 << dim_output
) - 1;
726 ureg_release_temporary(ureg
, input_coord
);
729 ureg_MOV(ureg
, ureg_writemask(oTex
, writemask
), ureg_src(transformed
));
730 ureg_release_temporary(ureg
, transformed
);
731 ureg_release_temporary(ureg
, tmp
);
736 * DIRECTIONAL: Light at infinite distance, parallel rays, no attenuation.
737 * POINT: Finite distance to scene, divergent rays, isotropic, attenuation.
738 * SPOT: Finite distance, divergent rays, angular dependence, attenuation.
740 * vec3 normal = normalize(in.Normal * NormalMatrix);
741 * vec3 hitDir = light.direction;
744 * if (light.type != DIRECTIONAL)
746 * vec3 hitVec = light.position - eyeVertex;
747 * float d = length(hitVec);
748 * hitDir = hitVec / d;
749 * atten = 1 / ((light.atten2 * d + light.atten1) * d + light.atten0);
752 * if (light.type == SPOTLIGHT)
754 * float rho = dp3(-hitVec, light.direction);
755 * if (rho < cos(light.phi / 2))
757 * if (rho < cos(light.theta / 2))
758 * atten *= pow(some_func(rho), light.falloff);
761 * float nDotHit = dp3_sat(normal, hitVec);
762 * float powFact = 0.0;
766 * vec3 midVec = normalize(hitDir + eye);
767 * float nDotMid = dp3_sat(normal, midVec);
768 * pFact = pow(nDotMid, material.power);
771 * ambient += light.ambient * atten;
772 * diffuse += light.diffuse * atten * nDotHit;
773 * specular += light.specular * atten * powFact;
776 struct ureg_dst tmp
= ureg_DECL_temporary(ureg
);
777 struct ureg_dst tmp_x
= ureg_writemask(tmp
, TGSI_WRITEMASK_X
);
778 struct ureg_dst tmp_y
= ureg_writemask(tmp
, TGSI_WRITEMASK_Y
);
779 struct ureg_dst tmp_z
= ureg_writemask(tmp
, TGSI_WRITEMASK_Z
);
780 struct ureg_dst rAtt
= ureg_writemask(ureg_DECL_temporary(ureg
), TGSI_WRITEMASK_W
);
781 struct ureg_dst rHit
= ureg_writemask(ureg_DECL_temporary(ureg
), TGSI_WRITEMASK_XYZ
);
782 struct ureg_dst rMid
= ureg_writemask(ureg_DECL_temporary(ureg
), TGSI_WRITEMASK_XYZ
);
784 struct ureg_dst rCtr
= ureg_writemask(ureg_DECL_temporary(ureg
), TGSI_WRITEMASK_W
);
786 struct ureg_dst AL
= ureg_writemask(AR
, TGSI_WRITEMASK_X
);
788 /* Light.*.Alpha is not used. */
789 struct ureg_dst rD
= ureg_writemask(ureg_DECL_temporary(ureg
), TGSI_WRITEMASK_XYZ
);
790 struct ureg_dst rA
= ureg_writemask(ureg_DECL_temporary(ureg
), TGSI_WRITEMASK_XYZ
);
791 struct ureg_dst rS
= ureg_DECL_temporary(ureg
);
793 struct ureg_src mtlP
= _XXXX(MATERIAL_CONST(4));
795 struct ureg_src cLKind
= _XXXX(LIGHT_CONST(0));
796 struct ureg_src cLAtt0
= _YYYY(LIGHT_CONST(0));
797 struct ureg_src cLAtt1
= _ZZZZ(LIGHT_CONST(0));
798 struct ureg_src cLAtt2
= _WWWW(LIGHT_CONST(0));
799 struct ureg_src cLColD
= _XYZW(LIGHT_CONST(1));
800 struct ureg_src cLColS
= _XYZW(LIGHT_CONST(2));
801 struct ureg_src cLColA
= _XYZW(LIGHT_CONST(3));
802 struct ureg_src cLPos
= _XYZW(LIGHT_CONST(4));
803 struct ureg_src cLRng
= _WWWW(LIGHT_CONST(4));
804 struct ureg_src cLDir
= _XYZW(LIGHT_CONST(5));
805 struct ureg_src cLFOff
= _WWWW(LIGHT_CONST(5));
806 struct ureg_src cLTht
= _XXXX(LIGHT_CONST(6));
807 struct ureg_src cLPhi
= _YYYY(LIGHT_CONST(6));
808 struct ureg_src cLSDiv
= _ZZZZ(LIGHT_CONST(6));
809 struct ureg_src cLLast
= _WWWW(LIGHT_CONST(7));
811 const unsigned loop_label
= l
++;
813 /* Declare all light constants to allow indirect adressing */
814 for (i
= 32; i
< 96; i
++)
815 ureg_DECL_constant(ureg
, i
);
817 ureg_MOV(ureg
, rCtr
, ureg_imm1f(ureg
, 32.0f
)); /* &lightconst(0) */
818 ureg_MOV(ureg
, rD
, ureg_imm1f(ureg
, 0.0f
));
819 ureg_MOV(ureg
, rA
, ureg_imm1f(ureg
, 0.0f
));
820 ureg_MOV(ureg
, rS
, ureg_imm1f(ureg
, 0.0f
));
822 /* loop management */
823 ureg_BGNLOOP(ureg
, &label
[loop_label
]);
824 ureg_ARL(ureg
, AL
, _W(rCtr
));
826 /* if (not DIRECTIONAL light): */
827 ureg_SNE(ureg
, tmp_x
, cLKind
, ureg_imm1f(ureg
, D3DLIGHT_DIRECTIONAL
));
828 ureg_MOV(ureg
, rHit
, ureg_negate(cLDir
));
829 ureg_MOV(ureg
, rAtt
, ureg_imm1f(ureg
, 1.0f
));
830 ureg_IF(ureg
, _X(tmp
), &label
[l
++]);
832 /* hitDir = light.position - eyeVtx
835 ureg_ADD(ureg
, rHit
, cLPos
, ureg_negate(vs
->aVtx
));
836 ureg_DP3(ureg
, tmp_x
, ureg_src(rHit
), ureg_src(rHit
));
837 ureg_RSQ(ureg
, tmp_y
, _X(tmp
));
838 ureg_MUL(ureg
, tmp_x
, _X(tmp
), _Y(tmp
)); /* length */
840 /* att = 1.0 / (light.att0 + (light.att1 + light.att2 * d) * d) */
841 ureg_MAD(ureg
, rAtt
, _X(tmp
), cLAtt2
, cLAtt1
);
842 ureg_MAD(ureg
, rAtt
, _X(tmp
), _W(rAtt
), cLAtt0
);
843 ureg_RCP(ureg
, rAtt
, _W(rAtt
));
844 /* cut-off if distance exceeds Light.Range */
845 ureg_SLT(ureg
, tmp_x
, _X(tmp
), cLRng
);
846 ureg_MUL(ureg
, rAtt
, _W(rAtt
), _X(tmp
));
848 ureg_fixup_label(ureg
, label
[l
-1], ureg_get_instruction_number(ureg
));
851 /* normalize hitDir */
852 ureg_normalize3(ureg
, rHit
, ureg_src(rHit
));
854 /* if (SPOT light) */
855 ureg_SEQ(ureg
, tmp_x
, cLKind
, ureg_imm1f(ureg
, D3DLIGHT_SPOT
));
856 ureg_IF(ureg
, _X(tmp
), &label
[l
++]);
858 /* rho = dp3(-hitDir, light.spotDir)
860 * if (rho > light.ctht2) NOTE: 0 <= phi <= pi, 0 <= theta <= phi
863 * if (rho <= light.cphi2)
866 * spotAtt = (rho - light.cphi2) / (light.ctht2 - light.cphi2) ^ light.falloff
868 ureg_DP3(ureg
, tmp_y
, ureg_negate(ureg_src(rHit
)), cLDir
); /* rho */
869 ureg_ADD(ureg
, tmp_x
, _Y(tmp
), ureg_negate(cLPhi
));
870 ureg_MUL(ureg
, tmp_x
, _X(tmp
), cLSDiv
);
871 ureg_POW(ureg
, tmp_x
, _X(tmp
), cLFOff
); /* spotAtten */
872 ureg_SGE(ureg
, tmp_z
, _Y(tmp
), cLTht
); /* if inside theta && phi */
873 ureg_SGE(ureg
, tmp_y
, _Y(tmp
), cLPhi
); /* if inside phi */
874 ureg_MAD(ureg
, ureg_saturate(tmp_x
), _X(tmp
), _Y(tmp
), _Z(tmp
));
875 ureg_MUL(ureg
, rAtt
, _W(rAtt
), _X(tmp
));
877 ureg_fixup_label(ureg
, label
[l
-1], ureg_get_instruction_number(ureg
));
880 /* directional factors, let's not use LIT because of clarity */
883 if (key
->localviewer
) {
884 ureg_normalize3(ureg
, rMid
, vs
->aVtx
);
885 ureg_ADD(ureg
, rMid
, ureg_src(rHit
), ureg_negate(ureg_src(rMid
)));
887 ureg_ADD(ureg
, rMid
, ureg_src(rHit
), ureg_imm3f(ureg
, 0.0f
, 0.0f
, -1.0f
));
889 ureg_normalize3(ureg
, rMid
, ureg_src(rMid
));
890 ureg_DP3(ureg
, ureg_saturate(tmp_x
), vs
->aNrm
, ureg_src(rHit
));
891 ureg_DP3(ureg
, ureg_saturate(tmp_y
), vs
->aNrm
, ureg_src(rMid
));
892 ureg_MUL(ureg
, tmp_z
, _X(tmp
), _Y(tmp
));
893 /* Tests show that specular is computed only if (dp3(normal,hitDir) > 0).
894 * For front facing, it is more restrictive than test (dp3(normal,mid) > 0).
895 * No tests were made for backfacing, so add the two conditions */
896 ureg_IF(ureg
, _Z(tmp
), &label
[l
++]);
898 ureg_DP3(ureg
, ureg_saturate(tmp_y
), vs
->aNrm
, ureg_src(rMid
));
899 ureg_POW(ureg
, tmp_y
, _Y(tmp
), mtlP
);
900 ureg_MUL(ureg
, tmp_y
, _W(rAtt
), _Y(tmp
)); /* power factor * att */
901 ureg_MAD(ureg
, rS
, cLColS
, _Y(tmp
), ureg_src(rS
)); /* accumulate specular */
903 ureg_fixup_label(ureg
, label
[l
-1], ureg_get_instruction_number(ureg
));
906 ureg_MUL(ureg
, tmp_x
, _W(rAtt
), _X(tmp
)); /* dp3(normal,hitDir) * att */
907 ureg_MAD(ureg
, rD
, cLColD
, _X(tmp
), ureg_src(rD
)); /* accumulate diffuse */
910 ureg_MAD(ureg
, rA
, cLColA
, _W(rAtt
), ureg_src(rA
)); /* accumulate ambient */
912 /* break if this was the last light */
913 ureg_IF(ureg
, cLLast
, &label
[l
++]);
916 ureg_fixup_label(ureg
, label
[l
-1], ureg_get_instruction_number(ureg
));
918 ureg_ADD(ureg
, rCtr
, _W(rCtr
), ureg_imm1f(ureg
, 8.0f
));
919 ureg_fixup_label(ureg
, label
[loop_label
], ureg_get_instruction_number(ureg
));
920 ureg_ENDLOOP(ureg
, &label
[loop_label
]);
922 /* Apply to material:
924 * oCol[0] = (material.emissive + material.ambient * rs.ambient) +
925 * material.ambient * ambient +
926 * material.diffuse * diffuse +
927 * oCol[1] = material.specular * specular;
929 if (key
->mtl_emissive
== 0 && key
->mtl_ambient
== 0)
930 ureg_MAD(ureg
, ureg_writemask(tmp
, TGSI_WRITEMASK_XYZ
), ureg_src(rA
), vs
->mtlA
, _CONST(19));
932 ureg_ADD(ureg
, ureg_writemask(tmp
, TGSI_WRITEMASK_XYZ
), ureg_src(rA
), _CONST(25));
933 ureg_MAD(ureg
, ureg_writemask(tmp
, TGSI_WRITEMASK_XYZ
), vs
->mtlA
, ureg_src(tmp
), vs
->mtlE
);
936 ureg_MAD(ureg
, ureg_writemask(oCol
[0], TGSI_WRITEMASK_XYZ
), ureg_src(rD
), vs
->mtlD
, ureg_src(tmp
));
937 ureg_MOV(ureg
, ureg_writemask(oCol
[0], TGSI_WRITEMASK_W
), vs
->mtlD
);
938 ureg_MUL(ureg
, oCol
[1], ureg_src(rS
), vs
->mtlS
);
939 ureg_release_temporary(ureg
, rAtt
);
940 ureg_release_temporary(ureg
, rHit
);
941 ureg_release_temporary(ureg
, rMid
);
942 ureg_release_temporary(ureg
, rCtr
);
943 ureg_release_temporary(ureg
, rD
);
944 ureg_release_temporary(ureg
, rA
);
945 ureg_release_temporary(ureg
, rS
);
946 ureg_release_temporary(ureg
, rAtt
);
947 ureg_release_temporary(ureg
, tmp
);
951 if (key
->mtl_emissive
== 0 && key
->mtl_ambient
== 0)
952 ureg_MOV(ureg
, ureg_writemask(oCol
[0], TGSI_WRITEMASK_XYZ
), _CONST(19));
954 ureg_MAD(ureg
, ureg_writemask(oCol
[0], TGSI_WRITEMASK_XYZ
), vs
->mtlA
, _CONST(25), vs
->mtlE
);
955 ureg_MOV(ureg
, ureg_writemask(oCol
[0], TGSI_WRITEMASK_W
), vs
->mtlD
);
956 ureg_MOV(ureg
, oCol
[1], ureg_imm1f(ureg
, 0.0f
));
958 ureg_MOV(ureg
, oCol
[0], vs
->aCol
[0]);
959 ureg_MOV(ureg
, oCol
[1], vs
->aCol
[1]);
964 * exp(x) = ex2(log2(e) * x)
967 struct ureg_dst tmp
= ureg_DECL_temporary(ureg
);
968 struct ureg_dst tmp_x
= ureg_writemask(tmp
, TGSI_WRITEMASK_X
);
969 struct ureg_dst tmp_z
= ureg_writemask(tmp
, TGSI_WRITEMASK_Z
);
970 if (key
->fog_range
) {
971 ureg_DP3(ureg
, tmp_x
, vs
->aVtx
, vs
->aVtx
);
972 ureg_RSQ(ureg
, tmp_z
, _X(tmp
));
973 ureg_MUL(ureg
, tmp_z
, _Z(tmp
), _X(tmp
));
975 ureg_MOV(ureg
, tmp_z
, ureg_abs(_ZZZZ(vs
->aVtx
)));
978 if (key
->fog_mode
== D3DFOG_EXP
) {
979 ureg_MUL(ureg
, tmp_x
, _Z(tmp
), _ZZZZ(_CONST(28)));
980 ureg_MUL(ureg
, tmp_x
, _X(tmp
), ureg_imm1f(ureg
, -1.442695f
));
981 ureg_EX2(ureg
, tmp_x
, _X(tmp
));
983 if (key
->fog_mode
== D3DFOG_EXP2
) {
984 ureg_MUL(ureg
, tmp_x
, _Z(tmp
), _ZZZZ(_CONST(28)));
985 ureg_MUL(ureg
, tmp_x
, _X(tmp
), _X(tmp
));
986 ureg_MUL(ureg
, tmp_x
, _X(tmp
), ureg_imm1f(ureg
, -1.442695f
));
987 ureg_EX2(ureg
, tmp_x
, _X(tmp
));
989 if (key
->fog_mode
== D3DFOG_LINEAR
) {
990 ureg_ADD(ureg
, tmp_x
, _XXXX(_CONST(28)), ureg_negate(_Z(tmp
)));
991 ureg_MUL(ureg
, ureg_saturate(tmp_x
), _X(tmp
), _YYYY(_CONST(28)));
993 ureg_MOV(ureg
, oFog
, _X(tmp
));
994 ureg_release_temporary(ureg
, tmp
);
995 } else if (key
->fog
&& !(key
->passthrough
& (1 << NINE_DECLUSAGE_FOG
))) {
996 ureg_MOV(ureg
, oFog
, ureg_scalar(vs
->aCol
[1], TGSI_SWIZZLE_W
));
999 if (key
->passthrough
& (1 << NINE_DECLUSAGE_BLENDWEIGHT
)) {
1000 struct ureg_src input
;
1001 struct ureg_dst output
;
1003 output
= ureg_DECL_output(ureg
, TGSI_SEMANTIC_GENERIC
, 18);
1004 ureg_MOV(ureg
, output
, input
);
1006 if (key
->passthrough
& (1 << NINE_DECLUSAGE_BLENDINDICES
)) {
1007 struct ureg_src input
;
1008 struct ureg_dst output
;
1010 output
= ureg_DECL_output(ureg
, TGSI_SEMANTIC_GENERIC
, 19);
1011 ureg_MOV(ureg
, output
, input
);
1013 if (key
->passthrough
& (1 << NINE_DECLUSAGE_NORMAL
)) {
1014 struct ureg_src input
;
1015 struct ureg_dst output
;
1017 output
= ureg_DECL_output(ureg
, TGSI_SEMANTIC_GENERIC
, 20);
1018 ureg_MOV(ureg
, output
, input
);
1020 if (key
->passthrough
& (1 << NINE_DECLUSAGE_TANGENT
)) {
1021 struct ureg_src input
;
1022 struct ureg_dst output
;
1023 input
= build_vs_add_input(vs
, NINE_DECLUSAGE_TANGENT
);
1024 output
= ureg_DECL_output(ureg
, TGSI_SEMANTIC_GENERIC
, 21);
1025 ureg_MOV(ureg
, output
, input
);
1027 if (key
->passthrough
& (1 << NINE_DECLUSAGE_BINORMAL
)) {
1028 struct ureg_src input
;
1029 struct ureg_dst output
;
1030 input
= build_vs_add_input(vs
, NINE_DECLUSAGE_BINORMAL
);
1031 output
= ureg_DECL_output(ureg
, TGSI_SEMANTIC_GENERIC
, 22);
1032 ureg_MOV(ureg
, output
, input
);
1034 if (key
->passthrough
& (1 << NINE_DECLUSAGE_FOG
)) {
1035 struct ureg_src input
;
1036 struct ureg_dst output
;
1037 input
= build_vs_add_input(vs
, NINE_DECLUSAGE_FOG
);
1038 input
= ureg_scalar(input
, TGSI_SWIZZLE_X
);
1040 ureg_MOV(ureg
, output
, input
);
1042 if (key
->passthrough
& (1 << NINE_DECLUSAGE_DEPTH
)) {
1043 (void) 0; /* TODO: replace z of position output ? */
1046 /* ucp for ff applies on world coordinates.
1047 * aVtx is in worldview coordinates. */
1049 struct ureg_dst clipVect
= ureg_DECL_output(ureg
, TGSI_SEMANTIC_CLIPVERTEX
, 0);
1050 struct ureg_dst tmp
= ureg_DECL_temporary(ureg
);
1051 ureg_MUL(ureg
, tmp
, _XXXX(vs
->aVtx
), _CONST(12));
1052 ureg_MAD(ureg
, tmp
, _YYYY(vs
->aVtx
), _CONST(13), ureg_src(tmp
));
1053 ureg_MAD(ureg
, tmp
, _ZZZZ(vs
->aVtx
), _CONST(14), ureg_src(tmp
));
1054 ureg_ADD(ureg
, clipVect
, _CONST(15), ureg_src(tmp
));
1055 ureg_release_temporary(ureg
, tmp
);
1058 if (key
->position_t
&& device
->driver_caps
.window_space_position_support
)
1059 ureg_property(ureg
, TGSI_PROPERTY_VS_WINDOW_SPACE_POSITION
, TRUE
);
1062 nine_ureg_tgsi_dump(ureg
, FALSE
);
1063 return ureg_create_shader_and_destroy(ureg
, device
->context
.pipe
);
1066 /* PS FF constants layout:
1068 * CONST[ 0.. 7] stage[i].D3DTSS_CONSTANT
1069 * CONST[ 8..15].x___ stage[i].D3DTSS_BUMPENVMAT00
1070 * CONST[ 8..15]._y__ stage[i].D3DTSS_BUMPENVMAT01
1071 * CONST[ 8..15].__z_ stage[i].D3DTSS_BUMPENVMAT10
1072 * CONST[ 8..15].___w stage[i].D3DTSS_BUMPENVMAT11
1073 * CONST[16..19].x_z_ stage[i].D3DTSS_BUMPENVLSCALE
1074 * CONST[17..19]._y_w stage[i].D3DTSS_BUMPENVLOFFSET
1076 * CONST[20] D3DRS_TEXTUREFACTOR
1077 * CONST[21] D3DRS_FOGCOLOR
1078 * CONST[22].x___ RS.FogEnd
1079 * CONST[22]._y__ 1.0f / (RS.FogEnd - RS.FogStart)
1080 * CONST[22].__z_ RS.FogDensity
1084 struct ureg_program
*ureg
;
1086 struct ureg_src vC
[2]; /* DIFFUSE, SPECULAR */
1087 struct ureg_src vT
[8]; /* TEXCOORD[i] */
1088 struct ureg_dst rCur
; /* D3DTA_CURRENT */
1089 struct ureg_dst rMod
;
1090 struct ureg_src rCurSrc
;
1091 struct ureg_dst rTmp
; /* D3DTA_TEMP */
1092 struct ureg_src rTmpSrc
;
1093 struct ureg_dst rTex
;
1094 struct ureg_src rTexSrc
;
1095 struct ureg_src cBEM
[8];
1096 struct ureg_src s
[8];
1100 unsigned index_pre_mod
;
1104 static struct ureg_src
1105 ps_get_ts_arg(struct ps_build_ctx
*ps
, unsigned ta
)
1107 struct ureg_src reg
;
1109 switch (ta
& D3DTA_SELECTMASK
) {
1110 case D3DTA_CONSTANT
:
1111 reg
= ureg_DECL_constant(ps
->ureg
, ps
->stage
.index
);
1114 reg
= (ps
->stage
.index
== ps
->stage
.index_pre_mod
) ? ureg_src(ps
->rMod
) : ps
->rCurSrc
;
1117 reg
= ureg_DECL_fs_input(ps
->ureg
, TGSI_SEMANTIC_COLOR
, 0, TGSI_INTERPOLATE_COLOR
);
1119 case D3DTA_SPECULAR
:
1120 reg
= ureg_DECL_fs_input(ps
->ureg
, TGSI_SEMANTIC_COLOR
, 1, TGSI_INTERPOLATE_COLOR
);
1129 reg
= ureg_DECL_constant(ps
->ureg
, 20);
1133 reg
= ureg_src_undef();
1136 if (ta
& D3DTA_COMPLEMENT
) {
1137 struct ureg_dst dst
= ureg_DECL_temporary(ps
->ureg
);
1138 ureg_ADD(ps
->ureg
, dst
, ureg_imm1f(ps
->ureg
, 1.0f
), ureg_negate(reg
));
1139 reg
= ureg_src(dst
);
1141 if (ta
& D3DTA_ALPHAREPLICATE
)
1146 static struct ureg_dst
1147 ps_get_ts_dst(struct ps_build_ctx
*ps
, unsigned ta
)
1149 assert(!(ta
& (D3DTA_COMPLEMENT
| D3DTA_ALPHAREPLICATE
)));
1151 switch (ta
& D3DTA_SELECTMASK
) {
1158 return ureg_dst_undef();
1162 static uint8_t ps_d3dtop_args_mask(D3DTEXTUREOP top
)
1165 case D3DTOP_DISABLE
:
1167 case D3DTOP_SELECTARG1
:
1168 case D3DTOP_PREMODULATE
:
1170 case D3DTOP_SELECTARG2
:
1172 case D3DTOP_MULTIPLYADD
:
1180 static inline boolean
1181 is_MOV_no_op(struct ureg_dst dst
, struct ureg_src src
)
1183 return !dst
.WriteMask
||
1184 (dst
.File
== src
.File
&&
1185 dst
.Index
== src
.Index
&&
1191 (!(dst
.WriteMask
& TGSI_WRITEMASK_X
) || (src
.SwizzleX
== TGSI_SWIZZLE_X
)) &&
1192 (!(dst
.WriteMask
& TGSI_WRITEMASK_Y
) || (src
.SwizzleY
== TGSI_SWIZZLE_Y
)) &&
1193 (!(dst
.WriteMask
& TGSI_WRITEMASK_Z
) || (src
.SwizzleZ
== TGSI_SWIZZLE_Z
)) &&
1194 (!(dst
.WriteMask
& TGSI_WRITEMASK_W
) || (src
.SwizzleW
== TGSI_SWIZZLE_W
)));
1199 ps_do_ts_op(struct ps_build_ctx
*ps
, unsigned top
, struct ureg_dst dst
, struct ureg_src
*arg
)
1201 struct ureg_program
*ureg
= ps
->ureg
;
1202 struct ureg_dst tmp
= ureg_DECL_temporary(ureg
);
1203 struct ureg_dst tmp2
= ureg_DECL_temporary(ureg
);
1204 struct ureg_dst tmp_x
= ureg_writemask(tmp
, TGSI_WRITEMASK_X
);
1206 tmp
.WriteMask
= dst
.WriteMask
;
1208 if (top
!= D3DTOP_SELECTARG1
&& top
!= D3DTOP_SELECTARG2
&&
1209 top
!= D3DTOP_MODULATE
&& top
!= D3DTOP_PREMODULATE
&&
1210 top
!= D3DTOP_BLENDDIFFUSEALPHA
&& top
!= D3DTOP_BLENDTEXTUREALPHA
&&
1211 top
!= D3DTOP_BLENDFACTORALPHA
&& top
!= D3DTOP_BLENDCURRENTALPHA
&&
1212 top
!= D3DTOP_BUMPENVMAP
&& top
!= D3DTOP_BUMPENVMAPLUMINANCE
&&
1214 dst
= ureg_saturate(dst
);
1217 case D3DTOP_SELECTARG1
:
1218 if (!is_MOV_no_op(dst
, arg
[1]))
1219 ureg_MOV(ureg
, dst
, arg
[1]);
1221 case D3DTOP_SELECTARG2
:
1222 if (!is_MOV_no_op(dst
, arg
[2]))
1223 ureg_MOV(ureg
, dst
, arg
[2]);
1225 case D3DTOP_MODULATE
:
1226 ureg_MUL(ureg
, dst
, arg
[1], arg
[2]);
1228 case D3DTOP_MODULATE2X
:
1229 ureg_MUL(ureg
, tmp
, arg
[1], arg
[2]);
1230 ureg_ADD(ureg
, dst
, ureg_src(tmp
), ureg_src(tmp
));
1232 case D3DTOP_MODULATE4X
:
1233 ureg_MUL(ureg
, tmp
, arg
[1], arg
[2]);
1234 ureg_MUL(ureg
, dst
, ureg_src(tmp
), ureg_imm1f(ureg
, 4.0f
));
1237 ureg_ADD(ureg
, dst
, arg
[1], arg
[2]);
1239 case D3DTOP_ADDSIGNED
:
1240 ureg_ADD(ureg
, tmp
, arg
[1], arg
[2]);
1241 ureg_ADD(ureg
, dst
, ureg_src(tmp
), ureg_imm1f(ureg
, -0.5f
));
1243 case D3DTOP_ADDSIGNED2X
:
1244 ureg_ADD(ureg
, tmp
, arg
[1], arg
[2]);
1245 ureg_MAD(ureg
, dst
, ureg_src(tmp
), ureg_imm1f(ureg
, 2.0f
), ureg_imm1f(ureg
, -1.0f
));
1247 case D3DTOP_SUBTRACT
:
1248 ureg_ADD(ureg
, dst
, arg
[1], ureg_negate(arg
[2]));
1250 case D3DTOP_ADDSMOOTH
:
1251 ureg_ADD(ureg
, tmp
, ureg_imm1f(ureg
, 1.0f
), ureg_negate(arg
[1]));
1252 ureg_MAD(ureg
, dst
, ureg_src(tmp
), arg
[2], arg
[1]);
1254 case D3DTOP_BLENDDIFFUSEALPHA
:
1255 ureg_LRP(ureg
, dst
, _WWWW(ps
->vC
[0]), arg
[1], arg
[2]);
1257 case D3DTOP_BLENDTEXTUREALPHA
:
1258 /* XXX: alpha taken from previous stage, texture or result ? */
1259 ureg_LRP(ureg
, dst
, _W(ps
->rTex
), arg
[1], arg
[2]);
1261 case D3DTOP_BLENDFACTORALPHA
:
1262 ureg_LRP(ureg
, dst
, _WWWW(_CONST(20)), arg
[1], arg
[2]);
1264 case D3DTOP_BLENDTEXTUREALPHAPM
:
1265 ureg_ADD(ureg
, tmp_x
, ureg_imm1f(ureg
, 1.0f
), ureg_negate(_W(ps
->rTex
)));
1266 ureg_MAD(ureg
, dst
, arg
[2], _X(tmp
), arg
[1]);
1268 case D3DTOP_BLENDCURRENTALPHA
:
1269 ureg_LRP(ureg
, dst
, _WWWW(ps
->rCurSrc
), arg
[1], arg
[2]);
1271 case D3DTOP_PREMODULATE
:
1272 ureg_MOV(ureg
, dst
, arg
[1]);
1273 ps
->stage
.index_pre_mod
= ps
->stage
.index
+ 1;
1275 case D3DTOP_MODULATEALPHA_ADDCOLOR
:
1276 ureg_MAD(ureg
, dst
, _WWWW(arg
[1]), arg
[2], arg
[1]);
1278 case D3DTOP_MODULATECOLOR_ADDALPHA
:
1279 ureg_MAD(ureg
, dst
, arg
[1], arg
[2], _WWWW(arg
[1]));
1281 case D3DTOP_MODULATEINVALPHA_ADDCOLOR
:
1282 ureg_ADD(ureg
, tmp_x
, ureg_imm1f(ureg
, 1.0f
), ureg_negate(_WWWW(arg
[1])));
1283 ureg_MAD(ureg
, dst
, _X(tmp
), arg
[2], arg
[1]);
1285 case D3DTOP_MODULATEINVCOLOR_ADDALPHA
:
1286 ureg_ADD(ureg
, tmp
, ureg_imm1f(ureg
, 1.0f
), ureg_negate(arg
[1]));
1287 ureg_MAD(ureg
, dst
, ureg_src(tmp
), arg
[2], _WWWW(arg
[1]));
1289 case D3DTOP_BUMPENVMAP
:
1291 case D3DTOP_BUMPENVMAPLUMINANCE
:
1293 case D3DTOP_DOTPRODUCT3
:
1294 ureg_ADD(ureg
, tmp
, arg
[1], ureg_imm4f(ureg
,-0.5,-0.5,-0.5,-0.5));
1295 ureg_ADD(ureg
, tmp2
, arg
[2] , ureg_imm4f(ureg
,-0.5,-0.5,-0.5,-0.5));
1296 ureg_DP3(ureg
, tmp
, ureg_src(tmp
), ureg_src(tmp2
));
1297 ureg_MUL(ureg
, ureg_saturate(dst
), ureg_src(tmp
), ureg_imm4f(ureg
,4.0,4.0,4.0,4.0));
1299 case D3DTOP_MULTIPLYADD
:
1300 ureg_MAD(ureg
, dst
, arg
[1], arg
[2], arg
[0]);
1303 ureg_LRP(ureg
, dst
, arg
[0], arg
[1], arg
[2]);
1305 case D3DTOP_DISABLE
:
1309 assert(!"invalid D3DTOP");
1312 ureg_release_temporary(ureg
, tmp
);
1313 ureg_release_temporary(ureg
, tmp2
);
1317 nine_ff_build_ps(struct NineDevice9
*device
, struct nine_ff_ps_key
*key
)
1319 struct ps_build_ctx ps
;
1320 struct ureg_program
*ureg
= ureg_create(PIPE_SHADER_FRAGMENT
);
1321 struct ureg_dst oCol
;
1323 const unsigned texcoord_sn
= get_texcoord_sn(device
->screen
);
1325 memset(&ps
, 0, sizeof(ps
));
1327 ps
.stage
.index_pre_mod
= -1;
1329 ps
.vC
[0] = ureg_DECL_fs_input(ureg
, TGSI_SEMANTIC_COLOR
, 0, TGSI_INTERPOLATE_COLOR
);
1331 ps
.rCur
= ureg_DECL_temporary(ureg
);
1332 ps
.rTmp
= ureg_DECL_temporary(ureg
);
1333 ps
.rTex
= ureg_DECL_temporary(ureg
);
1334 ps
.rCurSrc
= ureg_src(ps
.rCur
);
1335 ps
.rTmpSrc
= ureg_src(ps
.rTmp
);
1336 ps
.rTexSrc
= ureg_src(ps
.rTex
);
1338 /* Initial values */
1339 ureg_MOV(ureg
, ps
.rCur
, ps
.vC
[0]);
1340 ureg_MOV(ureg
, ps
.rTmp
, ureg_imm1f(ureg
, 0.0f
));
1341 ureg_MOV(ureg
, ps
.rTex
, ureg_imm1f(ureg
, 0.0f
));
1343 for (s
= 0; s
< 8; ++s
) {
1344 ps
.s
[s
] = ureg_src_undef();
1346 if (key
->ts
[s
].colorop
!= D3DTOP_DISABLE
) {
1347 if (key
->ts
[s
].colorarg0
== D3DTA_SPECULAR
||
1348 key
->ts
[s
].colorarg1
== D3DTA_SPECULAR
||
1349 key
->ts
[s
].colorarg2
== D3DTA_SPECULAR
)
1350 ps
.vC
[1] = ureg_DECL_fs_input(ureg
, TGSI_SEMANTIC_COLOR
, 1, TGSI_INTERPOLATE_COLOR
);
1352 if (key
->ts
[s
].colorarg0
== D3DTA_TEXTURE
||
1353 key
->ts
[s
].colorarg1
== D3DTA_TEXTURE
||
1354 key
->ts
[s
].colorarg2
== D3DTA_TEXTURE
) {
1355 ps
.s
[s
] = ureg_DECL_sampler(ureg
, s
);
1356 ps
.vT
[s
] = ureg_DECL_fs_input(ureg
, texcoord_sn
, s
, TGSI_INTERPOLATE_PERSPECTIVE
);
1358 if (s
&& (key
->ts
[s
- 1].colorop
== D3DTOP_PREMODULATE
||
1359 key
->ts
[s
- 1].alphaop
== D3DTOP_PREMODULATE
))
1360 ps
.s
[s
] = ureg_DECL_sampler(ureg
, s
);
1363 if (key
->ts
[s
].alphaop
!= D3DTOP_DISABLE
) {
1364 if (key
->ts
[s
].alphaarg0
== D3DTA_SPECULAR
||
1365 key
->ts
[s
].alphaarg1
== D3DTA_SPECULAR
||
1366 key
->ts
[s
].alphaarg2
== D3DTA_SPECULAR
)
1367 ps
.vC
[1] = ureg_DECL_fs_input(ureg
, TGSI_SEMANTIC_COLOR
, 1, TGSI_INTERPOLATE_COLOR
);
1369 if (key
->ts
[s
].alphaarg0
== D3DTA_TEXTURE
||
1370 key
->ts
[s
].alphaarg1
== D3DTA_TEXTURE
||
1371 key
->ts
[s
].alphaarg2
== D3DTA_TEXTURE
) {
1372 ps
.s
[s
] = ureg_DECL_sampler(ureg
, s
);
1373 ps
.vT
[s
] = ureg_DECL_fs_input(ureg
, texcoord_sn
, s
, TGSI_INTERPOLATE_PERSPECTIVE
);
1378 ps
.vC
[1] = ureg_DECL_fs_input(ureg
, TGSI_SEMANTIC_COLOR
, 1, TGSI_INTERPOLATE_COLOR
);
1380 oCol
= ureg_DECL_output(ureg
, TGSI_SEMANTIC_COLOR
, 0);
1384 for (s
= 0; s
< 8; ++s
) {
1385 unsigned colorarg
[3];
1386 unsigned alphaarg
[3];
1387 const uint8_t used_c
= ps_d3dtop_args_mask(key
->ts
[s
].colorop
);
1388 const uint8_t used_a
= ps_d3dtop_args_mask(key
->ts
[s
].alphaop
);
1389 struct ureg_dst dst
;
1390 struct ureg_src arg
[3];
1392 if (key
->ts
[s
].colorop
== D3DTOP_DISABLE
) {
1393 assert (key
->ts
[s
].alphaop
== D3DTOP_DISABLE
);
1398 DBG("STAGE[%u]: colorop=%s alphaop=%s\n", s
,
1399 nine_D3DTOP_to_str(key
->ts
[s
].colorop
),
1400 nine_D3DTOP_to_str(key
->ts
[s
].alphaop
));
1402 if (!ureg_src_is_undef(ps
.s
[s
])) {
1404 struct ureg_src texture_coord
= ps
.vT
[s
];
1405 struct ureg_dst delta
;
1406 switch (key
->ts
[s
].textarget
) {
1407 case 0: target
= TGSI_TEXTURE_1D
; break;
1408 case 1: target
= TGSI_TEXTURE_2D
; break;
1409 case 2: target
= TGSI_TEXTURE_3D
; break;
1410 case 3: target
= TGSI_TEXTURE_CUBE
; break;
1411 /* this is a 2 bit bitfield, do I really need a default case ? */
1414 /* Modify coordinates */
1416 (key
->ts
[s
-1].colorop
== D3DTOP_BUMPENVMAP
||
1417 key
->ts
[s
-1].colorop
== D3DTOP_BUMPENVMAPLUMINANCE
)) {
1418 delta
= ureg_DECL_temporary(ureg
);
1419 /* Du' = D3DTSS_BUMPENVMAT00(stage s-1)*t(s-1)R + D3DTSS_BUMPENVMAT10(stage s-1)*t(s-1)G */
1420 ureg_MUL(ureg
, ureg_writemask(delta
, TGSI_WRITEMASK_X
), _X(ps
.rTex
), _XXXX(_CONST(8 + s
- 1)));
1421 ureg_MAD(ureg
, ureg_writemask(delta
, TGSI_WRITEMASK_X
), _Y(ps
.rTex
), _ZZZZ(_CONST(8 + s
- 1)), ureg_src(delta
));
1422 /* Dv' = D3DTSS_BUMPENVMAT01(stage s-1)*t(s-1)R + D3DTSS_BUMPENVMAT11(stage s-1)*t(s-1)G */
1423 ureg_MUL(ureg
, ureg_writemask(delta
, TGSI_WRITEMASK_Y
), _X(ps
.rTex
), _YYYY(_CONST(8 + s
- 1)));
1424 ureg_MAD(ureg
, ureg_writemask(delta
, TGSI_WRITEMASK_Y
), _Y(ps
.rTex
), _WWWW(_CONST(8 + s
- 1)), ureg_src(delta
));
1425 texture_coord
= ureg_src(ureg_DECL_temporary(ureg
));
1426 ureg_MOV(ureg
, ureg_writemask(ureg_dst(texture_coord
), ureg_dst(ps
.vT
[s
]).WriteMask
), ps
.vT
[s
]);
1427 ureg_ADD(ureg
, ureg_writemask(ureg_dst(texture_coord
), TGSI_WRITEMASK_XY
), texture_coord
, ureg_src(delta
));
1428 /* Prepare luminance multiplier
1429 * t(s)RGBA = t(s)RGBA * clamp[(t(s-1)B * D3DTSS_BUMPENVLSCALE(stage s-1)) + D3DTSS_BUMPENVLOFFSET(stage s-1)] */
1430 if (key
->ts
[s
-1].colorop
== D3DTOP_BUMPENVMAPLUMINANCE
) {
1431 struct ureg_src bumpenvlscale
= ((s
-1) & 1) ? _ZZZZ(_CONST(16 + (s
-1) / 2)) : _XXXX(_CONST(16 + (s
-1) / 2));
1432 struct ureg_src bumpenvloffset
= ((s
-1) & 1) ? _WWWW(_CONST(16 + (s
-1) / 2)) : _YYYY(_CONST(16 + (s
-1) / 2));
1434 ureg_MAD(ureg
, ureg_saturate(ureg_writemask(delta
, TGSI_WRITEMASK_X
)), _Z(ps
.rTex
), bumpenvlscale
, bumpenvloffset
);
1437 if (key
->projected
& (3 << (s
*2))) {
1438 unsigned dim
= 1 + ((key
->projected
>> (2 * s
)) & 3);
1440 ureg_TXP(ureg
, ps
.rTex
, target
, texture_coord
, ps
.s
[s
]);
1442 struct ureg_dst tmp
= ureg_DECL_temporary(ureg
);
1443 ureg_RCP(ureg
, ureg_writemask(tmp
, TGSI_WRITEMASK_X
), ureg_scalar(texture_coord
, dim
-1));
1444 ureg_MUL(ureg
, ps
.rTmp
, _X(tmp
), texture_coord
);
1445 ureg_TEX(ureg
, ps
.rTex
, target
, ps
.rTmpSrc
, ps
.s
[s
]);
1446 ureg_release_temporary(ureg
, tmp
);
1449 ureg_TEX(ureg
, ps
.rTex
, target
, texture_coord
, ps
.s
[s
]);
1451 if (s
>= 1 && key
->ts
[s
-1].colorop
== D3DTOP_BUMPENVMAPLUMINANCE
)
1452 ureg_MUL(ureg
, ps
.rTex
, ureg_src(ps
.rTex
), _X(delta
));
1455 if (key
->ts
[s
].colorop
== D3DTOP_BUMPENVMAP
||
1456 key
->ts
[s
].colorop
== D3DTOP_BUMPENVMAPLUMINANCE
)
1459 dst
= ps_get_ts_dst(&ps
, key
->ts
[s
].resultarg
? D3DTA_TEMP
: D3DTA_CURRENT
);
1461 if (ps
.stage
.index_pre_mod
== ps
.stage
.index
) {
1462 ps
.rMod
= ureg_DECL_temporary(ureg
);
1463 ureg_MUL(ureg
, ps
.rMod
, ps
.rCurSrc
, ps
.rTexSrc
);
1466 colorarg
[0] = (key
->ts
[s
].colorarg0
| (((key
->colorarg_b4
[0] >> s
) & 0x1) << 4) | ((key
->colorarg_b5
[0] >> s
) << 5)) & 0x3f;
1467 colorarg
[1] = (key
->ts
[s
].colorarg1
| (((key
->colorarg_b4
[1] >> s
) & 0x1) << 4) | ((key
->colorarg_b5
[1] >> s
) << 5)) & 0x3f;
1468 colorarg
[2] = (key
->ts
[s
].colorarg2
| (((key
->colorarg_b4
[2] >> s
) & 0x1) << 4) | ((key
->colorarg_b5
[2] >> s
) << 5)) & 0x3f;
1469 alphaarg
[0] = (key
->ts
[s
].alphaarg0
| ((key
->alphaarg_b4
[0] >> s
) << 4)) & 0x1f;
1470 alphaarg
[1] = (key
->ts
[s
].alphaarg1
| ((key
->alphaarg_b4
[1] >> s
) << 4)) & 0x1f;
1471 alphaarg
[2] = (key
->ts
[s
].alphaarg2
| ((key
->alphaarg_b4
[2] >> s
) << 4)) & 0x1f;
1473 if (key
->ts
[s
].colorop
!= key
->ts
[s
].alphaop
||
1474 colorarg
[0] != alphaarg
[0] ||
1475 colorarg
[1] != alphaarg
[1] ||
1476 colorarg
[2] != alphaarg
[2])
1477 dst
.WriteMask
= TGSI_WRITEMASK_XYZ
;
1479 /* Special DOTPRODUCT behaviour (see wine tests) */
1480 if (key
->ts
[s
].colorop
== D3DTOP_DOTPRODUCT3
)
1481 dst
.WriteMask
= TGSI_WRITEMASK_XYZW
;
1483 if (used_c
& 0x1) arg
[0] = ps_get_ts_arg(&ps
, colorarg
[0]);
1484 if (used_c
& 0x2) arg
[1] = ps_get_ts_arg(&ps
, colorarg
[1]);
1485 if (used_c
& 0x4) arg
[2] = ps_get_ts_arg(&ps
, colorarg
[2]);
1486 ps_do_ts_op(&ps
, key
->ts
[s
].colorop
, dst
, arg
);
1488 if (dst
.WriteMask
!= TGSI_WRITEMASK_XYZW
) {
1489 dst
.WriteMask
= TGSI_WRITEMASK_W
;
1491 if (used_a
& 0x1) arg
[0] = ps_get_ts_arg(&ps
, alphaarg
[0]);
1492 if (used_a
& 0x2) arg
[1] = ps_get_ts_arg(&ps
, alphaarg
[1]);
1493 if (used_a
& 0x4) arg
[2] = ps_get_ts_arg(&ps
, alphaarg
[2]);
1494 ps_do_ts_op(&ps
, key
->ts
[s
].alphaop
, dst
, arg
);
1499 ureg_ADD(ureg
, ureg_writemask(ps
.rCur
, TGSI_WRITEMASK_XYZ
), ps
.rCurSrc
, ps
.vC
[1]);
1503 if (key
->fog_mode
) {
1504 struct ureg_dst rFog
= ureg_writemask(ps
.rTmp
, TGSI_WRITEMASK_X
);
1505 struct ureg_src vPos
;
1506 if (device
->screen
->get_param(device
->screen
,
1507 PIPE_CAP_TGSI_FS_POSITION_IS_SYSVAL
)) {
1508 vPos
= ureg_DECL_system_value(ureg
, TGSI_SEMANTIC_POSITION
, 0);
1510 vPos
= ureg_DECL_fs_input(ureg
, TGSI_SEMANTIC_POSITION
, 0,
1511 TGSI_INTERPOLATE_LINEAR
);
1514 /* Source is either W or Z.
1515 * When we use vs ff,
1516 * Z is when an orthogonal projection matrix is detected,
1518 * Z is used for programmable vs.
1519 * Note: Tests indicate that the projection matrix coefficients do
1520 * actually affect pixel fog (and not vertex fog) when vs ff is used,
1521 * which justifies taking the position's w instead of taking the z coordinate
1522 * before the projection in the vs shader.
1524 if (!key
->fog_source
)
1525 ureg_MOV(ureg
, rFog
, _ZZZZ(vPos
));
1527 /* Position's w is 1/w */
1528 ureg_RCP(ureg
, rFog
, _WWWW(vPos
));
1530 if (key
->fog_mode
== D3DFOG_EXP
) {
1531 ureg_MUL(ureg
, rFog
, _X(rFog
), _ZZZZ(_CONST(22)));
1532 ureg_MUL(ureg
, rFog
, _X(rFog
), ureg_imm1f(ureg
, -1.442695f
));
1533 ureg_EX2(ureg
, rFog
, _X(rFog
));
1535 if (key
->fog_mode
== D3DFOG_EXP2
) {
1536 ureg_MUL(ureg
, rFog
, _X(rFog
), _ZZZZ(_CONST(22)));
1537 ureg_MUL(ureg
, rFog
, _X(rFog
), _X(rFog
));
1538 ureg_MUL(ureg
, rFog
, _X(rFog
), ureg_imm1f(ureg
, -1.442695f
));
1539 ureg_EX2(ureg
, rFog
, _X(rFog
));
1541 if (key
->fog_mode
== D3DFOG_LINEAR
) {
1542 ureg_ADD(ureg
, rFog
, _XXXX(_CONST(22)), ureg_negate(_X(rFog
)));
1543 ureg_MUL(ureg
, ureg_saturate(rFog
), _X(rFog
), _YYYY(_CONST(22)));
1545 ureg_LRP(ureg
, ureg_writemask(oCol
, TGSI_WRITEMASK_XYZ
), _X(rFog
), ps
.rCurSrc
, _CONST(21));
1546 ureg_MOV(ureg
, ureg_writemask(oCol
, TGSI_WRITEMASK_W
), ps
.rCurSrc
);
1549 struct ureg_src vFog
= ureg_DECL_fs_input(ureg
, TGSI_SEMANTIC_FOG
, 0, TGSI_INTERPOLATE_PERSPECTIVE
);
1550 ureg_LRP(ureg
, ureg_writemask(oCol
, TGSI_WRITEMASK_XYZ
), _XXXX(vFog
), ps
.rCurSrc
, _CONST(21));
1551 ureg_MOV(ureg
, ureg_writemask(oCol
, TGSI_WRITEMASK_W
), ps
.rCurSrc
);
1553 ureg_MOV(ureg
, oCol
, ps
.rCurSrc
);
1557 nine_ureg_tgsi_dump(ureg
, FALSE
);
1558 return ureg_create_shader_and_destroy(ureg
, device
->context
.pipe
);
1561 static struct NineVertexShader9
*
1562 nine_ff_get_vs(struct NineDevice9
*device
)
1564 const struct nine_context
*context
= &device
->context
;
1565 struct NineVertexShader9
*vs
;
1566 enum pipe_error err
;
1567 struct vs_build_ctx bld
;
1568 struct nine_ff_vs_key key
;
1570 boolean has_indexes
= false;
1571 boolean has_weights
= false;
1572 char input_texture_coord
[8];
1574 assert(sizeof(key
) <= sizeof(key
.value32
));
1576 memset(&key
, 0, sizeof(key
));
1577 memset(&bld
, 0, sizeof(bld
));
1578 memset(&input_texture_coord
, 0, sizeof(input_texture_coord
));
1582 /* FIXME: this shouldn't be NULL, but it is on init */
1583 if (context
->vdecl
) {
1584 key
.color0in_one
= 1;
1585 key
.color1in_zero
= 1;
1586 for (i
= 0; i
< context
->vdecl
->nelems
; i
++) {
1587 uint16_t usage
= context
->vdecl
->usage_map
[i
];
1588 if (usage
== NINE_DECLUSAGE_POSITIONT
)
1590 else if (usage
== NINE_DECLUSAGE_i(COLOR
, 0))
1591 key
.color0in_one
= 0;
1592 else if (usage
== NINE_DECLUSAGE_i(COLOR
, 1))
1593 key
.color1in_zero
= 0;
1594 else if (usage
== NINE_DECLUSAGE_i(BLENDINDICES
, 0)) {
1596 key
.passthrough
|= 1 << usage
;
1597 } else if (usage
== NINE_DECLUSAGE_i(BLENDWEIGHT
, 0)) {
1599 key
.passthrough
|= 1 << usage
;
1600 } else if (usage
== NINE_DECLUSAGE_i(NORMAL
, 0)) {
1602 key
.passthrough
|= 1 << usage
;
1603 } else if (usage
== NINE_DECLUSAGE_PSIZE
)
1604 key
.vertexpointsize
= 1;
1605 else if (usage
% NINE_DECLUSAGE_COUNT
== NINE_DECLUSAGE_TEXCOORD
) {
1606 s
= usage
/ NINE_DECLUSAGE_COUNT
;
1608 input_texture_coord
[s
] = nine_decltype_get_dim(context
->vdecl
->decls
[i
].Type
);
1610 DBG("FF given texture coordinate >= 8. Ignoring\n");
1611 } else if (usage
< NINE_DECLUSAGE_NONE
)
1612 key
.passthrough
|= 1 << usage
;
1615 /* ff vs + ps 3.0: some elements are passed to the ps (wine test).
1616 * We do restrict to indices 0 */
1617 key
.passthrough
&= ~((1 << NINE_DECLUSAGE_POSITION
) | (1 << NINE_DECLUSAGE_PSIZE
) |
1618 (1 << NINE_DECLUSAGE_TEXCOORD
) | (1 << NINE_DECLUSAGE_POSITIONT
) |
1619 (1 << NINE_DECLUSAGE_TESSFACTOR
) | (1 << NINE_DECLUSAGE_SAMPLE
));
1620 if (!key
.position_t
)
1621 key
.passthrough
= 0;
1622 key
.pointscale
= !!context
->rs
[D3DRS_POINTSCALEENABLE
];
1624 key
.lighting
= !!context
->rs
[D3DRS_LIGHTING
] && context
->ff
.num_lights_active
;
1625 key
.darkness
= !!context
->rs
[D3DRS_LIGHTING
] && !context
->ff
.num_lights_active
;
1626 if (key
.position_t
) {
1627 key
.darkness
= 0; /* |= key.lighting; */ /* XXX ? */
1630 if ((key
.lighting
| key
.darkness
) && context
->rs
[D3DRS_COLORVERTEX
]) {
1631 uint32_t mask
= (key
.color0in_one
? 0 : 1) | (key
.color1in_zero
? 0 : 2);
1632 key
.mtl_diffuse
= context
->rs
[D3DRS_DIFFUSEMATERIALSOURCE
] & mask
;
1633 key
.mtl_ambient
= context
->rs
[D3DRS_AMBIENTMATERIALSOURCE
] & mask
;
1634 key
.mtl_specular
= context
->rs
[D3DRS_SPECULARMATERIALSOURCE
] & mask
;
1635 key
.mtl_emissive
= context
->rs
[D3DRS_EMISSIVEMATERIALSOURCE
] & mask
;
1637 key
.fog
= !!context
->rs
[D3DRS_FOGENABLE
];
1638 key
.fog_mode
= (!key
.position_t
&& context
->rs
[D3DRS_FOGENABLE
]) ? context
->rs
[D3DRS_FOGVERTEXMODE
] : 0;
1640 key
.fog_range
= context
->rs
[D3DRS_RANGEFOGENABLE
];
1642 key
.localviewer
= !!context
->rs
[D3DRS_LOCALVIEWER
];
1643 key
.normalizenormals
= !!context
->rs
[D3DRS_NORMALIZENORMALS
];
1644 key
.ucp
= !!context
->rs
[D3DRS_CLIPPLANEENABLE
];
1646 if (context
->rs
[D3DRS_VERTEXBLEND
] != D3DVBF_DISABLE
) {
1647 key
.vertexblend_indexed
= !!context
->rs
[D3DRS_INDEXEDVERTEXBLENDENABLE
] && has_indexes
;
1649 switch (context
->rs
[D3DRS_VERTEXBLEND
]) {
1650 case D3DVBF_0WEIGHTS
: key
.vertexblend
= key
.vertexblend_indexed
; break;
1651 case D3DVBF_1WEIGHTS
: key
.vertexblend
= 2; break;
1652 case D3DVBF_2WEIGHTS
: key
.vertexblend
= 3; break;
1653 case D3DVBF_3WEIGHTS
: key
.vertexblend
= 4; break;
1654 case D3DVBF_TWEENING
: key
.vertextween
= 1; break;
1656 assert(!"invalid D3DVBF");
1659 if (!has_weights
&& context
->rs
[D3DRS_VERTEXBLEND
] != D3DVBF_0WEIGHTS
)
1660 key
.vertexblend
= 0; /* TODO: if key.vertexblend_indexed, perhaps it should use 1.0 as weight, or revert to D3DVBF_0WEIGHTS */
1663 for (s
= 0; s
< 8; ++s
) {
1664 unsigned gen
= (context
->ff
.tex_stage
[s
][D3DTSS_TEXCOORDINDEX
] >> 16) + 1;
1665 unsigned idx
= context
->ff
.tex_stage
[s
][D3DTSS_TEXCOORDINDEX
] & 7;
1668 if (key
.position_t
&& gen
> NINED3DTSS_TCI_PASSTHRU
)
1669 gen
= NINED3DTSS_TCI_PASSTHRU
;
1671 if (!input_texture_coord
[idx
] && gen
== NINED3DTSS_TCI_PASSTHRU
)
1672 gen
= NINED3DTSS_TCI_DISABLE
;
1674 key
.tc_gen
|= gen
<< (s
* 3);
1675 key
.tc_idx
|= idx
<< (s
* 3);
1676 key
.tc_dim_input
|= ((input_texture_coord
[idx
]-1) & 0x3) << (s
* 2);
1678 dim
= context
->ff
.tex_stage
[s
][D3DTSS_TEXTURETRANSFORMFLAGS
] & 0x7;
1680 dim
= input_texture_coord
[idx
];
1681 if (dim
== 1) /* NV behaviour */
1683 key
.tc_dim_output
|= dim
<< (s
* 3);
1686 DBG("VS ff key hash: %x\n", nine_ff_vs_key_hash(&key
));
1687 vs
= util_hash_table_get(device
->ff
.ht_vs
, &key
);
1690 NineVertexShader9_new(device
, &vs
, NULL
, nine_ff_build_vs(device
, &bld
));
1692 nine_ff_prune_vs(device
);
1696 memcpy(&vs
->ff_key
, &key
, sizeof(vs
->ff_key
));
1698 err
= util_hash_table_set(device
->ff
.ht_vs
, &vs
->ff_key
, vs
);
1700 assert(err
== PIPE_OK
);
1701 device
->ff
.num_vs
++;
1703 vs
->num_inputs
= bld
.num_inputs
;
1704 for (n
= 0; n
< bld
.num_inputs
; ++n
)
1705 vs
->input_map
[n
].ndecl
= bld
.input
[n
];
1707 vs
->position_t
= key
.position_t
;
1708 vs
->point_size
= key
.vertexpointsize
| key
.pointscale
;
1713 #define GET_D3DTS(n) nine_state_access_transform(&context->ff, D3DTS_##n, FALSE)
1714 #define IS_D3DTS_DIRTY(s,n) ((s)->ff.changed.transform[(D3DTS_##n) / 32] & (1 << ((D3DTS_##n) % 32)))
1716 static struct NinePixelShader9
*
1717 nine_ff_get_ps(struct NineDevice9
*device
)
1719 struct nine_context
*context
= &device
->context
;
1720 D3DMATRIX
*projection_matrix
= GET_D3DTS(PROJECTION
);
1721 struct NinePixelShader9
*ps
;
1722 enum pipe_error err
;
1723 struct nine_ff_ps_key key
;
1725 uint8_t sampler_mask
= 0;
1727 assert(sizeof(key
) <= sizeof(key
.value32
));
1729 memset(&key
, 0, sizeof(key
));
1730 for (s
= 0; s
< 8; ++s
) {
1731 key
.ts
[s
].colorop
= context
->ff
.tex_stage
[s
][D3DTSS_COLOROP
];
1732 key
.ts
[s
].alphaop
= context
->ff
.tex_stage
[s
][D3DTSS_ALPHAOP
];
1733 const uint8_t used_c
= ps_d3dtop_args_mask(key
.ts
[s
].colorop
);
1734 const uint8_t used_a
= ps_d3dtop_args_mask(key
.ts
[s
].alphaop
);
1735 /* MSDN says D3DTOP_DISABLE disables this and all subsequent stages.
1736 * ALPHAOP cannot be enabled if COLOROP is disabled.
1737 * Verified on Windows. */
1738 if (key
.ts
[s
].colorop
== D3DTOP_DISABLE
) {
1739 key
.ts
[s
].alphaop
= D3DTOP_DISABLE
; /* DISABLE == 1, avoid degenerate keys */
1743 if (!context
->texture
[s
].enabled
&&
1744 ((context
->ff
.tex_stage
[s
][D3DTSS_COLORARG0
] == D3DTA_TEXTURE
&&
1746 (context
->ff
.tex_stage
[s
][D3DTSS_COLORARG1
] == D3DTA_TEXTURE
&&
1748 (context
->ff
.tex_stage
[s
][D3DTSS_COLORARG2
] == D3DTA_TEXTURE
&&
1750 /* Tested on Windows: Invalid texture read disables the stage
1751 * and the subsequent ones, but only for colorop. For alpha,
1752 * it's as if the texture had alpha of 1.0, which is what
1753 * has our dummy texture in that case. Invalid color also
1754 * disabled the following alpha stages. */
1755 key
.ts
[s
].colorop
= key
.ts
[s
].alphaop
= D3DTOP_DISABLE
;
1759 if (context
->ff
.tex_stage
[s
][D3DTSS_COLORARG0
] == D3DTA_TEXTURE
||
1760 context
->ff
.tex_stage
[s
][D3DTSS_COLORARG1
] == D3DTA_TEXTURE
||
1761 context
->ff
.tex_stage
[s
][D3DTSS_COLORARG2
] == D3DTA_TEXTURE
||
1762 context
->ff
.tex_stage
[s
][D3DTSS_ALPHAARG0
] == D3DTA_TEXTURE
||
1763 context
->ff
.tex_stage
[s
][D3DTSS_ALPHAARG1
] == D3DTA_TEXTURE
||
1764 context
->ff
.tex_stage
[s
][D3DTSS_ALPHAARG2
] == D3DTA_TEXTURE
)
1765 sampler_mask
|= (1 << s
);
1767 if (key
.ts
[s
].colorop
!= D3DTOP_DISABLE
) {
1768 if (used_c
& 0x1) key
.ts
[s
].colorarg0
= context
->ff
.tex_stage
[s
][D3DTSS_COLORARG0
] & 0x7;
1769 if (used_c
& 0x2) key
.ts
[s
].colorarg1
= context
->ff
.tex_stage
[s
][D3DTSS_COLORARG1
] & 0x7;
1770 if (used_c
& 0x4) key
.ts
[s
].colorarg2
= context
->ff
.tex_stage
[s
][D3DTSS_COLORARG2
] & 0x7;
1771 if (used_c
& 0x1) key
.colorarg_b4
[0] |= ((context
->ff
.tex_stage
[s
][D3DTSS_COLORARG0
] >> 4) & 0x1) << s
;
1772 if (used_c
& 0x1) key
.colorarg_b5
[0] |= ((context
->ff
.tex_stage
[s
][D3DTSS_COLORARG0
] >> 5) & 0x1) << s
;
1773 if (used_c
& 0x2) key
.colorarg_b4
[1] |= ((context
->ff
.tex_stage
[s
][D3DTSS_COLORARG1
] >> 4) & 0x1) << s
;
1774 if (used_c
& 0x2) key
.colorarg_b5
[1] |= ((context
->ff
.tex_stage
[s
][D3DTSS_COLORARG1
] >> 5) & 0x1) << s
;
1775 if (used_c
& 0x4) key
.colorarg_b4
[2] |= ((context
->ff
.tex_stage
[s
][D3DTSS_COLORARG2
] >> 4) & 0x1) << s
;
1776 if (used_c
& 0x4) key
.colorarg_b5
[2] |= ((context
->ff
.tex_stage
[s
][D3DTSS_COLORARG2
] >> 5) & 0x1) << s
;
1778 if (key
.ts
[s
].alphaop
!= D3DTOP_DISABLE
) {
1779 if (used_a
& 0x1) key
.ts
[s
].alphaarg0
= context
->ff
.tex_stage
[s
][D3DTSS_ALPHAARG0
] & 0x7;
1780 if (used_a
& 0x2) key
.ts
[s
].alphaarg1
= context
->ff
.tex_stage
[s
][D3DTSS_ALPHAARG1
] & 0x7;
1781 if (used_a
& 0x4) key
.ts
[s
].alphaarg2
= context
->ff
.tex_stage
[s
][D3DTSS_ALPHAARG2
] & 0x7;
1782 if (used_a
& 0x1) key
.alphaarg_b4
[0] |= ((context
->ff
.tex_stage
[s
][D3DTSS_ALPHAARG0
] >> 4) & 0x1) << s
;
1783 if (used_a
& 0x2) key
.alphaarg_b4
[1] |= ((context
->ff
.tex_stage
[s
][D3DTSS_ALPHAARG1
] >> 4) & 0x1) << s
;
1784 if (used_a
& 0x4) key
.alphaarg_b4
[2] |= ((context
->ff
.tex_stage
[s
][D3DTSS_ALPHAARG2
] >> 4) & 0x1) << s
;
1786 key
.ts
[s
].resultarg
= context
->ff
.tex_stage
[s
][D3DTSS_RESULTARG
] == D3DTA_TEMP
;
1788 if (context
->texture
[s
].enabled
) {
1789 switch (context
->texture
[s
].type
) {
1790 case D3DRTYPE_TEXTURE
: key
.ts
[s
].textarget
= 1; break;
1791 case D3DRTYPE_VOLUMETEXTURE
: key
.ts
[s
].textarget
= 2; break;
1792 case D3DRTYPE_CUBETEXTURE
: key
.ts
[s
].textarget
= 3; break;
1794 assert(!"unexpected texture type");
1798 key
.ts
[s
].textarget
= 1;
1802 /* Note: If colorop is D3DTOP_DISABLE for the first stage
1803 * (which implies alphaop is too), nothing particular happens,
1804 * that is, current is equal to diffuse (which is the case anyway,
1805 * because it is how it is initialized).
1806 * Special case seems if alphaop is D3DTOP_DISABLE and not colorop,
1807 * because then if the resultarg is TEMP, then diffuse alpha is written
1809 if (key
.ts
[0].colorop
!= D3DTOP_DISABLE
&&
1810 key
.ts
[0].alphaop
== D3DTOP_DISABLE
&&
1811 key
.ts
[0].resultarg
!= 0) {
1812 key
.ts
[0].alphaop
= D3DTOP_SELECTARG1
;
1813 key
.ts
[0].alphaarg1
= D3DTA_DIFFUSE
;
1815 /* When no alpha stage writes to current, diffuse alpha is taken.
1816 * Since we initialize current to diffuse, we have the behaviour. */
1818 /* Last stage always writes to Current */
1820 key
.ts
[s
-1].resultarg
= 0;
1822 key
.projected
= nine_ff_get_projected_key(context
);
1823 key
.specular
= !!context
->rs
[D3DRS_SPECULARENABLE
];
1826 key
.ts
[s
].colorop
= key
.ts
[s
].alphaop
= D3DTOP_DISABLE
;
1827 if (context
->rs
[D3DRS_FOGENABLE
])
1828 key
.fog_mode
= context
->rs
[D3DRS_FOGTABLEMODE
];
1829 key
.fog
= !!context
->rs
[D3DRS_FOGENABLE
];
1830 /* Pixel fog (with WFOG advertised): source is either Z or W.
1831 * W is the source if vs ff is used, and the
1832 * projection matrix is not orthogonal.
1833 * Tests on Win 10 seem to indicate _34
1834 * and _33 are checked against 0, 1. */
1835 if (key
.fog_mode
&& key
.fog
)
1836 key
.fog_source
= !context
->programmable_vs
&&
1837 !(projection_matrix
->_34
== 0.0f
&&
1838 projection_matrix
->_44
== 1.0f
);
1840 DBG("PS ff key hash: %x\n", nine_ff_ps_key_hash(&key
));
1841 ps
= util_hash_table_get(device
->ff
.ht_ps
, &key
);
1844 NinePixelShader9_new(device
, &ps
, NULL
, nine_ff_build_ps(device
, &key
));
1846 nine_ff_prune_ps(device
);
1848 memcpy(&ps
->ff_key
, &key
, sizeof(ps
->ff_key
));
1850 err
= util_hash_table_set(device
->ff
.ht_ps
, &ps
->ff_key
, ps
);
1852 assert(err
== PIPE_OK
);
1853 device
->ff
.num_ps
++;
1856 ps
->sampler_mask
= sampler_mask
;
1862 nine_ff_load_vs_transforms(struct NineDevice9
*device
)
1864 struct nine_context
*context
= &device
->context
;
1866 D3DMATRIX
*M
= (D3DMATRIX
*)device
->ff
.vs_const
;
1869 /* TODO: make this nicer, and only upload the ones we need */
1870 /* TODO: use ff.vs_const as storage of W, V, P matrices */
1872 if (IS_D3DTS_DIRTY(context
, WORLD
) ||
1873 IS_D3DTS_DIRTY(context
, VIEW
) ||
1874 IS_D3DTS_DIRTY(context
, PROJECTION
)) {
1875 /* WVP, WV matrices */
1876 nine_d3d_matrix_matrix_mul(&M
[1], GET_D3DTS(WORLD
), GET_D3DTS(VIEW
));
1877 nine_d3d_matrix_matrix_mul(&M
[0], &M
[1], GET_D3DTS(PROJECTION
));
1879 /* normal matrix == transpose(inverse(WV)) */
1880 nine_d3d_matrix_inverse(&T
, &M
[1]);
1881 nine_d3d_matrix_transpose(&M
[4], &T
);
1884 M
[2] = *GET_D3DTS(PROJECTION
);
1886 /* V and W matrix */
1887 nine_d3d_matrix_inverse(&M
[3], GET_D3DTS(VIEW
));
1891 if (context
->rs
[D3DRS_VERTEXBLEND
] != D3DVBF_DISABLE
) {
1892 /* load other world matrices */
1893 for (i
= 1; i
<= 8; ++i
) {
1894 nine_d3d_matrix_matrix_mul(&M
[40 + i
], GET_D3DTS(WORLDMATRIX(i
)), GET_D3DTS(VIEW
));
1898 device
->ff
.vs_const
[30 * 4] = asfloat(context
->rs
[D3DRS_TWEENFACTOR
]);
1902 nine_ff_load_lights(struct NineDevice9
*device
)
1904 struct nine_context
*context
= &device
->context
;
1905 struct fvec4
*dst
= (struct fvec4
*)device
->ff
.vs_const
;
1908 if (context
->changed
.group
& NINE_STATE_FF_MATERIAL
) {
1909 const D3DMATERIAL9
*mtl
= &context
->ff
.material
;
1911 memcpy(&dst
[20], &mtl
->Diffuse
, 4 * sizeof(float));
1912 memcpy(&dst
[21], &mtl
->Ambient
, 4 * sizeof(float));
1913 memcpy(&dst
[22], &mtl
->Specular
, 4 * sizeof(float));
1914 dst
[23].x
= mtl
->Power
;
1915 memcpy(&dst
[24], &mtl
->Emissive
, 4 * sizeof(float));
1916 d3dcolor_to_rgba(&dst
[25].x
, context
->rs
[D3DRS_AMBIENT
]);
1917 dst
[19].x
= dst
[25].x
* mtl
->Ambient
.r
+ mtl
->Emissive
.r
;
1918 dst
[19].y
= dst
[25].y
* mtl
->Ambient
.g
+ mtl
->Emissive
.g
;
1919 dst
[19].z
= dst
[25].z
* mtl
->Ambient
.b
+ mtl
->Emissive
.b
;
1922 if (!(context
->changed
.group
& NINE_STATE_FF_LIGHTING
))
1925 for (l
= 0; l
< context
->ff
.num_lights_active
; ++l
) {
1926 const D3DLIGHT9
*light
= &context
->ff
.light
[context
->ff
.active_light
[l
]];
1928 dst
[32 + l
* 8].x
= light
->Type
;
1929 dst
[32 + l
* 8].y
= light
->Attenuation0
;
1930 dst
[32 + l
* 8].z
= light
->Attenuation1
;
1931 dst
[32 + l
* 8].w
= light
->Attenuation2
;
1932 memcpy(&dst
[33 + l
* 8].x
, &light
->Diffuse
, sizeof(light
->Diffuse
));
1933 memcpy(&dst
[34 + l
* 8].x
, &light
->Specular
, sizeof(light
->Specular
));
1934 memcpy(&dst
[35 + l
* 8].x
, &light
->Ambient
, sizeof(light
->Ambient
));
1935 nine_d3d_vector4_matrix_mul((D3DVECTOR
*)&dst
[36 + l
* 8].x
, &light
->Position
, GET_D3DTS(VIEW
));
1936 nine_d3d_vector3_matrix_mul((D3DVECTOR
*)&dst
[37 + l
* 8].x
, &light
->Direction
, GET_D3DTS(VIEW
));
1937 dst
[36 + l
* 8].w
= light
->Type
== D3DLIGHT_DIRECTIONAL
? 1e9f
: light
->Range
;
1938 dst
[37 + l
* 8].w
= light
->Falloff
;
1939 dst
[38 + l
* 8].x
= cosf(light
->Theta
* 0.5f
);
1940 dst
[38 + l
* 8].y
= cosf(light
->Phi
* 0.5f
);
1941 dst
[38 + l
* 8].z
= 1.0f
/ (dst
[38 + l
* 8].x
- dst
[38 + l
* 8].y
);
1942 dst
[39 + l
* 8].w
= (float)((l
+ 1) == context
->ff
.num_lights_active
);
1947 nine_ff_load_point_and_fog_params(struct NineDevice9
*device
)
1949 struct nine_context
*context
= &device
->context
;
1950 struct fvec4
*dst
= (struct fvec4
*)device
->ff
.vs_const
;
1952 if (!(context
->changed
.group
& NINE_STATE_FF_VS_OTHER
))
1954 dst
[26].x
= asfloat(context
->rs
[D3DRS_POINTSIZE_MIN
]);
1955 dst
[26].y
= asfloat(context
->rs
[D3DRS_POINTSIZE_MAX
]);
1956 dst
[26].z
= asfloat(context
->rs
[D3DRS_POINTSIZE
]);
1957 dst
[26].w
= asfloat(context
->rs
[D3DRS_POINTSCALE_A
]);
1958 dst
[27].x
= asfloat(context
->rs
[D3DRS_POINTSCALE_B
]);
1959 dst
[27].y
= asfloat(context
->rs
[D3DRS_POINTSCALE_C
]);
1960 dst
[28].x
= asfloat(context
->rs
[D3DRS_FOGEND
]);
1961 dst
[28].y
= 1.0f
/ (asfloat(context
->rs
[D3DRS_FOGEND
]) - asfloat(context
->rs
[D3DRS_FOGSTART
]));
1962 if (isinf(dst
[28].y
))
1964 dst
[28].z
= asfloat(context
->rs
[D3DRS_FOGDENSITY
]);
1968 nine_ff_load_tex_matrices(struct NineDevice9
*device
)
1970 struct nine_context
*context
= &device
->context
;
1971 D3DMATRIX
*M
= (D3DMATRIX
*)device
->ff
.vs_const
;
1974 if (!(context
->ff
.changed
.transform
[0] & 0xff0000))
1976 for (s
= 0; s
< 8; ++s
) {
1977 if (IS_D3DTS_DIRTY(context
, TEXTURE0
+ s
))
1978 nine_d3d_matrix_transpose(&M
[32 + s
], nine_state_access_transform(&context
->ff
, D3DTS_TEXTURE0
+ s
, FALSE
));
1983 nine_ff_load_ps_params(struct NineDevice9
*device
)
1985 struct nine_context
*context
= &device
->context
;
1986 struct fvec4
*dst
= (struct fvec4
*)device
->ff
.ps_const
;
1989 if (!(context
->changed
.group
& NINE_STATE_FF_PS_CONSTS
))
1992 for (s
= 0; s
< 8; ++s
)
1993 d3dcolor_to_rgba(&dst
[s
].x
, context
->ff
.tex_stage
[s
][D3DTSS_CONSTANT
]);
1995 for (s
= 0; s
< 8; ++s
) {
1996 dst
[8 + s
].x
= asfloat(context
->ff
.tex_stage
[s
][D3DTSS_BUMPENVMAT00
]);
1997 dst
[8 + s
].y
= asfloat(context
->ff
.tex_stage
[s
][D3DTSS_BUMPENVMAT01
]);
1998 dst
[8 + s
].z
= asfloat(context
->ff
.tex_stage
[s
][D3DTSS_BUMPENVMAT10
]);
1999 dst
[8 + s
].w
= asfloat(context
->ff
.tex_stage
[s
][D3DTSS_BUMPENVMAT11
]);
2001 dst
[16 + s
/ 2].z
= asfloat(context
->ff
.tex_stage
[s
][D3DTSS_BUMPENVLSCALE
]);
2002 dst
[16 + s
/ 2].w
= asfloat(context
->ff
.tex_stage
[s
][D3DTSS_BUMPENVLOFFSET
]);
2004 dst
[16 + s
/ 2].x
= asfloat(context
->ff
.tex_stage
[s
][D3DTSS_BUMPENVLSCALE
]);
2005 dst
[16 + s
/ 2].y
= asfloat(context
->ff
.tex_stage
[s
][D3DTSS_BUMPENVLOFFSET
]);
2009 d3dcolor_to_rgba(&dst
[20].x
, context
->rs
[D3DRS_TEXTUREFACTOR
]);
2010 d3dcolor_to_rgba(&dst
[21].x
, context
->rs
[D3DRS_FOGCOLOR
]);
2011 dst
[22].x
= asfloat(context
->rs
[D3DRS_FOGEND
]);
2012 dst
[22].y
= 1.0f
/ (asfloat(context
->rs
[D3DRS_FOGEND
]) - asfloat(context
->rs
[D3DRS_FOGSTART
]));
2013 dst
[22].z
= asfloat(context
->rs
[D3DRS_FOGDENSITY
]);
2017 nine_ff_load_viewport_info(struct NineDevice9
*device
)
2019 D3DVIEWPORT9
*viewport
= &device
->context
.viewport
;
2020 struct fvec4
*dst
= (struct fvec4
*)device
->ff
.vs_const
;
2021 float diffZ
= viewport
->MaxZ
- viewport
->MinZ
;
2023 /* Note: the other functions avoids to fill the const again if nothing changed.
2024 * But we don't have much to fill, and adding code to allow that may be complex
2025 * so just fill it always */
2026 dst
[100].x
= 2.0f
/ (float)(viewport
->Width
);
2027 dst
[100].y
= 2.0f
/ (float)(viewport
->Height
);
2028 dst
[100].z
= (diffZ
== 0.0f
) ? 0.0f
: (1.0f
/ diffZ
);
2029 dst
[100].w
= (float)(viewport
->Width
);
2030 dst
[101].x
= (float)(viewport
->X
);
2031 dst
[101].y
= (float)(viewport
->Y
);
2032 dst
[101].z
= (float)(viewport
->MinZ
);
2036 nine_ff_update(struct NineDevice9
*device
)
2038 struct nine_context
*context
= &device
->context
;
2039 struct pipe_constant_buffer cb
;
2041 DBG("vs=%p ps=%p\n", context
->vs
, context
->ps
);
2043 /* NOTE: the only reference belongs to the hash table */
2044 if (!context
->programmable_vs
) {
2045 device
->ff
.vs
= nine_ff_get_vs(device
);
2046 context
->changed
.group
|= NINE_STATE_VS
;
2049 device
->ff
.ps
= nine_ff_get_ps(device
);
2050 context
->changed
.group
|= NINE_STATE_PS
;
2053 if (!context
->programmable_vs
) {
2054 nine_ff_load_vs_transforms(device
);
2055 nine_ff_load_tex_matrices(device
);
2056 nine_ff_load_lights(device
);
2057 nine_ff_load_point_and_fog_params(device
);
2058 nine_ff_load_viewport_info(device
);
2060 memset(context
->ff
.changed
.transform
, 0, sizeof(context
->ff
.changed
.transform
));
2062 cb
.buffer_offset
= 0;
2064 cb
.user_buffer
= device
->ff
.vs_const
;
2065 cb
.buffer_size
= NINE_FF_NUM_VS_CONST
* 4 * sizeof(float);
2067 context
->pipe_data
.cb_vs_ff
= cb
;
2068 context
->commit
|= NINE_STATE_COMMIT_CONST_VS
;
2070 context
->changed
.group
&= ~NINE_STATE_FF_VS
;
2074 nine_ff_load_ps_params(device
);
2076 cb
.buffer_offset
= 0;
2078 cb
.user_buffer
= device
->ff
.ps_const
;
2079 cb
.buffer_size
= NINE_FF_NUM_PS_CONST
* 4 * sizeof(float);
2081 context
->pipe_data
.cb_ps_ff
= cb
;
2082 context
->commit
|= NINE_STATE_COMMIT_CONST_PS
;
2084 context
->changed
.group
&= ~NINE_STATE_FF_PS
;
2090 nine_ff_init(struct NineDevice9
*device
)
2092 device
->ff
.ht_vs
= util_hash_table_create(nine_ff_vs_key_hash
,
2093 nine_ff_vs_key_comp
);
2094 device
->ff
.ht_ps
= util_hash_table_create(nine_ff_ps_key_hash
,
2095 nine_ff_ps_key_comp
);
2097 device
->ff
.ht_fvf
= util_hash_table_create(nine_ff_fvf_key_hash
,
2098 nine_ff_fvf_key_comp
);
2100 device
->ff
.vs_const
= CALLOC(NINE_FF_NUM_VS_CONST
, 4 * sizeof(float));
2101 device
->ff
.ps_const
= CALLOC(NINE_FF_NUM_PS_CONST
, 4 * sizeof(float));
2103 return device
->ff
.ht_vs
&& device
->ff
.ht_ps
&&
2104 device
->ff
.ht_fvf
&&
2105 device
->ff
.vs_const
&& device
->ff
.ps_const
;
2108 static enum pipe_error
nine_ff_ht_delete_cb(void *key
, void *value
, void *data
)
2110 NineUnknown_Unbind(NineUnknown(value
));
2115 nine_ff_fini(struct NineDevice9
*device
)
2117 if (device
->ff
.ht_vs
) {
2118 util_hash_table_foreach(device
->ff
.ht_vs
, nine_ff_ht_delete_cb
, NULL
);
2119 util_hash_table_destroy(device
->ff
.ht_vs
);
2121 if (device
->ff
.ht_ps
) {
2122 util_hash_table_foreach(device
->ff
.ht_ps
, nine_ff_ht_delete_cb
, NULL
);
2123 util_hash_table_destroy(device
->ff
.ht_ps
);
2125 if (device
->ff
.ht_fvf
) {
2126 util_hash_table_foreach(device
->ff
.ht_fvf
, nine_ff_ht_delete_cb
, NULL
);
2127 util_hash_table_destroy(device
->ff
.ht_fvf
);
2129 device
->ff
.vs
= NULL
; /* destroyed by unbinding from hash table */
2130 device
->ff
.ps
= NULL
;
2132 FREE(device
->ff
.vs_const
);
2133 FREE(device
->ff
.ps_const
);
2137 nine_ff_prune_vs(struct NineDevice9
*device
)
2139 struct nine_context
*context
= &device
->context
;
2141 if (device
->ff
.num_vs
> 1024) {
2142 /* could destroy the bound one here, so unbind */
2143 context
->pipe
->bind_vs_state(context
->pipe
, NULL
);
2144 util_hash_table_foreach(device
->ff
.ht_vs
, nine_ff_ht_delete_cb
, NULL
);
2145 util_hash_table_clear(device
->ff
.ht_vs
);
2146 device
->ff
.num_vs
= 0;
2147 context
->changed
.group
|= NINE_STATE_VS
;
2151 nine_ff_prune_ps(struct NineDevice9
*device
)
2153 struct nine_context
*context
= &device
->context
;
2155 if (device
->ff
.num_ps
> 1024) {
2156 /* could destroy the bound one here, so unbind */
2157 context
->pipe
->bind_fs_state(context
->pipe
, NULL
);
2158 util_hash_table_foreach(device
->ff
.ht_ps
, nine_ff_ht_delete_cb
, NULL
);
2159 util_hash_table_clear(device
->ff
.ht_ps
);
2160 device
->ff
.num_ps
= 0;
2161 context
->changed
.group
|= NINE_STATE_PS
;
2165 /* ========================================================================== */
2167 /* Matrix multiplication:
2169 * in memory: 0 1 2 3 (row major)
2175 * r0 = (r0 * cA) (r0 * cB) . .
2176 * r1 = (r1 * cA) (r1 * cB)
2180 * r: (11) (12) (13) (14)
2181 * (21) (22) (23) (24)
2182 * (31) (32) (33) (34)
2183 * (41) (42) (43) (44)
2191 * t.xyzw = MUL(v.xxxx, r[0]);
2192 * t.xyzw = MAD(v.yyyy, r[1], t.xyzw);
2193 * t.xyzw = MAD(v.zzzz, r[2], t.xyzw);
2194 * v.xyzw = MAD(v.wwww, r[3], t.xyzw);
2196 * v.x = DP4(v, c[0]);
2197 * v.y = DP4(v, c[1]);
2198 * v.z = DP4(v, c[2]);
2199 * v.w = DP4(v, c[3]) = 1
2204 nine_D3DMATRIX_print(const D3DMATRIX *M)
2206 DBG("\n(%f %f %f %f)\n"
2210 M->m[0][0], M->m[0][1], M->m[0][2], M->m[0][3],
2211 M->m[1][0], M->m[1][1], M->m[1][2], M->m[1][3],
2212 M->m[2][0], M->m[2][1], M->m[2][2], M->m[2][3],
2213 M->m[3][0], M->m[3][1], M->m[3][2], M->m[3][3]);
2218 nine_DP4_row_col(const D3DMATRIX
*A
, int r
, const D3DMATRIX
*B
, int c
)
2220 return A
->m
[r
][0] * B
->m
[0][c
] +
2221 A
->m
[r
][1] * B
->m
[1][c
] +
2222 A
->m
[r
][2] * B
->m
[2][c
] +
2223 A
->m
[r
][3] * B
->m
[3][c
];
2227 nine_DP4_vec_col(const D3DVECTOR
*v
, const D3DMATRIX
*M
, int c
)
2229 return v
->x
* M
->m
[0][c
] +
2236 nine_DP3_vec_col(const D3DVECTOR
*v
, const D3DMATRIX
*M
, int c
)
2238 return v
->x
* M
->m
[0][c
] +
2244 nine_d3d_matrix_matrix_mul(D3DMATRIX
*D
, const D3DMATRIX
*L
, const D3DMATRIX
*R
)
2246 D
->_11
= nine_DP4_row_col(L
, 0, R
, 0);
2247 D
->_12
= nine_DP4_row_col(L
, 0, R
, 1);
2248 D
->_13
= nine_DP4_row_col(L
, 0, R
, 2);
2249 D
->_14
= nine_DP4_row_col(L
, 0, R
, 3);
2251 D
->_21
= nine_DP4_row_col(L
, 1, R
, 0);
2252 D
->_22
= nine_DP4_row_col(L
, 1, R
, 1);
2253 D
->_23
= nine_DP4_row_col(L
, 1, R
, 2);
2254 D
->_24
= nine_DP4_row_col(L
, 1, R
, 3);
2256 D
->_31
= nine_DP4_row_col(L
, 2, R
, 0);
2257 D
->_32
= nine_DP4_row_col(L
, 2, R
, 1);
2258 D
->_33
= nine_DP4_row_col(L
, 2, R
, 2);
2259 D
->_34
= nine_DP4_row_col(L
, 2, R
, 3);
2261 D
->_41
= nine_DP4_row_col(L
, 3, R
, 0);
2262 D
->_42
= nine_DP4_row_col(L
, 3, R
, 1);
2263 D
->_43
= nine_DP4_row_col(L
, 3, R
, 2);
2264 D
->_44
= nine_DP4_row_col(L
, 3, R
, 3);
2268 nine_d3d_vector4_matrix_mul(D3DVECTOR
*d
, const D3DVECTOR
*v
, const D3DMATRIX
*M
)
2270 d
->x
= nine_DP4_vec_col(v
, M
, 0);
2271 d
->y
= nine_DP4_vec_col(v
, M
, 1);
2272 d
->z
= nine_DP4_vec_col(v
, M
, 2);
2276 nine_d3d_vector3_matrix_mul(D3DVECTOR
*d
, const D3DVECTOR
*v
, const D3DMATRIX
*M
)
2278 d
->x
= nine_DP3_vec_col(v
, M
, 0);
2279 d
->y
= nine_DP3_vec_col(v
, M
, 1);
2280 d
->z
= nine_DP3_vec_col(v
, M
, 2);
2284 nine_d3d_matrix_transpose(D3DMATRIX
*D
, const D3DMATRIX
*M
)
2287 for (i
= 0; i
< 4; ++i
)
2288 for (j
= 0; j
< 4; ++j
)
2289 D
->m
[i
][j
] = M
->m
[j
][i
];
2292 #define _M_ADD_PROD_1i_2j_3k_4l(i,j,k,l) do { \
2293 float t = M->_1##i * M->_2##j * M->_3##k * M->_4##l; \
2294 if (t > 0.0f) pos += t; else neg += t; } while(0)
2296 #define _M_SUB_PROD_1i_2j_3k_4l(i,j,k,l) do { \
2297 float t = M->_1##i * M->_2##j * M->_3##k * M->_4##l; \
2298 if (t > 0.0f) neg -= t; else pos -= t; } while(0)
2300 nine_d3d_matrix_det(const D3DMATRIX
*M
)
2305 _M_ADD_PROD_1i_2j_3k_4l(1, 2, 3, 4);
2306 _M_ADD_PROD_1i_2j_3k_4l(1, 3, 4, 2);
2307 _M_ADD_PROD_1i_2j_3k_4l(1, 4, 2, 3);
2309 _M_ADD_PROD_1i_2j_3k_4l(2, 1, 4, 3);
2310 _M_ADD_PROD_1i_2j_3k_4l(2, 3, 1, 4);
2311 _M_ADD_PROD_1i_2j_3k_4l(2, 4, 3, 1);
2313 _M_ADD_PROD_1i_2j_3k_4l(3, 1, 2, 4);
2314 _M_ADD_PROD_1i_2j_3k_4l(3, 2, 4, 1);
2315 _M_ADD_PROD_1i_2j_3k_4l(3, 4, 1, 2);
2317 _M_ADD_PROD_1i_2j_3k_4l(4, 1, 3, 2);
2318 _M_ADD_PROD_1i_2j_3k_4l(4, 2, 1, 3);
2319 _M_ADD_PROD_1i_2j_3k_4l(4, 3, 2, 1);
2321 _M_SUB_PROD_1i_2j_3k_4l(1, 2, 4, 3);
2322 _M_SUB_PROD_1i_2j_3k_4l(1, 3, 2, 4);
2323 _M_SUB_PROD_1i_2j_3k_4l(1, 4, 3, 2);
2325 _M_SUB_PROD_1i_2j_3k_4l(2, 1, 3, 4);
2326 _M_SUB_PROD_1i_2j_3k_4l(2, 3, 4, 1);
2327 _M_SUB_PROD_1i_2j_3k_4l(2, 4, 1, 3);
2329 _M_SUB_PROD_1i_2j_3k_4l(3, 1, 4, 2);
2330 _M_SUB_PROD_1i_2j_3k_4l(3, 2, 1, 4);
2331 _M_SUB_PROD_1i_2j_3k_4l(3, 4, 2, 1);
2333 _M_SUB_PROD_1i_2j_3k_4l(4, 1, 2, 3);
2334 _M_SUB_PROD_1i_2j_3k_4l(4, 2, 3, 1);
2335 _M_SUB_PROD_1i_2j_3k_4l(4, 3, 1, 2);
2340 /* XXX: Probably better to just use src/mesa/math/m_matrix.c because
2341 * I have no idea where this code came from.
2344 nine_d3d_matrix_inverse(D3DMATRIX
*D
, const D3DMATRIX
*M
)
2350 M
->m
[1][1] * M
->m
[2][2] * M
->m
[3][3] -
2351 M
->m
[1][1] * M
->m
[3][2] * M
->m
[2][3] -
2352 M
->m
[1][2] * M
->m
[2][1] * M
->m
[3][3] +
2353 M
->m
[1][2] * M
->m
[3][1] * M
->m
[2][3] +
2354 M
->m
[1][3] * M
->m
[2][1] * M
->m
[3][2] -
2355 M
->m
[1][3] * M
->m
[3][1] * M
->m
[2][2];
2358 -M
->m
[0][1] * M
->m
[2][2] * M
->m
[3][3] +
2359 M
->m
[0][1] * M
->m
[3][2] * M
->m
[2][3] +
2360 M
->m
[0][2] * M
->m
[2][1] * M
->m
[3][3] -
2361 M
->m
[0][2] * M
->m
[3][1] * M
->m
[2][3] -
2362 M
->m
[0][3] * M
->m
[2][1] * M
->m
[3][2] +
2363 M
->m
[0][3] * M
->m
[3][1] * M
->m
[2][2];
2366 M
->m
[0][1] * M
->m
[1][2] * M
->m
[3][3] -
2367 M
->m
[0][1] * M
->m
[3][2] * M
->m
[1][3] -
2368 M
->m
[0][2] * M
->m
[1][1] * M
->m
[3][3] +
2369 M
->m
[0][2] * M
->m
[3][1] * M
->m
[1][3] +
2370 M
->m
[0][3] * M
->m
[1][1] * M
->m
[3][2] -
2371 M
->m
[0][3] * M
->m
[3][1] * M
->m
[1][2];
2374 -M
->m
[0][1] * M
->m
[1][2] * M
->m
[2][3] +
2375 M
->m
[0][1] * M
->m
[2][2] * M
->m
[1][3] +
2376 M
->m
[0][2] * M
->m
[1][1] * M
->m
[2][3] -
2377 M
->m
[0][2] * M
->m
[2][1] * M
->m
[1][3] -
2378 M
->m
[0][3] * M
->m
[1][1] * M
->m
[2][2] +
2379 M
->m
[0][3] * M
->m
[2][1] * M
->m
[1][2];
2382 -M
->m
[1][0] * M
->m
[2][2] * M
->m
[3][3] +
2383 M
->m
[1][0] * M
->m
[3][2] * M
->m
[2][3] +
2384 M
->m
[1][2] * M
->m
[2][0] * M
->m
[3][3] -
2385 M
->m
[1][2] * M
->m
[3][0] * M
->m
[2][3] -
2386 M
->m
[1][3] * M
->m
[2][0] * M
->m
[3][2] +
2387 M
->m
[1][3] * M
->m
[3][0] * M
->m
[2][2];
2390 M
->m
[0][0] * M
->m
[2][2] * M
->m
[3][3] -
2391 M
->m
[0][0] * M
->m
[3][2] * M
->m
[2][3] -
2392 M
->m
[0][2] * M
->m
[2][0] * M
->m
[3][3] +
2393 M
->m
[0][2] * M
->m
[3][0] * M
->m
[2][3] +
2394 M
->m
[0][3] * M
->m
[2][0] * M
->m
[3][2] -
2395 M
->m
[0][3] * M
->m
[3][0] * M
->m
[2][2];
2398 -M
->m
[0][0] * M
->m
[1][2] * M
->m
[3][3] +
2399 M
->m
[0][0] * M
->m
[3][2] * M
->m
[1][3] +
2400 M
->m
[0][2] * M
->m
[1][0] * M
->m
[3][3] -
2401 M
->m
[0][2] * M
->m
[3][0] * M
->m
[1][3] -
2402 M
->m
[0][3] * M
->m
[1][0] * M
->m
[3][2] +
2403 M
->m
[0][3] * M
->m
[3][0] * M
->m
[1][2];
2406 M
->m
[0][0] * M
->m
[1][2] * M
->m
[2][3] -
2407 M
->m
[0][0] * M
->m
[2][2] * M
->m
[1][3] -
2408 M
->m
[0][2] * M
->m
[1][0] * M
->m
[2][3] +
2409 M
->m
[0][2] * M
->m
[2][0] * M
->m
[1][3] +
2410 M
->m
[0][3] * M
->m
[1][0] * M
->m
[2][2] -
2411 M
->m
[0][3] * M
->m
[2][0] * M
->m
[1][2];
2414 M
->m
[1][0] * M
->m
[2][1] * M
->m
[3][3] -
2415 M
->m
[1][0] * M
->m
[3][1] * M
->m
[2][3] -
2416 M
->m
[1][1] * M
->m
[2][0] * M
->m
[3][3] +
2417 M
->m
[1][1] * M
->m
[3][0] * M
->m
[2][3] +
2418 M
->m
[1][3] * M
->m
[2][0] * M
->m
[3][1] -
2419 M
->m
[1][3] * M
->m
[3][0] * M
->m
[2][1];
2422 -M
->m
[0][0] * M
->m
[2][1] * M
->m
[3][3] +
2423 M
->m
[0][0] * M
->m
[3][1] * M
->m
[2][3] +
2424 M
->m
[0][1] * M
->m
[2][0] * M
->m
[3][3] -
2425 M
->m
[0][1] * M
->m
[3][0] * M
->m
[2][3] -
2426 M
->m
[0][3] * M
->m
[2][0] * M
->m
[3][1] +
2427 M
->m
[0][3] * M
->m
[3][0] * M
->m
[2][1];
2430 M
->m
[0][0] * M
->m
[1][1] * M
->m
[3][3] -
2431 M
->m
[0][0] * M
->m
[3][1] * M
->m
[1][3] -
2432 M
->m
[0][1] * M
->m
[1][0] * M
->m
[3][3] +
2433 M
->m
[0][1] * M
->m
[3][0] * M
->m
[1][3] +
2434 M
->m
[0][3] * M
->m
[1][0] * M
->m
[3][1] -
2435 M
->m
[0][3] * M
->m
[3][0] * M
->m
[1][1];
2438 -M
->m
[0][0] * M
->m
[1][1] * M
->m
[2][3] +
2439 M
->m
[0][0] * M
->m
[2][1] * M
->m
[1][3] +
2440 M
->m
[0][1] * M
->m
[1][0] * M
->m
[2][3] -
2441 M
->m
[0][1] * M
->m
[2][0] * M
->m
[1][3] -
2442 M
->m
[0][3] * M
->m
[1][0] * M
->m
[2][1] +
2443 M
->m
[0][3] * M
->m
[2][0] * M
->m
[1][1];
2446 -M
->m
[1][0] * M
->m
[2][1] * M
->m
[3][2] +
2447 M
->m
[1][0] * M
->m
[3][1] * M
->m
[2][2] +
2448 M
->m
[1][1] * M
->m
[2][0] * M
->m
[3][2] -
2449 M
->m
[1][1] * M
->m
[3][0] * M
->m
[2][2] -
2450 M
->m
[1][2] * M
->m
[2][0] * M
->m
[3][1] +
2451 M
->m
[1][2] * M
->m
[3][0] * M
->m
[2][1];
2454 M
->m
[0][0] * M
->m
[2][1] * M
->m
[3][2] -
2455 M
->m
[0][0] * M
->m
[3][1] * M
->m
[2][2] -
2456 M
->m
[0][1] * M
->m
[2][0] * M
->m
[3][2] +
2457 M
->m
[0][1] * M
->m
[3][0] * M
->m
[2][2] +
2458 M
->m
[0][2] * M
->m
[2][0] * M
->m
[3][1] -
2459 M
->m
[0][2] * M
->m
[3][0] * M
->m
[2][1];
2462 -M
->m
[0][0] * M
->m
[1][1] * M
->m
[3][2] +
2463 M
->m
[0][0] * M
->m
[3][1] * M
->m
[1][2] +
2464 M
->m
[0][1] * M
->m
[1][0] * M
->m
[3][2] -
2465 M
->m
[0][1] * M
->m
[3][0] * M
->m
[1][2] -
2466 M
->m
[0][2] * M
->m
[1][0] * M
->m
[3][1] +
2467 M
->m
[0][2] * M
->m
[3][0] * M
->m
[1][1];
2470 M
->m
[0][0] * M
->m
[1][1] * M
->m
[2][2] -
2471 M
->m
[0][0] * M
->m
[2][1] * M
->m
[1][2] -
2472 M
->m
[0][1] * M
->m
[1][0] * M
->m
[2][2] +
2473 M
->m
[0][1] * M
->m
[2][0] * M
->m
[1][2] +
2474 M
->m
[0][2] * M
->m
[1][0] * M
->m
[2][1] -
2475 M
->m
[0][2] * M
->m
[2][0] * M
->m
[1][1];
2478 M
->m
[0][0] * D
->m
[0][0] +
2479 M
->m
[1][0] * D
->m
[0][1] +
2480 M
->m
[2][0] * D
->m
[0][2] +
2481 M
->m
[3][0] * D
->m
[0][3];
2483 if (fabsf(det
) < 1e-30) {/* non inversible */
2484 *D
= *M
; /* wine tests */
2490 for (i
= 0; i
< 4; i
++)
2491 for (k
= 0; k
< 4; k
++)
2498 nine_d3d_matrix_matrix_mul(&I
, D
, M
);
2500 for (i
= 0; i
< 4; ++i
)
2501 for (k
= 0; k
< 4; ++k
)
2502 if (fabsf(I
.m
[i
][k
] - (float)(i
== k
)) > 1e-3)
2503 DBG("Matrix inversion check FAILED !\n");