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.
33 #include "wined3d_private.h"
35 WINE_DEFAULT_DEBUG_CHANNEL(d3d_shader
);
36 WINE_DECLARE_DEBUG_CHANNEL(d3d_constants
);
37 WINE_DECLARE_DEBUG_CHANNEL(d3d_caps
);
38 WINE_DECLARE_DEBUG_CHANNEL(d3d
);
40 #define GLINFO_LOCATION (*gl_info)
55 } glsl_sample_function_t
;
57 /* GLSL shader private data */
58 struct shader_glsl_priv
{
59 struct hash_table_t
*glsl_program_lookup
;
60 const struct glsl_shader_prog_link
*glsl_program
;
61 GLhandleARB depth_blt_program
[tex_type_count
];
64 /* Struct to maintain data about a linked GLSL program */
65 struct glsl_shader_prog_link
{
66 struct list vshader_entry
;
67 struct list pshader_entry
;
68 GLhandleARB programId
;
69 GLhandleARB
*vuniformF_locations
;
70 GLhandleARB
*puniformF_locations
;
71 GLhandleARB vuniformI_locations
[MAX_CONST_I
];
72 GLhandleARB puniformI_locations
[MAX_CONST_I
];
73 GLhandleARB posFixup_location
;
74 GLhandleARB bumpenvmat_location
[MAX_TEXTURES
];
75 GLhandleARB luminancescale_location
[MAX_TEXTURES
];
76 GLhandleARB luminanceoffset_location
[MAX_TEXTURES
];
77 GLhandleARB ycorrection_location
;
78 GLenum vertex_color_clamp
;
80 IWineD3DPixelShader
*pshader
;
81 struct ps_compile_args ps_args
;
86 IWineD3DPixelShader
*pshader
;
87 struct ps_compile_args ps_args
;
91 /** Prints the GLSL info log which will contain error messages if they exist */
92 static void print_glsl_info_log(const WineD3D_GL_Info
*gl_info
, GLhandleARB obj
)
94 int infologLength
= 0;
99 static const char * const spam
[] =
101 "Vertex shader was successfully compiled to run on hardware.\n", /* fglrx */
102 "Fragment shader was successfully compiled to run on hardware.\n", /* fglrx */
103 "Fragment shader(s) linked, vertex shader(s) linked. \n ", /* fglrx, with \n */
104 "Fragment shader(s) linked, vertex shader(s) linked.", /* fglrx, no \n */
105 "Vertex shader(s) linked, no fragment shader(s) defined. \n ", /* fglrx, with \n */
106 "Vertex shader(s) linked, no fragment shader(s) defined.", /* fglrx, no \n */
107 "Fragment shader was successfully compiled to run on hardware.\nWARNING: 0:1: extension 'GL_ARB_draw_buffers' is not supported",
108 "Fragment shader(s) linked, no vertex shader(s) defined.", /* fglrx, no \n */
109 "Fragment shader(s) linked, no vertex shader(s) defined. \n ", /* fglrx, with \n */
110 "WARNING: 0:2: extension 'GL_ARB_draw_buffers' is not supported\n" /* MacOS ati */
113 GL_EXTCALL(glGetObjectParameterivARB(obj
,
114 GL_OBJECT_INFO_LOG_LENGTH_ARB
,
117 /* A size of 1 is just a null-terminated string, so the log should be bigger than
118 * that if there are errors. */
119 if (infologLength
> 1)
121 /* Fglrx doesn't terminate the string properly, but it tells us the proper length.
122 * So use HEAP_ZERO_MEMORY to avoid uninitialized bytes
124 infoLog
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, infologLength
);
125 GL_EXTCALL(glGetInfoLogARB(obj
, infologLength
, NULL
, infoLog
));
128 for(i
= 0; i
< sizeof(spam
) / sizeof(spam
[0]); i
++) {
129 if(strcmp(infoLog
, spam
[i
]) == 0) {
135 TRACE("Spam received from GLSL shader #%u: %s\n", obj
, debugstr_a(infoLog
));
137 FIXME("Error received from GLSL shader #%u: %s\n", obj
, debugstr_a(infoLog
));
139 HeapFree(GetProcessHeap(), 0, infoLog
);
144 * Loads (pixel shader) samplers
146 static void shader_glsl_load_psamplers(const WineD3D_GL_Info
*gl_info
, IWineD3DStateBlock
*iface
, GLhandleARB programId
)
148 IWineD3DStateBlockImpl
* stateBlock
= (IWineD3DStateBlockImpl
*) iface
;
149 GLhandleARB name_loc
;
151 char sampler_name
[20];
153 for (i
= 0; i
< MAX_FRAGMENT_SAMPLERS
; ++i
) {
154 snprintf(sampler_name
, sizeof(sampler_name
), "Psampler%d", i
);
155 name_loc
= GL_EXTCALL(glGetUniformLocationARB(programId
, sampler_name
));
156 if (name_loc
!= -1) {
157 int mapped_unit
= stateBlock
->wineD3DDevice
->texUnitMap
[i
];
158 if (mapped_unit
!= -1 && mapped_unit
< GL_LIMITS(fragment_samplers
)) {
159 TRACE("Loading %s for texture %d\n", sampler_name
, mapped_unit
);
160 GL_EXTCALL(glUniform1iARB(name_loc
, mapped_unit
));
161 checkGLcall("glUniform1iARB");
163 ERR("Trying to load sampler %s on unsupported unit %d\n", sampler_name
, mapped_unit
);
169 static void shader_glsl_load_vsamplers(const WineD3D_GL_Info
*gl_info
, IWineD3DStateBlock
*iface
, GLhandleARB programId
)
171 IWineD3DStateBlockImpl
* stateBlock
= (IWineD3DStateBlockImpl
*) iface
;
172 GLhandleARB name_loc
;
173 char sampler_name
[20];
176 for (i
= 0; i
< MAX_VERTEX_SAMPLERS
; ++i
) {
177 snprintf(sampler_name
, sizeof(sampler_name
), "Vsampler%d", i
);
178 name_loc
= GL_EXTCALL(glGetUniformLocationARB(programId
, sampler_name
));
179 if (name_loc
!= -1) {
180 int mapped_unit
= stateBlock
->wineD3DDevice
->texUnitMap
[MAX_FRAGMENT_SAMPLERS
+ i
];
181 if (mapped_unit
!= -1 && mapped_unit
< GL_LIMITS(combined_samplers
)) {
182 TRACE("Loading %s for texture %d\n", sampler_name
, mapped_unit
);
183 GL_EXTCALL(glUniform1iARB(name_loc
, mapped_unit
));
184 checkGLcall("glUniform1iARB");
186 ERR("Trying to load sampler %s on unsupported unit %d\n", sampler_name
, mapped_unit
);
193 * Loads floating point constants (aka uniforms) into the currently set GLSL program.
194 * When constant_list == NULL, it will load all the constants.
196 static void shader_glsl_load_constantsF(IWineD3DBaseShaderImpl
*This
, const WineD3D_GL_Info
*gl_info
,
197 unsigned int max_constants
, const float *constants
, const GLhandleARB
*constant_locations
,
198 const struct list
*constant_list
)
200 const constants_entry
*constant
;
201 const local_constant
*lconst
;
206 if (TRACE_ON(d3d_shader
)) {
207 LIST_FOR_EACH_ENTRY(constant
, constant_list
, constants_entry
, entry
) {
212 tmp_loc
= constant_locations
[i
];
214 TRACE_(d3d_constants
)("Loading constants %i: %f, %f, %f, %f\n", i
,
215 constants
[i
* 4 + 0], constants
[i
* 4 + 1],
216 constants
[i
* 4 + 2], constants
[i
* 4 + 3]);
222 /* 1.X pshaders have the constants clamped to [-1;1] implicitly. */
223 if (WINED3DSHADER_VERSION_MAJOR(This
->baseShader
.reg_maps
.shader_version
) == 1
224 && shader_is_pshader_version(This
->baseShader
.reg_maps
.shader_version
))
228 LIST_FOR_EACH_ENTRY(constant
, constant_list
, constants_entry
, entry
) {
233 tmp_loc
= constant_locations
[i
];
235 /* We found this uniform name in the program - go ahead and send the data */
237 if(constants
[k
+ 0] < -1.0) lcl_const
[0] = -1.0;
238 else if(constants
[k
+ 0] > 1.0) lcl_const
[0] = 1.0;
239 else lcl_const
[0] = constants
[k
+ 0];
240 if(constants
[k
+ 1] < -1.0) lcl_const
[1] = -1.0;
241 else if(constants
[k
+ 1] > 1.0) lcl_const
[1] = 1.0;
242 else lcl_const
[1] = constants
[k
+ 1];
243 if(constants
[k
+ 2] < -1.0) lcl_const
[2] = -1.0;
244 else if(constants
[k
+ 2] > 1.0) lcl_const
[2] = 1.0;
245 else lcl_const
[2] = constants
[k
+ 2];
246 if(constants
[k
+ 3] < -1.0) lcl_const
[3] = -1.0;
247 else if(constants
[k
+ 3] > 1.0) lcl_const
[3] = 1.0;
248 else lcl_const
[3] = constants
[k
+ 3];
250 GL_EXTCALL(glUniform4fvARB(tmp_loc
, 1, lcl_const
));
255 LIST_FOR_EACH_ENTRY(constant
, constant_list
, constants_entry
, entry
) {
260 tmp_loc
= constant_locations
[i
];
262 /* We found this uniform name in the program - go ahead and send the data */
263 GL_EXTCALL(glUniform4fvARB(tmp_loc
, 1, constants
+ (i
* 4)));
268 checkGLcall("glUniform4fvARB()");
270 if(!This
->baseShader
.load_local_constsF
) {
271 TRACE("No need to load local float constants for this shader\n");
275 /* Load immediate constants */
276 if (TRACE_ON(d3d_shader
)) {
277 LIST_FOR_EACH_ENTRY(lconst
, &This
->baseShader
.constantsF
, local_constant
, entry
) {
278 tmp_loc
= constant_locations
[lconst
->idx
];
280 const GLfloat
*values
= (const GLfloat
*)lconst
->value
;
281 TRACE_(d3d_constants
)("Loading local constants %i: %f, %f, %f, %f\n", lconst
->idx
,
282 values
[0], values
[1], values
[2], values
[3]);
286 /* Immediate constants are clamped to [-1;1] at shader creation time if needed */
287 LIST_FOR_EACH_ENTRY(lconst
, &This
->baseShader
.constantsF
, local_constant
, entry
) {
288 tmp_loc
= constant_locations
[lconst
->idx
];
290 /* We found this uniform name in the program - go ahead and send the data */
291 GL_EXTCALL(glUniform4fvARB(tmp_loc
, 1, (const GLfloat
*)lconst
->value
));
294 checkGLcall("glUniform4fvARB()");
297 /* Loads integer constants (aka uniforms) into the currently set GLSL program. */
298 static void shader_glsl_load_constantsI(IWineD3DBaseShaderImpl
*This
, const WineD3D_GL_Info
*gl_info
,
299 const GLhandleARB locations
[MAX_CONST_I
], const int *constants
, WORD constants_set
)
304 for (i
= 0; constants_set
; constants_set
>>= 1, ++i
)
306 if (!(constants_set
& 1)) continue;
308 TRACE_(d3d_constants
)("Loading constants %u: %i, %i, %i, %i\n",
309 i
, constants
[i
*4], constants
[i
*4+1], constants
[i
*4+2], constants
[i
*4+3]);
311 /* We found this uniform name in the program - go ahead and send the data */
312 GL_EXTCALL(glUniform4ivARB(locations
[i
], 1, &constants
[i
*4]));
313 checkGLcall("glUniform4ivARB");
316 /* Load immediate constants */
317 ptr
= list_head(&This
->baseShader
.constantsI
);
319 const struct local_constant
*lconst
= LIST_ENTRY(ptr
, const struct local_constant
, entry
);
320 unsigned int idx
= lconst
->idx
;
321 const GLint
*values
= (const GLint
*)lconst
->value
;
323 TRACE_(d3d_constants
)("Loading local constants %i: %i, %i, %i, %i\n", idx
,
324 values
[0], values
[1], values
[2], values
[3]);
326 /* We found this uniform name in the program - go ahead and send the data */
327 GL_EXTCALL(glUniform4ivARB(locations
[idx
], 1, values
));
328 checkGLcall("glUniform4ivARB");
329 ptr
= list_next(&This
->baseShader
.constantsI
, ptr
);
333 /* Loads boolean constants (aka uniforms) into the currently set GLSL program. */
334 static void shader_glsl_load_constantsB(IWineD3DBaseShaderImpl
*This
, const WineD3D_GL_Info
*gl_info
,
335 GLhandleARB programId
, const BOOL
*constants
, WORD constants_set
)
340 char is_pshader
= shader_is_pshader_version(This
->baseShader
.reg_maps
.shader_version
);
341 const char* prefix
= is_pshader
? "PB":"VB";
344 /* TODO: Benchmark and see if it would be beneficial to store the
345 * locations of the constants to avoid looking up each time */
346 for (i
= 0; constants_set
; constants_set
>>= 1, ++i
)
348 if (!(constants_set
& 1)) continue;
350 TRACE_(d3d_constants
)("Loading constants %i: %i;\n", i
, constants
[i
]);
352 /* TODO: Benchmark and see if it would be beneficial to store the
353 * locations of the constants to avoid looking up each time */
354 snprintf(tmp_name
, sizeof(tmp_name
), "%s[%i]", prefix
, i
);
355 tmp_loc
= GL_EXTCALL(glGetUniformLocationARB(programId
, tmp_name
));
358 /* We found this uniform name in the program - go ahead and send the data */
359 GL_EXTCALL(glUniform1ivARB(tmp_loc
, 1, &constants
[i
]));
360 checkGLcall("glUniform1ivARB");
364 /* Load immediate constants */
365 ptr
= list_head(&This
->baseShader
.constantsB
);
367 const struct local_constant
*lconst
= LIST_ENTRY(ptr
, const struct local_constant
, entry
);
368 unsigned int idx
= lconst
->idx
;
369 const GLint
*values
= (const GLint
*)lconst
->value
;
371 TRACE_(d3d_constants
)("Loading local constants %i: %i\n", idx
, values
[0]);
373 snprintf(tmp_name
, sizeof(tmp_name
), "%s[%i]", prefix
, idx
);
374 tmp_loc
= GL_EXTCALL(glGetUniformLocationARB(programId
, tmp_name
));
376 /* We found this uniform name in the program - go ahead and send the data */
377 GL_EXTCALL(glUniform1ivARB(tmp_loc
, 1, values
));
378 checkGLcall("glUniform1ivARB");
380 ptr
= list_next(&This
->baseShader
.constantsB
, ptr
);
387 * Loads the app-supplied constants into the currently set GLSL program.
389 static void shader_glsl_load_constants(
390 IWineD3DDevice
* device
,
392 char useVertexShader
) {
394 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) device
;
395 const struct shader_glsl_priv
*priv
= (struct shader_glsl_priv
*)deviceImpl
->shader_priv
;
396 IWineD3DStateBlockImpl
* stateBlock
= deviceImpl
->stateBlock
;
397 const WineD3D_GL_Info
*gl_info
= &deviceImpl
->adapter
->gl_info
;
399 const GLhandleARB
*constant_locations
;
400 const struct list
*constant_list
;
401 GLhandleARB programId
;
402 const struct glsl_shader_prog_link
*prog
= priv
->glsl_program
;
406 /* No GLSL program set - nothing to do. */
409 programId
= prog
->programId
;
411 if (useVertexShader
) {
412 IWineD3DBaseShaderImpl
* vshader
= (IWineD3DBaseShaderImpl
*) stateBlock
->vertexShader
;
414 constant_locations
= prog
->vuniformF_locations
;
415 constant_list
= &stateBlock
->set_vconstantsF
;
417 /* Load DirectX 9 float constants/uniforms for vertex shader */
418 shader_glsl_load_constantsF(vshader
, gl_info
, GL_LIMITS(vshader_constantsF
),
419 stateBlock
->vertexShaderConstantF
, constant_locations
, constant_list
);
421 /* Load DirectX 9 integer constants/uniforms for vertex shader */
422 if(vshader
->baseShader
.uses_int_consts
) {
423 shader_glsl_load_constantsI(vshader
, gl_info
, prog
->vuniformI_locations
,
424 stateBlock
->vertexShaderConstantI
, stateBlock
->changed
.vertexShaderConstantsI
);
427 /* Load DirectX 9 boolean constants/uniforms for vertex shader */
428 if(vshader
->baseShader
.uses_bool_consts
) {
429 shader_glsl_load_constantsB(vshader
, gl_info
, programId
,
430 stateBlock
->vertexShaderConstantB
, stateBlock
->changed
.vertexShaderConstantsB
);
433 /* Upload the position fixup params */
434 GL_EXTCALL(glUniform4fvARB(prog
->posFixup_location
, 1, &deviceImpl
->posFixup
[0]));
435 checkGLcall("glUniform4fvARB");
438 if (usePixelShader
) {
440 IWineD3DBaseShaderImpl
* pshader
= (IWineD3DBaseShaderImpl
*) stateBlock
->pixelShader
;
442 constant_locations
= prog
->puniformF_locations
;
443 constant_list
= &stateBlock
->set_pconstantsF
;
445 /* Load DirectX 9 float constants/uniforms for pixel shader */
446 shader_glsl_load_constantsF(pshader
, gl_info
, GL_LIMITS(pshader_constantsF
),
447 stateBlock
->pixelShaderConstantF
, constant_locations
, constant_list
);
449 /* Load DirectX 9 integer constants/uniforms for pixel shader */
450 if(pshader
->baseShader
.uses_int_consts
) {
451 shader_glsl_load_constantsI(pshader
, gl_info
, prog
->puniformI_locations
,
452 stateBlock
->pixelShaderConstantI
, stateBlock
->changed
.pixelShaderConstantsI
);
455 /* Load DirectX 9 boolean constants/uniforms for pixel shader */
456 if(pshader
->baseShader
.uses_bool_consts
) {
457 shader_glsl_load_constantsB(pshader
, gl_info
, programId
,
458 stateBlock
->pixelShaderConstantB
, stateBlock
->changed
.pixelShaderConstantsB
);
461 /* Upload the environment bump map matrix if needed. The needsbumpmat member specifies the texture stage to load the matrix from.
462 * It can't be 0 for a valid texbem instruction.
464 for(i
= 0; i
< ((IWineD3DPixelShaderImpl
*) pshader
)->numbumpenvmatconsts
; i
++) {
465 IWineD3DPixelShaderImpl
*ps
= (IWineD3DPixelShaderImpl
*) pshader
;
466 int stage
= ps
->luminanceconst
[i
].texunit
;
468 const float *data
= (const float *)&stateBlock
->textureState
[(int)ps
->bumpenvmatconst
[i
].texunit
][WINED3DTSS_BUMPENVMAT00
];
469 GL_EXTCALL(glUniformMatrix2fvARB(prog
->bumpenvmat_location
[i
], 1, 0, data
));
470 checkGLcall("glUniformMatrix2fvARB");
472 /* texbeml needs the luminance scale and offset too. If texbeml is used, needsbumpmat
473 * is set too, so we can check that in the needsbumpmat check
475 if(ps
->baseShader
.reg_maps
.luminanceparams
[stage
]) {
476 const GLfloat
*scale
= (const GLfloat
*)&stateBlock
->textureState
[stage
][WINED3DTSS_BUMPENVLSCALE
];
477 const GLfloat
*offset
= (const GLfloat
*)&stateBlock
->textureState
[stage
][WINED3DTSS_BUMPENVLOFFSET
];
479 GL_EXTCALL(glUniform1fvARB(prog
->luminancescale_location
[i
], 1, scale
));
480 checkGLcall("glUniform1fvARB");
481 GL_EXTCALL(glUniform1fvARB(prog
->luminanceoffset_location
[i
], 1, offset
));
482 checkGLcall("glUniform1fvARB");
486 if(((IWineD3DPixelShaderImpl
*) pshader
)->vpos_uniform
) {
487 float correction_params
[4];
488 if(deviceImpl
->render_offscreen
) {
489 correction_params
[0] = 0.0;
490 correction_params
[1] = 1.0;
492 /* position is window relative, not viewport relative */
493 correction_params
[0] = ((IWineD3DSurfaceImpl
*) deviceImpl
->render_targets
[0])->currentDesc
.Height
;
494 correction_params
[1] = -1.0;
496 GL_EXTCALL(glUniform4fvARB(prog
->ycorrection_location
, 1, correction_params
));
501 static void shader_glsl_update_float_vertex_constants(IWineD3DDevice
*iface
, UINT start
, UINT count
)
503 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
506 for (i
= start
; i
< count
+ start
; ++i
)
508 if (!This
->stateBlock
->changed
.vertexShaderConstantsF
[i
])
510 constants_entry
*ptr
= LIST_ENTRY(list_head(&This
->stateBlock
->set_vconstantsF
),
511 constants_entry
, entry
);
513 if (!ptr
|| ptr
->count
>= sizeof(ptr
->idx
) / sizeof(*ptr
->idx
))
515 ptr
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(constants_entry
));
516 list_add_head(&This
->stateBlock
->set_vconstantsF
, &ptr
->entry
);
518 ptr
->idx
[ptr
->count
++] = i
;
523 static void shader_glsl_update_float_pixel_constants(IWineD3DDevice
*iface
, UINT start
, UINT count
)
525 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
528 for (i
= start
; i
< count
+ start
; ++i
)
530 if (!This
->stateBlock
->changed
.pixelShaderConstantsF
[i
])
532 constants_entry
*ptr
= LIST_ENTRY(list_head(&This
->stateBlock
->set_pconstantsF
),
533 constants_entry
, entry
);
535 if (!ptr
|| ptr
->count
>= sizeof(ptr
->idx
) / sizeof(*ptr
->idx
))
537 ptr
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(constants_entry
));
538 list_add_head(&This
->stateBlock
->set_pconstantsF
, &ptr
->entry
);
540 ptr
->idx
[ptr
->count
++] = i
;
545 /** Generate the variable & register declarations for the GLSL output target */
546 static void shader_generate_glsl_declarations(IWineD3DBaseShader
*iface
, const shader_reg_maps
*reg_maps
,
547 SHADER_BUFFER
*buffer
, const WineD3D_GL_Info
*gl_info
,
548 const struct ps_compile_args
*ps_args
)
550 IWineD3DBaseShaderImpl
* This
= (IWineD3DBaseShaderImpl
*) iface
;
551 IWineD3DDeviceImpl
*device
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
552 DWORD shader_version
= reg_maps
->shader_version
;
553 unsigned int i
, extra_constants_needed
= 0;
554 const local_constant
*lconst
;
556 /* There are some minor differences between pixel and vertex shaders */
557 char pshader
= shader_is_pshader_version(shader_version
);
558 char prefix
= pshader
? 'P' : 'V';
560 /* Prototype the subroutines */
561 for (i
= 0; i
< This
->baseShader
.limits
.label
; i
++) {
562 if (reg_maps
->labels
[i
])
563 shader_addline(buffer
, "void subroutine%u();\n", i
);
566 /* Declare the constants (aka uniforms) */
567 if (This
->baseShader
.limits
.constant_float
> 0) {
568 unsigned max_constantsF
= min(This
->baseShader
.limits
.constant_float
,
569 (pshader
? GL_LIMITS(pshader_constantsF
) : GL_LIMITS(vshader_constantsF
)));
570 shader_addline(buffer
, "uniform vec4 %cC[%u];\n", prefix
, max_constantsF
);
573 if (This
->baseShader
.limits
.constant_int
> 0)
574 shader_addline(buffer
, "uniform ivec4 %cI[%u];\n", prefix
, This
->baseShader
.limits
.constant_int
);
576 if (This
->baseShader
.limits
.constant_bool
> 0)
577 shader_addline(buffer
, "uniform bool %cB[%u];\n", prefix
, This
->baseShader
.limits
.constant_bool
);
580 shader_addline(buffer
, "uniform vec4 posFixup;\n");
581 /* Predeclaration; This function is added at link time based on the pixel shader.
582 * VS 3.0 shaders have an array OUT[] the shader writes to, earlier versions don't have
583 * that. We know the input to the reorder function at vertex shader compile time, so
584 * we can deal with that. The reorder function for a 1.x and 2.x vertex shader can just
585 * read gl_FrontColor. The output depends on the pixel shader. The reorder function for a
586 * 1.x and 2.x pshader or for fixed function will write gl_FrontColor, and for a 3.0 shader
587 * it will write to the varying array. Here we depend on the shader optimizer on sorting that
588 * out. The nvidia driver only does that if the parameter is inout instead of out, hence the
591 if (shader_version
>= WINED3DVS_VERSION(3, 0))
593 shader_addline(buffer
, "void order_ps_input(in vec4[%u]);\n", MAX_REG_OUTPUT
);
595 shader_addline(buffer
, "void order_ps_input();\n");
598 IWineD3DPixelShaderImpl
*ps_impl
= (IWineD3DPixelShaderImpl
*) This
;
600 ps_impl
->numbumpenvmatconsts
= 0;
601 for(i
= 0; i
< (sizeof(reg_maps
->bumpmat
) / sizeof(reg_maps
->bumpmat
[0])); i
++) {
602 if(!reg_maps
->bumpmat
[i
]) {
606 ps_impl
->bumpenvmatconst
[(int) ps_impl
->numbumpenvmatconsts
].texunit
= i
;
607 shader_addline(buffer
, "uniform mat2 bumpenvmat%d;\n", i
);
609 if(reg_maps
->luminanceparams
) {
610 ps_impl
->luminanceconst
[(int) ps_impl
->numbumpenvmatconsts
].texunit
= i
;
611 shader_addline(buffer
, "uniform float luminancescale%d;\n", i
);
612 shader_addline(buffer
, "uniform float luminanceoffset%d;\n", i
);
613 extra_constants_needed
++;
615 ps_impl
->luminanceconst
[(int) ps_impl
->numbumpenvmatconsts
].texunit
= -1;
618 extra_constants_needed
++;
619 ps_impl
->numbumpenvmatconsts
++;
622 if(ps_args
->srgb_correction
) {
623 shader_addline(buffer
, "const vec4 srgb_mul_low = vec4(%f, %f, %f, %f);\n",
624 srgb_mul_low
, srgb_mul_low
, srgb_mul_low
, srgb_mul_low
);
625 shader_addline(buffer
, "const vec4 srgb_comparison = vec4(%f, %f, %f, %f);\n",
626 srgb_cmp
, srgb_cmp
, srgb_cmp
, srgb_cmp
);
628 if(reg_maps
->vpos
|| reg_maps
->usesdsy
) {
629 if(This
->baseShader
.limits
.constant_float
+ extra_constants_needed
+ 1 < GL_LIMITS(pshader_constantsF
)) {
630 shader_addline(buffer
, "uniform vec4 ycorrection;\n");
631 ((IWineD3DPixelShaderImpl
*) This
)->vpos_uniform
= 1;
632 extra_constants_needed
++;
634 /* This happens because we do not have proper tracking of the constant registers that are
635 * actually used, only the max limit of the shader version
637 FIXME("Cannot find a free uniform for vpos correction params\n");
638 shader_addline(buffer
, "const vec4 ycorrection = vec4(%f, %f, 0.0, 0.0);\n",
639 device
->render_offscreen
? 0.0 : ((IWineD3DSurfaceImpl
*) device
->render_targets
[0])->currentDesc
.Height
,
640 device
->render_offscreen
? 1.0 : -1.0);
642 shader_addline(buffer
, "vec4 vpos;\n");
646 /* Declare texture samplers */
647 for (i
= 0; i
< This
->baseShader
.limits
.sampler
; i
++) {
648 if (reg_maps
->samplers
[i
]) {
650 DWORD stype
= reg_maps
->samplers
[i
] & WINED3DSP_TEXTURETYPE_MASK
;
654 shader_addline(buffer
, "uniform sampler1D %csampler%u;\n", prefix
, i
);
657 if(device
->stateBlock
->textures
[i
] &&
658 IWineD3DBaseTexture_GetTextureDimensions(device
->stateBlock
->textures
[i
]) == GL_TEXTURE_RECTANGLE_ARB
) {
659 shader_addline(buffer
, "uniform sampler2DRect %csampler%u;\n", prefix
, i
);
661 shader_addline(buffer
, "uniform sampler2D %csampler%u;\n", prefix
, i
);
664 case WINED3DSTT_CUBE
:
665 shader_addline(buffer
, "uniform samplerCube %csampler%u;\n", prefix
, i
);
667 case WINED3DSTT_VOLUME
:
668 shader_addline(buffer
, "uniform sampler3D %csampler%u;\n", prefix
, i
);
671 shader_addline(buffer
, "uniform unsupported_sampler %csampler%u;\n", prefix
, i
);
672 FIXME("Unrecognized sampler type: %#x\n", stype
);
678 /* Declare address variables */
679 for (i
= 0; i
< This
->baseShader
.limits
.address
; i
++) {
680 if (reg_maps
->address
[i
])
681 shader_addline(buffer
, "ivec4 A%d;\n", i
);
684 /* Declare texture coordinate temporaries and initialize them */
685 for (i
= 0; i
< This
->baseShader
.limits
.texcoord
; i
++) {
686 if (reg_maps
->texcoord
[i
])
687 shader_addline(buffer
, "vec4 T%u = gl_TexCoord[%u];\n", i
, i
);
690 /* Declare input register varyings. Only pixel shader, vertex shaders have that declared in the
691 * helper function shader that is linked in at link time
693 if (pshader
&& shader_version
>= WINED3DPS_VERSION(3, 0))
696 shader_addline(buffer
, "varying vec4 IN[%u];\n", GL_LIMITS(glsl_varyings
) / 4);
698 /* TODO: Write a replacement shader for the fixed function vertex pipeline, so this isn't needed.
699 * For fixed function vertex processing + 3.0 pixel shader we need a separate function in the
700 * pixel shader that reads the fixed function color into the packed input registers.
702 shader_addline(buffer
, "vec4 IN[%u];\n", GL_LIMITS(glsl_varyings
) / 4);
706 /* Declare output register temporaries */
707 if(This
->baseShader
.limits
.packed_output
) {
708 shader_addline(buffer
, "vec4 OUT[%u];\n", This
->baseShader
.limits
.packed_output
);
711 /* Declare temporary variables */
712 for(i
= 0; i
< This
->baseShader
.limits
.temporary
; i
++) {
713 if (reg_maps
->temporary
[i
])
714 shader_addline(buffer
, "vec4 R%u;\n", i
);
717 /* Declare attributes */
718 for (i
= 0; i
< This
->baseShader
.limits
.attributes
; i
++) {
719 if (reg_maps
->attributes
[i
])
720 shader_addline(buffer
, "attribute vec4 attrib%i;\n", i
);
723 /* Declare loop registers aLx */
724 for (i
= 0; i
< reg_maps
->loop_depth
; i
++) {
725 shader_addline(buffer
, "int aL%u;\n", i
);
726 shader_addline(buffer
, "int tmpInt%u;\n", i
);
729 /* Temporary variables for matrix operations */
730 shader_addline(buffer
, "vec4 tmp0;\n");
731 shader_addline(buffer
, "vec4 tmp1;\n");
733 /* Local constants use a different name so they can be loaded once at shader link time
734 * They can't be hardcoded into the shader text via LC = {x, y, z, w}; because the
735 * float -> string conversion can cause precision loss.
737 if(!This
->baseShader
.load_local_constsF
) {
738 LIST_FOR_EACH_ENTRY(lconst
, &This
->baseShader
.constantsF
, local_constant
, entry
) {
739 shader_addline(buffer
, "uniform vec4 %cLC%u;\n", prefix
, lconst
->idx
);
743 /* Start the main program */
744 shader_addline(buffer
, "void main() {\n");
745 if(pshader
&& reg_maps
->vpos
) {
746 /* DirectX apps expect integer values, while OpenGL drivers add approximately 0.5. This causes
747 * off-by-one problems as spotted by the vPos d3d9 visual test. Unfortunately the ATI cards do
748 * not add exactly 0.5, but rather something like 0.49999999 or 0.50000001, which still causes
749 * precision troubles when we just substract 0.5.
751 * To deal with that just floor() the position. This will eliminate the fraction on all cards.
753 * TODO: Test how that behaves with multisampling once we can enable multisampling in winex11.
755 * An advantage of floor is that it works even if the driver doesn't add 1/2. It is somewhat
756 * questionable if 1.5, 2.5, ... are the proper values to return in gl_FragCoord, even though
757 * coordinates specify the pixel centers instead of the pixel corners. This code will behave
758 * correctly on drivers that returns integer values.
760 shader_addline(buffer
, "vpos = floor(vec4(0, ycorrection[0], 0, 0) + gl_FragCoord * vec4(1, ycorrection[1], 1, 1));\n");
764 /*****************************************************************************
765 * Functions to generate GLSL strings from DirectX Shader bytecode begin here.
767 * For more information, see http://wiki.winehq.org/DirectX-Shaders
768 ****************************************************************************/
771 static void shader_glsl_add_src_param(const SHADER_OPCODE_ARG
*arg
, const DWORD param
,
772 const DWORD addr_token
, DWORD mask
, glsl_src_param_t
*src_param
);
774 /** Used for opcode modifiers - They multiply the result by the specified amount */
775 static const char * const shift_glsl_tab
[] = {
777 "2.0 * ", /* 1 (x2) */
778 "4.0 * ", /* 2 (x4) */
779 "8.0 * ", /* 3 (x8) */
780 "16.0 * ", /* 4 (x16) */
781 "32.0 * ", /* 5 (x32) */
788 "0.0625 * ", /* 12 (d16) */
789 "0.125 * ", /* 13 (d8) */
790 "0.25 * ", /* 14 (d4) */
791 "0.5 * " /* 15 (d2) */
794 /* Generate a GLSL parameter that does the input modifier computation and return the input register/mask to use */
795 static void shader_glsl_gen_modifier (
798 const char *in_regswizzle
,
803 if (instr
== WINED3DSIO_TEXKILL
)
806 switch (instr
& WINED3DSP_SRCMOD_MASK
) {
807 case WINED3DSPSM_DZ
: /* Need to handle this in the instructions itself (texld & texcrd). */
809 case WINED3DSPSM_NONE
:
810 sprintf(out_str
, "%s%s", in_reg
, in_regswizzle
);
812 case WINED3DSPSM_NEG
:
813 sprintf(out_str
, "-%s%s", in_reg
, in_regswizzle
);
815 case WINED3DSPSM_NOT
:
816 sprintf(out_str
, "!%s%s", in_reg
, in_regswizzle
);
818 case WINED3DSPSM_BIAS
:
819 sprintf(out_str
, "(%s%s - vec4(0.5)%s)", in_reg
, in_regswizzle
, in_regswizzle
);
821 case WINED3DSPSM_BIASNEG
:
822 sprintf(out_str
, "-(%s%s - vec4(0.5)%s)", in_reg
, in_regswizzle
, in_regswizzle
);
824 case WINED3DSPSM_SIGN
:
825 sprintf(out_str
, "(2.0 * (%s%s - 0.5))", in_reg
, in_regswizzle
);
827 case WINED3DSPSM_SIGNNEG
:
828 sprintf(out_str
, "-(2.0 * (%s%s - 0.5))", in_reg
, in_regswizzle
);
830 case WINED3DSPSM_COMP
:
831 sprintf(out_str
, "(1.0 - %s%s)", in_reg
, in_regswizzle
);
834 sprintf(out_str
, "(2.0 * %s%s)", in_reg
, in_regswizzle
);
836 case WINED3DSPSM_X2NEG
:
837 sprintf(out_str
, "-(2.0 * %s%s)", in_reg
, in_regswizzle
);
839 case WINED3DSPSM_ABS
:
840 sprintf(out_str
, "abs(%s%s)", in_reg
, in_regswizzle
);
842 case WINED3DSPSM_ABSNEG
:
843 sprintf(out_str
, "-abs(%s%s)", in_reg
, in_regswizzle
);
846 FIXME("Unhandled modifier %u\n", (instr
& WINED3DSP_SRCMOD_MASK
));
847 sprintf(out_str
, "%s%s", in_reg
, in_regswizzle
);
851 /** Writes the GLSL variable name that corresponds to the register that the
852 * DX opcode parameter is trying to access */
853 static void shader_glsl_get_register_name(const DWORD param
, const DWORD addr_token
,
854 char *regstr
, BOOL
*is_color
, const SHADER_OPCODE_ARG
*arg
)
856 /* oPos, oFog and oPts in D3D */
857 static const char * const hwrastout_reg_names
[] = { "gl_Position", "gl_FogFragCoord", "gl_PointSize" };
859 DWORD reg
= param
& WINED3DSP_REGNUM_MASK
;
860 DWORD regtype
= shader_get_regtype(param
);
861 IWineD3DBaseShaderImpl
* This
= (IWineD3DBaseShaderImpl
*) arg
->shader
;
862 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
863 const WineD3D_GL_Info
* gl_info
= &deviceImpl
->adapter
->gl_info
;
864 DWORD shader_version
= This
->baseShader
.reg_maps
.shader_version
;
865 char pshader
= shader_is_pshader_version(shader_version
);
871 case WINED3DSPR_TEMP
:
872 sprintf(tmpStr
, "R%u", reg
);
874 case WINED3DSPR_INPUT
:
876 /* Pixel shaders >= 3.0 */
877 if (WINED3DSHADER_VERSION_MAJOR(shader_version
) >= 3)
879 DWORD in_count
= GL_LIMITS(glsl_varyings
) / 4;
881 if (param
& WINED3DSHADER_ADDRMODE_RELATIVE
) {
882 glsl_src_param_t rel_param
;
883 shader_glsl_add_src_param(arg
, addr_token
, 0, WINED3DSP_WRITEMASK_0
, &rel_param
);
885 /* Removing a + 0 would be an obvious optimization, but macos doesn't see the NOP
888 if(((IWineD3DPixelShaderImpl
*) This
)->input_reg_map
[reg
]) {
889 if (((IWineD3DPixelShaderImpl
*)This
)->declared_in_count
> in_count
) {
890 sprintf(tmpStr
, "((%s + %u) > %d ? (%s + %u) > %d ? gl_SecondaryColor : gl_Color : IN[%s + %u])",
891 rel_param
.param_str
, ((IWineD3DPixelShaderImpl
*)This
)->input_reg_map
[reg
], in_count
- 1,
892 rel_param
.param_str
, ((IWineD3DPixelShaderImpl
*)This
)->input_reg_map
[reg
], in_count
,
893 rel_param
.param_str
, ((IWineD3DPixelShaderImpl
*)This
)->input_reg_map
[reg
]);
895 sprintf(tmpStr
, "IN[%s + %u]", rel_param
.param_str
, ((IWineD3DPixelShaderImpl
*)This
)->input_reg_map
[reg
]);
898 if (((IWineD3DPixelShaderImpl
*)This
)->declared_in_count
> in_count
) {
899 sprintf(tmpStr
, "((%s) > %d ? (%s) > %d ? gl_SecondaryColor : gl_Color : IN[%s])",
900 rel_param
.param_str
, in_count
- 1,
901 rel_param
.param_str
, in_count
,
902 rel_param
.param_str
);
904 sprintf(tmpStr
, "IN[%s]", rel_param
.param_str
);
908 DWORD idx
= ((IWineD3DPixelShaderImpl
*) This
)->input_reg_map
[reg
];
909 if (idx
== in_count
) {
910 sprintf(tmpStr
, "gl_Color");
911 } else if (idx
== in_count
+ 1) {
912 sprintf(tmpStr
, "gl_SecondaryColor");
914 sprintf(tmpStr
, "IN[%u]", idx
);
919 strcpy(tmpStr
, "gl_Color");
921 strcpy(tmpStr
, "gl_SecondaryColor");
924 if (vshader_input_is_color((IWineD3DVertexShader
*) This
, reg
))
926 sprintf(tmpStr
, "attrib%u", reg
);
929 case WINED3DSPR_CONST
:
931 const char prefix
= pshader
? 'P':'V';
933 /* Relative addressing */
934 if (param
& WINED3DSHADER_ADDRMODE_RELATIVE
) {
936 /* Relative addressing on shaders 2.0+ have a relative address token,
937 * prior to that, it was hard-coded as "A0.x" because there's only 1 register */
938 if (WINED3DSHADER_VERSION_MAJOR(shader_version
) >= 2)
940 glsl_src_param_t rel_param
;
941 shader_glsl_add_src_param(arg
, addr_token
, 0, WINED3DSP_WRITEMASK_0
, &rel_param
);
943 sprintf(tmpStr
, "%cC[%s + %u]", prefix
, rel_param
.param_str
, reg
);
945 sprintf(tmpStr
, "%cC[%s]", prefix
, rel_param
.param_str
);
949 sprintf(tmpStr
, "%cC[A0.x + %u]", prefix
, reg
);
951 sprintf(tmpStr
, "%cC[A0.x]", prefix
);
956 if(shader_constant_is_local(This
, reg
)) {
957 sprintf(tmpStr
, "%cLC%u", prefix
, reg
);
959 sprintf(tmpStr
, "%cC[%u]", prefix
, reg
);
965 case WINED3DSPR_CONSTINT
:
967 sprintf(tmpStr
, "PI[%u]", reg
);
969 sprintf(tmpStr
, "VI[%u]", reg
);
971 case WINED3DSPR_CONSTBOOL
:
973 sprintf(tmpStr
, "PB[%u]", reg
);
975 sprintf(tmpStr
, "VB[%u]", reg
);
977 case WINED3DSPR_TEXTURE
: /* case WINED3DSPR_ADDR: */
979 sprintf(tmpStr
, "T%u", reg
);
981 sprintf(tmpStr
, "A%u", reg
);
984 case WINED3DSPR_LOOP
:
985 sprintf(tmpStr
, "aL%u", This
->baseShader
.cur_loop_regno
- 1);
987 case WINED3DSPR_SAMPLER
:
989 sprintf(tmpStr
, "Psampler%u", reg
);
991 sprintf(tmpStr
, "Vsampler%u", reg
);
993 case WINED3DSPR_COLOROUT
:
994 if (reg
>= GL_LIMITS(buffers
)) {
995 WARN("Write to render target %u, only %d supported\n", reg
, 4);
997 if (GL_SUPPORT(ARB_DRAW_BUFFERS
)) {
998 sprintf(tmpStr
, "gl_FragData[%u]", reg
);
999 } else { /* On older cards with GLSL support like the GeforceFX there's only one buffer. */
1000 sprintf(tmpStr
, "gl_FragColor");
1003 case WINED3DSPR_RASTOUT
:
1004 sprintf(tmpStr
, "%s", hwrastout_reg_names
[reg
]);
1006 case WINED3DSPR_DEPTHOUT
:
1007 sprintf(tmpStr
, "gl_FragDepth");
1009 case WINED3DSPR_ATTROUT
:
1011 sprintf(tmpStr
, "gl_FrontColor");
1013 sprintf(tmpStr
, "gl_FrontSecondaryColor");
1016 case WINED3DSPR_TEXCRDOUT
:
1017 /* Vertex shaders >= 3.0: WINED3DSPR_OUTPUT */
1018 if (WINED3DSHADER_VERSION_MAJOR(shader_version
) >= 3) sprintf(tmpStr
, "OUT[%u]", reg
);
1019 else sprintf(tmpStr
, "gl_TexCoord[%u]", reg
);
1021 case WINED3DSPR_MISCTYPE
:
1024 sprintf(tmpStr
, "vpos");
1025 } else if (reg
== 1){
1026 /* Note that gl_FrontFacing is a bool, while vFace is
1027 * a float for which the sign determines front/back
1029 sprintf(tmpStr
, "(gl_FrontFacing ? 1.0 : -1.0)");
1031 FIXME("Unhandled misctype register %d\n", reg
);
1032 sprintf(tmpStr
, "unrecognized_register");
1036 FIXME("Unhandled register name Type(%d)\n", regtype
);
1037 sprintf(tmpStr
, "unrecognized_register");
1041 strcat(regstr
, tmpStr
);
1044 /* Get the GLSL write mask for the destination register */
1045 static DWORD
shader_glsl_get_write_mask(const DWORD param
, char *write_mask
) {
1046 char *ptr
= write_mask
;
1047 DWORD mask
= param
& WINED3DSP_WRITEMASK_ALL
;
1049 if (shader_is_scalar(param
)) {
1050 mask
= WINED3DSP_WRITEMASK_0
;
1053 if (param
& WINED3DSP_WRITEMASK_0
) *ptr
++ = 'x';
1054 if (param
& WINED3DSP_WRITEMASK_1
) *ptr
++ = 'y';
1055 if (param
& WINED3DSP_WRITEMASK_2
) *ptr
++ = 'z';
1056 if (param
& WINED3DSP_WRITEMASK_3
) *ptr
++ = 'w';
1064 static unsigned int shader_glsl_get_write_mask_size(DWORD write_mask
) {
1065 unsigned int size
= 0;
1067 if (write_mask
& WINED3DSP_WRITEMASK_0
) ++size
;
1068 if (write_mask
& WINED3DSP_WRITEMASK_1
) ++size
;
1069 if (write_mask
& WINED3DSP_WRITEMASK_2
) ++size
;
1070 if (write_mask
& WINED3DSP_WRITEMASK_3
) ++size
;
1075 static void shader_glsl_get_swizzle(const DWORD param
, BOOL fixup
, DWORD mask
, char *swizzle_str
) {
1076 /* For registers of type WINED3DDECLTYPE_D3DCOLOR, data is stored as "bgra",
1077 * but addressed as "rgba". To fix this we need to swap the register's x
1078 * and z components. */
1079 DWORD swizzle
= (param
& WINED3DSP_SWIZZLE_MASK
) >> WINED3DSP_SWIZZLE_SHIFT
;
1080 const char *swizzle_chars
= fixup
? "zyxw" : "xyzw";
1081 char *ptr
= swizzle_str
;
1083 if (!shader_is_scalar(param
)) {
1085 /* swizzle bits fields: wwzzyyxx */
1086 if (mask
& WINED3DSP_WRITEMASK_0
) *ptr
++ = swizzle_chars
[swizzle
& 0x03];
1087 if (mask
& WINED3DSP_WRITEMASK_1
) *ptr
++ = swizzle_chars
[(swizzle
>> 2) & 0x03];
1088 if (mask
& WINED3DSP_WRITEMASK_2
) *ptr
++ = swizzle_chars
[(swizzle
>> 4) & 0x03];
1089 if (mask
& WINED3DSP_WRITEMASK_3
) *ptr
++ = swizzle_chars
[(swizzle
>> 6) & 0x03];
1095 /* From a given parameter token, generate the corresponding GLSL string.
1096 * Also, return the actual register name and swizzle in case the
1097 * caller needs this information as well. */
1098 static void shader_glsl_add_src_param(const SHADER_OPCODE_ARG
*arg
, const DWORD param
,
1099 const DWORD addr_token
, DWORD mask
, glsl_src_param_t
*src_param
)
1101 BOOL is_color
= FALSE
;
1102 char swizzle_str
[6];
1104 src_param
->reg_name
[0] = '\0';
1105 src_param
->param_str
[0] = '\0';
1106 swizzle_str
[0] = '\0';
1108 shader_glsl_get_register_name(param
, addr_token
, src_param
->reg_name
, &is_color
, arg
);
1110 shader_glsl_get_swizzle(param
, is_color
, mask
, swizzle_str
);
1111 shader_glsl_gen_modifier(param
, src_param
->reg_name
, swizzle_str
, src_param
->param_str
);
1114 /* From a given parameter token, generate the corresponding GLSL string.
1115 * Also, return the actual register name and swizzle in case the
1116 * caller needs this information as well. */
1117 static DWORD
shader_glsl_add_dst_param(const SHADER_OPCODE_ARG
* arg
, const DWORD param
,
1118 const DWORD addr_token
, glsl_dst_param_t
*dst_param
)
1120 BOOL is_color
= FALSE
;
1122 dst_param
->mask_str
[0] = '\0';
1123 dst_param
->reg_name
[0] = '\0';
1125 shader_glsl_get_register_name(param
, addr_token
, dst_param
->reg_name
, &is_color
, arg
);
1126 return shader_glsl_get_write_mask(param
, dst_param
->mask_str
);
1129 /* Append the destination part of the instruction to the buffer, return the effective write mask */
1130 static DWORD
shader_glsl_append_dst_ext(SHADER_BUFFER
*buffer
, const SHADER_OPCODE_ARG
*arg
, const DWORD param
)
1132 glsl_dst_param_t dst_param
;
1136 mask
= shader_glsl_add_dst_param(arg
, param
, arg
->dst_addr
, &dst_param
);
1139 shift
= (param
& WINED3DSP_DSTSHIFT_MASK
) >> WINED3DSP_DSTSHIFT_SHIFT
;
1140 shader_addline(buffer
, "%s%s = %s(", dst_param
.reg_name
, dst_param
.mask_str
, shift_glsl_tab
[shift
]);
1146 /* Append the destination part of the instruction to the buffer, return the effective write mask */
1147 static DWORD
shader_glsl_append_dst(SHADER_BUFFER
*buffer
, const SHADER_OPCODE_ARG
*arg
)
1149 return shader_glsl_append_dst_ext(buffer
, arg
, arg
->dst
);
1152 /** Process GLSL instruction modifiers */
1153 void shader_glsl_add_instruction_modifiers(const SHADER_OPCODE_ARG
* arg
)
1155 DWORD mask
= arg
->dst
& WINED3DSP_DSTMOD_MASK
;
1157 if (arg
->opcode
->dst_token
&& mask
!= 0) {
1158 glsl_dst_param_t dst_param
;
1160 shader_glsl_add_dst_param(arg
, arg
->dst
, 0, &dst_param
);
1162 if (mask
& WINED3DSPDM_SATURATE
) {
1163 /* _SAT means to clamp the value of the register to between 0 and 1 */
1164 shader_addline(arg
->buffer
, "%s%s = clamp(%s%s, 0.0, 1.0);\n", dst_param
.reg_name
,
1165 dst_param
.mask_str
, dst_param
.reg_name
, dst_param
.mask_str
);
1167 if (mask
& WINED3DSPDM_MSAMPCENTROID
) {
1168 FIXME("_centroid modifier not handled\n");
1170 if (mask
& WINED3DSPDM_PARTIALPRECISION
) {
1171 /* MSDN says this modifier can be safely ignored, so that's what we'll do. */
1176 static inline const char* shader_get_comp_op(
1177 const DWORD opcode
) {
1179 DWORD op
= (opcode
& INST_CONTROLS_MASK
) >> INST_CONTROLS_SHIFT
;
1181 case COMPARISON_GT
: return ">";
1182 case COMPARISON_EQ
: return "==";
1183 case COMPARISON_GE
: return ">=";
1184 case COMPARISON_LT
: return "<";
1185 case COMPARISON_NE
: return "!=";
1186 case COMPARISON_LE
: return "<=";
1188 FIXME("Unrecognized comparison value: %u\n", op
);
1193 static void shader_glsl_get_sample_function(DWORD sampler_type
, BOOL projected
, BOOL texrect
, glsl_sample_function_t
*sample_function
) {
1194 /* Note that there's no such thing as a projected cube texture. */
1195 switch(sampler_type
) {
1197 sample_function
->name
= projected
? "texture1DProj" : "texture1D";
1198 sample_function
->coord_mask
= WINED3DSP_WRITEMASK_0
;
1202 sample_function
->name
= projected
? "texture2DRectProj" : "texture2DRect";
1204 sample_function
->name
= projected
? "texture2DProj" : "texture2D";
1206 sample_function
->coord_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
;
1208 case WINED3DSTT_CUBE
:
1209 sample_function
->name
= "textureCube";
1210 sample_function
->coord_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
1212 case WINED3DSTT_VOLUME
:
1213 sample_function
->name
= projected
? "texture3DProj" : "texture3D";
1214 sample_function
->coord_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
1217 sample_function
->name
= "";
1218 sample_function
->coord_mask
= 0;
1219 FIXME("Unrecognized sampler type: %#x;\n", sampler_type
);
1224 static void shader_glsl_append_fixup_arg(char *arguments
, const char *reg_name
,
1225 BOOL sign_fixup
, enum fixup_channel_source channel_source
)
1227 switch(channel_source
)
1229 case CHANNEL_SOURCE_ZERO
:
1230 strcat(arguments
, "0.0");
1233 case CHANNEL_SOURCE_ONE
:
1234 strcat(arguments
, "1.0");
1237 case CHANNEL_SOURCE_X
:
1238 strcat(arguments
, reg_name
);
1239 strcat(arguments
, ".x");
1242 case CHANNEL_SOURCE_Y
:
1243 strcat(arguments
, reg_name
);
1244 strcat(arguments
, ".y");
1247 case CHANNEL_SOURCE_Z
:
1248 strcat(arguments
, reg_name
);
1249 strcat(arguments
, ".z");
1252 case CHANNEL_SOURCE_W
:
1253 strcat(arguments
, reg_name
);
1254 strcat(arguments
, ".w");
1258 FIXME("Unhandled channel source %#x\n", channel_source
);
1259 strcat(arguments
, "undefined");
1263 if (sign_fixup
) strcat(arguments
, " * 2.0 - 1.0");
1266 static void shader_glsl_color_correction(const struct SHADER_OPCODE_ARG
*arg
, struct color_fixup_desc fixup
)
1268 unsigned int mask_size
, remaining
;
1269 glsl_dst_param_t dst_param
;
1270 char arguments
[256];
1275 if (fixup
.x_sign_fixup
|| fixup
.x_source
!= CHANNEL_SOURCE_X
) mask
|= WINED3DSP_WRITEMASK_0
;
1276 if (fixup
.y_sign_fixup
|| fixup
.y_source
!= CHANNEL_SOURCE_Y
) mask
|= WINED3DSP_WRITEMASK_1
;
1277 if (fixup
.z_sign_fixup
|| fixup
.z_source
!= CHANNEL_SOURCE_Z
) mask
|= WINED3DSP_WRITEMASK_2
;
1278 if (fixup
.w_sign_fixup
|| fixup
.w_source
!= CHANNEL_SOURCE_W
) mask
|= WINED3DSP_WRITEMASK_3
;
1281 if (!mask
) return; /* Nothing to do */
1283 if (is_yuv_fixup(fixup
))
1285 enum yuv_fixup yuv_fixup
= get_yuv_fixup(fixup
);
1286 FIXME("YUV fixup (%#x) not supported\n", yuv_fixup
);
1290 mask_size
= shader_glsl_get_write_mask_size(mask
);
1292 dst_param
.mask_str
[0] = '\0';
1293 shader_glsl_get_write_mask(mask
, dst_param
.mask_str
);
1295 dst_param
.reg_name
[0] = '\0';
1296 shader_glsl_get_register_name(arg
->dst
, arg
->dst_addr
, dst_param
.reg_name
, &dummy
, arg
);
1298 arguments
[0] = '\0';
1299 remaining
= mask_size
;
1300 if (mask
& WINED3DSP_WRITEMASK_0
)
1302 shader_glsl_append_fixup_arg(arguments
, dst_param
.reg_name
, fixup
.x_sign_fixup
, fixup
.x_source
);
1303 if (--remaining
) strcat(arguments
, ", ");
1305 if (mask
& WINED3DSP_WRITEMASK_1
)
1307 shader_glsl_append_fixup_arg(arguments
, dst_param
.reg_name
, fixup
.y_sign_fixup
, fixup
.y_source
);
1308 if (--remaining
) strcat(arguments
, ", ");
1310 if (mask
& WINED3DSP_WRITEMASK_2
)
1312 shader_glsl_append_fixup_arg(arguments
, dst_param
.reg_name
, fixup
.z_sign_fixup
, fixup
.z_source
);
1313 if (--remaining
) strcat(arguments
, ", ");
1315 if (mask
& WINED3DSP_WRITEMASK_3
)
1317 shader_glsl_append_fixup_arg(arguments
, dst_param
.reg_name
, fixup
.w_sign_fixup
, fixup
.w_source
);
1318 if (--remaining
) strcat(arguments
, ", ");
1323 shader_addline(arg
->buffer
, "%s%s = vec%u(%s);\n",
1324 dst_param
.reg_name
, dst_param
.mask_str
, mask_size
, arguments
);
1328 shader_addline(arg
->buffer
, "%s%s = %s;\n", dst_param
.reg_name
, dst_param
.mask_str
, arguments
);
1332 /*****************************************************************************
1334 * Begin processing individual instruction opcodes
1336 ****************************************************************************/
1338 /* Generate GLSL arithmetic functions (dst = src1 + src2) */
1339 static void shader_glsl_arith(const SHADER_OPCODE_ARG
*arg
)
1341 CONST SHADER_OPCODE
* curOpcode
= arg
->opcode
;
1342 SHADER_BUFFER
* buffer
= arg
->buffer
;
1343 glsl_src_param_t src0_param
;
1344 glsl_src_param_t src1_param
;
1348 /* Determine the GLSL operator to use based on the opcode */
1349 switch (curOpcode
->opcode
) {
1350 case WINED3DSIO_MUL
: op
= '*'; break;
1351 case WINED3DSIO_ADD
: op
= '+'; break;
1352 case WINED3DSIO_SUB
: op
= '-'; break;
1355 FIXME("Opcode %s not yet handled in GLSL\n", curOpcode
->name
);
1359 write_mask
= shader_glsl_append_dst(buffer
, arg
);
1360 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], write_mask
, &src0_param
);
1361 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], write_mask
, &src1_param
);
1362 shader_addline(buffer
, "%s %c %s);\n", src0_param
.param_str
, op
, src1_param
.param_str
);
1365 /* Process the WINED3DSIO_MOV opcode using GLSL (dst = src) */
1366 static void shader_glsl_mov(const SHADER_OPCODE_ARG
*arg
)
1368 SHADER_BUFFER
* buffer
= arg
->buffer
;
1369 glsl_src_param_t src0_param
;
1372 write_mask
= shader_glsl_append_dst(buffer
, arg
);
1373 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], write_mask
, &src0_param
);
1375 /* In vs_1_1 WINED3DSIO_MOV can write to the address register. In later
1376 * shader versions WINED3DSIO_MOVA is used for this. */
1377 if ((WINED3DSHADER_VERSION_MAJOR(arg
->reg_maps
->shader_version
) == 1
1378 && !shader_is_pshader_version(arg
->reg_maps
->shader_version
)
1379 && shader_get_regtype(arg
->dst
) == WINED3DSPR_ADDR
))
1381 /* This is a simple floor() */
1382 unsigned int mask_size
= shader_glsl_get_write_mask_size(write_mask
);
1383 if (mask_size
> 1) {
1384 shader_addline(buffer
, "ivec%d(floor(%s)));\n", mask_size
, src0_param
.param_str
);
1386 shader_addline(buffer
, "int(floor(%s)));\n", src0_param
.param_str
);
1388 } else if(arg
->opcode
->opcode
== WINED3DSIO_MOVA
) {
1389 /* We need to *round* to the nearest int here. */
1390 unsigned int mask_size
= shader_glsl_get_write_mask_size(write_mask
);
1391 if (mask_size
> 1) {
1392 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
);
1394 shader_addline(buffer
, "int(floor(abs(%s) + 0.5) * sign(%s)));\n", src0_param
.param_str
, src0_param
.param_str
);
1397 shader_addline(buffer
, "%s);\n", src0_param
.param_str
);
1401 /* Process the dot product operators DP3 and DP4 in GLSL (dst = dot(src0, src1)) */
1402 static void shader_glsl_dot(const SHADER_OPCODE_ARG
*arg
)
1404 CONST SHADER_OPCODE
* curOpcode
= arg
->opcode
;
1405 SHADER_BUFFER
* buffer
= arg
->buffer
;
1406 glsl_src_param_t src0_param
;
1407 glsl_src_param_t src1_param
;
1408 DWORD dst_write_mask
, src_write_mask
;
1409 unsigned int dst_size
= 0;
1411 dst_write_mask
= shader_glsl_append_dst(buffer
, arg
);
1412 dst_size
= shader_glsl_get_write_mask_size(dst_write_mask
);
1414 /* dp3 works on vec3, dp4 on vec4 */
1415 if (curOpcode
->opcode
== WINED3DSIO_DP4
) {
1416 src_write_mask
= WINED3DSP_WRITEMASK_ALL
;
1418 src_write_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
1421 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], src_write_mask
, &src0_param
);
1422 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], src_write_mask
, &src1_param
);
1425 shader_addline(buffer
, "vec%d(dot(%s, %s)));\n", dst_size
, src0_param
.param_str
, src1_param
.param_str
);
1427 shader_addline(buffer
, "dot(%s, %s));\n", src0_param
.param_str
, src1_param
.param_str
);
1431 /* Note that this instruction has some restrictions. The destination write mask
1432 * can't contain the w component, and the source swizzles have to be .xyzw */
1433 static void shader_glsl_cross(const SHADER_OPCODE_ARG
*arg
)
1435 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
1436 glsl_src_param_t src0_param
;
1437 glsl_src_param_t src1_param
;
1440 shader_glsl_get_write_mask(arg
->dst
, dst_mask
);
1441 shader_glsl_append_dst(arg
->buffer
, arg
);
1442 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], src_mask
, &src0_param
);
1443 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], src_mask
, &src1_param
);
1444 shader_addline(arg
->buffer
, "cross(%s, %s)%s);\n", src0_param
.param_str
, src1_param
.param_str
, dst_mask
);
1447 /* Process the WINED3DSIO_POW instruction in GLSL (dst = |src0|^src1)
1448 * Src0 and src1 are scalars. Note that D3D uses the absolute of src0, while
1449 * GLSL uses the value as-is. */
1450 static void shader_glsl_pow(const SHADER_OPCODE_ARG
*arg
)
1452 SHADER_BUFFER
*buffer
= arg
->buffer
;
1453 glsl_src_param_t src0_param
;
1454 glsl_src_param_t src1_param
;
1455 DWORD dst_write_mask
;
1456 unsigned int dst_size
;
1458 dst_write_mask
= shader_glsl_append_dst(buffer
, arg
);
1459 dst_size
= shader_glsl_get_write_mask_size(dst_write_mask
);
1461 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
1462 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], WINED3DSP_WRITEMASK_0
, &src1_param
);
1465 shader_addline(buffer
, "vec%d(pow(abs(%s), %s)));\n", dst_size
, src0_param
.param_str
, src1_param
.param_str
);
1467 shader_addline(buffer
, "pow(abs(%s), %s));\n", src0_param
.param_str
, src1_param
.param_str
);
1471 /* Process the WINED3DSIO_LOG instruction in GLSL (dst = log2(|src0|))
1472 * Src0 is a scalar. Note that D3D uses the absolute of src0, while
1473 * GLSL uses the value as-is. */
1474 static void shader_glsl_log(const SHADER_OPCODE_ARG
*arg
)
1476 SHADER_BUFFER
*buffer
= arg
->buffer
;
1477 glsl_src_param_t src0_param
;
1478 DWORD dst_write_mask
;
1479 unsigned int dst_size
;
1481 dst_write_mask
= shader_glsl_append_dst(buffer
, arg
);
1482 dst_size
= shader_glsl_get_write_mask_size(dst_write_mask
);
1484 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
1487 shader_addline(buffer
, "vec%d(log2(abs(%s))));\n", dst_size
, src0_param
.param_str
);
1489 shader_addline(buffer
, "log2(abs(%s)));\n", src0_param
.param_str
);
1493 /* Map the opcode 1-to-1 to the GL code (arg->dst = instruction(src0, src1, ...) */
1494 static void shader_glsl_map2gl(const SHADER_OPCODE_ARG
*arg
)
1496 CONST SHADER_OPCODE
* curOpcode
= arg
->opcode
;
1497 SHADER_BUFFER
* buffer
= arg
->buffer
;
1498 glsl_src_param_t src_param
;
1499 const char *instruction
;
1500 char arguments
[256];
1504 /* Determine the GLSL function to use based on the opcode */
1505 /* TODO: Possibly make this a table for faster lookups */
1506 switch (curOpcode
->opcode
) {
1507 case WINED3DSIO_MIN
: instruction
= "min"; break;
1508 case WINED3DSIO_MAX
: instruction
= "max"; break;
1509 case WINED3DSIO_ABS
: instruction
= "abs"; break;
1510 case WINED3DSIO_FRC
: instruction
= "fract"; break;
1511 case WINED3DSIO_NRM
: instruction
= "normalize"; break;
1512 case WINED3DSIO_EXP
: instruction
= "exp2"; break;
1513 case WINED3DSIO_SGN
: instruction
= "sign"; break;
1514 case WINED3DSIO_DSX
: instruction
= "dFdx"; break;
1515 case WINED3DSIO_DSY
: instruction
= "ycorrection.y * dFdy"; break;
1516 default: instruction
= "";
1517 FIXME("Opcode %s not yet handled in GLSL\n", curOpcode
->name
);
1521 write_mask
= shader_glsl_append_dst(buffer
, arg
);
1523 arguments
[0] = '\0';
1524 if (curOpcode
->num_params
> 0) {
1525 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], write_mask
, &src_param
);
1526 strcat(arguments
, src_param
.param_str
);
1527 for (i
= 2; i
< curOpcode
->num_params
; ++i
) {
1528 strcat(arguments
, ", ");
1529 shader_glsl_add_src_param(arg
, arg
->src
[i
-1], arg
->src_addr
[i
-1], write_mask
, &src_param
);
1530 strcat(arguments
, src_param
.param_str
);
1534 shader_addline(buffer
, "%s(%s));\n", instruction
, arguments
);
1537 /** Process the WINED3DSIO_EXPP instruction in GLSL:
1538 * For shader model 1.x, do the following (and honor the writemask, so use a temporary variable):
1539 * dst.x = 2^(floor(src))
1540 * dst.y = src - floor(src)
1541 * dst.z = 2^src (partial precision is allowed, but optional)
1543 * For 2.0 shaders, just do this (honoring writemask and swizzle):
1544 * dst = 2^src; (partial precision is allowed, but optional)
1546 static void shader_glsl_expp(const SHADER_OPCODE_ARG
*arg
)
1548 glsl_src_param_t src_param
;
1550 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src_param
);
1552 if (arg
->reg_maps
->shader_version
< WINED3DPS_VERSION(2,0))
1556 shader_addline(arg
->buffer
, "tmp0.x = exp2(floor(%s));\n", src_param
.param_str
);
1557 shader_addline(arg
->buffer
, "tmp0.y = %s - floor(%s);\n", src_param
.param_str
, src_param
.param_str
);
1558 shader_addline(arg
->buffer
, "tmp0.z = exp2(%s);\n", src_param
.param_str
);
1559 shader_addline(arg
->buffer
, "tmp0.w = 1.0;\n");
1561 shader_glsl_append_dst(arg
->buffer
, arg
);
1562 shader_glsl_get_write_mask(arg
->dst
, dst_mask
);
1563 shader_addline(arg
->buffer
, "tmp0%s);\n", dst_mask
);
1566 unsigned int mask_size
;
1568 write_mask
= shader_glsl_append_dst(arg
->buffer
, arg
);
1569 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
1571 if (mask_size
> 1) {
1572 shader_addline(arg
->buffer
, "vec%d(exp2(%s)));\n", mask_size
, src_param
.param_str
);
1574 shader_addline(arg
->buffer
, "exp2(%s));\n", src_param
.param_str
);
1579 /** Process the RCP (reciprocal or inverse) opcode in GLSL (dst = 1 / src) */
1580 static void shader_glsl_rcp(const SHADER_OPCODE_ARG
*arg
)
1582 glsl_src_param_t src_param
;
1584 unsigned int mask_size
;
1586 write_mask
= shader_glsl_append_dst(arg
->buffer
, arg
);
1587 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
1588 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_3
, &src_param
);
1590 if (mask_size
> 1) {
1591 shader_addline(arg
->buffer
, "vec%d(1.0 / %s));\n", mask_size
, src_param
.param_str
);
1593 shader_addline(arg
->buffer
, "1.0 / %s);\n", src_param
.param_str
);
1597 static void shader_glsl_rsq(const SHADER_OPCODE_ARG
*arg
)
1599 SHADER_BUFFER
* buffer
= arg
->buffer
;
1600 glsl_src_param_t src_param
;
1602 unsigned int mask_size
;
1604 write_mask
= shader_glsl_append_dst(buffer
, arg
);
1605 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
1607 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_3
, &src_param
);
1609 if (mask_size
> 1) {
1610 shader_addline(buffer
, "vec%d(inversesqrt(%s)));\n", mask_size
, src_param
.param_str
);
1612 shader_addline(buffer
, "inversesqrt(%s));\n", src_param
.param_str
);
1616 /** Process signed comparison opcodes in GLSL. */
1617 static void shader_glsl_compare(const SHADER_OPCODE_ARG
*arg
)
1619 glsl_src_param_t src0_param
;
1620 glsl_src_param_t src1_param
;
1622 unsigned int mask_size
;
1624 write_mask
= shader_glsl_append_dst(arg
->buffer
, arg
);
1625 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
1626 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], write_mask
, &src0_param
);
1627 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], write_mask
, &src1_param
);
1629 if (mask_size
> 1) {
1630 const char *compare
;
1632 switch(arg
->opcode
->opcode
) {
1633 case WINED3DSIO_SLT
: compare
= "lessThan"; break;
1634 case WINED3DSIO_SGE
: compare
= "greaterThanEqual"; break;
1635 default: compare
= "";
1636 FIXME("Can't handle opcode %s\n", arg
->opcode
->name
);
1639 shader_addline(arg
->buffer
, "vec%d(%s(%s, %s)));\n", mask_size
, compare
,
1640 src0_param
.param_str
, src1_param
.param_str
);
1642 switch(arg
->opcode
->opcode
) {
1643 case WINED3DSIO_SLT
:
1644 /* Step(src0, src1) is not suitable here because if src0 == src1 SLT is supposed,
1645 * to return 0.0 but step returns 1.0 because step is not < x
1646 * An alternative is a bvec compare padded with an unused second component.
1647 * step(src1 * -1.0, src0 * -1.0) is not an option because it suffers from the same
1648 * issue. Playing with not() is not possible either because not() does not accept
1651 shader_addline(arg
->buffer
, "(%s < %s) ? 1.0 : 0.0);\n", src0_param
.param_str
, src1_param
.param_str
);
1653 case WINED3DSIO_SGE
:
1654 /* Here we can use the step() function and safe a conditional */
1655 shader_addline(arg
->buffer
, "step(%s, %s));\n", src1_param
.param_str
, src0_param
.param_str
);
1658 FIXME("Can't handle opcode %s\n", arg
->opcode
->name
);
1664 /** Process CMP instruction in GLSL (dst = src0 >= 0.0 ? src1 : src2), per channel */
1665 static void shader_glsl_cmp(const SHADER_OPCODE_ARG
*arg
)
1667 glsl_src_param_t src0_param
;
1668 glsl_src_param_t src1_param
;
1669 glsl_src_param_t src2_param
;
1670 DWORD write_mask
, cmp_channel
= 0;
1673 BOOL temp_destination
= FALSE
;
1675 if(shader_is_scalar(arg
->src
[0])) {
1676 write_mask
= shader_glsl_append_dst(arg
->buffer
, arg
);
1678 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_ALL
, &src0_param
);
1679 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], write_mask
, &src1_param
);
1680 shader_glsl_add_src_param(arg
, arg
->src
[2], arg
->src_addr
[2], write_mask
, &src2_param
);
1682 shader_addline(arg
->buffer
, "%s >= 0.0 ? %s : %s);\n",
1683 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
1685 DWORD src0reg
= arg
->src
[0] & WINED3DSP_REGNUM_MASK
;
1686 DWORD src1reg
= arg
->src
[1] & WINED3DSP_REGNUM_MASK
;
1687 DWORD src2reg
= arg
->src
[2] & WINED3DSP_REGNUM_MASK
;
1688 DWORD src0regtype
= shader_get_regtype(arg
->src
[0]);
1689 DWORD src1regtype
= shader_get_regtype(arg
->src
[1]);
1690 DWORD src2regtype
= shader_get_regtype(arg
->src
[2]);
1691 DWORD dstreg
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
1692 DWORD dstregtype
= shader_get_regtype(arg
->dst
);
1694 /* Cycle through all source0 channels */
1695 for (i
=0; i
<4; i
++) {
1697 /* Find the destination channels which use the current source0 channel */
1698 for (j
=0; j
<4; j
++) {
1699 if ( ((arg
->src
[0] >> (WINED3DSP_SWIZZLE_SHIFT
+ 2*j
)) & 0x3) == i
) {
1700 write_mask
|= WINED3DSP_WRITEMASK_0
<< j
;
1701 cmp_channel
= WINED3DSP_WRITEMASK_0
<< j
;
1705 /* Splitting the cmp instruction up in multiple lines imposes a problem:
1706 * The first lines may overwrite source parameters of the following lines.
1707 * Deal with that by using a temporary destination register if needed
1709 if((src0reg
== dstreg
&& src0regtype
== dstregtype
) ||
1710 (src1reg
== dstreg
&& src1regtype
== dstregtype
) ||
1711 (src2reg
== dstreg
&& src2regtype
== dstregtype
)) {
1713 write_mask
= shader_glsl_get_write_mask(arg
->dst
& (~WINED3DSP_SWIZZLE_MASK
| write_mask
), mask_char
);
1714 if (!write_mask
) continue;
1715 shader_addline(arg
->buffer
, "tmp0%s = (", mask_char
);
1716 temp_destination
= TRUE
;
1718 write_mask
= shader_glsl_append_dst_ext(arg
->buffer
, arg
, arg
->dst
& (~WINED3DSP_SWIZZLE_MASK
| write_mask
));
1719 if (!write_mask
) continue;
1722 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], cmp_channel
, &src0_param
);
1723 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], write_mask
, &src1_param
);
1724 shader_glsl_add_src_param(arg
, arg
->src
[2], arg
->src_addr
[2], write_mask
, &src2_param
);
1726 shader_addline(arg
->buffer
, "%s >= 0.0 ? %s : %s);\n",
1727 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
1730 if(temp_destination
) {
1731 shader_glsl_get_write_mask(arg
->dst
, mask_char
);
1732 shader_glsl_append_dst_ext(arg
->buffer
, arg
, arg
->dst
);
1733 shader_addline(arg
->buffer
, "tmp0%s);\n", mask_char
);
1739 /** Process the CND opcode in GLSL (dst = (src0 > 0.5) ? src1 : src2) */
1740 /* For ps 1.1-1.3, only a single component of src0 is used. For ps 1.4
1741 * the compare is done per component of src0. */
1742 static void shader_glsl_cnd(const SHADER_OPCODE_ARG
*arg
)
1744 glsl_src_param_t src0_param
;
1745 glsl_src_param_t src1_param
;
1746 glsl_src_param_t src2_param
;
1747 DWORD write_mask
, cmp_channel
= 0;
1750 if (arg
->reg_maps
->shader_version
< WINED3DPS_VERSION(1, 4))
1752 write_mask
= shader_glsl_append_dst(arg
->buffer
, arg
);
1753 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
1754 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], write_mask
, &src1_param
);
1755 shader_glsl_add_src_param(arg
, arg
->src
[2], arg
->src_addr
[2], write_mask
, &src2_param
);
1757 /* Fun: The D3DSI_COISSUE flag changes the semantic of the cnd instruction for < 1.4 shaders */
1758 if(arg
->opcode_token
& WINED3DSI_COISSUE
) {
1759 shader_addline(arg
->buffer
, "%s /* COISSUE! */);\n", src1_param
.param_str
);
1761 shader_addline(arg
->buffer
, "%s > 0.5 ? %s : %s);\n",
1762 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
1766 /* Cycle through all source0 channels */
1767 for (i
=0; i
<4; i
++) {
1769 /* Find the destination channels which use the current source0 channel */
1770 for (j
=0; j
<4; j
++) {
1771 if ( ((arg
->src
[0] >> (WINED3DSP_SWIZZLE_SHIFT
+ 2*j
)) & 0x3) == i
) {
1772 write_mask
|= WINED3DSP_WRITEMASK_0
<< j
;
1773 cmp_channel
= WINED3DSP_WRITEMASK_0
<< j
;
1776 write_mask
= shader_glsl_append_dst_ext(arg
->buffer
, arg
, arg
->dst
& (~WINED3DSP_SWIZZLE_MASK
| write_mask
));
1777 if (!write_mask
) continue;
1779 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], cmp_channel
, &src0_param
);
1780 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], write_mask
, &src1_param
);
1781 shader_glsl_add_src_param(arg
, arg
->src
[2], arg
->src_addr
[2], write_mask
, &src2_param
);
1783 shader_addline(arg
->buffer
, "%s > 0.5 ? %s : %s);\n",
1784 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
1788 /** GLSL code generation for WINED3DSIO_MAD: Multiply the first 2 opcodes, then add the last */
1789 static void shader_glsl_mad(const SHADER_OPCODE_ARG
*arg
)
1791 glsl_src_param_t src0_param
;
1792 glsl_src_param_t src1_param
;
1793 glsl_src_param_t src2_param
;
1796 write_mask
= shader_glsl_append_dst(arg
->buffer
, arg
);
1797 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], write_mask
, &src0_param
);
1798 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], write_mask
, &src1_param
);
1799 shader_glsl_add_src_param(arg
, arg
->src
[2], arg
->src_addr
[2], write_mask
, &src2_param
);
1800 shader_addline(arg
->buffer
, "(%s * %s) + %s);\n",
1801 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
1804 /** Handles transforming all WINED3DSIO_M?x? opcodes for
1805 Vertex shaders to GLSL codes */
1806 static void shader_glsl_mnxn(const SHADER_OPCODE_ARG
*arg
)
1808 IWineD3DBaseShaderImpl
*shader
= (IWineD3DBaseShaderImpl
*)arg
->shader
;
1809 const SHADER_OPCODE
*opcode_table
= shader
->baseShader
.shader_ins
;
1810 DWORD shader_version
= arg
->reg_maps
->shader_version
;
1812 int nComponents
= 0;
1813 SHADER_OPCODE_ARG tmpArg
;
1815 memset(&tmpArg
, 0, sizeof(SHADER_OPCODE_ARG
));
1817 /* Set constants for the temporary argument */
1818 tmpArg
.shader
= arg
->shader
;
1819 tmpArg
.buffer
= arg
->buffer
;
1820 tmpArg
.src
[0] = arg
->src
[0];
1821 tmpArg
.src_addr
[0] = arg
->src_addr
[0];
1822 tmpArg
.src_addr
[1] = arg
->src_addr
[1];
1823 tmpArg
.reg_maps
= arg
->reg_maps
;
1825 switch(arg
->opcode
->opcode
) {
1826 case WINED3DSIO_M4x4
:
1828 tmpArg
.opcode
= shader_get_opcode(opcode_table
, shader_version
, WINED3DSIO_DP4
);
1830 case WINED3DSIO_M4x3
:
1832 tmpArg
.opcode
= shader_get_opcode(opcode_table
, shader_version
, WINED3DSIO_DP4
);
1834 case WINED3DSIO_M3x4
:
1836 tmpArg
.opcode
= shader_get_opcode(opcode_table
, shader_version
, WINED3DSIO_DP3
);
1838 case WINED3DSIO_M3x3
:
1840 tmpArg
.opcode
= shader_get_opcode(opcode_table
, shader_version
, WINED3DSIO_DP3
);
1842 case WINED3DSIO_M3x2
:
1844 tmpArg
.opcode
= shader_get_opcode(opcode_table
, shader_version
, WINED3DSIO_DP3
);
1850 for (i
= 0; i
< nComponents
; i
++) {
1851 tmpArg
.dst
= ((arg
->dst
) & ~WINED3DSP_WRITEMASK_ALL
)|(WINED3DSP_WRITEMASK_0
<<i
);
1852 tmpArg
.src
[1] = arg
->src
[1]+i
;
1853 shader_glsl_dot(&tmpArg
);
1858 The LRP instruction performs a component-wise linear interpolation
1859 between the second and third operands using the first operand as the
1860 blend factor. Equation: (dst = src2 + src0 * (src1 - src2))
1861 This is equivalent to mix(src2, src1, src0);
1863 static void shader_glsl_lrp(const SHADER_OPCODE_ARG
*arg
)
1865 glsl_src_param_t src0_param
;
1866 glsl_src_param_t src1_param
;
1867 glsl_src_param_t src2_param
;
1870 write_mask
= shader_glsl_append_dst(arg
->buffer
, arg
);
1872 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], write_mask
, &src0_param
);
1873 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], write_mask
, &src1_param
);
1874 shader_glsl_add_src_param(arg
, arg
->src
[2], arg
->src_addr
[2], write_mask
, &src2_param
);
1876 shader_addline(arg
->buffer
, "mix(%s, %s, %s));\n",
1877 src2_param
.param_str
, src1_param
.param_str
, src0_param
.param_str
);
1880 /** Process the WINED3DSIO_LIT instruction in GLSL:
1881 * dst.x = dst.w = 1.0
1882 * dst.y = (src0.x > 0) ? src0.x
1883 * dst.z = (src0.x > 0) ? ((src0.y > 0) ? pow(src0.y, src.w) : 0) : 0
1884 * where src.w is clamped at +- 128
1886 static void shader_glsl_lit(const SHADER_OPCODE_ARG
*arg
)
1888 glsl_src_param_t src0_param
;
1889 glsl_src_param_t src1_param
;
1890 glsl_src_param_t src3_param
;
1893 shader_glsl_append_dst(arg
->buffer
, arg
);
1894 shader_glsl_get_write_mask(arg
->dst
, dst_mask
);
1896 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
1897 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_1
, &src1_param
);
1898 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_3
, &src3_param
);
1900 /* The sdk specifies the instruction like this
1902 * if(src.x > 0.0) dst.y = src.x
1904 * if(src.x > 0.0 && src.y > 0.0) dst.z = pow(src.y, power);
1908 * Obviously that has quite a few conditionals in it which we don't like. So the first step is this:
1909 * dst.x = 1.0 ... No further explanation needed
1910 * dst.y = max(src.y, 0.0); ... If x < 0.0, use 0.0, otherwise x. Same as the conditional
1911 * dst.z = x > 0.0 ? pow(max(y, 0.0), p) : 0; ... 0 ^ power is 0, and otherwise we use y anyway
1912 * dst.w = 1.0. ... Nothing fancy.
1914 * So we still have one conditional in there. So do this:
1915 * dst.z = pow(max(0.0, src.y) * step(0.0, src.x), power);
1917 * 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),
1918 * 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.
1919 * if both x and y are > 0, we get pow(y * 1.0, power), as it is supposed to
1921 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",
1922 src0_param
.param_str
, src1_param
.param_str
, src0_param
.param_str
, src3_param
.param_str
, dst_mask
);
1925 /** Process the WINED3DSIO_DST instruction in GLSL:
1927 * dst.y = src0.x * src0.y
1931 static void shader_glsl_dst(const SHADER_OPCODE_ARG
*arg
)
1933 glsl_src_param_t src0y_param
;
1934 glsl_src_param_t src0z_param
;
1935 glsl_src_param_t src1y_param
;
1936 glsl_src_param_t src1w_param
;
1939 shader_glsl_append_dst(arg
->buffer
, arg
);
1940 shader_glsl_get_write_mask(arg
->dst
, dst_mask
);
1942 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_1
, &src0y_param
);
1943 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_2
, &src0z_param
);
1944 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], WINED3DSP_WRITEMASK_1
, &src1y_param
);
1945 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], WINED3DSP_WRITEMASK_3
, &src1w_param
);
1947 shader_addline(arg
->buffer
, "vec4(1.0, %s * %s, %s, %s))%s;\n",
1948 src0y_param
.param_str
, src1y_param
.param_str
, src0z_param
.param_str
, src1w_param
.param_str
, dst_mask
);
1951 /** Process the WINED3DSIO_SINCOS instruction in GLSL:
1952 * VS 2.0 requires that specific cosine and sine constants be passed to this instruction so the hardware
1953 * can handle it. But, these functions are built-in for GLSL, so we can just ignore the last 2 params.
1955 * dst.x = cos(src0.?)
1956 * dst.y = sin(src0.?)
1960 static void shader_glsl_sincos(const SHADER_OPCODE_ARG
*arg
)
1962 glsl_src_param_t src0_param
;
1965 write_mask
= shader_glsl_append_dst(arg
->buffer
, arg
);
1966 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
1968 switch (write_mask
) {
1969 case WINED3DSP_WRITEMASK_0
:
1970 shader_addline(arg
->buffer
, "cos(%s));\n", src0_param
.param_str
);
1973 case WINED3DSP_WRITEMASK_1
:
1974 shader_addline(arg
->buffer
, "sin(%s));\n", src0_param
.param_str
);
1977 case (WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
):
1978 shader_addline(arg
->buffer
, "vec2(cos(%s), sin(%s)));\n", src0_param
.param_str
, src0_param
.param_str
);
1982 ERR("Write mask should be .x, .y or .xy\n");
1987 /** Process the WINED3DSIO_LOOP instruction in GLSL:
1988 * Start a for() loop where src1.y is the initial value of aL,
1989 * increment aL by src1.z for a total of src1.x iterations.
1990 * Need to use a temporary variable for this operation.
1992 /* FIXME: I don't think nested loops will work correctly this way. */
1993 static void shader_glsl_loop(const SHADER_OPCODE_ARG
*arg
)
1995 glsl_src_param_t src1_param
;
1996 IWineD3DBaseShaderImpl
* shader
= (IWineD3DBaseShaderImpl
*) arg
->shader
;
1997 DWORD regtype
= shader_get_regtype(arg
->src
[1]);
1998 DWORD reg
= arg
->src
[1] & WINED3DSP_REGNUM_MASK
;
1999 const DWORD
*control_values
= NULL
;
2000 const local_constant
*constant
;
2002 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], WINED3DSP_WRITEMASK_ALL
, &src1_param
);
2004 /* Try to hardcode the loop control parameters if possible. Direct3D 9 class hardware doesn't support real
2005 * varying indexing, but Microsoft designed this feature for Shader model 2.x+. If the loop control is
2006 * known at compile time, the GLSL compiler can unroll the loop, and replace indirect addressing with direct
2009 if(regtype
== WINED3DSPR_CONSTINT
) {
2010 LIST_FOR_EACH_ENTRY(constant
, &shader
->baseShader
.constantsI
, local_constant
, entry
) {
2011 if(constant
->idx
== reg
) {
2012 control_values
= constant
->value
;
2018 if(control_values
) {
2019 if(control_values
[2] > 0) {
2020 shader_addline(arg
->buffer
, "for (aL%u = %d; aL%u < (%d * %d + %d); aL%u += %d) {\n",
2021 shader
->baseShader
.cur_loop_depth
, control_values
[1],
2022 shader
->baseShader
.cur_loop_depth
, control_values
[0], control_values
[2], control_values
[1],
2023 shader
->baseShader
.cur_loop_depth
, control_values
[2]);
2024 } else if(control_values
[2] == 0) {
2025 shader_addline(arg
->buffer
, "for (aL%u = %d, tmpInt%u = 0; tmpInt%u < %d; tmpInt%u++) {\n",
2026 shader
->baseShader
.cur_loop_depth
, control_values
[1], shader
->baseShader
.cur_loop_depth
,
2027 shader
->baseShader
.cur_loop_depth
, control_values
[0],
2028 shader
->baseShader
.cur_loop_depth
);
2030 shader_addline(arg
->buffer
, "for (aL%u = %d; aL%u > (%d * %d + %d); aL%u += %d) {\n",
2031 shader
->baseShader
.cur_loop_depth
, control_values
[1],
2032 shader
->baseShader
.cur_loop_depth
, control_values
[0], control_values
[2], control_values
[1],
2033 shader
->baseShader
.cur_loop_depth
, control_values
[2]);
2036 shader_addline(arg
->buffer
, "for (tmpInt%u = 0, aL%u = %s.y; tmpInt%u < %s.x; tmpInt%u++, aL%u += %s.z) {\n",
2037 shader
->baseShader
.cur_loop_depth
, shader
->baseShader
.cur_loop_regno
,
2038 src1_param
.reg_name
, shader
->baseShader
.cur_loop_depth
, src1_param
.reg_name
,
2039 shader
->baseShader
.cur_loop_depth
, shader
->baseShader
.cur_loop_regno
, src1_param
.reg_name
);
2042 shader
->baseShader
.cur_loop_depth
++;
2043 shader
->baseShader
.cur_loop_regno
++;
2046 static void shader_glsl_end(const SHADER_OPCODE_ARG
*arg
)
2048 IWineD3DBaseShaderImpl
* shader
= (IWineD3DBaseShaderImpl
*) arg
->shader
;
2050 shader_addline(arg
->buffer
, "}\n");
2052 if(arg
->opcode
->opcode
== WINED3DSIO_ENDLOOP
) {
2053 shader
->baseShader
.cur_loop_depth
--;
2054 shader
->baseShader
.cur_loop_regno
--;
2056 if(arg
->opcode
->opcode
== WINED3DSIO_ENDREP
) {
2057 shader
->baseShader
.cur_loop_depth
--;
2061 static void shader_glsl_rep(const SHADER_OPCODE_ARG
*arg
)
2063 IWineD3DBaseShaderImpl
* shader
= (IWineD3DBaseShaderImpl
*) arg
->shader
;
2064 glsl_src_param_t src0_param
;
2066 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2067 shader_addline(arg
->buffer
, "for (tmpInt%d = 0; tmpInt%d < %s; tmpInt%d++) {\n",
2068 shader
->baseShader
.cur_loop_depth
, shader
->baseShader
.cur_loop_depth
,
2069 src0_param
.param_str
, shader
->baseShader
.cur_loop_depth
);
2070 shader
->baseShader
.cur_loop_depth
++;
2073 static void shader_glsl_if(const SHADER_OPCODE_ARG
*arg
)
2075 glsl_src_param_t src0_param
;
2077 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2078 shader_addline(arg
->buffer
, "if (%s) {\n", src0_param
.param_str
);
2081 static void shader_glsl_ifc(const SHADER_OPCODE_ARG
*arg
)
2083 glsl_src_param_t src0_param
;
2084 glsl_src_param_t src1_param
;
2086 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2087 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], WINED3DSP_WRITEMASK_0
, &src1_param
);
2089 shader_addline(arg
->buffer
, "if (%s %s %s) {\n",
2090 src0_param
.param_str
, shader_get_comp_op(arg
->opcode_token
), src1_param
.param_str
);
2093 static void shader_glsl_else(const SHADER_OPCODE_ARG
*arg
)
2095 shader_addline(arg
->buffer
, "} else {\n");
2098 static void shader_glsl_break(const SHADER_OPCODE_ARG
*arg
)
2100 shader_addline(arg
->buffer
, "break;\n");
2103 /* FIXME: According to MSDN the compare is done per component. */
2104 static void shader_glsl_breakc(const SHADER_OPCODE_ARG
*arg
)
2106 glsl_src_param_t src0_param
;
2107 glsl_src_param_t src1_param
;
2109 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2110 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], WINED3DSP_WRITEMASK_0
, &src1_param
);
2112 shader_addline(arg
->buffer
, "if (%s %s %s) break;\n",
2113 src0_param
.param_str
, shader_get_comp_op(arg
->opcode_token
), src1_param
.param_str
);
2116 static void shader_glsl_label(const SHADER_OPCODE_ARG
*arg
)
2119 DWORD snum
= (arg
->src
[0]) & WINED3DSP_REGNUM_MASK
;
2120 shader_addline(arg
->buffer
, "}\n");
2121 shader_addline(arg
->buffer
, "void subroutine%u () {\n", snum
);
2124 static void shader_glsl_call(const SHADER_OPCODE_ARG
*arg
)
2126 DWORD snum
= (arg
->src
[0]) & WINED3DSP_REGNUM_MASK
;
2127 shader_addline(arg
->buffer
, "subroutine%u();\n", snum
);
2130 static void shader_glsl_callnz(const SHADER_OPCODE_ARG
*arg
)
2132 glsl_src_param_t src1_param
;
2134 DWORD snum
= (arg
->src
[0]) & WINED3DSP_REGNUM_MASK
;
2135 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], WINED3DSP_WRITEMASK_0
, &src1_param
);
2136 shader_addline(arg
->buffer
, "if (%s) subroutine%u();\n", src1_param
.param_str
, snum
);
2139 /*********************************************
2140 * Pixel Shader Specific Code begins here
2141 ********************************************/
2142 static void pshader_glsl_tex(const SHADER_OPCODE_ARG
*arg
)
2144 IWineD3DPixelShaderImpl
* This
= (IWineD3DPixelShaderImpl
*) arg
->shader
;
2145 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
2146 DWORD shader_version
= arg
->reg_maps
->shader_version
;
2147 char dst_swizzle
[6];
2148 glsl_sample_function_t sample_function
;
2151 BOOL projected
, texrect
= FALSE
;
2154 /* All versions have a destination register */
2155 shader_glsl_append_dst(arg
->buffer
, arg
);
2157 /* 1.0-1.4: Use destination register as sampler source.
2158 * 2.0+: Use provided sampler source. */
2159 if (shader_version
< WINED3DPS_VERSION(2,0)) sampler_idx
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2160 else sampler_idx
= arg
->src
[1] & WINED3DSP_REGNUM_MASK
;
2161 sampler_type
= arg
->reg_maps
->samplers
[sampler_idx
] & WINED3DSP_TEXTURETYPE_MASK
;
2163 if (shader_version
< WINED3DPS_VERSION(1,4))
2165 DWORD flags
= deviceImpl
->stateBlock
->textureState
[sampler_idx
][WINED3DTSS_TEXTURETRANSFORMFLAGS
];
2167 /* Projected cube textures don't make a lot of sense, the resulting coordinates stay the same. */
2168 if (flags
& WINED3DTTFF_PROJECTED
&& sampler_type
!= WINED3DSTT_CUBE
) {
2170 switch (flags
& ~WINED3DTTFF_PROJECTED
) {
2171 case WINED3DTTFF_COUNT1
: FIXME("WINED3DTTFF_PROJECTED with WINED3DTTFF_COUNT1?\n"); break;
2172 case WINED3DTTFF_COUNT2
: mask
= WINED3DSP_WRITEMASK_1
; break;
2173 case WINED3DTTFF_COUNT3
: mask
= WINED3DSP_WRITEMASK_2
; break;
2174 case WINED3DTTFF_COUNT4
:
2175 case WINED3DTTFF_DISABLE
: mask
= WINED3DSP_WRITEMASK_3
; break;
2181 else if (shader_version
< WINED3DPS_VERSION(2,0))
2183 DWORD src_mod
= arg
->src
[0] & WINED3DSP_SRCMOD_MASK
;
2185 if (src_mod
== WINED3DSPSM_DZ
) {
2187 mask
= WINED3DSP_WRITEMASK_2
;
2188 } else if (src_mod
== WINED3DSPSM_DW
) {
2190 mask
= WINED3DSP_WRITEMASK_3
;
2195 if(arg
->opcode_token
& WINED3DSI_TEXLD_PROJECT
) {
2196 /* ps 2.0 texldp instruction always divides by the fourth component. */
2198 mask
= WINED3DSP_WRITEMASK_3
;
2204 if(deviceImpl
->stateBlock
->textures
[sampler_idx
] &&
2205 IWineD3DBaseTexture_GetTextureDimensions(deviceImpl
->stateBlock
->textures
[sampler_idx
]) == GL_TEXTURE_RECTANGLE_ARB
) {
2209 shader_glsl_get_sample_function(sampler_type
, projected
, texrect
, &sample_function
);
2210 mask
|= sample_function
.coord_mask
;
2212 if (shader_version
< WINED3DPS_VERSION(2,0)) shader_glsl_get_write_mask(arg
->dst
, dst_swizzle
);
2213 else shader_glsl_get_swizzle(arg
->src
[1], FALSE
, arg
->dst
, dst_swizzle
);
2215 /* 1.0-1.3: Use destination register as coordinate source.
2216 1.4+: Use provided coordinate source register. */
2217 if (shader_version
< WINED3DPS_VERSION(1,4))
2220 shader_glsl_get_write_mask(mask
, coord_mask
);
2221 shader_addline(arg
->buffer
, "%s(Psampler%u, T%u%s)%s);\n",
2222 sample_function
.name
, sampler_idx
, sampler_idx
, coord_mask
, dst_swizzle
);
2224 glsl_src_param_t coord_param
;
2225 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], mask
, &coord_param
);
2226 if(arg
->opcode_token
& WINED3DSI_TEXLD_BIAS
) {
2227 glsl_src_param_t bias
;
2228 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_3
, &bias
);
2230 shader_addline(arg
->buffer
, "%s(Psampler%u, %s, %s)%s);\n",
2231 sample_function
.name
, sampler_idx
, coord_param
.param_str
,
2232 bias
.param_str
, dst_swizzle
);
2234 shader_addline(arg
->buffer
, "%s(Psampler%u, %s)%s);\n",
2235 sample_function
.name
, sampler_idx
, coord_param
.param_str
, dst_swizzle
);
2240 static void shader_glsl_texldl(const SHADER_OPCODE_ARG
*arg
)
2242 IWineD3DBaseShaderImpl
* This
= (IWineD3DBaseShaderImpl
*)arg
->shader
;
2243 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
2244 glsl_sample_function_t sample_function
;
2245 glsl_src_param_t coord_param
, lod_param
;
2246 char dst_swizzle
[6];
2249 BOOL texrect
= FALSE
;
2251 shader_glsl_append_dst(arg
->buffer
, arg
);
2252 shader_glsl_get_swizzle(arg
->src
[1], FALSE
, arg
->dst
, dst_swizzle
);
2254 sampler_idx
= arg
->src
[1] & WINED3DSP_REGNUM_MASK
;
2255 sampler_type
= arg
->reg_maps
->samplers
[sampler_idx
] & WINED3DSP_TEXTURETYPE_MASK
;
2256 if(deviceImpl
->stateBlock
->textures
[sampler_idx
] &&
2257 IWineD3DBaseTexture_GetTextureDimensions(deviceImpl
->stateBlock
->textures
[sampler_idx
]) == GL_TEXTURE_RECTANGLE_ARB
) {
2260 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
);
2262 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_3
, &lod_param
);
2264 if (shader_is_pshader_version(arg
->reg_maps
->shader_version
))
2266 /* The GLSL spec claims the Lod sampling functions are only supported in vertex shaders.
2267 * However, they seem to work just fine in fragment shaders as well. */
2268 WARN("Using %sLod in fragment shader.\n", sample_function
.name
);
2269 shader_addline(arg
->buffer
, "%sLod(Psampler%u, %s, %s)%s);\n",
2270 sample_function
.name
, sampler_idx
, coord_param
.param_str
, lod_param
.param_str
, dst_swizzle
);
2272 shader_addline(arg
->buffer
, "%sLod(Vsampler%u, %s, %s)%s);\n",
2273 sample_function
.name
, sampler_idx
, coord_param
.param_str
, lod_param
.param_str
, dst_swizzle
);
2277 static void pshader_glsl_texcoord(const SHADER_OPCODE_ARG
*arg
)
2279 /* FIXME: Make this work for more than just 2D textures */
2280 SHADER_BUFFER
* buffer
= arg
->buffer
;
2284 write_mask
= shader_glsl_append_dst(arg
->buffer
, arg
);
2285 shader_glsl_get_write_mask(write_mask
, dst_mask
);
2287 if (arg
->reg_maps
->shader_version
!= WINED3DPS_VERSION(1,4))
2289 DWORD reg
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2290 shader_addline(buffer
, "clamp(gl_TexCoord[%u], 0.0, 1.0)%s);\n", reg
, dst_mask
);
2292 DWORD reg
= arg
->src
[0] & WINED3DSP_REGNUM_MASK
;
2293 DWORD src_mod
= arg
->src
[0] & WINED3DSP_SRCMOD_MASK
;
2294 char dst_swizzle
[6];
2296 shader_glsl_get_swizzle(arg
->src
[0], FALSE
, write_mask
, dst_swizzle
);
2298 if (src_mod
== WINED3DSPSM_DZ
) {
2299 glsl_src_param_t div_param
;
2300 unsigned int mask_size
= shader_glsl_get_write_mask_size(write_mask
);
2301 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_2
, &div_param
);
2303 if (mask_size
> 1) {
2304 shader_addline(buffer
, "gl_TexCoord[%u]%s / vec%d(%s));\n", reg
, dst_swizzle
, mask_size
, div_param
.param_str
);
2306 shader_addline(buffer
, "gl_TexCoord[%u]%s / %s);\n", reg
, dst_swizzle
, div_param
.param_str
);
2308 } else if (src_mod
== WINED3DSPSM_DW
) {
2309 glsl_src_param_t div_param
;
2310 unsigned int mask_size
= shader_glsl_get_write_mask_size(write_mask
);
2311 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_3
, &div_param
);
2313 if (mask_size
> 1) {
2314 shader_addline(buffer
, "gl_TexCoord[%u]%s / vec%d(%s));\n", reg
, dst_swizzle
, mask_size
, div_param
.param_str
);
2316 shader_addline(buffer
, "gl_TexCoord[%u]%s / %s);\n", reg
, dst_swizzle
, div_param
.param_str
);
2319 shader_addline(buffer
, "gl_TexCoord[%u]%s);\n", reg
, dst_swizzle
);
2324 /** Process the WINED3DSIO_TEXDP3TEX instruction in GLSL:
2325 * Take a 3-component dot product of the TexCoord[dstreg] and src,
2326 * then perform a 1D texture lookup from stage dstregnum, place into dst. */
2327 static void pshader_glsl_texdp3tex(const SHADER_OPCODE_ARG
*arg
)
2329 glsl_src_param_t src0_param
;
2331 glsl_sample_function_t sample_function
;
2332 DWORD sampler_idx
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2333 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2334 DWORD sampler_type
= arg
->reg_maps
->samplers
[sampler_idx
] & WINED3DSP_TEXTURETYPE_MASK
;
2336 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], src_mask
, &src0_param
);
2338 shader_glsl_append_dst(arg
->buffer
, arg
);
2339 shader_glsl_get_write_mask(arg
->dst
, dst_mask
);
2341 /* Do I have to take care about the projected bit? I don't think so, since the dp3 returns only one
2342 * scalar, and projected sampling would require 4.
2344 * It is a dependent read - not valid with conditional NP2 textures
2346 shader_glsl_get_sample_function(sampler_type
, FALSE
, FALSE
, &sample_function
);
2348 switch(count_bits(sample_function
.coord_mask
)) {
2350 shader_addline(arg
->buffer
, "%s(Psampler%u, dot(gl_TexCoord[%u].xyz, %s))%s);\n",
2351 sample_function
.name
, sampler_idx
, sampler_idx
, src0_param
.param_str
, dst_mask
);
2355 shader_addline(arg
->buffer
, "%s(Psampler%u, vec2(dot(gl_TexCoord[%u].xyz, %s), 0.0))%s);\n",
2356 sample_function
.name
, sampler_idx
, sampler_idx
, src0_param
.param_str
, dst_mask
);
2360 shader_addline(arg
->buffer
, "%s(Psampler%u, vec3(dot(gl_TexCoord[%u].xyz, %s), 0.0, 0.0))%s);\n",
2361 sample_function
.name
, sampler_idx
, sampler_idx
, src0_param
.param_str
, dst_mask
);
2364 FIXME("Unexpected mask bitcount %d\n", count_bits(sample_function
.coord_mask
));
2368 /** Process the WINED3DSIO_TEXDP3 instruction in GLSL:
2369 * Take a 3-component dot product of the TexCoord[dstreg] and src. */
2370 static void pshader_glsl_texdp3(const SHADER_OPCODE_ARG
*arg
)
2372 glsl_src_param_t src0_param
;
2373 DWORD dstreg
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2374 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2376 unsigned int mask_size
;
2378 dst_mask
= shader_glsl_append_dst(arg
->buffer
, arg
);
2379 mask_size
= shader_glsl_get_write_mask_size(dst_mask
);
2380 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], src_mask
, &src0_param
);
2382 if (mask_size
> 1) {
2383 shader_addline(arg
->buffer
, "vec%d(dot(T%u.xyz, %s)));\n", mask_size
, dstreg
, src0_param
.param_str
);
2385 shader_addline(arg
->buffer
, "dot(T%u.xyz, %s));\n", dstreg
, src0_param
.param_str
);
2389 /** Process the WINED3DSIO_TEXDEPTH instruction in GLSL:
2390 * Calculate the depth as dst.x / dst.y */
2391 static void pshader_glsl_texdepth(const SHADER_OPCODE_ARG
*arg
)
2393 glsl_dst_param_t dst_param
;
2395 shader_glsl_add_dst_param(arg
, arg
->dst
, 0, &dst_param
);
2397 /* Tests show that texdepth never returns anything below 0.0, and that r5.y is clamped to 1.0.
2398 * Negative input is accepted, -0.25 / -0.5 returns 0.5. GL should clamp gl_FragDepth to [0;1], but
2399 * this doesn't always work, so clamp the results manually. Whether or not the x value is clamped at 1
2400 * too is irrelevant, since if x = 0, any y value < 1.0 (and > 1.0 is not allowed) results in a result
2403 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
);
2406 /** Process the WINED3DSIO_TEXM3X2DEPTH instruction in GLSL:
2407 * Last row of a 3x2 matrix multiply, use the result to calculate the depth:
2408 * Calculate tmp0.y = TexCoord[dstreg] . src.xyz; (tmp0.x has already been calculated)
2409 * depth = (tmp0.y == 0.0) ? 1.0 : tmp0.x / tmp0.y
2411 static void pshader_glsl_texm3x2depth(const SHADER_OPCODE_ARG
*arg
)
2413 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2414 DWORD dstreg
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2415 glsl_src_param_t src0_param
;
2417 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], src_mask
, &src0_param
);
2419 shader_addline(arg
->buffer
, "tmp0.y = dot(T%u.xyz, %s);\n", dstreg
, src0_param
.param_str
);
2420 shader_addline(arg
->buffer
, "gl_FragDepth = (tmp0.y == 0.0) ? 1.0 : clamp(tmp0.x / tmp0.y, 0.0, 1.0);\n");
2423 /** Process the WINED3DSIO_TEXM3X2PAD instruction in GLSL
2424 * Calculate the 1st of a 2-row matrix multiplication. */
2425 static void pshader_glsl_texm3x2pad(const SHADER_OPCODE_ARG
*arg
)
2427 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2428 DWORD reg
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2429 SHADER_BUFFER
* buffer
= arg
->buffer
;
2430 glsl_src_param_t src0_param
;
2432 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], src_mask
, &src0_param
);
2433 shader_addline(buffer
, "tmp0.x = dot(T%u.xyz, %s);\n", reg
, src0_param
.param_str
);
2436 /** Process the WINED3DSIO_TEXM3X3PAD instruction in GLSL
2437 * Calculate the 1st or 2nd row of a 3-row matrix multiplication. */
2438 static void pshader_glsl_texm3x3pad(const SHADER_OPCODE_ARG
* arg
)
2440 IWineD3DPixelShaderImpl
* shader
= (IWineD3DPixelShaderImpl
*) arg
->shader
;
2441 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2442 DWORD reg
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2443 SHADER_BUFFER
* buffer
= arg
->buffer
;
2444 SHADER_PARSE_STATE
* current_state
= &shader
->baseShader
.parse_state
;
2445 glsl_src_param_t src0_param
;
2447 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], src_mask
, &src0_param
);
2448 shader_addline(buffer
, "tmp0.%c = dot(T%u.xyz, %s);\n", 'x' + current_state
->current_row
, reg
, src0_param
.param_str
);
2449 current_state
->texcoord_w
[current_state
->current_row
++] = reg
;
2452 static void pshader_glsl_texm3x2tex(const SHADER_OPCODE_ARG
*arg
)
2454 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2455 DWORD reg
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2456 SHADER_BUFFER
* buffer
= arg
->buffer
;
2457 glsl_src_param_t src0_param
;
2460 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], src_mask
, &src0_param
);
2461 shader_addline(buffer
, "tmp0.y = dot(T%u.xyz, %s);\n", reg
, src0_param
.param_str
);
2463 shader_glsl_append_dst(buffer
, arg
);
2464 shader_glsl_get_write_mask(arg
->dst
, dst_mask
);
2466 /* Sample the texture using the calculated coordinates */
2467 shader_addline(buffer
, "texture2D(Psampler%u, tmp0.xy)%s);\n", reg
, dst_mask
);
2470 /** Process the WINED3DSIO_TEXM3X3TEX instruction in GLSL
2471 * Perform the 3rd row of a 3x3 matrix multiply, then sample the texture using the calculated coordinates */
2472 static void pshader_glsl_texm3x3tex(const SHADER_OPCODE_ARG
*arg
)
2474 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2475 glsl_src_param_t src0_param
;
2477 DWORD reg
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2478 IWineD3DPixelShaderImpl
* This
= (IWineD3DPixelShaderImpl
*) arg
->shader
;
2479 SHADER_PARSE_STATE
* current_state
= &This
->baseShader
.parse_state
;
2480 DWORD sampler_type
= arg
->reg_maps
->samplers
[reg
] & WINED3DSP_TEXTURETYPE_MASK
;
2481 glsl_sample_function_t sample_function
;
2483 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], src_mask
, &src0_param
);
2484 shader_addline(arg
->buffer
, "tmp0.z = dot(T%u.xyz, %s);\n", reg
, src0_param
.param_str
);
2486 shader_glsl_append_dst(arg
->buffer
, arg
);
2487 shader_glsl_get_write_mask(arg
->dst
, dst_mask
);
2488 /* Dependent read, not valid with conditional NP2 */
2489 shader_glsl_get_sample_function(sampler_type
, FALSE
, FALSE
, &sample_function
);
2491 /* Sample the texture using the calculated coordinates */
2492 shader_addline(arg
->buffer
, "%s(Psampler%u, tmp0.xyz)%s);\n", sample_function
.name
, reg
, dst_mask
);
2494 current_state
->current_row
= 0;
2497 /** Process the WINED3DSIO_TEXM3X3 instruction in GLSL
2498 * Perform the 3rd row of a 3x3 matrix multiply */
2499 static void pshader_glsl_texm3x3(const SHADER_OPCODE_ARG
*arg
)
2501 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2502 glsl_src_param_t src0_param
;
2504 DWORD reg
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2505 IWineD3DPixelShaderImpl
* This
= (IWineD3DPixelShaderImpl
*) arg
->shader
;
2506 SHADER_PARSE_STATE
* current_state
= &This
->baseShader
.parse_state
;
2508 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], src_mask
, &src0_param
);
2510 shader_glsl_append_dst(arg
->buffer
, arg
);
2511 shader_glsl_get_write_mask(arg
->dst
, dst_mask
);
2512 shader_addline(arg
->buffer
, "vec4(tmp0.xy, dot(T%u.xyz, %s), 1.0)%s);\n", reg
, src0_param
.param_str
, dst_mask
);
2514 current_state
->current_row
= 0;
2517 /** Process the WINED3DSIO_TEXM3X3SPEC instruction in GLSL
2518 * Perform the final texture lookup based on the previous 2 3x3 matrix multiplies */
2519 static void pshader_glsl_texm3x3spec(const SHADER_OPCODE_ARG
*arg
)
2521 IWineD3DPixelShaderImpl
* shader
= (IWineD3DPixelShaderImpl
*) arg
->shader
;
2522 DWORD reg
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2523 glsl_src_param_t src0_param
;
2524 glsl_src_param_t src1_param
;
2526 SHADER_BUFFER
* buffer
= arg
->buffer
;
2527 SHADER_PARSE_STATE
* current_state
= &shader
->baseShader
.parse_state
;
2528 DWORD stype
= arg
->reg_maps
->samplers
[reg
] & WINED3DSP_TEXTURETYPE_MASK
;
2529 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2530 glsl_sample_function_t sample_function
;
2532 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], src_mask
, &src0_param
);
2533 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], src_mask
, &src1_param
);
2535 /* Perform the last matrix multiply operation */
2536 shader_addline(buffer
, "tmp0.z = dot(T%u.xyz, %s);\n", reg
, src0_param
.param_str
);
2537 /* Reflection calculation */
2538 shader_addline(buffer
, "tmp0.xyz = -reflect((%s), normalize(tmp0.xyz));\n", src1_param
.param_str
);
2540 shader_glsl_append_dst(buffer
, arg
);
2541 shader_glsl_get_write_mask(arg
->dst
, dst_mask
);
2542 /* Dependent read, not valid with conditional NP2 */
2543 shader_glsl_get_sample_function(stype
, FALSE
, FALSE
, &sample_function
);
2545 /* Sample the texture */
2546 shader_addline(buffer
, "%s(Psampler%u, tmp0.xyz)%s);\n", sample_function
.name
, reg
, dst_mask
);
2548 current_state
->current_row
= 0;
2551 /** Process the WINED3DSIO_TEXM3X3VSPEC instruction in GLSL
2552 * Perform the final texture lookup based on the previous 2 3x3 matrix multiplies */
2553 static void pshader_glsl_texm3x3vspec(const SHADER_OPCODE_ARG
*arg
)
2555 IWineD3DPixelShaderImpl
* shader
= (IWineD3DPixelShaderImpl
*) arg
->shader
;
2556 DWORD reg
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2557 SHADER_BUFFER
* buffer
= arg
->buffer
;
2558 SHADER_PARSE_STATE
* current_state
= &shader
->baseShader
.parse_state
;
2559 glsl_src_param_t src0_param
;
2561 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2562 DWORD sampler_type
= arg
->reg_maps
->samplers
[reg
] & WINED3DSP_TEXTURETYPE_MASK
;
2563 glsl_sample_function_t sample_function
;
2565 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], src_mask
, &src0_param
);
2567 /* Perform the last matrix multiply operation */
2568 shader_addline(buffer
, "tmp0.z = dot(vec3(T%u), vec3(%s));\n", reg
, src0_param
.param_str
);
2570 /* Construct the eye-ray vector from w coordinates */
2571 shader_addline(buffer
, "tmp1.xyz = normalize(vec3(gl_TexCoord[%u].w, gl_TexCoord[%u].w, gl_TexCoord[%u].w));\n",
2572 current_state
->texcoord_w
[0], current_state
->texcoord_w
[1], reg
);
2573 shader_addline(buffer
, "tmp0.xyz = -reflect(tmp1.xyz, normalize(tmp0.xyz));\n");
2575 shader_glsl_append_dst(buffer
, arg
);
2576 shader_glsl_get_write_mask(arg
->dst
, dst_mask
);
2577 /* Dependent read, not valid with conditional NP2 */
2578 shader_glsl_get_sample_function(sampler_type
, FALSE
, FALSE
, &sample_function
);
2580 /* Sample the texture using the calculated coordinates */
2581 shader_addline(buffer
, "%s(Psampler%u, tmp0.xyz)%s);\n", sample_function
.name
, reg
, dst_mask
);
2583 current_state
->current_row
= 0;
2586 /** Process the WINED3DSIO_TEXBEM instruction in GLSL.
2587 * Apply a fake bump map transform.
2588 * texbem is pshader <= 1.3 only, this saves a few version checks
2590 static void pshader_glsl_texbem(const SHADER_OPCODE_ARG
*arg
)
2592 IWineD3DPixelShaderImpl
* This
= (IWineD3DPixelShaderImpl
*) arg
->shader
;
2593 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
2594 char dst_swizzle
[6];
2595 glsl_sample_function_t sample_function
;
2596 glsl_src_param_t coord_param
;
2603 sampler_idx
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2604 flags
= deviceImpl
->stateBlock
->textureState
[sampler_idx
][WINED3DTSS_TEXTURETRANSFORMFLAGS
];
2606 sampler_type
= arg
->reg_maps
->samplers
[sampler_idx
] & WINED3DSP_TEXTURETYPE_MASK
;
2607 /* Dependent read, not valid with conditional NP2 */
2608 shader_glsl_get_sample_function(sampler_type
, FALSE
, FALSE
, &sample_function
);
2609 mask
= sample_function
.coord_mask
;
2611 shader_glsl_get_write_mask(arg
->dst
, dst_swizzle
);
2613 shader_glsl_get_write_mask(mask
, coord_mask
);
2615 /* with projective textures, texbem only divides the static texture coord, not the displacement,
2616 * so we can't let the GL handle this.
2618 if (flags
& WINED3DTTFF_PROJECTED
) {
2620 char coord_div_mask
[3];
2621 switch (flags
& ~WINED3DTTFF_PROJECTED
) {
2622 case WINED3DTTFF_COUNT1
: FIXME("WINED3DTTFF_PROJECTED with WINED3DTTFF_COUNT1?\n"); break;
2623 case WINED3DTTFF_COUNT2
: div_mask
= WINED3DSP_WRITEMASK_1
; break;
2624 case WINED3DTTFF_COUNT3
: div_mask
= WINED3DSP_WRITEMASK_2
; break;
2625 case WINED3DTTFF_COUNT4
:
2626 case WINED3DTTFF_DISABLE
: div_mask
= WINED3DSP_WRITEMASK_3
; break;
2628 shader_glsl_get_write_mask(div_mask
, coord_div_mask
);
2629 shader_addline(arg
->buffer
, "T%u%s /= T%u%s;\n", sampler_idx
, coord_mask
, sampler_idx
, coord_div_mask
);
2632 shader_glsl_append_dst(arg
->buffer
, arg
);
2633 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_0
|WINED3DSP_WRITEMASK_1
, &coord_param
);
2634 if(arg
->opcode
->opcode
== WINED3DSIO_TEXBEML
) {
2635 glsl_src_param_t luminance_param
;
2636 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_2
, &luminance_param
);
2637 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",
2638 sample_function
.name
, sampler_idx
, sampler_idx
, coord_mask
, sampler_idx
, coord_param
.param_str
, coord_mask
,
2639 luminance_param
.param_str
, sampler_idx
, sampler_idx
, dst_swizzle
);
2641 shader_addline(arg
->buffer
, "%s(Psampler%u, T%u%s + vec4(bumpenvmat%d * %s, 0.0, 0.0)%s )%s);\n",
2642 sample_function
.name
, sampler_idx
, sampler_idx
, coord_mask
, sampler_idx
, coord_param
.param_str
, coord_mask
, dst_swizzle
);
2646 static void pshader_glsl_bem(const SHADER_OPCODE_ARG
*arg
)
2648 glsl_src_param_t src0_param
, src1_param
;
2649 DWORD sampler_idx
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2651 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_0
|WINED3DSP_WRITEMASK_1
, &src0_param
);
2652 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], WINED3DSP_WRITEMASK_0
|WINED3DSP_WRITEMASK_1
, &src1_param
);
2654 shader_glsl_append_dst(arg
->buffer
, arg
);
2655 shader_addline(arg
->buffer
, "%s + bumpenvmat%d * %s);\n",
2656 src0_param
.param_str
, sampler_idx
, src1_param
.param_str
);
2659 /** Process the WINED3DSIO_TEXREG2AR instruction in GLSL
2660 * Sample 2D texture at dst using the alpha & red (wx) components of src as texture coordinates */
2661 static void pshader_glsl_texreg2ar(const SHADER_OPCODE_ARG
*arg
)
2663 glsl_src_param_t src0_param
;
2664 DWORD sampler_idx
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2667 shader_glsl_append_dst(arg
->buffer
, arg
);
2668 shader_glsl_get_write_mask(arg
->dst
, dst_mask
);
2669 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_ALL
, &src0_param
);
2671 shader_addline(arg
->buffer
, "texture2D(Psampler%u, %s.wx)%s);\n", sampler_idx
, src0_param
.reg_name
, dst_mask
);
2674 /** Process the WINED3DSIO_TEXREG2GB instruction in GLSL
2675 * Sample 2D texture at dst using the green & blue (yz) components of src as texture coordinates */
2676 static void pshader_glsl_texreg2gb(const SHADER_OPCODE_ARG
*arg
)
2678 glsl_src_param_t src0_param
;
2679 DWORD sampler_idx
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2682 shader_glsl_append_dst(arg
->buffer
, arg
);
2683 shader_glsl_get_write_mask(arg
->dst
, dst_mask
);
2684 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_ALL
, &src0_param
);
2686 shader_addline(arg
->buffer
, "texture2D(Psampler%u, %s.yz)%s);\n", sampler_idx
, src0_param
.reg_name
, dst_mask
);
2689 /** Process the WINED3DSIO_TEXREG2RGB instruction in GLSL
2690 * Sample texture at dst using the rgb (xyz) components of src as texture coordinates */
2691 static void pshader_glsl_texreg2rgb(const SHADER_OPCODE_ARG
*arg
)
2693 glsl_src_param_t src0_param
;
2695 DWORD sampler_idx
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2696 DWORD sampler_type
= arg
->reg_maps
->samplers
[sampler_idx
] & WINED3DSP_TEXTURETYPE_MASK
;
2697 glsl_sample_function_t sample_function
;
2699 shader_glsl_append_dst(arg
->buffer
, arg
);
2700 shader_glsl_get_write_mask(arg
->dst
, dst_mask
);
2701 /* Dependent read, not valid with conditional NP2 */
2702 shader_glsl_get_sample_function(sampler_type
, FALSE
, FALSE
, &sample_function
);
2703 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], sample_function
.coord_mask
, &src0_param
);
2705 shader_addline(arg
->buffer
, "%s(Psampler%u, %s)%s);\n", sample_function
.name
, sampler_idx
, src0_param
.param_str
, dst_mask
);
2708 /** Process the WINED3DSIO_TEXKILL instruction in GLSL.
2709 * If any of the first 3 components are < 0, discard this pixel */
2710 static void pshader_glsl_texkill(const SHADER_OPCODE_ARG
*arg
)
2712 glsl_dst_param_t dst_param
;
2714 /* The argument is a destination parameter, and no writemasks are allowed */
2715 shader_glsl_add_dst_param(arg
, arg
->dst
, 0, &dst_param
);
2716 if ((arg
->reg_maps
->shader_version
>= WINED3DPS_VERSION(2,0)))
2718 /* 2.0 shaders compare all 4 components in texkill */
2719 shader_addline(arg
->buffer
, "if (any(lessThan(%s.xyzw, vec4(0.0)))) discard;\n", dst_param
.reg_name
);
2721 /* 1.X shaders only compare the first 3 components, probably due to the nature of the texkill
2722 * instruction as a tex* instruction, and phase, which kills all a / w components. Even if all
2723 * 4 components are defined, only the first 3 are used
2725 shader_addline(arg
->buffer
, "if (any(lessThan(%s.xyz, vec3(0.0)))) discard;\n", dst_param
.reg_name
);
2729 /** Process the WINED3DSIO_DP2ADD instruction in GLSL.
2730 * dst = dot2(src0, src1) + src2 */
2731 static void pshader_glsl_dp2add(const SHADER_OPCODE_ARG
*arg
)
2733 glsl_src_param_t src0_param
;
2734 glsl_src_param_t src1_param
;
2735 glsl_src_param_t src2_param
;
2737 unsigned int mask_size
;
2739 write_mask
= shader_glsl_append_dst(arg
->buffer
, arg
);
2740 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
2742 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
, &src0_param
);
2743 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
, &src1_param
);
2744 shader_glsl_add_src_param(arg
, arg
->src
[2], arg
->src_addr
[2], WINED3DSP_WRITEMASK_0
, &src2_param
);
2746 if (mask_size
> 1) {
2747 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
);
2749 shader_addline(arg
->buffer
, "dot(%s, %s) + %s);\n", src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
2753 static void pshader_glsl_input_pack(SHADER_BUFFER
* buffer
, const struct semantic
* semantics_in
,
2754 IWineD3DPixelShader
*iface
, enum vertexprocessing_mode vertexprocessing
)
2757 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*) iface
;
2759 for (i
= 0; i
< MAX_REG_INPUT
; i
++) {
2761 DWORD usage_token
= semantics_in
[i
].usage
;
2762 DWORD register_token
= semantics_in
[i
].reg
;
2763 DWORD usage
, usage_idx
;
2767 if (!usage_token
) continue;
2768 usage
= (usage_token
& WINED3DSP_DCL_USAGE_MASK
) >> WINED3DSP_DCL_USAGE_SHIFT
;
2769 usage_idx
= (usage_token
& WINED3DSP_DCL_USAGEINDEX_MASK
) >> WINED3DSP_DCL_USAGEINDEX_SHIFT
;
2770 shader_glsl_get_write_mask(register_token
, reg_mask
);
2774 case WINED3DDECLUSAGE_TEXCOORD
:
2775 if(usage_idx
< 8 && vertexprocessing
== pretransformed
) {
2776 shader_addline(buffer
, "IN[%u]%s = gl_TexCoord[%u]%s;\n",
2777 This
->input_reg_map
[i
], reg_mask
, usage_idx
, reg_mask
);
2779 shader_addline(buffer
, "IN[%u]%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
2780 This
->input_reg_map
[i
], reg_mask
, reg_mask
);
2784 case WINED3DDECLUSAGE_COLOR
:
2786 shader_addline(buffer
, "IN[%u]%s = vec4(gl_Color)%s;\n",
2787 This
->input_reg_map
[i
], reg_mask
, reg_mask
);
2788 else if (usage_idx
== 1)
2789 shader_addline(buffer
, "IN[%u]%s = vec4(gl_SecondaryColor)%s;\n",
2790 This
->input_reg_map
[i
], reg_mask
, reg_mask
);
2792 shader_addline(buffer
, "IN[%u]%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
2793 This
->input_reg_map
[i
], reg_mask
, reg_mask
);
2797 shader_addline(buffer
, "IN[%u]%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
2798 This
->input_reg_map
[i
], reg_mask
, reg_mask
);
2803 /*********************************************
2804 * Vertex Shader Specific Code begins here
2805 ********************************************/
2807 static void add_glsl_program_entry(struct shader_glsl_priv
*priv
, struct glsl_shader_prog_link
*entry
) {
2808 glsl_program_key_t
*key
;
2810 key
= HeapAlloc(GetProcessHeap(), 0, sizeof(glsl_program_key_t
));
2811 key
->vshader
= entry
->vshader
;
2812 key
->pshader
= entry
->pshader
;
2813 key
->ps_args
= entry
->ps_args
;
2815 hash_table_put(priv
->glsl_program_lookup
, key
, entry
);
2818 static struct glsl_shader_prog_link
*get_glsl_program_entry(struct shader_glsl_priv
*priv
,
2819 GLhandleARB vshader
, IWineD3DPixelShader
*pshader
, struct ps_compile_args
*ps_args
) {
2820 glsl_program_key_t key
;
2822 key
.vshader
= vshader
;
2823 key
.pshader
= pshader
;
2824 key
.ps_args
= *ps_args
;
2826 return (struct glsl_shader_prog_link
*)hash_table_get(priv
->glsl_program_lookup
, &key
);
2829 static void delete_glsl_program_entry(struct shader_glsl_priv
*priv
, const WineD3D_GL_Info
*gl_info
,
2830 struct glsl_shader_prog_link
*entry
)
2832 glsl_program_key_t
*key
;
2834 key
= HeapAlloc(GetProcessHeap(), 0, sizeof(glsl_program_key_t
));
2835 key
->vshader
= entry
->vshader
;
2836 key
->pshader
= entry
->pshader
;
2837 key
->ps_args
= entry
->ps_args
;
2838 hash_table_remove(priv
->glsl_program_lookup
, key
);
2840 GL_EXTCALL(glDeleteObjectARB(entry
->programId
));
2841 if (entry
->vshader
) list_remove(&entry
->vshader_entry
);
2842 if (entry
->pshader
) list_remove(&entry
->pshader_entry
);
2843 HeapFree(GetProcessHeap(), 0, entry
->vuniformF_locations
);
2844 HeapFree(GetProcessHeap(), 0, entry
->puniformF_locations
);
2845 HeapFree(GetProcessHeap(), 0, entry
);
2848 static void handle_ps3_input(SHADER_BUFFER
*buffer
, const struct semantic
*semantics_in
,
2849 const struct semantic
*semantics_out
, const WineD3D_GL_Info
*gl_info
, const DWORD
*map
)
2852 DWORD usage_token
, usage_token_out
;
2853 DWORD register_token
, register_token_out
;
2854 DWORD usage
, usage_idx
, usage_out
, usage_idx_out
;
2857 DWORD in_count
= GL_LIMITS(glsl_varyings
) / 4;
2858 char reg_mask
[6], reg_mask_out
[6];
2859 char destination
[50];
2861 set
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*set
) * (in_count
+ 2));
2863 if (!semantics_out
) {
2864 /* Save gl_FrontColor & gl_FrontSecondaryColor before overwriting them. */
2865 shader_addline(buffer
, "vec4 front_color = gl_FrontColor;\n");
2866 shader_addline(buffer
, "vec4 front_secondary_color = gl_FrontSecondaryColor;\n");
2869 for(i
= 0; i
< MAX_REG_INPUT
; i
++) {
2870 usage_token
= semantics_in
[i
].usage
;
2871 if (!usage_token
) continue;
2874 if (in_idx
>= (in_count
+ 2)) {
2875 FIXME("More input varyings declared than supported, expect issues\n");
2877 } else if(map
[i
] == -1) {
2878 /* Declared, but not read register */
2882 if (in_idx
== in_count
) {
2883 sprintf(destination
, "gl_FrontColor");
2884 } else if (in_idx
== in_count
+ 1) {
2885 sprintf(destination
, "gl_FrontSecondaryColor");
2887 sprintf(destination
, "IN[%u]", in_idx
);
2890 register_token
= semantics_in
[i
].reg
;
2892 usage
= (usage_token
& WINED3DSP_DCL_USAGE_MASK
) >> WINED3DSP_DCL_USAGE_SHIFT
;
2893 usage_idx
= (usage_token
& WINED3DSP_DCL_USAGEINDEX_MASK
) >> WINED3DSP_DCL_USAGEINDEX_SHIFT
;
2894 set
[map
[i
]] = shader_glsl_get_write_mask(register_token
, reg_mask
);
2896 if(!semantics_out
) {
2898 case WINED3DDECLUSAGE_COLOR
:
2900 shader_addline(buffer
, "%s%s = front_color%s;\n",
2901 destination
, reg_mask
, reg_mask
);
2902 else if (usage_idx
== 1)
2903 shader_addline(buffer
, "%s%s = front_secondary_color%s;\n",
2904 destination
, reg_mask
, reg_mask
);
2906 shader_addline(buffer
, "%s%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
2907 destination
, reg_mask
, reg_mask
);
2910 case WINED3DDECLUSAGE_TEXCOORD
:
2911 if (usage_idx
< 8) {
2912 shader_addline(buffer
, "%s%s = gl_TexCoord[%u]%s;\n",
2913 destination
, reg_mask
, usage_idx
, reg_mask
);
2915 shader_addline(buffer
, "%s%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
2916 destination
, reg_mask
, reg_mask
);
2920 case WINED3DDECLUSAGE_FOG
:
2921 shader_addline(buffer
, "%s%s = vec4(gl_FogFragCoord, 0.0, 0.0, 0.0)%s;\n",
2922 destination
, reg_mask
, reg_mask
);
2926 shader_addline(buffer
, "%s%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
2927 destination
, reg_mask
, reg_mask
);
2931 for(j
= 0; j
< MAX_REG_OUTPUT
; j
++) {
2932 usage_token_out
= semantics_out
[j
].usage
;
2933 if (!usage_token_out
) continue;
2934 register_token_out
= semantics_out
[j
].reg
;
2936 usage_out
= (usage_token_out
& WINED3DSP_DCL_USAGE_MASK
) >> WINED3DSP_DCL_USAGE_SHIFT
;
2937 usage_idx_out
= (usage_token_out
& WINED3DSP_DCL_USAGEINDEX_MASK
) >> WINED3DSP_DCL_USAGEINDEX_SHIFT
;
2938 shader_glsl_get_write_mask(register_token_out
, reg_mask_out
);
2940 if(usage
== usage_out
&&
2941 usage_idx
== usage_idx_out
) {
2942 shader_addline(buffer
, "%s%s = OUT[%u]%s;\n",
2943 destination
, reg_mask
, j
, reg_mask
);
2948 shader_addline(buffer
, "%s%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
2949 destination
, reg_mask
, reg_mask
);
2954 /* This is solely to make the compiler / linker happy and avoid warning about undefined
2955 * varyings. It shouldn't result in any real code executed on the GPU, since all read
2956 * input varyings are assigned above, if the optimizer works properly.
2958 for(i
= 0; i
< in_count
+ 2; i
++) {
2959 if(set
[i
] != WINED3DSP_WRITEMASK_ALL
) {
2960 unsigned int size
= 0;
2961 memset(reg_mask
, 0, sizeof(reg_mask
));
2962 if(!(set
[i
] & WINED3DSP_WRITEMASK_0
)) {
2963 reg_mask
[size
] = 'x';
2966 if(!(set
[i
] & WINED3DSP_WRITEMASK_1
)) {
2967 reg_mask
[size
] = 'y';
2970 if(!(set
[i
] & WINED3DSP_WRITEMASK_2
)) {
2971 reg_mask
[size
] = 'z';
2974 if(!(set
[i
] & WINED3DSP_WRITEMASK_3
)) {
2975 reg_mask
[size
] = 'w';
2979 if (i
== in_count
) {
2980 sprintf(destination
, "gl_FrontColor");
2981 } else if (i
== in_count
+ 1) {
2982 sprintf(destination
, "gl_FrontSecondaryColor");
2984 sprintf(destination
, "IN[%u]", i
);
2988 shader_addline(buffer
, "%s.%s = 0.0;\n", destination
, reg_mask
);
2990 shader_addline(buffer
, "%s.%s = vec%u(0.0);\n", destination
, reg_mask
, size
);
2995 HeapFree(GetProcessHeap(), 0, set
);
2998 static GLhandleARB
generate_param_reorder_function(IWineD3DVertexShader
*vertexshader
,
2999 IWineD3DPixelShader
*pixelshader
, const WineD3D_GL_Info
*gl_info
)
3001 GLhandleARB ret
= 0;
3002 IWineD3DVertexShaderImpl
*vs
= (IWineD3DVertexShaderImpl
*) vertexshader
;
3003 IWineD3DPixelShaderImpl
*ps
= (IWineD3DPixelShaderImpl
*) pixelshader
;
3004 IWineD3DDeviceImpl
*device
;
3005 DWORD vs_major
= WINED3DSHADER_VERSION_MAJOR(vs
->baseShader
.reg_maps
.shader_version
);
3006 DWORD ps_major
= ps
? WINED3DSHADER_VERSION_MAJOR(ps
->baseShader
.reg_maps
.shader_version
) : 0;
3008 SHADER_BUFFER buffer
;
3010 DWORD register_token
;
3011 DWORD usage
, usage_idx
, writemask
;
3013 const struct semantic
*semantics_out
, *semantics_in
;
3015 shader_buffer_init(&buffer
);
3017 shader_addline(&buffer
, "#version 120\n");
3019 if(vs_major
< 3 && ps_major
< 3) {
3020 /* That one is easy: The vertex shader writes to the builtin varyings, the pixel shader reads from them.
3021 * Take care about the texcoord .w fixup though if we're using the fixed function fragment pipeline
3023 device
= (IWineD3DDeviceImpl
*) vs
->baseShader
.device
;
3024 if((GLINFO_LOCATION
).set_texcoord_w
&& ps_major
== 0 && vs_major
> 0 &&
3025 !device
->frag_pipe
->ffp_proj_control
) {
3026 shader_addline(&buffer
, "void order_ps_input() {\n");
3027 for(i
= 0; i
< min(8, MAX_REG_TEXCRD
); i
++) {
3028 if(vs
->baseShader
.reg_maps
.texcoord_mask
[i
] != 0 &&
3029 vs
->baseShader
.reg_maps
.texcoord_mask
[i
] != WINED3DSP_WRITEMASK_ALL
) {
3030 shader_addline(&buffer
, "gl_TexCoord[%u].w = 1.0;\n", i
);
3033 shader_addline(&buffer
, "}\n");
3035 shader_addline(&buffer
, "void order_ps_input() { /* do nothing */ }\n");
3037 } else if(ps_major
< 3 && vs_major
>= 3) {
3038 /* The vertex shader writes to its own varyings, the pixel shader needs them in the builtin ones */
3039 semantics_out
= vs
->semantics_out
;
3041 shader_addline(&buffer
, "void order_ps_input(in vec4 OUT[%u]) {\n", MAX_REG_OUTPUT
);
3042 for(i
= 0; i
< MAX_REG_OUTPUT
; i
++) {
3043 usage_token
= semantics_out
[i
].usage
;
3044 if (!usage_token
) continue;
3045 register_token
= semantics_out
[i
].reg
;
3047 usage
= (usage_token
& WINED3DSP_DCL_USAGE_MASK
) >> WINED3DSP_DCL_USAGE_SHIFT
;
3048 usage_idx
= (usage_token
& WINED3DSP_DCL_USAGEINDEX_MASK
) >> WINED3DSP_DCL_USAGEINDEX_SHIFT
;
3049 writemask
= shader_glsl_get_write_mask(register_token
, reg_mask
);
3052 case WINED3DDECLUSAGE_COLOR
:
3054 shader_addline(&buffer
, "gl_FrontColor%s = OUT[%u]%s;\n", reg_mask
, i
, reg_mask
);
3055 else if (usage_idx
== 1)
3056 shader_addline(&buffer
, "gl_FrontSecondaryColor%s = OUT[%u]%s;\n", reg_mask
, i
, reg_mask
);
3059 case WINED3DDECLUSAGE_POSITION
:
3060 shader_addline(&buffer
, "gl_Position%s = OUT[%u]%s;\n", reg_mask
, i
, reg_mask
);
3063 case WINED3DDECLUSAGE_TEXCOORD
:
3064 if (usage_idx
< 8) {
3065 if(!(GLINFO_LOCATION
).set_texcoord_w
|| ps_major
> 0) writemask
|= WINED3DSP_WRITEMASK_3
;
3067 shader_addline(&buffer
, "gl_TexCoord[%u]%s = OUT[%u]%s;\n",
3068 usage_idx
, reg_mask
, i
, reg_mask
);
3069 if(!(writemask
& WINED3DSP_WRITEMASK_3
)) {
3070 shader_addline(&buffer
, "gl_TexCoord[%u].w = 1.0;\n", usage_idx
);
3075 case WINED3DDECLUSAGE_PSIZE
:
3076 shader_addline(&buffer
, "gl_PointSize = OUT[%u].x;\n", i
);
3079 case WINED3DDECLUSAGE_FOG
:
3080 shader_addline(&buffer
, "gl_FogFragCoord = OUT[%u].%c;\n", i
, reg_mask
[1]);
3087 shader_addline(&buffer
, "}\n");
3089 } else if(ps_major
>= 3 && vs_major
>= 3) {
3090 semantics_out
= vs
->semantics_out
;
3091 semantics_in
= ps
->semantics_in
;
3093 /* This one is tricky: a 3.0 pixel shader reads from a 3.0 vertex shader */
3094 shader_addline(&buffer
, "varying vec4 IN[%u];\n", GL_LIMITS(glsl_varyings
) / 4);
3095 shader_addline(&buffer
, "void order_ps_input(in vec4 OUT[%u]) {\n", MAX_REG_OUTPUT
);
3097 /* First, sort out position and point size. Those are not passed to the pixel shader */
3098 for(i
= 0; i
< MAX_REG_OUTPUT
; i
++) {
3099 usage_token
= semantics_out
[i
].usage
;
3100 if (!usage_token
) continue;
3101 register_token
= semantics_out
[i
].reg
;
3103 usage
= (usage_token
& WINED3DSP_DCL_USAGE_MASK
) >> WINED3DSP_DCL_USAGE_SHIFT
;
3104 usage_idx
= (usage_token
& WINED3DSP_DCL_USAGEINDEX_MASK
) >> WINED3DSP_DCL_USAGEINDEX_SHIFT
;
3105 shader_glsl_get_write_mask(register_token
, reg_mask
);
3108 case WINED3DDECLUSAGE_POSITION
:
3109 shader_addline(&buffer
, "gl_Position%s = OUT[%u]%s;\n", reg_mask
, i
, reg_mask
);
3112 case WINED3DDECLUSAGE_PSIZE
:
3113 shader_addline(&buffer
, "gl_PointSize = OUT[%u].x;\n", i
);
3121 /* Then, fix the pixel shader input */
3122 handle_ps3_input(&buffer
, semantics_in
, semantics_out
, gl_info
, ps
->input_reg_map
);
3124 shader_addline(&buffer
, "}\n");
3125 } else if(ps_major
>= 3 && vs_major
< 3) {
3126 semantics_in
= ps
->semantics_in
;
3128 shader_addline(&buffer
, "varying vec4 IN[%u];\n", GL_LIMITS(glsl_varyings
) / 4);
3129 shader_addline(&buffer
, "void order_ps_input() {\n");
3130 /* The vertex shader wrote to the builtin varyings. There is no need to figure out position and
3131 * point size, but we depend on the optimizers kindness to find out that the pixel shader doesn't
3132 * read gl_TexCoord and gl_ColorX, otherwise we'll run out of varyings
3134 handle_ps3_input(&buffer
, semantics_in
, NULL
, gl_info
, ps
->input_reg_map
);
3135 shader_addline(&buffer
, "}\n");
3137 ERR("Unexpected vertex and pixel shader version condition: vs: %d, ps: %d\n", vs_major
, ps_major
);
3140 ret
= GL_EXTCALL(glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB
));
3141 checkGLcall("glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB)");
3142 GL_EXTCALL(glShaderSourceARB(ret
, 1, (const char**)&buffer
.buffer
, NULL
));
3143 checkGLcall("glShaderSourceARB(ret, 1, (const char**)&buffer.buffer, NULL)");
3144 GL_EXTCALL(glCompileShaderARB(ret
));
3145 checkGLcall("glCompileShaderARB(ret)");
3147 shader_buffer_free(&buffer
);
3151 static void hardcode_local_constants(IWineD3DBaseShaderImpl
*shader
, const WineD3D_GL_Info
*gl_info
,
3152 GLhandleARB programId
, char prefix
)
3154 const local_constant
*lconst
;
3159 LIST_FOR_EACH_ENTRY(lconst
, &shader
->baseShader
.constantsF
, local_constant
, entry
) {
3160 value
= (const float *)lconst
->value
;
3161 snprintf(glsl_name
, sizeof(glsl_name
), "%cLC%u", prefix
, lconst
->idx
);
3162 tmp_loc
= GL_EXTCALL(glGetUniformLocationARB(programId
, glsl_name
));
3163 GL_EXTCALL(glUniform4fvARB(tmp_loc
, 1, value
));
3165 checkGLcall("Hardcoding local constants\n");
3168 /** Sets the GLSL program ID for the given pixel and vertex shader combination.
3169 * It sets the programId on the current StateBlock (because it should be called
3170 * inside of the DrawPrimitive() part of the render loop).
3172 * If a program for the given combination does not exist, create one, and store
3173 * the program in the hash table. If it creates a program, it will link the
3174 * given objects, too.
3176 static void set_glsl_shader_program(IWineD3DDevice
*iface
, BOOL use_ps
, BOOL use_vs
) {
3177 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
3178 struct shader_glsl_priv
*priv
= (struct shader_glsl_priv
*)This
->shader_priv
;
3179 const WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
3180 IWineD3DPixelShader
*pshader
= This
->stateBlock
->pixelShader
;
3181 IWineD3DVertexShader
*vshader
= This
->stateBlock
->vertexShader
;
3182 struct glsl_shader_prog_link
*entry
= NULL
;
3183 GLhandleARB programId
= 0;
3184 GLhandleARB reorder_shader_id
= 0;
3187 GLhandleARB vshader_id
, pshader_id
;
3188 struct ps_compile_args compile_args
;
3191 IWineD3DVertexShaderImpl_CompileShader(vshader
);
3192 vshader_id
= ((IWineD3DVertexShaderImpl
*)vshader
)->prgId
;
3197 find_ps_compile_args((IWineD3DPixelShaderImpl
*)This
->stateBlock
->pixelShader
, This
->stateBlock
, &compile_args
);
3199 /* FIXME: Do we really have to spend CPU cycles to generate a few zeroed bytes? */
3200 memset(&compile_args
, 0, sizeof(compile_args
));
3202 entry
= get_glsl_program_entry(priv
, vshader_id
, pshader
, &compile_args
);
3204 priv
->glsl_program
= entry
;
3208 /* If we get to this point, then no matching program exists, so we create one */
3209 programId
= GL_EXTCALL(glCreateProgramObjectARB());
3210 TRACE("Created new GLSL shader program %u\n", programId
);
3212 /* Create the entry */
3213 entry
= HeapAlloc(GetProcessHeap(), 0, sizeof(struct glsl_shader_prog_link
));
3214 entry
->programId
= programId
;
3215 entry
->vshader
= vshader_id
;
3216 entry
->pshader
= pshader
;
3217 entry
->ps_args
= compile_args
;
3218 /* Add the hash table entry */
3219 add_glsl_program_entry(priv
, entry
);
3221 /* Set the current program */
3222 priv
->glsl_program
= entry
;
3224 /* Attach GLSL vshader */
3226 int max_attribs
= 16; /* TODO: Will this always be the case? It is at the moment... */
3229 reorder_shader_id
= generate_param_reorder_function(vshader
, pshader
, gl_info
);
3230 TRACE("Attaching GLSL shader object %u to program %u\n", reorder_shader_id
, programId
);
3231 GL_EXTCALL(glAttachObjectARB(programId
, reorder_shader_id
));
3232 checkGLcall("glAttachObjectARB");
3233 /* Flag the reorder function for deletion, then it will be freed automatically when the program
3236 GL_EXTCALL(glDeleteObjectARB(reorder_shader_id
));
3238 TRACE("Attaching GLSL shader object %u to program %u\n", vshader_id
, programId
);
3239 GL_EXTCALL(glAttachObjectARB(programId
, vshader_id
));
3240 checkGLcall("glAttachObjectARB");
3242 /* Bind vertex attributes to a corresponding index number to match
3243 * the same index numbers as ARB_vertex_programs (makes loading
3244 * vertex attributes simpler). With this method, we can use the
3245 * exact same code to load the attributes later for both ARB and
3248 * We have to do this here because we need to know the Program ID
3249 * in order to make the bindings work, and it has to be done prior
3250 * to linking the GLSL program. */
3251 for (i
= 0; i
< max_attribs
; ++i
) {
3252 if (((IWineD3DBaseShaderImpl
*)vshader
)->baseShader
.reg_maps
.attributes
[i
]) {
3253 snprintf(tmp_name
, sizeof(tmp_name
), "attrib%i", i
);
3254 GL_EXTCALL(glBindAttribLocationARB(programId
, i
, tmp_name
));
3257 checkGLcall("glBindAttribLocationARB");
3259 list_add_head(&((IWineD3DBaseShaderImpl
*)vshader
)->baseShader
.linked_programs
, &entry
->vshader_entry
);
3263 pshader_id
= find_gl_pshader((IWineD3DPixelShaderImpl
*) pshader
, &compile_args
);
3268 /* Attach GLSL pshader */
3270 TRACE("Attaching GLSL shader object %u to program %u\n", pshader_id
, programId
);
3271 GL_EXTCALL(glAttachObjectARB(programId
, pshader_id
));
3272 checkGLcall("glAttachObjectARB");
3274 list_add_head(&((IWineD3DBaseShaderImpl
*)pshader
)->baseShader
.linked_programs
, &entry
->pshader_entry
);
3277 /* Link the program */
3278 TRACE("Linking GLSL shader program %u\n", programId
);
3279 GL_EXTCALL(glLinkProgramARB(programId
));
3280 print_glsl_info_log(&GLINFO_LOCATION
, programId
);
3282 entry
->vuniformF_locations
= HeapAlloc(GetProcessHeap(), 0, sizeof(GLhandleARB
) * GL_LIMITS(vshader_constantsF
));
3283 for (i
= 0; i
< GL_LIMITS(vshader_constantsF
); ++i
) {
3284 snprintf(glsl_name
, sizeof(glsl_name
), "VC[%i]", i
);
3285 entry
->vuniformF_locations
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, glsl_name
));
3287 for (i
= 0; i
< MAX_CONST_I
; ++i
) {
3288 snprintf(glsl_name
, sizeof(glsl_name
), "VI[%i]", i
);
3289 entry
->vuniformI_locations
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, glsl_name
));
3291 entry
->puniformF_locations
= HeapAlloc(GetProcessHeap(), 0, sizeof(GLhandleARB
) * GL_LIMITS(pshader_constantsF
));
3292 for (i
= 0; i
< GL_LIMITS(pshader_constantsF
); ++i
) {
3293 snprintf(glsl_name
, sizeof(glsl_name
), "PC[%i]", i
);
3294 entry
->puniformF_locations
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, glsl_name
));
3296 for (i
= 0; i
< MAX_CONST_I
; ++i
) {
3297 snprintf(glsl_name
, sizeof(glsl_name
), "PI[%i]", i
);
3298 entry
->puniformI_locations
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, glsl_name
));
3302 for(i
= 0; i
< ((IWineD3DPixelShaderImpl
*)pshader
)->numbumpenvmatconsts
; i
++) {
3304 sprintf(name
, "bumpenvmat%d", ((IWineD3DPixelShaderImpl
*)pshader
)->bumpenvmatconst
[i
].texunit
);
3305 entry
->bumpenvmat_location
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, name
));
3306 sprintf(name
, "luminancescale%d", ((IWineD3DPixelShaderImpl
*)pshader
)->luminanceconst
[i
].texunit
);
3307 entry
->luminancescale_location
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, name
));
3308 sprintf(name
, "luminanceoffset%d", ((IWineD3DPixelShaderImpl
*)pshader
)->luminanceconst
[i
].texunit
);
3309 entry
->luminanceoffset_location
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, name
));
3314 entry
->posFixup_location
= GL_EXTCALL(glGetUniformLocationARB(programId
, "posFixup"));
3315 entry
->ycorrection_location
= GL_EXTCALL(glGetUniformLocationARB(programId
, "ycorrection"));
3316 checkGLcall("Find glsl program uniform locations");
3319 && WINED3DSHADER_VERSION_MAJOR(((IWineD3DPixelShaderImpl
*)pshader
)->baseShader
.reg_maps
.shader_version
) >= 3
3320 && ((IWineD3DPixelShaderImpl
*)pshader
)->declared_in_count
> GL_LIMITS(glsl_varyings
) / 4)
3322 TRACE("Shader %d needs vertex color clamping disabled\n", programId
);
3323 entry
->vertex_color_clamp
= GL_FALSE
;
3325 entry
->vertex_color_clamp
= GL_FIXED_ONLY_ARB
;
3328 /* Set the shader to allow uniform loading on it */
3329 GL_EXTCALL(glUseProgramObjectARB(programId
));
3330 checkGLcall("glUseProgramObjectARB(programId)");
3332 /* Load the vertex and pixel samplers now. The function that finds the mappings makes sure
3333 * that it stays the same for each vertexshader-pixelshader pair(=linked glsl program). If
3334 * a pshader with fixed function pipeline is used there are no vertex samplers, and if a
3335 * vertex shader with fixed function pixel processing is used we make sure that the card
3336 * supports enough samplers to allow the max number of vertex samplers with all possible
3337 * fixed function fragment processing setups. So once the program is linked these samplers
3341 /* Load vertex shader samplers */
3342 shader_glsl_load_vsamplers(gl_info
, (IWineD3DStateBlock
*)This
->stateBlock
, programId
);
3345 /* Load pixel shader samplers */
3346 shader_glsl_load_psamplers(gl_info
, (IWineD3DStateBlock
*)This
->stateBlock
, programId
);
3349 /* If the local constants do not have to be loaded with the environment constants,
3350 * load them now to have them hardcoded in the GLSL program. This saves some CPU cycles
3353 if(pshader
&& !((IWineD3DPixelShaderImpl
*)pshader
)->baseShader
.load_local_constsF
) {
3354 hardcode_local_constants((IWineD3DBaseShaderImpl
*) pshader
, gl_info
, programId
, 'P');
3356 if(vshader
&& !((IWineD3DVertexShaderImpl
*)vshader
)->baseShader
.load_local_constsF
) {
3357 hardcode_local_constants((IWineD3DBaseShaderImpl
*) vshader
, gl_info
, programId
, 'V');
3361 static GLhandleARB
create_glsl_blt_shader(const WineD3D_GL_Info
*gl_info
, enum tex_types tex_type
)
3363 GLhandleARB program_id
;
3364 GLhandleARB vshader_id
, pshader_id
;
3365 static const char *blt_vshader
[] =
3370 " gl_Position = gl_Vertex;\n"
3371 " gl_FrontColor = vec4(1.0);\n"
3372 " gl_TexCoord[0] = gl_MultiTexCoord0;\n"
3376 static const char *blt_pshaders
[tex_type_count
] =
3382 "uniform sampler2D sampler;\n"
3385 " gl_FragDepth = texture2D(sampler, gl_TexCoord[0].xy).x;\n"
3391 "uniform samplerCube sampler;\n"
3394 " gl_FragDepth = textureCube(sampler, gl_TexCoord[0].xyz).x;\n"
3398 "#extension GL_ARB_texture_rectangle : enable\n"
3399 "uniform sampler2DRect sampler;\n"
3402 " gl_FragDepth = texture2DRect(sampler, gl_TexCoord[0].xy).x;\n"
3406 if (!blt_pshaders
[tex_type
])
3408 FIXME("tex_type %#x not supported\n", tex_type
);
3412 vshader_id
= GL_EXTCALL(glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB
));
3413 GL_EXTCALL(glShaderSourceARB(vshader_id
, 1, blt_vshader
, NULL
));
3414 GL_EXTCALL(glCompileShaderARB(vshader_id
));
3416 pshader_id
= GL_EXTCALL(glCreateShaderObjectARB(GL_FRAGMENT_SHADER_ARB
));
3417 GL_EXTCALL(glShaderSourceARB(pshader_id
, 1, &blt_pshaders
[tex_type
], NULL
));
3418 GL_EXTCALL(glCompileShaderARB(pshader_id
));
3420 program_id
= GL_EXTCALL(glCreateProgramObjectARB());
3421 GL_EXTCALL(glAttachObjectARB(program_id
, vshader_id
));
3422 GL_EXTCALL(glAttachObjectARB(program_id
, pshader_id
));
3423 GL_EXTCALL(glLinkProgramARB(program_id
));
3425 print_glsl_info_log(&GLINFO_LOCATION
, program_id
);
3427 /* Once linked we can mark the shaders for deletion. They will be deleted once the program
3430 GL_EXTCALL(glDeleteObjectARB(vshader_id
));
3431 GL_EXTCALL(glDeleteObjectARB(pshader_id
));
3435 static void shader_glsl_select(IWineD3DDevice
*iface
, BOOL usePS
, BOOL useVS
) {
3436 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
3437 struct shader_glsl_priv
*priv
= (struct shader_glsl_priv
*)This
->shader_priv
;
3438 const WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
3439 GLhandleARB program_id
= 0;
3440 GLenum old_vertex_color_clamp
, current_vertex_color_clamp
;
3442 old_vertex_color_clamp
= priv
->glsl_program
? priv
->glsl_program
->vertex_color_clamp
: GL_FIXED_ONLY_ARB
;
3444 if (useVS
|| usePS
) set_glsl_shader_program(iface
, usePS
, useVS
);
3445 else priv
->glsl_program
= NULL
;
3447 current_vertex_color_clamp
= priv
->glsl_program
? priv
->glsl_program
->vertex_color_clamp
: GL_FIXED_ONLY_ARB
;
3449 if (old_vertex_color_clamp
!= current_vertex_color_clamp
) {
3450 if (GL_SUPPORT(ARB_COLOR_BUFFER_FLOAT
)) {
3451 GL_EXTCALL(glClampColorARB(GL_CLAMP_VERTEX_COLOR_ARB
, current_vertex_color_clamp
));
3452 checkGLcall("glClampColorARB");
3454 FIXME("vertex color clamp needs to be changed, but extension not supported.\n");
3458 program_id
= priv
->glsl_program
? priv
->glsl_program
->programId
: 0;
3459 if (program_id
) TRACE("Using GLSL program %u\n", program_id
);
3460 GL_EXTCALL(glUseProgramObjectARB(program_id
));
3461 checkGLcall("glUseProgramObjectARB");
3464 static void shader_glsl_select_depth_blt(IWineD3DDevice
*iface
, enum tex_types tex_type
) {
3465 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
3466 const WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
3467 struct shader_glsl_priv
*priv
= (struct shader_glsl_priv
*) This
->shader_priv
;
3468 GLhandleARB
*blt_program
= &priv
->depth_blt_program
[tex_type
];
3470 if (!*blt_program
) {
3472 *blt_program
= create_glsl_blt_shader(gl_info
, tex_type
);
3473 loc
= GL_EXTCALL(glGetUniformLocationARB(*blt_program
, "sampler"));
3474 GL_EXTCALL(glUseProgramObjectARB(*blt_program
));
3475 GL_EXTCALL(glUniform1iARB(loc
, 0));
3477 GL_EXTCALL(glUseProgramObjectARB(*blt_program
));
3481 static void shader_glsl_deselect_depth_blt(IWineD3DDevice
*iface
) {
3482 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
3483 const WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
3484 struct shader_glsl_priv
*priv
= (struct shader_glsl_priv
*) This
->shader_priv
;
3485 GLhandleARB program_id
;
3487 program_id
= priv
->glsl_program
? priv
->glsl_program
->programId
: 0;
3488 if (program_id
) TRACE("Using GLSL program %u\n", program_id
);
3490 GL_EXTCALL(glUseProgramObjectARB(program_id
));
3491 checkGLcall("glUseProgramObjectARB");
3494 static void shader_glsl_destroy(IWineD3DBaseShader
*iface
) {
3495 const struct list
*linked_programs
;
3496 IWineD3DBaseShaderImpl
*This
= (IWineD3DBaseShaderImpl
*) iface
;
3497 IWineD3DDeviceImpl
*device
= (IWineD3DDeviceImpl
*)This
->baseShader
.device
;
3498 struct shader_glsl_priv
*priv
= (struct shader_glsl_priv
*)device
->shader_priv
;
3499 const WineD3D_GL_Info
*gl_info
= &device
->adapter
->gl_info
;
3500 IWineD3DPixelShaderImpl
*ps
= NULL
;
3501 IWineD3DVertexShaderImpl
*vs
= NULL
;
3503 /* Note: Do not use QueryInterface here to find out which shader type this is because this code
3504 * can be called from IWineD3DBaseShader::Release
3506 char pshader
= shader_is_pshader_version(This
->baseShader
.reg_maps
.shader_version
);
3509 ps
= (IWineD3DPixelShaderImpl
*) This
;
3510 if(ps
->num_gl_shaders
== 0) return;
3512 vs
= (IWineD3DVertexShaderImpl
*) This
;
3513 if(vs
->prgId
== 0) return;
3516 linked_programs
= &This
->baseShader
.linked_programs
;
3518 TRACE("Deleting linked programs\n");
3519 if (linked_programs
->next
) {
3520 struct glsl_shader_prog_link
*entry
, *entry2
;
3523 LIST_FOR_EACH_ENTRY_SAFE(entry
, entry2
, linked_programs
, struct glsl_shader_prog_link
, pshader_entry
) {
3524 delete_glsl_program_entry(priv
, gl_info
, entry
);
3527 LIST_FOR_EACH_ENTRY_SAFE(entry
, entry2
, linked_programs
, struct glsl_shader_prog_link
, vshader_entry
) {
3528 delete_glsl_program_entry(priv
, gl_info
, entry
);
3537 for(i
= 0; i
< ps
->num_gl_shaders
; i
++) {
3538 TRACE("deleting pshader %u\n", ps
->gl_shaders
[i
].prgId
);
3539 GL_EXTCALL(glDeleteObjectARB(ps
->gl_shaders
[i
].prgId
));
3540 checkGLcall("glDeleteObjectARB");
3543 HeapFree(GetProcessHeap(), 0, ps
->gl_shaders
);
3544 ps
->gl_shaders
= NULL
;
3545 ps
->num_gl_shaders
= 0;
3547 TRACE("Deleting shader object %u\n", vs
->prgId
);
3549 GL_EXTCALL(glDeleteObjectARB(vs
->prgId
));
3550 checkGLcall("glDeleteObjectARB");
3553 vs
->baseShader
.is_compiled
= FALSE
;
3557 static unsigned int glsl_program_key_hash(const void *key
)
3559 const glsl_program_key_t
*k
= (const glsl_program_key_t
*)key
;
3561 unsigned int hash
= k
->vshader
| ((DWORD_PTR
) k
->pshader
) << 16;
3562 hash
+= ~(hash
<< 15);
3563 hash
^= (hash
>> 10);
3564 hash
+= (hash
<< 3);
3565 hash
^= (hash
>> 6);
3566 hash
+= ~(hash
<< 11);
3567 hash
^= (hash
>> 16);
3572 static BOOL
glsl_program_key_compare(const void *keya
, const void *keyb
)
3574 const glsl_program_key_t
*ka
= (const glsl_program_key_t
*)keya
;
3575 const glsl_program_key_t
*kb
= (const glsl_program_key_t
*)keyb
;
3577 return ka
->vshader
== kb
->vshader
&& ka
->pshader
== kb
->pshader
&&
3578 (memcmp(&ka
->ps_args
, &kb
->ps_args
, sizeof(kb
->ps_args
)) == 0);
3581 static HRESULT
shader_glsl_alloc(IWineD3DDevice
*iface
) {
3582 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
3583 struct shader_glsl_priv
*priv
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(struct shader_glsl_priv
));
3584 priv
->glsl_program_lookup
= hash_table_create(glsl_program_key_hash
, glsl_program_key_compare
);
3585 This
->shader_priv
= priv
;
3589 static void shader_glsl_free(IWineD3DDevice
*iface
) {
3590 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
3591 const WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
3592 struct shader_glsl_priv
*priv
= (struct shader_glsl_priv
*)This
->shader_priv
;
3595 for (i
= 0; i
< tex_type_count
; ++i
)
3597 if (priv
->depth_blt_program
[i
])
3599 GL_EXTCALL(glDeleteObjectARB(priv
->depth_blt_program
[i
]));
3603 hash_table_destroy(priv
->glsl_program_lookup
, NULL
, NULL
);
3605 HeapFree(GetProcessHeap(), 0, This
->shader_priv
);
3606 This
->shader_priv
= NULL
;
3609 static BOOL
shader_glsl_dirty_const(IWineD3DDevice
*iface
) {
3610 /* TODO: GL_EXT_bindable_uniform can be used to share constants across shaders */
3614 static GLuint
shader_glsl_generate_pshader(IWineD3DPixelShader
*iface
, SHADER_BUFFER
*buffer
, const struct ps_compile_args
*args
) {
3615 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*)iface
;
3616 const struct shader_reg_maps
*reg_maps
= &This
->baseShader
.reg_maps
;
3617 CONST DWORD
*function
= This
->baseShader
.function
;
3618 const char *fragcolor
;
3619 const WineD3D_GL_Info
*gl_info
= &((IWineD3DDeviceImpl
*)This
->baseShader
.device
)->adapter
->gl_info
;
3621 /* Create the hw GLSL shader object and assign it as the shader->prgId */
3622 GLhandleARB shader_obj
= GL_EXTCALL(glCreateShaderObjectARB(GL_FRAGMENT_SHADER_ARB
));
3624 shader_addline(buffer
, "#version 120\n");
3626 if (GL_SUPPORT(ARB_DRAW_BUFFERS
)) {
3627 shader_addline(buffer
, "#extension GL_ARB_draw_buffers : enable\n");
3629 if (GL_SUPPORT(ARB_TEXTURE_RECTANGLE
)) {
3630 /* The spec says that it doesn't have to be explicitly enabled, but the nvidia
3631 * drivers write a warning if we don't do so
3633 shader_addline(buffer
, "#extension GL_ARB_texture_rectangle : enable\n");
3636 /* Base Declarations */
3637 shader_generate_glsl_declarations( (IWineD3DBaseShader
*) This
, reg_maps
, buffer
, &GLINFO_LOCATION
, args
);
3639 /* Pack 3.0 inputs */
3640 if (reg_maps
->shader_version
>= WINED3DPS_VERSION(3,0) && args
->vp_mode
!= vertexshader
) {
3641 pshader_glsl_input_pack(buffer
, This
->semantics_in
, iface
, args
->vp_mode
);
3644 /* Base Shader Body */
3645 shader_generate_main( (IWineD3DBaseShader
*) This
, buffer
, reg_maps
, function
);
3647 /* Pixel shaders < 2.0 place the resulting color in R0 implicitly */
3648 if (reg_maps
->shader_version
< WINED3DPS_VERSION(2,0))
3650 /* Some older cards like GeforceFX ones don't support multiple buffers, so also not gl_FragData */
3651 if(GL_SUPPORT(ARB_DRAW_BUFFERS
))
3652 shader_addline(buffer
, "gl_FragData[0] = R0;\n");
3654 shader_addline(buffer
, "gl_FragColor = R0;\n");
3657 if(GL_SUPPORT(ARB_DRAW_BUFFERS
)) {
3658 fragcolor
= "gl_FragData[0]";
3660 fragcolor
= "gl_FragColor";
3662 if(args
->srgb_correction
) {
3663 shader_addline(buffer
, "tmp0.xyz = pow(%s.xyz, vec3(%f, %f, %f)) * vec3(%f, %f, %f) - vec3(%f, %f, %f);\n",
3664 fragcolor
, srgb_pow
, srgb_pow
, srgb_pow
, srgb_mul_high
, srgb_mul_high
, srgb_mul_high
,
3665 srgb_sub_high
, srgb_sub_high
, srgb_sub_high
);
3666 shader_addline(buffer
, "tmp1.xyz = %s.xyz * srgb_mul_low.xyz;\n", fragcolor
);
3667 shader_addline(buffer
, "%s.x = %s.x < srgb_comparison.x ? tmp1.x : tmp0.x;\n", fragcolor
, fragcolor
);
3668 shader_addline(buffer
, "%s.y = %s.y < srgb_comparison.y ? tmp1.y : tmp0.y;\n", fragcolor
, fragcolor
);
3669 shader_addline(buffer
, "%s.z = %s.z < srgb_comparison.z ? tmp1.z : tmp0.z;\n", fragcolor
, fragcolor
);
3670 shader_addline(buffer
, "%s = clamp(%s, 0.0, 1.0);\n", fragcolor
, fragcolor
);
3672 /* Pixel shader < 3.0 do not replace the fog stage.
3673 * This implements linear fog computation and blending.
3674 * TODO: non linear fog
3675 * NOTE: gl_Fog.start and gl_Fog.end don't hold fog start s and end e but
3676 * -1/(e-s) and e/(e-s) respectively.
3678 if (reg_maps
->shader_version
< WINED3DPS_VERSION(3,0))
3680 shader_addline(buffer
, "float Fog = clamp(gl_FogFragCoord * gl_Fog.start + gl_Fog.end, 0.0, 1.0);\n");
3681 shader_addline(buffer
, "%s.xyz = mix(gl_Fog.color.xyz, %s.xyz, Fog);\n", fragcolor
, fragcolor
);
3684 shader_addline(buffer
, "}\n");
3686 TRACE("Compiling shader object %u\n", shader_obj
);
3687 GL_EXTCALL(glShaderSourceARB(shader_obj
, 1, (const char**)&buffer
->buffer
, NULL
));
3688 GL_EXTCALL(glCompileShaderARB(shader_obj
));
3689 print_glsl_info_log(&GLINFO_LOCATION
, shader_obj
);
3691 /* Store the shader object */
3695 static void shader_glsl_generate_vshader(IWineD3DVertexShader
*iface
, SHADER_BUFFER
*buffer
) {
3696 IWineD3DVertexShaderImpl
*This
= (IWineD3DVertexShaderImpl
*)iface
;
3697 const struct shader_reg_maps
*reg_maps
= &This
->baseShader
.reg_maps
;
3698 CONST DWORD
*function
= This
->baseShader
.function
;
3699 const WineD3D_GL_Info
*gl_info
= &((IWineD3DDeviceImpl
*)This
->baseShader
.device
)->adapter
->gl_info
;
3701 /* Create the hw GLSL shader program and assign it as the shader->prgId */
3702 GLhandleARB shader_obj
= GL_EXTCALL(glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB
));
3704 shader_addline(buffer
, "#version 120\n");
3706 /* Base Declarations */
3707 shader_generate_glsl_declarations( (IWineD3DBaseShader
*) This
, reg_maps
, buffer
, &GLINFO_LOCATION
, NULL
);
3709 /* Base Shader Body */
3710 shader_generate_main( (IWineD3DBaseShader
*) This
, buffer
, reg_maps
, function
);
3712 /* Unpack 3.0 outputs */
3713 if (reg_maps
->shader_version
>= WINED3DVS_VERSION(3,0)) shader_addline(buffer
, "order_ps_input(OUT);\n");
3714 else shader_addline(buffer
, "order_ps_input();\n");
3716 /* If this shader doesn't use fog copy the z coord to the fog coord so that we can use table fog */
3718 shader_addline(buffer
, "gl_FogFragCoord = gl_Position.z;\n");
3720 /* Write the final position.
3722 * OpenGL coordinates specify the center of the pixel while d3d coords specify
3723 * the corner. The offsets are stored in z and w in posFixup. posFixup.y contains
3724 * 1.0 or -1.0 to turn the rendering upside down for offscreen rendering. PosFixup.x
3725 * contains 1.0 to allow a mad.
3727 shader_addline(buffer
, "gl_Position.y = gl_Position.y * posFixup.y;\n");
3728 shader_addline(buffer
, "gl_Position.xy += posFixup.zw * gl_Position.ww;\n");
3730 /* Z coord [0;1]->[-1;1] mapping, see comment in transform_projection in state.c
3732 * Basically we want (in homogeneous coordinates) z = z * 2 - 1. However, shaders are run
3733 * before the homogeneous divide, so we have to take the w into account: z = ((z / w) * 2 - 1) * w,
3734 * which is the same as z = z * 2 - w.
3736 shader_addline(buffer
, "gl_Position.z = gl_Position.z * 2.0 - gl_Position.w;\n");
3738 shader_addline(buffer
, "}\n");
3740 TRACE("Compiling shader object %u\n", shader_obj
);
3741 GL_EXTCALL(glShaderSourceARB(shader_obj
, 1, (const char**)&buffer
->buffer
, NULL
));
3742 GL_EXTCALL(glCompileShaderARB(shader_obj
));
3743 print_glsl_info_log(&GLINFO_LOCATION
, shader_obj
);
3745 /* Store the shader object */
3746 This
->prgId
= shader_obj
;
3749 static void shader_glsl_get_caps(WINED3DDEVTYPE devtype
, const WineD3D_GL_Info
*gl_info
, struct shader_caps
*pCaps
)
3751 /* Nvidia Geforce6/7 or Ati R4xx/R5xx cards with GLSL support, support VS 3.0 but older Nvidia/Ati
3752 * models with GLSL support only support 2.0. In case of nvidia we can detect VS 2.0 support using
3753 * vs_nv_version which is based on NV_vertex_program.
3754 * For Ati cards there's no way using glsl (it abstracts the lowlevel info away) and also not
3755 * using ARB_vertex_program. It is safe to assume that when a card supports pixel shader 2.0 it
3756 * supports vertex shader 2.0 too and the way around. We can detect ps2.0 using the maximum number
3757 * of native instructions, so use that here. For more info see the pixel shader versioning code below.
3759 if((GLINFO_LOCATION
.vs_nv_version
== VS_VERSION_20
) || (GLINFO_LOCATION
.ps_arb_max_instructions
<= 512))
3760 pCaps
->VertexShaderVersion
= WINED3DVS_VERSION(2,0);
3762 pCaps
->VertexShaderVersion
= WINED3DVS_VERSION(3,0);
3763 TRACE_(d3d_caps
)("Hardware vertex shader version %d.%d enabled (GLSL)\n", (pCaps
->VertexShaderVersion
>> 8) & 0xff, pCaps
->VertexShaderVersion
& 0xff);
3764 pCaps
->MaxVertexShaderConst
= GL_LIMITS(vshader_constantsF
);
3766 /* Older DX9-class videocards (GeforceFX / Radeon >9500/X*00) only support pixel shader 2.0/2.0a/2.0b.
3767 * In OpenGL the extensions related to GLSL abstract lowlevel GL info away which is needed
3768 * to distinguish between 2.0 and 3.0 (and 2.0a/2.0b). In case of Nvidia we use their fragment
3769 * program extensions. On other hardware including ATI GL_ARB_fragment_program offers the info
3770 * in max native instructions. Intel and others also offer the info in this extension but they
3771 * don't support GLSL (at least on Windows).
3773 * PS2.0 requires at least 96 instructions, 2.0a/2.0b go up to 512. Assume that if the number
3774 * of instructions is 512 or less we have to do with ps2.0 hardware.
3775 * 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.
3777 if((GLINFO_LOCATION
.ps_nv_version
== PS_VERSION_20
) || (GLINFO_LOCATION
.ps_arb_max_instructions
<= 512))
3778 pCaps
->PixelShaderVersion
= WINED3DPS_VERSION(2,0);
3780 pCaps
->PixelShaderVersion
= WINED3DPS_VERSION(3,0);
3782 /* FIXME: The following line is card dependent. -8.0 to 8.0 is the
3783 * Direct3D minimum requirement.
3785 * Both GL_ARB_fragment_program and GLSL require a "maximum representable magnitude"
3786 * of colors to be 2^10, and 2^32 for other floats. Should we use 1024 here?
3788 * The problem is that the refrast clamps temporary results in the shader to
3789 * [-MaxValue;+MaxValue]. If the card's max value is bigger than the one we advertize here,
3790 * then applications may miss the clamping behavior. On the other hand, if it is smaller,
3791 * the shader will generate incorrect results too. Unfortunately, GL deliberately doesn't
3792 * offer a way to query this.
3794 pCaps
->PixelShader1xMaxValue
= 8.0;
3795 TRACE_(d3d_caps
)("Hardware pixel shader version %d.%d enabled (GLSL)\n", (pCaps
->PixelShaderVersion
>> 8) & 0xff, pCaps
->PixelShaderVersion
& 0xff);
3798 static BOOL
shader_glsl_color_fixup_supported(struct color_fixup_desc fixup
)
3800 if (TRACE_ON(d3d_shader
) && TRACE_ON(d3d
))
3802 TRACE("Checking support for fixup:\n");
3803 dump_color_fixup_desc(fixup
);
3806 /* We support everything except YUV conversions. */
3807 if (!is_yuv_fixup(fixup
))
3813 TRACE("[FAILED]\n");
3817 static const SHADER_HANDLER shader_glsl_instruction_handler_table
[WINED3DSIH_TABLE_SIZE
] =
3819 /* WINED3DSIH_ABS */ shader_glsl_map2gl
,
3820 /* WINED3DSIH_ADD */ shader_glsl_arith
,
3821 /* WINED3DSIH_BEM */ pshader_glsl_bem
,
3822 /* WINED3DSIH_BREAK */ shader_glsl_break
,
3823 /* WINED3DSIH_BREAKC */ shader_glsl_breakc
,
3824 /* WINED3DSIH_BREAKP */ NULL
,
3825 /* WINED3DSIH_CALL */ shader_glsl_call
,
3826 /* WINED3DSIH_CALLNZ */ shader_glsl_callnz
,
3827 /* WINED3DSIH_CMP */ shader_glsl_cmp
,
3828 /* WINED3DSIH_CND */ shader_glsl_cnd
,
3829 /* WINED3DSIH_CRS */ shader_glsl_cross
,
3830 /* WINED3DSIH_DCL */ NULL
,
3831 /* WINED3DSIH_DEF */ NULL
,
3832 /* WINED3DSIH_DEFB */ NULL
,
3833 /* WINED3DSIH_DEFI */ NULL
,
3834 /* WINED3DSIH_DP2ADD */ pshader_glsl_dp2add
,
3835 /* WINED3DSIH_DP3 */ shader_glsl_dot
,
3836 /* WINED3DSIH_DP4 */ shader_glsl_dot
,
3837 /* WINED3DSIH_DST */ shader_glsl_dst
,
3838 /* WINED3DSIH_DSX */ shader_glsl_map2gl
,
3839 /* WINED3DSIH_DSY */ shader_glsl_map2gl
,
3840 /* WINED3DSIH_ELSE */ shader_glsl_else
,
3841 /* WINED3DSIH_ENDIF */ shader_glsl_end
,
3842 /* WINED3DSIH_ENDLOOP */ shader_glsl_end
,
3843 /* WINED3DSIH_ENDREP */ shader_glsl_end
,
3844 /* WINED3DSIH_EXP */ shader_glsl_map2gl
,
3845 /* WINED3DSIH_EXPP */ shader_glsl_expp
,
3846 /* WINED3DSIH_FRC */ shader_glsl_map2gl
,
3847 /* WINED3DSIH_IF */ shader_glsl_if
,
3848 /* WINED3DSIH_IFC */ shader_glsl_ifc
,
3849 /* WINED3DSIH_LABEL */ shader_glsl_label
,
3850 /* WINED3DSIH_LIT */ shader_glsl_lit
,
3851 /* WINED3DSIH_LOG */ shader_glsl_log
,
3852 /* WINED3DSIH_LOGP */ shader_glsl_log
,
3853 /* WINED3DSIH_LOOP */ shader_glsl_loop
,
3854 /* WINED3DSIH_LRP */ shader_glsl_lrp
,
3855 /* WINED3DSIH_M3x2 */ shader_glsl_mnxn
,
3856 /* WINED3DSIH_M3x3 */ shader_glsl_mnxn
,
3857 /* WINED3DSIH_M3x4 */ shader_glsl_mnxn
,
3858 /* WINED3DSIH_M4x3 */ shader_glsl_mnxn
,
3859 /* WINED3DSIH_M4x4 */ shader_glsl_mnxn
,
3860 /* WINED3DSIH_MAD */ shader_glsl_mad
,
3861 /* WINED3DSIH_MAX */ shader_glsl_map2gl
,
3862 /* WINED3DSIH_MIN */ shader_glsl_map2gl
,
3863 /* WINED3DSIH_MOV */ shader_glsl_mov
,
3864 /* WINED3DSIH_MOVA */ shader_glsl_mov
,
3865 /* WINED3DSIH_MUL */ shader_glsl_arith
,
3866 /* WINED3DSIH_NOP */ NULL
,
3867 /* WINED3DSIH_NRM */ shader_glsl_map2gl
,
3868 /* WINED3DSIH_PHASE */ NULL
,
3869 /* WINED3DSIH_POW */ shader_glsl_pow
,
3870 /* WINED3DSIH_RCP */ shader_glsl_rcp
,
3871 /* WINED3DSIH_REP */ shader_glsl_rep
,
3872 /* WINED3DSIH_RET */ NULL
,
3873 /* WINED3DSIH_RSQ */ shader_glsl_rsq
,
3874 /* WINED3DSIH_SETP */ NULL
,
3875 /* WINED3DSIH_SGE */ shader_glsl_compare
,
3876 /* WINED3DSIH_SGN */ shader_glsl_map2gl
,
3877 /* WINED3DSIH_SINCOS */ shader_glsl_sincos
,
3878 /* WINED3DSIH_SLT */ shader_glsl_compare
,
3879 /* WINED3DSIH_SUB */ shader_glsl_arith
,
3880 /* WINED3DSIH_TEX */ pshader_glsl_tex
,
3881 /* WINED3DSIH_TEXBEM */ pshader_glsl_texbem
,
3882 /* WINED3DSIH_TEXBEML */ pshader_glsl_texbem
,
3883 /* WINED3DSIH_TEXCOORD */ pshader_glsl_texcoord
,
3884 /* WINED3DSIH_TEXDEPTH */ pshader_glsl_texdepth
,
3885 /* WINED3DSIH_TEXDP3 */ pshader_glsl_texdp3
,
3886 /* WINED3DSIH_TEXDP3TEX */ pshader_glsl_texdp3tex
,
3887 /* WINED3DSIH_TEXKILL */ pshader_glsl_texkill
,
3888 /* WINED3DSIH_TEXLDD */ NULL
,
3889 /* WINED3DSIH_TEXLDL */ shader_glsl_texldl
,
3890 /* WINED3DSIH_TEXM3x2DEPTH */ pshader_glsl_texm3x2depth
,
3891 /* WINED3DSIH_TEXM3x2PAD */ pshader_glsl_texm3x2pad
,
3892 /* WINED3DSIH_TEXM3x2TEX */ pshader_glsl_texm3x2tex
,
3893 /* WINED3DSIH_TEXM3x3 */ pshader_glsl_texm3x3
,
3894 /* WINED3DSIH_TEXM3x3DIFF */ NULL
,
3895 /* WINED3DSIH_TEXM3x3PAD */ pshader_glsl_texm3x3pad
,
3896 /* WINED3DSIH_TEXM3x3SPEC */ pshader_glsl_texm3x3spec
,
3897 /* WINED3DSIH_TEXM3x3TEX */ pshader_glsl_texm3x3tex
,
3898 /* WINED3DSIH_TEXM3x3VSPEC */ pshader_glsl_texm3x3vspec
,
3899 /* WINED3DSIH_TEXREG2AR */ pshader_glsl_texreg2ar
,
3900 /* WINED3DSIH_TEXREG2GB */ pshader_glsl_texreg2gb
,
3901 /* WINED3DSIH_TEXREG2RGB */ pshader_glsl_texreg2rgb
,
3904 const shader_backend_t glsl_shader_backend
= {
3905 shader_glsl_instruction_handler_table
,
3907 shader_glsl_select_depth_blt
,
3908 shader_glsl_deselect_depth_blt
,
3909 shader_glsl_update_float_vertex_constants
,
3910 shader_glsl_update_float_pixel_constants
,
3911 shader_glsl_load_constants
,
3912 shader_glsl_color_correction
,
3913 shader_glsl_destroy
,
3916 shader_glsl_dirty_const
,
3917 shader_glsl_generate_pshader
,
3918 shader_glsl_generate_vshader
,
3919 shader_glsl_get_caps
,
3920 shader_glsl_color_fixup_supported
,