2 * GLSL pixel and vertex shader implementation
4 * Copyright 2006 Jason Green
5 * Copyright 2006-2007 Henri Verbeet
6 * Copyright 2007-2008 Stefan Dösinger for CodeWeavers
8 * This library is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU Lesser General Public
10 * License as published by the Free Software Foundation; either
11 * version 2.1 of the License, or (at your option) any later version.
13 * This library is distributed in the hope that it will be useful,
14 * but WITHOUT ANY WARRANTY; without even the implied warranty of
15 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
16 * Lesser General Public License for more details.
18 * You should have received a copy of the GNU Lesser General Public
19 * License along with this library; if not, write to the Free Software
20 * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301, USA
24 * D3D shader asm has swizzles on source parameters, and write masks for
25 * destination parameters. GLSL uses swizzles for both. The result of this is
26 * that for example "mov dst.xw, src.zyxw" becomes "dst.xw = src.zw" in GLSL.
27 * Ie, to generate a proper GLSL source swizzle, we need to take the D3D write
28 * mask for the destination parameter into account.
34 #include "wined3d_private.h"
36 WINE_DEFAULT_DEBUG_CHANNEL(d3d_shader
);
37 WINE_DECLARE_DEBUG_CHANNEL(d3d_constants
);
38 WINE_DECLARE_DEBUG_CHANNEL(d3d_caps
);
39 WINE_DECLARE_DEBUG_CHANNEL(d3d
);
41 #define GLINFO_LOCATION (*gl_info)
43 #define WINED3D_GLSL_SAMPLE_PROJECTED 0x1
44 #define WINED3D_GLSL_SAMPLE_RECT 0x2
45 #define WINED3D_GLSL_SAMPLE_LOD 0x4
60 } glsl_sample_function_t
;
64 HEAP_NODE_TRAVERSE_LEFT
,
65 HEAP_NODE_TRAVERSE_RIGHT
,
77 struct constant_entry
*entries
;
78 unsigned int *positions
;
82 /* GLSL shader private data */
83 struct shader_glsl_priv
{
84 struct hash_table_t
*glsl_program_lookup
;
85 struct glsl_shader_prog_link
*glsl_program
;
86 struct constant_heap vconst_heap
;
87 struct constant_heap pconst_heap
;
89 GLhandleARB depth_blt_program
[tex_type_count
];
90 UINT next_constant_version
;
93 /* Struct to maintain data about a linked GLSL program */
94 struct glsl_shader_prog_link
{
95 struct list vshader_entry
;
96 struct list pshader_entry
;
97 GLhandleARB programId
;
98 GLhandleARB
*vuniformF_locations
;
99 GLhandleARB
*puniformF_locations
;
100 GLhandleARB vuniformI_locations
[MAX_CONST_I
];
101 GLhandleARB puniformI_locations
[MAX_CONST_I
];
102 GLhandleARB posFixup_location
;
103 GLhandleARB bumpenvmat_location
[MAX_TEXTURES
];
104 GLhandleARB luminancescale_location
[MAX_TEXTURES
];
105 GLhandleARB luminanceoffset_location
[MAX_TEXTURES
];
106 GLhandleARB ycorrection_location
;
107 GLenum vertex_color_clamp
;
108 IWineD3DVertexShader
*vshader
;
109 IWineD3DPixelShader
*pshader
;
110 struct vs_compile_args vs_args
;
111 struct ps_compile_args ps_args
;
112 UINT constant_version
;
116 IWineD3DVertexShader
*vshader
;
117 IWineD3DPixelShader
*pshader
;
118 struct ps_compile_args ps_args
;
119 struct vs_compile_args vs_args
;
120 } glsl_program_key_t
;
123 /** Prints the GLSL info log which will contain error messages if they exist */
124 static void print_glsl_info_log(const WineD3D_GL_Info
*gl_info
, GLhandleARB obj
)
126 int infologLength
= 0;
131 static const char * const spam
[] =
133 "Vertex shader was successfully compiled to run on hardware.\n", /* fglrx */
134 "Fragment shader was successfully compiled to run on hardware.\n", /* fglrx */
135 "Fragment shader(s) linked, vertex shader(s) linked. \n ", /* fglrx, with \n */
136 "Fragment shader(s) linked, vertex shader(s) linked.", /* fglrx, no \n */
137 "Vertex shader(s) linked, no fragment shader(s) defined. \n ", /* fglrx, with \n */
138 "Vertex shader(s) linked, no fragment shader(s) defined.", /* fglrx, no \n */
139 "Fragment shader was successfully compiled to run on hardware.\n"
140 "WARNING: 0:2: extension 'GL_ARB_draw_buffers' is not supported",
141 "Fragment shader(s) linked, no vertex shader(s) defined.", /* fglrx, no \n */
142 "Fragment shader(s) linked, no vertex shader(s) defined. \n ", /* fglrx, with \n */
143 "WARNING: 0:2: extension 'GL_ARB_draw_buffers' is not supported\n" /* MacOS ati */
146 if (!TRACE_ON(d3d_shader
) && !FIXME_ON(d3d_shader
)) return;
148 GL_EXTCALL(glGetObjectParameterivARB(obj
,
149 GL_OBJECT_INFO_LOG_LENGTH_ARB
,
152 /* A size of 1 is just a null-terminated string, so the log should be bigger than
153 * that if there are errors. */
154 if (infologLength
> 1)
156 /* Fglrx doesn't terminate the string properly, but it tells us the proper length.
157 * So use HEAP_ZERO_MEMORY to avoid uninitialized bytes
159 infoLog
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, infologLength
);
160 GL_EXTCALL(glGetInfoLogARB(obj
, infologLength
, NULL
, infoLog
));
163 for(i
= 0; i
< sizeof(spam
) / sizeof(spam
[0]); i
++) {
164 if(strcmp(infoLog
, spam
[i
]) == 0) {
170 TRACE("Spam received from GLSL shader #%u: %s\n", obj
, debugstr_a(infoLog
));
172 FIXME("Error received from GLSL shader #%u: %s\n", obj
, debugstr_a(infoLog
));
174 HeapFree(GetProcessHeap(), 0, infoLog
);
179 * Loads (pixel shader) samplers
181 static void shader_glsl_load_psamplers(const WineD3D_GL_Info
*gl_info
, IWineD3DStateBlock
*iface
, GLhandleARB programId
)
183 IWineD3DStateBlockImpl
* stateBlock
= (IWineD3DStateBlockImpl
*) iface
;
184 GLhandleARB name_loc
;
186 char sampler_name
[20];
188 for (i
= 0; i
< MAX_FRAGMENT_SAMPLERS
; ++i
) {
189 snprintf(sampler_name
, sizeof(sampler_name
), "Psampler%d", i
);
190 name_loc
= GL_EXTCALL(glGetUniformLocationARB(programId
, sampler_name
));
191 if (name_loc
!= -1) {
192 DWORD mapped_unit
= stateBlock
->wineD3DDevice
->texUnitMap
[i
];
193 if (mapped_unit
!= WINED3D_UNMAPPED_STAGE
&& mapped_unit
< GL_LIMITS(fragment_samplers
))
195 TRACE("Loading %s for texture %d\n", sampler_name
, mapped_unit
);
196 GL_EXTCALL(glUniform1iARB(name_loc
, mapped_unit
));
197 checkGLcall("glUniform1iARB");
199 ERR("Trying to load sampler %s on unsupported unit %d\n", sampler_name
, mapped_unit
);
205 static void shader_glsl_load_vsamplers(const WineD3D_GL_Info
*gl_info
, IWineD3DStateBlock
*iface
, GLhandleARB programId
)
207 IWineD3DStateBlockImpl
* stateBlock
= (IWineD3DStateBlockImpl
*) iface
;
208 GLhandleARB name_loc
;
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
= stateBlock
->wineD3DDevice
->texUnitMap
[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 GLhandleARB
*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 GLhandleARB
*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 GLhandleARB
*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 GLhandleARB 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 GLhandleARB 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 app-supplied constants into the currently set GLSL program.
489 static void shader_glsl_load_constants(
490 IWineD3DDevice
* device
,
492 char useVertexShader
) {
494 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) device
;
495 struct shader_glsl_priv
*priv
= deviceImpl
->shader_priv
;
496 IWineD3DStateBlockImpl
* stateBlock
= deviceImpl
->stateBlock
;
497 const WineD3D_GL_Info
*gl_info
= &deviceImpl
->adapter
->gl_info
;
499 GLhandleARB programId
;
500 struct glsl_shader_prog_link
*prog
= priv
->glsl_program
;
501 UINT constant_version
;
505 /* No GLSL program set - nothing to do. */
508 programId
= prog
->programId
;
509 constant_version
= prog
->constant_version
;
511 if (useVertexShader
) {
512 IWineD3DBaseShaderImpl
* vshader
= (IWineD3DBaseShaderImpl
*) stateBlock
->vertexShader
;
514 /* Load DirectX 9 float constants/uniforms for vertex shader */
515 shader_glsl_load_constantsF(vshader
, gl_info
, stateBlock
->vertexShaderConstantF
,
516 prog
->vuniformF_locations
, &priv
->vconst_heap
, priv
->stack
, constant_version
);
518 /* Load DirectX 9 integer constants/uniforms for vertex shader */
519 if(vshader
->baseShader
.uses_int_consts
) {
520 shader_glsl_load_constantsI(vshader
, gl_info
, prog
->vuniformI_locations
,
521 stateBlock
->vertexShaderConstantI
, stateBlock
->changed
.vertexShaderConstantsI
);
524 /* Load DirectX 9 boolean constants/uniforms for vertex shader */
525 if(vshader
->baseShader
.uses_bool_consts
) {
526 shader_glsl_load_constantsB(vshader
, gl_info
, programId
,
527 stateBlock
->vertexShaderConstantB
, stateBlock
->changed
.vertexShaderConstantsB
);
530 /* Upload the position fixup params */
531 GL_EXTCALL(glUniform4fvARB(prog
->posFixup_location
, 1, &deviceImpl
->posFixup
[0]));
532 checkGLcall("glUniform4fvARB");
535 if (usePixelShader
) {
537 IWineD3DBaseShaderImpl
* pshader
= (IWineD3DBaseShaderImpl
*) stateBlock
->pixelShader
;
539 /* Load DirectX 9 float constants/uniforms for pixel shader */
540 shader_glsl_load_constantsF(pshader
, gl_info
, stateBlock
->pixelShaderConstantF
,
541 prog
->puniformF_locations
, &priv
->pconst_heap
, priv
->stack
, constant_version
);
543 /* Load DirectX 9 integer constants/uniforms for pixel shader */
544 if(pshader
->baseShader
.uses_int_consts
) {
545 shader_glsl_load_constantsI(pshader
, gl_info
, prog
->puniformI_locations
,
546 stateBlock
->pixelShaderConstantI
, stateBlock
->changed
.pixelShaderConstantsI
);
549 /* Load DirectX 9 boolean constants/uniforms for pixel shader */
550 if(pshader
->baseShader
.uses_bool_consts
) {
551 shader_glsl_load_constantsB(pshader
, gl_info
, programId
,
552 stateBlock
->pixelShaderConstantB
, stateBlock
->changed
.pixelShaderConstantsB
);
555 /* Upload the environment bump map matrix if needed. The needsbumpmat member specifies the texture stage to load the matrix from.
556 * It can't be 0 for a valid texbem instruction.
558 for(i
= 0; i
< ((IWineD3DPixelShaderImpl
*) pshader
)->numbumpenvmatconsts
; i
++) {
559 IWineD3DPixelShaderImpl
*ps
= (IWineD3DPixelShaderImpl
*) pshader
;
560 int stage
= ps
->luminanceconst
[i
].texunit
;
562 const float *data
= (const float *)&stateBlock
->textureState
[(int)ps
->bumpenvmatconst
[i
].texunit
][WINED3DTSS_BUMPENVMAT00
];
563 GL_EXTCALL(glUniformMatrix2fvARB(prog
->bumpenvmat_location
[i
], 1, 0, data
));
564 checkGLcall("glUniformMatrix2fvARB");
566 /* texbeml needs the luminance scale and offset too. If texbeml is used, needsbumpmat
567 * is set too, so we can check that in the needsbumpmat check
569 if(ps
->baseShader
.reg_maps
.luminanceparams
[stage
]) {
570 const GLfloat
*scale
= (const GLfloat
*)&stateBlock
->textureState
[stage
][WINED3DTSS_BUMPENVLSCALE
];
571 const GLfloat
*offset
= (const GLfloat
*)&stateBlock
->textureState
[stage
][WINED3DTSS_BUMPENVLOFFSET
];
573 GL_EXTCALL(glUniform1fvARB(prog
->luminancescale_location
[i
], 1, scale
));
574 checkGLcall("glUniform1fvARB");
575 GL_EXTCALL(glUniform1fvARB(prog
->luminanceoffset_location
[i
], 1, offset
));
576 checkGLcall("glUniform1fvARB");
580 if(((IWineD3DPixelShaderImpl
*) pshader
)->vpos_uniform
) {
581 float correction_params
[4];
582 if(deviceImpl
->render_offscreen
) {
583 correction_params
[0] = 0.0;
584 correction_params
[1] = 1.0;
586 /* position is window relative, not viewport relative */
587 correction_params
[0] = ((IWineD3DSurfaceImpl
*) deviceImpl
->render_targets
[0])->currentDesc
.Height
;
588 correction_params
[1] = -1.0;
590 GL_EXTCALL(glUniform4fvARB(prog
->ycorrection_location
, 1, correction_params
));
594 if (priv
->next_constant_version
== UINT_MAX
)
596 TRACE("Max constant version reached, resetting to 0.\n");
597 hash_table_for_each_entry(priv
->glsl_program_lookup
, reset_program_constant_version
, NULL
);
598 priv
->next_constant_version
= 1;
602 prog
->constant_version
= priv
->next_constant_version
++;
606 static inline void update_heap_entry(struct constant_heap
*heap
, unsigned int idx
,
607 unsigned int heap_idx
, DWORD new_version
)
609 struct constant_entry
*entries
= heap
->entries
;
610 unsigned int *positions
= heap
->positions
;
611 unsigned int parent_idx
;
615 parent_idx
= heap_idx
>> 1;
617 if (new_version
<= entries
[parent_idx
].version
) break;
619 entries
[heap_idx
] = entries
[parent_idx
];
620 positions
[entries
[parent_idx
].idx
] = heap_idx
;
621 heap_idx
= parent_idx
;
624 entries
[heap_idx
].version
= new_version
;
625 entries
[heap_idx
].idx
= idx
;
626 positions
[idx
] = heap_idx
;
629 static void shader_glsl_update_float_vertex_constants(IWineD3DDevice
*iface
, UINT start
, UINT count
)
631 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
632 struct shader_glsl_priv
*priv
= This
->shader_priv
;
633 struct constant_heap
*heap
= &priv
->vconst_heap
;
636 for (i
= start
; i
< count
+ start
; ++i
)
638 if (!This
->stateBlock
->changed
.vertexShaderConstantsF
[i
])
639 update_heap_entry(heap
, i
, heap
->size
++, priv
->next_constant_version
);
641 update_heap_entry(heap
, i
, heap
->positions
[i
], priv
->next_constant_version
);
645 static void shader_glsl_update_float_pixel_constants(IWineD3DDevice
*iface
, UINT start
, UINT count
)
647 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
648 struct shader_glsl_priv
*priv
= This
->shader_priv
;
649 struct constant_heap
*heap
= &priv
->pconst_heap
;
652 for (i
= start
; i
< count
+ start
; ++i
)
654 if (!This
->stateBlock
->changed
.pixelShaderConstantsF
[i
])
655 update_heap_entry(heap
, i
, heap
->size
++, priv
->next_constant_version
);
657 update_heap_entry(heap
, i
, heap
->positions
[i
], priv
->next_constant_version
);
661 /** Generate the variable & register declarations for the GLSL output target */
662 static void shader_generate_glsl_declarations(IWineD3DBaseShader
*iface
, const shader_reg_maps
*reg_maps
,
663 SHADER_BUFFER
*buffer
, const WineD3D_GL_Info
*gl_info
,
664 const struct ps_compile_args
*ps_args
)
666 IWineD3DBaseShaderImpl
* This
= (IWineD3DBaseShaderImpl
*) iface
;
667 IWineD3DDeviceImpl
*device
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
668 DWORD shader_version
= reg_maps
->shader_version
;
669 unsigned int i
, extra_constants_needed
= 0;
670 const local_constant
*lconst
;
672 /* There are some minor differences between pixel and vertex shaders */
673 char pshader
= shader_is_pshader_version(shader_version
);
674 char prefix
= pshader
? 'P' : 'V';
676 /* Prototype the subroutines */
677 for (i
= 0; i
< This
->baseShader
.limits
.label
; i
++) {
678 if (reg_maps
->labels
[i
])
679 shader_addline(buffer
, "void subroutine%u();\n", i
);
682 /* Declare the constants (aka uniforms) */
683 if (This
->baseShader
.limits
.constant_float
> 0) {
684 unsigned max_constantsF
= min(This
->baseShader
.limits
.constant_float
,
685 (pshader
? GL_LIMITS(pshader_constantsF
) : GL_LIMITS(vshader_constantsF
)));
686 shader_addline(buffer
, "uniform vec4 %cC[%u];\n", prefix
, max_constantsF
);
689 if (This
->baseShader
.limits
.constant_int
> 0)
690 shader_addline(buffer
, "uniform ivec4 %cI[%u];\n", prefix
, This
->baseShader
.limits
.constant_int
);
692 if (This
->baseShader
.limits
.constant_bool
> 0)
693 shader_addline(buffer
, "uniform bool %cB[%u];\n", prefix
, This
->baseShader
.limits
.constant_bool
);
696 shader_addline(buffer
, "uniform vec4 posFixup;\n");
697 /* Predeclaration; This function is added at link time based on the pixel shader.
698 * VS 3.0 shaders have an array OUT[] the shader writes to, earlier versions don't have
699 * that. We know the input to the reorder function at vertex shader compile time, so
700 * we can deal with that. The reorder function for a 1.x and 2.x vertex shader can just
701 * read gl_FrontColor. The output depends on the pixel shader. The reorder function for a
702 * 1.x and 2.x pshader or for fixed function will write gl_FrontColor, and for a 3.0 shader
703 * it will write to the varying array. Here we depend on the shader optimizer on sorting that
704 * out. The nvidia driver only does that if the parameter is inout instead of out, hence the
707 if (shader_version
>= WINED3DVS_VERSION(3, 0))
709 shader_addline(buffer
, "void order_ps_input(in vec4[%u]);\n", MAX_REG_OUTPUT
);
711 shader_addline(buffer
, "void order_ps_input();\n");
714 IWineD3DPixelShaderImpl
*ps_impl
= (IWineD3DPixelShaderImpl
*) This
;
716 ps_impl
->numbumpenvmatconsts
= 0;
717 for(i
= 0; i
< (sizeof(reg_maps
->bumpmat
) / sizeof(reg_maps
->bumpmat
[0])); i
++) {
718 if(!reg_maps
->bumpmat
[i
]) {
722 ps_impl
->bumpenvmatconst
[(int) ps_impl
->numbumpenvmatconsts
].texunit
= i
;
723 shader_addline(buffer
, "uniform mat2 bumpenvmat%d;\n", i
);
725 if(reg_maps
->luminanceparams
) {
726 ps_impl
->luminanceconst
[(int) ps_impl
->numbumpenvmatconsts
].texunit
= i
;
727 shader_addline(buffer
, "uniform float luminancescale%d;\n", i
);
728 shader_addline(buffer
, "uniform float luminanceoffset%d;\n", i
);
729 extra_constants_needed
++;
731 ps_impl
->luminanceconst
[(int) ps_impl
->numbumpenvmatconsts
].texunit
= -1;
734 extra_constants_needed
++;
735 ps_impl
->numbumpenvmatconsts
++;
738 if(ps_args
->srgb_correction
) {
739 shader_addline(buffer
, "const vec4 srgb_mul_low = vec4(%f, %f, %f, %f);\n",
740 srgb_mul_low
, srgb_mul_low
, srgb_mul_low
, srgb_mul_low
);
741 shader_addline(buffer
, "const vec4 srgb_comparison = vec4(%f, %f, %f, %f);\n",
742 srgb_cmp
, srgb_cmp
, srgb_cmp
, srgb_cmp
);
744 if(reg_maps
->vpos
|| reg_maps
->usesdsy
) {
745 if(This
->baseShader
.limits
.constant_float
+ extra_constants_needed
+ 1 < GL_LIMITS(pshader_constantsF
)) {
746 shader_addline(buffer
, "uniform vec4 ycorrection;\n");
747 ((IWineD3DPixelShaderImpl
*) This
)->vpos_uniform
= 1;
748 extra_constants_needed
++;
750 /* This happens because we do not have proper tracking of the constant registers that are
751 * actually used, only the max limit of the shader version
753 FIXME("Cannot find a free uniform for vpos correction params\n");
754 shader_addline(buffer
, "const vec4 ycorrection = vec4(%f, %f, 0.0, 0.0);\n",
755 device
->render_offscreen
? 0.0 : ((IWineD3DSurfaceImpl
*) device
->render_targets
[0])->currentDesc
.Height
,
756 device
->render_offscreen
? 1.0 : -1.0);
758 shader_addline(buffer
, "vec4 vpos;\n");
762 /* Declare texture samplers */
763 for (i
= 0; i
< This
->baseShader
.limits
.sampler
; i
++) {
764 if (reg_maps
->samplers
[i
]) {
766 DWORD stype
= reg_maps
->samplers
[i
] & WINED3DSP_TEXTURETYPE_MASK
;
770 shader_addline(buffer
, "uniform sampler1D %csampler%u;\n", prefix
, i
);
773 if(device
->stateBlock
->textures
[i
] &&
774 IWineD3DBaseTexture_GetTextureDimensions(device
->stateBlock
->textures
[i
]) == GL_TEXTURE_RECTANGLE_ARB
) {
775 shader_addline(buffer
, "uniform sampler2DRect %csampler%u;\n", prefix
, i
);
777 shader_addline(buffer
, "uniform sampler2D %csampler%u;\n", prefix
, i
);
780 case WINED3DSTT_CUBE
:
781 shader_addline(buffer
, "uniform samplerCube %csampler%u;\n", prefix
, i
);
783 case WINED3DSTT_VOLUME
:
784 shader_addline(buffer
, "uniform sampler3D %csampler%u;\n", prefix
, i
);
787 shader_addline(buffer
, "uniform unsupported_sampler %csampler%u;\n", prefix
, i
);
788 FIXME("Unrecognized sampler type: %#x\n", stype
);
794 /* Declare address variables */
795 for (i
= 0; i
< This
->baseShader
.limits
.address
; i
++) {
796 if (reg_maps
->address
[i
])
797 shader_addline(buffer
, "ivec4 A%d;\n", i
);
800 /* Declare texture coordinate temporaries and initialize them */
801 for (i
= 0; i
< This
->baseShader
.limits
.texcoord
; i
++) {
802 if (reg_maps
->texcoord
[i
])
803 shader_addline(buffer
, "vec4 T%u = gl_TexCoord[%u];\n", i
, i
);
806 /* Declare input register varyings. Only pixel shader, vertex shaders have that declared in the
807 * helper function shader that is linked in at link time
809 if (pshader
&& shader_version
>= WINED3DPS_VERSION(3, 0))
811 if (use_vs(device
->stateBlock
))
813 shader_addline(buffer
, "varying vec4 IN[%u];\n", GL_LIMITS(glsl_varyings
) / 4);
815 /* TODO: Write a replacement shader for the fixed function vertex pipeline, so this isn't needed.
816 * For fixed function vertex processing + 3.0 pixel shader we need a separate function in the
817 * pixel shader that reads the fixed function color into the packed input registers.
819 shader_addline(buffer
, "vec4 IN[%u];\n", GL_LIMITS(glsl_varyings
) / 4);
823 /* Declare output register temporaries */
824 if(This
->baseShader
.limits
.packed_output
) {
825 shader_addline(buffer
, "vec4 OUT[%u];\n", This
->baseShader
.limits
.packed_output
);
828 /* Declare temporary variables */
829 for(i
= 0; i
< This
->baseShader
.limits
.temporary
; i
++) {
830 if (reg_maps
->temporary
[i
])
831 shader_addline(buffer
, "vec4 R%u;\n", i
);
834 /* Declare attributes */
835 for (i
= 0; i
< This
->baseShader
.limits
.attributes
; i
++) {
836 if (reg_maps
->attributes
[i
])
837 shader_addline(buffer
, "attribute vec4 attrib%i;\n", i
);
840 /* Declare loop registers aLx */
841 for (i
= 0; i
< reg_maps
->loop_depth
; i
++) {
842 shader_addline(buffer
, "int aL%u;\n", i
);
843 shader_addline(buffer
, "int tmpInt%u;\n", i
);
846 /* Temporary variables for matrix operations */
847 shader_addline(buffer
, "vec4 tmp0;\n");
848 shader_addline(buffer
, "vec4 tmp1;\n");
850 /* Local constants use a different name so they can be loaded once at shader link time
851 * They can't be hardcoded into the shader text via LC = {x, y, z, w}; because the
852 * float -> string conversion can cause precision loss.
854 if(!This
->baseShader
.load_local_constsF
) {
855 LIST_FOR_EACH_ENTRY(lconst
, &This
->baseShader
.constantsF
, local_constant
, entry
) {
856 shader_addline(buffer
, "uniform vec4 %cLC%u;\n", prefix
, lconst
->idx
);
860 /* Start the main program */
861 shader_addline(buffer
, "void main() {\n");
862 if(pshader
&& reg_maps
->vpos
) {
863 /* DirectX apps expect integer values, while OpenGL drivers add approximately 0.5. This causes
864 * off-by-one problems as spotted by the vPos d3d9 visual test. Unfortunately the ATI cards do
865 * not add exactly 0.5, but rather something like 0.49999999 or 0.50000001, which still causes
866 * precision troubles when we just substract 0.5.
868 * To deal with that just floor() the position. This will eliminate the fraction on all cards.
870 * TODO: Test how that behaves with multisampling once we can enable multisampling in winex11.
872 * An advantage of floor is that it works even if the driver doesn't add 1/2. It is somewhat
873 * questionable if 1.5, 2.5, ... are the proper values to return in gl_FragCoord, even though
874 * coordinates specify the pixel centers instead of the pixel corners. This code will behave
875 * correctly on drivers that returns integer values.
877 shader_addline(buffer
, "vpos = floor(vec4(0, ycorrection[0], 0, 0) + gl_FragCoord * vec4(1, ycorrection[1], 1, 1));\n");
881 /*****************************************************************************
882 * Functions to generate GLSL strings from DirectX Shader bytecode begin here.
884 * For more information, see http://wiki.winehq.org/DirectX-Shaders
885 ****************************************************************************/
888 static void shader_glsl_add_src_param(const SHADER_OPCODE_ARG
*arg
, const DWORD param
,
889 const DWORD addr_token
, DWORD mask
, glsl_src_param_t
*src_param
);
891 /** Used for opcode modifiers - They multiply the result by the specified amount */
892 static const char * const shift_glsl_tab
[] = {
894 "2.0 * ", /* 1 (x2) */
895 "4.0 * ", /* 2 (x4) */
896 "8.0 * ", /* 3 (x8) */
897 "16.0 * ", /* 4 (x16) */
898 "32.0 * ", /* 5 (x32) */
905 "0.0625 * ", /* 12 (d16) */
906 "0.125 * ", /* 13 (d8) */
907 "0.25 * ", /* 14 (d4) */
908 "0.5 * " /* 15 (d2) */
911 /* Generate a GLSL parameter that does the input modifier computation and return the input register/mask to use */
912 static void shader_glsl_gen_modifier (
915 const char *in_regswizzle
,
920 if (instr
== WINED3DSIO_TEXKILL
)
923 switch (instr
& WINED3DSP_SRCMOD_MASK
) {
924 case WINED3DSPSM_DZ
: /* Need to handle this in the instructions itself (texld & texcrd). */
926 case WINED3DSPSM_NONE
:
927 sprintf(out_str
, "%s%s", in_reg
, in_regswizzle
);
929 case WINED3DSPSM_NEG
:
930 sprintf(out_str
, "-%s%s", in_reg
, in_regswizzle
);
932 case WINED3DSPSM_NOT
:
933 sprintf(out_str
, "!%s%s", in_reg
, in_regswizzle
);
935 case WINED3DSPSM_BIAS
:
936 sprintf(out_str
, "(%s%s - vec4(0.5)%s)", in_reg
, in_regswizzle
, in_regswizzle
);
938 case WINED3DSPSM_BIASNEG
:
939 sprintf(out_str
, "-(%s%s - vec4(0.5)%s)", in_reg
, in_regswizzle
, in_regswizzle
);
941 case WINED3DSPSM_SIGN
:
942 sprintf(out_str
, "(2.0 * (%s%s - 0.5))", in_reg
, in_regswizzle
);
944 case WINED3DSPSM_SIGNNEG
:
945 sprintf(out_str
, "-(2.0 * (%s%s - 0.5))", in_reg
, in_regswizzle
);
947 case WINED3DSPSM_COMP
:
948 sprintf(out_str
, "(1.0 - %s%s)", in_reg
, in_regswizzle
);
951 sprintf(out_str
, "(2.0 * %s%s)", in_reg
, in_regswizzle
);
953 case WINED3DSPSM_X2NEG
:
954 sprintf(out_str
, "-(2.0 * %s%s)", in_reg
, in_regswizzle
);
956 case WINED3DSPSM_ABS
:
957 sprintf(out_str
, "abs(%s%s)", in_reg
, in_regswizzle
);
959 case WINED3DSPSM_ABSNEG
:
960 sprintf(out_str
, "-abs(%s%s)", in_reg
, in_regswizzle
);
963 FIXME("Unhandled modifier %u\n", (instr
& WINED3DSP_SRCMOD_MASK
));
964 sprintf(out_str
, "%s%s", in_reg
, in_regswizzle
);
968 /** Writes the GLSL variable name that corresponds to the register that the
969 * DX opcode parameter is trying to access */
970 static void shader_glsl_get_register_name(const DWORD param
, const DWORD addr_token
,
971 char *regstr
, BOOL
*is_color
, const SHADER_OPCODE_ARG
*arg
)
973 /* oPos, oFog and oPts in D3D */
974 static const char * const hwrastout_reg_names
[] = { "gl_Position", "gl_FogFragCoord", "gl_PointSize" };
976 DWORD reg
= param
& WINED3DSP_REGNUM_MASK
;
977 DWORD regtype
= shader_get_regtype(param
);
978 IWineD3DBaseShaderImpl
* This
= (IWineD3DBaseShaderImpl
*) arg
->shader
;
979 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
980 const WineD3D_GL_Info
* gl_info
= &deviceImpl
->adapter
->gl_info
;
981 DWORD shader_version
= This
->baseShader
.reg_maps
.shader_version
;
982 char pshader
= shader_is_pshader_version(shader_version
);
988 case WINED3DSPR_TEMP
:
989 sprintf(tmpStr
, "R%u", reg
);
991 case WINED3DSPR_INPUT
:
993 /* Pixel shaders >= 3.0 */
994 if (WINED3DSHADER_VERSION_MAJOR(shader_version
) >= 3)
996 DWORD in_count
= GL_LIMITS(glsl_varyings
) / 4;
998 if (param
& WINED3DSHADER_ADDRMODE_RELATIVE
) {
999 glsl_src_param_t rel_param
;
1000 shader_glsl_add_src_param(arg
, addr_token
, 0, WINED3DSP_WRITEMASK_0
, &rel_param
);
1002 /* Removing a + 0 would be an obvious optimization, but macos doesn't see the NOP
1005 if(((IWineD3DPixelShaderImpl
*) This
)->input_reg_map
[reg
]) {
1006 if (((IWineD3DPixelShaderImpl
*)This
)->declared_in_count
> in_count
) {
1007 sprintf(tmpStr
, "((%s + %u) > %d ? (%s + %u) > %d ? gl_SecondaryColor : gl_Color : IN[%s + %u])",
1008 rel_param
.param_str
, ((IWineD3DPixelShaderImpl
*)This
)->input_reg_map
[reg
], in_count
- 1,
1009 rel_param
.param_str
, ((IWineD3DPixelShaderImpl
*)This
)->input_reg_map
[reg
], in_count
,
1010 rel_param
.param_str
, ((IWineD3DPixelShaderImpl
*)This
)->input_reg_map
[reg
]);
1012 sprintf(tmpStr
, "IN[%s + %u]", rel_param
.param_str
, ((IWineD3DPixelShaderImpl
*)This
)->input_reg_map
[reg
]);
1015 if (((IWineD3DPixelShaderImpl
*)This
)->declared_in_count
> in_count
) {
1016 sprintf(tmpStr
, "((%s) > %d ? (%s) > %d ? gl_SecondaryColor : gl_Color : IN[%s])",
1017 rel_param
.param_str
, in_count
- 1,
1018 rel_param
.param_str
, in_count
,
1019 rel_param
.param_str
);
1021 sprintf(tmpStr
, "IN[%s]", rel_param
.param_str
);
1025 DWORD idx
= ((IWineD3DPixelShaderImpl
*) This
)->input_reg_map
[reg
];
1026 if (idx
== in_count
) {
1027 sprintf(tmpStr
, "gl_Color");
1028 } else if (idx
== in_count
+ 1) {
1029 sprintf(tmpStr
, "gl_SecondaryColor");
1031 sprintf(tmpStr
, "IN[%u]", idx
);
1036 strcpy(tmpStr
, "gl_Color");
1038 strcpy(tmpStr
, "gl_SecondaryColor");
1041 if (((IWineD3DVertexShaderImpl
*)This
)->cur_args
->swizzle_map
& (1 << reg
)) *is_color
= TRUE
;
1042 sprintf(tmpStr
, "attrib%u", reg
);
1045 case WINED3DSPR_CONST
:
1047 const char prefix
= pshader
? 'P':'V';
1049 /* Relative addressing */
1050 if (param
& WINED3DSHADER_ADDRMODE_RELATIVE
) {
1052 /* Relative addressing on shaders 2.0+ have a relative address token,
1053 * prior to that, it was hard-coded as "A0.x" because there's only 1 register */
1054 if (WINED3DSHADER_VERSION_MAJOR(shader_version
) >= 2)
1056 glsl_src_param_t rel_param
;
1057 shader_glsl_add_src_param(arg
, addr_token
, 0, WINED3DSP_WRITEMASK_0
, &rel_param
);
1059 sprintf(tmpStr
, "%cC[%s + %u]", prefix
, rel_param
.param_str
, reg
);
1061 sprintf(tmpStr
, "%cC[%s]", prefix
, rel_param
.param_str
);
1065 sprintf(tmpStr
, "%cC[A0.x + %u]", prefix
, reg
);
1067 sprintf(tmpStr
, "%cC[A0.x]", prefix
);
1072 if(shader_constant_is_local(This
, reg
)) {
1073 sprintf(tmpStr
, "%cLC%u", prefix
, reg
);
1075 sprintf(tmpStr
, "%cC[%u]", prefix
, reg
);
1081 case WINED3DSPR_CONSTINT
:
1083 sprintf(tmpStr
, "PI[%u]", reg
);
1085 sprintf(tmpStr
, "VI[%u]", reg
);
1087 case WINED3DSPR_CONSTBOOL
:
1089 sprintf(tmpStr
, "PB[%u]", reg
);
1091 sprintf(tmpStr
, "VB[%u]", reg
);
1093 case WINED3DSPR_TEXTURE
: /* case WINED3DSPR_ADDR: */
1095 sprintf(tmpStr
, "T%u", reg
);
1097 sprintf(tmpStr
, "A%u", reg
);
1100 case WINED3DSPR_LOOP
:
1101 sprintf(tmpStr
, "aL%u", This
->baseShader
.cur_loop_regno
- 1);
1103 case WINED3DSPR_SAMPLER
:
1105 sprintf(tmpStr
, "Psampler%u", reg
);
1107 sprintf(tmpStr
, "Vsampler%u", reg
);
1109 case WINED3DSPR_COLOROUT
:
1110 if (reg
>= GL_LIMITS(buffers
)) {
1111 WARN("Write to render target %u, only %d supported\n", reg
, 4);
1113 if (GL_SUPPORT(ARB_DRAW_BUFFERS
)) {
1114 sprintf(tmpStr
, "gl_FragData[%u]", reg
);
1115 } else { /* On older cards with GLSL support like the GeforceFX there's only one buffer. */
1116 sprintf(tmpStr
, "gl_FragColor");
1119 case WINED3DSPR_RASTOUT
:
1120 sprintf(tmpStr
, "%s", hwrastout_reg_names
[reg
]);
1122 case WINED3DSPR_DEPTHOUT
:
1123 sprintf(tmpStr
, "gl_FragDepth");
1125 case WINED3DSPR_ATTROUT
:
1127 sprintf(tmpStr
, "gl_FrontColor");
1129 sprintf(tmpStr
, "gl_FrontSecondaryColor");
1132 case WINED3DSPR_TEXCRDOUT
:
1133 /* Vertex shaders >= 3.0: WINED3DSPR_OUTPUT */
1134 if (WINED3DSHADER_VERSION_MAJOR(shader_version
) >= 3) sprintf(tmpStr
, "OUT[%u]", reg
);
1135 else sprintf(tmpStr
, "gl_TexCoord[%u]", reg
);
1137 case WINED3DSPR_MISCTYPE
:
1140 sprintf(tmpStr
, "vpos");
1141 } else if (reg
== 1){
1142 /* Note that gl_FrontFacing is a bool, while vFace is
1143 * a float for which the sign determines front/back
1145 sprintf(tmpStr
, "(gl_FrontFacing ? 1.0 : -1.0)");
1147 FIXME("Unhandled misctype register %d\n", reg
);
1148 sprintf(tmpStr
, "unrecognized_register");
1152 FIXME("Unhandled register name Type(%d)\n", regtype
);
1153 sprintf(tmpStr
, "unrecognized_register");
1157 strcat(regstr
, tmpStr
);
1160 /* Get the GLSL write mask for the destination register */
1161 static DWORD
shader_glsl_get_write_mask(const DWORD param
, char *write_mask
) {
1162 char *ptr
= write_mask
;
1163 DWORD mask
= param
& WINED3DSP_WRITEMASK_ALL
;
1165 if (shader_is_scalar(param
)) {
1166 mask
= WINED3DSP_WRITEMASK_0
;
1169 if (param
& WINED3DSP_WRITEMASK_0
) *ptr
++ = 'x';
1170 if (param
& WINED3DSP_WRITEMASK_1
) *ptr
++ = 'y';
1171 if (param
& WINED3DSP_WRITEMASK_2
) *ptr
++ = 'z';
1172 if (param
& WINED3DSP_WRITEMASK_3
) *ptr
++ = 'w';
1180 static unsigned int shader_glsl_get_write_mask_size(DWORD write_mask
) {
1181 unsigned int size
= 0;
1183 if (write_mask
& WINED3DSP_WRITEMASK_0
) ++size
;
1184 if (write_mask
& WINED3DSP_WRITEMASK_1
) ++size
;
1185 if (write_mask
& WINED3DSP_WRITEMASK_2
) ++size
;
1186 if (write_mask
& WINED3DSP_WRITEMASK_3
) ++size
;
1191 static void shader_glsl_get_swizzle(const DWORD param
, BOOL fixup
, DWORD mask
, char *swizzle_str
) {
1192 /* For registers of type WINED3DDECLTYPE_D3DCOLOR, data is stored as "bgra",
1193 * but addressed as "rgba". To fix this we need to swap the register's x
1194 * and z components. */
1195 DWORD swizzle
= (param
& WINED3DSP_SWIZZLE_MASK
) >> WINED3DSP_SWIZZLE_SHIFT
;
1196 const char *swizzle_chars
= fixup
? "zyxw" : "xyzw";
1197 char *ptr
= swizzle_str
;
1199 if (!shader_is_scalar(param
)) {
1201 /* swizzle bits fields: wwzzyyxx */
1202 if (mask
& WINED3DSP_WRITEMASK_0
) *ptr
++ = swizzle_chars
[swizzle
& 0x03];
1203 if (mask
& WINED3DSP_WRITEMASK_1
) *ptr
++ = swizzle_chars
[(swizzle
>> 2) & 0x03];
1204 if (mask
& WINED3DSP_WRITEMASK_2
) *ptr
++ = swizzle_chars
[(swizzle
>> 4) & 0x03];
1205 if (mask
& WINED3DSP_WRITEMASK_3
) *ptr
++ = swizzle_chars
[(swizzle
>> 6) & 0x03];
1211 /* From a given parameter token, generate the corresponding GLSL string.
1212 * Also, return the actual register name and swizzle in case the
1213 * caller needs this information as well. */
1214 static void shader_glsl_add_src_param(const SHADER_OPCODE_ARG
*arg
, const DWORD param
,
1215 const DWORD addr_token
, DWORD mask
, glsl_src_param_t
*src_param
)
1217 BOOL is_color
= FALSE
;
1218 char swizzle_str
[6];
1220 src_param
->reg_name
[0] = '\0';
1221 src_param
->param_str
[0] = '\0';
1222 swizzle_str
[0] = '\0';
1224 shader_glsl_get_register_name(param
, addr_token
, src_param
->reg_name
, &is_color
, arg
);
1226 shader_glsl_get_swizzle(param
, is_color
, mask
, swizzle_str
);
1227 shader_glsl_gen_modifier(param
, src_param
->reg_name
, swizzle_str
, src_param
->param_str
);
1230 /* From a given parameter token, generate the corresponding GLSL string.
1231 * Also, return the actual register name and swizzle in case the
1232 * caller needs this information as well. */
1233 static DWORD
shader_glsl_add_dst_param(const SHADER_OPCODE_ARG
* arg
, const DWORD param
,
1234 const DWORD addr_token
, glsl_dst_param_t
*dst_param
)
1236 BOOL is_color
= FALSE
;
1238 dst_param
->mask_str
[0] = '\0';
1239 dst_param
->reg_name
[0] = '\0';
1241 shader_glsl_get_register_name(param
, addr_token
, dst_param
->reg_name
, &is_color
, arg
);
1242 return shader_glsl_get_write_mask(param
, dst_param
->mask_str
);
1245 /* Append the destination part of the instruction to the buffer, return the effective write mask */
1246 static DWORD
shader_glsl_append_dst_ext(SHADER_BUFFER
*buffer
, const SHADER_OPCODE_ARG
*arg
, const DWORD param
)
1248 glsl_dst_param_t dst_param
;
1252 mask
= shader_glsl_add_dst_param(arg
, param
, arg
->dst_addr
, &dst_param
);
1255 shift
= (param
& WINED3DSP_DSTSHIFT_MASK
) >> WINED3DSP_DSTSHIFT_SHIFT
;
1256 shader_addline(buffer
, "%s%s = %s(", dst_param
.reg_name
, dst_param
.mask_str
, shift_glsl_tab
[shift
]);
1262 /* Append the destination part of the instruction to the buffer, return the effective write mask */
1263 static DWORD
shader_glsl_append_dst(SHADER_BUFFER
*buffer
, const SHADER_OPCODE_ARG
*arg
)
1265 return shader_glsl_append_dst_ext(buffer
, arg
, arg
->dst
);
1268 /** Process GLSL instruction modifiers */
1269 void shader_glsl_add_instruction_modifiers(const SHADER_OPCODE_ARG
* arg
)
1271 DWORD mask
= arg
->dst
& WINED3DSP_DSTMOD_MASK
;
1273 if (arg
->opcode
->dst_token
&& mask
!= 0) {
1274 glsl_dst_param_t dst_param
;
1276 shader_glsl_add_dst_param(arg
, arg
->dst
, 0, &dst_param
);
1278 if (mask
& WINED3DSPDM_SATURATE
) {
1279 /* _SAT means to clamp the value of the register to between 0 and 1 */
1280 shader_addline(arg
->buffer
, "%s%s = clamp(%s%s, 0.0, 1.0);\n", dst_param
.reg_name
,
1281 dst_param
.mask_str
, dst_param
.reg_name
, dst_param
.mask_str
);
1283 if (mask
& WINED3DSPDM_MSAMPCENTROID
) {
1284 FIXME("_centroid modifier not handled\n");
1286 if (mask
& WINED3DSPDM_PARTIALPRECISION
) {
1287 /* MSDN says this modifier can be safely ignored, so that's what we'll do. */
1292 static inline const char* shader_get_comp_op(
1293 const DWORD opcode
) {
1295 DWORD op
= (opcode
& INST_CONTROLS_MASK
) >> INST_CONTROLS_SHIFT
;
1297 case COMPARISON_GT
: return ">";
1298 case COMPARISON_EQ
: return "==";
1299 case COMPARISON_GE
: return ">=";
1300 case COMPARISON_LT
: return "<";
1301 case COMPARISON_NE
: return "!=";
1302 case COMPARISON_LE
: return "<=";
1304 FIXME("Unrecognized comparison value: %u\n", op
);
1309 static void shader_glsl_get_sample_function(DWORD sampler_type
, DWORD flags
, glsl_sample_function_t
*sample_function
)
1311 BOOL projected
= flags
& WINED3D_GLSL_SAMPLE_PROJECTED
;
1312 BOOL texrect
= flags
& WINED3D_GLSL_SAMPLE_RECT
;
1313 BOOL lod
= flags
& WINED3D_GLSL_SAMPLE_LOD
;
1315 /* Note that there's no such thing as a projected cube texture. */
1316 switch(sampler_type
) {
1319 sample_function
->name
= projected
? "texture1DProjLod" : "texture1DLod";
1321 sample_function
->name
= projected
? "texture1DProj" : "texture1D";
1323 sample_function
->coord_mask
= WINED3DSP_WRITEMASK_0
;
1328 sample_function
->name
= projected
? "texture2DRectProjLod" : "texture2DRectLod";
1330 sample_function
->name
= projected
? "texture2DRectProj" : "texture2DRect";
1334 sample_function
->name
= projected
? "texture2DProjLod" : "texture2DLod";
1336 sample_function
->name
= projected
? "texture2DProj" : "texture2D";
1339 sample_function
->coord_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
;
1341 case WINED3DSTT_CUBE
:
1343 sample_function
->name
= "textureCubeLod";
1345 sample_function
->name
= "textureCube";
1347 sample_function
->coord_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
1349 case WINED3DSTT_VOLUME
:
1351 sample_function
->name
= projected
? "texture3DProjLod" : "texture3DLod";
1353 sample_function
->name
= projected
? "texture3DProj" : "texture3D";
1355 sample_function
->coord_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
1358 sample_function
->name
= "";
1359 sample_function
->coord_mask
= 0;
1360 FIXME("Unrecognized sampler type: %#x;\n", sampler_type
);
1365 static void shader_glsl_append_fixup_arg(char *arguments
, const char *reg_name
,
1366 BOOL sign_fixup
, enum fixup_channel_source channel_source
)
1368 switch(channel_source
)
1370 case CHANNEL_SOURCE_ZERO
:
1371 strcat(arguments
, "0.0");
1374 case CHANNEL_SOURCE_ONE
:
1375 strcat(arguments
, "1.0");
1378 case CHANNEL_SOURCE_X
:
1379 strcat(arguments
, reg_name
);
1380 strcat(arguments
, ".x");
1383 case CHANNEL_SOURCE_Y
:
1384 strcat(arguments
, reg_name
);
1385 strcat(arguments
, ".y");
1388 case CHANNEL_SOURCE_Z
:
1389 strcat(arguments
, reg_name
);
1390 strcat(arguments
, ".z");
1393 case CHANNEL_SOURCE_W
:
1394 strcat(arguments
, reg_name
);
1395 strcat(arguments
, ".w");
1399 FIXME("Unhandled channel source %#x\n", channel_source
);
1400 strcat(arguments
, "undefined");
1404 if (sign_fixup
) strcat(arguments
, " * 2.0 - 1.0");
1407 static void shader_glsl_color_correction(const struct SHADER_OPCODE_ARG
*arg
, struct color_fixup_desc fixup
)
1409 unsigned int mask_size
, remaining
;
1410 glsl_dst_param_t dst_param
;
1411 char arguments
[256];
1416 if (fixup
.x_sign_fixup
|| fixup
.x_source
!= CHANNEL_SOURCE_X
) mask
|= WINED3DSP_WRITEMASK_0
;
1417 if (fixup
.y_sign_fixup
|| fixup
.y_source
!= CHANNEL_SOURCE_Y
) mask
|= WINED3DSP_WRITEMASK_1
;
1418 if (fixup
.z_sign_fixup
|| fixup
.z_source
!= CHANNEL_SOURCE_Z
) mask
|= WINED3DSP_WRITEMASK_2
;
1419 if (fixup
.w_sign_fixup
|| fixup
.w_source
!= CHANNEL_SOURCE_W
) mask
|= WINED3DSP_WRITEMASK_3
;
1422 if (!mask
) return; /* Nothing to do */
1424 if (is_yuv_fixup(fixup
))
1426 enum yuv_fixup yuv_fixup
= get_yuv_fixup(fixup
);
1427 FIXME("YUV fixup (%#x) not supported\n", yuv_fixup
);
1431 mask_size
= shader_glsl_get_write_mask_size(mask
);
1433 dst_param
.mask_str
[0] = '\0';
1434 shader_glsl_get_write_mask(mask
, dst_param
.mask_str
);
1436 dst_param
.reg_name
[0] = '\0';
1437 shader_glsl_get_register_name(arg
->dst
, arg
->dst_addr
, dst_param
.reg_name
, &dummy
, arg
);
1439 arguments
[0] = '\0';
1440 remaining
= mask_size
;
1441 if (mask
& WINED3DSP_WRITEMASK_0
)
1443 shader_glsl_append_fixup_arg(arguments
, dst_param
.reg_name
, fixup
.x_sign_fixup
, fixup
.x_source
);
1444 if (--remaining
) strcat(arguments
, ", ");
1446 if (mask
& WINED3DSP_WRITEMASK_1
)
1448 shader_glsl_append_fixup_arg(arguments
, dst_param
.reg_name
, fixup
.y_sign_fixup
, fixup
.y_source
);
1449 if (--remaining
) strcat(arguments
, ", ");
1451 if (mask
& WINED3DSP_WRITEMASK_2
)
1453 shader_glsl_append_fixup_arg(arguments
, dst_param
.reg_name
, fixup
.z_sign_fixup
, fixup
.z_source
);
1454 if (--remaining
) strcat(arguments
, ", ");
1456 if (mask
& WINED3DSP_WRITEMASK_3
)
1458 shader_glsl_append_fixup_arg(arguments
, dst_param
.reg_name
, fixup
.w_sign_fixup
, fixup
.w_source
);
1459 if (--remaining
) strcat(arguments
, ", ");
1464 shader_addline(arg
->buffer
, "%s%s = vec%u(%s);\n",
1465 dst_param
.reg_name
, dst_param
.mask_str
, mask_size
, arguments
);
1469 shader_addline(arg
->buffer
, "%s%s = %s;\n", dst_param
.reg_name
, dst_param
.mask_str
, arguments
);
1473 static void shader_glsl_gen_sample_code(const SHADER_OPCODE_ARG
*arg
, DWORD sampler
, const char *coord_reg
,
1474 const glsl_sample_function_t
*sample_function
, DWORD swizzle
,
1476 const char *sampler_base
;
1477 char dst_swizzle
[6];
1478 struct color_fixup_desc fixup
;
1480 shader_glsl_get_swizzle(swizzle
, FALSE
, arg
->dst
, dst_swizzle
);
1482 if(shader_is_pshader_version(arg
->reg_maps
->shader_version
)) {
1483 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*) arg
->shader
;
1484 fixup
= This
->cur_args
->color_fixup
[sampler
];
1485 sampler_base
= "Psampler";
1487 sampler_base
= "Vsampler";
1488 fixup
= COLOR_FIXUP_IDENTITY
; /* FIXME: Vshader color fixup */
1491 shader_glsl_append_dst(arg
->buffer
, arg
);
1494 shader_addline(arg
->buffer
, "%s(%s%u, %s, %s)%s);\n",
1495 sample_function
->name
, sampler_base
, sampler
, coord_reg
, bias
, dst_swizzle
);
1497 shader_addline(arg
->buffer
, "%s(%s%u, %s)%s);\n",
1498 sample_function
->name
, sampler_base
, sampler
, coord_reg
, dst_swizzle
);
1501 if(!is_identity_fixup(fixup
)) {
1502 shader_glsl_color_correction(arg
, fixup
);
1506 /*****************************************************************************
1508 * Begin processing individual instruction opcodes
1510 ****************************************************************************/
1512 /* Generate GLSL arithmetic functions (dst = src1 + src2) */
1513 static void shader_glsl_arith(const SHADER_OPCODE_ARG
*arg
)
1515 CONST SHADER_OPCODE
* curOpcode
= arg
->opcode
;
1516 SHADER_BUFFER
* buffer
= arg
->buffer
;
1517 glsl_src_param_t src0_param
;
1518 glsl_src_param_t src1_param
;
1522 /* Determine the GLSL operator to use based on the opcode */
1523 switch (curOpcode
->opcode
) {
1524 case WINED3DSIO_MUL
: op
= '*'; break;
1525 case WINED3DSIO_ADD
: op
= '+'; break;
1526 case WINED3DSIO_SUB
: op
= '-'; break;
1529 FIXME("Opcode %s not yet handled in GLSL\n", curOpcode
->name
);
1533 write_mask
= shader_glsl_append_dst(buffer
, arg
);
1534 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], write_mask
, &src0_param
);
1535 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], write_mask
, &src1_param
);
1536 shader_addline(buffer
, "%s %c %s);\n", src0_param
.param_str
, op
, src1_param
.param_str
);
1539 /* Process the WINED3DSIO_MOV opcode using GLSL (dst = src) */
1540 static void shader_glsl_mov(const SHADER_OPCODE_ARG
*arg
)
1542 SHADER_BUFFER
* buffer
= arg
->buffer
;
1543 glsl_src_param_t src0_param
;
1546 write_mask
= shader_glsl_append_dst(buffer
, arg
);
1547 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], write_mask
, &src0_param
);
1549 /* In vs_1_1 WINED3DSIO_MOV can write to the address register. In later
1550 * shader versions WINED3DSIO_MOVA is used for this. */
1551 if ((WINED3DSHADER_VERSION_MAJOR(arg
->reg_maps
->shader_version
) == 1
1552 && !shader_is_pshader_version(arg
->reg_maps
->shader_version
)
1553 && shader_get_regtype(arg
->dst
) == WINED3DSPR_ADDR
))
1555 /* This is a simple floor() */
1556 unsigned int mask_size
= shader_glsl_get_write_mask_size(write_mask
);
1557 if (mask_size
> 1) {
1558 shader_addline(buffer
, "ivec%d(floor(%s)));\n", mask_size
, src0_param
.param_str
);
1560 shader_addline(buffer
, "int(floor(%s)));\n", src0_param
.param_str
);
1562 } else if(arg
->opcode
->opcode
== WINED3DSIO_MOVA
) {
1563 /* We need to *round* to the nearest int here. */
1564 unsigned int mask_size
= shader_glsl_get_write_mask_size(write_mask
);
1565 if (mask_size
> 1) {
1566 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
);
1568 shader_addline(buffer
, "int(floor(abs(%s) + 0.5) * sign(%s)));\n", src0_param
.param_str
, src0_param
.param_str
);
1571 shader_addline(buffer
, "%s);\n", src0_param
.param_str
);
1575 /* Process the dot product operators DP3 and DP4 in GLSL (dst = dot(src0, src1)) */
1576 static void shader_glsl_dot(const SHADER_OPCODE_ARG
*arg
)
1578 CONST SHADER_OPCODE
* curOpcode
= arg
->opcode
;
1579 SHADER_BUFFER
* buffer
= arg
->buffer
;
1580 glsl_src_param_t src0_param
;
1581 glsl_src_param_t src1_param
;
1582 DWORD dst_write_mask
, src_write_mask
;
1583 unsigned int dst_size
= 0;
1585 dst_write_mask
= shader_glsl_append_dst(buffer
, arg
);
1586 dst_size
= shader_glsl_get_write_mask_size(dst_write_mask
);
1588 /* dp3 works on vec3, dp4 on vec4 */
1589 if (curOpcode
->opcode
== WINED3DSIO_DP4
) {
1590 src_write_mask
= WINED3DSP_WRITEMASK_ALL
;
1592 src_write_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
1595 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], src_write_mask
, &src0_param
);
1596 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], src_write_mask
, &src1_param
);
1599 shader_addline(buffer
, "vec%d(dot(%s, %s)));\n", dst_size
, src0_param
.param_str
, src1_param
.param_str
);
1601 shader_addline(buffer
, "dot(%s, %s));\n", src0_param
.param_str
, src1_param
.param_str
);
1605 /* Note that this instruction has some restrictions. The destination write mask
1606 * can't contain the w component, and the source swizzles have to be .xyzw */
1607 static void shader_glsl_cross(const SHADER_OPCODE_ARG
*arg
)
1609 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
1610 glsl_src_param_t src0_param
;
1611 glsl_src_param_t src1_param
;
1614 shader_glsl_get_write_mask(arg
->dst
, dst_mask
);
1615 shader_glsl_append_dst(arg
->buffer
, arg
);
1616 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], src_mask
, &src0_param
);
1617 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], src_mask
, &src1_param
);
1618 shader_addline(arg
->buffer
, "cross(%s, %s)%s);\n", src0_param
.param_str
, src1_param
.param_str
, dst_mask
);
1621 /* Process the WINED3DSIO_POW instruction in GLSL (dst = |src0|^src1)
1622 * Src0 and src1 are scalars. Note that D3D uses the absolute of src0, while
1623 * GLSL uses the value as-is. */
1624 static void shader_glsl_pow(const SHADER_OPCODE_ARG
*arg
)
1626 SHADER_BUFFER
*buffer
= arg
->buffer
;
1627 glsl_src_param_t src0_param
;
1628 glsl_src_param_t src1_param
;
1629 DWORD dst_write_mask
;
1630 unsigned int dst_size
;
1632 dst_write_mask
= shader_glsl_append_dst(buffer
, arg
);
1633 dst_size
= shader_glsl_get_write_mask_size(dst_write_mask
);
1635 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
1636 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], WINED3DSP_WRITEMASK_0
, &src1_param
);
1639 shader_addline(buffer
, "vec%d(pow(abs(%s), %s)));\n", dst_size
, src0_param
.param_str
, src1_param
.param_str
);
1641 shader_addline(buffer
, "pow(abs(%s), %s));\n", src0_param
.param_str
, src1_param
.param_str
);
1645 /* Process the WINED3DSIO_LOG instruction in GLSL (dst = log2(|src0|))
1646 * Src0 is a scalar. Note that D3D uses the absolute of src0, while
1647 * GLSL uses the value as-is. */
1648 static void shader_glsl_log(const SHADER_OPCODE_ARG
*arg
)
1650 SHADER_BUFFER
*buffer
= arg
->buffer
;
1651 glsl_src_param_t src0_param
;
1652 DWORD dst_write_mask
;
1653 unsigned int dst_size
;
1655 dst_write_mask
= shader_glsl_append_dst(buffer
, arg
);
1656 dst_size
= shader_glsl_get_write_mask_size(dst_write_mask
);
1658 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
1661 shader_addline(buffer
, "vec%d(log2(abs(%s))));\n", dst_size
, src0_param
.param_str
);
1663 shader_addline(buffer
, "log2(abs(%s)));\n", src0_param
.param_str
);
1667 /* Map the opcode 1-to-1 to the GL code (arg->dst = instruction(src0, src1, ...) */
1668 static void shader_glsl_map2gl(const SHADER_OPCODE_ARG
*arg
)
1670 CONST SHADER_OPCODE
* curOpcode
= arg
->opcode
;
1671 SHADER_BUFFER
* buffer
= arg
->buffer
;
1672 glsl_src_param_t src_param
;
1673 const char *instruction
;
1674 char arguments
[256];
1678 /* Determine the GLSL function to use based on the opcode */
1679 /* TODO: Possibly make this a table for faster lookups */
1680 switch (curOpcode
->opcode
) {
1681 case WINED3DSIO_MIN
: instruction
= "min"; break;
1682 case WINED3DSIO_MAX
: instruction
= "max"; break;
1683 case WINED3DSIO_ABS
: instruction
= "abs"; break;
1684 case WINED3DSIO_FRC
: instruction
= "fract"; break;
1685 case WINED3DSIO_NRM
: instruction
= "normalize"; break;
1686 case WINED3DSIO_EXP
: instruction
= "exp2"; break;
1687 case WINED3DSIO_SGN
: instruction
= "sign"; break;
1688 case WINED3DSIO_DSX
: instruction
= "dFdx"; break;
1689 case WINED3DSIO_DSY
: instruction
= "ycorrection.y * dFdy"; break;
1690 default: instruction
= "";
1691 FIXME("Opcode %s not yet handled in GLSL\n", curOpcode
->name
);
1695 write_mask
= shader_glsl_append_dst(buffer
, arg
);
1697 arguments
[0] = '\0';
1698 if (curOpcode
->num_params
> 0) {
1699 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], write_mask
, &src_param
);
1700 strcat(arguments
, src_param
.param_str
);
1701 for (i
= 2; i
< curOpcode
->num_params
; ++i
) {
1702 strcat(arguments
, ", ");
1703 shader_glsl_add_src_param(arg
, arg
->src
[i
-1], arg
->src_addr
[i
-1], write_mask
, &src_param
);
1704 strcat(arguments
, src_param
.param_str
);
1708 shader_addline(buffer
, "%s(%s));\n", instruction
, arguments
);
1711 /** Process the WINED3DSIO_EXPP instruction in GLSL:
1712 * For shader model 1.x, do the following (and honor the writemask, so use a temporary variable):
1713 * dst.x = 2^(floor(src))
1714 * dst.y = src - floor(src)
1715 * dst.z = 2^src (partial precision is allowed, but optional)
1717 * For 2.0 shaders, just do this (honoring writemask and swizzle):
1718 * dst = 2^src; (partial precision is allowed, but optional)
1720 static void shader_glsl_expp(const SHADER_OPCODE_ARG
*arg
)
1722 glsl_src_param_t src_param
;
1724 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src_param
);
1726 if (arg
->reg_maps
->shader_version
< WINED3DPS_VERSION(2,0))
1730 shader_addline(arg
->buffer
, "tmp0.x = exp2(floor(%s));\n", src_param
.param_str
);
1731 shader_addline(arg
->buffer
, "tmp0.y = %s - floor(%s);\n", src_param
.param_str
, src_param
.param_str
);
1732 shader_addline(arg
->buffer
, "tmp0.z = exp2(%s);\n", src_param
.param_str
);
1733 shader_addline(arg
->buffer
, "tmp0.w = 1.0;\n");
1735 shader_glsl_append_dst(arg
->buffer
, arg
);
1736 shader_glsl_get_write_mask(arg
->dst
, dst_mask
);
1737 shader_addline(arg
->buffer
, "tmp0%s);\n", dst_mask
);
1740 unsigned int mask_size
;
1742 write_mask
= shader_glsl_append_dst(arg
->buffer
, arg
);
1743 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
1745 if (mask_size
> 1) {
1746 shader_addline(arg
->buffer
, "vec%d(exp2(%s)));\n", mask_size
, src_param
.param_str
);
1748 shader_addline(arg
->buffer
, "exp2(%s));\n", src_param
.param_str
);
1753 /** Process the RCP (reciprocal or inverse) opcode in GLSL (dst = 1 / src) */
1754 static void shader_glsl_rcp(const SHADER_OPCODE_ARG
*arg
)
1756 glsl_src_param_t src_param
;
1758 unsigned int mask_size
;
1760 write_mask
= shader_glsl_append_dst(arg
->buffer
, arg
);
1761 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
1762 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_3
, &src_param
);
1764 if (mask_size
> 1) {
1765 shader_addline(arg
->buffer
, "vec%d(1.0 / %s));\n", mask_size
, src_param
.param_str
);
1767 shader_addline(arg
->buffer
, "1.0 / %s);\n", src_param
.param_str
);
1771 static void shader_glsl_rsq(const SHADER_OPCODE_ARG
*arg
)
1773 SHADER_BUFFER
* buffer
= arg
->buffer
;
1774 glsl_src_param_t src_param
;
1776 unsigned int mask_size
;
1778 write_mask
= shader_glsl_append_dst(buffer
, arg
);
1779 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
1781 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_3
, &src_param
);
1783 if (mask_size
> 1) {
1784 shader_addline(buffer
, "vec%d(inversesqrt(%s)));\n", mask_size
, src_param
.param_str
);
1786 shader_addline(buffer
, "inversesqrt(%s));\n", src_param
.param_str
);
1790 /** Process signed comparison opcodes in GLSL. */
1791 static void shader_glsl_compare(const SHADER_OPCODE_ARG
*arg
)
1793 glsl_src_param_t src0_param
;
1794 glsl_src_param_t src1_param
;
1796 unsigned int mask_size
;
1798 write_mask
= shader_glsl_append_dst(arg
->buffer
, arg
);
1799 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
1800 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], write_mask
, &src0_param
);
1801 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], write_mask
, &src1_param
);
1803 if (mask_size
> 1) {
1804 const char *compare
;
1806 switch(arg
->opcode
->opcode
) {
1807 case WINED3DSIO_SLT
: compare
= "lessThan"; break;
1808 case WINED3DSIO_SGE
: compare
= "greaterThanEqual"; break;
1809 default: compare
= "";
1810 FIXME("Can't handle opcode %s\n", arg
->opcode
->name
);
1813 shader_addline(arg
->buffer
, "vec%d(%s(%s, %s)));\n", mask_size
, compare
,
1814 src0_param
.param_str
, src1_param
.param_str
);
1816 switch(arg
->opcode
->opcode
) {
1817 case WINED3DSIO_SLT
:
1818 /* Step(src0, src1) is not suitable here because if src0 == src1 SLT is supposed,
1819 * to return 0.0 but step returns 1.0 because step is not < x
1820 * An alternative is a bvec compare padded with an unused second component.
1821 * step(src1 * -1.0, src0 * -1.0) is not an option because it suffers from the same
1822 * issue. Playing with not() is not possible either because not() does not accept
1825 shader_addline(arg
->buffer
, "(%s < %s) ? 1.0 : 0.0);\n", src0_param
.param_str
, src1_param
.param_str
);
1827 case WINED3DSIO_SGE
:
1828 /* Here we can use the step() function and safe a conditional */
1829 shader_addline(arg
->buffer
, "step(%s, %s));\n", src1_param
.param_str
, src0_param
.param_str
);
1832 FIXME("Can't handle opcode %s\n", arg
->opcode
->name
);
1838 /** Process CMP instruction in GLSL (dst = src0 >= 0.0 ? src1 : src2), per channel */
1839 static void shader_glsl_cmp(const SHADER_OPCODE_ARG
*arg
)
1841 glsl_src_param_t src0_param
;
1842 glsl_src_param_t src1_param
;
1843 glsl_src_param_t src2_param
;
1844 DWORD write_mask
, cmp_channel
= 0;
1847 BOOL temp_destination
= FALSE
;
1849 if(shader_is_scalar(arg
->src
[0])) {
1850 write_mask
= shader_glsl_append_dst(arg
->buffer
, arg
);
1852 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_ALL
, &src0_param
);
1853 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], write_mask
, &src1_param
);
1854 shader_glsl_add_src_param(arg
, arg
->src
[2], arg
->src_addr
[2], write_mask
, &src2_param
);
1856 shader_addline(arg
->buffer
, "%s >= 0.0 ? %s : %s);\n",
1857 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
1859 DWORD src0reg
= arg
->src
[0] & WINED3DSP_REGNUM_MASK
;
1860 DWORD src1reg
= arg
->src
[1] & WINED3DSP_REGNUM_MASK
;
1861 DWORD src2reg
= arg
->src
[2] & WINED3DSP_REGNUM_MASK
;
1862 DWORD src0regtype
= shader_get_regtype(arg
->src
[0]);
1863 DWORD src1regtype
= shader_get_regtype(arg
->src
[1]);
1864 DWORD src2regtype
= shader_get_regtype(arg
->src
[2]);
1865 DWORD dstreg
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
1866 DWORD dstregtype
= shader_get_regtype(arg
->dst
);
1868 /* Cycle through all source0 channels */
1869 for (i
=0; i
<4; i
++) {
1871 /* Find the destination channels which use the current source0 channel */
1872 for (j
=0; j
<4; j
++) {
1873 if ( ((arg
->src
[0] >> (WINED3DSP_SWIZZLE_SHIFT
+ 2*j
)) & 0x3) == i
) {
1874 write_mask
|= WINED3DSP_WRITEMASK_0
<< j
;
1875 cmp_channel
= WINED3DSP_WRITEMASK_0
<< j
;
1879 /* Splitting the cmp instruction up in multiple lines imposes a problem:
1880 * The first lines may overwrite source parameters of the following lines.
1881 * Deal with that by using a temporary destination register if needed
1883 if((src0reg
== dstreg
&& src0regtype
== dstregtype
) ||
1884 (src1reg
== dstreg
&& src1regtype
== dstregtype
) ||
1885 (src2reg
== dstreg
&& src2regtype
== dstregtype
)) {
1887 write_mask
= shader_glsl_get_write_mask(arg
->dst
& (~WINED3DSP_SWIZZLE_MASK
| write_mask
), mask_char
);
1888 if (!write_mask
) continue;
1889 shader_addline(arg
->buffer
, "tmp0%s = (", mask_char
);
1890 temp_destination
= TRUE
;
1892 write_mask
= shader_glsl_append_dst_ext(arg
->buffer
, arg
, arg
->dst
& (~WINED3DSP_SWIZZLE_MASK
| write_mask
));
1893 if (!write_mask
) continue;
1896 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], cmp_channel
, &src0_param
);
1897 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], write_mask
, &src1_param
);
1898 shader_glsl_add_src_param(arg
, arg
->src
[2], arg
->src_addr
[2], write_mask
, &src2_param
);
1900 shader_addline(arg
->buffer
, "%s >= 0.0 ? %s : %s);\n",
1901 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
1904 if(temp_destination
) {
1905 shader_glsl_get_write_mask(arg
->dst
, mask_char
);
1906 shader_glsl_append_dst_ext(arg
->buffer
, arg
, arg
->dst
);
1907 shader_addline(arg
->buffer
, "tmp0%s);\n", mask_char
);
1913 /** Process the CND opcode in GLSL (dst = (src0 > 0.5) ? src1 : src2) */
1914 /* For ps 1.1-1.3, only a single component of src0 is used. For ps 1.4
1915 * the compare is done per component of src0. */
1916 static void shader_glsl_cnd(const SHADER_OPCODE_ARG
*arg
)
1918 glsl_src_param_t src0_param
;
1919 glsl_src_param_t src1_param
;
1920 glsl_src_param_t src2_param
;
1921 DWORD write_mask
, cmp_channel
= 0;
1924 if (arg
->reg_maps
->shader_version
< WINED3DPS_VERSION(1, 4))
1926 write_mask
= shader_glsl_append_dst(arg
->buffer
, arg
);
1927 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
1928 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], write_mask
, &src1_param
);
1929 shader_glsl_add_src_param(arg
, arg
->src
[2], arg
->src_addr
[2], write_mask
, &src2_param
);
1931 /* Fun: The D3DSI_COISSUE flag changes the semantic of the cnd instruction for < 1.4 shaders */
1932 if(arg
->opcode_token
& WINED3DSI_COISSUE
) {
1933 shader_addline(arg
->buffer
, "%s /* COISSUE! */);\n", src1_param
.param_str
);
1935 shader_addline(arg
->buffer
, "%s > 0.5 ? %s : %s);\n",
1936 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
1940 /* Cycle through all source0 channels */
1941 for (i
=0; i
<4; i
++) {
1943 /* Find the destination channels which use the current source0 channel */
1944 for (j
=0; j
<4; j
++) {
1945 if ( ((arg
->src
[0] >> (WINED3DSP_SWIZZLE_SHIFT
+ 2*j
)) & 0x3) == i
) {
1946 write_mask
|= WINED3DSP_WRITEMASK_0
<< j
;
1947 cmp_channel
= WINED3DSP_WRITEMASK_0
<< j
;
1950 write_mask
= shader_glsl_append_dst_ext(arg
->buffer
, arg
, arg
->dst
& (~WINED3DSP_SWIZZLE_MASK
| write_mask
));
1951 if (!write_mask
) continue;
1953 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], cmp_channel
, &src0_param
);
1954 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], write_mask
, &src1_param
);
1955 shader_glsl_add_src_param(arg
, arg
->src
[2], arg
->src_addr
[2], write_mask
, &src2_param
);
1957 shader_addline(arg
->buffer
, "%s > 0.5 ? %s : %s);\n",
1958 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
1962 /** GLSL code generation for WINED3DSIO_MAD: Multiply the first 2 opcodes, then add the last */
1963 static void shader_glsl_mad(const SHADER_OPCODE_ARG
*arg
)
1965 glsl_src_param_t src0_param
;
1966 glsl_src_param_t src1_param
;
1967 glsl_src_param_t src2_param
;
1970 write_mask
= shader_glsl_append_dst(arg
->buffer
, arg
);
1971 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], write_mask
, &src0_param
);
1972 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], write_mask
, &src1_param
);
1973 shader_glsl_add_src_param(arg
, arg
->src
[2], arg
->src_addr
[2], write_mask
, &src2_param
);
1974 shader_addline(arg
->buffer
, "(%s * %s) + %s);\n",
1975 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
1978 /** Handles transforming all WINED3DSIO_M?x? opcodes for
1979 Vertex shaders to GLSL codes */
1980 static void shader_glsl_mnxn(const SHADER_OPCODE_ARG
*arg
)
1982 IWineD3DBaseShaderImpl
*shader
= (IWineD3DBaseShaderImpl
*)arg
->shader
;
1983 const SHADER_OPCODE
*opcode_table
= shader
->baseShader
.shader_ins
;
1984 DWORD shader_version
= arg
->reg_maps
->shader_version
;
1986 int nComponents
= 0;
1987 SHADER_OPCODE_ARG tmpArg
;
1989 memset(&tmpArg
, 0, sizeof(SHADER_OPCODE_ARG
));
1991 /* Set constants for the temporary argument */
1992 tmpArg
.shader
= arg
->shader
;
1993 tmpArg
.buffer
= arg
->buffer
;
1994 tmpArg
.src
[0] = arg
->src
[0];
1995 tmpArg
.src_addr
[0] = arg
->src_addr
[0];
1996 tmpArg
.src_addr
[1] = arg
->src_addr
[1];
1997 tmpArg
.reg_maps
= arg
->reg_maps
;
1999 switch(arg
->opcode
->opcode
) {
2000 case WINED3DSIO_M4x4
:
2002 tmpArg
.opcode
= shader_get_opcode(opcode_table
, shader_version
, WINED3DSIO_DP4
);
2004 case WINED3DSIO_M4x3
:
2006 tmpArg
.opcode
= shader_get_opcode(opcode_table
, shader_version
, WINED3DSIO_DP4
);
2008 case WINED3DSIO_M3x4
:
2010 tmpArg
.opcode
= shader_get_opcode(opcode_table
, shader_version
, WINED3DSIO_DP3
);
2012 case WINED3DSIO_M3x3
:
2014 tmpArg
.opcode
= shader_get_opcode(opcode_table
, shader_version
, WINED3DSIO_DP3
);
2016 case WINED3DSIO_M3x2
:
2018 tmpArg
.opcode
= shader_get_opcode(opcode_table
, shader_version
, WINED3DSIO_DP3
);
2024 for (i
= 0; i
< nComponents
; i
++) {
2025 tmpArg
.dst
= ((arg
->dst
) & ~WINED3DSP_WRITEMASK_ALL
)|(WINED3DSP_WRITEMASK_0
<<i
);
2026 tmpArg
.src
[1] = arg
->src
[1]+i
;
2027 shader_glsl_dot(&tmpArg
);
2032 The LRP instruction performs a component-wise linear interpolation
2033 between the second and third operands using the first operand as the
2034 blend factor. Equation: (dst = src2 + src0 * (src1 - src2))
2035 This is equivalent to mix(src2, src1, src0);
2037 static void shader_glsl_lrp(const SHADER_OPCODE_ARG
*arg
)
2039 glsl_src_param_t src0_param
;
2040 glsl_src_param_t src1_param
;
2041 glsl_src_param_t src2_param
;
2044 write_mask
= shader_glsl_append_dst(arg
->buffer
, arg
);
2046 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], write_mask
, &src0_param
);
2047 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], write_mask
, &src1_param
);
2048 shader_glsl_add_src_param(arg
, arg
->src
[2], arg
->src_addr
[2], write_mask
, &src2_param
);
2050 shader_addline(arg
->buffer
, "mix(%s, %s, %s));\n",
2051 src2_param
.param_str
, src1_param
.param_str
, src0_param
.param_str
);
2054 /** Process the WINED3DSIO_LIT instruction in GLSL:
2055 * dst.x = dst.w = 1.0
2056 * dst.y = (src0.x > 0) ? src0.x
2057 * dst.z = (src0.x > 0) ? ((src0.y > 0) ? pow(src0.y, src.w) : 0) : 0
2058 * where src.w is clamped at +- 128
2060 static void shader_glsl_lit(const SHADER_OPCODE_ARG
*arg
)
2062 glsl_src_param_t src0_param
;
2063 glsl_src_param_t src1_param
;
2064 glsl_src_param_t src3_param
;
2067 shader_glsl_append_dst(arg
->buffer
, arg
);
2068 shader_glsl_get_write_mask(arg
->dst
, dst_mask
);
2070 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2071 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_1
, &src1_param
);
2072 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_3
, &src3_param
);
2074 /* The sdk specifies the instruction like this
2076 * if(src.x > 0.0) dst.y = src.x
2078 * if(src.x > 0.0 && src.y > 0.0) dst.z = pow(src.y, power);
2082 * Obviously that has quite a few conditionals in it which we don't like. So the first step is this:
2083 * dst.x = 1.0 ... No further explanation needed
2084 * dst.y = max(src.y, 0.0); ... If x < 0.0, use 0.0, otherwise x. Same as the conditional
2085 * dst.z = x > 0.0 ? pow(max(y, 0.0), p) : 0; ... 0 ^ power is 0, and otherwise we use y anyway
2086 * dst.w = 1.0. ... Nothing fancy.
2088 * So we still have one conditional in there. So do this:
2089 * dst.z = pow(max(0.0, src.y) * step(0.0, src.x), power);
2091 * 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),
2092 * 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.
2093 * if both x and y are > 0, we get pow(y * 1.0, power), as it is supposed to
2095 shader_addline(arg
->buffer
, "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",
2096 src0_param
.param_str
, src1_param
.param_str
, src0_param
.param_str
, src3_param
.param_str
, dst_mask
);
2099 /** Process the WINED3DSIO_DST instruction in GLSL:
2101 * dst.y = src0.x * src0.y
2105 static void shader_glsl_dst(const SHADER_OPCODE_ARG
*arg
)
2107 glsl_src_param_t src0y_param
;
2108 glsl_src_param_t src0z_param
;
2109 glsl_src_param_t src1y_param
;
2110 glsl_src_param_t src1w_param
;
2113 shader_glsl_append_dst(arg
->buffer
, arg
);
2114 shader_glsl_get_write_mask(arg
->dst
, dst_mask
);
2116 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_1
, &src0y_param
);
2117 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_2
, &src0z_param
);
2118 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], WINED3DSP_WRITEMASK_1
, &src1y_param
);
2119 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], WINED3DSP_WRITEMASK_3
, &src1w_param
);
2121 shader_addline(arg
->buffer
, "vec4(1.0, %s * %s, %s, %s))%s;\n",
2122 src0y_param
.param_str
, src1y_param
.param_str
, src0z_param
.param_str
, src1w_param
.param_str
, dst_mask
);
2125 /** Process the WINED3DSIO_SINCOS instruction in GLSL:
2126 * VS 2.0 requires that specific cosine and sine constants be passed to this instruction so the hardware
2127 * can handle it. But, these functions are built-in for GLSL, so we can just ignore the last 2 params.
2129 * dst.x = cos(src0.?)
2130 * dst.y = sin(src0.?)
2134 static void shader_glsl_sincos(const SHADER_OPCODE_ARG
*arg
)
2136 glsl_src_param_t src0_param
;
2139 write_mask
= shader_glsl_append_dst(arg
->buffer
, arg
);
2140 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2142 switch (write_mask
) {
2143 case WINED3DSP_WRITEMASK_0
:
2144 shader_addline(arg
->buffer
, "cos(%s));\n", src0_param
.param_str
);
2147 case WINED3DSP_WRITEMASK_1
:
2148 shader_addline(arg
->buffer
, "sin(%s));\n", src0_param
.param_str
);
2151 case (WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
):
2152 shader_addline(arg
->buffer
, "vec2(cos(%s), sin(%s)));\n", src0_param
.param_str
, src0_param
.param_str
);
2156 ERR("Write mask should be .x, .y or .xy\n");
2161 /** Process the WINED3DSIO_LOOP instruction in GLSL:
2162 * Start a for() loop where src1.y is the initial value of aL,
2163 * increment aL by src1.z for a total of src1.x iterations.
2164 * Need to use a temporary variable for this operation.
2166 /* FIXME: I don't think nested loops will work correctly this way. */
2167 static void shader_glsl_loop(const SHADER_OPCODE_ARG
*arg
)
2169 glsl_src_param_t src1_param
;
2170 IWineD3DBaseShaderImpl
* shader
= (IWineD3DBaseShaderImpl
*) arg
->shader
;
2171 DWORD regtype
= shader_get_regtype(arg
->src
[1]);
2172 DWORD reg
= arg
->src
[1] & WINED3DSP_REGNUM_MASK
;
2173 const DWORD
*control_values
= NULL
;
2174 const local_constant
*constant
;
2176 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], WINED3DSP_WRITEMASK_ALL
, &src1_param
);
2178 /* Try to hardcode the loop control parameters if possible. Direct3D 9 class hardware doesn't support real
2179 * varying indexing, but Microsoft designed this feature for Shader model 2.x+. If the loop control is
2180 * known at compile time, the GLSL compiler can unroll the loop, and replace indirect addressing with direct
2183 if(regtype
== WINED3DSPR_CONSTINT
) {
2184 LIST_FOR_EACH_ENTRY(constant
, &shader
->baseShader
.constantsI
, local_constant
, entry
) {
2185 if(constant
->idx
== reg
) {
2186 control_values
= constant
->value
;
2192 if(control_values
) {
2193 if(control_values
[2] > 0) {
2194 shader_addline(arg
->buffer
, "for (aL%u = %d; aL%u < (%d * %d + %d); aL%u += %d) {\n",
2195 shader
->baseShader
.cur_loop_depth
, control_values
[1],
2196 shader
->baseShader
.cur_loop_depth
, control_values
[0], control_values
[2], control_values
[1],
2197 shader
->baseShader
.cur_loop_depth
, control_values
[2]);
2198 } else if(control_values
[2] == 0) {
2199 shader_addline(arg
->buffer
, "for (aL%u = %d, tmpInt%u = 0; tmpInt%u < %d; tmpInt%u++) {\n",
2200 shader
->baseShader
.cur_loop_depth
, control_values
[1], shader
->baseShader
.cur_loop_depth
,
2201 shader
->baseShader
.cur_loop_depth
, control_values
[0],
2202 shader
->baseShader
.cur_loop_depth
);
2204 shader_addline(arg
->buffer
, "for (aL%u = %d; aL%u > (%d * %d + %d); aL%u += %d) {\n",
2205 shader
->baseShader
.cur_loop_depth
, control_values
[1],
2206 shader
->baseShader
.cur_loop_depth
, control_values
[0], control_values
[2], control_values
[1],
2207 shader
->baseShader
.cur_loop_depth
, control_values
[2]);
2210 shader_addline(arg
->buffer
, "for (tmpInt%u = 0, aL%u = %s.y; tmpInt%u < %s.x; tmpInt%u++, aL%u += %s.z) {\n",
2211 shader
->baseShader
.cur_loop_depth
, shader
->baseShader
.cur_loop_regno
,
2212 src1_param
.reg_name
, shader
->baseShader
.cur_loop_depth
, src1_param
.reg_name
,
2213 shader
->baseShader
.cur_loop_depth
, shader
->baseShader
.cur_loop_regno
, src1_param
.reg_name
);
2216 shader
->baseShader
.cur_loop_depth
++;
2217 shader
->baseShader
.cur_loop_regno
++;
2220 static void shader_glsl_end(const SHADER_OPCODE_ARG
*arg
)
2222 IWineD3DBaseShaderImpl
* shader
= (IWineD3DBaseShaderImpl
*) arg
->shader
;
2224 shader_addline(arg
->buffer
, "}\n");
2226 if(arg
->opcode
->opcode
== WINED3DSIO_ENDLOOP
) {
2227 shader
->baseShader
.cur_loop_depth
--;
2228 shader
->baseShader
.cur_loop_regno
--;
2230 if(arg
->opcode
->opcode
== WINED3DSIO_ENDREP
) {
2231 shader
->baseShader
.cur_loop_depth
--;
2235 static void shader_glsl_rep(const SHADER_OPCODE_ARG
*arg
)
2237 IWineD3DBaseShaderImpl
* shader
= (IWineD3DBaseShaderImpl
*) arg
->shader
;
2238 glsl_src_param_t src0_param
;
2240 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2241 shader_addline(arg
->buffer
, "for (tmpInt%d = 0; tmpInt%d < %s; tmpInt%d++) {\n",
2242 shader
->baseShader
.cur_loop_depth
, shader
->baseShader
.cur_loop_depth
,
2243 src0_param
.param_str
, shader
->baseShader
.cur_loop_depth
);
2244 shader
->baseShader
.cur_loop_depth
++;
2247 static void shader_glsl_if(const SHADER_OPCODE_ARG
*arg
)
2249 glsl_src_param_t src0_param
;
2251 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2252 shader_addline(arg
->buffer
, "if (%s) {\n", src0_param
.param_str
);
2255 static void shader_glsl_ifc(const SHADER_OPCODE_ARG
*arg
)
2257 glsl_src_param_t src0_param
;
2258 glsl_src_param_t src1_param
;
2260 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2261 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], WINED3DSP_WRITEMASK_0
, &src1_param
);
2263 shader_addline(arg
->buffer
, "if (%s %s %s) {\n",
2264 src0_param
.param_str
, shader_get_comp_op(arg
->opcode_token
), src1_param
.param_str
);
2267 static void shader_glsl_else(const SHADER_OPCODE_ARG
*arg
)
2269 shader_addline(arg
->buffer
, "} else {\n");
2272 static void shader_glsl_break(const SHADER_OPCODE_ARG
*arg
)
2274 shader_addline(arg
->buffer
, "break;\n");
2277 /* FIXME: According to MSDN the compare is done per component. */
2278 static void shader_glsl_breakc(const SHADER_OPCODE_ARG
*arg
)
2280 glsl_src_param_t src0_param
;
2281 glsl_src_param_t src1_param
;
2283 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2284 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], WINED3DSP_WRITEMASK_0
, &src1_param
);
2286 shader_addline(arg
->buffer
, "if (%s %s %s) break;\n",
2287 src0_param
.param_str
, shader_get_comp_op(arg
->opcode_token
), src1_param
.param_str
);
2290 static void shader_glsl_label(const SHADER_OPCODE_ARG
*arg
)
2293 DWORD snum
= (arg
->src
[0]) & WINED3DSP_REGNUM_MASK
;
2294 shader_addline(arg
->buffer
, "}\n");
2295 shader_addline(arg
->buffer
, "void subroutine%u () {\n", snum
);
2298 static void shader_glsl_call(const SHADER_OPCODE_ARG
*arg
)
2300 DWORD snum
= (arg
->src
[0]) & WINED3DSP_REGNUM_MASK
;
2301 shader_addline(arg
->buffer
, "subroutine%u();\n", snum
);
2304 static void shader_glsl_callnz(const SHADER_OPCODE_ARG
*arg
)
2306 glsl_src_param_t src1_param
;
2308 DWORD snum
= (arg
->src
[0]) & WINED3DSP_REGNUM_MASK
;
2309 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], WINED3DSP_WRITEMASK_0
, &src1_param
);
2310 shader_addline(arg
->buffer
, "if (%s) subroutine%u();\n", src1_param
.param_str
, snum
);
2313 /*********************************************
2314 * Pixel Shader Specific Code begins here
2315 ********************************************/
2316 static void pshader_glsl_tex(const SHADER_OPCODE_ARG
*arg
)
2318 IWineD3DPixelShaderImpl
* This
= (IWineD3DPixelShaderImpl
*) arg
->shader
;
2319 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
2320 DWORD shader_version
= arg
->reg_maps
->shader_version
;
2321 glsl_sample_function_t sample_function
;
2322 DWORD sample_flags
= 0;
2325 DWORD mask
= 0, swizzle
;
2327 /* 1.0-1.4: Use destination register as sampler source.
2328 * 2.0+: Use provided sampler source. */
2329 if (shader_version
< WINED3DPS_VERSION(2,0)) sampler_idx
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2330 else sampler_idx
= arg
->src
[1] & WINED3DSP_REGNUM_MASK
;
2331 sampler_type
= arg
->reg_maps
->samplers
[sampler_idx
] & WINED3DSP_TEXTURETYPE_MASK
;
2333 if (shader_version
< WINED3DPS_VERSION(1,4))
2335 DWORD flags
= deviceImpl
->stateBlock
->textureState
[sampler_idx
][WINED3DTSS_TEXTURETRANSFORMFLAGS
];
2337 /* Projected cube textures don't make a lot of sense, the resulting coordinates stay the same. */
2338 if (flags
& WINED3DTTFF_PROJECTED
&& sampler_type
!= WINED3DSTT_CUBE
) {
2339 sample_flags
|= WINED3D_GLSL_SAMPLE_PROJECTED
;
2340 switch (flags
& ~WINED3DTTFF_PROJECTED
) {
2341 case WINED3DTTFF_COUNT1
: FIXME("WINED3DTTFF_PROJECTED with WINED3DTTFF_COUNT1?\n"); break;
2342 case WINED3DTTFF_COUNT2
: mask
= WINED3DSP_WRITEMASK_1
; break;
2343 case WINED3DTTFF_COUNT3
: mask
= WINED3DSP_WRITEMASK_2
; break;
2344 case WINED3DTTFF_COUNT4
:
2345 case WINED3DTTFF_DISABLE
: mask
= WINED3DSP_WRITEMASK_3
; break;
2349 else if (shader_version
< WINED3DPS_VERSION(2,0))
2351 DWORD src_mod
= arg
->src
[0] & WINED3DSP_SRCMOD_MASK
;
2353 if (src_mod
== WINED3DSPSM_DZ
) {
2354 sample_flags
|= WINED3D_GLSL_SAMPLE_PROJECTED
;
2355 mask
= WINED3DSP_WRITEMASK_2
;
2356 } else if (src_mod
== WINED3DSPSM_DW
) {
2357 sample_flags
|= WINED3D_GLSL_SAMPLE_PROJECTED
;
2358 mask
= WINED3DSP_WRITEMASK_3
;
2361 if(arg
->opcode_token
& WINED3DSI_TEXLD_PROJECT
) {
2362 /* ps 2.0 texldp instruction always divides by the fourth component. */
2363 sample_flags
|= WINED3D_GLSL_SAMPLE_PROJECTED
;
2364 mask
= WINED3DSP_WRITEMASK_3
;
2368 if(deviceImpl
->stateBlock
->textures
[sampler_idx
] &&
2369 IWineD3DBaseTexture_GetTextureDimensions(deviceImpl
->stateBlock
->textures
[sampler_idx
]) == GL_TEXTURE_RECTANGLE_ARB
) {
2370 sample_flags
|= WINED3D_GLSL_SAMPLE_RECT
;
2373 shader_glsl_get_sample_function(sampler_type
, sample_flags
, &sample_function
);
2374 mask
|= sample_function
.coord_mask
;
2376 if (shader_version
< WINED3DPS_VERSION(2,0)) swizzle
= WINED3DVS_NOSWIZZLE
;
2377 else swizzle
= arg
->src
[1] & WINED3DSP_SWIZZLE_MASK
;
2379 /* 1.0-1.3: Use destination register as coordinate source.
2380 1.4+: Use provided coordinate source register. */
2381 if (shader_version
< WINED3DPS_VERSION(1,4))
2383 char coord_mask
[6], srcreg
[12];
2384 shader_glsl_get_write_mask(mask
, coord_mask
);
2385 sprintf(srcreg
, "T%u%s", sampler_idx
, coord_mask
);
2386 shader_glsl_gen_sample_code(arg
, sampler_idx
, srcreg
, &sample_function
, swizzle
, NULL
);
2388 glsl_src_param_t coord_param
;
2389 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], mask
, &coord_param
);
2390 if(arg
->opcode_token
& WINED3DSI_TEXLD_BIAS
) {
2391 glsl_src_param_t bias
;
2392 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_3
, &bias
);
2393 shader_glsl_gen_sample_code(arg
, sampler_idx
, coord_param
.param_str
, &sample_function
,
2394 swizzle
, bias
.param_str
);
2396 shader_glsl_gen_sample_code(arg
, sampler_idx
, coord_param
.param_str
, &sample_function
,
2402 static void shader_glsl_texldl(const SHADER_OPCODE_ARG
*arg
)
2404 IWineD3DBaseShaderImpl
* This
= (IWineD3DBaseShaderImpl
*)arg
->shader
;
2405 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
2406 glsl_sample_function_t sample_function
;
2407 glsl_src_param_t coord_param
, lod_param
;
2408 DWORD sample_flags
= WINED3D_GLSL_SAMPLE_LOD
;
2411 DWORD swizzle
= arg
->src
[1] & WINED3DSP_SWIZZLE_MASK
;
2413 shader_glsl_append_dst(arg
->buffer
, arg
);
2415 sampler_idx
= arg
->src
[1] & WINED3DSP_REGNUM_MASK
;
2416 sampler_type
= arg
->reg_maps
->samplers
[sampler_idx
] & WINED3DSP_TEXTURETYPE_MASK
;
2417 if(deviceImpl
->stateBlock
->textures
[sampler_idx
] &&
2418 IWineD3DBaseTexture_GetTextureDimensions(deviceImpl
->stateBlock
->textures
[sampler_idx
]) == GL_TEXTURE_RECTANGLE_ARB
) {
2419 sample_flags
|= WINED3D_GLSL_SAMPLE_RECT
;
2421 shader_glsl_get_sample_function(sampler_type
, sample_flags
, &sample_function
);
2422 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], sample_function
.coord_mask
, &coord_param
);
2424 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_3
, &lod_param
);
2426 if (shader_is_pshader_version(arg
->reg_maps
->shader_version
))
2428 /* The GLSL spec claims the Lod sampling functions are only supported in vertex shaders.
2429 * However, they seem to work just fine in fragment shaders as well. */
2430 WARN("Using %s in fragment shader.\n", sample_function
.name
);
2432 shader_glsl_gen_sample_code(arg
, sampler_idx
, coord_param
.param_str
,
2433 &sample_function
, swizzle
,
2434 lod_param
.param_str
);
2437 static void pshader_glsl_texcoord(const SHADER_OPCODE_ARG
*arg
)
2439 /* FIXME: Make this work for more than just 2D textures */
2440 SHADER_BUFFER
* buffer
= arg
->buffer
;
2444 write_mask
= shader_glsl_append_dst(arg
->buffer
, arg
);
2445 shader_glsl_get_write_mask(write_mask
, dst_mask
);
2447 if (arg
->reg_maps
->shader_version
!= WINED3DPS_VERSION(1,4))
2449 DWORD reg
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2450 shader_addline(buffer
, "clamp(gl_TexCoord[%u], 0.0, 1.0)%s);\n", reg
, dst_mask
);
2452 DWORD reg
= arg
->src
[0] & WINED3DSP_REGNUM_MASK
;
2453 DWORD src_mod
= arg
->src
[0] & WINED3DSP_SRCMOD_MASK
;
2454 char dst_swizzle
[6];
2456 shader_glsl_get_swizzle(arg
->src
[0], FALSE
, write_mask
, dst_swizzle
);
2458 if (src_mod
== WINED3DSPSM_DZ
) {
2459 glsl_src_param_t div_param
;
2460 unsigned int mask_size
= shader_glsl_get_write_mask_size(write_mask
);
2461 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_2
, &div_param
);
2463 if (mask_size
> 1) {
2464 shader_addline(buffer
, "gl_TexCoord[%u]%s / vec%d(%s));\n", reg
, dst_swizzle
, mask_size
, div_param
.param_str
);
2466 shader_addline(buffer
, "gl_TexCoord[%u]%s / %s);\n", reg
, dst_swizzle
, div_param
.param_str
);
2468 } else if (src_mod
== WINED3DSPSM_DW
) {
2469 glsl_src_param_t div_param
;
2470 unsigned int mask_size
= shader_glsl_get_write_mask_size(write_mask
);
2471 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_3
, &div_param
);
2473 if (mask_size
> 1) {
2474 shader_addline(buffer
, "gl_TexCoord[%u]%s / vec%d(%s));\n", reg
, dst_swizzle
, mask_size
, div_param
.param_str
);
2476 shader_addline(buffer
, "gl_TexCoord[%u]%s / %s);\n", reg
, dst_swizzle
, div_param
.param_str
);
2479 shader_addline(buffer
, "gl_TexCoord[%u]%s);\n", reg
, dst_swizzle
);
2484 /** Process the WINED3DSIO_TEXDP3TEX instruction in GLSL:
2485 * Take a 3-component dot product of the TexCoord[dstreg] and src,
2486 * then perform a 1D texture lookup from stage dstregnum, place into dst. */
2487 static void pshader_glsl_texdp3tex(const SHADER_OPCODE_ARG
*arg
)
2489 glsl_src_param_t src0_param
;
2490 char coord_param
[64];
2491 glsl_sample_function_t sample_function
;
2492 DWORD sampler_idx
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2493 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2494 DWORD sampler_type
= arg
->reg_maps
->samplers
[sampler_idx
] & WINED3DSP_TEXTURETYPE_MASK
;
2497 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], src_mask
, &src0_param
);
2499 /* Do I have to take care about the projected bit? I don't think so, since the dp3 returns only one
2500 * scalar, and projected sampling would require 4.
2502 * It is a dependent read - not valid with conditional NP2 textures
2504 shader_glsl_get_sample_function(sampler_type
, 0, &sample_function
);
2505 mask_size
= shader_glsl_get_write_mask_size(sample_function
.coord_mask
);
2510 sprintf(coord_param
, "dot(gl_TexCoord[%u].xyz, %s)",
2511 sampler_idx
, src0_param
.param_str
);
2515 sprintf(coord_param
, "vec2(dot(gl_TexCoord[%u].xyz, %s), 0.0)",
2516 sampler_idx
, src0_param
.param_str
);
2520 sprintf(coord_param
, "vec3(dot(gl_TexCoord[%u].xyz, %s), 0.0, 0.0)",
2521 sampler_idx
, src0_param
.param_str
);
2525 FIXME("Unexpected mask size %u\n", mask_size
);
2528 shader_glsl_gen_sample_code(arg
, sampler_idx
, coord_param
,
2529 &sample_function
, WINED3DVS_NOSWIZZLE
,
2533 /** Process the WINED3DSIO_TEXDP3 instruction in GLSL:
2534 * Take a 3-component dot product of the TexCoord[dstreg] and src. */
2535 static void pshader_glsl_texdp3(const SHADER_OPCODE_ARG
*arg
)
2537 glsl_src_param_t src0_param
;
2538 DWORD dstreg
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2539 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2541 unsigned int mask_size
;
2543 dst_mask
= shader_glsl_append_dst(arg
->buffer
, arg
);
2544 mask_size
= shader_glsl_get_write_mask_size(dst_mask
);
2545 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], src_mask
, &src0_param
);
2547 if (mask_size
> 1) {
2548 shader_addline(arg
->buffer
, "vec%d(dot(T%u.xyz, %s)));\n", mask_size
, dstreg
, src0_param
.param_str
);
2550 shader_addline(arg
->buffer
, "dot(T%u.xyz, %s));\n", dstreg
, src0_param
.param_str
);
2554 /** Process the WINED3DSIO_TEXDEPTH instruction in GLSL:
2555 * Calculate the depth as dst.x / dst.y */
2556 static void pshader_glsl_texdepth(const SHADER_OPCODE_ARG
*arg
)
2558 glsl_dst_param_t dst_param
;
2560 shader_glsl_add_dst_param(arg
, arg
->dst
, 0, &dst_param
);
2562 /* Tests show that texdepth never returns anything below 0.0, and that r5.y is clamped to 1.0.
2563 * Negative input is accepted, -0.25 / -0.5 returns 0.5. GL should clamp gl_FragDepth to [0;1], but
2564 * this doesn't always work, so clamp the results manually. Whether or not the x value is clamped at 1
2565 * too is irrelevant, since if x = 0, any y value < 1.0 (and > 1.0 is not allowed) results in a result
2568 shader_addline(arg
->buffer
, "gl_FragDepth = clamp((%s.x / min(%s.y, 1.0)), 0.0, 1.0);\n", dst_param
.reg_name
, dst_param
.reg_name
);
2571 /** Process the WINED3DSIO_TEXM3X2DEPTH instruction in GLSL:
2572 * Last row of a 3x2 matrix multiply, use the result to calculate the depth:
2573 * Calculate tmp0.y = TexCoord[dstreg] . src.xyz; (tmp0.x has already been calculated)
2574 * depth = (tmp0.y == 0.0) ? 1.0 : tmp0.x / tmp0.y
2576 static void pshader_glsl_texm3x2depth(const SHADER_OPCODE_ARG
*arg
)
2578 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2579 DWORD dstreg
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2580 glsl_src_param_t src0_param
;
2582 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], src_mask
, &src0_param
);
2584 shader_addline(arg
->buffer
, "tmp0.y = dot(T%u.xyz, %s);\n", dstreg
, src0_param
.param_str
);
2585 shader_addline(arg
->buffer
, "gl_FragDepth = (tmp0.y == 0.0) ? 1.0 : clamp(tmp0.x / tmp0.y, 0.0, 1.0);\n");
2588 /** Process the WINED3DSIO_TEXM3X2PAD instruction in GLSL
2589 * Calculate the 1st of a 2-row matrix multiplication. */
2590 static void pshader_glsl_texm3x2pad(const SHADER_OPCODE_ARG
*arg
)
2592 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2593 DWORD reg
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2594 SHADER_BUFFER
* buffer
= arg
->buffer
;
2595 glsl_src_param_t src0_param
;
2597 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], src_mask
, &src0_param
);
2598 shader_addline(buffer
, "tmp0.x = dot(T%u.xyz, %s);\n", reg
, src0_param
.param_str
);
2601 /** Process the WINED3DSIO_TEXM3X3PAD instruction in GLSL
2602 * Calculate the 1st or 2nd row of a 3-row matrix multiplication. */
2603 static void pshader_glsl_texm3x3pad(const SHADER_OPCODE_ARG
* arg
)
2605 IWineD3DPixelShaderImpl
* shader
= (IWineD3DPixelShaderImpl
*) arg
->shader
;
2606 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2607 DWORD reg
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2608 SHADER_BUFFER
* buffer
= arg
->buffer
;
2609 SHADER_PARSE_STATE
* current_state
= &shader
->baseShader
.parse_state
;
2610 glsl_src_param_t src0_param
;
2612 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], src_mask
, &src0_param
);
2613 shader_addline(buffer
, "tmp0.%c = dot(T%u.xyz, %s);\n", 'x' + current_state
->current_row
, reg
, src0_param
.param_str
);
2614 current_state
->texcoord_w
[current_state
->current_row
++] = reg
;
2617 static void pshader_glsl_texm3x2tex(const SHADER_OPCODE_ARG
*arg
)
2619 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2620 DWORD reg
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2621 SHADER_BUFFER
* buffer
= arg
->buffer
;
2622 glsl_src_param_t src0_param
;
2623 DWORD sampler_type
= arg
->reg_maps
->samplers
[reg
] & WINED3DSP_TEXTURETYPE_MASK
;
2624 glsl_sample_function_t sample_function
;
2626 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], src_mask
, &src0_param
);
2627 shader_addline(buffer
, "tmp0.y = dot(T%u.xyz, %s);\n", reg
, src0_param
.param_str
);
2629 shader_glsl_get_sample_function(sampler_type
, 0, &sample_function
);
2631 /* Sample the texture using the calculated coordinates */
2632 shader_glsl_gen_sample_code(arg
, reg
, "tmp0.xy",
2633 &sample_function
, WINED3DVS_NOSWIZZLE
,
2637 /** Process the WINED3DSIO_TEXM3X3TEX instruction in GLSL
2638 * Perform the 3rd row of a 3x3 matrix multiply, then sample the texture using the calculated coordinates */
2639 static void pshader_glsl_texm3x3tex(const SHADER_OPCODE_ARG
*arg
)
2641 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2642 glsl_src_param_t src0_param
;
2643 DWORD reg
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2644 IWineD3DPixelShaderImpl
* This
= (IWineD3DPixelShaderImpl
*) arg
->shader
;
2645 SHADER_PARSE_STATE
* current_state
= &This
->baseShader
.parse_state
;
2646 DWORD sampler_type
= arg
->reg_maps
->samplers
[reg
] & WINED3DSP_TEXTURETYPE_MASK
;
2647 glsl_sample_function_t sample_function
;
2649 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], src_mask
, &src0_param
);
2650 shader_addline(arg
->buffer
, "tmp0.z = dot(T%u.xyz, %s);\n", reg
, src0_param
.param_str
);
2652 /* Dependent read, not valid with conditional NP2 */
2653 shader_glsl_get_sample_function(sampler_type
, 0, &sample_function
);
2655 /* Sample the texture using the calculated coordinates */
2656 shader_glsl_gen_sample_code(arg
, reg
, "tmp0.xyz",
2657 &sample_function
, WINED3DVS_NOSWIZZLE
,
2660 current_state
->current_row
= 0;
2663 /** Process the WINED3DSIO_TEXM3X3 instruction in GLSL
2664 * Perform the 3rd row of a 3x3 matrix multiply */
2665 static void pshader_glsl_texm3x3(const SHADER_OPCODE_ARG
*arg
)
2667 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2668 glsl_src_param_t src0_param
;
2670 DWORD reg
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2671 IWineD3DPixelShaderImpl
* This
= (IWineD3DPixelShaderImpl
*) arg
->shader
;
2672 SHADER_PARSE_STATE
* current_state
= &This
->baseShader
.parse_state
;
2674 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], src_mask
, &src0_param
);
2676 shader_glsl_append_dst(arg
->buffer
, arg
);
2677 shader_glsl_get_write_mask(arg
->dst
, dst_mask
);
2678 shader_addline(arg
->buffer
, "vec4(tmp0.xy, dot(T%u.xyz, %s), 1.0)%s);\n", reg
, src0_param
.param_str
, dst_mask
);
2680 current_state
->current_row
= 0;
2683 /** Process the WINED3DSIO_TEXM3X3SPEC instruction in GLSL
2684 * Perform the final texture lookup based on the previous 2 3x3 matrix multiplies */
2685 static void pshader_glsl_texm3x3spec(const SHADER_OPCODE_ARG
*arg
)
2687 IWineD3DPixelShaderImpl
* shader
= (IWineD3DPixelShaderImpl
*) arg
->shader
;
2688 DWORD reg
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2689 glsl_src_param_t src0_param
;
2690 glsl_src_param_t src1_param
;
2691 SHADER_BUFFER
* buffer
= arg
->buffer
;
2692 SHADER_PARSE_STATE
* current_state
= &shader
->baseShader
.parse_state
;
2693 DWORD stype
= arg
->reg_maps
->samplers
[reg
] & WINED3DSP_TEXTURETYPE_MASK
;
2694 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2695 glsl_sample_function_t sample_function
;
2697 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], src_mask
, &src0_param
);
2698 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], src_mask
, &src1_param
);
2700 /* Perform the last matrix multiply operation */
2701 shader_addline(buffer
, "tmp0.z = dot(T%u.xyz, %s);\n", reg
, src0_param
.param_str
);
2702 /* Reflection calculation */
2703 shader_addline(buffer
, "tmp0.xyz = -reflect((%s), normalize(tmp0.xyz));\n", src1_param
.param_str
);
2705 /* Dependent read, not valid with conditional NP2 */
2706 shader_glsl_get_sample_function(stype
, 0, &sample_function
);
2708 /* Sample the texture */
2709 shader_glsl_gen_sample_code(arg
, reg
, "tmp0.xyz",
2710 &sample_function
, WINED3DVS_NOSWIZZLE
,
2713 current_state
->current_row
= 0;
2716 /** Process the WINED3DSIO_TEXM3X3VSPEC instruction in GLSL
2717 * Perform the final texture lookup based on the previous 2 3x3 matrix multiplies */
2718 static void pshader_glsl_texm3x3vspec(const SHADER_OPCODE_ARG
*arg
)
2720 IWineD3DPixelShaderImpl
* shader
= (IWineD3DPixelShaderImpl
*) arg
->shader
;
2721 DWORD reg
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2722 SHADER_BUFFER
* buffer
= arg
->buffer
;
2723 SHADER_PARSE_STATE
* current_state
= &shader
->baseShader
.parse_state
;
2724 glsl_src_param_t src0_param
;
2725 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2726 DWORD sampler_type
= arg
->reg_maps
->samplers
[reg
] & WINED3DSP_TEXTURETYPE_MASK
;
2727 glsl_sample_function_t sample_function
;
2729 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], src_mask
, &src0_param
);
2731 /* Perform the last matrix multiply operation */
2732 shader_addline(buffer
, "tmp0.z = dot(vec3(T%u), vec3(%s));\n", reg
, src0_param
.param_str
);
2734 /* Construct the eye-ray vector from w coordinates */
2735 shader_addline(buffer
, "tmp1.xyz = normalize(vec3(gl_TexCoord[%u].w, gl_TexCoord[%u].w, gl_TexCoord[%u].w));\n",
2736 current_state
->texcoord_w
[0], current_state
->texcoord_w
[1], reg
);
2737 shader_addline(buffer
, "tmp0.xyz = -reflect(tmp1.xyz, normalize(tmp0.xyz));\n");
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(arg
, reg
, "tmp0.xyz",
2744 &sample_function
, WINED3DVS_NOSWIZZLE
,
2747 current_state
->current_row
= 0;
2750 /** Process the WINED3DSIO_TEXBEM instruction in GLSL.
2751 * Apply a fake bump map transform.
2752 * texbem is pshader <= 1.3 only, this saves a few version checks
2754 static void pshader_glsl_texbem(const SHADER_OPCODE_ARG
*arg
)
2756 IWineD3DPixelShaderImpl
* This
= (IWineD3DPixelShaderImpl
*) arg
->shader
;
2757 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
2759 glsl_sample_function_t sample_function
;
2760 glsl_src_param_t coord_param
;
2767 sampler_idx
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2768 flags
= deviceImpl
->stateBlock
->textureState
[sampler_idx
][WINED3DTSS_TEXTURETRANSFORMFLAGS
];
2770 sampler_type
= arg
->reg_maps
->samplers
[sampler_idx
] & WINED3DSP_TEXTURETYPE_MASK
;
2771 /* Dependent read, not valid with conditional NP2 */
2772 shader_glsl_get_sample_function(sampler_type
, 0, &sample_function
);
2773 mask
= sample_function
.coord_mask
;
2775 shader_glsl_get_write_mask(mask
, coord_mask
);
2777 /* with projective textures, texbem only divides the static texture coord, not the displacement,
2778 * so we can't let the GL handle this.
2780 if (flags
& WINED3DTTFF_PROJECTED
) {
2782 char coord_div_mask
[3];
2783 switch (flags
& ~WINED3DTTFF_PROJECTED
) {
2784 case WINED3DTTFF_COUNT1
: FIXME("WINED3DTTFF_PROJECTED with WINED3DTTFF_COUNT1?\n"); break;
2785 case WINED3DTTFF_COUNT2
: div_mask
= WINED3DSP_WRITEMASK_1
; break;
2786 case WINED3DTTFF_COUNT3
: div_mask
= WINED3DSP_WRITEMASK_2
; break;
2787 case WINED3DTTFF_COUNT4
:
2788 case WINED3DTTFF_DISABLE
: div_mask
= WINED3DSP_WRITEMASK_3
; break;
2790 shader_glsl_get_write_mask(div_mask
, coord_div_mask
);
2791 shader_addline(arg
->buffer
, "T%u%s /= T%u%s;\n", sampler_idx
, coord_mask
, sampler_idx
, coord_div_mask
);
2794 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_0
|WINED3DSP_WRITEMASK_1
, &coord_param
);
2796 sprintf(coord
, "T%u%s + vec4(bumpenvmat%d * %s, 0.0, 0.0)%s",
2797 sampler_idx
, coord_mask
, sampler_idx
, coord_param
.param_str
, coord_mask
);
2798 shader_glsl_gen_sample_code(arg
, sampler_idx
, coord
,
2799 &sample_function
, WINED3DVS_NOSWIZZLE
,
2802 if(arg
->opcode
->opcode
== WINED3DSIO_TEXBEML
) {
2803 glsl_src_param_t luminance_param
;
2804 glsl_dst_param_t dst_param
;
2806 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_2
, &luminance_param
);
2807 shader_glsl_add_dst_param(arg
, arg
->dst
, arg
->dst_addr
, &dst_param
);
2809 shader_addline(arg
->buffer
, "%s%s *= (%s * luminancescale%d + luminanceoffset%d);\n",
2810 dst_param
.reg_name
, dst_param
.mask_str
,
2811 luminance_param
.param_str
, sampler_idx
, sampler_idx
);
2815 static void pshader_glsl_bem(const SHADER_OPCODE_ARG
*arg
)
2817 glsl_src_param_t src0_param
, src1_param
;
2818 DWORD sampler_idx
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2820 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_0
|WINED3DSP_WRITEMASK_1
, &src0_param
);
2821 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], WINED3DSP_WRITEMASK_0
|WINED3DSP_WRITEMASK_1
, &src1_param
);
2823 shader_glsl_append_dst(arg
->buffer
, arg
);
2824 shader_addline(arg
->buffer
, "%s + bumpenvmat%d * %s);\n",
2825 src0_param
.param_str
, sampler_idx
, src1_param
.param_str
);
2828 /** Process the WINED3DSIO_TEXREG2AR instruction in GLSL
2829 * Sample 2D texture at dst using the alpha & red (wx) components of src as texture coordinates */
2830 static void pshader_glsl_texreg2ar(const SHADER_OPCODE_ARG
*arg
)
2832 glsl_src_param_t src0_param
;
2833 DWORD sampler_idx
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2835 DWORD sampler_type
= arg
->reg_maps
->samplers
[sampler_idx
] & WINED3DSP_TEXTURETYPE_MASK
;
2836 glsl_sample_function_t sample_function
;
2838 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_ALL
, &src0_param
);
2840 shader_glsl_get_sample_function(sampler_type
, 0, &sample_function
);
2841 sprintf(src
, "%s.wx", src0_param
.reg_name
);
2842 shader_glsl_gen_sample_code(arg
, sampler_idx
, src
,
2843 &sample_function
, WINED3DVS_NOSWIZZLE
,
2847 /** Process the WINED3DSIO_TEXREG2GB instruction in GLSL
2848 * Sample 2D texture at dst using the green & blue (yz) components of src as texture coordinates */
2849 static void pshader_glsl_texreg2gb(const SHADER_OPCODE_ARG
*arg
)
2851 glsl_src_param_t src0_param
;
2852 DWORD sampler_idx
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2854 DWORD sampler_type
= arg
->reg_maps
->samplers
[sampler_idx
] & WINED3DSP_TEXTURETYPE_MASK
;
2855 glsl_sample_function_t sample_function
;
2857 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_ALL
, &src0_param
);
2859 shader_glsl_get_sample_function(sampler_type
, 0, &sample_function
);
2860 sprintf(src
, "%s.yz", src0_param
.reg_name
);
2861 shader_glsl_gen_sample_code(arg
, sampler_idx
, src
,
2862 &sample_function
, WINED3DVS_NOSWIZZLE
,
2866 /** Process the WINED3DSIO_TEXREG2RGB instruction in GLSL
2867 * Sample texture at dst using the rgb (xyz) components of src as texture coordinates */
2868 static void pshader_glsl_texreg2rgb(const SHADER_OPCODE_ARG
*arg
)
2870 glsl_src_param_t src0_param
;
2871 DWORD sampler_idx
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2872 DWORD sampler_type
= arg
->reg_maps
->samplers
[sampler_idx
] & WINED3DSP_TEXTURETYPE_MASK
;
2873 glsl_sample_function_t sample_function
;
2875 /* Dependent read, not valid with conditional NP2 */
2876 shader_glsl_get_sample_function(sampler_type
, 0, &sample_function
);
2877 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], sample_function
.coord_mask
, &src0_param
);
2879 shader_glsl_gen_sample_code(arg
, sampler_idx
, src0_param
.param_str
,
2880 &sample_function
, WINED3DVS_NOSWIZZLE
,
2884 /** Process the WINED3DSIO_TEXKILL instruction in GLSL.
2885 * If any of the first 3 components are < 0, discard this pixel */
2886 static void pshader_glsl_texkill(const SHADER_OPCODE_ARG
*arg
)
2888 glsl_dst_param_t dst_param
;
2890 /* The argument is a destination parameter, and no writemasks are allowed */
2891 shader_glsl_add_dst_param(arg
, arg
->dst
, 0, &dst_param
);
2892 if ((arg
->reg_maps
->shader_version
>= WINED3DPS_VERSION(2,0)))
2894 /* 2.0 shaders compare all 4 components in texkill */
2895 shader_addline(arg
->buffer
, "if (any(lessThan(%s.xyzw, vec4(0.0)))) discard;\n", dst_param
.reg_name
);
2897 /* 1.X shaders only compare the first 3 components, probably due to the nature of the texkill
2898 * instruction as a tex* instruction, and phase, which kills all a / w components. Even if all
2899 * 4 components are defined, only the first 3 are used
2901 shader_addline(arg
->buffer
, "if (any(lessThan(%s.xyz, vec3(0.0)))) discard;\n", dst_param
.reg_name
);
2905 /** Process the WINED3DSIO_DP2ADD instruction in GLSL.
2906 * dst = dot2(src0, src1) + src2 */
2907 static void pshader_glsl_dp2add(const SHADER_OPCODE_ARG
*arg
)
2909 glsl_src_param_t src0_param
;
2910 glsl_src_param_t src1_param
;
2911 glsl_src_param_t src2_param
;
2913 unsigned int mask_size
;
2915 write_mask
= shader_glsl_append_dst(arg
->buffer
, arg
);
2916 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
2918 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
, &src0_param
);
2919 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
, &src1_param
);
2920 shader_glsl_add_src_param(arg
, arg
->src
[2], arg
->src_addr
[2], WINED3DSP_WRITEMASK_0
, &src2_param
);
2922 if (mask_size
> 1) {
2923 shader_addline(arg
->buffer
, "vec%d(dot(%s, %s) + %s));\n", mask_size
, src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
2925 shader_addline(arg
->buffer
, "dot(%s, %s) + %s);\n", src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
2929 static void pshader_glsl_input_pack(SHADER_BUFFER
* buffer
, const struct semantic
* semantics_in
,
2930 IWineD3DPixelShader
*iface
, enum vertexprocessing_mode vertexprocessing
)
2933 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*) iface
;
2935 for (i
= 0; i
< MAX_REG_INPUT
; i
++) {
2937 DWORD usage_token
= semantics_in
[i
].usage
;
2938 DWORD register_token
= semantics_in
[i
].reg
;
2939 DWORD usage
, usage_idx
;
2943 if (!usage_token
) continue;
2944 usage
= (usage_token
& WINED3DSP_DCL_USAGE_MASK
) >> WINED3DSP_DCL_USAGE_SHIFT
;
2945 usage_idx
= (usage_token
& WINED3DSP_DCL_USAGEINDEX_MASK
) >> WINED3DSP_DCL_USAGEINDEX_SHIFT
;
2946 shader_glsl_get_write_mask(register_token
, reg_mask
);
2950 case WINED3DDECLUSAGE_TEXCOORD
:
2951 if(usage_idx
< 8 && vertexprocessing
== pretransformed
) {
2952 shader_addline(buffer
, "IN[%u]%s = gl_TexCoord[%u]%s;\n",
2953 This
->input_reg_map
[i
], reg_mask
, usage_idx
, reg_mask
);
2955 shader_addline(buffer
, "IN[%u]%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
2956 This
->input_reg_map
[i
], reg_mask
, reg_mask
);
2960 case WINED3DDECLUSAGE_COLOR
:
2962 shader_addline(buffer
, "IN[%u]%s = vec4(gl_Color)%s;\n",
2963 This
->input_reg_map
[i
], reg_mask
, reg_mask
);
2964 else if (usage_idx
== 1)
2965 shader_addline(buffer
, "IN[%u]%s = vec4(gl_SecondaryColor)%s;\n",
2966 This
->input_reg_map
[i
], reg_mask
, reg_mask
);
2968 shader_addline(buffer
, "IN[%u]%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
2969 This
->input_reg_map
[i
], reg_mask
, reg_mask
);
2973 shader_addline(buffer
, "IN[%u]%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
2974 This
->input_reg_map
[i
], reg_mask
, reg_mask
);
2979 /*********************************************
2980 * Vertex Shader Specific Code begins here
2981 ********************************************/
2983 static void add_glsl_program_entry(struct shader_glsl_priv
*priv
, struct glsl_shader_prog_link
*entry
) {
2984 glsl_program_key_t
*key
;
2986 key
= HeapAlloc(GetProcessHeap(), 0, sizeof(glsl_program_key_t
));
2987 key
->vshader
= entry
->vshader
;
2988 key
->pshader
= entry
->pshader
;
2989 key
->vs_args
= entry
->vs_args
;
2990 key
->ps_args
= entry
->ps_args
;
2992 hash_table_put(priv
->glsl_program_lookup
, key
, entry
);
2995 static struct glsl_shader_prog_link
*get_glsl_program_entry(struct shader_glsl_priv
*priv
,
2996 IWineD3DVertexShader
*vshader
, IWineD3DPixelShader
*pshader
, struct vs_compile_args
*vs_args
,
2997 struct ps_compile_args
*ps_args
) {
2998 glsl_program_key_t key
;
3000 key
.vshader
= vshader
;
3001 key
.pshader
= pshader
;
3002 key
.vs_args
= *vs_args
;
3003 key
.ps_args
= *ps_args
;
3005 return hash_table_get(priv
->glsl_program_lookup
, &key
);
3008 static void delete_glsl_program_entry(struct shader_glsl_priv
*priv
, const WineD3D_GL_Info
*gl_info
,
3009 struct glsl_shader_prog_link
*entry
)
3011 glsl_program_key_t
*key
;
3013 key
= HeapAlloc(GetProcessHeap(), 0, sizeof(glsl_program_key_t
));
3014 key
->vshader
= entry
->vshader
;
3015 key
->pshader
= entry
->pshader
;
3016 key
->vs_args
= entry
->vs_args
;
3017 key
->ps_args
= entry
->ps_args
;
3018 hash_table_remove(priv
->glsl_program_lookup
, key
);
3020 GL_EXTCALL(glDeleteObjectARB(entry
->programId
));
3021 if (entry
->vshader
) list_remove(&entry
->vshader_entry
);
3022 if (entry
->pshader
) list_remove(&entry
->pshader_entry
);
3023 HeapFree(GetProcessHeap(), 0, entry
->vuniformF_locations
);
3024 HeapFree(GetProcessHeap(), 0, entry
->puniformF_locations
);
3025 HeapFree(GetProcessHeap(), 0, entry
);
3028 static void handle_ps3_input(SHADER_BUFFER
*buffer
, const struct semantic
*semantics_in
,
3029 const struct semantic
*semantics_out
, const WineD3D_GL_Info
*gl_info
, const DWORD
*map
)
3032 DWORD usage_token
, usage_token_out
;
3033 DWORD register_token
, register_token_out
;
3034 DWORD usage
, usage_idx
, usage_out
, usage_idx_out
;
3037 DWORD in_count
= GL_LIMITS(glsl_varyings
) / 4;
3038 char reg_mask
[6], reg_mask_out
[6];
3039 char destination
[50];
3041 set
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*set
) * (in_count
+ 2));
3043 if (!semantics_out
) {
3044 /* Save gl_FrontColor & gl_FrontSecondaryColor before overwriting them. */
3045 shader_addline(buffer
, "vec4 front_color = gl_FrontColor;\n");
3046 shader_addline(buffer
, "vec4 front_secondary_color = gl_FrontSecondaryColor;\n");
3049 for(i
= 0; i
< MAX_REG_INPUT
; i
++) {
3050 usage_token
= semantics_in
[i
].usage
;
3051 if (!usage_token
) continue;
3054 if (in_idx
>= (in_count
+ 2)) {
3055 FIXME("More input varyings declared than supported, expect issues\n");
3057 } else if(map
[i
] == -1) {
3058 /* Declared, but not read register */
3062 if (in_idx
== in_count
) {
3063 sprintf(destination
, "gl_FrontColor");
3064 } else if (in_idx
== in_count
+ 1) {
3065 sprintf(destination
, "gl_FrontSecondaryColor");
3067 sprintf(destination
, "IN[%u]", in_idx
);
3070 register_token
= semantics_in
[i
].reg
;
3072 usage
= (usage_token
& WINED3DSP_DCL_USAGE_MASK
) >> WINED3DSP_DCL_USAGE_SHIFT
;
3073 usage_idx
= (usage_token
& WINED3DSP_DCL_USAGEINDEX_MASK
) >> WINED3DSP_DCL_USAGEINDEX_SHIFT
;
3074 set
[map
[i
]] = shader_glsl_get_write_mask(register_token
, reg_mask
);
3076 if(!semantics_out
) {
3078 case WINED3DDECLUSAGE_COLOR
:
3080 shader_addline(buffer
, "%s%s = front_color%s;\n",
3081 destination
, reg_mask
, reg_mask
);
3082 else if (usage_idx
== 1)
3083 shader_addline(buffer
, "%s%s = front_secondary_color%s;\n",
3084 destination
, reg_mask
, reg_mask
);
3086 shader_addline(buffer
, "%s%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
3087 destination
, reg_mask
, reg_mask
);
3090 case WINED3DDECLUSAGE_TEXCOORD
:
3091 if (usage_idx
< 8) {
3092 shader_addline(buffer
, "%s%s = gl_TexCoord[%u]%s;\n",
3093 destination
, reg_mask
, usage_idx
, reg_mask
);
3095 shader_addline(buffer
, "%s%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
3096 destination
, reg_mask
, reg_mask
);
3100 case WINED3DDECLUSAGE_FOG
:
3101 shader_addline(buffer
, "%s%s = vec4(gl_FogFragCoord, 0.0, 0.0, 0.0)%s;\n",
3102 destination
, reg_mask
, reg_mask
);
3106 shader_addline(buffer
, "%s%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
3107 destination
, reg_mask
, reg_mask
);
3111 for(j
= 0; j
< MAX_REG_OUTPUT
; j
++) {
3112 usage_token_out
= semantics_out
[j
].usage
;
3113 if (!usage_token_out
) continue;
3114 register_token_out
= semantics_out
[j
].reg
;
3116 usage_out
= (usage_token_out
& WINED3DSP_DCL_USAGE_MASK
) >> WINED3DSP_DCL_USAGE_SHIFT
;
3117 usage_idx_out
= (usage_token_out
& WINED3DSP_DCL_USAGEINDEX_MASK
) >> WINED3DSP_DCL_USAGEINDEX_SHIFT
;
3118 shader_glsl_get_write_mask(register_token_out
, reg_mask_out
);
3120 if(usage
== usage_out
&&
3121 usage_idx
== usage_idx_out
) {
3122 shader_addline(buffer
, "%s%s = OUT[%u]%s;\n",
3123 destination
, reg_mask
, j
, reg_mask
);
3128 shader_addline(buffer
, "%s%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
3129 destination
, reg_mask
, reg_mask
);
3134 /* This is solely to make the compiler / linker happy and avoid warning about undefined
3135 * varyings. It shouldn't result in any real code executed on the GPU, since all read
3136 * input varyings are assigned above, if the optimizer works properly.
3138 for(i
= 0; i
< in_count
+ 2; i
++) {
3139 if(set
[i
] != WINED3DSP_WRITEMASK_ALL
) {
3140 unsigned int size
= 0;
3141 memset(reg_mask
, 0, sizeof(reg_mask
));
3142 if(!(set
[i
] & WINED3DSP_WRITEMASK_0
)) {
3143 reg_mask
[size
] = 'x';
3146 if(!(set
[i
] & WINED3DSP_WRITEMASK_1
)) {
3147 reg_mask
[size
] = 'y';
3150 if(!(set
[i
] & WINED3DSP_WRITEMASK_2
)) {
3151 reg_mask
[size
] = 'z';
3154 if(!(set
[i
] & WINED3DSP_WRITEMASK_3
)) {
3155 reg_mask
[size
] = 'w';
3159 if (i
== in_count
) {
3160 sprintf(destination
, "gl_FrontColor");
3161 } else if (i
== in_count
+ 1) {
3162 sprintf(destination
, "gl_FrontSecondaryColor");
3164 sprintf(destination
, "IN[%u]", i
);
3168 shader_addline(buffer
, "%s.%s = 0.0;\n", destination
, reg_mask
);
3170 shader_addline(buffer
, "%s.%s = vec%u(0.0);\n", destination
, reg_mask
, size
);
3175 HeapFree(GetProcessHeap(), 0, set
);
3178 static GLhandleARB
generate_param_reorder_function(IWineD3DVertexShader
*vertexshader
,
3179 IWineD3DPixelShader
*pixelshader
, const WineD3D_GL_Info
*gl_info
)
3181 GLhandleARB ret
= 0;
3182 IWineD3DVertexShaderImpl
*vs
= (IWineD3DVertexShaderImpl
*) vertexshader
;
3183 IWineD3DPixelShaderImpl
*ps
= (IWineD3DPixelShaderImpl
*) pixelshader
;
3184 IWineD3DDeviceImpl
*device
;
3185 DWORD vs_major
= WINED3DSHADER_VERSION_MAJOR(vs
->baseShader
.reg_maps
.shader_version
);
3186 DWORD ps_major
= ps
? WINED3DSHADER_VERSION_MAJOR(ps
->baseShader
.reg_maps
.shader_version
) : 0;
3188 SHADER_BUFFER buffer
;
3190 DWORD register_token
;
3191 DWORD usage
, usage_idx
, writemask
;
3193 const struct semantic
*semantics_out
, *semantics_in
;
3195 shader_buffer_init(&buffer
);
3197 shader_addline(&buffer
, "#version 120\n");
3199 if(vs_major
< 3 && ps_major
< 3) {
3200 /* That one is easy: The vertex shader writes to the builtin varyings, the pixel shader reads from them.
3201 * Take care about the texcoord .w fixup though if we're using the fixed function fragment pipeline
3203 device
= (IWineD3DDeviceImpl
*) vs
->baseShader
.device
;
3204 if((GLINFO_LOCATION
).set_texcoord_w
&& ps_major
== 0 && vs_major
> 0 &&
3205 !device
->frag_pipe
->ffp_proj_control
) {
3206 shader_addline(&buffer
, "void order_ps_input() {\n");
3207 for(i
= 0; i
< min(8, MAX_REG_TEXCRD
); i
++) {
3208 if(vs
->baseShader
.reg_maps
.texcoord_mask
[i
] != 0 &&
3209 vs
->baseShader
.reg_maps
.texcoord_mask
[i
] != WINED3DSP_WRITEMASK_ALL
) {
3210 shader_addline(&buffer
, "gl_TexCoord[%u].w = 1.0;\n", i
);
3213 shader_addline(&buffer
, "}\n");
3215 shader_addline(&buffer
, "void order_ps_input() { /* do nothing */ }\n");
3217 } else if(ps_major
< 3 && vs_major
>= 3) {
3218 /* The vertex shader writes to its own varyings, the pixel shader needs them in the builtin ones */
3219 semantics_out
= vs
->semantics_out
;
3221 shader_addline(&buffer
, "void order_ps_input(in vec4 OUT[%u]) {\n", MAX_REG_OUTPUT
);
3222 for(i
= 0; i
< MAX_REG_OUTPUT
; i
++) {
3223 usage_token
= semantics_out
[i
].usage
;
3224 if (!usage_token
) continue;
3225 register_token
= semantics_out
[i
].reg
;
3227 usage
= (usage_token
& WINED3DSP_DCL_USAGE_MASK
) >> WINED3DSP_DCL_USAGE_SHIFT
;
3228 usage_idx
= (usage_token
& WINED3DSP_DCL_USAGEINDEX_MASK
) >> WINED3DSP_DCL_USAGEINDEX_SHIFT
;
3229 writemask
= shader_glsl_get_write_mask(register_token
, reg_mask
);
3232 case WINED3DDECLUSAGE_COLOR
:
3234 shader_addline(&buffer
, "gl_FrontColor%s = OUT[%u]%s;\n", reg_mask
, i
, reg_mask
);
3235 else if (usage_idx
== 1)
3236 shader_addline(&buffer
, "gl_FrontSecondaryColor%s = OUT[%u]%s;\n", reg_mask
, i
, reg_mask
);
3239 case WINED3DDECLUSAGE_POSITION
:
3240 shader_addline(&buffer
, "gl_Position%s = OUT[%u]%s;\n", reg_mask
, i
, reg_mask
);
3243 case WINED3DDECLUSAGE_TEXCOORD
:
3244 if (usage_idx
< 8) {
3245 if(!(GLINFO_LOCATION
).set_texcoord_w
|| ps_major
> 0) writemask
|= WINED3DSP_WRITEMASK_3
;
3247 shader_addline(&buffer
, "gl_TexCoord[%u]%s = OUT[%u]%s;\n",
3248 usage_idx
, reg_mask
, i
, reg_mask
);
3249 if(!(writemask
& WINED3DSP_WRITEMASK_3
)) {
3250 shader_addline(&buffer
, "gl_TexCoord[%u].w = 1.0;\n", usage_idx
);
3255 case WINED3DDECLUSAGE_PSIZE
:
3256 shader_addline(&buffer
, "gl_PointSize = OUT[%u].x;\n", i
);
3259 case WINED3DDECLUSAGE_FOG
:
3260 shader_addline(&buffer
, "gl_FogFragCoord = OUT[%u].%c;\n", i
, reg_mask
[1]);
3267 shader_addline(&buffer
, "}\n");
3269 } else if(ps_major
>= 3 && vs_major
>= 3) {
3270 semantics_out
= vs
->semantics_out
;
3271 semantics_in
= ps
->semantics_in
;
3273 /* This one is tricky: a 3.0 pixel shader reads from a 3.0 vertex shader */
3274 shader_addline(&buffer
, "varying vec4 IN[%u];\n", GL_LIMITS(glsl_varyings
) / 4);
3275 shader_addline(&buffer
, "void order_ps_input(in vec4 OUT[%u]) {\n", MAX_REG_OUTPUT
);
3277 /* First, sort out position and point size. Those are not passed to the pixel shader */
3278 for(i
= 0; i
< MAX_REG_OUTPUT
; i
++) {
3279 usage_token
= semantics_out
[i
].usage
;
3280 if (!usage_token
) continue;
3281 register_token
= semantics_out
[i
].reg
;
3283 usage
= (usage_token
& WINED3DSP_DCL_USAGE_MASK
) >> WINED3DSP_DCL_USAGE_SHIFT
;
3284 usage_idx
= (usage_token
& WINED3DSP_DCL_USAGEINDEX_MASK
) >> WINED3DSP_DCL_USAGEINDEX_SHIFT
;
3285 shader_glsl_get_write_mask(register_token
, reg_mask
);
3288 case WINED3DDECLUSAGE_POSITION
:
3289 shader_addline(&buffer
, "gl_Position%s = OUT[%u]%s;\n", reg_mask
, i
, reg_mask
);
3292 case WINED3DDECLUSAGE_PSIZE
:
3293 shader_addline(&buffer
, "gl_PointSize = OUT[%u].x;\n", i
);
3301 /* Then, fix the pixel shader input */
3302 handle_ps3_input(&buffer
, semantics_in
, semantics_out
, gl_info
, ps
->input_reg_map
);
3304 shader_addline(&buffer
, "}\n");
3305 } else if(ps_major
>= 3 && vs_major
< 3) {
3306 semantics_in
= ps
->semantics_in
;
3308 shader_addline(&buffer
, "varying vec4 IN[%u];\n", GL_LIMITS(glsl_varyings
) / 4);
3309 shader_addline(&buffer
, "void order_ps_input() {\n");
3310 /* The vertex shader wrote to the builtin varyings. There is no need to figure out position and
3311 * point size, but we depend on the optimizers kindness to find out that the pixel shader doesn't
3312 * read gl_TexCoord and gl_ColorX, otherwise we'll run out of varyings
3314 handle_ps3_input(&buffer
, semantics_in
, NULL
, gl_info
, ps
->input_reg_map
);
3315 shader_addline(&buffer
, "}\n");
3317 ERR("Unexpected vertex and pixel shader version condition: vs: %d, ps: %d\n", vs_major
, ps_major
);
3320 ret
= GL_EXTCALL(glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB
));
3321 checkGLcall("glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB)");
3322 GL_EXTCALL(glShaderSourceARB(ret
, 1, (const char**)&buffer
.buffer
, NULL
));
3323 checkGLcall("glShaderSourceARB(ret, 1, &buffer.buffer, NULL)");
3324 GL_EXTCALL(glCompileShaderARB(ret
));
3325 checkGLcall("glCompileShaderARB(ret)");
3327 shader_buffer_free(&buffer
);
3331 static void hardcode_local_constants(IWineD3DBaseShaderImpl
*shader
, const WineD3D_GL_Info
*gl_info
,
3332 GLhandleARB programId
, char prefix
)
3334 const local_constant
*lconst
;
3339 LIST_FOR_EACH_ENTRY(lconst
, &shader
->baseShader
.constantsF
, local_constant
, entry
) {
3340 value
= (const float *)lconst
->value
;
3341 snprintf(glsl_name
, sizeof(glsl_name
), "%cLC%u", prefix
, lconst
->idx
);
3342 tmp_loc
= GL_EXTCALL(glGetUniformLocationARB(programId
, glsl_name
));
3343 GL_EXTCALL(glUniform4fvARB(tmp_loc
, 1, value
));
3345 checkGLcall("Hardcoding local constants\n");
3348 /** Sets the GLSL program ID for the given pixel and vertex shader combination.
3349 * It sets the programId on the current StateBlock (because it should be called
3350 * inside of the DrawPrimitive() part of the render loop).
3352 * If a program for the given combination does not exist, create one, and store
3353 * the program in the hash table. If it creates a program, it will link the
3354 * given objects, too.
3356 static void set_glsl_shader_program(IWineD3DDevice
*iface
, BOOL use_ps
, BOOL use_vs
) {
3357 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
3358 struct shader_glsl_priv
*priv
= This
->shader_priv
;
3359 const WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
3360 IWineD3DPixelShader
*pshader
= This
->stateBlock
->pixelShader
;
3361 IWineD3DVertexShader
*vshader
= This
->stateBlock
->vertexShader
;
3362 struct glsl_shader_prog_link
*entry
= NULL
;
3363 GLhandleARB programId
= 0;
3364 GLhandleARB reorder_shader_id
= 0;
3367 GLhandleARB vshader_id
, pshader_id
;
3368 struct ps_compile_args ps_compile_args
;
3369 struct vs_compile_args vs_compile_args
;
3372 find_vs_compile_args((IWineD3DVertexShaderImpl
*)This
->stateBlock
->vertexShader
, This
->stateBlock
, &vs_compile_args
);
3374 /* FIXME: Do we really have to spend CPU cycles to generate a few zeroed bytes? */
3375 memset(&vs_compile_args
, 0, sizeof(vs_compile_args
));
3378 find_ps_compile_args((IWineD3DPixelShaderImpl
*)This
->stateBlock
->pixelShader
, This
->stateBlock
, &ps_compile_args
);
3380 /* FIXME: Do we really have to spend CPU cycles to generate a few zeroed bytes? */
3381 memset(&ps_compile_args
, 0, sizeof(ps_compile_args
));
3383 entry
= get_glsl_program_entry(priv
, vshader
, pshader
, &vs_compile_args
, &ps_compile_args
);
3385 priv
->glsl_program
= entry
;
3389 /* If we get to this point, then no matching program exists, so we create one */
3390 programId
= GL_EXTCALL(glCreateProgramObjectARB());
3391 TRACE("Created new GLSL shader program %u\n", programId
);
3393 /* Create the entry */
3394 entry
= HeapAlloc(GetProcessHeap(), 0, sizeof(struct glsl_shader_prog_link
));
3395 entry
->programId
= programId
;
3396 entry
->vshader
= vshader
;
3397 entry
->pshader
= pshader
;
3398 entry
->vs_args
= vs_compile_args
;
3399 entry
->ps_args
= ps_compile_args
;
3400 entry
->constant_version
= 0;
3401 /* Add the hash table entry */
3402 add_glsl_program_entry(priv
, entry
);
3404 /* Set the current program */
3405 priv
->glsl_program
= entry
;
3408 vshader_id
= find_gl_vshader((IWineD3DVertexShaderImpl
*) vshader
, &vs_compile_args
);
3413 /* Attach GLSL vshader */
3415 int max_attribs
= 16; /* TODO: Will this always be the case? It is at the moment... */
3418 reorder_shader_id
= generate_param_reorder_function(vshader
, pshader
, gl_info
);
3419 TRACE("Attaching GLSL shader object %u to program %u\n", reorder_shader_id
, programId
);
3420 GL_EXTCALL(glAttachObjectARB(programId
, reorder_shader_id
));
3421 checkGLcall("glAttachObjectARB");
3422 /* Flag the reorder function for deletion, then it will be freed automatically when the program
3425 GL_EXTCALL(glDeleteObjectARB(reorder_shader_id
));
3427 TRACE("Attaching GLSL shader object %u to program %u\n", vshader_id
, programId
);
3428 GL_EXTCALL(glAttachObjectARB(programId
, vshader_id
));
3429 checkGLcall("glAttachObjectARB");
3431 /* Bind vertex attributes to a corresponding index number to match
3432 * the same index numbers as ARB_vertex_programs (makes loading
3433 * vertex attributes simpler). With this method, we can use the
3434 * exact same code to load the attributes later for both ARB and
3437 * We have to do this here because we need to know the Program ID
3438 * in order to make the bindings work, and it has to be done prior
3439 * to linking the GLSL program. */
3440 for (i
= 0; i
< max_attribs
; ++i
) {
3441 if (((IWineD3DBaseShaderImpl
*)vshader
)->baseShader
.reg_maps
.attributes
[i
]) {
3442 snprintf(tmp_name
, sizeof(tmp_name
), "attrib%i", i
);
3443 GL_EXTCALL(glBindAttribLocationARB(programId
, i
, tmp_name
));
3446 checkGLcall("glBindAttribLocationARB");
3448 list_add_head(&((IWineD3DBaseShaderImpl
*)vshader
)->baseShader
.linked_programs
, &entry
->vshader_entry
);
3452 pshader_id
= find_gl_pshader((IWineD3DPixelShaderImpl
*) pshader
, &ps_compile_args
);
3457 /* Attach GLSL pshader */
3459 TRACE("Attaching GLSL shader object %u to program %u\n", pshader_id
, programId
);
3460 GL_EXTCALL(glAttachObjectARB(programId
, pshader_id
));
3461 checkGLcall("glAttachObjectARB");
3463 list_add_head(&((IWineD3DBaseShaderImpl
*)pshader
)->baseShader
.linked_programs
, &entry
->pshader_entry
);
3466 /* Link the program */
3467 TRACE("Linking GLSL shader program %u\n", programId
);
3468 GL_EXTCALL(glLinkProgramARB(programId
));
3469 print_glsl_info_log(&GLINFO_LOCATION
, programId
);
3471 entry
->vuniformF_locations
= HeapAlloc(GetProcessHeap(), 0, sizeof(GLhandleARB
) * GL_LIMITS(vshader_constantsF
));
3472 for (i
= 0; i
< GL_LIMITS(vshader_constantsF
); ++i
) {
3473 snprintf(glsl_name
, sizeof(glsl_name
), "VC[%i]", i
);
3474 entry
->vuniformF_locations
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, glsl_name
));
3476 for (i
= 0; i
< MAX_CONST_I
; ++i
) {
3477 snprintf(glsl_name
, sizeof(glsl_name
), "VI[%i]", i
);
3478 entry
->vuniformI_locations
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, glsl_name
));
3480 entry
->puniformF_locations
= HeapAlloc(GetProcessHeap(), 0, sizeof(GLhandleARB
) * GL_LIMITS(pshader_constantsF
));
3481 for (i
= 0; i
< GL_LIMITS(pshader_constantsF
); ++i
) {
3482 snprintf(glsl_name
, sizeof(glsl_name
), "PC[%i]", i
);
3483 entry
->puniformF_locations
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, glsl_name
));
3485 for (i
= 0; i
< MAX_CONST_I
; ++i
) {
3486 snprintf(glsl_name
, sizeof(glsl_name
), "PI[%i]", i
);
3487 entry
->puniformI_locations
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, glsl_name
));
3491 for(i
= 0; i
< ((IWineD3DPixelShaderImpl
*)pshader
)->numbumpenvmatconsts
; i
++) {
3493 sprintf(name
, "bumpenvmat%d", ((IWineD3DPixelShaderImpl
*)pshader
)->bumpenvmatconst
[i
].texunit
);
3494 entry
->bumpenvmat_location
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, name
));
3495 sprintf(name
, "luminancescale%d", ((IWineD3DPixelShaderImpl
*)pshader
)->luminanceconst
[i
].texunit
);
3496 entry
->luminancescale_location
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, name
));
3497 sprintf(name
, "luminanceoffset%d", ((IWineD3DPixelShaderImpl
*)pshader
)->luminanceconst
[i
].texunit
);
3498 entry
->luminanceoffset_location
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, name
));
3503 entry
->posFixup_location
= GL_EXTCALL(glGetUniformLocationARB(programId
, "posFixup"));
3504 entry
->ycorrection_location
= GL_EXTCALL(glGetUniformLocationARB(programId
, "ycorrection"));
3505 checkGLcall("Find glsl program uniform locations");
3508 && WINED3DSHADER_VERSION_MAJOR(((IWineD3DPixelShaderImpl
*)pshader
)->baseShader
.reg_maps
.shader_version
) >= 3
3509 && ((IWineD3DPixelShaderImpl
*)pshader
)->declared_in_count
> GL_LIMITS(glsl_varyings
) / 4)
3511 TRACE("Shader %d needs vertex color clamping disabled\n", programId
);
3512 entry
->vertex_color_clamp
= GL_FALSE
;
3514 entry
->vertex_color_clamp
= GL_FIXED_ONLY_ARB
;
3517 /* Set the shader to allow uniform loading on it */
3518 GL_EXTCALL(glUseProgramObjectARB(programId
));
3519 checkGLcall("glUseProgramObjectARB(programId)");
3521 /* Load the vertex and pixel samplers now. The function that finds the mappings makes sure
3522 * that it stays the same for each vertexshader-pixelshader pair(=linked glsl program). If
3523 * a pshader with fixed function pipeline is used there are no vertex samplers, and if a
3524 * vertex shader with fixed function pixel processing is used we make sure that the card
3525 * supports enough samplers to allow the max number of vertex samplers with all possible
3526 * fixed function fragment processing setups. So once the program is linked these samplers
3530 /* Load vertex shader samplers */
3531 shader_glsl_load_vsamplers(gl_info
, (IWineD3DStateBlock
*)This
->stateBlock
, programId
);
3534 /* Load pixel shader samplers */
3535 shader_glsl_load_psamplers(gl_info
, (IWineD3DStateBlock
*)This
->stateBlock
, programId
);
3538 /* If the local constants do not have to be loaded with the environment constants,
3539 * load them now to have them hardcoded in the GLSL program. This saves some CPU cycles
3542 if(pshader
&& !((IWineD3DPixelShaderImpl
*)pshader
)->baseShader
.load_local_constsF
) {
3543 hardcode_local_constants((IWineD3DBaseShaderImpl
*) pshader
, gl_info
, programId
, 'P');
3545 if(vshader
&& !((IWineD3DVertexShaderImpl
*)vshader
)->baseShader
.load_local_constsF
) {
3546 hardcode_local_constants((IWineD3DBaseShaderImpl
*) vshader
, gl_info
, programId
, 'V');
3550 static GLhandleARB
create_glsl_blt_shader(const WineD3D_GL_Info
*gl_info
, enum tex_types tex_type
)
3552 GLhandleARB program_id
;
3553 GLhandleARB vshader_id
, pshader_id
;
3554 static const char *blt_vshader
[] =
3559 " gl_Position = gl_Vertex;\n"
3560 " gl_FrontColor = vec4(1.0);\n"
3561 " gl_TexCoord[0] = gl_MultiTexCoord0;\n"
3565 static const char *blt_pshaders
[tex_type_count
] =
3571 "uniform sampler2D sampler;\n"
3574 " gl_FragDepth = texture2D(sampler, gl_TexCoord[0].xy).x;\n"
3580 "uniform samplerCube sampler;\n"
3583 " gl_FragDepth = textureCube(sampler, gl_TexCoord[0].xyz).x;\n"
3587 "#extension GL_ARB_texture_rectangle : enable\n"
3588 "uniform sampler2DRect sampler;\n"
3591 " gl_FragDepth = texture2DRect(sampler, gl_TexCoord[0].xy).x;\n"
3595 if (!blt_pshaders
[tex_type
])
3597 FIXME("tex_type %#x not supported\n", tex_type
);
3601 vshader_id
= GL_EXTCALL(glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB
));
3602 GL_EXTCALL(glShaderSourceARB(vshader_id
, 1, blt_vshader
, NULL
));
3603 GL_EXTCALL(glCompileShaderARB(vshader_id
));
3605 pshader_id
= GL_EXTCALL(glCreateShaderObjectARB(GL_FRAGMENT_SHADER_ARB
));
3606 GL_EXTCALL(glShaderSourceARB(pshader_id
, 1, &blt_pshaders
[tex_type
], NULL
));
3607 GL_EXTCALL(glCompileShaderARB(pshader_id
));
3609 program_id
= GL_EXTCALL(glCreateProgramObjectARB());
3610 GL_EXTCALL(glAttachObjectARB(program_id
, vshader_id
));
3611 GL_EXTCALL(glAttachObjectARB(program_id
, pshader_id
));
3612 GL_EXTCALL(glLinkProgramARB(program_id
));
3614 print_glsl_info_log(&GLINFO_LOCATION
, program_id
);
3616 /* Once linked we can mark the shaders for deletion. They will be deleted once the program
3619 GL_EXTCALL(glDeleteObjectARB(vshader_id
));
3620 GL_EXTCALL(glDeleteObjectARB(pshader_id
));
3624 static void shader_glsl_select(IWineD3DDevice
*iface
, BOOL usePS
, BOOL useVS
) {
3625 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
3626 struct shader_glsl_priv
*priv
= This
->shader_priv
;
3627 const WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
3628 GLhandleARB program_id
= 0;
3629 GLenum old_vertex_color_clamp
, current_vertex_color_clamp
;
3631 old_vertex_color_clamp
= priv
->glsl_program
? priv
->glsl_program
->vertex_color_clamp
: GL_FIXED_ONLY_ARB
;
3633 if (useVS
|| usePS
) set_glsl_shader_program(iface
, usePS
, useVS
);
3634 else priv
->glsl_program
= NULL
;
3636 current_vertex_color_clamp
= priv
->glsl_program
? priv
->glsl_program
->vertex_color_clamp
: GL_FIXED_ONLY_ARB
;
3638 if (old_vertex_color_clamp
!= current_vertex_color_clamp
) {
3639 if (GL_SUPPORT(ARB_COLOR_BUFFER_FLOAT
)) {
3640 GL_EXTCALL(glClampColorARB(GL_CLAMP_VERTEX_COLOR_ARB
, current_vertex_color_clamp
));
3641 checkGLcall("glClampColorARB");
3643 FIXME("vertex color clamp needs to be changed, but extension not supported.\n");
3647 program_id
= priv
->glsl_program
? priv
->glsl_program
->programId
: 0;
3648 if (program_id
) TRACE("Using GLSL program %u\n", program_id
);
3649 GL_EXTCALL(glUseProgramObjectARB(program_id
));
3650 checkGLcall("glUseProgramObjectARB");
3653 static void shader_glsl_select_depth_blt(IWineD3DDevice
*iface
, enum tex_types tex_type
) {
3654 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
3655 const WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
3656 struct shader_glsl_priv
*priv
= This
->shader_priv
;
3657 GLhandleARB
*blt_program
= &priv
->depth_blt_program
[tex_type
];
3659 if (!*blt_program
) {
3661 *blt_program
= create_glsl_blt_shader(gl_info
, tex_type
);
3662 loc
= GL_EXTCALL(glGetUniformLocationARB(*blt_program
, "sampler"));
3663 GL_EXTCALL(glUseProgramObjectARB(*blt_program
));
3664 GL_EXTCALL(glUniform1iARB(loc
, 0));
3666 GL_EXTCALL(glUseProgramObjectARB(*blt_program
));
3670 static void shader_glsl_deselect_depth_blt(IWineD3DDevice
*iface
) {
3671 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
3672 const WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
3673 struct shader_glsl_priv
*priv
= This
->shader_priv
;
3674 GLhandleARB program_id
;
3676 program_id
= priv
->glsl_program
? priv
->glsl_program
->programId
: 0;
3677 if (program_id
) TRACE("Using GLSL program %u\n", program_id
);
3679 GL_EXTCALL(glUseProgramObjectARB(program_id
));
3680 checkGLcall("glUseProgramObjectARB");
3683 static void shader_glsl_destroy(IWineD3DBaseShader
*iface
) {
3684 const struct list
*linked_programs
;
3685 IWineD3DBaseShaderImpl
*This
= (IWineD3DBaseShaderImpl
*) iface
;
3686 IWineD3DDeviceImpl
*device
= (IWineD3DDeviceImpl
*)This
->baseShader
.device
;
3687 struct shader_glsl_priv
*priv
= device
->shader_priv
;
3688 const WineD3D_GL_Info
*gl_info
= &device
->adapter
->gl_info
;
3689 IWineD3DPixelShaderImpl
*ps
= NULL
;
3690 IWineD3DVertexShaderImpl
*vs
= NULL
;
3692 /* Note: Do not use QueryInterface here to find out which shader type this is because this code
3693 * can be called from IWineD3DBaseShader::Release
3695 char pshader
= shader_is_pshader_version(This
->baseShader
.reg_maps
.shader_version
);
3698 ps
= (IWineD3DPixelShaderImpl
*) This
;
3699 if(ps
->num_gl_shaders
== 0) return;
3701 vs
= (IWineD3DVertexShaderImpl
*) This
;
3702 if(vs
->num_gl_shaders
== 0) return;
3705 linked_programs
= &This
->baseShader
.linked_programs
;
3707 TRACE("Deleting linked programs\n");
3708 if (linked_programs
->next
) {
3709 struct glsl_shader_prog_link
*entry
, *entry2
;
3712 LIST_FOR_EACH_ENTRY_SAFE(entry
, entry2
, linked_programs
, struct glsl_shader_prog_link
, pshader_entry
) {
3713 delete_glsl_program_entry(priv
, gl_info
, entry
);
3716 LIST_FOR_EACH_ENTRY_SAFE(entry
, entry2
, linked_programs
, struct glsl_shader_prog_link
, vshader_entry
) {
3717 delete_glsl_program_entry(priv
, gl_info
, entry
);
3726 for(i
= 0; i
< ps
->num_gl_shaders
; i
++) {
3727 TRACE("deleting pshader %u\n", ps
->gl_shaders
[i
].prgId
);
3728 GL_EXTCALL(glDeleteObjectARB(ps
->gl_shaders
[i
].prgId
));
3729 checkGLcall("glDeleteObjectARB");
3732 HeapFree(GetProcessHeap(), 0, ps
->gl_shaders
);
3733 ps
->gl_shaders
= NULL
;
3734 ps
->num_gl_shaders
= 0;
3735 ps
->shader_array_size
= 0;
3740 for(i
= 0; i
< vs
->num_gl_shaders
; i
++) {
3741 TRACE("deleting vshader %u\n", vs
->gl_shaders
[i
].prgId
);
3742 GL_EXTCALL(glDeleteObjectARB(vs
->gl_shaders
[i
].prgId
));
3743 checkGLcall("glDeleteObjectARB");
3746 HeapFree(GetProcessHeap(), 0, vs
->gl_shaders
);
3747 vs
->gl_shaders
= NULL
;
3748 vs
->num_gl_shaders
= 0;
3749 vs
->shader_array_size
= 0;
3753 static unsigned int glsl_program_key_hash(const void *key
)
3755 const glsl_program_key_t
*k
= key
;
3757 unsigned int hash
= ((DWORD_PTR
) k
->vshader
) | ((DWORD_PTR
) k
->pshader
) << 16;
3758 hash
+= ~(hash
<< 15);
3759 hash
^= (hash
>> 10);
3760 hash
+= (hash
<< 3);
3761 hash
^= (hash
>> 6);
3762 hash
+= ~(hash
<< 11);
3763 hash
^= (hash
>> 16);
3768 static BOOL
glsl_program_key_compare(const void *keya
, const void *keyb
)
3770 const glsl_program_key_t
*ka
= keya
;
3771 const glsl_program_key_t
*kb
= keyb
;
3773 return ka
->vshader
== kb
->vshader
&& ka
->pshader
== kb
->pshader
&&
3774 (memcmp(&ka
->ps_args
, &kb
->ps_args
, sizeof(kb
->ps_args
)) == 0) &&
3775 (memcmp(&ka
->vs_args
, &kb
->vs_args
, sizeof(kb
->vs_args
)) == 0);
3778 static BOOL
constant_heap_init(struct constant_heap
*heap
, unsigned int constant_count
)
3780 SIZE_T size
= (constant_count
+ 1) * sizeof(*heap
->entries
) + constant_count
* sizeof(*heap
->positions
);
3781 void *mem
= HeapAlloc(GetProcessHeap(), 0, size
);
3785 ERR("Failed to allocate memory\n");
3789 heap
->entries
= mem
;
3790 heap
->entries
[1].version
= 0;
3791 heap
->positions
= (unsigned int *)(heap
->entries
+ constant_count
+ 1);
3797 static void constant_heap_free(struct constant_heap
*heap
)
3799 HeapFree(GetProcessHeap(), 0, heap
->entries
);
3802 static HRESULT
shader_glsl_alloc(IWineD3DDevice
*iface
) {
3803 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
3804 const WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
3805 struct shader_glsl_priv
*priv
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(struct shader_glsl_priv
));
3806 SIZE_T stack_size
= wined3d_log2i(max(GL_LIMITS(vshader_constantsF
), GL_LIMITS(pshader_constantsF
))) + 1;
3808 priv
->stack
= HeapAlloc(GetProcessHeap(), 0, stack_size
* sizeof(*priv
->stack
));
3811 ERR("Failed to allocate memory.\n");
3812 HeapFree(GetProcessHeap(), 0, priv
);
3813 return E_OUTOFMEMORY
;
3816 if (!constant_heap_init(&priv
->vconst_heap
, GL_LIMITS(vshader_constantsF
)))
3818 ERR("Failed to initialize vertex shader constant heap\n");
3819 HeapFree(GetProcessHeap(), 0, priv
->stack
);
3820 HeapFree(GetProcessHeap(), 0, priv
);
3821 return E_OUTOFMEMORY
;
3824 if (!constant_heap_init(&priv
->pconst_heap
, GL_LIMITS(pshader_constantsF
)))
3826 ERR("Failed to initialize pixel shader constant heap\n");
3827 constant_heap_free(&priv
->vconst_heap
);
3828 HeapFree(GetProcessHeap(), 0, priv
->stack
);
3829 HeapFree(GetProcessHeap(), 0, priv
);
3830 return E_OUTOFMEMORY
;
3833 priv
->glsl_program_lookup
= hash_table_create(glsl_program_key_hash
, glsl_program_key_compare
);
3834 priv
->next_constant_version
= 1;
3836 This
->shader_priv
= priv
;
3840 static void shader_glsl_free(IWineD3DDevice
*iface
) {
3841 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
3842 const WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
3843 struct shader_glsl_priv
*priv
= This
->shader_priv
;
3846 for (i
= 0; i
< tex_type_count
; ++i
)
3848 if (priv
->depth_blt_program
[i
])
3850 GL_EXTCALL(glDeleteObjectARB(priv
->depth_blt_program
[i
]));
3854 hash_table_destroy(priv
->glsl_program_lookup
, NULL
, NULL
);
3855 constant_heap_free(&priv
->pconst_heap
);
3856 constant_heap_free(&priv
->vconst_heap
);
3858 HeapFree(GetProcessHeap(), 0, This
->shader_priv
);
3859 This
->shader_priv
= NULL
;
3862 static BOOL
shader_glsl_dirty_const(IWineD3DDevice
*iface
) {
3863 /* TODO: GL_EXT_bindable_uniform can be used to share constants across shaders */
3867 static GLuint
shader_glsl_generate_pshader(IWineD3DPixelShader
*iface
, SHADER_BUFFER
*buffer
, const struct ps_compile_args
*args
) {
3868 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*)iface
;
3869 const struct shader_reg_maps
*reg_maps
= &This
->baseShader
.reg_maps
;
3870 CONST DWORD
*function
= This
->baseShader
.function
;
3871 const char *fragcolor
;
3872 const WineD3D_GL_Info
*gl_info
= &((IWineD3DDeviceImpl
*)This
->baseShader
.device
)->adapter
->gl_info
;
3874 /* Create the hw GLSL shader object and assign it as the shader->prgId */
3875 GLhandleARB shader_obj
= GL_EXTCALL(glCreateShaderObjectARB(GL_FRAGMENT_SHADER_ARB
));
3877 shader_addline(buffer
, "#version 120\n");
3879 if (GL_SUPPORT(ARB_DRAW_BUFFERS
)) {
3880 shader_addline(buffer
, "#extension GL_ARB_draw_buffers : enable\n");
3882 if (GL_SUPPORT(ARB_TEXTURE_RECTANGLE
)) {
3883 /* The spec says that it doesn't have to be explicitly enabled, but the nvidia
3884 * drivers write a warning if we don't do so
3886 shader_addline(buffer
, "#extension GL_ARB_texture_rectangle : enable\n");
3889 /* Base Declarations */
3890 shader_generate_glsl_declarations( (IWineD3DBaseShader
*) This
, reg_maps
, buffer
, &GLINFO_LOCATION
, args
);
3892 /* Pack 3.0 inputs */
3893 if (reg_maps
->shader_version
>= WINED3DPS_VERSION(3,0) && args
->vp_mode
!= vertexshader
) {
3894 pshader_glsl_input_pack(buffer
, This
->semantics_in
, iface
, args
->vp_mode
);
3897 /* Base Shader Body */
3898 shader_generate_main( (IWineD3DBaseShader
*) This
, buffer
, reg_maps
, function
);
3900 /* Pixel shaders < 2.0 place the resulting color in R0 implicitly */
3901 if (reg_maps
->shader_version
< WINED3DPS_VERSION(2,0))
3903 /* Some older cards like GeforceFX ones don't support multiple buffers, so also not gl_FragData */
3904 if(GL_SUPPORT(ARB_DRAW_BUFFERS
))
3905 shader_addline(buffer
, "gl_FragData[0] = R0;\n");
3907 shader_addline(buffer
, "gl_FragColor = R0;\n");
3910 if(GL_SUPPORT(ARB_DRAW_BUFFERS
)) {
3911 fragcolor
= "gl_FragData[0]";
3913 fragcolor
= "gl_FragColor";
3915 if(args
->srgb_correction
) {
3916 shader_addline(buffer
, "tmp0.xyz = pow(%s.xyz, vec3(%f, %f, %f)) * vec3(%f, %f, %f) - vec3(%f, %f, %f);\n",
3917 fragcolor
, srgb_pow
, srgb_pow
, srgb_pow
, srgb_mul_high
, srgb_mul_high
, srgb_mul_high
,
3918 srgb_sub_high
, srgb_sub_high
, srgb_sub_high
);
3919 shader_addline(buffer
, "tmp1.xyz = %s.xyz * srgb_mul_low.xyz;\n", fragcolor
);
3920 shader_addline(buffer
, "%s.x = %s.x < srgb_comparison.x ? tmp1.x : tmp0.x;\n", fragcolor
, fragcolor
);
3921 shader_addline(buffer
, "%s.y = %s.y < srgb_comparison.y ? tmp1.y : tmp0.y;\n", fragcolor
, fragcolor
);
3922 shader_addline(buffer
, "%s.z = %s.z < srgb_comparison.z ? tmp1.z : tmp0.z;\n", fragcolor
, fragcolor
);
3923 shader_addline(buffer
, "%s = clamp(%s, 0.0, 1.0);\n", fragcolor
, fragcolor
);
3925 /* Pixel shader < 3.0 do not replace the fog stage.
3926 * This implements linear fog computation and blending.
3927 * TODO: non linear fog
3928 * NOTE: gl_Fog.start and gl_Fog.end don't hold fog start s and end e but
3929 * -1/(e-s) and e/(e-s) respectively.
3931 if(reg_maps
->shader_version
< WINED3DPS_VERSION(3,0)) {
3933 case FOG_OFF
: break;
3935 shader_addline(buffer
, "float fogstart = -1.0 / (gl_Fog.end - gl_Fog.start);\n");
3936 shader_addline(buffer
, "float fogend = gl_Fog.end * -fogstart;\n");
3937 shader_addline(buffer
, "float Fog = clamp(gl_FogFragCoord * fogstart + fogend, 0.0, 1.0);\n");
3938 shader_addline(buffer
, "%s.xyz = mix(gl_Fog.color.xyz, %s.xyz, Fog);\n", fragcolor
, fragcolor
);
3941 /* Fog = e^(-gl_Fog.density * gl_FogFragCoord) */
3942 shader_addline(buffer
, "float Fog = exp(-gl_Fog.density * gl_FogFragCoord);\n");
3943 shader_addline(buffer
, "Fog = clamp(Fog, 0.0, 1.0);\n");
3944 shader_addline(buffer
, "%s.xyz = mix(gl_Fog.color.xyz, %s.xyz, Fog);\n", fragcolor
, fragcolor
);
3947 /* Fog = e^(-(gl_Fog.density * gl_FogFragCoord)^2) */
3948 shader_addline(buffer
, "float Fog = exp(-gl_Fog.density * gl_Fog.density * gl_FogFragCoord * gl_FogFragCoord);\n");
3949 shader_addline(buffer
, "Fog = clamp(Fog, 0.0, 1.0);\n");
3950 shader_addline(buffer
, "%s.xyz = mix(gl_Fog.color.xyz, %s.xyz, Fog);\n", fragcolor
, fragcolor
);
3955 shader_addline(buffer
, "}\n");
3957 TRACE("Compiling shader object %u\n", shader_obj
);
3958 GL_EXTCALL(glShaderSourceARB(shader_obj
, 1, (const char**)&buffer
->buffer
, NULL
));
3959 GL_EXTCALL(glCompileShaderARB(shader_obj
));
3960 print_glsl_info_log(&GLINFO_LOCATION
, shader_obj
);
3962 /* Store the shader object */
3966 static GLuint
shader_glsl_generate_vshader(IWineD3DVertexShader
*iface
, SHADER_BUFFER
*buffer
, const struct vs_compile_args
*args
) {
3967 IWineD3DVertexShaderImpl
*This
= (IWineD3DVertexShaderImpl
*)iface
;
3968 const struct shader_reg_maps
*reg_maps
= &This
->baseShader
.reg_maps
;
3969 CONST DWORD
*function
= This
->baseShader
.function
;
3970 const WineD3D_GL_Info
*gl_info
= &((IWineD3DDeviceImpl
*)This
->baseShader
.device
)->adapter
->gl_info
;
3972 /* Create the hw GLSL shader program and assign it as the shader->prgId */
3973 GLhandleARB shader_obj
= GL_EXTCALL(glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB
));
3975 shader_addline(buffer
, "#version 120\n");
3977 /* Base Declarations */
3978 shader_generate_glsl_declarations( (IWineD3DBaseShader
*) This
, reg_maps
, buffer
, &GLINFO_LOCATION
, NULL
);
3980 /* Base Shader Body */
3981 shader_generate_main( (IWineD3DBaseShader
*) This
, buffer
, reg_maps
, function
);
3983 /* Unpack 3.0 outputs */
3984 if (reg_maps
->shader_version
>= WINED3DVS_VERSION(3,0)) shader_addline(buffer
, "order_ps_input(OUT);\n");
3985 else shader_addline(buffer
, "order_ps_input();\n");
3987 /* The D3DRS_FOGTABLEMODE render state defines if the shader-generated fog coord is used
3988 * or if the fragment depth is used. If the fragment depth is used(FOGTABLEMODE != NONE),
3989 * the fog frag coord is thrown away. If the fog frag coord is used, but not written by
3990 * the shader, it is set to 0.0(fully fogged, since start = 1.0, end = 0.0)
3992 if(args
->fog_src
== VS_FOG_Z
) {
3993 shader_addline(buffer
, "gl_FogFragCoord = gl_Position.z;\n");
3994 } else if (!reg_maps
->fog
) {
3995 shader_addline(buffer
, "gl_FogFragCoord = 0.0;\n");
3998 /* Write the final position.
4000 * OpenGL coordinates specify the center of the pixel while d3d coords specify
4001 * the corner. The offsets are stored in z and w in posFixup. posFixup.y contains
4002 * 1.0 or -1.0 to turn the rendering upside down for offscreen rendering. PosFixup.x
4003 * contains 1.0 to allow a mad.
4005 shader_addline(buffer
, "gl_Position.y = gl_Position.y * posFixup.y;\n");
4006 shader_addline(buffer
, "gl_Position.xy += posFixup.zw * gl_Position.ww;\n");
4008 /* Z coord [0;1]->[-1;1] mapping, see comment in transform_projection in state.c
4010 * Basically we want (in homogeneous coordinates) z = z * 2 - 1. However, shaders are run
4011 * before the homogeneous divide, so we have to take the w into account: z = ((z / w) * 2 - 1) * w,
4012 * which is the same as z = z * 2 - w.
4014 shader_addline(buffer
, "gl_Position.z = gl_Position.z * 2.0 - gl_Position.w;\n");
4016 shader_addline(buffer
, "}\n");
4018 TRACE("Compiling shader object %u\n", shader_obj
);
4019 GL_EXTCALL(glShaderSourceARB(shader_obj
, 1, (const char**)&buffer
->buffer
, NULL
));
4020 GL_EXTCALL(glCompileShaderARB(shader_obj
));
4021 print_glsl_info_log(&GLINFO_LOCATION
, shader_obj
);
4026 static void shader_glsl_get_caps(WINED3DDEVTYPE devtype
, const WineD3D_GL_Info
*gl_info
, struct shader_caps
*pCaps
)
4028 /* Nvidia Geforce6/7 or Ati R4xx/R5xx cards with GLSL support, support VS 3.0 but older Nvidia/Ati
4029 * models with GLSL support only support 2.0. In case of nvidia we can detect VS 2.0 support using
4030 * vs_nv_version which is based on NV_vertex_program.
4031 * For Ati cards there's no way using glsl (it abstracts the lowlevel info away) and also not
4032 * using ARB_vertex_program. It is safe to assume that when a card supports pixel shader 2.0 it
4033 * supports vertex shader 2.0 too and the way around. We can detect ps2.0 using the maximum number
4034 * of native instructions, so use that here. For more info see the pixel shader versioning code below.
4036 if((GLINFO_LOCATION
.vs_nv_version
== VS_VERSION_20
) || (GLINFO_LOCATION
.ps_arb_max_instructions
<= 512))
4037 pCaps
->VertexShaderVersion
= WINED3DVS_VERSION(2,0);
4039 pCaps
->VertexShaderVersion
= WINED3DVS_VERSION(3,0);
4040 TRACE_(d3d_caps
)("Hardware vertex shader version %d.%d enabled (GLSL)\n", (pCaps
->VertexShaderVersion
>> 8) & 0xff, pCaps
->VertexShaderVersion
& 0xff);
4041 pCaps
->MaxVertexShaderConst
= GL_LIMITS(vshader_constantsF
);
4043 /* Older DX9-class videocards (GeforceFX / Radeon >9500/X*00) only support pixel shader 2.0/2.0a/2.0b.
4044 * In OpenGL the extensions related to GLSL abstract lowlevel GL info away which is needed
4045 * to distinguish between 2.0 and 3.0 (and 2.0a/2.0b). In case of Nvidia we use their fragment
4046 * program extensions. On other hardware including ATI GL_ARB_fragment_program offers the info
4047 * in max native instructions. Intel and others also offer the info in this extension but they
4048 * don't support GLSL (at least on Windows).
4050 * PS2.0 requires at least 96 instructions, 2.0a/2.0b go up to 512. Assume that if the number
4051 * of instructions is 512 or less we have to do with ps2.0 hardware.
4052 * 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.
4054 if((GLINFO_LOCATION
.ps_nv_version
== PS_VERSION_20
) || (GLINFO_LOCATION
.ps_arb_max_instructions
<= 512))
4055 pCaps
->PixelShaderVersion
= WINED3DPS_VERSION(2,0);
4057 pCaps
->PixelShaderVersion
= WINED3DPS_VERSION(3,0);
4059 /* FIXME: The following line is card dependent. -8.0 to 8.0 is the
4060 * Direct3D minimum requirement.
4062 * Both GL_ARB_fragment_program and GLSL require a "maximum representable magnitude"
4063 * of colors to be 2^10, and 2^32 for other floats. Should we use 1024 here?
4065 * The problem is that the refrast clamps temporary results in the shader to
4066 * [-MaxValue;+MaxValue]. If the card's max value is bigger than the one we advertize here,
4067 * then applications may miss the clamping behavior. On the other hand, if it is smaller,
4068 * the shader will generate incorrect results too. Unfortunately, GL deliberately doesn't
4069 * offer a way to query this.
4071 pCaps
->PixelShader1xMaxValue
= 8.0;
4072 TRACE_(d3d_caps
)("Hardware pixel shader version %d.%d enabled (GLSL)\n", (pCaps
->PixelShaderVersion
>> 8) & 0xff, pCaps
->PixelShaderVersion
& 0xff);
4075 static BOOL
shader_glsl_color_fixup_supported(struct color_fixup_desc fixup
)
4077 if (TRACE_ON(d3d_shader
) && TRACE_ON(d3d
))
4079 TRACE("Checking support for fixup:\n");
4080 dump_color_fixup_desc(fixup
);
4083 /* We support everything except YUV conversions. */
4084 if (!is_yuv_fixup(fixup
))
4090 TRACE("[FAILED]\n");
4094 static const SHADER_HANDLER shader_glsl_instruction_handler_table
[WINED3DSIH_TABLE_SIZE
] =
4096 /* WINED3DSIH_ABS */ shader_glsl_map2gl
,
4097 /* WINED3DSIH_ADD */ shader_glsl_arith
,
4098 /* WINED3DSIH_BEM */ pshader_glsl_bem
,
4099 /* WINED3DSIH_BREAK */ shader_glsl_break
,
4100 /* WINED3DSIH_BREAKC */ shader_glsl_breakc
,
4101 /* WINED3DSIH_BREAKP */ NULL
,
4102 /* WINED3DSIH_CALL */ shader_glsl_call
,
4103 /* WINED3DSIH_CALLNZ */ shader_glsl_callnz
,
4104 /* WINED3DSIH_CMP */ shader_glsl_cmp
,
4105 /* WINED3DSIH_CND */ shader_glsl_cnd
,
4106 /* WINED3DSIH_CRS */ shader_glsl_cross
,
4107 /* WINED3DSIH_DCL */ NULL
,
4108 /* WINED3DSIH_DEF */ NULL
,
4109 /* WINED3DSIH_DEFB */ NULL
,
4110 /* WINED3DSIH_DEFI */ NULL
,
4111 /* WINED3DSIH_DP2ADD */ pshader_glsl_dp2add
,
4112 /* WINED3DSIH_DP3 */ shader_glsl_dot
,
4113 /* WINED3DSIH_DP4 */ shader_glsl_dot
,
4114 /* WINED3DSIH_DST */ shader_glsl_dst
,
4115 /* WINED3DSIH_DSX */ shader_glsl_map2gl
,
4116 /* WINED3DSIH_DSY */ shader_glsl_map2gl
,
4117 /* WINED3DSIH_ELSE */ shader_glsl_else
,
4118 /* WINED3DSIH_ENDIF */ shader_glsl_end
,
4119 /* WINED3DSIH_ENDLOOP */ shader_glsl_end
,
4120 /* WINED3DSIH_ENDREP */ shader_glsl_end
,
4121 /* WINED3DSIH_EXP */ shader_glsl_map2gl
,
4122 /* WINED3DSIH_EXPP */ shader_glsl_expp
,
4123 /* WINED3DSIH_FRC */ shader_glsl_map2gl
,
4124 /* WINED3DSIH_IF */ shader_glsl_if
,
4125 /* WINED3DSIH_IFC */ shader_glsl_ifc
,
4126 /* WINED3DSIH_LABEL */ shader_glsl_label
,
4127 /* WINED3DSIH_LIT */ shader_glsl_lit
,
4128 /* WINED3DSIH_LOG */ shader_glsl_log
,
4129 /* WINED3DSIH_LOGP */ shader_glsl_log
,
4130 /* WINED3DSIH_LOOP */ shader_glsl_loop
,
4131 /* WINED3DSIH_LRP */ shader_glsl_lrp
,
4132 /* WINED3DSIH_M3x2 */ shader_glsl_mnxn
,
4133 /* WINED3DSIH_M3x3 */ shader_glsl_mnxn
,
4134 /* WINED3DSIH_M3x4 */ shader_glsl_mnxn
,
4135 /* WINED3DSIH_M4x3 */ shader_glsl_mnxn
,
4136 /* WINED3DSIH_M4x4 */ shader_glsl_mnxn
,
4137 /* WINED3DSIH_MAD */ shader_glsl_mad
,
4138 /* WINED3DSIH_MAX */ shader_glsl_map2gl
,
4139 /* WINED3DSIH_MIN */ shader_glsl_map2gl
,
4140 /* WINED3DSIH_MOV */ shader_glsl_mov
,
4141 /* WINED3DSIH_MOVA */ shader_glsl_mov
,
4142 /* WINED3DSIH_MUL */ shader_glsl_arith
,
4143 /* WINED3DSIH_NOP */ NULL
,
4144 /* WINED3DSIH_NRM */ shader_glsl_map2gl
,
4145 /* WINED3DSIH_PHASE */ NULL
,
4146 /* WINED3DSIH_POW */ shader_glsl_pow
,
4147 /* WINED3DSIH_RCP */ shader_glsl_rcp
,
4148 /* WINED3DSIH_REP */ shader_glsl_rep
,
4149 /* WINED3DSIH_RET */ NULL
,
4150 /* WINED3DSIH_RSQ */ shader_glsl_rsq
,
4151 /* WINED3DSIH_SETP */ NULL
,
4152 /* WINED3DSIH_SGE */ shader_glsl_compare
,
4153 /* WINED3DSIH_SGN */ shader_glsl_map2gl
,
4154 /* WINED3DSIH_SINCOS */ shader_glsl_sincos
,
4155 /* WINED3DSIH_SLT */ shader_glsl_compare
,
4156 /* WINED3DSIH_SUB */ shader_glsl_arith
,
4157 /* WINED3DSIH_TEX */ pshader_glsl_tex
,
4158 /* WINED3DSIH_TEXBEM */ pshader_glsl_texbem
,
4159 /* WINED3DSIH_TEXBEML */ pshader_glsl_texbem
,
4160 /* WINED3DSIH_TEXCOORD */ pshader_glsl_texcoord
,
4161 /* WINED3DSIH_TEXDEPTH */ pshader_glsl_texdepth
,
4162 /* WINED3DSIH_TEXDP3 */ pshader_glsl_texdp3
,
4163 /* WINED3DSIH_TEXDP3TEX */ pshader_glsl_texdp3tex
,
4164 /* WINED3DSIH_TEXKILL */ pshader_glsl_texkill
,
4165 /* WINED3DSIH_TEXLDD */ NULL
,
4166 /* WINED3DSIH_TEXLDL */ shader_glsl_texldl
,
4167 /* WINED3DSIH_TEXM3x2DEPTH */ pshader_glsl_texm3x2depth
,
4168 /* WINED3DSIH_TEXM3x2PAD */ pshader_glsl_texm3x2pad
,
4169 /* WINED3DSIH_TEXM3x2TEX */ pshader_glsl_texm3x2tex
,
4170 /* WINED3DSIH_TEXM3x3 */ pshader_glsl_texm3x3
,
4171 /* WINED3DSIH_TEXM3x3DIFF */ NULL
,
4172 /* WINED3DSIH_TEXM3x3PAD */ pshader_glsl_texm3x3pad
,
4173 /* WINED3DSIH_TEXM3x3SPEC */ pshader_glsl_texm3x3spec
,
4174 /* WINED3DSIH_TEXM3x3TEX */ pshader_glsl_texm3x3tex
,
4175 /* WINED3DSIH_TEXM3x3VSPEC */ pshader_glsl_texm3x3vspec
,
4176 /* WINED3DSIH_TEXREG2AR */ pshader_glsl_texreg2ar
,
4177 /* WINED3DSIH_TEXREG2GB */ pshader_glsl_texreg2gb
,
4178 /* WINED3DSIH_TEXREG2RGB */ pshader_glsl_texreg2rgb
,
4181 const shader_backend_t glsl_shader_backend
= {
4182 shader_glsl_instruction_handler_table
,
4184 shader_glsl_select_depth_blt
,
4185 shader_glsl_deselect_depth_blt
,
4186 shader_glsl_update_float_vertex_constants
,
4187 shader_glsl_update_float_pixel_constants
,
4188 shader_glsl_load_constants
,
4189 shader_glsl_destroy
,
4192 shader_glsl_dirty_const
,
4193 shader_glsl_generate_pshader
,
4194 shader_glsl_generate_vshader
,
4195 shader_glsl_get_caps
,
4196 shader_glsl_color_fixup_supported
,