2 * GLSL pixel and vertex shader implementation
4 * Copyright 2006 Jason Green
5 * Copyright 2006-2007 Henri Verbeet
6 * Copyright 2007-2008 Stefan Dösinger for CodeWeavers
8 * This library is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU Lesser General Public
10 * License as published by the Free Software Foundation; either
11 * version 2.1 of the License, or (at your option) any later version.
13 * This library is distributed in the hope that it will be useful,
14 * but WITHOUT ANY WARRANTY; without even the implied warranty of
15 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
16 * Lesser General Public License for more details.
18 * You should have received a copy of the GNU Lesser General Public
19 * License along with this library; if not, write to the Free Software
20 * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301, USA
24 * D3D shader asm has swizzles on source parameters, and write masks for
25 * destination parameters. GLSL uses swizzles for both. The result of this is
26 * that for example "mov dst.xw, src.zyxw" becomes "dst.xw = src.zw" in GLSL.
27 * Ie, to generate a proper GLSL source swizzle, we need to take the D3D write
28 * mask for the destination parameter into account.
34 #include "wined3d_private.h"
36 WINE_DEFAULT_DEBUG_CHANNEL(d3d_shader
);
37 WINE_DECLARE_DEBUG_CHANNEL(d3d_constants
);
38 WINE_DECLARE_DEBUG_CHANNEL(d3d_caps
);
39 WINE_DECLARE_DEBUG_CHANNEL(d3d
);
41 #define GLINFO_LOCATION (*gl_info)
56 } glsl_sample_function_t
;
60 HEAP_NODE_TRAVERSE_LEFT
,
61 HEAP_NODE_TRAVERSE_RIGHT
,
73 struct constant_entry
*entries
;
74 unsigned int *positions
;
78 /* GLSL shader private data */
79 struct shader_glsl_priv
{
80 struct hash_table_t
*glsl_program_lookup
;
81 struct glsl_shader_prog_link
*glsl_program
;
82 struct constant_heap vconst_heap
;
83 struct constant_heap pconst_heap
;
85 GLhandleARB depth_blt_program
[tex_type_count
];
86 UINT next_constant_version
;
89 /* Struct to maintain data about a linked GLSL program */
90 struct glsl_shader_prog_link
{
91 struct list vshader_entry
;
92 struct list pshader_entry
;
93 GLhandleARB programId
;
94 GLhandleARB
*vuniformF_locations
;
95 GLhandleARB
*puniformF_locations
;
96 GLhandleARB vuniformI_locations
[MAX_CONST_I
];
97 GLhandleARB puniformI_locations
[MAX_CONST_I
];
98 GLhandleARB posFixup_location
;
99 GLhandleARB bumpenvmat_location
[MAX_TEXTURES
];
100 GLhandleARB luminancescale_location
[MAX_TEXTURES
];
101 GLhandleARB luminanceoffset_location
[MAX_TEXTURES
];
102 GLhandleARB ycorrection_location
;
103 GLenum vertex_color_clamp
;
105 IWineD3DPixelShader
*pshader
;
106 struct ps_compile_args ps_args
;
107 UINT constant_version
;
112 IWineD3DPixelShader
*pshader
;
113 struct ps_compile_args ps_args
;
114 } glsl_program_key_t
;
117 /** Prints the GLSL info log which will contain error messages if they exist */
118 static void print_glsl_info_log(const WineD3D_GL_Info
*gl_info
, GLhandleARB obj
)
120 int infologLength
= 0;
125 static const char * const spam
[] =
127 "Vertex shader was successfully compiled to run on hardware.\n", /* fglrx */
128 "Fragment shader was successfully compiled to run on hardware.\n", /* fglrx */
129 "Fragment shader(s) linked, vertex shader(s) linked. \n ", /* fglrx, with \n */
130 "Fragment shader(s) linked, vertex shader(s) linked.", /* fglrx, no \n */
131 "Vertex shader(s) linked, no fragment shader(s) defined. \n ", /* fglrx, with \n */
132 "Vertex shader(s) linked, no fragment shader(s) defined.", /* fglrx, no \n */
133 "Fragment shader was successfully compiled to run on hardware.\nWARNING: 0:1: extension 'GL_ARB_draw_buffers' is not supported",
134 "Fragment shader(s) linked, no vertex shader(s) defined.", /* fglrx, no \n */
135 "Fragment shader(s) linked, no vertex shader(s) defined. \n ", /* fglrx, with \n */
136 "WARNING: 0:2: extension 'GL_ARB_draw_buffers' is not supported\n" /* MacOS ati */
139 GL_EXTCALL(glGetObjectParameterivARB(obj
,
140 GL_OBJECT_INFO_LOG_LENGTH_ARB
,
143 /* A size of 1 is just a null-terminated string, so the log should be bigger than
144 * that if there are errors. */
145 if (infologLength
> 1)
147 /* Fglrx doesn't terminate the string properly, but it tells us the proper length.
148 * So use HEAP_ZERO_MEMORY to avoid uninitialized bytes
150 infoLog
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, infologLength
);
151 GL_EXTCALL(glGetInfoLogARB(obj
, infologLength
, NULL
, infoLog
));
154 for(i
= 0; i
< sizeof(spam
) / sizeof(spam
[0]); i
++) {
155 if(strcmp(infoLog
, spam
[i
]) == 0) {
161 TRACE("Spam received from GLSL shader #%u: %s\n", obj
, debugstr_a(infoLog
));
163 FIXME("Error received from GLSL shader #%u: %s\n", obj
, debugstr_a(infoLog
));
165 HeapFree(GetProcessHeap(), 0, infoLog
);
170 * Loads (pixel shader) samplers
172 static void shader_glsl_load_psamplers(const WineD3D_GL_Info
*gl_info
, IWineD3DStateBlock
*iface
, GLhandleARB programId
)
174 IWineD3DStateBlockImpl
* stateBlock
= (IWineD3DStateBlockImpl
*) iface
;
175 GLhandleARB name_loc
;
177 char sampler_name
[20];
179 for (i
= 0; i
< MAX_FRAGMENT_SAMPLERS
; ++i
) {
180 snprintf(sampler_name
, sizeof(sampler_name
), "Psampler%d", i
);
181 name_loc
= GL_EXTCALL(glGetUniformLocationARB(programId
, sampler_name
));
182 if (name_loc
!= -1) {
183 int mapped_unit
= stateBlock
->wineD3DDevice
->texUnitMap
[i
];
184 if (mapped_unit
!= -1 && mapped_unit
< GL_LIMITS(fragment_samplers
)) {
185 TRACE("Loading %s for texture %d\n", sampler_name
, mapped_unit
);
186 GL_EXTCALL(glUniform1iARB(name_loc
, mapped_unit
));
187 checkGLcall("glUniform1iARB");
189 ERR("Trying to load sampler %s on unsupported unit %d\n", sampler_name
, mapped_unit
);
195 static void shader_glsl_load_vsamplers(const WineD3D_GL_Info
*gl_info
, IWineD3DStateBlock
*iface
, GLhandleARB programId
)
197 IWineD3DStateBlockImpl
* stateBlock
= (IWineD3DStateBlockImpl
*) iface
;
198 GLhandleARB name_loc
;
199 char sampler_name
[20];
202 for (i
= 0; i
< MAX_VERTEX_SAMPLERS
; ++i
) {
203 snprintf(sampler_name
, sizeof(sampler_name
), "Vsampler%d", i
);
204 name_loc
= GL_EXTCALL(glGetUniformLocationARB(programId
, sampler_name
));
205 if (name_loc
!= -1) {
206 int mapped_unit
= stateBlock
->wineD3DDevice
->texUnitMap
[MAX_FRAGMENT_SAMPLERS
+ i
];
207 if (mapped_unit
!= -1 && mapped_unit
< GL_LIMITS(combined_samplers
)) {
208 TRACE("Loading %s for texture %d\n", sampler_name
, mapped_unit
);
209 GL_EXTCALL(glUniform1iARB(name_loc
, mapped_unit
));
210 checkGLcall("glUniform1iARB");
212 ERR("Trying to load sampler %s on unsupported unit %d\n", sampler_name
, mapped_unit
);
218 static inline void walk_constant_heap(const WineD3D_GL_Info
*gl_info
, const float *constants
,
219 const GLhandleARB
*constant_locations
, const struct constant_heap
*heap
, unsigned char *stack
, DWORD version
)
222 unsigned int heap_idx
= 1;
225 if (heap
->entries
[heap_idx
].version
<= version
) return;
227 idx
= heap
->entries
[heap_idx
].idx
;
228 if (constant_locations
[idx
] != -1) GL_EXTCALL(glUniform4fvARB(constant_locations
[idx
], 1, &constants
[idx
* 4]));
229 stack
[stack_idx
] = HEAP_NODE_TRAVERSE_LEFT
;
231 while (stack_idx
>= 0)
233 /* Note that we fall through to the next case statement. */
234 switch(stack
[stack_idx
])
236 case HEAP_NODE_TRAVERSE_LEFT
:
238 unsigned int left_idx
= heap_idx
<< 1;
239 if (left_idx
< heap
->size
&& heap
->entries
[left_idx
].version
> version
)
242 idx
= heap
->entries
[heap_idx
].idx
;
243 if (constant_locations
[idx
] != -1)
244 GL_EXTCALL(glUniform4fvARB(constant_locations
[idx
], 1, &constants
[idx
* 4]));
246 stack
[stack_idx
++] = HEAP_NODE_TRAVERSE_RIGHT
;
247 stack
[stack_idx
] = HEAP_NODE_TRAVERSE_LEFT
;
252 case HEAP_NODE_TRAVERSE_RIGHT
:
254 unsigned int right_idx
= (heap_idx
<< 1) + 1;
255 if (right_idx
< heap
->size
&& heap
->entries
[right_idx
].version
> version
)
257 heap_idx
= right_idx
;
258 idx
= heap
->entries
[heap_idx
].idx
;
259 if (constant_locations
[idx
] != -1)
260 GL_EXTCALL(glUniform4fvARB(constant_locations
[idx
], 1, &constants
[idx
* 4]));
262 stack
[stack_idx
++] = HEAP_NODE_POP
;
263 stack
[stack_idx
] = HEAP_NODE_TRAVERSE_LEFT
;
276 checkGLcall("walk_constant_heap()");
279 static inline void apply_clamped_constant(const WineD3D_GL_Info
*gl_info
, GLint location
, const GLfloat
*data
)
281 GLfloat clamped_constant
[4];
283 if (location
== -1) return;
285 clamped_constant
[0] = data
[0] < -1.0f
? -1.0f
: data
[0] > 1.0 ? 1.0 : data
[0];
286 clamped_constant
[1] = data
[1] < -1.0f
? -1.0f
: data
[1] > 1.0 ? 1.0 : data
[1];
287 clamped_constant
[2] = data
[2] < -1.0f
? -1.0f
: data
[2] > 1.0 ? 1.0 : data
[2];
288 clamped_constant
[3] = data
[3] < -1.0f
? -1.0f
: data
[3] > 1.0 ? 1.0 : data
[3];
290 GL_EXTCALL(glUniform4fvARB(location
, 1, clamped_constant
));
293 static inline void walk_constant_heap_clamped(const WineD3D_GL_Info
*gl_info
, const float *constants
,
294 const GLhandleARB
*constant_locations
, const struct constant_heap
*heap
, unsigned char *stack
, DWORD version
)
297 unsigned int heap_idx
= 1;
300 if (heap
->entries
[heap_idx
].version
<= version
) return;
302 idx
= heap
->entries
[heap_idx
].idx
;
303 apply_clamped_constant(gl_info
, constant_locations
[idx
], &constants
[idx
* 4]);
304 stack
[stack_idx
] = HEAP_NODE_TRAVERSE_LEFT
;
306 while (stack_idx
>= 0)
308 /* Note that we fall through to the next case statement. */
309 switch(stack
[stack_idx
])
311 case HEAP_NODE_TRAVERSE_LEFT
:
313 unsigned int left_idx
= heap_idx
<< 1;
314 if (left_idx
< heap
->size
&& heap
->entries
[left_idx
].version
> version
)
317 idx
= heap
->entries
[heap_idx
].idx
;
318 apply_clamped_constant(gl_info
, constant_locations
[idx
], &constants
[idx
* 4]);
320 stack
[stack_idx
++] = HEAP_NODE_TRAVERSE_RIGHT
;
321 stack
[stack_idx
] = HEAP_NODE_TRAVERSE_LEFT
;
326 case HEAP_NODE_TRAVERSE_RIGHT
:
328 unsigned int right_idx
= (heap_idx
<< 1) + 1;
329 if (right_idx
< heap
->size
&& heap
->entries
[right_idx
].version
> version
)
331 heap_idx
= right_idx
;
332 idx
= heap
->entries
[heap_idx
].idx
;
333 apply_clamped_constant(gl_info
, constant_locations
[idx
], &constants
[idx
* 4]);
335 stack
[stack_idx
++] = HEAP_NODE_POP
;
336 stack
[stack_idx
] = HEAP_NODE_TRAVERSE_LEFT
;
349 checkGLcall("walk_constant_heap_clamped()");
352 /* Loads floating point constants (aka uniforms) into the currently set GLSL program. */
353 static void shader_glsl_load_constantsF(IWineD3DBaseShaderImpl
*This
, const WineD3D_GL_Info
*gl_info
,
354 const float *constants
, const GLhandleARB
*constant_locations
, const struct constant_heap
*heap
,
355 unsigned char *stack
, UINT version
)
357 const local_constant
*lconst
;
359 /* 1.X pshaders have the constants clamped to [-1;1] implicitly. */
360 if (WINED3DSHADER_VERSION_MAJOR(This
->baseShader
.reg_maps
.shader_version
) == 1
361 && shader_is_pshader_version(This
->baseShader
.reg_maps
.shader_version
))
362 walk_constant_heap_clamped(gl_info
, constants
, constant_locations
, heap
, stack
, version
);
364 walk_constant_heap(gl_info
, constants
, constant_locations
, heap
, stack
, version
);
366 if (!This
->baseShader
.load_local_constsF
)
368 TRACE("No need to load local float constants for this shader\n");
372 /* Immediate constants are clamped to [-1;1] at shader creation time if needed */
373 LIST_FOR_EACH_ENTRY(lconst
, &This
->baseShader
.constantsF
, local_constant
, entry
)
375 GLhandleARB location
= constant_locations
[lconst
->idx
];
376 /* We found this uniform name in the program - go ahead and send the data */
377 if (location
!= -1) GL_EXTCALL(glUniform4fvARB(location
, 1, (const GLfloat
*)lconst
->value
));
379 checkGLcall("glUniform4fvARB()");
382 /* Loads integer constants (aka uniforms) into the currently set GLSL program. */
383 static void shader_glsl_load_constantsI(IWineD3DBaseShaderImpl
*This
, const WineD3D_GL_Info
*gl_info
,
384 const GLhandleARB locations
[MAX_CONST_I
], const int *constants
, WORD constants_set
)
389 for (i
= 0; constants_set
; constants_set
>>= 1, ++i
)
391 if (!(constants_set
& 1)) continue;
393 TRACE_(d3d_constants
)("Loading constants %u: %i, %i, %i, %i\n",
394 i
, constants
[i
*4], constants
[i
*4+1], constants
[i
*4+2], constants
[i
*4+3]);
396 /* We found this uniform name in the program - go ahead and send the data */
397 GL_EXTCALL(glUniform4ivARB(locations
[i
], 1, &constants
[i
*4]));
398 checkGLcall("glUniform4ivARB");
401 /* Load immediate constants */
402 ptr
= list_head(&This
->baseShader
.constantsI
);
404 const struct local_constant
*lconst
= LIST_ENTRY(ptr
, const struct local_constant
, entry
);
405 unsigned int idx
= lconst
->idx
;
406 const GLint
*values
= (const GLint
*)lconst
->value
;
408 TRACE_(d3d_constants
)("Loading local constants %i: %i, %i, %i, %i\n", idx
,
409 values
[0], values
[1], values
[2], values
[3]);
411 /* We found this uniform name in the program - go ahead and send the data */
412 GL_EXTCALL(glUniform4ivARB(locations
[idx
], 1, values
));
413 checkGLcall("glUniform4ivARB");
414 ptr
= list_next(&This
->baseShader
.constantsI
, ptr
);
418 /* Loads boolean constants (aka uniforms) into the currently set GLSL program. */
419 static void shader_glsl_load_constantsB(IWineD3DBaseShaderImpl
*This
, const WineD3D_GL_Info
*gl_info
,
420 GLhandleARB programId
, const BOOL
*constants
, WORD constants_set
)
425 char is_pshader
= shader_is_pshader_version(This
->baseShader
.reg_maps
.shader_version
);
426 const char* prefix
= is_pshader
? "PB":"VB";
429 /* TODO: Benchmark and see if it would be beneficial to store the
430 * locations of the constants to avoid looking up each time */
431 for (i
= 0; constants_set
; constants_set
>>= 1, ++i
)
433 if (!(constants_set
& 1)) continue;
435 TRACE_(d3d_constants
)("Loading constants %i: %i;\n", i
, constants
[i
]);
437 /* TODO: Benchmark and see if it would be beneficial to store the
438 * locations of the constants to avoid looking up each time */
439 snprintf(tmp_name
, sizeof(tmp_name
), "%s[%i]", prefix
, i
);
440 tmp_loc
= GL_EXTCALL(glGetUniformLocationARB(programId
, tmp_name
));
443 /* We found this uniform name in the program - go ahead and send the data */
444 GL_EXTCALL(glUniform1ivARB(tmp_loc
, 1, &constants
[i
]));
445 checkGLcall("glUniform1ivARB");
449 /* Load immediate constants */
450 ptr
= list_head(&This
->baseShader
.constantsB
);
452 const struct local_constant
*lconst
= LIST_ENTRY(ptr
, const struct local_constant
, entry
);
453 unsigned int idx
= lconst
->idx
;
454 const GLint
*values
= (const GLint
*)lconst
->value
;
456 TRACE_(d3d_constants
)("Loading local constants %i: %i\n", idx
, values
[0]);
458 snprintf(tmp_name
, sizeof(tmp_name
), "%s[%i]", prefix
, idx
);
459 tmp_loc
= GL_EXTCALL(glGetUniformLocationARB(programId
, tmp_name
));
461 /* We found this uniform name in the program - go ahead and send the data */
462 GL_EXTCALL(glUniform1ivARB(tmp_loc
, 1, values
));
463 checkGLcall("glUniform1ivARB");
465 ptr
= list_next(&This
->baseShader
.constantsB
, ptr
);
469 static void reset_program_constant_version(void *value
, void *context
)
471 struct glsl_shader_prog_link
*entry
= (struct glsl_shader_prog_link
*)value
;
472 entry
->constant_version
= 0;
476 * Loads the app-supplied constants into the currently set GLSL program.
478 static void shader_glsl_load_constants(
479 IWineD3DDevice
* device
,
481 char useVertexShader
) {
483 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) device
;
484 struct shader_glsl_priv
*priv
= (struct shader_glsl_priv
*)deviceImpl
->shader_priv
;
485 IWineD3DStateBlockImpl
* stateBlock
= deviceImpl
->stateBlock
;
486 const WineD3D_GL_Info
*gl_info
= &deviceImpl
->adapter
->gl_info
;
488 GLhandleARB programId
;
489 struct glsl_shader_prog_link
*prog
= priv
->glsl_program
;
490 UINT constant_version
;
494 /* No GLSL program set - nothing to do. */
497 programId
= prog
->programId
;
498 constant_version
= prog
->constant_version
;
500 if (useVertexShader
) {
501 IWineD3DBaseShaderImpl
* vshader
= (IWineD3DBaseShaderImpl
*) stateBlock
->vertexShader
;
503 /* Load DirectX 9 float constants/uniforms for vertex shader */
504 shader_glsl_load_constantsF(vshader
, gl_info
, stateBlock
->vertexShaderConstantF
,
505 prog
->vuniformF_locations
, &priv
->vconst_heap
, priv
->stack
, constant_version
);
507 /* Load DirectX 9 integer constants/uniforms for vertex shader */
508 if(vshader
->baseShader
.uses_int_consts
) {
509 shader_glsl_load_constantsI(vshader
, gl_info
, prog
->vuniformI_locations
,
510 stateBlock
->vertexShaderConstantI
, stateBlock
->changed
.vertexShaderConstantsI
);
513 /* Load DirectX 9 boolean constants/uniforms for vertex shader */
514 if(vshader
->baseShader
.uses_bool_consts
) {
515 shader_glsl_load_constantsB(vshader
, gl_info
, programId
,
516 stateBlock
->vertexShaderConstantB
, stateBlock
->changed
.vertexShaderConstantsB
);
519 /* Upload the position fixup params */
520 GL_EXTCALL(glUniform4fvARB(prog
->posFixup_location
, 1, &deviceImpl
->posFixup
[0]));
521 checkGLcall("glUniform4fvARB");
524 if (usePixelShader
) {
526 IWineD3DBaseShaderImpl
* pshader
= (IWineD3DBaseShaderImpl
*) stateBlock
->pixelShader
;
528 /* Load DirectX 9 float constants/uniforms for pixel shader */
529 shader_glsl_load_constantsF(pshader
, gl_info
, stateBlock
->pixelShaderConstantF
,
530 prog
->puniformF_locations
, &priv
->pconst_heap
, priv
->stack
, constant_version
);
532 /* Load DirectX 9 integer constants/uniforms for pixel shader */
533 if(pshader
->baseShader
.uses_int_consts
) {
534 shader_glsl_load_constantsI(pshader
, gl_info
, prog
->puniformI_locations
,
535 stateBlock
->pixelShaderConstantI
, stateBlock
->changed
.pixelShaderConstantsI
);
538 /* Load DirectX 9 boolean constants/uniforms for pixel shader */
539 if(pshader
->baseShader
.uses_bool_consts
) {
540 shader_glsl_load_constantsB(pshader
, gl_info
, programId
,
541 stateBlock
->pixelShaderConstantB
, stateBlock
->changed
.pixelShaderConstantsB
);
544 /* Upload the environment bump map matrix if needed. The needsbumpmat member specifies the texture stage to load the matrix from.
545 * It can't be 0 for a valid texbem instruction.
547 for(i
= 0; i
< ((IWineD3DPixelShaderImpl
*) pshader
)->numbumpenvmatconsts
; i
++) {
548 IWineD3DPixelShaderImpl
*ps
= (IWineD3DPixelShaderImpl
*) pshader
;
549 int stage
= ps
->luminanceconst
[i
].texunit
;
551 const float *data
= (const float *)&stateBlock
->textureState
[(int)ps
->bumpenvmatconst
[i
].texunit
][WINED3DTSS_BUMPENVMAT00
];
552 GL_EXTCALL(glUniformMatrix2fvARB(prog
->bumpenvmat_location
[i
], 1, 0, data
));
553 checkGLcall("glUniformMatrix2fvARB");
555 /* texbeml needs the luminance scale and offset too. If texbeml is used, needsbumpmat
556 * is set too, so we can check that in the needsbumpmat check
558 if(ps
->baseShader
.reg_maps
.luminanceparams
[stage
]) {
559 const GLfloat
*scale
= (const GLfloat
*)&stateBlock
->textureState
[stage
][WINED3DTSS_BUMPENVLSCALE
];
560 const GLfloat
*offset
= (const GLfloat
*)&stateBlock
->textureState
[stage
][WINED3DTSS_BUMPENVLOFFSET
];
562 GL_EXTCALL(glUniform1fvARB(prog
->luminancescale_location
[i
], 1, scale
));
563 checkGLcall("glUniform1fvARB");
564 GL_EXTCALL(glUniform1fvARB(prog
->luminanceoffset_location
[i
], 1, offset
));
565 checkGLcall("glUniform1fvARB");
569 if(((IWineD3DPixelShaderImpl
*) pshader
)->vpos_uniform
) {
570 float correction_params
[4];
571 if(deviceImpl
->render_offscreen
) {
572 correction_params
[0] = 0.0;
573 correction_params
[1] = 1.0;
575 /* position is window relative, not viewport relative */
576 correction_params
[0] = ((IWineD3DSurfaceImpl
*) deviceImpl
->render_targets
[0])->currentDesc
.Height
;
577 correction_params
[1] = -1.0;
579 GL_EXTCALL(glUniform4fvARB(prog
->ycorrection_location
, 1, correction_params
));
583 if (priv
->next_constant_version
== UINT_MAX
)
585 TRACE("Max constant version reached, resetting to 0.\n");
586 hash_table_for_each_entry(priv
->glsl_program_lookup
, reset_program_constant_version
, NULL
);
587 priv
->next_constant_version
= 1;
591 prog
->constant_version
= priv
->next_constant_version
++;
595 static inline void update_heap_entry(struct constant_heap
*heap
, unsigned int idx
,
596 unsigned int heap_idx
, DWORD new_version
)
598 struct constant_entry
*entries
= heap
->entries
;
599 unsigned int *positions
= heap
->positions
;
600 unsigned int parent_idx
;
604 parent_idx
= heap_idx
>> 1;
606 if (new_version
<= entries
[parent_idx
].version
) break;
608 entries
[heap_idx
] = entries
[parent_idx
];
609 positions
[entries
[parent_idx
].idx
] = heap_idx
;
610 heap_idx
= parent_idx
;
613 entries
[heap_idx
].version
= new_version
;
614 entries
[heap_idx
].idx
= idx
;
615 positions
[idx
] = heap_idx
;
618 static void shader_glsl_update_float_vertex_constants(IWineD3DDevice
*iface
, UINT start
, UINT count
)
620 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
621 struct shader_glsl_priv
*priv
= (struct shader_glsl_priv
*)This
->shader_priv
;
622 struct constant_heap
*heap
= &priv
->vconst_heap
;
625 for (i
= start
; i
< count
+ start
; ++i
)
627 if (!This
->stateBlock
->changed
.vertexShaderConstantsF
[i
])
628 update_heap_entry(heap
, i
, heap
->size
++, priv
->next_constant_version
);
630 update_heap_entry(heap
, i
, heap
->positions
[i
], priv
->next_constant_version
);
634 static void shader_glsl_update_float_pixel_constants(IWineD3DDevice
*iface
, UINT start
, UINT count
)
636 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
637 struct shader_glsl_priv
*priv
= (struct shader_glsl_priv
*)This
->shader_priv
;
638 struct constant_heap
*heap
= &priv
->pconst_heap
;
641 for (i
= start
; i
< count
+ start
; ++i
)
643 if (!This
->stateBlock
->changed
.pixelShaderConstantsF
[i
])
644 update_heap_entry(heap
, i
, heap
->size
++, priv
->next_constant_version
);
646 update_heap_entry(heap
, i
, heap
->positions
[i
], priv
->next_constant_version
);
650 /** Generate the variable & register declarations for the GLSL output target */
651 static void shader_generate_glsl_declarations(IWineD3DBaseShader
*iface
, const shader_reg_maps
*reg_maps
,
652 SHADER_BUFFER
*buffer
, const WineD3D_GL_Info
*gl_info
,
653 const struct ps_compile_args
*ps_args
)
655 IWineD3DBaseShaderImpl
* This
= (IWineD3DBaseShaderImpl
*) iface
;
656 IWineD3DDeviceImpl
*device
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
657 DWORD shader_version
= reg_maps
->shader_version
;
658 unsigned int i
, extra_constants_needed
= 0;
659 const local_constant
*lconst
;
661 /* There are some minor differences between pixel and vertex shaders */
662 char pshader
= shader_is_pshader_version(shader_version
);
663 char prefix
= pshader
? 'P' : 'V';
665 /* Prototype the subroutines */
666 for (i
= 0; i
< This
->baseShader
.limits
.label
; i
++) {
667 if (reg_maps
->labels
[i
])
668 shader_addline(buffer
, "void subroutine%u();\n", i
);
671 /* Declare the constants (aka uniforms) */
672 if (This
->baseShader
.limits
.constant_float
> 0) {
673 unsigned max_constantsF
= min(This
->baseShader
.limits
.constant_float
,
674 (pshader
? GL_LIMITS(pshader_constantsF
) : GL_LIMITS(vshader_constantsF
)));
675 shader_addline(buffer
, "uniform vec4 %cC[%u];\n", prefix
, max_constantsF
);
678 if (This
->baseShader
.limits
.constant_int
> 0)
679 shader_addline(buffer
, "uniform ivec4 %cI[%u];\n", prefix
, This
->baseShader
.limits
.constant_int
);
681 if (This
->baseShader
.limits
.constant_bool
> 0)
682 shader_addline(buffer
, "uniform bool %cB[%u];\n", prefix
, This
->baseShader
.limits
.constant_bool
);
685 shader_addline(buffer
, "uniform vec4 posFixup;\n");
686 /* Predeclaration; This function is added at link time based on the pixel shader.
687 * VS 3.0 shaders have an array OUT[] the shader writes to, earlier versions don't have
688 * that. We know the input to the reorder function at vertex shader compile time, so
689 * we can deal with that. The reorder function for a 1.x and 2.x vertex shader can just
690 * read gl_FrontColor. The output depends on the pixel shader. The reorder function for a
691 * 1.x and 2.x pshader or for fixed function will write gl_FrontColor, and for a 3.0 shader
692 * it will write to the varying array. Here we depend on the shader optimizer on sorting that
693 * out. The nvidia driver only does that if the parameter is inout instead of out, hence the
696 if (shader_version
>= WINED3DVS_VERSION(3, 0))
698 shader_addline(buffer
, "void order_ps_input(in vec4[%u]);\n", MAX_REG_OUTPUT
);
700 shader_addline(buffer
, "void order_ps_input();\n");
703 IWineD3DPixelShaderImpl
*ps_impl
= (IWineD3DPixelShaderImpl
*) This
;
705 ps_impl
->numbumpenvmatconsts
= 0;
706 for(i
= 0; i
< (sizeof(reg_maps
->bumpmat
) / sizeof(reg_maps
->bumpmat
[0])); i
++) {
707 if(!reg_maps
->bumpmat
[i
]) {
711 ps_impl
->bumpenvmatconst
[(int) ps_impl
->numbumpenvmatconsts
].texunit
= i
;
712 shader_addline(buffer
, "uniform mat2 bumpenvmat%d;\n", i
);
714 if(reg_maps
->luminanceparams
) {
715 ps_impl
->luminanceconst
[(int) ps_impl
->numbumpenvmatconsts
].texunit
= i
;
716 shader_addline(buffer
, "uniform float luminancescale%d;\n", i
);
717 shader_addline(buffer
, "uniform float luminanceoffset%d;\n", i
);
718 extra_constants_needed
++;
720 ps_impl
->luminanceconst
[(int) ps_impl
->numbumpenvmatconsts
].texunit
= -1;
723 extra_constants_needed
++;
724 ps_impl
->numbumpenvmatconsts
++;
727 if(ps_args
->srgb_correction
) {
728 shader_addline(buffer
, "const vec4 srgb_mul_low = vec4(%f, %f, %f, %f);\n",
729 srgb_mul_low
, srgb_mul_low
, srgb_mul_low
, srgb_mul_low
);
730 shader_addline(buffer
, "const vec4 srgb_comparison = vec4(%f, %f, %f, %f);\n",
731 srgb_cmp
, srgb_cmp
, srgb_cmp
, srgb_cmp
);
733 if(reg_maps
->vpos
|| reg_maps
->usesdsy
) {
734 if(This
->baseShader
.limits
.constant_float
+ extra_constants_needed
+ 1 < GL_LIMITS(pshader_constantsF
)) {
735 shader_addline(buffer
, "uniform vec4 ycorrection;\n");
736 ((IWineD3DPixelShaderImpl
*) This
)->vpos_uniform
= 1;
737 extra_constants_needed
++;
739 /* This happens because we do not have proper tracking of the constant registers that are
740 * actually used, only the max limit of the shader version
742 FIXME("Cannot find a free uniform for vpos correction params\n");
743 shader_addline(buffer
, "const vec4 ycorrection = vec4(%f, %f, 0.0, 0.0);\n",
744 device
->render_offscreen
? 0.0 : ((IWineD3DSurfaceImpl
*) device
->render_targets
[0])->currentDesc
.Height
,
745 device
->render_offscreen
? 1.0 : -1.0);
747 shader_addline(buffer
, "vec4 vpos;\n");
751 /* Declare texture samplers */
752 for (i
= 0; i
< This
->baseShader
.limits
.sampler
; i
++) {
753 if (reg_maps
->samplers
[i
]) {
755 DWORD stype
= reg_maps
->samplers
[i
] & WINED3DSP_TEXTURETYPE_MASK
;
759 shader_addline(buffer
, "uniform sampler1D %csampler%u;\n", prefix
, i
);
762 if(device
->stateBlock
->textures
[i
] &&
763 IWineD3DBaseTexture_GetTextureDimensions(device
->stateBlock
->textures
[i
]) == GL_TEXTURE_RECTANGLE_ARB
) {
764 shader_addline(buffer
, "uniform sampler2DRect %csampler%u;\n", prefix
, i
);
766 shader_addline(buffer
, "uniform sampler2D %csampler%u;\n", prefix
, i
);
769 case WINED3DSTT_CUBE
:
770 shader_addline(buffer
, "uniform samplerCube %csampler%u;\n", prefix
, i
);
772 case WINED3DSTT_VOLUME
:
773 shader_addline(buffer
, "uniform sampler3D %csampler%u;\n", prefix
, i
);
776 shader_addline(buffer
, "uniform unsupported_sampler %csampler%u;\n", prefix
, i
);
777 FIXME("Unrecognized sampler type: %#x\n", stype
);
783 /* Declare address variables */
784 for (i
= 0; i
< This
->baseShader
.limits
.address
; i
++) {
785 if (reg_maps
->address
[i
])
786 shader_addline(buffer
, "ivec4 A%d;\n", i
);
789 /* Declare texture coordinate temporaries and initialize them */
790 for (i
= 0; i
< This
->baseShader
.limits
.texcoord
; i
++) {
791 if (reg_maps
->texcoord
[i
])
792 shader_addline(buffer
, "vec4 T%u = gl_TexCoord[%u];\n", i
, i
);
795 /* Declare input register varyings. Only pixel shader, vertex shaders have that declared in the
796 * helper function shader that is linked in at link time
798 if (pshader
&& shader_version
>= WINED3DPS_VERSION(3, 0))
800 if (use_vs(device
->stateBlock
))
802 shader_addline(buffer
, "varying vec4 IN[%u];\n", GL_LIMITS(glsl_varyings
) / 4);
804 /* TODO: Write a replacement shader for the fixed function vertex pipeline, so this isn't needed.
805 * For fixed function vertex processing + 3.0 pixel shader we need a separate function in the
806 * pixel shader that reads the fixed function color into the packed input registers.
808 shader_addline(buffer
, "vec4 IN[%u];\n", GL_LIMITS(glsl_varyings
) / 4);
812 /* Declare output register temporaries */
813 if(This
->baseShader
.limits
.packed_output
) {
814 shader_addline(buffer
, "vec4 OUT[%u];\n", This
->baseShader
.limits
.packed_output
);
817 /* Declare temporary variables */
818 for(i
= 0; i
< This
->baseShader
.limits
.temporary
; i
++) {
819 if (reg_maps
->temporary
[i
])
820 shader_addline(buffer
, "vec4 R%u;\n", i
);
823 /* Declare attributes */
824 for (i
= 0; i
< This
->baseShader
.limits
.attributes
; i
++) {
825 if (reg_maps
->attributes
[i
])
826 shader_addline(buffer
, "attribute vec4 attrib%i;\n", i
);
829 /* Declare loop registers aLx */
830 for (i
= 0; i
< reg_maps
->loop_depth
; i
++) {
831 shader_addline(buffer
, "int aL%u;\n", i
);
832 shader_addline(buffer
, "int tmpInt%u;\n", i
);
835 /* Temporary variables for matrix operations */
836 shader_addline(buffer
, "vec4 tmp0;\n");
837 shader_addline(buffer
, "vec4 tmp1;\n");
839 /* Local constants use a different name so they can be loaded once at shader link time
840 * They can't be hardcoded into the shader text via LC = {x, y, z, w}; because the
841 * float -> string conversion can cause precision loss.
843 if(!This
->baseShader
.load_local_constsF
) {
844 LIST_FOR_EACH_ENTRY(lconst
, &This
->baseShader
.constantsF
, local_constant
, entry
) {
845 shader_addline(buffer
, "uniform vec4 %cLC%u;\n", prefix
, lconst
->idx
);
849 /* Start the main program */
850 shader_addline(buffer
, "void main() {\n");
851 if(pshader
&& reg_maps
->vpos
) {
852 /* DirectX apps expect integer values, while OpenGL drivers add approximately 0.5. This causes
853 * off-by-one problems as spotted by the vPos d3d9 visual test. Unfortunately the ATI cards do
854 * not add exactly 0.5, but rather something like 0.49999999 or 0.50000001, which still causes
855 * precision troubles when we just substract 0.5.
857 * To deal with that just floor() the position. This will eliminate the fraction on all cards.
859 * TODO: Test how that behaves with multisampling once we can enable multisampling in winex11.
861 * An advantage of floor is that it works even if the driver doesn't add 1/2. It is somewhat
862 * questionable if 1.5, 2.5, ... are the proper values to return in gl_FragCoord, even though
863 * coordinates specify the pixel centers instead of the pixel corners. This code will behave
864 * correctly on drivers that returns integer values.
866 shader_addline(buffer
, "vpos = floor(vec4(0, ycorrection[0], 0, 0) + gl_FragCoord * vec4(1, ycorrection[1], 1, 1));\n");
870 /*****************************************************************************
871 * Functions to generate GLSL strings from DirectX Shader bytecode begin here.
873 * For more information, see http://wiki.winehq.org/DirectX-Shaders
874 ****************************************************************************/
877 static void shader_glsl_add_src_param(const SHADER_OPCODE_ARG
*arg
, const DWORD param
,
878 const DWORD addr_token
, DWORD mask
, glsl_src_param_t
*src_param
);
880 /** Used for opcode modifiers - They multiply the result by the specified amount */
881 static const char * const shift_glsl_tab
[] = {
883 "2.0 * ", /* 1 (x2) */
884 "4.0 * ", /* 2 (x4) */
885 "8.0 * ", /* 3 (x8) */
886 "16.0 * ", /* 4 (x16) */
887 "32.0 * ", /* 5 (x32) */
894 "0.0625 * ", /* 12 (d16) */
895 "0.125 * ", /* 13 (d8) */
896 "0.25 * ", /* 14 (d4) */
897 "0.5 * " /* 15 (d2) */
900 /* Generate a GLSL parameter that does the input modifier computation and return the input register/mask to use */
901 static void shader_glsl_gen_modifier (
904 const char *in_regswizzle
,
909 if (instr
== WINED3DSIO_TEXKILL
)
912 switch (instr
& WINED3DSP_SRCMOD_MASK
) {
913 case WINED3DSPSM_DZ
: /* Need to handle this in the instructions itself (texld & texcrd). */
915 case WINED3DSPSM_NONE
:
916 sprintf(out_str
, "%s%s", in_reg
, in_regswizzle
);
918 case WINED3DSPSM_NEG
:
919 sprintf(out_str
, "-%s%s", in_reg
, in_regswizzle
);
921 case WINED3DSPSM_NOT
:
922 sprintf(out_str
, "!%s%s", in_reg
, in_regswizzle
);
924 case WINED3DSPSM_BIAS
:
925 sprintf(out_str
, "(%s%s - vec4(0.5)%s)", in_reg
, in_regswizzle
, in_regswizzle
);
927 case WINED3DSPSM_BIASNEG
:
928 sprintf(out_str
, "-(%s%s - vec4(0.5)%s)", in_reg
, in_regswizzle
, in_regswizzle
);
930 case WINED3DSPSM_SIGN
:
931 sprintf(out_str
, "(2.0 * (%s%s - 0.5))", in_reg
, in_regswizzle
);
933 case WINED3DSPSM_SIGNNEG
:
934 sprintf(out_str
, "-(2.0 * (%s%s - 0.5))", in_reg
, in_regswizzle
);
936 case WINED3DSPSM_COMP
:
937 sprintf(out_str
, "(1.0 - %s%s)", in_reg
, in_regswizzle
);
940 sprintf(out_str
, "(2.0 * %s%s)", in_reg
, in_regswizzle
);
942 case WINED3DSPSM_X2NEG
:
943 sprintf(out_str
, "-(2.0 * %s%s)", in_reg
, in_regswizzle
);
945 case WINED3DSPSM_ABS
:
946 sprintf(out_str
, "abs(%s%s)", in_reg
, in_regswizzle
);
948 case WINED3DSPSM_ABSNEG
:
949 sprintf(out_str
, "-abs(%s%s)", in_reg
, in_regswizzle
);
952 FIXME("Unhandled modifier %u\n", (instr
& WINED3DSP_SRCMOD_MASK
));
953 sprintf(out_str
, "%s%s", in_reg
, in_regswizzle
);
957 /** Writes the GLSL variable name that corresponds to the register that the
958 * DX opcode parameter is trying to access */
959 static void shader_glsl_get_register_name(const DWORD param
, const DWORD addr_token
,
960 char *regstr
, BOOL
*is_color
, const SHADER_OPCODE_ARG
*arg
)
962 /* oPos, oFog and oPts in D3D */
963 static const char * const hwrastout_reg_names
[] = { "gl_Position", "gl_FogFragCoord", "gl_PointSize" };
965 DWORD reg
= param
& WINED3DSP_REGNUM_MASK
;
966 DWORD regtype
= shader_get_regtype(param
);
967 IWineD3DBaseShaderImpl
* This
= (IWineD3DBaseShaderImpl
*) arg
->shader
;
968 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
969 const WineD3D_GL_Info
* gl_info
= &deviceImpl
->adapter
->gl_info
;
970 DWORD shader_version
= This
->baseShader
.reg_maps
.shader_version
;
971 char pshader
= shader_is_pshader_version(shader_version
);
977 case WINED3DSPR_TEMP
:
978 sprintf(tmpStr
, "R%u", reg
);
980 case WINED3DSPR_INPUT
:
982 /* Pixel shaders >= 3.0 */
983 if (WINED3DSHADER_VERSION_MAJOR(shader_version
) >= 3)
985 DWORD in_count
= GL_LIMITS(glsl_varyings
) / 4;
987 if (param
& WINED3DSHADER_ADDRMODE_RELATIVE
) {
988 glsl_src_param_t rel_param
;
989 shader_glsl_add_src_param(arg
, addr_token
, 0, WINED3DSP_WRITEMASK_0
, &rel_param
);
991 /* Removing a + 0 would be an obvious optimization, but macos doesn't see the NOP
994 if(((IWineD3DPixelShaderImpl
*) This
)->input_reg_map
[reg
]) {
995 if (((IWineD3DPixelShaderImpl
*)This
)->declared_in_count
> in_count
) {
996 sprintf(tmpStr
, "((%s + %u) > %d ? (%s + %u) > %d ? gl_SecondaryColor : gl_Color : IN[%s + %u])",
997 rel_param
.param_str
, ((IWineD3DPixelShaderImpl
*)This
)->input_reg_map
[reg
], in_count
- 1,
998 rel_param
.param_str
, ((IWineD3DPixelShaderImpl
*)This
)->input_reg_map
[reg
], in_count
,
999 rel_param
.param_str
, ((IWineD3DPixelShaderImpl
*)This
)->input_reg_map
[reg
]);
1001 sprintf(tmpStr
, "IN[%s + %u]", rel_param
.param_str
, ((IWineD3DPixelShaderImpl
*)This
)->input_reg_map
[reg
]);
1004 if (((IWineD3DPixelShaderImpl
*)This
)->declared_in_count
> in_count
) {
1005 sprintf(tmpStr
, "((%s) > %d ? (%s) > %d ? gl_SecondaryColor : gl_Color : IN[%s])",
1006 rel_param
.param_str
, in_count
- 1,
1007 rel_param
.param_str
, in_count
,
1008 rel_param
.param_str
);
1010 sprintf(tmpStr
, "IN[%s]", rel_param
.param_str
);
1014 DWORD idx
= ((IWineD3DPixelShaderImpl
*) This
)->input_reg_map
[reg
];
1015 if (idx
== in_count
) {
1016 sprintf(tmpStr
, "gl_Color");
1017 } else if (idx
== in_count
+ 1) {
1018 sprintf(tmpStr
, "gl_SecondaryColor");
1020 sprintf(tmpStr
, "IN[%u]", idx
);
1025 strcpy(tmpStr
, "gl_Color");
1027 strcpy(tmpStr
, "gl_SecondaryColor");
1030 if (vshader_input_is_color((IWineD3DVertexShader
*) This
, reg
))
1032 sprintf(tmpStr
, "attrib%u", reg
);
1035 case WINED3DSPR_CONST
:
1037 const char prefix
= pshader
? 'P':'V';
1039 /* Relative addressing */
1040 if (param
& WINED3DSHADER_ADDRMODE_RELATIVE
) {
1042 /* Relative addressing on shaders 2.0+ have a relative address token,
1043 * prior to that, it was hard-coded as "A0.x" because there's only 1 register */
1044 if (WINED3DSHADER_VERSION_MAJOR(shader_version
) >= 2)
1046 glsl_src_param_t rel_param
;
1047 shader_glsl_add_src_param(arg
, addr_token
, 0, WINED3DSP_WRITEMASK_0
, &rel_param
);
1049 sprintf(tmpStr
, "%cC[%s + %u]", prefix
, rel_param
.param_str
, reg
);
1051 sprintf(tmpStr
, "%cC[%s]", prefix
, rel_param
.param_str
);
1055 sprintf(tmpStr
, "%cC[A0.x + %u]", prefix
, reg
);
1057 sprintf(tmpStr
, "%cC[A0.x]", prefix
);
1062 if(shader_constant_is_local(This
, reg
)) {
1063 sprintf(tmpStr
, "%cLC%u", prefix
, reg
);
1065 sprintf(tmpStr
, "%cC[%u]", prefix
, reg
);
1071 case WINED3DSPR_CONSTINT
:
1073 sprintf(tmpStr
, "PI[%u]", reg
);
1075 sprintf(tmpStr
, "VI[%u]", reg
);
1077 case WINED3DSPR_CONSTBOOL
:
1079 sprintf(tmpStr
, "PB[%u]", reg
);
1081 sprintf(tmpStr
, "VB[%u]", reg
);
1083 case WINED3DSPR_TEXTURE
: /* case WINED3DSPR_ADDR: */
1085 sprintf(tmpStr
, "T%u", reg
);
1087 sprintf(tmpStr
, "A%u", reg
);
1090 case WINED3DSPR_LOOP
:
1091 sprintf(tmpStr
, "aL%u", This
->baseShader
.cur_loop_regno
- 1);
1093 case WINED3DSPR_SAMPLER
:
1095 sprintf(tmpStr
, "Psampler%u", reg
);
1097 sprintf(tmpStr
, "Vsampler%u", reg
);
1099 case WINED3DSPR_COLOROUT
:
1100 if (reg
>= GL_LIMITS(buffers
)) {
1101 WARN("Write to render target %u, only %d supported\n", reg
, 4);
1103 if (GL_SUPPORT(ARB_DRAW_BUFFERS
)) {
1104 sprintf(tmpStr
, "gl_FragData[%u]", reg
);
1105 } else { /* On older cards with GLSL support like the GeforceFX there's only one buffer. */
1106 sprintf(tmpStr
, "gl_FragColor");
1109 case WINED3DSPR_RASTOUT
:
1110 sprintf(tmpStr
, "%s", hwrastout_reg_names
[reg
]);
1112 case WINED3DSPR_DEPTHOUT
:
1113 sprintf(tmpStr
, "gl_FragDepth");
1115 case WINED3DSPR_ATTROUT
:
1117 sprintf(tmpStr
, "gl_FrontColor");
1119 sprintf(tmpStr
, "gl_FrontSecondaryColor");
1122 case WINED3DSPR_TEXCRDOUT
:
1123 /* Vertex shaders >= 3.0: WINED3DSPR_OUTPUT */
1124 if (WINED3DSHADER_VERSION_MAJOR(shader_version
) >= 3) sprintf(tmpStr
, "OUT[%u]", reg
);
1125 else sprintf(tmpStr
, "gl_TexCoord[%u]", reg
);
1127 case WINED3DSPR_MISCTYPE
:
1130 sprintf(tmpStr
, "vpos");
1131 } else if (reg
== 1){
1132 /* Note that gl_FrontFacing is a bool, while vFace is
1133 * a float for which the sign determines front/back
1135 sprintf(tmpStr
, "(gl_FrontFacing ? 1.0 : -1.0)");
1137 FIXME("Unhandled misctype register %d\n", reg
);
1138 sprintf(tmpStr
, "unrecognized_register");
1142 FIXME("Unhandled register name Type(%d)\n", regtype
);
1143 sprintf(tmpStr
, "unrecognized_register");
1147 strcat(regstr
, tmpStr
);
1150 /* Get the GLSL write mask for the destination register */
1151 static DWORD
shader_glsl_get_write_mask(const DWORD param
, char *write_mask
) {
1152 char *ptr
= write_mask
;
1153 DWORD mask
= param
& WINED3DSP_WRITEMASK_ALL
;
1155 if (shader_is_scalar(param
)) {
1156 mask
= WINED3DSP_WRITEMASK_0
;
1159 if (param
& WINED3DSP_WRITEMASK_0
) *ptr
++ = 'x';
1160 if (param
& WINED3DSP_WRITEMASK_1
) *ptr
++ = 'y';
1161 if (param
& WINED3DSP_WRITEMASK_2
) *ptr
++ = 'z';
1162 if (param
& WINED3DSP_WRITEMASK_3
) *ptr
++ = 'w';
1170 static unsigned int shader_glsl_get_write_mask_size(DWORD write_mask
) {
1171 unsigned int size
= 0;
1173 if (write_mask
& WINED3DSP_WRITEMASK_0
) ++size
;
1174 if (write_mask
& WINED3DSP_WRITEMASK_1
) ++size
;
1175 if (write_mask
& WINED3DSP_WRITEMASK_2
) ++size
;
1176 if (write_mask
& WINED3DSP_WRITEMASK_3
) ++size
;
1181 static void shader_glsl_get_swizzle(const DWORD param
, BOOL fixup
, DWORD mask
, char *swizzle_str
) {
1182 /* For registers of type WINED3DDECLTYPE_D3DCOLOR, data is stored as "bgra",
1183 * but addressed as "rgba". To fix this we need to swap the register's x
1184 * and z components. */
1185 DWORD swizzle
= (param
& WINED3DSP_SWIZZLE_MASK
) >> WINED3DSP_SWIZZLE_SHIFT
;
1186 const char *swizzle_chars
= fixup
? "zyxw" : "xyzw";
1187 char *ptr
= swizzle_str
;
1189 if (!shader_is_scalar(param
)) {
1191 /* swizzle bits fields: wwzzyyxx */
1192 if (mask
& WINED3DSP_WRITEMASK_0
) *ptr
++ = swizzle_chars
[swizzle
& 0x03];
1193 if (mask
& WINED3DSP_WRITEMASK_1
) *ptr
++ = swizzle_chars
[(swizzle
>> 2) & 0x03];
1194 if (mask
& WINED3DSP_WRITEMASK_2
) *ptr
++ = swizzle_chars
[(swizzle
>> 4) & 0x03];
1195 if (mask
& WINED3DSP_WRITEMASK_3
) *ptr
++ = swizzle_chars
[(swizzle
>> 6) & 0x03];
1201 /* From a given parameter token, generate the corresponding GLSL string.
1202 * Also, return the actual register name and swizzle in case the
1203 * caller needs this information as well. */
1204 static void shader_glsl_add_src_param(const SHADER_OPCODE_ARG
*arg
, const DWORD param
,
1205 const DWORD addr_token
, DWORD mask
, glsl_src_param_t
*src_param
)
1207 BOOL is_color
= FALSE
;
1208 char swizzle_str
[6];
1210 src_param
->reg_name
[0] = '\0';
1211 src_param
->param_str
[0] = '\0';
1212 swizzle_str
[0] = '\0';
1214 shader_glsl_get_register_name(param
, addr_token
, src_param
->reg_name
, &is_color
, arg
);
1216 shader_glsl_get_swizzle(param
, is_color
, mask
, swizzle_str
);
1217 shader_glsl_gen_modifier(param
, src_param
->reg_name
, swizzle_str
, src_param
->param_str
);
1220 /* From a given parameter token, generate the corresponding GLSL string.
1221 * Also, return the actual register name and swizzle in case the
1222 * caller needs this information as well. */
1223 static DWORD
shader_glsl_add_dst_param(const SHADER_OPCODE_ARG
* arg
, const DWORD param
,
1224 const DWORD addr_token
, glsl_dst_param_t
*dst_param
)
1226 BOOL is_color
= FALSE
;
1228 dst_param
->mask_str
[0] = '\0';
1229 dst_param
->reg_name
[0] = '\0';
1231 shader_glsl_get_register_name(param
, addr_token
, dst_param
->reg_name
, &is_color
, arg
);
1232 return shader_glsl_get_write_mask(param
, dst_param
->mask_str
);
1235 /* Append the destination part of the instruction to the buffer, return the effective write mask */
1236 static DWORD
shader_glsl_append_dst_ext(SHADER_BUFFER
*buffer
, const SHADER_OPCODE_ARG
*arg
, const DWORD param
)
1238 glsl_dst_param_t dst_param
;
1242 mask
= shader_glsl_add_dst_param(arg
, param
, arg
->dst_addr
, &dst_param
);
1245 shift
= (param
& WINED3DSP_DSTSHIFT_MASK
) >> WINED3DSP_DSTSHIFT_SHIFT
;
1246 shader_addline(buffer
, "%s%s = %s(", dst_param
.reg_name
, dst_param
.mask_str
, shift_glsl_tab
[shift
]);
1252 /* Append the destination part of the instruction to the buffer, return the effective write mask */
1253 static DWORD
shader_glsl_append_dst(SHADER_BUFFER
*buffer
, const SHADER_OPCODE_ARG
*arg
)
1255 return shader_glsl_append_dst_ext(buffer
, arg
, arg
->dst
);
1258 /** Process GLSL instruction modifiers */
1259 void shader_glsl_add_instruction_modifiers(const SHADER_OPCODE_ARG
* arg
)
1261 DWORD mask
= arg
->dst
& WINED3DSP_DSTMOD_MASK
;
1263 if (arg
->opcode
->dst_token
&& mask
!= 0) {
1264 glsl_dst_param_t dst_param
;
1266 shader_glsl_add_dst_param(arg
, arg
->dst
, 0, &dst_param
);
1268 if (mask
& WINED3DSPDM_SATURATE
) {
1269 /* _SAT means to clamp the value of the register to between 0 and 1 */
1270 shader_addline(arg
->buffer
, "%s%s = clamp(%s%s, 0.0, 1.0);\n", dst_param
.reg_name
,
1271 dst_param
.mask_str
, dst_param
.reg_name
, dst_param
.mask_str
);
1273 if (mask
& WINED3DSPDM_MSAMPCENTROID
) {
1274 FIXME("_centroid modifier not handled\n");
1276 if (mask
& WINED3DSPDM_PARTIALPRECISION
) {
1277 /* MSDN says this modifier can be safely ignored, so that's what we'll do. */
1282 static inline const char* shader_get_comp_op(
1283 const DWORD opcode
) {
1285 DWORD op
= (opcode
& INST_CONTROLS_MASK
) >> INST_CONTROLS_SHIFT
;
1287 case COMPARISON_GT
: return ">";
1288 case COMPARISON_EQ
: return "==";
1289 case COMPARISON_GE
: return ">=";
1290 case COMPARISON_LT
: return "<";
1291 case COMPARISON_NE
: return "!=";
1292 case COMPARISON_LE
: return "<=";
1294 FIXME("Unrecognized comparison value: %u\n", op
);
1299 static void shader_glsl_get_sample_function(DWORD sampler_type
, BOOL projected
, BOOL texrect
, glsl_sample_function_t
*sample_function
) {
1300 /* Note that there's no such thing as a projected cube texture. */
1301 switch(sampler_type
) {
1303 sample_function
->name
= projected
? "texture1DProj" : "texture1D";
1304 sample_function
->coord_mask
= WINED3DSP_WRITEMASK_0
;
1308 sample_function
->name
= projected
? "texture2DRectProj" : "texture2DRect";
1310 sample_function
->name
= projected
? "texture2DProj" : "texture2D";
1312 sample_function
->coord_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
;
1314 case WINED3DSTT_CUBE
:
1315 sample_function
->name
= "textureCube";
1316 sample_function
->coord_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
1318 case WINED3DSTT_VOLUME
:
1319 sample_function
->name
= projected
? "texture3DProj" : "texture3D";
1320 sample_function
->coord_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
1323 sample_function
->name
= "";
1324 sample_function
->coord_mask
= 0;
1325 FIXME("Unrecognized sampler type: %#x;\n", sampler_type
);
1330 static void shader_glsl_append_fixup_arg(char *arguments
, const char *reg_name
,
1331 BOOL sign_fixup
, enum fixup_channel_source channel_source
)
1333 switch(channel_source
)
1335 case CHANNEL_SOURCE_ZERO
:
1336 strcat(arguments
, "0.0");
1339 case CHANNEL_SOURCE_ONE
:
1340 strcat(arguments
, "1.0");
1343 case CHANNEL_SOURCE_X
:
1344 strcat(arguments
, reg_name
);
1345 strcat(arguments
, ".x");
1348 case CHANNEL_SOURCE_Y
:
1349 strcat(arguments
, reg_name
);
1350 strcat(arguments
, ".y");
1353 case CHANNEL_SOURCE_Z
:
1354 strcat(arguments
, reg_name
);
1355 strcat(arguments
, ".z");
1358 case CHANNEL_SOURCE_W
:
1359 strcat(arguments
, reg_name
);
1360 strcat(arguments
, ".w");
1364 FIXME("Unhandled channel source %#x\n", channel_source
);
1365 strcat(arguments
, "undefined");
1369 if (sign_fixup
) strcat(arguments
, " * 2.0 - 1.0");
1372 static void shader_glsl_color_correction(const struct SHADER_OPCODE_ARG
*arg
, struct color_fixup_desc fixup
)
1374 unsigned int mask_size
, remaining
;
1375 glsl_dst_param_t dst_param
;
1376 char arguments
[256];
1381 if (fixup
.x_sign_fixup
|| fixup
.x_source
!= CHANNEL_SOURCE_X
) mask
|= WINED3DSP_WRITEMASK_0
;
1382 if (fixup
.y_sign_fixup
|| fixup
.y_source
!= CHANNEL_SOURCE_Y
) mask
|= WINED3DSP_WRITEMASK_1
;
1383 if (fixup
.z_sign_fixup
|| fixup
.z_source
!= CHANNEL_SOURCE_Z
) mask
|= WINED3DSP_WRITEMASK_2
;
1384 if (fixup
.w_sign_fixup
|| fixup
.w_source
!= CHANNEL_SOURCE_W
) mask
|= WINED3DSP_WRITEMASK_3
;
1387 if (!mask
) return; /* Nothing to do */
1389 if (is_yuv_fixup(fixup
))
1391 enum yuv_fixup yuv_fixup
= get_yuv_fixup(fixup
);
1392 FIXME("YUV fixup (%#x) not supported\n", yuv_fixup
);
1396 mask_size
= shader_glsl_get_write_mask_size(mask
);
1398 dst_param
.mask_str
[0] = '\0';
1399 shader_glsl_get_write_mask(mask
, dst_param
.mask_str
);
1401 dst_param
.reg_name
[0] = '\0';
1402 shader_glsl_get_register_name(arg
->dst
, arg
->dst_addr
, dst_param
.reg_name
, &dummy
, arg
);
1404 arguments
[0] = '\0';
1405 remaining
= mask_size
;
1406 if (mask
& WINED3DSP_WRITEMASK_0
)
1408 shader_glsl_append_fixup_arg(arguments
, dst_param
.reg_name
, fixup
.x_sign_fixup
, fixup
.x_source
);
1409 if (--remaining
) strcat(arguments
, ", ");
1411 if (mask
& WINED3DSP_WRITEMASK_1
)
1413 shader_glsl_append_fixup_arg(arguments
, dst_param
.reg_name
, fixup
.y_sign_fixup
, fixup
.y_source
);
1414 if (--remaining
) strcat(arguments
, ", ");
1416 if (mask
& WINED3DSP_WRITEMASK_2
)
1418 shader_glsl_append_fixup_arg(arguments
, dst_param
.reg_name
, fixup
.z_sign_fixup
, fixup
.z_source
);
1419 if (--remaining
) strcat(arguments
, ", ");
1421 if (mask
& WINED3DSP_WRITEMASK_3
)
1423 shader_glsl_append_fixup_arg(arguments
, dst_param
.reg_name
, fixup
.w_sign_fixup
, fixup
.w_source
);
1424 if (--remaining
) strcat(arguments
, ", ");
1429 shader_addline(arg
->buffer
, "%s%s = vec%u(%s);\n",
1430 dst_param
.reg_name
, dst_param
.mask_str
, mask_size
, arguments
);
1434 shader_addline(arg
->buffer
, "%s%s = %s;\n", dst_param
.reg_name
, dst_param
.mask_str
, arguments
);
1438 /*****************************************************************************
1440 * Begin processing individual instruction opcodes
1442 ****************************************************************************/
1444 /* Generate GLSL arithmetic functions (dst = src1 + src2) */
1445 static void shader_glsl_arith(const SHADER_OPCODE_ARG
*arg
)
1447 CONST SHADER_OPCODE
* curOpcode
= arg
->opcode
;
1448 SHADER_BUFFER
* buffer
= arg
->buffer
;
1449 glsl_src_param_t src0_param
;
1450 glsl_src_param_t src1_param
;
1454 /* Determine the GLSL operator to use based on the opcode */
1455 switch (curOpcode
->opcode
) {
1456 case WINED3DSIO_MUL
: op
= '*'; break;
1457 case WINED3DSIO_ADD
: op
= '+'; break;
1458 case WINED3DSIO_SUB
: op
= '-'; break;
1461 FIXME("Opcode %s not yet handled in GLSL\n", curOpcode
->name
);
1465 write_mask
= shader_glsl_append_dst(buffer
, arg
);
1466 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], write_mask
, &src0_param
);
1467 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], write_mask
, &src1_param
);
1468 shader_addline(buffer
, "%s %c %s);\n", src0_param
.param_str
, op
, src1_param
.param_str
);
1471 /* Process the WINED3DSIO_MOV opcode using GLSL (dst = src) */
1472 static void shader_glsl_mov(const SHADER_OPCODE_ARG
*arg
)
1474 SHADER_BUFFER
* buffer
= arg
->buffer
;
1475 glsl_src_param_t src0_param
;
1478 write_mask
= shader_glsl_append_dst(buffer
, arg
);
1479 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], write_mask
, &src0_param
);
1481 /* In vs_1_1 WINED3DSIO_MOV can write to the address register. In later
1482 * shader versions WINED3DSIO_MOVA is used for this. */
1483 if ((WINED3DSHADER_VERSION_MAJOR(arg
->reg_maps
->shader_version
) == 1
1484 && !shader_is_pshader_version(arg
->reg_maps
->shader_version
)
1485 && shader_get_regtype(arg
->dst
) == WINED3DSPR_ADDR
))
1487 /* This is a simple floor() */
1488 unsigned int mask_size
= shader_glsl_get_write_mask_size(write_mask
);
1489 if (mask_size
> 1) {
1490 shader_addline(buffer
, "ivec%d(floor(%s)));\n", mask_size
, src0_param
.param_str
);
1492 shader_addline(buffer
, "int(floor(%s)));\n", src0_param
.param_str
);
1494 } else if(arg
->opcode
->opcode
== WINED3DSIO_MOVA
) {
1495 /* We need to *round* to the nearest int here. */
1496 unsigned int mask_size
= shader_glsl_get_write_mask_size(write_mask
);
1497 if (mask_size
> 1) {
1498 shader_addline(buffer
, "ivec%d(floor(abs(%s) + vec%d(0.5)) * sign(%s)));\n", mask_size
, src0_param
.param_str
, mask_size
, src0_param
.param_str
);
1500 shader_addline(buffer
, "int(floor(abs(%s) + 0.5) * sign(%s)));\n", src0_param
.param_str
, src0_param
.param_str
);
1503 shader_addline(buffer
, "%s);\n", src0_param
.param_str
);
1507 /* Process the dot product operators DP3 and DP4 in GLSL (dst = dot(src0, src1)) */
1508 static void shader_glsl_dot(const SHADER_OPCODE_ARG
*arg
)
1510 CONST SHADER_OPCODE
* curOpcode
= arg
->opcode
;
1511 SHADER_BUFFER
* buffer
= arg
->buffer
;
1512 glsl_src_param_t src0_param
;
1513 glsl_src_param_t src1_param
;
1514 DWORD dst_write_mask
, src_write_mask
;
1515 unsigned int dst_size
= 0;
1517 dst_write_mask
= shader_glsl_append_dst(buffer
, arg
);
1518 dst_size
= shader_glsl_get_write_mask_size(dst_write_mask
);
1520 /* dp3 works on vec3, dp4 on vec4 */
1521 if (curOpcode
->opcode
== WINED3DSIO_DP4
) {
1522 src_write_mask
= WINED3DSP_WRITEMASK_ALL
;
1524 src_write_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
1527 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], src_write_mask
, &src0_param
);
1528 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], src_write_mask
, &src1_param
);
1531 shader_addline(buffer
, "vec%d(dot(%s, %s)));\n", dst_size
, src0_param
.param_str
, src1_param
.param_str
);
1533 shader_addline(buffer
, "dot(%s, %s));\n", src0_param
.param_str
, src1_param
.param_str
);
1537 /* Note that this instruction has some restrictions. The destination write mask
1538 * can't contain the w component, and the source swizzles have to be .xyzw */
1539 static void shader_glsl_cross(const SHADER_OPCODE_ARG
*arg
)
1541 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
1542 glsl_src_param_t src0_param
;
1543 glsl_src_param_t src1_param
;
1546 shader_glsl_get_write_mask(arg
->dst
, dst_mask
);
1547 shader_glsl_append_dst(arg
->buffer
, arg
);
1548 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], src_mask
, &src0_param
);
1549 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], src_mask
, &src1_param
);
1550 shader_addline(arg
->buffer
, "cross(%s, %s)%s);\n", src0_param
.param_str
, src1_param
.param_str
, dst_mask
);
1553 /* Process the WINED3DSIO_POW instruction in GLSL (dst = |src0|^src1)
1554 * Src0 and src1 are scalars. Note that D3D uses the absolute of src0, while
1555 * GLSL uses the value as-is. */
1556 static void shader_glsl_pow(const SHADER_OPCODE_ARG
*arg
)
1558 SHADER_BUFFER
*buffer
= arg
->buffer
;
1559 glsl_src_param_t src0_param
;
1560 glsl_src_param_t src1_param
;
1561 DWORD dst_write_mask
;
1562 unsigned int dst_size
;
1564 dst_write_mask
= shader_glsl_append_dst(buffer
, arg
);
1565 dst_size
= shader_glsl_get_write_mask_size(dst_write_mask
);
1567 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
1568 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], WINED3DSP_WRITEMASK_0
, &src1_param
);
1571 shader_addline(buffer
, "vec%d(pow(abs(%s), %s)));\n", dst_size
, src0_param
.param_str
, src1_param
.param_str
);
1573 shader_addline(buffer
, "pow(abs(%s), %s));\n", src0_param
.param_str
, src1_param
.param_str
);
1577 /* Process the WINED3DSIO_LOG instruction in GLSL (dst = log2(|src0|))
1578 * Src0 is a scalar. Note that D3D uses the absolute of src0, while
1579 * GLSL uses the value as-is. */
1580 static void shader_glsl_log(const SHADER_OPCODE_ARG
*arg
)
1582 SHADER_BUFFER
*buffer
= arg
->buffer
;
1583 glsl_src_param_t src0_param
;
1584 DWORD dst_write_mask
;
1585 unsigned int dst_size
;
1587 dst_write_mask
= shader_glsl_append_dst(buffer
, arg
);
1588 dst_size
= shader_glsl_get_write_mask_size(dst_write_mask
);
1590 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
1593 shader_addline(buffer
, "vec%d(log2(abs(%s))));\n", dst_size
, src0_param
.param_str
);
1595 shader_addline(buffer
, "log2(abs(%s)));\n", src0_param
.param_str
);
1599 /* Map the opcode 1-to-1 to the GL code (arg->dst = instruction(src0, src1, ...) */
1600 static void shader_glsl_map2gl(const SHADER_OPCODE_ARG
*arg
)
1602 CONST SHADER_OPCODE
* curOpcode
= arg
->opcode
;
1603 SHADER_BUFFER
* buffer
= arg
->buffer
;
1604 glsl_src_param_t src_param
;
1605 const char *instruction
;
1606 char arguments
[256];
1610 /* Determine the GLSL function to use based on the opcode */
1611 /* TODO: Possibly make this a table for faster lookups */
1612 switch (curOpcode
->opcode
) {
1613 case WINED3DSIO_MIN
: instruction
= "min"; break;
1614 case WINED3DSIO_MAX
: instruction
= "max"; break;
1615 case WINED3DSIO_ABS
: instruction
= "abs"; break;
1616 case WINED3DSIO_FRC
: instruction
= "fract"; break;
1617 case WINED3DSIO_NRM
: instruction
= "normalize"; break;
1618 case WINED3DSIO_EXP
: instruction
= "exp2"; break;
1619 case WINED3DSIO_SGN
: instruction
= "sign"; break;
1620 case WINED3DSIO_DSX
: instruction
= "dFdx"; break;
1621 case WINED3DSIO_DSY
: instruction
= "ycorrection.y * dFdy"; break;
1622 default: instruction
= "";
1623 FIXME("Opcode %s not yet handled in GLSL\n", curOpcode
->name
);
1627 write_mask
= shader_glsl_append_dst(buffer
, arg
);
1629 arguments
[0] = '\0';
1630 if (curOpcode
->num_params
> 0) {
1631 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], write_mask
, &src_param
);
1632 strcat(arguments
, src_param
.param_str
);
1633 for (i
= 2; i
< curOpcode
->num_params
; ++i
) {
1634 strcat(arguments
, ", ");
1635 shader_glsl_add_src_param(arg
, arg
->src
[i
-1], arg
->src_addr
[i
-1], write_mask
, &src_param
);
1636 strcat(arguments
, src_param
.param_str
);
1640 shader_addline(buffer
, "%s(%s));\n", instruction
, arguments
);
1643 /** Process the WINED3DSIO_EXPP instruction in GLSL:
1644 * For shader model 1.x, do the following (and honor the writemask, so use a temporary variable):
1645 * dst.x = 2^(floor(src))
1646 * dst.y = src - floor(src)
1647 * dst.z = 2^src (partial precision is allowed, but optional)
1649 * For 2.0 shaders, just do this (honoring writemask and swizzle):
1650 * dst = 2^src; (partial precision is allowed, but optional)
1652 static void shader_glsl_expp(const SHADER_OPCODE_ARG
*arg
)
1654 glsl_src_param_t src_param
;
1656 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src_param
);
1658 if (arg
->reg_maps
->shader_version
< WINED3DPS_VERSION(2,0))
1662 shader_addline(arg
->buffer
, "tmp0.x = exp2(floor(%s));\n", src_param
.param_str
);
1663 shader_addline(arg
->buffer
, "tmp0.y = %s - floor(%s);\n", src_param
.param_str
, src_param
.param_str
);
1664 shader_addline(arg
->buffer
, "tmp0.z = exp2(%s);\n", src_param
.param_str
);
1665 shader_addline(arg
->buffer
, "tmp0.w = 1.0;\n");
1667 shader_glsl_append_dst(arg
->buffer
, arg
);
1668 shader_glsl_get_write_mask(arg
->dst
, dst_mask
);
1669 shader_addline(arg
->buffer
, "tmp0%s);\n", dst_mask
);
1672 unsigned int mask_size
;
1674 write_mask
= shader_glsl_append_dst(arg
->buffer
, arg
);
1675 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
1677 if (mask_size
> 1) {
1678 shader_addline(arg
->buffer
, "vec%d(exp2(%s)));\n", mask_size
, src_param
.param_str
);
1680 shader_addline(arg
->buffer
, "exp2(%s));\n", src_param
.param_str
);
1685 /** Process the RCP (reciprocal or inverse) opcode in GLSL (dst = 1 / src) */
1686 static void shader_glsl_rcp(const SHADER_OPCODE_ARG
*arg
)
1688 glsl_src_param_t src_param
;
1690 unsigned int mask_size
;
1692 write_mask
= shader_glsl_append_dst(arg
->buffer
, arg
);
1693 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
1694 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_3
, &src_param
);
1696 if (mask_size
> 1) {
1697 shader_addline(arg
->buffer
, "vec%d(1.0 / %s));\n", mask_size
, src_param
.param_str
);
1699 shader_addline(arg
->buffer
, "1.0 / %s);\n", src_param
.param_str
);
1703 static void shader_glsl_rsq(const SHADER_OPCODE_ARG
*arg
)
1705 SHADER_BUFFER
* buffer
= arg
->buffer
;
1706 glsl_src_param_t src_param
;
1708 unsigned int mask_size
;
1710 write_mask
= shader_glsl_append_dst(buffer
, arg
);
1711 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
1713 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_3
, &src_param
);
1715 if (mask_size
> 1) {
1716 shader_addline(buffer
, "vec%d(inversesqrt(%s)));\n", mask_size
, src_param
.param_str
);
1718 shader_addline(buffer
, "inversesqrt(%s));\n", src_param
.param_str
);
1722 /** Process signed comparison opcodes in GLSL. */
1723 static void shader_glsl_compare(const SHADER_OPCODE_ARG
*arg
)
1725 glsl_src_param_t src0_param
;
1726 glsl_src_param_t src1_param
;
1728 unsigned int mask_size
;
1730 write_mask
= shader_glsl_append_dst(arg
->buffer
, arg
);
1731 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
1732 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], write_mask
, &src0_param
);
1733 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], write_mask
, &src1_param
);
1735 if (mask_size
> 1) {
1736 const char *compare
;
1738 switch(arg
->opcode
->opcode
) {
1739 case WINED3DSIO_SLT
: compare
= "lessThan"; break;
1740 case WINED3DSIO_SGE
: compare
= "greaterThanEqual"; break;
1741 default: compare
= "";
1742 FIXME("Can't handle opcode %s\n", arg
->opcode
->name
);
1745 shader_addline(arg
->buffer
, "vec%d(%s(%s, %s)));\n", mask_size
, compare
,
1746 src0_param
.param_str
, src1_param
.param_str
);
1748 switch(arg
->opcode
->opcode
) {
1749 case WINED3DSIO_SLT
:
1750 /* Step(src0, src1) is not suitable here because if src0 == src1 SLT is supposed,
1751 * to return 0.0 but step returns 1.0 because step is not < x
1752 * An alternative is a bvec compare padded with an unused second component.
1753 * step(src1 * -1.0, src0 * -1.0) is not an option because it suffers from the same
1754 * issue. Playing with not() is not possible either because not() does not accept
1757 shader_addline(arg
->buffer
, "(%s < %s) ? 1.0 : 0.0);\n", src0_param
.param_str
, src1_param
.param_str
);
1759 case WINED3DSIO_SGE
:
1760 /* Here we can use the step() function and safe a conditional */
1761 shader_addline(arg
->buffer
, "step(%s, %s));\n", src1_param
.param_str
, src0_param
.param_str
);
1764 FIXME("Can't handle opcode %s\n", arg
->opcode
->name
);
1770 /** Process CMP instruction in GLSL (dst = src0 >= 0.0 ? src1 : src2), per channel */
1771 static void shader_glsl_cmp(const SHADER_OPCODE_ARG
*arg
)
1773 glsl_src_param_t src0_param
;
1774 glsl_src_param_t src1_param
;
1775 glsl_src_param_t src2_param
;
1776 DWORD write_mask
, cmp_channel
= 0;
1779 BOOL temp_destination
= FALSE
;
1781 if(shader_is_scalar(arg
->src
[0])) {
1782 write_mask
= shader_glsl_append_dst(arg
->buffer
, arg
);
1784 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_ALL
, &src0_param
);
1785 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], write_mask
, &src1_param
);
1786 shader_glsl_add_src_param(arg
, arg
->src
[2], arg
->src_addr
[2], write_mask
, &src2_param
);
1788 shader_addline(arg
->buffer
, "%s >= 0.0 ? %s : %s);\n",
1789 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
1791 DWORD src0reg
= arg
->src
[0] & WINED3DSP_REGNUM_MASK
;
1792 DWORD src1reg
= arg
->src
[1] & WINED3DSP_REGNUM_MASK
;
1793 DWORD src2reg
= arg
->src
[2] & WINED3DSP_REGNUM_MASK
;
1794 DWORD src0regtype
= shader_get_regtype(arg
->src
[0]);
1795 DWORD src1regtype
= shader_get_regtype(arg
->src
[1]);
1796 DWORD src2regtype
= shader_get_regtype(arg
->src
[2]);
1797 DWORD dstreg
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
1798 DWORD dstregtype
= shader_get_regtype(arg
->dst
);
1800 /* Cycle through all source0 channels */
1801 for (i
=0; i
<4; i
++) {
1803 /* Find the destination channels which use the current source0 channel */
1804 for (j
=0; j
<4; j
++) {
1805 if ( ((arg
->src
[0] >> (WINED3DSP_SWIZZLE_SHIFT
+ 2*j
)) & 0x3) == i
) {
1806 write_mask
|= WINED3DSP_WRITEMASK_0
<< j
;
1807 cmp_channel
= WINED3DSP_WRITEMASK_0
<< j
;
1811 /* Splitting the cmp instruction up in multiple lines imposes a problem:
1812 * The first lines may overwrite source parameters of the following lines.
1813 * Deal with that by using a temporary destination register if needed
1815 if((src0reg
== dstreg
&& src0regtype
== dstregtype
) ||
1816 (src1reg
== dstreg
&& src1regtype
== dstregtype
) ||
1817 (src2reg
== dstreg
&& src2regtype
== dstregtype
)) {
1819 write_mask
= shader_glsl_get_write_mask(arg
->dst
& (~WINED3DSP_SWIZZLE_MASK
| write_mask
), mask_char
);
1820 if (!write_mask
) continue;
1821 shader_addline(arg
->buffer
, "tmp0%s = (", mask_char
);
1822 temp_destination
= TRUE
;
1824 write_mask
= shader_glsl_append_dst_ext(arg
->buffer
, arg
, arg
->dst
& (~WINED3DSP_SWIZZLE_MASK
| write_mask
));
1825 if (!write_mask
) continue;
1828 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], cmp_channel
, &src0_param
);
1829 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], write_mask
, &src1_param
);
1830 shader_glsl_add_src_param(arg
, arg
->src
[2], arg
->src_addr
[2], write_mask
, &src2_param
);
1832 shader_addline(arg
->buffer
, "%s >= 0.0 ? %s : %s);\n",
1833 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
1836 if(temp_destination
) {
1837 shader_glsl_get_write_mask(arg
->dst
, mask_char
);
1838 shader_glsl_append_dst_ext(arg
->buffer
, arg
, arg
->dst
);
1839 shader_addline(arg
->buffer
, "tmp0%s);\n", mask_char
);
1845 /** Process the CND opcode in GLSL (dst = (src0 > 0.5) ? src1 : src2) */
1846 /* For ps 1.1-1.3, only a single component of src0 is used. For ps 1.4
1847 * the compare is done per component of src0. */
1848 static void shader_glsl_cnd(const SHADER_OPCODE_ARG
*arg
)
1850 glsl_src_param_t src0_param
;
1851 glsl_src_param_t src1_param
;
1852 glsl_src_param_t src2_param
;
1853 DWORD write_mask
, cmp_channel
= 0;
1856 if (arg
->reg_maps
->shader_version
< WINED3DPS_VERSION(1, 4))
1858 write_mask
= shader_glsl_append_dst(arg
->buffer
, arg
);
1859 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
1860 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], write_mask
, &src1_param
);
1861 shader_glsl_add_src_param(arg
, arg
->src
[2], arg
->src_addr
[2], write_mask
, &src2_param
);
1863 /* Fun: The D3DSI_COISSUE flag changes the semantic of the cnd instruction for < 1.4 shaders */
1864 if(arg
->opcode_token
& WINED3DSI_COISSUE
) {
1865 shader_addline(arg
->buffer
, "%s /* COISSUE! */);\n", src1_param
.param_str
);
1867 shader_addline(arg
->buffer
, "%s > 0.5 ? %s : %s);\n",
1868 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
1872 /* Cycle through all source0 channels */
1873 for (i
=0; i
<4; i
++) {
1875 /* Find the destination channels which use the current source0 channel */
1876 for (j
=0; j
<4; j
++) {
1877 if ( ((arg
->src
[0] >> (WINED3DSP_SWIZZLE_SHIFT
+ 2*j
)) & 0x3) == i
) {
1878 write_mask
|= WINED3DSP_WRITEMASK_0
<< j
;
1879 cmp_channel
= WINED3DSP_WRITEMASK_0
<< j
;
1882 write_mask
= shader_glsl_append_dst_ext(arg
->buffer
, arg
, arg
->dst
& (~WINED3DSP_SWIZZLE_MASK
| write_mask
));
1883 if (!write_mask
) continue;
1885 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], cmp_channel
, &src0_param
);
1886 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], write_mask
, &src1_param
);
1887 shader_glsl_add_src_param(arg
, arg
->src
[2], arg
->src_addr
[2], write_mask
, &src2_param
);
1889 shader_addline(arg
->buffer
, "%s > 0.5 ? %s : %s);\n",
1890 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
1894 /** GLSL code generation for WINED3DSIO_MAD: Multiply the first 2 opcodes, then add the last */
1895 static void shader_glsl_mad(const SHADER_OPCODE_ARG
*arg
)
1897 glsl_src_param_t src0_param
;
1898 glsl_src_param_t src1_param
;
1899 glsl_src_param_t src2_param
;
1902 write_mask
= shader_glsl_append_dst(arg
->buffer
, arg
);
1903 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], write_mask
, &src0_param
);
1904 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], write_mask
, &src1_param
);
1905 shader_glsl_add_src_param(arg
, arg
->src
[2], arg
->src_addr
[2], write_mask
, &src2_param
);
1906 shader_addline(arg
->buffer
, "(%s * %s) + %s);\n",
1907 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
1910 /** Handles transforming all WINED3DSIO_M?x? opcodes for
1911 Vertex shaders to GLSL codes */
1912 static void shader_glsl_mnxn(const SHADER_OPCODE_ARG
*arg
)
1914 IWineD3DBaseShaderImpl
*shader
= (IWineD3DBaseShaderImpl
*)arg
->shader
;
1915 const SHADER_OPCODE
*opcode_table
= shader
->baseShader
.shader_ins
;
1916 DWORD shader_version
= arg
->reg_maps
->shader_version
;
1918 int nComponents
= 0;
1919 SHADER_OPCODE_ARG tmpArg
;
1921 memset(&tmpArg
, 0, sizeof(SHADER_OPCODE_ARG
));
1923 /* Set constants for the temporary argument */
1924 tmpArg
.shader
= arg
->shader
;
1925 tmpArg
.buffer
= arg
->buffer
;
1926 tmpArg
.src
[0] = arg
->src
[0];
1927 tmpArg
.src_addr
[0] = arg
->src_addr
[0];
1928 tmpArg
.src_addr
[1] = arg
->src_addr
[1];
1929 tmpArg
.reg_maps
= arg
->reg_maps
;
1931 switch(arg
->opcode
->opcode
) {
1932 case WINED3DSIO_M4x4
:
1934 tmpArg
.opcode
= shader_get_opcode(opcode_table
, shader_version
, WINED3DSIO_DP4
);
1936 case WINED3DSIO_M4x3
:
1938 tmpArg
.opcode
= shader_get_opcode(opcode_table
, shader_version
, WINED3DSIO_DP4
);
1940 case WINED3DSIO_M3x4
:
1942 tmpArg
.opcode
= shader_get_opcode(opcode_table
, shader_version
, WINED3DSIO_DP3
);
1944 case WINED3DSIO_M3x3
:
1946 tmpArg
.opcode
= shader_get_opcode(opcode_table
, shader_version
, WINED3DSIO_DP3
);
1948 case WINED3DSIO_M3x2
:
1950 tmpArg
.opcode
= shader_get_opcode(opcode_table
, shader_version
, WINED3DSIO_DP3
);
1956 for (i
= 0; i
< nComponents
; i
++) {
1957 tmpArg
.dst
= ((arg
->dst
) & ~WINED3DSP_WRITEMASK_ALL
)|(WINED3DSP_WRITEMASK_0
<<i
);
1958 tmpArg
.src
[1] = arg
->src
[1]+i
;
1959 shader_glsl_dot(&tmpArg
);
1964 The LRP instruction performs a component-wise linear interpolation
1965 between the second and third operands using the first operand as the
1966 blend factor. Equation: (dst = src2 + src0 * (src1 - src2))
1967 This is equivalent to mix(src2, src1, src0);
1969 static void shader_glsl_lrp(const SHADER_OPCODE_ARG
*arg
)
1971 glsl_src_param_t src0_param
;
1972 glsl_src_param_t src1_param
;
1973 glsl_src_param_t src2_param
;
1976 write_mask
= shader_glsl_append_dst(arg
->buffer
, arg
);
1978 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], write_mask
, &src0_param
);
1979 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], write_mask
, &src1_param
);
1980 shader_glsl_add_src_param(arg
, arg
->src
[2], arg
->src_addr
[2], write_mask
, &src2_param
);
1982 shader_addline(arg
->buffer
, "mix(%s, %s, %s));\n",
1983 src2_param
.param_str
, src1_param
.param_str
, src0_param
.param_str
);
1986 /** Process the WINED3DSIO_LIT instruction in GLSL:
1987 * dst.x = dst.w = 1.0
1988 * dst.y = (src0.x > 0) ? src0.x
1989 * dst.z = (src0.x > 0) ? ((src0.y > 0) ? pow(src0.y, src.w) : 0) : 0
1990 * where src.w is clamped at +- 128
1992 static void shader_glsl_lit(const SHADER_OPCODE_ARG
*arg
)
1994 glsl_src_param_t src0_param
;
1995 glsl_src_param_t src1_param
;
1996 glsl_src_param_t src3_param
;
1999 shader_glsl_append_dst(arg
->buffer
, arg
);
2000 shader_glsl_get_write_mask(arg
->dst
, dst_mask
);
2002 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2003 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_1
, &src1_param
);
2004 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_3
, &src3_param
);
2006 /* The sdk specifies the instruction like this
2008 * if(src.x > 0.0) dst.y = src.x
2010 * if(src.x > 0.0 && src.y > 0.0) dst.z = pow(src.y, power);
2014 * Obviously that has quite a few conditionals in it which we don't like. So the first step is this:
2015 * dst.x = 1.0 ... No further explanation needed
2016 * dst.y = max(src.y, 0.0); ... If x < 0.0, use 0.0, otherwise x. Same as the conditional
2017 * dst.z = x > 0.0 ? pow(max(y, 0.0), p) : 0; ... 0 ^ power is 0, and otherwise we use y anyway
2018 * dst.w = 1.0. ... Nothing fancy.
2020 * So we still have one conditional in there. So do this:
2021 * dst.z = pow(max(0.0, src.y) * step(0.0, src.x), power);
2023 * step(0.0, x) will return 1 if src.x > 0.0, and 0 otherwise. So if y is 0 we get pow(0.0 * 1.0, power),
2024 * which sets dst.z to 0. If y > 0, but x = 0.0, we get pow(y * 0.0, power), which results in 0 too.
2025 * if both x and y are > 0, we get pow(y * 1.0, power), as it is supposed to
2027 shader_addline(arg
->buffer
, "vec4(1.0, max(%s, 0.0), pow(max(0.0, %s) * step(0.0, %s), clamp(%s, -128.0, 128.0)), 1.0)%s);\n",
2028 src0_param
.param_str
, src1_param
.param_str
, src0_param
.param_str
, src3_param
.param_str
, dst_mask
);
2031 /** Process the WINED3DSIO_DST instruction in GLSL:
2033 * dst.y = src0.x * src0.y
2037 static void shader_glsl_dst(const SHADER_OPCODE_ARG
*arg
)
2039 glsl_src_param_t src0y_param
;
2040 glsl_src_param_t src0z_param
;
2041 glsl_src_param_t src1y_param
;
2042 glsl_src_param_t src1w_param
;
2045 shader_glsl_append_dst(arg
->buffer
, arg
);
2046 shader_glsl_get_write_mask(arg
->dst
, dst_mask
);
2048 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_1
, &src0y_param
);
2049 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_2
, &src0z_param
);
2050 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], WINED3DSP_WRITEMASK_1
, &src1y_param
);
2051 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], WINED3DSP_WRITEMASK_3
, &src1w_param
);
2053 shader_addline(arg
->buffer
, "vec4(1.0, %s * %s, %s, %s))%s;\n",
2054 src0y_param
.param_str
, src1y_param
.param_str
, src0z_param
.param_str
, src1w_param
.param_str
, dst_mask
);
2057 /** Process the WINED3DSIO_SINCOS instruction in GLSL:
2058 * VS 2.0 requires that specific cosine and sine constants be passed to this instruction so the hardware
2059 * can handle it. But, these functions are built-in for GLSL, so we can just ignore the last 2 params.
2061 * dst.x = cos(src0.?)
2062 * dst.y = sin(src0.?)
2066 static void shader_glsl_sincos(const SHADER_OPCODE_ARG
*arg
)
2068 glsl_src_param_t src0_param
;
2071 write_mask
= shader_glsl_append_dst(arg
->buffer
, arg
);
2072 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2074 switch (write_mask
) {
2075 case WINED3DSP_WRITEMASK_0
:
2076 shader_addline(arg
->buffer
, "cos(%s));\n", src0_param
.param_str
);
2079 case WINED3DSP_WRITEMASK_1
:
2080 shader_addline(arg
->buffer
, "sin(%s));\n", src0_param
.param_str
);
2083 case (WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
):
2084 shader_addline(arg
->buffer
, "vec2(cos(%s), sin(%s)));\n", src0_param
.param_str
, src0_param
.param_str
);
2088 ERR("Write mask should be .x, .y or .xy\n");
2093 /** Process the WINED3DSIO_LOOP instruction in GLSL:
2094 * Start a for() loop where src1.y is the initial value of aL,
2095 * increment aL by src1.z for a total of src1.x iterations.
2096 * Need to use a temporary variable for this operation.
2098 /* FIXME: I don't think nested loops will work correctly this way. */
2099 static void shader_glsl_loop(const SHADER_OPCODE_ARG
*arg
)
2101 glsl_src_param_t src1_param
;
2102 IWineD3DBaseShaderImpl
* shader
= (IWineD3DBaseShaderImpl
*) arg
->shader
;
2103 DWORD regtype
= shader_get_regtype(arg
->src
[1]);
2104 DWORD reg
= arg
->src
[1] & WINED3DSP_REGNUM_MASK
;
2105 const DWORD
*control_values
= NULL
;
2106 const local_constant
*constant
;
2108 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], WINED3DSP_WRITEMASK_ALL
, &src1_param
);
2110 /* Try to hardcode the loop control parameters if possible. Direct3D 9 class hardware doesn't support real
2111 * varying indexing, but Microsoft designed this feature for Shader model 2.x+. If the loop control is
2112 * known at compile time, the GLSL compiler can unroll the loop, and replace indirect addressing with direct
2115 if(regtype
== WINED3DSPR_CONSTINT
) {
2116 LIST_FOR_EACH_ENTRY(constant
, &shader
->baseShader
.constantsI
, local_constant
, entry
) {
2117 if(constant
->idx
== reg
) {
2118 control_values
= constant
->value
;
2124 if(control_values
) {
2125 if(control_values
[2] > 0) {
2126 shader_addline(arg
->buffer
, "for (aL%u = %d; aL%u < (%d * %d + %d); aL%u += %d) {\n",
2127 shader
->baseShader
.cur_loop_depth
, control_values
[1],
2128 shader
->baseShader
.cur_loop_depth
, control_values
[0], control_values
[2], control_values
[1],
2129 shader
->baseShader
.cur_loop_depth
, control_values
[2]);
2130 } else if(control_values
[2] == 0) {
2131 shader_addline(arg
->buffer
, "for (aL%u = %d, tmpInt%u = 0; tmpInt%u < %d; tmpInt%u++) {\n",
2132 shader
->baseShader
.cur_loop_depth
, control_values
[1], shader
->baseShader
.cur_loop_depth
,
2133 shader
->baseShader
.cur_loop_depth
, control_values
[0],
2134 shader
->baseShader
.cur_loop_depth
);
2136 shader_addline(arg
->buffer
, "for (aL%u = %d; aL%u > (%d * %d + %d); aL%u += %d) {\n",
2137 shader
->baseShader
.cur_loop_depth
, control_values
[1],
2138 shader
->baseShader
.cur_loop_depth
, control_values
[0], control_values
[2], control_values
[1],
2139 shader
->baseShader
.cur_loop_depth
, control_values
[2]);
2142 shader_addline(arg
->buffer
, "for (tmpInt%u = 0, aL%u = %s.y; tmpInt%u < %s.x; tmpInt%u++, aL%u += %s.z) {\n",
2143 shader
->baseShader
.cur_loop_depth
, shader
->baseShader
.cur_loop_regno
,
2144 src1_param
.reg_name
, shader
->baseShader
.cur_loop_depth
, src1_param
.reg_name
,
2145 shader
->baseShader
.cur_loop_depth
, shader
->baseShader
.cur_loop_regno
, src1_param
.reg_name
);
2148 shader
->baseShader
.cur_loop_depth
++;
2149 shader
->baseShader
.cur_loop_regno
++;
2152 static void shader_glsl_end(const SHADER_OPCODE_ARG
*arg
)
2154 IWineD3DBaseShaderImpl
* shader
= (IWineD3DBaseShaderImpl
*) arg
->shader
;
2156 shader_addline(arg
->buffer
, "}\n");
2158 if(arg
->opcode
->opcode
== WINED3DSIO_ENDLOOP
) {
2159 shader
->baseShader
.cur_loop_depth
--;
2160 shader
->baseShader
.cur_loop_regno
--;
2162 if(arg
->opcode
->opcode
== WINED3DSIO_ENDREP
) {
2163 shader
->baseShader
.cur_loop_depth
--;
2167 static void shader_glsl_rep(const SHADER_OPCODE_ARG
*arg
)
2169 IWineD3DBaseShaderImpl
* shader
= (IWineD3DBaseShaderImpl
*) arg
->shader
;
2170 glsl_src_param_t src0_param
;
2172 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2173 shader_addline(arg
->buffer
, "for (tmpInt%d = 0; tmpInt%d < %s; tmpInt%d++) {\n",
2174 shader
->baseShader
.cur_loop_depth
, shader
->baseShader
.cur_loop_depth
,
2175 src0_param
.param_str
, shader
->baseShader
.cur_loop_depth
);
2176 shader
->baseShader
.cur_loop_depth
++;
2179 static void shader_glsl_if(const SHADER_OPCODE_ARG
*arg
)
2181 glsl_src_param_t src0_param
;
2183 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2184 shader_addline(arg
->buffer
, "if (%s) {\n", src0_param
.param_str
);
2187 static void shader_glsl_ifc(const SHADER_OPCODE_ARG
*arg
)
2189 glsl_src_param_t src0_param
;
2190 glsl_src_param_t src1_param
;
2192 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2193 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], WINED3DSP_WRITEMASK_0
, &src1_param
);
2195 shader_addline(arg
->buffer
, "if (%s %s %s) {\n",
2196 src0_param
.param_str
, shader_get_comp_op(arg
->opcode_token
), src1_param
.param_str
);
2199 static void shader_glsl_else(const SHADER_OPCODE_ARG
*arg
)
2201 shader_addline(arg
->buffer
, "} else {\n");
2204 static void shader_glsl_break(const SHADER_OPCODE_ARG
*arg
)
2206 shader_addline(arg
->buffer
, "break;\n");
2209 /* FIXME: According to MSDN the compare is done per component. */
2210 static void shader_glsl_breakc(const SHADER_OPCODE_ARG
*arg
)
2212 glsl_src_param_t src0_param
;
2213 glsl_src_param_t src1_param
;
2215 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2216 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], WINED3DSP_WRITEMASK_0
, &src1_param
);
2218 shader_addline(arg
->buffer
, "if (%s %s %s) break;\n",
2219 src0_param
.param_str
, shader_get_comp_op(arg
->opcode_token
), src1_param
.param_str
);
2222 static void shader_glsl_label(const SHADER_OPCODE_ARG
*arg
)
2225 DWORD snum
= (arg
->src
[0]) & WINED3DSP_REGNUM_MASK
;
2226 shader_addline(arg
->buffer
, "}\n");
2227 shader_addline(arg
->buffer
, "void subroutine%u () {\n", snum
);
2230 static void shader_glsl_call(const SHADER_OPCODE_ARG
*arg
)
2232 DWORD snum
= (arg
->src
[0]) & WINED3DSP_REGNUM_MASK
;
2233 shader_addline(arg
->buffer
, "subroutine%u();\n", snum
);
2236 static void shader_glsl_callnz(const SHADER_OPCODE_ARG
*arg
)
2238 glsl_src_param_t src1_param
;
2240 DWORD snum
= (arg
->src
[0]) & WINED3DSP_REGNUM_MASK
;
2241 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], WINED3DSP_WRITEMASK_0
, &src1_param
);
2242 shader_addline(arg
->buffer
, "if (%s) subroutine%u();\n", src1_param
.param_str
, snum
);
2245 /*********************************************
2246 * Pixel Shader Specific Code begins here
2247 ********************************************/
2248 static void pshader_glsl_tex(const SHADER_OPCODE_ARG
*arg
)
2250 IWineD3DPixelShaderImpl
* This
= (IWineD3DPixelShaderImpl
*) arg
->shader
;
2251 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
2252 DWORD shader_version
= arg
->reg_maps
->shader_version
;
2253 char dst_swizzle
[6];
2254 glsl_sample_function_t sample_function
;
2257 BOOL projected
, texrect
= FALSE
;
2260 /* All versions have a destination register */
2261 shader_glsl_append_dst(arg
->buffer
, arg
);
2263 /* 1.0-1.4: Use destination register as sampler source.
2264 * 2.0+: Use provided sampler source. */
2265 if (shader_version
< WINED3DPS_VERSION(2,0)) sampler_idx
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2266 else sampler_idx
= arg
->src
[1] & WINED3DSP_REGNUM_MASK
;
2267 sampler_type
= arg
->reg_maps
->samplers
[sampler_idx
] & WINED3DSP_TEXTURETYPE_MASK
;
2269 if (shader_version
< WINED3DPS_VERSION(1,4))
2271 DWORD flags
= deviceImpl
->stateBlock
->textureState
[sampler_idx
][WINED3DTSS_TEXTURETRANSFORMFLAGS
];
2273 /* Projected cube textures don't make a lot of sense, the resulting coordinates stay the same. */
2274 if (flags
& WINED3DTTFF_PROJECTED
&& sampler_type
!= WINED3DSTT_CUBE
) {
2276 switch (flags
& ~WINED3DTTFF_PROJECTED
) {
2277 case WINED3DTTFF_COUNT1
: FIXME("WINED3DTTFF_PROJECTED with WINED3DTTFF_COUNT1?\n"); break;
2278 case WINED3DTTFF_COUNT2
: mask
= WINED3DSP_WRITEMASK_1
; break;
2279 case WINED3DTTFF_COUNT3
: mask
= WINED3DSP_WRITEMASK_2
; break;
2280 case WINED3DTTFF_COUNT4
:
2281 case WINED3DTTFF_DISABLE
: mask
= WINED3DSP_WRITEMASK_3
; break;
2287 else if (shader_version
< WINED3DPS_VERSION(2,0))
2289 DWORD src_mod
= arg
->src
[0] & WINED3DSP_SRCMOD_MASK
;
2291 if (src_mod
== WINED3DSPSM_DZ
) {
2293 mask
= WINED3DSP_WRITEMASK_2
;
2294 } else if (src_mod
== WINED3DSPSM_DW
) {
2296 mask
= WINED3DSP_WRITEMASK_3
;
2301 if(arg
->opcode_token
& WINED3DSI_TEXLD_PROJECT
) {
2302 /* ps 2.0 texldp instruction always divides by the fourth component. */
2304 mask
= WINED3DSP_WRITEMASK_3
;
2310 if(deviceImpl
->stateBlock
->textures
[sampler_idx
] &&
2311 IWineD3DBaseTexture_GetTextureDimensions(deviceImpl
->stateBlock
->textures
[sampler_idx
]) == GL_TEXTURE_RECTANGLE_ARB
) {
2315 shader_glsl_get_sample_function(sampler_type
, projected
, texrect
, &sample_function
);
2316 mask
|= sample_function
.coord_mask
;
2318 if (shader_version
< WINED3DPS_VERSION(2,0)) shader_glsl_get_write_mask(arg
->dst
, dst_swizzle
);
2319 else shader_glsl_get_swizzle(arg
->src
[1], FALSE
, arg
->dst
, dst_swizzle
);
2321 /* 1.0-1.3: Use destination register as coordinate source.
2322 1.4+: Use provided coordinate source register. */
2323 if (shader_version
< WINED3DPS_VERSION(1,4))
2326 shader_glsl_get_write_mask(mask
, coord_mask
);
2327 shader_addline(arg
->buffer
, "%s(Psampler%u, T%u%s)%s);\n",
2328 sample_function
.name
, sampler_idx
, sampler_idx
, coord_mask
, dst_swizzle
);
2330 glsl_src_param_t coord_param
;
2331 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], mask
, &coord_param
);
2332 if(arg
->opcode_token
& WINED3DSI_TEXLD_BIAS
) {
2333 glsl_src_param_t bias
;
2334 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_3
, &bias
);
2336 shader_addline(arg
->buffer
, "%s(Psampler%u, %s, %s)%s);\n",
2337 sample_function
.name
, sampler_idx
, coord_param
.param_str
,
2338 bias
.param_str
, dst_swizzle
);
2340 shader_addline(arg
->buffer
, "%s(Psampler%u, %s)%s);\n",
2341 sample_function
.name
, sampler_idx
, coord_param
.param_str
, dst_swizzle
);
2346 static void shader_glsl_texldl(const SHADER_OPCODE_ARG
*arg
)
2348 IWineD3DBaseShaderImpl
* This
= (IWineD3DBaseShaderImpl
*)arg
->shader
;
2349 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
2350 glsl_sample_function_t sample_function
;
2351 glsl_src_param_t coord_param
, lod_param
;
2352 char dst_swizzle
[6];
2355 BOOL texrect
= FALSE
;
2357 shader_glsl_append_dst(arg
->buffer
, arg
);
2358 shader_glsl_get_swizzle(arg
->src
[1], FALSE
, arg
->dst
, dst_swizzle
);
2360 sampler_idx
= arg
->src
[1] & WINED3DSP_REGNUM_MASK
;
2361 sampler_type
= arg
->reg_maps
->samplers
[sampler_idx
] & WINED3DSP_TEXTURETYPE_MASK
;
2362 if(deviceImpl
->stateBlock
->textures
[sampler_idx
] &&
2363 IWineD3DBaseTexture_GetTextureDimensions(deviceImpl
->stateBlock
->textures
[sampler_idx
]) == GL_TEXTURE_RECTANGLE_ARB
) {
2366 shader_glsl_get_sample_function(sampler_type
, FALSE
, texrect
, &sample_function
); shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], sample_function
.coord_mask
, &coord_param
);
2368 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_3
, &lod_param
);
2370 if (shader_is_pshader_version(arg
->reg_maps
->shader_version
))
2372 /* The GLSL spec claims the Lod sampling functions are only supported in vertex shaders.
2373 * However, they seem to work just fine in fragment shaders as well. */
2374 WARN("Using %sLod in fragment shader.\n", sample_function
.name
);
2375 shader_addline(arg
->buffer
, "%sLod(Psampler%u, %s, %s)%s);\n",
2376 sample_function
.name
, sampler_idx
, coord_param
.param_str
, lod_param
.param_str
, dst_swizzle
);
2378 shader_addline(arg
->buffer
, "%sLod(Vsampler%u, %s, %s)%s);\n",
2379 sample_function
.name
, sampler_idx
, coord_param
.param_str
, lod_param
.param_str
, dst_swizzle
);
2383 static void pshader_glsl_texcoord(const SHADER_OPCODE_ARG
*arg
)
2385 /* FIXME: Make this work for more than just 2D textures */
2386 SHADER_BUFFER
* buffer
= arg
->buffer
;
2390 write_mask
= shader_glsl_append_dst(arg
->buffer
, arg
);
2391 shader_glsl_get_write_mask(write_mask
, dst_mask
);
2393 if (arg
->reg_maps
->shader_version
!= WINED3DPS_VERSION(1,4))
2395 DWORD reg
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2396 shader_addline(buffer
, "clamp(gl_TexCoord[%u], 0.0, 1.0)%s);\n", reg
, dst_mask
);
2398 DWORD reg
= arg
->src
[0] & WINED3DSP_REGNUM_MASK
;
2399 DWORD src_mod
= arg
->src
[0] & WINED3DSP_SRCMOD_MASK
;
2400 char dst_swizzle
[6];
2402 shader_glsl_get_swizzle(arg
->src
[0], FALSE
, write_mask
, dst_swizzle
);
2404 if (src_mod
== WINED3DSPSM_DZ
) {
2405 glsl_src_param_t div_param
;
2406 unsigned int mask_size
= shader_glsl_get_write_mask_size(write_mask
);
2407 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_2
, &div_param
);
2409 if (mask_size
> 1) {
2410 shader_addline(buffer
, "gl_TexCoord[%u]%s / vec%d(%s));\n", reg
, dst_swizzle
, mask_size
, div_param
.param_str
);
2412 shader_addline(buffer
, "gl_TexCoord[%u]%s / %s);\n", reg
, dst_swizzle
, div_param
.param_str
);
2414 } else if (src_mod
== WINED3DSPSM_DW
) {
2415 glsl_src_param_t div_param
;
2416 unsigned int mask_size
= shader_glsl_get_write_mask_size(write_mask
);
2417 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_3
, &div_param
);
2419 if (mask_size
> 1) {
2420 shader_addline(buffer
, "gl_TexCoord[%u]%s / vec%d(%s));\n", reg
, dst_swizzle
, mask_size
, div_param
.param_str
);
2422 shader_addline(buffer
, "gl_TexCoord[%u]%s / %s);\n", reg
, dst_swizzle
, div_param
.param_str
);
2425 shader_addline(buffer
, "gl_TexCoord[%u]%s);\n", reg
, dst_swizzle
);
2430 /** Process the WINED3DSIO_TEXDP3TEX instruction in GLSL:
2431 * Take a 3-component dot product of the TexCoord[dstreg] and src,
2432 * then perform a 1D texture lookup from stage dstregnum, place into dst. */
2433 static void pshader_glsl_texdp3tex(const SHADER_OPCODE_ARG
*arg
)
2435 glsl_src_param_t src0_param
;
2437 glsl_sample_function_t sample_function
;
2438 DWORD sampler_idx
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2439 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2440 DWORD sampler_type
= arg
->reg_maps
->samplers
[sampler_idx
] & WINED3DSP_TEXTURETYPE_MASK
;
2442 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], src_mask
, &src0_param
);
2444 shader_glsl_append_dst(arg
->buffer
, arg
);
2445 shader_glsl_get_write_mask(arg
->dst
, dst_mask
);
2447 /* Do I have to take care about the projected bit? I don't think so, since the dp3 returns only one
2448 * scalar, and projected sampling would require 4.
2450 * It is a dependent read - not valid with conditional NP2 textures
2452 shader_glsl_get_sample_function(sampler_type
, FALSE
, FALSE
, &sample_function
);
2454 switch(count_bits(sample_function
.coord_mask
)) {
2456 shader_addline(arg
->buffer
, "%s(Psampler%u, dot(gl_TexCoord[%u].xyz, %s))%s);\n",
2457 sample_function
.name
, sampler_idx
, sampler_idx
, src0_param
.param_str
, dst_mask
);
2461 shader_addline(arg
->buffer
, "%s(Psampler%u, vec2(dot(gl_TexCoord[%u].xyz, %s), 0.0))%s);\n",
2462 sample_function
.name
, sampler_idx
, sampler_idx
, src0_param
.param_str
, dst_mask
);
2466 shader_addline(arg
->buffer
, "%s(Psampler%u, vec3(dot(gl_TexCoord[%u].xyz, %s), 0.0, 0.0))%s);\n",
2467 sample_function
.name
, sampler_idx
, sampler_idx
, src0_param
.param_str
, dst_mask
);
2470 FIXME("Unexpected mask bitcount %d\n", count_bits(sample_function
.coord_mask
));
2474 /** Process the WINED3DSIO_TEXDP3 instruction in GLSL:
2475 * Take a 3-component dot product of the TexCoord[dstreg] and src. */
2476 static void pshader_glsl_texdp3(const SHADER_OPCODE_ARG
*arg
)
2478 glsl_src_param_t src0_param
;
2479 DWORD dstreg
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2480 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2482 unsigned int mask_size
;
2484 dst_mask
= shader_glsl_append_dst(arg
->buffer
, arg
);
2485 mask_size
= shader_glsl_get_write_mask_size(dst_mask
);
2486 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], src_mask
, &src0_param
);
2488 if (mask_size
> 1) {
2489 shader_addline(arg
->buffer
, "vec%d(dot(T%u.xyz, %s)));\n", mask_size
, dstreg
, src0_param
.param_str
);
2491 shader_addline(arg
->buffer
, "dot(T%u.xyz, %s));\n", dstreg
, src0_param
.param_str
);
2495 /** Process the WINED3DSIO_TEXDEPTH instruction in GLSL:
2496 * Calculate the depth as dst.x / dst.y */
2497 static void pshader_glsl_texdepth(const SHADER_OPCODE_ARG
*arg
)
2499 glsl_dst_param_t dst_param
;
2501 shader_glsl_add_dst_param(arg
, arg
->dst
, 0, &dst_param
);
2503 /* Tests show that texdepth never returns anything below 0.0, and that r5.y is clamped to 1.0.
2504 * Negative input is accepted, -0.25 / -0.5 returns 0.5. GL should clamp gl_FragDepth to [0;1], but
2505 * this doesn't always work, so clamp the results manually. Whether or not the x value is clamped at 1
2506 * too is irrelevant, since if x = 0, any y value < 1.0 (and > 1.0 is not allowed) results in a result
2509 shader_addline(arg
->buffer
, "gl_FragDepth = clamp((%s.x / min(%s.y, 1.0)), 0.0, 1.0);\n", dst_param
.reg_name
, dst_param
.reg_name
);
2512 /** Process the WINED3DSIO_TEXM3X2DEPTH instruction in GLSL:
2513 * Last row of a 3x2 matrix multiply, use the result to calculate the depth:
2514 * Calculate tmp0.y = TexCoord[dstreg] . src.xyz; (tmp0.x has already been calculated)
2515 * depth = (tmp0.y == 0.0) ? 1.0 : tmp0.x / tmp0.y
2517 static void pshader_glsl_texm3x2depth(const SHADER_OPCODE_ARG
*arg
)
2519 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2520 DWORD dstreg
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2521 glsl_src_param_t src0_param
;
2523 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], src_mask
, &src0_param
);
2525 shader_addline(arg
->buffer
, "tmp0.y = dot(T%u.xyz, %s);\n", dstreg
, src0_param
.param_str
);
2526 shader_addline(arg
->buffer
, "gl_FragDepth = (tmp0.y == 0.0) ? 1.0 : clamp(tmp0.x / tmp0.y, 0.0, 1.0);\n");
2529 /** Process the WINED3DSIO_TEXM3X2PAD instruction in GLSL
2530 * Calculate the 1st of a 2-row matrix multiplication. */
2531 static void pshader_glsl_texm3x2pad(const SHADER_OPCODE_ARG
*arg
)
2533 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2534 DWORD reg
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2535 SHADER_BUFFER
* buffer
= arg
->buffer
;
2536 glsl_src_param_t src0_param
;
2538 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], src_mask
, &src0_param
);
2539 shader_addline(buffer
, "tmp0.x = dot(T%u.xyz, %s);\n", reg
, src0_param
.param_str
);
2542 /** Process the WINED3DSIO_TEXM3X3PAD instruction in GLSL
2543 * Calculate the 1st or 2nd row of a 3-row matrix multiplication. */
2544 static void pshader_glsl_texm3x3pad(const SHADER_OPCODE_ARG
* arg
)
2546 IWineD3DPixelShaderImpl
* shader
= (IWineD3DPixelShaderImpl
*) arg
->shader
;
2547 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2548 DWORD reg
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2549 SHADER_BUFFER
* buffer
= arg
->buffer
;
2550 SHADER_PARSE_STATE
* current_state
= &shader
->baseShader
.parse_state
;
2551 glsl_src_param_t src0_param
;
2553 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], src_mask
, &src0_param
);
2554 shader_addline(buffer
, "tmp0.%c = dot(T%u.xyz, %s);\n", 'x' + current_state
->current_row
, reg
, src0_param
.param_str
);
2555 current_state
->texcoord_w
[current_state
->current_row
++] = reg
;
2558 static void pshader_glsl_texm3x2tex(const SHADER_OPCODE_ARG
*arg
)
2560 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2561 DWORD reg
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2562 SHADER_BUFFER
* buffer
= arg
->buffer
;
2563 glsl_src_param_t src0_param
;
2566 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], src_mask
, &src0_param
);
2567 shader_addline(buffer
, "tmp0.y = dot(T%u.xyz, %s);\n", reg
, src0_param
.param_str
);
2569 shader_glsl_append_dst(buffer
, arg
);
2570 shader_glsl_get_write_mask(arg
->dst
, dst_mask
);
2572 /* Sample the texture using the calculated coordinates */
2573 shader_addline(buffer
, "texture2D(Psampler%u, tmp0.xy)%s);\n", reg
, dst_mask
);
2576 /** Process the WINED3DSIO_TEXM3X3TEX instruction in GLSL
2577 * Perform the 3rd row of a 3x3 matrix multiply, then sample the texture using the calculated coordinates */
2578 static void pshader_glsl_texm3x3tex(const SHADER_OPCODE_ARG
*arg
)
2580 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2581 glsl_src_param_t src0_param
;
2583 DWORD reg
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2584 IWineD3DPixelShaderImpl
* This
= (IWineD3DPixelShaderImpl
*) arg
->shader
;
2585 SHADER_PARSE_STATE
* current_state
= &This
->baseShader
.parse_state
;
2586 DWORD sampler_type
= arg
->reg_maps
->samplers
[reg
] & WINED3DSP_TEXTURETYPE_MASK
;
2587 glsl_sample_function_t sample_function
;
2589 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], src_mask
, &src0_param
);
2590 shader_addline(arg
->buffer
, "tmp0.z = dot(T%u.xyz, %s);\n", reg
, src0_param
.param_str
);
2592 shader_glsl_append_dst(arg
->buffer
, arg
);
2593 shader_glsl_get_write_mask(arg
->dst
, dst_mask
);
2594 /* Dependent read, not valid with conditional NP2 */
2595 shader_glsl_get_sample_function(sampler_type
, FALSE
, FALSE
, &sample_function
);
2597 /* Sample the texture using the calculated coordinates */
2598 shader_addline(arg
->buffer
, "%s(Psampler%u, tmp0.xyz)%s);\n", sample_function
.name
, reg
, dst_mask
);
2600 current_state
->current_row
= 0;
2603 /** Process the WINED3DSIO_TEXM3X3 instruction in GLSL
2604 * Perform the 3rd row of a 3x3 matrix multiply */
2605 static void pshader_glsl_texm3x3(const SHADER_OPCODE_ARG
*arg
)
2607 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2608 glsl_src_param_t src0_param
;
2610 DWORD reg
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2611 IWineD3DPixelShaderImpl
* This
= (IWineD3DPixelShaderImpl
*) arg
->shader
;
2612 SHADER_PARSE_STATE
* current_state
= &This
->baseShader
.parse_state
;
2614 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], src_mask
, &src0_param
);
2616 shader_glsl_append_dst(arg
->buffer
, arg
);
2617 shader_glsl_get_write_mask(arg
->dst
, dst_mask
);
2618 shader_addline(arg
->buffer
, "vec4(tmp0.xy, dot(T%u.xyz, %s), 1.0)%s);\n", reg
, src0_param
.param_str
, dst_mask
);
2620 current_state
->current_row
= 0;
2623 /** Process the WINED3DSIO_TEXM3X3SPEC instruction in GLSL
2624 * Perform the final texture lookup based on the previous 2 3x3 matrix multiplies */
2625 static void pshader_glsl_texm3x3spec(const SHADER_OPCODE_ARG
*arg
)
2627 IWineD3DPixelShaderImpl
* shader
= (IWineD3DPixelShaderImpl
*) arg
->shader
;
2628 DWORD reg
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2629 glsl_src_param_t src0_param
;
2630 glsl_src_param_t src1_param
;
2632 SHADER_BUFFER
* buffer
= arg
->buffer
;
2633 SHADER_PARSE_STATE
* current_state
= &shader
->baseShader
.parse_state
;
2634 DWORD stype
= arg
->reg_maps
->samplers
[reg
] & WINED3DSP_TEXTURETYPE_MASK
;
2635 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2636 glsl_sample_function_t sample_function
;
2638 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], src_mask
, &src0_param
);
2639 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], src_mask
, &src1_param
);
2641 /* Perform the last matrix multiply operation */
2642 shader_addline(buffer
, "tmp0.z = dot(T%u.xyz, %s);\n", reg
, src0_param
.param_str
);
2643 /* Reflection calculation */
2644 shader_addline(buffer
, "tmp0.xyz = -reflect((%s), normalize(tmp0.xyz));\n", src1_param
.param_str
);
2646 shader_glsl_append_dst(buffer
, arg
);
2647 shader_glsl_get_write_mask(arg
->dst
, dst_mask
);
2648 /* Dependent read, not valid with conditional NP2 */
2649 shader_glsl_get_sample_function(stype
, FALSE
, FALSE
, &sample_function
);
2651 /* Sample the texture */
2652 shader_addline(buffer
, "%s(Psampler%u, tmp0.xyz)%s);\n", sample_function
.name
, reg
, dst_mask
);
2654 current_state
->current_row
= 0;
2657 /** Process the WINED3DSIO_TEXM3X3VSPEC instruction in GLSL
2658 * Perform the final texture lookup based on the previous 2 3x3 matrix multiplies */
2659 static void pshader_glsl_texm3x3vspec(const SHADER_OPCODE_ARG
*arg
)
2661 IWineD3DPixelShaderImpl
* shader
= (IWineD3DPixelShaderImpl
*) arg
->shader
;
2662 DWORD reg
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2663 SHADER_BUFFER
* buffer
= arg
->buffer
;
2664 SHADER_PARSE_STATE
* current_state
= &shader
->baseShader
.parse_state
;
2665 glsl_src_param_t src0_param
;
2667 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2668 DWORD sampler_type
= arg
->reg_maps
->samplers
[reg
] & WINED3DSP_TEXTURETYPE_MASK
;
2669 glsl_sample_function_t sample_function
;
2671 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], src_mask
, &src0_param
);
2673 /* Perform the last matrix multiply operation */
2674 shader_addline(buffer
, "tmp0.z = dot(vec3(T%u), vec3(%s));\n", reg
, src0_param
.param_str
);
2676 /* Construct the eye-ray vector from w coordinates */
2677 shader_addline(buffer
, "tmp1.xyz = normalize(vec3(gl_TexCoord[%u].w, gl_TexCoord[%u].w, gl_TexCoord[%u].w));\n",
2678 current_state
->texcoord_w
[0], current_state
->texcoord_w
[1], reg
);
2679 shader_addline(buffer
, "tmp0.xyz = -reflect(tmp1.xyz, normalize(tmp0.xyz));\n");
2681 shader_glsl_append_dst(buffer
, arg
);
2682 shader_glsl_get_write_mask(arg
->dst
, dst_mask
);
2683 /* Dependent read, not valid with conditional NP2 */
2684 shader_glsl_get_sample_function(sampler_type
, FALSE
, FALSE
, &sample_function
);
2686 /* Sample the texture using the calculated coordinates */
2687 shader_addline(buffer
, "%s(Psampler%u, tmp0.xyz)%s);\n", sample_function
.name
, reg
, dst_mask
);
2689 current_state
->current_row
= 0;
2692 /** Process the WINED3DSIO_TEXBEM instruction in GLSL.
2693 * Apply a fake bump map transform.
2694 * texbem is pshader <= 1.3 only, this saves a few version checks
2696 static void pshader_glsl_texbem(const SHADER_OPCODE_ARG
*arg
)
2698 IWineD3DPixelShaderImpl
* This
= (IWineD3DPixelShaderImpl
*) arg
->shader
;
2699 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
2700 char dst_swizzle
[6];
2701 glsl_sample_function_t sample_function
;
2702 glsl_src_param_t coord_param
;
2709 sampler_idx
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2710 flags
= deviceImpl
->stateBlock
->textureState
[sampler_idx
][WINED3DTSS_TEXTURETRANSFORMFLAGS
];
2712 sampler_type
= arg
->reg_maps
->samplers
[sampler_idx
] & WINED3DSP_TEXTURETYPE_MASK
;
2713 /* Dependent read, not valid with conditional NP2 */
2714 shader_glsl_get_sample_function(sampler_type
, FALSE
, FALSE
, &sample_function
);
2715 mask
= sample_function
.coord_mask
;
2717 shader_glsl_get_write_mask(arg
->dst
, dst_swizzle
);
2719 shader_glsl_get_write_mask(mask
, coord_mask
);
2721 /* with projective textures, texbem only divides the static texture coord, not the displacement,
2722 * so we can't let the GL handle this.
2724 if (flags
& WINED3DTTFF_PROJECTED
) {
2726 char coord_div_mask
[3];
2727 switch (flags
& ~WINED3DTTFF_PROJECTED
) {
2728 case WINED3DTTFF_COUNT1
: FIXME("WINED3DTTFF_PROJECTED with WINED3DTTFF_COUNT1?\n"); break;
2729 case WINED3DTTFF_COUNT2
: div_mask
= WINED3DSP_WRITEMASK_1
; break;
2730 case WINED3DTTFF_COUNT3
: div_mask
= WINED3DSP_WRITEMASK_2
; break;
2731 case WINED3DTTFF_COUNT4
:
2732 case WINED3DTTFF_DISABLE
: div_mask
= WINED3DSP_WRITEMASK_3
; break;
2734 shader_glsl_get_write_mask(div_mask
, coord_div_mask
);
2735 shader_addline(arg
->buffer
, "T%u%s /= T%u%s;\n", sampler_idx
, coord_mask
, sampler_idx
, coord_div_mask
);
2738 shader_glsl_append_dst(arg
->buffer
, arg
);
2739 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_0
|WINED3DSP_WRITEMASK_1
, &coord_param
);
2740 if(arg
->opcode
->opcode
== WINED3DSIO_TEXBEML
) {
2741 glsl_src_param_t luminance_param
;
2742 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_2
, &luminance_param
);
2743 shader_addline(arg
->buffer
, "(%s(Psampler%u, T%u%s + vec4(bumpenvmat%d * %s, 0.0, 0.0)%s )*(%s * luminancescale%d + luminanceoffset%d))%s);\n",
2744 sample_function
.name
, sampler_idx
, sampler_idx
, coord_mask
, sampler_idx
, coord_param
.param_str
, coord_mask
,
2745 luminance_param
.param_str
, sampler_idx
, sampler_idx
, dst_swizzle
);
2747 shader_addline(arg
->buffer
, "%s(Psampler%u, T%u%s + vec4(bumpenvmat%d * %s, 0.0, 0.0)%s )%s);\n",
2748 sample_function
.name
, sampler_idx
, sampler_idx
, coord_mask
, sampler_idx
, coord_param
.param_str
, coord_mask
, dst_swizzle
);
2752 static void pshader_glsl_bem(const SHADER_OPCODE_ARG
*arg
)
2754 glsl_src_param_t src0_param
, src1_param
;
2755 DWORD sampler_idx
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2757 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_0
|WINED3DSP_WRITEMASK_1
, &src0_param
);
2758 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], WINED3DSP_WRITEMASK_0
|WINED3DSP_WRITEMASK_1
, &src1_param
);
2760 shader_glsl_append_dst(arg
->buffer
, arg
);
2761 shader_addline(arg
->buffer
, "%s + bumpenvmat%d * %s);\n",
2762 src0_param
.param_str
, sampler_idx
, src1_param
.param_str
);
2765 /** Process the WINED3DSIO_TEXREG2AR instruction in GLSL
2766 * Sample 2D texture at dst using the alpha & red (wx) components of src as texture coordinates */
2767 static void pshader_glsl_texreg2ar(const SHADER_OPCODE_ARG
*arg
)
2769 glsl_src_param_t src0_param
;
2770 DWORD sampler_idx
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2773 shader_glsl_append_dst(arg
->buffer
, arg
);
2774 shader_glsl_get_write_mask(arg
->dst
, dst_mask
);
2775 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_ALL
, &src0_param
);
2777 shader_addline(arg
->buffer
, "texture2D(Psampler%u, %s.wx)%s);\n", sampler_idx
, src0_param
.reg_name
, dst_mask
);
2780 /** Process the WINED3DSIO_TEXREG2GB instruction in GLSL
2781 * Sample 2D texture at dst using the green & blue (yz) components of src as texture coordinates */
2782 static void pshader_glsl_texreg2gb(const SHADER_OPCODE_ARG
*arg
)
2784 glsl_src_param_t src0_param
;
2785 DWORD sampler_idx
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2788 shader_glsl_append_dst(arg
->buffer
, arg
);
2789 shader_glsl_get_write_mask(arg
->dst
, dst_mask
);
2790 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_ALL
, &src0_param
);
2792 shader_addline(arg
->buffer
, "texture2D(Psampler%u, %s.yz)%s);\n", sampler_idx
, src0_param
.reg_name
, dst_mask
);
2795 /** Process the WINED3DSIO_TEXREG2RGB instruction in GLSL
2796 * Sample texture at dst using the rgb (xyz) components of src as texture coordinates */
2797 static void pshader_glsl_texreg2rgb(const SHADER_OPCODE_ARG
*arg
)
2799 glsl_src_param_t src0_param
;
2801 DWORD sampler_idx
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2802 DWORD sampler_type
= arg
->reg_maps
->samplers
[sampler_idx
] & WINED3DSP_TEXTURETYPE_MASK
;
2803 glsl_sample_function_t sample_function
;
2805 shader_glsl_append_dst(arg
->buffer
, arg
);
2806 shader_glsl_get_write_mask(arg
->dst
, dst_mask
);
2807 /* Dependent read, not valid with conditional NP2 */
2808 shader_glsl_get_sample_function(sampler_type
, FALSE
, FALSE
, &sample_function
);
2809 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], sample_function
.coord_mask
, &src0_param
);
2811 shader_addline(arg
->buffer
, "%s(Psampler%u, %s)%s);\n", sample_function
.name
, sampler_idx
, src0_param
.param_str
, dst_mask
);
2814 /** Process the WINED3DSIO_TEXKILL instruction in GLSL.
2815 * If any of the first 3 components are < 0, discard this pixel */
2816 static void pshader_glsl_texkill(const SHADER_OPCODE_ARG
*arg
)
2818 glsl_dst_param_t dst_param
;
2820 /* The argument is a destination parameter, and no writemasks are allowed */
2821 shader_glsl_add_dst_param(arg
, arg
->dst
, 0, &dst_param
);
2822 if ((arg
->reg_maps
->shader_version
>= WINED3DPS_VERSION(2,0)))
2824 /* 2.0 shaders compare all 4 components in texkill */
2825 shader_addline(arg
->buffer
, "if (any(lessThan(%s.xyzw, vec4(0.0)))) discard;\n", dst_param
.reg_name
);
2827 /* 1.X shaders only compare the first 3 components, probably due to the nature of the texkill
2828 * instruction as a tex* instruction, and phase, which kills all a / w components. Even if all
2829 * 4 components are defined, only the first 3 are used
2831 shader_addline(arg
->buffer
, "if (any(lessThan(%s.xyz, vec3(0.0)))) discard;\n", dst_param
.reg_name
);
2835 /** Process the WINED3DSIO_DP2ADD instruction in GLSL.
2836 * dst = dot2(src0, src1) + src2 */
2837 static void pshader_glsl_dp2add(const SHADER_OPCODE_ARG
*arg
)
2839 glsl_src_param_t src0_param
;
2840 glsl_src_param_t src1_param
;
2841 glsl_src_param_t src2_param
;
2843 unsigned int mask_size
;
2845 write_mask
= shader_glsl_append_dst(arg
->buffer
, arg
);
2846 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
2848 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
, &src0_param
);
2849 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
, &src1_param
);
2850 shader_glsl_add_src_param(arg
, arg
->src
[2], arg
->src_addr
[2], WINED3DSP_WRITEMASK_0
, &src2_param
);
2852 if (mask_size
> 1) {
2853 shader_addline(arg
->buffer
, "vec%d(dot(%s, %s) + %s));\n", mask_size
, src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
2855 shader_addline(arg
->buffer
, "dot(%s, %s) + %s);\n", src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
2859 static void pshader_glsl_input_pack(SHADER_BUFFER
* buffer
, const struct semantic
* semantics_in
,
2860 IWineD3DPixelShader
*iface
, enum vertexprocessing_mode vertexprocessing
)
2863 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*) iface
;
2865 for (i
= 0; i
< MAX_REG_INPUT
; i
++) {
2867 DWORD usage_token
= semantics_in
[i
].usage
;
2868 DWORD register_token
= semantics_in
[i
].reg
;
2869 DWORD usage
, usage_idx
;
2873 if (!usage_token
) continue;
2874 usage
= (usage_token
& WINED3DSP_DCL_USAGE_MASK
) >> WINED3DSP_DCL_USAGE_SHIFT
;
2875 usage_idx
= (usage_token
& WINED3DSP_DCL_USAGEINDEX_MASK
) >> WINED3DSP_DCL_USAGEINDEX_SHIFT
;
2876 shader_glsl_get_write_mask(register_token
, reg_mask
);
2880 case WINED3DDECLUSAGE_TEXCOORD
:
2881 if(usage_idx
< 8 && vertexprocessing
== pretransformed
) {
2882 shader_addline(buffer
, "IN[%u]%s = gl_TexCoord[%u]%s;\n",
2883 This
->input_reg_map
[i
], reg_mask
, usage_idx
, reg_mask
);
2885 shader_addline(buffer
, "IN[%u]%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
2886 This
->input_reg_map
[i
], reg_mask
, reg_mask
);
2890 case WINED3DDECLUSAGE_COLOR
:
2892 shader_addline(buffer
, "IN[%u]%s = vec4(gl_Color)%s;\n",
2893 This
->input_reg_map
[i
], reg_mask
, reg_mask
);
2894 else if (usage_idx
== 1)
2895 shader_addline(buffer
, "IN[%u]%s = vec4(gl_SecondaryColor)%s;\n",
2896 This
->input_reg_map
[i
], reg_mask
, reg_mask
);
2898 shader_addline(buffer
, "IN[%u]%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
2899 This
->input_reg_map
[i
], reg_mask
, reg_mask
);
2903 shader_addline(buffer
, "IN[%u]%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
2904 This
->input_reg_map
[i
], reg_mask
, reg_mask
);
2909 /*********************************************
2910 * Vertex Shader Specific Code begins here
2911 ********************************************/
2913 static void add_glsl_program_entry(struct shader_glsl_priv
*priv
, struct glsl_shader_prog_link
*entry
) {
2914 glsl_program_key_t
*key
;
2916 key
= HeapAlloc(GetProcessHeap(), 0, sizeof(glsl_program_key_t
));
2917 key
->vshader
= entry
->vshader
;
2918 key
->pshader
= entry
->pshader
;
2919 key
->ps_args
= entry
->ps_args
;
2921 hash_table_put(priv
->glsl_program_lookup
, key
, entry
);
2924 static struct glsl_shader_prog_link
*get_glsl_program_entry(struct shader_glsl_priv
*priv
,
2925 GLhandleARB vshader
, IWineD3DPixelShader
*pshader
, struct ps_compile_args
*ps_args
) {
2926 glsl_program_key_t key
;
2928 key
.vshader
= vshader
;
2929 key
.pshader
= pshader
;
2930 key
.ps_args
= *ps_args
;
2932 return (struct glsl_shader_prog_link
*)hash_table_get(priv
->glsl_program_lookup
, &key
);
2935 static void delete_glsl_program_entry(struct shader_glsl_priv
*priv
, const WineD3D_GL_Info
*gl_info
,
2936 struct glsl_shader_prog_link
*entry
)
2938 glsl_program_key_t
*key
;
2940 key
= HeapAlloc(GetProcessHeap(), 0, sizeof(glsl_program_key_t
));
2941 key
->vshader
= entry
->vshader
;
2942 key
->pshader
= entry
->pshader
;
2943 key
->ps_args
= entry
->ps_args
;
2944 hash_table_remove(priv
->glsl_program_lookup
, key
);
2946 GL_EXTCALL(glDeleteObjectARB(entry
->programId
));
2947 if (entry
->vshader
) list_remove(&entry
->vshader_entry
);
2948 if (entry
->pshader
) list_remove(&entry
->pshader_entry
);
2949 HeapFree(GetProcessHeap(), 0, entry
->vuniformF_locations
);
2950 HeapFree(GetProcessHeap(), 0, entry
->puniformF_locations
);
2951 HeapFree(GetProcessHeap(), 0, entry
);
2954 static void handle_ps3_input(SHADER_BUFFER
*buffer
, const struct semantic
*semantics_in
,
2955 const struct semantic
*semantics_out
, const WineD3D_GL_Info
*gl_info
, const DWORD
*map
)
2958 DWORD usage_token
, usage_token_out
;
2959 DWORD register_token
, register_token_out
;
2960 DWORD usage
, usage_idx
, usage_out
, usage_idx_out
;
2963 DWORD in_count
= GL_LIMITS(glsl_varyings
) / 4;
2964 char reg_mask
[6], reg_mask_out
[6];
2965 char destination
[50];
2967 set
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*set
) * (in_count
+ 2));
2969 if (!semantics_out
) {
2970 /* Save gl_FrontColor & gl_FrontSecondaryColor before overwriting them. */
2971 shader_addline(buffer
, "vec4 front_color = gl_FrontColor;\n");
2972 shader_addline(buffer
, "vec4 front_secondary_color = gl_FrontSecondaryColor;\n");
2975 for(i
= 0; i
< MAX_REG_INPUT
; i
++) {
2976 usage_token
= semantics_in
[i
].usage
;
2977 if (!usage_token
) continue;
2980 if (in_idx
>= (in_count
+ 2)) {
2981 FIXME("More input varyings declared than supported, expect issues\n");
2983 } else if(map
[i
] == -1) {
2984 /* Declared, but not read register */
2988 if (in_idx
== in_count
) {
2989 sprintf(destination
, "gl_FrontColor");
2990 } else if (in_idx
== in_count
+ 1) {
2991 sprintf(destination
, "gl_FrontSecondaryColor");
2993 sprintf(destination
, "IN[%u]", in_idx
);
2996 register_token
= semantics_in
[i
].reg
;
2998 usage
= (usage_token
& WINED3DSP_DCL_USAGE_MASK
) >> WINED3DSP_DCL_USAGE_SHIFT
;
2999 usage_idx
= (usage_token
& WINED3DSP_DCL_USAGEINDEX_MASK
) >> WINED3DSP_DCL_USAGEINDEX_SHIFT
;
3000 set
[map
[i
]] = shader_glsl_get_write_mask(register_token
, reg_mask
);
3002 if(!semantics_out
) {
3004 case WINED3DDECLUSAGE_COLOR
:
3006 shader_addline(buffer
, "%s%s = front_color%s;\n",
3007 destination
, reg_mask
, reg_mask
);
3008 else if (usage_idx
== 1)
3009 shader_addline(buffer
, "%s%s = front_secondary_color%s;\n",
3010 destination
, reg_mask
, reg_mask
);
3012 shader_addline(buffer
, "%s%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
3013 destination
, reg_mask
, reg_mask
);
3016 case WINED3DDECLUSAGE_TEXCOORD
:
3017 if (usage_idx
< 8) {
3018 shader_addline(buffer
, "%s%s = gl_TexCoord[%u]%s;\n",
3019 destination
, reg_mask
, usage_idx
, reg_mask
);
3021 shader_addline(buffer
, "%s%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
3022 destination
, reg_mask
, reg_mask
);
3026 case WINED3DDECLUSAGE_FOG
:
3027 shader_addline(buffer
, "%s%s = vec4(gl_FogFragCoord, 0.0, 0.0, 0.0)%s;\n",
3028 destination
, reg_mask
, reg_mask
);
3032 shader_addline(buffer
, "%s%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
3033 destination
, reg_mask
, reg_mask
);
3037 for(j
= 0; j
< MAX_REG_OUTPUT
; j
++) {
3038 usage_token_out
= semantics_out
[j
].usage
;
3039 if (!usage_token_out
) continue;
3040 register_token_out
= semantics_out
[j
].reg
;
3042 usage_out
= (usage_token_out
& WINED3DSP_DCL_USAGE_MASK
) >> WINED3DSP_DCL_USAGE_SHIFT
;
3043 usage_idx_out
= (usage_token_out
& WINED3DSP_DCL_USAGEINDEX_MASK
) >> WINED3DSP_DCL_USAGEINDEX_SHIFT
;
3044 shader_glsl_get_write_mask(register_token_out
, reg_mask_out
);
3046 if(usage
== usage_out
&&
3047 usage_idx
== usage_idx_out
) {
3048 shader_addline(buffer
, "%s%s = OUT[%u]%s;\n",
3049 destination
, reg_mask
, j
, reg_mask
);
3054 shader_addline(buffer
, "%s%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
3055 destination
, reg_mask
, reg_mask
);
3060 /* This is solely to make the compiler / linker happy and avoid warning about undefined
3061 * varyings. It shouldn't result in any real code executed on the GPU, since all read
3062 * input varyings are assigned above, if the optimizer works properly.
3064 for(i
= 0; i
< in_count
+ 2; i
++) {
3065 if(set
[i
] != WINED3DSP_WRITEMASK_ALL
) {
3066 unsigned int size
= 0;
3067 memset(reg_mask
, 0, sizeof(reg_mask
));
3068 if(!(set
[i
] & WINED3DSP_WRITEMASK_0
)) {
3069 reg_mask
[size
] = 'x';
3072 if(!(set
[i
] & WINED3DSP_WRITEMASK_1
)) {
3073 reg_mask
[size
] = 'y';
3076 if(!(set
[i
] & WINED3DSP_WRITEMASK_2
)) {
3077 reg_mask
[size
] = 'z';
3080 if(!(set
[i
] & WINED3DSP_WRITEMASK_3
)) {
3081 reg_mask
[size
] = 'w';
3085 if (i
== in_count
) {
3086 sprintf(destination
, "gl_FrontColor");
3087 } else if (i
== in_count
+ 1) {
3088 sprintf(destination
, "gl_FrontSecondaryColor");
3090 sprintf(destination
, "IN[%u]", i
);
3094 shader_addline(buffer
, "%s.%s = 0.0;\n", destination
, reg_mask
);
3096 shader_addline(buffer
, "%s.%s = vec%u(0.0);\n", destination
, reg_mask
, size
);
3101 HeapFree(GetProcessHeap(), 0, set
);
3104 static GLhandleARB
generate_param_reorder_function(IWineD3DVertexShader
*vertexshader
,
3105 IWineD3DPixelShader
*pixelshader
, const WineD3D_GL_Info
*gl_info
)
3107 GLhandleARB ret
= 0;
3108 IWineD3DVertexShaderImpl
*vs
= (IWineD3DVertexShaderImpl
*) vertexshader
;
3109 IWineD3DPixelShaderImpl
*ps
= (IWineD3DPixelShaderImpl
*) pixelshader
;
3110 IWineD3DDeviceImpl
*device
;
3111 DWORD vs_major
= WINED3DSHADER_VERSION_MAJOR(vs
->baseShader
.reg_maps
.shader_version
);
3112 DWORD ps_major
= ps
? WINED3DSHADER_VERSION_MAJOR(ps
->baseShader
.reg_maps
.shader_version
) : 0;
3114 SHADER_BUFFER buffer
;
3116 DWORD register_token
;
3117 DWORD usage
, usage_idx
, writemask
;
3119 const struct semantic
*semantics_out
, *semantics_in
;
3121 shader_buffer_init(&buffer
);
3123 shader_addline(&buffer
, "#version 120\n");
3125 if(vs_major
< 3 && ps_major
< 3) {
3126 /* That one is easy: The vertex shader writes to the builtin varyings, the pixel shader reads from them.
3127 * Take care about the texcoord .w fixup though if we're using the fixed function fragment pipeline
3129 device
= (IWineD3DDeviceImpl
*) vs
->baseShader
.device
;
3130 if((GLINFO_LOCATION
).set_texcoord_w
&& ps_major
== 0 && vs_major
> 0 &&
3131 !device
->frag_pipe
->ffp_proj_control
) {
3132 shader_addline(&buffer
, "void order_ps_input() {\n");
3133 for(i
= 0; i
< min(8, MAX_REG_TEXCRD
); i
++) {
3134 if(vs
->baseShader
.reg_maps
.texcoord_mask
[i
] != 0 &&
3135 vs
->baseShader
.reg_maps
.texcoord_mask
[i
] != WINED3DSP_WRITEMASK_ALL
) {
3136 shader_addline(&buffer
, "gl_TexCoord[%u].w = 1.0;\n", i
);
3139 shader_addline(&buffer
, "}\n");
3141 shader_addline(&buffer
, "void order_ps_input() { /* do nothing */ }\n");
3143 } else if(ps_major
< 3 && vs_major
>= 3) {
3144 /* The vertex shader writes to its own varyings, the pixel shader needs them in the builtin ones */
3145 semantics_out
= vs
->semantics_out
;
3147 shader_addline(&buffer
, "void order_ps_input(in vec4 OUT[%u]) {\n", MAX_REG_OUTPUT
);
3148 for(i
= 0; i
< MAX_REG_OUTPUT
; i
++) {
3149 usage_token
= semantics_out
[i
].usage
;
3150 if (!usage_token
) continue;
3151 register_token
= semantics_out
[i
].reg
;
3153 usage
= (usage_token
& WINED3DSP_DCL_USAGE_MASK
) >> WINED3DSP_DCL_USAGE_SHIFT
;
3154 usage_idx
= (usage_token
& WINED3DSP_DCL_USAGEINDEX_MASK
) >> WINED3DSP_DCL_USAGEINDEX_SHIFT
;
3155 writemask
= shader_glsl_get_write_mask(register_token
, reg_mask
);
3158 case WINED3DDECLUSAGE_COLOR
:
3160 shader_addline(&buffer
, "gl_FrontColor%s = OUT[%u]%s;\n", reg_mask
, i
, reg_mask
);
3161 else if (usage_idx
== 1)
3162 shader_addline(&buffer
, "gl_FrontSecondaryColor%s = OUT[%u]%s;\n", reg_mask
, i
, reg_mask
);
3165 case WINED3DDECLUSAGE_POSITION
:
3166 shader_addline(&buffer
, "gl_Position%s = OUT[%u]%s;\n", reg_mask
, i
, reg_mask
);
3169 case WINED3DDECLUSAGE_TEXCOORD
:
3170 if (usage_idx
< 8) {
3171 if(!(GLINFO_LOCATION
).set_texcoord_w
|| ps_major
> 0) writemask
|= WINED3DSP_WRITEMASK_3
;
3173 shader_addline(&buffer
, "gl_TexCoord[%u]%s = OUT[%u]%s;\n",
3174 usage_idx
, reg_mask
, i
, reg_mask
);
3175 if(!(writemask
& WINED3DSP_WRITEMASK_3
)) {
3176 shader_addline(&buffer
, "gl_TexCoord[%u].w = 1.0;\n", usage_idx
);
3181 case WINED3DDECLUSAGE_PSIZE
:
3182 shader_addline(&buffer
, "gl_PointSize = OUT[%u].x;\n", i
);
3185 case WINED3DDECLUSAGE_FOG
:
3186 shader_addline(&buffer
, "gl_FogFragCoord = OUT[%u].%c;\n", i
, reg_mask
[1]);
3193 shader_addline(&buffer
, "}\n");
3195 } else if(ps_major
>= 3 && vs_major
>= 3) {
3196 semantics_out
= vs
->semantics_out
;
3197 semantics_in
= ps
->semantics_in
;
3199 /* This one is tricky: a 3.0 pixel shader reads from a 3.0 vertex shader */
3200 shader_addline(&buffer
, "varying vec4 IN[%u];\n", GL_LIMITS(glsl_varyings
) / 4);
3201 shader_addline(&buffer
, "void order_ps_input(in vec4 OUT[%u]) {\n", MAX_REG_OUTPUT
);
3203 /* First, sort out position and point size. Those are not passed to the pixel shader */
3204 for(i
= 0; i
< MAX_REG_OUTPUT
; i
++) {
3205 usage_token
= semantics_out
[i
].usage
;
3206 if (!usage_token
) continue;
3207 register_token
= semantics_out
[i
].reg
;
3209 usage
= (usage_token
& WINED3DSP_DCL_USAGE_MASK
) >> WINED3DSP_DCL_USAGE_SHIFT
;
3210 usage_idx
= (usage_token
& WINED3DSP_DCL_USAGEINDEX_MASK
) >> WINED3DSP_DCL_USAGEINDEX_SHIFT
;
3211 shader_glsl_get_write_mask(register_token
, reg_mask
);
3214 case WINED3DDECLUSAGE_POSITION
:
3215 shader_addline(&buffer
, "gl_Position%s = OUT[%u]%s;\n", reg_mask
, i
, reg_mask
);
3218 case WINED3DDECLUSAGE_PSIZE
:
3219 shader_addline(&buffer
, "gl_PointSize = OUT[%u].x;\n", i
);
3227 /* Then, fix the pixel shader input */
3228 handle_ps3_input(&buffer
, semantics_in
, semantics_out
, gl_info
, ps
->input_reg_map
);
3230 shader_addline(&buffer
, "}\n");
3231 } else if(ps_major
>= 3 && vs_major
< 3) {
3232 semantics_in
= ps
->semantics_in
;
3234 shader_addline(&buffer
, "varying vec4 IN[%u];\n", GL_LIMITS(glsl_varyings
) / 4);
3235 shader_addline(&buffer
, "void order_ps_input() {\n");
3236 /* The vertex shader wrote to the builtin varyings. There is no need to figure out position and
3237 * point size, but we depend on the optimizers kindness to find out that the pixel shader doesn't
3238 * read gl_TexCoord and gl_ColorX, otherwise we'll run out of varyings
3240 handle_ps3_input(&buffer
, semantics_in
, NULL
, gl_info
, ps
->input_reg_map
);
3241 shader_addline(&buffer
, "}\n");
3243 ERR("Unexpected vertex and pixel shader version condition: vs: %d, ps: %d\n", vs_major
, ps_major
);
3246 ret
= GL_EXTCALL(glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB
));
3247 checkGLcall("glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB)");
3248 GL_EXTCALL(glShaderSourceARB(ret
, 1, (const char**)&buffer
.buffer
, NULL
));
3249 checkGLcall("glShaderSourceARB(ret, 1, (const char**)&buffer.buffer, NULL)");
3250 GL_EXTCALL(glCompileShaderARB(ret
));
3251 checkGLcall("glCompileShaderARB(ret)");
3253 shader_buffer_free(&buffer
);
3257 static void hardcode_local_constants(IWineD3DBaseShaderImpl
*shader
, const WineD3D_GL_Info
*gl_info
,
3258 GLhandleARB programId
, char prefix
)
3260 const local_constant
*lconst
;
3265 LIST_FOR_EACH_ENTRY(lconst
, &shader
->baseShader
.constantsF
, local_constant
, entry
) {
3266 value
= (const float *)lconst
->value
;
3267 snprintf(glsl_name
, sizeof(glsl_name
), "%cLC%u", prefix
, lconst
->idx
);
3268 tmp_loc
= GL_EXTCALL(glGetUniformLocationARB(programId
, glsl_name
));
3269 GL_EXTCALL(glUniform4fvARB(tmp_loc
, 1, value
));
3271 checkGLcall("Hardcoding local constants\n");
3274 /** Sets the GLSL program ID for the given pixel and vertex shader combination.
3275 * It sets the programId on the current StateBlock (because it should be called
3276 * inside of the DrawPrimitive() part of the render loop).
3278 * If a program for the given combination does not exist, create one, and store
3279 * the program in the hash table. If it creates a program, it will link the
3280 * given objects, too.
3282 static void set_glsl_shader_program(IWineD3DDevice
*iface
, BOOL use_ps
, BOOL use_vs
) {
3283 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
3284 struct shader_glsl_priv
*priv
= (struct shader_glsl_priv
*)This
->shader_priv
;
3285 const WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
3286 IWineD3DPixelShader
*pshader
= This
->stateBlock
->pixelShader
;
3287 IWineD3DVertexShader
*vshader
= This
->stateBlock
->vertexShader
;
3288 struct glsl_shader_prog_link
*entry
= NULL
;
3289 GLhandleARB programId
= 0;
3290 GLhandleARB reorder_shader_id
= 0;
3293 GLhandleARB vshader_id
, pshader_id
;
3294 struct ps_compile_args compile_args
;
3297 IWineD3DVertexShaderImpl_CompileShader(vshader
);
3298 vshader_id
= ((IWineD3DVertexShaderImpl
*)vshader
)->prgId
;
3303 find_ps_compile_args((IWineD3DPixelShaderImpl
*)This
->stateBlock
->pixelShader
, This
->stateBlock
, &compile_args
);
3305 /* FIXME: Do we really have to spend CPU cycles to generate a few zeroed bytes? */
3306 memset(&compile_args
, 0, sizeof(compile_args
));
3308 entry
= get_glsl_program_entry(priv
, vshader_id
, pshader
, &compile_args
);
3310 priv
->glsl_program
= entry
;
3314 /* If we get to this point, then no matching program exists, so we create one */
3315 programId
= GL_EXTCALL(glCreateProgramObjectARB());
3316 TRACE("Created new GLSL shader program %u\n", programId
);
3318 /* Create the entry */
3319 entry
= HeapAlloc(GetProcessHeap(), 0, sizeof(struct glsl_shader_prog_link
));
3320 entry
->programId
= programId
;
3321 entry
->vshader
= vshader_id
;
3322 entry
->pshader
= pshader
;
3323 entry
->ps_args
= compile_args
;
3324 entry
->constant_version
= 0;
3325 /* Add the hash table entry */
3326 add_glsl_program_entry(priv
, entry
);
3328 /* Set the current program */
3329 priv
->glsl_program
= entry
;
3331 /* Attach GLSL vshader */
3333 int max_attribs
= 16; /* TODO: Will this always be the case? It is at the moment... */
3336 reorder_shader_id
= generate_param_reorder_function(vshader
, pshader
, gl_info
);
3337 TRACE("Attaching GLSL shader object %u to program %u\n", reorder_shader_id
, programId
);
3338 GL_EXTCALL(glAttachObjectARB(programId
, reorder_shader_id
));
3339 checkGLcall("glAttachObjectARB");
3340 /* Flag the reorder function for deletion, then it will be freed automatically when the program
3343 GL_EXTCALL(glDeleteObjectARB(reorder_shader_id
));
3345 TRACE("Attaching GLSL shader object %u to program %u\n", vshader_id
, programId
);
3346 GL_EXTCALL(glAttachObjectARB(programId
, vshader_id
));
3347 checkGLcall("glAttachObjectARB");
3349 /* Bind vertex attributes to a corresponding index number to match
3350 * the same index numbers as ARB_vertex_programs (makes loading
3351 * vertex attributes simpler). With this method, we can use the
3352 * exact same code to load the attributes later for both ARB and
3355 * We have to do this here because we need to know the Program ID
3356 * in order to make the bindings work, and it has to be done prior
3357 * to linking the GLSL program. */
3358 for (i
= 0; i
< max_attribs
; ++i
) {
3359 if (((IWineD3DBaseShaderImpl
*)vshader
)->baseShader
.reg_maps
.attributes
[i
]) {
3360 snprintf(tmp_name
, sizeof(tmp_name
), "attrib%i", i
);
3361 GL_EXTCALL(glBindAttribLocationARB(programId
, i
, tmp_name
));
3364 checkGLcall("glBindAttribLocationARB");
3366 list_add_head(&((IWineD3DBaseShaderImpl
*)vshader
)->baseShader
.linked_programs
, &entry
->vshader_entry
);
3370 pshader_id
= find_gl_pshader((IWineD3DPixelShaderImpl
*) pshader
, &compile_args
);
3375 /* Attach GLSL pshader */
3377 TRACE("Attaching GLSL shader object %u to program %u\n", pshader_id
, programId
);
3378 GL_EXTCALL(glAttachObjectARB(programId
, pshader_id
));
3379 checkGLcall("glAttachObjectARB");
3381 list_add_head(&((IWineD3DBaseShaderImpl
*)pshader
)->baseShader
.linked_programs
, &entry
->pshader_entry
);
3384 /* Link the program */
3385 TRACE("Linking GLSL shader program %u\n", programId
);
3386 GL_EXTCALL(glLinkProgramARB(programId
));
3387 print_glsl_info_log(&GLINFO_LOCATION
, programId
);
3389 entry
->vuniformF_locations
= HeapAlloc(GetProcessHeap(), 0, sizeof(GLhandleARB
) * GL_LIMITS(vshader_constantsF
));
3390 for (i
= 0; i
< GL_LIMITS(vshader_constantsF
); ++i
) {
3391 snprintf(glsl_name
, sizeof(glsl_name
), "VC[%i]", i
);
3392 entry
->vuniformF_locations
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, glsl_name
));
3394 for (i
= 0; i
< MAX_CONST_I
; ++i
) {
3395 snprintf(glsl_name
, sizeof(glsl_name
), "VI[%i]", i
);
3396 entry
->vuniformI_locations
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, glsl_name
));
3398 entry
->puniformF_locations
= HeapAlloc(GetProcessHeap(), 0, sizeof(GLhandleARB
) * GL_LIMITS(pshader_constantsF
));
3399 for (i
= 0; i
< GL_LIMITS(pshader_constantsF
); ++i
) {
3400 snprintf(glsl_name
, sizeof(glsl_name
), "PC[%i]", i
);
3401 entry
->puniformF_locations
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, glsl_name
));
3403 for (i
= 0; i
< MAX_CONST_I
; ++i
) {
3404 snprintf(glsl_name
, sizeof(glsl_name
), "PI[%i]", i
);
3405 entry
->puniformI_locations
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, glsl_name
));
3409 for(i
= 0; i
< ((IWineD3DPixelShaderImpl
*)pshader
)->numbumpenvmatconsts
; i
++) {
3411 sprintf(name
, "bumpenvmat%d", ((IWineD3DPixelShaderImpl
*)pshader
)->bumpenvmatconst
[i
].texunit
);
3412 entry
->bumpenvmat_location
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, name
));
3413 sprintf(name
, "luminancescale%d", ((IWineD3DPixelShaderImpl
*)pshader
)->luminanceconst
[i
].texunit
);
3414 entry
->luminancescale_location
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, name
));
3415 sprintf(name
, "luminanceoffset%d", ((IWineD3DPixelShaderImpl
*)pshader
)->luminanceconst
[i
].texunit
);
3416 entry
->luminanceoffset_location
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, name
));
3421 entry
->posFixup_location
= GL_EXTCALL(glGetUniformLocationARB(programId
, "posFixup"));
3422 entry
->ycorrection_location
= GL_EXTCALL(glGetUniformLocationARB(programId
, "ycorrection"));
3423 checkGLcall("Find glsl program uniform locations");
3426 && WINED3DSHADER_VERSION_MAJOR(((IWineD3DPixelShaderImpl
*)pshader
)->baseShader
.reg_maps
.shader_version
) >= 3
3427 && ((IWineD3DPixelShaderImpl
*)pshader
)->declared_in_count
> GL_LIMITS(glsl_varyings
) / 4)
3429 TRACE("Shader %d needs vertex color clamping disabled\n", programId
);
3430 entry
->vertex_color_clamp
= GL_FALSE
;
3432 entry
->vertex_color_clamp
= GL_FIXED_ONLY_ARB
;
3435 /* Set the shader to allow uniform loading on it */
3436 GL_EXTCALL(glUseProgramObjectARB(programId
));
3437 checkGLcall("glUseProgramObjectARB(programId)");
3439 /* Load the vertex and pixel samplers now. The function that finds the mappings makes sure
3440 * that it stays the same for each vertexshader-pixelshader pair(=linked glsl program). If
3441 * a pshader with fixed function pipeline is used there are no vertex samplers, and if a
3442 * vertex shader with fixed function pixel processing is used we make sure that the card
3443 * supports enough samplers to allow the max number of vertex samplers with all possible
3444 * fixed function fragment processing setups. So once the program is linked these samplers
3448 /* Load vertex shader samplers */
3449 shader_glsl_load_vsamplers(gl_info
, (IWineD3DStateBlock
*)This
->stateBlock
, programId
);
3452 /* Load pixel shader samplers */
3453 shader_glsl_load_psamplers(gl_info
, (IWineD3DStateBlock
*)This
->stateBlock
, programId
);
3456 /* If the local constants do not have to be loaded with the environment constants,
3457 * load them now to have them hardcoded in the GLSL program. This saves some CPU cycles
3460 if(pshader
&& !((IWineD3DPixelShaderImpl
*)pshader
)->baseShader
.load_local_constsF
) {
3461 hardcode_local_constants((IWineD3DBaseShaderImpl
*) pshader
, gl_info
, programId
, 'P');
3463 if(vshader
&& !((IWineD3DVertexShaderImpl
*)vshader
)->baseShader
.load_local_constsF
) {
3464 hardcode_local_constants((IWineD3DBaseShaderImpl
*) vshader
, gl_info
, programId
, 'V');
3468 static GLhandleARB
create_glsl_blt_shader(const WineD3D_GL_Info
*gl_info
, enum tex_types tex_type
)
3470 GLhandleARB program_id
;
3471 GLhandleARB vshader_id
, pshader_id
;
3472 static const char *blt_vshader
[] =
3477 " gl_Position = gl_Vertex;\n"
3478 " gl_FrontColor = vec4(1.0);\n"
3479 " gl_TexCoord[0] = gl_MultiTexCoord0;\n"
3483 static const char *blt_pshaders
[tex_type_count
] =
3489 "uniform sampler2D sampler;\n"
3492 " gl_FragDepth = texture2D(sampler, gl_TexCoord[0].xy).x;\n"
3498 "uniform samplerCube sampler;\n"
3501 " gl_FragDepth = textureCube(sampler, gl_TexCoord[0].xyz).x;\n"
3505 "#extension GL_ARB_texture_rectangle : enable\n"
3506 "uniform sampler2DRect sampler;\n"
3509 " gl_FragDepth = texture2DRect(sampler, gl_TexCoord[0].xy).x;\n"
3513 if (!blt_pshaders
[tex_type
])
3515 FIXME("tex_type %#x not supported\n", tex_type
);
3519 vshader_id
= GL_EXTCALL(glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB
));
3520 GL_EXTCALL(glShaderSourceARB(vshader_id
, 1, blt_vshader
, NULL
));
3521 GL_EXTCALL(glCompileShaderARB(vshader_id
));
3523 pshader_id
= GL_EXTCALL(glCreateShaderObjectARB(GL_FRAGMENT_SHADER_ARB
));
3524 GL_EXTCALL(glShaderSourceARB(pshader_id
, 1, &blt_pshaders
[tex_type
], NULL
));
3525 GL_EXTCALL(glCompileShaderARB(pshader_id
));
3527 program_id
= GL_EXTCALL(glCreateProgramObjectARB());
3528 GL_EXTCALL(glAttachObjectARB(program_id
, vshader_id
));
3529 GL_EXTCALL(glAttachObjectARB(program_id
, pshader_id
));
3530 GL_EXTCALL(glLinkProgramARB(program_id
));
3532 print_glsl_info_log(&GLINFO_LOCATION
, program_id
);
3534 /* Once linked we can mark the shaders for deletion. They will be deleted once the program
3537 GL_EXTCALL(glDeleteObjectARB(vshader_id
));
3538 GL_EXTCALL(glDeleteObjectARB(pshader_id
));
3542 static void shader_glsl_select(IWineD3DDevice
*iface
, BOOL usePS
, BOOL useVS
) {
3543 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
3544 struct shader_glsl_priv
*priv
= (struct shader_glsl_priv
*)This
->shader_priv
;
3545 const WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
3546 GLhandleARB program_id
= 0;
3547 GLenum old_vertex_color_clamp
, current_vertex_color_clamp
;
3549 old_vertex_color_clamp
= priv
->glsl_program
? priv
->glsl_program
->vertex_color_clamp
: GL_FIXED_ONLY_ARB
;
3551 if (useVS
|| usePS
) set_glsl_shader_program(iface
, usePS
, useVS
);
3552 else priv
->glsl_program
= NULL
;
3554 current_vertex_color_clamp
= priv
->glsl_program
? priv
->glsl_program
->vertex_color_clamp
: GL_FIXED_ONLY_ARB
;
3556 if (old_vertex_color_clamp
!= current_vertex_color_clamp
) {
3557 if (GL_SUPPORT(ARB_COLOR_BUFFER_FLOAT
)) {
3558 GL_EXTCALL(glClampColorARB(GL_CLAMP_VERTEX_COLOR_ARB
, current_vertex_color_clamp
));
3559 checkGLcall("glClampColorARB");
3561 FIXME("vertex color clamp needs to be changed, but extension not supported.\n");
3565 program_id
= priv
->glsl_program
? priv
->glsl_program
->programId
: 0;
3566 if (program_id
) TRACE("Using GLSL program %u\n", program_id
);
3567 GL_EXTCALL(glUseProgramObjectARB(program_id
));
3568 checkGLcall("glUseProgramObjectARB");
3571 static void shader_glsl_select_depth_blt(IWineD3DDevice
*iface
, enum tex_types tex_type
) {
3572 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
3573 const WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
3574 struct shader_glsl_priv
*priv
= (struct shader_glsl_priv
*) This
->shader_priv
;
3575 GLhandleARB
*blt_program
= &priv
->depth_blt_program
[tex_type
];
3577 if (!*blt_program
) {
3579 *blt_program
= create_glsl_blt_shader(gl_info
, tex_type
);
3580 loc
= GL_EXTCALL(glGetUniformLocationARB(*blt_program
, "sampler"));
3581 GL_EXTCALL(glUseProgramObjectARB(*blt_program
));
3582 GL_EXTCALL(glUniform1iARB(loc
, 0));
3584 GL_EXTCALL(glUseProgramObjectARB(*blt_program
));
3588 static void shader_glsl_deselect_depth_blt(IWineD3DDevice
*iface
) {
3589 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
3590 const WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
3591 struct shader_glsl_priv
*priv
= (struct shader_glsl_priv
*) This
->shader_priv
;
3592 GLhandleARB program_id
;
3594 program_id
= priv
->glsl_program
? priv
->glsl_program
->programId
: 0;
3595 if (program_id
) TRACE("Using GLSL program %u\n", program_id
);
3597 GL_EXTCALL(glUseProgramObjectARB(program_id
));
3598 checkGLcall("glUseProgramObjectARB");
3601 static void shader_glsl_destroy(IWineD3DBaseShader
*iface
) {
3602 const struct list
*linked_programs
;
3603 IWineD3DBaseShaderImpl
*This
= (IWineD3DBaseShaderImpl
*) iface
;
3604 IWineD3DDeviceImpl
*device
= (IWineD3DDeviceImpl
*)This
->baseShader
.device
;
3605 struct shader_glsl_priv
*priv
= (struct shader_glsl_priv
*)device
->shader_priv
;
3606 const WineD3D_GL_Info
*gl_info
= &device
->adapter
->gl_info
;
3607 IWineD3DPixelShaderImpl
*ps
= NULL
;
3608 IWineD3DVertexShaderImpl
*vs
= NULL
;
3610 /* Note: Do not use QueryInterface here to find out which shader type this is because this code
3611 * can be called from IWineD3DBaseShader::Release
3613 char pshader
= shader_is_pshader_version(This
->baseShader
.reg_maps
.shader_version
);
3616 ps
= (IWineD3DPixelShaderImpl
*) This
;
3617 if(ps
->num_gl_shaders
== 0) return;
3619 vs
= (IWineD3DVertexShaderImpl
*) This
;
3620 if(vs
->prgId
== 0) return;
3623 linked_programs
= &This
->baseShader
.linked_programs
;
3625 TRACE("Deleting linked programs\n");
3626 if (linked_programs
->next
) {
3627 struct glsl_shader_prog_link
*entry
, *entry2
;
3630 LIST_FOR_EACH_ENTRY_SAFE(entry
, entry2
, linked_programs
, struct glsl_shader_prog_link
, pshader_entry
) {
3631 delete_glsl_program_entry(priv
, gl_info
, entry
);
3634 LIST_FOR_EACH_ENTRY_SAFE(entry
, entry2
, linked_programs
, struct glsl_shader_prog_link
, vshader_entry
) {
3635 delete_glsl_program_entry(priv
, gl_info
, entry
);
3644 for(i
= 0; i
< ps
->num_gl_shaders
; i
++) {
3645 TRACE("deleting pshader %u\n", ps
->gl_shaders
[i
].prgId
);
3646 GL_EXTCALL(glDeleteObjectARB(ps
->gl_shaders
[i
].prgId
));
3647 checkGLcall("glDeleteObjectARB");
3650 HeapFree(GetProcessHeap(), 0, ps
->gl_shaders
);
3651 ps
->gl_shaders
= NULL
;
3652 ps
->num_gl_shaders
= 0;
3654 TRACE("Deleting shader object %u\n", vs
->prgId
);
3656 GL_EXTCALL(glDeleteObjectARB(vs
->prgId
));
3657 checkGLcall("glDeleteObjectARB");
3660 vs
->baseShader
.is_compiled
= FALSE
;
3664 static unsigned int glsl_program_key_hash(const void *key
)
3666 const glsl_program_key_t
*k
= (const glsl_program_key_t
*)key
;
3668 unsigned int hash
= k
->vshader
| ((DWORD_PTR
) k
->pshader
) << 16;
3669 hash
+= ~(hash
<< 15);
3670 hash
^= (hash
>> 10);
3671 hash
+= (hash
<< 3);
3672 hash
^= (hash
>> 6);
3673 hash
+= ~(hash
<< 11);
3674 hash
^= (hash
>> 16);
3679 static BOOL
glsl_program_key_compare(const void *keya
, const void *keyb
)
3681 const glsl_program_key_t
*ka
= (const glsl_program_key_t
*)keya
;
3682 const glsl_program_key_t
*kb
= (const glsl_program_key_t
*)keyb
;
3684 return ka
->vshader
== kb
->vshader
&& ka
->pshader
== kb
->pshader
&&
3685 (memcmp(&ka
->ps_args
, &kb
->ps_args
, sizeof(kb
->ps_args
)) == 0);
3688 static BOOL
constant_heap_init(struct constant_heap
*heap
, unsigned int constant_count
)
3690 SIZE_T size
= (constant_count
+ 1) * sizeof(*heap
->entries
) + constant_count
* sizeof(*heap
->positions
);
3691 void *mem
= HeapAlloc(GetProcessHeap(), 0, size
);
3695 ERR("Failed to allocate memory\n");
3699 heap
->entries
= mem
;
3700 heap
->entries
[1].version
= 0;
3701 heap
->positions
= (unsigned int *)(heap
->entries
+ constant_count
+ 1);
3707 static void constant_heap_free(struct constant_heap
*heap
)
3709 HeapFree(GetProcessHeap(), 0, heap
->entries
);
3712 static HRESULT
shader_glsl_alloc(IWineD3DDevice
*iface
) {
3713 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
3714 const WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
3715 struct shader_glsl_priv
*priv
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(struct shader_glsl_priv
));
3716 SIZE_T stack_size
= wined3d_log2i(max(GL_LIMITS(vshader_constantsF
), GL_LIMITS(pshader_constantsF
))) + 1;
3718 priv
->stack
= HeapAlloc(GetProcessHeap(), 0, stack_size
* sizeof(*priv
->stack
));
3721 ERR("Failed to allocate memory.\n");
3722 HeapFree(GetProcessHeap(), 0, priv
);
3723 return E_OUTOFMEMORY
;
3726 if (!constant_heap_init(&priv
->vconst_heap
, GL_LIMITS(vshader_constantsF
)))
3728 ERR("Failed to initialize vertex shader constant heap\n");
3729 HeapFree(GetProcessHeap(), 0, priv
->stack
);
3730 HeapFree(GetProcessHeap(), 0, priv
);
3731 return E_OUTOFMEMORY
;
3734 if (!constant_heap_init(&priv
->pconst_heap
, GL_LIMITS(pshader_constantsF
)))
3736 ERR("Failed to initialize pixel shader constant heap\n");
3737 constant_heap_free(&priv
->vconst_heap
);
3738 HeapFree(GetProcessHeap(), 0, priv
->stack
);
3739 HeapFree(GetProcessHeap(), 0, priv
);
3740 return E_OUTOFMEMORY
;
3743 priv
->glsl_program_lookup
= hash_table_create(glsl_program_key_hash
, glsl_program_key_compare
);
3744 priv
->next_constant_version
= 1;
3746 This
->shader_priv
= priv
;
3750 static void shader_glsl_free(IWineD3DDevice
*iface
) {
3751 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
3752 const WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
3753 struct shader_glsl_priv
*priv
= (struct shader_glsl_priv
*)This
->shader_priv
;
3756 for (i
= 0; i
< tex_type_count
; ++i
)
3758 if (priv
->depth_blt_program
[i
])
3760 GL_EXTCALL(glDeleteObjectARB(priv
->depth_blt_program
[i
]));
3764 hash_table_destroy(priv
->glsl_program_lookup
, NULL
, NULL
);
3765 constant_heap_free(&priv
->pconst_heap
);
3766 constant_heap_free(&priv
->vconst_heap
);
3768 HeapFree(GetProcessHeap(), 0, This
->shader_priv
);
3769 This
->shader_priv
= NULL
;
3772 static BOOL
shader_glsl_dirty_const(IWineD3DDevice
*iface
) {
3773 /* TODO: GL_EXT_bindable_uniform can be used to share constants across shaders */
3777 static GLuint
shader_glsl_generate_pshader(IWineD3DPixelShader
*iface
, SHADER_BUFFER
*buffer
, const struct ps_compile_args
*args
) {
3778 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*)iface
;
3779 const struct shader_reg_maps
*reg_maps
= &This
->baseShader
.reg_maps
;
3780 CONST DWORD
*function
= This
->baseShader
.function
;
3781 const char *fragcolor
;
3782 const WineD3D_GL_Info
*gl_info
= &((IWineD3DDeviceImpl
*)This
->baseShader
.device
)->adapter
->gl_info
;
3784 /* Create the hw GLSL shader object and assign it as the shader->prgId */
3785 GLhandleARB shader_obj
= GL_EXTCALL(glCreateShaderObjectARB(GL_FRAGMENT_SHADER_ARB
));
3787 shader_addline(buffer
, "#version 120\n");
3789 if (GL_SUPPORT(ARB_DRAW_BUFFERS
)) {
3790 shader_addline(buffer
, "#extension GL_ARB_draw_buffers : enable\n");
3792 if (GL_SUPPORT(ARB_TEXTURE_RECTANGLE
)) {
3793 /* The spec says that it doesn't have to be explicitly enabled, but the nvidia
3794 * drivers write a warning if we don't do so
3796 shader_addline(buffer
, "#extension GL_ARB_texture_rectangle : enable\n");
3799 /* Base Declarations */
3800 shader_generate_glsl_declarations( (IWineD3DBaseShader
*) This
, reg_maps
, buffer
, &GLINFO_LOCATION
, args
);
3802 /* Pack 3.0 inputs */
3803 if (reg_maps
->shader_version
>= WINED3DPS_VERSION(3,0) && args
->vp_mode
!= vertexshader
) {
3804 pshader_glsl_input_pack(buffer
, This
->semantics_in
, iface
, args
->vp_mode
);
3807 /* Base Shader Body */
3808 shader_generate_main( (IWineD3DBaseShader
*) This
, buffer
, reg_maps
, function
);
3810 /* Pixel shaders < 2.0 place the resulting color in R0 implicitly */
3811 if (reg_maps
->shader_version
< WINED3DPS_VERSION(2,0))
3813 /* Some older cards like GeforceFX ones don't support multiple buffers, so also not gl_FragData */
3814 if(GL_SUPPORT(ARB_DRAW_BUFFERS
))
3815 shader_addline(buffer
, "gl_FragData[0] = R0;\n");
3817 shader_addline(buffer
, "gl_FragColor = R0;\n");
3820 if(GL_SUPPORT(ARB_DRAW_BUFFERS
)) {
3821 fragcolor
= "gl_FragData[0]";
3823 fragcolor
= "gl_FragColor";
3825 if(args
->srgb_correction
) {
3826 shader_addline(buffer
, "tmp0.xyz = pow(%s.xyz, vec3(%f, %f, %f)) * vec3(%f, %f, %f) - vec3(%f, %f, %f);\n",
3827 fragcolor
, srgb_pow
, srgb_pow
, srgb_pow
, srgb_mul_high
, srgb_mul_high
, srgb_mul_high
,
3828 srgb_sub_high
, srgb_sub_high
, srgb_sub_high
);
3829 shader_addline(buffer
, "tmp1.xyz = %s.xyz * srgb_mul_low.xyz;\n", fragcolor
);
3830 shader_addline(buffer
, "%s.x = %s.x < srgb_comparison.x ? tmp1.x : tmp0.x;\n", fragcolor
, fragcolor
);
3831 shader_addline(buffer
, "%s.y = %s.y < srgb_comparison.y ? tmp1.y : tmp0.y;\n", fragcolor
, fragcolor
);
3832 shader_addline(buffer
, "%s.z = %s.z < srgb_comparison.z ? tmp1.z : tmp0.z;\n", fragcolor
, fragcolor
);
3833 shader_addline(buffer
, "%s = clamp(%s, 0.0, 1.0);\n", fragcolor
, fragcolor
);
3835 /* Pixel shader < 3.0 do not replace the fog stage.
3836 * This implements linear fog computation and blending.
3837 * TODO: non linear fog
3838 * NOTE: gl_Fog.start and gl_Fog.end don't hold fog start s and end e but
3839 * -1/(e-s) and e/(e-s) respectively.
3841 if(reg_maps
->shader_version
< WINED3DPS_VERSION(3,0)) {
3843 case FOG_OFF
: break;
3845 shader_addline(buffer
, "float Fog = clamp(gl_FogFragCoord * gl_Fog.start + gl_Fog.end, 0.0, 1.0);\n");
3846 shader_addline(buffer
, "%s.xyz = mix(gl_Fog.color.xyz, %s.xyz, Fog);\n", fragcolor
, fragcolor
);
3849 FIXME("Implement EXP fog in glsl\n");
3852 FIXME("Implement EXP2 fog in glsl\n");
3857 shader_addline(buffer
, "}\n");
3859 TRACE("Compiling shader object %u\n", shader_obj
);
3860 GL_EXTCALL(glShaderSourceARB(shader_obj
, 1, (const char**)&buffer
->buffer
, NULL
));
3861 GL_EXTCALL(glCompileShaderARB(shader_obj
));
3862 print_glsl_info_log(&GLINFO_LOCATION
, shader_obj
);
3864 /* Store the shader object */
3868 static void shader_glsl_generate_vshader(IWineD3DVertexShader
*iface
, SHADER_BUFFER
*buffer
) {
3869 IWineD3DVertexShaderImpl
*This
= (IWineD3DVertexShaderImpl
*)iface
;
3870 const struct shader_reg_maps
*reg_maps
= &This
->baseShader
.reg_maps
;
3871 CONST DWORD
*function
= This
->baseShader
.function
;
3872 const WineD3D_GL_Info
*gl_info
= &((IWineD3DDeviceImpl
*)This
->baseShader
.device
)->adapter
->gl_info
;
3874 /* Create the hw GLSL shader program and assign it as the shader->prgId */
3875 GLhandleARB shader_obj
= GL_EXTCALL(glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB
));
3877 shader_addline(buffer
, "#version 120\n");
3879 /* Base Declarations */
3880 shader_generate_glsl_declarations( (IWineD3DBaseShader
*) This
, reg_maps
, buffer
, &GLINFO_LOCATION
, NULL
);
3882 /* Base Shader Body */
3883 shader_generate_main( (IWineD3DBaseShader
*) This
, buffer
, reg_maps
, function
);
3885 /* Unpack 3.0 outputs */
3886 if (reg_maps
->shader_version
>= WINED3DVS_VERSION(3,0)) shader_addline(buffer
, "order_ps_input(OUT);\n");
3887 else shader_addline(buffer
, "order_ps_input();\n");
3889 /* If this shader doesn't use fog copy the z coord to the fog coord so that we can use table fog */
3891 shader_addline(buffer
, "gl_FogFragCoord = gl_Position.z;\n");
3893 /* Write the final position.
3895 * OpenGL coordinates specify the center of the pixel while d3d coords specify
3896 * the corner. The offsets are stored in z and w in posFixup. posFixup.y contains
3897 * 1.0 or -1.0 to turn the rendering upside down for offscreen rendering. PosFixup.x
3898 * contains 1.0 to allow a mad.
3900 shader_addline(buffer
, "gl_Position.y = gl_Position.y * posFixup.y;\n");
3901 shader_addline(buffer
, "gl_Position.xy += posFixup.zw * gl_Position.ww;\n");
3903 /* Z coord [0;1]->[-1;1] mapping, see comment in transform_projection in state.c
3905 * Basically we want (in homogeneous coordinates) z = z * 2 - 1. However, shaders are run
3906 * before the homogeneous divide, so we have to take the w into account: z = ((z / w) * 2 - 1) * w,
3907 * which is the same as z = z * 2 - w.
3909 shader_addline(buffer
, "gl_Position.z = gl_Position.z * 2.0 - gl_Position.w;\n");
3911 shader_addline(buffer
, "}\n");
3913 TRACE("Compiling shader object %u\n", shader_obj
);
3914 GL_EXTCALL(glShaderSourceARB(shader_obj
, 1, (const char**)&buffer
->buffer
, NULL
));
3915 GL_EXTCALL(glCompileShaderARB(shader_obj
));
3916 print_glsl_info_log(&GLINFO_LOCATION
, shader_obj
);
3918 /* Store the shader object */
3919 This
->prgId
= shader_obj
;
3922 static void shader_glsl_get_caps(WINED3DDEVTYPE devtype
, const WineD3D_GL_Info
*gl_info
, struct shader_caps
*pCaps
)
3924 /* Nvidia Geforce6/7 or Ati R4xx/R5xx cards with GLSL support, support VS 3.0 but older Nvidia/Ati
3925 * models with GLSL support only support 2.0. In case of nvidia we can detect VS 2.0 support using
3926 * vs_nv_version which is based on NV_vertex_program.
3927 * For Ati cards there's no way using glsl (it abstracts the lowlevel info away) and also not
3928 * using ARB_vertex_program. It is safe to assume that when a card supports pixel shader 2.0 it
3929 * supports vertex shader 2.0 too and the way around. We can detect ps2.0 using the maximum number
3930 * of native instructions, so use that here. For more info see the pixel shader versioning code below.
3932 if((GLINFO_LOCATION
.vs_nv_version
== VS_VERSION_20
) || (GLINFO_LOCATION
.ps_arb_max_instructions
<= 512))
3933 pCaps
->VertexShaderVersion
= WINED3DVS_VERSION(2,0);
3935 pCaps
->VertexShaderVersion
= WINED3DVS_VERSION(3,0);
3936 TRACE_(d3d_caps
)("Hardware vertex shader version %d.%d enabled (GLSL)\n", (pCaps
->VertexShaderVersion
>> 8) & 0xff, pCaps
->VertexShaderVersion
& 0xff);
3937 pCaps
->MaxVertexShaderConst
= GL_LIMITS(vshader_constantsF
);
3939 /* Older DX9-class videocards (GeforceFX / Radeon >9500/X*00) only support pixel shader 2.0/2.0a/2.0b.
3940 * In OpenGL the extensions related to GLSL abstract lowlevel GL info away which is needed
3941 * to distinguish between 2.0 and 3.0 (and 2.0a/2.0b). In case of Nvidia we use their fragment
3942 * program extensions. On other hardware including ATI GL_ARB_fragment_program offers the info
3943 * in max native instructions. Intel and others also offer the info in this extension but they
3944 * don't support GLSL (at least on Windows).
3946 * PS2.0 requires at least 96 instructions, 2.0a/2.0b go up to 512. Assume that if the number
3947 * of instructions is 512 or less we have to do with ps2.0 hardware.
3948 * NOTE: ps3.0 hardware requires 512 or more instructions but ati and nvidia offer 'enough' (1024 vs 4096) on their most basic ps3.0 hardware.
3950 if((GLINFO_LOCATION
.ps_nv_version
== PS_VERSION_20
) || (GLINFO_LOCATION
.ps_arb_max_instructions
<= 512))
3951 pCaps
->PixelShaderVersion
= WINED3DPS_VERSION(2,0);
3953 pCaps
->PixelShaderVersion
= WINED3DPS_VERSION(3,0);
3955 /* FIXME: The following line is card dependent. -8.0 to 8.0 is the
3956 * Direct3D minimum requirement.
3958 * Both GL_ARB_fragment_program and GLSL require a "maximum representable magnitude"
3959 * of colors to be 2^10, and 2^32 for other floats. Should we use 1024 here?
3961 * The problem is that the refrast clamps temporary results in the shader to
3962 * [-MaxValue;+MaxValue]. If the card's max value is bigger than the one we advertize here,
3963 * then applications may miss the clamping behavior. On the other hand, if it is smaller,
3964 * the shader will generate incorrect results too. Unfortunately, GL deliberately doesn't
3965 * offer a way to query this.
3967 pCaps
->PixelShader1xMaxValue
= 8.0;
3968 TRACE_(d3d_caps
)("Hardware pixel shader version %d.%d enabled (GLSL)\n", (pCaps
->PixelShaderVersion
>> 8) & 0xff, pCaps
->PixelShaderVersion
& 0xff);
3971 static BOOL
shader_glsl_color_fixup_supported(struct color_fixup_desc fixup
)
3973 if (TRACE_ON(d3d_shader
) && TRACE_ON(d3d
))
3975 TRACE("Checking support for fixup:\n");
3976 dump_color_fixup_desc(fixup
);
3979 /* We support everything except YUV conversions. */
3980 if (!is_yuv_fixup(fixup
))
3986 TRACE("[FAILED]\n");
3990 static const SHADER_HANDLER shader_glsl_instruction_handler_table
[WINED3DSIH_TABLE_SIZE
] =
3992 /* WINED3DSIH_ABS */ shader_glsl_map2gl
,
3993 /* WINED3DSIH_ADD */ shader_glsl_arith
,
3994 /* WINED3DSIH_BEM */ pshader_glsl_bem
,
3995 /* WINED3DSIH_BREAK */ shader_glsl_break
,
3996 /* WINED3DSIH_BREAKC */ shader_glsl_breakc
,
3997 /* WINED3DSIH_BREAKP */ NULL
,
3998 /* WINED3DSIH_CALL */ shader_glsl_call
,
3999 /* WINED3DSIH_CALLNZ */ shader_glsl_callnz
,
4000 /* WINED3DSIH_CMP */ shader_glsl_cmp
,
4001 /* WINED3DSIH_CND */ shader_glsl_cnd
,
4002 /* WINED3DSIH_CRS */ shader_glsl_cross
,
4003 /* WINED3DSIH_DCL */ NULL
,
4004 /* WINED3DSIH_DEF */ NULL
,
4005 /* WINED3DSIH_DEFB */ NULL
,
4006 /* WINED3DSIH_DEFI */ NULL
,
4007 /* WINED3DSIH_DP2ADD */ pshader_glsl_dp2add
,
4008 /* WINED3DSIH_DP3 */ shader_glsl_dot
,
4009 /* WINED3DSIH_DP4 */ shader_glsl_dot
,
4010 /* WINED3DSIH_DST */ shader_glsl_dst
,
4011 /* WINED3DSIH_DSX */ shader_glsl_map2gl
,
4012 /* WINED3DSIH_DSY */ shader_glsl_map2gl
,
4013 /* WINED3DSIH_ELSE */ shader_glsl_else
,
4014 /* WINED3DSIH_ENDIF */ shader_glsl_end
,
4015 /* WINED3DSIH_ENDLOOP */ shader_glsl_end
,
4016 /* WINED3DSIH_ENDREP */ shader_glsl_end
,
4017 /* WINED3DSIH_EXP */ shader_glsl_map2gl
,
4018 /* WINED3DSIH_EXPP */ shader_glsl_expp
,
4019 /* WINED3DSIH_FRC */ shader_glsl_map2gl
,
4020 /* WINED3DSIH_IF */ shader_glsl_if
,
4021 /* WINED3DSIH_IFC */ shader_glsl_ifc
,
4022 /* WINED3DSIH_LABEL */ shader_glsl_label
,
4023 /* WINED3DSIH_LIT */ shader_glsl_lit
,
4024 /* WINED3DSIH_LOG */ shader_glsl_log
,
4025 /* WINED3DSIH_LOGP */ shader_glsl_log
,
4026 /* WINED3DSIH_LOOP */ shader_glsl_loop
,
4027 /* WINED3DSIH_LRP */ shader_glsl_lrp
,
4028 /* WINED3DSIH_M3x2 */ shader_glsl_mnxn
,
4029 /* WINED3DSIH_M3x3 */ shader_glsl_mnxn
,
4030 /* WINED3DSIH_M3x4 */ shader_glsl_mnxn
,
4031 /* WINED3DSIH_M4x3 */ shader_glsl_mnxn
,
4032 /* WINED3DSIH_M4x4 */ shader_glsl_mnxn
,
4033 /* WINED3DSIH_MAD */ shader_glsl_mad
,
4034 /* WINED3DSIH_MAX */ shader_glsl_map2gl
,
4035 /* WINED3DSIH_MIN */ shader_glsl_map2gl
,
4036 /* WINED3DSIH_MOV */ shader_glsl_mov
,
4037 /* WINED3DSIH_MOVA */ shader_glsl_mov
,
4038 /* WINED3DSIH_MUL */ shader_glsl_arith
,
4039 /* WINED3DSIH_NOP */ NULL
,
4040 /* WINED3DSIH_NRM */ shader_glsl_map2gl
,
4041 /* WINED3DSIH_PHASE */ NULL
,
4042 /* WINED3DSIH_POW */ shader_glsl_pow
,
4043 /* WINED3DSIH_RCP */ shader_glsl_rcp
,
4044 /* WINED3DSIH_REP */ shader_glsl_rep
,
4045 /* WINED3DSIH_RET */ NULL
,
4046 /* WINED3DSIH_RSQ */ shader_glsl_rsq
,
4047 /* WINED3DSIH_SETP */ NULL
,
4048 /* WINED3DSIH_SGE */ shader_glsl_compare
,
4049 /* WINED3DSIH_SGN */ shader_glsl_map2gl
,
4050 /* WINED3DSIH_SINCOS */ shader_glsl_sincos
,
4051 /* WINED3DSIH_SLT */ shader_glsl_compare
,
4052 /* WINED3DSIH_SUB */ shader_glsl_arith
,
4053 /* WINED3DSIH_TEX */ pshader_glsl_tex
,
4054 /* WINED3DSIH_TEXBEM */ pshader_glsl_texbem
,
4055 /* WINED3DSIH_TEXBEML */ pshader_glsl_texbem
,
4056 /* WINED3DSIH_TEXCOORD */ pshader_glsl_texcoord
,
4057 /* WINED3DSIH_TEXDEPTH */ pshader_glsl_texdepth
,
4058 /* WINED3DSIH_TEXDP3 */ pshader_glsl_texdp3
,
4059 /* WINED3DSIH_TEXDP3TEX */ pshader_glsl_texdp3tex
,
4060 /* WINED3DSIH_TEXKILL */ pshader_glsl_texkill
,
4061 /* WINED3DSIH_TEXLDD */ NULL
,
4062 /* WINED3DSIH_TEXLDL */ shader_glsl_texldl
,
4063 /* WINED3DSIH_TEXM3x2DEPTH */ pshader_glsl_texm3x2depth
,
4064 /* WINED3DSIH_TEXM3x2PAD */ pshader_glsl_texm3x2pad
,
4065 /* WINED3DSIH_TEXM3x2TEX */ pshader_glsl_texm3x2tex
,
4066 /* WINED3DSIH_TEXM3x3 */ pshader_glsl_texm3x3
,
4067 /* WINED3DSIH_TEXM3x3DIFF */ NULL
,
4068 /* WINED3DSIH_TEXM3x3PAD */ pshader_glsl_texm3x3pad
,
4069 /* WINED3DSIH_TEXM3x3SPEC */ pshader_glsl_texm3x3spec
,
4070 /* WINED3DSIH_TEXM3x3TEX */ pshader_glsl_texm3x3tex
,
4071 /* WINED3DSIH_TEXM3x3VSPEC */ pshader_glsl_texm3x3vspec
,
4072 /* WINED3DSIH_TEXREG2AR */ pshader_glsl_texreg2ar
,
4073 /* WINED3DSIH_TEXREG2GB */ pshader_glsl_texreg2gb
,
4074 /* WINED3DSIH_TEXREG2RGB */ pshader_glsl_texreg2rgb
,
4077 const shader_backend_t glsl_shader_backend
= {
4078 shader_glsl_instruction_handler_table
,
4080 shader_glsl_select_depth_blt
,
4081 shader_glsl_deselect_depth_blt
,
4082 shader_glsl_update_float_vertex_constants
,
4083 shader_glsl_update_float_pixel_constants
,
4084 shader_glsl_load_constants
,
4085 shader_glsl_color_correction
,
4086 shader_glsl_destroy
,
4089 shader_glsl_dirty_const
,
4090 shader_glsl_generate_pshader
,
4091 shader_glsl_generate_vshader
,
4092 shader_glsl_get_caps
,
4093 shader_glsl_color_fixup_supported
,