2 * GLSL pixel and vertex shader implementation
4 * Copyright 2006 Jason Green
5 * Copyright 2006-2007 Henri Verbeet
6 * Copyright 2007-2008 Stefan Dösinger for CodeWeavers
8 * This library is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU Lesser General Public
10 * License as published by the Free Software Foundation; either
11 * version 2.1 of the License, or (at your option) any later version.
13 * This library is distributed in the hope that it will be useful,
14 * but WITHOUT ANY WARRANTY; without even the implied warranty of
15 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
16 * Lesser General Public License for more details.
18 * You should have received a copy of the GNU Lesser General Public
19 * License along with this library; if not, write to the Free Software
20 * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301, USA
24 * D3D shader asm has swizzles on source parameters, and write masks for
25 * destination parameters. GLSL uses swizzles for both. The result of this is
26 * that for example "mov dst.xw, src.zyxw" becomes "dst.xw = src.zw" in GLSL.
27 * Ie, to generate a proper GLSL source swizzle, we need to take the D3D write
28 * mask for the destination parameter into account.
34 #include "wined3d_private.h"
36 WINE_DEFAULT_DEBUG_CHANNEL(d3d_shader
);
37 WINE_DECLARE_DEBUG_CHANNEL(d3d_constants
);
38 WINE_DECLARE_DEBUG_CHANNEL(d3d_caps
);
39 WINE_DECLARE_DEBUG_CHANNEL(d3d
);
41 #define GLINFO_LOCATION (*gl_info)
43 #define WINED3D_GLSL_SAMPLE_PROJECTED 0x1
44 #define WINED3D_GLSL_SAMPLE_RECT 0x2
45 #define WINED3D_GLSL_SAMPLE_LOD 0x4
60 } glsl_sample_function_t
;
64 HEAP_NODE_TRAVERSE_LEFT
,
65 HEAP_NODE_TRAVERSE_RIGHT
,
77 struct constant_entry
*entries
;
78 unsigned int *positions
;
82 /* GLSL shader private data */
83 struct shader_glsl_priv
{
84 struct hash_table_t
*glsl_program_lookup
;
85 struct glsl_shader_prog_link
*glsl_program
;
86 struct constant_heap vconst_heap
;
87 struct constant_heap pconst_heap
;
89 GLhandleARB depth_blt_program
[tex_type_count
];
90 UINT next_constant_version
;
93 /* Struct to maintain data about a linked GLSL program */
94 struct glsl_shader_prog_link
{
95 struct list vshader_entry
;
96 struct list pshader_entry
;
97 GLhandleARB programId
;
98 GLint
*vuniformF_locations
;
99 GLint
*puniformF_locations
;
100 GLint vuniformI_locations
[MAX_CONST_I
];
101 GLint puniformI_locations
[MAX_CONST_I
];
102 GLint posFixup_location
;
103 GLint rectFixup_location
[MAX_FRAGMENT_SAMPLERS
];
104 GLint bumpenvmat_location
[MAX_TEXTURES
];
105 GLint luminancescale_location
[MAX_TEXTURES
];
106 GLint luminanceoffset_location
[MAX_TEXTURES
];
107 GLint ycorrection_location
;
108 GLenum vertex_color_clamp
;
109 IWineD3DVertexShader
*vshader
;
110 IWineD3DPixelShader
*pshader
;
111 struct vs_compile_args vs_args
;
112 struct ps_compile_args ps_args
;
113 UINT constant_version
;
117 IWineD3DVertexShader
*vshader
;
118 IWineD3DPixelShader
*pshader
;
119 struct ps_compile_args ps_args
;
120 struct vs_compile_args vs_args
;
121 } glsl_program_key_t
;
124 /** Prints the GLSL info log which will contain error messages if they exist */
125 static void print_glsl_info_log(const WineD3D_GL_Info
*gl_info
, GLhandleARB obj
)
127 int infologLength
= 0;
132 static const char * const spam
[] =
134 "Vertex shader was successfully compiled to run on hardware.\n", /* fglrx */
135 "Fragment shader was successfully compiled to run on hardware.\n", /* fglrx */
136 "Fragment shader(s) linked, vertex shader(s) linked. \n ", /* fglrx, with \n */
137 "Fragment shader(s) linked, vertex shader(s) linked.", /* fglrx, no \n */
138 "Vertex shader(s) linked, no fragment shader(s) defined. \n ", /* fglrx, with \n */
139 "Vertex shader(s) linked, no fragment shader(s) defined.", /* fglrx, no \n */
140 "Fragment shader was successfully compiled to run on hardware.\n"
141 "WARNING: 0:2: extension 'GL_ARB_draw_buffers' is not supported",
142 "Fragment shader(s) linked, no vertex shader(s) defined.", /* fglrx, no \n */
143 "Fragment shader(s) linked, no vertex shader(s) defined. \n ", /* fglrx, with \n */
144 "WARNING: 0:2: extension 'GL_ARB_draw_buffers' is not supported\n" /* MacOS ati */
147 if (!TRACE_ON(d3d_shader
) && !FIXME_ON(d3d_shader
)) return;
149 GL_EXTCALL(glGetObjectParameterivARB(obj
,
150 GL_OBJECT_INFO_LOG_LENGTH_ARB
,
153 /* A size of 1 is just a null-terminated string, so the log should be bigger than
154 * that if there are errors. */
155 if (infologLength
> 1)
157 /* Fglrx doesn't terminate the string properly, but it tells us the proper length.
158 * So use HEAP_ZERO_MEMORY to avoid uninitialized bytes
160 infoLog
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, infologLength
);
161 GL_EXTCALL(glGetInfoLogARB(obj
, infologLength
, NULL
, infoLog
));
164 for(i
= 0; i
< sizeof(spam
) / sizeof(spam
[0]); i
++) {
165 if(strcmp(infoLog
, spam
[i
]) == 0) {
171 TRACE("Spam received from GLSL shader #%u: %s\n", obj
, debugstr_a(infoLog
));
173 FIXME("Error received from GLSL shader #%u: %s\n", obj
, debugstr_a(infoLog
));
175 HeapFree(GetProcessHeap(), 0, infoLog
);
180 * Loads (pixel shader) samplers
182 static void shader_glsl_load_psamplers(const WineD3D_GL_Info
*gl_info
, DWORD
*tex_unit_map
, GLhandleARB programId
)
186 char sampler_name
[20];
188 for (i
= 0; i
< MAX_FRAGMENT_SAMPLERS
; ++i
) {
189 snprintf(sampler_name
, sizeof(sampler_name
), "Psampler%d", i
);
190 name_loc
= GL_EXTCALL(glGetUniformLocationARB(programId
, sampler_name
));
191 if (name_loc
!= -1) {
192 DWORD mapped_unit
= tex_unit_map
[i
];
193 if (mapped_unit
!= WINED3D_UNMAPPED_STAGE
&& mapped_unit
< GL_LIMITS(fragment_samplers
))
195 TRACE("Loading %s for texture %d\n", sampler_name
, mapped_unit
);
196 GL_EXTCALL(glUniform1iARB(name_loc
, mapped_unit
));
197 checkGLcall("glUniform1iARB");
199 ERR("Trying to load sampler %s on unsupported unit %d\n", sampler_name
, mapped_unit
);
205 static void shader_glsl_load_vsamplers(const WineD3D_GL_Info
*gl_info
, DWORD
*tex_unit_map
, GLhandleARB programId
)
208 char sampler_name
[20];
211 for (i
= 0; i
< MAX_VERTEX_SAMPLERS
; ++i
) {
212 snprintf(sampler_name
, sizeof(sampler_name
), "Vsampler%d", i
);
213 name_loc
= GL_EXTCALL(glGetUniformLocationARB(programId
, sampler_name
));
214 if (name_loc
!= -1) {
215 DWORD mapped_unit
= tex_unit_map
[MAX_FRAGMENT_SAMPLERS
+ i
];
216 if (mapped_unit
!= WINED3D_UNMAPPED_STAGE
&& mapped_unit
< GL_LIMITS(combined_samplers
))
218 TRACE("Loading %s for texture %d\n", sampler_name
, mapped_unit
);
219 GL_EXTCALL(glUniform1iARB(name_loc
, mapped_unit
));
220 checkGLcall("glUniform1iARB");
222 ERR("Trying to load sampler %s on unsupported unit %d\n", sampler_name
, mapped_unit
);
228 static inline void walk_constant_heap(const WineD3D_GL_Info
*gl_info
, const float *constants
,
229 const GLint
*constant_locations
, const struct constant_heap
*heap
, unsigned char *stack
, DWORD version
)
232 unsigned int heap_idx
= 1;
235 if (heap
->entries
[heap_idx
].version
<= version
) return;
237 idx
= heap
->entries
[heap_idx
].idx
;
238 if (constant_locations
[idx
] != -1) GL_EXTCALL(glUniform4fvARB(constant_locations
[idx
], 1, &constants
[idx
* 4]));
239 stack
[stack_idx
] = HEAP_NODE_TRAVERSE_LEFT
;
241 while (stack_idx
>= 0)
243 /* Note that we fall through to the next case statement. */
244 switch(stack
[stack_idx
])
246 case HEAP_NODE_TRAVERSE_LEFT
:
248 unsigned int left_idx
= heap_idx
<< 1;
249 if (left_idx
< heap
->size
&& heap
->entries
[left_idx
].version
> version
)
252 idx
= heap
->entries
[heap_idx
].idx
;
253 if (constant_locations
[idx
] != -1)
254 GL_EXTCALL(glUniform4fvARB(constant_locations
[idx
], 1, &constants
[idx
* 4]));
256 stack
[stack_idx
++] = HEAP_NODE_TRAVERSE_RIGHT
;
257 stack
[stack_idx
] = HEAP_NODE_TRAVERSE_LEFT
;
262 case HEAP_NODE_TRAVERSE_RIGHT
:
264 unsigned int right_idx
= (heap_idx
<< 1) + 1;
265 if (right_idx
< heap
->size
&& heap
->entries
[right_idx
].version
> version
)
267 heap_idx
= right_idx
;
268 idx
= heap
->entries
[heap_idx
].idx
;
269 if (constant_locations
[idx
] != -1)
270 GL_EXTCALL(glUniform4fvARB(constant_locations
[idx
], 1, &constants
[idx
* 4]));
272 stack
[stack_idx
++] = HEAP_NODE_POP
;
273 stack
[stack_idx
] = HEAP_NODE_TRAVERSE_LEFT
;
286 checkGLcall("walk_constant_heap()");
289 static inline void apply_clamped_constant(const WineD3D_GL_Info
*gl_info
, GLint location
, const GLfloat
*data
)
291 GLfloat clamped_constant
[4];
293 if (location
== -1) return;
295 clamped_constant
[0] = data
[0] < -1.0f
? -1.0f
: data
[0] > 1.0 ? 1.0 : data
[0];
296 clamped_constant
[1] = data
[1] < -1.0f
? -1.0f
: data
[1] > 1.0 ? 1.0 : data
[1];
297 clamped_constant
[2] = data
[2] < -1.0f
? -1.0f
: data
[2] > 1.0 ? 1.0 : data
[2];
298 clamped_constant
[3] = data
[3] < -1.0f
? -1.0f
: data
[3] > 1.0 ? 1.0 : data
[3];
300 GL_EXTCALL(glUniform4fvARB(location
, 1, clamped_constant
));
303 static inline void walk_constant_heap_clamped(const WineD3D_GL_Info
*gl_info
, const float *constants
,
304 const GLint
*constant_locations
, const struct constant_heap
*heap
, unsigned char *stack
, DWORD version
)
307 unsigned int heap_idx
= 1;
310 if (heap
->entries
[heap_idx
].version
<= version
) return;
312 idx
= heap
->entries
[heap_idx
].idx
;
313 apply_clamped_constant(gl_info
, constant_locations
[idx
], &constants
[idx
* 4]);
314 stack
[stack_idx
] = HEAP_NODE_TRAVERSE_LEFT
;
316 while (stack_idx
>= 0)
318 /* Note that we fall through to the next case statement. */
319 switch(stack
[stack_idx
])
321 case HEAP_NODE_TRAVERSE_LEFT
:
323 unsigned int left_idx
= heap_idx
<< 1;
324 if (left_idx
< heap
->size
&& heap
->entries
[left_idx
].version
> version
)
327 idx
= heap
->entries
[heap_idx
].idx
;
328 apply_clamped_constant(gl_info
, constant_locations
[idx
], &constants
[idx
* 4]);
330 stack
[stack_idx
++] = HEAP_NODE_TRAVERSE_RIGHT
;
331 stack
[stack_idx
] = HEAP_NODE_TRAVERSE_LEFT
;
336 case HEAP_NODE_TRAVERSE_RIGHT
:
338 unsigned int right_idx
= (heap_idx
<< 1) + 1;
339 if (right_idx
< heap
->size
&& heap
->entries
[right_idx
].version
> version
)
341 heap_idx
= right_idx
;
342 idx
= heap
->entries
[heap_idx
].idx
;
343 apply_clamped_constant(gl_info
, constant_locations
[idx
], &constants
[idx
* 4]);
345 stack
[stack_idx
++] = HEAP_NODE_POP
;
346 stack
[stack_idx
] = HEAP_NODE_TRAVERSE_LEFT
;
359 checkGLcall("walk_constant_heap_clamped()");
362 /* Loads floating point constants (aka uniforms) into the currently set GLSL program. */
363 static void shader_glsl_load_constantsF(IWineD3DBaseShaderImpl
*This
, const WineD3D_GL_Info
*gl_info
,
364 const float *constants
, const GLint
*constant_locations
, const struct constant_heap
*heap
,
365 unsigned char *stack
, UINT version
)
367 const local_constant
*lconst
;
369 /* 1.X pshaders have the constants clamped to [-1;1] implicitly. */
370 if (WINED3DSHADER_VERSION_MAJOR(This
->baseShader
.reg_maps
.shader_version
) == 1
371 && shader_is_pshader_version(This
->baseShader
.reg_maps
.shader_version
))
372 walk_constant_heap_clamped(gl_info
, constants
, constant_locations
, heap
, stack
, version
);
374 walk_constant_heap(gl_info
, constants
, constant_locations
, heap
, stack
, version
);
376 if (!This
->baseShader
.load_local_constsF
)
378 TRACE("No need to load local float constants for this shader\n");
382 /* Immediate constants are clamped to [-1;1] at shader creation time if needed */
383 LIST_FOR_EACH_ENTRY(lconst
, &This
->baseShader
.constantsF
, local_constant
, entry
)
385 GLint location
= constant_locations
[lconst
->idx
];
386 /* We found this uniform name in the program - go ahead and send the data */
387 if (location
!= -1) GL_EXTCALL(glUniform4fvARB(location
, 1, (const GLfloat
*)lconst
->value
));
389 checkGLcall("glUniform4fvARB()");
392 /* Loads integer constants (aka uniforms) into the currently set GLSL program. */
393 static void shader_glsl_load_constantsI(IWineD3DBaseShaderImpl
*This
, const WineD3D_GL_Info
*gl_info
,
394 const GLint locations
[MAX_CONST_I
], const int *constants
, WORD constants_set
)
399 for (i
= 0; constants_set
; constants_set
>>= 1, ++i
)
401 if (!(constants_set
& 1)) continue;
403 TRACE_(d3d_constants
)("Loading constants %u: %i, %i, %i, %i\n",
404 i
, constants
[i
*4], constants
[i
*4+1], constants
[i
*4+2], constants
[i
*4+3]);
406 /* We found this uniform name in the program - go ahead and send the data */
407 GL_EXTCALL(glUniform4ivARB(locations
[i
], 1, &constants
[i
*4]));
408 checkGLcall("glUniform4ivARB");
411 /* Load immediate constants */
412 ptr
= list_head(&This
->baseShader
.constantsI
);
414 const struct local_constant
*lconst
= LIST_ENTRY(ptr
, const struct local_constant
, entry
);
415 unsigned int idx
= lconst
->idx
;
416 const GLint
*values
= (const GLint
*)lconst
->value
;
418 TRACE_(d3d_constants
)("Loading local constants %i: %i, %i, %i, %i\n", idx
,
419 values
[0], values
[1], values
[2], values
[3]);
421 /* We found this uniform name in the program - go ahead and send the data */
422 GL_EXTCALL(glUniform4ivARB(locations
[idx
], 1, values
));
423 checkGLcall("glUniform4ivARB");
424 ptr
= list_next(&This
->baseShader
.constantsI
, ptr
);
428 /* Loads boolean constants (aka uniforms) into the currently set GLSL program. */
429 static void shader_glsl_load_constantsB(IWineD3DBaseShaderImpl
*This
, const WineD3D_GL_Info
*gl_info
,
430 GLhandleARB programId
, const BOOL
*constants
, WORD constants_set
)
435 char is_pshader
= shader_is_pshader_version(This
->baseShader
.reg_maps
.shader_version
);
436 const char* prefix
= is_pshader
? "PB":"VB";
439 /* TODO: Benchmark and see if it would be beneficial to store the
440 * locations of the constants to avoid looking up each time */
441 for (i
= 0; constants_set
; constants_set
>>= 1, ++i
)
443 if (!(constants_set
& 1)) continue;
445 TRACE_(d3d_constants
)("Loading constants %i: %i;\n", i
, constants
[i
]);
447 /* TODO: Benchmark and see if it would be beneficial to store the
448 * locations of the constants to avoid looking up each time */
449 snprintf(tmp_name
, sizeof(tmp_name
), "%s[%i]", prefix
, i
);
450 tmp_loc
= GL_EXTCALL(glGetUniformLocationARB(programId
, tmp_name
));
453 /* We found this uniform name in the program - go ahead and send the data */
454 GL_EXTCALL(glUniform1ivARB(tmp_loc
, 1, &constants
[i
]));
455 checkGLcall("glUniform1ivARB");
459 /* Load immediate constants */
460 ptr
= list_head(&This
->baseShader
.constantsB
);
462 const struct local_constant
*lconst
= LIST_ENTRY(ptr
, const struct local_constant
, entry
);
463 unsigned int idx
= lconst
->idx
;
464 const GLint
*values
= (const GLint
*)lconst
->value
;
466 TRACE_(d3d_constants
)("Loading local constants %i: %i\n", idx
, values
[0]);
468 snprintf(tmp_name
, sizeof(tmp_name
), "%s[%i]", prefix
, idx
);
469 tmp_loc
= GL_EXTCALL(glGetUniformLocationARB(programId
, tmp_name
));
471 /* We found this uniform name in the program - go ahead and send the data */
472 GL_EXTCALL(glUniform1ivARB(tmp_loc
, 1, values
));
473 checkGLcall("glUniform1ivARB");
475 ptr
= list_next(&This
->baseShader
.constantsB
, ptr
);
479 static void reset_program_constant_version(void *value
, void *context
)
481 struct glsl_shader_prog_link
*entry
= value
;
482 entry
->constant_version
= 0;
486 * Loads the app-supplied constants into the currently set GLSL program.
488 static void shader_glsl_load_constants(
489 IWineD3DDevice
* device
,
491 char useVertexShader
) {
493 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) device
;
494 struct shader_glsl_priv
*priv
= deviceImpl
->shader_priv
;
495 IWineD3DStateBlockImpl
* stateBlock
= deviceImpl
->stateBlock
;
496 const WineD3D_GL_Info
*gl_info
= &deviceImpl
->adapter
->gl_info
;
498 GLhandleARB programId
;
499 struct glsl_shader_prog_link
*prog
= priv
->glsl_program
;
500 UINT constant_version
;
504 /* No GLSL program set - nothing to do. */
507 programId
= prog
->programId
;
508 constant_version
= prog
->constant_version
;
510 if (useVertexShader
) {
511 IWineD3DBaseShaderImpl
* vshader
= (IWineD3DBaseShaderImpl
*) stateBlock
->vertexShader
;
513 /* Load DirectX 9 float constants/uniforms for vertex shader */
514 shader_glsl_load_constantsF(vshader
, gl_info
, stateBlock
->vertexShaderConstantF
,
515 prog
->vuniformF_locations
, &priv
->vconst_heap
, priv
->stack
, constant_version
);
517 /* Load DirectX 9 integer constants/uniforms for vertex shader */
518 if(vshader
->baseShader
.uses_int_consts
) {
519 shader_glsl_load_constantsI(vshader
, gl_info
, prog
->vuniformI_locations
,
520 stateBlock
->vertexShaderConstantI
, stateBlock
->changed
.vertexShaderConstantsI
);
523 /* Load DirectX 9 boolean constants/uniforms for vertex shader */
524 if(vshader
->baseShader
.uses_bool_consts
) {
525 shader_glsl_load_constantsB(vshader
, gl_info
, programId
,
526 stateBlock
->vertexShaderConstantB
, stateBlock
->changed
.vertexShaderConstantsB
);
529 /* Upload the position fixup params */
530 GL_EXTCALL(glUniform4fvARB(prog
->posFixup_location
, 1, &deviceImpl
->posFixup
[0]));
531 checkGLcall("glUniform4fvARB");
534 if (usePixelShader
) {
536 IWineD3DBaseShaderImpl
* pshader
= (IWineD3DBaseShaderImpl
*) stateBlock
->pixelShader
;
538 /* Load DirectX 9 float constants/uniforms for pixel shader */
539 shader_glsl_load_constantsF(pshader
, gl_info
, stateBlock
->pixelShaderConstantF
,
540 prog
->puniformF_locations
, &priv
->pconst_heap
, priv
->stack
, constant_version
);
542 /* Load DirectX 9 integer constants/uniforms for pixel shader */
543 if(pshader
->baseShader
.uses_int_consts
) {
544 shader_glsl_load_constantsI(pshader
, gl_info
, prog
->puniformI_locations
,
545 stateBlock
->pixelShaderConstantI
, stateBlock
->changed
.pixelShaderConstantsI
);
548 /* Load DirectX 9 boolean constants/uniforms for pixel shader */
549 if(pshader
->baseShader
.uses_bool_consts
) {
550 shader_glsl_load_constantsB(pshader
, gl_info
, programId
,
551 stateBlock
->pixelShaderConstantB
, stateBlock
->changed
.pixelShaderConstantsB
);
554 /* Upload the environment bump map matrix if needed. The needsbumpmat member specifies the texture stage to load the matrix from.
555 * It can't be 0 for a valid texbem instruction.
557 for(i
= 0; i
< ((IWineD3DPixelShaderImpl
*) pshader
)->numbumpenvmatconsts
; i
++) {
558 IWineD3DPixelShaderImpl
*ps
= (IWineD3DPixelShaderImpl
*) pshader
;
559 int stage
= ps
->luminanceconst
[i
].texunit
;
561 const float *data
= (const float *)&stateBlock
->textureState
[(int)ps
->bumpenvmatconst
[i
].texunit
][WINED3DTSS_BUMPENVMAT00
];
562 GL_EXTCALL(glUniformMatrix2fvARB(prog
->bumpenvmat_location
[i
], 1, 0, data
));
563 checkGLcall("glUniformMatrix2fvARB");
565 /* texbeml needs the luminance scale and offset too. If texbeml is used, needsbumpmat
566 * is set too, so we can check that in the needsbumpmat check
568 if(ps
->baseShader
.reg_maps
.luminanceparams
[stage
]) {
569 const GLfloat
*scale
= (const GLfloat
*)&stateBlock
->textureState
[stage
][WINED3DTSS_BUMPENVLSCALE
];
570 const GLfloat
*offset
= (const GLfloat
*)&stateBlock
->textureState
[stage
][WINED3DTSS_BUMPENVLOFFSET
];
572 GL_EXTCALL(glUniform1fvARB(prog
->luminancescale_location
[i
], 1, scale
));
573 checkGLcall("glUniform1fvARB");
574 GL_EXTCALL(glUniform1fvARB(prog
->luminanceoffset_location
[i
], 1, offset
));
575 checkGLcall("glUniform1fvARB");
579 if(((IWineD3DPixelShaderImpl
*) pshader
)->vpos_uniform
) {
580 float correction_params
[4];
581 if(deviceImpl
->render_offscreen
) {
582 correction_params
[0] = 0.0;
583 correction_params
[1] = 1.0;
585 /* position is window relative, not viewport relative */
586 correction_params
[0] = ((IWineD3DSurfaceImpl
*) deviceImpl
->render_targets
[0])->currentDesc
.Height
;
587 correction_params
[1] = -1.0;
589 GL_EXTCALL(glUniform4fvARB(prog
->ycorrection_location
, 1, correction_params
));
592 /* Constant loading for texture rect coord fixup. */
593 if (prog
->ps_args
.texrect_fixup
) {
594 UINT fixup
= prog
->ps_args
.texrect_fixup
;
596 for (i
= 0; fixup
; fixup
>>= 1, ++i
) {
597 if (-1 != prog
->rectFixup_location
[i
]) {
598 const IWineD3DBaseTextureImpl
* const tex
= (const IWineD3DBaseTextureImpl
*) stateBlock
->textures
[i
];
600 FIXME("Non-existant texture is flagged for NP2 texcoord fixup\n");
603 const float tex_dim
[2] = {tex
->baseTexture
.pow2Matrix
[0], tex
->baseTexture
.pow2Matrix
[5]};
604 GL_EXTCALL(glUniform2fvARB(prog
->rectFixup_location
[i
], 1, tex_dim
));
611 if (priv
->next_constant_version
== UINT_MAX
)
613 TRACE("Max constant version reached, resetting to 0.\n");
614 hash_table_for_each_entry(priv
->glsl_program_lookup
, reset_program_constant_version
, NULL
);
615 priv
->next_constant_version
= 1;
619 prog
->constant_version
= priv
->next_constant_version
++;
623 static inline void update_heap_entry(struct constant_heap
*heap
, unsigned int idx
,
624 unsigned int heap_idx
, DWORD new_version
)
626 struct constant_entry
*entries
= heap
->entries
;
627 unsigned int *positions
= heap
->positions
;
628 unsigned int parent_idx
;
632 parent_idx
= heap_idx
>> 1;
634 if (new_version
<= entries
[parent_idx
].version
) break;
636 entries
[heap_idx
] = entries
[parent_idx
];
637 positions
[entries
[parent_idx
].idx
] = heap_idx
;
638 heap_idx
= parent_idx
;
641 entries
[heap_idx
].version
= new_version
;
642 entries
[heap_idx
].idx
= idx
;
643 positions
[idx
] = heap_idx
;
646 static void shader_glsl_update_float_vertex_constants(IWineD3DDevice
*iface
, UINT start
, UINT count
)
648 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
649 struct shader_glsl_priv
*priv
= This
->shader_priv
;
650 struct constant_heap
*heap
= &priv
->vconst_heap
;
653 for (i
= start
; i
< count
+ start
; ++i
)
655 if (!This
->stateBlock
->changed
.vertexShaderConstantsF
[i
])
656 update_heap_entry(heap
, i
, heap
->size
++, priv
->next_constant_version
);
658 update_heap_entry(heap
, i
, heap
->positions
[i
], priv
->next_constant_version
);
662 static void shader_glsl_update_float_pixel_constants(IWineD3DDevice
*iface
, UINT start
, UINT count
)
664 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
665 struct shader_glsl_priv
*priv
= This
->shader_priv
;
666 struct constant_heap
*heap
= &priv
->pconst_heap
;
669 for (i
= start
; i
< count
+ start
; ++i
)
671 if (!This
->stateBlock
->changed
.pixelShaderConstantsF
[i
])
672 update_heap_entry(heap
, i
, heap
->size
++, priv
->next_constant_version
);
674 update_heap_entry(heap
, i
, heap
->positions
[i
], priv
->next_constant_version
);
678 /** Generate the variable & register declarations for the GLSL output target */
679 static void shader_generate_glsl_declarations(IWineD3DBaseShader
*iface
, const shader_reg_maps
*reg_maps
,
680 SHADER_BUFFER
*buffer
, const WineD3D_GL_Info
*gl_info
,
681 const struct ps_compile_args
*ps_args
)
683 IWineD3DBaseShaderImpl
* This
= (IWineD3DBaseShaderImpl
*) iface
;
684 IWineD3DDeviceImpl
*device
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
685 DWORD shader_version
= reg_maps
->shader_version
;
686 unsigned int i
, extra_constants_needed
= 0;
687 const local_constant
*lconst
;
689 /* There are some minor differences between pixel and vertex shaders */
690 char pshader
= shader_is_pshader_version(shader_version
);
691 char prefix
= pshader
? 'P' : 'V';
693 /* Prototype the subroutines */
694 for (i
= 0; i
< This
->baseShader
.limits
.label
; i
++) {
695 if (reg_maps
->labels
[i
])
696 shader_addline(buffer
, "void subroutine%u();\n", i
);
699 /* Declare the constants (aka uniforms) */
700 if (This
->baseShader
.limits
.constant_float
> 0) {
701 unsigned max_constantsF
= min(This
->baseShader
.limits
.constant_float
,
702 (pshader
? GL_LIMITS(pshader_constantsF
) : GL_LIMITS(vshader_constantsF
)));
703 shader_addline(buffer
, "uniform vec4 %cC[%u];\n", prefix
, max_constantsF
);
706 if (This
->baseShader
.limits
.constant_int
> 0)
707 shader_addline(buffer
, "uniform ivec4 %cI[%u];\n", prefix
, This
->baseShader
.limits
.constant_int
);
709 if (This
->baseShader
.limits
.constant_bool
> 0)
710 shader_addline(buffer
, "uniform bool %cB[%u];\n", prefix
, This
->baseShader
.limits
.constant_bool
);
713 shader_addline(buffer
, "uniform vec4 posFixup;\n");
714 /* Predeclaration; This function is added at link time based on the pixel shader.
715 * VS 3.0 shaders have an array OUT[] the shader writes to, earlier versions don't have
716 * that. We know the input to the reorder function at vertex shader compile time, so
717 * we can deal with that. The reorder function for a 1.x and 2.x vertex shader can just
718 * read gl_FrontColor. The output depends on the pixel shader. The reorder function for a
719 * 1.x and 2.x pshader or for fixed function will write gl_FrontColor, and for a 3.0 shader
720 * it will write to the varying array. Here we depend on the shader optimizer on sorting that
721 * out. The nvidia driver only does that if the parameter is inout instead of out, hence the
724 if (shader_version
>= WINED3DVS_VERSION(3, 0))
726 shader_addline(buffer
, "void order_ps_input(in vec4[%u]);\n", MAX_REG_OUTPUT
);
728 shader_addline(buffer
, "void order_ps_input();\n");
731 IWineD3DPixelShaderImpl
*ps_impl
= (IWineD3DPixelShaderImpl
*) This
;
733 ps_impl
->numbumpenvmatconsts
= 0;
734 for(i
= 0; i
< (sizeof(reg_maps
->bumpmat
) / sizeof(reg_maps
->bumpmat
[0])); i
++) {
735 if(!reg_maps
->bumpmat
[i
]) {
739 ps_impl
->bumpenvmatconst
[(int) ps_impl
->numbumpenvmatconsts
].texunit
= i
;
740 shader_addline(buffer
, "uniform mat2 bumpenvmat%d;\n", i
);
742 if(reg_maps
->luminanceparams
) {
743 ps_impl
->luminanceconst
[(int) ps_impl
->numbumpenvmatconsts
].texunit
= i
;
744 shader_addline(buffer
, "uniform float luminancescale%d;\n", i
);
745 shader_addline(buffer
, "uniform float luminanceoffset%d;\n", i
);
746 extra_constants_needed
++;
748 ps_impl
->luminanceconst
[(int) ps_impl
->numbumpenvmatconsts
].texunit
= -1;
751 extra_constants_needed
++;
752 ps_impl
->numbumpenvmatconsts
++;
755 if(ps_args
->srgb_correction
) {
756 shader_addline(buffer
, "const vec4 srgb_mul_low = vec4(%f, %f, %f, %f);\n",
757 srgb_mul_low
, srgb_mul_low
, srgb_mul_low
, srgb_mul_low
);
758 shader_addline(buffer
, "const vec4 srgb_comparison = vec4(%f, %f, %f, %f);\n",
759 srgb_cmp
, srgb_cmp
, srgb_cmp
, srgb_cmp
);
761 if(reg_maps
->vpos
|| reg_maps
->usesdsy
) {
762 if(This
->baseShader
.limits
.constant_float
+ extra_constants_needed
+ 1 < GL_LIMITS(pshader_constantsF
)) {
763 shader_addline(buffer
, "uniform vec4 ycorrection;\n");
764 ((IWineD3DPixelShaderImpl
*) This
)->vpos_uniform
= 1;
765 extra_constants_needed
++;
767 /* This happens because we do not have proper tracking of the constant registers that are
768 * actually used, only the max limit of the shader version
770 FIXME("Cannot find a free uniform for vpos correction params\n");
771 shader_addline(buffer
, "const vec4 ycorrection = vec4(%f, %f, 0.0, 0.0);\n",
772 device
->render_offscreen
? 0.0 : ((IWineD3DSurfaceImpl
*) device
->render_targets
[0])->currentDesc
.Height
,
773 device
->render_offscreen
? 1.0 : -1.0);
775 shader_addline(buffer
, "vec4 vpos;\n");
779 /* Declare texture samplers */
780 for (i
= 0; i
< This
->baseShader
.limits
.sampler
; i
++) {
781 if (reg_maps
->samplers
[i
]) {
783 DWORD stype
= reg_maps
->samplers
[i
] & WINED3DSP_TEXTURETYPE_MASK
;
787 shader_addline(buffer
, "uniform sampler1D %csampler%u;\n", prefix
, i
);
790 if(device
->stateBlock
->textures
[i
] &&
791 IWineD3DBaseTexture_GetTextureDimensions(device
->stateBlock
->textures
[i
]) == GL_TEXTURE_RECTANGLE_ARB
) {
792 shader_addline(buffer
, "uniform sampler2DRect %csampler%u;\n", prefix
, i
);
794 shader_addline(buffer
, "uniform sampler2D %csampler%u;\n", prefix
, i
);
797 if(ps_args
->texrect_fixup
& (1 << i
)) {
798 /* RECT textures in OpenGL use texcoords in the range [0,width]x[0,height]
799 * while D3D has them in the (normalized) [0,1]x[0,1] range.
800 * samplerRectFixup stores texture dimensions and is updated through
801 * shader_glsl_load_constants when the sampler changes. */
802 shader_addline(buffer
, "uniform vec2 %csamplerRectFixup%u;\n", prefix
, i
);
805 case WINED3DSTT_CUBE
:
806 shader_addline(buffer
, "uniform samplerCube %csampler%u;\n", prefix
, i
);
808 case WINED3DSTT_VOLUME
:
809 shader_addline(buffer
, "uniform sampler3D %csampler%u;\n", prefix
, i
);
812 shader_addline(buffer
, "uniform unsupported_sampler %csampler%u;\n", prefix
, i
);
813 FIXME("Unrecognized sampler type: %#x\n", stype
);
819 /* Declare address variables */
820 for (i
= 0; i
< This
->baseShader
.limits
.address
; i
++) {
821 if (reg_maps
->address
[i
])
822 shader_addline(buffer
, "ivec4 A%d;\n", i
);
825 /* Declare texture coordinate temporaries and initialize them */
826 for (i
= 0; i
< This
->baseShader
.limits
.texcoord
; i
++) {
827 if (reg_maps
->texcoord
[i
])
828 shader_addline(buffer
, "vec4 T%u = gl_TexCoord[%u];\n", i
, i
);
831 /* Declare input register varyings. Only pixel shader, vertex shaders have that declared in the
832 * helper function shader that is linked in at link time
834 if (pshader
&& shader_version
>= WINED3DPS_VERSION(3, 0))
836 if (use_vs(device
->stateBlock
))
838 shader_addline(buffer
, "varying vec4 IN[%u];\n", GL_LIMITS(glsl_varyings
) / 4);
840 /* TODO: Write a replacement shader for the fixed function vertex pipeline, so this isn't needed.
841 * For fixed function vertex processing + 3.0 pixel shader we need a separate function in the
842 * pixel shader that reads the fixed function color into the packed input registers.
844 shader_addline(buffer
, "vec4 IN[%u];\n", GL_LIMITS(glsl_varyings
) / 4);
848 /* Declare output register temporaries */
849 if(This
->baseShader
.limits
.packed_output
) {
850 shader_addline(buffer
, "vec4 OUT[%u];\n", This
->baseShader
.limits
.packed_output
);
853 /* Declare temporary variables */
854 for(i
= 0; i
< This
->baseShader
.limits
.temporary
; i
++) {
855 if (reg_maps
->temporary
[i
])
856 shader_addline(buffer
, "vec4 R%u;\n", i
);
859 /* Declare attributes */
860 for (i
= 0; i
< This
->baseShader
.limits
.attributes
; i
++) {
861 if (reg_maps
->attributes
[i
])
862 shader_addline(buffer
, "attribute vec4 attrib%i;\n", i
);
865 /* Declare loop registers aLx */
866 for (i
= 0; i
< reg_maps
->loop_depth
; i
++) {
867 shader_addline(buffer
, "int aL%u;\n", i
);
868 shader_addline(buffer
, "int tmpInt%u;\n", i
);
871 /* Temporary variables for matrix operations */
872 shader_addline(buffer
, "vec4 tmp0;\n");
873 shader_addline(buffer
, "vec4 tmp1;\n");
875 /* Local constants use a different name so they can be loaded once at shader link time
876 * They can't be hardcoded into the shader text via LC = {x, y, z, w}; because the
877 * float -> string conversion can cause precision loss.
879 if(!This
->baseShader
.load_local_constsF
) {
880 LIST_FOR_EACH_ENTRY(lconst
, &This
->baseShader
.constantsF
, local_constant
, entry
) {
881 shader_addline(buffer
, "uniform vec4 %cLC%u;\n", prefix
, lconst
->idx
);
885 /* Start the main program */
886 shader_addline(buffer
, "void main() {\n");
887 if(pshader
&& reg_maps
->vpos
) {
888 /* DirectX apps expect integer values, while OpenGL drivers add approximately 0.5. This causes
889 * off-by-one problems as spotted by the vPos d3d9 visual test. Unfortunately the ATI cards do
890 * not add exactly 0.5, but rather something like 0.49999999 or 0.50000001, which still causes
891 * precision troubles when we just substract 0.5.
893 * To deal with that just floor() the position. This will eliminate the fraction on all cards.
895 * TODO: Test how that behaves with multisampling once we can enable multisampling in winex11.
897 * An advantage of floor is that it works even if the driver doesn't add 1/2. It is somewhat
898 * questionable if 1.5, 2.5, ... are the proper values to return in gl_FragCoord, even though
899 * coordinates specify the pixel centers instead of the pixel corners. This code will behave
900 * correctly on drivers that returns integer values.
902 shader_addline(buffer
, "vpos = floor(vec4(0, ycorrection[0], 0, 0) + gl_FragCoord * vec4(1, ycorrection[1], 1, 1));\n");
906 /*****************************************************************************
907 * Functions to generate GLSL strings from DirectX Shader bytecode begin here.
909 * For more information, see http://wiki.winehq.org/DirectX-Shaders
910 ****************************************************************************/
913 static void shader_glsl_add_src_param(const struct wined3d_shader_instruction
*ins
, const DWORD param
,
914 const DWORD addr_token
, DWORD mask
, glsl_src_param_t
*src_param
);
916 /** Used for opcode modifiers - They multiply the result by the specified amount */
917 static const char * const shift_glsl_tab
[] = {
919 "2.0 * ", /* 1 (x2) */
920 "4.0 * ", /* 2 (x4) */
921 "8.0 * ", /* 3 (x8) */
922 "16.0 * ", /* 4 (x16) */
923 "32.0 * ", /* 5 (x32) */
930 "0.0625 * ", /* 12 (d16) */
931 "0.125 * ", /* 13 (d8) */
932 "0.25 * ", /* 14 (d4) */
933 "0.5 * " /* 15 (d2) */
936 /* Generate a GLSL parameter that does the input modifier computation and return the input register/mask to use */
937 static void shader_glsl_gen_modifier (
940 const char *in_regswizzle
,
945 switch (instr
& WINED3DSP_SRCMOD_MASK
) {
946 case WINED3DSPSM_DZ
: /* Need to handle this in the instructions itself (texld & texcrd). */
948 case WINED3DSPSM_NONE
:
949 sprintf(out_str
, "%s%s", in_reg
, in_regswizzle
);
951 case WINED3DSPSM_NEG
:
952 sprintf(out_str
, "-%s%s", in_reg
, in_regswizzle
);
954 case WINED3DSPSM_NOT
:
955 sprintf(out_str
, "!%s%s", in_reg
, in_regswizzle
);
957 case WINED3DSPSM_BIAS
:
958 sprintf(out_str
, "(%s%s - vec4(0.5)%s)", in_reg
, in_regswizzle
, in_regswizzle
);
960 case WINED3DSPSM_BIASNEG
:
961 sprintf(out_str
, "-(%s%s - vec4(0.5)%s)", in_reg
, in_regswizzle
, in_regswizzle
);
963 case WINED3DSPSM_SIGN
:
964 sprintf(out_str
, "(2.0 * (%s%s - 0.5))", in_reg
, in_regswizzle
);
966 case WINED3DSPSM_SIGNNEG
:
967 sprintf(out_str
, "-(2.0 * (%s%s - 0.5))", in_reg
, in_regswizzle
);
969 case WINED3DSPSM_COMP
:
970 sprintf(out_str
, "(1.0 - %s%s)", in_reg
, in_regswizzle
);
973 sprintf(out_str
, "(2.0 * %s%s)", in_reg
, in_regswizzle
);
975 case WINED3DSPSM_X2NEG
:
976 sprintf(out_str
, "-(2.0 * %s%s)", in_reg
, in_regswizzle
);
978 case WINED3DSPSM_ABS
:
979 sprintf(out_str
, "abs(%s%s)", in_reg
, in_regswizzle
);
981 case WINED3DSPSM_ABSNEG
:
982 sprintf(out_str
, "-abs(%s%s)", in_reg
, in_regswizzle
);
985 FIXME("Unhandled modifier %u\n", (instr
& WINED3DSP_SRCMOD_MASK
));
986 sprintf(out_str
, "%s%s", in_reg
, in_regswizzle
);
990 /** Writes the GLSL variable name that corresponds to the register that the
991 * DX opcode parameter is trying to access */
992 static void shader_glsl_get_register_name(const DWORD param
, const DWORD addr_token
,
993 char *regstr
, BOOL
*is_color
, const struct wined3d_shader_instruction
*ins
)
995 /* oPos, oFog and oPts in D3D */
996 static const char * const hwrastout_reg_names
[] = { "gl_Position", "gl_FogFragCoord", "gl_PointSize" };
998 DWORD reg
= param
& WINED3DSP_REGNUM_MASK
;
999 DWORD regtype
= shader_get_regtype(param
);
1000 IWineD3DBaseShaderImpl
*This
= (IWineD3DBaseShaderImpl
*)ins
->shader
;
1001 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
1002 const WineD3D_GL_Info
* gl_info
= &deviceImpl
->adapter
->gl_info
;
1003 DWORD shader_version
= This
->baseShader
.reg_maps
.shader_version
;
1004 char pshader
= shader_is_pshader_version(shader_version
);
1010 case WINED3DSPR_TEMP
:
1011 sprintf(tmpStr
, "R%u", reg
);
1013 case WINED3DSPR_INPUT
:
1015 /* Pixel shaders >= 3.0 */
1016 if (WINED3DSHADER_VERSION_MAJOR(shader_version
) >= 3)
1018 DWORD in_count
= GL_LIMITS(glsl_varyings
) / 4;
1020 if (param
& WINED3DSHADER_ADDRMODE_RELATIVE
) {
1021 glsl_src_param_t rel_param
;
1022 shader_glsl_add_src_param(ins
, addr_token
, 0, WINED3DSP_WRITEMASK_0
, &rel_param
);
1024 /* Removing a + 0 would be an obvious optimization, but macos doesn't see the NOP
1027 if(((IWineD3DPixelShaderImpl
*) This
)->input_reg_map
[reg
]) {
1028 if (((IWineD3DPixelShaderImpl
*)This
)->declared_in_count
> in_count
) {
1029 sprintf(tmpStr
, "((%s + %u) > %d ? (%s + %u) > %d ? gl_SecondaryColor : gl_Color : IN[%s + %u])",
1030 rel_param
.param_str
, ((IWineD3DPixelShaderImpl
*)This
)->input_reg_map
[reg
], in_count
- 1,
1031 rel_param
.param_str
, ((IWineD3DPixelShaderImpl
*)This
)->input_reg_map
[reg
], in_count
,
1032 rel_param
.param_str
, ((IWineD3DPixelShaderImpl
*)This
)->input_reg_map
[reg
]);
1034 sprintf(tmpStr
, "IN[%s + %u]", rel_param
.param_str
, ((IWineD3DPixelShaderImpl
*)This
)->input_reg_map
[reg
]);
1037 if (((IWineD3DPixelShaderImpl
*)This
)->declared_in_count
> in_count
) {
1038 sprintf(tmpStr
, "((%s) > %d ? (%s) > %d ? gl_SecondaryColor : gl_Color : IN[%s])",
1039 rel_param
.param_str
, in_count
- 1,
1040 rel_param
.param_str
, in_count
,
1041 rel_param
.param_str
);
1043 sprintf(tmpStr
, "IN[%s]", rel_param
.param_str
);
1047 DWORD idx
= ((IWineD3DPixelShaderImpl
*) This
)->input_reg_map
[reg
];
1048 if (idx
== in_count
) {
1049 sprintf(tmpStr
, "gl_Color");
1050 } else if (idx
== in_count
+ 1) {
1051 sprintf(tmpStr
, "gl_SecondaryColor");
1053 sprintf(tmpStr
, "IN[%u]", idx
);
1058 strcpy(tmpStr
, "gl_Color");
1060 strcpy(tmpStr
, "gl_SecondaryColor");
1063 if (((IWineD3DVertexShaderImpl
*)This
)->cur_args
->swizzle_map
& (1 << reg
)) *is_color
= TRUE
;
1064 sprintf(tmpStr
, "attrib%u", reg
);
1067 case WINED3DSPR_CONST
:
1069 const char prefix
= pshader
? 'P':'V';
1071 /* Relative addressing */
1072 if (param
& WINED3DSHADER_ADDRMODE_RELATIVE
) {
1074 /* Relative addressing on shaders 2.0+ have a relative address token,
1075 * prior to that, it was hard-coded as "A0.x" because there's only 1 register */
1076 if (WINED3DSHADER_VERSION_MAJOR(shader_version
) >= 2)
1078 glsl_src_param_t rel_param
;
1079 shader_glsl_add_src_param(ins
, addr_token
, 0, WINED3DSP_WRITEMASK_0
, &rel_param
);
1081 sprintf(tmpStr
, "%cC[%s + %u]", prefix
, rel_param
.param_str
, reg
);
1083 sprintf(tmpStr
, "%cC[%s]", prefix
, rel_param
.param_str
);
1087 sprintf(tmpStr
, "%cC[A0.x + %u]", prefix
, reg
);
1089 sprintf(tmpStr
, "%cC[A0.x]", prefix
);
1094 if(shader_constant_is_local(This
, reg
)) {
1095 sprintf(tmpStr
, "%cLC%u", prefix
, reg
);
1097 sprintf(tmpStr
, "%cC[%u]", prefix
, reg
);
1103 case WINED3DSPR_CONSTINT
:
1105 sprintf(tmpStr
, "PI[%u]", reg
);
1107 sprintf(tmpStr
, "VI[%u]", reg
);
1109 case WINED3DSPR_CONSTBOOL
:
1111 sprintf(tmpStr
, "PB[%u]", reg
);
1113 sprintf(tmpStr
, "VB[%u]", reg
);
1115 case WINED3DSPR_TEXTURE
: /* case WINED3DSPR_ADDR: */
1117 sprintf(tmpStr
, "T%u", reg
);
1119 sprintf(tmpStr
, "A%u", reg
);
1122 case WINED3DSPR_LOOP
:
1123 sprintf(tmpStr
, "aL%u", This
->baseShader
.cur_loop_regno
- 1);
1125 case WINED3DSPR_SAMPLER
:
1127 sprintf(tmpStr
, "Psampler%u", reg
);
1129 sprintf(tmpStr
, "Vsampler%u", reg
);
1131 case WINED3DSPR_COLOROUT
:
1132 if (reg
>= GL_LIMITS(buffers
)) {
1133 WARN("Write to render target %u, only %d supported\n", reg
, 4);
1135 if (GL_SUPPORT(ARB_DRAW_BUFFERS
)) {
1136 sprintf(tmpStr
, "gl_FragData[%u]", reg
);
1137 } else { /* On older cards with GLSL support like the GeforceFX there's only one buffer. */
1138 sprintf(tmpStr
, "gl_FragColor");
1141 case WINED3DSPR_RASTOUT
:
1142 sprintf(tmpStr
, "%s", hwrastout_reg_names
[reg
]);
1144 case WINED3DSPR_DEPTHOUT
:
1145 sprintf(tmpStr
, "gl_FragDepth");
1147 case WINED3DSPR_ATTROUT
:
1149 sprintf(tmpStr
, "gl_FrontColor");
1151 sprintf(tmpStr
, "gl_FrontSecondaryColor");
1154 case WINED3DSPR_TEXCRDOUT
:
1155 /* Vertex shaders >= 3.0: WINED3DSPR_OUTPUT */
1156 if (WINED3DSHADER_VERSION_MAJOR(shader_version
) >= 3) sprintf(tmpStr
, "OUT[%u]", reg
);
1157 else sprintf(tmpStr
, "gl_TexCoord[%u]", reg
);
1159 case WINED3DSPR_MISCTYPE
:
1162 sprintf(tmpStr
, "vpos");
1163 } else if (reg
== 1){
1164 /* Note that gl_FrontFacing is a bool, while vFace is
1165 * a float for which the sign determines front/back
1167 sprintf(tmpStr
, "(gl_FrontFacing ? 1.0 : -1.0)");
1169 FIXME("Unhandled misctype register %d\n", reg
);
1170 sprintf(tmpStr
, "unrecognized_register");
1174 FIXME("Unhandled register name Type(%d)\n", regtype
);
1175 sprintf(tmpStr
, "unrecognized_register");
1179 strcat(regstr
, tmpStr
);
1182 /* Get the GLSL write mask for the destination register */
1183 static DWORD
shader_glsl_get_write_mask(const DWORD param
, char *write_mask
) {
1184 char *ptr
= write_mask
;
1185 DWORD mask
= param
& WINED3DSP_WRITEMASK_ALL
;
1187 if (shader_is_scalar(shader_get_regtype(param
), param
& WINED3DSP_REGNUM_MASK
))
1189 mask
= WINED3DSP_WRITEMASK_0
;
1192 if (param
& WINED3DSP_WRITEMASK_0
) *ptr
++ = 'x';
1193 if (param
& WINED3DSP_WRITEMASK_1
) *ptr
++ = 'y';
1194 if (param
& WINED3DSP_WRITEMASK_2
) *ptr
++ = 'z';
1195 if (param
& WINED3DSP_WRITEMASK_3
) *ptr
++ = 'w';
1203 static unsigned int shader_glsl_get_write_mask_size(DWORD write_mask
) {
1204 unsigned int size
= 0;
1206 if (write_mask
& WINED3DSP_WRITEMASK_0
) ++size
;
1207 if (write_mask
& WINED3DSP_WRITEMASK_1
) ++size
;
1208 if (write_mask
& WINED3DSP_WRITEMASK_2
) ++size
;
1209 if (write_mask
& WINED3DSP_WRITEMASK_3
) ++size
;
1214 static void shader_glsl_get_swizzle(const DWORD param
, BOOL fixup
, DWORD mask
, char *swizzle_str
) {
1215 /* For registers of type WINED3DDECLTYPE_D3DCOLOR, data is stored as "bgra",
1216 * but addressed as "rgba". To fix this we need to swap the register's x
1217 * and z components. */
1218 DWORD swizzle
= (param
& WINED3DSP_SWIZZLE_MASK
) >> WINED3DSP_SWIZZLE_SHIFT
;
1219 const char *swizzle_chars
= fixup
? "zyxw" : "xyzw";
1220 char *ptr
= swizzle_str
;
1222 if (!shader_is_scalar(shader_get_regtype(param
), param
& WINED3DSP_REGNUM_MASK
))
1225 /* swizzle bits fields: wwzzyyxx */
1226 if (mask
& WINED3DSP_WRITEMASK_0
) *ptr
++ = swizzle_chars
[swizzle
& 0x03];
1227 if (mask
& WINED3DSP_WRITEMASK_1
) *ptr
++ = swizzle_chars
[(swizzle
>> 2) & 0x03];
1228 if (mask
& WINED3DSP_WRITEMASK_2
) *ptr
++ = swizzle_chars
[(swizzle
>> 4) & 0x03];
1229 if (mask
& WINED3DSP_WRITEMASK_3
) *ptr
++ = swizzle_chars
[(swizzle
>> 6) & 0x03];
1235 /* From a given parameter token, generate the corresponding GLSL string.
1236 * Also, return the actual register name and swizzle in case the
1237 * caller needs this information as well. */
1238 static void shader_glsl_add_src_param(const struct wined3d_shader_instruction
*ins
,
1239 const DWORD param
, const DWORD addr_token
, DWORD mask
, glsl_src_param_t
*src_param
)
1241 BOOL is_color
= FALSE
;
1242 char swizzle_str
[6];
1244 src_param
->reg_name
[0] = '\0';
1245 src_param
->param_str
[0] = '\0';
1246 swizzle_str
[0] = '\0';
1248 shader_glsl_get_register_name(param
, addr_token
, src_param
->reg_name
, &is_color
, ins
);
1250 shader_glsl_get_swizzle(param
, is_color
, mask
, swizzle_str
);
1251 shader_glsl_gen_modifier(param
, src_param
->reg_name
, swizzle_str
, src_param
->param_str
);
1254 /* From a given parameter token, generate the corresponding GLSL string.
1255 * Also, return the actual register name and swizzle in case the
1256 * caller needs this information as well. */
1257 static DWORD
shader_glsl_add_dst_param(const struct wined3d_shader_instruction
*ins
,
1258 const struct wined3d_shader_dst_param
*wined3d_dst
, glsl_dst_param_t
*glsl_dst
)
1260 BOOL is_color
= FALSE
;
1262 glsl_dst
->mask_str
[0] = '\0';
1263 glsl_dst
->reg_name
[0] = '\0';
1265 shader_glsl_get_register_name(wined3d_dst
->token
, wined3d_dst
->addr_token
, glsl_dst
->reg_name
, &is_color
, ins
);
1266 return shader_glsl_get_write_mask(wined3d_dst
->token
, glsl_dst
->mask_str
);
1269 /* Append the destination part of the instruction to the buffer, return the effective write mask */
1270 static DWORD
shader_glsl_append_dst_ext(SHADER_BUFFER
*buffer
,
1271 const struct wined3d_shader_instruction
*ins
, const struct wined3d_shader_dst_param
*dst
)
1273 glsl_dst_param_t glsl_dst
;
1277 mask
= shader_glsl_add_dst_param(ins
, dst
, &glsl_dst
);
1280 shift
= (dst
->token
& WINED3DSP_DSTSHIFT_MASK
) >> WINED3DSP_DSTSHIFT_SHIFT
;
1281 shader_addline(buffer
, "%s%s = %s(", glsl_dst
.reg_name
, glsl_dst
.mask_str
, shift_glsl_tab
[shift
]);
1287 /* Append the destination part of the instruction to the buffer, return the effective write mask */
1288 static DWORD
shader_glsl_append_dst(SHADER_BUFFER
*buffer
, const struct wined3d_shader_instruction
*ins
)
1290 return shader_glsl_append_dst_ext(buffer
, ins
, &ins
->dst
[0]);
1293 /** Process GLSL instruction modifiers */
1294 void shader_glsl_add_instruction_modifiers(const struct wined3d_shader_instruction
*ins
)
1296 glsl_dst_param_t dst_param
;
1299 if (!ins
->dst_count
) return;
1301 modifiers
= ins
->dst
[0].modifiers
;
1302 if (!modifiers
) return;
1304 shader_glsl_add_dst_param(ins
, &ins
->dst
[0], &dst_param
);
1306 if (modifiers
& WINED3DSPDM_SATURATE
)
1308 /* _SAT means to clamp the value of the register to between 0 and 1 */
1309 shader_addline(ins
->buffer
, "%s%s = clamp(%s%s, 0.0, 1.0);\n", dst_param
.reg_name
,
1310 dst_param
.mask_str
, dst_param
.reg_name
, dst_param
.mask_str
);
1313 if (modifiers
& WINED3DSPDM_MSAMPCENTROID
)
1315 FIXME("_centroid modifier not handled\n");
1318 if (modifiers
& WINED3DSPDM_PARTIALPRECISION
)
1320 /* MSDN says this modifier can be safely ignored, so that's what we'll do. */
1324 static inline const char *shader_get_comp_op(DWORD flags
)
1326 DWORD op
= (flags
& INST_CONTROLS_MASK
) >> INST_CONTROLS_SHIFT
;
1328 case COMPARISON_GT
: return ">";
1329 case COMPARISON_EQ
: return "==";
1330 case COMPARISON_GE
: return ">=";
1331 case COMPARISON_LT
: return "<";
1332 case COMPARISON_NE
: return "!=";
1333 case COMPARISON_LE
: return "<=";
1335 FIXME("Unrecognized comparison value: %u\n", op
);
1340 static void shader_glsl_get_sample_function(DWORD sampler_type
, DWORD flags
, glsl_sample_function_t
*sample_function
)
1342 BOOL projected
= flags
& WINED3D_GLSL_SAMPLE_PROJECTED
;
1343 BOOL texrect
= flags
& WINED3D_GLSL_SAMPLE_RECT
;
1344 BOOL lod
= flags
& WINED3D_GLSL_SAMPLE_LOD
;
1346 /* Note that there's no such thing as a projected cube texture. */
1347 switch(sampler_type
) {
1350 sample_function
->name
= projected
? "texture1DProjLod" : "texture1DLod";
1352 sample_function
->name
= projected
? "texture1DProj" : "texture1D";
1354 sample_function
->coord_mask
= WINED3DSP_WRITEMASK_0
;
1359 sample_function
->name
= projected
? "texture2DRectProjLod" : "texture2DRectLod";
1361 sample_function
->name
= projected
? "texture2DRectProj" : "texture2DRect";
1365 sample_function
->name
= projected
? "texture2DProjLod" : "texture2DLod";
1367 sample_function
->name
= projected
? "texture2DProj" : "texture2D";
1370 sample_function
->coord_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
;
1372 case WINED3DSTT_CUBE
:
1374 sample_function
->name
= "textureCubeLod";
1376 sample_function
->name
= "textureCube";
1378 sample_function
->coord_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
1380 case WINED3DSTT_VOLUME
:
1382 sample_function
->name
= projected
? "texture3DProjLod" : "texture3DLod";
1384 sample_function
->name
= projected
? "texture3DProj" : "texture3D";
1386 sample_function
->coord_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
1389 sample_function
->name
= "";
1390 sample_function
->coord_mask
= 0;
1391 FIXME("Unrecognized sampler type: %#x;\n", sampler_type
);
1396 static void shader_glsl_append_fixup_arg(char *arguments
, const char *reg_name
,
1397 BOOL sign_fixup
, enum fixup_channel_source channel_source
)
1399 switch(channel_source
)
1401 case CHANNEL_SOURCE_ZERO
:
1402 strcat(arguments
, "0.0");
1405 case CHANNEL_SOURCE_ONE
:
1406 strcat(arguments
, "1.0");
1409 case CHANNEL_SOURCE_X
:
1410 strcat(arguments
, reg_name
);
1411 strcat(arguments
, ".x");
1414 case CHANNEL_SOURCE_Y
:
1415 strcat(arguments
, reg_name
);
1416 strcat(arguments
, ".y");
1419 case CHANNEL_SOURCE_Z
:
1420 strcat(arguments
, reg_name
);
1421 strcat(arguments
, ".z");
1424 case CHANNEL_SOURCE_W
:
1425 strcat(arguments
, reg_name
);
1426 strcat(arguments
, ".w");
1430 FIXME("Unhandled channel source %#x\n", channel_source
);
1431 strcat(arguments
, "undefined");
1435 if (sign_fixup
) strcat(arguments
, " * 2.0 - 1.0");
1438 static void shader_glsl_color_correction(const struct wined3d_shader_instruction
*ins
, struct color_fixup_desc fixup
)
1440 struct wined3d_shader_dst_param dst
;
1441 unsigned int mask_size
, remaining
;
1442 glsl_dst_param_t dst_param
;
1443 char arguments
[256];
1447 if (fixup
.x_sign_fixup
|| fixup
.x_source
!= CHANNEL_SOURCE_X
) mask
|= WINED3DSP_WRITEMASK_0
;
1448 if (fixup
.y_sign_fixup
|| fixup
.y_source
!= CHANNEL_SOURCE_Y
) mask
|= WINED3DSP_WRITEMASK_1
;
1449 if (fixup
.z_sign_fixup
|| fixup
.z_source
!= CHANNEL_SOURCE_Z
) mask
|= WINED3DSP_WRITEMASK_2
;
1450 if (fixup
.w_sign_fixup
|| fixup
.w_source
!= CHANNEL_SOURCE_W
) mask
|= WINED3DSP_WRITEMASK_3
;
1451 mask
&= ins
->dst
[0].write_mask
;
1453 if (!mask
) return; /* Nothing to do */
1455 if (is_yuv_fixup(fixup
))
1457 enum yuv_fixup yuv_fixup
= get_yuv_fixup(fixup
);
1458 FIXME("YUV fixup (%#x) not supported\n", yuv_fixup
);
1462 mask_size
= shader_glsl_get_write_mask_size(mask
);
1465 dst
.write_mask
= mask
;
1466 dst
.token
= (dst
.token
& ~WINED3DSP_WRITEMASK_ALL
) | dst
.write_mask
;
1467 shader_glsl_add_dst_param(ins
, &dst
, &dst_param
);
1469 arguments
[0] = '\0';
1470 remaining
= mask_size
;
1471 if (mask
& WINED3DSP_WRITEMASK_0
)
1473 shader_glsl_append_fixup_arg(arguments
, dst_param
.reg_name
, fixup
.x_sign_fixup
, fixup
.x_source
);
1474 if (--remaining
) strcat(arguments
, ", ");
1476 if (mask
& WINED3DSP_WRITEMASK_1
)
1478 shader_glsl_append_fixup_arg(arguments
, dst_param
.reg_name
, fixup
.y_sign_fixup
, fixup
.y_source
);
1479 if (--remaining
) strcat(arguments
, ", ");
1481 if (mask
& WINED3DSP_WRITEMASK_2
)
1483 shader_glsl_append_fixup_arg(arguments
, dst_param
.reg_name
, fixup
.z_sign_fixup
, fixup
.z_source
);
1484 if (--remaining
) strcat(arguments
, ", ");
1486 if (mask
& WINED3DSP_WRITEMASK_3
)
1488 shader_glsl_append_fixup_arg(arguments
, dst_param
.reg_name
, fixup
.w_sign_fixup
, fixup
.w_source
);
1489 if (--remaining
) strcat(arguments
, ", ");
1494 shader_addline(ins
->buffer
, "%s%s = vec%u(%s);\n",
1495 dst_param
.reg_name
, dst_param
.mask_str
, mask_size
, arguments
);
1499 shader_addline(ins
->buffer
, "%s%s = %s;\n", dst_param
.reg_name
, dst_param
.mask_str
, arguments
);
1503 static void PRINTF_ATTR(6, 7) shader_glsl_gen_sample_code(const struct wined3d_shader_instruction
*ins
,
1504 DWORD sampler
, const glsl_sample_function_t
*sample_function
, DWORD swizzle
,
1505 const char *bias
, const char *coord_reg_fmt
, ...)
1507 const char *sampler_base
;
1508 char dst_swizzle
[6];
1509 struct color_fixup_desc fixup
;
1510 BOOL rect_fixup
= FALSE
;
1513 shader_glsl_get_swizzle(swizzle
, FALSE
, ins
->dst
[0].write_mask
, dst_swizzle
);
1515 if (shader_is_pshader_version(ins
->reg_maps
->shader_version
))
1517 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*)ins
->shader
;
1518 fixup
= This
->cur_args
->color_fixup
[sampler
];
1519 sampler_base
= "Psampler";
1521 if(This
->cur_args
->texrect_fixup
& (1 << sampler
)) {
1523 FIXME("Biased sampling from RECT textures is unsupported\n");
1529 sampler_base
= "Vsampler";
1530 fixup
= COLOR_FIXUP_IDENTITY
; /* FIXME: Vshader color fixup */
1533 shader_glsl_append_dst(ins
->buffer
, ins
);
1535 shader_addline(ins
->buffer
, "%s(%s%u, ", sample_function
->name
, sampler_base
, sampler
);
1537 va_start(args
, coord_reg_fmt
);
1538 shader_vaddline(ins
->buffer
, coord_reg_fmt
, args
);
1542 shader_addline(ins
->buffer
, ", %s)%s);\n", bias
, dst_swizzle
);
1545 shader_addline(ins
->buffer
, " * PsamplerRectFixup%u)%s);\n", sampler
, dst_swizzle
);
1547 shader_addline(ins
->buffer
, ")%s);\n", dst_swizzle
);
1551 if(!is_identity_fixup(fixup
)) {
1552 shader_glsl_color_correction(ins
, fixup
);
1556 /*****************************************************************************
1558 * Begin processing individual instruction opcodes
1560 ****************************************************************************/
1562 /* Generate GLSL arithmetic functions (dst = src1 + src2) */
1563 static void shader_glsl_arith(const struct wined3d_shader_instruction
*ins
)
1565 SHADER_BUFFER
*buffer
= ins
->buffer
;
1566 glsl_src_param_t src0_param
;
1567 glsl_src_param_t src1_param
;
1571 /* Determine the GLSL operator to use based on the opcode */
1572 switch (ins
->handler_idx
)
1574 case WINED3DSIH_MUL
: op
= '*'; break;
1575 case WINED3DSIH_ADD
: op
= '+'; break;
1576 case WINED3DSIH_SUB
: op
= '-'; break;
1579 FIXME("Opcode %#x not yet handled in GLSL\n", ins
->handler_idx
);
1583 write_mask
= shader_glsl_append_dst(buffer
, ins
);
1584 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], write_mask
, &src0_param
);
1585 shader_glsl_add_src_param(ins
, ins
->src
[1], ins
->src_addr
[1], write_mask
, &src1_param
);
1586 shader_addline(buffer
, "%s %c %s);\n", src0_param
.param_str
, op
, src1_param
.param_str
);
1589 /* Process the WINED3DSIO_MOV opcode using GLSL (dst = src) */
1590 static void shader_glsl_mov(const struct wined3d_shader_instruction
*ins
)
1592 SHADER_BUFFER
*buffer
= ins
->buffer
;
1593 glsl_src_param_t src0_param
;
1596 write_mask
= shader_glsl_append_dst(buffer
, ins
);
1597 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], write_mask
, &src0_param
);
1599 /* In vs_1_1 WINED3DSIO_MOV can write to the address register. In later
1600 * shader versions WINED3DSIO_MOVA is used for this. */
1601 if ((WINED3DSHADER_VERSION_MAJOR(ins
->reg_maps
->shader_version
) == 1
1602 && !shader_is_pshader_version(ins
->reg_maps
->shader_version
)
1603 && ins
->dst
[0].register_type
== WINED3DSPR_ADDR
))
1605 /* This is a simple floor() */
1606 unsigned int mask_size
= shader_glsl_get_write_mask_size(write_mask
);
1607 if (mask_size
> 1) {
1608 shader_addline(buffer
, "ivec%d(floor(%s)));\n", mask_size
, src0_param
.param_str
);
1610 shader_addline(buffer
, "int(floor(%s)));\n", src0_param
.param_str
);
1613 else if(ins
->handler_idx
== WINED3DSIH_MOVA
)
1615 /* We need to *round* to the nearest int here. */
1616 unsigned int mask_size
= shader_glsl_get_write_mask_size(write_mask
);
1617 if (mask_size
> 1) {
1618 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
);
1620 shader_addline(buffer
, "int(floor(abs(%s) + 0.5) * sign(%s)));\n", src0_param
.param_str
, src0_param
.param_str
);
1623 shader_addline(buffer
, "%s);\n", src0_param
.param_str
);
1627 /* Process the dot product operators DP3 and DP4 in GLSL (dst = dot(src0, src1)) */
1628 static void shader_glsl_dot(const struct wined3d_shader_instruction
*ins
)
1630 SHADER_BUFFER
*buffer
= ins
->buffer
;
1631 glsl_src_param_t src0_param
;
1632 glsl_src_param_t src1_param
;
1633 DWORD dst_write_mask
, src_write_mask
;
1634 unsigned int dst_size
= 0;
1636 dst_write_mask
= shader_glsl_append_dst(buffer
, ins
);
1637 dst_size
= shader_glsl_get_write_mask_size(dst_write_mask
);
1639 /* dp3 works on vec3, dp4 on vec4 */
1640 if (ins
->handler_idx
== WINED3DSIH_DP4
)
1642 src_write_mask
= WINED3DSP_WRITEMASK_ALL
;
1644 src_write_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
1647 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], src_write_mask
, &src0_param
);
1648 shader_glsl_add_src_param(ins
, ins
->src
[1], ins
->src_addr
[1], src_write_mask
, &src1_param
);
1651 shader_addline(buffer
, "vec%d(dot(%s, %s)));\n", dst_size
, src0_param
.param_str
, src1_param
.param_str
);
1653 shader_addline(buffer
, "dot(%s, %s));\n", src0_param
.param_str
, src1_param
.param_str
);
1657 /* Note that this instruction has some restrictions. The destination write mask
1658 * can't contain the w component, and the source swizzles have to be .xyzw */
1659 static void shader_glsl_cross(const struct wined3d_shader_instruction
*ins
)
1661 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
1662 glsl_src_param_t src0_param
;
1663 glsl_src_param_t src1_param
;
1666 shader_glsl_get_write_mask(ins
->dst
[0].token
, dst_mask
);
1667 shader_glsl_append_dst(ins
->buffer
, ins
);
1668 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], src_mask
, &src0_param
);
1669 shader_glsl_add_src_param(ins
, ins
->src
[1], ins
->src_addr
[1], src_mask
, &src1_param
);
1670 shader_addline(ins
->buffer
, "cross(%s, %s)%s);\n", src0_param
.param_str
, src1_param
.param_str
, dst_mask
);
1673 /* Process the WINED3DSIO_POW instruction in GLSL (dst = |src0|^src1)
1674 * Src0 and src1 are scalars. Note that D3D uses the absolute of src0, while
1675 * GLSL uses the value as-is. */
1676 static void shader_glsl_pow(const struct wined3d_shader_instruction
*ins
)
1678 SHADER_BUFFER
*buffer
= ins
->buffer
;
1679 glsl_src_param_t src0_param
;
1680 glsl_src_param_t src1_param
;
1681 DWORD dst_write_mask
;
1682 unsigned int dst_size
;
1684 dst_write_mask
= shader_glsl_append_dst(buffer
, ins
);
1685 dst_size
= shader_glsl_get_write_mask_size(dst_write_mask
);
1687 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
1688 shader_glsl_add_src_param(ins
, ins
->src
[1], ins
->src_addr
[1], WINED3DSP_WRITEMASK_0
, &src1_param
);
1691 shader_addline(buffer
, "vec%d(pow(abs(%s), %s)));\n", dst_size
, src0_param
.param_str
, src1_param
.param_str
);
1693 shader_addline(buffer
, "pow(abs(%s), %s));\n", src0_param
.param_str
, src1_param
.param_str
);
1697 /* Process the WINED3DSIO_LOG instruction in GLSL (dst = log2(|src0|))
1698 * Src0 is a scalar. Note that D3D uses the absolute of src0, while
1699 * GLSL uses the value as-is. */
1700 static void shader_glsl_log(const struct wined3d_shader_instruction
*ins
)
1702 SHADER_BUFFER
*buffer
= ins
->buffer
;
1703 glsl_src_param_t src0_param
;
1704 DWORD dst_write_mask
;
1705 unsigned int dst_size
;
1707 dst_write_mask
= shader_glsl_append_dst(buffer
, ins
);
1708 dst_size
= shader_glsl_get_write_mask_size(dst_write_mask
);
1710 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
1713 shader_addline(buffer
, "vec%d(log2(abs(%s))));\n", dst_size
, src0_param
.param_str
);
1715 shader_addline(buffer
, "log2(abs(%s)));\n", src0_param
.param_str
);
1719 /* Map the opcode 1-to-1 to the GL code (arg->dst = instruction(src0, src1, ...) */
1720 static void shader_glsl_map2gl(const struct wined3d_shader_instruction
*ins
)
1722 SHADER_BUFFER
*buffer
= ins
->buffer
;
1723 glsl_src_param_t src_param
;
1724 const char *instruction
;
1728 /* Determine the GLSL function to use based on the opcode */
1729 /* TODO: Possibly make this a table for faster lookups */
1730 switch (ins
->handler_idx
)
1732 case WINED3DSIH_MIN
: instruction
= "min"; break;
1733 case WINED3DSIH_MAX
: instruction
= "max"; break;
1734 case WINED3DSIH_ABS
: instruction
= "abs"; break;
1735 case WINED3DSIH_FRC
: instruction
= "fract"; break;
1736 case WINED3DSIH_NRM
: instruction
= "normalize"; break;
1737 case WINED3DSIH_EXP
: instruction
= "exp2"; break;
1738 case WINED3DSIH_SGN
: instruction
= "sign"; break;
1739 case WINED3DSIH_DSX
: instruction
= "dFdx"; break;
1740 case WINED3DSIH_DSY
: instruction
= "ycorrection.y * dFdy"; break;
1741 default: instruction
= "";
1742 FIXME("Opcode %#x not yet handled in GLSL\n", ins
->handler_idx
);
1746 write_mask
= shader_glsl_append_dst(buffer
, ins
);
1748 shader_addline(buffer
, "%s(", instruction
);
1752 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], write_mask
, &src_param
);
1753 shader_addline(buffer
, "%s", src_param
.param_str
);
1754 for (i
= 1; i
< ins
->src_count
; ++i
)
1756 shader_glsl_add_src_param(ins
, ins
->src
[i
], ins
->src_addr
[i
], write_mask
, &src_param
);
1757 shader_addline(buffer
, ", %s", src_param
.param_str
);
1761 shader_addline(buffer
, "));\n");
1764 /** Process the WINED3DSIO_EXPP instruction in GLSL:
1765 * For shader model 1.x, do the following (and honor the writemask, so use a temporary variable):
1766 * dst.x = 2^(floor(src))
1767 * dst.y = src - floor(src)
1768 * dst.z = 2^src (partial precision is allowed, but optional)
1770 * For 2.0 shaders, just do this (honoring writemask and swizzle):
1771 * dst = 2^src; (partial precision is allowed, but optional)
1773 static void shader_glsl_expp(const struct wined3d_shader_instruction
*ins
)
1775 glsl_src_param_t src_param
;
1777 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src_param
);
1779 if (ins
->reg_maps
->shader_version
< WINED3DPS_VERSION(2,0))
1783 shader_addline(ins
->buffer
, "tmp0.x = exp2(floor(%s));\n", src_param
.param_str
);
1784 shader_addline(ins
->buffer
, "tmp0.y = %s - floor(%s);\n", src_param
.param_str
, src_param
.param_str
);
1785 shader_addline(ins
->buffer
, "tmp0.z = exp2(%s);\n", src_param
.param_str
);
1786 shader_addline(ins
->buffer
, "tmp0.w = 1.0;\n");
1788 shader_glsl_append_dst(ins
->buffer
, ins
);
1789 shader_glsl_get_write_mask(ins
->dst
[0].token
, dst_mask
);
1790 shader_addline(ins
->buffer
, "tmp0%s);\n", dst_mask
);
1793 unsigned int mask_size
;
1795 write_mask
= shader_glsl_append_dst(ins
->buffer
, ins
);
1796 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
1798 if (mask_size
> 1) {
1799 shader_addline(ins
->buffer
, "vec%d(exp2(%s)));\n", mask_size
, src_param
.param_str
);
1801 shader_addline(ins
->buffer
, "exp2(%s));\n", src_param
.param_str
);
1806 /** Process the RCP (reciprocal or inverse) opcode in GLSL (dst = 1 / src) */
1807 static void shader_glsl_rcp(const struct wined3d_shader_instruction
*ins
)
1809 glsl_src_param_t src_param
;
1811 unsigned int mask_size
;
1813 write_mask
= shader_glsl_append_dst(ins
->buffer
, ins
);
1814 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
1815 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], WINED3DSP_WRITEMASK_3
, &src_param
);
1817 if (mask_size
> 1) {
1818 shader_addline(ins
->buffer
, "vec%d(1.0 / %s));\n", mask_size
, src_param
.param_str
);
1820 shader_addline(ins
->buffer
, "1.0 / %s);\n", src_param
.param_str
);
1824 static void shader_glsl_rsq(const struct wined3d_shader_instruction
*ins
)
1826 SHADER_BUFFER
*buffer
= ins
->buffer
;
1827 glsl_src_param_t src_param
;
1829 unsigned int mask_size
;
1831 write_mask
= shader_glsl_append_dst(buffer
, ins
);
1832 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
1834 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], WINED3DSP_WRITEMASK_3
, &src_param
);
1836 if (mask_size
> 1) {
1837 shader_addline(buffer
, "vec%d(inversesqrt(%s)));\n", mask_size
, src_param
.param_str
);
1839 shader_addline(buffer
, "inversesqrt(%s));\n", src_param
.param_str
);
1843 /** Process signed comparison opcodes in GLSL. */
1844 static void shader_glsl_compare(const struct wined3d_shader_instruction
*ins
)
1846 glsl_src_param_t src0_param
;
1847 glsl_src_param_t src1_param
;
1849 unsigned int mask_size
;
1851 write_mask
= shader_glsl_append_dst(ins
->buffer
, ins
);
1852 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
1853 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], write_mask
, &src0_param
);
1854 shader_glsl_add_src_param(ins
, ins
->src
[1], ins
->src_addr
[1], write_mask
, &src1_param
);
1856 if (mask_size
> 1) {
1857 const char *compare
;
1859 switch(ins
->handler_idx
)
1861 case WINED3DSIH_SLT
: compare
= "lessThan"; break;
1862 case WINED3DSIH_SGE
: compare
= "greaterThanEqual"; break;
1863 default: compare
= "";
1864 FIXME("Can't handle opcode %#x\n", ins
->handler_idx
);
1867 shader_addline(ins
->buffer
, "vec%d(%s(%s, %s)));\n", mask_size
, compare
,
1868 src0_param
.param_str
, src1_param
.param_str
);
1870 switch(ins
->handler_idx
)
1872 case WINED3DSIH_SLT
:
1873 /* Step(src0, src1) is not suitable here because if src0 == src1 SLT is supposed,
1874 * to return 0.0 but step returns 1.0 because step is not < x
1875 * An alternative is a bvec compare padded with an unused second component.
1876 * step(src1 * -1.0, src0 * -1.0) is not an option because it suffers from the same
1877 * issue. Playing with not() is not possible either because not() does not accept
1880 shader_addline(ins
->buffer
, "(%s < %s) ? 1.0 : 0.0);\n", src0_param
.param_str
, src1_param
.param_str
);
1882 case WINED3DSIH_SGE
:
1883 /* Here we can use the step() function and safe a conditional */
1884 shader_addline(ins
->buffer
, "step(%s, %s));\n", src1_param
.param_str
, src0_param
.param_str
);
1887 FIXME("Can't handle opcode %#x\n", ins
->handler_idx
);
1893 /** Process CMP instruction in GLSL (dst = src0 >= 0.0 ? src1 : src2), per channel */
1894 static void shader_glsl_cmp(const struct wined3d_shader_instruction
*ins
)
1896 glsl_src_param_t src0_param
;
1897 glsl_src_param_t src1_param
;
1898 glsl_src_param_t src2_param
;
1899 DWORD write_mask
, cmp_channel
= 0;
1902 BOOL temp_destination
= FALSE
;
1904 if (shader_is_scalar(shader_get_regtype(ins
->src
[0]), ins
->src
[0] & WINED3DSP_REGNUM_MASK
))
1906 write_mask
= shader_glsl_append_dst(ins
->buffer
, ins
);
1908 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], WINED3DSP_WRITEMASK_ALL
, &src0_param
);
1909 shader_glsl_add_src_param(ins
, ins
->src
[1], ins
->src_addr
[1], write_mask
, &src1_param
);
1910 shader_glsl_add_src_param(ins
, ins
->src
[2], ins
->src_addr
[2], write_mask
, &src2_param
);
1912 shader_addline(ins
->buffer
, "%s >= 0.0 ? %s : %s);\n",
1913 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
1915 DWORD src0reg
= ins
->src
[0] & WINED3DSP_REGNUM_MASK
;
1916 DWORD src1reg
= ins
->src
[1] & WINED3DSP_REGNUM_MASK
;
1917 DWORD src2reg
= ins
->src
[2] & WINED3DSP_REGNUM_MASK
;
1918 DWORD src0regtype
= shader_get_regtype(ins
->src
[0]);
1919 DWORD src1regtype
= shader_get_regtype(ins
->src
[1]);
1920 DWORD src2regtype
= shader_get_regtype(ins
->src
[2]);
1921 DWORD dstreg
= ins
->dst
[0].register_idx
;
1922 DWORD dstregtype
= ins
->dst
[0].register_type
;
1923 DWORD dst_mask
= ins
->dst
[0].write_mask
;
1924 struct wined3d_shader_dst_param dst
= ins
->dst
[0];
1926 /* Cycle through all source0 channels */
1927 for (i
=0; i
<4; i
++) {
1929 /* Find the destination channels which use the current source0 channel */
1930 for (j
=0; j
<4; j
++) {
1931 if (((ins
->src
[0] >> (WINED3DSP_SWIZZLE_SHIFT
+ 2 * j
)) & 0x3) == i
)
1933 write_mask
|= WINED3DSP_WRITEMASK_0
<< j
;
1934 cmp_channel
= WINED3DSP_WRITEMASK_0
<< j
;
1938 dst
.write_mask
= dst_mask
& write_mask
;
1939 dst
.token
= (dst
.token
& ~WINED3DSP_WRITEMASK_ALL
) | dst
.write_mask
;
1941 /* Splitting the cmp instruction up in multiple lines imposes a problem:
1942 * The first lines may overwrite source parameters of the following lines.
1943 * Deal with that by using a temporary destination register if needed
1945 if ((src0reg
== dstreg
&& src0regtype
== dstregtype
)
1946 || (src1reg
== dstreg
&& src1regtype
== dstregtype
)
1947 || (src2reg
== dstreg
&& src2regtype
== dstregtype
))
1949 write_mask
= shader_glsl_get_write_mask(dst
.token
, mask_char
);
1950 if (!write_mask
) continue;
1951 shader_addline(ins
->buffer
, "tmp0%s = (", mask_char
);
1952 temp_destination
= TRUE
;
1954 write_mask
= shader_glsl_append_dst_ext(ins
->buffer
, ins
, &dst
);
1955 if (!write_mask
) continue;
1958 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], cmp_channel
, &src0_param
);
1959 shader_glsl_add_src_param(ins
, ins
->src
[1], ins
->src_addr
[1], write_mask
, &src1_param
);
1960 shader_glsl_add_src_param(ins
, ins
->src
[2], ins
->src_addr
[2], write_mask
, &src2_param
);
1962 shader_addline(ins
->buffer
, "%s >= 0.0 ? %s : %s);\n",
1963 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
1966 if(temp_destination
) {
1967 shader_glsl_get_write_mask(ins
->dst
[0].token
, mask_char
);
1968 shader_glsl_append_dst(ins
->buffer
, ins
);
1969 shader_addline(ins
->buffer
, "tmp0%s);\n", mask_char
);
1975 /** Process the CND opcode in GLSL (dst = (src0 > 0.5) ? src1 : src2) */
1976 /* For ps 1.1-1.3, only a single component of src0 is used. For ps 1.4
1977 * the compare is done per component of src0. */
1978 static void shader_glsl_cnd(const struct wined3d_shader_instruction
*ins
)
1980 struct wined3d_shader_dst_param dst
;
1981 glsl_src_param_t src0_param
;
1982 glsl_src_param_t src1_param
;
1983 glsl_src_param_t src2_param
;
1984 DWORD write_mask
, cmp_channel
= 0;
1988 if (ins
->reg_maps
->shader_version
< WINED3DPS_VERSION(1, 4))
1990 write_mask
= shader_glsl_append_dst(ins
->buffer
, ins
);
1991 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
1992 shader_glsl_add_src_param(ins
, ins
->src
[1], ins
->src_addr
[1], write_mask
, &src1_param
);
1993 shader_glsl_add_src_param(ins
, ins
->src
[2], ins
->src_addr
[2], write_mask
, &src2_param
);
1995 /* Fun: The D3DSI_COISSUE flag changes the semantic of the cnd instruction for < 1.4 shaders */
1998 shader_addline(ins
->buffer
, "%s /* COISSUE! */);\n", src1_param
.param_str
);
2000 shader_addline(ins
->buffer
, "%s > 0.5 ? %s : %s);\n",
2001 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
2005 /* Cycle through all source0 channels */
2006 dst_mask
= ins
->dst
[0].write_mask
;
2008 for (i
=0; i
<4; i
++) {
2010 /* Find the destination channels which use the current source0 channel */
2011 for (j
=0; j
<4; j
++) {
2012 if (((ins
->src
[0] >> (WINED3DSP_SWIZZLE_SHIFT
+ 2 * j
)) & 0x3) == i
)
2014 write_mask
|= WINED3DSP_WRITEMASK_0
<< j
;
2015 cmp_channel
= WINED3DSP_WRITEMASK_0
<< j
;
2019 dst
.write_mask
= dst_mask
& write_mask
;
2020 dst
.token
= (dst
.token
& ~WINED3DSP_WRITEMASK_ALL
) | dst
.write_mask
;
2021 write_mask
= shader_glsl_append_dst_ext(ins
->buffer
, ins
, &dst
);
2022 if (!write_mask
) continue;
2024 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], cmp_channel
, &src0_param
);
2025 shader_glsl_add_src_param(ins
, ins
->src
[1], ins
->src_addr
[1], write_mask
, &src1_param
);
2026 shader_glsl_add_src_param(ins
, ins
->src
[2], ins
->src_addr
[2], write_mask
, &src2_param
);
2028 shader_addline(ins
->buffer
, "%s > 0.5 ? %s : %s);\n",
2029 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
2033 /** GLSL code generation for WINED3DSIO_MAD: Multiply the first 2 opcodes, then add the last */
2034 static void shader_glsl_mad(const struct wined3d_shader_instruction
*ins
)
2036 glsl_src_param_t src0_param
;
2037 glsl_src_param_t src1_param
;
2038 glsl_src_param_t src2_param
;
2041 write_mask
= shader_glsl_append_dst(ins
->buffer
, ins
);
2042 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], write_mask
, &src0_param
);
2043 shader_glsl_add_src_param(ins
, ins
->src
[1], ins
->src_addr
[1], write_mask
, &src1_param
);
2044 shader_glsl_add_src_param(ins
, ins
->src
[2], ins
->src_addr
[2], write_mask
, &src2_param
);
2045 shader_addline(ins
->buffer
, "(%s * %s) + %s);\n",
2046 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
2049 /** Handles transforming all WINED3DSIO_M?x? opcodes for
2050 Vertex shaders to GLSL codes */
2051 static void shader_glsl_mnxn(const struct wined3d_shader_instruction
*ins
)
2054 int nComponents
= 0;
2055 struct wined3d_shader_dst_param tmp_dst
= {0};
2056 struct wined3d_shader_instruction tmp_ins
;
2058 memset(&tmp_ins
, 0, sizeof(tmp_ins
));
2060 /* Set constants for the temporary argument */
2061 tmp_ins
.shader
= ins
->shader
;
2062 tmp_ins
.buffer
= ins
->buffer
;
2063 tmp_ins
.src
[0] = ins
->src
[0];
2064 tmp_ins
.src_addr
[0] = ins
->src_addr
[0];
2065 tmp_ins
.src_addr
[1] = ins
->src_addr
[1];
2066 tmp_ins
.reg_maps
= ins
->reg_maps
;
2067 tmp_ins
.dst_count
= 1;
2068 tmp_ins
.dst
= &tmp_dst
;
2069 tmp_ins
.src_count
= 2;
2071 switch(ins
->handler_idx
)
2073 case WINED3DSIH_M4x4
:
2075 tmp_ins
.handler_idx
= WINED3DSIH_DP4
;
2077 case WINED3DSIH_M4x3
:
2079 tmp_ins
.handler_idx
= WINED3DSIH_DP4
;
2081 case WINED3DSIH_M3x4
:
2083 tmp_ins
.handler_idx
= WINED3DSIH_DP3
;
2085 case WINED3DSIH_M3x3
:
2087 tmp_ins
.handler_idx
= WINED3DSIH_DP3
;
2089 case WINED3DSIH_M3x2
:
2091 tmp_ins
.handler_idx
= WINED3DSIH_DP3
;
2097 tmp_dst
= ins
->dst
[0];
2098 for (i
= 0; i
< nComponents
; ++i
)
2100 tmp_dst
.write_mask
= WINED3DSP_WRITEMASK_0
<< i
;
2101 tmp_dst
.token
= (tmp_dst
.token
& ~WINED3DSP_WRITEMASK_ALL
) | tmp_dst
.write_mask
;
2102 tmp_ins
.src
[1] = ins
->src
[1] + i
;
2103 shader_glsl_dot(&tmp_ins
);
2108 The LRP instruction performs a component-wise linear interpolation
2109 between the second and third operands using the first operand as the
2110 blend factor. Equation: (dst = src2 + src0 * (src1 - src2))
2111 This is equivalent to mix(src2, src1, src0);
2113 static void shader_glsl_lrp(const struct wined3d_shader_instruction
*ins
)
2115 glsl_src_param_t src0_param
;
2116 glsl_src_param_t src1_param
;
2117 glsl_src_param_t src2_param
;
2120 write_mask
= shader_glsl_append_dst(ins
->buffer
, ins
);
2122 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], write_mask
, &src0_param
);
2123 shader_glsl_add_src_param(ins
, ins
->src
[1], ins
->src_addr
[1], write_mask
, &src1_param
);
2124 shader_glsl_add_src_param(ins
, ins
->src
[2], ins
->src_addr
[2], write_mask
, &src2_param
);
2126 shader_addline(ins
->buffer
, "mix(%s, %s, %s));\n",
2127 src2_param
.param_str
, src1_param
.param_str
, src0_param
.param_str
);
2130 /** Process the WINED3DSIO_LIT instruction in GLSL:
2131 * dst.x = dst.w = 1.0
2132 * dst.y = (src0.x > 0) ? src0.x
2133 * dst.z = (src0.x > 0) ? ((src0.y > 0) ? pow(src0.y, src.w) : 0) : 0
2134 * where src.w is clamped at +- 128
2136 static void shader_glsl_lit(const struct wined3d_shader_instruction
*ins
)
2138 glsl_src_param_t src0_param
;
2139 glsl_src_param_t src1_param
;
2140 glsl_src_param_t src3_param
;
2143 shader_glsl_append_dst(ins
->buffer
, ins
);
2144 shader_glsl_get_write_mask(ins
->dst
[0].token
, dst_mask
);
2146 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2147 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], WINED3DSP_WRITEMASK_1
, &src1_param
);
2148 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], WINED3DSP_WRITEMASK_3
, &src3_param
);
2150 /* The sdk specifies the instruction like this
2152 * if(src.x > 0.0) dst.y = src.x
2154 * if(src.x > 0.0 && src.y > 0.0) dst.z = pow(src.y, power);
2158 * Obviously that has quite a few conditionals in it which we don't like. So the first step is this:
2159 * dst.x = 1.0 ... No further explanation needed
2160 * dst.y = max(src.y, 0.0); ... If x < 0.0, use 0.0, otherwise x. Same as the conditional
2161 * dst.z = x > 0.0 ? pow(max(y, 0.0), p) : 0; ... 0 ^ power is 0, and otherwise we use y anyway
2162 * dst.w = 1.0. ... Nothing fancy.
2164 * So we still have one conditional in there. So do this:
2165 * dst.z = pow(max(0.0, src.y) * step(0.0, src.x), power);
2167 * 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),
2168 * 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.
2169 * if both x and y are > 0, we get pow(y * 1.0, power), as it is supposed to
2171 shader_addline(ins
->buffer
,
2172 "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",
2173 src0_param
.param_str
, src1_param
.param_str
, src0_param
.param_str
, src3_param
.param_str
, dst_mask
);
2176 /** Process the WINED3DSIO_DST instruction in GLSL:
2178 * dst.y = src0.x * src0.y
2182 static void shader_glsl_dst(const struct wined3d_shader_instruction
*ins
)
2184 glsl_src_param_t src0y_param
;
2185 glsl_src_param_t src0z_param
;
2186 glsl_src_param_t src1y_param
;
2187 glsl_src_param_t src1w_param
;
2190 shader_glsl_append_dst(ins
->buffer
, ins
);
2191 shader_glsl_get_write_mask(ins
->dst
[0].token
, dst_mask
);
2193 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], WINED3DSP_WRITEMASK_1
, &src0y_param
);
2194 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], WINED3DSP_WRITEMASK_2
, &src0z_param
);
2195 shader_glsl_add_src_param(ins
, ins
->src
[1], ins
->src_addr
[1], WINED3DSP_WRITEMASK_1
, &src1y_param
);
2196 shader_glsl_add_src_param(ins
, ins
->src
[1], ins
->src_addr
[1], WINED3DSP_WRITEMASK_3
, &src1w_param
);
2198 shader_addline(ins
->buffer
, "vec4(1.0, %s * %s, %s, %s))%s;\n",
2199 src0y_param
.param_str
, src1y_param
.param_str
, src0z_param
.param_str
, src1w_param
.param_str
, dst_mask
);
2202 /** Process the WINED3DSIO_SINCOS instruction in GLSL:
2203 * VS 2.0 requires that specific cosine and sine constants be passed to this instruction so the hardware
2204 * can handle it. But, these functions are built-in for GLSL, so we can just ignore the last 2 params.
2206 * dst.x = cos(src0.?)
2207 * dst.y = sin(src0.?)
2211 static void shader_glsl_sincos(const struct wined3d_shader_instruction
*ins
)
2213 glsl_src_param_t src0_param
;
2216 write_mask
= shader_glsl_append_dst(ins
->buffer
, ins
);
2217 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2219 switch (write_mask
) {
2220 case WINED3DSP_WRITEMASK_0
:
2221 shader_addline(ins
->buffer
, "cos(%s));\n", src0_param
.param_str
);
2224 case WINED3DSP_WRITEMASK_1
:
2225 shader_addline(ins
->buffer
, "sin(%s));\n", src0_param
.param_str
);
2228 case (WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
):
2229 shader_addline(ins
->buffer
, "vec2(cos(%s), sin(%s)));\n", src0_param
.param_str
, src0_param
.param_str
);
2233 ERR("Write mask should be .x, .y or .xy\n");
2238 /** Process the WINED3DSIO_LOOP instruction in GLSL:
2239 * Start a for() loop where src1.y is the initial value of aL,
2240 * increment aL by src1.z for a total of src1.x iterations.
2241 * Need to use a temporary variable for this operation.
2243 /* FIXME: I don't think nested loops will work correctly this way. */
2244 static void shader_glsl_loop(const struct wined3d_shader_instruction
*ins
)
2246 glsl_src_param_t src1_param
;
2247 IWineD3DBaseShaderImpl
*shader
= (IWineD3DBaseShaderImpl
*)ins
->shader
;
2248 DWORD regtype
= shader_get_regtype(ins
->src
[1]);
2249 DWORD reg
= ins
->src
[1] & WINED3DSP_REGNUM_MASK
;
2250 const DWORD
*control_values
= NULL
;
2251 const local_constant
*constant
;
2253 shader_glsl_add_src_param(ins
, ins
->src
[1], ins
->src_addr
[1], WINED3DSP_WRITEMASK_ALL
, &src1_param
);
2255 /* Try to hardcode the loop control parameters if possible. Direct3D 9 class hardware doesn't support real
2256 * varying indexing, but Microsoft designed this feature for Shader model 2.x+. If the loop control is
2257 * known at compile time, the GLSL compiler can unroll the loop, and replace indirect addressing with direct
2260 if(regtype
== WINED3DSPR_CONSTINT
) {
2261 LIST_FOR_EACH_ENTRY(constant
, &shader
->baseShader
.constantsI
, local_constant
, entry
) {
2262 if(constant
->idx
== reg
) {
2263 control_values
= constant
->value
;
2269 if(control_values
) {
2270 if(control_values
[2] > 0) {
2271 shader_addline(ins
->buffer
, "for (aL%u = %d; aL%u < (%d * %d + %d); aL%u += %d) {\n",
2272 shader
->baseShader
.cur_loop_depth
, control_values
[1],
2273 shader
->baseShader
.cur_loop_depth
, control_values
[0], control_values
[2], control_values
[1],
2274 shader
->baseShader
.cur_loop_depth
, control_values
[2]);
2275 } else if(control_values
[2] == 0) {
2276 shader_addline(ins
->buffer
, "for (aL%u = %d, tmpInt%u = 0; tmpInt%u < %d; tmpInt%u++) {\n",
2277 shader
->baseShader
.cur_loop_depth
, control_values
[1], shader
->baseShader
.cur_loop_depth
,
2278 shader
->baseShader
.cur_loop_depth
, control_values
[0],
2279 shader
->baseShader
.cur_loop_depth
);
2281 shader_addline(ins
->buffer
, "for (aL%u = %d; aL%u > (%d * %d + %d); aL%u += %d) {\n",
2282 shader
->baseShader
.cur_loop_depth
, control_values
[1],
2283 shader
->baseShader
.cur_loop_depth
, control_values
[0], control_values
[2], control_values
[1],
2284 shader
->baseShader
.cur_loop_depth
, control_values
[2]);
2287 shader_addline(ins
->buffer
, "for (tmpInt%u = 0, aL%u = %s.y; tmpInt%u < %s.x; tmpInt%u++, aL%u += %s.z) {\n",
2288 shader
->baseShader
.cur_loop_depth
, shader
->baseShader
.cur_loop_regno
,
2289 src1_param
.reg_name
, shader
->baseShader
.cur_loop_depth
, src1_param
.reg_name
,
2290 shader
->baseShader
.cur_loop_depth
, shader
->baseShader
.cur_loop_regno
, src1_param
.reg_name
);
2293 shader
->baseShader
.cur_loop_depth
++;
2294 shader
->baseShader
.cur_loop_regno
++;
2297 static void shader_glsl_end(const struct wined3d_shader_instruction
*ins
)
2299 IWineD3DBaseShaderImpl
*shader
= (IWineD3DBaseShaderImpl
*)ins
->shader
;
2301 shader_addline(ins
->buffer
, "}\n");
2303 if (ins
->handler_idx
== WINED3DSIH_ENDLOOP
)
2305 shader
->baseShader
.cur_loop_depth
--;
2306 shader
->baseShader
.cur_loop_regno
--;
2309 if (ins
->handler_idx
== WINED3DSIH_ENDREP
)
2311 shader
->baseShader
.cur_loop_depth
--;
2315 static void shader_glsl_rep(const struct wined3d_shader_instruction
*ins
)
2317 IWineD3DBaseShaderImpl
*shader
= (IWineD3DBaseShaderImpl
*)ins
->shader
;
2318 glsl_src_param_t src0_param
;
2320 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2321 shader_addline(ins
->buffer
, "for (tmpInt%d = 0; tmpInt%d < %s; tmpInt%d++) {\n",
2322 shader
->baseShader
.cur_loop_depth
, shader
->baseShader
.cur_loop_depth
,
2323 src0_param
.param_str
, shader
->baseShader
.cur_loop_depth
);
2324 shader
->baseShader
.cur_loop_depth
++;
2327 static void shader_glsl_if(const struct wined3d_shader_instruction
*ins
)
2329 glsl_src_param_t src0_param
;
2331 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2332 shader_addline(ins
->buffer
, "if (%s) {\n", src0_param
.param_str
);
2335 static void shader_glsl_ifc(const struct wined3d_shader_instruction
*ins
)
2337 glsl_src_param_t src0_param
;
2338 glsl_src_param_t src1_param
;
2340 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2341 shader_glsl_add_src_param(ins
, ins
->src
[1], ins
->src_addr
[1], WINED3DSP_WRITEMASK_0
, &src1_param
);
2343 shader_addline(ins
->buffer
, "if (%s %s %s) {\n",
2344 src0_param
.param_str
, shader_get_comp_op(ins
->flags
), src1_param
.param_str
);
2347 static void shader_glsl_else(const struct wined3d_shader_instruction
*ins
)
2349 shader_addline(ins
->buffer
, "} else {\n");
2352 static void shader_glsl_break(const struct wined3d_shader_instruction
*ins
)
2354 shader_addline(ins
->buffer
, "break;\n");
2357 /* FIXME: According to MSDN the compare is done per component. */
2358 static void shader_glsl_breakc(const struct wined3d_shader_instruction
*ins
)
2360 glsl_src_param_t src0_param
;
2361 glsl_src_param_t src1_param
;
2363 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2364 shader_glsl_add_src_param(ins
, ins
->src
[1], ins
->src_addr
[1], WINED3DSP_WRITEMASK_0
, &src1_param
);
2366 shader_addline(ins
->buffer
, "if (%s %s %s) break;\n",
2367 src0_param
.param_str
, shader_get_comp_op(ins
->flags
), src1_param
.param_str
);
2370 static void shader_glsl_label(const struct wined3d_shader_instruction
*ins
)
2373 DWORD snum
= (ins
->src
[0]) & WINED3DSP_REGNUM_MASK
;
2374 shader_addline(ins
->buffer
, "}\n");
2375 shader_addline(ins
->buffer
, "void subroutine%u () {\n", snum
);
2378 static void shader_glsl_call(const struct wined3d_shader_instruction
*ins
)
2380 DWORD snum
= (ins
->src
[0]) & WINED3DSP_REGNUM_MASK
;
2381 shader_addline(ins
->buffer
, "subroutine%u();\n", snum
);
2384 static void shader_glsl_callnz(const struct wined3d_shader_instruction
*ins
)
2386 glsl_src_param_t src1_param
;
2388 DWORD snum
= (ins
->src
[0]) & WINED3DSP_REGNUM_MASK
;
2389 shader_glsl_add_src_param(ins
, ins
->src
[1], ins
->src_addr
[1], WINED3DSP_WRITEMASK_0
, &src1_param
);
2390 shader_addline(ins
->buffer
, "if (%s) subroutine%u();\n", src1_param
.param_str
, snum
);
2393 /*********************************************
2394 * Pixel Shader Specific Code begins here
2395 ********************************************/
2396 static void pshader_glsl_tex(const struct wined3d_shader_instruction
*ins
)
2398 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*)ins
->shader
;
2399 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
2400 DWORD shader_version
= ins
->reg_maps
->shader_version
;
2401 glsl_sample_function_t sample_function
;
2402 DWORD sample_flags
= 0;
2405 DWORD mask
= 0, swizzle
;
2407 /* 1.0-1.4: Use destination register as sampler source.
2408 * 2.0+: Use provided sampler source. */
2409 if (shader_version
< WINED3DPS_VERSION(2,0)) sampler_idx
= ins
->dst
[0].register_idx
;
2410 else sampler_idx
= ins
->src
[1] & WINED3DSP_REGNUM_MASK
;
2411 sampler_type
= ins
->reg_maps
->samplers
[sampler_idx
] & WINED3DSP_TEXTURETYPE_MASK
;
2413 if (shader_version
< WINED3DPS_VERSION(1,4))
2415 DWORD flags
= deviceImpl
->stateBlock
->textureState
[sampler_idx
][WINED3DTSS_TEXTURETRANSFORMFLAGS
];
2417 /* Projected cube textures don't make a lot of sense, the resulting coordinates stay the same. */
2418 if (flags
& WINED3DTTFF_PROJECTED
&& sampler_type
!= WINED3DSTT_CUBE
) {
2419 sample_flags
|= WINED3D_GLSL_SAMPLE_PROJECTED
;
2420 switch (flags
& ~WINED3DTTFF_PROJECTED
) {
2421 case WINED3DTTFF_COUNT1
: FIXME("WINED3DTTFF_PROJECTED with WINED3DTTFF_COUNT1?\n"); break;
2422 case WINED3DTTFF_COUNT2
: mask
= WINED3DSP_WRITEMASK_1
; break;
2423 case WINED3DTTFF_COUNT3
: mask
= WINED3DSP_WRITEMASK_2
; break;
2424 case WINED3DTTFF_COUNT4
:
2425 case WINED3DTTFF_DISABLE
: mask
= WINED3DSP_WRITEMASK_3
; break;
2429 else if (shader_version
< WINED3DPS_VERSION(2,0))
2431 DWORD src_mod
= ins
->src
[0] & WINED3DSP_SRCMOD_MASK
;
2433 if (src_mod
== WINED3DSPSM_DZ
) {
2434 sample_flags
|= WINED3D_GLSL_SAMPLE_PROJECTED
;
2435 mask
= WINED3DSP_WRITEMASK_2
;
2436 } else if (src_mod
== WINED3DSPSM_DW
) {
2437 sample_flags
|= WINED3D_GLSL_SAMPLE_PROJECTED
;
2438 mask
= WINED3DSP_WRITEMASK_3
;
2441 if (ins
->flags
& WINED3DSI_TEXLD_PROJECT
)
2443 /* ps 2.0 texldp instruction always divides by the fourth component. */
2444 sample_flags
|= WINED3D_GLSL_SAMPLE_PROJECTED
;
2445 mask
= WINED3DSP_WRITEMASK_3
;
2449 if(deviceImpl
->stateBlock
->textures
[sampler_idx
] &&
2450 IWineD3DBaseTexture_GetTextureDimensions(deviceImpl
->stateBlock
->textures
[sampler_idx
]) == GL_TEXTURE_RECTANGLE_ARB
) {
2451 sample_flags
|= WINED3D_GLSL_SAMPLE_RECT
;
2454 shader_glsl_get_sample_function(sampler_type
, sample_flags
, &sample_function
);
2455 mask
|= sample_function
.coord_mask
;
2457 if (shader_version
< WINED3DPS_VERSION(2,0)) swizzle
= WINED3DVS_NOSWIZZLE
;
2458 else swizzle
= ins
->src
[1] & WINED3DSP_SWIZZLE_MASK
;
2460 /* 1.0-1.3: Use destination register as coordinate source.
2461 1.4+: Use provided coordinate source register. */
2462 if (shader_version
< WINED3DPS_VERSION(1,4))
2465 shader_glsl_get_write_mask(mask
, coord_mask
);
2466 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, swizzle
, NULL
,
2467 "T%u%s", sampler_idx
, coord_mask
);
2469 glsl_src_param_t coord_param
;
2470 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], mask
, &coord_param
);
2471 if (ins
->flags
& WINED3DSI_TEXLD_BIAS
)
2473 glsl_src_param_t bias
;
2474 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], WINED3DSP_WRITEMASK_3
, &bias
);
2475 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, swizzle
, bias
.param_str
,
2476 "%s", coord_param
.param_str
);
2478 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, swizzle
, NULL
,
2479 "%s", coord_param
.param_str
);
2484 static void shader_glsl_texldl(const struct wined3d_shader_instruction
*ins
)
2486 IWineD3DBaseShaderImpl
*This
= (IWineD3DBaseShaderImpl
*)ins
->shader
;
2487 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
2488 glsl_sample_function_t sample_function
;
2489 glsl_src_param_t coord_param
, lod_param
;
2490 DWORD sample_flags
= WINED3D_GLSL_SAMPLE_LOD
;
2493 DWORD swizzle
= ins
->src
[1] & WINED3DSP_SWIZZLE_MASK
;
2495 sampler_idx
= ins
->src
[1] & WINED3DSP_REGNUM_MASK
;
2496 sampler_type
= ins
->reg_maps
->samplers
[sampler_idx
] & WINED3DSP_TEXTURETYPE_MASK
;
2497 if(deviceImpl
->stateBlock
->textures
[sampler_idx
] &&
2498 IWineD3DBaseTexture_GetTextureDimensions(deviceImpl
->stateBlock
->textures
[sampler_idx
]) == GL_TEXTURE_RECTANGLE_ARB
) {
2499 sample_flags
|= WINED3D_GLSL_SAMPLE_RECT
;
2501 shader_glsl_get_sample_function(sampler_type
, sample_flags
, &sample_function
);
2502 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], sample_function
.coord_mask
, &coord_param
);
2504 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], WINED3DSP_WRITEMASK_3
, &lod_param
);
2506 if (shader_is_pshader_version(ins
->reg_maps
->shader_version
))
2508 /* The GLSL spec claims the Lod sampling functions are only supported in vertex shaders.
2509 * However, they seem to work just fine in fragment shaders as well. */
2510 WARN("Using %s in fragment shader.\n", sample_function
.name
);
2512 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, swizzle
, lod_param
.param_str
,
2513 "%s", coord_param
.param_str
);
2516 static void pshader_glsl_texcoord(const struct wined3d_shader_instruction
*ins
)
2518 /* FIXME: Make this work for more than just 2D textures */
2519 SHADER_BUFFER
*buffer
= ins
->buffer
;
2523 write_mask
= shader_glsl_append_dst(ins
->buffer
, ins
);
2524 shader_glsl_get_write_mask(write_mask
, dst_mask
);
2526 if (ins
->reg_maps
->shader_version
!= WINED3DPS_VERSION(1,4))
2528 DWORD reg
= ins
->dst
[0].register_idx
;
2529 shader_addline(buffer
, "clamp(gl_TexCoord[%u], 0.0, 1.0)%s);\n", reg
, dst_mask
);
2531 DWORD reg
= ins
->src
[0] & WINED3DSP_REGNUM_MASK
;
2532 DWORD src_mod
= ins
->src
[0] & WINED3DSP_SRCMOD_MASK
;
2533 char dst_swizzle
[6];
2535 shader_glsl_get_swizzle(ins
->src
[0], FALSE
, write_mask
, dst_swizzle
);
2537 if (src_mod
== WINED3DSPSM_DZ
) {
2538 glsl_src_param_t div_param
;
2539 unsigned int mask_size
= shader_glsl_get_write_mask_size(write_mask
);
2540 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], WINED3DSP_WRITEMASK_2
, &div_param
);
2542 if (mask_size
> 1) {
2543 shader_addline(buffer
, "gl_TexCoord[%u]%s / vec%d(%s));\n", reg
, dst_swizzle
, mask_size
, div_param
.param_str
);
2545 shader_addline(buffer
, "gl_TexCoord[%u]%s / %s);\n", reg
, dst_swizzle
, div_param
.param_str
);
2547 } else if (src_mod
== WINED3DSPSM_DW
) {
2548 glsl_src_param_t div_param
;
2549 unsigned int mask_size
= shader_glsl_get_write_mask_size(write_mask
);
2550 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], WINED3DSP_WRITEMASK_3
, &div_param
);
2552 if (mask_size
> 1) {
2553 shader_addline(buffer
, "gl_TexCoord[%u]%s / vec%d(%s));\n", reg
, dst_swizzle
, mask_size
, div_param
.param_str
);
2555 shader_addline(buffer
, "gl_TexCoord[%u]%s / %s);\n", reg
, dst_swizzle
, div_param
.param_str
);
2558 shader_addline(buffer
, "gl_TexCoord[%u]%s);\n", reg
, dst_swizzle
);
2563 /** Process the WINED3DSIO_TEXDP3TEX instruction in GLSL:
2564 * Take a 3-component dot product of the TexCoord[dstreg] and src,
2565 * then perform a 1D texture lookup from stage dstregnum, place into dst. */
2566 static void pshader_glsl_texdp3tex(const struct wined3d_shader_instruction
*ins
)
2568 glsl_src_param_t src0_param
;
2569 glsl_sample_function_t sample_function
;
2570 DWORD sampler_idx
= ins
->dst
[0].register_idx
;
2571 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2572 DWORD sampler_type
= ins
->reg_maps
->samplers
[sampler_idx
] & WINED3DSP_TEXTURETYPE_MASK
;
2575 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], src_mask
, &src0_param
);
2577 /* Do I have to take care about the projected bit? I don't think so, since the dp3 returns only one
2578 * scalar, and projected sampling would require 4.
2580 * It is a dependent read - not valid with conditional NP2 textures
2582 shader_glsl_get_sample_function(sampler_type
, 0, &sample_function
);
2583 mask_size
= shader_glsl_get_write_mask_size(sample_function
.coord_mask
);
2588 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, WINED3DVS_NOSWIZZLE
, NULL
,
2589 "dot(gl_TexCoord[%u].xyz, %s)", sampler_idx
, src0_param
.param_str
);
2593 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, WINED3DVS_NOSWIZZLE
, NULL
,
2594 "vec2(dot(gl_TexCoord[%u].xyz, %s), 0.0)", sampler_idx
, src0_param
.param_str
);
2598 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, WINED3DVS_NOSWIZZLE
, NULL
,
2599 "vec3(dot(gl_TexCoord[%u].xyz, %s), 0.0, 0.0)", sampler_idx
, src0_param
.param_str
);
2603 FIXME("Unexpected mask size %u\n", mask_size
);
2608 /** Process the WINED3DSIO_TEXDP3 instruction in GLSL:
2609 * Take a 3-component dot product of the TexCoord[dstreg] and src. */
2610 static void pshader_glsl_texdp3(const struct wined3d_shader_instruction
*ins
)
2612 glsl_src_param_t src0_param
;
2613 DWORD dstreg
= ins
->dst
[0].register_idx
;
2614 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2616 unsigned int mask_size
;
2618 dst_mask
= shader_glsl_append_dst(ins
->buffer
, ins
);
2619 mask_size
= shader_glsl_get_write_mask_size(dst_mask
);
2620 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], src_mask
, &src0_param
);
2622 if (mask_size
> 1) {
2623 shader_addline(ins
->buffer
, "vec%d(dot(T%u.xyz, %s)));\n", mask_size
, dstreg
, src0_param
.param_str
);
2625 shader_addline(ins
->buffer
, "dot(T%u.xyz, %s));\n", dstreg
, src0_param
.param_str
);
2629 /** Process the WINED3DSIO_TEXDEPTH instruction in GLSL:
2630 * Calculate the depth as dst.x / dst.y */
2631 static void pshader_glsl_texdepth(const struct wined3d_shader_instruction
*ins
)
2633 glsl_dst_param_t dst_param
;
2635 shader_glsl_add_dst_param(ins
, &ins
->dst
[0], &dst_param
);
2637 /* Tests show that texdepth never returns anything below 0.0, and that r5.y is clamped to 1.0.
2638 * Negative input is accepted, -0.25 / -0.5 returns 0.5. GL should clamp gl_FragDepth to [0;1], but
2639 * this doesn't always work, so clamp the results manually. Whether or not the x value is clamped at 1
2640 * too is irrelevant, since if x = 0, any y value < 1.0 (and > 1.0 is not allowed) results in a result
2643 shader_addline(ins
->buffer
, "gl_FragDepth = clamp((%s.x / min(%s.y, 1.0)), 0.0, 1.0);\n",
2644 dst_param
.reg_name
, dst_param
.reg_name
);
2647 /** Process the WINED3DSIO_TEXM3X2DEPTH instruction in GLSL:
2648 * Last row of a 3x2 matrix multiply, use the result to calculate the depth:
2649 * Calculate tmp0.y = TexCoord[dstreg] . src.xyz; (tmp0.x has already been calculated)
2650 * depth = (tmp0.y == 0.0) ? 1.0 : tmp0.x / tmp0.y
2652 static void pshader_glsl_texm3x2depth(const struct wined3d_shader_instruction
*ins
)
2654 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2655 DWORD dstreg
= ins
->dst
[0].register_idx
;
2656 glsl_src_param_t src0_param
;
2658 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], src_mask
, &src0_param
);
2660 shader_addline(ins
->buffer
, "tmp0.y = dot(T%u.xyz, %s);\n", dstreg
, src0_param
.param_str
);
2661 shader_addline(ins
->buffer
, "gl_FragDepth = (tmp0.y == 0.0) ? 1.0 : clamp(tmp0.x / tmp0.y, 0.0, 1.0);\n");
2664 /** Process the WINED3DSIO_TEXM3X2PAD instruction in GLSL
2665 * Calculate the 1st of a 2-row matrix multiplication. */
2666 static void pshader_glsl_texm3x2pad(const struct wined3d_shader_instruction
*ins
)
2668 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2669 DWORD reg
= ins
->dst
[0].register_idx
;
2670 SHADER_BUFFER
*buffer
= ins
->buffer
;
2671 glsl_src_param_t src0_param
;
2673 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], src_mask
, &src0_param
);
2674 shader_addline(buffer
, "tmp0.x = dot(T%u.xyz, %s);\n", reg
, src0_param
.param_str
);
2677 /** Process the WINED3DSIO_TEXM3X3PAD instruction in GLSL
2678 * Calculate the 1st or 2nd row of a 3-row matrix multiplication. */
2679 static void pshader_glsl_texm3x3pad(const struct wined3d_shader_instruction
*ins
)
2681 IWineD3DPixelShaderImpl
*shader
= (IWineD3DPixelShaderImpl
*)ins
->shader
;
2682 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2683 DWORD reg
= ins
->dst
[0].register_idx
;
2684 SHADER_BUFFER
*buffer
= ins
->buffer
;
2685 SHADER_PARSE_STATE
* current_state
= &shader
->baseShader
.parse_state
;
2686 glsl_src_param_t src0_param
;
2688 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], src_mask
, &src0_param
);
2689 shader_addline(buffer
, "tmp0.%c = dot(T%u.xyz, %s);\n", 'x' + current_state
->current_row
, reg
, src0_param
.param_str
);
2690 current_state
->texcoord_w
[current_state
->current_row
++] = reg
;
2693 static void pshader_glsl_texm3x2tex(const struct wined3d_shader_instruction
*ins
)
2695 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2696 DWORD reg
= ins
->dst
[0].register_idx
;
2697 SHADER_BUFFER
*buffer
= ins
->buffer
;
2698 glsl_src_param_t src0_param
;
2699 DWORD sampler_type
= ins
->reg_maps
->samplers
[reg
] & WINED3DSP_TEXTURETYPE_MASK
;
2700 glsl_sample_function_t sample_function
;
2702 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], src_mask
, &src0_param
);
2703 shader_addline(buffer
, "tmp0.y = dot(T%u.xyz, %s);\n", reg
, src0_param
.param_str
);
2705 shader_glsl_get_sample_function(sampler_type
, 0, &sample_function
);
2707 /* Sample the texture using the calculated coordinates */
2708 shader_glsl_gen_sample_code(ins
, reg
, &sample_function
, WINED3DVS_NOSWIZZLE
, NULL
, "tmp0.xy");
2711 /** Process the WINED3DSIO_TEXM3X3TEX instruction in GLSL
2712 * Perform the 3rd row of a 3x3 matrix multiply, then sample the texture using the calculated coordinates */
2713 static void pshader_glsl_texm3x3tex(const struct wined3d_shader_instruction
*ins
)
2715 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2716 glsl_src_param_t src0_param
;
2717 DWORD reg
= ins
->dst
[0].register_idx
;
2718 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*)ins
->shader
;
2719 SHADER_PARSE_STATE
* current_state
= &This
->baseShader
.parse_state
;
2720 DWORD sampler_type
= ins
->reg_maps
->samplers
[reg
] & WINED3DSP_TEXTURETYPE_MASK
;
2721 glsl_sample_function_t sample_function
;
2723 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], src_mask
, &src0_param
);
2724 shader_addline(ins
->buffer
, "tmp0.z = dot(T%u.xyz, %s);\n", reg
, src0_param
.param_str
);
2726 /* Dependent read, not valid with conditional NP2 */
2727 shader_glsl_get_sample_function(sampler_type
, 0, &sample_function
);
2729 /* Sample the texture using the calculated coordinates */
2730 shader_glsl_gen_sample_code(ins
, reg
, &sample_function
, WINED3DVS_NOSWIZZLE
, NULL
, "tmp0.xyz");
2732 current_state
->current_row
= 0;
2735 /** Process the WINED3DSIO_TEXM3X3 instruction in GLSL
2736 * Perform the 3rd row of a 3x3 matrix multiply */
2737 static void pshader_glsl_texm3x3(const struct wined3d_shader_instruction
*ins
)
2739 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2740 glsl_src_param_t src0_param
;
2742 DWORD reg
= ins
->dst
[0].register_idx
;
2743 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*)ins
->shader
;
2744 SHADER_PARSE_STATE
* current_state
= &This
->baseShader
.parse_state
;
2746 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], src_mask
, &src0_param
);
2748 shader_glsl_append_dst(ins
->buffer
, ins
);
2749 shader_glsl_get_write_mask(ins
->dst
[0].token
, dst_mask
);
2750 shader_addline(ins
->buffer
, "vec4(tmp0.xy, dot(T%u.xyz, %s), 1.0)%s);\n", reg
, src0_param
.param_str
, dst_mask
);
2752 current_state
->current_row
= 0;
2755 /** Process the WINED3DSIO_TEXM3X3SPEC instruction in GLSL
2756 * Perform the final texture lookup based on the previous 2 3x3 matrix multiplies */
2757 static void pshader_glsl_texm3x3spec(const struct wined3d_shader_instruction
*ins
)
2759 IWineD3DPixelShaderImpl
*shader
= (IWineD3DPixelShaderImpl
*)ins
->shader
;
2760 DWORD reg
= ins
->dst
[0].register_idx
;
2761 glsl_src_param_t src0_param
;
2762 glsl_src_param_t src1_param
;
2763 SHADER_BUFFER
*buffer
= ins
->buffer
;
2764 SHADER_PARSE_STATE
* current_state
= &shader
->baseShader
.parse_state
;
2765 DWORD stype
= ins
->reg_maps
->samplers
[reg
] & WINED3DSP_TEXTURETYPE_MASK
;
2766 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2767 glsl_sample_function_t sample_function
;
2769 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], src_mask
, &src0_param
);
2770 shader_glsl_add_src_param(ins
, ins
->src
[1], ins
->src_addr
[1], src_mask
, &src1_param
);
2772 /* Perform the last matrix multiply operation */
2773 shader_addline(buffer
, "tmp0.z = dot(T%u.xyz, %s);\n", reg
, src0_param
.param_str
);
2774 /* Reflection calculation */
2775 shader_addline(buffer
, "tmp0.xyz = -reflect((%s), normalize(tmp0.xyz));\n", src1_param
.param_str
);
2777 /* Dependent read, not valid with conditional NP2 */
2778 shader_glsl_get_sample_function(stype
, 0, &sample_function
);
2780 /* Sample the texture */
2781 shader_glsl_gen_sample_code(ins
, reg
, &sample_function
, WINED3DVS_NOSWIZZLE
, NULL
, "tmp0.xyz");
2783 current_state
->current_row
= 0;
2786 /** Process the WINED3DSIO_TEXM3X3VSPEC instruction in GLSL
2787 * Perform the final texture lookup based on the previous 2 3x3 matrix multiplies */
2788 static void pshader_glsl_texm3x3vspec(const struct wined3d_shader_instruction
*ins
)
2790 IWineD3DPixelShaderImpl
*shader
= (IWineD3DPixelShaderImpl
*)ins
->shader
;
2791 DWORD reg
= ins
->dst
[0].register_idx
;
2792 SHADER_BUFFER
*buffer
= ins
->buffer
;
2793 SHADER_PARSE_STATE
* current_state
= &shader
->baseShader
.parse_state
;
2794 glsl_src_param_t src0_param
;
2795 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2796 DWORD sampler_type
= ins
->reg_maps
->samplers
[reg
] & WINED3DSP_TEXTURETYPE_MASK
;
2797 glsl_sample_function_t sample_function
;
2799 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], src_mask
, &src0_param
);
2801 /* Perform the last matrix multiply operation */
2802 shader_addline(buffer
, "tmp0.z = dot(vec3(T%u), vec3(%s));\n", reg
, src0_param
.param_str
);
2804 /* Construct the eye-ray vector from w coordinates */
2805 shader_addline(buffer
, "tmp1.xyz = normalize(vec3(gl_TexCoord[%u].w, gl_TexCoord[%u].w, gl_TexCoord[%u].w));\n",
2806 current_state
->texcoord_w
[0], current_state
->texcoord_w
[1], reg
);
2807 shader_addline(buffer
, "tmp0.xyz = -reflect(tmp1.xyz, normalize(tmp0.xyz));\n");
2809 /* Dependent read, not valid with conditional NP2 */
2810 shader_glsl_get_sample_function(sampler_type
, 0, &sample_function
);
2812 /* Sample the texture using the calculated coordinates */
2813 shader_glsl_gen_sample_code(ins
, reg
, &sample_function
, WINED3DVS_NOSWIZZLE
, NULL
, "tmp0.xyz");
2815 current_state
->current_row
= 0;
2818 /** Process the WINED3DSIO_TEXBEM instruction in GLSL.
2819 * Apply a fake bump map transform.
2820 * texbem is pshader <= 1.3 only, this saves a few version checks
2822 static void pshader_glsl_texbem(const struct wined3d_shader_instruction
*ins
)
2824 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*)ins
->shader
;
2825 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
2826 glsl_sample_function_t sample_function
;
2827 glsl_src_param_t coord_param
;
2834 sampler_idx
= ins
->dst
[0].register_idx
;
2835 flags
= deviceImpl
->stateBlock
->textureState
[sampler_idx
][WINED3DTSS_TEXTURETRANSFORMFLAGS
];
2837 sampler_type
= ins
->reg_maps
->samplers
[sampler_idx
] & WINED3DSP_TEXTURETYPE_MASK
;
2838 /* Dependent read, not valid with conditional NP2 */
2839 shader_glsl_get_sample_function(sampler_type
, 0, &sample_function
);
2840 mask
= sample_function
.coord_mask
;
2842 shader_glsl_get_write_mask(mask
, coord_mask
);
2844 /* with projective textures, texbem only divides the static texture coord, not the displacement,
2845 * so we can't let the GL handle this.
2847 if (flags
& WINED3DTTFF_PROJECTED
) {
2849 char coord_div_mask
[3];
2850 switch (flags
& ~WINED3DTTFF_PROJECTED
) {
2851 case WINED3DTTFF_COUNT1
: FIXME("WINED3DTTFF_PROJECTED with WINED3DTTFF_COUNT1?\n"); break;
2852 case WINED3DTTFF_COUNT2
: div_mask
= WINED3DSP_WRITEMASK_1
; break;
2853 case WINED3DTTFF_COUNT3
: div_mask
= WINED3DSP_WRITEMASK_2
; break;
2854 case WINED3DTTFF_COUNT4
:
2855 case WINED3DTTFF_DISABLE
: div_mask
= WINED3DSP_WRITEMASK_3
; break;
2857 shader_glsl_get_write_mask(div_mask
, coord_div_mask
);
2858 shader_addline(ins
->buffer
, "T%u%s /= T%u%s;\n", sampler_idx
, coord_mask
, sampler_idx
, coord_div_mask
);
2861 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0],
2862 WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
, &coord_param
);
2864 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, WINED3DVS_NOSWIZZLE
, NULL
,
2865 "T%u%s + vec4(bumpenvmat%d * %s, 0.0, 0.0)%s", sampler_idx
, coord_mask
, sampler_idx
,
2866 coord_param
.param_str
, coord_mask
);
2868 if (ins
->handler_idx
== WINED3DSIH_TEXBEML
)
2870 glsl_src_param_t luminance_param
;
2871 glsl_dst_param_t dst_param
;
2873 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], WINED3DSP_WRITEMASK_2
, &luminance_param
);
2874 shader_glsl_add_dst_param(ins
, &ins
->dst
[0], &dst_param
);
2876 shader_addline(ins
->buffer
, "%s%s *= (%s * luminancescale%d + luminanceoffset%d);\n",
2877 dst_param
.reg_name
, dst_param
.mask_str
,
2878 luminance_param
.param_str
, sampler_idx
, sampler_idx
);
2882 static void pshader_glsl_bem(const struct wined3d_shader_instruction
*ins
)
2884 glsl_src_param_t src0_param
, src1_param
;
2885 DWORD sampler_idx
= ins
->dst
[0].register_idx
;
2887 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0],
2888 WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
, &src0_param
);
2889 shader_glsl_add_src_param(ins
, ins
->src
[1], ins
->src_addr
[1],
2890 WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
, &src1_param
);
2892 shader_glsl_append_dst(ins
->buffer
, ins
);
2893 shader_addline(ins
->buffer
, "%s + bumpenvmat%d * %s);\n",
2894 src0_param
.param_str
, sampler_idx
, src1_param
.param_str
);
2897 /** Process the WINED3DSIO_TEXREG2AR instruction in GLSL
2898 * Sample 2D texture at dst using the alpha & red (wx) components of src as texture coordinates */
2899 static void pshader_glsl_texreg2ar(const struct wined3d_shader_instruction
*ins
)
2901 glsl_src_param_t src0_param
;
2902 DWORD sampler_idx
= ins
->dst
[0].register_idx
;
2903 DWORD sampler_type
= ins
->reg_maps
->samplers
[sampler_idx
] & WINED3DSP_TEXTURETYPE_MASK
;
2904 glsl_sample_function_t sample_function
;
2906 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], WINED3DSP_WRITEMASK_ALL
, &src0_param
);
2908 shader_glsl_get_sample_function(sampler_type
, 0, &sample_function
);
2909 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, WINED3DVS_NOSWIZZLE
, NULL
,
2910 "%s.wx", src0_param
.reg_name
);
2913 /** Process the WINED3DSIO_TEXREG2GB instruction in GLSL
2914 * Sample 2D texture at dst using the green & blue (yz) components of src as texture coordinates */
2915 static void pshader_glsl_texreg2gb(const struct wined3d_shader_instruction
*ins
)
2917 glsl_src_param_t src0_param
;
2918 DWORD sampler_idx
= ins
->dst
[0].register_idx
;
2919 DWORD sampler_type
= ins
->reg_maps
->samplers
[sampler_idx
] & WINED3DSP_TEXTURETYPE_MASK
;
2920 glsl_sample_function_t sample_function
;
2922 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], WINED3DSP_WRITEMASK_ALL
, &src0_param
);
2924 shader_glsl_get_sample_function(sampler_type
, 0, &sample_function
);
2925 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, WINED3DVS_NOSWIZZLE
, NULL
,
2926 "%s.yz", src0_param
.reg_name
);
2929 /** Process the WINED3DSIO_TEXREG2RGB instruction in GLSL
2930 * Sample texture at dst using the rgb (xyz) components of src as texture coordinates */
2931 static void pshader_glsl_texreg2rgb(const struct wined3d_shader_instruction
*ins
)
2933 glsl_src_param_t src0_param
;
2934 DWORD sampler_idx
= ins
->dst
[0].register_idx
;
2935 DWORD sampler_type
= ins
->reg_maps
->samplers
[sampler_idx
] & WINED3DSP_TEXTURETYPE_MASK
;
2936 glsl_sample_function_t sample_function
;
2938 /* Dependent read, not valid with conditional NP2 */
2939 shader_glsl_get_sample_function(sampler_type
, 0, &sample_function
);
2940 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], sample_function
.coord_mask
, &src0_param
);
2942 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, WINED3DVS_NOSWIZZLE
, NULL
,
2943 "%s", src0_param
.param_str
);
2946 /** Process the WINED3DSIO_TEXKILL instruction in GLSL.
2947 * If any of the first 3 components are < 0, discard this pixel */
2948 static void pshader_glsl_texkill(const struct wined3d_shader_instruction
*ins
)
2950 glsl_dst_param_t dst_param
;
2952 /* The argument is a destination parameter, and no writemasks are allowed */
2953 shader_glsl_add_dst_param(ins
, &ins
->dst
[0], &dst_param
);
2954 if ((ins
->reg_maps
->shader_version
>= WINED3DPS_VERSION(2,0)))
2956 /* 2.0 shaders compare all 4 components in texkill */
2957 shader_addline(ins
->buffer
, "if (any(lessThan(%s.xyzw, vec4(0.0)))) discard;\n", dst_param
.reg_name
);
2959 /* 1.X shaders only compare the first 3 components, probably due to the nature of the texkill
2960 * instruction as a tex* instruction, and phase, which kills all a / w components. Even if all
2961 * 4 components are defined, only the first 3 are used
2963 shader_addline(ins
->buffer
, "if (any(lessThan(%s.xyz, vec3(0.0)))) discard;\n", dst_param
.reg_name
);
2967 /** Process the WINED3DSIO_DP2ADD instruction in GLSL.
2968 * dst = dot2(src0, src1) + src2 */
2969 static void pshader_glsl_dp2add(const struct wined3d_shader_instruction
*ins
)
2971 glsl_src_param_t src0_param
;
2972 glsl_src_param_t src1_param
;
2973 glsl_src_param_t src2_param
;
2975 unsigned int mask_size
;
2977 write_mask
= shader_glsl_append_dst(ins
->buffer
, ins
);
2978 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
2980 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0],
2981 WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
, &src0_param
);
2982 shader_glsl_add_src_param(ins
, ins
->src
[1], ins
->src_addr
[1],
2983 WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
, &src1_param
);
2984 shader_glsl_add_src_param(ins
, ins
->src
[2], ins
->src_addr
[2],
2985 WINED3DSP_WRITEMASK_0
, &src2_param
);
2987 if (mask_size
> 1) {
2988 shader_addline(ins
->buffer
, "vec%d(dot(%s, %s) + %s));\n",
2989 mask_size
, src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
2991 shader_addline(ins
->buffer
, "dot(%s, %s) + %s);\n",
2992 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
2996 static void pshader_glsl_input_pack(SHADER_BUFFER
* buffer
, const struct semantic
* semantics_in
,
2997 IWineD3DPixelShader
*iface
, enum vertexprocessing_mode vertexprocessing
)
3000 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*) iface
;
3002 for (i
= 0; i
< MAX_REG_INPUT
; i
++) {
3004 DWORD usage_token
= semantics_in
[i
].usage
;
3005 DWORD register_token
= semantics_in
[i
].reg
;
3006 DWORD usage
, usage_idx
;
3010 if (!usage_token
) continue;
3011 usage
= (usage_token
& WINED3DSP_DCL_USAGE_MASK
) >> WINED3DSP_DCL_USAGE_SHIFT
;
3012 usage_idx
= (usage_token
& WINED3DSP_DCL_USAGEINDEX_MASK
) >> WINED3DSP_DCL_USAGEINDEX_SHIFT
;
3013 shader_glsl_get_write_mask(register_token
, reg_mask
);
3017 case WINED3DDECLUSAGE_TEXCOORD
:
3018 if(usage_idx
< 8 && vertexprocessing
== pretransformed
) {
3019 shader_addline(buffer
, "IN[%u]%s = gl_TexCoord[%u]%s;\n",
3020 This
->input_reg_map
[i
], reg_mask
, usage_idx
, reg_mask
);
3022 shader_addline(buffer
, "IN[%u]%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
3023 This
->input_reg_map
[i
], reg_mask
, reg_mask
);
3027 case WINED3DDECLUSAGE_COLOR
:
3029 shader_addline(buffer
, "IN[%u]%s = vec4(gl_Color)%s;\n",
3030 This
->input_reg_map
[i
], reg_mask
, reg_mask
);
3031 else if (usage_idx
== 1)
3032 shader_addline(buffer
, "IN[%u]%s = vec4(gl_SecondaryColor)%s;\n",
3033 This
->input_reg_map
[i
], reg_mask
, reg_mask
);
3035 shader_addline(buffer
, "IN[%u]%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
3036 This
->input_reg_map
[i
], reg_mask
, reg_mask
);
3040 shader_addline(buffer
, "IN[%u]%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
3041 This
->input_reg_map
[i
], reg_mask
, reg_mask
);
3046 /*********************************************
3047 * Vertex Shader Specific Code begins here
3048 ********************************************/
3050 static void add_glsl_program_entry(struct shader_glsl_priv
*priv
, struct glsl_shader_prog_link
*entry
) {
3051 glsl_program_key_t
*key
;
3053 key
= HeapAlloc(GetProcessHeap(), 0, sizeof(glsl_program_key_t
));
3054 key
->vshader
= entry
->vshader
;
3055 key
->pshader
= entry
->pshader
;
3056 key
->vs_args
= entry
->vs_args
;
3057 key
->ps_args
= entry
->ps_args
;
3059 hash_table_put(priv
->glsl_program_lookup
, key
, entry
);
3062 static struct glsl_shader_prog_link
*get_glsl_program_entry(struct shader_glsl_priv
*priv
,
3063 IWineD3DVertexShader
*vshader
, IWineD3DPixelShader
*pshader
, struct vs_compile_args
*vs_args
,
3064 struct ps_compile_args
*ps_args
) {
3065 glsl_program_key_t key
;
3067 key
.vshader
= vshader
;
3068 key
.pshader
= pshader
;
3069 key
.vs_args
= *vs_args
;
3070 key
.ps_args
= *ps_args
;
3072 return hash_table_get(priv
->glsl_program_lookup
, &key
);
3075 static void delete_glsl_program_entry(struct shader_glsl_priv
*priv
, const WineD3D_GL_Info
*gl_info
,
3076 struct glsl_shader_prog_link
*entry
)
3078 glsl_program_key_t
*key
;
3080 key
= HeapAlloc(GetProcessHeap(), 0, sizeof(glsl_program_key_t
));
3081 key
->vshader
= entry
->vshader
;
3082 key
->pshader
= entry
->pshader
;
3083 key
->vs_args
= entry
->vs_args
;
3084 key
->ps_args
= entry
->ps_args
;
3085 hash_table_remove(priv
->glsl_program_lookup
, key
);
3087 GL_EXTCALL(glDeleteObjectARB(entry
->programId
));
3088 if (entry
->vshader
) list_remove(&entry
->vshader_entry
);
3089 if (entry
->pshader
) list_remove(&entry
->pshader_entry
);
3090 HeapFree(GetProcessHeap(), 0, entry
->vuniformF_locations
);
3091 HeapFree(GetProcessHeap(), 0, entry
->puniformF_locations
);
3092 HeapFree(GetProcessHeap(), 0, entry
);
3095 static void handle_ps3_input(SHADER_BUFFER
*buffer
, const struct semantic
*semantics_in
,
3096 const struct semantic
*semantics_out
, const WineD3D_GL_Info
*gl_info
, const DWORD
*map
)
3099 DWORD usage_token
, usage_token_out
;
3100 DWORD register_token
, register_token_out
;
3101 DWORD usage
, usage_idx
, usage_out
, usage_idx_out
;
3104 DWORD in_count
= GL_LIMITS(glsl_varyings
) / 4;
3105 char reg_mask
[6], reg_mask_out
[6];
3106 char destination
[50];
3108 set
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*set
) * (in_count
+ 2));
3110 if (!semantics_out
) {
3111 /* Save gl_FrontColor & gl_FrontSecondaryColor before overwriting them. */
3112 shader_addline(buffer
, "vec4 front_color = gl_FrontColor;\n");
3113 shader_addline(buffer
, "vec4 front_secondary_color = gl_FrontSecondaryColor;\n");
3116 for(i
= 0; i
< MAX_REG_INPUT
; i
++) {
3117 usage_token
= semantics_in
[i
].usage
;
3118 if (!usage_token
) continue;
3121 if (in_idx
>= (in_count
+ 2)) {
3122 FIXME("More input varyings declared than supported, expect issues\n");
3125 else if (map
[i
] == ~0U)
3127 /* Declared, but not read register */
3131 if (in_idx
== in_count
) {
3132 sprintf(destination
, "gl_FrontColor");
3133 } else if (in_idx
== in_count
+ 1) {
3134 sprintf(destination
, "gl_FrontSecondaryColor");
3136 sprintf(destination
, "IN[%u]", in_idx
);
3139 register_token
= semantics_in
[i
].reg
;
3141 usage
= (usage_token
& WINED3DSP_DCL_USAGE_MASK
) >> WINED3DSP_DCL_USAGE_SHIFT
;
3142 usage_idx
= (usage_token
& WINED3DSP_DCL_USAGEINDEX_MASK
) >> WINED3DSP_DCL_USAGEINDEX_SHIFT
;
3143 set
[map
[i
]] = shader_glsl_get_write_mask(register_token
, reg_mask
);
3145 if(!semantics_out
) {
3147 case WINED3DDECLUSAGE_COLOR
:
3149 shader_addline(buffer
, "%s%s = front_color%s;\n",
3150 destination
, reg_mask
, reg_mask
);
3151 else if (usage_idx
== 1)
3152 shader_addline(buffer
, "%s%s = front_secondary_color%s;\n",
3153 destination
, reg_mask
, reg_mask
);
3155 shader_addline(buffer
, "%s%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
3156 destination
, reg_mask
, reg_mask
);
3159 case WINED3DDECLUSAGE_TEXCOORD
:
3160 if (usage_idx
< 8) {
3161 shader_addline(buffer
, "%s%s = gl_TexCoord[%u]%s;\n",
3162 destination
, reg_mask
, usage_idx
, reg_mask
);
3164 shader_addline(buffer
, "%s%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
3165 destination
, reg_mask
, reg_mask
);
3169 case WINED3DDECLUSAGE_FOG
:
3170 shader_addline(buffer
, "%s%s = vec4(gl_FogFragCoord, 0.0, 0.0, 0.0)%s;\n",
3171 destination
, reg_mask
, reg_mask
);
3175 shader_addline(buffer
, "%s%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
3176 destination
, reg_mask
, reg_mask
);
3180 for(j
= 0; j
< MAX_REG_OUTPUT
; j
++) {
3181 usage_token_out
= semantics_out
[j
].usage
;
3182 if (!usage_token_out
) continue;
3183 register_token_out
= semantics_out
[j
].reg
;
3185 usage_out
= (usage_token_out
& WINED3DSP_DCL_USAGE_MASK
) >> WINED3DSP_DCL_USAGE_SHIFT
;
3186 usage_idx_out
= (usage_token_out
& WINED3DSP_DCL_USAGEINDEX_MASK
) >> WINED3DSP_DCL_USAGEINDEX_SHIFT
;
3187 shader_glsl_get_write_mask(register_token_out
, reg_mask_out
);
3189 if(usage
== usage_out
&&
3190 usage_idx
== usage_idx_out
) {
3191 shader_addline(buffer
, "%s%s = OUT[%u]%s;\n",
3192 destination
, reg_mask
, j
, reg_mask
);
3197 shader_addline(buffer
, "%s%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
3198 destination
, reg_mask
, reg_mask
);
3203 /* This is solely to make the compiler / linker happy and avoid warning about undefined
3204 * varyings. It shouldn't result in any real code executed on the GPU, since all read
3205 * input varyings are assigned above, if the optimizer works properly.
3207 for(i
= 0; i
< in_count
+ 2; i
++) {
3208 if(set
[i
] != WINED3DSP_WRITEMASK_ALL
) {
3209 unsigned int size
= 0;
3210 memset(reg_mask
, 0, sizeof(reg_mask
));
3211 if(!(set
[i
] & WINED3DSP_WRITEMASK_0
)) {
3212 reg_mask
[size
] = 'x';
3215 if(!(set
[i
] & WINED3DSP_WRITEMASK_1
)) {
3216 reg_mask
[size
] = 'y';
3219 if(!(set
[i
] & WINED3DSP_WRITEMASK_2
)) {
3220 reg_mask
[size
] = 'z';
3223 if(!(set
[i
] & WINED3DSP_WRITEMASK_3
)) {
3224 reg_mask
[size
] = 'w';
3228 if (i
== in_count
) {
3229 sprintf(destination
, "gl_FrontColor");
3230 } else if (i
== in_count
+ 1) {
3231 sprintf(destination
, "gl_FrontSecondaryColor");
3233 sprintf(destination
, "IN[%u]", i
);
3237 shader_addline(buffer
, "%s.%s = 0.0;\n", destination
, reg_mask
);
3239 shader_addline(buffer
, "%s.%s = vec%u(0.0);\n", destination
, reg_mask
, size
);
3244 HeapFree(GetProcessHeap(), 0, set
);
3247 static GLhandleARB
generate_param_reorder_function(IWineD3DVertexShader
*vertexshader
,
3248 IWineD3DPixelShader
*pixelshader
, const WineD3D_GL_Info
*gl_info
)
3250 GLhandleARB ret
= 0;
3251 IWineD3DVertexShaderImpl
*vs
= (IWineD3DVertexShaderImpl
*) vertexshader
;
3252 IWineD3DPixelShaderImpl
*ps
= (IWineD3DPixelShaderImpl
*) pixelshader
;
3253 IWineD3DDeviceImpl
*device
;
3254 DWORD vs_major
= WINED3DSHADER_VERSION_MAJOR(vs
->baseShader
.reg_maps
.shader_version
);
3255 DWORD ps_major
= ps
? WINED3DSHADER_VERSION_MAJOR(ps
->baseShader
.reg_maps
.shader_version
) : 0;
3257 SHADER_BUFFER buffer
;
3259 DWORD register_token
;
3260 DWORD usage
, usage_idx
, writemask
;
3262 const struct semantic
*semantics_out
, *semantics_in
;
3264 shader_buffer_init(&buffer
);
3266 shader_addline(&buffer
, "#version 120\n");
3268 if(vs_major
< 3 && ps_major
< 3) {
3269 /* That one is easy: The vertex shader writes to the builtin varyings, the pixel shader reads from them.
3270 * Take care about the texcoord .w fixup though if we're using the fixed function fragment pipeline
3272 device
= (IWineD3DDeviceImpl
*) vs
->baseShader
.device
;
3273 if((GLINFO_LOCATION
).set_texcoord_w
&& ps_major
== 0 && vs_major
> 0 &&
3274 !device
->frag_pipe
->ffp_proj_control
) {
3275 shader_addline(&buffer
, "void order_ps_input() {\n");
3276 for(i
= 0; i
< min(8, MAX_REG_TEXCRD
); i
++) {
3277 if(vs
->baseShader
.reg_maps
.texcoord_mask
[i
] != 0 &&
3278 vs
->baseShader
.reg_maps
.texcoord_mask
[i
] != WINED3DSP_WRITEMASK_ALL
) {
3279 shader_addline(&buffer
, "gl_TexCoord[%u].w = 1.0;\n", i
);
3282 shader_addline(&buffer
, "}\n");
3284 shader_addline(&buffer
, "void order_ps_input() { /* do nothing */ }\n");
3286 } else if(ps_major
< 3 && vs_major
>= 3) {
3287 /* The vertex shader writes to its own varyings, the pixel shader needs them in the builtin ones */
3288 semantics_out
= vs
->semantics_out
;
3290 shader_addline(&buffer
, "void order_ps_input(in vec4 OUT[%u]) {\n", MAX_REG_OUTPUT
);
3291 for(i
= 0; i
< MAX_REG_OUTPUT
; i
++) {
3292 usage_token
= semantics_out
[i
].usage
;
3293 if (!usage_token
) continue;
3294 register_token
= semantics_out
[i
].reg
;
3296 usage
= (usage_token
& WINED3DSP_DCL_USAGE_MASK
) >> WINED3DSP_DCL_USAGE_SHIFT
;
3297 usage_idx
= (usage_token
& WINED3DSP_DCL_USAGEINDEX_MASK
) >> WINED3DSP_DCL_USAGEINDEX_SHIFT
;
3298 writemask
= shader_glsl_get_write_mask(register_token
, reg_mask
);
3301 case WINED3DDECLUSAGE_COLOR
:
3303 shader_addline(&buffer
, "gl_FrontColor%s = OUT[%u]%s;\n", reg_mask
, i
, reg_mask
);
3304 else if (usage_idx
== 1)
3305 shader_addline(&buffer
, "gl_FrontSecondaryColor%s = OUT[%u]%s;\n", reg_mask
, i
, reg_mask
);
3308 case WINED3DDECLUSAGE_POSITION
:
3309 shader_addline(&buffer
, "gl_Position%s = OUT[%u]%s;\n", reg_mask
, i
, reg_mask
);
3312 case WINED3DDECLUSAGE_TEXCOORD
:
3313 if (usage_idx
< 8) {
3314 if(!(GLINFO_LOCATION
).set_texcoord_w
|| ps_major
> 0) writemask
|= WINED3DSP_WRITEMASK_3
;
3316 shader_addline(&buffer
, "gl_TexCoord[%u]%s = OUT[%u]%s;\n",
3317 usage_idx
, reg_mask
, i
, reg_mask
);
3318 if(!(writemask
& WINED3DSP_WRITEMASK_3
)) {
3319 shader_addline(&buffer
, "gl_TexCoord[%u].w = 1.0;\n", usage_idx
);
3324 case WINED3DDECLUSAGE_PSIZE
:
3325 shader_addline(&buffer
, "gl_PointSize = OUT[%u].x;\n", i
);
3328 case WINED3DDECLUSAGE_FOG
:
3329 shader_addline(&buffer
, "gl_FogFragCoord = OUT[%u].%c;\n", i
, reg_mask
[1]);
3336 shader_addline(&buffer
, "}\n");
3338 } else if(ps_major
>= 3 && vs_major
>= 3) {
3339 semantics_out
= vs
->semantics_out
;
3340 semantics_in
= ps
->semantics_in
;
3342 /* This one is tricky: a 3.0 pixel shader reads from a 3.0 vertex shader */
3343 shader_addline(&buffer
, "varying vec4 IN[%u];\n", GL_LIMITS(glsl_varyings
) / 4);
3344 shader_addline(&buffer
, "void order_ps_input(in vec4 OUT[%u]) {\n", MAX_REG_OUTPUT
);
3346 /* First, sort out position and point size. Those are not passed to the pixel shader */
3347 for(i
= 0; i
< MAX_REG_OUTPUT
; i
++) {
3348 usage_token
= semantics_out
[i
].usage
;
3349 if (!usage_token
) continue;
3350 register_token
= semantics_out
[i
].reg
;
3352 usage
= (usage_token
& WINED3DSP_DCL_USAGE_MASK
) >> WINED3DSP_DCL_USAGE_SHIFT
;
3353 usage_idx
= (usage_token
& WINED3DSP_DCL_USAGEINDEX_MASK
) >> WINED3DSP_DCL_USAGEINDEX_SHIFT
;
3354 shader_glsl_get_write_mask(register_token
, reg_mask
);
3357 case WINED3DDECLUSAGE_POSITION
:
3358 shader_addline(&buffer
, "gl_Position%s = OUT[%u]%s;\n", reg_mask
, i
, reg_mask
);
3361 case WINED3DDECLUSAGE_PSIZE
:
3362 shader_addline(&buffer
, "gl_PointSize = OUT[%u].x;\n", i
);
3370 /* Then, fix the pixel shader input */
3371 handle_ps3_input(&buffer
, semantics_in
, semantics_out
, gl_info
, ps
->input_reg_map
);
3373 shader_addline(&buffer
, "}\n");
3374 } else if(ps_major
>= 3 && vs_major
< 3) {
3375 semantics_in
= ps
->semantics_in
;
3377 shader_addline(&buffer
, "varying vec4 IN[%u];\n", GL_LIMITS(glsl_varyings
) / 4);
3378 shader_addline(&buffer
, "void order_ps_input() {\n");
3379 /* The vertex shader wrote to the builtin varyings. There is no need to figure out position and
3380 * point size, but we depend on the optimizers kindness to find out that the pixel shader doesn't
3381 * read gl_TexCoord and gl_ColorX, otherwise we'll run out of varyings
3383 handle_ps3_input(&buffer
, semantics_in
, NULL
, gl_info
, ps
->input_reg_map
);
3384 shader_addline(&buffer
, "}\n");
3386 ERR("Unexpected vertex and pixel shader version condition: vs: %d, ps: %d\n", vs_major
, ps_major
);
3389 ret
= GL_EXTCALL(glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB
));
3390 checkGLcall("glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB)");
3391 GL_EXTCALL(glShaderSourceARB(ret
, 1, (const char**)&buffer
.buffer
, NULL
));
3392 checkGLcall("glShaderSourceARB(ret, 1, &buffer.buffer, NULL)");
3393 GL_EXTCALL(glCompileShaderARB(ret
));
3394 checkGLcall("glCompileShaderARB(ret)");
3396 shader_buffer_free(&buffer
);
3400 static void hardcode_local_constants(IWineD3DBaseShaderImpl
*shader
, const WineD3D_GL_Info
*gl_info
,
3401 GLhandleARB programId
, char prefix
)
3403 const local_constant
*lconst
;
3408 LIST_FOR_EACH_ENTRY(lconst
, &shader
->baseShader
.constantsF
, local_constant
, entry
) {
3409 value
= (const float *)lconst
->value
;
3410 snprintf(glsl_name
, sizeof(glsl_name
), "%cLC%u", prefix
, lconst
->idx
);
3411 tmp_loc
= GL_EXTCALL(glGetUniformLocationARB(programId
, glsl_name
));
3412 GL_EXTCALL(glUniform4fvARB(tmp_loc
, 1, value
));
3414 checkGLcall("Hardcoding local constants\n");
3417 /** Sets the GLSL program ID for the given pixel and vertex shader combination.
3418 * It sets the programId on the current StateBlock (because it should be called
3419 * inside of the DrawPrimitive() part of the render loop).
3421 * If a program for the given combination does not exist, create one, and store
3422 * the program in the hash table. If it creates a program, it will link the
3423 * given objects, too.
3425 static void set_glsl_shader_program(IWineD3DDevice
*iface
, BOOL use_ps
, BOOL use_vs
) {
3426 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
3427 struct shader_glsl_priv
*priv
= This
->shader_priv
;
3428 const WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
3429 IWineD3DPixelShader
*pshader
= This
->stateBlock
->pixelShader
;
3430 IWineD3DVertexShader
*vshader
= This
->stateBlock
->vertexShader
;
3431 struct glsl_shader_prog_link
*entry
= NULL
;
3432 GLhandleARB programId
= 0;
3433 GLhandleARB reorder_shader_id
= 0;
3436 GLhandleARB vshader_id
, pshader_id
;
3437 struct ps_compile_args ps_compile_args
;
3438 struct vs_compile_args vs_compile_args
;
3441 find_vs_compile_args((IWineD3DVertexShaderImpl
*)This
->stateBlock
->vertexShader
, This
->stateBlock
, &vs_compile_args
);
3443 /* FIXME: Do we really have to spend CPU cycles to generate a few zeroed bytes? */
3444 memset(&vs_compile_args
, 0, sizeof(vs_compile_args
));
3447 find_ps_compile_args((IWineD3DPixelShaderImpl
*)This
->stateBlock
->pixelShader
, This
->stateBlock
, &ps_compile_args
);
3449 /* FIXME: Do we really have to spend CPU cycles to generate a few zeroed bytes? */
3450 memset(&ps_compile_args
, 0, sizeof(ps_compile_args
));
3452 entry
= get_glsl_program_entry(priv
, vshader
, pshader
, &vs_compile_args
, &ps_compile_args
);
3454 priv
->glsl_program
= entry
;
3458 /* If we get to this point, then no matching program exists, so we create one */
3459 programId
= GL_EXTCALL(glCreateProgramObjectARB());
3460 TRACE("Created new GLSL shader program %u\n", programId
);
3462 /* Create the entry */
3463 entry
= HeapAlloc(GetProcessHeap(), 0, sizeof(struct glsl_shader_prog_link
));
3464 entry
->programId
= programId
;
3465 entry
->vshader
= vshader
;
3466 entry
->pshader
= pshader
;
3467 entry
->vs_args
= vs_compile_args
;
3468 entry
->ps_args
= ps_compile_args
;
3469 entry
->constant_version
= 0;
3470 /* Add the hash table entry */
3471 add_glsl_program_entry(priv
, entry
);
3473 /* Set the current program */
3474 priv
->glsl_program
= entry
;
3477 vshader_id
= find_gl_vshader((IWineD3DVertexShaderImpl
*) vshader
, &vs_compile_args
);
3482 /* Attach GLSL vshader */
3484 const unsigned int max_attribs
= 16; /* TODO: Will this always be the case? It is at the moment... */
3487 reorder_shader_id
= generate_param_reorder_function(vshader
, pshader
, gl_info
);
3488 TRACE("Attaching GLSL shader object %u to program %u\n", reorder_shader_id
, programId
);
3489 GL_EXTCALL(glAttachObjectARB(programId
, reorder_shader_id
));
3490 checkGLcall("glAttachObjectARB");
3491 /* Flag the reorder function for deletion, then it will be freed automatically when the program
3494 GL_EXTCALL(glDeleteObjectARB(reorder_shader_id
));
3496 TRACE("Attaching GLSL shader object %u to program %u\n", vshader_id
, programId
);
3497 GL_EXTCALL(glAttachObjectARB(programId
, vshader_id
));
3498 checkGLcall("glAttachObjectARB");
3500 /* Bind vertex attributes to a corresponding index number to match
3501 * the same index numbers as ARB_vertex_programs (makes loading
3502 * vertex attributes simpler). With this method, we can use the
3503 * exact same code to load the attributes later for both ARB and
3506 * We have to do this here because we need to know the Program ID
3507 * in order to make the bindings work, and it has to be done prior
3508 * to linking the GLSL program. */
3509 for (i
= 0; i
< max_attribs
; ++i
) {
3510 if (((IWineD3DBaseShaderImpl
*)vshader
)->baseShader
.reg_maps
.attributes
[i
]) {
3511 snprintf(tmp_name
, sizeof(tmp_name
), "attrib%i", i
);
3512 GL_EXTCALL(glBindAttribLocationARB(programId
, i
, tmp_name
));
3515 checkGLcall("glBindAttribLocationARB");
3517 list_add_head(&((IWineD3DBaseShaderImpl
*)vshader
)->baseShader
.linked_programs
, &entry
->vshader_entry
);
3521 pshader_id
= find_gl_pshader((IWineD3DPixelShaderImpl
*) pshader
, &ps_compile_args
);
3526 /* Attach GLSL pshader */
3528 TRACE("Attaching GLSL shader object %u to program %u\n", pshader_id
, programId
);
3529 GL_EXTCALL(glAttachObjectARB(programId
, pshader_id
));
3530 checkGLcall("glAttachObjectARB");
3532 list_add_head(&((IWineD3DBaseShaderImpl
*)pshader
)->baseShader
.linked_programs
, &entry
->pshader_entry
);
3535 /* Link the program */
3536 TRACE("Linking GLSL shader program %u\n", programId
);
3537 GL_EXTCALL(glLinkProgramARB(programId
));
3538 print_glsl_info_log(&GLINFO_LOCATION
, programId
);
3540 entry
->vuniformF_locations
= HeapAlloc(GetProcessHeap(), 0, sizeof(GLhandleARB
) * GL_LIMITS(vshader_constantsF
));
3541 for (i
= 0; i
< GL_LIMITS(vshader_constantsF
); ++i
) {
3542 snprintf(glsl_name
, sizeof(glsl_name
), "VC[%i]", i
);
3543 entry
->vuniformF_locations
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, glsl_name
));
3545 for (i
= 0; i
< MAX_CONST_I
; ++i
) {
3546 snprintf(glsl_name
, sizeof(glsl_name
), "VI[%i]", i
);
3547 entry
->vuniformI_locations
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, glsl_name
));
3549 entry
->puniformF_locations
= HeapAlloc(GetProcessHeap(), 0, sizeof(GLhandleARB
) * GL_LIMITS(pshader_constantsF
));
3550 for (i
= 0; i
< GL_LIMITS(pshader_constantsF
); ++i
) {
3551 snprintf(glsl_name
, sizeof(glsl_name
), "PC[%i]", i
);
3552 entry
->puniformF_locations
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, glsl_name
));
3554 for (i
= 0; i
< MAX_CONST_I
; ++i
) {
3555 snprintf(glsl_name
, sizeof(glsl_name
), "PI[%i]", i
);
3556 entry
->puniformI_locations
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, glsl_name
));
3560 for(i
= 0; i
< ((IWineD3DPixelShaderImpl
*)pshader
)->numbumpenvmatconsts
; i
++) {
3562 sprintf(name
, "bumpenvmat%d", ((IWineD3DPixelShaderImpl
*)pshader
)->bumpenvmatconst
[i
].texunit
);
3563 entry
->bumpenvmat_location
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, name
));
3564 sprintf(name
, "luminancescale%d", ((IWineD3DPixelShaderImpl
*)pshader
)->luminanceconst
[i
].texunit
);
3565 entry
->luminancescale_location
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, name
));
3566 sprintf(name
, "luminanceoffset%d", ((IWineD3DPixelShaderImpl
*)pshader
)->luminanceconst
[i
].texunit
);
3567 entry
->luminanceoffset_location
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, name
));
3571 if (use_ps
&& ps_compile_args
.texrect_fixup
) {
3573 for (i
= 0; i
< MAX_FRAGMENT_SAMPLERS
; ++i
) {
3574 if (ps_compile_args
.texrect_fixup
& (1 << i
)) {
3575 sprintf(name
, "PsamplerRectFixup%u", i
);
3576 entry
->rectFixup_location
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, name
));
3578 entry
->rectFixup_location
[i
] = -1;
3583 entry
->posFixup_location
= GL_EXTCALL(glGetUniformLocationARB(programId
, "posFixup"));
3584 entry
->ycorrection_location
= GL_EXTCALL(glGetUniformLocationARB(programId
, "ycorrection"));
3585 checkGLcall("Find glsl program uniform locations");
3588 && WINED3DSHADER_VERSION_MAJOR(((IWineD3DPixelShaderImpl
*)pshader
)->baseShader
.reg_maps
.shader_version
) >= 3
3589 && ((IWineD3DPixelShaderImpl
*)pshader
)->declared_in_count
> GL_LIMITS(glsl_varyings
) / 4)
3591 TRACE("Shader %d needs vertex color clamping disabled\n", programId
);
3592 entry
->vertex_color_clamp
= GL_FALSE
;
3594 entry
->vertex_color_clamp
= GL_FIXED_ONLY_ARB
;
3597 /* Set the shader to allow uniform loading on it */
3598 GL_EXTCALL(glUseProgramObjectARB(programId
));
3599 checkGLcall("glUseProgramObjectARB(programId)");
3601 /* Load the vertex and pixel samplers now. The function that finds the mappings makes sure
3602 * that it stays the same for each vertexshader-pixelshader pair(=linked glsl program). If
3603 * a pshader with fixed function pipeline is used there are no vertex samplers, and if a
3604 * vertex shader with fixed function pixel processing is used we make sure that the card
3605 * supports enough samplers to allow the max number of vertex samplers with all possible
3606 * fixed function fragment processing setups. So once the program is linked these samplers
3610 /* Load vertex shader samplers */
3611 shader_glsl_load_vsamplers(gl_info
, This
->texUnitMap
, programId
);
3614 /* Load pixel shader samplers */
3615 shader_glsl_load_psamplers(gl_info
, This
->texUnitMap
, programId
);
3618 /* If the local constants do not have to be loaded with the environment constants,
3619 * load them now to have them hardcoded in the GLSL program. This saves some CPU cycles
3622 if(pshader
&& !((IWineD3DPixelShaderImpl
*)pshader
)->baseShader
.load_local_constsF
) {
3623 hardcode_local_constants((IWineD3DBaseShaderImpl
*) pshader
, gl_info
, programId
, 'P');
3625 if(vshader
&& !((IWineD3DVertexShaderImpl
*)vshader
)->baseShader
.load_local_constsF
) {
3626 hardcode_local_constants((IWineD3DBaseShaderImpl
*) vshader
, gl_info
, programId
, 'V');
3630 static GLhandleARB
create_glsl_blt_shader(const WineD3D_GL_Info
*gl_info
, enum tex_types tex_type
)
3632 GLhandleARB program_id
;
3633 GLhandleARB vshader_id
, pshader_id
;
3634 static const char *blt_vshader
[] =
3639 " gl_Position = gl_Vertex;\n"
3640 " gl_FrontColor = vec4(1.0);\n"
3641 " gl_TexCoord[0] = gl_MultiTexCoord0;\n"
3645 static const char *blt_pshaders
[tex_type_count
] =
3651 "uniform sampler2D sampler;\n"
3654 " gl_FragDepth = texture2D(sampler, gl_TexCoord[0].xy).x;\n"
3660 "uniform samplerCube sampler;\n"
3663 " gl_FragDepth = textureCube(sampler, gl_TexCoord[0].xyz).x;\n"
3667 "#extension GL_ARB_texture_rectangle : enable\n"
3668 "uniform sampler2DRect sampler;\n"
3671 " gl_FragDepth = texture2DRect(sampler, gl_TexCoord[0].xy).x;\n"
3675 if (!blt_pshaders
[tex_type
])
3677 FIXME("tex_type %#x not supported\n", tex_type
);
3681 vshader_id
= GL_EXTCALL(glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB
));
3682 GL_EXTCALL(glShaderSourceARB(vshader_id
, 1, blt_vshader
, NULL
));
3683 GL_EXTCALL(glCompileShaderARB(vshader_id
));
3685 pshader_id
= GL_EXTCALL(glCreateShaderObjectARB(GL_FRAGMENT_SHADER_ARB
));
3686 GL_EXTCALL(glShaderSourceARB(pshader_id
, 1, &blt_pshaders
[tex_type
], NULL
));
3687 GL_EXTCALL(glCompileShaderARB(pshader_id
));
3689 program_id
= GL_EXTCALL(glCreateProgramObjectARB());
3690 GL_EXTCALL(glAttachObjectARB(program_id
, vshader_id
));
3691 GL_EXTCALL(glAttachObjectARB(program_id
, pshader_id
));
3692 GL_EXTCALL(glLinkProgramARB(program_id
));
3694 print_glsl_info_log(&GLINFO_LOCATION
, program_id
);
3696 /* Once linked we can mark the shaders for deletion. They will be deleted once the program
3699 GL_EXTCALL(glDeleteObjectARB(vshader_id
));
3700 GL_EXTCALL(glDeleteObjectARB(pshader_id
));
3704 static void shader_glsl_select(IWineD3DDevice
*iface
, BOOL usePS
, BOOL useVS
) {
3705 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
3706 struct shader_glsl_priv
*priv
= This
->shader_priv
;
3707 const WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
3708 GLhandleARB program_id
= 0;
3709 GLenum old_vertex_color_clamp
, current_vertex_color_clamp
;
3711 old_vertex_color_clamp
= priv
->glsl_program
? priv
->glsl_program
->vertex_color_clamp
: GL_FIXED_ONLY_ARB
;
3713 if (useVS
|| usePS
) set_glsl_shader_program(iface
, usePS
, useVS
);
3714 else priv
->glsl_program
= NULL
;
3716 current_vertex_color_clamp
= priv
->glsl_program
? priv
->glsl_program
->vertex_color_clamp
: GL_FIXED_ONLY_ARB
;
3718 if (old_vertex_color_clamp
!= current_vertex_color_clamp
) {
3719 if (GL_SUPPORT(ARB_COLOR_BUFFER_FLOAT
)) {
3720 GL_EXTCALL(glClampColorARB(GL_CLAMP_VERTEX_COLOR_ARB
, current_vertex_color_clamp
));
3721 checkGLcall("glClampColorARB");
3723 FIXME("vertex color clamp needs to be changed, but extension not supported.\n");
3727 program_id
= priv
->glsl_program
? priv
->glsl_program
->programId
: 0;
3728 if (program_id
) TRACE("Using GLSL program %u\n", program_id
);
3729 GL_EXTCALL(glUseProgramObjectARB(program_id
));
3730 checkGLcall("glUseProgramObjectARB");
3733 static void shader_glsl_select_depth_blt(IWineD3DDevice
*iface
, enum tex_types tex_type
) {
3734 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
3735 const WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
3736 struct shader_glsl_priv
*priv
= This
->shader_priv
;
3737 GLhandleARB
*blt_program
= &priv
->depth_blt_program
[tex_type
];
3739 if (!*blt_program
) {
3741 *blt_program
= create_glsl_blt_shader(gl_info
, tex_type
);
3742 loc
= GL_EXTCALL(glGetUniformLocationARB(*blt_program
, "sampler"));
3743 GL_EXTCALL(glUseProgramObjectARB(*blt_program
));
3744 GL_EXTCALL(glUniform1iARB(loc
, 0));
3746 GL_EXTCALL(glUseProgramObjectARB(*blt_program
));
3750 static void shader_glsl_deselect_depth_blt(IWineD3DDevice
*iface
) {
3751 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
3752 const WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
3753 struct shader_glsl_priv
*priv
= This
->shader_priv
;
3754 GLhandleARB program_id
;
3756 program_id
= priv
->glsl_program
? priv
->glsl_program
->programId
: 0;
3757 if (program_id
) TRACE("Using GLSL program %u\n", program_id
);
3759 GL_EXTCALL(glUseProgramObjectARB(program_id
));
3760 checkGLcall("glUseProgramObjectARB");
3763 static void shader_glsl_destroy(IWineD3DBaseShader
*iface
) {
3764 const struct list
*linked_programs
;
3765 IWineD3DBaseShaderImpl
*This
= (IWineD3DBaseShaderImpl
*) iface
;
3766 IWineD3DDeviceImpl
*device
= (IWineD3DDeviceImpl
*)This
->baseShader
.device
;
3767 struct shader_glsl_priv
*priv
= device
->shader_priv
;
3768 const WineD3D_GL_Info
*gl_info
= &device
->adapter
->gl_info
;
3769 IWineD3DPixelShaderImpl
*ps
= NULL
;
3770 IWineD3DVertexShaderImpl
*vs
= NULL
;
3772 /* Note: Do not use QueryInterface here to find out which shader type this is because this code
3773 * can be called from IWineD3DBaseShader::Release
3775 char pshader
= shader_is_pshader_version(This
->baseShader
.reg_maps
.shader_version
);
3778 ps
= (IWineD3DPixelShaderImpl
*) This
;
3779 if(ps
->num_gl_shaders
== 0) return;
3780 if (priv
->glsl_program
&& (IWineD3DBaseShader
*)priv
->glsl_program
->pshader
== iface
)
3781 shader_glsl_select(This
->baseShader
.device
, FALSE
, FALSE
);
3783 vs
= (IWineD3DVertexShaderImpl
*) This
;
3784 if(vs
->num_gl_shaders
== 0) return;
3785 if (priv
->glsl_program
&& (IWineD3DBaseShader
*)priv
->glsl_program
->vshader
== iface
)
3786 shader_glsl_select(This
->baseShader
.device
, FALSE
, FALSE
);
3789 linked_programs
= &This
->baseShader
.linked_programs
;
3791 TRACE("Deleting linked programs\n");
3792 if (linked_programs
->next
) {
3793 struct glsl_shader_prog_link
*entry
, *entry2
;
3796 LIST_FOR_EACH_ENTRY_SAFE(entry
, entry2
, linked_programs
, struct glsl_shader_prog_link
, pshader_entry
) {
3797 delete_glsl_program_entry(priv
, gl_info
, entry
);
3800 LIST_FOR_EACH_ENTRY_SAFE(entry
, entry2
, linked_programs
, struct glsl_shader_prog_link
, vshader_entry
) {
3801 delete_glsl_program_entry(priv
, gl_info
, entry
);
3810 for(i
= 0; i
< ps
->num_gl_shaders
; i
++) {
3811 TRACE("deleting pshader %u\n", ps
->gl_shaders
[i
].prgId
);
3812 GL_EXTCALL(glDeleteObjectARB(ps
->gl_shaders
[i
].prgId
));
3813 checkGLcall("glDeleteObjectARB");
3816 HeapFree(GetProcessHeap(), 0, ps
->gl_shaders
);
3817 ps
->gl_shaders
= NULL
;
3818 ps
->num_gl_shaders
= 0;
3819 ps
->shader_array_size
= 0;
3824 for(i
= 0; i
< vs
->num_gl_shaders
; i
++) {
3825 TRACE("deleting vshader %u\n", vs
->gl_shaders
[i
].prgId
);
3826 GL_EXTCALL(glDeleteObjectARB(vs
->gl_shaders
[i
].prgId
));
3827 checkGLcall("glDeleteObjectARB");
3830 HeapFree(GetProcessHeap(), 0, vs
->gl_shaders
);
3831 vs
->gl_shaders
= NULL
;
3832 vs
->num_gl_shaders
= 0;
3833 vs
->shader_array_size
= 0;
3837 static unsigned int glsl_program_key_hash(const void *key
)
3839 const glsl_program_key_t
*k
= key
;
3841 unsigned int hash
= ((DWORD_PTR
) k
->vshader
) | ((DWORD_PTR
) k
->pshader
) << 16;
3842 hash
+= ~(hash
<< 15);
3843 hash
^= (hash
>> 10);
3844 hash
+= (hash
<< 3);
3845 hash
^= (hash
>> 6);
3846 hash
+= ~(hash
<< 11);
3847 hash
^= (hash
>> 16);
3852 static BOOL
glsl_program_key_compare(const void *keya
, const void *keyb
)
3854 const glsl_program_key_t
*ka
= keya
;
3855 const glsl_program_key_t
*kb
= keyb
;
3857 return ka
->vshader
== kb
->vshader
&& ka
->pshader
== kb
->pshader
&&
3858 (memcmp(&ka
->ps_args
, &kb
->ps_args
, sizeof(kb
->ps_args
)) == 0) &&
3859 (memcmp(&ka
->vs_args
, &kb
->vs_args
, sizeof(kb
->vs_args
)) == 0);
3862 static BOOL
constant_heap_init(struct constant_heap
*heap
, unsigned int constant_count
)
3864 SIZE_T size
= (constant_count
+ 1) * sizeof(*heap
->entries
) + constant_count
* sizeof(*heap
->positions
);
3865 void *mem
= HeapAlloc(GetProcessHeap(), 0, size
);
3869 ERR("Failed to allocate memory\n");
3873 heap
->entries
= mem
;
3874 heap
->entries
[1].version
= 0;
3875 heap
->positions
= (unsigned int *)(heap
->entries
+ constant_count
+ 1);
3881 static void constant_heap_free(struct constant_heap
*heap
)
3883 HeapFree(GetProcessHeap(), 0, heap
->entries
);
3886 static HRESULT
shader_glsl_alloc(IWineD3DDevice
*iface
) {
3887 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
3888 const WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
3889 struct shader_glsl_priv
*priv
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(struct shader_glsl_priv
));
3890 SIZE_T stack_size
= wined3d_log2i(max(GL_LIMITS(vshader_constantsF
), GL_LIMITS(pshader_constantsF
))) + 1;
3892 priv
->stack
= HeapAlloc(GetProcessHeap(), 0, stack_size
* sizeof(*priv
->stack
));
3895 ERR("Failed to allocate memory.\n");
3896 HeapFree(GetProcessHeap(), 0, priv
);
3897 return E_OUTOFMEMORY
;
3900 if (!constant_heap_init(&priv
->vconst_heap
, GL_LIMITS(vshader_constantsF
)))
3902 ERR("Failed to initialize vertex shader constant heap\n");
3903 HeapFree(GetProcessHeap(), 0, priv
->stack
);
3904 HeapFree(GetProcessHeap(), 0, priv
);
3905 return E_OUTOFMEMORY
;
3908 if (!constant_heap_init(&priv
->pconst_heap
, GL_LIMITS(pshader_constantsF
)))
3910 ERR("Failed to initialize pixel shader constant heap\n");
3911 constant_heap_free(&priv
->vconst_heap
);
3912 HeapFree(GetProcessHeap(), 0, priv
->stack
);
3913 HeapFree(GetProcessHeap(), 0, priv
);
3914 return E_OUTOFMEMORY
;
3917 priv
->glsl_program_lookup
= hash_table_create(glsl_program_key_hash
, glsl_program_key_compare
);
3918 priv
->next_constant_version
= 1;
3920 This
->shader_priv
= priv
;
3924 static void shader_glsl_free(IWineD3DDevice
*iface
) {
3925 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
3926 const WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
3927 struct shader_glsl_priv
*priv
= This
->shader_priv
;
3930 for (i
= 0; i
< tex_type_count
; ++i
)
3932 if (priv
->depth_blt_program
[i
])
3934 GL_EXTCALL(glDeleteObjectARB(priv
->depth_blt_program
[i
]));
3938 hash_table_destroy(priv
->glsl_program_lookup
, NULL
, NULL
);
3939 constant_heap_free(&priv
->pconst_heap
);
3940 constant_heap_free(&priv
->vconst_heap
);
3942 HeapFree(GetProcessHeap(), 0, This
->shader_priv
);
3943 This
->shader_priv
= NULL
;
3946 static BOOL
shader_glsl_dirty_const(IWineD3DDevice
*iface
) {
3947 /* TODO: GL_EXT_bindable_uniform can be used to share constants across shaders */
3951 static GLuint
shader_glsl_generate_pshader(IWineD3DPixelShader
*iface
, SHADER_BUFFER
*buffer
, const struct ps_compile_args
*args
) {
3952 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*)iface
;
3953 const struct shader_reg_maps
*reg_maps
= &This
->baseShader
.reg_maps
;
3954 CONST DWORD
*function
= This
->baseShader
.function
;
3955 const char *fragcolor
;
3956 const WineD3D_GL_Info
*gl_info
= &((IWineD3DDeviceImpl
*)This
->baseShader
.device
)->adapter
->gl_info
;
3958 /* Create the hw GLSL shader object and assign it as the shader->prgId */
3959 GLhandleARB shader_obj
= GL_EXTCALL(glCreateShaderObjectARB(GL_FRAGMENT_SHADER_ARB
));
3961 shader_addline(buffer
, "#version 120\n");
3963 if (GL_SUPPORT(ARB_DRAW_BUFFERS
)) {
3964 shader_addline(buffer
, "#extension GL_ARB_draw_buffers : enable\n");
3966 if (GL_SUPPORT(ARB_TEXTURE_RECTANGLE
)) {
3967 /* The spec says that it doesn't have to be explicitly enabled, but the nvidia
3968 * drivers write a warning if we don't do so
3970 shader_addline(buffer
, "#extension GL_ARB_texture_rectangle : enable\n");
3973 /* Base Declarations */
3974 shader_generate_glsl_declarations( (IWineD3DBaseShader
*) This
, reg_maps
, buffer
, &GLINFO_LOCATION
, args
);
3976 /* Pack 3.0 inputs */
3977 if (reg_maps
->shader_version
>= WINED3DPS_VERSION(3,0) && args
->vp_mode
!= vertexshader
) {
3978 pshader_glsl_input_pack(buffer
, This
->semantics_in
, iface
, args
->vp_mode
);
3981 /* Base Shader Body */
3982 shader_generate_main( (IWineD3DBaseShader
*) This
, buffer
, reg_maps
, function
);
3984 /* Pixel shaders < 2.0 place the resulting color in R0 implicitly */
3985 if (reg_maps
->shader_version
< WINED3DPS_VERSION(2,0))
3987 /* Some older cards like GeforceFX ones don't support multiple buffers, so also not gl_FragData */
3988 if(GL_SUPPORT(ARB_DRAW_BUFFERS
))
3989 shader_addline(buffer
, "gl_FragData[0] = R0;\n");
3991 shader_addline(buffer
, "gl_FragColor = R0;\n");
3994 if(GL_SUPPORT(ARB_DRAW_BUFFERS
)) {
3995 fragcolor
= "gl_FragData[0]";
3997 fragcolor
= "gl_FragColor";
3999 if(args
->srgb_correction
) {
4000 shader_addline(buffer
, "tmp0.xyz = pow(%s.xyz, vec3(%f, %f, %f)) * vec3(%f, %f, %f) - vec3(%f, %f, %f);\n",
4001 fragcolor
, srgb_pow
, srgb_pow
, srgb_pow
, srgb_mul_high
, srgb_mul_high
, srgb_mul_high
,
4002 srgb_sub_high
, srgb_sub_high
, srgb_sub_high
);
4003 shader_addline(buffer
, "tmp1.xyz = %s.xyz * srgb_mul_low.xyz;\n", fragcolor
);
4004 shader_addline(buffer
, "%s.x = %s.x < srgb_comparison.x ? tmp1.x : tmp0.x;\n", fragcolor
, fragcolor
);
4005 shader_addline(buffer
, "%s.y = %s.y < srgb_comparison.y ? tmp1.y : tmp0.y;\n", fragcolor
, fragcolor
);
4006 shader_addline(buffer
, "%s.z = %s.z < srgb_comparison.z ? tmp1.z : tmp0.z;\n", fragcolor
, fragcolor
);
4007 shader_addline(buffer
, "%s = clamp(%s, 0.0, 1.0);\n", fragcolor
, fragcolor
);
4009 /* Pixel shader < 3.0 do not replace the fog stage.
4010 * This implements linear fog computation and blending.
4011 * TODO: non linear fog
4012 * NOTE: gl_Fog.start and gl_Fog.end don't hold fog start s and end e but
4013 * -1/(e-s) and e/(e-s) respectively.
4015 if(reg_maps
->shader_version
< WINED3DPS_VERSION(3,0)) {
4017 case FOG_OFF
: break;
4019 shader_addline(buffer
, "float fogstart = -1.0 / (gl_Fog.end - gl_Fog.start);\n");
4020 shader_addline(buffer
, "float fogend = gl_Fog.end * -fogstart;\n");
4021 shader_addline(buffer
, "float Fog = clamp(gl_FogFragCoord * fogstart + fogend, 0.0, 1.0);\n");
4022 shader_addline(buffer
, "%s.xyz = mix(gl_Fog.color.xyz, %s.xyz, Fog);\n", fragcolor
, fragcolor
);
4025 /* Fog = e^(-gl_Fog.density * gl_FogFragCoord) */
4026 shader_addline(buffer
, "float Fog = exp(-gl_Fog.density * gl_FogFragCoord);\n");
4027 shader_addline(buffer
, "Fog = clamp(Fog, 0.0, 1.0);\n");
4028 shader_addline(buffer
, "%s.xyz = mix(gl_Fog.color.xyz, %s.xyz, Fog);\n", fragcolor
, fragcolor
);
4031 /* Fog = e^(-(gl_Fog.density * gl_FogFragCoord)^2) */
4032 shader_addline(buffer
, "float Fog = exp(-gl_Fog.density * gl_Fog.density * gl_FogFragCoord * gl_FogFragCoord);\n");
4033 shader_addline(buffer
, "Fog = clamp(Fog, 0.0, 1.0);\n");
4034 shader_addline(buffer
, "%s.xyz = mix(gl_Fog.color.xyz, %s.xyz, Fog);\n", fragcolor
, fragcolor
);
4039 shader_addline(buffer
, "}\n");
4041 TRACE("Compiling shader object %u\n", shader_obj
);
4042 GL_EXTCALL(glShaderSourceARB(shader_obj
, 1, (const char**)&buffer
->buffer
, NULL
));
4043 GL_EXTCALL(glCompileShaderARB(shader_obj
));
4044 print_glsl_info_log(&GLINFO_LOCATION
, shader_obj
);
4046 /* Store the shader object */
4050 static GLuint
shader_glsl_generate_vshader(IWineD3DVertexShader
*iface
, SHADER_BUFFER
*buffer
, const struct vs_compile_args
*args
) {
4051 IWineD3DVertexShaderImpl
*This
= (IWineD3DVertexShaderImpl
*)iface
;
4052 const struct shader_reg_maps
*reg_maps
= &This
->baseShader
.reg_maps
;
4053 CONST DWORD
*function
= This
->baseShader
.function
;
4054 const WineD3D_GL_Info
*gl_info
= &((IWineD3DDeviceImpl
*)This
->baseShader
.device
)->adapter
->gl_info
;
4056 /* Create the hw GLSL shader program and assign it as the shader->prgId */
4057 GLhandleARB shader_obj
= GL_EXTCALL(glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB
));
4059 shader_addline(buffer
, "#version 120\n");
4061 /* Base Declarations */
4062 shader_generate_glsl_declarations( (IWineD3DBaseShader
*) This
, reg_maps
, buffer
, &GLINFO_LOCATION
, NULL
);
4064 /* Base Shader Body */
4065 shader_generate_main( (IWineD3DBaseShader
*) This
, buffer
, reg_maps
, function
);
4067 /* Unpack 3.0 outputs */
4068 if (reg_maps
->shader_version
>= WINED3DVS_VERSION(3,0)) shader_addline(buffer
, "order_ps_input(OUT);\n");
4069 else shader_addline(buffer
, "order_ps_input();\n");
4071 /* The D3DRS_FOGTABLEMODE render state defines if the shader-generated fog coord is used
4072 * or if the fragment depth is used. If the fragment depth is used(FOGTABLEMODE != NONE),
4073 * the fog frag coord is thrown away. If the fog frag coord is used, but not written by
4074 * the shader, it is set to 0.0(fully fogged, since start = 1.0, end = 0.0)
4076 if(args
->fog_src
== VS_FOG_Z
) {
4077 shader_addline(buffer
, "gl_FogFragCoord = gl_Position.z;\n");
4078 } else if (!reg_maps
->fog
) {
4079 shader_addline(buffer
, "gl_FogFragCoord = 0.0;\n");
4082 /* Write the final position.
4084 * OpenGL coordinates specify the center of the pixel while d3d coords specify
4085 * the corner. The offsets are stored in z and w in posFixup. posFixup.y contains
4086 * 1.0 or -1.0 to turn the rendering upside down for offscreen rendering. PosFixup.x
4087 * contains 1.0 to allow a mad.
4089 shader_addline(buffer
, "gl_Position.y = gl_Position.y * posFixup.y;\n");
4090 shader_addline(buffer
, "gl_Position.xy += posFixup.zw * gl_Position.ww;\n");
4092 /* Z coord [0;1]->[-1;1] mapping, see comment in transform_projection in state.c
4094 * Basically we want (in homogeneous coordinates) z = z * 2 - 1. However, shaders are run
4095 * before the homogeneous divide, so we have to take the w into account: z = ((z / w) * 2 - 1) * w,
4096 * which is the same as z = z * 2 - w.
4098 shader_addline(buffer
, "gl_Position.z = gl_Position.z * 2.0 - gl_Position.w;\n");
4100 shader_addline(buffer
, "}\n");
4102 TRACE("Compiling shader object %u\n", shader_obj
);
4103 GL_EXTCALL(glShaderSourceARB(shader_obj
, 1, (const char**)&buffer
->buffer
, NULL
));
4104 GL_EXTCALL(glCompileShaderARB(shader_obj
));
4105 print_glsl_info_log(&GLINFO_LOCATION
, shader_obj
);
4110 static void shader_glsl_get_caps(WINED3DDEVTYPE devtype
, const WineD3D_GL_Info
*gl_info
, struct shader_caps
*pCaps
)
4112 /* Nvidia Geforce6/7 or Ati R4xx/R5xx cards with GLSL support, support VS 3.0 but older Nvidia/Ati
4113 * models with GLSL support only support 2.0. In case of nvidia we can detect VS 2.0 support using
4114 * vs_nv_version which is based on NV_vertex_program.
4115 * For Ati cards there's no way using glsl (it abstracts the lowlevel info away) and also not
4116 * using ARB_vertex_program. It is safe to assume that when a card supports pixel shader 2.0 it
4117 * supports vertex shader 2.0 too and the way around. We can detect ps2.0 using the maximum number
4118 * of native instructions, so use that here. For more info see the pixel shader versioning code below.
4120 if((GLINFO_LOCATION
.vs_nv_version
== VS_VERSION_20
) || (GLINFO_LOCATION
.ps_arb_max_instructions
<= 512))
4121 pCaps
->VertexShaderVersion
= WINED3DVS_VERSION(2,0);
4123 pCaps
->VertexShaderVersion
= WINED3DVS_VERSION(3,0);
4124 TRACE_(d3d_caps
)("Hardware vertex shader version %d.%d enabled (GLSL)\n", (pCaps
->VertexShaderVersion
>> 8) & 0xff, pCaps
->VertexShaderVersion
& 0xff);
4125 pCaps
->MaxVertexShaderConst
= GL_LIMITS(vshader_constantsF
);
4127 /* Older DX9-class videocards (GeforceFX / Radeon >9500/X*00) only support pixel shader 2.0/2.0a/2.0b.
4128 * In OpenGL the extensions related to GLSL abstract lowlevel GL info away which is needed
4129 * to distinguish between 2.0 and 3.0 (and 2.0a/2.0b). In case of Nvidia we use their fragment
4130 * program extensions. On other hardware including ATI GL_ARB_fragment_program offers the info
4131 * in max native instructions. Intel and others also offer the info in this extension but they
4132 * don't support GLSL (at least on Windows).
4134 * PS2.0 requires at least 96 instructions, 2.0a/2.0b go up to 512. Assume that if the number
4135 * of instructions is 512 or less we have to do with ps2.0 hardware.
4136 * 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.
4138 if((GLINFO_LOCATION
.ps_nv_version
== PS_VERSION_20
) || (GLINFO_LOCATION
.ps_arb_max_instructions
<= 512))
4139 pCaps
->PixelShaderVersion
= WINED3DPS_VERSION(2,0);
4141 pCaps
->PixelShaderVersion
= WINED3DPS_VERSION(3,0);
4143 /* FIXME: The following line is card dependent. -8.0 to 8.0 is the
4144 * Direct3D minimum requirement.
4146 * Both GL_ARB_fragment_program and GLSL require a "maximum representable magnitude"
4147 * of colors to be 2^10, and 2^32 for other floats. Should we use 1024 here?
4149 * The problem is that the refrast clamps temporary results in the shader to
4150 * [-MaxValue;+MaxValue]. If the card's max value is bigger than the one we advertize here,
4151 * then applications may miss the clamping behavior. On the other hand, if it is smaller,
4152 * the shader will generate incorrect results too. Unfortunately, GL deliberately doesn't
4153 * offer a way to query this.
4155 pCaps
->PixelShader1xMaxValue
= 8.0;
4156 TRACE_(d3d_caps
)("Hardware pixel shader version %d.%d enabled (GLSL)\n", (pCaps
->PixelShaderVersion
>> 8) & 0xff, pCaps
->PixelShaderVersion
& 0xff);
4159 static BOOL
shader_glsl_color_fixup_supported(struct color_fixup_desc fixup
)
4161 if (TRACE_ON(d3d_shader
) && TRACE_ON(d3d
))
4163 TRACE("Checking support for fixup:\n");
4164 dump_color_fixup_desc(fixup
);
4167 /* We support everything except YUV conversions. */
4168 if (!is_yuv_fixup(fixup
))
4174 TRACE("[FAILED]\n");
4178 static const SHADER_HANDLER shader_glsl_instruction_handler_table
[WINED3DSIH_TABLE_SIZE
] =
4180 /* WINED3DSIH_ABS */ shader_glsl_map2gl
,
4181 /* WINED3DSIH_ADD */ shader_glsl_arith
,
4182 /* WINED3DSIH_BEM */ pshader_glsl_bem
,
4183 /* WINED3DSIH_BREAK */ shader_glsl_break
,
4184 /* WINED3DSIH_BREAKC */ shader_glsl_breakc
,
4185 /* WINED3DSIH_BREAKP */ NULL
,
4186 /* WINED3DSIH_CALL */ shader_glsl_call
,
4187 /* WINED3DSIH_CALLNZ */ shader_glsl_callnz
,
4188 /* WINED3DSIH_CMP */ shader_glsl_cmp
,
4189 /* WINED3DSIH_CND */ shader_glsl_cnd
,
4190 /* WINED3DSIH_CRS */ shader_glsl_cross
,
4191 /* WINED3DSIH_DCL */ NULL
,
4192 /* WINED3DSIH_DEF */ NULL
,
4193 /* WINED3DSIH_DEFB */ NULL
,
4194 /* WINED3DSIH_DEFI */ NULL
,
4195 /* WINED3DSIH_DP2ADD */ pshader_glsl_dp2add
,
4196 /* WINED3DSIH_DP3 */ shader_glsl_dot
,
4197 /* WINED3DSIH_DP4 */ shader_glsl_dot
,
4198 /* WINED3DSIH_DST */ shader_glsl_dst
,
4199 /* WINED3DSIH_DSX */ shader_glsl_map2gl
,
4200 /* WINED3DSIH_DSY */ shader_glsl_map2gl
,
4201 /* WINED3DSIH_ELSE */ shader_glsl_else
,
4202 /* WINED3DSIH_ENDIF */ shader_glsl_end
,
4203 /* WINED3DSIH_ENDLOOP */ shader_glsl_end
,
4204 /* WINED3DSIH_ENDREP */ shader_glsl_end
,
4205 /* WINED3DSIH_EXP */ shader_glsl_map2gl
,
4206 /* WINED3DSIH_EXPP */ shader_glsl_expp
,
4207 /* WINED3DSIH_FRC */ shader_glsl_map2gl
,
4208 /* WINED3DSIH_IF */ shader_glsl_if
,
4209 /* WINED3DSIH_IFC */ shader_glsl_ifc
,
4210 /* WINED3DSIH_LABEL */ shader_glsl_label
,
4211 /* WINED3DSIH_LIT */ shader_glsl_lit
,
4212 /* WINED3DSIH_LOG */ shader_glsl_log
,
4213 /* WINED3DSIH_LOGP */ shader_glsl_log
,
4214 /* WINED3DSIH_LOOP */ shader_glsl_loop
,
4215 /* WINED3DSIH_LRP */ shader_glsl_lrp
,
4216 /* WINED3DSIH_M3x2 */ shader_glsl_mnxn
,
4217 /* WINED3DSIH_M3x3 */ shader_glsl_mnxn
,
4218 /* WINED3DSIH_M3x4 */ shader_glsl_mnxn
,
4219 /* WINED3DSIH_M4x3 */ shader_glsl_mnxn
,
4220 /* WINED3DSIH_M4x4 */ shader_glsl_mnxn
,
4221 /* WINED3DSIH_MAD */ shader_glsl_mad
,
4222 /* WINED3DSIH_MAX */ shader_glsl_map2gl
,
4223 /* WINED3DSIH_MIN */ shader_glsl_map2gl
,
4224 /* WINED3DSIH_MOV */ shader_glsl_mov
,
4225 /* WINED3DSIH_MOVA */ shader_glsl_mov
,
4226 /* WINED3DSIH_MUL */ shader_glsl_arith
,
4227 /* WINED3DSIH_NOP */ NULL
,
4228 /* WINED3DSIH_NRM */ shader_glsl_map2gl
,
4229 /* WINED3DSIH_PHASE */ NULL
,
4230 /* WINED3DSIH_POW */ shader_glsl_pow
,
4231 /* WINED3DSIH_RCP */ shader_glsl_rcp
,
4232 /* WINED3DSIH_REP */ shader_glsl_rep
,
4233 /* WINED3DSIH_RET */ NULL
,
4234 /* WINED3DSIH_RSQ */ shader_glsl_rsq
,
4235 /* WINED3DSIH_SETP */ NULL
,
4236 /* WINED3DSIH_SGE */ shader_glsl_compare
,
4237 /* WINED3DSIH_SGN */ shader_glsl_map2gl
,
4238 /* WINED3DSIH_SINCOS */ shader_glsl_sincos
,
4239 /* WINED3DSIH_SLT */ shader_glsl_compare
,
4240 /* WINED3DSIH_SUB */ shader_glsl_arith
,
4241 /* WINED3DSIH_TEX */ pshader_glsl_tex
,
4242 /* WINED3DSIH_TEXBEM */ pshader_glsl_texbem
,
4243 /* WINED3DSIH_TEXBEML */ pshader_glsl_texbem
,
4244 /* WINED3DSIH_TEXCOORD */ pshader_glsl_texcoord
,
4245 /* WINED3DSIH_TEXDEPTH */ pshader_glsl_texdepth
,
4246 /* WINED3DSIH_TEXDP3 */ pshader_glsl_texdp3
,
4247 /* WINED3DSIH_TEXDP3TEX */ pshader_glsl_texdp3tex
,
4248 /* WINED3DSIH_TEXKILL */ pshader_glsl_texkill
,
4249 /* WINED3DSIH_TEXLDD */ NULL
,
4250 /* WINED3DSIH_TEXLDL */ shader_glsl_texldl
,
4251 /* WINED3DSIH_TEXM3x2DEPTH */ pshader_glsl_texm3x2depth
,
4252 /* WINED3DSIH_TEXM3x2PAD */ pshader_glsl_texm3x2pad
,
4253 /* WINED3DSIH_TEXM3x2TEX */ pshader_glsl_texm3x2tex
,
4254 /* WINED3DSIH_TEXM3x3 */ pshader_glsl_texm3x3
,
4255 /* WINED3DSIH_TEXM3x3DIFF */ NULL
,
4256 /* WINED3DSIH_TEXM3x3PAD */ pshader_glsl_texm3x3pad
,
4257 /* WINED3DSIH_TEXM3x3SPEC */ pshader_glsl_texm3x3spec
,
4258 /* WINED3DSIH_TEXM3x3TEX */ pshader_glsl_texm3x3tex
,
4259 /* WINED3DSIH_TEXM3x3VSPEC */ pshader_glsl_texm3x3vspec
,
4260 /* WINED3DSIH_TEXREG2AR */ pshader_glsl_texreg2ar
,
4261 /* WINED3DSIH_TEXREG2GB */ pshader_glsl_texreg2gb
,
4262 /* WINED3DSIH_TEXREG2RGB */ pshader_glsl_texreg2rgb
,
4265 const shader_backend_t glsl_shader_backend
= {
4266 shader_glsl_instruction_handler_table
,
4268 shader_glsl_select_depth_blt
,
4269 shader_glsl_deselect_depth_blt
,
4270 shader_glsl_update_float_vertex_constants
,
4271 shader_glsl_update_float_pixel_constants
,
4272 shader_glsl_load_constants
,
4273 shader_glsl_destroy
,
4276 shader_glsl_dirty_const
,
4277 shader_glsl_generate_pshader
,
4278 shader_glsl_generate_vshader
,
4279 shader_glsl_get_caps
,
4280 shader_glsl_color_fixup_supported
,