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))
801 shader_addline(buffer
, "varying vec4 IN[%u];\n", GL_LIMITS(glsl_varyings
) / 4);
803 /* TODO: Write a replacement shader for the fixed function vertex pipeline, so this isn't needed.
804 * For fixed function vertex processing + 3.0 pixel shader we need a separate function in the
805 * pixel shader that reads the fixed function color into the packed input registers.
807 shader_addline(buffer
, "vec4 IN[%u];\n", GL_LIMITS(glsl_varyings
) / 4);
811 /* Declare output register temporaries */
812 if(This
->baseShader
.limits
.packed_output
) {
813 shader_addline(buffer
, "vec4 OUT[%u];\n", This
->baseShader
.limits
.packed_output
);
816 /* Declare temporary variables */
817 for(i
= 0; i
< This
->baseShader
.limits
.temporary
; i
++) {
818 if (reg_maps
->temporary
[i
])
819 shader_addline(buffer
, "vec4 R%u;\n", i
);
822 /* Declare attributes */
823 for (i
= 0; i
< This
->baseShader
.limits
.attributes
; i
++) {
824 if (reg_maps
->attributes
[i
])
825 shader_addline(buffer
, "attribute vec4 attrib%i;\n", i
);
828 /* Declare loop registers aLx */
829 for (i
= 0; i
< reg_maps
->loop_depth
; i
++) {
830 shader_addline(buffer
, "int aL%u;\n", i
);
831 shader_addline(buffer
, "int tmpInt%u;\n", i
);
834 /* Temporary variables for matrix operations */
835 shader_addline(buffer
, "vec4 tmp0;\n");
836 shader_addline(buffer
, "vec4 tmp1;\n");
838 /* Local constants use a different name so they can be loaded once at shader link time
839 * They can't be hardcoded into the shader text via LC = {x, y, z, w}; because the
840 * float -> string conversion can cause precision loss.
842 if(!This
->baseShader
.load_local_constsF
) {
843 LIST_FOR_EACH_ENTRY(lconst
, &This
->baseShader
.constantsF
, local_constant
, entry
) {
844 shader_addline(buffer
, "uniform vec4 %cLC%u;\n", prefix
, lconst
->idx
);
848 /* Start the main program */
849 shader_addline(buffer
, "void main() {\n");
850 if(pshader
&& reg_maps
->vpos
) {
851 /* DirectX apps expect integer values, while OpenGL drivers add approximately 0.5. This causes
852 * off-by-one problems as spotted by the vPos d3d9 visual test. Unfortunately the ATI cards do
853 * not add exactly 0.5, but rather something like 0.49999999 or 0.50000001, which still causes
854 * precision troubles when we just substract 0.5.
856 * To deal with that just floor() the position. This will eliminate the fraction on all cards.
858 * TODO: Test how that behaves with multisampling once we can enable multisampling in winex11.
860 * An advantage of floor is that it works even if the driver doesn't add 1/2. It is somewhat
861 * questionable if 1.5, 2.5, ... are the proper values to return in gl_FragCoord, even though
862 * coordinates specify the pixel centers instead of the pixel corners. This code will behave
863 * correctly on drivers that returns integer values.
865 shader_addline(buffer
, "vpos = floor(vec4(0, ycorrection[0], 0, 0) + gl_FragCoord * vec4(1, ycorrection[1], 1, 1));\n");
869 /*****************************************************************************
870 * Functions to generate GLSL strings from DirectX Shader bytecode begin here.
872 * For more information, see http://wiki.winehq.org/DirectX-Shaders
873 ****************************************************************************/
876 static void shader_glsl_add_src_param(const SHADER_OPCODE_ARG
*arg
, const DWORD param
,
877 const DWORD addr_token
, DWORD mask
, glsl_src_param_t
*src_param
);
879 /** Used for opcode modifiers - They multiply the result by the specified amount */
880 static const char * const shift_glsl_tab
[] = {
882 "2.0 * ", /* 1 (x2) */
883 "4.0 * ", /* 2 (x4) */
884 "8.0 * ", /* 3 (x8) */
885 "16.0 * ", /* 4 (x16) */
886 "32.0 * ", /* 5 (x32) */
893 "0.0625 * ", /* 12 (d16) */
894 "0.125 * ", /* 13 (d8) */
895 "0.25 * ", /* 14 (d4) */
896 "0.5 * " /* 15 (d2) */
899 /* Generate a GLSL parameter that does the input modifier computation and return the input register/mask to use */
900 static void shader_glsl_gen_modifier (
903 const char *in_regswizzle
,
908 if (instr
== WINED3DSIO_TEXKILL
)
911 switch (instr
& WINED3DSP_SRCMOD_MASK
) {
912 case WINED3DSPSM_DZ
: /* Need to handle this in the instructions itself (texld & texcrd). */
914 case WINED3DSPSM_NONE
:
915 sprintf(out_str
, "%s%s", in_reg
, in_regswizzle
);
917 case WINED3DSPSM_NEG
:
918 sprintf(out_str
, "-%s%s", in_reg
, in_regswizzle
);
920 case WINED3DSPSM_NOT
:
921 sprintf(out_str
, "!%s%s", in_reg
, in_regswizzle
);
923 case WINED3DSPSM_BIAS
:
924 sprintf(out_str
, "(%s%s - vec4(0.5)%s)", in_reg
, in_regswizzle
, in_regswizzle
);
926 case WINED3DSPSM_BIASNEG
:
927 sprintf(out_str
, "-(%s%s - vec4(0.5)%s)", in_reg
, in_regswizzle
, in_regswizzle
);
929 case WINED3DSPSM_SIGN
:
930 sprintf(out_str
, "(2.0 * (%s%s - 0.5))", in_reg
, in_regswizzle
);
932 case WINED3DSPSM_SIGNNEG
:
933 sprintf(out_str
, "-(2.0 * (%s%s - 0.5))", in_reg
, in_regswizzle
);
935 case WINED3DSPSM_COMP
:
936 sprintf(out_str
, "(1.0 - %s%s)", in_reg
, in_regswizzle
);
939 sprintf(out_str
, "(2.0 * %s%s)", in_reg
, in_regswizzle
);
941 case WINED3DSPSM_X2NEG
:
942 sprintf(out_str
, "-(2.0 * %s%s)", in_reg
, in_regswizzle
);
944 case WINED3DSPSM_ABS
:
945 sprintf(out_str
, "abs(%s%s)", in_reg
, in_regswizzle
);
947 case WINED3DSPSM_ABSNEG
:
948 sprintf(out_str
, "-abs(%s%s)", in_reg
, in_regswizzle
);
951 FIXME("Unhandled modifier %u\n", (instr
& WINED3DSP_SRCMOD_MASK
));
952 sprintf(out_str
, "%s%s", in_reg
, in_regswizzle
);
956 /** Writes the GLSL variable name that corresponds to the register that the
957 * DX opcode parameter is trying to access */
958 static void shader_glsl_get_register_name(const DWORD param
, const DWORD addr_token
,
959 char *regstr
, BOOL
*is_color
, const SHADER_OPCODE_ARG
*arg
)
961 /* oPos, oFog and oPts in D3D */
962 static const char * const hwrastout_reg_names
[] = { "gl_Position", "gl_FogFragCoord", "gl_PointSize" };
964 DWORD reg
= param
& WINED3DSP_REGNUM_MASK
;
965 DWORD regtype
= shader_get_regtype(param
);
966 IWineD3DBaseShaderImpl
* This
= (IWineD3DBaseShaderImpl
*) arg
->shader
;
967 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
968 const WineD3D_GL_Info
* gl_info
= &deviceImpl
->adapter
->gl_info
;
969 DWORD shader_version
= This
->baseShader
.reg_maps
.shader_version
;
970 char pshader
= shader_is_pshader_version(shader_version
);
976 case WINED3DSPR_TEMP
:
977 sprintf(tmpStr
, "R%u", reg
);
979 case WINED3DSPR_INPUT
:
981 /* Pixel shaders >= 3.0 */
982 if (WINED3DSHADER_VERSION_MAJOR(shader_version
) >= 3)
984 DWORD in_count
= GL_LIMITS(glsl_varyings
) / 4;
986 if (param
& WINED3DSHADER_ADDRMODE_RELATIVE
) {
987 glsl_src_param_t rel_param
;
988 shader_glsl_add_src_param(arg
, addr_token
, 0, WINED3DSP_WRITEMASK_0
, &rel_param
);
990 /* Removing a + 0 would be an obvious optimization, but macos doesn't see the NOP
993 if(((IWineD3DPixelShaderImpl
*) This
)->input_reg_map
[reg
]) {
994 if (((IWineD3DPixelShaderImpl
*)This
)->declared_in_count
> in_count
) {
995 sprintf(tmpStr
, "((%s + %u) > %d ? (%s + %u) > %d ? gl_SecondaryColor : gl_Color : IN[%s + %u])",
996 rel_param
.param_str
, ((IWineD3DPixelShaderImpl
*)This
)->input_reg_map
[reg
], in_count
- 1,
997 rel_param
.param_str
, ((IWineD3DPixelShaderImpl
*)This
)->input_reg_map
[reg
], in_count
,
998 rel_param
.param_str
, ((IWineD3DPixelShaderImpl
*)This
)->input_reg_map
[reg
]);
1000 sprintf(tmpStr
, "IN[%s + %u]", rel_param
.param_str
, ((IWineD3DPixelShaderImpl
*)This
)->input_reg_map
[reg
]);
1003 if (((IWineD3DPixelShaderImpl
*)This
)->declared_in_count
> in_count
) {
1004 sprintf(tmpStr
, "((%s) > %d ? (%s) > %d ? gl_SecondaryColor : gl_Color : IN[%s])",
1005 rel_param
.param_str
, in_count
- 1,
1006 rel_param
.param_str
, in_count
,
1007 rel_param
.param_str
);
1009 sprintf(tmpStr
, "IN[%s]", rel_param
.param_str
);
1013 DWORD idx
= ((IWineD3DPixelShaderImpl
*) This
)->input_reg_map
[reg
];
1014 if (idx
== in_count
) {
1015 sprintf(tmpStr
, "gl_Color");
1016 } else if (idx
== in_count
+ 1) {
1017 sprintf(tmpStr
, "gl_SecondaryColor");
1019 sprintf(tmpStr
, "IN[%u]", idx
);
1024 strcpy(tmpStr
, "gl_Color");
1026 strcpy(tmpStr
, "gl_SecondaryColor");
1029 if (vshader_input_is_color((IWineD3DVertexShader
*) This
, reg
))
1031 sprintf(tmpStr
, "attrib%u", reg
);
1034 case WINED3DSPR_CONST
:
1036 const char prefix
= pshader
? 'P':'V';
1038 /* Relative addressing */
1039 if (param
& WINED3DSHADER_ADDRMODE_RELATIVE
) {
1041 /* Relative addressing on shaders 2.0+ have a relative address token,
1042 * prior to that, it was hard-coded as "A0.x" because there's only 1 register */
1043 if (WINED3DSHADER_VERSION_MAJOR(shader_version
) >= 2)
1045 glsl_src_param_t rel_param
;
1046 shader_glsl_add_src_param(arg
, addr_token
, 0, WINED3DSP_WRITEMASK_0
, &rel_param
);
1048 sprintf(tmpStr
, "%cC[%s + %u]", prefix
, rel_param
.param_str
, reg
);
1050 sprintf(tmpStr
, "%cC[%s]", prefix
, rel_param
.param_str
);
1054 sprintf(tmpStr
, "%cC[A0.x + %u]", prefix
, reg
);
1056 sprintf(tmpStr
, "%cC[A0.x]", prefix
);
1061 if(shader_constant_is_local(This
, reg
)) {
1062 sprintf(tmpStr
, "%cLC%u", prefix
, reg
);
1064 sprintf(tmpStr
, "%cC[%u]", prefix
, reg
);
1070 case WINED3DSPR_CONSTINT
:
1072 sprintf(tmpStr
, "PI[%u]", reg
);
1074 sprintf(tmpStr
, "VI[%u]", reg
);
1076 case WINED3DSPR_CONSTBOOL
:
1078 sprintf(tmpStr
, "PB[%u]", reg
);
1080 sprintf(tmpStr
, "VB[%u]", reg
);
1082 case WINED3DSPR_TEXTURE
: /* case WINED3DSPR_ADDR: */
1084 sprintf(tmpStr
, "T%u", reg
);
1086 sprintf(tmpStr
, "A%u", reg
);
1089 case WINED3DSPR_LOOP
:
1090 sprintf(tmpStr
, "aL%u", This
->baseShader
.cur_loop_regno
- 1);
1092 case WINED3DSPR_SAMPLER
:
1094 sprintf(tmpStr
, "Psampler%u", reg
);
1096 sprintf(tmpStr
, "Vsampler%u", reg
);
1098 case WINED3DSPR_COLOROUT
:
1099 if (reg
>= GL_LIMITS(buffers
)) {
1100 WARN("Write to render target %u, only %d supported\n", reg
, 4);
1102 if (GL_SUPPORT(ARB_DRAW_BUFFERS
)) {
1103 sprintf(tmpStr
, "gl_FragData[%u]", reg
);
1104 } else { /* On older cards with GLSL support like the GeforceFX there's only one buffer. */
1105 sprintf(tmpStr
, "gl_FragColor");
1108 case WINED3DSPR_RASTOUT
:
1109 sprintf(tmpStr
, "%s", hwrastout_reg_names
[reg
]);
1111 case WINED3DSPR_DEPTHOUT
:
1112 sprintf(tmpStr
, "gl_FragDepth");
1114 case WINED3DSPR_ATTROUT
:
1116 sprintf(tmpStr
, "gl_FrontColor");
1118 sprintf(tmpStr
, "gl_FrontSecondaryColor");
1121 case WINED3DSPR_TEXCRDOUT
:
1122 /* Vertex shaders >= 3.0: WINED3DSPR_OUTPUT */
1123 if (WINED3DSHADER_VERSION_MAJOR(shader_version
) >= 3) sprintf(tmpStr
, "OUT[%u]", reg
);
1124 else sprintf(tmpStr
, "gl_TexCoord[%u]", reg
);
1126 case WINED3DSPR_MISCTYPE
:
1129 sprintf(tmpStr
, "vpos");
1130 } else if (reg
== 1){
1131 /* Note that gl_FrontFacing is a bool, while vFace is
1132 * a float for which the sign determines front/back
1134 sprintf(tmpStr
, "(gl_FrontFacing ? 1.0 : -1.0)");
1136 FIXME("Unhandled misctype register %d\n", reg
);
1137 sprintf(tmpStr
, "unrecognized_register");
1141 FIXME("Unhandled register name Type(%d)\n", regtype
);
1142 sprintf(tmpStr
, "unrecognized_register");
1146 strcat(regstr
, tmpStr
);
1149 /* Get the GLSL write mask for the destination register */
1150 static DWORD
shader_glsl_get_write_mask(const DWORD param
, char *write_mask
) {
1151 char *ptr
= write_mask
;
1152 DWORD mask
= param
& WINED3DSP_WRITEMASK_ALL
;
1154 if (shader_is_scalar(param
)) {
1155 mask
= WINED3DSP_WRITEMASK_0
;
1158 if (param
& WINED3DSP_WRITEMASK_0
) *ptr
++ = 'x';
1159 if (param
& WINED3DSP_WRITEMASK_1
) *ptr
++ = 'y';
1160 if (param
& WINED3DSP_WRITEMASK_2
) *ptr
++ = 'z';
1161 if (param
& WINED3DSP_WRITEMASK_3
) *ptr
++ = 'w';
1169 static unsigned int shader_glsl_get_write_mask_size(DWORD write_mask
) {
1170 unsigned int size
= 0;
1172 if (write_mask
& WINED3DSP_WRITEMASK_0
) ++size
;
1173 if (write_mask
& WINED3DSP_WRITEMASK_1
) ++size
;
1174 if (write_mask
& WINED3DSP_WRITEMASK_2
) ++size
;
1175 if (write_mask
& WINED3DSP_WRITEMASK_3
) ++size
;
1180 static void shader_glsl_get_swizzle(const DWORD param
, BOOL fixup
, DWORD mask
, char *swizzle_str
) {
1181 /* For registers of type WINED3DDECLTYPE_D3DCOLOR, data is stored as "bgra",
1182 * but addressed as "rgba". To fix this we need to swap the register's x
1183 * and z components. */
1184 DWORD swizzle
= (param
& WINED3DSP_SWIZZLE_MASK
) >> WINED3DSP_SWIZZLE_SHIFT
;
1185 const char *swizzle_chars
= fixup
? "zyxw" : "xyzw";
1186 char *ptr
= swizzle_str
;
1188 if (!shader_is_scalar(param
)) {
1190 /* swizzle bits fields: wwzzyyxx */
1191 if (mask
& WINED3DSP_WRITEMASK_0
) *ptr
++ = swizzle_chars
[swizzle
& 0x03];
1192 if (mask
& WINED3DSP_WRITEMASK_1
) *ptr
++ = swizzle_chars
[(swizzle
>> 2) & 0x03];
1193 if (mask
& WINED3DSP_WRITEMASK_2
) *ptr
++ = swizzle_chars
[(swizzle
>> 4) & 0x03];
1194 if (mask
& WINED3DSP_WRITEMASK_3
) *ptr
++ = swizzle_chars
[(swizzle
>> 6) & 0x03];
1200 /* From a given parameter token, generate the corresponding GLSL string.
1201 * Also, return the actual register name and swizzle in case the
1202 * caller needs this information as well. */
1203 static void shader_glsl_add_src_param(const SHADER_OPCODE_ARG
*arg
, const DWORD param
,
1204 const DWORD addr_token
, DWORD mask
, glsl_src_param_t
*src_param
)
1206 BOOL is_color
= FALSE
;
1207 char swizzle_str
[6];
1209 src_param
->reg_name
[0] = '\0';
1210 src_param
->param_str
[0] = '\0';
1211 swizzle_str
[0] = '\0';
1213 shader_glsl_get_register_name(param
, addr_token
, src_param
->reg_name
, &is_color
, arg
);
1215 shader_glsl_get_swizzle(param
, is_color
, mask
, swizzle_str
);
1216 shader_glsl_gen_modifier(param
, src_param
->reg_name
, swizzle_str
, src_param
->param_str
);
1219 /* From a given parameter token, generate the corresponding GLSL string.
1220 * Also, return the actual register name and swizzle in case the
1221 * caller needs this information as well. */
1222 static DWORD
shader_glsl_add_dst_param(const SHADER_OPCODE_ARG
* arg
, const DWORD param
,
1223 const DWORD addr_token
, glsl_dst_param_t
*dst_param
)
1225 BOOL is_color
= FALSE
;
1227 dst_param
->mask_str
[0] = '\0';
1228 dst_param
->reg_name
[0] = '\0';
1230 shader_glsl_get_register_name(param
, addr_token
, dst_param
->reg_name
, &is_color
, arg
);
1231 return shader_glsl_get_write_mask(param
, dst_param
->mask_str
);
1234 /* Append the destination part of the instruction to the buffer, return the effective write mask */
1235 static DWORD
shader_glsl_append_dst_ext(SHADER_BUFFER
*buffer
, const SHADER_OPCODE_ARG
*arg
, const DWORD param
)
1237 glsl_dst_param_t dst_param
;
1241 mask
= shader_glsl_add_dst_param(arg
, param
, arg
->dst_addr
, &dst_param
);
1244 shift
= (param
& WINED3DSP_DSTSHIFT_MASK
) >> WINED3DSP_DSTSHIFT_SHIFT
;
1245 shader_addline(buffer
, "%s%s = %s(", dst_param
.reg_name
, dst_param
.mask_str
, shift_glsl_tab
[shift
]);
1251 /* Append the destination part of the instruction to the buffer, return the effective write mask */
1252 static DWORD
shader_glsl_append_dst(SHADER_BUFFER
*buffer
, const SHADER_OPCODE_ARG
*arg
)
1254 return shader_glsl_append_dst_ext(buffer
, arg
, arg
->dst
);
1257 /** Process GLSL instruction modifiers */
1258 void shader_glsl_add_instruction_modifiers(const SHADER_OPCODE_ARG
* arg
)
1260 DWORD mask
= arg
->dst
& WINED3DSP_DSTMOD_MASK
;
1262 if (arg
->opcode
->dst_token
&& mask
!= 0) {
1263 glsl_dst_param_t dst_param
;
1265 shader_glsl_add_dst_param(arg
, arg
->dst
, 0, &dst_param
);
1267 if (mask
& WINED3DSPDM_SATURATE
) {
1268 /* _SAT means to clamp the value of the register to between 0 and 1 */
1269 shader_addline(arg
->buffer
, "%s%s = clamp(%s%s, 0.0, 1.0);\n", dst_param
.reg_name
,
1270 dst_param
.mask_str
, dst_param
.reg_name
, dst_param
.mask_str
);
1272 if (mask
& WINED3DSPDM_MSAMPCENTROID
) {
1273 FIXME("_centroid modifier not handled\n");
1275 if (mask
& WINED3DSPDM_PARTIALPRECISION
) {
1276 /* MSDN says this modifier can be safely ignored, so that's what we'll do. */
1281 static inline const char* shader_get_comp_op(
1282 const DWORD opcode
) {
1284 DWORD op
= (opcode
& INST_CONTROLS_MASK
) >> INST_CONTROLS_SHIFT
;
1286 case COMPARISON_GT
: return ">";
1287 case COMPARISON_EQ
: return "==";
1288 case COMPARISON_GE
: return ">=";
1289 case COMPARISON_LT
: return "<";
1290 case COMPARISON_NE
: return "!=";
1291 case COMPARISON_LE
: return "<=";
1293 FIXME("Unrecognized comparison value: %u\n", op
);
1298 static void shader_glsl_get_sample_function(DWORD sampler_type
, BOOL projected
, BOOL texrect
, glsl_sample_function_t
*sample_function
) {
1299 /* Note that there's no such thing as a projected cube texture. */
1300 switch(sampler_type
) {
1302 sample_function
->name
= projected
? "texture1DProj" : "texture1D";
1303 sample_function
->coord_mask
= WINED3DSP_WRITEMASK_0
;
1307 sample_function
->name
= projected
? "texture2DRectProj" : "texture2DRect";
1309 sample_function
->name
= projected
? "texture2DProj" : "texture2D";
1311 sample_function
->coord_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
;
1313 case WINED3DSTT_CUBE
:
1314 sample_function
->name
= "textureCube";
1315 sample_function
->coord_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
1317 case WINED3DSTT_VOLUME
:
1318 sample_function
->name
= projected
? "texture3DProj" : "texture3D";
1319 sample_function
->coord_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
1322 sample_function
->name
= "";
1323 sample_function
->coord_mask
= 0;
1324 FIXME("Unrecognized sampler type: %#x;\n", sampler_type
);
1329 static void shader_glsl_append_fixup_arg(char *arguments
, const char *reg_name
,
1330 BOOL sign_fixup
, enum fixup_channel_source channel_source
)
1332 switch(channel_source
)
1334 case CHANNEL_SOURCE_ZERO
:
1335 strcat(arguments
, "0.0");
1338 case CHANNEL_SOURCE_ONE
:
1339 strcat(arguments
, "1.0");
1342 case CHANNEL_SOURCE_X
:
1343 strcat(arguments
, reg_name
);
1344 strcat(arguments
, ".x");
1347 case CHANNEL_SOURCE_Y
:
1348 strcat(arguments
, reg_name
);
1349 strcat(arguments
, ".y");
1352 case CHANNEL_SOURCE_Z
:
1353 strcat(arguments
, reg_name
);
1354 strcat(arguments
, ".z");
1357 case CHANNEL_SOURCE_W
:
1358 strcat(arguments
, reg_name
);
1359 strcat(arguments
, ".w");
1363 FIXME("Unhandled channel source %#x\n", channel_source
);
1364 strcat(arguments
, "undefined");
1368 if (sign_fixup
) strcat(arguments
, " * 2.0 - 1.0");
1371 static void shader_glsl_color_correction(const struct SHADER_OPCODE_ARG
*arg
, struct color_fixup_desc fixup
)
1373 unsigned int mask_size
, remaining
;
1374 glsl_dst_param_t dst_param
;
1375 char arguments
[256];
1380 if (fixup
.x_sign_fixup
|| fixup
.x_source
!= CHANNEL_SOURCE_X
) mask
|= WINED3DSP_WRITEMASK_0
;
1381 if (fixup
.y_sign_fixup
|| fixup
.y_source
!= CHANNEL_SOURCE_Y
) mask
|= WINED3DSP_WRITEMASK_1
;
1382 if (fixup
.z_sign_fixup
|| fixup
.z_source
!= CHANNEL_SOURCE_Z
) mask
|= WINED3DSP_WRITEMASK_2
;
1383 if (fixup
.w_sign_fixup
|| fixup
.w_source
!= CHANNEL_SOURCE_W
) mask
|= WINED3DSP_WRITEMASK_3
;
1386 if (!mask
) return; /* Nothing to do */
1388 if (is_yuv_fixup(fixup
))
1390 enum yuv_fixup yuv_fixup
= get_yuv_fixup(fixup
);
1391 FIXME("YUV fixup (%#x) not supported\n", yuv_fixup
);
1395 mask_size
= shader_glsl_get_write_mask_size(mask
);
1397 dst_param
.mask_str
[0] = '\0';
1398 shader_glsl_get_write_mask(mask
, dst_param
.mask_str
);
1400 dst_param
.reg_name
[0] = '\0';
1401 shader_glsl_get_register_name(arg
->dst
, arg
->dst_addr
, dst_param
.reg_name
, &dummy
, arg
);
1403 arguments
[0] = '\0';
1404 remaining
= mask_size
;
1405 if (mask
& WINED3DSP_WRITEMASK_0
)
1407 shader_glsl_append_fixup_arg(arguments
, dst_param
.reg_name
, fixup
.x_sign_fixup
, fixup
.x_source
);
1408 if (--remaining
) strcat(arguments
, ", ");
1410 if (mask
& WINED3DSP_WRITEMASK_1
)
1412 shader_glsl_append_fixup_arg(arguments
, dst_param
.reg_name
, fixup
.y_sign_fixup
, fixup
.y_source
);
1413 if (--remaining
) strcat(arguments
, ", ");
1415 if (mask
& WINED3DSP_WRITEMASK_2
)
1417 shader_glsl_append_fixup_arg(arguments
, dst_param
.reg_name
, fixup
.z_sign_fixup
, fixup
.z_source
);
1418 if (--remaining
) strcat(arguments
, ", ");
1420 if (mask
& WINED3DSP_WRITEMASK_3
)
1422 shader_glsl_append_fixup_arg(arguments
, dst_param
.reg_name
, fixup
.w_sign_fixup
, fixup
.w_source
);
1423 if (--remaining
) strcat(arguments
, ", ");
1428 shader_addline(arg
->buffer
, "%s%s = vec%u(%s);\n",
1429 dst_param
.reg_name
, dst_param
.mask_str
, mask_size
, arguments
);
1433 shader_addline(arg
->buffer
, "%s%s = %s;\n", dst_param
.reg_name
, dst_param
.mask_str
, arguments
);
1437 /*****************************************************************************
1439 * Begin processing individual instruction opcodes
1441 ****************************************************************************/
1443 /* Generate GLSL arithmetic functions (dst = src1 + src2) */
1444 static void shader_glsl_arith(const SHADER_OPCODE_ARG
*arg
)
1446 CONST SHADER_OPCODE
* curOpcode
= arg
->opcode
;
1447 SHADER_BUFFER
* buffer
= arg
->buffer
;
1448 glsl_src_param_t src0_param
;
1449 glsl_src_param_t src1_param
;
1453 /* Determine the GLSL operator to use based on the opcode */
1454 switch (curOpcode
->opcode
) {
1455 case WINED3DSIO_MUL
: op
= '*'; break;
1456 case WINED3DSIO_ADD
: op
= '+'; break;
1457 case WINED3DSIO_SUB
: op
= '-'; break;
1460 FIXME("Opcode %s not yet handled in GLSL\n", curOpcode
->name
);
1464 write_mask
= shader_glsl_append_dst(buffer
, arg
);
1465 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], write_mask
, &src0_param
);
1466 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], write_mask
, &src1_param
);
1467 shader_addline(buffer
, "%s %c %s);\n", src0_param
.param_str
, op
, src1_param
.param_str
);
1470 /* Process the WINED3DSIO_MOV opcode using GLSL (dst = src) */
1471 static void shader_glsl_mov(const SHADER_OPCODE_ARG
*arg
)
1473 SHADER_BUFFER
* buffer
= arg
->buffer
;
1474 glsl_src_param_t src0_param
;
1477 write_mask
= shader_glsl_append_dst(buffer
, arg
);
1478 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], write_mask
, &src0_param
);
1480 /* In vs_1_1 WINED3DSIO_MOV can write to the address register. In later
1481 * shader versions WINED3DSIO_MOVA is used for this. */
1482 if ((WINED3DSHADER_VERSION_MAJOR(arg
->reg_maps
->shader_version
) == 1
1483 && !shader_is_pshader_version(arg
->reg_maps
->shader_version
)
1484 && shader_get_regtype(arg
->dst
) == WINED3DSPR_ADDR
))
1486 /* This is a simple floor() */
1487 unsigned int mask_size
= shader_glsl_get_write_mask_size(write_mask
);
1488 if (mask_size
> 1) {
1489 shader_addline(buffer
, "ivec%d(floor(%s)));\n", mask_size
, src0_param
.param_str
);
1491 shader_addline(buffer
, "int(floor(%s)));\n", src0_param
.param_str
);
1493 } else if(arg
->opcode
->opcode
== WINED3DSIO_MOVA
) {
1494 /* We need to *round* to the nearest int here. */
1495 unsigned int mask_size
= shader_glsl_get_write_mask_size(write_mask
);
1496 if (mask_size
> 1) {
1497 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
);
1499 shader_addline(buffer
, "int(floor(abs(%s) + 0.5) * sign(%s)));\n", src0_param
.param_str
, src0_param
.param_str
);
1502 shader_addline(buffer
, "%s);\n", src0_param
.param_str
);
1506 /* Process the dot product operators DP3 and DP4 in GLSL (dst = dot(src0, src1)) */
1507 static void shader_glsl_dot(const SHADER_OPCODE_ARG
*arg
)
1509 CONST SHADER_OPCODE
* curOpcode
= arg
->opcode
;
1510 SHADER_BUFFER
* buffer
= arg
->buffer
;
1511 glsl_src_param_t src0_param
;
1512 glsl_src_param_t src1_param
;
1513 DWORD dst_write_mask
, src_write_mask
;
1514 unsigned int dst_size
= 0;
1516 dst_write_mask
= shader_glsl_append_dst(buffer
, arg
);
1517 dst_size
= shader_glsl_get_write_mask_size(dst_write_mask
);
1519 /* dp3 works on vec3, dp4 on vec4 */
1520 if (curOpcode
->opcode
== WINED3DSIO_DP4
) {
1521 src_write_mask
= WINED3DSP_WRITEMASK_ALL
;
1523 src_write_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
1526 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], src_write_mask
, &src0_param
);
1527 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], src_write_mask
, &src1_param
);
1530 shader_addline(buffer
, "vec%d(dot(%s, %s)));\n", dst_size
, src0_param
.param_str
, src1_param
.param_str
);
1532 shader_addline(buffer
, "dot(%s, %s));\n", src0_param
.param_str
, src1_param
.param_str
);
1536 /* Note that this instruction has some restrictions. The destination write mask
1537 * can't contain the w component, and the source swizzles have to be .xyzw */
1538 static void shader_glsl_cross(const SHADER_OPCODE_ARG
*arg
)
1540 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
1541 glsl_src_param_t src0_param
;
1542 glsl_src_param_t src1_param
;
1545 shader_glsl_get_write_mask(arg
->dst
, dst_mask
);
1546 shader_glsl_append_dst(arg
->buffer
, arg
);
1547 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], src_mask
, &src0_param
);
1548 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], src_mask
, &src1_param
);
1549 shader_addline(arg
->buffer
, "cross(%s, %s)%s);\n", src0_param
.param_str
, src1_param
.param_str
, dst_mask
);
1552 /* Process the WINED3DSIO_POW instruction in GLSL (dst = |src0|^src1)
1553 * Src0 and src1 are scalars. Note that D3D uses the absolute of src0, while
1554 * GLSL uses the value as-is. */
1555 static void shader_glsl_pow(const SHADER_OPCODE_ARG
*arg
)
1557 SHADER_BUFFER
*buffer
= arg
->buffer
;
1558 glsl_src_param_t src0_param
;
1559 glsl_src_param_t src1_param
;
1560 DWORD dst_write_mask
;
1561 unsigned int dst_size
;
1563 dst_write_mask
= shader_glsl_append_dst(buffer
, arg
);
1564 dst_size
= shader_glsl_get_write_mask_size(dst_write_mask
);
1566 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
1567 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], WINED3DSP_WRITEMASK_0
, &src1_param
);
1570 shader_addline(buffer
, "vec%d(pow(abs(%s), %s)));\n", dst_size
, src0_param
.param_str
, src1_param
.param_str
);
1572 shader_addline(buffer
, "pow(abs(%s), %s));\n", src0_param
.param_str
, src1_param
.param_str
);
1576 /* Process the WINED3DSIO_LOG instruction in GLSL (dst = log2(|src0|))
1577 * Src0 is a scalar. Note that D3D uses the absolute of src0, while
1578 * GLSL uses the value as-is. */
1579 static void shader_glsl_log(const SHADER_OPCODE_ARG
*arg
)
1581 SHADER_BUFFER
*buffer
= arg
->buffer
;
1582 glsl_src_param_t src0_param
;
1583 DWORD dst_write_mask
;
1584 unsigned int dst_size
;
1586 dst_write_mask
= shader_glsl_append_dst(buffer
, arg
);
1587 dst_size
= shader_glsl_get_write_mask_size(dst_write_mask
);
1589 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
1592 shader_addline(buffer
, "vec%d(log2(abs(%s))));\n", dst_size
, src0_param
.param_str
);
1594 shader_addline(buffer
, "log2(abs(%s)));\n", src0_param
.param_str
);
1598 /* Map the opcode 1-to-1 to the GL code (arg->dst = instruction(src0, src1, ...) */
1599 static void shader_glsl_map2gl(const SHADER_OPCODE_ARG
*arg
)
1601 CONST SHADER_OPCODE
* curOpcode
= arg
->opcode
;
1602 SHADER_BUFFER
* buffer
= arg
->buffer
;
1603 glsl_src_param_t src_param
;
1604 const char *instruction
;
1605 char arguments
[256];
1609 /* Determine the GLSL function to use based on the opcode */
1610 /* TODO: Possibly make this a table for faster lookups */
1611 switch (curOpcode
->opcode
) {
1612 case WINED3DSIO_MIN
: instruction
= "min"; break;
1613 case WINED3DSIO_MAX
: instruction
= "max"; break;
1614 case WINED3DSIO_ABS
: instruction
= "abs"; break;
1615 case WINED3DSIO_FRC
: instruction
= "fract"; break;
1616 case WINED3DSIO_NRM
: instruction
= "normalize"; break;
1617 case WINED3DSIO_EXP
: instruction
= "exp2"; break;
1618 case WINED3DSIO_SGN
: instruction
= "sign"; break;
1619 case WINED3DSIO_DSX
: instruction
= "dFdx"; break;
1620 case WINED3DSIO_DSY
: instruction
= "ycorrection.y * dFdy"; break;
1621 default: instruction
= "";
1622 FIXME("Opcode %s not yet handled in GLSL\n", curOpcode
->name
);
1626 write_mask
= shader_glsl_append_dst(buffer
, arg
);
1628 arguments
[0] = '\0';
1629 if (curOpcode
->num_params
> 0) {
1630 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], write_mask
, &src_param
);
1631 strcat(arguments
, src_param
.param_str
);
1632 for (i
= 2; i
< curOpcode
->num_params
; ++i
) {
1633 strcat(arguments
, ", ");
1634 shader_glsl_add_src_param(arg
, arg
->src
[i
-1], arg
->src_addr
[i
-1], write_mask
, &src_param
);
1635 strcat(arguments
, src_param
.param_str
);
1639 shader_addline(buffer
, "%s(%s));\n", instruction
, arguments
);
1642 /** Process the WINED3DSIO_EXPP instruction in GLSL:
1643 * For shader model 1.x, do the following (and honor the writemask, so use a temporary variable):
1644 * dst.x = 2^(floor(src))
1645 * dst.y = src - floor(src)
1646 * dst.z = 2^src (partial precision is allowed, but optional)
1648 * For 2.0 shaders, just do this (honoring writemask and swizzle):
1649 * dst = 2^src; (partial precision is allowed, but optional)
1651 static void shader_glsl_expp(const SHADER_OPCODE_ARG
*arg
)
1653 glsl_src_param_t src_param
;
1655 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src_param
);
1657 if (arg
->reg_maps
->shader_version
< WINED3DPS_VERSION(2,0))
1661 shader_addline(arg
->buffer
, "tmp0.x = exp2(floor(%s));\n", src_param
.param_str
);
1662 shader_addline(arg
->buffer
, "tmp0.y = %s - floor(%s);\n", src_param
.param_str
, src_param
.param_str
);
1663 shader_addline(arg
->buffer
, "tmp0.z = exp2(%s);\n", src_param
.param_str
);
1664 shader_addline(arg
->buffer
, "tmp0.w = 1.0;\n");
1666 shader_glsl_append_dst(arg
->buffer
, arg
);
1667 shader_glsl_get_write_mask(arg
->dst
, dst_mask
);
1668 shader_addline(arg
->buffer
, "tmp0%s);\n", dst_mask
);
1671 unsigned int mask_size
;
1673 write_mask
= shader_glsl_append_dst(arg
->buffer
, arg
);
1674 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
1676 if (mask_size
> 1) {
1677 shader_addline(arg
->buffer
, "vec%d(exp2(%s)));\n", mask_size
, src_param
.param_str
);
1679 shader_addline(arg
->buffer
, "exp2(%s));\n", src_param
.param_str
);
1684 /** Process the RCP (reciprocal or inverse) opcode in GLSL (dst = 1 / src) */
1685 static void shader_glsl_rcp(const SHADER_OPCODE_ARG
*arg
)
1687 glsl_src_param_t src_param
;
1689 unsigned int mask_size
;
1691 write_mask
= shader_glsl_append_dst(arg
->buffer
, arg
);
1692 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
1693 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_3
, &src_param
);
1695 if (mask_size
> 1) {
1696 shader_addline(arg
->buffer
, "vec%d(1.0 / %s));\n", mask_size
, src_param
.param_str
);
1698 shader_addline(arg
->buffer
, "1.0 / %s);\n", src_param
.param_str
);
1702 static void shader_glsl_rsq(const SHADER_OPCODE_ARG
*arg
)
1704 SHADER_BUFFER
* buffer
= arg
->buffer
;
1705 glsl_src_param_t src_param
;
1707 unsigned int mask_size
;
1709 write_mask
= shader_glsl_append_dst(buffer
, arg
);
1710 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
1712 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_3
, &src_param
);
1714 if (mask_size
> 1) {
1715 shader_addline(buffer
, "vec%d(inversesqrt(%s)));\n", mask_size
, src_param
.param_str
);
1717 shader_addline(buffer
, "inversesqrt(%s));\n", src_param
.param_str
);
1721 /** Process signed comparison opcodes in GLSL. */
1722 static void shader_glsl_compare(const SHADER_OPCODE_ARG
*arg
)
1724 glsl_src_param_t src0_param
;
1725 glsl_src_param_t src1_param
;
1727 unsigned int mask_size
;
1729 write_mask
= shader_glsl_append_dst(arg
->buffer
, arg
);
1730 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
1731 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], write_mask
, &src0_param
);
1732 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], write_mask
, &src1_param
);
1734 if (mask_size
> 1) {
1735 const char *compare
;
1737 switch(arg
->opcode
->opcode
) {
1738 case WINED3DSIO_SLT
: compare
= "lessThan"; break;
1739 case WINED3DSIO_SGE
: compare
= "greaterThanEqual"; break;
1740 default: compare
= "";
1741 FIXME("Can't handle opcode %s\n", arg
->opcode
->name
);
1744 shader_addline(arg
->buffer
, "vec%d(%s(%s, %s)));\n", mask_size
, compare
,
1745 src0_param
.param_str
, src1_param
.param_str
);
1747 switch(arg
->opcode
->opcode
) {
1748 case WINED3DSIO_SLT
:
1749 /* Step(src0, src1) is not suitable here because if src0 == src1 SLT is supposed,
1750 * to return 0.0 but step returns 1.0 because step is not < x
1751 * An alternative is a bvec compare padded with an unused second component.
1752 * step(src1 * -1.0, src0 * -1.0) is not an option because it suffers from the same
1753 * issue. Playing with not() is not possible either because not() does not accept
1756 shader_addline(arg
->buffer
, "(%s < %s) ? 1.0 : 0.0);\n", src0_param
.param_str
, src1_param
.param_str
);
1758 case WINED3DSIO_SGE
:
1759 /* Here we can use the step() function and safe a conditional */
1760 shader_addline(arg
->buffer
, "step(%s, %s));\n", src1_param
.param_str
, src0_param
.param_str
);
1763 FIXME("Can't handle opcode %s\n", arg
->opcode
->name
);
1769 /** Process CMP instruction in GLSL (dst = src0 >= 0.0 ? src1 : src2), per channel */
1770 static void shader_glsl_cmp(const SHADER_OPCODE_ARG
*arg
)
1772 glsl_src_param_t src0_param
;
1773 glsl_src_param_t src1_param
;
1774 glsl_src_param_t src2_param
;
1775 DWORD write_mask
, cmp_channel
= 0;
1778 BOOL temp_destination
= FALSE
;
1780 if(shader_is_scalar(arg
->src
[0])) {
1781 write_mask
= shader_glsl_append_dst(arg
->buffer
, arg
);
1783 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_ALL
, &src0_param
);
1784 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], write_mask
, &src1_param
);
1785 shader_glsl_add_src_param(arg
, arg
->src
[2], arg
->src_addr
[2], write_mask
, &src2_param
);
1787 shader_addline(arg
->buffer
, "%s >= 0.0 ? %s : %s);\n",
1788 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
1790 DWORD src0reg
= arg
->src
[0] & WINED3DSP_REGNUM_MASK
;
1791 DWORD src1reg
= arg
->src
[1] & WINED3DSP_REGNUM_MASK
;
1792 DWORD src2reg
= arg
->src
[2] & WINED3DSP_REGNUM_MASK
;
1793 DWORD src0regtype
= shader_get_regtype(arg
->src
[0]);
1794 DWORD src1regtype
= shader_get_regtype(arg
->src
[1]);
1795 DWORD src2regtype
= shader_get_regtype(arg
->src
[2]);
1796 DWORD dstreg
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
1797 DWORD dstregtype
= shader_get_regtype(arg
->dst
);
1799 /* Cycle through all source0 channels */
1800 for (i
=0; i
<4; i
++) {
1802 /* Find the destination channels which use the current source0 channel */
1803 for (j
=0; j
<4; j
++) {
1804 if ( ((arg
->src
[0] >> (WINED3DSP_SWIZZLE_SHIFT
+ 2*j
)) & 0x3) == i
) {
1805 write_mask
|= WINED3DSP_WRITEMASK_0
<< j
;
1806 cmp_channel
= WINED3DSP_WRITEMASK_0
<< j
;
1810 /* Splitting the cmp instruction up in multiple lines imposes a problem:
1811 * The first lines may overwrite source parameters of the following lines.
1812 * Deal with that by using a temporary destination register if needed
1814 if((src0reg
== dstreg
&& src0regtype
== dstregtype
) ||
1815 (src1reg
== dstreg
&& src1regtype
== dstregtype
) ||
1816 (src2reg
== dstreg
&& src2regtype
== dstregtype
)) {
1818 write_mask
= shader_glsl_get_write_mask(arg
->dst
& (~WINED3DSP_SWIZZLE_MASK
| write_mask
), mask_char
);
1819 if (!write_mask
) continue;
1820 shader_addline(arg
->buffer
, "tmp0%s = (", mask_char
);
1821 temp_destination
= TRUE
;
1823 write_mask
= shader_glsl_append_dst_ext(arg
->buffer
, arg
, arg
->dst
& (~WINED3DSP_SWIZZLE_MASK
| write_mask
));
1824 if (!write_mask
) continue;
1827 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], cmp_channel
, &src0_param
);
1828 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], write_mask
, &src1_param
);
1829 shader_glsl_add_src_param(arg
, arg
->src
[2], arg
->src_addr
[2], write_mask
, &src2_param
);
1831 shader_addline(arg
->buffer
, "%s >= 0.0 ? %s : %s);\n",
1832 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
1835 if(temp_destination
) {
1836 shader_glsl_get_write_mask(arg
->dst
, mask_char
);
1837 shader_glsl_append_dst_ext(arg
->buffer
, arg
, arg
->dst
);
1838 shader_addline(arg
->buffer
, "tmp0%s);\n", mask_char
);
1844 /** Process the CND opcode in GLSL (dst = (src0 > 0.5) ? src1 : src2) */
1845 /* For ps 1.1-1.3, only a single component of src0 is used. For ps 1.4
1846 * the compare is done per component of src0. */
1847 static void shader_glsl_cnd(const SHADER_OPCODE_ARG
*arg
)
1849 glsl_src_param_t src0_param
;
1850 glsl_src_param_t src1_param
;
1851 glsl_src_param_t src2_param
;
1852 DWORD write_mask
, cmp_channel
= 0;
1855 if (arg
->reg_maps
->shader_version
< WINED3DPS_VERSION(1, 4))
1857 write_mask
= shader_glsl_append_dst(arg
->buffer
, arg
);
1858 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
1859 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], write_mask
, &src1_param
);
1860 shader_glsl_add_src_param(arg
, arg
->src
[2], arg
->src_addr
[2], write_mask
, &src2_param
);
1862 /* Fun: The D3DSI_COISSUE flag changes the semantic of the cnd instruction for < 1.4 shaders */
1863 if(arg
->opcode_token
& WINED3DSI_COISSUE
) {
1864 shader_addline(arg
->buffer
, "%s /* COISSUE! */);\n", src1_param
.param_str
);
1866 shader_addline(arg
->buffer
, "%s > 0.5 ? %s : %s);\n",
1867 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
1871 /* Cycle through all source0 channels */
1872 for (i
=0; i
<4; i
++) {
1874 /* Find the destination channels which use the current source0 channel */
1875 for (j
=0; j
<4; j
++) {
1876 if ( ((arg
->src
[0] >> (WINED3DSP_SWIZZLE_SHIFT
+ 2*j
)) & 0x3) == i
) {
1877 write_mask
|= WINED3DSP_WRITEMASK_0
<< j
;
1878 cmp_channel
= WINED3DSP_WRITEMASK_0
<< j
;
1881 write_mask
= shader_glsl_append_dst_ext(arg
->buffer
, arg
, arg
->dst
& (~WINED3DSP_SWIZZLE_MASK
| write_mask
));
1882 if (!write_mask
) continue;
1884 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], cmp_channel
, &src0_param
);
1885 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], write_mask
, &src1_param
);
1886 shader_glsl_add_src_param(arg
, arg
->src
[2], arg
->src_addr
[2], write_mask
, &src2_param
);
1888 shader_addline(arg
->buffer
, "%s > 0.5 ? %s : %s);\n",
1889 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
1893 /** GLSL code generation for WINED3DSIO_MAD: Multiply the first 2 opcodes, then add the last */
1894 static void shader_glsl_mad(const SHADER_OPCODE_ARG
*arg
)
1896 glsl_src_param_t src0_param
;
1897 glsl_src_param_t src1_param
;
1898 glsl_src_param_t src2_param
;
1901 write_mask
= shader_glsl_append_dst(arg
->buffer
, arg
);
1902 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], write_mask
, &src0_param
);
1903 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], write_mask
, &src1_param
);
1904 shader_glsl_add_src_param(arg
, arg
->src
[2], arg
->src_addr
[2], write_mask
, &src2_param
);
1905 shader_addline(arg
->buffer
, "(%s * %s) + %s);\n",
1906 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
1909 /** Handles transforming all WINED3DSIO_M?x? opcodes for
1910 Vertex shaders to GLSL codes */
1911 static void shader_glsl_mnxn(const SHADER_OPCODE_ARG
*arg
)
1913 IWineD3DBaseShaderImpl
*shader
= (IWineD3DBaseShaderImpl
*)arg
->shader
;
1914 const SHADER_OPCODE
*opcode_table
= shader
->baseShader
.shader_ins
;
1915 DWORD shader_version
= arg
->reg_maps
->shader_version
;
1917 int nComponents
= 0;
1918 SHADER_OPCODE_ARG tmpArg
;
1920 memset(&tmpArg
, 0, sizeof(SHADER_OPCODE_ARG
));
1922 /* Set constants for the temporary argument */
1923 tmpArg
.shader
= arg
->shader
;
1924 tmpArg
.buffer
= arg
->buffer
;
1925 tmpArg
.src
[0] = arg
->src
[0];
1926 tmpArg
.src_addr
[0] = arg
->src_addr
[0];
1927 tmpArg
.src_addr
[1] = arg
->src_addr
[1];
1928 tmpArg
.reg_maps
= arg
->reg_maps
;
1930 switch(arg
->opcode
->opcode
) {
1931 case WINED3DSIO_M4x4
:
1933 tmpArg
.opcode
= shader_get_opcode(opcode_table
, shader_version
, WINED3DSIO_DP4
);
1935 case WINED3DSIO_M4x3
:
1937 tmpArg
.opcode
= shader_get_opcode(opcode_table
, shader_version
, WINED3DSIO_DP4
);
1939 case WINED3DSIO_M3x4
:
1941 tmpArg
.opcode
= shader_get_opcode(opcode_table
, shader_version
, WINED3DSIO_DP3
);
1943 case WINED3DSIO_M3x3
:
1945 tmpArg
.opcode
= shader_get_opcode(opcode_table
, shader_version
, WINED3DSIO_DP3
);
1947 case WINED3DSIO_M3x2
:
1949 tmpArg
.opcode
= shader_get_opcode(opcode_table
, shader_version
, WINED3DSIO_DP3
);
1955 for (i
= 0; i
< nComponents
; i
++) {
1956 tmpArg
.dst
= ((arg
->dst
) & ~WINED3DSP_WRITEMASK_ALL
)|(WINED3DSP_WRITEMASK_0
<<i
);
1957 tmpArg
.src
[1] = arg
->src
[1]+i
;
1958 shader_glsl_dot(&tmpArg
);
1963 The LRP instruction performs a component-wise linear interpolation
1964 between the second and third operands using the first operand as the
1965 blend factor. Equation: (dst = src2 + src0 * (src1 - src2))
1966 This is equivalent to mix(src2, src1, src0);
1968 static void shader_glsl_lrp(const SHADER_OPCODE_ARG
*arg
)
1970 glsl_src_param_t src0_param
;
1971 glsl_src_param_t src1_param
;
1972 glsl_src_param_t src2_param
;
1975 write_mask
= shader_glsl_append_dst(arg
->buffer
, arg
);
1977 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], write_mask
, &src0_param
);
1978 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], write_mask
, &src1_param
);
1979 shader_glsl_add_src_param(arg
, arg
->src
[2], arg
->src_addr
[2], write_mask
, &src2_param
);
1981 shader_addline(arg
->buffer
, "mix(%s, %s, %s));\n",
1982 src2_param
.param_str
, src1_param
.param_str
, src0_param
.param_str
);
1985 /** Process the WINED3DSIO_LIT instruction in GLSL:
1986 * dst.x = dst.w = 1.0
1987 * dst.y = (src0.x > 0) ? src0.x
1988 * dst.z = (src0.x > 0) ? ((src0.y > 0) ? pow(src0.y, src.w) : 0) : 0
1989 * where src.w is clamped at +- 128
1991 static void shader_glsl_lit(const SHADER_OPCODE_ARG
*arg
)
1993 glsl_src_param_t src0_param
;
1994 glsl_src_param_t src1_param
;
1995 glsl_src_param_t src3_param
;
1998 shader_glsl_append_dst(arg
->buffer
, arg
);
1999 shader_glsl_get_write_mask(arg
->dst
, dst_mask
);
2001 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2002 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_1
, &src1_param
);
2003 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_3
, &src3_param
);
2005 /* The sdk specifies the instruction like this
2007 * if(src.x > 0.0) dst.y = src.x
2009 * if(src.x > 0.0 && src.y > 0.0) dst.z = pow(src.y, power);
2013 * Obviously that has quite a few conditionals in it which we don't like. So the first step is this:
2014 * dst.x = 1.0 ... No further explanation needed
2015 * dst.y = max(src.y, 0.0); ... If x < 0.0, use 0.0, otherwise x. Same as the conditional
2016 * dst.z = x > 0.0 ? pow(max(y, 0.0), p) : 0; ... 0 ^ power is 0, and otherwise we use y anyway
2017 * dst.w = 1.0. ... Nothing fancy.
2019 * So we still have one conditional in there. So do this:
2020 * dst.z = pow(max(0.0, src.y) * step(0.0, src.x), power);
2022 * 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),
2023 * 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.
2024 * if both x and y are > 0, we get pow(y * 1.0, power), as it is supposed to
2026 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",
2027 src0_param
.param_str
, src1_param
.param_str
, src0_param
.param_str
, src3_param
.param_str
, dst_mask
);
2030 /** Process the WINED3DSIO_DST instruction in GLSL:
2032 * dst.y = src0.x * src0.y
2036 static void shader_glsl_dst(const SHADER_OPCODE_ARG
*arg
)
2038 glsl_src_param_t src0y_param
;
2039 glsl_src_param_t src0z_param
;
2040 glsl_src_param_t src1y_param
;
2041 glsl_src_param_t src1w_param
;
2044 shader_glsl_append_dst(arg
->buffer
, arg
);
2045 shader_glsl_get_write_mask(arg
->dst
, dst_mask
);
2047 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_1
, &src0y_param
);
2048 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_2
, &src0z_param
);
2049 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], WINED3DSP_WRITEMASK_1
, &src1y_param
);
2050 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], WINED3DSP_WRITEMASK_3
, &src1w_param
);
2052 shader_addline(arg
->buffer
, "vec4(1.0, %s * %s, %s, %s))%s;\n",
2053 src0y_param
.param_str
, src1y_param
.param_str
, src0z_param
.param_str
, src1w_param
.param_str
, dst_mask
);
2056 /** Process the WINED3DSIO_SINCOS instruction in GLSL:
2057 * VS 2.0 requires that specific cosine and sine constants be passed to this instruction so the hardware
2058 * can handle it. But, these functions are built-in for GLSL, so we can just ignore the last 2 params.
2060 * dst.x = cos(src0.?)
2061 * dst.y = sin(src0.?)
2065 static void shader_glsl_sincos(const SHADER_OPCODE_ARG
*arg
)
2067 glsl_src_param_t src0_param
;
2070 write_mask
= shader_glsl_append_dst(arg
->buffer
, arg
);
2071 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2073 switch (write_mask
) {
2074 case WINED3DSP_WRITEMASK_0
:
2075 shader_addline(arg
->buffer
, "cos(%s));\n", src0_param
.param_str
);
2078 case WINED3DSP_WRITEMASK_1
:
2079 shader_addline(arg
->buffer
, "sin(%s));\n", src0_param
.param_str
);
2082 case (WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
):
2083 shader_addline(arg
->buffer
, "vec2(cos(%s), sin(%s)));\n", src0_param
.param_str
, src0_param
.param_str
);
2087 ERR("Write mask should be .x, .y or .xy\n");
2092 /** Process the WINED3DSIO_LOOP instruction in GLSL:
2093 * Start a for() loop where src1.y is the initial value of aL,
2094 * increment aL by src1.z for a total of src1.x iterations.
2095 * Need to use a temporary variable for this operation.
2097 /* FIXME: I don't think nested loops will work correctly this way. */
2098 static void shader_glsl_loop(const SHADER_OPCODE_ARG
*arg
)
2100 glsl_src_param_t src1_param
;
2101 IWineD3DBaseShaderImpl
* shader
= (IWineD3DBaseShaderImpl
*) arg
->shader
;
2102 DWORD regtype
= shader_get_regtype(arg
->src
[1]);
2103 DWORD reg
= arg
->src
[1] & WINED3DSP_REGNUM_MASK
;
2104 const DWORD
*control_values
= NULL
;
2105 const local_constant
*constant
;
2107 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], WINED3DSP_WRITEMASK_ALL
, &src1_param
);
2109 /* Try to hardcode the loop control parameters if possible. Direct3D 9 class hardware doesn't support real
2110 * varying indexing, but Microsoft designed this feature for Shader model 2.x+. If the loop control is
2111 * known at compile time, the GLSL compiler can unroll the loop, and replace indirect addressing with direct
2114 if(regtype
== WINED3DSPR_CONSTINT
) {
2115 LIST_FOR_EACH_ENTRY(constant
, &shader
->baseShader
.constantsI
, local_constant
, entry
) {
2116 if(constant
->idx
== reg
) {
2117 control_values
= constant
->value
;
2123 if(control_values
) {
2124 if(control_values
[2] > 0) {
2125 shader_addline(arg
->buffer
, "for (aL%u = %d; aL%u < (%d * %d + %d); aL%u += %d) {\n",
2126 shader
->baseShader
.cur_loop_depth
, control_values
[1],
2127 shader
->baseShader
.cur_loop_depth
, control_values
[0], control_values
[2], control_values
[1],
2128 shader
->baseShader
.cur_loop_depth
, control_values
[2]);
2129 } else if(control_values
[2] == 0) {
2130 shader_addline(arg
->buffer
, "for (aL%u = %d, tmpInt%u = 0; tmpInt%u < %d; tmpInt%u++) {\n",
2131 shader
->baseShader
.cur_loop_depth
, control_values
[1], shader
->baseShader
.cur_loop_depth
,
2132 shader
->baseShader
.cur_loop_depth
, control_values
[0],
2133 shader
->baseShader
.cur_loop_depth
);
2135 shader_addline(arg
->buffer
, "for (aL%u = %d; aL%u > (%d * %d + %d); aL%u += %d) {\n",
2136 shader
->baseShader
.cur_loop_depth
, control_values
[1],
2137 shader
->baseShader
.cur_loop_depth
, control_values
[0], control_values
[2], control_values
[1],
2138 shader
->baseShader
.cur_loop_depth
, control_values
[2]);
2141 shader_addline(arg
->buffer
, "for (tmpInt%u = 0, aL%u = %s.y; tmpInt%u < %s.x; tmpInt%u++, aL%u += %s.z) {\n",
2142 shader
->baseShader
.cur_loop_depth
, shader
->baseShader
.cur_loop_regno
,
2143 src1_param
.reg_name
, shader
->baseShader
.cur_loop_depth
, src1_param
.reg_name
,
2144 shader
->baseShader
.cur_loop_depth
, shader
->baseShader
.cur_loop_regno
, src1_param
.reg_name
);
2147 shader
->baseShader
.cur_loop_depth
++;
2148 shader
->baseShader
.cur_loop_regno
++;
2151 static void shader_glsl_end(const SHADER_OPCODE_ARG
*arg
)
2153 IWineD3DBaseShaderImpl
* shader
= (IWineD3DBaseShaderImpl
*) arg
->shader
;
2155 shader_addline(arg
->buffer
, "}\n");
2157 if(arg
->opcode
->opcode
== WINED3DSIO_ENDLOOP
) {
2158 shader
->baseShader
.cur_loop_depth
--;
2159 shader
->baseShader
.cur_loop_regno
--;
2161 if(arg
->opcode
->opcode
== WINED3DSIO_ENDREP
) {
2162 shader
->baseShader
.cur_loop_depth
--;
2166 static void shader_glsl_rep(const SHADER_OPCODE_ARG
*arg
)
2168 IWineD3DBaseShaderImpl
* shader
= (IWineD3DBaseShaderImpl
*) arg
->shader
;
2169 glsl_src_param_t src0_param
;
2171 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2172 shader_addline(arg
->buffer
, "for (tmpInt%d = 0; tmpInt%d < %s; tmpInt%d++) {\n",
2173 shader
->baseShader
.cur_loop_depth
, shader
->baseShader
.cur_loop_depth
,
2174 src0_param
.param_str
, shader
->baseShader
.cur_loop_depth
);
2175 shader
->baseShader
.cur_loop_depth
++;
2178 static void shader_glsl_if(const SHADER_OPCODE_ARG
*arg
)
2180 glsl_src_param_t src0_param
;
2182 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2183 shader_addline(arg
->buffer
, "if (%s) {\n", src0_param
.param_str
);
2186 static void shader_glsl_ifc(const SHADER_OPCODE_ARG
*arg
)
2188 glsl_src_param_t src0_param
;
2189 glsl_src_param_t src1_param
;
2191 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2192 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], WINED3DSP_WRITEMASK_0
, &src1_param
);
2194 shader_addline(arg
->buffer
, "if (%s %s %s) {\n",
2195 src0_param
.param_str
, shader_get_comp_op(arg
->opcode_token
), src1_param
.param_str
);
2198 static void shader_glsl_else(const SHADER_OPCODE_ARG
*arg
)
2200 shader_addline(arg
->buffer
, "} else {\n");
2203 static void shader_glsl_break(const SHADER_OPCODE_ARG
*arg
)
2205 shader_addline(arg
->buffer
, "break;\n");
2208 /* FIXME: According to MSDN the compare is done per component. */
2209 static void shader_glsl_breakc(const SHADER_OPCODE_ARG
*arg
)
2211 glsl_src_param_t src0_param
;
2212 glsl_src_param_t src1_param
;
2214 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2215 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], WINED3DSP_WRITEMASK_0
, &src1_param
);
2217 shader_addline(arg
->buffer
, "if (%s %s %s) break;\n",
2218 src0_param
.param_str
, shader_get_comp_op(arg
->opcode_token
), src1_param
.param_str
);
2221 static void shader_glsl_label(const SHADER_OPCODE_ARG
*arg
)
2224 DWORD snum
= (arg
->src
[0]) & WINED3DSP_REGNUM_MASK
;
2225 shader_addline(arg
->buffer
, "}\n");
2226 shader_addline(arg
->buffer
, "void subroutine%u () {\n", snum
);
2229 static void shader_glsl_call(const SHADER_OPCODE_ARG
*arg
)
2231 DWORD snum
= (arg
->src
[0]) & WINED3DSP_REGNUM_MASK
;
2232 shader_addline(arg
->buffer
, "subroutine%u();\n", snum
);
2235 static void shader_glsl_callnz(const SHADER_OPCODE_ARG
*arg
)
2237 glsl_src_param_t src1_param
;
2239 DWORD snum
= (arg
->src
[0]) & WINED3DSP_REGNUM_MASK
;
2240 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], WINED3DSP_WRITEMASK_0
, &src1_param
);
2241 shader_addline(arg
->buffer
, "if (%s) subroutine%u();\n", src1_param
.param_str
, snum
);
2244 /*********************************************
2245 * Pixel Shader Specific Code begins here
2246 ********************************************/
2247 static void pshader_glsl_tex(const SHADER_OPCODE_ARG
*arg
)
2249 IWineD3DPixelShaderImpl
* This
= (IWineD3DPixelShaderImpl
*) arg
->shader
;
2250 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
2251 DWORD shader_version
= arg
->reg_maps
->shader_version
;
2252 char dst_swizzle
[6];
2253 glsl_sample_function_t sample_function
;
2256 BOOL projected
, texrect
= FALSE
;
2259 /* All versions have a destination register */
2260 shader_glsl_append_dst(arg
->buffer
, arg
);
2262 /* 1.0-1.4: Use destination register as sampler source.
2263 * 2.0+: Use provided sampler source. */
2264 if (shader_version
< WINED3DPS_VERSION(2,0)) sampler_idx
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2265 else sampler_idx
= arg
->src
[1] & WINED3DSP_REGNUM_MASK
;
2266 sampler_type
= arg
->reg_maps
->samplers
[sampler_idx
] & WINED3DSP_TEXTURETYPE_MASK
;
2268 if (shader_version
< WINED3DPS_VERSION(1,4))
2270 DWORD flags
= deviceImpl
->stateBlock
->textureState
[sampler_idx
][WINED3DTSS_TEXTURETRANSFORMFLAGS
];
2272 /* Projected cube textures don't make a lot of sense, the resulting coordinates stay the same. */
2273 if (flags
& WINED3DTTFF_PROJECTED
&& sampler_type
!= WINED3DSTT_CUBE
) {
2275 switch (flags
& ~WINED3DTTFF_PROJECTED
) {
2276 case WINED3DTTFF_COUNT1
: FIXME("WINED3DTTFF_PROJECTED with WINED3DTTFF_COUNT1?\n"); break;
2277 case WINED3DTTFF_COUNT2
: mask
= WINED3DSP_WRITEMASK_1
; break;
2278 case WINED3DTTFF_COUNT3
: mask
= WINED3DSP_WRITEMASK_2
; break;
2279 case WINED3DTTFF_COUNT4
:
2280 case WINED3DTTFF_DISABLE
: mask
= WINED3DSP_WRITEMASK_3
; break;
2286 else if (shader_version
< WINED3DPS_VERSION(2,0))
2288 DWORD src_mod
= arg
->src
[0] & WINED3DSP_SRCMOD_MASK
;
2290 if (src_mod
== WINED3DSPSM_DZ
) {
2292 mask
= WINED3DSP_WRITEMASK_2
;
2293 } else if (src_mod
== WINED3DSPSM_DW
) {
2295 mask
= WINED3DSP_WRITEMASK_3
;
2300 if(arg
->opcode_token
& WINED3DSI_TEXLD_PROJECT
) {
2301 /* ps 2.0 texldp instruction always divides by the fourth component. */
2303 mask
= WINED3DSP_WRITEMASK_3
;
2309 if(deviceImpl
->stateBlock
->textures
[sampler_idx
] &&
2310 IWineD3DBaseTexture_GetTextureDimensions(deviceImpl
->stateBlock
->textures
[sampler_idx
]) == GL_TEXTURE_RECTANGLE_ARB
) {
2314 shader_glsl_get_sample_function(sampler_type
, projected
, texrect
, &sample_function
);
2315 mask
|= sample_function
.coord_mask
;
2317 if (shader_version
< WINED3DPS_VERSION(2,0)) shader_glsl_get_write_mask(arg
->dst
, dst_swizzle
);
2318 else shader_glsl_get_swizzle(arg
->src
[1], FALSE
, arg
->dst
, dst_swizzle
);
2320 /* 1.0-1.3: Use destination register as coordinate source.
2321 1.4+: Use provided coordinate source register. */
2322 if (shader_version
< WINED3DPS_VERSION(1,4))
2325 shader_glsl_get_write_mask(mask
, coord_mask
);
2326 shader_addline(arg
->buffer
, "%s(Psampler%u, T%u%s)%s);\n",
2327 sample_function
.name
, sampler_idx
, sampler_idx
, coord_mask
, dst_swizzle
);
2329 glsl_src_param_t coord_param
;
2330 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], mask
, &coord_param
);
2331 if(arg
->opcode_token
& WINED3DSI_TEXLD_BIAS
) {
2332 glsl_src_param_t bias
;
2333 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_3
, &bias
);
2335 shader_addline(arg
->buffer
, "%s(Psampler%u, %s, %s)%s);\n",
2336 sample_function
.name
, sampler_idx
, coord_param
.param_str
,
2337 bias
.param_str
, dst_swizzle
);
2339 shader_addline(arg
->buffer
, "%s(Psampler%u, %s)%s);\n",
2340 sample_function
.name
, sampler_idx
, coord_param
.param_str
, dst_swizzle
);
2345 static void shader_glsl_texldl(const SHADER_OPCODE_ARG
*arg
)
2347 IWineD3DBaseShaderImpl
* This
= (IWineD3DBaseShaderImpl
*)arg
->shader
;
2348 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
2349 glsl_sample_function_t sample_function
;
2350 glsl_src_param_t coord_param
, lod_param
;
2351 char dst_swizzle
[6];
2354 BOOL texrect
= FALSE
;
2356 shader_glsl_append_dst(arg
->buffer
, arg
);
2357 shader_glsl_get_swizzle(arg
->src
[1], FALSE
, arg
->dst
, dst_swizzle
);
2359 sampler_idx
= arg
->src
[1] & WINED3DSP_REGNUM_MASK
;
2360 sampler_type
= arg
->reg_maps
->samplers
[sampler_idx
] & WINED3DSP_TEXTURETYPE_MASK
;
2361 if(deviceImpl
->stateBlock
->textures
[sampler_idx
] &&
2362 IWineD3DBaseTexture_GetTextureDimensions(deviceImpl
->stateBlock
->textures
[sampler_idx
]) == GL_TEXTURE_RECTANGLE_ARB
) {
2365 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
);
2367 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_3
, &lod_param
);
2369 if (shader_is_pshader_version(arg
->reg_maps
->shader_version
))
2371 /* The GLSL spec claims the Lod sampling functions are only supported in vertex shaders.
2372 * However, they seem to work just fine in fragment shaders as well. */
2373 WARN("Using %sLod in fragment shader.\n", sample_function
.name
);
2374 shader_addline(arg
->buffer
, "%sLod(Psampler%u, %s, %s)%s);\n",
2375 sample_function
.name
, sampler_idx
, coord_param
.param_str
, lod_param
.param_str
, dst_swizzle
);
2377 shader_addline(arg
->buffer
, "%sLod(Vsampler%u, %s, %s)%s);\n",
2378 sample_function
.name
, sampler_idx
, coord_param
.param_str
, lod_param
.param_str
, dst_swizzle
);
2382 static void pshader_glsl_texcoord(const SHADER_OPCODE_ARG
*arg
)
2384 /* FIXME: Make this work for more than just 2D textures */
2385 SHADER_BUFFER
* buffer
= arg
->buffer
;
2389 write_mask
= shader_glsl_append_dst(arg
->buffer
, arg
);
2390 shader_glsl_get_write_mask(write_mask
, dst_mask
);
2392 if (arg
->reg_maps
->shader_version
!= WINED3DPS_VERSION(1,4))
2394 DWORD reg
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2395 shader_addline(buffer
, "clamp(gl_TexCoord[%u], 0.0, 1.0)%s);\n", reg
, dst_mask
);
2397 DWORD reg
= arg
->src
[0] & WINED3DSP_REGNUM_MASK
;
2398 DWORD src_mod
= arg
->src
[0] & WINED3DSP_SRCMOD_MASK
;
2399 char dst_swizzle
[6];
2401 shader_glsl_get_swizzle(arg
->src
[0], FALSE
, write_mask
, dst_swizzle
);
2403 if (src_mod
== WINED3DSPSM_DZ
) {
2404 glsl_src_param_t div_param
;
2405 unsigned int mask_size
= shader_glsl_get_write_mask_size(write_mask
);
2406 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_2
, &div_param
);
2408 if (mask_size
> 1) {
2409 shader_addline(buffer
, "gl_TexCoord[%u]%s / vec%d(%s));\n", reg
, dst_swizzle
, mask_size
, div_param
.param_str
);
2411 shader_addline(buffer
, "gl_TexCoord[%u]%s / %s);\n", reg
, dst_swizzle
, div_param
.param_str
);
2413 } else if (src_mod
== WINED3DSPSM_DW
) {
2414 glsl_src_param_t div_param
;
2415 unsigned int mask_size
= shader_glsl_get_write_mask_size(write_mask
);
2416 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_3
, &div_param
);
2418 if (mask_size
> 1) {
2419 shader_addline(buffer
, "gl_TexCoord[%u]%s / vec%d(%s));\n", reg
, dst_swizzle
, mask_size
, div_param
.param_str
);
2421 shader_addline(buffer
, "gl_TexCoord[%u]%s / %s);\n", reg
, dst_swizzle
, div_param
.param_str
);
2424 shader_addline(buffer
, "gl_TexCoord[%u]%s);\n", reg
, dst_swizzle
);
2429 /** Process the WINED3DSIO_TEXDP3TEX instruction in GLSL:
2430 * Take a 3-component dot product of the TexCoord[dstreg] and src,
2431 * then perform a 1D texture lookup from stage dstregnum, place into dst. */
2432 static void pshader_glsl_texdp3tex(const SHADER_OPCODE_ARG
*arg
)
2434 glsl_src_param_t src0_param
;
2436 glsl_sample_function_t sample_function
;
2437 DWORD sampler_idx
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2438 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2439 DWORD sampler_type
= arg
->reg_maps
->samplers
[sampler_idx
] & WINED3DSP_TEXTURETYPE_MASK
;
2441 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], src_mask
, &src0_param
);
2443 shader_glsl_append_dst(arg
->buffer
, arg
);
2444 shader_glsl_get_write_mask(arg
->dst
, dst_mask
);
2446 /* Do I have to take care about the projected bit? I don't think so, since the dp3 returns only one
2447 * scalar, and projected sampling would require 4.
2449 * It is a dependent read - not valid with conditional NP2 textures
2451 shader_glsl_get_sample_function(sampler_type
, FALSE
, FALSE
, &sample_function
);
2453 switch(count_bits(sample_function
.coord_mask
)) {
2455 shader_addline(arg
->buffer
, "%s(Psampler%u, dot(gl_TexCoord[%u].xyz, %s))%s);\n",
2456 sample_function
.name
, sampler_idx
, sampler_idx
, src0_param
.param_str
, dst_mask
);
2460 shader_addline(arg
->buffer
, "%s(Psampler%u, vec2(dot(gl_TexCoord[%u].xyz, %s), 0.0))%s);\n",
2461 sample_function
.name
, sampler_idx
, sampler_idx
, src0_param
.param_str
, dst_mask
);
2465 shader_addline(arg
->buffer
, "%s(Psampler%u, vec3(dot(gl_TexCoord[%u].xyz, %s), 0.0, 0.0))%s);\n",
2466 sample_function
.name
, sampler_idx
, sampler_idx
, src0_param
.param_str
, dst_mask
);
2469 FIXME("Unexpected mask bitcount %d\n", count_bits(sample_function
.coord_mask
));
2473 /** Process the WINED3DSIO_TEXDP3 instruction in GLSL:
2474 * Take a 3-component dot product of the TexCoord[dstreg] and src. */
2475 static void pshader_glsl_texdp3(const SHADER_OPCODE_ARG
*arg
)
2477 glsl_src_param_t src0_param
;
2478 DWORD dstreg
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2479 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2481 unsigned int mask_size
;
2483 dst_mask
= shader_glsl_append_dst(arg
->buffer
, arg
);
2484 mask_size
= shader_glsl_get_write_mask_size(dst_mask
);
2485 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], src_mask
, &src0_param
);
2487 if (mask_size
> 1) {
2488 shader_addline(arg
->buffer
, "vec%d(dot(T%u.xyz, %s)));\n", mask_size
, dstreg
, src0_param
.param_str
);
2490 shader_addline(arg
->buffer
, "dot(T%u.xyz, %s));\n", dstreg
, src0_param
.param_str
);
2494 /** Process the WINED3DSIO_TEXDEPTH instruction in GLSL:
2495 * Calculate the depth as dst.x / dst.y */
2496 static void pshader_glsl_texdepth(const SHADER_OPCODE_ARG
*arg
)
2498 glsl_dst_param_t dst_param
;
2500 shader_glsl_add_dst_param(arg
, arg
->dst
, 0, &dst_param
);
2502 /* Tests show that texdepth never returns anything below 0.0, and that r5.y is clamped to 1.0.
2503 * Negative input is accepted, -0.25 / -0.5 returns 0.5. GL should clamp gl_FragDepth to [0;1], but
2504 * this doesn't always work, so clamp the results manually. Whether or not the x value is clamped at 1
2505 * too is irrelevant, since if x = 0, any y value < 1.0 (and > 1.0 is not allowed) results in a result
2508 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
);
2511 /** Process the WINED3DSIO_TEXM3X2DEPTH instruction in GLSL:
2512 * Last row of a 3x2 matrix multiply, use the result to calculate the depth:
2513 * Calculate tmp0.y = TexCoord[dstreg] . src.xyz; (tmp0.x has already been calculated)
2514 * depth = (tmp0.y == 0.0) ? 1.0 : tmp0.x / tmp0.y
2516 static void pshader_glsl_texm3x2depth(const SHADER_OPCODE_ARG
*arg
)
2518 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2519 DWORD dstreg
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2520 glsl_src_param_t src0_param
;
2522 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], src_mask
, &src0_param
);
2524 shader_addline(arg
->buffer
, "tmp0.y = dot(T%u.xyz, %s);\n", dstreg
, src0_param
.param_str
);
2525 shader_addline(arg
->buffer
, "gl_FragDepth = (tmp0.y == 0.0) ? 1.0 : clamp(tmp0.x / tmp0.y, 0.0, 1.0);\n");
2528 /** Process the WINED3DSIO_TEXM3X2PAD instruction in GLSL
2529 * Calculate the 1st of a 2-row matrix multiplication. */
2530 static void pshader_glsl_texm3x2pad(const SHADER_OPCODE_ARG
*arg
)
2532 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2533 DWORD reg
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2534 SHADER_BUFFER
* buffer
= arg
->buffer
;
2535 glsl_src_param_t src0_param
;
2537 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], src_mask
, &src0_param
);
2538 shader_addline(buffer
, "tmp0.x = dot(T%u.xyz, %s);\n", reg
, src0_param
.param_str
);
2541 /** Process the WINED3DSIO_TEXM3X3PAD instruction in GLSL
2542 * Calculate the 1st or 2nd row of a 3-row matrix multiplication. */
2543 static void pshader_glsl_texm3x3pad(const SHADER_OPCODE_ARG
* arg
)
2545 IWineD3DPixelShaderImpl
* shader
= (IWineD3DPixelShaderImpl
*) arg
->shader
;
2546 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2547 DWORD reg
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2548 SHADER_BUFFER
* buffer
= arg
->buffer
;
2549 SHADER_PARSE_STATE
* current_state
= &shader
->baseShader
.parse_state
;
2550 glsl_src_param_t src0_param
;
2552 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], src_mask
, &src0_param
);
2553 shader_addline(buffer
, "tmp0.%c = dot(T%u.xyz, %s);\n", 'x' + current_state
->current_row
, reg
, src0_param
.param_str
);
2554 current_state
->texcoord_w
[current_state
->current_row
++] = reg
;
2557 static void pshader_glsl_texm3x2tex(const SHADER_OPCODE_ARG
*arg
)
2559 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2560 DWORD reg
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2561 SHADER_BUFFER
* buffer
= arg
->buffer
;
2562 glsl_src_param_t src0_param
;
2565 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], src_mask
, &src0_param
);
2566 shader_addline(buffer
, "tmp0.y = dot(T%u.xyz, %s);\n", reg
, src0_param
.param_str
);
2568 shader_glsl_append_dst(buffer
, arg
);
2569 shader_glsl_get_write_mask(arg
->dst
, dst_mask
);
2571 /* Sample the texture using the calculated coordinates */
2572 shader_addline(buffer
, "texture2D(Psampler%u, tmp0.xy)%s);\n", reg
, dst_mask
);
2575 /** Process the WINED3DSIO_TEXM3X3TEX instruction in GLSL
2576 * Perform the 3rd row of a 3x3 matrix multiply, then sample the texture using the calculated coordinates */
2577 static void pshader_glsl_texm3x3tex(const SHADER_OPCODE_ARG
*arg
)
2579 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2580 glsl_src_param_t src0_param
;
2582 DWORD reg
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2583 IWineD3DPixelShaderImpl
* This
= (IWineD3DPixelShaderImpl
*) arg
->shader
;
2584 SHADER_PARSE_STATE
* current_state
= &This
->baseShader
.parse_state
;
2585 DWORD sampler_type
= arg
->reg_maps
->samplers
[reg
] & WINED3DSP_TEXTURETYPE_MASK
;
2586 glsl_sample_function_t sample_function
;
2588 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], src_mask
, &src0_param
);
2589 shader_addline(arg
->buffer
, "tmp0.z = dot(T%u.xyz, %s);\n", reg
, src0_param
.param_str
);
2591 shader_glsl_append_dst(arg
->buffer
, arg
);
2592 shader_glsl_get_write_mask(arg
->dst
, dst_mask
);
2593 /* Dependent read, not valid with conditional NP2 */
2594 shader_glsl_get_sample_function(sampler_type
, FALSE
, FALSE
, &sample_function
);
2596 /* Sample the texture using the calculated coordinates */
2597 shader_addline(arg
->buffer
, "%s(Psampler%u, tmp0.xyz)%s);\n", sample_function
.name
, reg
, dst_mask
);
2599 current_state
->current_row
= 0;
2602 /** Process the WINED3DSIO_TEXM3X3 instruction in GLSL
2603 * Perform the 3rd row of a 3x3 matrix multiply */
2604 static void pshader_glsl_texm3x3(const SHADER_OPCODE_ARG
*arg
)
2606 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2607 glsl_src_param_t src0_param
;
2609 DWORD reg
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2610 IWineD3DPixelShaderImpl
* This
= (IWineD3DPixelShaderImpl
*) arg
->shader
;
2611 SHADER_PARSE_STATE
* current_state
= &This
->baseShader
.parse_state
;
2613 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], src_mask
, &src0_param
);
2615 shader_glsl_append_dst(arg
->buffer
, arg
);
2616 shader_glsl_get_write_mask(arg
->dst
, dst_mask
);
2617 shader_addline(arg
->buffer
, "vec4(tmp0.xy, dot(T%u.xyz, %s), 1.0)%s);\n", reg
, src0_param
.param_str
, dst_mask
);
2619 current_state
->current_row
= 0;
2622 /** Process the WINED3DSIO_TEXM3X3SPEC instruction in GLSL
2623 * Perform the final texture lookup based on the previous 2 3x3 matrix multiplies */
2624 static void pshader_glsl_texm3x3spec(const SHADER_OPCODE_ARG
*arg
)
2626 IWineD3DPixelShaderImpl
* shader
= (IWineD3DPixelShaderImpl
*) arg
->shader
;
2627 DWORD reg
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2628 glsl_src_param_t src0_param
;
2629 glsl_src_param_t src1_param
;
2631 SHADER_BUFFER
* buffer
= arg
->buffer
;
2632 SHADER_PARSE_STATE
* current_state
= &shader
->baseShader
.parse_state
;
2633 DWORD stype
= arg
->reg_maps
->samplers
[reg
] & WINED3DSP_TEXTURETYPE_MASK
;
2634 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2635 glsl_sample_function_t sample_function
;
2637 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], src_mask
, &src0_param
);
2638 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], src_mask
, &src1_param
);
2640 /* Perform the last matrix multiply operation */
2641 shader_addline(buffer
, "tmp0.z = dot(T%u.xyz, %s);\n", reg
, src0_param
.param_str
);
2642 /* Reflection calculation */
2643 shader_addline(buffer
, "tmp0.xyz = -reflect((%s), normalize(tmp0.xyz));\n", src1_param
.param_str
);
2645 shader_glsl_append_dst(buffer
, arg
);
2646 shader_glsl_get_write_mask(arg
->dst
, dst_mask
);
2647 /* Dependent read, not valid with conditional NP2 */
2648 shader_glsl_get_sample_function(stype
, FALSE
, FALSE
, &sample_function
);
2650 /* Sample the texture */
2651 shader_addline(buffer
, "%s(Psampler%u, tmp0.xyz)%s);\n", sample_function
.name
, reg
, dst_mask
);
2653 current_state
->current_row
= 0;
2656 /** Process the WINED3DSIO_TEXM3X3VSPEC instruction in GLSL
2657 * Perform the final texture lookup based on the previous 2 3x3 matrix multiplies */
2658 static void pshader_glsl_texm3x3vspec(const SHADER_OPCODE_ARG
*arg
)
2660 IWineD3DPixelShaderImpl
* shader
= (IWineD3DPixelShaderImpl
*) arg
->shader
;
2661 DWORD reg
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2662 SHADER_BUFFER
* buffer
= arg
->buffer
;
2663 SHADER_PARSE_STATE
* current_state
= &shader
->baseShader
.parse_state
;
2664 glsl_src_param_t src0_param
;
2666 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2667 DWORD sampler_type
= arg
->reg_maps
->samplers
[reg
] & WINED3DSP_TEXTURETYPE_MASK
;
2668 glsl_sample_function_t sample_function
;
2670 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], src_mask
, &src0_param
);
2672 /* Perform the last matrix multiply operation */
2673 shader_addline(buffer
, "tmp0.z = dot(vec3(T%u), vec3(%s));\n", reg
, src0_param
.param_str
);
2675 /* Construct the eye-ray vector from w coordinates */
2676 shader_addline(buffer
, "tmp1.xyz = normalize(vec3(gl_TexCoord[%u].w, gl_TexCoord[%u].w, gl_TexCoord[%u].w));\n",
2677 current_state
->texcoord_w
[0], current_state
->texcoord_w
[1], reg
);
2678 shader_addline(buffer
, "tmp0.xyz = -reflect(tmp1.xyz, normalize(tmp0.xyz));\n");
2680 shader_glsl_append_dst(buffer
, arg
);
2681 shader_glsl_get_write_mask(arg
->dst
, dst_mask
);
2682 /* Dependent read, not valid with conditional NP2 */
2683 shader_glsl_get_sample_function(sampler_type
, FALSE
, FALSE
, &sample_function
);
2685 /* Sample the texture using the calculated coordinates */
2686 shader_addline(buffer
, "%s(Psampler%u, tmp0.xyz)%s);\n", sample_function
.name
, reg
, dst_mask
);
2688 current_state
->current_row
= 0;
2691 /** Process the WINED3DSIO_TEXBEM instruction in GLSL.
2692 * Apply a fake bump map transform.
2693 * texbem is pshader <= 1.3 only, this saves a few version checks
2695 static void pshader_glsl_texbem(const SHADER_OPCODE_ARG
*arg
)
2697 IWineD3DPixelShaderImpl
* This
= (IWineD3DPixelShaderImpl
*) arg
->shader
;
2698 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
2699 char dst_swizzle
[6];
2700 glsl_sample_function_t sample_function
;
2701 glsl_src_param_t coord_param
;
2708 sampler_idx
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2709 flags
= deviceImpl
->stateBlock
->textureState
[sampler_idx
][WINED3DTSS_TEXTURETRANSFORMFLAGS
];
2711 sampler_type
= arg
->reg_maps
->samplers
[sampler_idx
] & WINED3DSP_TEXTURETYPE_MASK
;
2712 /* Dependent read, not valid with conditional NP2 */
2713 shader_glsl_get_sample_function(sampler_type
, FALSE
, FALSE
, &sample_function
);
2714 mask
= sample_function
.coord_mask
;
2716 shader_glsl_get_write_mask(arg
->dst
, dst_swizzle
);
2718 shader_glsl_get_write_mask(mask
, coord_mask
);
2720 /* with projective textures, texbem only divides the static texture coord, not the displacement,
2721 * so we can't let the GL handle this.
2723 if (flags
& WINED3DTTFF_PROJECTED
) {
2725 char coord_div_mask
[3];
2726 switch (flags
& ~WINED3DTTFF_PROJECTED
) {
2727 case WINED3DTTFF_COUNT1
: FIXME("WINED3DTTFF_PROJECTED with WINED3DTTFF_COUNT1?\n"); break;
2728 case WINED3DTTFF_COUNT2
: div_mask
= WINED3DSP_WRITEMASK_1
; break;
2729 case WINED3DTTFF_COUNT3
: div_mask
= WINED3DSP_WRITEMASK_2
; break;
2730 case WINED3DTTFF_COUNT4
:
2731 case WINED3DTTFF_DISABLE
: div_mask
= WINED3DSP_WRITEMASK_3
; break;
2733 shader_glsl_get_write_mask(div_mask
, coord_div_mask
);
2734 shader_addline(arg
->buffer
, "T%u%s /= T%u%s;\n", sampler_idx
, coord_mask
, sampler_idx
, coord_div_mask
);
2737 shader_glsl_append_dst(arg
->buffer
, arg
);
2738 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_0
|WINED3DSP_WRITEMASK_1
, &coord_param
);
2739 if(arg
->opcode
->opcode
== WINED3DSIO_TEXBEML
) {
2740 glsl_src_param_t luminance_param
;
2741 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_2
, &luminance_param
);
2742 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",
2743 sample_function
.name
, sampler_idx
, sampler_idx
, coord_mask
, sampler_idx
, coord_param
.param_str
, coord_mask
,
2744 luminance_param
.param_str
, sampler_idx
, sampler_idx
, dst_swizzle
);
2746 shader_addline(arg
->buffer
, "%s(Psampler%u, T%u%s + vec4(bumpenvmat%d * %s, 0.0, 0.0)%s )%s);\n",
2747 sample_function
.name
, sampler_idx
, sampler_idx
, coord_mask
, sampler_idx
, coord_param
.param_str
, coord_mask
, dst_swizzle
);
2751 static void pshader_glsl_bem(const SHADER_OPCODE_ARG
*arg
)
2753 glsl_src_param_t src0_param
, src1_param
;
2754 DWORD sampler_idx
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2756 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_0
|WINED3DSP_WRITEMASK_1
, &src0_param
);
2757 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], WINED3DSP_WRITEMASK_0
|WINED3DSP_WRITEMASK_1
, &src1_param
);
2759 shader_glsl_append_dst(arg
->buffer
, arg
);
2760 shader_addline(arg
->buffer
, "%s + bumpenvmat%d * %s);\n",
2761 src0_param
.param_str
, sampler_idx
, src1_param
.param_str
);
2764 /** Process the WINED3DSIO_TEXREG2AR instruction in GLSL
2765 * Sample 2D texture at dst using the alpha & red (wx) components of src as texture coordinates */
2766 static void pshader_glsl_texreg2ar(const SHADER_OPCODE_ARG
*arg
)
2768 glsl_src_param_t src0_param
;
2769 DWORD sampler_idx
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2772 shader_glsl_append_dst(arg
->buffer
, arg
);
2773 shader_glsl_get_write_mask(arg
->dst
, dst_mask
);
2774 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_ALL
, &src0_param
);
2776 shader_addline(arg
->buffer
, "texture2D(Psampler%u, %s.wx)%s);\n", sampler_idx
, src0_param
.reg_name
, dst_mask
);
2779 /** Process the WINED3DSIO_TEXREG2GB instruction in GLSL
2780 * Sample 2D texture at dst using the green & blue (yz) components of src as texture coordinates */
2781 static void pshader_glsl_texreg2gb(const SHADER_OPCODE_ARG
*arg
)
2783 glsl_src_param_t src0_param
;
2784 DWORD sampler_idx
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2787 shader_glsl_append_dst(arg
->buffer
, arg
);
2788 shader_glsl_get_write_mask(arg
->dst
, dst_mask
);
2789 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_ALL
, &src0_param
);
2791 shader_addline(arg
->buffer
, "texture2D(Psampler%u, %s.yz)%s);\n", sampler_idx
, src0_param
.reg_name
, dst_mask
);
2794 /** Process the WINED3DSIO_TEXREG2RGB instruction in GLSL
2795 * Sample texture at dst using the rgb (xyz) components of src as texture coordinates */
2796 static void pshader_glsl_texreg2rgb(const SHADER_OPCODE_ARG
*arg
)
2798 glsl_src_param_t src0_param
;
2800 DWORD sampler_idx
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2801 DWORD sampler_type
= arg
->reg_maps
->samplers
[sampler_idx
] & WINED3DSP_TEXTURETYPE_MASK
;
2802 glsl_sample_function_t sample_function
;
2804 shader_glsl_append_dst(arg
->buffer
, arg
);
2805 shader_glsl_get_write_mask(arg
->dst
, dst_mask
);
2806 /* Dependent read, not valid with conditional NP2 */
2807 shader_glsl_get_sample_function(sampler_type
, FALSE
, FALSE
, &sample_function
);
2808 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], sample_function
.coord_mask
, &src0_param
);
2810 shader_addline(arg
->buffer
, "%s(Psampler%u, %s)%s);\n", sample_function
.name
, sampler_idx
, src0_param
.param_str
, dst_mask
);
2813 /** Process the WINED3DSIO_TEXKILL instruction in GLSL.
2814 * If any of the first 3 components are < 0, discard this pixel */
2815 static void pshader_glsl_texkill(const SHADER_OPCODE_ARG
*arg
)
2817 glsl_dst_param_t dst_param
;
2819 /* The argument is a destination parameter, and no writemasks are allowed */
2820 shader_glsl_add_dst_param(arg
, arg
->dst
, 0, &dst_param
);
2821 if ((arg
->reg_maps
->shader_version
>= WINED3DPS_VERSION(2,0)))
2823 /* 2.0 shaders compare all 4 components in texkill */
2824 shader_addline(arg
->buffer
, "if (any(lessThan(%s.xyzw, vec4(0.0)))) discard;\n", dst_param
.reg_name
);
2826 /* 1.X shaders only compare the first 3 components, probably due to the nature of the texkill
2827 * instruction as a tex* instruction, and phase, which kills all a / w components. Even if all
2828 * 4 components are defined, only the first 3 are used
2830 shader_addline(arg
->buffer
, "if (any(lessThan(%s.xyz, vec3(0.0)))) discard;\n", dst_param
.reg_name
);
2834 /** Process the WINED3DSIO_DP2ADD instruction in GLSL.
2835 * dst = dot2(src0, src1) + src2 */
2836 static void pshader_glsl_dp2add(const SHADER_OPCODE_ARG
*arg
)
2838 glsl_src_param_t src0_param
;
2839 glsl_src_param_t src1_param
;
2840 glsl_src_param_t src2_param
;
2842 unsigned int mask_size
;
2844 write_mask
= shader_glsl_append_dst(arg
->buffer
, arg
);
2845 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
2847 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
, &src0_param
);
2848 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
, &src1_param
);
2849 shader_glsl_add_src_param(arg
, arg
->src
[2], arg
->src_addr
[2], WINED3DSP_WRITEMASK_0
, &src2_param
);
2851 if (mask_size
> 1) {
2852 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
);
2854 shader_addline(arg
->buffer
, "dot(%s, %s) + %s);\n", src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
2858 static void pshader_glsl_input_pack(SHADER_BUFFER
* buffer
, const struct semantic
* semantics_in
,
2859 IWineD3DPixelShader
*iface
, enum vertexprocessing_mode vertexprocessing
)
2862 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*) iface
;
2864 for (i
= 0; i
< MAX_REG_INPUT
; i
++) {
2866 DWORD usage_token
= semantics_in
[i
].usage
;
2867 DWORD register_token
= semantics_in
[i
].reg
;
2868 DWORD usage
, usage_idx
;
2872 if (!usage_token
) continue;
2873 usage
= (usage_token
& WINED3DSP_DCL_USAGE_MASK
) >> WINED3DSP_DCL_USAGE_SHIFT
;
2874 usage_idx
= (usage_token
& WINED3DSP_DCL_USAGEINDEX_MASK
) >> WINED3DSP_DCL_USAGEINDEX_SHIFT
;
2875 shader_glsl_get_write_mask(register_token
, reg_mask
);
2879 case WINED3DDECLUSAGE_TEXCOORD
:
2880 if(usage_idx
< 8 && vertexprocessing
== pretransformed
) {
2881 shader_addline(buffer
, "IN[%u]%s = gl_TexCoord[%u]%s;\n",
2882 This
->input_reg_map
[i
], reg_mask
, usage_idx
, reg_mask
);
2884 shader_addline(buffer
, "IN[%u]%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
2885 This
->input_reg_map
[i
], reg_mask
, reg_mask
);
2889 case WINED3DDECLUSAGE_COLOR
:
2891 shader_addline(buffer
, "IN[%u]%s = vec4(gl_Color)%s;\n",
2892 This
->input_reg_map
[i
], reg_mask
, reg_mask
);
2893 else if (usage_idx
== 1)
2894 shader_addline(buffer
, "IN[%u]%s = vec4(gl_SecondaryColor)%s;\n",
2895 This
->input_reg_map
[i
], reg_mask
, reg_mask
);
2897 shader_addline(buffer
, "IN[%u]%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
2898 This
->input_reg_map
[i
], reg_mask
, reg_mask
);
2902 shader_addline(buffer
, "IN[%u]%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
2903 This
->input_reg_map
[i
], reg_mask
, reg_mask
);
2908 /*********************************************
2909 * Vertex Shader Specific Code begins here
2910 ********************************************/
2912 static void add_glsl_program_entry(struct shader_glsl_priv
*priv
, struct glsl_shader_prog_link
*entry
) {
2913 glsl_program_key_t
*key
;
2915 key
= HeapAlloc(GetProcessHeap(), 0, sizeof(glsl_program_key_t
));
2916 key
->vshader
= entry
->vshader
;
2917 key
->pshader
= entry
->pshader
;
2918 key
->ps_args
= entry
->ps_args
;
2920 hash_table_put(priv
->glsl_program_lookup
, key
, entry
);
2923 static struct glsl_shader_prog_link
*get_glsl_program_entry(struct shader_glsl_priv
*priv
,
2924 GLhandleARB vshader
, IWineD3DPixelShader
*pshader
, struct ps_compile_args
*ps_args
) {
2925 glsl_program_key_t key
;
2927 key
.vshader
= vshader
;
2928 key
.pshader
= pshader
;
2929 key
.ps_args
= *ps_args
;
2931 return (struct glsl_shader_prog_link
*)hash_table_get(priv
->glsl_program_lookup
, &key
);
2934 static void delete_glsl_program_entry(struct shader_glsl_priv
*priv
, const WineD3D_GL_Info
*gl_info
,
2935 struct glsl_shader_prog_link
*entry
)
2937 glsl_program_key_t
*key
;
2939 key
= HeapAlloc(GetProcessHeap(), 0, sizeof(glsl_program_key_t
));
2940 key
->vshader
= entry
->vshader
;
2941 key
->pshader
= entry
->pshader
;
2942 key
->ps_args
= entry
->ps_args
;
2943 hash_table_remove(priv
->glsl_program_lookup
, key
);
2945 GL_EXTCALL(glDeleteObjectARB(entry
->programId
));
2946 if (entry
->vshader
) list_remove(&entry
->vshader_entry
);
2947 if (entry
->pshader
) list_remove(&entry
->pshader_entry
);
2948 HeapFree(GetProcessHeap(), 0, entry
->vuniformF_locations
);
2949 HeapFree(GetProcessHeap(), 0, entry
->puniformF_locations
);
2950 HeapFree(GetProcessHeap(), 0, entry
);
2953 static void handle_ps3_input(SHADER_BUFFER
*buffer
, const struct semantic
*semantics_in
,
2954 const struct semantic
*semantics_out
, const WineD3D_GL_Info
*gl_info
, const DWORD
*map
)
2957 DWORD usage_token
, usage_token_out
;
2958 DWORD register_token
, register_token_out
;
2959 DWORD usage
, usage_idx
, usage_out
, usage_idx_out
;
2962 DWORD in_count
= GL_LIMITS(glsl_varyings
) / 4;
2963 char reg_mask
[6], reg_mask_out
[6];
2964 char destination
[50];
2966 set
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*set
) * (in_count
+ 2));
2968 if (!semantics_out
) {
2969 /* Save gl_FrontColor & gl_FrontSecondaryColor before overwriting them. */
2970 shader_addline(buffer
, "vec4 front_color = gl_FrontColor;\n");
2971 shader_addline(buffer
, "vec4 front_secondary_color = gl_FrontSecondaryColor;\n");
2974 for(i
= 0; i
< MAX_REG_INPUT
; i
++) {
2975 usage_token
= semantics_in
[i
].usage
;
2976 if (!usage_token
) continue;
2979 if (in_idx
>= (in_count
+ 2)) {
2980 FIXME("More input varyings declared than supported, expect issues\n");
2982 } else if(map
[i
] == -1) {
2983 /* Declared, but not read register */
2987 if (in_idx
== in_count
) {
2988 sprintf(destination
, "gl_FrontColor");
2989 } else if (in_idx
== in_count
+ 1) {
2990 sprintf(destination
, "gl_FrontSecondaryColor");
2992 sprintf(destination
, "IN[%u]", in_idx
);
2995 register_token
= semantics_in
[i
].reg
;
2997 usage
= (usage_token
& WINED3DSP_DCL_USAGE_MASK
) >> WINED3DSP_DCL_USAGE_SHIFT
;
2998 usage_idx
= (usage_token
& WINED3DSP_DCL_USAGEINDEX_MASK
) >> WINED3DSP_DCL_USAGEINDEX_SHIFT
;
2999 set
[map
[i
]] = shader_glsl_get_write_mask(register_token
, reg_mask
);
3001 if(!semantics_out
) {
3003 case WINED3DDECLUSAGE_COLOR
:
3005 shader_addline(buffer
, "%s%s = front_color%s;\n",
3006 destination
, reg_mask
, reg_mask
);
3007 else if (usage_idx
== 1)
3008 shader_addline(buffer
, "%s%s = front_secondary_color%s;\n",
3009 destination
, reg_mask
, reg_mask
);
3011 shader_addline(buffer
, "%s%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
3012 destination
, reg_mask
, reg_mask
);
3015 case WINED3DDECLUSAGE_TEXCOORD
:
3016 if (usage_idx
< 8) {
3017 shader_addline(buffer
, "%s%s = gl_TexCoord[%u]%s;\n",
3018 destination
, reg_mask
, usage_idx
, reg_mask
);
3020 shader_addline(buffer
, "%s%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
3021 destination
, reg_mask
, reg_mask
);
3025 case WINED3DDECLUSAGE_FOG
:
3026 shader_addline(buffer
, "%s%s = vec4(gl_FogFragCoord, 0.0, 0.0, 0.0)%s;\n",
3027 destination
, reg_mask
, reg_mask
);
3031 shader_addline(buffer
, "%s%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
3032 destination
, reg_mask
, reg_mask
);
3036 for(j
= 0; j
< MAX_REG_OUTPUT
; j
++) {
3037 usage_token_out
= semantics_out
[j
].usage
;
3038 if (!usage_token_out
) continue;
3039 register_token_out
= semantics_out
[j
].reg
;
3041 usage_out
= (usage_token_out
& WINED3DSP_DCL_USAGE_MASK
) >> WINED3DSP_DCL_USAGE_SHIFT
;
3042 usage_idx_out
= (usage_token_out
& WINED3DSP_DCL_USAGEINDEX_MASK
) >> WINED3DSP_DCL_USAGEINDEX_SHIFT
;
3043 shader_glsl_get_write_mask(register_token_out
, reg_mask_out
);
3045 if(usage
== usage_out
&&
3046 usage_idx
== usage_idx_out
) {
3047 shader_addline(buffer
, "%s%s = OUT[%u]%s;\n",
3048 destination
, reg_mask
, j
, reg_mask
);
3053 shader_addline(buffer
, "%s%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
3054 destination
, reg_mask
, reg_mask
);
3059 /* This is solely to make the compiler / linker happy and avoid warning about undefined
3060 * varyings. It shouldn't result in any real code executed on the GPU, since all read
3061 * input varyings are assigned above, if the optimizer works properly.
3063 for(i
= 0; i
< in_count
+ 2; i
++) {
3064 if(set
[i
] != WINED3DSP_WRITEMASK_ALL
) {
3065 unsigned int size
= 0;
3066 memset(reg_mask
, 0, sizeof(reg_mask
));
3067 if(!(set
[i
] & WINED3DSP_WRITEMASK_0
)) {
3068 reg_mask
[size
] = 'x';
3071 if(!(set
[i
] & WINED3DSP_WRITEMASK_1
)) {
3072 reg_mask
[size
] = 'y';
3075 if(!(set
[i
] & WINED3DSP_WRITEMASK_2
)) {
3076 reg_mask
[size
] = 'z';
3079 if(!(set
[i
] & WINED3DSP_WRITEMASK_3
)) {
3080 reg_mask
[size
] = 'w';
3084 if (i
== in_count
) {
3085 sprintf(destination
, "gl_FrontColor");
3086 } else if (i
== in_count
+ 1) {
3087 sprintf(destination
, "gl_FrontSecondaryColor");
3089 sprintf(destination
, "IN[%u]", i
);
3093 shader_addline(buffer
, "%s.%s = 0.0;\n", destination
, reg_mask
);
3095 shader_addline(buffer
, "%s.%s = vec%u(0.0);\n", destination
, reg_mask
, size
);
3100 HeapFree(GetProcessHeap(), 0, set
);
3103 static GLhandleARB
generate_param_reorder_function(IWineD3DVertexShader
*vertexshader
,
3104 IWineD3DPixelShader
*pixelshader
, const WineD3D_GL_Info
*gl_info
)
3106 GLhandleARB ret
= 0;
3107 IWineD3DVertexShaderImpl
*vs
= (IWineD3DVertexShaderImpl
*) vertexshader
;
3108 IWineD3DPixelShaderImpl
*ps
= (IWineD3DPixelShaderImpl
*) pixelshader
;
3109 IWineD3DDeviceImpl
*device
;
3110 DWORD vs_major
= WINED3DSHADER_VERSION_MAJOR(vs
->baseShader
.reg_maps
.shader_version
);
3111 DWORD ps_major
= ps
? WINED3DSHADER_VERSION_MAJOR(ps
->baseShader
.reg_maps
.shader_version
) : 0;
3113 SHADER_BUFFER buffer
;
3115 DWORD register_token
;
3116 DWORD usage
, usage_idx
, writemask
;
3118 const struct semantic
*semantics_out
, *semantics_in
;
3120 shader_buffer_init(&buffer
);
3122 shader_addline(&buffer
, "#version 120\n");
3124 if(vs_major
< 3 && ps_major
< 3) {
3125 /* That one is easy: The vertex shader writes to the builtin varyings, the pixel shader reads from them.
3126 * Take care about the texcoord .w fixup though if we're using the fixed function fragment pipeline
3128 device
= (IWineD3DDeviceImpl
*) vs
->baseShader
.device
;
3129 if((GLINFO_LOCATION
).set_texcoord_w
&& ps_major
== 0 && vs_major
> 0 &&
3130 !device
->frag_pipe
->ffp_proj_control
) {
3131 shader_addline(&buffer
, "void order_ps_input() {\n");
3132 for(i
= 0; i
< min(8, MAX_REG_TEXCRD
); i
++) {
3133 if(vs
->baseShader
.reg_maps
.texcoord_mask
[i
] != 0 &&
3134 vs
->baseShader
.reg_maps
.texcoord_mask
[i
] != WINED3DSP_WRITEMASK_ALL
) {
3135 shader_addline(&buffer
, "gl_TexCoord[%u].w = 1.0;\n", i
);
3138 shader_addline(&buffer
, "}\n");
3140 shader_addline(&buffer
, "void order_ps_input() { /* do nothing */ }\n");
3142 } else if(ps_major
< 3 && vs_major
>= 3) {
3143 /* The vertex shader writes to its own varyings, the pixel shader needs them in the builtin ones */
3144 semantics_out
= vs
->semantics_out
;
3146 shader_addline(&buffer
, "void order_ps_input(in vec4 OUT[%u]) {\n", MAX_REG_OUTPUT
);
3147 for(i
= 0; i
< MAX_REG_OUTPUT
; i
++) {
3148 usage_token
= semantics_out
[i
].usage
;
3149 if (!usage_token
) continue;
3150 register_token
= semantics_out
[i
].reg
;
3152 usage
= (usage_token
& WINED3DSP_DCL_USAGE_MASK
) >> WINED3DSP_DCL_USAGE_SHIFT
;
3153 usage_idx
= (usage_token
& WINED3DSP_DCL_USAGEINDEX_MASK
) >> WINED3DSP_DCL_USAGEINDEX_SHIFT
;
3154 writemask
= shader_glsl_get_write_mask(register_token
, reg_mask
);
3157 case WINED3DDECLUSAGE_COLOR
:
3159 shader_addline(&buffer
, "gl_FrontColor%s = OUT[%u]%s;\n", reg_mask
, i
, reg_mask
);
3160 else if (usage_idx
== 1)
3161 shader_addline(&buffer
, "gl_FrontSecondaryColor%s = OUT[%u]%s;\n", reg_mask
, i
, reg_mask
);
3164 case WINED3DDECLUSAGE_POSITION
:
3165 shader_addline(&buffer
, "gl_Position%s = OUT[%u]%s;\n", reg_mask
, i
, reg_mask
);
3168 case WINED3DDECLUSAGE_TEXCOORD
:
3169 if (usage_idx
< 8) {
3170 if(!(GLINFO_LOCATION
).set_texcoord_w
|| ps_major
> 0) writemask
|= WINED3DSP_WRITEMASK_3
;
3172 shader_addline(&buffer
, "gl_TexCoord[%u]%s = OUT[%u]%s;\n",
3173 usage_idx
, reg_mask
, i
, reg_mask
);
3174 if(!(writemask
& WINED3DSP_WRITEMASK_3
)) {
3175 shader_addline(&buffer
, "gl_TexCoord[%u].w = 1.0;\n", usage_idx
);
3180 case WINED3DDECLUSAGE_PSIZE
:
3181 shader_addline(&buffer
, "gl_PointSize = OUT[%u].x;\n", i
);
3184 case WINED3DDECLUSAGE_FOG
:
3185 shader_addline(&buffer
, "gl_FogFragCoord = OUT[%u].%c;\n", i
, reg_mask
[1]);
3192 shader_addline(&buffer
, "}\n");
3194 } else if(ps_major
>= 3 && vs_major
>= 3) {
3195 semantics_out
= vs
->semantics_out
;
3196 semantics_in
= ps
->semantics_in
;
3198 /* This one is tricky: a 3.0 pixel shader reads from a 3.0 vertex shader */
3199 shader_addline(&buffer
, "varying vec4 IN[%u];\n", GL_LIMITS(glsl_varyings
) / 4);
3200 shader_addline(&buffer
, "void order_ps_input(in vec4 OUT[%u]) {\n", MAX_REG_OUTPUT
);
3202 /* First, sort out position and point size. Those are not passed to the pixel shader */
3203 for(i
= 0; i
< MAX_REG_OUTPUT
; i
++) {
3204 usage_token
= semantics_out
[i
].usage
;
3205 if (!usage_token
) continue;
3206 register_token
= semantics_out
[i
].reg
;
3208 usage
= (usage_token
& WINED3DSP_DCL_USAGE_MASK
) >> WINED3DSP_DCL_USAGE_SHIFT
;
3209 usage_idx
= (usage_token
& WINED3DSP_DCL_USAGEINDEX_MASK
) >> WINED3DSP_DCL_USAGEINDEX_SHIFT
;
3210 shader_glsl_get_write_mask(register_token
, reg_mask
);
3213 case WINED3DDECLUSAGE_POSITION
:
3214 shader_addline(&buffer
, "gl_Position%s = OUT[%u]%s;\n", reg_mask
, i
, reg_mask
);
3217 case WINED3DDECLUSAGE_PSIZE
:
3218 shader_addline(&buffer
, "gl_PointSize = OUT[%u].x;\n", i
);
3226 /* Then, fix the pixel shader input */
3227 handle_ps3_input(&buffer
, semantics_in
, semantics_out
, gl_info
, ps
->input_reg_map
);
3229 shader_addline(&buffer
, "}\n");
3230 } else if(ps_major
>= 3 && vs_major
< 3) {
3231 semantics_in
= ps
->semantics_in
;
3233 shader_addline(&buffer
, "varying vec4 IN[%u];\n", GL_LIMITS(glsl_varyings
) / 4);
3234 shader_addline(&buffer
, "void order_ps_input() {\n");
3235 /* The vertex shader wrote to the builtin varyings. There is no need to figure out position and
3236 * point size, but we depend on the optimizers kindness to find out that the pixel shader doesn't
3237 * read gl_TexCoord and gl_ColorX, otherwise we'll run out of varyings
3239 handle_ps3_input(&buffer
, semantics_in
, NULL
, gl_info
, ps
->input_reg_map
);
3240 shader_addline(&buffer
, "}\n");
3242 ERR("Unexpected vertex and pixel shader version condition: vs: %d, ps: %d\n", vs_major
, ps_major
);
3245 ret
= GL_EXTCALL(glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB
));
3246 checkGLcall("glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB)");
3247 GL_EXTCALL(glShaderSourceARB(ret
, 1, (const char**)&buffer
.buffer
, NULL
));
3248 checkGLcall("glShaderSourceARB(ret, 1, (const char**)&buffer.buffer, NULL)");
3249 GL_EXTCALL(glCompileShaderARB(ret
));
3250 checkGLcall("glCompileShaderARB(ret)");
3252 shader_buffer_free(&buffer
);
3256 static void hardcode_local_constants(IWineD3DBaseShaderImpl
*shader
, const WineD3D_GL_Info
*gl_info
,
3257 GLhandleARB programId
, char prefix
)
3259 const local_constant
*lconst
;
3264 LIST_FOR_EACH_ENTRY(lconst
, &shader
->baseShader
.constantsF
, local_constant
, entry
) {
3265 value
= (const float *)lconst
->value
;
3266 snprintf(glsl_name
, sizeof(glsl_name
), "%cLC%u", prefix
, lconst
->idx
);
3267 tmp_loc
= GL_EXTCALL(glGetUniformLocationARB(programId
, glsl_name
));
3268 GL_EXTCALL(glUniform4fvARB(tmp_loc
, 1, value
));
3270 checkGLcall("Hardcoding local constants\n");
3273 /** Sets the GLSL program ID for the given pixel and vertex shader combination.
3274 * It sets the programId on the current StateBlock (because it should be called
3275 * inside of the DrawPrimitive() part of the render loop).
3277 * If a program for the given combination does not exist, create one, and store
3278 * the program in the hash table. If it creates a program, it will link the
3279 * given objects, too.
3281 static void set_glsl_shader_program(IWineD3DDevice
*iface
, BOOL use_ps
, BOOL use_vs
) {
3282 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
3283 struct shader_glsl_priv
*priv
= (struct shader_glsl_priv
*)This
->shader_priv
;
3284 const WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
3285 IWineD3DPixelShader
*pshader
= This
->stateBlock
->pixelShader
;
3286 IWineD3DVertexShader
*vshader
= This
->stateBlock
->vertexShader
;
3287 struct glsl_shader_prog_link
*entry
= NULL
;
3288 GLhandleARB programId
= 0;
3289 GLhandleARB reorder_shader_id
= 0;
3292 GLhandleARB vshader_id
, pshader_id
;
3293 struct ps_compile_args compile_args
;
3296 IWineD3DVertexShaderImpl_CompileShader(vshader
);
3297 vshader_id
= ((IWineD3DVertexShaderImpl
*)vshader
)->prgId
;
3302 find_ps_compile_args((IWineD3DPixelShaderImpl
*)This
->stateBlock
->pixelShader
, This
->stateBlock
, &compile_args
);
3304 /* FIXME: Do we really have to spend CPU cycles to generate a few zeroed bytes? */
3305 memset(&compile_args
, 0, sizeof(compile_args
));
3307 entry
= get_glsl_program_entry(priv
, vshader_id
, pshader
, &compile_args
);
3309 priv
->glsl_program
= entry
;
3313 /* If we get to this point, then no matching program exists, so we create one */
3314 programId
= GL_EXTCALL(glCreateProgramObjectARB());
3315 TRACE("Created new GLSL shader program %u\n", programId
);
3317 /* Create the entry */
3318 entry
= HeapAlloc(GetProcessHeap(), 0, sizeof(struct glsl_shader_prog_link
));
3319 entry
->programId
= programId
;
3320 entry
->vshader
= vshader_id
;
3321 entry
->pshader
= pshader
;
3322 entry
->ps_args
= compile_args
;
3323 entry
->constant_version
= 0;
3324 /* Add the hash table entry */
3325 add_glsl_program_entry(priv
, entry
);
3327 /* Set the current program */
3328 priv
->glsl_program
= entry
;
3330 /* Attach GLSL vshader */
3332 int max_attribs
= 16; /* TODO: Will this always be the case? It is at the moment... */
3335 reorder_shader_id
= generate_param_reorder_function(vshader
, pshader
, gl_info
);
3336 TRACE("Attaching GLSL shader object %u to program %u\n", reorder_shader_id
, programId
);
3337 GL_EXTCALL(glAttachObjectARB(programId
, reorder_shader_id
));
3338 checkGLcall("glAttachObjectARB");
3339 /* Flag the reorder function for deletion, then it will be freed automatically when the program
3342 GL_EXTCALL(glDeleteObjectARB(reorder_shader_id
));
3344 TRACE("Attaching GLSL shader object %u to program %u\n", vshader_id
, programId
);
3345 GL_EXTCALL(glAttachObjectARB(programId
, vshader_id
));
3346 checkGLcall("glAttachObjectARB");
3348 /* Bind vertex attributes to a corresponding index number to match
3349 * the same index numbers as ARB_vertex_programs (makes loading
3350 * vertex attributes simpler). With this method, we can use the
3351 * exact same code to load the attributes later for both ARB and
3354 * We have to do this here because we need to know the Program ID
3355 * in order to make the bindings work, and it has to be done prior
3356 * to linking the GLSL program. */
3357 for (i
= 0; i
< max_attribs
; ++i
) {
3358 if (((IWineD3DBaseShaderImpl
*)vshader
)->baseShader
.reg_maps
.attributes
[i
]) {
3359 snprintf(tmp_name
, sizeof(tmp_name
), "attrib%i", i
);
3360 GL_EXTCALL(glBindAttribLocationARB(programId
, i
, tmp_name
));
3363 checkGLcall("glBindAttribLocationARB");
3365 list_add_head(&((IWineD3DBaseShaderImpl
*)vshader
)->baseShader
.linked_programs
, &entry
->vshader_entry
);
3369 pshader_id
= find_gl_pshader((IWineD3DPixelShaderImpl
*) pshader
, &compile_args
);
3374 /* Attach GLSL pshader */
3376 TRACE("Attaching GLSL shader object %u to program %u\n", pshader_id
, programId
);
3377 GL_EXTCALL(glAttachObjectARB(programId
, pshader_id
));
3378 checkGLcall("glAttachObjectARB");
3380 list_add_head(&((IWineD3DBaseShaderImpl
*)pshader
)->baseShader
.linked_programs
, &entry
->pshader_entry
);
3383 /* Link the program */
3384 TRACE("Linking GLSL shader program %u\n", programId
);
3385 GL_EXTCALL(glLinkProgramARB(programId
));
3386 print_glsl_info_log(&GLINFO_LOCATION
, programId
);
3388 entry
->vuniformF_locations
= HeapAlloc(GetProcessHeap(), 0, sizeof(GLhandleARB
) * GL_LIMITS(vshader_constantsF
));
3389 for (i
= 0; i
< GL_LIMITS(vshader_constantsF
); ++i
) {
3390 snprintf(glsl_name
, sizeof(glsl_name
), "VC[%i]", i
);
3391 entry
->vuniformF_locations
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, glsl_name
));
3393 for (i
= 0; i
< MAX_CONST_I
; ++i
) {
3394 snprintf(glsl_name
, sizeof(glsl_name
), "VI[%i]", i
);
3395 entry
->vuniformI_locations
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, glsl_name
));
3397 entry
->puniformF_locations
= HeapAlloc(GetProcessHeap(), 0, sizeof(GLhandleARB
) * GL_LIMITS(pshader_constantsF
));
3398 for (i
= 0; i
< GL_LIMITS(pshader_constantsF
); ++i
) {
3399 snprintf(glsl_name
, sizeof(glsl_name
), "PC[%i]", i
);
3400 entry
->puniformF_locations
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, glsl_name
));
3402 for (i
= 0; i
< MAX_CONST_I
; ++i
) {
3403 snprintf(glsl_name
, sizeof(glsl_name
), "PI[%i]", i
);
3404 entry
->puniformI_locations
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, glsl_name
));
3408 for(i
= 0; i
< ((IWineD3DPixelShaderImpl
*)pshader
)->numbumpenvmatconsts
; i
++) {
3410 sprintf(name
, "bumpenvmat%d", ((IWineD3DPixelShaderImpl
*)pshader
)->bumpenvmatconst
[i
].texunit
);
3411 entry
->bumpenvmat_location
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, name
));
3412 sprintf(name
, "luminancescale%d", ((IWineD3DPixelShaderImpl
*)pshader
)->luminanceconst
[i
].texunit
);
3413 entry
->luminancescale_location
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, name
));
3414 sprintf(name
, "luminanceoffset%d", ((IWineD3DPixelShaderImpl
*)pshader
)->luminanceconst
[i
].texunit
);
3415 entry
->luminanceoffset_location
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, name
));
3420 entry
->posFixup_location
= GL_EXTCALL(glGetUniformLocationARB(programId
, "posFixup"));
3421 entry
->ycorrection_location
= GL_EXTCALL(glGetUniformLocationARB(programId
, "ycorrection"));
3422 checkGLcall("Find glsl program uniform locations");
3425 && WINED3DSHADER_VERSION_MAJOR(((IWineD3DPixelShaderImpl
*)pshader
)->baseShader
.reg_maps
.shader_version
) >= 3
3426 && ((IWineD3DPixelShaderImpl
*)pshader
)->declared_in_count
> GL_LIMITS(glsl_varyings
) / 4)
3428 TRACE("Shader %d needs vertex color clamping disabled\n", programId
);
3429 entry
->vertex_color_clamp
= GL_FALSE
;
3431 entry
->vertex_color_clamp
= GL_FIXED_ONLY_ARB
;
3434 /* Set the shader to allow uniform loading on it */
3435 GL_EXTCALL(glUseProgramObjectARB(programId
));
3436 checkGLcall("glUseProgramObjectARB(programId)");
3438 /* Load the vertex and pixel samplers now. The function that finds the mappings makes sure
3439 * that it stays the same for each vertexshader-pixelshader pair(=linked glsl program). If
3440 * a pshader with fixed function pipeline is used there are no vertex samplers, and if a
3441 * vertex shader with fixed function pixel processing is used we make sure that the card
3442 * supports enough samplers to allow the max number of vertex samplers with all possible
3443 * fixed function fragment processing setups. So once the program is linked these samplers
3447 /* Load vertex shader samplers */
3448 shader_glsl_load_vsamplers(gl_info
, (IWineD3DStateBlock
*)This
->stateBlock
, programId
);
3451 /* Load pixel shader samplers */
3452 shader_glsl_load_psamplers(gl_info
, (IWineD3DStateBlock
*)This
->stateBlock
, programId
);
3455 /* If the local constants do not have to be loaded with the environment constants,
3456 * load them now to have them hardcoded in the GLSL program. This saves some CPU cycles
3459 if(pshader
&& !((IWineD3DPixelShaderImpl
*)pshader
)->baseShader
.load_local_constsF
) {
3460 hardcode_local_constants((IWineD3DBaseShaderImpl
*) pshader
, gl_info
, programId
, 'P');
3462 if(vshader
&& !((IWineD3DVertexShaderImpl
*)vshader
)->baseShader
.load_local_constsF
) {
3463 hardcode_local_constants((IWineD3DBaseShaderImpl
*) vshader
, gl_info
, programId
, 'V');
3467 static GLhandleARB
create_glsl_blt_shader(const WineD3D_GL_Info
*gl_info
, enum tex_types tex_type
)
3469 GLhandleARB program_id
;
3470 GLhandleARB vshader_id
, pshader_id
;
3471 static const char *blt_vshader
[] =
3476 " gl_Position = gl_Vertex;\n"
3477 " gl_FrontColor = vec4(1.0);\n"
3478 " gl_TexCoord[0] = gl_MultiTexCoord0;\n"
3482 static const char *blt_pshaders
[tex_type_count
] =
3488 "uniform sampler2D sampler;\n"
3491 " gl_FragDepth = texture2D(sampler, gl_TexCoord[0].xy).x;\n"
3497 "uniform samplerCube sampler;\n"
3500 " gl_FragDepth = textureCube(sampler, gl_TexCoord[0].xyz).x;\n"
3504 "#extension GL_ARB_texture_rectangle : enable\n"
3505 "uniform sampler2DRect sampler;\n"
3508 " gl_FragDepth = texture2DRect(sampler, gl_TexCoord[0].xy).x;\n"
3512 if (!blt_pshaders
[tex_type
])
3514 FIXME("tex_type %#x not supported\n", tex_type
);
3518 vshader_id
= GL_EXTCALL(glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB
));
3519 GL_EXTCALL(glShaderSourceARB(vshader_id
, 1, blt_vshader
, NULL
));
3520 GL_EXTCALL(glCompileShaderARB(vshader_id
));
3522 pshader_id
= GL_EXTCALL(glCreateShaderObjectARB(GL_FRAGMENT_SHADER_ARB
));
3523 GL_EXTCALL(glShaderSourceARB(pshader_id
, 1, &blt_pshaders
[tex_type
], NULL
));
3524 GL_EXTCALL(glCompileShaderARB(pshader_id
));
3526 program_id
= GL_EXTCALL(glCreateProgramObjectARB());
3527 GL_EXTCALL(glAttachObjectARB(program_id
, vshader_id
));
3528 GL_EXTCALL(glAttachObjectARB(program_id
, pshader_id
));
3529 GL_EXTCALL(glLinkProgramARB(program_id
));
3531 print_glsl_info_log(&GLINFO_LOCATION
, program_id
);
3533 /* Once linked we can mark the shaders for deletion. They will be deleted once the program
3536 GL_EXTCALL(glDeleteObjectARB(vshader_id
));
3537 GL_EXTCALL(glDeleteObjectARB(pshader_id
));
3541 static void shader_glsl_select(IWineD3DDevice
*iface
, BOOL usePS
, BOOL useVS
) {
3542 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
3543 struct shader_glsl_priv
*priv
= (struct shader_glsl_priv
*)This
->shader_priv
;
3544 const WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
3545 GLhandleARB program_id
= 0;
3546 GLenum old_vertex_color_clamp
, current_vertex_color_clamp
;
3548 old_vertex_color_clamp
= priv
->glsl_program
? priv
->glsl_program
->vertex_color_clamp
: GL_FIXED_ONLY_ARB
;
3550 if (useVS
|| usePS
) set_glsl_shader_program(iface
, usePS
, useVS
);
3551 else priv
->glsl_program
= NULL
;
3553 current_vertex_color_clamp
= priv
->glsl_program
? priv
->glsl_program
->vertex_color_clamp
: GL_FIXED_ONLY_ARB
;
3555 if (old_vertex_color_clamp
!= current_vertex_color_clamp
) {
3556 if (GL_SUPPORT(ARB_COLOR_BUFFER_FLOAT
)) {
3557 GL_EXTCALL(glClampColorARB(GL_CLAMP_VERTEX_COLOR_ARB
, current_vertex_color_clamp
));
3558 checkGLcall("glClampColorARB");
3560 FIXME("vertex color clamp needs to be changed, but extension not supported.\n");
3564 program_id
= priv
->glsl_program
? priv
->glsl_program
->programId
: 0;
3565 if (program_id
) TRACE("Using GLSL program %u\n", program_id
);
3566 GL_EXTCALL(glUseProgramObjectARB(program_id
));
3567 checkGLcall("glUseProgramObjectARB");
3570 static void shader_glsl_select_depth_blt(IWineD3DDevice
*iface
, enum tex_types tex_type
) {
3571 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
3572 const WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
3573 struct shader_glsl_priv
*priv
= (struct shader_glsl_priv
*) This
->shader_priv
;
3574 GLhandleARB
*blt_program
= &priv
->depth_blt_program
[tex_type
];
3576 if (!*blt_program
) {
3578 *blt_program
= create_glsl_blt_shader(gl_info
, tex_type
);
3579 loc
= GL_EXTCALL(glGetUniformLocationARB(*blt_program
, "sampler"));
3580 GL_EXTCALL(glUseProgramObjectARB(*blt_program
));
3581 GL_EXTCALL(glUniform1iARB(loc
, 0));
3583 GL_EXTCALL(glUseProgramObjectARB(*blt_program
));
3587 static void shader_glsl_deselect_depth_blt(IWineD3DDevice
*iface
) {
3588 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
3589 const WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
3590 struct shader_glsl_priv
*priv
= (struct shader_glsl_priv
*) This
->shader_priv
;
3591 GLhandleARB program_id
;
3593 program_id
= priv
->glsl_program
? priv
->glsl_program
->programId
: 0;
3594 if (program_id
) TRACE("Using GLSL program %u\n", program_id
);
3596 GL_EXTCALL(glUseProgramObjectARB(program_id
));
3597 checkGLcall("glUseProgramObjectARB");
3600 static void shader_glsl_destroy(IWineD3DBaseShader
*iface
) {
3601 const struct list
*linked_programs
;
3602 IWineD3DBaseShaderImpl
*This
= (IWineD3DBaseShaderImpl
*) iface
;
3603 IWineD3DDeviceImpl
*device
= (IWineD3DDeviceImpl
*)This
->baseShader
.device
;
3604 struct shader_glsl_priv
*priv
= (struct shader_glsl_priv
*)device
->shader_priv
;
3605 const WineD3D_GL_Info
*gl_info
= &device
->adapter
->gl_info
;
3606 IWineD3DPixelShaderImpl
*ps
= NULL
;
3607 IWineD3DVertexShaderImpl
*vs
= NULL
;
3609 /* Note: Do not use QueryInterface here to find out which shader type this is because this code
3610 * can be called from IWineD3DBaseShader::Release
3612 char pshader
= shader_is_pshader_version(This
->baseShader
.reg_maps
.shader_version
);
3615 ps
= (IWineD3DPixelShaderImpl
*) This
;
3616 if(ps
->num_gl_shaders
== 0) return;
3618 vs
= (IWineD3DVertexShaderImpl
*) This
;
3619 if(vs
->prgId
== 0) return;
3622 linked_programs
= &This
->baseShader
.linked_programs
;
3624 TRACE("Deleting linked programs\n");
3625 if (linked_programs
->next
) {
3626 struct glsl_shader_prog_link
*entry
, *entry2
;
3629 LIST_FOR_EACH_ENTRY_SAFE(entry
, entry2
, linked_programs
, struct glsl_shader_prog_link
, pshader_entry
) {
3630 delete_glsl_program_entry(priv
, gl_info
, entry
);
3633 LIST_FOR_EACH_ENTRY_SAFE(entry
, entry2
, linked_programs
, struct glsl_shader_prog_link
, vshader_entry
) {
3634 delete_glsl_program_entry(priv
, gl_info
, entry
);
3643 for(i
= 0; i
< ps
->num_gl_shaders
; i
++) {
3644 TRACE("deleting pshader %u\n", ps
->gl_shaders
[i
].prgId
);
3645 GL_EXTCALL(glDeleteObjectARB(ps
->gl_shaders
[i
].prgId
));
3646 checkGLcall("glDeleteObjectARB");
3649 HeapFree(GetProcessHeap(), 0, ps
->gl_shaders
);
3650 ps
->gl_shaders
= NULL
;
3651 ps
->num_gl_shaders
= 0;
3653 TRACE("Deleting shader object %u\n", vs
->prgId
);
3655 GL_EXTCALL(glDeleteObjectARB(vs
->prgId
));
3656 checkGLcall("glDeleteObjectARB");
3659 vs
->baseShader
.is_compiled
= FALSE
;
3663 static unsigned int glsl_program_key_hash(const void *key
)
3665 const glsl_program_key_t
*k
= (const glsl_program_key_t
*)key
;
3667 unsigned int hash
= k
->vshader
| ((DWORD_PTR
) k
->pshader
) << 16;
3668 hash
+= ~(hash
<< 15);
3669 hash
^= (hash
>> 10);
3670 hash
+= (hash
<< 3);
3671 hash
^= (hash
>> 6);
3672 hash
+= ~(hash
<< 11);
3673 hash
^= (hash
>> 16);
3678 static BOOL
glsl_program_key_compare(const void *keya
, const void *keyb
)
3680 const glsl_program_key_t
*ka
= (const glsl_program_key_t
*)keya
;
3681 const glsl_program_key_t
*kb
= (const glsl_program_key_t
*)keyb
;
3683 return ka
->vshader
== kb
->vshader
&& ka
->pshader
== kb
->pshader
&&
3684 (memcmp(&ka
->ps_args
, &kb
->ps_args
, sizeof(kb
->ps_args
)) == 0);
3687 static BOOL
constant_heap_init(struct constant_heap
*heap
, unsigned int constant_count
)
3689 SIZE_T size
= (constant_count
+ 1) * sizeof(*heap
->entries
) + constant_count
* sizeof(*heap
->positions
);
3690 void *mem
= HeapAlloc(GetProcessHeap(), 0, size
);
3694 ERR("Failed to allocate memory\n");
3698 heap
->entries
= mem
;
3699 heap
->entries
[1].version
= 0;
3700 heap
->positions
= (unsigned int *)(heap
->entries
+ constant_count
+ 1);
3706 static void constant_heap_free(struct constant_heap
*heap
)
3708 HeapFree(GetProcessHeap(), 0, heap
->entries
);
3711 static HRESULT
shader_glsl_alloc(IWineD3DDevice
*iface
) {
3712 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
3713 const WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
3714 struct shader_glsl_priv
*priv
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(struct shader_glsl_priv
));
3715 SIZE_T stack_size
= wined3d_log2i(max(GL_LIMITS(vshader_constantsF
), GL_LIMITS(pshader_constantsF
))) + 1;
3717 priv
->stack
= HeapAlloc(GetProcessHeap(), 0, stack_size
* sizeof(*priv
->stack
));
3720 ERR("Failed to allocate memory.\n");
3721 HeapFree(GetProcessHeap(), 0, priv
);
3722 return E_OUTOFMEMORY
;
3725 if (!constant_heap_init(&priv
->vconst_heap
, GL_LIMITS(vshader_constantsF
)))
3727 ERR("Failed to initialize vertex shader constant heap\n");
3728 HeapFree(GetProcessHeap(), 0, priv
->stack
);
3729 HeapFree(GetProcessHeap(), 0, priv
);
3730 return E_OUTOFMEMORY
;
3733 if (!constant_heap_init(&priv
->pconst_heap
, GL_LIMITS(pshader_constantsF
)))
3735 ERR("Failed to initialize pixel shader constant heap\n");
3736 constant_heap_free(&priv
->vconst_heap
);
3737 HeapFree(GetProcessHeap(), 0, priv
->stack
);
3738 HeapFree(GetProcessHeap(), 0, priv
);
3739 return E_OUTOFMEMORY
;
3742 priv
->glsl_program_lookup
= hash_table_create(glsl_program_key_hash
, glsl_program_key_compare
);
3743 priv
->next_constant_version
= 1;
3745 This
->shader_priv
= priv
;
3749 static void shader_glsl_free(IWineD3DDevice
*iface
) {
3750 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
3751 const WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
3752 struct shader_glsl_priv
*priv
= (struct shader_glsl_priv
*)This
->shader_priv
;
3755 for (i
= 0; i
< tex_type_count
; ++i
)
3757 if (priv
->depth_blt_program
[i
])
3759 GL_EXTCALL(glDeleteObjectARB(priv
->depth_blt_program
[i
]));
3763 hash_table_destroy(priv
->glsl_program_lookup
, NULL
, NULL
);
3764 constant_heap_free(&priv
->pconst_heap
);
3765 constant_heap_free(&priv
->vconst_heap
);
3767 HeapFree(GetProcessHeap(), 0, This
->shader_priv
);
3768 This
->shader_priv
= NULL
;
3771 static BOOL
shader_glsl_dirty_const(IWineD3DDevice
*iface
) {
3772 /* TODO: GL_EXT_bindable_uniform can be used to share constants across shaders */
3776 static GLuint
shader_glsl_generate_pshader(IWineD3DPixelShader
*iface
, SHADER_BUFFER
*buffer
, const struct ps_compile_args
*args
) {
3777 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*)iface
;
3778 const struct shader_reg_maps
*reg_maps
= &This
->baseShader
.reg_maps
;
3779 CONST DWORD
*function
= This
->baseShader
.function
;
3780 const char *fragcolor
;
3781 const WineD3D_GL_Info
*gl_info
= &((IWineD3DDeviceImpl
*)This
->baseShader
.device
)->adapter
->gl_info
;
3783 /* Create the hw GLSL shader object and assign it as the shader->prgId */
3784 GLhandleARB shader_obj
= GL_EXTCALL(glCreateShaderObjectARB(GL_FRAGMENT_SHADER_ARB
));
3786 shader_addline(buffer
, "#version 120\n");
3788 if (GL_SUPPORT(ARB_DRAW_BUFFERS
)) {
3789 shader_addline(buffer
, "#extension GL_ARB_draw_buffers : enable\n");
3791 if (GL_SUPPORT(ARB_TEXTURE_RECTANGLE
)) {
3792 /* The spec says that it doesn't have to be explicitly enabled, but the nvidia
3793 * drivers write a warning if we don't do so
3795 shader_addline(buffer
, "#extension GL_ARB_texture_rectangle : enable\n");
3798 /* Base Declarations */
3799 shader_generate_glsl_declarations( (IWineD3DBaseShader
*) This
, reg_maps
, buffer
, &GLINFO_LOCATION
, args
);
3801 /* Pack 3.0 inputs */
3802 if (reg_maps
->shader_version
>= WINED3DPS_VERSION(3,0) && args
->vp_mode
!= vertexshader
) {
3803 pshader_glsl_input_pack(buffer
, This
->semantics_in
, iface
, args
->vp_mode
);
3806 /* Base Shader Body */
3807 shader_generate_main( (IWineD3DBaseShader
*) This
, buffer
, reg_maps
, function
);
3809 /* Pixel shaders < 2.0 place the resulting color in R0 implicitly */
3810 if (reg_maps
->shader_version
< WINED3DPS_VERSION(2,0))
3812 /* Some older cards like GeforceFX ones don't support multiple buffers, so also not gl_FragData */
3813 if(GL_SUPPORT(ARB_DRAW_BUFFERS
))
3814 shader_addline(buffer
, "gl_FragData[0] = R0;\n");
3816 shader_addline(buffer
, "gl_FragColor = R0;\n");
3819 if(GL_SUPPORT(ARB_DRAW_BUFFERS
)) {
3820 fragcolor
= "gl_FragData[0]";
3822 fragcolor
= "gl_FragColor";
3824 if(args
->srgb_correction
) {
3825 shader_addline(buffer
, "tmp0.xyz = pow(%s.xyz, vec3(%f, %f, %f)) * vec3(%f, %f, %f) - vec3(%f, %f, %f);\n",
3826 fragcolor
, srgb_pow
, srgb_pow
, srgb_pow
, srgb_mul_high
, srgb_mul_high
, srgb_mul_high
,
3827 srgb_sub_high
, srgb_sub_high
, srgb_sub_high
);
3828 shader_addline(buffer
, "tmp1.xyz = %s.xyz * srgb_mul_low.xyz;\n", fragcolor
);
3829 shader_addline(buffer
, "%s.x = %s.x < srgb_comparison.x ? tmp1.x : tmp0.x;\n", fragcolor
, fragcolor
);
3830 shader_addline(buffer
, "%s.y = %s.y < srgb_comparison.y ? tmp1.y : tmp0.y;\n", fragcolor
, fragcolor
);
3831 shader_addline(buffer
, "%s.z = %s.z < srgb_comparison.z ? tmp1.z : tmp0.z;\n", fragcolor
, fragcolor
);
3832 shader_addline(buffer
, "%s = clamp(%s, 0.0, 1.0);\n", fragcolor
, fragcolor
);
3834 /* Pixel shader < 3.0 do not replace the fog stage.
3835 * This implements linear fog computation and blending.
3836 * TODO: non linear fog
3837 * NOTE: gl_Fog.start and gl_Fog.end don't hold fog start s and end e but
3838 * -1/(e-s) and e/(e-s) respectively.
3840 if(reg_maps
->shader_version
< WINED3DPS_VERSION(3,0)) {
3842 case FOG_OFF
: break;
3844 shader_addline(buffer
, "float Fog = clamp(gl_FogFragCoord * gl_Fog.start + gl_Fog.end, 0.0, 1.0);\n");
3845 shader_addline(buffer
, "%s.xyz = mix(gl_Fog.color.xyz, %s.xyz, Fog);\n", fragcolor
, fragcolor
);
3848 FIXME("Implement EXP fog in glsl\n");
3851 FIXME("Implement EXP2 fog in glsl\n");
3856 shader_addline(buffer
, "}\n");
3858 TRACE("Compiling shader object %u\n", shader_obj
);
3859 GL_EXTCALL(glShaderSourceARB(shader_obj
, 1, (const char**)&buffer
->buffer
, NULL
));
3860 GL_EXTCALL(glCompileShaderARB(shader_obj
));
3861 print_glsl_info_log(&GLINFO_LOCATION
, shader_obj
);
3863 /* Store the shader object */
3867 static void shader_glsl_generate_vshader(IWineD3DVertexShader
*iface
, SHADER_BUFFER
*buffer
) {
3868 IWineD3DVertexShaderImpl
*This
= (IWineD3DVertexShaderImpl
*)iface
;
3869 const struct shader_reg_maps
*reg_maps
= &This
->baseShader
.reg_maps
;
3870 CONST DWORD
*function
= This
->baseShader
.function
;
3871 const WineD3D_GL_Info
*gl_info
= &((IWineD3DDeviceImpl
*)This
->baseShader
.device
)->adapter
->gl_info
;
3873 /* Create the hw GLSL shader program and assign it as the shader->prgId */
3874 GLhandleARB shader_obj
= GL_EXTCALL(glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB
));
3876 shader_addline(buffer
, "#version 120\n");
3878 /* Base Declarations */
3879 shader_generate_glsl_declarations( (IWineD3DBaseShader
*) This
, reg_maps
, buffer
, &GLINFO_LOCATION
, NULL
);
3881 /* Base Shader Body */
3882 shader_generate_main( (IWineD3DBaseShader
*) This
, buffer
, reg_maps
, function
);
3884 /* Unpack 3.0 outputs */
3885 if (reg_maps
->shader_version
>= WINED3DVS_VERSION(3,0)) shader_addline(buffer
, "order_ps_input(OUT);\n");
3886 else shader_addline(buffer
, "order_ps_input();\n");
3888 /* If this shader doesn't use fog copy the z coord to the fog coord so that we can use table fog */
3890 shader_addline(buffer
, "gl_FogFragCoord = gl_Position.z;\n");
3892 /* Write the final position.
3894 * OpenGL coordinates specify the center of the pixel while d3d coords specify
3895 * the corner. The offsets are stored in z and w in posFixup. posFixup.y contains
3896 * 1.0 or -1.0 to turn the rendering upside down for offscreen rendering. PosFixup.x
3897 * contains 1.0 to allow a mad.
3899 shader_addline(buffer
, "gl_Position.y = gl_Position.y * posFixup.y;\n");
3900 shader_addline(buffer
, "gl_Position.xy += posFixup.zw * gl_Position.ww;\n");
3902 /* Z coord [0;1]->[-1;1] mapping, see comment in transform_projection in state.c
3904 * Basically we want (in homogeneous coordinates) z = z * 2 - 1. However, shaders are run
3905 * before the homogeneous divide, so we have to take the w into account: z = ((z / w) * 2 - 1) * w,
3906 * which is the same as z = z * 2 - w.
3908 shader_addline(buffer
, "gl_Position.z = gl_Position.z * 2.0 - gl_Position.w;\n");
3910 shader_addline(buffer
, "}\n");
3912 TRACE("Compiling shader object %u\n", shader_obj
);
3913 GL_EXTCALL(glShaderSourceARB(shader_obj
, 1, (const char**)&buffer
->buffer
, NULL
));
3914 GL_EXTCALL(glCompileShaderARB(shader_obj
));
3915 print_glsl_info_log(&GLINFO_LOCATION
, shader_obj
);
3917 /* Store the shader object */
3918 This
->prgId
= shader_obj
;
3921 static void shader_glsl_get_caps(WINED3DDEVTYPE devtype
, const WineD3D_GL_Info
*gl_info
, struct shader_caps
*pCaps
)
3923 /* Nvidia Geforce6/7 or Ati R4xx/R5xx cards with GLSL support, support VS 3.0 but older Nvidia/Ati
3924 * models with GLSL support only support 2.0. In case of nvidia we can detect VS 2.0 support using
3925 * vs_nv_version which is based on NV_vertex_program.
3926 * For Ati cards there's no way using glsl (it abstracts the lowlevel info away) and also not
3927 * using ARB_vertex_program. It is safe to assume that when a card supports pixel shader 2.0 it
3928 * supports vertex shader 2.0 too and the way around. We can detect ps2.0 using the maximum number
3929 * of native instructions, so use that here. For more info see the pixel shader versioning code below.
3931 if((GLINFO_LOCATION
.vs_nv_version
== VS_VERSION_20
) || (GLINFO_LOCATION
.ps_arb_max_instructions
<= 512))
3932 pCaps
->VertexShaderVersion
= WINED3DVS_VERSION(2,0);
3934 pCaps
->VertexShaderVersion
= WINED3DVS_VERSION(3,0);
3935 TRACE_(d3d_caps
)("Hardware vertex shader version %d.%d enabled (GLSL)\n", (pCaps
->VertexShaderVersion
>> 8) & 0xff, pCaps
->VertexShaderVersion
& 0xff);
3936 pCaps
->MaxVertexShaderConst
= GL_LIMITS(vshader_constantsF
);
3938 /* Older DX9-class videocards (GeforceFX / Radeon >9500/X*00) only support pixel shader 2.0/2.0a/2.0b.
3939 * In OpenGL the extensions related to GLSL abstract lowlevel GL info away which is needed
3940 * to distinguish between 2.0 and 3.0 (and 2.0a/2.0b). In case of Nvidia we use their fragment
3941 * program extensions. On other hardware including ATI GL_ARB_fragment_program offers the info
3942 * in max native instructions. Intel and others also offer the info in this extension but they
3943 * don't support GLSL (at least on Windows).
3945 * PS2.0 requires at least 96 instructions, 2.0a/2.0b go up to 512. Assume that if the number
3946 * of instructions is 512 or less we have to do with ps2.0 hardware.
3947 * 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.
3949 if((GLINFO_LOCATION
.ps_nv_version
== PS_VERSION_20
) || (GLINFO_LOCATION
.ps_arb_max_instructions
<= 512))
3950 pCaps
->PixelShaderVersion
= WINED3DPS_VERSION(2,0);
3952 pCaps
->PixelShaderVersion
= WINED3DPS_VERSION(3,0);
3954 /* FIXME: The following line is card dependent. -8.0 to 8.0 is the
3955 * Direct3D minimum requirement.
3957 * Both GL_ARB_fragment_program and GLSL require a "maximum representable magnitude"
3958 * of colors to be 2^10, and 2^32 for other floats. Should we use 1024 here?
3960 * The problem is that the refrast clamps temporary results in the shader to
3961 * [-MaxValue;+MaxValue]. If the card's max value is bigger than the one we advertize here,
3962 * then applications may miss the clamping behavior. On the other hand, if it is smaller,
3963 * the shader will generate incorrect results too. Unfortunately, GL deliberately doesn't
3964 * offer a way to query this.
3966 pCaps
->PixelShader1xMaxValue
= 8.0;
3967 TRACE_(d3d_caps
)("Hardware pixel shader version %d.%d enabled (GLSL)\n", (pCaps
->PixelShaderVersion
>> 8) & 0xff, pCaps
->PixelShaderVersion
& 0xff);
3970 static BOOL
shader_glsl_color_fixup_supported(struct color_fixup_desc fixup
)
3972 if (TRACE_ON(d3d_shader
) && TRACE_ON(d3d
))
3974 TRACE("Checking support for fixup:\n");
3975 dump_color_fixup_desc(fixup
);
3978 /* We support everything except YUV conversions. */
3979 if (!is_yuv_fixup(fixup
))
3985 TRACE("[FAILED]\n");
3989 static const SHADER_HANDLER shader_glsl_instruction_handler_table
[WINED3DSIH_TABLE_SIZE
] =
3991 /* WINED3DSIH_ABS */ shader_glsl_map2gl
,
3992 /* WINED3DSIH_ADD */ shader_glsl_arith
,
3993 /* WINED3DSIH_BEM */ pshader_glsl_bem
,
3994 /* WINED3DSIH_BREAK */ shader_glsl_break
,
3995 /* WINED3DSIH_BREAKC */ shader_glsl_breakc
,
3996 /* WINED3DSIH_BREAKP */ NULL
,
3997 /* WINED3DSIH_CALL */ shader_glsl_call
,
3998 /* WINED3DSIH_CALLNZ */ shader_glsl_callnz
,
3999 /* WINED3DSIH_CMP */ shader_glsl_cmp
,
4000 /* WINED3DSIH_CND */ shader_glsl_cnd
,
4001 /* WINED3DSIH_CRS */ shader_glsl_cross
,
4002 /* WINED3DSIH_DCL */ NULL
,
4003 /* WINED3DSIH_DEF */ NULL
,
4004 /* WINED3DSIH_DEFB */ NULL
,
4005 /* WINED3DSIH_DEFI */ NULL
,
4006 /* WINED3DSIH_DP2ADD */ pshader_glsl_dp2add
,
4007 /* WINED3DSIH_DP3 */ shader_glsl_dot
,
4008 /* WINED3DSIH_DP4 */ shader_glsl_dot
,
4009 /* WINED3DSIH_DST */ shader_glsl_dst
,
4010 /* WINED3DSIH_DSX */ shader_glsl_map2gl
,
4011 /* WINED3DSIH_DSY */ shader_glsl_map2gl
,
4012 /* WINED3DSIH_ELSE */ shader_glsl_else
,
4013 /* WINED3DSIH_ENDIF */ shader_glsl_end
,
4014 /* WINED3DSIH_ENDLOOP */ shader_glsl_end
,
4015 /* WINED3DSIH_ENDREP */ shader_glsl_end
,
4016 /* WINED3DSIH_EXP */ shader_glsl_map2gl
,
4017 /* WINED3DSIH_EXPP */ shader_glsl_expp
,
4018 /* WINED3DSIH_FRC */ shader_glsl_map2gl
,
4019 /* WINED3DSIH_IF */ shader_glsl_if
,
4020 /* WINED3DSIH_IFC */ shader_glsl_ifc
,
4021 /* WINED3DSIH_LABEL */ shader_glsl_label
,
4022 /* WINED3DSIH_LIT */ shader_glsl_lit
,
4023 /* WINED3DSIH_LOG */ shader_glsl_log
,
4024 /* WINED3DSIH_LOGP */ shader_glsl_log
,
4025 /* WINED3DSIH_LOOP */ shader_glsl_loop
,
4026 /* WINED3DSIH_LRP */ shader_glsl_lrp
,
4027 /* WINED3DSIH_M3x2 */ shader_glsl_mnxn
,
4028 /* WINED3DSIH_M3x3 */ shader_glsl_mnxn
,
4029 /* WINED3DSIH_M3x4 */ shader_glsl_mnxn
,
4030 /* WINED3DSIH_M4x3 */ shader_glsl_mnxn
,
4031 /* WINED3DSIH_M4x4 */ shader_glsl_mnxn
,
4032 /* WINED3DSIH_MAD */ shader_glsl_mad
,
4033 /* WINED3DSIH_MAX */ shader_glsl_map2gl
,
4034 /* WINED3DSIH_MIN */ shader_glsl_map2gl
,
4035 /* WINED3DSIH_MOV */ shader_glsl_mov
,
4036 /* WINED3DSIH_MOVA */ shader_glsl_mov
,
4037 /* WINED3DSIH_MUL */ shader_glsl_arith
,
4038 /* WINED3DSIH_NOP */ NULL
,
4039 /* WINED3DSIH_NRM */ shader_glsl_map2gl
,
4040 /* WINED3DSIH_PHASE */ NULL
,
4041 /* WINED3DSIH_POW */ shader_glsl_pow
,
4042 /* WINED3DSIH_RCP */ shader_glsl_rcp
,
4043 /* WINED3DSIH_REP */ shader_glsl_rep
,
4044 /* WINED3DSIH_RET */ NULL
,
4045 /* WINED3DSIH_RSQ */ shader_glsl_rsq
,
4046 /* WINED3DSIH_SETP */ NULL
,
4047 /* WINED3DSIH_SGE */ shader_glsl_compare
,
4048 /* WINED3DSIH_SGN */ shader_glsl_map2gl
,
4049 /* WINED3DSIH_SINCOS */ shader_glsl_sincos
,
4050 /* WINED3DSIH_SLT */ shader_glsl_compare
,
4051 /* WINED3DSIH_SUB */ shader_glsl_arith
,
4052 /* WINED3DSIH_TEX */ pshader_glsl_tex
,
4053 /* WINED3DSIH_TEXBEM */ pshader_glsl_texbem
,
4054 /* WINED3DSIH_TEXBEML */ pshader_glsl_texbem
,
4055 /* WINED3DSIH_TEXCOORD */ pshader_glsl_texcoord
,
4056 /* WINED3DSIH_TEXDEPTH */ pshader_glsl_texdepth
,
4057 /* WINED3DSIH_TEXDP3 */ pshader_glsl_texdp3
,
4058 /* WINED3DSIH_TEXDP3TEX */ pshader_glsl_texdp3tex
,
4059 /* WINED3DSIH_TEXKILL */ pshader_glsl_texkill
,
4060 /* WINED3DSIH_TEXLDD */ NULL
,
4061 /* WINED3DSIH_TEXLDL */ shader_glsl_texldl
,
4062 /* WINED3DSIH_TEXM3x2DEPTH */ pshader_glsl_texm3x2depth
,
4063 /* WINED3DSIH_TEXM3x2PAD */ pshader_glsl_texm3x2pad
,
4064 /* WINED3DSIH_TEXM3x2TEX */ pshader_glsl_texm3x2tex
,
4065 /* WINED3DSIH_TEXM3x3 */ pshader_glsl_texm3x3
,
4066 /* WINED3DSIH_TEXM3x3DIFF */ NULL
,
4067 /* WINED3DSIH_TEXM3x3PAD */ pshader_glsl_texm3x3pad
,
4068 /* WINED3DSIH_TEXM3x3SPEC */ pshader_glsl_texm3x3spec
,
4069 /* WINED3DSIH_TEXM3x3TEX */ pshader_glsl_texm3x3tex
,
4070 /* WINED3DSIH_TEXM3x3VSPEC */ pshader_glsl_texm3x3vspec
,
4071 /* WINED3DSIH_TEXREG2AR */ pshader_glsl_texreg2ar
,
4072 /* WINED3DSIH_TEXREG2GB */ pshader_glsl_texreg2gb
,
4073 /* WINED3DSIH_TEXREG2RGB */ pshader_glsl_texreg2rgb
,
4076 const shader_backend_t glsl_shader_backend
= {
4077 shader_glsl_instruction_handler_table
,
4079 shader_glsl_select_depth_blt
,
4080 shader_glsl_deselect_depth_blt
,
4081 shader_glsl_update_float_vertex_constants
,
4082 shader_glsl_update_float_pixel_constants
,
4083 shader_glsl_load_constants
,
4084 shader_glsl_color_correction
,
4085 shader_glsl_destroy
,
4088 shader_glsl_dirty_const
,
4089 shader_glsl_generate_pshader
,
4090 shader_glsl_generate_vshader
,
4091 shader_glsl_get_caps
,
4092 shader_glsl_color_fixup_supported
,