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(DWORD src_modifier
, const char *in_reg
, const char *in_regswizzle
, char *out_str
)
974 switch (src_modifier
)
976 case WINED3DSPSM_DZ
: /* Need to handle this in the instructions itself (texld & texcrd). */
978 case WINED3DSPSM_NONE
:
979 sprintf(out_str
, "%s%s", in_reg
, in_regswizzle
);
981 case WINED3DSPSM_NEG
:
982 sprintf(out_str
, "-%s%s", in_reg
, in_regswizzle
);
984 case WINED3DSPSM_NOT
:
985 sprintf(out_str
, "!%s%s", in_reg
, in_regswizzle
);
987 case WINED3DSPSM_BIAS
:
988 sprintf(out_str
, "(%s%s - vec4(0.5)%s)", in_reg
, in_regswizzle
, in_regswizzle
);
990 case WINED3DSPSM_BIASNEG
:
991 sprintf(out_str
, "-(%s%s - vec4(0.5)%s)", in_reg
, in_regswizzle
, in_regswizzle
);
993 case WINED3DSPSM_SIGN
:
994 sprintf(out_str
, "(2.0 * (%s%s - 0.5))", in_reg
, in_regswizzle
);
996 case WINED3DSPSM_SIGNNEG
:
997 sprintf(out_str
, "-(2.0 * (%s%s - 0.5))", in_reg
, in_regswizzle
);
999 case WINED3DSPSM_COMP
:
1000 sprintf(out_str
, "(1.0 - %s%s)", in_reg
, in_regswizzle
);
1002 case WINED3DSPSM_X2
:
1003 sprintf(out_str
, "(2.0 * %s%s)", in_reg
, in_regswizzle
);
1005 case WINED3DSPSM_X2NEG
:
1006 sprintf(out_str
, "-(2.0 * %s%s)", in_reg
, in_regswizzle
);
1008 case WINED3DSPSM_ABS
:
1009 sprintf(out_str
, "abs(%s%s)", in_reg
, in_regswizzle
);
1011 case WINED3DSPSM_ABSNEG
:
1012 sprintf(out_str
, "-abs(%s%s)", in_reg
, in_regswizzle
);
1015 FIXME("Unhandled modifier %u\n", src_modifier
);
1016 sprintf(out_str
, "%s%s", in_reg
, in_regswizzle
);
1020 /** Writes the GLSL variable name that corresponds to the register that the
1021 * DX opcode parameter is trying to access */
1022 static void shader_glsl_get_register_name(WINED3DSHADER_PARAM_REGISTER_TYPE register_type
, UINT register_idx
,
1023 const struct wined3d_shader_src_param
*rel_addr
, char *register_name
, BOOL
*is_color
,
1024 const struct wined3d_shader_instruction
*ins
)
1026 /* oPos, oFog and oPts in D3D */
1027 static const char * const hwrastout_reg_names
[] = { "gl_Position", "gl_FogFragCoord", "gl_PointSize" };
1029 IWineD3DBaseShaderImpl
*This
= (IWineD3DBaseShaderImpl
*)ins
->ctx
->shader
;
1030 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
1031 const WineD3D_GL_Info
* gl_info
= &deviceImpl
->adapter
->gl_info
;
1032 DWORD shader_version
= This
->baseShader
.reg_maps
.shader_version
;
1033 char pshader
= shader_is_pshader_version(shader_version
);
1037 switch (register_type
)
1039 case WINED3DSPR_TEMP
:
1040 sprintf(register_name
, "R%u", register_idx
);
1042 case WINED3DSPR_INPUT
:
1044 /* Pixel shaders >= 3.0 */
1045 if (WINED3DSHADER_VERSION_MAJOR(shader_version
) >= 3)
1047 DWORD in_count
= GL_LIMITS(glsl_varyings
) / 4;
1051 glsl_src_param_t rel_param
;
1052 shader_glsl_add_src_param(ins
, rel_addr
, WINED3DSP_WRITEMASK_0
, &rel_param
);
1054 /* Removing a + 0 would be an obvious optimization, but macos doesn't see the NOP
1057 if (((IWineD3DPixelShaderImpl
*)This
)->input_reg_map
[register_idx
])
1059 if (((IWineD3DPixelShaderImpl
*)This
)->declared_in_count
> in_count
) {
1060 sprintf(register_name
, "((%s + %u) > %d ? (%s + %u) > %d ? gl_SecondaryColor : gl_Color : IN[%s + %u])",
1061 rel_param
.param_str
, ((IWineD3DPixelShaderImpl
*)This
)->input_reg_map
[register_idx
], in_count
- 1,
1062 rel_param
.param_str
, ((IWineD3DPixelShaderImpl
*)This
)->input_reg_map
[register_idx
], in_count
,
1063 rel_param
.param_str
, ((IWineD3DPixelShaderImpl
*)This
)->input_reg_map
[register_idx
]);
1065 sprintf(register_name
, "IN[%s + %u]", rel_param
.param_str
,
1066 ((IWineD3DPixelShaderImpl
*)This
)->input_reg_map
[register_idx
]);
1069 if (((IWineD3DPixelShaderImpl
*)This
)->declared_in_count
> in_count
) {
1070 sprintf(register_name
, "((%s) > %d ? (%s) > %d ? gl_SecondaryColor : gl_Color : IN[%s])",
1071 rel_param
.param_str
, in_count
- 1,
1072 rel_param
.param_str
, in_count
,
1073 rel_param
.param_str
);
1075 sprintf(register_name
, "IN[%s]", rel_param
.param_str
);
1079 DWORD idx
= ((IWineD3DPixelShaderImpl
*) This
)->input_reg_map
[register_idx
];
1080 if (idx
== in_count
) {
1081 sprintf(register_name
, "gl_Color");
1082 } else if (idx
== in_count
+ 1) {
1083 sprintf(register_name
, "gl_SecondaryColor");
1085 sprintf(register_name
, "IN[%u]", idx
);
1089 if (register_idx
== 0)
1090 strcpy(register_name
, "gl_Color");
1092 strcpy(register_name
, "gl_SecondaryColor");
1095 if (((IWineD3DVertexShaderImpl
*)This
)->cur_args
->swizzle_map
& (1 << register_idx
)) *is_color
= TRUE
;
1096 sprintf(register_name
, "attrib%u", register_idx
);
1099 case WINED3DSPR_CONST
:
1101 const char prefix
= pshader
? 'P':'V';
1103 /* Relative addressing */
1106 glsl_src_param_t rel_param
;
1107 shader_glsl_add_src_param(ins
, rel_addr
, WINED3DSP_WRITEMASK_0
, &rel_param
);
1109 sprintf(register_name
, "%cC[%s + %u]", prefix
, rel_param
.param_str
, register_idx
);
1111 sprintf(register_name
, "%cC[%s]", prefix
, rel_param
.param_str
);
1113 if (shader_constant_is_local(This
, register_idx
))
1115 sprintf(register_name
, "%cLC%u", prefix
, register_idx
);
1117 sprintf(register_name
, "%cC[%u]", prefix
, register_idx
);
1123 case WINED3DSPR_CONSTINT
:
1125 sprintf(register_name
, "PI[%u]", register_idx
);
1127 sprintf(register_name
, "VI[%u]", register_idx
);
1129 case WINED3DSPR_CONSTBOOL
:
1131 sprintf(register_name
, "PB[%u]", register_idx
);
1133 sprintf(register_name
, "VB[%u]", register_idx
);
1135 case WINED3DSPR_TEXTURE
: /* case WINED3DSPR_ADDR: */
1137 sprintf(register_name
, "T%u", register_idx
);
1139 sprintf(register_name
, "A%u", register_idx
);
1142 case WINED3DSPR_LOOP
:
1143 sprintf(register_name
, "aL%u", This
->baseShader
.cur_loop_regno
- 1);
1145 case WINED3DSPR_SAMPLER
:
1147 sprintf(register_name
, "Psampler%u", register_idx
);
1149 sprintf(register_name
, "Vsampler%u", register_idx
);
1151 case WINED3DSPR_COLOROUT
:
1152 if (register_idx
>= GL_LIMITS(buffers
))
1153 WARN("Write to render target %u, only %d supported\n", register_idx
, 4);
1155 if (GL_SUPPORT(ARB_DRAW_BUFFERS
)) {
1156 sprintf(register_name
, "gl_FragData[%u]", register_idx
);
1157 } else { /* On older cards with GLSL support like the GeforceFX there's only one buffer. */
1158 sprintf(register_name
, "gl_FragColor");
1161 case WINED3DSPR_RASTOUT
:
1162 sprintf(register_name
, "%s", hwrastout_reg_names
[register_idx
]);
1164 case WINED3DSPR_DEPTHOUT
:
1165 sprintf(register_name
, "gl_FragDepth");
1167 case WINED3DSPR_ATTROUT
:
1168 if (register_idx
== 0)
1170 sprintf(register_name
, "gl_FrontColor");
1172 sprintf(register_name
, "gl_FrontSecondaryColor");
1175 case WINED3DSPR_TEXCRDOUT
:
1176 /* Vertex shaders >= 3.0: WINED3DSPR_OUTPUT */
1177 if (WINED3DSHADER_VERSION_MAJOR(shader_version
) >= 3) sprintf(register_name
, "OUT[%u]", register_idx
);
1178 else sprintf(register_name
, "gl_TexCoord[%u]", register_idx
);
1180 case WINED3DSPR_MISCTYPE
:
1181 if (register_idx
== 0)
1184 sprintf(register_name
, "vpos");
1186 else if (register_idx
== 1)
1188 /* Note that gl_FrontFacing is a bool, while vFace is
1189 * a float for which the sign determines front/back
1191 sprintf(register_name
, "(gl_FrontFacing ? 1.0 : -1.0)");
1193 FIXME("Unhandled misctype register %d\n", register_idx
);
1194 sprintf(register_name
, "unrecognized_register");
1198 FIXME("Unhandled register name Type(%d)\n", register_type
);
1199 sprintf(register_name
, "unrecognized_register");
1204 static void shader_glsl_write_mask_to_str(DWORD write_mask
, char *str
)
1207 if (write_mask
& WINED3DSP_WRITEMASK_0
) *str
++ = 'x';
1208 if (write_mask
& WINED3DSP_WRITEMASK_1
) *str
++ = 'y';
1209 if (write_mask
& WINED3DSP_WRITEMASK_2
) *str
++ = 'z';
1210 if (write_mask
& WINED3DSP_WRITEMASK_3
) *str
++ = 'w';
1214 /* Get the GLSL write mask for the destination register */
1215 static DWORD
shader_glsl_get_write_mask(const struct wined3d_shader_dst_param
*param
, char *write_mask
)
1217 DWORD mask
= param
->write_mask
;
1219 if (shader_is_scalar(param
->register_type
, param
->register_idx
))
1221 mask
= WINED3DSP_WRITEMASK_0
;
1226 shader_glsl_write_mask_to_str(mask
, write_mask
);
1232 static unsigned int shader_glsl_get_write_mask_size(DWORD write_mask
) {
1233 unsigned int size
= 0;
1235 if (write_mask
& WINED3DSP_WRITEMASK_0
) ++size
;
1236 if (write_mask
& WINED3DSP_WRITEMASK_1
) ++size
;
1237 if (write_mask
& WINED3DSP_WRITEMASK_2
) ++size
;
1238 if (write_mask
& WINED3DSP_WRITEMASK_3
) ++size
;
1243 static void shader_glsl_get_swizzle(const DWORD param
, BOOL fixup
, DWORD mask
, char *swizzle_str
) {
1244 /* For registers of type WINED3DDECLTYPE_D3DCOLOR, data is stored as "bgra",
1245 * but addressed as "rgba". To fix this we need to swap the register's x
1246 * and z components. */
1247 DWORD swizzle
= (param
& WINED3DSP_SWIZZLE_MASK
) >> WINED3DSP_SWIZZLE_SHIFT
;
1248 const char *swizzle_chars
= fixup
? "zyxw" : "xyzw";
1249 char *ptr
= swizzle_str
;
1251 if (!shader_is_scalar(shader_get_regtype(param
), param
& WINED3DSP_REGNUM_MASK
))
1254 /* swizzle bits fields: wwzzyyxx */
1255 if (mask
& WINED3DSP_WRITEMASK_0
) *ptr
++ = swizzle_chars
[swizzle
& 0x03];
1256 if (mask
& WINED3DSP_WRITEMASK_1
) *ptr
++ = swizzle_chars
[(swizzle
>> 2) & 0x03];
1257 if (mask
& WINED3DSP_WRITEMASK_2
) *ptr
++ = swizzle_chars
[(swizzle
>> 4) & 0x03];
1258 if (mask
& WINED3DSP_WRITEMASK_3
) *ptr
++ = swizzle_chars
[(swizzle
>> 6) & 0x03];
1264 /* From a given parameter token, generate the corresponding GLSL string.
1265 * Also, return the actual register name and swizzle in case the
1266 * caller needs this information as well. */
1267 static void shader_glsl_add_src_param(const struct wined3d_shader_instruction
*ins
,
1268 const struct wined3d_shader_src_param
*wined3d_src
, DWORD mask
, glsl_src_param_t
*glsl_src
)
1270 BOOL is_color
= FALSE
;
1271 char swizzle_str
[6];
1273 glsl_src
->reg_name
[0] = '\0';
1274 glsl_src
->param_str
[0] = '\0';
1275 swizzle_str
[0] = '\0';
1277 shader_glsl_get_register_name(wined3d_src
->register_type
, wined3d_src
->register_idx
,
1278 wined3d_src
->rel_addr
, glsl_src
->reg_name
, &is_color
, ins
);
1280 shader_glsl_get_swizzle(wined3d_src
->token
, is_color
, mask
, swizzle_str
);
1281 shader_glsl_gen_modifier(wined3d_src
->modifiers
, glsl_src
->reg_name
, swizzle_str
, glsl_src
->param_str
);
1284 /* From a given parameter token, generate the corresponding GLSL string.
1285 * Also, return the actual register name and swizzle in case the
1286 * caller needs this information as well. */
1287 static DWORD
shader_glsl_add_dst_param(const struct wined3d_shader_instruction
*ins
,
1288 const struct wined3d_shader_dst_param
*wined3d_dst
, glsl_dst_param_t
*glsl_dst
)
1290 BOOL is_color
= FALSE
;
1292 glsl_dst
->mask_str
[0] = '\0';
1293 glsl_dst
->reg_name
[0] = '\0';
1295 shader_glsl_get_register_name(wined3d_dst
->register_type
, wined3d_dst
->register_idx
,
1296 wined3d_dst
->rel_addr
, glsl_dst
->reg_name
, &is_color
, ins
);
1297 return shader_glsl_get_write_mask(wined3d_dst
, glsl_dst
->mask_str
);
1300 /* Append the destination part of the instruction to the buffer, return the effective write mask */
1301 static DWORD
shader_glsl_append_dst_ext(SHADER_BUFFER
*buffer
,
1302 const struct wined3d_shader_instruction
*ins
, const struct wined3d_shader_dst_param
*dst
)
1304 glsl_dst_param_t glsl_dst
;
1307 mask
= shader_glsl_add_dst_param(ins
, dst
, &glsl_dst
);
1308 if (mask
) shader_addline(buffer
, "%s%s = %s(", glsl_dst
.reg_name
, glsl_dst
.mask_str
, shift_glsl_tab
[dst
->shift
]);
1313 /* Append the destination part of the instruction to the buffer, return the effective write mask */
1314 static DWORD
shader_glsl_append_dst(SHADER_BUFFER
*buffer
, const struct wined3d_shader_instruction
*ins
)
1316 return shader_glsl_append_dst_ext(buffer
, ins
, &ins
->dst
[0]);
1319 /** Process GLSL instruction modifiers */
1320 void shader_glsl_add_instruction_modifiers(const struct wined3d_shader_instruction
*ins
)
1322 glsl_dst_param_t dst_param
;
1325 if (!ins
->dst_count
) return;
1327 modifiers
= ins
->dst
[0].modifiers
;
1328 if (!modifiers
) return;
1330 shader_glsl_add_dst_param(ins
, &ins
->dst
[0], &dst_param
);
1332 if (modifiers
& WINED3DSPDM_SATURATE
)
1334 /* _SAT means to clamp the value of the register to between 0 and 1 */
1335 shader_addline(ins
->ctx
->buffer
, "%s%s = clamp(%s%s, 0.0, 1.0);\n", dst_param
.reg_name
,
1336 dst_param
.mask_str
, dst_param
.reg_name
, dst_param
.mask_str
);
1339 if (modifiers
& WINED3DSPDM_MSAMPCENTROID
)
1341 FIXME("_centroid modifier not handled\n");
1344 if (modifiers
& WINED3DSPDM_PARTIALPRECISION
)
1346 /* MSDN says this modifier can be safely ignored, so that's what we'll do. */
1350 static inline const char *shader_get_comp_op(DWORD flags
)
1352 DWORD op
= (flags
& INST_CONTROLS_MASK
) >> INST_CONTROLS_SHIFT
;
1354 case COMPARISON_GT
: return ">";
1355 case COMPARISON_EQ
: return "==";
1356 case COMPARISON_GE
: return ">=";
1357 case COMPARISON_LT
: return "<";
1358 case COMPARISON_NE
: return "!=";
1359 case COMPARISON_LE
: return "<=";
1361 FIXME("Unrecognized comparison value: %u\n", op
);
1366 static void shader_glsl_get_sample_function(DWORD sampler_type
, DWORD flags
, glsl_sample_function_t
*sample_function
)
1368 BOOL projected
= flags
& WINED3D_GLSL_SAMPLE_PROJECTED
;
1369 BOOL texrect
= flags
& WINED3D_GLSL_SAMPLE_RECT
;
1370 BOOL lod
= flags
& WINED3D_GLSL_SAMPLE_LOD
;
1372 /* Note that there's no such thing as a projected cube texture. */
1373 switch(sampler_type
) {
1376 sample_function
->name
= projected
? "texture1DProjLod" : "texture1DLod";
1378 sample_function
->name
= projected
? "texture1DProj" : "texture1D";
1380 sample_function
->coord_mask
= WINED3DSP_WRITEMASK_0
;
1385 sample_function
->name
= projected
? "texture2DRectProjLod" : "texture2DRectLod";
1387 sample_function
->name
= projected
? "texture2DRectProj" : "texture2DRect";
1391 sample_function
->name
= projected
? "texture2DProjLod" : "texture2DLod";
1393 sample_function
->name
= projected
? "texture2DProj" : "texture2D";
1396 sample_function
->coord_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
;
1398 case WINED3DSTT_CUBE
:
1400 sample_function
->name
= "textureCubeLod";
1402 sample_function
->name
= "textureCube";
1404 sample_function
->coord_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
1406 case WINED3DSTT_VOLUME
:
1408 sample_function
->name
= projected
? "texture3DProjLod" : "texture3DLod";
1410 sample_function
->name
= projected
? "texture3DProj" : "texture3D";
1412 sample_function
->coord_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
1415 sample_function
->name
= "";
1416 sample_function
->coord_mask
= 0;
1417 FIXME("Unrecognized sampler type: %#x;\n", sampler_type
);
1422 static void shader_glsl_append_fixup_arg(char *arguments
, const char *reg_name
,
1423 BOOL sign_fixup
, enum fixup_channel_source channel_source
)
1425 switch(channel_source
)
1427 case CHANNEL_SOURCE_ZERO
:
1428 strcat(arguments
, "0.0");
1431 case CHANNEL_SOURCE_ONE
:
1432 strcat(arguments
, "1.0");
1435 case CHANNEL_SOURCE_X
:
1436 strcat(arguments
, reg_name
);
1437 strcat(arguments
, ".x");
1440 case CHANNEL_SOURCE_Y
:
1441 strcat(arguments
, reg_name
);
1442 strcat(arguments
, ".y");
1445 case CHANNEL_SOURCE_Z
:
1446 strcat(arguments
, reg_name
);
1447 strcat(arguments
, ".z");
1450 case CHANNEL_SOURCE_W
:
1451 strcat(arguments
, reg_name
);
1452 strcat(arguments
, ".w");
1456 FIXME("Unhandled channel source %#x\n", channel_source
);
1457 strcat(arguments
, "undefined");
1461 if (sign_fixup
) strcat(arguments
, " * 2.0 - 1.0");
1464 static void shader_glsl_color_correction(const struct wined3d_shader_instruction
*ins
, struct color_fixup_desc fixup
)
1466 struct wined3d_shader_dst_param dst
;
1467 unsigned int mask_size
, remaining
;
1468 glsl_dst_param_t dst_param
;
1469 char arguments
[256];
1473 if (fixup
.x_sign_fixup
|| fixup
.x_source
!= CHANNEL_SOURCE_X
) mask
|= WINED3DSP_WRITEMASK_0
;
1474 if (fixup
.y_sign_fixup
|| fixup
.y_source
!= CHANNEL_SOURCE_Y
) mask
|= WINED3DSP_WRITEMASK_1
;
1475 if (fixup
.z_sign_fixup
|| fixup
.z_source
!= CHANNEL_SOURCE_Z
) mask
|= WINED3DSP_WRITEMASK_2
;
1476 if (fixup
.w_sign_fixup
|| fixup
.w_source
!= CHANNEL_SOURCE_W
) mask
|= WINED3DSP_WRITEMASK_3
;
1477 mask
&= ins
->dst
[0].write_mask
;
1479 if (!mask
) return; /* Nothing to do */
1481 if (is_yuv_fixup(fixup
))
1483 enum yuv_fixup yuv_fixup
= get_yuv_fixup(fixup
);
1484 FIXME("YUV fixup (%#x) not supported\n", yuv_fixup
);
1488 mask_size
= shader_glsl_get_write_mask_size(mask
);
1491 dst
.write_mask
= mask
;
1492 shader_glsl_add_dst_param(ins
, &dst
, &dst_param
);
1494 arguments
[0] = '\0';
1495 remaining
= mask_size
;
1496 if (mask
& WINED3DSP_WRITEMASK_0
)
1498 shader_glsl_append_fixup_arg(arguments
, dst_param
.reg_name
, fixup
.x_sign_fixup
, fixup
.x_source
);
1499 if (--remaining
) strcat(arguments
, ", ");
1501 if (mask
& WINED3DSP_WRITEMASK_1
)
1503 shader_glsl_append_fixup_arg(arguments
, dst_param
.reg_name
, fixup
.y_sign_fixup
, fixup
.y_source
);
1504 if (--remaining
) strcat(arguments
, ", ");
1506 if (mask
& WINED3DSP_WRITEMASK_2
)
1508 shader_glsl_append_fixup_arg(arguments
, dst_param
.reg_name
, fixup
.z_sign_fixup
, fixup
.z_source
);
1509 if (--remaining
) strcat(arguments
, ", ");
1511 if (mask
& WINED3DSP_WRITEMASK_3
)
1513 shader_glsl_append_fixup_arg(arguments
, dst_param
.reg_name
, fixup
.w_sign_fixup
, fixup
.w_source
);
1514 if (--remaining
) strcat(arguments
, ", ");
1519 shader_addline(ins
->ctx
->buffer
, "%s%s = vec%u(%s);\n",
1520 dst_param
.reg_name
, dst_param
.mask_str
, mask_size
, arguments
);
1524 shader_addline(ins
->ctx
->buffer
, "%s%s = %s;\n", dst_param
.reg_name
, dst_param
.mask_str
, arguments
);
1528 static void PRINTF_ATTR(6, 7) shader_glsl_gen_sample_code(const struct wined3d_shader_instruction
*ins
,
1529 DWORD sampler
, const glsl_sample_function_t
*sample_function
, DWORD swizzle
,
1530 const char *bias
, const char *coord_reg_fmt
, ...)
1532 const char *sampler_base
;
1533 char dst_swizzle
[6];
1534 struct color_fixup_desc fixup
;
1535 BOOL np2_fixup
= FALSE
;
1538 shader_glsl_get_swizzle(swizzle
, FALSE
, ins
->dst
[0].write_mask
, dst_swizzle
);
1540 if (shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
))
1542 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*)ins
->ctx
->shader
;
1543 fixup
= This
->cur_args
->color_fixup
[sampler
];
1544 sampler_base
= "Psampler";
1546 if(This
->cur_args
->np2_fixup
& (1 << sampler
)) {
1548 FIXME("Biased sampling from NP2 textures is unsupported\n");
1554 sampler_base
= "Vsampler";
1555 fixup
= COLOR_FIXUP_IDENTITY
; /* FIXME: Vshader color fixup */
1558 shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
1560 shader_addline(ins
->ctx
->buffer
, "%s(%s%u, ", sample_function
->name
, sampler_base
, sampler
);
1562 va_start(args
, coord_reg_fmt
);
1563 shader_vaddline(ins
->ctx
->buffer
, coord_reg_fmt
, args
);
1567 shader_addline(ins
->ctx
->buffer
, ", %s)%s);\n", bias
, dst_swizzle
);
1570 shader_addline(ins
->ctx
->buffer
, " * PsamplerNP2Fixup%u)%s);\n", sampler
, dst_swizzle
);
1572 shader_addline(ins
->ctx
->buffer
, ")%s);\n", dst_swizzle
);
1576 if(!is_identity_fixup(fixup
)) {
1577 shader_glsl_color_correction(ins
, fixup
);
1581 /*****************************************************************************
1583 * Begin processing individual instruction opcodes
1585 ****************************************************************************/
1587 /* Generate GLSL arithmetic functions (dst = src1 + src2) */
1588 static void shader_glsl_arith(const struct wined3d_shader_instruction
*ins
)
1590 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
1591 glsl_src_param_t src0_param
;
1592 glsl_src_param_t src1_param
;
1596 /* Determine the GLSL operator to use based on the opcode */
1597 switch (ins
->handler_idx
)
1599 case WINED3DSIH_MUL
: op
= '*'; break;
1600 case WINED3DSIH_ADD
: op
= '+'; break;
1601 case WINED3DSIH_SUB
: op
= '-'; break;
1604 FIXME("Opcode %#x not yet handled in GLSL\n", ins
->handler_idx
);
1608 write_mask
= shader_glsl_append_dst(buffer
, ins
);
1609 shader_glsl_add_src_param(ins
, &ins
->src
[0], write_mask
, &src0_param
);
1610 shader_glsl_add_src_param(ins
, &ins
->src
[1], write_mask
, &src1_param
);
1611 shader_addline(buffer
, "%s %c %s);\n", src0_param
.param_str
, op
, src1_param
.param_str
);
1614 /* Process the WINED3DSIO_MOV opcode using GLSL (dst = src) */
1615 static void shader_glsl_mov(const struct wined3d_shader_instruction
*ins
)
1617 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
1618 glsl_src_param_t src0_param
;
1621 write_mask
= shader_glsl_append_dst(buffer
, ins
);
1622 shader_glsl_add_src_param(ins
, &ins
->src
[0], write_mask
, &src0_param
);
1624 /* In vs_1_1 WINED3DSIO_MOV can write to the address register. In later
1625 * shader versions WINED3DSIO_MOVA is used for this. */
1626 if ((WINED3DSHADER_VERSION_MAJOR(ins
->ctx
->reg_maps
->shader_version
) == 1
1627 && !shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
)
1628 && ins
->dst
[0].register_type
== WINED3DSPR_ADDR
))
1630 /* This is a simple floor() */
1631 unsigned int mask_size
= shader_glsl_get_write_mask_size(write_mask
);
1632 if (mask_size
> 1) {
1633 shader_addline(buffer
, "ivec%d(floor(%s)));\n", mask_size
, src0_param
.param_str
);
1635 shader_addline(buffer
, "int(floor(%s)));\n", src0_param
.param_str
);
1638 else if(ins
->handler_idx
== WINED3DSIH_MOVA
)
1640 /* We need to *round* to the nearest int here. */
1641 unsigned int mask_size
= shader_glsl_get_write_mask_size(write_mask
);
1642 if (mask_size
> 1) {
1643 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
);
1645 shader_addline(buffer
, "int(floor(abs(%s) + 0.5) * sign(%s)));\n", src0_param
.param_str
, src0_param
.param_str
);
1648 shader_addline(buffer
, "%s);\n", src0_param
.param_str
);
1652 /* Process the dot product operators DP3 and DP4 in GLSL (dst = dot(src0, src1)) */
1653 static void shader_glsl_dot(const struct wined3d_shader_instruction
*ins
)
1655 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
1656 glsl_src_param_t src0_param
;
1657 glsl_src_param_t src1_param
;
1658 DWORD dst_write_mask
, src_write_mask
;
1659 unsigned int dst_size
= 0;
1661 dst_write_mask
= shader_glsl_append_dst(buffer
, ins
);
1662 dst_size
= shader_glsl_get_write_mask_size(dst_write_mask
);
1664 /* dp3 works on vec3, dp4 on vec4 */
1665 if (ins
->handler_idx
== WINED3DSIH_DP4
)
1667 src_write_mask
= WINED3DSP_WRITEMASK_ALL
;
1669 src_write_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
1672 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_write_mask
, &src0_param
);
1673 shader_glsl_add_src_param(ins
, &ins
->src
[1], src_write_mask
, &src1_param
);
1676 shader_addline(buffer
, "vec%d(dot(%s, %s)));\n", dst_size
, src0_param
.param_str
, src1_param
.param_str
);
1678 shader_addline(buffer
, "dot(%s, %s));\n", src0_param
.param_str
, src1_param
.param_str
);
1682 /* Note that this instruction has some restrictions. The destination write mask
1683 * can't contain the w component, and the source swizzles have to be .xyzw */
1684 static void shader_glsl_cross(const struct wined3d_shader_instruction
*ins
)
1686 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
1687 glsl_src_param_t src0_param
;
1688 glsl_src_param_t src1_param
;
1691 shader_glsl_get_write_mask(&ins
->dst
[0], dst_mask
);
1692 shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
1693 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
1694 shader_glsl_add_src_param(ins
, &ins
->src
[1], src_mask
, &src1_param
);
1695 shader_addline(ins
->ctx
->buffer
, "cross(%s, %s)%s);\n", src0_param
.param_str
, src1_param
.param_str
, dst_mask
);
1698 /* Process the WINED3DSIO_POW instruction in GLSL (dst = |src0|^src1)
1699 * Src0 and src1 are scalars. Note that D3D uses the absolute of src0, while
1700 * GLSL uses the value as-is. */
1701 static void shader_glsl_pow(const struct wined3d_shader_instruction
*ins
)
1703 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
1704 glsl_src_param_t src0_param
;
1705 glsl_src_param_t src1_param
;
1706 DWORD dst_write_mask
;
1707 unsigned int dst_size
;
1709 dst_write_mask
= shader_glsl_append_dst(buffer
, ins
);
1710 dst_size
= shader_glsl_get_write_mask_size(dst_write_mask
);
1712 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
1713 shader_glsl_add_src_param(ins
, &ins
->src
[1], WINED3DSP_WRITEMASK_0
, &src1_param
);
1716 shader_addline(buffer
, "vec%d(pow(abs(%s), %s)));\n", dst_size
, src0_param
.param_str
, src1_param
.param_str
);
1718 shader_addline(buffer
, "pow(abs(%s), %s));\n", src0_param
.param_str
, src1_param
.param_str
);
1722 /* Process the WINED3DSIO_LOG instruction in GLSL (dst = log2(|src0|))
1723 * Src0 is a scalar. Note that D3D uses the absolute of src0, while
1724 * GLSL uses the value as-is. */
1725 static void shader_glsl_log(const struct wined3d_shader_instruction
*ins
)
1727 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
1728 glsl_src_param_t src0_param
;
1729 DWORD dst_write_mask
;
1730 unsigned int dst_size
;
1732 dst_write_mask
= shader_glsl_append_dst(buffer
, ins
);
1733 dst_size
= shader_glsl_get_write_mask_size(dst_write_mask
);
1735 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
1738 shader_addline(buffer
, "vec%d(log2(abs(%s))));\n", dst_size
, src0_param
.param_str
);
1740 shader_addline(buffer
, "log2(abs(%s)));\n", src0_param
.param_str
);
1744 /* Map the opcode 1-to-1 to the GL code (arg->dst = instruction(src0, src1, ...) */
1745 static void shader_glsl_map2gl(const struct wined3d_shader_instruction
*ins
)
1747 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
1748 glsl_src_param_t src_param
;
1749 const char *instruction
;
1753 /* Determine the GLSL function to use based on the opcode */
1754 /* TODO: Possibly make this a table for faster lookups */
1755 switch (ins
->handler_idx
)
1757 case WINED3DSIH_MIN
: instruction
= "min"; break;
1758 case WINED3DSIH_MAX
: instruction
= "max"; break;
1759 case WINED3DSIH_ABS
: instruction
= "abs"; break;
1760 case WINED3DSIH_FRC
: instruction
= "fract"; break;
1761 case WINED3DSIH_NRM
: instruction
= "normalize"; break;
1762 case WINED3DSIH_EXP
: instruction
= "exp2"; break;
1763 case WINED3DSIH_SGN
: instruction
= "sign"; break;
1764 case WINED3DSIH_DSX
: instruction
= "dFdx"; break;
1765 case WINED3DSIH_DSY
: instruction
= "ycorrection.y * dFdy"; break;
1766 default: instruction
= "";
1767 FIXME("Opcode %#x not yet handled in GLSL\n", ins
->handler_idx
);
1771 write_mask
= shader_glsl_append_dst(buffer
, ins
);
1773 shader_addline(buffer
, "%s(", instruction
);
1777 shader_glsl_add_src_param(ins
, &ins
->src
[0], write_mask
, &src_param
);
1778 shader_addline(buffer
, "%s", src_param
.param_str
);
1779 for (i
= 1; i
< ins
->src_count
; ++i
)
1781 shader_glsl_add_src_param(ins
, &ins
->src
[i
], write_mask
, &src_param
);
1782 shader_addline(buffer
, ", %s", src_param
.param_str
);
1786 shader_addline(buffer
, "));\n");
1789 /** Process the WINED3DSIO_EXPP instruction in GLSL:
1790 * For shader model 1.x, do the following (and honor the writemask, so use a temporary variable):
1791 * dst.x = 2^(floor(src))
1792 * dst.y = src - floor(src)
1793 * dst.z = 2^src (partial precision is allowed, but optional)
1795 * For 2.0 shaders, just do this (honoring writemask and swizzle):
1796 * dst = 2^src; (partial precision is allowed, but optional)
1798 static void shader_glsl_expp(const struct wined3d_shader_instruction
*ins
)
1800 glsl_src_param_t src_param
;
1802 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
, &src_param
);
1804 if (ins
->ctx
->reg_maps
->shader_version
< WINED3DPS_VERSION(2,0))
1808 shader_addline(ins
->ctx
->buffer
, "tmp0.x = exp2(floor(%s));\n", src_param
.param_str
);
1809 shader_addline(ins
->ctx
->buffer
, "tmp0.y = %s - floor(%s);\n", src_param
.param_str
, src_param
.param_str
);
1810 shader_addline(ins
->ctx
->buffer
, "tmp0.z = exp2(%s);\n", src_param
.param_str
);
1811 shader_addline(ins
->ctx
->buffer
, "tmp0.w = 1.0;\n");
1813 shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
1814 shader_glsl_get_write_mask(&ins
->dst
[0], dst_mask
);
1815 shader_addline(ins
->ctx
->buffer
, "tmp0%s);\n", dst_mask
);
1818 unsigned int mask_size
;
1820 write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
1821 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
1823 if (mask_size
> 1) {
1824 shader_addline(ins
->ctx
->buffer
, "vec%d(exp2(%s)));\n", mask_size
, src_param
.param_str
);
1826 shader_addline(ins
->ctx
->buffer
, "exp2(%s));\n", src_param
.param_str
);
1831 /** Process the RCP (reciprocal or inverse) opcode in GLSL (dst = 1 / src) */
1832 static void shader_glsl_rcp(const struct wined3d_shader_instruction
*ins
)
1834 glsl_src_param_t src_param
;
1836 unsigned int mask_size
;
1838 write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
1839 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
1840 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_3
, &src_param
);
1842 if (mask_size
> 1) {
1843 shader_addline(ins
->ctx
->buffer
, "vec%d(1.0 / %s));\n", mask_size
, src_param
.param_str
);
1845 shader_addline(ins
->ctx
->buffer
, "1.0 / %s);\n", src_param
.param_str
);
1849 static void shader_glsl_rsq(const struct wined3d_shader_instruction
*ins
)
1851 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
1852 glsl_src_param_t src_param
;
1854 unsigned int mask_size
;
1856 write_mask
= shader_glsl_append_dst(buffer
, ins
);
1857 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
1859 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_3
, &src_param
);
1861 if (mask_size
> 1) {
1862 shader_addline(buffer
, "vec%d(inversesqrt(%s)));\n", mask_size
, src_param
.param_str
);
1864 shader_addline(buffer
, "inversesqrt(%s));\n", src_param
.param_str
);
1868 /** Process signed comparison opcodes in GLSL. */
1869 static void shader_glsl_compare(const struct wined3d_shader_instruction
*ins
)
1871 glsl_src_param_t src0_param
;
1872 glsl_src_param_t src1_param
;
1874 unsigned int mask_size
;
1876 write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
1877 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
1878 shader_glsl_add_src_param(ins
, &ins
->src
[0], write_mask
, &src0_param
);
1879 shader_glsl_add_src_param(ins
, &ins
->src
[1], write_mask
, &src1_param
);
1881 if (mask_size
> 1) {
1882 const char *compare
;
1884 switch(ins
->handler_idx
)
1886 case WINED3DSIH_SLT
: compare
= "lessThan"; break;
1887 case WINED3DSIH_SGE
: compare
= "greaterThanEqual"; break;
1888 default: compare
= "";
1889 FIXME("Can't handle opcode %#x\n", ins
->handler_idx
);
1892 shader_addline(ins
->ctx
->buffer
, "vec%d(%s(%s, %s)));\n", mask_size
, compare
,
1893 src0_param
.param_str
, src1_param
.param_str
);
1895 switch(ins
->handler_idx
)
1897 case WINED3DSIH_SLT
:
1898 /* Step(src0, src1) is not suitable here because if src0 == src1 SLT is supposed,
1899 * to return 0.0 but step returns 1.0 because step is not < x
1900 * An alternative is a bvec compare padded with an unused second component.
1901 * step(src1 * -1.0, src0 * -1.0) is not an option because it suffers from the same
1902 * issue. Playing with not() is not possible either because not() does not accept
1905 shader_addline(ins
->ctx
->buffer
, "(%s < %s) ? 1.0 : 0.0);\n",
1906 src0_param
.param_str
, src1_param
.param_str
);
1908 case WINED3DSIH_SGE
:
1909 /* Here we can use the step() function and safe a conditional */
1910 shader_addline(ins
->ctx
->buffer
, "step(%s, %s));\n", src1_param
.param_str
, src0_param
.param_str
);
1913 FIXME("Can't handle opcode %#x\n", ins
->handler_idx
);
1919 /** Process CMP instruction in GLSL (dst = src0 >= 0.0 ? src1 : src2), per channel */
1920 static void shader_glsl_cmp(const struct wined3d_shader_instruction
*ins
)
1922 glsl_src_param_t src0_param
;
1923 glsl_src_param_t src1_param
;
1924 glsl_src_param_t src2_param
;
1925 DWORD write_mask
, cmp_channel
= 0;
1928 BOOL temp_destination
= FALSE
;
1930 if (shader_is_scalar(ins
->src
[0].register_type
, ins
->src
[0].register_idx
))
1932 write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
1934 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_ALL
, &src0_param
);
1935 shader_glsl_add_src_param(ins
, &ins
->src
[1], write_mask
, &src1_param
);
1936 shader_glsl_add_src_param(ins
, &ins
->src
[2], write_mask
, &src2_param
);
1938 shader_addline(ins
->ctx
->buffer
, "%s >= 0.0 ? %s : %s);\n",
1939 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
1941 DWORD dst_mask
= ins
->dst
[0].write_mask
;
1942 struct wined3d_shader_dst_param dst
= ins
->dst
[0];
1944 /* Cycle through all source0 channels */
1945 for (i
=0; i
<4; i
++) {
1947 /* Find the destination channels which use the current source0 channel */
1948 for (j
=0; j
<4; j
++) {
1949 if (((ins
->src
[0].swizzle
>> (WINED3DSP_SWIZZLE_SHIFT
+ 2 * j
)) & 0x3) == i
)
1951 write_mask
|= WINED3DSP_WRITEMASK_0
<< j
;
1952 cmp_channel
= WINED3DSP_WRITEMASK_0
<< j
;
1955 dst
.write_mask
= dst_mask
& write_mask
;
1957 /* Splitting the cmp instruction up in multiple lines imposes a problem:
1958 * The first lines may overwrite source parameters of the following lines.
1959 * Deal with that by using a temporary destination register if needed
1961 if ((ins
->src
[0].register_idx
== ins
->dst
[0].register_idx
1962 && ins
->src
[0].register_type
== ins
->dst
[0].register_type
)
1963 || (ins
->src
[1].register_idx
== ins
->dst
[0].register_idx
1964 && ins
->src
[1].register_type
== ins
->dst
[0].register_type
)
1965 || (ins
->src
[2].register_idx
== ins
->dst
[0].register_idx
1966 && ins
->src
[2].register_type
== ins
->dst
[0].register_type
))
1968 write_mask
= shader_glsl_get_write_mask(&dst
, mask_char
);
1969 if (!write_mask
) continue;
1970 shader_addline(ins
->ctx
->buffer
, "tmp0%s = (", mask_char
);
1971 temp_destination
= TRUE
;
1973 write_mask
= shader_glsl_append_dst_ext(ins
->ctx
->buffer
, ins
, &dst
);
1974 if (!write_mask
) continue;
1977 shader_glsl_add_src_param(ins
, &ins
->src
[0], cmp_channel
, &src0_param
);
1978 shader_glsl_add_src_param(ins
, &ins
->src
[1], write_mask
, &src1_param
);
1979 shader_glsl_add_src_param(ins
, &ins
->src
[2], write_mask
, &src2_param
);
1981 shader_addline(ins
->ctx
->buffer
, "%s >= 0.0 ? %s : %s);\n",
1982 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
1985 if(temp_destination
) {
1986 shader_glsl_get_write_mask(&ins
->dst
[0], mask_char
);
1987 shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
1988 shader_addline(ins
->ctx
->buffer
, "tmp0%s);\n", mask_char
);
1994 /** Process the CND opcode in GLSL (dst = (src0 > 0.5) ? src1 : src2) */
1995 /* For ps 1.1-1.3, only a single component of src0 is used. For ps 1.4
1996 * the compare is done per component of src0. */
1997 static void shader_glsl_cnd(const struct wined3d_shader_instruction
*ins
)
1999 struct wined3d_shader_dst_param dst
;
2000 glsl_src_param_t src0_param
;
2001 glsl_src_param_t src1_param
;
2002 glsl_src_param_t src2_param
;
2003 DWORD write_mask
, cmp_channel
= 0;
2007 if (ins
->ctx
->reg_maps
->shader_version
< WINED3DPS_VERSION(1, 4))
2009 write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2010 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2011 shader_glsl_add_src_param(ins
, &ins
->src
[1], write_mask
, &src1_param
);
2012 shader_glsl_add_src_param(ins
, &ins
->src
[2], write_mask
, &src2_param
);
2014 /* Fun: The D3DSI_COISSUE flag changes the semantic of the cnd instruction for < 1.4 shaders */
2017 shader_addline(ins
->ctx
->buffer
, "%s /* COISSUE! */);\n", src1_param
.param_str
);
2019 shader_addline(ins
->ctx
->buffer
, "%s > 0.5 ? %s : %s);\n",
2020 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
2024 /* Cycle through all source0 channels */
2025 dst_mask
= ins
->dst
[0].write_mask
;
2027 for (i
=0; i
<4; i
++) {
2029 /* Find the destination channels which use the current source0 channel */
2030 for (j
=0; j
<4; j
++) {
2031 if (((ins
->src
[0].swizzle
>> (WINED3DSP_SWIZZLE_SHIFT
+ 2 * j
)) & 0x3) == i
)
2033 write_mask
|= WINED3DSP_WRITEMASK_0
<< j
;
2034 cmp_channel
= WINED3DSP_WRITEMASK_0
<< j
;
2038 dst
.write_mask
= dst_mask
& write_mask
;
2039 write_mask
= shader_glsl_append_dst_ext(ins
->ctx
->buffer
, ins
, &dst
);
2040 if (!write_mask
) continue;
2042 shader_glsl_add_src_param(ins
, &ins
->src
[0], cmp_channel
, &src0_param
);
2043 shader_glsl_add_src_param(ins
, &ins
->src
[1], write_mask
, &src1_param
);
2044 shader_glsl_add_src_param(ins
, &ins
->src
[2], write_mask
, &src2_param
);
2046 shader_addline(ins
->ctx
->buffer
, "%s > 0.5 ? %s : %s);\n",
2047 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
2051 /** GLSL code generation for WINED3DSIO_MAD: Multiply the first 2 opcodes, then add the last */
2052 static void shader_glsl_mad(const struct wined3d_shader_instruction
*ins
)
2054 glsl_src_param_t src0_param
;
2055 glsl_src_param_t src1_param
;
2056 glsl_src_param_t src2_param
;
2059 write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2060 shader_glsl_add_src_param(ins
, &ins
->src
[0], write_mask
, &src0_param
);
2061 shader_glsl_add_src_param(ins
, &ins
->src
[1], write_mask
, &src1_param
);
2062 shader_glsl_add_src_param(ins
, &ins
->src
[2], write_mask
, &src2_param
);
2063 shader_addline(ins
->ctx
->buffer
, "(%s * %s) + %s);\n",
2064 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
2067 /** Handles transforming all WINED3DSIO_M?x? opcodes for
2068 Vertex shaders to GLSL codes */
2069 static void shader_glsl_mnxn(const struct wined3d_shader_instruction
*ins
)
2072 int nComponents
= 0;
2073 struct wined3d_shader_dst_param tmp_dst
= {0};
2074 struct wined3d_shader_src_param tmp_src
[2] = {{0}};
2075 struct wined3d_shader_instruction tmp_ins
;
2077 memset(&tmp_ins
, 0, sizeof(tmp_ins
));
2079 /* Set constants for the temporary argument */
2080 tmp_ins
.ctx
= ins
->ctx
;
2081 tmp_ins
.dst_count
= 1;
2082 tmp_ins
.dst
= &tmp_dst
;
2083 tmp_ins
.src_count
= 2;
2084 tmp_ins
.src
= tmp_src
;
2086 switch(ins
->handler_idx
)
2088 case WINED3DSIH_M4x4
:
2090 tmp_ins
.handler_idx
= WINED3DSIH_DP4
;
2092 case WINED3DSIH_M4x3
:
2094 tmp_ins
.handler_idx
= WINED3DSIH_DP4
;
2096 case WINED3DSIH_M3x4
:
2098 tmp_ins
.handler_idx
= WINED3DSIH_DP3
;
2100 case WINED3DSIH_M3x3
:
2102 tmp_ins
.handler_idx
= WINED3DSIH_DP3
;
2104 case WINED3DSIH_M3x2
:
2106 tmp_ins
.handler_idx
= WINED3DSIH_DP3
;
2112 tmp_dst
= ins
->dst
[0];
2113 tmp_src
[0] = ins
->src
[0];
2114 tmp_src
[1] = ins
->src
[1];
2115 for (i
= 0; i
< nComponents
; ++i
)
2117 tmp_dst
.write_mask
= WINED3DSP_WRITEMASK_0
<< i
;
2118 shader_glsl_dot(&tmp_ins
);
2119 ++tmp_src
[1].register_idx
;
2125 The LRP instruction performs a component-wise linear interpolation
2126 between the second and third operands using the first operand as the
2127 blend factor. Equation: (dst = src2 + src0 * (src1 - src2))
2128 This is equivalent to mix(src2, src1, src0);
2130 static void shader_glsl_lrp(const struct wined3d_shader_instruction
*ins
)
2132 glsl_src_param_t src0_param
;
2133 glsl_src_param_t src1_param
;
2134 glsl_src_param_t src2_param
;
2137 write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2139 shader_glsl_add_src_param(ins
, &ins
->src
[0], write_mask
, &src0_param
);
2140 shader_glsl_add_src_param(ins
, &ins
->src
[1], write_mask
, &src1_param
);
2141 shader_glsl_add_src_param(ins
, &ins
->src
[2], write_mask
, &src2_param
);
2143 shader_addline(ins
->ctx
->buffer
, "mix(%s, %s, %s));\n",
2144 src2_param
.param_str
, src1_param
.param_str
, src0_param
.param_str
);
2147 /** Process the WINED3DSIO_LIT instruction in GLSL:
2148 * dst.x = dst.w = 1.0
2149 * dst.y = (src0.x > 0) ? src0.x
2150 * dst.z = (src0.x > 0) ? ((src0.y > 0) ? pow(src0.y, src.w) : 0) : 0
2151 * where src.w is clamped at +- 128
2153 static void shader_glsl_lit(const struct wined3d_shader_instruction
*ins
)
2155 glsl_src_param_t src0_param
;
2156 glsl_src_param_t src1_param
;
2157 glsl_src_param_t src3_param
;
2160 shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2161 shader_glsl_get_write_mask(&ins
->dst
[0], dst_mask
);
2163 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2164 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_1
, &src1_param
);
2165 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_3
, &src3_param
);
2167 /* The sdk specifies the instruction like this
2169 * if(src.x > 0.0) dst.y = src.x
2171 * if(src.x > 0.0 && src.y > 0.0) dst.z = pow(src.y, power);
2175 * Obviously that has quite a few conditionals in it which we don't like. So the first step is this:
2176 * dst.x = 1.0 ... No further explanation needed
2177 * dst.y = max(src.y, 0.0); ... If x < 0.0, use 0.0, otherwise x. Same as the conditional
2178 * dst.z = x > 0.0 ? pow(max(y, 0.0), p) : 0; ... 0 ^ power is 0, and otherwise we use y anyway
2179 * dst.w = 1.0. ... Nothing fancy.
2181 * So we still have one conditional in there. So do this:
2182 * dst.z = pow(max(0.0, src.y) * step(0.0, src.x), power);
2184 * 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),
2185 * 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.
2186 * if both x and y are > 0, we get pow(y * 1.0, power), as it is supposed to
2188 shader_addline(ins
->ctx
->buffer
,
2189 "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",
2190 src0_param
.param_str
, src1_param
.param_str
, src0_param
.param_str
, src3_param
.param_str
, dst_mask
);
2193 /** Process the WINED3DSIO_DST instruction in GLSL:
2195 * dst.y = src0.x * src0.y
2199 static void shader_glsl_dst(const struct wined3d_shader_instruction
*ins
)
2201 glsl_src_param_t src0y_param
;
2202 glsl_src_param_t src0z_param
;
2203 glsl_src_param_t src1y_param
;
2204 glsl_src_param_t src1w_param
;
2207 shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2208 shader_glsl_get_write_mask(&ins
->dst
[0], dst_mask
);
2210 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_1
, &src0y_param
);
2211 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_2
, &src0z_param
);
2212 shader_glsl_add_src_param(ins
, &ins
->src
[1], WINED3DSP_WRITEMASK_1
, &src1y_param
);
2213 shader_glsl_add_src_param(ins
, &ins
->src
[1], WINED3DSP_WRITEMASK_3
, &src1w_param
);
2215 shader_addline(ins
->ctx
->buffer
, "vec4(1.0, %s * %s, %s, %s))%s;\n",
2216 src0y_param
.param_str
, src1y_param
.param_str
, src0z_param
.param_str
, src1w_param
.param_str
, dst_mask
);
2219 /** Process the WINED3DSIO_SINCOS instruction in GLSL:
2220 * VS 2.0 requires that specific cosine and sine constants be passed to this instruction so the hardware
2221 * can handle it. But, these functions are built-in for GLSL, so we can just ignore the last 2 params.
2223 * dst.x = cos(src0.?)
2224 * dst.y = sin(src0.?)
2228 static void shader_glsl_sincos(const struct wined3d_shader_instruction
*ins
)
2230 glsl_src_param_t src0_param
;
2233 write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2234 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2236 switch (write_mask
) {
2237 case WINED3DSP_WRITEMASK_0
:
2238 shader_addline(ins
->ctx
->buffer
, "cos(%s));\n", src0_param
.param_str
);
2241 case WINED3DSP_WRITEMASK_1
:
2242 shader_addline(ins
->ctx
->buffer
, "sin(%s));\n", src0_param
.param_str
);
2245 case (WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
):
2246 shader_addline(ins
->ctx
->buffer
, "vec2(cos(%s), sin(%s)));\n", src0_param
.param_str
, src0_param
.param_str
);
2250 ERR("Write mask should be .x, .y or .xy\n");
2255 /** Process the WINED3DSIO_LOOP instruction in GLSL:
2256 * Start a for() loop where src1.y is the initial value of aL,
2257 * increment aL by src1.z for a total of src1.x iterations.
2258 * Need to use a temporary variable for this operation.
2260 /* FIXME: I don't think nested loops will work correctly this way. */
2261 static void shader_glsl_loop(const struct wined3d_shader_instruction
*ins
)
2263 glsl_src_param_t src1_param
;
2264 IWineD3DBaseShaderImpl
*shader
= (IWineD3DBaseShaderImpl
*)ins
->ctx
->shader
;
2265 const DWORD
*control_values
= NULL
;
2266 const local_constant
*constant
;
2268 shader_glsl_add_src_param(ins
, &ins
->src
[1], WINED3DSP_WRITEMASK_ALL
, &src1_param
);
2270 /* Try to hardcode the loop control parameters if possible. Direct3D 9 class hardware doesn't support real
2271 * varying indexing, but Microsoft designed this feature for Shader model 2.x+. If the loop control is
2272 * known at compile time, the GLSL compiler can unroll the loop, and replace indirect addressing with direct
2275 if (ins
->src
[1].register_type
== WINED3DSPR_CONSTINT
)
2277 LIST_FOR_EACH_ENTRY(constant
, &shader
->baseShader
.constantsI
, local_constant
, entry
) {
2278 if (constant
->idx
== ins
->src
[1].register_idx
)
2280 control_values
= constant
->value
;
2286 if(control_values
) {
2287 if(control_values
[2] > 0) {
2288 shader_addline(ins
->ctx
->buffer
, "for (aL%u = %d; aL%u < (%d * %d + %d); aL%u += %d) {\n",
2289 shader
->baseShader
.cur_loop_depth
, control_values
[1],
2290 shader
->baseShader
.cur_loop_depth
, control_values
[0], control_values
[2], control_values
[1],
2291 shader
->baseShader
.cur_loop_depth
, control_values
[2]);
2292 } else if(control_values
[2] == 0) {
2293 shader_addline(ins
->ctx
->buffer
, "for (aL%u = %d, tmpInt%u = 0; tmpInt%u < %d; tmpInt%u++) {\n",
2294 shader
->baseShader
.cur_loop_depth
, control_values
[1], shader
->baseShader
.cur_loop_depth
,
2295 shader
->baseShader
.cur_loop_depth
, control_values
[0],
2296 shader
->baseShader
.cur_loop_depth
);
2298 shader_addline(ins
->ctx
->buffer
, "for (aL%u = %d; aL%u > (%d * %d + %d); aL%u += %d) {\n",
2299 shader
->baseShader
.cur_loop_depth
, control_values
[1],
2300 shader
->baseShader
.cur_loop_depth
, control_values
[0], control_values
[2], control_values
[1],
2301 shader
->baseShader
.cur_loop_depth
, control_values
[2]);
2304 shader_addline(ins
->ctx
->buffer
,
2305 "for (tmpInt%u = 0, aL%u = %s.y; tmpInt%u < %s.x; tmpInt%u++, aL%u += %s.z) {\n",
2306 shader
->baseShader
.cur_loop_depth
, shader
->baseShader
.cur_loop_regno
,
2307 src1_param
.reg_name
, shader
->baseShader
.cur_loop_depth
, src1_param
.reg_name
,
2308 shader
->baseShader
.cur_loop_depth
, shader
->baseShader
.cur_loop_regno
, src1_param
.reg_name
);
2311 shader
->baseShader
.cur_loop_depth
++;
2312 shader
->baseShader
.cur_loop_regno
++;
2315 static void shader_glsl_end(const struct wined3d_shader_instruction
*ins
)
2317 IWineD3DBaseShaderImpl
*shader
= (IWineD3DBaseShaderImpl
*)ins
->ctx
->shader
;
2319 shader_addline(ins
->ctx
->buffer
, "}\n");
2321 if (ins
->handler_idx
== WINED3DSIH_ENDLOOP
)
2323 shader
->baseShader
.cur_loop_depth
--;
2324 shader
->baseShader
.cur_loop_regno
--;
2327 if (ins
->handler_idx
== WINED3DSIH_ENDREP
)
2329 shader
->baseShader
.cur_loop_depth
--;
2333 static void shader_glsl_rep(const struct wined3d_shader_instruction
*ins
)
2335 IWineD3DBaseShaderImpl
*shader
= (IWineD3DBaseShaderImpl
*)ins
->ctx
->shader
;
2336 glsl_src_param_t src0_param
;
2338 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2339 shader_addline(ins
->ctx
->buffer
, "for (tmpInt%d = 0; tmpInt%d < %s; tmpInt%d++) {\n",
2340 shader
->baseShader
.cur_loop_depth
, shader
->baseShader
.cur_loop_depth
,
2341 src0_param
.param_str
, shader
->baseShader
.cur_loop_depth
);
2342 shader
->baseShader
.cur_loop_depth
++;
2345 static void shader_glsl_if(const struct wined3d_shader_instruction
*ins
)
2347 glsl_src_param_t src0_param
;
2349 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2350 shader_addline(ins
->ctx
->buffer
, "if (%s) {\n", src0_param
.param_str
);
2353 static void shader_glsl_ifc(const struct wined3d_shader_instruction
*ins
)
2355 glsl_src_param_t src0_param
;
2356 glsl_src_param_t src1_param
;
2358 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2359 shader_glsl_add_src_param(ins
, &ins
->src
[1], WINED3DSP_WRITEMASK_0
, &src1_param
);
2361 shader_addline(ins
->ctx
->buffer
, "if (%s %s %s) {\n",
2362 src0_param
.param_str
, shader_get_comp_op(ins
->flags
), src1_param
.param_str
);
2365 static void shader_glsl_else(const struct wined3d_shader_instruction
*ins
)
2367 shader_addline(ins
->ctx
->buffer
, "} else {\n");
2370 static void shader_glsl_break(const struct wined3d_shader_instruction
*ins
)
2372 shader_addline(ins
->ctx
->buffer
, "break;\n");
2375 /* FIXME: According to MSDN the compare is done per component. */
2376 static void shader_glsl_breakc(const struct wined3d_shader_instruction
*ins
)
2378 glsl_src_param_t src0_param
;
2379 glsl_src_param_t src1_param
;
2381 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2382 shader_glsl_add_src_param(ins
, &ins
->src
[1], WINED3DSP_WRITEMASK_0
, &src1_param
);
2384 shader_addline(ins
->ctx
->buffer
, "if (%s %s %s) break;\n",
2385 src0_param
.param_str
, shader_get_comp_op(ins
->flags
), src1_param
.param_str
);
2388 static void shader_glsl_label(const struct wined3d_shader_instruction
*ins
)
2390 shader_addline(ins
->ctx
->buffer
, "}\n");
2391 shader_addline(ins
->ctx
->buffer
, "void subroutine%u () {\n", ins
->src
[0].register_idx
);
2394 static void shader_glsl_call(const struct wined3d_shader_instruction
*ins
)
2396 shader_addline(ins
->ctx
->buffer
, "subroutine%u();\n", ins
->src
[0].register_idx
);
2399 static void shader_glsl_callnz(const struct wined3d_shader_instruction
*ins
)
2401 glsl_src_param_t src1_param
;
2403 shader_glsl_add_src_param(ins
, &ins
->src
[1], WINED3DSP_WRITEMASK_0
, &src1_param
);
2404 shader_addline(ins
->ctx
->buffer
, "if (%s) subroutine%u();\n", src1_param
.param_str
, ins
->src
[0].register_idx
);
2407 /*********************************************
2408 * Pixel Shader Specific Code begins here
2409 ********************************************/
2410 static void pshader_glsl_tex(const struct wined3d_shader_instruction
*ins
)
2412 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*)ins
->ctx
->shader
;
2413 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
2414 DWORD shader_version
= ins
->ctx
->reg_maps
->shader_version
;
2415 glsl_sample_function_t sample_function
;
2416 DWORD sample_flags
= 0;
2419 DWORD mask
= 0, swizzle
;
2421 /* 1.0-1.4: Use destination register as sampler source.
2422 * 2.0+: Use provided sampler source. */
2423 if (shader_version
< WINED3DPS_VERSION(2,0)) sampler_idx
= ins
->dst
[0].register_idx
;
2424 else sampler_idx
= ins
->src
[1].register_idx
;
2425 sampler_type
= ins
->ctx
->reg_maps
->samplers
[sampler_idx
] & WINED3DSP_TEXTURETYPE_MASK
;
2427 if (shader_version
< WINED3DPS_VERSION(1,4))
2429 DWORD flags
= deviceImpl
->stateBlock
->textureState
[sampler_idx
][WINED3DTSS_TEXTURETRANSFORMFLAGS
];
2431 /* Projected cube textures don't make a lot of sense, the resulting coordinates stay the same. */
2432 if (flags
& WINED3DTTFF_PROJECTED
&& sampler_type
!= WINED3DSTT_CUBE
) {
2433 sample_flags
|= WINED3D_GLSL_SAMPLE_PROJECTED
;
2434 switch (flags
& ~WINED3DTTFF_PROJECTED
) {
2435 case WINED3DTTFF_COUNT1
: FIXME("WINED3DTTFF_PROJECTED with WINED3DTTFF_COUNT1?\n"); break;
2436 case WINED3DTTFF_COUNT2
: mask
= WINED3DSP_WRITEMASK_1
; break;
2437 case WINED3DTTFF_COUNT3
: mask
= WINED3DSP_WRITEMASK_2
; break;
2438 case WINED3DTTFF_COUNT4
:
2439 case WINED3DTTFF_DISABLE
: mask
= WINED3DSP_WRITEMASK_3
; break;
2443 else if (shader_version
< WINED3DPS_VERSION(2,0))
2445 DWORD src_mod
= ins
->src
[0].modifiers
;
2447 if (src_mod
== WINED3DSPSM_DZ
) {
2448 sample_flags
|= WINED3D_GLSL_SAMPLE_PROJECTED
;
2449 mask
= WINED3DSP_WRITEMASK_2
;
2450 } else if (src_mod
== WINED3DSPSM_DW
) {
2451 sample_flags
|= WINED3D_GLSL_SAMPLE_PROJECTED
;
2452 mask
= WINED3DSP_WRITEMASK_3
;
2455 if (ins
->flags
& WINED3DSI_TEXLD_PROJECT
)
2457 /* ps 2.0 texldp instruction always divides by the fourth component. */
2458 sample_flags
|= WINED3D_GLSL_SAMPLE_PROJECTED
;
2459 mask
= WINED3DSP_WRITEMASK_3
;
2463 if(deviceImpl
->stateBlock
->textures
[sampler_idx
] &&
2464 IWineD3DBaseTexture_GetTextureDimensions(deviceImpl
->stateBlock
->textures
[sampler_idx
]) == GL_TEXTURE_RECTANGLE_ARB
) {
2465 sample_flags
|= WINED3D_GLSL_SAMPLE_RECT
;
2468 shader_glsl_get_sample_function(sampler_type
, sample_flags
, &sample_function
);
2469 mask
|= sample_function
.coord_mask
;
2471 if (shader_version
< WINED3DPS_VERSION(2,0)) swizzle
= WINED3DVS_NOSWIZZLE
;
2472 else swizzle
= ins
->src
[1].swizzle
;
2474 /* 1.0-1.3: Use destination register as coordinate source.
2475 1.4+: Use provided coordinate source register. */
2476 if (shader_version
< WINED3DPS_VERSION(1,4))
2479 shader_glsl_write_mask_to_str(mask
, coord_mask
);
2480 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, swizzle
, NULL
,
2481 "T%u%s", sampler_idx
, coord_mask
);
2483 glsl_src_param_t coord_param
;
2484 shader_glsl_add_src_param(ins
, &ins
->src
[0], mask
, &coord_param
);
2485 if (ins
->flags
& WINED3DSI_TEXLD_BIAS
)
2487 glsl_src_param_t bias
;
2488 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_3
, &bias
);
2489 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, swizzle
, bias
.param_str
,
2490 "%s", coord_param
.param_str
);
2492 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, swizzle
, NULL
,
2493 "%s", coord_param
.param_str
);
2498 static void shader_glsl_texldl(const struct wined3d_shader_instruction
*ins
)
2500 IWineD3DBaseShaderImpl
*This
= (IWineD3DBaseShaderImpl
*)ins
->ctx
->shader
;
2501 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
2502 glsl_sample_function_t sample_function
;
2503 glsl_src_param_t coord_param
, lod_param
;
2504 DWORD sample_flags
= WINED3D_GLSL_SAMPLE_LOD
;
2507 DWORD swizzle
= ins
->src
[1].swizzle
;
2509 sampler_idx
= ins
->src
[1].register_idx
;
2510 sampler_type
= ins
->ctx
->reg_maps
->samplers
[sampler_idx
] & WINED3DSP_TEXTURETYPE_MASK
;
2511 if(deviceImpl
->stateBlock
->textures
[sampler_idx
] &&
2512 IWineD3DBaseTexture_GetTextureDimensions(deviceImpl
->stateBlock
->textures
[sampler_idx
]) == GL_TEXTURE_RECTANGLE_ARB
) {
2513 sample_flags
|= WINED3D_GLSL_SAMPLE_RECT
;
2515 shader_glsl_get_sample_function(sampler_type
, sample_flags
, &sample_function
);
2516 shader_glsl_add_src_param(ins
, &ins
->src
[0], sample_function
.coord_mask
, &coord_param
);
2518 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_3
, &lod_param
);
2520 if (shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
))
2522 /* The GLSL spec claims the Lod sampling functions are only supported in vertex shaders.
2523 * However, they seem to work just fine in fragment shaders as well. */
2524 WARN("Using %s in fragment shader.\n", sample_function
.name
);
2526 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, swizzle
, lod_param
.param_str
,
2527 "%s", coord_param
.param_str
);
2530 static void pshader_glsl_texcoord(const struct wined3d_shader_instruction
*ins
)
2532 /* FIXME: Make this work for more than just 2D textures */
2533 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
2534 DWORD write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2536 if (ins
->ctx
->reg_maps
->shader_version
!= WINED3DPS_VERSION(1,4))
2540 shader_glsl_get_write_mask(&ins
->dst
[0], dst_mask
);
2541 shader_addline(buffer
, "clamp(gl_TexCoord[%u], 0.0, 1.0)%s);\n",
2542 ins
->dst
[0].register_idx
, dst_mask
);
2544 DWORD reg
= ins
->src
[0].register_idx
;
2545 DWORD src_mod
= ins
->src
[0].modifiers
;
2546 char dst_swizzle
[6];
2548 shader_glsl_get_swizzle(ins
->src
[0].token
, FALSE
, write_mask
, dst_swizzle
);
2550 if (src_mod
== WINED3DSPSM_DZ
) {
2551 glsl_src_param_t div_param
;
2552 unsigned int mask_size
= shader_glsl_get_write_mask_size(write_mask
);
2553 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_2
, &div_param
);
2555 if (mask_size
> 1) {
2556 shader_addline(buffer
, "gl_TexCoord[%u]%s / vec%d(%s));\n", reg
, dst_swizzle
, mask_size
, div_param
.param_str
);
2558 shader_addline(buffer
, "gl_TexCoord[%u]%s / %s);\n", reg
, dst_swizzle
, div_param
.param_str
);
2560 } else if (src_mod
== WINED3DSPSM_DW
) {
2561 glsl_src_param_t div_param
;
2562 unsigned int mask_size
= shader_glsl_get_write_mask_size(write_mask
);
2563 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_3
, &div_param
);
2565 if (mask_size
> 1) {
2566 shader_addline(buffer
, "gl_TexCoord[%u]%s / vec%d(%s));\n", reg
, dst_swizzle
, mask_size
, div_param
.param_str
);
2568 shader_addline(buffer
, "gl_TexCoord[%u]%s / %s);\n", reg
, dst_swizzle
, div_param
.param_str
);
2571 shader_addline(buffer
, "gl_TexCoord[%u]%s);\n", reg
, dst_swizzle
);
2576 /** Process the WINED3DSIO_TEXDP3TEX instruction in GLSL:
2577 * Take a 3-component dot product of the TexCoord[dstreg] and src,
2578 * then perform a 1D texture lookup from stage dstregnum, place into dst. */
2579 static void pshader_glsl_texdp3tex(const struct wined3d_shader_instruction
*ins
)
2581 glsl_src_param_t src0_param
;
2582 glsl_sample_function_t sample_function
;
2583 DWORD sampler_idx
= ins
->dst
[0].register_idx
;
2584 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2585 DWORD sampler_type
= ins
->ctx
->reg_maps
->samplers
[sampler_idx
] & WINED3DSP_TEXTURETYPE_MASK
;
2588 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
2590 /* Do I have to take care about the projected bit? I don't think so, since the dp3 returns only one
2591 * scalar, and projected sampling would require 4.
2593 * It is a dependent read - not valid with conditional NP2 textures
2595 shader_glsl_get_sample_function(sampler_type
, 0, &sample_function
);
2596 mask_size
= shader_glsl_get_write_mask_size(sample_function
.coord_mask
);
2601 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, WINED3DVS_NOSWIZZLE
, NULL
,
2602 "dot(gl_TexCoord[%u].xyz, %s)", sampler_idx
, src0_param
.param_str
);
2606 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, WINED3DVS_NOSWIZZLE
, NULL
,
2607 "vec2(dot(gl_TexCoord[%u].xyz, %s), 0.0)", sampler_idx
, src0_param
.param_str
);
2611 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, WINED3DVS_NOSWIZZLE
, NULL
,
2612 "vec3(dot(gl_TexCoord[%u].xyz, %s), 0.0, 0.0)", sampler_idx
, src0_param
.param_str
);
2616 FIXME("Unexpected mask size %u\n", mask_size
);
2621 /** Process the WINED3DSIO_TEXDP3 instruction in GLSL:
2622 * Take a 3-component dot product of the TexCoord[dstreg] and src. */
2623 static void pshader_glsl_texdp3(const struct wined3d_shader_instruction
*ins
)
2625 glsl_src_param_t src0_param
;
2626 DWORD dstreg
= ins
->dst
[0].register_idx
;
2627 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2629 unsigned int mask_size
;
2631 dst_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2632 mask_size
= shader_glsl_get_write_mask_size(dst_mask
);
2633 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
2635 if (mask_size
> 1) {
2636 shader_addline(ins
->ctx
->buffer
, "vec%d(dot(T%u.xyz, %s)));\n", mask_size
, dstreg
, src0_param
.param_str
);
2638 shader_addline(ins
->ctx
->buffer
, "dot(T%u.xyz, %s));\n", dstreg
, src0_param
.param_str
);
2642 /** Process the WINED3DSIO_TEXDEPTH instruction in GLSL:
2643 * Calculate the depth as dst.x / dst.y */
2644 static void pshader_glsl_texdepth(const struct wined3d_shader_instruction
*ins
)
2646 glsl_dst_param_t dst_param
;
2648 shader_glsl_add_dst_param(ins
, &ins
->dst
[0], &dst_param
);
2650 /* Tests show that texdepth never returns anything below 0.0, and that r5.y is clamped to 1.0.
2651 * Negative input is accepted, -0.25 / -0.5 returns 0.5. GL should clamp gl_FragDepth to [0;1], but
2652 * this doesn't always work, so clamp the results manually. Whether or not the x value is clamped at 1
2653 * too is irrelevant, since if x = 0, any y value < 1.0 (and > 1.0 is not allowed) results in a result
2656 shader_addline(ins
->ctx
->buffer
, "gl_FragDepth = clamp((%s.x / min(%s.y, 1.0)), 0.0, 1.0);\n",
2657 dst_param
.reg_name
, dst_param
.reg_name
);
2660 /** Process the WINED3DSIO_TEXM3X2DEPTH instruction in GLSL:
2661 * Last row of a 3x2 matrix multiply, use the result to calculate the depth:
2662 * Calculate tmp0.y = TexCoord[dstreg] . src.xyz; (tmp0.x has already been calculated)
2663 * depth = (tmp0.y == 0.0) ? 1.0 : tmp0.x / tmp0.y
2665 static void pshader_glsl_texm3x2depth(const struct wined3d_shader_instruction
*ins
)
2667 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2668 DWORD dstreg
= ins
->dst
[0].register_idx
;
2669 glsl_src_param_t src0_param
;
2671 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
2673 shader_addline(ins
->ctx
->buffer
, "tmp0.y = dot(T%u.xyz, %s);\n", dstreg
, src0_param
.param_str
);
2674 shader_addline(ins
->ctx
->buffer
, "gl_FragDepth = (tmp0.y == 0.0) ? 1.0 : clamp(tmp0.x / tmp0.y, 0.0, 1.0);\n");
2677 /** Process the WINED3DSIO_TEXM3X2PAD instruction in GLSL
2678 * Calculate the 1st of a 2-row matrix multiplication. */
2679 static void pshader_glsl_texm3x2pad(const struct wined3d_shader_instruction
*ins
)
2681 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2682 DWORD reg
= ins
->dst
[0].register_idx
;
2683 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
2684 glsl_src_param_t src0_param
;
2686 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
2687 shader_addline(buffer
, "tmp0.x = dot(T%u.xyz, %s);\n", reg
, src0_param
.param_str
);
2690 /** Process the WINED3DSIO_TEXM3X3PAD instruction in GLSL
2691 * Calculate the 1st or 2nd row of a 3-row matrix multiplication. */
2692 static void pshader_glsl_texm3x3pad(const struct wined3d_shader_instruction
*ins
)
2694 IWineD3DPixelShaderImpl
*shader
= (IWineD3DPixelShaderImpl
*)ins
->ctx
->shader
;
2695 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2696 DWORD reg
= ins
->dst
[0].register_idx
;
2697 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
2698 SHADER_PARSE_STATE
* current_state
= &shader
->baseShader
.parse_state
;
2699 glsl_src_param_t src0_param
;
2701 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
2702 shader_addline(buffer
, "tmp0.%c = dot(T%u.xyz, %s);\n", 'x' + current_state
->current_row
, reg
, src0_param
.param_str
);
2703 current_state
->texcoord_w
[current_state
->current_row
++] = reg
;
2706 static void pshader_glsl_texm3x2tex(const struct wined3d_shader_instruction
*ins
)
2708 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2709 DWORD reg
= ins
->dst
[0].register_idx
;
2710 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
2711 glsl_src_param_t src0_param
;
2712 DWORD sampler_type
= ins
->ctx
->reg_maps
->samplers
[reg
] & WINED3DSP_TEXTURETYPE_MASK
;
2713 glsl_sample_function_t sample_function
;
2715 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
2716 shader_addline(buffer
, "tmp0.y = dot(T%u.xyz, %s);\n", reg
, src0_param
.param_str
);
2718 shader_glsl_get_sample_function(sampler_type
, 0, &sample_function
);
2720 /* Sample the texture using the calculated coordinates */
2721 shader_glsl_gen_sample_code(ins
, reg
, &sample_function
, WINED3DVS_NOSWIZZLE
, NULL
, "tmp0.xy");
2724 /** Process the WINED3DSIO_TEXM3X3TEX instruction in GLSL
2725 * Perform the 3rd row of a 3x3 matrix multiply, then sample the texture using the calculated coordinates */
2726 static void pshader_glsl_texm3x3tex(const struct wined3d_shader_instruction
*ins
)
2728 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2729 glsl_src_param_t src0_param
;
2730 DWORD reg
= ins
->dst
[0].register_idx
;
2731 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*)ins
->ctx
->shader
;
2732 SHADER_PARSE_STATE
* current_state
= &This
->baseShader
.parse_state
;
2733 DWORD sampler_type
= ins
->ctx
->reg_maps
->samplers
[reg
] & WINED3DSP_TEXTURETYPE_MASK
;
2734 glsl_sample_function_t sample_function
;
2736 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
2737 shader_addline(ins
->ctx
->buffer
, "tmp0.z = dot(T%u.xyz, %s);\n", reg
, src0_param
.param_str
);
2739 /* Dependent read, not valid with conditional NP2 */
2740 shader_glsl_get_sample_function(sampler_type
, 0, &sample_function
);
2742 /* Sample the texture using the calculated coordinates */
2743 shader_glsl_gen_sample_code(ins
, reg
, &sample_function
, WINED3DVS_NOSWIZZLE
, NULL
, "tmp0.xyz");
2745 current_state
->current_row
= 0;
2748 /** Process the WINED3DSIO_TEXM3X3 instruction in GLSL
2749 * Perform the 3rd row of a 3x3 matrix multiply */
2750 static void pshader_glsl_texm3x3(const struct wined3d_shader_instruction
*ins
)
2752 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2753 glsl_src_param_t src0_param
;
2755 DWORD reg
= ins
->dst
[0].register_idx
;
2756 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*)ins
->ctx
->shader
;
2757 SHADER_PARSE_STATE
* current_state
= &This
->baseShader
.parse_state
;
2759 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
2761 shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2762 shader_glsl_get_write_mask(&ins
->dst
[0], dst_mask
);
2763 shader_addline(ins
->ctx
->buffer
, "vec4(tmp0.xy, dot(T%u.xyz, %s), 1.0)%s);\n", reg
, src0_param
.param_str
, dst_mask
);
2765 current_state
->current_row
= 0;
2768 /** Process the WINED3DSIO_TEXM3X3SPEC instruction in GLSL
2769 * Perform the final texture lookup based on the previous 2 3x3 matrix multiplies */
2770 static void pshader_glsl_texm3x3spec(const struct wined3d_shader_instruction
*ins
)
2772 IWineD3DPixelShaderImpl
*shader
= (IWineD3DPixelShaderImpl
*)ins
->ctx
->shader
;
2773 DWORD reg
= ins
->dst
[0].register_idx
;
2774 glsl_src_param_t src0_param
;
2775 glsl_src_param_t src1_param
;
2776 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
2777 SHADER_PARSE_STATE
* current_state
= &shader
->baseShader
.parse_state
;
2778 DWORD stype
= ins
->ctx
->reg_maps
->samplers
[reg
] & WINED3DSP_TEXTURETYPE_MASK
;
2779 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2780 glsl_sample_function_t sample_function
;
2782 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
2783 shader_glsl_add_src_param(ins
, &ins
->src
[1], src_mask
, &src1_param
);
2785 /* Perform the last matrix multiply operation */
2786 shader_addline(buffer
, "tmp0.z = dot(T%u.xyz, %s);\n", reg
, src0_param
.param_str
);
2787 /* Reflection calculation */
2788 shader_addline(buffer
, "tmp0.xyz = -reflect((%s), normalize(tmp0.xyz));\n", src1_param
.param_str
);
2790 /* Dependent read, not valid with conditional NP2 */
2791 shader_glsl_get_sample_function(stype
, 0, &sample_function
);
2793 /* Sample the texture */
2794 shader_glsl_gen_sample_code(ins
, reg
, &sample_function
, WINED3DVS_NOSWIZZLE
, NULL
, "tmp0.xyz");
2796 current_state
->current_row
= 0;
2799 /** Process the WINED3DSIO_TEXM3X3VSPEC instruction in GLSL
2800 * Perform the final texture lookup based on the previous 2 3x3 matrix multiplies */
2801 static void pshader_glsl_texm3x3vspec(const struct wined3d_shader_instruction
*ins
)
2803 IWineD3DPixelShaderImpl
*shader
= (IWineD3DPixelShaderImpl
*)ins
->ctx
->shader
;
2804 DWORD reg
= ins
->dst
[0].register_idx
;
2805 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
2806 SHADER_PARSE_STATE
* current_state
= &shader
->baseShader
.parse_state
;
2807 glsl_src_param_t src0_param
;
2808 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2809 DWORD sampler_type
= ins
->ctx
->reg_maps
->samplers
[reg
] & WINED3DSP_TEXTURETYPE_MASK
;
2810 glsl_sample_function_t sample_function
;
2812 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
2814 /* Perform the last matrix multiply operation */
2815 shader_addline(buffer
, "tmp0.z = dot(vec3(T%u), vec3(%s));\n", reg
, src0_param
.param_str
);
2817 /* Construct the eye-ray vector from w coordinates */
2818 shader_addline(buffer
, "tmp1.xyz = normalize(vec3(gl_TexCoord[%u].w, gl_TexCoord[%u].w, gl_TexCoord[%u].w));\n",
2819 current_state
->texcoord_w
[0], current_state
->texcoord_w
[1], reg
);
2820 shader_addline(buffer
, "tmp0.xyz = -reflect(tmp1.xyz, normalize(tmp0.xyz));\n");
2822 /* Dependent read, not valid with conditional NP2 */
2823 shader_glsl_get_sample_function(sampler_type
, 0, &sample_function
);
2825 /* Sample the texture using the calculated coordinates */
2826 shader_glsl_gen_sample_code(ins
, reg
, &sample_function
, WINED3DVS_NOSWIZZLE
, NULL
, "tmp0.xyz");
2828 current_state
->current_row
= 0;
2831 /** Process the WINED3DSIO_TEXBEM instruction in GLSL.
2832 * Apply a fake bump map transform.
2833 * texbem is pshader <= 1.3 only, this saves a few version checks
2835 static void pshader_glsl_texbem(const struct wined3d_shader_instruction
*ins
)
2837 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*)ins
->ctx
->shader
;
2838 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
2839 glsl_sample_function_t sample_function
;
2840 glsl_src_param_t coord_param
;
2847 sampler_idx
= ins
->dst
[0].register_idx
;
2848 flags
= deviceImpl
->stateBlock
->textureState
[sampler_idx
][WINED3DTSS_TEXTURETRANSFORMFLAGS
];
2850 sampler_type
= ins
->ctx
->reg_maps
->samplers
[sampler_idx
] & WINED3DSP_TEXTURETYPE_MASK
;
2851 /* Dependent read, not valid with conditional NP2 */
2852 shader_glsl_get_sample_function(sampler_type
, 0, &sample_function
);
2853 mask
= sample_function
.coord_mask
;
2855 shader_glsl_write_mask_to_str(mask
, coord_mask
);
2857 /* with projective textures, texbem only divides the static texture coord, not the displacement,
2858 * so we can't let the GL handle this.
2860 if (flags
& WINED3DTTFF_PROJECTED
) {
2862 char coord_div_mask
[3];
2863 switch (flags
& ~WINED3DTTFF_PROJECTED
) {
2864 case WINED3DTTFF_COUNT1
: FIXME("WINED3DTTFF_PROJECTED with WINED3DTTFF_COUNT1?\n"); break;
2865 case WINED3DTTFF_COUNT2
: div_mask
= WINED3DSP_WRITEMASK_1
; break;
2866 case WINED3DTTFF_COUNT3
: div_mask
= WINED3DSP_WRITEMASK_2
; break;
2867 case WINED3DTTFF_COUNT4
:
2868 case WINED3DTTFF_DISABLE
: div_mask
= WINED3DSP_WRITEMASK_3
; break;
2870 shader_glsl_write_mask_to_str(div_mask
, coord_div_mask
);
2871 shader_addline(ins
->ctx
->buffer
, "T%u%s /= T%u%s;\n", sampler_idx
, coord_mask
, sampler_idx
, coord_div_mask
);
2874 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
, &coord_param
);
2876 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, WINED3DVS_NOSWIZZLE
, NULL
,
2877 "T%u%s + vec4(bumpenvmat%d * %s, 0.0, 0.0)%s", sampler_idx
, coord_mask
, sampler_idx
,
2878 coord_param
.param_str
, coord_mask
);
2880 if (ins
->handler_idx
== WINED3DSIH_TEXBEML
)
2882 glsl_src_param_t luminance_param
;
2883 glsl_dst_param_t dst_param
;
2885 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_2
, &luminance_param
);
2886 shader_glsl_add_dst_param(ins
, &ins
->dst
[0], &dst_param
);
2888 shader_addline(ins
->ctx
->buffer
, "%s%s *= (%s * luminancescale%d + luminanceoffset%d);\n",
2889 dst_param
.reg_name
, dst_param
.mask_str
,
2890 luminance_param
.param_str
, sampler_idx
, sampler_idx
);
2894 static void pshader_glsl_bem(const struct wined3d_shader_instruction
*ins
)
2896 glsl_src_param_t src0_param
, src1_param
;
2897 DWORD sampler_idx
= ins
->dst
[0].register_idx
;
2899 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
, &src0_param
);
2900 shader_glsl_add_src_param(ins
, &ins
->src
[1], WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
, &src1_param
);
2902 shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2903 shader_addline(ins
->ctx
->buffer
, "%s + bumpenvmat%d * %s);\n",
2904 src0_param
.param_str
, sampler_idx
, src1_param
.param_str
);
2907 /** Process the WINED3DSIO_TEXREG2AR instruction in GLSL
2908 * Sample 2D texture at dst using the alpha & red (wx) components of src as texture coordinates */
2909 static void pshader_glsl_texreg2ar(const struct wined3d_shader_instruction
*ins
)
2911 glsl_src_param_t src0_param
;
2912 DWORD sampler_idx
= ins
->dst
[0].register_idx
;
2913 DWORD sampler_type
= ins
->ctx
->reg_maps
->samplers
[sampler_idx
] & WINED3DSP_TEXTURETYPE_MASK
;
2914 glsl_sample_function_t sample_function
;
2916 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_ALL
, &src0_param
);
2918 shader_glsl_get_sample_function(sampler_type
, 0, &sample_function
);
2919 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, WINED3DVS_NOSWIZZLE
, NULL
,
2920 "%s.wx", src0_param
.reg_name
);
2923 /** Process the WINED3DSIO_TEXREG2GB instruction in GLSL
2924 * Sample 2D texture at dst using the green & blue (yz) components of src as texture coordinates */
2925 static void pshader_glsl_texreg2gb(const struct wined3d_shader_instruction
*ins
)
2927 glsl_src_param_t src0_param
;
2928 DWORD sampler_idx
= ins
->dst
[0].register_idx
;
2929 DWORD sampler_type
= ins
->ctx
->reg_maps
->samplers
[sampler_idx
] & WINED3DSP_TEXTURETYPE_MASK
;
2930 glsl_sample_function_t sample_function
;
2932 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_ALL
, &src0_param
);
2934 shader_glsl_get_sample_function(sampler_type
, 0, &sample_function
);
2935 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, WINED3DVS_NOSWIZZLE
, NULL
,
2936 "%s.yz", src0_param
.reg_name
);
2939 /** Process the WINED3DSIO_TEXREG2RGB instruction in GLSL
2940 * Sample texture at dst using the rgb (xyz) components of src as texture coordinates */
2941 static void pshader_glsl_texreg2rgb(const struct wined3d_shader_instruction
*ins
)
2943 glsl_src_param_t src0_param
;
2944 DWORD sampler_idx
= ins
->dst
[0].register_idx
;
2945 DWORD sampler_type
= ins
->ctx
->reg_maps
->samplers
[sampler_idx
] & WINED3DSP_TEXTURETYPE_MASK
;
2946 glsl_sample_function_t sample_function
;
2948 /* Dependent read, not valid with conditional NP2 */
2949 shader_glsl_get_sample_function(sampler_type
, 0, &sample_function
);
2950 shader_glsl_add_src_param(ins
, &ins
->src
[0], sample_function
.coord_mask
, &src0_param
);
2952 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, WINED3DVS_NOSWIZZLE
, NULL
,
2953 "%s", src0_param
.param_str
);
2956 /** Process the WINED3DSIO_TEXKILL instruction in GLSL.
2957 * If any of the first 3 components are < 0, discard this pixel */
2958 static void pshader_glsl_texkill(const struct wined3d_shader_instruction
*ins
)
2960 glsl_dst_param_t dst_param
;
2962 /* The argument is a destination parameter, and no writemasks are allowed */
2963 shader_glsl_add_dst_param(ins
, &ins
->dst
[0], &dst_param
);
2964 if ((ins
->ctx
->reg_maps
->shader_version
>= WINED3DPS_VERSION(2,0)))
2966 /* 2.0 shaders compare all 4 components in texkill */
2967 shader_addline(ins
->ctx
->buffer
, "if (any(lessThan(%s.xyzw, vec4(0.0)))) discard;\n", dst_param
.reg_name
);
2969 /* 1.X shaders only compare the first 3 components, probably due to the nature of the texkill
2970 * instruction as a tex* instruction, and phase, which kills all a / w components. Even if all
2971 * 4 components are defined, only the first 3 are used
2973 shader_addline(ins
->ctx
->buffer
, "if (any(lessThan(%s.xyz, vec3(0.0)))) discard;\n", dst_param
.reg_name
);
2977 /** Process the WINED3DSIO_DP2ADD instruction in GLSL.
2978 * dst = dot2(src0, src1) + src2 */
2979 static void pshader_glsl_dp2add(const struct wined3d_shader_instruction
*ins
)
2981 glsl_src_param_t src0_param
;
2982 glsl_src_param_t src1_param
;
2983 glsl_src_param_t src2_param
;
2985 unsigned int mask_size
;
2987 write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2988 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
2990 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
, &src0_param
);
2991 shader_glsl_add_src_param(ins
, &ins
->src
[1], WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
, &src1_param
);
2992 shader_glsl_add_src_param(ins
, &ins
->src
[2], WINED3DSP_WRITEMASK_0
, &src2_param
);
2994 if (mask_size
> 1) {
2995 shader_addline(ins
->ctx
->buffer
, "vec%d(dot(%s, %s) + %s));\n",
2996 mask_size
, src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
2998 shader_addline(ins
->ctx
->buffer
, "dot(%s, %s) + %s);\n",
2999 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
3003 static void pshader_glsl_input_pack(IWineD3DPixelShader
*iface
, SHADER_BUFFER
*buffer
,
3004 const struct wined3d_shader_semantic
*semantics_in
, const struct shader_reg_maps
*reg_maps
,
3005 enum vertexprocessing_mode vertexprocessing
)
3008 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*)iface
;
3010 for (i
= 0; i
< MAX_REG_INPUT
; ++i
)
3012 DWORD usage
, usage_idx
;
3016 if (!reg_maps
->packed_input
[i
]) continue;
3018 usage
= semantics_in
[i
].usage
;
3019 usage_idx
= semantics_in
[i
].usage_idx
;
3020 shader_glsl_get_write_mask(&semantics_in
[i
].reg
, reg_mask
);
3024 case WINED3DDECLUSAGE_TEXCOORD
:
3025 if (usage_idx
< 8 && vertexprocessing
== pretransformed
)
3026 shader_addline(buffer
, "IN[%u]%s = gl_TexCoord[%u]%s;\n",
3027 This
->input_reg_map
[i
], reg_mask
, usage_idx
, reg_mask
);
3029 shader_addline(buffer
, "IN[%u]%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
3030 This
->input_reg_map
[i
], reg_mask
, reg_mask
);
3033 case WINED3DDECLUSAGE_COLOR
:
3035 shader_addline(buffer
, "IN[%u]%s = vec4(gl_Color)%s;\n",
3036 This
->input_reg_map
[i
], reg_mask
, reg_mask
);
3037 else if (usage_idx
== 1)
3038 shader_addline(buffer
, "IN[%u]%s = vec4(gl_SecondaryColor)%s;\n",
3039 This
->input_reg_map
[i
], reg_mask
, reg_mask
);
3041 shader_addline(buffer
, "IN[%u]%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
3042 This
->input_reg_map
[i
], reg_mask
, reg_mask
);
3046 shader_addline(buffer
, "IN[%u]%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
3047 This
->input_reg_map
[i
], reg_mask
, reg_mask
);
3053 /*********************************************
3054 * Vertex Shader Specific Code begins here
3055 ********************************************/
3057 static void add_glsl_program_entry(struct shader_glsl_priv
*priv
, struct glsl_shader_prog_link
*entry
) {
3058 glsl_program_key_t
*key
;
3060 key
= HeapAlloc(GetProcessHeap(), 0, sizeof(glsl_program_key_t
));
3061 key
->vshader
= entry
->vshader
;
3062 key
->pshader
= entry
->pshader
;
3063 key
->vs_args
= entry
->vs_args
;
3064 key
->ps_args
= entry
->ps_args
;
3066 hash_table_put(priv
->glsl_program_lookup
, key
, entry
);
3069 static struct glsl_shader_prog_link
*get_glsl_program_entry(struct shader_glsl_priv
*priv
,
3070 IWineD3DVertexShader
*vshader
, IWineD3DPixelShader
*pshader
, struct vs_compile_args
*vs_args
,
3071 struct ps_compile_args
*ps_args
) {
3072 glsl_program_key_t key
;
3074 key
.vshader
= vshader
;
3075 key
.pshader
= pshader
;
3076 key
.vs_args
= *vs_args
;
3077 key
.ps_args
= *ps_args
;
3079 return hash_table_get(priv
->glsl_program_lookup
, &key
);
3082 static void delete_glsl_program_entry(struct shader_glsl_priv
*priv
, const WineD3D_GL_Info
*gl_info
,
3083 struct glsl_shader_prog_link
*entry
)
3085 glsl_program_key_t
*key
;
3087 key
= HeapAlloc(GetProcessHeap(), 0, sizeof(glsl_program_key_t
));
3088 key
->vshader
= entry
->vshader
;
3089 key
->pshader
= entry
->pshader
;
3090 key
->vs_args
= entry
->vs_args
;
3091 key
->ps_args
= entry
->ps_args
;
3092 hash_table_remove(priv
->glsl_program_lookup
, key
);
3094 GL_EXTCALL(glDeleteObjectARB(entry
->programId
));
3095 if (entry
->vshader
) list_remove(&entry
->vshader_entry
);
3096 if (entry
->pshader
) list_remove(&entry
->pshader_entry
);
3097 HeapFree(GetProcessHeap(), 0, entry
->vuniformF_locations
);
3098 HeapFree(GetProcessHeap(), 0, entry
->puniformF_locations
);
3099 HeapFree(GetProcessHeap(), 0, entry
);
3102 static void handle_ps3_input(SHADER_BUFFER
*buffer
, const WineD3D_GL_Info
*gl_info
, const DWORD
*map
,
3103 const struct wined3d_shader_semantic
*semantics_in
, const struct shader_reg_maps
*reg_maps_in
,
3104 const struct wined3d_shader_semantic
*semantics_out
, const struct shader_reg_maps
*reg_maps_out
)
3107 DWORD usage
, usage_idx
, usage_out
, usage_idx_out
;
3110 DWORD in_count
= GL_LIMITS(glsl_varyings
) / 4;
3111 char reg_mask
[6], reg_mask_out
[6];
3112 char destination
[50];
3114 set
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*set
) * (in_count
+ 2));
3116 if (!semantics_out
) {
3117 /* Save gl_FrontColor & gl_FrontSecondaryColor before overwriting them. */
3118 shader_addline(buffer
, "vec4 front_color = gl_FrontColor;\n");
3119 shader_addline(buffer
, "vec4 front_secondary_color = gl_FrontSecondaryColor;\n");
3122 for(i
= 0; i
< MAX_REG_INPUT
; i
++) {
3123 if (!reg_maps_in
->packed_input
[i
]) continue;
3126 if (in_idx
>= (in_count
+ 2)) {
3127 FIXME("More input varyings declared than supported, expect issues\n");
3130 else if (map
[i
] == ~0U)
3132 /* Declared, but not read register */
3136 if (in_idx
== in_count
) {
3137 sprintf(destination
, "gl_FrontColor");
3138 } else if (in_idx
== in_count
+ 1) {
3139 sprintf(destination
, "gl_FrontSecondaryColor");
3141 sprintf(destination
, "IN[%u]", in_idx
);
3144 usage
= semantics_in
[i
].usage
;
3145 usage_idx
= semantics_in
[i
].usage_idx
;
3146 set
[map
[i
]] = shader_glsl_get_write_mask(&semantics_in
[i
].reg
, reg_mask
);
3148 if(!semantics_out
) {
3150 case WINED3DDECLUSAGE_COLOR
:
3152 shader_addline(buffer
, "%s%s = front_color%s;\n",
3153 destination
, reg_mask
, reg_mask
);
3154 else if (usage_idx
== 1)
3155 shader_addline(buffer
, "%s%s = front_secondary_color%s;\n",
3156 destination
, reg_mask
, reg_mask
);
3158 shader_addline(buffer
, "%s%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
3159 destination
, reg_mask
, reg_mask
);
3162 case WINED3DDECLUSAGE_TEXCOORD
:
3163 if (usage_idx
< 8) {
3164 shader_addline(buffer
, "%s%s = gl_TexCoord[%u]%s;\n",
3165 destination
, reg_mask
, usage_idx
, reg_mask
);
3167 shader_addline(buffer
, "%s%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
3168 destination
, reg_mask
, reg_mask
);
3172 case WINED3DDECLUSAGE_FOG
:
3173 shader_addline(buffer
, "%s%s = vec4(gl_FogFragCoord, 0.0, 0.0, 0.0)%s;\n",
3174 destination
, reg_mask
, reg_mask
);
3178 shader_addline(buffer
, "%s%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
3179 destination
, reg_mask
, reg_mask
);
3183 for(j
= 0; j
< MAX_REG_OUTPUT
; j
++) {
3184 if (!reg_maps_out
->packed_output
[j
]) continue;
3186 usage_out
= semantics_out
[j
].usage
;
3187 usage_idx_out
= semantics_out
[j
].usage_idx
;
3188 shader_glsl_get_write_mask(&semantics_out
[j
].reg
, reg_mask_out
);
3190 if(usage
== usage_out
&&
3191 usage_idx
== usage_idx_out
) {
3192 shader_addline(buffer
, "%s%s = OUT[%u]%s;\n",
3193 destination
, reg_mask
, j
, reg_mask
);
3198 shader_addline(buffer
, "%s%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
3199 destination
, reg_mask
, reg_mask
);
3204 /* This is solely to make the compiler / linker happy and avoid warning about undefined
3205 * varyings. It shouldn't result in any real code executed on the GPU, since all read
3206 * input varyings are assigned above, if the optimizer works properly.
3208 for(i
= 0; i
< in_count
+ 2; i
++) {
3209 if(set
[i
] != WINED3DSP_WRITEMASK_ALL
) {
3210 unsigned int size
= 0;
3211 memset(reg_mask
, 0, sizeof(reg_mask
));
3212 if(!(set
[i
] & WINED3DSP_WRITEMASK_0
)) {
3213 reg_mask
[size
] = 'x';
3216 if(!(set
[i
] & WINED3DSP_WRITEMASK_1
)) {
3217 reg_mask
[size
] = 'y';
3220 if(!(set
[i
] & WINED3DSP_WRITEMASK_2
)) {
3221 reg_mask
[size
] = 'z';
3224 if(!(set
[i
] & WINED3DSP_WRITEMASK_3
)) {
3225 reg_mask
[size
] = 'w';
3229 if (i
== in_count
) {
3230 sprintf(destination
, "gl_FrontColor");
3231 } else if (i
== in_count
+ 1) {
3232 sprintf(destination
, "gl_FrontSecondaryColor");
3234 sprintf(destination
, "IN[%u]", i
);
3238 shader_addline(buffer
, "%s.%s = 0.0;\n", destination
, reg_mask
);
3240 shader_addline(buffer
, "%s.%s = vec%u(0.0);\n", destination
, reg_mask
, size
);
3245 HeapFree(GetProcessHeap(), 0, set
);
3248 static GLhandleARB
generate_param_reorder_function(IWineD3DVertexShader
*vertexshader
,
3249 IWineD3DPixelShader
*pixelshader
, const WineD3D_GL_Info
*gl_info
)
3251 GLhandleARB ret
= 0;
3252 IWineD3DVertexShaderImpl
*vs
= (IWineD3DVertexShaderImpl
*) vertexshader
;
3253 IWineD3DPixelShaderImpl
*ps
= (IWineD3DPixelShaderImpl
*) pixelshader
;
3254 IWineD3DDeviceImpl
*device
;
3255 DWORD vs_major
= WINED3DSHADER_VERSION_MAJOR(vs
->baseShader
.reg_maps
.shader_version
);
3256 DWORD ps_major
= ps
? WINED3DSHADER_VERSION_MAJOR(ps
->baseShader
.reg_maps
.shader_version
) : 0;
3258 SHADER_BUFFER buffer
;
3259 DWORD usage
, usage_idx
, writemask
;
3261 const struct wined3d_shader_semantic
*semantics_out
;
3263 shader_buffer_init(&buffer
);
3265 shader_addline(&buffer
, "#version 120\n");
3267 if(vs_major
< 3 && ps_major
< 3) {
3268 /* That one is easy: The vertex shader writes to the builtin varyings, the pixel shader reads from them.
3269 * Take care about the texcoord .w fixup though if we're using the fixed function fragment pipeline
3271 device
= (IWineD3DDeviceImpl
*) vs
->baseShader
.device
;
3272 if((GLINFO_LOCATION
).set_texcoord_w
&& ps_major
== 0 && vs_major
> 0 &&
3273 !device
->frag_pipe
->ffp_proj_control
) {
3274 shader_addline(&buffer
, "void order_ps_input() {\n");
3275 for(i
= 0; i
< min(8, MAX_REG_TEXCRD
); i
++) {
3276 if(vs
->baseShader
.reg_maps
.texcoord_mask
[i
] != 0 &&
3277 vs
->baseShader
.reg_maps
.texcoord_mask
[i
] != WINED3DSP_WRITEMASK_ALL
) {
3278 shader_addline(&buffer
, "gl_TexCoord[%u].w = 1.0;\n", i
);
3281 shader_addline(&buffer
, "}\n");
3283 shader_addline(&buffer
, "void order_ps_input() { /* do nothing */ }\n");
3285 } else if(ps_major
< 3 && vs_major
>= 3) {
3286 /* The vertex shader writes to its own varyings, the pixel shader needs them in the builtin ones */
3287 semantics_out
= vs
->semantics_out
;
3289 shader_addline(&buffer
, "void order_ps_input(in vec4 OUT[%u]) {\n", MAX_REG_OUTPUT
);
3290 for(i
= 0; i
< MAX_REG_OUTPUT
; i
++) {
3291 if (!vs
->baseShader
.reg_maps
.packed_output
[i
]) continue;
3293 usage
= semantics_out
[i
].usage
;
3294 usage_idx
= semantics_out
[i
].usage_idx
;
3295 writemask
= shader_glsl_get_write_mask(&semantics_out
[i
].reg
, reg_mask
);
3298 case WINED3DDECLUSAGE_COLOR
:
3300 shader_addline(&buffer
, "gl_FrontColor%s = OUT[%u]%s;\n", reg_mask
, i
, reg_mask
);
3301 else if (usage_idx
== 1)
3302 shader_addline(&buffer
, "gl_FrontSecondaryColor%s = OUT[%u]%s;\n", reg_mask
, i
, reg_mask
);
3305 case WINED3DDECLUSAGE_POSITION
:
3306 shader_addline(&buffer
, "gl_Position%s = OUT[%u]%s;\n", reg_mask
, i
, reg_mask
);
3309 case WINED3DDECLUSAGE_TEXCOORD
:
3310 if (usage_idx
< 8) {
3311 if(!(GLINFO_LOCATION
).set_texcoord_w
|| ps_major
> 0) writemask
|= WINED3DSP_WRITEMASK_3
;
3313 shader_addline(&buffer
, "gl_TexCoord[%u]%s = OUT[%u]%s;\n",
3314 usage_idx
, reg_mask
, i
, reg_mask
);
3315 if(!(writemask
& WINED3DSP_WRITEMASK_3
)) {
3316 shader_addline(&buffer
, "gl_TexCoord[%u].w = 1.0;\n", usage_idx
);
3321 case WINED3DDECLUSAGE_PSIZE
:
3322 shader_addline(&buffer
, "gl_PointSize = OUT[%u].x;\n", i
);
3325 case WINED3DDECLUSAGE_FOG
:
3326 shader_addline(&buffer
, "gl_FogFragCoord = OUT[%u].%c;\n", i
, reg_mask
[1]);
3333 shader_addline(&buffer
, "}\n");
3335 } else if(ps_major
>= 3 && vs_major
>= 3) {
3336 semantics_out
= vs
->semantics_out
;
3338 /* This one is tricky: a 3.0 pixel shader reads from a 3.0 vertex shader */
3339 shader_addline(&buffer
, "varying vec4 IN[%u];\n", GL_LIMITS(glsl_varyings
) / 4);
3340 shader_addline(&buffer
, "void order_ps_input(in vec4 OUT[%u]) {\n", MAX_REG_OUTPUT
);
3342 /* First, sort out position and point size. Those are not passed to the pixel shader */
3343 for(i
= 0; i
< MAX_REG_OUTPUT
; i
++) {
3344 if (!vs
->baseShader
.reg_maps
.packed_output
[i
]) continue;
3346 usage
= semantics_out
[i
].usage
;
3347 usage_idx
= semantics_out
[i
].usage_idx
;
3348 shader_glsl_get_write_mask(&semantics_out
[i
].reg
, reg_mask
);
3351 case WINED3DDECLUSAGE_POSITION
:
3352 shader_addline(&buffer
, "gl_Position%s = OUT[%u]%s;\n", reg_mask
, i
, reg_mask
);
3355 case WINED3DDECLUSAGE_PSIZE
:
3356 shader_addline(&buffer
, "gl_PointSize = OUT[%u].x;\n", i
);
3364 /* Then, fix the pixel shader input */
3365 handle_ps3_input(&buffer
, gl_info
, ps
->input_reg_map
,
3366 ps
->semantics_in
, &ps
->baseShader
.reg_maps
, semantics_out
, &vs
->baseShader
.reg_maps
);
3368 shader_addline(&buffer
, "}\n");
3369 } else if(ps_major
>= 3 && vs_major
< 3) {
3370 shader_addline(&buffer
, "varying vec4 IN[%u];\n", GL_LIMITS(glsl_varyings
) / 4);
3371 shader_addline(&buffer
, "void order_ps_input() {\n");
3372 /* The vertex shader wrote to the builtin varyings. There is no need to figure out position and
3373 * point size, but we depend on the optimizers kindness to find out that the pixel shader doesn't
3374 * read gl_TexCoord and gl_ColorX, otherwise we'll run out of varyings
3376 handle_ps3_input(&buffer
, gl_info
, ps
->input_reg_map
, ps
->semantics_in
, &ps
->baseShader
.reg_maps
, NULL
, NULL
);
3377 shader_addline(&buffer
, "}\n");
3379 ERR("Unexpected vertex and pixel shader version condition: vs: %d, ps: %d\n", vs_major
, ps_major
);
3382 ret
= GL_EXTCALL(glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB
));
3383 checkGLcall("glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB)");
3384 GL_EXTCALL(glShaderSourceARB(ret
, 1, (const char**)&buffer
.buffer
, NULL
));
3385 checkGLcall("glShaderSourceARB(ret, 1, &buffer.buffer, NULL)");
3386 GL_EXTCALL(glCompileShaderARB(ret
));
3387 checkGLcall("glCompileShaderARB(ret)");
3389 shader_buffer_free(&buffer
);
3393 static void hardcode_local_constants(IWineD3DBaseShaderImpl
*shader
, const WineD3D_GL_Info
*gl_info
,
3394 GLhandleARB programId
, char prefix
)
3396 const local_constant
*lconst
;
3401 LIST_FOR_EACH_ENTRY(lconst
, &shader
->baseShader
.constantsF
, local_constant
, entry
) {
3402 value
= (const float *)lconst
->value
;
3403 snprintf(glsl_name
, sizeof(glsl_name
), "%cLC%u", prefix
, lconst
->idx
);
3404 tmp_loc
= GL_EXTCALL(glGetUniformLocationARB(programId
, glsl_name
));
3405 GL_EXTCALL(glUniform4fvARB(tmp_loc
, 1, value
));
3407 checkGLcall("Hardcoding local constants\n");
3410 /** Sets the GLSL program ID for the given pixel and vertex shader combination.
3411 * It sets the programId on the current StateBlock (because it should be called
3412 * inside of the DrawPrimitive() part of the render loop).
3414 * If a program for the given combination does not exist, create one, and store
3415 * the program in the hash table. If it creates a program, it will link the
3416 * given objects, too.
3418 static void set_glsl_shader_program(IWineD3DDevice
*iface
, BOOL use_ps
, BOOL use_vs
) {
3419 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
3420 struct shader_glsl_priv
*priv
= This
->shader_priv
;
3421 const WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
3422 IWineD3DPixelShader
*pshader
= This
->stateBlock
->pixelShader
;
3423 IWineD3DVertexShader
*vshader
= This
->stateBlock
->vertexShader
;
3424 struct glsl_shader_prog_link
*entry
= NULL
;
3425 GLhandleARB programId
= 0;
3426 GLhandleARB reorder_shader_id
= 0;
3429 GLhandleARB vshader_id
, pshader_id
;
3430 struct ps_compile_args ps_compile_args
;
3431 struct vs_compile_args vs_compile_args
;
3434 find_vs_compile_args((IWineD3DVertexShaderImpl
*)This
->stateBlock
->vertexShader
, This
->stateBlock
, &vs_compile_args
);
3436 /* FIXME: Do we really have to spend CPU cycles to generate a few zeroed bytes? */
3437 memset(&vs_compile_args
, 0, sizeof(vs_compile_args
));
3440 find_ps_compile_args((IWineD3DPixelShaderImpl
*)This
->stateBlock
->pixelShader
, This
->stateBlock
, &ps_compile_args
);
3442 /* FIXME: Do we really have to spend CPU cycles to generate a few zeroed bytes? */
3443 memset(&ps_compile_args
, 0, sizeof(ps_compile_args
));
3445 entry
= get_glsl_program_entry(priv
, vshader
, pshader
, &vs_compile_args
, &ps_compile_args
);
3447 priv
->glsl_program
= entry
;
3451 /* If we get to this point, then no matching program exists, so we create one */
3452 programId
= GL_EXTCALL(glCreateProgramObjectARB());
3453 TRACE("Created new GLSL shader program %u\n", programId
);
3455 /* Create the entry */
3456 entry
= HeapAlloc(GetProcessHeap(), 0, sizeof(struct glsl_shader_prog_link
));
3457 entry
->programId
= programId
;
3458 entry
->vshader
= vshader
;
3459 entry
->pshader
= pshader
;
3460 entry
->vs_args
= vs_compile_args
;
3461 entry
->ps_args
= ps_compile_args
;
3462 entry
->constant_version
= 0;
3463 /* Add the hash table entry */
3464 add_glsl_program_entry(priv
, entry
);
3466 /* Set the current program */
3467 priv
->glsl_program
= entry
;
3470 vshader_id
= find_gl_vshader((IWineD3DVertexShaderImpl
*) vshader
, &vs_compile_args
);
3475 /* Attach GLSL vshader */
3477 const unsigned int max_attribs
= 16; /* TODO: Will this always be the case? It is at the moment... */
3480 reorder_shader_id
= generate_param_reorder_function(vshader
, pshader
, gl_info
);
3481 TRACE("Attaching GLSL shader object %u to program %u\n", reorder_shader_id
, programId
);
3482 GL_EXTCALL(glAttachObjectARB(programId
, reorder_shader_id
));
3483 checkGLcall("glAttachObjectARB");
3484 /* Flag the reorder function for deletion, then it will be freed automatically when the program
3487 GL_EXTCALL(glDeleteObjectARB(reorder_shader_id
));
3489 TRACE("Attaching GLSL shader object %u to program %u\n", vshader_id
, programId
);
3490 GL_EXTCALL(glAttachObjectARB(programId
, vshader_id
));
3491 checkGLcall("glAttachObjectARB");
3493 /* Bind vertex attributes to a corresponding index number to match
3494 * the same index numbers as ARB_vertex_programs (makes loading
3495 * vertex attributes simpler). With this method, we can use the
3496 * exact same code to load the attributes later for both ARB and
3499 * We have to do this here because we need to know the Program ID
3500 * in order to make the bindings work, and it has to be done prior
3501 * to linking the GLSL program. */
3502 for (i
= 0; i
< max_attribs
; ++i
) {
3503 if (((IWineD3DBaseShaderImpl
*)vshader
)->baseShader
.reg_maps
.attributes
[i
]) {
3504 snprintf(tmp_name
, sizeof(tmp_name
), "attrib%i", i
);
3505 GL_EXTCALL(glBindAttribLocationARB(programId
, i
, tmp_name
));
3508 checkGLcall("glBindAttribLocationARB");
3510 list_add_head(&((IWineD3DBaseShaderImpl
*)vshader
)->baseShader
.linked_programs
, &entry
->vshader_entry
);
3514 pshader_id
= find_gl_pshader((IWineD3DPixelShaderImpl
*) pshader
, &ps_compile_args
);
3519 /* Attach GLSL pshader */
3521 TRACE("Attaching GLSL shader object %u to program %u\n", pshader_id
, programId
);
3522 GL_EXTCALL(glAttachObjectARB(programId
, pshader_id
));
3523 checkGLcall("glAttachObjectARB");
3525 list_add_head(&((IWineD3DBaseShaderImpl
*)pshader
)->baseShader
.linked_programs
, &entry
->pshader_entry
);
3528 /* Link the program */
3529 TRACE("Linking GLSL shader program %u\n", programId
);
3530 GL_EXTCALL(glLinkProgramARB(programId
));
3531 print_glsl_info_log(&GLINFO_LOCATION
, programId
);
3533 entry
->vuniformF_locations
= HeapAlloc(GetProcessHeap(), 0, sizeof(GLhandleARB
) * GL_LIMITS(vshader_constantsF
));
3534 for (i
= 0; i
< GL_LIMITS(vshader_constantsF
); ++i
) {
3535 snprintf(glsl_name
, sizeof(glsl_name
), "VC[%i]", i
);
3536 entry
->vuniformF_locations
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, glsl_name
));
3538 for (i
= 0; i
< MAX_CONST_I
; ++i
) {
3539 snprintf(glsl_name
, sizeof(glsl_name
), "VI[%i]", i
);
3540 entry
->vuniformI_locations
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, glsl_name
));
3542 entry
->puniformF_locations
= HeapAlloc(GetProcessHeap(), 0, sizeof(GLhandleARB
) * GL_LIMITS(pshader_constantsF
));
3543 for (i
= 0; i
< GL_LIMITS(pshader_constantsF
); ++i
) {
3544 snprintf(glsl_name
, sizeof(glsl_name
), "PC[%i]", i
);
3545 entry
->puniformF_locations
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, glsl_name
));
3547 for (i
= 0; i
< MAX_CONST_I
; ++i
) {
3548 snprintf(glsl_name
, sizeof(glsl_name
), "PI[%i]", i
);
3549 entry
->puniformI_locations
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, glsl_name
));
3553 for(i
= 0; i
< ((IWineD3DPixelShaderImpl
*)pshader
)->numbumpenvmatconsts
; i
++) {
3555 sprintf(name
, "bumpenvmat%d", ((IWineD3DPixelShaderImpl
*)pshader
)->bumpenvmatconst
[i
].texunit
);
3556 entry
->bumpenvmat_location
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, name
));
3557 sprintf(name
, "luminancescale%d", ((IWineD3DPixelShaderImpl
*)pshader
)->luminanceconst
[i
].texunit
);
3558 entry
->luminancescale_location
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, name
));
3559 sprintf(name
, "luminanceoffset%d", ((IWineD3DPixelShaderImpl
*)pshader
)->luminanceconst
[i
].texunit
);
3560 entry
->luminanceoffset_location
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, name
));
3564 if (use_ps
&& ps_compile_args
.np2_fixup
) {
3566 for (i
= 0; i
< MAX_FRAGMENT_SAMPLERS
; ++i
) {
3567 if (ps_compile_args
.np2_fixup
& (1 << i
)) {
3568 sprintf(name
, "PsamplerNP2Fixup%u", i
);
3569 entry
->np2Fixup_location
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, name
));
3571 entry
->np2Fixup_location
[i
] = -1;
3576 entry
->posFixup_location
= GL_EXTCALL(glGetUniformLocationARB(programId
, "posFixup"));
3577 entry
->ycorrection_location
= GL_EXTCALL(glGetUniformLocationARB(programId
, "ycorrection"));
3578 checkGLcall("Find glsl program uniform locations");
3581 && WINED3DSHADER_VERSION_MAJOR(((IWineD3DPixelShaderImpl
*)pshader
)->baseShader
.reg_maps
.shader_version
) >= 3
3582 && ((IWineD3DPixelShaderImpl
*)pshader
)->declared_in_count
> GL_LIMITS(glsl_varyings
) / 4)
3584 TRACE("Shader %d needs vertex color clamping disabled\n", programId
);
3585 entry
->vertex_color_clamp
= GL_FALSE
;
3587 entry
->vertex_color_clamp
= GL_FIXED_ONLY_ARB
;
3590 /* Set the shader to allow uniform loading on it */
3591 GL_EXTCALL(glUseProgramObjectARB(programId
));
3592 checkGLcall("glUseProgramObjectARB(programId)");
3594 /* Load the vertex and pixel samplers now. The function that finds the mappings makes sure
3595 * that it stays the same for each vertexshader-pixelshader pair(=linked glsl program). If
3596 * a pshader with fixed function pipeline is used there are no vertex samplers, and if a
3597 * vertex shader with fixed function pixel processing is used we make sure that the card
3598 * supports enough samplers to allow the max number of vertex samplers with all possible
3599 * fixed function fragment processing setups. So once the program is linked these samplers
3603 /* Load vertex shader samplers */
3604 shader_glsl_load_vsamplers(gl_info
, This
->texUnitMap
, programId
);
3607 /* Load pixel shader samplers */
3608 shader_glsl_load_psamplers(gl_info
, This
->texUnitMap
, programId
);
3611 /* If the local constants do not have to be loaded with the environment constants,
3612 * load them now to have them hardcoded in the GLSL program. This saves some CPU cycles
3615 if(pshader
&& !((IWineD3DPixelShaderImpl
*)pshader
)->baseShader
.load_local_constsF
) {
3616 hardcode_local_constants((IWineD3DBaseShaderImpl
*) pshader
, gl_info
, programId
, 'P');
3618 if(vshader
&& !((IWineD3DVertexShaderImpl
*)vshader
)->baseShader
.load_local_constsF
) {
3619 hardcode_local_constants((IWineD3DBaseShaderImpl
*) vshader
, gl_info
, programId
, 'V');
3623 static GLhandleARB
create_glsl_blt_shader(const WineD3D_GL_Info
*gl_info
, enum tex_types tex_type
)
3625 GLhandleARB program_id
;
3626 GLhandleARB vshader_id
, pshader_id
;
3627 static const char *blt_vshader
[] =
3632 " gl_Position = gl_Vertex;\n"
3633 " gl_FrontColor = vec4(1.0);\n"
3634 " gl_TexCoord[0] = gl_MultiTexCoord0;\n"
3638 static const char *blt_pshaders
[tex_type_count
] =
3644 "uniform sampler2D sampler;\n"
3647 " gl_FragDepth = texture2D(sampler, gl_TexCoord[0].xy).x;\n"
3653 "uniform samplerCube sampler;\n"
3656 " gl_FragDepth = textureCube(sampler, gl_TexCoord[0].xyz).x;\n"
3660 "#extension GL_ARB_texture_rectangle : enable\n"
3661 "uniform sampler2DRect sampler;\n"
3664 " gl_FragDepth = texture2DRect(sampler, gl_TexCoord[0].xy).x;\n"
3668 if (!blt_pshaders
[tex_type
])
3670 FIXME("tex_type %#x not supported\n", tex_type
);
3674 vshader_id
= GL_EXTCALL(glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB
));
3675 GL_EXTCALL(glShaderSourceARB(vshader_id
, 1, blt_vshader
, NULL
));
3676 GL_EXTCALL(glCompileShaderARB(vshader_id
));
3678 pshader_id
= GL_EXTCALL(glCreateShaderObjectARB(GL_FRAGMENT_SHADER_ARB
));
3679 GL_EXTCALL(glShaderSourceARB(pshader_id
, 1, &blt_pshaders
[tex_type
], NULL
));
3680 GL_EXTCALL(glCompileShaderARB(pshader_id
));
3682 program_id
= GL_EXTCALL(glCreateProgramObjectARB());
3683 GL_EXTCALL(glAttachObjectARB(program_id
, vshader_id
));
3684 GL_EXTCALL(glAttachObjectARB(program_id
, pshader_id
));
3685 GL_EXTCALL(glLinkProgramARB(program_id
));
3687 print_glsl_info_log(&GLINFO_LOCATION
, program_id
);
3689 /* Once linked we can mark the shaders for deletion. They will be deleted once the program
3692 GL_EXTCALL(glDeleteObjectARB(vshader_id
));
3693 GL_EXTCALL(glDeleteObjectARB(pshader_id
));
3697 static void shader_glsl_select(IWineD3DDevice
*iface
, BOOL usePS
, BOOL useVS
) {
3698 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
3699 struct shader_glsl_priv
*priv
= This
->shader_priv
;
3700 const WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
3701 GLhandleARB program_id
= 0;
3702 GLenum old_vertex_color_clamp
, current_vertex_color_clamp
;
3704 old_vertex_color_clamp
= priv
->glsl_program
? priv
->glsl_program
->vertex_color_clamp
: GL_FIXED_ONLY_ARB
;
3706 if (useVS
|| usePS
) set_glsl_shader_program(iface
, usePS
, useVS
);
3707 else priv
->glsl_program
= NULL
;
3709 current_vertex_color_clamp
= priv
->glsl_program
? priv
->glsl_program
->vertex_color_clamp
: GL_FIXED_ONLY_ARB
;
3711 if (old_vertex_color_clamp
!= current_vertex_color_clamp
) {
3712 if (GL_SUPPORT(ARB_COLOR_BUFFER_FLOAT
)) {
3713 GL_EXTCALL(glClampColorARB(GL_CLAMP_VERTEX_COLOR_ARB
, current_vertex_color_clamp
));
3714 checkGLcall("glClampColorARB");
3716 FIXME("vertex color clamp needs to be changed, but extension not supported.\n");
3720 program_id
= priv
->glsl_program
? priv
->glsl_program
->programId
: 0;
3721 if (program_id
) TRACE("Using GLSL program %u\n", program_id
);
3722 GL_EXTCALL(glUseProgramObjectARB(program_id
));
3723 checkGLcall("glUseProgramObjectARB");
3726 static void shader_glsl_select_depth_blt(IWineD3DDevice
*iface
, enum tex_types tex_type
) {
3727 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
3728 const WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
3729 struct shader_glsl_priv
*priv
= This
->shader_priv
;
3730 GLhandleARB
*blt_program
= &priv
->depth_blt_program
[tex_type
];
3732 if (!*blt_program
) {
3734 *blt_program
= create_glsl_blt_shader(gl_info
, tex_type
);
3735 loc
= GL_EXTCALL(glGetUniformLocationARB(*blt_program
, "sampler"));
3736 GL_EXTCALL(glUseProgramObjectARB(*blt_program
));
3737 GL_EXTCALL(glUniform1iARB(loc
, 0));
3739 GL_EXTCALL(glUseProgramObjectARB(*blt_program
));
3743 static void shader_glsl_deselect_depth_blt(IWineD3DDevice
*iface
) {
3744 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
3745 const WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
3746 struct shader_glsl_priv
*priv
= This
->shader_priv
;
3747 GLhandleARB program_id
;
3749 program_id
= priv
->glsl_program
? priv
->glsl_program
->programId
: 0;
3750 if (program_id
) TRACE("Using GLSL program %u\n", program_id
);
3752 GL_EXTCALL(glUseProgramObjectARB(program_id
));
3753 checkGLcall("glUseProgramObjectARB");
3756 static void shader_glsl_destroy(IWineD3DBaseShader
*iface
) {
3757 const struct list
*linked_programs
;
3758 IWineD3DBaseShaderImpl
*This
= (IWineD3DBaseShaderImpl
*) iface
;
3759 IWineD3DDeviceImpl
*device
= (IWineD3DDeviceImpl
*)This
->baseShader
.device
;
3760 struct shader_glsl_priv
*priv
= device
->shader_priv
;
3761 const WineD3D_GL_Info
*gl_info
= &device
->adapter
->gl_info
;
3762 IWineD3DPixelShaderImpl
*ps
= NULL
;
3763 IWineD3DVertexShaderImpl
*vs
= NULL
;
3765 /* Note: Do not use QueryInterface here to find out which shader type this is because this code
3766 * can be called from IWineD3DBaseShader::Release
3768 char pshader
= shader_is_pshader_version(This
->baseShader
.reg_maps
.shader_version
);
3771 ps
= (IWineD3DPixelShaderImpl
*) This
;
3772 if(ps
->num_gl_shaders
== 0) return;
3773 if (priv
->glsl_program
&& (IWineD3DBaseShader
*)priv
->glsl_program
->pshader
== iface
)
3774 shader_glsl_select(This
->baseShader
.device
, FALSE
, FALSE
);
3776 vs
= (IWineD3DVertexShaderImpl
*) This
;
3777 if(vs
->num_gl_shaders
== 0) return;
3778 if (priv
->glsl_program
&& (IWineD3DBaseShader
*)priv
->glsl_program
->vshader
== iface
)
3779 shader_glsl_select(This
->baseShader
.device
, FALSE
, FALSE
);
3782 linked_programs
= &This
->baseShader
.linked_programs
;
3784 TRACE("Deleting linked programs\n");
3785 if (linked_programs
->next
) {
3786 struct glsl_shader_prog_link
*entry
, *entry2
;
3789 LIST_FOR_EACH_ENTRY_SAFE(entry
, entry2
, linked_programs
, struct glsl_shader_prog_link
, pshader_entry
) {
3790 delete_glsl_program_entry(priv
, gl_info
, entry
);
3793 LIST_FOR_EACH_ENTRY_SAFE(entry
, entry2
, linked_programs
, struct glsl_shader_prog_link
, vshader_entry
) {
3794 delete_glsl_program_entry(priv
, gl_info
, entry
);
3803 for(i
= 0; i
< ps
->num_gl_shaders
; i
++) {
3804 TRACE("deleting pshader %u\n", ps
->gl_shaders
[i
].prgId
);
3805 GL_EXTCALL(glDeleteObjectARB(ps
->gl_shaders
[i
].prgId
));
3806 checkGLcall("glDeleteObjectARB");
3809 HeapFree(GetProcessHeap(), 0, ps
->gl_shaders
);
3810 ps
->gl_shaders
= NULL
;
3811 ps
->num_gl_shaders
= 0;
3812 ps
->shader_array_size
= 0;
3817 for(i
= 0; i
< vs
->num_gl_shaders
; i
++) {
3818 TRACE("deleting vshader %u\n", vs
->gl_shaders
[i
].prgId
);
3819 GL_EXTCALL(glDeleteObjectARB(vs
->gl_shaders
[i
].prgId
));
3820 checkGLcall("glDeleteObjectARB");
3823 HeapFree(GetProcessHeap(), 0, vs
->gl_shaders
);
3824 vs
->gl_shaders
= NULL
;
3825 vs
->num_gl_shaders
= 0;
3826 vs
->shader_array_size
= 0;
3830 static unsigned int glsl_program_key_hash(const void *key
)
3832 const glsl_program_key_t
*k
= key
;
3834 unsigned int hash
= ((DWORD_PTR
) k
->vshader
) | ((DWORD_PTR
) k
->pshader
) << 16;
3835 hash
+= ~(hash
<< 15);
3836 hash
^= (hash
>> 10);
3837 hash
+= (hash
<< 3);
3838 hash
^= (hash
>> 6);
3839 hash
+= ~(hash
<< 11);
3840 hash
^= (hash
>> 16);
3845 static BOOL
glsl_program_key_compare(const void *keya
, const void *keyb
)
3847 const glsl_program_key_t
*ka
= keya
;
3848 const glsl_program_key_t
*kb
= keyb
;
3850 return ka
->vshader
== kb
->vshader
&& ka
->pshader
== kb
->pshader
&&
3851 (memcmp(&ka
->ps_args
, &kb
->ps_args
, sizeof(kb
->ps_args
)) == 0) &&
3852 (memcmp(&ka
->vs_args
, &kb
->vs_args
, sizeof(kb
->vs_args
)) == 0);
3855 static BOOL
constant_heap_init(struct constant_heap
*heap
, unsigned int constant_count
)
3857 SIZE_T size
= (constant_count
+ 1) * sizeof(*heap
->entries
) + constant_count
* sizeof(*heap
->positions
);
3858 void *mem
= HeapAlloc(GetProcessHeap(), 0, size
);
3862 ERR("Failed to allocate memory\n");
3866 heap
->entries
= mem
;
3867 heap
->entries
[1].version
= 0;
3868 heap
->positions
= (unsigned int *)(heap
->entries
+ constant_count
+ 1);
3874 static void constant_heap_free(struct constant_heap
*heap
)
3876 HeapFree(GetProcessHeap(), 0, heap
->entries
);
3879 static HRESULT
shader_glsl_alloc(IWineD3DDevice
*iface
) {
3880 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
3881 const WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
3882 struct shader_glsl_priv
*priv
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(struct shader_glsl_priv
));
3883 SIZE_T stack_size
= wined3d_log2i(max(GL_LIMITS(vshader_constantsF
), GL_LIMITS(pshader_constantsF
))) + 1;
3885 priv
->stack
= HeapAlloc(GetProcessHeap(), 0, stack_size
* sizeof(*priv
->stack
));
3888 ERR("Failed to allocate memory.\n");
3889 HeapFree(GetProcessHeap(), 0, priv
);
3890 return E_OUTOFMEMORY
;
3893 if (!constant_heap_init(&priv
->vconst_heap
, GL_LIMITS(vshader_constantsF
)))
3895 ERR("Failed to initialize vertex shader constant heap\n");
3896 HeapFree(GetProcessHeap(), 0, priv
->stack
);
3897 HeapFree(GetProcessHeap(), 0, priv
);
3898 return E_OUTOFMEMORY
;
3901 if (!constant_heap_init(&priv
->pconst_heap
, GL_LIMITS(pshader_constantsF
)))
3903 ERR("Failed to initialize pixel shader constant heap\n");
3904 constant_heap_free(&priv
->vconst_heap
);
3905 HeapFree(GetProcessHeap(), 0, priv
->stack
);
3906 HeapFree(GetProcessHeap(), 0, priv
);
3907 return E_OUTOFMEMORY
;
3910 priv
->glsl_program_lookup
= hash_table_create(glsl_program_key_hash
, glsl_program_key_compare
);
3911 priv
->next_constant_version
= 1;
3913 This
->shader_priv
= priv
;
3917 static void shader_glsl_free(IWineD3DDevice
*iface
) {
3918 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
3919 const WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
3920 struct shader_glsl_priv
*priv
= This
->shader_priv
;
3923 for (i
= 0; i
< tex_type_count
; ++i
)
3925 if (priv
->depth_blt_program
[i
])
3927 GL_EXTCALL(glDeleteObjectARB(priv
->depth_blt_program
[i
]));
3931 hash_table_destroy(priv
->glsl_program_lookup
, NULL
, NULL
);
3932 constant_heap_free(&priv
->pconst_heap
);
3933 constant_heap_free(&priv
->vconst_heap
);
3935 HeapFree(GetProcessHeap(), 0, This
->shader_priv
);
3936 This
->shader_priv
= NULL
;
3939 static BOOL
shader_glsl_dirty_const(IWineD3DDevice
*iface
) {
3940 /* TODO: GL_EXT_bindable_uniform can be used to share constants across shaders */
3944 static GLuint
shader_glsl_generate_pshader(IWineD3DPixelShader
*iface
, SHADER_BUFFER
*buffer
, const struct ps_compile_args
*args
) {
3945 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*)iface
;
3946 const struct shader_reg_maps
*reg_maps
= &This
->baseShader
.reg_maps
;
3947 CONST DWORD
*function
= This
->baseShader
.function
;
3948 const char *fragcolor
;
3949 const WineD3D_GL_Info
*gl_info
= &((IWineD3DDeviceImpl
*)This
->baseShader
.device
)->adapter
->gl_info
;
3951 /* Create the hw GLSL shader object and assign it as the shader->prgId */
3952 GLhandleARB shader_obj
= GL_EXTCALL(glCreateShaderObjectARB(GL_FRAGMENT_SHADER_ARB
));
3954 shader_addline(buffer
, "#version 120\n");
3956 if (GL_SUPPORT(ARB_DRAW_BUFFERS
)) {
3957 shader_addline(buffer
, "#extension GL_ARB_draw_buffers : enable\n");
3959 if (GL_SUPPORT(ARB_TEXTURE_RECTANGLE
)) {
3960 /* The spec says that it doesn't have to be explicitly enabled, but the nvidia
3961 * drivers write a warning if we don't do so
3963 shader_addline(buffer
, "#extension GL_ARB_texture_rectangle : enable\n");
3966 /* Base Declarations */
3967 shader_generate_glsl_declarations( (IWineD3DBaseShader
*) This
, reg_maps
, buffer
, &GLINFO_LOCATION
, args
);
3969 /* Pack 3.0 inputs */
3970 if (reg_maps
->shader_version
>= WINED3DPS_VERSION(3,0) && args
->vp_mode
!= vertexshader
) {
3971 pshader_glsl_input_pack(iface
, buffer
, This
->semantics_in
, reg_maps
, args
->vp_mode
);
3974 /* Base Shader Body */
3975 shader_generate_main( (IWineD3DBaseShader
*) This
, buffer
, reg_maps
, function
);
3977 /* Pixel shaders < 2.0 place the resulting color in R0 implicitly */
3978 if (reg_maps
->shader_version
< WINED3DPS_VERSION(2,0))
3980 /* Some older cards like GeforceFX ones don't support multiple buffers, so also not gl_FragData */
3981 if(GL_SUPPORT(ARB_DRAW_BUFFERS
))
3982 shader_addline(buffer
, "gl_FragData[0] = R0;\n");
3984 shader_addline(buffer
, "gl_FragColor = R0;\n");
3987 if(GL_SUPPORT(ARB_DRAW_BUFFERS
)) {
3988 fragcolor
= "gl_FragData[0]";
3990 fragcolor
= "gl_FragColor";
3992 if(args
->srgb_correction
) {
3993 shader_addline(buffer
, "tmp0.xyz = pow(%s.xyz, vec3(%f, %f, %f)) * vec3(%f, %f, %f) - vec3(%f, %f, %f);\n",
3994 fragcolor
, srgb_pow
, srgb_pow
, srgb_pow
, srgb_mul_high
, srgb_mul_high
, srgb_mul_high
,
3995 srgb_sub_high
, srgb_sub_high
, srgb_sub_high
);
3996 shader_addline(buffer
, "tmp1.xyz = %s.xyz * srgb_mul_low.xyz;\n", fragcolor
);
3997 shader_addline(buffer
, "%s.x = %s.x < srgb_comparison.x ? tmp1.x : tmp0.x;\n", fragcolor
, fragcolor
);
3998 shader_addline(buffer
, "%s.y = %s.y < srgb_comparison.y ? tmp1.y : tmp0.y;\n", fragcolor
, fragcolor
);
3999 shader_addline(buffer
, "%s.z = %s.z < srgb_comparison.z ? tmp1.z : tmp0.z;\n", fragcolor
, fragcolor
);
4000 shader_addline(buffer
, "%s = clamp(%s, 0.0, 1.0);\n", fragcolor
, fragcolor
);
4002 /* Pixel shader < 3.0 do not replace the fog stage.
4003 * This implements linear fog computation and blending.
4004 * TODO: non linear fog
4005 * NOTE: gl_Fog.start and gl_Fog.end don't hold fog start s and end e but
4006 * -1/(e-s) and e/(e-s) respectively.
4008 if(reg_maps
->shader_version
< WINED3DPS_VERSION(3,0)) {
4010 case FOG_OFF
: break;
4012 shader_addline(buffer
, "float fogstart = -1.0 / (gl_Fog.end - gl_Fog.start);\n");
4013 shader_addline(buffer
, "float fogend = gl_Fog.end * -fogstart;\n");
4014 shader_addline(buffer
, "float Fog = clamp(gl_FogFragCoord * fogstart + fogend, 0.0, 1.0);\n");
4015 shader_addline(buffer
, "%s.xyz = mix(gl_Fog.color.xyz, %s.xyz, Fog);\n", fragcolor
, fragcolor
);
4018 /* Fog = e^(-gl_Fog.density * gl_FogFragCoord) */
4019 shader_addline(buffer
, "float Fog = exp(-gl_Fog.density * gl_FogFragCoord);\n");
4020 shader_addline(buffer
, "Fog = clamp(Fog, 0.0, 1.0);\n");
4021 shader_addline(buffer
, "%s.xyz = mix(gl_Fog.color.xyz, %s.xyz, Fog);\n", fragcolor
, fragcolor
);
4024 /* Fog = e^(-(gl_Fog.density * gl_FogFragCoord)^2) */
4025 shader_addline(buffer
, "float Fog = exp(-gl_Fog.density * gl_Fog.density * gl_FogFragCoord * gl_FogFragCoord);\n");
4026 shader_addline(buffer
, "Fog = clamp(Fog, 0.0, 1.0);\n");
4027 shader_addline(buffer
, "%s.xyz = mix(gl_Fog.color.xyz, %s.xyz, Fog);\n", fragcolor
, fragcolor
);
4032 shader_addline(buffer
, "}\n");
4034 TRACE("Compiling shader object %u\n", shader_obj
);
4035 GL_EXTCALL(glShaderSourceARB(shader_obj
, 1, (const char**)&buffer
->buffer
, NULL
));
4036 GL_EXTCALL(glCompileShaderARB(shader_obj
));
4037 print_glsl_info_log(&GLINFO_LOCATION
, shader_obj
);
4039 /* Store the shader object */
4043 static GLuint
shader_glsl_generate_vshader(IWineD3DVertexShader
*iface
, SHADER_BUFFER
*buffer
, const struct vs_compile_args
*args
) {
4044 IWineD3DVertexShaderImpl
*This
= (IWineD3DVertexShaderImpl
*)iface
;
4045 const struct shader_reg_maps
*reg_maps
= &This
->baseShader
.reg_maps
;
4046 CONST DWORD
*function
= This
->baseShader
.function
;
4047 const WineD3D_GL_Info
*gl_info
= &((IWineD3DDeviceImpl
*)This
->baseShader
.device
)->adapter
->gl_info
;
4049 /* Create the hw GLSL shader program and assign it as the shader->prgId */
4050 GLhandleARB shader_obj
= GL_EXTCALL(glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB
));
4052 shader_addline(buffer
, "#version 120\n");
4054 /* Base Declarations */
4055 shader_generate_glsl_declarations( (IWineD3DBaseShader
*) This
, reg_maps
, buffer
, &GLINFO_LOCATION
, NULL
);
4057 /* Base Shader Body */
4058 shader_generate_main( (IWineD3DBaseShader
*) This
, buffer
, reg_maps
, function
);
4060 /* Unpack 3.0 outputs */
4061 if (reg_maps
->shader_version
>= WINED3DVS_VERSION(3,0)) shader_addline(buffer
, "order_ps_input(OUT);\n");
4062 else shader_addline(buffer
, "order_ps_input();\n");
4064 /* The D3DRS_FOGTABLEMODE render state defines if the shader-generated fog coord is used
4065 * or if the fragment depth is used. If the fragment depth is used(FOGTABLEMODE != NONE),
4066 * the fog frag coord is thrown away. If the fog frag coord is used, but not written by
4067 * the shader, it is set to 0.0(fully fogged, since start = 1.0, end = 0.0)
4069 if(args
->fog_src
== VS_FOG_Z
) {
4070 shader_addline(buffer
, "gl_FogFragCoord = gl_Position.z;\n");
4071 } else if (!reg_maps
->fog
) {
4072 shader_addline(buffer
, "gl_FogFragCoord = 0.0;\n");
4075 /* Write the final position.
4077 * OpenGL coordinates specify the center of the pixel while d3d coords specify
4078 * the corner. The offsets are stored in z and w in posFixup. posFixup.y contains
4079 * 1.0 or -1.0 to turn the rendering upside down for offscreen rendering. PosFixup.x
4080 * contains 1.0 to allow a mad.
4082 shader_addline(buffer
, "gl_Position.y = gl_Position.y * posFixup.y;\n");
4083 shader_addline(buffer
, "gl_Position.xy += posFixup.zw * gl_Position.ww;\n");
4085 /* Z coord [0;1]->[-1;1] mapping, see comment in transform_projection in state.c
4087 * Basically we want (in homogeneous coordinates) z = z * 2 - 1. However, shaders are run
4088 * before the homogeneous divide, so we have to take the w into account: z = ((z / w) * 2 - 1) * w,
4089 * which is the same as z = z * 2 - w.
4091 shader_addline(buffer
, "gl_Position.z = gl_Position.z * 2.0 - gl_Position.w;\n");
4093 shader_addline(buffer
, "}\n");
4095 TRACE("Compiling shader object %u\n", shader_obj
);
4096 GL_EXTCALL(glShaderSourceARB(shader_obj
, 1, (const char**)&buffer
->buffer
, NULL
));
4097 GL_EXTCALL(glCompileShaderARB(shader_obj
));
4098 print_glsl_info_log(&GLINFO_LOCATION
, shader_obj
);
4103 static void shader_glsl_get_caps(WINED3DDEVTYPE devtype
, const WineD3D_GL_Info
*gl_info
, struct shader_caps
*pCaps
)
4105 /* Nvidia Geforce6/7 or Ati R4xx/R5xx cards with GLSL support, support VS 3.0 but older Nvidia/Ati
4106 * models with GLSL support only support 2.0. In case of nvidia we can detect VS 2.0 support using
4107 * vs_nv_version which is based on NV_vertex_program.
4108 * For Ati cards there's no way using glsl (it abstracts the lowlevel info away) and also not
4109 * using ARB_vertex_program. It is safe to assume that when a card supports pixel shader 2.0 it
4110 * supports vertex shader 2.0 too and the way around. We can detect ps2.0 using the maximum number
4111 * of native instructions, so use that here. For more info see the pixel shader versioning code below.
4113 if((GLINFO_LOCATION
.vs_nv_version
== VS_VERSION_20
) || (GLINFO_LOCATION
.ps_arb_max_instructions
<= 512))
4114 pCaps
->VertexShaderVersion
= WINED3DVS_VERSION(2,0);
4116 pCaps
->VertexShaderVersion
= WINED3DVS_VERSION(3,0);
4117 TRACE_(d3d_caps
)("Hardware vertex shader version %d.%d enabled (GLSL)\n", (pCaps
->VertexShaderVersion
>> 8) & 0xff, pCaps
->VertexShaderVersion
& 0xff);
4118 /* Subtract the other potential uniforms from the max available (bools, ints, and 1 row of projection matrix) */
4119 pCaps
->MaxVertexShaderConst
= GL_LIMITS(vshader_constantsF
) - (MAX_CONST_B
/ 4) - MAX_CONST_I
- 1;
4121 /* Older DX9-class videocards (GeforceFX / Radeon >9500/X*00) only support pixel shader 2.0/2.0a/2.0b.
4122 * In OpenGL the extensions related to GLSL abstract lowlevel GL info away which is needed
4123 * to distinguish between 2.0 and 3.0 (and 2.0a/2.0b). In case of Nvidia we use their fragment
4124 * program extensions. On other hardware including ATI GL_ARB_fragment_program offers the info
4125 * in max native instructions. Intel and others also offer the info in this extension but they
4126 * don't support GLSL (at least on Windows).
4128 * PS2.0 requires at least 96 instructions, 2.0a/2.0b go up to 512. Assume that if the number
4129 * of instructions is 512 or less we have to do with ps2.0 hardware.
4130 * 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.
4132 if((GLINFO_LOCATION
.ps_nv_version
== PS_VERSION_20
) || (GLINFO_LOCATION
.ps_arb_max_instructions
<= 512))
4133 pCaps
->PixelShaderVersion
= WINED3DPS_VERSION(2,0);
4135 pCaps
->PixelShaderVersion
= WINED3DPS_VERSION(3,0);
4137 /* Subtract the other potential uniforms from the max available (bools & ints), and 2 states for fog.
4138 * In theory the texbem instruction may need one more shader constant too. But lets assume
4139 * that a sm <= 1.3 shader does not need all the uniforms provided by a glsl-capable card,
4140 * and lets not take away a uniform needlessly from all other shaders.
4142 pCaps
->MaxPixelShaderConst
= GL_LIMITS(pshader_constantsF
) - (MAX_CONST_B
/ 4) - MAX_CONST_I
- 2;
4144 /* FIXME: The following line is card dependent. -8.0 to 8.0 is the
4145 * Direct3D minimum requirement.
4147 * Both GL_ARB_fragment_program and GLSL require a "maximum representable magnitude"
4148 * of colors to be 2^10, and 2^32 for other floats. Should we use 1024 here?
4150 * The problem is that the refrast clamps temporary results in the shader to
4151 * [-MaxValue;+MaxValue]. If the card's max value is bigger than the one we advertize here,
4152 * then applications may miss the clamping behavior. On the other hand, if it is smaller,
4153 * the shader will generate incorrect results too. Unfortunately, GL deliberately doesn't
4154 * offer a way to query this.
4156 pCaps
->PixelShader1xMaxValue
= 8.0;
4157 TRACE_(d3d_caps
)("Hardware pixel shader version %d.%d enabled (GLSL)\n", (pCaps
->PixelShaderVersion
>> 8) & 0xff, pCaps
->PixelShaderVersion
& 0xff);
4160 static BOOL
shader_glsl_color_fixup_supported(struct color_fixup_desc fixup
)
4162 if (TRACE_ON(d3d_shader
) && TRACE_ON(d3d
))
4164 TRACE("Checking support for fixup:\n");
4165 dump_color_fixup_desc(fixup
);
4168 /* We support everything except YUV conversions. */
4169 if (!is_yuv_fixup(fixup
))
4175 TRACE("[FAILED]\n");
4179 static const SHADER_HANDLER shader_glsl_instruction_handler_table
[WINED3DSIH_TABLE_SIZE
] =
4181 /* WINED3DSIH_ABS */ shader_glsl_map2gl
,
4182 /* WINED3DSIH_ADD */ shader_glsl_arith
,
4183 /* WINED3DSIH_BEM */ pshader_glsl_bem
,
4184 /* WINED3DSIH_BREAK */ shader_glsl_break
,
4185 /* WINED3DSIH_BREAKC */ shader_glsl_breakc
,
4186 /* WINED3DSIH_BREAKP */ NULL
,
4187 /* WINED3DSIH_CALL */ shader_glsl_call
,
4188 /* WINED3DSIH_CALLNZ */ shader_glsl_callnz
,
4189 /* WINED3DSIH_CMP */ shader_glsl_cmp
,
4190 /* WINED3DSIH_CND */ shader_glsl_cnd
,
4191 /* WINED3DSIH_CRS */ shader_glsl_cross
,
4192 /* WINED3DSIH_DCL */ NULL
,
4193 /* WINED3DSIH_DEF */ NULL
,
4194 /* WINED3DSIH_DEFB */ NULL
,
4195 /* WINED3DSIH_DEFI */ NULL
,
4196 /* WINED3DSIH_DP2ADD */ pshader_glsl_dp2add
,
4197 /* WINED3DSIH_DP3 */ shader_glsl_dot
,
4198 /* WINED3DSIH_DP4 */ shader_glsl_dot
,
4199 /* WINED3DSIH_DST */ shader_glsl_dst
,
4200 /* WINED3DSIH_DSX */ shader_glsl_map2gl
,
4201 /* WINED3DSIH_DSY */ shader_glsl_map2gl
,
4202 /* WINED3DSIH_ELSE */ shader_glsl_else
,
4203 /* WINED3DSIH_ENDIF */ shader_glsl_end
,
4204 /* WINED3DSIH_ENDLOOP */ shader_glsl_end
,
4205 /* WINED3DSIH_ENDREP */ shader_glsl_end
,
4206 /* WINED3DSIH_EXP */ shader_glsl_map2gl
,
4207 /* WINED3DSIH_EXPP */ shader_glsl_expp
,
4208 /* WINED3DSIH_FRC */ shader_glsl_map2gl
,
4209 /* WINED3DSIH_IF */ shader_glsl_if
,
4210 /* WINED3DSIH_IFC */ shader_glsl_ifc
,
4211 /* WINED3DSIH_LABEL */ shader_glsl_label
,
4212 /* WINED3DSIH_LIT */ shader_glsl_lit
,
4213 /* WINED3DSIH_LOG */ shader_glsl_log
,
4214 /* WINED3DSIH_LOGP */ shader_glsl_log
,
4215 /* WINED3DSIH_LOOP */ shader_glsl_loop
,
4216 /* WINED3DSIH_LRP */ shader_glsl_lrp
,
4217 /* WINED3DSIH_M3x2 */ shader_glsl_mnxn
,
4218 /* WINED3DSIH_M3x3 */ shader_glsl_mnxn
,
4219 /* WINED3DSIH_M3x4 */ shader_glsl_mnxn
,
4220 /* WINED3DSIH_M4x3 */ shader_glsl_mnxn
,
4221 /* WINED3DSIH_M4x4 */ shader_glsl_mnxn
,
4222 /* WINED3DSIH_MAD */ shader_glsl_mad
,
4223 /* WINED3DSIH_MAX */ shader_glsl_map2gl
,
4224 /* WINED3DSIH_MIN */ shader_glsl_map2gl
,
4225 /* WINED3DSIH_MOV */ shader_glsl_mov
,
4226 /* WINED3DSIH_MOVA */ shader_glsl_mov
,
4227 /* WINED3DSIH_MUL */ shader_glsl_arith
,
4228 /* WINED3DSIH_NOP */ NULL
,
4229 /* WINED3DSIH_NRM */ shader_glsl_map2gl
,
4230 /* WINED3DSIH_PHASE */ NULL
,
4231 /* WINED3DSIH_POW */ shader_glsl_pow
,
4232 /* WINED3DSIH_RCP */ shader_glsl_rcp
,
4233 /* WINED3DSIH_REP */ shader_glsl_rep
,
4234 /* WINED3DSIH_RET */ NULL
,
4235 /* WINED3DSIH_RSQ */ shader_glsl_rsq
,
4236 /* WINED3DSIH_SETP */ NULL
,
4237 /* WINED3DSIH_SGE */ shader_glsl_compare
,
4238 /* WINED3DSIH_SGN */ shader_glsl_map2gl
,
4239 /* WINED3DSIH_SINCOS */ shader_glsl_sincos
,
4240 /* WINED3DSIH_SLT */ shader_glsl_compare
,
4241 /* WINED3DSIH_SUB */ shader_glsl_arith
,
4242 /* WINED3DSIH_TEX */ pshader_glsl_tex
,
4243 /* WINED3DSIH_TEXBEM */ pshader_glsl_texbem
,
4244 /* WINED3DSIH_TEXBEML */ pshader_glsl_texbem
,
4245 /* WINED3DSIH_TEXCOORD */ pshader_glsl_texcoord
,
4246 /* WINED3DSIH_TEXDEPTH */ pshader_glsl_texdepth
,
4247 /* WINED3DSIH_TEXDP3 */ pshader_glsl_texdp3
,
4248 /* WINED3DSIH_TEXDP3TEX */ pshader_glsl_texdp3tex
,
4249 /* WINED3DSIH_TEXKILL */ pshader_glsl_texkill
,
4250 /* WINED3DSIH_TEXLDD */ NULL
,
4251 /* WINED3DSIH_TEXLDL */ shader_glsl_texldl
,
4252 /* WINED3DSIH_TEXM3x2DEPTH */ pshader_glsl_texm3x2depth
,
4253 /* WINED3DSIH_TEXM3x2PAD */ pshader_glsl_texm3x2pad
,
4254 /* WINED3DSIH_TEXM3x2TEX */ pshader_glsl_texm3x2tex
,
4255 /* WINED3DSIH_TEXM3x3 */ pshader_glsl_texm3x3
,
4256 /* WINED3DSIH_TEXM3x3DIFF */ NULL
,
4257 /* WINED3DSIH_TEXM3x3PAD */ pshader_glsl_texm3x3pad
,
4258 /* WINED3DSIH_TEXM3x3SPEC */ pshader_glsl_texm3x3spec
,
4259 /* WINED3DSIH_TEXM3x3TEX */ pshader_glsl_texm3x3tex
,
4260 /* WINED3DSIH_TEXM3x3VSPEC */ pshader_glsl_texm3x3vspec
,
4261 /* WINED3DSIH_TEXREG2AR */ pshader_glsl_texreg2ar
,
4262 /* WINED3DSIH_TEXREG2GB */ pshader_glsl_texreg2gb
,
4263 /* WINED3DSIH_TEXREG2RGB */ pshader_glsl_texreg2rgb
,
4266 const shader_backend_t glsl_shader_backend
= {
4267 shader_glsl_instruction_handler_table
,
4269 shader_glsl_select_depth_blt
,
4270 shader_glsl_deselect_depth_blt
,
4271 shader_glsl_update_float_vertex_constants
,
4272 shader_glsl_update_float_pixel_constants
,
4273 shader_glsl_load_constants
,
4274 shader_glsl_load_np2fixup_constants
,
4275 shader_glsl_destroy
,
4278 shader_glsl_dirty_const
,
4279 shader_glsl_generate_pshader
,
4280 shader_glsl_generate_vshader
,
4281 shader_glsl_get_caps
,
4282 shader_glsl_color_fixup_supported
,