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
7 * Copyright 2009 Henri Verbeet for CodeWeavers
9 * This library is free software; you can redistribute it and/or
10 * modify it under the terms of the GNU Lesser General Public
11 * License as published by the Free Software Foundation; either
12 * version 2.1 of the License, or (at your option) any later version.
14 * This library is distributed in the hope that it will be useful,
15 * but WITHOUT ANY WARRANTY; without even the implied warranty of
16 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
17 * Lesser General Public License for more details.
19 * You should have received a copy of the GNU Lesser General Public
20 * License along with this library; if not, write to the Free Software
21 * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301, USA
25 * D3D shader asm has swizzles on source parameters, and write masks for
26 * destination parameters. GLSL uses swizzles for both. The result of this is
27 * that for example "mov dst.xw, src.zyxw" becomes "dst.xw = src.zw" in GLSL.
28 * Ie, to generate a proper GLSL source swizzle, we need to take the D3D write
29 * mask for the destination parameter into account.
35 #include "wined3d_private.h"
37 WINE_DEFAULT_DEBUG_CHANNEL(d3d_shader
);
38 WINE_DECLARE_DEBUG_CHANNEL(d3d_constants
);
39 WINE_DECLARE_DEBUG_CHANNEL(d3d_caps
);
40 WINE_DECLARE_DEBUG_CHANNEL(d3d
);
42 #define GLINFO_LOCATION (*gl_info)
44 #define WINED3D_GLSL_SAMPLE_PROJECTED 0x1
45 #define WINED3D_GLSL_SAMPLE_RECT 0x2
46 #define WINED3D_GLSL_SAMPLE_LOD 0x4
61 } glsl_sample_function_t
;
65 HEAP_NODE_TRAVERSE_LEFT
,
66 HEAP_NODE_TRAVERSE_RIGHT
,
78 struct constant_entry
*entries
;
79 unsigned int *positions
;
83 /* GLSL shader private data */
84 struct shader_glsl_priv
{
85 struct hash_table_t
*glsl_program_lookup
;
86 struct glsl_shader_prog_link
*glsl_program
;
87 struct constant_heap vconst_heap
;
88 struct constant_heap pconst_heap
;
90 GLhandleARB depth_blt_program
[tex_type_count
];
91 UINT next_constant_version
;
94 /* Struct to maintain data about a linked GLSL program */
95 struct glsl_shader_prog_link
{
96 struct list vshader_entry
;
97 struct list pshader_entry
;
98 GLhandleARB programId
;
99 GLint
*vuniformF_locations
;
100 GLint
*puniformF_locations
;
101 GLint vuniformI_locations
[MAX_CONST_I
];
102 GLint puniformI_locations
[MAX_CONST_I
];
103 GLint posFixup_location
;
104 GLint np2Fixup_location
[MAX_FRAGMENT_SAMPLERS
];
105 GLint bumpenvmat_location
[MAX_TEXTURES
];
106 GLint luminancescale_location
[MAX_TEXTURES
];
107 GLint luminanceoffset_location
[MAX_TEXTURES
];
108 GLint ycorrection_location
;
109 GLenum vertex_color_clamp
;
110 IWineD3DVertexShader
*vshader
;
111 IWineD3DPixelShader
*pshader
;
112 struct vs_compile_args vs_args
;
113 struct ps_compile_args ps_args
;
114 UINT constant_version
;
118 IWineD3DVertexShader
*vshader
;
119 IWineD3DPixelShader
*pshader
;
120 struct ps_compile_args ps_args
;
121 struct vs_compile_args vs_args
;
122 } glsl_program_key_t
;
125 /** Prints the GLSL info log which will contain error messages if they exist */
126 static void print_glsl_info_log(const WineD3D_GL_Info
*gl_info
, GLhandleARB obj
)
128 int infologLength
= 0;
133 static const char * const spam
[] =
135 "Vertex shader was successfully compiled to run on hardware.\n", /* fglrx */
136 "Fragment shader was successfully compiled to run on hardware.\n", /* fglrx */
137 "Fragment shader(s) linked, vertex shader(s) linked. \n ", /* fglrx, with \n */
138 "Fragment shader(s) linked, vertex shader(s) linked.", /* fglrx, no \n */
139 "Vertex shader(s) linked, no fragment shader(s) defined. \n ", /* fglrx, with \n */
140 "Vertex shader(s) linked, no fragment shader(s) defined.", /* fglrx, no \n */
141 "Fragment shader was successfully compiled to run on hardware.\n"
142 "WARNING: 0:2: extension 'GL_ARB_draw_buffers' is not supported",
143 "Fragment shader(s) linked, no vertex shader(s) defined.", /* fglrx, no \n */
144 "Fragment shader(s) linked, no vertex shader(s) defined. \n ", /* fglrx, with \n */
145 "WARNING: 0:2: extension 'GL_ARB_draw_buffers' is not supported\n" /* MacOS ati */
148 if (!TRACE_ON(d3d_shader
) && !FIXME_ON(d3d_shader
)) return;
150 GL_EXTCALL(glGetObjectParameterivARB(obj
,
151 GL_OBJECT_INFO_LOG_LENGTH_ARB
,
154 /* A size of 1 is just a null-terminated string, so the log should be bigger than
155 * that if there are errors. */
156 if (infologLength
> 1)
158 /* Fglrx doesn't terminate the string properly, but it tells us the proper length.
159 * So use HEAP_ZERO_MEMORY to avoid uninitialized bytes
161 infoLog
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, infologLength
);
162 GL_EXTCALL(glGetInfoLogARB(obj
, infologLength
, NULL
, infoLog
));
165 for(i
= 0; i
< sizeof(spam
) / sizeof(spam
[0]); i
++) {
166 if(strcmp(infoLog
, spam
[i
]) == 0) {
172 TRACE("Spam received from GLSL shader #%u: %s\n", obj
, debugstr_a(infoLog
));
174 FIXME("Error received from GLSL shader #%u: %s\n", obj
, debugstr_a(infoLog
));
176 HeapFree(GetProcessHeap(), 0, infoLog
);
181 * Loads (pixel shader) samplers
183 static void shader_glsl_load_psamplers(const WineD3D_GL_Info
*gl_info
, DWORD
*tex_unit_map
, GLhandleARB programId
)
187 char sampler_name
[20];
189 for (i
= 0; i
< MAX_FRAGMENT_SAMPLERS
; ++i
) {
190 snprintf(sampler_name
, sizeof(sampler_name
), "Psampler%d", i
);
191 name_loc
= GL_EXTCALL(glGetUniformLocationARB(programId
, sampler_name
));
192 if (name_loc
!= -1) {
193 DWORD mapped_unit
= tex_unit_map
[i
];
194 if (mapped_unit
!= WINED3D_UNMAPPED_STAGE
&& mapped_unit
< GL_LIMITS(fragment_samplers
))
196 TRACE("Loading %s for texture %d\n", sampler_name
, mapped_unit
);
197 GL_EXTCALL(glUniform1iARB(name_loc
, mapped_unit
));
198 checkGLcall("glUniform1iARB");
200 ERR("Trying to load sampler %s on unsupported unit %d\n", sampler_name
, mapped_unit
);
206 static void shader_glsl_load_vsamplers(const WineD3D_GL_Info
*gl_info
, DWORD
*tex_unit_map
, GLhandleARB programId
)
209 char sampler_name
[20];
212 for (i
= 0; i
< MAX_VERTEX_SAMPLERS
; ++i
) {
213 snprintf(sampler_name
, sizeof(sampler_name
), "Vsampler%d", i
);
214 name_loc
= GL_EXTCALL(glGetUniformLocationARB(programId
, sampler_name
));
215 if (name_loc
!= -1) {
216 DWORD mapped_unit
= tex_unit_map
[MAX_FRAGMENT_SAMPLERS
+ i
];
217 if (mapped_unit
!= WINED3D_UNMAPPED_STAGE
&& mapped_unit
< GL_LIMITS(combined_samplers
))
219 TRACE("Loading %s for texture %d\n", sampler_name
, mapped_unit
);
220 GL_EXTCALL(glUniform1iARB(name_loc
, mapped_unit
));
221 checkGLcall("glUniform1iARB");
223 ERR("Trying to load sampler %s on unsupported unit %d\n", sampler_name
, mapped_unit
);
229 static inline void walk_constant_heap(const WineD3D_GL_Info
*gl_info
, const float *constants
,
230 const GLint
*constant_locations
, const struct constant_heap
*heap
, unsigned char *stack
, DWORD version
)
233 unsigned int heap_idx
= 1;
236 if (heap
->entries
[heap_idx
].version
<= version
) return;
238 idx
= heap
->entries
[heap_idx
].idx
;
239 if (constant_locations
[idx
] != -1) GL_EXTCALL(glUniform4fvARB(constant_locations
[idx
], 1, &constants
[idx
* 4]));
240 stack
[stack_idx
] = HEAP_NODE_TRAVERSE_LEFT
;
242 while (stack_idx
>= 0)
244 /* Note that we fall through to the next case statement. */
245 switch(stack
[stack_idx
])
247 case HEAP_NODE_TRAVERSE_LEFT
:
249 unsigned int left_idx
= heap_idx
<< 1;
250 if (left_idx
< heap
->size
&& heap
->entries
[left_idx
].version
> version
)
253 idx
= heap
->entries
[heap_idx
].idx
;
254 if (constant_locations
[idx
] != -1)
255 GL_EXTCALL(glUniform4fvARB(constant_locations
[idx
], 1, &constants
[idx
* 4]));
257 stack
[stack_idx
++] = HEAP_NODE_TRAVERSE_RIGHT
;
258 stack
[stack_idx
] = HEAP_NODE_TRAVERSE_LEFT
;
263 case HEAP_NODE_TRAVERSE_RIGHT
:
265 unsigned int right_idx
= (heap_idx
<< 1) + 1;
266 if (right_idx
< heap
->size
&& heap
->entries
[right_idx
].version
> version
)
268 heap_idx
= right_idx
;
269 idx
= heap
->entries
[heap_idx
].idx
;
270 if (constant_locations
[idx
] != -1)
271 GL_EXTCALL(glUniform4fvARB(constant_locations
[idx
], 1, &constants
[idx
* 4]));
273 stack
[stack_idx
++] = HEAP_NODE_POP
;
274 stack
[stack_idx
] = HEAP_NODE_TRAVERSE_LEFT
;
287 checkGLcall("walk_constant_heap()");
290 static inline void apply_clamped_constant(const WineD3D_GL_Info
*gl_info
, GLint location
, const GLfloat
*data
)
292 GLfloat clamped_constant
[4];
294 if (location
== -1) return;
296 clamped_constant
[0] = data
[0] < -1.0f
? -1.0f
: data
[0] > 1.0 ? 1.0 : data
[0];
297 clamped_constant
[1] = data
[1] < -1.0f
? -1.0f
: data
[1] > 1.0 ? 1.0 : data
[1];
298 clamped_constant
[2] = data
[2] < -1.0f
? -1.0f
: data
[2] > 1.0 ? 1.0 : data
[2];
299 clamped_constant
[3] = data
[3] < -1.0f
? -1.0f
: data
[3] > 1.0 ? 1.0 : data
[3];
301 GL_EXTCALL(glUniform4fvARB(location
, 1, clamped_constant
));
304 static inline void walk_constant_heap_clamped(const WineD3D_GL_Info
*gl_info
, const float *constants
,
305 const GLint
*constant_locations
, const struct constant_heap
*heap
, unsigned char *stack
, DWORD version
)
308 unsigned int heap_idx
= 1;
311 if (heap
->entries
[heap_idx
].version
<= version
) return;
313 idx
= heap
->entries
[heap_idx
].idx
;
314 apply_clamped_constant(gl_info
, constant_locations
[idx
], &constants
[idx
* 4]);
315 stack
[stack_idx
] = HEAP_NODE_TRAVERSE_LEFT
;
317 while (stack_idx
>= 0)
319 /* Note that we fall through to the next case statement. */
320 switch(stack
[stack_idx
])
322 case HEAP_NODE_TRAVERSE_LEFT
:
324 unsigned int left_idx
= heap_idx
<< 1;
325 if (left_idx
< heap
->size
&& heap
->entries
[left_idx
].version
> version
)
328 idx
= heap
->entries
[heap_idx
].idx
;
329 apply_clamped_constant(gl_info
, constant_locations
[idx
], &constants
[idx
* 4]);
331 stack
[stack_idx
++] = HEAP_NODE_TRAVERSE_RIGHT
;
332 stack
[stack_idx
] = HEAP_NODE_TRAVERSE_LEFT
;
337 case HEAP_NODE_TRAVERSE_RIGHT
:
339 unsigned int right_idx
= (heap_idx
<< 1) + 1;
340 if (right_idx
< heap
->size
&& heap
->entries
[right_idx
].version
> version
)
342 heap_idx
= right_idx
;
343 idx
= heap
->entries
[heap_idx
].idx
;
344 apply_clamped_constant(gl_info
, constant_locations
[idx
], &constants
[idx
* 4]);
346 stack
[stack_idx
++] = HEAP_NODE_POP
;
347 stack
[stack_idx
] = HEAP_NODE_TRAVERSE_LEFT
;
360 checkGLcall("walk_constant_heap_clamped()");
363 /* Loads floating point constants (aka uniforms) into the currently set GLSL program. */
364 static void shader_glsl_load_constantsF(IWineD3DBaseShaderImpl
*This
, const WineD3D_GL_Info
*gl_info
,
365 const float *constants
, const GLint
*constant_locations
, const struct constant_heap
*heap
,
366 unsigned char *stack
, UINT version
)
368 const local_constant
*lconst
;
370 /* 1.X pshaders have the constants clamped to [-1;1] implicitly. */
371 if (WINED3DSHADER_VERSION_MAJOR(This
->baseShader
.reg_maps
.shader_version
) == 1
372 && shader_is_pshader_version(This
->baseShader
.reg_maps
.shader_version
))
373 walk_constant_heap_clamped(gl_info
, constants
, constant_locations
, heap
, stack
, version
);
375 walk_constant_heap(gl_info
, constants
, constant_locations
, heap
, stack
, version
);
377 if (!This
->baseShader
.load_local_constsF
)
379 TRACE("No need to load local float constants for this shader\n");
383 /* Immediate constants are clamped to [-1;1] at shader creation time if needed */
384 LIST_FOR_EACH_ENTRY(lconst
, &This
->baseShader
.constantsF
, local_constant
, entry
)
386 GLint location
= constant_locations
[lconst
->idx
];
387 /* We found this uniform name in the program - go ahead and send the data */
388 if (location
!= -1) GL_EXTCALL(glUniform4fvARB(location
, 1, (const GLfloat
*)lconst
->value
));
390 checkGLcall("glUniform4fvARB()");
393 /* Loads integer constants (aka uniforms) into the currently set GLSL program. */
394 static void shader_glsl_load_constantsI(IWineD3DBaseShaderImpl
*This
, const WineD3D_GL_Info
*gl_info
,
395 const GLint locations
[MAX_CONST_I
], const int *constants
, WORD constants_set
)
400 for (i
= 0; constants_set
; constants_set
>>= 1, ++i
)
402 if (!(constants_set
& 1)) continue;
404 TRACE_(d3d_constants
)("Loading constants %u: %i, %i, %i, %i\n",
405 i
, constants
[i
*4], constants
[i
*4+1], constants
[i
*4+2], constants
[i
*4+3]);
407 /* We found this uniform name in the program - go ahead and send the data */
408 GL_EXTCALL(glUniform4ivARB(locations
[i
], 1, &constants
[i
*4]));
409 checkGLcall("glUniform4ivARB");
412 /* Load immediate constants */
413 ptr
= list_head(&This
->baseShader
.constantsI
);
415 const struct local_constant
*lconst
= LIST_ENTRY(ptr
, const struct local_constant
, entry
);
416 unsigned int idx
= lconst
->idx
;
417 const GLint
*values
= (const GLint
*)lconst
->value
;
419 TRACE_(d3d_constants
)("Loading local constants %i: %i, %i, %i, %i\n", idx
,
420 values
[0], values
[1], values
[2], values
[3]);
422 /* We found this uniform name in the program - go ahead and send the data */
423 GL_EXTCALL(glUniform4ivARB(locations
[idx
], 1, values
));
424 checkGLcall("glUniform4ivARB");
425 ptr
= list_next(&This
->baseShader
.constantsI
, ptr
);
429 /* Loads boolean constants (aka uniforms) into the currently set GLSL program. */
430 static void shader_glsl_load_constantsB(IWineD3DBaseShaderImpl
*This
, const WineD3D_GL_Info
*gl_info
,
431 GLhandleARB programId
, const BOOL
*constants
, WORD constants_set
)
436 char is_pshader
= shader_is_pshader_version(This
->baseShader
.reg_maps
.shader_version
);
437 const char* prefix
= is_pshader
? "PB":"VB";
440 /* TODO: Benchmark and see if it would be beneficial to store the
441 * locations of the constants to avoid looking up each time */
442 for (i
= 0; constants_set
; constants_set
>>= 1, ++i
)
444 if (!(constants_set
& 1)) continue;
446 TRACE_(d3d_constants
)("Loading constants %i: %i;\n", i
, constants
[i
]);
448 /* TODO: Benchmark and see if it would be beneficial to store the
449 * locations of the constants to avoid looking up each time */
450 snprintf(tmp_name
, sizeof(tmp_name
), "%s[%i]", prefix
, i
);
451 tmp_loc
= GL_EXTCALL(glGetUniformLocationARB(programId
, tmp_name
));
454 /* We found this uniform name in the program - go ahead and send the data */
455 GL_EXTCALL(glUniform1ivARB(tmp_loc
, 1, &constants
[i
]));
456 checkGLcall("glUniform1ivARB");
460 /* Load immediate constants */
461 ptr
= list_head(&This
->baseShader
.constantsB
);
463 const struct local_constant
*lconst
= LIST_ENTRY(ptr
, const struct local_constant
, entry
);
464 unsigned int idx
= lconst
->idx
;
465 const GLint
*values
= (const GLint
*)lconst
->value
;
467 TRACE_(d3d_constants
)("Loading local constants %i: %i\n", idx
, values
[0]);
469 snprintf(tmp_name
, sizeof(tmp_name
), "%s[%i]", prefix
, idx
);
470 tmp_loc
= GL_EXTCALL(glGetUniformLocationARB(programId
, tmp_name
));
472 /* We found this uniform name in the program - go ahead and send the data */
473 GL_EXTCALL(glUniform1ivARB(tmp_loc
, 1, values
));
474 checkGLcall("glUniform1ivARB");
476 ptr
= list_next(&This
->baseShader
.constantsB
, ptr
);
480 static void reset_program_constant_version(void *value
, void *context
)
482 struct glsl_shader_prog_link
*entry
= value
;
483 entry
->constant_version
= 0;
487 * Loads the texture dimensions for NP2 fixup into the currently set GLSL program.
489 static void shader_glsl_load_np2fixup_constants(
490 IWineD3DDevice
* device
,
492 char useVertexShader
) {
494 const IWineD3DDeviceImpl
* deviceImpl
= (const IWineD3DDeviceImpl
*) device
;
495 const struct glsl_shader_prog_link
* prog
= ((struct shader_glsl_priv
*)(deviceImpl
->shader_priv
))->glsl_program
;
498 /* No GLSL program set - nothing to do. */
502 if (!usePixelShader
) {
503 /* NP2 texcoord fixup is (currently) only done for pixelshaders. */
507 if (prog
->ps_args
.np2_fixup
) {
509 UINT fixup
= prog
->ps_args
.np2_fixup
;
510 const WineD3D_GL_Info
* gl_info
= &deviceImpl
->adapter
->gl_info
;
511 const IWineD3DStateBlockImpl
* stateBlock
= (const IWineD3DStateBlockImpl
*) deviceImpl
->stateBlock
;
513 for (i
= 0; fixup
; fixup
>>= 1, ++i
) {
514 if (-1 != prog
->np2Fixup_location
[i
]) {
515 const IWineD3DBaseTextureImpl
* const tex
= (const IWineD3DBaseTextureImpl
*) stateBlock
->textures
[i
];
517 FIXME("Nonexistent texture is flagged for NP2 texcoord fixup\n");
520 const float tex_dim
[2] = {tex
->baseTexture
.pow2Matrix
[0], tex
->baseTexture
.pow2Matrix
[5]};
521 GL_EXTCALL(glUniform2fvARB(prog
->np2Fixup_location
[i
], 1, tex_dim
));
529 * Loads the app-supplied constants into the currently set GLSL program.
531 static void shader_glsl_load_constants(
532 IWineD3DDevice
* device
,
534 char useVertexShader
) {
536 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) device
;
537 struct shader_glsl_priv
*priv
= deviceImpl
->shader_priv
;
538 IWineD3DStateBlockImpl
* stateBlock
= deviceImpl
->stateBlock
;
539 const WineD3D_GL_Info
*gl_info
= &deviceImpl
->adapter
->gl_info
;
541 GLhandleARB programId
;
542 struct glsl_shader_prog_link
*prog
= priv
->glsl_program
;
543 UINT constant_version
;
547 /* No GLSL program set - nothing to do. */
550 programId
= prog
->programId
;
551 constant_version
= prog
->constant_version
;
553 if (useVertexShader
) {
554 IWineD3DBaseShaderImpl
* vshader
= (IWineD3DBaseShaderImpl
*) stateBlock
->vertexShader
;
556 /* Load DirectX 9 float constants/uniforms for vertex shader */
557 shader_glsl_load_constantsF(vshader
, gl_info
, stateBlock
->vertexShaderConstantF
,
558 prog
->vuniformF_locations
, &priv
->vconst_heap
, priv
->stack
, constant_version
);
560 /* Load DirectX 9 integer constants/uniforms for vertex shader */
561 shader_glsl_load_constantsI(vshader
, gl_info
, prog
->vuniformI_locations
, stateBlock
->vertexShaderConstantI
,
562 stateBlock
->changed
.vertexShaderConstantsI
& vshader
->baseShader
.reg_maps
.integer_constants
);
564 /* Load DirectX 9 boolean constants/uniforms for vertex shader */
565 shader_glsl_load_constantsB(vshader
, gl_info
, programId
, stateBlock
->vertexShaderConstantB
,
566 stateBlock
->changed
.vertexShaderConstantsB
& vshader
->baseShader
.reg_maps
.boolean_constants
);
568 /* Upload the position fixup params */
569 GL_EXTCALL(glUniform4fvARB(prog
->posFixup_location
, 1, &deviceImpl
->posFixup
[0]));
570 checkGLcall("glUniform4fvARB");
573 if (usePixelShader
) {
575 IWineD3DBaseShaderImpl
* pshader
= (IWineD3DBaseShaderImpl
*) stateBlock
->pixelShader
;
577 /* Load DirectX 9 float constants/uniforms for pixel shader */
578 shader_glsl_load_constantsF(pshader
, gl_info
, stateBlock
->pixelShaderConstantF
,
579 prog
->puniformF_locations
, &priv
->pconst_heap
, priv
->stack
, constant_version
);
581 /* Load DirectX 9 integer constants/uniforms for pixel shader */
582 shader_glsl_load_constantsI(pshader
, gl_info
, prog
->puniformI_locations
, stateBlock
->pixelShaderConstantI
,
583 stateBlock
->changed
.pixelShaderConstantsI
& pshader
->baseShader
.reg_maps
.integer_constants
);
585 /* Load DirectX 9 boolean constants/uniforms for pixel shader */
586 shader_glsl_load_constantsB(pshader
, gl_info
, programId
, stateBlock
->pixelShaderConstantB
,
587 stateBlock
->changed
.pixelShaderConstantsB
& pshader
->baseShader
.reg_maps
.boolean_constants
);
589 /* Upload the environment bump map matrix if needed. The needsbumpmat member specifies the texture stage to load the matrix from.
590 * It can't be 0 for a valid texbem instruction.
592 for(i
= 0; i
< ((IWineD3DPixelShaderImpl
*) pshader
)->numbumpenvmatconsts
; i
++) {
593 IWineD3DPixelShaderImpl
*ps
= (IWineD3DPixelShaderImpl
*) pshader
;
594 int stage
= ps
->luminanceconst
[i
].texunit
;
596 const float *data
= (const float *)&stateBlock
->textureState
[(int)ps
->bumpenvmatconst
[i
].texunit
][WINED3DTSS_BUMPENVMAT00
];
597 GL_EXTCALL(glUniformMatrix2fvARB(prog
->bumpenvmat_location
[i
], 1, 0, data
));
598 checkGLcall("glUniformMatrix2fvARB");
600 /* texbeml needs the luminance scale and offset too. If texbeml is used, needsbumpmat
601 * is set too, so we can check that in the needsbumpmat check
603 if(ps
->baseShader
.reg_maps
.luminanceparams
[stage
]) {
604 const GLfloat
*scale
= (const GLfloat
*)&stateBlock
->textureState
[stage
][WINED3DTSS_BUMPENVLSCALE
];
605 const GLfloat
*offset
= (const GLfloat
*)&stateBlock
->textureState
[stage
][WINED3DTSS_BUMPENVLOFFSET
];
607 GL_EXTCALL(glUniform1fvARB(prog
->luminancescale_location
[i
], 1, scale
));
608 checkGLcall("glUniform1fvARB");
609 GL_EXTCALL(glUniform1fvARB(prog
->luminanceoffset_location
[i
], 1, offset
));
610 checkGLcall("glUniform1fvARB");
614 if(((IWineD3DPixelShaderImpl
*) pshader
)->vpos_uniform
) {
615 float correction_params
[4];
616 if(deviceImpl
->render_offscreen
) {
617 correction_params
[0] = 0.0;
618 correction_params
[1] = 1.0;
620 /* position is window relative, not viewport relative */
621 correction_params
[0] = ((IWineD3DSurfaceImpl
*) deviceImpl
->render_targets
[0])->currentDesc
.Height
;
622 correction_params
[1] = -1.0;
624 GL_EXTCALL(glUniform4fvARB(prog
->ycorrection_location
, 1, correction_params
));
628 if (priv
->next_constant_version
== UINT_MAX
)
630 TRACE("Max constant version reached, resetting to 0.\n");
631 hash_table_for_each_entry(priv
->glsl_program_lookup
, reset_program_constant_version
, NULL
);
632 priv
->next_constant_version
= 1;
636 prog
->constant_version
= priv
->next_constant_version
++;
640 static inline void update_heap_entry(struct constant_heap
*heap
, unsigned int idx
,
641 unsigned int heap_idx
, DWORD new_version
)
643 struct constant_entry
*entries
= heap
->entries
;
644 unsigned int *positions
= heap
->positions
;
645 unsigned int parent_idx
;
649 parent_idx
= heap_idx
>> 1;
651 if (new_version
<= entries
[parent_idx
].version
) break;
653 entries
[heap_idx
] = entries
[parent_idx
];
654 positions
[entries
[parent_idx
].idx
] = heap_idx
;
655 heap_idx
= parent_idx
;
658 entries
[heap_idx
].version
= new_version
;
659 entries
[heap_idx
].idx
= idx
;
660 positions
[idx
] = heap_idx
;
663 static void shader_glsl_update_float_vertex_constants(IWineD3DDevice
*iface
, UINT start
, UINT count
)
665 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
666 struct shader_glsl_priv
*priv
= This
->shader_priv
;
667 struct constant_heap
*heap
= &priv
->vconst_heap
;
670 for (i
= start
; i
< count
+ start
; ++i
)
672 if (!This
->stateBlock
->changed
.vertexShaderConstantsF
[i
])
673 update_heap_entry(heap
, i
, heap
->size
++, priv
->next_constant_version
);
675 update_heap_entry(heap
, i
, heap
->positions
[i
], priv
->next_constant_version
);
679 static void shader_glsl_update_float_pixel_constants(IWineD3DDevice
*iface
, UINT start
, UINT count
)
681 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
682 struct shader_glsl_priv
*priv
= This
->shader_priv
;
683 struct constant_heap
*heap
= &priv
->pconst_heap
;
686 for (i
= start
; i
< count
+ start
; ++i
)
688 if (!This
->stateBlock
->changed
.pixelShaderConstantsF
[i
])
689 update_heap_entry(heap
, i
, heap
->size
++, priv
->next_constant_version
);
691 update_heap_entry(heap
, i
, heap
->positions
[i
], priv
->next_constant_version
);
695 /** Generate the variable & register declarations for the GLSL output target */
696 static void shader_generate_glsl_declarations(IWineD3DBaseShader
*iface
, const shader_reg_maps
*reg_maps
,
697 SHADER_BUFFER
*buffer
, const WineD3D_GL_Info
*gl_info
,
698 const struct ps_compile_args
*ps_args
)
700 IWineD3DBaseShaderImpl
* This
= (IWineD3DBaseShaderImpl
*) iface
;
701 IWineD3DDeviceImpl
*device
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
702 DWORD shader_version
= reg_maps
->shader_version
;
703 unsigned int i
, extra_constants_needed
= 0;
704 const local_constant
*lconst
;
706 /* There are some minor differences between pixel and vertex shaders */
707 char pshader
= shader_is_pshader_version(shader_version
);
708 char prefix
= pshader
? 'P' : 'V';
710 /* Prototype the subroutines */
711 for (i
= 0; i
< This
->baseShader
.limits
.label
; i
++) {
712 if (reg_maps
->labels
[i
])
713 shader_addline(buffer
, "void subroutine%u();\n", i
);
716 /* Declare the constants (aka uniforms) */
717 if (This
->baseShader
.limits
.constant_float
> 0) {
718 unsigned max_constantsF
;
719 /* Unless the shader uses indirect addressing, always declare the maximum array size and ignore that we need some
720 * uniforms privately. E.g. if GL supports 256 uniforms, and we need 2 for the pos fixup and immediate values, still
721 * declare VC[256]. If the shader needs more uniforms than we have it won't work in any case. If it uses less, the
722 * compiler will figure out which uniforms are really used and strip them out. This allows a shader to use c255 on
723 * a dx9 card, as long as it doesn't also use all the other constants.
725 * If the shader uses indirect addressing the compiler must assume that all declared uniforms are used. In this case,
726 * declare only the amount that we're assured to have.
728 * Thus we run into problems in these two cases:
729 * 1) The shader really uses more uniforms than supported
730 * 2) The shader uses indirect addressing, less constants than supported, but uses a constant index > #supported consts
733 /* No indirect addressing here */
734 max_constantsF
= GL_LIMITS(pshader_constantsF
);
736 if(This
->baseShader
.reg_maps
.usesrelconstF
) {
737 /* Subtract the other potential uniforms from the max available (bools, ints, and 1 row of projection matrix).
738 * Subtract another uniform for immediate values, which have to be loaded via uniform by the driver as well.
739 * The shader code only uses 0.5, 2.0, 1.0, 128 and -128 in vertex shader code, so one vec4 should be enough
740 * (Unfortunately the Nvidia driver doesn't store 128 and -128 in one float
742 max_constantsF
= GL_LIMITS(vshader_constantsF
) - 3;
743 max_constantsF
-= count_bits(This
->baseShader
.reg_maps
.integer_constants
);
744 /* Strictly speaking a bool only uses one scalar, but the nvidia(Linux) compiler doesn't pack them properly,
745 * so each scalar requires a full vec4. We could work around this by packing the booleans ourselves, but
746 * for now take this into account when calculating the number of available constants
748 max_constantsF
-= count_bits(This
->baseShader
.reg_maps
.boolean_constants
);
749 /* Set by driver quirks in directx.c */
750 max_constantsF
-= GLINFO_LOCATION
.reserved_glsl_constants
;
752 max_constantsF
= GL_LIMITS(vshader_constantsF
);
755 max_constantsF
= min(This
->baseShader
.limits
.constant_float
, max_constantsF
);
756 shader_addline(buffer
, "uniform vec4 %cC[%u];\n", prefix
, max_constantsF
);
759 /* Always declare the full set of constants, the compiler can remove the unused ones because d3d doesn't(yet)
760 * support indirect int and bool constant addressing. This avoids problems if the app uses e.g. i0 and i9.
762 if (This
->baseShader
.limits
.constant_int
> 0 && This
->baseShader
.reg_maps
.integer_constants
)
763 shader_addline(buffer
, "uniform ivec4 %cI[%u];\n", prefix
, This
->baseShader
.limits
.constant_int
);
765 if (This
->baseShader
.limits
.constant_bool
> 0 && This
->baseShader
.reg_maps
.boolean_constants
)
766 shader_addline(buffer
, "uniform bool %cB[%u];\n", prefix
, This
->baseShader
.limits
.constant_bool
);
769 shader_addline(buffer
, "uniform vec4 posFixup;\n");
770 /* Predeclaration; This function is added at link time based on the pixel shader.
771 * VS 3.0 shaders have an array OUT[] the shader writes to, earlier versions don't have
772 * that. We know the input to the reorder function at vertex shader compile time, so
773 * we can deal with that. The reorder function for a 1.x and 2.x vertex shader can just
774 * read gl_FrontColor. The output depends on the pixel shader. The reorder function for a
775 * 1.x and 2.x pshader or for fixed function will write gl_FrontColor, and for a 3.0 shader
776 * it will write to the varying array. Here we depend on the shader optimizer on sorting that
777 * out. The nvidia driver only does that if the parameter is inout instead of out, hence the
780 if (shader_version
>= WINED3DVS_VERSION(3, 0))
782 shader_addline(buffer
, "void order_ps_input(in vec4[%u]);\n", MAX_REG_OUTPUT
);
784 shader_addline(buffer
, "void order_ps_input();\n");
787 IWineD3DPixelShaderImpl
*ps_impl
= (IWineD3DPixelShaderImpl
*) This
;
789 ps_impl
->numbumpenvmatconsts
= 0;
790 for(i
= 0; i
< (sizeof(reg_maps
->bumpmat
) / sizeof(reg_maps
->bumpmat
[0])); i
++) {
791 if(!reg_maps
->bumpmat
[i
]) {
795 ps_impl
->bumpenvmatconst
[(int) ps_impl
->numbumpenvmatconsts
].texunit
= i
;
796 shader_addline(buffer
, "uniform mat2 bumpenvmat%d;\n", i
);
798 if(reg_maps
->luminanceparams
) {
799 ps_impl
->luminanceconst
[(int) ps_impl
->numbumpenvmatconsts
].texunit
= i
;
800 shader_addline(buffer
, "uniform float luminancescale%d;\n", i
);
801 shader_addline(buffer
, "uniform float luminanceoffset%d;\n", i
);
802 extra_constants_needed
++;
804 ps_impl
->luminanceconst
[(int) ps_impl
->numbumpenvmatconsts
].texunit
= -1;
807 extra_constants_needed
++;
808 ps_impl
->numbumpenvmatconsts
++;
811 if(ps_args
->srgb_correction
) {
812 shader_addline(buffer
, "const vec4 srgb_mul_low = vec4(%f, %f, %f, %f);\n",
813 srgb_mul_low
, srgb_mul_low
, srgb_mul_low
, srgb_mul_low
);
814 shader_addline(buffer
, "const vec4 srgb_comparison = vec4(%f, %f, %f, %f);\n",
815 srgb_cmp
, srgb_cmp
, srgb_cmp
, srgb_cmp
);
817 if(reg_maps
->vpos
|| reg_maps
->usesdsy
) {
818 if(This
->baseShader
.limits
.constant_float
+ extra_constants_needed
+ 1 < GL_LIMITS(pshader_constantsF
)) {
819 shader_addline(buffer
, "uniform vec4 ycorrection;\n");
820 ((IWineD3DPixelShaderImpl
*) This
)->vpos_uniform
= 1;
821 extra_constants_needed
++;
823 /* This happens because we do not have proper tracking of the constant registers that are
824 * actually used, only the max limit of the shader version
826 FIXME("Cannot find a free uniform for vpos correction params\n");
827 shader_addline(buffer
, "const vec4 ycorrection = vec4(%f, %f, 0.0, 0.0);\n",
828 device
->render_offscreen
? 0.0 : ((IWineD3DSurfaceImpl
*) device
->render_targets
[0])->currentDesc
.Height
,
829 device
->render_offscreen
? 1.0 : -1.0);
831 shader_addline(buffer
, "vec4 vpos;\n");
835 /* Declare texture samplers */
836 for (i
= 0; i
< This
->baseShader
.limits
.sampler
; i
++) {
837 if (reg_maps
->sampler_type
[i
])
839 switch (reg_maps
->sampler_type
[i
])
842 shader_addline(buffer
, "uniform sampler1D %csampler%u;\n", prefix
, i
);
845 if(device
->stateBlock
->textures
[i
] &&
846 IWineD3DBaseTexture_GetTextureDimensions(device
->stateBlock
->textures
[i
]) == GL_TEXTURE_RECTANGLE_ARB
) {
847 shader_addline(buffer
, "uniform sampler2DRect %csampler%u;\n", prefix
, i
);
849 shader_addline(buffer
, "uniform sampler2D %csampler%u;\n", prefix
, i
);
852 if (pshader
&& ps_args
->np2_fixup
& (1 << i
))
854 /* NP2/RECT textures in OpenGL use texcoords in the range [0,width]x[0,height]
855 * while D3D has them in the (normalized) [0,1]x[0,1] range.
856 * samplerNP2Fixup stores texture dimensions and is updated through
857 * shader_glsl_load_np2fixup_constants when the sampler changes. */
858 shader_addline(buffer
, "uniform vec2 %csamplerNP2Fixup%u;\n", prefix
, i
);
861 case WINED3DSTT_CUBE
:
862 shader_addline(buffer
, "uniform samplerCube %csampler%u;\n", prefix
, i
);
864 case WINED3DSTT_VOLUME
:
865 shader_addline(buffer
, "uniform sampler3D %csampler%u;\n", prefix
, i
);
868 shader_addline(buffer
, "uniform unsupported_sampler %csampler%u;\n", prefix
, i
);
869 FIXME("Unrecognized sampler type: %#x\n", reg_maps
->sampler_type
[i
]);
875 /* Declare address variables */
876 for (i
= 0; i
< This
->baseShader
.limits
.address
; i
++) {
877 if (reg_maps
->address
[i
])
878 shader_addline(buffer
, "ivec4 A%d;\n", i
);
881 /* Declare texture coordinate temporaries and initialize them */
882 for (i
= 0; i
< This
->baseShader
.limits
.texcoord
; i
++) {
883 if (reg_maps
->texcoord
[i
])
884 shader_addline(buffer
, "vec4 T%u = gl_TexCoord[%u];\n", i
, i
);
887 /* Declare input register varyings. Only pixel shader, vertex shaders have that declared in the
888 * helper function shader that is linked in at link time
890 if (pshader
&& shader_version
>= WINED3DPS_VERSION(3, 0))
892 if (use_vs(device
->stateBlock
))
894 shader_addline(buffer
, "varying vec4 IN[%u];\n", GL_LIMITS(glsl_varyings
) / 4);
896 /* TODO: Write a replacement shader for the fixed function vertex pipeline, so this isn't needed.
897 * For fixed function vertex processing + 3.0 pixel shader we need a separate function in the
898 * pixel shader that reads the fixed function color into the packed input registers.
900 shader_addline(buffer
, "vec4 IN[%u];\n", GL_LIMITS(glsl_varyings
) / 4);
904 /* Declare output register temporaries */
905 if(This
->baseShader
.limits
.packed_output
) {
906 shader_addline(buffer
, "vec4 OUT[%u];\n", This
->baseShader
.limits
.packed_output
);
909 /* Declare temporary variables */
910 for(i
= 0; i
< This
->baseShader
.limits
.temporary
; i
++) {
911 if (reg_maps
->temporary
[i
])
912 shader_addline(buffer
, "vec4 R%u;\n", i
);
915 /* Declare attributes */
916 for (i
= 0; i
< This
->baseShader
.limits
.attributes
; i
++) {
917 if (reg_maps
->attributes
[i
])
918 shader_addline(buffer
, "attribute vec4 attrib%i;\n", i
);
921 /* Declare loop registers aLx */
922 for (i
= 0; i
< reg_maps
->loop_depth
; i
++) {
923 shader_addline(buffer
, "int aL%u;\n", i
);
924 shader_addline(buffer
, "int tmpInt%u;\n", i
);
927 /* Temporary variables for matrix operations */
928 shader_addline(buffer
, "vec4 tmp0;\n");
929 shader_addline(buffer
, "vec4 tmp1;\n");
931 /* Local constants use a different name so they can be loaded once at shader link time
932 * They can't be hardcoded into the shader text via LC = {x, y, z, w}; because the
933 * float -> string conversion can cause precision loss.
935 if(!This
->baseShader
.load_local_constsF
) {
936 LIST_FOR_EACH_ENTRY(lconst
, &This
->baseShader
.constantsF
, local_constant
, entry
) {
937 shader_addline(buffer
, "uniform vec4 %cLC%u;\n", prefix
, lconst
->idx
);
941 /* Start the main program */
942 shader_addline(buffer
, "void main() {\n");
943 if(pshader
&& reg_maps
->vpos
) {
944 /* DirectX apps expect integer values, while OpenGL drivers add approximately 0.5. This causes
945 * off-by-one problems as spotted by the vPos d3d9 visual test. Unfortunately the ATI cards do
946 * not add exactly 0.5, but rather something like 0.49999999 or 0.50000001, which still causes
947 * precision troubles when we just substract 0.5.
949 * To deal with that just floor() the position. This will eliminate the fraction on all cards.
951 * TODO: Test how that behaves with multisampling once we can enable multisampling in winex11.
953 * An advantage of floor is that it works even if the driver doesn't add 1/2. It is somewhat
954 * questionable if 1.5, 2.5, ... are the proper values to return in gl_FragCoord, even though
955 * coordinates specify the pixel centers instead of the pixel corners. This code will behave
956 * correctly on drivers that returns integer values.
958 shader_addline(buffer
, "vpos = floor(vec4(0, ycorrection[0], 0, 0) + gl_FragCoord * vec4(1, ycorrection[1], 1, 1));\n");
962 /*****************************************************************************
963 * Functions to generate GLSL strings from DirectX Shader bytecode begin here.
965 * For more information, see http://wiki.winehq.org/DirectX-Shaders
966 ****************************************************************************/
969 static void shader_glsl_add_src_param(const struct wined3d_shader_instruction
*ins
,
970 const struct wined3d_shader_src_param
*wined3d_src
, DWORD mask
, glsl_src_param_t
*glsl_src
);
972 /** Used for opcode modifiers - They multiply the result by the specified amount */
973 static const char * const shift_glsl_tab
[] = {
975 "2.0 * ", /* 1 (x2) */
976 "4.0 * ", /* 2 (x4) */
977 "8.0 * ", /* 3 (x8) */
978 "16.0 * ", /* 4 (x16) */
979 "32.0 * ", /* 5 (x32) */
986 "0.0625 * ", /* 12 (d16) */
987 "0.125 * ", /* 13 (d8) */
988 "0.25 * ", /* 14 (d4) */
989 "0.5 * " /* 15 (d2) */
992 /* Generate a GLSL parameter that does the input modifier computation and return the input register/mask to use */
993 static void shader_glsl_gen_modifier(DWORD src_modifier
, const char *in_reg
, const char *in_regswizzle
, char *out_str
)
997 switch (src_modifier
)
999 case WINED3DSPSM_DZ
: /* Need to handle this in the instructions itself (texld & texcrd). */
1000 case WINED3DSPSM_DW
:
1001 case WINED3DSPSM_NONE
:
1002 sprintf(out_str
, "%s%s", in_reg
, in_regswizzle
);
1004 case WINED3DSPSM_NEG
:
1005 sprintf(out_str
, "-%s%s", in_reg
, in_regswizzle
);
1007 case WINED3DSPSM_NOT
:
1008 sprintf(out_str
, "!%s%s", in_reg
, in_regswizzle
);
1010 case WINED3DSPSM_BIAS
:
1011 sprintf(out_str
, "(%s%s - vec4(0.5)%s)", in_reg
, in_regswizzle
, in_regswizzle
);
1013 case WINED3DSPSM_BIASNEG
:
1014 sprintf(out_str
, "-(%s%s - vec4(0.5)%s)", in_reg
, in_regswizzle
, in_regswizzle
);
1016 case WINED3DSPSM_SIGN
:
1017 sprintf(out_str
, "(2.0 * (%s%s - 0.5))", in_reg
, in_regswizzle
);
1019 case WINED3DSPSM_SIGNNEG
:
1020 sprintf(out_str
, "-(2.0 * (%s%s - 0.5))", in_reg
, in_regswizzle
);
1022 case WINED3DSPSM_COMP
:
1023 sprintf(out_str
, "(1.0 - %s%s)", in_reg
, in_regswizzle
);
1025 case WINED3DSPSM_X2
:
1026 sprintf(out_str
, "(2.0 * %s%s)", in_reg
, in_regswizzle
);
1028 case WINED3DSPSM_X2NEG
:
1029 sprintf(out_str
, "-(2.0 * %s%s)", in_reg
, in_regswizzle
);
1031 case WINED3DSPSM_ABS
:
1032 sprintf(out_str
, "abs(%s%s)", in_reg
, in_regswizzle
);
1034 case WINED3DSPSM_ABSNEG
:
1035 sprintf(out_str
, "-abs(%s%s)", in_reg
, in_regswizzle
);
1038 FIXME("Unhandled modifier %u\n", src_modifier
);
1039 sprintf(out_str
, "%s%s", in_reg
, in_regswizzle
);
1043 /** Writes the GLSL variable name that corresponds to the register that the
1044 * DX opcode parameter is trying to access */
1045 static void shader_glsl_get_register_name(WINED3DSHADER_PARAM_REGISTER_TYPE register_type
, UINT register_idx
,
1046 const struct wined3d_shader_src_param
*rel_addr
, char *register_name
, BOOL
*is_color
,
1047 const struct wined3d_shader_instruction
*ins
)
1049 /* oPos, oFog and oPts in D3D */
1050 static const char * const hwrastout_reg_names
[] = { "gl_Position", "gl_FogFragCoord", "gl_PointSize" };
1052 IWineD3DBaseShaderImpl
*This
= (IWineD3DBaseShaderImpl
*)ins
->ctx
->shader
;
1053 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
1054 const WineD3D_GL_Info
* gl_info
= &deviceImpl
->adapter
->gl_info
;
1055 DWORD shader_version
= This
->baseShader
.reg_maps
.shader_version
;
1056 char pshader
= shader_is_pshader_version(shader_version
);
1060 switch (register_type
)
1062 case WINED3DSPR_TEMP
:
1063 sprintf(register_name
, "R%u", register_idx
);
1065 case WINED3DSPR_INPUT
:
1067 /* Pixel shaders >= 3.0 */
1068 if (WINED3DSHADER_VERSION_MAJOR(shader_version
) >= 3)
1070 DWORD in_count
= GL_LIMITS(glsl_varyings
) / 4;
1074 glsl_src_param_t rel_param
;
1075 shader_glsl_add_src_param(ins
, rel_addr
, WINED3DSP_WRITEMASK_0
, &rel_param
);
1077 /* Removing a + 0 would be an obvious optimization, but macos doesn't see the NOP
1080 if (((IWineD3DPixelShaderImpl
*)This
)->input_reg_map
[register_idx
])
1082 if (((IWineD3DPixelShaderImpl
*)This
)->declared_in_count
> in_count
) {
1083 sprintf(register_name
, "((%s + %u) > %d ? (%s + %u) > %d ? gl_SecondaryColor : gl_Color : IN[%s + %u])",
1084 rel_param
.param_str
, ((IWineD3DPixelShaderImpl
*)This
)->input_reg_map
[register_idx
], in_count
- 1,
1085 rel_param
.param_str
, ((IWineD3DPixelShaderImpl
*)This
)->input_reg_map
[register_idx
], in_count
,
1086 rel_param
.param_str
, ((IWineD3DPixelShaderImpl
*)This
)->input_reg_map
[register_idx
]);
1088 sprintf(register_name
, "IN[%s + %u]", rel_param
.param_str
,
1089 ((IWineD3DPixelShaderImpl
*)This
)->input_reg_map
[register_idx
]);
1092 if (((IWineD3DPixelShaderImpl
*)This
)->declared_in_count
> in_count
) {
1093 sprintf(register_name
, "((%s) > %d ? (%s) > %d ? gl_SecondaryColor : gl_Color : IN[%s])",
1094 rel_param
.param_str
, in_count
- 1,
1095 rel_param
.param_str
, in_count
,
1096 rel_param
.param_str
);
1098 sprintf(register_name
, "IN[%s]", rel_param
.param_str
);
1102 DWORD idx
= ((IWineD3DPixelShaderImpl
*) This
)->input_reg_map
[register_idx
];
1103 if (idx
== in_count
) {
1104 sprintf(register_name
, "gl_Color");
1105 } else if (idx
== in_count
+ 1) {
1106 sprintf(register_name
, "gl_SecondaryColor");
1108 sprintf(register_name
, "IN[%u]", idx
);
1112 if (register_idx
== 0)
1113 strcpy(register_name
, "gl_Color");
1115 strcpy(register_name
, "gl_SecondaryColor");
1118 if (((IWineD3DVertexShaderImpl
*)This
)->cur_args
->swizzle_map
& (1 << register_idx
)) *is_color
= TRUE
;
1119 sprintf(register_name
, "attrib%u", register_idx
);
1122 case WINED3DSPR_CONST
:
1124 const char prefix
= pshader
? 'P':'V';
1126 /* Relative addressing */
1129 glsl_src_param_t rel_param
;
1130 shader_glsl_add_src_param(ins
, rel_addr
, WINED3DSP_WRITEMASK_0
, &rel_param
);
1132 sprintf(register_name
, "%cC[%s + %u]", prefix
, rel_param
.param_str
, register_idx
);
1134 sprintf(register_name
, "%cC[%s]", prefix
, rel_param
.param_str
);
1136 if (shader_constant_is_local(This
, register_idx
))
1138 sprintf(register_name
, "%cLC%u", prefix
, register_idx
);
1140 sprintf(register_name
, "%cC[%u]", prefix
, register_idx
);
1146 case WINED3DSPR_CONSTINT
:
1148 sprintf(register_name
, "PI[%u]", register_idx
);
1150 sprintf(register_name
, "VI[%u]", register_idx
);
1152 case WINED3DSPR_CONSTBOOL
:
1154 sprintf(register_name
, "PB[%u]", register_idx
);
1156 sprintf(register_name
, "VB[%u]", register_idx
);
1158 case WINED3DSPR_TEXTURE
: /* case WINED3DSPR_ADDR: */
1160 sprintf(register_name
, "T%u", register_idx
);
1162 sprintf(register_name
, "A%u", register_idx
);
1165 case WINED3DSPR_LOOP
:
1166 sprintf(register_name
, "aL%u", This
->baseShader
.cur_loop_regno
- 1);
1168 case WINED3DSPR_SAMPLER
:
1170 sprintf(register_name
, "Psampler%u", register_idx
);
1172 sprintf(register_name
, "Vsampler%u", register_idx
);
1174 case WINED3DSPR_COLOROUT
:
1175 if (register_idx
>= GL_LIMITS(buffers
))
1176 WARN("Write to render target %u, only %d supported\n", register_idx
, 4);
1178 if (GL_SUPPORT(ARB_DRAW_BUFFERS
)) {
1179 sprintf(register_name
, "gl_FragData[%u]", register_idx
);
1180 } else { /* On older cards with GLSL support like the GeforceFX there's only one buffer. */
1181 sprintf(register_name
, "gl_FragColor");
1184 case WINED3DSPR_RASTOUT
:
1185 sprintf(register_name
, "%s", hwrastout_reg_names
[register_idx
]);
1187 case WINED3DSPR_DEPTHOUT
:
1188 sprintf(register_name
, "gl_FragDepth");
1190 case WINED3DSPR_ATTROUT
:
1191 if (register_idx
== 0)
1193 sprintf(register_name
, "gl_FrontColor");
1195 sprintf(register_name
, "gl_FrontSecondaryColor");
1198 case WINED3DSPR_TEXCRDOUT
:
1199 /* Vertex shaders >= 3.0: WINED3DSPR_OUTPUT */
1200 if (WINED3DSHADER_VERSION_MAJOR(shader_version
) >= 3) sprintf(register_name
, "OUT[%u]", register_idx
);
1201 else sprintf(register_name
, "gl_TexCoord[%u]", register_idx
);
1203 case WINED3DSPR_MISCTYPE
:
1204 if (register_idx
== 0)
1207 sprintf(register_name
, "vpos");
1209 else if (register_idx
== 1)
1211 /* Note that gl_FrontFacing is a bool, while vFace is
1212 * a float for which the sign determines front/back
1214 sprintf(register_name
, "(gl_FrontFacing ? 1.0 : -1.0)");
1216 FIXME("Unhandled misctype register %d\n", register_idx
);
1217 sprintf(register_name
, "unrecognized_register");
1221 FIXME("Unhandled register name Type(%d)\n", register_type
);
1222 sprintf(register_name
, "unrecognized_register");
1227 static void shader_glsl_write_mask_to_str(DWORD write_mask
, char *str
)
1230 if (write_mask
& WINED3DSP_WRITEMASK_0
) *str
++ = 'x';
1231 if (write_mask
& WINED3DSP_WRITEMASK_1
) *str
++ = 'y';
1232 if (write_mask
& WINED3DSP_WRITEMASK_2
) *str
++ = 'z';
1233 if (write_mask
& WINED3DSP_WRITEMASK_3
) *str
++ = 'w';
1237 /* Get the GLSL write mask for the destination register */
1238 static DWORD
shader_glsl_get_write_mask(const struct wined3d_shader_dst_param
*param
, char *write_mask
)
1240 DWORD mask
= param
->write_mask
;
1242 if (shader_is_scalar(param
->register_type
, param
->register_idx
))
1244 mask
= WINED3DSP_WRITEMASK_0
;
1249 shader_glsl_write_mask_to_str(mask
, write_mask
);
1255 static unsigned int shader_glsl_get_write_mask_size(DWORD write_mask
) {
1256 unsigned int size
= 0;
1258 if (write_mask
& WINED3DSP_WRITEMASK_0
) ++size
;
1259 if (write_mask
& WINED3DSP_WRITEMASK_1
) ++size
;
1260 if (write_mask
& WINED3DSP_WRITEMASK_2
) ++size
;
1261 if (write_mask
& WINED3DSP_WRITEMASK_3
) ++size
;
1266 static void shader_glsl_swizzle_to_str(const DWORD swizzle
, BOOL fixup
, DWORD mask
, char *str
)
1268 /* For registers of type WINED3DDECLTYPE_D3DCOLOR, data is stored as "bgra",
1269 * but addressed as "rgba". To fix this we need to swap the register's x
1270 * and z components. */
1271 const char *swizzle_chars
= fixup
? "zyxw" : "xyzw";
1274 /* swizzle bits fields: wwzzyyxx */
1275 if (mask
& WINED3DSP_WRITEMASK_0
) *str
++ = swizzle_chars
[swizzle
& 0x03];
1276 if (mask
& WINED3DSP_WRITEMASK_1
) *str
++ = swizzle_chars
[(swizzle
>> 2) & 0x03];
1277 if (mask
& WINED3DSP_WRITEMASK_2
) *str
++ = swizzle_chars
[(swizzle
>> 4) & 0x03];
1278 if (mask
& WINED3DSP_WRITEMASK_3
) *str
++ = swizzle_chars
[(swizzle
>> 6) & 0x03];
1282 static void shader_glsl_get_swizzle(const struct wined3d_shader_src_param
*param
,
1283 BOOL fixup
, DWORD mask
, char *swizzle_str
)
1285 if (shader_is_scalar(param
->register_type
, param
->register_idx
))
1286 *swizzle_str
= '\0';
1288 shader_glsl_swizzle_to_str(param
->swizzle
, fixup
, mask
, swizzle_str
);
1291 /* From a given parameter token, generate the corresponding GLSL string.
1292 * Also, return the actual register name and swizzle in case the
1293 * caller needs this information as well. */
1294 static void shader_glsl_add_src_param(const struct wined3d_shader_instruction
*ins
,
1295 const struct wined3d_shader_src_param
*wined3d_src
, DWORD mask
, glsl_src_param_t
*glsl_src
)
1297 BOOL is_color
= FALSE
;
1298 char swizzle_str
[6];
1300 glsl_src
->reg_name
[0] = '\0';
1301 glsl_src
->param_str
[0] = '\0';
1302 swizzle_str
[0] = '\0';
1304 shader_glsl_get_register_name(wined3d_src
->register_type
, wined3d_src
->register_idx
,
1305 wined3d_src
->rel_addr
, glsl_src
->reg_name
, &is_color
, ins
);
1307 shader_glsl_get_swizzle(wined3d_src
, is_color
, mask
, swizzle_str
);
1308 shader_glsl_gen_modifier(wined3d_src
->modifiers
, glsl_src
->reg_name
, swizzle_str
, glsl_src
->param_str
);
1311 /* From a given parameter token, generate the corresponding GLSL string.
1312 * Also, return the actual register name and swizzle in case the
1313 * caller needs this information as well. */
1314 static DWORD
shader_glsl_add_dst_param(const struct wined3d_shader_instruction
*ins
,
1315 const struct wined3d_shader_dst_param
*wined3d_dst
, glsl_dst_param_t
*glsl_dst
)
1317 BOOL is_color
= FALSE
;
1319 glsl_dst
->mask_str
[0] = '\0';
1320 glsl_dst
->reg_name
[0] = '\0';
1322 shader_glsl_get_register_name(wined3d_dst
->register_type
, wined3d_dst
->register_idx
,
1323 wined3d_dst
->rel_addr
, glsl_dst
->reg_name
, &is_color
, ins
);
1324 return shader_glsl_get_write_mask(wined3d_dst
, glsl_dst
->mask_str
);
1327 /* Append the destination part of the instruction to the buffer, return the effective write mask */
1328 static DWORD
shader_glsl_append_dst_ext(SHADER_BUFFER
*buffer
,
1329 const struct wined3d_shader_instruction
*ins
, const struct wined3d_shader_dst_param
*dst
)
1331 glsl_dst_param_t glsl_dst
;
1334 mask
= shader_glsl_add_dst_param(ins
, dst
, &glsl_dst
);
1335 if (mask
) shader_addline(buffer
, "%s%s = %s(", glsl_dst
.reg_name
, glsl_dst
.mask_str
, shift_glsl_tab
[dst
->shift
]);
1340 /* Append the destination part of the instruction to the buffer, return the effective write mask */
1341 static DWORD
shader_glsl_append_dst(SHADER_BUFFER
*buffer
, const struct wined3d_shader_instruction
*ins
)
1343 return shader_glsl_append_dst_ext(buffer
, ins
, &ins
->dst
[0]);
1346 /** Process GLSL instruction modifiers */
1347 void shader_glsl_add_instruction_modifiers(const struct wined3d_shader_instruction
*ins
)
1349 glsl_dst_param_t dst_param
;
1352 if (!ins
->dst_count
) return;
1354 modifiers
= ins
->dst
[0].modifiers
;
1355 if (!modifiers
) return;
1357 shader_glsl_add_dst_param(ins
, &ins
->dst
[0], &dst_param
);
1359 if (modifiers
& WINED3DSPDM_SATURATE
)
1361 /* _SAT means to clamp the value of the register to between 0 and 1 */
1362 shader_addline(ins
->ctx
->buffer
, "%s%s = clamp(%s%s, 0.0, 1.0);\n", dst_param
.reg_name
,
1363 dst_param
.mask_str
, dst_param
.reg_name
, dst_param
.mask_str
);
1366 if (modifiers
& WINED3DSPDM_MSAMPCENTROID
)
1368 FIXME("_centroid modifier not handled\n");
1371 if (modifiers
& WINED3DSPDM_PARTIALPRECISION
)
1373 /* MSDN says this modifier can be safely ignored, so that's what we'll do. */
1377 static inline const char *shader_get_comp_op(DWORD op
)
1380 case COMPARISON_GT
: return ">";
1381 case COMPARISON_EQ
: return "==";
1382 case COMPARISON_GE
: return ">=";
1383 case COMPARISON_LT
: return "<";
1384 case COMPARISON_NE
: return "!=";
1385 case COMPARISON_LE
: return "<=";
1387 FIXME("Unrecognized comparison value: %u\n", op
);
1392 static void shader_glsl_get_sample_function(DWORD sampler_type
, DWORD flags
, glsl_sample_function_t
*sample_function
)
1394 BOOL projected
= flags
& WINED3D_GLSL_SAMPLE_PROJECTED
;
1395 BOOL texrect
= flags
& WINED3D_GLSL_SAMPLE_RECT
;
1396 BOOL lod
= flags
& WINED3D_GLSL_SAMPLE_LOD
;
1398 /* Note that there's no such thing as a projected cube texture. */
1399 switch(sampler_type
) {
1402 sample_function
->name
= projected
? "texture1DProjLod" : "texture1DLod";
1404 sample_function
->name
= projected
? "texture1DProj" : "texture1D";
1406 sample_function
->coord_mask
= WINED3DSP_WRITEMASK_0
;
1411 sample_function
->name
= projected
? "texture2DRectProjLod" : "texture2DRectLod";
1413 sample_function
->name
= projected
? "texture2DRectProj" : "texture2DRect";
1417 sample_function
->name
= projected
? "texture2DProjLod" : "texture2DLod";
1419 sample_function
->name
= projected
? "texture2DProj" : "texture2D";
1422 sample_function
->coord_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
;
1424 case WINED3DSTT_CUBE
:
1426 sample_function
->name
= "textureCubeLod";
1428 sample_function
->name
= "textureCube";
1430 sample_function
->coord_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
1432 case WINED3DSTT_VOLUME
:
1434 sample_function
->name
= projected
? "texture3DProjLod" : "texture3DLod";
1436 sample_function
->name
= projected
? "texture3DProj" : "texture3D";
1438 sample_function
->coord_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
1441 sample_function
->name
= "";
1442 sample_function
->coord_mask
= 0;
1443 FIXME("Unrecognized sampler type: %#x;\n", sampler_type
);
1448 static void shader_glsl_append_fixup_arg(char *arguments
, const char *reg_name
,
1449 BOOL sign_fixup
, enum fixup_channel_source channel_source
)
1451 switch(channel_source
)
1453 case CHANNEL_SOURCE_ZERO
:
1454 strcat(arguments
, "0.0");
1457 case CHANNEL_SOURCE_ONE
:
1458 strcat(arguments
, "1.0");
1461 case CHANNEL_SOURCE_X
:
1462 strcat(arguments
, reg_name
);
1463 strcat(arguments
, ".x");
1466 case CHANNEL_SOURCE_Y
:
1467 strcat(arguments
, reg_name
);
1468 strcat(arguments
, ".y");
1471 case CHANNEL_SOURCE_Z
:
1472 strcat(arguments
, reg_name
);
1473 strcat(arguments
, ".z");
1476 case CHANNEL_SOURCE_W
:
1477 strcat(arguments
, reg_name
);
1478 strcat(arguments
, ".w");
1482 FIXME("Unhandled channel source %#x\n", channel_source
);
1483 strcat(arguments
, "undefined");
1487 if (sign_fixup
) strcat(arguments
, " * 2.0 - 1.0");
1490 static void shader_glsl_color_correction(const struct wined3d_shader_instruction
*ins
, struct color_fixup_desc fixup
)
1492 struct wined3d_shader_dst_param dst
;
1493 unsigned int mask_size
, remaining
;
1494 glsl_dst_param_t dst_param
;
1495 char arguments
[256];
1499 if (fixup
.x_sign_fixup
|| fixup
.x_source
!= CHANNEL_SOURCE_X
) mask
|= WINED3DSP_WRITEMASK_0
;
1500 if (fixup
.y_sign_fixup
|| fixup
.y_source
!= CHANNEL_SOURCE_Y
) mask
|= WINED3DSP_WRITEMASK_1
;
1501 if (fixup
.z_sign_fixup
|| fixup
.z_source
!= CHANNEL_SOURCE_Z
) mask
|= WINED3DSP_WRITEMASK_2
;
1502 if (fixup
.w_sign_fixup
|| fixup
.w_source
!= CHANNEL_SOURCE_W
) mask
|= WINED3DSP_WRITEMASK_3
;
1503 mask
&= ins
->dst
[0].write_mask
;
1505 if (!mask
) return; /* Nothing to do */
1507 if (is_yuv_fixup(fixup
))
1509 enum yuv_fixup yuv_fixup
= get_yuv_fixup(fixup
);
1510 FIXME("YUV fixup (%#x) not supported\n", yuv_fixup
);
1514 mask_size
= shader_glsl_get_write_mask_size(mask
);
1517 dst
.write_mask
= mask
;
1518 shader_glsl_add_dst_param(ins
, &dst
, &dst_param
);
1520 arguments
[0] = '\0';
1521 remaining
= mask_size
;
1522 if (mask
& WINED3DSP_WRITEMASK_0
)
1524 shader_glsl_append_fixup_arg(arguments
, dst_param
.reg_name
, fixup
.x_sign_fixup
, fixup
.x_source
);
1525 if (--remaining
) strcat(arguments
, ", ");
1527 if (mask
& WINED3DSP_WRITEMASK_1
)
1529 shader_glsl_append_fixup_arg(arguments
, dst_param
.reg_name
, fixup
.y_sign_fixup
, fixup
.y_source
);
1530 if (--remaining
) strcat(arguments
, ", ");
1532 if (mask
& WINED3DSP_WRITEMASK_2
)
1534 shader_glsl_append_fixup_arg(arguments
, dst_param
.reg_name
, fixup
.z_sign_fixup
, fixup
.z_source
);
1535 if (--remaining
) strcat(arguments
, ", ");
1537 if (mask
& WINED3DSP_WRITEMASK_3
)
1539 shader_glsl_append_fixup_arg(arguments
, dst_param
.reg_name
, fixup
.w_sign_fixup
, fixup
.w_source
);
1540 if (--remaining
) strcat(arguments
, ", ");
1545 shader_addline(ins
->ctx
->buffer
, "%s%s = vec%u(%s);\n",
1546 dst_param
.reg_name
, dst_param
.mask_str
, mask_size
, arguments
);
1550 shader_addline(ins
->ctx
->buffer
, "%s%s = %s;\n", dst_param
.reg_name
, dst_param
.mask_str
, arguments
);
1554 static void PRINTF_ATTR(6, 7) shader_glsl_gen_sample_code(const struct wined3d_shader_instruction
*ins
,
1555 DWORD sampler
, const glsl_sample_function_t
*sample_function
, DWORD swizzle
,
1556 const char *bias
, const char *coord_reg_fmt
, ...)
1558 const char *sampler_base
;
1559 char dst_swizzle
[6];
1560 struct color_fixup_desc fixup
;
1561 BOOL np2_fixup
= FALSE
;
1564 shader_glsl_swizzle_to_str(swizzle
, FALSE
, ins
->dst
[0].write_mask
, dst_swizzle
);
1566 if (shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
))
1568 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*)ins
->ctx
->shader
;
1569 fixup
= This
->cur_args
->color_fixup
[sampler
];
1570 sampler_base
= "Psampler";
1572 if(This
->cur_args
->np2_fixup
& (1 << sampler
)) {
1574 FIXME("Biased sampling from NP2 textures is unsupported\n");
1580 sampler_base
= "Vsampler";
1581 fixup
= COLOR_FIXUP_IDENTITY
; /* FIXME: Vshader color fixup */
1584 shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
1586 shader_addline(ins
->ctx
->buffer
, "%s(%s%u, ", sample_function
->name
, sampler_base
, sampler
);
1588 va_start(args
, coord_reg_fmt
);
1589 shader_vaddline(ins
->ctx
->buffer
, coord_reg_fmt
, args
);
1593 shader_addline(ins
->ctx
->buffer
, ", %s)%s);\n", bias
, dst_swizzle
);
1596 shader_addline(ins
->ctx
->buffer
, " * PsamplerNP2Fixup%u)%s);\n", sampler
, dst_swizzle
);
1598 shader_addline(ins
->ctx
->buffer
, ")%s);\n", dst_swizzle
);
1602 if(!is_identity_fixup(fixup
)) {
1603 shader_glsl_color_correction(ins
, fixup
);
1607 /*****************************************************************************
1609 * Begin processing individual instruction opcodes
1611 ****************************************************************************/
1613 /* Generate GLSL arithmetic functions (dst = src1 + src2) */
1614 static void shader_glsl_arith(const struct wined3d_shader_instruction
*ins
)
1616 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
1617 glsl_src_param_t src0_param
;
1618 glsl_src_param_t src1_param
;
1622 /* Determine the GLSL operator to use based on the opcode */
1623 switch (ins
->handler_idx
)
1625 case WINED3DSIH_MUL
: op
= '*'; break;
1626 case WINED3DSIH_ADD
: op
= '+'; break;
1627 case WINED3DSIH_SUB
: op
= '-'; break;
1630 FIXME("Opcode %#x not yet handled in GLSL\n", ins
->handler_idx
);
1634 write_mask
= shader_glsl_append_dst(buffer
, ins
);
1635 shader_glsl_add_src_param(ins
, &ins
->src
[0], write_mask
, &src0_param
);
1636 shader_glsl_add_src_param(ins
, &ins
->src
[1], write_mask
, &src1_param
);
1637 shader_addline(buffer
, "%s %c %s);\n", src0_param
.param_str
, op
, src1_param
.param_str
);
1640 /* Process the WINED3DSIO_MOV opcode using GLSL (dst = src) */
1641 static void shader_glsl_mov(const struct wined3d_shader_instruction
*ins
)
1643 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
1644 glsl_src_param_t src0_param
;
1647 write_mask
= shader_glsl_append_dst(buffer
, ins
);
1648 shader_glsl_add_src_param(ins
, &ins
->src
[0], write_mask
, &src0_param
);
1650 /* In vs_1_1 WINED3DSIO_MOV can write to the address register. In later
1651 * shader versions WINED3DSIO_MOVA is used for this. */
1652 if ((WINED3DSHADER_VERSION_MAJOR(ins
->ctx
->reg_maps
->shader_version
) == 1
1653 && !shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
)
1654 && ins
->dst
[0].register_type
== WINED3DSPR_ADDR
))
1656 /* This is a simple floor() */
1657 unsigned int mask_size
= shader_glsl_get_write_mask_size(write_mask
);
1658 if (mask_size
> 1) {
1659 shader_addline(buffer
, "ivec%d(floor(%s)));\n", mask_size
, src0_param
.param_str
);
1661 shader_addline(buffer
, "int(floor(%s)));\n", src0_param
.param_str
);
1664 else if(ins
->handler_idx
== WINED3DSIH_MOVA
)
1666 /* We need to *round* to the nearest int here. */
1667 unsigned int mask_size
= shader_glsl_get_write_mask_size(write_mask
);
1668 if (mask_size
> 1) {
1669 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
);
1671 shader_addline(buffer
, "int(floor(abs(%s) + 0.5) * sign(%s)));\n", src0_param
.param_str
, src0_param
.param_str
);
1674 shader_addline(buffer
, "%s);\n", src0_param
.param_str
);
1678 /* Process the dot product operators DP3 and DP4 in GLSL (dst = dot(src0, src1)) */
1679 static void shader_glsl_dot(const struct wined3d_shader_instruction
*ins
)
1681 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
1682 glsl_src_param_t src0_param
;
1683 glsl_src_param_t src1_param
;
1684 DWORD dst_write_mask
, src_write_mask
;
1685 unsigned int dst_size
= 0;
1687 dst_write_mask
= shader_glsl_append_dst(buffer
, ins
);
1688 dst_size
= shader_glsl_get_write_mask_size(dst_write_mask
);
1690 /* dp3 works on vec3, dp4 on vec4 */
1691 if (ins
->handler_idx
== WINED3DSIH_DP4
)
1693 src_write_mask
= WINED3DSP_WRITEMASK_ALL
;
1695 src_write_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
1698 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_write_mask
, &src0_param
);
1699 shader_glsl_add_src_param(ins
, &ins
->src
[1], src_write_mask
, &src1_param
);
1702 shader_addline(buffer
, "vec%d(dot(%s, %s)));\n", dst_size
, src0_param
.param_str
, src1_param
.param_str
);
1704 shader_addline(buffer
, "dot(%s, %s));\n", src0_param
.param_str
, src1_param
.param_str
);
1708 /* Note that this instruction has some restrictions. The destination write mask
1709 * can't contain the w component, and the source swizzles have to be .xyzw */
1710 static void shader_glsl_cross(const struct wined3d_shader_instruction
*ins
)
1712 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
1713 glsl_src_param_t src0_param
;
1714 glsl_src_param_t src1_param
;
1717 shader_glsl_get_write_mask(&ins
->dst
[0], dst_mask
);
1718 shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
1719 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
1720 shader_glsl_add_src_param(ins
, &ins
->src
[1], src_mask
, &src1_param
);
1721 shader_addline(ins
->ctx
->buffer
, "cross(%s, %s)%s);\n", src0_param
.param_str
, src1_param
.param_str
, dst_mask
);
1724 /* Process the WINED3DSIO_POW instruction in GLSL (dst = |src0|^src1)
1725 * Src0 and src1 are scalars. Note that D3D uses the absolute of src0, while
1726 * GLSL uses the value as-is. */
1727 static void shader_glsl_pow(const struct wined3d_shader_instruction
*ins
)
1729 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
1730 glsl_src_param_t src0_param
;
1731 glsl_src_param_t src1_param
;
1732 DWORD dst_write_mask
;
1733 unsigned int dst_size
;
1735 dst_write_mask
= shader_glsl_append_dst(buffer
, ins
);
1736 dst_size
= shader_glsl_get_write_mask_size(dst_write_mask
);
1738 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
1739 shader_glsl_add_src_param(ins
, &ins
->src
[1], WINED3DSP_WRITEMASK_0
, &src1_param
);
1742 shader_addline(buffer
, "vec%d(pow(abs(%s), %s)));\n", dst_size
, src0_param
.param_str
, src1_param
.param_str
);
1744 shader_addline(buffer
, "pow(abs(%s), %s));\n", src0_param
.param_str
, src1_param
.param_str
);
1748 /* Process the WINED3DSIO_LOG instruction in GLSL (dst = log2(|src0|))
1749 * Src0 is a scalar. Note that D3D uses the absolute of src0, while
1750 * GLSL uses the value as-is. */
1751 static void shader_glsl_log(const struct wined3d_shader_instruction
*ins
)
1753 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
1754 glsl_src_param_t src0_param
;
1755 DWORD dst_write_mask
;
1756 unsigned int dst_size
;
1758 dst_write_mask
= shader_glsl_append_dst(buffer
, ins
);
1759 dst_size
= shader_glsl_get_write_mask_size(dst_write_mask
);
1761 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
1764 shader_addline(buffer
, "vec%d(log2(abs(%s))));\n", dst_size
, src0_param
.param_str
);
1766 shader_addline(buffer
, "log2(abs(%s)));\n", src0_param
.param_str
);
1770 /* Map the opcode 1-to-1 to the GL code (arg->dst = instruction(src0, src1, ...) */
1771 static void shader_glsl_map2gl(const struct wined3d_shader_instruction
*ins
)
1773 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
1774 glsl_src_param_t src_param
;
1775 const char *instruction
;
1779 /* Determine the GLSL function to use based on the opcode */
1780 /* TODO: Possibly make this a table for faster lookups */
1781 switch (ins
->handler_idx
)
1783 case WINED3DSIH_MIN
: instruction
= "min"; break;
1784 case WINED3DSIH_MAX
: instruction
= "max"; break;
1785 case WINED3DSIH_ABS
: instruction
= "abs"; break;
1786 case WINED3DSIH_FRC
: instruction
= "fract"; break;
1787 case WINED3DSIH_NRM
: instruction
= "normalize"; break;
1788 case WINED3DSIH_EXP
: instruction
= "exp2"; break;
1789 case WINED3DSIH_SGN
: instruction
= "sign"; break;
1790 case WINED3DSIH_DSX
: instruction
= "dFdx"; break;
1791 case WINED3DSIH_DSY
: instruction
= "ycorrection.y * dFdy"; break;
1792 default: instruction
= "";
1793 FIXME("Opcode %#x not yet handled in GLSL\n", ins
->handler_idx
);
1797 write_mask
= shader_glsl_append_dst(buffer
, ins
);
1799 shader_addline(buffer
, "%s(", instruction
);
1803 shader_glsl_add_src_param(ins
, &ins
->src
[0], write_mask
, &src_param
);
1804 shader_addline(buffer
, "%s", src_param
.param_str
);
1805 for (i
= 1; i
< ins
->src_count
; ++i
)
1807 shader_glsl_add_src_param(ins
, &ins
->src
[i
], write_mask
, &src_param
);
1808 shader_addline(buffer
, ", %s", src_param
.param_str
);
1812 shader_addline(buffer
, "));\n");
1815 /** Process the WINED3DSIO_EXPP instruction in GLSL:
1816 * For shader model 1.x, do the following (and honor the writemask, so use a temporary variable):
1817 * dst.x = 2^(floor(src))
1818 * dst.y = src - floor(src)
1819 * dst.z = 2^src (partial precision is allowed, but optional)
1821 * For 2.0 shaders, just do this (honoring writemask and swizzle):
1822 * dst = 2^src; (partial precision is allowed, but optional)
1824 static void shader_glsl_expp(const struct wined3d_shader_instruction
*ins
)
1826 glsl_src_param_t src_param
;
1828 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
, &src_param
);
1830 if (ins
->ctx
->reg_maps
->shader_version
< WINED3DPS_VERSION(2,0))
1834 shader_addline(ins
->ctx
->buffer
, "tmp0.x = exp2(floor(%s));\n", src_param
.param_str
);
1835 shader_addline(ins
->ctx
->buffer
, "tmp0.y = %s - floor(%s);\n", src_param
.param_str
, src_param
.param_str
);
1836 shader_addline(ins
->ctx
->buffer
, "tmp0.z = exp2(%s);\n", src_param
.param_str
);
1837 shader_addline(ins
->ctx
->buffer
, "tmp0.w = 1.0;\n");
1839 shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
1840 shader_glsl_get_write_mask(&ins
->dst
[0], dst_mask
);
1841 shader_addline(ins
->ctx
->buffer
, "tmp0%s);\n", dst_mask
);
1844 unsigned int mask_size
;
1846 write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
1847 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
1849 if (mask_size
> 1) {
1850 shader_addline(ins
->ctx
->buffer
, "vec%d(exp2(%s)));\n", mask_size
, src_param
.param_str
);
1852 shader_addline(ins
->ctx
->buffer
, "exp2(%s));\n", src_param
.param_str
);
1857 /** Process the RCP (reciprocal or inverse) opcode in GLSL (dst = 1 / src) */
1858 static void shader_glsl_rcp(const struct wined3d_shader_instruction
*ins
)
1860 glsl_src_param_t src_param
;
1862 unsigned int mask_size
;
1864 write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
1865 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
1866 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_3
, &src_param
);
1868 if (mask_size
> 1) {
1869 shader_addline(ins
->ctx
->buffer
, "vec%d(1.0 / %s));\n", mask_size
, src_param
.param_str
);
1871 shader_addline(ins
->ctx
->buffer
, "1.0 / %s);\n", src_param
.param_str
);
1875 static void shader_glsl_rsq(const struct wined3d_shader_instruction
*ins
)
1877 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
1878 glsl_src_param_t src_param
;
1880 unsigned int mask_size
;
1882 write_mask
= shader_glsl_append_dst(buffer
, ins
);
1883 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
1885 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_3
, &src_param
);
1887 if (mask_size
> 1) {
1888 shader_addline(buffer
, "vec%d(inversesqrt(%s)));\n", mask_size
, src_param
.param_str
);
1890 shader_addline(buffer
, "inversesqrt(%s));\n", src_param
.param_str
);
1894 /** Process signed comparison opcodes in GLSL. */
1895 static void shader_glsl_compare(const struct wined3d_shader_instruction
*ins
)
1897 glsl_src_param_t src0_param
;
1898 glsl_src_param_t src1_param
;
1900 unsigned int mask_size
;
1902 write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
1903 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
1904 shader_glsl_add_src_param(ins
, &ins
->src
[0], write_mask
, &src0_param
);
1905 shader_glsl_add_src_param(ins
, &ins
->src
[1], write_mask
, &src1_param
);
1907 if (mask_size
> 1) {
1908 const char *compare
;
1910 switch(ins
->handler_idx
)
1912 case WINED3DSIH_SLT
: compare
= "lessThan"; break;
1913 case WINED3DSIH_SGE
: compare
= "greaterThanEqual"; break;
1914 default: compare
= "";
1915 FIXME("Can't handle opcode %#x\n", ins
->handler_idx
);
1918 shader_addline(ins
->ctx
->buffer
, "vec%d(%s(%s, %s)));\n", mask_size
, compare
,
1919 src0_param
.param_str
, src1_param
.param_str
);
1921 switch(ins
->handler_idx
)
1923 case WINED3DSIH_SLT
:
1924 /* Step(src0, src1) is not suitable here because if src0 == src1 SLT is supposed,
1925 * to return 0.0 but step returns 1.0 because step is not < x
1926 * An alternative is a bvec compare padded with an unused second component.
1927 * step(src1 * -1.0, src0 * -1.0) is not an option because it suffers from the same
1928 * issue. Playing with not() is not possible either because not() does not accept
1931 shader_addline(ins
->ctx
->buffer
, "(%s < %s) ? 1.0 : 0.0);\n",
1932 src0_param
.param_str
, src1_param
.param_str
);
1934 case WINED3DSIH_SGE
:
1935 /* Here we can use the step() function and safe a conditional */
1936 shader_addline(ins
->ctx
->buffer
, "step(%s, %s));\n", src1_param
.param_str
, src0_param
.param_str
);
1939 FIXME("Can't handle opcode %#x\n", ins
->handler_idx
);
1945 /** Process CMP instruction in GLSL (dst = src0 >= 0.0 ? src1 : src2), per channel */
1946 static void shader_glsl_cmp(const struct wined3d_shader_instruction
*ins
)
1948 glsl_src_param_t src0_param
;
1949 glsl_src_param_t src1_param
;
1950 glsl_src_param_t src2_param
;
1951 DWORD write_mask
, cmp_channel
= 0;
1954 BOOL temp_destination
= FALSE
;
1956 if (shader_is_scalar(ins
->src
[0].register_type
, ins
->src
[0].register_idx
))
1958 write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
1960 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_ALL
, &src0_param
);
1961 shader_glsl_add_src_param(ins
, &ins
->src
[1], write_mask
, &src1_param
);
1962 shader_glsl_add_src_param(ins
, &ins
->src
[2], write_mask
, &src2_param
);
1964 shader_addline(ins
->ctx
->buffer
, "%s >= 0.0 ? %s : %s);\n",
1965 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
1967 DWORD dst_mask
= ins
->dst
[0].write_mask
;
1968 struct wined3d_shader_dst_param dst
= ins
->dst
[0];
1970 /* Cycle through all source0 channels */
1971 for (i
=0; i
<4; i
++) {
1973 /* Find the destination channels which use the current source0 channel */
1974 for (j
=0; j
<4; j
++) {
1975 if (((ins
->src
[0].swizzle
>> (2 * j
)) & 0x3) == i
)
1977 write_mask
|= WINED3DSP_WRITEMASK_0
<< j
;
1978 cmp_channel
= WINED3DSP_WRITEMASK_0
<< j
;
1981 dst
.write_mask
= dst_mask
& write_mask
;
1983 /* Splitting the cmp instruction up in multiple lines imposes a problem:
1984 * The first lines may overwrite source parameters of the following lines.
1985 * Deal with that by using a temporary destination register if needed
1987 if ((ins
->src
[0].register_idx
== ins
->dst
[0].register_idx
1988 && ins
->src
[0].register_type
== ins
->dst
[0].register_type
)
1989 || (ins
->src
[1].register_idx
== ins
->dst
[0].register_idx
1990 && ins
->src
[1].register_type
== ins
->dst
[0].register_type
)
1991 || (ins
->src
[2].register_idx
== ins
->dst
[0].register_idx
1992 && ins
->src
[2].register_type
== ins
->dst
[0].register_type
))
1994 write_mask
= shader_glsl_get_write_mask(&dst
, mask_char
);
1995 if (!write_mask
) continue;
1996 shader_addline(ins
->ctx
->buffer
, "tmp0%s = (", mask_char
);
1997 temp_destination
= TRUE
;
1999 write_mask
= shader_glsl_append_dst_ext(ins
->ctx
->buffer
, ins
, &dst
);
2000 if (!write_mask
) continue;
2003 shader_glsl_add_src_param(ins
, &ins
->src
[0], cmp_channel
, &src0_param
);
2004 shader_glsl_add_src_param(ins
, &ins
->src
[1], write_mask
, &src1_param
);
2005 shader_glsl_add_src_param(ins
, &ins
->src
[2], write_mask
, &src2_param
);
2007 shader_addline(ins
->ctx
->buffer
, "%s >= 0.0 ? %s : %s);\n",
2008 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
2011 if(temp_destination
) {
2012 shader_glsl_get_write_mask(&ins
->dst
[0], mask_char
);
2013 shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2014 shader_addline(ins
->ctx
->buffer
, "tmp0%s);\n", mask_char
);
2020 /** Process the CND opcode in GLSL (dst = (src0 > 0.5) ? src1 : src2) */
2021 /* For ps 1.1-1.3, only a single component of src0 is used. For ps 1.4
2022 * the compare is done per component of src0. */
2023 static void shader_glsl_cnd(const struct wined3d_shader_instruction
*ins
)
2025 struct wined3d_shader_dst_param dst
;
2026 glsl_src_param_t src0_param
;
2027 glsl_src_param_t src1_param
;
2028 glsl_src_param_t src2_param
;
2029 DWORD write_mask
, cmp_channel
= 0;
2033 if (ins
->ctx
->reg_maps
->shader_version
< WINED3DPS_VERSION(1, 4))
2035 write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2036 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2037 shader_glsl_add_src_param(ins
, &ins
->src
[1], write_mask
, &src1_param
);
2038 shader_glsl_add_src_param(ins
, &ins
->src
[2], write_mask
, &src2_param
);
2040 /* Fun: The D3DSI_COISSUE flag changes the semantic of the cnd instruction for < 1.4 shaders */
2043 shader_addline(ins
->ctx
->buffer
, "%s /* COISSUE! */);\n", src1_param
.param_str
);
2045 shader_addline(ins
->ctx
->buffer
, "%s > 0.5 ? %s : %s);\n",
2046 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
2050 /* Cycle through all source0 channels */
2051 dst_mask
= ins
->dst
[0].write_mask
;
2053 for (i
=0; i
<4; i
++) {
2055 /* Find the destination channels which use the current source0 channel */
2056 for (j
=0; j
<4; j
++) {
2057 if (((ins
->src
[0].swizzle
>> (2 * j
)) & 0x3) == i
)
2059 write_mask
|= WINED3DSP_WRITEMASK_0
<< j
;
2060 cmp_channel
= WINED3DSP_WRITEMASK_0
<< j
;
2064 dst
.write_mask
= dst_mask
& write_mask
;
2065 write_mask
= shader_glsl_append_dst_ext(ins
->ctx
->buffer
, ins
, &dst
);
2066 if (!write_mask
) continue;
2068 shader_glsl_add_src_param(ins
, &ins
->src
[0], cmp_channel
, &src0_param
);
2069 shader_glsl_add_src_param(ins
, &ins
->src
[1], write_mask
, &src1_param
);
2070 shader_glsl_add_src_param(ins
, &ins
->src
[2], write_mask
, &src2_param
);
2072 shader_addline(ins
->ctx
->buffer
, "%s > 0.5 ? %s : %s);\n",
2073 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
2077 /** GLSL code generation for WINED3DSIO_MAD: Multiply the first 2 opcodes, then add the last */
2078 static void shader_glsl_mad(const struct wined3d_shader_instruction
*ins
)
2080 glsl_src_param_t src0_param
;
2081 glsl_src_param_t src1_param
;
2082 glsl_src_param_t src2_param
;
2085 write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2086 shader_glsl_add_src_param(ins
, &ins
->src
[0], write_mask
, &src0_param
);
2087 shader_glsl_add_src_param(ins
, &ins
->src
[1], write_mask
, &src1_param
);
2088 shader_glsl_add_src_param(ins
, &ins
->src
[2], write_mask
, &src2_param
);
2089 shader_addline(ins
->ctx
->buffer
, "(%s * %s) + %s);\n",
2090 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
2093 /** Handles transforming all WINED3DSIO_M?x? opcodes for
2094 Vertex shaders to GLSL codes */
2095 static void shader_glsl_mnxn(const struct wined3d_shader_instruction
*ins
)
2098 int nComponents
= 0;
2099 struct wined3d_shader_dst_param tmp_dst
= {0};
2100 struct wined3d_shader_src_param tmp_src
[2] = {{0}};
2101 struct wined3d_shader_instruction tmp_ins
;
2103 memset(&tmp_ins
, 0, sizeof(tmp_ins
));
2105 /* Set constants for the temporary argument */
2106 tmp_ins
.ctx
= ins
->ctx
;
2107 tmp_ins
.dst_count
= 1;
2108 tmp_ins
.dst
= &tmp_dst
;
2109 tmp_ins
.src_count
= 2;
2110 tmp_ins
.src
= tmp_src
;
2112 switch(ins
->handler_idx
)
2114 case WINED3DSIH_M4x4
:
2116 tmp_ins
.handler_idx
= WINED3DSIH_DP4
;
2118 case WINED3DSIH_M4x3
:
2120 tmp_ins
.handler_idx
= WINED3DSIH_DP4
;
2122 case WINED3DSIH_M3x4
:
2124 tmp_ins
.handler_idx
= WINED3DSIH_DP3
;
2126 case WINED3DSIH_M3x3
:
2128 tmp_ins
.handler_idx
= WINED3DSIH_DP3
;
2130 case WINED3DSIH_M3x2
:
2132 tmp_ins
.handler_idx
= WINED3DSIH_DP3
;
2138 tmp_dst
= ins
->dst
[0];
2139 tmp_src
[0] = ins
->src
[0];
2140 tmp_src
[1] = ins
->src
[1];
2141 for (i
= 0; i
< nComponents
; ++i
)
2143 tmp_dst
.write_mask
= WINED3DSP_WRITEMASK_0
<< i
;
2144 shader_glsl_dot(&tmp_ins
);
2145 ++tmp_src
[1].register_idx
;
2150 The LRP instruction performs a component-wise linear interpolation
2151 between the second and third operands using the first operand as the
2152 blend factor. Equation: (dst = src2 + src0 * (src1 - src2))
2153 This is equivalent to mix(src2, src1, src0);
2155 static void shader_glsl_lrp(const struct wined3d_shader_instruction
*ins
)
2157 glsl_src_param_t src0_param
;
2158 glsl_src_param_t src1_param
;
2159 glsl_src_param_t src2_param
;
2162 write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2164 shader_glsl_add_src_param(ins
, &ins
->src
[0], write_mask
, &src0_param
);
2165 shader_glsl_add_src_param(ins
, &ins
->src
[1], write_mask
, &src1_param
);
2166 shader_glsl_add_src_param(ins
, &ins
->src
[2], write_mask
, &src2_param
);
2168 shader_addline(ins
->ctx
->buffer
, "mix(%s, %s, %s));\n",
2169 src2_param
.param_str
, src1_param
.param_str
, src0_param
.param_str
);
2172 /** Process the WINED3DSIO_LIT instruction in GLSL:
2173 * dst.x = dst.w = 1.0
2174 * dst.y = (src0.x > 0) ? src0.x
2175 * dst.z = (src0.x > 0) ? ((src0.y > 0) ? pow(src0.y, src.w) : 0) : 0
2176 * where src.w is clamped at +- 128
2178 static void shader_glsl_lit(const struct wined3d_shader_instruction
*ins
)
2180 glsl_src_param_t src0_param
;
2181 glsl_src_param_t src1_param
;
2182 glsl_src_param_t src3_param
;
2185 shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2186 shader_glsl_get_write_mask(&ins
->dst
[0], dst_mask
);
2188 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2189 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_1
, &src1_param
);
2190 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_3
, &src3_param
);
2192 /* The sdk specifies the instruction like this
2194 * if(src.x > 0.0) dst.y = src.x
2196 * if(src.x > 0.0 && src.y > 0.0) dst.z = pow(src.y, power);
2200 * Obviously that has quite a few conditionals in it which we don't like. So the first step is this:
2201 * dst.x = 1.0 ... No further explanation needed
2202 * dst.y = max(src.y, 0.0); ... If x < 0.0, use 0.0, otherwise x. Same as the conditional
2203 * dst.z = x > 0.0 ? pow(max(y, 0.0), p) : 0; ... 0 ^ power is 0, and otherwise we use y anyway
2204 * dst.w = 1.0. ... Nothing fancy.
2206 * So we still have one conditional in there. So do this:
2207 * dst.z = pow(max(0.0, src.y) * step(0.0, src.x), power);
2209 * 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),
2210 * 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.
2211 * if both x and y are > 0, we get pow(y * 1.0, power), as it is supposed to
2213 shader_addline(ins
->ctx
->buffer
,
2214 "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",
2215 src0_param
.param_str
, src1_param
.param_str
, src0_param
.param_str
, src3_param
.param_str
, dst_mask
);
2218 /** Process the WINED3DSIO_DST instruction in GLSL:
2220 * dst.y = src0.x * src0.y
2224 static void shader_glsl_dst(const struct wined3d_shader_instruction
*ins
)
2226 glsl_src_param_t src0y_param
;
2227 glsl_src_param_t src0z_param
;
2228 glsl_src_param_t src1y_param
;
2229 glsl_src_param_t src1w_param
;
2232 shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2233 shader_glsl_get_write_mask(&ins
->dst
[0], dst_mask
);
2235 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_1
, &src0y_param
);
2236 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_2
, &src0z_param
);
2237 shader_glsl_add_src_param(ins
, &ins
->src
[1], WINED3DSP_WRITEMASK_1
, &src1y_param
);
2238 shader_glsl_add_src_param(ins
, &ins
->src
[1], WINED3DSP_WRITEMASK_3
, &src1w_param
);
2240 shader_addline(ins
->ctx
->buffer
, "vec4(1.0, %s * %s, %s, %s))%s;\n",
2241 src0y_param
.param_str
, src1y_param
.param_str
, src0z_param
.param_str
, src1w_param
.param_str
, dst_mask
);
2244 /** Process the WINED3DSIO_SINCOS instruction in GLSL:
2245 * VS 2.0 requires that specific cosine and sine constants be passed to this instruction so the hardware
2246 * can handle it. But, these functions are built-in for GLSL, so we can just ignore the last 2 params.
2248 * dst.x = cos(src0.?)
2249 * dst.y = sin(src0.?)
2253 static void shader_glsl_sincos(const struct wined3d_shader_instruction
*ins
)
2255 glsl_src_param_t src0_param
;
2258 write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2259 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2261 switch (write_mask
) {
2262 case WINED3DSP_WRITEMASK_0
:
2263 shader_addline(ins
->ctx
->buffer
, "cos(%s));\n", src0_param
.param_str
);
2266 case WINED3DSP_WRITEMASK_1
:
2267 shader_addline(ins
->ctx
->buffer
, "sin(%s));\n", src0_param
.param_str
);
2270 case (WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
):
2271 shader_addline(ins
->ctx
->buffer
, "vec2(cos(%s), sin(%s)));\n", src0_param
.param_str
, src0_param
.param_str
);
2275 ERR("Write mask should be .x, .y or .xy\n");
2280 /** Process the WINED3DSIO_LOOP instruction in GLSL:
2281 * Start a for() loop where src1.y is the initial value of aL,
2282 * increment aL by src1.z for a total of src1.x iterations.
2283 * Need to use a temporary variable for this operation.
2285 /* FIXME: I don't think nested loops will work correctly this way. */
2286 static void shader_glsl_loop(const struct wined3d_shader_instruction
*ins
)
2288 glsl_src_param_t src1_param
;
2289 IWineD3DBaseShaderImpl
*shader
= (IWineD3DBaseShaderImpl
*)ins
->ctx
->shader
;
2290 const DWORD
*control_values
= NULL
;
2291 const local_constant
*constant
;
2293 shader_glsl_add_src_param(ins
, &ins
->src
[1], WINED3DSP_WRITEMASK_ALL
, &src1_param
);
2295 /* Try to hardcode the loop control parameters if possible. Direct3D 9 class hardware doesn't support real
2296 * varying indexing, but Microsoft designed this feature for Shader model 2.x+. If the loop control is
2297 * known at compile time, the GLSL compiler can unroll the loop, and replace indirect addressing with direct
2300 if (ins
->src
[1].register_type
== WINED3DSPR_CONSTINT
)
2302 LIST_FOR_EACH_ENTRY(constant
, &shader
->baseShader
.constantsI
, local_constant
, entry
) {
2303 if (constant
->idx
== ins
->src
[1].register_idx
)
2305 control_values
= constant
->value
;
2311 if(control_values
) {
2312 if(control_values
[2] > 0) {
2313 shader_addline(ins
->ctx
->buffer
, "for (aL%u = %d; aL%u < (%d * %d + %d); aL%u += %d) {\n",
2314 shader
->baseShader
.cur_loop_depth
, control_values
[1],
2315 shader
->baseShader
.cur_loop_depth
, control_values
[0], control_values
[2], control_values
[1],
2316 shader
->baseShader
.cur_loop_depth
, control_values
[2]);
2317 } else if(control_values
[2] == 0) {
2318 shader_addline(ins
->ctx
->buffer
, "for (aL%u = %d, tmpInt%u = 0; tmpInt%u < %d; tmpInt%u++) {\n",
2319 shader
->baseShader
.cur_loop_depth
, control_values
[1], shader
->baseShader
.cur_loop_depth
,
2320 shader
->baseShader
.cur_loop_depth
, control_values
[0],
2321 shader
->baseShader
.cur_loop_depth
);
2323 shader_addline(ins
->ctx
->buffer
, "for (aL%u = %d; aL%u > (%d * %d + %d); aL%u += %d) {\n",
2324 shader
->baseShader
.cur_loop_depth
, control_values
[1],
2325 shader
->baseShader
.cur_loop_depth
, control_values
[0], control_values
[2], control_values
[1],
2326 shader
->baseShader
.cur_loop_depth
, control_values
[2]);
2329 shader_addline(ins
->ctx
->buffer
,
2330 "for (tmpInt%u = 0, aL%u = %s.y; tmpInt%u < %s.x; tmpInt%u++, aL%u += %s.z) {\n",
2331 shader
->baseShader
.cur_loop_depth
, shader
->baseShader
.cur_loop_regno
,
2332 src1_param
.reg_name
, shader
->baseShader
.cur_loop_depth
, src1_param
.reg_name
,
2333 shader
->baseShader
.cur_loop_depth
, shader
->baseShader
.cur_loop_regno
, src1_param
.reg_name
);
2336 shader
->baseShader
.cur_loop_depth
++;
2337 shader
->baseShader
.cur_loop_regno
++;
2340 static void shader_glsl_end(const struct wined3d_shader_instruction
*ins
)
2342 IWineD3DBaseShaderImpl
*shader
= (IWineD3DBaseShaderImpl
*)ins
->ctx
->shader
;
2344 shader_addline(ins
->ctx
->buffer
, "}\n");
2346 if (ins
->handler_idx
== WINED3DSIH_ENDLOOP
)
2348 shader
->baseShader
.cur_loop_depth
--;
2349 shader
->baseShader
.cur_loop_regno
--;
2352 if (ins
->handler_idx
== WINED3DSIH_ENDREP
)
2354 shader
->baseShader
.cur_loop_depth
--;
2358 static void shader_glsl_rep(const struct wined3d_shader_instruction
*ins
)
2360 IWineD3DBaseShaderImpl
*shader
= (IWineD3DBaseShaderImpl
*)ins
->ctx
->shader
;
2361 glsl_src_param_t src0_param
;
2362 const DWORD
*control_values
= NULL
;
2363 const local_constant
*constant
;
2365 /* Try to hardcode local values to help the GLSL compiler to unroll and optimize the loop */
2366 if(ins
->src
[0].register_type
== WINED3DSPR_CONSTINT
) {
2367 LIST_FOR_EACH_ENTRY(constant
, &shader
->baseShader
.constantsI
, local_constant
, entry
) {
2368 if(constant
->idx
== ins
->src
[0].register_idx
) {
2369 control_values
= constant
->value
;
2375 if(control_values
) {
2376 shader_addline(ins
->ctx
->buffer
, "for (tmpInt%d = 0; tmpInt%d < %d; tmpInt%d++) {\n",
2377 shader
->baseShader
.cur_loop_depth
, shader
->baseShader
.cur_loop_depth
,
2378 control_values
[0], shader
->baseShader
.cur_loop_depth
);
2380 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2381 shader_addline(ins
->ctx
->buffer
, "for (tmpInt%d = 0; tmpInt%d < %s; tmpInt%d++) {\n",
2382 shader
->baseShader
.cur_loop_depth
, shader
->baseShader
.cur_loop_depth
,
2383 src0_param
.param_str
, shader
->baseShader
.cur_loop_depth
);
2385 shader
->baseShader
.cur_loop_depth
++;
2388 static void shader_glsl_if(const struct wined3d_shader_instruction
*ins
)
2390 glsl_src_param_t src0_param
;
2392 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2393 shader_addline(ins
->ctx
->buffer
, "if (%s) {\n", src0_param
.param_str
);
2396 static void shader_glsl_ifc(const struct wined3d_shader_instruction
*ins
)
2398 glsl_src_param_t src0_param
;
2399 glsl_src_param_t src1_param
;
2401 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2402 shader_glsl_add_src_param(ins
, &ins
->src
[1], WINED3DSP_WRITEMASK_0
, &src1_param
);
2404 shader_addline(ins
->ctx
->buffer
, "if (%s %s %s) {\n",
2405 src0_param
.param_str
, shader_get_comp_op(ins
->flags
), src1_param
.param_str
);
2408 static void shader_glsl_else(const struct wined3d_shader_instruction
*ins
)
2410 shader_addline(ins
->ctx
->buffer
, "} else {\n");
2413 static void shader_glsl_break(const struct wined3d_shader_instruction
*ins
)
2415 shader_addline(ins
->ctx
->buffer
, "break;\n");
2418 /* FIXME: According to MSDN the compare is done per component. */
2419 static void shader_glsl_breakc(const struct wined3d_shader_instruction
*ins
)
2421 glsl_src_param_t src0_param
;
2422 glsl_src_param_t src1_param
;
2424 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2425 shader_glsl_add_src_param(ins
, &ins
->src
[1], WINED3DSP_WRITEMASK_0
, &src1_param
);
2427 shader_addline(ins
->ctx
->buffer
, "if (%s %s %s) break;\n",
2428 src0_param
.param_str
, shader_get_comp_op(ins
->flags
), src1_param
.param_str
);
2431 static void shader_glsl_label(const struct wined3d_shader_instruction
*ins
)
2433 shader_addline(ins
->ctx
->buffer
, "}\n");
2434 shader_addline(ins
->ctx
->buffer
, "void subroutine%u () {\n", ins
->src
[0].register_idx
);
2437 static void shader_glsl_call(const struct wined3d_shader_instruction
*ins
)
2439 shader_addline(ins
->ctx
->buffer
, "subroutine%u();\n", ins
->src
[0].register_idx
);
2442 static void shader_glsl_callnz(const struct wined3d_shader_instruction
*ins
)
2444 glsl_src_param_t src1_param
;
2446 shader_glsl_add_src_param(ins
, &ins
->src
[1], WINED3DSP_WRITEMASK_0
, &src1_param
);
2447 shader_addline(ins
->ctx
->buffer
, "if (%s) subroutine%u();\n", src1_param
.param_str
, ins
->src
[0].register_idx
);
2450 /*********************************************
2451 * Pixel Shader Specific Code begins here
2452 ********************************************/
2453 static void pshader_glsl_tex(const struct wined3d_shader_instruction
*ins
)
2455 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*)ins
->ctx
->shader
;
2456 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
2457 DWORD shader_version
= ins
->ctx
->reg_maps
->shader_version
;
2458 glsl_sample_function_t sample_function
;
2459 DWORD sample_flags
= 0;
2460 WINED3DSAMPLER_TEXTURE_TYPE sampler_type
;
2462 DWORD mask
= 0, swizzle
;
2464 /* 1.0-1.4: Use destination register as sampler source.
2465 * 2.0+: Use provided sampler source. */
2466 if (shader_version
< WINED3DPS_VERSION(2,0)) sampler_idx
= ins
->dst
[0].register_idx
;
2467 else sampler_idx
= ins
->src
[1].register_idx
;
2468 sampler_type
= ins
->ctx
->reg_maps
->sampler_type
[sampler_idx
];
2470 if (shader_version
< WINED3DPS_VERSION(1,4))
2472 DWORD flags
= deviceImpl
->stateBlock
->textureState
[sampler_idx
][WINED3DTSS_TEXTURETRANSFORMFLAGS
];
2474 /* Projected cube textures don't make a lot of sense, the resulting coordinates stay the same. */
2475 if (flags
& WINED3DTTFF_PROJECTED
&& sampler_type
!= WINED3DSTT_CUBE
) {
2476 sample_flags
|= WINED3D_GLSL_SAMPLE_PROJECTED
;
2477 switch (flags
& ~WINED3DTTFF_PROJECTED
) {
2478 case WINED3DTTFF_COUNT1
: FIXME("WINED3DTTFF_PROJECTED with WINED3DTTFF_COUNT1?\n"); break;
2479 case WINED3DTTFF_COUNT2
: mask
= WINED3DSP_WRITEMASK_1
; break;
2480 case WINED3DTTFF_COUNT3
: mask
= WINED3DSP_WRITEMASK_2
; break;
2481 case WINED3DTTFF_COUNT4
:
2482 case WINED3DTTFF_DISABLE
: mask
= WINED3DSP_WRITEMASK_3
; break;
2486 else if (shader_version
< WINED3DPS_VERSION(2,0))
2488 DWORD src_mod
= ins
->src
[0].modifiers
;
2490 if (src_mod
== WINED3DSPSM_DZ
) {
2491 sample_flags
|= WINED3D_GLSL_SAMPLE_PROJECTED
;
2492 mask
= WINED3DSP_WRITEMASK_2
;
2493 } else if (src_mod
== WINED3DSPSM_DW
) {
2494 sample_flags
|= WINED3D_GLSL_SAMPLE_PROJECTED
;
2495 mask
= WINED3DSP_WRITEMASK_3
;
2498 if (ins
->flags
& WINED3DSI_TEXLD_PROJECT
)
2500 /* ps 2.0 texldp instruction always divides by the fourth component. */
2501 sample_flags
|= WINED3D_GLSL_SAMPLE_PROJECTED
;
2502 mask
= WINED3DSP_WRITEMASK_3
;
2506 if(deviceImpl
->stateBlock
->textures
[sampler_idx
] &&
2507 IWineD3DBaseTexture_GetTextureDimensions(deviceImpl
->stateBlock
->textures
[sampler_idx
]) == GL_TEXTURE_RECTANGLE_ARB
) {
2508 sample_flags
|= WINED3D_GLSL_SAMPLE_RECT
;
2511 shader_glsl_get_sample_function(sampler_type
, sample_flags
, &sample_function
);
2512 mask
|= sample_function
.coord_mask
;
2514 if (shader_version
< WINED3DPS_VERSION(2,0)) swizzle
= WINED3DSP_NOSWIZZLE
;
2515 else swizzle
= ins
->src
[1].swizzle
;
2517 /* 1.0-1.3: Use destination register as coordinate source.
2518 1.4+: Use provided coordinate source register. */
2519 if (shader_version
< WINED3DPS_VERSION(1,4))
2522 shader_glsl_write_mask_to_str(mask
, coord_mask
);
2523 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, swizzle
, NULL
,
2524 "T%u%s", sampler_idx
, coord_mask
);
2526 glsl_src_param_t coord_param
;
2527 shader_glsl_add_src_param(ins
, &ins
->src
[0], mask
, &coord_param
);
2528 if (ins
->flags
& WINED3DSI_TEXLD_BIAS
)
2530 glsl_src_param_t bias
;
2531 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_3
, &bias
);
2532 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, swizzle
, bias
.param_str
,
2533 "%s", coord_param
.param_str
);
2535 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, swizzle
, NULL
,
2536 "%s", coord_param
.param_str
);
2541 static void shader_glsl_texldl(const struct wined3d_shader_instruction
*ins
)
2543 IWineD3DBaseShaderImpl
*This
= (IWineD3DBaseShaderImpl
*)ins
->ctx
->shader
;
2544 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
2545 glsl_sample_function_t sample_function
;
2546 glsl_src_param_t coord_param
, lod_param
;
2547 DWORD sample_flags
= WINED3D_GLSL_SAMPLE_LOD
;
2550 DWORD swizzle
= ins
->src
[1].swizzle
;
2552 sampler_idx
= ins
->src
[1].register_idx
;
2553 sampler_type
= ins
->ctx
->reg_maps
->sampler_type
[sampler_idx
];
2554 if(deviceImpl
->stateBlock
->textures
[sampler_idx
] &&
2555 IWineD3DBaseTexture_GetTextureDimensions(deviceImpl
->stateBlock
->textures
[sampler_idx
]) == GL_TEXTURE_RECTANGLE_ARB
) {
2556 sample_flags
|= WINED3D_GLSL_SAMPLE_RECT
;
2558 shader_glsl_get_sample_function(sampler_type
, sample_flags
, &sample_function
);
2559 shader_glsl_add_src_param(ins
, &ins
->src
[0], sample_function
.coord_mask
, &coord_param
);
2561 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_3
, &lod_param
);
2563 if (shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
))
2565 /* The GLSL spec claims the Lod sampling functions are only supported in vertex shaders.
2566 * However, they seem to work just fine in fragment shaders as well. */
2567 WARN("Using %s in fragment shader.\n", sample_function
.name
);
2569 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, swizzle
, lod_param
.param_str
,
2570 "%s", coord_param
.param_str
);
2573 static void pshader_glsl_texcoord(const struct wined3d_shader_instruction
*ins
)
2575 /* FIXME: Make this work for more than just 2D textures */
2576 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
2577 DWORD write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2579 if (ins
->ctx
->reg_maps
->shader_version
!= WINED3DPS_VERSION(1,4))
2583 shader_glsl_get_write_mask(&ins
->dst
[0], dst_mask
);
2584 shader_addline(buffer
, "clamp(gl_TexCoord[%u], 0.0, 1.0)%s);\n",
2585 ins
->dst
[0].register_idx
, dst_mask
);
2587 DWORD reg
= ins
->src
[0].register_idx
;
2588 DWORD src_mod
= ins
->src
[0].modifiers
;
2589 char dst_swizzle
[6];
2591 shader_glsl_get_swizzle(&ins
->src
[0], FALSE
, write_mask
, dst_swizzle
);
2593 if (src_mod
== WINED3DSPSM_DZ
) {
2594 glsl_src_param_t div_param
;
2595 unsigned int mask_size
= shader_glsl_get_write_mask_size(write_mask
);
2596 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_2
, &div_param
);
2598 if (mask_size
> 1) {
2599 shader_addline(buffer
, "gl_TexCoord[%u]%s / vec%d(%s));\n", reg
, dst_swizzle
, mask_size
, div_param
.param_str
);
2601 shader_addline(buffer
, "gl_TexCoord[%u]%s / %s);\n", reg
, dst_swizzle
, div_param
.param_str
);
2603 } else if (src_mod
== WINED3DSPSM_DW
) {
2604 glsl_src_param_t div_param
;
2605 unsigned int mask_size
= shader_glsl_get_write_mask_size(write_mask
);
2606 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_3
, &div_param
);
2608 if (mask_size
> 1) {
2609 shader_addline(buffer
, "gl_TexCoord[%u]%s / vec%d(%s));\n", reg
, dst_swizzle
, mask_size
, div_param
.param_str
);
2611 shader_addline(buffer
, "gl_TexCoord[%u]%s / %s);\n", reg
, dst_swizzle
, div_param
.param_str
);
2614 shader_addline(buffer
, "gl_TexCoord[%u]%s);\n", reg
, dst_swizzle
);
2619 /** Process the WINED3DSIO_TEXDP3TEX instruction in GLSL:
2620 * Take a 3-component dot product of the TexCoord[dstreg] and src,
2621 * then perform a 1D texture lookup from stage dstregnum, place into dst. */
2622 static void pshader_glsl_texdp3tex(const struct wined3d_shader_instruction
*ins
)
2624 glsl_src_param_t src0_param
;
2625 glsl_sample_function_t sample_function
;
2626 DWORD sampler_idx
= ins
->dst
[0].register_idx
;
2627 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2628 WINED3DSAMPLER_TEXTURE_TYPE sampler_type
= ins
->ctx
->reg_maps
->sampler_type
[sampler_idx
];
2631 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
2633 /* Do I have to take care about the projected bit? I don't think so, since the dp3 returns only one
2634 * scalar, and projected sampling would require 4.
2636 * It is a dependent read - not valid with conditional NP2 textures
2638 shader_glsl_get_sample_function(sampler_type
, 0, &sample_function
);
2639 mask_size
= shader_glsl_get_write_mask_size(sample_function
.coord_mask
);
2644 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, WINED3DSP_NOSWIZZLE
, NULL
,
2645 "dot(gl_TexCoord[%u].xyz, %s)", sampler_idx
, src0_param
.param_str
);
2649 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, WINED3DSP_NOSWIZZLE
, NULL
,
2650 "vec2(dot(gl_TexCoord[%u].xyz, %s), 0.0)", sampler_idx
, src0_param
.param_str
);
2654 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, WINED3DSP_NOSWIZZLE
, NULL
,
2655 "vec3(dot(gl_TexCoord[%u].xyz, %s), 0.0, 0.0)", sampler_idx
, src0_param
.param_str
);
2659 FIXME("Unexpected mask size %u\n", mask_size
);
2664 /** Process the WINED3DSIO_TEXDP3 instruction in GLSL:
2665 * Take a 3-component dot product of the TexCoord[dstreg] and src. */
2666 static void pshader_glsl_texdp3(const struct wined3d_shader_instruction
*ins
)
2668 glsl_src_param_t src0_param
;
2669 DWORD dstreg
= ins
->dst
[0].register_idx
;
2670 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2672 unsigned int mask_size
;
2674 dst_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2675 mask_size
= shader_glsl_get_write_mask_size(dst_mask
);
2676 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
2678 if (mask_size
> 1) {
2679 shader_addline(ins
->ctx
->buffer
, "vec%d(dot(T%u.xyz, %s)));\n", mask_size
, dstreg
, src0_param
.param_str
);
2681 shader_addline(ins
->ctx
->buffer
, "dot(T%u.xyz, %s));\n", dstreg
, src0_param
.param_str
);
2685 /** Process the WINED3DSIO_TEXDEPTH instruction in GLSL:
2686 * Calculate the depth as dst.x / dst.y */
2687 static void pshader_glsl_texdepth(const struct wined3d_shader_instruction
*ins
)
2689 glsl_dst_param_t dst_param
;
2691 shader_glsl_add_dst_param(ins
, &ins
->dst
[0], &dst_param
);
2693 /* Tests show that texdepth never returns anything below 0.0, and that r5.y is clamped to 1.0.
2694 * Negative input is accepted, -0.25 / -0.5 returns 0.5. GL should clamp gl_FragDepth to [0;1], but
2695 * this doesn't always work, so clamp the results manually. Whether or not the x value is clamped at 1
2696 * too is irrelevant, since if x = 0, any y value < 1.0 (and > 1.0 is not allowed) results in a result
2699 shader_addline(ins
->ctx
->buffer
, "gl_FragDepth = clamp((%s.x / min(%s.y, 1.0)), 0.0, 1.0);\n",
2700 dst_param
.reg_name
, dst_param
.reg_name
);
2703 /** Process the WINED3DSIO_TEXM3X2DEPTH instruction in GLSL:
2704 * Last row of a 3x2 matrix multiply, use the result to calculate the depth:
2705 * Calculate tmp0.y = TexCoord[dstreg] . src.xyz; (tmp0.x has already been calculated)
2706 * depth = (tmp0.y == 0.0) ? 1.0 : tmp0.x / tmp0.y
2708 static void pshader_glsl_texm3x2depth(const struct wined3d_shader_instruction
*ins
)
2710 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2711 DWORD dstreg
= ins
->dst
[0].register_idx
;
2712 glsl_src_param_t src0_param
;
2714 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
2716 shader_addline(ins
->ctx
->buffer
, "tmp0.y = dot(T%u.xyz, %s);\n", dstreg
, src0_param
.param_str
);
2717 shader_addline(ins
->ctx
->buffer
, "gl_FragDepth = (tmp0.y == 0.0) ? 1.0 : clamp(tmp0.x / tmp0.y, 0.0, 1.0);\n");
2720 /** Process the WINED3DSIO_TEXM3X2PAD instruction in GLSL
2721 * Calculate the 1st of a 2-row matrix multiplication. */
2722 static void pshader_glsl_texm3x2pad(const struct wined3d_shader_instruction
*ins
)
2724 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2725 DWORD reg
= ins
->dst
[0].register_idx
;
2726 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
2727 glsl_src_param_t src0_param
;
2729 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
2730 shader_addline(buffer
, "tmp0.x = dot(T%u.xyz, %s);\n", reg
, src0_param
.param_str
);
2733 /** Process the WINED3DSIO_TEXM3X3PAD instruction in GLSL
2734 * Calculate the 1st or 2nd row of a 3-row matrix multiplication. */
2735 static void pshader_glsl_texm3x3pad(const struct wined3d_shader_instruction
*ins
)
2737 IWineD3DPixelShaderImpl
*shader
= (IWineD3DPixelShaderImpl
*)ins
->ctx
->shader
;
2738 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2739 DWORD reg
= ins
->dst
[0].register_idx
;
2740 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
2741 SHADER_PARSE_STATE
* current_state
= &shader
->baseShader
.parse_state
;
2742 glsl_src_param_t src0_param
;
2744 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
2745 shader_addline(buffer
, "tmp0.%c = dot(T%u.xyz, %s);\n", 'x' + current_state
->current_row
, reg
, src0_param
.param_str
);
2746 current_state
->texcoord_w
[current_state
->current_row
++] = reg
;
2749 static void pshader_glsl_texm3x2tex(const struct wined3d_shader_instruction
*ins
)
2751 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2752 DWORD reg
= ins
->dst
[0].register_idx
;
2753 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
2754 glsl_src_param_t src0_param
;
2755 WINED3DSAMPLER_TEXTURE_TYPE sampler_type
= ins
->ctx
->reg_maps
->sampler_type
[reg
];
2756 glsl_sample_function_t sample_function
;
2758 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
2759 shader_addline(buffer
, "tmp0.y = dot(T%u.xyz, %s);\n", reg
, src0_param
.param_str
);
2761 shader_glsl_get_sample_function(sampler_type
, 0, &sample_function
);
2763 /* Sample the texture using the calculated coordinates */
2764 shader_glsl_gen_sample_code(ins
, reg
, &sample_function
, WINED3DSP_NOSWIZZLE
, NULL
, "tmp0.xy");
2767 /** Process the WINED3DSIO_TEXM3X3TEX instruction in GLSL
2768 * Perform the 3rd row of a 3x3 matrix multiply, then sample the texture using the calculated coordinates */
2769 static void pshader_glsl_texm3x3tex(const struct wined3d_shader_instruction
*ins
)
2771 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2772 glsl_src_param_t src0_param
;
2773 DWORD reg
= ins
->dst
[0].register_idx
;
2774 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*)ins
->ctx
->shader
;
2775 SHADER_PARSE_STATE
* current_state
= &This
->baseShader
.parse_state
;
2776 WINED3DSAMPLER_TEXTURE_TYPE sampler_type
= ins
->ctx
->reg_maps
->sampler_type
[reg
];
2777 glsl_sample_function_t sample_function
;
2779 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
2780 shader_addline(ins
->ctx
->buffer
, "tmp0.z = dot(T%u.xyz, %s);\n", reg
, src0_param
.param_str
);
2782 /* Dependent read, not valid with conditional NP2 */
2783 shader_glsl_get_sample_function(sampler_type
, 0, &sample_function
);
2785 /* Sample the texture using the calculated coordinates */
2786 shader_glsl_gen_sample_code(ins
, reg
, &sample_function
, WINED3DSP_NOSWIZZLE
, NULL
, "tmp0.xyz");
2788 current_state
->current_row
= 0;
2791 /** Process the WINED3DSIO_TEXM3X3 instruction in GLSL
2792 * Perform the 3rd row of a 3x3 matrix multiply */
2793 static void pshader_glsl_texm3x3(const struct wined3d_shader_instruction
*ins
)
2795 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2796 glsl_src_param_t src0_param
;
2798 DWORD reg
= ins
->dst
[0].register_idx
;
2799 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*)ins
->ctx
->shader
;
2800 SHADER_PARSE_STATE
* current_state
= &This
->baseShader
.parse_state
;
2802 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
2804 shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2805 shader_glsl_get_write_mask(&ins
->dst
[0], dst_mask
);
2806 shader_addline(ins
->ctx
->buffer
, "vec4(tmp0.xy, dot(T%u.xyz, %s), 1.0)%s);\n", reg
, src0_param
.param_str
, dst_mask
);
2808 current_state
->current_row
= 0;
2811 /** Process the WINED3DSIO_TEXM3X3SPEC instruction in GLSL
2812 * Perform the final texture lookup based on the previous 2 3x3 matrix multiplies */
2813 static void pshader_glsl_texm3x3spec(const struct wined3d_shader_instruction
*ins
)
2815 IWineD3DPixelShaderImpl
*shader
= (IWineD3DPixelShaderImpl
*)ins
->ctx
->shader
;
2816 DWORD reg
= ins
->dst
[0].register_idx
;
2817 glsl_src_param_t src0_param
;
2818 glsl_src_param_t src1_param
;
2819 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
2820 SHADER_PARSE_STATE
* current_state
= &shader
->baseShader
.parse_state
;
2821 WINED3DSAMPLER_TEXTURE_TYPE stype
= ins
->ctx
->reg_maps
->sampler_type
[reg
];
2822 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2823 glsl_sample_function_t sample_function
;
2825 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
2826 shader_glsl_add_src_param(ins
, &ins
->src
[1], src_mask
, &src1_param
);
2828 /* Perform the last matrix multiply operation */
2829 shader_addline(buffer
, "tmp0.z = dot(T%u.xyz, %s);\n", reg
, src0_param
.param_str
);
2830 /* Reflection calculation */
2831 shader_addline(buffer
, "tmp0.xyz = -reflect((%s), normalize(tmp0.xyz));\n", src1_param
.param_str
);
2833 /* Dependent read, not valid with conditional NP2 */
2834 shader_glsl_get_sample_function(stype
, 0, &sample_function
);
2836 /* Sample the texture */
2837 shader_glsl_gen_sample_code(ins
, reg
, &sample_function
, WINED3DSP_NOSWIZZLE
, NULL
, "tmp0.xyz");
2839 current_state
->current_row
= 0;
2842 /** Process the WINED3DSIO_TEXM3X3VSPEC instruction in GLSL
2843 * Perform the final texture lookup based on the previous 2 3x3 matrix multiplies */
2844 static void pshader_glsl_texm3x3vspec(const struct wined3d_shader_instruction
*ins
)
2846 IWineD3DPixelShaderImpl
*shader
= (IWineD3DPixelShaderImpl
*)ins
->ctx
->shader
;
2847 DWORD reg
= ins
->dst
[0].register_idx
;
2848 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
2849 SHADER_PARSE_STATE
* current_state
= &shader
->baseShader
.parse_state
;
2850 glsl_src_param_t src0_param
;
2851 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2852 WINED3DSAMPLER_TEXTURE_TYPE sampler_type
= ins
->ctx
->reg_maps
->sampler_type
[reg
];
2853 glsl_sample_function_t sample_function
;
2855 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
2857 /* Perform the last matrix multiply operation */
2858 shader_addline(buffer
, "tmp0.z = dot(vec3(T%u), vec3(%s));\n", reg
, src0_param
.param_str
);
2860 /* Construct the eye-ray vector from w coordinates */
2861 shader_addline(buffer
, "tmp1.xyz = normalize(vec3(gl_TexCoord[%u].w, gl_TexCoord[%u].w, gl_TexCoord[%u].w));\n",
2862 current_state
->texcoord_w
[0], current_state
->texcoord_w
[1], reg
);
2863 shader_addline(buffer
, "tmp0.xyz = -reflect(tmp1.xyz, normalize(tmp0.xyz));\n");
2865 /* Dependent read, not valid with conditional NP2 */
2866 shader_glsl_get_sample_function(sampler_type
, 0, &sample_function
);
2868 /* Sample the texture using the calculated coordinates */
2869 shader_glsl_gen_sample_code(ins
, reg
, &sample_function
, WINED3DSP_NOSWIZZLE
, NULL
, "tmp0.xyz");
2871 current_state
->current_row
= 0;
2874 /** Process the WINED3DSIO_TEXBEM instruction in GLSL.
2875 * Apply a fake bump map transform.
2876 * texbem is pshader <= 1.3 only, this saves a few version checks
2878 static void pshader_glsl_texbem(const struct wined3d_shader_instruction
*ins
)
2880 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*)ins
->ctx
->shader
;
2881 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
2882 glsl_sample_function_t sample_function
;
2883 glsl_src_param_t coord_param
;
2884 WINED3DSAMPLER_TEXTURE_TYPE sampler_type
;
2890 sampler_idx
= ins
->dst
[0].register_idx
;
2891 flags
= deviceImpl
->stateBlock
->textureState
[sampler_idx
][WINED3DTSS_TEXTURETRANSFORMFLAGS
];
2893 sampler_type
= ins
->ctx
->reg_maps
->sampler_type
[sampler_idx
];
2894 /* Dependent read, not valid with conditional NP2 */
2895 shader_glsl_get_sample_function(sampler_type
, 0, &sample_function
);
2896 mask
= sample_function
.coord_mask
;
2898 shader_glsl_write_mask_to_str(mask
, coord_mask
);
2900 /* with projective textures, texbem only divides the static texture coord, not the displacement,
2901 * so we can't let the GL handle this.
2903 if (flags
& WINED3DTTFF_PROJECTED
) {
2905 char coord_div_mask
[3];
2906 switch (flags
& ~WINED3DTTFF_PROJECTED
) {
2907 case WINED3DTTFF_COUNT1
: FIXME("WINED3DTTFF_PROJECTED with WINED3DTTFF_COUNT1?\n"); break;
2908 case WINED3DTTFF_COUNT2
: div_mask
= WINED3DSP_WRITEMASK_1
; break;
2909 case WINED3DTTFF_COUNT3
: div_mask
= WINED3DSP_WRITEMASK_2
; break;
2910 case WINED3DTTFF_COUNT4
:
2911 case WINED3DTTFF_DISABLE
: div_mask
= WINED3DSP_WRITEMASK_3
; break;
2913 shader_glsl_write_mask_to_str(div_mask
, coord_div_mask
);
2914 shader_addline(ins
->ctx
->buffer
, "T%u%s /= T%u%s;\n", sampler_idx
, coord_mask
, sampler_idx
, coord_div_mask
);
2917 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
, &coord_param
);
2919 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, WINED3DSP_NOSWIZZLE
, NULL
,
2920 "T%u%s + vec4(bumpenvmat%d * %s, 0.0, 0.0)%s", sampler_idx
, coord_mask
, sampler_idx
,
2921 coord_param
.param_str
, coord_mask
);
2923 if (ins
->handler_idx
== WINED3DSIH_TEXBEML
)
2925 glsl_src_param_t luminance_param
;
2926 glsl_dst_param_t dst_param
;
2928 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_2
, &luminance_param
);
2929 shader_glsl_add_dst_param(ins
, &ins
->dst
[0], &dst_param
);
2931 shader_addline(ins
->ctx
->buffer
, "%s%s *= (%s * luminancescale%d + luminanceoffset%d);\n",
2932 dst_param
.reg_name
, dst_param
.mask_str
,
2933 luminance_param
.param_str
, sampler_idx
, sampler_idx
);
2937 static void pshader_glsl_bem(const struct wined3d_shader_instruction
*ins
)
2939 glsl_src_param_t src0_param
, src1_param
;
2940 DWORD sampler_idx
= ins
->dst
[0].register_idx
;
2942 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
, &src0_param
);
2943 shader_glsl_add_src_param(ins
, &ins
->src
[1], WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
, &src1_param
);
2945 shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2946 shader_addline(ins
->ctx
->buffer
, "%s + bumpenvmat%d * %s);\n",
2947 src0_param
.param_str
, sampler_idx
, src1_param
.param_str
);
2950 /** Process the WINED3DSIO_TEXREG2AR instruction in GLSL
2951 * Sample 2D texture at dst using the alpha & red (wx) components of src as texture coordinates */
2952 static void pshader_glsl_texreg2ar(const struct wined3d_shader_instruction
*ins
)
2954 glsl_src_param_t src0_param
;
2955 DWORD sampler_idx
= ins
->dst
[0].register_idx
;
2956 WINED3DSAMPLER_TEXTURE_TYPE sampler_type
= ins
->ctx
->reg_maps
->sampler_type
[sampler_idx
];
2957 glsl_sample_function_t sample_function
;
2959 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_ALL
, &src0_param
);
2961 shader_glsl_get_sample_function(sampler_type
, 0, &sample_function
);
2962 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, WINED3DSP_NOSWIZZLE
, NULL
,
2963 "%s.wx", src0_param
.reg_name
);
2966 /** Process the WINED3DSIO_TEXREG2GB instruction in GLSL
2967 * Sample 2D texture at dst using the green & blue (yz) components of src as texture coordinates */
2968 static void pshader_glsl_texreg2gb(const struct wined3d_shader_instruction
*ins
)
2970 glsl_src_param_t src0_param
;
2971 DWORD sampler_idx
= ins
->dst
[0].register_idx
;
2972 WINED3DSAMPLER_TEXTURE_TYPE sampler_type
= ins
->ctx
->reg_maps
->sampler_type
[sampler_idx
];
2973 glsl_sample_function_t sample_function
;
2975 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_ALL
, &src0_param
);
2977 shader_glsl_get_sample_function(sampler_type
, 0, &sample_function
);
2978 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, WINED3DSP_NOSWIZZLE
, NULL
,
2979 "%s.yz", src0_param
.reg_name
);
2982 /** Process the WINED3DSIO_TEXREG2RGB instruction in GLSL
2983 * Sample texture at dst using the rgb (xyz) components of src as texture coordinates */
2984 static void pshader_glsl_texreg2rgb(const struct wined3d_shader_instruction
*ins
)
2986 glsl_src_param_t src0_param
;
2987 DWORD sampler_idx
= ins
->dst
[0].register_idx
;
2988 WINED3DSAMPLER_TEXTURE_TYPE sampler_type
= ins
->ctx
->reg_maps
->sampler_type
[sampler_idx
];
2989 glsl_sample_function_t sample_function
;
2991 /* Dependent read, not valid with conditional NP2 */
2992 shader_glsl_get_sample_function(sampler_type
, 0, &sample_function
);
2993 shader_glsl_add_src_param(ins
, &ins
->src
[0], sample_function
.coord_mask
, &src0_param
);
2995 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, WINED3DSP_NOSWIZZLE
, NULL
,
2996 "%s", src0_param
.param_str
);
2999 /** Process the WINED3DSIO_TEXKILL instruction in GLSL.
3000 * If any of the first 3 components are < 0, discard this pixel */
3001 static void pshader_glsl_texkill(const struct wined3d_shader_instruction
*ins
)
3003 glsl_dst_param_t dst_param
;
3005 /* The argument is a destination parameter, and no writemasks are allowed */
3006 shader_glsl_add_dst_param(ins
, &ins
->dst
[0], &dst_param
);
3007 if ((ins
->ctx
->reg_maps
->shader_version
>= WINED3DPS_VERSION(2,0)))
3009 /* 2.0 shaders compare all 4 components in texkill */
3010 shader_addline(ins
->ctx
->buffer
, "if (any(lessThan(%s.xyzw, vec4(0.0)))) discard;\n", dst_param
.reg_name
);
3012 /* 1.X shaders only compare the first 3 components, probably due to the nature of the texkill
3013 * instruction as a tex* instruction, and phase, which kills all a / w components. Even if all
3014 * 4 components are defined, only the first 3 are used
3016 shader_addline(ins
->ctx
->buffer
, "if (any(lessThan(%s.xyz, vec3(0.0)))) discard;\n", dst_param
.reg_name
);
3020 /** Process the WINED3DSIO_DP2ADD instruction in GLSL.
3021 * dst = dot2(src0, src1) + src2 */
3022 static void pshader_glsl_dp2add(const struct wined3d_shader_instruction
*ins
)
3024 glsl_src_param_t src0_param
;
3025 glsl_src_param_t src1_param
;
3026 glsl_src_param_t src2_param
;
3028 unsigned int mask_size
;
3030 write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
3031 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
3033 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
, &src0_param
);
3034 shader_glsl_add_src_param(ins
, &ins
->src
[1], WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
, &src1_param
);
3035 shader_glsl_add_src_param(ins
, &ins
->src
[2], WINED3DSP_WRITEMASK_0
, &src2_param
);
3037 if (mask_size
> 1) {
3038 shader_addline(ins
->ctx
->buffer
, "vec%d(dot(%s, %s) + %s));\n",
3039 mask_size
, src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
3041 shader_addline(ins
->ctx
->buffer
, "dot(%s, %s) + %s);\n",
3042 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
3046 static void pshader_glsl_input_pack(IWineD3DPixelShader
*iface
, SHADER_BUFFER
*buffer
,
3047 const struct wined3d_shader_semantic
*semantics_in
, const struct shader_reg_maps
*reg_maps
,
3048 enum vertexprocessing_mode vertexprocessing
)
3051 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*)iface
;
3053 for (i
= 0; i
< MAX_REG_INPUT
; ++i
)
3055 DWORD usage
, usage_idx
;
3059 if (!reg_maps
->packed_input
[i
]) continue;
3061 usage
= semantics_in
[i
].usage
;
3062 usage_idx
= semantics_in
[i
].usage_idx
;
3063 shader_glsl_get_write_mask(&semantics_in
[i
].reg
, reg_mask
);
3067 case WINED3DDECLUSAGE_TEXCOORD
:
3068 if (usage_idx
< 8 && vertexprocessing
== pretransformed
)
3069 shader_addline(buffer
, "IN[%u]%s = gl_TexCoord[%u]%s;\n",
3070 This
->input_reg_map
[i
], reg_mask
, usage_idx
, reg_mask
);
3072 shader_addline(buffer
, "IN[%u]%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
3073 This
->input_reg_map
[i
], reg_mask
, reg_mask
);
3076 case WINED3DDECLUSAGE_COLOR
:
3078 shader_addline(buffer
, "IN[%u]%s = vec4(gl_Color)%s;\n",
3079 This
->input_reg_map
[i
], reg_mask
, reg_mask
);
3080 else if (usage_idx
== 1)
3081 shader_addline(buffer
, "IN[%u]%s = vec4(gl_SecondaryColor)%s;\n",
3082 This
->input_reg_map
[i
], reg_mask
, reg_mask
);
3084 shader_addline(buffer
, "IN[%u]%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
3085 This
->input_reg_map
[i
], reg_mask
, reg_mask
);
3089 shader_addline(buffer
, "IN[%u]%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
3090 This
->input_reg_map
[i
], reg_mask
, reg_mask
);
3096 /*********************************************
3097 * Vertex Shader Specific Code begins here
3098 ********************************************/
3100 static void add_glsl_program_entry(struct shader_glsl_priv
*priv
, struct glsl_shader_prog_link
*entry
) {
3101 glsl_program_key_t
*key
;
3103 key
= HeapAlloc(GetProcessHeap(), 0, sizeof(glsl_program_key_t
));
3104 key
->vshader
= entry
->vshader
;
3105 key
->pshader
= entry
->pshader
;
3106 key
->vs_args
= entry
->vs_args
;
3107 key
->ps_args
= entry
->ps_args
;
3109 hash_table_put(priv
->glsl_program_lookup
, key
, entry
);
3112 static struct glsl_shader_prog_link
*get_glsl_program_entry(struct shader_glsl_priv
*priv
,
3113 IWineD3DVertexShader
*vshader
, IWineD3DPixelShader
*pshader
, struct vs_compile_args
*vs_args
,
3114 struct ps_compile_args
*ps_args
) {
3115 glsl_program_key_t key
;
3117 key
.vshader
= vshader
;
3118 key
.pshader
= pshader
;
3119 key
.vs_args
= *vs_args
;
3120 key
.ps_args
= *ps_args
;
3122 return hash_table_get(priv
->glsl_program_lookup
, &key
);
3125 static void delete_glsl_program_entry(struct shader_glsl_priv
*priv
, const WineD3D_GL_Info
*gl_info
,
3126 struct glsl_shader_prog_link
*entry
)
3128 glsl_program_key_t
*key
;
3130 key
= HeapAlloc(GetProcessHeap(), 0, sizeof(glsl_program_key_t
));
3131 key
->vshader
= entry
->vshader
;
3132 key
->pshader
= entry
->pshader
;
3133 key
->vs_args
= entry
->vs_args
;
3134 key
->ps_args
= entry
->ps_args
;
3135 hash_table_remove(priv
->glsl_program_lookup
, key
);
3137 GL_EXTCALL(glDeleteObjectARB(entry
->programId
));
3138 if (entry
->vshader
) list_remove(&entry
->vshader_entry
);
3139 if (entry
->pshader
) list_remove(&entry
->pshader_entry
);
3140 HeapFree(GetProcessHeap(), 0, entry
->vuniformF_locations
);
3141 HeapFree(GetProcessHeap(), 0, entry
->puniformF_locations
);
3142 HeapFree(GetProcessHeap(), 0, entry
);
3145 static void handle_ps3_input(SHADER_BUFFER
*buffer
, const WineD3D_GL_Info
*gl_info
, const DWORD
*map
,
3146 const struct wined3d_shader_semantic
*semantics_in
, const struct shader_reg_maps
*reg_maps_in
,
3147 const struct wined3d_shader_semantic
*semantics_out
, const struct shader_reg_maps
*reg_maps_out
)
3150 DWORD usage
, usage_idx
, usage_out
, usage_idx_out
;
3153 DWORD in_count
= GL_LIMITS(glsl_varyings
) / 4;
3154 char reg_mask
[6], reg_mask_out
[6];
3155 char destination
[50];
3157 set
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*set
) * (in_count
+ 2));
3159 if (!semantics_out
) {
3160 /* Save gl_FrontColor & gl_FrontSecondaryColor before overwriting them. */
3161 shader_addline(buffer
, "vec4 front_color = gl_FrontColor;\n");
3162 shader_addline(buffer
, "vec4 front_secondary_color = gl_FrontSecondaryColor;\n");
3165 for(i
= 0; i
< MAX_REG_INPUT
; i
++) {
3166 if (!reg_maps_in
->packed_input
[i
]) continue;
3169 if (in_idx
>= (in_count
+ 2)) {
3170 FIXME("More input varyings declared than supported, expect issues\n");
3173 else if (map
[i
] == ~0U)
3175 /* Declared, but not read register */
3179 if (in_idx
== in_count
) {
3180 sprintf(destination
, "gl_FrontColor");
3181 } else if (in_idx
== in_count
+ 1) {
3182 sprintf(destination
, "gl_FrontSecondaryColor");
3184 sprintf(destination
, "IN[%u]", in_idx
);
3187 usage
= semantics_in
[i
].usage
;
3188 usage_idx
= semantics_in
[i
].usage_idx
;
3189 set
[map
[i
]] = shader_glsl_get_write_mask(&semantics_in
[i
].reg
, reg_mask
);
3191 if(!semantics_out
) {
3193 case WINED3DDECLUSAGE_COLOR
:
3195 shader_addline(buffer
, "%s%s = front_color%s;\n",
3196 destination
, reg_mask
, reg_mask
);
3197 else if (usage_idx
== 1)
3198 shader_addline(buffer
, "%s%s = front_secondary_color%s;\n",
3199 destination
, reg_mask
, reg_mask
);
3201 shader_addline(buffer
, "%s%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
3202 destination
, reg_mask
, reg_mask
);
3205 case WINED3DDECLUSAGE_TEXCOORD
:
3206 if (usage_idx
< 8) {
3207 shader_addline(buffer
, "%s%s = gl_TexCoord[%u]%s;\n",
3208 destination
, reg_mask
, usage_idx
, reg_mask
);
3210 shader_addline(buffer
, "%s%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
3211 destination
, reg_mask
, reg_mask
);
3215 case WINED3DDECLUSAGE_FOG
:
3216 shader_addline(buffer
, "%s%s = vec4(gl_FogFragCoord, 0.0, 0.0, 0.0)%s;\n",
3217 destination
, reg_mask
, reg_mask
);
3221 shader_addline(buffer
, "%s%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
3222 destination
, reg_mask
, reg_mask
);
3226 for(j
= 0; j
< MAX_REG_OUTPUT
; j
++) {
3227 if (!reg_maps_out
->packed_output
[j
]) continue;
3229 usage_out
= semantics_out
[j
].usage
;
3230 usage_idx_out
= semantics_out
[j
].usage_idx
;
3231 shader_glsl_get_write_mask(&semantics_out
[j
].reg
, reg_mask_out
);
3233 if(usage
== usage_out
&&
3234 usage_idx
== usage_idx_out
) {
3235 shader_addline(buffer
, "%s%s = OUT[%u]%s;\n",
3236 destination
, reg_mask
, j
, reg_mask
);
3241 shader_addline(buffer
, "%s%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
3242 destination
, reg_mask
, reg_mask
);
3247 /* This is solely to make the compiler / linker happy and avoid warning about undefined
3248 * varyings. It shouldn't result in any real code executed on the GPU, since all read
3249 * input varyings are assigned above, if the optimizer works properly.
3251 for(i
= 0; i
< in_count
+ 2; i
++) {
3252 if(set
[i
] != WINED3DSP_WRITEMASK_ALL
) {
3253 unsigned int size
= 0;
3254 memset(reg_mask
, 0, sizeof(reg_mask
));
3255 if(!(set
[i
] & WINED3DSP_WRITEMASK_0
)) {
3256 reg_mask
[size
] = 'x';
3259 if(!(set
[i
] & WINED3DSP_WRITEMASK_1
)) {
3260 reg_mask
[size
] = 'y';
3263 if(!(set
[i
] & WINED3DSP_WRITEMASK_2
)) {
3264 reg_mask
[size
] = 'z';
3267 if(!(set
[i
] & WINED3DSP_WRITEMASK_3
)) {
3268 reg_mask
[size
] = 'w';
3272 if (i
== in_count
) {
3273 sprintf(destination
, "gl_FrontColor");
3274 } else if (i
== in_count
+ 1) {
3275 sprintf(destination
, "gl_FrontSecondaryColor");
3277 sprintf(destination
, "IN[%u]", i
);
3281 shader_addline(buffer
, "%s.%s = 0.0;\n", destination
, reg_mask
);
3283 shader_addline(buffer
, "%s.%s = vec%u(0.0);\n", destination
, reg_mask
, size
);
3288 HeapFree(GetProcessHeap(), 0, set
);
3291 static GLhandleARB
generate_param_reorder_function(IWineD3DVertexShader
*vertexshader
,
3292 IWineD3DPixelShader
*pixelshader
, const WineD3D_GL_Info
*gl_info
)
3294 GLhandleARB ret
= 0;
3295 IWineD3DVertexShaderImpl
*vs
= (IWineD3DVertexShaderImpl
*) vertexshader
;
3296 IWineD3DPixelShaderImpl
*ps
= (IWineD3DPixelShaderImpl
*) pixelshader
;
3297 IWineD3DDeviceImpl
*device
;
3298 DWORD vs_major
= WINED3DSHADER_VERSION_MAJOR(vs
->baseShader
.reg_maps
.shader_version
);
3299 DWORD ps_major
= ps
? WINED3DSHADER_VERSION_MAJOR(ps
->baseShader
.reg_maps
.shader_version
) : 0;
3301 SHADER_BUFFER buffer
;
3302 DWORD usage
, usage_idx
, writemask
;
3304 const struct wined3d_shader_semantic
*semantics_out
;
3306 shader_buffer_init(&buffer
);
3308 shader_addline(&buffer
, "#version 120\n");
3310 if(vs_major
< 3 && ps_major
< 3) {
3311 /* That one is easy: The vertex shader writes to the builtin varyings, the pixel shader reads from them.
3312 * Take care about the texcoord .w fixup though if we're using the fixed function fragment pipeline
3314 device
= (IWineD3DDeviceImpl
*) vs
->baseShader
.device
;
3315 if((GLINFO_LOCATION
).set_texcoord_w
&& ps_major
== 0 && vs_major
> 0 &&
3316 !device
->frag_pipe
->ffp_proj_control
) {
3317 shader_addline(&buffer
, "void order_ps_input() {\n");
3318 for(i
= 0; i
< min(8, MAX_REG_TEXCRD
); i
++) {
3319 if(vs
->baseShader
.reg_maps
.texcoord_mask
[i
] != 0 &&
3320 vs
->baseShader
.reg_maps
.texcoord_mask
[i
] != WINED3DSP_WRITEMASK_ALL
) {
3321 shader_addline(&buffer
, "gl_TexCoord[%u].w = 1.0;\n", i
);
3324 shader_addline(&buffer
, "}\n");
3326 shader_addline(&buffer
, "void order_ps_input() { /* do nothing */ }\n");
3328 } else if(ps_major
< 3 && vs_major
>= 3) {
3329 /* The vertex shader writes to its own varyings, the pixel shader needs them in the builtin ones */
3330 semantics_out
= vs
->semantics_out
;
3332 shader_addline(&buffer
, "void order_ps_input(in vec4 OUT[%u]) {\n", MAX_REG_OUTPUT
);
3333 for(i
= 0; i
< MAX_REG_OUTPUT
; i
++) {
3334 if (!vs
->baseShader
.reg_maps
.packed_output
[i
]) continue;
3336 usage
= semantics_out
[i
].usage
;
3337 usage_idx
= semantics_out
[i
].usage_idx
;
3338 writemask
= shader_glsl_get_write_mask(&semantics_out
[i
].reg
, reg_mask
);
3341 case WINED3DDECLUSAGE_COLOR
:
3343 shader_addline(&buffer
, "gl_FrontColor%s = OUT[%u]%s;\n", reg_mask
, i
, reg_mask
);
3344 else if (usage_idx
== 1)
3345 shader_addline(&buffer
, "gl_FrontSecondaryColor%s = OUT[%u]%s;\n", reg_mask
, i
, reg_mask
);
3348 case WINED3DDECLUSAGE_POSITION
:
3349 shader_addline(&buffer
, "gl_Position%s = OUT[%u]%s;\n", reg_mask
, i
, reg_mask
);
3352 case WINED3DDECLUSAGE_TEXCOORD
:
3353 if (usage_idx
< 8) {
3354 if(!(GLINFO_LOCATION
).set_texcoord_w
|| ps_major
> 0) writemask
|= WINED3DSP_WRITEMASK_3
;
3356 shader_addline(&buffer
, "gl_TexCoord[%u]%s = OUT[%u]%s;\n",
3357 usage_idx
, reg_mask
, i
, reg_mask
);
3358 if(!(writemask
& WINED3DSP_WRITEMASK_3
)) {
3359 shader_addline(&buffer
, "gl_TexCoord[%u].w = 1.0;\n", usage_idx
);
3364 case WINED3DDECLUSAGE_PSIZE
:
3365 shader_addline(&buffer
, "gl_PointSize = OUT[%u].x;\n", i
);
3368 case WINED3DDECLUSAGE_FOG
:
3369 shader_addline(&buffer
, "gl_FogFragCoord = OUT[%u].%c;\n", i
, reg_mask
[1]);
3376 shader_addline(&buffer
, "}\n");
3378 } else if(ps_major
>= 3 && vs_major
>= 3) {
3379 semantics_out
= vs
->semantics_out
;
3381 /* This one is tricky: a 3.0 pixel shader reads from a 3.0 vertex shader */
3382 shader_addline(&buffer
, "varying vec4 IN[%u];\n", GL_LIMITS(glsl_varyings
) / 4);
3383 shader_addline(&buffer
, "void order_ps_input(in vec4 OUT[%u]) {\n", MAX_REG_OUTPUT
);
3385 /* First, sort out position and point size. Those are not passed to the pixel shader */
3386 for(i
= 0; i
< MAX_REG_OUTPUT
; i
++) {
3387 if (!vs
->baseShader
.reg_maps
.packed_output
[i
]) continue;
3389 usage
= semantics_out
[i
].usage
;
3390 usage_idx
= semantics_out
[i
].usage_idx
;
3391 shader_glsl_get_write_mask(&semantics_out
[i
].reg
, reg_mask
);
3394 case WINED3DDECLUSAGE_POSITION
:
3395 shader_addline(&buffer
, "gl_Position%s = OUT[%u]%s;\n", reg_mask
, i
, reg_mask
);
3398 case WINED3DDECLUSAGE_PSIZE
:
3399 shader_addline(&buffer
, "gl_PointSize = OUT[%u].x;\n", i
);
3407 /* Then, fix the pixel shader input */
3408 handle_ps3_input(&buffer
, gl_info
, ps
->input_reg_map
,
3409 ps
->semantics_in
, &ps
->baseShader
.reg_maps
, semantics_out
, &vs
->baseShader
.reg_maps
);
3411 shader_addline(&buffer
, "}\n");
3412 } else if(ps_major
>= 3 && vs_major
< 3) {
3413 shader_addline(&buffer
, "varying vec4 IN[%u];\n", GL_LIMITS(glsl_varyings
) / 4);
3414 shader_addline(&buffer
, "void order_ps_input() {\n");
3415 /* The vertex shader wrote to the builtin varyings. There is no need to figure out position and
3416 * point size, but we depend on the optimizers kindness to find out that the pixel shader doesn't
3417 * read gl_TexCoord and gl_ColorX, otherwise we'll run out of varyings
3419 handle_ps3_input(&buffer
, gl_info
, ps
->input_reg_map
, ps
->semantics_in
, &ps
->baseShader
.reg_maps
, NULL
, NULL
);
3420 shader_addline(&buffer
, "}\n");
3422 ERR("Unexpected vertex and pixel shader version condition: vs: %d, ps: %d\n", vs_major
, ps_major
);
3425 ret
= GL_EXTCALL(glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB
));
3426 checkGLcall("glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB)");
3427 GL_EXTCALL(glShaderSourceARB(ret
, 1, (const char**)&buffer
.buffer
, NULL
));
3428 checkGLcall("glShaderSourceARB(ret, 1, &buffer.buffer, NULL)");
3429 GL_EXTCALL(glCompileShaderARB(ret
));
3430 checkGLcall("glCompileShaderARB(ret)");
3432 shader_buffer_free(&buffer
);
3436 static void hardcode_local_constants(IWineD3DBaseShaderImpl
*shader
, const WineD3D_GL_Info
*gl_info
,
3437 GLhandleARB programId
, char prefix
)
3439 const local_constant
*lconst
;
3444 LIST_FOR_EACH_ENTRY(lconst
, &shader
->baseShader
.constantsF
, local_constant
, entry
) {
3445 value
= (const float *)lconst
->value
;
3446 snprintf(glsl_name
, sizeof(glsl_name
), "%cLC%u", prefix
, lconst
->idx
);
3447 tmp_loc
= GL_EXTCALL(glGetUniformLocationARB(programId
, glsl_name
));
3448 GL_EXTCALL(glUniform4fvARB(tmp_loc
, 1, value
));
3450 checkGLcall("Hardcoding local constants\n");
3453 /** Sets the GLSL program ID for the given pixel and vertex shader combination.
3454 * It sets the programId on the current StateBlock (because it should be called
3455 * inside of the DrawPrimitive() part of the render loop).
3457 * If a program for the given combination does not exist, create one, and store
3458 * the program in the hash table. If it creates a program, it will link the
3459 * given objects, too.
3461 static void set_glsl_shader_program(IWineD3DDevice
*iface
, BOOL use_ps
, BOOL use_vs
) {
3462 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
3463 struct shader_glsl_priv
*priv
= This
->shader_priv
;
3464 const WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
3465 IWineD3DPixelShader
*pshader
= This
->stateBlock
->pixelShader
;
3466 IWineD3DVertexShader
*vshader
= This
->stateBlock
->vertexShader
;
3467 struct glsl_shader_prog_link
*entry
= NULL
;
3468 GLhandleARB programId
= 0;
3469 GLhandleARB reorder_shader_id
= 0;
3472 GLhandleARB vshader_id
, pshader_id
;
3473 struct ps_compile_args ps_compile_args
;
3474 struct vs_compile_args vs_compile_args
;
3477 find_vs_compile_args((IWineD3DVertexShaderImpl
*)This
->stateBlock
->vertexShader
, This
->stateBlock
, &vs_compile_args
);
3479 /* FIXME: Do we really have to spend CPU cycles to generate a few zeroed bytes? */
3480 memset(&vs_compile_args
, 0, sizeof(vs_compile_args
));
3483 find_ps_compile_args((IWineD3DPixelShaderImpl
*)This
->stateBlock
->pixelShader
, This
->stateBlock
, &ps_compile_args
);
3485 /* FIXME: Do we really have to spend CPU cycles to generate a few zeroed bytes? */
3486 memset(&ps_compile_args
, 0, sizeof(ps_compile_args
));
3488 entry
= get_glsl_program_entry(priv
, vshader
, pshader
, &vs_compile_args
, &ps_compile_args
);
3490 priv
->glsl_program
= entry
;
3494 /* If we get to this point, then no matching program exists, so we create one */
3495 programId
= GL_EXTCALL(glCreateProgramObjectARB());
3496 TRACE("Created new GLSL shader program %u\n", programId
);
3498 /* Create the entry */
3499 entry
= HeapAlloc(GetProcessHeap(), 0, sizeof(struct glsl_shader_prog_link
));
3500 entry
->programId
= programId
;
3501 entry
->vshader
= vshader
;
3502 entry
->pshader
= pshader
;
3503 entry
->vs_args
= vs_compile_args
;
3504 entry
->ps_args
= ps_compile_args
;
3505 entry
->constant_version
= 0;
3506 /* Add the hash table entry */
3507 add_glsl_program_entry(priv
, entry
);
3509 /* Set the current program */
3510 priv
->glsl_program
= entry
;
3513 vshader_id
= find_gl_vshader((IWineD3DVertexShaderImpl
*) vshader
, &vs_compile_args
);
3518 /* Attach GLSL vshader */
3520 const unsigned int max_attribs
= 16; /* TODO: Will this always be the case? It is at the moment... */
3523 reorder_shader_id
= generate_param_reorder_function(vshader
, pshader
, gl_info
);
3524 TRACE("Attaching GLSL shader object %u to program %u\n", reorder_shader_id
, programId
);
3525 GL_EXTCALL(glAttachObjectARB(programId
, reorder_shader_id
));
3526 checkGLcall("glAttachObjectARB");
3527 /* Flag the reorder function for deletion, then it will be freed automatically when the program
3530 GL_EXTCALL(glDeleteObjectARB(reorder_shader_id
));
3532 TRACE("Attaching GLSL shader object %u to program %u\n", vshader_id
, programId
);
3533 GL_EXTCALL(glAttachObjectARB(programId
, vshader_id
));
3534 checkGLcall("glAttachObjectARB");
3536 /* Bind vertex attributes to a corresponding index number to match
3537 * the same index numbers as ARB_vertex_programs (makes loading
3538 * vertex attributes simpler). With this method, we can use the
3539 * exact same code to load the attributes later for both ARB and
3542 * We have to do this here because we need to know the Program ID
3543 * in order to make the bindings work, and it has to be done prior
3544 * to linking the GLSL program. */
3545 for (i
= 0; i
< max_attribs
; ++i
) {
3546 if (((IWineD3DBaseShaderImpl
*)vshader
)->baseShader
.reg_maps
.attributes
[i
]) {
3547 snprintf(tmp_name
, sizeof(tmp_name
), "attrib%i", i
);
3548 GL_EXTCALL(glBindAttribLocationARB(programId
, i
, tmp_name
));
3551 checkGLcall("glBindAttribLocationARB");
3553 list_add_head(&((IWineD3DBaseShaderImpl
*)vshader
)->baseShader
.linked_programs
, &entry
->vshader_entry
);
3557 pshader_id
= find_gl_pshader((IWineD3DPixelShaderImpl
*) pshader
, &ps_compile_args
);
3562 /* Attach GLSL pshader */
3564 TRACE("Attaching GLSL shader object %u to program %u\n", pshader_id
, programId
);
3565 GL_EXTCALL(glAttachObjectARB(programId
, pshader_id
));
3566 checkGLcall("glAttachObjectARB");
3568 list_add_head(&((IWineD3DBaseShaderImpl
*)pshader
)->baseShader
.linked_programs
, &entry
->pshader_entry
);
3571 /* Link the program */
3572 TRACE("Linking GLSL shader program %u\n", programId
);
3573 GL_EXTCALL(glLinkProgramARB(programId
));
3574 print_glsl_info_log(&GLINFO_LOCATION
, programId
);
3576 entry
->vuniformF_locations
= HeapAlloc(GetProcessHeap(), 0, sizeof(GLhandleARB
) * GL_LIMITS(vshader_constantsF
));
3577 for (i
= 0; i
< GL_LIMITS(vshader_constantsF
); ++i
) {
3578 snprintf(glsl_name
, sizeof(glsl_name
), "VC[%i]", i
);
3579 entry
->vuniformF_locations
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, glsl_name
));
3581 for (i
= 0; i
< MAX_CONST_I
; ++i
) {
3582 snprintf(glsl_name
, sizeof(glsl_name
), "VI[%i]", i
);
3583 entry
->vuniformI_locations
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, glsl_name
));
3585 entry
->puniformF_locations
= HeapAlloc(GetProcessHeap(), 0, sizeof(GLhandleARB
) * GL_LIMITS(pshader_constantsF
));
3586 for (i
= 0; i
< GL_LIMITS(pshader_constantsF
); ++i
) {
3587 snprintf(glsl_name
, sizeof(glsl_name
), "PC[%i]", i
);
3588 entry
->puniformF_locations
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, glsl_name
));
3590 for (i
= 0; i
< MAX_CONST_I
; ++i
) {
3591 snprintf(glsl_name
, sizeof(glsl_name
), "PI[%i]", i
);
3592 entry
->puniformI_locations
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, glsl_name
));
3596 for(i
= 0; i
< ((IWineD3DPixelShaderImpl
*)pshader
)->numbumpenvmatconsts
; i
++) {
3598 sprintf(name
, "bumpenvmat%d", ((IWineD3DPixelShaderImpl
*)pshader
)->bumpenvmatconst
[i
].texunit
);
3599 entry
->bumpenvmat_location
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, name
));
3600 sprintf(name
, "luminancescale%d", ((IWineD3DPixelShaderImpl
*)pshader
)->luminanceconst
[i
].texunit
);
3601 entry
->luminancescale_location
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, name
));
3602 sprintf(name
, "luminanceoffset%d", ((IWineD3DPixelShaderImpl
*)pshader
)->luminanceconst
[i
].texunit
);
3603 entry
->luminanceoffset_location
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, name
));
3607 if (use_ps
&& ps_compile_args
.np2_fixup
) {
3609 for (i
= 0; i
< MAX_FRAGMENT_SAMPLERS
; ++i
) {
3610 if (ps_compile_args
.np2_fixup
& (1 << i
)) {
3611 sprintf(name
, "PsamplerNP2Fixup%u", i
);
3612 entry
->np2Fixup_location
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, name
));
3614 entry
->np2Fixup_location
[i
] = -1;
3619 entry
->posFixup_location
= GL_EXTCALL(glGetUniformLocationARB(programId
, "posFixup"));
3620 entry
->ycorrection_location
= GL_EXTCALL(glGetUniformLocationARB(programId
, "ycorrection"));
3621 checkGLcall("Find glsl program uniform locations");
3624 && WINED3DSHADER_VERSION_MAJOR(((IWineD3DPixelShaderImpl
*)pshader
)->baseShader
.reg_maps
.shader_version
) >= 3
3625 && ((IWineD3DPixelShaderImpl
*)pshader
)->declared_in_count
> GL_LIMITS(glsl_varyings
) / 4)
3627 TRACE("Shader %d needs vertex color clamping disabled\n", programId
);
3628 entry
->vertex_color_clamp
= GL_FALSE
;
3630 entry
->vertex_color_clamp
= GL_FIXED_ONLY_ARB
;
3633 /* Set the shader to allow uniform loading on it */
3634 GL_EXTCALL(glUseProgramObjectARB(programId
));
3635 checkGLcall("glUseProgramObjectARB(programId)");
3637 /* Load the vertex and pixel samplers now. The function that finds the mappings makes sure
3638 * that it stays the same for each vertexshader-pixelshader pair(=linked glsl program). If
3639 * a pshader with fixed function pipeline is used there are no vertex samplers, and if a
3640 * vertex shader with fixed function pixel processing is used we make sure that the card
3641 * supports enough samplers to allow the max number of vertex samplers with all possible
3642 * fixed function fragment processing setups. So once the program is linked these samplers
3646 /* Load vertex shader samplers */
3647 shader_glsl_load_vsamplers(gl_info
, This
->texUnitMap
, programId
);
3650 /* Load pixel shader samplers */
3651 shader_glsl_load_psamplers(gl_info
, This
->texUnitMap
, programId
);
3654 /* If the local constants do not have to be loaded with the environment constants,
3655 * load them now to have them hardcoded in the GLSL program. This saves some CPU cycles
3658 if(pshader
&& !((IWineD3DPixelShaderImpl
*)pshader
)->baseShader
.load_local_constsF
) {
3659 hardcode_local_constants((IWineD3DBaseShaderImpl
*) pshader
, gl_info
, programId
, 'P');
3661 if(vshader
&& !((IWineD3DVertexShaderImpl
*)vshader
)->baseShader
.load_local_constsF
) {
3662 hardcode_local_constants((IWineD3DBaseShaderImpl
*) vshader
, gl_info
, programId
, 'V');
3666 static GLhandleARB
create_glsl_blt_shader(const WineD3D_GL_Info
*gl_info
, enum tex_types tex_type
)
3668 GLhandleARB program_id
;
3669 GLhandleARB vshader_id
, pshader_id
;
3670 static const char *blt_vshader
[] =
3675 " gl_Position = gl_Vertex;\n"
3676 " gl_FrontColor = vec4(1.0);\n"
3677 " gl_TexCoord[0] = gl_MultiTexCoord0;\n"
3681 static const char *blt_pshaders
[tex_type_count
] =
3687 "uniform sampler2D sampler;\n"
3690 " gl_FragDepth = texture2D(sampler, gl_TexCoord[0].xy).x;\n"
3696 "uniform samplerCube sampler;\n"
3699 " gl_FragDepth = textureCube(sampler, gl_TexCoord[0].xyz).x;\n"
3703 "#extension GL_ARB_texture_rectangle : enable\n"
3704 "uniform sampler2DRect sampler;\n"
3707 " gl_FragDepth = texture2DRect(sampler, gl_TexCoord[0].xy).x;\n"
3711 if (!blt_pshaders
[tex_type
])
3713 FIXME("tex_type %#x not supported\n", tex_type
);
3717 vshader_id
= GL_EXTCALL(glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB
));
3718 GL_EXTCALL(glShaderSourceARB(vshader_id
, 1, blt_vshader
, NULL
));
3719 GL_EXTCALL(glCompileShaderARB(vshader_id
));
3721 pshader_id
= GL_EXTCALL(glCreateShaderObjectARB(GL_FRAGMENT_SHADER_ARB
));
3722 GL_EXTCALL(glShaderSourceARB(pshader_id
, 1, &blt_pshaders
[tex_type
], NULL
));
3723 GL_EXTCALL(glCompileShaderARB(pshader_id
));
3725 program_id
= GL_EXTCALL(glCreateProgramObjectARB());
3726 GL_EXTCALL(glAttachObjectARB(program_id
, vshader_id
));
3727 GL_EXTCALL(glAttachObjectARB(program_id
, pshader_id
));
3728 GL_EXTCALL(glLinkProgramARB(program_id
));
3730 print_glsl_info_log(&GLINFO_LOCATION
, program_id
);
3732 /* Once linked we can mark the shaders for deletion. They will be deleted once the program
3735 GL_EXTCALL(glDeleteObjectARB(vshader_id
));
3736 GL_EXTCALL(glDeleteObjectARB(pshader_id
));
3740 static void shader_glsl_select(IWineD3DDevice
*iface
, BOOL usePS
, BOOL useVS
) {
3741 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
3742 struct shader_glsl_priv
*priv
= This
->shader_priv
;
3743 const WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
3744 GLhandleARB program_id
= 0;
3745 GLenum old_vertex_color_clamp
, current_vertex_color_clamp
;
3747 old_vertex_color_clamp
= priv
->glsl_program
? priv
->glsl_program
->vertex_color_clamp
: GL_FIXED_ONLY_ARB
;
3749 if (useVS
|| usePS
) set_glsl_shader_program(iface
, usePS
, useVS
);
3750 else priv
->glsl_program
= NULL
;
3752 current_vertex_color_clamp
= priv
->glsl_program
? priv
->glsl_program
->vertex_color_clamp
: GL_FIXED_ONLY_ARB
;
3754 if (old_vertex_color_clamp
!= current_vertex_color_clamp
) {
3755 if (GL_SUPPORT(ARB_COLOR_BUFFER_FLOAT
)) {
3756 GL_EXTCALL(glClampColorARB(GL_CLAMP_VERTEX_COLOR_ARB
, current_vertex_color_clamp
));
3757 checkGLcall("glClampColorARB");
3759 FIXME("vertex color clamp needs to be changed, but extension not supported.\n");
3763 program_id
= priv
->glsl_program
? priv
->glsl_program
->programId
: 0;
3764 if (program_id
) TRACE("Using GLSL program %u\n", program_id
);
3765 GL_EXTCALL(glUseProgramObjectARB(program_id
));
3766 checkGLcall("glUseProgramObjectARB");
3769 static void shader_glsl_select_depth_blt(IWineD3DDevice
*iface
, enum tex_types tex_type
) {
3770 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
3771 const WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
3772 struct shader_glsl_priv
*priv
= This
->shader_priv
;
3773 GLhandleARB
*blt_program
= &priv
->depth_blt_program
[tex_type
];
3775 if (!*blt_program
) {
3777 *blt_program
= create_glsl_blt_shader(gl_info
, tex_type
);
3778 loc
= GL_EXTCALL(glGetUniformLocationARB(*blt_program
, "sampler"));
3779 GL_EXTCALL(glUseProgramObjectARB(*blt_program
));
3780 GL_EXTCALL(glUniform1iARB(loc
, 0));
3782 GL_EXTCALL(glUseProgramObjectARB(*blt_program
));
3786 static void shader_glsl_deselect_depth_blt(IWineD3DDevice
*iface
) {
3787 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
3788 const WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
3789 struct shader_glsl_priv
*priv
= This
->shader_priv
;
3790 GLhandleARB program_id
;
3792 program_id
= priv
->glsl_program
? priv
->glsl_program
->programId
: 0;
3793 if (program_id
) TRACE("Using GLSL program %u\n", program_id
);
3795 GL_EXTCALL(glUseProgramObjectARB(program_id
));
3796 checkGLcall("glUseProgramObjectARB");
3799 static void shader_glsl_destroy(IWineD3DBaseShader
*iface
) {
3800 const struct list
*linked_programs
;
3801 IWineD3DBaseShaderImpl
*This
= (IWineD3DBaseShaderImpl
*) iface
;
3802 IWineD3DDeviceImpl
*device
= (IWineD3DDeviceImpl
*)This
->baseShader
.device
;
3803 struct shader_glsl_priv
*priv
= device
->shader_priv
;
3804 const WineD3D_GL_Info
*gl_info
= &device
->adapter
->gl_info
;
3805 IWineD3DPixelShaderImpl
*ps
= NULL
;
3806 IWineD3DVertexShaderImpl
*vs
= NULL
;
3808 /* Note: Do not use QueryInterface here to find out which shader type this is because this code
3809 * can be called from IWineD3DBaseShader::Release
3811 char pshader
= shader_is_pshader_version(This
->baseShader
.reg_maps
.shader_version
);
3814 ps
= (IWineD3DPixelShaderImpl
*) This
;
3815 if(ps
->num_gl_shaders
== 0) return;
3816 if (priv
->glsl_program
&& (IWineD3DBaseShader
*)priv
->glsl_program
->pshader
== iface
)
3817 shader_glsl_select(This
->baseShader
.device
, FALSE
, FALSE
);
3819 vs
= (IWineD3DVertexShaderImpl
*) This
;
3820 if(vs
->num_gl_shaders
== 0) return;
3821 if (priv
->glsl_program
&& (IWineD3DBaseShader
*)priv
->glsl_program
->vshader
== iface
)
3822 shader_glsl_select(This
->baseShader
.device
, FALSE
, FALSE
);
3825 linked_programs
= &This
->baseShader
.linked_programs
;
3827 TRACE("Deleting linked programs\n");
3828 if (linked_programs
->next
) {
3829 struct glsl_shader_prog_link
*entry
, *entry2
;
3832 LIST_FOR_EACH_ENTRY_SAFE(entry
, entry2
, linked_programs
, struct glsl_shader_prog_link
, pshader_entry
) {
3833 delete_glsl_program_entry(priv
, gl_info
, entry
);
3836 LIST_FOR_EACH_ENTRY_SAFE(entry
, entry2
, linked_programs
, struct glsl_shader_prog_link
, vshader_entry
) {
3837 delete_glsl_program_entry(priv
, gl_info
, entry
);
3846 for(i
= 0; i
< ps
->num_gl_shaders
; i
++) {
3847 TRACE("deleting pshader %u\n", ps
->gl_shaders
[i
].prgId
);
3848 GL_EXTCALL(glDeleteObjectARB(ps
->gl_shaders
[i
].prgId
));
3849 checkGLcall("glDeleteObjectARB");
3852 HeapFree(GetProcessHeap(), 0, ps
->gl_shaders
);
3853 ps
->gl_shaders
= NULL
;
3854 ps
->num_gl_shaders
= 0;
3855 ps
->shader_array_size
= 0;
3860 for(i
= 0; i
< vs
->num_gl_shaders
; i
++) {
3861 TRACE("deleting vshader %u\n", vs
->gl_shaders
[i
].prgId
);
3862 GL_EXTCALL(glDeleteObjectARB(vs
->gl_shaders
[i
].prgId
));
3863 checkGLcall("glDeleteObjectARB");
3866 HeapFree(GetProcessHeap(), 0, vs
->gl_shaders
);
3867 vs
->gl_shaders
= NULL
;
3868 vs
->num_gl_shaders
= 0;
3869 vs
->shader_array_size
= 0;
3873 static unsigned int glsl_program_key_hash(const void *key
)
3875 const glsl_program_key_t
*k
= key
;
3877 unsigned int hash
= ((DWORD_PTR
) k
->vshader
) | ((DWORD_PTR
) k
->pshader
) << 16;
3878 hash
+= ~(hash
<< 15);
3879 hash
^= (hash
>> 10);
3880 hash
+= (hash
<< 3);
3881 hash
^= (hash
>> 6);
3882 hash
+= ~(hash
<< 11);
3883 hash
^= (hash
>> 16);
3888 static BOOL
glsl_program_key_compare(const void *keya
, const void *keyb
)
3890 const glsl_program_key_t
*ka
= keya
;
3891 const glsl_program_key_t
*kb
= keyb
;
3893 return ka
->vshader
== kb
->vshader
&& ka
->pshader
== kb
->pshader
&&
3894 (memcmp(&ka
->ps_args
, &kb
->ps_args
, sizeof(kb
->ps_args
)) == 0) &&
3895 (memcmp(&ka
->vs_args
, &kb
->vs_args
, sizeof(kb
->vs_args
)) == 0);
3898 static BOOL
constant_heap_init(struct constant_heap
*heap
, unsigned int constant_count
)
3900 SIZE_T size
= (constant_count
+ 1) * sizeof(*heap
->entries
) + constant_count
* sizeof(*heap
->positions
);
3901 void *mem
= HeapAlloc(GetProcessHeap(), 0, size
);
3905 ERR("Failed to allocate memory\n");
3909 heap
->entries
= mem
;
3910 heap
->entries
[1].version
= 0;
3911 heap
->positions
= (unsigned int *)(heap
->entries
+ constant_count
+ 1);
3917 static void constant_heap_free(struct constant_heap
*heap
)
3919 HeapFree(GetProcessHeap(), 0, heap
->entries
);
3922 static HRESULT
shader_glsl_alloc(IWineD3DDevice
*iface
) {
3923 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
3924 const WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
3925 struct shader_glsl_priv
*priv
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(struct shader_glsl_priv
));
3926 SIZE_T stack_size
= wined3d_log2i(max(GL_LIMITS(vshader_constantsF
), GL_LIMITS(pshader_constantsF
))) + 1;
3928 priv
->stack
= HeapAlloc(GetProcessHeap(), 0, stack_size
* sizeof(*priv
->stack
));
3931 ERR("Failed to allocate memory.\n");
3932 HeapFree(GetProcessHeap(), 0, priv
);
3933 return E_OUTOFMEMORY
;
3936 if (!constant_heap_init(&priv
->vconst_heap
, GL_LIMITS(vshader_constantsF
)))
3938 ERR("Failed to initialize vertex shader constant heap\n");
3939 HeapFree(GetProcessHeap(), 0, priv
->stack
);
3940 HeapFree(GetProcessHeap(), 0, priv
);
3941 return E_OUTOFMEMORY
;
3944 if (!constant_heap_init(&priv
->pconst_heap
, GL_LIMITS(pshader_constantsF
)))
3946 ERR("Failed to initialize pixel shader constant heap\n");
3947 constant_heap_free(&priv
->vconst_heap
);
3948 HeapFree(GetProcessHeap(), 0, priv
->stack
);
3949 HeapFree(GetProcessHeap(), 0, priv
);
3950 return E_OUTOFMEMORY
;
3953 priv
->glsl_program_lookup
= hash_table_create(glsl_program_key_hash
, glsl_program_key_compare
);
3954 priv
->next_constant_version
= 1;
3956 This
->shader_priv
= priv
;
3960 static void shader_glsl_free(IWineD3DDevice
*iface
) {
3961 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
3962 const WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
3963 struct shader_glsl_priv
*priv
= This
->shader_priv
;
3966 for (i
= 0; i
< tex_type_count
; ++i
)
3968 if (priv
->depth_blt_program
[i
])
3970 GL_EXTCALL(glDeleteObjectARB(priv
->depth_blt_program
[i
]));
3974 hash_table_destroy(priv
->glsl_program_lookup
, NULL
, NULL
);
3975 constant_heap_free(&priv
->pconst_heap
);
3976 constant_heap_free(&priv
->vconst_heap
);
3978 HeapFree(GetProcessHeap(), 0, This
->shader_priv
);
3979 This
->shader_priv
= NULL
;
3982 static BOOL
shader_glsl_dirty_const(IWineD3DDevice
*iface
) {
3983 /* TODO: GL_EXT_bindable_uniform can be used to share constants across shaders */
3987 static GLuint
shader_glsl_generate_pshader(IWineD3DPixelShader
*iface
, SHADER_BUFFER
*buffer
, const struct ps_compile_args
*args
) {
3988 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*)iface
;
3989 const struct shader_reg_maps
*reg_maps
= &This
->baseShader
.reg_maps
;
3990 CONST DWORD
*function
= This
->baseShader
.function
;
3991 const char *fragcolor
;
3992 const WineD3D_GL_Info
*gl_info
= &((IWineD3DDeviceImpl
*)This
->baseShader
.device
)->adapter
->gl_info
;
3994 /* Create the hw GLSL shader object and assign it as the shader->prgId */
3995 GLhandleARB shader_obj
= GL_EXTCALL(glCreateShaderObjectARB(GL_FRAGMENT_SHADER_ARB
));
3997 shader_addline(buffer
, "#version 120\n");
3999 if (GL_SUPPORT(ARB_DRAW_BUFFERS
)) {
4000 shader_addline(buffer
, "#extension GL_ARB_draw_buffers : enable\n");
4002 if (GL_SUPPORT(ARB_TEXTURE_RECTANGLE
)) {
4003 /* The spec says that it doesn't have to be explicitly enabled, but the nvidia
4004 * drivers write a warning if we don't do so
4006 shader_addline(buffer
, "#extension GL_ARB_texture_rectangle : enable\n");
4009 /* Base Declarations */
4010 shader_generate_glsl_declarations( (IWineD3DBaseShader
*) This
, reg_maps
, buffer
, &GLINFO_LOCATION
, args
);
4012 /* Pack 3.0 inputs */
4013 if (reg_maps
->shader_version
>= WINED3DPS_VERSION(3,0) && args
->vp_mode
!= vertexshader
) {
4014 pshader_glsl_input_pack(iface
, buffer
, This
->semantics_in
, reg_maps
, args
->vp_mode
);
4017 /* Base Shader Body */
4018 shader_generate_main( (IWineD3DBaseShader
*) This
, buffer
, reg_maps
, function
);
4020 /* Pixel shaders < 2.0 place the resulting color in R0 implicitly */
4021 if (reg_maps
->shader_version
< WINED3DPS_VERSION(2,0))
4023 /* Some older cards like GeforceFX ones don't support multiple buffers, so also not gl_FragData */
4024 if(GL_SUPPORT(ARB_DRAW_BUFFERS
))
4025 shader_addline(buffer
, "gl_FragData[0] = R0;\n");
4027 shader_addline(buffer
, "gl_FragColor = R0;\n");
4030 if(GL_SUPPORT(ARB_DRAW_BUFFERS
)) {
4031 fragcolor
= "gl_FragData[0]";
4033 fragcolor
= "gl_FragColor";
4035 if(args
->srgb_correction
) {
4036 shader_addline(buffer
, "tmp0.xyz = pow(%s.xyz, vec3(%f, %f, %f)) * vec3(%f, %f, %f) - vec3(%f, %f, %f);\n",
4037 fragcolor
, srgb_pow
, srgb_pow
, srgb_pow
, srgb_mul_high
, srgb_mul_high
, srgb_mul_high
,
4038 srgb_sub_high
, srgb_sub_high
, srgb_sub_high
);
4039 shader_addline(buffer
, "tmp1.xyz = %s.xyz * srgb_mul_low.xyz;\n", fragcolor
);
4040 shader_addline(buffer
, "%s.x = %s.x < srgb_comparison.x ? tmp1.x : tmp0.x;\n", fragcolor
, fragcolor
);
4041 shader_addline(buffer
, "%s.y = %s.y < srgb_comparison.y ? tmp1.y : tmp0.y;\n", fragcolor
, fragcolor
);
4042 shader_addline(buffer
, "%s.z = %s.z < srgb_comparison.z ? tmp1.z : tmp0.z;\n", fragcolor
, fragcolor
);
4043 shader_addline(buffer
, "%s = clamp(%s, 0.0, 1.0);\n", fragcolor
, fragcolor
);
4045 /* Pixel shader < 3.0 do not replace the fog stage.
4046 * This implements linear fog computation and blending.
4047 * TODO: non linear fog
4048 * NOTE: gl_Fog.start and gl_Fog.end don't hold fog start s and end e but
4049 * -1/(e-s) and e/(e-s) respectively.
4051 if(reg_maps
->shader_version
< WINED3DPS_VERSION(3,0)) {
4053 case FOG_OFF
: break;
4055 shader_addline(buffer
, "float fogstart = -1.0 / (gl_Fog.end - gl_Fog.start);\n");
4056 shader_addline(buffer
, "float fogend = gl_Fog.end * -fogstart;\n");
4057 shader_addline(buffer
, "float Fog = clamp(gl_FogFragCoord * fogstart + fogend, 0.0, 1.0);\n");
4058 shader_addline(buffer
, "%s.xyz = mix(gl_Fog.color.xyz, %s.xyz, Fog);\n", fragcolor
, fragcolor
);
4061 /* Fog = e^(-gl_Fog.density * gl_FogFragCoord) */
4062 shader_addline(buffer
, "float Fog = exp(-gl_Fog.density * gl_FogFragCoord);\n");
4063 shader_addline(buffer
, "Fog = clamp(Fog, 0.0, 1.0);\n");
4064 shader_addline(buffer
, "%s.xyz = mix(gl_Fog.color.xyz, %s.xyz, Fog);\n", fragcolor
, fragcolor
);
4067 /* Fog = e^(-(gl_Fog.density * gl_FogFragCoord)^2) */
4068 shader_addline(buffer
, "float Fog = exp(-gl_Fog.density * gl_Fog.density * gl_FogFragCoord * gl_FogFragCoord);\n");
4069 shader_addline(buffer
, "Fog = clamp(Fog, 0.0, 1.0);\n");
4070 shader_addline(buffer
, "%s.xyz = mix(gl_Fog.color.xyz, %s.xyz, Fog);\n", fragcolor
, fragcolor
);
4075 shader_addline(buffer
, "}\n");
4077 TRACE("Compiling shader object %u\n", shader_obj
);
4078 GL_EXTCALL(glShaderSourceARB(shader_obj
, 1, (const char**)&buffer
->buffer
, NULL
));
4079 GL_EXTCALL(glCompileShaderARB(shader_obj
));
4080 print_glsl_info_log(&GLINFO_LOCATION
, shader_obj
);
4082 /* Store the shader object */
4086 static GLuint
shader_glsl_generate_vshader(IWineD3DVertexShader
*iface
, SHADER_BUFFER
*buffer
, const struct vs_compile_args
*args
) {
4087 IWineD3DVertexShaderImpl
*This
= (IWineD3DVertexShaderImpl
*)iface
;
4088 const struct shader_reg_maps
*reg_maps
= &This
->baseShader
.reg_maps
;
4089 CONST DWORD
*function
= This
->baseShader
.function
;
4090 const WineD3D_GL_Info
*gl_info
= &((IWineD3DDeviceImpl
*)This
->baseShader
.device
)->adapter
->gl_info
;
4092 /* Create the hw GLSL shader program and assign it as the shader->prgId */
4093 GLhandleARB shader_obj
= GL_EXTCALL(glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB
));
4095 shader_addline(buffer
, "#version 120\n");
4097 /* Base Declarations */
4098 shader_generate_glsl_declarations( (IWineD3DBaseShader
*) This
, reg_maps
, buffer
, &GLINFO_LOCATION
, NULL
);
4100 /* Base Shader Body */
4101 shader_generate_main( (IWineD3DBaseShader
*) This
, buffer
, reg_maps
, function
);
4103 /* Unpack 3.0 outputs */
4104 if (reg_maps
->shader_version
>= WINED3DVS_VERSION(3,0)) shader_addline(buffer
, "order_ps_input(OUT);\n");
4105 else shader_addline(buffer
, "order_ps_input();\n");
4107 /* The D3DRS_FOGTABLEMODE render state defines if the shader-generated fog coord is used
4108 * or if the fragment depth is used. If the fragment depth is used(FOGTABLEMODE != NONE),
4109 * the fog frag coord is thrown away. If the fog frag coord is used, but not written by
4110 * the shader, it is set to 0.0(fully fogged, since start = 1.0, end = 0.0)
4112 if(args
->fog_src
== VS_FOG_Z
) {
4113 shader_addline(buffer
, "gl_FogFragCoord = gl_Position.z;\n");
4114 } else if (!reg_maps
->fog
) {
4115 shader_addline(buffer
, "gl_FogFragCoord = 0.0;\n");
4118 /* Write the final position.
4120 * OpenGL coordinates specify the center of the pixel while d3d coords specify
4121 * the corner. The offsets are stored in z and w in posFixup. posFixup.y contains
4122 * 1.0 or -1.0 to turn the rendering upside down for offscreen rendering. PosFixup.x
4123 * contains 1.0 to allow a mad.
4125 shader_addline(buffer
, "gl_Position.y = gl_Position.y * posFixup.y;\n");
4126 shader_addline(buffer
, "gl_Position.xy += posFixup.zw * gl_Position.ww;\n");
4128 /* Z coord [0;1]->[-1;1] mapping, see comment in transform_projection in state.c
4130 * Basically we want (in homogeneous coordinates) z = z * 2 - 1. However, shaders are run
4131 * before the homogeneous divide, so we have to take the w into account: z = ((z / w) * 2 - 1) * w,
4132 * which is the same as z = z * 2 - w.
4134 shader_addline(buffer
, "gl_Position.z = gl_Position.z * 2.0 - gl_Position.w;\n");
4136 shader_addline(buffer
, "}\n");
4138 TRACE("Compiling shader object %u\n", shader_obj
);
4139 GL_EXTCALL(glShaderSourceARB(shader_obj
, 1, (const char**)&buffer
->buffer
, NULL
));
4140 GL_EXTCALL(glCompileShaderARB(shader_obj
));
4141 print_glsl_info_log(&GLINFO_LOCATION
, shader_obj
);
4146 static void shader_glsl_get_caps(WINED3DDEVTYPE devtype
, const WineD3D_GL_Info
*gl_info
, struct shader_caps
*pCaps
)
4148 /* Nvidia Geforce6/7 or Ati R4xx/R5xx cards with GLSL support, support VS 3.0 but older Nvidia/Ati
4149 * models with GLSL support only support 2.0. In case of nvidia we can detect VS 2.0 support using
4150 * vs_nv_version which is based on NV_vertex_program.
4151 * For Ati cards there's no way using glsl (it abstracts the lowlevel info away) and also not
4152 * using ARB_vertex_program. It is safe to assume that when a card supports pixel shader 2.0 it
4153 * supports vertex shader 2.0 too and the way around. We can detect ps2.0 using the maximum number
4154 * of native instructions, so use that here. For more info see the pixel shader versioning code below.
4156 if((GLINFO_LOCATION
.vs_nv_version
== VS_VERSION_20
) || (GLINFO_LOCATION
.ps_arb_max_instructions
<= 512))
4157 pCaps
->VertexShaderVersion
= WINED3DVS_VERSION(2,0);
4159 pCaps
->VertexShaderVersion
= WINED3DVS_VERSION(3,0);
4160 TRACE_(d3d_caps
)("Hardware vertex shader version %d.%d enabled (GLSL)\n", (pCaps
->VertexShaderVersion
>> 8) & 0xff, pCaps
->VertexShaderVersion
& 0xff);
4161 pCaps
->MaxVertexShaderConst
= GL_LIMITS(vshader_constantsF
);
4163 /* Older DX9-class videocards (GeforceFX / Radeon >9500/X*00) only support pixel shader 2.0/2.0a/2.0b.
4164 * In OpenGL the extensions related to GLSL abstract lowlevel GL info away which is needed
4165 * to distinguish between 2.0 and 3.0 (and 2.0a/2.0b). In case of Nvidia we use their fragment
4166 * program extensions. On other hardware including ATI GL_ARB_fragment_program offers the info
4167 * in max native instructions. Intel and others also offer the info in this extension but they
4168 * don't support GLSL (at least on Windows).
4170 * PS2.0 requires at least 96 instructions, 2.0a/2.0b go up to 512. Assume that if the number
4171 * of instructions is 512 or less we have to do with ps2.0 hardware.
4172 * 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.
4174 if((GLINFO_LOCATION
.ps_nv_version
== PS_VERSION_20
) || (GLINFO_LOCATION
.ps_arb_max_instructions
<= 512))
4175 pCaps
->PixelShaderVersion
= WINED3DPS_VERSION(2,0);
4177 pCaps
->PixelShaderVersion
= WINED3DPS_VERSION(3,0);
4179 pCaps
->MaxPixelShaderConst
= GL_LIMITS(pshader_constantsF
);
4181 /* FIXME: The following line is card dependent. -8.0 to 8.0 is the
4182 * Direct3D minimum requirement.
4184 * Both GL_ARB_fragment_program and GLSL require a "maximum representable magnitude"
4185 * of colors to be 2^10, and 2^32 for other floats. Should we use 1024 here?
4187 * The problem is that the refrast clamps temporary results in the shader to
4188 * [-MaxValue;+MaxValue]. If the card's max value is bigger than the one we advertize here,
4189 * then applications may miss the clamping behavior. On the other hand, if it is smaller,
4190 * the shader will generate incorrect results too. Unfortunately, GL deliberately doesn't
4191 * offer a way to query this.
4193 pCaps
->PixelShader1xMaxValue
= 8.0;
4194 TRACE_(d3d_caps
)("Hardware pixel shader version %d.%d enabled (GLSL)\n", (pCaps
->PixelShaderVersion
>> 8) & 0xff, pCaps
->PixelShaderVersion
& 0xff);
4197 static BOOL
shader_glsl_color_fixup_supported(struct color_fixup_desc fixup
)
4199 if (TRACE_ON(d3d_shader
) && TRACE_ON(d3d
))
4201 TRACE("Checking support for fixup:\n");
4202 dump_color_fixup_desc(fixup
);
4205 /* We support everything except YUV conversions. */
4206 if (!is_yuv_fixup(fixup
))
4212 TRACE("[FAILED]\n");
4216 static const SHADER_HANDLER shader_glsl_instruction_handler_table
[WINED3DSIH_TABLE_SIZE
] =
4218 /* WINED3DSIH_ABS */ shader_glsl_map2gl
,
4219 /* WINED3DSIH_ADD */ shader_glsl_arith
,
4220 /* WINED3DSIH_BEM */ pshader_glsl_bem
,
4221 /* WINED3DSIH_BREAK */ shader_glsl_break
,
4222 /* WINED3DSIH_BREAKC */ shader_glsl_breakc
,
4223 /* WINED3DSIH_BREAKP */ NULL
,
4224 /* WINED3DSIH_CALL */ shader_glsl_call
,
4225 /* WINED3DSIH_CALLNZ */ shader_glsl_callnz
,
4226 /* WINED3DSIH_CMP */ shader_glsl_cmp
,
4227 /* WINED3DSIH_CND */ shader_glsl_cnd
,
4228 /* WINED3DSIH_CRS */ shader_glsl_cross
,
4229 /* WINED3DSIH_DCL */ NULL
,
4230 /* WINED3DSIH_DEF */ NULL
,
4231 /* WINED3DSIH_DEFB */ NULL
,
4232 /* WINED3DSIH_DEFI */ NULL
,
4233 /* WINED3DSIH_DP2ADD */ pshader_glsl_dp2add
,
4234 /* WINED3DSIH_DP3 */ shader_glsl_dot
,
4235 /* WINED3DSIH_DP4 */ shader_glsl_dot
,
4236 /* WINED3DSIH_DST */ shader_glsl_dst
,
4237 /* WINED3DSIH_DSX */ shader_glsl_map2gl
,
4238 /* WINED3DSIH_DSY */ shader_glsl_map2gl
,
4239 /* WINED3DSIH_ELSE */ shader_glsl_else
,
4240 /* WINED3DSIH_ENDIF */ shader_glsl_end
,
4241 /* WINED3DSIH_ENDLOOP */ shader_glsl_end
,
4242 /* WINED3DSIH_ENDREP */ shader_glsl_end
,
4243 /* WINED3DSIH_EXP */ shader_glsl_map2gl
,
4244 /* WINED3DSIH_EXPP */ shader_glsl_expp
,
4245 /* WINED3DSIH_FRC */ shader_glsl_map2gl
,
4246 /* WINED3DSIH_IF */ shader_glsl_if
,
4247 /* WINED3DSIH_IFC */ shader_glsl_ifc
,
4248 /* WINED3DSIH_LABEL */ shader_glsl_label
,
4249 /* WINED3DSIH_LIT */ shader_glsl_lit
,
4250 /* WINED3DSIH_LOG */ shader_glsl_log
,
4251 /* WINED3DSIH_LOGP */ shader_glsl_log
,
4252 /* WINED3DSIH_LOOP */ shader_glsl_loop
,
4253 /* WINED3DSIH_LRP */ shader_glsl_lrp
,
4254 /* WINED3DSIH_M3x2 */ shader_glsl_mnxn
,
4255 /* WINED3DSIH_M3x3 */ shader_glsl_mnxn
,
4256 /* WINED3DSIH_M3x4 */ shader_glsl_mnxn
,
4257 /* WINED3DSIH_M4x3 */ shader_glsl_mnxn
,
4258 /* WINED3DSIH_M4x4 */ shader_glsl_mnxn
,
4259 /* WINED3DSIH_MAD */ shader_glsl_mad
,
4260 /* WINED3DSIH_MAX */ shader_glsl_map2gl
,
4261 /* WINED3DSIH_MIN */ shader_glsl_map2gl
,
4262 /* WINED3DSIH_MOV */ shader_glsl_mov
,
4263 /* WINED3DSIH_MOVA */ shader_glsl_mov
,
4264 /* WINED3DSIH_MUL */ shader_glsl_arith
,
4265 /* WINED3DSIH_NOP */ NULL
,
4266 /* WINED3DSIH_NRM */ shader_glsl_map2gl
,
4267 /* WINED3DSIH_PHASE */ NULL
,
4268 /* WINED3DSIH_POW */ shader_glsl_pow
,
4269 /* WINED3DSIH_RCP */ shader_glsl_rcp
,
4270 /* WINED3DSIH_REP */ shader_glsl_rep
,
4271 /* WINED3DSIH_RET */ NULL
,
4272 /* WINED3DSIH_RSQ */ shader_glsl_rsq
,
4273 /* WINED3DSIH_SETP */ NULL
,
4274 /* WINED3DSIH_SGE */ shader_glsl_compare
,
4275 /* WINED3DSIH_SGN */ shader_glsl_map2gl
,
4276 /* WINED3DSIH_SINCOS */ shader_glsl_sincos
,
4277 /* WINED3DSIH_SLT */ shader_glsl_compare
,
4278 /* WINED3DSIH_SUB */ shader_glsl_arith
,
4279 /* WINED3DSIH_TEX */ pshader_glsl_tex
,
4280 /* WINED3DSIH_TEXBEM */ pshader_glsl_texbem
,
4281 /* WINED3DSIH_TEXBEML */ pshader_glsl_texbem
,
4282 /* WINED3DSIH_TEXCOORD */ pshader_glsl_texcoord
,
4283 /* WINED3DSIH_TEXDEPTH */ pshader_glsl_texdepth
,
4284 /* WINED3DSIH_TEXDP3 */ pshader_glsl_texdp3
,
4285 /* WINED3DSIH_TEXDP3TEX */ pshader_glsl_texdp3tex
,
4286 /* WINED3DSIH_TEXKILL */ pshader_glsl_texkill
,
4287 /* WINED3DSIH_TEXLDD */ NULL
,
4288 /* WINED3DSIH_TEXLDL */ shader_glsl_texldl
,
4289 /* WINED3DSIH_TEXM3x2DEPTH */ pshader_glsl_texm3x2depth
,
4290 /* WINED3DSIH_TEXM3x2PAD */ pshader_glsl_texm3x2pad
,
4291 /* WINED3DSIH_TEXM3x2TEX */ pshader_glsl_texm3x2tex
,
4292 /* WINED3DSIH_TEXM3x3 */ pshader_glsl_texm3x3
,
4293 /* WINED3DSIH_TEXM3x3DIFF */ NULL
,
4294 /* WINED3DSIH_TEXM3x3PAD */ pshader_glsl_texm3x3pad
,
4295 /* WINED3DSIH_TEXM3x3SPEC */ pshader_glsl_texm3x3spec
,
4296 /* WINED3DSIH_TEXM3x3TEX */ pshader_glsl_texm3x3tex
,
4297 /* WINED3DSIH_TEXM3x3VSPEC */ pshader_glsl_texm3x3vspec
,
4298 /* WINED3DSIH_TEXREG2AR */ pshader_glsl_texreg2ar
,
4299 /* WINED3DSIH_TEXREG2GB */ pshader_glsl_texreg2gb
,
4300 /* WINED3DSIH_TEXREG2RGB */ pshader_glsl_texreg2rgb
,
4303 const shader_backend_t glsl_shader_backend
= {
4304 shader_glsl_instruction_handler_table
,
4306 shader_glsl_select_depth_blt
,
4307 shader_glsl_deselect_depth_blt
,
4308 shader_glsl_update_float_vertex_constants
,
4309 shader_glsl_update_float_pixel_constants
,
4310 shader_glsl_load_constants
,
4311 shader_glsl_load_np2fixup_constants
,
4312 shader_glsl_destroy
,
4315 shader_glsl_dirty_const
,
4316 shader_glsl_generate_pshader
,
4317 shader_glsl_generate_vshader
,
4318 shader_glsl_get_caps
,
4319 shader_glsl_color_fixup_supported
,