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
);
39 #define GLINFO_LOCATION (*gl_info)
54 } glsl_sample_function_t
;
56 /* GLSL shader private data */
57 struct shader_glsl_priv
{
58 struct hash_table_t
*glsl_program_lookup
;
59 const struct glsl_shader_prog_link
*glsl_program
;
60 GLhandleARB depth_blt_program
[tex_type_count
];
63 /* Struct to maintain data about a linked GLSL program */
64 struct glsl_shader_prog_link
{
65 struct list vshader_entry
;
66 struct list pshader_entry
;
67 GLhandleARB programId
;
68 GLhandleARB
*vuniformF_locations
;
69 GLhandleARB
*puniformF_locations
;
70 GLhandleARB vuniformI_locations
[MAX_CONST_I
];
71 GLhandleARB puniformI_locations
[MAX_CONST_I
];
72 GLhandleARB posFixup_location
;
73 GLhandleARB bumpenvmat_location
[MAX_TEXTURES
];
74 GLhandleARB luminancescale_location
[MAX_TEXTURES
];
75 GLhandleARB luminanceoffset_location
[MAX_TEXTURES
];
76 GLhandleARB ycorrection_location
;
77 GLenum vertex_color_clamp
;
79 IWineD3DPixelShader
*pshader
;
80 struct ps_compile_args ps_args
;
85 IWineD3DPixelShader
*pshader
;
86 struct ps_compile_args ps_args
;
90 /** Prints the GLSL info log which will contain error messages if they exist */
91 static void print_glsl_info_log(const WineD3D_GL_Info
*gl_info
, GLhandleARB obj
)
93 int infologLength
= 0;
98 const char *spam
[] = {
99 "Vertex shader was successfully compiled to run on hardware.\n", /* fglrx */
100 "Fragment shader was successfully compiled to run on hardware.\n", /* fglrx */
101 "Fragment shader(s) linked, vertex shader(s) linked. \n ", /* fglrx, with \n */
102 "Fragment shader(s) linked, vertex shader(s) linked.", /* fglrx, no \n */
103 "Vertex shader(s) linked, no fragment shader(s) defined. \n ", /* fglrx, with \n */
104 "Vertex shader(s) linked, no fragment shader(s) defined.", /* fglrx, no \n */
105 "Fragment shader was successfully compiled to run on hardware.\nWARNING: 0:1: extension 'GL_ARB_draw_buffers' is not supported",
106 "Fragment shader(s) linked, no vertex shader(s) defined.", /* fglrx, no \n */
107 "Fragment shader(s) linked, no vertex shader(s) defined. \n ", /* fglrx, with \n */
108 "WARNING: 0:2: extension 'GL_ARB_draw_buffers' is not supported\n" /* MacOS ati */
111 GL_EXTCALL(glGetObjectParameterivARB(obj
,
112 GL_OBJECT_INFO_LOG_LENGTH_ARB
,
115 /* A size of 1 is just a null-terminated string, so the log should be bigger than
116 * that if there are errors. */
117 if (infologLength
> 1)
119 /* Fglrx doesn't terminate the string properly, but it tells us the proper length.
120 * So use HEAP_ZERO_MEMORY to avoid uninitialized bytes
122 infoLog
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, infologLength
);
123 GL_EXTCALL(glGetInfoLogARB(obj
, infologLength
, NULL
, infoLog
));
126 for(i
= 0; i
< sizeof(spam
) / sizeof(spam
[0]); i
++) {
127 if(strcmp(infoLog
, spam
[i
]) == 0) {
133 TRACE("Spam received from GLSL shader #%u: %s\n", obj
, debugstr_a(infoLog
));
135 FIXME("Error received from GLSL shader #%u: %s\n", obj
, debugstr_a(infoLog
));
137 HeapFree(GetProcessHeap(), 0, infoLog
);
142 * Loads (pixel shader) samplers
144 static void shader_glsl_load_psamplers(const WineD3D_GL_Info
*gl_info
, IWineD3DStateBlock
*iface
, GLhandleARB programId
)
146 IWineD3DStateBlockImpl
* stateBlock
= (IWineD3DStateBlockImpl
*) iface
;
147 GLhandleARB name_loc
;
149 char sampler_name
[20];
151 for (i
= 0; i
< MAX_FRAGMENT_SAMPLERS
; ++i
) {
152 snprintf(sampler_name
, sizeof(sampler_name
), "Psampler%d", i
);
153 name_loc
= GL_EXTCALL(glGetUniformLocationARB(programId
, sampler_name
));
154 if (name_loc
!= -1) {
155 int mapped_unit
= stateBlock
->wineD3DDevice
->texUnitMap
[i
];
156 if (mapped_unit
!= -1 && mapped_unit
< GL_LIMITS(fragment_samplers
)) {
157 TRACE("Loading %s for texture %d\n", sampler_name
, mapped_unit
);
158 GL_EXTCALL(glUniform1iARB(name_loc
, mapped_unit
));
159 checkGLcall("glUniform1iARB");
161 ERR("Trying to load sampler %s on unsupported unit %d\n", sampler_name
, mapped_unit
);
167 static void shader_glsl_load_vsamplers(const WineD3D_GL_Info
*gl_info
, IWineD3DStateBlock
*iface
, GLhandleARB programId
)
169 IWineD3DStateBlockImpl
* stateBlock
= (IWineD3DStateBlockImpl
*) iface
;
170 GLhandleARB name_loc
;
171 char sampler_name
[20];
174 for (i
= 0; i
< MAX_VERTEX_SAMPLERS
; ++i
) {
175 snprintf(sampler_name
, sizeof(sampler_name
), "Vsampler%d", i
);
176 name_loc
= GL_EXTCALL(glGetUniformLocationARB(programId
, sampler_name
));
177 if (name_loc
!= -1) {
178 int mapped_unit
= stateBlock
->wineD3DDevice
->texUnitMap
[MAX_FRAGMENT_SAMPLERS
+ i
];
179 if (mapped_unit
!= -1 && mapped_unit
< GL_LIMITS(combined_samplers
)) {
180 TRACE("Loading %s for texture %d\n", sampler_name
, mapped_unit
);
181 GL_EXTCALL(glUniform1iARB(name_loc
, mapped_unit
));
182 checkGLcall("glUniform1iARB");
184 ERR("Trying to load sampler %s on unsupported unit %d\n", sampler_name
, mapped_unit
);
191 * Loads floating point constants (aka uniforms) into the currently set GLSL program.
192 * When constant_list == NULL, it will load all the constants.
194 static void shader_glsl_load_constantsF(IWineD3DBaseShaderImpl
*This
, const WineD3D_GL_Info
*gl_info
,
195 unsigned int max_constants
, const float *constants
, const GLhandleARB
*constant_locations
,
196 const struct list
*constant_list
)
198 const constants_entry
*constant
;
199 const local_constant
*lconst
;
204 if (TRACE_ON(d3d_shader
)) {
205 LIST_FOR_EACH_ENTRY(constant
, constant_list
, constants_entry
, entry
) {
210 tmp_loc
= constant_locations
[i
];
212 TRACE_(d3d_constants
)("Loading constants %i: %f, %f, %f, %f\n", i
,
213 constants
[i
* 4 + 0], constants
[i
* 4 + 1],
214 constants
[i
* 4 + 2], constants
[i
* 4 + 3]);
220 /* 1.X pshaders have the constants clamped to [-1;1] implicitly. */
221 if(WINED3DSHADER_VERSION_MAJOR(This
->baseShader
.hex_version
) == 1 &&
222 shader_is_pshader_version(This
->baseShader
.hex_version
)) {
225 LIST_FOR_EACH_ENTRY(constant
, constant_list
, constants_entry
, entry
) {
230 tmp_loc
= constant_locations
[i
];
232 /* We found this uniform name in the program - go ahead and send the data */
234 if(constants
[k
+ 0] < -1.0) lcl_const
[0] = -1.0;
235 else if(constants
[k
+ 0] > 1.0) lcl_const
[0] = 1.0;
236 else lcl_const
[0] = constants
[k
+ 0];
237 if(constants
[k
+ 1] < -1.0) lcl_const
[1] = -1.0;
238 else if(constants
[k
+ 1] > 1.0) lcl_const
[1] = 1.0;
239 else lcl_const
[1] = constants
[k
+ 1];
240 if(constants
[k
+ 2] < -1.0) lcl_const
[2] = -1.0;
241 else if(constants
[k
+ 2] > 1.0) lcl_const
[2] = 1.0;
242 else lcl_const
[2] = constants
[k
+ 2];
243 if(constants
[k
+ 3] < -1.0) lcl_const
[3] = -1.0;
244 else if(constants
[k
+ 3] > 1.0) lcl_const
[3] = 1.0;
245 else lcl_const
[3] = constants
[k
+ 3];
247 GL_EXTCALL(glUniform4fvARB(tmp_loc
, 1, lcl_const
));
252 LIST_FOR_EACH_ENTRY(constant
, constant_list
, constants_entry
, entry
) {
257 tmp_loc
= constant_locations
[i
];
259 /* We found this uniform name in the program - go ahead and send the data */
260 GL_EXTCALL(glUniform4fvARB(tmp_loc
, 1, constants
+ (i
* 4)));
265 checkGLcall("glUniform4fvARB()");
267 if(!This
->baseShader
.load_local_constsF
) {
268 TRACE("No need to load local float constants for this shader\n");
272 /* Load immediate constants */
273 if (TRACE_ON(d3d_shader
)) {
274 LIST_FOR_EACH_ENTRY(lconst
, &This
->baseShader
.constantsF
, local_constant
, entry
) {
275 tmp_loc
= constant_locations
[lconst
->idx
];
277 const GLfloat
*values
= (const GLfloat
*)lconst
->value
;
278 TRACE_(d3d_constants
)("Loading local constants %i: %f, %f, %f, %f\n", lconst
->idx
,
279 values
[0], values
[1], values
[2], values
[3]);
283 /* Immediate constants are clamped to [-1;1] at shader creation time if needed */
284 LIST_FOR_EACH_ENTRY(lconst
, &This
->baseShader
.constantsF
, local_constant
, entry
) {
285 tmp_loc
= constant_locations
[lconst
->idx
];
287 /* We found this uniform name in the program - go ahead and send the data */
288 GL_EXTCALL(glUniform4fvARB(tmp_loc
, 1, (const GLfloat
*)lconst
->value
));
291 checkGLcall("glUniform4fvARB()");
295 * Loads integer constants (aka uniforms) into the currently set GLSL program.
296 * When @constants_set == NULL, it will load all the constants.
298 static void shader_glsl_load_constantsI(IWineD3DBaseShaderImpl
*This
, const WineD3D_GL_Info
*gl_info
,
299 GLhandleARB programId
, const GLhandleARB locations
[MAX_CONST_I
], unsigned int max_constants
,
300 const int *constants
, const BOOL
*constants_set
)
305 for (i
=0; i
<max_constants
; ++i
) {
306 if (NULL
== constants_set
|| constants_set
[i
]) {
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");
317 /* Load immediate constants */
318 ptr
= list_head(&This
->baseShader
.constantsI
);
320 const struct local_constant
*lconst
= LIST_ENTRY(ptr
, const struct local_constant
, entry
);
321 unsigned int idx
= lconst
->idx
;
322 const GLint
*values
= (const GLint
*)lconst
->value
;
324 TRACE_(d3d_constants
)("Loading local constants %i: %i, %i, %i, %i\n", idx
,
325 values
[0], values
[1], values
[2], values
[3]);
327 /* We found this uniform name in the program - go ahead and send the data */
328 GL_EXTCALL(glUniform4ivARB(locations
[idx
], 1, values
));
329 checkGLcall("glUniform4ivARB");
330 ptr
= list_next(&This
->baseShader
.constantsI
, ptr
);
335 * Loads boolean constants (aka uniforms) into the currently set GLSL program.
336 * When @constants_set == NULL, it will load all the constants.
338 static void shader_glsl_load_constantsB(IWineD3DBaseShaderImpl
*This
, const WineD3D_GL_Info
*gl_info
,
339 GLhandleARB programId
, unsigned int max_constants
, const BOOL
*constants
, const BOOL
*constants_set
)
344 char is_pshader
= shader_is_pshader_version(This
->baseShader
.hex_version
);
345 const char* prefix
= is_pshader
? "PB":"VB";
348 for (i
=0; i
<max_constants
; ++i
) {
349 if (NULL
== constants_set
|| constants_set
[i
]) {
351 TRACE_(d3d_constants
)("Loading constants %u: %i;\n", i
, constants
[i
]);
353 /* TODO: Benchmark and see if it would be beneficial to store the
354 * locations of the constants to avoid looking up each time */
355 snprintf(tmp_name
, sizeof(tmp_name
), "%s[%u]", prefix
, i
);
356 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");
365 /* Load immediate constants */
366 ptr
= list_head(&This
->baseShader
.constantsB
);
368 const struct local_constant
*lconst
= LIST_ENTRY(ptr
, const struct local_constant
, entry
);
369 unsigned int idx
= lconst
->idx
;
370 const GLint
*values
= (const GLint
*)lconst
->value
;
372 TRACE_(d3d_constants
)("Loading local constants %i: %i\n", idx
, values
[0]);
374 snprintf(tmp_name
, sizeof(tmp_name
), "%s[%i]", prefix
, idx
);
375 tmp_loc
= GL_EXTCALL(glGetUniformLocationARB(programId
, tmp_name
));
377 /* We found this uniform name in the program - go ahead and send the data */
378 GL_EXTCALL(glUniform1ivARB(tmp_loc
, 1, values
));
379 checkGLcall("glUniform1ivARB");
381 ptr
= list_next(&This
->baseShader
.constantsB
, ptr
);
388 * Loads the app-supplied constants into the currently set GLSL program.
390 static void shader_glsl_load_constants(
391 IWineD3DDevice
* device
,
393 char useVertexShader
) {
395 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) device
;
396 const struct shader_glsl_priv
*priv
= (struct shader_glsl_priv
*)deviceImpl
->shader_priv
;
397 IWineD3DStateBlockImpl
* stateBlock
= deviceImpl
->stateBlock
;
398 const WineD3D_GL_Info
*gl_info
= &deviceImpl
->adapter
->gl_info
;
400 const GLhandleARB
*constant_locations
;
401 const struct list
*constant_list
;
402 GLhandleARB programId
;
403 const struct glsl_shader_prog_link
*prog
= priv
->glsl_program
;
407 /* No GLSL program set - nothing to do. */
410 programId
= prog
->programId
;
412 if (useVertexShader
) {
413 IWineD3DBaseShaderImpl
* vshader
= (IWineD3DBaseShaderImpl
*) stateBlock
->vertexShader
;
415 constant_locations
= prog
->vuniformF_locations
;
416 constant_list
= &stateBlock
->set_vconstantsF
;
418 /* Load DirectX 9 float constants/uniforms for vertex shader */
419 shader_glsl_load_constantsF(vshader
, gl_info
, GL_LIMITS(vshader_constantsF
),
420 stateBlock
->vertexShaderConstantF
, constant_locations
, constant_list
);
422 /* Load DirectX 9 integer constants/uniforms for vertex shader */
423 shader_glsl_load_constantsI(vshader
, gl_info
, programId
,
424 prog
->vuniformI_locations
, MAX_CONST_I
,
425 stateBlock
->vertexShaderConstantI
,
426 stateBlock
->changed
.vertexShaderConstantsI
);
428 /* Load DirectX 9 boolean constants/uniforms for vertex shader */
429 shader_glsl_load_constantsB(vshader
, gl_info
, programId
, MAX_CONST_B
,
430 stateBlock
->vertexShaderConstantB
,
431 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 shader_glsl_load_constantsI(pshader
, gl_info
, programId
,
451 prog
->puniformI_locations
, MAX_CONST_I
,
452 stateBlock
->pixelShaderConstantI
,
453 stateBlock
->changed
.pixelShaderConstantsI
);
455 /* Load DirectX 9 boolean constants/uniforms for pixel shader */
456 shader_glsl_load_constantsB(pshader
, gl_info
, programId
, MAX_CONST_B
,
457 stateBlock
->pixelShaderConstantB
,
458 stateBlock
->changed
.pixelShaderConstantsB
);
460 /* Upload the environment bump map matrix if needed. The needsbumpmat member specifies the texture stage to load the matrix from.
461 * It can't be 0 for a valid texbem instruction.
463 for(i
= 0; i
< ((IWineD3DPixelShaderImpl
*) pshader
)->numbumpenvmatconsts
; i
++) {
464 IWineD3DPixelShaderImpl
*ps
= (IWineD3DPixelShaderImpl
*) pshader
;
465 int stage
= ps
->luminanceconst
[i
].texunit
;
467 const float *data
= (const float *)&stateBlock
->textureState
[(int)ps
->bumpenvmatconst
[i
].texunit
][WINED3DTSS_BUMPENVMAT00
];
468 GL_EXTCALL(glUniformMatrix2fvARB(prog
->bumpenvmat_location
[i
], 1, 0, data
));
469 checkGLcall("glUniformMatrix2fvARB");
471 /* texbeml needs the luminance scale and offset too. If texbeml is used, needsbumpmat
472 * is set too, so we can check that in the needsbumpmat check
474 if(ps
->baseShader
.reg_maps
.luminanceparams
[stage
]) {
475 const GLfloat
*scale
= (const GLfloat
*)&stateBlock
->textureState
[stage
][WINED3DTSS_BUMPENVLSCALE
];
476 const GLfloat
*offset
= (const GLfloat
*)&stateBlock
->textureState
[stage
][WINED3DTSS_BUMPENVLOFFSET
];
478 GL_EXTCALL(glUniform1fvARB(prog
->luminancescale_location
[i
], 1, scale
));
479 checkGLcall("glUniform1fvARB");
480 GL_EXTCALL(glUniform1fvARB(prog
->luminanceoffset_location
[i
], 1, offset
));
481 checkGLcall("glUniform1fvARB");
485 if(((IWineD3DPixelShaderImpl
*) pshader
)->vpos_uniform
) {
486 float correction_params
[4];
487 if(deviceImpl
->render_offscreen
) {
488 correction_params
[0] = 0.0;
489 correction_params
[1] = 1.0;
491 /* position is window relative, not viewport relative */
492 correction_params
[0] = ((IWineD3DSurfaceImpl
*) deviceImpl
->render_targets
[0])->currentDesc
.Height
;
493 correction_params
[1] = -1.0;
495 GL_EXTCALL(glUniform4fvARB(prog
->ycorrection_location
, 1, correction_params
));
500 /** Generate the variable & register declarations for the GLSL output target */
501 static void shader_generate_glsl_declarations(IWineD3DBaseShader
*iface
, const shader_reg_maps
*reg_maps
,
502 SHADER_BUFFER
*buffer
, const WineD3D_GL_Info
*gl_info
)
504 IWineD3DBaseShaderImpl
* This
= (IWineD3DBaseShaderImpl
*) iface
;
505 IWineD3DDeviceImpl
*device
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
506 unsigned int i
, extra_constants_needed
= 0;
507 const local_constant
*lconst
;
509 /* There are some minor differences between pixel and vertex shaders */
510 char pshader
= shader_is_pshader_version(This
->baseShader
.hex_version
);
511 char prefix
= pshader
? 'P' : 'V';
513 /* Prototype the subroutines */
514 for (i
= 0; i
< This
->baseShader
.limits
.label
; i
++) {
515 if (reg_maps
->labels
[i
])
516 shader_addline(buffer
, "void subroutine%u();\n", i
);
519 /* Declare the constants (aka uniforms) */
520 if (This
->baseShader
.limits
.constant_float
> 0) {
521 unsigned max_constantsF
= min(This
->baseShader
.limits
.constant_float
,
522 (pshader
? GL_LIMITS(pshader_constantsF
) : GL_LIMITS(vshader_constantsF
)));
523 shader_addline(buffer
, "uniform vec4 %cC[%u];\n", prefix
, max_constantsF
);
526 if (This
->baseShader
.limits
.constant_int
> 0)
527 shader_addline(buffer
, "uniform ivec4 %cI[%u];\n", prefix
, This
->baseShader
.limits
.constant_int
);
529 if (This
->baseShader
.limits
.constant_bool
> 0)
530 shader_addline(buffer
, "uniform bool %cB[%u];\n", prefix
, This
->baseShader
.limits
.constant_bool
);
533 shader_addline(buffer
, "uniform vec4 posFixup;\n");
534 /* Predeclaration; This function is added at link time based on the pixel shader.
535 * VS 3.0 shaders have an array OUT[] the shader writes to, earlier versions don't have
536 * that. We know the input to the reorder function at vertex shader compile time, so
537 * we can deal with that. The reorder function for a 1.x and 2.x vertex shader can just
538 * read gl_FrontColor. The output depends on the pixel shader. The reorder function for a
539 * 1.x and 2.x pshader or for fixed function will write gl_FrontColor, and for a 3.0 shader
540 * it will write to the varying array. Here we depend on the shader optimizer on sorting that
541 * out. The nvidia driver only does that if the parameter is inout instead of out, hence the
544 if(This
->baseShader
.hex_version
>= WINED3DVS_VERSION(3, 0)) {
545 shader_addline(buffer
, "void order_ps_input(in vec4[%u]);\n", MAX_REG_OUTPUT
);
547 shader_addline(buffer
, "void order_ps_input();\n");
550 IWineD3DPixelShaderImpl
*ps_impl
= (IWineD3DPixelShaderImpl
*) This
;
552 ps_impl
->numbumpenvmatconsts
= 0;
553 for(i
= 0; i
< (sizeof(reg_maps
->bumpmat
) / sizeof(reg_maps
->bumpmat
[0])); i
++) {
554 if(!reg_maps
->bumpmat
[i
]) {
558 ps_impl
->bumpenvmatconst
[(int) ps_impl
->numbumpenvmatconsts
].texunit
= i
;
559 shader_addline(buffer
, "uniform mat2 bumpenvmat%d;\n", i
);
561 if(reg_maps
->luminanceparams
) {
562 ps_impl
->luminanceconst
[(int) ps_impl
->numbumpenvmatconsts
].texunit
= i
;
563 shader_addline(buffer
, "uniform float luminancescale%d;\n", i
);
564 shader_addline(buffer
, "uniform float luminanceoffset%d;\n", i
);
565 extra_constants_needed
++;
567 ps_impl
->luminanceconst
[(int) ps_impl
->numbumpenvmatconsts
].texunit
= -1;
570 extra_constants_needed
++;
571 ps_impl
->numbumpenvmatconsts
++;
574 if(device
->stateBlock
->renderState
[WINED3DRS_SRGBWRITEENABLE
]) {
575 shader_addline(buffer
, "const vec4 srgb_mul_low = vec4(%f, %f, %f, %f);\n",
576 srgb_mul_low
, srgb_mul_low
, srgb_mul_low
, srgb_mul_low
);
577 shader_addline(buffer
, "const vec4 srgb_comparison = vec4(%f, %f, %f, %f);\n",
578 srgb_cmp
, srgb_cmp
, srgb_cmp
, srgb_cmp
);
580 if(reg_maps
->vpos
|| reg_maps
->usesdsy
) {
581 if(This
->baseShader
.limits
.constant_float
+ extra_constants_needed
+ 1 < GL_LIMITS(pshader_constantsF
)) {
582 shader_addline(buffer
, "uniform vec4 ycorrection;\n");
583 ((IWineD3DPixelShaderImpl
*) This
)->vpos_uniform
= 1;
584 extra_constants_needed
++;
586 /* This happens because we do not have proper tracking of the constant registers that are
587 * actually used, only the max limit of the shader version
589 FIXME("Cannot find a free uniform for vpos correction params\n");
590 shader_addline(buffer
, "const vec4 ycorrection = vec4(%f, %f, 0.0, 0.0);\n",
591 device
->render_offscreen
? 0.0 : ((IWineD3DSurfaceImpl
*) device
->render_targets
[0])->currentDesc
.Height
,
592 device
->render_offscreen
? 1.0 : -1.0);
594 shader_addline(buffer
, "vec4 vpos;\n");
598 /* Declare texture samplers */
599 for (i
= 0; i
< This
->baseShader
.limits
.sampler
; i
++) {
600 if (reg_maps
->samplers
[i
]) {
602 DWORD stype
= reg_maps
->samplers
[i
] & WINED3DSP_TEXTURETYPE_MASK
;
606 shader_addline(buffer
, "uniform sampler1D %csampler%u;\n", prefix
, i
);
609 if(device
->stateBlock
->textures
[i
] &&
610 IWineD3DBaseTexture_GetTextureDimensions(device
->stateBlock
->textures
[i
]) == GL_TEXTURE_RECTANGLE_ARB
) {
611 shader_addline(buffer
, "uniform sampler2DRect %csampler%u;\n", prefix
, i
);
613 shader_addline(buffer
, "uniform sampler2D %csampler%u;\n", prefix
, i
);
616 case WINED3DSTT_CUBE
:
617 shader_addline(buffer
, "uniform samplerCube %csampler%u;\n", prefix
, i
);
619 case WINED3DSTT_VOLUME
:
620 shader_addline(buffer
, "uniform sampler3D %csampler%u;\n", prefix
, i
);
623 shader_addline(buffer
, "uniform unsupported_sampler %csampler%u;\n", prefix
, i
);
624 FIXME("Unrecognized sampler type: %#x\n", stype
);
630 /* Declare address variables */
631 for (i
= 0; i
< This
->baseShader
.limits
.address
; i
++) {
632 if (reg_maps
->address
[i
])
633 shader_addline(buffer
, "ivec4 A%d;\n", i
);
636 /* Declare texture coordinate temporaries and initialize them */
637 for (i
= 0; i
< This
->baseShader
.limits
.texcoord
; i
++) {
638 if (reg_maps
->texcoord
[i
])
639 shader_addline(buffer
, "vec4 T%u = gl_TexCoord[%u];\n", i
, i
);
642 /* Declare input register varyings. Only pixel shader, vertex shaders have that declared in the
643 * helper function shader that is linked in at link time
645 if(pshader
&& This
->baseShader
.hex_version
>= WINED3DPS_VERSION(3, 0)) {
647 shader_addline(buffer
, "varying vec4 IN[%u];\n", GL_LIMITS(glsl_varyings
) / 4);
649 /* TODO: Write a replacement shader for the fixed function vertex pipeline, so this isn't needed.
650 * For fixed function vertex processing + 3.0 pixel shader we need a separate function in the
651 * pixel shader that reads the fixed function color into the packed input registers.
653 shader_addline(buffer
, "vec4 IN[%u];\n", GL_LIMITS(glsl_varyings
) / 4);
657 /* Declare output register temporaries */
658 if(This
->baseShader
.limits
.packed_output
) {
659 shader_addline(buffer
, "vec4 OUT[%u];\n", This
->baseShader
.limits
.packed_output
);
662 /* Declare temporary variables */
663 for(i
= 0; i
< This
->baseShader
.limits
.temporary
; i
++) {
664 if (reg_maps
->temporary
[i
])
665 shader_addline(buffer
, "vec4 R%u;\n", i
);
668 /* Declare attributes */
669 for (i
= 0; i
< This
->baseShader
.limits
.attributes
; i
++) {
670 if (reg_maps
->attributes
[i
])
671 shader_addline(buffer
, "attribute vec4 attrib%i;\n", i
);
674 /* Declare loop registers aLx */
675 for (i
= 0; i
< reg_maps
->loop_depth
; i
++) {
676 shader_addline(buffer
, "int aL%u;\n", i
);
677 shader_addline(buffer
, "int tmpInt%u;\n", i
);
680 /* Temporary variables for matrix operations */
681 shader_addline(buffer
, "vec4 tmp0;\n");
682 shader_addline(buffer
, "vec4 tmp1;\n");
684 /* Local constants use a different name so they can be loaded once at shader link time
685 * They can't be hardcoded into the shader text via LC = {x, y, z, w}; because the
686 * float -> string conversion can cause precision loss.
688 if(!This
->baseShader
.load_local_constsF
) {
689 LIST_FOR_EACH_ENTRY(lconst
, &This
->baseShader
.constantsF
, local_constant
, entry
) {
690 shader_addline(buffer
, "uniform vec4 %cLC%u;\n", prefix
, lconst
->idx
);
694 /* Start the main program */
695 shader_addline(buffer
, "void main() {\n");
696 if(pshader
&& reg_maps
->vpos
) {
697 /* DirectX apps expect integer values, while OpenGL drivers add approximately 0.5. This causes
698 * off-by-one problems as spotted by the vPos d3d9 visual test. Unfortunately the ATI cards do
699 * not add exactly 0.5, but rather something like 0.49999999 or 0.50000001, which still causes
700 * precision troubles when we just substract 0.5.
702 * To deal with that just floor() the position. This will eliminate the fraction on all cards.
704 * TODO: Test how that behaves with multisampling once we can enable multisampling in winex11.
706 * An advantage of floor is that it works even if the driver doesn't add 1/2. It is somewhat
707 * questionable if 1.5, 2.5, ... are the proper values to return in gl_FragCoord, even though
708 * coordinates specify the pixel centers instead of the pixel corners. This code will behave
709 * correctly on drivers that returns integer values.
711 shader_addline(buffer
, "vpos = floor(vec4(0, ycorrection[0], 0, 0) + gl_FragCoord * vec4(1, ycorrection[1], 1, 1));\n");
715 /*****************************************************************************
716 * Functions to generate GLSL strings from DirectX Shader bytecode begin here.
718 * For more information, see http://wiki.winehq.org/DirectX-Shaders
719 ****************************************************************************/
722 static void shader_glsl_add_src_param(const SHADER_OPCODE_ARG
*arg
, const DWORD param
,
723 const DWORD addr_token
, DWORD mask
, glsl_src_param_t
*src_param
);
725 /** Used for opcode modifiers - They multiply the result by the specified amount */
726 static const char * const shift_glsl_tab
[] = {
728 "2.0 * ", /* 1 (x2) */
729 "4.0 * ", /* 2 (x4) */
730 "8.0 * ", /* 3 (x8) */
731 "16.0 * ", /* 4 (x16) */
732 "32.0 * ", /* 5 (x32) */
739 "0.0625 * ", /* 12 (d16) */
740 "0.125 * ", /* 13 (d8) */
741 "0.25 * ", /* 14 (d4) */
742 "0.5 * " /* 15 (d2) */
745 /* Generate a GLSL parameter that does the input modifier computation and return the input register/mask to use */
746 static void shader_glsl_gen_modifier (
749 const char *in_regswizzle
,
754 if (instr
== WINED3DSIO_TEXKILL
)
757 switch (instr
& WINED3DSP_SRCMOD_MASK
) {
758 case WINED3DSPSM_DZ
: /* Need to handle this in the instructions itself (texld & texcrd). */
760 case WINED3DSPSM_NONE
:
761 sprintf(out_str
, "%s%s", in_reg
, in_regswizzle
);
763 case WINED3DSPSM_NEG
:
764 sprintf(out_str
, "-%s%s", in_reg
, in_regswizzle
);
766 case WINED3DSPSM_NOT
:
767 sprintf(out_str
, "!%s%s", in_reg
, in_regswizzle
);
769 case WINED3DSPSM_BIAS
:
770 sprintf(out_str
, "(%s%s - vec4(0.5)%s)", in_reg
, in_regswizzle
, in_regswizzle
);
772 case WINED3DSPSM_BIASNEG
:
773 sprintf(out_str
, "-(%s%s - vec4(0.5)%s)", in_reg
, in_regswizzle
, in_regswizzle
);
775 case WINED3DSPSM_SIGN
:
776 sprintf(out_str
, "(2.0 * (%s%s - 0.5))", in_reg
, in_regswizzle
);
778 case WINED3DSPSM_SIGNNEG
:
779 sprintf(out_str
, "-(2.0 * (%s%s - 0.5))", in_reg
, in_regswizzle
);
781 case WINED3DSPSM_COMP
:
782 sprintf(out_str
, "(1.0 - %s%s)", in_reg
, in_regswizzle
);
785 sprintf(out_str
, "(2.0 * %s%s)", in_reg
, in_regswizzle
);
787 case WINED3DSPSM_X2NEG
:
788 sprintf(out_str
, "-(2.0 * %s%s)", in_reg
, in_regswizzle
);
790 case WINED3DSPSM_ABS
:
791 sprintf(out_str
, "abs(%s%s)", in_reg
, in_regswizzle
);
793 case WINED3DSPSM_ABSNEG
:
794 sprintf(out_str
, "-abs(%s%s)", in_reg
, in_regswizzle
);
797 FIXME("Unhandled modifier %u\n", (instr
& WINED3DSP_SRCMOD_MASK
));
798 sprintf(out_str
, "%s%s", in_reg
, in_regswizzle
);
802 /** Writes the GLSL variable name that corresponds to the register that the
803 * DX opcode parameter is trying to access */
804 static void shader_glsl_get_register_name(const DWORD param
, const DWORD addr_token
,
805 char *regstr
, BOOL
*is_color
, const SHADER_OPCODE_ARG
*arg
)
807 /* oPos, oFog and oPts in D3D */
808 static const char * const hwrastout_reg_names
[] = { "gl_Position", "gl_FogFragCoord", "gl_PointSize" };
810 DWORD reg
= param
& WINED3DSP_REGNUM_MASK
;
811 DWORD regtype
= shader_get_regtype(param
);
812 IWineD3DBaseShaderImpl
* This
= (IWineD3DBaseShaderImpl
*) arg
->shader
;
813 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
814 const WineD3D_GL_Info
* gl_info
= &deviceImpl
->adapter
->gl_info
;
816 char pshader
= shader_is_pshader_version(This
->baseShader
.hex_version
);
822 case WINED3DSPR_TEMP
:
823 sprintf(tmpStr
, "R%u", reg
);
825 case WINED3DSPR_INPUT
:
827 /* Pixel shaders >= 3.0 */
828 if (WINED3DSHADER_VERSION_MAJOR(This
->baseShader
.hex_version
) >= 3) {
829 DWORD in_count
= GL_LIMITS(glsl_varyings
) / 4;
831 if (param
& WINED3DSHADER_ADDRMODE_RELATIVE
) {
832 glsl_src_param_t rel_param
;
833 shader_glsl_add_src_param(arg
, addr_token
, 0, WINED3DSP_WRITEMASK_0
, &rel_param
);
835 /* Removing a + 0 would be an obvious optimization, but macos doesn't see the NOP
838 if(((IWineD3DPixelShaderImpl
*) This
)->input_reg_map
[reg
]) {
839 if (((IWineD3DPixelShaderImpl
*)This
)->declared_in_count
> in_count
) {
840 sprintf(tmpStr
, "((%s + %u) > %d ? (%s + %u) > %d ? gl_SecondaryColor : gl_Color : IN[%s + %u])",
841 rel_param
.param_str
, ((IWineD3DPixelShaderImpl
*)This
)->input_reg_map
[reg
], in_count
- 1,
842 rel_param
.param_str
, ((IWineD3DPixelShaderImpl
*)This
)->input_reg_map
[reg
], in_count
,
843 rel_param
.param_str
, ((IWineD3DPixelShaderImpl
*)This
)->input_reg_map
[reg
]);
845 sprintf(tmpStr
, "IN[%s + %u]", rel_param
.param_str
, ((IWineD3DPixelShaderImpl
*)This
)->input_reg_map
[reg
]);
848 if (((IWineD3DPixelShaderImpl
*)This
)->declared_in_count
> in_count
) {
849 sprintf(tmpStr
, "((%s) > %d ? (%s) > %d ? gl_SecondaryColor : gl_Color : IN[%s])",
850 rel_param
.param_str
, in_count
- 1,
851 rel_param
.param_str
, in_count
,
852 rel_param
.param_str
);
854 sprintf(tmpStr
, "IN[%s]", rel_param
.param_str
);
858 DWORD idx
= ((IWineD3DPixelShaderImpl
*) This
)->input_reg_map
[reg
];
859 if (idx
== in_count
) {
860 sprintf(tmpStr
, "gl_Color");
861 } else if (idx
== in_count
+ 1) {
862 sprintf(tmpStr
, "gl_SecondaryColor");
864 sprintf(tmpStr
, "IN[%u]", idx
);
869 strcpy(tmpStr
, "gl_Color");
871 strcpy(tmpStr
, "gl_SecondaryColor");
874 if (vshader_input_is_color((IWineD3DVertexShader
*) This
, reg
))
876 sprintf(tmpStr
, "attrib%u", reg
);
879 case WINED3DSPR_CONST
:
881 const char prefix
= pshader
? 'P':'V';
883 /* Relative addressing */
884 if (param
& WINED3DSHADER_ADDRMODE_RELATIVE
) {
886 /* Relative addressing on shaders 2.0+ have a relative address token,
887 * prior to that, it was hard-coded as "A0.x" because there's only 1 register */
888 if (WINED3DSHADER_VERSION_MAJOR(This
->baseShader
.hex_version
) >= 2) {
889 glsl_src_param_t rel_param
;
890 shader_glsl_add_src_param(arg
, addr_token
, 0, WINED3DSP_WRITEMASK_0
, &rel_param
);
892 sprintf(tmpStr
, "%cC[%s + %u]", prefix
, rel_param
.param_str
, reg
);
894 sprintf(tmpStr
, "%cC[%s]", prefix
, rel_param
.param_str
);
898 sprintf(tmpStr
, "%cC[A0.x + %u]", prefix
, reg
);
900 sprintf(tmpStr
, "%cC[A0.x]", prefix
);
905 if(shader_constant_is_local(This
, reg
)) {
906 sprintf(tmpStr
, "%cLC%u", prefix
, reg
);
908 sprintf(tmpStr
, "%cC[%u]", prefix
, reg
);
914 case WINED3DSPR_CONSTINT
:
916 sprintf(tmpStr
, "PI[%u]", reg
);
918 sprintf(tmpStr
, "VI[%u]", reg
);
920 case WINED3DSPR_CONSTBOOL
:
922 sprintf(tmpStr
, "PB[%u]", reg
);
924 sprintf(tmpStr
, "VB[%u]", reg
);
926 case WINED3DSPR_TEXTURE
: /* case WINED3DSPR_ADDR: */
928 sprintf(tmpStr
, "T%u", reg
);
930 sprintf(tmpStr
, "A%u", reg
);
933 case WINED3DSPR_LOOP
:
934 sprintf(tmpStr
, "aL%u", This
->baseShader
.cur_loop_regno
- 1);
936 case WINED3DSPR_SAMPLER
:
938 sprintf(tmpStr
, "Psampler%u", reg
);
940 sprintf(tmpStr
, "Vsampler%u", reg
);
942 case WINED3DSPR_COLOROUT
:
943 if (reg
>= GL_LIMITS(buffers
)) {
944 WARN("Write to render target %u, only %d supported\n", reg
, 4);
946 if (GL_SUPPORT(ARB_DRAW_BUFFERS
)) {
947 sprintf(tmpStr
, "gl_FragData[%u]", reg
);
948 } else { /* On older cards with GLSL support like the GeforceFX there's only one buffer. */
949 sprintf(tmpStr
, "gl_FragColor");
952 case WINED3DSPR_RASTOUT
:
953 sprintf(tmpStr
, "%s", hwrastout_reg_names
[reg
]);
955 case WINED3DSPR_DEPTHOUT
:
956 sprintf(tmpStr
, "gl_FragDepth");
958 case WINED3DSPR_ATTROUT
:
960 sprintf(tmpStr
, "gl_FrontColor");
962 sprintf(tmpStr
, "gl_FrontSecondaryColor");
965 case WINED3DSPR_TEXCRDOUT
:
966 /* Vertex shaders >= 3.0: WINED3DSPR_OUTPUT */
967 if (WINED3DSHADER_VERSION_MAJOR(This
->baseShader
.hex_version
) >= 3)
968 sprintf(tmpStr
, "OUT[%u]", reg
);
970 sprintf(tmpStr
, "gl_TexCoord[%u]", reg
);
972 case WINED3DSPR_MISCTYPE
:
975 sprintf(tmpStr
, "vpos");
976 } else if (reg
== 1){
977 /* Note that gl_FrontFacing is a bool, while vFace is
978 * a float for which the sign determines front/back
980 sprintf(tmpStr
, "(gl_FrontFacing ? 1.0 : -1.0)");
982 FIXME("Unhandled misctype register %d\n", reg
);
983 sprintf(tmpStr
, "unrecognized_register");
987 FIXME("Unhandled register name Type(%d)\n", regtype
);
988 sprintf(tmpStr
, "unrecognized_register");
992 strcat(regstr
, tmpStr
);
995 /* Get the GLSL write mask for the destination register */
996 static DWORD
shader_glsl_get_write_mask(const DWORD param
, char *write_mask
) {
997 char *ptr
= write_mask
;
998 DWORD mask
= param
& WINED3DSP_WRITEMASK_ALL
;
1000 if (shader_is_scalar(param
)) {
1001 mask
= WINED3DSP_WRITEMASK_0
;
1004 if (param
& WINED3DSP_WRITEMASK_0
) *ptr
++ = 'x';
1005 if (param
& WINED3DSP_WRITEMASK_1
) *ptr
++ = 'y';
1006 if (param
& WINED3DSP_WRITEMASK_2
) *ptr
++ = 'z';
1007 if (param
& WINED3DSP_WRITEMASK_3
) *ptr
++ = 'w';
1015 static unsigned int shader_glsl_get_write_mask_size(DWORD write_mask
) {
1016 unsigned int size
= 0;
1018 if (write_mask
& WINED3DSP_WRITEMASK_0
) ++size
;
1019 if (write_mask
& WINED3DSP_WRITEMASK_1
) ++size
;
1020 if (write_mask
& WINED3DSP_WRITEMASK_2
) ++size
;
1021 if (write_mask
& WINED3DSP_WRITEMASK_3
) ++size
;
1026 static void shader_glsl_get_swizzle(const DWORD param
, BOOL fixup
, DWORD mask
, char *swizzle_str
) {
1027 /* For registers of type WINED3DDECLTYPE_D3DCOLOR, data is stored as "bgra",
1028 * but addressed as "rgba". To fix this we need to swap the register's x
1029 * and z components. */
1030 DWORD swizzle
= (param
& WINED3DSP_SWIZZLE_MASK
) >> WINED3DSP_SWIZZLE_SHIFT
;
1031 const char *swizzle_chars
= fixup
? "zyxw" : "xyzw";
1032 char *ptr
= swizzle_str
;
1034 if (!shader_is_scalar(param
)) {
1036 /* swizzle bits fields: wwzzyyxx */
1037 if (mask
& WINED3DSP_WRITEMASK_0
) *ptr
++ = swizzle_chars
[swizzle
& 0x03];
1038 if (mask
& WINED3DSP_WRITEMASK_1
) *ptr
++ = swizzle_chars
[(swizzle
>> 2) & 0x03];
1039 if (mask
& WINED3DSP_WRITEMASK_2
) *ptr
++ = swizzle_chars
[(swizzle
>> 4) & 0x03];
1040 if (mask
& WINED3DSP_WRITEMASK_3
) *ptr
++ = swizzle_chars
[(swizzle
>> 6) & 0x03];
1046 /* From a given parameter token, generate the corresponding GLSL string.
1047 * Also, return the actual register name and swizzle in case the
1048 * caller needs this information as well. */
1049 static void shader_glsl_add_src_param(const SHADER_OPCODE_ARG
*arg
, const DWORD param
,
1050 const DWORD addr_token
, DWORD mask
, glsl_src_param_t
*src_param
)
1052 BOOL is_color
= FALSE
;
1053 char swizzle_str
[6];
1055 src_param
->reg_name
[0] = '\0';
1056 src_param
->param_str
[0] = '\0';
1057 swizzle_str
[0] = '\0';
1059 shader_glsl_get_register_name(param
, addr_token
, src_param
->reg_name
, &is_color
, arg
);
1061 shader_glsl_get_swizzle(param
, is_color
, mask
, swizzle_str
);
1062 shader_glsl_gen_modifier(param
, src_param
->reg_name
, swizzle_str
, src_param
->param_str
);
1065 /* From a given parameter token, generate the corresponding GLSL string.
1066 * Also, return the actual register name and swizzle in case the
1067 * caller needs this information as well. */
1068 static DWORD
shader_glsl_add_dst_param(const SHADER_OPCODE_ARG
* arg
, const DWORD param
,
1069 const DWORD addr_token
, glsl_dst_param_t
*dst_param
)
1071 BOOL is_color
= FALSE
;
1073 dst_param
->mask_str
[0] = '\0';
1074 dst_param
->reg_name
[0] = '\0';
1076 shader_glsl_get_register_name(param
, addr_token
, dst_param
->reg_name
, &is_color
, arg
);
1077 return shader_glsl_get_write_mask(param
, dst_param
->mask_str
);
1080 /* Append the destination part of the instruction to the buffer, return the effective write mask */
1081 static DWORD
shader_glsl_append_dst_ext(SHADER_BUFFER
*buffer
, const SHADER_OPCODE_ARG
*arg
, const DWORD param
)
1083 glsl_dst_param_t dst_param
;
1087 mask
= shader_glsl_add_dst_param(arg
, param
, arg
->dst_addr
, &dst_param
);
1090 shift
= (param
& WINED3DSP_DSTSHIFT_MASK
) >> WINED3DSP_DSTSHIFT_SHIFT
;
1091 shader_addline(buffer
, "%s%s = %s(", dst_param
.reg_name
, dst_param
.mask_str
, shift_glsl_tab
[shift
]);
1097 /* Append the destination part of the instruction to the buffer, return the effective write mask */
1098 static DWORD
shader_glsl_append_dst(SHADER_BUFFER
*buffer
, const SHADER_OPCODE_ARG
*arg
)
1100 return shader_glsl_append_dst_ext(buffer
, arg
, arg
->dst
);
1103 /** Process GLSL instruction modifiers */
1104 void shader_glsl_add_instruction_modifiers(const SHADER_OPCODE_ARG
* arg
)
1106 DWORD mask
= arg
->dst
& WINED3DSP_DSTMOD_MASK
;
1108 if (arg
->opcode
->dst_token
&& mask
!= 0) {
1109 glsl_dst_param_t dst_param
;
1111 shader_glsl_add_dst_param(arg
, arg
->dst
, 0, &dst_param
);
1113 if (mask
& WINED3DSPDM_SATURATE
) {
1114 /* _SAT means to clamp the value of the register to between 0 and 1 */
1115 shader_addline(arg
->buffer
, "%s%s = clamp(%s%s, 0.0, 1.0);\n", dst_param
.reg_name
,
1116 dst_param
.mask_str
, dst_param
.reg_name
, dst_param
.mask_str
);
1118 if (mask
& WINED3DSPDM_MSAMPCENTROID
) {
1119 FIXME("_centroid modifier not handled\n");
1121 if (mask
& WINED3DSPDM_PARTIALPRECISION
) {
1122 /* MSDN says this modifier can be safely ignored, so that's what we'll do. */
1127 static inline const char* shader_get_comp_op(
1128 const DWORD opcode
) {
1130 DWORD op
= (opcode
& INST_CONTROLS_MASK
) >> INST_CONTROLS_SHIFT
;
1132 case COMPARISON_GT
: return ">";
1133 case COMPARISON_EQ
: return "==";
1134 case COMPARISON_GE
: return ">=";
1135 case COMPARISON_LT
: return "<";
1136 case COMPARISON_NE
: return "!=";
1137 case COMPARISON_LE
: return "<=";
1139 FIXME("Unrecognized comparison value: %u\n", op
);
1144 static void shader_glsl_get_sample_function(DWORD sampler_type
, BOOL projected
, BOOL texrect
, glsl_sample_function_t
*sample_function
) {
1145 /* Note that there's no such thing as a projected cube texture. */
1146 switch(sampler_type
) {
1148 sample_function
->name
= projected
? "texture1DProj" : "texture1D";
1149 sample_function
->coord_mask
= WINED3DSP_WRITEMASK_0
;
1153 sample_function
->name
= projected
? "texture2DRectProj" : "texture2DRect";
1155 sample_function
->name
= projected
? "texture2DProj" : "texture2D";
1157 sample_function
->coord_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
;
1159 case WINED3DSTT_CUBE
:
1160 sample_function
->name
= "textureCube";
1161 sample_function
->coord_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
1163 case WINED3DSTT_VOLUME
:
1164 sample_function
->name
= projected
? "texture3DProj" : "texture3D";
1165 sample_function
->coord_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
1168 sample_function
->name
= "";
1169 sample_function
->coord_mask
= 0;
1170 FIXME("Unrecognized sampler type: %#x;\n", sampler_type
);
1175 static void shader_glsl_color_correction(const SHADER_OPCODE_ARG
*arg
)
1177 IWineD3DBaseShaderImpl
* shader
= (IWineD3DBaseShaderImpl
*) arg
->shader
;
1178 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) shader
->baseShader
.device
;
1179 const WineD3D_GL_Info
*gl_info
= &deviceImpl
->adapter
->gl_info
;
1180 glsl_dst_param_t dst_param
;
1181 glsl_dst_param_t dst_param2
;
1183 WINED3DFORMAT conversion_group
;
1184 IWineD3DBaseTextureImpl
*texture
;
1185 DWORD mask
, mask_size
;
1187 BOOL recorded
= FALSE
;
1189 DWORD hex_version
= shader
->baseShader
.hex_version
;
1191 switch(arg
->opcode
->opcode
) {
1192 case WINED3DSIO_TEX
:
1193 if (hex_version
< WINED3DPS_VERSION(2,0)) {
1194 sampler_idx
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
1196 sampler_idx
= arg
->src
[1] & WINED3DSP_REGNUM_MASK
;
1200 case WINED3DSIO_TEXLDL
:
1201 FIXME("Add color fixup for vertex texture WINED3DSIO_TEXLDL\n");
1204 case WINED3DSIO_TEXDP3TEX
:
1205 case WINED3DSIO_TEXM3x3TEX
:
1206 case WINED3DSIO_TEXM3x3SPEC
:
1207 case WINED3DSIO_TEXM3x3VSPEC
:
1208 case WINED3DSIO_TEXBEM
:
1209 case WINED3DSIO_TEXREG2AR
:
1210 case WINED3DSIO_TEXREG2GB
:
1211 case WINED3DSIO_TEXREG2RGB
:
1212 sampler_idx
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
1216 /* Not a texture sampling instruction, nothing to do */
1220 texture
= (IWineD3DBaseTextureImpl
*) deviceImpl
->stateBlock
->textures
[sampler_idx
];
1222 fmt
= texture
->resource
.format
;
1223 conversion_group
= texture
->baseTexture
.shader_conversion_group
;
1225 fmt
= WINED3DFMT_UNKNOWN
;
1226 conversion_group
= WINED3DFMT_UNKNOWN
;
1229 /* before doing anything, record the sampler with the format in the format conversion list,
1230 * but check if it's not there already
1232 for(i
= 0; i
< shader
->baseShader
.num_sampled_samplers
; i
++) {
1233 if(shader
->baseShader
.sampled_samplers
[i
] == sampler_idx
) {
1239 shader
->baseShader
.sampled_samplers
[shader
->baseShader
.num_sampled_samplers
] = sampler_idx
;
1240 shader
->baseShader
.num_sampled_samplers
++;
1241 shader
->baseShader
.sampled_format
[sampler_idx
] = conversion_group
;
1245 case WINED3DFMT_V8U8
:
1246 case WINED3DFMT_V16U16
:
1247 if(GL_SUPPORT(NV_TEXTURE_SHADER
) && fmt
== WINED3DFMT_V8U8
) {
1248 /* The 3rd channel returns 1.0 in d3d, but 0.0 in gl. Fix this while we're at it :-) */
1249 mask
= shader_glsl_add_dst_param(arg
, arg
->dst
, WINED3DSP_WRITEMASK_2
, &dst_param
);
1250 mask_size
= shader_glsl_get_write_mask_size(mask
);
1251 if(mask_size
>= 3) {
1252 shader_addline(arg
->buffer
, "%s.%c = 1.0;\n", dst_param
.reg_name
, dst_param
.mask_str
[3]);
1255 /* Correct the sign, but leave the blue as it is - it was loaded correctly already */
1256 mask
= shader_glsl_add_dst_param(arg
, arg
->dst
,
1257 WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
,
1259 mask_size
= shader_glsl_get_write_mask_size(mask
);
1260 if(mask_size
>= 2) {
1261 shader_addline(arg
->buffer
, "%s.%c%c = %s.%c%c * 2.0 - 1.0;\n",
1262 dst_param
.reg_name
, dst_param
.mask_str
[1], dst_param
.mask_str
[2],
1263 dst_param
.reg_name
, dst_param
.mask_str
[1], dst_param
.mask_str
[2]);
1264 } else if(mask_size
== 1) {
1265 shader_addline(arg
->buffer
, "%s.%c = %s.%c * 2.0 - 1.0;\n", dst_param
.reg_name
, dst_param
.mask_str
[1],
1266 dst_param
.reg_name
, dst_param
.mask_str
[1]);
1271 case WINED3DFMT_X8L8V8U8
:
1272 if(!GL_SUPPORT(NV_TEXTURE_SHADER
)) {
1273 /* Red and blue are the signed channels, fix them up; Blue(=L) is correct already,
1274 * and a(X) is always 1.0
1276 mask
= shader_glsl_add_dst_param(arg
, arg
->dst
, WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
, &dst_param
);
1277 mask_size
= shader_glsl_get_write_mask_size(mask
);
1278 if(mask_size
>= 2) {
1279 shader_addline(arg
->buffer
, "%s.%c%c = %s.%c%c * 2.0 - 1.0;\n",
1280 dst_param
.reg_name
, dst_param
.mask_str
[1], dst_param
.mask_str
[2],
1281 dst_param
.reg_name
, dst_param
.mask_str
[1], dst_param
.mask_str
[2]);
1282 } else if(mask_size
== 1) {
1283 shader_addline(arg
->buffer
, "%s.%c = %s.%c * 2.0 - 1.0;\n",
1284 dst_param
.reg_name
, dst_param
.mask_str
[1],
1285 dst_param
.reg_name
, dst_param
.mask_str
[1]);
1290 case WINED3DFMT_L6V5U5
:
1291 if(!GL_SUPPORT(NV_TEXTURE_SHADER
)) {
1292 mask
= shader_glsl_add_dst_param(arg
, arg
->dst
, WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
, &dst_param
);
1293 mask_size
= shader_glsl_get_write_mask_size(mask
);
1294 shader_glsl_add_dst_param(arg
, arg
->dst
, WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_2
, &dst_param2
);
1295 if(mask_size
>= 3) {
1296 /* Swap y and z (U and L), and do a sign conversion on x and the new y(V and U) */
1297 shader_addline(arg
->buffer
, "tmp0.g = %s.%c;\n",
1298 dst_param
.reg_name
, dst_param
.mask_str
[2]);
1299 shader_addline(arg
->buffer
, "%s.%c%c = %s.%c%c * 2.0 - 1.0;\n",
1300 dst_param
.reg_name
, dst_param
.mask_str
[2], dst_param
.mask_str
[1],
1301 dst_param2
.reg_name
, dst_param
.mask_str
[1], dst_param
.mask_str
[3]);
1302 shader_addline(arg
->buffer
, "%s.%c = tmp0.g;\n", dst_param
.reg_name
,
1303 dst_param
.mask_str
[3]);
1304 } else if(mask_size
== 2) {
1305 /* This is bad: We have VL, but we need VU */
1306 FIXME("2 components sampled from a converted L6V5U5 texture\n");
1308 shader_addline(arg
->buffer
, "%s.%c = %s.%c * 2.0 - 1.0;\n",
1309 dst_param
.reg_name
, dst_param
.mask_str
[1],
1310 dst_param2
.reg_name
, dst_param
.mask_str
[1]);
1315 case WINED3DFMT_Q8W8V8U8
:
1316 if(!GL_SUPPORT(NV_TEXTURE_SHADER
)) {
1317 /* Correct the sign in all channels. The writemask just applies as-is, no
1318 * need for checking the mask size
1320 shader_glsl_add_dst_param(arg
, arg
->dst
,
1321 WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
|
1322 WINED3DSP_WRITEMASK_2
| WINED3DSP_WRITEMASK_3
,
1324 shader_addline(arg
->buffer
, "%s%s = %s%s * 2.0 - 1.0;\n", dst_param
.reg_name
, dst_param
.mask_str
,
1325 dst_param
.reg_name
, dst_param
.mask_str
);
1329 case WINED3DFMT_ATI2N
:
1330 /* GL_ATI_texture_compression_3dc returns the two channels as luminance-alpha,
1331 * which means the first one is replicated across .rgb, and the 2nd one is in
1332 * .a. We need the 2nd in .g
1334 * GL_EXT_texture_compression_rgtc returns the values in .rg, however, they
1335 * are swapped compared to d3d. So swap red and green.
1337 mask
= shader_glsl_add_dst_param(arg
, arg
->dst
, WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
, &dst_param
);
1338 mask_size
= shader_glsl_get_write_mask_size(mask
);
1339 if(GL_SUPPORT(EXT_TEXTURE_COMPRESSION_RGTC
)) {
1340 if(mask_size
>= 2) {
1341 shader_addline(arg
->buffer
, "%s.%c%c = %s.%c%c;\n",
1342 dst_param
.reg_name
, dst_param
.mask_str
[1],
1343 dst_param
.mask_str
[2],
1344 dst_param
.reg_name
, dst_param
.mask_str
[2],
1345 dst_param
.mask_str
[1]);
1347 FIXME("%u components sampled from a converted ATI2N texture\n", mask_size
);
1350 if(mask_size
== 4) {
1351 /* Swap y and z (U and L), and do a sign conversion on x and the new y(V and U) */
1352 shader_addline(arg
->buffer
, "%s.%c = %s.%c;\n",
1353 dst_param
.reg_name
, dst_param
.mask_str
[2],
1354 dst_param
.reg_name
, dst_param
.mask_str
[4]);
1355 } else if(mask_size
== 1) {
1358 FIXME("%u components sampled from a converted ATI2N texture\n", mask_size
);
1359 /* This is bad: We have .r[gb], but we need .ra */
1364 /* stupid compiler */
1370 /*****************************************************************************
1372 * Begin processing individual instruction opcodes
1374 ****************************************************************************/
1376 /* Generate GLSL arithmetic functions (dst = src1 + src2) */
1377 static void shader_glsl_arith(const SHADER_OPCODE_ARG
*arg
)
1379 CONST SHADER_OPCODE
* curOpcode
= arg
->opcode
;
1380 SHADER_BUFFER
* buffer
= arg
->buffer
;
1381 glsl_src_param_t src0_param
;
1382 glsl_src_param_t src1_param
;
1386 /* Determine the GLSL operator to use based on the opcode */
1387 switch (curOpcode
->opcode
) {
1388 case WINED3DSIO_MUL
: op
= '*'; break;
1389 case WINED3DSIO_ADD
: op
= '+'; break;
1390 case WINED3DSIO_SUB
: op
= '-'; break;
1393 FIXME("Opcode %s not yet handled in GLSL\n", curOpcode
->name
);
1397 write_mask
= shader_glsl_append_dst(buffer
, arg
);
1398 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], write_mask
, &src0_param
);
1399 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], write_mask
, &src1_param
);
1400 shader_addline(buffer
, "%s %c %s);\n", src0_param
.param_str
, op
, src1_param
.param_str
);
1403 /* Process the WINED3DSIO_MOV opcode using GLSL (dst = src) */
1404 static void shader_glsl_mov(const SHADER_OPCODE_ARG
*arg
)
1406 IWineD3DBaseShaderImpl
* shader
= (IWineD3DBaseShaderImpl
*) arg
->shader
;
1407 SHADER_BUFFER
* buffer
= arg
->buffer
;
1408 glsl_src_param_t src0_param
;
1411 write_mask
= shader_glsl_append_dst(buffer
, arg
);
1412 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], write_mask
, &src0_param
);
1414 /* In vs_1_1 WINED3DSIO_MOV can write to the address register. In later
1415 * shader versions WINED3DSIO_MOVA is used for this. */
1416 if ((WINED3DSHADER_VERSION_MAJOR(shader
->baseShader
.hex_version
) == 1 &&
1417 !shader_is_pshader_version(shader
->baseShader
.hex_version
) &&
1418 shader_get_regtype(arg
->dst
) == WINED3DSPR_ADDR
)) {
1419 /* This is a simple floor() */
1420 unsigned int mask_size
= shader_glsl_get_write_mask_size(write_mask
);
1421 if (mask_size
> 1) {
1422 shader_addline(buffer
, "ivec%d(floor(%s)));\n", mask_size
, src0_param
.param_str
);
1424 shader_addline(buffer
, "int(floor(%s)));\n", src0_param
.param_str
);
1426 } else if(arg
->opcode
->opcode
== WINED3DSIO_MOVA
) {
1427 /* We need to *round* to the nearest int here. */
1428 unsigned int mask_size
= shader_glsl_get_write_mask_size(write_mask
);
1429 if (mask_size
> 1) {
1430 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
);
1432 shader_addline(buffer
, "int(floor(abs(%s) + 0.5) * sign(%s)));\n", src0_param
.param_str
, src0_param
.param_str
);
1435 shader_addline(buffer
, "%s);\n", src0_param
.param_str
);
1439 /* Process the dot product operators DP3 and DP4 in GLSL (dst = dot(src0, src1)) */
1440 static void shader_glsl_dot(const SHADER_OPCODE_ARG
*arg
)
1442 CONST SHADER_OPCODE
* curOpcode
= arg
->opcode
;
1443 SHADER_BUFFER
* buffer
= arg
->buffer
;
1444 glsl_src_param_t src0_param
;
1445 glsl_src_param_t src1_param
;
1446 DWORD dst_write_mask
, src_write_mask
;
1447 unsigned int dst_size
= 0;
1449 dst_write_mask
= shader_glsl_append_dst(buffer
, arg
);
1450 dst_size
= shader_glsl_get_write_mask_size(dst_write_mask
);
1452 /* dp3 works on vec3, dp4 on vec4 */
1453 if (curOpcode
->opcode
== WINED3DSIO_DP4
) {
1454 src_write_mask
= WINED3DSP_WRITEMASK_ALL
;
1456 src_write_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
1459 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], src_write_mask
, &src0_param
);
1460 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], src_write_mask
, &src1_param
);
1463 shader_addline(buffer
, "vec%d(dot(%s, %s)));\n", dst_size
, src0_param
.param_str
, src1_param
.param_str
);
1465 shader_addline(buffer
, "dot(%s, %s));\n", src0_param
.param_str
, src1_param
.param_str
);
1469 /* Note that this instruction has some restrictions. The destination write mask
1470 * can't contain the w component, and the source swizzles have to be .xyzw */
1471 static void shader_glsl_cross(const SHADER_OPCODE_ARG
*arg
)
1473 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
1474 glsl_src_param_t src0_param
;
1475 glsl_src_param_t src1_param
;
1478 shader_glsl_get_write_mask(arg
->dst
, dst_mask
);
1479 shader_glsl_append_dst(arg
->buffer
, arg
);
1480 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], src_mask
, &src0_param
);
1481 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], src_mask
, &src1_param
);
1482 shader_addline(arg
->buffer
, "cross(%s, %s)%s);\n", src0_param
.param_str
, src1_param
.param_str
, dst_mask
);
1485 /* Process the WINED3DSIO_POW instruction in GLSL (dst = |src0|^src1)
1486 * Src0 and src1 are scalars. Note that D3D uses the absolute of src0, while
1487 * GLSL uses the value as-is. */
1488 static void shader_glsl_pow(const SHADER_OPCODE_ARG
*arg
)
1490 SHADER_BUFFER
*buffer
= arg
->buffer
;
1491 glsl_src_param_t src0_param
;
1492 glsl_src_param_t src1_param
;
1493 DWORD dst_write_mask
;
1494 unsigned int dst_size
;
1496 dst_write_mask
= shader_glsl_append_dst(buffer
, arg
);
1497 dst_size
= shader_glsl_get_write_mask_size(dst_write_mask
);
1499 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
1500 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], WINED3DSP_WRITEMASK_0
, &src1_param
);
1503 shader_addline(buffer
, "vec%d(pow(abs(%s), %s)));\n", dst_size
, src0_param
.param_str
, src1_param
.param_str
);
1505 shader_addline(buffer
, "pow(abs(%s), %s));\n", src0_param
.param_str
, src1_param
.param_str
);
1509 /* Process the WINED3DSIO_LOG instruction in GLSL (dst = log2(|src0|))
1510 * Src0 is a scalar. Note that D3D uses the absolute of src0, while
1511 * GLSL uses the value as-is. */
1512 static void shader_glsl_log(const SHADER_OPCODE_ARG
*arg
)
1514 SHADER_BUFFER
*buffer
= arg
->buffer
;
1515 glsl_src_param_t src0_param
;
1516 DWORD dst_write_mask
;
1517 unsigned int dst_size
;
1519 dst_write_mask
= shader_glsl_append_dst(buffer
, arg
);
1520 dst_size
= shader_glsl_get_write_mask_size(dst_write_mask
);
1522 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
1525 shader_addline(buffer
, "vec%d(log2(abs(%s))));\n", dst_size
, src0_param
.param_str
);
1527 shader_addline(buffer
, "log2(abs(%s)));\n", src0_param
.param_str
);
1531 /* Map the opcode 1-to-1 to the GL code (arg->dst = instruction(src0, src1, ...) */
1532 static void shader_glsl_map2gl(const SHADER_OPCODE_ARG
*arg
)
1534 CONST SHADER_OPCODE
* curOpcode
= arg
->opcode
;
1535 SHADER_BUFFER
* buffer
= arg
->buffer
;
1536 glsl_src_param_t src_param
;
1537 const char *instruction
;
1538 char arguments
[256];
1542 /* Determine the GLSL function to use based on the opcode */
1543 /* TODO: Possibly make this a table for faster lookups */
1544 switch (curOpcode
->opcode
) {
1545 case WINED3DSIO_MIN
: instruction
= "min"; break;
1546 case WINED3DSIO_MAX
: instruction
= "max"; break;
1547 case WINED3DSIO_ABS
: instruction
= "abs"; break;
1548 case WINED3DSIO_FRC
: instruction
= "fract"; break;
1549 case WINED3DSIO_NRM
: instruction
= "normalize"; break;
1550 case WINED3DSIO_EXP
: instruction
= "exp2"; break;
1551 case WINED3DSIO_SGN
: instruction
= "sign"; break;
1552 case WINED3DSIO_DSX
: instruction
= "dFdx"; break;
1553 case WINED3DSIO_DSY
: instruction
= "ycorrection.y * dFdy"; break;
1554 default: instruction
= "";
1555 FIXME("Opcode %s not yet handled in GLSL\n", curOpcode
->name
);
1559 write_mask
= shader_glsl_append_dst(buffer
, arg
);
1561 arguments
[0] = '\0';
1562 if (curOpcode
->num_params
> 0) {
1563 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], write_mask
, &src_param
);
1564 strcat(arguments
, src_param
.param_str
);
1565 for (i
= 2; i
< curOpcode
->num_params
; ++i
) {
1566 strcat(arguments
, ", ");
1567 shader_glsl_add_src_param(arg
, arg
->src
[i
-1], arg
->src_addr
[i
-1], write_mask
, &src_param
);
1568 strcat(arguments
, src_param
.param_str
);
1572 shader_addline(buffer
, "%s(%s));\n", instruction
, arguments
);
1575 /** Process the WINED3DSIO_EXPP instruction in GLSL:
1576 * For shader model 1.x, do the following (and honor the writemask, so use a temporary variable):
1577 * dst.x = 2^(floor(src))
1578 * dst.y = src - floor(src)
1579 * dst.z = 2^src (partial precision is allowed, but optional)
1581 * For 2.0 shaders, just do this (honoring writemask and swizzle):
1582 * dst = 2^src; (partial precision is allowed, but optional)
1584 static void shader_glsl_expp(const SHADER_OPCODE_ARG
*arg
)
1586 IWineD3DBaseShaderImpl
*shader
= (IWineD3DBaseShaderImpl
*)arg
->shader
;
1587 glsl_src_param_t src_param
;
1589 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src_param
);
1591 if (shader
->baseShader
.hex_version
< WINED3DPS_VERSION(2,0)) {
1594 shader_addline(arg
->buffer
, "tmp0.x = exp2(floor(%s));\n", src_param
.param_str
);
1595 shader_addline(arg
->buffer
, "tmp0.y = %s - floor(%s);\n", src_param
.param_str
, src_param
.param_str
);
1596 shader_addline(arg
->buffer
, "tmp0.z = exp2(%s);\n", src_param
.param_str
);
1597 shader_addline(arg
->buffer
, "tmp0.w = 1.0;\n");
1599 shader_glsl_append_dst(arg
->buffer
, arg
);
1600 shader_glsl_get_write_mask(arg
->dst
, dst_mask
);
1601 shader_addline(arg
->buffer
, "tmp0%s);\n", dst_mask
);
1604 unsigned int mask_size
;
1606 write_mask
= shader_glsl_append_dst(arg
->buffer
, arg
);
1607 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
1609 if (mask_size
> 1) {
1610 shader_addline(arg
->buffer
, "vec%d(exp2(%s)));\n", mask_size
, src_param
.param_str
);
1612 shader_addline(arg
->buffer
, "exp2(%s));\n", src_param
.param_str
);
1617 /** Process the RCP (reciprocal or inverse) opcode in GLSL (dst = 1 / src) */
1618 static void shader_glsl_rcp(const SHADER_OPCODE_ARG
*arg
)
1620 glsl_src_param_t src_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], WINED3DSP_WRITEMASK_3
, &src_param
);
1628 if (mask_size
> 1) {
1629 shader_addline(arg
->buffer
, "vec%d(1.0 / %s));\n", mask_size
, src_param
.param_str
);
1631 shader_addline(arg
->buffer
, "1.0 / %s);\n", src_param
.param_str
);
1635 static void shader_glsl_rsq(const SHADER_OPCODE_ARG
*arg
)
1637 SHADER_BUFFER
* buffer
= arg
->buffer
;
1638 glsl_src_param_t src_param
;
1640 unsigned int mask_size
;
1642 write_mask
= shader_glsl_append_dst(buffer
, arg
);
1643 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
1645 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_3
, &src_param
);
1647 if (mask_size
> 1) {
1648 shader_addline(buffer
, "vec%d(inversesqrt(%s)));\n", mask_size
, src_param
.param_str
);
1650 shader_addline(buffer
, "inversesqrt(%s));\n", src_param
.param_str
);
1654 /** Process signed comparison opcodes in GLSL. */
1655 static void shader_glsl_compare(const SHADER_OPCODE_ARG
*arg
)
1657 glsl_src_param_t src0_param
;
1658 glsl_src_param_t src1_param
;
1660 unsigned int mask_size
;
1662 write_mask
= shader_glsl_append_dst(arg
->buffer
, arg
);
1663 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
1664 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], write_mask
, &src0_param
);
1665 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], write_mask
, &src1_param
);
1667 if (mask_size
> 1) {
1668 const char *compare
;
1670 switch(arg
->opcode
->opcode
) {
1671 case WINED3DSIO_SLT
: compare
= "lessThan"; break;
1672 case WINED3DSIO_SGE
: compare
= "greaterThanEqual"; break;
1673 default: compare
= "";
1674 FIXME("Can't handle opcode %s\n", arg
->opcode
->name
);
1677 shader_addline(arg
->buffer
, "vec%d(%s(%s, %s)));\n", mask_size
, compare
,
1678 src0_param
.param_str
, src1_param
.param_str
);
1680 switch(arg
->opcode
->opcode
) {
1681 case WINED3DSIO_SLT
:
1682 /* Step(src0, src1) is not suitable here because if src0 == src1 SLT is supposed,
1683 * to return 0.0 but step returns 1.0 because step is not < x
1684 * An alternative is a bvec compare padded with an unused second component.
1685 * step(src1 * -1.0, src0 * -1.0) is not an option because it suffers from the same
1686 * issue. Playing with not() is not possible either because not() does not accept
1689 shader_addline(arg
->buffer
, "(%s < %s) ? 1.0 : 0.0);\n", src0_param
.param_str
, src1_param
.param_str
);
1691 case WINED3DSIO_SGE
:
1692 /* Here we can use the step() function and safe a conditional */
1693 shader_addline(arg
->buffer
, "step(%s, %s));\n", src1_param
.param_str
, src0_param
.param_str
);
1696 FIXME("Can't handle opcode %s\n", arg
->opcode
->name
);
1702 /** Process CMP instruction in GLSL (dst = src0 >= 0.0 ? src1 : src2), per channel */
1703 static void shader_glsl_cmp(const SHADER_OPCODE_ARG
*arg
)
1705 glsl_src_param_t src0_param
;
1706 glsl_src_param_t src1_param
;
1707 glsl_src_param_t src2_param
;
1708 DWORD write_mask
, cmp_channel
= 0;
1711 BOOL temp_destination
= FALSE
;
1713 if(shader_is_scalar(arg
->src
[0])) {
1714 write_mask
= shader_glsl_append_dst(arg
->buffer
, arg
);
1716 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_ALL
, &src0_param
);
1717 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], write_mask
, &src1_param
);
1718 shader_glsl_add_src_param(arg
, arg
->src
[2], arg
->src_addr
[2], write_mask
, &src2_param
);
1720 shader_addline(arg
->buffer
, "%s >= 0.0 ? %s : %s);\n",
1721 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
1723 DWORD src0reg
= arg
->src
[0] & WINED3DSP_REGNUM_MASK
;
1724 DWORD src1reg
= arg
->src
[1] & WINED3DSP_REGNUM_MASK
;
1725 DWORD src2reg
= arg
->src
[2] & WINED3DSP_REGNUM_MASK
;
1726 DWORD src0regtype
= shader_get_regtype(arg
->src
[0]);
1727 DWORD src1regtype
= shader_get_regtype(arg
->src
[1]);
1728 DWORD src2regtype
= shader_get_regtype(arg
->src
[2]);
1729 DWORD dstreg
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
1730 DWORD dstregtype
= shader_get_regtype(arg
->dst
);
1732 /* Cycle through all source0 channels */
1733 for (i
=0; i
<4; i
++) {
1735 /* Find the destination channels which use the current source0 channel */
1736 for (j
=0; j
<4; j
++) {
1737 if ( ((arg
->src
[0] >> (WINED3DSP_SWIZZLE_SHIFT
+ 2*j
)) & 0x3) == i
) {
1738 write_mask
|= WINED3DSP_WRITEMASK_0
<< j
;
1739 cmp_channel
= WINED3DSP_WRITEMASK_0
<< j
;
1743 /* Splitting the cmp instruction up in multiple lines imposes a problem:
1744 * The first lines may overwrite source parameters of the following lines.
1745 * Deal with that by using a temporary destination register if needed
1747 if((src0reg
== dstreg
&& src0regtype
== dstregtype
) ||
1748 (src1reg
== dstreg
&& src1regtype
== dstregtype
) ||
1749 (src2reg
== dstreg
&& src2regtype
== dstregtype
)) {
1751 write_mask
= shader_glsl_get_write_mask(arg
->dst
& (~WINED3DSP_SWIZZLE_MASK
| write_mask
), mask_char
);
1752 if (!write_mask
) continue;
1753 shader_addline(arg
->buffer
, "tmp0%s = (", mask_char
);
1754 temp_destination
= TRUE
;
1756 write_mask
= shader_glsl_append_dst_ext(arg
->buffer
, arg
, arg
->dst
& (~WINED3DSP_SWIZZLE_MASK
| write_mask
));
1757 if (!write_mask
) continue;
1760 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], cmp_channel
, &src0_param
);
1761 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], write_mask
, &src1_param
);
1762 shader_glsl_add_src_param(arg
, arg
->src
[2], arg
->src_addr
[2], write_mask
, &src2_param
);
1764 shader_addline(arg
->buffer
, "%s >= 0.0 ? %s : %s);\n",
1765 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
1768 if(temp_destination
) {
1769 shader_glsl_get_write_mask(arg
->dst
, mask_char
);
1770 shader_glsl_append_dst_ext(arg
->buffer
, arg
, arg
->dst
);
1771 shader_addline(arg
->buffer
, "tmp0%s);\n", mask_char
);
1777 /** Process the CND opcode in GLSL (dst = (src0 > 0.5) ? src1 : src2) */
1778 /* For ps 1.1-1.3, only a single component of src0 is used. For ps 1.4
1779 * the compare is done per component of src0. */
1780 static void shader_glsl_cnd(const SHADER_OPCODE_ARG
*arg
)
1782 IWineD3DBaseShaderImpl
* shader
= (IWineD3DBaseShaderImpl
*) arg
->shader
;
1783 glsl_src_param_t src0_param
;
1784 glsl_src_param_t src1_param
;
1785 glsl_src_param_t src2_param
;
1786 DWORD write_mask
, cmp_channel
= 0;
1789 if (shader
->baseShader
.hex_version
< WINED3DPS_VERSION(1, 4)) {
1790 write_mask
= shader_glsl_append_dst(arg
->buffer
, arg
);
1791 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
1792 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], write_mask
, &src1_param
);
1793 shader_glsl_add_src_param(arg
, arg
->src
[2], arg
->src_addr
[2], write_mask
, &src2_param
);
1795 /* Fun: The D3DSI_COISSUE flag changes the semantic of the cnd instruction for < 1.4 shaders */
1796 if(arg
->opcode_token
& WINED3DSI_COISSUE
) {
1797 shader_addline(arg
->buffer
, "%s /* COISSUE! */);\n", src1_param
.param_str
);
1799 shader_addline(arg
->buffer
, "%s > 0.5 ? %s : %s);\n",
1800 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
1804 /* Cycle through all source0 channels */
1805 for (i
=0; i
<4; i
++) {
1807 /* Find the destination channels which use the current source0 channel */
1808 for (j
=0; j
<4; j
++) {
1809 if ( ((arg
->src
[0] >> (WINED3DSP_SWIZZLE_SHIFT
+ 2*j
)) & 0x3) == i
) {
1810 write_mask
|= WINED3DSP_WRITEMASK_0
<< j
;
1811 cmp_channel
= WINED3DSP_WRITEMASK_0
<< j
;
1814 write_mask
= shader_glsl_append_dst_ext(arg
->buffer
, arg
, arg
->dst
& (~WINED3DSP_SWIZZLE_MASK
| write_mask
));
1815 if (!write_mask
) continue;
1817 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], cmp_channel
, &src0_param
);
1818 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], write_mask
, &src1_param
);
1819 shader_glsl_add_src_param(arg
, arg
->src
[2], arg
->src_addr
[2], write_mask
, &src2_param
);
1821 shader_addline(arg
->buffer
, "%s > 0.5 ? %s : %s);\n",
1822 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
1826 /** GLSL code generation for WINED3DSIO_MAD: Multiply the first 2 opcodes, then add the last */
1827 static void shader_glsl_mad(const SHADER_OPCODE_ARG
*arg
)
1829 glsl_src_param_t src0_param
;
1830 glsl_src_param_t src1_param
;
1831 glsl_src_param_t src2_param
;
1834 write_mask
= shader_glsl_append_dst(arg
->buffer
, arg
);
1835 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], write_mask
, &src0_param
);
1836 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], write_mask
, &src1_param
);
1837 shader_glsl_add_src_param(arg
, arg
->src
[2], arg
->src_addr
[2], write_mask
, &src2_param
);
1838 shader_addline(arg
->buffer
, "(%s * %s) + %s);\n",
1839 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
1842 /** Handles transforming all WINED3DSIO_M?x? opcodes for
1843 Vertex shaders to GLSL codes */
1844 static void shader_glsl_mnxn(const SHADER_OPCODE_ARG
*arg
)
1847 int nComponents
= 0;
1848 SHADER_OPCODE_ARG tmpArg
;
1850 memset(&tmpArg
, 0, sizeof(SHADER_OPCODE_ARG
));
1852 /* Set constants for the temporary argument */
1853 tmpArg
.shader
= arg
->shader
;
1854 tmpArg
.buffer
= arg
->buffer
;
1855 tmpArg
.src
[0] = arg
->src
[0];
1856 tmpArg
.src_addr
[0] = arg
->src_addr
[0];
1857 tmpArg
.src_addr
[1] = arg
->src_addr
[1];
1858 tmpArg
.reg_maps
= arg
->reg_maps
;
1860 switch(arg
->opcode
->opcode
) {
1861 case WINED3DSIO_M4x4
:
1863 tmpArg
.opcode
= shader_get_opcode(arg
->shader
, WINED3DSIO_DP4
);
1865 case WINED3DSIO_M4x3
:
1867 tmpArg
.opcode
= shader_get_opcode(arg
->shader
, WINED3DSIO_DP4
);
1869 case WINED3DSIO_M3x4
:
1871 tmpArg
.opcode
= shader_get_opcode(arg
->shader
, WINED3DSIO_DP3
);
1873 case WINED3DSIO_M3x3
:
1875 tmpArg
.opcode
= shader_get_opcode(arg
->shader
, WINED3DSIO_DP3
);
1877 case WINED3DSIO_M3x2
:
1879 tmpArg
.opcode
= shader_get_opcode(arg
->shader
, WINED3DSIO_DP3
);
1885 for (i
= 0; i
< nComponents
; i
++) {
1886 tmpArg
.dst
= ((arg
->dst
) & ~WINED3DSP_WRITEMASK_ALL
)|(WINED3DSP_WRITEMASK_0
<<i
);
1887 tmpArg
.src
[1] = arg
->src
[1]+i
;
1888 shader_glsl_dot(&tmpArg
);
1893 The LRP instruction performs a component-wise linear interpolation
1894 between the second and third operands using the first operand as the
1895 blend factor. Equation: (dst = src2 + src0 * (src1 - src2))
1896 This is equivalent to mix(src2, src1, src0);
1898 static void shader_glsl_lrp(const SHADER_OPCODE_ARG
*arg
)
1900 glsl_src_param_t src0_param
;
1901 glsl_src_param_t src1_param
;
1902 glsl_src_param_t src2_param
;
1905 write_mask
= shader_glsl_append_dst(arg
->buffer
, arg
);
1907 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], write_mask
, &src0_param
);
1908 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], write_mask
, &src1_param
);
1909 shader_glsl_add_src_param(arg
, arg
->src
[2], arg
->src_addr
[2], write_mask
, &src2_param
);
1911 shader_addline(arg
->buffer
, "mix(%s, %s, %s));\n",
1912 src2_param
.param_str
, src1_param
.param_str
, src0_param
.param_str
);
1915 /** Process the WINED3DSIO_LIT instruction in GLSL:
1916 * dst.x = dst.w = 1.0
1917 * dst.y = (src0.x > 0) ? src0.x
1918 * dst.z = (src0.x > 0) ? ((src0.y > 0) ? pow(src0.y, src.w) : 0) : 0
1919 * where src.w is clamped at +- 128
1921 static void shader_glsl_lit(const SHADER_OPCODE_ARG
*arg
)
1923 glsl_src_param_t src0_param
;
1924 glsl_src_param_t src1_param
;
1925 glsl_src_param_t src3_param
;
1928 shader_glsl_append_dst(arg
->buffer
, arg
);
1929 shader_glsl_get_write_mask(arg
->dst
, dst_mask
);
1931 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
1932 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_1
, &src1_param
);
1933 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_3
, &src3_param
);
1935 /* The sdk specifies the instruction like this
1937 * if(src.x > 0.0) dst.y = src.x
1939 * if(src.x > 0.0 && src.y > 0.0) dst.z = pow(src.y, power);
1943 * Obviously that has quite a few conditionals in it which we don't like. So the first step is this:
1944 * dst.x = 1.0 ... No further explanation needed
1945 * dst.y = max(src.y, 0.0); ... If x < 0.0, use 0.0, otherwise x. Same as the conditional
1946 * dst.z = x > 0.0 ? pow(max(y, 0.0), p) : 0; ... 0 ^ power is 0, and otherwise we use y anyway
1947 * dst.w = 1.0. ... Nothing fancy.
1949 * So we still have one conditional in there. So do this:
1950 * dst.z = pow(max(0.0, src.y) * step(0.0, src.x), power);
1952 * 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),
1953 * 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.
1954 * if both x and y are > 0, we get pow(y * 1.0, power), as it is supposed to
1956 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",
1957 src0_param
.param_str
, src1_param
.param_str
, src0_param
.param_str
, src3_param
.param_str
, dst_mask
);
1960 /** Process the WINED3DSIO_DST instruction in GLSL:
1962 * dst.y = src0.x * src0.y
1966 static void shader_glsl_dst(const SHADER_OPCODE_ARG
*arg
)
1968 glsl_src_param_t src0y_param
;
1969 glsl_src_param_t src0z_param
;
1970 glsl_src_param_t src1y_param
;
1971 glsl_src_param_t src1w_param
;
1974 shader_glsl_append_dst(arg
->buffer
, arg
);
1975 shader_glsl_get_write_mask(arg
->dst
, dst_mask
);
1977 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_1
, &src0y_param
);
1978 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_2
, &src0z_param
);
1979 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], WINED3DSP_WRITEMASK_1
, &src1y_param
);
1980 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], WINED3DSP_WRITEMASK_3
, &src1w_param
);
1982 shader_addline(arg
->buffer
, "vec4(1.0, %s * %s, %s, %s))%s;\n",
1983 src0y_param
.param_str
, src1y_param
.param_str
, src0z_param
.param_str
, src1w_param
.param_str
, dst_mask
);
1986 /** Process the WINED3DSIO_SINCOS instruction in GLSL:
1987 * VS 2.0 requires that specific cosine and sine constants be passed to this instruction so the hardware
1988 * can handle it. But, these functions are built-in for GLSL, so we can just ignore the last 2 params.
1990 * dst.x = cos(src0.?)
1991 * dst.y = sin(src0.?)
1995 static void shader_glsl_sincos(const SHADER_OPCODE_ARG
*arg
)
1997 glsl_src_param_t src0_param
;
2000 write_mask
= shader_glsl_append_dst(arg
->buffer
, arg
);
2001 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2003 switch (write_mask
) {
2004 case WINED3DSP_WRITEMASK_0
:
2005 shader_addline(arg
->buffer
, "cos(%s));\n", src0_param
.param_str
);
2008 case WINED3DSP_WRITEMASK_1
:
2009 shader_addline(arg
->buffer
, "sin(%s));\n", src0_param
.param_str
);
2012 case (WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
):
2013 shader_addline(arg
->buffer
, "vec2(cos(%s), sin(%s)));\n", src0_param
.param_str
, src0_param
.param_str
);
2017 ERR("Write mask should be .x, .y or .xy\n");
2022 /** Process the WINED3DSIO_LOOP instruction in GLSL:
2023 * Start a for() loop where src1.y is the initial value of aL,
2024 * increment aL by src1.z for a total of src1.x iterations.
2025 * Need to use a temporary variable for this operation.
2027 /* FIXME: I don't think nested loops will work correctly this way. */
2028 static void shader_glsl_loop(const SHADER_OPCODE_ARG
*arg
)
2030 glsl_src_param_t src1_param
;
2031 IWineD3DBaseShaderImpl
* shader
= (IWineD3DBaseShaderImpl
*) arg
->shader
;
2032 DWORD regtype
= shader_get_regtype(arg
->src
[1]);
2033 DWORD reg
= arg
->src
[1] & WINED3DSP_REGNUM_MASK
;
2034 const DWORD
*control_values
= NULL
;
2035 const local_constant
*constant
;
2037 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], WINED3DSP_WRITEMASK_ALL
, &src1_param
);
2039 /* Try to hardcode the loop control parameters if possible. Direct3D 9 class hardware doesn't support real
2040 * varying indexing, but Microsoft designed this feature for Shader model 2.x+. If the loop control is
2041 * known at compile time, the GLSL compiler can unroll the loop, and replace indirect addressing with direct
2044 if(regtype
== WINED3DSPR_CONSTINT
) {
2045 LIST_FOR_EACH_ENTRY(constant
, &shader
->baseShader
.constantsI
, local_constant
, entry
) {
2046 if(constant
->idx
== reg
) {
2047 control_values
= constant
->value
;
2053 if(control_values
) {
2054 if(control_values
[2] > 0) {
2055 shader_addline(arg
->buffer
, "for (aL%u = %d; aL%u < (%d * %d + %d); aL%u += %d) {\n",
2056 shader
->baseShader
.cur_loop_depth
, control_values
[1],
2057 shader
->baseShader
.cur_loop_depth
, control_values
[0], control_values
[2], control_values
[1],
2058 shader
->baseShader
.cur_loop_depth
, control_values
[2]);
2059 } else if(control_values
[2] == 0) {
2060 shader_addline(arg
->buffer
, "for (aL%u = %d, tmpInt%u = 0; tmpInt%u < %d; tmpInt%u++) {\n",
2061 shader
->baseShader
.cur_loop_depth
, control_values
[1], shader
->baseShader
.cur_loop_depth
,
2062 shader
->baseShader
.cur_loop_depth
, control_values
[0],
2063 shader
->baseShader
.cur_loop_depth
);
2065 shader_addline(arg
->buffer
, "for (aL%u = %d; aL%u > (%d * %d + %d); aL%u += %d) {\n",
2066 shader
->baseShader
.cur_loop_depth
, control_values
[1],
2067 shader
->baseShader
.cur_loop_depth
, control_values
[0], control_values
[2], control_values
[1],
2068 shader
->baseShader
.cur_loop_depth
, control_values
[2]);
2071 shader_addline(arg
->buffer
, "for (tmpInt%u = 0, aL%u = %s.y; tmpInt%u < %s.x; tmpInt%u++, aL%u += %s.z) {\n",
2072 shader
->baseShader
.cur_loop_depth
, shader
->baseShader
.cur_loop_regno
,
2073 src1_param
.reg_name
, shader
->baseShader
.cur_loop_depth
, src1_param
.reg_name
,
2074 shader
->baseShader
.cur_loop_depth
, shader
->baseShader
.cur_loop_regno
, src1_param
.reg_name
);
2077 shader
->baseShader
.cur_loop_depth
++;
2078 shader
->baseShader
.cur_loop_regno
++;
2081 static void shader_glsl_end(const SHADER_OPCODE_ARG
*arg
)
2083 IWineD3DBaseShaderImpl
* shader
= (IWineD3DBaseShaderImpl
*) arg
->shader
;
2085 shader_addline(arg
->buffer
, "}\n");
2087 if(arg
->opcode
->opcode
== WINED3DSIO_ENDLOOP
) {
2088 shader
->baseShader
.cur_loop_depth
--;
2089 shader
->baseShader
.cur_loop_regno
--;
2091 if(arg
->opcode
->opcode
== WINED3DSIO_ENDREP
) {
2092 shader
->baseShader
.cur_loop_depth
--;
2096 static void shader_glsl_rep(const SHADER_OPCODE_ARG
*arg
)
2098 IWineD3DBaseShaderImpl
* shader
= (IWineD3DBaseShaderImpl
*) arg
->shader
;
2099 glsl_src_param_t src0_param
;
2101 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2102 shader_addline(arg
->buffer
, "for (tmpInt%d = 0; tmpInt%d < %s; tmpInt%d++) {\n",
2103 shader
->baseShader
.cur_loop_depth
, shader
->baseShader
.cur_loop_depth
,
2104 src0_param
.param_str
, shader
->baseShader
.cur_loop_depth
);
2105 shader
->baseShader
.cur_loop_depth
++;
2108 static void shader_glsl_if(const SHADER_OPCODE_ARG
*arg
)
2110 glsl_src_param_t src0_param
;
2112 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2113 shader_addline(arg
->buffer
, "if (%s) {\n", src0_param
.param_str
);
2116 static void shader_glsl_ifc(const SHADER_OPCODE_ARG
*arg
)
2118 glsl_src_param_t src0_param
;
2119 glsl_src_param_t src1_param
;
2121 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2122 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], WINED3DSP_WRITEMASK_0
, &src1_param
);
2124 shader_addline(arg
->buffer
, "if (%s %s %s) {\n",
2125 src0_param
.param_str
, shader_get_comp_op(arg
->opcode_token
), src1_param
.param_str
);
2128 static void shader_glsl_else(const SHADER_OPCODE_ARG
*arg
)
2130 shader_addline(arg
->buffer
, "} else {\n");
2133 static void shader_glsl_break(const SHADER_OPCODE_ARG
*arg
)
2135 shader_addline(arg
->buffer
, "break;\n");
2138 /* FIXME: According to MSDN the compare is done per component. */
2139 static void shader_glsl_breakc(const SHADER_OPCODE_ARG
*arg
)
2141 glsl_src_param_t src0_param
;
2142 glsl_src_param_t src1_param
;
2144 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2145 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], WINED3DSP_WRITEMASK_0
, &src1_param
);
2147 shader_addline(arg
->buffer
, "if (%s %s %s) break;\n",
2148 src0_param
.param_str
, shader_get_comp_op(arg
->opcode_token
), src1_param
.param_str
);
2151 static void shader_glsl_label(const SHADER_OPCODE_ARG
*arg
)
2154 DWORD snum
= (arg
->src
[0]) & WINED3DSP_REGNUM_MASK
;
2155 shader_addline(arg
->buffer
, "}\n");
2156 shader_addline(arg
->buffer
, "void subroutine%u () {\n", snum
);
2159 static void shader_glsl_call(const SHADER_OPCODE_ARG
*arg
)
2161 DWORD snum
= (arg
->src
[0]) & WINED3DSP_REGNUM_MASK
;
2162 shader_addline(arg
->buffer
, "subroutine%u();\n", snum
);
2165 static void shader_glsl_callnz(const SHADER_OPCODE_ARG
*arg
)
2167 glsl_src_param_t src1_param
;
2169 DWORD snum
= (arg
->src
[0]) & WINED3DSP_REGNUM_MASK
;
2170 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], WINED3DSP_WRITEMASK_0
, &src1_param
);
2171 shader_addline(arg
->buffer
, "if (%s) subroutine%u();\n", src1_param
.param_str
, snum
);
2174 /*********************************************
2175 * Pixel Shader Specific Code begins here
2176 ********************************************/
2177 static void pshader_glsl_tex(const SHADER_OPCODE_ARG
*arg
)
2179 IWineD3DPixelShaderImpl
* This
= (IWineD3DPixelShaderImpl
*) arg
->shader
;
2180 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
2181 DWORD hex_version
= This
->baseShader
.hex_version
;
2182 char dst_swizzle
[6];
2183 glsl_sample_function_t sample_function
;
2186 BOOL projected
, texrect
= FALSE
;
2189 /* All versions have a destination register */
2190 shader_glsl_append_dst(arg
->buffer
, arg
);
2192 /* 1.0-1.4: Use destination register as sampler source.
2193 * 2.0+: Use provided sampler source. */
2194 if (hex_version
< WINED3DPS_VERSION(2,0)) {
2195 sampler_idx
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2197 sampler_idx
= arg
->src
[1] & WINED3DSP_REGNUM_MASK
;
2199 sampler_type
= arg
->reg_maps
->samplers
[sampler_idx
] & WINED3DSP_TEXTURETYPE_MASK
;
2201 if (hex_version
< WINED3DPS_VERSION(1,4)) {
2202 DWORD flags
= deviceImpl
->stateBlock
->textureState
[sampler_idx
][WINED3DTSS_TEXTURETRANSFORMFLAGS
];
2204 /* Projected cube textures don't make a lot of sense, the resulting coordinates stay the same. */
2205 if (flags
& WINED3DTTFF_PROJECTED
&& sampler_type
!= WINED3DSTT_CUBE
) {
2207 switch (flags
& ~WINED3DTTFF_PROJECTED
) {
2208 case WINED3DTTFF_COUNT1
: FIXME("WINED3DTTFF_PROJECTED with WINED3DTTFF_COUNT1?\n"); break;
2209 case WINED3DTTFF_COUNT2
: mask
= WINED3DSP_WRITEMASK_1
; break;
2210 case WINED3DTTFF_COUNT3
: mask
= WINED3DSP_WRITEMASK_2
; break;
2211 case WINED3DTTFF_COUNT4
:
2212 case WINED3DTTFF_DISABLE
: mask
= WINED3DSP_WRITEMASK_3
; break;
2217 } else if (hex_version
< WINED3DPS_VERSION(2,0)) {
2218 DWORD src_mod
= arg
->src
[0] & WINED3DSP_SRCMOD_MASK
;
2220 if (src_mod
== WINED3DSPSM_DZ
) {
2222 mask
= WINED3DSP_WRITEMASK_2
;
2223 } else if (src_mod
== WINED3DSPSM_DW
) {
2225 mask
= WINED3DSP_WRITEMASK_3
;
2230 if(arg
->opcode_token
& WINED3DSI_TEXLD_PROJECT
) {
2231 /* ps 2.0 texldp instruction always divides by the fourth component. */
2233 mask
= WINED3DSP_WRITEMASK_3
;
2239 if(deviceImpl
->stateBlock
->textures
[sampler_idx
] &&
2240 IWineD3DBaseTexture_GetTextureDimensions(deviceImpl
->stateBlock
->textures
[sampler_idx
]) == GL_TEXTURE_RECTANGLE_ARB
) {
2244 shader_glsl_get_sample_function(sampler_type
, projected
, texrect
, &sample_function
);
2245 mask
|= sample_function
.coord_mask
;
2247 if (hex_version
< WINED3DPS_VERSION(2,0)) {
2248 shader_glsl_get_write_mask(arg
->dst
, dst_swizzle
);
2250 shader_glsl_get_swizzle(arg
->src
[1], FALSE
, arg
->dst
, dst_swizzle
);
2253 /* 1.0-1.3: Use destination register as coordinate source.
2254 1.4+: Use provided coordinate source register. */
2255 if (hex_version
< WINED3DPS_VERSION(1,4)) {
2257 shader_glsl_get_write_mask(mask
, coord_mask
);
2258 shader_addline(arg
->buffer
, "%s(Psampler%u, T%u%s)%s);\n",
2259 sample_function
.name
, sampler_idx
, sampler_idx
, coord_mask
, dst_swizzle
);
2261 glsl_src_param_t coord_param
;
2262 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], mask
, &coord_param
);
2263 if(arg
->opcode_token
& WINED3DSI_TEXLD_BIAS
) {
2264 glsl_src_param_t bias
;
2265 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_3
, &bias
);
2267 shader_addline(arg
->buffer
, "%s(Psampler%u, %s, %s)%s);\n",
2268 sample_function
.name
, sampler_idx
, coord_param
.param_str
,
2269 bias
.param_str
, dst_swizzle
);
2271 shader_addline(arg
->buffer
, "%s(Psampler%u, %s)%s);\n",
2272 sample_function
.name
, sampler_idx
, coord_param
.param_str
, dst_swizzle
);
2277 static void shader_glsl_texldl(const SHADER_OPCODE_ARG
*arg
)
2279 IWineD3DBaseShaderImpl
* This
= (IWineD3DBaseShaderImpl
*)arg
->shader
;
2280 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
2281 glsl_sample_function_t sample_function
;
2282 glsl_src_param_t coord_param
, lod_param
;
2283 char dst_swizzle
[6];
2286 BOOL texrect
= FALSE
;
2288 shader_glsl_append_dst(arg
->buffer
, arg
);
2289 shader_glsl_get_swizzle(arg
->src
[1], FALSE
, arg
->dst
, dst_swizzle
);
2291 sampler_idx
= arg
->src
[1] & WINED3DSP_REGNUM_MASK
;
2292 sampler_type
= arg
->reg_maps
->samplers
[sampler_idx
] & WINED3DSP_TEXTURETYPE_MASK
;
2293 if(deviceImpl
->stateBlock
->textures
[sampler_idx
] &&
2294 IWineD3DBaseTexture_GetTextureDimensions(deviceImpl
->stateBlock
->textures
[sampler_idx
]) == GL_TEXTURE_RECTANGLE_ARB
) {
2297 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
);
2299 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_3
, &lod_param
);
2301 if (shader_is_pshader_version(This
->baseShader
.hex_version
)) {
2302 /* The GLSL spec claims the Lod sampling functions are only supported in vertex shaders.
2303 * However, they seem to work just fine in fragment shaders as well. */
2304 WARN("Using %sLod in fragment shader.\n", sample_function
.name
);
2305 shader_addline(arg
->buffer
, "%sLod(Psampler%u, %s, %s)%s);\n",
2306 sample_function
.name
, sampler_idx
, coord_param
.param_str
, lod_param
.param_str
, dst_swizzle
);
2308 shader_addline(arg
->buffer
, "%sLod(Vsampler%u, %s, %s)%s);\n",
2309 sample_function
.name
, sampler_idx
, coord_param
.param_str
, lod_param
.param_str
, dst_swizzle
);
2313 static void pshader_glsl_texcoord(const SHADER_OPCODE_ARG
*arg
)
2315 /* FIXME: Make this work for more than just 2D textures */
2317 IWineD3DPixelShaderImpl
* This
= (IWineD3DPixelShaderImpl
*) arg
->shader
;
2318 SHADER_BUFFER
* buffer
= arg
->buffer
;
2319 DWORD hex_version
= This
->baseShader
.hex_version
;
2323 write_mask
= shader_glsl_append_dst(arg
->buffer
, arg
);
2324 shader_glsl_get_write_mask(write_mask
, dst_mask
);
2326 if (hex_version
!= WINED3DPS_VERSION(1,4)) {
2327 DWORD reg
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2328 shader_addline(buffer
, "clamp(gl_TexCoord[%u], 0.0, 1.0)%s);\n", reg
, dst_mask
);
2330 DWORD reg
= arg
->src
[0] & WINED3DSP_REGNUM_MASK
;
2331 DWORD src_mod
= arg
->src
[0] & WINED3DSP_SRCMOD_MASK
;
2332 char dst_swizzle
[6];
2334 shader_glsl_get_swizzle(arg
->src
[0], FALSE
, write_mask
, dst_swizzle
);
2336 if (src_mod
== WINED3DSPSM_DZ
) {
2337 glsl_src_param_t div_param
;
2338 unsigned int mask_size
= shader_glsl_get_write_mask_size(write_mask
);
2339 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_2
, &div_param
);
2341 if (mask_size
> 1) {
2342 shader_addline(buffer
, "gl_TexCoord[%u]%s / vec%d(%s));\n", reg
, dst_swizzle
, mask_size
, div_param
.param_str
);
2344 shader_addline(buffer
, "gl_TexCoord[%u]%s / %s);\n", reg
, dst_swizzle
, div_param
.param_str
);
2346 } else if (src_mod
== WINED3DSPSM_DW
) {
2347 glsl_src_param_t div_param
;
2348 unsigned int mask_size
= shader_glsl_get_write_mask_size(write_mask
);
2349 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_3
, &div_param
);
2351 if (mask_size
> 1) {
2352 shader_addline(buffer
, "gl_TexCoord[%u]%s / vec%d(%s));\n", reg
, dst_swizzle
, mask_size
, div_param
.param_str
);
2354 shader_addline(buffer
, "gl_TexCoord[%u]%s / %s);\n", reg
, dst_swizzle
, div_param
.param_str
);
2357 shader_addline(buffer
, "gl_TexCoord[%u]%s);\n", reg
, dst_swizzle
);
2362 /** Process the WINED3DSIO_TEXDP3TEX instruction in GLSL:
2363 * Take a 3-component dot product of the TexCoord[dstreg] and src,
2364 * then perform a 1D texture lookup from stage dstregnum, place into dst. */
2365 static void pshader_glsl_texdp3tex(const SHADER_OPCODE_ARG
*arg
)
2367 glsl_src_param_t src0_param
;
2369 glsl_sample_function_t sample_function
;
2370 DWORD sampler_idx
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2371 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2372 DWORD sampler_type
= arg
->reg_maps
->samplers
[sampler_idx
] & WINED3DSP_TEXTURETYPE_MASK
;
2374 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], src_mask
, &src0_param
);
2376 shader_glsl_append_dst(arg
->buffer
, arg
);
2377 shader_glsl_get_write_mask(arg
->dst
, dst_mask
);
2379 /* Do I have to take care about the projected bit? I don't think so, since the dp3 returns only one
2380 * scalar, and projected sampling would require 4.
2382 * It is a dependent read - not valid with conditional NP2 textures
2384 shader_glsl_get_sample_function(sampler_type
, FALSE
, FALSE
, &sample_function
);
2386 switch(count_bits(sample_function
.coord_mask
)) {
2388 shader_addline(arg
->buffer
, "%s(Psampler%u, dot(gl_TexCoord[%u].xyz, %s))%s);\n",
2389 sample_function
.name
, sampler_idx
, sampler_idx
, src0_param
.param_str
, dst_mask
);
2393 shader_addline(arg
->buffer
, "%s(Psampler%u, vec2(dot(gl_TexCoord[%u].xyz, %s), 0.0))%s);\n",
2394 sample_function
.name
, sampler_idx
, sampler_idx
, src0_param
.param_str
, dst_mask
);
2398 shader_addline(arg
->buffer
, "%s(Psampler%u, vec3(dot(gl_TexCoord[%u].xyz, %s), 0.0, 0.0))%s);\n",
2399 sample_function
.name
, sampler_idx
, sampler_idx
, src0_param
.param_str
, dst_mask
);
2402 FIXME("Unexpected mask bitcount %d\n", count_bits(sample_function
.coord_mask
));
2406 /** Process the WINED3DSIO_TEXDP3 instruction in GLSL:
2407 * Take a 3-component dot product of the TexCoord[dstreg] and src. */
2408 static void pshader_glsl_texdp3(const SHADER_OPCODE_ARG
*arg
)
2410 glsl_src_param_t src0_param
;
2411 DWORD dstreg
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2412 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2414 unsigned int mask_size
;
2416 dst_mask
= shader_glsl_append_dst(arg
->buffer
, arg
);
2417 mask_size
= shader_glsl_get_write_mask_size(dst_mask
);
2418 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], src_mask
, &src0_param
);
2420 if (mask_size
> 1) {
2421 shader_addline(arg
->buffer
, "vec%d(dot(T%u.xyz, %s)));\n", mask_size
, dstreg
, src0_param
.param_str
);
2423 shader_addline(arg
->buffer
, "dot(T%u.xyz, %s));\n", dstreg
, src0_param
.param_str
);
2427 /** Process the WINED3DSIO_TEXDEPTH instruction in GLSL:
2428 * Calculate the depth as dst.x / dst.y */
2429 static void pshader_glsl_texdepth(const SHADER_OPCODE_ARG
*arg
)
2431 glsl_dst_param_t dst_param
;
2433 shader_glsl_add_dst_param(arg
, arg
->dst
, 0, &dst_param
);
2435 /* Tests show that texdepth never returns anything below 0.0, and that r5.y is clamped to 1.0.
2436 * Negative input is accepted, -0.25 / -0.5 returns 0.5. GL should clamp gl_FragDepth to [0;1], but
2437 * this doesn't always work, so clamp the results manually. Whether or not the x value is clamped at 1
2438 * too is irrelevant, since if x = 0, any y value < 1.0 (and > 1.0 is not allowed) results in a result
2441 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
);
2444 /** Process the WINED3DSIO_TEXM3X2DEPTH instruction in GLSL:
2445 * Last row of a 3x2 matrix multiply, use the result to calculate the depth:
2446 * Calculate tmp0.y = TexCoord[dstreg] . src.xyz; (tmp0.x has already been calculated)
2447 * depth = (tmp0.y == 0.0) ? 1.0 : tmp0.x / tmp0.y
2449 static void pshader_glsl_texm3x2depth(const SHADER_OPCODE_ARG
*arg
)
2451 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2452 DWORD dstreg
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2453 glsl_src_param_t src0_param
;
2455 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], src_mask
, &src0_param
);
2457 shader_addline(arg
->buffer
, "tmp0.y = dot(T%u.xyz, %s);\n", dstreg
, src0_param
.param_str
);
2458 shader_addline(arg
->buffer
, "gl_FragDepth = (tmp0.y == 0.0) ? 1.0 : clamp(tmp0.x / tmp0.y, 0.0, 1.0);\n");
2461 /** Process the WINED3DSIO_TEXM3X2PAD instruction in GLSL
2462 * Calculate the 1st of a 2-row matrix multiplication. */
2463 static void pshader_glsl_texm3x2pad(const SHADER_OPCODE_ARG
*arg
)
2465 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2466 DWORD reg
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2467 SHADER_BUFFER
* buffer
= arg
->buffer
;
2468 glsl_src_param_t src0_param
;
2470 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], src_mask
, &src0_param
);
2471 shader_addline(buffer
, "tmp0.x = dot(T%u.xyz, %s);\n", reg
, src0_param
.param_str
);
2474 /** Process the WINED3DSIO_TEXM3X3PAD instruction in GLSL
2475 * Calculate the 1st or 2nd row of a 3-row matrix multiplication. */
2476 static void pshader_glsl_texm3x3pad(const SHADER_OPCODE_ARG
* arg
)
2478 IWineD3DPixelShaderImpl
* shader
= (IWineD3DPixelShaderImpl
*) arg
->shader
;
2479 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2480 DWORD reg
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2481 SHADER_BUFFER
* buffer
= arg
->buffer
;
2482 SHADER_PARSE_STATE
* current_state
= &shader
->baseShader
.parse_state
;
2483 glsl_src_param_t src0_param
;
2485 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], src_mask
, &src0_param
);
2486 shader_addline(buffer
, "tmp0.%c = dot(T%u.xyz, %s);\n", 'x' + current_state
->current_row
, reg
, src0_param
.param_str
);
2487 current_state
->texcoord_w
[current_state
->current_row
++] = reg
;
2490 static void pshader_glsl_texm3x2tex(const SHADER_OPCODE_ARG
*arg
)
2492 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2493 DWORD reg
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2494 SHADER_BUFFER
* buffer
= arg
->buffer
;
2495 glsl_src_param_t src0_param
;
2498 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], src_mask
, &src0_param
);
2499 shader_addline(buffer
, "tmp0.y = dot(T%u.xyz, %s);\n", reg
, src0_param
.param_str
);
2501 shader_glsl_append_dst(buffer
, arg
);
2502 shader_glsl_get_write_mask(arg
->dst
, dst_mask
);
2504 /* Sample the texture using the calculated coordinates */
2505 shader_addline(buffer
, "texture2D(Psampler%u, tmp0.xy)%s);\n", reg
, dst_mask
);
2508 /** Process the WINED3DSIO_TEXM3X3TEX instruction in GLSL
2509 * Perform the 3rd row of a 3x3 matrix multiply, then sample the texture using the calculated coordinates */
2510 static void pshader_glsl_texm3x3tex(const SHADER_OPCODE_ARG
*arg
)
2512 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2513 glsl_src_param_t src0_param
;
2515 DWORD reg
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2516 IWineD3DPixelShaderImpl
* This
= (IWineD3DPixelShaderImpl
*) arg
->shader
;
2517 SHADER_PARSE_STATE
* current_state
= &This
->baseShader
.parse_state
;
2518 DWORD sampler_type
= arg
->reg_maps
->samplers
[reg
] & WINED3DSP_TEXTURETYPE_MASK
;
2519 glsl_sample_function_t sample_function
;
2521 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], src_mask
, &src0_param
);
2522 shader_addline(arg
->buffer
, "tmp0.z = dot(T%u.xyz, %s);\n", reg
, src0_param
.param_str
);
2524 shader_glsl_append_dst(arg
->buffer
, arg
);
2525 shader_glsl_get_write_mask(arg
->dst
, dst_mask
);
2526 /* Dependent read, not valid with conditional NP2 */
2527 shader_glsl_get_sample_function(sampler_type
, FALSE
, FALSE
, &sample_function
);
2529 /* Sample the texture using the calculated coordinates */
2530 shader_addline(arg
->buffer
, "%s(Psampler%u, tmp0.xyz)%s);\n", sample_function
.name
, reg
, dst_mask
);
2532 current_state
->current_row
= 0;
2535 /** Process the WINED3DSIO_TEXM3X3 instruction in GLSL
2536 * Perform the 3rd row of a 3x3 matrix multiply */
2537 static void pshader_glsl_texm3x3(const SHADER_OPCODE_ARG
*arg
)
2539 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2540 glsl_src_param_t src0_param
;
2542 DWORD reg
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2543 IWineD3DPixelShaderImpl
* This
= (IWineD3DPixelShaderImpl
*) arg
->shader
;
2544 SHADER_PARSE_STATE
* current_state
= &This
->baseShader
.parse_state
;
2546 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], src_mask
, &src0_param
);
2548 shader_glsl_append_dst(arg
->buffer
, arg
);
2549 shader_glsl_get_write_mask(arg
->dst
, dst_mask
);
2550 shader_addline(arg
->buffer
, "vec4(tmp0.xy, dot(T%u.xyz, %s), 1.0)%s);\n", reg
, src0_param
.param_str
, dst_mask
);
2552 current_state
->current_row
= 0;
2555 /** Process the WINED3DSIO_TEXM3X3SPEC instruction in GLSL
2556 * Perform the final texture lookup based on the previous 2 3x3 matrix multiplies */
2557 static void pshader_glsl_texm3x3spec(const SHADER_OPCODE_ARG
*arg
)
2559 IWineD3DPixelShaderImpl
* shader
= (IWineD3DPixelShaderImpl
*) arg
->shader
;
2560 DWORD reg
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2561 glsl_src_param_t src0_param
;
2562 glsl_src_param_t src1_param
;
2564 SHADER_BUFFER
* buffer
= arg
->buffer
;
2565 SHADER_PARSE_STATE
* current_state
= &shader
->baseShader
.parse_state
;
2566 DWORD stype
= arg
->reg_maps
->samplers
[reg
] & WINED3DSP_TEXTURETYPE_MASK
;
2567 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2568 glsl_sample_function_t sample_function
;
2570 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], src_mask
, &src0_param
);
2571 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], src_mask
, &src1_param
);
2573 /* Perform the last matrix multiply operation */
2574 shader_addline(buffer
, "tmp0.z = dot(T%u.xyz, %s);\n", reg
, src0_param
.param_str
);
2575 /* Reflection calculation */
2576 shader_addline(buffer
, "tmp0.xyz = -reflect((%s), normalize(tmp0.xyz));\n", src1_param
.param_str
);
2578 shader_glsl_append_dst(buffer
, arg
);
2579 shader_glsl_get_write_mask(arg
->dst
, dst_mask
);
2580 /* Dependent read, not valid with conditional NP2 */
2581 shader_glsl_get_sample_function(stype
, FALSE
, FALSE
, &sample_function
);
2583 /* Sample the texture */
2584 shader_addline(buffer
, "%s(Psampler%u, tmp0.xyz)%s);\n", sample_function
.name
, reg
, dst_mask
);
2586 current_state
->current_row
= 0;
2589 /** Process the WINED3DSIO_TEXM3X3VSPEC instruction in GLSL
2590 * Perform the final texture lookup based on the previous 2 3x3 matrix multiplies */
2591 static void pshader_glsl_texm3x3vspec(const SHADER_OPCODE_ARG
*arg
)
2593 IWineD3DPixelShaderImpl
* shader
= (IWineD3DPixelShaderImpl
*) arg
->shader
;
2594 DWORD reg
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2595 SHADER_BUFFER
* buffer
= arg
->buffer
;
2596 SHADER_PARSE_STATE
* current_state
= &shader
->baseShader
.parse_state
;
2597 glsl_src_param_t src0_param
;
2599 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2600 DWORD sampler_type
= arg
->reg_maps
->samplers
[reg
] & WINED3DSP_TEXTURETYPE_MASK
;
2601 glsl_sample_function_t sample_function
;
2603 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], src_mask
, &src0_param
);
2605 /* Perform the last matrix multiply operation */
2606 shader_addline(buffer
, "tmp0.z = dot(vec3(T%u), vec3(%s));\n", reg
, src0_param
.param_str
);
2608 /* Construct the eye-ray vector from w coordinates */
2609 shader_addline(buffer
, "tmp1.xyz = normalize(vec3(gl_TexCoord[%u].w, gl_TexCoord[%u].w, gl_TexCoord[%u].w));\n",
2610 current_state
->texcoord_w
[0], current_state
->texcoord_w
[1], reg
);
2611 shader_addline(buffer
, "tmp0.xyz = -reflect(tmp1.xyz, normalize(tmp0.xyz));\n");
2613 shader_glsl_append_dst(buffer
, arg
);
2614 shader_glsl_get_write_mask(arg
->dst
, dst_mask
);
2615 /* Dependent read, not valid with conditional NP2 */
2616 shader_glsl_get_sample_function(sampler_type
, FALSE
, FALSE
, &sample_function
);
2618 /* Sample the texture using the calculated coordinates */
2619 shader_addline(buffer
, "%s(Psampler%u, tmp0.xyz)%s);\n", sample_function
.name
, reg
, dst_mask
);
2621 current_state
->current_row
= 0;
2624 /** Process the WINED3DSIO_TEXBEM instruction in GLSL.
2625 * Apply a fake bump map transform.
2626 * texbem is pshader <= 1.3 only, this saves a few version checks
2628 static void pshader_glsl_texbem(const SHADER_OPCODE_ARG
*arg
)
2630 IWineD3DPixelShaderImpl
* This
= (IWineD3DPixelShaderImpl
*) arg
->shader
;
2631 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
2632 char dst_swizzle
[6];
2633 glsl_sample_function_t sample_function
;
2634 glsl_src_param_t coord_param
;
2641 sampler_idx
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2642 flags
= deviceImpl
->stateBlock
->textureState
[sampler_idx
][WINED3DTSS_TEXTURETRANSFORMFLAGS
];
2644 sampler_type
= arg
->reg_maps
->samplers
[sampler_idx
] & WINED3DSP_TEXTURETYPE_MASK
;
2645 /* Dependent read, not valid with conditional NP2 */
2646 shader_glsl_get_sample_function(sampler_type
, FALSE
, FALSE
, &sample_function
);
2647 mask
= sample_function
.coord_mask
;
2649 shader_glsl_get_write_mask(arg
->dst
, dst_swizzle
);
2651 shader_glsl_get_write_mask(mask
, coord_mask
);
2653 /* with projective textures, texbem only divides the static texture coord, not the displacement,
2654 * so we can't let the GL handle this.
2656 if (flags
& WINED3DTTFF_PROJECTED
) {
2658 char coord_div_mask
[3];
2659 switch (flags
& ~WINED3DTTFF_PROJECTED
) {
2660 case WINED3DTTFF_COUNT1
: FIXME("WINED3DTTFF_PROJECTED with WINED3DTTFF_COUNT1?\n"); break;
2661 case WINED3DTTFF_COUNT2
: div_mask
= WINED3DSP_WRITEMASK_1
; break;
2662 case WINED3DTTFF_COUNT3
: div_mask
= WINED3DSP_WRITEMASK_2
; break;
2663 case WINED3DTTFF_COUNT4
:
2664 case WINED3DTTFF_DISABLE
: div_mask
= WINED3DSP_WRITEMASK_3
; break;
2666 shader_glsl_get_write_mask(div_mask
, coord_div_mask
);
2667 shader_addline(arg
->buffer
, "T%u%s /= T%u%s;\n", sampler_idx
, coord_mask
, sampler_idx
, coord_div_mask
);
2670 shader_glsl_append_dst(arg
->buffer
, arg
);
2671 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_0
|WINED3DSP_WRITEMASK_1
, &coord_param
);
2672 if(arg
->opcode
->opcode
== WINED3DSIO_TEXBEML
) {
2673 glsl_src_param_t luminance_param
;
2674 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_2
, &luminance_param
);
2675 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",
2676 sample_function
.name
, sampler_idx
, sampler_idx
, coord_mask
, sampler_idx
, coord_param
.param_str
, coord_mask
,
2677 luminance_param
.param_str
, sampler_idx
, sampler_idx
, dst_swizzle
);
2679 shader_addline(arg
->buffer
, "%s(Psampler%u, T%u%s + vec4(bumpenvmat%d * %s, 0.0, 0.0)%s )%s);\n",
2680 sample_function
.name
, sampler_idx
, sampler_idx
, coord_mask
, sampler_idx
, coord_param
.param_str
, coord_mask
, dst_swizzle
);
2684 static void pshader_glsl_bem(const SHADER_OPCODE_ARG
*arg
)
2686 glsl_src_param_t src0_param
, src1_param
;
2687 DWORD sampler_idx
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2689 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_0
|WINED3DSP_WRITEMASK_1
, &src0_param
);
2690 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], WINED3DSP_WRITEMASK_0
|WINED3DSP_WRITEMASK_1
, &src1_param
);
2692 shader_glsl_append_dst(arg
->buffer
, arg
);
2693 shader_addline(arg
->buffer
, "%s + bumpenvmat%d * %s);\n",
2694 src0_param
.param_str
, sampler_idx
, src1_param
.param_str
);
2697 /** Process the WINED3DSIO_TEXREG2AR instruction in GLSL
2698 * Sample 2D texture at dst using the alpha & red (wx) components of src as texture coordinates */
2699 static void pshader_glsl_texreg2ar(const SHADER_OPCODE_ARG
*arg
)
2701 glsl_src_param_t src0_param
;
2702 DWORD sampler_idx
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2705 shader_glsl_append_dst(arg
->buffer
, arg
);
2706 shader_glsl_get_write_mask(arg
->dst
, dst_mask
);
2707 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_ALL
, &src0_param
);
2709 shader_addline(arg
->buffer
, "texture2D(Psampler%u, %s.wx)%s);\n", sampler_idx
, src0_param
.reg_name
, dst_mask
);
2712 /** Process the WINED3DSIO_TEXREG2GB instruction in GLSL
2713 * Sample 2D texture at dst using the green & blue (yz) components of src as texture coordinates */
2714 static void pshader_glsl_texreg2gb(const SHADER_OPCODE_ARG
*arg
)
2716 glsl_src_param_t src0_param
;
2717 DWORD sampler_idx
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2720 shader_glsl_append_dst(arg
->buffer
, arg
);
2721 shader_glsl_get_write_mask(arg
->dst
, dst_mask
);
2722 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_ALL
, &src0_param
);
2724 shader_addline(arg
->buffer
, "texture2D(Psampler%u, %s.yz)%s);\n", sampler_idx
, src0_param
.reg_name
, dst_mask
);
2727 /** Process the WINED3DSIO_TEXREG2RGB instruction in GLSL
2728 * Sample texture at dst using the rgb (xyz) components of src as texture coordinates */
2729 static void pshader_glsl_texreg2rgb(const SHADER_OPCODE_ARG
*arg
)
2731 glsl_src_param_t src0_param
;
2733 DWORD sampler_idx
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2734 DWORD sampler_type
= arg
->reg_maps
->samplers
[sampler_idx
] & WINED3DSP_TEXTURETYPE_MASK
;
2735 glsl_sample_function_t sample_function
;
2737 shader_glsl_append_dst(arg
->buffer
, arg
);
2738 shader_glsl_get_write_mask(arg
->dst
, dst_mask
);
2739 /* Dependent read, not valid with conditional NP2 */
2740 shader_glsl_get_sample_function(sampler_type
, FALSE
, FALSE
, &sample_function
);
2741 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], sample_function
.coord_mask
, &src0_param
);
2743 shader_addline(arg
->buffer
, "%s(Psampler%u, %s)%s);\n", sample_function
.name
, sampler_idx
, src0_param
.param_str
, dst_mask
);
2746 /** Process the WINED3DSIO_TEXKILL instruction in GLSL.
2747 * If any of the first 3 components are < 0, discard this pixel */
2748 static void pshader_glsl_texkill(const SHADER_OPCODE_ARG
*arg
)
2750 IWineD3DPixelShaderImpl
* This
= (IWineD3DPixelShaderImpl
*) arg
->shader
;
2751 DWORD hex_version
= This
->baseShader
.hex_version
;
2752 glsl_dst_param_t dst_param
;
2754 /* The argument is a destination parameter, and no writemasks are allowed */
2755 shader_glsl_add_dst_param(arg
, arg
->dst
, 0, &dst_param
);
2756 if((hex_version
>= WINED3DPS_VERSION(2,0))) {
2757 /* 2.0 shaders compare all 4 components in texkill */
2758 shader_addline(arg
->buffer
, "if (any(lessThan(%s.xyzw, vec4(0.0)))) discard;\n", dst_param
.reg_name
);
2760 /* 1.X shaders only compare the first 3 components, probably due to the nature of the texkill
2761 * instruction as a tex* instruction, and phase, which kills all a / w components. Even if all
2762 * 4 components are defined, only the first 3 are used
2764 shader_addline(arg
->buffer
, "if (any(lessThan(%s.xyz, vec3(0.0)))) discard;\n", dst_param
.reg_name
);
2768 /** Process the WINED3DSIO_DP2ADD instruction in GLSL.
2769 * dst = dot2(src0, src1) + src2 */
2770 static void pshader_glsl_dp2add(const SHADER_OPCODE_ARG
*arg
)
2772 glsl_src_param_t src0_param
;
2773 glsl_src_param_t src1_param
;
2774 glsl_src_param_t src2_param
;
2776 unsigned int mask_size
;
2778 write_mask
= shader_glsl_append_dst(arg
->buffer
, arg
);
2779 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
2781 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
, &src0_param
);
2782 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
, &src1_param
);
2783 shader_glsl_add_src_param(arg
, arg
->src
[2], arg
->src_addr
[2], WINED3DSP_WRITEMASK_0
, &src2_param
);
2785 if (mask_size
> 1) {
2786 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
);
2788 shader_addline(arg
->buffer
, "dot(%s, %s) + %s);\n", src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
2792 static void pshader_glsl_input_pack(SHADER_BUFFER
* buffer
, const struct semantic
* semantics_in
,
2793 IWineD3DPixelShader
*iface
, enum vertexprocessing_mode vertexprocessing
)
2796 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*) iface
;
2798 for (i
= 0; i
< MAX_REG_INPUT
; i
++) {
2800 DWORD usage_token
= semantics_in
[i
].usage
;
2801 DWORD register_token
= semantics_in
[i
].reg
;
2802 DWORD usage
, usage_idx
;
2806 if (!usage_token
) continue;
2807 usage
= (usage_token
& WINED3DSP_DCL_USAGE_MASK
) >> WINED3DSP_DCL_USAGE_SHIFT
;
2808 usage_idx
= (usage_token
& WINED3DSP_DCL_USAGEINDEX_MASK
) >> WINED3DSP_DCL_USAGEINDEX_SHIFT
;
2809 shader_glsl_get_write_mask(register_token
, reg_mask
);
2813 case WINED3DDECLUSAGE_TEXCOORD
:
2814 if(usage_idx
< 8 && vertexprocessing
== pretransformed
) {
2815 shader_addline(buffer
, "IN[%u]%s = gl_TexCoord[%u]%s;\n",
2816 This
->input_reg_map
[i
], reg_mask
, usage_idx
, reg_mask
);
2818 shader_addline(buffer
, "IN[%u]%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
2819 This
->input_reg_map
[i
], reg_mask
, reg_mask
);
2823 case WINED3DDECLUSAGE_COLOR
:
2825 shader_addline(buffer
, "IN[%u]%s = vec4(gl_Color)%s;\n",
2826 This
->input_reg_map
[i
], reg_mask
, reg_mask
);
2827 else if (usage_idx
== 1)
2828 shader_addline(buffer
, "IN[%u]%s = vec4(gl_SecondaryColor)%s;\n",
2829 This
->input_reg_map
[i
], reg_mask
, reg_mask
);
2831 shader_addline(buffer
, "IN[%u]%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
2832 This
->input_reg_map
[i
], reg_mask
, reg_mask
);
2836 shader_addline(buffer
, "IN[%u]%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
2837 This
->input_reg_map
[i
], reg_mask
, reg_mask
);
2842 /*********************************************
2843 * Vertex Shader Specific Code begins here
2844 ********************************************/
2846 static void add_glsl_program_entry(struct shader_glsl_priv
*priv
, struct glsl_shader_prog_link
*entry
) {
2847 glsl_program_key_t
*key
;
2849 key
= HeapAlloc(GetProcessHeap(), 0, sizeof(glsl_program_key_t
));
2850 key
->vshader
= entry
->vshader
;
2851 key
->pshader
= entry
->pshader
;
2852 key
->ps_args
= entry
->ps_args
;
2854 hash_table_put(priv
->glsl_program_lookup
, key
, entry
);
2857 static struct glsl_shader_prog_link
*get_glsl_program_entry(struct shader_glsl_priv
*priv
,
2858 GLhandleARB vshader
, IWineD3DPixelShader
*pshader
, struct ps_compile_args
*ps_args
) {
2859 glsl_program_key_t key
;
2861 key
.vshader
= vshader
;
2862 key
.pshader
= pshader
;
2863 key
.ps_args
= *ps_args
;
2865 return (struct glsl_shader_prog_link
*)hash_table_get(priv
->glsl_program_lookup
, &key
);
2868 static void delete_glsl_program_entry(struct shader_glsl_priv
*priv
, const WineD3D_GL_Info
*gl_info
,
2869 struct glsl_shader_prog_link
*entry
)
2871 glsl_program_key_t
*key
;
2873 key
= HeapAlloc(GetProcessHeap(), 0, sizeof(glsl_program_key_t
));
2874 key
->vshader
= entry
->vshader
;
2875 key
->pshader
= entry
->pshader
;
2876 key
->ps_args
= entry
->ps_args
;
2877 hash_table_remove(priv
->glsl_program_lookup
, key
);
2879 GL_EXTCALL(glDeleteObjectARB(entry
->programId
));
2880 if (entry
->vshader
) list_remove(&entry
->vshader_entry
);
2881 if (entry
->pshader
) list_remove(&entry
->pshader_entry
);
2882 HeapFree(GetProcessHeap(), 0, entry
->vuniformF_locations
);
2883 HeapFree(GetProcessHeap(), 0, entry
->puniformF_locations
);
2884 HeapFree(GetProcessHeap(), 0, entry
);
2887 static void handle_ps3_input(SHADER_BUFFER
*buffer
, const struct semantic
*semantics_in
,
2888 const struct semantic
*semantics_out
, const WineD3D_GL_Info
*gl_info
, const DWORD
*map
)
2891 DWORD usage_token
, usage_token_out
;
2892 DWORD register_token
, register_token_out
;
2893 DWORD usage
, usage_idx
, usage_out
, usage_idx_out
;
2896 DWORD in_count
= GL_LIMITS(glsl_varyings
) / 4;
2897 char reg_mask
[6], reg_mask_out
[6];
2898 char destination
[50];
2900 set
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*set
) * (in_count
+ 2));
2902 if (!semantics_out
) {
2903 /* Save gl_FrontColor & gl_FrontSecondaryColor before overwriting them. */
2904 shader_addline(buffer
, "vec4 front_color = gl_FrontColor;\n");
2905 shader_addline(buffer
, "vec4 front_secondary_color = gl_FrontSecondaryColor;\n");
2908 for(i
= 0; i
< MAX_REG_INPUT
; i
++) {
2909 usage_token
= semantics_in
[i
].usage
;
2910 if (!usage_token
) continue;
2913 if (in_idx
>= (in_count
+ 2)) {
2914 FIXME("More input varyings declared than supported, expect issues\n");
2916 } else if(map
[i
] == -1) {
2917 /* Declared, but not read register */
2921 if (in_idx
== in_count
) {
2922 sprintf(destination
, "gl_FrontColor");
2923 } else if (in_idx
== in_count
+ 1) {
2924 sprintf(destination
, "gl_FrontSecondaryColor");
2926 sprintf(destination
, "IN[%u]", in_idx
);
2929 register_token
= semantics_in
[i
].reg
;
2931 usage
= (usage_token
& WINED3DSP_DCL_USAGE_MASK
) >> WINED3DSP_DCL_USAGE_SHIFT
;
2932 usage_idx
= (usage_token
& WINED3DSP_DCL_USAGEINDEX_MASK
) >> WINED3DSP_DCL_USAGEINDEX_SHIFT
;
2933 set
[map
[i
]] = shader_glsl_get_write_mask(register_token
, reg_mask
);
2935 if(!semantics_out
) {
2937 case WINED3DDECLUSAGE_COLOR
:
2939 shader_addline(buffer
, "%s%s = front_color%s;\n",
2940 destination
, reg_mask
, reg_mask
);
2941 else if (usage_idx
== 1)
2942 shader_addline(buffer
, "%s%s = front_secondary_color%s;\n",
2943 destination
, reg_mask
, reg_mask
);
2945 shader_addline(buffer
, "%s%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
2946 destination
, reg_mask
, reg_mask
);
2949 case WINED3DDECLUSAGE_TEXCOORD
:
2950 if (usage_idx
< 8) {
2951 shader_addline(buffer
, "%s%s = gl_TexCoord[%u]%s;\n",
2952 destination
, reg_mask
, usage_idx
, reg_mask
);
2954 shader_addline(buffer
, "%s%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
2955 destination
, reg_mask
, reg_mask
);
2959 case WINED3DDECLUSAGE_FOG
:
2960 shader_addline(buffer
, "%s%s = vec4(gl_FogFragCoord, 0.0, 0.0, 0.0)%s;\n",
2961 destination
, reg_mask
, reg_mask
);
2965 shader_addline(buffer
, "%s%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
2966 destination
, reg_mask
, reg_mask
);
2970 for(j
= 0; j
< MAX_REG_OUTPUT
; j
++) {
2971 usage_token_out
= semantics_out
[j
].usage
;
2972 if (!usage_token_out
) continue;
2973 register_token_out
= semantics_out
[j
].reg
;
2975 usage_out
= (usage_token_out
& WINED3DSP_DCL_USAGE_MASK
) >> WINED3DSP_DCL_USAGE_SHIFT
;
2976 usage_idx_out
= (usage_token_out
& WINED3DSP_DCL_USAGEINDEX_MASK
) >> WINED3DSP_DCL_USAGEINDEX_SHIFT
;
2977 shader_glsl_get_write_mask(register_token_out
, reg_mask_out
);
2979 if(usage
== usage_out
&&
2980 usage_idx
== usage_idx_out
) {
2981 shader_addline(buffer
, "%s%s = OUT[%u]%s;\n",
2982 destination
, reg_mask
, j
, reg_mask
);
2987 shader_addline(buffer
, "%s%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
2988 destination
, reg_mask
, reg_mask
);
2993 /* This is solely to make the compiler / linker happy and avoid warning about undefined
2994 * varyings. It shouldn't result in any real code executed on the GPU, since all read
2995 * input varyings are assigned above, if the optimizer works properly.
2997 for(i
= 0; i
< in_count
+ 2; i
++) {
2998 if(set
[i
] != WINED3DSP_WRITEMASK_ALL
) {
2999 unsigned int size
= 0;
3000 memset(reg_mask
, 0, sizeof(reg_mask
));
3001 if(!(set
[i
] & WINED3DSP_WRITEMASK_0
)) {
3002 reg_mask
[size
] = 'x';
3005 if(!(set
[i
] & WINED3DSP_WRITEMASK_1
)) {
3006 reg_mask
[size
] = 'y';
3009 if(!(set
[i
] & WINED3DSP_WRITEMASK_2
)) {
3010 reg_mask
[size
] = 'z';
3013 if(!(set
[i
] & WINED3DSP_WRITEMASK_3
)) {
3014 reg_mask
[size
] = 'w';
3018 if (i
== in_count
) {
3019 sprintf(destination
, "gl_FrontColor");
3020 } else if (i
== in_count
+ 1) {
3021 sprintf(destination
, "gl_FrontSecondaryColor");
3023 sprintf(destination
, "IN[%u]", i
);
3027 shader_addline(buffer
, "%s.%s = 0.0;\n", destination
, reg_mask
);
3029 shader_addline(buffer
, "%s.%s = vec%u(0.0);\n", destination
, reg_mask
, size
);
3034 HeapFree(GetProcessHeap(), 0, set
);
3037 static GLhandleARB
generate_param_reorder_function(IWineD3DVertexShader
*vertexshader
,
3038 IWineD3DPixelShader
*pixelshader
, const WineD3D_GL_Info
*gl_info
)
3040 GLhandleARB ret
= 0;
3041 IWineD3DVertexShaderImpl
*vs
= (IWineD3DVertexShaderImpl
*) vertexshader
;
3042 IWineD3DPixelShaderImpl
*ps
= (IWineD3DPixelShaderImpl
*) pixelshader
;
3043 IWineD3DDeviceImpl
*device
;
3044 DWORD vs_major
= WINED3DSHADER_VERSION_MAJOR(vs
->baseShader
.hex_version
);
3045 DWORD ps_major
= ps
? WINED3DSHADER_VERSION_MAJOR(ps
->baseShader
.hex_version
) : 0;
3047 SHADER_BUFFER buffer
;
3049 DWORD register_token
;
3050 DWORD usage
, usage_idx
, writemask
;
3052 const struct semantic
*semantics_out
, *semantics_in
;
3054 buffer
.buffer
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, SHADER_PGMSIZE
);
3057 buffer
.newline
= TRUE
;
3059 shader_addline(&buffer
, "#version 120\n");
3061 if(vs_major
< 3 && ps_major
< 3) {
3062 /* That one is easy: The vertex shader writes to the builtin varyings, the pixel shader reads from them.
3063 * Take care about the texcoord .w fixup though if we're using the fixed function fragment pipeline
3065 device
= (IWineD3DDeviceImpl
*) vs
->baseShader
.device
;
3066 if((GLINFO_LOCATION
).set_texcoord_w
&& ps_major
== 0 && vs_major
> 0 &&
3067 !device
->frag_pipe
->ffp_proj_control
) {
3068 shader_addline(&buffer
, "void order_ps_input() {\n");
3069 for(i
= 0; i
< min(8, MAX_REG_TEXCRD
); i
++) {
3070 if(vs
->baseShader
.reg_maps
.texcoord_mask
[i
] != 0 &&
3071 vs
->baseShader
.reg_maps
.texcoord_mask
[i
] != WINED3DSP_WRITEMASK_ALL
) {
3072 shader_addline(&buffer
, "gl_TexCoord[%u].w = 1.0;\n", i
);
3075 shader_addline(&buffer
, "}\n");
3077 shader_addline(&buffer
, "void order_ps_input() { /* do nothing */ }\n");
3079 } else if(ps_major
< 3 && vs_major
>= 3) {
3080 /* The vertex shader writes to its own varyings, the pixel shader needs them in the builtin ones */
3081 semantics_out
= vs
->semantics_out
;
3083 shader_addline(&buffer
, "void order_ps_input(in vec4 OUT[%u]) {\n", MAX_REG_OUTPUT
);
3084 for(i
= 0; i
< MAX_REG_OUTPUT
; i
++) {
3085 usage_token
= semantics_out
[i
].usage
;
3086 if (!usage_token
) continue;
3087 register_token
= semantics_out
[i
].reg
;
3089 usage
= (usage_token
& WINED3DSP_DCL_USAGE_MASK
) >> WINED3DSP_DCL_USAGE_SHIFT
;
3090 usage_idx
= (usage_token
& WINED3DSP_DCL_USAGEINDEX_MASK
) >> WINED3DSP_DCL_USAGEINDEX_SHIFT
;
3091 writemask
= shader_glsl_get_write_mask(register_token
, reg_mask
);
3094 case WINED3DDECLUSAGE_COLOR
:
3096 shader_addline(&buffer
, "gl_FrontColor%s = OUT[%u]%s;\n", reg_mask
, i
, reg_mask
);
3097 else if (usage_idx
== 1)
3098 shader_addline(&buffer
, "gl_FrontSecondaryColor%s = OUT[%u]%s;\n", reg_mask
, i
, reg_mask
);
3101 case WINED3DDECLUSAGE_POSITION
:
3102 shader_addline(&buffer
, "gl_Position%s = OUT[%u]%s;\n", reg_mask
, i
, reg_mask
);
3105 case WINED3DDECLUSAGE_TEXCOORD
:
3106 if (usage_idx
< 8) {
3107 if(!(GLINFO_LOCATION
).set_texcoord_w
|| ps_major
> 0) writemask
|= WINED3DSP_WRITEMASK_3
;
3109 shader_addline(&buffer
, "gl_TexCoord[%u]%s = OUT[%u]%s;\n",
3110 usage_idx
, reg_mask
, i
, reg_mask
);
3111 if(!(writemask
& WINED3DSP_WRITEMASK_3
)) {
3112 shader_addline(&buffer
, "gl_TexCoord[%u].w = 1.0;\n", usage_idx
);
3117 case WINED3DDECLUSAGE_PSIZE
:
3118 shader_addline(&buffer
, "gl_PointSize = OUT[%u].x;\n", i
);
3121 case WINED3DDECLUSAGE_FOG
:
3122 shader_addline(&buffer
, "gl_FogFragCoord = OUT[%u].%c;\n", i
, reg_mask
[1]);
3129 shader_addline(&buffer
, "}\n");
3131 } else if(ps_major
>= 3 && vs_major
>= 3) {
3132 semantics_out
= vs
->semantics_out
;
3133 semantics_in
= ps
->semantics_in
;
3135 /* This one is tricky: a 3.0 pixel shader reads from a 3.0 vertex shader */
3136 shader_addline(&buffer
, "varying vec4 IN[%u];\n", GL_LIMITS(glsl_varyings
) / 4);
3137 shader_addline(&buffer
, "void order_ps_input(in vec4 OUT[%u]) {\n", MAX_REG_OUTPUT
);
3139 /* First, sort out position and point size. Those are not passed to the pixel shader */
3140 for(i
= 0; i
< MAX_REG_OUTPUT
; i
++) {
3141 usage_token
= semantics_out
[i
].usage
;
3142 if (!usage_token
) continue;
3143 register_token
= semantics_out
[i
].reg
;
3145 usage
= (usage_token
& WINED3DSP_DCL_USAGE_MASK
) >> WINED3DSP_DCL_USAGE_SHIFT
;
3146 usage_idx
= (usage_token
& WINED3DSP_DCL_USAGEINDEX_MASK
) >> WINED3DSP_DCL_USAGEINDEX_SHIFT
;
3147 shader_glsl_get_write_mask(register_token
, reg_mask
);
3150 case WINED3DDECLUSAGE_POSITION
:
3151 shader_addline(&buffer
, "gl_Position%s = OUT[%u]%s;\n", reg_mask
, i
, reg_mask
);
3154 case WINED3DDECLUSAGE_PSIZE
:
3155 shader_addline(&buffer
, "gl_PointSize = OUT[%u].x;\n", i
);
3163 /* Then, fix the pixel shader input */
3164 handle_ps3_input(&buffer
, semantics_in
, semantics_out
, gl_info
, ps
->input_reg_map
);
3166 shader_addline(&buffer
, "}\n");
3167 } else if(ps_major
>= 3 && vs_major
< 3) {
3168 semantics_in
= ps
->semantics_in
;
3170 shader_addline(&buffer
, "varying vec4 IN[%u];\n", GL_LIMITS(glsl_varyings
) / 4);
3171 shader_addline(&buffer
, "void order_ps_input() {\n");
3172 /* The vertex shader wrote to the builtin varyings. There is no need to figure out position and
3173 * point size, but we depend on the optimizers kindness to find out that the pixel shader doesn't
3174 * read gl_TexCoord and gl_ColorX, otherwise we'll run out of varyings
3176 handle_ps3_input(&buffer
, semantics_in
, NULL
, gl_info
, ps
->input_reg_map
);
3177 shader_addline(&buffer
, "}\n");
3179 ERR("Unexpected vertex and pixel shader version condition: vs: %d, ps: %d\n", vs_major
, ps_major
);
3182 ret
= GL_EXTCALL(glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB
));
3183 checkGLcall("glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB)");
3184 GL_EXTCALL(glShaderSourceARB(ret
, 1, (const char**)&buffer
.buffer
, NULL
));
3185 checkGLcall("glShaderSourceARB(ret, 1, (const char**)&buffer.buffer, NULL)");
3186 GL_EXTCALL(glCompileShaderARB(ret
));
3187 checkGLcall("glCompileShaderARB(ret)");
3189 HeapFree(GetProcessHeap(), 0, buffer
.buffer
);
3193 static void hardcode_local_constants(IWineD3DBaseShaderImpl
*shader
, const WineD3D_GL_Info
*gl_info
,
3194 GLhandleARB programId
, char prefix
)
3196 const local_constant
*lconst
;
3201 LIST_FOR_EACH_ENTRY(lconst
, &shader
->baseShader
.constantsF
, local_constant
, entry
) {
3202 value
= (const float *)lconst
->value
;
3203 snprintf(glsl_name
, sizeof(glsl_name
), "%cLC%u", prefix
, lconst
->idx
);
3204 tmp_loc
= GL_EXTCALL(glGetUniformLocationARB(programId
, glsl_name
));
3205 GL_EXTCALL(glUniform4fvARB(tmp_loc
, 1, value
));
3207 checkGLcall("Hardcoding local constants\n");
3210 /** Sets the GLSL program ID for the given pixel and vertex shader combination.
3211 * It sets the programId on the current StateBlock (because it should be called
3212 * inside of the DrawPrimitive() part of the render loop).
3214 * If a program for the given combination does not exist, create one, and store
3215 * the program in the hash table. If it creates a program, it will link the
3216 * given objects, too.
3218 static void set_glsl_shader_program(IWineD3DDevice
*iface
, BOOL use_ps
, BOOL use_vs
) {
3219 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
3220 struct shader_glsl_priv
*priv
= (struct shader_glsl_priv
*)This
->shader_priv
;
3221 const WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
3222 IWineD3DPixelShader
*pshader
= This
->stateBlock
->pixelShader
;
3223 IWineD3DVertexShader
*vshader
= This
->stateBlock
->vertexShader
;
3224 struct glsl_shader_prog_link
*entry
= NULL
;
3225 GLhandleARB programId
= 0;
3226 GLhandleARB reorder_shader_id
= 0;
3229 GLhandleARB vshader_id
, pshader_id
;
3230 struct ps_compile_args compile_args
;
3233 IWineD3DVertexShaderImpl_CompileShader(vshader
);
3234 vshader_id
= ((IWineD3DVertexShaderImpl
*)vshader
)->prgId
;
3239 find_ps_compile_args((IWineD3DPixelShaderImpl
*)This
->stateBlock
->pixelShader
, This
->stateBlock
, &compile_args
);
3241 /* FIXME: Do we really have to spend CPU cycles to generate a few zeroed bytes? */
3242 memset(&compile_args
, 0, sizeof(compile_args
));
3244 entry
= get_glsl_program_entry(priv
, vshader_id
, pshader
, &compile_args
);
3246 priv
->glsl_program
= entry
;
3250 /* If we get to this point, then no matching program exists, so we create one */
3251 programId
= GL_EXTCALL(glCreateProgramObjectARB());
3252 TRACE("Created new GLSL shader program %u\n", programId
);
3254 /* Create the entry */
3255 entry
= HeapAlloc(GetProcessHeap(), 0, sizeof(struct glsl_shader_prog_link
));
3256 entry
->programId
= programId
;
3257 entry
->vshader
= vshader_id
;
3258 entry
->pshader
= pshader
;
3259 entry
->ps_args
= compile_args
;
3260 /* Add the hash table entry */
3261 add_glsl_program_entry(priv
, entry
);
3263 /* Set the current program */
3264 priv
->glsl_program
= entry
;
3266 /* Attach GLSL vshader */
3268 int max_attribs
= 16; /* TODO: Will this always be the case? It is at the moment... */
3271 reorder_shader_id
= generate_param_reorder_function(vshader
, pshader
, gl_info
);
3272 TRACE("Attaching GLSL shader object %u to program %u\n", reorder_shader_id
, programId
);
3273 GL_EXTCALL(glAttachObjectARB(programId
, reorder_shader_id
));
3274 checkGLcall("glAttachObjectARB");
3275 /* Flag the reorder function for deletion, then it will be freed automatically when the program
3278 GL_EXTCALL(glDeleteObjectARB(reorder_shader_id
));
3280 TRACE("Attaching GLSL shader object %u to program %u\n", vshader_id
, programId
);
3281 GL_EXTCALL(glAttachObjectARB(programId
, vshader_id
));
3282 checkGLcall("glAttachObjectARB");
3284 /* Bind vertex attributes to a corresponding index number to match
3285 * the same index numbers as ARB_vertex_programs (makes loading
3286 * vertex attributes simpler). With this method, we can use the
3287 * exact same code to load the attributes later for both ARB and
3290 * We have to do this here because we need to know the Program ID
3291 * in order to make the bindings work, and it has to be done prior
3292 * to linking the GLSL program. */
3293 for (i
= 0; i
< max_attribs
; ++i
) {
3294 if (((IWineD3DBaseShaderImpl
*)vshader
)->baseShader
.reg_maps
.attributes
[i
]) {
3295 snprintf(tmp_name
, sizeof(tmp_name
), "attrib%i", i
);
3296 GL_EXTCALL(glBindAttribLocationARB(programId
, i
, tmp_name
));
3299 checkGLcall("glBindAttribLocationARB");
3301 list_add_head(&((IWineD3DBaseShaderImpl
*)vshader
)->baseShader
.linked_programs
, &entry
->vshader_entry
);
3305 pshader_id
= find_gl_pshader((IWineD3DPixelShaderImpl
*) pshader
, &compile_args
);
3310 /* Attach GLSL pshader */
3312 TRACE("Attaching GLSL shader object %u to program %u\n", pshader_id
, programId
);
3313 GL_EXTCALL(glAttachObjectARB(programId
, pshader_id
));
3314 checkGLcall("glAttachObjectARB");
3316 list_add_head(&((IWineD3DBaseShaderImpl
*)pshader
)->baseShader
.linked_programs
, &entry
->pshader_entry
);
3319 /* Link the program */
3320 TRACE("Linking GLSL shader program %u\n", programId
);
3321 GL_EXTCALL(glLinkProgramARB(programId
));
3322 print_glsl_info_log(&GLINFO_LOCATION
, programId
);
3324 entry
->vuniformF_locations
= HeapAlloc(GetProcessHeap(), 0, sizeof(GLhandleARB
) * GL_LIMITS(vshader_constantsF
));
3325 for (i
= 0; i
< GL_LIMITS(vshader_constantsF
); ++i
) {
3326 snprintf(glsl_name
, sizeof(glsl_name
), "VC[%i]", i
);
3327 entry
->vuniformF_locations
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, glsl_name
));
3329 for (i
= 0; i
< MAX_CONST_I
; ++i
) {
3330 snprintf(glsl_name
, sizeof(glsl_name
), "VI[%i]", i
);
3331 entry
->vuniformI_locations
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, glsl_name
));
3333 entry
->puniformF_locations
= HeapAlloc(GetProcessHeap(), 0, sizeof(GLhandleARB
) * GL_LIMITS(pshader_constantsF
));
3334 for (i
= 0; i
< GL_LIMITS(pshader_constantsF
); ++i
) {
3335 snprintf(glsl_name
, sizeof(glsl_name
), "PC[%i]", i
);
3336 entry
->puniformF_locations
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, glsl_name
));
3338 for (i
= 0; i
< MAX_CONST_I
; ++i
) {
3339 snprintf(glsl_name
, sizeof(glsl_name
), "PI[%i]", i
);
3340 entry
->puniformI_locations
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, glsl_name
));
3344 for(i
= 0; i
< ((IWineD3DPixelShaderImpl
*)pshader
)->numbumpenvmatconsts
; i
++) {
3346 sprintf(name
, "bumpenvmat%d", ((IWineD3DPixelShaderImpl
*)pshader
)->bumpenvmatconst
[i
].texunit
);
3347 entry
->bumpenvmat_location
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, name
));
3348 sprintf(name
, "luminancescale%d", ((IWineD3DPixelShaderImpl
*)pshader
)->luminanceconst
[i
].texunit
);
3349 entry
->luminancescale_location
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, name
));
3350 sprintf(name
, "luminanceoffset%d", ((IWineD3DPixelShaderImpl
*)pshader
)->luminanceconst
[i
].texunit
);
3351 entry
->luminanceoffset_location
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, name
));
3356 entry
->posFixup_location
= GL_EXTCALL(glGetUniformLocationARB(programId
, "posFixup"));
3357 entry
->ycorrection_location
= GL_EXTCALL(glGetUniformLocationARB(programId
, "ycorrection"));
3358 checkGLcall("Find glsl program uniform locations");
3360 if (pshader
&& WINED3DSHADER_VERSION_MAJOR(((IWineD3DPixelShaderImpl
*)pshader
)->baseShader
.hex_version
) >= 3
3361 && ((IWineD3DPixelShaderImpl
*)pshader
)->declared_in_count
> GL_LIMITS(glsl_varyings
) / 4) {
3362 TRACE("Shader %d needs vertex color clamping disabled\n", programId
);
3363 entry
->vertex_color_clamp
= GL_FALSE
;
3365 entry
->vertex_color_clamp
= GL_FIXED_ONLY_ARB
;
3368 /* Set the shader to allow uniform loading on it */
3369 GL_EXTCALL(glUseProgramObjectARB(programId
));
3370 checkGLcall("glUseProgramObjectARB(programId)");
3372 /* Load the vertex and pixel samplers now. The function that finds the mappings makes sure
3373 * that it stays the same for each vertexshader-pixelshader pair(=linked glsl program). If
3374 * a pshader with fixed function pipeline is used there are no vertex samplers, and if a
3375 * vertex shader with fixed function pixel processing is used we make sure that the card
3376 * supports enough samplers to allow the max number of vertex samplers with all possible
3377 * fixed function fragment processing setups. So once the program is linked these samplers
3381 /* Load vertex shader samplers */
3382 shader_glsl_load_vsamplers(gl_info
, (IWineD3DStateBlock
*)This
->stateBlock
, programId
);
3385 /* Load pixel shader samplers */
3386 shader_glsl_load_psamplers(gl_info
, (IWineD3DStateBlock
*)This
->stateBlock
, programId
);
3389 /* If the local constants do not have to be loaded with the environment constants,
3390 * load them now to have them hardcoded in the GLSL program. This saves some CPU cycles
3393 if(pshader
&& !((IWineD3DPixelShaderImpl
*)pshader
)->baseShader
.load_local_constsF
) {
3394 hardcode_local_constants((IWineD3DBaseShaderImpl
*) pshader
, gl_info
, programId
, 'P');
3396 if(vshader
&& !((IWineD3DVertexShaderImpl
*)vshader
)->baseShader
.load_local_constsF
) {
3397 hardcode_local_constants((IWineD3DBaseShaderImpl
*) vshader
, gl_info
, programId
, 'V');
3401 static GLhandleARB
create_glsl_blt_shader(const WineD3D_GL_Info
*gl_info
, enum tex_types tex_type
)
3403 GLhandleARB program_id
;
3404 GLhandleARB vshader_id
, pshader_id
;
3405 const char *blt_vshader
[] = {
3409 " gl_Position = gl_Vertex;\n"
3410 " gl_FrontColor = vec4(1.0);\n"
3411 " gl_TexCoord[0] = gl_MultiTexCoord0;\n"
3415 const char *blt_pshaders
[tex_type_count
] = {
3420 "uniform sampler2D sampler;\n"
3423 " gl_FragDepth = texture2D(sampler, gl_TexCoord[0].xy).x;\n"
3429 "uniform samplerCube sampler;\n"
3432 " gl_FragDepth = textureCube(sampler, gl_TexCoord[0].xyz).x;\n"
3436 "#extension GL_ARB_texture_rectangle : enable\n"
3437 "uniform sampler2DRect sampler;\n"
3440 " gl_FragDepth = texture2DRect(sampler, gl_TexCoord[0].xy).x;\n"
3444 if (!blt_pshaders
[tex_type
])
3446 FIXME("tex_type %#x not supported\n", tex_type
);
3450 vshader_id
= GL_EXTCALL(glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB
));
3451 GL_EXTCALL(glShaderSourceARB(vshader_id
, 1, blt_vshader
, NULL
));
3452 GL_EXTCALL(glCompileShaderARB(vshader_id
));
3454 pshader_id
= GL_EXTCALL(glCreateShaderObjectARB(GL_FRAGMENT_SHADER_ARB
));
3455 GL_EXTCALL(glShaderSourceARB(pshader_id
, 1, &blt_pshaders
[tex_type
], NULL
));
3456 GL_EXTCALL(glCompileShaderARB(pshader_id
));
3458 program_id
= GL_EXTCALL(glCreateProgramObjectARB());
3459 GL_EXTCALL(glAttachObjectARB(program_id
, vshader_id
));
3460 GL_EXTCALL(glAttachObjectARB(program_id
, pshader_id
));
3461 GL_EXTCALL(glLinkProgramARB(program_id
));
3463 print_glsl_info_log(&GLINFO_LOCATION
, program_id
);
3465 /* Once linked we can mark the shaders for deletion. They will be deleted once the program
3468 GL_EXTCALL(glDeleteObjectARB(vshader_id
));
3469 GL_EXTCALL(glDeleteObjectARB(pshader_id
));
3473 static void shader_glsl_select(IWineD3DDevice
*iface
, BOOL usePS
, BOOL useVS
) {
3474 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
3475 struct shader_glsl_priv
*priv
= (struct shader_glsl_priv
*)This
->shader_priv
;
3476 const WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
3477 GLhandleARB program_id
= 0;
3478 GLenum old_vertex_color_clamp
, current_vertex_color_clamp
;
3480 old_vertex_color_clamp
= priv
->glsl_program
? priv
->glsl_program
->vertex_color_clamp
: GL_FIXED_ONLY_ARB
;
3482 if (useVS
|| usePS
) set_glsl_shader_program(iface
, usePS
, useVS
);
3483 else priv
->glsl_program
= NULL
;
3485 current_vertex_color_clamp
= priv
->glsl_program
? priv
->glsl_program
->vertex_color_clamp
: GL_FIXED_ONLY_ARB
;
3487 if (old_vertex_color_clamp
!= current_vertex_color_clamp
) {
3488 if (GL_SUPPORT(ARB_COLOR_BUFFER_FLOAT
)) {
3489 GL_EXTCALL(glClampColorARB(GL_CLAMP_VERTEX_COLOR_ARB
, current_vertex_color_clamp
));
3490 checkGLcall("glClampColorARB");
3492 FIXME("vertex color clamp needs to be changed, but extension not supported.\n");
3496 program_id
= priv
->glsl_program
? priv
->glsl_program
->programId
: 0;
3497 if (program_id
) TRACE("Using GLSL program %u\n", program_id
);
3498 GL_EXTCALL(glUseProgramObjectARB(program_id
));
3499 checkGLcall("glUseProgramObjectARB");
3502 static void shader_glsl_select_depth_blt(IWineD3DDevice
*iface
, enum tex_types tex_type
) {
3503 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
3504 const WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
3505 struct shader_glsl_priv
*priv
= (struct shader_glsl_priv
*) This
->shader_priv
;
3506 GLhandleARB
*blt_program
= &priv
->depth_blt_program
[tex_type
];
3508 if (!*blt_program
) {
3510 *blt_program
= create_glsl_blt_shader(gl_info
, tex_type
);
3511 loc
= GL_EXTCALL(glGetUniformLocationARB(*blt_program
, "sampler"));
3512 GL_EXTCALL(glUseProgramObjectARB(*blt_program
));
3513 GL_EXTCALL(glUniform1iARB(loc
, 0));
3515 GL_EXTCALL(glUseProgramObjectARB(*blt_program
));
3519 static void shader_glsl_deselect_depth_blt(IWineD3DDevice
*iface
) {
3520 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
3521 const WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
3522 struct shader_glsl_priv
*priv
= (struct shader_glsl_priv
*) This
->shader_priv
;
3523 GLhandleARB program_id
;
3525 program_id
= priv
->glsl_program
? priv
->glsl_program
->programId
: 0;
3526 if (program_id
) TRACE("Using GLSL program %u\n", program_id
);
3528 GL_EXTCALL(glUseProgramObjectARB(program_id
));
3529 checkGLcall("glUseProgramObjectARB");
3532 static void shader_glsl_cleanup(IWineD3DDevice
*iface
) {
3533 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
3534 const WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
3535 GL_EXTCALL(glUseProgramObjectARB(0));
3538 static void shader_glsl_destroy(IWineD3DBaseShader
*iface
) {
3539 const struct list
*linked_programs
;
3540 IWineD3DBaseShaderImpl
*This
= (IWineD3DBaseShaderImpl
*) iface
;
3541 IWineD3DDeviceImpl
*device
= (IWineD3DDeviceImpl
*)This
->baseShader
.device
;
3542 struct shader_glsl_priv
*priv
= (struct shader_glsl_priv
*)device
->shader_priv
;
3543 const WineD3D_GL_Info
*gl_info
= &device
->adapter
->gl_info
;
3544 IWineD3DPixelShaderImpl
*ps
= NULL
;
3545 IWineD3DVertexShaderImpl
*vs
= NULL
;
3547 /* Note: Do not use QueryInterface here to find out which shader type this is because this code
3548 * can be called from IWineD3DBaseShader::Release
3550 char pshader
= shader_is_pshader_version(This
->baseShader
.hex_version
);
3553 ps
= (IWineD3DPixelShaderImpl
*) This
;
3554 if(ps
->num_gl_shaders
== 0) return;
3556 vs
= (IWineD3DVertexShaderImpl
*) This
;
3557 if(vs
->prgId
== 0) return;
3560 linked_programs
= &This
->baseShader
.linked_programs
;
3562 TRACE("Deleting linked programs\n");
3563 if (linked_programs
->next
) {
3564 struct glsl_shader_prog_link
*entry
, *entry2
;
3567 LIST_FOR_EACH_ENTRY_SAFE(entry
, entry2
, linked_programs
, struct glsl_shader_prog_link
, pshader_entry
) {
3568 delete_glsl_program_entry(priv
, gl_info
, entry
);
3571 LIST_FOR_EACH_ENTRY_SAFE(entry
, entry2
, linked_programs
, struct glsl_shader_prog_link
, vshader_entry
) {
3572 delete_glsl_program_entry(priv
, gl_info
, entry
);
3581 for(i
= 0; i
< ps
->num_gl_shaders
; i
++) {
3582 TRACE("deleting pshader %u\n", ps
->gl_shaders
[i
].prgId
);
3583 GL_EXTCALL(glDeleteObjectARB(ps
->gl_shaders
[i
].prgId
));
3584 checkGLcall("glDeleteObjectARB");
3587 HeapFree(GetProcessHeap(), 0, ps
->gl_shaders
);
3588 ps
->gl_shaders
= NULL
;
3589 ps
->num_gl_shaders
= 0;
3591 TRACE("Deleting shader object %u\n", vs
->prgId
);
3593 GL_EXTCALL(glDeleteObjectARB(vs
->prgId
));
3594 checkGLcall("glDeleteObjectARB");
3597 vs
->baseShader
.is_compiled
= FALSE
;
3601 static unsigned int glsl_program_key_hash(const void *key
)
3603 const glsl_program_key_t
*k
= (const glsl_program_key_t
*)key
;
3605 unsigned int hash
= k
->vshader
| ((DWORD_PTR
) k
->pshader
) << 16;
3606 hash
+= ~(hash
<< 15);
3607 hash
^= (hash
>> 10);
3608 hash
+= (hash
<< 3);
3609 hash
^= (hash
>> 6);
3610 hash
+= ~(hash
<< 11);
3611 hash
^= (hash
>> 16);
3616 static BOOL
glsl_program_key_compare(const void *keya
, const void *keyb
)
3618 const glsl_program_key_t
*ka
= (const glsl_program_key_t
*)keya
;
3619 const glsl_program_key_t
*kb
= (const glsl_program_key_t
*)keyb
;
3621 return ka
->vshader
== kb
->vshader
&& ka
->pshader
== kb
->pshader
&&
3622 (memcmp(&ka
->ps_args
, &kb
->ps_args
, sizeof(kb
->ps_args
)) == 0);
3625 static HRESULT
shader_glsl_alloc(IWineD3DDevice
*iface
) {
3626 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
3627 struct shader_glsl_priv
*priv
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(struct shader_glsl_priv
));
3628 priv
->glsl_program_lookup
= hash_table_create(glsl_program_key_hash
, glsl_program_key_compare
);
3629 This
->shader_priv
= priv
;
3633 static void shader_glsl_free(IWineD3DDevice
*iface
) {
3634 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
3635 const WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
3636 struct shader_glsl_priv
*priv
= (struct shader_glsl_priv
*)This
->shader_priv
;
3639 for (i
= 0; i
< tex_type_count
; ++i
)
3641 if (priv
->depth_blt_program
[i
])
3643 GL_EXTCALL(glDeleteObjectARB(priv
->depth_blt_program
[i
]));
3647 hash_table_destroy(priv
->glsl_program_lookup
, NULL
, NULL
);
3649 HeapFree(GetProcessHeap(), 0, This
->shader_priv
);
3650 This
->shader_priv
= NULL
;
3653 static BOOL
shader_glsl_dirty_const(IWineD3DDevice
*iface
) {
3654 /* TODO: GL_EXT_bindable_uniform can be used to share constants across shaders */
3658 static GLuint
shader_glsl_generate_pshader(IWineD3DPixelShader
*iface
, SHADER_BUFFER
*buffer
) {
3659 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*)iface
;
3660 const struct shader_reg_maps
*reg_maps
= &This
->baseShader
.reg_maps
;
3661 CONST DWORD
*function
= This
->baseShader
.function
;
3662 const char *fragcolor
;
3663 const WineD3D_GL_Info
*gl_info
= &((IWineD3DDeviceImpl
*)This
->baseShader
.device
)->adapter
->gl_info
;
3665 /* Create the hw GLSL shader object and assign it as the shader->prgId */
3666 GLhandleARB shader_obj
= GL_EXTCALL(glCreateShaderObjectARB(GL_FRAGMENT_SHADER_ARB
));
3668 shader_addline(buffer
, "#version 120\n");
3670 if (GL_SUPPORT(ARB_DRAW_BUFFERS
)) {
3671 shader_addline(buffer
, "#extension GL_ARB_draw_buffers : enable\n");
3673 if (GL_SUPPORT(ARB_TEXTURE_RECTANGLE
)) {
3674 /* The spec says that it doesn't have to be explicitly enabled, but the nvidia
3675 * drivers write a warning if we don't do so
3677 shader_addline(buffer
, "#extension GL_ARB_texture_rectangle : enable\n");
3680 /* Base Declarations */
3681 shader_generate_glsl_declarations( (IWineD3DBaseShader
*) This
, reg_maps
, buffer
, &GLINFO_LOCATION
);
3683 /* Pack 3.0 inputs */
3684 if (This
->baseShader
.hex_version
>= WINED3DPS_VERSION(3,0)) {
3686 if(((IWineD3DDeviceImpl
*) This
->baseShader
.device
)->strided_streams
.u
.s
.position_transformed
) {
3687 pshader_glsl_input_pack(buffer
, This
->semantics_in
, iface
, pretransformed
);
3688 } else if(!use_vs((IWineD3DDeviceImpl
*) This
->baseShader
.device
)) {
3689 pshader_glsl_input_pack(buffer
, This
->semantics_in
, iface
, fixedfunction
);
3693 /* Base Shader Body */
3694 shader_generate_main( (IWineD3DBaseShader
*) This
, buffer
, reg_maps
, function
);
3696 /* Pixel shaders < 2.0 place the resulting color in R0 implicitly */
3697 if (This
->baseShader
.hex_version
< WINED3DPS_VERSION(2,0)) {
3698 /* Some older cards like GeforceFX ones don't support multiple buffers, so also not gl_FragData */
3699 if(GL_SUPPORT(ARB_DRAW_BUFFERS
))
3700 shader_addline(buffer
, "gl_FragData[0] = R0;\n");
3702 shader_addline(buffer
, "gl_FragColor = R0;\n");
3705 if(GL_SUPPORT(ARB_DRAW_BUFFERS
)) {
3706 fragcolor
= "gl_FragData[0]";
3708 fragcolor
= "gl_FragColor";
3710 if(((IWineD3DDeviceImpl
*)This
->baseShader
.device
)->stateBlock
->renderState
[WINED3DRS_SRGBWRITEENABLE
]) {
3711 shader_addline(buffer
, "tmp0.xyz = pow(%s.xyz, vec3(%f, %f, %f)) * vec3(%f, %f, %f) - vec3(%f, %f, %f);\n",
3712 fragcolor
, srgb_pow
, srgb_pow
, srgb_pow
, srgb_mul_high
, srgb_mul_high
, srgb_mul_high
,
3713 srgb_sub_high
, srgb_sub_high
, srgb_sub_high
);
3714 shader_addline(buffer
, "tmp1.xyz = %s.xyz * srgb_mul_low.xyz;\n", fragcolor
);
3715 shader_addline(buffer
, "%s.x = %s.x < srgb_comparison.x ? tmp1.x : tmp0.x;\n", fragcolor
, fragcolor
);
3716 shader_addline(buffer
, "%s.y = %s.y < srgb_comparison.y ? tmp1.y : tmp0.y;\n", fragcolor
, fragcolor
);
3717 shader_addline(buffer
, "%s.z = %s.z < srgb_comparison.z ? tmp1.z : tmp0.z;\n", fragcolor
, fragcolor
);
3718 shader_addline(buffer
, "%s = clamp(%s, 0.0, 1.0);\n", fragcolor
, fragcolor
);
3720 /* Pixel shader < 3.0 do not replace the fog stage.
3721 * This implements linear fog computation and blending.
3722 * TODO: non linear fog
3723 * NOTE: gl_Fog.start and gl_Fog.end don't hold fog start s and end e but
3724 * -1/(e-s) and e/(e-s) respectively.
3726 if(This
->baseShader
.hex_version
< WINED3DPS_VERSION(3,0)) {
3727 shader_addline(buffer
, "float Fog = clamp(gl_FogFragCoord * gl_Fog.start + gl_Fog.end, 0.0, 1.0);\n");
3728 shader_addline(buffer
, "%s.xyz = mix(gl_Fog.color.xyz, %s.xyz, Fog);\n", fragcolor
, fragcolor
);
3731 shader_addline(buffer
, "}\n");
3733 TRACE("Compiling shader object %u\n", shader_obj
);
3734 GL_EXTCALL(glShaderSourceARB(shader_obj
, 1, (const char**)&buffer
->buffer
, NULL
));
3735 GL_EXTCALL(glCompileShaderARB(shader_obj
));
3736 print_glsl_info_log(&GLINFO_LOCATION
, shader_obj
);
3738 /* Store the shader object */
3742 static void shader_glsl_generate_vshader(IWineD3DVertexShader
*iface
, SHADER_BUFFER
*buffer
) {
3743 IWineD3DVertexShaderImpl
*This
= (IWineD3DVertexShaderImpl
*)iface
;
3744 const struct shader_reg_maps
*reg_maps
= &This
->baseShader
.reg_maps
;
3745 CONST DWORD
*function
= This
->baseShader
.function
;
3746 const WineD3D_GL_Info
*gl_info
= &((IWineD3DDeviceImpl
*)This
->baseShader
.device
)->adapter
->gl_info
;
3748 /* Create the hw GLSL shader program and assign it as the shader->prgId */
3749 GLhandleARB shader_obj
= GL_EXTCALL(glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB
));
3751 shader_addline(buffer
, "#version 120\n");
3753 /* Base Declarations */
3754 shader_generate_glsl_declarations( (IWineD3DBaseShader
*) This
, reg_maps
, buffer
, &GLINFO_LOCATION
);
3756 /* Base Shader Body */
3757 shader_generate_main( (IWineD3DBaseShader
*) This
, buffer
, reg_maps
, function
);
3759 /* Unpack 3.0 outputs */
3760 if (This
->baseShader
.hex_version
>= WINED3DVS_VERSION(3,0)) {
3761 shader_addline(buffer
, "order_ps_input(OUT);\n");
3763 shader_addline(buffer
, "order_ps_input();\n");
3766 /* If this shader doesn't use fog copy the z coord to the fog coord so that we can use table fog */
3768 shader_addline(buffer
, "gl_FogFragCoord = gl_Position.z;\n");
3770 /* Write the final position.
3772 * OpenGL coordinates specify the center of the pixel while d3d coords specify
3773 * the corner. The offsets are stored in z and w in posFixup. posFixup.y contains
3774 * 1.0 or -1.0 to turn the rendering upside down for offscreen rendering. PosFixup.x
3775 * contains 1.0 to allow a mad.
3777 shader_addline(buffer
, "gl_Position.y = gl_Position.y * posFixup.y;\n");
3778 shader_addline(buffer
, "gl_Position.xy += posFixup.zw * gl_Position.ww;\n");
3780 /* Z coord [0;1]->[-1;1] mapping, see comment in transform_projection in state.c
3782 * Basically we want (in homogeneous coordinates) z = z * 2 - 1. However, shaders are run
3783 * before the homogeneous divide, so we have to take the w into account: z = ((z / w) * 2 - 1) * w,
3784 * which is the same as z = z * 2 - w.
3786 shader_addline(buffer
, "gl_Position.z = gl_Position.z * 2.0 - gl_Position.w;\n");
3788 shader_addline(buffer
, "}\n");
3790 TRACE("Compiling shader object %u\n", shader_obj
);
3791 GL_EXTCALL(glShaderSourceARB(shader_obj
, 1, (const char**)&buffer
->buffer
, NULL
));
3792 GL_EXTCALL(glCompileShaderARB(shader_obj
));
3793 print_glsl_info_log(&GLINFO_LOCATION
, shader_obj
);
3795 /* Store the shader object */
3796 This
->prgId
= shader_obj
;
3799 static void shader_glsl_get_caps(WINED3DDEVTYPE devtype
, const WineD3D_GL_Info
*gl_info
, struct shader_caps
*pCaps
)
3801 /* Nvidia Geforce6/7 or Ati R4xx/R5xx cards with GLSL support, support VS 3.0 but older Nvidia/Ati
3802 * models with GLSL support only support 2.0. In case of nvidia we can detect VS 2.0 support using
3803 * vs_nv_version which is based on NV_vertex_program.
3804 * For Ati cards there's no way using glsl (it abstracts the lowlevel info away) and also not
3805 * using ARB_vertex_program. It is safe to assume that when a card supports pixel shader 2.0 it
3806 * supports vertex shader 2.0 too and the way around. We can detect ps2.0 using the maximum number
3807 * of native instructions, so use that here. For more info see the pixel shader versioning code below.
3809 if((GLINFO_LOCATION
.vs_nv_version
== VS_VERSION_20
) || (GLINFO_LOCATION
.ps_arb_max_instructions
<= 512))
3810 pCaps
->VertexShaderVersion
= WINED3DVS_VERSION(2,0);
3812 pCaps
->VertexShaderVersion
= WINED3DVS_VERSION(3,0);
3813 TRACE_(d3d_caps
)("Hardware vertex shader version %d.%d enabled (GLSL)\n", (pCaps
->VertexShaderVersion
>> 8) & 0xff, pCaps
->VertexShaderVersion
& 0xff);
3814 pCaps
->MaxVertexShaderConst
= GL_LIMITS(vshader_constantsF
);
3816 /* Older DX9-class videocards (GeforceFX / Radeon >9500/X*00) only support pixel shader 2.0/2.0a/2.0b.
3817 * In OpenGL the extensions related to GLSL abstract lowlevel GL info away which is needed
3818 * to distinguish between 2.0 and 3.0 (and 2.0a/2.0b). In case of Nvidia we use their fragment
3819 * program extensions. On other hardware including ATI GL_ARB_fragment_program offers the info
3820 * in max native instructions. Intel and others also offer the info in this extension but they
3821 * don't support GLSL (at least on Windows).
3823 * PS2.0 requires at least 96 instructions, 2.0a/2.0b go up to 512. Assume that if the number
3824 * of instructions is 512 or less we have to do with ps2.0 hardware.
3825 * 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.
3827 if((GLINFO_LOCATION
.ps_nv_version
== PS_VERSION_20
) || (GLINFO_LOCATION
.ps_arb_max_instructions
<= 512))
3828 pCaps
->PixelShaderVersion
= WINED3DPS_VERSION(2,0);
3830 pCaps
->PixelShaderVersion
= WINED3DPS_VERSION(3,0);
3832 /* FIXME: The following line is card dependent. -8.0 to 8.0 is the
3833 * Direct3D minimum requirement.
3835 * Both GL_ARB_fragment_program and GLSL require a "maximum representable magnitude"
3836 * of colors to be 2^10, and 2^32 for other floats. Should we use 1024 here?
3838 * The problem is that the refrast clamps temporary results in the shader to
3839 * [-MaxValue;+MaxValue]. If the card's max value is bigger than the one we advertize here,
3840 * then applications may miss the clamping behavior. On the other hand, if it is smaller,
3841 * the shader will generate incorrect results too. Unfortunately, GL deliberately doesn't
3842 * offer a way to query this.
3844 pCaps
->PixelShader1xMaxValue
= 8.0;
3845 TRACE_(d3d_caps
)("Hardware pixel shader version %d.%d enabled (GLSL)\n", (pCaps
->PixelShaderVersion
>> 8) & 0xff, pCaps
->PixelShaderVersion
& 0xff);
3848 static BOOL
shader_glsl_conv_supported(WINED3DFORMAT fmt
) {
3849 TRACE("Checking shader format support for format %s:", debug_d3dformat(fmt
));
3851 case WINED3DFMT_V8U8
:
3852 case WINED3DFMT_V16U16
:
3853 case WINED3DFMT_X8L8V8U8
:
3854 case WINED3DFMT_L6V5U5
:
3855 case WINED3DFMT_Q8W8V8U8
:
3856 case WINED3DFMT_ATI2N
:
3865 static const SHADER_HANDLER shader_glsl_instruction_handler_table
[WINED3DSIH_TABLE_SIZE
] =
3867 /* WINED3DSIH_ABS */ shader_glsl_map2gl
,
3868 /* WINED3DSIH_ADD */ shader_glsl_arith
,
3869 /* WINED3DSIH_BEM */ pshader_glsl_bem
,
3870 /* WINED3DSIH_BREAK */ shader_glsl_break
,
3871 /* WINED3DSIH_BREAKC */ shader_glsl_breakc
,
3872 /* WINED3DSIH_BREAKP */ NULL
,
3873 /* WINED3DSIH_CALL */ shader_glsl_call
,
3874 /* WINED3DSIH_CALLNZ */ shader_glsl_callnz
,
3875 /* WINED3DSIH_CMP */ shader_glsl_cmp
,
3876 /* WINED3DSIH_CND */ shader_glsl_cnd
,
3877 /* WINED3DSIH_CRS */ shader_glsl_cross
,
3878 /* WINED3DSIH_DCL */ NULL
,
3879 /* WINED3DSIH_DEF */ NULL
,
3880 /* WINED3DSIH_DEFB */ NULL
,
3881 /* WINED3DSIH_DEFI */ NULL
,
3882 /* WINED3DSIH_DP2ADD */ pshader_glsl_dp2add
,
3883 /* WINED3DSIH_DP3 */ shader_glsl_dot
,
3884 /* WINED3DSIH_DP4 */ shader_glsl_dot
,
3885 /* WINED3DSIH_DST */ shader_glsl_dst
,
3886 /* WINED3DSIH_DSX */ shader_glsl_map2gl
,
3887 /* WINED3DSIH_DSY */ shader_glsl_map2gl
,
3888 /* WINED3DSIH_ELSE */ shader_glsl_else
,
3889 /* WINED3DSIH_ENDIF */ shader_glsl_end
,
3890 /* WINED3DSIH_ENDLOOP */ shader_glsl_end
,
3891 /* WINED3DSIH_ENDREP */ shader_glsl_end
,
3892 /* WINED3DSIH_EXP */ shader_glsl_map2gl
,
3893 /* WINED3DSIH_EXPP */ shader_glsl_expp
,
3894 /* WINED3DSIH_FRC */ shader_glsl_map2gl
,
3895 /* WINED3DSIH_IF */ shader_glsl_if
,
3896 /* WINED3DSIH_IFC */ shader_glsl_ifc
,
3897 /* WINED3DSIH_LABEL */ shader_glsl_label
,
3898 /* WINED3DSIH_LIT */ shader_glsl_lit
,
3899 /* WINED3DSIH_LOG */ shader_glsl_log
,
3900 /* WINED3DSIH_LOGP */ shader_glsl_log
,
3901 /* WINED3DSIH_LOOP */ shader_glsl_loop
,
3902 /* WINED3DSIH_LRP */ shader_glsl_lrp
,
3903 /* WINED3DSIH_M3x2 */ shader_glsl_mnxn
,
3904 /* WINED3DSIH_M3x3 */ shader_glsl_mnxn
,
3905 /* WINED3DSIH_M3x4 */ shader_glsl_mnxn
,
3906 /* WINED3DSIH_M4x3 */ shader_glsl_mnxn
,
3907 /* WINED3DSIH_M4x4 */ shader_glsl_mnxn
,
3908 /* WINED3DSIH_MAD */ shader_glsl_mad
,
3909 /* WINED3DSIH_MAX */ shader_glsl_map2gl
,
3910 /* WINED3DSIH_MIN */ shader_glsl_map2gl
,
3911 /* WINED3DSIH_MOV */ shader_glsl_mov
,
3912 /* WINED3DSIH_MOVA */ shader_glsl_mov
,
3913 /* WINED3DSIH_MUL */ shader_glsl_arith
,
3914 /* WINED3DSIH_NOP */ NULL
,
3915 /* WINED3DSIH_NRM */ shader_glsl_map2gl
,
3916 /* WINED3DSIH_PHASE */ NULL
,
3917 /* WINED3DSIH_POW */ shader_glsl_pow
,
3918 /* WINED3DSIH_RCP */ shader_glsl_rcp
,
3919 /* WINED3DSIH_REP */ shader_glsl_rep
,
3920 /* WINED3DSIH_RET */ NULL
,
3921 /* WINED3DSIH_RSQ */ shader_glsl_rsq
,
3922 /* WINED3DSIH_SETP */ NULL
,
3923 /* WINED3DSIH_SGE */ shader_glsl_compare
,
3924 /* WINED3DSIH_SGN */ shader_glsl_map2gl
,
3925 /* WINED3DSIH_SINCOS */ shader_glsl_sincos
,
3926 /* WINED3DSIH_SLT */ shader_glsl_compare
,
3927 /* WINED3DSIH_SUB */ shader_glsl_arith
,
3928 /* WINED3DSIH_TEX */ pshader_glsl_tex
,
3929 /* WINED3DSIH_TEXBEM */ pshader_glsl_texbem
,
3930 /* WINED3DSIH_TEXBEML */ pshader_glsl_texbem
,
3931 /* WINED3DSIH_TEXCOORD */ pshader_glsl_texcoord
,
3932 /* WINED3DSIH_TEXDEPTH */ pshader_glsl_texdepth
,
3933 /* WINED3DSIH_TEXDP3 */ pshader_glsl_texdp3
,
3934 /* WINED3DSIH_TEXDP3TEX */ pshader_glsl_texdp3tex
,
3935 /* WINED3DSIH_TEXKILL */ pshader_glsl_texkill
,
3936 /* WINED3DSIH_TEXLDD */ NULL
,
3937 /* WINED3DSIH_TEXLDL */ shader_glsl_texldl
,
3938 /* WINED3DSIH_TEXM3x2DEPTH */ pshader_glsl_texm3x2depth
,
3939 /* WINED3DSIH_TEXM3x2PAD */ pshader_glsl_texm3x2pad
,
3940 /* WINED3DSIH_TEXM3x2TEX */ pshader_glsl_texm3x2tex
,
3941 /* WINED3DSIH_TEXM3x3 */ pshader_glsl_texm3x3
,
3942 /* WINED3DSIH_TEXM3x3DIFF */ NULL
,
3943 /* WINED3DSIH_TEXM3x3PAD */ pshader_glsl_texm3x3pad
,
3944 /* WINED3DSIH_TEXM3x3SPEC */ pshader_glsl_texm3x3spec
,
3945 /* WINED3DSIH_TEXM3x3TEX */ pshader_glsl_texm3x3tex
,
3946 /* WINED3DSIH_TEXM3x3VSPEC */ pshader_glsl_texm3x3vspec
,
3947 /* WINED3DSIH_TEXREG2AR */ pshader_glsl_texreg2ar
,
3948 /* WINED3DSIH_TEXREG2GB */ pshader_glsl_texreg2gb
,
3949 /* WINED3DSIH_TEXREG2RGB */ pshader_glsl_texreg2rgb
,
3952 const shader_backend_t glsl_shader_backend
= {
3953 shader_glsl_instruction_handler_table
,
3955 shader_glsl_select_depth_blt
,
3956 shader_glsl_deselect_depth_blt
,
3957 shader_glsl_load_constants
,
3958 shader_glsl_cleanup
,
3959 shader_glsl_color_correction
,
3960 shader_glsl_destroy
,
3963 shader_glsl_dirty_const
,
3964 shader_glsl_generate_pshader
,
3965 shader_glsl_generate_vshader
,
3966 shader_glsl_get_caps
,
3967 shader_glsl_conv_supported
,