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 int mapped_unit
= stateBlock
->wineD3DDevice
->texUnitMap
[i
];
193 if (mapped_unit
!= -1 && mapped_unit
< GL_LIMITS(fragment_samplers
)) {
194 TRACE("Loading %s for texture %d\n", sampler_name
, mapped_unit
);
195 GL_EXTCALL(glUniform1iARB(name_loc
, mapped_unit
));
196 checkGLcall("glUniform1iARB");
198 ERR("Trying to load sampler %s on unsupported unit %d\n", sampler_name
, mapped_unit
);
204 static void shader_glsl_load_vsamplers(const WineD3D_GL_Info
*gl_info
, IWineD3DStateBlock
*iface
, GLhandleARB programId
)
206 IWineD3DStateBlockImpl
* stateBlock
= (IWineD3DStateBlockImpl
*) iface
;
207 GLhandleARB name_loc
;
208 char sampler_name
[20];
211 for (i
= 0; i
< MAX_VERTEX_SAMPLERS
; ++i
) {
212 snprintf(sampler_name
, sizeof(sampler_name
), "Vsampler%d", i
);
213 name_loc
= GL_EXTCALL(glGetUniformLocationARB(programId
, sampler_name
));
214 if (name_loc
!= -1) {
215 int mapped_unit
= stateBlock
->wineD3DDevice
->texUnitMap
[MAX_FRAGMENT_SAMPLERS
+ i
];
216 if (mapped_unit
!= -1 && mapped_unit
< GL_LIMITS(combined_samplers
)) {
217 TRACE("Loading %s for texture %d\n", sampler_name
, mapped_unit
);
218 GL_EXTCALL(glUniform1iARB(name_loc
, mapped_unit
));
219 checkGLcall("glUniform1iARB");
221 ERR("Trying to load sampler %s on unsupported unit %d\n", sampler_name
, mapped_unit
);
227 static inline void walk_constant_heap(const WineD3D_GL_Info
*gl_info
, const float *constants
,
228 const GLhandleARB
*constant_locations
, const struct constant_heap
*heap
, unsigned char *stack
, DWORD version
)
231 unsigned int heap_idx
= 1;
234 if (heap
->entries
[heap_idx
].version
<= version
) return;
236 idx
= heap
->entries
[heap_idx
].idx
;
237 if (constant_locations
[idx
] != -1) GL_EXTCALL(glUniform4fvARB(constant_locations
[idx
], 1, &constants
[idx
* 4]));
238 stack
[stack_idx
] = HEAP_NODE_TRAVERSE_LEFT
;
240 while (stack_idx
>= 0)
242 /* Note that we fall through to the next case statement. */
243 switch(stack
[stack_idx
])
245 case HEAP_NODE_TRAVERSE_LEFT
:
247 unsigned int left_idx
= heap_idx
<< 1;
248 if (left_idx
< heap
->size
&& heap
->entries
[left_idx
].version
> version
)
251 idx
= heap
->entries
[heap_idx
].idx
;
252 if (constant_locations
[idx
] != -1)
253 GL_EXTCALL(glUniform4fvARB(constant_locations
[idx
], 1, &constants
[idx
* 4]));
255 stack
[stack_idx
++] = HEAP_NODE_TRAVERSE_RIGHT
;
256 stack
[stack_idx
] = HEAP_NODE_TRAVERSE_LEFT
;
261 case HEAP_NODE_TRAVERSE_RIGHT
:
263 unsigned int right_idx
= (heap_idx
<< 1) + 1;
264 if (right_idx
< heap
->size
&& heap
->entries
[right_idx
].version
> version
)
266 heap_idx
= right_idx
;
267 idx
= heap
->entries
[heap_idx
].idx
;
268 if (constant_locations
[idx
] != -1)
269 GL_EXTCALL(glUniform4fvARB(constant_locations
[idx
], 1, &constants
[idx
* 4]));
271 stack
[stack_idx
++] = HEAP_NODE_POP
;
272 stack
[stack_idx
] = HEAP_NODE_TRAVERSE_LEFT
;
285 checkGLcall("walk_constant_heap()");
288 static inline void apply_clamped_constant(const WineD3D_GL_Info
*gl_info
, GLint location
, const GLfloat
*data
)
290 GLfloat clamped_constant
[4];
292 if (location
== -1) return;
294 clamped_constant
[0] = data
[0] < -1.0f
? -1.0f
: data
[0] > 1.0 ? 1.0 : data
[0];
295 clamped_constant
[1] = data
[1] < -1.0f
? -1.0f
: data
[1] > 1.0 ? 1.0 : data
[1];
296 clamped_constant
[2] = data
[2] < -1.0f
? -1.0f
: data
[2] > 1.0 ? 1.0 : data
[2];
297 clamped_constant
[3] = data
[3] < -1.0f
? -1.0f
: data
[3] > 1.0 ? 1.0 : data
[3];
299 GL_EXTCALL(glUniform4fvARB(location
, 1, clamped_constant
));
302 static inline void walk_constant_heap_clamped(const WineD3D_GL_Info
*gl_info
, const float *constants
,
303 const GLhandleARB
*constant_locations
, const struct constant_heap
*heap
, unsigned char *stack
, DWORD version
)
306 unsigned int heap_idx
= 1;
309 if (heap
->entries
[heap_idx
].version
<= version
) return;
311 idx
= heap
->entries
[heap_idx
].idx
;
312 apply_clamped_constant(gl_info
, constant_locations
[idx
], &constants
[idx
* 4]);
313 stack
[stack_idx
] = HEAP_NODE_TRAVERSE_LEFT
;
315 while (stack_idx
>= 0)
317 /* Note that we fall through to the next case statement. */
318 switch(stack
[stack_idx
])
320 case HEAP_NODE_TRAVERSE_LEFT
:
322 unsigned int left_idx
= heap_idx
<< 1;
323 if (left_idx
< heap
->size
&& heap
->entries
[left_idx
].version
> version
)
326 idx
= heap
->entries
[heap_idx
].idx
;
327 apply_clamped_constant(gl_info
, constant_locations
[idx
], &constants
[idx
* 4]);
329 stack
[stack_idx
++] = HEAP_NODE_TRAVERSE_RIGHT
;
330 stack
[stack_idx
] = HEAP_NODE_TRAVERSE_LEFT
;
335 case HEAP_NODE_TRAVERSE_RIGHT
:
337 unsigned int right_idx
= (heap_idx
<< 1) + 1;
338 if (right_idx
< heap
->size
&& heap
->entries
[right_idx
].version
> version
)
340 heap_idx
= right_idx
;
341 idx
= heap
->entries
[heap_idx
].idx
;
342 apply_clamped_constant(gl_info
, constant_locations
[idx
], &constants
[idx
* 4]);
344 stack
[stack_idx
++] = HEAP_NODE_POP
;
345 stack
[stack_idx
] = HEAP_NODE_TRAVERSE_LEFT
;
358 checkGLcall("walk_constant_heap_clamped()");
361 /* Loads floating point constants (aka uniforms) into the currently set GLSL program. */
362 static void shader_glsl_load_constantsF(IWineD3DBaseShaderImpl
*This
, const WineD3D_GL_Info
*gl_info
,
363 const float *constants
, const GLhandleARB
*constant_locations
, const struct constant_heap
*heap
,
364 unsigned char *stack
, UINT version
)
366 const local_constant
*lconst
;
368 /* 1.X pshaders have the constants clamped to [-1;1] implicitly. */
369 if (WINED3DSHADER_VERSION_MAJOR(This
->baseShader
.reg_maps
.shader_version
) == 1
370 && shader_is_pshader_version(This
->baseShader
.reg_maps
.shader_version
))
371 walk_constant_heap_clamped(gl_info
, constants
, constant_locations
, heap
, stack
, version
);
373 walk_constant_heap(gl_info
, constants
, constant_locations
, heap
, stack
, version
);
375 if (!This
->baseShader
.load_local_constsF
)
377 TRACE("No need to load local float constants for this shader\n");
381 /* Immediate constants are clamped to [-1;1] at shader creation time if needed */
382 LIST_FOR_EACH_ENTRY(lconst
, &This
->baseShader
.constantsF
, local_constant
, entry
)
384 GLhandleARB location
= constant_locations
[lconst
->idx
];
385 /* We found this uniform name in the program - go ahead and send the data */
386 if (location
!= -1) GL_EXTCALL(glUniform4fvARB(location
, 1, (const GLfloat
*)lconst
->value
));
388 checkGLcall("glUniform4fvARB()");
391 /* Loads integer constants (aka uniforms) into the currently set GLSL program. */
392 static void shader_glsl_load_constantsI(IWineD3DBaseShaderImpl
*This
, const WineD3D_GL_Info
*gl_info
,
393 const GLhandleARB locations
[MAX_CONST_I
], const int *constants
, WORD constants_set
)
398 for (i
= 0; constants_set
; constants_set
>>= 1, ++i
)
400 if (!(constants_set
& 1)) continue;
402 TRACE_(d3d_constants
)("Loading constants %u: %i, %i, %i, %i\n",
403 i
, constants
[i
*4], constants
[i
*4+1], constants
[i
*4+2], constants
[i
*4+3]);
405 /* We found this uniform name in the program - go ahead and send the data */
406 GL_EXTCALL(glUniform4ivARB(locations
[i
], 1, &constants
[i
*4]));
407 checkGLcall("glUniform4ivARB");
410 /* Load immediate constants */
411 ptr
= list_head(&This
->baseShader
.constantsI
);
413 const struct local_constant
*lconst
= LIST_ENTRY(ptr
, const struct local_constant
, entry
);
414 unsigned int idx
= lconst
->idx
;
415 const GLint
*values
= (const GLint
*)lconst
->value
;
417 TRACE_(d3d_constants
)("Loading local constants %i: %i, %i, %i, %i\n", idx
,
418 values
[0], values
[1], values
[2], values
[3]);
420 /* We found this uniform name in the program - go ahead and send the data */
421 GL_EXTCALL(glUniform4ivARB(locations
[idx
], 1, values
));
422 checkGLcall("glUniform4ivARB");
423 ptr
= list_next(&This
->baseShader
.constantsI
, ptr
);
427 /* Loads boolean constants (aka uniforms) into the currently set GLSL program. */
428 static void shader_glsl_load_constantsB(IWineD3DBaseShaderImpl
*This
, const WineD3D_GL_Info
*gl_info
,
429 GLhandleARB programId
, const BOOL
*constants
, WORD constants_set
)
434 char is_pshader
= shader_is_pshader_version(This
->baseShader
.reg_maps
.shader_version
);
435 const char* prefix
= is_pshader
? "PB":"VB";
438 /* TODO: Benchmark and see if it would be beneficial to store the
439 * locations of the constants to avoid looking up each time */
440 for (i
= 0; constants_set
; constants_set
>>= 1, ++i
)
442 if (!(constants_set
& 1)) continue;
444 TRACE_(d3d_constants
)("Loading constants %i: %i;\n", i
, constants
[i
]);
446 /* TODO: Benchmark and see if it would be beneficial to store the
447 * locations of the constants to avoid looking up each time */
448 snprintf(tmp_name
, sizeof(tmp_name
), "%s[%i]", prefix
, i
);
449 tmp_loc
= GL_EXTCALL(glGetUniformLocationARB(programId
, tmp_name
));
452 /* We found this uniform name in the program - go ahead and send the data */
453 GL_EXTCALL(glUniform1ivARB(tmp_loc
, 1, &constants
[i
]));
454 checkGLcall("glUniform1ivARB");
458 /* Load immediate constants */
459 ptr
= list_head(&This
->baseShader
.constantsB
);
461 const struct local_constant
*lconst
= LIST_ENTRY(ptr
, const struct local_constant
, entry
);
462 unsigned int idx
= lconst
->idx
;
463 const GLint
*values
= (const GLint
*)lconst
->value
;
465 TRACE_(d3d_constants
)("Loading local constants %i: %i\n", idx
, values
[0]);
467 snprintf(tmp_name
, sizeof(tmp_name
), "%s[%i]", prefix
, idx
);
468 tmp_loc
= GL_EXTCALL(glGetUniformLocationARB(programId
, tmp_name
));
470 /* We found this uniform name in the program - go ahead and send the data */
471 GL_EXTCALL(glUniform1ivARB(tmp_loc
, 1, values
));
472 checkGLcall("glUniform1ivARB");
474 ptr
= list_next(&This
->baseShader
.constantsB
, ptr
);
478 static void reset_program_constant_version(void *value
, void *context
)
480 struct glsl_shader_prog_link
*entry
= value
;
481 entry
->constant_version
= 0;
485 * Loads the app-supplied constants into the currently set GLSL program.
487 static void shader_glsl_load_constants(
488 IWineD3DDevice
* device
,
490 char useVertexShader
) {
492 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) device
;
493 struct shader_glsl_priv
*priv
= deviceImpl
->shader_priv
;
494 IWineD3DStateBlockImpl
* stateBlock
= deviceImpl
->stateBlock
;
495 const WineD3D_GL_Info
*gl_info
= &deviceImpl
->adapter
->gl_info
;
497 GLhandleARB programId
;
498 struct glsl_shader_prog_link
*prog
= priv
->glsl_program
;
499 UINT constant_version
;
503 /* No GLSL program set - nothing to do. */
506 programId
= prog
->programId
;
507 constant_version
= prog
->constant_version
;
509 if (useVertexShader
) {
510 IWineD3DBaseShaderImpl
* vshader
= (IWineD3DBaseShaderImpl
*) stateBlock
->vertexShader
;
512 /* Load DirectX 9 float constants/uniforms for vertex shader */
513 shader_glsl_load_constantsF(vshader
, gl_info
, stateBlock
->vertexShaderConstantF
,
514 prog
->vuniformF_locations
, &priv
->vconst_heap
, priv
->stack
, constant_version
);
516 /* Load DirectX 9 integer constants/uniforms for vertex shader */
517 if(vshader
->baseShader
.uses_int_consts
) {
518 shader_glsl_load_constantsI(vshader
, gl_info
, prog
->vuniformI_locations
,
519 stateBlock
->vertexShaderConstantI
, stateBlock
->changed
.vertexShaderConstantsI
);
522 /* Load DirectX 9 boolean constants/uniforms for vertex shader */
523 if(vshader
->baseShader
.uses_bool_consts
) {
524 shader_glsl_load_constantsB(vshader
, gl_info
, programId
,
525 stateBlock
->vertexShaderConstantB
, stateBlock
->changed
.vertexShaderConstantsB
);
528 /* Upload the position fixup params */
529 GL_EXTCALL(glUniform4fvARB(prog
->posFixup_location
, 1, &deviceImpl
->posFixup
[0]));
530 checkGLcall("glUniform4fvARB");
533 if (usePixelShader
) {
535 IWineD3DBaseShaderImpl
* pshader
= (IWineD3DBaseShaderImpl
*) stateBlock
->pixelShader
;
537 /* Load DirectX 9 float constants/uniforms for pixel shader */
538 shader_glsl_load_constantsF(pshader
, gl_info
, stateBlock
->pixelShaderConstantF
,
539 prog
->puniformF_locations
, &priv
->pconst_heap
, priv
->stack
, constant_version
);
541 /* Load DirectX 9 integer constants/uniforms for pixel shader */
542 if(pshader
->baseShader
.uses_int_consts
) {
543 shader_glsl_load_constantsI(pshader
, gl_info
, prog
->puniformI_locations
,
544 stateBlock
->pixelShaderConstantI
, stateBlock
->changed
.pixelShaderConstantsI
);
547 /* Load DirectX 9 boolean constants/uniforms for pixel shader */
548 if(pshader
->baseShader
.uses_bool_consts
) {
549 shader_glsl_load_constantsB(pshader
, gl_info
, programId
,
550 stateBlock
->pixelShaderConstantB
, stateBlock
->changed
.pixelShaderConstantsB
);
553 /* Upload the environment bump map matrix if needed. The needsbumpmat member specifies the texture stage to load the matrix from.
554 * It can't be 0 for a valid texbem instruction.
556 for(i
= 0; i
< ((IWineD3DPixelShaderImpl
*) pshader
)->numbumpenvmatconsts
; i
++) {
557 IWineD3DPixelShaderImpl
*ps
= (IWineD3DPixelShaderImpl
*) pshader
;
558 int stage
= ps
->luminanceconst
[i
].texunit
;
560 const float *data
= (const float *)&stateBlock
->textureState
[(int)ps
->bumpenvmatconst
[i
].texunit
][WINED3DTSS_BUMPENVMAT00
];
561 GL_EXTCALL(glUniformMatrix2fvARB(prog
->bumpenvmat_location
[i
], 1, 0, data
));
562 checkGLcall("glUniformMatrix2fvARB");
564 /* texbeml needs the luminance scale and offset too. If texbeml is used, needsbumpmat
565 * is set too, so we can check that in the needsbumpmat check
567 if(ps
->baseShader
.reg_maps
.luminanceparams
[stage
]) {
568 const GLfloat
*scale
= (const GLfloat
*)&stateBlock
->textureState
[stage
][WINED3DTSS_BUMPENVLSCALE
];
569 const GLfloat
*offset
= (const GLfloat
*)&stateBlock
->textureState
[stage
][WINED3DTSS_BUMPENVLOFFSET
];
571 GL_EXTCALL(glUniform1fvARB(prog
->luminancescale_location
[i
], 1, scale
));
572 checkGLcall("glUniform1fvARB");
573 GL_EXTCALL(glUniform1fvARB(prog
->luminanceoffset_location
[i
], 1, offset
));
574 checkGLcall("glUniform1fvARB");
578 if(((IWineD3DPixelShaderImpl
*) pshader
)->vpos_uniform
) {
579 float correction_params
[4];
580 if(deviceImpl
->render_offscreen
) {
581 correction_params
[0] = 0.0;
582 correction_params
[1] = 1.0;
584 /* position is window relative, not viewport relative */
585 correction_params
[0] = ((IWineD3DSurfaceImpl
*) deviceImpl
->render_targets
[0])->currentDesc
.Height
;
586 correction_params
[1] = -1.0;
588 GL_EXTCALL(glUniform4fvARB(prog
->ycorrection_location
, 1, correction_params
));
592 if (priv
->next_constant_version
== UINT_MAX
)
594 TRACE("Max constant version reached, resetting to 0.\n");
595 hash_table_for_each_entry(priv
->glsl_program_lookup
, reset_program_constant_version
, NULL
);
596 priv
->next_constant_version
= 1;
600 prog
->constant_version
= priv
->next_constant_version
++;
604 static inline void update_heap_entry(struct constant_heap
*heap
, unsigned int idx
,
605 unsigned int heap_idx
, DWORD new_version
)
607 struct constant_entry
*entries
= heap
->entries
;
608 unsigned int *positions
= heap
->positions
;
609 unsigned int parent_idx
;
613 parent_idx
= heap_idx
>> 1;
615 if (new_version
<= entries
[parent_idx
].version
) break;
617 entries
[heap_idx
] = entries
[parent_idx
];
618 positions
[entries
[parent_idx
].idx
] = heap_idx
;
619 heap_idx
= parent_idx
;
622 entries
[heap_idx
].version
= new_version
;
623 entries
[heap_idx
].idx
= idx
;
624 positions
[idx
] = heap_idx
;
627 static void shader_glsl_update_float_vertex_constants(IWineD3DDevice
*iface
, UINT start
, UINT count
)
629 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
630 struct shader_glsl_priv
*priv
= This
->shader_priv
;
631 struct constant_heap
*heap
= &priv
->vconst_heap
;
634 for (i
= start
; i
< count
+ start
; ++i
)
636 if (!This
->stateBlock
->changed
.vertexShaderConstantsF
[i
])
637 update_heap_entry(heap
, i
, heap
->size
++, priv
->next_constant_version
);
639 update_heap_entry(heap
, i
, heap
->positions
[i
], priv
->next_constant_version
);
643 static void shader_glsl_update_float_pixel_constants(IWineD3DDevice
*iface
, UINT start
, UINT count
)
645 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
646 struct shader_glsl_priv
*priv
= This
->shader_priv
;
647 struct constant_heap
*heap
= &priv
->pconst_heap
;
650 for (i
= start
; i
< count
+ start
; ++i
)
652 if (!This
->stateBlock
->changed
.pixelShaderConstantsF
[i
])
653 update_heap_entry(heap
, i
, heap
->size
++, priv
->next_constant_version
);
655 update_heap_entry(heap
, i
, heap
->positions
[i
], priv
->next_constant_version
);
659 /** Generate the variable & register declarations for the GLSL output target */
660 static void shader_generate_glsl_declarations(IWineD3DBaseShader
*iface
, const shader_reg_maps
*reg_maps
,
661 SHADER_BUFFER
*buffer
, const WineD3D_GL_Info
*gl_info
,
662 const struct ps_compile_args
*ps_args
)
664 IWineD3DBaseShaderImpl
* This
= (IWineD3DBaseShaderImpl
*) iface
;
665 IWineD3DDeviceImpl
*device
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
666 DWORD shader_version
= reg_maps
->shader_version
;
667 unsigned int i
, extra_constants_needed
= 0;
668 const local_constant
*lconst
;
670 /* There are some minor differences between pixel and vertex shaders */
671 char pshader
= shader_is_pshader_version(shader_version
);
672 char prefix
= pshader
? 'P' : 'V';
674 /* Prototype the subroutines */
675 for (i
= 0; i
< This
->baseShader
.limits
.label
; i
++) {
676 if (reg_maps
->labels
[i
])
677 shader_addline(buffer
, "void subroutine%u();\n", i
);
680 /* Declare the constants (aka uniforms) */
681 if (This
->baseShader
.limits
.constant_float
> 0) {
682 unsigned max_constantsF
= min(This
->baseShader
.limits
.constant_float
,
683 (pshader
? GL_LIMITS(pshader_constantsF
) : GL_LIMITS(vshader_constantsF
)));
684 shader_addline(buffer
, "uniform vec4 %cC[%u];\n", prefix
, max_constantsF
);
687 if (This
->baseShader
.limits
.constant_int
> 0)
688 shader_addline(buffer
, "uniform ivec4 %cI[%u];\n", prefix
, This
->baseShader
.limits
.constant_int
);
690 if (This
->baseShader
.limits
.constant_bool
> 0)
691 shader_addline(buffer
, "uniform bool %cB[%u];\n", prefix
, This
->baseShader
.limits
.constant_bool
);
694 shader_addline(buffer
, "uniform vec4 posFixup;\n");
695 /* Predeclaration; This function is added at link time based on the pixel shader.
696 * VS 3.0 shaders have an array OUT[] the shader writes to, earlier versions don't have
697 * that. We know the input to the reorder function at vertex shader compile time, so
698 * we can deal with that. The reorder function for a 1.x and 2.x vertex shader can just
699 * read gl_FrontColor. The output depends on the pixel shader. The reorder function for a
700 * 1.x and 2.x pshader or for fixed function will write gl_FrontColor, and for a 3.0 shader
701 * it will write to the varying array. Here we depend on the shader optimizer on sorting that
702 * out. The nvidia driver only does that if the parameter is inout instead of out, hence the
705 if (shader_version
>= WINED3DVS_VERSION(3, 0))
707 shader_addline(buffer
, "void order_ps_input(in vec4[%u]);\n", MAX_REG_OUTPUT
);
709 shader_addline(buffer
, "void order_ps_input();\n");
712 IWineD3DPixelShaderImpl
*ps_impl
= (IWineD3DPixelShaderImpl
*) This
;
714 ps_impl
->numbumpenvmatconsts
= 0;
715 for(i
= 0; i
< (sizeof(reg_maps
->bumpmat
) / sizeof(reg_maps
->bumpmat
[0])); i
++) {
716 if(!reg_maps
->bumpmat
[i
]) {
720 ps_impl
->bumpenvmatconst
[(int) ps_impl
->numbumpenvmatconsts
].texunit
= i
;
721 shader_addline(buffer
, "uniform mat2 bumpenvmat%d;\n", i
);
723 if(reg_maps
->luminanceparams
) {
724 ps_impl
->luminanceconst
[(int) ps_impl
->numbumpenvmatconsts
].texunit
= i
;
725 shader_addline(buffer
, "uniform float luminancescale%d;\n", i
);
726 shader_addline(buffer
, "uniform float luminanceoffset%d;\n", i
);
727 extra_constants_needed
++;
729 ps_impl
->luminanceconst
[(int) ps_impl
->numbumpenvmatconsts
].texunit
= -1;
732 extra_constants_needed
++;
733 ps_impl
->numbumpenvmatconsts
++;
736 if(ps_args
->srgb_correction
) {
737 shader_addline(buffer
, "const vec4 srgb_mul_low = vec4(%f, %f, %f, %f);\n",
738 srgb_mul_low
, srgb_mul_low
, srgb_mul_low
, srgb_mul_low
);
739 shader_addline(buffer
, "const vec4 srgb_comparison = vec4(%f, %f, %f, %f);\n",
740 srgb_cmp
, srgb_cmp
, srgb_cmp
, srgb_cmp
);
742 if(reg_maps
->vpos
|| reg_maps
->usesdsy
) {
743 if(This
->baseShader
.limits
.constant_float
+ extra_constants_needed
+ 1 < GL_LIMITS(pshader_constantsF
)) {
744 shader_addline(buffer
, "uniform vec4 ycorrection;\n");
745 ((IWineD3DPixelShaderImpl
*) This
)->vpos_uniform
= 1;
746 extra_constants_needed
++;
748 /* This happens because we do not have proper tracking of the constant registers that are
749 * actually used, only the max limit of the shader version
751 FIXME("Cannot find a free uniform for vpos correction params\n");
752 shader_addline(buffer
, "const vec4 ycorrection = vec4(%f, %f, 0.0, 0.0);\n",
753 device
->render_offscreen
? 0.0 : ((IWineD3DSurfaceImpl
*) device
->render_targets
[0])->currentDesc
.Height
,
754 device
->render_offscreen
? 1.0 : -1.0);
756 shader_addline(buffer
, "vec4 vpos;\n");
760 /* Declare texture samplers */
761 for (i
= 0; i
< This
->baseShader
.limits
.sampler
; i
++) {
762 if (reg_maps
->samplers
[i
]) {
764 DWORD stype
= reg_maps
->samplers
[i
] & WINED3DSP_TEXTURETYPE_MASK
;
768 shader_addline(buffer
, "uniform sampler1D %csampler%u;\n", prefix
, i
);
771 if(device
->stateBlock
->textures
[i
] &&
772 IWineD3DBaseTexture_GetTextureDimensions(device
->stateBlock
->textures
[i
]) == GL_TEXTURE_RECTANGLE_ARB
) {
773 shader_addline(buffer
, "uniform sampler2DRect %csampler%u;\n", prefix
, i
);
775 shader_addline(buffer
, "uniform sampler2D %csampler%u;\n", prefix
, i
);
778 case WINED3DSTT_CUBE
:
779 shader_addline(buffer
, "uniform samplerCube %csampler%u;\n", prefix
, i
);
781 case WINED3DSTT_VOLUME
:
782 shader_addline(buffer
, "uniform sampler3D %csampler%u;\n", prefix
, i
);
785 shader_addline(buffer
, "uniform unsupported_sampler %csampler%u;\n", prefix
, i
);
786 FIXME("Unrecognized sampler type: %#x\n", stype
);
792 /* Declare address variables */
793 for (i
= 0; i
< This
->baseShader
.limits
.address
; i
++) {
794 if (reg_maps
->address
[i
])
795 shader_addline(buffer
, "ivec4 A%d;\n", i
);
798 /* Declare texture coordinate temporaries and initialize them */
799 for (i
= 0; i
< This
->baseShader
.limits
.texcoord
; i
++) {
800 if (reg_maps
->texcoord
[i
])
801 shader_addline(buffer
, "vec4 T%u = gl_TexCoord[%u];\n", i
, i
);
804 /* Declare input register varyings. Only pixel shader, vertex shaders have that declared in the
805 * helper function shader that is linked in at link time
807 if (pshader
&& shader_version
>= WINED3DPS_VERSION(3, 0))
809 if (use_vs(device
->stateBlock
))
811 shader_addline(buffer
, "varying vec4 IN[%u];\n", GL_LIMITS(glsl_varyings
) / 4);
813 /* TODO: Write a replacement shader for the fixed function vertex pipeline, so this isn't needed.
814 * For fixed function vertex processing + 3.0 pixel shader we need a separate function in the
815 * pixel shader that reads the fixed function color into the packed input registers.
817 shader_addline(buffer
, "vec4 IN[%u];\n", GL_LIMITS(glsl_varyings
) / 4);
821 /* Declare output register temporaries */
822 if(This
->baseShader
.limits
.packed_output
) {
823 shader_addline(buffer
, "vec4 OUT[%u];\n", This
->baseShader
.limits
.packed_output
);
826 /* Declare temporary variables */
827 for(i
= 0; i
< This
->baseShader
.limits
.temporary
; i
++) {
828 if (reg_maps
->temporary
[i
])
829 shader_addline(buffer
, "vec4 R%u;\n", i
);
832 /* Declare attributes */
833 for (i
= 0; i
< This
->baseShader
.limits
.attributes
; i
++) {
834 if (reg_maps
->attributes
[i
])
835 shader_addline(buffer
, "attribute vec4 attrib%i;\n", i
);
838 /* Declare loop registers aLx */
839 for (i
= 0; i
< reg_maps
->loop_depth
; i
++) {
840 shader_addline(buffer
, "int aL%u;\n", i
);
841 shader_addline(buffer
, "int tmpInt%u;\n", i
);
844 /* Temporary variables for matrix operations */
845 shader_addline(buffer
, "vec4 tmp0;\n");
846 shader_addline(buffer
, "vec4 tmp1;\n");
848 /* Local constants use a different name so they can be loaded once at shader link time
849 * They can't be hardcoded into the shader text via LC = {x, y, z, w}; because the
850 * float -> string conversion can cause precision loss.
852 if(!This
->baseShader
.load_local_constsF
) {
853 LIST_FOR_EACH_ENTRY(lconst
, &This
->baseShader
.constantsF
, local_constant
, entry
) {
854 shader_addline(buffer
, "uniform vec4 %cLC%u;\n", prefix
, lconst
->idx
);
858 /* Start the main program */
859 shader_addline(buffer
, "void main() {\n");
860 if(pshader
&& reg_maps
->vpos
) {
861 /* DirectX apps expect integer values, while OpenGL drivers add approximately 0.5. This causes
862 * off-by-one problems as spotted by the vPos d3d9 visual test. Unfortunately the ATI cards do
863 * not add exactly 0.5, but rather something like 0.49999999 or 0.50000001, which still causes
864 * precision troubles when we just substract 0.5.
866 * To deal with that just floor() the position. This will eliminate the fraction on all cards.
868 * TODO: Test how that behaves with multisampling once we can enable multisampling in winex11.
870 * An advantage of floor is that it works even if the driver doesn't add 1/2. It is somewhat
871 * questionable if 1.5, 2.5, ... are the proper values to return in gl_FragCoord, even though
872 * coordinates specify the pixel centers instead of the pixel corners. This code will behave
873 * correctly on drivers that returns integer values.
875 shader_addline(buffer
, "vpos = floor(vec4(0, ycorrection[0], 0, 0) + gl_FragCoord * vec4(1, ycorrection[1], 1, 1));\n");
879 /*****************************************************************************
880 * Functions to generate GLSL strings from DirectX Shader bytecode begin here.
882 * For more information, see http://wiki.winehq.org/DirectX-Shaders
883 ****************************************************************************/
886 static void shader_glsl_add_src_param(const SHADER_OPCODE_ARG
*arg
, const DWORD param
,
887 const DWORD addr_token
, DWORD mask
, glsl_src_param_t
*src_param
);
889 /** Used for opcode modifiers - They multiply the result by the specified amount */
890 static const char * const shift_glsl_tab
[] = {
892 "2.0 * ", /* 1 (x2) */
893 "4.0 * ", /* 2 (x4) */
894 "8.0 * ", /* 3 (x8) */
895 "16.0 * ", /* 4 (x16) */
896 "32.0 * ", /* 5 (x32) */
903 "0.0625 * ", /* 12 (d16) */
904 "0.125 * ", /* 13 (d8) */
905 "0.25 * ", /* 14 (d4) */
906 "0.5 * " /* 15 (d2) */
909 /* Generate a GLSL parameter that does the input modifier computation and return the input register/mask to use */
910 static void shader_glsl_gen_modifier (
913 const char *in_regswizzle
,
918 if (instr
== WINED3DSIO_TEXKILL
)
921 switch (instr
& WINED3DSP_SRCMOD_MASK
) {
922 case WINED3DSPSM_DZ
: /* Need to handle this in the instructions itself (texld & texcrd). */
924 case WINED3DSPSM_NONE
:
925 sprintf(out_str
, "%s%s", in_reg
, in_regswizzle
);
927 case WINED3DSPSM_NEG
:
928 sprintf(out_str
, "-%s%s", in_reg
, in_regswizzle
);
930 case WINED3DSPSM_NOT
:
931 sprintf(out_str
, "!%s%s", in_reg
, in_regswizzle
);
933 case WINED3DSPSM_BIAS
:
934 sprintf(out_str
, "(%s%s - vec4(0.5)%s)", in_reg
, in_regswizzle
, in_regswizzle
);
936 case WINED3DSPSM_BIASNEG
:
937 sprintf(out_str
, "-(%s%s - vec4(0.5)%s)", in_reg
, in_regswizzle
, in_regswizzle
);
939 case WINED3DSPSM_SIGN
:
940 sprintf(out_str
, "(2.0 * (%s%s - 0.5))", in_reg
, in_regswizzle
);
942 case WINED3DSPSM_SIGNNEG
:
943 sprintf(out_str
, "-(2.0 * (%s%s - 0.5))", in_reg
, in_regswizzle
);
945 case WINED3DSPSM_COMP
:
946 sprintf(out_str
, "(1.0 - %s%s)", in_reg
, in_regswizzle
);
949 sprintf(out_str
, "(2.0 * %s%s)", in_reg
, in_regswizzle
);
951 case WINED3DSPSM_X2NEG
:
952 sprintf(out_str
, "-(2.0 * %s%s)", in_reg
, in_regswizzle
);
954 case WINED3DSPSM_ABS
:
955 sprintf(out_str
, "abs(%s%s)", in_reg
, in_regswizzle
);
957 case WINED3DSPSM_ABSNEG
:
958 sprintf(out_str
, "-abs(%s%s)", in_reg
, in_regswizzle
);
961 FIXME("Unhandled modifier %u\n", (instr
& WINED3DSP_SRCMOD_MASK
));
962 sprintf(out_str
, "%s%s", in_reg
, in_regswizzle
);
966 /** Writes the GLSL variable name that corresponds to the register that the
967 * DX opcode parameter is trying to access */
968 static void shader_glsl_get_register_name(const DWORD param
, const DWORD addr_token
,
969 char *regstr
, BOOL
*is_color
, const SHADER_OPCODE_ARG
*arg
)
971 /* oPos, oFog and oPts in D3D */
972 static const char * const hwrastout_reg_names
[] = { "gl_Position", "gl_FogFragCoord", "gl_PointSize" };
974 DWORD reg
= param
& WINED3DSP_REGNUM_MASK
;
975 DWORD regtype
= shader_get_regtype(param
);
976 IWineD3DBaseShaderImpl
* This
= (IWineD3DBaseShaderImpl
*) arg
->shader
;
977 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
978 const WineD3D_GL_Info
* gl_info
= &deviceImpl
->adapter
->gl_info
;
979 DWORD shader_version
= This
->baseShader
.reg_maps
.shader_version
;
980 char pshader
= shader_is_pshader_version(shader_version
);
986 case WINED3DSPR_TEMP
:
987 sprintf(tmpStr
, "R%u", reg
);
989 case WINED3DSPR_INPUT
:
991 /* Pixel shaders >= 3.0 */
992 if (WINED3DSHADER_VERSION_MAJOR(shader_version
) >= 3)
994 DWORD in_count
= GL_LIMITS(glsl_varyings
) / 4;
996 if (param
& WINED3DSHADER_ADDRMODE_RELATIVE
) {
997 glsl_src_param_t rel_param
;
998 shader_glsl_add_src_param(arg
, addr_token
, 0, WINED3DSP_WRITEMASK_0
, &rel_param
);
1000 /* Removing a + 0 would be an obvious optimization, but macos doesn't see the NOP
1003 if(((IWineD3DPixelShaderImpl
*) This
)->input_reg_map
[reg
]) {
1004 if (((IWineD3DPixelShaderImpl
*)This
)->declared_in_count
> in_count
) {
1005 sprintf(tmpStr
, "((%s + %u) > %d ? (%s + %u) > %d ? gl_SecondaryColor : gl_Color : IN[%s + %u])",
1006 rel_param
.param_str
, ((IWineD3DPixelShaderImpl
*)This
)->input_reg_map
[reg
], in_count
- 1,
1007 rel_param
.param_str
, ((IWineD3DPixelShaderImpl
*)This
)->input_reg_map
[reg
], in_count
,
1008 rel_param
.param_str
, ((IWineD3DPixelShaderImpl
*)This
)->input_reg_map
[reg
]);
1010 sprintf(tmpStr
, "IN[%s + %u]", rel_param
.param_str
, ((IWineD3DPixelShaderImpl
*)This
)->input_reg_map
[reg
]);
1013 if (((IWineD3DPixelShaderImpl
*)This
)->declared_in_count
> in_count
) {
1014 sprintf(tmpStr
, "((%s) > %d ? (%s) > %d ? gl_SecondaryColor : gl_Color : IN[%s])",
1015 rel_param
.param_str
, in_count
- 1,
1016 rel_param
.param_str
, in_count
,
1017 rel_param
.param_str
);
1019 sprintf(tmpStr
, "IN[%s]", rel_param
.param_str
);
1023 DWORD idx
= ((IWineD3DPixelShaderImpl
*) This
)->input_reg_map
[reg
];
1024 if (idx
== in_count
) {
1025 sprintf(tmpStr
, "gl_Color");
1026 } else if (idx
== in_count
+ 1) {
1027 sprintf(tmpStr
, "gl_SecondaryColor");
1029 sprintf(tmpStr
, "IN[%u]", idx
);
1034 strcpy(tmpStr
, "gl_Color");
1036 strcpy(tmpStr
, "gl_SecondaryColor");
1039 if (((IWineD3DVertexShaderImpl
*)This
)->cur_args
->swizzle_map
& (1 << reg
)) *is_color
= TRUE
;
1040 sprintf(tmpStr
, "attrib%u", reg
);
1043 case WINED3DSPR_CONST
:
1045 const char prefix
= pshader
? 'P':'V';
1047 /* Relative addressing */
1048 if (param
& WINED3DSHADER_ADDRMODE_RELATIVE
) {
1050 /* Relative addressing on shaders 2.0+ have a relative address token,
1051 * prior to that, it was hard-coded as "A0.x" because there's only 1 register */
1052 if (WINED3DSHADER_VERSION_MAJOR(shader_version
) >= 2)
1054 glsl_src_param_t rel_param
;
1055 shader_glsl_add_src_param(arg
, addr_token
, 0, WINED3DSP_WRITEMASK_0
, &rel_param
);
1057 sprintf(tmpStr
, "%cC[%s + %u]", prefix
, rel_param
.param_str
, reg
);
1059 sprintf(tmpStr
, "%cC[%s]", prefix
, rel_param
.param_str
);
1063 sprintf(tmpStr
, "%cC[A0.x + %u]", prefix
, reg
);
1065 sprintf(tmpStr
, "%cC[A0.x]", prefix
);
1070 if(shader_constant_is_local(This
, reg
)) {
1071 sprintf(tmpStr
, "%cLC%u", prefix
, reg
);
1073 sprintf(tmpStr
, "%cC[%u]", prefix
, reg
);
1079 case WINED3DSPR_CONSTINT
:
1081 sprintf(tmpStr
, "PI[%u]", reg
);
1083 sprintf(tmpStr
, "VI[%u]", reg
);
1085 case WINED3DSPR_CONSTBOOL
:
1087 sprintf(tmpStr
, "PB[%u]", reg
);
1089 sprintf(tmpStr
, "VB[%u]", reg
);
1091 case WINED3DSPR_TEXTURE
: /* case WINED3DSPR_ADDR: */
1093 sprintf(tmpStr
, "T%u", reg
);
1095 sprintf(tmpStr
, "A%u", reg
);
1098 case WINED3DSPR_LOOP
:
1099 sprintf(tmpStr
, "aL%u", This
->baseShader
.cur_loop_regno
- 1);
1101 case WINED3DSPR_SAMPLER
:
1103 sprintf(tmpStr
, "Psampler%u", reg
);
1105 sprintf(tmpStr
, "Vsampler%u", reg
);
1107 case WINED3DSPR_COLOROUT
:
1108 if (reg
>= GL_LIMITS(buffers
)) {
1109 WARN("Write to render target %u, only %d supported\n", reg
, 4);
1111 if (GL_SUPPORT(ARB_DRAW_BUFFERS
)) {
1112 sprintf(tmpStr
, "gl_FragData[%u]", reg
);
1113 } else { /* On older cards with GLSL support like the GeforceFX there's only one buffer. */
1114 sprintf(tmpStr
, "gl_FragColor");
1117 case WINED3DSPR_RASTOUT
:
1118 sprintf(tmpStr
, "%s", hwrastout_reg_names
[reg
]);
1120 case WINED3DSPR_DEPTHOUT
:
1121 sprintf(tmpStr
, "gl_FragDepth");
1123 case WINED3DSPR_ATTROUT
:
1125 sprintf(tmpStr
, "gl_FrontColor");
1127 sprintf(tmpStr
, "gl_FrontSecondaryColor");
1130 case WINED3DSPR_TEXCRDOUT
:
1131 /* Vertex shaders >= 3.0: WINED3DSPR_OUTPUT */
1132 if (WINED3DSHADER_VERSION_MAJOR(shader_version
) >= 3) sprintf(tmpStr
, "OUT[%u]", reg
);
1133 else sprintf(tmpStr
, "gl_TexCoord[%u]", reg
);
1135 case WINED3DSPR_MISCTYPE
:
1138 sprintf(tmpStr
, "vpos");
1139 } else if (reg
== 1){
1140 /* Note that gl_FrontFacing is a bool, while vFace is
1141 * a float for which the sign determines front/back
1143 sprintf(tmpStr
, "(gl_FrontFacing ? 1.0 : -1.0)");
1145 FIXME("Unhandled misctype register %d\n", reg
);
1146 sprintf(tmpStr
, "unrecognized_register");
1150 FIXME("Unhandled register name Type(%d)\n", regtype
);
1151 sprintf(tmpStr
, "unrecognized_register");
1155 strcat(regstr
, tmpStr
);
1158 /* Get the GLSL write mask for the destination register */
1159 static DWORD
shader_glsl_get_write_mask(const DWORD param
, char *write_mask
) {
1160 char *ptr
= write_mask
;
1161 DWORD mask
= param
& WINED3DSP_WRITEMASK_ALL
;
1163 if (shader_is_scalar(param
)) {
1164 mask
= WINED3DSP_WRITEMASK_0
;
1167 if (param
& WINED3DSP_WRITEMASK_0
) *ptr
++ = 'x';
1168 if (param
& WINED3DSP_WRITEMASK_1
) *ptr
++ = 'y';
1169 if (param
& WINED3DSP_WRITEMASK_2
) *ptr
++ = 'z';
1170 if (param
& WINED3DSP_WRITEMASK_3
) *ptr
++ = 'w';
1178 static unsigned int shader_glsl_get_write_mask_size(DWORD write_mask
) {
1179 unsigned int size
= 0;
1181 if (write_mask
& WINED3DSP_WRITEMASK_0
) ++size
;
1182 if (write_mask
& WINED3DSP_WRITEMASK_1
) ++size
;
1183 if (write_mask
& WINED3DSP_WRITEMASK_2
) ++size
;
1184 if (write_mask
& WINED3DSP_WRITEMASK_3
) ++size
;
1189 static void shader_glsl_get_swizzle(const DWORD param
, BOOL fixup
, DWORD mask
, char *swizzle_str
) {
1190 /* For registers of type WINED3DDECLTYPE_D3DCOLOR, data is stored as "bgra",
1191 * but addressed as "rgba". To fix this we need to swap the register's x
1192 * and z components. */
1193 DWORD swizzle
= (param
& WINED3DSP_SWIZZLE_MASK
) >> WINED3DSP_SWIZZLE_SHIFT
;
1194 const char *swizzle_chars
= fixup
? "zyxw" : "xyzw";
1195 char *ptr
= swizzle_str
;
1197 if (!shader_is_scalar(param
)) {
1199 /* swizzle bits fields: wwzzyyxx */
1200 if (mask
& WINED3DSP_WRITEMASK_0
) *ptr
++ = swizzle_chars
[swizzle
& 0x03];
1201 if (mask
& WINED3DSP_WRITEMASK_1
) *ptr
++ = swizzle_chars
[(swizzle
>> 2) & 0x03];
1202 if (mask
& WINED3DSP_WRITEMASK_2
) *ptr
++ = swizzle_chars
[(swizzle
>> 4) & 0x03];
1203 if (mask
& WINED3DSP_WRITEMASK_3
) *ptr
++ = swizzle_chars
[(swizzle
>> 6) & 0x03];
1209 /* From a given parameter token, generate the corresponding GLSL string.
1210 * Also, return the actual register name and swizzle in case the
1211 * caller needs this information as well. */
1212 static void shader_glsl_add_src_param(const SHADER_OPCODE_ARG
*arg
, const DWORD param
,
1213 const DWORD addr_token
, DWORD mask
, glsl_src_param_t
*src_param
)
1215 BOOL is_color
= FALSE
;
1216 char swizzle_str
[6];
1218 src_param
->reg_name
[0] = '\0';
1219 src_param
->param_str
[0] = '\0';
1220 swizzle_str
[0] = '\0';
1222 shader_glsl_get_register_name(param
, addr_token
, src_param
->reg_name
, &is_color
, arg
);
1224 shader_glsl_get_swizzle(param
, is_color
, mask
, swizzle_str
);
1225 shader_glsl_gen_modifier(param
, src_param
->reg_name
, swizzle_str
, src_param
->param_str
);
1228 /* From a given parameter token, generate the corresponding GLSL string.
1229 * Also, return the actual register name and swizzle in case the
1230 * caller needs this information as well. */
1231 static DWORD
shader_glsl_add_dst_param(const SHADER_OPCODE_ARG
* arg
, const DWORD param
,
1232 const DWORD addr_token
, glsl_dst_param_t
*dst_param
)
1234 BOOL is_color
= FALSE
;
1236 dst_param
->mask_str
[0] = '\0';
1237 dst_param
->reg_name
[0] = '\0';
1239 shader_glsl_get_register_name(param
, addr_token
, dst_param
->reg_name
, &is_color
, arg
);
1240 return shader_glsl_get_write_mask(param
, dst_param
->mask_str
);
1243 /* Append the destination part of the instruction to the buffer, return the effective write mask */
1244 static DWORD
shader_glsl_append_dst_ext(SHADER_BUFFER
*buffer
, const SHADER_OPCODE_ARG
*arg
, const DWORD param
)
1246 glsl_dst_param_t dst_param
;
1250 mask
= shader_glsl_add_dst_param(arg
, param
, arg
->dst_addr
, &dst_param
);
1253 shift
= (param
& WINED3DSP_DSTSHIFT_MASK
) >> WINED3DSP_DSTSHIFT_SHIFT
;
1254 shader_addline(buffer
, "%s%s = %s(", dst_param
.reg_name
, dst_param
.mask_str
, shift_glsl_tab
[shift
]);
1260 /* Append the destination part of the instruction to the buffer, return the effective write mask */
1261 static DWORD
shader_glsl_append_dst(SHADER_BUFFER
*buffer
, const SHADER_OPCODE_ARG
*arg
)
1263 return shader_glsl_append_dst_ext(buffer
, arg
, arg
->dst
);
1266 /** Process GLSL instruction modifiers */
1267 void shader_glsl_add_instruction_modifiers(const SHADER_OPCODE_ARG
* arg
)
1269 DWORD mask
= arg
->dst
& WINED3DSP_DSTMOD_MASK
;
1271 if (arg
->opcode
->dst_token
&& mask
!= 0) {
1272 glsl_dst_param_t dst_param
;
1274 shader_glsl_add_dst_param(arg
, arg
->dst
, 0, &dst_param
);
1276 if (mask
& WINED3DSPDM_SATURATE
) {
1277 /* _SAT means to clamp the value of the register to between 0 and 1 */
1278 shader_addline(arg
->buffer
, "%s%s = clamp(%s%s, 0.0, 1.0);\n", dst_param
.reg_name
,
1279 dst_param
.mask_str
, dst_param
.reg_name
, dst_param
.mask_str
);
1281 if (mask
& WINED3DSPDM_MSAMPCENTROID
) {
1282 FIXME("_centroid modifier not handled\n");
1284 if (mask
& WINED3DSPDM_PARTIALPRECISION
) {
1285 /* MSDN says this modifier can be safely ignored, so that's what we'll do. */
1290 static inline const char* shader_get_comp_op(
1291 const DWORD opcode
) {
1293 DWORD op
= (opcode
& INST_CONTROLS_MASK
) >> INST_CONTROLS_SHIFT
;
1295 case COMPARISON_GT
: return ">";
1296 case COMPARISON_EQ
: return "==";
1297 case COMPARISON_GE
: return ">=";
1298 case COMPARISON_LT
: return "<";
1299 case COMPARISON_NE
: return "!=";
1300 case COMPARISON_LE
: return "<=";
1302 FIXME("Unrecognized comparison value: %u\n", op
);
1307 static void shader_glsl_get_sample_function(DWORD sampler_type
, DWORD flags
, glsl_sample_function_t
*sample_function
)
1309 BOOL projected
= flags
& WINED3D_GLSL_SAMPLE_PROJECTED
;
1310 BOOL texrect
= flags
& WINED3D_GLSL_SAMPLE_RECT
;
1311 BOOL lod
= flags
& WINED3D_GLSL_SAMPLE_LOD
;
1313 /* Note that there's no such thing as a projected cube texture. */
1314 switch(sampler_type
) {
1317 sample_function
->name
= projected
? "texture1DProjLod" : "texture1DLod";
1319 sample_function
->name
= projected
? "texture1DProj" : "texture1D";
1321 sample_function
->coord_mask
= WINED3DSP_WRITEMASK_0
;
1326 sample_function
->name
= projected
? "texture2DRectProjLod" : "texture2DRectLod";
1328 sample_function
->name
= projected
? "texture2DRectProj" : "texture2DRect";
1332 sample_function
->name
= projected
? "texture2DProjLod" : "texture2DLod";
1334 sample_function
->name
= projected
? "texture2DProj" : "texture2D";
1337 sample_function
->coord_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
;
1339 case WINED3DSTT_CUBE
:
1341 sample_function
->name
= "textureCubeLod";
1343 sample_function
->name
= "textureCube";
1345 sample_function
->coord_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
1347 case WINED3DSTT_VOLUME
:
1349 sample_function
->name
= projected
? "texture3DProjLod" : "texture3DLod";
1351 sample_function
->name
= projected
? "texture3DProj" : "texture3D";
1353 sample_function
->coord_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
1356 sample_function
->name
= "";
1357 sample_function
->coord_mask
= 0;
1358 FIXME("Unrecognized sampler type: %#x;\n", sampler_type
);
1363 static void shader_glsl_append_fixup_arg(char *arguments
, const char *reg_name
,
1364 BOOL sign_fixup
, enum fixup_channel_source channel_source
)
1366 switch(channel_source
)
1368 case CHANNEL_SOURCE_ZERO
:
1369 strcat(arguments
, "0.0");
1372 case CHANNEL_SOURCE_ONE
:
1373 strcat(arguments
, "1.0");
1376 case CHANNEL_SOURCE_X
:
1377 strcat(arguments
, reg_name
);
1378 strcat(arguments
, ".x");
1381 case CHANNEL_SOURCE_Y
:
1382 strcat(arguments
, reg_name
);
1383 strcat(arguments
, ".y");
1386 case CHANNEL_SOURCE_Z
:
1387 strcat(arguments
, reg_name
);
1388 strcat(arguments
, ".z");
1391 case CHANNEL_SOURCE_W
:
1392 strcat(arguments
, reg_name
);
1393 strcat(arguments
, ".w");
1397 FIXME("Unhandled channel source %#x\n", channel_source
);
1398 strcat(arguments
, "undefined");
1402 if (sign_fixup
) strcat(arguments
, " * 2.0 - 1.0");
1405 static void shader_glsl_color_correction(const struct SHADER_OPCODE_ARG
*arg
, struct color_fixup_desc fixup
)
1407 unsigned int mask_size
, remaining
;
1408 glsl_dst_param_t dst_param
;
1409 char arguments
[256];
1414 if (fixup
.x_sign_fixup
|| fixup
.x_source
!= CHANNEL_SOURCE_X
) mask
|= WINED3DSP_WRITEMASK_0
;
1415 if (fixup
.y_sign_fixup
|| fixup
.y_source
!= CHANNEL_SOURCE_Y
) mask
|= WINED3DSP_WRITEMASK_1
;
1416 if (fixup
.z_sign_fixup
|| fixup
.z_source
!= CHANNEL_SOURCE_Z
) mask
|= WINED3DSP_WRITEMASK_2
;
1417 if (fixup
.w_sign_fixup
|| fixup
.w_source
!= CHANNEL_SOURCE_W
) mask
|= WINED3DSP_WRITEMASK_3
;
1420 if (!mask
) return; /* Nothing to do */
1422 if (is_yuv_fixup(fixup
))
1424 enum yuv_fixup yuv_fixup
= get_yuv_fixup(fixup
);
1425 FIXME("YUV fixup (%#x) not supported\n", yuv_fixup
);
1429 mask_size
= shader_glsl_get_write_mask_size(mask
);
1431 dst_param
.mask_str
[0] = '\0';
1432 shader_glsl_get_write_mask(mask
, dst_param
.mask_str
);
1434 dst_param
.reg_name
[0] = '\0';
1435 shader_glsl_get_register_name(arg
->dst
, arg
->dst_addr
, dst_param
.reg_name
, &dummy
, arg
);
1437 arguments
[0] = '\0';
1438 remaining
= mask_size
;
1439 if (mask
& WINED3DSP_WRITEMASK_0
)
1441 shader_glsl_append_fixup_arg(arguments
, dst_param
.reg_name
, fixup
.x_sign_fixup
, fixup
.x_source
);
1442 if (--remaining
) strcat(arguments
, ", ");
1444 if (mask
& WINED3DSP_WRITEMASK_1
)
1446 shader_glsl_append_fixup_arg(arguments
, dst_param
.reg_name
, fixup
.y_sign_fixup
, fixup
.y_source
);
1447 if (--remaining
) strcat(arguments
, ", ");
1449 if (mask
& WINED3DSP_WRITEMASK_2
)
1451 shader_glsl_append_fixup_arg(arguments
, dst_param
.reg_name
, fixup
.z_sign_fixup
, fixup
.z_source
);
1452 if (--remaining
) strcat(arguments
, ", ");
1454 if (mask
& WINED3DSP_WRITEMASK_3
)
1456 shader_glsl_append_fixup_arg(arguments
, dst_param
.reg_name
, fixup
.w_sign_fixup
, fixup
.w_source
);
1457 if (--remaining
) strcat(arguments
, ", ");
1462 shader_addline(arg
->buffer
, "%s%s = vec%u(%s);\n",
1463 dst_param
.reg_name
, dst_param
.mask_str
, mask_size
, arguments
);
1467 shader_addline(arg
->buffer
, "%s%s = %s;\n", dst_param
.reg_name
, dst_param
.mask_str
, arguments
);
1471 static void shader_glsl_gen_sample_code(const SHADER_OPCODE_ARG
*arg
, DWORD sampler
, const char *coord_reg
,
1472 const glsl_sample_function_t
*sample_function
, DWORD swizzle
,
1474 const char *sampler_base
;
1475 char dst_swizzle
[6];
1476 struct color_fixup_desc fixup
;
1478 shader_glsl_get_swizzle(swizzle
, FALSE
, arg
->dst
, dst_swizzle
);
1480 if(shader_is_pshader_version(arg
->reg_maps
->shader_version
)) {
1481 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*) arg
->shader
;
1482 fixup
= This
->cur_args
->color_fixup
[sampler
];
1483 sampler_base
= "Psampler";
1485 sampler_base
= "Vsampler";
1486 fixup
= COLOR_FIXUP_IDENTITY
; /* FIXME: Vshader color fixup */
1489 shader_glsl_append_dst(arg
->buffer
, arg
);
1492 shader_addline(arg
->buffer
, "%s(%s%u, %s, %s)%s);\n",
1493 sample_function
->name
, sampler_base
, sampler
, coord_reg
, bias
, dst_swizzle
);
1495 shader_addline(arg
->buffer
, "%s(%s%u, %s)%s);\n",
1496 sample_function
->name
, sampler_base
, sampler
, coord_reg
, dst_swizzle
);
1499 if(!is_identity_fixup(fixup
)) {
1500 shader_glsl_color_correction(arg
, fixup
);
1504 /*****************************************************************************
1506 * Begin processing individual instruction opcodes
1508 ****************************************************************************/
1510 /* Generate GLSL arithmetic functions (dst = src1 + src2) */
1511 static void shader_glsl_arith(const SHADER_OPCODE_ARG
*arg
)
1513 CONST SHADER_OPCODE
* curOpcode
= arg
->opcode
;
1514 SHADER_BUFFER
* buffer
= arg
->buffer
;
1515 glsl_src_param_t src0_param
;
1516 glsl_src_param_t src1_param
;
1520 /* Determine the GLSL operator to use based on the opcode */
1521 switch (curOpcode
->opcode
) {
1522 case WINED3DSIO_MUL
: op
= '*'; break;
1523 case WINED3DSIO_ADD
: op
= '+'; break;
1524 case WINED3DSIO_SUB
: op
= '-'; break;
1527 FIXME("Opcode %s not yet handled in GLSL\n", curOpcode
->name
);
1531 write_mask
= shader_glsl_append_dst(buffer
, arg
);
1532 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], write_mask
, &src0_param
);
1533 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], write_mask
, &src1_param
);
1534 shader_addline(buffer
, "%s %c %s);\n", src0_param
.param_str
, op
, src1_param
.param_str
);
1537 /* Process the WINED3DSIO_MOV opcode using GLSL (dst = src) */
1538 static void shader_glsl_mov(const SHADER_OPCODE_ARG
*arg
)
1540 SHADER_BUFFER
* buffer
= arg
->buffer
;
1541 glsl_src_param_t src0_param
;
1544 write_mask
= shader_glsl_append_dst(buffer
, arg
);
1545 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], write_mask
, &src0_param
);
1547 /* In vs_1_1 WINED3DSIO_MOV can write to the address register. In later
1548 * shader versions WINED3DSIO_MOVA is used for this. */
1549 if ((WINED3DSHADER_VERSION_MAJOR(arg
->reg_maps
->shader_version
) == 1
1550 && !shader_is_pshader_version(arg
->reg_maps
->shader_version
)
1551 && shader_get_regtype(arg
->dst
) == WINED3DSPR_ADDR
))
1553 /* This is a simple floor() */
1554 unsigned int mask_size
= shader_glsl_get_write_mask_size(write_mask
);
1555 if (mask_size
> 1) {
1556 shader_addline(buffer
, "ivec%d(floor(%s)));\n", mask_size
, src0_param
.param_str
);
1558 shader_addline(buffer
, "int(floor(%s)));\n", src0_param
.param_str
);
1560 } else if(arg
->opcode
->opcode
== WINED3DSIO_MOVA
) {
1561 /* We need to *round* to the nearest int here. */
1562 unsigned int mask_size
= shader_glsl_get_write_mask_size(write_mask
);
1563 if (mask_size
> 1) {
1564 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
);
1566 shader_addline(buffer
, "int(floor(abs(%s) + 0.5) * sign(%s)));\n", src0_param
.param_str
, src0_param
.param_str
);
1569 shader_addline(buffer
, "%s);\n", src0_param
.param_str
);
1573 /* Process the dot product operators DP3 and DP4 in GLSL (dst = dot(src0, src1)) */
1574 static void shader_glsl_dot(const SHADER_OPCODE_ARG
*arg
)
1576 CONST SHADER_OPCODE
* curOpcode
= arg
->opcode
;
1577 SHADER_BUFFER
* buffer
= arg
->buffer
;
1578 glsl_src_param_t src0_param
;
1579 glsl_src_param_t src1_param
;
1580 DWORD dst_write_mask
, src_write_mask
;
1581 unsigned int dst_size
= 0;
1583 dst_write_mask
= shader_glsl_append_dst(buffer
, arg
);
1584 dst_size
= shader_glsl_get_write_mask_size(dst_write_mask
);
1586 /* dp3 works on vec3, dp4 on vec4 */
1587 if (curOpcode
->opcode
== WINED3DSIO_DP4
) {
1588 src_write_mask
= WINED3DSP_WRITEMASK_ALL
;
1590 src_write_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
1593 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], src_write_mask
, &src0_param
);
1594 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], src_write_mask
, &src1_param
);
1597 shader_addline(buffer
, "vec%d(dot(%s, %s)));\n", dst_size
, src0_param
.param_str
, src1_param
.param_str
);
1599 shader_addline(buffer
, "dot(%s, %s));\n", src0_param
.param_str
, src1_param
.param_str
);
1603 /* Note that this instruction has some restrictions. The destination write mask
1604 * can't contain the w component, and the source swizzles have to be .xyzw */
1605 static void shader_glsl_cross(const SHADER_OPCODE_ARG
*arg
)
1607 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
1608 glsl_src_param_t src0_param
;
1609 glsl_src_param_t src1_param
;
1612 shader_glsl_get_write_mask(arg
->dst
, dst_mask
);
1613 shader_glsl_append_dst(arg
->buffer
, arg
);
1614 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], src_mask
, &src0_param
);
1615 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], src_mask
, &src1_param
);
1616 shader_addline(arg
->buffer
, "cross(%s, %s)%s);\n", src0_param
.param_str
, src1_param
.param_str
, dst_mask
);
1619 /* Process the WINED3DSIO_POW instruction in GLSL (dst = |src0|^src1)
1620 * Src0 and src1 are scalars. Note that D3D uses the absolute of src0, while
1621 * GLSL uses the value as-is. */
1622 static void shader_glsl_pow(const SHADER_OPCODE_ARG
*arg
)
1624 SHADER_BUFFER
*buffer
= arg
->buffer
;
1625 glsl_src_param_t src0_param
;
1626 glsl_src_param_t src1_param
;
1627 DWORD dst_write_mask
;
1628 unsigned int dst_size
;
1630 dst_write_mask
= shader_glsl_append_dst(buffer
, arg
);
1631 dst_size
= shader_glsl_get_write_mask_size(dst_write_mask
);
1633 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
1634 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], WINED3DSP_WRITEMASK_0
, &src1_param
);
1637 shader_addline(buffer
, "vec%d(pow(abs(%s), %s)));\n", dst_size
, src0_param
.param_str
, src1_param
.param_str
);
1639 shader_addline(buffer
, "pow(abs(%s), %s));\n", src0_param
.param_str
, src1_param
.param_str
);
1643 /* Process the WINED3DSIO_LOG instruction in GLSL (dst = log2(|src0|))
1644 * Src0 is a scalar. Note that D3D uses the absolute of src0, while
1645 * GLSL uses the value as-is. */
1646 static void shader_glsl_log(const SHADER_OPCODE_ARG
*arg
)
1648 SHADER_BUFFER
*buffer
= arg
->buffer
;
1649 glsl_src_param_t src0_param
;
1650 DWORD dst_write_mask
;
1651 unsigned int dst_size
;
1653 dst_write_mask
= shader_glsl_append_dst(buffer
, arg
);
1654 dst_size
= shader_glsl_get_write_mask_size(dst_write_mask
);
1656 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
1659 shader_addline(buffer
, "vec%d(log2(abs(%s))));\n", dst_size
, src0_param
.param_str
);
1661 shader_addline(buffer
, "log2(abs(%s)));\n", src0_param
.param_str
);
1665 /* Map the opcode 1-to-1 to the GL code (arg->dst = instruction(src0, src1, ...) */
1666 static void shader_glsl_map2gl(const SHADER_OPCODE_ARG
*arg
)
1668 CONST SHADER_OPCODE
* curOpcode
= arg
->opcode
;
1669 SHADER_BUFFER
* buffer
= arg
->buffer
;
1670 glsl_src_param_t src_param
;
1671 const char *instruction
;
1672 char arguments
[256];
1676 /* Determine the GLSL function to use based on the opcode */
1677 /* TODO: Possibly make this a table for faster lookups */
1678 switch (curOpcode
->opcode
) {
1679 case WINED3DSIO_MIN
: instruction
= "min"; break;
1680 case WINED3DSIO_MAX
: instruction
= "max"; break;
1681 case WINED3DSIO_ABS
: instruction
= "abs"; break;
1682 case WINED3DSIO_FRC
: instruction
= "fract"; break;
1683 case WINED3DSIO_NRM
: instruction
= "normalize"; break;
1684 case WINED3DSIO_EXP
: instruction
= "exp2"; break;
1685 case WINED3DSIO_SGN
: instruction
= "sign"; break;
1686 case WINED3DSIO_DSX
: instruction
= "dFdx"; break;
1687 case WINED3DSIO_DSY
: instruction
= "ycorrection.y * dFdy"; break;
1688 default: instruction
= "";
1689 FIXME("Opcode %s not yet handled in GLSL\n", curOpcode
->name
);
1693 write_mask
= shader_glsl_append_dst(buffer
, arg
);
1695 arguments
[0] = '\0';
1696 if (curOpcode
->num_params
> 0) {
1697 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], write_mask
, &src_param
);
1698 strcat(arguments
, src_param
.param_str
);
1699 for (i
= 2; i
< curOpcode
->num_params
; ++i
) {
1700 strcat(arguments
, ", ");
1701 shader_glsl_add_src_param(arg
, arg
->src
[i
-1], arg
->src_addr
[i
-1], write_mask
, &src_param
);
1702 strcat(arguments
, src_param
.param_str
);
1706 shader_addline(buffer
, "%s(%s));\n", instruction
, arguments
);
1709 /** Process the WINED3DSIO_EXPP instruction in GLSL:
1710 * For shader model 1.x, do the following (and honor the writemask, so use a temporary variable):
1711 * dst.x = 2^(floor(src))
1712 * dst.y = src - floor(src)
1713 * dst.z = 2^src (partial precision is allowed, but optional)
1715 * For 2.0 shaders, just do this (honoring writemask and swizzle):
1716 * dst = 2^src; (partial precision is allowed, but optional)
1718 static void shader_glsl_expp(const SHADER_OPCODE_ARG
*arg
)
1720 glsl_src_param_t src_param
;
1722 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src_param
);
1724 if (arg
->reg_maps
->shader_version
< WINED3DPS_VERSION(2,0))
1728 shader_addline(arg
->buffer
, "tmp0.x = exp2(floor(%s));\n", src_param
.param_str
);
1729 shader_addline(arg
->buffer
, "tmp0.y = %s - floor(%s);\n", src_param
.param_str
, src_param
.param_str
);
1730 shader_addline(arg
->buffer
, "tmp0.z = exp2(%s);\n", src_param
.param_str
);
1731 shader_addline(arg
->buffer
, "tmp0.w = 1.0;\n");
1733 shader_glsl_append_dst(arg
->buffer
, arg
);
1734 shader_glsl_get_write_mask(arg
->dst
, dst_mask
);
1735 shader_addline(arg
->buffer
, "tmp0%s);\n", dst_mask
);
1738 unsigned int mask_size
;
1740 write_mask
= shader_glsl_append_dst(arg
->buffer
, arg
);
1741 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
1743 if (mask_size
> 1) {
1744 shader_addline(arg
->buffer
, "vec%d(exp2(%s)));\n", mask_size
, src_param
.param_str
);
1746 shader_addline(arg
->buffer
, "exp2(%s));\n", src_param
.param_str
);
1751 /** Process the RCP (reciprocal or inverse) opcode in GLSL (dst = 1 / src) */
1752 static void shader_glsl_rcp(const SHADER_OPCODE_ARG
*arg
)
1754 glsl_src_param_t src_param
;
1756 unsigned int mask_size
;
1758 write_mask
= shader_glsl_append_dst(arg
->buffer
, arg
);
1759 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
1760 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_3
, &src_param
);
1762 if (mask_size
> 1) {
1763 shader_addline(arg
->buffer
, "vec%d(1.0 / %s));\n", mask_size
, src_param
.param_str
);
1765 shader_addline(arg
->buffer
, "1.0 / %s);\n", src_param
.param_str
);
1769 static void shader_glsl_rsq(const SHADER_OPCODE_ARG
*arg
)
1771 SHADER_BUFFER
* buffer
= arg
->buffer
;
1772 glsl_src_param_t src_param
;
1774 unsigned int mask_size
;
1776 write_mask
= shader_glsl_append_dst(buffer
, arg
);
1777 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
1779 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_3
, &src_param
);
1781 if (mask_size
> 1) {
1782 shader_addline(buffer
, "vec%d(inversesqrt(%s)));\n", mask_size
, src_param
.param_str
);
1784 shader_addline(buffer
, "inversesqrt(%s));\n", src_param
.param_str
);
1788 /** Process signed comparison opcodes in GLSL. */
1789 static void shader_glsl_compare(const SHADER_OPCODE_ARG
*arg
)
1791 glsl_src_param_t src0_param
;
1792 glsl_src_param_t src1_param
;
1794 unsigned int mask_size
;
1796 write_mask
= shader_glsl_append_dst(arg
->buffer
, arg
);
1797 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
1798 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], write_mask
, &src0_param
);
1799 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], write_mask
, &src1_param
);
1801 if (mask_size
> 1) {
1802 const char *compare
;
1804 switch(arg
->opcode
->opcode
) {
1805 case WINED3DSIO_SLT
: compare
= "lessThan"; break;
1806 case WINED3DSIO_SGE
: compare
= "greaterThanEqual"; break;
1807 default: compare
= "";
1808 FIXME("Can't handle opcode %s\n", arg
->opcode
->name
);
1811 shader_addline(arg
->buffer
, "vec%d(%s(%s, %s)));\n", mask_size
, compare
,
1812 src0_param
.param_str
, src1_param
.param_str
);
1814 switch(arg
->opcode
->opcode
) {
1815 case WINED3DSIO_SLT
:
1816 /* Step(src0, src1) is not suitable here because if src0 == src1 SLT is supposed,
1817 * to return 0.0 but step returns 1.0 because step is not < x
1818 * An alternative is a bvec compare padded with an unused second component.
1819 * step(src1 * -1.0, src0 * -1.0) is not an option because it suffers from the same
1820 * issue. Playing with not() is not possible either because not() does not accept
1823 shader_addline(arg
->buffer
, "(%s < %s) ? 1.0 : 0.0);\n", src0_param
.param_str
, src1_param
.param_str
);
1825 case WINED3DSIO_SGE
:
1826 /* Here we can use the step() function and safe a conditional */
1827 shader_addline(arg
->buffer
, "step(%s, %s));\n", src1_param
.param_str
, src0_param
.param_str
);
1830 FIXME("Can't handle opcode %s\n", arg
->opcode
->name
);
1836 /** Process CMP instruction in GLSL (dst = src0 >= 0.0 ? src1 : src2), per channel */
1837 static void shader_glsl_cmp(const SHADER_OPCODE_ARG
*arg
)
1839 glsl_src_param_t src0_param
;
1840 glsl_src_param_t src1_param
;
1841 glsl_src_param_t src2_param
;
1842 DWORD write_mask
, cmp_channel
= 0;
1845 BOOL temp_destination
= FALSE
;
1847 if(shader_is_scalar(arg
->src
[0])) {
1848 write_mask
= shader_glsl_append_dst(arg
->buffer
, arg
);
1850 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_ALL
, &src0_param
);
1851 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], write_mask
, &src1_param
);
1852 shader_glsl_add_src_param(arg
, arg
->src
[2], arg
->src_addr
[2], write_mask
, &src2_param
);
1854 shader_addline(arg
->buffer
, "%s >= 0.0 ? %s : %s);\n",
1855 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
1857 DWORD src0reg
= arg
->src
[0] & WINED3DSP_REGNUM_MASK
;
1858 DWORD src1reg
= arg
->src
[1] & WINED3DSP_REGNUM_MASK
;
1859 DWORD src2reg
= arg
->src
[2] & WINED3DSP_REGNUM_MASK
;
1860 DWORD src0regtype
= shader_get_regtype(arg
->src
[0]);
1861 DWORD src1regtype
= shader_get_regtype(arg
->src
[1]);
1862 DWORD src2regtype
= shader_get_regtype(arg
->src
[2]);
1863 DWORD dstreg
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
1864 DWORD dstregtype
= shader_get_regtype(arg
->dst
);
1866 /* Cycle through all source0 channels */
1867 for (i
=0; i
<4; i
++) {
1869 /* Find the destination channels which use the current source0 channel */
1870 for (j
=0; j
<4; j
++) {
1871 if ( ((arg
->src
[0] >> (WINED3DSP_SWIZZLE_SHIFT
+ 2*j
)) & 0x3) == i
) {
1872 write_mask
|= WINED3DSP_WRITEMASK_0
<< j
;
1873 cmp_channel
= WINED3DSP_WRITEMASK_0
<< j
;
1877 /* Splitting the cmp instruction up in multiple lines imposes a problem:
1878 * The first lines may overwrite source parameters of the following lines.
1879 * Deal with that by using a temporary destination register if needed
1881 if((src0reg
== dstreg
&& src0regtype
== dstregtype
) ||
1882 (src1reg
== dstreg
&& src1regtype
== dstregtype
) ||
1883 (src2reg
== dstreg
&& src2regtype
== dstregtype
)) {
1885 write_mask
= shader_glsl_get_write_mask(arg
->dst
& (~WINED3DSP_SWIZZLE_MASK
| write_mask
), mask_char
);
1886 if (!write_mask
) continue;
1887 shader_addline(arg
->buffer
, "tmp0%s = (", mask_char
);
1888 temp_destination
= TRUE
;
1890 write_mask
= shader_glsl_append_dst_ext(arg
->buffer
, arg
, arg
->dst
& (~WINED3DSP_SWIZZLE_MASK
| write_mask
));
1891 if (!write_mask
) continue;
1894 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], cmp_channel
, &src0_param
);
1895 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], write_mask
, &src1_param
);
1896 shader_glsl_add_src_param(arg
, arg
->src
[2], arg
->src_addr
[2], write_mask
, &src2_param
);
1898 shader_addline(arg
->buffer
, "%s >= 0.0 ? %s : %s);\n",
1899 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
1902 if(temp_destination
) {
1903 shader_glsl_get_write_mask(arg
->dst
, mask_char
);
1904 shader_glsl_append_dst_ext(arg
->buffer
, arg
, arg
->dst
);
1905 shader_addline(arg
->buffer
, "tmp0%s);\n", mask_char
);
1911 /** Process the CND opcode in GLSL (dst = (src0 > 0.5) ? src1 : src2) */
1912 /* For ps 1.1-1.3, only a single component of src0 is used. For ps 1.4
1913 * the compare is done per component of src0. */
1914 static void shader_glsl_cnd(const SHADER_OPCODE_ARG
*arg
)
1916 glsl_src_param_t src0_param
;
1917 glsl_src_param_t src1_param
;
1918 glsl_src_param_t src2_param
;
1919 DWORD write_mask
, cmp_channel
= 0;
1922 if (arg
->reg_maps
->shader_version
< WINED3DPS_VERSION(1, 4))
1924 write_mask
= shader_glsl_append_dst(arg
->buffer
, arg
);
1925 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
1926 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], write_mask
, &src1_param
);
1927 shader_glsl_add_src_param(arg
, arg
->src
[2], arg
->src_addr
[2], write_mask
, &src2_param
);
1929 /* Fun: The D3DSI_COISSUE flag changes the semantic of the cnd instruction for < 1.4 shaders */
1930 if(arg
->opcode_token
& WINED3DSI_COISSUE
) {
1931 shader_addline(arg
->buffer
, "%s /* COISSUE! */);\n", src1_param
.param_str
);
1933 shader_addline(arg
->buffer
, "%s > 0.5 ? %s : %s);\n",
1934 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
1938 /* Cycle through all source0 channels */
1939 for (i
=0; i
<4; i
++) {
1941 /* Find the destination channels which use the current source0 channel */
1942 for (j
=0; j
<4; j
++) {
1943 if ( ((arg
->src
[0] >> (WINED3DSP_SWIZZLE_SHIFT
+ 2*j
)) & 0x3) == i
) {
1944 write_mask
|= WINED3DSP_WRITEMASK_0
<< j
;
1945 cmp_channel
= WINED3DSP_WRITEMASK_0
<< j
;
1948 write_mask
= shader_glsl_append_dst_ext(arg
->buffer
, arg
, arg
->dst
& (~WINED3DSP_SWIZZLE_MASK
| write_mask
));
1949 if (!write_mask
) continue;
1951 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], cmp_channel
, &src0_param
);
1952 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], write_mask
, &src1_param
);
1953 shader_glsl_add_src_param(arg
, arg
->src
[2], arg
->src_addr
[2], write_mask
, &src2_param
);
1955 shader_addline(arg
->buffer
, "%s > 0.5 ? %s : %s);\n",
1956 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
1960 /** GLSL code generation for WINED3DSIO_MAD: Multiply the first 2 opcodes, then add the last */
1961 static void shader_glsl_mad(const SHADER_OPCODE_ARG
*arg
)
1963 glsl_src_param_t src0_param
;
1964 glsl_src_param_t src1_param
;
1965 glsl_src_param_t src2_param
;
1968 write_mask
= shader_glsl_append_dst(arg
->buffer
, arg
);
1969 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], write_mask
, &src0_param
);
1970 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], write_mask
, &src1_param
);
1971 shader_glsl_add_src_param(arg
, arg
->src
[2], arg
->src_addr
[2], write_mask
, &src2_param
);
1972 shader_addline(arg
->buffer
, "(%s * %s) + %s);\n",
1973 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
1976 /** Handles transforming all WINED3DSIO_M?x? opcodes for
1977 Vertex shaders to GLSL codes */
1978 static void shader_glsl_mnxn(const SHADER_OPCODE_ARG
*arg
)
1980 IWineD3DBaseShaderImpl
*shader
= (IWineD3DBaseShaderImpl
*)arg
->shader
;
1981 const SHADER_OPCODE
*opcode_table
= shader
->baseShader
.shader_ins
;
1982 DWORD shader_version
= arg
->reg_maps
->shader_version
;
1984 int nComponents
= 0;
1985 SHADER_OPCODE_ARG tmpArg
;
1987 memset(&tmpArg
, 0, sizeof(SHADER_OPCODE_ARG
));
1989 /* Set constants for the temporary argument */
1990 tmpArg
.shader
= arg
->shader
;
1991 tmpArg
.buffer
= arg
->buffer
;
1992 tmpArg
.src
[0] = arg
->src
[0];
1993 tmpArg
.src_addr
[0] = arg
->src_addr
[0];
1994 tmpArg
.src_addr
[1] = arg
->src_addr
[1];
1995 tmpArg
.reg_maps
= arg
->reg_maps
;
1997 switch(arg
->opcode
->opcode
) {
1998 case WINED3DSIO_M4x4
:
2000 tmpArg
.opcode
= shader_get_opcode(opcode_table
, shader_version
, WINED3DSIO_DP4
);
2002 case WINED3DSIO_M4x3
:
2004 tmpArg
.opcode
= shader_get_opcode(opcode_table
, shader_version
, WINED3DSIO_DP4
);
2006 case WINED3DSIO_M3x4
:
2008 tmpArg
.opcode
= shader_get_opcode(opcode_table
, shader_version
, WINED3DSIO_DP3
);
2010 case WINED3DSIO_M3x3
:
2012 tmpArg
.opcode
= shader_get_opcode(opcode_table
, shader_version
, WINED3DSIO_DP3
);
2014 case WINED3DSIO_M3x2
:
2016 tmpArg
.opcode
= shader_get_opcode(opcode_table
, shader_version
, WINED3DSIO_DP3
);
2022 for (i
= 0; i
< nComponents
; i
++) {
2023 tmpArg
.dst
= ((arg
->dst
) & ~WINED3DSP_WRITEMASK_ALL
)|(WINED3DSP_WRITEMASK_0
<<i
);
2024 tmpArg
.src
[1] = arg
->src
[1]+i
;
2025 shader_glsl_dot(&tmpArg
);
2030 The LRP instruction performs a component-wise linear interpolation
2031 between the second and third operands using the first operand as the
2032 blend factor. Equation: (dst = src2 + src0 * (src1 - src2))
2033 This is equivalent to mix(src2, src1, src0);
2035 static void shader_glsl_lrp(const SHADER_OPCODE_ARG
*arg
)
2037 glsl_src_param_t src0_param
;
2038 glsl_src_param_t src1_param
;
2039 glsl_src_param_t src2_param
;
2042 write_mask
= shader_glsl_append_dst(arg
->buffer
, arg
);
2044 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], write_mask
, &src0_param
);
2045 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], write_mask
, &src1_param
);
2046 shader_glsl_add_src_param(arg
, arg
->src
[2], arg
->src_addr
[2], write_mask
, &src2_param
);
2048 shader_addline(arg
->buffer
, "mix(%s, %s, %s));\n",
2049 src2_param
.param_str
, src1_param
.param_str
, src0_param
.param_str
);
2052 /** Process the WINED3DSIO_LIT instruction in GLSL:
2053 * dst.x = dst.w = 1.0
2054 * dst.y = (src0.x > 0) ? src0.x
2055 * dst.z = (src0.x > 0) ? ((src0.y > 0) ? pow(src0.y, src.w) : 0) : 0
2056 * where src.w is clamped at +- 128
2058 static void shader_glsl_lit(const SHADER_OPCODE_ARG
*arg
)
2060 glsl_src_param_t src0_param
;
2061 glsl_src_param_t src1_param
;
2062 glsl_src_param_t src3_param
;
2065 shader_glsl_append_dst(arg
->buffer
, arg
);
2066 shader_glsl_get_write_mask(arg
->dst
, dst_mask
);
2068 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2069 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_1
, &src1_param
);
2070 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_3
, &src3_param
);
2072 /* The sdk specifies the instruction like this
2074 * if(src.x > 0.0) dst.y = src.x
2076 * if(src.x > 0.0 && src.y > 0.0) dst.z = pow(src.y, power);
2080 * Obviously that has quite a few conditionals in it which we don't like. So the first step is this:
2081 * dst.x = 1.0 ... No further explanation needed
2082 * dst.y = max(src.y, 0.0); ... If x < 0.0, use 0.0, otherwise x. Same as the conditional
2083 * dst.z = x > 0.0 ? pow(max(y, 0.0), p) : 0; ... 0 ^ power is 0, and otherwise we use y anyway
2084 * dst.w = 1.0. ... Nothing fancy.
2086 * So we still have one conditional in there. So do this:
2087 * dst.z = pow(max(0.0, src.y) * step(0.0, src.x), power);
2089 * 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),
2090 * 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.
2091 * if both x and y are > 0, we get pow(y * 1.0, power), as it is supposed to
2093 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",
2094 src0_param
.param_str
, src1_param
.param_str
, src0_param
.param_str
, src3_param
.param_str
, dst_mask
);
2097 /** Process the WINED3DSIO_DST instruction in GLSL:
2099 * dst.y = src0.x * src0.y
2103 static void shader_glsl_dst(const SHADER_OPCODE_ARG
*arg
)
2105 glsl_src_param_t src0y_param
;
2106 glsl_src_param_t src0z_param
;
2107 glsl_src_param_t src1y_param
;
2108 glsl_src_param_t src1w_param
;
2111 shader_glsl_append_dst(arg
->buffer
, arg
);
2112 shader_glsl_get_write_mask(arg
->dst
, dst_mask
);
2114 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_1
, &src0y_param
);
2115 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_2
, &src0z_param
);
2116 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], WINED3DSP_WRITEMASK_1
, &src1y_param
);
2117 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], WINED3DSP_WRITEMASK_3
, &src1w_param
);
2119 shader_addline(arg
->buffer
, "vec4(1.0, %s * %s, %s, %s))%s;\n",
2120 src0y_param
.param_str
, src1y_param
.param_str
, src0z_param
.param_str
, src1w_param
.param_str
, dst_mask
);
2123 /** Process the WINED3DSIO_SINCOS instruction in GLSL:
2124 * VS 2.0 requires that specific cosine and sine constants be passed to this instruction so the hardware
2125 * can handle it. But, these functions are built-in for GLSL, so we can just ignore the last 2 params.
2127 * dst.x = cos(src0.?)
2128 * dst.y = sin(src0.?)
2132 static void shader_glsl_sincos(const SHADER_OPCODE_ARG
*arg
)
2134 glsl_src_param_t src0_param
;
2137 write_mask
= shader_glsl_append_dst(arg
->buffer
, arg
);
2138 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2140 switch (write_mask
) {
2141 case WINED3DSP_WRITEMASK_0
:
2142 shader_addline(arg
->buffer
, "cos(%s));\n", src0_param
.param_str
);
2145 case WINED3DSP_WRITEMASK_1
:
2146 shader_addline(arg
->buffer
, "sin(%s));\n", src0_param
.param_str
);
2149 case (WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
):
2150 shader_addline(arg
->buffer
, "vec2(cos(%s), sin(%s)));\n", src0_param
.param_str
, src0_param
.param_str
);
2154 ERR("Write mask should be .x, .y or .xy\n");
2159 /** Process the WINED3DSIO_LOOP instruction in GLSL:
2160 * Start a for() loop where src1.y is the initial value of aL,
2161 * increment aL by src1.z for a total of src1.x iterations.
2162 * Need to use a temporary variable for this operation.
2164 /* FIXME: I don't think nested loops will work correctly this way. */
2165 static void shader_glsl_loop(const SHADER_OPCODE_ARG
*arg
)
2167 glsl_src_param_t src1_param
;
2168 IWineD3DBaseShaderImpl
* shader
= (IWineD3DBaseShaderImpl
*) arg
->shader
;
2169 DWORD regtype
= shader_get_regtype(arg
->src
[1]);
2170 DWORD reg
= arg
->src
[1] & WINED3DSP_REGNUM_MASK
;
2171 const DWORD
*control_values
= NULL
;
2172 const local_constant
*constant
;
2174 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], WINED3DSP_WRITEMASK_ALL
, &src1_param
);
2176 /* Try to hardcode the loop control parameters if possible. Direct3D 9 class hardware doesn't support real
2177 * varying indexing, but Microsoft designed this feature for Shader model 2.x+. If the loop control is
2178 * known at compile time, the GLSL compiler can unroll the loop, and replace indirect addressing with direct
2181 if(regtype
== WINED3DSPR_CONSTINT
) {
2182 LIST_FOR_EACH_ENTRY(constant
, &shader
->baseShader
.constantsI
, local_constant
, entry
) {
2183 if(constant
->idx
== reg
) {
2184 control_values
= constant
->value
;
2190 if(control_values
) {
2191 if(control_values
[2] > 0) {
2192 shader_addline(arg
->buffer
, "for (aL%u = %d; aL%u < (%d * %d + %d); aL%u += %d) {\n",
2193 shader
->baseShader
.cur_loop_depth
, control_values
[1],
2194 shader
->baseShader
.cur_loop_depth
, control_values
[0], control_values
[2], control_values
[1],
2195 shader
->baseShader
.cur_loop_depth
, control_values
[2]);
2196 } else if(control_values
[2] == 0) {
2197 shader_addline(arg
->buffer
, "for (aL%u = %d, tmpInt%u = 0; tmpInt%u < %d; tmpInt%u++) {\n",
2198 shader
->baseShader
.cur_loop_depth
, control_values
[1], shader
->baseShader
.cur_loop_depth
,
2199 shader
->baseShader
.cur_loop_depth
, control_values
[0],
2200 shader
->baseShader
.cur_loop_depth
);
2202 shader_addline(arg
->buffer
, "for (aL%u = %d; aL%u > (%d * %d + %d); aL%u += %d) {\n",
2203 shader
->baseShader
.cur_loop_depth
, control_values
[1],
2204 shader
->baseShader
.cur_loop_depth
, control_values
[0], control_values
[2], control_values
[1],
2205 shader
->baseShader
.cur_loop_depth
, control_values
[2]);
2208 shader_addline(arg
->buffer
, "for (tmpInt%u = 0, aL%u = %s.y; tmpInt%u < %s.x; tmpInt%u++, aL%u += %s.z) {\n",
2209 shader
->baseShader
.cur_loop_depth
, shader
->baseShader
.cur_loop_regno
,
2210 src1_param
.reg_name
, shader
->baseShader
.cur_loop_depth
, src1_param
.reg_name
,
2211 shader
->baseShader
.cur_loop_depth
, shader
->baseShader
.cur_loop_regno
, src1_param
.reg_name
);
2214 shader
->baseShader
.cur_loop_depth
++;
2215 shader
->baseShader
.cur_loop_regno
++;
2218 static void shader_glsl_end(const SHADER_OPCODE_ARG
*arg
)
2220 IWineD3DBaseShaderImpl
* shader
= (IWineD3DBaseShaderImpl
*) arg
->shader
;
2222 shader_addline(arg
->buffer
, "}\n");
2224 if(arg
->opcode
->opcode
== WINED3DSIO_ENDLOOP
) {
2225 shader
->baseShader
.cur_loop_depth
--;
2226 shader
->baseShader
.cur_loop_regno
--;
2228 if(arg
->opcode
->opcode
== WINED3DSIO_ENDREP
) {
2229 shader
->baseShader
.cur_loop_depth
--;
2233 static void shader_glsl_rep(const SHADER_OPCODE_ARG
*arg
)
2235 IWineD3DBaseShaderImpl
* shader
= (IWineD3DBaseShaderImpl
*) arg
->shader
;
2236 glsl_src_param_t src0_param
;
2238 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2239 shader_addline(arg
->buffer
, "for (tmpInt%d = 0; tmpInt%d < %s; tmpInt%d++) {\n",
2240 shader
->baseShader
.cur_loop_depth
, shader
->baseShader
.cur_loop_depth
,
2241 src0_param
.param_str
, shader
->baseShader
.cur_loop_depth
);
2242 shader
->baseShader
.cur_loop_depth
++;
2245 static void shader_glsl_if(const SHADER_OPCODE_ARG
*arg
)
2247 glsl_src_param_t src0_param
;
2249 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2250 shader_addline(arg
->buffer
, "if (%s) {\n", src0_param
.param_str
);
2253 static void shader_glsl_ifc(const SHADER_OPCODE_ARG
*arg
)
2255 glsl_src_param_t src0_param
;
2256 glsl_src_param_t src1_param
;
2258 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2259 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], WINED3DSP_WRITEMASK_0
, &src1_param
);
2261 shader_addline(arg
->buffer
, "if (%s %s %s) {\n",
2262 src0_param
.param_str
, shader_get_comp_op(arg
->opcode_token
), src1_param
.param_str
);
2265 static void shader_glsl_else(const SHADER_OPCODE_ARG
*arg
)
2267 shader_addline(arg
->buffer
, "} else {\n");
2270 static void shader_glsl_break(const SHADER_OPCODE_ARG
*arg
)
2272 shader_addline(arg
->buffer
, "break;\n");
2275 /* FIXME: According to MSDN the compare is done per component. */
2276 static void shader_glsl_breakc(const SHADER_OPCODE_ARG
*arg
)
2278 glsl_src_param_t src0_param
;
2279 glsl_src_param_t src1_param
;
2281 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2282 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], WINED3DSP_WRITEMASK_0
, &src1_param
);
2284 shader_addline(arg
->buffer
, "if (%s %s %s) break;\n",
2285 src0_param
.param_str
, shader_get_comp_op(arg
->opcode_token
), src1_param
.param_str
);
2288 static void shader_glsl_label(const SHADER_OPCODE_ARG
*arg
)
2291 DWORD snum
= (arg
->src
[0]) & WINED3DSP_REGNUM_MASK
;
2292 shader_addline(arg
->buffer
, "}\n");
2293 shader_addline(arg
->buffer
, "void subroutine%u () {\n", snum
);
2296 static void shader_glsl_call(const SHADER_OPCODE_ARG
*arg
)
2298 DWORD snum
= (arg
->src
[0]) & WINED3DSP_REGNUM_MASK
;
2299 shader_addline(arg
->buffer
, "subroutine%u();\n", snum
);
2302 static void shader_glsl_callnz(const SHADER_OPCODE_ARG
*arg
)
2304 glsl_src_param_t src1_param
;
2306 DWORD snum
= (arg
->src
[0]) & WINED3DSP_REGNUM_MASK
;
2307 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], WINED3DSP_WRITEMASK_0
, &src1_param
);
2308 shader_addline(arg
->buffer
, "if (%s) subroutine%u();\n", src1_param
.param_str
, snum
);
2311 /*********************************************
2312 * Pixel Shader Specific Code begins here
2313 ********************************************/
2314 static void pshader_glsl_tex(const SHADER_OPCODE_ARG
*arg
)
2316 IWineD3DPixelShaderImpl
* This
= (IWineD3DPixelShaderImpl
*) arg
->shader
;
2317 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
2318 DWORD shader_version
= arg
->reg_maps
->shader_version
;
2319 glsl_sample_function_t sample_function
;
2320 DWORD sample_flags
= 0;
2323 DWORD mask
= 0, swizzle
;
2325 /* 1.0-1.4: Use destination register as sampler source.
2326 * 2.0+: Use provided sampler source. */
2327 if (shader_version
< WINED3DPS_VERSION(2,0)) sampler_idx
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2328 else sampler_idx
= arg
->src
[1] & WINED3DSP_REGNUM_MASK
;
2329 sampler_type
= arg
->reg_maps
->samplers
[sampler_idx
] & WINED3DSP_TEXTURETYPE_MASK
;
2331 if (shader_version
< WINED3DPS_VERSION(1,4))
2333 DWORD flags
= deviceImpl
->stateBlock
->textureState
[sampler_idx
][WINED3DTSS_TEXTURETRANSFORMFLAGS
];
2335 /* Projected cube textures don't make a lot of sense, the resulting coordinates stay the same. */
2336 if (flags
& WINED3DTTFF_PROJECTED
&& sampler_type
!= WINED3DSTT_CUBE
) {
2337 sample_flags
|= WINED3D_GLSL_SAMPLE_PROJECTED
;
2338 switch (flags
& ~WINED3DTTFF_PROJECTED
) {
2339 case WINED3DTTFF_COUNT1
: FIXME("WINED3DTTFF_PROJECTED with WINED3DTTFF_COUNT1?\n"); break;
2340 case WINED3DTTFF_COUNT2
: mask
= WINED3DSP_WRITEMASK_1
; break;
2341 case WINED3DTTFF_COUNT3
: mask
= WINED3DSP_WRITEMASK_2
; break;
2342 case WINED3DTTFF_COUNT4
:
2343 case WINED3DTTFF_DISABLE
: mask
= WINED3DSP_WRITEMASK_3
; break;
2347 else if (shader_version
< WINED3DPS_VERSION(2,0))
2349 DWORD src_mod
= arg
->src
[0] & WINED3DSP_SRCMOD_MASK
;
2351 if (src_mod
== WINED3DSPSM_DZ
) {
2352 sample_flags
|= WINED3D_GLSL_SAMPLE_PROJECTED
;
2353 mask
= WINED3DSP_WRITEMASK_2
;
2354 } else if (src_mod
== WINED3DSPSM_DW
) {
2355 sample_flags
|= WINED3D_GLSL_SAMPLE_PROJECTED
;
2356 mask
= WINED3DSP_WRITEMASK_3
;
2359 if(arg
->opcode_token
& WINED3DSI_TEXLD_PROJECT
) {
2360 /* ps 2.0 texldp instruction always divides by the fourth component. */
2361 sample_flags
|= WINED3D_GLSL_SAMPLE_PROJECTED
;
2362 mask
= WINED3DSP_WRITEMASK_3
;
2366 if(deviceImpl
->stateBlock
->textures
[sampler_idx
] &&
2367 IWineD3DBaseTexture_GetTextureDimensions(deviceImpl
->stateBlock
->textures
[sampler_idx
]) == GL_TEXTURE_RECTANGLE_ARB
) {
2368 sample_flags
|= WINED3D_GLSL_SAMPLE_RECT
;
2371 shader_glsl_get_sample_function(sampler_type
, sample_flags
, &sample_function
);
2372 mask
|= sample_function
.coord_mask
;
2374 if (shader_version
< WINED3DPS_VERSION(2,0)) swizzle
= WINED3DVS_NOSWIZZLE
;
2375 else swizzle
= arg
->src
[1] & WINED3DSP_SWIZZLE_MASK
;
2377 /* 1.0-1.3: Use destination register as coordinate source.
2378 1.4+: Use provided coordinate source register. */
2379 if (shader_version
< WINED3DPS_VERSION(1,4))
2381 char coord_mask
[6], srcreg
[12];
2382 shader_glsl_get_write_mask(mask
, coord_mask
);
2383 sprintf(srcreg
, "T%u%s", sampler_idx
, coord_mask
);
2384 shader_glsl_gen_sample_code(arg
, sampler_idx
, srcreg
, &sample_function
, swizzle
, NULL
);
2386 glsl_src_param_t coord_param
;
2387 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], mask
, &coord_param
);
2388 if(arg
->opcode_token
& WINED3DSI_TEXLD_BIAS
) {
2389 glsl_src_param_t bias
;
2390 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_3
, &bias
);
2391 shader_glsl_gen_sample_code(arg
, sampler_idx
, coord_param
.param_str
, &sample_function
,
2392 swizzle
, bias
.param_str
);
2394 shader_glsl_gen_sample_code(arg
, sampler_idx
, coord_param
.param_str
, &sample_function
,
2400 static void shader_glsl_texldl(const SHADER_OPCODE_ARG
*arg
)
2402 IWineD3DBaseShaderImpl
* This
= (IWineD3DBaseShaderImpl
*)arg
->shader
;
2403 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
2404 glsl_sample_function_t sample_function
;
2405 glsl_src_param_t coord_param
, lod_param
;
2406 DWORD sample_flags
= WINED3D_GLSL_SAMPLE_LOD
;
2409 DWORD swizzle
= arg
->src
[1] & WINED3DSP_SWIZZLE_MASK
;
2411 shader_glsl_append_dst(arg
->buffer
, arg
);
2413 sampler_idx
= arg
->src
[1] & WINED3DSP_REGNUM_MASK
;
2414 sampler_type
= arg
->reg_maps
->samplers
[sampler_idx
] & WINED3DSP_TEXTURETYPE_MASK
;
2415 if(deviceImpl
->stateBlock
->textures
[sampler_idx
] &&
2416 IWineD3DBaseTexture_GetTextureDimensions(deviceImpl
->stateBlock
->textures
[sampler_idx
]) == GL_TEXTURE_RECTANGLE_ARB
) {
2417 sample_flags
|= WINED3D_GLSL_SAMPLE_RECT
;
2419 shader_glsl_get_sample_function(sampler_type
, sample_flags
, &sample_function
);
2420 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], sample_function
.coord_mask
, &coord_param
);
2422 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_3
, &lod_param
);
2424 if (shader_is_pshader_version(arg
->reg_maps
->shader_version
))
2426 /* The GLSL spec claims the Lod sampling functions are only supported in vertex shaders.
2427 * However, they seem to work just fine in fragment shaders as well. */
2428 WARN("Using %s in fragment shader.\n", sample_function
.name
);
2430 shader_glsl_gen_sample_code(arg
, sampler_idx
, coord_param
.param_str
,
2431 &sample_function
, swizzle
,
2432 lod_param
.param_str
);
2435 static void pshader_glsl_texcoord(const SHADER_OPCODE_ARG
*arg
)
2437 /* FIXME: Make this work for more than just 2D textures */
2438 SHADER_BUFFER
* buffer
= arg
->buffer
;
2442 write_mask
= shader_glsl_append_dst(arg
->buffer
, arg
);
2443 shader_glsl_get_write_mask(write_mask
, dst_mask
);
2445 if (arg
->reg_maps
->shader_version
!= WINED3DPS_VERSION(1,4))
2447 DWORD reg
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2448 shader_addline(buffer
, "clamp(gl_TexCoord[%u], 0.0, 1.0)%s);\n", reg
, dst_mask
);
2450 DWORD reg
= arg
->src
[0] & WINED3DSP_REGNUM_MASK
;
2451 DWORD src_mod
= arg
->src
[0] & WINED3DSP_SRCMOD_MASK
;
2452 char dst_swizzle
[6];
2454 shader_glsl_get_swizzle(arg
->src
[0], FALSE
, write_mask
, dst_swizzle
);
2456 if (src_mod
== WINED3DSPSM_DZ
) {
2457 glsl_src_param_t div_param
;
2458 unsigned int mask_size
= shader_glsl_get_write_mask_size(write_mask
);
2459 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_2
, &div_param
);
2461 if (mask_size
> 1) {
2462 shader_addline(buffer
, "gl_TexCoord[%u]%s / vec%d(%s));\n", reg
, dst_swizzle
, mask_size
, div_param
.param_str
);
2464 shader_addline(buffer
, "gl_TexCoord[%u]%s / %s);\n", reg
, dst_swizzle
, div_param
.param_str
);
2466 } else if (src_mod
== WINED3DSPSM_DW
) {
2467 glsl_src_param_t div_param
;
2468 unsigned int mask_size
= shader_glsl_get_write_mask_size(write_mask
);
2469 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_3
, &div_param
);
2471 if (mask_size
> 1) {
2472 shader_addline(buffer
, "gl_TexCoord[%u]%s / vec%d(%s));\n", reg
, dst_swizzle
, mask_size
, div_param
.param_str
);
2474 shader_addline(buffer
, "gl_TexCoord[%u]%s / %s);\n", reg
, dst_swizzle
, div_param
.param_str
);
2477 shader_addline(buffer
, "gl_TexCoord[%u]%s);\n", reg
, dst_swizzle
);
2482 /** Process the WINED3DSIO_TEXDP3TEX instruction in GLSL:
2483 * Take a 3-component dot product of the TexCoord[dstreg] and src,
2484 * then perform a 1D texture lookup from stage dstregnum, place into dst. */
2485 static void pshader_glsl_texdp3tex(const SHADER_OPCODE_ARG
*arg
)
2487 glsl_src_param_t src0_param
;
2488 char coord_param
[64];
2489 glsl_sample_function_t sample_function
;
2490 DWORD sampler_idx
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2491 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2492 DWORD sampler_type
= arg
->reg_maps
->samplers
[sampler_idx
] & WINED3DSP_TEXTURETYPE_MASK
;
2494 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], src_mask
, &src0_param
);
2496 /* Do I have to take care about the projected bit? I don't think so, since the dp3 returns only one
2497 * scalar, and projected sampling would require 4.
2499 * It is a dependent read - not valid with conditional NP2 textures
2501 shader_glsl_get_sample_function(sampler_type
, 0, &sample_function
);
2503 switch(count_bits(sample_function
.coord_mask
)) {
2505 sprintf(coord_param
, "dot(gl_TexCoord[%u].xyz, %s)",
2506 sampler_idx
, src0_param
.param_str
);
2510 sprintf(coord_param
, "vec2(dot(gl_TexCoord[%u].xyz, %s), 0.0)",
2511 sampler_idx
, src0_param
.param_str
);
2515 sprintf(coord_param
, "vec3(dot(gl_TexCoord[%u].xyz, %s), 0.0, 0.0)",
2516 sampler_idx
, src0_param
.param_str
);
2519 FIXME("Unexpected mask bitcount %d\n", count_bits(sample_function
.coord_mask
));
2521 shader_glsl_gen_sample_code(arg
, sampler_idx
, coord_param
,
2522 &sample_function
, WINED3DVS_NOSWIZZLE
,
2526 /** Process the WINED3DSIO_TEXDP3 instruction in GLSL:
2527 * Take a 3-component dot product of the TexCoord[dstreg] and src. */
2528 static void pshader_glsl_texdp3(const SHADER_OPCODE_ARG
*arg
)
2530 glsl_src_param_t src0_param
;
2531 DWORD dstreg
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2532 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2534 unsigned int mask_size
;
2536 dst_mask
= shader_glsl_append_dst(arg
->buffer
, arg
);
2537 mask_size
= shader_glsl_get_write_mask_size(dst_mask
);
2538 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], src_mask
, &src0_param
);
2540 if (mask_size
> 1) {
2541 shader_addline(arg
->buffer
, "vec%d(dot(T%u.xyz, %s)));\n", mask_size
, dstreg
, src0_param
.param_str
);
2543 shader_addline(arg
->buffer
, "dot(T%u.xyz, %s));\n", dstreg
, src0_param
.param_str
);
2547 /** Process the WINED3DSIO_TEXDEPTH instruction in GLSL:
2548 * Calculate the depth as dst.x / dst.y */
2549 static void pshader_glsl_texdepth(const SHADER_OPCODE_ARG
*arg
)
2551 glsl_dst_param_t dst_param
;
2553 shader_glsl_add_dst_param(arg
, arg
->dst
, 0, &dst_param
);
2555 /* Tests show that texdepth never returns anything below 0.0, and that r5.y is clamped to 1.0.
2556 * Negative input is accepted, -0.25 / -0.5 returns 0.5. GL should clamp gl_FragDepth to [0;1], but
2557 * this doesn't always work, so clamp the results manually. Whether or not the x value is clamped at 1
2558 * too is irrelevant, since if x = 0, any y value < 1.0 (and > 1.0 is not allowed) results in a result
2561 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
);
2564 /** Process the WINED3DSIO_TEXM3X2DEPTH instruction in GLSL:
2565 * Last row of a 3x2 matrix multiply, use the result to calculate the depth:
2566 * Calculate tmp0.y = TexCoord[dstreg] . src.xyz; (tmp0.x has already been calculated)
2567 * depth = (tmp0.y == 0.0) ? 1.0 : tmp0.x / tmp0.y
2569 static void pshader_glsl_texm3x2depth(const SHADER_OPCODE_ARG
*arg
)
2571 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2572 DWORD dstreg
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2573 glsl_src_param_t src0_param
;
2575 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], src_mask
, &src0_param
);
2577 shader_addline(arg
->buffer
, "tmp0.y = dot(T%u.xyz, %s);\n", dstreg
, src0_param
.param_str
);
2578 shader_addline(arg
->buffer
, "gl_FragDepth = (tmp0.y == 0.0) ? 1.0 : clamp(tmp0.x / tmp0.y, 0.0, 1.0);\n");
2581 /** Process the WINED3DSIO_TEXM3X2PAD instruction in GLSL
2582 * Calculate the 1st of a 2-row matrix multiplication. */
2583 static void pshader_glsl_texm3x2pad(const SHADER_OPCODE_ARG
*arg
)
2585 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2586 DWORD reg
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2587 SHADER_BUFFER
* buffer
= arg
->buffer
;
2588 glsl_src_param_t src0_param
;
2590 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], src_mask
, &src0_param
);
2591 shader_addline(buffer
, "tmp0.x = dot(T%u.xyz, %s);\n", reg
, src0_param
.param_str
);
2594 /** Process the WINED3DSIO_TEXM3X3PAD instruction in GLSL
2595 * Calculate the 1st or 2nd row of a 3-row matrix multiplication. */
2596 static void pshader_glsl_texm3x3pad(const SHADER_OPCODE_ARG
* arg
)
2598 IWineD3DPixelShaderImpl
* shader
= (IWineD3DPixelShaderImpl
*) arg
->shader
;
2599 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2600 DWORD reg
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2601 SHADER_BUFFER
* buffer
= arg
->buffer
;
2602 SHADER_PARSE_STATE
* current_state
= &shader
->baseShader
.parse_state
;
2603 glsl_src_param_t src0_param
;
2605 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], src_mask
, &src0_param
);
2606 shader_addline(buffer
, "tmp0.%c = dot(T%u.xyz, %s);\n", 'x' + current_state
->current_row
, reg
, src0_param
.param_str
);
2607 current_state
->texcoord_w
[current_state
->current_row
++] = reg
;
2610 static void pshader_glsl_texm3x2tex(const SHADER_OPCODE_ARG
*arg
)
2612 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2613 DWORD reg
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2614 SHADER_BUFFER
* buffer
= arg
->buffer
;
2615 glsl_src_param_t src0_param
;
2616 DWORD sampler_type
= arg
->reg_maps
->samplers
[reg
] & WINED3DSP_TEXTURETYPE_MASK
;
2617 glsl_sample_function_t sample_function
;
2619 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], src_mask
, &src0_param
);
2620 shader_addline(buffer
, "tmp0.y = dot(T%u.xyz, %s);\n", reg
, src0_param
.param_str
);
2622 shader_glsl_get_sample_function(sampler_type
, 0, &sample_function
);
2624 /* Sample the texture using the calculated coordinates */
2625 shader_glsl_gen_sample_code(arg
, reg
, "tmp0.xy",
2626 &sample_function
, WINED3DVS_NOSWIZZLE
,
2630 /** Process the WINED3DSIO_TEXM3X3TEX instruction in GLSL
2631 * Perform the 3rd row of a 3x3 matrix multiply, then sample the texture using the calculated coordinates */
2632 static void pshader_glsl_texm3x3tex(const SHADER_OPCODE_ARG
*arg
)
2634 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2635 glsl_src_param_t src0_param
;
2636 DWORD reg
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2637 IWineD3DPixelShaderImpl
* This
= (IWineD3DPixelShaderImpl
*) arg
->shader
;
2638 SHADER_PARSE_STATE
* current_state
= &This
->baseShader
.parse_state
;
2639 DWORD sampler_type
= arg
->reg_maps
->samplers
[reg
] & WINED3DSP_TEXTURETYPE_MASK
;
2640 glsl_sample_function_t sample_function
;
2642 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], src_mask
, &src0_param
);
2643 shader_addline(arg
->buffer
, "tmp0.z = dot(T%u.xyz, %s);\n", reg
, src0_param
.param_str
);
2645 /* Dependent read, not valid with conditional NP2 */
2646 shader_glsl_get_sample_function(sampler_type
, 0, &sample_function
);
2648 /* Sample the texture using the calculated coordinates */
2649 shader_glsl_gen_sample_code(arg
, reg
, "tmp0.xyz",
2650 &sample_function
, WINED3DVS_NOSWIZZLE
,
2653 current_state
->current_row
= 0;
2656 /** Process the WINED3DSIO_TEXM3X3 instruction in GLSL
2657 * Perform the 3rd row of a 3x3 matrix multiply */
2658 static void pshader_glsl_texm3x3(const SHADER_OPCODE_ARG
*arg
)
2660 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2661 glsl_src_param_t src0_param
;
2663 DWORD reg
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2664 IWineD3DPixelShaderImpl
* This
= (IWineD3DPixelShaderImpl
*) arg
->shader
;
2665 SHADER_PARSE_STATE
* current_state
= &This
->baseShader
.parse_state
;
2667 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], src_mask
, &src0_param
);
2669 shader_glsl_append_dst(arg
->buffer
, arg
);
2670 shader_glsl_get_write_mask(arg
->dst
, dst_mask
);
2671 shader_addline(arg
->buffer
, "vec4(tmp0.xy, dot(T%u.xyz, %s), 1.0)%s);\n", reg
, src0_param
.param_str
, dst_mask
);
2673 current_state
->current_row
= 0;
2676 /** Process the WINED3DSIO_TEXM3X3SPEC instruction in GLSL
2677 * Perform the final texture lookup based on the previous 2 3x3 matrix multiplies */
2678 static void pshader_glsl_texm3x3spec(const SHADER_OPCODE_ARG
*arg
)
2680 IWineD3DPixelShaderImpl
* shader
= (IWineD3DPixelShaderImpl
*) arg
->shader
;
2681 DWORD reg
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2682 glsl_src_param_t src0_param
;
2683 glsl_src_param_t src1_param
;
2684 SHADER_BUFFER
* buffer
= arg
->buffer
;
2685 SHADER_PARSE_STATE
* current_state
= &shader
->baseShader
.parse_state
;
2686 DWORD stype
= arg
->reg_maps
->samplers
[reg
] & WINED3DSP_TEXTURETYPE_MASK
;
2687 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2688 glsl_sample_function_t sample_function
;
2690 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], src_mask
, &src0_param
);
2691 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], src_mask
, &src1_param
);
2693 /* Perform the last matrix multiply operation */
2694 shader_addline(buffer
, "tmp0.z = dot(T%u.xyz, %s);\n", reg
, src0_param
.param_str
);
2695 /* Reflection calculation */
2696 shader_addline(buffer
, "tmp0.xyz = -reflect((%s), normalize(tmp0.xyz));\n", src1_param
.param_str
);
2698 /* Dependent read, not valid with conditional NP2 */
2699 shader_glsl_get_sample_function(stype
, 0, &sample_function
);
2701 /* Sample the texture */
2702 shader_glsl_gen_sample_code(arg
, reg
, "tmp0.xyz",
2703 &sample_function
, WINED3DVS_NOSWIZZLE
,
2706 current_state
->current_row
= 0;
2709 /** Process the WINED3DSIO_TEXM3X3VSPEC instruction in GLSL
2710 * Perform the final texture lookup based on the previous 2 3x3 matrix multiplies */
2711 static void pshader_glsl_texm3x3vspec(const SHADER_OPCODE_ARG
*arg
)
2713 IWineD3DPixelShaderImpl
* shader
= (IWineD3DPixelShaderImpl
*) arg
->shader
;
2714 DWORD reg
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2715 SHADER_BUFFER
* buffer
= arg
->buffer
;
2716 SHADER_PARSE_STATE
* current_state
= &shader
->baseShader
.parse_state
;
2717 glsl_src_param_t src0_param
;
2718 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2719 DWORD sampler_type
= arg
->reg_maps
->samplers
[reg
] & WINED3DSP_TEXTURETYPE_MASK
;
2720 glsl_sample_function_t sample_function
;
2722 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], src_mask
, &src0_param
);
2724 /* Perform the last matrix multiply operation */
2725 shader_addline(buffer
, "tmp0.z = dot(vec3(T%u), vec3(%s));\n", reg
, src0_param
.param_str
);
2727 /* Construct the eye-ray vector from w coordinates */
2728 shader_addline(buffer
, "tmp1.xyz = normalize(vec3(gl_TexCoord[%u].w, gl_TexCoord[%u].w, gl_TexCoord[%u].w));\n",
2729 current_state
->texcoord_w
[0], current_state
->texcoord_w
[1], reg
);
2730 shader_addline(buffer
, "tmp0.xyz = -reflect(tmp1.xyz, normalize(tmp0.xyz));\n");
2732 /* Dependent read, not valid with conditional NP2 */
2733 shader_glsl_get_sample_function(sampler_type
, 0, &sample_function
);
2735 /* Sample the texture using the calculated coordinates */
2736 shader_glsl_gen_sample_code(arg
, reg
, "tmp0.xyz",
2737 &sample_function
, WINED3DVS_NOSWIZZLE
,
2740 current_state
->current_row
= 0;
2743 /** Process the WINED3DSIO_TEXBEM instruction in GLSL.
2744 * Apply a fake bump map transform.
2745 * texbem is pshader <= 1.3 only, this saves a few version checks
2747 static void pshader_glsl_texbem(const SHADER_OPCODE_ARG
*arg
)
2749 IWineD3DPixelShaderImpl
* This
= (IWineD3DPixelShaderImpl
*) arg
->shader
;
2750 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
2752 glsl_sample_function_t sample_function
;
2753 glsl_src_param_t coord_param
;
2760 sampler_idx
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2761 flags
= deviceImpl
->stateBlock
->textureState
[sampler_idx
][WINED3DTSS_TEXTURETRANSFORMFLAGS
];
2763 sampler_type
= arg
->reg_maps
->samplers
[sampler_idx
] & WINED3DSP_TEXTURETYPE_MASK
;
2764 /* Dependent read, not valid with conditional NP2 */
2765 shader_glsl_get_sample_function(sampler_type
, 0, &sample_function
);
2766 mask
= sample_function
.coord_mask
;
2768 shader_glsl_get_write_mask(mask
, coord_mask
);
2770 /* with projective textures, texbem only divides the static texture coord, not the displacement,
2771 * so we can't let the GL handle this.
2773 if (flags
& WINED3DTTFF_PROJECTED
) {
2775 char coord_div_mask
[3];
2776 switch (flags
& ~WINED3DTTFF_PROJECTED
) {
2777 case WINED3DTTFF_COUNT1
: FIXME("WINED3DTTFF_PROJECTED with WINED3DTTFF_COUNT1?\n"); break;
2778 case WINED3DTTFF_COUNT2
: div_mask
= WINED3DSP_WRITEMASK_1
; break;
2779 case WINED3DTTFF_COUNT3
: div_mask
= WINED3DSP_WRITEMASK_2
; break;
2780 case WINED3DTTFF_COUNT4
:
2781 case WINED3DTTFF_DISABLE
: div_mask
= WINED3DSP_WRITEMASK_3
; break;
2783 shader_glsl_get_write_mask(div_mask
, coord_div_mask
);
2784 shader_addline(arg
->buffer
, "T%u%s /= T%u%s;\n", sampler_idx
, coord_mask
, sampler_idx
, coord_div_mask
);
2787 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_0
|WINED3DSP_WRITEMASK_1
, &coord_param
);
2789 sprintf(coord
, "T%u%s + vec4(bumpenvmat%d * %s, 0.0, 0.0)%s",
2790 sampler_idx
, coord_mask
, sampler_idx
, coord_param
.param_str
, coord_mask
);
2791 shader_glsl_gen_sample_code(arg
, sampler_idx
, coord
,
2792 &sample_function
, WINED3DVS_NOSWIZZLE
,
2795 if(arg
->opcode
->opcode
== WINED3DSIO_TEXBEML
) {
2796 glsl_src_param_t luminance_param
;
2797 glsl_dst_param_t dst_param
;
2799 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_2
, &luminance_param
);
2800 shader_glsl_add_dst_param(arg
, arg
->dst
, arg
->dst_addr
, &dst_param
);
2802 shader_addline(arg
->buffer
, "%s%s *= (%s * luminancescale%d + luminanceoffset%d);\n",
2803 dst_param
.reg_name
, dst_param
.mask_str
,
2804 luminance_param
.param_str
, sampler_idx
, sampler_idx
);
2808 static void pshader_glsl_bem(const SHADER_OPCODE_ARG
*arg
)
2810 glsl_src_param_t src0_param
, src1_param
;
2811 DWORD sampler_idx
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2813 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_0
|WINED3DSP_WRITEMASK_1
, &src0_param
);
2814 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], WINED3DSP_WRITEMASK_0
|WINED3DSP_WRITEMASK_1
, &src1_param
);
2816 shader_glsl_append_dst(arg
->buffer
, arg
);
2817 shader_addline(arg
->buffer
, "%s + bumpenvmat%d * %s);\n",
2818 src0_param
.param_str
, sampler_idx
, src1_param
.param_str
);
2821 /** Process the WINED3DSIO_TEXREG2AR instruction in GLSL
2822 * Sample 2D texture at dst using the alpha & red (wx) components of src as texture coordinates */
2823 static void pshader_glsl_texreg2ar(const SHADER_OPCODE_ARG
*arg
)
2825 glsl_src_param_t src0_param
;
2826 DWORD sampler_idx
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2828 DWORD sampler_type
= arg
->reg_maps
->samplers
[sampler_idx
] & WINED3DSP_TEXTURETYPE_MASK
;
2829 glsl_sample_function_t sample_function
;
2831 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_ALL
, &src0_param
);
2833 shader_glsl_get_sample_function(sampler_type
, 0, &sample_function
);
2834 sprintf(src
, "%s.wx", src0_param
.reg_name
);
2835 shader_glsl_gen_sample_code(arg
, sampler_idx
, src
,
2836 &sample_function
, WINED3DVS_NOSWIZZLE
,
2840 /** Process the WINED3DSIO_TEXREG2GB instruction in GLSL
2841 * Sample 2D texture at dst using the green & blue (yz) components of src as texture coordinates */
2842 static void pshader_glsl_texreg2gb(const SHADER_OPCODE_ARG
*arg
)
2844 glsl_src_param_t src0_param
;
2845 DWORD sampler_idx
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2847 DWORD sampler_type
= arg
->reg_maps
->samplers
[sampler_idx
] & WINED3DSP_TEXTURETYPE_MASK
;
2848 glsl_sample_function_t sample_function
;
2850 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_ALL
, &src0_param
);
2852 shader_glsl_get_sample_function(sampler_type
, 0, &sample_function
);
2853 sprintf(src
, "%s.yz", src0_param
.reg_name
);
2854 shader_glsl_gen_sample_code(arg
, sampler_idx
, src
,
2855 &sample_function
, WINED3DVS_NOSWIZZLE
,
2859 /** Process the WINED3DSIO_TEXREG2RGB instruction in GLSL
2860 * Sample texture at dst using the rgb (xyz) components of src as texture coordinates */
2861 static void pshader_glsl_texreg2rgb(const SHADER_OPCODE_ARG
*arg
)
2863 glsl_src_param_t src0_param
;
2864 DWORD sampler_idx
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
2865 DWORD sampler_type
= arg
->reg_maps
->samplers
[sampler_idx
] & WINED3DSP_TEXTURETYPE_MASK
;
2866 glsl_sample_function_t sample_function
;
2868 /* Dependent read, not valid with conditional NP2 */
2869 shader_glsl_get_sample_function(sampler_type
, 0, &sample_function
);
2870 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], sample_function
.coord_mask
, &src0_param
);
2872 shader_glsl_gen_sample_code(arg
, sampler_idx
, src0_param
.param_str
,
2873 &sample_function
, WINED3DVS_NOSWIZZLE
,
2877 /** Process the WINED3DSIO_TEXKILL instruction in GLSL.
2878 * If any of the first 3 components are < 0, discard this pixel */
2879 static void pshader_glsl_texkill(const SHADER_OPCODE_ARG
*arg
)
2881 glsl_dst_param_t dst_param
;
2883 /* The argument is a destination parameter, and no writemasks are allowed */
2884 shader_glsl_add_dst_param(arg
, arg
->dst
, 0, &dst_param
);
2885 if ((arg
->reg_maps
->shader_version
>= WINED3DPS_VERSION(2,0)))
2887 /* 2.0 shaders compare all 4 components in texkill */
2888 shader_addline(arg
->buffer
, "if (any(lessThan(%s.xyzw, vec4(0.0)))) discard;\n", dst_param
.reg_name
);
2890 /* 1.X shaders only compare the first 3 components, probably due to the nature of the texkill
2891 * instruction as a tex* instruction, and phase, which kills all a / w components. Even if all
2892 * 4 components are defined, only the first 3 are used
2894 shader_addline(arg
->buffer
, "if (any(lessThan(%s.xyz, vec3(0.0)))) discard;\n", dst_param
.reg_name
);
2898 /** Process the WINED3DSIO_DP2ADD instruction in GLSL.
2899 * dst = dot2(src0, src1) + src2 */
2900 static void pshader_glsl_dp2add(const SHADER_OPCODE_ARG
*arg
)
2902 glsl_src_param_t src0_param
;
2903 glsl_src_param_t src1_param
;
2904 glsl_src_param_t src2_param
;
2906 unsigned int mask_size
;
2908 write_mask
= shader_glsl_append_dst(arg
->buffer
, arg
);
2909 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
2911 shader_glsl_add_src_param(arg
, arg
->src
[0], arg
->src_addr
[0], WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
, &src0_param
);
2912 shader_glsl_add_src_param(arg
, arg
->src
[1], arg
->src_addr
[1], WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
, &src1_param
);
2913 shader_glsl_add_src_param(arg
, arg
->src
[2], arg
->src_addr
[2], WINED3DSP_WRITEMASK_0
, &src2_param
);
2915 if (mask_size
> 1) {
2916 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
);
2918 shader_addline(arg
->buffer
, "dot(%s, %s) + %s);\n", src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
2922 static void pshader_glsl_input_pack(SHADER_BUFFER
* buffer
, const struct semantic
* semantics_in
,
2923 IWineD3DPixelShader
*iface
, enum vertexprocessing_mode vertexprocessing
)
2926 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*) iface
;
2928 for (i
= 0; i
< MAX_REG_INPUT
; i
++) {
2930 DWORD usage_token
= semantics_in
[i
].usage
;
2931 DWORD register_token
= semantics_in
[i
].reg
;
2932 DWORD usage
, usage_idx
;
2936 if (!usage_token
) continue;
2937 usage
= (usage_token
& WINED3DSP_DCL_USAGE_MASK
) >> WINED3DSP_DCL_USAGE_SHIFT
;
2938 usage_idx
= (usage_token
& WINED3DSP_DCL_USAGEINDEX_MASK
) >> WINED3DSP_DCL_USAGEINDEX_SHIFT
;
2939 shader_glsl_get_write_mask(register_token
, reg_mask
);
2943 case WINED3DDECLUSAGE_TEXCOORD
:
2944 if(usage_idx
< 8 && vertexprocessing
== pretransformed
) {
2945 shader_addline(buffer
, "IN[%u]%s = gl_TexCoord[%u]%s;\n",
2946 This
->input_reg_map
[i
], reg_mask
, usage_idx
, reg_mask
);
2948 shader_addline(buffer
, "IN[%u]%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
2949 This
->input_reg_map
[i
], reg_mask
, reg_mask
);
2953 case WINED3DDECLUSAGE_COLOR
:
2955 shader_addline(buffer
, "IN[%u]%s = vec4(gl_Color)%s;\n",
2956 This
->input_reg_map
[i
], reg_mask
, reg_mask
);
2957 else if (usage_idx
== 1)
2958 shader_addline(buffer
, "IN[%u]%s = vec4(gl_SecondaryColor)%s;\n",
2959 This
->input_reg_map
[i
], reg_mask
, reg_mask
);
2961 shader_addline(buffer
, "IN[%u]%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
2962 This
->input_reg_map
[i
], reg_mask
, reg_mask
);
2966 shader_addline(buffer
, "IN[%u]%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
2967 This
->input_reg_map
[i
], reg_mask
, reg_mask
);
2972 /*********************************************
2973 * Vertex Shader Specific Code begins here
2974 ********************************************/
2976 static void add_glsl_program_entry(struct shader_glsl_priv
*priv
, struct glsl_shader_prog_link
*entry
) {
2977 glsl_program_key_t
*key
;
2979 key
= HeapAlloc(GetProcessHeap(), 0, sizeof(glsl_program_key_t
));
2980 key
->vshader
= entry
->vshader
;
2981 key
->pshader
= entry
->pshader
;
2982 key
->vs_args
= entry
->vs_args
;
2983 key
->ps_args
= entry
->ps_args
;
2985 hash_table_put(priv
->glsl_program_lookup
, key
, entry
);
2988 static struct glsl_shader_prog_link
*get_glsl_program_entry(struct shader_glsl_priv
*priv
,
2989 IWineD3DVertexShader
*vshader
, IWineD3DPixelShader
*pshader
, struct vs_compile_args
*vs_args
,
2990 struct ps_compile_args
*ps_args
) {
2991 glsl_program_key_t key
;
2993 key
.vshader
= vshader
;
2994 key
.pshader
= pshader
;
2995 key
.vs_args
= *vs_args
;
2996 key
.ps_args
= *ps_args
;
2998 return hash_table_get(priv
->glsl_program_lookup
, &key
);
3001 static void delete_glsl_program_entry(struct shader_glsl_priv
*priv
, const WineD3D_GL_Info
*gl_info
,
3002 struct glsl_shader_prog_link
*entry
)
3004 glsl_program_key_t
*key
;
3006 key
= HeapAlloc(GetProcessHeap(), 0, sizeof(glsl_program_key_t
));
3007 key
->vshader
= entry
->vshader
;
3008 key
->pshader
= entry
->pshader
;
3009 key
->vs_args
= entry
->vs_args
;
3010 key
->ps_args
= entry
->ps_args
;
3011 hash_table_remove(priv
->glsl_program_lookup
, key
);
3013 GL_EXTCALL(glDeleteObjectARB(entry
->programId
));
3014 if (entry
->vshader
) list_remove(&entry
->vshader_entry
);
3015 if (entry
->pshader
) list_remove(&entry
->pshader_entry
);
3016 HeapFree(GetProcessHeap(), 0, entry
->vuniformF_locations
);
3017 HeapFree(GetProcessHeap(), 0, entry
->puniformF_locations
);
3018 HeapFree(GetProcessHeap(), 0, entry
);
3021 static void handle_ps3_input(SHADER_BUFFER
*buffer
, const struct semantic
*semantics_in
,
3022 const struct semantic
*semantics_out
, const WineD3D_GL_Info
*gl_info
, const DWORD
*map
)
3025 DWORD usage_token
, usage_token_out
;
3026 DWORD register_token
, register_token_out
;
3027 DWORD usage
, usage_idx
, usage_out
, usage_idx_out
;
3030 DWORD in_count
= GL_LIMITS(glsl_varyings
) / 4;
3031 char reg_mask
[6], reg_mask_out
[6];
3032 char destination
[50];
3034 set
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*set
) * (in_count
+ 2));
3036 if (!semantics_out
) {
3037 /* Save gl_FrontColor & gl_FrontSecondaryColor before overwriting them. */
3038 shader_addline(buffer
, "vec4 front_color = gl_FrontColor;\n");
3039 shader_addline(buffer
, "vec4 front_secondary_color = gl_FrontSecondaryColor;\n");
3042 for(i
= 0; i
< MAX_REG_INPUT
; i
++) {
3043 usage_token
= semantics_in
[i
].usage
;
3044 if (!usage_token
) continue;
3047 if (in_idx
>= (in_count
+ 2)) {
3048 FIXME("More input varyings declared than supported, expect issues\n");
3050 } else if(map
[i
] == -1) {
3051 /* Declared, but not read register */
3055 if (in_idx
== in_count
) {
3056 sprintf(destination
, "gl_FrontColor");
3057 } else if (in_idx
== in_count
+ 1) {
3058 sprintf(destination
, "gl_FrontSecondaryColor");
3060 sprintf(destination
, "IN[%u]", in_idx
);
3063 register_token
= semantics_in
[i
].reg
;
3065 usage
= (usage_token
& WINED3DSP_DCL_USAGE_MASK
) >> WINED3DSP_DCL_USAGE_SHIFT
;
3066 usage_idx
= (usage_token
& WINED3DSP_DCL_USAGEINDEX_MASK
) >> WINED3DSP_DCL_USAGEINDEX_SHIFT
;
3067 set
[map
[i
]] = shader_glsl_get_write_mask(register_token
, reg_mask
);
3069 if(!semantics_out
) {
3071 case WINED3DDECLUSAGE_COLOR
:
3073 shader_addline(buffer
, "%s%s = front_color%s;\n",
3074 destination
, reg_mask
, reg_mask
);
3075 else if (usage_idx
== 1)
3076 shader_addline(buffer
, "%s%s = front_secondary_color%s;\n",
3077 destination
, reg_mask
, reg_mask
);
3079 shader_addline(buffer
, "%s%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
3080 destination
, reg_mask
, reg_mask
);
3083 case WINED3DDECLUSAGE_TEXCOORD
:
3084 if (usage_idx
< 8) {
3085 shader_addline(buffer
, "%s%s = gl_TexCoord[%u]%s;\n",
3086 destination
, reg_mask
, usage_idx
, reg_mask
);
3088 shader_addline(buffer
, "%s%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
3089 destination
, reg_mask
, reg_mask
);
3093 case WINED3DDECLUSAGE_FOG
:
3094 shader_addline(buffer
, "%s%s = vec4(gl_FogFragCoord, 0.0, 0.0, 0.0)%s;\n",
3095 destination
, reg_mask
, reg_mask
);
3099 shader_addline(buffer
, "%s%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
3100 destination
, reg_mask
, reg_mask
);
3104 for(j
= 0; j
< MAX_REG_OUTPUT
; j
++) {
3105 usage_token_out
= semantics_out
[j
].usage
;
3106 if (!usage_token_out
) continue;
3107 register_token_out
= semantics_out
[j
].reg
;
3109 usage_out
= (usage_token_out
& WINED3DSP_DCL_USAGE_MASK
) >> WINED3DSP_DCL_USAGE_SHIFT
;
3110 usage_idx_out
= (usage_token_out
& WINED3DSP_DCL_USAGEINDEX_MASK
) >> WINED3DSP_DCL_USAGEINDEX_SHIFT
;
3111 shader_glsl_get_write_mask(register_token_out
, reg_mask_out
);
3113 if(usage
== usage_out
&&
3114 usage_idx
== usage_idx_out
) {
3115 shader_addline(buffer
, "%s%s = OUT[%u]%s;\n",
3116 destination
, reg_mask
, j
, reg_mask
);
3121 shader_addline(buffer
, "%s%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
3122 destination
, reg_mask
, reg_mask
);
3127 /* This is solely to make the compiler / linker happy and avoid warning about undefined
3128 * varyings. It shouldn't result in any real code executed on the GPU, since all read
3129 * input varyings are assigned above, if the optimizer works properly.
3131 for(i
= 0; i
< in_count
+ 2; i
++) {
3132 if(set
[i
] != WINED3DSP_WRITEMASK_ALL
) {
3133 unsigned int size
= 0;
3134 memset(reg_mask
, 0, sizeof(reg_mask
));
3135 if(!(set
[i
] & WINED3DSP_WRITEMASK_0
)) {
3136 reg_mask
[size
] = 'x';
3139 if(!(set
[i
] & WINED3DSP_WRITEMASK_1
)) {
3140 reg_mask
[size
] = 'y';
3143 if(!(set
[i
] & WINED3DSP_WRITEMASK_2
)) {
3144 reg_mask
[size
] = 'z';
3147 if(!(set
[i
] & WINED3DSP_WRITEMASK_3
)) {
3148 reg_mask
[size
] = 'w';
3152 if (i
== in_count
) {
3153 sprintf(destination
, "gl_FrontColor");
3154 } else if (i
== in_count
+ 1) {
3155 sprintf(destination
, "gl_FrontSecondaryColor");
3157 sprintf(destination
, "IN[%u]", i
);
3161 shader_addline(buffer
, "%s.%s = 0.0;\n", destination
, reg_mask
);
3163 shader_addline(buffer
, "%s.%s = vec%u(0.0);\n", destination
, reg_mask
, size
);
3168 HeapFree(GetProcessHeap(), 0, set
);
3171 static GLhandleARB
generate_param_reorder_function(IWineD3DVertexShader
*vertexshader
,
3172 IWineD3DPixelShader
*pixelshader
, const WineD3D_GL_Info
*gl_info
)
3174 GLhandleARB ret
= 0;
3175 IWineD3DVertexShaderImpl
*vs
= (IWineD3DVertexShaderImpl
*) vertexshader
;
3176 IWineD3DPixelShaderImpl
*ps
= (IWineD3DPixelShaderImpl
*) pixelshader
;
3177 IWineD3DDeviceImpl
*device
;
3178 DWORD vs_major
= WINED3DSHADER_VERSION_MAJOR(vs
->baseShader
.reg_maps
.shader_version
);
3179 DWORD ps_major
= ps
? WINED3DSHADER_VERSION_MAJOR(ps
->baseShader
.reg_maps
.shader_version
) : 0;
3181 SHADER_BUFFER buffer
;
3183 DWORD register_token
;
3184 DWORD usage
, usage_idx
, writemask
;
3186 const struct semantic
*semantics_out
, *semantics_in
;
3188 shader_buffer_init(&buffer
);
3190 shader_addline(&buffer
, "#version 120\n");
3192 if(vs_major
< 3 && ps_major
< 3) {
3193 /* That one is easy: The vertex shader writes to the builtin varyings, the pixel shader reads from them.
3194 * Take care about the texcoord .w fixup though if we're using the fixed function fragment pipeline
3196 device
= (IWineD3DDeviceImpl
*) vs
->baseShader
.device
;
3197 if((GLINFO_LOCATION
).set_texcoord_w
&& ps_major
== 0 && vs_major
> 0 &&
3198 !device
->frag_pipe
->ffp_proj_control
) {
3199 shader_addline(&buffer
, "void order_ps_input() {\n");
3200 for(i
= 0; i
< min(8, MAX_REG_TEXCRD
); i
++) {
3201 if(vs
->baseShader
.reg_maps
.texcoord_mask
[i
] != 0 &&
3202 vs
->baseShader
.reg_maps
.texcoord_mask
[i
] != WINED3DSP_WRITEMASK_ALL
) {
3203 shader_addline(&buffer
, "gl_TexCoord[%u].w = 1.0;\n", i
);
3206 shader_addline(&buffer
, "}\n");
3208 shader_addline(&buffer
, "void order_ps_input() { /* do nothing */ }\n");
3210 } else if(ps_major
< 3 && vs_major
>= 3) {
3211 /* The vertex shader writes to its own varyings, the pixel shader needs them in the builtin ones */
3212 semantics_out
= vs
->semantics_out
;
3214 shader_addline(&buffer
, "void order_ps_input(in vec4 OUT[%u]) {\n", MAX_REG_OUTPUT
);
3215 for(i
= 0; i
< MAX_REG_OUTPUT
; i
++) {
3216 usage_token
= semantics_out
[i
].usage
;
3217 if (!usage_token
) continue;
3218 register_token
= semantics_out
[i
].reg
;
3220 usage
= (usage_token
& WINED3DSP_DCL_USAGE_MASK
) >> WINED3DSP_DCL_USAGE_SHIFT
;
3221 usage_idx
= (usage_token
& WINED3DSP_DCL_USAGEINDEX_MASK
) >> WINED3DSP_DCL_USAGEINDEX_SHIFT
;
3222 writemask
= shader_glsl_get_write_mask(register_token
, reg_mask
);
3225 case WINED3DDECLUSAGE_COLOR
:
3227 shader_addline(&buffer
, "gl_FrontColor%s = OUT[%u]%s;\n", reg_mask
, i
, reg_mask
);
3228 else if (usage_idx
== 1)
3229 shader_addline(&buffer
, "gl_FrontSecondaryColor%s = OUT[%u]%s;\n", reg_mask
, i
, reg_mask
);
3232 case WINED3DDECLUSAGE_POSITION
:
3233 shader_addline(&buffer
, "gl_Position%s = OUT[%u]%s;\n", reg_mask
, i
, reg_mask
);
3236 case WINED3DDECLUSAGE_TEXCOORD
:
3237 if (usage_idx
< 8) {
3238 if(!(GLINFO_LOCATION
).set_texcoord_w
|| ps_major
> 0) writemask
|= WINED3DSP_WRITEMASK_3
;
3240 shader_addline(&buffer
, "gl_TexCoord[%u]%s = OUT[%u]%s;\n",
3241 usage_idx
, reg_mask
, i
, reg_mask
);
3242 if(!(writemask
& WINED3DSP_WRITEMASK_3
)) {
3243 shader_addline(&buffer
, "gl_TexCoord[%u].w = 1.0;\n", usage_idx
);
3248 case WINED3DDECLUSAGE_PSIZE
:
3249 shader_addline(&buffer
, "gl_PointSize = OUT[%u].x;\n", i
);
3252 case WINED3DDECLUSAGE_FOG
:
3253 shader_addline(&buffer
, "gl_FogFragCoord = OUT[%u].%c;\n", i
, reg_mask
[1]);
3260 shader_addline(&buffer
, "}\n");
3262 } else if(ps_major
>= 3 && vs_major
>= 3) {
3263 semantics_out
= vs
->semantics_out
;
3264 semantics_in
= ps
->semantics_in
;
3266 /* This one is tricky: a 3.0 pixel shader reads from a 3.0 vertex shader */
3267 shader_addline(&buffer
, "varying vec4 IN[%u];\n", GL_LIMITS(glsl_varyings
) / 4);
3268 shader_addline(&buffer
, "void order_ps_input(in vec4 OUT[%u]) {\n", MAX_REG_OUTPUT
);
3270 /* First, sort out position and point size. Those are not passed to the pixel shader */
3271 for(i
= 0; i
< MAX_REG_OUTPUT
; i
++) {
3272 usage_token
= semantics_out
[i
].usage
;
3273 if (!usage_token
) continue;
3274 register_token
= semantics_out
[i
].reg
;
3276 usage
= (usage_token
& WINED3DSP_DCL_USAGE_MASK
) >> WINED3DSP_DCL_USAGE_SHIFT
;
3277 usage_idx
= (usage_token
& WINED3DSP_DCL_USAGEINDEX_MASK
) >> WINED3DSP_DCL_USAGEINDEX_SHIFT
;
3278 shader_glsl_get_write_mask(register_token
, reg_mask
);
3281 case WINED3DDECLUSAGE_POSITION
:
3282 shader_addline(&buffer
, "gl_Position%s = OUT[%u]%s;\n", reg_mask
, i
, reg_mask
);
3285 case WINED3DDECLUSAGE_PSIZE
:
3286 shader_addline(&buffer
, "gl_PointSize = OUT[%u].x;\n", i
);
3294 /* Then, fix the pixel shader input */
3295 handle_ps3_input(&buffer
, semantics_in
, semantics_out
, gl_info
, ps
->input_reg_map
);
3297 shader_addline(&buffer
, "}\n");
3298 } else if(ps_major
>= 3 && vs_major
< 3) {
3299 semantics_in
= ps
->semantics_in
;
3301 shader_addline(&buffer
, "varying vec4 IN[%u];\n", GL_LIMITS(glsl_varyings
) / 4);
3302 shader_addline(&buffer
, "void order_ps_input() {\n");
3303 /* The vertex shader wrote to the builtin varyings. There is no need to figure out position and
3304 * point size, but we depend on the optimizers kindness to find out that the pixel shader doesn't
3305 * read gl_TexCoord and gl_ColorX, otherwise we'll run out of varyings
3307 handle_ps3_input(&buffer
, semantics_in
, NULL
, gl_info
, ps
->input_reg_map
);
3308 shader_addline(&buffer
, "}\n");
3310 ERR("Unexpected vertex and pixel shader version condition: vs: %d, ps: %d\n", vs_major
, ps_major
);
3313 ret
= GL_EXTCALL(glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB
));
3314 checkGLcall("glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB)");
3315 GL_EXTCALL(glShaderSourceARB(ret
, 1, (const char**)&buffer
.buffer
, NULL
));
3316 checkGLcall("glShaderSourceARB(ret, 1, &buffer.buffer, NULL)");
3317 GL_EXTCALL(glCompileShaderARB(ret
));
3318 checkGLcall("glCompileShaderARB(ret)");
3320 shader_buffer_free(&buffer
);
3324 static void hardcode_local_constants(IWineD3DBaseShaderImpl
*shader
, const WineD3D_GL_Info
*gl_info
,
3325 GLhandleARB programId
, char prefix
)
3327 const local_constant
*lconst
;
3332 LIST_FOR_EACH_ENTRY(lconst
, &shader
->baseShader
.constantsF
, local_constant
, entry
) {
3333 value
= (const float *)lconst
->value
;
3334 snprintf(glsl_name
, sizeof(glsl_name
), "%cLC%u", prefix
, lconst
->idx
);
3335 tmp_loc
= GL_EXTCALL(glGetUniformLocationARB(programId
, glsl_name
));
3336 GL_EXTCALL(glUniform4fvARB(tmp_loc
, 1, value
));
3338 checkGLcall("Hardcoding local constants\n");
3341 /** Sets the GLSL program ID for the given pixel and vertex shader combination.
3342 * It sets the programId on the current StateBlock (because it should be called
3343 * inside of the DrawPrimitive() part of the render loop).
3345 * If a program for the given combination does not exist, create one, and store
3346 * the program in the hash table. If it creates a program, it will link the
3347 * given objects, too.
3349 static void set_glsl_shader_program(IWineD3DDevice
*iface
, BOOL use_ps
, BOOL use_vs
) {
3350 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
3351 struct shader_glsl_priv
*priv
= This
->shader_priv
;
3352 const WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
3353 IWineD3DPixelShader
*pshader
= This
->stateBlock
->pixelShader
;
3354 IWineD3DVertexShader
*vshader
= This
->stateBlock
->vertexShader
;
3355 struct glsl_shader_prog_link
*entry
= NULL
;
3356 GLhandleARB programId
= 0;
3357 GLhandleARB reorder_shader_id
= 0;
3360 GLhandleARB vshader_id
, pshader_id
;
3361 struct ps_compile_args ps_compile_args
;
3362 struct vs_compile_args vs_compile_args
;
3365 find_vs_compile_args((IWineD3DVertexShaderImpl
*)This
->stateBlock
->vertexShader
, This
->stateBlock
, &vs_compile_args
);
3367 /* FIXME: Do we really have to spend CPU cycles to generate a few zeroed bytes? */
3368 memset(&vs_compile_args
, 0, sizeof(vs_compile_args
));
3371 find_ps_compile_args((IWineD3DPixelShaderImpl
*)This
->stateBlock
->pixelShader
, This
->stateBlock
, &ps_compile_args
);
3373 /* FIXME: Do we really have to spend CPU cycles to generate a few zeroed bytes? */
3374 memset(&ps_compile_args
, 0, sizeof(ps_compile_args
));
3376 entry
= get_glsl_program_entry(priv
, vshader
, pshader
, &vs_compile_args
, &ps_compile_args
);
3378 priv
->glsl_program
= entry
;
3382 /* If we get to this point, then no matching program exists, so we create one */
3383 programId
= GL_EXTCALL(glCreateProgramObjectARB());
3384 TRACE("Created new GLSL shader program %u\n", programId
);
3386 /* Create the entry */
3387 entry
= HeapAlloc(GetProcessHeap(), 0, sizeof(struct glsl_shader_prog_link
));
3388 entry
->programId
= programId
;
3389 entry
->vshader
= vshader
;
3390 entry
->pshader
= pshader
;
3391 entry
->vs_args
= vs_compile_args
;
3392 entry
->ps_args
= ps_compile_args
;
3393 entry
->constant_version
= 0;
3394 /* Add the hash table entry */
3395 add_glsl_program_entry(priv
, entry
);
3397 /* Set the current program */
3398 priv
->glsl_program
= entry
;
3401 vshader_id
= find_gl_vshader((IWineD3DVertexShaderImpl
*) vshader
, &vs_compile_args
);
3406 /* Attach GLSL vshader */
3408 int max_attribs
= 16; /* TODO: Will this always be the case? It is at the moment... */
3411 reorder_shader_id
= generate_param_reorder_function(vshader
, pshader
, gl_info
);
3412 TRACE("Attaching GLSL shader object %u to program %u\n", reorder_shader_id
, programId
);
3413 GL_EXTCALL(glAttachObjectARB(programId
, reorder_shader_id
));
3414 checkGLcall("glAttachObjectARB");
3415 /* Flag the reorder function for deletion, then it will be freed automatically when the program
3418 GL_EXTCALL(glDeleteObjectARB(reorder_shader_id
));
3420 TRACE("Attaching GLSL shader object %u to program %u\n", vshader_id
, programId
);
3421 GL_EXTCALL(glAttachObjectARB(programId
, vshader_id
));
3422 checkGLcall("glAttachObjectARB");
3424 /* Bind vertex attributes to a corresponding index number to match
3425 * the same index numbers as ARB_vertex_programs (makes loading
3426 * vertex attributes simpler). With this method, we can use the
3427 * exact same code to load the attributes later for both ARB and
3430 * We have to do this here because we need to know the Program ID
3431 * in order to make the bindings work, and it has to be done prior
3432 * to linking the GLSL program. */
3433 for (i
= 0; i
< max_attribs
; ++i
) {
3434 if (((IWineD3DBaseShaderImpl
*)vshader
)->baseShader
.reg_maps
.attributes
[i
]) {
3435 snprintf(tmp_name
, sizeof(tmp_name
), "attrib%i", i
);
3436 GL_EXTCALL(glBindAttribLocationARB(programId
, i
, tmp_name
));
3439 checkGLcall("glBindAttribLocationARB");
3441 list_add_head(&((IWineD3DBaseShaderImpl
*)vshader
)->baseShader
.linked_programs
, &entry
->vshader_entry
);
3445 pshader_id
= find_gl_pshader((IWineD3DPixelShaderImpl
*) pshader
, &ps_compile_args
);
3450 /* Attach GLSL pshader */
3452 TRACE("Attaching GLSL shader object %u to program %u\n", pshader_id
, programId
);
3453 GL_EXTCALL(glAttachObjectARB(programId
, pshader_id
));
3454 checkGLcall("glAttachObjectARB");
3456 list_add_head(&((IWineD3DBaseShaderImpl
*)pshader
)->baseShader
.linked_programs
, &entry
->pshader_entry
);
3459 /* Link the program */
3460 TRACE("Linking GLSL shader program %u\n", programId
);
3461 GL_EXTCALL(glLinkProgramARB(programId
));
3462 print_glsl_info_log(&GLINFO_LOCATION
, programId
);
3464 entry
->vuniformF_locations
= HeapAlloc(GetProcessHeap(), 0, sizeof(GLhandleARB
) * GL_LIMITS(vshader_constantsF
));
3465 for (i
= 0; i
< GL_LIMITS(vshader_constantsF
); ++i
) {
3466 snprintf(glsl_name
, sizeof(glsl_name
), "VC[%i]", i
);
3467 entry
->vuniformF_locations
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, glsl_name
));
3469 for (i
= 0; i
< MAX_CONST_I
; ++i
) {
3470 snprintf(glsl_name
, sizeof(glsl_name
), "VI[%i]", i
);
3471 entry
->vuniformI_locations
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, glsl_name
));
3473 entry
->puniformF_locations
= HeapAlloc(GetProcessHeap(), 0, sizeof(GLhandleARB
) * GL_LIMITS(pshader_constantsF
));
3474 for (i
= 0; i
< GL_LIMITS(pshader_constantsF
); ++i
) {
3475 snprintf(glsl_name
, sizeof(glsl_name
), "PC[%i]", i
);
3476 entry
->puniformF_locations
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, glsl_name
));
3478 for (i
= 0; i
< MAX_CONST_I
; ++i
) {
3479 snprintf(glsl_name
, sizeof(glsl_name
), "PI[%i]", i
);
3480 entry
->puniformI_locations
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, glsl_name
));
3484 for(i
= 0; i
< ((IWineD3DPixelShaderImpl
*)pshader
)->numbumpenvmatconsts
; i
++) {
3486 sprintf(name
, "bumpenvmat%d", ((IWineD3DPixelShaderImpl
*)pshader
)->bumpenvmatconst
[i
].texunit
);
3487 entry
->bumpenvmat_location
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, name
));
3488 sprintf(name
, "luminancescale%d", ((IWineD3DPixelShaderImpl
*)pshader
)->luminanceconst
[i
].texunit
);
3489 entry
->luminancescale_location
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, name
));
3490 sprintf(name
, "luminanceoffset%d", ((IWineD3DPixelShaderImpl
*)pshader
)->luminanceconst
[i
].texunit
);
3491 entry
->luminanceoffset_location
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, name
));
3496 entry
->posFixup_location
= GL_EXTCALL(glGetUniformLocationARB(programId
, "posFixup"));
3497 entry
->ycorrection_location
= GL_EXTCALL(glGetUniformLocationARB(programId
, "ycorrection"));
3498 checkGLcall("Find glsl program uniform locations");
3501 && WINED3DSHADER_VERSION_MAJOR(((IWineD3DPixelShaderImpl
*)pshader
)->baseShader
.reg_maps
.shader_version
) >= 3
3502 && ((IWineD3DPixelShaderImpl
*)pshader
)->declared_in_count
> GL_LIMITS(glsl_varyings
) / 4)
3504 TRACE("Shader %d needs vertex color clamping disabled\n", programId
);
3505 entry
->vertex_color_clamp
= GL_FALSE
;
3507 entry
->vertex_color_clamp
= GL_FIXED_ONLY_ARB
;
3510 /* Set the shader to allow uniform loading on it */
3511 GL_EXTCALL(glUseProgramObjectARB(programId
));
3512 checkGLcall("glUseProgramObjectARB(programId)");
3514 /* Load the vertex and pixel samplers now. The function that finds the mappings makes sure
3515 * that it stays the same for each vertexshader-pixelshader pair(=linked glsl program). If
3516 * a pshader with fixed function pipeline is used there are no vertex samplers, and if a
3517 * vertex shader with fixed function pixel processing is used we make sure that the card
3518 * supports enough samplers to allow the max number of vertex samplers with all possible
3519 * fixed function fragment processing setups. So once the program is linked these samplers
3523 /* Load vertex shader samplers */
3524 shader_glsl_load_vsamplers(gl_info
, (IWineD3DStateBlock
*)This
->stateBlock
, programId
);
3527 /* Load pixel shader samplers */
3528 shader_glsl_load_psamplers(gl_info
, (IWineD3DStateBlock
*)This
->stateBlock
, programId
);
3531 /* If the local constants do not have to be loaded with the environment constants,
3532 * load them now to have them hardcoded in the GLSL program. This saves some CPU cycles
3535 if(pshader
&& !((IWineD3DPixelShaderImpl
*)pshader
)->baseShader
.load_local_constsF
) {
3536 hardcode_local_constants((IWineD3DBaseShaderImpl
*) pshader
, gl_info
, programId
, 'P');
3538 if(vshader
&& !((IWineD3DVertexShaderImpl
*)vshader
)->baseShader
.load_local_constsF
) {
3539 hardcode_local_constants((IWineD3DBaseShaderImpl
*) vshader
, gl_info
, programId
, 'V');
3543 static GLhandleARB
create_glsl_blt_shader(const WineD3D_GL_Info
*gl_info
, enum tex_types tex_type
)
3545 GLhandleARB program_id
;
3546 GLhandleARB vshader_id
, pshader_id
;
3547 static const char *blt_vshader
[] =
3552 " gl_Position = gl_Vertex;\n"
3553 " gl_FrontColor = vec4(1.0);\n"
3554 " gl_TexCoord[0] = gl_MultiTexCoord0;\n"
3558 static const char *blt_pshaders
[tex_type_count
] =
3564 "uniform sampler2D sampler;\n"
3567 " gl_FragDepth = texture2D(sampler, gl_TexCoord[0].xy).x;\n"
3573 "uniform samplerCube sampler;\n"
3576 " gl_FragDepth = textureCube(sampler, gl_TexCoord[0].xyz).x;\n"
3580 "#extension GL_ARB_texture_rectangle : enable\n"
3581 "uniform sampler2DRect sampler;\n"
3584 " gl_FragDepth = texture2DRect(sampler, gl_TexCoord[0].xy).x;\n"
3588 if (!blt_pshaders
[tex_type
])
3590 FIXME("tex_type %#x not supported\n", tex_type
);
3594 vshader_id
= GL_EXTCALL(glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB
));
3595 GL_EXTCALL(glShaderSourceARB(vshader_id
, 1, blt_vshader
, NULL
));
3596 GL_EXTCALL(glCompileShaderARB(vshader_id
));
3598 pshader_id
= GL_EXTCALL(glCreateShaderObjectARB(GL_FRAGMENT_SHADER_ARB
));
3599 GL_EXTCALL(glShaderSourceARB(pshader_id
, 1, &blt_pshaders
[tex_type
], NULL
));
3600 GL_EXTCALL(glCompileShaderARB(pshader_id
));
3602 program_id
= GL_EXTCALL(glCreateProgramObjectARB());
3603 GL_EXTCALL(glAttachObjectARB(program_id
, vshader_id
));
3604 GL_EXTCALL(glAttachObjectARB(program_id
, pshader_id
));
3605 GL_EXTCALL(glLinkProgramARB(program_id
));
3607 print_glsl_info_log(&GLINFO_LOCATION
, program_id
);
3609 /* Once linked we can mark the shaders for deletion. They will be deleted once the program
3612 GL_EXTCALL(glDeleteObjectARB(vshader_id
));
3613 GL_EXTCALL(glDeleteObjectARB(pshader_id
));
3617 static void shader_glsl_select(IWineD3DDevice
*iface
, BOOL usePS
, BOOL useVS
) {
3618 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
3619 struct shader_glsl_priv
*priv
= This
->shader_priv
;
3620 const WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
3621 GLhandleARB program_id
= 0;
3622 GLenum old_vertex_color_clamp
, current_vertex_color_clamp
;
3624 old_vertex_color_clamp
= priv
->glsl_program
? priv
->glsl_program
->vertex_color_clamp
: GL_FIXED_ONLY_ARB
;
3626 if (useVS
|| usePS
) set_glsl_shader_program(iface
, usePS
, useVS
);
3627 else priv
->glsl_program
= NULL
;
3629 current_vertex_color_clamp
= priv
->glsl_program
? priv
->glsl_program
->vertex_color_clamp
: GL_FIXED_ONLY_ARB
;
3631 if (old_vertex_color_clamp
!= current_vertex_color_clamp
) {
3632 if (GL_SUPPORT(ARB_COLOR_BUFFER_FLOAT
)) {
3633 GL_EXTCALL(glClampColorARB(GL_CLAMP_VERTEX_COLOR_ARB
, current_vertex_color_clamp
));
3634 checkGLcall("glClampColorARB");
3636 FIXME("vertex color clamp needs to be changed, but extension not supported.\n");
3640 program_id
= priv
->glsl_program
? priv
->glsl_program
->programId
: 0;
3641 if (program_id
) TRACE("Using GLSL program %u\n", program_id
);
3642 GL_EXTCALL(glUseProgramObjectARB(program_id
));
3643 checkGLcall("glUseProgramObjectARB");
3646 static void shader_glsl_select_depth_blt(IWineD3DDevice
*iface
, enum tex_types tex_type
) {
3647 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
3648 const WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
3649 struct shader_glsl_priv
*priv
= This
->shader_priv
;
3650 GLhandleARB
*blt_program
= &priv
->depth_blt_program
[tex_type
];
3652 if (!*blt_program
) {
3654 *blt_program
= create_glsl_blt_shader(gl_info
, tex_type
);
3655 loc
= GL_EXTCALL(glGetUniformLocationARB(*blt_program
, "sampler"));
3656 GL_EXTCALL(glUseProgramObjectARB(*blt_program
));
3657 GL_EXTCALL(glUniform1iARB(loc
, 0));
3659 GL_EXTCALL(glUseProgramObjectARB(*blt_program
));
3663 static void shader_glsl_deselect_depth_blt(IWineD3DDevice
*iface
) {
3664 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
3665 const WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
3666 struct shader_glsl_priv
*priv
= This
->shader_priv
;
3667 GLhandleARB program_id
;
3669 program_id
= priv
->glsl_program
? priv
->glsl_program
->programId
: 0;
3670 if (program_id
) TRACE("Using GLSL program %u\n", program_id
);
3672 GL_EXTCALL(glUseProgramObjectARB(program_id
));
3673 checkGLcall("glUseProgramObjectARB");
3676 static void shader_glsl_destroy(IWineD3DBaseShader
*iface
) {
3677 const struct list
*linked_programs
;
3678 IWineD3DBaseShaderImpl
*This
= (IWineD3DBaseShaderImpl
*) iface
;
3679 IWineD3DDeviceImpl
*device
= (IWineD3DDeviceImpl
*)This
->baseShader
.device
;
3680 struct shader_glsl_priv
*priv
= device
->shader_priv
;
3681 const WineD3D_GL_Info
*gl_info
= &device
->adapter
->gl_info
;
3682 IWineD3DPixelShaderImpl
*ps
= NULL
;
3683 IWineD3DVertexShaderImpl
*vs
= NULL
;
3685 /* Note: Do not use QueryInterface here to find out which shader type this is because this code
3686 * can be called from IWineD3DBaseShader::Release
3688 char pshader
= shader_is_pshader_version(This
->baseShader
.reg_maps
.shader_version
);
3691 ps
= (IWineD3DPixelShaderImpl
*) This
;
3692 if(ps
->num_gl_shaders
== 0) return;
3694 vs
= (IWineD3DVertexShaderImpl
*) This
;
3695 if(vs
->num_gl_shaders
== 0) return;
3698 linked_programs
= &This
->baseShader
.linked_programs
;
3700 TRACE("Deleting linked programs\n");
3701 if (linked_programs
->next
) {
3702 struct glsl_shader_prog_link
*entry
, *entry2
;
3705 LIST_FOR_EACH_ENTRY_SAFE(entry
, entry2
, linked_programs
, struct glsl_shader_prog_link
, pshader_entry
) {
3706 delete_glsl_program_entry(priv
, gl_info
, entry
);
3709 LIST_FOR_EACH_ENTRY_SAFE(entry
, entry2
, linked_programs
, struct glsl_shader_prog_link
, vshader_entry
) {
3710 delete_glsl_program_entry(priv
, gl_info
, entry
);
3719 for(i
= 0; i
< ps
->num_gl_shaders
; i
++) {
3720 TRACE("deleting pshader %u\n", ps
->gl_shaders
[i
].prgId
);
3721 GL_EXTCALL(glDeleteObjectARB(ps
->gl_shaders
[i
].prgId
));
3722 checkGLcall("glDeleteObjectARB");
3725 HeapFree(GetProcessHeap(), 0, ps
->gl_shaders
);
3726 ps
->gl_shaders
= NULL
;
3727 ps
->num_gl_shaders
= 0;
3728 ps
->shader_array_size
= 0;
3733 for(i
= 0; i
< vs
->num_gl_shaders
; i
++) {
3734 TRACE("deleting vshader %u\n", vs
->gl_shaders
[i
].prgId
);
3735 GL_EXTCALL(glDeleteObjectARB(vs
->gl_shaders
[i
].prgId
));
3736 checkGLcall("glDeleteObjectARB");
3739 HeapFree(GetProcessHeap(), 0, vs
->gl_shaders
);
3740 vs
->gl_shaders
= NULL
;
3741 vs
->num_gl_shaders
= 0;
3742 vs
->shader_array_size
= 0;
3746 static unsigned int glsl_program_key_hash(const void *key
)
3748 const glsl_program_key_t
*k
= key
;
3750 unsigned int hash
= ((DWORD_PTR
) k
->vshader
) | ((DWORD_PTR
) k
->pshader
) << 16;
3751 hash
+= ~(hash
<< 15);
3752 hash
^= (hash
>> 10);
3753 hash
+= (hash
<< 3);
3754 hash
^= (hash
>> 6);
3755 hash
+= ~(hash
<< 11);
3756 hash
^= (hash
>> 16);
3761 static BOOL
glsl_program_key_compare(const void *keya
, const void *keyb
)
3763 const glsl_program_key_t
*ka
= keya
;
3764 const glsl_program_key_t
*kb
= keyb
;
3766 return ka
->vshader
== kb
->vshader
&& ka
->pshader
== kb
->pshader
&&
3767 (memcmp(&ka
->ps_args
, &kb
->ps_args
, sizeof(kb
->ps_args
)) == 0) &&
3768 (memcmp(&ka
->vs_args
, &kb
->vs_args
, sizeof(kb
->vs_args
)) == 0);
3771 static BOOL
constant_heap_init(struct constant_heap
*heap
, unsigned int constant_count
)
3773 SIZE_T size
= (constant_count
+ 1) * sizeof(*heap
->entries
) + constant_count
* sizeof(*heap
->positions
);
3774 void *mem
= HeapAlloc(GetProcessHeap(), 0, size
);
3778 ERR("Failed to allocate memory\n");
3782 heap
->entries
= mem
;
3783 heap
->entries
[1].version
= 0;
3784 heap
->positions
= (unsigned int *)(heap
->entries
+ constant_count
+ 1);
3790 static void constant_heap_free(struct constant_heap
*heap
)
3792 HeapFree(GetProcessHeap(), 0, heap
->entries
);
3795 static HRESULT
shader_glsl_alloc(IWineD3DDevice
*iface
) {
3796 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
3797 const WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
3798 struct shader_glsl_priv
*priv
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(struct shader_glsl_priv
));
3799 SIZE_T stack_size
= wined3d_log2i(max(GL_LIMITS(vshader_constantsF
), GL_LIMITS(pshader_constantsF
))) + 1;
3801 priv
->stack
= HeapAlloc(GetProcessHeap(), 0, stack_size
* sizeof(*priv
->stack
));
3804 ERR("Failed to allocate memory.\n");
3805 HeapFree(GetProcessHeap(), 0, priv
);
3806 return E_OUTOFMEMORY
;
3809 if (!constant_heap_init(&priv
->vconst_heap
, GL_LIMITS(vshader_constantsF
)))
3811 ERR("Failed to initialize vertex shader constant heap\n");
3812 HeapFree(GetProcessHeap(), 0, priv
->stack
);
3813 HeapFree(GetProcessHeap(), 0, priv
);
3814 return E_OUTOFMEMORY
;
3817 if (!constant_heap_init(&priv
->pconst_heap
, GL_LIMITS(pshader_constantsF
)))
3819 ERR("Failed to initialize pixel shader constant heap\n");
3820 constant_heap_free(&priv
->vconst_heap
);
3821 HeapFree(GetProcessHeap(), 0, priv
->stack
);
3822 HeapFree(GetProcessHeap(), 0, priv
);
3823 return E_OUTOFMEMORY
;
3826 priv
->glsl_program_lookup
= hash_table_create(glsl_program_key_hash
, glsl_program_key_compare
);
3827 priv
->next_constant_version
= 1;
3829 This
->shader_priv
= priv
;
3833 static void shader_glsl_free(IWineD3DDevice
*iface
) {
3834 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
3835 const WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
3836 struct shader_glsl_priv
*priv
= This
->shader_priv
;
3839 for (i
= 0; i
< tex_type_count
; ++i
)
3841 if (priv
->depth_blt_program
[i
])
3843 GL_EXTCALL(glDeleteObjectARB(priv
->depth_blt_program
[i
]));
3847 hash_table_destroy(priv
->glsl_program_lookup
, NULL
, NULL
);
3848 constant_heap_free(&priv
->pconst_heap
);
3849 constant_heap_free(&priv
->vconst_heap
);
3851 HeapFree(GetProcessHeap(), 0, This
->shader_priv
);
3852 This
->shader_priv
= NULL
;
3855 static BOOL
shader_glsl_dirty_const(IWineD3DDevice
*iface
) {
3856 /* TODO: GL_EXT_bindable_uniform can be used to share constants across shaders */
3860 static GLuint
shader_glsl_generate_pshader(IWineD3DPixelShader
*iface
, SHADER_BUFFER
*buffer
, const struct ps_compile_args
*args
) {
3861 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*)iface
;
3862 const struct shader_reg_maps
*reg_maps
= &This
->baseShader
.reg_maps
;
3863 CONST DWORD
*function
= This
->baseShader
.function
;
3864 const char *fragcolor
;
3865 const WineD3D_GL_Info
*gl_info
= &((IWineD3DDeviceImpl
*)This
->baseShader
.device
)->adapter
->gl_info
;
3867 /* Create the hw GLSL shader object and assign it as the shader->prgId */
3868 GLhandleARB shader_obj
= GL_EXTCALL(glCreateShaderObjectARB(GL_FRAGMENT_SHADER_ARB
));
3870 shader_addline(buffer
, "#version 120\n");
3872 if (GL_SUPPORT(ARB_DRAW_BUFFERS
)) {
3873 shader_addline(buffer
, "#extension GL_ARB_draw_buffers : enable\n");
3875 if (GL_SUPPORT(ARB_TEXTURE_RECTANGLE
)) {
3876 /* The spec says that it doesn't have to be explicitly enabled, but the nvidia
3877 * drivers write a warning if we don't do so
3879 shader_addline(buffer
, "#extension GL_ARB_texture_rectangle : enable\n");
3882 /* Base Declarations */
3883 shader_generate_glsl_declarations( (IWineD3DBaseShader
*) This
, reg_maps
, buffer
, &GLINFO_LOCATION
, args
);
3885 /* Pack 3.0 inputs */
3886 if (reg_maps
->shader_version
>= WINED3DPS_VERSION(3,0) && args
->vp_mode
!= vertexshader
) {
3887 pshader_glsl_input_pack(buffer
, This
->semantics_in
, iface
, args
->vp_mode
);
3890 /* Base Shader Body */
3891 shader_generate_main( (IWineD3DBaseShader
*) This
, buffer
, reg_maps
, function
);
3893 /* Pixel shaders < 2.0 place the resulting color in R0 implicitly */
3894 if (reg_maps
->shader_version
< WINED3DPS_VERSION(2,0))
3896 /* Some older cards like GeforceFX ones don't support multiple buffers, so also not gl_FragData */
3897 if(GL_SUPPORT(ARB_DRAW_BUFFERS
))
3898 shader_addline(buffer
, "gl_FragData[0] = R0;\n");
3900 shader_addline(buffer
, "gl_FragColor = R0;\n");
3903 if(GL_SUPPORT(ARB_DRAW_BUFFERS
)) {
3904 fragcolor
= "gl_FragData[0]";
3906 fragcolor
= "gl_FragColor";
3908 if(args
->srgb_correction
) {
3909 shader_addline(buffer
, "tmp0.xyz = pow(%s.xyz, vec3(%f, %f, %f)) * vec3(%f, %f, %f) - vec3(%f, %f, %f);\n",
3910 fragcolor
, srgb_pow
, srgb_pow
, srgb_pow
, srgb_mul_high
, srgb_mul_high
, srgb_mul_high
,
3911 srgb_sub_high
, srgb_sub_high
, srgb_sub_high
);
3912 shader_addline(buffer
, "tmp1.xyz = %s.xyz * srgb_mul_low.xyz;\n", fragcolor
);
3913 shader_addline(buffer
, "%s.x = %s.x < srgb_comparison.x ? tmp1.x : tmp0.x;\n", fragcolor
, fragcolor
);
3914 shader_addline(buffer
, "%s.y = %s.y < srgb_comparison.y ? tmp1.y : tmp0.y;\n", fragcolor
, fragcolor
);
3915 shader_addline(buffer
, "%s.z = %s.z < srgb_comparison.z ? tmp1.z : tmp0.z;\n", fragcolor
, fragcolor
);
3916 shader_addline(buffer
, "%s = clamp(%s, 0.0, 1.0);\n", fragcolor
, fragcolor
);
3918 /* Pixel shader < 3.0 do not replace the fog stage.
3919 * This implements linear fog computation and blending.
3920 * TODO: non linear fog
3921 * NOTE: gl_Fog.start and gl_Fog.end don't hold fog start s and end e but
3922 * -1/(e-s) and e/(e-s) respectively.
3924 if(reg_maps
->shader_version
< WINED3DPS_VERSION(3,0)) {
3926 case FOG_OFF
: break;
3928 shader_addline(buffer
, "float fogstart = -1.0 / (gl_Fog.end - gl_Fog.start);\n");
3929 shader_addline(buffer
, "float fogend = gl_Fog.end * -fogstart;\n");
3930 shader_addline(buffer
, "float Fog = clamp(gl_FogFragCoord * fogstart + fogend, 0.0, 1.0);\n");
3931 shader_addline(buffer
, "%s.xyz = mix(gl_Fog.color.xyz, %s.xyz, Fog);\n", fragcolor
, fragcolor
);
3934 /* Fog = e^(-gl_Fog.density * gl_FogFragCoord) */
3935 shader_addline(buffer
, "float Fog = exp(-gl_Fog.density * gl_FogFragCoord);\n");
3936 shader_addline(buffer
, "Fog = clamp(Fog, 0.0, 1.0);\n");
3937 shader_addline(buffer
, "%s.xyz = mix(gl_Fog.color.xyz, %s.xyz, Fog);\n", fragcolor
, fragcolor
);
3940 /* Fog = e^(-(gl_Fog.density * gl_FogFragCoord)^2) */
3941 shader_addline(buffer
, "float Fog = exp(-gl_Fog.density * gl_Fog.density * gl_FogFragCoord * gl_FogFragCoord);\n");
3942 shader_addline(buffer
, "Fog = clamp(Fog, 0.0, 1.0);\n");
3943 shader_addline(buffer
, "%s.xyz = mix(gl_Fog.color.xyz, %s.xyz, Fog);\n", fragcolor
, fragcolor
);
3948 shader_addline(buffer
, "}\n");
3950 TRACE("Compiling shader object %u\n", shader_obj
);
3951 GL_EXTCALL(glShaderSourceARB(shader_obj
, 1, (const char**)&buffer
->buffer
, NULL
));
3952 GL_EXTCALL(glCompileShaderARB(shader_obj
));
3953 print_glsl_info_log(&GLINFO_LOCATION
, shader_obj
);
3955 /* Store the shader object */
3959 static GLuint
shader_glsl_generate_vshader(IWineD3DVertexShader
*iface
, SHADER_BUFFER
*buffer
, const struct vs_compile_args
*args
) {
3960 IWineD3DVertexShaderImpl
*This
= (IWineD3DVertexShaderImpl
*)iface
;
3961 const struct shader_reg_maps
*reg_maps
= &This
->baseShader
.reg_maps
;
3962 CONST DWORD
*function
= This
->baseShader
.function
;
3963 const WineD3D_GL_Info
*gl_info
= &((IWineD3DDeviceImpl
*)This
->baseShader
.device
)->adapter
->gl_info
;
3965 /* Create the hw GLSL shader program and assign it as the shader->prgId */
3966 GLhandleARB shader_obj
= GL_EXTCALL(glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB
));
3968 shader_addline(buffer
, "#version 120\n");
3970 /* Base Declarations */
3971 shader_generate_glsl_declarations( (IWineD3DBaseShader
*) This
, reg_maps
, buffer
, &GLINFO_LOCATION
, NULL
);
3973 /* Base Shader Body */
3974 shader_generate_main( (IWineD3DBaseShader
*) This
, buffer
, reg_maps
, function
);
3976 /* Unpack 3.0 outputs */
3977 if (reg_maps
->shader_version
>= WINED3DVS_VERSION(3,0)) shader_addline(buffer
, "order_ps_input(OUT);\n");
3978 else shader_addline(buffer
, "order_ps_input();\n");
3980 /* The D3DRS_FOGTABLEMODE render state defines if the shader-generated fog coord is used
3981 * or if the fragment depth is used. If the fragment depth is used(FOGTABLEMODE != NONE),
3982 * the fog frag coord is thrown away. If the fog frag coord is used, but not written by
3983 * the shader, it is set to 0.0(fully fogged, since start = 1.0, end = 0.0)
3985 if(args
->fog_src
== VS_FOG_Z
) {
3986 shader_addline(buffer
, "gl_FogFragCoord = gl_Position.z;\n");
3987 } else if (!reg_maps
->fog
) {
3988 shader_addline(buffer
, "gl_FogFragCoord = 0.0;\n");
3991 /* Write the final position.
3993 * OpenGL coordinates specify the center of the pixel while d3d coords specify
3994 * the corner. The offsets are stored in z and w in posFixup. posFixup.y contains
3995 * 1.0 or -1.0 to turn the rendering upside down for offscreen rendering. PosFixup.x
3996 * contains 1.0 to allow a mad.
3998 shader_addline(buffer
, "gl_Position.y = gl_Position.y * posFixup.y;\n");
3999 shader_addline(buffer
, "gl_Position.xy += posFixup.zw * gl_Position.ww;\n");
4001 /* Z coord [0;1]->[-1;1] mapping, see comment in transform_projection in state.c
4003 * Basically we want (in homogeneous coordinates) z = z * 2 - 1. However, shaders are run
4004 * before the homogeneous divide, so we have to take the w into account: z = ((z / w) * 2 - 1) * w,
4005 * which is the same as z = z * 2 - w.
4007 shader_addline(buffer
, "gl_Position.z = gl_Position.z * 2.0 - gl_Position.w;\n");
4009 shader_addline(buffer
, "}\n");
4011 TRACE("Compiling shader object %u\n", shader_obj
);
4012 GL_EXTCALL(glShaderSourceARB(shader_obj
, 1, (const char**)&buffer
->buffer
, NULL
));
4013 GL_EXTCALL(glCompileShaderARB(shader_obj
));
4014 print_glsl_info_log(&GLINFO_LOCATION
, shader_obj
);
4019 static void shader_glsl_get_caps(WINED3DDEVTYPE devtype
, const WineD3D_GL_Info
*gl_info
, struct shader_caps
*pCaps
)
4021 /* Nvidia Geforce6/7 or Ati R4xx/R5xx cards with GLSL support, support VS 3.0 but older Nvidia/Ati
4022 * models with GLSL support only support 2.0. In case of nvidia we can detect VS 2.0 support using
4023 * vs_nv_version which is based on NV_vertex_program.
4024 * For Ati cards there's no way using glsl (it abstracts the lowlevel info away) and also not
4025 * using ARB_vertex_program. It is safe to assume that when a card supports pixel shader 2.0 it
4026 * supports vertex shader 2.0 too and the way around. We can detect ps2.0 using the maximum number
4027 * of native instructions, so use that here. For more info see the pixel shader versioning code below.
4029 if((GLINFO_LOCATION
.vs_nv_version
== VS_VERSION_20
) || (GLINFO_LOCATION
.ps_arb_max_instructions
<= 512))
4030 pCaps
->VertexShaderVersion
= WINED3DVS_VERSION(2,0);
4032 pCaps
->VertexShaderVersion
= WINED3DVS_VERSION(3,0);
4033 TRACE_(d3d_caps
)("Hardware vertex shader version %d.%d enabled (GLSL)\n", (pCaps
->VertexShaderVersion
>> 8) & 0xff, pCaps
->VertexShaderVersion
& 0xff);
4034 pCaps
->MaxVertexShaderConst
= GL_LIMITS(vshader_constantsF
);
4036 /* Older DX9-class videocards (GeforceFX / Radeon >9500/X*00) only support pixel shader 2.0/2.0a/2.0b.
4037 * In OpenGL the extensions related to GLSL abstract lowlevel GL info away which is needed
4038 * to distinguish between 2.0 and 3.0 (and 2.0a/2.0b). In case of Nvidia we use their fragment
4039 * program extensions. On other hardware including ATI GL_ARB_fragment_program offers the info
4040 * in max native instructions. Intel and others also offer the info in this extension but they
4041 * don't support GLSL (at least on Windows).
4043 * PS2.0 requires at least 96 instructions, 2.0a/2.0b go up to 512. Assume that if the number
4044 * of instructions is 512 or less we have to do with ps2.0 hardware.
4045 * 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.
4047 if((GLINFO_LOCATION
.ps_nv_version
== PS_VERSION_20
) || (GLINFO_LOCATION
.ps_arb_max_instructions
<= 512))
4048 pCaps
->PixelShaderVersion
= WINED3DPS_VERSION(2,0);
4050 pCaps
->PixelShaderVersion
= WINED3DPS_VERSION(3,0);
4052 /* FIXME: The following line is card dependent. -8.0 to 8.0 is the
4053 * Direct3D minimum requirement.
4055 * Both GL_ARB_fragment_program and GLSL require a "maximum representable magnitude"
4056 * of colors to be 2^10, and 2^32 for other floats. Should we use 1024 here?
4058 * The problem is that the refrast clamps temporary results in the shader to
4059 * [-MaxValue;+MaxValue]. If the card's max value is bigger than the one we advertize here,
4060 * then applications may miss the clamping behavior. On the other hand, if it is smaller,
4061 * the shader will generate incorrect results too. Unfortunately, GL deliberately doesn't
4062 * offer a way to query this.
4064 pCaps
->PixelShader1xMaxValue
= 8.0;
4065 TRACE_(d3d_caps
)("Hardware pixel shader version %d.%d enabled (GLSL)\n", (pCaps
->PixelShaderVersion
>> 8) & 0xff, pCaps
->PixelShaderVersion
& 0xff);
4068 static BOOL
shader_glsl_color_fixup_supported(struct color_fixup_desc fixup
)
4070 if (TRACE_ON(d3d_shader
) && TRACE_ON(d3d
))
4072 TRACE("Checking support for fixup:\n");
4073 dump_color_fixup_desc(fixup
);
4076 /* We support everything except YUV conversions. */
4077 if (!is_yuv_fixup(fixup
))
4083 TRACE("[FAILED]\n");
4087 static const SHADER_HANDLER shader_glsl_instruction_handler_table
[WINED3DSIH_TABLE_SIZE
] =
4089 /* WINED3DSIH_ABS */ shader_glsl_map2gl
,
4090 /* WINED3DSIH_ADD */ shader_glsl_arith
,
4091 /* WINED3DSIH_BEM */ pshader_glsl_bem
,
4092 /* WINED3DSIH_BREAK */ shader_glsl_break
,
4093 /* WINED3DSIH_BREAKC */ shader_glsl_breakc
,
4094 /* WINED3DSIH_BREAKP */ NULL
,
4095 /* WINED3DSIH_CALL */ shader_glsl_call
,
4096 /* WINED3DSIH_CALLNZ */ shader_glsl_callnz
,
4097 /* WINED3DSIH_CMP */ shader_glsl_cmp
,
4098 /* WINED3DSIH_CND */ shader_glsl_cnd
,
4099 /* WINED3DSIH_CRS */ shader_glsl_cross
,
4100 /* WINED3DSIH_DCL */ NULL
,
4101 /* WINED3DSIH_DEF */ NULL
,
4102 /* WINED3DSIH_DEFB */ NULL
,
4103 /* WINED3DSIH_DEFI */ NULL
,
4104 /* WINED3DSIH_DP2ADD */ pshader_glsl_dp2add
,
4105 /* WINED3DSIH_DP3 */ shader_glsl_dot
,
4106 /* WINED3DSIH_DP4 */ shader_glsl_dot
,
4107 /* WINED3DSIH_DST */ shader_glsl_dst
,
4108 /* WINED3DSIH_DSX */ shader_glsl_map2gl
,
4109 /* WINED3DSIH_DSY */ shader_glsl_map2gl
,
4110 /* WINED3DSIH_ELSE */ shader_glsl_else
,
4111 /* WINED3DSIH_ENDIF */ shader_glsl_end
,
4112 /* WINED3DSIH_ENDLOOP */ shader_glsl_end
,
4113 /* WINED3DSIH_ENDREP */ shader_glsl_end
,
4114 /* WINED3DSIH_EXP */ shader_glsl_map2gl
,
4115 /* WINED3DSIH_EXPP */ shader_glsl_expp
,
4116 /* WINED3DSIH_FRC */ shader_glsl_map2gl
,
4117 /* WINED3DSIH_IF */ shader_glsl_if
,
4118 /* WINED3DSIH_IFC */ shader_glsl_ifc
,
4119 /* WINED3DSIH_LABEL */ shader_glsl_label
,
4120 /* WINED3DSIH_LIT */ shader_glsl_lit
,
4121 /* WINED3DSIH_LOG */ shader_glsl_log
,
4122 /* WINED3DSIH_LOGP */ shader_glsl_log
,
4123 /* WINED3DSIH_LOOP */ shader_glsl_loop
,
4124 /* WINED3DSIH_LRP */ shader_glsl_lrp
,
4125 /* WINED3DSIH_M3x2 */ shader_glsl_mnxn
,
4126 /* WINED3DSIH_M3x3 */ shader_glsl_mnxn
,
4127 /* WINED3DSIH_M3x4 */ shader_glsl_mnxn
,
4128 /* WINED3DSIH_M4x3 */ shader_glsl_mnxn
,
4129 /* WINED3DSIH_M4x4 */ shader_glsl_mnxn
,
4130 /* WINED3DSIH_MAD */ shader_glsl_mad
,
4131 /* WINED3DSIH_MAX */ shader_glsl_map2gl
,
4132 /* WINED3DSIH_MIN */ shader_glsl_map2gl
,
4133 /* WINED3DSIH_MOV */ shader_glsl_mov
,
4134 /* WINED3DSIH_MOVA */ shader_glsl_mov
,
4135 /* WINED3DSIH_MUL */ shader_glsl_arith
,
4136 /* WINED3DSIH_NOP */ NULL
,
4137 /* WINED3DSIH_NRM */ shader_glsl_map2gl
,
4138 /* WINED3DSIH_PHASE */ NULL
,
4139 /* WINED3DSIH_POW */ shader_glsl_pow
,
4140 /* WINED3DSIH_RCP */ shader_glsl_rcp
,
4141 /* WINED3DSIH_REP */ shader_glsl_rep
,
4142 /* WINED3DSIH_RET */ NULL
,
4143 /* WINED3DSIH_RSQ */ shader_glsl_rsq
,
4144 /* WINED3DSIH_SETP */ NULL
,
4145 /* WINED3DSIH_SGE */ shader_glsl_compare
,
4146 /* WINED3DSIH_SGN */ shader_glsl_map2gl
,
4147 /* WINED3DSIH_SINCOS */ shader_glsl_sincos
,
4148 /* WINED3DSIH_SLT */ shader_glsl_compare
,
4149 /* WINED3DSIH_SUB */ shader_glsl_arith
,
4150 /* WINED3DSIH_TEX */ pshader_glsl_tex
,
4151 /* WINED3DSIH_TEXBEM */ pshader_glsl_texbem
,
4152 /* WINED3DSIH_TEXBEML */ pshader_glsl_texbem
,
4153 /* WINED3DSIH_TEXCOORD */ pshader_glsl_texcoord
,
4154 /* WINED3DSIH_TEXDEPTH */ pshader_glsl_texdepth
,
4155 /* WINED3DSIH_TEXDP3 */ pshader_glsl_texdp3
,
4156 /* WINED3DSIH_TEXDP3TEX */ pshader_glsl_texdp3tex
,
4157 /* WINED3DSIH_TEXKILL */ pshader_glsl_texkill
,
4158 /* WINED3DSIH_TEXLDD */ NULL
,
4159 /* WINED3DSIH_TEXLDL */ shader_glsl_texldl
,
4160 /* WINED3DSIH_TEXM3x2DEPTH */ pshader_glsl_texm3x2depth
,
4161 /* WINED3DSIH_TEXM3x2PAD */ pshader_glsl_texm3x2pad
,
4162 /* WINED3DSIH_TEXM3x2TEX */ pshader_glsl_texm3x2tex
,
4163 /* WINED3DSIH_TEXM3x3 */ pshader_glsl_texm3x3
,
4164 /* WINED3DSIH_TEXM3x3DIFF */ NULL
,
4165 /* WINED3DSIH_TEXM3x3PAD */ pshader_glsl_texm3x3pad
,
4166 /* WINED3DSIH_TEXM3x3SPEC */ pshader_glsl_texm3x3spec
,
4167 /* WINED3DSIH_TEXM3x3TEX */ pshader_glsl_texm3x3tex
,
4168 /* WINED3DSIH_TEXM3x3VSPEC */ pshader_glsl_texm3x3vspec
,
4169 /* WINED3DSIH_TEXREG2AR */ pshader_glsl_texreg2ar
,
4170 /* WINED3DSIH_TEXREG2GB */ pshader_glsl_texreg2gb
,
4171 /* WINED3DSIH_TEXREG2RGB */ pshader_glsl_texreg2rgb
,
4174 const shader_backend_t glsl_shader_backend
= {
4175 shader_glsl_instruction_handler_table
,
4177 shader_glsl_select_depth_blt
,
4178 shader_glsl_deselect_depth_blt
,
4179 shader_glsl_update_float_vertex_constants
,
4180 shader_glsl_update_float_pixel_constants
,
4181 shader_glsl_load_constants
,
4182 shader_glsl_destroy
,
4185 shader_glsl_dirty_const
,
4186 shader_glsl_generate_pshader
,
4187 shader_glsl_generate_vshader
,
4188 shader_glsl_get_caps
,
4189 shader_glsl_color_fixup_supported
,