2 * GLSL pixel and vertex shader implementation
4 * Copyright 2006 Jason Green
5 * Copyright 2006-2007 Henri Verbeet
6 * Copyright 2007-2008 Stefan Dösinger for CodeWeavers
7 * Copyright 2009 Henri Verbeet for CodeWeavers
9 * This library is free software; you can redistribute it and/or
10 * modify it under the terms of the GNU Lesser General Public
11 * License as published by the Free Software Foundation; either
12 * version 2.1 of the License, or (at your option) any later version.
14 * This library is distributed in the hope that it will be useful,
15 * but WITHOUT ANY WARRANTY; without even the implied warranty of
16 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
17 * Lesser General Public License for more details.
19 * You should have received a copy of the GNU Lesser General Public
20 * License along with this library; if not, write to the Free Software
21 * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301, USA
25 * D3D shader asm has swizzles on source parameters, and write masks for
26 * destination parameters. GLSL uses swizzles for both. The result of this is
27 * that for example "mov dst.xw, src.zyxw" becomes "dst.xw = src.zw" in GLSL.
28 * Ie, to generate a proper GLSL source swizzle, we need to take the D3D write
29 * mask for the destination parameter into account.
35 #include "wined3d_private.h"
37 WINE_DEFAULT_DEBUG_CHANNEL(d3d_shader
);
38 WINE_DECLARE_DEBUG_CHANNEL(d3d_constants
);
39 WINE_DECLARE_DEBUG_CHANNEL(d3d_caps
);
40 WINE_DECLARE_DEBUG_CHANNEL(d3d
);
42 #define GLINFO_LOCATION (*gl_info)
44 #define WINED3D_GLSL_SAMPLE_PROJECTED 0x1
45 #define WINED3D_GLSL_SAMPLE_RECT 0x2
46 #define WINED3D_GLSL_SAMPLE_LOD 0x4
61 } glsl_sample_function_t
;
65 HEAP_NODE_TRAVERSE_LEFT
,
66 HEAP_NODE_TRAVERSE_RIGHT
,
78 struct constant_entry
*entries
;
79 unsigned int *positions
;
83 /* GLSL shader private data */
84 struct shader_glsl_priv
{
85 struct hash_table_t
*glsl_program_lookup
;
86 struct glsl_shader_prog_link
*glsl_program
;
87 struct constant_heap vconst_heap
;
88 struct constant_heap pconst_heap
;
90 GLhandleARB depth_blt_program
[tex_type_count
];
91 UINT next_constant_version
;
94 /* Struct to maintain data about a linked GLSL program */
95 struct glsl_shader_prog_link
{
96 struct list vshader_entry
;
97 struct list pshader_entry
;
98 GLhandleARB programId
;
99 GLint
*vuniformF_locations
;
100 GLint
*puniformF_locations
;
101 GLint vuniformI_locations
[MAX_CONST_I
];
102 GLint puniformI_locations
[MAX_CONST_I
];
103 GLint posFixup_location
;
104 GLint np2Fixup_location
[MAX_FRAGMENT_SAMPLERS
];
105 GLint bumpenvmat_location
[MAX_TEXTURES
];
106 GLint luminancescale_location
[MAX_TEXTURES
];
107 GLint luminanceoffset_location
[MAX_TEXTURES
];
108 GLint ycorrection_location
;
109 GLenum vertex_color_clamp
;
110 IWineD3DVertexShader
*vshader
;
111 IWineD3DPixelShader
*pshader
;
112 struct vs_compile_args vs_args
;
113 struct ps_compile_args ps_args
;
114 UINT constant_version
;
118 IWineD3DVertexShader
*vshader
;
119 IWineD3DPixelShader
*pshader
;
120 struct ps_compile_args ps_args
;
121 struct vs_compile_args vs_args
;
122 } glsl_program_key_t
;
125 /** Prints the GLSL info log which will contain error messages if they exist */
126 static void print_glsl_info_log(const WineD3D_GL_Info
*gl_info
, GLhandleARB obj
)
128 int infologLength
= 0;
133 static const char * const spam
[] =
135 "Vertex shader was successfully compiled to run on hardware.\n", /* fglrx */
136 "Fragment shader was successfully compiled to run on hardware.\n", /* fglrx */
137 "Fragment shader(s) linked, vertex shader(s) linked. \n ", /* fglrx, with \n */
138 "Fragment shader(s) linked, vertex shader(s) linked.", /* fglrx, no \n */
139 "Vertex shader(s) linked, no fragment shader(s) defined. \n ", /* fglrx, with \n */
140 "Vertex shader(s) linked, no fragment shader(s) defined.", /* fglrx, no \n */
141 "Fragment shader was successfully compiled to run on hardware.\n"
142 "WARNING: 0:2: extension 'GL_ARB_draw_buffers' is not supported",
143 "Fragment shader(s) linked, no vertex shader(s) defined.", /* fglrx, no \n */
144 "Fragment shader(s) linked, no vertex shader(s) defined. \n ", /* fglrx, with \n */
145 "WARNING: 0:2: extension 'GL_ARB_draw_buffers' is not supported\n" /* MacOS ati */
148 if (!TRACE_ON(d3d_shader
) && !FIXME_ON(d3d_shader
)) return;
150 GL_EXTCALL(glGetObjectParameterivARB(obj
,
151 GL_OBJECT_INFO_LOG_LENGTH_ARB
,
154 /* A size of 1 is just a null-terminated string, so the log should be bigger than
155 * that if there are errors. */
156 if (infologLength
> 1)
158 /* Fglrx doesn't terminate the string properly, but it tells us the proper length.
159 * So use HEAP_ZERO_MEMORY to avoid uninitialized bytes
161 infoLog
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, infologLength
);
162 GL_EXTCALL(glGetInfoLogARB(obj
, infologLength
, NULL
, infoLog
));
165 for(i
= 0; i
< sizeof(spam
) / sizeof(spam
[0]); i
++) {
166 if(strcmp(infoLog
, spam
[i
]) == 0) {
172 TRACE("Spam received from GLSL shader #%u: %s\n", obj
, debugstr_a(infoLog
));
174 FIXME("Error received from GLSL shader #%u: %s\n", obj
, debugstr_a(infoLog
));
176 HeapFree(GetProcessHeap(), 0, infoLog
);
181 * Loads (pixel shader) samplers
183 static void shader_glsl_load_psamplers(const WineD3D_GL_Info
*gl_info
, DWORD
*tex_unit_map
, GLhandleARB programId
)
187 char sampler_name
[20];
189 for (i
= 0; i
< MAX_FRAGMENT_SAMPLERS
; ++i
) {
190 snprintf(sampler_name
, sizeof(sampler_name
), "Psampler%d", i
);
191 name_loc
= GL_EXTCALL(glGetUniformLocationARB(programId
, sampler_name
));
192 if (name_loc
!= -1) {
193 DWORD mapped_unit
= tex_unit_map
[i
];
194 if (mapped_unit
!= WINED3D_UNMAPPED_STAGE
&& mapped_unit
< GL_LIMITS(fragment_samplers
))
196 TRACE("Loading %s for texture %d\n", sampler_name
, mapped_unit
);
197 GL_EXTCALL(glUniform1iARB(name_loc
, mapped_unit
));
198 checkGLcall("glUniform1iARB");
200 ERR("Trying to load sampler %s on unsupported unit %d\n", sampler_name
, mapped_unit
);
206 static void shader_glsl_load_vsamplers(const WineD3D_GL_Info
*gl_info
, DWORD
*tex_unit_map
, GLhandleARB programId
)
209 char sampler_name
[20];
212 for (i
= 0; i
< MAX_VERTEX_SAMPLERS
; ++i
) {
213 snprintf(sampler_name
, sizeof(sampler_name
), "Vsampler%d", i
);
214 name_loc
= GL_EXTCALL(glGetUniformLocationARB(programId
, sampler_name
));
215 if (name_loc
!= -1) {
216 DWORD mapped_unit
= tex_unit_map
[MAX_FRAGMENT_SAMPLERS
+ i
];
217 if (mapped_unit
!= WINED3D_UNMAPPED_STAGE
&& mapped_unit
< GL_LIMITS(combined_samplers
))
219 TRACE("Loading %s for texture %d\n", sampler_name
, mapped_unit
);
220 GL_EXTCALL(glUniform1iARB(name_loc
, mapped_unit
));
221 checkGLcall("glUniform1iARB");
223 ERR("Trying to load sampler %s on unsupported unit %d\n", sampler_name
, mapped_unit
);
229 static inline void walk_constant_heap(const WineD3D_GL_Info
*gl_info
, const float *constants
,
230 const GLint
*constant_locations
, const struct constant_heap
*heap
, unsigned char *stack
, DWORD version
)
233 unsigned int heap_idx
= 1;
236 if (heap
->entries
[heap_idx
].version
<= version
) return;
238 idx
= heap
->entries
[heap_idx
].idx
;
239 if (constant_locations
[idx
] != -1) GL_EXTCALL(glUniform4fvARB(constant_locations
[idx
], 1, &constants
[idx
* 4]));
240 stack
[stack_idx
] = HEAP_NODE_TRAVERSE_LEFT
;
242 while (stack_idx
>= 0)
244 /* Note that we fall through to the next case statement. */
245 switch(stack
[stack_idx
])
247 case HEAP_NODE_TRAVERSE_LEFT
:
249 unsigned int left_idx
= heap_idx
<< 1;
250 if (left_idx
< heap
->size
&& heap
->entries
[left_idx
].version
> version
)
253 idx
= heap
->entries
[heap_idx
].idx
;
254 if (constant_locations
[idx
] != -1)
255 GL_EXTCALL(glUniform4fvARB(constant_locations
[idx
], 1, &constants
[idx
* 4]));
257 stack
[stack_idx
++] = HEAP_NODE_TRAVERSE_RIGHT
;
258 stack
[stack_idx
] = HEAP_NODE_TRAVERSE_LEFT
;
263 case HEAP_NODE_TRAVERSE_RIGHT
:
265 unsigned int right_idx
= (heap_idx
<< 1) + 1;
266 if (right_idx
< heap
->size
&& heap
->entries
[right_idx
].version
> version
)
268 heap_idx
= right_idx
;
269 idx
= heap
->entries
[heap_idx
].idx
;
270 if (constant_locations
[idx
] != -1)
271 GL_EXTCALL(glUniform4fvARB(constant_locations
[idx
], 1, &constants
[idx
* 4]));
273 stack
[stack_idx
++] = HEAP_NODE_POP
;
274 stack
[stack_idx
] = HEAP_NODE_TRAVERSE_LEFT
;
287 checkGLcall("walk_constant_heap()");
290 static inline void apply_clamped_constant(const WineD3D_GL_Info
*gl_info
, GLint location
, const GLfloat
*data
)
292 GLfloat clamped_constant
[4];
294 if (location
== -1) return;
296 clamped_constant
[0] = data
[0] < -1.0f
? -1.0f
: data
[0] > 1.0 ? 1.0 : data
[0];
297 clamped_constant
[1] = data
[1] < -1.0f
? -1.0f
: data
[1] > 1.0 ? 1.0 : data
[1];
298 clamped_constant
[2] = data
[2] < -1.0f
? -1.0f
: data
[2] > 1.0 ? 1.0 : data
[2];
299 clamped_constant
[3] = data
[3] < -1.0f
? -1.0f
: data
[3] > 1.0 ? 1.0 : data
[3];
301 GL_EXTCALL(glUniform4fvARB(location
, 1, clamped_constant
));
304 static inline void walk_constant_heap_clamped(const WineD3D_GL_Info
*gl_info
, const float *constants
,
305 const GLint
*constant_locations
, const struct constant_heap
*heap
, unsigned char *stack
, DWORD version
)
308 unsigned int heap_idx
= 1;
311 if (heap
->entries
[heap_idx
].version
<= version
) return;
313 idx
= heap
->entries
[heap_idx
].idx
;
314 apply_clamped_constant(gl_info
, constant_locations
[idx
], &constants
[idx
* 4]);
315 stack
[stack_idx
] = HEAP_NODE_TRAVERSE_LEFT
;
317 while (stack_idx
>= 0)
319 /* Note that we fall through to the next case statement. */
320 switch(stack
[stack_idx
])
322 case HEAP_NODE_TRAVERSE_LEFT
:
324 unsigned int left_idx
= heap_idx
<< 1;
325 if (left_idx
< heap
->size
&& heap
->entries
[left_idx
].version
> version
)
328 idx
= heap
->entries
[heap_idx
].idx
;
329 apply_clamped_constant(gl_info
, constant_locations
[idx
], &constants
[idx
* 4]);
331 stack
[stack_idx
++] = HEAP_NODE_TRAVERSE_RIGHT
;
332 stack
[stack_idx
] = HEAP_NODE_TRAVERSE_LEFT
;
337 case HEAP_NODE_TRAVERSE_RIGHT
:
339 unsigned int right_idx
= (heap_idx
<< 1) + 1;
340 if (right_idx
< heap
->size
&& heap
->entries
[right_idx
].version
> version
)
342 heap_idx
= right_idx
;
343 idx
= heap
->entries
[heap_idx
].idx
;
344 apply_clamped_constant(gl_info
, constant_locations
[idx
], &constants
[idx
* 4]);
346 stack
[stack_idx
++] = HEAP_NODE_POP
;
347 stack
[stack_idx
] = HEAP_NODE_TRAVERSE_LEFT
;
360 checkGLcall("walk_constant_heap_clamped()");
363 /* Loads floating point constants (aka uniforms) into the currently set GLSL program. */
364 static void shader_glsl_load_constantsF(IWineD3DBaseShaderImpl
*This
, const WineD3D_GL_Info
*gl_info
,
365 const float *constants
, const GLint
*constant_locations
, const struct constant_heap
*heap
,
366 unsigned char *stack
, UINT version
)
368 const local_constant
*lconst
;
370 /* 1.X pshaders have the constants clamped to [-1;1] implicitly. */
371 if (WINED3DSHADER_VERSION_MAJOR(This
->baseShader
.reg_maps
.shader_version
) == 1
372 && shader_is_pshader_version(This
->baseShader
.reg_maps
.shader_version
))
373 walk_constant_heap_clamped(gl_info
, constants
, constant_locations
, heap
, stack
, version
);
375 walk_constant_heap(gl_info
, constants
, constant_locations
, heap
, stack
, version
);
377 if (!This
->baseShader
.load_local_constsF
)
379 TRACE("No need to load local float constants for this shader\n");
383 /* Immediate constants are clamped to [-1;1] at shader creation time if needed */
384 LIST_FOR_EACH_ENTRY(lconst
, &This
->baseShader
.constantsF
, local_constant
, entry
)
386 GLint location
= constant_locations
[lconst
->idx
];
387 /* We found this uniform name in the program - go ahead and send the data */
388 if (location
!= -1) GL_EXTCALL(glUniform4fvARB(location
, 1, (const GLfloat
*)lconst
->value
));
390 checkGLcall("glUniform4fvARB()");
393 /* Loads integer constants (aka uniforms) into the currently set GLSL program. */
394 static void shader_glsl_load_constantsI(IWineD3DBaseShaderImpl
*This
, const WineD3D_GL_Info
*gl_info
,
395 const GLint locations
[MAX_CONST_I
], const int *constants
, WORD constants_set
)
400 for (i
= 0; constants_set
; constants_set
>>= 1, ++i
)
402 if (!(constants_set
& 1)) continue;
404 TRACE_(d3d_constants
)("Loading constants %u: %i, %i, %i, %i\n",
405 i
, constants
[i
*4], constants
[i
*4+1], constants
[i
*4+2], constants
[i
*4+3]);
407 /* We found this uniform name in the program - go ahead and send the data */
408 GL_EXTCALL(glUniform4ivARB(locations
[i
], 1, &constants
[i
*4]));
409 checkGLcall("glUniform4ivARB");
412 /* Load immediate constants */
413 ptr
= list_head(&This
->baseShader
.constantsI
);
415 const struct local_constant
*lconst
= LIST_ENTRY(ptr
, const struct local_constant
, entry
);
416 unsigned int idx
= lconst
->idx
;
417 const GLint
*values
= (const GLint
*)lconst
->value
;
419 TRACE_(d3d_constants
)("Loading local constants %i: %i, %i, %i, %i\n", idx
,
420 values
[0], values
[1], values
[2], values
[3]);
422 /* We found this uniform name in the program - go ahead and send the data */
423 GL_EXTCALL(glUniform4ivARB(locations
[idx
], 1, values
));
424 checkGLcall("glUniform4ivARB");
425 ptr
= list_next(&This
->baseShader
.constantsI
, ptr
);
429 /* Loads boolean constants (aka uniforms) into the currently set GLSL program. */
430 static void shader_glsl_load_constantsB(IWineD3DBaseShaderImpl
*This
, const WineD3D_GL_Info
*gl_info
,
431 GLhandleARB programId
, const BOOL
*constants
, WORD constants_set
)
436 char is_pshader
= shader_is_pshader_version(This
->baseShader
.reg_maps
.shader_version
);
437 const char* prefix
= is_pshader
? "PB":"VB";
440 /* TODO: Benchmark and see if it would be beneficial to store the
441 * locations of the constants to avoid looking up each time */
442 for (i
= 0; constants_set
; constants_set
>>= 1, ++i
)
444 if (!(constants_set
& 1)) continue;
446 TRACE_(d3d_constants
)("Loading constants %i: %i;\n", i
, constants
[i
]);
448 /* TODO: Benchmark and see if it would be beneficial to store the
449 * locations of the constants to avoid looking up each time */
450 snprintf(tmp_name
, sizeof(tmp_name
), "%s[%i]", prefix
, i
);
451 tmp_loc
= GL_EXTCALL(glGetUniformLocationARB(programId
, tmp_name
));
454 /* We found this uniform name in the program - go ahead and send the data */
455 GL_EXTCALL(glUniform1ivARB(tmp_loc
, 1, &constants
[i
]));
456 checkGLcall("glUniform1ivARB");
460 /* Load immediate constants */
461 ptr
= list_head(&This
->baseShader
.constantsB
);
463 const struct local_constant
*lconst
= LIST_ENTRY(ptr
, const struct local_constant
, entry
);
464 unsigned int idx
= lconst
->idx
;
465 const GLint
*values
= (const GLint
*)lconst
->value
;
467 TRACE_(d3d_constants
)("Loading local constants %i: %i\n", idx
, values
[0]);
469 snprintf(tmp_name
, sizeof(tmp_name
), "%s[%i]", prefix
, idx
);
470 tmp_loc
= GL_EXTCALL(glGetUniformLocationARB(programId
, tmp_name
));
472 /* We found this uniform name in the program - go ahead and send the data */
473 GL_EXTCALL(glUniform1ivARB(tmp_loc
, 1, values
));
474 checkGLcall("glUniform1ivARB");
476 ptr
= list_next(&This
->baseShader
.constantsB
, ptr
);
480 static void reset_program_constant_version(void *value
, void *context
)
482 struct glsl_shader_prog_link
*entry
= value
;
483 entry
->constant_version
= 0;
487 * Loads the texture dimensions for NP2 fixup into the currently set GLSL program.
489 static void shader_glsl_load_np2fixup_constants(
490 IWineD3DDevice
* device
,
492 char useVertexShader
) {
494 const IWineD3DDeviceImpl
* deviceImpl
= (const IWineD3DDeviceImpl
*) device
;
495 const struct glsl_shader_prog_link
* prog
= ((struct shader_glsl_priv
*)(deviceImpl
->shader_priv
))->glsl_program
;
498 /* No GLSL program set - nothing to do. */
502 if (!usePixelShader
) {
503 /* NP2 texcoord fixup is (currently) only done for pixelshaders. */
507 if (prog
->ps_args
.np2_fixup
) {
509 UINT fixup
= prog
->ps_args
.np2_fixup
;
510 const WineD3D_GL_Info
* gl_info
= &deviceImpl
->adapter
->gl_info
;
511 const IWineD3DStateBlockImpl
* stateBlock
= (const IWineD3DStateBlockImpl
*) deviceImpl
->stateBlock
;
513 for (i
= 0; fixup
; fixup
>>= 1, ++i
) {
514 if (-1 != prog
->np2Fixup_location
[i
]) {
515 const IWineD3DBaseTextureImpl
* const tex
= (const IWineD3DBaseTextureImpl
*) stateBlock
->textures
[i
];
517 FIXME("Nonexistent texture is flagged for NP2 texcoord fixup\n");
520 const float tex_dim
[2] = {tex
->baseTexture
.pow2Matrix
[0], tex
->baseTexture
.pow2Matrix
[5]};
521 GL_EXTCALL(glUniform2fvARB(prog
->np2Fixup_location
[i
], 1, tex_dim
));
529 * Loads the app-supplied constants into the currently set GLSL program.
531 static void shader_glsl_load_constants(
532 IWineD3DDevice
* device
,
534 char useVertexShader
) {
536 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) device
;
537 struct shader_glsl_priv
*priv
= deviceImpl
->shader_priv
;
538 IWineD3DStateBlockImpl
* stateBlock
= deviceImpl
->stateBlock
;
539 const WineD3D_GL_Info
*gl_info
= &deviceImpl
->adapter
->gl_info
;
541 GLhandleARB programId
;
542 struct glsl_shader_prog_link
*prog
= priv
->glsl_program
;
543 UINT constant_version
;
547 /* No GLSL program set - nothing to do. */
550 programId
= prog
->programId
;
551 constant_version
= prog
->constant_version
;
553 if (useVertexShader
) {
554 IWineD3DBaseShaderImpl
* vshader
= (IWineD3DBaseShaderImpl
*) stateBlock
->vertexShader
;
556 /* Load DirectX 9 float constants/uniforms for vertex shader */
557 shader_glsl_load_constantsF(vshader
, gl_info
, stateBlock
->vertexShaderConstantF
,
558 prog
->vuniformF_locations
, &priv
->vconst_heap
, priv
->stack
, constant_version
);
560 /* Load DirectX 9 integer constants/uniforms for vertex shader */
561 if(vshader
->baseShader
.uses_int_consts
) {
562 shader_glsl_load_constantsI(vshader
, gl_info
, prog
->vuniformI_locations
,
563 stateBlock
->vertexShaderConstantI
, stateBlock
->changed
.vertexShaderConstantsI
);
566 /* Load DirectX 9 boolean constants/uniforms for vertex shader */
567 if(vshader
->baseShader
.uses_bool_consts
) {
568 shader_glsl_load_constantsB(vshader
, gl_info
, programId
,
569 stateBlock
->vertexShaderConstantB
, stateBlock
->changed
.vertexShaderConstantsB
);
572 /* Upload the position fixup params */
573 GL_EXTCALL(glUniform4fvARB(prog
->posFixup_location
, 1, &deviceImpl
->posFixup
[0]));
574 checkGLcall("glUniform4fvARB");
577 if (usePixelShader
) {
579 IWineD3DBaseShaderImpl
* pshader
= (IWineD3DBaseShaderImpl
*) stateBlock
->pixelShader
;
581 /* Load DirectX 9 float constants/uniforms for pixel shader */
582 shader_glsl_load_constantsF(pshader
, gl_info
, stateBlock
->pixelShaderConstantF
,
583 prog
->puniformF_locations
, &priv
->pconst_heap
, priv
->stack
, constant_version
);
585 /* Load DirectX 9 integer constants/uniforms for pixel shader */
586 if(pshader
->baseShader
.uses_int_consts
) {
587 shader_glsl_load_constantsI(pshader
, gl_info
, prog
->puniformI_locations
,
588 stateBlock
->pixelShaderConstantI
, stateBlock
->changed
.pixelShaderConstantsI
);
591 /* Load DirectX 9 boolean constants/uniforms for pixel shader */
592 if(pshader
->baseShader
.uses_bool_consts
) {
593 shader_glsl_load_constantsB(pshader
, gl_info
, programId
,
594 stateBlock
->pixelShaderConstantB
, stateBlock
->changed
.pixelShaderConstantsB
);
597 /* Upload the environment bump map matrix if needed. The needsbumpmat member specifies the texture stage to load the matrix from.
598 * It can't be 0 for a valid texbem instruction.
600 for(i
= 0; i
< ((IWineD3DPixelShaderImpl
*) pshader
)->numbumpenvmatconsts
; i
++) {
601 IWineD3DPixelShaderImpl
*ps
= (IWineD3DPixelShaderImpl
*) pshader
;
602 int stage
= ps
->luminanceconst
[i
].texunit
;
604 const float *data
= (const float *)&stateBlock
->textureState
[(int)ps
->bumpenvmatconst
[i
].texunit
][WINED3DTSS_BUMPENVMAT00
];
605 GL_EXTCALL(glUniformMatrix2fvARB(prog
->bumpenvmat_location
[i
], 1, 0, data
));
606 checkGLcall("glUniformMatrix2fvARB");
608 /* texbeml needs the luminance scale and offset too. If texbeml is used, needsbumpmat
609 * is set too, so we can check that in the needsbumpmat check
611 if(ps
->baseShader
.reg_maps
.luminanceparams
[stage
]) {
612 const GLfloat
*scale
= (const GLfloat
*)&stateBlock
->textureState
[stage
][WINED3DTSS_BUMPENVLSCALE
];
613 const GLfloat
*offset
= (const GLfloat
*)&stateBlock
->textureState
[stage
][WINED3DTSS_BUMPENVLOFFSET
];
615 GL_EXTCALL(glUniform1fvARB(prog
->luminancescale_location
[i
], 1, scale
));
616 checkGLcall("glUniform1fvARB");
617 GL_EXTCALL(glUniform1fvARB(prog
->luminanceoffset_location
[i
], 1, offset
));
618 checkGLcall("glUniform1fvARB");
622 if(((IWineD3DPixelShaderImpl
*) pshader
)->vpos_uniform
) {
623 float correction_params
[4];
624 if(deviceImpl
->render_offscreen
) {
625 correction_params
[0] = 0.0;
626 correction_params
[1] = 1.0;
628 /* position is window relative, not viewport relative */
629 correction_params
[0] = ((IWineD3DSurfaceImpl
*) deviceImpl
->render_targets
[0])->currentDesc
.Height
;
630 correction_params
[1] = -1.0;
632 GL_EXTCALL(glUniform4fvARB(prog
->ycorrection_location
, 1, correction_params
));
636 if (priv
->next_constant_version
== UINT_MAX
)
638 TRACE("Max constant version reached, resetting to 0.\n");
639 hash_table_for_each_entry(priv
->glsl_program_lookup
, reset_program_constant_version
, NULL
);
640 priv
->next_constant_version
= 1;
644 prog
->constant_version
= priv
->next_constant_version
++;
648 static inline void update_heap_entry(struct constant_heap
*heap
, unsigned int idx
,
649 unsigned int heap_idx
, DWORD new_version
)
651 struct constant_entry
*entries
= heap
->entries
;
652 unsigned int *positions
= heap
->positions
;
653 unsigned int parent_idx
;
657 parent_idx
= heap_idx
>> 1;
659 if (new_version
<= entries
[parent_idx
].version
) break;
661 entries
[heap_idx
] = entries
[parent_idx
];
662 positions
[entries
[parent_idx
].idx
] = heap_idx
;
663 heap_idx
= parent_idx
;
666 entries
[heap_idx
].version
= new_version
;
667 entries
[heap_idx
].idx
= idx
;
668 positions
[idx
] = heap_idx
;
671 static void shader_glsl_update_float_vertex_constants(IWineD3DDevice
*iface
, UINT start
, UINT count
)
673 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
674 struct shader_glsl_priv
*priv
= This
->shader_priv
;
675 struct constant_heap
*heap
= &priv
->vconst_heap
;
678 for (i
= start
; i
< count
+ start
; ++i
)
680 if (!This
->stateBlock
->changed
.vertexShaderConstantsF
[i
])
681 update_heap_entry(heap
, i
, heap
->size
++, priv
->next_constant_version
);
683 update_heap_entry(heap
, i
, heap
->positions
[i
], priv
->next_constant_version
);
687 static void shader_glsl_update_float_pixel_constants(IWineD3DDevice
*iface
, UINT start
, UINT count
)
689 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
690 struct shader_glsl_priv
*priv
= This
->shader_priv
;
691 struct constant_heap
*heap
= &priv
->pconst_heap
;
694 for (i
= start
; i
< count
+ start
; ++i
)
696 if (!This
->stateBlock
->changed
.pixelShaderConstantsF
[i
])
697 update_heap_entry(heap
, i
, heap
->size
++, priv
->next_constant_version
);
699 update_heap_entry(heap
, i
, heap
->positions
[i
], priv
->next_constant_version
);
703 /** Generate the variable & register declarations for the GLSL output target */
704 static void shader_generate_glsl_declarations(IWineD3DBaseShader
*iface
, const shader_reg_maps
*reg_maps
,
705 SHADER_BUFFER
*buffer
, const WineD3D_GL_Info
*gl_info
,
706 const struct ps_compile_args
*ps_args
)
708 IWineD3DBaseShaderImpl
* This
= (IWineD3DBaseShaderImpl
*) iface
;
709 IWineD3DDeviceImpl
*device
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
710 DWORD shader_version
= reg_maps
->shader_version
;
711 unsigned int i
, extra_constants_needed
= 0;
712 const local_constant
*lconst
;
714 /* There are some minor differences between pixel and vertex shaders */
715 char pshader
= shader_is_pshader_version(shader_version
);
716 char prefix
= pshader
? 'P' : 'V';
718 /* Prototype the subroutines */
719 for (i
= 0; i
< This
->baseShader
.limits
.label
; i
++) {
720 if (reg_maps
->labels
[i
])
721 shader_addline(buffer
, "void subroutine%u();\n", i
);
724 /* Declare the constants (aka uniforms) */
725 if (This
->baseShader
.limits
.constant_float
> 0) {
726 unsigned max_constantsF
;
728 max_constantsF
= GL_LIMITS(pshader_constantsF
) - (MAX_CONST_B
/ 4) - MAX_CONST_I
- 2;
729 max_constantsF
= min(This
->baseShader
.limits
.constant_float
, max_constantsF
);
731 /* Subtract the other potential uniforms from the max available (bools, ints, and 1 row of projection matrix) */
732 max_constantsF
= GL_LIMITS(vshader_constantsF
) - (MAX_CONST_B
/ 4) - MAX_CONST_I
- 1;
733 max_constantsF
= min(This
->baseShader
.limits
.constant_float
, max_constantsF
);
735 shader_addline(buffer
, "uniform vec4 %cC[%u];\n", prefix
, max_constantsF
);
738 if (This
->baseShader
.limits
.constant_int
> 0)
739 shader_addline(buffer
, "uniform ivec4 %cI[%u];\n", prefix
, This
->baseShader
.limits
.constant_int
);
741 if (This
->baseShader
.limits
.constant_bool
> 0)
742 shader_addline(buffer
, "uniform bool %cB[%u];\n", prefix
, This
->baseShader
.limits
.constant_bool
);
745 shader_addline(buffer
, "uniform vec4 posFixup;\n");
746 /* Predeclaration; This function is added at link time based on the pixel shader.
747 * VS 3.0 shaders have an array OUT[] the shader writes to, earlier versions don't have
748 * that. We know the input to the reorder function at vertex shader compile time, so
749 * we can deal with that. The reorder function for a 1.x and 2.x vertex shader can just
750 * read gl_FrontColor. The output depends on the pixel shader. The reorder function for a
751 * 1.x and 2.x pshader or for fixed function will write gl_FrontColor, and for a 3.0 shader
752 * it will write to the varying array. Here we depend on the shader optimizer on sorting that
753 * out. The nvidia driver only does that if the parameter is inout instead of out, hence the
756 if (shader_version
>= WINED3DVS_VERSION(3, 0))
758 shader_addline(buffer
, "void order_ps_input(in vec4[%u]);\n", MAX_REG_OUTPUT
);
760 shader_addline(buffer
, "void order_ps_input();\n");
763 IWineD3DPixelShaderImpl
*ps_impl
= (IWineD3DPixelShaderImpl
*) This
;
765 ps_impl
->numbumpenvmatconsts
= 0;
766 for(i
= 0; i
< (sizeof(reg_maps
->bumpmat
) / sizeof(reg_maps
->bumpmat
[0])); i
++) {
767 if(!reg_maps
->bumpmat
[i
]) {
771 ps_impl
->bumpenvmatconst
[(int) ps_impl
->numbumpenvmatconsts
].texunit
= i
;
772 shader_addline(buffer
, "uniform mat2 bumpenvmat%d;\n", i
);
774 if(reg_maps
->luminanceparams
) {
775 ps_impl
->luminanceconst
[(int) ps_impl
->numbumpenvmatconsts
].texunit
= i
;
776 shader_addline(buffer
, "uniform float luminancescale%d;\n", i
);
777 shader_addline(buffer
, "uniform float luminanceoffset%d;\n", i
);
778 extra_constants_needed
++;
780 ps_impl
->luminanceconst
[(int) ps_impl
->numbumpenvmatconsts
].texunit
= -1;
783 extra_constants_needed
++;
784 ps_impl
->numbumpenvmatconsts
++;
787 if(ps_args
->srgb_correction
) {
788 shader_addline(buffer
, "const vec4 srgb_mul_low = vec4(%f, %f, %f, %f);\n",
789 srgb_mul_low
, srgb_mul_low
, srgb_mul_low
, srgb_mul_low
);
790 shader_addline(buffer
, "const vec4 srgb_comparison = vec4(%f, %f, %f, %f);\n",
791 srgb_cmp
, srgb_cmp
, srgb_cmp
, srgb_cmp
);
793 if(reg_maps
->vpos
|| reg_maps
->usesdsy
) {
794 if(This
->baseShader
.limits
.constant_float
+ extra_constants_needed
+ 1 < GL_LIMITS(pshader_constantsF
)) {
795 shader_addline(buffer
, "uniform vec4 ycorrection;\n");
796 ((IWineD3DPixelShaderImpl
*) This
)->vpos_uniform
= 1;
797 extra_constants_needed
++;
799 /* This happens because we do not have proper tracking of the constant registers that are
800 * actually used, only the max limit of the shader version
802 FIXME("Cannot find a free uniform for vpos correction params\n");
803 shader_addline(buffer
, "const vec4 ycorrection = vec4(%f, %f, 0.0, 0.0);\n",
804 device
->render_offscreen
? 0.0 : ((IWineD3DSurfaceImpl
*) device
->render_targets
[0])->currentDesc
.Height
,
805 device
->render_offscreen
? 1.0 : -1.0);
807 shader_addline(buffer
, "vec4 vpos;\n");
811 /* Declare texture samplers */
812 for (i
= 0; i
< This
->baseShader
.limits
.sampler
; i
++) {
813 if (reg_maps
->samplers
[i
]) {
815 DWORD stype
= reg_maps
->samplers
[i
] & WINED3DSP_TEXTURETYPE_MASK
;
819 shader_addline(buffer
, "uniform sampler1D %csampler%u;\n", prefix
, i
);
822 if(device
->stateBlock
->textures
[i
] &&
823 IWineD3DBaseTexture_GetTextureDimensions(device
->stateBlock
->textures
[i
]) == GL_TEXTURE_RECTANGLE_ARB
) {
824 shader_addline(buffer
, "uniform sampler2DRect %csampler%u;\n", prefix
, i
);
826 shader_addline(buffer
, "uniform sampler2D %csampler%u;\n", prefix
, i
);
829 if (pshader
&& ps_args
->np2_fixup
& (1 << i
))
831 /* NP2/RECT textures in OpenGL use texcoords in the range [0,width]x[0,height]
832 * while D3D has them in the (normalized) [0,1]x[0,1] range.
833 * samplerNP2Fixup stores texture dimensions and is updated through
834 * shader_glsl_load_np2fixup_constants when the sampler changes. */
835 shader_addline(buffer
, "uniform vec2 %csamplerNP2Fixup%u;\n", prefix
, i
);
838 case WINED3DSTT_CUBE
:
839 shader_addline(buffer
, "uniform samplerCube %csampler%u;\n", prefix
, i
);
841 case WINED3DSTT_VOLUME
:
842 shader_addline(buffer
, "uniform sampler3D %csampler%u;\n", prefix
, i
);
845 shader_addline(buffer
, "uniform unsupported_sampler %csampler%u;\n", prefix
, i
);
846 FIXME("Unrecognized sampler type: %#x\n", stype
);
852 /* Declare address variables */
853 for (i
= 0; i
< This
->baseShader
.limits
.address
; i
++) {
854 if (reg_maps
->address
[i
])
855 shader_addline(buffer
, "ivec4 A%d;\n", i
);
858 /* Declare texture coordinate temporaries and initialize them */
859 for (i
= 0; i
< This
->baseShader
.limits
.texcoord
; i
++) {
860 if (reg_maps
->texcoord
[i
])
861 shader_addline(buffer
, "vec4 T%u = gl_TexCoord[%u];\n", i
, i
);
864 /* Declare input register varyings. Only pixel shader, vertex shaders have that declared in the
865 * helper function shader that is linked in at link time
867 if (pshader
&& shader_version
>= WINED3DPS_VERSION(3, 0))
869 if (use_vs(device
->stateBlock
))
871 shader_addline(buffer
, "varying vec4 IN[%u];\n", GL_LIMITS(glsl_varyings
) / 4);
873 /* TODO: Write a replacement shader for the fixed function vertex pipeline, so this isn't needed.
874 * For fixed function vertex processing + 3.0 pixel shader we need a separate function in the
875 * pixel shader that reads the fixed function color into the packed input registers.
877 shader_addline(buffer
, "vec4 IN[%u];\n", GL_LIMITS(glsl_varyings
) / 4);
881 /* Declare output register temporaries */
882 if(This
->baseShader
.limits
.packed_output
) {
883 shader_addline(buffer
, "vec4 OUT[%u];\n", This
->baseShader
.limits
.packed_output
);
886 /* Declare temporary variables */
887 for(i
= 0; i
< This
->baseShader
.limits
.temporary
; i
++) {
888 if (reg_maps
->temporary
[i
])
889 shader_addline(buffer
, "vec4 R%u;\n", i
);
892 /* Declare attributes */
893 for (i
= 0; i
< This
->baseShader
.limits
.attributes
; i
++) {
894 if (reg_maps
->attributes
[i
])
895 shader_addline(buffer
, "attribute vec4 attrib%i;\n", i
);
898 /* Declare loop registers aLx */
899 for (i
= 0; i
< reg_maps
->loop_depth
; i
++) {
900 shader_addline(buffer
, "int aL%u;\n", i
);
901 shader_addline(buffer
, "int tmpInt%u;\n", i
);
904 /* Temporary variables for matrix operations */
905 shader_addline(buffer
, "vec4 tmp0;\n");
906 shader_addline(buffer
, "vec4 tmp1;\n");
908 /* Local constants use a different name so they can be loaded once at shader link time
909 * They can't be hardcoded into the shader text via LC = {x, y, z, w}; because the
910 * float -> string conversion can cause precision loss.
912 if(!This
->baseShader
.load_local_constsF
) {
913 LIST_FOR_EACH_ENTRY(lconst
, &This
->baseShader
.constantsF
, local_constant
, entry
) {
914 shader_addline(buffer
, "uniform vec4 %cLC%u;\n", prefix
, lconst
->idx
);
918 /* Start the main program */
919 shader_addline(buffer
, "void main() {\n");
920 if(pshader
&& reg_maps
->vpos
) {
921 /* DirectX apps expect integer values, while OpenGL drivers add approximately 0.5. This causes
922 * off-by-one problems as spotted by the vPos d3d9 visual test. Unfortunately the ATI cards do
923 * not add exactly 0.5, but rather something like 0.49999999 or 0.50000001, which still causes
924 * precision troubles when we just substract 0.5.
926 * To deal with that just floor() the position. This will eliminate the fraction on all cards.
928 * TODO: Test how that behaves with multisampling once we can enable multisampling in winex11.
930 * An advantage of floor is that it works even if the driver doesn't add 1/2. It is somewhat
931 * questionable if 1.5, 2.5, ... are the proper values to return in gl_FragCoord, even though
932 * coordinates specify the pixel centers instead of the pixel corners. This code will behave
933 * correctly on drivers that returns integer values.
935 shader_addline(buffer
, "vpos = floor(vec4(0, ycorrection[0], 0, 0) + gl_FragCoord * vec4(1, ycorrection[1], 1, 1));\n");
939 /*****************************************************************************
940 * Functions to generate GLSL strings from DirectX Shader bytecode begin here.
942 * For more information, see http://wiki.winehq.org/DirectX-Shaders
943 ****************************************************************************/
946 static void shader_glsl_add_src_param(const struct wined3d_shader_instruction
*ins
,
947 const struct wined3d_shader_src_param
*wined3d_src
, DWORD mask
, glsl_src_param_t
*glsl_src
);
949 /** Used for opcode modifiers - They multiply the result by the specified amount */
950 static const char * const shift_glsl_tab
[] = {
952 "2.0 * ", /* 1 (x2) */
953 "4.0 * ", /* 2 (x4) */
954 "8.0 * ", /* 3 (x8) */
955 "16.0 * ", /* 4 (x16) */
956 "32.0 * ", /* 5 (x32) */
963 "0.0625 * ", /* 12 (d16) */
964 "0.125 * ", /* 13 (d8) */
965 "0.25 * ", /* 14 (d4) */
966 "0.5 * " /* 15 (d2) */
969 /* Generate a GLSL parameter that does the input modifier computation and return the input register/mask to use */
970 static void shader_glsl_gen_modifier (
973 const char *in_regswizzle
,
978 switch (instr
& WINED3DSP_SRCMOD_MASK
) {
979 case WINED3DSPSM_DZ
: /* Need to handle this in the instructions itself (texld & texcrd). */
981 case WINED3DSPSM_NONE
:
982 sprintf(out_str
, "%s%s", in_reg
, in_regswizzle
);
984 case WINED3DSPSM_NEG
:
985 sprintf(out_str
, "-%s%s", in_reg
, in_regswizzle
);
987 case WINED3DSPSM_NOT
:
988 sprintf(out_str
, "!%s%s", in_reg
, in_regswizzle
);
990 case WINED3DSPSM_BIAS
:
991 sprintf(out_str
, "(%s%s - vec4(0.5)%s)", in_reg
, in_regswizzle
, in_regswizzle
);
993 case WINED3DSPSM_BIASNEG
:
994 sprintf(out_str
, "-(%s%s - vec4(0.5)%s)", in_reg
, in_regswizzle
, in_regswizzle
);
996 case WINED3DSPSM_SIGN
:
997 sprintf(out_str
, "(2.0 * (%s%s - 0.5))", in_reg
, in_regswizzle
);
999 case WINED3DSPSM_SIGNNEG
:
1000 sprintf(out_str
, "-(2.0 * (%s%s - 0.5))", in_reg
, in_regswizzle
);
1002 case WINED3DSPSM_COMP
:
1003 sprintf(out_str
, "(1.0 - %s%s)", in_reg
, in_regswizzle
);
1005 case WINED3DSPSM_X2
:
1006 sprintf(out_str
, "(2.0 * %s%s)", in_reg
, in_regswizzle
);
1008 case WINED3DSPSM_X2NEG
:
1009 sprintf(out_str
, "-(2.0 * %s%s)", in_reg
, in_regswizzle
);
1011 case WINED3DSPSM_ABS
:
1012 sprintf(out_str
, "abs(%s%s)", in_reg
, in_regswizzle
);
1014 case WINED3DSPSM_ABSNEG
:
1015 sprintf(out_str
, "-abs(%s%s)", in_reg
, in_regswizzle
);
1018 FIXME("Unhandled modifier %u\n", (instr
& WINED3DSP_SRCMOD_MASK
));
1019 sprintf(out_str
, "%s%s", in_reg
, in_regswizzle
);
1023 /** Writes the GLSL variable name that corresponds to the register that the
1024 * DX opcode parameter is trying to access */
1025 static void shader_glsl_get_register_name(WINED3DSHADER_PARAM_REGISTER_TYPE register_type
, UINT register_idx
,
1026 const struct wined3d_shader_src_param
*rel_addr
, char *register_name
, BOOL
*is_color
,
1027 const struct wined3d_shader_instruction
*ins
)
1029 /* oPos, oFog and oPts in D3D */
1030 static const char * const hwrastout_reg_names
[] = { "gl_Position", "gl_FogFragCoord", "gl_PointSize" };
1032 IWineD3DBaseShaderImpl
*This
= (IWineD3DBaseShaderImpl
*)ins
->ctx
->shader
;
1033 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
1034 const WineD3D_GL_Info
* gl_info
= &deviceImpl
->adapter
->gl_info
;
1035 DWORD shader_version
= This
->baseShader
.reg_maps
.shader_version
;
1036 char pshader
= shader_is_pshader_version(shader_version
);
1040 switch (register_type
)
1042 case WINED3DSPR_TEMP
:
1043 sprintf(register_name
, "R%u", register_idx
);
1045 case WINED3DSPR_INPUT
:
1047 /* Pixel shaders >= 3.0 */
1048 if (WINED3DSHADER_VERSION_MAJOR(shader_version
) >= 3)
1050 DWORD in_count
= GL_LIMITS(glsl_varyings
) / 4;
1054 glsl_src_param_t rel_param
;
1055 shader_glsl_add_src_param(ins
, rel_addr
, WINED3DSP_WRITEMASK_0
, &rel_param
);
1057 /* Removing a + 0 would be an obvious optimization, but macos doesn't see the NOP
1060 if (((IWineD3DPixelShaderImpl
*)This
)->input_reg_map
[register_idx
])
1062 if (((IWineD3DPixelShaderImpl
*)This
)->declared_in_count
> in_count
) {
1063 sprintf(register_name
, "((%s + %u) > %d ? (%s + %u) > %d ? gl_SecondaryColor : gl_Color : IN[%s + %u])",
1064 rel_param
.param_str
, ((IWineD3DPixelShaderImpl
*)This
)->input_reg_map
[register_idx
], in_count
- 1,
1065 rel_param
.param_str
, ((IWineD3DPixelShaderImpl
*)This
)->input_reg_map
[register_idx
], in_count
,
1066 rel_param
.param_str
, ((IWineD3DPixelShaderImpl
*)This
)->input_reg_map
[register_idx
]);
1068 sprintf(register_name
, "IN[%s + %u]", rel_param
.param_str
,
1069 ((IWineD3DPixelShaderImpl
*)This
)->input_reg_map
[register_idx
]);
1072 if (((IWineD3DPixelShaderImpl
*)This
)->declared_in_count
> in_count
) {
1073 sprintf(register_name
, "((%s) > %d ? (%s) > %d ? gl_SecondaryColor : gl_Color : IN[%s])",
1074 rel_param
.param_str
, in_count
- 1,
1075 rel_param
.param_str
, in_count
,
1076 rel_param
.param_str
);
1078 sprintf(register_name
, "IN[%s]", rel_param
.param_str
);
1082 DWORD idx
= ((IWineD3DPixelShaderImpl
*) This
)->input_reg_map
[register_idx
];
1083 if (idx
== in_count
) {
1084 sprintf(register_name
, "gl_Color");
1085 } else if (idx
== in_count
+ 1) {
1086 sprintf(register_name
, "gl_SecondaryColor");
1088 sprintf(register_name
, "IN[%u]", idx
);
1092 if (register_idx
== 0)
1093 strcpy(register_name
, "gl_Color");
1095 strcpy(register_name
, "gl_SecondaryColor");
1098 if (((IWineD3DVertexShaderImpl
*)This
)->cur_args
->swizzle_map
& (1 << register_idx
)) *is_color
= TRUE
;
1099 sprintf(register_name
, "attrib%u", register_idx
);
1102 case WINED3DSPR_CONST
:
1104 const char prefix
= pshader
? 'P':'V';
1106 /* Relative addressing */
1109 glsl_src_param_t rel_param
;
1110 shader_glsl_add_src_param(ins
, rel_addr
, WINED3DSP_WRITEMASK_0
, &rel_param
);
1112 sprintf(register_name
, "%cC[%s + %u]", prefix
, rel_param
.param_str
, register_idx
);
1114 sprintf(register_name
, "%cC[%s]", prefix
, rel_param
.param_str
);
1116 if (shader_constant_is_local(This
, register_idx
))
1118 sprintf(register_name
, "%cLC%u", prefix
, register_idx
);
1120 sprintf(register_name
, "%cC[%u]", prefix
, register_idx
);
1126 case WINED3DSPR_CONSTINT
:
1128 sprintf(register_name
, "PI[%u]", register_idx
);
1130 sprintf(register_name
, "VI[%u]", register_idx
);
1132 case WINED3DSPR_CONSTBOOL
:
1134 sprintf(register_name
, "PB[%u]", register_idx
);
1136 sprintf(register_name
, "VB[%u]", register_idx
);
1138 case WINED3DSPR_TEXTURE
: /* case WINED3DSPR_ADDR: */
1140 sprintf(register_name
, "T%u", register_idx
);
1142 sprintf(register_name
, "A%u", register_idx
);
1145 case WINED3DSPR_LOOP
:
1146 sprintf(register_name
, "aL%u", This
->baseShader
.cur_loop_regno
- 1);
1148 case WINED3DSPR_SAMPLER
:
1150 sprintf(register_name
, "Psampler%u", register_idx
);
1152 sprintf(register_name
, "Vsampler%u", register_idx
);
1154 case WINED3DSPR_COLOROUT
:
1155 if (register_idx
>= GL_LIMITS(buffers
))
1156 WARN("Write to render target %u, only %d supported\n", register_idx
, 4);
1158 if (GL_SUPPORT(ARB_DRAW_BUFFERS
)) {
1159 sprintf(register_name
, "gl_FragData[%u]", register_idx
);
1160 } else { /* On older cards with GLSL support like the GeforceFX there's only one buffer. */
1161 sprintf(register_name
, "gl_FragColor");
1164 case WINED3DSPR_RASTOUT
:
1165 sprintf(register_name
, "%s", hwrastout_reg_names
[register_idx
]);
1167 case WINED3DSPR_DEPTHOUT
:
1168 sprintf(register_name
, "gl_FragDepth");
1170 case WINED3DSPR_ATTROUT
:
1171 if (register_idx
== 0)
1173 sprintf(register_name
, "gl_FrontColor");
1175 sprintf(register_name
, "gl_FrontSecondaryColor");
1178 case WINED3DSPR_TEXCRDOUT
:
1179 /* Vertex shaders >= 3.0: WINED3DSPR_OUTPUT */
1180 if (WINED3DSHADER_VERSION_MAJOR(shader_version
) >= 3) sprintf(register_name
, "OUT[%u]", register_idx
);
1181 else sprintf(register_name
, "gl_TexCoord[%u]", register_idx
);
1183 case WINED3DSPR_MISCTYPE
:
1184 if (register_idx
== 0)
1187 sprintf(register_name
, "vpos");
1189 else if (register_idx
== 1)
1191 /* Note that gl_FrontFacing is a bool, while vFace is
1192 * a float for which the sign determines front/back
1194 sprintf(register_name
, "(gl_FrontFacing ? 1.0 : -1.0)");
1196 FIXME("Unhandled misctype register %d\n", register_idx
);
1197 sprintf(register_name
, "unrecognized_register");
1201 FIXME("Unhandled register name Type(%d)\n", register_type
);
1202 sprintf(register_name
, "unrecognized_register");
1207 static void shader_glsl_write_mask_to_str(DWORD write_mask
, char *str
)
1210 if (write_mask
& WINED3DSP_WRITEMASK_0
) *str
++ = 'x';
1211 if (write_mask
& WINED3DSP_WRITEMASK_1
) *str
++ = 'y';
1212 if (write_mask
& WINED3DSP_WRITEMASK_2
) *str
++ = 'z';
1213 if (write_mask
& WINED3DSP_WRITEMASK_3
) *str
++ = 'w';
1217 /* Get the GLSL write mask for the destination register */
1218 static DWORD
shader_glsl_get_write_mask(const struct wined3d_shader_dst_param
*param
, char *write_mask
)
1220 DWORD mask
= param
->write_mask
;
1222 if (shader_is_scalar(param
->register_type
, param
->register_idx
))
1224 mask
= WINED3DSP_WRITEMASK_0
;
1229 shader_glsl_write_mask_to_str(mask
, write_mask
);
1235 static unsigned int shader_glsl_get_write_mask_size(DWORD write_mask
) {
1236 unsigned int size
= 0;
1238 if (write_mask
& WINED3DSP_WRITEMASK_0
) ++size
;
1239 if (write_mask
& WINED3DSP_WRITEMASK_1
) ++size
;
1240 if (write_mask
& WINED3DSP_WRITEMASK_2
) ++size
;
1241 if (write_mask
& WINED3DSP_WRITEMASK_3
) ++size
;
1246 static void shader_glsl_get_swizzle(const DWORD param
, BOOL fixup
, DWORD mask
, char *swizzle_str
) {
1247 /* For registers of type WINED3DDECLTYPE_D3DCOLOR, data is stored as "bgra",
1248 * but addressed as "rgba". To fix this we need to swap the register's x
1249 * and z components. */
1250 DWORD swizzle
= (param
& WINED3DSP_SWIZZLE_MASK
) >> WINED3DSP_SWIZZLE_SHIFT
;
1251 const char *swizzle_chars
= fixup
? "zyxw" : "xyzw";
1252 char *ptr
= swizzle_str
;
1254 if (!shader_is_scalar(shader_get_regtype(param
), param
& WINED3DSP_REGNUM_MASK
))
1257 /* swizzle bits fields: wwzzyyxx */
1258 if (mask
& WINED3DSP_WRITEMASK_0
) *ptr
++ = swizzle_chars
[swizzle
& 0x03];
1259 if (mask
& WINED3DSP_WRITEMASK_1
) *ptr
++ = swizzle_chars
[(swizzle
>> 2) & 0x03];
1260 if (mask
& WINED3DSP_WRITEMASK_2
) *ptr
++ = swizzle_chars
[(swizzle
>> 4) & 0x03];
1261 if (mask
& WINED3DSP_WRITEMASK_3
) *ptr
++ = swizzle_chars
[(swizzle
>> 6) & 0x03];
1267 /* From a given parameter token, generate the corresponding GLSL string.
1268 * Also, return the actual register name and swizzle in case the
1269 * caller needs this information as well. */
1270 static void shader_glsl_add_src_param(const struct wined3d_shader_instruction
*ins
,
1271 const struct wined3d_shader_src_param
*wined3d_src
, DWORD mask
, glsl_src_param_t
*glsl_src
)
1273 BOOL is_color
= FALSE
;
1274 char swizzle_str
[6];
1276 glsl_src
->reg_name
[0] = '\0';
1277 glsl_src
->param_str
[0] = '\0';
1278 swizzle_str
[0] = '\0';
1280 shader_glsl_get_register_name(wined3d_src
->register_type
, wined3d_src
->register_idx
,
1281 wined3d_src
->rel_addr
, glsl_src
->reg_name
, &is_color
, ins
);
1283 shader_glsl_get_swizzle(wined3d_src
->token
, is_color
, mask
, swizzle_str
);
1284 shader_glsl_gen_modifier(wined3d_src
->token
, glsl_src
->reg_name
, swizzle_str
, glsl_src
->param_str
);
1287 /* From a given parameter token, generate the corresponding GLSL string.
1288 * Also, return the actual register name and swizzle in case the
1289 * caller needs this information as well. */
1290 static DWORD
shader_glsl_add_dst_param(const struct wined3d_shader_instruction
*ins
,
1291 const struct wined3d_shader_dst_param
*wined3d_dst
, glsl_dst_param_t
*glsl_dst
)
1293 BOOL is_color
= FALSE
;
1295 glsl_dst
->mask_str
[0] = '\0';
1296 glsl_dst
->reg_name
[0] = '\0';
1298 shader_glsl_get_register_name(wined3d_dst
->register_type
, wined3d_dst
->register_idx
,
1299 wined3d_dst
->rel_addr
, glsl_dst
->reg_name
, &is_color
, ins
);
1300 return shader_glsl_get_write_mask(wined3d_dst
, glsl_dst
->mask_str
);
1303 /* Append the destination part of the instruction to the buffer, return the effective write mask */
1304 static DWORD
shader_glsl_append_dst_ext(SHADER_BUFFER
*buffer
,
1305 const struct wined3d_shader_instruction
*ins
, const struct wined3d_shader_dst_param
*dst
)
1307 glsl_dst_param_t glsl_dst
;
1310 mask
= shader_glsl_add_dst_param(ins
, dst
, &glsl_dst
);
1311 if (mask
) shader_addline(buffer
, "%s%s = %s(", glsl_dst
.reg_name
, glsl_dst
.mask_str
, shift_glsl_tab
[dst
->shift
]);
1316 /* Append the destination part of the instruction to the buffer, return the effective write mask */
1317 static DWORD
shader_glsl_append_dst(SHADER_BUFFER
*buffer
, const struct wined3d_shader_instruction
*ins
)
1319 return shader_glsl_append_dst_ext(buffer
, ins
, &ins
->dst
[0]);
1322 /** Process GLSL instruction modifiers */
1323 void shader_glsl_add_instruction_modifiers(const struct wined3d_shader_instruction
*ins
)
1325 glsl_dst_param_t dst_param
;
1328 if (!ins
->dst_count
) return;
1330 modifiers
= ins
->dst
[0].modifiers
;
1331 if (!modifiers
) return;
1333 shader_glsl_add_dst_param(ins
, &ins
->dst
[0], &dst_param
);
1335 if (modifiers
& WINED3DSPDM_SATURATE
)
1337 /* _SAT means to clamp the value of the register to between 0 and 1 */
1338 shader_addline(ins
->ctx
->buffer
, "%s%s = clamp(%s%s, 0.0, 1.0);\n", dst_param
.reg_name
,
1339 dst_param
.mask_str
, dst_param
.reg_name
, dst_param
.mask_str
);
1342 if (modifiers
& WINED3DSPDM_MSAMPCENTROID
)
1344 FIXME("_centroid modifier not handled\n");
1347 if (modifiers
& WINED3DSPDM_PARTIALPRECISION
)
1349 /* MSDN says this modifier can be safely ignored, so that's what we'll do. */
1353 static inline const char *shader_get_comp_op(DWORD flags
)
1355 DWORD op
= (flags
& INST_CONTROLS_MASK
) >> INST_CONTROLS_SHIFT
;
1357 case COMPARISON_GT
: return ">";
1358 case COMPARISON_EQ
: return "==";
1359 case COMPARISON_GE
: return ">=";
1360 case COMPARISON_LT
: return "<";
1361 case COMPARISON_NE
: return "!=";
1362 case COMPARISON_LE
: return "<=";
1364 FIXME("Unrecognized comparison value: %u\n", op
);
1369 static void shader_glsl_get_sample_function(DWORD sampler_type
, DWORD flags
, glsl_sample_function_t
*sample_function
)
1371 BOOL projected
= flags
& WINED3D_GLSL_SAMPLE_PROJECTED
;
1372 BOOL texrect
= flags
& WINED3D_GLSL_SAMPLE_RECT
;
1373 BOOL lod
= flags
& WINED3D_GLSL_SAMPLE_LOD
;
1375 /* Note that there's no such thing as a projected cube texture. */
1376 switch(sampler_type
) {
1379 sample_function
->name
= projected
? "texture1DProjLod" : "texture1DLod";
1381 sample_function
->name
= projected
? "texture1DProj" : "texture1D";
1383 sample_function
->coord_mask
= WINED3DSP_WRITEMASK_0
;
1388 sample_function
->name
= projected
? "texture2DRectProjLod" : "texture2DRectLod";
1390 sample_function
->name
= projected
? "texture2DRectProj" : "texture2DRect";
1394 sample_function
->name
= projected
? "texture2DProjLod" : "texture2DLod";
1396 sample_function
->name
= projected
? "texture2DProj" : "texture2D";
1399 sample_function
->coord_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
;
1401 case WINED3DSTT_CUBE
:
1403 sample_function
->name
= "textureCubeLod";
1405 sample_function
->name
= "textureCube";
1407 sample_function
->coord_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
1409 case WINED3DSTT_VOLUME
:
1411 sample_function
->name
= projected
? "texture3DProjLod" : "texture3DLod";
1413 sample_function
->name
= projected
? "texture3DProj" : "texture3D";
1415 sample_function
->coord_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
1418 sample_function
->name
= "";
1419 sample_function
->coord_mask
= 0;
1420 FIXME("Unrecognized sampler type: %#x;\n", sampler_type
);
1425 static void shader_glsl_append_fixup_arg(char *arguments
, const char *reg_name
,
1426 BOOL sign_fixup
, enum fixup_channel_source channel_source
)
1428 switch(channel_source
)
1430 case CHANNEL_SOURCE_ZERO
:
1431 strcat(arguments
, "0.0");
1434 case CHANNEL_SOURCE_ONE
:
1435 strcat(arguments
, "1.0");
1438 case CHANNEL_SOURCE_X
:
1439 strcat(arguments
, reg_name
);
1440 strcat(arguments
, ".x");
1443 case CHANNEL_SOURCE_Y
:
1444 strcat(arguments
, reg_name
);
1445 strcat(arguments
, ".y");
1448 case CHANNEL_SOURCE_Z
:
1449 strcat(arguments
, reg_name
);
1450 strcat(arguments
, ".z");
1453 case CHANNEL_SOURCE_W
:
1454 strcat(arguments
, reg_name
);
1455 strcat(arguments
, ".w");
1459 FIXME("Unhandled channel source %#x\n", channel_source
);
1460 strcat(arguments
, "undefined");
1464 if (sign_fixup
) strcat(arguments
, " * 2.0 - 1.0");
1467 static void shader_glsl_color_correction(const struct wined3d_shader_instruction
*ins
, struct color_fixup_desc fixup
)
1469 struct wined3d_shader_dst_param dst
;
1470 unsigned int mask_size
, remaining
;
1471 glsl_dst_param_t dst_param
;
1472 char arguments
[256];
1476 if (fixup
.x_sign_fixup
|| fixup
.x_source
!= CHANNEL_SOURCE_X
) mask
|= WINED3DSP_WRITEMASK_0
;
1477 if (fixup
.y_sign_fixup
|| fixup
.y_source
!= CHANNEL_SOURCE_Y
) mask
|= WINED3DSP_WRITEMASK_1
;
1478 if (fixup
.z_sign_fixup
|| fixup
.z_source
!= CHANNEL_SOURCE_Z
) mask
|= WINED3DSP_WRITEMASK_2
;
1479 if (fixup
.w_sign_fixup
|| fixup
.w_source
!= CHANNEL_SOURCE_W
) mask
|= WINED3DSP_WRITEMASK_3
;
1480 mask
&= ins
->dst
[0].write_mask
;
1482 if (!mask
) return; /* Nothing to do */
1484 if (is_yuv_fixup(fixup
))
1486 enum yuv_fixup yuv_fixup
= get_yuv_fixup(fixup
);
1487 FIXME("YUV fixup (%#x) not supported\n", yuv_fixup
);
1491 mask_size
= shader_glsl_get_write_mask_size(mask
);
1494 dst
.write_mask
= mask
;
1495 shader_glsl_add_dst_param(ins
, &dst
, &dst_param
);
1497 arguments
[0] = '\0';
1498 remaining
= mask_size
;
1499 if (mask
& WINED3DSP_WRITEMASK_0
)
1501 shader_glsl_append_fixup_arg(arguments
, dst_param
.reg_name
, fixup
.x_sign_fixup
, fixup
.x_source
);
1502 if (--remaining
) strcat(arguments
, ", ");
1504 if (mask
& WINED3DSP_WRITEMASK_1
)
1506 shader_glsl_append_fixup_arg(arguments
, dst_param
.reg_name
, fixup
.y_sign_fixup
, fixup
.y_source
);
1507 if (--remaining
) strcat(arguments
, ", ");
1509 if (mask
& WINED3DSP_WRITEMASK_2
)
1511 shader_glsl_append_fixup_arg(arguments
, dst_param
.reg_name
, fixup
.z_sign_fixup
, fixup
.z_source
);
1512 if (--remaining
) strcat(arguments
, ", ");
1514 if (mask
& WINED3DSP_WRITEMASK_3
)
1516 shader_glsl_append_fixup_arg(arguments
, dst_param
.reg_name
, fixup
.w_sign_fixup
, fixup
.w_source
);
1517 if (--remaining
) strcat(arguments
, ", ");
1522 shader_addline(ins
->ctx
->buffer
, "%s%s = vec%u(%s);\n",
1523 dst_param
.reg_name
, dst_param
.mask_str
, mask_size
, arguments
);
1527 shader_addline(ins
->ctx
->buffer
, "%s%s = %s;\n", dst_param
.reg_name
, dst_param
.mask_str
, arguments
);
1531 static void PRINTF_ATTR(6, 7) shader_glsl_gen_sample_code(const struct wined3d_shader_instruction
*ins
,
1532 DWORD sampler
, const glsl_sample_function_t
*sample_function
, DWORD swizzle
,
1533 const char *bias
, const char *coord_reg_fmt
, ...)
1535 const char *sampler_base
;
1536 char dst_swizzle
[6];
1537 struct color_fixup_desc fixup
;
1538 BOOL np2_fixup
= FALSE
;
1541 shader_glsl_get_swizzle(swizzle
, FALSE
, ins
->dst
[0].write_mask
, dst_swizzle
);
1543 if (shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
))
1545 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*)ins
->ctx
->shader
;
1546 fixup
= This
->cur_args
->color_fixup
[sampler
];
1547 sampler_base
= "Psampler";
1549 if(This
->cur_args
->np2_fixup
& (1 << sampler
)) {
1551 FIXME("Biased sampling from NP2 textures is unsupported\n");
1557 sampler_base
= "Vsampler";
1558 fixup
= COLOR_FIXUP_IDENTITY
; /* FIXME: Vshader color fixup */
1561 shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
1563 shader_addline(ins
->ctx
->buffer
, "%s(%s%u, ", sample_function
->name
, sampler_base
, sampler
);
1565 va_start(args
, coord_reg_fmt
);
1566 shader_vaddline(ins
->ctx
->buffer
, coord_reg_fmt
, args
);
1570 shader_addline(ins
->ctx
->buffer
, ", %s)%s);\n", bias
, dst_swizzle
);
1573 shader_addline(ins
->ctx
->buffer
, " * PsamplerNP2Fixup%u)%s);\n", sampler
, dst_swizzle
);
1575 shader_addline(ins
->ctx
->buffer
, ")%s);\n", dst_swizzle
);
1579 if(!is_identity_fixup(fixup
)) {
1580 shader_glsl_color_correction(ins
, fixup
);
1584 /*****************************************************************************
1586 * Begin processing individual instruction opcodes
1588 ****************************************************************************/
1590 /* Generate GLSL arithmetic functions (dst = src1 + src2) */
1591 static void shader_glsl_arith(const struct wined3d_shader_instruction
*ins
)
1593 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
1594 glsl_src_param_t src0_param
;
1595 glsl_src_param_t src1_param
;
1599 /* Determine the GLSL operator to use based on the opcode */
1600 switch (ins
->handler_idx
)
1602 case WINED3DSIH_MUL
: op
= '*'; break;
1603 case WINED3DSIH_ADD
: op
= '+'; break;
1604 case WINED3DSIH_SUB
: op
= '-'; break;
1607 FIXME("Opcode %#x not yet handled in GLSL\n", ins
->handler_idx
);
1611 write_mask
= shader_glsl_append_dst(buffer
, ins
);
1612 shader_glsl_add_src_param(ins
, &ins
->src
[0], write_mask
, &src0_param
);
1613 shader_glsl_add_src_param(ins
, &ins
->src
[1], write_mask
, &src1_param
);
1614 shader_addline(buffer
, "%s %c %s);\n", src0_param
.param_str
, op
, src1_param
.param_str
);
1617 /* Process the WINED3DSIO_MOV opcode using GLSL (dst = src) */
1618 static void shader_glsl_mov(const struct wined3d_shader_instruction
*ins
)
1620 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
1621 glsl_src_param_t src0_param
;
1624 write_mask
= shader_glsl_append_dst(buffer
, ins
);
1625 shader_glsl_add_src_param(ins
, &ins
->src
[0], write_mask
, &src0_param
);
1627 /* In vs_1_1 WINED3DSIO_MOV can write to the address register. In later
1628 * shader versions WINED3DSIO_MOVA is used for this. */
1629 if ((WINED3DSHADER_VERSION_MAJOR(ins
->ctx
->reg_maps
->shader_version
) == 1
1630 && !shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
)
1631 && ins
->dst
[0].register_type
== WINED3DSPR_ADDR
))
1633 /* This is a simple floor() */
1634 unsigned int mask_size
= shader_glsl_get_write_mask_size(write_mask
);
1635 if (mask_size
> 1) {
1636 shader_addline(buffer
, "ivec%d(floor(%s)));\n", mask_size
, src0_param
.param_str
);
1638 shader_addline(buffer
, "int(floor(%s)));\n", src0_param
.param_str
);
1641 else if(ins
->handler_idx
== WINED3DSIH_MOVA
)
1643 /* We need to *round* to the nearest int here. */
1644 unsigned int mask_size
= shader_glsl_get_write_mask_size(write_mask
);
1645 if (mask_size
> 1) {
1646 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
);
1648 shader_addline(buffer
, "int(floor(abs(%s) + 0.5) * sign(%s)));\n", src0_param
.param_str
, src0_param
.param_str
);
1651 shader_addline(buffer
, "%s);\n", src0_param
.param_str
);
1655 /* Process the dot product operators DP3 and DP4 in GLSL (dst = dot(src0, src1)) */
1656 static void shader_glsl_dot(const struct wined3d_shader_instruction
*ins
)
1658 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
1659 glsl_src_param_t src0_param
;
1660 glsl_src_param_t src1_param
;
1661 DWORD dst_write_mask
, src_write_mask
;
1662 unsigned int dst_size
= 0;
1664 dst_write_mask
= shader_glsl_append_dst(buffer
, ins
);
1665 dst_size
= shader_glsl_get_write_mask_size(dst_write_mask
);
1667 /* dp3 works on vec3, dp4 on vec4 */
1668 if (ins
->handler_idx
== WINED3DSIH_DP4
)
1670 src_write_mask
= WINED3DSP_WRITEMASK_ALL
;
1672 src_write_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
1675 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_write_mask
, &src0_param
);
1676 shader_glsl_add_src_param(ins
, &ins
->src
[1], src_write_mask
, &src1_param
);
1679 shader_addline(buffer
, "vec%d(dot(%s, %s)));\n", dst_size
, src0_param
.param_str
, src1_param
.param_str
);
1681 shader_addline(buffer
, "dot(%s, %s));\n", src0_param
.param_str
, src1_param
.param_str
);
1685 /* Note that this instruction has some restrictions. The destination write mask
1686 * can't contain the w component, and the source swizzles have to be .xyzw */
1687 static void shader_glsl_cross(const struct wined3d_shader_instruction
*ins
)
1689 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
1690 glsl_src_param_t src0_param
;
1691 glsl_src_param_t src1_param
;
1694 shader_glsl_get_write_mask(&ins
->dst
[0], dst_mask
);
1695 shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
1696 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
1697 shader_glsl_add_src_param(ins
, &ins
->src
[1], src_mask
, &src1_param
);
1698 shader_addline(ins
->ctx
->buffer
, "cross(%s, %s)%s);\n", src0_param
.param_str
, src1_param
.param_str
, dst_mask
);
1701 /* Process the WINED3DSIO_POW instruction in GLSL (dst = |src0|^src1)
1702 * Src0 and src1 are scalars. Note that D3D uses the absolute of src0, while
1703 * GLSL uses the value as-is. */
1704 static void shader_glsl_pow(const struct wined3d_shader_instruction
*ins
)
1706 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
1707 glsl_src_param_t src0_param
;
1708 glsl_src_param_t src1_param
;
1709 DWORD dst_write_mask
;
1710 unsigned int dst_size
;
1712 dst_write_mask
= shader_glsl_append_dst(buffer
, ins
);
1713 dst_size
= shader_glsl_get_write_mask_size(dst_write_mask
);
1715 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
1716 shader_glsl_add_src_param(ins
, &ins
->src
[1], WINED3DSP_WRITEMASK_0
, &src1_param
);
1719 shader_addline(buffer
, "vec%d(pow(abs(%s), %s)));\n", dst_size
, src0_param
.param_str
, src1_param
.param_str
);
1721 shader_addline(buffer
, "pow(abs(%s), %s));\n", src0_param
.param_str
, src1_param
.param_str
);
1725 /* Process the WINED3DSIO_LOG instruction in GLSL (dst = log2(|src0|))
1726 * Src0 is a scalar. Note that D3D uses the absolute of src0, while
1727 * GLSL uses the value as-is. */
1728 static void shader_glsl_log(const struct wined3d_shader_instruction
*ins
)
1730 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
1731 glsl_src_param_t src0_param
;
1732 DWORD dst_write_mask
;
1733 unsigned int dst_size
;
1735 dst_write_mask
= shader_glsl_append_dst(buffer
, ins
);
1736 dst_size
= shader_glsl_get_write_mask_size(dst_write_mask
);
1738 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
1741 shader_addline(buffer
, "vec%d(log2(abs(%s))));\n", dst_size
, src0_param
.param_str
);
1743 shader_addline(buffer
, "log2(abs(%s)));\n", src0_param
.param_str
);
1747 /* Map the opcode 1-to-1 to the GL code (arg->dst = instruction(src0, src1, ...) */
1748 static void shader_glsl_map2gl(const struct wined3d_shader_instruction
*ins
)
1750 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
1751 glsl_src_param_t src_param
;
1752 const char *instruction
;
1756 /* Determine the GLSL function to use based on the opcode */
1757 /* TODO: Possibly make this a table for faster lookups */
1758 switch (ins
->handler_idx
)
1760 case WINED3DSIH_MIN
: instruction
= "min"; break;
1761 case WINED3DSIH_MAX
: instruction
= "max"; break;
1762 case WINED3DSIH_ABS
: instruction
= "abs"; break;
1763 case WINED3DSIH_FRC
: instruction
= "fract"; break;
1764 case WINED3DSIH_NRM
: instruction
= "normalize"; break;
1765 case WINED3DSIH_EXP
: instruction
= "exp2"; break;
1766 case WINED3DSIH_SGN
: instruction
= "sign"; break;
1767 case WINED3DSIH_DSX
: instruction
= "dFdx"; break;
1768 case WINED3DSIH_DSY
: instruction
= "ycorrection.y * dFdy"; break;
1769 default: instruction
= "";
1770 FIXME("Opcode %#x not yet handled in GLSL\n", ins
->handler_idx
);
1774 write_mask
= shader_glsl_append_dst(buffer
, ins
);
1776 shader_addline(buffer
, "%s(", instruction
);
1780 shader_glsl_add_src_param(ins
, &ins
->src
[0], write_mask
, &src_param
);
1781 shader_addline(buffer
, "%s", src_param
.param_str
);
1782 for (i
= 1; i
< ins
->src_count
; ++i
)
1784 shader_glsl_add_src_param(ins
, &ins
->src
[i
], write_mask
, &src_param
);
1785 shader_addline(buffer
, ", %s", src_param
.param_str
);
1789 shader_addline(buffer
, "));\n");
1792 /** Process the WINED3DSIO_EXPP instruction in GLSL:
1793 * For shader model 1.x, do the following (and honor the writemask, so use a temporary variable):
1794 * dst.x = 2^(floor(src))
1795 * dst.y = src - floor(src)
1796 * dst.z = 2^src (partial precision is allowed, but optional)
1798 * For 2.0 shaders, just do this (honoring writemask and swizzle):
1799 * dst = 2^src; (partial precision is allowed, but optional)
1801 static void shader_glsl_expp(const struct wined3d_shader_instruction
*ins
)
1803 glsl_src_param_t src_param
;
1805 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
, &src_param
);
1807 if (ins
->ctx
->reg_maps
->shader_version
< WINED3DPS_VERSION(2,0))
1811 shader_addline(ins
->ctx
->buffer
, "tmp0.x = exp2(floor(%s));\n", src_param
.param_str
);
1812 shader_addline(ins
->ctx
->buffer
, "tmp0.y = %s - floor(%s);\n", src_param
.param_str
, src_param
.param_str
);
1813 shader_addline(ins
->ctx
->buffer
, "tmp0.z = exp2(%s);\n", src_param
.param_str
);
1814 shader_addline(ins
->ctx
->buffer
, "tmp0.w = 1.0;\n");
1816 shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
1817 shader_glsl_get_write_mask(&ins
->dst
[0], dst_mask
);
1818 shader_addline(ins
->ctx
->buffer
, "tmp0%s);\n", dst_mask
);
1821 unsigned int mask_size
;
1823 write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
1824 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
1826 if (mask_size
> 1) {
1827 shader_addline(ins
->ctx
->buffer
, "vec%d(exp2(%s)));\n", mask_size
, src_param
.param_str
);
1829 shader_addline(ins
->ctx
->buffer
, "exp2(%s));\n", src_param
.param_str
);
1834 /** Process the RCP (reciprocal or inverse) opcode in GLSL (dst = 1 / src) */
1835 static void shader_glsl_rcp(const struct wined3d_shader_instruction
*ins
)
1837 glsl_src_param_t src_param
;
1839 unsigned int mask_size
;
1841 write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
1842 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
1843 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_3
, &src_param
);
1845 if (mask_size
> 1) {
1846 shader_addline(ins
->ctx
->buffer
, "vec%d(1.0 / %s));\n", mask_size
, src_param
.param_str
);
1848 shader_addline(ins
->ctx
->buffer
, "1.0 / %s);\n", src_param
.param_str
);
1852 static void shader_glsl_rsq(const struct wined3d_shader_instruction
*ins
)
1854 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
1855 glsl_src_param_t src_param
;
1857 unsigned int mask_size
;
1859 write_mask
= shader_glsl_append_dst(buffer
, ins
);
1860 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
1862 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_3
, &src_param
);
1864 if (mask_size
> 1) {
1865 shader_addline(buffer
, "vec%d(inversesqrt(%s)));\n", mask_size
, src_param
.param_str
);
1867 shader_addline(buffer
, "inversesqrt(%s));\n", src_param
.param_str
);
1871 /** Process signed comparison opcodes in GLSL. */
1872 static void shader_glsl_compare(const struct wined3d_shader_instruction
*ins
)
1874 glsl_src_param_t src0_param
;
1875 glsl_src_param_t src1_param
;
1877 unsigned int mask_size
;
1879 write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
1880 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
1881 shader_glsl_add_src_param(ins
, &ins
->src
[0], write_mask
, &src0_param
);
1882 shader_glsl_add_src_param(ins
, &ins
->src
[1], write_mask
, &src1_param
);
1884 if (mask_size
> 1) {
1885 const char *compare
;
1887 switch(ins
->handler_idx
)
1889 case WINED3DSIH_SLT
: compare
= "lessThan"; break;
1890 case WINED3DSIH_SGE
: compare
= "greaterThanEqual"; break;
1891 default: compare
= "";
1892 FIXME("Can't handle opcode %#x\n", ins
->handler_idx
);
1895 shader_addline(ins
->ctx
->buffer
, "vec%d(%s(%s, %s)));\n", mask_size
, compare
,
1896 src0_param
.param_str
, src1_param
.param_str
);
1898 switch(ins
->handler_idx
)
1900 case WINED3DSIH_SLT
:
1901 /* Step(src0, src1) is not suitable here because if src0 == src1 SLT is supposed,
1902 * to return 0.0 but step returns 1.0 because step is not < x
1903 * An alternative is a bvec compare padded with an unused second component.
1904 * step(src1 * -1.0, src0 * -1.0) is not an option because it suffers from the same
1905 * issue. Playing with not() is not possible either because not() does not accept
1908 shader_addline(ins
->ctx
->buffer
, "(%s < %s) ? 1.0 : 0.0);\n",
1909 src0_param
.param_str
, src1_param
.param_str
);
1911 case WINED3DSIH_SGE
:
1912 /* Here we can use the step() function and safe a conditional */
1913 shader_addline(ins
->ctx
->buffer
, "step(%s, %s));\n", src1_param
.param_str
, src0_param
.param_str
);
1916 FIXME("Can't handle opcode %#x\n", ins
->handler_idx
);
1922 /** Process CMP instruction in GLSL (dst = src0 >= 0.0 ? src1 : src2), per channel */
1923 static void shader_glsl_cmp(const struct wined3d_shader_instruction
*ins
)
1925 glsl_src_param_t src0_param
;
1926 glsl_src_param_t src1_param
;
1927 glsl_src_param_t src2_param
;
1928 DWORD write_mask
, cmp_channel
= 0;
1931 BOOL temp_destination
= FALSE
;
1933 if (shader_is_scalar(ins
->src
[0].register_type
, ins
->src
[0].register_idx
))
1935 write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
1937 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_ALL
, &src0_param
);
1938 shader_glsl_add_src_param(ins
, &ins
->src
[1], write_mask
, &src1_param
);
1939 shader_glsl_add_src_param(ins
, &ins
->src
[2], write_mask
, &src2_param
);
1941 shader_addline(ins
->ctx
->buffer
, "%s >= 0.0 ? %s : %s);\n",
1942 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
1944 DWORD dst_mask
= ins
->dst
[0].write_mask
;
1945 struct wined3d_shader_dst_param dst
= ins
->dst
[0];
1947 /* Cycle through all source0 channels */
1948 for (i
=0; i
<4; i
++) {
1950 /* Find the destination channels which use the current source0 channel */
1951 for (j
=0; j
<4; j
++) {
1952 if (((ins
->src
[0].token
>> (WINED3DSP_SWIZZLE_SHIFT
+ 2 * j
)) & 0x3) == i
)
1954 write_mask
|= WINED3DSP_WRITEMASK_0
<< j
;
1955 cmp_channel
= WINED3DSP_WRITEMASK_0
<< j
;
1958 dst
.write_mask
= dst_mask
& write_mask
;
1960 /* Splitting the cmp instruction up in multiple lines imposes a problem:
1961 * The first lines may overwrite source parameters of the following lines.
1962 * Deal with that by using a temporary destination register if needed
1964 if ((ins
->src
[0].register_idx
== ins
->dst
[0].register_idx
1965 && ins
->src
[0].register_type
== ins
->dst
[0].register_type
)
1966 || (ins
->src
[1].register_idx
== ins
->dst
[0].register_idx
1967 && ins
->src
[1].register_type
== ins
->dst
[0].register_type
)
1968 || (ins
->src
[2].register_idx
== ins
->dst
[0].register_idx
1969 && ins
->src
[2].register_type
== ins
->dst
[0].register_type
))
1971 write_mask
= shader_glsl_get_write_mask(&dst
, mask_char
);
1972 if (!write_mask
) continue;
1973 shader_addline(ins
->ctx
->buffer
, "tmp0%s = (", mask_char
);
1974 temp_destination
= TRUE
;
1976 write_mask
= shader_glsl_append_dst_ext(ins
->ctx
->buffer
, ins
, &dst
);
1977 if (!write_mask
) continue;
1980 shader_glsl_add_src_param(ins
, &ins
->src
[0], cmp_channel
, &src0_param
);
1981 shader_glsl_add_src_param(ins
, &ins
->src
[1], write_mask
, &src1_param
);
1982 shader_glsl_add_src_param(ins
, &ins
->src
[2], write_mask
, &src2_param
);
1984 shader_addline(ins
->ctx
->buffer
, "%s >= 0.0 ? %s : %s);\n",
1985 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
1988 if(temp_destination
) {
1989 shader_glsl_get_write_mask(&ins
->dst
[0], mask_char
);
1990 shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
1991 shader_addline(ins
->ctx
->buffer
, "tmp0%s);\n", mask_char
);
1997 /** Process the CND opcode in GLSL (dst = (src0 > 0.5) ? src1 : src2) */
1998 /* For ps 1.1-1.3, only a single component of src0 is used. For ps 1.4
1999 * the compare is done per component of src0. */
2000 static void shader_glsl_cnd(const struct wined3d_shader_instruction
*ins
)
2002 struct wined3d_shader_dst_param dst
;
2003 glsl_src_param_t src0_param
;
2004 glsl_src_param_t src1_param
;
2005 glsl_src_param_t src2_param
;
2006 DWORD write_mask
, cmp_channel
= 0;
2010 if (ins
->ctx
->reg_maps
->shader_version
< WINED3DPS_VERSION(1, 4))
2012 write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2013 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2014 shader_glsl_add_src_param(ins
, &ins
->src
[1], write_mask
, &src1_param
);
2015 shader_glsl_add_src_param(ins
, &ins
->src
[2], write_mask
, &src2_param
);
2017 /* Fun: The D3DSI_COISSUE flag changes the semantic of the cnd instruction for < 1.4 shaders */
2020 shader_addline(ins
->ctx
->buffer
, "%s /* COISSUE! */);\n", src1_param
.param_str
);
2022 shader_addline(ins
->ctx
->buffer
, "%s > 0.5 ? %s : %s);\n",
2023 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
2027 /* Cycle through all source0 channels */
2028 dst_mask
= ins
->dst
[0].write_mask
;
2030 for (i
=0; i
<4; i
++) {
2032 /* Find the destination channels which use the current source0 channel */
2033 for (j
=0; j
<4; j
++) {
2034 if (((ins
->src
[0].token
>> (WINED3DSP_SWIZZLE_SHIFT
+ 2 * j
)) & 0x3) == i
)
2036 write_mask
|= WINED3DSP_WRITEMASK_0
<< j
;
2037 cmp_channel
= WINED3DSP_WRITEMASK_0
<< j
;
2041 dst
.write_mask
= dst_mask
& write_mask
;
2042 write_mask
= shader_glsl_append_dst_ext(ins
->ctx
->buffer
, ins
, &dst
);
2043 if (!write_mask
) continue;
2045 shader_glsl_add_src_param(ins
, &ins
->src
[0], cmp_channel
, &src0_param
);
2046 shader_glsl_add_src_param(ins
, &ins
->src
[1], write_mask
, &src1_param
);
2047 shader_glsl_add_src_param(ins
, &ins
->src
[2], write_mask
, &src2_param
);
2049 shader_addline(ins
->ctx
->buffer
, "%s > 0.5 ? %s : %s);\n",
2050 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
2054 /** GLSL code generation for WINED3DSIO_MAD: Multiply the first 2 opcodes, then add the last */
2055 static void shader_glsl_mad(const struct wined3d_shader_instruction
*ins
)
2057 glsl_src_param_t src0_param
;
2058 glsl_src_param_t src1_param
;
2059 glsl_src_param_t src2_param
;
2062 write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2063 shader_glsl_add_src_param(ins
, &ins
->src
[0], write_mask
, &src0_param
);
2064 shader_glsl_add_src_param(ins
, &ins
->src
[1], write_mask
, &src1_param
);
2065 shader_glsl_add_src_param(ins
, &ins
->src
[2], write_mask
, &src2_param
);
2066 shader_addline(ins
->ctx
->buffer
, "(%s * %s) + %s);\n",
2067 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
2070 /** Handles transforming all WINED3DSIO_M?x? opcodes for
2071 Vertex shaders to GLSL codes */
2072 static void shader_glsl_mnxn(const struct wined3d_shader_instruction
*ins
)
2075 int nComponents
= 0;
2076 struct wined3d_shader_dst_param tmp_dst
= {0};
2077 struct wined3d_shader_src_param tmp_src
[2] = {{0}};
2078 struct wined3d_shader_instruction tmp_ins
;
2080 memset(&tmp_ins
, 0, sizeof(tmp_ins
));
2082 /* Set constants for the temporary argument */
2083 tmp_ins
.ctx
= ins
->ctx
;
2084 tmp_ins
.dst_count
= 1;
2085 tmp_ins
.dst
= &tmp_dst
;
2086 tmp_ins
.src_count
= 2;
2087 tmp_ins
.src
= tmp_src
;
2089 switch(ins
->handler_idx
)
2091 case WINED3DSIH_M4x4
:
2093 tmp_ins
.handler_idx
= WINED3DSIH_DP4
;
2095 case WINED3DSIH_M4x3
:
2097 tmp_ins
.handler_idx
= WINED3DSIH_DP4
;
2099 case WINED3DSIH_M3x4
:
2101 tmp_ins
.handler_idx
= WINED3DSIH_DP3
;
2103 case WINED3DSIH_M3x3
:
2105 tmp_ins
.handler_idx
= WINED3DSIH_DP3
;
2107 case WINED3DSIH_M3x2
:
2109 tmp_ins
.handler_idx
= WINED3DSIH_DP3
;
2115 tmp_dst
= ins
->dst
[0];
2116 tmp_src
[0] = ins
->src
[0];
2117 tmp_src
[1] = ins
->src
[1];
2118 for (i
= 0; i
< nComponents
; ++i
)
2120 tmp_dst
.write_mask
= WINED3DSP_WRITEMASK_0
<< i
;
2121 shader_glsl_dot(&tmp_ins
);
2122 ++tmp_src
[1].register_idx
;
2128 The LRP instruction performs a component-wise linear interpolation
2129 between the second and third operands using the first operand as the
2130 blend factor. Equation: (dst = src2 + src0 * (src1 - src2))
2131 This is equivalent to mix(src2, src1, src0);
2133 static void shader_glsl_lrp(const struct wined3d_shader_instruction
*ins
)
2135 glsl_src_param_t src0_param
;
2136 glsl_src_param_t src1_param
;
2137 glsl_src_param_t src2_param
;
2140 write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2142 shader_glsl_add_src_param(ins
, &ins
->src
[0], write_mask
, &src0_param
);
2143 shader_glsl_add_src_param(ins
, &ins
->src
[1], write_mask
, &src1_param
);
2144 shader_glsl_add_src_param(ins
, &ins
->src
[2], write_mask
, &src2_param
);
2146 shader_addline(ins
->ctx
->buffer
, "mix(%s, %s, %s));\n",
2147 src2_param
.param_str
, src1_param
.param_str
, src0_param
.param_str
);
2150 /** Process the WINED3DSIO_LIT instruction in GLSL:
2151 * dst.x = dst.w = 1.0
2152 * dst.y = (src0.x > 0) ? src0.x
2153 * dst.z = (src0.x > 0) ? ((src0.y > 0) ? pow(src0.y, src.w) : 0) : 0
2154 * where src.w is clamped at +- 128
2156 static void shader_glsl_lit(const struct wined3d_shader_instruction
*ins
)
2158 glsl_src_param_t src0_param
;
2159 glsl_src_param_t src1_param
;
2160 glsl_src_param_t src3_param
;
2163 shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2164 shader_glsl_get_write_mask(&ins
->dst
[0], dst_mask
);
2166 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2167 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_1
, &src1_param
);
2168 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_3
, &src3_param
);
2170 /* The sdk specifies the instruction like this
2172 * if(src.x > 0.0) dst.y = src.x
2174 * if(src.x > 0.0 && src.y > 0.0) dst.z = pow(src.y, power);
2178 * Obviously that has quite a few conditionals in it which we don't like. So the first step is this:
2179 * dst.x = 1.0 ... No further explanation needed
2180 * dst.y = max(src.y, 0.0); ... If x < 0.0, use 0.0, otherwise x. Same as the conditional
2181 * dst.z = x > 0.0 ? pow(max(y, 0.0), p) : 0; ... 0 ^ power is 0, and otherwise we use y anyway
2182 * dst.w = 1.0. ... Nothing fancy.
2184 * So we still have one conditional in there. So do this:
2185 * dst.z = pow(max(0.0, src.y) * step(0.0, src.x), power);
2187 * 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),
2188 * 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.
2189 * if both x and y are > 0, we get pow(y * 1.0, power), as it is supposed to
2191 shader_addline(ins
->ctx
->buffer
,
2192 "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",
2193 src0_param
.param_str
, src1_param
.param_str
, src0_param
.param_str
, src3_param
.param_str
, dst_mask
);
2196 /** Process the WINED3DSIO_DST instruction in GLSL:
2198 * dst.y = src0.x * src0.y
2202 static void shader_glsl_dst(const struct wined3d_shader_instruction
*ins
)
2204 glsl_src_param_t src0y_param
;
2205 glsl_src_param_t src0z_param
;
2206 glsl_src_param_t src1y_param
;
2207 glsl_src_param_t src1w_param
;
2210 shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2211 shader_glsl_get_write_mask(&ins
->dst
[0], dst_mask
);
2213 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_1
, &src0y_param
);
2214 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_2
, &src0z_param
);
2215 shader_glsl_add_src_param(ins
, &ins
->src
[1], WINED3DSP_WRITEMASK_1
, &src1y_param
);
2216 shader_glsl_add_src_param(ins
, &ins
->src
[1], WINED3DSP_WRITEMASK_3
, &src1w_param
);
2218 shader_addline(ins
->ctx
->buffer
, "vec4(1.0, %s * %s, %s, %s))%s;\n",
2219 src0y_param
.param_str
, src1y_param
.param_str
, src0z_param
.param_str
, src1w_param
.param_str
, dst_mask
);
2222 /** Process the WINED3DSIO_SINCOS instruction in GLSL:
2223 * VS 2.0 requires that specific cosine and sine constants be passed to this instruction so the hardware
2224 * can handle it. But, these functions are built-in for GLSL, so we can just ignore the last 2 params.
2226 * dst.x = cos(src0.?)
2227 * dst.y = sin(src0.?)
2231 static void shader_glsl_sincos(const struct wined3d_shader_instruction
*ins
)
2233 glsl_src_param_t src0_param
;
2236 write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2237 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2239 switch (write_mask
) {
2240 case WINED3DSP_WRITEMASK_0
:
2241 shader_addline(ins
->ctx
->buffer
, "cos(%s));\n", src0_param
.param_str
);
2244 case WINED3DSP_WRITEMASK_1
:
2245 shader_addline(ins
->ctx
->buffer
, "sin(%s));\n", src0_param
.param_str
);
2248 case (WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
):
2249 shader_addline(ins
->ctx
->buffer
, "vec2(cos(%s), sin(%s)));\n", src0_param
.param_str
, src0_param
.param_str
);
2253 ERR("Write mask should be .x, .y or .xy\n");
2258 /** Process the WINED3DSIO_LOOP instruction in GLSL:
2259 * Start a for() loop where src1.y is the initial value of aL,
2260 * increment aL by src1.z for a total of src1.x iterations.
2261 * Need to use a temporary variable for this operation.
2263 /* FIXME: I don't think nested loops will work correctly this way. */
2264 static void shader_glsl_loop(const struct wined3d_shader_instruction
*ins
)
2266 glsl_src_param_t src1_param
;
2267 IWineD3DBaseShaderImpl
*shader
= (IWineD3DBaseShaderImpl
*)ins
->ctx
->shader
;
2268 const DWORD
*control_values
= NULL
;
2269 const local_constant
*constant
;
2271 shader_glsl_add_src_param(ins
, &ins
->src
[1], WINED3DSP_WRITEMASK_ALL
, &src1_param
);
2273 /* Try to hardcode the loop control parameters if possible. Direct3D 9 class hardware doesn't support real
2274 * varying indexing, but Microsoft designed this feature for Shader model 2.x+. If the loop control is
2275 * known at compile time, the GLSL compiler can unroll the loop, and replace indirect addressing with direct
2278 if (ins
->src
[1].register_type
== WINED3DSPR_CONSTINT
)
2280 LIST_FOR_EACH_ENTRY(constant
, &shader
->baseShader
.constantsI
, local_constant
, entry
) {
2281 if (constant
->idx
== ins
->src
[1].register_idx
)
2283 control_values
= constant
->value
;
2289 if(control_values
) {
2290 if(control_values
[2] > 0) {
2291 shader_addline(ins
->ctx
->buffer
, "for (aL%u = %d; aL%u < (%d * %d + %d); aL%u += %d) {\n",
2292 shader
->baseShader
.cur_loop_depth
, control_values
[1],
2293 shader
->baseShader
.cur_loop_depth
, control_values
[0], control_values
[2], control_values
[1],
2294 shader
->baseShader
.cur_loop_depth
, control_values
[2]);
2295 } else if(control_values
[2] == 0) {
2296 shader_addline(ins
->ctx
->buffer
, "for (aL%u = %d, tmpInt%u = 0; tmpInt%u < %d; tmpInt%u++) {\n",
2297 shader
->baseShader
.cur_loop_depth
, control_values
[1], shader
->baseShader
.cur_loop_depth
,
2298 shader
->baseShader
.cur_loop_depth
, control_values
[0],
2299 shader
->baseShader
.cur_loop_depth
);
2301 shader_addline(ins
->ctx
->buffer
, "for (aL%u = %d; aL%u > (%d * %d + %d); aL%u += %d) {\n",
2302 shader
->baseShader
.cur_loop_depth
, control_values
[1],
2303 shader
->baseShader
.cur_loop_depth
, control_values
[0], control_values
[2], control_values
[1],
2304 shader
->baseShader
.cur_loop_depth
, control_values
[2]);
2307 shader_addline(ins
->ctx
->buffer
,
2308 "for (tmpInt%u = 0, aL%u = %s.y; tmpInt%u < %s.x; tmpInt%u++, aL%u += %s.z) {\n",
2309 shader
->baseShader
.cur_loop_depth
, shader
->baseShader
.cur_loop_regno
,
2310 src1_param
.reg_name
, shader
->baseShader
.cur_loop_depth
, src1_param
.reg_name
,
2311 shader
->baseShader
.cur_loop_depth
, shader
->baseShader
.cur_loop_regno
, src1_param
.reg_name
);
2314 shader
->baseShader
.cur_loop_depth
++;
2315 shader
->baseShader
.cur_loop_regno
++;
2318 static void shader_glsl_end(const struct wined3d_shader_instruction
*ins
)
2320 IWineD3DBaseShaderImpl
*shader
= (IWineD3DBaseShaderImpl
*)ins
->ctx
->shader
;
2322 shader_addline(ins
->ctx
->buffer
, "}\n");
2324 if (ins
->handler_idx
== WINED3DSIH_ENDLOOP
)
2326 shader
->baseShader
.cur_loop_depth
--;
2327 shader
->baseShader
.cur_loop_regno
--;
2330 if (ins
->handler_idx
== WINED3DSIH_ENDREP
)
2332 shader
->baseShader
.cur_loop_depth
--;
2336 static void shader_glsl_rep(const struct wined3d_shader_instruction
*ins
)
2338 IWineD3DBaseShaderImpl
*shader
= (IWineD3DBaseShaderImpl
*)ins
->ctx
->shader
;
2339 glsl_src_param_t src0_param
;
2341 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2342 shader_addline(ins
->ctx
->buffer
, "for (tmpInt%d = 0; tmpInt%d < %s; tmpInt%d++) {\n",
2343 shader
->baseShader
.cur_loop_depth
, shader
->baseShader
.cur_loop_depth
,
2344 src0_param
.param_str
, shader
->baseShader
.cur_loop_depth
);
2345 shader
->baseShader
.cur_loop_depth
++;
2348 static void shader_glsl_if(const struct wined3d_shader_instruction
*ins
)
2350 glsl_src_param_t src0_param
;
2352 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2353 shader_addline(ins
->ctx
->buffer
, "if (%s) {\n", src0_param
.param_str
);
2356 static void shader_glsl_ifc(const struct wined3d_shader_instruction
*ins
)
2358 glsl_src_param_t src0_param
;
2359 glsl_src_param_t src1_param
;
2361 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2362 shader_glsl_add_src_param(ins
, &ins
->src
[1], WINED3DSP_WRITEMASK_0
, &src1_param
);
2364 shader_addline(ins
->ctx
->buffer
, "if (%s %s %s) {\n",
2365 src0_param
.param_str
, shader_get_comp_op(ins
->flags
), src1_param
.param_str
);
2368 static void shader_glsl_else(const struct wined3d_shader_instruction
*ins
)
2370 shader_addline(ins
->ctx
->buffer
, "} else {\n");
2373 static void shader_glsl_break(const struct wined3d_shader_instruction
*ins
)
2375 shader_addline(ins
->ctx
->buffer
, "break;\n");
2378 /* FIXME: According to MSDN the compare is done per component. */
2379 static void shader_glsl_breakc(const struct wined3d_shader_instruction
*ins
)
2381 glsl_src_param_t src0_param
;
2382 glsl_src_param_t src1_param
;
2384 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2385 shader_glsl_add_src_param(ins
, &ins
->src
[1], WINED3DSP_WRITEMASK_0
, &src1_param
);
2387 shader_addline(ins
->ctx
->buffer
, "if (%s %s %s) break;\n",
2388 src0_param
.param_str
, shader_get_comp_op(ins
->flags
), src1_param
.param_str
);
2391 static void shader_glsl_label(const struct wined3d_shader_instruction
*ins
)
2393 shader_addline(ins
->ctx
->buffer
, "}\n");
2394 shader_addline(ins
->ctx
->buffer
, "void subroutine%u () {\n", ins
->src
[0].register_idx
);
2397 static void shader_glsl_call(const struct wined3d_shader_instruction
*ins
)
2399 shader_addline(ins
->ctx
->buffer
, "subroutine%u();\n", ins
->src
[0].register_idx
);
2402 static void shader_glsl_callnz(const struct wined3d_shader_instruction
*ins
)
2404 glsl_src_param_t src1_param
;
2406 shader_glsl_add_src_param(ins
, &ins
->src
[1], WINED3DSP_WRITEMASK_0
, &src1_param
);
2407 shader_addline(ins
->ctx
->buffer
, "if (%s) subroutine%u();\n", src1_param
.param_str
, ins
->src
[0].register_idx
);
2410 /*********************************************
2411 * Pixel Shader Specific Code begins here
2412 ********************************************/
2413 static void pshader_glsl_tex(const struct wined3d_shader_instruction
*ins
)
2415 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*)ins
->ctx
->shader
;
2416 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
2417 DWORD shader_version
= ins
->ctx
->reg_maps
->shader_version
;
2418 glsl_sample_function_t sample_function
;
2419 DWORD sample_flags
= 0;
2422 DWORD mask
= 0, swizzle
;
2424 /* 1.0-1.4: Use destination register as sampler source.
2425 * 2.0+: Use provided sampler source. */
2426 if (shader_version
< WINED3DPS_VERSION(2,0)) sampler_idx
= ins
->dst
[0].register_idx
;
2427 else sampler_idx
= ins
->src
[1].register_idx
;
2428 sampler_type
= ins
->ctx
->reg_maps
->samplers
[sampler_idx
] & WINED3DSP_TEXTURETYPE_MASK
;
2430 if (shader_version
< WINED3DPS_VERSION(1,4))
2432 DWORD flags
= deviceImpl
->stateBlock
->textureState
[sampler_idx
][WINED3DTSS_TEXTURETRANSFORMFLAGS
];
2434 /* Projected cube textures don't make a lot of sense, the resulting coordinates stay the same. */
2435 if (flags
& WINED3DTTFF_PROJECTED
&& sampler_type
!= WINED3DSTT_CUBE
) {
2436 sample_flags
|= WINED3D_GLSL_SAMPLE_PROJECTED
;
2437 switch (flags
& ~WINED3DTTFF_PROJECTED
) {
2438 case WINED3DTTFF_COUNT1
: FIXME("WINED3DTTFF_PROJECTED with WINED3DTTFF_COUNT1?\n"); break;
2439 case WINED3DTTFF_COUNT2
: mask
= WINED3DSP_WRITEMASK_1
; break;
2440 case WINED3DTTFF_COUNT3
: mask
= WINED3DSP_WRITEMASK_2
; break;
2441 case WINED3DTTFF_COUNT4
:
2442 case WINED3DTTFF_DISABLE
: mask
= WINED3DSP_WRITEMASK_3
; break;
2446 else if (shader_version
< WINED3DPS_VERSION(2,0))
2448 DWORD src_mod
= ins
->src
[0].modifiers
;
2450 if (src_mod
== WINED3DSPSM_DZ
) {
2451 sample_flags
|= WINED3D_GLSL_SAMPLE_PROJECTED
;
2452 mask
= WINED3DSP_WRITEMASK_2
;
2453 } else if (src_mod
== WINED3DSPSM_DW
) {
2454 sample_flags
|= WINED3D_GLSL_SAMPLE_PROJECTED
;
2455 mask
= WINED3DSP_WRITEMASK_3
;
2458 if (ins
->flags
& WINED3DSI_TEXLD_PROJECT
)
2460 /* ps 2.0 texldp instruction always divides by the fourth component. */
2461 sample_flags
|= WINED3D_GLSL_SAMPLE_PROJECTED
;
2462 mask
= WINED3DSP_WRITEMASK_3
;
2466 if(deviceImpl
->stateBlock
->textures
[sampler_idx
] &&
2467 IWineD3DBaseTexture_GetTextureDimensions(deviceImpl
->stateBlock
->textures
[sampler_idx
]) == GL_TEXTURE_RECTANGLE_ARB
) {
2468 sample_flags
|= WINED3D_GLSL_SAMPLE_RECT
;
2471 shader_glsl_get_sample_function(sampler_type
, sample_flags
, &sample_function
);
2472 mask
|= sample_function
.coord_mask
;
2474 if (shader_version
< WINED3DPS_VERSION(2,0)) swizzle
= WINED3DVS_NOSWIZZLE
;
2475 else swizzle
= ins
->src
[1].token
& WINED3DSP_SWIZZLE_MASK
;
2477 /* 1.0-1.3: Use destination register as coordinate source.
2478 1.4+: Use provided coordinate source register. */
2479 if (shader_version
< WINED3DPS_VERSION(1,4))
2482 shader_glsl_write_mask_to_str(mask
, coord_mask
);
2483 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, swizzle
, NULL
,
2484 "T%u%s", sampler_idx
, coord_mask
);
2486 glsl_src_param_t coord_param
;
2487 shader_glsl_add_src_param(ins
, &ins
->src
[0], mask
, &coord_param
);
2488 if (ins
->flags
& WINED3DSI_TEXLD_BIAS
)
2490 glsl_src_param_t bias
;
2491 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_3
, &bias
);
2492 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, swizzle
, bias
.param_str
,
2493 "%s", coord_param
.param_str
);
2495 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, swizzle
, NULL
,
2496 "%s", coord_param
.param_str
);
2501 static void shader_glsl_texldl(const struct wined3d_shader_instruction
*ins
)
2503 IWineD3DBaseShaderImpl
*This
= (IWineD3DBaseShaderImpl
*)ins
->ctx
->shader
;
2504 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
2505 glsl_sample_function_t sample_function
;
2506 glsl_src_param_t coord_param
, lod_param
;
2507 DWORD sample_flags
= WINED3D_GLSL_SAMPLE_LOD
;
2510 DWORD swizzle
= ins
->src
[1].token
& WINED3DSP_SWIZZLE_MASK
;
2512 sampler_idx
= ins
->src
[1].register_idx
;
2513 sampler_type
= ins
->ctx
->reg_maps
->samplers
[sampler_idx
] & WINED3DSP_TEXTURETYPE_MASK
;
2514 if(deviceImpl
->stateBlock
->textures
[sampler_idx
] &&
2515 IWineD3DBaseTexture_GetTextureDimensions(deviceImpl
->stateBlock
->textures
[sampler_idx
]) == GL_TEXTURE_RECTANGLE_ARB
) {
2516 sample_flags
|= WINED3D_GLSL_SAMPLE_RECT
;
2518 shader_glsl_get_sample_function(sampler_type
, sample_flags
, &sample_function
);
2519 shader_glsl_add_src_param(ins
, &ins
->src
[0], sample_function
.coord_mask
, &coord_param
);
2521 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_3
, &lod_param
);
2523 if (shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
))
2525 /* The GLSL spec claims the Lod sampling functions are only supported in vertex shaders.
2526 * However, they seem to work just fine in fragment shaders as well. */
2527 WARN("Using %s in fragment shader.\n", sample_function
.name
);
2529 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, swizzle
, lod_param
.param_str
,
2530 "%s", coord_param
.param_str
);
2533 static void pshader_glsl_texcoord(const struct wined3d_shader_instruction
*ins
)
2535 /* FIXME: Make this work for more than just 2D textures */
2536 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
2537 DWORD write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2539 if (ins
->ctx
->reg_maps
->shader_version
!= WINED3DPS_VERSION(1,4))
2543 shader_glsl_get_write_mask(&ins
->dst
[0], dst_mask
);
2544 shader_addline(buffer
, "clamp(gl_TexCoord[%u], 0.0, 1.0)%s);\n",
2545 ins
->dst
[0].register_idx
, dst_mask
);
2547 DWORD reg
= ins
->src
[0].register_idx
;
2548 DWORD src_mod
= ins
->src
[0].modifiers
;
2549 char dst_swizzle
[6];
2551 shader_glsl_get_swizzle(ins
->src
[0].token
, FALSE
, write_mask
, dst_swizzle
);
2553 if (src_mod
== WINED3DSPSM_DZ
) {
2554 glsl_src_param_t div_param
;
2555 unsigned int mask_size
= shader_glsl_get_write_mask_size(write_mask
);
2556 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_2
, &div_param
);
2558 if (mask_size
> 1) {
2559 shader_addline(buffer
, "gl_TexCoord[%u]%s / vec%d(%s));\n", reg
, dst_swizzle
, mask_size
, div_param
.param_str
);
2561 shader_addline(buffer
, "gl_TexCoord[%u]%s / %s);\n", reg
, dst_swizzle
, div_param
.param_str
);
2563 } else if (src_mod
== WINED3DSPSM_DW
) {
2564 glsl_src_param_t div_param
;
2565 unsigned int mask_size
= shader_glsl_get_write_mask_size(write_mask
);
2566 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_3
, &div_param
);
2568 if (mask_size
> 1) {
2569 shader_addline(buffer
, "gl_TexCoord[%u]%s / vec%d(%s));\n", reg
, dst_swizzle
, mask_size
, div_param
.param_str
);
2571 shader_addline(buffer
, "gl_TexCoord[%u]%s / %s);\n", reg
, dst_swizzle
, div_param
.param_str
);
2574 shader_addline(buffer
, "gl_TexCoord[%u]%s);\n", reg
, dst_swizzle
);
2579 /** Process the WINED3DSIO_TEXDP3TEX instruction in GLSL:
2580 * Take a 3-component dot product of the TexCoord[dstreg] and src,
2581 * then perform a 1D texture lookup from stage dstregnum, place into dst. */
2582 static void pshader_glsl_texdp3tex(const struct wined3d_shader_instruction
*ins
)
2584 glsl_src_param_t src0_param
;
2585 glsl_sample_function_t sample_function
;
2586 DWORD sampler_idx
= ins
->dst
[0].register_idx
;
2587 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2588 DWORD sampler_type
= ins
->ctx
->reg_maps
->samplers
[sampler_idx
] & WINED3DSP_TEXTURETYPE_MASK
;
2591 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
2593 /* Do I have to take care about the projected bit? I don't think so, since the dp3 returns only one
2594 * scalar, and projected sampling would require 4.
2596 * It is a dependent read - not valid with conditional NP2 textures
2598 shader_glsl_get_sample_function(sampler_type
, 0, &sample_function
);
2599 mask_size
= shader_glsl_get_write_mask_size(sample_function
.coord_mask
);
2604 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, WINED3DVS_NOSWIZZLE
, NULL
,
2605 "dot(gl_TexCoord[%u].xyz, %s)", sampler_idx
, src0_param
.param_str
);
2609 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, WINED3DVS_NOSWIZZLE
, NULL
,
2610 "vec2(dot(gl_TexCoord[%u].xyz, %s), 0.0)", sampler_idx
, src0_param
.param_str
);
2614 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, WINED3DVS_NOSWIZZLE
, NULL
,
2615 "vec3(dot(gl_TexCoord[%u].xyz, %s), 0.0, 0.0)", sampler_idx
, src0_param
.param_str
);
2619 FIXME("Unexpected mask size %u\n", mask_size
);
2624 /** Process the WINED3DSIO_TEXDP3 instruction in GLSL:
2625 * Take a 3-component dot product of the TexCoord[dstreg] and src. */
2626 static void pshader_glsl_texdp3(const struct wined3d_shader_instruction
*ins
)
2628 glsl_src_param_t src0_param
;
2629 DWORD dstreg
= ins
->dst
[0].register_idx
;
2630 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2632 unsigned int mask_size
;
2634 dst_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2635 mask_size
= shader_glsl_get_write_mask_size(dst_mask
);
2636 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
2638 if (mask_size
> 1) {
2639 shader_addline(ins
->ctx
->buffer
, "vec%d(dot(T%u.xyz, %s)));\n", mask_size
, dstreg
, src0_param
.param_str
);
2641 shader_addline(ins
->ctx
->buffer
, "dot(T%u.xyz, %s));\n", dstreg
, src0_param
.param_str
);
2645 /** Process the WINED3DSIO_TEXDEPTH instruction in GLSL:
2646 * Calculate the depth as dst.x / dst.y */
2647 static void pshader_glsl_texdepth(const struct wined3d_shader_instruction
*ins
)
2649 glsl_dst_param_t dst_param
;
2651 shader_glsl_add_dst_param(ins
, &ins
->dst
[0], &dst_param
);
2653 /* Tests show that texdepth never returns anything below 0.0, and that r5.y is clamped to 1.0.
2654 * Negative input is accepted, -0.25 / -0.5 returns 0.5. GL should clamp gl_FragDepth to [0;1], but
2655 * this doesn't always work, so clamp the results manually. Whether or not the x value is clamped at 1
2656 * too is irrelevant, since if x = 0, any y value < 1.0 (and > 1.0 is not allowed) results in a result
2659 shader_addline(ins
->ctx
->buffer
, "gl_FragDepth = clamp((%s.x / min(%s.y, 1.0)), 0.0, 1.0);\n",
2660 dst_param
.reg_name
, dst_param
.reg_name
);
2663 /** Process the WINED3DSIO_TEXM3X2DEPTH instruction in GLSL:
2664 * Last row of a 3x2 matrix multiply, use the result to calculate the depth:
2665 * Calculate tmp0.y = TexCoord[dstreg] . src.xyz; (tmp0.x has already been calculated)
2666 * depth = (tmp0.y == 0.0) ? 1.0 : tmp0.x / tmp0.y
2668 static void pshader_glsl_texm3x2depth(const struct wined3d_shader_instruction
*ins
)
2670 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2671 DWORD dstreg
= ins
->dst
[0].register_idx
;
2672 glsl_src_param_t src0_param
;
2674 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
2676 shader_addline(ins
->ctx
->buffer
, "tmp0.y = dot(T%u.xyz, %s);\n", dstreg
, src0_param
.param_str
);
2677 shader_addline(ins
->ctx
->buffer
, "gl_FragDepth = (tmp0.y == 0.0) ? 1.0 : clamp(tmp0.x / tmp0.y, 0.0, 1.0);\n");
2680 /** Process the WINED3DSIO_TEXM3X2PAD instruction in GLSL
2681 * Calculate the 1st of a 2-row matrix multiplication. */
2682 static void pshader_glsl_texm3x2pad(const struct wined3d_shader_instruction
*ins
)
2684 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2685 DWORD reg
= ins
->dst
[0].register_idx
;
2686 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
2687 glsl_src_param_t src0_param
;
2689 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
2690 shader_addline(buffer
, "tmp0.x = dot(T%u.xyz, %s);\n", reg
, src0_param
.param_str
);
2693 /** Process the WINED3DSIO_TEXM3X3PAD instruction in GLSL
2694 * Calculate the 1st or 2nd row of a 3-row matrix multiplication. */
2695 static void pshader_glsl_texm3x3pad(const struct wined3d_shader_instruction
*ins
)
2697 IWineD3DPixelShaderImpl
*shader
= (IWineD3DPixelShaderImpl
*)ins
->ctx
->shader
;
2698 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2699 DWORD reg
= ins
->dst
[0].register_idx
;
2700 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
2701 SHADER_PARSE_STATE
* current_state
= &shader
->baseShader
.parse_state
;
2702 glsl_src_param_t src0_param
;
2704 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
2705 shader_addline(buffer
, "tmp0.%c = dot(T%u.xyz, %s);\n", 'x' + current_state
->current_row
, reg
, src0_param
.param_str
);
2706 current_state
->texcoord_w
[current_state
->current_row
++] = reg
;
2709 static void pshader_glsl_texm3x2tex(const struct wined3d_shader_instruction
*ins
)
2711 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2712 DWORD reg
= ins
->dst
[0].register_idx
;
2713 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
2714 glsl_src_param_t src0_param
;
2715 DWORD sampler_type
= ins
->ctx
->reg_maps
->samplers
[reg
] & WINED3DSP_TEXTURETYPE_MASK
;
2716 glsl_sample_function_t sample_function
;
2718 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
2719 shader_addline(buffer
, "tmp0.y = dot(T%u.xyz, %s);\n", reg
, src0_param
.param_str
);
2721 shader_glsl_get_sample_function(sampler_type
, 0, &sample_function
);
2723 /* Sample the texture using the calculated coordinates */
2724 shader_glsl_gen_sample_code(ins
, reg
, &sample_function
, WINED3DVS_NOSWIZZLE
, NULL
, "tmp0.xy");
2727 /** Process the WINED3DSIO_TEXM3X3TEX instruction in GLSL
2728 * Perform the 3rd row of a 3x3 matrix multiply, then sample the texture using the calculated coordinates */
2729 static void pshader_glsl_texm3x3tex(const struct wined3d_shader_instruction
*ins
)
2731 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2732 glsl_src_param_t src0_param
;
2733 DWORD reg
= ins
->dst
[0].register_idx
;
2734 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*)ins
->ctx
->shader
;
2735 SHADER_PARSE_STATE
* current_state
= &This
->baseShader
.parse_state
;
2736 DWORD sampler_type
= ins
->ctx
->reg_maps
->samplers
[reg
] & WINED3DSP_TEXTURETYPE_MASK
;
2737 glsl_sample_function_t sample_function
;
2739 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
2740 shader_addline(ins
->ctx
->buffer
, "tmp0.z = dot(T%u.xyz, %s);\n", reg
, src0_param
.param_str
);
2742 /* Dependent read, not valid with conditional NP2 */
2743 shader_glsl_get_sample_function(sampler_type
, 0, &sample_function
);
2745 /* Sample the texture using the calculated coordinates */
2746 shader_glsl_gen_sample_code(ins
, reg
, &sample_function
, WINED3DVS_NOSWIZZLE
, NULL
, "tmp0.xyz");
2748 current_state
->current_row
= 0;
2751 /** Process the WINED3DSIO_TEXM3X3 instruction in GLSL
2752 * Perform the 3rd row of a 3x3 matrix multiply */
2753 static void pshader_glsl_texm3x3(const struct wined3d_shader_instruction
*ins
)
2755 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2756 glsl_src_param_t src0_param
;
2758 DWORD reg
= ins
->dst
[0].register_idx
;
2759 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*)ins
->ctx
->shader
;
2760 SHADER_PARSE_STATE
* current_state
= &This
->baseShader
.parse_state
;
2762 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
2764 shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2765 shader_glsl_get_write_mask(&ins
->dst
[0], dst_mask
);
2766 shader_addline(ins
->ctx
->buffer
, "vec4(tmp0.xy, dot(T%u.xyz, %s), 1.0)%s);\n", reg
, src0_param
.param_str
, dst_mask
);
2768 current_state
->current_row
= 0;
2771 /** Process the WINED3DSIO_TEXM3X3SPEC instruction in GLSL
2772 * Perform the final texture lookup based on the previous 2 3x3 matrix multiplies */
2773 static void pshader_glsl_texm3x3spec(const struct wined3d_shader_instruction
*ins
)
2775 IWineD3DPixelShaderImpl
*shader
= (IWineD3DPixelShaderImpl
*)ins
->ctx
->shader
;
2776 DWORD reg
= ins
->dst
[0].register_idx
;
2777 glsl_src_param_t src0_param
;
2778 glsl_src_param_t src1_param
;
2779 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
2780 SHADER_PARSE_STATE
* current_state
= &shader
->baseShader
.parse_state
;
2781 DWORD stype
= ins
->ctx
->reg_maps
->samplers
[reg
] & WINED3DSP_TEXTURETYPE_MASK
;
2782 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2783 glsl_sample_function_t sample_function
;
2785 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
2786 shader_glsl_add_src_param(ins
, &ins
->src
[1], src_mask
, &src1_param
);
2788 /* Perform the last matrix multiply operation */
2789 shader_addline(buffer
, "tmp0.z = dot(T%u.xyz, %s);\n", reg
, src0_param
.param_str
);
2790 /* Reflection calculation */
2791 shader_addline(buffer
, "tmp0.xyz = -reflect((%s), normalize(tmp0.xyz));\n", src1_param
.param_str
);
2793 /* Dependent read, not valid with conditional NP2 */
2794 shader_glsl_get_sample_function(stype
, 0, &sample_function
);
2796 /* Sample the texture */
2797 shader_glsl_gen_sample_code(ins
, reg
, &sample_function
, WINED3DVS_NOSWIZZLE
, NULL
, "tmp0.xyz");
2799 current_state
->current_row
= 0;
2802 /** Process the WINED3DSIO_TEXM3X3VSPEC instruction in GLSL
2803 * Perform the final texture lookup based on the previous 2 3x3 matrix multiplies */
2804 static void pshader_glsl_texm3x3vspec(const struct wined3d_shader_instruction
*ins
)
2806 IWineD3DPixelShaderImpl
*shader
= (IWineD3DPixelShaderImpl
*)ins
->ctx
->shader
;
2807 DWORD reg
= ins
->dst
[0].register_idx
;
2808 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
2809 SHADER_PARSE_STATE
* current_state
= &shader
->baseShader
.parse_state
;
2810 glsl_src_param_t src0_param
;
2811 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2812 DWORD sampler_type
= ins
->ctx
->reg_maps
->samplers
[reg
] & WINED3DSP_TEXTURETYPE_MASK
;
2813 glsl_sample_function_t sample_function
;
2815 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
2817 /* Perform the last matrix multiply operation */
2818 shader_addline(buffer
, "tmp0.z = dot(vec3(T%u), vec3(%s));\n", reg
, src0_param
.param_str
);
2820 /* Construct the eye-ray vector from w coordinates */
2821 shader_addline(buffer
, "tmp1.xyz = normalize(vec3(gl_TexCoord[%u].w, gl_TexCoord[%u].w, gl_TexCoord[%u].w));\n",
2822 current_state
->texcoord_w
[0], current_state
->texcoord_w
[1], reg
);
2823 shader_addline(buffer
, "tmp0.xyz = -reflect(tmp1.xyz, normalize(tmp0.xyz));\n");
2825 /* Dependent read, not valid with conditional NP2 */
2826 shader_glsl_get_sample_function(sampler_type
, 0, &sample_function
);
2828 /* Sample the texture using the calculated coordinates */
2829 shader_glsl_gen_sample_code(ins
, reg
, &sample_function
, WINED3DVS_NOSWIZZLE
, NULL
, "tmp0.xyz");
2831 current_state
->current_row
= 0;
2834 /** Process the WINED3DSIO_TEXBEM instruction in GLSL.
2835 * Apply a fake bump map transform.
2836 * texbem is pshader <= 1.3 only, this saves a few version checks
2838 static void pshader_glsl_texbem(const struct wined3d_shader_instruction
*ins
)
2840 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*)ins
->ctx
->shader
;
2841 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
2842 glsl_sample_function_t sample_function
;
2843 glsl_src_param_t coord_param
;
2850 sampler_idx
= ins
->dst
[0].register_idx
;
2851 flags
= deviceImpl
->stateBlock
->textureState
[sampler_idx
][WINED3DTSS_TEXTURETRANSFORMFLAGS
];
2853 sampler_type
= ins
->ctx
->reg_maps
->samplers
[sampler_idx
] & WINED3DSP_TEXTURETYPE_MASK
;
2854 /* Dependent read, not valid with conditional NP2 */
2855 shader_glsl_get_sample_function(sampler_type
, 0, &sample_function
);
2856 mask
= sample_function
.coord_mask
;
2858 shader_glsl_write_mask_to_str(mask
, coord_mask
);
2860 /* with projective textures, texbem only divides the static texture coord, not the displacement,
2861 * so we can't let the GL handle this.
2863 if (flags
& WINED3DTTFF_PROJECTED
) {
2865 char coord_div_mask
[3];
2866 switch (flags
& ~WINED3DTTFF_PROJECTED
) {
2867 case WINED3DTTFF_COUNT1
: FIXME("WINED3DTTFF_PROJECTED with WINED3DTTFF_COUNT1?\n"); break;
2868 case WINED3DTTFF_COUNT2
: div_mask
= WINED3DSP_WRITEMASK_1
; break;
2869 case WINED3DTTFF_COUNT3
: div_mask
= WINED3DSP_WRITEMASK_2
; break;
2870 case WINED3DTTFF_COUNT4
:
2871 case WINED3DTTFF_DISABLE
: div_mask
= WINED3DSP_WRITEMASK_3
; break;
2873 shader_glsl_write_mask_to_str(div_mask
, coord_div_mask
);
2874 shader_addline(ins
->ctx
->buffer
, "T%u%s /= T%u%s;\n", sampler_idx
, coord_mask
, sampler_idx
, coord_div_mask
);
2877 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
, &coord_param
);
2879 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, WINED3DVS_NOSWIZZLE
, NULL
,
2880 "T%u%s + vec4(bumpenvmat%d * %s, 0.0, 0.0)%s", sampler_idx
, coord_mask
, sampler_idx
,
2881 coord_param
.param_str
, coord_mask
);
2883 if (ins
->handler_idx
== WINED3DSIH_TEXBEML
)
2885 glsl_src_param_t luminance_param
;
2886 glsl_dst_param_t dst_param
;
2888 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_2
, &luminance_param
);
2889 shader_glsl_add_dst_param(ins
, &ins
->dst
[0], &dst_param
);
2891 shader_addline(ins
->ctx
->buffer
, "%s%s *= (%s * luminancescale%d + luminanceoffset%d);\n",
2892 dst_param
.reg_name
, dst_param
.mask_str
,
2893 luminance_param
.param_str
, sampler_idx
, sampler_idx
);
2897 static void pshader_glsl_bem(const struct wined3d_shader_instruction
*ins
)
2899 glsl_src_param_t src0_param
, src1_param
;
2900 DWORD sampler_idx
= ins
->dst
[0].register_idx
;
2902 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
, &src0_param
);
2903 shader_glsl_add_src_param(ins
, &ins
->src
[1], WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
, &src1_param
);
2905 shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2906 shader_addline(ins
->ctx
->buffer
, "%s + bumpenvmat%d * %s);\n",
2907 src0_param
.param_str
, sampler_idx
, src1_param
.param_str
);
2910 /** Process the WINED3DSIO_TEXREG2AR instruction in GLSL
2911 * Sample 2D texture at dst using the alpha & red (wx) components of src as texture coordinates */
2912 static void pshader_glsl_texreg2ar(const struct wined3d_shader_instruction
*ins
)
2914 glsl_src_param_t src0_param
;
2915 DWORD sampler_idx
= ins
->dst
[0].register_idx
;
2916 DWORD sampler_type
= ins
->ctx
->reg_maps
->samplers
[sampler_idx
] & WINED3DSP_TEXTURETYPE_MASK
;
2917 glsl_sample_function_t sample_function
;
2919 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_ALL
, &src0_param
);
2921 shader_glsl_get_sample_function(sampler_type
, 0, &sample_function
);
2922 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, WINED3DVS_NOSWIZZLE
, NULL
,
2923 "%s.wx", src0_param
.reg_name
);
2926 /** Process the WINED3DSIO_TEXREG2GB instruction in GLSL
2927 * Sample 2D texture at dst using the green & blue (yz) components of src as texture coordinates */
2928 static void pshader_glsl_texreg2gb(const struct wined3d_shader_instruction
*ins
)
2930 glsl_src_param_t src0_param
;
2931 DWORD sampler_idx
= ins
->dst
[0].register_idx
;
2932 DWORD sampler_type
= ins
->ctx
->reg_maps
->samplers
[sampler_idx
] & WINED3DSP_TEXTURETYPE_MASK
;
2933 glsl_sample_function_t sample_function
;
2935 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_ALL
, &src0_param
);
2937 shader_glsl_get_sample_function(sampler_type
, 0, &sample_function
);
2938 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, WINED3DVS_NOSWIZZLE
, NULL
,
2939 "%s.yz", src0_param
.reg_name
);
2942 /** Process the WINED3DSIO_TEXREG2RGB instruction in GLSL
2943 * Sample texture at dst using the rgb (xyz) components of src as texture coordinates */
2944 static void pshader_glsl_texreg2rgb(const struct wined3d_shader_instruction
*ins
)
2946 glsl_src_param_t src0_param
;
2947 DWORD sampler_idx
= ins
->dst
[0].register_idx
;
2948 DWORD sampler_type
= ins
->ctx
->reg_maps
->samplers
[sampler_idx
] & WINED3DSP_TEXTURETYPE_MASK
;
2949 glsl_sample_function_t sample_function
;
2951 /* Dependent read, not valid with conditional NP2 */
2952 shader_glsl_get_sample_function(sampler_type
, 0, &sample_function
);
2953 shader_glsl_add_src_param(ins
, &ins
->src
[0], sample_function
.coord_mask
, &src0_param
);
2955 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, WINED3DVS_NOSWIZZLE
, NULL
,
2956 "%s", src0_param
.param_str
);
2959 /** Process the WINED3DSIO_TEXKILL instruction in GLSL.
2960 * If any of the first 3 components are < 0, discard this pixel */
2961 static void pshader_glsl_texkill(const struct wined3d_shader_instruction
*ins
)
2963 glsl_dst_param_t dst_param
;
2965 /* The argument is a destination parameter, and no writemasks are allowed */
2966 shader_glsl_add_dst_param(ins
, &ins
->dst
[0], &dst_param
);
2967 if ((ins
->ctx
->reg_maps
->shader_version
>= WINED3DPS_VERSION(2,0)))
2969 /* 2.0 shaders compare all 4 components in texkill */
2970 shader_addline(ins
->ctx
->buffer
, "if (any(lessThan(%s.xyzw, vec4(0.0)))) discard;\n", dst_param
.reg_name
);
2972 /* 1.X shaders only compare the first 3 components, probably due to the nature of the texkill
2973 * instruction as a tex* instruction, and phase, which kills all a / w components. Even if all
2974 * 4 components are defined, only the first 3 are used
2976 shader_addline(ins
->ctx
->buffer
, "if (any(lessThan(%s.xyz, vec3(0.0)))) discard;\n", dst_param
.reg_name
);
2980 /** Process the WINED3DSIO_DP2ADD instruction in GLSL.
2981 * dst = dot2(src0, src1) + src2 */
2982 static void pshader_glsl_dp2add(const struct wined3d_shader_instruction
*ins
)
2984 glsl_src_param_t src0_param
;
2985 glsl_src_param_t src1_param
;
2986 glsl_src_param_t src2_param
;
2988 unsigned int mask_size
;
2990 write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2991 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
2993 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
, &src0_param
);
2994 shader_glsl_add_src_param(ins
, &ins
->src
[1], WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
, &src1_param
);
2995 shader_glsl_add_src_param(ins
, &ins
->src
[2], WINED3DSP_WRITEMASK_0
, &src2_param
);
2997 if (mask_size
> 1) {
2998 shader_addline(ins
->ctx
->buffer
, "vec%d(dot(%s, %s) + %s));\n",
2999 mask_size
, src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
3001 shader_addline(ins
->ctx
->buffer
, "dot(%s, %s) + %s);\n",
3002 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
3006 static void pshader_glsl_input_pack(IWineD3DPixelShader
*iface
, SHADER_BUFFER
*buffer
,
3007 const struct wined3d_shader_semantic
*semantics_in
, const struct shader_reg_maps
*reg_maps
,
3008 enum vertexprocessing_mode vertexprocessing
)
3011 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*)iface
;
3013 for (i
= 0; i
< MAX_REG_INPUT
; ++i
)
3015 DWORD usage
, usage_idx
;
3019 if (!reg_maps
->packed_input
[i
]) continue;
3021 usage
= semantics_in
[i
].usage
;
3022 usage_idx
= semantics_in
[i
].usage_idx
;
3023 shader_glsl_get_write_mask(&semantics_in
[i
].reg
, reg_mask
);
3027 case WINED3DDECLUSAGE_TEXCOORD
:
3028 if (usage_idx
< 8 && vertexprocessing
== pretransformed
)
3029 shader_addline(buffer
, "IN[%u]%s = gl_TexCoord[%u]%s;\n",
3030 This
->input_reg_map
[i
], reg_mask
, usage_idx
, reg_mask
);
3032 shader_addline(buffer
, "IN[%u]%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
3033 This
->input_reg_map
[i
], reg_mask
, reg_mask
);
3036 case WINED3DDECLUSAGE_COLOR
:
3038 shader_addline(buffer
, "IN[%u]%s = vec4(gl_Color)%s;\n",
3039 This
->input_reg_map
[i
], reg_mask
, reg_mask
);
3040 else if (usage_idx
== 1)
3041 shader_addline(buffer
, "IN[%u]%s = vec4(gl_SecondaryColor)%s;\n",
3042 This
->input_reg_map
[i
], reg_mask
, reg_mask
);
3044 shader_addline(buffer
, "IN[%u]%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
3045 This
->input_reg_map
[i
], reg_mask
, reg_mask
);
3049 shader_addline(buffer
, "IN[%u]%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
3050 This
->input_reg_map
[i
], reg_mask
, reg_mask
);
3056 /*********************************************
3057 * Vertex Shader Specific Code begins here
3058 ********************************************/
3060 static void add_glsl_program_entry(struct shader_glsl_priv
*priv
, struct glsl_shader_prog_link
*entry
) {
3061 glsl_program_key_t
*key
;
3063 key
= HeapAlloc(GetProcessHeap(), 0, sizeof(glsl_program_key_t
));
3064 key
->vshader
= entry
->vshader
;
3065 key
->pshader
= entry
->pshader
;
3066 key
->vs_args
= entry
->vs_args
;
3067 key
->ps_args
= entry
->ps_args
;
3069 hash_table_put(priv
->glsl_program_lookup
, key
, entry
);
3072 static struct glsl_shader_prog_link
*get_glsl_program_entry(struct shader_glsl_priv
*priv
,
3073 IWineD3DVertexShader
*vshader
, IWineD3DPixelShader
*pshader
, struct vs_compile_args
*vs_args
,
3074 struct ps_compile_args
*ps_args
) {
3075 glsl_program_key_t key
;
3077 key
.vshader
= vshader
;
3078 key
.pshader
= pshader
;
3079 key
.vs_args
= *vs_args
;
3080 key
.ps_args
= *ps_args
;
3082 return hash_table_get(priv
->glsl_program_lookup
, &key
);
3085 static void delete_glsl_program_entry(struct shader_glsl_priv
*priv
, const WineD3D_GL_Info
*gl_info
,
3086 struct glsl_shader_prog_link
*entry
)
3088 glsl_program_key_t
*key
;
3090 key
= HeapAlloc(GetProcessHeap(), 0, sizeof(glsl_program_key_t
));
3091 key
->vshader
= entry
->vshader
;
3092 key
->pshader
= entry
->pshader
;
3093 key
->vs_args
= entry
->vs_args
;
3094 key
->ps_args
= entry
->ps_args
;
3095 hash_table_remove(priv
->glsl_program_lookup
, key
);
3097 GL_EXTCALL(glDeleteObjectARB(entry
->programId
));
3098 if (entry
->vshader
) list_remove(&entry
->vshader_entry
);
3099 if (entry
->pshader
) list_remove(&entry
->pshader_entry
);
3100 HeapFree(GetProcessHeap(), 0, entry
->vuniformF_locations
);
3101 HeapFree(GetProcessHeap(), 0, entry
->puniformF_locations
);
3102 HeapFree(GetProcessHeap(), 0, entry
);
3105 static void handle_ps3_input(SHADER_BUFFER
*buffer
, const WineD3D_GL_Info
*gl_info
, const DWORD
*map
,
3106 const struct wined3d_shader_semantic
*semantics_in
, const struct shader_reg_maps
*reg_maps_in
,
3107 const struct wined3d_shader_semantic
*semantics_out
, const struct shader_reg_maps
*reg_maps_out
)
3110 DWORD usage
, usage_idx
, usage_out
, usage_idx_out
;
3113 DWORD in_count
= GL_LIMITS(glsl_varyings
) / 4;
3114 char reg_mask
[6], reg_mask_out
[6];
3115 char destination
[50];
3117 set
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*set
) * (in_count
+ 2));
3119 if (!semantics_out
) {
3120 /* Save gl_FrontColor & gl_FrontSecondaryColor before overwriting them. */
3121 shader_addline(buffer
, "vec4 front_color = gl_FrontColor;\n");
3122 shader_addline(buffer
, "vec4 front_secondary_color = gl_FrontSecondaryColor;\n");
3125 for(i
= 0; i
< MAX_REG_INPUT
; i
++) {
3126 if (!reg_maps_in
->packed_input
[i
]) continue;
3129 if (in_idx
>= (in_count
+ 2)) {
3130 FIXME("More input varyings declared than supported, expect issues\n");
3133 else if (map
[i
] == ~0U)
3135 /* Declared, but not read register */
3139 if (in_idx
== in_count
) {
3140 sprintf(destination
, "gl_FrontColor");
3141 } else if (in_idx
== in_count
+ 1) {
3142 sprintf(destination
, "gl_FrontSecondaryColor");
3144 sprintf(destination
, "IN[%u]", in_idx
);
3147 usage
= semantics_in
[i
].usage
;
3148 usage_idx
= semantics_in
[i
].usage_idx
;
3149 set
[map
[i
]] = shader_glsl_get_write_mask(&semantics_in
[i
].reg
, reg_mask
);
3151 if(!semantics_out
) {
3153 case WINED3DDECLUSAGE_COLOR
:
3155 shader_addline(buffer
, "%s%s = front_color%s;\n",
3156 destination
, reg_mask
, reg_mask
);
3157 else if (usage_idx
== 1)
3158 shader_addline(buffer
, "%s%s = front_secondary_color%s;\n",
3159 destination
, reg_mask
, reg_mask
);
3161 shader_addline(buffer
, "%s%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
3162 destination
, reg_mask
, reg_mask
);
3165 case WINED3DDECLUSAGE_TEXCOORD
:
3166 if (usage_idx
< 8) {
3167 shader_addline(buffer
, "%s%s = gl_TexCoord[%u]%s;\n",
3168 destination
, reg_mask
, usage_idx
, reg_mask
);
3170 shader_addline(buffer
, "%s%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
3171 destination
, reg_mask
, reg_mask
);
3175 case WINED3DDECLUSAGE_FOG
:
3176 shader_addline(buffer
, "%s%s = vec4(gl_FogFragCoord, 0.0, 0.0, 0.0)%s;\n",
3177 destination
, reg_mask
, reg_mask
);
3181 shader_addline(buffer
, "%s%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
3182 destination
, reg_mask
, reg_mask
);
3186 for(j
= 0; j
< MAX_REG_OUTPUT
; j
++) {
3187 if (!reg_maps_out
->packed_output
[j
]) continue;
3189 usage_out
= semantics_out
[j
].usage
;
3190 usage_idx_out
= semantics_out
[j
].usage_idx
;
3191 shader_glsl_get_write_mask(&semantics_out
[j
].reg
, reg_mask_out
);
3193 if(usage
== usage_out
&&
3194 usage_idx
== usage_idx_out
) {
3195 shader_addline(buffer
, "%s%s = OUT[%u]%s;\n",
3196 destination
, reg_mask
, j
, reg_mask
);
3201 shader_addline(buffer
, "%s%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
3202 destination
, reg_mask
, reg_mask
);
3207 /* This is solely to make the compiler / linker happy and avoid warning about undefined
3208 * varyings. It shouldn't result in any real code executed on the GPU, since all read
3209 * input varyings are assigned above, if the optimizer works properly.
3211 for(i
= 0; i
< in_count
+ 2; i
++) {
3212 if(set
[i
] != WINED3DSP_WRITEMASK_ALL
) {
3213 unsigned int size
= 0;
3214 memset(reg_mask
, 0, sizeof(reg_mask
));
3215 if(!(set
[i
] & WINED3DSP_WRITEMASK_0
)) {
3216 reg_mask
[size
] = 'x';
3219 if(!(set
[i
] & WINED3DSP_WRITEMASK_1
)) {
3220 reg_mask
[size
] = 'y';
3223 if(!(set
[i
] & WINED3DSP_WRITEMASK_2
)) {
3224 reg_mask
[size
] = 'z';
3227 if(!(set
[i
] & WINED3DSP_WRITEMASK_3
)) {
3228 reg_mask
[size
] = 'w';
3232 if (i
== in_count
) {
3233 sprintf(destination
, "gl_FrontColor");
3234 } else if (i
== in_count
+ 1) {
3235 sprintf(destination
, "gl_FrontSecondaryColor");
3237 sprintf(destination
, "IN[%u]", i
);
3241 shader_addline(buffer
, "%s.%s = 0.0;\n", destination
, reg_mask
);
3243 shader_addline(buffer
, "%s.%s = vec%u(0.0);\n", destination
, reg_mask
, size
);
3248 HeapFree(GetProcessHeap(), 0, set
);
3251 static GLhandleARB
generate_param_reorder_function(IWineD3DVertexShader
*vertexshader
,
3252 IWineD3DPixelShader
*pixelshader
, const WineD3D_GL_Info
*gl_info
)
3254 GLhandleARB ret
= 0;
3255 IWineD3DVertexShaderImpl
*vs
= (IWineD3DVertexShaderImpl
*) vertexshader
;
3256 IWineD3DPixelShaderImpl
*ps
= (IWineD3DPixelShaderImpl
*) pixelshader
;
3257 IWineD3DDeviceImpl
*device
;
3258 DWORD vs_major
= WINED3DSHADER_VERSION_MAJOR(vs
->baseShader
.reg_maps
.shader_version
);
3259 DWORD ps_major
= ps
? WINED3DSHADER_VERSION_MAJOR(ps
->baseShader
.reg_maps
.shader_version
) : 0;
3261 SHADER_BUFFER buffer
;
3262 DWORD usage
, usage_idx
, writemask
;
3264 const struct wined3d_shader_semantic
*semantics_out
;
3266 shader_buffer_init(&buffer
);
3268 shader_addline(&buffer
, "#version 120\n");
3270 if(vs_major
< 3 && ps_major
< 3) {
3271 /* That one is easy: The vertex shader writes to the builtin varyings, the pixel shader reads from them.
3272 * Take care about the texcoord .w fixup though if we're using the fixed function fragment pipeline
3274 device
= (IWineD3DDeviceImpl
*) vs
->baseShader
.device
;
3275 if((GLINFO_LOCATION
).set_texcoord_w
&& ps_major
== 0 && vs_major
> 0 &&
3276 !device
->frag_pipe
->ffp_proj_control
) {
3277 shader_addline(&buffer
, "void order_ps_input() {\n");
3278 for(i
= 0; i
< min(8, MAX_REG_TEXCRD
); i
++) {
3279 if(vs
->baseShader
.reg_maps
.texcoord_mask
[i
] != 0 &&
3280 vs
->baseShader
.reg_maps
.texcoord_mask
[i
] != WINED3DSP_WRITEMASK_ALL
) {
3281 shader_addline(&buffer
, "gl_TexCoord[%u].w = 1.0;\n", i
);
3284 shader_addline(&buffer
, "}\n");
3286 shader_addline(&buffer
, "void order_ps_input() { /* do nothing */ }\n");
3288 } else if(ps_major
< 3 && vs_major
>= 3) {
3289 /* The vertex shader writes to its own varyings, the pixel shader needs them in the builtin ones */
3290 semantics_out
= vs
->semantics_out
;
3292 shader_addline(&buffer
, "void order_ps_input(in vec4 OUT[%u]) {\n", MAX_REG_OUTPUT
);
3293 for(i
= 0; i
< MAX_REG_OUTPUT
; i
++) {
3294 if (!vs
->baseShader
.reg_maps
.packed_output
[i
]) continue;
3296 usage
= semantics_out
[i
].usage
;
3297 usage_idx
= semantics_out
[i
].usage_idx
;
3298 writemask
= shader_glsl_get_write_mask(&semantics_out
[i
].reg
, reg_mask
);
3301 case WINED3DDECLUSAGE_COLOR
:
3303 shader_addline(&buffer
, "gl_FrontColor%s = OUT[%u]%s;\n", reg_mask
, i
, reg_mask
);
3304 else if (usage_idx
== 1)
3305 shader_addline(&buffer
, "gl_FrontSecondaryColor%s = OUT[%u]%s;\n", reg_mask
, i
, reg_mask
);
3308 case WINED3DDECLUSAGE_POSITION
:
3309 shader_addline(&buffer
, "gl_Position%s = OUT[%u]%s;\n", reg_mask
, i
, reg_mask
);
3312 case WINED3DDECLUSAGE_TEXCOORD
:
3313 if (usage_idx
< 8) {
3314 if(!(GLINFO_LOCATION
).set_texcoord_w
|| ps_major
> 0) writemask
|= WINED3DSP_WRITEMASK_3
;
3316 shader_addline(&buffer
, "gl_TexCoord[%u]%s = OUT[%u]%s;\n",
3317 usage_idx
, reg_mask
, i
, reg_mask
);
3318 if(!(writemask
& WINED3DSP_WRITEMASK_3
)) {
3319 shader_addline(&buffer
, "gl_TexCoord[%u].w = 1.0;\n", usage_idx
);
3324 case WINED3DDECLUSAGE_PSIZE
:
3325 shader_addline(&buffer
, "gl_PointSize = OUT[%u].x;\n", i
);
3328 case WINED3DDECLUSAGE_FOG
:
3329 shader_addline(&buffer
, "gl_FogFragCoord = OUT[%u].%c;\n", i
, reg_mask
[1]);
3336 shader_addline(&buffer
, "}\n");
3338 } else if(ps_major
>= 3 && vs_major
>= 3) {
3339 semantics_out
= vs
->semantics_out
;
3341 /* This one is tricky: a 3.0 pixel shader reads from a 3.0 vertex shader */
3342 shader_addline(&buffer
, "varying vec4 IN[%u];\n", GL_LIMITS(glsl_varyings
) / 4);
3343 shader_addline(&buffer
, "void order_ps_input(in vec4 OUT[%u]) {\n", MAX_REG_OUTPUT
);
3345 /* First, sort out position and point size. Those are not passed to the pixel shader */
3346 for(i
= 0; i
< MAX_REG_OUTPUT
; i
++) {
3347 if (!vs
->baseShader
.reg_maps
.packed_output
[i
]) continue;
3349 usage
= semantics_out
[i
].usage
;
3350 usage_idx
= semantics_out
[i
].usage_idx
;
3351 shader_glsl_get_write_mask(&semantics_out
[i
].reg
, reg_mask
);
3354 case WINED3DDECLUSAGE_POSITION
:
3355 shader_addline(&buffer
, "gl_Position%s = OUT[%u]%s;\n", reg_mask
, i
, reg_mask
);
3358 case WINED3DDECLUSAGE_PSIZE
:
3359 shader_addline(&buffer
, "gl_PointSize = OUT[%u].x;\n", i
);
3367 /* Then, fix the pixel shader input */
3368 handle_ps3_input(&buffer
, gl_info
, ps
->input_reg_map
,
3369 ps
->semantics_in
, &ps
->baseShader
.reg_maps
, semantics_out
, &vs
->baseShader
.reg_maps
);
3371 shader_addline(&buffer
, "}\n");
3372 } else if(ps_major
>= 3 && vs_major
< 3) {
3373 shader_addline(&buffer
, "varying vec4 IN[%u];\n", GL_LIMITS(glsl_varyings
) / 4);
3374 shader_addline(&buffer
, "void order_ps_input() {\n");
3375 /* The vertex shader wrote to the builtin varyings. There is no need to figure out position and
3376 * point size, but we depend on the optimizers kindness to find out that the pixel shader doesn't
3377 * read gl_TexCoord and gl_ColorX, otherwise we'll run out of varyings
3379 handle_ps3_input(&buffer
, gl_info
, ps
->input_reg_map
, ps
->semantics_in
, &ps
->baseShader
.reg_maps
, NULL
, NULL
);
3380 shader_addline(&buffer
, "}\n");
3382 ERR("Unexpected vertex and pixel shader version condition: vs: %d, ps: %d\n", vs_major
, ps_major
);
3385 ret
= GL_EXTCALL(glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB
));
3386 checkGLcall("glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB)");
3387 GL_EXTCALL(glShaderSourceARB(ret
, 1, (const char**)&buffer
.buffer
, NULL
));
3388 checkGLcall("glShaderSourceARB(ret, 1, &buffer.buffer, NULL)");
3389 GL_EXTCALL(glCompileShaderARB(ret
));
3390 checkGLcall("glCompileShaderARB(ret)");
3392 shader_buffer_free(&buffer
);
3396 static void hardcode_local_constants(IWineD3DBaseShaderImpl
*shader
, const WineD3D_GL_Info
*gl_info
,
3397 GLhandleARB programId
, char prefix
)
3399 const local_constant
*lconst
;
3404 LIST_FOR_EACH_ENTRY(lconst
, &shader
->baseShader
.constantsF
, local_constant
, entry
) {
3405 value
= (const float *)lconst
->value
;
3406 snprintf(glsl_name
, sizeof(glsl_name
), "%cLC%u", prefix
, lconst
->idx
);
3407 tmp_loc
= GL_EXTCALL(glGetUniformLocationARB(programId
, glsl_name
));
3408 GL_EXTCALL(glUniform4fvARB(tmp_loc
, 1, value
));
3410 checkGLcall("Hardcoding local constants\n");
3413 /** Sets the GLSL program ID for the given pixel and vertex shader combination.
3414 * It sets the programId on the current StateBlock (because it should be called
3415 * inside of the DrawPrimitive() part of the render loop).
3417 * If a program for the given combination does not exist, create one, and store
3418 * the program in the hash table. If it creates a program, it will link the
3419 * given objects, too.
3421 static void set_glsl_shader_program(IWineD3DDevice
*iface
, BOOL use_ps
, BOOL use_vs
) {
3422 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
3423 struct shader_glsl_priv
*priv
= This
->shader_priv
;
3424 const WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
3425 IWineD3DPixelShader
*pshader
= This
->stateBlock
->pixelShader
;
3426 IWineD3DVertexShader
*vshader
= This
->stateBlock
->vertexShader
;
3427 struct glsl_shader_prog_link
*entry
= NULL
;
3428 GLhandleARB programId
= 0;
3429 GLhandleARB reorder_shader_id
= 0;
3432 GLhandleARB vshader_id
, pshader_id
;
3433 struct ps_compile_args ps_compile_args
;
3434 struct vs_compile_args vs_compile_args
;
3437 find_vs_compile_args((IWineD3DVertexShaderImpl
*)This
->stateBlock
->vertexShader
, This
->stateBlock
, &vs_compile_args
);
3439 /* FIXME: Do we really have to spend CPU cycles to generate a few zeroed bytes? */
3440 memset(&vs_compile_args
, 0, sizeof(vs_compile_args
));
3443 find_ps_compile_args((IWineD3DPixelShaderImpl
*)This
->stateBlock
->pixelShader
, This
->stateBlock
, &ps_compile_args
);
3445 /* FIXME: Do we really have to spend CPU cycles to generate a few zeroed bytes? */
3446 memset(&ps_compile_args
, 0, sizeof(ps_compile_args
));
3448 entry
= get_glsl_program_entry(priv
, vshader
, pshader
, &vs_compile_args
, &ps_compile_args
);
3450 priv
->glsl_program
= entry
;
3454 /* If we get to this point, then no matching program exists, so we create one */
3455 programId
= GL_EXTCALL(glCreateProgramObjectARB());
3456 TRACE("Created new GLSL shader program %u\n", programId
);
3458 /* Create the entry */
3459 entry
= HeapAlloc(GetProcessHeap(), 0, sizeof(struct glsl_shader_prog_link
));
3460 entry
->programId
= programId
;
3461 entry
->vshader
= vshader
;
3462 entry
->pshader
= pshader
;
3463 entry
->vs_args
= vs_compile_args
;
3464 entry
->ps_args
= ps_compile_args
;
3465 entry
->constant_version
= 0;
3466 /* Add the hash table entry */
3467 add_glsl_program_entry(priv
, entry
);
3469 /* Set the current program */
3470 priv
->glsl_program
= entry
;
3473 vshader_id
= find_gl_vshader((IWineD3DVertexShaderImpl
*) vshader
, &vs_compile_args
);
3478 /* Attach GLSL vshader */
3480 const unsigned int max_attribs
= 16; /* TODO: Will this always be the case? It is at the moment... */
3483 reorder_shader_id
= generate_param_reorder_function(vshader
, pshader
, gl_info
);
3484 TRACE("Attaching GLSL shader object %u to program %u\n", reorder_shader_id
, programId
);
3485 GL_EXTCALL(glAttachObjectARB(programId
, reorder_shader_id
));
3486 checkGLcall("glAttachObjectARB");
3487 /* Flag the reorder function for deletion, then it will be freed automatically when the program
3490 GL_EXTCALL(glDeleteObjectARB(reorder_shader_id
));
3492 TRACE("Attaching GLSL shader object %u to program %u\n", vshader_id
, programId
);
3493 GL_EXTCALL(glAttachObjectARB(programId
, vshader_id
));
3494 checkGLcall("glAttachObjectARB");
3496 /* Bind vertex attributes to a corresponding index number to match
3497 * the same index numbers as ARB_vertex_programs (makes loading
3498 * vertex attributes simpler). With this method, we can use the
3499 * exact same code to load the attributes later for both ARB and
3502 * We have to do this here because we need to know the Program ID
3503 * in order to make the bindings work, and it has to be done prior
3504 * to linking the GLSL program. */
3505 for (i
= 0; i
< max_attribs
; ++i
) {
3506 if (((IWineD3DBaseShaderImpl
*)vshader
)->baseShader
.reg_maps
.attributes
[i
]) {
3507 snprintf(tmp_name
, sizeof(tmp_name
), "attrib%i", i
);
3508 GL_EXTCALL(glBindAttribLocationARB(programId
, i
, tmp_name
));
3511 checkGLcall("glBindAttribLocationARB");
3513 list_add_head(&((IWineD3DBaseShaderImpl
*)vshader
)->baseShader
.linked_programs
, &entry
->vshader_entry
);
3517 pshader_id
= find_gl_pshader((IWineD3DPixelShaderImpl
*) pshader
, &ps_compile_args
);
3522 /* Attach GLSL pshader */
3524 TRACE("Attaching GLSL shader object %u to program %u\n", pshader_id
, programId
);
3525 GL_EXTCALL(glAttachObjectARB(programId
, pshader_id
));
3526 checkGLcall("glAttachObjectARB");
3528 list_add_head(&((IWineD3DBaseShaderImpl
*)pshader
)->baseShader
.linked_programs
, &entry
->pshader_entry
);
3531 /* Link the program */
3532 TRACE("Linking GLSL shader program %u\n", programId
);
3533 GL_EXTCALL(glLinkProgramARB(programId
));
3534 print_glsl_info_log(&GLINFO_LOCATION
, programId
);
3536 entry
->vuniformF_locations
= HeapAlloc(GetProcessHeap(), 0, sizeof(GLhandleARB
) * GL_LIMITS(vshader_constantsF
));
3537 for (i
= 0; i
< GL_LIMITS(vshader_constantsF
); ++i
) {
3538 snprintf(glsl_name
, sizeof(glsl_name
), "VC[%i]", i
);
3539 entry
->vuniformF_locations
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, glsl_name
));
3541 for (i
= 0; i
< MAX_CONST_I
; ++i
) {
3542 snprintf(glsl_name
, sizeof(glsl_name
), "VI[%i]", i
);
3543 entry
->vuniformI_locations
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, glsl_name
));
3545 entry
->puniformF_locations
= HeapAlloc(GetProcessHeap(), 0, sizeof(GLhandleARB
) * GL_LIMITS(pshader_constantsF
));
3546 for (i
= 0; i
< GL_LIMITS(pshader_constantsF
); ++i
) {
3547 snprintf(glsl_name
, sizeof(glsl_name
), "PC[%i]", i
);
3548 entry
->puniformF_locations
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, glsl_name
));
3550 for (i
= 0; i
< MAX_CONST_I
; ++i
) {
3551 snprintf(glsl_name
, sizeof(glsl_name
), "PI[%i]", i
);
3552 entry
->puniformI_locations
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, glsl_name
));
3556 for(i
= 0; i
< ((IWineD3DPixelShaderImpl
*)pshader
)->numbumpenvmatconsts
; i
++) {
3558 sprintf(name
, "bumpenvmat%d", ((IWineD3DPixelShaderImpl
*)pshader
)->bumpenvmatconst
[i
].texunit
);
3559 entry
->bumpenvmat_location
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, name
));
3560 sprintf(name
, "luminancescale%d", ((IWineD3DPixelShaderImpl
*)pshader
)->luminanceconst
[i
].texunit
);
3561 entry
->luminancescale_location
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, name
));
3562 sprintf(name
, "luminanceoffset%d", ((IWineD3DPixelShaderImpl
*)pshader
)->luminanceconst
[i
].texunit
);
3563 entry
->luminanceoffset_location
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, name
));
3567 if (use_ps
&& ps_compile_args
.np2_fixup
) {
3569 for (i
= 0; i
< MAX_FRAGMENT_SAMPLERS
; ++i
) {
3570 if (ps_compile_args
.np2_fixup
& (1 << i
)) {
3571 sprintf(name
, "PsamplerNP2Fixup%u", i
);
3572 entry
->np2Fixup_location
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, name
));
3574 entry
->np2Fixup_location
[i
] = -1;
3579 entry
->posFixup_location
= GL_EXTCALL(glGetUniformLocationARB(programId
, "posFixup"));
3580 entry
->ycorrection_location
= GL_EXTCALL(glGetUniformLocationARB(programId
, "ycorrection"));
3581 checkGLcall("Find glsl program uniform locations");
3584 && WINED3DSHADER_VERSION_MAJOR(((IWineD3DPixelShaderImpl
*)pshader
)->baseShader
.reg_maps
.shader_version
) >= 3
3585 && ((IWineD3DPixelShaderImpl
*)pshader
)->declared_in_count
> GL_LIMITS(glsl_varyings
) / 4)
3587 TRACE("Shader %d needs vertex color clamping disabled\n", programId
);
3588 entry
->vertex_color_clamp
= GL_FALSE
;
3590 entry
->vertex_color_clamp
= GL_FIXED_ONLY_ARB
;
3593 /* Set the shader to allow uniform loading on it */
3594 GL_EXTCALL(glUseProgramObjectARB(programId
));
3595 checkGLcall("glUseProgramObjectARB(programId)");
3597 /* Load the vertex and pixel samplers now. The function that finds the mappings makes sure
3598 * that it stays the same for each vertexshader-pixelshader pair(=linked glsl program). If
3599 * a pshader with fixed function pipeline is used there are no vertex samplers, and if a
3600 * vertex shader with fixed function pixel processing is used we make sure that the card
3601 * supports enough samplers to allow the max number of vertex samplers with all possible
3602 * fixed function fragment processing setups. So once the program is linked these samplers
3606 /* Load vertex shader samplers */
3607 shader_glsl_load_vsamplers(gl_info
, This
->texUnitMap
, programId
);
3610 /* Load pixel shader samplers */
3611 shader_glsl_load_psamplers(gl_info
, This
->texUnitMap
, programId
);
3614 /* If the local constants do not have to be loaded with the environment constants,
3615 * load them now to have them hardcoded in the GLSL program. This saves some CPU cycles
3618 if(pshader
&& !((IWineD3DPixelShaderImpl
*)pshader
)->baseShader
.load_local_constsF
) {
3619 hardcode_local_constants((IWineD3DBaseShaderImpl
*) pshader
, gl_info
, programId
, 'P');
3621 if(vshader
&& !((IWineD3DVertexShaderImpl
*)vshader
)->baseShader
.load_local_constsF
) {
3622 hardcode_local_constants((IWineD3DBaseShaderImpl
*) vshader
, gl_info
, programId
, 'V');
3626 static GLhandleARB
create_glsl_blt_shader(const WineD3D_GL_Info
*gl_info
, enum tex_types tex_type
)
3628 GLhandleARB program_id
;
3629 GLhandleARB vshader_id
, pshader_id
;
3630 static const char *blt_vshader
[] =
3635 " gl_Position = gl_Vertex;\n"
3636 " gl_FrontColor = vec4(1.0);\n"
3637 " gl_TexCoord[0] = gl_MultiTexCoord0;\n"
3641 static const char *blt_pshaders
[tex_type_count
] =
3647 "uniform sampler2D sampler;\n"
3650 " gl_FragDepth = texture2D(sampler, gl_TexCoord[0].xy).x;\n"
3656 "uniform samplerCube sampler;\n"
3659 " gl_FragDepth = textureCube(sampler, gl_TexCoord[0].xyz).x;\n"
3663 "#extension GL_ARB_texture_rectangle : enable\n"
3664 "uniform sampler2DRect sampler;\n"
3667 " gl_FragDepth = texture2DRect(sampler, gl_TexCoord[0].xy).x;\n"
3671 if (!blt_pshaders
[tex_type
])
3673 FIXME("tex_type %#x not supported\n", tex_type
);
3677 vshader_id
= GL_EXTCALL(glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB
));
3678 GL_EXTCALL(glShaderSourceARB(vshader_id
, 1, blt_vshader
, NULL
));
3679 GL_EXTCALL(glCompileShaderARB(vshader_id
));
3681 pshader_id
= GL_EXTCALL(glCreateShaderObjectARB(GL_FRAGMENT_SHADER_ARB
));
3682 GL_EXTCALL(glShaderSourceARB(pshader_id
, 1, &blt_pshaders
[tex_type
], NULL
));
3683 GL_EXTCALL(glCompileShaderARB(pshader_id
));
3685 program_id
= GL_EXTCALL(glCreateProgramObjectARB());
3686 GL_EXTCALL(glAttachObjectARB(program_id
, vshader_id
));
3687 GL_EXTCALL(glAttachObjectARB(program_id
, pshader_id
));
3688 GL_EXTCALL(glLinkProgramARB(program_id
));
3690 print_glsl_info_log(&GLINFO_LOCATION
, program_id
);
3692 /* Once linked we can mark the shaders for deletion. They will be deleted once the program
3695 GL_EXTCALL(glDeleteObjectARB(vshader_id
));
3696 GL_EXTCALL(glDeleteObjectARB(pshader_id
));
3700 static void shader_glsl_select(IWineD3DDevice
*iface
, BOOL usePS
, BOOL useVS
) {
3701 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
3702 struct shader_glsl_priv
*priv
= This
->shader_priv
;
3703 const WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
3704 GLhandleARB program_id
= 0;
3705 GLenum old_vertex_color_clamp
, current_vertex_color_clamp
;
3707 old_vertex_color_clamp
= priv
->glsl_program
? priv
->glsl_program
->vertex_color_clamp
: GL_FIXED_ONLY_ARB
;
3709 if (useVS
|| usePS
) set_glsl_shader_program(iface
, usePS
, useVS
);
3710 else priv
->glsl_program
= NULL
;
3712 current_vertex_color_clamp
= priv
->glsl_program
? priv
->glsl_program
->vertex_color_clamp
: GL_FIXED_ONLY_ARB
;
3714 if (old_vertex_color_clamp
!= current_vertex_color_clamp
) {
3715 if (GL_SUPPORT(ARB_COLOR_BUFFER_FLOAT
)) {
3716 GL_EXTCALL(glClampColorARB(GL_CLAMP_VERTEX_COLOR_ARB
, current_vertex_color_clamp
));
3717 checkGLcall("glClampColorARB");
3719 FIXME("vertex color clamp needs to be changed, but extension not supported.\n");
3723 program_id
= priv
->glsl_program
? priv
->glsl_program
->programId
: 0;
3724 if (program_id
) TRACE("Using GLSL program %u\n", program_id
);
3725 GL_EXTCALL(glUseProgramObjectARB(program_id
));
3726 checkGLcall("glUseProgramObjectARB");
3729 static void shader_glsl_select_depth_blt(IWineD3DDevice
*iface
, enum tex_types tex_type
) {
3730 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
3731 const WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
3732 struct shader_glsl_priv
*priv
= This
->shader_priv
;
3733 GLhandleARB
*blt_program
= &priv
->depth_blt_program
[tex_type
];
3735 if (!*blt_program
) {
3737 *blt_program
= create_glsl_blt_shader(gl_info
, tex_type
);
3738 loc
= GL_EXTCALL(glGetUniformLocationARB(*blt_program
, "sampler"));
3739 GL_EXTCALL(glUseProgramObjectARB(*blt_program
));
3740 GL_EXTCALL(glUniform1iARB(loc
, 0));
3742 GL_EXTCALL(glUseProgramObjectARB(*blt_program
));
3746 static void shader_glsl_deselect_depth_blt(IWineD3DDevice
*iface
) {
3747 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
3748 const WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
3749 struct shader_glsl_priv
*priv
= This
->shader_priv
;
3750 GLhandleARB program_id
;
3752 program_id
= priv
->glsl_program
? priv
->glsl_program
->programId
: 0;
3753 if (program_id
) TRACE("Using GLSL program %u\n", program_id
);
3755 GL_EXTCALL(glUseProgramObjectARB(program_id
));
3756 checkGLcall("glUseProgramObjectARB");
3759 static void shader_glsl_destroy(IWineD3DBaseShader
*iface
) {
3760 const struct list
*linked_programs
;
3761 IWineD3DBaseShaderImpl
*This
= (IWineD3DBaseShaderImpl
*) iface
;
3762 IWineD3DDeviceImpl
*device
= (IWineD3DDeviceImpl
*)This
->baseShader
.device
;
3763 struct shader_glsl_priv
*priv
= device
->shader_priv
;
3764 const WineD3D_GL_Info
*gl_info
= &device
->adapter
->gl_info
;
3765 IWineD3DPixelShaderImpl
*ps
= NULL
;
3766 IWineD3DVertexShaderImpl
*vs
= NULL
;
3768 /* Note: Do not use QueryInterface here to find out which shader type this is because this code
3769 * can be called from IWineD3DBaseShader::Release
3771 char pshader
= shader_is_pshader_version(This
->baseShader
.reg_maps
.shader_version
);
3774 ps
= (IWineD3DPixelShaderImpl
*) This
;
3775 if(ps
->num_gl_shaders
== 0) return;
3776 if (priv
->glsl_program
&& (IWineD3DBaseShader
*)priv
->glsl_program
->pshader
== iface
)
3777 shader_glsl_select(This
->baseShader
.device
, FALSE
, FALSE
);
3779 vs
= (IWineD3DVertexShaderImpl
*) This
;
3780 if(vs
->num_gl_shaders
== 0) return;
3781 if (priv
->glsl_program
&& (IWineD3DBaseShader
*)priv
->glsl_program
->vshader
== iface
)
3782 shader_glsl_select(This
->baseShader
.device
, FALSE
, FALSE
);
3785 linked_programs
= &This
->baseShader
.linked_programs
;
3787 TRACE("Deleting linked programs\n");
3788 if (linked_programs
->next
) {
3789 struct glsl_shader_prog_link
*entry
, *entry2
;
3792 LIST_FOR_EACH_ENTRY_SAFE(entry
, entry2
, linked_programs
, struct glsl_shader_prog_link
, pshader_entry
) {
3793 delete_glsl_program_entry(priv
, gl_info
, entry
);
3796 LIST_FOR_EACH_ENTRY_SAFE(entry
, entry2
, linked_programs
, struct glsl_shader_prog_link
, vshader_entry
) {
3797 delete_glsl_program_entry(priv
, gl_info
, entry
);
3806 for(i
= 0; i
< ps
->num_gl_shaders
; i
++) {
3807 TRACE("deleting pshader %u\n", ps
->gl_shaders
[i
].prgId
);
3808 GL_EXTCALL(glDeleteObjectARB(ps
->gl_shaders
[i
].prgId
));
3809 checkGLcall("glDeleteObjectARB");
3812 HeapFree(GetProcessHeap(), 0, ps
->gl_shaders
);
3813 ps
->gl_shaders
= NULL
;
3814 ps
->num_gl_shaders
= 0;
3815 ps
->shader_array_size
= 0;
3820 for(i
= 0; i
< vs
->num_gl_shaders
; i
++) {
3821 TRACE("deleting vshader %u\n", vs
->gl_shaders
[i
].prgId
);
3822 GL_EXTCALL(glDeleteObjectARB(vs
->gl_shaders
[i
].prgId
));
3823 checkGLcall("glDeleteObjectARB");
3826 HeapFree(GetProcessHeap(), 0, vs
->gl_shaders
);
3827 vs
->gl_shaders
= NULL
;
3828 vs
->num_gl_shaders
= 0;
3829 vs
->shader_array_size
= 0;
3833 static unsigned int glsl_program_key_hash(const void *key
)
3835 const glsl_program_key_t
*k
= key
;
3837 unsigned int hash
= ((DWORD_PTR
) k
->vshader
) | ((DWORD_PTR
) k
->pshader
) << 16;
3838 hash
+= ~(hash
<< 15);
3839 hash
^= (hash
>> 10);
3840 hash
+= (hash
<< 3);
3841 hash
^= (hash
>> 6);
3842 hash
+= ~(hash
<< 11);
3843 hash
^= (hash
>> 16);
3848 static BOOL
glsl_program_key_compare(const void *keya
, const void *keyb
)
3850 const glsl_program_key_t
*ka
= keya
;
3851 const glsl_program_key_t
*kb
= keyb
;
3853 return ka
->vshader
== kb
->vshader
&& ka
->pshader
== kb
->pshader
&&
3854 (memcmp(&ka
->ps_args
, &kb
->ps_args
, sizeof(kb
->ps_args
)) == 0) &&
3855 (memcmp(&ka
->vs_args
, &kb
->vs_args
, sizeof(kb
->vs_args
)) == 0);
3858 static BOOL
constant_heap_init(struct constant_heap
*heap
, unsigned int constant_count
)
3860 SIZE_T size
= (constant_count
+ 1) * sizeof(*heap
->entries
) + constant_count
* sizeof(*heap
->positions
);
3861 void *mem
= HeapAlloc(GetProcessHeap(), 0, size
);
3865 ERR("Failed to allocate memory\n");
3869 heap
->entries
= mem
;
3870 heap
->entries
[1].version
= 0;
3871 heap
->positions
= (unsigned int *)(heap
->entries
+ constant_count
+ 1);
3877 static void constant_heap_free(struct constant_heap
*heap
)
3879 HeapFree(GetProcessHeap(), 0, heap
->entries
);
3882 static HRESULT
shader_glsl_alloc(IWineD3DDevice
*iface
) {
3883 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
3884 const WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
3885 struct shader_glsl_priv
*priv
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(struct shader_glsl_priv
));
3886 SIZE_T stack_size
= wined3d_log2i(max(GL_LIMITS(vshader_constantsF
), GL_LIMITS(pshader_constantsF
))) + 1;
3888 priv
->stack
= HeapAlloc(GetProcessHeap(), 0, stack_size
* sizeof(*priv
->stack
));
3891 ERR("Failed to allocate memory.\n");
3892 HeapFree(GetProcessHeap(), 0, priv
);
3893 return E_OUTOFMEMORY
;
3896 if (!constant_heap_init(&priv
->vconst_heap
, GL_LIMITS(vshader_constantsF
)))
3898 ERR("Failed to initialize vertex shader constant heap\n");
3899 HeapFree(GetProcessHeap(), 0, priv
->stack
);
3900 HeapFree(GetProcessHeap(), 0, priv
);
3901 return E_OUTOFMEMORY
;
3904 if (!constant_heap_init(&priv
->pconst_heap
, GL_LIMITS(pshader_constantsF
)))
3906 ERR("Failed to initialize pixel shader constant heap\n");
3907 constant_heap_free(&priv
->vconst_heap
);
3908 HeapFree(GetProcessHeap(), 0, priv
->stack
);
3909 HeapFree(GetProcessHeap(), 0, priv
);
3910 return E_OUTOFMEMORY
;
3913 priv
->glsl_program_lookup
= hash_table_create(glsl_program_key_hash
, glsl_program_key_compare
);
3914 priv
->next_constant_version
= 1;
3916 This
->shader_priv
= priv
;
3920 static void shader_glsl_free(IWineD3DDevice
*iface
) {
3921 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
3922 const WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
3923 struct shader_glsl_priv
*priv
= This
->shader_priv
;
3926 for (i
= 0; i
< tex_type_count
; ++i
)
3928 if (priv
->depth_blt_program
[i
])
3930 GL_EXTCALL(glDeleteObjectARB(priv
->depth_blt_program
[i
]));
3934 hash_table_destroy(priv
->glsl_program_lookup
, NULL
, NULL
);
3935 constant_heap_free(&priv
->pconst_heap
);
3936 constant_heap_free(&priv
->vconst_heap
);
3938 HeapFree(GetProcessHeap(), 0, This
->shader_priv
);
3939 This
->shader_priv
= NULL
;
3942 static BOOL
shader_glsl_dirty_const(IWineD3DDevice
*iface
) {
3943 /* TODO: GL_EXT_bindable_uniform can be used to share constants across shaders */
3947 static GLuint
shader_glsl_generate_pshader(IWineD3DPixelShader
*iface
, SHADER_BUFFER
*buffer
, const struct ps_compile_args
*args
) {
3948 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*)iface
;
3949 const struct shader_reg_maps
*reg_maps
= &This
->baseShader
.reg_maps
;
3950 CONST DWORD
*function
= This
->baseShader
.function
;
3951 const char *fragcolor
;
3952 const WineD3D_GL_Info
*gl_info
= &((IWineD3DDeviceImpl
*)This
->baseShader
.device
)->adapter
->gl_info
;
3954 /* Create the hw GLSL shader object and assign it as the shader->prgId */
3955 GLhandleARB shader_obj
= GL_EXTCALL(glCreateShaderObjectARB(GL_FRAGMENT_SHADER_ARB
));
3957 shader_addline(buffer
, "#version 120\n");
3959 if (GL_SUPPORT(ARB_DRAW_BUFFERS
)) {
3960 shader_addline(buffer
, "#extension GL_ARB_draw_buffers : enable\n");
3962 if (GL_SUPPORT(ARB_TEXTURE_RECTANGLE
)) {
3963 /* The spec says that it doesn't have to be explicitly enabled, but the nvidia
3964 * drivers write a warning if we don't do so
3966 shader_addline(buffer
, "#extension GL_ARB_texture_rectangle : enable\n");
3969 /* Base Declarations */
3970 shader_generate_glsl_declarations( (IWineD3DBaseShader
*) This
, reg_maps
, buffer
, &GLINFO_LOCATION
, args
);
3972 /* Pack 3.0 inputs */
3973 if (reg_maps
->shader_version
>= WINED3DPS_VERSION(3,0) && args
->vp_mode
!= vertexshader
) {
3974 pshader_glsl_input_pack(iface
, buffer
, This
->semantics_in
, reg_maps
, args
->vp_mode
);
3977 /* Base Shader Body */
3978 shader_generate_main( (IWineD3DBaseShader
*) This
, buffer
, reg_maps
, function
);
3980 /* Pixel shaders < 2.0 place the resulting color in R0 implicitly */
3981 if (reg_maps
->shader_version
< WINED3DPS_VERSION(2,0))
3983 /* Some older cards like GeforceFX ones don't support multiple buffers, so also not gl_FragData */
3984 if(GL_SUPPORT(ARB_DRAW_BUFFERS
))
3985 shader_addline(buffer
, "gl_FragData[0] = R0;\n");
3987 shader_addline(buffer
, "gl_FragColor = R0;\n");
3990 if(GL_SUPPORT(ARB_DRAW_BUFFERS
)) {
3991 fragcolor
= "gl_FragData[0]";
3993 fragcolor
= "gl_FragColor";
3995 if(args
->srgb_correction
) {
3996 shader_addline(buffer
, "tmp0.xyz = pow(%s.xyz, vec3(%f, %f, %f)) * vec3(%f, %f, %f) - vec3(%f, %f, %f);\n",
3997 fragcolor
, srgb_pow
, srgb_pow
, srgb_pow
, srgb_mul_high
, srgb_mul_high
, srgb_mul_high
,
3998 srgb_sub_high
, srgb_sub_high
, srgb_sub_high
);
3999 shader_addline(buffer
, "tmp1.xyz = %s.xyz * srgb_mul_low.xyz;\n", fragcolor
);
4000 shader_addline(buffer
, "%s.x = %s.x < srgb_comparison.x ? tmp1.x : tmp0.x;\n", fragcolor
, fragcolor
);
4001 shader_addline(buffer
, "%s.y = %s.y < srgb_comparison.y ? tmp1.y : tmp0.y;\n", fragcolor
, fragcolor
);
4002 shader_addline(buffer
, "%s.z = %s.z < srgb_comparison.z ? tmp1.z : tmp0.z;\n", fragcolor
, fragcolor
);
4003 shader_addline(buffer
, "%s = clamp(%s, 0.0, 1.0);\n", fragcolor
, fragcolor
);
4005 /* Pixel shader < 3.0 do not replace the fog stage.
4006 * This implements linear fog computation and blending.
4007 * TODO: non linear fog
4008 * NOTE: gl_Fog.start and gl_Fog.end don't hold fog start s and end e but
4009 * -1/(e-s) and e/(e-s) respectively.
4011 if(reg_maps
->shader_version
< WINED3DPS_VERSION(3,0)) {
4013 case FOG_OFF
: break;
4015 shader_addline(buffer
, "float fogstart = -1.0 / (gl_Fog.end - gl_Fog.start);\n");
4016 shader_addline(buffer
, "float fogend = gl_Fog.end * -fogstart;\n");
4017 shader_addline(buffer
, "float Fog = clamp(gl_FogFragCoord * fogstart + fogend, 0.0, 1.0);\n");
4018 shader_addline(buffer
, "%s.xyz = mix(gl_Fog.color.xyz, %s.xyz, Fog);\n", fragcolor
, fragcolor
);
4021 /* Fog = e^(-gl_Fog.density * gl_FogFragCoord) */
4022 shader_addline(buffer
, "float Fog = exp(-gl_Fog.density * gl_FogFragCoord);\n");
4023 shader_addline(buffer
, "Fog = clamp(Fog, 0.0, 1.0);\n");
4024 shader_addline(buffer
, "%s.xyz = mix(gl_Fog.color.xyz, %s.xyz, Fog);\n", fragcolor
, fragcolor
);
4027 /* Fog = e^(-(gl_Fog.density * gl_FogFragCoord)^2) */
4028 shader_addline(buffer
, "float Fog = exp(-gl_Fog.density * gl_Fog.density * gl_FogFragCoord * gl_FogFragCoord);\n");
4029 shader_addline(buffer
, "Fog = clamp(Fog, 0.0, 1.0);\n");
4030 shader_addline(buffer
, "%s.xyz = mix(gl_Fog.color.xyz, %s.xyz, Fog);\n", fragcolor
, fragcolor
);
4035 shader_addline(buffer
, "}\n");
4037 TRACE("Compiling shader object %u\n", shader_obj
);
4038 GL_EXTCALL(glShaderSourceARB(shader_obj
, 1, (const char**)&buffer
->buffer
, NULL
));
4039 GL_EXTCALL(glCompileShaderARB(shader_obj
));
4040 print_glsl_info_log(&GLINFO_LOCATION
, shader_obj
);
4042 /* Store the shader object */
4046 static GLuint
shader_glsl_generate_vshader(IWineD3DVertexShader
*iface
, SHADER_BUFFER
*buffer
, const struct vs_compile_args
*args
) {
4047 IWineD3DVertexShaderImpl
*This
= (IWineD3DVertexShaderImpl
*)iface
;
4048 const struct shader_reg_maps
*reg_maps
= &This
->baseShader
.reg_maps
;
4049 CONST DWORD
*function
= This
->baseShader
.function
;
4050 const WineD3D_GL_Info
*gl_info
= &((IWineD3DDeviceImpl
*)This
->baseShader
.device
)->adapter
->gl_info
;
4052 /* Create the hw GLSL shader program and assign it as the shader->prgId */
4053 GLhandleARB shader_obj
= GL_EXTCALL(glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB
));
4055 shader_addline(buffer
, "#version 120\n");
4057 /* Base Declarations */
4058 shader_generate_glsl_declarations( (IWineD3DBaseShader
*) This
, reg_maps
, buffer
, &GLINFO_LOCATION
, NULL
);
4060 /* Base Shader Body */
4061 shader_generate_main( (IWineD3DBaseShader
*) This
, buffer
, reg_maps
, function
);
4063 /* Unpack 3.0 outputs */
4064 if (reg_maps
->shader_version
>= WINED3DVS_VERSION(3,0)) shader_addline(buffer
, "order_ps_input(OUT);\n");
4065 else shader_addline(buffer
, "order_ps_input();\n");
4067 /* The D3DRS_FOGTABLEMODE render state defines if the shader-generated fog coord is used
4068 * or if the fragment depth is used. If the fragment depth is used(FOGTABLEMODE != NONE),
4069 * the fog frag coord is thrown away. If the fog frag coord is used, but not written by
4070 * the shader, it is set to 0.0(fully fogged, since start = 1.0, end = 0.0)
4072 if(args
->fog_src
== VS_FOG_Z
) {
4073 shader_addline(buffer
, "gl_FogFragCoord = gl_Position.z;\n");
4074 } else if (!reg_maps
->fog
) {
4075 shader_addline(buffer
, "gl_FogFragCoord = 0.0;\n");
4078 /* Write the final position.
4080 * OpenGL coordinates specify the center of the pixel while d3d coords specify
4081 * the corner. The offsets are stored in z and w in posFixup. posFixup.y contains
4082 * 1.0 or -1.0 to turn the rendering upside down for offscreen rendering. PosFixup.x
4083 * contains 1.0 to allow a mad.
4085 shader_addline(buffer
, "gl_Position.y = gl_Position.y * posFixup.y;\n");
4086 shader_addline(buffer
, "gl_Position.xy += posFixup.zw * gl_Position.ww;\n");
4088 /* Z coord [0;1]->[-1;1] mapping, see comment in transform_projection in state.c
4090 * Basically we want (in homogeneous coordinates) z = z * 2 - 1. However, shaders are run
4091 * before the homogeneous divide, so we have to take the w into account: z = ((z / w) * 2 - 1) * w,
4092 * which is the same as z = z * 2 - w.
4094 shader_addline(buffer
, "gl_Position.z = gl_Position.z * 2.0 - gl_Position.w;\n");
4096 shader_addline(buffer
, "}\n");
4098 TRACE("Compiling shader object %u\n", shader_obj
);
4099 GL_EXTCALL(glShaderSourceARB(shader_obj
, 1, (const char**)&buffer
->buffer
, NULL
));
4100 GL_EXTCALL(glCompileShaderARB(shader_obj
));
4101 print_glsl_info_log(&GLINFO_LOCATION
, shader_obj
);
4106 static void shader_glsl_get_caps(WINED3DDEVTYPE devtype
, const WineD3D_GL_Info
*gl_info
, struct shader_caps
*pCaps
)
4108 /* Nvidia Geforce6/7 or Ati R4xx/R5xx cards with GLSL support, support VS 3.0 but older Nvidia/Ati
4109 * models with GLSL support only support 2.0. In case of nvidia we can detect VS 2.0 support using
4110 * vs_nv_version which is based on NV_vertex_program.
4111 * For Ati cards there's no way using glsl (it abstracts the lowlevel info away) and also not
4112 * using ARB_vertex_program. It is safe to assume that when a card supports pixel shader 2.0 it
4113 * supports vertex shader 2.0 too and the way around. We can detect ps2.0 using the maximum number
4114 * of native instructions, so use that here. For more info see the pixel shader versioning code below.
4116 if((GLINFO_LOCATION
.vs_nv_version
== VS_VERSION_20
) || (GLINFO_LOCATION
.ps_arb_max_instructions
<= 512))
4117 pCaps
->VertexShaderVersion
= WINED3DVS_VERSION(2,0);
4119 pCaps
->VertexShaderVersion
= WINED3DVS_VERSION(3,0);
4120 TRACE_(d3d_caps
)("Hardware vertex shader version %d.%d enabled (GLSL)\n", (pCaps
->VertexShaderVersion
>> 8) & 0xff, pCaps
->VertexShaderVersion
& 0xff);
4121 /* Subtract the other potential uniforms from the max available (bools, ints, and 1 row of projection matrix) */
4122 pCaps
->MaxVertexShaderConst
= GL_LIMITS(vshader_constantsF
) - (MAX_CONST_B
/ 4) - MAX_CONST_I
- 1;
4124 /* Older DX9-class videocards (GeforceFX / Radeon >9500/X*00) only support pixel shader 2.0/2.0a/2.0b.
4125 * In OpenGL the extensions related to GLSL abstract lowlevel GL info away which is needed
4126 * to distinguish between 2.0 and 3.0 (and 2.0a/2.0b). In case of Nvidia we use their fragment
4127 * program extensions. On other hardware including ATI GL_ARB_fragment_program offers the info
4128 * in max native instructions. Intel and others also offer the info in this extension but they
4129 * don't support GLSL (at least on Windows).
4131 * PS2.0 requires at least 96 instructions, 2.0a/2.0b go up to 512. Assume that if the number
4132 * of instructions is 512 or less we have to do with ps2.0 hardware.
4133 * NOTE: ps3.0 hardware requires 512 or more instructions but ati and nvidia offer 'enough' (1024 vs 4096) on their most basic ps3.0 hardware.
4135 if((GLINFO_LOCATION
.ps_nv_version
== PS_VERSION_20
) || (GLINFO_LOCATION
.ps_arb_max_instructions
<= 512))
4136 pCaps
->PixelShaderVersion
= WINED3DPS_VERSION(2,0);
4138 pCaps
->PixelShaderVersion
= WINED3DPS_VERSION(3,0);
4140 /* Subtract the other potential uniforms from the max available (bools & ints), and 2 states for fog.
4141 * In theory the texbem instruction may need one more shader constant too. But lets assume
4142 * that a sm <= 1.3 shader does not need all the uniforms provided by a glsl-capable card,
4143 * and lets not take away a uniform needlessly from all other shaders.
4145 pCaps
->MaxPixelShaderConst
= GL_LIMITS(pshader_constantsF
) - (MAX_CONST_B
/ 4) - MAX_CONST_I
- 2;
4147 /* FIXME: The following line is card dependent. -8.0 to 8.0 is the
4148 * Direct3D minimum requirement.
4150 * Both GL_ARB_fragment_program and GLSL require a "maximum representable magnitude"
4151 * of colors to be 2^10, and 2^32 for other floats. Should we use 1024 here?
4153 * The problem is that the refrast clamps temporary results in the shader to
4154 * [-MaxValue;+MaxValue]. If the card's max value is bigger than the one we advertize here,
4155 * then applications may miss the clamping behavior. On the other hand, if it is smaller,
4156 * the shader will generate incorrect results too. Unfortunately, GL deliberately doesn't
4157 * offer a way to query this.
4159 pCaps
->PixelShader1xMaxValue
= 8.0;
4160 TRACE_(d3d_caps
)("Hardware pixel shader version %d.%d enabled (GLSL)\n", (pCaps
->PixelShaderVersion
>> 8) & 0xff, pCaps
->PixelShaderVersion
& 0xff);
4163 static BOOL
shader_glsl_color_fixup_supported(struct color_fixup_desc fixup
)
4165 if (TRACE_ON(d3d_shader
) && TRACE_ON(d3d
))
4167 TRACE("Checking support for fixup:\n");
4168 dump_color_fixup_desc(fixup
);
4171 /* We support everything except YUV conversions. */
4172 if (!is_yuv_fixup(fixup
))
4178 TRACE("[FAILED]\n");
4182 static const SHADER_HANDLER shader_glsl_instruction_handler_table
[WINED3DSIH_TABLE_SIZE
] =
4184 /* WINED3DSIH_ABS */ shader_glsl_map2gl
,
4185 /* WINED3DSIH_ADD */ shader_glsl_arith
,
4186 /* WINED3DSIH_BEM */ pshader_glsl_bem
,
4187 /* WINED3DSIH_BREAK */ shader_glsl_break
,
4188 /* WINED3DSIH_BREAKC */ shader_glsl_breakc
,
4189 /* WINED3DSIH_BREAKP */ NULL
,
4190 /* WINED3DSIH_CALL */ shader_glsl_call
,
4191 /* WINED3DSIH_CALLNZ */ shader_glsl_callnz
,
4192 /* WINED3DSIH_CMP */ shader_glsl_cmp
,
4193 /* WINED3DSIH_CND */ shader_glsl_cnd
,
4194 /* WINED3DSIH_CRS */ shader_glsl_cross
,
4195 /* WINED3DSIH_DCL */ NULL
,
4196 /* WINED3DSIH_DEF */ NULL
,
4197 /* WINED3DSIH_DEFB */ NULL
,
4198 /* WINED3DSIH_DEFI */ NULL
,
4199 /* WINED3DSIH_DP2ADD */ pshader_glsl_dp2add
,
4200 /* WINED3DSIH_DP3 */ shader_glsl_dot
,
4201 /* WINED3DSIH_DP4 */ shader_glsl_dot
,
4202 /* WINED3DSIH_DST */ shader_glsl_dst
,
4203 /* WINED3DSIH_DSX */ shader_glsl_map2gl
,
4204 /* WINED3DSIH_DSY */ shader_glsl_map2gl
,
4205 /* WINED3DSIH_ELSE */ shader_glsl_else
,
4206 /* WINED3DSIH_ENDIF */ shader_glsl_end
,
4207 /* WINED3DSIH_ENDLOOP */ shader_glsl_end
,
4208 /* WINED3DSIH_ENDREP */ shader_glsl_end
,
4209 /* WINED3DSIH_EXP */ shader_glsl_map2gl
,
4210 /* WINED3DSIH_EXPP */ shader_glsl_expp
,
4211 /* WINED3DSIH_FRC */ shader_glsl_map2gl
,
4212 /* WINED3DSIH_IF */ shader_glsl_if
,
4213 /* WINED3DSIH_IFC */ shader_glsl_ifc
,
4214 /* WINED3DSIH_LABEL */ shader_glsl_label
,
4215 /* WINED3DSIH_LIT */ shader_glsl_lit
,
4216 /* WINED3DSIH_LOG */ shader_glsl_log
,
4217 /* WINED3DSIH_LOGP */ shader_glsl_log
,
4218 /* WINED3DSIH_LOOP */ shader_glsl_loop
,
4219 /* WINED3DSIH_LRP */ shader_glsl_lrp
,
4220 /* WINED3DSIH_M3x2 */ shader_glsl_mnxn
,
4221 /* WINED3DSIH_M3x3 */ shader_glsl_mnxn
,
4222 /* WINED3DSIH_M3x4 */ shader_glsl_mnxn
,
4223 /* WINED3DSIH_M4x3 */ shader_glsl_mnxn
,
4224 /* WINED3DSIH_M4x4 */ shader_glsl_mnxn
,
4225 /* WINED3DSIH_MAD */ shader_glsl_mad
,
4226 /* WINED3DSIH_MAX */ shader_glsl_map2gl
,
4227 /* WINED3DSIH_MIN */ shader_glsl_map2gl
,
4228 /* WINED3DSIH_MOV */ shader_glsl_mov
,
4229 /* WINED3DSIH_MOVA */ shader_glsl_mov
,
4230 /* WINED3DSIH_MUL */ shader_glsl_arith
,
4231 /* WINED3DSIH_NOP */ NULL
,
4232 /* WINED3DSIH_NRM */ shader_glsl_map2gl
,
4233 /* WINED3DSIH_PHASE */ NULL
,
4234 /* WINED3DSIH_POW */ shader_glsl_pow
,
4235 /* WINED3DSIH_RCP */ shader_glsl_rcp
,
4236 /* WINED3DSIH_REP */ shader_glsl_rep
,
4237 /* WINED3DSIH_RET */ NULL
,
4238 /* WINED3DSIH_RSQ */ shader_glsl_rsq
,
4239 /* WINED3DSIH_SETP */ NULL
,
4240 /* WINED3DSIH_SGE */ shader_glsl_compare
,
4241 /* WINED3DSIH_SGN */ shader_glsl_map2gl
,
4242 /* WINED3DSIH_SINCOS */ shader_glsl_sincos
,
4243 /* WINED3DSIH_SLT */ shader_glsl_compare
,
4244 /* WINED3DSIH_SUB */ shader_glsl_arith
,
4245 /* WINED3DSIH_TEX */ pshader_glsl_tex
,
4246 /* WINED3DSIH_TEXBEM */ pshader_glsl_texbem
,
4247 /* WINED3DSIH_TEXBEML */ pshader_glsl_texbem
,
4248 /* WINED3DSIH_TEXCOORD */ pshader_glsl_texcoord
,
4249 /* WINED3DSIH_TEXDEPTH */ pshader_glsl_texdepth
,
4250 /* WINED3DSIH_TEXDP3 */ pshader_glsl_texdp3
,
4251 /* WINED3DSIH_TEXDP3TEX */ pshader_glsl_texdp3tex
,
4252 /* WINED3DSIH_TEXKILL */ pshader_glsl_texkill
,
4253 /* WINED3DSIH_TEXLDD */ NULL
,
4254 /* WINED3DSIH_TEXLDL */ shader_glsl_texldl
,
4255 /* WINED3DSIH_TEXM3x2DEPTH */ pshader_glsl_texm3x2depth
,
4256 /* WINED3DSIH_TEXM3x2PAD */ pshader_glsl_texm3x2pad
,
4257 /* WINED3DSIH_TEXM3x2TEX */ pshader_glsl_texm3x2tex
,
4258 /* WINED3DSIH_TEXM3x3 */ pshader_glsl_texm3x3
,
4259 /* WINED3DSIH_TEXM3x3DIFF */ NULL
,
4260 /* WINED3DSIH_TEXM3x3PAD */ pshader_glsl_texm3x3pad
,
4261 /* WINED3DSIH_TEXM3x3SPEC */ pshader_glsl_texm3x3spec
,
4262 /* WINED3DSIH_TEXM3x3TEX */ pshader_glsl_texm3x3tex
,
4263 /* WINED3DSIH_TEXM3x3VSPEC */ pshader_glsl_texm3x3vspec
,
4264 /* WINED3DSIH_TEXREG2AR */ pshader_glsl_texreg2ar
,
4265 /* WINED3DSIH_TEXREG2GB */ pshader_glsl_texreg2gb
,
4266 /* WINED3DSIH_TEXREG2RGB */ pshader_glsl_texreg2rgb
,
4269 const shader_backend_t glsl_shader_backend
= {
4270 shader_glsl_instruction_handler_table
,
4272 shader_glsl_select_depth_blt
,
4273 shader_glsl_deselect_depth_blt
,
4274 shader_glsl_update_float_vertex_constants
,
4275 shader_glsl_update_float_pixel_constants
,
4276 shader_glsl_load_constants
,
4277 shader_glsl_load_np2fixup_constants
,
4278 shader_glsl_destroy
,
4281 shader_glsl_dirty_const
,
4282 shader_glsl_generate_pshader
,
4283 shader_glsl_generate_vshader
,
4284 shader_glsl_get_caps
,
4285 shader_glsl_color_fixup_supported
,