2 * shaders implementation
4 * Copyright 2002-2003 Jason Edmeades
5 * Copyright 2002-2003 Raphael Junqueira
6 * Copyright 2004 Christian Costa
7 * Copyright 2005 Oliver Stieber
8 * Copyright 2006 Ivan Gyurdiev
9 * Copyright 2007-2008 Stefan Dösinger for CodeWeavers
11 * This library is free software; you can redistribute it and/or
12 * modify it under the terms of the GNU Lesser General Public
13 * License as published by the Free Software Foundation; either
14 * version 2.1 of the License, or (at your option) any later version.
16 * This library is distributed in the hope that it will be useful,
17 * but WITHOUT ANY WARRANTY; without even the implied warranty of
18 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
19 * Lesser General Public License for more details.
21 * You should have received a copy of the GNU Lesser General Public
22 * License along with this library; if not, write to the Free Software
23 * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301, USA
29 #include "wined3d_private.h"
31 WINE_DEFAULT_DEBUG_CHANNEL(d3d_shader
);
32 WINE_DECLARE_DEBUG_CHANNEL(d3d
);
34 #define GLNAME_REQUIRE_GLSL ((const char *)1)
36 static void shader_dump_param(const DWORD param
, const DWORD addr_token
, int input
, DWORD shader_version
);
38 static inline BOOL
shader_is_version_token(DWORD token
) {
39 return shader_is_pshader_version(token
) ||
40 shader_is_vshader_version(token
);
43 void shader_buffer_init(struct SHADER_BUFFER
*buffer
)
45 buffer
->buffer
= HeapAlloc(GetProcessHeap(), 0, SHADER_PGMSIZE
);
46 buffer
->buffer
[0] = '\0';
49 buffer
->newline
= TRUE
;
52 void shader_buffer_free(struct SHADER_BUFFER
*buffer
)
54 HeapFree(GetProcessHeap(), 0, buffer
->buffer
);
58 SHADER_BUFFER
* buffer
,
59 const char *format
, ...) {
61 char* base
= buffer
->buffer
+ buffer
->bsize
;
65 va_start(args
, format
);
66 rc
= vsnprintf(base
, SHADER_PGMSIZE
- 1 - buffer
->bsize
, format
, args
);
69 if (rc
< 0 || /* C89 */
70 rc
> SHADER_PGMSIZE
- 1 - buffer
->bsize
) { /* C99 */
72 ERR("The buffer allocated for the shader program string "
73 "is too small at %d bytes.\n", SHADER_PGMSIZE
);
74 buffer
->bsize
= SHADER_PGMSIZE
- 1;
78 if (buffer
->newline
) {
79 TRACE("GL HW (%u, %u) : %s", buffer
->lineNo
+ 1, buffer
->bsize
, base
);
80 buffer
->newline
= FALSE
;
86 if (buffer
->buffer
[buffer
->bsize
-1] == '\n') {
88 buffer
->newline
= TRUE
;
93 void shader_init(struct IWineD3DBaseShaderClass
*shader
,
94 IWineD3DDevice
*device
, const SHADER_OPCODE
*instruction_table
)
97 shader
->device
= device
;
98 shader
->shader_ins
= instruction_table
;
99 list_init(&shader
->linked_programs
);
102 const SHADER_OPCODE
*shader_get_opcode(const SHADER_OPCODE
*opcode_table
, DWORD shader_version
, DWORD code
)
106 /** TODO: use dichotomic search */
107 while (opcode_table
[i
].name
)
109 if ((code
& WINED3DSI_OPCODE_MASK
) == opcode_table
[i
].opcode
110 && shader_version
>= opcode_table
[i
].min_version
111 && (!opcode_table
[i
].max_version
|| shader_version
<= opcode_table
[i
].max_version
))
113 return &opcode_table
[i
];
118 FIXME("Unsupported opcode %#x(%d) masked %#x, shader version %#x\n",
119 code
, code
, code
& WINED3DSI_OPCODE_MASK
, shader_version
);
124 /* Read a parameter opcode from the input stream,
125 * and possibly a relative addressing token.
126 * Return the number of tokens read */
127 static int shader_get_param(const DWORD
*pToken
, DWORD shader_version
, DWORD
*param
, DWORD
*addr_token
)
129 /* PS >= 3.0 have relative addressing (with token)
130 * VS >= 2.0 have relative addressing (with token)
131 * VS >= 1.0 < 2.0 have relative addressing (without token)
132 * The version check below should work in general */
134 char rel_token
= WINED3DSHADER_VERSION_MAJOR(shader_version
) >= 2 &&
135 ((*pToken
& WINED3DSHADER_ADDRESSMODE_MASK
) == WINED3DSHADER_ADDRMODE_RELATIVE
);
138 *addr_token
= rel_token
? *(pToken
+ 1): 0;
139 return rel_token
? 2:1;
142 /* Return the number of parameters to skip for an opcode */
143 static inline int shader_skip_opcode(const SHADER_OPCODE
*curOpcode
, DWORD opcode_token
, DWORD shader_version
)
145 /* Shaders >= 2.0 may contain address tokens, but fortunately they
146 * have a useful length mask - use it here. Shaders 1.0 contain no such tokens */
147 return (WINED3DSHADER_VERSION_MAJOR(shader_version
) >= 2)
148 ? ((opcode_token
& WINED3DSI_INSTLENGTH_MASK
) >> WINED3DSI_INSTLENGTH_SHIFT
) : curOpcode
->num_params
;
151 /* Read the parameters of an unrecognized opcode from the input stream
152 * Return the number of tokens read.
154 * Note: This function assumes source or destination token format.
155 * It will not work with specially-formatted tokens like DEF or DCL,
156 * but hopefully those would be recognized */
157 static int shader_skip_unrecognized(const DWORD
*pToken
, DWORD shader_version
)
162 /* TODO: Think of a good name for 0x80000000 and replace it with a constant */
163 while (*pToken
& 0x80000000) {
165 DWORD param
, addr_token
;
166 tokens_read
+= shader_get_param(pToken
, shader_version
, ¶m
, &addr_token
);
167 pToken
+= tokens_read
;
169 FIXME("Unrecognized opcode param: token=0x%08x "
170 "addr_token=0x%08x name=", param
, addr_token
);
171 shader_dump_param(param
, addr_token
, i
, shader_version
);
178 /* Convert floating point offset relative
179 * to a register file to an absolute offset for float constants */
180 static unsigned int shader_get_float_offset(const DWORD reg
)
182 unsigned int regnum
= reg
& WINED3DSP_REGNUM_MASK
;
183 int regtype
= shader_get_regtype(reg
);
186 case WINED3DSPR_CONST
: return regnum
;
187 case WINED3DSPR_CONST2
: return 2048 + regnum
;
188 case WINED3DSPR_CONST3
: return 4096 + regnum
;
189 case WINED3DSPR_CONST4
: return 6144 + regnum
;
191 FIXME("Unsupported register type: %d\n", regtype
);
196 static void shader_delete_constant_list(struct list
* clist
) {
199 struct local_constant
* constant
;
201 ptr
= list_head(clist
);
203 constant
= LIST_ENTRY(ptr
, struct local_constant
, entry
);
204 ptr
= list_next(clist
, ptr
);
205 HeapFree(GetProcessHeap(), 0, constant
);
210 /* Note that this does not count the loop register
211 * as an address register. */
213 HRESULT
shader_get_registers_used(IWineD3DBaseShader
*iface
, struct shader_reg_maps
*reg_maps
,
214 struct semantic
*semantics_in
, struct semantic
*semantics_out
, const DWORD
*byte_code
)
216 IWineD3DBaseShaderImpl
* This
= (IWineD3DBaseShaderImpl
*) iface
;
217 const SHADER_OPCODE
*shader_ins
= This
->baseShader
.shader_ins
;
218 DWORD shader_version
;
219 unsigned int cur_loop_depth
= 0, max_loop_depth
= 0;
220 const DWORD
* pToken
= byte_code
;
223 /* There are some minor differences between pixel and vertex shaders */
225 memset(reg_maps
->bumpmat
, 0, sizeof(reg_maps
->bumpmat
));
226 memset(reg_maps
->luminanceparams
, 0, sizeof(reg_maps
->luminanceparams
));
228 /* get_registers_used is called on every compile on some 1.x shaders, which can result
229 * in stacking up a collection of local constants. Delete the old constants if existing
231 shader_delete_constant_list(&This
->baseShader
.constantsF
);
232 shader_delete_constant_list(&This
->baseShader
.constantsB
);
233 shader_delete_constant_list(&This
->baseShader
.constantsI
);
235 /* The version token is supposed to be the first token */
236 if (!shader_is_version_token(*pToken
))
238 FIXME("First token is not a version token, invalid shader.\n");
239 return WINED3DERR_INVALIDCALL
;
241 reg_maps
->shader_version
= shader_version
= *pToken
++;
242 pshader
= shader_is_pshader_version(shader_version
);
244 while (WINED3DVS_END() != *pToken
) {
245 CONST SHADER_OPCODE
* curOpcode
;
249 if (shader_is_comment(*pToken
))
251 DWORD comment_len
= (*pToken
& WINED3DSI_COMMENTSIZE_MASK
) >> WINED3DSI_COMMENTSIZE_SHIFT
;
253 pToken
+= comment_len
;
258 opcode_token
= *pToken
++;
259 curOpcode
= shader_get_opcode(shader_ins
, shader_version
, opcode_token
);
261 /* Unhandled opcode, and its parameters */
262 if (NULL
== curOpcode
) {
263 while (*pToken
& 0x80000000)
266 /* Handle declarations */
267 } else if (WINED3DSIO_DCL
== curOpcode
->opcode
) {
269 DWORD usage
= *pToken
++;
270 DWORD param
= *pToken
++;
271 DWORD regtype
= shader_get_regtype(param
);
272 unsigned int regnum
= param
& WINED3DSP_REGNUM_MASK
;
274 /* Vshader: mark attributes used
275 Pshader: mark 3.0 input registers used, save token */
276 if (WINED3DSPR_INPUT
== regtype
) {
279 reg_maps
->attributes
[regnum
] = 1;
281 reg_maps
->packed_input
[regnum
] = 1;
283 semantics_in
[regnum
].usage
= usage
;
284 semantics_in
[regnum
].reg
= param
;
286 /* Vshader: mark 3.0 output registers used, save token */
287 } else if (WINED3DSPR_OUTPUT
== regtype
) {
288 reg_maps
->packed_output
[regnum
] = 1;
289 semantics_out
[regnum
].usage
= usage
;
290 semantics_out
[regnum
].reg
= param
;
291 if (usage
& (WINED3DDECLUSAGE_FOG
<< WINED3DSP_DCL_USAGE_SHIFT
))
294 /* Save sampler usage token */
295 } else if (WINED3DSPR_SAMPLER
== regtype
)
296 reg_maps
->samplers
[regnum
] = usage
;
298 } else if (WINED3DSIO_DEF
== curOpcode
->opcode
) {
300 local_constant
* lconst
= HeapAlloc(GetProcessHeap(), 0, sizeof(local_constant
));
301 if (!lconst
) return E_OUTOFMEMORY
;
302 lconst
->idx
= *pToken
& WINED3DSP_REGNUM_MASK
;
303 memcpy(lconst
->value
, pToken
+ 1, 4 * sizeof(DWORD
));
305 /* In pixel shader 1.X shaders, the constants are clamped between [-1;1] */
306 if (WINED3DSHADER_VERSION_MAJOR(shader_version
) == 1 && pshader
)
308 float *value
= (float *) lconst
->value
;
309 if(value
[0] < -1.0) value
[0] = -1.0;
310 else if(value
[0] > 1.0) value
[0] = 1.0;
311 if(value
[1] < -1.0) value
[1] = -1.0;
312 else if(value
[1] > 1.0) value
[1] = 1.0;
313 if(value
[2] < -1.0) value
[2] = -1.0;
314 else if(value
[2] > 1.0) value
[2] = 1.0;
315 if(value
[3] < -1.0) value
[3] = -1.0;
316 else if(value
[3] > 1.0) value
[3] = 1.0;
319 list_add_head(&This
->baseShader
.constantsF
, &lconst
->entry
);
320 pToken
+= curOpcode
->num_params
;
322 } else if (WINED3DSIO_DEFI
== curOpcode
->opcode
) {
324 local_constant
* lconst
= HeapAlloc(GetProcessHeap(), 0, sizeof(local_constant
));
325 if (!lconst
) return E_OUTOFMEMORY
;
326 lconst
->idx
= *pToken
& WINED3DSP_REGNUM_MASK
;
327 memcpy(lconst
->value
, pToken
+ 1, 4 * sizeof(DWORD
));
328 list_add_head(&This
->baseShader
.constantsI
, &lconst
->entry
);
329 pToken
+= curOpcode
->num_params
;
331 } else if (WINED3DSIO_DEFB
== curOpcode
->opcode
) {
333 local_constant
* lconst
= HeapAlloc(GetProcessHeap(), 0, sizeof(local_constant
));
334 if (!lconst
) return E_OUTOFMEMORY
;
335 lconst
->idx
= *pToken
& WINED3DSP_REGNUM_MASK
;
336 memcpy(lconst
->value
, pToken
+ 1, 1 * sizeof(DWORD
));
337 list_add_head(&This
->baseShader
.constantsB
, &lconst
->entry
);
338 pToken
+= curOpcode
->num_params
;
340 /* If there's a loop in the shader */
341 } else if (WINED3DSIO_LOOP
== curOpcode
->opcode
||
342 WINED3DSIO_REP
== curOpcode
->opcode
) {
344 if(cur_loop_depth
> max_loop_depth
)
345 max_loop_depth
= cur_loop_depth
;
346 pToken
+= curOpcode
->num_params
;
348 /* Rep and Loop always use an integer constant for the control parameters */
349 This
->baseShader
.uses_int_consts
= TRUE
;
350 } else if (WINED3DSIO_ENDLOOP
== curOpcode
->opcode
||
351 WINED3DSIO_ENDREP
== curOpcode
->opcode
) {
354 /* For subroutine prototypes */
355 } else if (WINED3DSIO_LABEL
== curOpcode
->opcode
) {
357 DWORD snum
= *pToken
& WINED3DSP_REGNUM_MASK
;
358 reg_maps
->labels
[snum
] = 1;
359 pToken
+= curOpcode
->num_params
;
361 /* Set texture, address, temporary registers */
365 /* Declare 1.X samplers implicitly, based on the destination reg. number */
366 if (WINED3DSHADER_VERSION_MAJOR(shader_version
) == 1
367 && pshader
/* Filter different instructions with the same enum values in VS */
368 && (WINED3DSIO_TEX
== curOpcode
->opcode
369 || WINED3DSIO_TEXBEM
== curOpcode
->opcode
370 || WINED3DSIO_TEXBEML
== curOpcode
->opcode
371 || WINED3DSIO_TEXDP3TEX
== curOpcode
->opcode
372 || WINED3DSIO_TEXM3x2TEX
== curOpcode
->opcode
373 || WINED3DSIO_TEXM3x3SPEC
== curOpcode
->opcode
374 || WINED3DSIO_TEXM3x3TEX
== curOpcode
->opcode
375 || WINED3DSIO_TEXM3x3VSPEC
== curOpcode
->opcode
376 || WINED3DSIO_TEXREG2AR
== curOpcode
->opcode
377 || WINED3DSIO_TEXREG2GB
== curOpcode
->opcode
378 || WINED3DSIO_TEXREG2RGB
== curOpcode
->opcode
))
380 /* Fake sampler usage, only set reserved bit and ttype */
381 DWORD sampler_code
= *pToken
& WINED3DSP_REGNUM_MASK
;
383 TRACE("Setting fake 2D sampler for 1.x pixelshader\n");
384 reg_maps
->samplers
[sampler_code
] = (0x1 << 31) | WINED3DSTT_2D
;
386 /* texbem is only valid with < 1.4 pixel shaders */
387 if(WINED3DSIO_TEXBEM
== curOpcode
->opcode
||
388 WINED3DSIO_TEXBEML
== curOpcode
->opcode
) {
389 reg_maps
->bumpmat
[sampler_code
] = TRUE
;
390 if(WINED3DSIO_TEXBEML
== curOpcode
->opcode
) {
391 reg_maps
->luminanceparams
[sampler_code
] = TRUE
;
395 if(WINED3DSIO_NRM
== curOpcode
->opcode
) {
396 reg_maps
->usesnrm
= 1;
397 } else if(WINED3DSIO_BEM
== curOpcode
->opcode
&& pshader
) {
398 DWORD regnum
= *pToken
& WINED3DSP_REGNUM_MASK
;
399 reg_maps
->bumpmat
[regnum
] = TRUE
;
400 } else if(WINED3DSIO_DSY
== curOpcode
->opcode
) {
401 reg_maps
->usesdsy
= 1;
404 /* This will loop over all the registers and try to
405 * make a bitmask of the ones we're interested in.
407 * Relative addressing tokens are ignored, but that's
408 * okay, since we'll catch any address registers when
409 * they are initialized (required by spec) */
411 limit
= (opcode_token
& WINED3DSHADER_INSTRUCTION_PREDICATED
)?
412 curOpcode
->num_params
+ 1: curOpcode
->num_params
;
414 for (i
= 0; i
< limit
; ++i
) {
416 DWORD param
, addr_token
, reg
, regtype
;
417 pToken
+= shader_get_param(pToken
, shader_version
, ¶m
, &addr_token
);
419 regtype
= shader_get_regtype(param
);
420 reg
= param
& WINED3DSP_REGNUM_MASK
;
422 if (WINED3DSPR_TEXTURE
== regtype
) { /* vs: WINED3DSPR_ADDR */
425 reg_maps
->texcoord
[reg
] = 1;
427 reg_maps
->address
[reg
] = 1;
430 else if (WINED3DSPR_TEMP
== regtype
)
431 reg_maps
->temporary
[reg
] = 1;
433 else if (WINED3DSPR_INPUT
== regtype
) {
435 reg_maps
->attributes
[reg
] = 1;
437 if(param
& WINED3DSHADER_ADDRMODE_RELATIVE
) {
438 /* If relative addressing is used, we must assume that all registers
439 * are used. Even if it is a construct like v3[aL], we can't assume
440 * that v0, v1 and v2 aren't read because aL can be negative
443 for(i
= 0; i
< MAX_REG_INPUT
; i
++) {
444 ((IWineD3DPixelShaderImpl
*) This
)->input_reg_used
[i
] = TRUE
;
447 ((IWineD3DPixelShaderImpl
*) This
)->input_reg_used
[reg
] = TRUE
;
452 else if (WINED3DSPR_RASTOUT
== regtype
&& reg
== 1)
455 else if (WINED3DSPR_MISCTYPE
== regtype
&& reg
== 0 && pshader
)
458 else if(WINED3DSPR_CONST
== regtype
) {
459 if(param
& WINED3DSHADER_ADDRMODE_RELATIVE
) {
461 if(reg
<= ((IWineD3DVertexShaderImpl
*) This
)->min_rel_offset
) {
462 ((IWineD3DVertexShaderImpl
*) This
)->min_rel_offset
= reg
;
463 } else if(reg
>= ((IWineD3DVertexShaderImpl
*) This
)->max_rel_offset
) {
464 ((IWineD3DVertexShaderImpl
*) This
)->max_rel_offset
= reg
;
467 reg_maps
->usesrelconstF
= TRUE
;
470 else if(WINED3DSPR_CONSTINT
== regtype
) {
471 This
->baseShader
.uses_int_consts
= TRUE
;
473 else if(WINED3DSPR_CONSTBOOL
== regtype
) {
474 This
->baseShader
.uses_bool_consts
= TRUE
;
477 /* WINED3DSPR_TEXCRDOUT is the same as WINED3DSPR_OUTPUT. _OUTPUT can be > MAX_REG_TEXCRD and is used
478 * in >= 3.0 shaders. Filter 3.0 shaders to prevent overflows, and also filter pixel shaders because TECRDOUT
479 * isn't used in them, but future register types might cause issues
481 else if (WINED3DSPR_TEXCRDOUT
== regtype
&& i
== 0 /* Only look at writes */
482 && !pshader
&& WINED3DSHADER_VERSION_MAJOR(shader_version
) < 3)
484 reg_maps
->texcoord_mask
[reg
] |= shader_get_writemask(param
);
490 reg_maps
->loop_depth
= max_loop_depth
;
492 This
->baseShader
.functionLength
= ((char *)pToken
- (char *)byte_code
);
497 static void shader_dump_decl_usage(DWORD decl
, DWORD param
, DWORD shader_version
)
499 DWORD regtype
= shader_get_regtype(param
);
503 if (regtype
== WINED3DSPR_SAMPLER
) {
504 DWORD ttype
= decl
& WINED3DSP_TEXTURETYPE_MASK
;
507 case WINED3DSTT_2D
: TRACE("_2d"); break;
508 case WINED3DSTT_CUBE
: TRACE("_cube"); break;
509 case WINED3DSTT_VOLUME
: TRACE("_volume"); break;
510 default: TRACE("_unknown_ttype(0x%08x)", ttype
);
515 DWORD usage
= decl
& WINED3DSP_DCL_USAGE_MASK
;
516 DWORD idx
= (decl
& WINED3DSP_DCL_USAGEINDEX_MASK
) >> WINED3DSP_DCL_USAGEINDEX_SHIFT
;
518 /* Pixel shaders 3.0 don't have usage semantics */
519 if (shader_is_pshader_version(shader_version
) && shader_version
< WINED3DPS_VERSION(3,0))
525 case WINED3DDECLUSAGE_POSITION
:
526 TRACE("position%d", idx
);
528 case WINED3DDECLUSAGE_BLENDINDICES
:
531 case WINED3DDECLUSAGE_BLENDWEIGHT
:
534 case WINED3DDECLUSAGE_NORMAL
:
535 TRACE("normal%d", idx
);
537 case WINED3DDECLUSAGE_PSIZE
:
540 case WINED3DDECLUSAGE_COLOR
:
544 TRACE("specular%d", (idx
- 1));
547 case WINED3DDECLUSAGE_TEXCOORD
:
548 TRACE("texture%d", idx
);
550 case WINED3DDECLUSAGE_TANGENT
:
553 case WINED3DDECLUSAGE_BINORMAL
:
556 case WINED3DDECLUSAGE_TESSFACTOR
:
559 case WINED3DDECLUSAGE_POSITIONT
:
560 TRACE("positionT%d", idx
);
562 case WINED3DDECLUSAGE_FOG
:
565 case WINED3DDECLUSAGE_DEPTH
:
568 case WINED3DDECLUSAGE_SAMPLE
:
572 FIXME("unknown_semantics(0x%08x)", usage
);
577 static void shader_dump_arr_entry(const DWORD param
, const DWORD addr_token
,
578 unsigned int reg
, int input
, DWORD shader_version
)
581 ((param
& WINED3DSHADER_ADDRESSMODE_MASK
) == WINED3DSHADER_ADDRMODE_RELATIVE
);
586 shader_dump_param(addr_token
, 0, input
, shader_version
);
596 static void shader_dump_param(const DWORD param
, const DWORD addr_token
, int input
, DWORD shader_version
)
598 static const char * const rastout_reg_names
[] = { "oPos", "oFog", "oPts" };
599 static const char * const misctype_reg_names
[] = { "vPos", "vFace"};
600 const char *swizzle_reg_chars
= "xyzw";
602 DWORD reg
= param
& WINED3DSP_REGNUM_MASK
;
603 DWORD regtype
= shader_get_regtype(param
);
604 DWORD modifier
= param
& WINED3DSP_SRCMOD_MASK
;
606 /* There are some minor differences between pixel and vertex shaders */
607 char pshader
= shader_is_pshader_version(shader_version
);
610 if ( (modifier
== WINED3DSPSM_NEG
) ||
611 (modifier
== WINED3DSPSM_BIASNEG
) ||
612 (modifier
== WINED3DSPSM_SIGNNEG
) ||
613 (modifier
== WINED3DSPSM_X2NEG
) ||
614 (modifier
== WINED3DSPSM_ABSNEG
) )
616 else if (modifier
== WINED3DSPSM_COMP
)
618 else if (modifier
== WINED3DSPSM_NOT
)
621 if (modifier
== WINED3DSPSM_ABS
|| modifier
== WINED3DSPSM_ABSNEG
)
626 case WINED3DSPR_TEMP
:
629 case WINED3DSPR_INPUT
:
631 shader_dump_arr_entry(param
, addr_token
, reg
, input
, shader_version
);
633 case WINED3DSPR_CONST
:
634 case WINED3DSPR_CONST2
:
635 case WINED3DSPR_CONST3
:
636 case WINED3DSPR_CONST4
:
638 shader_dump_arr_entry(param
, addr_token
, shader_get_float_offset(param
), input
, shader_version
);
640 case WINED3DSPR_TEXTURE
: /* vs: case D3DSPR_ADDR */
641 TRACE("%c%u", (pshader
? 't':'a'), reg
);
643 case WINED3DSPR_RASTOUT
:
644 TRACE("%s", rastout_reg_names
[reg
]);
646 case WINED3DSPR_COLOROUT
:
649 case WINED3DSPR_DEPTHOUT
:
652 case WINED3DSPR_ATTROUT
:
655 case WINED3DSPR_TEXCRDOUT
:
657 /* Vertex shaders >= 3.0 use general purpose output registers
658 * (WINED3DSPR_OUTPUT), which can include an address token */
660 if (WINED3DSHADER_VERSION_MAJOR(shader_version
) >= 3) {
662 shader_dump_arr_entry(param
, addr_token
, reg
, input
, shader_version
);
667 case WINED3DSPR_CONSTINT
:
669 shader_dump_arr_entry(param
, addr_token
, reg
, input
, shader_version
);
671 case WINED3DSPR_CONSTBOOL
:
673 shader_dump_arr_entry(param
, addr_token
, reg
, input
, shader_version
);
675 case WINED3DSPR_LABEL
:
678 case WINED3DSPR_LOOP
:
681 case WINED3DSPR_SAMPLER
:
684 case WINED3DSPR_MISCTYPE
:
686 FIXME("Unhandled misctype register %d\n", reg
);
688 TRACE("%s", misctype_reg_names
[reg
]);
691 case WINED3DSPR_PREDICATE
:
695 TRACE("unhandled_rtype(%#x)", regtype
);
700 /* operand output (for modifiers and shift, see dump_ins_modifiers) */
702 if ((param
& WINED3DSP_WRITEMASK_ALL
) != WINED3DSP_WRITEMASK_ALL
) {
704 if (param
& WINED3DSP_WRITEMASK_0
) TRACE("%c", swizzle_reg_chars
[0]);
705 if (param
& WINED3DSP_WRITEMASK_1
) TRACE("%c", swizzle_reg_chars
[1]);
706 if (param
& WINED3DSP_WRITEMASK_2
) TRACE("%c", swizzle_reg_chars
[2]);
707 if (param
& WINED3DSP_WRITEMASK_3
) TRACE("%c", swizzle_reg_chars
[3]);
712 DWORD swizzle
= (param
& WINED3DSP_SWIZZLE_MASK
) >> WINED3DSP_SWIZZLE_SHIFT
;
713 DWORD swizzle_x
= swizzle
& 0x03;
714 DWORD swizzle_y
= (swizzle
>> 2) & 0x03;
715 DWORD swizzle_z
= (swizzle
>> 4) & 0x03;
716 DWORD swizzle_w
= (swizzle
>> 6) & 0x03;
720 case WINED3DSPSM_NONE
: break;
721 case WINED3DSPSM_NEG
: break;
722 case WINED3DSPSM_NOT
: break;
723 case WINED3DSPSM_BIAS
: TRACE("_bias"); break;
724 case WINED3DSPSM_BIASNEG
: TRACE("_bias"); break;
725 case WINED3DSPSM_SIGN
: TRACE("_bx2"); break;
726 case WINED3DSPSM_SIGNNEG
: TRACE("_bx2"); break;
727 case WINED3DSPSM_COMP
: break;
728 case WINED3DSPSM_X2
: TRACE("_x2"); break;
729 case WINED3DSPSM_X2NEG
: TRACE("_x2"); break;
730 case WINED3DSPSM_DZ
: TRACE("_dz"); break;
731 case WINED3DSPSM_DW
: TRACE("_dw"); break;
732 case WINED3DSPSM_ABSNEG
: TRACE(")"); break;
733 case WINED3DSPSM_ABS
: TRACE(")"); break;
735 TRACE("_unknown_modifier(%#x)", modifier
>> WINED3DSP_SRCMOD_SHIFT
);
740 * swizzle bits fields:
743 if ((WINED3DVS_NOSWIZZLE
>> WINED3DVS_SWIZZLE_SHIFT
) != swizzle
) {
744 if (swizzle_x
== swizzle_y
&&
745 swizzle_x
== swizzle_z
&&
746 swizzle_x
== swizzle_w
) {
747 TRACE(".%c", swizzle_reg_chars
[swizzle_x
]);
750 swizzle_reg_chars
[swizzle_x
],
751 swizzle_reg_chars
[swizzle_y
],
752 swizzle_reg_chars
[swizzle_z
],
753 swizzle_reg_chars
[swizzle_w
]);
759 static void shader_color_correction(IWineD3DBaseShaderImpl
*shader
,
760 IWineD3DDeviceImpl
*device
, const struct SHADER_OPCODE_ARG
*arg
, DWORD shader_version
)
762 IWineD3DBaseTextureImpl
*texture
;
763 struct color_fixup_desc fixup
;
764 BOOL recorded
= FALSE
;
768 switch(arg
->opcode
->opcode
)
771 if (WINED3DSHADER_VERSION_MAJOR(shader_version
) < 2) sampler_idx
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
772 else sampler_idx
= arg
->src
[1] & WINED3DSP_REGNUM_MASK
;
775 case WINED3DSIO_TEXLDL
:
776 FIXME("Add color fixup for vertex texture WINED3DSIO_TEXLDL\n");
779 case WINED3DSIO_TEXDP3TEX
:
780 case WINED3DSIO_TEXM3x3TEX
:
781 case WINED3DSIO_TEXM3x3SPEC
:
782 case WINED3DSIO_TEXM3x3VSPEC
:
783 case WINED3DSIO_TEXBEM
:
784 case WINED3DSIO_TEXREG2AR
:
785 case WINED3DSIO_TEXREG2GB
:
786 case WINED3DSIO_TEXREG2RGB
:
787 sampler_idx
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
791 /* Not a texture sampling instruction, nothing to do */
795 texture
= (IWineD3DBaseTextureImpl
*)device
->stateBlock
->textures
[sampler_idx
];
796 if (texture
) fixup
= texture
->baseTexture
.shader_color_fixup
;
797 else fixup
= COLOR_FIXUP_IDENTITY
;
799 /* before doing anything, record the sampler with the format in the format conversion list,
800 * but check if it's not there already */
801 for (i
= 0; i
< shader
->baseShader
.num_sampled_samplers
; ++i
)
803 if (shader
->baseShader
.sampled_samplers
[i
] == sampler_idx
)
812 shader
->baseShader
.sampled_samplers
[shader
->baseShader
.num_sampled_samplers
] = sampler_idx
;
813 ++shader
->baseShader
.num_sampled_samplers
;
816 device
->shader_backend
->shader_color_correction(arg
, fixup
);
819 /* Shared code in order to generate the bulk of the shader string.
820 * NOTE: A description of how to parse tokens can be found on msdn */
821 void shader_generate_main(IWineD3DBaseShader
*iface
, SHADER_BUFFER
* buffer
,
822 const shader_reg_maps
* reg_maps
, CONST DWORD
* pFunction
)
824 IWineD3DBaseShaderImpl
* This
= (IWineD3DBaseShaderImpl
*) iface
;
825 IWineD3DDeviceImpl
*device
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
; /* To access shader backend callbacks */
826 const SHADER_OPCODE
*opcode_table
= This
->baseShader
.shader_ins
;
827 const SHADER_HANDLER
*handler_table
= device
->shader_backend
->shader_instruction_handler_table
;
828 DWORD shader_version
= reg_maps
->shader_version
;
829 const DWORD
*pToken
= pFunction
;
830 const SHADER_OPCODE
*curOpcode
;
831 SHADER_HANDLER hw_fct
;
833 SHADER_OPCODE_ARG hw_arg
;
835 /* Initialize current parsing state */
836 hw_arg
.shader
= iface
;
837 hw_arg
.buffer
= buffer
;
838 hw_arg
.reg_maps
= reg_maps
;
839 This
->baseShader
.parse_state
.current_row
= 0;
841 while (WINED3DPS_END() != *pToken
)
843 /* Skip version token */
844 if (shader_is_version_token(*pToken
))
850 /* Skip comment tokens */
851 if (shader_is_comment(*pToken
))
853 pToken
+= (*pToken
& WINED3DSI_COMMENTSIZE_MASK
) >> WINED3DSI_COMMENTSIZE_SHIFT
;
859 hw_arg
.opcode_token
= *pToken
++;
860 curOpcode
= shader_get_opcode(opcode_table
, shader_version
, hw_arg
.opcode_token
);
862 /* Unknown opcode and its parameters */
865 FIXME("Unrecognized opcode: token=0x%08x\n", hw_arg
.opcode_token
);
866 pToken
+= shader_skip_unrecognized(pToken
, shader_version
);
871 if (WINED3DSIO_DCL
== curOpcode
->opcode
872 || WINED3DSIO_NOP
== curOpcode
->opcode
873 || WINED3DSIO_DEF
== curOpcode
->opcode
874 || WINED3DSIO_DEFI
== curOpcode
->opcode
875 || WINED3DSIO_DEFB
== curOpcode
->opcode
876 || WINED3DSIO_PHASE
== curOpcode
->opcode
877 || WINED3DSIO_RET
== curOpcode
->opcode
)
879 pToken
+= shader_skip_opcode(curOpcode
, hw_arg
.opcode_token
, shader_version
);
884 hw_fct
= handler_table
[curOpcode
->handler_idx
];
886 /* Unhandled opcode */
889 FIXME("Can't handle opcode %s in hwShader\n", curOpcode
->name
);
890 pToken
+= shader_skip_opcode(curOpcode
, hw_arg
.opcode_token
, shader_version
);
894 hw_arg
.opcode
= curOpcode
;
896 /* Destination token */
897 if (curOpcode
->dst_token
)
899 DWORD param
, addr_token
= 0;
900 pToken
+= shader_get_param(pToken
, shader_version
, ¶m
, &addr_token
);
902 hw_arg
.dst_addr
= addr_token
;
905 /* Predication token */
906 if (hw_arg
.opcode_token
& WINED3DSHADER_INSTRUCTION_PREDICATED
) hw_arg
.predicate
= *pToken
++;
908 /* Other source tokens */
909 for (i
= 0; i
< (curOpcode
->num_params
- curOpcode
->dst_token
); ++i
)
911 DWORD param
, addr_token
= 0;
912 pToken
+= shader_get_param(pToken
, shader_version
, ¶m
, &addr_token
);
913 hw_arg
.src
[i
] = param
;
914 hw_arg
.src_addr
[i
] = addr_token
;
917 /* Call appropriate function for output target */
920 /* Add color correction if needed */
921 shader_color_correction(This
, device
, &hw_arg
, shader_version
);
923 /* Process instruction modifiers for GLSL apps ( _sat, etc. ) */
924 /* FIXME: This should be internal to the shader backend.
925 * Also, right now this is the only reason "shader_mode" exists. */
926 if (This
->baseShader
.shader_mode
== SHADER_GLSL
) shader_glsl_add_instruction_modifiers(&hw_arg
);
930 static void shader_dump_ins_modifiers(const DWORD output
)
932 DWORD shift
= (output
& WINED3DSP_DSTSHIFT_MASK
) >> WINED3DSP_DSTSHIFT_SHIFT
;
933 DWORD mmask
= output
& WINED3DSP_DSTMOD_MASK
;
937 case 13: TRACE("_d8"); break;
938 case 14: TRACE("_d4"); break;
939 case 15: TRACE("_d2"); break;
940 case 1: TRACE("_x2"); break;
941 case 2: TRACE("_x4"); break;
942 case 3: TRACE("_x8"); break;
943 default: TRACE("_unhandled_shift(%d)", shift
); break;
946 if (mmask
& WINED3DSPDM_SATURATE
) TRACE("_sat");
947 if (mmask
& WINED3DSPDM_PARTIALPRECISION
) TRACE("_pp");
948 if (mmask
& WINED3DSPDM_MSAMPCENTROID
) TRACE("_centroid");
950 mmask
&= ~(WINED3DSPDM_SATURATE
| WINED3DSPDM_PARTIALPRECISION
| WINED3DSPDM_MSAMPCENTROID
);
952 FIXME("_unrecognized_modifier(%#x)", mmask
>> WINED3DSP_DSTMOD_SHIFT
);
955 void shader_trace_init(const DWORD
*pFunction
, const SHADER_OPCODE
*opcode_table
)
957 const DWORD
* pToken
= pFunction
;
958 const SHADER_OPCODE
* curOpcode
= NULL
;
959 DWORD shader_version
;
963 TRACE("Parsing %p\n", pFunction
);
965 /* The version token is supposed to be the first token */
966 if (!shader_is_version_token(*pToken
))
968 FIXME("First token is not a version token, invalid shader.\n");
971 shader_version
= *pToken
++;
972 TRACE("%s_%u_%u\n", shader_is_pshader_version(shader_version
) ? "ps": "vs",
973 WINED3DSHADER_VERSION_MAJOR(shader_version
), WINED3DSHADER_VERSION_MINOR(shader_version
));
975 while (WINED3DVS_END() != *pToken
)
977 if (shader_is_comment(*pToken
)) /* comment */
979 DWORD comment_len
= (*pToken
& WINED3DSI_COMMENTSIZE_MASK
) >> WINED3DSI_COMMENTSIZE_SHIFT
;
981 TRACE("//%s\n", (const char*)pToken
);
982 pToken
+= comment_len
;
985 opcode_token
= *pToken
++;
986 curOpcode
= shader_get_opcode(opcode_table
, shader_version
, opcode_token
);
991 FIXME("Unrecognized opcode: token=0x%08x\n", opcode_token
);
992 tokens_read
= shader_skip_unrecognized(pToken
, shader_version
);
993 pToken
+= tokens_read
;
997 if (curOpcode
->opcode
== WINED3DSIO_DCL
)
999 DWORD usage
= *pToken
;
1000 DWORD param
= *(pToken
+ 1);
1002 shader_dump_decl_usage(usage
, param
, shader_version
);
1003 shader_dump_ins_modifiers(param
);
1005 shader_dump_param(param
, 0, 0, shader_version
);
1008 else if (curOpcode
->opcode
== WINED3DSIO_DEF
)
1010 unsigned int offset
= shader_get_float_offset(*pToken
);
1012 TRACE("def c%u = %f, %f, %f, %f", offset
,
1013 *(const float *)(pToken
+ 1),
1014 *(const float *)(pToken
+ 2),
1015 *(const float *)(pToken
+ 3),
1016 *(const float *)(pToken
+ 4));
1019 else if (curOpcode
->opcode
== WINED3DSIO_DEFI
)
1021 TRACE("defi i%u = %d, %d, %d, %d", *pToken
& WINED3DSP_REGNUM_MASK
,
1028 else if (curOpcode
->opcode
== WINED3DSIO_DEFB
)
1030 TRACE("defb b%u = %s", *pToken
& WINED3DSP_REGNUM_MASK
,
1031 *(pToken
+ 1)? "true": "false");
1036 DWORD param
, addr_token
;
1039 /* Print out predication source token first - it follows
1040 * the destination token. */
1041 if (opcode_token
& WINED3DSHADER_INSTRUCTION_PREDICATED
)
1044 shader_dump_param(*(pToken
+ 2), 0, 1, shader_version
);
1047 if (opcode_token
& WINED3DSI_COISSUE
)
1049 /* PixWin marks instructions with the coissue flag with a '+' */
1053 TRACE("%s", curOpcode
->name
);
1055 if (curOpcode
->opcode
== WINED3DSIO_IFC
1056 || curOpcode
->opcode
== WINED3DSIO_BREAKC
)
1058 DWORD op
= (opcode_token
& INST_CONTROLS_MASK
) >> INST_CONTROLS_SHIFT
;
1062 case COMPARISON_GT
: TRACE("_gt"); break;
1063 case COMPARISON_EQ
: TRACE("_eq"); break;
1064 case COMPARISON_GE
: TRACE("_ge"); break;
1065 case COMPARISON_LT
: TRACE("_lt"); break;
1066 case COMPARISON_NE
: TRACE("_ne"); break;
1067 case COMPARISON_LE
: TRACE("_le"); break;
1068 default: TRACE("_(%u)", op
);
1071 else if (curOpcode
->opcode
== WINED3DSIO_TEX
1072 && shader_version
>= WINED3DPS_VERSION(2,0)
1073 && (opcode_token
& WINED3DSI_TEXLD_PROJECT
))
1078 /* Destination token */
1079 if (curOpcode
->dst_token
)
1081 tokens_read
= shader_get_param(pToken
, shader_version
, ¶m
, &addr_token
);
1082 pToken
+= tokens_read
;
1084 shader_dump_ins_modifiers(param
);
1086 shader_dump_param(param
, addr_token
, 0, shader_version
);
1089 /* Predication token - already printed out, just skip it */
1090 if (opcode_token
& WINED3DSHADER_INSTRUCTION_PREDICATED
)
1095 /* Other source tokens */
1096 for (i
= curOpcode
->dst_token
; i
< curOpcode
->num_params
; ++i
)
1098 tokens_read
= shader_get_param(pToken
, shader_version
, ¶m
, &addr_token
);
1099 pToken
+= tokens_read
;
1101 TRACE((i
== 0)? " " : ", ");
1102 shader_dump_param(param
, addr_token
, 1, shader_version
);
1110 void shader_cleanup(IWineD3DBaseShader
*iface
)
1112 IWineD3DBaseShaderImpl
*This
= (IWineD3DBaseShaderImpl
*)iface
;
1114 ((IWineD3DDeviceImpl
*)This
->baseShader
.device
)->shader_backend
->shader_destroy(iface
);
1115 HeapFree(GetProcessHeap(), 0, This
->baseShader
.function
);
1116 shader_delete_constant_list(&This
->baseShader
.constantsF
);
1117 shader_delete_constant_list(&This
->baseShader
.constantsB
);
1118 shader_delete_constant_list(&This
->baseShader
.constantsI
);
1119 list_remove(&This
->baseShader
.shader_list_entry
);
1122 static const SHADER_HANDLER shader_none_instruction_handler_table
[WINED3DSIH_TABLE_SIZE
] = {0};
1123 static void shader_none_select(IWineD3DDevice
*iface
, BOOL usePS
, BOOL useVS
) {}
1124 static void shader_none_select_depth_blt(IWineD3DDevice
*iface
, enum tex_types tex_type
) {}
1125 static void shader_none_deselect_depth_blt(IWineD3DDevice
*iface
) {}
1126 static void shader_none_update_float_vertex_constants(IWineD3DDevice
*iface
, UINT start
, UINT count
) {}
1127 static void shader_none_update_float_pixel_constants(IWineD3DDevice
*iface
, UINT start
, UINT count
) {}
1128 static void shader_none_load_constants(IWineD3DDevice
*iface
, char usePS
, char useVS
) {}
1129 static void shader_none_color_correction(const struct SHADER_OPCODE_ARG
*arg
, struct color_fixup_desc fixup
) {}
1130 static void shader_none_destroy(IWineD3DBaseShader
*iface
) {}
1131 static HRESULT
shader_none_alloc(IWineD3DDevice
*iface
) {return WINED3D_OK
;}
1132 static void shader_none_free(IWineD3DDevice
*iface
) {}
1133 static BOOL
shader_none_dirty_const(IWineD3DDevice
*iface
) {return FALSE
;}
1134 static GLuint
shader_none_generate_pshader(IWineD3DPixelShader
*iface
, SHADER_BUFFER
*buffer
, const struct ps_compile_args
*args
) {
1135 FIXME("NONE shader backend asked to generate a pixel shader\n");
1138 static void shader_none_generate_vshader(IWineD3DVertexShader
*iface
, SHADER_BUFFER
*buffer
) {
1139 FIXME("NONE shader backend asked to generate a vertex shader\n");
1142 #define GLINFO_LOCATION (*gl_info)
1143 static void shader_none_get_caps(WINED3DDEVTYPE devtype
, const WineD3D_GL_Info
*gl_info
, struct shader_caps
*pCaps
)
1145 /* Set the shader caps to 0 for the none shader backend */
1146 pCaps
->VertexShaderVersion
= 0;
1147 pCaps
->PixelShaderVersion
= 0;
1148 pCaps
->PixelShader1xMaxValue
= 0.0;
1150 #undef GLINFO_LOCATION
1151 static BOOL
shader_none_color_fixup_supported(struct color_fixup_desc fixup
)
1153 if (TRACE_ON(d3d_shader
) && TRACE_ON(d3d
))
1155 TRACE("Checking support for fixup:\n");
1156 dump_color_fixup_desc(fixup
);
1159 /* Faked to make some apps happy. */
1160 if (!is_yuv_fixup(fixup
))
1166 TRACE("[FAILED]\n");
1170 const shader_backend_t none_shader_backend
= {
1171 shader_none_instruction_handler_table
,
1173 shader_none_select_depth_blt
,
1174 shader_none_deselect_depth_blt
,
1175 shader_none_update_float_vertex_constants
,
1176 shader_none_update_float_pixel_constants
,
1177 shader_none_load_constants
,
1178 shader_none_color_correction
,
1179 shader_none_destroy
,
1182 shader_none_dirty_const
,
1183 shader_none_generate_pshader
,
1184 shader_none_generate_vshader
,
1185 shader_none_get_caps
,
1186 shader_none_color_fixup_supported
,