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(ps_args
->np2_fixup
& (1 << i
)) {
829 /* NP2/RECT textures in OpenGL use texcoords in the range [0,width]x[0,height]
830 * while D3D has them in the (normalized) [0,1]x[0,1] range.
831 * samplerNP2Fixup stores texture dimensions and is updated through
832 * shader_glsl_load_np2fixup_constants when the sampler changes. */
833 shader_addline(buffer
, "uniform vec2 %csamplerNP2Fixup%u;\n", prefix
, i
);
836 case WINED3DSTT_CUBE
:
837 shader_addline(buffer
, "uniform samplerCube %csampler%u;\n", prefix
, i
);
839 case WINED3DSTT_VOLUME
:
840 shader_addline(buffer
, "uniform sampler3D %csampler%u;\n", prefix
, i
);
843 shader_addline(buffer
, "uniform unsupported_sampler %csampler%u;\n", prefix
, i
);
844 FIXME("Unrecognized sampler type: %#x\n", stype
);
850 /* Declare address variables */
851 for (i
= 0; i
< This
->baseShader
.limits
.address
; i
++) {
852 if (reg_maps
->address
[i
])
853 shader_addline(buffer
, "ivec4 A%d;\n", i
);
856 /* Declare texture coordinate temporaries and initialize them */
857 for (i
= 0; i
< This
->baseShader
.limits
.texcoord
; i
++) {
858 if (reg_maps
->texcoord
[i
])
859 shader_addline(buffer
, "vec4 T%u = gl_TexCoord[%u];\n", i
, i
);
862 /* Declare input register varyings. Only pixel shader, vertex shaders have that declared in the
863 * helper function shader that is linked in at link time
865 if (pshader
&& shader_version
>= WINED3DPS_VERSION(3, 0))
867 if (use_vs(device
->stateBlock
))
869 shader_addline(buffer
, "varying vec4 IN[%u];\n", GL_LIMITS(glsl_varyings
) / 4);
871 /* TODO: Write a replacement shader for the fixed function vertex pipeline, so this isn't needed.
872 * For fixed function vertex processing + 3.0 pixel shader we need a separate function in the
873 * pixel shader that reads the fixed function color into the packed input registers.
875 shader_addline(buffer
, "vec4 IN[%u];\n", GL_LIMITS(glsl_varyings
) / 4);
879 /* Declare output register temporaries */
880 if(This
->baseShader
.limits
.packed_output
) {
881 shader_addline(buffer
, "vec4 OUT[%u];\n", This
->baseShader
.limits
.packed_output
);
884 /* Declare temporary variables */
885 for(i
= 0; i
< This
->baseShader
.limits
.temporary
; i
++) {
886 if (reg_maps
->temporary
[i
])
887 shader_addline(buffer
, "vec4 R%u;\n", i
);
890 /* Declare attributes */
891 for (i
= 0; i
< This
->baseShader
.limits
.attributes
; i
++) {
892 if (reg_maps
->attributes
[i
])
893 shader_addline(buffer
, "attribute vec4 attrib%i;\n", i
);
896 /* Declare loop registers aLx */
897 for (i
= 0; i
< reg_maps
->loop_depth
; i
++) {
898 shader_addline(buffer
, "int aL%u;\n", i
);
899 shader_addline(buffer
, "int tmpInt%u;\n", i
);
902 /* Temporary variables for matrix operations */
903 shader_addline(buffer
, "vec4 tmp0;\n");
904 shader_addline(buffer
, "vec4 tmp1;\n");
906 /* Local constants use a different name so they can be loaded once at shader link time
907 * They can't be hardcoded into the shader text via LC = {x, y, z, w}; because the
908 * float -> string conversion can cause precision loss.
910 if(!This
->baseShader
.load_local_constsF
) {
911 LIST_FOR_EACH_ENTRY(lconst
, &This
->baseShader
.constantsF
, local_constant
, entry
) {
912 shader_addline(buffer
, "uniform vec4 %cLC%u;\n", prefix
, lconst
->idx
);
916 /* Start the main program */
917 shader_addline(buffer
, "void main() {\n");
918 if(pshader
&& reg_maps
->vpos
) {
919 /* DirectX apps expect integer values, while OpenGL drivers add approximately 0.5. This causes
920 * off-by-one problems as spotted by the vPos d3d9 visual test. Unfortunately the ATI cards do
921 * not add exactly 0.5, but rather something like 0.49999999 or 0.50000001, which still causes
922 * precision troubles when we just substract 0.5.
924 * To deal with that just floor() the position. This will eliminate the fraction on all cards.
926 * TODO: Test how that behaves with multisampling once we can enable multisampling in winex11.
928 * An advantage of floor is that it works even if the driver doesn't add 1/2. It is somewhat
929 * questionable if 1.5, 2.5, ... are the proper values to return in gl_FragCoord, even though
930 * coordinates specify the pixel centers instead of the pixel corners. This code will behave
931 * correctly on drivers that returns integer values.
933 shader_addline(buffer
, "vpos = floor(vec4(0, ycorrection[0], 0, 0) + gl_FragCoord * vec4(1, ycorrection[1], 1, 1));\n");
937 /*****************************************************************************
938 * Functions to generate GLSL strings from DirectX Shader bytecode begin here.
940 * For more information, see http://wiki.winehq.org/DirectX-Shaders
941 ****************************************************************************/
944 static void shader_glsl_add_src_param(const struct wined3d_shader_instruction
*ins
, const DWORD param
,
945 const DWORD addr_token
, DWORD mask
, glsl_src_param_t
*src_param
);
947 /** Used for opcode modifiers - They multiply the result by the specified amount */
948 static const char * const shift_glsl_tab
[] = {
950 "2.0 * ", /* 1 (x2) */
951 "4.0 * ", /* 2 (x4) */
952 "8.0 * ", /* 3 (x8) */
953 "16.0 * ", /* 4 (x16) */
954 "32.0 * ", /* 5 (x32) */
961 "0.0625 * ", /* 12 (d16) */
962 "0.125 * ", /* 13 (d8) */
963 "0.25 * ", /* 14 (d4) */
964 "0.5 * " /* 15 (d2) */
967 /* Generate a GLSL parameter that does the input modifier computation and return the input register/mask to use */
968 static void shader_glsl_gen_modifier (
971 const char *in_regswizzle
,
976 switch (instr
& WINED3DSP_SRCMOD_MASK
) {
977 case WINED3DSPSM_DZ
: /* Need to handle this in the instructions itself (texld & texcrd). */
979 case WINED3DSPSM_NONE
:
980 sprintf(out_str
, "%s%s", in_reg
, in_regswizzle
);
982 case WINED3DSPSM_NEG
:
983 sprintf(out_str
, "-%s%s", in_reg
, in_regswizzle
);
985 case WINED3DSPSM_NOT
:
986 sprintf(out_str
, "!%s%s", in_reg
, in_regswizzle
);
988 case WINED3DSPSM_BIAS
:
989 sprintf(out_str
, "(%s%s - vec4(0.5)%s)", in_reg
, in_regswizzle
, in_regswizzle
);
991 case WINED3DSPSM_BIASNEG
:
992 sprintf(out_str
, "-(%s%s - vec4(0.5)%s)", in_reg
, in_regswizzle
, in_regswizzle
);
994 case WINED3DSPSM_SIGN
:
995 sprintf(out_str
, "(2.0 * (%s%s - 0.5))", in_reg
, in_regswizzle
);
997 case WINED3DSPSM_SIGNNEG
:
998 sprintf(out_str
, "-(2.0 * (%s%s - 0.5))", in_reg
, in_regswizzle
);
1000 case WINED3DSPSM_COMP
:
1001 sprintf(out_str
, "(1.0 - %s%s)", in_reg
, in_regswizzle
);
1003 case WINED3DSPSM_X2
:
1004 sprintf(out_str
, "(2.0 * %s%s)", in_reg
, in_regswizzle
);
1006 case WINED3DSPSM_X2NEG
:
1007 sprintf(out_str
, "-(2.0 * %s%s)", in_reg
, in_regswizzle
);
1009 case WINED3DSPSM_ABS
:
1010 sprintf(out_str
, "abs(%s%s)", in_reg
, in_regswizzle
);
1012 case WINED3DSPSM_ABSNEG
:
1013 sprintf(out_str
, "-abs(%s%s)", in_reg
, in_regswizzle
);
1016 FIXME("Unhandled modifier %u\n", (instr
& WINED3DSP_SRCMOD_MASK
));
1017 sprintf(out_str
, "%s%s", in_reg
, in_regswizzle
);
1021 /** Writes the GLSL variable name that corresponds to the register that the
1022 * DX opcode parameter is trying to access */
1023 static void shader_glsl_get_register_name(WINED3DSHADER_PARAM_REGISTER_TYPE register_type
, UINT register_idx
,
1024 BOOL rel_addr
, const DWORD addr_token
, char *register_name
, BOOL
*is_color
,
1025 const struct wined3d_shader_instruction
*ins
)
1027 /* oPos, oFog and oPts in D3D */
1028 static const char * const hwrastout_reg_names
[] = { "gl_Position", "gl_FogFragCoord", "gl_PointSize" };
1030 IWineD3DBaseShaderImpl
*This
= (IWineD3DBaseShaderImpl
*)ins
->shader
;
1031 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
1032 const WineD3D_GL_Info
* gl_info
= &deviceImpl
->adapter
->gl_info
;
1033 DWORD shader_version
= This
->baseShader
.reg_maps
.shader_version
;
1034 char pshader
= shader_is_pshader_version(shader_version
);
1038 switch (register_type
)
1040 case WINED3DSPR_TEMP
:
1041 sprintf(register_name
, "R%u", register_idx
);
1043 case WINED3DSPR_INPUT
:
1045 /* Pixel shaders >= 3.0 */
1046 if (WINED3DSHADER_VERSION_MAJOR(shader_version
) >= 3)
1048 DWORD in_count
= GL_LIMITS(glsl_varyings
) / 4;
1052 glsl_src_param_t rel_param
;
1053 shader_glsl_add_src_param(ins
, addr_token
, 0, WINED3DSP_WRITEMASK_0
, &rel_param
);
1055 /* Removing a + 0 would be an obvious optimization, but macos doesn't see the NOP
1058 if (((IWineD3DPixelShaderImpl
*)This
)->input_reg_map
[register_idx
])
1060 if (((IWineD3DPixelShaderImpl
*)This
)->declared_in_count
> in_count
) {
1061 sprintf(register_name
, "((%s + %u) > %d ? (%s + %u) > %d ? gl_SecondaryColor : gl_Color : IN[%s + %u])",
1062 rel_param
.param_str
, ((IWineD3DPixelShaderImpl
*)This
)->input_reg_map
[register_idx
], in_count
- 1,
1063 rel_param
.param_str
, ((IWineD3DPixelShaderImpl
*)This
)->input_reg_map
[register_idx
], in_count
,
1064 rel_param
.param_str
, ((IWineD3DPixelShaderImpl
*)This
)->input_reg_map
[register_idx
]);
1066 sprintf(register_name
, "IN[%s + %u]", rel_param
.param_str
,
1067 ((IWineD3DPixelShaderImpl
*)This
)->input_reg_map
[register_idx
]);
1070 if (((IWineD3DPixelShaderImpl
*)This
)->declared_in_count
> in_count
) {
1071 sprintf(register_name
, "((%s) > %d ? (%s) > %d ? gl_SecondaryColor : gl_Color : IN[%s])",
1072 rel_param
.param_str
, in_count
- 1,
1073 rel_param
.param_str
, in_count
,
1074 rel_param
.param_str
);
1076 sprintf(register_name
, "IN[%s]", rel_param
.param_str
);
1080 DWORD idx
= ((IWineD3DPixelShaderImpl
*) This
)->input_reg_map
[register_idx
];
1081 if (idx
== in_count
) {
1082 sprintf(register_name
, "gl_Color");
1083 } else if (idx
== in_count
+ 1) {
1084 sprintf(register_name
, "gl_SecondaryColor");
1086 sprintf(register_name
, "IN[%u]", idx
);
1090 if (register_idx
== 0)
1091 strcpy(register_name
, "gl_Color");
1093 strcpy(register_name
, "gl_SecondaryColor");
1096 if (((IWineD3DVertexShaderImpl
*)This
)->cur_args
->swizzle_map
& (1 << register_idx
)) *is_color
= TRUE
;
1097 sprintf(register_name
, "attrib%u", register_idx
);
1100 case WINED3DSPR_CONST
:
1102 const char prefix
= pshader
? 'P':'V';
1104 /* Relative addressing */
1107 /* Relative addressing on shaders 2.0+ have a relative address token,
1108 * prior to that, it was hard-coded as "A0.x" because there's only 1 register */
1109 if (WINED3DSHADER_VERSION_MAJOR(shader_version
) >= 2)
1111 glsl_src_param_t rel_param
;
1112 shader_glsl_add_src_param(ins
, addr_token
, 0, WINED3DSP_WRITEMASK_0
, &rel_param
);
1115 sprintf(register_name
, "%cC[%s + %u]", prefix
, rel_param
.param_str
, register_idx
);
1117 sprintf(register_name
, "%cC[%s]", prefix
, rel_param
.param_str
);
1122 sprintf(register_name
, "%cC[A0.x + %u]", prefix
, register_idx
);
1124 sprintf(register_name
, "%cC[A0.x]", prefix
);
1129 if (shader_constant_is_local(This
, register_idx
))
1131 sprintf(register_name
, "%cLC%u", prefix
, register_idx
);
1133 sprintf(register_name
, "%cC[%u]", prefix
, register_idx
);
1139 case WINED3DSPR_CONSTINT
:
1141 sprintf(register_name
, "PI[%u]", register_idx
);
1143 sprintf(register_name
, "VI[%u]", register_idx
);
1145 case WINED3DSPR_CONSTBOOL
:
1147 sprintf(register_name
, "PB[%u]", register_idx
);
1149 sprintf(register_name
, "VB[%u]", register_idx
);
1151 case WINED3DSPR_TEXTURE
: /* case WINED3DSPR_ADDR: */
1153 sprintf(register_name
, "T%u", register_idx
);
1155 sprintf(register_name
, "A%u", register_idx
);
1158 case WINED3DSPR_LOOP
:
1159 sprintf(register_name
, "aL%u", This
->baseShader
.cur_loop_regno
- 1);
1161 case WINED3DSPR_SAMPLER
:
1163 sprintf(register_name
, "Psampler%u", register_idx
);
1165 sprintf(register_name
, "Vsampler%u", register_idx
);
1167 case WINED3DSPR_COLOROUT
:
1168 if (register_idx
>= GL_LIMITS(buffers
))
1169 WARN("Write to render target %u, only %d supported\n", register_idx
, 4);
1171 if (GL_SUPPORT(ARB_DRAW_BUFFERS
)) {
1172 sprintf(register_name
, "gl_FragData[%u]", register_idx
);
1173 } else { /* On older cards with GLSL support like the GeforceFX there's only one buffer. */
1174 sprintf(register_name
, "gl_FragColor");
1177 case WINED3DSPR_RASTOUT
:
1178 sprintf(register_name
, "%s", hwrastout_reg_names
[register_idx
]);
1180 case WINED3DSPR_DEPTHOUT
:
1181 sprintf(register_name
, "gl_FragDepth");
1183 case WINED3DSPR_ATTROUT
:
1184 if (register_idx
== 0)
1186 sprintf(register_name
, "gl_FrontColor");
1188 sprintf(register_name
, "gl_FrontSecondaryColor");
1191 case WINED3DSPR_TEXCRDOUT
:
1192 /* Vertex shaders >= 3.0: WINED3DSPR_OUTPUT */
1193 if (WINED3DSHADER_VERSION_MAJOR(shader_version
) >= 3) sprintf(register_name
, "OUT[%u]", register_idx
);
1194 else sprintf(register_name
, "gl_TexCoord[%u]", register_idx
);
1196 case WINED3DSPR_MISCTYPE
:
1197 if (register_idx
== 0)
1200 sprintf(register_name
, "vpos");
1202 else if (register_idx
== 1)
1204 /* Note that gl_FrontFacing is a bool, while vFace is
1205 * a float for which the sign determines front/back
1207 sprintf(register_name
, "(gl_FrontFacing ? 1.0 : -1.0)");
1209 FIXME("Unhandled misctype register %d\n", register_idx
);
1210 sprintf(register_name
, "unrecognized_register");
1214 FIXME("Unhandled register name Type(%d)\n", register_type
);
1215 sprintf(register_name
, "unrecognized_register");
1220 static void shader_glsl_write_mask_to_str(DWORD write_mask
, char *str
)
1223 if (write_mask
& WINED3DSP_WRITEMASK_0
) *str
++ = 'x';
1224 if (write_mask
& WINED3DSP_WRITEMASK_1
) *str
++ = 'y';
1225 if (write_mask
& WINED3DSP_WRITEMASK_2
) *str
++ = 'z';
1226 if (write_mask
& WINED3DSP_WRITEMASK_3
) *str
++ = 'w';
1230 /* Get the GLSL write mask for the destination register */
1231 static DWORD
shader_glsl_get_write_mask(const struct wined3d_shader_dst_param
*param
, char *write_mask
)
1233 DWORD mask
= param
->write_mask
;
1235 if (shader_is_scalar(param
->register_type
, param
->register_idx
))
1237 mask
= WINED3DSP_WRITEMASK_0
;
1242 shader_glsl_write_mask_to_str(mask
, write_mask
);
1248 static unsigned int shader_glsl_get_write_mask_size(DWORD write_mask
) {
1249 unsigned int size
= 0;
1251 if (write_mask
& WINED3DSP_WRITEMASK_0
) ++size
;
1252 if (write_mask
& WINED3DSP_WRITEMASK_1
) ++size
;
1253 if (write_mask
& WINED3DSP_WRITEMASK_2
) ++size
;
1254 if (write_mask
& WINED3DSP_WRITEMASK_3
) ++size
;
1259 static void shader_glsl_get_swizzle(const DWORD param
, BOOL fixup
, DWORD mask
, char *swizzle_str
) {
1260 /* For registers of type WINED3DDECLTYPE_D3DCOLOR, data is stored as "bgra",
1261 * but addressed as "rgba". To fix this we need to swap the register's x
1262 * and z components. */
1263 DWORD swizzle
= (param
& WINED3DSP_SWIZZLE_MASK
) >> WINED3DSP_SWIZZLE_SHIFT
;
1264 const char *swizzle_chars
= fixup
? "zyxw" : "xyzw";
1265 char *ptr
= swizzle_str
;
1267 if (!shader_is_scalar(shader_get_regtype(param
), param
& WINED3DSP_REGNUM_MASK
))
1270 /* swizzle bits fields: wwzzyyxx */
1271 if (mask
& WINED3DSP_WRITEMASK_0
) *ptr
++ = swizzle_chars
[swizzle
& 0x03];
1272 if (mask
& WINED3DSP_WRITEMASK_1
) *ptr
++ = swizzle_chars
[(swizzle
>> 2) & 0x03];
1273 if (mask
& WINED3DSP_WRITEMASK_2
) *ptr
++ = swizzle_chars
[(swizzle
>> 4) & 0x03];
1274 if (mask
& WINED3DSP_WRITEMASK_3
) *ptr
++ = swizzle_chars
[(swizzle
>> 6) & 0x03];
1280 /* From a given parameter token, generate the corresponding GLSL string.
1281 * Also, return the actual register name and swizzle in case the
1282 * caller needs this information as well. */
1283 static void shader_glsl_add_src_param(const struct wined3d_shader_instruction
*ins
,
1284 const DWORD param
, const DWORD addr_token
, DWORD mask
, glsl_src_param_t
*src_param
)
1286 BOOL is_color
= FALSE
;
1287 char swizzle_str
[6];
1289 src_param
->reg_name
[0] = '\0';
1290 src_param
->param_str
[0] = '\0';
1291 swizzle_str
[0] = '\0';
1293 shader_glsl_get_register_name(shader_get_regtype(param
), param
& WINED3DSP_REGNUM_MASK
,
1294 param
& WINED3DSHADER_ADDRMODE_RELATIVE
, addr_token
, src_param
->reg_name
, &is_color
, ins
);
1296 shader_glsl_get_swizzle(param
, is_color
, mask
, swizzle_str
);
1297 shader_glsl_gen_modifier(param
, src_param
->reg_name
, swizzle_str
, src_param
->param_str
);
1300 /* From a given parameter token, generate the corresponding GLSL string.
1301 * Also, return the actual register name and swizzle in case the
1302 * caller needs this information as well. */
1303 static DWORD
shader_glsl_add_dst_param(const struct wined3d_shader_instruction
*ins
,
1304 const struct wined3d_shader_dst_param
*wined3d_dst
, glsl_dst_param_t
*glsl_dst
)
1306 BOOL is_color
= FALSE
;
1308 glsl_dst
->mask_str
[0] = '\0';
1309 glsl_dst
->reg_name
[0] = '\0';
1311 shader_glsl_get_register_name(wined3d_dst
->register_type
, wined3d_dst
->register_idx
,
1312 wined3d_dst
->has_rel_addr
, wined3d_dst
->addr_token
, glsl_dst
->reg_name
, &is_color
, ins
);
1313 return shader_glsl_get_write_mask(wined3d_dst
, glsl_dst
->mask_str
);
1316 /* Append the destination part of the instruction to the buffer, return the effective write mask */
1317 static DWORD
shader_glsl_append_dst_ext(SHADER_BUFFER
*buffer
,
1318 const struct wined3d_shader_instruction
*ins
, const struct wined3d_shader_dst_param
*dst
)
1320 glsl_dst_param_t glsl_dst
;
1323 mask
= shader_glsl_add_dst_param(ins
, dst
, &glsl_dst
);
1324 if (mask
) shader_addline(buffer
, "%s%s = %s(", glsl_dst
.reg_name
, glsl_dst
.mask_str
, shift_glsl_tab
[dst
->shift
]);
1329 /* Append the destination part of the instruction to the buffer, return the effective write mask */
1330 static DWORD
shader_glsl_append_dst(SHADER_BUFFER
*buffer
, const struct wined3d_shader_instruction
*ins
)
1332 return shader_glsl_append_dst_ext(buffer
, ins
, &ins
->dst
[0]);
1335 /** Process GLSL instruction modifiers */
1336 void shader_glsl_add_instruction_modifiers(const struct wined3d_shader_instruction
*ins
)
1338 glsl_dst_param_t dst_param
;
1341 if (!ins
->dst_count
) return;
1343 modifiers
= ins
->dst
[0].modifiers
;
1344 if (!modifiers
) return;
1346 shader_glsl_add_dst_param(ins
, &ins
->dst
[0], &dst_param
);
1348 if (modifiers
& WINED3DSPDM_SATURATE
)
1350 /* _SAT means to clamp the value of the register to between 0 and 1 */
1351 shader_addline(ins
->buffer
, "%s%s = clamp(%s%s, 0.0, 1.0);\n", dst_param
.reg_name
,
1352 dst_param
.mask_str
, dst_param
.reg_name
, dst_param
.mask_str
);
1355 if (modifiers
& WINED3DSPDM_MSAMPCENTROID
)
1357 FIXME("_centroid modifier not handled\n");
1360 if (modifiers
& WINED3DSPDM_PARTIALPRECISION
)
1362 /* MSDN says this modifier can be safely ignored, so that's what we'll do. */
1366 static inline const char *shader_get_comp_op(DWORD flags
)
1368 DWORD op
= (flags
& INST_CONTROLS_MASK
) >> INST_CONTROLS_SHIFT
;
1370 case COMPARISON_GT
: return ">";
1371 case COMPARISON_EQ
: return "==";
1372 case COMPARISON_GE
: return ">=";
1373 case COMPARISON_LT
: return "<";
1374 case COMPARISON_NE
: return "!=";
1375 case COMPARISON_LE
: return "<=";
1377 FIXME("Unrecognized comparison value: %u\n", op
);
1382 static void shader_glsl_get_sample_function(DWORD sampler_type
, DWORD flags
, glsl_sample_function_t
*sample_function
)
1384 BOOL projected
= flags
& WINED3D_GLSL_SAMPLE_PROJECTED
;
1385 BOOL texrect
= flags
& WINED3D_GLSL_SAMPLE_RECT
;
1386 BOOL lod
= flags
& WINED3D_GLSL_SAMPLE_LOD
;
1388 /* Note that there's no such thing as a projected cube texture. */
1389 switch(sampler_type
) {
1392 sample_function
->name
= projected
? "texture1DProjLod" : "texture1DLod";
1394 sample_function
->name
= projected
? "texture1DProj" : "texture1D";
1396 sample_function
->coord_mask
= WINED3DSP_WRITEMASK_0
;
1401 sample_function
->name
= projected
? "texture2DRectProjLod" : "texture2DRectLod";
1403 sample_function
->name
= projected
? "texture2DRectProj" : "texture2DRect";
1407 sample_function
->name
= projected
? "texture2DProjLod" : "texture2DLod";
1409 sample_function
->name
= projected
? "texture2DProj" : "texture2D";
1412 sample_function
->coord_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
;
1414 case WINED3DSTT_CUBE
:
1416 sample_function
->name
= "textureCubeLod";
1418 sample_function
->name
= "textureCube";
1420 sample_function
->coord_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
1422 case WINED3DSTT_VOLUME
:
1424 sample_function
->name
= projected
? "texture3DProjLod" : "texture3DLod";
1426 sample_function
->name
= projected
? "texture3DProj" : "texture3D";
1428 sample_function
->coord_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
1431 sample_function
->name
= "";
1432 sample_function
->coord_mask
= 0;
1433 FIXME("Unrecognized sampler type: %#x;\n", sampler_type
);
1438 static void shader_glsl_append_fixup_arg(char *arguments
, const char *reg_name
,
1439 BOOL sign_fixup
, enum fixup_channel_source channel_source
)
1441 switch(channel_source
)
1443 case CHANNEL_SOURCE_ZERO
:
1444 strcat(arguments
, "0.0");
1447 case CHANNEL_SOURCE_ONE
:
1448 strcat(arguments
, "1.0");
1451 case CHANNEL_SOURCE_X
:
1452 strcat(arguments
, reg_name
);
1453 strcat(arguments
, ".x");
1456 case CHANNEL_SOURCE_Y
:
1457 strcat(arguments
, reg_name
);
1458 strcat(arguments
, ".y");
1461 case CHANNEL_SOURCE_Z
:
1462 strcat(arguments
, reg_name
);
1463 strcat(arguments
, ".z");
1466 case CHANNEL_SOURCE_W
:
1467 strcat(arguments
, reg_name
);
1468 strcat(arguments
, ".w");
1472 FIXME("Unhandled channel source %#x\n", channel_source
);
1473 strcat(arguments
, "undefined");
1477 if (sign_fixup
) strcat(arguments
, " * 2.0 - 1.0");
1480 static void shader_glsl_color_correction(const struct wined3d_shader_instruction
*ins
, struct color_fixup_desc fixup
)
1482 struct wined3d_shader_dst_param dst
;
1483 unsigned int mask_size
, remaining
;
1484 glsl_dst_param_t dst_param
;
1485 char arguments
[256];
1489 if (fixup
.x_sign_fixup
|| fixup
.x_source
!= CHANNEL_SOURCE_X
) mask
|= WINED3DSP_WRITEMASK_0
;
1490 if (fixup
.y_sign_fixup
|| fixup
.y_source
!= CHANNEL_SOURCE_Y
) mask
|= WINED3DSP_WRITEMASK_1
;
1491 if (fixup
.z_sign_fixup
|| fixup
.z_source
!= CHANNEL_SOURCE_Z
) mask
|= WINED3DSP_WRITEMASK_2
;
1492 if (fixup
.w_sign_fixup
|| fixup
.w_source
!= CHANNEL_SOURCE_W
) mask
|= WINED3DSP_WRITEMASK_3
;
1493 mask
&= ins
->dst
[0].write_mask
;
1495 if (!mask
) return; /* Nothing to do */
1497 if (is_yuv_fixup(fixup
))
1499 enum yuv_fixup yuv_fixup
= get_yuv_fixup(fixup
);
1500 FIXME("YUV fixup (%#x) not supported\n", yuv_fixup
);
1504 mask_size
= shader_glsl_get_write_mask_size(mask
);
1507 dst
.write_mask
= mask
;
1508 shader_glsl_add_dst_param(ins
, &dst
, &dst_param
);
1510 arguments
[0] = '\0';
1511 remaining
= mask_size
;
1512 if (mask
& WINED3DSP_WRITEMASK_0
)
1514 shader_glsl_append_fixup_arg(arguments
, dst_param
.reg_name
, fixup
.x_sign_fixup
, fixup
.x_source
);
1515 if (--remaining
) strcat(arguments
, ", ");
1517 if (mask
& WINED3DSP_WRITEMASK_1
)
1519 shader_glsl_append_fixup_arg(arguments
, dst_param
.reg_name
, fixup
.y_sign_fixup
, fixup
.y_source
);
1520 if (--remaining
) strcat(arguments
, ", ");
1522 if (mask
& WINED3DSP_WRITEMASK_2
)
1524 shader_glsl_append_fixup_arg(arguments
, dst_param
.reg_name
, fixup
.z_sign_fixup
, fixup
.z_source
);
1525 if (--remaining
) strcat(arguments
, ", ");
1527 if (mask
& WINED3DSP_WRITEMASK_3
)
1529 shader_glsl_append_fixup_arg(arguments
, dst_param
.reg_name
, fixup
.w_sign_fixup
, fixup
.w_source
);
1530 if (--remaining
) strcat(arguments
, ", ");
1535 shader_addline(ins
->buffer
, "%s%s = vec%u(%s);\n",
1536 dst_param
.reg_name
, dst_param
.mask_str
, mask_size
, arguments
);
1540 shader_addline(ins
->buffer
, "%s%s = %s;\n", dst_param
.reg_name
, dst_param
.mask_str
, arguments
);
1544 static void PRINTF_ATTR(6, 7) shader_glsl_gen_sample_code(const struct wined3d_shader_instruction
*ins
,
1545 DWORD sampler
, const glsl_sample_function_t
*sample_function
, DWORD swizzle
,
1546 const char *bias
, const char *coord_reg_fmt
, ...)
1548 const char *sampler_base
;
1549 char dst_swizzle
[6];
1550 struct color_fixup_desc fixup
;
1551 BOOL np2_fixup
= FALSE
;
1554 shader_glsl_get_swizzle(swizzle
, FALSE
, ins
->dst
[0].write_mask
, dst_swizzle
);
1556 if (shader_is_pshader_version(ins
->reg_maps
->shader_version
))
1558 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*)ins
->shader
;
1559 fixup
= This
->cur_args
->color_fixup
[sampler
];
1560 sampler_base
= "Psampler";
1562 if(This
->cur_args
->np2_fixup
& (1 << sampler
)) {
1564 FIXME("Biased sampling from NP2 textures is unsupported\n");
1570 sampler_base
= "Vsampler";
1571 fixup
= COLOR_FIXUP_IDENTITY
; /* FIXME: Vshader color fixup */
1574 shader_glsl_append_dst(ins
->buffer
, ins
);
1576 shader_addline(ins
->buffer
, "%s(%s%u, ", sample_function
->name
, sampler_base
, sampler
);
1578 va_start(args
, coord_reg_fmt
);
1579 shader_vaddline(ins
->buffer
, coord_reg_fmt
, args
);
1583 shader_addline(ins
->buffer
, ", %s)%s);\n", bias
, dst_swizzle
);
1586 shader_addline(ins
->buffer
, " * PsamplerNP2Fixup%u)%s);\n", sampler
, dst_swizzle
);
1588 shader_addline(ins
->buffer
, ")%s);\n", dst_swizzle
);
1592 if(!is_identity_fixup(fixup
)) {
1593 shader_glsl_color_correction(ins
, fixup
);
1597 /*****************************************************************************
1599 * Begin processing individual instruction opcodes
1601 ****************************************************************************/
1603 /* Generate GLSL arithmetic functions (dst = src1 + src2) */
1604 static void shader_glsl_arith(const struct wined3d_shader_instruction
*ins
)
1606 SHADER_BUFFER
*buffer
= ins
->buffer
;
1607 glsl_src_param_t src0_param
;
1608 glsl_src_param_t src1_param
;
1612 /* Determine the GLSL operator to use based on the opcode */
1613 switch (ins
->handler_idx
)
1615 case WINED3DSIH_MUL
: op
= '*'; break;
1616 case WINED3DSIH_ADD
: op
= '+'; break;
1617 case WINED3DSIH_SUB
: op
= '-'; break;
1620 FIXME("Opcode %#x not yet handled in GLSL\n", ins
->handler_idx
);
1624 write_mask
= shader_glsl_append_dst(buffer
, ins
);
1625 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], write_mask
, &src0_param
);
1626 shader_glsl_add_src_param(ins
, ins
->src
[1], ins
->src_addr
[1], write_mask
, &src1_param
);
1627 shader_addline(buffer
, "%s %c %s);\n", src0_param
.param_str
, op
, src1_param
.param_str
);
1630 /* Process the WINED3DSIO_MOV opcode using GLSL (dst = src) */
1631 static void shader_glsl_mov(const struct wined3d_shader_instruction
*ins
)
1633 SHADER_BUFFER
*buffer
= ins
->buffer
;
1634 glsl_src_param_t src0_param
;
1637 write_mask
= shader_glsl_append_dst(buffer
, ins
);
1638 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], write_mask
, &src0_param
);
1640 /* In vs_1_1 WINED3DSIO_MOV can write to the address register. In later
1641 * shader versions WINED3DSIO_MOVA is used for this. */
1642 if ((WINED3DSHADER_VERSION_MAJOR(ins
->reg_maps
->shader_version
) == 1
1643 && !shader_is_pshader_version(ins
->reg_maps
->shader_version
)
1644 && ins
->dst
[0].register_type
== WINED3DSPR_ADDR
))
1646 /* This is a simple floor() */
1647 unsigned int mask_size
= shader_glsl_get_write_mask_size(write_mask
);
1648 if (mask_size
> 1) {
1649 shader_addline(buffer
, "ivec%d(floor(%s)));\n", mask_size
, src0_param
.param_str
);
1651 shader_addline(buffer
, "int(floor(%s)));\n", src0_param
.param_str
);
1654 else if(ins
->handler_idx
== WINED3DSIH_MOVA
)
1656 /* We need to *round* to the nearest int here. */
1657 unsigned int mask_size
= shader_glsl_get_write_mask_size(write_mask
);
1658 if (mask_size
> 1) {
1659 shader_addline(buffer
, "ivec%d(floor(abs(%s) + vec%d(0.5)) * sign(%s)));\n", mask_size
, src0_param
.param_str
, mask_size
, src0_param
.param_str
);
1661 shader_addline(buffer
, "int(floor(abs(%s) + 0.5) * sign(%s)));\n", src0_param
.param_str
, src0_param
.param_str
);
1664 shader_addline(buffer
, "%s);\n", src0_param
.param_str
);
1668 /* Process the dot product operators DP3 and DP4 in GLSL (dst = dot(src0, src1)) */
1669 static void shader_glsl_dot(const struct wined3d_shader_instruction
*ins
)
1671 SHADER_BUFFER
*buffer
= ins
->buffer
;
1672 glsl_src_param_t src0_param
;
1673 glsl_src_param_t src1_param
;
1674 DWORD dst_write_mask
, src_write_mask
;
1675 unsigned int dst_size
= 0;
1677 dst_write_mask
= shader_glsl_append_dst(buffer
, ins
);
1678 dst_size
= shader_glsl_get_write_mask_size(dst_write_mask
);
1680 /* dp3 works on vec3, dp4 on vec4 */
1681 if (ins
->handler_idx
== WINED3DSIH_DP4
)
1683 src_write_mask
= WINED3DSP_WRITEMASK_ALL
;
1685 src_write_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
1688 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], src_write_mask
, &src0_param
);
1689 shader_glsl_add_src_param(ins
, ins
->src
[1], ins
->src_addr
[1], src_write_mask
, &src1_param
);
1692 shader_addline(buffer
, "vec%d(dot(%s, %s)));\n", dst_size
, src0_param
.param_str
, src1_param
.param_str
);
1694 shader_addline(buffer
, "dot(%s, %s));\n", src0_param
.param_str
, src1_param
.param_str
);
1698 /* Note that this instruction has some restrictions. The destination write mask
1699 * can't contain the w component, and the source swizzles have to be .xyzw */
1700 static void shader_glsl_cross(const struct wined3d_shader_instruction
*ins
)
1702 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
1703 glsl_src_param_t src0_param
;
1704 glsl_src_param_t src1_param
;
1707 shader_glsl_get_write_mask(&ins
->dst
[0], dst_mask
);
1708 shader_glsl_append_dst(ins
->buffer
, ins
);
1709 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], src_mask
, &src0_param
);
1710 shader_glsl_add_src_param(ins
, ins
->src
[1], ins
->src_addr
[1], src_mask
, &src1_param
);
1711 shader_addline(ins
->buffer
, "cross(%s, %s)%s);\n", src0_param
.param_str
, src1_param
.param_str
, dst_mask
);
1714 /* Process the WINED3DSIO_POW instruction in GLSL (dst = |src0|^src1)
1715 * Src0 and src1 are scalars. Note that D3D uses the absolute of src0, while
1716 * GLSL uses the value as-is. */
1717 static void shader_glsl_pow(const struct wined3d_shader_instruction
*ins
)
1719 SHADER_BUFFER
*buffer
= ins
->buffer
;
1720 glsl_src_param_t src0_param
;
1721 glsl_src_param_t src1_param
;
1722 DWORD dst_write_mask
;
1723 unsigned int dst_size
;
1725 dst_write_mask
= shader_glsl_append_dst(buffer
, ins
);
1726 dst_size
= shader_glsl_get_write_mask_size(dst_write_mask
);
1728 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
1729 shader_glsl_add_src_param(ins
, ins
->src
[1], ins
->src_addr
[1], WINED3DSP_WRITEMASK_0
, &src1_param
);
1732 shader_addline(buffer
, "vec%d(pow(abs(%s), %s)));\n", dst_size
, src0_param
.param_str
, src1_param
.param_str
);
1734 shader_addline(buffer
, "pow(abs(%s), %s));\n", src0_param
.param_str
, src1_param
.param_str
);
1738 /* Process the WINED3DSIO_LOG instruction in GLSL (dst = log2(|src0|))
1739 * Src0 is a scalar. Note that D3D uses the absolute of src0, while
1740 * GLSL uses the value as-is. */
1741 static void shader_glsl_log(const struct wined3d_shader_instruction
*ins
)
1743 SHADER_BUFFER
*buffer
= ins
->buffer
;
1744 glsl_src_param_t src0_param
;
1745 DWORD dst_write_mask
;
1746 unsigned int dst_size
;
1748 dst_write_mask
= shader_glsl_append_dst(buffer
, ins
);
1749 dst_size
= shader_glsl_get_write_mask_size(dst_write_mask
);
1751 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
1754 shader_addline(buffer
, "vec%d(log2(abs(%s))));\n", dst_size
, src0_param
.param_str
);
1756 shader_addline(buffer
, "log2(abs(%s)));\n", src0_param
.param_str
);
1760 /* Map the opcode 1-to-1 to the GL code (arg->dst = instruction(src0, src1, ...) */
1761 static void shader_glsl_map2gl(const struct wined3d_shader_instruction
*ins
)
1763 SHADER_BUFFER
*buffer
= ins
->buffer
;
1764 glsl_src_param_t src_param
;
1765 const char *instruction
;
1769 /* Determine the GLSL function to use based on the opcode */
1770 /* TODO: Possibly make this a table for faster lookups */
1771 switch (ins
->handler_idx
)
1773 case WINED3DSIH_MIN
: instruction
= "min"; break;
1774 case WINED3DSIH_MAX
: instruction
= "max"; break;
1775 case WINED3DSIH_ABS
: instruction
= "abs"; break;
1776 case WINED3DSIH_FRC
: instruction
= "fract"; break;
1777 case WINED3DSIH_NRM
: instruction
= "normalize"; break;
1778 case WINED3DSIH_EXP
: instruction
= "exp2"; break;
1779 case WINED3DSIH_SGN
: instruction
= "sign"; break;
1780 case WINED3DSIH_DSX
: instruction
= "dFdx"; break;
1781 case WINED3DSIH_DSY
: instruction
= "ycorrection.y * dFdy"; break;
1782 default: instruction
= "";
1783 FIXME("Opcode %#x not yet handled in GLSL\n", ins
->handler_idx
);
1787 write_mask
= shader_glsl_append_dst(buffer
, ins
);
1789 shader_addline(buffer
, "%s(", instruction
);
1793 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], write_mask
, &src_param
);
1794 shader_addline(buffer
, "%s", src_param
.param_str
);
1795 for (i
= 1; i
< ins
->src_count
; ++i
)
1797 shader_glsl_add_src_param(ins
, ins
->src
[i
], ins
->src_addr
[i
], write_mask
, &src_param
);
1798 shader_addline(buffer
, ", %s", src_param
.param_str
);
1802 shader_addline(buffer
, "));\n");
1805 /** Process the WINED3DSIO_EXPP instruction in GLSL:
1806 * For shader model 1.x, do the following (and honor the writemask, so use a temporary variable):
1807 * dst.x = 2^(floor(src))
1808 * dst.y = src - floor(src)
1809 * dst.z = 2^src (partial precision is allowed, but optional)
1811 * For 2.0 shaders, just do this (honoring writemask and swizzle):
1812 * dst = 2^src; (partial precision is allowed, but optional)
1814 static void shader_glsl_expp(const struct wined3d_shader_instruction
*ins
)
1816 glsl_src_param_t src_param
;
1818 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src_param
);
1820 if (ins
->reg_maps
->shader_version
< WINED3DPS_VERSION(2,0))
1824 shader_addline(ins
->buffer
, "tmp0.x = exp2(floor(%s));\n", src_param
.param_str
);
1825 shader_addline(ins
->buffer
, "tmp0.y = %s - floor(%s);\n", src_param
.param_str
, src_param
.param_str
);
1826 shader_addline(ins
->buffer
, "tmp0.z = exp2(%s);\n", src_param
.param_str
);
1827 shader_addline(ins
->buffer
, "tmp0.w = 1.0;\n");
1829 shader_glsl_append_dst(ins
->buffer
, ins
);
1830 shader_glsl_get_write_mask(&ins
->dst
[0], dst_mask
);
1831 shader_addline(ins
->buffer
, "tmp0%s);\n", dst_mask
);
1834 unsigned int mask_size
;
1836 write_mask
= shader_glsl_append_dst(ins
->buffer
, ins
);
1837 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
1839 if (mask_size
> 1) {
1840 shader_addline(ins
->buffer
, "vec%d(exp2(%s)));\n", mask_size
, src_param
.param_str
);
1842 shader_addline(ins
->buffer
, "exp2(%s));\n", src_param
.param_str
);
1847 /** Process the RCP (reciprocal or inverse) opcode in GLSL (dst = 1 / src) */
1848 static void shader_glsl_rcp(const struct wined3d_shader_instruction
*ins
)
1850 glsl_src_param_t src_param
;
1852 unsigned int mask_size
;
1854 write_mask
= shader_glsl_append_dst(ins
->buffer
, ins
);
1855 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
1856 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], WINED3DSP_WRITEMASK_3
, &src_param
);
1858 if (mask_size
> 1) {
1859 shader_addline(ins
->buffer
, "vec%d(1.0 / %s));\n", mask_size
, src_param
.param_str
);
1861 shader_addline(ins
->buffer
, "1.0 / %s);\n", src_param
.param_str
);
1865 static void shader_glsl_rsq(const struct wined3d_shader_instruction
*ins
)
1867 SHADER_BUFFER
*buffer
= ins
->buffer
;
1868 glsl_src_param_t src_param
;
1870 unsigned int mask_size
;
1872 write_mask
= shader_glsl_append_dst(buffer
, ins
);
1873 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
1875 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], WINED3DSP_WRITEMASK_3
, &src_param
);
1877 if (mask_size
> 1) {
1878 shader_addline(buffer
, "vec%d(inversesqrt(%s)));\n", mask_size
, src_param
.param_str
);
1880 shader_addline(buffer
, "inversesqrt(%s));\n", src_param
.param_str
);
1884 /** Process signed comparison opcodes in GLSL. */
1885 static void shader_glsl_compare(const struct wined3d_shader_instruction
*ins
)
1887 glsl_src_param_t src0_param
;
1888 glsl_src_param_t src1_param
;
1890 unsigned int mask_size
;
1892 write_mask
= shader_glsl_append_dst(ins
->buffer
, ins
);
1893 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
1894 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], write_mask
, &src0_param
);
1895 shader_glsl_add_src_param(ins
, ins
->src
[1], ins
->src_addr
[1], write_mask
, &src1_param
);
1897 if (mask_size
> 1) {
1898 const char *compare
;
1900 switch(ins
->handler_idx
)
1902 case WINED3DSIH_SLT
: compare
= "lessThan"; break;
1903 case WINED3DSIH_SGE
: compare
= "greaterThanEqual"; break;
1904 default: compare
= "";
1905 FIXME("Can't handle opcode %#x\n", ins
->handler_idx
);
1908 shader_addline(ins
->buffer
, "vec%d(%s(%s, %s)));\n", mask_size
, compare
,
1909 src0_param
.param_str
, src1_param
.param_str
);
1911 switch(ins
->handler_idx
)
1913 case WINED3DSIH_SLT
:
1914 /* Step(src0, src1) is not suitable here because if src0 == src1 SLT is supposed,
1915 * to return 0.0 but step returns 1.0 because step is not < x
1916 * An alternative is a bvec compare padded with an unused second component.
1917 * step(src1 * -1.0, src0 * -1.0) is not an option because it suffers from the same
1918 * issue. Playing with not() is not possible either because not() does not accept
1921 shader_addline(ins
->buffer
, "(%s < %s) ? 1.0 : 0.0);\n", src0_param
.param_str
, src1_param
.param_str
);
1923 case WINED3DSIH_SGE
:
1924 /* Here we can use the step() function and safe a conditional */
1925 shader_addline(ins
->buffer
, "step(%s, %s));\n", src1_param
.param_str
, src0_param
.param_str
);
1928 FIXME("Can't handle opcode %#x\n", ins
->handler_idx
);
1934 /** Process CMP instruction in GLSL (dst = src0 >= 0.0 ? src1 : src2), per channel */
1935 static void shader_glsl_cmp(const struct wined3d_shader_instruction
*ins
)
1937 glsl_src_param_t src0_param
;
1938 glsl_src_param_t src1_param
;
1939 glsl_src_param_t src2_param
;
1940 DWORD write_mask
, cmp_channel
= 0;
1943 BOOL temp_destination
= FALSE
;
1945 if (shader_is_scalar(shader_get_regtype(ins
->src
[0]), ins
->src
[0] & WINED3DSP_REGNUM_MASK
))
1947 write_mask
= shader_glsl_append_dst(ins
->buffer
, ins
);
1949 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], WINED3DSP_WRITEMASK_ALL
, &src0_param
);
1950 shader_glsl_add_src_param(ins
, ins
->src
[1], ins
->src_addr
[1], write_mask
, &src1_param
);
1951 shader_glsl_add_src_param(ins
, ins
->src
[2], ins
->src_addr
[2], write_mask
, &src2_param
);
1953 shader_addline(ins
->buffer
, "%s >= 0.0 ? %s : %s);\n",
1954 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
1956 DWORD src0reg
= ins
->src
[0] & WINED3DSP_REGNUM_MASK
;
1957 DWORD src1reg
= ins
->src
[1] & WINED3DSP_REGNUM_MASK
;
1958 DWORD src2reg
= ins
->src
[2] & WINED3DSP_REGNUM_MASK
;
1959 DWORD src0regtype
= shader_get_regtype(ins
->src
[0]);
1960 DWORD src1regtype
= shader_get_regtype(ins
->src
[1]);
1961 DWORD src2regtype
= shader_get_regtype(ins
->src
[2]);
1962 DWORD dstreg
= ins
->dst
[0].register_idx
;
1963 DWORD dstregtype
= ins
->dst
[0].register_type
;
1964 DWORD dst_mask
= ins
->dst
[0].write_mask
;
1965 struct wined3d_shader_dst_param dst
= ins
->dst
[0];
1967 /* Cycle through all source0 channels */
1968 for (i
=0; i
<4; i
++) {
1970 /* Find the destination channels which use the current source0 channel */
1971 for (j
=0; j
<4; j
++) {
1972 if (((ins
->src
[0] >> (WINED3DSP_SWIZZLE_SHIFT
+ 2 * j
)) & 0x3) == i
)
1974 write_mask
|= WINED3DSP_WRITEMASK_0
<< j
;
1975 cmp_channel
= WINED3DSP_WRITEMASK_0
<< j
;
1978 dst
.write_mask
= dst_mask
& write_mask
;
1980 /* Splitting the cmp instruction up in multiple lines imposes a problem:
1981 * The first lines may overwrite source parameters of the following lines.
1982 * Deal with that by using a temporary destination register if needed
1984 if ((src0reg
== dstreg
&& src0regtype
== dstregtype
)
1985 || (src1reg
== dstreg
&& src1regtype
== dstregtype
)
1986 || (src2reg
== dstreg
&& src2regtype
== dstregtype
))
1988 write_mask
= shader_glsl_get_write_mask(&dst
, mask_char
);
1989 if (!write_mask
) continue;
1990 shader_addline(ins
->buffer
, "tmp0%s = (", mask_char
);
1991 temp_destination
= TRUE
;
1993 write_mask
= shader_glsl_append_dst_ext(ins
->buffer
, ins
, &dst
);
1994 if (!write_mask
) continue;
1997 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], cmp_channel
, &src0_param
);
1998 shader_glsl_add_src_param(ins
, ins
->src
[1], ins
->src_addr
[1], write_mask
, &src1_param
);
1999 shader_glsl_add_src_param(ins
, ins
->src
[2], ins
->src_addr
[2], write_mask
, &src2_param
);
2001 shader_addline(ins
->buffer
, "%s >= 0.0 ? %s : %s);\n",
2002 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
2005 if(temp_destination
) {
2006 shader_glsl_get_write_mask(&ins
->dst
[0], mask_char
);
2007 shader_glsl_append_dst(ins
->buffer
, ins
);
2008 shader_addline(ins
->buffer
, "tmp0%s);\n", mask_char
);
2014 /** Process the CND opcode in GLSL (dst = (src0 > 0.5) ? src1 : src2) */
2015 /* For ps 1.1-1.3, only a single component of src0 is used. For ps 1.4
2016 * the compare is done per component of src0. */
2017 static void shader_glsl_cnd(const struct wined3d_shader_instruction
*ins
)
2019 struct wined3d_shader_dst_param dst
;
2020 glsl_src_param_t src0_param
;
2021 glsl_src_param_t src1_param
;
2022 glsl_src_param_t src2_param
;
2023 DWORD write_mask
, cmp_channel
= 0;
2027 if (ins
->reg_maps
->shader_version
< WINED3DPS_VERSION(1, 4))
2029 write_mask
= shader_glsl_append_dst(ins
->buffer
, ins
);
2030 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2031 shader_glsl_add_src_param(ins
, ins
->src
[1], ins
->src_addr
[1], write_mask
, &src1_param
);
2032 shader_glsl_add_src_param(ins
, ins
->src
[2], ins
->src_addr
[2], write_mask
, &src2_param
);
2034 /* Fun: The D3DSI_COISSUE flag changes the semantic of the cnd instruction for < 1.4 shaders */
2037 shader_addline(ins
->buffer
, "%s /* COISSUE! */);\n", src1_param
.param_str
);
2039 shader_addline(ins
->buffer
, "%s > 0.5 ? %s : %s);\n",
2040 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
2044 /* Cycle through all source0 channels */
2045 dst_mask
= ins
->dst
[0].write_mask
;
2047 for (i
=0; i
<4; i
++) {
2049 /* Find the destination channels which use the current source0 channel */
2050 for (j
=0; j
<4; j
++) {
2051 if (((ins
->src
[0] >> (WINED3DSP_SWIZZLE_SHIFT
+ 2 * j
)) & 0x3) == i
)
2053 write_mask
|= WINED3DSP_WRITEMASK_0
<< j
;
2054 cmp_channel
= WINED3DSP_WRITEMASK_0
<< j
;
2058 dst
.write_mask
= dst_mask
& write_mask
;
2059 write_mask
= shader_glsl_append_dst_ext(ins
->buffer
, ins
, &dst
);
2060 if (!write_mask
) continue;
2062 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], cmp_channel
, &src0_param
);
2063 shader_glsl_add_src_param(ins
, ins
->src
[1], ins
->src_addr
[1], write_mask
, &src1_param
);
2064 shader_glsl_add_src_param(ins
, ins
->src
[2], ins
->src_addr
[2], write_mask
, &src2_param
);
2066 shader_addline(ins
->buffer
, "%s > 0.5 ? %s : %s);\n",
2067 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
2071 /** GLSL code generation for WINED3DSIO_MAD: Multiply the first 2 opcodes, then add the last */
2072 static void shader_glsl_mad(const struct wined3d_shader_instruction
*ins
)
2074 glsl_src_param_t src0_param
;
2075 glsl_src_param_t src1_param
;
2076 glsl_src_param_t src2_param
;
2079 write_mask
= shader_glsl_append_dst(ins
->buffer
, ins
);
2080 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], write_mask
, &src0_param
);
2081 shader_glsl_add_src_param(ins
, ins
->src
[1], ins
->src_addr
[1], write_mask
, &src1_param
);
2082 shader_glsl_add_src_param(ins
, ins
->src
[2], ins
->src_addr
[2], write_mask
, &src2_param
);
2083 shader_addline(ins
->buffer
, "(%s * %s) + %s);\n",
2084 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
2087 /** Handles transforming all WINED3DSIO_M?x? opcodes for
2088 Vertex shaders to GLSL codes */
2089 static void shader_glsl_mnxn(const struct wined3d_shader_instruction
*ins
)
2092 int nComponents
= 0;
2093 struct wined3d_shader_dst_param tmp_dst
= {0};
2094 struct wined3d_shader_instruction tmp_ins
;
2096 memset(&tmp_ins
, 0, sizeof(tmp_ins
));
2098 /* Set constants for the temporary argument */
2099 tmp_ins
.shader
= ins
->shader
;
2100 tmp_ins
.buffer
= ins
->buffer
;
2101 tmp_ins
.src
[0] = ins
->src
[0];
2102 tmp_ins
.src_addr
[0] = ins
->src_addr
[0];
2103 tmp_ins
.src_addr
[1] = ins
->src_addr
[1];
2104 tmp_ins
.reg_maps
= ins
->reg_maps
;
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
->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
->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
->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
->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
->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
->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
->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
->buffer
, "cos(%s));\n", src0_param
.param_str
);
2261 case WINED3DSP_WRITEMASK_1
:
2262 shader_addline(ins
->buffer
, "sin(%s));\n", src0_param
.param_str
);
2265 case (WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
):
2266 shader_addline(ins
->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
->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
->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
->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
->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
->buffer
, "for (tmpInt%u = 0, aL%u = %s.y; tmpInt%u < %s.x; tmpInt%u++, aL%u += %s.z) {\n",
2325 shader
->baseShader
.cur_loop_depth
, shader
->baseShader
.cur_loop_regno
,
2326 src1_param
.reg_name
, shader
->baseShader
.cur_loop_depth
, src1_param
.reg_name
,
2327 shader
->baseShader
.cur_loop_depth
, shader
->baseShader
.cur_loop_regno
, src1_param
.reg_name
);
2330 shader
->baseShader
.cur_loop_depth
++;
2331 shader
->baseShader
.cur_loop_regno
++;
2334 static void shader_glsl_end(const struct wined3d_shader_instruction
*ins
)
2336 IWineD3DBaseShaderImpl
*shader
= (IWineD3DBaseShaderImpl
*)ins
->shader
;
2338 shader_addline(ins
->buffer
, "}\n");
2340 if (ins
->handler_idx
== WINED3DSIH_ENDLOOP
)
2342 shader
->baseShader
.cur_loop_depth
--;
2343 shader
->baseShader
.cur_loop_regno
--;
2346 if (ins
->handler_idx
== WINED3DSIH_ENDREP
)
2348 shader
->baseShader
.cur_loop_depth
--;
2352 static void shader_glsl_rep(const struct wined3d_shader_instruction
*ins
)
2354 IWineD3DBaseShaderImpl
*shader
= (IWineD3DBaseShaderImpl
*)ins
->shader
;
2355 glsl_src_param_t src0_param
;
2357 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2358 shader_addline(ins
->buffer
, "for (tmpInt%d = 0; tmpInt%d < %s; tmpInt%d++) {\n",
2359 shader
->baseShader
.cur_loop_depth
, shader
->baseShader
.cur_loop_depth
,
2360 src0_param
.param_str
, shader
->baseShader
.cur_loop_depth
);
2361 shader
->baseShader
.cur_loop_depth
++;
2364 static void shader_glsl_if(const struct wined3d_shader_instruction
*ins
)
2366 glsl_src_param_t src0_param
;
2368 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2369 shader_addline(ins
->buffer
, "if (%s) {\n", src0_param
.param_str
);
2372 static void shader_glsl_ifc(const struct wined3d_shader_instruction
*ins
)
2374 glsl_src_param_t src0_param
;
2375 glsl_src_param_t src1_param
;
2377 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2378 shader_glsl_add_src_param(ins
, ins
->src
[1], ins
->src_addr
[1], WINED3DSP_WRITEMASK_0
, &src1_param
);
2380 shader_addline(ins
->buffer
, "if (%s %s %s) {\n",
2381 src0_param
.param_str
, shader_get_comp_op(ins
->flags
), src1_param
.param_str
);
2384 static void shader_glsl_else(const struct wined3d_shader_instruction
*ins
)
2386 shader_addline(ins
->buffer
, "} else {\n");
2389 static void shader_glsl_break(const struct wined3d_shader_instruction
*ins
)
2391 shader_addline(ins
->buffer
, "break;\n");
2394 /* FIXME: According to MSDN the compare is done per component. */
2395 static void shader_glsl_breakc(const struct wined3d_shader_instruction
*ins
)
2397 glsl_src_param_t src0_param
;
2398 glsl_src_param_t src1_param
;
2400 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2401 shader_glsl_add_src_param(ins
, ins
->src
[1], ins
->src_addr
[1], WINED3DSP_WRITEMASK_0
, &src1_param
);
2403 shader_addline(ins
->buffer
, "if (%s %s %s) break;\n",
2404 src0_param
.param_str
, shader_get_comp_op(ins
->flags
), src1_param
.param_str
);
2407 static void shader_glsl_label(const struct wined3d_shader_instruction
*ins
)
2410 DWORD snum
= (ins
->src
[0]) & WINED3DSP_REGNUM_MASK
;
2411 shader_addline(ins
->buffer
, "}\n");
2412 shader_addline(ins
->buffer
, "void subroutine%u () {\n", snum
);
2415 static void shader_glsl_call(const struct wined3d_shader_instruction
*ins
)
2417 DWORD snum
= (ins
->src
[0]) & WINED3DSP_REGNUM_MASK
;
2418 shader_addline(ins
->buffer
, "subroutine%u();\n", snum
);
2421 static void shader_glsl_callnz(const struct wined3d_shader_instruction
*ins
)
2423 glsl_src_param_t src1_param
;
2425 DWORD snum
= (ins
->src
[0]) & WINED3DSP_REGNUM_MASK
;
2426 shader_glsl_add_src_param(ins
, ins
->src
[1], ins
->src_addr
[1], WINED3DSP_WRITEMASK_0
, &src1_param
);
2427 shader_addline(ins
->buffer
, "if (%s) subroutine%u();\n", src1_param
.param_str
, snum
);
2430 /*********************************************
2431 * Pixel Shader Specific Code begins here
2432 ********************************************/
2433 static void pshader_glsl_tex(const struct wined3d_shader_instruction
*ins
)
2435 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*)ins
->shader
;
2436 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
2437 DWORD shader_version
= ins
->reg_maps
->shader_version
;
2438 glsl_sample_function_t sample_function
;
2439 DWORD sample_flags
= 0;
2442 DWORD mask
= 0, swizzle
;
2444 /* 1.0-1.4: Use destination register as sampler source.
2445 * 2.0+: Use provided sampler source. */
2446 if (shader_version
< WINED3DPS_VERSION(2,0)) sampler_idx
= ins
->dst
[0].register_idx
;
2447 else sampler_idx
= ins
->src
[1] & WINED3DSP_REGNUM_MASK
;
2448 sampler_type
= ins
->reg_maps
->samplers
[sampler_idx
] & WINED3DSP_TEXTURETYPE_MASK
;
2450 if (shader_version
< WINED3DPS_VERSION(1,4))
2452 DWORD flags
= deviceImpl
->stateBlock
->textureState
[sampler_idx
][WINED3DTSS_TEXTURETRANSFORMFLAGS
];
2454 /* Projected cube textures don't make a lot of sense, the resulting coordinates stay the same. */
2455 if (flags
& WINED3DTTFF_PROJECTED
&& sampler_type
!= WINED3DSTT_CUBE
) {
2456 sample_flags
|= WINED3D_GLSL_SAMPLE_PROJECTED
;
2457 switch (flags
& ~WINED3DTTFF_PROJECTED
) {
2458 case WINED3DTTFF_COUNT1
: FIXME("WINED3DTTFF_PROJECTED with WINED3DTTFF_COUNT1?\n"); break;
2459 case WINED3DTTFF_COUNT2
: mask
= WINED3DSP_WRITEMASK_1
; break;
2460 case WINED3DTTFF_COUNT3
: mask
= WINED3DSP_WRITEMASK_2
; break;
2461 case WINED3DTTFF_COUNT4
:
2462 case WINED3DTTFF_DISABLE
: mask
= WINED3DSP_WRITEMASK_3
; break;
2466 else if (shader_version
< WINED3DPS_VERSION(2,0))
2468 DWORD src_mod
= ins
->src
[0] & WINED3DSP_SRCMOD_MASK
;
2470 if (src_mod
== WINED3DSPSM_DZ
) {
2471 sample_flags
|= WINED3D_GLSL_SAMPLE_PROJECTED
;
2472 mask
= WINED3DSP_WRITEMASK_2
;
2473 } else if (src_mod
== WINED3DSPSM_DW
) {
2474 sample_flags
|= WINED3D_GLSL_SAMPLE_PROJECTED
;
2475 mask
= WINED3DSP_WRITEMASK_3
;
2478 if (ins
->flags
& WINED3DSI_TEXLD_PROJECT
)
2480 /* ps 2.0 texldp instruction always divides by the fourth component. */
2481 sample_flags
|= WINED3D_GLSL_SAMPLE_PROJECTED
;
2482 mask
= WINED3DSP_WRITEMASK_3
;
2486 if(deviceImpl
->stateBlock
->textures
[sampler_idx
] &&
2487 IWineD3DBaseTexture_GetTextureDimensions(deviceImpl
->stateBlock
->textures
[sampler_idx
]) == GL_TEXTURE_RECTANGLE_ARB
) {
2488 sample_flags
|= WINED3D_GLSL_SAMPLE_RECT
;
2491 shader_glsl_get_sample_function(sampler_type
, sample_flags
, &sample_function
);
2492 mask
|= sample_function
.coord_mask
;
2494 if (shader_version
< WINED3DPS_VERSION(2,0)) swizzle
= WINED3DVS_NOSWIZZLE
;
2495 else swizzle
= ins
->src
[1] & WINED3DSP_SWIZZLE_MASK
;
2497 /* 1.0-1.3: Use destination register as coordinate source.
2498 1.4+: Use provided coordinate source register. */
2499 if (shader_version
< WINED3DPS_VERSION(1,4))
2502 shader_glsl_write_mask_to_str(mask
, coord_mask
);
2503 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, swizzle
, NULL
,
2504 "T%u%s", sampler_idx
, coord_mask
);
2506 glsl_src_param_t coord_param
;
2507 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], mask
, &coord_param
);
2508 if (ins
->flags
& WINED3DSI_TEXLD_BIAS
)
2510 glsl_src_param_t bias
;
2511 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], WINED3DSP_WRITEMASK_3
, &bias
);
2512 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, swizzle
, bias
.param_str
,
2513 "%s", coord_param
.param_str
);
2515 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, swizzle
, NULL
,
2516 "%s", coord_param
.param_str
);
2521 static void shader_glsl_texldl(const struct wined3d_shader_instruction
*ins
)
2523 IWineD3DBaseShaderImpl
*This
= (IWineD3DBaseShaderImpl
*)ins
->shader
;
2524 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
2525 glsl_sample_function_t sample_function
;
2526 glsl_src_param_t coord_param
, lod_param
;
2527 DWORD sample_flags
= WINED3D_GLSL_SAMPLE_LOD
;
2530 DWORD swizzle
= ins
->src
[1] & WINED3DSP_SWIZZLE_MASK
;
2532 sampler_idx
= ins
->src
[1] & WINED3DSP_REGNUM_MASK
;
2533 sampler_type
= ins
->reg_maps
->samplers
[sampler_idx
] & WINED3DSP_TEXTURETYPE_MASK
;
2534 if(deviceImpl
->stateBlock
->textures
[sampler_idx
] &&
2535 IWineD3DBaseTexture_GetTextureDimensions(deviceImpl
->stateBlock
->textures
[sampler_idx
]) == GL_TEXTURE_RECTANGLE_ARB
) {
2536 sample_flags
|= WINED3D_GLSL_SAMPLE_RECT
;
2538 shader_glsl_get_sample_function(sampler_type
, sample_flags
, &sample_function
);
2539 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], sample_function
.coord_mask
, &coord_param
);
2541 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], WINED3DSP_WRITEMASK_3
, &lod_param
);
2543 if (shader_is_pshader_version(ins
->reg_maps
->shader_version
))
2545 /* The GLSL spec claims the Lod sampling functions are only supported in vertex shaders.
2546 * However, they seem to work just fine in fragment shaders as well. */
2547 WARN("Using %s in fragment shader.\n", sample_function
.name
);
2549 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, swizzle
, lod_param
.param_str
,
2550 "%s", coord_param
.param_str
);
2553 static void pshader_glsl_texcoord(const struct wined3d_shader_instruction
*ins
)
2555 /* FIXME: Make this work for more than just 2D textures */
2556 SHADER_BUFFER
*buffer
= ins
->buffer
;
2557 DWORD write_mask
= shader_glsl_append_dst(ins
->buffer
, ins
);
2559 if (ins
->reg_maps
->shader_version
!= WINED3DPS_VERSION(1,4))
2563 shader_glsl_get_write_mask(&ins
->dst
[0], dst_mask
);
2564 shader_addline(buffer
, "clamp(gl_TexCoord[%u], 0.0, 1.0)%s);\n",
2565 ins
->dst
[0].register_idx
, dst_mask
);
2567 DWORD reg
= ins
->src
[0] & WINED3DSP_REGNUM_MASK
;
2568 DWORD src_mod
= ins
->src
[0] & WINED3DSP_SRCMOD_MASK
;
2569 char dst_swizzle
[6];
2571 shader_glsl_get_swizzle(ins
->src
[0], FALSE
, write_mask
, dst_swizzle
);
2573 if (src_mod
== WINED3DSPSM_DZ
) {
2574 glsl_src_param_t div_param
;
2575 unsigned int mask_size
= shader_glsl_get_write_mask_size(write_mask
);
2576 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], WINED3DSP_WRITEMASK_2
, &div_param
);
2578 if (mask_size
> 1) {
2579 shader_addline(buffer
, "gl_TexCoord[%u]%s / vec%d(%s));\n", reg
, dst_swizzle
, mask_size
, div_param
.param_str
);
2581 shader_addline(buffer
, "gl_TexCoord[%u]%s / %s);\n", reg
, dst_swizzle
, div_param
.param_str
);
2583 } else if (src_mod
== WINED3DSPSM_DW
) {
2584 glsl_src_param_t div_param
;
2585 unsigned int mask_size
= shader_glsl_get_write_mask_size(write_mask
);
2586 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], WINED3DSP_WRITEMASK_3
, &div_param
);
2588 if (mask_size
> 1) {
2589 shader_addline(buffer
, "gl_TexCoord[%u]%s / vec%d(%s));\n", reg
, dst_swizzle
, mask_size
, div_param
.param_str
);
2591 shader_addline(buffer
, "gl_TexCoord[%u]%s / %s);\n", reg
, dst_swizzle
, div_param
.param_str
);
2594 shader_addline(buffer
, "gl_TexCoord[%u]%s);\n", reg
, dst_swizzle
);
2599 /** Process the WINED3DSIO_TEXDP3TEX instruction in GLSL:
2600 * Take a 3-component dot product of the TexCoord[dstreg] and src,
2601 * then perform a 1D texture lookup from stage dstregnum, place into dst. */
2602 static void pshader_glsl_texdp3tex(const struct wined3d_shader_instruction
*ins
)
2604 glsl_src_param_t src0_param
;
2605 glsl_sample_function_t sample_function
;
2606 DWORD sampler_idx
= ins
->dst
[0].register_idx
;
2607 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2608 DWORD sampler_type
= ins
->reg_maps
->samplers
[sampler_idx
] & WINED3DSP_TEXTURETYPE_MASK
;
2611 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], src_mask
, &src0_param
);
2613 /* Do I have to take care about the projected bit? I don't think so, since the dp3 returns only one
2614 * scalar, and projected sampling would require 4.
2616 * It is a dependent read - not valid with conditional NP2 textures
2618 shader_glsl_get_sample_function(sampler_type
, 0, &sample_function
);
2619 mask_size
= shader_glsl_get_write_mask_size(sample_function
.coord_mask
);
2624 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, WINED3DVS_NOSWIZZLE
, NULL
,
2625 "dot(gl_TexCoord[%u].xyz, %s)", sampler_idx
, src0_param
.param_str
);
2629 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, WINED3DVS_NOSWIZZLE
, NULL
,
2630 "vec2(dot(gl_TexCoord[%u].xyz, %s), 0.0)", sampler_idx
, src0_param
.param_str
);
2634 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, WINED3DVS_NOSWIZZLE
, NULL
,
2635 "vec3(dot(gl_TexCoord[%u].xyz, %s), 0.0, 0.0)", sampler_idx
, src0_param
.param_str
);
2639 FIXME("Unexpected mask size %u\n", mask_size
);
2644 /** Process the WINED3DSIO_TEXDP3 instruction in GLSL:
2645 * Take a 3-component dot product of the TexCoord[dstreg] and src. */
2646 static void pshader_glsl_texdp3(const struct wined3d_shader_instruction
*ins
)
2648 glsl_src_param_t src0_param
;
2649 DWORD dstreg
= ins
->dst
[0].register_idx
;
2650 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2652 unsigned int mask_size
;
2654 dst_mask
= shader_glsl_append_dst(ins
->buffer
, ins
);
2655 mask_size
= shader_glsl_get_write_mask_size(dst_mask
);
2656 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], src_mask
, &src0_param
);
2658 if (mask_size
> 1) {
2659 shader_addline(ins
->buffer
, "vec%d(dot(T%u.xyz, %s)));\n", mask_size
, dstreg
, src0_param
.param_str
);
2661 shader_addline(ins
->buffer
, "dot(T%u.xyz, %s));\n", dstreg
, src0_param
.param_str
);
2665 /** Process the WINED3DSIO_TEXDEPTH instruction in GLSL:
2666 * Calculate the depth as dst.x / dst.y */
2667 static void pshader_glsl_texdepth(const struct wined3d_shader_instruction
*ins
)
2669 glsl_dst_param_t dst_param
;
2671 shader_glsl_add_dst_param(ins
, &ins
->dst
[0], &dst_param
);
2673 /* Tests show that texdepth never returns anything below 0.0, and that r5.y is clamped to 1.0.
2674 * Negative input is accepted, -0.25 / -0.5 returns 0.5. GL should clamp gl_FragDepth to [0;1], but
2675 * this doesn't always work, so clamp the results manually. Whether or not the x value is clamped at 1
2676 * too is irrelevant, since if x = 0, any y value < 1.0 (and > 1.0 is not allowed) results in a result
2679 shader_addline(ins
->buffer
, "gl_FragDepth = clamp((%s.x / min(%s.y, 1.0)), 0.0, 1.0);\n",
2680 dst_param
.reg_name
, dst_param
.reg_name
);
2683 /** Process the WINED3DSIO_TEXM3X2DEPTH instruction in GLSL:
2684 * Last row of a 3x2 matrix multiply, use the result to calculate the depth:
2685 * Calculate tmp0.y = TexCoord[dstreg] . src.xyz; (tmp0.x has already been calculated)
2686 * depth = (tmp0.y == 0.0) ? 1.0 : tmp0.x / tmp0.y
2688 static void pshader_glsl_texm3x2depth(const struct wined3d_shader_instruction
*ins
)
2690 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2691 DWORD dstreg
= ins
->dst
[0].register_idx
;
2692 glsl_src_param_t src0_param
;
2694 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], src_mask
, &src0_param
);
2696 shader_addline(ins
->buffer
, "tmp0.y = dot(T%u.xyz, %s);\n", dstreg
, src0_param
.param_str
);
2697 shader_addline(ins
->buffer
, "gl_FragDepth = (tmp0.y == 0.0) ? 1.0 : clamp(tmp0.x / tmp0.y, 0.0, 1.0);\n");
2700 /** Process the WINED3DSIO_TEXM3X2PAD instruction in GLSL
2701 * Calculate the 1st of a 2-row matrix multiplication. */
2702 static void pshader_glsl_texm3x2pad(const struct wined3d_shader_instruction
*ins
)
2704 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2705 DWORD reg
= ins
->dst
[0].register_idx
;
2706 SHADER_BUFFER
*buffer
= ins
->buffer
;
2707 glsl_src_param_t src0_param
;
2709 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], src_mask
, &src0_param
);
2710 shader_addline(buffer
, "tmp0.x = dot(T%u.xyz, %s);\n", reg
, src0_param
.param_str
);
2713 /** Process the WINED3DSIO_TEXM3X3PAD instruction in GLSL
2714 * Calculate the 1st or 2nd row of a 3-row matrix multiplication. */
2715 static void pshader_glsl_texm3x3pad(const struct wined3d_shader_instruction
*ins
)
2717 IWineD3DPixelShaderImpl
*shader
= (IWineD3DPixelShaderImpl
*)ins
->shader
;
2718 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2719 DWORD reg
= ins
->dst
[0].register_idx
;
2720 SHADER_BUFFER
*buffer
= ins
->buffer
;
2721 SHADER_PARSE_STATE
* current_state
= &shader
->baseShader
.parse_state
;
2722 glsl_src_param_t src0_param
;
2724 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], src_mask
, &src0_param
);
2725 shader_addline(buffer
, "tmp0.%c = dot(T%u.xyz, %s);\n", 'x' + current_state
->current_row
, reg
, src0_param
.param_str
);
2726 current_state
->texcoord_w
[current_state
->current_row
++] = reg
;
2729 static void pshader_glsl_texm3x2tex(const struct wined3d_shader_instruction
*ins
)
2731 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2732 DWORD reg
= ins
->dst
[0].register_idx
;
2733 SHADER_BUFFER
*buffer
= ins
->buffer
;
2734 glsl_src_param_t src0_param
;
2735 DWORD sampler_type
= ins
->reg_maps
->samplers
[reg
] & WINED3DSP_TEXTURETYPE_MASK
;
2736 glsl_sample_function_t sample_function
;
2738 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], src_mask
, &src0_param
);
2739 shader_addline(buffer
, "tmp0.y = dot(T%u.xyz, %s);\n", reg
, src0_param
.param_str
);
2741 shader_glsl_get_sample_function(sampler_type
, 0, &sample_function
);
2743 /* Sample the texture using the calculated coordinates */
2744 shader_glsl_gen_sample_code(ins
, reg
, &sample_function
, WINED3DVS_NOSWIZZLE
, NULL
, "tmp0.xy");
2747 /** Process the WINED3DSIO_TEXM3X3TEX instruction in GLSL
2748 * Perform the 3rd row of a 3x3 matrix multiply, then sample the texture using the calculated coordinates */
2749 static void pshader_glsl_texm3x3tex(const struct wined3d_shader_instruction
*ins
)
2751 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2752 glsl_src_param_t src0_param
;
2753 DWORD reg
= ins
->dst
[0].register_idx
;
2754 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*)ins
->shader
;
2755 SHADER_PARSE_STATE
* current_state
= &This
->baseShader
.parse_state
;
2756 DWORD sampler_type
= ins
->reg_maps
->samplers
[reg
] & WINED3DSP_TEXTURETYPE_MASK
;
2757 glsl_sample_function_t sample_function
;
2759 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], src_mask
, &src0_param
);
2760 shader_addline(ins
->buffer
, "tmp0.z = dot(T%u.xyz, %s);\n", reg
, src0_param
.param_str
);
2762 /* Dependent read, not valid with conditional NP2 */
2763 shader_glsl_get_sample_function(sampler_type
, 0, &sample_function
);
2765 /* Sample the texture using the calculated coordinates */
2766 shader_glsl_gen_sample_code(ins
, reg
, &sample_function
, WINED3DVS_NOSWIZZLE
, NULL
, "tmp0.xyz");
2768 current_state
->current_row
= 0;
2771 /** Process the WINED3DSIO_TEXM3X3 instruction in GLSL
2772 * Perform the 3rd row of a 3x3 matrix multiply */
2773 static void pshader_glsl_texm3x3(const struct wined3d_shader_instruction
*ins
)
2775 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2776 glsl_src_param_t src0_param
;
2778 DWORD reg
= ins
->dst
[0].register_idx
;
2779 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*)ins
->shader
;
2780 SHADER_PARSE_STATE
* current_state
= &This
->baseShader
.parse_state
;
2782 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], src_mask
, &src0_param
);
2784 shader_glsl_append_dst(ins
->buffer
, ins
);
2785 shader_glsl_get_write_mask(&ins
->dst
[0], dst_mask
);
2786 shader_addline(ins
->buffer
, "vec4(tmp0.xy, dot(T%u.xyz, %s), 1.0)%s);\n", reg
, src0_param
.param_str
, dst_mask
);
2788 current_state
->current_row
= 0;
2791 /** Process the WINED3DSIO_TEXM3X3SPEC instruction in GLSL
2792 * Perform the final texture lookup based on the previous 2 3x3 matrix multiplies */
2793 static void pshader_glsl_texm3x3spec(const struct wined3d_shader_instruction
*ins
)
2795 IWineD3DPixelShaderImpl
*shader
= (IWineD3DPixelShaderImpl
*)ins
->shader
;
2796 DWORD reg
= ins
->dst
[0].register_idx
;
2797 glsl_src_param_t src0_param
;
2798 glsl_src_param_t src1_param
;
2799 SHADER_BUFFER
*buffer
= ins
->buffer
;
2800 SHADER_PARSE_STATE
* current_state
= &shader
->baseShader
.parse_state
;
2801 DWORD stype
= ins
->reg_maps
->samplers
[reg
] & WINED3DSP_TEXTURETYPE_MASK
;
2802 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2803 glsl_sample_function_t sample_function
;
2805 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], src_mask
, &src0_param
);
2806 shader_glsl_add_src_param(ins
, ins
->src
[1], ins
->src_addr
[1], src_mask
, &src1_param
);
2808 /* Perform the last matrix multiply operation */
2809 shader_addline(buffer
, "tmp0.z = dot(T%u.xyz, %s);\n", reg
, src0_param
.param_str
);
2810 /* Reflection calculation */
2811 shader_addline(buffer
, "tmp0.xyz = -reflect((%s), normalize(tmp0.xyz));\n", src1_param
.param_str
);
2813 /* Dependent read, not valid with conditional NP2 */
2814 shader_glsl_get_sample_function(stype
, 0, &sample_function
);
2816 /* Sample the texture */
2817 shader_glsl_gen_sample_code(ins
, reg
, &sample_function
, WINED3DVS_NOSWIZZLE
, NULL
, "tmp0.xyz");
2819 current_state
->current_row
= 0;
2822 /** Process the WINED3DSIO_TEXM3X3VSPEC instruction in GLSL
2823 * Perform the final texture lookup based on the previous 2 3x3 matrix multiplies */
2824 static void pshader_glsl_texm3x3vspec(const struct wined3d_shader_instruction
*ins
)
2826 IWineD3DPixelShaderImpl
*shader
= (IWineD3DPixelShaderImpl
*)ins
->shader
;
2827 DWORD reg
= ins
->dst
[0].register_idx
;
2828 SHADER_BUFFER
*buffer
= ins
->buffer
;
2829 SHADER_PARSE_STATE
* current_state
= &shader
->baseShader
.parse_state
;
2830 glsl_src_param_t src0_param
;
2831 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2832 DWORD sampler_type
= ins
->reg_maps
->samplers
[reg
] & WINED3DSP_TEXTURETYPE_MASK
;
2833 glsl_sample_function_t sample_function
;
2835 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], src_mask
, &src0_param
);
2837 /* Perform the last matrix multiply operation */
2838 shader_addline(buffer
, "tmp0.z = dot(vec3(T%u), vec3(%s));\n", reg
, src0_param
.param_str
);
2840 /* Construct the eye-ray vector from w coordinates */
2841 shader_addline(buffer
, "tmp1.xyz = normalize(vec3(gl_TexCoord[%u].w, gl_TexCoord[%u].w, gl_TexCoord[%u].w));\n",
2842 current_state
->texcoord_w
[0], current_state
->texcoord_w
[1], reg
);
2843 shader_addline(buffer
, "tmp0.xyz = -reflect(tmp1.xyz, normalize(tmp0.xyz));\n");
2845 /* Dependent read, not valid with conditional NP2 */
2846 shader_glsl_get_sample_function(sampler_type
, 0, &sample_function
);
2848 /* Sample the texture using the calculated coordinates */
2849 shader_glsl_gen_sample_code(ins
, reg
, &sample_function
, WINED3DVS_NOSWIZZLE
, NULL
, "tmp0.xyz");
2851 current_state
->current_row
= 0;
2854 /** Process the WINED3DSIO_TEXBEM instruction in GLSL.
2855 * Apply a fake bump map transform.
2856 * texbem is pshader <= 1.3 only, this saves a few version checks
2858 static void pshader_glsl_texbem(const struct wined3d_shader_instruction
*ins
)
2860 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*)ins
->shader
;
2861 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
2862 glsl_sample_function_t sample_function
;
2863 glsl_src_param_t coord_param
;
2870 sampler_idx
= ins
->dst
[0].register_idx
;
2871 flags
= deviceImpl
->stateBlock
->textureState
[sampler_idx
][WINED3DTSS_TEXTURETRANSFORMFLAGS
];
2873 sampler_type
= ins
->reg_maps
->samplers
[sampler_idx
] & WINED3DSP_TEXTURETYPE_MASK
;
2874 /* Dependent read, not valid with conditional NP2 */
2875 shader_glsl_get_sample_function(sampler_type
, 0, &sample_function
);
2876 mask
= sample_function
.coord_mask
;
2878 shader_glsl_write_mask_to_str(mask
, coord_mask
);
2880 /* with projective textures, texbem only divides the static texture coord, not the displacement,
2881 * so we can't let the GL handle this.
2883 if (flags
& WINED3DTTFF_PROJECTED
) {
2885 char coord_div_mask
[3];
2886 switch (flags
& ~WINED3DTTFF_PROJECTED
) {
2887 case WINED3DTTFF_COUNT1
: FIXME("WINED3DTTFF_PROJECTED with WINED3DTTFF_COUNT1?\n"); break;
2888 case WINED3DTTFF_COUNT2
: div_mask
= WINED3DSP_WRITEMASK_1
; break;
2889 case WINED3DTTFF_COUNT3
: div_mask
= WINED3DSP_WRITEMASK_2
; break;
2890 case WINED3DTTFF_COUNT4
:
2891 case WINED3DTTFF_DISABLE
: div_mask
= WINED3DSP_WRITEMASK_3
; break;
2893 shader_glsl_write_mask_to_str(div_mask
, coord_div_mask
);
2894 shader_addline(ins
->buffer
, "T%u%s /= T%u%s;\n", sampler_idx
, coord_mask
, sampler_idx
, coord_div_mask
);
2897 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0],
2898 WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
, &coord_param
);
2900 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, WINED3DVS_NOSWIZZLE
, NULL
,
2901 "T%u%s + vec4(bumpenvmat%d * %s, 0.0, 0.0)%s", sampler_idx
, coord_mask
, sampler_idx
,
2902 coord_param
.param_str
, coord_mask
);
2904 if (ins
->handler_idx
== WINED3DSIH_TEXBEML
)
2906 glsl_src_param_t luminance_param
;
2907 glsl_dst_param_t dst_param
;
2909 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], WINED3DSP_WRITEMASK_2
, &luminance_param
);
2910 shader_glsl_add_dst_param(ins
, &ins
->dst
[0], &dst_param
);
2912 shader_addline(ins
->buffer
, "%s%s *= (%s * luminancescale%d + luminanceoffset%d);\n",
2913 dst_param
.reg_name
, dst_param
.mask_str
,
2914 luminance_param
.param_str
, sampler_idx
, sampler_idx
);
2918 static void pshader_glsl_bem(const struct wined3d_shader_instruction
*ins
)
2920 glsl_src_param_t src0_param
, src1_param
;
2921 DWORD sampler_idx
= ins
->dst
[0].register_idx
;
2923 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0],
2924 WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
, &src0_param
);
2925 shader_glsl_add_src_param(ins
, ins
->src
[1], ins
->src_addr
[1],
2926 WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
, &src1_param
);
2928 shader_glsl_append_dst(ins
->buffer
, ins
);
2929 shader_addline(ins
->buffer
, "%s + bumpenvmat%d * %s);\n",
2930 src0_param
.param_str
, sampler_idx
, src1_param
.param_str
);
2933 /** Process the WINED3DSIO_TEXREG2AR instruction in GLSL
2934 * Sample 2D texture at dst using the alpha & red (wx) components of src as texture coordinates */
2935 static void pshader_glsl_texreg2ar(const struct wined3d_shader_instruction
*ins
)
2937 glsl_src_param_t src0_param
;
2938 DWORD sampler_idx
= ins
->dst
[0].register_idx
;
2939 DWORD sampler_type
= ins
->reg_maps
->samplers
[sampler_idx
] & WINED3DSP_TEXTURETYPE_MASK
;
2940 glsl_sample_function_t sample_function
;
2942 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], WINED3DSP_WRITEMASK_ALL
, &src0_param
);
2944 shader_glsl_get_sample_function(sampler_type
, 0, &sample_function
);
2945 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, WINED3DVS_NOSWIZZLE
, NULL
,
2946 "%s.wx", src0_param
.reg_name
);
2949 /** Process the WINED3DSIO_TEXREG2GB instruction in GLSL
2950 * Sample 2D texture at dst using the green & blue (yz) components of src as texture coordinates */
2951 static void pshader_glsl_texreg2gb(const struct wined3d_shader_instruction
*ins
)
2953 glsl_src_param_t src0_param
;
2954 DWORD sampler_idx
= ins
->dst
[0].register_idx
;
2955 DWORD sampler_type
= ins
->reg_maps
->samplers
[sampler_idx
] & WINED3DSP_TEXTURETYPE_MASK
;
2956 glsl_sample_function_t sample_function
;
2958 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], WINED3DSP_WRITEMASK_ALL
, &src0_param
);
2960 shader_glsl_get_sample_function(sampler_type
, 0, &sample_function
);
2961 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, WINED3DVS_NOSWIZZLE
, NULL
,
2962 "%s.yz", src0_param
.reg_name
);
2965 /** Process the WINED3DSIO_TEXREG2RGB instruction in GLSL
2966 * Sample texture at dst using the rgb (xyz) components of src as texture coordinates */
2967 static void pshader_glsl_texreg2rgb(const struct wined3d_shader_instruction
*ins
)
2969 glsl_src_param_t src0_param
;
2970 DWORD sampler_idx
= ins
->dst
[0].register_idx
;
2971 DWORD sampler_type
= ins
->reg_maps
->samplers
[sampler_idx
] & WINED3DSP_TEXTURETYPE_MASK
;
2972 glsl_sample_function_t sample_function
;
2974 /* Dependent read, not valid with conditional NP2 */
2975 shader_glsl_get_sample_function(sampler_type
, 0, &sample_function
);
2976 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0], sample_function
.coord_mask
, &src0_param
);
2978 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, WINED3DVS_NOSWIZZLE
, NULL
,
2979 "%s", src0_param
.param_str
);
2982 /** Process the WINED3DSIO_TEXKILL instruction in GLSL.
2983 * If any of the first 3 components are < 0, discard this pixel */
2984 static void pshader_glsl_texkill(const struct wined3d_shader_instruction
*ins
)
2986 glsl_dst_param_t dst_param
;
2988 /* The argument is a destination parameter, and no writemasks are allowed */
2989 shader_glsl_add_dst_param(ins
, &ins
->dst
[0], &dst_param
);
2990 if ((ins
->reg_maps
->shader_version
>= WINED3DPS_VERSION(2,0)))
2992 /* 2.0 shaders compare all 4 components in texkill */
2993 shader_addline(ins
->buffer
, "if (any(lessThan(%s.xyzw, vec4(0.0)))) discard;\n", dst_param
.reg_name
);
2995 /* 1.X shaders only compare the first 3 components, probably due to the nature of the texkill
2996 * instruction as a tex* instruction, and phase, which kills all a / w components. Even if all
2997 * 4 components are defined, only the first 3 are used
2999 shader_addline(ins
->buffer
, "if (any(lessThan(%s.xyz, vec3(0.0)))) discard;\n", dst_param
.reg_name
);
3003 /** Process the WINED3DSIO_DP2ADD instruction in GLSL.
3004 * dst = dot2(src0, src1) + src2 */
3005 static void pshader_glsl_dp2add(const struct wined3d_shader_instruction
*ins
)
3007 glsl_src_param_t src0_param
;
3008 glsl_src_param_t src1_param
;
3009 glsl_src_param_t src2_param
;
3011 unsigned int mask_size
;
3013 write_mask
= shader_glsl_append_dst(ins
->buffer
, ins
);
3014 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
3016 shader_glsl_add_src_param(ins
, ins
->src
[0], ins
->src_addr
[0],
3017 WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
, &src0_param
);
3018 shader_glsl_add_src_param(ins
, ins
->src
[1], ins
->src_addr
[1],
3019 WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
, &src1_param
);
3020 shader_glsl_add_src_param(ins
, ins
->src
[2], ins
->src_addr
[2],
3021 WINED3DSP_WRITEMASK_0
, &src2_param
);
3023 if (mask_size
> 1) {
3024 shader_addline(ins
->buffer
, "vec%d(dot(%s, %s) + %s));\n",
3025 mask_size
, src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
3027 shader_addline(ins
->buffer
, "dot(%s, %s) + %s);\n",
3028 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
3032 static void pshader_glsl_input_pack(IWineD3DPixelShader
*iface
, SHADER_BUFFER
*buffer
,
3033 const struct wined3d_shader_semantic
*semantics_in
, const struct shader_reg_maps
*reg_maps
,
3034 enum vertexprocessing_mode vertexprocessing
)
3037 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*)iface
;
3039 for (i
= 0; i
< MAX_REG_INPUT
; ++i
)
3041 DWORD usage
, usage_idx
;
3045 if (!reg_maps
->packed_input
[i
]) continue;
3047 usage
= semantics_in
[i
].usage
;
3048 usage_idx
= semantics_in
[i
].usage_idx
;
3049 shader_glsl_get_write_mask(&semantics_in
[i
].reg
, reg_mask
);
3053 case WINED3DDECLUSAGE_TEXCOORD
:
3054 if (usage_idx
< 8 && vertexprocessing
== pretransformed
)
3055 shader_addline(buffer
, "IN[%u]%s = gl_TexCoord[%u]%s;\n",
3056 This
->input_reg_map
[i
], reg_mask
, usage_idx
, reg_mask
);
3058 shader_addline(buffer
, "IN[%u]%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
3059 This
->input_reg_map
[i
], reg_mask
, reg_mask
);
3062 case WINED3DDECLUSAGE_COLOR
:
3064 shader_addline(buffer
, "IN[%u]%s = vec4(gl_Color)%s;\n",
3065 This
->input_reg_map
[i
], reg_mask
, reg_mask
);
3066 else if (usage_idx
== 1)
3067 shader_addline(buffer
, "IN[%u]%s = vec4(gl_SecondaryColor)%s;\n",
3068 This
->input_reg_map
[i
], reg_mask
, reg_mask
);
3070 shader_addline(buffer
, "IN[%u]%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
3071 This
->input_reg_map
[i
], reg_mask
, reg_mask
);
3075 shader_addline(buffer
, "IN[%u]%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
3076 This
->input_reg_map
[i
], reg_mask
, reg_mask
);
3082 /*********************************************
3083 * Vertex Shader Specific Code begins here
3084 ********************************************/
3086 static void add_glsl_program_entry(struct shader_glsl_priv
*priv
, struct glsl_shader_prog_link
*entry
) {
3087 glsl_program_key_t
*key
;
3089 key
= HeapAlloc(GetProcessHeap(), 0, sizeof(glsl_program_key_t
));
3090 key
->vshader
= entry
->vshader
;
3091 key
->pshader
= entry
->pshader
;
3092 key
->vs_args
= entry
->vs_args
;
3093 key
->ps_args
= entry
->ps_args
;
3095 hash_table_put(priv
->glsl_program_lookup
, key
, entry
);
3098 static struct glsl_shader_prog_link
*get_glsl_program_entry(struct shader_glsl_priv
*priv
,
3099 IWineD3DVertexShader
*vshader
, IWineD3DPixelShader
*pshader
, struct vs_compile_args
*vs_args
,
3100 struct ps_compile_args
*ps_args
) {
3101 glsl_program_key_t key
;
3103 key
.vshader
= vshader
;
3104 key
.pshader
= pshader
;
3105 key
.vs_args
= *vs_args
;
3106 key
.ps_args
= *ps_args
;
3108 return hash_table_get(priv
->glsl_program_lookup
, &key
);
3111 static void delete_glsl_program_entry(struct shader_glsl_priv
*priv
, const WineD3D_GL_Info
*gl_info
,
3112 struct glsl_shader_prog_link
*entry
)
3114 glsl_program_key_t
*key
;
3116 key
= HeapAlloc(GetProcessHeap(), 0, sizeof(glsl_program_key_t
));
3117 key
->vshader
= entry
->vshader
;
3118 key
->pshader
= entry
->pshader
;
3119 key
->vs_args
= entry
->vs_args
;
3120 key
->ps_args
= entry
->ps_args
;
3121 hash_table_remove(priv
->glsl_program_lookup
, key
);
3123 GL_EXTCALL(glDeleteObjectARB(entry
->programId
));
3124 if (entry
->vshader
) list_remove(&entry
->vshader_entry
);
3125 if (entry
->pshader
) list_remove(&entry
->pshader_entry
);
3126 HeapFree(GetProcessHeap(), 0, entry
->vuniformF_locations
);
3127 HeapFree(GetProcessHeap(), 0, entry
->puniformF_locations
);
3128 HeapFree(GetProcessHeap(), 0, entry
);
3131 static void handle_ps3_input(SHADER_BUFFER
*buffer
, const WineD3D_GL_Info
*gl_info
, const DWORD
*map
,
3132 const struct wined3d_shader_semantic
*semantics_in
, const struct shader_reg_maps
*reg_maps_in
,
3133 const struct wined3d_shader_semantic
*semantics_out
, const struct shader_reg_maps
*reg_maps_out
)
3136 DWORD usage
, usage_idx
, usage_out
, usage_idx_out
;
3139 DWORD in_count
= GL_LIMITS(glsl_varyings
) / 4;
3140 char reg_mask
[6], reg_mask_out
[6];
3141 char destination
[50];
3143 set
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*set
) * (in_count
+ 2));
3145 if (!semantics_out
) {
3146 /* Save gl_FrontColor & gl_FrontSecondaryColor before overwriting them. */
3147 shader_addline(buffer
, "vec4 front_color = gl_FrontColor;\n");
3148 shader_addline(buffer
, "vec4 front_secondary_color = gl_FrontSecondaryColor;\n");
3151 for(i
= 0; i
< MAX_REG_INPUT
; i
++) {
3152 if (!reg_maps_in
->packed_input
[i
]) continue;
3155 if (in_idx
>= (in_count
+ 2)) {
3156 FIXME("More input varyings declared than supported, expect issues\n");
3159 else if (map
[i
] == ~0U)
3161 /* Declared, but not read register */
3165 if (in_idx
== in_count
) {
3166 sprintf(destination
, "gl_FrontColor");
3167 } else if (in_idx
== in_count
+ 1) {
3168 sprintf(destination
, "gl_FrontSecondaryColor");
3170 sprintf(destination
, "IN[%u]", in_idx
);
3173 usage
= semantics_in
[i
].usage
;
3174 usage_idx
= semantics_in
[i
].usage_idx
;
3175 set
[map
[i
]] = shader_glsl_get_write_mask(&semantics_in
[i
].reg
, reg_mask
);
3177 if(!semantics_out
) {
3179 case WINED3DDECLUSAGE_COLOR
:
3181 shader_addline(buffer
, "%s%s = front_color%s;\n",
3182 destination
, reg_mask
, reg_mask
);
3183 else if (usage_idx
== 1)
3184 shader_addline(buffer
, "%s%s = front_secondary_color%s;\n",
3185 destination
, reg_mask
, reg_mask
);
3187 shader_addline(buffer
, "%s%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
3188 destination
, reg_mask
, reg_mask
);
3191 case WINED3DDECLUSAGE_TEXCOORD
:
3192 if (usage_idx
< 8) {
3193 shader_addline(buffer
, "%s%s = gl_TexCoord[%u]%s;\n",
3194 destination
, reg_mask
, usage_idx
, reg_mask
);
3196 shader_addline(buffer
, "%s%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
3197 destination
, reg_mask
, reg_mask
);
3201 case WINED3DDECLUSAGE_FOG
:
3202 shader_addline(buffer
, "%s%s = vec4(gl_FogFragCoord, 0.0, 0.0, 0.0)%s;\n",
3203 destination
, reg_mask
, reg_mask
);
3207 shader_addline(buffer
, "%s%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
3208 destination
, reg_mask
, reg_mask
);
3212 for(j
= 0; j
< MAX_REG_OUTPUT
; j
++) {
3213 if (!reg_maps_out
->packed_output
[j
]) continue;
3215 usage_out
= semantics_out
[j
].usage
;
3216 usage_idx_out
= semantics_out
[j
].usage_idx
;
3217 shader_glsl_get_write_mask(&semantics_out
[j
].reg
, reg_mask_out
);
3219 if(usage
== usage_out
&&
3220 usage_idx
== usage_idx_out
) {
3221 shader_addline(buffer
, "%s%s = OUT[%u]%s;\n",
3222 destination
, reg_mask
, j
, reg_mask
);
3227 shader_addline(buffer
, "%s%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
3228 destination
, reg_mask
, reg_mask
);
3233 /* This is solely to make the compiler / linker happy and avoid warning about undefined
3234 * varyings. It shouldn't result in any real code executed on the GPU, since all read
3235 * input varyings are assigned above, if the optimizer works properly.
3237 for(i
= 0; i
< in_count
+ 2; i
++) {
3238 if(set
[i
] != WINED3DSP_WRITEMASK_ALL
) {
3239 unsigned int size
= 0;
3240 memset(reg_mask
, 0, sizeof(reg_mask
));
3241 if(!(set
[i
] & WINED3DSP_WRITEMASK_0
)) {
3242 reg_mask
[size
] = 'x';
3245 if(!(set
[i
] & WINED3DSP_WRITEMASK_1
)) {
3246 reg_mask
[size
] = 'y';
3249 if(!(set
[i
] & WINED3DSP_WRITEMASK_2
)) {
3250 reg_mask
[size
] = 'z';
3253 if(!(set
[i
] & WINED3DSP_WRITEMASK_3
)) {
3254 reg_mask
[size
] = 'w';
3258 if (i
== in_count
) {
3259 sprintf(destination
, "gl_FrontColor");
3260 } else if (i
== in_count
+ 1) {
3261 sprintf(destination
, "gl_FrontSecondaryColor");
3263 sprintf(destination
, "IN[%u]", i
);
3267 shader_addline(buffer
, "%s.%s = 0.0;\n", destination
, reg_mask
);
3269 shader_addline(buffer
, "%s.%s = vec%u(0.0);\n", destination
, reg_mask
, size
);
3274 HeapFree(GetProcessHeap(), 0, set
);
3277 static GLhandleARB
generate_param_reorder_function(IWineD3DVertexShader
*vertexshader
,
3278 IWineD3DPixelShader
*pixelshader
, const WineD3D_GL_Info
*gl_info
)
3280 GLhandleARB ret
= 0;
3281 IWineD3DVertexShaderImpl
*vs
= (IWineD3DVertexShaderImpl
*) vertexshader
;
3282 IWineD3DPixelShaderImpl
*ps
= (IWineD3DPixelShaderImpl
*) pixelshader
;
3283 IWineD3DDeviceImpl
*device
;
3284 DWORD vs_major
= WINED3DSHADER_VERSION_MAJOR(vs
->baseShader
.reg_maps
.shader_version
);
3285 DWORD ps_major
= ps
? WINED3DSHADER_VERSION_MAJOR(ps
->baseShader
.reg_maps
.shader_version
) : 0;
3287 SHADER_BUFFER buffer
;
3288 DWORD usage
, usage_idx
, writemask
;
3290 const struct wined3d_shader_semantic
*semantics_out
;
3292 shader_buffer_init(&buffer
);
3294 shader_addline(&buffer
, "#version 120\n");
3296 if(vs_major
< 3 && ps_major
< 3) {
3297 /* That one is easy: The vertex shader writes to the builtin varyings, the pixel shader reads from them.
3298 * Take care about the texcoord .w fixup though if we're using the fixed function fragment pipeline
3300 device
= (IWineD3DDeviceImpl
*) vs
->baseShader
.device
;
3301 if((GLINFO_LOCATION
).set_texcoord_w
&& ps_major
== 0 && vs_major
> 0 &&
3302 !device
->frag_pipe
->ffp_proj_control
) {
3303 shader_addline(&buffer
, "void order_ps_input() {\n");
3304 for(i
= 0; i
< min(8, MAX_REG_TEXCRD
); i
++) {
3305 if(vs
->baseShader
.reg_maps
.texcoord_mask
[i
] != 0 &&
3306 vs
->baseShader
.reg_maps
.texcoord_mask
[i
] != WINED3DSP_WRITEMASK_ALL
) {
3307 shader_addline(&buffer
, "gl_TexCoord[%u].w = 1.0;\n", i
);
3310 shader_addline(&buffer
, "}\n");
3312 shader_addline(&buffer
, "void order_ps_input() { /* do nothing */ }\n");
3314 } else if(ps_major
< 3 && vs_major
>= 3) {
3315 /* The vertex shader writes to its own varyings, the pixel shader needs them in the builtin ones */
3316 semantics_out
= vs
->semantics_out
;
3318 shader_addline(&buffer
, "void order_ps_input(in vec4 OUT[%u]) {\n", MAX_REG_OUTPUT
);
3319 for(i
= 0; i
< MAX_REG_OUTPUT
; i
++) {
3320 if (!vs
->baseShader
.reg_maps
.packed_output
[i
]) continue;
3322 usage
= semantics_out
[i
].usage
;
3323 usage_idx
= semantics_out
[i
].usage_idx
;
3324 writemask
= shader_glsl_get_write_mask(&semantics_out
[i
].reg
, reg_mask
);
3327 case WINED3DDECLUSAGE_COLOR
:
3329 shader_addline(&buffer
, "gl_FrontColor%s = OUT[%u]%s;\n", reg_mask
, i
, reg_mask
);
3330 else if (usage_idx
== 1)
3331 shader_addline(&buffer
, "gl_FrontSecondaryColor%s = OUT[%u]%s;\n", reg_mask
, i
, reg_mask
);
3334 case WINED3DDECLUSAGE_POSITION
:
3335 shader_addline(&buffer
, "gl_Position%s = OUT[%u]%s;\n", reg_mask
, i
, reg_mask
);
3338 case WINED3DDECLUSAGE_TEXCOORD
:
3339 if (usage_idx
< 8) {
3340 if(!(GLINFO_LOCATION
).set_texcoord_w
|| ps_major
> 0) writemask
|= WINED3DSP_WRITEMASK_3
;
3342 shader_addline(&buffer
, "gl_TexCoord[%u]%s = OUT[%u]%s;\n",
3343 usage_idx
, reg_mask
, i
, reg_mask
);
3344 if(!(writemask
& WINED3DSP_WRITEMASK_3
)) {
3345 shader_addline(&buffer
, "gl_TexCoord[%u].w = 1.0;\n", usage_idx
);
3350 case WINED3DDECLUSAGE_PSIZE
:
3351 shader_addline(&buffer
, "gl_PointSize = OUT[%u].x;\n", i
);
3354 case WINED3DDECLUSAGE_FOG
:
3355 shader_addline(&buffer
, "gl_FogFragCoord = OUT[%u].%c;\n", i
, reg_mask
[1]);
3362 shader_addline(&buffer
, "}\n");
3364 } else if(ps_major
>= 3 && vs_major
>= 3) {
3365 semantics_out
= vs
->semantics_out
;
3367 /* This one is tricky: a 3.0 pixel shader reads from a 3.0 vertex shader */
3368 shader_addline(&buffer
, "varying vec4 IN[%u];\n", GL_LIMITS(glsl_varyings
) / 4);
3369 shader_addline(&buffer
, "void order_ps_input(in vec4 OUT[%u]) {\n", MAX_REG_OUTPUT
);
3371 /* First, sort out position and point size. Those are not passed to the pixel shader */
3372 for(i
= 0; i
< MAX_REG_OUTPUT
; i
++) {
3373 if (!vs
->baseShader
.reg_maps
.packed_output
[i
]) continue;
3375 usage
= semantics_out
[i
].usage
;
3376 usage_idx
= semantics_out
[i
].usage_idx
;
3377 shader_glsl_get_write_mask(&semantics_out
[i
].reg
, reg_mask
);
3380 case WINED3DDECLUSAGE_POSITION
:
3381 shader_addline(&buffer
, "gl_Position%s = OUT[%u]%s;\n", reg_mask
, i
, reg_mask
);
3384 case WINED3DDECLUSAGE_PSIZE
:
3385 shader_addline(&buffer
, "gl_PointSize = OUT[%u].x;\n", i
);
3393 /* Then, fix the pixel shader input */
3394 handle_ps3_input(&buffer
, gl_info
, ps
->input_reg_map
,
3395 ps
->semantics_in
, &ps
->baseShader
.reg_maps
, semantics_out
, &vs
->baseShader
.reg_maps
);
3397 shader_addline(&buffer
, "}\n");
3398 } else if(ps_major
>= 3 && vs_major
< 3) {
3399 shader_addline(&buffer
, "varying vec4 IN[%u];\n", GL_LIMITS(glsl_varyings
) / 4);
3400 shader_addline(&buffer
, "void order_ps_input() {\n");
3401 /* The vertex shader wrote to the builtin varyings. There is no need to figure out position and
3402 * point size, but we depend on the optimizers kindness to find out that the pixel shader doesn't
3403 * read gl_TexCoord and gl_ColorX, otherwise we'll run out of varyings
3405 handle_ps3_input(&buffer
, gl_info
, ps
->input_reg_map
, ps
->semantics_in
, &ps
->baseShader
.reg_maps
, NULL
, NULL
);
3406 shader_addline(&buffer
, "}\n");
3408 ERR("Unexpected vertex and pixel shader version condition: vs: %d, ps: %d\n", vs_major
, ps_major
);
3411 ret
= GL_EXTCALL(glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB
));
3412 checkGLcall("glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB)");
3413 GL_EXTCALL(glShaderSourceARB(ret
, 1, (const char**)&buffer
.buffer
, NULL
));
3414 checkGLcall("glShaderSourceARB(ret, 1, &buffer.buffer, NULL)");
3415 GL_EXTCALL(glCompileShaderARB(ret
));
3416 checkGLcall("glCompileShaderARB(ret)");
3418 shader_buffer_free(&buffer
);
3422 static void hardcode_local_constants(IWineD3DBaseShaderImpl
*shader
, const WineD3D_GL_Info
*gl_info
,
3423 GLhandleARB programId
, char prefix
)
3425 const local_constant
*lconst
;
3430 LIST_FOR_EACH_ENTRY(lconst
, &shader
->baseShader
.constantsF
, local_constant
, entry
) {
3431 value
= (const float *)lconst
->value
;
3432 snprintf(glsl_name
, sizeof(glsl_name
), "%cLC%u", prefix
, lconst
->idx
);
3433 tmp_loc
= GL_EXTCALL(glGetUniformLocationARB(programId
, glsl_name
));
3434 GL_EXTCALL(glUniform4fvARB(tmp_loc
, 1, value
));
3436 checkGLcall("Hardcoding local constants\n");
3439 /** Sets the GLSL program ID for the given pixel and vertex shader combination.
3440 * It sets the programId on the current StateBlock (because it should be called
3441 * inside of the DrawPrimitive() part of the render loop).
3443 * If a program for the given combination does not exist, create one, and store
3444 * the program in the hash table. If it creates a program, it will link the
3445 * given objects, too.
3447 static void set_glsl_shader_program(IWineD3DDevice
*iface
, BOOL use_ps
, BOOL use_vs
) {
3448 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
3449 struct shader_glsl_priv
*priv
= This
->shader_priv
;
3450 const WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
3451 IWineD3DPixelShader
*pshader
= This
->stateBlock
->pixelShader
;
3452 IWineD3DVertexShader
*vshader
= This
->stateBlock
->vertexShader
;
3453 struct glsl_shader_prog_link
*entry
= NULL
;
3454 GLhandleARB programId
= 0;
3455 GLhandleARB reorder_shader_id
= 0;
3458 GLhandleARB vshader_id
, pshader_id
;
3459 struct ps_compile_args ps_compile_args
;
3460 struct vs_compile_args vs_compile_args
;
3463 find_vs_compile_args((IWineD3DVertexShaderImpl
*)This
->stateBlock
->vertexShader
, This
->stateBlock
, &vs_compile_args
);
3465 /* FIXME: Do we really have to spend CPU cycles to generate a few zeroed bytes? */
3466 memset(&vs_compile_args
, 0, sizeof(vs_compile_args
));
3469 find_ps_compile_args((IWineD3DPixelShaderImpl
*)This
->stateBlock
->pixelShader
, This
->stateBlock
, &ps_compile_args
);
3471 /* FIXME: Do we really have to spend CPU cycles to generate a few zeroed bytes? */
3472 memset(&ps_compile_args
, 0, sizeof(ps_compile_args
));
3474 entry
= get_glsl_program_entry(priv
, vshader
, pshader
, &vs_compile_args
, &ps_compile_args
);
3476 priv
->glsl_program
= entry
;
3480 /* If we get to this point, then no matching program exists, so we create one */
3481 programId
= GL_EXTCALL(glCreateProgramObjectARB());
3482 TRACE("Created new GLSL shader program %u\n", programId
);
3484 /* Create the entry */
3485 entry
= HeapAlloc(GetProcessHeap(), 0, sizeof(struct glsl_shader_prog_link
));
3486 entry
->programId
= programId
;
3487 entry
->vshader
= vshader
;
3488 entry
->pshader
= pshader
;
3489 entry
->vs_args
= vs_compile_args
;
3490 entry
->ps_args
= ps_compile_args
;
3491 entry
->constant_version
= 0;
3492 /* Add the hash table entry */
3493 add_glsl_program_entry(priv
, entry
);
3495 /* Set the current program */
3496 priv
->glsl_program
= entry
;
3499 vshader_id
= find_gl_vshader((IWineD3DVertexShaderImpl
*) vshader
, &vs_compile_args
);
3504 /* Attach GLSL vshader */
3506 const unsigned int max_attribs
= 16; /* TODO: Will this always be the case? It is at the moment... */
3509 reorder_shader_id
= generate_param_reorder_function(vshader
, pshader
, gl_info
);
3510 TRACE("Attaching GLSL shader object %u to program %u\n", reorder_shader_id
, programId
);
3511 GL_EXTCALL(glAttachObjectARB(programId
, reorder_shader_id
));
3512 checkGLcall("glAttachObjectARB");
3513 /* Flag the reorder function for deletion, then it will be freed automatically when the program
3516 GL_EXTCALL(glDeleteObjectARB(reorder_shader_id
));
3518 TRACE("Attaching GLSL shader object %u to program %u\n", vshader_id
, programId
);
3519 GL_EXTCALL(glAttachObjectARB(programId
, vshader_id
));
3520 checkGLcall("glAttachObjectARB");
3522 /* Bind vertex attributes to a corresponding index number to match
3523 * the same index numbers as ARB_vertex_programs (makes loading
3524 * vertex attributes simpler). With this method, we can use the
3525 * exact same code to load the attributes later for both ARB and
3528 * We have to do this here because we need to know the Program ID
3529 * in order to make the bindings work, and it has to be done prior
3530 * to linking the GLSL program. */
3531 for (i
= 0; i
< max_attribs
; ++i
) {
3532 if (((IWineD3DBaseShaderImpl
*)vshader
)->baseShader
.reg_maps
.attributes
[i
]) {
3533 snprintf(tmp_name
, sizeof(tmp_name
), "attrib%i", i
);
3534 GL_EXTCALL(glBindAttribLocationARB(programId
, i
, tmp_name
));
3537 checkGLcall("glBindAttribLocationARB");
3539 list_add_head(&((IWineD3DBaseShaderImpl
*)vshader
)->baseShader
.linked_programs
, &entry
->vshader_entry
);
3543 pshader_id
= find_gl_pshader((IWineD3DPixelShaderImpl
*) pshader
, &ps_compile_args
);
3548 /* Attach GLSL pshader */
3550 TRACE("Attaching GLSL shader object %u to program %u\n", pshader_id
, programId
);
3551 GL_EXTCALL(glAttachObjectARB(programId
, pshader_id
));
3552 checkGLcall("glAttachObjectARB");
3554 list_add_head(&((IWineD3DBaseShaderImpl
*)pshader
)->baseShader
.linked_programs
, &entry
->pshader_entry
);
3557 /* Link the program */
3558 TRACE("Linking GLSL shader program %u\n", programId
);
3559 GL_EXTCALL(glLinkProgramARB(programId
));
3560 print_glsl_info_log(&GLINFO_LOCATION
, programId
);
3562 entry
->vuniformF_locations
= HeapAlloc(GetProcessHeap(), 0, sizeof(GLhandleARB
) * GL_LIMITS(vshader_constantsF
));
3563 for (i
= 0; i
< GL_LIMITS(vshader_constantsF
); ++i
) {
3564 snprintf(glsl_name
, sizeof(glsl_name
), "VC[%i]", i
);
3565 entry
->vuniformF_locations
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, glsl_name
));
3567 for (i
= 0; i
< MAX_CONST_I
; ++i
) {
3568 snprintf(glsl_name
, sizeof(glsl_name
), "VI[%i]", i
);
3569 entry
->vuniformI_locations
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, glsl_name
));
3571 entry
->puniformF_locations
= HeapAlloc(GetProcessHeap(), 0, sizeof(GLhandleARB
) * GL_LIMITS(pshader_constantsF
));
3572 for (i
= 0; i
< GL_LIMITS(pshader_constantsF
); ++i
) {
3573 snprintf(glsl_name
, sizeof(glsl_name
), "PC[%i]", i
);
3574 entry
->puniformF_locations
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, glsl_name
));
3576 for (i
= 0; i
< MAX_CONST_I
; ++i
) {
3577 snprintf(glsl_name
, sizeof(glsl_name
), "PI[%i]", i
);
3578 entry
->puniformI_locations
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, glsl_name
));
3582 for(i
= 0; i
< ((IWineD3DPixelShaderImpl
*)pshader
)->numbumpenvmatconsts
; i
++) {
3584 sprintf(name
, "bumpenvmat%d", ((IWineD3DPixelShaderImpl
*)pshader
)->bumpenvmatconst
[i
].texunit
);
3585 entry
->bumpenvmat_location
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, name
));
3586 sprintf(name
, "luminancescale%d", ((IWineD3DPixelShaderImpl
*)pshader
)->luminanceconst
[i
].texunit
);
3587 entry
->luminancescale_location
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, name
));
3588 sprintf(name
, "luminanceoffset%d", ((IWineD3DPixelShaderImpl
*)pshader
)->luminanceconst
[i
].texunit
);
3589 entry
->luminanceoffset_location
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, name
));
3593 if (use_ps
&& ps_compile_args
.np2_fixup
) {
3595 for (i
= 0; i
< MAX_FRAGMENT_SAMPLERS
; ++i
) {
3596 if (ps_compile_args
.np2_fixup
& (1 << i
)) {
3597 sprintf(name
, "PsamplerNP2Fixup%u", i
);
3598 entry
->np2Fixup_location
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, name
));
3600 entry
->np2Fixup_location
[i
] = -1;
3605 entry
->posFixup_location
= GL_EXTCALL(glGetUniformLocationARB(programId
, "posFixup"));
3606 entry
->ycorrection_location
= GL_EXTCALL(glGetUniformLocationARB(programId
, "ycorrection"));
3607 checkGLcall("Find glsl program uniform locations");
3610 && WINED3DSHADER_VERSION_MAJOR(((IWineD3DPixelShaderImpl
*)pshader
)->baseShader
.reg_maps
.shader_version
) >= 3
3611 && ((IWineD3DPixelShaderImpl
*)pshader
)->declared_in_count
> GL_LIMITS(glsl_varyings
) / 4)
3613 TRACE("Shader %d needs vertex color clamping disabled\n", programId
);
3614 entry
->vertex_color_clamp
= GL_FALSE
;
3616 entry
->vertex_color_clamp
= GL_FIXED_ONLY_ARB
;
3619 /* Set the shader to allow uniform loading on it */
3620 GL_EXTCALL(glUseProgramObjectARB(programId
));
3621 checkGLcall("glUseProgramObjectARB(programId)");
3623 /* Load the vertex and pixel samplers now. The function that finds the mappings makes sure
3624 * that it stays the same for each vertexshader-pixelshader pair(=linked glsl program). If
3625 * a pshader with fixed function pipeline is used there are no vertex samplers, and if a
3626 * vertex shader with fixed function pixel processing is used we make sure that the card
3627 * supports enough samplers to allow the max number of vertex samplers with all possible
3628 * fixed function fragment processing setups. So once the program is linked these samplers
3632 /* Load vertex shader samplers */
3633 shader_glsl_load_vsamplers(gl_info
, This
->texUnitMap
, programId
);
3636 /* Load pixel shader samplers */
3637 shader_glsl_load_psamplers(gl_info
, This
->texUnitMap
, programId
);
3640 /* If the local constants do not have to be loaded with the environment constants,
3641 * load them now to have them hardcoded in the GLSL program. This saves some CPU cycles
3644 if(pshader
&& !((IWineD3DPixelShaderImpl
*)pshader
)->baseShader
.load_local_constsF
) {
3645 hardcode_local_constants((IWineD3DBaseShaderImpl
*) pshader
, gl_info
, programId
, 'P');
3647 if(vshader
&& !((IWineD3DVertexShaderImpl
*)vshader
)->baseShader
.load_local_constsF
) {
3648 hardcode_local_constants((IWineD3DBaseShaderImpl
*) vshader
, gl_info
, programId
, 'V');
3652 static GLhandleARB
create_glsl_blt_shader(const WineD3D_GL_Info
*gl_info
, enum tex_types tex_type
)
3654 GLhandleARB program_id
;
3655 GLhandleARB vshader_id
, pshader_id
;
3656 static const char *blt_vshader
[] =
3661 " gl_Position = gl_Vertex;\n"
3662 " gl_FrontColor = vec4(1.0);\n"
3663 " gl_TexCoord[0] = gl_MultiTexCoord0;\n"
3667 static const char *blt_pshaders
[tex_type_count
] =
3673 "uniform sampler2D sampler;\n"
3676 " gl_FragDepth = texture2D(sampler, gl_TexCoord[0].xy).x;\n"
3682 "uniform samplerCube sampler;\n"
3685 " gl_FragDepth = textureCube(sampler, gl_TexCoord[0].xyz).x;\n"
3689 "#extension GL_ARB_texture_rectangle : enable\n"
3690 "uniform sampler2DRect sampler;\n"
3693 " gl_FragDepth = texture2DRect(sampler, gl_TexCoord[0].xy).x;\n"
3697 if (!blt_pshaders
[tex_type
])
3699 FIXME("tex_type %#x not supported\n", tex_type
);
3703 vshader_id
= GL_EXTCALL(glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB
));
3704 GL_EXTCALL(glShaderSourceARB(vshader_id
, 1, blt_vshader
, NULL
));
3705 GL_EXTCALL(glCompileShaderARB(vshader_id
));
3707 pshader_id
= GL_EXTCALL(glCreateShaderObjectARB(GL_FRAGMENT_SHADER_ARB
));
3708 GL_EXTCALL(glShaderSourceARB(pshader_id
, 1, &blt_pshaders
[tex_type
], NULL
));
3709 GL_EXTCALL(glCompileShaderARB(pshader_id
));
3711 program_id
= GL_EXTCALL(glCreateProgramObjectARB());
3712 GL_EXTCALL(glAttachObjectARB(program_id
, vshader_id
));
3713 GL_EXTCALL(glAttachObjectARB(program_id
, pshader_id
));
3714 GL_EXTCALL(glLinkProgramARB(program_id
));
3716 print_glsl_info_log(&GLINFO_LOCATION
, program_id
);
3718 /* Once linked we can mark the shaders for deletion. They will be deleted once the program
3721 GL_EXTCALL(glDeleteObjectARB(vshader_id
));
3722 GL_EXTCALL(glDeleteObjectARB(pshader_id
));
3726 static void shader_glsl_select(IWineD3DDevice
*iface
, BOOL usePS
, BOOL useVS
) {
3727 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
3728 struct shader_glsl_priv
*priv
= This
->shader_priv
;
3729 const WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
3730 GLhandleARB program_id
= 0;
3731 GLenum old_vertex_color_clamp
, current_vertex_color_clamp
;
3733 old_vertex_color_clamp
= priv
->glsl_program
? priv
->glsl_program
->vertex_color_clamp
: GL_FIXED_ONLY_ARB
;
3735 if (useVS
|| usePS
) set_glsl_shader_program(iface
, usePS
, useVS
);
3736 else priv
->glsl_program
= NULL
;
3738 current_vertex_color_clamp
= priv
->glsl_program
? priv
->glsl_program
->vertex_color_clamp
: GL_FIXED_ONLY_ARB
;
3740 if (old_vertex_color_clamp
!= current_vertex_color_clamp
) {
3741 if (GL_SUPPORT(ARB_COLOR_BUFFER_FLOAT
)) {
3742 GL_EXTCALL(glClampColorARB(GL_CLAMP_VERTEX_COLOR_ARB
, current_vertex_color_clamp
));
3743 checkGLcall("glClampColorARB");
3745 FIXME("vertex color clamp needs to be changed, but extension not supported.\n");
3749 program_id
= priv
->glsl_program
? priv
->glsl_program
->programId
: 0;
3750 if (program_id
) TRACE("Using GLSL program %u\n", program_id
);
3751 GL_EXTCALL(glUseProgramObjectARB(program_id
));
3752 checkGLcall("glUseProgramObjectARB");
3755 static void shader_glsl_select_depth_blt(IWineD3DDevice
*iface
, enum tex_types tex_type
) {
3756 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
3757 const WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
3758 struct shader_glsl_priv
*priv
= This
->shader_priv
;
3759 GLhandleARB
*blt_program
= &priv
->depth_blt_program
[tex_type
];
3761 if (!*blt_program
) {
3763 *blt_program
= create_glsl_blt_shader(gl_info
, tex_type
);
3764 loc
= GL_EXTCALL(glGetUniformLocationARB(*blt_program
, "sampler"));
3765 GL_EXTCALL(glUseProgramObjectARB(*blt_program
));
3766 GL_EXTCALL(glUniform1iARB(loc
, 0));
3768 GL_EXTCALL(glUseProgramObjectARB(*blt_program
));
3772 static void shader_glsl_deselect_depth_blt(IWineD3DDevice
*iface
) {
3773 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
3774 const WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
3775 struct shader_glsl_priv
*priv
= This
->shader_priv
;
3776 GLhandleARB program_id
;
3778 program_id
= priv
->glsl_program
? priv
->glsl_program
->programId
: 0;
3779 if (program_id
) TRACE("Using GLSL program %u\n", program_id
);
3781 GL_EXTCALL(glUseProgramObjectARB(program_id
));
3782 checkGLcall("glUseProgramObjectARB");
3785 static void shader_glsl_destroy(IWineD3DBaseShader
*iface
) {
3786 const struct list
*linked_programs
;
3787 IWineD3DBaseShaderImpl
*This
= (IWineD3DBaseShaderImpl
*) iface
;
3788 IWineD3DDeviceImpl
*device
= (IWineD3DDeviceImpl
*)This
->baseShader
.device
;
3789 struct shader_glsl_priv
*priv
= device
->shader_priv
;
3790 const WineD3D_GL_Info
*gl_info
= &device
->adapter
->gl_info
;
3791 IWineD3DPixelShaderImpl
*ps
= NULL
;
3792 IWineD3DVertexShaderImpl
*vs
= NULL
;
3794 /* Note: Do not use QueryInterface here to find out which shader type this is because this code
3795 * can be called from IWineD3DBaseShader::Release
3797 char pshader
= shader_is_pshader_version(This
->baseShader
.reg_maps
.shader_version
);
3800 ps
= (IWineD3DPixelShaderImpl
*) This
;
3801 if(ps
->num_gl_shaders
== 0) return;
3802 if (priv
->glsl_program
&& (IWineD3DBaseShader
*)priv
->glsl_program
->pshader
== iface
)
3803 shader_glsl_select(This
->baseShader
.device
, FALSE
, FALSE
);
3805 vs
= (IWineD3DVertexShaderImpl
*) This
;
3806 if(vs
->num_gl_shaders
== 0) return;
3807 if (priv
->glsl_program
&& (IWineD3DBaseShader
*)priv
->glsl_program
->vshader
== iface
)
3808 shader_glsl_select(This
->baseShader
.device
, FALSE
, FALSE
);
3811 linked_programs
= &This
->baseShader
.linked_programs
;
3813 TRACE("Deleting linked programs\n");
3814 if (linked_programs
->next
) {
3815 struct glsl_shader_prog_link
*entry
, *entry2
;
3818 LIST_FOR_EACH_ENTRY_SAFE(entry
, entry2
, linked_programs
, struct glsl_shader_prog_link
, pshader_entry
) {
3819 delete_glsl_program_entry(priv
, gl_info
, entry
);
3822 LIST_FOR_EACH_ENTRY_SAFE(entry
, entry2
, linked_programs
, struct glsl_shader_prog_link
, vshader_entry
) {
3823 delete_glsl_program_entry(priv
, gl_info
, entry
);
3832 for(i
= 0; i
< ps
->num_gl_shaders
; i
++) {
3833 TRACE("deleting pshader %u\n", ps
->gl_shaders
[i
].prgId
);
3834 GL_EXTCALL(glDeleteObjectARB(ps
->gl_shaders
[i
].prgId
));
3835 checkGLcall("glDeleteObjectARB");
3838 HeapFree(GetProcessHeap(), 0, ps
->gl_shaders
);
3839 ps
->gl_shaders
= NULL
;
3840 ps
->num_gl_shaders
= 0;
3841 ps
->shader_array_size
= 0;
3846 for(i
= 0; i
< vs
->num_gl_shaders
; i
++) {
3847 TRACE("deleting vshader %u\n", vs
->gl_shaders
[i
].prgId
);
3848 GL_EXTCALL(glDeleteObjectARB(vs
->gl_shaders
[i
].prgId
));
3849 checkGLcall("glDeleteObjectARB");
3852 HeapFree(GetProcessHeap(), 0, vs
->gl_shaders
);
3853 vs
->gl_shaders
= NULL
;
3854 vs
->num_gl_shaders
= 0;
3855 vs
->shader_array_size
= 0;
3859 static unsigned int glsl_program_key_hash(const void *key
)
3861 const glsl_program_key_t
*k
= key
;
3863 unsigned int hash
= ((DWORD_PTR
) k
->vshader
) | ((DWORD_PTR
) k
->pshader
) << 16;
3864 hash
+= ~(hash
<< 15);
3865 hash
^= (hash
>> 10);
3866 hash
+= (hash
<< 3);
3867 hash
^= (hash
>> 6);
3868 hash
+= ~(hash
<< 11);
3869 hash
^= (hash
>> 16);
3874 static BOOL
glsl_program_key_compare(const void *keya
, const void *keyb
)
3876 const glsl_program_key_t
*ka
= keya
;
3877 const glsl_program_key_t
*kb
= keyb
;
3879 return ka
->vshader
== kb
->vshader
&& ka
->pshader
== kb
->pshader
&&
3880 (memcmp(&ka
->ps_args
, &kb
->ps_args
, sizeof(kb
->ps_args
)) == 0) &&
3881 (memcmp(&ka
->vs_args
, &kb
->vs_args
, sizeof(kb
->vs_args
)) == 0);
3884 static BOOL
constant_heap_init(struct constant_heap
*heap
, unsigned int constant_count
)
3886 SIZE_T size
= (constant_count
+ 1) * sizeof(*heap
->entries
) + constant_count
* sizeof(*heap
->positions
);
3887 void *mem
= HeapAlloc(GetProcessHeap(), 0, size
);
3891 ERR("Failed to allocate memory\n");
3895 heap
->entries
= mem
;
3896 heap
->entries
[1].version
= 0;
3897 heap
->positions
= (unsigned int *)(heap
->entries
+ constant_count
+ 1);
3903 static void constant_heap_free(struct constant_heap
*heap
)
3905 HeapFree(GetProcessHeap(), 0, heap
->entries
);
3908 static HRESULT
shader_glsl_alloc(IWineD3DDevice
*iface
) {
3909 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
3910 const WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
3911 struct shader_glsl_priv
*priv
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(struct shader_glsl_priv
));
3912 SIZE_T stack_size
= wined3d_log2i(max(GL_LIMITS(vshader_constantsF
), GL_LIMITS(pshader_constantsF
))) + 1;
3914 priv
->stack
= HeapAlloc(GetProcessHeap(), 0, stack_size
* sizeof(*priv
->stack
));
3917 ERR("Failed to allocate memory.\n");
3918 HeapFree(GetProcessHeap(), 0, priv
);
3919 return E_OUTOFMEMORY
;
3922 if (!constant_heap_init(&priv
->vconst_heap
, GL_LIMITS(vshader_constantsF
)))
3924 ERR("Failed to initialize vertex shader constant heap\n");
3925 HeapFree(GetProcessHeap(), 0, priv
->stack
);
3926 HeapFree(GetProcessHeap(), 0, priv
);
3927 return E_OUTOFMEMORY
;
3930 if (!constant_heap_init(&priv
->pconst_heap
, GL_LIMITS(pshader_constantsF
)))
3932 ERR("Failed to initialize pixel shader constant heap\n");
3933 constant_heap_free(&priv
->vconst_heap
);
3934 HeapFree(GetProcessHeap(), 0, priv
->stack
);
3935 HeapFree(GetProcessHeap(), 0, priv
);
3936 return E_OUTOFMEMORY
;
3939 priv
->glsl_program_lookup
= hash_table_create(glsl_program_key_hash
, glsl_program_key_compare
);
3940 priv
->next_constant_version
= 1;
3942 This
->shader_priv
= priv
;
3946 static void shader_glsl_free(IWineD3DDevice
*iface
) {
3947 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
3948 const WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
3949 struct shader_glsl_priv
*priv
= This
->shader_priv
;
3952 for (i
= 0; i
< tex_type_count
; ++i
)
3954 if (priv
->depth_blt_program
[i
])
3956 GL_EXTCALL(glDeleteObjectARB(priv
->depth_blt_program
[i
]));
3960 hash_table_destroy(priv
->glsl_program_lookup
, NULL
, NULL
);
3961 constant_heap_free(&priv
->pconst_heap
);
3962 constant_heap_free(&priv
->vconst_heap
);
3964 HeapFree(GetProcessHeap(), 0, This
->shader_priv
);
3965 This
->shader_priv
= NULL
;
3968 static BOOL
shader_glsl_dirty_const(IWineD3DDevice
*iface
) {
3969 /* TODO: GL_EXT_bindable_uniform can be used to share constants across shaders */
3973 static GLuint
shader_glsl_generate_pshader(IWineD3DPixelShader
*iface
, SHADER_BUFFER
*buffer
, const struct ps_compile_args
*args
) {
3974 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*)iface
;
3975 const struct shader_reg_maps
*reg_maps
= &This
->baseShader
.reg_maps
;
3976 CONST DWORD
*function
= This
->baseShader
.function
;
3977 const char *fragcolor
;
3978 const WineD3D_GL_Info
*gl_info
= &((IWineD3DDeviceImpl
*)This
->baseShader
.device
)->adapter
->gl_info
;
3980 /* Create the hw GLSL shader object and assign it as the shader->prgId */
3981 GLhandleARB shader_obj
= GL_EXTCALL(glCreateShaderObjectARB(GL_FRAGMENT_SHADER_ARB
));
3983 shader_addline(buffer
, "#version 120\n");
3985 if (GL_SUPPORT(ARB_DRAW_BUFFERS
)) {
3986 shader_addline(buffer
, "#extension GL_ARB_draw_buffers : enable\n");
3988 if (GL_SUPPORT(ARB_TEXTURE_RECTANGLE
)) {
3989 /* The spec says that it doesn't have to be explicitly enabled, but the nvidia
3990 * drivers write a warning if we don't do so
3992 shader_addline(buffer
, "#extension GL_ARB_texture_rectangle : enable\n");
3995 /* Base Declarations */
3996 shader_generate_glsl_declarations( (IWineD3DBaseShader
*) This
, reg_maps
, buffer
, &GLINFO_LOCATION
, args
);
3998 /* Pack 3.0 inputs */
3999 if (reg_maps
->shader_version
>= WINED3DPS_VERSION(3,0) && args
->vp_mode
!= vertexshader
) {
4000 pshader_glsl_input_pack(iface
, buffer
, This
->semantics_in
, reg_maps
, args
->vp_mode
);
4003 /* Base Shader Body */
4004 shader_generate_main( (IWineD3DBaseShader
*) This
, buffer
, reg_maps
, function
);
4006 /* Pixel shaders < 2.0 place the resulting color in R0 implicitly */
4007 if (reg_maps
->shader_version
< WINED3DPS_VERSION(2,0))
4009 /* Some older cards like GeforceFX ones don't support multiple buffers, so also not gl_FragData */
4010 if(GL_SUPPORT(ARB_DRAW_BUFFERS
))
4011 shader_addline(buffer
, "gl_FragData[0] = R0;\n");
4013 shader_addline(buffer
, "gl_FragColor = R0;\n");
4016 if(GL_SUPPORT(ARB_DRAW_BUFFERS
)) {
4017 fragcolor
= "gl_FragData[0]";
4019 fragcolor
= "gl_FragColor";
4021 if(args
->srgb_correction
) {
4022 shader_addline(buffer
, "tmp0.xyz = pow(%s.xyz, vec3(%f, %f, %f)) * vec3(%f, %f, %f) - vec3(%f, %f, %f);\n",
4023 fragcolor
, srgb_pow
, srgb_pow
, srgb_pow
, srgb_mul_high
, srgb_mul_high
, srgb_mul_high
,
4024 srgb_sub_high
, srgb_sub_high
, srgb_sub_high
);
4025 shader_addline(buffer
, "tmp1.xyz = %s.xyz * srgb_mul_low.xyz;\n", fragcolor
);
4026 shader_addline(buffer
, "%s.x = %s.x < srgb_comparison.x ? tmp1.x : tmp0.x;\n", fragcolor
, fragcolor
);
4027 shader_addline(buffer
, "%s.y = %s.y < srgb_comparison.y ? tmp1.y : tmp0.y;\n", fragcolor
, fragcolor
);
4028 shader_addline(buffer
, "%s.z = %s.z < srgb_comparison.z ? tmp1.z : tmp0.z;\n", fragcolor
, fragcolor
);
4029 shader_addline(buffer
, "%s = clamp(%s, 0.0, 1.0);\n", fragcolor
, fragcolor
);
4031 /* Pixel shader < 3.0 do not replace the fog stage.
4032 * This implements linear fog computation and blending.
4033 * TODO: non linear fog
4034 * NOTE: gl_Fog.start and gl_Fog.end don't hold fog start s and end e but
4035 * -1/(e-s) and e/(e-s) respectively.
4037 if(reg_maps
->shader_version
< WINED3DPS_VERSION(3,0)) {
4039 case FOG_OFF
: break;
4041 shader_addline(buffer
, "float fogstart = -1.0 / (gl_Fog.end - gl_Fog.start);\n");
4042 shader_addline(buffer
, "float fogend = gl_Fog.end * -fogstart;\n");
4043 shader_addline(buffer
, "float Fog = clamp(gl_FogFragCoord * fogstart + fogend, 0.0, 1.0);\n");
4044 shader_addline(buffer
, "%s.xyz = mix(gl_Fog.color.xyz, %s.xyz, Fog);\n", fragcolor
, fragcolor
);
4047 /* Fog = e^(-gl_Fog.density * gl_FogFragCoord) */
4048 shader_addline(buffer
, "float Fog = exp(-gl_Fog.density * gl_FogFragCoord);\n");
4049 shader_addline(buffer
, "Fog = clamp(Fog, 0.0, 1.0);\n");
4050 shader_addline(buffer
, "%s.xyz = mix(gl_Fog.color.xyz, %s.xyz, Fog);\n", fragcolor
, fragcolor
);
4053 /* Fog = e^(-(gl_Fog.density * gl_FogFragCoord)^2) */
4054 shader_addline(buffer
, "float Fog = exp(-gl_Fog.density * gl_Fog.density * gl_FogFragCoord * gl_FogFragCoord);\n");
4055 shader_addline(buffer
, "Fog = clamp(Fog, 0.0, 1.0);\n");
4056 shader_addline(buffer
, "%s.xyz = mix(gl_Fog.color.xyz, %s.xyz, Fog);\n", fragcolor
, fragcolor
);
4061 shader_addline(buffer
, "}\n");
4063 TRACE("Compiling shader object %u\n", shader_obj
);
4064 GL_EXTCALL(glShaderSourceARB(shader_obj
, 1, (const char**)&buffer
->buffer
, NULL
));
4065 GL_EXTCALL(glCompileShaderARB(shader_obj
));
4066 print_glsl_info_log(&GLINFO_LOCATION
, shader_obj
);
4068 /* Store the shader object */
4072 static GLuint
shader_glsl_generate_vshader(IWineD3DVertexShader
*iface
, SHADER_BUFFER
*buffer
, const struct vs_compile_args
*args
) {
4073 IWineD3DVertexShaderImpl
*This
= (IWineD3DVertexShaderImpl
*)iface
;
4074 const struct shader_reg_maps
*reg_maps
= &This
->baseShader
.reg_maps
;
4075 CONST DWORD
*function
= This
->baseShader
.function
;
4076 const WineD3D_GL_Info
*gl_info
= &((IWineD3DDeviceImpl
*)This
->baseShader
.device
)->adapter
->gl_info
;
4078 /* Create the hw GLSL shader program and assign it as the shader->prgId */
4079 GLhandleARB shader_obj
= GL_EXTCALL(glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB
));
4081 shader_addline(buffer
, "#version 120\n");
4083 /* Base Declarations */
4084 shader_generate_glsl_declarations( (IWineD3DBaseShader
*) This
, reg_maps
, buffer
, &GLINFO_LOCATION
, NULL
);
4086 /* Base Shader Body */
4087 shader_generate_main( (IWineD3DBaseShader
*) This
, buffer
, reg_maps
, function
);
4089 /* Unpack 3.0 outputs */
4090 if (reg_maps
->shader_version
>= WINED3DVS_VERSION(3,0)) shader_addline(buffer
, "order_ps_input(OUT);\n");
4091 else shader_addline(buffer
, "order_ps_input();\n");
4093 /* The D3DRS_FOGTABLEMODE render state defines if the shader-generated fog coord is used
4094 * or if the fragment depth is used. If the fragment depth is used(FOGTABLEMODE != NONE),
4095 * the fog frag coord is thrown away. If the fog frag coord is used, but not written by
4096 * the shader, it is set to 0.0(fully fogged, since start = 1.0, end = 0.0)
4098 if(args
->fog_src
== VS_FOG_Z
) {
4099 shader_addline(buffer
, "gl_FogFragCoord = gl_Position.z;\n");
4100 } else if (!reg_maps
->fog
) {
4101 shader_addline(buffer
, "gl_FogFragCoord = 0.0;\n");
4104 /* Write the final position.
4106 * OpenGL coordinates specify the center of the pixel while d3d coords specify
4107 * the corner. The offsets are stored in z and w in posFixup. posFixup.y contains
4108 * 1.0 or -1.0 to turn the rendering upside down for offscreen rendering. PosFixup.x
4109 * contains 1.0 to allow a mad.
4111 shader_addline(buffer
, "gl_Position.y = gl_Position.y * posFixup.y;\n");
4112 shader_addline(buffer
, "gl_Position.xy += posFixup.zw * gl_Position.ww;\n");
4114 /* Z coord [0;1]->[-1;1] mapping, see comment in transform_projection in state.c
4116 * Basically we want (in homogeneous coordinates) z = z * 2 - 1. However, shaders are run
4117 * before the homogeneous divide, so we have to take the w into account: z = ((z / w) * 2 - 1) * w,
4118 * which is the same as z = z * 2 - w.
4120 shader_addline(buffer
, "gl_Position.z = gl_Position.z * 2.0 - gl_Position.w;\n");
4122 shader_addline(buffer
, "}\n");
4124 TRACE("Compiling shader object %u\n", shader_obj
);
4125 GL_EXTCALL(glShaderSourceARB(shader_obj
, 1, (const char**)&buffer
->buffer
, NULL
));
4126 GL_EXTCALL(glCompileShaderARB(shader_obj
));
4127 print_glsl_info_log(&GLINFO_LOCATION
, shader_obj
);
4132 static void shader_glsl_get_caps(WINED3DDEVTYPE devtype
, const WineD3D_GL_Info
*gl_info
, struct shader_caps
*pCaps
)
4134 /* Nvidia Geforce6/7 or Ati R4xx/R5xx cards with GLSL support, support VS 3.0 but older Nvidia/Ati
4135 * models with GLSL support only support 2.0. In case of nvidia we can detect VS 2.0 support using
4136 * vs_nv_version which is based on NV_vertex_program.
4137 * For Ati cards there's no way using glsl (it abstracts the lowlevel info away) and also not
4138 * using ARB_vertex_program. It is safe to assume that when a card supports pixel shader 2.0 it
4139 * supports vertex shader 2.0 too and the way around. We can detect ps2.0 using the maximum number
4140 * of native instructions, so use that here. For more info see the pixel shader versioning code below.
4142 if((GLINFO_LOCATION
.vs_nv_version
== VS_VERSION_20
) || (GLINFO_LOCATION
.ps_arb_max_instructions
<= 512))
4143 pCaps
->VertexShaderVersion
= WINED3DVS_VERSION(2,0);
4145 pCaps
->VertexShaderVersion
= WINED3DVS_VERSION(3,0);
4146 TRACE_(d3d_caps
)("Hardware vertex shader version %d.%d enabled (GLSL)\n", (pCaps
->VertexShaderVersion
>> 8) & 0xff, pCaps
->VertexShaderVersion
& 0xff);
4147 /* Subtract the other potential uniforms from the max available (bools, ints, and 1 row of projection matrix) */
4148 pCaps
->MaxVertexShaderConst
= GL_LIMITS(vshader_constantsF
) - (MAX_CONST_B
/ 4) - MAX_CONST_I
- 1;
4150 /* Older DX9-class videocards (GeforceFX / Radeon >9500/X*00) only support pixel shader 2.0/2.0a/2.0b.
4151 * In OpenGL the extensions related to GLSL abstract lowlevel GL info away which is needed
4152 * to distinguish between 2.0 and 3.0 (and 2.0a/2.0b). In case of Nvidia we use their fragment
4153 * program extensions. On other hardware including ATI GL_ARB_fragment_program offers the info
4154 * in max native instructions. Intel and others also offer the info in this extension but they
4155 * don't support GLSL (at least on Windows).
4157 * PS2.0 requires at least 96 instructions, 2.0a/2.0b go up to 512. Assume that if the number
4158 * of instructions is 512 or less we have to do with ps2.0 hardware.
4159 * 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.
4161 if((GLINFO_LOCATION
.ps_nv_version
== PS_VERSION_20
) || (GLINFO_LOCATION
.ps_arb_max_instructions
<= 512))
4162 pCaps
->PixelShaderVersion
= WINED3DPS_VERSION(2,0);
4164 pCaps
->PixelShaderVersion
= WINED3DPS_VERSION(3,0);
4166 /* Subtract the other potential uniforms from the max available (bools & ints), and 2 states for fog.
4167 * In theory the texbem instruction may need one more shader constant too. But lets assume
4168 * that a sm <= 1.3 shader does not need all the uniforms provided by a glsl-capable card,
4169 * and lets not take away a uniform needlessly from all other shaders.
4171 pCaps
->MaxPixelShaderConst
= GL_LIMITS(pshader_constantsF
) - (MAX_CONST_B
/ 4) - MAX_CONST_I
- 2;
4173 /* FIXME: The following line is card dependent. -8.0 to 8.0 is the
4174 * Direct3D minimum requirement.
4176 * Both GL_ARB_fragment_program and GLSL require a "maximum representable magnitude"
4177 * of colors to be 2^10, and 2^32 for other floats. Should we use 1024 here?
4179 * The problem is that the refrast clamps temporary results in the shader to
4180 * [-MaxValue;+MaxValue]. If the card's max value is bigger than the one we advertize here,
4181 * then applications may miss the clamping behavior. On the other hand, if it is smaller,
4182 * the shader will generate incorrect results too. Unfortunately, GL deliberately doesn't
4183 * offer a way to query this.
4185 pCaps
->PixelShader1xMaxValue
= 8.0;
4186 TRACE_(d3d_caps
)("Hardware pixel shader version %d.%d enabled (GLSL)\n", (pCaps
->PixelShaderVersion
>> 8) & 0xff, pCaps
->PixelShaderVersion
& 0xff);
4189 static BOOL
shader_glsl_color_fixup_supported(struct color_fixup_desc fixup
)
4191 if (TRACE_ON(d3d_shader
) && TRACE_ON(d3d
))
4193 TRACE("Checking support for fixup:\n");
4194 dump_color_fixup_desc(fixup
);
4197 /* We support everything except YUV conversions. */
4198 if (!is_yuv_fixup(fixup
))
4204 TRACE("[FAILED]\n");
4208 static const SHADER_HANDLER shader_glsl_instruction_handler_table
[WINED3DSIH_TABLE_SIZE
] =
4210 /* WINED3DSIH_ABS */ shader_glsl_map2gl
,
4211 /* WINED3DSIH_ADD */ shader_glsl_arith
,
4212 /* WINED3DSIH_BEM */ pshader_glsl_bem
,
4213 /* WINED3DSIH_BREAK */ shader_glsl_break
,
4214 /* WINED3DSIH_BREAKC */ shader_glsl_breakc
,
4215 /* WINED3DSIH_BREAKP */ NULL
,
4216 /* WINED3DSIH_CALL */ shader_glsl_call
,
4217 /* WINED3DSIH_CALLNZ */ shader_glsl_callnz
,
4218 /* WINED3DSIH_CMP */ shader_glsl_cmp
,
4219 /* WINED3DSIH_CND */ shader_glsl_cnd
,
4220 /* WINED3DSIH_CRS */ shader_glsl_cross
,
4221 /* WINED3DSIH_DCL */ NULL
,
4222 /* WINED3DSIH_DEF */ NULL
,
4223 /* WINED3DSIH_DEFB */ NULL
,
4224 /* WINED3DSIH_DEFI */ NULL
,
4225 /* WINED3DSIH_DP2ADD */ pshader_glsl_dp2add
,
4226 /* WINED3DSIH_DP3 */ shader_glsl_dot
,
4227 /* WINED3DSIH_DP4 */ shader_glsl_dot
,
4228 /* WINED3DSIH_DST */ shader_glsl_dst
,
4229 /* WINED3DSIH_DSX */ shader_glsl_map2gl
,
4230 /* WINED3DSIH_DSY */ shader_glsl_map2gl
,
4231 /* WINED3DSIH_ELSE */ shader_glsl_else
,
4232 /* WINED3DSIH_ENDIF */ shader_glsl_end
,
4233 /* WINED3DSIH_ENDLOOP */ shader_glsl_end
,
4234 /* WINED3DSIH_ENDREP */ shader_glsl_end
,
4235 /* WINED3DSIH_EXP */ shader_glsl_map2gl
,
4236 /* WINED3DSIH_EXPP */ shader_glsl_expp
,
4237 /* WINED3DSIH_FRC */ shader_glsl_map2gl
,
4238 /* WINED3DSIH_IF */ shader_glsl_if
,
4239 /* WINED3DSIH_IFC */ shader_glsl_ifc
,
4240 /* WINED3DSIH_LABEL */ shader_glsl_label
,
4241 /* WINED3DSIH_LIT */ shader_glsl_lit
,
4242 /* WINED3DSIH_LOG */ shader_glsl_log
,
4243 /* WINED3DSIH_LOGP */ shader_glsl_log
,
4244 /* WINED3DSIH_LOOP */ shader_glsl_loop
,
4245 /* WINED3DSIH_LRP */ shader_glsl_lrp
,
4246 /* WINED3DSIH_M3x2 */ shader_glsl_mnxn
,
4247 /* WINED3DSIH_M3x3 */ shader_glsl_mnxn
,
4248 /* WINED3DSIH_M3x4 */ shader_glsl_mnxn
,
4249 /* WINED3DSIH_M4x3 */ shader_glsl_mnxn
,
4250 /* WINED3DSIH_M4x4 */ shader_glsl_mnxn
,
4251 /* WINED3DSIH_MAD */ shader_glsl_mad
,
4252 /* WINED3DSIH_MAX */ shader_glsl_map2gl
,
4253 /* WINED3DSIH_MIN */ shader_glsl_map2gl
,
4254 /* WINED3DSIH_MOV */ shader_glsl_mov
,
4255 /* WINED3DSIH_MOVA */ shader_glsl_mov
,
4256 /* WINED3DSIH_MUL */ shader_glsl_arith
,
4257 /* WINED3DSIH_NOP */ NULL
,
4258 /* WINED3DSIH_NRM */ shader_glsl_map2gl
,
4259 /* WINED3DSIH_PHASE */ NULL
,
4260 /* WINED3DSIH_POW */ shader_glsl_pow
,
4261 /* WINED3DSIH_RCP */ shader_glsl_rcp
,
4262 /* WINED3DSIH_REP */ shader_glsl_rep
,
4263 /* WINED3DSIH_RET */ NULL
,
4264 /* WINED3DSIH_RSQ */ shader_glsl_rsq
,
4265 /* WINED3DSIH_SETP */ NULL
,
4266 /* WINED3DSIH_SGE */ shader_glsl_compare
,
4267 /* WINED3DSIH_SGN */ shader_glsl_map2gl
,
4268 /* WINED3DSIH_SINCOS */ shader_glsl_sincos
,
4269 /* WINED3DSIH_SLT */ shader_glsl_compare
,
4270 /* WINED3DSIH_SUB */ shader_glsl_arith
,
4271 /* WINED3DSIH_TEX */ pshader_glsl_tex
,
4272 /* WINED3DSIH_TEXBEM */ pshader_glsl_texbem
,
4273 /* WINED3DSIH_TEXBEML */ pshader_glsl_texbem
,
4274 /* WINED3DSIH_TEXCOORD */ pshader_glsl_texcoord
,
4275 /* WINED3DSIH_TEXDEPTH */ pshader_glsl_texdepth
,
4276 /* WINED3DSIH_TEXDP3 */ pshader_glsl_texdp3
,
4277 /* WINED3DSIH_TEXDP3TEX */ pshader_glsl_texdp3tex
,
4278 /* WINED3DSIH_TEXKILL */ pshader_glsl_texkill
,
4279 /* WINED3DSIH_TEXLDD */ NULL
,
4280 /* WINED3DSIH_TEXLDL */ shader_glsl_texldl
,
4281 /* WINED3DSIH_TEXM3x2DEPTH */ pshader_glsl_texm3x2depth
,
4282 /* WINED3DSIH_TEXM3x2PAD */ pshader_glsl_texm3x2pad
,
4283 /* WINED3DSIH_TEXM3x2TEX */ pshader_glsl_texm3x2tex
,
4284 /* WINED3DSIH_TEXM3x3 */ pshader_glsl_texm3x3
,
4285 /* WINED3DSIH_TEXM3x3DIFF */ NULL
,
4286 /* WINED3DSIH_TEXM3x3PAD */ pshader_glsl_texm3x3pad
,
4287 /* WINED3DSIH_TEXM3x3SPEC */ pshader_glsl_texm3x3spec
,
4288 /* WINED3DSIH_TEXM3x3TEX */ pshader_glsl_texm3x3tex
,
4289 /* WINED3DSIH_TEXM3x3VSPEC */ pshader_glsl_texm3x3vspec
,
4290 /* WINED3DSIH_TEXREG2AR */ pshader_glsl_texreg2ar
,
4291 /* WINED3DSIH_TEXREG2GB */ pshader_glsl_texreg2gb
,
4292 /* WINED3DSIH_TEXREG2RGB */ pshader_glsl_texreg2rgb
,
4295 const shader_backend_t glsl_shader_backend
= {
4296 shader_glsl_instruction_handler_table
,
4298 shader_glsl_select_depth_blt
,
4299 shader_glsl_deselect_depth_blt
,
4300 shader_glsl_update_float_vertex_constants
,
4301 shader_glsl_update_float_pixel_constants
,
4302 shader_glsl_load_constants
,
4303 shader_glsl_load_np2fixup_constants
,
4304 shader_glsl_destroy
,
4307 shader_glsl_dirty_const
,
4308 shader_glsl_generate_pshader
,
4309 shader_glsl_generate_vshader
,
4310 shader_glsl_get_caps
,
4311 shader_glsl_color_fixup_supported
,