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 SHADER_OPCODE_ARG
*arg
, 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 if (instr
== WINED3DSIO_TEXKILL
)
948 switch (instr
& WINED3DSP_SRCMOD_MASK
) {
949 case WINED3DSPSM_DZ
: /* Need to handle this in the instructions itself (texld & texcrd). */
951 case WINED3DSPSM_NONE
:
952 sprintf(out_str
, "%s%s", in_reg
, in_regswizzle
);
954 case WINED3DSPSM_NEG
:
955 sprintf(out_str
, "-%s%s", in_reg
, in_regswizzle
);
957 case WINED3DSPSM_NOT
:
958 sprintf(out_str
, "!%s%s", in_reg
, in_regswizzle
);
960 case WINED3DSPSM_BIAS
:
961 sprintf(out_str
, "(%s%s - vec4(0.5)%s)", in_reg
, in_regswizzle
, in_regswizzle
);
963 case WINED3DSPSM_BIASNEG
:
964 sprintf(out_str
, "-(%s%s - vec4(0.5)%s)", in_reg
, in_regswizzle
, in_regswizzle
);
966 case WINED3DSPSM_SIGN
:
967 sprintf(out_str
, "(2.0 * (%s%s - 0.5))", in_reg
, in_regswizzle
);
969 case WINED3DSPSM_SIGNNEG
:
970 sprintf(out_str
, "-(2.0 * (%s%s - 0.5))", in_reg
, in_regswizzle
);
972 case WINED3DSPSM_COMP
:
973 sprintf(out_str
, "(1.0 - %s%s)", in_reg
, in_regswizzle
);
976 sprintf(out_str
, "(2.0 * %s%s)", in_reg
, in_regswizzle
);
978 case WINED3DSPSM_X2NEG
:
979 sprintf(out_str
, "-(2.0 * %s%s)", in_reg
, in_regswizzle
);
981 case WINED3DSPSM_ABS
:
982 sprintf(out_str
, "abs(%s%s)", in_reg
, in_regswizzle
);
984 case WINED3DSPSM_ABSNEG
:
985 sprintf(out_str
, "-abs(%s%s)", in_reg
, in_regswizzle
);
988 FIXME("Unhandled modifier %u\n", (instr
& WINED3DSP_SRCMOD_MASK
));
989 sprintf(out_str
, "%s%s", in_reg
, in_regswizzle
);
993 /** Writes the GLSL variable name that corresponds to the register that the
994 * DX opcode parameter is trying to access */
995 static void shader_glsl_get_register_name(const DWORD param
, const DWORD addr_token
,
996 char *regstr
, BOOL
*is_color
, const SHADER_OPCODE_ARG
*arg
)
998 /* oPos, oFog and oPts in D3D */
999 static const char * const hwrastout_reg_names
[] = { "gl_Position", "gl_FogFragCoord", "gl_PointSize" };
1001 DWORD reg
= param
& WINED3DSP_REGNUM_MASK
;
1002 DWORD regtype
= shader_get_regtype(param
);
1003 IWineD3DBaseShaderImpl
* This
= (IWineD3DBaseShaderImpl
*) arg
->shader
;
1004 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
1005 const WineD3D_GL_Info
* gl_info
= &deviceImpl
->adapter
->gl_info
;
1006 DWORD shader_version
= This
->baseShader
.reg_maps
.shader_version
;
1007 char pshader
= shader_is_pshader_version(shader_version
);
1013 case WINED3DSPR_TEMP
:
1014 sprintf(tmpStr
, "R%u", reg
);
1016 case WINED3DSPR_INPUT
:
1018 /* Pixel shaders >= 3.0 */
1019 if (WINED3DSHADER_VERSION_MAJOR(shader_version
) >= 3)
1021 DWORD in_count
= GL_LIMITS(glsl_varyings
) / 4;
1023 if (param
& WINED3DSHADER_ADDRMODE_RELATIVE
) {
1024 glsl_src_param_t rel_param
;
1025 shader_glsl_add_src_param(arg
, addr_token
, 0, WINED3DSP_WRITEMASK_0
, &rel_param
);
1027 /* Removing a + 0 would be an obvious optimization, but macos doesn't see the NOP
1030 if(((IWineD3DPixelShaderImpl
*) This
)->input_reg_map
[reg
]) {
1031 if (((IWineD3DPixelShaderImpl
*)This
)->declared_in_count
> in_count
) {
1032 sprintf(tmpStr
, "((%s + %u) > %d ? (%s + %u) > %d ? gl_SecondaryColor : gl_Color : IN[%s + %u])",
1033 rel_param
.param_str
, ((IWineD3DPixelShaderImpl
*)This
)->input_reg_map
[reg
], in_count
- 1,
1034 rel_param
.param_str
, ((IWineD3DPixelShaderImpl
*)This
)->input_reg_map
[reg
], in_count
,
1035 rel_param
.param_str
, ((IWineD3DPixelShaderImpl
*)This
)->input_reg_map
[reg
]);
1037 sprintf(tmpStr
, "IN[%s + %u]", rel_param
.param_str
, ((IWineD3DPixelShaderImpl
*)This
)->input_reg_map
[reg
]);
1040 if (((IWineD3DPixelShaderImpl
*)This
)->declared_in_count
> in_count
) {
1041 sprintf(tmpStr
, "((%s) > %d ? (%s) > %d ? gl_SecondaryColor : gl_Color : IN[%s])",
1042 rel_param
.param_str
, in_count
- 1,
1043 rel_param
.param_str
, in_count
,
1044 rel_param
.param_str
);
1046 sprintf(tmpStr
, "IN[%s]", rel_param
.param_str
);
1050 DWORD idx
= ((IWineD3DPixelShaderImpl
*) This
)->input_reg_map
[reg
];
1051 if (idx
== in_count
) {
1052 sprintf(tmpStr
, "gl_Color");
1053 } else if (idx
== in_count
+ 1) {
1054 sprintf(tmpStr
, "gl_SecondaryColor");
1056 sprintf(tmpStr
, "IN[%u]", idx
);
1061 strcpy(tmpStr
, "gl_Color");
1063 strcpy(tmpStr
, "gl_SecondaryColor");
1066 if (((IWineD3DVertexShaderImpl
*)This
)->cur_args
->swizzle_map
& (1 << reg
)) *is_color
= TRUE
;
1067 sprintf(tmpStr
, "attrib%u", reg
);
1070 case WINED3DSPR_CONST
:
1072 const char prefix
= pshader
? 'P':'V';
1074 /* Relative addressing */
1075 if (param
& WINED3DSHADER_ADDRMODE_RELATIVE
) {
1077 /* Relative addressing on shaders 2.0+ have a relative address token,
1078 * prior to that, it was hard-coded as "A0.x" because there's only 1 register */
1079 if (WINED3DSHADER_VERSION_MAJOR(shader_version
) >= 2)
1081 glsl_src_param_t rel_param
;
1082 shader_glsl_add_src_param(arg
, addr_token
, 0, WINED3DSP_WRITEMASK_0
, &rel_param
);
1084 sprintf(tmpStr
, "%cC[%s + %u]", prefix
, rel_param
.param_str
, reg
);
1086 sprintf(tmpStr
, "%cC[%s]", prefix
, rel_param
.param_str
);
1090 sprintf(tmpStr
, "%cC[A0.x + %u]", prefix
, reg
);
1092 sprintf(tmpStr
, "%cC[A0.x]", prefix
);
1097 if(shader_constant_is_local(This
, reg
)) {
1098 sprintf(tmpStr
, "%cLC%u", prefix
, reg
);
1100 sprintf(tmpStr
, "%cC[%u]", prefix
, reg
);
1106 case WINED3DSPR_CONSTINT
:
1108 sprintf(tmpStr
, "PI[%u]", reg
);
1110 sprintf(tmpStr
, "VI[%u]", reg
);
1112 case WINED3DSPR_CONSTBOOL
:
1114 sprintf(tmpStr
, "PB[%u]", reg
);
1116 sprintf(tmpStr
, "VB[%u]", reg
);
1118 case WINED3DSPR_TEXTURE
: /* case WINED3DSPR_ADDR: */
1120 sprintf(tmpStr
, "T%u", reg
);
1122 sprintf(tmpStr
, "A%u", reg
);
1125 case WINED3DSPR_LOOP
:
1126 sprintf(tmpStr
, "aL%u", This
->baseShader
.cur_loop_regno
- 1);
1128 case WINED3DSPR_SAMPLER
:
1130 sprintf(tmpStr
, "Psampler%u", reg
);
1132 sprintf(tmpStr
, "Vsampler%u", reg
);
1134 case WINED3DSPR_COLOROUT
:
1135 if (reg
>= GL_LIMITS(buffers
)) {
1136 WARN("Write to render target %u, only %d supported\n", reg
, 4);
1138 if (GL_SUPPORT(ARB_DRAW_BUFFERS
)) {
1139 sprintf(tmpStr
, "gl_FragData[%u]", reg
);
1140 } else { /* On older cards with GLSL support like the GeforceFX there's only one buffer. */
1141 sprintf(tmpStr
, "gl_FragColor");
1144 case WINED3DSPR_RASTOUT
:
1145 sprintf(tmpStr
, "%s", hwrastout_reg_names
[reg
]);
1147 case WINED3DSPR_DEPTHOUT
:
1148 sprintf(tmpStr
, "gl_FragDepth");
1150 case WINED3DSPR_ATTROUT
:
1152 sprintf(tmpStr
, "gl_FrontColor");
1154 sprintf(tmpStr
, "gl_FrontSecondaryColor");
1157 case WINED3DSPR_TEXCRDOUT
:
1158 /* Vertex shaders >= 3.0: WINED3DSPR_OUTPUT */
1159 if (WINED3DSHADER_VERSION_MAJOR(shader_version
) >= 3) sprintf(tmpStr
, "OUT[%u]", reg
);
1160 else sprintf(tmpStr
, "gl_TexCoord[%u]", reg
);
1162 case WINED3DSPR_MISCTYPE
:
1165 sprintf(tmpStr
, "vpos");
1166 } else if (reg
== 1){
1167 /* Note that gl_FrontFacing is a bool, while vFace is
1168 * a float for which the sign determines front/back
1170 sprintf(tmpStr
, "(gl_FrontFacing ? 1.0 : -1.0)");
1172 FIXME("Unhandled misctype register %d\n", reg
);
1173 sprintf(tmpStr
, "unrecognized_register");
1177 FIXME("Unhandled register name Type(%d)\n", regtype
);
1178 sprintf(tmpStr
, "unrecognized_register");
1182 strcat(regstr
, tmpStr
);
1185 /* Get the GLSL write mask for the destination register */
1186 static DWORD
shader_glsl_get_write_mask(const DWORD param
, char *write_mask
) {
1187 char *ptr
= write_mask
;
1188 DWORD mask
= param
& WINED3DSP_WRITEMASK_ALL
;
1190 if (shader_is_scalar(param
)) {
1191 mask
= WINED3DSP_WRITEMASK_0
;
1194 if (param
& WINED3DSP_WRITEMASK_0
) *ptr
++ = 'x';
1195 if (param
& WINED3DSP_WRITEMASK_1
) *ptr
++ = 'y';
1196 if (param
& WINED3DSP_WRITEMASK_2
) *ptr
++ = 'z';
1197 if (param
& WINED3DSP_WRITEMASK_3
) *ptr
++ = 'w';
1205 static unsigned int shader_glsl_get_write_mask_size(DWORD write_mask
) {
1206 unsigned int size
= 0;
1208 if (write_mask
& WINED3DSP_WRITEMASK_0
) ++size
;
1209 if (write_mask
& WINED3DSP_WRITEMASK_1
) ++size
;
1210 if (write_mask
& WINED3DSP_WRITEMASK_2
) ++size
;
1211 if (write_mask
& WINED3DSP_WRITEMASK_3
) ++size
;
1216 static void shader_glsl_get_swizzle(const DWORD param
, BOOL fixup
, DWORD mask
, char *swizzle_str
) {
1217 /* For registers of type WINED3DDECLTYPE_D3DCOLOR, data is stored as "bgra",
1218 * but addressed as "rgba". To fix this we need to swap the register's x
1219 * and z components. */
1220 DWORD swizzle
= (param
& WINED3DSP_SWIZZLE_MASK
) >> WINED3DSP_SWIZZLE_SHIFT
;
1221 const char *swizzle_chars
= fixup
? "zyxw" : "xyzw";
1222 char *ptr
= swizzle_str
;
1224 if (!shader_is_scalar(param
)) {
1226 /* swizzle bits fields: wwzzyyxx */
1227 if (mask
& WINED3DSP_WRITEMASK_0
) *ptr
++ = swizzle_chars
[swizzle
& 0x03];
1228 if (mask
& WINED3DSP_WRITEMASK_1
) *ptr
++ = swizzle_chars
[(swizzle
>> 2) & 0x03];
1229 if (mask
& WINED3DSP_WRITEMASK_2
) *ptr
++ = swizzle_chars
[(swizzle
>> 4) & 0x03];
1230 if (mask
& WINED3DSP_WRITEMASK_3
) *ptr
++ = swizzle_chars
[(swizzle
>> 6) & 0x03];
1236 /* From a given parameter token, generate the corresponding GLSL string.
1237 * Also, return the actual register name and swizzle in case the
1238 * caller needs this information as well. */
1239 static void shader_glsl_add_src_param(const SHADER_OPCODE_ARG
*arg
, const DWORD param
,
1240 const DWORD addr_token
, DWORD mask
, glsl_src_param_t
*src_param
)
1242 BOOL is_color
= FALSE
;
1243 char swizzle_str
[6];
1245 src_param
->reg_name
[0] = '\0';
1246 src_param
->param_str
[0] = '\0';
1247 swizzle_str
[0] = '\0';
1249 shader_glsl_get_register_name(param
, addr_token
, src_param
->reg_name
, &is_color
, arg
);
1251 shader_glsl_get_swizzle(param
, is_color
, mask
, swizzle_str
);
1252 shader_glsl_gen_modifier(param
, src_param
->reg_name
, swizzle_str
, src_param
->param_str
);
1255 /* From a given parameter token, generate the corresponding GLSL string.
1256 * Also, return the actual register name and swizzle in case the
1257 * caller needs this information as well. */
1258 static DWORD
shader_glsl_add_dst_param(const SHADER_OPCODE_ARG
* arg
, const DWORD param
,
1259 const DWORD addr_token
, glsl_dst_param_t
*dst_param
)
1261 BOOL is_color
= FALSE
;
1263 dst_param
->mask_str
[0] = '\0';
1264 dst_param
->reg_name
[0] = '\0';
1266 shader_glsl_get_register_name(param
, addr_token
, dst_param
->reg_name
, &is_color
, arg
);
1267 return shader_glsl_get_write_mask(param
, dst_param
->mask_str
);
1270 /* Append the destination part of the instruction to the buffer, return the effective write mask */
1271 static DWORD
shader_glsl_append_dst_ext(SHADER_BUFFER
*buffer
, const SHADER_OPCODE_ARG
*arg
, const DWORD param
)
1273 glsl_dst_param_t dst_param
;
1277 mask
= shader_glsl_add_dst_param(arg
, param
, arg
->dst_addr
, &dst_param
);
1280 shift
= (param
& WINED3DSP_DSTSHIFT_MASK
) >> WINED3DSP_DSTSHIFT_SHIFT
;
1281 shader_addline(buffer
, "%s%s = %s(", dst_param
.reg_name
, dst_param
.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 SHADER_OPCODE_ARG
*arg
)
1290 return shader_glsl_append_dst_ext(buffer
, arg
, arg
->dst
);
1293 /** Process GLSL instruction modifiers */
1294 void shader_glsl_add_instruction_modifiers(const SHADER_OPCODE_ARG
* arg
)
1296 DWORD mask
= arg
->dst
& WINED3DSP_DSTMOD_MASK
;
1298 if (arg
->opcode
->dst_token
&& mask
!= 0) {
1299 glsl_dst_param_t dst_param
;
1301 shader_glsl_add_dst_param(arg
, arg
->dst
, 0, &dst_param
);
1303 if (mask
& WINED3DSPDM_SATURATE
) {
1304 /* _SAT means to clamp the value of the register to between 0 and 1 */
1305 shader_addline(arg
->buffer
, "%s%s = clamp(%s%s, 0.0, 1.0);\n", dst_param
.reg_name
,
1306 dst_param
.mask_str
, dst_param
.reg_name
, dst_param
.mask_str
);
1308 if (mask
& WINED3DSPDM_MSAMPCENTROID
) {
1309 FIXME("_centroid modifier not handled\n");
1311 if (mask
& WINED3DSPDM_PARTIALPRECISION
) {
1312 /* MSDN says this modifier can be safely ignored, so that's what we'll do. */
1317 static inline const char* shader_get_comp_op(
1318 const DWORD opcode
) {
1320 DWORD op
= (opcode
& INST_CONTROLS_MASK
) >> INST_CONTROLS_SHIFT
;
1322 case COMPARISON_GT
: return ">";
1323 case COMPARISON_EQ
: return "==";
1324 case COMPARISON_GE
: return ">=";
1325 case COMPARISON_LT
: return "<";
1326 case COMPARISON_NE
: return "!=";
1327 case COMPARISON_LE
: return "<=";
1329 FIXME("Unrecognized comparison value: %u\n", op
);
1334 static void shader_glsl_get_sample_function(DWORD sampler_type
, DWORD flags
, glsl_sample_function_t
*sample_function
)
1336 BOOL projected
= flags
& WINED3D_GLSL_SAMPLE_PROJECTED
;
1337 BOOL texrect
= flags
& WINED3D_GLSL_SAMPLE_RECT
;
1338 BOOL lod
= flags
& WINED3D_GLSL_SAMPLE_LOD
;
1340 /* Note that there's no such thing as a projected cube texture. */
1341 switch(sampler_type
) {
1344 sample_function
->name
= projected
? "texture1DProjLod" : "texture1DLod";
1346 sample_function
->name
= projected
? "texture1DProj" : "texture1D";
1348 sample_function
->coord_mask
= WINED3DSP_WRITEMASK_0
;
1353 sample_function
->name
= projected
? "texture2DRectProjLod" : "texture2DRectLod";
1355 sample_function
->name
= projected
? "texture2DRectProj" : "texture2DRect";
1359 sample_function
->name
= projected
? "texture2DProjLod" : "texture2DLod";
1361 sample_function
->name
= projected
? "texture2DProj" : "texture2D";
1364 sample_function
->coord_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
;
1366 case WINED3DSTT_CUBE
:
1368 sample_function
->name
= "textureCubeLod";
1370 sample_function
->name
= "textureCube";
1372 sample_function
->coord_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
1374 case WINED3DSTT_VOLUME
:
1376 sample_function
->name
= projected
? "texture3DProjLod" : "texture3DLod";
1378 sample_function
->name
= projected
? "texture3DProj" : "texture3D";
1380 sample_function
->coord_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
1383 sample_function
->name
= "";
1384 sample_function
->coord_mask
= 0;
1385 FIXME("Unrecognized sampler type: %#x;\n", sampler_type
);
1390 static void shader_glsl_append_fixup_arg(char *arguments
, const char *reg_name
,
1391 BOOL sign_fixup
, enum fixup_channel_source channel_source
)
1393 switch(channel_source
)
1395 case CHANNEL_SOURCE_ZERO
:
1396 strcat(arguments
, "0.0");
1399 case CHANNEL_SOURCE_ONE
:
1400 strcat(arguments
, "1.0");
1403 case CHANNEL_SOURCE_X
:
1404 strcat(arguments
, reg_name
);
1405 strcat(arguments
, ".x");
1408 case CHANNEL_SOURCE_Y
:
1409 strcat(arguments
, reg_name
);
1410 strcat(arguments
, ".y");
1413 case CHANNEL_SOURCE_Z
:
1414 strcat(arguments
, reg_name
);
1415 strcat(arguments
, ".z");
1418 case CHANNEL_SOURCE_W
:
1419 strcat(arguments
, reg_name
);
1420 strcat(arguments
, ".w");
1424 FIXME("Unhandled channel source %#x\n", channel_source
);
1425 strcat(arguments
, "undefined");
1429 if (sign_fixup
) strcat(arguments
, " * 2.0 - 1.0");
1432 static void shader_glsl_color_correction(const struct SHADER_OPCODE_ARG
*arg
, struct color_fixup_desc fixup
)
1434 unsigned int mask_size
, remaining
;
1435 glsl_dst_param_t dst_param
;
1436 char arguments
[256];
1441 if (fixup
.x_sign_fixup
|| fixup
.x_source
!= CHANNEL_SOURCE_X
) mask
|= WINED3DSP_WRITEMASK_0
;
1442 if (fixup
.y_sign_fixup
|| fixup
.y_source
!= CHANNEL_SOURCE_Y
) mask
|= WINED3DSP_WRITEMASK_1
;
1443 if (fixup
.z_sign_fixup
|| fixup
.z_source
!= CHANNEL_SOURCE_Z
) mask
|= WINED3DSP_WRITEMASK_2
;
1444 if (fixup
.w_sign_fixup
|| fixup
.w_source
!= CHANNEL_SOURCE_W
) mask
|= WINED3DSP_WRITEMASK_3
;
1447 if (!mask
) return; /* Nothing to do */
1449 if (is_yuv_fixup(fixup
))
1451 enum yuv_fixup yuv_fixup
= get_yuv_fixup(fixup
);
1452 FIXME("YUV fixup (%#x) not supported\n", yuv_fixup
);
1456 mask_size
= shader_glsl_get_write_mask_size(mask
);
1458 dst_param
.mask_str
[0] = '\0';
1459 shader_glsl_get_write_mask(mask
, dst_param
.mask_str
);
1461 dst_param
.reg_name
[0] = '\0';
1462 shader_glsl_get_register_name(arg
->dst
, arg
->dst_addr
, dst_param
.reg_name
, &dummy
, arg
);
1464 arguments
[0] = '\0';
1465 remaining
= mask_size
;
1466 if (mask
& WINED3DSP_WRITEMASK_0
)
1468 shader_glsl_append_fixup_arg(arguments
, dst_param
.reg_name
, fixup
.x_sign_fixup
, fixup
.x_source
);
1469 if (--remaining
) strcat(arguments
, ", ");
1471 if (mask
& WINED3DSP_WRITEMASK_1
)
1473 shader_glsl_append_fixup_arg(arguments
, dst_param
.reg_name
, fixup
.y_sign_fixup
, fixup
.y_source
);
1474 if (--remaining
) strcat(arguments
, ", ");
1476 if (mask
& WINED3DSP_WRITEMASK_2
)
1478 shader_glsl_append_fixup_arg(arguments
, dst_param
.reg_name
, fixup
.z_sign_fixup
, fixup
.z_source
);
1479 if (--remaining
) strcat(arguments
, ", ");
1481 if (mask
& WINED3DSP_WRITEMASK_3
)
1483 shader_glsl_append_fixup_arg(arguments
, dst_param
.reg_name
, fixup
.w_sign_fixup
, fixup
.w_source
);
1484 if (--remaining
) strcat(arguments
, ", ");
1489 shader_addline(arg
->buffer
, "%s%s = vec%u(%s);\n",
1490 dst_param
.reg_name
, dst_param
.mask_str
, mask_size
, arguments
);
1494 shader_addline(arg
->buffer
, "%s%s = %s;\n", dst_param
.reg_name
, dst_param
.mask_str
, arguments
);
1498 static void PRINTF_ATTR(6, 7) shader_glsl_gen_sample_code(const SHADER_OPCODE_ARG
*arg
,
1499 DWORD sampler
, const glsl_sample_function_t
*sample_function
, DWORD swizzle
,
1500 const char *bias
, const char *coord_reg_fmt
, ...)
1502 const char *sampler_base
;
1503 char dst_swizzle
[6];
1504 struct color_fixup_desc fixup
;
1505 BOOL rect_fixup
= FALSE
;
1508 shader_glsl_get_swizzle(swizzle
, FALSE
, arg
->dst
, dst_swizzle
);
1510 if(shader_is_pshader_version(arg
->reg_maps
->shader_version
)) {
1511 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*) arg
->shader
;
1512 fixup
= This
->cur_args
->color_fixup
[sampler
];
1513 sampler_base
= "Psampler";
1515 if(This
->cur_args
->texrect_fixup
& (1 << sampler
)) {
1517 FIXME("Biased sampling from RECT textures is unsupported\n");
1523 sampler_base
= "Vsampler";
1524 fixup
= COLOR_FIXUP_IDENTITY
; /* FIXME: Vshader color fixup */
1527 shader_glsl_append_dst(arg
->buffer
, arg
);
1529 shader_addline(arg
->buffer
, "%s(%s%u, ", sample_function
->name
, sampler_base
, sampler
);
1531 va_start(args
, coord_reg_fmt
);
1532 shader_vaddline(arg
->buffer
, coord_reg_fmt
, args
);
1536 shader_addline(arg
->buffer
, ", %s)%s);\n", bias
, dst_swizzle
);
1539 shader_addline(arg
->buffer
, " * PsamplerRectFixup%u)%s);\n", sampler
, dst_swizzle
);
1541 shader_addline(arg
->buffer
, ")%s);\n", dst_swizzle
);
1545 if(!is_identity_fixup(fixup
)) {
1546 shader_glsl_color_correction(arg
, fixup
);
1550 /*****************************************************************************
1552 * Begin processing individual instruction opcodes
1554 ****************************************************************************/
1556 /* Generate GLSL arithmetic functions (dst = src1 + src2) */
1557 static void shader_glsl_arith(const SHADER_OPCODE_ARG
*arg
)
1559 CONST SHADER_OPCODE
* curOpcode
= arg
->opcode
;
1560 SHADER_BUFFER
* buffer
= arg
->buffer
;
1561 glsl_src_param_t src0_param
;
1562 glsl_src_param_t src1_param
;
1566 /* Determine the GLSL operator to use based on the opcode */
1567 switch (curOpcode
->opcode
) {
1568 case WINED3DSIO_MUL
: op
= '*'; break;
1569 case WINED3DSIO_ADD
: op
= '+'; break;
1570 case WINED3DSIO_SUB
: op
= '-'; break;
1573 FIXME("Opcode %s not yet handled in GLSL\n", curOpcode
->name
);
1577 write_mask
= shader_glsl_append_dst(buffer
, arg
);
1578 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], write_mask
, &src0_param
);
1579 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], write_mask
, &src1_param
);
1580 shader_addline(buffer
, "%s %c %s);\n", src0_param
.param_str
, op
, src1_param
.param_str
);
1583 /* Process the WINED3DSIO_MOV opcode using GLSL (dst = src) */
1584 static void shader_glsl_mov(const SHADER_OPCODE_ARG
*arg
)
1586 SHADER_BUFFER
* buffer
= arg
->buffer
;
1587 glsl_src_param_t src0_param
;
1590 write_mask
= shader_glsl_append_dst(buffer
, arg
);
1591 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], write_mask
, &src0_param
);
1593 /* In vs_1_1 WINED3DSIO_MOV can write to the address register. In later
1594 * shader versions WINED3DSIO_MOVA is used for this. */
1595 if ((WINED3DSHADER_VERSION_MAJOR(arg
->reg_maps
->shader_version
) == 1
1596 && !shader_is_pshader_version(arg
->reg_maps
->shader_version
)
1597 && shader_get_regtype(arg
->dst
) == WINED3DSPR_ADDR
))
1599 /* This is a simple floor() */
1600 unsigned int mask_size
= shader_glsl_get_write_mask_size(write_mask
);
1601 if (mask_size
> 1) {
1602 shader_addline(buffer
, "ivec%d(floor(%s)));\n", mask_size
, src0_param
.param_str
);
1604 shader_addline(buffer
, "int(floor(%s)));\n", src0_param
.param_str
);
1606 } else if(arg
->opcode
->opcode
== WINED3DSIO_MOVA
) {
1607 /* We need to *round* to the nearest int here. */
1608 unsigned int mask_size
= shader_glsl_get_write_mask_size(write_mask
);
1609 if (mask_size
> 1) {
1610 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
);
1612 shader_addline(buffer
, "int(floor(abs(%s) + 0.5) * sign(%s)));\n", src0_param
.param_str
, src0_param
.param_str
);
1615 shader_addline(buffer
, "%s);\n", src0_param
.param_str
);
1619 /* Process the dot product operators DP3 and DP4 in GLSL (dst = dot(src0, src1)) */
1620 static void shader_glsl_dot(const SHADER_OPCODE_ARG
*arg
)
1622 CONST SHADER_OPCODE
* curOpcode
= arg
->opcode
;
1623 SHADER_BUFFER
* buffer
= arg
->buffer
;
1624 glsl_src_param_t src0_param
;
1625 glsl_src_param_t src1_param
;
1626 DWORD dst_write_mask
, src_write_mask
;
1627 unsigned int dst_size
= 0;
1629 dst_write_mask
= shader_glsl_append_dst(buffer
, arg
);
1630 dst_size
= shader_glsl_get_write_mask_size(dst_write_mask
);
1632 /* dp3 works on vec3, dp4 on vec4 */
1633 if (curOpcode
->opcode
== WINED3DSIO_DP4
) {
1634 src_write_mask
= WINED3DSP_WRITEMASK_ALL
;
1636 src_write_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
1639 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], src_write_mask
, &src0_param
);
1640 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], src_write_mask
, &src1_param
);
1643 shader_addline(buffer
, "vec%d(dot(%s, %s)));\n", dst_size
, src0_param
.param_str
, src1_param
.param_str
);
1645 shader_addline(buffer
, "dot(%s, %s));\n", src0_param
.param_str
, src1_param
.param_str
);
1649 /* Note that this instruction has some restrictions. The destination write mask
1650 * can't contain the w component, and the source swizzles have to be .xyzw */
1651 static void shader_glsl_cross(const SHADER_OPCODE_ARG
*arg
)
1653 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
1654 glsl_src_param_t src0_param
;
1655 glsl_src_param_t src1_param
;
1658 shader_glsl_get_write_mask(arg
->dst
, dst_mask
);
1659 shader_glsl_append_dst(arg
->buffer
, arg
);
1660 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], src_mask
, &src0_param
);
1661 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], src_mask
, &src1_param
);
1662 shader_addline(arg
->buffer
, "cross(%s, %s)%s);\n", src0_param
.param_str
, src1_param
.param_str
, dst_mask
);
1665 /* Process the WINED3DSIO_POW instruction in GLSL (dst = |src0|^src1)
1666 * Src0 and src1 are scalars. Note that D3D uses the absolute of src0, while
1667 * GLSL uses the value as-is. */
1668 static void shader_glsl_pow(const SHADER_OPCODE_ARG
*arg
)
1670 SHADER_BUFFER
*buffer
= arg
->buffer
;
1671 glsl_src_param_t src0_param
;
1672 glsl_src_param_t src1_param
;
1673 DWORD dst_write_mask
;
1674 unsigned int dst_size
;
1676 dst_write_mask
= shader_glsl_append_dst(buffer
, arg
);
1677 dst_size
= shader_glsl_get_write_mask_size(dst_write_mask
);
1679 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
1680 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], WINED3DSP_WRITEMASK_0
, &src1_param
);
1683 shader_addline(buffer
, "vec%d(pow(abs(%s), %s)));\n", dst_size
, src0_param
.param_str
, src1_param
.param_str
);
1685 shader_addline(buffer
, "pow(abs(%s), %s));\n", src0_param
.param_str
, src1_param
.param_str
);
1689 /* Process the WINED3DSIO_LOG instruction in GLSL (dst = log2(|src0|))
1690 * Src0 is a scalar. Note that D3D uses the absolute of src0, while
1691 * GLSL uses the value as-is. */
1692 static void shader_glsl_log(const SHADER_OPCODE_ARG
*arg
)
1694 SHADER_BUFFER
*buffer
= arg
->buffer
;
1695 glsl_src_param_t src0_param
;
1696 DWORD dst_write_mask
;
1697 unsigned int dst_size
;
1699 dst_write_mask
= shader_glsl_append_dst(buffer
, arg
);
1700 dst_size
= shader_glsl_get_write_mask_size(dst_write_mask
);
1702 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
1705 shader_addline(buffer
, "vec%d(log2(abs(%s))));\n", dst_size
, src0_param
.param_str
);
1707 shader_addline(buffer
, "log2(abs(%s)));\n", src0_param
.param_str
);
1711 /* Map the opcode 1-to-1 to the GL code (arg->dst = instruction(src0, src1, ...) */
1712 static void shader_glsl_map2gl(const SHADER_OPCODE_ARG
*arg
)
1714 CONST SHADER_OPCODE
* curOpcode
= arg
->opcode
;
1715 SHADER_BUFFER
* buffer
= arg
->buffer
;
1716 glsl_src_param_t src_param
;
1717 const char *instruction
;
1721 /* Determine the GLSL function to use based on the opcode */
1722 /* TODO: Possibly make this a table for faster lookups */
1723 switch (curOpcode
->opcode
) {
1724 case WINED3DSIO_MIN
: instruction
= "min"; break;
1725 case WINED3DSIO_MAX
: instruction
= "max"; break;
1726 case WINED3DSIO_ABS
: instruction
= "abs"; break;
1727 case WINED3DSIO_FRC
: instruction
= "fract"; break;
1728 case WINED3DSIO_NRM
: instruction
= "normalize"; break;
1729 case WINED3DSIO_EXP
: instruction
= "exp2"; break;
1730 case WINED3DSIO_SGN
: instruction
= "sign"; break;
1731 case WINED3DSIO_DSX
: instruction
= "dFdx"; break;
1732 case WINED3DSIO_DSY
: instruction
= "ycorrection.y * dFdy"; break;
1733 default: instruction
= "";
1734 FIXME("Opcode %s not yet handled in GLSL\n", curOpcode
->name
);
1738 write_mask
= shader_glsl_append_dst(buffer
, arg
);
1740 shader_addline(buffer
, "%s(", instruction
);
1742 if (curOpcode
->num_params
> 0) {
1743 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], write_mask
, &src_param
);
1744 shader_addline(buffer
, "%s", src_param
.param_str
);
1745 for (i
= 2; i
< curOpcode
->num_params
; ++i
) {
1746 shader_glsl_add_src_param(arg
, arg
->src
[i
-1], arg
->src_addr
[i
-1], write_mask
, &src_param
);
1747 shader_addline(buffer
, ", %s", src_param
.param_str
);
1751 shader_addline(buffer
, "));\n");
1754 /** Process the WINED3DSIO_EXPP instruction in GLSL:
1755 * For shader model 1.x, do the following (and honor the writemask, so use a temporary variable):
1756 * dst.x = 2^(floor(src))
1757 * dst.y = src - floor(src)
1758 * dst.z = 2^src (partial precision is allowed, but optional)
1760 * For 2.0 shaders, just do this (honoring writemask and swizzle):
1761 * dst = 2^src; (partial precision is allowed, but optional)
1763 static void shader_glsl_expp(const SHADER_OPCODE_ARG
*arg
)
1765 glsl_src_param_t src_param
;
1767 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src_param
);
1769 if (arg
->reg_maps
->shader_version
< WINED3DPS_VERSION(2,0))
1773 shader_addline(arg
->buffer
, "tmp0.x = exp2(floor(%s));\n", src_param
.param_str
);
1774 shader_addline(arg
->buffer
, "tmp0.y = %s - floor(%s);\n", src_param
.param_str
, src_param
.param_str
);
1775 shader_addline(arg
->buffer
, "tmp0.z = exp2(%s);\n", src_param
.param_str
);
1776 shader_addline(arg
->buffer
, "tmp0.w = 1.0;\n");
1778 shader_glsl_append_dst(arg
->buffer
, arg
);
1779 shader_glsl_get_write_mask(arg
->dst
, dst_mask
);
1780 shader_addline(arg
->buffer
, "tmp0%s);\n", dst_mask
);
1783 unsigned int mask_size
;
1785 write_mask
= shader_glsl_append_dst(arg
->buffer
, arg
);
1786 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
1788 if (mask_size
> 1) {
1789 shader_addline(arg
->buffer
, "vec%d(exp2(%s)));\n", mask_size
, src_param
.param_str
);
1791 shader_addline(arg
->buffer
, "exp2(%s));\n", src_param
.param_str
);
1796 /** Process the RCP (reciprocal or inverse) opcode in GLSL (dst = 1 / src) */
1797 static void shader_glsl_rcp(const SHADER_OPCODE_ARG
*arg
)
1799 glsl_src_param_t src_param
;
1801 unsigned int mask_size
;
1803 write_mask
= shader_glsl_append_dst(arg
->buffer
, arg
);
1804 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
1805 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_3
, &src_param
);
1807 if (mask_size
> 1) {
1808 shader_addline(arg
->buffer
, "vec%d(1.0 / %s));\n", mask_size
, src_param
.param_str
);
1810 shader_addline(arg
->buffer
, "1.0 / %s);\n", src_param
.param_str
);
1814 static void shader_glsl_rsq(const SHADER_OPCODE_ARG
*arg
)
1816 SHADER_BUFFER
* buffer
= arg
->buffer
;
1817 glsl_src_param_t src_param
;
1819 unsigned int mask_size
;
1821 write_mask
= shader_glsl_append_dst(buffer
, arg
);
1822 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
1824 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_3
, &src_param
);
1826 if (mask_size
> 1) {
1827 shader_addline(buffer
, "vec%d(inversesqrt(%s)));\n", mask_size
, src_param
.param_str
);
1829 shader_addline(buffer
, "inversesqrt(%s));\n", src_param
.param_str
);
1833 /** Process signed comparison opcodes in GLSL. */
1834 static void shader_glsl_compare(const SHADER_OPCODE_ARG
*arg
)
1836 glsl_src_param_t src0_param
;
1837 glsl_src_param_t src1_param
;
1839 unsigned int mask_size
;
1841 write_mask
= shader_glsl_append_dst(arg
->buffer
, arg
);
1842 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
1843 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], write_mask
, &src0_param
);
1844 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], write_mask
, &src1_param
);
1846 if (mask_size
> 1) {
1847 const char *compare
;
1849 switch(arg
->opcode
->opcode
) {
1850 case WINED3DSIO_SLT
: compare
= "lessThan"; break;
1851 case WINED3DSIO_SGE
: compare
= "greaterThanEqual"; break;
1852 default: compare
= "";
1853 FIXME("Can't handle opcode %s\n", arg
->opcode
->name
);
1856 shader_addline(arg
->buffer
, "vec%d(%s(%s, %s)));\n", mask_size
, compare
,
1857 src0_param
.param_str
, src1_param
.param_str
);
1859 switch(arg
->opcode
->opcode
) {
1860 case WINED3DSIO_SLT
:
1861 /* Step(src0, src1) is not suitable here because if src0 == src1 SLT is supposed,
1862 * to return 0.0 but step returns 1.0 because step is not < x
1863 * An alternative is a bvec compare padded with an unused second component.
1864 * step(src1 * -1.0, src0 * -1.0) is not an option because it suffers from the same
1865 * issue. Playing with not() is not possible either because not() does not accept
1868 shader_addline(arg
->buffer
, "(%s < %s) ? 1.0 : 0.0);\n", src0_param
.param_str
, src1_param
.param_str
);
1870 case WINED3DSIO_SGE
:
1871 /* Here we can use the step() function and safe a conditional */
1872 shader_addline(arg
->buffer
, "step(%s, %s));\n", src1_param
.param_str
, src0_param
.param_str
);
1875 FIXME("Can't handle opcode %s\n", arg
->opcode
->name
);
1881 /** Process CMP instruction in GLSL (dst = src0 >= 0.0 ? src1 : src2), per channel */
1882 static void shader_glsl_cmp(const SHADER_OPCODE_ARG
*arg
)
1884 glsl_src_param_t src0_param
;
1885 glsl_src_param_t src1_param
;
1886 glsl_src_param_t src2_param
;
1887 DWORD write_mask
, cmp_channel
= 0;
1890 BOOL temp_destination
= FALSE
;
1892 if(shader_is_scalar(arg
->src
[0])) {
1893 write_mask
= shader_glsl_append_dst(arg
->buffer
, arg
);
1895 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_ALL
, &src0_param
);
1896 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], write_mask
, &src1_param
);
1897 shader_glsl_add_src_param(arg
, arg
->src
[2], arg
->src_addr
[2], write_mask
, &src2_param
);
1899 shader_addline(arg
->buffer
, "%s >= 0.0 ? %s : %s);\n",
1900 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
1902 DWORD src0reg
= arg
->src
[0] & WINED3DSP_REGNUM_MASK
;
1903 DWORD src1reg
= arg
->src
[1] & WINED3DSP_REGNUM_MASK
;
1904 DWORD src2reg
= arg
->src
[2] & WINED3DSP_REGNUM_MASK
;
1905 DWORD src0regtype
= shader_get_regtype(arg
->src
[0]);
1906 DWORD src1regtype
= shader_get_regtype(arg
->src
[1]);
1907 DWORD src2regtype
= shader_get_regtype(arg
->src
[2]);
1908 DWORD dstreg
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
1909 DWORD dstregtype
= shader_get_regtype(arg
->dst
);
1911 /* Cycle through all source0 channels */
1912 for (i
=0; i
<4; i
++) {
1914 /* Find the destination channels which use the current source0 channel */
1915 for (j
=0; j
<4; j
++) {
1916 if ( ((arg
->src
[0] >> (WINED3DSP_SWIZZLE_SHIFT
+ 2*j
)) & 0x3) == i
) {
1917 write_mask
|= WINED3DSP_WRITEMASK_0
<< j
;
1918 cmp_channel
= WINED3DSP_WRITEMASK_0
<< j
;
1922 /* Splitting the cmp instruction up in multiple lines imposes a problem:
1923 * The first lines may overwrite source parameters of the following lines.
1924 * Deal with that by using a temporary destination register if needed
1926 if((src0reg
== dstreg
&& src0regtype
== dstregtype
) ||
1927 (src1reg
== dstreg
&& src1regtype
== dstregtype
) ||
1928 (src2reg
== dstreg
&& src2regtype
== dstregtype
)) {
1930 write_mask
= shader_glsl_get_write_mask(arg
->dst
& (~WINED3DSP_SWIZZLE_MASK
| write_mask
), mask_char
);
1931 if (!write_mask
) continue;
1932 shader_addline(arg
->buffer
, "tmp0%s = (", mask_char
);
1933 temp_destination
= TRUE
;
1935 write_mask
= shader_glsl_append_dst_ext(arg
->buffer
, arg
, arg
->dst
& (~WINED3DSP_SWIZZLE_MASK
| write_mask
));
1936 if (!write_mask
) continue;
1939 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], cmp_channel
, &src0_param
);
1940 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], write_mask
, &src1_param
);
1941 shader_glsl_add_src_param(arg
, arg
->src
[2], arg
->src_addr
[2], write_mask
, &src2_param
);
1943 shader_addline(arg
->buffer
, "%s >= 0.0 ? %s : %s);\n",
1944 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
1947 if(temp_destination
) {
1948 shader_glsl_get_write_mask(arg
->dst
, mask_char
);
1949 shader_glsl_append_dst_ext(arg
->buffer
, arg
, arg
->dst
);
1950 shader_addline(arg
->buffer
, "tmp0%s);\n", mask_char
);
1956 /** Process the CND opcode in GLSL (dst = (src0 > 0.5) ? src1 : src2) */
1957 /* For ps 1.1-1.3, only a single component of src0 is used. For ps 1.4
1958 * the compare is done per component of src0. */
1959 static void shader_glsl_cnd(const SHADER_OPCODE_ARG
*arg
)
1961 glsl_src_param_t src0_param
;
1962 glsl_src_param_t src1_param
;
1963 glsl_src_param_t src2_param
;
1964 DWORD write_mask
, cmp_channel
= 0;
1967 if (arg
->reg_maps
->shader_version
< WINED3DPS_VERSION(1, 4))
1969 write_mask
= shader_glsl_append_dst(arg
->buffer
, arg
);
1970 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
1971 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], write_mask
, &src1_param
);
1972 shader_glsl_add_src_param(arg
, arg
->src
[2], arg
->src_addr
[2], write_mask
, &src2_param
);
1974 /* Fun: The D3DSI_COISSUE flag changes the semantic of the cnd instruction for < 1.4 shaders */
1975 if(arg
->opcode_token
& WINED3DSI_COISSUE
) {
1976 shader_addline(arg
->buffer
, "%s /* COISSUE! */);\n", src1_param
.param_str
);
1978 shader_addline(arg
->buffer
, "%s > 0.5 ? %s : %s);\n",
1979 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
1983 /* Cycle through all source0 channels */
1984 for (i
=0; i
<4; i
++) {
1986 /* Find the destination channels which use the current source0 channel */
1987 for (j
=0; j
<4; j
++) {
1988 if ( ((arg
->src
[0] >> (WINED3DSP_SWIZZLE_SHIFT
+ 2*j
)) & 0x3) == i
) {
1989 write_mask
|= WINED3DSP_WRITEMASK_0
<< j
;
1990 cmp_channel
= WINED3DSP_WRITEMASK_0
<< j
;
1993 write_mask
= shader_glsl_append_dst_ext(arg
->buffer
, arg
, arg
->dst
& (~WINED3DSP_SWIZZLE_MASK
| write_mask
));
1994 if (!write_mask
) continue;
1996 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], cmp_channel
, &src0_param
);
1997 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], write_mask
, &src1_param
);
1998 shader_glsl_add_src_param(arg
, arg
->src
[2], arg
->src_addr
[2], write_mask
, &src2_param
);
2000 shader_addline(arg
->buffer
, "%s > 0.5 ? %s : %s);\n",
2001 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
2005 /** GLSL code generation for WINED3DSIO_MAD: Multiply the first 2 opcodes, then add the last */
2006 static void shader_glsl_mad(const SHADER_OPCODE_ARG
*arg
)
2008 glsl_src_param_t src0_param
;
2009 glsl_src_param_t src1_param
;
2010 glsl_src_param_t src2_param
;
2013 write_mask
= shader_glsl_append_dst(arg
->buffer
, arg
);
2014 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], write_mask
, &src0_param
);
2015 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], write_mask
, &src1_param
);
2016 shader_glsl_add_src_param(arg
, arg
->src
[2], arg
->src_addr
[2], write_mask
, &src2_param
);
2017 shader_addline(arg
->buffer
, "(%s * %s) + %s);\n",
2018 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
2021 /** Handles transforming all WINED3DSIO_M?x? opcodes for
2022 Vertex shaders to GLSL codes */
2023 static void shader_glsl_mnxn(const SHADER_OPCODE_ARG
*arg
)
2025 IWineD3DBaseShaderImpl
*shader
= (IWineD3DBaseShaderImpl
*)arg
->shader
;
2026 const SHADER_OPCODE
*opcode_table
= shader
->baseShader
.shader_ins
;
2027 DWORD shader_version
= arg
->reg_maps
->shader_version
;
2029 int nComponents
= 0;
2030 SHADER_OPCODE_ARG tmpArg
;
2032 memset(&tmpArg
, 0, sizeof(SHADER_OPCODE_ARG
));
2034 /* Set constants for the temporary argument */
2035 tmpArg
.shader
= arg
->shader
;
2036 tmpArg
.buffer
= arg
->buffer
;
2037 tmpArg
.src
[0] = arg
->src
[0];
2038 tmpArg
.src_addr
[0] = arg
->src_addr
[0];
2039 tmpArg
.src_addr
[1] = arg
->src_addr
[1];
2040 tmpArg
.reg_maps
= arg
->reg_maps
;
2042 switch(arg
->opcode
->opcode
) {
2043 case WINED3DSIO_M4x4
:
2045 tmpArg
.opcode
= shader_get_opcode(opcode_table
, shader_version
, WINED3DSIO_DP4
);
2047 case WINED3DSIO_M4x3
:
2049 tmpArg
.opcode
= shader_get_opcode(opcode_table
, shader_version
, WINED3DSIO_DP4
);
2051 case WINED3DSIO_M3x4
:
2053 tmpArg
.opcode
= shader_get_opcode(opcode_table
, shader_version
, WINED3DSIO_DP3
);
2055 case WINED3DSIO_M3x3
:
2057 tmpArg
.opcode
= shader_get_opcode(opcode_table
, shader_version
, WINED3DSIO_DP3
);
2059 case WINED3DSIO_M3x2
:
2061 tmpArg
.opcode
= shader_get_opcode(opcode_table
, shader_version
, WINED3DSIO_DP3
);
2067 for (i
= 0; i
< nComponents
; i
++) {
2068 tmpArg
.dst
= ((arg
->dst
) & ~WINED3DSP_WRITEMASK_ALL
)|(WINED3DSP_WRITEMASK_0
<<i
);
2069 tmpArg
.src
[1] = arg
->src
[1]+i
;
2070 shader_glsl_dot(&tmpArg
);
2075 The LRP instruction performs a component-wise linear interpolation
2076 between the second and third operands using the first operand as the
2077 blend factor. Equation: (dst = src2 + src0 * (src1 - src2))
2078 This is equivalent to mix(src2, src1, src0);
2080 static void shader_glsl_lrp(const SHADER_OPCODE_ARG
*arg
)
2082 glsl_src_param_t src0_param
;
2083 glsl_src_param_t src1_param
;
2084 glsl_src_param_t src2_param
;
2087 write_mask
= shader_glsl_append_dst(arg
->buffer
, arg
);
2089 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], write_mask
, &src0_param
);
2090 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], write_mask
, &src1_param
);
2091 shader_glsl_add_src_param(arg
, arg
->src
[2], arg
->src_addr
[2], write_mask
, &src2_param
);
2093 shader_addline(arg
->buffer
, "mix(%s, %s, %s));\n",
2094 src2_param
.param_str
, src1_param
.param_str
, src0_param
.param_str
);
2097 /** Process the WINED3DSIO_LIT instruction in GLSL:
2098 * dst.x = dst.w = 1.0
2099 * dst.y = (src0.x > 0) ? src0.x
2100 * dst.z = (src0.x > 0) ? ((src0.y > 0) ? pow(src0.y, src.w) : 0) : 0
2101 * where src.w is clamped at +- 128
2103 static void shader_glsl_lit(const SHADER_OPCODE_ARG
*arg
)
2105 glsl_src_param_t src0_param
;
2106 glsl_src_param_t src1_param
;
2107 glsl_src_param_t src3_param
;
2110 shader_glsl_append_dst(arg
->buffer
, arg
);
2111 shader_glsl_get_write_mask(arg
->dst
, dst_mask
);
2113 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2114 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_1
, &src1_param
);
2115 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_3
, &src3_param
);
2117 /* The sdk specifies the instruction like this
2119 * if(src.x > 0.0) dst.y = src.x
2121 * if(src.x > 0.0 && src.y > 0.0) dst.z = pow(src.y, power);
2125 * Obviously that has quite a few conditionals in it which we don't like. So the first step is this:
2126 * dst.x = 1.0 ... No further explanation needed
2127 * dst.y = max(src.y, 0.0); ... If x < 0.0, use 0.0, otherwise x. Same as the conditional
2128 * dst.z = x > 0.0 ? pow(max(y, 0.0), p) : 0; ... 0 ^ power is 0, and otherwise we use y anyway
2129 * dst.w = 1.0. ... Nothing fancy.
2131 * So we still have one conditional in there. So do this:
2132 * dst.z = pow(max(0.0, src.y) * step(0.0, src.x), power);
2134 * 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),
2135 * 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.
2136 * if both x and y are > 0, we get pow(y * 1.0, power), as it is supposed to
2138 shader_addline(arg
->buffer
, "vec4(1.0, max(%s, 0.0), pow(max(0.0, %s) * step(0.0, %s), clamp(%s, -128.0, 128.0)), 1.0)%s);\n",
2139 src0_param
.param_str
, src1_param
.param_str
, src0_param
.param_str
, src3_param
.param_str
, dst_mask
);
2142 /** Process the WINED3DSIO_DST instruction in GLSL:
2144 * dst.y = src0.x * src0.y
2148 static void shader_glsl_dst(const SHADER_OPCODE_ARG
*arg
)
2150 glsl_src_param_t src0y_param
;
2151 glsl_src_param_t src0z_param
;
2152 glsl_src_param_t src1y_param
;
2153 glsl_src_param_t src1w_param
;
2156 shader_glsl_append_dst(arg
->buffer
, arg
);
2157 shader_glsl_get_write_mask(arg
->dst
, dst_mask
);
2159 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_1
, &src0y_param
);
2160 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_2
, &src0z_param
);
2161 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], WINED3DSP_WRITEMASK_1
, &src1y_param
);
2162 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], WINED3DSP_WRITEMASK_3
, &src1w_param
);
2164 shader_addline(arg
->buffer
, "vec4(1.0, %s * %s, %s, %s))%s;\n",
2165 src0y_param
.param_str
, src1y_param
.param_str
, src0z_param
.param_str
, src1w_param
.param_str
, dst_mask
);
2168 /** Process the WINED3DSIO_SINCOS instruction in GLSL:
2169 * VS 2.0 requires that specific cosine and sine constants be passed to this instruction so the hardware
2170 * can handle it. But, these functions are built-in for GLSL, so we can just ignore the last 2 params.
2172 * dst.x = cos(src0.?)
2173 * dst.y = sin(src0.?)
2177 static void shader_glsl_sincos(const SHADER_OPCODE_ARG
*arg
)
2179 glsl_src_param_t src0_param
;
2182 write_mask
= shader_glsl_append_dst(arg
->buffer
, arg
);
2183 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2185 switch (write_mask
) {
2186 case WINED3DSP_WRITEMASK_0
:
2187 shader_addline(arg
->buffer
, "cos(%s));\n", src0_param
.param_str
);
2190 case WINED3DSP_WRITEMASK_1
:
2191 shader_addline(arg
->buffer
, "sin(%s));\n", src0_param
.param_str
);
2194 case (WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
):
2195 shader_addline(arg
->buffer
, "vec2(cos(%s), sin(%s)));\n", src0_param
.param_str
, src0_param
.param_str
);
2199 ERR("Write mask should be .x, .y or .xy\n");
2204 /** Process the WINED3DSIO_LOOP instruction in GLSL:
2205 * Start a for() loop where src1.y is the initial value of aL,
2206 * increment aL by src1.z for a total of src1.x iterations.
2207 * Need to use a temporary variable for this operation.
2209 /* FIXME: I don't think nested loops will work correctly this way. */
2210 static void shader_glsl_loop(const SHADER_OPCODE_ARG
*arg
)
2212 glsl_src_param_t src1_param
;
2213 IWineD3DBaseShaderImpl
* shader
= (IWineD3DBaseShaderImpl
*) arg
->shader
;
2214 DWORD regtype
= shader_get_regtype(arg
->src
[1]);
2215 DWORD reg
= arg
->src
[1] & WINED3DSP_REGNUM_MASK
;
2216 const DWORD
*control_values
= NULL
;
2217 const local_constant
*constant
;
2219 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], WINED3DSP_WRITEMASK_ALL
, &src1_param
);
2221 /* Try to hardcode the loop control parameters if possible. Direct3D 9 class hardware doesn't support real
2222 * varying indexing, but Microsoft designed this feature for Shader model 2.x+. If the loop control is
2223 * known at compile time, the GLSL compiler can unroll the loop, and replace indirect addressing with direct
2226 if(regtype
== WINED3DSPR_CONSTINT
) {
2227 LIST_FOR_EACH_ENTRY(constant
, &shader
->baseShader
.constantsI
, local_constant
, entry
) {
2228 if(constant
->idx
== reg
) {
2229 control_values
= constant
->value
;
2235 if(control_values
) {
2236 if(control_values
[2] > 0) {
2237 shader_addline(arg
->buffer
, "for (aL%u = %d; aL%u < (%d * %d + %d); aL%u += %d) {\n",
2238 shader
->baseShader
.cur_loop_depth
, control_values
[1],
2239 shader
->baseShader
.cur_loop_depth
, control_values
[0], control_values
[2], control_values
[1],
2240 shader
->baseShader
.cur_loop_depth
, control_values
[2]);
2241 } else if(control_values
[2] == 0) {
2242 shader_addline(arg
->buffer
, "for (aL%u = %d, tmpInt%u = 0; tmpInt%u < %d; tmpInt%u++) {\n",
2243 shader
->baseShader
.cur_loop_depth
, control_values
[1], shader
->baseShader
.cur_loop_depth
,
2244 shader
->baseShader
.cur_loop_depth
, control_values
[0],
2245 shader
->baseShader
.cur_loop_depth
);
2247 shader_addline(arg
->buffer
, "for (aL%u = %d; aL%u > (%d * %d + %d); aL%u += %d) {\n",
2248 shader
->baseShader
.cur_loop_depth
, control_values
[1],
2249 shader
->baseShader
.cur_loop_depth
, control_values
[0], control_values
[2], control_values
[1],
2250 shader
->baseShader
.cur_loop_depth
, control_values
[2]);
2253 shader_addline(arg
->buffer
, "for (tmpInt%u = 0, aL%u = %s.y; tmpInt%u < %s.x; tmpInt%u++, aL%u += %s.z) {\n",
2254 shader
->baseShader
.cur_loop_depth
, shader
->baseShader
.cur_loop_regno
,
2255 src1_param
.reg_name
, shader
->baseShader
.cur_loop_depth
, src1_param
.reg_name
,
2256 shader
->baseShader
.cur_loop_depth
, shader
->baseShader
.cur_loop_regno
, src1_param
.reg_name
);
2259 shader
->baseShader
.cur_loop_depth
++;
2260 shader
->baseShader
.cur_loop_regno
++;
2263 static void shader_glsl_end(const SHADER_OPCODE_ARG
*arg
)
2265 IWineD3DBaseShaderImpl
* shader
= (IWineD3DBaseShaderImpl
*) arg
->shader
;
2267 shader_addline(arg
->buffer
, "}\n");
2269 if(arg
->opcode
->opcode
== WINED3DSIO_ENDLOOP
) {
2270 shader
->baseShader
.cur_loop_depth
--;
2271 shader
->baseShader
.cur_loop_regno
--;
2273 if(arg
->opcode
->opcode
== WINED3DSIO_ENDREP
) {
2274 shader
->baseShader
.cur_loop_depth
--;
2278 static void shader_glsl_rep(const SHADER_OPCODE_ARG
*arg
)
2280 IWineD3DBaseShaderImpl
* shader
= (IWineD3DBaseShaderImpl
*) arg
->shader
;
2281 glsl_src_param_t src0_param
;
2283 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2284 shader_addline(arg
->buffer
, "for (tmpInt%d = 0; tmpInt%d < %s; tmpInt%d++) {\n",
2285 shader
->baseShader
.cur_loop_depth
, shader
->baseShader
.cur_loop_depth
,
2286 src0_param
.param_str
, shader
->baseShader
.cur_loop_depth
);
2287 shader
->baseShader
.cur_loop_depth
++;
2290 static void shader_glsl_if(const SHADER_OPCODE_ARG
*arg
)
2292 glsl_src_param_t src0_param
;
2294 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2295 shader_addline(arg
->buffer
, "if (%s) {\n", src0_param
.param_str
);
2298 static void shader_glsl_ifc(const SHADER_OPCODE_ARG
*arg
)
2300 glsl_src_param_t src0_param
;
2301 glsl_src_param_t src1_param
;
2303 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2304 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], WINED3DSP_WRITEMASK_0
, &src1_param
);
2306 shader_addline(arg
->buffer
, "if (%s %s %s) {\n",
2307 src0_param
.param_str
, shader_get_comp_op(arg
->opcode_token
), src1_param
.param_str
);
2310 static void shader_glsl_else(const SHADER_OPCODE_ARG
*arg
)
2312 shader_addline(arg
->buffer
, "} else {\n");
2315 static void shader_glsl_break(const SHADER_OPCODE_ARG
*arg
)
2317 shader_addline(arg
->buffer
, "break;\n");
2320 /* FIXME: According to MSDN the compare is done per component. */
2321 static void shader_glsl_breakc(const SHADER_OPCODE_ARG
*arg
)
2323 glsl_src_param_t src0_param
;
2324 glsl_src_param_t src1_param
;
2326 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2327 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], WINED3DSP_WRITEMASK_0
, &src1_param
);
2329 shader_addline(arg
->buffer
, "if (%s %s %s) break;\n",
2330 src0_param
.param_str
, shader_get_comp_op(arg
->opcode_token
), src1_param
.param_str
);
2333 static void shader_glsl_label(const SHADER_OPCODE_ARG
*arg
)
2336 DWORD snum
= (arg
->src
[0]) & WINED3DSP_REGNUM_MASK
;
2337 shader_addline(arg
->buffer
, "}\n");
2338 shader_addline(arg
->buffer
, "void subroutine%u () {\n", snum
);
2341 static void shader_glsl_call(const SHADER_OPCODE_ARG
*arg
)
2343 DWORD snum
= (arg
->src
[0]) & WINED3DSP_REGNUM_MASK
;
2344 shader_addline(arg
->buffer
, "subroutine%u();\n", snum
);
2347 static void shader_glsl_callnz(const SHADER_OPCODE_ARG
*arg
)
2349 glsl_src_param_t src1_param
;
2351 DWORD snum
= (arg
->src
[0]) & WINED3DSP_REGNUM_MASK
;
2352 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], WINED3DSP_WRITEMASK_0
, &src1_param
);
2353 shader_addline(arg
->buffer
, "if (%s) subroutine%u();\n", src1_param
.param_str
, snum
);
2356 /*********************************************
2357 * Pixel Shader Specific Code begins here
2358 ********************************************/
2359 static void pshader_glsl_tex(const SHADER_OPCODE_ARG
*arg
)
2361 IWineD3DPixelShaderImpl
* This
= (IWineD3DPixelShaderImpl
*) arg
->shader
;
2362 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
2363 DWORD shader_version
= arg
->reg_maps
->shader_version
;
2364 glsl_sample_function_t sample_function
;
2365 DWORD sample_flags
= 0;
2368 DWORD mask
= 0, swizzle
;
2370 /* 1.0-1.4: Use destination register as sampler source.
2371 * 2.0+: Use provided sampler source. */
2372 if (shader_version
< WINED3DPS_VERSION(2,0)) sampler_idx
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2373 else sampler_idx
= arg
->src
[1] & WINED3DSP_REGNUM_MASK
;
2374 sampler_type
= arg
->reg_maps
->samplers
[sampler_idx
] & WINED3DSP_TEXTURETYPE_MASK
;
2376 if (shader_version
< WINED3DPS_VERSION(1,4))
2378 DWORD flags
= deviceImpl
->stateBlock
->textureState
[sampler_idx
][WINED3DTSS_TEXTURETRANSFORMFLAGS
];
2380 /* Projected cube textures don't make a lot of sense, the resulting coordinates stay the same. */
2381 if (flags
& WINED3DTTFF_PROJECTED
&& sampler_type
!= WINED3DSTT_CUBE
) {
2382 sample_flags
|= WINED3D_GLSL_SAMPLE_PROJECTED
;
2383 switch (flags
& ~WINED3DTTFF_PROJECTED
) {
2384 case WINED3DTTFF_COUNT1
: FIXME("WINED3DTTFF_PROJECTED with WINED3DTTFF_COUNT1?\n"); break;
2385 case WINED3DTTFF_COUNT2
: mask
= WINED3DSP_WRITEMASK_1
; break;
2386 case WINED3DTTFF_COUNT3
: mask
= WINED3DSP_WRITEMASK_2
; break;
2387 case WINED3DTTFF_COUNT4
:
2388 case WINED3DTTFF_DISABLE
: mask
= WINED3DSP_WRITEMASK_3
; break;
2392 else if (shader_version
< WINED3DPS_VERSION(2,0))
2394 DWORD src_mod
= arg
->src
[0] & WINED3DSP_SRCMOD_MASK
;
2396 if (src_mod
== WINED3DSPSM_DZ
) {
2397 sample_flags
|= WINED3D_GLSL_SAMPLE_PROJECTED
;
2398 mask
= WINED3DSP_WRITEMASK_2
;
2399 } else if (src_mod
== WINED3DSPSM_DW
) {
2400 sample_flags
|= WINED3D_GLSL_SAMPLE_PROJECTED
;
2401 mask
= WINED3DSP_WRITEMASK_3
;
2404 if(arg
->opcode_token
& WINED3DSI_TEXLD_PROJECT
) {
2405 /* ps 2.0 texldp instruction always divides by the fourth component. */
2406 sample_flags
|= WINED3D_GLSL_SAMPLE_PROJECTED
;
2407 mask
= WINED3DSP_WRITEMASK_3
;
2411 if(deviceImpl
->stateBlock
->textures
[sampler_idx
] &&
2412 IWineD3DBaseTexture_GetTextureDimensions(deviceImpl
->stateBlock
->textures
[sampler_idx
]) == GL_TEXTURE_RECTANGLE_ARB
) {
2413 sample_flags
|= WINED3D_GLSL_SAMPLE_RECT
;
2416 shader_glsl_get_sample_function(sampler_type
, sample_flags
, &sample_function
);
2417 mask
|= sample_function
.coord_mask
;
2419 if (shader_version
< WINED3DPS_VERSION(2,0)) swizzle
= WINED3DVS_NOSWIZZLE
;
2420 else swizzle
= arg
->src
[1] & WINED3DSP_SWIZZLE_MASK
;
2422 /* 1.0-1.3: Use destination register as coordinate source.
2423 1.4+: Use provided coordinate source register. */
2424 if (shader_version
< WINED3DPS_VERSION(1,4))
2427 shader_glsl_get_write_mask(mask
, coord_mask
);
2428 shader_glsl_gen_sample_code(arg
, sampler_idx
, &sample_function
, swizzle
, NULL
,
2429 "T%u%s", sampler_idx
, coord_mask
);
2431 glsl_src_param_t coord_param
;
2432 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], mask
, &coord_param
);
2433 if(arg
->opcode_token
& WINED3DSI_TEXLD_BIAS
) {
2434 glsl_src_param_t bias
;
2435 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_3
, &bias
);
2436 shader_glsl_gen_sample_code(arg
, sampler_idx
, &sample_function
, swizzle
, bias
.param_str
,
2437 "%s", coord_param
.param_str
);
2439 shader_glsl_gen_sample_code(arg
, sampler_idx
, &sample_function
, swizzle
, NULL
,
2440 "%s", coord_param
.param_str
);
2445 static void shader_glsl_texldl(const SHADER_OPCODE_ARG
*arg
)
2447 IWineD3DBaseShaderImpl
* This
= (IWineD3DBaseShaderImpl
*)arg
->shader
;
2448 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
2449 glsl_sample_function_t sample_function
;
2450 glsl_src_param_t coord_param
, lod_param
;
2451 DWORD sample_flags
= WINED3D_GLSL_SAMPLE_LOD
;
2454 DWORD swizzle
= arg
->src
[1] & WINED3DSP_SWIZZLE_MASK
;
2456 sampler_idx
= arg
->src
[1] & WINED3DSP_REGNUM_MASK
;
2457 sampler_type
= arg
->reg_maps
->samplers
[sampler_idx
] & WINED3DSP_TEXTURETYPE_MASK
;
2458 if(deviceImpl
->stateBlock
->textures
[sampler_idx
] &&
2459 IWineD3DBaseTexture_GetTextureDimensions(deviceImpl
->stateBlock
->textures
[sampler_idx
]) == GL_TEXTURE_RECTANGLE_ARB
) {
2460 sample_flags
|= WINED3D_GLSL_SAMPLE_RECT
;
2462 shader_glsl_get_sample_function(sampler_type
, sample_flags
, &sample_function
);
2463 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], sample_function
.coord_mask
, &coord_param
);
2465 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_3
, &lod_param
);
2467 if (shader_is_pshader_version(arg
->reg_maps
->shader_version
))
2469 /* The GLSL spec claims the Lod sampling functions are only supported in vertex shaders.
2470 * However, they seem to work just fine in fragment shaders as well. */
2471 WARN("Using %s in fragment shader.\n", sample_function
.name
);
2473 shader_glsl_gen_sample_code(arg
, sampler_idx
, &sample_function
, swizzle
, lod_param
.param_str
,
2474 "%s", coord_param
.param_str
);
2477 static void pshader_glsl_texcoord(const SHADER_OPCODE_ARG
*arg
)
2479 /* FIXME: Make this work for more than just 2D textures */
2480 SHADER_BUFFER
* buffer
= arg
->buffer
;
2484 write_mask
= shader_glsl_append_dst(arg
->buffer
, arg
);
2485 shader_glsl_get_write_mask(write_mask
, dst_mask
);
2487 if (arg
->reg_maps
->shader_version
!= WINED3DPS_VERSION(1,4))
2489 DWORD reg
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2490 shader_addline(buffer
, "clamp(gl_TexCoord[%u], 0.0, 1.0)%s);\n", reg
, dst_mask
);
2492 DWORD reg
= arg
->src
[0] & WINED3DSP_REGNUM_MASK
;
2493 DWORD src_mod
= arg
->src
[0] & WINED3DSP_SRCMOD_MASK
;
2494 char dst_swizzle
[6];
2496 shader_glsl_get_swizzle(arg
->src
[0], FALSE
, write_mask
, dst_swizzle
);
2498 if (src_mod
== WINED3DSPSM_DZ
) {
2499 glsl_src_param_t div_param
;
2500 unsigned int mask_size
= shader_glsl_get_write_mask_size(write_mask
);
2501 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_2
, &div_param
);
2503 if (mask_size
> 1) {
2504 shader_addline(buffer
, "gl_TexCoord[%u]%s / vec%d(%s));\n", reg
, dst_swizzle
, mask_size
, div_param
.param_str
);
2506 shader_addline(buffer
, "gl_TexCoord[%u]%s / %s);\n", reg
, dst_swizzle
, div_param
.param_str
);
2508 } else if (src_mod
== WINED3DSPSM_DW
) {
2509 glsl_src_param_t div_param
;
2510 unsigned int mask_size
= shader_glsl_get_write_mask_size(write_mask
);
2511 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_3
, &div_param
);
2513 if (mask_size
> 1) {
2514 shader_addline(buffer
, "gl_TexCoord[%u]%s / vec%d(%s));\n", reg
, dst_swizzle
, mask_size
, div_param
.param_str
);
2516 shader_addline(buffer
, "gl_TexCoord[%u]%s / %s);\n", reg
, dst_swizzle
, div_param
.param_str
);
2519 shader_addline(buffer
, "gl_TexCoord[%u]%s);\n", reg
, dst_swizzle
);
2524 /** Process the WINED3DSIO_TEXDP3TEX instruction in GLSL:
2525 * Take a 3-component dot product of the TexCoord[dstreg] and src,
2526 * then perform a 1D texture lookup from stage dstregnum, place into dst. */
2527 static void pshader_glsl_texdp3tex(const SHADER_OPCODE_ARG
*arg
)
2529 glsl_src_param_t src0_param
;
2530 glsl_sample_function_t sample_function
;
2531 DWORD sampler_idx
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2532 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2533 DWORD sampler_type
= arg
->reg_maps
->samplers
[sampler_idx
] & WINED3DSP_TEXTURETYPE_MASK
;
2536 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], src_mask
, &src0_param
);
2538 /* Do I have to take care about the projected bit? I don't think so, since the dp3 returns only one
2539 * scalar, and projected sampling would require 4.
2541 * It is a dependent read - not valid with conditional NP2 textures
2543 shader_glsl_get_sample_function(sampler_type
, 0, &sample_function
);
2544 mask_size
= shader_glsl_get_write_mask_size(sample_function
.coord_mask
);
2549 shader_glsl_gen_sample_code(arg
, sampler_idx
, &sample_function
, WINED3DVS_NOSWIZZLE
, NULL
,
2550 "dot(gl_TexCoord[%u].xyz, %s)", sampler_idx
, src0_param
.param_str
);
2554 shader_glsl_gen_sample_code(arg
, sampler_idx
, &sample_function
, WINED3DVS_NOSWIZZLE
, NULL
,
2555 "vec2(dot(gl_TexCoord[%u].xyz, %s), 0.0)", sampler_idx
, src0_param
.param_str
);
2559 shader_glsl_gen_sample_code(arg
, sampler_idx
, &sample_function
, WINED3DVS_NOSWIZZLE
, NULL
,
2560 "vec3(dot(gl_TexCoord[%u].xyz, %s), 0.0, 0.0)", sampler_idx
, src0_param
.param_str
);
2564 FIXME("Unexpected mask size %u\n", mask_size
);
2569 /** Process the WINED3DSIO_TEXDP3 instruction in GLSL:
2570 * Take a 3-component dot product of the TexCoord[dstreg] and src. */
2571 static void pshader_glsl_texdp3(const SHADER_OPCODE_ARG
*arg
)
2573 glsl_src_param_t src0_param
;
2574 DWORD dstreg
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2575 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2577 unsigned int mask_size
;
2579 dst_mask
= shader_glsl_append_dst(arg
->buffer
, arg
);
2580 mask_size
= shader_glsl_get_write_mask_size(dst_mask
);
2581 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], src_mask
, &src0_param
);
2583 if (mask_size
> 1) {
2584 shader_addline(arg
->buffer
, "vec%d(dot(T%u.xyz, %s)));\n", mask_size
, dstreg
, src0_param
.param_str
);
2586 shader_addline(arg
->buffer
, "dot(T%u.xyz, %s));\n", dstreg
, src0_param
.param_str
);
2590 /** Process the WINED3DSIO_TEXDEPTH instruction in GLSL:
2591 * Calculate the depth as dst.x / dst.y */
2592 static void pshader_glsl_texdepth(const SHADER_OPCODE_ARG
*arg
)
2594 glsl_dst_param_t dst_param
;
2596 shader_glsl_add_dst_param(arg
, arg
->dst
, 0, &dst_param
);
2598 /* Tests show that texdepth never returns anything below 0.0, and that r5.y is clamped to 1.0.
2599 * Negative input is accepted, -0.25 / -0.5 returns 0.5. GL should clamp gl_FragDepth to [0;1], but
2600 * this doesn't always work, so clamp the results manually. Whether or not the x value is clamped at 1
2601 * too is irrelevant, since if x = 0, any y value < 1.0 (and > 1.0 is not allowed) results in a result
2604 shader_addline(arg
->buffer
, "gl_FragDepth = clamp((%s.x / min(%s.y, 1.0)), 0.0, 1.0);\n", dst_param
.reg_name
, dst_param
.reg_name
);
2607 /** Process the WINED3DSIO_TEXM3X2DEPTH instruction in GLSL:
2608 * Last row of a 3x2 matrix multiply, use the result to calculate the depth:
2609 * Calculate tmp0.y = TexCoord[dstreg] . src.xyz; (tmp0.x has already been calculated)
2610 * depth = (tmp0.y == 0.0) ? 1.0 : tmp0.x / tmp0.y
2612 static void pshader_glsl_texm3x2depth(const SHADER_OPCODE_ARG
*arg
)
2614 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2615 DWORD dstreg
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2616 glsl_src_param_t src0_param
;
2618 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], src_mask
, &src0_param
);
2620 shader_addline(arg
->buffer
, "tmp0.y = dot(T%u.xyz, %s);\n", dstreg
, src0_param
.param_str
);
2621 shader_addline(arg
->buffer
, "gl_FragDepth = (tmp0.y == 0.0) ? 1.0 : clamp(tmp0.x / tmp0.y, 0.0, 1.0);\n");
2624 /** Process the WINED3DSIO_TEXM3X2PAD instruction in GLSL
2625 * Calculate the 1st of a 2-row matrix multiplication. */
2626 static void pshader_glsl_texm3x2pad(const SHADER_OPCODE_ARG
*arg
)
2628 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2629 DWORD reg
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2630 SHADER_BUFFER
* buffer
= arg
->buffer
;
2631 glsl_src_param_t src0_param
;
2633 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], src_mask
, &src0_param
);
2634 shader_addline(buffer
, "tmp0.x = dot(T%u.xyz, %s);\n", reg
, src0_param
.param_str
);
2637 /** Process the WINED3DSIO_TEXM3X3PAD instruction in GLSL
2638 * Calculate the 1st or 2nd row of a 3-row matrix multiplication. */
2639 static void pshader_glsl_texm3x3pad(const SHADER_OPCODE_ARG
* arg
)
2641 IWineD3DPixelShaderImpl
* shader
= (IWineD3DPixelShaderImpl
*) arg
->shader
;
2642 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2643 DWORD reg
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2644 SHADER_BUFFER
* buffer
= arg
->buffer
;
2645 SHADER_PARSE_STATE
* current_state
= &shader
->baseShader
.parse_state
;
2646 glsl_src_param_t src0_param
;
2648 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], src_mask
, &src0_param
);
2649 shader_addline(buffer
, "tmp0.%c = dot(T%u.xyz, %s);\n", 'x' + current_state
->current_row
, reg
, src0_param
.param_str
);
2650 current_state
->texcoord_w
[current_state
->current_row
++] = reg
;
2653 static void pshader_glsl_texm3x2tex(const SHADER_OPCODE_ARG
*arg
)
2655 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2656 DWORD reg
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2657 SHADER_BUFFER
* buffer
= arg
->buffer
;
2658 glsl_src_param_t src0_param
;
2659 DWORD sampler_type
= arg
->reg_maps
->samplers
[reg
] & WINED3DSP_TEXTURETYPE_MASK
;
2660 glsl_sample_function_t sample_function
;
2662 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], src_mask
, &src0_param
);
2663 shader_addline(buffer
, "tmp0.y = dot(T%u.xyz, %s);\n", reg
, src0_param
.param_str
);
2665 shader_glsl_get_sample_function(sampler_type
, 0, &sample_function
);
2667 /* Sample the texture using the calculated coordinates */
2668 shader_glsl_gen_sample_code(arg
, reg
, &sample_function
, WINED3DVS_NOSWIZZLE
, NULL
, "tmp0.xy");
2671 /** Process the WINED3DSIO_TEXM3X3TEX instruction in GLSL
2672 * Perform the 3rd row of a 3x3 matrix multiply, then sample the texture using the calculated coordinates */
2673 static void pshader_glsl_texm3x3tex(const SHADER_OPCODE_ARG
*arg
)
2675 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2676 glsl_src_param_t src0_param
;
2677 DWORD reg
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2678 IWineD3DPixelShaderImpl
* This
= (IWineD3DPixelShaderImpl
*) arg
->shader
;
2679 SHADER_PARSE_STATE
* current_state
= &This
->baseShader
.parse_state
;
2680 DWORD sampler_type
= arg
->reg_maps
->samplers
[reg
] & WINED3DSP_TEXTURETYPE_MASK
;
2681 glsl_sample_function_t sample_function
;
2683 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], src_mask
, &src0_param
);
2684 shader_addline(arg
->buffer
, "tmp0.z = dot(T%u.xyz, %s);\n", reg
, src0_param
.param_str
);
2686 /* Dependent read, not valid with conditional NP2 */
2687 shader_glsl_get_sample_function(sampler_type
, 0, &sample_function
);
2689 /* Sample the texture using the calculated coordinates */
2690 shader_glsl_gen_sample_code(arg
, reg
, &sample_function
, WINED3DVS_NOSWIZZLE
, NULL
, "tmp0.xyz");
2692 current_state
->current_row
= 0;
2695 /** Process the WINED3DSIO_TEXM3X3 instruction in GLSL
2696 * Perform the 3rd row of a 3x3 matrix multiply */
2697 static void pshader_glsl_texm3x3(const SHADER_OPCODE_ARG
*arg
)
2699 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2700 glsl_src_param_t src0_param
;
2702 DWORD reg
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2703 IWineD3DPixelShaderImpl
* This
= (IWineD3DPixelShaderImpl
*) arg
->shader
;
2704 SHADER_PARSE_STATE
* current_state
= &This
->baseShader
.parse_state
;
2706 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], src_mask
, &src0_param
);
2708 shader_glsl_append_dst(arg
->buffer
, arg
);
2709 shader_glsl_get_write_mask(arg
->dst
, dst_mask
);
2710 shader_addline(arg
->buffer
, "vec4(tmp0.xy, dot(T%u.xyz, %s), 1.0)%s);\n", reg
, src0_param
.param_str
, dst_mask
);
2712 current_state
->current_row
= 0;
2715 /** Process the WINED3DSIO_TEXM3X3SPEC instruction in GLSL
2716 * Perform the final texture lookup based on the previous 2 3x3 matrix multiplies */
2717 static void pshader_glsl_texm3x3spec(const SHADER_OPCODE_ARG
*arg
)
2719 IWineD3DPixelShaderImpl
* shader
= (IWineD3DPixelShaderImpl
*) arg
->shader
;
2720 DWORD reg
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2721 glsl_src_param_t src0_param
;
2722 glsl_src_param_t src1_param
;
2723 SHADER_BUFFER
* buffer
= arg
->buffer
;
2724 SHADER_PARSE_STATE
* current_state
= &shader
->baseShader
.parse_state
;
2725 DWORD stype
= arg
->reg_maps
->samplers
[reg
] & WINED3DSP_TEXTURETYPE_MASK
;
2726 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2727 glsl_sample_function_t sample_function
;
2729 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], src_mask
, &src0_param
);
2730 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], src_mask
, &src1_param
);
2732 /* Perform the last matrix multiply operation */
2733 shader_addline(buffer
, "tmp0.z = dot(T%u.xyz, %s);\n", reg
, src0_param
.param_str
);
2734 /* Reflection calculation */
2735 shader_addline(buffer
, "tmp0.xyz = -reflect((%s), normalize(tmp0.xyz));\n", src1_param
.param_str
);
2737 /* Dependent read, not valid with conditional NP2 */
2738 shader_glsl_get_sample_function(stype
, 0, &sample_function
);
2740 /* Sample the texture */
2741 shader_glsl_gen_sample_code(arg
, reg
, &sample_function
, WINED3DVS_NOSWIZZLE
, NULL
, "tmp0.xyz");
2743 current_state
->current_row
= 0;
2746 /** Process the WINED3DSIO_TEXM3X3VSPEC instruction in GLSL
2747 * Perform the final texture lookup based on the previous 2 3x3 matrix multiplies */
2748 static void pshader_glsl_texm3x3vspec(const SHADER_OPCODE_ARG
*arg
)
2750 IWineD3DPixelShaderImpl
* shader
= (IWineD3DPixelShaderImpl
*) arg
->shader
;
2751 DWORD reg
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2752 SHADER_BUFFER
* buffer
= arg
->buffer
;
2753 SHADER_PARSE_STATE
* current_state
= &shader
->baseShader
.parse_state
;
2754 glsl_src_param_t src0_param
;
2755 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2756 DWORD sampler_type
= arg
->reg_maps
->samplers
[reg
] & WINED3DSP_TEXTURETYPE_MASK
;
2757 glsl_sample_function_t sample_function
;
2759 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], src_mask
, &src0_param
);
2761 /* Perform the last matrix multiply operation */
2762 shader_addline(buffer
, "tmp0.z = dot(vec3(T%u), vec3(%s));\n", reg
, src0_param
.param_str
);
2764 /* Construct the eye-ray vector from w coordinates */
2765 shader_addline(buffer
, "tmp1.xyz = normalize(vec3(gl_TexCoord[%u].w, gl_TexCoord[%u].w, gl_TexCoord[%u].w));\n",
2766 current_state
->texcoord_w
[0], current_state
->texcoord_w
[1], reg
);
2767 shader_addline(buffer
, "tmp0.xyz = -reflect(tmp1.xyz, normalize(tmp0.xyz));\n");
2769 /* Dependent read, not valid with conditional NP2 */
2770 shader_glsl_get_sample_function(sampler_type
, 0, &sample_function
);
2772 /* Sample the texture using the calculated coordinates */
2773 shader_glsl_gen_sample_code(arg
, reg
, &sample_function
, WINED3DVS_NOSWIZZLE
, NULL
, "tmp0.xyz");
2775 current_state
->current_row
= 0;
2778 /** Process the WINED3DSIO_TEXBEM instruction in GLSL.
2779 * Apply a fake bump map transform.
2780 * texbem is pshader <= 1.3 only, this saves a few version checks
2782 static void pshader_glsl_texbem(const SHADER_OPCODE_ARG
*arg
)
2784 IWineD3DPixelShaderImpl
* This
= (IWineD3DPixelShaderImpl
*) arg
->shader
;
2785 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
2786 glsl_sample_function_t sample_function
;
2787 glsl_src_param_t coord_param
;
2794 sampler_idx
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2795 flags
= deviceImpl
->stateBlock
->textureState
[sampler_idx
][WINED3DTSS_TEXTURETRANSFORMFLAGS
];
2797 sampler_type
= arg
->reg_maps
->samplers
[sampler_idx
] & WINED3DSP_TEXTURETYPE_MASK
;
2798 /* Dependent read, not valid with conditional NP2 */
2799 shader_glsl_get_sample_function(sampler_type
, 0, &sample_function
);
2800 mask
= sample_function
.coord_mask
;
2802 shader_glsl_get_write_mask(mask
, coord_mask
);
2804 /* with projective textures, texbem only divides the static texture coord, not the displacement,
2805 * so we can't let the GL handle this.
2807 if (flags
& WINED3DTTFF_PROJECTED
) {
2809 char coord_div_mask
[3];
2810 switch (flags
& ~WINED3DTTFF_PROJECTED
) {
2811 case WINED3DTTFF_COUNT1
: FIXME("WINED3DTTFF_PROJECTED with WINED3DTTFF_COUNT1?\n"); break;
2812 case WINED3DTTFF_COUNT2
: div_mask
= WINED3DSP_WRITEMASK_1
; break;
2813 case WINED3DTTFF_COUNT3
: div_mask
= WINED3DSP_WRITEMASK_2
; break;
2814 case WINED3DTTFF_COUNT4
:
2815 case WINED3DTTFF_DISABLE
: div_mask
= WINED3DSP_WRITEMASK_3
; break;
2817 shader_glsl_get_write_mask(div_mask
, coord_div_mask
);
2818 shader_addline(arg
->buffer
, "T%u%s /= T%u%s;\n", sampler_idx
, coord_mask
, sampler_idx
, coord_div_mask
);
2821 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_0
|WINED3DSP_WRITEMASK_1
, &coord_param
);
2823 shader_glsl_gen_sample_code(arg
, sampler_idx
, &sample_function
, WINED3DVS_NOSWIZZLE
, NULL
,
2824 "T%u%s + vec4(bumpenvmat%d * %s, 0.0, 0.0)%s", sampler_idx
, coord_mask
, sampler_idx
,
2825 coord_param
.param_str
, coord_mask
);
2827 if(arg
->opcode
->opcode
== WINED3DSIO_TEXBEML
) {
2828 glsl_src_param_t luminance_param
;
2829 glsl_dst_param_t dst_param
;
2831 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_2
, &luminance_param
);
2832 shader_glsl_add_dst_param(arg
, arg
->dst
, arg
->dst_addr
, &dst_param
);
2834 shader_addline(arg
->buffer
, "%s%s *= (%s * luminancescale%d + luminanceoffset%d);\n",
2835 dst_param
.reg_name
, dst_param
.mask_str
,
2836 luminance_param
.param_str
, sampler_idx
, sampler_idx
);
2840 static void pshader_glsl_bem(const SHADER_OPCODE_ARG
*arg
)
2842 glsl_src_param_t src0_param
, src1_param
;
2843 DWORD sampler_idx
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2845 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_0
|WINED3DSP_WRITEMASK_1
, &src0_param
);
2846 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], WINED3DSP_WRITEMASK_0
|WINED3DSP_WRITEMASK_1
, &src1_param
);
2848 shader_glsl_append_dst(arg
->buffer
, arg
);
2849 shader_addline(arg
->buffer
, "%s + bumpenvmat%d * %s);\n",
2850 src0_param
.param_str
, sampler_idx
, src1_param
.param_str
);
2853 /** Process the WINED3DSIO_TEXREG2AR instruction in GLSL
2854 * Sample 2D texture at dst using the alpha & red (wx) components of src as texture coordinates */
2855 static void pshader_glsl_texreg2ar(const SHADER_OPCODE_ARG
*arg
)
2857 glsl_src_param_t src0_param
;
2858 DWORD sampler_idx
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2859 DWORD sampler_type
= arg
->reg_maps
->samplers
[sampler_idx
] & WINED3DSP_TEXTURETYPE_MASK
;
2860 glsl_sample_function_t sample_function
;
2862 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_ALL
, &src0_param
);
2864 shader_glsl_get_sample_function(sampler_type
, 0, &sample_function
);
2865 shader_glsl_gen_sample_code(arg
, sampler_idx
, &sample_function
, WINED3DVS_NOSWIZZLE
, NULL
,
2866 "%s.wx", src0_param
.reg_name
);
2869 /** Process the WINED3DSIO_TEXREG2GB instruction in GLSL
2870 * Sample 2D texture at dst using the green & blue (yz) components of src as texture coordinates */
2871 static void pshader_glsl_texreg2gb(const SHADER_OPCODE_ARG
*arg
)
2873 glsl_src_param_t src0_param
;
2874 DWORD sampler_idx
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2875 DWORD sampler_type
= arg
->reg_maps
->samplers
[sampler_idx
] & WINED3DSP_TEXTURETYPE_MASK
;
2876 glsl_sample_function_t sample_function
;
2878 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_ALL
, &src0_param
);
2880 shader_glsl_get_sample_function(sampler_type
, 0, &sample_function
);
2881 shader_glsl_gen_sample_code(arg
, sampler_idx
, &sample_function
, WINED3DVS_NOSWIZZLE
, NULL
,
2882 "%s.yz", src0_param
.reg_name
);
2885 /** Process the WINED3DSIO_TEXREG2RGB instruction in GLSL
2886 * Sample texture at dst using the rgb (xyz) components of src as texture coordinates */
2887 static void pshader_glsl_texreg2rgb(const SHADER_OPCODE_ARG
*arg
)
2889 glsl_src_param_t src0_param
;
2890 DWORD sampler_idx
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2891 DWORD sampler_type
= arg
->reg_maps
->samplers
[sampler_idx
] & WINED3DSP_TEXTURETYPE_MASK
;
2892 glsl_sample_function_t sample_function
;
2894 /* Dependent read, not valid with conditional NP2 */
2895 shader_glsl_get_sample_function(sampler_type
, 0, &sample_function
);
2896 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], sample_function
.coord_mask
, &src0_param
);
2898 shader_glsl_gen_sample_code(arg
, sampler_idx
, &sample_function
, WINED3DVS_NOSWIZZLE
, NULL
,
2899 "%s", src0_param
.param_str
);
2902 /** Process the WINED3DSIO_TEXKILL instruction in GLSL.
2903 * If any of the first 3 components are < 0, discard this pixel */
2904 static void pshader_glsl_texkill(const SHADER_OPCODE_ARG
*arg
)
2906 glsl_dst_param_t dst_param
;
2908 /* The argument is a destination parameter, and no writemasks are allowed */
2909 shader_glsl_add_dst_param(arg
, arg
->dst
, 0, &dst_param
);
2910 if ((arg
->reg_maps
->shader_version
>= WINED3DPS_VERSION(2,0)))
2912 /* 2.0 shaders compare all 4 components in texkill */
2913 shader_addline(arg
->buffer
, "if (any(lessThan(%s.xyzw, vec4(0.0)))) discard;\n", dst_param
.reg_name
);
2915 /* 1.X shaders only compare the first 3 components, probably due to the nature of the texkill
2916 * instruction as a tex* instruction, and phase, which kills all a / w components. Even if all
2917 * 4 components are defined, only the first 3 are used
2919 shader_addline(arg
->buffer
, "if (any(lessThan(%s.xyz, vec3(0.0)))) discard;\n", dst_param
.reg_name
);
2923 /** Process the WINED3DSIO_DP2ADD instruction in GLSL.
2924 * dst = dot2(src0, src1) + src2 */
2925 static void pshader_glsl_dp2add(const SHADER_OPCODE_ARG
*arg
)
2927 glsl_src_param_t src0_param
;
2928 glsl_src_param_t src1_param
;
2929 glsl_src_param_t src2_param
;
2931 unsigned int mask_size
;
2933 write_mask
= shader_glsl_append_dst(arg
->buffer
, arg
);
2934 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
2936 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
, &src0_param
);
2937 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
, &src1_param
);
2938 shader_glsl_add_src_param(arg
, arg
->src
[2], arg
->src_addr
[2], WINED3DSP_WRITEMASK_0
, &src2_param
);
2940 if (mask_size
> 1) {
2941 shader_addline(arg
->buffer
, "vec%d(dot(%s, %s) + %s));\n", mask_size
, src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
2943 shader_addline(arg
->buffer
, "dot(%s, %s) + %s);\n", src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
2947 static void pshader_glsl_input_pack(SHADER_BUFFER
* buffer
, const struct semantic
* semantics_in
,
2948 IWineD3DPixelShader
*iface
, enum vertexprocessing_mode vertexprocessing
)
2951 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*) iface
;
2953 for (i
= 0; i
< MAX_REG_INPUT
; i
++) {
2955 DWORD usage_token
= semantics_in
[i
].usage
;
2956 DWORD register_token
= semantics_in
[i
].reg
;
2957 DWORD usage
, usage_idx
;
2961 if (!usage_token
) continue;
2962 usage
= (usage_token
& WINED3DSP_DCL_USAGE_MASK
) >> WINED3DSP_DCL_USAGE_SHIFT
;
2963 usage_idx
= (usage_token
& WINED3DSP_DCL_USAGEINDEX_MASK
) >> WINED3DSP_DCL_USAGEINDEX_SHIFT
;
2964 shader_glsl_get_write_mask(register_token
, reg_mask
);
2968 case WINED3DDECLUSAGE_TEXCOORD
:
2969 if(usage_idx
< 8 && vertexprocessing
== pretransformed
) {
2970 shader_addline(buffer
, "IN[%u]%s = gl_TexCoord[%u]%s;\n",
2971 This
->input_reg_map
[i
], reg_mask
, usage_idx
, reg_mask
);
2973 shader_addline(buffer
, "IN[%u]%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
2974 This
->input_reg_map
[i
], reg_mask
, reg_mask
);
2978 case WINED3DDECLUSAGE_COLOR
:
2980 shader_addline(buffer
, "IN[%u]%s = vec4(gl_Color)%s;\n",
2981 This
->input_reg_map
[i
], reg_mask
, reg_mask
);
2982 else if (usage_idx
== 1)
2983 shader_addline(buffer
, "IN[%u]%s = vec4(gl_SecondaryColor)%s;\n",
2984 This
->input_reg_map
[i
], reg_mask
, reg_mask
);
2986 shader_addline(buffer
, "IN[%u]%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
2987 This
->input_reg_map
[i
], reg_mask
, reg_mask
);
2991 shader_addline(buffer
, "IN[%u]%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
2992 This
->input_reg_map
[i
], reg_mask
, reg_mask
);
2997 /*********************************************
2998 * Vertex Shader Specific Code begins here
2999 ********************************************/
3001 static void add_glsl_program_entry(struct shader_glsl_priv
*priv
, struct glsl_shader_prog_link
*entry
) {
3002 glsl_program_key_t
*key
;
3004 key
= HeapAlloc(GetProcessHeap(), 0, sizeof(glsl_program_key_t
));
3005 key
->vshader
= entry
->vshader
;
3006 key
->pshader
= entry
->pshader
;
3007 key
->vs_args
= entry
->vs_args
;
3008 key
->ps_args
= entry
->ps_args
;
3010 hash_table_put(priv
->glsl_program_lookup
, key
, entry
);
3013 static struct glsl_shader_prog_link
*get_glsl_program_entry(struct shader_glsl_priv
*priv
,
3014 IWineD3DVertexShader
*vshader
, IWineD3DPixelShader
*pshader
, struct vs_compile_args
*vs_args
,
3015 struct ps_compile_args
*ps_args
) {
3016 glsl_program_key_t key
;
3018 key
.vshader
= vshader
;
3019 key
.pshader
= pshader
;
3020 key
.vs_args
= *vs_args
;
3021 key
.ps_args
= *ps_args
;
3023 return hash_table_get(priv
->glsl_program_lookup
, &key
);
3026 static void delete_glsl_program_entry(struct shader_glsl_priv
*priv
, const WineD3D_GL_Info
*gl_info
,
3027 struct glsl_shader_prog_link
*entry
)
3029 glsl_program_key_t
*key
;
3031 key
= HeapAlloc(GetProcessHeap(), 0, sizeof(glsl_program_key_t
));
3032 key
->vshader
= entry
->vshader
;
3033 key
->pshader
= entry
->pshader
;
3034 key
->vs_args
= entry
->vs_args
;
3035 key
->ps_args
= entry
->ps_args
;
3036 hash_table_remove(priv
->glsl_program_lookup
, key
);
3038 GL_EXTCALL(glDeleteObjectARB(entry
->programId
));
3039 if (entry
->vshader
) list_remove(&entry
->vshader_entry
);
3040 if (entry
->pshader
) list_remove(&entry
->pshader_entry
);
3041 HeapFree(GetProcessHeap(), 0, entry
->vuniformF_locations
);
3042 HeapFree(GetProcessHeap(), 0, entry
->puniformF_locations
);
3043 HeapFree(GetProcessHeap(), 0, entry
);
3046 static void handle_ps3_input(SHADER_BUFFER
*buffer
, const struct semantic
*semantics_in
,
3047 const struct semantic
*semantics_out
, const WineD3D_GL_Info
*gl_info
, const DWORD
*map
)
3050 DWORD usage_token
, usage_token_out
;
3051 DWORD register_token
, register_token_out
;
3052 DWORD usage
, usage_idx
, usage_out
, usage_idx_out
;
3055 DWORD in_count
= GL_LIMITS(glsl_varyings
) / 4;
3056 char reg_mask
[6], reg_mask_out
[6];
3057 char destination
[50];
3059 set
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*set
) * (in_count
+ 2));
3061 if (!semantics_out
) {
3062 /* Save gl_FrontColor & gl_FrontSecondaryColor before overwriting them. */
3063 shader_addline(buffer
, "vec4 front_color = gl_FrontColor;\n");
3064 shader_addline(buffer
, "vec4 front_secondary_color = gl_FrontSecondaryColor;\n");
3067 for(i
= 0; i
< MAX_REG_INPUT
; i
++) {
3068 usage_token
= semantics_in
[i
].usage
;
3069 if (!usage_token
) continue;
3072 if (in_idx
>= (in_count
+ 2)) {
3073 FIXME("More input varyings declared than supported, expect issues\n");
3076 else if (map
[i
] == ~0U)
3078 /* Declared, but not read register */
3082 if (in_idx
== in_count
) {
3083 sprintf(destination
, "gl_FrontColor");
3084 } else if (in_idx
== in_count
+ 1) {
3085 sprintf(destination
, "gl_FrontSecondaryColor");
3087 sprintf(destination
, "IN[%u]", in_idx
);
3090 register_token
= semantics_in
[i
].reg
;
3092 usage
= (usage_token
& WINED3DSP_DCL_USAGE_MASK
) >> WINED3DSP_DCL_USAGE_SHIFT
;
3093 usage_idx
= (usage_token
& WINED3DSP_DCL_USAGEINDEX_MASK
) >> WINED3DSP_DCL_USAGEINDEX_SHIFT
;
3094 set
[map
[i
]] = shader_glsl_get_write_mask(register_token
, reg_mask
);
3096 if(!semantics_out
) {
3098 case WINED3DDECLUSAGE_COLOR
:
3100 shader_addline(buffer
, "%s%s = front_color%s;\n",
3101 destination
, reg_mask
, reg_mask
);
3102 else if (usage_idx
== 1)
3103 shader_addline(buffer
, "%s%s = front_secondary_color%s;\n",
3104 destination
, reg_mask
, reg_mask
);
3106 shader_addline(buffer
, "%s%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
3107 destination
, reg_mask
, reg_mask
);
3110 case WINED3DDECLUSAGE_TEXCOORD
:
3111 if (usage_idx
< 8) {
3112 shader_addline(buffer
, "%s%s = gl_TexCoord[%u]%s;\n",
3113 destination
, reg_mask
, usage_idx
, reg_mask
);
3115 shader_addline(buffer
, "%s%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
3116 destination
, reg_mask
, reg_mask
);
3120 case WINED3DDECLUSAGE_FOG
:
3121 shader_addline(buffer
, "%s%s = vec4(gl_FogFragCoord, 0.0, 0.0, 0.0)%s;\n",
3122 destination
, reg_mask
, reg_mask
);
3126 shader_addline(buffer
, "%s%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
3127 destination
, reg_mask
, reg_mask
);
3131 for(j
= 0; j
< MAX_REG_OUTPUT
; j
++) {
3132 usage_token_out
= semantics_out
[j
].usage
;
3133 if (!usage_token_out
) continue;
3134 register_token_out
= semantics_out
[j
].reg
;
3136 usage_out
= (usage_token_out
& WINED3DSP_DCL_USAGE_MASK
) >> WINED3DSP_DCL_USAGE_SHIFT
;
3137 usage_idx_out
= (usage_token_out
& WINED3DSP_DCL_USAGEINDEX_MASK
) >> WINED3DSP_DCL_USAGEINDEX_SHIFT
;
3138 shader_glsl_get_write_mask(register_token_out
, reg_mask_out
);
3140 if(usage
== usage_out
&&
3141 usage_idx
== usage_idx_out
) {
3142 shader_addline(buffer
, "%s%s = OUT[%u]%s;\n",
3143 destination
, reg_mask
, j
, reg_mask
);
3148 shader_addline(buffer
, "%s%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
3149 destination
, reg_mask
, reg_mask
);
3154 /* This is solely to make the compiler / linker happy and avoid warning about undefined
3155 * varyings. It shouldn't result in any real code executed on the GPU, since all read
3156 * input varyings are assigned above, if the optimizer works properly.
3158 for(i
= 0; i
< in_count
+ 2; i
++) {
3159 if(set
[i
] != WINED3DSP_WRITEMASK_ALL
) {
3160 unsigned int size
= 0;
3161 memset(reg_mask
, 0, sizeof(reg_mask
));
3162 if(!(set
[i
] & WINED3DSP_WRITEMASK_0
)) {
3163 reg_mask
[size
] = 'x';
3166 if(!(set
[i
] & WINED3DSP_WRITEMASK_1
)) {
3167 reg_mask
[size
] = 'y';
3170 if(!(set
[i
] & WINED3DSP_WRITEMASK_2
)) {
3171 reg_mask
[size
] = 'z';
3174 if(!(set
[i
] & WINED3DSP_WRITEMASK_3
)) {
3175 reg_mask
[size
] = 'w';
3179 if (i
== in_count
) {
3180 sprintf(destination
, "gl_FrontColor");
3181 } else if (i
== in_count
+ 1) {
3182 sprintf(destination
, "gl_FrontSecondaryColor");
3184 sprintf(destination
, "IN[%u]", i
);
3188 shader_addline(buffer
, "%s.%s = 0.0;\n", destination
, reg_mask
);
3190 shader_addline(buffer
, "%s.%s = vec%u(0.0);\n", destination
, reg_mask
, size
);
3195 HeapFree(GetProcessHeap(), 0, set
);
3198 static GLhandleARB
generate_param_reorder_function(IWineD3DVertexShader
*vertexshader
,
3199 IWineD3DPixelShader
*pixelshader
, const WineD3D_GL_Info
*gl_info
)
3201 GLhandleARB ret
= 0;
3202 IWineD3DVertexShaderImpl
*vs
= (IWineD3DVertexShaderImpl
*) vertexshader
;
3203 IWineD3DPixelShaderImpl
*ps
= (IWineD3DPixelShaderImpl
*) pixelshader
;
3204 IWineD3DDeviceImpl
*device
;
3205 DWORD vs_major
= WINED3DSHADER_VERSION_MAJOR(vs
->baseShader
.reg_maps
.shader_version
);
3206 DWORD ps_major
= ps
? WINED3DSHADER_VERSION_MAJOR(ps
->baseShader
.reg_maps
.shader_version
) : 0;
3208 SHADER_BUFFER buffer
;
3210 DWORD register_token
;
3211 DWORD usage
, usage_idx
, writemask
;
3213 const struct semantic
*semantics_out
, *semantics_in
;
3215 shader_buffer_init(&buffer
);
3217 shader_addline(&buffer
, "#version 120\n");
3219 if(vs_major
< 3 && ps_major
< 3) {
3220 /* That one is easy: The vertex shader writes to the builtin varyings, the pixel shader reads from them.
3221 * Take care about the texcoord .w fixup though if we're using the fixed function fragment pipeline
3223 device
= (IWineD3DDeviceImpl
*) vs
->baseShader
.device
;
3224 if((GLINFO_LOCATION
).set_texcoord_w
&& ps_major
== 0 && vs_major
> 0 &&
3225 !device
->frag_pipe
->ffp_proj_control
) {
3226 shader_addline(&buffer
, "void order_ps_input() {\n");
3227 for(i
= 0; i
< min(8, MAX_REG_TEXCRD
); i
++) {
3228 if(vs
->baseShader
.reg_maps
.texcoord_mask
[i
] != 0 &&
3229 vs
->baseShader
.reg_maps
.texcoord_mask
[i
] != WINED3DSP_WRITEMASK_ALL
) {
3230 shader_addline(&buffer
, "gl_TexCoord[%u].w = 1.0;\n", i
);
3233 shader_addline(&buffer
, "}\n");
3235 shader_addline(&buffer
, "void order_ps_input() { /* do nothing */ }\n");
3237 } else if(ps_major
< 3 && vs_major
>= 3) {
3238 /* The vertex shader writes to its own varyings, the pixel shader needs them in the builtin ones */
3239 semantics_out
= vs
->semantics_out
;
3241 shader_addline(&buffer
, "void order_ps_input(in vec4 OUT[%u]) {\n", MAX_REG_OUTPUT
);
3242 for(i
= 0; i
< MAX_REG_OUTPUT
; i
++) {
3243 usage_token
= semantics_out
[i
].usage
;
3244 if (!usage_token
) continue;
3245 register_token
= semantics_out
[i
].reg
;
3247 usage
= (usage_token
& WINED3DSP_DCL_USAGE_MASK
) >> WINED3DSP_DCL_USAGE_SHIFT
;
3248 usage_idx
= (usage_token
& WINED3DSP_DCL_USAGEINDEX_MASK
) >> WINED3DSP_DCL_USAGEINDEX_SHIFT
;
3249 writemask
= shader_glsl_get_write_mask(register_token
, reg_mask
);
3252 case WINED3DDECLUSAGE_COLOR
:
3254 shader_addline(&buffer
, "gl_FrontColor%s = OUT[%u]%s;\n", reg_mask
, i
, reg_mask
);
3255 else if (usage_idx
== 1)
3256 shader_addline(&buffer
, "gl_FrontSecondaryColor%s = OUT[%u]%s;\n", reg_mask
, i
, reg_mask
);
3259 case WINED3DDECLUSAGE_POSITION
:
3260 shader_addline(&buffer
, "gl_Position%s = OUT[%u]%s;\n", reg_mask
, i
, reg_mask
);
3263 case WINED3DDECLUSAGE_TEXCOORD
:
3264 if (usage_idx
< 8) {
3265 if(!(GLINFO_LOCATION
).set_texcoord_w
|| ps_major
> 0) writemask
|= WINED3DSP_WRITEMASK_3
;
3267 shader_addline(&buffer
, "gl_TexCoord[%u]%s = OUT[%u]%s;\n",
3268 usage_idx
, reg_mask
, i
, reg_mask
);
3269 if(!(writemask
& WINED3DSP_WRITEMASK_3
)) {
3270 shader_addline(&buffer
, "gl_TexCoord[%u].w = 1.0;\n", usage_idx
);
3275 case WINED3DDECLUSAGE_PSIZE
:
3276 shader_addline(&buffer
, "gl_PointSize = OUT[%u].x;\n", i
);
3279 case WINED3DDECLUSAGE_FOG
:
3280 shader_addline(&buffer
, "gl_FogFragCoord = OUT[%u].%c;\n", i
, reg_mask
[1]);
3287 shader_addline(&buffer
, "}\n");
3289 } else if(ps_major
>= 3 && vs_major
>= 3) {
3290 semantics_out
= vs
->semantics_out
;
3291 semantics_in
= ps
->semantics_in
;
3293 /* This one is tricky: a 3.0 pixel shader reads from a 3.0 vertex shader */
3294 shader_addline(&buffer
, "varying vec4 IN[%u];\n", GL_LIMITS(glsl_varyings
) / 4);
3295 shader_addline(&buffer
, "void order_ps_input(in vec4 OUT[%u]) {\n", MAX_REG_OUTPUT
);
3297 /* First, sort out position and point size. Those are not passed to the pixel shader */
3298 for(i
= 0; i
< MAX_REG_OUTPUT
; i
++) {
3299 usage_token
= semantics_out
[i
].usage
;
3300 if (!usage_token
) continue;
3301 register_token
= semantics_out
[i
].reg
;
3303 usage
= (usage_token
& WINED3DSP_DCL_USAGE_MASK
) >> WINED3DSP_DCL_USAGE_SHIFT
;
3304 usage_idx
= (usage_token
& WINED3DSP_DCL_USAGEINDEX_MASK
) >> WINED3DSP_DCL_USAGEINDEX_SHIFT
;
3305 shader_glsl_get_write_mask(register_token
, reg_mask
);
3308 case WINED3DDECLUSAGE_POSITION
:
3309 shader_addline(&buffer
, "gl_Position%s = OUT[%u]%s;\n", reg_mask
, i
, reg_mask
);
3312 case WINED3DDECLUSAGE_PSIZE
:
3313 shader_addline(&buffer
, "gl_PointSize = OUT[%u].x;\n", i
);
3321 /* Then, fix the pixel shader input */
3322 handle_ps3_input(&buffer
, semantics_in
, semantics_out
, gl_info
, ps
->input_reg_map
);
3324 shader_addline(&buffer
, "}\n");
3325 } else if(ps_major
>= 3 && vs_major
< 3) {
3326 semantics_in
= ps
->semantics_in
;
3328 shader_addline(&buffer
, "varying vec4 IN[%u];\n", GL_LIMITS(glsl_varyings
) / 4);
3329 shader_addline(&buffer
, "void order_ps_input() {\n");
3330 /* The vertex shader wrote to the builtin varyings. There is no need to figure out position and
3331 * point size, but we depend on the optimizers kindness to find out that the pixel shader doesn't
3332 * read gl_TexCoord and gl_ColorX, otherwise we'll run out of varyings
3334 handle_ps3_input(&buffer
, semantics_in
, NULL
, gl_info
, ps
->input_reg_map
);
3335 shader_addline(&buffer
, "}\n");
3337 ERR("Unexpected vertex and pixel shader version condition: vs: %d, ps: %d\n", vs_major
, ps_major
);
3340 ret
= GL_EXTCALL(glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB
));
3341 checkGLcall("glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB)");
3342 GL_EXTCALL(glShaderSourceARB(ret
, 1, (const char**)&buffer
.buffer
, NULL
));
3343 checkGLcall("glShaderSourceARB(ret, 1, &buffer.buffer, NULL)");
3344 GL_EXTCALL(glCompileShaderARB(ret
));
3345 checkGLcall("glCompileShaderARB(ret)");
3347 shader_buffer_free(&buffer
);
3351 static void hardcode_local_constants(IWineD3DBaseShaderImpl
*shader
, const WineD3D_GL_Info
*gl_info
,
3352 GLhandleARB programId
, char prefix
)
3354 const local_constant
*lconst
;
3359 LIST_FOR_EACH_ENTRY(lconst
, &shader
->baseShader
.constantsF
, local_constant
, entry
) {
3360 value
= (const float *)lconst
->value
;
3361 snprintf(glsl_name
, sizeof(glsl_name
), "%cLC%u", prefix
, lconst
->idx
);
3362 tmp_loc
= GL_EXTCALL(glGetUniformLocationARB(programId
, glsl_name
));
3363 GL_EXTCALL(glUniform4fvARB(tmp_loc
, 1, value
));
3365 checkGLcall("Hardcoding local constants\n");
3368 /** Sets the GLSL program ID for the given pixel and vertex shader combination.
3369 * It sets the programId on the current StateBlock (because it should be called
3370 * inside of the DrawPrimitive() part of the render loop).
3372 * If a program for the given combination does not exist, create one, and store
3373 * the program in the hash table. If it creates a program, it will link the
3374 * given objects, too.
3376 static void set_glsl_shader_program(IWineD3DDevice
*iface
, BOOL use_ps
, BOOL use_vs
) {
3377 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
3378 struct shader_glsl_priv
*priv
= This
->shader_priv
;
3379 const WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
3380 IWineD3DPixelShader
*pshader
= This
->stateBlock
->pixelShader
;
3381 IWineD3DVertexShader
*vshader
= This
->stateBlock
->vertexShader
;
3382 struct glsl_shader_prog_link
*entry
= NULL
;
3383 GLhandleARB programId
= 0;
3384 GLhandleARB reorder_shader_id
= 0;
3387 GLhandleARB vshader_id
, pshader_id
;
3388 struct ps_compile_args ps_compile_args
;
3389 struct vs_compile_args vs_compile_args
;
3392 find_vs_compile_args((IWineD3DVertexShaderImpl
*)This
->stateBlock
->vertexShader
, This
->stateBlock
, &vs_compile_args
);
3394 /* FIXME: Do we really have to spend CPU cycles to generate a few zeroed bytes? */
3395 memset(&vs_compile_args
, 0, sizeof(vs_compile_args
));
3398 find_ps_compile_args((IWineD3DPixelShaderImpl
*)This
->stateBlock
->pixelShader
, This
->stateBlock
, &ps_compile_args
);
3400 /* FIXME: Do we really have to spend CPU cycles to generate a few zeroed bytes? */
3401 memset(&ps_compile_args
, 0, sizeof(ps_compile_args
));
3403 entry
= get_glsl_program_entry(priv
, vshader
, pshader
, &vs_compile_args
, &ps_compile_args
);
3405 priv
->glsl_program
= entry
;
3409 /* If we get to this point, then no matching program exists, so we create one */
3410 programId
= GL_EXTCALL(glCreateProgramObjectARB());
3411 TRACE("Created new GLSL shader program %u\n", programId
);
3413 /* Create the entry */
3414 entry
= HeapAlloc(GetProcessHeap(), 0, sizeof(struct glsl_shader_prog_link
));
3415 entry
->programId
= programId
;
3416 entry
->vshader
= vshader
;
3417 entry
->pshader
= pshader
;
3418 entry
->vs_args
= vs_compile_args
;
3419 entry
->ps_args
= ps_compile_args
;
3420 entry
->constant_version
= 0;
3421 /* Add the hash table entry */
3422 add_glsl_program_entry(priv
, entry
);
3424 /* Set the current program */
3425 priv
->glsl_program
= entry
;
3428 vshader_id
= find_gl_vshader((IWineD3DVertexShaderImpl
*) vshader
, &vs_compile_args
);
3433 /* Attach GLSL vshader */
3435 const unsigned int max_attribs
= 16; /* TODO: Will this always be the case? It is at the moment... */
3438 reorder_shader_id
= generate_param_reorder_function(vshader
, pshader
, gl_info
);
3439 TRACE("Attaching GLSL shader object %u to program %u\n", reorder_shader_id
, programId
);
3440 GL_EXTCALL(glAttachObjectARB(programId
, reorder_shader_id
));
3441 checkGLcall("glAttachObjectARB");
3442 /* Flag the reorder function for deletion, then it will be freed automatically when the program
3445 GL_EXTCALL(glDeleteObjectARB(reorder_shader_id
));
3447 TRACE("Attaching GLSL shader object %u to program %u\n", vshader_id
, programId
);
3448 GL_EXTCALL(glAttachObjectARB(programId
, vshader_id
));
3449 checkGLcall("glAttachObjectARB");
3451 /* Bind vertex attributes to a corresponding index number to match
3452 * the same index numbers as ARB_vertex_programs (makes loading
3453 * vertex attributes simpler). With this method, we can use the
3454 * exact same code to load the attributes later for both ARB and
3457 * We have to do this here because we need to know the Program ID
3458 * in order to make the bindings work, and it has to be done prior
3459 * to linking the GLSL program. */
3460 for (i
= 0; i
< max_attribs
; ++i
) {
3461 if (((IWineD3DBaseShaderImpl
*)vshader
)->baseShader
.reg_maps
.attributes
[i
]) {
3462 snprintf(tmp_name
, sizeof(tmp_name
), "attrib%i", i
);
3463 GL_EXTCALL(glBindAttribLocationARB(programId
, i
, tmp_name
));
3466 checkGLcall("glBindAttribLocationARB");
3468 list_add_head(&((IWineD3DBaseShaderImpl
*)vshader
)->baseShader
.linked_programs
, &entry
->vshader_entry
);
3472 pshader_id
= find_gl_pshader((IWineD3DPixelShaderImpl
*) pshader
, &ps_compile_args
);
3477 /* Attach GLSL pshader */
3479 TRACE("Attaching GLSL shader object %u to program %u\n", pshader_id
, programId
);
3480 GL_EXTCALL(glAttachObjectARB(programId
, pshader_id
));
3481 checkGLcall("glAttachObjectARB");
3483 list_add_head(&((IWineD3DBaseShaderImpl
*)pshader
)->baseShader
.linked_programs
, &entry
->pshader_entry
);
3486 /* Link the program */
3487 TRACE("Linking GLSL shader program %u\n", programId
);
3488 GL_EXTCALL(glLinkProgramARB(programId
));
3489 print_glsl_info_log(&GLINFO_LOCATION
, programId
);
3491 entry
->vuniformF_locations
= HeapAlloc(GetProcessHeap(), 0, sizeof(GLhandleARB
) * GL_LIMITS(vshader_constantsF
));
3492 for (i
= 0; i
< GL_LIMITS(vshader_constantsF
); ++i
) {
3493 snprintf(glsl_name
, sizeof(glsl_name
), "VC[%i]", i
);
3494 entry
->vuniformF_locations
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, glsl_name
));
3496 for (i
= 0; i
< MAX_CONST_I
; ++i
) {
3497 snprintf(glsl_name
, sizeof(glsl_name
), "VI[%i]", i
);
3498 entry
->vuniformI_locations
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, glsl_name
));
3500 entry
->puniformF_locations
= HeapAlloc(GetProcessHeap(), 0, sizeof(GLhandleARB
) * GL_LIMITS(pshader_constantsF
));
3501 for (i
= 0; i
< GL_LIMITS(pshader_constantsF
); ++i
) {
3502 snprintf(glsl_name
, sizeof(glsl_name
), "PC[%i]", i
);
3503 entry
->puniformF_locations
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, glsl_name
));
3505 for (i
= 0; i
< MAX_CONST_I
; ++i
) {
3506 snprintf(glsl_name
, sizeof(glsl_name
), "PI[%i]", i
);
3507 entry
->puniformI_locations
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, glsl_name
));
3511 for(i
= 0; i
< ((IWineD3DPixelShaderImpl
*)pshader
)->numbumpenvmatconsts
; i
++) {
3513 sprintf(name
, "bumpenvmat%d", ((IWineD3DPixelShaderImpl
*)pshader
)->bumpenvmatconst
[i
].texunit
);
3514 entry
->bumpenvmat_location
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, name
));
3515 sprintf(name
, "luminancescale%d", ((IWineD3DPixelShaderImpl
*)pshader
)->luminanceconst
[i
].texunit
);
3516 entry
->luminancescale_location
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, name
));
3517 sprintf(name
, "luminanceoffset%d", ((IWineD3DPixelShaderImpl
*)pshader
)->luminanceconst
[i
].texunit
);
3518 entry
->luminanceoffset_location
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, name
));
3522 if (use_ps
&& ps_compile_args
.texrect_fixup
) {
3524 for (i
= 0; i
< MAX_FRAGMENT_SAMPLERS
; ++i
) {
3525 if (ps_compile_args
.texrect_fixup
& (1 << i
)) {
3526 sprintf(name
, "PsamplerRectFixup%u", i
);
3527 entry
->rectFixup_location
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, name
));
3529 entry
->rectFixup_location
[i
] = -1;
3534 entry
->posFixup_location
= GL_EXTCALL(glGetUniformLocationARB(programId
, "posFixup"));
3535 entry
->ycorrection_location
= GL_EXTCALL(glGetUniformLocationARB(programId
, "ycorrection"));
3536 checkGLcall("Find glsl program uniform locations");
3539 && WINED3DSHADER_VERSION_MAJOR(((IWineD3DPixelShaderImpl
*)pshader
)->baseShader
.reg_maps
.shader_version
) >= 3
3540 && ((IWineD3DPixelShaderImpl
*)pshader
)->declared_in_count
> GL_LIMITS(glsl_varyings
) / 4)
3542 TRACE("Shader %d needs vertex color clamping disabled\n", programId
);
3543 entry
->vertex_color_clamp
= GL_FALSE
;
3545 entry
->vertex_color_clamp
= GL_FIXED_ONLY_ARB
;
3548 /* Set the shader to allow uniform loading on it */
3549 GL_EXTCALL(glUseProgramObjectARB(programId
));
3550 checkGLcall("glUseProgramObjectARB(programId)");
3552 /* Load the vertex and pixel samplers now. The function that finds the mappings makes sure
3553 * that it stays the same for each vertexshader-pixelshader pair(=linked glsl program). If
3554 * a pshader with fixed function pipeline is used there are no vertex samplers, and if a
3555 * vertex shader with fixed function pixel processing is used we make sure that the card
3556 * supports enough samplers to allow the max number of vertex samplers with all possible
3557 * fixed function fragment processing setups. So once the program is linked these samplers
3561 /* Load vertex shader samplers */
3562 shader_glsl_load_vsamplers(gl_info
, This
->texUnitMap
, programId
);
3565 /* Load pixel shader samplers */
3566 shader_glsl_load_psamplers(gl_info
, This
->texUnitMap
, programId
);
3569 /* If the local constants do not have to be loaded with the environment constants,
3570 * load them now to have them hardcoded in the GLSL program. This saves some CPU cycles
3573 if(pshader
&& !((IWineD3DPixelShaderImpl
*)pshader
)->baseShader
.load_local_constsF
) {
3574 hardcode_local_constants((IWineD3DBaseShaderImpl
*) pshader
, gl_info
, programId
, 'P');
3576 if(vshader
&& !((IWineD3DVertexShaderImpl
*)vshader
)->baseShader
.load_local_constsF
) {
3577 hardcode_local_constants((IWineD3DBaseShaderImpl
*) vshader
, gl_info
, programId
, 'V');
3581 static GLhandleARB
create_glsl_blt_shader(const WineD3D_GL_Info
*gl_info
, enum tex_types tex_type
)
3583 GLhandleARB program_id
;
3584 GLhandleARB vshader_id
, pshader_id
;
3585 static const char *blt_vshader
[] =
3590 " gl_Position = gl_Vertex;\n"
3591 " gl_FrontColor = vec4(1.0);\n"
3592 " gl_TexCoord[0] = gl_MultiTexCoord0;\n"
3596 static const char *blt_pshaders
[tex_type_count
] =
3602 "uniform sampler2D sampler;\n"
3605 " gl_FragDepth = texture2D(sampler, gl_TexCoord[0].xy).x;\n"
3611 "uniform samplerCube sampler;\n"
3614 " gl_FragDepth = textureCube(sampler, gl_TexCoord[0].xyz).x;\n"
3618 "#extension GL_ARB_texture_rectangle : enable\n"
3619 "uniform sampler2DRect sampler;\n"
3622 " gl_FragDepth = texture2DRect(sampler, gl_TexCoord[0].xy).x;\n"
3626 if (!blt_pshaders
[tex_type
])
3628 FIXME("tex_type %#x not supported\n", tex_type
);
3632 vshader_id
= GL_EXTCALL(glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB
));
3633 GL_EXTCALL(glShaderSourceARB(vshader_id
, 1, blt_vshader
, NULL
));
3634 GL_EXTCALL(glCompileShaderARB(vshader_id
));
3636 pshader_id
= GL_EXTCALL(glCreateShaderObjectARB(GL_FRAGMENT_SHADER_ARB
));
3637 GL_EXTCALL(glShaderSourceARB(pshader_id
, 1, &blt_pshaders
[tex_type
], NULL
));
3638 GL_EXTCALL(glCompileShaderARB(pshader_id
));
3640 program_id
= GL_EXTCALL(glCreateProgramObjectARB());
3641 GL_EXTCALL(glAttachObjectARB(program_id
, vshader_id
));
3642 GL_EXTCALL(glAttachObjectARB(program_id
, pshader_id
));
3643 GL_EXTCALL(glLinkProgramARB(program_id
));
3645 print_glsl_info_log(&GLINFO_LOCATION
, program_id
);
3647 /* Once linked we can mark the shaders for deletion. They will be deleted once the program
3650 GL_EXTCALL(glDeleteObjectARB(vshader_id
));
3651 GL_EXTCALL(glDeleteObjectARB(pshader_id
));
3655 static void shader_glsl_select(IWineD3DDevice
*iface
, BOOL usePS
, BOOL useVS
) {
3656 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
3657 struct shader_glsl_priv
*priv
= This
->shader_priv
;
3658 const WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
3659 GLhandleARB program_id
= 0;
3660 GLenum old_vertex_color_clamp
, current_vertex_color_clamp
;
3662 old_vertex_color_clamp
= priv
->glsl_program
? priv
->glsl_program
->vertex_color_clamp
: GL_FIXED_ONLY_ARB
;
3664 if (useVS
|| usePS
) set_glsl_shader_program(iface
, usePS
, useVS
);
3665 else priv
->glsl_program
= NULL
;
3667 current_vertex_color_clamp
= priv
->glsl_program
? priv
->glsl_program
->vertex_color_clamp
: GL_FIXED_ONLY_ARB
;
3669 if (old_vertex_color_clamp
!= current_vertex_color_clamp
) {
3670 if (GL_SUPPORT(ARB_COLOR_BUFFER_FLOAT
)) {
3671 GL_EXTCALL(glClampColorARB(GL_CLAMP_VERTEX_COLOR_ARB
, current_vertex_color_clamp
));
3672 checkGLcall("glClampColorARB");
3674 FIXME("vertex color clamp needs to be changed, but extension not supported.\n");
3678 program_id
= priv
->glsl_program
? priv
->glsl_program
->programId
: 0;
3679 if (program_id
) TRACE("Using GLSL program %u\n", program_id
);
3680 GL_EXTCALL(glUseProgramObjectARB(program_id
));
3681 checkGLcall("glUseProgramObjectARB");
3684 static void shader_glsl_select_depth_blt(IWineD3DDevice
*iface
, enum tex_types tex_type
) {
3685 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
3686 const WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
3687 struct shader_glsl_priv
*priv
= This
->shader_priv
;
3688 GLhandleARB
*blt_program
= &priv
->depth_blt_program
[tex_type
];
3690 if (!*blt_program
) {
3692 *blt_program
= create_glsl_blt_shader(gl_info
, tex_type
);
3693 loc
= GL_EXTCALL(glGetUniformLocationARB(*blt_program
, "sampler"));
3694 GL_EXTCALL(glUseProgramObjectARB(*blt_program
));
3695 GL_EXTCALL(glUniform1iARB(loc
, 0));
3697 GL_EXTCALL(glUseProgramObjectARB(*blt_program
));
3701 static void shader_glsl_deselect_depth_blt(IWineD3DDevice
*iface
) {
3702 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
3703 const WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
3704 struct shader_glsl_priv
*priv
= This
->shader_priv
;
3705 GLhandleARB program_id
;
3707 program_id
= priv
->glsl_program
? priv
->glsl_program
->programId
: 0;
3708 if (program_id
) TRACE("Using GLSL program %u\n", program_id
);
3710 GL_EXTCALL(glUseProgramObjectARB(program_id
));
3711 checkGLcall("glUseProgramObjectARB");
3714 static void shader_glsl_destroy(IWineD3DBaseShader
*iface
) {
3715 const struct list
*linked_programs
;
3716 IWineD3DBaseShaderImpl
*This
= (IWineD3DBaseShaderImpl
*) iface
;
3717 IWineD3DDeviceImpl
*device
= (IWineD3DDeviceImpl
*)This
->baseShader
.device
;
3718 struct shader_glsl_priv
*priv
= device
->shader_priv
;
3719 const WineD3D_GL_Info
*gl_info
= &device
->adapter
->gl_info
;
3720 IWineD3DPixelShaderImpl
*ps
= NULL
;
3721 IWineD3DVertexShaderImpl
*vs
= NULL
;
3723 /* Note: Do not use QueryInterface here to find out which shader type this is because this code
3724 * can be called from IWineD3DBaseShader::Release
3726 char pshader
= shader_is_pshader_version(This
->baseShader
.reg_maps
.shader_version
);
3729 ps
= (IWineD3DPixelShaderImpl
*) This
;
3730 if(ps
->num_gl_shaders
== 0) return;
3732 vs
= (IWineD3DVertexShaderImpl
*) This
;
3733 if(vs
->num_gl_shaders
== 0) return;
3736 linked_programs
= &This
->baseShader
.linked_programs
;
3738 TRACE("Deleting linked programs\n");
3739 if (linked_programs
->next
) {
3740 struct glsl_shader_prog_link
*entry
, *entry2
;
3743 LIST_FOR_EACH_ENTRY_SAFE(entry
, entry2
, linked_programs
, struct glsl_shader_prog_link
, pshader_entry
) {
3744 delete_glsl_program_entry(priv
, gl_info
, entry
);
3747 LIST_FOR_EACH_ENTRY_SAFE(entry
, entry2
, linked_programs
, struct glsl_shader_prog_link
, vshader_entry
) {
3748 delete_glsl_program_entry(priv
, gl_info
, entry
);
3757 for(i
= 0; i
< ps
->num_gl_shaders
; i
++) {
3758 TRACE("deleting pshader %u\n", ps
->gl_shaders
[i
].prgId
);
3759 GL_EXTCALL(glDeleteObjectARB(ps
->gl_shaders
[i
].prgId
));
3760 checkGLcall("glDeleteObjectARB");
3763 HeapFree(GetProcessHeap(), 0, ps
->gl_shaders
);
3764 ps
->gl_shaders
= NULL
;
3765 ps
->num_gl_shaders
= 0;
3766 ps
->shader_array_size
= 0;
3771 for(i
= 0; i
< vs
->num_gl_shaders
; i
++) {
3772 TRACE("deleting vshader %u\n", vs
->gl_shaders
[i
].prgId
);
3773 GL_EXTCALL(glDeleteObjectARB(vs
->gl_shaders
[i
].prgId
));
3774 checkGLcall("glDeleteObjectARB");
3777 HeapFree(GetProcessHeap(), 0, vs
->gl_shaders
);
3778 vs
->gl_shaders
= NULL
;
3779 vs
->num_gl_shaders
= 0;
3780 vs
->shader_array_size
= 0;
3784 static unsigned int glsl_program_key_hash(const void *key
)
3786 const glsl_program_key_t
*k
= key
;
3788 unsigned int hash
= ((DWORD_PTR
) k
->vshader
) | ((DWORD_PTR
) k
->pshader
) << 16;
3789 hash
+= ~(hash
<< 15);
3790 hash
^= (hash
>> 10);
3791 hash
+= (hash
<< 3);
3792 hash
^= (hash
>> 6);
3793 hash
+= ~(hash
<< 11);
3794 hash
^= (hash
>> 16);
3799 static BOOL
glsl_program_key_compare(const void *keya
, const void *keyb
)
3801 const glsl_program_key_t
*ka
= keya
;
3802 const glsl_program_key_t
*kb
= keyb
;
3804 return ka
->vshader
== kb
->vshader
&& ka
->pshader
== kb
->pshader
&&
3805 (memcmp(&ka
->ps_args
, &kb
->ps_args
, sizeof(kb
->ps_args
)) == 0) &&
3806 (memcmp(&ka
->vs_args
, &kb
->vs_args
, sizeof(kb
->vs_args
)) == 0);
3809 static BOOL
constant_heap_init(struct constant_heap
*heap
, unsigned int constant_count
)
3811 SIZE_T size
= (constant_count
+ 1) * sizeof(*heap
->entries
) + constant_count
* sizeof(*heap
->positions
);
3812 void *mem
= HeapAlloc(GetProcessHeap(), 0, size
);
3816 ERR("Failed to allocate memory\n");
3820 heap
->entries
= mem
;
3821 heap
->entries
[1].version
= 0;
3822 heap
->positions
= (unsigned int *)(heap
->entries
+ constant_count
+ 1);
3828 static void constant_heap_free(struct constant_heap
*heap
)
3830 HeapFree(GetProcessHeap(), 0, heap
->entries
);
3833 static HRESULT
shader_glsl_alloc(IWineD3DDevice
*iface
) {
3834 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
3835 const WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
3836 struct shader_glsl_priv
*priv
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(struct shader_glsl_priv
));
3837 SIZE_T stack_size
= wined3d_log2i(max(GL_LIMITS(vshader_constantsF
), GL_LIMITS(pshader_constantsF
))) + 1;
3839 priv
->stack
= HeapAlloc(GetProcessHeap(), 0, stack_size
* sizeof(*priv
->stack
));
3842 ERR("Failed to allocate memory.\n");
3843 HeapFree(GetProcessHeap(), 0, priv
);
3844 return E_OUTOFMEMORY
;
3847 if (!constant_heap_init(&priv
->vconst_heap
, GL_LIMITS(vshader_constantsF
)))
3849 ERR("Failed to initialize vertex shader constant heap\n");
3850 HeapFree(GetProcessHeap(), 0, priv
->stack
);
3851 HeapFree(GetProcessHeap(), 0, priv
);
3852 return E_OUTOFMEMORY
;
3855 if (!constant_heap_init(&priv
->pconst_heap
, GL_LIMITS(pshader_constantsF
)))
3857 ERR("Failed to initialize pixel shader constant heap\n");
3858 constant_heap_free(&priv
->vconst_heap
);
3859 HeapFree(GetProcessHeap(), 0, priv
->stack
);
3860 HeapFree(GetProcessHeap(), 0, priv
);
3861 return E_OUTOFMEMORY
;
3864 priv
->glsl_program_lookup
= hash_table_create(glsl_program_key_hash
, glsl_program_key_compare
);
3865 priv
->next_constant_version
= 1;
3867 This
->shader_priv
= priv
;
3871 static void shader_glsl_free(IWineD3DDevice
*iface
) {
3872 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
3873 const WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
3874 struct shader_glsl_priv
*priv
= This
->shader_priv
;
3877 for (i
= 0; i
< tex_type_count
; ++i
)
3879 if (priv
->depth_blt_program
[i
])
3881 GL_EXTCALL(glDeleteObjectARB(priv
->depth_blt_program
[i
]));
3885 hash_table_destroy(priv
->glsl_program_lookup
, NULL
, NULL
);
3886 constant_heap_free(&priv
->pconst_heap
);
3887 constant_heap_free(&priv
->vconst_heap
);
3889 HeapFree(GetProcessHeap(), 0, This
->shader_priv
);
3890 This
->shader_priv
= NULL
;
3893 static BOOL
shader_glsl_dirty_const(IWineD3DDevice
*iface
) {
3894 /* TODO: GL_EXT_bindable_uniform can be used to share constants across shaders */
3898 static GLuint
shader_glsl_generate_pshader(IWineD3DPixelShader
*iface
, SHADER_BUFFER
*buffer
, const struct ps_compile_args
*args
) {
3899 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*)iface
;
3900 const struct shader_reg_maps
*reg_maps
= &This
->baseShader
.reg_maps
;
3901 CONST DWORD
*function
= This
->baseShader
.function
;
3902 const char *fragcolor
;
3903 const WineD3D_GL_Info
*gl_info
= &((IWineD3DDeviceImpl
*)This
->baseShader
.device
)->adapter
->gl_info
;
3905 /* Create the hw GLSL shader object and assign it as the shader->prgId */
3906 GLhandleARB shader_obj
= GL_EXTCALL(glCreateShaderObjectARB(GL_FRAGMENT_SHADER_ARB
));
3908 shader_addline(buffer
, "#version 120\n");
3910 if (GL_SUPPORT(ARB_DRAW_BUFFERS
)) {
3911 shader_addline(buffer
, "#extension GL_ARB_draw_buffers : enable\n");
3913 if (GL_SUPPORT(ARB_TEXTURE_RECTANGLE
)) {
3914 /* The spec says that it doesn't have to be explicitly enabled, but the nvidia
3915 * drivers write a warning if we don't do so
3917 shader_addline(buffer
, "#extension GL_ARB_texture_rectangle : enable\n");
3920 /* Base Declarations */
3921 shader_generate_glsl_declarations( (IWineD3DBaseShader
*) This
, reg_maps
, buffer
, &GLINFO_LOCATION
, args
);
3923 /* Pack 3.0 inputs */
3924 if (reg_maps
->shader_version
>= WINED3DPS_VERSION(3,0) && args
->vp_mode
!= vertexshader
) {
3925 pshader_glsl_input_pack(buffer
, This
->semantics_in
, iface
, args
->vp_mode
);
3928 /* Base Shader Body */
3929 shader_generate_main( (IWineD3DBaseShader
*) This
, buffer
, reg_maps
, function
);
3931 /* Pixel shaders < 2.0 place the resulting color in R0 implicitly */
3932 if (reg_maps
->shader_version
< WINED3DPS_VERSION(2,0))
3934 /* Some older cards like GeforceFX ones don't support multiple buffers, so also not gl_FragData */
3935 if(GL_SUPPORT(ARB_DRAW_BUFFERS
))
3936 shader_addline(buffer
, "gl_FragData[0] = R0;\n");
3938 shader_addline(buffer
, "gl_FragColor = R0;\n");
3941 if(GL_SUPPORT(ARB_DRAW_BUFFERS
)) {
3942 fragcolor
= "gl_FragData[0]";
3944 fragcolor
= "gl_FragColor";
3946 if(args
->srgb_correction
) {
3947 shader_addline(buffer
, "tmp0.xyz = pow(%s.xyz, vec3(%f, %f, %f)) * vec3(%f, %f, %f) - vec3(%f, %f, %f);\n",
3948 fragcolor
, srgb_pow
, srgb_pow
, srgb_pow
, srgb_mul_high
, srgb_mul_high
, srgb_mul_high
,
3949 srgb_sub_high
, srgb_sub_high
, srgb_sub_high
);
3950 shader_addline(buffer
, "tmp1.xyz = %s.xyz * srgb_mul_low.xyz;\n", fragcolor
);
3951 shader_addline(buffer
, "%s.x = %s.x < srgb_comparison.x ? tmp1.x : tmp0.x;\n", fragcolor
, fragcolor
);
3952 shader_addline(buffer
, "%s.y = %s.y < srgb_comparison.y ? tmp1.y : tmp0.y;\n", fragcolor
, fragcolor
);
3953 shader_addline(buffer
, "%s.z = %s.z < srgb_comparison.z ? tmp1.z : tmp0.z;\n", fragcolor
, fragcolor
);
3954 shader_addline(buffer
, "%s = clamp(%s, 0.0, 1.0);\n", fragcolor
, fragcolor
);
3956 /* Pixel shader < 3.0 do not replace the fog stage.
3957 * This implements linear fog computation and blending.
3958 * TODO: non linear fog
3959 * NOTE: gl_Fog.start and gl_Fog.end don't hold fog start s and end e but
3960 * -1/(e-s) and e/(e-s) respectively.
3962 if(reg_maps
->shader_version
< WINED3DPS_VERSION(3,0)) {
3964 case FOG_OFF
: break;
3966 shader_addline(buffer
, "float fogstart = -1.0 / (gl_Fog.end - gl_Fog.start);\n");
3967 shader_addline(buffer
, "float fogend = gl_Fog.end * -fogstart;\n");
3968 shader_addline(buffer
, "float Fog = clamp(gl_FogFragCoord * fogstart + fogend, 0.0, 1.0);\n");
3969 shader_addline(buffer
, "%s.xyz = mix(gl_Fog.color.xyz, %s.xyz, Fog);\n", fragcolor
, fragcolor
);
3972 /* Fog = e^(-gl_Fog.density * gl_FogFragCoord) */
3973 shader_addline(buffer
, "float Fog = exp(-gl_Fog.density * gl_FogFragCoord);\n");
3974 shader_addline(buffer
, "Fog = clamp(Fog, 0.0, 1.0);\n");
3975 shader_addline(buffer
, "%s.xyz = mix(gl_Fog.color.xyz, %s.xyz, Fog);\n", fragcolor
, fragcolor
);
3978 /* Fog = e^(-(gl_Fog.density * gl_FogFragCoord)^2) */
3979 shader_addline(buffer
, "float Fog = exp(-gl_Fog.density * gl_Fog.density * gl_FogFragCoord * gl_FogFragCoord);\n");
3980 shader_addline(buffer
, "Fog = clamp(Fog, 0.0, 1.0);\n");
3981 shader_addline(buffer
, "%s.xyz = mix(gl_Fog.color.xyz, %s.xyz, Fog);\n", fragcolor
, fragcolor
);
3986 shader_addline(buffer
, "}\n");
3988 TRACE("Compiling shader object %u\n", shader_obj
);
3989 GL_EXTCALL(glShaderSourceARB(shader_obj
, 1, (const char**)&buffer
->buffer
, NULL
));
3990 GL_EXTCALL(glCompileShaderARB(shader_obj
));
3991 print_glsl_info_log(&GLINFO_LOCATION
, shader_obj
);
3993 /* Store the shader object */
3997 static GLuint
shader_glsl_generate_vshader(IWineD3DVertexShader
*iface
, SHADER_BUFFER
*buffer
, const struct vs_compile_args
*args
) {
3998 IWineD3DVertexShaderImpl
*This
= (IWineD3DVertexShaderImpl
*)iface
;
3999 const struct shader_reg_maps
*reg_maps
= &This
->baseShader
.reg_maps
;
4000 CONST DWORD
*function
= This
->baseShader
.function
;
4001 const WineD3D_GL_Info
*gl_info
= &((IWineD3DDeviceImpl
*)This
->baseShader
.device
)->adapter
->gl_info
;
4003 /* Create the hw GLSL shader program and assign it as the shader->prgId */
4004 GLhandleARB shader_obj
= GL_EXTCALL(glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB
));
4006 shader_addline(buffer
, "#version 120\n");
4008 /* Base Declarations */
4009 shader_generate_glsl_declarations( (IWineD3DBaseShader
*) This
, reg_maps
, buffer
, &GLINFO_LOCATION
, NULL
);
4011 /* Base Shader Body */
4012 shader_generate_main( (IWineD3DBaseShader
*) This
, buffer
, reg_maps
, function
);
4014 /* Unpack 3.0 outputs */
4015 if (reg_maps
->shader_version
>= WINED3DVS_VERSION(3,0)) shader_addline(buffer
, "order_ps_input(OUT);\n");
4016 else shader_addline(buffer
, "order_ps_input();\n");
4018 /* The D3DRS_FOGTABLEMODE render state defines if the shader-generated fog coord is used
4019 * or if the fragment depth is used. If the fragment depth is used(FOGTABLEMODE != NONE),
4020 * the fog frag coord is thrown away. If the fog frag coord is used, but not written by
4021 * the shader, it is set to 0.0(fully fogged, since start = 1.0, end = 0.0)
4023 if(args
->fog_src
== VS_FOG_Z
) {
4024 shader_addline(buffer
, "gl_FogFragCoord = gl_Position.z;\n");
4025 } else if (!reg_maps
->fog
) {
4026 shader_addline(buffer
, "gl_FogFragCoord = 0.0;\n");
4029 /* Write the final position.
4031 * OpenGL coordinates specify the center of the pixel while d3d coords specify
4032 * the corner. The offsets are stored in z and w in posFixup. posFixup.y contains
4033 * 1.0 or -1.0 to turn the rendering upside down for offscreen rendering. PosFixup.x
4034 * contains 1.0 to allow a mad.
4036 shader_addline(buffer
, "gl_Position.y = gl_Position.y * posFixup.y;\n");
4037 shader_addline(buffer
, "gl_Position.xy += posFixup.zw * gl_Position.ww;\n");
4039 /* Z coord [0;1]->[-1;1] mapping, see comment in transform_projection in state.c
4041 * Basically we want (in homogeneous coordinates) z = z * 2 - 1. However, shaders are run
4042 * before the homogeneous divide, so we have to take the w into account: z = ((z / w) * 2 - 1) * w,
4043 * which is the same as z = z * 2 - w.
4045 shader_addline(buffer
, "gl_Position.z = gl_Position.z * 2.0 - gl_Position.w;\n");
4047 shader_addline(buffer
, "}\n");
4049 TRACE("Compiling shader object %u\n", shader_obj
);
4050 GL_EXTCALL(glShaderSourceARB(shader_obj
, 1, (const char**)&buffer
->buffer
, NULL
));
4051 GL_EXTCALL(glCompileShaderARB(shader_obj
));
4052 print_glsl_info_log(&GLINFO_LOCATION
, shader_obj
);
4057 static void shader_glsl_get_caps(WINED3DDEVTYPE devtype
, const WineD3D_GL_Info
*gl_info
, struct shader_caps
*pCaps
)
4059 /* Nvidia Geforce6/7 or Ati R4xx/R5xx cards with GLSL support, support VS 3.0 but older Nvidia/Ati
4060 * models with GLSL support only support 2.0. In case of nvidia we can detect VS 2.0 support using
4061 * vs_nv_version which is based on NV_vertex_program.
4062 * For Ati cards there's no way using glsl (it abstracts the lowlevel info away) and also not
4063 * using ARB_vertex_program. It is safe to assume that when a card supports pixel shader 2.0 it
4064 * supports vertex shader 2.0 too and the way around. We can detect ps2.0 using the maximum number
4065 * of native instructions, so use that here. For more info see the pixel shader versioning code below.
4067 if((GLINFO_LOCATION
.vs_nv_version
== VS_VERSION_20
) || (GLINFO_LOCATION
.ps_arb_max_instructions
<= 512))
4068 pCaps
->VertexShaderVersion
= WINED3DVS_VERSION(2,0);
4070 pCaps
->VertexShaderVersion
= WINED3DVS_VERSION(3,0);
4071 TRACE_(d3d_caps
)("Hardware vertex shader version %d.%d enabled (GLSL)\n", (pCaps
->VertexShaderVersion
>> 8) & 0xff, pCaps
->VertexShaderVersion
& 0xff);
4072 pCaps
->MaxVertexShaderConst
= GL_LIMITS(vshader_constantsF
);
4074 /* Older DX9-class videocards (GeforceFX / Radeon >9500/X*00) only support pixel shader 2.0/2.0a/2.0b.
4075 * In OpenGL the extensions related to GLSL abstract lowlevel GL info away which is needed
4076 * to distinguish between 2.0 and 3.0 (and 2.0a/2.0b). In case of Nvidia we use their fragment
4077 * program extensions. On other hardware including ATI GL_ARB_fragment_program offers the info
4078 * in max native instructions. Intel and others also offer the info in this extension but they
4079 * don't support GLSL (at least on Windows).
4081 * PS2.0 requires at least 96 instructions, 2.0a/2.0b go up to 512. Assume that if the number
4082 * of instructions is 512 or less we have to do with ps2.0 hardware.
4083 * 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.
4085 if((GLINFO_LOCATION
.ps_nv_version
== PS_VERSION_20
) || (GLINFO_LOCATION
.ps_arb_max_instructions
<= 512))
4086 pCaps
->PixelShaderVersion
= WINED3DPS_VERSION(2,0);
4088 pCaps
->PixelShaderVersion
= WINED3DPS_VERSION(3,0);
4090 /* FIXME: The following line is card dependent. -8.0 to 8.0 is the
4091 * Direct3D minimum requirement.
4093 * Both GL_ARB_fragment_program and GLSL require a "maximum representable magnitude"
4094 * of colors to be 2^10, and 2^32 for other floats. Should we use 1024 here?
4096 * The problem is that the refrast clamps temporary results in the shader to
4097 * [-MaxValue;+MaxValue]. If the card's max value is bigger than the one we advertize here,
4098 * then applications may miss the clamping behavior. On the other hand, if it is smaller,
4099 * the shader will generate incorrect results too. Unfortunately, GL deliberately doesn't
4100 * offer a way to query this.
4102 pCaps
->PixelShader1xMaxValue
= 8.0;
4103 TRACE_(d3d_caps
)("Hardware pixel shader version %d.%d enabled (GLSL)\n", (pCaps
->PixelShaderVersion
>> 8) & 0xff, pCaps
->PixelShaderVersion
& 0xff);
4106 static BOOL
shader_glsl_color_fixup_supported(struct color_fixup_desc fixup
)
4108 if (TRACE_ON(d3d_shader
) && TRACE_ON(d3d
))
4110 TRACE("Checking support for fixup:\n");
4111 dump_color_fixup_desc(fixup
);
4114 /* We support everything except YUV conversions. */
4115 if (!is_yuv_fixup(fixup
))
4121 TRACE("[FAILED]\n");
4125 static const SHADER_HANDLER shader_glsl_instruction_handler_table
[WINED3DSIH_TABLE_SIZE
] =
4127 /* WINED3DSIH_ABS */ shader_glsl_map2gl
,
4128 /* WINED3DSIH_ADD */ shader_glsl_arith
,
4129 /* WINED3DSIH_BEM */ pshader_glsl_bem
,
4130 /* WINED3DSIH_BREAK */ shader_glsl_break
,
4131 /* WINED3DSIH_BREAKC */ shader_glsl_breakc
,
4132 /* WINED3DSIH_BREAKP */ NULL
,
4133 /* WINED3DSIH_CALL */ shader_glsl_call
,
4134 /* WINED3DSIH_CALLNZ */ shader_glsl_callnz
,
4135 /* WINED3DSIH_CMP */ shader_glsl_cmp
,
4136 /* WINED3DSIH_CND */ shader_glsl_cnd
,
4137 /* WINED3DSIH_CRS */ shader_glsl_cross
,
4138 /* WINED3DSIH_DCL */ NULL
,
4139 /* WINED3DSIH_DEF */ NULL
,
4140 /* WINED3DSIH_DEFB */ NULL
,
4141 /* WINED3DSIH_DEFI */ NULL
,
4142 /* WINED3DSIH_DP2ADD */ pshader_glsl_dp2add
,
4143 /* WINED3DSIH_DP3 */ shader_glsl_dot
,
4144 /* WINED3DSIH_DP4 */ shader_glsl_dot
,
4145 /* WINED3DSIH_DST */ shader_glsl_dst
,
4146 /* WINED3DSIH_DSX */ shader_glsl_map2gl
,
4147 /* WINED3DSIH_DSY */ shader_glsl_map2gl
,
4148 /* WINED3DSIH_ELSE */ shader_glsl_else
,
4149 /* WINED3DSIH_ENDIF */ shader_glsl_end
,
4150 /* WINED3DSIH_ENDLOOP */ shader_glsl_end
,
4151 /* WINED3DSIH_ENDREP */ shader_glsl_end
,
4152 /* WINED3DSIH_EXP */ shader_glsl_map2gl
,
4153 /* WINED3DSIH_EXPP */ shader_glsl_expp
,
4154 /* WINED3DSIH_FRC */ shader_glsl_map2gl
,
4155 /* WINED3DSIH_IF */ shader_glsl_if
,
4156 /* WINED3DSIH_IFC */ shader_glsl_ifc
,
4157 /* WINED3DSIH_LABEL */ shader_glsl_label
,
4158 /* WINED3DSIH_LIT */ shader_glsl_lit
,
4159 /* WINED3DSIH_LOG */ shader_glsl_log
,
4160 /* WINED3DSIH_LOGP */ shader_glsl_log
,
4161 /* WINED3DSIH_LOOP */ shader_glsl_loop
,
4162 /* WINED3DSIH_LRP */ shader_glsl_lrp
,
4163 /* WINED3DSIH_M3x2 */ shader_glsl_mnxn
,
4164 /* WINED3DSIH_M3x3 */ shader_glsl_mnxn
,
4165 /* WINED3DSIH_M3x4 */ shader_glsl_mnxn
,
4166 /* WINED3DSIH_M4x3 */ shader_glsl_mnxn
,
4167 /* WINED3DSIH_M4x4 */ shader_glsl_mnxn
,
4168 /* WINED3DSIH_MAD */ shader_glsl_mad
,
4169 /* WINED3DSIH_MAX */ shader_glsl_map2gl
,
4170 /* WINED3DSIH_MIN */ shader_glsl_map2gl
,
4171 /* WINED3DSIH_MOV */ shader_glsl_mov
,
4172 /* WINED3DSIH_MOVA */ shader_glsl_mov
,
4173 /* WINED3DSIH_MUL */ shader_glsl_arith
,
4174 /* WINED3DSIH_NOP */ NULL
,
4175 /* WINED3DSIH_NRM */ shader_glsl_map2gl
,
4176 /* WINED3DSIH_PHASE */ NULL
,
4177 /* WINED3DSIH_POW */ shader_glsl_pow
,
4178 /* WINED3DSIH_RCP */ shader_glsl_rcp
,
4179 /* WINED3DSIH_REP */ shader_glsl_rep
,
4180 /* WINED3DSIH_RET */ NULL
,
4181 /* WINED3DSIH_RSQ */ shader_glsl_rsq
,
4182 /* WINED3DSIH_SETP */ NULL
,
4183 /* WINED3DSIH_SGE */ shader_glsl_compare
,
4184 /* WINED3DSIH_SGN */ shader_glsl_map2gl
,
4185 /* WINED3DSIH_SINCOS */ shader_glsl_sincos
,
4186 /* WINED3DSIH_SLT */ shader_glsl_compare
,
4187 /* WINED3DSIH_SUB */ shader_glsl_arith
,
4188 /* WINED3DSIH_TEX */ pshader_glsl_tex
,
4189 /* WINED3DSIH_TEXBEM */ pshader_glsl_texbem
,
4190 /* WINED3DSIH_TEXBEML */ pshader_glsl_texbem
,
4191 /* WINED3DSIH_TEXCOORD */ pshader_glsl_texcoord
,
4192 /* WINED3DSIH_TEXDEPTH */ pshader_glsl_texdepth
,
4193 /* WINED3DSIH_TEXDP3 */ pshader_glsl_texdp3
,
4194 /* WINED3DSIH_TEXDP3TEX */ pshader_glsl_texdp3tex
,
4195 /* WINED3DSIH_TEXKILL */ pshader_glsl_texkill
,
4196 /* WINED3DSIH_TEXLDD */ NULL
,
4197 /* WINED3DSIH_TEXLDL */ shader_glsl_texldl
,
4198 /* WINED3DSIH_TEXM3x2DEPTH */ pshader_glsl_texm3x2depth
,
4199 /* WINED3DSIH_TEXM3x2PAD */ pshader_glsl_texm3x2pad
,
4200 /* WINED3DSIH_TEXM3x2TEX */ pshader_glsl_texm3x2tex
,
4201 /* WINED3DSIH_TEXM3x3 */ pshader_glsl_texm3x3
,
4202 /* WINED3DSIH_TEXM3x3DIFF */ NULL
,
4203 /* WINED3DSIH_TEXM3x3PAD */ pshader_glsl_texm3x3pad
,
4204 /* WINED3DSIH_TEXM3x3SPEC */ pshader_glsl_texm3x3spec
,
4205 /* WINED3DSIH_TEXM3x3TEX */ pshader_glsl_texm3x3tex
,
4206 /* WINED3DSIH_TEXM3x3VSPEC */ pshader_glsl_texm3x3vspec
,
4207 /* WINED3DSIH_TEXREG2AR */ pshader_glsl_texreg2ar
,
4208 /* WINED3DSIH_TEXREG2GB */ pshader_glsl_texreg2gb
,
4209 /* WINED3DSIH_TEXREG2RGB */ pshader_glsl_texreg2rgb
,
4212 const shader_backend_t glsl_shader_backend
= {
4213 shader_glsl_instruction_handler_table
,
4215 shader_glsl_select_depth_blt
,
4216 shader_glsl_deselect_depth_blt
,
4217 shader_glsl_update_float_vertex_constants
,
4218 shader_glsl_update_float_pixel_constants
,
4219 shader_glsl_load_constants
,
4220 shader_glsl_destroy
,
4223 shader_glsl_dirty_const
,
4224 shader_glsl_generate_pshader
,
4225 shader_glsl_generate_vshader
,
4226 shader_glsl_get_caps
,
4227 shader_glsl_color_fixup_supported
,