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
10 * Copyright 2009 Henri Verbeet for CodeWeavers
12 * This library is free software; you can redistribute it and/or
13 * modify it under the terms of the GNU Lesser General Public
14 * License as published by the Free Software Foundation; either
15 * version 2.1 of the License, or (at your option) any later version.
17 * This library is distributed in the hope that it will be useful,
18 * but WITHOUT ANY WARRANTY; without even the implied warranty of
19 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
20 * Lesser General Public License for more details.
22 * You should have received a copy of the GNU Lesser General Public
23 * License along with this library; if not, write to the Free Software
24 * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301, USA
30 #include "wined3d_private.h"
32 WINE_DEFAULT_DEBUG_CHANNEL(d3d_shader
);
33 WINE_DECLARE_DEBUG_CHANNEL(d3d
);
35 static const char *shader_opcode_names
[] =
37 /* WINED3DSIH_ABS */ "abs",
38 /* WINED3DSIH_ADD */ "add",
39 /* WINED3DSIH_BEM */ "bem",
40 /* WINED3DSIH_BREAK */ "break",
41 /* WINED3DSIH_BREAKC */ "breakc",
42 /* WINED3DSIH_BREAKP */ "breakp",
43 /* WINED3DSIH_CALL */ "call",
44 /* WINED3DSIH_CALLNZ */ "callnz",
45 /* WINED3DSIH_CMP */ "cmp",
46 /* WINED3DSIH_CND */ "cnd",
47 /* WINED3DSIH_CRS */ "crs",
48 /* WINED3DSIH_DCL */ "dcl",
49 /* WINED3DSIH_DEF */ "def",
50 /* WINED3DSIH_DEFB */ "defb",
51 /* WINED3DSIH_DEFI */ "defi",
52 /* WINED3DSIH_DP2ADD */ "dp2add",
53 /* WINED3DSIH_DP3 */ "dp3",
54 /* WINED3DSIH_DP4 */ "dp4",
55 /* WINED3DSIH_DST */ "dst",
56 /* WINED3DSIH_DSX */ "dsx",
57 /* WINED3DSIH_DSY */ "dsy",
58 /* WINED3DSIH_ELSE */ "else",
59 /* WINED3DSIH_ENDIF */ "endif",
60 /* WINED3DSIH_ENDLOOP */ "endloop",
61 /* WINED3DSIH_ENDREP */ "endrep",
62 /* WINED3DSIH_EXP */ "exp",
63 /* WINED3DSIH_EXPP */ "expp",
64 /* WINED3DSIH_FRC */ "frc",
65 /* WINED3DSIH_IF */ "if",
66 /* WINED3DSIH_IFC */ "ifc",
67 /* WINED3DSIH_LABEL */ "label",
68 /* WINED3DSIH_LIT */ "lit",
69 /* WINED3DSIH_LOG */ "log",
70 /* WINED3DSIH_LOGP */ "logp",
71 /* WINED3DSIH_LOOP */ "loop",
72 /* WINED3DSIH_LRP */ "lrp",
73 /* WINED3DSIH_M3x2 */ "m3x2",
74 /* WINED3DSIH_M3x3 */ "m3x3",
75 /* WINED3DSIH_M3x4 */ "m3x4",
76 /* WINED3DSIH_M4x3 */ "m4x3",
77 /* WINED3DSIH_M4x4 */ "m4x4",
78 /* WINED3DSIH_MAD */ "mad",
79 /* WINED3DSIH_MAX */ "max",
80 /* WINED3DSIH_MIN */ "min",
81 /* WINED3DSIH_MOV */ "mov",
82 /* WINED3DSIH_MOVA */ "mova",
83 /* WINED3DSIH_MUL */ "mul",
84 /* WINED3DSIH_NOP */ "nop",
85 /* WINED3DSIH_NRM */ "nrm",
86 /* WINED3DSIH_PHASE */ "phase",
87 /* WINED3DSIH_POW */ "pow",
88 /* WINED3DSIH_RCP */ "rcp",
89 /* WINED3DSIH_REP */ "rep",
90 /* WINED3DSIH_RET */ "ret",
91 /* WINED3DSIH_RSQ */ "rsq",
92 /* WINED3DSIH_SETP */ "setp",
93 /* WINED3DSIH_SGE */ "sge",
94 /* WINED3DSIH_SGN */ "sgn",
95 /* WINED3DSIH_SINCOS */ "sincos",
96 /* WINED3DSIH_SLT */ "slt",
97 /* WINED3DSIH_SUB */ "sub",
98 /* WINED3DSIH_TEX */ "texld",
99 /* WINED3DSIH_TEXBEM */ "texbem",
100 /* WINED3DSIH_TEXBEML */ "texbeml",
101 /* WINED3DSIH_TEXCOORD */ "texcrd",
102 /* WINED3DSIH_TEXDEPTH */ "texdepth",
103 /* WINED3DSIH_TEXDP3 */ "texdp3",
104 /* WINED3DSIH_TEXDP3TEX */ "texdp3tex",
105 /* WINED3DSIH_TEXKILL */ "texkill",
106 /* WINED3DSIH_TEXLDD */ "texldd",
107 /* WINED3DSIH_TEXLDL */ "texldl",
108 /* WINED3DSIH_TEXM3x2DEPTH */ "texm3x2depth",
109 /* WINED3DSIH_TEXM3x2PAD */ "texm3x2pad",
110 /* WINED3DSIH_TEXM3x2TEX */ "texm3x2tex",
111 /* WINED3DSIH_TEXM3x3 */ "texm3x3",
112 /* WINED3DSIH_TEXM3x3DIFF */ "texm3x3diff",
113 /* WINED3DSIH_TEXM3x3PAD */ "texm3x3pad",
114 /* WINED3DSIH_TEXM3x3SPEC */ "texm3x3spec",
115 /* WINED3DSIH_TEXM3x3TEX */ "texm3x3tex",
116 /* WINED3DSIH_TEXM3x3VSPEC */ "texm3x3vspec",
117 /* WINED3DSIH_TEXREG2AR */ "texreg2ar",
118 /* WINED3DSIH_TEXREG2GB */ "texreg2gb",
119 /* WINED3DSIH_TEXREG2RGB */ "texreg2rgb",
122 const struct wined3d_shader_frontend
*shader_select_frontend(DWORD version_token
)
124 switch (version_token
>> 16)
128 return &sm1_shader_frontend
;
133 return &sm4_shader_frontend
;
136 FIXME("Unrecognised version token %#x\n", version_token
);
141 static inline BOOL
shader_is_version_token(DWORD token
) {
142 return shader_is_pshader_version(token
) ||
143 shader_is_vshader_version(token
);
146 void shader_buffer_init(struct SHADER_BUFFER
*buffer
)
148 buffer
->buffer
= HeapAlloc(GetProcessHeap(), 0, SHADER_PGMSIZE
);
149 buffer
->buffer
[0] = '\0';
152 buffer
->newline
= TRUE
;
155 void shader_buffer_free(struct SHADER_BUFFER
*buffer
)
157 HeapFree(GetProcessHeap(), 0, buffer
->buffer
);
160 int shader_vaddline(SHADER_BUFFER
* buffer
, const char *format
, va_list args
)
162 char* base
= buffer
->buffer
+ buffer
->bsize
;
165 rc
= vsnprintf(base
, SHADER_PGMSIZE
- 1 - buffer
->bsize
, format
, args
);
167 if (rc
< 0 || /* C89 */
168 rc
> SHADER_PGMSIZE
- 1 - buffer
->bsize
) { /* C99 */
170 ERR("The buffer allocated for the shader program string "
171 "is too small at %d bytes.\n", SHADER_PGMSIZE
);
172 buffer
->bsize
= SHADER_PGMSIZE
- 1;
176 if (buffer
->newline
) {
177 TRACE("GL HW (%u, %u) : %s", buffer
->lineNo
+ 1, buffer
->bsize
, base
);
178 buffer
->newline
= FALSE
;
184 if (buffer
->buffer
[buffer
->bsize
-1] == '\n') {
186 buffer
->newline
= TRUE
;
191 int shader_addline(SHADER_BUFFER
* buffer
, const char *format
, ...)
196 va_start(args
, format
);
197 ret
= shader_vaddline(buffer
, format
, args
);
203 void shader_init(struct IWineD3DBaseShaderClass
*shader
, IWineD3DDevice
*device
)
206 shader
->device
= device
;
207 list_init(&shader
->linked_programs
);
210 /* Convert floating point offset relative
211 * to a register file to an absolute offset for float constants */
212 static unsigned int shader_get_float_offset(WINED3DSHADER_PARAM_REGISTER_TYPE register_type
, UINT register_idx
)
214 switch (register_type
)
216 case WINED3DSPR_CONST
: return register_idx
;
217 case WINED3DSPR_CONST2
: return 2048 + register_idx
;
218 case WINED3DSPR_CONST3
: return 4096 + register_idx
;
219 case WINED3DSPR_CONST4
: return 6144 + register_idx
;
221 FIXME("Unsupported register type: %d\n", register_type
);
226 static void shader_delete_constant_list(struct list
* clist
) {
229 struct local_constant
* constant
;
231 ptr
= list_head(clist
);
233 constant
= LIST_ENTRY(ptr
, struct local_constant
, entry
);
234 ptr
= list_next(clist
, ptr
);
235 HeapFree(GetProcessHeap(), 0, constant
);
240 static inline void set_bitmap_bit(DWORD
*bitmap
, DWORD bit
)
245 bitmap
[idx
] |= (1 << shift
);
248 static void shader_record_register_usage(IWineD3DBaseShaderImpl
*This
, struct shader_reg_maps
*reg_maps
,
249 const struct wined3d_shader_register
*reg
, BOOL pshader
)
253 case WINED3DSPR_TEXTURE
: /* WINED3DSPR_ADDR */
254 if (pshader
) reg_maps
->texcoord
[reg
->idx
] = 1;
255 else reg_maps
->address
[reg
->idx
] = 1;
258 case WINED3DSPR_TEMP
:
259 reg_maps
->temporary
[reg
->idx
] = 1;
262 case WINED3DSPR_INPUT
:
263 if (!pshader
) reg_maps
->attributes
[reg
->idx
] = 1;
268 /* If relative addressing is used, we must assume that all registers
269 * are used. Even if it is a construct like v3[aL], we can't assume
270 * that v0, v1 and v2 aren't read because aL can be negative */
272 for (i
= 0; i
< MAX_REG_INPUT
; ++i
)
274 ((IWineD3DPixelShaderImpl
*)This
)->input_reg_used
[i
] = TRUE
;
279 ((IWineD3DPixelShaderImpl
*)This
)->input_reg_used
[reg
->idx
] = TRUE
;
284 case WINED3DSPR_RASTOUT
:
285 if (reg
->idx
== 1) reg_maps
->fog
= 1;
288 case WINED3DSPR_MISCTYPE
:
289 if (pshader
&& reg
->idx
== 0) reg_maps
->vpos
= 1;
292 case WINED3DSPR_CONST
:
297 if (reg
->idx
<= ((IWineD3DVertexShaderImpl
*)This
)->min_rel_offset
)
298 ((IWineD3DVertexShaderImpl
*)This
)->min_rel_offset
= reg
->idx
;
299 else if (reg
->idx
>= ((IWineD3DVertexShaderImpl
*)This
)->max_rel_offset
)
300 ((IWineD3DVertexShaderImpl
*)This
)->max_rel_offset
= reg
->idx
;
302 reg_maps
->usesrelconstF
= TRUE
;
306 set_bitmap_bit(reg_maps
->constf
, reg
->idx
);
310 case WINED3DSPR_CONSTINT
:
311 reg_maps
->integer_constants
|= (1 << reg
->idx
);
314 case WINED3DSPR_CONSTBOOL
:
315 reg_maps
->boolean_constants
|= (1 << reg
->idx
);
319 TRACE("Not recording register of type %#x and idx %u\n", reg
->type
, reg
->idx
);
324 static unsigned int get_instr_extra_regcount(enum WINED3D_SHADER_INSTRUCTION_HANDLER instr
, unsigned int param
)
328 case WINED3DSIH_M4x4
:
329 case WINED3DSIH_M3x4
:
330 return param
== 1 ? 3 : 0;
332 case WINED3DSIH_M4x3
:
333 case WINED3DSIH_M3x3
:
334 return param
== 1 ? 2 : 0;
336 case WINED3DSIH_M3x2
:
337 return param
== 1 ? 1 : 0;
344 /* Note that this does not count the loop register
345 * as an address register. */
347 HRESULT
shader_get_registers_used(IWineD3DBaseShader
*iface
, const struct wined3d_shader_frontend
*fe
,
348 struct shader_reg_maps
*reg_maps
, struct wined3d_shader_semantic
*semantics_in
,
349 struct wined3d_shader_semantic
*semantics_out
, const DWORD
*byte_code
, DWORD constf_size
)
351 IWineD3DBaseShaderImpl
* This
= (IWineD3DBaseShaderImpl
*) iface
;
352 void *fe_data
= This
->baseShader
.frontend_data
;
353 struct wined3d_shader_version shader_version
;
354 unsigned int cur_loop_depth
= 0, max_loop_depth
= 0;
355 const DWORD
* pToken
= byte_code
;
358 /* There are some minor differences between pixel and vertex shaders */
360 memset(reg_maps
, 0, sizeof(*reg_maps
));
362 /* get_registers_used is called on every compile on some 1.x shaders, which can result
363 * in stacking up a collection of local constants. Delete the old constants if existing
365 shader_delete_constant_list(&This
->baseShader
.constantsF
);
366 shader_delete_constant_list(&This
->baseShader
.constantsB
);
367 shader_delete_constant_list(&This
->baseShader
.constantsI
);
369 fe
->shader_read_header(fe_data
, &pToken
, &shader_version
);
370 reg_maps
->shader_version
= shader_version
;
371 pshader
= shader_is_pshader_version(shader_version
.type
);
373 reg_maps
->constf
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
,
374 sizeof(*reg_maps
->constf
) * ((constf_size
+ 31) / 32));
375 if(!reg_maps
->constf
) {
376 ERR("Out of memory\n");
377 return E_OUTOFMEMORY
;
380 while (!fe
->shader_is_end(fe_data
, &pToken
))
382 struct wined3d_shader_instruction ins
;
387 fe
->shader_read_comment(&pToken
, &comment
);
388 if (comment
) continue;
391 fe
->shader_read_opcode(fe_data
, &pToken
, &ins
, ¶m_size
);
393 /* Unhandled opcode, and its parameters */
394 if (ins
.handler_idx
== WINED3DSIH_TABLE_SIZE
)
396 TRACE("Skipping unrecognized instruction.\n");
397 pToken
+= param_size
;
401 /* Handle declarations */
402 if (ins
.handler_idx
== WINED3DSIH_DCL
)
404 struct wined3d_shader_semantic semantic
;
406 fe
->shader_read_semantic(&pToken
, &semantic
);
408 switch (semantic
.reg
.reg
.type
)
410 /* Vshader: mark attributes used
411 * Pshader: mark 3.0 input registers used, save token */
412 case WINED3DSPR_INPUT
:
413 if (!pshader
) reg_maps
->attributes
[semantic
.reg
.reg
.idx
] = 1;
414 else reg_maps
->packed_input
[semantic
.reg
.reg
.idx
] = 1;
415 semantics_in
[semantic
.reg
.reg
.idx
] = semantic
;
418 /* Vshader: mark 3.0 output registers used, save token */
419 case WINED3DSPR_OUTPUT
:
420 reg_maps
->packed_output
[semantic
.reg
.reg
.idx
] = 1;
421 semantics_out
[semantic
.reg
.reg
.idx
] = semantic
;
422 if (semantic
.usage
== WINED3DDECLUSAGE_FOG
) reg_maps
->fog
= 1;
425 /* Save sampler usage token */
426 case WINED3DSPR_SAMPLER
:
427 reg_maps
->sampler_type
[semantic
.reg
.reg
.idx
] = semantic
.sampler_type
;
431 TRACE("Not recording DCL register type %#x.\n", semantic
.reg
.reg
.type
);
435 else if (ins
.handler_idx
== WINED3DSIH_DEF
)
437 struct wined3d_shader_dst_param dst
;
438 struct wined3d_shader_src_param rel_addr
;
440 local_constant
* lconst
= HeapAlloc(GetProcessHeap(), 0, sizeof(local_constant
));
441 if (!lconst
) return E_OUTOFMEMORY
;
443 fe
->shader_read_dst_param(fe_data
, &pToken
, &dst
, &rel_addr
);
444 lconst
->idx
= dst
.reg
.idx
;
446 memcpy(lconst
->value
, pToken
, 4 * sizeof(DWORD
));
449 /* In pixel shader 1.X shaders, the constants are clamped between [-1;1] */
450 if (shader_version
.major
== 1 && pshader
)
452 float *value
= (float *) lconst
->value
;
453 if(value
[0] < -1.0) value
[0] = -1.0;
454 else if(value
[0] > 1.0) value
[0] = 1.0;
455 if(value
[1] < -1.0) value
[1] = -1.0;
456 else if(value
[1] > 1.0) value
[1] = 1.0;
457 if(value
[2] < -1.0) value
[2] = -1.0;
458 else if(value
[2] > 1.0) value
[2] = 1.0;
459 if(value
[3] < -1.0) value
[3] = -1.0;
460 else if(value
[3] > 1.0) value
[3] = 1.0;
463 list_add_head(&This
->baseShader
.constantsF
, &lconst
->entry
);
465 else if (ins
.handler_idx
== WINED3DSIH_DEFI
)
467 struct wined3d_shader_dst_param dst
;
468 struct wined3d_shader_src_param rel_addr
;
470 local_constant
* lconst
= HeapAlloc(GetProcessHeap(), 0, sizeof(local_constant
));
471 if (!lconst
) return E_OUTOFMEMORY
;
473 fe
->shader_read_dst_param(fe_data
, &pToken
, &dst
, &rel_addr
);
474 lconst
->idx
= dst
.reg
.idx
;
476 memcpy(lconst
->value
, pToken
, 4 * sizeof(DWORD
));
479 list_add_head(&This
->baseShader
.constantsI
, &lconst
->entry
);
481 else if (ins
.handler_idx
== WINED3DSIH_DEFB
)
483 struct wined3d_shader_dst_param dst
;
484 struct wined3d_shader_src_param rel_addr
;
486 local_constant
* lconst
= HeapAlloc(GetProcessHeap(), 0, sizeof(local_constant
));
487 if (!lconst
) return E_OUTOFMEMORY
;
489 fe
->shader_read_dst_param(fe_data
, &pToken
, &dst
, &rel_addr
);
490 lconst
->idx
= dst
.reg
.idx
;
492 memcpy(lconst
->value
, pToken
, sizeof(DWORD
));
495 list_add_head(&This
->baseShader
.constantsB
, &lconst
->entry
);
497 /* If there's a loop in the shader */
498 else if (ins
.handler_idx
== WINED3DSIH_LOOP
499 || ins
.handler_idx
== WINED3DSIH_REP
)
501 struct wined3d_shader_src_param src
, rel_addr
;
503 fe
->shader_read_src_param(fe_data
, &pToken
, &src
, &rel_addr
);
505 /* Rep and Loop always use an integer constant for the control parameters */
506 if (ins
.handler_idx
== WINED3DSIH_REP
)
508 reg_maps
->integer_constants
|= 1 << src
.reg
.idx
;
512 fe
->shader_read_src_param(fe_data
, &pToken
, &src
, &rel_addr
);
513 reg_maps
->integer_constants
|= 1 << src
.reg
.idx
;
517 if(cur_loop_depth
> max_loop_depth
)
518 max_loop_depth
= cur_loop_depth
;
520 else if (ins
.handler_idx
== WINED3DSIH_ENDLOOP
521 || ins
.handler_idx
== WINED3DSIH_ENDREP
)
525 /* For subroutine prototypes */
526 else if (ins
.handler_idx
== WINED3DSIH_LABEL
)
528 struct wined3d_shader_src_param src
, rel_addr
;
530 fe
->shader_read_src_param(fe_data
, &pToken
, &src
, &rel_addr
);
531 reg_maps
->labels
[src
.reg
.idx
] = 1;
533 /* Set texture, address, temporary registers */
538 /* This will loop over all the registers and try to
539 * make a bitmask of the ones we're interested in.
541 * Relative addressing tokens are ignored, but that's
542 * okay, since we'll catch any address registers when
543 * they are initialized (required by spec) */
547 struct wined3d_shader_dst_param dst_param
;
548 struct wined3d_shader_src_param dst_rel_addr
;
550 fe
->shader_read_dst_param(fe_data
, &pToken
, &dst_param
, &dst_rel_addr
);
552 /* WINED3DSPR_TEXCRDOUT is the same as WINED3DSPR_OUTPUT. _OUTPUT can be > MAX_REG_TEXCRD and
553 * is used in >= 3.0 shaders. Filter 3.0 shaders to prevent overflows, and also filter pixel
554 * shaders because TECRDOUT isn't used in them, but future register types might cause issues */
555 if (!pshader
&& shader_version
.major
< 3 && dst_param
.reg
.type
== WINED3DSPR_TEXCRDOUT
)
557 reg_maps
->texcoord_mask
[dst_param
.reg
.type
] |= dst_param
.write_mask
;
561 shader_record_register_usage(This
, reg_maps
, &dst_param
.reg
, pshader
);
564 /* Declare 1.X samplers implicitly, based on the destination reg. number */
565 if (shader_version
.major
== 1
566 && pshader
/* Filter different instructions with the same enum values in VS */
567 && (ins
.handler_idx
== WINED3DSIH_TEX
568 || ins
.handler_idx
== WINED3DSIH_TEXBEM
569 || ins
.handler_idx
== WINED3DSIH_TEXBEML
570 || ins
.handler_idx
== WINED3DSIH_TEXDP3TEX
571 || ins
.handler_idx
== WINED3DSIH_TEXM3x2TEX
572 || ins
.handler_idx
== WINED3DSIH_TEXM3x3SPEC
573 || ins
.handler_idx
== WINED3DSIH_TEXM3x3TEX
574 || ins
.handler_idx
== WINED3DSIH_TEXM3x3VSPEC
575 || ins
.handler_idx
== WINED3DSIH_TEXREG2AR
576 || ins
.handler_idx
== WINED3DSIH_TEXREG2GB
577 || ins
.handler_idx
== WINED3DSIH_TEXREG2RGB
))
579 /* Fake sampler usage, only set reserved bit and ttype */
580 DWORD sampler_code
= dst_param
.reg
.idx
;
582 TRACE("Setting fake 2D sampler for 1.x pixelshader\n");
583 reg_maps
->sampler_type
[sampler_code
] = WINED3DSTT_2D
;
585 /* texbem is only valid with < 1.4 pixel shaders */
586 if (ins
.handler_idx
== WINED3DSIH_TEXBEM
587 || ins
.handler_idx
== WINED3DSIH_TEXBEML
)
589 reg_maps
->bumpmat
[sampler_code
] = TRUE
;
590 if (ins
.handler_idx
== WINED3DSIH_TEXBEML
)
592 reg_maps
->luminanceparams
[sampler_code
] = TRUE
;
596 else if (pshader
&& ins
.handler_idx
== WINED3DSIH_BEM
)
598 reg_maps
->bumpmat
[dst_param
.reg
.idx
] = TRUE
;
602 if (ins
.handler_idx
== WINED3DSIH_NRM
)
604 reg_maps
->usesnrm
= 1;
606 else if (ins
.handler_idx
== WINED3DSIH_DSY
)
608 reg_maps
->usesdsy
= 1;
610 else if(ins
.handler_idx
== WINED3DSIH_TEXLDD
)
612 reg_maps
->usestexldd
= 1;
614 else if(ins
.handler_idx
== WINED3DSIH_MOVA
)
616 reg_maps
->usesmova
= 1;
619 limit
= ins
.src_count
+ (ins
.predicate
? 1 : 0);
620 for (i
= 0; i
< limit
; ++i
)
622 struct wined3d_shader_src_param src_param
, src_rel_addr
;
625 fe
->shader_read_src_param(fe_data
, &pToken
, &src_param
, &src_rel_addr
);
626 count
= get_instr_extra_regcount(ins
.handler_idx
, i
);
628 shader_record_register_usage(This
, reg_maps
, &src_param
.reg
, pshader
);
632 shader_record_register_usage(This
, reg_maps
, &src_param
.reg
, pshader
);
638 reg_maps
->loop_depth
= max_loop_depth
;
640 This
->baseShader
.functionLength
= ((char *)pToken
- (char *)byte_code
);
645 static void shader_dump_decl_usage(const struct wined3d_shader_semantic
*semantic
,
646 const struct wined3d_shader_version
*shader_version
)
650 if (semantic
->reg
.reg
.type
== WINED3DSPR_SAMPLER
)
652 switch (semantic
->sampler_type
)
654 case WINED3DSTT_2D
: TRACE("_2d"); break;
655 case WINED3DSTT_CUBE
: TRACE("_cube"); break;
656 case WINED3DSTT_VOLUME
: TRACE("_volume"); break;
657 default: TRACE("_unknown_ttype(0x%08x)", semantic
->sampler_type
);
662 /* Pixel shaders 3.0 don't have usage semantics */
663 if (shader_is_pshader_version(shader_version
->type
) && shader_version
->major
< 3)
668 switch (semantic
->usage
)
670 case WINED3DDECLUSAGE_POSITION
:
671 TRACE("position%d", semantic
->usage_idx
);
673 case WINED3DDECLUSAGE_BLENDINDICES
:
676 case WINED3DDECLUSAGE_BLENDWEIGHT
:
679 case WINED3DDECLUSAGE_NORMAL
:
680 TRACE("normal%d", semantic
->usage_idx
);
682 case WINED3DDECLUSAGE_PSIZE
:
685 case WINED3DDECLUSAGE_COLOR
:
686 if (semantic
->usage_idx
== 0) TRACE("color");
687 else TRACE("specular%d", (semantic
->usage_idx
- 1));
689 case WINED3DDECLUSAGE_TEXCOORD
:
690 TRACE("texture%d", semantic
->usage_idx
);
692 case WINED3DDECLUSAGE_TANGENT
:
695 case WINED3DDECLUSAGE_BINORMAL
:
698 case WINED3DDECLUSAGE_TESSFACTOR
:
701 case WINED3DDECLUSAGE_POSITIONT
:
702 TRACE("positionT%d", semantic
->usage_idx
);
704 case WINED3DDECLUSAGE_FOG
:
707 case WINED3DDECLUSAGE_DEPTH
:
710 case WINED3DDECLUSAGE_SAMPLE
:
714 FIXME("unknown_semantics(0x%08x)", semantic
->usage
);
719 static void shader_dump_register(const struct wined3d_shader_register
*reg
,
720 const struct wined3d_shader_version
*shader_version
)
722 static const char * const rastout_reg_names
[] = {"oPos", "oFog", "oPts"};
723 static const char * const misctype_reg_names
[] = {"vPos", "vFace"};
724 UINT offset
= reg
->idx
;
728 case WINED3DSPR_TEMP
:
732 case WINED3DSPR_INPUT
:
736 case WINED3DSPR_CONST
:
737 case WINED3DSPR_CONST2
:
738 case WINED3DSPR_CONST3
:
739 case WINED3DSPR_CONST4
:
741 offset
= shader_get_float_offset(reg
->type
, reg
->idx
);
744 case WINED3DSPR_TEXTURE
: /* vs: case WINED3DSPR_ADDR */
745 TRACE("%c", shader_is_pshader_version(shader_version
->type
) ? 't' : 'a');
748 case WINED3DSPR_RASTOUT
:
749 TRACE("%s", rastout_reg_names
[reg
->idx
]);
752 case WINED3DSPR_COLOROUT
:
756 case WINED3DSPR_DEPTHOUT
:
760 case WINED3DSPR_ATTROUT
:
764 case WINED3DSPR_TEXCRDOUT
:
765 /* Vertex shaders >= 3.0 use general purpose output registers
766 * (WINED3DSPR_OUTPUT), which can include an address token */
767 if (shader_version
->major
>= 3) TRACE("o");
771 case WINED3DSPR_CONSTINT
:
775 case WINED3DSPR_CONSTBOOL
:
779 case WINED3DSPR_LABEL
:
783 case WINED3DSPR_LOOP
:
787 case WINED3DSPR_SAMPLER
:
791 case WINED3DSPR_MISCTYPE
:
792 if (reg
->idx
> 1) FIXME("Unhandled misctype register %d\n", reg
->idx
);
793 else TRACE("%s", misctype_reg_names
[reg
->idx
]);
796 case WINED3DSPR_PREDICATE
:
800 case WINED3DSPR_IMMCONST
:
805 TRACE("unhandled_rtype(%#x)", reg
->type
);
809 if (reg
->type
== WINED3DSPR_IMMCONST
)
812 switch (reg
->immconst_type
)
814 case WINED3D_IMMCONST_FLOAT
:
815 TRACE("%.8e", *(float *)reg
->immconst_data
);
818 case WINED3D_IMMCONST_FLOAT4
:
819 TRACE("%.8e, %.8e, %.8e, %.8e",
820 *(float *)®
->immconst_data
[0], *(float *)®
->immconst_data
[1],
821 *(float *)®
->immconst_data
[2], *(float *)®
->immconst_data
[3]);
825 TRACE("<unhandled immconst_type %#x>", reg
->immconst_type
);
830 else if (reg
->type
!= WINED3DSPR_RASTOUT
&& reg
->type
!= WINED3DSPR_MISCTYPE
)
835 shader_dump_src_param(reg
->rel_addr
, shader_version
);
839 if (reg
->rel_addr
) TRACE("]");
843 void shader_dump_dst_param(const struct wined3d_shader_dst_param
*param
,
844 const struct wined3d_shader_version
*shader_version
)
846 DWORD write_mask
= param
->write_mask
;
848 shader_dump_register(¶m
->reg
, shader_version
);
850 if (write_mask
!= WINED3DSP_WRITEMASK_ALL
)
852 static const char *write_mask_chars
= "xyzw";
855 if (write_mask
& WINED3DSP_WRITEMASK_0
) TRACE("%c", write_mask_chars
[0]);
856 if (write_mask
& WINED3DSP_WRITEMASK_1
) TRACE("%c", write_mask_chars
[1]);
857 if (write_mask
& WINED3DSP_WRITEMASK_2
) TRACE("%c", write_mask_chars
[2]);
858 if (write_mask
& WINED3DSP_WRITEMASK_3
) TRACE("%c", write_mask_chars
[3]);
862 void shader_dump_src_param(const struct wined3d_shader_src_param
*param
,
863 const struct wined3d_shader_version
*shader_version
)
865 DWORD src_modifier
= param
->modifiers
;
866 DWORD swizzle
= param
->swizzle
;
868 if (src_modifier
== WINED3DSPSM_NEG
869 || src_modifier
== WINED3DSPSM_BIASNEG
870 || src_modifier
== WINED3DSPSM_SIGNNEG
871 || src_modifier
== WINED3DSPSM_X2NEG
872 || src_modifier
== WINED3DSPSM_ABSNEG
)
874 else if (src_modifier
== WINED3DSPSM_COMP
)
876 else if (src_modifier
== WINED3DSPSM_NOT
)
879 if (src_modifier
== WINED3DSPSM_ABS
|| src_modifier
== WINED3DSPSM_ABSNEG
)
882 shader_dump_register(¶m
->reg
, shader_version
);
886 switch (src_modifier
)
888 case WINED3DSPSM_NONE
: break;
889 case WINED3DSPSM_NEG
: break;
890 case WINED3DSPSM_NOT
: break;
891 case WINED3DSPSM_BIAS
: TRACE("_bias"); break;
892 case WINED3DSPSM_BIASNEG
: TRACE("_bias"); break;
893 case WINED3DSPSM_SIGN
: TRACE("_bx2"); break;
894 case WINED3DSPSM_SIGNNEG
: TRACE("_bx2"); break;
895 case WINED3DSPSM_COMP
: break;
896 case WINED3DSPSM_X2
: TRACE("_x2"); break;
897 case WINED3DSPSM_X2NEG
: TRACE("_x2"); break;
898 case WINED3DSPSM_DZ
: TRACE("_dz"); break;
899 case WINED3DSPSM_DW
: TRACE("_dw"); break;
900 case WINED3DSPSM_ABSNEG
: TRACE(")"); break;
901 case WINED3DSPSM_ABS
: TRACE(")"); break;
903 TRACE("_unknown_modifier(%#x)", src_modifier
);
907 if (swizzle
!= WINED3DSP_NOSWIZZLE
)
909 static const char *swizzle_chars
= "xyzw";
910 DWORD swizzle_x
= swizzle
& 0x03;
911 DWORD swizzle_y
= (swizzle
>> 2) & 0x03;
912 DWORD swizzle_z
= (swizzle
>> 4) & 0x03;
913 DWORD swizzle_w
= (swizzle
>> 6) & 0x03;
915 if (swizzle_x
== swizzle_y
916 && swizzle_x
== swizzle_z
917 && swizzle_x
== swizzle_w
)
919 TRACE(".%c", swizzle_chars
[swizzle_x
]);
923 TRACE(".%c%c%c%c", swizzle_chars
[swizzle_x
], swizzle_chars
[swizzle_y
],
924 swizzle_chars
[swizzle_z
], swizzle_chars
[swizzle_w
]);
929 /* Shared code in order to generate the bulk of the shader string.
930 * NOTE: A description of how to parse tokens can be found on msdn */
931 void shader_generate_main(IWineD3DBaseShader
*iface
, SHADER_BUFFER
*buffer
,
932 const shader_reg_maps
*reg_maps
, const DWORD
*pFunction
, void *backend_ctx
)
934 IWineD3DBaseShaderImpl
* This
= (IWineD3DBaseShaderImpl
*) iface
;
935 IWineD3DDeviceImpl
*device
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
; /* To access shader backend callbacks */
936 const SHADER_HANDLER
*handler_table
= device
->shader_backend
->shader_instruction_handler_table
;
937 const struct wined3d_shader_frontend
*fe
= This
->baseShader
.frontend
;
938 void *fe_data
= This
->baseShader
.frontend_data
;
939 struct wined3d_shader_src_param src_rel_addr
[4];
940 struct wined3d_shader_src_param src_param
[4];
941 struct wined3d_shader_version shader_version
;
942 struct wined3d_shader_src_param dst_rel_addr
;
943 struct wined3d_shader_dst_param dst_param
;
944 struct wined3d_shader_instruction ins
;
945 struct wined3d_shader_context ctx
;
946 const DWORD
*pToken
= pFunction
;
947 SHADER_HANDLER hw_fct
;
950 /* Initialize current parsing state */
952 ctx
.reg_maps
= reg_maps
;
954 ctx
.backend_data
= backend_ctx
;
957 ins
.dst
= &dst_param
;
959 This
->baseShader
.parse_state
.current_row
= 0;
961 fe
->shader_read_header(fe_data
, &pToken
, &shader_version
);
963 while (!fe
->shader_is_end(fe_data
, &pToken
))
968 /* Skip comment tokens */
969 fe
->shader_read_comment(&pToken
, &comment
);
970 if (comment
) continue;
973 fe
->shader_read_opcode(fe_data
, &pToken
, &ins
, ¶m_size
);
975 /* Unknown opcode and its parameters */
976 if (ins
.handler_idx
== WINED3DSIH_TABLE_SIZE
)
978 TRACE("Skipping unrecognized instruction.\n");
979 pToken
+= param_size
;
984 if (ins
.handler_idx
== WINED3DSIH_DCL
985 || ins
.handler_idx
== WINED3DSIH_NOP
986 || ins
.handler_idx
== WINED3DSIH_DEF
987 || ins
.handler_idx
== WINED3DSIH_DEFI
988 || ins
.handler_idx
== WINED3DSIH_DEFB
989 || ins
.handler_idx
== WINED3DSIH_PHASE
990 || ins
.handler_idx
== WINED3DSIH_RET
)
992 pToken
+= param_size
;
997 hw_fct
= handler_table
[ins
.handler_idx
];
999 /* Unhandled opcode */
1002 FIXME("Backend can't handle opcode %#x\n", ins
.handler_idx
);
1003 pToken
+= param_size
;
1007 /* Destination token */
1008 if (ins
.dst_count
) fe
->shader_read_dst_param(fe_data
, &pToken
, &dst_param
, &dst_rel_addr
);
1010 /* Predication token */
1011 if (ins
.predicate
) ins
.predicate
= *pToken
++;
1013 /* Other source tokens */
1014 for (i
= 0; i
< ins
.src_count
; ++i
)
1016 fe
->shader_read_src_param(fe_data
, &pToken
, &src_param
[i
], &src_rel_addr
[i
]);
1019 /* Call appropriate function for output target */
1022 /* Process instruction modifiers for GLSL apps ( _sat, etc. ) */
1023 device
->shader_backend
->shader_add_instruction_modifiers(&ins
);
1027 static void shader_dump_ins_modifiers(const struct wined3d_shader_dst_param
*dst
)
1029 DWORD mmask
= dst
->modifiers
;
1034 case 13: TRACE("_d8"); break;
1035 case 14: TRACE("_d4"); break;
1036 case 15: TRACE("_d2"); break;
1037 case 1: TRACE("_x2"); break;
1038 case 2: TRACE("_x4"); break;
1039 case 3: TRACE("_x8"); break;
1040 default: TRACE("_unhandled_shift(%d)", dst
->shift
); break;
1043 if (mmask
& WINED3DSPDM_SATURATE
) TRACE("_sat");
1044 if (mmask
& WINED3DSPDM_PARTIALPRECISION
) TRACE("_pp");
1045 if (mmask
& WINED3DSPDM_MSAMPCENTROID
) TRACE("_centroid");
1047 mmask
&= ~(WINED3DSPDM_SATURATE
| WINED3DSPDM_PARTIALPRECISION
| WINED3DSPDM_MSAMPCENTROID
);
1049 FIXME("_unrecognized_modifier(%#x)", mmask
);
1052 void shader_trace_init(const struct wined3d_shader_frontend
*fe
, void *fe_data
, const DWORD
*pFunction
)
1054 struct wined3d_shader_version shader_version
;
1055 const DWORD
* pToken
= pFunction
;
1058 TRACE("Parsing %p\n", pFunction
);
1060 fe
->shader_read_header(fe_data
, &pToken
, &shader_version
);
1062 TRACE("%s_%u_%u\n", shader_is_pshader_version(shader_version
.type
) ? "ps": "vs",
1063 shader_version
.major
, shader_version
.minor
);
1065 while (!fe
->shader_is_end(fe_data
, &pToken
))
1067 struct wined3d_shader_instruction ins
;
1068 const char *comment
;
1072 fe
->shader_read_comment(&pToken
, &comment
);
1075 TRACE("//%s\n", comment
);
1079 fe
->shader_read_opcode(fe_data
, &pToken
, &ins
, ¶m_size
);
1080 if (ins
.handler_idx
== WINED3DSIH_TABLE_SIZE
)
1082 TRACE("Skipping unrecognized instruction.\n");
1083 pToken
+= param_size
;
1087 if (ins
.handler_idx
== WINED3DSIH_DCL
)
1089 struct wined3d_shader_semantic semantic
;
1091 fe
->shader_read_semantic(&pToken
, &semantic
);
1093 shader_dump_decl_usage(&semantic
, &shader_version
);
1094 shader_dump_ins_modifiers(&semantic
.reg
);
1096 shader_dump_dst_param(&semantic
.reg
, &shader_version
);
1098 else if (ins
.handler_idx
== WINED3DSIH_DEF
)
1100 struct wined3d_shader_dst_param dst
;
1101 struct wined3d_shader_src_param rel_addr
;
1103 fe
->shader_read_dst_param(fe_data
, &pToken
, &dst
, &rel_addr
);
1105 TRACE("def c%u = %f, %f, %f, %f", shader_get_float_offset(dst
.reg
.type
, dst
.reg
.idx
),
1106 *(const float *)(pToken
),
1107 *(const float *)(pToken
+ 1),
1108 *(const float *)(pToken
+ 2),
1109 *(const float *)(pToken
+ 3));
1112 else if (ins
.handler_idx
== WINED3DSIH_DEFI
)
1114 struct wined3d_shader_dst_param dst
;
1115 struct wined3d_shader_src_param rel_addr
;
1117 fe
->shader_read_dst_param(fe_data
, &pToken
, &dst
, &rel_addr
);
1119 TRACE("defi i%u = %d, %d, %d, %d", dst
.reg
.idx
,
1126 else if (ins
.handler_idx
== WINED3DSIH_DEFB
)
1128 struct wined3d_shader_dst_param dst
;
1129 struct wined3d_shader_src_param rel_addr
;
1131 fe
->shader_read_dst_param(fe_data
, &pToken
, &dst
, &rel_addr
);
1133 TRACE("defb b%u = %s", dst
.reg
.idx
, *pToken
? "true" : "false");
1138 struct wined3d_shader_src_param dst_rel_addr
, src_rel_addr
;
1139 struct wined3d_shader_dst_param dst_param
;
1140 struct wined3d_shader_src_param src_param
;
1144 fe
->shader_read_dst_param(fe_data
, &pToken
, &dst_param
, &dst_rel_addr
);
1147 /* Print out predication source token first - it follows
1148 * the destination token. */
1151 fe
->shader_read_src_param(fe_data
, &pToken
, &src_param
, &src_rel_addr
);
1153 shader_dump_src_param(&src_param
, &shader_version
);
1157 /* PixWin marks instructions with the coissue flag with a '+' */
1158 if (ins
.coissue
) TRACE("+");
1160 TRACE("%s", shader_opcode_names
[ins
.handler_idx
]);
1162 if (ins
.handler_idx
== WINED3DSIH_IFC
1163 || ins
.handler_idx
== WINED3DSIH_BREAKC
)
1167 case COMPARISON_GT
: TRACE("_gt"); break;
1168 case COMPARISON_EQ
: TRACE("_eq"); break;
1169 case COMPARISON_GE
: TRACE("_ge"); break;
1170 case COMPARISON_LT
: TRACE("_lt"); break;
1171 case COMPARISON_NE
: TRACE("_ne"); break;
1172 case COMPARISON_LE
: TRACE("_le"); break;
1173 default: TRACE("_(%u)", ins
.flags
);
1176 else if (ins
.handler_idx
== WINED3DSIH_TEX
1177 && shader_version
.major
>= 2
1178 && (ins
.flags
& WINED3DSI_TEXLD_PROJECT
))
1183 /* We already read the destination token, print it. */
1186 shader_dump_ins_modifiers(&dst_param
);
1188 shader_dump_dst_param(&dst_param
, &shader_version
);
1191 /* Other source tokens */
1192 for (i
= ins
.dst_count
; i
< (ins
.dst_count
+ ins
.src_count
); ++i
)
1194 fe
->shader_read_src_param(fe_data
, &pToken
, &src_param
, &src_rel_addr
);
1195 TRACE(!i
? " " : ", ");
1196 shader_dump_src_param(&src_param
, &shader_version
);
1203 void shader_cleanup(IWineD3DBaseShader
*iface
)
1205 IWineD3DBaseShaderImpl
*This
= (IWineD3DBaseShaderImpl
*)iface
;
1207 ((IWineD3DDeviceImpl
*)This
->baseShader
.device
)->shader_backend
->shader_destroy(iface
);
1208 HeapFree(GetProcessHeap(), 0, This
->baseShader
.reg_maps
.constf
);
1209 HeapFree(GetProcessHeap(), 0, This
->baseShader
.function
);
1210 shader_delete_constant_list(&This
->baseShader
.constantsF
);
1211 shader_delete_constant_list(&This
->baseShader
.constantsB
);
1212 shader_delete_constant_list(&This
->baseShader
.constantsI
);
1213 list_remove(&This
->baseShader
.shader_list_entry
);
1215 if (This
->baseShader
.frontend
&& This
->baseShader
.frontend_data
)
1217 This
->baseShader
.frontend
->shader_free(This
->baseShader
.frontend_data
);
1221 static const SHADER_HANDLER shader_none_instruction_handler_table
[WINED3DSIH_TABLE_SIZE
] = {0};
1222 static void shader_none_select(IWineD3DDevice
*iface
, BOOL usePS
, BOOL useVS
) {}
1223 static void shader_none_select_depth_blt(IWineD3DDevice
*iface
, enum tex_types tex_type
) {}
1224 static void shader_none_deselect_depth_blt(IWineD3DDevice
*iface
) {}
1225 static void shader_none_update_float_vertex_constants(IWineD3DDevice
*iface
, UINT start
, UINT count
) {}
1226 static void shader_none_update_float_pixel_constants(IWineD3DDevice
*iface
, UINT start
, UINT count
) {}
1227 static void shader_none_load_constants(IWineD3DDevice
*iface
, char usePS
, char useVS
) {}
1228 static void shader_none_load_np2fixup_constants(IWineD3DDevice
*iface
, char usePS
, char useVS
) {}
1229 static void shader_none_destroy(IWineD3DBaseShader
*iface
) {}
1230 static HRESULT
shader_none_alloc(IWineD3DDevice
*iface
) {return WINED3D_OK
;}
1231 static void shader_none_free(IWineD3DDevice
*iface
) {}
1232 static BOOL
shader_none_dirty_const(IWineD3DDevice
*iface
) {return FALSE
;}
1233 static GLuint
shader_none_generate_pshader(IWineD3DPixelShader
*iface
,
1234 SHADER_BUFFER
*buffer
, const struct ps_compile_args
*args
)
1236 FIXME("NONE shader backend asked to generate a pixel shader\n");
1239 static GLuint
shader_none_generate_vshader(IWineD3DVertexShader
*iface
,
1240 SHADER_BUFFER
*buffer
, const struct vs_compile_args
*args
)
1242 FIXME("NONE shader backend asked to generate a vertex shader\n");
1245 static void shader_none_add_instruction_modifiers(const struct wined3d_shader_instruction
*ins
) {}
1247 #define GLINFO_LOCATION (*gl_info)
1248 static void shader_none_get_caps(WINED3DDEVTYPE devtype
, const WineD3D_GL_Info
*gl_info
, struct shader_caps
*pCaps
)
1250 /* Set the shader caps to 0 for the none shader backend */
1251 pCaps
->VertexShaderVersion
= 0;
1252 pCaps
->PixelShaderVersion
= 0;
1253 pCaps
->PixelShader1xMaxValue
= 0.0;
1255 #undef GLINFO_LOCATION
1256 static BOOL
shader_none_color_fixup_supported(struct color_fixup_desc fixup
)
1258 if (TRACE_ON(d3d_shader
) && TRACE_ON(d3d
))
1260 TRACE("Checking support for fixup:\n");
1261 dump_color_fixup_desc(fixup
);
1264 /* Faked to make some apps happy. */
1265 if (!is_yuv_fixup(fixup
))
1271 TRACE("[FAILED]\n");
1275 const shader_backend_t none_shader_backend
= {
1276 shader_none_instruction_handler_table
,
1278 shader_none_select_depth_blt
,
1279 shader_none_deselect_depth_blt
,
1280 shader_none_update_float_vertex_constants
,
1281 shader_none_update_float_pixel_constants
,
1282 shader_none_load_constants
,
1283 shader_none_load_np2fixup_constants
,
1284 shader_none_destroy
,
1287 shader_none_dirty_const
,
1288 shader_none_generate_pshader
,
1289 shader_none_generate_vshader
,
1290 shader_none_get_caps
,
1291 shader_none_color_fixup_supported
,
1292 shader_none_add_instruction_modifiers
,