2 * GLSL pixel and vertex shader implementation
4 * Copyright 2006 Jason Green
5 * Copyright 2006-2007 Henri Verbeet
6 * Copyright 2007-2008 Stefan Dösinger for CodeWeavers
8 * This library is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU Lesser General Public
10 * License as published by the Free Software Foundation; either
11 * version 2.1 of the License, or (at your option) any later version.
13 * This library is distributed in the hope that it will be useful,
14 * but WITHOUT ANY WARRANTY; without even the implied warranty of
15 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
16 * Lesser General Public License for more details.
18 * You should have received a copy of the GNU Lesser General Public
19 * License along with this library; if not, write to the Free Software
20 * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301, USA
24 * D3D shader asm has swizzles on source parameters, and write masks for
25 * destination parameters. GLSL uses swizzles for both. The result of this is
26 * that for example "mov dst.xw, src.zyxw" becomes "dst.xw = src.zw" in GLSL.
27 * Ie, to generate a proper GLSL source swizzle, we need to take the D3D write
28 * mask for the destination parameter into account.
34 #include "wined3d_private.h"
36 WINE_DEFAULT_DEBUG_CHANNEL(d3d_shader
);
37 WINE_DECLARE_DEBUG_CHANNEL(d3d_constants
);
38 WINE_DECLARE_DEBUG_CHANNEL(d3d_caps
);
39 WINE_DECLARE_DEBUG_CHANNEL(d3d
);
41 #define GLINFO_LOCATION (*gl_info)
43 #define WINED3D_GLSL_SAMPLE_PROJECTED 0x1
44 #define WINED3D_GLSL_SAMPLE_RECT 0x2
45 #define WINED3D_GLSL_SAMPLE_LOD 0x4
60 } glsl_sample_function_t
;
64 HEAP_NODE_TRAVERSE_LEFT
,
65 HEAP_NODE_TRAVERSE_RIGHT
,
77 struct constant_entry
*entries
;
78 unsigned int *positions
;
82 /* GLSL shader private data */
83 struct shader_glsl_priv
{
84 struct hash_table_t
*glsl_program_lookup
;
85 struct glsl_shader_prog_link
*glsl_program
;
86 struct constant_heap vconst_heap
;
87 struct constant_heap pconst_heap
;
89 GLhandleARB depth_blt_program
[tex_type_count
];
90 UINT next_constant_version
;
93 /* Struct to maintain data about a linked GLSL program */
94 struct glsl_shader_prog_link
{
95 struct list vshader_entry
;
96 struct list pshader_entry
;
97 GLhandleARB programId
;
98 GLint
*vuniformF_locations
;
99 GLint
*puniformF_locations
;
100 GLint vuniformI_locations
[MAX_CONST_I
];
101 GLint puniformI_locations
[MAX_CONST_I
];
102 GLint posFixup_location
;
103 GLint np2Fixup_location
[MAX_FRAGMENT_SAMPLERS
];
104 GLint bumpenvmat_location
[MAX_TEXTURES
];
105 GLint luminancescale_location
[MAX_TEXTURES
];
106 GLint luminanceoffset_location
[MAX_TEXTURES
];
107 GLint ycorrection_location
;
108 GLenum vertex_color_clamp
;
109 IWineD3DVertexShader
*vshader
;
110 IWineD3DPixelShader
*pshader
;
111 struct vs_compile_args vs_args
;
112 struct ps_compile_args ps_args
;
113 UINT constant_version
;
117 IWineD3DVertexShader
*vshader
;
118 IWineD3DPixelShader
*pshader
;
119 struct ps_compile_args ps_args
;
120 struct vs_compile_args vs_args
;
121 } glsl_program_key_t
;
124 /** Prints the GLSL info log which will contain error messages if they exist */
125 static void print_glsl_info_log(const WineD3D_GL_Info
*gl_info
, GLhandleARB obj
)
127 int infologLength
= 0;
132 static const char * const spam
[] =
134 "Vertex shader was successfully compiled to run on hardware.\n", /* fglrx */
135 "Fragment shader was successfully compiled to run on hardware.\n", /* fglrx */
136 "Fragment shader(s) linked, vertex shader(s) linked. \n ", /* fglrx, with \n */
137 "Fragment shader(s) linked, vertex shader(s) linked.", /* fglrx, no \n */
138 "Vertex shader(s) linked, no fragment shader(s) defined. \n ", /* fglrx, with \n */
139 "Vertex shader(s) linked, no fragment shader(s) defined.", /* fglrx, no \n */
140 "Fragment shader was successfully compiled to run on hardware.\n"
141 "WARNING: 0:2: extension 'GL_ARB_draw_buffers' is not supported",
142 "Fragment shader(s) linked, no vertex shader(s) defined.", /* fglrx, no \n */
143 "Fragment shader(s) linked, no vertex shader(s) defined. \n ", /* fglrx, with \n */
144 "WARNING: 0:2: extension 'GL_ARB_draw_buffers' is not supported\n" /* MacOS ati */
147 if (!TRACE_ON(d3d_shader
) && !FIXME_ON(d3d_shader
)) return;
149 GL_EXTCALL(glGetObjectParameterivARB(obj
,
150 GL_OBJECT_INFO_LOG_LENGTH_ARB
,
153 /* A size of 1 is just a null-terminated string, so the log should be bigger than
154 * that if there are errors. */
155 if (infologLength
> 1)
157 /* Fglrx doesn't terminate the string properly, but it tells us the proper length.
158 * So use HEAP_ZERO_MEMORY to avoid uninitialized bytes
160 infoLog
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, infologLength
);
161 GL_EXTCALL(glGetInfoLogARB(obj
, infologLength
, NULL
, infoLog
));
164 for(i
= 0; i
< sizeof(spam
) / sizeof(spam
[0]); i
++) {
165 if(strcmp(infoLog
, spam
[i
]) == 0) {
171 TRACE("Spam received from GLSL shader #%u: %s\n", obj
, debugstr_a(infoLog
));
173 FIXME("Error received from GLSL shader #%u: %s\n", obj
, debugstr_a(infoLog
));
175 HeapFree(GetProcessHeap(), 0, infoLog
);
180 * Loads (pixel shader) samplers
182 static void shader_glsl_load_psamplers(const WineD3D_GL_Info
*gl_info
, DWORD
*tex_unit_map
, GLhandleARB programId
)
186 char sampler_name
[20];
188 for (i
= 0; i
< MAX_FRAGMENT_SAMPLERS
; ++i
) {
189 snprintf(sampler_name
, sizeof(sampler_name
), "Psampler%d", i
);
190 name_loc
= GL_EXTCALL(glGetUniformLocationARB(programId
, sampler_name
));
191 if (name_loc
!= -1) {
192 DWORD mapped_unit
= tex_unit_map
[i
];
193 if (mapped_unit
!= WINED3D_UNMAPPED_STAGE
&& mapped_unit
< GL_LIMITS(fragment_samplers
))
195 TRACE("Loading %s for texture %d\n", sampler_name
, mapped_unit
);
196 GL_EXTCALL(glUniform1iARB(name_loc
, mapped_unit
));
197 checkGLcall("glUniform1iARB");
199 ERR("Trying to load sampler %s on unsupported unit %d\n", sampler_name
, mapped_unit
);
205 static void shader_glsl_load_vsamplers(const WineD3D_GL_Info
*gl_info
, DWORD
*tex_unit_map
, GLhandleARB programId
)
208 char sampler_name
[20];
211 for (i
= 0; i
< MAX_VERTEX_SAMPLERS
; ++i
) {
212 snprintf(sampler_name
, sizeof(sampler_name
), "Vsampler%d", i
);
213 name_loc
= GL_EXTCALL(glGetUniformLocationARB(programId
, sampler_name
));
214 if (name_loc
!= -1) {
215 DWORD mapped_unit
= tex_unit_map
[MAX_FRAGMENT_SAMPLERS
+ i
];
216 if (mapped_unit
!= WINED3D_UNMAPPED_STAGE
&& mapped_unit
< GL_LIMITS(combined_samplers
))
218 TRACE("Loading %s for texture %d\n", sampler_name
, mapped_unit
);
219 GL_EXTCALL(glUniform1iARB(name_loc
, mapped_unit
));
220 checkGLcall("glUniform1iARB");
222 ERR("Trying to load sampler %s on unsupported unit %d\n", sampler_name
, mapped_unit
);
228 static inline void walk_constant_heap(const WineD3D_GL_Info
*gl_info
, const float *constants
,
229 const GLint
*constant_locations
, const struct constant_heap
*heap
, unsigned char *stack
, DWORD version
)
232 unsigned int heap_idx
= 1;
235 if (heap
->entries
[heap_idx
].version
<= version
) return;
237 idx
= heap
->entries
[heap_idx
].idx
;
238 if (constant_locations
[idx
] != -1) GL_EXTCALL(glUniform4fvARB(constant_locations
[idx
], 1, &constants
[idx
* 4]));
239 stack
[stack_idx
] = HEAP_NODE_TRAVERSE_LEFT
;
241 while (stack_idx
>= 0)
243 /* Note that we fall through to the next case statement. */
244 switch(stack
[stack_idx
])
246 case HEAP_NODE_TRAVERSE_LEFT
:
248 unsigned int left_idx
= heap_idx
<< 1;
249 if (left_idx
< heap
->size
&& heap
->entries
[left_idx
].version
> version
)
252 idx
= heap
->entries
[heap_idx
].idx
;
253 if (constant_locations
[idx
] != -1)
254 GL_EXTCALL(glUniform4fvARB(constant_locations
[idx
], 1, &constants
[idx
* 4]));
256 stack
[stack_idx
++] = HEAP_NODE_TRAVERSE_RIGHT
;
257 stack
[stack_idx
] = HEAP_NODE_TRAVERSE_LEFT
;
262 case HEAP_NODE_TRAVERSE_RIGHT
:
264 unsigned int right_idx
= (heap_idx
<< 1) + 1;
265 if (right_idx
< heap
->size
&& heap
->entries
[right_idx
].version
> version
)
267 heap_idx
= right_idx
;
268 idx
= heap
->entries
[heap_idx
].idx
;
269 if (constant_locations
[idx
] != -1)
270 GL_EXTCALL(glUniform4fvARB(constant_locations
[idx
], 1, &constants
[idx
* 4]));
272 stack
[stack_idx
++] = HEAP_NODE_POP
;
273 stack
[stack_idx
] = HEAP_NODE_TRAVERSE_LEFT
;
286 checkGLcall("walk_constant_heap()");
289 static inline void apply_clamped_constant(const WineD3D_GL_Info
*gl_info
, GLint location
, const GLfloat
*data
)
291 GLfloat clamped_constant
[4];
293 if (location
== -1) return;
295 clamped_constant
[0] = data
[0] < -1.0f
? -1.0f
: data
[0] > 1.0 ? 1.0 : data
[0];
296 clamped_constant
[1] = data
[1] < -1.0f
? -1.0f
: data
[1] > 1.0 ? 1.0 : data
[1];
297 clamped_constant
[2] = data
[2] < -1.0f
? -1.0f
: data
[2] > 1.0 ? 1.0 : data
[2];
298 clamped_constant
[3] = data
[3] < -1.0f
? -1.0f
: data
[3] > 1.0 ? 1.0 : data
[3];
300 GL_EXTCALL(glUniform4fvARB(location
, 1, clamped_constant
));
303 static inline void walk_constant_heap_clamped(const WineD3D_GL_Info
*gl_info
, const float *constants
,
304 const GLint
*constant_locations
, const struct constant_heap
*heap
, unsigned char *stack
, DWORD version
)
307 unsigned int heap_idx
= 1;
310 if (heap
->entries
[heap_idx
].version
<= version
) return;
312 idx
= heap
->entries
[heap_idx
].idx
;
313 apply_clamped_constant(gl_info
, constant_locations
[idx
], &constants
[idx
* 4]);
314 stack
[stack_idx
] = HEAP_NODE_TRAVERSE_LEFT
;
316 while (stack_idx
>= 0)
318 /* Note that we fall through to the next case statement. */
319 switch(stack
[stack_idx
])
321 case HEAP_NODE_TRAVERSE_LEFT
:
323 unsigned int left_idx
= heap_idx
<< 1;
324 if (left_idx
< heap
->size
&& heap
->entries
[left_idx
].version
> version
)
327 idx
= heap
->entries
[heap_idx
].idx
;
328 apply_clamped_constant(gl_info
, constant_locations
[idx
], &constants
[idx
* 4]);
330 stack
[stack_idx
++] = HEAP_NODE_TRAVERSE_RIGHT
;
331 stack
[stack_idx
] = HEAP_NODE_TRAVERSE_LEFT
;
336 case HEAP_NODE_TRAVERSE_RIGHT
:
338 unsigned int right_idx
= (heap_idx
<< 1) + 1;
339 if (right_idx
< heap
->size
&& heap
->entries
[right_idx
].version
> version
)
341 heap_idx
= right_idx
;
342 idx
= heap
->entries
[heap_idx
].idx
;
343 apply_clamped_constant(gl_info
, constant_locations
[idx
], &constants
[idx
* 4]);
345 stack
[stack_idx
++] = HEAP_NODE_POP
;
346 stack
[stack_idx
] = HEAP_NODE_TRAVERSE_LEFT
;
359 checkGLcall("walk_constant_heap_clamped()");
362 /* Loads floating point constants (aka uniforms) into the currently set GLSL program. */
363 static void shader_glsl_load_constantsF(IWineD3DBaseShaderImpl
*This
, const WineD3D_GL_Info
*gl_info
,
364 const float *constants
, const GLint
*constant_locations
, const struct constant_heap
*heap
,
365 unsigned char *stack
, UINT version
)
367 const local_constant
*lconst
;
369 /* 1.X pshaders have the constants clamped to [-1;1] implicitly. */
370 if (WINED3DSHADER_VERSION_MAJOR(This
->baseShader
.reg_maps
.shader_version
) == 1
371 && shader_is_pshader_version(This
->baseShader
.reg_maps
.shader_version
))
372 walk_constant_heap_clamped(gl_info
, constants
, constant_locations
, heap
, stack
, version
);
374 walk_constant_heap(gl_info
, constants
, constant_locations
, heap
, stack
, version
);
376 if (!This
->baseShader
.load_local_constsF
)
378 TRACE("No need to load local float constants for this shader\n");
382 /* Immediate constants are clamped to [-1;1] at shader creation time if needed */
383 LIST_FOR_EACH_ENTRY(lconst
, &This
->baseShader
.constantsF
, local_constant
, entry
)
385 GLint location
= constant_locations
[lconst
->idx
];
386 /* We found this uniform name in the program - go ahead and send the data */
387 if (location
!= -1) GL_EXTCALL(glUniform4fvARB(location
, 1, (const GLfloat
*)lconst
->value
));
389 checkGLcall("glUniform4fvARB()");
392 /* Loads integer constants (aka uniforms) into the currently set GLSL program. */
393 static void shader_glsl_load_constantsI(IWineD3DBaseShaderImpl
*This
, const WineD3D_GL_Info
*gl_info
,
394 const GLint locations
[MAX_CONST_I
], const int *constants
, WORD constants_set
)
399 for (i
= 0; constants_set
; constants_set
>>= 1, ++i
)
401 if (!(constants_set
& 1)) continue;
403 TRACE_(d3d_constants
)("Loading constants %u: %i, %i, %i, %i\n",
404 i
, constants
[i
*4], constants
[i
*4+1], constants
[i
*4+2], constants
[i
*4+3]);
406 /* We found this uniform name in the program - go ahead and send the data */
407 GL_EXTCALL(glUniform4ivARB(locations
[i
], 1, &constants
[i
*4]));
408 checkGLcall("glUniform4ivARB");
411 /* Load immediate constants */
412 ptr
= list_head(&This
->baseShader
.constantsI
);
414 const struct local_constant
*lconst
= LIST_ENTRY(ptr
, const struct local_constant
, entry
);
415 unsigned int idx
= lconst
->idx
;
416 const GLint
*values
= (const GLint
*)lconst
->value
;
418 TRACE_(d3d_constants
)("Loading local constants %i: %i, %i, %i, %i\n", idx
,
419 values
[0], values
[1], values
[2], values
[3]);
421 /* We found this uniform name in the program - go ahead and send the data */
422 GL_EXTCALL(glUniform4ivARB(locations
[idx
], 1, values
));
423 checkGLcall("glUniform4ivARB");
424 ptr
= list_next(&This
->baseShader
.constantsI
, ptr
);
428 /* Loads boolean constants (aka uniforms) into the currently set GLSL program. */
429 static void shader_glsl_load_constantsB(IWineD3DBaseShaderImpl
*This
, const WineD3D_GL_Info
*gl_info
,
430 GLhandleARB programId
, const BOOL
*constants
, WORD constants_set
)
435 char is_pshader
= shader_is_pshader_version(This
->baseShader
.reg_maps
.shader_version
);
436 const char* prefix
= is_pshader
? "PB":"VB";
439 /* TODO: Benchmark and see if it would be beneficial to store the
440 * locations of the constants to avoid looking up each time */
441 for (i
= 0; constants_set
; constants_set
>>= 1, ++i
)
443 if (!(constants_set
& 1)) continue;
445 TRACE_(d3d_constants
)("Loading constants %i: %i;\n", i
, constants
[i
]);
447 /* TODO: Benchmark and see if it would be beneficial to store the
448 * locations of the constants to avoid looking up each time */
449 snprintf(tmp_name
, sizeof(tmp_name
), "%s[%i]", prefix
, i
);
450 tmp_loc
= GL_EXTCALL(glGetUniformLocationARB(programId
, tmp_name
));
453 /* We found this uniform name in the program - go ahead and send the data */
454 GL_EXTCALL(glUniform1ivARB(tmp_loc
, 1, &constants
[i
]));
455 checkGLcall("glUniform1ivARB");
459 /* Load immediate constants */
460 ptr
= list_head(&This
->baseShader
.constantsB
);
462 const struct local_constant
*lconst
= LIST_ENTRY(ptr
, const struct local_constant
, entry
);
463 unsigned int idx
= lconst
->idx
;
464 const GLint
*values
= (const GLint
*)lconst
->value
;
466 TRACE_(d3d_constants
)("Loading local constants %i: %i\n", idx
, values
[0]);
468 snprintf(tmp_name
, sizeof(tmp_name
), "%s[%i]", prefix
, idx
);
469 tmp_loc
= GL_EXTCALL(glGetUniformLocationARB(programId
, tmp_name
));
471 /* We found this uniform name in the program - go ahead and send the data */
472 GL_EXTCALL(glUniform1ivARB(tmp_loc
, 1, values
));
473 checkGLcall("glUniform1ivARB");
475 ptr
= list_next(&This
->baseShader
.constantsB
, ptr
);
479 static void reset_program_constant_version(void *value
, void *context
)
481 struct glsl_shader_prog_link
*entry
= value
;
482 entry
->constant_version
= 0;
486 * Loads the texture dimensions for NP2 fixup into the currently set GLSL program.
488 static void shader_glsl_load_np2fixup_constants(
489 IWineD3DDevice
* device
,
491 char useVertexShader
) {
493 const IWineD3DDeviceImpl
* deviceImpl
= (const IWineD3DDeviceImpl
*) device
;
494 const struct glsl_shader_prog_link
* prog
= ((struct shader_glsl_priv
*)(deviceImpl
->shader_priv
))->glsl_program
;
497 /* No GLSL program set - nothing to do. */
501 if (!usePixelShader
) {
502 /* NP2 texcoord fixup is (currently) only done for pixelshaders. */
506 if (prog
->ps_args
.np2_fixup
) {
508 UINT fixup
= prog
->ps_args
.np2_fixup
;
509 const WineD3D_GL_Info
* gl_info
= &deviceImpl
->adapter
->gl_info
;
510 const IWineD3DStateBlockImpl
* stateBlock
= (const IWineD3DStateBlockImpl
*) deviceImpl
->stateBlock
;
512 for (i
= 0; fixup
; fixup
>>= 1, ++i
) {
513 if (-1 != prog
->np2Fixup_location
[i
]) {
514 const IWineD3DBaseTextureImpl
* const tex
= (const IWineD3DBaseTextureImpl
*) stateBlock
->textures
[i
];
516 FIXME("Non-existant texture is flagged for NP2 texcoord fixup\n");
519 const float tex_dim
[2] = {tex
->baseTexture
.pow2Matrix
[0], tex
->baseTexture
.pow2Matrix
[5]};
520 GL_EXTCALL(glUniform2fvARB(prog
->np2Fixup_location
[i
], 1, tex_dim
));
528 * Loads the app-supplied constants into the currently set GLSL program.
530 static void shader_glsl_load_constants(
531 IWineD3DDevice
* device
,
533 char useVertexShader
) {
535 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) device
;
536 struct shader_glsl_priv
*priv
= deviceImpl
->shader_priv
;
537 IWineD3DStateBlockImpl
* stateBlock
= deviceImpl
->stateBlock
;
538 const WineD3D_GL_Info
*gl_info
= &deviceImpl
->adapter
->gl_info
;
540 GLhandleARB programId
;
541 struct glsl_shader_prog_link
*prog
= priv
->glsl_program
;
542 UINT constant_version
;
546 /* No GLSL program set - nothing to do. */
549 programId
= prog
->programId
;
550 constant_version
= prog
->constant_version
;
552 if (useVertexShader
) {
553 IWineD3DBaseShaderImpl
* vshader
= (IWineD3DBaseShaderImpl
*) stateBlock
->vertexShader
;
555 /* Load DirectX 9 float constants/uniforms for vertex shader */
556 shader_glsl_load_constantsF(vshader
, gl_info
, stateBlock
->vertexShaderConstantF
,
557 prog
->vuniformF_locations
, &priv
->vconst_heap
, priv
->stack
, constant_version
);
559 /* Load DirectX 9 integer constants/uniforms for vertex shader */
560 if(vshader
->baseShader
.uses_int_consts
) {
561 shader_glsl_load_constantsI(vshader
, gl_info
, prog
->vuniformI_locations
,
562 stateBlock
->vertexShaderConstantI
, stateBlock
->changed
.vertexShaderConstantsI
);
565 /* Load DirectX 9 boolean constants/uniforms for vertex shader */
566 if(vshader
->baseShader
.uses_bool_consts
) {
567 shader_glsl_load_constantsB(vshader
, gl_info
, programId
,
568 stateBlock
->vertexShaderConstantB
, stateBlock
->changed
.vertexShaderConstantsB
);
571 /* Upload the position fixup params */
572 GL_EXTCALL(glUniform4fvARB(prog
->posFixup_location
, 1, &deviceImpl
->posFixup
[0]));
573 checkGLcall("glUniform4fvARB");
576 if (usePixelShader
) {
578 IWineD3DBaseShaderImpl
* pshader
= (IWineD3DBaseShaderImpl
*) stateBlock
->pixelShader
;
580 /* Load DirectX 9 float constants/uniforms for pixel shader */
581 shader_glsl_load_constantsF(pshader
, gl_info
, stateBlock
->pixelShaderConstantF
,
582 prog
->puniformF_locations
, &priv
->pconst_heap
, priv
->stack
, constant_version
);
584 /* Load DirectX 9 integer constants/uniforms for pixel shader */
585 if(pshader
->baseShader
.uses_int_consts
) {
586 shader_glsl_load_constantsI(pshader
, gl_info
, prog
->puniformI_locations
,
587 stateBlock
->pixelShaderConstantI
, stateBlock
->changed
.pixelShaderConstantsI
);
590 /* Load DirectX 9 boolean constants/uniforms for pixel shader */
591 if(pshader
->baseShader
.uses_bool_consts
) {
592 shader_glsl_load_constantsB(pshader
, gl_info
, programId
,
593 stateBlock
->pixelShaderConstantB
, stateBlock
->changed
.pixelShaderConstantsB
);
596 /* Upload the environment bump map matrix if needed. The needsbumpmat member specifies the texture stage to load the matrix from.
597 * It can't be 0 for a valid texbem instruction.
599 for(i
= 0; i
< ((IWineD3DPixelShaderImpl
*) pshader
)->numbumpenvmatconsts
; i
++) {
600 IWineD3DPixelShaderImpl
*ps
= (IWineD3DPixelShaderImpl
*) pshader
;
601 int stage
= ps
->luminanceconst
[i
].texunit
;
603 const float *data
= (const float *)&stateBlock
->textureState
[(int)ps
->bumpenvmatconst
[i
].texunit
][WINED3DTSS_BUMPENVMAT00
];
604 GL_EXTCALL(glUniformMatrix2fvARB(prog
->bumpenvmat_location
[i
], 1, 0, data
));
605 checkGLcall("glUniformMatrix2fvARB");
607 /* texbeml needs the luminance scale and offset too. If texbeml is used, needsbumpmat
608 * is set too, so we can check that in the needsbumpmat check
610 if(ps
->baseShader
.reg_maps
.luminanceparams
[stage
]) {
611 const GLfloat
*scale
= (const GLfloat
*)&stateBlock
->textureState
[stage
][WINED3DTSS_BUMPENVLSCALE
];
612 const GLfloat
*offset
= (const GLfloat
*)&stateBlock
->textureState
[stage
][WINED3DTSS_BUMPENVLOFFSET
];
614 GL_EXTCALL(glUniform1fvARB(prog
->luminancescale_location
[i
], 1, scale
));
615 checkGLcall("glUniform1fvARB");
616 GL_EXTCALL(glUniform1fvARB(prog
->luminanceoffset_location
[i
], 1, offset
));
617 checkGLcall("glUniform1fvARB");
621 if(((IWineD3DPixelShaderImpl
*) pshader
)->vpos_uniform
) {
622 float correction_params
[4];
623 if(deviceImpl
->render_offscreen
) {
624 correction_params
[0] = 0.0;
625 correction_params
[1] = 1.0;
627 /* position is window relative, not viewport relative */
628 correction_params
[0] = ((IWineD3DSurfaceImpl
*) deviceImpl
->render_targets
[0])->currentDesc
.Height
;
629 correction_params
[1] = -1.0;
631 GL_EXTCALL(glUniform4fvARB(prog
->ycorrection_location
, 1, correction_params
));
635 if (priv
->next_constant_version
== UINT_MAX
)
637 TRACE("Max constant version reached, resetting to 0.\n");
638 hash_table_for_each_entry(priv
->glsl_program_lookup
, reset_program_constant_version
, NULL
);
639 priv
->next_constant_version
= 1;
643 prog
->constant_version
= priv
->next_constant_version
++;
647 static inline void update_heap_entry(struct constant_heap
*heap
, unsigned int idx
,
648 unsigned int heap_idx
, DWORD new_version
)
650 struct constant_entry
*entries
= heap
->entries
;
651 unsigned int *positions
= heap
->positions
;
652 unsigned int parent_idx
;
656 parent_idx
= heap_idx
>> 1;
658 if (new_version
<= entries
[parent_idx
].version
) break;
660 entries
[heap_idx
] = entries
[parent_idx
];
661 positions
[entries
[parent_idx
].idx
] = heap_idx
;
662 heap_idx
= parent_idx
;
665 entries
[heap_idx
].version
= new_version
;
666 entries
[heap_idx
].idx
= idx
;
667 positions
[idx
] = heap_idx
;
670 static void shader_glsl_update_float_vertex_constants(IWineD3DDevice
*iface
, UINT start
, UINT count
)
672 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
673 struct shader_glsl_priv
*priv
= This
->shader_priv
;
674 struct constant_heap
*heap
= &priv
->vconst_heap
;
677 for (i
= start
; i
< count
+ start
; ++i
)
679 if (!This
->stateBlock
->changed
.vertexShaderConstantsF
[i
])
680 update_heap_entry(heap
, i
, heap
->size
++, priv
->next_constant_version
);
682 update_heap_entry(heap
, i
, heap
->positions
[i
], priv
->next_constant_version
);
686 static void shader_glsl_update_float_pixel_constants(IWineD3DDevice
*iface
, UINT start
, UINT count
)
688 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
689 struct shader_glsl_priv
*priv
= This
->shader_priv
;
690 struct constant_heap
*heap
= &priv
->pconst_heap
;
693 for (i
= start
; i
< count
+ start
; ++i
)
695 if (!This
->stateBlock
->changed
.pixelShaderConstantsF
[i
])
696 update_heap_entry(heap
, i
, heap
->size
++, priv
->next_constant_version
);
698 update_heap_entry(heap
, i
, heap
->positions
[i
], priv
->next_constant_version
);
702 /** Generate the variable & register declarations for the GLSL output target */
703 static void shader_generate_glsl_declarations(IWineD3DBaseShader
*iface
, const shader_reg_maps
*reg_maps
,
704 SHADER_BUFFER
*buffer
, const WineD3D_GL_Info
*gl_info
,
705 const struct ps_compile_args
*ps_args
)
707 IWineD3DBaseShaderImpl
* This
= (IWineD3DBaseShaderImpl
*) iface
;
708 IWineD3DDeviceImpl
*device
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
709 DWORD shader_version
= reg_maps
->shader_version
;
710 unsigned int i
, extra_constants_needed
= 0;
711 const local_constant
*lconst
;
713 /* There are some minor differences between pixel and vertex shaders */
714 char pshader
= shader_is_pshader_version(shader_version
);
715 char prefix
= pshader
? 'P' : 'V';
717 /* Prototype the subroutines */
718 for (i
= 0; i
< This
->baseShader
.limits
.label
; i
++) {
719 if (reg_maps
->labels
[i
])
720 shader_addline(buffer
, "void subroutine%u();\n", i
);
723 /* Declare the constants (aka uniforms) */
724 if (This
->baseShader
.limits
.constant_float
> 0) {
725 unsigned max_constantsF
;
727 max_constantsF
= GL_LIMITS(pshader_constantsF
) - (MAX_CONST_B
/ 4) - MAX_CONST_I
- 2;
728 max_constantsF
= min(This
->baseShader
.limits
.constant_float
, max_constantsF
);
730 /* Subtract the other potential uniforms from the max available (bools, ints, and 1 row of projection matrix) */
731 max_constantsF
= GL_LIMITS(vshader_constantsF
) - (MAX_CONST_B
/ 4) - MAX_CONST_I
- 1;
732 max_constantsF
= min(This
->baseShader
.limits
.constant_float
, max_constantsF
);
734 shader_addline(buffer
, "uniform vec4 %cC[%u];\n", prefix
, max_constantsF
);
737 if (This
->baseShader
.limits
.constant_int
> 0)
738 shader_addline(buffer
, "uniform ivec4 %cI[%u];\n", prefix
, This
->baseShader
.limits
.constant_int
);
740 if (This
->baseShader
.limits
.constant_bool
> 0)
741 shader_addline(buffer
, "uniform bool %cB[%u];\n", prefix
, This
->baseShader
.limits
.constant_bool
);
744 shader_addline(buffer
, "uniform vec4 posFixup;\n");
745 /* Predeclaration; This function is added at link time based on the pixel shader.
746 * VS 3.0 shaders have an array OUT[] the shader writes to, earlier versions don't have
747 * that. We know the input to the reorder function at vertex shader compile time, so
748 * we can deal with that. The reorder function for a 1.x and 2.x vertex shader can just
749 * read gl_FrontColor. The output depends on the pixel shader. The reorder function for a
750 * 1.x and 2.x pshader or for fixed function will write gl_FrontColor, and for a 3.0 shader
751 * it will write to the varying array. Here we depend on the shader optimizer on sorting that
752 * out. The nvidia driver only does that if the parameter is inout instead of out, hence the
755 if (shader_version
>= WINED3DVS_VERSION(3, 0))
757 shader_addline(buffer
, "void order_ps_input(in vec4[%u]);\n", MAX_REG_OUTPUT
);
759 shader_addline(buffer
, "void order_ps_input();\n");
762 IWineD3DPixelShaderImpl
*ps_impl
= (IWineD3DPixelShaderImpl
*) This
;
764 ps_impl
->numbumpenvmatconsts
= 0;
765 for(i
= 0; i
< (sizeof(reg_maps
->bumpmat
) / sizeof(reg_maps
->bumpmat
[0])); i
++) {
766 if(!reg_maps
->bumpmat
[i
]) {
770 ps_impl
->bumpenvmatconst
[(int) ps_impl
->numbumpenvmatconsts
].texunit
= i
;
771 shader_addline(buffer
, "uniform mat2 bumpenvmat%d;\n", i
);
773 if(reg_maps
->luminanceparams
) {
774 ps_impl
->luminanceconst
[(int) ps_impl
->numbumpenvmatconsts
].texunit
= i
;
775 shader_addline(buffer
, "uniform float luminancescale%d;\n", i
);
776 shader_addline(buffer
, "uniform float luminanceoffset%d;\n", i
);
777 extra_constants_needed
++;
779 ps_impl
->luminanceconst
[(int) ps_impl
->numbumpenvmatconsts
].texunit
= -1;
782 extra_constants_needed
++;
783 ps_impl
->numbumpenvmatconsts
++;
786 if(ps_args
->srgb_correction
) {
787 shader_addline(buffer
, "const vec4 srgb_mul_low = vec4(%f, %f, %f, %f);\n",
788 srgb_mul_low
, srgb_mul_low
, srgb_mul_low
, srgb_mul_low
);
789 shader_addline(buffer
, "const vec4 srgb_comparison = vec4(%f, %f, %f, %f);\n",
790 srgb_cmp
, srgb_cmp
, srgb_cmp
, srgb_cmp
);
792 if(reg_maps
->vpos
|| reg_maps
->usesdsy
) {
793 if(This
->baseShader
.limits
.constant_float
+ extra_constants_needed
+ 1 < GL_LIMITS(pshader_constantsF
)) {
794 shader_addline(buffer
, "uniform vec4 ycorrection;\n");
795 ((IWineD3DPixelShaderImpl
*) This
)->vpos_uniform
= 1;
796 extra_constants_needed
++;
798 /* This happens because we do not have proper tracking of the constant registers that are
799 * actually used, only the max limit of the shader version
801 FIXME("Cannot find a free uniform for vpos correction params\n");
802 shader_addline(buffer
, "const vec4 ycorrection = vec4(%f, %f, 0.0, 0.0);\n",
803 device
->render_offscreen
? 0.0 : ((IWineD3DSurfaceImpl
*) device
->render_targets
[0])->currentDesc
.Height
,
804 device
->render_offscreen
? 1.0 : -1.0);
806 shader_addline(buffer
, "vec4 vpos;\n");
810 /* Declare texture samplers */
811 for (i
= 0; i
< This
->baseShader
.limits
.sampler
; i
++) {
812 if (reg_maps
->samplers
[i
]) {
814 DWORD stype
= reg_maps
->samplers
[i
] & WINED3DSP_TEXTURETYPE_MASK
;
818 shader_addline(buffer
, "uniform sampler1D %csampler%u;\n", prefix
, i
);
821 if(device
->stateBlock
->textures
[i
] &&
822 IWineD3DBaseTexture_GetTextureDimensions(device
->stateBlock
->textures
[i
]) == GL_TEXTURE_RECTANGLE_ARB
) {
823 shader_addline(buffer
, "uniform sampler2DRect %csampler%u;\n", prefix
, i
);
825 shader_addline(buffer
, "uniform sampler2D %csampler%u;\n", prefix
, i
);
828 if (pshader
&& ps_args
->np2_fixup
& (1 << i
))
830 /* NP2/RECT textures in OpenGL use texcoords in the range [0,width]x[0,height]
831 * while D3D has them in the (normalized) [0,1]x[0,1] range.
832 * samplerNP2Fixup stores texture dimensions and is updated through
833 * shader_glsl_load_np2fixup_constants when the sampler changes. */
834 shader_addline(buffer
, "uniform vec2 %csamplerNP2Fixup%u;\n", prefix
, i
);
837 case WINED3DSTT_CUBE
:
838 shader_addline(buffer
, "uniform samplerCube %csampler%u;\n", prefix
, i
);
840 case WINED3DSTT_VOLUME
:
841 shader_addline(buffer
, "uniform sampler3D %csampler%u;\n", prefix
, i
);
844 shader_addline(buffer
, "uniform unsupported_sampler %csampler%u;\n", prefix
, i
);
845 FIXME("Unrecognized sampler type: %#x\n", stype
);
851 /* Declare address variables */
852 for (i
= 0; i
< This
->baseShader
.limits
.address
; i
++) {
853 if (reg_maps
->address
[i
])
854 shader_addline(buffer
, "ivec4 A%d;\n", i
);
857 /* Declare texture coordinate temporaries and initialize them */
858 for (i
= 0; i
< This
->baseShader
.limits
.texcoord
; i
++) {
859 if (reg_maps
->texcoord
[i
])
860 shader_addline(buffer
, "vec4 T%u = gl_TexCoord[%u];\n", i
, i
);
863 /* Declare input register varyings. Only pixel shader, vertex shaders have that declared in the
864 * helper function shader that is linked in at link time
866 if (pshader
&& shader_version
>= WINED3DPS_VERSION(3, 0))
868 if (use_vs(device
->stateBlock
))
870 shader_addline(buffer
, "varying vec4 IN[%u];\n", GL_LIMITS(glsl_varyings
) / 4);
872 /* TODO: Write a replacement shader for the fixed function vertex pipeline, so this isn't needed.
873 * For fixed function vertex processing + 3.0 pixel shader we need a separate function in the
874 * pixel shader that reads the fixed function color into the packed input registers.
876 shader_addline(buffer
, "vec4 IN[%u];\n", GL_LIMITS(glsl_varyings
) / 4);
880 /* Declare output register temporaries */
881 if(This
->baseShader
.limits
.packed_output
) {
882 shader_addline(buffer
, "vec4 OUT[%u];\n", This
->baseShader
.limits
.packed_output
);
885 /* Declare temporary variables */
886 for(i
= 0; i
< This
->baseShader
.limits
.temporary
; i
++) {
887 if (reg_maps
->temporary
[i
])
888 shader_addline(buffer
, "vec4 R%u;\n", i
);
891 /* Declare attributes */
892 for (i
= 0; i
< This
->baseShader
.limits
.attributes
; i
++) {
893 if (reg_maps
->attributes
[i
])
894 shader_addline(buffer
, "attribute vec4 attrib%i;\n", i
);
897 /* Declare loop registers aLx */
898 for (i
= 0; i
< reg_maps
->loop_depth
; i
++) {
899 shader_addline(buffer
, "int aL%u;\n", i
);
900 shader_addline(buffer
, "int tmpInt%u;\n", i
);
903 /* Temporary variables for matrix operations */
904 shader_addline(buffer
, "vec4 tmp0;\n");
905 shader_addline(buffer
, "vec4 tmp1;\n");
907 /* Local constants use a different name so they can be loaded once at shader link time
908 * They can't be hardcoded into the shader text via LC = {x, y, z, w}; because the
909 * float -> string conversion can cause precision loss.
911 if(!This
->baseShader
.load_local_constsF
) {
912 LIST_FOR_EACH_ENTRY(lconst
, &This
->baseShader
.constantsF
, local_constant
, entry
) {
913 shader_addline(buffer
, "uniform vec4 %cLC%u;\n", prefix
, lconst
->idx
);
917 /* Start the main program */
918 shader_addline(buffer
, "void main() {\n");
919 if(pshader
&& reg_maps
->vpos
) {
920 /* DirectX apps expect integer values, while OpenGL drivers add approximately 0.5. This causes
921 * off-by-one problems as spotted by the vPos d3d9 visual test. Unfortunately the ATI cards do
922 * not add exactly 0.5, but rather something like 0.49999999 or 0.50000001, which still causes
923 * precision troubles when we just substract 0.5.
925 * To deal with that just floor() the position. This will eliminate the fraction on all cards.
927 * TODO: Test how that behaves with multisampling once we can enable multisampling in winex11.
929 * An advantage of floor is that it works even if the driver doesn't add 1/2. It is somewhat
930 * questionable if 1.5, 2.5, ... are the proper values to return in gl_FragCoord, even though
931 * coordinates specify the pixel centers instead of the pixel corners. This code will behave
932 * correctly on drivers that returns integer values.
934 shader_addline(buffer
, "vpos = floor(vec4(0, ycorrection[0], 0, 0) + gl_FragCoord * vec4(1, ycorrection[1], 1, 1));\n");
938 /*****************************************************************************
939 * Functions to generate GLSL strings from DirectX Shader bytecode begin here.
941 * For more information, see http://wiki.winehq.org/DirectX-Shaders
942 ****************************************************************************/
945 static void shader_glsl_add_src_param(const struct wined3d_shader_instruction
*ins
, const DWORD param
,
946 const DWORD addr_token
, DWORD mask
, glsl_src_param_t
*src_param
);
948 /** Used for opcode modifiers - They multiply the result by the specified amount */
949 static const char * const shift_glsl_tab
[] = {
951 "2.0 * ", /* 1 (x2) */
952 "4.0 * ", /* 2 (x4) */
953 "8.0 * ", /* 3 (x8) */
954 "16.0 * ", /* 4 (x16) */
955 "32.0 * ", /* 5 (x32) */
962 "0.0625 * ", /* 12 (d16) */
963 "0.125 * ", /* 13 (d8) */
964 "0.25 * ", /* 14 (d4) */
965 "0.5 * " /* 15 (d2) */
968 /* Generate a GLSL parameter that does the input modifier computation and return the input register/mask to use */
969 static void shader_glsl_gen_modifier (
972 const char *in_regswizzle
,
977 switch (instr
& WINED3DSP_SRCMOD_MASK
) {
978 case WINED3DSPSM_DZ
: /* Need to handle this in the instructions itself (texld & texcrd). */
980 case WINED3DSPSM_NONE
:
981 sprintf(out_str
, "%s%s", in_reg
, in_regswizzle
);
983 case WINED3DSPSM_NEG
:
984 sprintf(out_str
, "-%s%s", in_reg
, in_regswizzle
);
986 case WINED3DSPSM_NOT
:
987 sprintf(out_str
, "!%s%s", in_reg
, in_regswizzle
);
989 case WINED3DSPSM_BIAS
:
990 sprintf(out_str
, "(%s%s - vec4(0.5)%s)", in_reg
, in_regswizzle
, in_regswizzle
);
992 case WINED3DSPSM_BIASNEG
:
993 sprintf(out_str
, "-(%s%s - vec4(0.5)%s)", in_reg
, in_regswizzle
, in_regswizzle
);
995 case WINED3DSPSM_SIGN
:
996 sprintf(out_str
, "(2.0 * (%s%s - 0.5))", in_reg
, in_regswizzle
);
998 case WINED3DSPSM_SIGNNEG
:
999 sprintf(out_str
, "-(2.0 * (%s%s - 0.5))", in_reg
, in_regswizzle
);
1001 case WINED3DSPSM_COMP
:
1002 sprintf(out_str
, "(1.0 - %s%s)", in_reg
, in_regswizzle
);
1004 case WINED3DSPSM_X2
:
1005 sprintf(out_str
, "(2.0 * %s%s)", in_reg
, in_regswizzle
);
1007 case WINED3DSPSM_X2NEG
:
1008 sprintf(out_str
, "-(2.0 * %s%s)", in_reg
, in_regswizzle
);
1010 case WINED3DSPSM_ABS
:
1011 sprintf(out_str
, "abs(%s%s)", in_reg
, in_regswizzle
);
1013 case WINED3DSPSM_ABSNEG
:
1014 sprintf(out_str
, "-abs(%s%s)", in_reg
, in_regswizzle
);
1017 FIXME("Unhandled modifier %u\n", (instr
& WINED3DSP_SRCMOD_MASK
));
1018 sprintf(out_str
, "%s%s", in_reg
, in_regswizzle
);
1022 /** Writes the GLSL variable name that corresponds to the register that the
1023 * DX opcode parameter is trying to access */
1024 static void shader_glsl_get_register_name(WINED3DSHADER_PARAM_REGISTER_TYPE register_type
, UINT register_idx
,
1025 BOOL rel_addr
, const DWORD addr_token
, char *register_name
, BOOL
*is_color
,
1026 const struct wined3d_shader_instruction
*ins
)
1028 /* oPos, oFog and oPts in D3D */
1029 static const char * const hwrastout_reg_names
[] = { "gl_Position", "gl_FogFragCoord", "gl_PointSize" };
1031 IWineD3DBaseShaderImpl
*This
= (IWineD3DBaseShaderImpl
*)ins
->ctx
->shader
;
1032 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
1033 const WineD3D_GL_Info
* gl_info
= &deviceImpl
->adapter
->gl_info
;
1034 DWORD shader_version
= This
->baseShader
.reg_maps
.shader_version
;
1035 char pshader
= shader_is_pshader_version(shader_version
);
1039 switch (register_type
)
1041 case WINED3DSPR_TEMP
:
1042 sprintf(register_name
, "R%u", register_idx
);
1044 case WINED3DSPR_INPUT
:
1046 /* Pixel shaders >= 3.0 */
1047 if (WINED3DSHADER_VERSION_MAJOR(shader_version
) >= 3)
1049 DWORD in_count
= GL_LIMITS(glsl_varyings
) / 4;
1053 glsl_src_param_t rel_param
;
1054 shader_glsl_add_src_param(ins
, addr_token
, 0, WINED3DSP_WRITEMASK_0
, &rel_param
);
1056 /* Removing a + 0 would be an obvious optimization, but macos doesn't see the NOP
1059 if (((IWineD3DPixelShaderImpl
*)This
)->input_reg_map
[register_idx
])
1061 if (((IWineD3DPixelShaderImpl
*)This
)->declared_in_count
> in_count
) {
1062 sprintf(register_name
, "((%s + %u) > %d ? (%s + %u) > %d ? gl_SecondaryColor : gl_Color : IN[%s + %u])",
1063 rel_param
.param_str
, ((IWineD3DPixelShaderImpl
*)This
)->input_reg_map
[register_idx
], in_count
- 1,
1064 rel_param
.param_str
, ((IWineD3DPixelShaderImpl
*)This
)->input_reg_map
[register_idx
], in_count
,
1065 rel_param
.param_str
, ((IWineD3DPixelShaderImpl
*)This
)->input_reg_map
[register_idx
]);
1067 sprintf(register_name
, "IN[%s + %u]", rel_param
.param_str
,
1068 ((IWineD3DPixelShaderImpl
*)This
)->input_reg_map
[register_idx
]);
1071 if (((IWineD3DPixelShaderImpl
*)This
)->declared_in_count
> in_count
) {
1072 sprintf(register_name
, "((%s) > %d ? (%s) > %d ? gl_SecondaryColor : gl_Color : IN[%s])",
1073 rel_param
.param_str
, in_count
- 1,
1074 rel_param
.param_str
, in_count
,
1075 rel_param
.param_str
);
1077 sprintf(register_name
, "IN[%s]", rel_param
.param_str
);
1081 DWORD idx
= ((IWineD3DPixelShaderImpl
*) This
)->input_reg_map
[register_idx
];
1082 if (idx
== in_count
) {
1083 sprintf(register_name
, "gl_Color");
1084 } else if (idx
== in_count
+ 1) {
1085 sprintf(register_name
, "gl_SecondaryColor");
1087 sprintf(register_name
, "IN[%u]", idx
);
1091 if (register_idx
== 0)
1092 strcpy(register_name
, "gl_Color");
1094 strcpy(register_name
, "gl_SecondaryColor");
1097 if (((IWineD3DVertexShaderImpl
*)This
)->cur_args
->swizzle_map
& (1 << register_idx
)) *is_color
= TRUE
;
1098 sprintf(register_name
, "attrib%u", register_idx
);
1101 case WINED3DSPR_CONST
:
1103 const char prefix
= pshader
? 'P':'V';
1105 /* Relative addressing */
1108 /* Relative addressing on shaders 2.0+ have a relative address token,
1109 * prior to that, it was hard-coded as "A0.x" because there's only 1 register */
1110 if (WINED3DSHADER_VERSION_MAJOR(shader_version
) >= 2)
1112 glsl_src_param_t rel_param
;
1113 shader_glsl_add_src_param(ins
, addr_token
, 0, WINED3DSP_WRITEMASK_0
, &rel_param
);
1116 sprintf(register_name
, "%cC[%s + %u]", prefix
, rel_param
.param_str
, register_idx
);
1118 sprintf(register_name
, "%cC[%s]", prefix
, rel_param
.param_str
);
1123 sprintf(register_name
, "%cC[A0.x + %u]", prefix
, register_idx
);
1125 sprintf(register_name
, "%cC[A0.x]", prefix
);
1130 if (shader_constant_is_local(This
, register_idx
))
1132 sprintf(register_name
, "%cLC%u", prefix
, register_idx
);
1134 sprintf(register_name
, "%cC[%u]", prefix
, register_idx
);
1140 case WINED3DSPR_CONSTINT
:
1142 sprintf(register_name
, "PI[%u]", register_idx
);
1144 sprintf(register_name
, "VI[%u]", register_idx
);
1146 case WINED3DSPR_CONSTBOOL
:
1148 sprintf(register_name
, "PB[%u]", register_idx
);
1150 sprintf(register_name
, "VB[%u]", register_idx
);
1152 case WINED3DSPR_TEXTURE
: /* case WINED3DSPR_ADDR: */
1154 sprintf(register_name
, "T%u", register_idx
);
1156 sprintf(register_name
, "A%u", register_idx
);
1159 case WINED3DSPR_LOOP
:
1160 sprintf(register_name
, "aL%u", This
->baseShader
.cur_loop_regno
- 1);
1162 case WINED3DSPR_SAMPLER
:
1164 sprintf(register_name
, "Psampler%u", register_idx
);
1166 sprintf(register_name
, "Vsampler%u", register_idx
);
1168 case WINED3DSPR_COLOROUT
:
1169 if (register_idx
>= GL_LIMITS(buffers
))
1170 WARN("Write to render target %u, only %d supported\n", register_idx
, 4);
1172 if (GL_SUPPORT(ARB_DRAW_BUFFERS
)) {
1173 sprintf(register_name
, "gl_FragData[%u]", register_idx
);
1174 } else { /* On older cards with GLSL support like the GeforceFX there's only one buffer. */
1175 sprintf(register_name
, "gl_FragColor");
1178 case WINED3DSPR_RASTOUT
:
1179 sprintf(register_name
, "%s", hwrastout_reg_names
[register_idx
]);
1181 case WINED3DSPR_DEPTHOUT
:
1182 sprintf(register_name
, "gl_FragDepth");
1184 case WINED3DSPR_ATTROUT
:
1185 if (register_idx
== 0)
1187 sprintf(register_name
, "gl_FrontColor");
1189 sprintf(register_name
, "gl_FrontSecondaryColor");
1192 case WINED3DSPR_TEXCRDOUT
:
1193 /* Vertex shaders >= 3.0: WINED3DSPR_OUTPUT */
1194 if (WINED3DSHADER_VERSION_MAJOR(shader_version
) >= 3) sprintf(register_name
, "OUT[%u]", register_idx
);
1195 else sprintf(register_name
, "gl_TexCoord[%u]", register_idx
);
1197 case WINED3DSPR_MISCTYPE
:
1198 if (register_idx
== 0)
1201 sprintf(register_name
, "vpos");
1203 else if (register_idx
== 1)
1205 /* Note that gl_FrontFacing is a bool, while vFace is
1206 * a float for which the sign determines front/back
1208 sprintf(register_name
, "(gl_FrontFacing ? 1.0 : -1.0)");
1210 FIXME("Unhandled misctype register %d\n", register_idx
);
1211 sprintf(register_name
, "unrecognized_register");
1215 FIXME("Unhandled register name Type(%d)\n", register_type
);
1216 sprintf(register_name
, "unrecognized_register");
1221 static void shader_glsl_write_mask_to_str(DWORD write_mask
, char *str
)
1224 if (write_mask
& WINED3DSP_WRITEMASK_0
) *str
++ = 'x';
1225 if (write_mask
& WINED3DSP_WRITEMASK_1
) *str
++ = 'y';
1226 if (write_mask
& WINED3DSP_WRITEMASK_2
) *str
++ = 'z';
1227 if (write_mask
& WINED3DSP_WRITEMASK_3
) *str
++ = 'w';
1231 /* Get the GLSL write mask for the destination register */
1232 static DWORD
shader_glsl_get_write_mask(const struct wined3d_shader_dst_param
*param
, char *write_mask
)
1234 DWORD mask
= param
->write_mask
;
1236 if (shader_is_scalar(param
->register_type
, param
->register_idx
))
1238 mask
= WINED3DSP_WRITEMASK_0
;
1243 shader_glsl_write_mask_to_str(mask
, write_mask
);
1249 static unsigned int shader_glsl_get_write_mask_size(DWORD write_mask
) {
1250 unsigned int size
= 0;
1252 if (write_mask
& WINED3DSP_WRITEMASK_0
) ++size
;
1253 if (write_mask
& WINED3DSP_WRITEMASK_1
) ++size
;
1254 if (write_mask
& WINED3DSP_WRITEMASK_2
) ++size
;
1255 if (write_mask
& WINED3DSP_WRITEMASK_3
) ++size
;
1260 static void shader_glsl_get_swizzle(const DWORD param
, BOOL fixup
, DWORD mask
, char *swizzle_str
) {
1261 /* For registers of type WINED3DDECLTYPE_D3DCOLOR, data is stored as "bgra",
1262 * but addressed as "rgba". To fix this we need to swap the register's x
1263 * and z components. */
1264 DWORD swizzle
= (param
& WINED3DSP_SWIZZLE_MASK
) >> WINED3DSP_SWIZZLE_SHIFT
;
1265 const char *swizzle_chars
= fixup
? "zyxw" : "xyzw";
1266 char *ptr
= swizzle_str
;
1268 if (!shader_is_scalar(shader_get_regtype(param
), param
& WINED3DSP_REGNUM_MASK
))
1271 /* swizzle bits fields: wwzzyyxx */
1272 if (mask
& WINED3DSP_WRITEMASK_0
) *ptr
++ = swizzle_chars
[swizzle
& 0x03];
1273 if (mask
& WINED3DSP_WRITEMASK_1
) *ptr
++ = swizzle_chars
[(swizzle
>> 2) & 0x03];
1274 if (mask
& WINED3DSP_WRITEMASK_2
) *ptr
++ = swizzle_chars
[(swizzle
>> 4) & 0x03];
1275 if (mask
& WINED3DSP_WRITEMASK_3
) *ptr
++ = swizzle_chars
[(swizzle
>> 6) & 0x03];
1281 /* From a given parameter token, generate the corresponding GLSL string.
1282 * Also, return the actual register name and swizzle in case the
1283 * caller needs this information as well. */
1284 static void shader_glsl_add_src_param(const struct wined3d_shader_instruction
*ins
,
1285 const DWORD param
, const DWORD addr_token
, DWORD mask
, glsl_src_param_t
*src_param
)
1287 BOOL is_color
= FALSE
;
1288 char swizzle_str
[6];
1290 src_param
->reg_name
[0] = '\0';
1291 src_param
->param_str
[0] = '\0';
1292 swizzle_str
[0] = '\0';
1294 shader_glsl_get_register_name(shader_get_regtype(param
), param
& WINED3DSP_REGNUM_MASK
,
1295 param
& WINED3DSHADER_ADDRMODE_RELATIVE
, addr_token
, src_param
->reg_name
, &is_color
, ins
);
1297 shader_glsl_get_swizzle(param
, is_color
, mask
, swizzle_str
);
1298 shader_glsl_gen_modifier(param
, src_param
->reg_name
, swizzle_str
, src_param
->param_str
);
1301 /* From a given parameter token, generate the corresponding GLSL string.
1302 * Also, return the actual register name and swizzle in case the
1303 * caller needs this information as well. */
1304 static DWORD
shader_glsl_add_dst_param(const struct wined3d_shader_instruction
*ins
,
1305 const struct wined3d_shader_dst_param
*wined3d_dst
, glsl_dst_param_t
*glsl_dst
)
1307 BOOL is_color
= FALSE
;
1309 glsl_dst
->mask_str
[0] = '\0';
1310 glsl_dst
->reg_name
[0] = '\0';
1312 shader_glsl_get_register_name(wined3d_dst
->register_type
, wined3d_dst
->register_idx
,
1313 wined3d_dst
->has_rel_addr
, wined3d_dst
->addr_token
, glsl_dst
->reg_name
, &is_color
, ins
);
1314 return shader_glsl_get_write_mask(wined3d_dst
, glsl_dst
->mask_str
);
1317 /* Append the destination part of the instruction to the buffer, return the effective write mask */
1318 static DWORD
shader_glsl_append_dst_ext(SHADER_BUFFER
*buffer
,
1319 const struct wined3d_shader_instruction
*ins
, const struct wined3d_shader_dst_param
*dst
)
1321 glsl_dst_param_t glsl_dst
;
1324 mask
= shader_glsl_add_dst_param(ins
, dst
, &glsl_dst
);
1325 if (mask
) shader_addline(buffer
, "%s%s = %s(", glsl_dst
.reg_name
, glsl_dst
.mask_str
, shift_glsl_tab
[dst
->shift
]);
1330 /* Append the destination part of the instruction to the buffer, return the effective write mask */
1331 static DWORD
shader_glsl_append_dst(SHADER_BUFFER
*buffer
, const struct wined3d_shader_instruction
*ins
)
1333 return shader_glsl_append_dst_ext(buffer
, ins
, &ins
->dst
[0]);
1336 /** Process GLSL instruction modifiers */
1337 void shader_glsl_add_instruction_modifiers(const struct wined3d_shader_instruction
*ins
)
1339 glsl_dst_param_t dst_param
;
1342 if (!ins
->dst_count
) return;
1344 modifiers
= ins
->dst
[0].modifiers
;
1345 if (!modifiers
) return;
1347 shader_glsl_add_dst_param(ins
, &ins
->dst
[0], &dst_param
);
1349 if (modifiers
& WINED3DSPDM_SATURATE
)
1351 /* _SAT means to clamp the value of the register to between 0 and 1 */
1352 shader_addline(ins
->ctx
->buffer
, "%s%s = clamp(%s%s, 0.0, 1.0);\n", dst_param
.reg_name
,
1353 dst_param
.mask_str
, dst_param
.reg_name
, dst_param
.mask_str
);
1356 if (modifiers
& WINED3DSPDM_MSAMPCENTROID
)
1358 FIXME("_centroid modifier not handled\n");
1361 if (modifiers
& WINED3DSPDM_PARTIALPRECISION
)
1363 /* MSDN says this modifier can be safely ignored, so that's what we'll do. */
1367 static inline const char *shader_get_comp_op(DWORD flags
)
1369 DWORD op
= (flags
& INST_CONTROLS_MASK
) >> INST_CONTROLS_SHIFT
;
1371 case COMPARISON_GT
: return ">";
1372 case COMPARISON_EQ
: return "==";
1373 case COMPARISON_GE
: return ">=";
1374 case COMPARISON_LT
: return "<";
1375 case COMPARISON_NE
: return "!=";
1376 case COMPARISON_LE
: return "<=";
1378 FIXME("Unrecognized comparison value: %u\n", op
);
1383 static void shader_glsl_get_sample_function(DWORD sampler_type
, DWORD flags
, glsl_sample_function_t
*sample_function
)
1385 BOOL projected
= flags
& WINED3D_GLSL_SAMPLE_PROJECTED
;
1386 BOOL texrect
= flags
& WINED3D_GLSL_SAMPLE_RECT
;
1387 BOOL lod
= flags
& WINED3D_GLSL_SAMPLE_LOD
;
1389 /* Note that there's no such thing as a projected cube texture. */
1390 switch(sampler_type
) {
1393 sample_function
->name
= projected
? "texture1DProjLod" : "texture1DLod";
1395 sample_function
->name
= projected
? "texture1DProj" : "texture1D";
1397 sample_function
->coord_mask
= WINED3DSP_WRITEMASK_0
;
1402 sample_function
->name
= projected
? "texture2DRectProjLod" : "texture2DRectLod";
1404 sample_function
->name
= projected
? "texture2DRectProj" : "texture2DRect";
1408 sample_function
->name
= projected
? "texture2DProjLod" : "texture2DLod";
1410 sample_function
->name
= projected
? "texture2DProj" : "texture2D";
1413 sample_function
->coord_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
;
1415 case WINED3DSTT_CUBE
:
1417 sample_function
->name
= "textureCubeLod";
1419 sample_function
->name
= "textureCube";
1421 sample_function
->coord_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
1423 case WINED3DSTT_VOLUME
:
1425 sample_function
->name
= projected
? "texture3DProjLod" : "texture3DLod";
1427 sample_function
->name
= projected
? "texture3DProj" : "texture3D";
1429 sample_function
->coord_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
1432 sample_function
->name
= "";
1433 sample_function
->coord_mask
= 0;
1434 FIXME("Unrecognized sampler type: %#x;\n", sampler_type
);
1439 static void shader_glsl_append_fixup_arg(char *arguments
, const char *reg_name
,
1440 BOOL sign_fixup
, enum fixup_channel_source channel_source
)
1442 switch(channel_source
)
1444 case CHANNEL_SOURCE_ZERO
:
1445 strcat(arguments
, "0.0");
1448 case CHANNEL_SOURCE_ONE
:
1449 strcat(arguments
, "1.0");
1452 case CHANNEL_SOURCE_X
:
1453 strcat(arguments
, reg_name
);
1454 strcat(arguments
, ".x");
1457 case CHANNEL_SOURCE_Y
:
1458 strcat(arguments
, reg_name
);
1459 strcat(arguments
, ".y");
1462 case CHANNEL_SOURCE_Z
:
1463 strcat(arguments
, reg_name
);
1464 strcat(arguments
, ".z");
1467 case CHANNEL_SOURCE_W
:
1468 strcat(arguments
, reg_name
);
1469 strcat(arguments
, ".w");
1473 FIXME("Unhandled channel source %#x\n", channel_source
);
1474 strcat(arguments
, "undefined");
1478 if (sign_fixup
) strcat(arguments
, " * 2.0 - 1.0");
1481 static void shader_glsl_color_correction(const struct wined3d_shader_instruction
*ins
, struct color_fixup_desc fixup
)
1483 struct wined3d_shader_dst_param dst
;
1484 unsigned int mask_size
, remaining
;
1485 glsl_dst_param_t dst_param
;
1486 char arguments
[256];
1490 if (fixup
.x_sign_fixup
|| fixup
.x_source
!= CHANNEL_SOURCE_X
) mask
|= WINED3DSP_WRITEMASK_0
;
1491 if (fixup
.y_sign_fixup
|| fixup
.y_source
!= CHANNEL_SOURCE_Y
) mask
|= WINED3DSP_WRITEMASK_1
;
1492 if (fixup
.z_sign_fixup
|| fixup
.z_source
!= CHANNEL_SOURCE_Z
) mask
|= WINED3DSP_WRITEMASK_2
;
1493 if (fixup
.w_sign_fixup
|| fixup
.w_source
!= CHANNEL_SOURCE_W
) mask
|= WINED3DSP_WRITEMASK_3
;
1494 mask
&= ins
->dst
[0].write_mask
;
1496 if (!mask
) return; /* Nothing to do */
1498 if (is_yuv_fixup(fixup
))
1500 enum yuv_fixup yuv_fixup
= get_yuv_fixup(fixup
);
1501 FIXME("YUV fixup (%#x) not supported\n", yuv_fixup
);
1505 mask_size
= shader_glsl_get_write_mask_size(mask
);
1508 dst
.write_mask
= mask
;
1509 shader_glsl_add_dst_param(ins
, &dst
, &dst_param
);
1511 arguments
[0] = '\0';
1512 remaining
= mask_size
;
1513 if (mask
& WINED3DSP_WRITEMASK_0
)
1515 shader_glsl_append_fixup_arg(arguments
, dst_param
.reg_name
, fixup
.x_sign_fixup
, fixup
.x_source
);
1516 if (--remaining
) strcat(arguments
, ", ");
1518 if (mask
& WINED3DSP_WRITEMASK_1
)
1520 shader_glsl_append_fixup_arg(arguments
, dst_param
.reg_name
, fixup
.y_sign_fixup
, fixup
.y_source
);
1521 if (--remaining
) strcat(arguments
, ", ");
1523 if (mask
& WINED3DSP_WRITEMASK_2
)
1525 shader_glsl_append_fixup_arg(arguments
, dst_param
.reg_name
, fixup
.z_sign_fixup
, fixup
.z_source
);
1526 if (--remaining
) strcat(arguments
, ", ");
1528 if (mask
& WINED3DSP_WRITEMASK_3
)
1530 shader_glsl_append_fixup_arg(arguments
, dst_param
.reg_name
, fixup
.w_sign_fixup
, fixup
.w_source
);
1531 if (--remaining
) strcat(arguments
, ", ");
1536 shader_addline(ins
->ctx
->buffer
, "%s%s = vec%u(%s);\n",
1537 dst_param
.reg_name
, dst_param
.mask_str
, mask_size
, arguments
);
1541 shader_addline(ins
->ctx
->buffer
, "%s%s = %s;\n", dst_param
.reg_name
, dst_param
.mask_str
, arguments
);
1545 static void PRINTF_ATTR(6, 7) shader_glsl_gen_sample_code(const struct wined3d_shader_instruction
*ins
,
1546 DWORD sampler
, const glsl_sample_function_t
*sample_function
, DWORD swizzle
,
1547 const char *bias
, const char *coord_reg_fmt
, ...)
1549 const char *sampler_base
;
1550 char dst_swizzle
[6];
1551 struct color_fixup_desc fixup
;
1552 BOOL np2_fixup
= FALSE
;
1555 shader_glsl_get_swizzle(swizzle
, FALSE
, ins
->dst
[0].write_mask
, dst_swizzle
);
1557 if (shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
))
1559 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*)ins
->ctx
->shader
;
1560 fixup
= This
->cur_args
->color_fixup
[sampler
];
1561 sampler_base
= "Psampler";
1563 if(This
->cur_args
->np2_fixup
& (1 << sampler
)) {
1565 FIXME("Biased sampling from NP2 textures is unsupported\n");
1571 sampler_base
= "Vsampler";
1572 fixup
= COLOR_FIXUP_IDENTITY
; /* FIXME: Vshader color fixup */
1575 shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
1577 shader_addline(ins
->ctx
->buffer
, "%s(%s%u, ", sample_function
->name
, sampler_base
, sampler
);
1579 va_start(args
, coord_reg_fmt
);
1580 shader_vaddline(ins
->ctx
->buffer
, coord_reg_fmt
, args
);
1584 shader_addline(ins
->ctx
->buffer
, ", %s)%s);\n", bias
, dst_swizzle
);
1587 shader_addline(ins
->ctx
->buffer
, " * PsamplerNP2Fixup%u)%s);\n", sampler
, dst_swizzle
);
1589 shader_addline(ins
->ctx
->buffer
, ")%s);\n", dst_swizzle
);
1593 if(!is_identity_fixup(fixup
)) {
1594 shader_glsl_color_correction(ins
, fixup
);
1598 /*****************************************************************************
1600 * Begin processing individual instruction opcodes
1602 ****************************************************************************/
1604 /* Generate GLSL arithmetic functions (dst = src1 + src2) */
1605 static void shader_glsl_arith(const struct wined3d_shader_instruction
*ins
)
1607 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
1608 glsl_src_param_t src0_param
;
1609 glsl_src_param_t src1_param
;
1613 /* Determine the GLSL operator to use based on the opcode */
1614 switch (ins
->handler_idx
)
1616 case WINED3DSIH_MUL
: op
= '*'; break;
1617 case WINED3DSIH_ADD
: op
= '+'; break;
1618 case WINED3DSIH_SUB
: op
= '-'; break;
1621 FIXME("Opcode %#x not yet handled in GLSL\n", ins
->handler_idx
);
1625 write_mask
= shader_glsl_append_dst(buffer
, ins
);
1626 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], write_mask
, &src0_param
);
1627 shader_glsl_add_src_param(ins
, ins
->src
[1], ins
->src_addr
[1], write_mask
, &src1_param
);
1628 shader_addline(buffer
, "%s %c %s);\n", src0_param
.param_str
, op
, src1_param
.param_str
);
1631 /* Process the WINED3DSIO_MOV opcode using GLSL (dst = src) */
1632 static void shader_glsl_mov(const struct wined3d_shader_instruction
*ins
)
1634 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
1635 glsl_src_param_t src0_param
;
1638 write_mask
= shader_glsl_append_dst(buffer
, ins
);
1639 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], write_mask
, &src0_param
);
1641 /* In vs_1_1 WINED3DSIO_MOV can write to the address register. In later
1642 * shader versions WINED3DSIO_MOVA is used for this. */
1643 if ((WINED3DSHADER_VERSION_MAJOR(ins
->ctx
->reg_maps
->shader_version
) == 1
1644 && !shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
)
1645 && ins
->dst
[0].register_type
== WINED3DSPR_ADDR
))
1647 /* This is a simple floor() */
1648 unsigned int mask_size
= shader_glsl_get_write_mask_size(write_mask
);
1649 if (mask_size
> 1) {
1650 shader_addline(buffer
, "ivec%d(floor(%s)));\n", mask_size
, src0_param
.param_str
);
1652 shader_addline(buffer
, "int(floor(%s)));\n", src0_param
.param_str
);
1655 else if(ins
->handler_idx
== WINED3DSIH_MOVA
)
1657 /* We need to *round* to the nearest int here. */
1658 unsigned int mask_size
= shader_glsl_get_write_mask_size(write_mask
);
1659 if (mask_size
> 1) {
1660 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
);
1662 shader_addline(buffer
, "int(floor(abs(%s) + 0.5) * sign(%s)));\n", src0_param
.param_str
, src0_param
.param_str
);
1665 shader_addline(buffer
, "%s);\n", src0_param
.param_str
);
1669 /* Process the dot product operators DP3 and DP4 in GLSL (dst = dot(src0, src1)) */
1670 static void shader_glsl_dot(const struct wined3d_shader_instruction
*ins
)
1672 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
1673 glsl_src_param_t src0_param
;
1674 glsl_src_param_t src1_param
;
1675 DWORD dst_write_mask
, src_write_mask
;
1676 unsigned int dst_size
= 0;
1678 dst_write_mask
= shader_glsl_append_dst(buffer
, ins
);
1679 dst_size
= shader_glsl_get_write_mask_size(dst_write_mask
);
1681 /* dp3 works on vec3, dp4 on vec4 */
1682 if (ins
->handler_idx
== WINED3DSIH_DP4
)
1684 src_write_mask
= WINED3DSP_WRITEMASK_ALL
;
1686 src_write_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
1689 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], src_write_mask
, &src0_param
);
1690 shader_glsl_add_src_param(ins
, ins
->src
[1], ins
->src_addr
[1], src_write_mask
, &src1_param
);
1693 shader_addline(buffer
, "vec%d(dot(%s, %s)));\n", dst_size
, src0_param
.param_str
, src1_param
.param_str
);
1695 shader_addline(buffer
, "dot(%s, %s));\n", src0_param
.param_str
, src1_param
.param_str
);
1699 /* Note that this instruction has some restrictions. The destination write mask
1700 * can't contain the w component, and the source swizzles have to be .xyzw */
1701 static void shader_glsl_cross(const struct wined3d_shader_instruction
*ins
)
1703 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
1704 glsl_src_param_t src0_param
;
1705 glsl_src_param_t src1_param
;
1708 shader_glsl_get_write_mask(&ins
->dst
[0], dst_mask
);
1709 shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
1710 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], src_mask
, &src0_param
);
1711 shader_glsl_add_src_param(ins
, ins
->src
[1], ins
->src_addr
[1], src_mask
, &src1_param
);
1712 shader_addline(ins
->ctx
->buffer
, "cross(%s, %s)%s);\n", src0_param
.param_str
, src1_param
.param_str
, dst_mask
);
1715 /* Process the WINED3DSIO_POW instruction in GLSL (dst = |src0|^src1)
1716 * Src0 and src1 are scalars. Note that D3D uses the absolute of src0, while
1717 * GLSL uses the value as-is. */
1718 static void shader_glsl_pow(const struct wined3d_shader_instruction
*ins
)
1720 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
1721 glsl_src_param_t src0_param
;
1722 glsl_src_param_t src1_param
;
1723 DWORD dst_write_mask
;
1724 unsigned int dst_size
;
1726 dst_write_mask
= shader_glsl_append_dst(buffer
, ins
);
1727 dst_size
= shader_glsl_get_write_mask_size(dst_write_mask
);
1729 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
1730 shader_glsl_add_src_param(ins
, ins
->src
[1], ins
->src_addr
[1], WINED3DSP_WRITEMASK_0
, &src1_param
);
1733 shader_addline(buffer
, "vec%d(pow(abs(%s), %s)));\n", dst_size
, src0_param
.param_str
, src1_param
.param_str
);
1735 shader_addline(buffer
, "pow(abs(%s), %s));\n", src0_param
.param_str
, src1_param
.param_str
);
1739 /* Process the WINED3DSIO_LOG instruction in GLSL (dst = log2(|src0|))
1740 * Src0 is a scalar. Note that D3D uses the absolute of src0, while
1741 * GLSL uses the value as-is. */
1742 static void shader_glsl_log(const struct wined3d_shader_instruction
*ins
)
1744 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
1745 glsl_src_param_t src0_param
;
1746 DWORD dst_write_mask
;
1747 unsigned int dst_size
;
1749 dst_write_mask
= shader_glsl_append_dst(buffer
, ins
);
1750 dst_size
= shader_glsl_get_write_mask_size(dst_write_mask
);
1752 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
1755 shader_addline(buffer
, "vec%d(log2(abs(%s))));\n", dst_size
, src0_param
.param_str
);
1757 shader_addline(buffer
, "log2(abs(%s)));\n", src0_param
.param_str
);
1761 /* Map the opcode 1-to-1 to the GL code (arg->dst = instruction(src0, src1, ...) */
1762 static void shader_glsl_map2gl(const struct wined3d_shader_instruction
*ins
)
1764 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
1765 glsl_src_param_t src_param
;
1766 const char *instruction
;
1770 /* Determine the GLSL function to use based on the opcode */
1771 /* TODO: Possibly make this a table for faster lookups */
1772 switch (ins
->handler_idx
)
1774 case WINED3DSIH_MIN
: instruction
= "min"; break;
1775 case WINED3DSIH_MAX
: instruction
= "max"; break;
1776 case WINED3DSIH_ABS
: instruction
= "abs"; break;
1777 case WINED3DSIH_FRC
: instruction
= "fract"; break;
1778 case WINED3DSIH_NRM
: instruction
= "normalize"; break;
1779 case WINED3DSIH_EXP
: instruction
= "exp2"; break;
1780 case WINED3DSIH_SGN
: instruction
= "sign"; break;
1781 case WINED3DSIH_DSX
: instruction
= "dFdx"; break;
1782 case WINED3DSIH_DSY
: instruction
= "ycorrection.y * dFdy"; break;
1783 default: instruction
= "";
1784 FIXME("Opcode %#x not yet handled in GLSL\n", ins
->handler_idx
);
1788 write_mask
= shader_glsl_append_dst(buffer
, ins
);
1790 shader_addline(buffer
, "%s(", instruction
);
1794 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], write_mask
, &src_param
);
1795 shader_addline(buffer
, "%s", src_param
.param_str
);
1796 for (i
= 1; i
< ins
->src_count
; ++i
)
1798 shader_glsl_add_src_param(ins
, ins
->src
[i
], ins
->src_addr
[i
], write_mask
, &src_param
);
1799 shader_addline(buffer
, ", %s", src_param
.param_str
);
1803 shader_addline(buffer
, "));\n");
1806 /** Process the WINED3DSIO_EXPP instruction in GLSL:
1807 * For shader model 1.x, do the following (and honor the writemask, so use a temporary variable):
1808 * dst.x = 2^(floor(src))
1809 * dst.y = src - floor(src)
1810 * dst.z = 2^src (partial precision is allowed, but optional)
1812 * For 2.0 shaders, just do this (honoring writemask and swizzle):
1813 * dst = 2^src; (partial precision is allowed, but optional)
1815 static void shader_glsl_expp(const struct wined3d_shader_instruction
*ins
)
1817 glsl_src_param_t src_param
;
1819 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src_param
);
1821 if (ins
->ctx
->reg_maps
->shader_version
< WINED3DPS_VERSION(2,0))
1825 shader_addline(ins
->ctx
->buffer
, "tmp0.x = exp2(floor(%s));\n", src_param
.param_str
);
1826 shader_addline(ins
->ctx
->buffer
, "tmp0.y = %s - floor(%s);\n", src_param
.param_str
, src_param
.param_str
);
1827 shader_addline(ins
->ctx
->buffer
, "tmp0.z = exp2(%s);\n", src_param
.param_str
);
1828 shader_addline(ins
->ctx
->buffer
, "tmp0.w = 1.0;\n");
1830 shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
1831 shader_glsl_get_write_mask(&ins
->dst
[0], dst_mask
);
1832 shader_addline(ins
->ctx
->buffer
, "tmp0%s);\n", dst_mask
);
1835 unsigned int mask_size
;
1837 write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
1838 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
1840 if (mask_size
> 1) {
1841 shader_addline(ins
->ctx
->buffer
, "vec%d(exp2(%s)));\n", mask_size
, src_param
.param_str
);
1843 shader_addline(ins
->ctx
->buffer
, "exp2(%s));\n", src_param
.param_str
);
1848 /** Process the RCP (reciprocal or inverse) opcode in GLSL (dst = 1 / src) */
1849 static void shader_glsl_rcp(const struct wined3d_shader_instruction
*ins
)
1851 glsl_src_param_t src_param
;
1853 unsigned int mask_size
;
1855 write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
1856 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
1857 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], WINED3DSP_WRITEMASK_3
, &src_param
);
1859 if (mask_size
> 1) {
1860 shader_addline(ins
->ctx
->buffer
, "vec%d(1.0 / %s));\n", mask_size
, src_param
.param_str
);
1862 shader_addline(ins
->ctx
->buffer
, "1.0 / %s);\n", src_param
.param_str
);
1866 static void shader_glsl_rsq(const struct wined3d_shader_instruction
*ins
)
1868 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
1869 glsl_src_param_t src_param
;
1871 unsigned int mask_size
;
1873 write_mask
= shader_glsl_append_dst(buffer
, ins
);
1874 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
1876 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], WINED3DSP_WRITEMASK_3
, &src_param
);
1878 if (mask_size
> 1) {
1879 shader_addline(buffer
, "vec%d(inversesqrt(%s)));\n", mask_size
, src_param
.param_str
);
1881 shader_addline(buffer
, "inversesqrt(%s));\n", src_param
.param_str
);
1885 /** Process signed comparison opcodes in GLSL. */
1886 static void shader_glsl_compare(const struct wined3d_shader_instruction
*ins
)
1888 glsl_src_param_t src0_param
;
1889 glsl_src_param_t src1_param
;
1891 unsigned int mask_size
;
1893 write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
1894 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
1895 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], write_mask
, &src0_param
);
1896 shader_glsl_add_src_param(ins
, ins
->src
[1], ins
->src_addr
[1], write_mask
, &src1_param
);
1898 if (mask_size
> 1) {
1899 const char *compare
;
1901 switch(ins
->handler_idx
)
1903 case WINED3DSIH_SLT
: compare
= "lessThan"; break;
1904 case WINED3DSIH_SGE
: compare
= "greaterThanEqual"; break;
1905 default: compare
= "";
1906 FIXME("Can't handle opcode %#x\n", ins
->handler_idx
);
1909 shader_addline(ins
->ctx
->buffer
, "vec%d(%s(%s, %s)));\n", mask_size
, compare
,
1910 src0_param
.param_str
, src1_param
.param_str
);
1912 switch(ins
->handler_idx
)
1914 case WINED3DSIH_SLT
:
1915 /* Step(src0, src1) is not suitable here because if src0 == src1 SLT is supposed,
1916 * to return 0.0 but step returns 1.0 because step is not < x
1917 * An alternative is a bvec compare padded with an unused second component.
1918 * step(src1 * -1.0, src0 * -1.0) is not an option because it suffers from the same
1919 * issue. Playing with not() is not possible either because not() does not accept
1922 shader_addline(ins
->ctx
->buffer
, "(%s < %s) ? 1.0 : 0.0);\n",
1923 src0_param
.param_str
, src1_param
.param_str
);
1925 case WINED3DSIH_SGE
:
1926 /* Here we can use the step() function and safe a conditional */
1927 shader_addline(ins
->ctx
->buffer
, "step(%s, %s));\n", src1_param
.param_str
, src0_param
.param_str
);
1930 FIXME("Can't handle opcode %#x\n", ins
->handler_idx
);
1936 /** Process CMP instruction in GLSL (dst = src0 >= 0.0 ? src1 : src2), per channel */
1937 static void shader_glsl_cmp(const struct wined3d_shader_instruction
*ins
)
1939 glsl_src_param_t src0_param
;
1940 glsl_src_param_t src1_param
;
1941 glsl_src_param_t src2_param
;
1942 DWORD write_mask
, cmp_channel
= 0;
1945 BOOL temp_destination
= FALSE
;
1947 if (shader_is_scalar(shader_get_regtype(ins
->src
[0]), ins
->src
[0] & WINED3DSP_REGNUM_MASK
))
1949 write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
1951 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], WINED3DSP_WRITEMASK_ALL
, &src0_param
);
1952 shader_glsl_add_src_param(ins
, ins
->src
[1], ins
->src_addr
[1], write_mask
, &src1_param
);
1953 shader_glsl_add_src_param(ins
, ins
->src
[2], ins
->src_addr
[2], write_mask
, &src2_param
);
1955 shader_addline(ins
->ctx
->buffer
, "%s >= 0.0 ? %s : %s);\n",
1956 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
1958 DWORD src0reg
= ins
->src
[0] & WINED3DSP_REGNUM_MASK
;
1959 DWORD src1reg
= ins
->src
[1] & WINED3DSP_REGNUM_MASK
;
1960 DWORD src2reg
= ins
->src
[2] & WINED3DSP_REGNUM_MASK
;
1961 DWORD src0regtype
= shader_get_regtype(ins
->src
[0]);
1962 DWORD src1regtype
= shader_get_regtype(ins
->src
[1]);
1963 DWORD src2regtype
= shader_get_regtype(ins
->src
[2]);
1964 DWORD dstreg
= ins
->dst
[0].register_idx
;
1965 DWORD dstregtype
= ins
->dst
[0].register_type
;
1966 DWORD dst_mask
= ins
->dst
[0].write_mask
;
1967 struct wined3d_shader_dst_param dst
= ins
->dst
[0];
1969 /* Cycle through all source0 channels */
1970 for (i
=0; i
<4; i
++) {
1972 /* Find the destination channels which use the current source0 channel */
1973 for (j
=0; j
<4; j
++) {
1974 if (((ins
->src
[0] >> (WINED3DSP_SWIZZLE_SHIFT
+ 2 * j
)) & 0x3) == i
)
1976 write_mask
|= WINED3DSP_WRITEMASK_0
<< j
;
1977 cmp_channel
= WINED3DSP_WRITEMASK_0
<< j
;
1980 dst
.write_mask
= dst_mask
& write_mask
;
1982 /* Splitting the cmp instruction up in multiple lines imposes a problem:
1983 * The first lines may overwrite source parameters of the following lines.
1984 * Deal with that by using a temporary destination register if needed
1986 if ((src0reg
== dstreg
&& src0regtype
== dstregtype
)
1987 || (src1reg
== dstreg
&& src1regtype
== dstregtype
)
1988 || (src2reg
== dstreg
&& src2regtype
== dstregtype
))
1990 write_mask
= shader_glsl_get_write_mask(&dst
, mask_char
);
1991 if (!write_mask
) continue;
1992 shader_addline(ins
->ctx
->buffer
, "tmp0%s = (", mask_char
);
1993 temp_destination
= TRUE
;
1995 write_mask
= shader_glsl_append_dst_ext(ins
->ctx
->buffer
, ins
, &dst
);
1996 if (!write_mask
) continue;
1999 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], cmp_channel
, &src0_param
);
2000 shader_glsl_add_src_param(ins
, ins
->src
[1], ins
->src_addr
[1], write_mask
, &src1_param
);
2001 shader_glsl_add_src_param(ins
, ins
->src
[2], ins
->src_addr
[2], write_mask
, &src2_param
);
2003 shader_addline(ins
->ctx
->buffer
, "%s >= 0.0 ? %s : %s);\n",
2004 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
2007 if(temp_destination
) {
2008 shader_glsl_get_write_mask(&ins
->dst
[0], mask_char
);
2009 shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2010 shader_addline(ins
->ctx
->buffer
, "tmp0%s);\n", mask_char
);
2016 /** Process the CND opcode in GLSL (dst = (src0 > 0.5) ? src1 : src2) */
2017 /* For ps 1.1-1.3, only a single component of src0 is used. For ps 1.4
2018 * the compare is done per component of src0. */
2019 static void shader_glsl_cnd(const struct wined3d_shader_instruction
*ins
)
2021 struct wined3d_shader_dst_param dst
;
2022 glsl_src_param_t src0_param
;
2023 glsl_src_param_t src1_param
;
2024 glsl_src_param_t src2_param
;
2025 DWORD write_mask
, cmp_channel
= 0;
2029 if (ins
->ctx
->reg_maps
->shader_version
< WINED3DPS_VERSION(1, 4))
2031 write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2032 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2033 shader_glsl_add_src_param(ins
, ins
->src
[1], ins
->src_addr
[1], write_mask
, &src1_param
);
2034 shader_glsl_add_src_param(ins
, ins
->src
[2], ins
->src_addr
[2], write_mask
, &src2_param
);
2036 /* Fun: The D3DSI_COISSUE flag changes the semantic of the cnd instruction for < 1.4 shaders */
2039 shader_addline(ins
->ctx
->buffer
, "%s /* COISSUE! */);\n", src1_param
.param_str
);
2041 shader_addline(ins
->ctx
->buffer
, "%s > 0.5 ? %s : %s);\n",
2042 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
2046 /* Cycle through all source0 channels */
2047 dst_mask
= ins
->dst
[0].write_mask
;
2049 for (i
=0; i
<4; i
++) {
2051 /* Find the destination channels which use the current source0 channel */
2052 for (j
=0; j
<4; j
++) {
2053 if (((ins
->src
[0] >> (WINED3DSP_SWIZZLE_SHIFT
+ 2 * j
)) & 0x3) == i
)
2055 write_mask
|= WINED3DSP_WRITEMASK_0
<< j
;
2056 cmp_channel
= WINED3DSP_WRITEMASK_0
<< j
;
2060 dst
.write_mask
= dst_mask
& write_mask
;
2061 write_mask
= shader_glsl_append_dst_ext(ins
->ctx
->buffer
, ins
, &dst
);
2062 if (!write_mask
) continue;
2064 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], cmp_channel
, &src0_param
);
2065 shader_glsl_add_src_param(ins
, ins
->src
[1], ins
->src_addr
[1], write_mask
, &src1_param
);
2066 shader_glsl_add_src_param(ins
, ins
->src
[2], ins
->src_addr
[2], write_mask
, &src2_param
);
2068 shader_addline(ins
->ctx
->buffer
, "%s > 0.5 ? %s : %s);\n",
2069 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
2073 /** GLSL code generation for WINED3DSIO_MAD: Multiply the first 2 opcodes, then add the last */
2074 static void shader_glsl_mad(const struct wined3d_shader_instruction
*ins
)
2076 glsl_src_param_t src0_param
;
2077 glsl_src_param_t src1_param
;
2078 glsl_src_param_t src2_param
;
2081 write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2082 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], write_mask
, &src0_param
);
2083 shader_glsl_add_src_param(ins
, ins
->src
[1], ins
->src_addr
[1], write_mask
, &src1_param
);
2084 shader_glsl_add_src_param(ins
, ins
->src
[2], ins
->src_addr
[2], write_mask
, &src2_param
);
2085 shader_addline(ins
->ctx
->buffer
, "(%s * %s) + %s);\n",
2086 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
2089 /** Handles transforming all WINED3DSIO_M?x? opcodes for
2090 Vertex shaders to GLSL codes */
2091 static void shader_glsl_mnxn(const struct wined3d_shader_instruction
*ins
)
2094 int nComponents
= 0;
2095 struct wined3d_shader_dst_param tmp_dst
= {0};
2096 struct wined3d_shader_instruction tmp_ins
;
2098 memset(&tmp_ins
, 0, sizeof(tmp_ins
));
2100 /* Set constants for the temporary argument */
2101 tmp_ins
.ctx
= ins
->ctx
;
2102 tmp_ins
.src
[0] = ins
->src
[0];
2103 tmp_ins
.src_addr
[0] = ins
->src_addr
[0];
2104 tmp_ins
.src_addr
[1] = ins
->src_addr
[1];
2105 tmp_ins
.dst_count
= 1;
2106 tmp_ins
.dst
= &tmp_dst
;
2107 tmp_ins
.src_count
= 2;
2109 switch(ins
->handler_idx
)
2111 case WINED3DSIH_M4x4
:
2113 tmp_ins
.handler_idx
= WINED3DSIH_DP4
;
2115 case WINED3DSIH_M4x3
:
2117 tmp_ins
.handler_idx
= WINED3DSIH_DP4
;
2119 case WINED3DSIH_M3x4
:
2121 tmp_ins
.handler_idx
= WINED3DSIH_DP3
;
2123 case WINED3DSIH_M3x3
:
2125 tmp_ins
.handler_idx
= WINED3DSIH_DP3
;
2127 case WINED3DSIH_M3x2
:
2129 tmp_ins
.handler_idx
= WINED3DSIH_DP3
;
2135 tmp_dst
= ins
->dst
[0];
2136 for (i
= 0; i
< nComponents
; ++i
)
2138 tmp_dst
.write_mask
= WINED3DSP_WRITEMASK_0
<< i
;
2139 tmp_ins
.src
[1] = ins
->src
[1] + i
;
2140 shader_glsl_dot(&tmp_ins
);
2145 The LRP instruction performs a component-wise linear interpolation
2146 between the second and third operands using the first operand as the
2147 blend factor. Equation: (dst = src2 + src0 * (src1 - src2))
2148 This is equivalent to mix(src2, src1, src0);
2150 static void shader_glsl_lrp(const struct wined3d_shader_instruction
*ins
)
2152 glsl_src_param_t src0_param
;
2153 glsl_src_param_t src1_param
;
2154 glsl_src_param_t src2_param
;
2157 write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2159 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], write_mask
, &src0_param
);
2160 shader_glsl_add_src_param(ins
, ins
->src
[1], ins
->src_addr
[1], write_mask
, &src1_param
);
2161 shader_glsl_add_src_param(ins
, ins
->src
[2], ins
->src_addr
[2], write_mask
, &src2_param
);
2163 shader_addline(ins
->ctx
->buffer
, "mix(%s, %s, %s));\n",
2164 src2_param
.param_str
, src1_param
.param_str
, src0_param
.param_str
);
2167 /** Process the WINED3DSIO_LIT instruction in GLSL:
2168 * dst.x = dst.w = 1.0
2169 * dst.y = (src0.x > 0) ? src0.x
2170 * dst.z = (src0.x > 0) ? ((src0.y > 0) ? pow(src0.y, src.w) : 0) : 0
2171 * where src.w is clamped at +- 128
2173 static void shader_glsl_lit(const struct wined3d_shader_instruction
*ins
)
2175 glsl_src_param_t src0_param
;
2176 glsl_src_param_t src1_param
;
2177 glsl_src_param_t src3_param
;
2180 shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2181 shader_glsl_get_write_mask(&ins
->dst
[0], dst_mask
);
2183 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2184 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], WINED3DSP_WRITEMASK_1
, &src1_param
);
2185 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], WINED3DSP_WRITEMASK_3
, &src3_param
);
2187 /* The sdk specifies the instruction like this
2189 * if(src.x > 0.0) dst.y = src.x
2191 * if(src.x > 0.0 && src.y > 0.0) dst.z = pow(src.y, power);
2195 * Obviously that has quite a few conditionals in it which we don't like. So the first step is this:
2196 * dst.x = 1.0 ... No further explanation needed
2197 * dst.y = max(src.y, 0.0); ... If x < 0.0, use 0.0, otherwise x. Same as the conditional
2198 * dst.z = x > 0.0 ? pow(max(y, 0.0), p) : 0; ... 0 ^ power is 0, and otherwise we use y anyway
2199 * dst.w = 1.0. ... Nothing fancy.
2201 * So we still have one conditional in there. So do this:
2202 * dst.z = pow(max(0.0, src.y) * step(0.0, src.x), power);
2204 * 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),
2205 * 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.
2206 * if both x and y are > 0, we get pow(y * 1.0, power), as it is supposed to
2208 shader_addline(ins
->ctx
->buffer
,
2209 "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",
2210 src0_param
.param_str
, src1_param
.param_str
, src0_param
.param_str
, src3_param
.param_str
, dst_mask
);
2213 /** Process the WINED3DSIO_DST instruction in GLSL:
2215 * dst.y = src0.x * src0.y
2219 static void shader_glsl_dst(const struct wined3d_shader_instruction
*ins
)
2221 glsl_src_param_t src0y_param
;
2222 glsl_src_param_t src0z_param
;
2223 glsl_src_param_t src1y_param
;
2224 glsl_src_param_t src1w_param
;
2227 shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2228 shader_glsl_get_write_mask(&ins
->dst
[0], dst_mask
);
2230 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], WINED3DSP_WRITEMASK_1
, &src0y_param
);
2231 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], WINED3DSP_WRITEMASK_2
, &src0z_param
);
2232 shader_glsl_add_src_param(ins
, ins
->src
[1], ins
->src_addr
[1], WINED3DSP_WRITEMASK_1
, &src1y_param
);
2233 shader_glsl_add_src_param(ins
, ins
->src
[1], ins
->src_addr
[1], WINED3DSP_WRITEMASK_3
, &src1w_param
);
2235 shader_addline(ins
->ctx
->buffer
, "vec4(1.0, %s * %s, %s, %s))%s;\n",
2236 src0y_param
.param_str
, src1y_param
.param_str
, src0z_param
.param_str
, src1w_param
.param_str
, dst_mask
);
2239 /** Process the WINED3DSIO_SINCOS instruction in GLSL:
2240 * VS 2.0 requires that specific cosine and sine constants be passed to this instruction so the hardware
2241 * can handle it. But, these functions are built-in for GLSL, so we can just ignore the last 2 params.
2243 * dst.x = cos(src0.?)
2244 * dst.y = sin(src0.?)
2248 static void shader_glsl_sincos(const struct wined3d_shader_instruction
*ins
)
2250 glsl_src_param_t src0_param
;
2253 write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2254 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2256 switch (write_mask
) {
2257 case WINED3DSP_WRITEMASK_0
:
2258 shader_addline(ins
->ctx
->buffer
, "cos(%s));\n", src0_param
.param_str
);
2261 case WINED3DSP_WRITEMASK_1
:
2262 shader_addline(ins
->ctx
->buffer
, "sin(%s));\n", src0_param
.param_str
);
2265 case (WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
):
2266 shader_addline(ins
->ctx
->buffer
, "vec2(cos(%s), sin(%s)));\n", src0_param
.param_str
, src0_param
.param_str
);
2270 ERR("Write mask should be .x, .y or .xy\n");
2275 /** Process the WINED3DSIO_LOOP instruction in GLSL:
2276 * Start a for() loop where src1.y is the initial value of aL,
2277 * increment aL by src1.z for a total of src1.x iterations.
2278 * Need to use a temporary variable for this operation.
2280 /* FIXME: I don't think nested loops will work correctly this way. */
2281 static void shader_glsl_loop(const struct wined3d_shader_instruction
*ins
)
2283 glsl_src_param_t src1_param
;
2284 IWineD3DBaseShaderImpl
*shader
= (IWineD3DBaseShaderImpl
*)ins
->ctx
->shader
;
2285 DWORD regtype
= shader_get_regtype(ins
->src
[1]);
2286 DWORD reg
= ins
->src
[1] & WINED3DSP_REGNUM_MASK
;
2287 const DWORD
*control_values
= NULL
;
2288 const local_constant
*constant
;
2290 shader_glsl_add_src_param(ins
, ins
->src
[1], ins
->src_addr
[1], WINED3DSP_WRITEMASK_ALL
, &src1_param
);
2292 /* Try to hardcode the loop control parameters if possible. Direct3D 9 class hardware doesn't support real
2293 * varying indexing, but Microsoft designed this feature for Shader model 2.x+. If the loop control is
2294 * known at compile time, the GLSL compiler can unroll the loop, and replace indirect addressing with direct
2297 if(regtype
== WINED3DSPR_CONSTINT
) {
2298 LIST_FOR_EACH_ENTRY(constant
, &shader
->baseShader
.constantsI
, local_constant
, entry
) {
2299 if(constant
->idx
== reg
) {
2300 control_values
= constant
->value
;
2306 if(control_values
) {
2307 if(control_values
[2] > 0) {
2308 shader_addline(ins
->ctx
->buffer
, "for (aL%u = %d; aL%u < (%d * %d + %d); aL%u += %d) {\n",
2309 shader
->baseShader
.cur_loop_depth
, control_values
[1],
2310 shader
->baseShader
.cur_loop_depth
, control_values
[0], control_values
[2], control_values
[1],
2311 shader
->baseShader
.cur_loop_depth
, control_values
[2]);
2312 } else if(control_values
[2] == 0) {
2313 shader_addline(ins
->ctx
->buffer
, "for (aL%u = %d, tmpInt%u = 0; tmpInt%u < %d; tmpInt%u++) {\n",
2314 shader
->baseShader
.cur_loop_depth
, control_values
[1], shader
->baseShader
.cur_loop_depth
,
2315 shader
->baseShader
.cur_loop_depth
, control_values
[0],
2316 shader
->baseShader
.cur_loop_depth
);
2318 shader_addline(ins
->ctx
->buffer
, "for (aL%u = %d; aL%u > (%d * %d + %d); aL%u += %d) {\n",
2319 shader
->baseShader
.cur_loop_depth
, control_values
[1],
2320 shader
->baseShader
.cur_loop_depth
, control_values
[0], control_values
[2], control_values
[1],
2321 shader
->baseShader
.cur_loop_depth
, control_values
[2]);
2324 shader_addline(ins
->ctx
->buffer
,
2325 "for (tmpInt%u = 0, aL%u = %s.y; tmpInt%u < %s.x; tmpInt%u++, aL%u += %s.z) {\n",
2326 shader
->baseShader
.cur_loop_depth
, shader
->baseShader
.cur_loop_regno
,
2327 src1_param
.reg_name
, shader
->baseShader
.cur_loop_depth
, src1_param
.reg_name
,
2328 shader
->baseShader
.cur_loop_depth
, shader
->baseShader
.cur_loop_regno
, src1_param
.reg_name
);
2331 shader
->baseShader
.cur_loop_depth
++;
2332 shader
->baseShader
.cur_loop_regno
++;
2335 static void shader_glsl_end(const struct wined3d_shader_instruction
*ins
)
2337 IWineD3DBaseShaderImpl
*shader
= (IWineD3DBaseShaderImpl
*)ins
->ctx
->shader
;
2339 shader_addline(ins
->ctx
->buffer
, "}\n");
2341 if (ins
->handler_idx
== WINED3DSIH_ENDLOOP
)
2343 shader
->baseShader
.cur_loop_depth
--;
2344 shader
->baseShader
.cur_loop_regno
--;
2347 if (ins
->handler_idx
== WINED3DSIH_ENDREP
)
2349 shader
->baseShader
.cur_loop_depth
--;
2353 static void shader_glsl_rep(const struct wined3d_shader_instruction
*ins
)
2355 IWineD3DBaseShaderImpl
*shader
= (IWineD3DBaseShaderImpl
*)ins
->ctx
->shader
;
2356 glsl_src_param_t src0_param
;
2358 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2359 shader_addline(ins
->ctx
->buffer
, "for (tmpInt%d = 0; tmpInt%d < %s; tmpInt%d++) {\n",
2360 shader
->baseShader
.cur_loop_depth
, shader
->baseShader
.cur_loop_depth
,
2361 src0_param
.param_str
, shader
->baseShader
.cur_loop_depth
);
2362 shader
->baseShader
.cur_loop_depth
++;
2365 static void shader_glsl_if(const struct wined3d_shader_instruction
*ins
)
2367 glsl_src_param_t src0_param
;
2369 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2370 shader_addline(ins
->ctx
->buffer
, "if (%s) {\n", src0_param
.param_str
);
2373 static void shader_glsl_ifc(const struct wined3d_shader_instruction
*ins
)
2375 glsl_src_param_t src0_param
;
2376 glsl_src_param_t src1_param
;
2378 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2379 shader_glsl_add_src_param(ins
, ins
->src
[1], ins
->src_addr
[1], WINED3DSP_WRITEMASK_0
, &src1_param
);
2381 shader_addline(ins
->ctx
->buffer
, "if (%s %s %s) {\n",
2382 src0_param
.param_str
, shader_get_comp_op(ins
->flags
), src1_param
.param_str
);
2385 static void shader_glsl_else(const struct wined3d_shader_instruction
*ins
)
2387 shader_addline(ins
->ctx
->buffer
, "} else {\n");
2390 static void shader_glsl_break(const struct wined3d_shader_instruction
*ins
)
2392 shader_addline(ins
->ctx
->buffer
, "break;\n");
2395 /* FIXME: According to MSDN the compare is done per component. */
2396 static void shader_glsl_breakc(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], ins
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2402 shader_glsl_add_src_param(ins
, ins
->src
[1], ins
->src_addr
[1], WINED3DSP_WRITEMASK_0
, &src1_param
);
2404 shader_addline(ins
->ctx
->buffer
, "if (%s %s %s) break;\n",
2405 src0_param
.param_str
, shader_get_comp_op(ins
->flags
), src1_param
.param_str
);
2408 static void shader_glsl_label(const struct wined3d_shader_instruction
*ins
)
2411 DWORD snum
= (ins
->src
[0]) & WINED3DSP_REGNUM_MASK
;
2412 shader_addline(ins
->ctx
->buffer
, "}\n");
2413 shader_addline(ins
->ctx
->buffer
, "void subroutine%u () {\n", snum
);
2416 static void shader_glsl_call(const struct wined3d_shader_instruction
*ins
)
2418 DWORD snum
= (ins
->src
[0]) & WINED3DSP_REGNUM_MASK
;
2419 shader_addline(ins
->ctx
->buffer
, "subroutine%u();\n", snum
);
2422 static void shader_glsl_callnz(const struct wined3d_shader_instruction
*ins
)
2424 glsl_src_param_t src1_param
;
2426 DWORD snum
= (ins
->src
[0]) & WINED3DSP_REGNUM_MASK
;
2427 shader_glsl_add_src_param(ins
, ins
->src
[1], ins
->src_addr
[1], WINED3DSP_WRITEMASK_0
, &src1_param
);
2428 shader_addline(ins
->ctx
->buffer
, "if (%s) subroutine%u();\n", src1_param
.param_str
, snum
);
2431 /*********************************************
2432 * Pixel Shader Specific Code begins here
2433 ********************************************/
2434 static void pshader_glsl_tex(const struct wined3d_shader_instruction
*ins
)
2436 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*)ins
->ctx
->shader
;
2437 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
2438 DWORD shader_version
= ins
->ctx
->reg_maps
->shader_version
;
2439 glsl_sample_function_t sample_function
;
2440 DWORD sample_flags
= 0;
2443 DWORD mask
= 0, swizzle
;
2445 /* 1.0-1.4: Use destination register as sampler source.
2446 * 2.0+: Use provided sampler source. */
2447 if (shader_version
< WINED3DPS_VERSION(2,0)) sampler_idx
= ins
->dst
[0].register_idx
;
2448 else sampler_idx
= ins
->src
[1] & WINED3DSP_REGNUM_MASK
;
2449 sampler_type
= ins
->ctx
->reg_maps
->samplers
[sampler_idx
] & WINED3DSP_TEXTURETYPE_MASK
;
2451 if (shader_version
< WINED3DPS_VERSION(1,4))
2453 DWORD flags
= deviceImpl
->stateBlock
->textureState
[sampler_idx
][WINED3DTSS_TEXTURETRANSFORMFLAGS
];
2455 /* Projected cube textures don't make a lot of sense, the resulting coordinates stay the same. */
2456 if (flags
& WINED3DTTFF_PROJECTED
&& sampler_type
!= WINED3DSTT_CUBE
) {
2457 sample_flags
|= WINED3D_GLSL_SAMPLE_PROJECTED
;
2458 switch (flags
& ~WINED3DTTFF_PROJECTED
) {
2459 case WINED3DTTFF_COUNT1
: FIXME("WINED3DTTFF_PROJECTED with WINED3DTTFF_COUNT1?\n"); break;
2460 case WINED3DTTFF_COUNT2
: mask
= WINED3DSP_WRITEMASK_1
; break;
2461 case WINED3DTTFF_COUNT3
: mask
= WINED3DSP_WRITEMASK_2
; break;
2462 case WINED3DTTFF_COUNT4
:
2463 case WINED3DTTFF_DISABLE
: mask
= WINED3DSP_WRITEMASK_3
; break;
2467 else if (shader_version
< WINED3DPS_VERSION(2,0))
2469 DWORD src_mod
= ins
->src
[0] & WINED3DSP_SRCMOD_MASK
;
2471 if (src_mod
== WINED3DSPSM_DZ
) {
2472 sample_flags
|= WINED3D_GLSL_SAMPLE_PROJECTED
;
2473 mask
= WINED3DSP_WRITEMASK_2
;
2474 } else if (src_mod
== WINED3DSPSM_DW
) {
2475 sample_flags
|= WINED3D_GLSL_SAMPLE_PROJECTED
;
2476 mask
= WINED3DSP_WRITEMASK_3
;
2479 if (ins
->flags
& WINED3DSI_TEXLD_PROJECT
)
2481 /* ps 2.0 texldp instruction always divides by the fourth component. */
2482 sample_flags
|= WINED3D_GLSL_SAMPLE_PROJECTED
;
2483 mask
= WINED3DSP_WRITEMASK_3
;
2487 if(deviceImpl
->stateBlock
->textures
[sampler_idx
] &&
2488 IWineD3DBaseTexture_GetTextureDimensions(deviceImpl
->stateBlock
->textures
[sampler_idx
]) == GL_TEXTURE_RECTANGLE_ARB
) {
2489 sample_flags
|= WINED3D_GLSL_SAMPLE_RECT
;
2492 shader_glsl_get_sample_function(sampler_type
, sample_flags
, &sample_function
);
2493 mask
|= sample_function
.coord_mask
;
2495 if (shader_version
< WINED3DPS_VERSION(2,0)) swizzle
= WINED3DVS_NOSWIZZLE
;
2496 else swizzle
= ins
->src
[1] & WINED3DSP_SWIZZLE_MASK
;
2498 /* 1.0-1.3: Use destination register as coordinate source.
2499 1.4+: Use provided coordinate source register. */
2500 if (shader_version
< WINED3DPS_VERSION(1,4))
2503 shader_glsl_write_mask_to_str(mask
, coord_mask
);
2504 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, swizzle
, NULL
,
2505 "T%u%s", sampler_idx
, coord_mask
);
2507 glsl_src_param_t coord_param
;
2508 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], mask
, &coord_param
);
2509 if (ins
->flags
& WINED3DSI_TEXLD_BIAS
)
2511 glsl_src_param_t bias
;
2512 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], WINED3DSP_WRITEMASK_3
, &bias
);
2513 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, swizzle
, bias
.param_str
,
2514 "%s", coord_param
.param_str
);
2516 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, swizzle
, NULL
,
2517 "%s", coord_param
.param_str
);
2522 static void shader_glsl_texldl(const struct wined3d_shader_instruction
*ins
)
2524 IWineD3DBaseShaderImpl
*This
= (IWineD3DBaseShaderImpl
*)ins
->ctx
->shader
;
2525 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
2526 glsl_sample_function_t sample_function
;
2527 glsl_src_param_t coord_param
, lod_param
;
2528 DWORD sample_flags
= WINED3D_GLSL_SAMPLE_LOD
;
2531 DWORD swizzle
= ins
->src
[1] & WINED3DSP_SWIZZLE_MASK
;
2533 sampler_idx
= ins
->src
[1] & WINED3DSP_REGNUM_MASK
;
2534 sampler_type
= ins
->ctx
->reg_maps
->samplers
[sampler_idx
] & WINED3DSP_TEXTURETYPE_MASK
;
2535 if(deviceImpl
->stateBlock
->textures
[sampler_idx
] &&
2536 IWineD3DBaseTexture_GetTextureDimensions(deviceImpl
->stateBlock
->textures
[sampler_idx
]) == GL_TEXTURE_RECTANGLE_ARB
) {
2537 sample_flags
|= WINED3D_GLSL_SAMPLE_RECT
;
2539 shader_glsl_get_sample_function(sampler_type
, sample_flags
, &sample_function
);
2540 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], sample_function
.coord_mask
, &coord_param
);
2542 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], WINED3DSP_WRITEMASK_3
, &lod_param
);
2544 if (shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
))
2546 /* The GLSL spec claims the Lod sampling functions are only supported in vertex shaders.
2547 * However, they seem to work just fine in fragment shaders as well. */
2548 WARN("Using %s in fragment shader.\n", sample_function
.name
);
2550 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, swizzle
, lod_param
.param_str
,
2551 "%s", coord_param
.param_str
);
2554 static void pshader_glsl_texcoord(const struct wined3d_shader_instruction
*ins
)
2556 /* FIXME: Make this work for more than just 2D textures */
2557 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
2558 DWORD write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2560 if (ins
->ctx
->reg_maps
->shader_version
!= WINED3DPS_VERSION(1,4))
2564 shader_glsl_get_write_mask(&ins
->dst
[0], dst_mask
);
2565 shader_addline(buffer
, "clamp(gl_TexCoord[%u], 0.0, 1.0)%s);\n",
2566 ins
->dst
[0].register_idx
, dst_mask
);
2568 DWORD reg
= ins
->src
[0] & WINED3DSP_REGNUM_MASK
;
2569 DWORD src_mod
= ins
->src
[0] & WINED3DSP_SRCMOD_MASK
;
2570 char dst_swizzle
[6];
2572 shader_glsl_get_swizzle(ins
->src
[0], FALSE
, write_mask
, dst_swizzle
);
2574 if (src_mod
== WINED3DSPSM_DZ
) {
2575 glsl_src_param_t div_param
;
2576 unsigned int mask_size
= shader_glsl_get_write_mask_size(write_mask
);
2577 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], WINED3DSP_WRITEMASK_2
, &div_param
);
2579 if (mask_size
> 1) {
2580 shader_addline(buffer
, "gl_TexCoord[%u]%s / vec%d(%s));\n", reg
, dst_swizzle
, mask_size
, div_param
.param_str
);
2582 shader_addline(buffer
, "gl_TexCoord[%u]%s / %s);\n", reg
, dst_swizzle
, div_param
.param_str
);
2584 } else if (src_mod
== WINED3DSPSM_DW
) {
2585 glsl_src_param_t div_param
;
2586 unsigned int mask_size
= shader_glsl_get_write_mask_size(write_mask
);
2587 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], WINED3DSP_WRITEMASK_3
, &div_param
);
2589 if (mask_size
> 1) {
2590 shader_addline(buffer
, "gl_TexCoord[%u]%s / vec%d(%s));\n", reg
, dst_swizzle
, mask_size
, div_param
.param_str
);
2592 shader_addline(buffer
, "gl_TexCoord[%u]%s / %s);\n", reg
, dst_swizzle
, div_param
.param_str
);
2595 shader_addline(buffer
, "gl_TexCoord[%u]%s);\n", reg
, dst_swizzle
);
2600 /** Process the WINED3DSIO_TEXDP3TEX instruction in GLSL:
2601 * Take a 3-component dot product of the TexCoord[dstreg] and src,
2602 * then perform a 1D texture lookup from stage dstregnum, place into dst. */
2603 static void pshader_glsl_texdp3tex(const struct wined3d_shader_instruction
*ins
)
2605 glsl_src_param_t src0_param
;
2606 glsl_sample_function_t sample_function
;
2607 DWORD sampler_idx
= ins
->dst
[0].register_idx
;
2608 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2609 DWORD sampler_type
= ins
->ctx
->reg_maps
->samplers
[sampler_idx
] & WINED3DSP_TEXTURETYPE_MASK
;
2612 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], src_mask
, &src0_param
);
2614 /* Do I have to take care about the projected bit? I don't think so, since the dp3 returns only one
2615 * scalar, and projected sampling would require 4.
2617 * It is a dependent read - not valid with conditional NP2 textures
2619 shader_glsl_get_sample_function(sampler_type
, 0, &sample_function
);
2620 mask_size
= shader_glsl_get_write_mask_size(sample_function
.coord_mask
);
2625 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, WINED3DVS_NOSWIZZLE
, NULL
,
2626 "dot(gl_TexCoord[%u].xyz, %s)", sampler_idx
, src0_param
.param_str
);
2630 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, WINED3DVS_NOSWIZZLE
, NULL
,
2631 "vec2(dot(gl_TexCoord[%u].xyz, %s), 0.0)", sampler_idx
, src0_param
.param_str
);
2635 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, WINED3DVS_NOSWIZZLE
, NULL
,
2636 "vec3(dot(gl_TexCoord[%u].xyz, %s), 0.0, 0.0)", sampler_idx
, src0_param
.param_str
);
2640 FIXME("Unexpected mask size %u\n", mask_size
);
2645 /** Process the WINED3DSIO_TEXDP3 instruction in GLSL:
2646 * Take a 3-component dot product of the TexCoord[dstreg] and src. */
2647 static void pshader_glsl_texdp3(const struct wined3d_shader_instruction
*ins
)
2649 glsl_src_param_t src0_param
;
2650 DWORD dstreg
= ins
->dst
[0].register_idx
;
2651 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2653 unsigned int mask_size
;
2655 dst_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2656 mask_size
= shader_glsl_get_write_mask_size(dst_mask
);
2657 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], src_mask
, &src0_param
);
2659 if (mask_size
> 1) {
2660 shader_addline(ins
->ctx
->buffer
, "vec%d(dot(T%u.xyz, %s)));\n", mask_size
, dstreg
, src0_param
.param_str
);
2662 shader_addline(ins
->ctx
->buffer
, "dot(T%u.xyz, %s));\n", dstreg
, src0_param
.param_str
);
2666 /** Process the WINED3DSIO_TEXDEPTH instruction in GLSL:
2667 * Calculate the depth as dst.x / dst.y */
2668 static void pshader_glsl_texdepth(const struct wined3d_shader_instruction
*ins
)
2670 glsl_dst_param_t dst_param
;
2672 shader_glsl_add_dst_param(ins
, &ins
->dst
[0], &dst_param
);
2674 /* Tests show that texdepth never returns anything below 0.0, and that r5.y is clamped to 1.0.
2675 * Negative input is accepted, -0.25 / -0.5 returns 0.5. GL should clamp gl_FragDepth to [0;1], but
2676 * this doesn't always work, so clamp the results manually. Whether or not the x value is clamped at 1
2677 * too is irrelevant, since if x = 0, any y value < 1.0 (and > 1.0 is not allowed) results in a result
2680 shader_addline(ins
->ctx
->buffer
, "gl_FragDepth = clamp((%s.x / min(%s.y, 1.0)), 0.0, 1.0);\n",
2681 dst_param
.reg_name
, dst_param
.reg_name
);
2684 /** Process the WINED3DSIO_TEXM3X2DEPTH instruction in GLSL:
2685 * Last row of a 3x2 matrix multiply, use the result to calculate the depth:
2686 * Calculate tmp0.y = TexCoord[dstreg] . src.xyz; (tmp0.x has already been calculated)
2687 * depth = (tmp0.y == 0.0) ? 1.0 : tmp0.x / tmp0.y
2689 static void pshader_glsl_texm3x2depth(const struct wined3d_shader_instruction
*ins
)
2691 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2692 DWORD dstreg
= ins
->dst
[0].register_idx
;
2693 glsl_src_param_t src0_param
;
2695 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], src_mask
, &src0_param
);
2697 shader_addline(ins
->ctx
->buffer
, "tmp0.y = dot(T%u.xyz, %s);\n", dstreg
, src0_param
.param_str
);
2698 shader_addline(ins
->ctx
->buffer
, "gl_FragDepth = (tmp0.y == 0.0) ? 1.0 : clamp(tmp0.x / tmp0.y, 0.0, 1.0);\n");
2701 /** Process the WINED3DSIO_TEXM3X2PAD instruction in GLSL
2702 * Calculate the 1st of a 2-row matrix multiplication. */
2703 static void pshader_glsl_texm3x2pad(const struct wined3d_shader_instruction
*ins
)
2705 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2706 DWORD reg
= ins
->dst
[0].register_idx
;
2707 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
2708 glsl_src_param_t src0_param
;
2710 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], src_mask
, &src0_param
);
2711 shader_addline(buffer
, "tmp0.x = dot(T%u.xyz, %s);\n", reg
, src0_param
.param_str
);
2714 /** Process the WINED3DSIO_TEXM3X3PAD instruction in GLSL
2715 * Calculate the 1st or 2nd row of a 3-row matrix multiplication. */
2716 static void pshader_glsl_texm3x3pad(const struct wined3d_shader_instruction
*ins
)
2718 IWineD3DPixelShaderImpl
*shader
= (IWineD3DPixelShaderImpl
*)ins
->ctx
->shader
;
2719 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2720 DWORD reg
= ins
->dst
[0].register_idx
;
2721 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
2722 SHADER_PARSE_STATE
* current_state
= &shader
->baseShader
.parse_state
;
2723 glsl_src_param_t src0_param
;
2725 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], src_mask
, &src0_param
);
2726 shader_addline(buffer
, "tmp0.%c = dot(T%u.xyz, %s);\n", 'x' + current_state
->current_row
, reg
, src0_param
.param_str
);
2727 current_state
->texcoord_w
[current_state
->current_row
++] = reg
;
2730 static void pshader_glsl_texm3x2tex(const struct wined3d_shader_instruction
*ins
)
2732 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2733 DWORD reg
= ins
->dst
[0].register_idx
;
2734 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
2735 glsl_src_param_t src0_param
;
2736 DWORD sampler_type
= ins
->ctx
->reg_maps
->samplers
[reg
] & WINED3DSP_TEXTURETYPE_MASK
;
2737 glsl_sample_function_t sample_function
;
2739 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], src_mask
, &src0_param
);
2740 shader_addline(buffer
, "tmp0.y = dot(T%u.xyz, %s);\n", reg
, src0_param
.param_str
);
2742 shader_glsl_get_sample_function(sampler_type
, 0, &sample_function
);
2744 /* Sample the texture using the calculated coordinates */
2745 shader_glsl_gen_sample_code(ins
, reg
, &sample_function
, WINED3DVS_NOSWIZZLE
, NULL
, "tmp0.xy");
2748 /** Process the WINED3DSIO_TEXM3X3TEX instruction in GLSL
2749 * Perform the 3rd row of a 3x3 matrix multiply, then sample the texture using the calculated coordinates */
2750 static void pshader_glsl_texm3x3tex(const struct wined3d_shader_instruction
*ins
)
2752 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2753 glsl_src_param_t src0_param
;
2754 DWORD reg
= ins
->dst
[0].register_idx
;
2755 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*)ins
->ctx
->shader
;
2756 SHADER_PARSE_STATE
* current_state
= &This
->baseShader
.parse_state
;
2757 DWORD sampler_type
= ins
->ctx
->reg_maps
->samplers
[reg
] & WINED3DSP_TEXTURETYPE_MASK
;
2758 glsl_sample_function_t sample_function
;
2760 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], src_mask
, &src0_param
);
2761 shader_addline(ins
->ctx
->buffer
, "tmp0.z = dot(T%u.xyz, %s);\n", reg
, src0_param
.param_str
);
2763 /* Dependent read, not valid with conditional NP2 */
2764 shader_glsl_get_sample_function(sampler_type
, 0, &sample_function
);
2766 /* Sample the texture using the calculated coordinates */
2767 shader_glsl_gen_sample_code(ins
, reg
, &sample_function
, WINED3DVS_NOSWIZZLE
, NULL
, "tmp0.xyz");
2769 current_state
->current_row
= 0;
2772 /** Process the WINED3DSIO_TEXM3X3 instruction in GLSL
2773 * Perform the 3rd row of a 3x3 matrix multiply */
2774 static void pshader_glsl_texm3x3(const struct wined3d_shader_instruction
*ins
)
2776 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2777 glsl_src_param_t src0_param
;
2779 DWORD reg
= ins
->dst
[0].register_idx
;
2780 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*)ins
->ctx
->shader
;
2781 SHADER_PARSE_STATE
* current_state
= &This
->baseShader
.parse_state
;
2783 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], src_mask
, &src0_param
);
2785 shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2786 shader_glsl_get_write_mask(&ins
->dst
[0], dst_mask
);
2787 shader_addline(ins
->ctx
->buffer
, "vec4(tmp0.xy, dot(T%u.xyz, %s), 1.0)%s);\n", reg
, src0_param
.param_str
, dst_mask
);
2789 current_state
->current_row
= 0;
2792 /** Process the WINED3DSIO_TEXM3X3SPEC instruction in GLSL
2793 * Perform the final texture lookup based on the previous 2 3x3 matrix multiplies */
2794 static void pshader_glsl_texm3x3spec(const struct wined3d_shader_instruction
*ins
)
2796 IWineD3DPixelShaderImpl
*shader
= (IWineD3DPixelShaderImpl
*)ins
->ctx
->shader
;
2797 DWORD reg
= ins
->dst
[0].register_idx
;
2798 glsl_src_param_t src0_param
;
2799 glsl_src_param_t src1_param
;
2800 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
2801 SHADER_PARSE_STATE
* current_state
= &shader
->baseShader
.parse_state
;
2802 DWORD stype
= ins
->ctx
->reg_maps
->samplers
[reg
] & WINED3DSP_TEXTURETYPE_MASK
;
2803 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2804 glsl_sample_function_t sample_function
;
2806 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], src_mask
, &src0_param
);
2807 shader_glsl_add_src_param(ins
, ins
->src
[1], ins
->src_addr
[1], src_mask
, &src1_param
);
2809 /* Perform the last matrix multiply operation */
2810 shader_addline(buffer
, "tmp0.z = dot(T%u.xyz, %s);\n", reg
, src0_param
.param_str
);
2811 /* Reflection calculation */
2812 shader_addline(buffer
, "tmp0.xyz = -reflect((%s), normalize(tmp0.xyz));\n", src1_param
.param_str
);
2814 /* Dependent read, not valid with conditional NP2 */
2815 shader_glsl_get_sample_function(stype
, 0, &sample_function
);
2817 /* Sample the texture */
2818 shader_glsl_gen_sample_code(ins
, reg
, &sample_function
, WINED3DVS_NOSWIZZLE
, NULL
, "tmp0.xyz");
2820 current_state
->current_row
= 0;
2823 /** Process the WINED3DSIO_TEXM3X3VSPEC instruction in GLSL
2824 * Perform the final texture lookup based on the previous 2 3x3 matrix multiplies */
2825 static void pshader_glsl_texm3x3vspec(const struct wined3d_shader_instruction
*ins
)
2827 IWineD3DPixelShaderImpl
*shader
= (IWineD3DPixelShaderImpl
*)ins
->ctx
->shader
;
2828 DWORD reg
= ins
->dst
[0].register_idx
;
2829 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
2830 SHADER_PARSE_STATE
* current_state
= &shader
->baseShader
.parse_state
;
2831 glsl_src_param_t src0_param
;
2832 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2833 DWORD sampler_type
= ins
->ctx
->reg_maps
->samplers
[reg
] & WINED3DSP_TEXTURETYPE_MASK
;
2834 glsl_sample_function_t sample_function
;
2836 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], src_mask
, &src0_param
);
2838 /* Perform the last matrix multiply operation */
2839 shader_addline(buffer
, "tmp0.z = dot(vec3(T%u), vec3(%s));\n", reg
, src0_param
.param_str
);
2841 /* Construct the eye-ray vector from w coordinates */
2842 shader_addline(buffer
, "tmp1.xyz = normalize(vec3(gl_TexCoord[%u].w, gl_TexCoord[%u].w, gl_TexCoord[%u].w));\n",
2843 current_state
->texcoord_w
[0], current_state
->texcoord_w
[1], reg
);
2844 shader_addline(buffer
, "tmp0.xyz = -reflect(tmp1.xyz, normalize(tmp0.xyz));\n");
2846 /* Dependent read, not valid with conditional NP2 */
2847 shader_glsl_get_sample_function(sampler_type
, 0, &sample_function
);
2849 /* Sample the texture using the calculated coordinates */
2850 shader_glsl_gen_sample_code(ins
, reg
, &sample_function
, WINED3DVS_NOSWIZZLE
, NULL
, "tmp0.xyz");
2852 current_state
->current_row
= 0;
2855 /** Process the WINED3DSIO_TEXBEM instruction in GLSL.
2856 * Apply a fake bump map transform.
2857 * texbem is pshader <= 1.3 only, this saves a few version checks
2859 static void pshader_glsl_texbem(const struct wined3d_shader_instruction
*ins
)
2861 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*)ins
->ctx
->shader
;
2862 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
2863 glsl_sample_function_t sample_function
;
2864 glsl_src_param_t coord_param
;
2871 sampler_idx
= ins
->dst
[0].register_idx
;
2872 flags
= deviceImpl
->stateBlock
->textureState
[sampler_idx
][WINED3DTSS_TEXTURETRANSFORMFLAGS
];
2874 sampler_type
= ins
->ctx
->reg_maps
->samplers
[sampler_idx
] & WINED3DSP_TEXTURETYPE_MASK
;
2875 /* Dependent read, not valid with conditional NP2 */
2876 shader_glsl_get_sample_function(sampler_type
, 0, &sample_function
);
2877 mask
= sample_function
.coord_mask
;
2879 shader_glsl_write_mask_to_str(mask
, coord_mask
);
2881 /* with projective textures, texbem only divides the static texture coord, not the displacement,
2882 * so we can't let the GL handle this.
2884 if (flags
& WINED3DTTFF_PROJECTED
) {
2886 char coord_div_mask
[3];
2887 switch (flags
& ~WINED3DTTFF_PROJECTED
) {
2888 case WINED3DTTFF_COUNT1
: FIXME("WINED3DTTFF_PROJECTED with WINED3DTTFF_COUNT1?\n"); break;
2889 case WINED3DTTFF_COUNT2
: div_mask
= WINED3DSP_WRITEMASK_1
; break;
2890 case WINED3DTTFF_COUNT3
: div_mask
= WINED3DSP_WRITEMASK_2
; break;
2891 case WINED3DTTFF_COUNT4
:
2892 case WINED3DTTFF_DISABLE
: div_mask
= WINED3DSP_WRITEMASK_3
; break;
2894 shader_glsl_write_mask_to_str(div_mask
, coord_div_mask
);
2895 shader_addline(ins
->ctx
->buffer
, "T%u%s /= T%u%s;\n", sampler_idx
, coord_mask
, sampler_idx
, coord_div_mask
);
2898 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0],
2899 WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
, &coord_param
);
2901 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, WINED3DVS_NOSWIZZLE
, NULL
,
2902 "T%u%s + vec4(bumpenvmat%d * %s, 0.0, 0.0)%s", sampler_idx
, coord_mask
, sampler_idx
,
2903 coord_param
.param_str
, coord_mask
);
2905 if (ins
->handler_idx
== WINED3DSIH_TEXBEML
)
2907 glsl_src_param_t luminance_param
;
2908 glsl_dst_param_t dst_param
;
2910 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], WINED3DSP_WRITEMASK_2
, &luminance_param
);
2911 shader_glsl_add_dst_param(ins
, &ins
->dst
[0], &dst_param
);
2913 shader_addline(ins
->ctx
->buffer
, "%s%s *= (%s * luminancescale%d + luminanceoffset%d);\n",
2914 dst_param
.reg_name
, dst_param
.mask_str
,
2915 luminance_param
.param_str
, sampler_idx
, sampler_idx
);
2919 static void pshader_glsl_bem(const struct wined3d_shader_instruction
*ins
)
2921 glsl_src_param_t src0_param
, src1_param
;
2922 DWORD sampler_idx
= ins
->dst
[0].register_idx
;
2924 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0],
2925 WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
, &src0_param
);
2926 shader_glsl_add_src_param(ins
, ins
->src
[1], ins
->src_addr
[1],
2927 WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
, &src1_param
);
2929 shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2930 shader_addline(ins
->ctx
->buffer
, "%s + bumpenvmat%d * %s);\n",
2931 src0_param
.param_str
, sampler_idx
, src1_param
.param_str
);
2934 /** Process the WINED3DSIO_TEXREG2AR instruction in GLSL
2935 * Sample 2D texture at dst using the alpha & red (wx) components of src as texture coordinates */
2936 static void pshader_glsl_texreg2ar(const struct wined3d_shader_instruction
*ins
)
2938 glsl_src_param_t src0_param
;
2939 DWORD sampler_idx
= ins
->dst
[0].register_idx
;
2940 DWORD sampler_type
= ins
->ctx
->reg_maps
->samplers
[sampler_idx
] & WINED3DSP_TEXTURETYPE_MASK
;
2941 glsl_sample_function_t sample_function
;
2943 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], WINED3DSP_WRITEMASK_ALL
, &src0_param
);
2945 shader_glsl_get_sample_function(sampler_type
, 0, &sample_function
);
2946 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, WINED3DVS_NOSWIZZLE
, NULL
,
2947 "%s.wx", src0_param
.reg_name
);
2950 /** Process the WINED3DSIO_TEXREG2GB instruction in GLSL
2951 * Sample 2D texture at dst using the green & blue (yz) components of src as texture coordinates */
2952 static void pshader_glsl_texreg2gb(const struct wined3d_shader_instruction
*ins
)
2954 glsl_src_param_t src0_param
;
2955 DWORD sampler_idx
= ins
->dst
[0].register_idx
;
2956 DWORD sampler_type
= ins
->ctx
->reg_maps
->samplers
[sampler_idx
] & WINED3DSP_TEXTURETYPE_MASK
;
2957 glsl_sample_function_t sample_function
;
2959 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[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
, WINED3DVS_NOSWIZZLE
, NULL
,
2963 "%s.yz", src0_param
.reg_name
);
2966 /** Process the WINED3DSIO_TEXREG2RGB instruction in GLSL
2967 * Sample texture at dst using the rgb (xyz) components of src as texture coordinates */
2968 static void pshader_glsl_texreg2rgb(const struct wined3d_shader_instruction
*ins
)
2970 glsl_src_param_t src0_param
;
2971 DWORD sampler_idx
= ins
->dst
[0].register_idx
;
2972 DWORD sampler_type
= ins
->ctx
->reg_maps
->samplers
[sampler_idx
] & WINED3DSP_TEXTURETYPE_MASK
;
2973 glsl_sample_function_t sample_function
;
2975 /* Dependent read, not valid with conditional NP2 */
2976 shader_glsl_get_sample_function(sampler_type
, 0, &sample_function
);
2977 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], sample_function
.coord_mask
, &src0_param
);
2979 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, WINED3DVS_NOSWIZZLE
, NULL
,
2980 "%s", src0_param
.param_str
);
2983 /** Process the WINED3DSIO_TEXKILL instruction in GLSL.
2984 * If any of the first 3 components are < 0, discard this pixel */
2985 static void pshader_glsl_texkill(const struct wined3d_shader_instruction
*ins
)
2987 glsl_dst_param_t dst_param
;
2989 /* The argument is a destination parameter, and no writemasks are allowed */
2990 shader_glsl_add_dst_param(ins
, &ins
->dst
[0], &dst_param
);
2991 if ((ins
->ctx
->reg_maps
->shader_version
>= WINED3DPS_VERSION(2,0)))
2993 /* 2.0 shaders compare all 4 components in texkill */
2994 shader_addline(ins
->ctx
->buffer
, "if (any(lessThan(%s.xyzw, vec4(0.0)))) discard;\n", dst_param
.reg_name
);
2996 /* 1.X shaders only compare the first 3 components, probably due to the nature of the texkill
2997 * instruction as a tex* instruction, and phase, which kills all a / w components. Even if all
2998 * 4 components are defined, only the first 3 are used
3000 shader_addline(ins
->ctx
->buffer
, "if (any(lessThan(%s.xyz, vec3(0.0)))) discard;\n", dst_param
.reg_name
);
3004 /** Process the WINED3DSIO_DP2ADD instruction in GLSL.
3005 * dst = dot2(src0, src1) + src2 */
3006 static void pshader_glsl_dp2add(const struct wined3d_shader_instruction
*ins
)
3008 glsl_src_param_t src0_param
;
3009 glsl_src_param_t src1_param
;
3010 glsl_src_param_t src2_param
;
3012 unsigned int mask_size
;
3014 write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
3015 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
3017 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0],
3018 WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
, &src0_param
);
3019 shader_glsl_add_src_param(ins
, ins
->src
[1], ins
->src_addr
[1],
3020 WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
, &src1_param
);
3021 shader_glsl_add_src_param(ins
, ins
->src
[2], ins
->src_addr
[2],
3022 WINED3DSP_WRITEMASK_0
, &src2_param
);
3024 if (mask_size
> 1) {
3025 shader_addline(ins
->ctx
->buffer
, "vec%d(dot(%s, %s) + %s));\n",
3026 mask_size
, src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
3028 shader_addline(ins
->ctx
->buffer
, "dot(%s, %s) + %s);\n",
3029 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
3033 static void pshader_glsl_input_pack(IWineD3DPixelShader
*iface
, SHADER_BUFFER
*buffer
,
3034 const struct wined3d_shader_semantic
*semantics_in
, const struct shader_reg_maps
*reg_maps
,
3035 enum vertexprocessing_mode vertexprocessing
)
3038 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*)iface
;
3040 for (i
= 0; i
< MAX_REG_INPUT
; ++i
)
3042 DWORD usage
, usage_idx
;
3046 if (!reg_maps
->packed_input
[i
]) continue;
3048 usage
= semantics_in
[i
].usage
;
3049 usage_idx
= semantics_in
[i
].usage_idx
;
3050 shader_glsl_get_write_mask(&semantics_in
[i
].reg
, reg_mask
);
3054 case WINED3DDECLUSAGE_TEXCOORD
:
3055 if (usage_idx
< 8 && vertexprocessing
== pretransformed
)
3056 shader_addline(buffer
, "IN[%u]%s = gl_TexCoord[%u]%s;\n",
3057 This
->input_reg_map
[i
], reg_mask
, usage_idx
, reg_mask
);
3059 shader_addline(buffer
, "IN[%u]%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
3060 This
->input_reg_map
[i
], reg_mask
, reg_mask
);
3063 case WINED3DDECLUSAGE_COLOR
:
3065 shader_addline(buffer
, "IN[%u]%s = vec4(gl_Color)%s;\n",
3066 This
->input_reg_map
[i
], reg_mask
, reg_mask
);
3067 else if (usage_idx
== 1)
3068 shader_addline(buffer
, "IN[%u]%s = vec4(gl_SecondaryColor)%s;\n",
3069 This
->input_reg_map
[i
], reg_mask
, reg_mask
);
3071 shader_addline(buffer
, "IN[%u]%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
3072 This
->input_reg_map
[i
], reg_mask
, reg_mask
);
3076 shader_addline(buffer
, "IN[%u]%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
3077 This
->input_reg_map
[i
], reg_mask
, reg_mask
);
3083 /*********************************************
3084 * Vertex Shader Specific Code begins here
3085 ********************************************/
3087 static void add_glsl_program_entry(struct shader_glsl_priv
*priv
, struct glsl_shader_prog_link
*entry
) {
3088 glsl_program_key_t
*key
;
3090 key
= HeapAlloc(GetProcessHeap(), 0, sizeof(glsl_program_key_t
));
3091 key
->vshader
= entry
->vshader
;
3092 key
->pshader
= entry
->pshader
;
3093 key
->vs_args
= entry
->vs_args
;
3094 key
->ps_args
= entry
->ps_args
;
3096 hash_table_put(priv
->glsl_program_lookup
, key
, entry
);
3099 static struct glsl_shader_prog_link
*get_glsl_program_entry(struct shader_glsl_priv
*priv
,
3100 IWineD3DVertexShader
*vshader
, IWineD3DPixelShader
*pshader
, struct vs_compile_args
*vs_args
,
3101 struct ps_compile_args
*ps_args
) {
3102 glsl_program_key_t key
;
3104 key
.vshader
= vshader
;
3105 key
.pshader
= pshader
;
3106 key
.vs_args
= *vs_args
;
3107 key
.ps_args
= *ps_args
;
3109 return hash_table_get(priv
->glsl_program_lookup
, &key
);
3112 static void delete_glsl_program_entry(struct shader_glsl_priv
*priv
, const WineD3D_GL_Info
*gl_info
,
3113 struct glsl_shader_prog_link
*entry
)
3115 glsl_program_key_t
*key
;
3117 key
= HeapAlloc(GetProcessHeap(), 0, sizeof(glsl_program_key_t
));
3118 key
->vshader
= entry
->vshader
;
3119 key
->pshader
= entry
->pshader
;
3120 key
->vs_args
= entry
->vs_args
;
3121 key
->ps_args
= entry
->ps_args
;
3122 hash_table_remove(priv
->glsl_program_lookup
, key
);
3124 GL_EXTCALL(glDeleteObjectARB(entry
->programId
));
3125 if (entry
->vshader
) list_remove(&entry
->vshader_entry
);
3126 if (entry
->pshader
) list_remove(&entry
->pshader_entry
);
3127 HeapFree(GetProcessHeap(), 0, entry
->vuniformF_locations
);
3128 HeapFree(GetProcessHeap(), 0, entry
->puniformF_locations
);
3129 HeapFree(GetProcessHeap(), 0, entry
);
3132 static void handle_ps3_input(SHADER_BUFFER
*buffer
, const WineD3D_GL_Info
*gl_info
, const DWORD
*map
,
3133 const struct wined3d_shader_semantic
*semantics_in
, const struct shader_reg_maps
*reg_maps_in
,
3134 const struct wined3d_shader_semantic
*semantics_out
, const struct shader_reg_maps
*reg_maps_out
)
3137 DWORD usage
, usage_idx
, usage_out
, usage_idx_out
;
3140 DWORD in_count
= GL_LIMITS(glsl_varyings
) / 4;
3141 char reg_mask
[6], reg_mask_out
[6];
3142 char destination
[50];
3144 set
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*set
) * (in_count
+ 2));
3146 if (!semantics_out
) {
3147 /* Save gl_FrontColor & gl_FrontSecondaryColor before overwriting them. */
3148 shader_addline(buffer
, "vec4 front_color = gl_FrontColor;\n");
3149 shader_addline(buffer
, "vec4 front_secondary_color = gl_FrontSecondaryColor;\n");
3152 for(i
= 0; i
< MAX_REG_INPUT
; i
++) {
3153 if (!reg_maps_in
->packed_input
[i
]) continue;
3156 if (in_idx
>= (in_count
+ 2)) {
3157 FIXME("More input varyings declared than supported, expect issues\n");
3160 else if (map
[i
] == ~0U)
3162 /* Declared, but not read register */
3166 if (in_idx
== in_count
) {
3167 sprintf(destination
, "gl_FrontColor");
3168 } else if (in_idx
== in_count
+ 1) {
3169 sprintf(destination
, "gl_FrontSecondaryColor");
3171 sprintf(destination
, "IN[%u]", in_idx
);
3174 usage
= semantics_in
[i
].usage
;
3175 usage_idx
= semantics_in
[i
].usage_idx
;
3176 set
[map
[i
]] = shader_glsl_get_write_mask(&semantics_in
[i
].reg
, reg_mask
);
3178 if(!semantics_out
) {
3180 case WINED3DDECLUSAGE_COLOR
:
3182 shader_addline(buffer
, "%s%s = front_color%s;\n",
3183 destination
, reg_mask
, reg_mask
);
3184 else if (usage_idx
== 1)
3185 shader_addline(buffer
, "%s%s = front_secondary_color%s;\n",
3186 destination
, reg_mask
, reg_mask
);
3188 shader_addline(buffer
, "%s%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
3189 destination
, reg_mask
, reg_mask
);
3192 case WINED3DDECLUSAGE_TEXCOORD
:
3193 if (usage_idx
< 8) {
3194 shader_addline(buffer
, "%s%s = gl_TexCoord[%u]%s;\n",
3195 destination
, reg_mask
, usage_idx
, reg_mask
);
3197 shader_addline(buffer
, "%s%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
3198 destination
, reg_mask
, reg_mask
);
3202 case WINED3DDECLUSAGE_FOG
:
3203 shader_addline(buffer
, "%s%s = vec4(gl_FogFragCoord, 0.0, 0.0, 0.0)%s;\n",
3204 destination
, reg_mask
, reg_mask
);
3208 shader_addline(buffer
, "%s%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
3209 destination
, reg_mask
, reg_mask
);
3213 for(j
= 0; j
< MAX_REG_OUTPUT
; j
++) {
3214 if (!reg_maps_out
->packed_output
[j
]) continue;
3216 usage_out
= semantics_out
[j
].usage
;
3217 usage_idx_out
= semantics_out
[j
].usage_idx
;
3218 shader_glsl_get_write_mask(&semantics_out
[j
].reg
, reg_mask_out
);
3220 if(usage
== usage_out
&&
3221 usage_idx
== usage_idx_out
) {
3222 shader_addline(buffer
, "%s%s = OUT[%u]%s;\n",
3223 destination
, reg_mask
, j
, reg_mask
);
3228 shader_addline(buffer
, "%s%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
3229 destination
, reg_mask
, reg_mask
);
3234 /* This is solely to make the compiler / linker happy and avoid warning about undefined
3235 * varyings. It shouldn't result in any real code executed on the GPU, since all read
3236 * input varyings are assigned above, if the optimizer works properly.
3238 for(i
= 0; i
< in_count
+ 2; i
++) {
3239 if(set
[i
] != WINED3DSP_WRITEMASK_ALL
) {
3240 unsigned int size
= 0;
3241 memset(reg_mask
, 0, sizeof(reg_mask
));
3242 if(!(set
[i
] & WINED3DSP_WRITEMASK_0
)) {
3243 reg_mask
[size
] = 'x';
3246 if(!(set
[i
] & WINED3DSP_WRITEMASK_1
)) {
3247 reg_mask
[size
] = 'y';
3250 if(!(set
[i
] & WINED3DSP_WRITEMASK_2
)) {
3251 reg_mask
[size
] = 'z';
3254 if(!(set
[i
] & WINED3DSP_WRITEMASK_3
)) {
3255 reg_mask
[size
] = 'w';
3259 if (i
== in_count
) {
3260 sprintf(destination
, "gl_FrontColor");
3261 } else if (i
== in_count
+ 1) {
3262 sprintf(destination
, "gl_FrontSecondaryColor");
3264 sprintf(destination
, "IN[%u]", i
);
3268 shader_addline(buffer
, "%s.%s = 0.0;\n", destination
, reg_mask
);
3270 shader_addline(buffer
, "%s.%s = vec%u(0.0);\n", destination
, reg_mask
, size
);
3275 HeapFree(GetProcessHeap(), 0, set
);
3278 static GLhandleARB
generate_param_reorder_function(IWineD3DVertexShader
*vertexshader
,
3279 IWineD3DPixelShader
*pixelshader
, const WineD3D_GL_Info
*gl_info
)
3281 GLhandleARB ret
= 0;
3282 IWineD3DVertexShaderImpl
*vs
= (IWineD3DVertexShaderImpl
*) vertexshader
;
3283 IWineD3DPixelShaderImpl
*ps
= (IWineD3DPixelShaderImpl
*) pixelshader
;
3284 IWineD3DDeviceImpl
*device
;
3285 DWORD vs_major
= WINED3DSHADER_VERSION_MAJOR(vs
->baseShader
.reg_maps
.shader_version
);
3286 DWORD ps_major
= ps
? WINED3DSHADER_VERSION_MAJOR(ps
->baseShader
.reg_maps
.shader_version
) : 0;
3288 SHADER_BUFFER buffer
;
3289 DWORD usage
, usage_idx
, writemask
;
3291 const struct wined3d_shader_semantic
*semantics_out
;
3293 shader_buffer_init(&buffer
);
3295 shader_addline(&buffer
, "#version 120\n");
3297 if(vs_major
< 3 && ps_major
< 3) {
3298 /* That one is easy: The vertex shader writes to the builtin varyings, the pixel shader reads from them.
3299 * Take care about the texcoord .w fixup though if we're using the fixed function fragment pipeline
3301 device
= (IWineD3DDeviceImpl
*) vs
->baseShader
.device
;
3302 if((GLINFO_LOCATION
).set_texcoord_w
&& ps_major
== 0 && vs_major
> 0 &&
3303 !device
->frag_pipe
->ffp_proj_control
) {
3304 shader_addline(&buffer
, "void order_ps_input() {\n");
3305 for(i
= 0; i
< min(8, MAX_REG_TEXCRD
); i
++) {
3306 if(vs
->baseShader
.reg_maps
.texcoord_mask
[i
] != 0 &&
3307 vs
->baseShader
.reg_maps
.texcoord_mask
[i
] != WINED3DSP_WRITEMASK_ALL
) {
3308 shader_addline(&buffer
, "gl_TexCoord[%u].w = 1.0;\n", i
);
3311 shader_addline(&buffer
, "}\n");
3313 shader_addline(&buffer
, "void order_ps_input() { /* do nothing */ }\n");
3315 } else if(ps_major
< 3 && vs_major
>= 3) {
3316 /* The vertex shader writes to its own varyings, the pixel shader needs them in the builtin ones */
3317 semantics_out
= vs
->semantics_out
;
3319 shader_addline(&buffer
, "void order_ps_input(in vec4 OUT[%u]) {\n", MAX_REG_OUTPUT
);
3320 for(i
= 0; i
< MAX_REG_OUTPUT
; i
++) {
3321 if (!vs
->baseShader
.reg_maps
.packed_output
[i
]) continue;
3323 usage
= semantics_out
[i
].usage
;
3324 usage_idx
= semantics_out
[i
].usage_idx
;
3325 writemask
= shader_glsl_get_write_mask(&semantics_out
[i
].reg
, reg_mask
);
3328 case WINED3DDECLUSAGE_COLOR
:
3330 shader_addline(&buffer
, "gl_FrontColor%s = OUT[%u]%s;\n", reg_mask
, i
, reg_mask
);
3331 else if (usage_idx
== 1)
3332 shader_addline(&buffer
, "gl_FrontSecondaryColor%s = OUT[%u]%s;\n", reg_mask
, i
, reg_mask
);
3335 case WINED3DDECLUSAGE_POSITION
:
3336 shader_addline(&buffer
, "gl_Position%s = OUT[%u]%s;\n", reg_mask
, i
, reg_mask
);
3339 case WINED3DDECLUSAGE_TEXCOORD
:
3340 if (usage_idx
< 8) {
3341 if(!(GLINFO_LOCATION
).set_texcoord_w
|| ps_major
> 0) writemask
|= WINED3DSP_WRITEMASK_3
;
3343 shader_addline(&buffer
, "gl_TexCoord[%u]%s = OUT[%u]%s;\n",
3344 usage_idx
, reg_mask
, i
, reg_mask
);
3345 if(!(writemask
& WINED3DSP_WRITEMASK_3
)) {
3346 shader_addline(&buffer
, "gl_TexCoord[%u].w = 1.0;\n", usage_idx
);
3351 case WINED3DDECLUSAGE_PSIZE
:
3352 shader_addline(&buffer
, "gl_PointSize = OUT[%u].x;\n", i
);
3355 case WINED3DDECLUSAGE_FOG
:
3356 shader_addline(&buffer
, "gl_FogFragCoord = OUT[%u].%c;\n", i
, reg_mask
[1]);
3363 shader_addline(&buffer
, "}\n");
3365 } else if(ps_major
>= 3 && vs_major
>= 3) {
3366 semantics_out
= vs
->semantics_out
;
3368 /* This one is tricky: a 3.0 pixel shader reads from a 3.0 vertex shader */
3369 shader_addline(&buffer
, "varying vec4 IN[%u];\n", GL_LIMITS(glsl_varyings
) / 4);
3370 shader_addline(&buffer
, "void order_ps_input(in vec4 OUT[%u]) {\n", MAX_REG_OUTPUT
);
3372 /* First, sort out position and point size. Those are not passed to the pixel shader */
3373 for(i
= 0; i
< MAX_REG_OUTPUT
; i
++) {
3374 if (!vs
->baseShader
.reg_maps
.packed_output
[i
]) continue;
3376 usage
= semantics_out
[i
].usage
;
3377 usage_idx
= semantics_out
[i
].usage_idx
;
3378 shader_glsl_get_write_mask(&semantics_out
[i
].reg
, reg_mask
);
3381 case WINED3DDECLUSAGE_POSITION
:
3382 shader_addline(&buffer
, "gl_Position%s = OUT[%u]%s;\n", reg_mask
, i
, reg_mask
);
3385 case WINED3DDECLUSAGE_PSIZE
:
3386 shader_addline(&buffer
, "gl_PointSize = OUT[%u].x;\n", i
);
3394 /* Then, fix the pixel shader input */
3395 handle_ps3_input(&buffer
, gl_info
, ps
->input_reg_map
,
3396 ps
->semantics_in
, &ps
->baseShader
.reg_maps
, semantics_out
, &vs
->baseShader
.reg_maps
);
3398 shader_addline(&buffer
, "}\n");
3399 } else if(ps_major
>= 3 && vs_major
< 3) {
3400 shader_addline(&buffer
, "varying vec4 IN[%u];\n", GL_LIMITS(glsl_varyings
) / 4);
3401 shader_addline(&buffer
, "void order_ps_input() {\n");
3402 /* The vertex shader wrote to the builtin varyings. There is no need to figure out position and
3403 * point size, but we depend on the optimizers kindness to find out that the pixel shader doesn't
3404 * read gl_TexCoord and gl_ColorX, otherwise we'll run out of varyings
3406 handle_ps3_input(&buffer
, gl_info
, ps
->input_reg_map
, ps
->semantics_in
, &ps
->baseShader
.reg_maps
, NULL
, NULL
);
3407 shader_addline(&buffer
, "}\n");
3409 ERR("Unexpected vertex and pixel shader version condition: vs: %d, ps: %d\n", vs_major
, ps_major
);
3412 ret
= GL_EXTCALL(glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB
));
3413 checkGLcall("glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB)");
3414 GL_EXTCALL(glShaderSourceARB(ret
, 1, (const char**)&buffer
.buffer
, NULL
));
3415 checkGLcall("glShaderSourceARB(ret, 1, &buffer.buffer, NULL)");
3416 GL_EXTCALL(glCompileShaderARB(ret
));
3417 checkGLcall("glCompileShaderARB(ret)");
3419 shader_buffer_free(&buffer
);
3423 static void hardcode_local_constants(IWineD3DBaseShaderImpl
*shader
, const WineD3D_GL_Info
*gl_info
,
3424 GLhandleARB programId
, char prefix
)
3426 const local_constant
*lconst
;
3431 LIST_FOR_EACH_ENTRY(lconst
, &shader
->baseShader
.constantsF
, local_constant
, entry
) {
3432 value
= (const float *)lconst
->value
;
3433 snprintf(glsl_name
, sizeof(glsl_name
), "%cLC%u", prefix
, lconst
->idx
);
3434 tmp_loc
= GL_EXTCALL(glGetUniformLocationARB(programId
, glsl_name
));
3435 GL_EXTCALL(glUniform4fvARB(tmp_loc
, 1, value
));
3437 checkGLcall("Hardcoding local constants\n");
3440 /** Sets the GLSL program ID for the given pixel and vertex shader combination.
3441 * It sets the programId on the current StateBlock (because it should be called
3442 * inside of the DrawPrimitive() part of the render loop).
3444 * If a program for the given combination does not exist, create one, and store
3445 * the program in the hash table. If it creates a program, it will link the
3446 * given objects, too.
3448 static void set_glsl_shader_program(IWineD3DDevice
*iface
, BOOL use_ps
, BOOL use_vs
) {
3449 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
3450 struct shader_glsl_priv
*priv
= This
->shader_priv
;
3451 const WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
3452 IWineD3DPixelShader
*pshader
= This
->stateBlock
->pixelShader
;
3453 IWineD3DVertexShader
*vshader
= This
->stateBlock
->vertexShader
;
3454 struct glsl_shader_prog_link
*entry
= NULL
;
3455 GLhandleARB programId
= 0;
3456 GLhandleARB reorder_shader_id
= 0;
3459 GLhandleARB vshader_id
, pshader_id
;
3460 struct ps_compile_args ps_compile_args
;
3461 struct vs_compile_args vs_compile_args
;
3464 find_vs_compile_args((IWineD3DVertexShaderImpl
*)This
->stateBlock
->vertexShader
, This
->stateBlock
, &vs_compile_args
);
3466 /* FIXME: Do we really have to spend CPU cycles to generate a few zeroed bytes? */
3467 memset(&vs_compile_args
, 0, sizeof(vs_compile_args
));
3470 find_ps_compile_args((IWineD3DPixelShaderImpl
*)This
->stateBlock
->pixelShader
, This
->stateBlock
, &ps_compile_args
);
3472 /* FIXME: Do we really have to spend CPU cycles to generate a few zeroed bytes? */
3473 memset(&ps_compile_args
, 0, sizeof(ps_compile_args
));
3475 entry
= get_glsl_program_entry(priv
, vshader
, pshader
, &vs_compile_args
, &ps_compile_args
);
3477 priv
->glsl_program
= entry
;
3481 /* If we get to this point, then no matching program exists, so we create one */
3482 programId
= GL_EXTCALL(glCreateProgramObjectARB());
3483 TRACE("Created new GLSL shader program %u\n", programId
);
3485 /* Create the entry */
3486 entry
= HeapAlloc(GetProcessHeap(), 0, sizeof(struct glsl_shader_prog_link
));
3487 entry
->programId
= programId
;
3488 entry
->vshader
= vshader
;
3489 entry
->pshader
= pshader
;
3490 entry
->vs_args
= vs_compile_args
;
3491 entry
->ps_args
= ps_compile_args
;
3492 entry
->constant_version
= 0;
3493 /* Add the hash table entry */
3494 add_glsl_program_entry(priv
, entry
);
3496 /* Set the current program */
3497 priv
->glsl_program
= entry
;
3500 vshader_id
= find_gl_vshader((IWineD3DVertexShaderImpl
*) vshader
, &vs_compile_args
);
3505 /* Attach GLSL vshader */
3507 const unsigned int max_attribs
= 16; /* TODO: Will this always be the case? It is at the moment... */
3510 reorder_shader_id
= generate_param_reorder_function(vshader
, pshader
, gl_info
);
3511 TRACE("Attaching GLSL shader object %u to program %u\n", reorder_shader_id
, programId
);
3512 GL_EXTCALL(glAttachObjectARB(programId
, reorder_shader_id
));
3513 checkGLcall("glAttachObjectARB");
3514 /* Flag the reorder function for deletion, then it will be freed automatically when the program
3517 GL_EXTCALL(glDeleteObjectARB(reorder_shader_id
));
3519 TRACE("Attaching GLSL shader object %u to program %u\n", vshader_id
, programId
);
3520 GL_EXTCALL(glAttachObjectARB(programId
, vshader_id
));
3521 checkGLcall("glAttachObjectARB");
3523 /* Bind vertex attributes to a corresponding index number to match
3524 * the same index numbers as ARB_vertex_programs (makes loading
3525 * vertex attributes simpler). With this method, we can use the
3526 * exact same code to load the attributes later for both ARB and
3529 * We have to do this here because we need to know the Program ID
3530 * in order to make the bindings work, and it has to be done prior
3531 * to linking the GLSL program. */
3532 for (i
= 0; i
< max_attribs
; ++i
) {
3533 if (((IWineD3DBaseShaderImpl
*)vshader
)->baseShader
.reg_maps
.attributes
[i
]) {
3534 snprintf(tmp_name
, sizeof(tmp_name
), "attrib%i", i
);
3535 GL_EXTCALL(glBindAttribLocationARB(programId
, i
, tmp_name
));
3538 checkGLcall("glBindAttribLocationARB");
3540 list_add_head(&((IWineD3DBaseShaderImpl
*)vshader
)->baseShader
.linked_programs
, &entry
->vshader_entry
);
3544 pshader_id
= find_gl_pshader((IWineD3DPixelShaderImpl
*) pshader
, &ps_compile_args
);
3549 /* Attach GLSL pshader */
3551 TRACE("Attaching GLSL shader object %u to program %u\n", pshader_id
, programId
);
3552 GL_EXTCALL(glAttachObjectARB(programId
, pshader_id
));
3553 checkGLcall("glAttachObjectARB");
3555 list_add_head(&((IWineD3DBaseShaderImpl
*)pshader
)->baseShader
.linked_programs
, &entry
->pshader_entry
);
3558 /* Link the program */
3559 TRACE("Linking GLSL shader program %u\n", programId
);
3560 GL_EXTCALL(glLinkProgramARB(programId
));
3561 print_glsl_info_log(&GLINFO_LOCATION
, programId
);
3563 entry
->vuniformF_locations
= HeapAlloc(GetProcessHeap(), 0, sizeof(GLhandleARB
) * GL_LIMITS(vshader_constantsF
));
3564 for (i
= 0; i
< GL_LIMITS(vshader_constantsF
); ++i
) {
3565 snprintf(glsl_name
, sizeof(glsl_name
), "VC[%i]", i
);
3566 entry
->vuniformF_locations
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, glsl_name
));
3568 for (i
= 0; i
< MAX_CONST_I
; ++i
) {
3569 snprintf(glsl_name
, sizeof(glsl_name
), "VI[%i]", i
);
3570 entry
->vuniformI_locations
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, glsl_name
));
3572 entry
->puniformF_locations
= HeapAlloc(GetProcessHeap(), 0, sizeof(GLhandleARB
) * GL_LIMITS(pshader_constantsF
));
3573 for (i
= 0; i
< GL_LIMITS(pshader_constantsF
); ++i
) {
3574 snprintf(glsl_name
, sizeof(glsl_name
), "PC[%i]", i
);
3575 entry
->puniformF_locations
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, glsl_name
));
3577 for (i
= 0; i
< MAX_CONST_I
; ++i
) {
3578 snprintf(glsl_name
, sizeof(glsl_name
), "PI[%i]", i
);
3579 entry
->puniformI_locations
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, glsl_name
));
3583 for(i
= 0; i
< ((IWineD3DPixelShaderImpl
*)pshader
)->numbumpenvmatconsts
; i
++) {
3585 sprintf(name
, "bumpenvmat%d", ((IWineD3DPixelShaderImpl
*)pshader
)->bumpenvmatconst
[i
].texunit
);
3586 entry
->bumpenvmat_location
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, name
));
3587 sprintf(name
, "luminancescale%d", ((IWineD3DPixelShaderImpl
*)pshader
)->luminanceconst
[i
].texunit
);
3588 entry
->luminancescale_location
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, name
));
3589 sprintf(name
, "luminanceoffset%d", ((IWineD3DPixelShaderImpl
*)pshader
)->luminanceconst
[i
].texunit
);
3590 entry
->luminanceoffset_location
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, name
));
3594 if (use_ps
&& ps_compile_args
.np2_fixup
) {
3596 for (i
= 0; i
< MAX_FRAGMENT_SAMPLERS
; ++i
) {
3597 if (ps_compile_args
.np2_fixup
& (1 << i
)) {
3598 sprintf(name
, "PsamplerNP2Fixup%u", i
);
3599 entry
->np2Fixup_location
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, name
));
3601 entry
->np2Fixup_location
[i
] = -1;
3606 entry
->posFixup_location
= GL_EXTCALL(glGetUniformLocationARB(programId
, "posFixup"));
3607 entry
->ycorrection_location
= GL_EXTCALL(glGetUniformLocationARB(programId
, "ycorrection"));
3608 checkGLcall("Find glsl program uniform locations");
3611 && WINED3DSHADER_VERSION_MAJOR(((IWineD3DPixelShaderImpl
*)pshader
)->baseShader
.reg_maps
.shader_version
) >= 3
3612 && ((IWineD3DPixelShaderImpl
*)pshader
)->declared_in_count
> GL_LIMITS(glsl_varyings
) / 4)
3614 TRACE("Shader %d needs vertex color clamping disabled\n", programId
);
3615 entry
->vertex_color_clamp
= GL_FALSE
;
3617 entry
->vertex_color_clamp
= GL_FIXED_ONLY_ARB
;
3620 /* Set the shader to allow uniform loading on it */
3621 GL_EXTCALL(glUseProgramObjectARB(programId
));
3622 checkGLcall("glUseProgramObjectARB(programId)");
3624 /* Load the vertex and pixel samplers now. The function that finds the mappings makes sure
3625 * that it stays the same for each vertexshader-pixelshader pair(=linked glsl program). If
3626 * a pshader with fixed function pipeline is used there are no vertex samplers, and if a
3627 * vertex shader with fixed function pixel processing is used we make sure that the card
3628 * supports enough samplers to allow the max number of vertex samplers with all possible
3629 * fixed function fragment processing setups. So once the program is linked these samplers
3633 /* Load vertex shader samplers */
3634 shader_glsl_load_vsamplers(gl_info
, This
->texUnitMap
, programId
);
3637 /* Load pixel shader samplers */
3638 shader_glsl_load_psamplers(gl_info
, This
->texUnitMap
, programId
);
3641 /* If the local constants do not have to be loaded with the environment constants,
3642 * load them now to have them hardcoded in the GLSL program. This saves some CPU cycles
3645 if(pshader
&& !((IWineD3DPixelShaderImpl
*)pshader
)->baseShader
.load_local_constsF
) {
3646 hardcode_local_constants((IWineD3DBaseShaderImpl
*) pshader
, gl_info
, programId
, 'P');
3648 if(vshader
&& !((IWineD3DVertexShaderImpl
*)vshader
)->baseShader
.load_local_constsF
) {
3649 hardcode_local_constants((IWineD3DBaseShaderImpl
*) vshader
, gl_info
, programId
, 'V');
3653 static GLhandleARB
create_glsl_blt_shader(const WineD3D_GL_Info
*gl_info
, enum tex_types tex_type
)
3655 GLhandleARB program_id
;
3656 GLhandleARB vshader_id
, pshader_id
;
3657 static const char *blt_vshader
[] =
3662 " gl_Position = gl_Vertex;\n"
3663 " gl_FrontColor = vec4(1.0);\n"
3664 " gl_TexCoord[0] = gl_MultiTexCoord0;\n"
3668 static const char *blt_pshaders
[tex_type_count
] =
3674 "uniform sampler2D sampler;\n"
3677 " gl_FragDepth = texture2D(sampler, gl_TexCoord[0].xy).x;\n"
3683 "uniform samplerCube sampler;\n"
3686 " gl_FragDepth = textureCube(sampler, gl_TexCoord[0].xyz).x;\n"
3690 "#extension GL_ARB_texture_rectangle : enable\n"
3691 "uniform sampler2DRect sampler;\n"
3694 " gl_FragDepth = texture2DRect(sampler, gl_TexCoord[0].xy).x;\n"
3698 if (!blt_pshaders
[tex_type
])
3700 FIXME("tex_type %#x not supported\n", tex_type
);
3704 vshader_id
= GL_EXTCALL(glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB
));
3705 GL_EXTCALL(glShaderSourceARB(vshader_id
, 1, blt_vshader
, NULL
));
3706 GL_EXTCALL(glCompileShaderARB(vshader_id
));
3708 pshader_id
= GL_EXTCALL(glCreateShaderObjectARB(GL_FRAGMENT_SHADER_ARB
));
3709 GL_EXTCALL(glShaderSourceARB(pshader_id
, 1, &blt_pshaders
[tex_type
], NULL
));
3710 GL_EXTCALL(glCompileShaderARB(pshader_id
));
3712 program_id
= GL_EXTCALL(glCreateProgramObjectARB());
3713 GL_EXTCALL(glAttachObjectARB(program_id
, vshader_id
));
3714 GL_EXTCALL(glAttachObjectARB(program_id
, pshader_id
));
3715 GL_EXTCALL(glLinkProgramARB(program_id
));
3717 print_glsl_info_log(&GLINFO_LOCATION
, program_id
);
3719 /* Once linked we can mark the shaders for deletion. They will be deleted once the program
3722 GL_EXTCALL(glDeleteObjectARB(vshader_id
));
3723 GL_EXTCALL(glDeleteObjectARB(pshader_id
));
3727 static void shader_glsl_select(IWineD3DDevice
*iface
, BOOL usePS
, BOOL useVS
) {
3728 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
3729 struct shader_glsl_priv
*priv
= This
->shader_priv
;
3730 const WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
3731 GLhandleARB program_id
= 0;
3732 GLenum old_vertex_color_clamp
, current_vertex_color_clamp
;
3734 old_vertex_color_clamp
= priv
->glsl_program
? priv
->glsl_program
->vertex_color_clamp
: GL_FIXED_ONLY_ARB
;
3736 if (useVS
|| usePS
) set_glsl_shader_program(iface
, usePS
, useVS
);
3737 else priv
->glsl_program
= NULL
;
3739 current_vertex_color_clamp
= priv
->glsl_program
? priv
->glsl_program
->vertex_color_clamp
: GL_FIXED_ONLY_ARB
;
3741 if (old_vertex_color_clamp
!= current_vertex_color_clamp
) {
3742 if (GL_SUPPORT(ARB_COLOR_BUFFER_FLOAT
)) {
3743 GL_EXTCALL(glClampColorARB(GL_CLAMP_VERTEX_COLOR_ARB
, current_vertex_color_clamp
));
3744 checkGLcall("glClampColorARB");
3746 FIXME("vertex color clamp needs to be changed, but extension not supported.\n");
3750 program_id
= priv
->glsl_program
? priv
->glsl_program
->programId
: 0;
3751 if (program_id
) TRACE("Using GLSL program %u\n", program_id
);
3752 GL_EXTCALL(glUseProgramObjectARB(program_id
));
3753 checkGLcall("glUseProgramObjectARB");
3756 static void shader_glsl_select_depth_blt(IWineD3DDevice
*iface
, enum tex_types tex_type
) {
3757 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
3758 const WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
3759 struct shader_glsl_priv
*priv
= This
->shader_priv
;
3760 GLhandleARB
*blt_program
= &priv
->depth_blt_program
[tex_type
];
3762 if (!*blt_program
) {
3764 *blt_program
= create_glsl_blt_shader(gl_info
, tex_type
);
3765 loc
= GL_EXTCALL(glGetUniformLocationARB(*blt_program
, "sampler"));
3766 GL_EXTCALL(glUseProgramObjectARB(*blt_program
));
3767 GL_EXTCALL(glUniform1iARB(loc
, 0));
3769 GL_EXTCALL(glUseProgramObjectARB(*blt_program
));
3773 static void shader_glsl_deselect_depth_blt(IWineD3DDevice
*iface
) {
3774 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
3775 const WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
3776 struct shader_glsl_priv
*priv
= This
->shader_priv
;
3777 GLhandleARB program_id
;
3779 program_id
= priv
->glsl_program
? priv
->glsl_program
->programId
: 0;
3780 if (program_id
) TRACE("Using GLSL program %u\n", program_id
);
3782 GL_EXTCALL(glUseProgramObjectARB(program_id
));
3783 checkGLcall("glUseProgramObjectARB");
3786 static void shader_glsl_destroy(IWineD3DBaseShader
*iface
) {
3787 const struct list
*linked_programs
;
3788 IWineD3DBaseShaderImpl
*This
= (IWineD3DBaseShaderImpl
*) iface
;
3789 IWineD3DDeviceImpl
*device
= (IWineD3DDeviceImpl
*)This
->baseShader
.device
;
3790 struct shader_glsl_priv
*priv
= device
->shader_priv
;
3791 const WineD3D_GL_Info
*gl_info
= &device
->adapter
->gl_info
;
3792 IWineD3DPixelShaderImpl
*ps
= NULL
;
3793 IWineD3DVertexShaderImpl
*vs
= NULL
;
3795 /* Note: Do not use QueryInterface here to find out which shader type this is because this code
3796 * can be called from IWineD3DBaseShader::Release
3798 char pshader
= shader_is_pshader_version(This
->baseShader
.reg_maps
.shader_version
);
3801 ps
= (IWineD3DPixelShaderImpl
*) This
;
3802 if(ps
->num_gl_shaders
== 0) return;
3803 if (priv
->glsl_program
&& (IWineD3DBaseShader
*)priv
->glsl_program
->pshader
== iface
)
3804 shader_glsl_select(This
->baseShader
.device
, FALSE
, FALSE
);
3806 vs
= (IWineD3DVertexShaderImpl
*) This
;
3807 if(vs
->num_gl_shaders
== 0) return;
3808 if (priv
->glsl_program
&& (IWineD3DBaseShader
*)priv
->glsl_program
->vshader
== iface
)
3809 shader_glsl_select(This
->baseShader
.device
, FALSE
, FALSE
);
3812 linked_programs
= &This
->baseShader
.linked_programs
;
3814 TRACE("Deleting linked programs\n");
3815 if (linked_programs
->next
) {
3816 struct glsl_shader_prog_link
*entry
, *entry2
;
3819 LIST_FOR_EACH_ENTRY_SAFE(entry
, entry2
, linked_programs
, struct glsl_shader_prog_link
, pshader_entry
) {
3820 delete_glsl_program_entry(priv
, gl_info
, entry
);
3823 LIST_FOR_EACH_ENTRY_SAFE(entry
, entry2
, linked_programs
, struct glsl_shader_prog_link
, vshader_entry
) {
3824 delete_glsl_program_entry(priv
, gl_info
, entry
);
3833 for(i
= 0; i
< ps
->num_gl_shaders
; i
++) {
3834 TRACE("deleting pshader %u\n", ps
->gl_shaders
[i
].prgId
);
3835 GL_EXTCALL(glDeleteObjectARB(ps
->gl_shaders
[i
].prgId
));
3836 checkGLcall("glDeleteObjectARB");
3839 HeapFree(GetProcessHeap(), 0, ps
->gl_shaders
);
3840 ps
->gl_shaders
= NULL
;
3841 ps
->num_gl_shaders
= 0;
3842 ps
->shader_array_size
= 0;
3847 for(i
= 0; i
< vs
->num_gl_shaders
; i
++) {
3848 TRACE("deleting vshader %u\n", vs
->gl_shaders
[i
].prgId
);
3849 GL_EXTCALL(glDeleteObjectARB(vs
->gl_shaders
[i
].prgId
));
3850 checkGLcall("glDeleteObjectARB");
3853 HeapFree(GetProcessHeap(), 0, vs
->gl_shaders
);
3854 vs
->gl_shaders
= NULL
;
3855 vs
->num_gl_shaders
= 0;
3856 vs
->shader_array_size
= 0;
3860 static unsigned int glsl_program_key_hash(const void *key
)
3862 const glsl_program_key_t
*k
= key
;
3864 unsigned int hash
= ((DWORD_PTR
) k
->vshader
) | ((DWORD_PTR
) k
->pshader
) << 16;
3865 hash
+= ~(hash
<< 15);
3866 hash
^= (hash
>> 10);
3867 hash
+= (hash
<< 3);
3868 hash
^= (hash
>> 6);
3869 hash
+= ~(hash
<< 11);
3870 hash
^= (hash
>> 16);
3875 static BOOL
glsl_program_key_compare(const void *keya
, const void *keyb
)
3877 const glsl_program_key_t
*ka
= keya
;
3878 const glsl_program_key_t
*kb
= keyb
;
3880 return ka
->vshader
== kb
->vshader
&& ka
->pshader
== kb
->pshader
&&
3881 (memcmp(&ka
->ps_args
, &kb
->ps_args
, sizeof(kb
->ps_args
)) == 0) &&
3882 (memcmp(&ka
->vs_args
, &kb
->vs_args
, sizeof(kb
->vs_args
)) == 0);
3885 static BOOL
constant_heap_init(struct constant_heap
*heap
, unsigned int constant_count
)
3887 SIZE_T size
= (constant_count
+ 1) * sizeof(*heap
->entries
) + constant_count
* sizeof(*heap
->positions
);
3888 void *mem
= HeapAlloc(GetProcessHeap(), 0, size
);
3892 ERR("Failed to allocate memory\n");
3896 heap
->entries
= mem
;
3897 heap
->entries
[1].version
= 0;
3898 heap
->positions
= (unsigned int *)(heap
->entries
+ constant_count
+ 1);
3904 static void constant_heap_free(struct constant_heap
*heap
)
3906 HeapFree(GetProcessHeap(), 0, heap
->entries
);
3909 static HRESULT
shader_glsl_alloc(IWineD3DDevice
*iface
) {
3910 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
3911 const WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
3912 struct shader_glsl_priv
*priv
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(struct shader_glsl_priv
));
3913 SIZE_T stack_size
= wined3d_log2i(max(GL_LIMITS(vshader_constantsF
), GL_LIMITS(pshader_constantsF
))) + 1;
3915 priv
->stack
= HeapAlloc(GetProcessHeap(), 0, stack_size
* sizeof(*priv
->stack
));
3918 ERR("Failed to allocate memory.\n");
3919 HeapFree(GetProcessHeap(), 0, priv
);
3920 return E_OUTOFMEMORY
;
3923 if (!constant_heap_init(&priv
->vconst_heap
, GL_LIMITS(vshader_constantsF
)))
3925 ERR("Failed to initialize vertex shader constant heap\n");
3926 HeapFree(GetProcessHeap(), 0, priv
->stack
);
3927 HeapFree(GetProcessHeap(), 0, priv
);
3928 return E_OUTOFMEMORY
;
3931 if (!constant_heap_init(&priv
->pconst_heap
, GL_LIMITS(pshader_constantsF
)))
3933 ERR("Failed to initialize pixel shader constant heap\n");
3934 constant_heap_free(&priv
->vconst_heap
);
3935 HeapFree(GetProcessHeap(), 0, priv
->stack
);
3936 HeapFree(GetProcessHeap(), 0, priv
);
3937 return E_OUTOFMEMORY
;
3940 priv
->glsl_program_lookup
= hash_table_create(glsl_program_key_hash
, glsl_program_key_compare
);
3941 priv
->next_constant_version
= 1;
3943 This
->shader_priv
= priv
;
3947 static void shader_glsl_free(IWineD3DDevice
*iface
) {
3948 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
3949 const WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
3950 struct shader_glsl_priv
*priv
= This
->shader_priv
;
3953 for (i
= 0; i
< tex_type_count
; ++i
)
3955 if (priv
->depth_blt_program
[i
])
3957 GL_EXTCALL(glDeleteObjectARB(priv
->depth_blt_program
[i
]));
3961 hash_table_destroy(priv
->glsl_program_lookup
, NULL
, NULL
);
3962 constant_heap_free(&priv
->pconst_heap
);
3963 constant_heap_free(&priv
->vconst_heap
);
3965 HeapFree(GetProcessHeap(), 0, This
->shader_priv
);
3966 This
->shader_priv
= NULL
;
3969 static BOOL
shader_glsl_dirty_const(IWineD3DDevice
*iface
) {
3970 /* TODO: GL_EXT_bindable_uniform can be used to share constants across shaders */
3974 static GLuint
shader_glsl_generate_pshader(IWineD3DPixelShader
*iface
, SHADER_BUFFER
*buffer
, const struct ps_compile_args
*args
) {
3975 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*)iface
;
3976 const struct shader_reg_maps
*reg_maps
= &This
->baseShader
.reg_maps
;
3977 CONST DWORD
*function
= This
->baseShader
.function
;
3978 const char *fragcolor
;
3979 const WineD3D_GL_Info
*gl_info
= &((IWineD3DDeviceImpl
*)This
->baseShader
.device
)->adapter
->gl_info
;
3981 /* Create the hw GLSL shader object and assign it as the shader->prgId */
3982 GLhandleARB shader_obj
= GL_EXTCALL(glCreateShaderObjectARB(GL_FRAGMENT_SHADER_ARB
));
3984 shader_addline(buffer
, "#version 120\n");
3986 if (GL_SUPPORT(ARB_DRAW_BUFFERS
)) {
3987 shader_addline(buffer
, "#extension GL_ARB_draw_buffers : enable\n");
3989 if (GL_SUPPORT(ARB_TEXTURE_RECTANGLE
)) {
3990 /* The spec says that it doesn't have to be explicitly enabled, but the nvidia
3991 * drivers write a warning if we don't do so
3993 shader_addline(buffer
, "#extension GL_ARB_texture_rectangle : enable\n");
3996 /* Base Declarations */
3997 shader_generate_glsl_declarations( (IWineD3DBaseShader
*) This
, reg_maps
, buffer
, &GLINFO_LOCATION
, args
);
3999 /* Pack 3.0 inputs */
4000 if (reg_maps
->shader_version
>= WINED3DPS_VERSION(3,0) && args
->vp_mode
!= vertexshader
) {
4001 pshader_glsl_input_pack(iface
, buffer
, This
->semantics_in
, reg_maps
, args
->vp_mode
);
4004 /* Base Shader Body */
4005 shader_generate_main( (IWineD3DBaseShader
*) This
, buffer
, reg_maps
, function
);
4007 /* Pixel shaders < 2.0 place the resulting color in R0 implicitly */
4008 if (reg_maps
->shader_version
< WINED3DPS_VERSION(2,0))
4010 /* Some older cards like GeforceFX ones don't support multiple buffers, so also not gl_FragData */
4011 if(GL_SUPPORT(ARB_DRAW_BUFFERS
))
4012 shader_addline(buffer
, "gl_FragData[0] = R0;\n");
4014 shader_addline(buffer
, "gl_FragColor = R0;\n");
4017 if(GL_SUPPORT(ARB_DRAW_BUFFERS
)) {
4018 fragcolor
= "gl_FragData[0]";
4020 fragcolor
= "gl_FragColor";
4022 if(args
->srgb_correction
) {
4023 shader_addline(buffer
, "tmp0.xyz = pow(%s.xyz, vec3(%f, %f, %f)) * vec3(%f, %f, %f) - vec3(%f, %f, %f);\n",
4024 fragcolor
, srgb_pow
, srgb_pow
, srgb_pow
, srgb_mul_high
, srgb_mul_high
, srgb_mul_high
,
4025 srgb_sub_high
, srgb_sub_high
, srgb_sub_high
);
4026 shader_addline(buffer
, "tmp1.xyz = %s.xyz * srgb_mul_low.xyz;\n", fragcolor
);
4027 shader_addline(buffer
, "%s.x = %s.x < srgb_comparison.x ? tmp1.x : tmp0.x;\n", fragcolor
, fragcolor
);
4028 shader_addline(buffer
, "%s.y = %s.y < srgb_comparison.y ? tmp1.y : tmp0.y;\n", fragcolor
, fragcolor
);
4029 shader_addline(buffer
, "%s.z = %s.z < srgb_comparison.z ? tmp1.z : tmp0.z;\n", fragcolor
, fragcolor
);
4030 shader_addline(buffer
, "%s = clamp(%s, 0.0, 1.0);\n", fragcolor
, fragcolor
);
4032 /* Pixel shader < 3.0 do not replace the fog stage.
4033 * This implements linear fog computation and blending.
4034 * TODO: non linear fog
4035 * NOTE: gl_Fog.start and gl_Fog.end don't hold fog start s and end e but
4036 * -1/(e-s) and e/(e-s) respectively.
4038 if(reg_maps
->shader_version
< WINED3DPS_VERSION(3,0)) {
4040 case FOG_OFF
: break;
4042 shader_addline(buffer
, "float fogstart = -1.0 / (gl_Fog.end - gl_Fog.start);\n");
4043 shader_addline(buffer
, "float fogend = gl_Fog.end * -fogstart;\n");
4044 shader_addline(buffer
, "float Fog = clamp(gl_FogFragCoord * fogstart + fogend, 0.0, 1.0);\n");
4045 shader_addline(buffer
, "%s.xyz = mix(gl_Fog.color.xyz, %s.xyz, Fog);\n", fragcolor
, fragcolor
);
4048 /* Fog = e^(-gl_Fog.density * gl_FogFragCoord) */
4049 shader_addline(buffer
, "float Fog = exp(-gl_Fog.density * gl_FogFragCoord);\n");
4050 shader_addline(buffer
, "Fog = clamp(Fog, 0.0, 1.0);\n");
4051 shader_addline(buffer
, "%s.xyz = mix(gl_Fog.color.xyz, %s.xyz, Fog);\n", fragcolor
, fragcolor
);
4054 /* Fog = e^(-(gl_Fog.density * gl_FogFragCoord)^2) */
4055 shader_addline(buffer
, "float Fog = exp(-gl_Fog.density * gl_Fog.density * gl_FogFragCoord * gl_FogFragCoord);\n");
4056 shader_addline(buffer
, "Fog = clamp(Fog, 0.0, 1.0);\n");
4057 shader_addline(buffer
, "%s.xyz = mix(gl_Fog.color.xyz, %s.xyz, Fog);\n", fragcolor
, fragcolor
);
4062 shader_addline(buffer
, "}\n");
4064 TRACE("Compiling shader object %u\n", shader_obj
);
4065 GL_EXTCALL(glShaderSourceARB(shader_obj
, 1, (const char**)&buffer
->buffer
, NULL
));
4066 GL_EXTCALL(glCompileShaderARB(shader_obj
));
4067 print_glsl_info_log(&GLINFO_LOCATION
, shader_obj
);
4069 /* Store the shader object */
4073 static GLuint
shader_glsl_generate_vshader(IWineD3DVertexShader
*iface
, SHADER_BUFFER
*buffer
, const struct vs_compile_args
*args
) {
4074 IWineD3DVertexShaderImpl
*This
= (IWineD3DVertexShaderImpl
*)iface
;
4075 const struct shader_reg_maps
*reg_maps
= &This
->baseShader
.reg_maps
;
4076 CONST DWORD
*function
= This
->baseShader
.function
;
4077 const WineD3D_GL_Info
*gl_info
= &((IWineD3DDeviceImpl
*)This
->baseShader
.device
)->adapter
->gl_info
;
4079 /* Create the hw GLSL shader program and assign it as the shader->prgId */
4080 GLhandleARB shader_obj
= GL_EXTCALL(glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB
));
4082 shader_addline(buffer
, "#version 120\n");
4084 /* Base Declarations */
4085 shader_generate_glsl_declarations( (IWineD3DBaseShader
*) This
, reg_maps
, buffer
, &GLINFO_LOCATION
, NULL
);
4087 /* Base Shader Body */
4088 shader_generate_main( (IWineD3DBaseShader
*) This
, buffer
, reg_maps
, function
);
4090 /* Unpack 3.0 outputs */
4091 if (reg_maps
->shader_version
>= WINED3DVS_VERSION(3,0)) shader_addline(buffer
, "order_ps_input(OUT);\n");
4092 else shader_addline(buffer
, "order_ps_input();\n");
4094 /* The D3DRS_FOGTABLEMODE render state defines if the shader-generated fog coord is used
4095 * or if the fragment depth is used. If the fragment depth is used(FOGTABLEMODE != NONE),
4096 * the fog frag coord is thrown away. If the fog frag coord is used, but not written by
4097 * the shader, it is set to 0.0(fully fogged, since start = 1.0, end = 0.0)
4099 if(args
->fog_src
== VS_FOG_Z
) {
4100 shader_addline(buffer
, "gl_FogFragCoord = gl_Position.z;\n");
4101 } else if (!reg_maps
->fog
) {
4102 shader_addline(buffer
, "gl_FogFragCoord = 0.0;\n");
4105 /* Write the final position.
4107 * OpenGL coordinates specify the center of the pixel while d3d coords specify
4108 * the corner. The offsets are stored in z and w in posFixup. posFixup.y contains
4109 * 1.0 or -1.0 to turn the rendering upside down for offscreen rendering. PosFixup.x
4110 * contains 1.0 to allow a mad.
4112 shader_addline(buffer
, "gl_Position.y = gl_Position.y * posFixup.y;\n");
4113 shader_addline(buffer
, "gl_Position.xy += posFixup.zw * gl_Position.ww;\n");
4115 /* Z coord [0;1]->[-1;1] mapping, see comment in transform_projection in state.c
4117 * Basically we want (in homogeneous coordinates) z = z * 2 - 1. However, shaders are run
4118 * before the homogeneous divide, so we have to take the w into account: z = ((z / w) * 2 - 1) * w,
4119 * which is the same as z = z * 2 - w.
4121 shader_addline(buffer
, "gl_Position.z = gl_Position.z * 2.0 - gl_Position.w;\n");
4123 shader_addline(buffer
, "}\n");
4125 TRACE("Compiling shader object %u\n", shader_obj
);
4126 GL_EXTCALL(glShaderSourceARB(shader_obj
, 1, (const char**)&buffer
->buffer
, NULL
));
4127 GL_EXTCALL(glCompileShaderARB(shader_obj
));
4128 print_glsl_info_log(&GLINFO_LOCATION
, shader_obj
);
4133 static void shader_glsl_get_caps(WINED3DDEVTYPE devtype
, const WineD3D_GL_Info
*gl_info
, struct shader_caps
*pCaps
)
4135 /* Nvidia Geforce6/7 or Ati R4xx/R5xx cards with GLSL support, support VS 3.0 but older Nvidia/Ati
4136 * models with GLSL support only support 2.0. In case of nvidia we can detect VS 2.0 support using
4137 * vs_nv_version which is based on NV_vertex_program.
4138 * For Ati cards there's no way using glsl (it abstracts the lowlevel info away) and also not
4139 * using ARB_vertex_program. It is safe to assume that when a card supports pixel shader 2.0 it
4140 * supports vertex shader 2.0 too and the way around. We can detect ps2.0 using the maximum number
4141 * of native instructions, so use that here. For more info see the pixel shader versioning code below.
4143 if((GLINFO_LOCATION
.vs_nv_version
== VS_VERSION_20
) || (GLINFO_LOCATION
.ps_arb_max_instructions
<= 512))
4144 pCaps
->VertexShaderVersion
= WINED3DVS_VERSION(2,0);
4146 pCaps
->VertexShaderVersion
= WINED3DVS_VERSION(3,0);
4147 TRACE_(d3d_caps
)("Hardware vertex shader version %d.%d enabled (GLSL)\n", (pCaps
->VertexShaderVersion
>> 8) & 0xff, pCaps
->VertexShaderVersion
& 0xff);
4148 /* Subtract the other potential uniforms from the max available (bools, ints, and 1 row of projection matrix) */
4149 pCaps
->MaxVertexShaderConst
= GL_LIMITS(vshader_constantsF
) - (MAX_CONST_B
/ 4) - MAX_CONST_I
- 1;
4151 /* Older DX9-class videocards (GeforceFX / Radeon >9500/X*00) only support pixel shader 2.0/2.0a/2.0b.
4152 * In OpenGL the extensions related to GLSL abstract lowlevel GL info away which is needed
4153 * to distinguish between 2.0 and 3.0 (and 2.0a/2.0b). In case of Nvidia we use their fragment
4154 * program extensions. On other hardware including ATI GL_ARB_fragment_program offers the info
4155 * in max native instructions. Intel and others also offer the info in this extension but they
4156 * don't support GLSL (at least on Windows).
4158 * PS2.0 requires at least 96 instructions, 2.0a/2.0b go up to 512. Assume that if the number
4159 * of instructions is 512 or less we have to do with ps2.0 hardware.
4160 * 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.
4162 if((GLINFO_LOCATION
.ps_nv_version
== PS_VERSION_20
) || (GLINFO_LOCATION
.ps_arb_max_instructions
<= 512))
4163 pCaps
->PixelShaderVersion
= WINED3DPS_VERSION(2,0);
4165 pCaps
->PixelShaderVersion
= WINED3DPS_VERSION(3,0);
4167 /* Subtract the other potential uniforms from the max available (bools & ints), and 2 states for fog.
4168 * In theory the texbem instruction may need one more shader constant too. But lets assume
4169 * that a sm <= 1.3 shader does not need all the uniforms provided by a glsl-capable card,
4170 * and lets not take away a uniform needlessly from all other shaders.
4172 pCaps
->MaxPixelShaderConst
= GL_LIMITS(pshader_constantsF
) - (MAX_CONST_B
/ 4) - MAX_CONST_I
- 2;
4174 /* FIXME: The following line is card dependent. -8.0 to 8.0 is the
4175 * Direct3D minimum requirement.
4177 * Both GL_ARB_fragment_program and GLSL require a "maximum representable magnitude"
4178 * of colors to be 2^10, and 2^32 for other floats. Should we use 1024 here?
4180 * The problem is that the refrast clamps temporary results in the shader to
4181 * [-MaxValue;+MaxValue]. If the card's max value is bigger than the one we advertize here,
4182 * then applications may miss the clamping behavior. On the other hand, if it is smaller,
4183 * the shader will generate incorrect results too. Unfortunately, GL deliberately doesn't
4184 * offer a way to query this.
4186 pCaps
->PixelShader1xMaxValue
= 8.0;
4187 TRACE_(d3d_caps
)("Hardware pixel shader version %d.%d enabled (GLSL)\n", (pCaps
->PixelShaderVersion
>> 8) & 0xff, pCaps
->PixelShaderVersion
& 0xff);
4190 static BOOL
shader_glsl_color_fixup_supported(struct color_fixup_desc fixup
)
4192 if (TRACE_ON(d3d_shader
) && TRACE_ON(d3d
))
4194 TRACE("Checking support for fixup:\n");
4195 dump_color_fixup_desc(fixup
);
4198 /* We support everything except YUV conversions. */
4199 if (!is_yuv_fixup(fixup
))
4205 TRACE("[FAILED]\n");
4209 static const SHADER_HANDLER shader_glsl_instruction_handler_table
[WINED3DSIH_TABLE_SIZE
] =
4211 /* WINED3DSIH_ABS */ shader_glsl_map2gl
,
4212 /* WINED3DSIH_ADD */ shader_glsl_arith
,
4213 /* WINED3DSIH_BEM */ pshader_glsl_bem
,
4214 /* WINED3DSIH_BREAK */ shader_glsl_break
,
4215 /* WINED3DSIH_BREAKC */ shader_glsl_breakc
,
4216 /* WINED3DSIH_BREAKP */ NULL
,
4217 /* WINED3DSIH_CALL */ shader_glsl_call
,
4218 /* WINED3DSIH_CALLNZ */ shader_glsl_callnz
,
4219 /* WINED3DSIH_CMP */ shader_glsl_cmp
,
4220 /* WINED3DSIH_CND */ shader_glsl_cnd
,
4221 /* WINED3DSIH_CRS */ shader_glsl_cross
,
4222 /* WINED3DSIH_DCL */ NULL
,
4223 /* WINED3DSIH_DEF */ NULL
,
4224 /* WINED3DSIH_DEFB */ NULL
,
4225 /* WINED3DSIH_DEFI */ NULL
,
4226 /* WINED3DSIH_DP2ADD */ pshader_glsl_dp2add
,
4227 /* WINED3DSIH_DP3 */ shader_glsl_dot
,
4228 /* WINED3DSIH_DP4 */ shader_glsl_dot
,
4229 /* WINED3DSIH_DST */ shader_glsl_dst
,
4230 /* WINED3DSIH_DSX */ shader_glsl_map2gl
,
4231 /* WINED3DSIH_DSY */ shader_glsl_map2gl
,
4232 /* WINED3DSIH_ELSE */ shader_glsl_else
,
4233 /* WINED3DSIH_ENDIF */ shader_glsl_end
,
4234 /* WINED3DSIH_ENDLOOP */ shader_glsl_end
,
4235 /* WINED3DSIH_ENDREP */ shader_glsl_end
,
4236 /* WINED3DSIH_EXP */ shader_glsl_map2gl
,
4237 /* WINED3DSIH_EXPP */ shader_glsl_expp
,
4238 /* WINED3DSIH_FRC */ shader_glsl_map2gl
,
4239 /* WINED3DSIH_IF */ shader_glsl_if
,
4240 /* WINED3DSIH_IFC */ shader_glsl_ifc
,
4241 /* WINED3DSIH_LABEL */ shader_glsl_label
,
4242 /* WINED3DSIH_LIT */ shader_glsl_lit
,
4243 /* WINED3DSIH_LOG */ shader_glsl_log
,
4244 /* WINED3DSIH_LOGP */ shader_glsl_log
,
4245 /* WINED3DSIH_LOOP */ shader_glsl_loop
,
4246 /* WINED3DSIH_LRP */ shader_glsl_lrp
,
4247 /* WINED3DSIH_M3x2 */ shader_glsl_mnxn
,
4248 /* WINED3DSIH_M3x3 */ shader_glsl_mnxn
,
4249 /* WINED3DSIH_M3x4 */ shader_glsl_mnxn
,
4250 /* WINED3DSIH_M4x3 */ shader_glsl_mnxn
,
4251 /* WINED3DSIH_M4x4 */ shader_glsl_mnxn
,
4252 /* WINED3DSIH_MAD */ shader_glsl_mad
,
4253 /* WINED3DSIH_MAX */ shader_glsl_map2gl
,
4254 /* WINED3DSIH_MIN */ shader_glsl_map2gl
,
4255 /* WINED3DSIH_MOV */ shader_glsl_mov
,
4256 /* WINED3DSIH_MOVA */ shader_glsl_mov
,
4257 /* WINED3DSIH_MUL */ shader_glsl_arith
,
4258 /* WINED3DSIH_NOP */ NULL
,
4259 /* WINED3DSIH_NRM */ shader_glsl_map2gl
,
4260 /* WINED3DSIH_PHASE */ NULL
,
4261 /* WINED3DSIH_POW */ shader_glsl_pow
,
4262 /* WINED3DSIH_RCP */ shader_glsl_rcp
,
4263 /* WINED3DSIH_REP */ shader_glsl_rep
,
4264 /* WINED3DSIH_RET */ NULL
,
4265 /* WINED3DSIH_RSQ */ shader_glsl_rsq
,
4266 /* WINED3DSIH_SETP */ NULL
,
4267 /* WINED3DSIH_SGE */ shader_glsl_compare
,
4268 /* WINED3DSIH_SGN */ shader_glsl_map2gl
,
4269 /* WINED3DSIH_SINCOS */ shader_glsl_sincos
,
4270 /* WINED3DSIH_SLT */ shader_glsl_compare
,
4271 /* WINED3DSIH_SUB */ shader_glsl_arith
,
4272 /* WINED3DSIH_TEX */ pshader_glsl_tex
,
4273 /* WINED3DSIH_TEXBEM */ pshader_glsl_texbem
,
4274 /* WINED3DSIH_TEXBEML */ pshader_glsl_texbem
,
4275 /* WINED3DSIH_TEXCOORD */ pshader_glsl_texcoord
,
4276 /* WINED3DSIH_TEXDEPTH */ pshader_glsl_texdepth
,
4277 /* WINED3DSIH_TEXDP3 */ pshader_glsl_texdp3
,
4278 /* WINED3DSIH_TEXDP3TEX */ pshader_glsl_texdp3tex
,
4279 /* WINED3DSIH_TEXKILL */ pshader_glsl_texkill
,
4280 /* WINED3DSIH_TEXLDD */ NULL
,
4281 /* WINED3DSIH_TEXLDL */ shader_glsl_texldl
,
4282 /* WINED3DSIH_TEXM3x2DEPTH */ pshader_glsl_texm3x2depth
,
4283 /* WINED3DSIH_TEXM3x2PAD */ pshader_glsl_texm3x2pad
,
4284 /* WINED3DSIH_TEXM3x2TEX */ pshader_glsl_texm3x2tex
,
4285 /* WINED3DSIH_TEXM3x3 */ pshader_glsl_texm3x3
,
4286 /* WINED3DSIH_TEXM3x3DIFF */ NULL
,
4287 /* WINED3DSIH_TEXM3x3PAD */ pshader_glsl_texm3x3pad
,
4288 /* WINED3DSIH_TEXM3x3SPEC */ pshader_glsl_texm3x3spec
,
4289 /* WINED3DSIH_TEXM3x3TEX */ pshader_glsl_texm3x3tex
,
4290 /* WINED3DSIH_TEXM3x3VSPEC */ pshader_glsl_texm3x3vspec
,
4291 /* WINED3DSIH_TEXREG2AR */ pshader_glsl_texreg2ar
,
4292 /* WINED3DSIH_TEXREG2GB */ pshader_glsl_texreg2gb
,
4293 /* WINED3DSIH_TEXREG2RGB */ pshader_glsl_texreg2rgb
,
4296 const shader_backend_t glsl_shader_backend
= {
4297 shader_glsl_instruction_handler_table
,
4299 shader_glsl_select_depth_blt
,
4300 shader_glsl_deselect_depth_blt
,
4301 shader_glsl_update_float_vertex_constants
,
4302 shader_glsl_update_float_pixel_constants
,
4303 shader_glsl_load_constants
,
4304 shader_glsl_load_np2fixup_constants
,
4305 shader_glsl_destroy
,
4308 shader_glsl_dirty_const
,
4309 shader_glsl_generate_pshader
,
4310 shader_glsl_generate_vshader
,
4311 shader_glsl_get_caps
,
4312 shader_glsl_color_fixup_supported
,