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 unsigned int mask_size
, remaining
;
1441 glsl_dst_param_t dst_param
;
1442 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
);
1464 dst_param
.mask_str
[0] = '\0';
1465 shader_glsl_get_write_mask(mask
, dst_param
.mask_str
);
1467 dst_param
.reg_name
[0] = '\0';
1468 shader_glsl_get_register_name(ins
->dst
[0].token
, ins
->dst
[0].addr_token
, dst_param
.reg_name
, &dummy
, ins
);
1470 arguments
[0] = '\0';
1471 remaining
= mask_size
;
1472 if (mask
& WINED3DSP_WRITEMASK_0
)
1474 shader_glsl_append_fixup_arg(arguments
, dst_param
.reg_name
, fixup
.x_sign_fixup
, fixup
.x_source
);
1475 if (--remaining
) strcat(arguments
, ", ");
1477 if (mask
& WINED3DSP_WRITEMASK_1
)
1479 shader_glsl_append_fixup_arg(arguments
, dst_param
.reg_name
, fixup
.y_sign_fixup
, fixup
.y_source
);
1480 if (--remaining
) strcat(arguments
, ", ");
1482 if (mask
& WINED3DSP_WRITEMASK_2
)
1484 shader_glsl_append_fixup_arg(arguments
, dst_param
.reg_name
, fixup
.z_sign_fixup
, fixup
.z_source
);
1485 if (--remaining
) strcat(arguments
, ", ");
1487 if (mask
& WINED3DSP_WRITEMASK_3
)
1489 shader_glsl_append_fixup_arg(arguments
, dst_param
.reg_name
, fixup
.w_sign_fixup
, fixup
.w_source
);
1490 if (--remaining
) strcat(arguments
, ", ");
1495 shader_addline(ins
->buffer
, "%s%s = vec%u(%s);\n",
1496 dst_param
.reg_name
, dst_param
.mask_str
, mask_size
, arguments
);
1500 shader_addline(ins
->buffer
, "%s%s = %s;\n", dst_param
.reg_name
, dst_param
.mask_str
, arguments
);
1504 static void PRINTF_ATTR(6, 7) shader_glsl_gen_sample_code(const struct wined3d_shader_instruction
*ins
,
1505 DWORD sampler
, const glsl_sample_function_t
*sample_function
, DWORD swizzle
,
1506 const char *bias
, const char *coord_reg_fmt
, ...)
1508 const char *sampler_base
;
1509 char dst_swizzle
[6];
1510 struct color_fixup_desc fixup
;
1511 BOOL rect_fixup
= FALSE
;
1514 shader_glsl_get_swizzle(swizzle
, FALSE
, ins
->dst
[0].write_mask
, dst_swizzle
);
1516 if (shader_is_pshader_version(ins
->reg_maps
->shader_version
))
1518 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*)ins
->shader
;
1519 fixup
= This
->cur_args
->color_fixup
[sampler
];
1520 sampler_base
= "Psampler";
1522 if(This
->cur_args
->texrect_fixup
& (1 << sampler
)) {
1524 FIXME("Biased sampling from RECT textures is unsupported\n");
1530 sampler_base
= "Vsampler";
1531 fixup
= COLOR_FIXUP_IDENTITY
; /* FIXME: Vshader color fixup */
1534 shader_glsl_append_dst(ins
->buffer
, ins
);
1536 shader_addline(ins
->buffer
, "%s(%s%u, ", sample_function
->name
, sampler_base
, sampler
);
1538 va_start(args
, coord_reg_fmt
);
1539 shader_vaddline(ins
->buffer
, coord_reg_fmt
, args
);
1543 shader_addline(ins
->buffer
, ", %s)%s);\n", bias
, dst_swizzle
);
1546 shader_addline(ins
->buffer
, " * PsamplerRectFixup%u)%s);\n", sampler
, dst_swizzle
);
1548 shader_addline(ins
->buffer
, ")%s);\n", dst_swizzle
);
1552 if(!is_identity_fixup(fixup
)) {
1553 shader_glsl_color_correction(ins
, fixup
);
1557 /*****************************************************************************
1559 * Begin processing individual instruction opcodes
1561 ****************************************************************************/
1563 /* Generate GLSL arithmetic functions (dst = src1 + src2) */
1564 static void shader_glsl_arith(const struct wined3d_shader_instruction
*ins
)
1566 SHADER_BUFFER
*buffer
= ins
->buffer
;
1567 glsl_src_param_t src0_param
;
1568 glsl_src_param_t src1_param
;
1572 /* Determine the GLSL operator to use based on the opcode */
1573 switch (ins
->handler_idx
)
1575 case WINED3DSIH_MUL
: op
= '*'; break;
1576 case WINED3DSIH_ADD
: op
= '+'; break;
1577 case WINED3DSIH_SUB
: op
= '-'; break;
1580 FIXME("Opcode %#x not yet handled in GLSL\n", ins
->handler_idx
);
1584 write_mask
= shader_glsl_append_dst(buffer
, ins
);
1585 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], write_mask
, &src0_param
);
1586 shader_glsl_add_src_param(ins
, ins
->src
[1], ins
->src_addr
[1], write_mask
, &src1_param
);
1587 shader_addline(buffer
, "%s %c %s);\n", src0_param
.param_str
, op
, src1_param
.param_str
);
1590 /* Process the WINED3DSIO_MOV opcode using GLSL (dst = src) */
1591 static void shader_glsl_mov(const struct wined3d_shader_instruction
*ins
)
1593 SHADER_BUFFER
*buffer
= ins
->buffer
;
1594 glsl_src_param_t src0_param
;
1597 write_mask
= shader_glsl_append_dst(buffer
, ins
);
1598 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], write_mask
, &src0_param
);
1600 /* In vs_1_1 WINED3DSIO_MOV can write to the address register. In later
1601 * shader versions WINED3DSIO_MOVA is used for this. */
1602 if ((WINED3DSHADER_VERSION_MAJOR(ins
->reg_maps
->shader_version
) == 1
1603 && !shader_is_pshader_version(ins
->reg_maps
->shader_version
)
1604 && shader_get_regtype(ins
->dst
[0].token
) == WINED3DSPR_ADDR
))
1606 /* This is a simple floor() */
1607 unsigned int mask_size
= shader_glsl_get_write_mask_size(write_mask
);
1608 if (mask_size
> 1) {
1609 shader_addline(buffer
, "ivec%d(floor(%s)));\n", mask_size
, src0_param
.param_str
);
1611 shader_addline(buffer
, "int(floor(%s)));\n", src0_param
.param_str
);
1614 else if(ins
->handler_idx
== WINED3DSIH_MOVA
)
1616 /* We need to *round* to the nearest int here. */
1617 unsigned int mask_size
= shader_glsl_get_write_mask_size(write_mask
);
1618 if (mask_size
> 1) {
1619 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
);
1621 shader_addline(buffer
, "int(floor(abs(%s) + 0.5) * sign(%s)));\n", src0_param
.param_str
, src0_param
.param_str
);
1624 shader_addline(buffer
, "%s);\n", src0_param
.param_str
);
1628 /* Process the dot product operators DP3 and DP4 in GLSL (dst = dot(src0, src1)) */
1629 static void shader_glsl_dot(const struct wined3d_shader_instruction
*ins
)
1631 SHADER_BUFFER
*buffer
= ins
->buffer
;
1632 glsl_src_param_t src0_param
;
1633 glsl_src_param_t src1_param
;
1634 DWORD dst_write_mask
, src_write_mask
;
1635 unsigned int dst_size
= 0;
1637 dst_write_mask
= shader_glsl_append_dst(buffer
, ins
);
1638 dst_size
= shader_glsl_get_write_mask_size(dst_write_mask
);
1640 /* dp3 works on vec3, dp4 on vec4 */
1641 if (ins
->handler_idx
== WINED3DSIH_DP4
)
1643 src_write_mask
= WINED3DSP_WRITEMASK_ALL
;
1645 src_write_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
1648 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], src_write_mask
, &src0_param
);
1649 shader_glsl_add_src_param(ins
, ins
->src
[1], ins
->src_addr
[1], src_write_mask
, &src1_param
);
1652 shader_addline(buffer
, "vec%d(dot(%s, %s)));\n", dst_size
, src0_param
.param_str
, src1_param
.param_str
);
1654 shader_addline(buffer
, "dot(%s, %s));\n", src0_param
.param_str
, src1_param
.param_str
);
1658 /* Note that this instruction has some restrictions. The destination write mask
1659 * can't contain the w component, and the source swizzles have to be .xyzw */
1660 static void shader_glsl_cross(const struct wined3d_shader_instruction
*ins
)
1662 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
1663 glsl_src_param_t src0_param
;
1664 glsl_src_param_t src1_param
;
1667 shader_glsl_get_write_mask(ins
->dst
[0].token
, dst_mask
);
1668 shader_glsl_append_dst(ins
->buffer
, ins
);
1669 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], src_mask
, &src0_param
);
1670 shader_glsl_add_src_param(ins
, ins
->src
[1], ins
->src_addr
[1], src_mask
, &src1_param
);
1671 shader_addline(ins
->buffer
, "cross(%s, %s)%s);\n", src0_param
.param_str
, src1_param
.param_str
, dst_mask
);
1674 /* Process the WINED3DSIO_POW instruction in GLSL (dst = |src0|^src1)
1675 * Src0 and src1 are scalars. Note that D3D uses the absolute of src0, while
1676 * GLSL uses the value as-is. */
1677 static void shader_glsl_pow(const struct wined3d_shader_instruction
*ins
)
1679 SHADER_BUFFER
*buffer
= ins
->buffer
;
1680 glsl_src_param_t src0_param
;
1681 glsl_src_param_t src1_param
;
1682 DWORD dst_write_mask
;
1683 unsigned int dst_size
;
1685 dst_write_mask
= shader_glsl_append_dst(buffer
, ins
);
1686 dst_size
= shader_glsl_get_write_mask_size(dst_write_mask
);
1688 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
1689 shader_glsl_add_src_param(ins
, ins
->src
[1], ins
->src_addr
[1], WINED3DSP_WRITEMASK_0
, &src1_param
);
1692 shader_addline(buffer
, "vec%d(pow(abs(%s), %s)));\n", dst_size
, src0_param
.param_str
, src1_param
.param_str
);
1694 shader_addline(buffer
, "pow(abs(%s), %s));\n", src0_param
.param_str
, src1_param
.param_str
);
1698 /* Process the WINED3DSIO_LOG instruction in GLSL (dst = log2(|src0|))
1699 * Src0 is a scalar. Note that D3D uses the absolute of src0, while
1700 * GLSL uses the value as-is. */
1701 static void shader_glsl_log(const struct wined3d_shader_instruction
*ins
)
1703 SHADER_BUFFER
*buffer
= ins
->buffer
;
1704 glsl_src_param_t src0_param
;
1705 DWORD dst_write_mask
;
1706 unsigned int dst_size
;
1708 dst_write_mask
= shader_glsl_append_dst(buffer
, ins
);
1709 dst_size
= shader_glsl_get_write_mask_size(dst_write_mask
);
1711 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
1714 shader_addline(buffer
, "vec%d(log2(abs(%s))));\n", dst_size
, src0_param
.param_str
);
1716 shader_addline(buffer
, "log2(abs(%s)));\n", src0_param
.param_str
);
1720 /* Map the opcode 1-to-1 to the GL code (arg->dst = instruction(src0, src1, ...) */
1721 static void shader_glsl_map2gl(const struct wined3d_shader_instruction
*ins
)
1723 SHADER_BUFFER
*buffer
= ins
->buffer
;
1724 glsl_src_param_t src_param
;
1725 const char *instruction
;
1729 /* Determine the GLSL function to use based on the opcode */
1730 /* TODO: Possibly make this a table for faster lookups */
1731 switch (ins
->handler_idx
)
1733 case WINED3DSIH_MIN
: instruction
= "min"; break;
1734 case WINED3DSIH_MAX
: instruction
= "max"; break;
1735 case WINED3DSIH_ABS
: instruction
= "abs"; break;
1736 case WINED3DSIH_FRC
: instruction
= "fract"; break;
1737 case WINED3DSIH_NRM
: instruction
= "normalize"; break;
1738 case WINED3DSIH_EXP
: instruction
= "exp2"; break;
1739 case WINED3DSIH_SGN
: instruction
= "sign"; break;
1740 case WINED3DSIH_DSX
: instruction
= "dFdx"; break;
1741 case WINED3DSIH_DSY
: instruction
= "ycorrection.y * dFdy"; break;
1742 default: instruction
= "";
1743 FIXME("Opcode %#x not yet handled in GLSL\n", ins
->handler_idx
);
1747 write_mask
= shader_glsl_append_dst(buffer
, ins
);
1749 shader_addline(buffer
, "%s(", instruction
);
1753 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], write_mask
, &src_param
);
1754 shader_addline(buffer
, "%s", src_param
.param_str
);
1755 for (i
= 1; i
< ins
->src_count
; ++i
)
1757 shader_glsl_add_src_param(ins
, ins
->src
[i
], ins
->src_addr
[i
], write_mask
, &src_param
);
1758 shader_addline(buffer
, ", %s", src_param
.param_str
);
1762 shader_addline(buffer
, "));\n");
1765 /** Process the WINED3DSIO_EXPP instruction in GLSL:
1766 * For shader model 1.x, do the following (and honor the writemask, so use a temporary variable):
1767 * dst.x = 2^(floor(src))
1768 * dst.y = src - floor(src)
1769 * dst.z = 2^src (partial precision is allowed, but optional)
1771 * For 2.0 shaders, just do this (honoring writemask and swizzle):
1772 * dst = 2^src; (partial precision is allowed, but optional)
1774 static void shader_glsl_expp(const struct wined3d_shader_instruction
*ins
)
1776 glsl_src_param_t src_param
;
1778 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src_param
);
1780 if (ins
->reg_maps
->shader_version
< WINED3DPS_VERSION(2,0))
1784 shader_addline(ins
->buffer
, "tmp0.x = exp2(floor(%s));\n", src_param
.param_str
);
1785 shader_addline(ins
->buffer
, "tmp0.y = %s - floor(%s);\n", src_param
.param_str
, src_param
.param_str
);
1786 shader_addline(ins
->buffer
, "tmp0.z = exp2(%s);\n", src_param
.param_str
);
1787 shader_addline(ins
->buffer
, "tmp0.w = 1.0;\n");
1789 shader_glsl_append_dst(ins
->buffer
, ins
);
1790 shader_glsl_get_write_mask(ins
->dst
[0].token
, dst_mask
);
1791 shader_addline(ins
->buffer
, "tmp0%s);\n", dst_mask
);
1794 unsigned int mask_size
;
1796 write_mask
= shader_glsl_append_dst(ins
->buffer
, ins
);
1797 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
1799 if (mask_size
> 1) {
1800 shader_addline(ins
->buffer
, "vec%d(exp2(%s)));\n", mask_size
, src_param
.param_str
);
1802 shader_addline(ins
->buffer
, "exp2(%s));\n", src_param
.param_str
);
1807 /** Process the RCP (reciprocal or inverse) opcode in GLSL (dst = 1 / src) */
1808 static void shader_glsl_rcp(const struct wined3d_shader_instruction
*ins
)
1810 glsl_src_param_t src_param
;
1812 unsigned int mask_size
;
1814 write_mask
= shader_glsl_append_dst(ins
->buffer
, ins
);
1815 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
1816 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], WINED3DSP_WRITEMASK_3
, &src_param
);
1818 if (mask_size
> 1) {
1819 shader_addline(ins
->buffer
, "vec%d(1.0 / %s));\n", mask_size
, src_param
.param_str
);
1821 shader_addline(ins
->buffer
, "1.0 / %s);\n", src_param
.param_str
);
1825 static void shader_glsl_rsq(const struct wined3d_shader_instruction
*ins
)
1827 SHADER_BUFFER
*buffer
= ins
->buffer
;
1828 glsl_src_param_t src_param
;
1830 unsigned int mask_size
;
1832 write_mask
= shader_glsl_append_dst(buffer
, ins
);
1833 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
1835 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], WINED3DSP_WRITEMASK_3
, &src_param
);
1837 if (mask_size
> 1) {
1838 shader_addline(buffer
, "vec%d(inversesqrt(%s)));\n", mask_size
, src_param
.param_str
);
1840 shader_addline(buffer
, "inversesqrt(%s));\n", src_param
.param_str
);
1844 /** Process signed comparison opcodes in GLSL. */
1845 static void shader_glsl_compare(const struct wined3d_shader_instruction
*ins
)
1847 glsl_src_param_t src0_param
;
1848 glsl_src_param_t src1_param
;
1850 unsigned int mask_size
;
1852 write_mask
= shader_glsl_append_dst(ins
->buffer
, ins
);
1853 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
1854 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], write_mask
, &src0_param
);
1855 shader_glsl_add_src_param(ins
, ins
->src
[1], ins
->src_addr
[1], write_mask
, &src1_param
);
1857 if (mask_size
> 1) {
1858 const char *compare
;
1860 switch(ins
->handler_idx
)
1862 case WINED3DSIH_SLT
: compare
= "lessThan"; break;
1863 case WINED3DSIH_SGE
: compare
= "greaterThanEqual"; break;
1864 default: compare
= "";
1865 FIXME("Can't handle opcode %#x\n", ins
->handler_idx
);
1868 shader_addline(ins
->buffer
, "vec%d(%s(%s, %s)));\n", mask_size
, compare
,
1869 src0_param
.param_str
, src1_param
.param_str
);
1871 switch(ins
->handler_idx
)
1873 case WINED3DSIH_SLT
:
1874 /* Step(src0, src1) is not suitable here because if src0 == src1 SLT is supposed,
1875 * to return 0.0 but step returns 1.0 because step is not < x
1876 * An alternative is a bvec compare padded with an unused second component.
1877 * step(src1 * -1.0, src0 * -1.0) is not an option because it suffers from the same
1878 * issue. Playing with not() is not possible either because not() does not accept
1881 shader_addline(ins
->buffer
, "(%s < %s) ? 1.0 : 0.0);\n", src0_param
.param_str
, src1_param
.param_str
);
1883 case WINED3DSIH_SGE
:
1884 /* Here we can use the step() function and safe a conditional */
1885 shader_addline(ins
->buffer
, "step(%s, %s));\n", src1_param
.param_str
, src0_param
.param_str
);
1888 FIXME("Can't handle opcode %#x\n", ins
->handler_idx
);
1894 /** Process CMP instruction in GLSL (dst = src0 >= 0.0 ? src1 : src2), per channel */
1895 static void shader_glsl_cmp(const struct wined3d_shader_instruction
*ins
)
1897 glsl_src_param_t src0_param
;
1898 glsl_src_param_t src1_param
;
1899 glsl_src_param_t src2_param
;
1900 DWORD write_mask
, cmp_channel
= 0;
1903 BOOL temp_destination
= FALSE
;
1905 if (shader_is_scalar(shader_get_regtype(ins
->src
[0]), ins
->src
[0] & WINED3DSP_REGNUM_MASK
))
1907 write_mask
= shader_glsl_append_dst(ins
->buffer
, ins
);
1909 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], WINED3DSP_WRITEMASK_ALL
, &src0_param
);
1910 shader_glsl_add_src_param(ins
, ins
->src
[1], ins
->src_addr
[1], write_mask
, &src1_param
);
1911 shader_glsl_add_src_param(ins
, ins
->src
[2], ins
->src_addr
[2], write_mask
, &src2_param
);
1913 shader_addline(ins
->buffer
, "%s >= 0.0 ? %s : %s);\n",
1914 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
1916 DWORD src0reg
= ins
->src
[0] & WINED3DSP_REGNUM_MASK
;
1917 DWORD src1reg
= ins
->src
[1] & WINED3DSP_REGNUM_MASK
;
1918 DWORD src2reg
= ins
->src
[2] & WINED3DSP_REGNUM_MASK
;
1919 DWORD src0regtype
= shader_get_regtype(ins
->src
[0]);
1920 DWORD src1regtype
= shader_get_regtype(ins
->src
[1]);
1921 DWORD src2regtype
= shader_get_regtype(ins
->src
[2]);
1922 DWORD dstreg
= ins
->dst
[0].register_idx
;
1923 DWORD dstregtype
= shader_get_regtype(ins
->dst
[0].token
);
1925 /* Cycle through all source0 channels */
1926 for (i
=0; i
<4; i
++) {
1928 /* Find the destination channels which use the current source0 channel */
1929 for (j
=0; j
<4; j
++) {
1930 if (((ins
->src
[0] >> (WINED3DSP_SWIZZLE_SHIFT
+ 2 * j
)) & 0x3) == i
)
1932 write_mask
|= WINED3DSP_WRITEMASK_0
<< j
;
1933 cmp_channel
= WINED3DSP_WRITEMASK_0
<< j
;
1937 /* Splitting the cmp instruction up in multiple lines imposes a problem:
1938 * The first lines may overwrite source parameters of the following lines.
1939 * Deal with that by using a temporary destination register if needed
1941 if((src0reg
== dstreg
&& src0regtype
== dstregtype
) ||
1942 (src1reg
== dstreg
&& src1regtype
== dstregtype
) ||
1943 (src2reg
== dstreg
&& src2regtype
== dstregtype
)) {
1945 write_mask
= shader_glsl_get_write_mask(ins
->dst
[0].token
& (~WINED3DSP_WRITEMASK_ALL
| write_mask
),
1947 if (!write_mask
) continue;
1948 shader_addline(ins
->buffer
, "tmp0%s = (", mask_char
);
1949 temp_destination
= TRUE
;
1951 struct wined3d_shader_dst_param dst
= ins
->dst
[0];
1952 dst
.write_mask
&= write_mask
;
1953 dst
.token
&= ~WINED3DSP_WRITEMASK_ALL
| write_mask
;
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 glsl_src_param_t src0_param
;
1981 glsl_src_param_t src1_param
;
1982 glsl_src_param_t src2_param
;
1983 DWORD write_mask
, cmp_channel
= 0;
1986 if (ins
->reg_maps
->shader_version
< WINED3DPS_VERSION(1, 4))
1988 write_mask
= shader_glsl_append_dst(ins
->buffer
, ins
);
1989 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
1990 shader_glsl_add_src_param(ins
, ins
->src
[1], ins
->src_addr
[1], write_mask
, &src1_param
);
1991 shader_glsl_add_src_param(ins
, ins
->src
[2], ins
->src_addr
[2], write_mask
, &src2_param
);
1993 /* Fun: The D3DSI_COISSUE flag changes the semantic of the cnd instruction for < 1.4 shaders */
1996 shader_addline(ins
->buffer
, "%s /* COISSUE! */);\n", src1_param
.param_str
);
1998 shader_addline(ins
->buffer
, "%s > 0.5 ? %s : %s);\n",
1999 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
2003 /* Cycle through all source0 channels */
2004 for (i
=0; i
<4; i
++) {
2005 struct wined3d_shader_dst_param dst
;
2007 /* Find the destination channels which use the current source0 channel */
2008 for (j
=0; j
<4; j
++) {
2009 if (((ins
->src
[0] >> (WINED3DSP_SWIZZLE_SHIFT
+ 2 * j
)) & 0x3) == i
)
2011 write_mask
|= WINED3DSP_WRITEMASK_0
<< j
;
2012 cmp_channel
= WINED3DSP_WRITEMASK_0
<< j
;
2016 dst
.write_mask
&= write_mask
;
2017 dst
.token
&= ~WINED3DSP_WRITEMASK_ALL
| write_mask
;
2018 write_mask
= shader_glsl_append_dst_ext(ins
->buffer
, ins
, &dst
);
2019 if (!write_mask
) continue;
2021 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], cmp_channel
, &src0_param
);
2022 shader_glsl_add_src_param(ins
, ins
->src
[1], ins
->src_addr
[1], write_mask
, &src1_param
);
2023 shader_glsl_add_src_param(ins
, ins
->src
[2], ins
->src_addr
[2], write_mask
, &src2_param
);
2025 shader_addline(ins
->buffer
, "%s > 0.5 ? %s : %s);\n",
2026 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
2030 /** GLSL code generation for WINED3DSIO_MAD: Multiply the first 2 opcodes, then add the last */
2031 static void shader_glsl_mad(const struct wined3d_shader_instruction
*ins
)
2033 glsl_src_param_t src0_param
;
2034 glsl_src_param_t src1_param
;
2035 glsl_src_param_t src2_param
;
2038 write_mask
= shader_glsl_append_dst(ins
->buffer
, ins
);
2039 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], write_mask
, &src0_param
);
2040 shader_glsl_add_src_param(ins
, ins
->src
[1], ins
->src_addr
[1], write_mask
, &src1_param
);
2041 shader_glsl_add_src_param(ins
, ins
->src
[2], ins
->src_addr
[2], write_mask
, &src2_param
);
2042 shader_addline(ins
->buffer
, "(%s * %s) + %s);\n",
2043 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
2046 /** Handles transforming all WINED3DSIO_M?x? opcodes for
2047 Vertex shaders to GLSL codes */
2048 static void shader_glsl_mnxn(const struct wined3d_shader_instruction
*ins
)
2051 int nComponents
= 0;
2052 struct wined3d_shader_dst_param tmp_dst
= {0};
2053 struct wined3d_shader_instruction tmp_ins
;
2055 memset(&tmp_ins
, 0, sizeof(tmp_ins
));
2057 /* Set constants for the temporary argument */
2058 tmp_ins
.shader
= ins
->shader
;
2059 tmp_ins
.buffer
= ins
->buffer
;
2060 tmp_ins
.src
[0] = ins
->src
[0];
2061 tmp_ins
.src_addr
[0] = ins
->src_addr
[0];
2062 tmp_ins
.src_addr
[1] = ins
->src_addr
[1];
2063 tmp_ins
.reg_maps
= ins
->reg_maps
;
2064 tmp_ins
.dst_count
= 1;
2065 tmp_ins
.dst
= &tmp_dst
;
2066 tmp_ins
.src_count
= 2;
2068 switch(ins
->handler_idx
)
2070 case WINED3DSIH_M4x4
:
2072 tmp_ins
.handler_idx
= WINED3DSIH_DP4
;
2074 case WINED3DSIH_M4x3
:
2076 tmp_ins
.handler_idx
= WINED3DSIH_DP4
;
2078 case WINED3DSIH_M3x4
:
2080 tmp_ins
.handler_idx
= WINED3DSIH_DP3
;
2082 case WINED3DSIH_M3x3
:
2084 tmp_ins
.handler_idx
= WINED3DSIH_DP3
;
2086 case WINED3DSIH_M3x2
:
2088 tmp_ins
.handler_idx
= WINED3DSIH_DP3
;
2094 tmp_dst
= ins
->dst
[0];
2095 for (i
= 0; i
< nComponents
; ++i
)
2097 tmp_dst
.write_mask
= WINED3DSP_WRITEMASK_0
<< i
;
2098 tmp_dst
.token
= (tmp_dst
.token
& ~WINED3DSP_WRITEMASK_ALL
) | tmp_dst
.write_mask
;
2099 tmp_ins
.src
[1] = ins
->src
[1] + i
;
2100 shader_glsl_dot(&tmp_ins
);
2105 The LRP instruction performs a component-wise linear interpolation
2106 between the second and third operands using the first operand as the
2107 blend factor. Equation: (dst = src2 + src0 * (src1 - src2))
2108 This is equivalent to mix(src2, src1, src0);
2110 static void shader_glsl_lrp(const struct wined3d_shader_instruction
*ins
)
2112 glsl_src_param_t src0_param
;
2113 glsl_src_param_t src1_param
;
2114 glsl_src_param_t src2_param
;
2117 write_mask
= shader_glsl_append_dst(ins
->buffer
, ins
);
2119 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], write_mask
, &src0_param
);
2120 shader_glsl_add_src_param(ins
, ins
->src
[1], ins
->src_addr
[1], write_mask
, &src1_param
);
2121 shader_glsl_add_src_param(ins
, ins
->src
[2], ins
->src_addr
[2], write_mask
, &src2_param
);
2123 shader_addline(ins
->buffer
, "mix(%s, %s, %s));\n",
2124 src2_param
.param_str
, src1_param
.param_str
, src0_param
.param_str
);
2127 /** Process the WINED3DSIO_LIT instruction in GLSL:
2128 * dst.x = dst.w = 1.0
2129 * dst.y = (src0.x > 0) ? src0.x
2130 * dst.z = (src0.x > 0) ? ((src0.y > 0) ? pow(src0.y, src.w) : 0) : 0
2131 * where src.w is clamped at +- 128
2133 static void shader_glsl_lit(const struct wined3d_shader_instruction
*ins
)
2135 glsl_src_param_t src0_param
;
2136 glsl_src_param_t src1_param
;
2137 glsl_src_param_t src3_param
;
2140 shader_glsl_append_dst(ins
->buffer
, ins
);
2141 shader_glsl_get_write_mask(ins
->dst
[0].token
, dst_mask
);
2143 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2144 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], WINED3DSP_WRITEMASK_1
, &src1_param
);
2145 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], WINED3DSP_WRITEMASK_3
, &src3_param
);
2147 /* The sdk specifies the instruction like this
2149 * if(src.x > 0.0) dst.y = src.x
2151 * if(src.x > 0.0 && src.y > 0.0) dst.z = pow(src.y, power);
2155 * Obviously that has quite a few conditionals in it which we don't like. So the first step is this:
2156 * dst.x = 1.0 ... No further explanation needed
2157 * dst.y = max(src.y, 0.0); ... If x < 0.0, use 0.0, otherwise x. Same as the conditional
2158 * dst.z = x > 0.0 ? pow(max(y, 0.0), p) : 0; ... 0 ^ power is 0, and otherwise we use y anyway
2159 * dst.w = 1.0. ... Nothing fancy.
2161 * So we still have one conditional in there. So do this:
2162 * dst.z = pow(max(0.0, src.y) * step(0.0, src.x), power);
2164 * 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),
2165 * 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.
2166 * if both x and y are > 0, we get pow(y * 1.0, power), as it is supposed to
2168 shader_addline(ins
->buffer
,
2169 "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",
2170 src0_param
.param_str
, src1_param
.param_str
, src0_param
.param_str
, src3_param
.param_str
, dst_mask
);
2173 /** Process the WINED3DSIO_DST instruction in GLSL:
2175 * dst.y = src0.x * src0.y
2179 static void shader_glsl_dst(const struct wined3d_shader_instruction
*ins
)
2181 glsl_src_param_t src0y_param
;
2182 glsl_src_param_t src0z_param
;
2183 glsl_src_param_t src1y_param
;
2184 glsl_src_param_t src1w_param
;
2187 shader_glsl_append_dst(ins
->buffer
, ins
);
2188 shader_glsl_get_write_mask(ins
->dst
[0].token
, dst_mask
);
2190 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], WINED3DSP_WRITEMASK_1
, &src0y_param
);
2191 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], WINED3DSP_WRITEMASK_2
, &src0z_param
);
2192 shader_glsl_add_src_param(ins
, ins
->src
[1], ins
->src_addr
[1], WINED3DSP_WRITEMASK_1
, &src1y_param
);
2193 shader_glsl_add_src_param(ins
, ins
->src
[1], ins
->src_addr
[1], WINED3DSP_WRITEMASK_3
, &src1w_param
);
2195 shader_addline(ins
->buffer
, "vec4(1.0, %s * %s, %s, %s))%s;\n",
2196 src0y_param
.param_str
, src1y_param
.param_str
, src0z_param
.param_str
, src1w_param
.param_str
, dst_mask
);
2199 /** Process the WINED3DSIO_SINCOS instruction in GLSL:
2200 * VS 2.0 requires that specific cosine and sine constants be passed to this instruction so the hardware
2201 * can handle it. But, these functions are built-in for GLSL, so we can just ignore the last 2 params.
2203 * dst.x = cos(src0.?)
2204 * dst.y = sin(src0.?)
2208 static void shader_glsl_sincos(const struct wined3d_shader_instruction
*ins
)
2210 glsl_src_param_t src0_param
;
2213 write_mask
= shader_glsl_append_dst(ins
->buffer
, ins
);
2214 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2216 switch (write_mask
) {
2217 case WINED3DSP_WRITEMASK_0
:
2218 shader_addline(ins
->buffer
, "cos(%s));\n", src0_param
.param_str
);
2221 case WINED3DSP_WRITEMASK_1
:
2222 shader_addline(ins
->buffer
, "sin(%s));\n", src0_param
.param_str
);
2225 case (WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
):
2226 shader_addline(ins
->buffer
, "vec2(cos(%s), sin(%s)));\n", src0_param
.param_str
, src0_param
.param_str
);
2230 ERR("Write mask should be .x, .y or .xy\n");
2235 /** Process the WINED3DSIO_LOOP instruction in GLSL:
2236 * Start a for() loop where src1.y is the initial value of aL,
2237 * increment aL by src1.z for a total of src1.x iterations.
2238 * Need to use a temporary variable for this operation.
2240 /* FIXME: I don't think nested loops will work correctly this way. */
2241 static void shader_glsl_loop(const struct wined3d_shader_instruction
*ins
)
2243 glsl_src_param_t src1_param
;
2244 IWineD3DBaseShaderImpl
*shader
= (IWineD3DBaseShaderImpl
*)ins
->shader
;
2245 DWORD regtype
= shader_get_regtype(ins
->src
[1]);
2246 DWORD reg
= ins
->src
[1] & WINED3DSP_REGNUM_MASK
;
2247 const DWORD
*control_values
= NULL
;
2248 const local_constant
*constant
;
2250 shader_glsl_add_src_param(ins
, ins
->src
[1], ins
->src_addr
[1], WINED3DSP_WRITEMASK_ALL
, &src1_param
);
2252 /* Try to hardcode the loop control parameters if possible. Direct3D 9 class hardware doesn't support real
2253 * varying indexing, but Microsoft designed this feature for Shader model 2.x+. If the loop control is
2254 * known at compile time, the GLSL compiler can unroll the loop, and replace indirect addressing with direct
2257 if(regtype
== WINED3DSPR_CONSTINT
) {
2258 LIST_FOR_EACH_ENTRY(constant
, &shader
->baseShader
.constantsI
, local_constant
, entry
) {
2259 if(constant
->idx
== reg
) {
2260 control_values
= constant
->value
;
2266 if(control_values
) {
2267 if(control_values
[2] > 0) {
2268 shader_addline(ins
->buffer
, "for (aL%u = %d; aL%u < (%d * %d + %d); aL%u += %d) {\n",
2269 shader
->baseShader
.cur_loop_depth
, control_values
[1],
2270 shader
->baseShader
.cur_loop_depth
, control_values
[0], control_values
[2], control_values
[1],
2271 shader
->baseShader
.cur_loop_depth
, control_values
[2]);
2272 } else if(control_values
[2] == 0) {
2273 shader_addline(ins
->buffer
, "for (aL%u = %d, tmpInt%u = 0; tmpInt%u < %d; tmpInt%u++) {\n",
2274 shader
->baseShader
.cur_loop_depth
, control_values
[1], shader
->baseShader
.cur_loop_depth
,
2275 shader
->baseShader
.cur_loop_depth
, control_values
[0],
2276 shader
->baseShader
.cur_loop_depth
);
2278 shader_addline(ins
->buffer
, "for (aL%u = %d; aL%u > (%d * %d + %d); aL%u += %d) {\n",
2279 shader
->baseShader
.cur_loop_depth
, control_values
[1],
2280 shader
->baseShader
.cur_loop_depth
, control_values
[0], control_values
[2], control_values
[1],
2281 shader
->baseShader
.cur_loop_depth
, control_values
[2]);
2284 shader_addline(ins
->buffer
, "for (tmpInt%u = 0, aL%u = %s.y; tmpInt%u < %s.x; tmpInt%u++, aL%u += %s.z) {\n",
2285 shader
->baseShader
.cur_loop_depth
, shader
->baseShader
.cur_loop_regno
,
2286 src1_param
.reg_name
, shader
->baseShader
.cur_loop_depth
, src1_param
.reg_name
,
2287 shader
->baseShader
.cur_loop_depth
, shader
->baseShader
.cur_loop_regno
, src1_param
.reg_name
);
2290 shader
->baseShader
.cur_loop_depth
++;
2291 shader
->baseShader
.cur_loop_regno
++;
2294 static void shader_glsl_end(const struct wined3d_shader_instruction
*ins
)
2296 IWineD3DBaseShaderImpl
*shader
= (IWineD3DBaseShaderImpl
*)ins
->shader
;
2298 shader_addline(ins
->buffer
, "}\n");
2300 if (ins
->handler_idx
== WINED3DSIH_ENDLOOP
)
2302 shader
->baseShader
.cur_loop_depth
--;
2303 shader
->baseShader
.cur_loop_regno
--;
2306 if (ins
->handler_idx
== WINED3DSIH_ENDREP
)
2308 shader
->baseShader
.cur_loop_depth
--;
2312 static void shader_glsl_rep(const struct wined3d_shader_instruction
*ins
)
2314 IWineD3DBaseShaderImpl
*shader
= (IWineD3DBaseShaderImpl
*)ins
->shader
;
2315 glsl_src_param_t src0_param
;
2317 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2318 shader_addline(ins
->buffer
, "for (tmpInt%d = 0; tmpInt%d < %s; tmpInt%d++) {\n",
2319 shader
->baseShader
.cur_loop_depth
, shader
->baseShader
.cur_loop_depth
,
2320 src0_param
.param_str
, shader
->baseShader
.cur_loop_depth
);
2321 shader
->baseShader
.cur_loop_depth
++;
2324 static void shader_glsl_if(const struct wined3d_shader_instruction
*ins
)
2326 glsl_src_param_t src0_param
;
2328 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2329 shader_addline(ins
->buffer
, "if (%s) {\n", src0_param
.param_str
);
2332 static void shader_glsl_ifc(const struct wined3d_shader_instruction
*ins
)
2334 glsl_src_param_t src0_param
;
2335 glsl_src_param_t src1_param
;
2337 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2338 shader_glsl_add_src_param(ins
, ins
->src
[1], ins
->src_addr
[1], WINED3DSP_WRITEMASK_0
, &src1_param
);
2340 shader_addline(ins
->buffer
, "if (%s %s %s) {\n",
2341 src0_param
.param_str
, shader_get_comp_op(ins
->flags
), src1_param
.param_str
);
2344 static void shader_glsl_else(const struct wined3d_shader_instruction
*ins
)
2346 shader_addline(ins
->buffer
, "} else {\n");
2349 static void shader_glsl_break(const struct wined3d_shader_instruction
*ins
)
2351 shader_addline(ins
->buffer
, "break;\n");
2354 /* FIXME: According to MSDN the compare is done per component. */
2355 static void shader_glsl_breakc(const struct wined3d_shader_instruction
*ins
)
2357 glsl_src_param_t src0_param
;
2358 glsl_src_param_t src1_param
;
2360 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2361 shader_glsl_add_src_param(ins
, ins
->src
[1], ins
->src_addr
[1], WINED3DSP_WRITEMASK_0
, &src1_param
);
2363 shader_addline(ins
->buffer
, "if (%s %s %s) break;\n",
2364 src0_param
.param_str
, shader_get_comp_op(ins
->flags
), src1_param
.param_str
);
2367 static void shader_glsl_label(const struct wined3d_shader_instruction
*ins
)
2370 DWORD snum
= (ins
->src
[0]) & WINED3DSP_REGNUM_MASK
;
2371 shader_addline(ins
->buffer
, "}\n");
2372 shader_addline(ins
->buffer
, "void subroutine%u () {\n", snum
);
2375 static void shader_glsl_call(const struct wined3d_shader_instruction
*ins
)
2377 DWORD snum
= (ins
->src
[0]) & WINED3DSP_REGNUM_MASK
;
2378 shader_addline(ins
->buffer
, "subroutine%u();\n", snum
);
2381 static void shader_glsl_callnz(const struct wined3d_shader_instruction
*ins
)
2383 glsl_src_param_t src1_param
;
2385 DWORD snum
= (ins
->src
[0]) & WINED3DSP_REGNUM_MASK
;
2386 shader_glsl_add_src_param(ins
, ins
->src
[1], ins
->src_addr
[1], WINED3DSP_WRITEMASK_0
, &src1_param
);
2387 shader_addline(ins
->buffer
, "if (%s) subroutine%u();\n", src1_param
.param_str
, snum
);
2390 /*********************************************
2391 * Pixel Shader Specific Code begins here
2392 ********************************************/
2393 static void pshader_glsl_tex(const struct wined3d_shader_instruction
*ins
)
2395 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*)ins
->shader
;
2396 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
2397 DWORD shader_version
= ins
->reg_maps
->shader_version
;
2398 glsl_sample_function_t sample_function
;
2399 DWORD sample_flags
= 0;
2402 DWORD mask
= 0, swizzle
;
2404 /* 1.0-1.4: Use destination register as sampler source.
2405 * 2.0+: Use provided sampler source. */
2406 if (shader_version
< WINED3DPS_VERSION(2,0)) sampler_idx
= ins
->dst
[0].register_idx
;
2407 else sampler_idx
= ins
->src
[1] & WINED3DSP_REGNUM_MASK
;
2408 sampler_type
= ins
->reg_maps
->samplers
[sampler_idx
] & WINED3DSP_TEXTURETYPE_MASK
;
2410 if (shader_version
< WINED3DPS_VERSION(1,4))
2412 DWORD flags
= deviceImpl
->stateBlock
->textureState
[sampler_idx
][WINED3DTSS_TEXTURETRANSFORMFLAGS
];
2414 /* Projected cube textures don't make a lot of sense, the resulting coordinates stay the same. */
2415 if (flags
& WINED3DTTFF_PROJECTED
&& sampler_type
!= WINED3DSTT_CUBE
) {
2416 sample_flags
|= WINED3D_GLSL_SAMPLE_PROJECTED
;
2417 switch (flags
& ~WINED3DTTFF_PROJECTED
) {
2418 case WINED3DTTFF_COUNT1
: FIXME("WINED3DTTFF_PROJECTED with WINED3DTTFF_COUNT1?\n"); break;
2419 case WINED3DTTFF_COUNT2
: mask
= WINED3DSP_WRITEMASK_1
; break;
2420 case WINED3DTTFF_COUNT3
: mask
= WINED3DSP_WRITEMASK_2
; break;
2421 case WINED3DTTFF_COUNT4
:
2422 case WINED3DTTFF_DISABLE
: mask
= WINED3DSP_WRITEMASK_3
; break;
2426 else if (shader_version
< WINED3DPS_VERSION(2,0))
2428 DWORD src_mod
= ins
->src
[0] & WINED3DSP_SRCMOD_MASK
;
2430 if (src_mod
== WINED3DSPSM_DZ
) {
2431 sample_flags
|= WINED3D_GLSL_SAMPLE_PROJECTED
;
2432 mask
= WINED3DSP_WRITEMASK_2
;
2433 } else if (src_mod
== WINED3DSPSM_DW
) {
2434 sample_flags
|= WINED3D_GLSL_SAMPLE_PROJECTED
;
2435 mask
= WINED3DSP_WRITEMASK_3
;
2438 if (ins
->flags
& WINED3DSI_TEXLD_PROJECT
)
2440 /* ps 2.0 texldp instruction always divides by the fourth component. */
2441 sample_flags
|= WINED3D_GLSL_SAMPLE_PROJECTED
;
2442 mask
= WINED3DSP_WRITEMASK_3
;
2446 if(deviceImpl
->stateBlock
->textures
[sampler_idx
] &&
2447 IWineD3DBaseTexture_GetTextureDimensions(deviceImpl
->stateBlock
->textures
[sampler_idx
]) == GL_TEXTURE_RECTANGLE_ARB
) {
2448 sample_flags
|= WINED3D_GLSL_SAMPLE_RECT
;
2451 shader_glsl_get_sample_function(sampler_type
, sample_flags
, &sample_function
);
2452 mask
|= sample_function
.coord_mask
;
2454 if (shader_version
< WINED3DPS_VERSION(2,0)) swizzle
= WINED3DVS_NOSWIZZLE
;
2455 else swizzle
= ins
->src
[1] & WINED3DSP_SWIZZLE_MASK
;
2457 /* 1.0-1.3: Use destination register as coordinate source.
2458 1.4+: Use provided coordinate source register. */
2459 if (shader_version
< WINED3DPS_VERSION(1,4))
2462 shader_glsl_get_write_mask(mask
, coord_mask
);
2463 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, swizzle
, NULL
,
2464 "T%u%s", sampler_idx
, coord_mask
);
2466 glsl_src_param_t coord_param
;
2467 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], mask
, &coord_param
);
2468 if (ins
->flags
& WINED3DSI_TEXLD_BIAS
)
2470 glsl_src_param_t bias
;
2471 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], WINED3DSP_WRITEMASK_3
, &bias
);
2472 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, swizzle
, bias
.param_str
,
2473 "%s", coord_param
.param_str
);
2475 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, swizzle
, NULL
,
2476 "%s", coord_param
.param_str
);
2481 static void shader_glsl_texldl(const struct wined3d_shader_instruction
*ins
)
2483 IWineD3DBaseShaderImpl
*This
= (IWineD3DBaseShaderImpl
*)ins
->shader
;
2484 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
2485 glsl_sample_function_t sample_function
;
2486 glsl_src_param_t coord_param
, lod_param
;
2487 DWORD sample_flags
= WINED3D_GLSL_SAMPLE_LOD
;
2490 DWORD swizzle
= ins
->src
[1] & WINED3DSP_SWIZZLE_MASK
;
2492 sampler_idx
= ins
->src
[1] & WINED3DSP_REGNUM_MASK
;
2493 sampler_type
= ins
->reg_maps
->samplers
[sampler_idx
] & WINED3DSP_TEXTURETYPE_MASK
;
2494 if(deviceImpl
->stateBlock
->textures
[sampler_idx
] &&
2495 IWineD3DBaseTexture_GetTextureDimensions(deviceImpl
->stateBlock
->textures
[sampler_idx
]) == GL_TEXTURE_RECTANGLE_ARB
) {
2496 sample_flags
|= WINED3D_GLSL_SAMPLE_RECT
;
2498 shader_glsl_get_sample_function(sampler_type
, sample_flags
, &sample_function
);
2499 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], sample_function
.coord_mask
, &coord_param
);
2501 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], WINED3DSP_WRITEMASK_3
, &lod_param
);
2503 if (shader_is_pshader_version(ins
->reg_maps
->shader_version
))
2505 /* The GLSL spec claims the Lod sampling functions are only supported in vertex shaders.
2506 * However, they seem to work just fine in fragment shaders as well. */
2507 WARN("Using %s in fragment shader.\n", sample_function
.name
);
2509 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, swizzle
, lod_param
.param_str
,
2510 "%s", coord_param
.param_str
);
2513 static void pshader_glsl_texcoord(const struct wined3d_shader_instruction
*ins
)
2515 /* FIXME: Make this work for more than just 2D textures */
2516 SHADER_BUFFER
*buffer
= ins
->buffer
;
2520 write_mask
= shader_glsl_append_dst(ins
->buffer
, ins
);
2521 shader_glsl_get_write_mask(write_mask
, dst_mask
);
2523 if (ins
->reg_maps
->shader_version
!= WINED3DPS_VERSION(1,4))
2525 DWORD reg
= ins
->dst
[0].register_idx
;
2526 shader_addline(buffer
, "clamp(gl_TexCoord[%u], 0.0, 1.0)%s);\n", reg
, dst_mask
);
2528 DWORD reg
= ins
->src
[0] & WINED3DSP_REGNUM_MASK
;
2529 DWORD src_mod
= ins
->src
[0] & WINED3DSP_SRCMOD_MASK
;
2530 char dst_swizzle
[6];
2532 shader_glsl_get_swizzle(ins
->src
[0], FALSE
, write_mask
, dst_swizzle
);
2534 if (src_mod
== WINED3DSPSM_DZ
) {
2535 glsl_src_param_t div_param
;
2536 unsigned int mask_size
= shader_glsl_get_write_mask_size(write_mask
);
2537 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], WINED3DSP_WRITEMASK_2
, &div_param
);
2539 if (mask_size
> 1) {
2540 shader_addline(buffer
, "gl_TexCoord[%u]%s / vec%d(%s));\n", reg
, dst_swizzle
, mask_size
, div_param
.param_str
);
2542 shader_addline(buffer
, "gl_TexCoord[%u]%s / %s);\n", reg
, dst_swizzle
, div_param
.param_str
);
2544 } else if (src_mod
== WINED3DSPSM_DW
) {
2545 glsl_src_param_t div_param
;
2546 unsigned int mask_size
= shader_glsl_get_write_mask_size(write_mask
);
2547 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], WINED3DSP_WRITEMASK_3
, &div_param
);
2549 if (mask_size
> 1) {
2550 shader_addline(buffer
, "gl_TexCoord[%u]%s / vec%d(%s));\n", reg
, dst_swizzle
, mask_size
, div_param
.param_str
);
2552 shader_addline(buffer
, "gl_TexCoord[%u]%s / %s);\n", reg
, dst_swizzle
, div_param
.param_str
);
2555 shader_addline(buffer
, "gl_TexCoord[%u]%s);\n", reg
, dst_swizzle
);
2560 /** Process the WINED3DSIO_TEXDP3TEX instruction in GLSL:
2561 * Take a 3-component dot product of the TexCoord[dstreg] and src,
2562 * then perform a 1D texture lookup from stage dstregnum, place into dst. */
2563 static void pshader_glsl_texdp3tex(const struct wined3d_shader_instruction
*ins
)
2565 glsl_src_param_t src0_param
;
2566 glsl_sample_function_t sample_function
;
2567 DWORD sampler_idx
= ins
->dst
[0].register_idx
;
2568 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2569 DWORD sampler_type
= ins
->reg_maps
->samplers
[sampler_idx
] & WINED3DSP_TEXTURETYPE_MASK
;
2572 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], src_mask
, &src0_param
);
2574 /* Do I have to take care about the projected bit? I don't think so, since the dp3 returns only one
2575 * scalar, and projected sampling would require 4.
2577 * It is a dependent read - not valid with conditional NP2 textures
2579 shader_glsl_get_sample_function(sampler_type
, 0, &sample_function
);
2580 mask_size
= shader_glsl_get_write_mask_size(sample_function
.coord_mask
);
2585 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, WINED3DVS_NOSWIZZLE
, NULL
,
2586 "dot(gl_TexCoord[%u].xyz, %s)", sampler_idx
, src0_param
.param_str
);
2590 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, WINED3DVS_NOSWIZZLE
, NULL
,
2591 "vec2(dot(gl_TexCoord[%u].xyz, %s), 0.0)", sampler_idx
, src0_param
.param_str
);
2595 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, WINED3DVS_NOSWIZZLE
, NULL
,
2596 "vec3(dot(gl_TexCoord[%u].xyz, %s), 0.0, 0.0)", sampler_idx
, src0_param
.param_str
);
2600 FIXME("Unexpected mask size %u\n", mask_size
);
2605 /** Process the WINED3DSIO_TEXDP3 instruction in GLSL:
2606 * Take a 3-component dot product of the TexCoord[dstreg] and src. */
2607 static void pshader_glsl_texdp3(const struct wined3d_shader_instruction
*ins
)
2609 glsl_src_param_t src0_param
;
2610 DWORD dstreg
= ins
->dst
[0].register_idx
;
2611 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2613 unsigned int mask_size
;
2615 dst_mask
= shader_glsl_append_dst(ins
->buffer
, ins
);
2616 mask_size
= shader_glsl_get_write_mask_size(dst_mask
);
2617 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], src_mask
, &src0_param
);
2619 if (mask_size
> 1) {
2620 shader_addline(ins
->buffer
, "vec%d(dot(T%u.xyz, %s)));\n", mask_size
, dstreg
, src0_param
.param_str
);
2622 shader_addline(ins
->buffer
, "dot(T%u.xyz, %s));\n", dstreg
, src0_param
.param_str
);
2626 /** Process the WINED3DSIO_TEXDEPTH instruction in GLSL:
2627 * Calculate the depth as dst.x / dst.y */
2628 static void pshader_glsl_texdepth(const struct wined3d_shader_instruction
*ins
)
2630 glsl_dst_param_t dst_param
;
2632 shader_glsl_add_dst_param(ins
, &ins
->dst
[0], &dst_param
);
2634 /* Tests show that texdepth never returns anything below 0.0, and that r5.y is clamped to 1.0.
2635 * Negative input is accepted, -0.25 / -0.5 returns 0.5. GL should clamp gl_FragDepth to [0;1], but
2636 * this doesn't always work, so clamp the results manually. Whether or not the x value is clamped at 1
2637 * too is irrelevant, since if x = 0, any y value < 1.0 (and > 1.0 is not allowed) results in a result
2640 shader_addline(ins
->buffer
, "gl_FragDepth = clamp((%s.x / min(%s.y, 1.0)), 0.0, 1.0);\n",
2641 dst_param
.reg_name
, dst_param
.reg_name
);
2644 /** Process the WINED3DSIO_TEXM3X2DEPTH instruction in GLSL:
2645 * Last row of a 3x2 matrix multiply, use the result to calculate the depth:
2646 * Calculate tmp0.y = TexCoord[dstreg] . src.xyz; (tmp0.x has already been calculated)
2647 * depth = (tmp0.y == 0.0) ? 1.0 : tmp0.x / tmp0.y
2649 static void pshader_glsl_texm3x2depth(const struct wined3d_shader_instruction
*ins
)
2651 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2652 DWORD dstreg
= ins
->dst
[0].register_idx
;
2653 glsl_src_param_t src0_param
;
2655 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], src_mask
, &src0_param
);
2657 shader_addline(ins
->buffer
, "tmp0.y = dot(T%u.xyz, %s);\n", dstreg
, src0_param
.param_str
);
2658 shader_addline(ins
->buffer
, "gl_FragDepth = (tmp0.y == 0.0) ? 1.0 : clamp(tmp0.x / tmp0.y, 0.0, 1.0);\n");
2661 /** Process the WINED3DSIO_TEXM3X2PAD instruction in GLSL
2662 * Calculate the 1st of a 2-row matrix multiplication. */
2663 static void pshader_glsl_texm3x2pad(const struct wined3d_shader_instruction
*ins
)
2665 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2666 DWORD reg
= ins
->dst
[0].register_idx
;
2667 SHADER_BUFFER
*buffer
= ins
->buffer
;
2668 glsl_src_param_t src0_param
;
2670 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], src_mask
, &src0_param
);
2671 shader_addline(buffer
, "tmp0.x = dot(T%u.xyz, %s);\n", reg
, src0_param
.param_str
);
2674 /** Process the WINED3DSIO_TEXM3X3PAD instruction in GLSL
2675 * Calculate the 1st or 2nd row of a 3-row matrix multiplication. */
2676 static void pshader_glsl_texm3x3pad(const struct wined3d_shader_instruction
*ins
)
2678 IWineD3DPixelShaderImpl
*shader
= (IWineD3DPixelShaderImpl
*)ins
->shader
;
2679 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2680 DWORD reg
= ins
->dst
[0].register_idx
;
2681 SHADER_BUFFER
*buffer
= ins
->buffer
;
2682 SHADER_PARSE_STATE
* current_state
= &shader
->baseShader
.parse_state
;
2683 glsl_src_param_t src0_param
;
2685 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], src_mask
, &src0_param
);
2686 shader_addline(buffer
, "tmp0.%c = dot(T%u.xyz, %s);\n", 'x' + current_state
->current_row
, reg
, src0_param
.param_str
);
2687 current_state
->texcoord_w
[current_state
->current_row
++] = reg
;
2690 static void pshader_glsl_texm3x2tex(const struct wined3d_shader_instruction
*ins
)
2692 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2693 DWORD reg
= ins
->dst
[0].register_idx
;
2694 SHADER_BUFFER
*buffer
= ins
->buffer
;
2695 glsl_src_param_t src0_param
;
2696 DWORD sampler_type
= ins
->reg_maps
->samplers
[reg
] & WINED3DSP_TEXTURETYPE_MASK
;
2697 glsl_sample_function_t sample_function
;
2699 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], src_mask
, &src0_param
);
2700 shader_addline(buffer
, "tmp0.y = dot(T%u.xyz, %s);\n", reg
, src0_param
.param_str
);
2702 shader_glsl_get_sample_function(sampler_type
, 0, &sample_function
);
2704 /* Sample the texture using the calculated coordinates */
2705 shader_glsl_gen_sample_code(ins
, reg
, &sample_function
, WINED3DVS_NOSWIZZLE
, NULL
, "tmp0.xy");
2708 /** Process the WINED3DSIO_TEXM3X3TEX instruction in GLSL
2709 * Perform the 3rd row of a 3x3 matrix multiply, then sample the texture using the calculated coordinates */
2710 static void pshader_glsl_texm3x3tex(const struct wined3d_shader_instruction
*ins
)
2712 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2713 glsl_src_param_t src0_param
;
2714 DWORD reg
= ins
->dst
[0].register_idx
;
2715 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*)ins
->shader
;
2716 SHADER_PARSE_STATE
* current_state
= &This
->baseShader
.parse_state
;
2717 DWORD sampler_type
= ins
->reg_maps
->samplers
[reg
] & WINED3DSP_TEXTURETYPE_MASK
;
2718 glsl_sample_function_t sample_function
;
2720 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], src_mask
, &src0_param
);
2721 shader_addline(ins
->buffer
, "tmp0.z = dot(T%u.xyz, %s);\n", reg
, src0_param
.param_str
);
2723 /* Dependent read, not valid with conditional NP2 */
2724 shader_glsl_get_sample_function(sampler_type
, 0, &sample_function
);
2726 /* Sample the texture using the calculated coordinates */
2727 shader_glsl_gen_sample_code(ins
, reg
, &sample_function
, WINED3DVS_NOSWIZZLE
, NULL
, "tmp0.xyz");
2729 current_state
->current_row
= 0;
2732 /** Process the WINED3DSIO_TEXM3X3 instruction in GLSL
2733 * Perform the 3rd row of a 3x3 matrix multiply */
2734 static void pshader_glsl_texm3x3(const struct wined3d_shader_instruction
*ins
)
2736 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2737 glsl_src_param_t src0_param
;
2739 DWORD reg
= ins
->dst
[0].register_idx
;
2740 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*)ins
->shader
;
2741 SHADER_PARSE_STATE
* current_state
= &This
->baseShader
.parse_state
;
2743 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], src_mask
, &src0_param
);
2745 shader_glsl_append_dst(ins
->buffer
, ins
);
2746 shader_glsl_get_write_mask(ins
->dst
[0].token
, dst_mask
);
2747 shader_addline(ins
->buffer
, "vec4(tmp0.xy, dot(T%u.xyz, %s), 1.0)%s);\n", reg
, src0_param
.param_str
, dst_mask
);
2749 current_state
->current_row
= 0;
2752 /** Process the WINED3DSIO_TEXM3X3SPEC instruction in GLSL
2753 * Perform the final texture lookup based on the previous 2 3x3 matrix multiplies */
2754 static void pshader_glsl_texm3x3spec(const struct wined3d_shader_instruction
*ins
)
2756 IWineD3DPixelShaderImpl
*shader
= (IWineD3DPixelShaderImpl
*)ins
->shader
;
2757 DWORD reg
= ins
->dst
[0].register_idx
;
2758 glsl_src_param_t src0_param
;
2759 glsl_src_param_t src1_param
;
2760 SHADER_BUFFER
*buffer
= ins
->buffer
;
2761 SHADER_PARSE_STATE
* current_state
= &shader
->baseShader
.parse_state
;
2762 DWORD stype
= ins
->reg_maps
->samplers
[reg
] & WINED3DSP_TEXTURETYPE_MASK
;
2763 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2764 glsl_sample_function_t sample_function
;
2766 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], src_mask
, &src0_param
);
2767 shader_glsl_add_src_param(ins
, ins
->src
[1], ins
->src_addr
[1], src_mask
, &src1_param
);
2769 /* Perform the last matrix multiply operation */
2770 shader_addline(buffer
, "tmp0.z = dot(T%u.xyz, %s);\n", reg
, src0_param
.param_str
);
2771 /* Reflection calculation */
2772 shader_addline(buffer
, "tmp0.xyz = -reflect((%s), normalize(tmp0.xyz));\n", src1_param
.param_str
);
2774 /* Dependent read, not valid with conditional NP2 */
2775 shader_glsl_get_sample_function(stype
, 0, &sample_function
);
2777 /* Sample the texture */
2778 shader_glsl_gen_sample_code(ins
, reg
, &sample_function
, WINED3DVS_NOSWIZZLE
, NULL
, "tmp0.xyz");
2780 current_state
->current_row
= 0;
2783 /** Process the WINED3DSIO_TEXM3X3VSPEC instruction in GLSL
2784 * Perform the final texture lookup based on the previous 2 3x3 matrix multiplies */
2785 static void pshader_glsl_texm3x3vspec(const struct wined3d_shader_instruction
*ins
)
2787 IWineD3DPixelShaderImpl
*shader
= (IWineD3DPixelShaderImpl
*)ins
->shader
;
2788 DWORD reg
= ins
->dst
[0].register_idx
;
2789 SHADER_BUFFER
*buffer
= ins
->buffer
;
2790 SHADER_PARSE_STATE
* current_state
= &shader
->baseShader
.parse_state
;
2791 glsl_src_param_t src0_param
;
2792 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2793 DWORD sampler_type
= ins
->reg_maps
->samplers
[reg
] & WINED3DSP_TEXTURETYPE_MASK
;
2794 glsl_sample_function_t sample_function
;
2796 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], src_mask
, &src0_param
);
2798 /* Perform the last matrix multiply operation */
2799 shader_addline(buffer
, "tmp0.z = dot(vec3(T%u), vec3(%s));\n", reg
, src0_param
.param_str
);
2801 /* Construct the eye-ray vector from w coordinates */
2802 shader_addline(buffer
, "tmp1.xyz = normalize(vec3(gl_TexCoord[%u].w, gl_TexCoord[%u].w, gl_TexCoord[%u].w));\n",
2803 current_state
->texcoord_w
[0], current_state
->texcoord_w
[1], reg
);
2804 shader_addline(buffer
, "tmp0.xyz = -reflect(tmp1.xyz, normalize(tmp0.xyz));\n");
2806 /* Dependent read, not valid with conditional NP2 */
2807 shader_glsl_get_sample_function(sampler_type
, 0, &sample_function
);
2809 /* Sample the texture using the calculated coordinates */
2810 shader_glsl_gen_sample_code(ins
, reg
, &sample_function
, WINED3DVS_NOSWIZZLE
, NULL
, "tmp0.xyz");
2812 current_state
->current_row
= 0;
2815 /** Process the WINED3DSIO_TEXBEM instruction in GLSL.
2816 * Apply a fake bump map transform.
2817 * texbem is pshader <= 1.3 only, this saves a few version checks
2819 static void pshader_glsl_texbem(const struct wined3d_shader_instruction
*ins
)
2821 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*)ins
->shader
;
2822 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
2823 glsl_sample_function_t sample_function
;
2824 glsl_src_param_t coord_param
;
2831 sampler_idx
= ins
->dst
[0].register_idx
;
2832 flags
= deviceImpl
->stateBlock
->textureState
[sampler_idx
][WINED3DTSS_TEXTURETRANSFORMFLAGS
];
2834 sampler_type
= ins
->reg_maps
->samplers
[sampler_idx
] & WINED3DSP_TEXTURETYPE_MASK
;
2835 /* Dependent read, not valid with conditional NP2 */
2836 shader_glsl_get_sample_function(sampler_type
, 0, &sample_function
);
2837 mask
= sample_function
.coord_mask
;
2839 shader_glsl_get_write_mask(mask
, coord_mask
);
2841 /* with projective textures, texbem only divides the static texture coord, not the displacement,
2842 * so we can't let the GL handle this.
2844 if (flags
& WINED3DTTFF_PROJECTED
) {
2846 char coord_div_mask
[3];
2847 switch (flags
& ~WINED3DTTFF_PROJECTED
) {
2848 case WINED3DTTFF_COUNT1
: FIXME("WINED3DTTFF_PROJECTED with WINED3DTTFF_COUNT1?\n"); break;
2849 case WINED3DTTFF_COUNT2
: div_mask
= WINED3DSP_WRITEMASK_1
; break;
2850 case WINED3DTTFF_COUNT3
: div_mask
= WINED3DSP_WRITEMASK_2
; break;
2851 case WINED3DTTFF_COUNT4
:
2852 case WINED3DTTFF_DISABLE
: div_mask
= WINED3DSP_WRITEMASK_3
; break;
2854 shader_glsl_get_write_mask(div_mask
, coord_div_mask
);
2855 shader_addline(ins
->buffer
, "T%u%s /= T%u%s;\n", sampler_idx
, coord_mask
, sampler_idx
, coord_div_mask
);
2858 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0],
2859 WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
, &coord_param
);
2861 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, WINED3DVS_NOSWIZZLE
, NULL
,
2862 "T%u%s + vec4(bumpenvmat%d * %s, 0.0, 0.0)%s", sampler_idx
, coord_mask
, sampler_idx
,
2863 coord_param
.param_str
, coord_mask
);
2865 if (ins
->handler_idx
== WINED3DSIH_TEXBEML
)
2867 glsl_src_param_t luminance_param
;
2868 glsl_dst_param_t dst_param
;
2870 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], WINED3DSP_WRITEMASK_2
, &luminance_param
);
2871 shader_glsl_add_dst_param(ins
, &ins
->dst
[0], &dst_param
);
2873 shader_addline(ins
->buffer
, "%s%s *= (%s * luminancescale%d + luminanceoffset%d);\n",
2874 dst_param
.reg_name
, dst_param
.mask_str
,
2875 luminance_param
.param_str
, sampler_idx
, sampler_idx
);
2879 static void pshader_glsl_bem(const struct wined3d_shader_instruction
*ins
)
2881 glsl_src_param_t src0_param
, src1_param
;
2882 DWORD sampler_idx
= ins
->dst
[0].register_idx
;
2884 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0],
2885 WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
, &src0_param
);
2886 shader_glsl_add_src_param(ins
, ins
->src
[1], ins
->src_addr
[1],
2887 WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
, &src1_param
);
2889 shader_glsl_append_dst(ins
->buffer
, ins
);
2890 shader_addline(ins
->buffer
, "%s + bumpenvmat%d * %s);\n",
2891 src0_param
.param_str
, sampler_idx
, src1_param
.param_str
);
2894 /** Process the WINED3DSIO_TEXREG2AR instruction in GLSL
2895 * Sample 2D texture at dst using the alpha & red (wx) components of src as texture coordinates */
2896 static void pshader_glsl_texreg2ar(const struct wined3d_shader_instruction
*ins
)
2898 glsl_src_param_t src0_param
;
2899 DWORD sampler_idx
= ins
->dst
[0].register_idx
;
2900 DWORD sampler_type
= ins
->reg_maps
->samplers
[sampler_idx
] & WINED3DSP_TEXTURETYPE_MASK
;
2901 glsl_sample_function_t sample_function
;
2903 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], WINED3DSP_WRITEMASK_ALL
, &src0_param
);
2905 shader_glsl_get_sample_function(sampler_type
, 0, &sample_function
);
2906 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, WINED3DVS_NOSWIZZLE
, NULL
,
2907 "%s.wx", src0_param
.reg_name
);
2910 /** Process the WINED3DSIO_TEXREG2GB instruction in GLSL
2911 * Sample 2D texture at dst using the green & blue (yz) components of src as texture coordinates */
2912 static void pshader_glsl_texreg2gb(const struct wined3d_shader_instruction
*ins
)
2914 glsl_src_param_t src0_param
;
2915 DWORD sampler_idx
= ins
->dst
[0].register_idx
;
2916 DWORD sampler_type
= ins
->reg_maps
->samplers
[sampler_idx
] & WINED3DSP_TEXTURETYPE_MASK
;
2917 glsl_sample_function_t sample_function
;
2919 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], WINED3DSP_WRITEMASK_ALL
, &src0_param
);
2921 shader_glsl_get_sample_function(sampler_type
, 0, &sample_function
);
2922 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, WINED3DVS_NOSWIZZLE
, NULL
,
2923 "%s.yz", src0_param
.reg_name
);
2926 /** Process the WINED3DSIO_TEXREG2RGB instruction in GLSL
2927 * Sample texture at dst using the rgb (xyz) components of src as texture coordinates */
2928 static void pshader_glsl_texreg2rgb(const struct wined3d_shader_instruction
*ins
)
2930 glsl_src_param_t src0_param
;
2931 DWORD sampler_idx
= ins
->dst
[0].register_idx
;
2932 DWORD sampler_type
= ins
->reg_maps
->samplers
[sampler_idx
] & WINED3DSP_TEXTURETYPE_MASK
;
2933 glsl_sample_function_t sample_function
;
2935 /* Dependent read, not valid with conditional NP2 */
2936 shader_glsl_get_sample_function(sampler_type
, 0, &sample_function
);
2937 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], sample_function
.coord_mask
, &src0_param
);
2939 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, WINED3DVS_NOSWIZZLE
, NULL
,
2940 "%s", src0_param
.param_str
);
2943 /** Process the WINED3DSIO_TEXKILL instruction in GLSL.
2944 * If any of the first 3 components are < 0, discard this pixel */
2945 static void pshader_glsl_texkill(const struct wined3d_shader_instruction
*ins
)
2947 glsl_dst_param_t dst_param
;
2949 /* The argument is a destination parameter, and no writemasks are allowed */
2950 shader_glsl_add_dst_param(ins
, &ins
->dst
[0], &dst_param
);
2951 if ((ins
->reg_maps
->shader_version
>= WINED3DPS_VERSION(2,0)))
2953 /* 2.0 shaders compare all 4 components in texkill */
2954 shader_addline(ins
->buffer
, "if (any(lessThan(%s.xyzw, vec4(0.0)))) discard;\n", dst_param
.reg_name
);
2956 /* 1.X shaders only compare the first 3 components, probably due to the nature of the texkill
2957 * instruction as a tex* instruction, and phase, which kills all a / w components. Even if all
2958 * 4 components are defined, only the first 3 are used
2960 shader_addline(ins
->buffer
, "if (any(lessThan(%s.xyz, vec3(0.0)))) discard;\n", dst_param
.reg_name
);
2964 /** Process the WINED3DSIO_DP2ADD instruction in GLSL.
2965 * dst = dot2(src0, src1) + src2 */
2966 static void pshader_glsl_dp2add(const struct wined3d_shader_instruction
*ins
)
2968 glsl_src_param_t src0_param
;
2969 glsl_src_param_t src1_param
;
2970 glsl_src_param_t src2_param
;
2972 unsigned int mask_size
;
2974 write_mask
= shader_glsl_append_dst(ins
->buffer
, ins
);
2975 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
2977 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0],
2978 WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
, &src0_param
);
2979 shader_glsl_add_src_param(ins
, ins
->src
[1], ins
->src_addr
[1],
2980 WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
, &src1_param
);
2981 shader_glsl_add_src_param(ins
, ins
->src
[2], ins
->src_addr
[2],
2982 WINED3DSP_WRITEMASK_0
, &src2_param
);
2984 if (mask_size
> 1) {
2985 shader_addline(ins
->buffer
, "vec%d(dot(%s, %s) + %s));\n",
2986 mask_size
, src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
2988 shader_addline(ins
->buffer
, "dot(%s, %s) + %s);\n",
2989 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
2993 static void pshader_glsl_input_pack(SHADER_BUFFER
* buffer
, const struct semantic
* semantics_in
,
2994 IWineD3DPixelShader
*iface
, enum vertexprocessing_mode vertexprocessing
)
2997 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*) iface
;
2999 for (i
= 0; i
< MAX_REG_INPUT
; i
++) {
3001 DWORD usage_token
= semantics_in
[i
].usage
;
3002 DWORD register_token
= semantics_in
[i
].reg
;
3003 DWORD usage
, usage_idx
;
3007 if (!usage_token
) continue;
3008 usage
= (usage_token
& WINED3DSP_DCL_USAGE_MASK
) >> WINED3DSP_DCL_USAGE_SHIFT
;
3009 usage_idx
= (usage_token
& WINED3DSP_DCL_USAGEINDEX_MASK
) >> WINED3DSP_DCL_USAGEINDEX_SHIFT
;
3010 shader_glsl_get_write_mask(register_token
, reg_mask
);
3014 case WINED3DDECLUSAGE_TEXCOORD
:
3015 if(usage_idx
< 8 && vertexprocessing
== pretransformed
) {
3016 shader_addline(buffer
, "IN[%u]%s = gl_TexCoord[%u]%s;\n",
3017 This
->input_reg_map
[i
], reg_mask
, usage_idx
, reg_mask
);
3019 shader_addline(buffer
, "IN[%u]%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
3020 This
->input_reg_map
[i
], reg_mask
, reg_mask
);
3024 case WINED3DDECLUSAGE_COLOR
:
3026 shader_addline(buffer
, "IN[%u]%s = vec4(gl_Color)%s;\n",
3027 This
->input_reg_map
[i
], reg_mask
, reg_mask
);
3028 else if (usage_idx
== 1)
3029 shader_addline(buffer
, "IN[%u]%s = vec4(gl_SecondaryColor)%s;\n",
3030 This
->input_reg_map
[i
], reg_mask
, reg_mask
);
3032 shader_addline(buffer
, "IN[%u]%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
3033 This
->input_reg_map
[i
], reg_mask
, reg_mask
);
3037 shader_addline(buffer
, "IN[%u]%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
3038 This
->input_reg_map
[i
], reg_mask
, reg_mask
);
3043 /*********************************************
3044 * Vertex Shader Specific Code begins here
3045 ********************************************/
3047 static void add_glsl_program_entry(struct shader_glsl_priv
*priv
, struct glsl_shader_prog_link
*entry
) {
3048 glsl_program_key_t
*key
;
3050 key
= HeapAlloc(GetProcessHeap(), 0, sizeof(glsl_program_key_t
));
3051 key
->vshader
= entry
->vshader
;
3052 key
->pshader
= entry
->pshader
;
3053 key
->vs_args
= entry
->vs_args
;
3054 key
->ps_args
= entry
->ps_args
;
3056 hash_table_put(priv
->glsl_program_lookup
, key
, entry
);
3059 static struct glsl_shader_prog_link
*get_glsl_program_entry(struct shader_glsl_priv
*priv
,
3060 IWineD3DVertexShader
*vshader
, IWineD3DPixelShader
*pshader
, struct vs_compile_args
*vs_args
,
3061 struct ps_compile_args
*ps_args
) {
3062 glsl_program_key_t key
;
3064 key
.vshader
= vshader
;
3065 key
.pshader
= pshader
;
3066 key
.vs_args
= *vs_args
;
3067 key
.ps_args
= *ps_args
;
3069 return hash_table_get(priv
->glsl_program_lookup
, &key
);
3072 static void delete_glsl_program_entry(struct shader_glsl_priv
*priv
, const WineD3D_GL_Info
*gl_info
,
3073 struct glsl_shader_prog_link
*entry
)
3075 glsl_program_key_t
*key
;
3077 key
= HeapAlloc(GetProcessHeap(), 0, sizeof(glsl_program_key_t
));
3078 key
->vshader
= entry
->vshader
;
3079 key
->pshader
= entry
->pshader
;
3080 key
->vs_args
= entry
->vs_args
;
3081 key
->ps_args
= entry
->ps_args
;
3082 hash_table_remove(priv
->glsl_program_lookup
, key
);
3084 GL_EXTCALL(glDeleteObjectARB(entry
->programId
));
3085 if (entry
->vshader
) list_remove(&entry
->vshader_entry
);
3086 if (entry
->pshader
) list_remove(&entry
->pshader_entry
);
3087 HeapFree(GetProcessHeap(), 0, entry
->vuniformF_locations
);
3088 HeapFree(GetProcessHeap(), 0, entry
->puniformF_locations
);
3089 HeapFree(GetProcessHeap(), 0, entry
);
3092 static void handle_ps3_input(SHADER_BUFFER
*buffer
, const struct semantic
*semantics_in
,
3093 const struct semantic
*semantics_out
, const WineD3D_GL_Info
*gl_info
, const DWORD
*map
)
3096 DWORD usage_token
, usage_token_out
;
3097 DWORD register_token
, register_token_out
;
3098 DWORD usage
, usage_idx
, usage_out
, usage_idx_out
;
3101 DWORD in_count
= GL_LIMITS(glsl_varyings
) / 4;
3102 char reg_mask
[6], reg_mask_out
[6];
3103 char destination
[50];
3105 set
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*set
) * (in_count
+ 2));
3107 if (!semantics_out
) {
3108 /* Save gl_FrontColor & gl_FrontSecondaryColor before overwriting them. */
3109 shader_addline(buffer
, "vec4 front_color = gl_FrontColor;\n");
3110 shader_addline(buffer
, "vec4 front_secondary_color = gl_FrontSecondaryColor;\n");
3113 for(i
= 0; i
< MAX_REG_INPUT
; i
++) {
3114 usage_token
= semantics_in
[i
].usage
;
3115 if (!usage_token
) continue;
3118 if (in_idx
>= (in_count
+ 2)) {
3119 FIXME("More input varyings declared than supported, expect issues\n");
3122 else if (map
[i
] == ~0U)
3124 /* Declared, but not read register */
3128 if (in_idx
== in_count
) {
3129 sprintf(destination
, "gl_FrontColor");
3130 } else if (in_idx
== in_count
+ 1) {
3131 sprintf(destination
, "gl_FrontSecondaryColor");
3133 sprintf(destination
, "IN[%u]", in_idx
);
3136 register_token
= semantics_in
[i
].reg
;
3138 usage
= (usage_token
& WINED3DSP_DCL_USAGE_MASK
) >> WINED3DSP_DCL_USAGE_SHIFT
;
3139 usage_idx
= (usage_token
& WINED3DSP_DCL_USAGEINDEX_MASK
) >> WINED3DSP_DCL_USAGEINDEX_SHIFT
;
3140 set
[map
[i
]] = shader_glsl_get_write_mask(register_token
, reg_mask
);
3142 if(!semantics_out
) {
3144 case WINED3DDECLUSAGE_COLOR
:
3146 shader_addline(buffer
, "%s%s = front_color%s;\n",
3147 destination
, reg_mask
, reg_mask
);
3148 else if (usage_idx
== 1)
3149 shader_addline(buffer
, "%s%s = front_secondary_color%s;\n",
3150 destination
, reg_mask
, reg_mask
);
3152 shader_addline(buffer
, "%s%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
3153 destination
, reg_mask
, reg_mask
);
3156 case WINED3DDECLUSAGE_TEXCOORD
:
3157 if (usage_idx
< 8) {
3158 shader_addline(buffer
, "%s%s = gl_TexCoord[%u]%s;\n",
3159 destination
, reg_mask
, usage_idx
, reg_mask
);
3161 shader_addline(buffer
, "%s%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
3162 destination
, reg_mask
, reg_mask
);
3166 case WINED3DDECLUSAGE_FOG
:
3167 shader_addline(buffer
, "%s%s = vec4(gl_FogFragCoord, 0.0, 0.0, 0.0)%s;\n",
3168 destination
, reg_mask
, reg_mask
);
3172 shader_addline(buffer
, "%s%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
3173 destination
, reg_mask
, reg_mask
);
3177 for(j
= 0; j
< MAX_REG_OUTPUT
; j
++) {
3178 usage_token_out
= semantics_out
[j
].usage
;
3179 if (!usage_token_out
) continue;
3180 register_token_out
= semantics_out
[j
].reg
;
3182 usage_out
= (usage_token_out
& WINED3DSP_DCL_USAGE_MASK
) >> WINED3DSP_DCL_USAGE_SHIFT
;
3183 usage_idx_out
= (usage_token_out
& WINED3DSP_DCL_USAGEINDEX_MASK
) >> WINED3DSP_DCL_USAGEINDEX_SHIFT
;
3184 shader_glsl_get_write_mask(register_token_out
, reg_mask_out
);
3186 if(usage
== usage_out
&&
3187 usage_idx
== usage_idx_out
) {
3188 shader_addline(buffer
, "%s%s = OUT[%u]%s;\n",
3189 destination
, reg_mask
, j
, reg_mask
);
3194 shader_addline(buffer
, "%s%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
3195 destination
, reg_mask
, reg_mask
);
3200 /* This is solely to make the compiler / linker happy and avoid warning about undefined
3201 * varyings. It shouldn't result in any real code executed on the GPU, since all read
3202 * input varyings are assigned above, if the optimizer works properly.
3204 for(i
= 0; i
< in_count
+ 2; i
++) {
3205 if(set
[i
] != WINED3DSP_WRITEMASK_ALL
) {
3206 unsigned int size
= 0;
3207 memset(reg_mask
, 0, sizeof(reg_mask
));
3208 if(!(set
[i
] & WINED3DSP_WRITEMASK_0
)) {
3209 reg_mask
[size
] = 'x';
3212 if(!(set
[i
] & WINED3DSP_WRITEMASK_1
)) {
3213 reg_mask
[size
] = 'y';
3216 if(!(set
[i
] & WINED3DSP_WRITEMASK_2
)) {
3217 reg_mask
[size
] = 'z';
3220 if(!(set
[i
] & WINED3DSP_WRITEMASK_3
)) {
3221 reg_mask
[size
] = 'w';
3225 if (i
== in_count
) {
3226 sprintf(destination
, "gl_FrontColor");
3227 } else if (i
== in_count
+ 1) {
3228 sprintf(destination
, "gl_FrontSecondaryColor");
3230 sprintf(destination
, "IN[%u]", i
);
3234 shader_addline(buffer
, "%s.%s = 0.0;\n", destination
, reg_mask
);
3236 shader_addline(buffer
, "%s.%s = vec%u(0.0);\n", destination
, reg_mask
, size
);
3241 HeapFree(GetProcessHeap(), 0, set
);
3244 static GLhandleARB
generate_param_reorder_function(IWineD3DVertexShader
*vertexshader
,
3245 IWineD3DPixelShader
*pixelshader
, const WineD3D_GL_Info
*gl_info
)
3247 GLhandleARB ret
= 0;
3248 IWineD3DVertexShaderImpl
*vs
= (IWineD3DVertexShaderImpl
*) vertexshader
;
3249 IWineD3DPixelShaderImpl
*ps
= (IWineD3DPixelShaderImpl
*) pixelshader
;
3250 IWineD3DDeviceImpl
*device
;
3251 DWORD vs_major
= WINED3DSHADER_VERSION_MAJOR(vs
->baseShader
.reg_maps
.shader_version
);
3252 DWORD ps_major
= ps
? WINED3DSHADER_VERSION_MAJOR(ps
->baseShader
.reg_maps
.shader_version
) : 0;
3254 SHADER_BUFFER buffer
;
3256 DWORD register_token
;
3257 DWORD usage
, usage_idx
, writemask
;
3259 const struct semantic
*semantics_out
, *semantics_in
;
3261 shader_buffer_init(&buffer
);
3263 shader_addline(&buffer
, "#version 120\n");
3265 if(vs_major
< 3 && ps_major
< 3) {
3266 /* That one is easy: The vertex shader writes to the builtin varyings, the pixel shader reads from them.
3267 * Take care about the texcoord .w fixup though if we're using the fixed function fragment pipeline
3269 device
= (IWineD3DDeviceImpl
*) vs
->baseShader
.device
;
3270 if((GLINFO_LOCATION
).set_texcoord_w
&& ps_major
== 0 && vs_major
> 0 &&
3271 !device
->frag_pipe
->ffp_proj_control
) {
3272 shader_addline(&buffer
, "void order_ps_input() {\n");
3273 for(i
= 0; i
< min(8, MAX_REG_TEXCRD
); i
++) {
3274 if(vs
->baseShader
.reg_maps
.texcoord_mask
[i
] != 0 &&
3275 vs
->baseShader
.reg_maps
.texcoord_mask
[i
] != WINED3DSP_WRITEMASK_ALL
) {
3276 shader_addline(&buffer
, "gl_TexCoord[%u].w = 1.0;\n", i
);
3279 shader_addline(&buffer
, "}\n");
3281 shader_addline(&buffer
, "void order_ps_input() { /* do nothing */ }\n");
3283 } else if(ps_major
< 3 && vs_major
>= 3) {
3284 /* The vertex shader writes to its own varyings, the pixel shader needs them in the builtin ones */
3285 semantics_out
= vs
->semantics_out
;
3287 shader_addline(&buffer
, "void order_ps_input(in vec4 OUT[%u]) {\n", MAX_REG_OUTPUT
);
3288 for(i
= 0; i
< MAX_REG_OUTPUT
; i
++) {
3289 usage_token
= semantics_out
[i
].usage
;
3290 if (!usage_token
) continue;
3291 register_token
= semantics_out
[i
].reg
;
3293 usage
= (usage_token
& WINED3DSP_DCL_USAGE_MASK
) >> WINED3DSP_DCL_USAGE_SHIFT
;
3294 usage_idx
= (usage_token
& WINED3DSP_DCL_USAGEINDEX_MASK
) >> WINED3DSP_DCL_USAGEINDEX_SHIFT
;
3295 writemask
= shader_glsl_get_write_mask(register_token
, reg_mask
);
3298 case WINED3DDECLUSAGE_COLOR
:
3300 shader_addline(&buffer
, "gl_FrontColor%s = OUT[%u]%s;\n", reg_mask
, i
, reg_mask
);
3301 else if (usage_idx
== 1)
3302 shader_addline(&buffer
, "gl_FrontSecondaryColor%s = OUT[%u]%s;\n", reg_mask
, i
, reg_mask
);
3305 case WINED3DDECLUSAGE_POSITION
:
3306 shader_addline(&buffer
, "gl_Position%s = OUT[%u]%s;\n", reg_mask
, i
, reg_mask
);
3309 case WINED3DDECLUSAGE_TEXCOORD
:
3310 if (usage_idx
< 8) {
3311 if(!(GLINFO_LOCATION
).set_texcoord_w
|| ps_major
> 0) writemask
|= WINED3DSP_WRITEMASK_3
;
3313 shader_addline(&buffer
, "gl_TexCoord[%u]%s = OUT[%u]%s;\n",
3314 usage_idx
, reg_mask
, i
, reg_mask
);
3315 if(!(writemask
& WINED3DSP_WRITEMASK_3
)) {
3316 shader_addline(&buffer
, "gl_TexCoord[%u].w = 1.0;\n", usage_idx
);
3321 case WINED3DDECLUSAGE_PSIZE
:
3322 shader_addline(&buffer
, "gl_PointSize = OUT[%u].x;\n", i
);
3325 case WINED3DDECLUSAGE_FOG
:
3326 shader_addline(&buffer
, "gl_FogFragCoord = OUT[%u].%c;\n", i
, reg_mask
[1]);
3333 shader_addline(&buffer
, "}\n");
3335 } else if(ps_major
>= 3 && vs_major
>= 3) {
3336 semantics_out
= vs
->semantics_out
;
3337 semantics_in
= ps
->semantics_in
;
3339 /* This one is tricky: a 3.0 pixel shader reads from a 3.0 vertex shader */
3340 shader_addline(&buffer
, "varying vec4 IN[%u];\n", GL_LIMITS(glsl_varyings
) / 4);
3341 shader_addline(&buffer
, "void order_ps_input(in vec4 OUT[%u]) {\n", MAX_REG_OUTPUT
);
3343 /* First, sort out position and point size. Those are not passed to the pixel shader */
3344 for(i
= 0; i
< MAX_REG_OUTPUT
; i
++) {
3345 usage_token
= semantics_out
[i
].usage
;
3346 if (!usage_token
) continue;
3347 register_token
= semantics_out
[i
].reg
;
3349 usage
= (usage_token
& WINED3DSP_DCL_USAGE_MASK
) >> WINED3DSP_DCL_USAGE_SHIFT
;
3350 usage_idx
= (usage_token
& WINED3DSP_DCL_USAGEINDEX_MASK
) >> WINED3DSP_DCL_USAGEINDEX_SHIFT
;
3351 shader_glsl_get_write_mask(register_token
, reg_mask
);
3354 case WINED3DDECLUSAGE_POSITION
:
3355 shader_addline(&buffer
, "gl_Position%s = OUT[%u]%s;\n", reg_mask
, i
, reg_mask
);
3358 case WINED3DDECLUSAGE_PSIZE
:
3359 shader_addline(&buffer
, "gl_PointSize = OUT[%u].x;\n", i
);
3367 /* Then, fix the pixel shader input */
3368 handle_ps3_input(&buffer
, semantics_in
, semantics_out
, gl_info
, ps
->input_reg_map
);
3370 shader_addline(&buffer
, "}\n");
3371 } else if(ps_major
>= 3 && vs_major
< 3) {
3372 semantics_in
= ps
->semantics_in
;
3374 shader_addline(&buffer
, "varying vec4 IN[%u];\n", GL_LIMITS(glsl_varyings
) / 4);
3375 shader_addline(&buffer
, "void order_ps_input() {\n");
3376 /* The vertex shader wrote to the builtin varyings. There is no need to figure out position and
3377 * point size, but we depend on the optimizers kindness to find out that the pixel shader doesn't
3378 * read gl_TexCoord and gl_ColorX, otherwise we'll run out of varyings
3380 handle_ps3_input(&buffer
, semantics_in
, NULL
, gl_info
, ps
->input_reg_map
);
3381 shader_addline(&buffer
, "}\n");
3383 ERR("Unexpected vertex and pixel shader version condition: vs: %d, ps: %d\n", vs_major
, ps_major
);
3386 ret
= GL_EXTCALL(glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB
));
3387 checkGLcall("glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB)");
3388 GL_EXTCALL(glShaderSourceARB(ret
, 1, (const char**)&buffer
.buffer
, NULL
));
3389 checkGLcall("glShaderSourceARB(ret, 1, &buffer.buffer, NULL)");
3390 GL_EXTCALL(glCompileShaderARB(ret
));
3391 checkGLcall("glCompileShaderARB(ret)");
3393 shader_buffer_free(&buffer
);
3397 static void hardcode_local_constants(IWineD3DBaseShaderImpl
*shader
, const WineD3D_GL_Info
*gl_info
,
3398 GLhandleARB programId
, char prefix
)
3400 const local_constant
*lconst
;
3405 LIST_FOR_EACH_ENTRY(lconst
, &shader
->baseShader
.constantsF
, local_constant
, entry
) {
3406 value
= (const float *)lconst
->value
;
3407 snprintf(glsl_name
, sizeof(glsl_name
), "%cLC%u", prefix
, lconst
->idx
);
3408 tmp_loc
= GL_EXTCALL(glGetUniformLocationARB(programId
, glsl_name
));
3409 GL_EXTCALL(glUniform4fvARB(tmp_loc
, 1, value
));
3411 checkGLcall("Hardcoding local constants\n");
3414 /** Sets the GLSL program ID for the given pixel and vertex shader combination.
3415 * It sets the programId on the current StateBlock (because it should be called
3416 * inside of the DrawPrimitive() part of the render loop).
3418 * If a program for the given combination does not exist, create one, and store
3419 * the program in the hash table. If it creates a program, it will link the
3420 * given objects, too.
3422 static void set_glsl_shader_program(IWineD3DDevice
*iface
, BOOL use_ps
, BOOL use_vs
) {
3423 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
3424 struct shader_glsl_priv
*priv
= This
->shader_priv
;
3425 const WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
3426 IWineD3DPixelShader
*pshader
= This
->stateBlock
->pixelShader
;
3427 IWineD3DVertexShader
*vshader
= This
->stateBlock
->vertexShader
;
3428 struct glsl_shader_prog_link
*entry
= NULL
;
3429 GLhandleARB programId
= 0;
3430 GLhandleARB reorder_shader_id
= 0;
3433 GLhandleARB vshader_id
, pshader_id
;
3434 struct ps_compile_args ps_compile_args
;
3435 struct vs_compile_args vs_compile_args
;
3438 find_vs_compile_args((IWineD3DVertexShaderImpl
*)This
->stateBlock
->vertexShader
, This
->stateBlock
, &vs_compile_args
);
3440 /* FIXME: Do we really have to spend CPU cycles to generate a few zeroed bytes? */
3441 memset(&vs_compile_args
, 0, sizeof(vs_compile_args
));
3444 find_ps_compile_args((IWineD3DPixelShaderImpl
*)This
->stateBlock
->pixelShader
, This
->stateBlock
, &ps_compile_args
);
3446 /* FIXME: Do we really have to spend CPU cycles to generate a few zeroed bytes? */
3447 memset(&ps_compile_args
, 0, sizeof(ps_compile_args
));
3449 entry
= get_glsl_program_entry(priv
, vshader
, pshader
, &vs_compile_args
, &ps_compile_args
);
3451 priv
->glsl_program
= entry
;
3455 /* If we get to this point, then no matching program exists, so we create one */
3456 programId
= GL_EXTCALL(glCreateProgramObjectARB());
3457 TRACE("Created new GLSL shader program %u\n", programId
);
3459 /* Create the entry */
3460 entry
= HeapAlloc(GetProcessHeap(), 0, sizeof(struct glsl_shader_prog_link
));
3461 entry
->programId
= programId
;
3462 entry
->vshader
= vshader
;
3463 entry
->pshader
= pshader
;
3464 entry
->vs_args
= vs_compile_args
;
3465 entry
->ps_args
= ps_compile_args
;
3466 entry
->constant_version
= 0;
3467 /* Add the hash table entry */
3468 add_glsl_program_entry(priv
, entry
);
3470 /* Set the current program */
3471 priv
->glsl_program
= entry
;
3474 vshader_id
= find_gl_vshader((IWineD3DVertexShaderImpl
*) vshader
, &vs_compile_args
);
3479 /* Attach GLSL vshader */
3481 const unsigned int max_attribs
= 16; /* TODO: Will this always be the case? It is at the moment... */
3484 reorder_shader_id
= generate_param_reorder_function(vshader
, pshader
, gl_info
);
3485 TRACE("Attaching GLSL shader object %u to program %u\n", reorder_shader_id
, programId
);
3486 GL_EXTCALL(glAttachObjectARB(programId
, reorder_shader_id
));
3487 checkGLcall("glAttachObjectARB");
3488 /* Flag the reorder function for deletion, then it will be freed automatically when the program
3491 GL_EXTCALL(glDeleteObjectARB(reorder_shader_id
));
3493 TRACE("Attaching GLSL shader object %u to program %u\n", vshader_id
, programId
);
3494 GL_EXTCALL(glAttachObjectARB(programId
, vshader_id
));
3495 checkGLcall("glAttachObjectARB");
3497 /* Bind vertex attributes to a corresponding index number to match
3498 * the same index numbers as ARB_vertex_programs (makes loading
3499 * vertex attributes simpler). With this method, we can use the
3500 * exact same code to load the attributes later for both ARB and
3503 * We have to do this here because we need to know the Program ID
3504 * in order to make the bindings work, and it has to be done prior
3505 * to linking the GLSL program. */
3506 for (i
= 0; i
< max_attribs
; ++i
) {
3507 if (((IWineD3DBaseShaderImpl
*)vshader
)->baseShader
.reg_maps
.attributes
[i
]) {
3508 snprintf(tmp_name
, sizeof(tmp_name
), "attrib%i", i
);
3509 GL_EXTCALL(glBindAttribLocationARB(programId
, i
, tmp_name
));
3512 checkGLcall("glBindAttribLocationARB");
3514 list_add_head(&((IWineD3DBaseShaderImpl
*)vshader
)->baseShader
.linked_programs
, &entry
->vshader_entry
);
3518 pshader_id
= find_gl_pshader((IWineD3DPixelShaderImpl
*) pshader
, &ps_compile_args
);
3523 /* Attach GLSL pshader */
3525 TRACE("Attaching GLSL shader object %u to program %u\n", pshader_id
, programId
);
3526 GL_EXTCALL(glAttachObjectARB(programId
, pshader_id
));
3527 checkGLcall("glAttachObjectARB");
3529 list_add_head(&((IWineD3DBaseShaderImpl
*)pshader
)->baseShader
.linked_programs
, &entry
->pshader_entry
);
3532 /* Link the program */
3533 TRACE("Linking GLSL shader program %u\n", programId
);
3534 GL_EXTCALL(glLinkProgramARB(programId
));
3535 print_glsl_info_log(&GLINFO_LOCATION
, programId
);
3537 entry
->vuniformF_locations
= HeapAlloc(GetProcessHeap(), 0, sizeof(GLhandleARB
) * GL_LIMITS(vshader_constantsF
));
3538 for (i
= 0; i
< GL_LIMITS(vshader_constantsF
); ++i
) {
3539 snprintf(glsl_name
, sizeof(glsl_name
), "VC[%i]", i
);
3540 entry
->vuniformF_locations
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, glsl_name
));
3542 for (i
= 0; i
< MAX_CONST_I
; ++i
) {
3543 snprintf(glsl_name
, sizeof(glsl_name
), "VI[%i]", i
);
3544 entry
->vuniformI_locations
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, glsl_name
));
3546 entry
->puniformF_locations
= HeapAlloc(GetProcessHeap(), 0, sizeof(GLhandleARB
) * GL_LIMITS(pshader_constantsF
));
3547 for (i
= 0; i
< GL_LIMITS(pshader_constantsF
); ++i
) {
3548 snprintf(glsl_name
, sizeof(glsl_name
), "PC[%i]", i
);
3549 entry
->puniformF_locations
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, glsl_name
));
3551 for (i
= 0; i
< MAX_CONST_I
; ++i
) {
3552 snprintf(glsl_name
, sizeof(glsl_name
), "PI[%i]", i
);
3553 entry
->puniformI_locations
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, glsl_name
));
3557 for(i
= 0; i
< ((IWineD3DPixelShaderImpl
*)pshader
)->numbumpenvmatconsts
; i
++) {
3559 sprintf(name
, "bumpenvmat%d", ((IWineD3DPixelShaderImpl
*)pshader
)->bumpenvmatconst
[i
].texunit
);
3560 entry
->bumpenvmat_location
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, name
));
3561 sprintf(name
, "luminancescale%d", ((IWineD3DPixelShaderImpl
*)pshader
)->luminanceconst
[i
].texunit
);
3562 entry
->luminancescale_location
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, name
));
3563 sprintf(name
, "luminanceoffset%d", ((IWineD3DPixelShaderImpl
*)pshader
)->luminanceconst
[i
].texunit
);
3564 entry
->luminanceoffset_location
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, name
));
3568 if (use_ps
&& ps_compile_args
.texrect_fixup
) {
3570 for (i
= 0; i
< MAX_FRAGMENT_SAMPLERS
; ++i
) {
3571 if (ps_compile_args
.texrect_fixup
& (1 << i
)) {
3572 sprintf(name
, "PsamplerRectFixup%u", i
);
3573 entry
->rectFixup_location
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, name
));
3575 entry
->rectFixup_location
[i
] = -1;
3580 entry
->posFixup_location
= GL_EXTCALL(glGetUniformLocationARB(programId
, "posFixup"));
3581 entry
->ycorrection_location
= GL_EXTCALL(glGetUniformLocationARB(programId
, "ycorrection"));
3582 checkGLcall("Find glsl program uniform locations");
3585 && WINED3DSHADER_VERSION_MAJOR(((IWineD3DPixelShaderImpl
*)pshader
)->baseShader
.reg_maps
.shader_version
) >= 3
3586 && ((IWineD3DPixelShaderImpl
*)pshader
)->declared_in_count
> GL_LIMITS(glsl_varyings
) / 4)
3588 TRACE("Shader %d needs vertex color clamping disabled\n", programId
);
3589 entry
->vertex_color_clamp
= GL_FALSE
;
3591 entry
->vertex_color_clamp
= GL_FIXED_ONLY_ARB
;
3594 /* Set the shader to allow uniform loading on it */
3595 GL_EXTCALL(glUseProgramObjectARB(programId
));
3596 checkGLcall("glUseProgramObjectARB(programId)");
3598 /* Load the vertex and pixel samplers now. The function that finds the mappings makes sure
3599 * that it stays the same for each vertexshader-pixelshader pair(=linked glsl program). If
3600 * a pshader with fixed function pipeline is used there are no vertex samplers, and if a
3601 * vertex shader with fixed function pixel processing is used we make sure that the card
3602 * supports enough samplers to allow the max number of vertex samplers with all possible
3603 * fixed function fragment processing setups. So once the program is linked these samplers
3607 /* Load vertex shader samplers */
3608 shader_glsl_load_vsamplers(gl_info
, This
->texUnitMap
, programId
);
3611 /* Load pixel shader samplers */
3612 shader_glsl_load_psamplers(gl_info
, This
->texUnitMap
, programId
);
3615 /* If the local constants do not have to be loaded with the environment constants,
3616 * load them now to have them hardcoded in the GLSL program. This saves some CPU cycles
3619 if(pshader
&& !((IWineD3DPixelShaderImpl
*)pshader
)->baseShader
.load_local_constsF
) {
3620 hardcode_local_constants((IWineD3DBaseShaderImpl
*) pshader
, gl_info
, programId
, 'P');
3622 if(vshader
&& !((IWineD3DVertexShaderImpl
*)vshader
)->baseShader
.load_local_constsF
) {
3623 hardcode_local_constants((IWineD3DBaseShaderImpl
*) vshader
, gl_info
, programId
, 'V');
3627 static GLhandleARB
create_glsl_blt_shader(const WineD3D_GL_Info
*gl_info
, enum tex_types tex_type
)
3629 GLhandleARB program_id
;
3630 GLhandleARB vshader_id
, pshader_id
;
3631 static const char *blt_vshader
[] =
3636 " gl_Position = gl_Vertex;\n"
3637 " gl_FrontColor = vec4(1.0);\n"
3638 " gl_TexCoord[0] = gl_MultiTexCoord0;\n"
3642 static const char *blt_pshaders
[tex_type_count
] =
3648 "uniform sampler2D sampler;\n"
3651 " gl_FragDepth = texture2D(sampler, gl_TexCoord[0].xy).x;\n"
3657 "uniform samplerCube sampler;\n"
3660 " gl_FragDepth = textureCube(sampler, gl_TexCoord[0].xyz).x;\n"
3664 "#extension GL_ARB_texture_rectangle : enable\n"
3665 "uniform sampler2DRect sampler;\n"
3668 " gl_FragDepth = texture2DRect(sampler, gl_TexCoord[0].xy).x;\n"
3672 if (!blt_pshaders
[tex_type
])
3674 FIXME("tex_type %#x not supported\n", tex_type
);
3678 vshader_id
= GL_EXTCALL(glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB
));
3679 GL_EXTCALL(glShaderSourceARB(vshader_id
, 1, blt_vshader
, NULL
));
3680 GL_EXTCALL(glCompileShaderARB(vshader_id
));
3682 pshader_id
= GL_EXTCALL(glCreateShaderObjectARB(GL_FRAGMENT_SHADER_ARB
));
3683 GL_EXTCALL(glShaderSourceARB(pshader_id
, 1, &blt_pshaders
[tex_type
], NULL
));
3684 GL_EXTCALL(glCompileShaderARB(pshader_id
));
3686 program_id
= GL_EXTCALL(glCreateProgramObjectARB());
3687 GL_EXTCALL(glAttachObjectARB(program_id
, vshader_id
));
3688 GL_EXTCALL(glAttachObjectARB(program_id
, pshader_id
));
3689 GL_EXTCALL(glLinkProgramARB(program_id
));
3691 print_glsl_info_log(&GLINFO_LOCATION
, program_id
);
3693 /* Once linked we can mark the shaders for deletion. They will be deleted once the program
3696 GL_EXTCALL(glDeleteObjectARB(vshader_id
));
3697 GL_EXTCALL(glDeleteObjectARB(pshader_id
));
3701 static void shader_glsl_select(IWineD3DDevice
*iface
, BOOL usePS
, BOOL useVS
) {
3702 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
3703 struct shader_glsl_priv
*priv
= This
->shader_priv
;
3704 const WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
3705 GLhandleARB program_id
= 0;
3706 GLenum old_vertex_color_clamp
, current_vertex_color_clamp
;
3708 old_vertex_color_clamp
= priv
->glsl_program
? priv
->glsl_program
->vertex_color_clamp
: GL_FIXED_ONLY_ARB
;
3710 if (useVS
|| usePS
) set_glsl_shader_program(iface
, usePS
, useVS
);
3711 else priv
->glsl_program
= NULL
;
3713 current_vertex_color_clamp
= priv
->glsl_program
? priv
->glsl_program
->vertex_color_clamp
: GL_FIXED_ONLY_ARB
;
3715 if (old_vertex_color_clamp
!= current_vertex_color_clamp
) {
3716 if (GL_SUPPORT(ARB_COLOR_BUFFER_FLOAT
)) {
3717 GL_EXTCALL(glClampColorARB(GL_CLAMP_VERTEX_COLOR_ARB
, current_vertex_color_clamp
));
3718 checkGLcall("glClampColorARB");
3720 FIXME("vertex color clamp needs to be changed, but extension not supported.\n");
3724 program_id
= priv
->glsl_program
? priv
->glsl_program
->programId
: 0;
3725 if (program_id
) TRACE("Using GLSL program %u\n", program_id
);
3726 GL_EXTCALL(glUseProgramObjectARB(program_id
));
3727 checkGLcall("glUseProgramObjectARB");
3730 static void shader_glsl_select_depth_blt(IWineD3DDevice
*iface
, enum tex_types tex_type
) {
3731 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
3732 const WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
3733 struct shader_glsl_priv
*priv
= This
->shader_priv
;
3734 GLhandleARB
*blt_program
= &priv
->depth_blt_program
[tex_type
];
3736 if (!*blt_program
) {
3738 *blt_program
= create_glsl_blt_shader(gl_info
, tex_type
);
3739 loc
= GL_EXTCALL(glGetUniformLocationARB(*blt_program
, "sampler"));
3740 GL_EXTCALL(glUseProgramObjectARB(*blt_program
));
3741 GL_EXTCALL(glUniform1iARB(loc
, 0));
3743 GL_EXTCALL(glUseProgramObjectARB(*blt_program
));
3747 static void shader_glsl_deselect_depth_blt(IWineD3DDevice
*iface
) {
3748 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
3749 const WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
3750 struct shader_glsl_priv
*priv
= This
->shader_priv
;
3751 GLhandleARB program_id
;
3753 program_id
= priv
->glsl_program
? priv
->glsl_program
->programId
: 0;
3754 if (program_id
) TRACE("Using GLSL program %u\n", program_id
);
3756 GL_EXTCALL(glUseProgramObjectARB(program_id
));
3757 checkGLcall("glUseProgramObjectARB");
3760 static void shader_glsl_destroy(IWineD3DBaseShader
*iface
) {
3761 const struct list
*linked_programs
;
3762 IWineD3DBaseShaderImpl
*This
= (IWineD3DBaseShaderImpl
*) iface
;
3763 IWineD3DDeviceImpl
*device
= (IWineD3DDeviceImpl
*)This
->baseShader
.device
;
3764 struct shader_glsl_priv
*priv
= device
->shader_priv
;
3765 const WineD3D_GL_Info
*gl_info
= &device
->adapter
->gl_info
;
3766 IWineD3DPixelShaderImpl
*ps
= NULL
;
3767 IWineD3DVertexShaderImpl
*vs
= NULL
;
3769 /* Note: Do not use QueryInterface here to find out which shader type this is because this code
3770 * can be called from IWineD3DBaseShader::Release
3772 char pshader
= shader_is_pshader_version(This
->baseShader
.reg_maps
.shader_version
);
3775 ps
= (IWineD3DPixelShaderImpl
*) This
;
3776 if(ps
->num_gl_shaders
== 0) return;
3777 if (priv
->glsl_program
&& (IWineD3DBaseShader
*)priv
->glsl_program
->pshader
== iface
)
3778 shader_glsl_select(This
->baseShader
.device
, FALSE
, FALSE
);
3780 vs
= (IWineD3DVertexShaderImpl
*) This
;
3781 if(vs
->num_gl_shaders
== 0) return;
3782 if (priv
->glsl_program
&& (IWineD3DBaseShader
*)priv
->glsl_program
->vshader
== iface
)
3783 shader_glsl_select(This
->baseShader
.device
, FALSE
, FALSE
);
3786 linked_programs
= &This
->baseShader
.linked_programs
;
3788 TRACE("Deleting linked programs\n");
3789 if (linked_programs
->next
) {
3790 struct glsl_shader_prog_link
*entry
, *entry2
;
3793 LIST_FOR_EACH_ENTRY_SAFE(entry
, entry2
, linked_programs
, struct glsl_shader_prog_link
, pshader_entry
) {
3794 delete_glsl_program_entry(priv
, gl_info
, entry
);
3797 LIST_FOR_EACH_ENTRY_SAFE(entry
, entry2
, linked_programs
, struct glsl_shader_prog_link
, vshader_entry
) {
3798 delete_glsl_program_entry(priv
, gl_info
, entry
);
3807 for(i
= 0; i
< ps
->num_gl_shaders
; i
++) {
3808 TRACE("deleting pshader %u\n", ps
->gl_shaders
[i
].prgId
);
3809 GL_EXTCALL(glDeleteObjectARB(ps
->gl_shaders
[i
].prgId
));
3810 checkGLcall("glDeleteObjectARB");
3813 HeapFree(GetProcessHeap(), 0, ps
->gl_shaders
);
3814 ps
->gl_shaders
= NULL
;
3815 ps
->num_gl_shaders
= 0;
3816 ps
->shader_array_size
= 0;
3821 for(i
= 0; i
< vs
->num_gl_shaders
; i
++) {
3822 TRACE("deleting vshader %u\n", vs
->gl_shaders
[i
].prgId
);
3823 GL_EXTCALL(glDeleteObjectARB(vs
->gl_shaders
[i
].prgId
));
3824 checkGLcall("glDeleteObjectARB");
3827 HeapFree(GetProcessHeap(), 0, vs
->gl_shaders
);
3828 vs
->gl_shaders
= NULL
;
3829 vs
->num_gl_shaders
= 0;
3830 vs
->shader_array_size
= 0;
3834 static unsigned int glsl_program_key_hash(const void *key
)
3836 const glsl_program_key_t
*k
= key
;
3838 unsigned int hash
= ((DWORD_PTR
) k
->vshader
) | ((DWORD_PTR
) k
->pshader
) << 16;
3839 hash
+= ~(hash
<< 15);
3840 hash
^= (hash
>> 10);
3841 hash
+= (hash
<< 3);
3842 hash
^= (hash
>> 6);
3843 hash
+= ~(hash
<< 11);
3844 hash
^= (hash
>> 16);
3849 static BOOL
glsl_program_key_compare(const void *keya
, const void *keyb
)
3851 const glsl_program_key_t
*ka
= keya
;
3852 const glsl_program_key_t
*kb
= keyb
;
3854 return ka
->vshader
== kb
->vshader
&& ka
->pshader
== kb
->pshader
&&
3855 (memcmp(&ka
->ps_args
, &kb
->ps_args
, sizeof(kb
->ps_args
)) == 0) &&
3856 (memcmp(&ka
->vs_args
, &kb
->vs_args
, sizeof(kb
->vs_args
)) == 0);
3859 static BOOL
constant_heap_init(struct constant_heap
*heap
, unsigned int constant_count
)
3861 SIZE_T size
= (constant_count
+ 1) * sizeof(*heap
->entries
) + constant_count
* sizeof(*heap
->positions
);
3862 void *mem
= HeapAlloc(GetProcessHeap(), 0, size
);
3866 ERR("Failed to allocate memory\n");
3870 heap
->entries
= mem
;
3871 heap
->entries
[1].version
= 0;
3872 heap
->positions
= (unsigned int *)(heap
->entries
+ constant_count
+ 1);
3878 static void constant_heap_free(struct constant_heap
*heap
)
3880 HeapFree(GetProcessHeap(), 0, heap
->entries
);
3883 static HRESULT
shader_glsl_alloc(IWineD3DDevice
*iface
) {
3884 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
3885 const WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
3886 struct shader_glsl_priv
*priv
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(struct shader_glsl_priv
));
3887 SIZE_T stack_size
= wined3d_log2i(max(GL_LIMITS(vshader_constantsF
), GL_LIMITS(pshader_constantsF
))) + 1;
3889 priv
->stack
= HeapAlloc(GetProcessHeap(), 0, stack_size
* sizeof(*priv
->stack
));
3892 ERR("Failed to allocate memory.\n");
3893 HeapFree(GetProcessHeap(), 0, priv
);
3894 return E_OUTOFMEMORY
;
3897 if (!constant_heap_init(&priv
->vconst_heap
, GL_LIMITS(vshader_constantsF
)))
3899 ERR("Failed to initialize vertex shader constant heap\n");
3900 HeapFree(GetProcessHeap(), 0, priv
->stack
);
3901 HeapFree(GetProcessHeap(), 0, priv
);
3902 return E_OUTOFMEMORY
;
3905 if (!constant_heap_init(&priv
->pconst_heap
, GL_LIMITS(pshader_constantsF
)))
3907 ERR("Failed to initialize pixel shader constant heap\n");
3908 constant_heap_free(&priv
->vconst_heap
);
3909 HeapFree(GetProcessHeap(), 0, priv
->stack
);
3910 HeapFree(GetProcessHeap(), 0, priv
);
3911 return E_OUTOFMEMORY
;
3914 priv
->glsl_program_lookup
= hash_table_create(glsl_program_key_hash
, glsl_program_key_compare
);
3915 priv
->next_constant_version
= 1;
3917 This
->shader_priv
= priv
;
3921 static void shader_glsl_free(IWineD3DDevice
*iface
) {
3922 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
3923 const WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
3924 struct shader_glsl_priv
*priv
= This
->shader_priv
;
3927 for (i
= 0; i
< tex_type_count
; ++i
)
3929 if (priv
->depth_blt_program
[i
])
3931 GL_EXTCALL(glDeleteObjectARB(priv
->depth_blt_program
[i
]));
3935 hash_table_destroy(priv
->glsl_program_lookup
, NULL
, NULL
);
3936 constant_heap_free(&priv
->pconst_heap
);
3937 constant_heap_free(&priv
->vconst_heap
);
3939 HeapFree(GetProcessHeap(), 0, This
->shader_priv
);
3940 This
->shader_priv
= NULL
;
3943 static BOOL
shader_glsl_dirty_const(IWineD3DDevice
*iface
) {
3944 /* TODO: GL_EXT_bindable_uniform can be used to share constants across shaders */
3948 static GLuint
shader_glsl_generate_pshader(IWineD3DPixelShader
*iface
, SHADER_BUFFER
*buffer
, const struct ps_compile_args
*args
) {
3949 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*)iface
;
3950 const struct shader_reg_maps
*reg_maps
= &This
->baseShader
.reg_maps
;
3951 CONST DWORD
*function
= This
->baseShader
.function
;
3952 const char *fragcolor
;
3953 const WineD3D_GL_Info
*gl_info
= &((IWineD3DDeviceImpl
*)This
->baseShader
.device
)->adapter
->gl_info
;
3955 /* Create the hw GLSL shader object and assign it as the shader->prgId */
3956 GLhandleARB shader_obj
= GL_EXTCALL(glCreateShaderObjectARB(GL_FRAGMENT_SHADER_ARB
));
3958 shader_addline(buffer
, "#version 120\n");
3960 if (GL_SUPPORT(ARB_DRAW_BUFFERS
)) {
3961 shader_addline(buffer
, "#extension GL_ARB_draw_buffers : enable\n");
3963 if (GL_SUPPORT(ARB_TEXTURE_RECTANGLE
)) {
3964 /* The spec says that it doesn't have to be explicitly enabled, but the nvidia
3965 * drivers write a warning if we don't do so
3967 shader_addline(buffer
, "#extension GL_ARB_texture_rectangle : enable\n");
3970 /* Base Declarations */
3971 shader_generate_glsl_declarations( (IWineD3DBaseShader
*) This
, reg_maps
, buffer
, &GLINFO_LOCATION
, args
);
3973 /* Pack 3.0 inputs */
3974 if (reg_maps
->shader_version
>= WINED3DPS_VERSION(3,0) && args
->vp_mode
!= vertexshader
) {
3975 pshader_glsl_input_pack(buffer
, This
->semantics_in
, iface
, args
->vp_mode
);
3978 /* Base Shader Body */
3979 shader_generate_main( (IWineD3DBaseShader
*) This
, buffer
, reg_maps
, function
);
3981 /* Pixel shaders < 2.0 place the resulting color in R0 implicitly */
3982 if (reg_maps
->shader_version
< WINED3DPS_VERSION(2,0))
3984 /* Some older cards like GeforceFX ones don't support multiple buffers, so also not gl_FragData */
3985 if(GL_SUPPORT(ARB_DRAW_BUFFERS
))
3986 shader_addline(buffer
, "gl_FragData[0] = R0;\n");
3988 shader_addline(buffer
, "gl_FragColor = R0;\n");
3991 if(GL_SUPPORT(ARB_DRAW_BUFFERS
)) {
3992 fragcolor
= "gl_FragData[0]";
3994 fragcolor
= "gl_FragColor";
3996 if(args
->srgb_correction
) {
3997 shader_addline(buffer
, "tmp0.xyz = pow(%s.xyz, vec3(%f, %f, %f)) * vec3(%f, %f, %f) - vec3(%f, %f, %f);\n",
3998 fragcolor
, srgb_pow
, srgb_pow
, srgb_pow
, srgb_mul_high
, srgb_mul_high
, srgb_mul_high
,
3999 srgb_sub_high
, srgb_sub_high
, srgb_sub_high
);
4000 shader_addline(buffer
, "tmp1.xyz = %s.xyz * srgb_mul_low.xyz;\n", fragcolor
);
4001 shader_addline(buffer
, "%s.x = %s.x < srgb_comparison.x ? tmp1.x : tmp0.x;\n", fragcolor
, fragcolor
);
4002 shader_addline(buffer
, "%s.y = %s.y < srgb_comparison.y ? tmp1.y : tmp0.y;\n", fragcolor
, fragcolor
);
4003 shader_addline(buffer
, "%s.z = %s.z < srgb_comparison.z ? tmp1.z : tmp0.z;\n", fragcolor
, fragcolor
);
4004 shader_addline(buffer
, "%s = clamp(%s, 0.0, 1.0);\n", fragcolor
, fragcolor
);
4006 /* Pixel shader < 3.0 do not replace the fog stage.
4007 * This implements linear fog computation and blending.
4008 * TODO: non linear fog
4009 * NOTE: gl_Fog.start and gl_Fog.end don't hold fog start s and end e but
4010 * -1/(e-s) and e/(e-s) respectively.
4012 if(reg_maps
->shader_version
< WINED3DPS_VERSION(3,0)) {
4014 case FOG_OFF
: break;
4016 shader_addline(buffer
, "float fogstart = -1.0 / (gl_Fog.end - gl_Fog.start);\n");
4017 shader_addline(buffer
, "float fogend = gl_Fog.end * -fogstart;\n");
4018 shader_addline(buffer
, "float Fog = clamp(gl_FogFragCoord * fogstart + fogend, 0.0, 1.0);\n");
4019 shader_addline(buffer
, "%s.xyz = mix(gl_Fog.color.xyz, %s.xyz, Fog);\n", fragcolor
, fragcolor
);
4022 /* Fog = e^(-gl_Fog.density * gl_FogFragCoord) */
4023 shader_addline(buffer
, "float Fog = exp(-gl_Fog.density * gl_FogFragCoord);\n");
4024 shader_addline(buffer
, "Fog = clamp(Fog, 0.0, 1.0);\n");
4025 shader_addline(buffer
, "%s.xyz = mix(gl_Fog.color.xyz, %s.xyz, Fog);\n", fragcolor
, fragcolor
);
4028 /* Fog = e^(-(gl_Fog.density * gl_FogFragCoord)^2) */
4029 shader_addline(buffer
, "float Fog = exp(-gl_Fog.density * gl_Fog.density * gl_FogFragCoord * gl_FogFragCoord);\n");
4030 shader_addline(buffer
, "Fog = clamp(Fog, 0.0, 1.0);\n");
4031 shader_addline(buffer
, "%s.xyz = mix(gl_Fog.color.xyz, %s.xyz, Fog);\n", fragcolor
, fragcolor
);
4036 shader_addline(buffer
, "}\n");
4038 TRACE("Compiling shader object %u\n", shader_obj
);
4039 GL_EXTCALL(glShaderSourceARB(shader_obj
, 1, (const char**)&buffer
->buffer
, NULL
));
4040 GL_EXTCALL(glCompileShaderARB(shader_obj
));
4041 print_glsl_info_log(&GLINFO_LOCATION
, shader_obj
);
4043 /* Store the shader object */
4047 static GLuint
shader_glsl_generate_vshader(IWineD3DVertexShader
*iface
, SHADER_BUFFER
*buffer
, const struct vs_compile_args
*args
) {
4048 IWineD3DVertexShaderImpl
*This
= (IWineD3DVertexShaderImpl
*)iface
;
4049 const struct shader_reg_maps
*reg_maps
= &This
->baseShader
.reg_maps
;
4050 CONST DWORD
*function
= This
->baseShader
.function
;
4051 const WineD3D_GL_Info
*gl_info
= &((IWineD3DDeviceImpl
*)This
->baseShader
.device
)->adapter
->gl_info
;
4053 /* Create the hw GLSL shader program and assign it as the shader->prgId */
4054 GLhandleARB shader_obj
= GL_EXTCALL(glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB
));
4056 shader_addline(buffer
, "#version 120\n");
4058 /* Base Declarations */
4059 shader_generate_glsl_declarations( (IWineD3DBaseShader
*) This
, reg_maps
, buffer
, &GLINFO_LOCATION
, NULL
);
4061 /* Base Shader Body */
4062 shader_generate_main( (IWineD3DBaseShader
*) This
, buffer
, reg_maps
, function
);
4064 /* Unpack 3.0 outputs */
4065 if (reg_maps
->shader_version
>= WINED3DVS_VERSION(3,0)) shader_addline(buffer
, "order_ps_input(OUT);\n");
4066 else shader_addline(buffer
, "order_ps_input();\n");
4068 /* The D3DRS_FOGTABLEMODE render state defines if the shader-generated fog coord is used
4069 * or if the fragment depth is used. If the fragment depth is used(FOGTABLEMODE != NONE),
4070 * the fog frag coord is thrown away. If the fog frag coord is used, but not written by
4071 * the shader, it is set to 0.0(fully fogged, since start = 1.0, end = 0.0)
4073 if(args
->fog_src
== VS_FOG_Z
) {
4074 shader_addline(buffer
, "gl_FogFragCoord = gl_Position.z;\n");
4075 } else if (!reg_maps
->fog
) {
4076 shader_addline(buffer
, "gl_FogFragCoord = 0.0;\n");
4079 /* Write the final position.
4081 * OpenGL coordinates specify the center of the pixel while d3d coords specify
4082 * the corner. The offsets are stored in z and w in posFixup. posFixup.y contains
4083 * 1.0 or -1.0 to turn the rendering upside down for offscreen rendering. PosFixup.x
4084 * contains 1.0 to allow a mad.
4086 shader_addline(buffer
, "gl_Position.y = gl_Position.y * posFixup.y;\n");
4087 shader_addline(buffer
, "gl_Position.xy += posFixup.zw * gl_Position.ww;\n");
4089 /* Z coord [0;1]->[-1;1] mapping, see comment in transform_projection in state.c
4091 * Basically we want (in homogeneous coordinates) z = z * 2 - 1. However, shaders are run
4092 * before the homogeneous divide, so we have to take the w into account: z = ((z / w) * 2 - 1) * w,
4093 * which is the same as z = z * 2 - w.
4095 shader_addline(buffer
, "gl_Position.z = gl_Position.z * 2.0 - gl_Position.w;\n");
4097 shader_addline(buffer
, "}\n");
4099 TRACE("Compiling shader object %u\n", shader_obj
);
4100 GL_EXTCALL(glShaderSourceARB(shader_obj
, 1, (const char**)&buffer
->buffer
, NULL
));
4101 GL_EXTCALL(glCompileShaderARB(shader_obj
));
4102 print_glsl_info_log(&GLINFO_LOCATION
, shader_obj
);
4107 static void shader_glsl_get_caps(WINED3DDEVTYPE devtype
, const WineD3D_GL_Info
*gl_info
, struct shader_caps
*pCaps
)
4109 /* Nvidia Geforce6/7 or Ati R4xx/R5xx cards with GLSL support, support VS 3.0 but older Nvidia/Ati
4110 * models with GLSL support only support 2.0. In case of nvidia we can detect VS 2.0 support using
4111 * vs_nv_version which is based on NV_vertex_program.
4112 * For Ati cards there's no way using glsl (it abstracts the lowlevel info away) and also not
4113 * using ARB_vertex_program. It is safe to assume that when a card supports pixel shader 2.0 it
4114 * supports vertex shader 2.0 too and the way around. We can detect ps2.0 using the maximum number
4115 * of native instructions, so use that here. For more info see the pixel shader versioning code below.
4117 if((GLINFO_LOCATION
.vs_nv_version
== VS_VERSION_20
) || (GLINFO_LOCATION
.ps_arb_max_instructions
<= 512))
4118 pCaps
->VertexShaderVersion
= WINED3DVS_VERSION(2,0);
4120 pCaps
->VertexShaderVersion
= WINED3DVS_VERSION(3,0);
4121 TRACE_(d3d_caps
)("Hardware vertex shader version %d.%d enabled (GLSL)\n", (pCaps
->VertexShaderVersion
>> 8) & 0xff, pCaps
->VertexShaderVersion
& 0xff);
4122 pCaps
->MaxVertexShaderConst
= GL_LIMITS(vshader_constantsF
);
4124 /* Older DX9-class videocards (GeforceFX / Radeon >9500/X*00) only support pixel shader 2.0/2.0a/2.0b.
4125 * In OpenGL the extensions related to GLSL abstract lowlevel GL info away which is needed
4126 * to distinguish between 2.0 and 3.0 (and 2.0a/2.0b). In case of Nvidia we use their fragment
4127 * program extensions. On other hardware including ATI GL_ARB_fragment_program offers the info
4128 * in max native instructions. Intel and others also offer the info in this extension but they
4129 * don't support GLSL (at least on Windows).
4131 * PS2.0 requires at least 96 instructions, 2.0a/2.0b go up to 512. Assume that if the number
4132 * of instructions is 512 or less we have to do with ps2.0 hardware.
4133 * 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.
4135 if((GLINFO_LOCATION
.ps_nv_version
== PS_VERSION_20
) || (GLINFO_LOCATION
.ps_arb_max_instructions
<= 512))
4136 pCaps
->PixelShaderVersion
= WINED3DPS_VERSION(2,0);
4138 pCaps
->PixelShaderVersion
= WINED3DPS_VERSION(3,0);
4140 /* FIXME: The following line is card dependent. -8.0 to 8.0 is the
4141 * Direct3D minimum requirement.
4143 * Both GL_ARB_fragment_program and GLSL require a "maximum representable magnitude"
4144 * of colors to be 2^10, and 2^32 for other floats. Should we use 1024 here?
4146 * The problem is that the refrast clamps temporary results in the shader to
4147 * [-MaxValue;+MaxValue]. If the card's max value is bigger than the one we advertize here,
4148 * then applications may miss the clamping behavior. On the other hand, if it is smaller,
4149 * the shader will generate incorrect results too. Unfortunately, GL deliberately doesn't
4150 * offer a way to query this.
4152 pCaps
->PixelShader1xMaxValue
= 8.0;
4153 TRACE_(d3d_caps
)("Hardware pixel shader version %d.%d enabled (GLSL)\n", (pCaps
->PixelShaderVersion
>> 8) & 0xff, pCaps
->PixelShaderVersion
& 0xff);
4156 static BOOL
shader_glsl_color_fixup_supported(struct color_fixup_desc fixup
)
4158 if (TRACE_ON(d3d_shader
) && TRACE_ON(d3d
))
4160 TRACE("Checking support for fixup:\n");
4161 dump_color_fixup_desc(fixup
);
4164 /* We support everything except YUV conversions. */
4165 if (!is_yuv_fixup(fixup
))
4171 TRACE("[FAILED]\n");
4175 static const SHADER_HANDLER shader_glsl_instruction_handler_table
[WINED3DSIH_TABLE_SIZE
] =
4177 /* WINED3DSIH_ABS */ shader_glsl_map2gl
,
4178 /* WINED3DSIH_ADD */ shader_glsl_arith
,
4179 /* WINED3DSIH_BEM */ pshader_glsl_bem
,
4180 /* WINED3DSIH_BREAK */ shader_glsl_break
,
4181 /* WINED3DSIH_BREAKC */ shader_glsl_breakc
,
4182 /* WINED3DSIH_BREAKP */ NULL
,
4183 /* WINED3DSIH_CALL */ shader_glsl_call
,
4184 /* WINED3DSIH_CALLNZ */ shader_glsl_callnz
,
4185 /* WINED3DSIH_CMP */ shader_glsl_cmp
,
4186 /* WINED3DSIH_CND */ shader_glsl_cnd
,
4187 /* WINED3DSIH_CRS */ shader_glsl_cross
,
4188 /* WINED3DSIH_DCL */ NULL
,
4189 /* WINED3DSIH_DEF */ NULL
,
4190 /* WINED3DSIH_DEFB */ NULL
,
4191 /* WINED3DSIH_DEFI */ NULL
,
4192 /* WINED3DSIH_DP2ADD */ pshader_glsl_dp2add
,
4193 /* WINED3DSIH_DP3 */ shader_glsl_dot
,
4194 /* WINED3DSIH_DP4 */ shader_glsl_dot
,
4195 /* WINED3DSIH_DST */ shader_glsl_dst
,
4196 /* WINED3DSIH_DSX */ shader_glsl_map2gl
,
4197 /* WINED3DSIH_DSY */ shader_glsl_map2gl
,
4198 /* WINED3DSIH_ELSE */ shader_glsl_else
,
4199 /* WINED3DSIH_ENDIF */ shader_glsl_end
,
4200 /* WINED3DSIH_ENDLOOP */ shader_glsl_end
,
4201 /* WINED3DSIH_ENDREP */ shader_glsl_end
,
4202 /* WINED3DSIH_EXP */ shader_glsl_map2gl
,
4203 /* WINED3DSIH_EXPP */ shader_glsl_expp
,
4204 /* WINED3DSIH_FRC */ shader_glsl_map2gl
,
4205 /* WINED3DSIH_IF */ shader_glsl_if
,
4206 /* WINED3DSIH_IFC */ shader_glsl_ifc
,
4207 /* WINED3DSIH_LABEL */ shader_glsl_label
,
4208 /* WINED3DSIH_LIT */ shader_glsl_lit
,
4209 /* WINED3DSIH_LOG */ shader_glsl_log
,
4210 /* WINED3DSIH_LOGP */ shader_glsl_log
,
4211 /* WINED3DSIH_LOOP */ shader_glsl_loop
,
4212 /* WINED3DSIH_LRP */ shader_glsl_lrp
,
4213 /* WINED3DSIH_M3x2 */ shader_glsl_mnxn
,
4214 /* WINED3DSIH_M3x3 */ shader_glsl_mnxn
,
4215 /* WINED3DSIH_M3x4 */ shader_glsl_mnxn
,
4216 /* WINED3DSIH_M4x3 */ shader_glsl_mnxn
,
4217 /* WINED3DSIH_M4x4 */ shader_glsl_mnxn
,
4218 /* WINED3DSIH_MAD */ shader_glsl_mad
,
4219 /* WINED3DSIH_MAX */ shader_glsl_map2gl
,
4220 /* WINED3DSIH_MIN */ shader_glsl_map2gl
,
4221 /* WINED3DSIH_MOV */ shader_glsl_mov
,
4222 /* WINED3DSIH_MOVA */ shader_glsl_mov
,
4223 /* WINED3DSIH_MUL */ shader_glsl_arith
,
4224 /* WINED3DSIH_NOP */ NULL
,
4225 /* WINED3DSIH_NRM */ shader_glsl_map2gl
,
4226 /* WINED3DSIH_PHASE */ NULL
,
4227 /* WINED3DSIH_POW */ shader_glsl_pow
,
4228 /* WINED3DSIH_RCP */ shader_glsl_rcp
,
4229 /* WINED3DSIH_REP */ shader_glsl_rep
,
4230 /* WINED3DSIH_RET */ NULL
,
4231 /* WINED3DSIH_RSQ */ shader_glsl_rsq
,
4232 /* WINED3DSIH_SETP */ NULL
,
4233 /* WINED3DSIH_SGE */ shader_glsl_compare
,
4234 /* WINED3DSIH_SGN */ shader_glsl_map2gl
,
4235 /* WINED3DSIH_SINCOS */ shader_glsl_sincos
,
4236 /* WINED3DSIH_SLT */ shader_glsl_compare
,
4237 /* WINED3DSIH_SUB */ shader_glsl_arith
,
4238 /* WINED3DSIH_TEX */ pshader_glsl_tex
,
4239 /* WINED3DSIH_TEXBEM */ pshader_glsl_texbem
,
4240 /* WINED3DSIH_TEXBEML */ pshader_glsl_texbem
,
4241 /* WINED3DSIH_TEXCOORD */ pshader_glsl_texcoord
,
4242 /* WINED3DSIH_TEXDEPTH */ pshader_glsl_texdepth
,
4243 /* WINED3DSIH_TEXDP3 */ pshader_glsl_texdp3
,
4244 /* WINED3DSIH_TEXDP3TEX */ pshader_glsl_texdp3tex
,
4245 /* WINED3DSIH_TEXKILL */ pshader_glsl_texkill
,
4246 /* WINED3DSIH_TEXLDD */ NULL
,
4247 /* WINED3DSIH_TEXLDL */ shader_glsl_texldl
,
4248 /* WINED3DSIH_TEXM3x2DEPTH */ pshader_glsl_texm3x2depth
,
4249 /* WINED3DSIH_TEXM3x2PAD */ pshader_glsl_texm3x2pad
,
4250 /* WINED3DSIH_TEXM3x2TEX */ pshader_glsl_texm3x2tex
,
4251 /* WINED3DSIH_TEXM3x3 */ pshader_glsl_texm3x3
,
4252 /* WINED3DSIH_TEXM3x3DIFF */ NULL
,
4253 /* WINED3DSIH_TEXM3x3PAD */ pshader_glsl_texm3x3pad
,
4254 /* WINED3DSIH_TEXM3x3SPEC */ pshader_glsl_texm3x3spec
,
4255 /* WINED3DSIH_TEXM3x3TEX */ pshader_glsl_texm3x3tex
,
4256 /* WINED3DSIH_TEXM3x3VSPEC */ pshader_glsl_texm3x3vspec
,
4257 /* WINED3DSIH_TEXREG2AR */ pshader_glsl_texreg2ar
,
4258 /* WINED3DSIH_TEXREG2GB */ pshader_glsl_texreg2gb
,
4259 /* WINED3DSIH_TEXREG2RGB */ pshader_glsl_texreg2rgb
,
4262 const shader_backend_t glsl_shader_backend
= {
4263 shader_glsl_instruction_handler_table
,
4265 shader_glsl_select_depth_blt
,
4266 shader_glsl_deselect_depth_blt
,
4267 shader_glsl_update_float_vertex_constants
,
4268 shader_glsl_update_float_pixel_constants
,
4269 shader_glsl_load_constants
,
4270 shader_glsl_destroy
,
4273 shader_glsl_dirty_const
,
4274 shader_glsl_generate_pshader
,
4275 shader_glsl_generate_vshader
,
4276 shader_glsl_get_caps
,
4277 shader_glsl_color_fixup_supported
,