2 * GLSL pixel and vertex shader implementation
4 * Copyright 2006 Jason Green
5 * Copyright 2006-2007 Henri Verbeet
6 * Copyright 2007-2008 Stefan Dösinger for CodeWeavers
7 * Copyright 2009 Henri Verbeet for CodeWeavers
9 * This library is free software; you can redistribute it and/or
10 * modify it under the terms of the GNU Lesser General Public
11 * License as published by the Free Software Foundation; either
12 * version 2.1 of the License, or (at your option) any later version.
14 * This library is distributed in the hope that it will be useful,
15 * but WITHOUT ANY WARRANTY; without even the implied warranty of
16 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
17 * Lesser General Public License for more details.
19 * You should have received a copy of the GNU Lesser General Public
20 * License along with this library; if not, write to the Free Software
21 * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301, USA
25 * D3D shader asm has swizzles on source parameters, and write masks for
26 * destination parameters. GLSL uses swizzles for both. The result of this is
27 * that for example "mov dst.xw, src.zyxw" becomes "dst.xw = src.zw" in GLSL.
28 * Ie, to generate a proper GLSL source swizzle, we need to take the D3D write
29 * mask for the destination parameter into account.
35 #include "wined3d_private.h"
37 WINE_DEFAULT_DEBUG_CHANNEL(d3d_shader
);
38 WINE_DECLARE_DEBUG_CHANNEL(d3d_constants
);
39 WINE_DECLARE_DEBUG_CHANNEL(d3d_caps
);
40 WINE_DECLARE_DEBUG_CHANNEL(d3d
);
42 #define GLINFO_LOCATION (*gl_info)
44 #define WINED3D_GLSL_SAMPLE_PROJECTED 0x1
45 #define WINED3D_GLSL_SAMPLE_RECT 0x2
46 #define WINED3D_GLSL_SAMPLE_LOD 0x4
47 #define WINED3D_GLSL_SAMPLE_GRAD 0x8
62 } glsl_sample_function_t
;
66 HEAP_NODE_TRAVERSE_LEFT
,
67 HEAP_NODE_TRAVERSE_RIGHT
,
79 struct constant_entry
*entries
;
80 unsigned int *positions
;
84 /* GLSL shader private data */
85 struct shader_glsl_priv
{
86 struct hash_table_t
*glsl_program_lookup
;
87 struct glsl_shader_prog_link
*glsl_program
;
88 struct constant_heap vconst_heap
;
89 struct constant_heap pconst_heap
;
91 GLhandleARB depth_blt_program
[tex_type_count
];
92 UINT next_constant_version
;
95 /* Struct to maintain data about a linked GLSL program */
96 struct glsl_shader_prog_link
{
97 struct list vshader_entry
;
98 struct list pshader_entry
;
99 GLhandleARB programId
;
100 GLint
*vuniformF_locations
;
101 GLint
*puniformF_locations
;
102 GLint vuniformI_locations
[MAX_CONST_I
];
103 GLint puniformI_locations
[MAX_CONST_I
];
104 GLint posFixup_location
;
105 GLint np2Fixup_location
[MAX_FRAGMENT_SAMPLERS
];
106 GLint bumpenvmat_location
[MAX_TEXTURES
];
107 GLint luminancescale_location
[MAX_TEXTURES
];
108 GLint luminanceoffset_location
[MAX_TEXTURES
];
109 GLint ycorrection_location
;
110 GLenum vertex_color_clamp
;
111 IWineD3DVertexShader
*vshader
;
112 IWineD3DPixelShader
*pshader
;
113 struct vs_compile_args vs_args
;
114 struct ps_compile_args ps_args
;
115 UINT constant_version
;
119 IWineD3DVertexShader
*vshader
;
120 IWineD3DPixelShader
*pshader
;
121 struct ps_compile_args ps_args
;
122 struct vs_compile_args vs_args
;
123 } glsl_program_key_t
;
126 /** Prints the GLSL info log which will contain error messages if they exist */
127 static void print_glsl_info_log(const WineD3D_GL_Info
*gl_info
, GLhandleARB obj
)
129 int infologLength
= 0;
134 static const char * const spam
[] =
136 "Vertex shader was successfully compiled to run on hardware.\n", /* fglrx */
137 "Fragment shader was successfully compiled to run on hardware.\n", /* fglrx */
138 "Fragment shader(s) linked, vertex shader(s) linked. \n ", /* fglrx, with \n */
139 "Fragment shader(s) linked, vertex shader(s) linked.", /* fglrx, no \n */
140 "Vertex shader(s) linked, no fragment shader(s) defined. \n ", /* fglrx, with \n */
141 "Vertex shader(s) linked, no fragment shader(s) defined.", /* fglrx, no \n */
142 "Fragment shader was successfully compiled to run on hardware.\n"
143 "WARNING: 0:2: extension 'GL_ARB_draw_buffers' is not supported",
144 "Fragment shader(s) linked, no vertex shader(s) defined.", /* fglrx, no \n */
145 "Fragment shader(s) linked, no vertex shader(s) defined. \n ", /* fglrx, with \n */
146 "WARNING: 0:2: extension 'GL_ARB_draw_buffers' is not supported\n" /* MacOS ati */
149 if (!TRACE_ON(d3d_shader
) && !FIXME_ON(d3d_shader
)) return;
151 GL_EXTCALL(glGetObjectParameterivARB(obj
,
152 GL_OBJECT_INFO_LOG_LENGTH_ARB
,
155 /* A size of 1 is just a null-terminated string, so the log should be bigger than
156 * that if there are errors. */
157 if (infologLength
> 1)
159 /* Fglrx doesn't terminate the string properly, but it tells us the proper length.
160 * So use HEAP_ZERO_MEMORY to avoid uninitialized bytes
162 infoLog
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, infologLength
);
163 GL_EXTCALL(glGetInfoLogARB(obj
, infologLength
, NULL
, infoLog
));
166 for(i
= 0; i
< sizeof(spam
) / sizeof(spam
[0]); i
++) {
167 if(strcmp(infoLog
, spam
[i
]) == 0) {
173 TRACE("Spam received from GLSL shader #%u: %s\n", obj
, debugstr_a(infoLog
));
175 FIXME("Error received from GLSL shader #%u: %s\n", obj
, debugstr_a(infoLog
));
177 HeapFree(GetProcessHeap(), 0, infoLog
);
182 * Loads (pixel shader) samplers
184 static void shader_glsl_load_psamplers(const WineD3D_GL_Info
*gl_info
, DWORD
*tex_unit_map
, GLhandleARB programId
)
188 char sampler_name
[20];
190 for (i
= 0; i
< MAX_FRAGMENT_SAMPLERS
; ++i
) {
191 snprintf(sampler_name
, sizeof(sampler_name
), "Psampler%d", i
);
192 name_loc
= GL_EXTCALL(glGetUniformLocationARB(programId
, sampler_name
));
193 if (name_loc
!= -1) {
194 DWORD mapped_unit
= tex_unit_map
[i
];
195 if (mapped_unit
!= WINED3D_UNMAPPED_STAGE
&& mapped_unit
< GL_LIMITS(fragment_samplers
))
197 TRACE("Loading %s for texture %d\n", sampler_name
, mapped_unit
);
198 GL_EXTCALL(glUniform1iARB(name_loc
, mapped_unit
));
199 checkGLcall("glUniform1iARB");
201 ERR("Trying to load sampler %s on unsupported unit %d\n", sampler_name
, mapped_unit
);
207 static void shader_glsl_load_vsamplers(const WineD3D_GL_Info
*gl_info
, DWORD
*tex_unit_map
, GLhandleARB programId
)
210 char sampler_name
[20];
213 for (i
= 0; i
< MAX_VERTEX_SAMPLERS
; ++i
) {
214 snprintf(sampler_name
, sizeof(sampler_name
), "Vsampler%d", i
);
215 name_loc
= GL_EXTCALL(glGetUniformLocationARB(programId
, sampler_name
));
216 if (name_loc
!= -1) {
217 DWORD mapped_unit
= tex_unit_map
[MAX_FRAGMENT_SAMPLERS
+ i
];
218 if (mapped_unit
!= WINED3D_UNMAPPED_STAGE
&& mapped_unit
< GL_LIMITS(combined_samplers
))
220 TRACE("Loading %s for texture %d\n", sampler_name
, mapped_unit
);
221 GL_EXTCALL(glUniform1iARB(name_loc
, mapped_unit
));
222 checkGLcall("glUniform1iARB");
224 ERR("Trying to load sampler %s on unsupported unit %d\n", sampler_name
, mapped_unit
);
230 static inline void walk_constant_heap(const WineD3D_GL_Info
*gl_info
, const float *constants
,
231 const GLint
*constant_locations
, const struct constant_heap
*heap
, unsigned char *stack
, DWORD version
)
234 unsigned int heap_idx
= 1;
237 if (heap
->entries
[heap_idx
].version
<= version
) return;
239 idx
= heap
->entries
[heap_idx
].idx
;
240 if (constant_locations
[idx
] != -1) GL_EXTCALL(glUniform4fvARB(constant_locations
[idx
], 1, &constants
[idx
* 4]));
241 stack
[stack_idx
] = HEAP_NODE_TRAVERSE_LEFT
;
243 while (stack_idx
>= 0)
245 /* Note that we fall through to the next case statement. */
246 switch(stack
[stack_idx
])
248 case HEAP_NODE_TRAVERSE_LEFT
:
250 unsigned int left_idx
= heap_idx
<< 1;
251 if (left_idx
< heap
->size
&& heap
->entries
[left_idx
].version
> version
)
254 idx
= heap
->entries
[heap_idx
].idx
;
255 if (constant_locations
[idx
] != -1)
256 GL_EXTCALL(glUniform4fvARB(constant_locations
[idx
], 1, &constants
[idx
* 4]));
258 stack
[stack_idx
++] = HEAP_NODE_TRAVERSE_RIGHT
;
259 stack
[stack_idx
] = HEAP_NODE_TRAVERSE_LEFT
;
264 case HEAP_NODE_TRAVERSE_RIGHT
:
266 unsigned int right_idx
= (heap_idx
<< 1) + 1;
267 if (right_idx
< heap
->size
&& heap
->entries
[right_idx
].version
> version
)
269 heap_idx
= right_idx
;
270 idx
= heap
->entries
[heap_idx
].idx
;
271 if (constant_locations
[idx
] != -1)
272 GL_EXTCALL(glUniform4fvARB(constant_locations
[idx
], 1, &constants
[idx
* 4]));
274 stack
[stack_idx
++] = HEAP_NODE_POP
;
275 stack
[stack_idx
] = HEAP_NODE_TRAVERSE_LEFT
;
288 checkGLcall("walk_constant_heap()");
291 static inline void apply_clamped_constant(const WineD3D_GL_Info
*gl_info
, GLint location
, const GLfloat
*data
)
293 GLfloat clamped_constant
[4];
295 if (location
== -1) return;
297 clamped_constant
[0] = data
[0] < -1.0f
? -1.0f
: data
[0] > 1.0 ? 1.0 : data
[0];
298 clamped_constant
[1] = data
[1] < -1.0f
? -1.0f
: data
[1] > 1.0 ? 1.0 : data
[1];
299 clamped_constant
[2] = data
[2] < -1.0f
? -1.0f
: data
[2] > 1.0 ? 1.0 : data
[2];
300 clamped_constant
[3] = data
[3] < -1.0f
? -1.0f
: data
[3] > 1.0 ? 1.0 : data
[3];
302 GL_EXTCALL(glUniform4fvARB(location
, 1, clamped_constant
));
305 static inline void walk_constant_heap_clamped(const WineD3D_GL_Info
*gl_info
, const float *constants
,
306 const GLint
*constant_locations
, const struct constant_heap
*heap
, unsigned char *stack
, DWORD version
)
309 unsigned int heap_idx
= 1;
312 if (heap
->entries
[heap_idx
].version
<= version
) return;
314 idx
= heap
->entries
[heap_idx
].idx
;
315 apply_clamped_constant(gl_info
, constant_locations
[idx
], &constants
[idx
* 4]);
316 stack
[stack_idx
] = HEAP_NODE_TRAVERSE_LEFT
;
318 while (stack_idx
>= 0)
320 /* Note that we fall through to the next case statement. */
321 switch(stack
[stack_idx
])
323 case HEAP_NODE_TRAVERSE_LEFT
:
325 unsigned int left_idx
= heap_idx
<< 1;
326 if (left_idx
< heap
->size
&& heap
->entries
[left_idx
].version
> version
)
329 idx
= heap
->entries
[heap_idx
].idx
;
330 apply_clamped_constant(gl_info
, constant_locations
[idx
], &constants
[idx
* 4]);
332 stack
[stack_idx
++] = HEAP_NODE_TRAVERSE_RIGHT
;
333 stack
[stack_idx
] = HEAP_NODE_TRAVERSE_LEFT
;
338 case HEAP_NODE_TRAVERSE_RIGHT
:
340 unsigned int right_idx
= (heap_idx
<< 1) + 1;
341 if (right_idx
< heap
->size
&& heap
->entries
[right_idx
].version
> version
)
343 heap_idx
= right_idx
;
344 idx
= heap
->entries
[heap_idx
].idx
;
345 apply_clamped_constant(gl_info
, constant_locations
[idx
], &constants
[idx
* 4]);
347 stack
[stack_idx
++] = HEAP_NODE_POP
;
348 stack
[stack_idx
] = HEAP_NODE_TRAVERSE_LEFT
;
361 checkGLcall("walk_constant_heap_clamped()");
364 /* Loads floating point constants (aka uniforms) into the currently set GLSL program. */
365 static void shader_glsl_load_constantsF(IWineD3DBaseShaderImpl
*This
, const WineD3D_GL_Info
*gl_info
,
366 const float *constants
, const GLint
*constant_locations
, const struct constant_heap
*heap
,
367 unsigned char *stack
, UINT version
)
369 const local_constant
*lconst
;
371 /* 1.X pshaders have the constants clamped to [-1;1] implicitly. */
372 if (This
->baseShader
.reg_maps
.shader_version
.major
== 1
373 && shader_is_pshader_version(This
->baseShader
.reg_maps
.shader_version
.type
))
374 walk_constant_heap_clamped(gl_info
, constants
, constant_locations
, heap
, stack
, version
);
376 walk_constant_heap(gl_info
, constants
, constant_locations
, heap
, stack
, version
);
378 if (!This
->baseShader
.load_local_constsF
)
380 TRACE("No need to load local float constants for this shader\n");
384 /* Immediate constants are clamped to [-1;1] at shader creation time if needed */
385 LIST_FOR_EACH_ENTRY(lconst
, &This
->baseShader
.constantsF
, local_constant
, entry
)
387 GLint location
= constant_locations
[lconst
->idx
];
388 /* We found this uniform name in the program - go ahead and send the data */
389 if (location
!= -1) GL_EXTCALL(glUniform4fvARB(location
, 1, (const GLfloat
*)lconst
->value
));
391 checkGLcall("glUniform4fvARB()");
394 /* Loads integer constants (aka uniforms) into the currently set GLSL program. */
395 static void shader_glsl_load_constantsI(IWineD3DBaseShaderImpl
*This
, const WineD3D_GL_Info
*gl_info
,
396 const GLint locations
[MAX_CONST_I
], const int *constants
, WORD constants_set
)
401 for (i
= 0; constants_set
; constants_set
>>= 1, ++i
)
403 if (!(constants_set
& 1)) continue;
405 TRACE_(d3d_constants
)("Loading constants %u: %i, %i, %i, %i\n",
406 i
, constants
[i
*4], constants
[i
*4+1], constants
[i
*4+2], constants
[i
*4+3]);
408 /* We found this uniform name in the program - go ahead and send the data */
409 GL_EXTCALL(glUniform4ivARB(locations
[i
], 1, &constants
[i
*4]));
410 checkGLcall("glUniform4ivARB");
413 /* Load immediate constants */
414 ptr
= list_head(&This
->baseShader
.constantsI
);
416 const struct local_constant
*lconst
= LIST_ENTRY(ptr
, const struct local_constant
, entry
);
417 unsigned int idx
= lconst
->idx
;
418 const GLint
*values
= (const GLint
*)lconst
->value
;
420 TRACE_(d3d_constants
)("Loading local constants %i: %i, %i, %i, %i\n", idx
,
421 values
[0], values
[1], values
[2], values
[3]);
423 /* We found this uniform name in the program - go ahead and send the data */
424 GL_EXTCALL(glUniform4ivARB(locations
[idx
], 1, values
));
425 checkGLcall("glUniform4ivARB");
426 ptr
= list_next(&This
->baseShader
.constantsI
, ptr
);
430 /* Loads boolean constants (aka uniforms) into the currently set GLSL program. */
431 static void shader_glsl_load_constantsB(IWineD3DBaseShaderImpl
*This
, const WineD3D_GL_Info
*gl_info
,
432 GLhandleARB programId
, const BOOL
*constants
, WORD constants_set
)
437 char is_pshader
= shader_is_pshader_version(This
->baseShader
.reg_maps
.shader_version
.type
);
438 const char* prefix
= is_pshader
? "PB":"VB";
441 /* TODO: Benchmark and see if it would be beneficial to store the
442 * locations of the constants to avoid looking up each time */
443 for (i
= 0; constants_set
; constants_set
>>= 1, ++i
)
445 if (!(constants_set
& 1)) continue;
447 TRACE_(d3d_constants
)("Loading constants %i: %i;\n", i
, constants
[i
]);
449 /* TODO: Benchmark and see if it would be beneficial to store the
450 * locations of the constants to avoid looking up each time */
451 snprintf(tmp_name
, sizeof(tmp_name
), "%s[%i]", prefix
, i
);
452 tmp_loc
= GL_EXTCALL(glGetUniformLocationARB(programId
, tmp_name
));
455 /* We found this uniform name in the program - go ahead and send the data */
456 GL_EXTCALL(glUniform1ivARB(tmp_loc
, 1, &constants
[i
]));
457 checkGLcall("glUniform1ivARB");
461 /* Load immediate constants */
462 ptr
= list_head(&This
->baseShader
.constantsB
);
464 const struct local_constant
*lconst
= LIST_ENTRY(ptr
, const struct local_constant
, entry
);
465 unsigned int idx
= lconst
->idx
;
466 const GLint
*values
= (const GLint
*)lconst
->value
;
468 TRACE_(d3d_constants
)("Loading local constants %i: %i\n", idx
, values
[0]);
470 snprintf(tmp_name
, sizeof(tmp_name
), "%s[%i]", prefix
, idx
);
471 tmp_loc
= GL_EXTCALL(glGetUniformLocationARB(programId
, tmp_name
));
473 /* We found this uniform name in the program - go ahead and send the data */
474 GL_EXTCALL(glUniform1ivARB(tmp_loc
, 1, values
));
475 checkGLcall("glUniform1ivARB");
477 ptr
= list_next(&This
->baseShader
.constantsB
, ptr
);
481 static void reset_program_constant_version(void *value
, void *context
)
483 struct glsl_shader_prog_link
*entry
= value
;
484 entry
->constant_version
= 0;
488 * Loads the texture dimensions for NP2 fixup into the currently set GLSL program.
490 static void shader_glsl_load_np2fixup_constants(
491 IWineD3DDevice
* device
,
493 char useVertexShader
) {
495 const IWineD3DDeviceImpl
* deviceImpl
= (const IWineD3DDeviceImpl
*) device
;
496 const struct glsl_shader_prog_link
* prog
= ((struct shader_glsl_priv
*)(deviceImpl
->shader_priv
))->glsl_program
;
499 /* No GLSL program set - nothing to do. */
503 if (!usePixelShader
) {
504 /* NP2 texcoord fixup is (currently) only done for pixelshaders. */
508 if (prog
->ps_args
.np2_fixup
) {
510 UINT fixup
= prog
->ps_args
.np2_fixup
;
511 const WineD3D_GL_Info
* gl_info
= &deviceImpl
->adapter
->gl_info
;
512 const IWineD3DStateBlockImpl
* stateBlock
= (const IWineD3DStateBlockImpl
*) deviceImpl
->stateBlock
;
514 for (i
= 0; fixup
; fixup
>>= 1, ++i
) {
515 if (-1 != prog
->np2Fixup_location
[i
]) {
516 const IWineD3DBaseTextureImpl
* const tex
= (const IWineD3DBaseTextureImpl
*) stateBlock
->textures
[i
];
518 FIXME("Nonexistent texture is flagged for NP2 texcoord fixup\n");
521 const float tex_dim
[2] = {tex
->baseTexture
.pow2Matrix
[0], tex
->baseTexture
.pow2Matrix
[5]};
522 GL_EXTCALL(glUniform2fvARB(prog
->np2Fixup_location
[i
], 1, tex_dim
));
530 * Loads the app-supplied constants into the currently set GLSL program.
532 static void shader_glsl_load_constants(
533 IWineD3DDevice
* device
,
535 char useVertexShader
) {
537 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) device
;
538 struct shader_glsl_priv
*priv
= deviceImpl
->shader_priv
;
539 IWineD3DStateBlockImpl
* stateBlock
= deviceImpl
->stateBlock
;
540 const WineD3D_GL_Info
*gl_info
= &deviceImpl
->adapter
->gl_info
;
542 GLhandleARB programId
;
543 struct glsl_shader_prog_link
*prog
= priv
->glsl_program
;
544 UINT constant_version
;
548 /* No GLSL program set - nothing to do. */
551 programId
= prog
->programId
;
552 constant_version
= prog
->constant_version
;
554 if (useVertexShader
) {
555 IWineD3DBaseShaderImpl
* vshader
= (IWineD3DBaseShaderImpl
*) stateBlock
->vertexShader
;
557 /* Load DirectX 9 float constants/uniforms for vertex shader */
558 shader_glsl_load_constantsF(vshader
, gl_info
, stateBlock
->vertexShaderConstantF
,
559 prog
->vuniformF_locations
, &priv
->vconst_heap
, priv
->stack
, constant_version
);
561 /* Load DirectX 9 integer constants/uniforms for vertex shader */
562 shader_glsl_load_constantsI(vshader
, gl_info
, prog
->vuniformI_locations
, stateBlock
->vertexShaderConstantI
,
563 stateBlock
->changed
.vertexShaderConstantsI
& vshader
->baseShader
.reg_maps
.integer_constants
);
565 /* Load DirectX 9 boolean constants/uniforms for vertex shader */
566 shader_glsl_load_constantsB(vshader
, gl_info
, programId
, stateBlock
->vertexShaderConstantB
,
567 stateBlock
->changed
.vertexShaderConstantsB
& vshader
->baseShader
.reg_maps
.boolean_constants
);
569 /* Upload the position fixup params */
570 GL_EXTCALL(glUniform4fvARB(prog
->posFixup_location
, 1, &deviceImpl
->posFixup
[0]));
571 checkGLcall("glUniform4fvARB");
574 if (usePixelShader
) {
576 IWineD3DBaseShaderImpl
* pshader
= (IWineD3DBaseShaderImpl
*) stateBlock
->pixelShader
;
578 /* Load DirectX 9 float constants/uniforms for pixel shader */
579 shader_glsl_load_constantsF(pshader
, gl_info
, stateBlock
->pixelShaderConstantF
,
580 prog
->puniformF_locations
, &priv
->pconst_heap
, priv
->stack
, constant_version
);
582 /* Load DirectX 9 integer constants/uniforms for pixel shader */
583 shader_glsl_load_constantsI(pshader
, gl_info
, prog
->puniformI_locations
, stateBlock
->pixelShaderConstantI
,
584 stateBlock
->changed
.pixelShaderConstantsI
& pshader
->baseShader
.reg_maps
.integer_constants
);
586 /* Load DirectX 9 boolean constants/uniforms for pixel shader */
587 shader_glsl_load_constantsB(pshader
, gl_info
, programId
, stateBlock
->pixelShaderConstantB
,
588 stateBlock
->changed
.pixelShaderConstantsB
& pshader
->baseShader
.reg_maps
.boolean_constants
);
590 /* Upload the environment bump map matrix if needed. The needsbumpmat member specifies the texture stage to load the matrix from.
591 * It can't be 0 for a valid texbem instruction.
593 for(i
= 0; i
< ((IWineD3DPixelShaderImpl
*) pshader
)->numbumpenvmatconsts
; i
++) {
594 IWineD3DPixelShaderImpl
*ps
= (IWineD3DPixelShaderImpl
*) pshader
;
595 int stage
= ps
->luminanceconst
[i
].texunit
;
597 const float *data
= (const float *)&stateBlock
->textureState
[(int)ps
->bumpenvmatconst
[i
].texunit
][WINED3DTSS_BUMPENVMAT00
];
598 GL_EXTCALL(glUniformMatrix2fvARB(prog
->bumpenvmat_location
[i
], 1, 0, data
));
599 checkGLcall("glUniformMatrix2fvARB");
601 /* texbeml needs the luminance scale and offset too. If texbeml is used, needsbumpmat
602 * is set too, so we can check that in the needsbumpmat check
604 if(ps
->baseShader
.reg_maps
.luminanceparams
[stage
]) {
605 const GLfloat
*scale
= (const GLfloat
*)&stateBlock
->textureState
[stage
][WINED3DTSS_BUMPENVLSCALE
];
606 const GLfloat
*offset
= (const GLfloat
*)&stateBlock
->textureState
[stage
][WINED3DTSS_BUMPENVLOFFSET
];
608 GL_EXTCALL(glUniform1fvARB(prog
->luminancescale_location
[i
], 1, scale
));
609 checkGLcall("glUniform1fvARB");
610 GL_EXTCALL(glUniform1fvARB(prog
->luminanceoffset_location
[i
], 1, offset
));
611 checkGLcall("glUniform1fvARB");
615 if(((IWineD3DPixelShaderImpl
*) pshader
)->vpos_uniform
) {
616 float correction_params
[4];
617 if(deviceImpl
->render_offscreen
) {
618 correction_params
[0] = 0.0;
619 correction_params
[1] = 1.0;
621 /* position is window relative, not viewport relative */
622 correction_params
[0] = ((IWineD3DSurfaceImpl
*) deviceImpl
->render_targets
[0])->currentDesc
.Height
;
623 correction_params
[1] = -1.0;
625 GL_EXTCALL(glUniform4fvARB(prog
->ycorrection_location
, 1, correction_params
));
629 if (priv
->next_constant_version
== UINT_MAX
)
631 TRACE("Max constant version reached, resetting to 0.\n");
632 hash_table_for_each_entry(priv
->glsl_program_lookup
, reset_program_constant_version
, NULL
);
633 priv
->next_constant_version
= 1;
637 prog
->constant_version
= priv
->next_constant_version
++;
641 static inline void update_heap_entry(struct constant_heap
*heap
, unsigned int idx
,
642 unsigned int heap_idx
, DWORD new_version
)
644 struct constant_entry
*entries
= heap
->entries
;
645 unsigned int *positions
= heap
->positions
;
646 unsigned int parent_idx
;
650 parent_idx
= heap_idx
>> 1;
652 if (new_version
<= entries
[parent_idx
].version
) break;
654 entries
[heap_idx
] = entries
[parent_idx
];
655 positions
[entries
[parent_idx
].idx
] = heap_idx
;
656 heap_idx
= parent_idx
;
659 entries
[heap_idx
].version
= new_version
;
660 entries
[heap_idx
].idx
= idx
;
661 positions
[idx
] = heap_idx
;
664 static void shader_glsl_update_float_vertex_constants(IWineD3DDevice
*iface
, UINT start
, UINT count
)
666 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
667 struct shader_glsl_priv
*priv
= This
->shader_priv
;
668 struct constant_heap
*heap
= &priv
->vconst_heap
;
671 for (i
= start
; i
< count
+ start
; ++i
)
673 if (!This
->stateBlock
->changed
.vertexShaderConstantsF
[i
])
674 update_heap_entry(heap
, i
, heap
->size
++, priv
->next_constant_version
);
676 update_heap_entry(heap
, i
, heap
->positions
[i
], priv
->next_constant_version
);
680 static void shader_glsl_update_float_pixel_constants(IWineD3DDevice
*iface
, UINT start
, UINT count
)
682 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
683 struct shader_glsl_priv
*priv
= This
->shader_priv
;
684 struct constant_heap
*heap
= &priv
->pconst_heap
;
687 for (i
= start
; i
< count
+ start
; ++i
)
689 if (!This
->stateBlock
->changed
.pixelShaderConstantsF
[i
])
690 update_heap_entry(heap
, i
, heap
->size
++, priv
->next_constant_version
);
692 update_heap_entry(heap
, i
, heap
->positions
[i
], priv
->next_constant_version
);
696 /** Generate the variable & register declarations for the GLSL output target */
697 static void shader_generate_glsl_declarations(IWineD3DBaseShader
*iface
, const shader_reg_maps
*reg_maps
,
698 SHADER_BUFFER
*buffer
, const WineD3D_GL_Info
*gl_info
,
699 const struct ps_compile_args
*ps_args
)
701 IWineD3DBaseShaderImpl
* This
= (IWineD3DBaseShaderImpl
*) iface
;
702 IWineD3DDeviceImpl
*device
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
703 unsigned int i
, extra_constants_needed
= 0;
704 const local_constant
*lconst
;
706 /* There are some minor differences between pixel and vertex shaders */
707 char pshader
= shader_is_pshader_version(reg_maps
->shader_version
.type
);
708 char prefix
= pshader
? 'P' : 'V';
710 /* Prototype the subroutines */
711 for (i
= 0; i
< This
->baseShader
.limits
.label
; i
++) {
712 if (reg_maps
->labels
[i
])
713 shader_addline(buffer
, "void subroutine%u();\n", i
);
716 /* Declare the constants (aka uniforms) */
717 if (This
->baseShader
.limits
.constant_float
> 0) {
718 unsigned max_constantsF
;
719 /* Unless the shader uses indirect addressing, always declare the maximum array size and ignore that we need some
720 * uniforms privately. E.g. if GL supports 256 uniforms, and we need 2 for the pos fixup and immediate values, still
721 * declare VC[256]. If the shader needs more uniforms than we have it won't work in any case. If it uses less, the
722 * compiler will figure out which uniforms are really used and strip them out. This allows a shader to use c255 on
723 * a dx9 card, as long as it doesn't also use all the other constants.
725 * If the shader uses indirect addressing the compiler must assume that all declared uniforms are used. In this case,
726 * declare only the amount that we're assured to have.
728 * Thus we run into problems in these two cases:
729 * 1) The shader really uses more uniforms than supported
730 * 2) The shader uses indirect addressing, less constants than supported, but uses a constant index > #supported consts
733 /* No indirect addressing here */
734 max_constantsF
= GL_LIMITS(pshader_constantsF
);
736 if(This
->baseShader
.reg_maps
.usesrelconstF
) {
737 /* Subtract the other potential uniforms from the max available (bools, ints, and 1 row of projection matrix).
738 * Subtract another uniform for immediate values, which have to be loaded via uniform by the driver as well.
739 * The shader code only uses 0.5, 2.0, 1.0, 128 and -128 in vertex shader code, so one vec4 should be enough
740 * (Unfortunately the Nvidia driver doesn't store 128 and -128 in one float
742 max_constantsF
= GL_LIMITS(vshader_constantsF
) - 3;
743 max_constantsF
-= count_bits(This
->baseShader
.reg_maps
.integer_constants
);
744 /* Strictly speaking a bool only uses one scalar, but the nvidia(Linux) compiler doesn't pack them properly,
745 * so each scalar requires a full vec4. We could work around this by packing the booleans ourselves, but
746 * for now take this into account when calculating the number of available constants
748 max_constantsF
-= count_bits(This
->baseShader
.reg_maps
.boolean_constants
);
749 /* Set by driver quirks in directx.c */
750 max_constantsF
-= GLINFO_LOCATION
.reserved_glsl_constants
;
752 max_constantsF
= GL_LIMITS(vshader_constantsF
);
755 max_constantsF
= min(This
->baseShader
.limits
.constant_float
, max_constantsF
);
756 shader_addline(buffer
, "uniform vec4 %cC[%u];\n", prefix
, max_constantsF
);
759 /* Always declare the full set of constants, the compiler can remove the unused ones because d3d doesn't(yet)
760 * support indirect int and bool constant addressing. This avoids problems if the app uses e.g. i0 and i9.
762 if (This
->baseShader
.limits
.constant_int
> 0 && This
->baseShader
.reg_maps
.integer_constants
)
763 shader_addline(buffer
, "uniform ivec4 %cI[%u];\n", prefix
, This
->baseShader
.limits
.constant_int
);
765 if (This
->baseShader
.limits
.constant_bool
> 0 && This
->baseShader
.reg_maps
.boolean_constants
)
766 shader_addline(buffer
, "uniform bool %cB[%u];\n", prefix
, This
->baseShader
.limits
.constant_bool
);
769 shader_addline(buffer
, "uniform vec4 posFixup;\n");
770 /* Predeclaration; This function is added at link time based on the pixel shader.
771 * VS 3.0 shaders have an array OUT[] the shader writes to, earlier versions don't have
772 * that. We know the input to the reorder function at vertex shader compile time, so
773 * we can deal with that. The reorder function for a 1.x and 2.x vertex shader can just
774 * read gl_FrontColor. The output depends on the pixel shader. The reorder function for a
775 * 1.x and 2.x pshader or for fixed function will write gl_FrontColor, and for a 3.0 shader
776 * it will write to the varying array. Here we depend on the shader optimizer on sorting that
777 * out. The nvidia driver only does that if the parameter is inout instead of out, hence the
780 if (reg_maps
->shader_version
.major
>= 3)
782 shader_addline(buffer
, "void order_ps_input(in vec4[%u]);\n", MAX_REG_OUTPUT
);
784 shader_addline(buffer
, "void order_ps_input();\n");
787 IWineD3DPixelShaderImpl
*ps_impl
= (IWineD3DPixelShaderImpl
*) This
;
789 ps_impl
->numbumpenvmatconsts
= 0;
790 for(i
= 0; i
< (sizeof(reg_maps
->bumpmat
) / sizeof(reg_maps
->bumpmat
[0])); i
++) {
791 if(!reg_maps
->bumpmat
[i
]) {
795 ps_impl
->bumpenvmatconst
[(int) ps_impl
->numbumpenvmatconsts
].texunit
= i
;
796 shader_addline(buffer
, "uniform mat2 bumpenvmat%d;\n", i
);
798 if(reg_maps
->luminanceparams
) {
799 ps_impl
->luminanceconst
[(int) ps_impl
->numbumpenvmatconsts
].texunit
= i
;
800 shader_addline(buffer
, "uniform float luminancescale%d;\n", i
);
801 shader_addline(buffer
, "uniform float luminanceoffset%d;\n", i
);
802 extra_constants_needed
++;
804 ps_impl
->luminanceconst
[(int) ps_impl
->numbumpenvmatconsts
].texunit
= -1;
807 extra_constants_needed
++;
808 ps_impl
->numbumpenvmatconsts
++;
811 if(ps_args
->srgb_correction
) {
812 shader_addline(buffer
, "const vec4 srgb_mul_low = vec4(%f, %f, %f, %f);\n",
813 srgb_mul_low
, srgb_mul_low
, srgb_mul_low
, srgb_mul_low
);
814 shader_addline(buffer
, "const vec4 srgb_comparison = vec4(%f, %f, %f, %f);\n",
815 srgb_cmp
, srgb_cmp
, srgb_cmp
, srgb_cmp
);
817 if(reg_maps
->vpos
|| reg_maps
->usesdsy
) {
818 if(This
->baseShader
.limits
.constant_float
+ extra_constants_needed
+ 1 < GL_LIMITS(pshader_constantsF
)) {
819 shader_addline(buffer
, "uniform vec4 ycorrection;\n");
820 ((IWineD3DPixelShaderImpl
*) This
)->vpos_uniform
= 1;
821 extra_constants_needed
++;
823 /* This happens because we do not have proper tracking of the constant registers that are
824 * actually used, only the max limit of the shader version
826 FIXME("Cannot find a free uniform for vpos correction params\n");
827 shader_addline(buffer
, "const vec4 ycorrection = vec4(%f, %f, 0.0, 0.0);\n",
828 device
->render_offscreen
? 0.0 : ((IWineD3DSurfaceImpl
*) device
->render_targets
[0])->currentDesc
.Height
,
829 device
->render_offscreen
? 1.0 : -1.0);
831 shader_addline(buffer
, "vec4 vpos;\n");
835 /* Declare texture samplers */
836 for (i
= 0; i
< This
->baseShader
.limits
.sampler
; i
++) {
837 if (reg_maps
->sampler_type
[i
])
839 switch (reg_maps
->sampler_type
[i
])
842 shader_addline(buffer
, "uniform sampler1D %csampler%u;\n", prefix
, i
);
845 if(device
->stateBlock
->textures
[i
] &&
846 IWineD3DBaseTexture_GetTextureDimensions(device
->stateBlock
->textures
[i
]) == GL_TEXTURE_RECTANGLE_ARB
) {
847 shader_addline(buffer
, "uniform sampler2DRect %csampler%u;\n", prefix
, i
);
849 shader_addline(buffer
, "uniform sampler2D %csampler%u;\n", prefix
, i
);
852 if (pshader
&& ps_args
->np2_fixup
& (1 << i
))
854 /* NP2/RECT textures in OpenGL use texcoords in the range [0,width]x[0,height]
855 * while D3D has them in the (normalized) [0,1]x[0,1] range.
856 * samplerNP2Fixup stores texture dimensions and is updated through
857 * shader_glsl_load_np2fixup_constants when the sampler changes. */
858 shader_addline(buffer
, "uniform vec2 %csamplerNP2Fixup%u;\n", prefix
, i
);
861 case WINED3DSTT_CUBE
:
862 shader_addline(buffer
, "uniform samplerCube %csampler%u;\n", prefix
, i
);
864 case WINED3DSTT_VOLUME
:
865 shader_addline(buffer
, "uniform sampler3D %csampler%u;\n", prefix
, i
);
868 shader_addline(buffer
, "uniform unsupported_sampler %csampler%u;\n", prefix
, i
);
869 FIXME("Unrecognized sampler type: %#x\n", reg_maps
->sampler_type
[i
]);
875 /* Declare address variables */
876 for (i
= 0; i
< This
->baseShader
.limits
.address
; i
++) {
877 if (reg_maps
->address
[i
])
878 shader_addline(buffer
, "ivec4 A%d;\n", i
);
881 /* Declare texture coordinate temporaries and initialize them */
882 for (i
= 0; i
< This
->baseShader
.limits
.texcoord
; i
++) {
883 if (reg_maps
->texcoord
[i
])
884 shader_addline(buffer
, "vec4 T%u = gl_TexCoord[%u];\n", i
, i
);
887 /* Declare input register varyings. Only pixel shader, vertex shaders have that declared in the
888 * helper function shader that is linked in at link time
890 if (pshader
&& reg_maps
->shader_version
.major
>= 3)
892 if (use_vs(device
->stateBlock
))
894 shader_addline(buffer
, "varying vec4 IN[%u];\n", GL_LIMITS(glsl_varyings
) / 4);
896 /* TODO: Write a replacement shader for the fixed function vertex pipeline, so this isn't needed.
897 * For fixed function vertex processing + 3.0 pixel shader we need a separate function in the
898 * pixel shader that reads the fixed function color into the packed input registers.
900 shader_addline(buffer
, "vec4 IN[%u];\n", GL_LIMITS(glsl_varyings
) / 4);
904 /* Declare output register temporaries */
905 if(This
->baseShader
.limits
.packed_output
) {
906 shader_addline(buffer
, "vec4 OUT[%u];\n", This
->baseShader
.limits
.packed_output
);
909 /* Declare temporary variables */
910 for(i
= 0; i
< This
->baseShader
.limits
.temporary
; i
++) {
911 if (reg_maps
->temporary
[i
])
912 shader_addline(buffer
, "vec4 R%u;\n", i
);
915 /* Declare attributes */
916 for (i
= 0; i
< This
->baseShader
.limits
.attributes
; i
++) {
917 if (reg_maps
->attributes
[i
])
918 shader_addline(buffer
, "attribute vec4 attrib%i;\n", i
);
921 /* Declare loop registers aLx */
922 for (i
= 0; i
< reg_maps
->loop_depth
; i
++) {
923 shader_addline(buffer
, "int aL%u;\n", i
);
924 shader_addline(buffer
, "int tmpInt%u;\n", i
);
927 /* Temporary variables for matrix operations */
928 shader_addline(buffer
, "vec4 tmp0;\n");
929 shader_addline(buffer
, "vec4 tmp1;\n");
931 /* Local constants use a different name so they can be loaded once at shader link time
932 * They can't be hardcoded into the shader text via LC = {x, y, z, w}; because the
933 * float -> string conversion can cause precision loss.
935 if(!This
->baseShader
.load_local_constsF
) {
936 LIST_FOR_EACH_ENTRY(lconst
, &This
->baseShader
.constantsF
, local_constant
, entry
) {
937 shader_addline(buffer
, "uniform vec4 %cLC%u;\n", prefix
, lconst
->idx
);
941 /* Start the main program */
942 shader_addline(buffer
, "void main() {\n");
943 if(pshader
&& reg_maps
->vpos
) {
944 /* DirectX apps expect integer values, while OpenGL drivers add approximately 0.5. This causes
945 * off-by-one problems as spotted by the vPos d3d9 visual test. Unfortunately the ATI cards do
946 * not add exactly 0.5, but rather something like 0.49999999 or 0.50000001, which still causes
947 * precision troubles when we just substract 0.5.
949 * To deal with that just floor() the position. This will eliminate the fraction on all cards.
951 * TODO: Test how that behaves with multisampling once we can enable multisampling in winex11.
953 * An advantage of floor is that it works even if the driver doesn't add 1/2. It is somewhat
954 * questionable if 1.5, 2.5, ... are the proper values to return in gl_FragCoord, even though
955 * coordinates specify the pixel centers instead of the pixel corners. This code will behave
956 * correctly on drivers that returns integer values.
958 shader_addline(buffer
, "vpos = floor(vec4(0, ycorrection[0], 0, 0) + gl_FragCoord * vec4(1, ycorrection[1], 1, 1));\n");
962 /*****************************************************************************
963 * Functions to generate GLSL strings from DirectX Shader bytecode begin here.
965 * For more information, see http://wiki.winehq.org/DirectX-Shaders
966 ****************************************************************************/
969 static void shader_glsl_add_src_param(const struct wined3d_shader_instruction
*ins
,
970 const struct wined3d_shader_src_param
*wined3d_src
, DWORD mask
, glsl_src_param_t
*glsl_src
);
972 /** Used for opcode modifiers - They multiply the result by the specified amount */
973 static const char * const shift_glsl_tab
[] = {
975 "2.0 * ", /* 1 (x2) */
976 "4.0 * ", /* 2 (x4) */
977 "8.0 * ", /* 3 (x8) */
978 "16.0 * ", /* 4 (x16) */
979 "32.0 * ", /* 5 (x32) */
986 "0.0625 * ", /* 12 (d16) */
987 "0.125 * ", /* 13 (d8) */
988 "0.25 * ", /* 14 (d4) */
989 "0.5 * " /* 15 (d2) */
992 /* Generate a GLSL parameter that does the input modifier computation and return the input register/mask to use */
993 static void shader_glsl_gen_modifier(DWORD src_modifier
, const char *in_reg
, const char *in_regswizzle
, char *out_str
)
997 switch (src_modifier
)
999 case WINED3DSPSM_DZ
: /* Need to handle this in the instructions itself (texld & texcrd). */
1000 case WINED3DSPSM_DW
:
1001 case WINED3DSPSM_NONE
:
1002 sprintf(out_str
, "%s%s", in_reg
, in_regswizzle
);
1004 case WINED3DSPSM_NEG
:
1005 sprintf(out_str
, "-%s%s", in_reg
, in_regswizzle
);
1007 case WINED3DSPSM_NOT
:
1008 sprintf(out_str
, "!%s%s", in_reg
, in_regswizzle
);
1010 case WINED3DSPSM_BIAS
:
1011 sprintf(out_str
, "(%s%s - vec4(0.5)%s)", in_reg
, in_regswizzle
, in_regswizzle
);
1013 case WINED3DSPSM_BIASNEG
:
1014 sprintf(out_str
, "-(%s%s - vec4(0.5)%s)", in_reg
, in_regswizzle
, in_regswizzle
);
1016 case WINED3DSPSM_SIGN
:
1017 sprintf(out_str
, "(2.0 * (%s%s - 0.5))", in_reg
, in_regswizzle
);
1019 case WINED3DSPSM_SIGNNEG
:
1020 sprintf(out_str
, "-(2.0 * (%s%s - 0.5))", in_reg
, in_regswizzle
);
1022 case WINED3DSPSM_COMP
:
1023 sprintf(out_str
, "(1.0 - %s%s)", in_reg
, in_regswizzle
);
1025 case WINED3DSPSM_X2
:
1026 sprintf(out_str
, "(2.0 * %s%s)", in_reg
, in_regswizzle
);
1028 case WINED3DSPSM_X2NEG
:
1029 sprintf(out_str
, "-(2.0 * %s%s)", in_reg
, in_regswizzle
);
1031 case WINED3DSPSM_ABS
:
1032 sprintf(out_str
, "abs(%s%s)", in_reg
, in_regswizzle
);
1034 case WINED3DSPSM_ABSNEG
:
1035 sprintf(out_str
, "-abs(%s%s)", in_reg
, in_regswizzle
);
1038 FIXME("Unhandled modifier %u\n", src_modifier
);
1039 sprintf(out_str
, "%s%s", in_reg
, in_regswizzle
);
1043 /** Writes the GLSL variable name that corresponds to the register that the
1044 * DX opcode parameter is trying to access */
1045 static void shader_glsl_get_register_name(WINED3DSHADER_PARAM_REGISTER_TYPE register_type
, UINT register_idx
,
1046 const struct wined3d_shader_src_param
*rel_addr
, char *register_name
, BOOL
*is_color
,
1047 const struct wined3d_shader_instruction
*ins
)
1049 /* oPos, oFog and oPts in D3D */
1050 static const char * const hwrastout_reg_names
[] = { "gl_Position", "gl_FogFragCoord", "gl_PointSize" };
1052 IWineD3DBaseShaderImpl
*This
= (IWineD3DBaseShaderImpl
*)ins
->ctx
->shader
;
1053 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
1054 const WineD3D_GL_Info
* gl_info
= &deviceImpl
->adapter
->gl_info
;
1055 char pshader
= shader_is_pshader_version(This
->baseShader
.reg_maps
.shader_version
.type
);
1059 switch (register_type
)
1061 case WINED3DSPR_TEMP
:
1062 sprintf(register_name
, "R%u", register_idx
);
1064 case WINED3DSPR_INPUT
:
1066 /* Pixel shaders >= 3.0 */
1067 if (This
->baseShader
.reg_maps
.shader_version
.major
>= 3)
1069 DWORD in_count
= GL_LIMITS(glsl_varyings
) / 4;
1073 glsl_src_param_t rel_param
;
1074 shader_glsl_add_src_param(ins
, rel_addr
, WINED3DSP_WRITEMASK_0
, &rel_param
);
1076 /* Removing a + 0 would be an obvious optimization, but macos doesn't see the NOP
1079 if (((IWineD3DPixelShaderImpl
*)This
)->input_reg_map
[register_idx
])
1081 if (((IWineD3DPixelShaderImpl
*)This
)->declared_in_count
> in_count
) {
1082 sprintf(register_name
, "((%s + %u) > %d ? (%s + %u) > %d ? gl_SecondaryColor : gl_Color : IN[%s + %u])",
1083 rel_param
.param_str
, ((IWineD3DPixelShaderImpl
*)This
)->input_reg_map
[register_idx
], in_count
- 1,
1084 rel_param
.param_str
, ((IWineD3DPixelShaderImpl
*)This
)->input_reg_map
[register_idx
], in_count
,
1085 rel_param
.param_str
, ((IWineD3DPixelShaderImpl
*)This
)->input_reg_map
[register_idx
]);
1087 sprintf(register_name
, "IN[%s + %u]", rel_param
.param_str
,
1088 ((IWineD3DPixelShaderImpl
*)This
)->input_reg_map
[register_idx
]);
1091 if (((IWineD3DPixelShaderImpl
*)This
)->declared_in_count
> in_count
) {
1092 sprintf(register_name
, "((%s) > %d ? (%s) > %d ? gl_SecondaryColor : gl_Color : IN[%s])",
1093 rel_param
.param_str
, in_count
- 1,
1094 rel_param
.param_str
, in_count
,
1095 rel_param
.param_str
);
1097 sprintf(register_name
, "IN[%s]", rel_param
.param_str
);
1101 DWORD idx
= ((IWineD3DPixelShaderImpl
*) This
)->input_reg_map
[register_idx
];
1102 if (idx
== in_count
) {
1103 sprintf(register_name
, "gl_Color");
1104 } else if (idx
== in_count
+ 1) {
1105 sprintf(register_name
, "gl_SecondaryColor");
1107 sprintf(register_name
, "IN[%u]", idx
);
1111 if (register_idx
== 0)
1112 strcpy(register_name
, "gl_Color");
1114 strcpy(register_name
, "gl_SecondaryColor");
1117 if (((IWineD3DVertexShaderImpl
*)This
)->cur_args
->swizzle_map
& (1 << register_idx
)) *is_color
= TRUE
;
1118 sprintf(register_name
, "attrib%u", register_idx
);
1121 case WINED3DSPR_CONST
:
1123 const char prefix
= pshader
? 'P':'V';
1125 /* Relative addressing */
1128 glsl_src_param_t rel_param
;
1129 shader_glsl_add_src_param(ins
, rel_addr
, WINED3DSP_WRITEMASK_0
, &rel_param
);
1131 sprintf(register_name
, "%cC[%s + %u]", prefix
, rel_param
.param_str
, register_idx
);
1133 sprintf(register_name
, "%cC[%s]", prefix
, rel_param
.param_str
);
1135 if (shader_constant_is_local(This
, register_idx
))
1137 sprintf(register_name
, "%cLC%u", prefix
, register_idx
);
1139 sprintf(register_name
, "%cC[%u]", prefix
, register_idx
);
1145 case WINED3DSPR_CONSTINT
:
1147 sprintf(register_name
, "PI[%u]", register_idx
);
1149 sprintf(register_name
, "VI[%u]", register_idx
);
1151 case WINED3DSPR_CONSTBOOL
:
1153 sprintf(register_name
, "PB[%u]", register_idx
);
1155 sprintf(register_name
, "VB[%u]", register_idx
);
1157 case WINED3DSPR_TEXTURE
: /* case WINED3DSPR_ADDR: */
1159 sprintf(register_name
, "T%u", register_idx
);
1161 sprintf(register_name
, "A%u", register_idx
);
1164 case WINED3DSPR_LOOP
:
1165 sprintf(register_name
, "aL%u", This
->baseShader
.cur_loop_regno
- 1);
1167 case WINED3DSPR_SAMPLER
:
1169 sprintf(register_name
, "Psampler%u", register_idx
);
1171 sprintf(register_name
, "Vsampler%u", register_idx
);
1173 case WINED3DSPR_COLOROUT
:
1174 if (register_idx
>= GL_LIMITS(buffers
))
1175 WARN("Write to render target %u, only %d supported\n", register_idx
, 4);
1177 if (GL_SUPPORT(ARB_DRAW_BUFFERS
)) {
1178 sprintf(register_name
, "gl_FragData[%u]", register_idx
);
1179 } else { /* On older cards with GLSL support like the GeforceFX there's only one buffer. */
1180 sprintf(register_name
, "gl_FragColor");
1183 case WINED3DSPR_RASTOUT
:
1184 sprintf(register_name
, "%s", hwrastout_reg_names
[register_idx
]);
1186 case WINED3DSPR_DEPTHOUT
:
1187 sprintf(register_name
, "gl_FragDepth");
1189 case WINED3DSPR_ATTROUT
:
1190 if (register_idx
== 0)
1192 sprintf(register_name
, "gl_FrontColor");
1194 sprintf(register_name
, "gl_FrontSecondaryColor");
1197 case WINED3DSPR_TEXCRDOUT
:
1198 /* Vertex shaders >= 3.0: WINED3DSPR_OUTPUT */
1199 if (This
->baseShader
.reg_maps
.shader_version
.major
>= 3) sprintf(register_name
, "OUT[%u]", register_idx
);
1200 else sprintf(register_name
, "gl_TexCoord[%u]", register_idx
);
1202 case WINED3DSPR_MISCTYPE
:
1203 if (register_idx
== 0)
1206 sprintf(register_name
, "vpos");
1208 else if (register_idx
== 1)
1210 /* Note that gl_FrontFacing is a bool, while vFace is
1211 * a float for which the sign determines front/back
1213 sprintf(register_name
, "(gl_FrontFacing ? 1.0 : -1.0)");
1215 FIXME("Unhandled misctype register %d\n", register_idx
);
1216 sprintf(register_name
, "unrecognized_register");
1220 FIXME("Unhandled register name Type(%d)\n", register_type
);
1221 sprintf(register_name
, "unrecognized_register");
1226 static void shader_glsl_write_mask_to_str(DWORD write_mask
, char *str
)
1229 if (write_mask
& WINED3DSP_WRITEMASK_0
) *str
++ = 'x';
1230 if (write_mask
& WINED3DSP_WRITEMASK_1
) *str
++ = 'y';
1231 if (write_mask
& WINED3DSP_WRITEMASK_2
) *str
++ = 'z';
1232 if (write_mask
& WINED3DSP_WRITEMASK_3
) *str
++ = 'w';
1236 /* Get the GLSL write mask for the destination register */
1237 static DWORD
shader_glsl_get_write_mask(const struct wined3d_shader_dst_param
*param
, char *write_mask
)
1239 DWORD mask
= param
->write_mask
;
1241 if (shader_is_scalar(param
->register_type
, param
->register_idx
))
1243 mask
= WINED3DSP_WRITEMASK_0
;
1248 shader_glsl_write_mask_to_str(mask
, write_mask
);
1254 static unsigned int shader_glsl_get_write_mask_size(DWORD write_mask
) {
1255 unsigned int size
= 0;
1257 if (write_mask
& WINED3DSP_WRITEMASK_0
) ++size
;
1258 if (write_mask
& WINED3DSP_WRITEMASK_1
) ++size
;
1259 if (write_mask
& WINED3DSP_WRITEMASK_2
) ++size
;
1260 if (write_mask
& WINED3DSP_WRITEMASK_3
) ++size
;
1265 static void shader_glsl_swizzle_to_str(const DWORD swizzle
, BOOL fixup
, DWORD mask
, char *str
)
1267 /* For registers of type WINED3DDECLTYPE_D3DCOLOR, data is stored as "bgra",
1268 * but addressed as "rgba". To fix this we need to swap the register's x
1269 * and z components. */
1270 const char *swizzle_chars
= fixup
? "zyxw" : "xyzw";
1273 /* swizzle bits fields: wwzzyyxx */
1274 if (mask
& WINED3DSP_WRITEMASK_0
) *str
++ = swizzle_chars
[swizzle
& 0x03];
1275 if (mask
& WINED3DSP_WRITEMASK_1
) *str
++ = swizzle_chars
[(swizzle
>> 2) & 0x03];
1276 if (mask
& WINED3DSP_WRITEMASK_2
) *str
++ = swizzle_chars
[(swizzle
>> 4) & 0x03];
1277 if (mask
& WINED3DSP_WRITEMASK_3
) *str
++ = swizzle_chars
[(swizzle
>> 6) & 0x03];
1281 static void shader_glsl_get_swizzle(const struct wined3d_shader_src_param
*param
,
1282 BOOL fixup
, DWORD mask
, char *swizzle_str
)
1284 if (shader_is_scalar(param
->register_type
, param
->register_idx
))
1285 *swizzle_str
= '\0';
1287 shader_glsl_swizzle_to_str(param
->swizzle
, fixup
, mask
, swizzle_str
);
1290 /* From a given parameter token, generate the corresponding GLSL string.
1291 * Also, return the actual register name and swizzle in case the
1292 * caller needs this information as well. */
1293 static void shader_glsl_add_src_param(const struct wined3d_shader_instruction
*ins
,
1294 const struct wined3d_shader_src_param
*wined3d_src
, DWORD mask
, glsl_src_param_t
*glsl_src
)
1296 BOOL is_color
= FALSE
;
1297 char swizzle_str
[6];
1299 glsl_src
->reg_name
[0] = '\0';
1300 glsl_src
->param_str
[0] = '\0';
1301 swizzle_str
[0] = '\0';
1303 shader_glsl_get_register_name(wined3d_src
->register_type
, wined3d_src
->register_idx
,
1304 wined3d_src
->rel_addr
, glsl_src
->reg_name
, &is_color
, ins
);
1306 shader_glsl_get_swizzle(wined3d_src
, is_color
, mask
, swizzle_str
);
1307 shader_glsl_gen_modifier(wined3d_src
->modifiers
, glsl_src
->reg_name
, swizzle_str
, glsl_src
->param_str
);
1310 /* From a given parameter token, generate the corresponding GLSL string.
1311 * Also, return the actual register name and swizzle in case the
1312 * caller needs this information as well. */
1313 static DWORD
shader_glsl_add_dst_param(const struct wined3d_shader_instruction
*ins
,
1314 const struct wined3d_shader_dst_param
*wined3d_dst
, glsl_dst_param_t
*glsl_dst
)
1316 BOOL is_color
= FALSE
;
1318 glsl_dst
->mask_str
[0] = '\0';
1319 glsl_dst
->reg_name
[0] = '\0';
1321 shader_glsl_get_register_name(wined3d_dst
->register_type
, wined3d_dst
->register_idx
,
1322 wined3d_dst
->rel_addr
, glsl_dst
->reg_name
, &is_color
, ins
);
1323 return shader_glsl_get_write_mask(wined3d_dst
, glsl_dst
->mask_str
);
1326 /* Append the destination part of the instruction to the buffer, return the effective write mask */
1327 static DWORD
shader_glsl_append_dst_ext(SHADER_BUFFER
*buffer
,
1328 const struct wined3d_shader_instruction
*ins
, const struct wined3d_shader_dst_param
*dst
)
1330 glsl_dst_param_t glsl_dst
;
1333 mask
= shader_glsl_add_dst_param(ins
, dst
, &glsl_dst
);
1334 if (mask
) shader_addline(buffer
, "%s%s = %s(", glsl_dst
.reg_name
, glsl_dst
.mask_str
, shift_glsl_tab
[dst
->shift
]);
1339 /* Append the destination part of the instruction to the buffer, return the effective write mask */
1340 static DWORD
shader_glsl_append_dst(SHADER_BUFFER
*buffer
, const struct wined3d_shader_instruction
*ins
)
1342 return shader_glsl_append_dst_ext(buffer
, ins
, &ins
->dst
[0]);
1345 /** Process GLSL instruction modifiers */
1346 void shader_glsl_add_instruction_modifiers(const struct wined3d_shader_instruction
*ins
)
1348 glsl_dst_param_t dst_param
;
1351 if (!ins
->dst_count
) return;
1353 modifiers
= ins
->dst
[0].modifiers
;
1354 if (!modifiers
) return;
1356 shader_glsl_add_dst_param(ins
, &ins
->dst
[0], &dst_param
);
1358 if (modifiers
& WINED3DSPDM_SATURATE
)
1360 /* _SAT means to clamp the value of the register to between 0 and 1 */
1361 shader_addline(ins
->ctx
->buffer
, "%s%s = clamp(%s%s, 0.0, 1.0);\n", dst_param
.reg_name
,
1362 dst_param
.mask_str
, dst_param
.reg_name
, dst_param
.mask_str
);
1365 if (modifiers
& WINED3DSPDM_MSAMPCENTROID
)
1367 FIXME("_centroid modifier not handled\n");
1370 if (modifiers
& WINED3DSPDM_PARTIALPRECISION
)
1372 /* MSDN says this modifier can be safely ignored, so that's what we'll do. */
1376 static inline const char *shader_get_comp_op(DWORD op
)
1379 case COMPARISON_GT
: return ">";
1380 case COMPARISON_EQ
: return "==";
1381 case COMPARISON_GE
: return ">=";
1382 case COMPARISON_LT
: return "<";
1383 case COMPARISON_NE
: return "!=";
1384 case COMPARISON_LE
: return "<=";
1386 FIXME("Unrecognized comparison value: %u\n", op
);
1391 static void shader_glsl_get_sample_function(DWORD sampler_type
, DWORD flags
, glsl_sample_function_t
*sample_function
)
1393 BOOL projected
= flags
& WINED3D_GLSL_SAMPLE_PROJECTED
;
1394 BOOL texrect
= flags
& WINED3D_GLSL_SAMPLE_RECT
;
1395 BOOL lod
= flags
& WINED3D_GLSL_SAMPLE_LOD
;
1396 BOOL grad
= flags
& WINED3D_GLSL_SAMPLE_GRAD
;
1398 /* Note that there's no such thing as a projected cube texture. */
1399 switch(sampler_type
) {
1402 sample_function
->name
= projected
? "texture1DProjLod" : "texture1DLod";
1404 sample_function
->name
= projected
? "texture1DProjGradARB" : "texture1DGradARB";
1406 sample_function
->name
= projected
? "texture1DProj" : "texture1D";
1408 sample_function
->coord_mask
= WINED3DSP_WRITEMASK_0
;
1413 sample_function
->name
= projected
? "texture2DRectProjLod" : "texture2DRectLod";
1415 /* What good are texrect grad functions? I don't know, but GL_EXT_gpu_shader4 defines them.
1416 * There is no GL_ARB_shader_texture_lod spec yet, so I don't know if they're defined there
1418 sample_function
->name
= projected
? "shadow2DRectProjGradARB" : "shadow2DRectGradARB";
1420 sample_function
->name
= projected
? "texture2DRectProj" : "texture2DRect";
1424 sample_function
->name
= projected
? "texture2DProjLod" : "texture2DLod";
1426 sample_function
->name
= projected
? "texture2DProjGradARB" : "texture2DGradARB";
1428 sample_function
->name
= projected
? "texture2DProj" : "texture2D";
1431 sample_function
->coord_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
;
1433 case WINED3DSTT_CUBE
:
1435 sample_function
->name
= "textureCubeLod";
1437 sample_function
->name
= "textureCubeGradARB";
1439 sample_function
->name
= "textureCube";
1441 sample_function
->coord_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
1443 case WINED3DSTT_VOLUME
:
1445 sample_function
->name
= projected
? "texture3DProjLod" : "texture3DLod";
1447 sample_function
->name
= projected
? "texture3DProjGradARB" : "texture3DGradARB";
1449 sample_function
->name
= projected
? "texture3DProj" : "texture3D";
1451 sample_function
->coord_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
1454 sample_function
->name
= "";
1455 sample_function
->coord_mask
= 0;
1456 FIXME("Unrecognized sampler type: %#x;\n", sampler_type
);
1461 static void shader_glsl_append_fixup_arg(char *arguments
, const char *reg_name
,
1462 BOOL sign_fixup
, enum fixup_channel_source channel_source
)
1464 switch(channel_source
)
1466 case CHANNEL_SOURCE_ZERO
:
1467 strcat(arguments
, "0.0");
1470 case CHANNEL_SOURCE_ONE
:
1471 strcat(arguments
, "1.0");
1474 case CHANNEL_SOURCE_X
:
1475 strcat(arguments
, reg_name
);
1476 strcat(arguments
, ".x");
1479 case CHANNEL_SOURCE_Y
:
1480 strcat(arguments
, reg_name
);
1481 strcat(arguments
, ".y");
1484 case CHANNEL_SOURCE_Z
:
1485 strcat(arguments
, reg_name
);
1486 strcat(arguments
, ".z");
1489 case CHANNEL_SOURCE_W
:
1490 strcat(arguments
, reg_name
);
1491 strcat(arguments
, ".w");
1495 FIXME("Unhandled channel source %#x\n", channel_source
);
1496 strcat(arguments
, "undefined");
1500 if (sign_fixup
) strcat(arguments
, " * 2.0 - 1.0");
1503 static void shader_glsl_color_correction(const struct wined3d_shader_instruction
*ins
, struct color_fixup_desc fixup
)
1505 struct wined3d_shader_dst_param dst
;
1506 unsigned int mask_size
, remaining
;
1507 glsl_dst_param_t dst_param
;
1508 char arguments
[256];
1512 if (fixup
.x_sign_fixup
|| fixup
.x_source
!= CHANNEL_SOURCE_X
) mask
|= WINED3DSP_WRITEMASK_0
;
1513 if (fixup
.y_sign_fixup
|| fixup
.y_source
!= CHANNEL_SOURCE_Y
) mask
|= WINED3DSP_WRITEMASK_1
;
1514 if (fixup
.z_sign_fixup
|| fixup
.z_source
!= CHANNEL_SOURCE_Z
) mask
|= WINED3DSP_WRITEMASK_2
;
1515 if (fixup
.w_sign_fixup
|| fixup
.w_source
!= CHANNEL_SOURCE_W
) mask
|= WINED3DSP_WRITEMASK_3
;
1516 mask
&= ins
->dst
[0].write_mask
;
1518 if (!mask
) return; /* Nothing to do */
1520 if (is_yuv_fixup(fixup
))
1522 enum yuv_fixup yuv_fixup
= get_yuv_fixup(fixup
);
1523 FIXME("YUV fixup (%#x) not supported\n", yuv_fixup
);
1527 mask_size
= shader_glsl_get_write_mask_size(mask
);
1530 dst
.write_mask
= mask
;
1531 shader_glsl_add_dst_param(ins
, &dst
, &dst_param
);
1533 arguments
[0] = '\0';
1534 remaining
= mask_size
;
1535 if (mask
& WINED3DSP_WRITEMASK_0
)
1537 shader_glsl_append_fixup_arg(arguments
, dst_param
.reg_name
, fixup
.x_sign_fixup
, fixup
.x_source
);
1538 if (--remaining
) strcat(arguments
, ", ");
1540 if (mask
& WINED3DSP_WRITEMASK_1
)
1542 shader_glsl_append_fixup_arg(arguments
, dst_param
.reg_name
, fixup
.y_sign_fixup
, fixup
.y_source
);
1543 if (--remaining
) strcat(arguments
, ", ");
1545 if (mask
& WINED3DSP_WRITEMASK_2
)
1547 shader_glsl_append_fixup_arg(arguments
, dst_param
.reg_name
, fixup
.z_sign_fixup
, fixup
.z_source
);
1548 if (--remaining
) strcat(arguments
, ", ");
1550 if (mask
& WINED3DSP_WRITEMASK_3
)
1552 shader_glsl_append_fixup_arg(arguments
, dst_param
.reg_name
, fixup
.w_sign_fixup
, fixup
.w_source
);
1553 if (--remaining
) strcat(arguments
, ", ");
1558 shader_addline(ins
->ctx
->buffer
, "%s%s = vec%u(%s);\n",
1559 dst_param
.reg_name
, dst_param
.mask_str
, mask_size
, arguments
);
1563 shader_addline(ins
->ctx
->buffer
, "%s%s = %s;\n", dst_param
.reg_name
, dst_param
.mask_str
, arguments
);
1567 static void PRINTF_ATTR(8, 9) shader_glsl_gen_sample_code(const struct wined3d_shader_instruction
*ins
,
1568 DWORD sampler
, const glsl_sample_function_t
*sample_function
, DWORD swizzle
,
1569 const char *dx
, const char *dy
,
1570 const char *bias
, const char *coord_reg_fmt
, ...)
1572 const char *sampler_base
;
1573 char dst_swizzle
[6];
1574 struct color_fixup_desc fixup
;
1575 BOOL np2_fixup
= FALSE
;
1578 shader_glsl_swizzle_to_str(swizzle
, FALSE
, ins
->dst
[0].write_mask
, dst_swizzle
);
1580 if (shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
.type
))
1582 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*)ins
->ctx
->shader
;
1583 fixup
= This
->cur_args
->color_fixup
[sampler
];
1584 sampler_base
= "Psampler";
1586 if(This
->cur_args
->np2_fixup
& (1 << sampler
)) {
1588 FIXME("Biased sampling from NP2 textures is unsupported\n");
1594 sampler_base
= "Vsampler";
1595 fixup
= COLOR_FIXUP_IDENTITY
; /* FIXME: Vshader color fixup */
1598 shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
1600 shader_addline(ins
->ctx
->buffer
, "%s(%s%u, ", sample_function
->name
, sampler_base
, sampler
);
1602 va_start(args
, coord_reg_fmt
);
1603 shader_vaddline(ins
->ctx
->buffer
, coord_reg_fmt
, args
);
1607 shader_addline(ins
->ctx
->buffer
, ", %s)%s);\n", bias
, dst_swizzle
);
1610 shader_addline(ins
->ctx
->buffer
, " * PsamplerNP2Fixup%u)%s);\n", sampler
, dst_swizzle
);
1611 } else if(dx
&& dy
) {
1612 shader_addline(ins
->ctx
->buffer
, ", %s, %s)%s);\n", dx
, dy
, dst_swizzle
);
1614 shader_addline(ins
->ctx
->buffer
, ")%s);\n", dst_swizzle
);
1618 if(!is_identity_fixup(fixup
)) {
1619 shader_glsl_color_correction(ins
, fixup
);
1623 /*****************************************************************************
1625 * Begin processing individual instruction opcodes
1627 ****************************************************************************/
1629 /* Generate GLSL arithmetic functions (dst = src1 + src2) */
1630 static void shader_glsl_arith(const struct wined3d_shader_instruction
*ins
)
1632 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
1633 glsl_src_param_t src0_param
;
1634 glsl_src_param_t src1_param
;
1638 /* Determine the GLSL operator to use based on the opcode */
1639 switch (ins
->handler_idx
)
1641 case WINED3DSIH_MUL
: op
= '*'; break;
1642 case WINED3DSIH_ADD
: op
= '+'; break;
1643 case WINED3DSIH_SUB
: op
= '-'; break;
1646 FIXME("Opcode %#x not yet handled in GLSL\n", ins
->handler_idx
);
1650 write_mask
= shader_glsl_append_dst(buffer
, ins
);
1651 shader_glsl_add_src_param(ins
, &ins
->src
[0], write_mask
, &src0_param
);
1652 shader_glsl_add_src_param(ins
, &ins
->src
[1], write_mask
, &src1_param
);
1653 shader_addline(buffer
, "%s %c %s);\n", src0_param
.param_str
, op
, src1_param
.param_str
);
1656 /* Process the WINED3DSIO_MOV opcode using GLSL (dst = src) */
1657 static void shader_glsl_mov(const struct wined3d_shader_instruction
*ins
)
1659 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
1660 glsl_src_param_t src0_param
;
1663 write_mask
= shader_glsl_append_dst(buffer
, ins
);
1664 shader_glsl_add_src_param(ins
, &ins
->src
[0], write_mask
, &src0_param
);
1666 /* In vs_1_1 WINED3DSIO_MOV can write to the address register. In later
1667 * shader versions WINED3DSIO_MOVA is used for this. */
1668 if (ins
->ctx
->reg_maps
->shader_version
.major
== 1
1669 && !shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
.type
)
1670 && ins
->dst
[0].register_type
== WINED3DSPR_ADDR
)
1672 /* This is a simple floor() */
1673 unsigned int mask_size
= shader_glsl_get_write_mask_size(write_mask
);
1674 if (mask_size
> 1) {
1675 shader_addline(buffer
, "ivec%d(floor(%s)));\n", mask_size
, src0_param
.param_str
);
1677 shader_addline(buffer
, "int(floor(%s)));\n", src0_param
.param_str
);
1680 else if(ins
->handler_idx
== WINED3DSIH_MOVA
)
1682 /* We need to *round* to the nearest int here. */
1683 unsigned int mask_size
= shader_glsl_get_write_mask_size(write_mask
);
1684 if (mask_size
> 1) {
1685 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
);
1687 shader_addline(buffer
, "int(floor(abs(%s) + 0.5) * sign(%s)));\n", src0_param
.param_str
, src0_param
.param_str
);
1690 shader_addline(buffer
, "%s);\n", src0_param
.param_str
);
1694 /* Process the dot product operators DP3 and DP4 in GLSL (dst = dot(src0, src1)) */
1695 static void shader_glsl_dot(const struct wined3d_shader_instruction
*ins
)
1697 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
1698 glsl_src_param_t src0_param
;
1699 glsl_src_param_t src1_param
;
1700 DWORD dst_write_mask
, src_write_mask
;
1701 unsigned int dst_size
= 0;
1703 dst_write_mask
= shader_glsl_append_dst(buffer
, ins
);
1704 dst_size
= shader_glsl_get_write_mask_size(dst_write_mask
);
1706 /* dp3 works on vec3, dp4 on vec4 */
1707 if (ins
->handler_idx
== WINED3DSIH_DP4
)
1709 src_write_mask
= WINED3DSP_WRITEMASK_ALL
;
1711 src_write_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
1714 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_write_mask
, &src0_param
);
1715 shader_glsl_add_src_param(ins
, &ins
->src
[1], src_write_mask
, &src1_param
);
1718 shader_addline(buffer
, "vec%d(dot(%s, %s)));\n", dst_size
, src0_param
.param_str
, src1_param
.param_str
);
1720 shader_addline(buffer
, "dot(%s, %s));\n", src0_param
.param_str
, src1_param
.param_str
);
1724 /* Note that this instruction has some restrictions. The destination write mask
1725 * can't contain the w component, and the source swizzles have to be .xyzw */
1726 static void shader_glsl_cross(const struct wined3d_shader_instruction
*ins
)
1728 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
1729 glsl_src_param_t src0_param
;
1730 glsl_src_param_t src1_param
;
1733 shader_glsl_get_write_mask(&ins
->dst
[0], dst_mask
);
1734 shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
1735 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
1736 shader_glsl_add_src_param(ins
, &ins
->src
[1], src_mask
, &src1_param
);
1737 shader_addline(ins
->ctx
->buffer
, "cross(%s, %s)%s);\n", src0_param
.param_str
, src1_param
.param_str
, dst_mask
);
1740 /* Process the WINED3DSIO_POW instruction in GLSL (dst = |src0|^src1)
1741 * Src0 and src1 are scalars. Note that D3D uses the absolute of src0, while
1742 * GLSL uses the value as-is. */
1743 static void shader_glsl_pow(const struct wined3d_shader_instruction
*ins
)
1745 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
1746 glsl_src_param_t src0_param
;
1747 glsl_src_param_t src1_param
;
1748 DWORD dst_write_mask
;
1749 unsigned int dst_size
;
1751 dst_write_mask
= shader_glsl_append_dst(buffer
, ins
);
1752 dst_size
= shader_glsl_get_write_mask_size(dst_write_mask
);
1754 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
1755 shader_glsl_add_src_param(ins
, &ins
->src
[1], WINED3DSP_WRITEMASK_0
, &src1_param
);
1758 shader_addline(buffer
, "vec%d(pow(abs(%s), %s)));\n", dst_size
, src0_param
.param_str
, src1_param
.param_str
);
1760 shader_addline(buffer
, "pow(abs(%s), %s));\n", src0_param
.param_str
, src1_param
.param_str
);
1764 /* Process the WINED3DSIO_LOG instruction in GLSL (dst = log2(|src0|))
1765 * Src0 is a scalar. Note that D3D uses the absolute of src0, while
1766 * GLSL uses the value as-is. */
1767 static void shader_glsl_log(const struct wined3d_shader_instruction
*ins
)
1769 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
1770 glsl_src_param_t src0_param
;
1771 DWORD dst_write_mask
;
1772 unsigned int dst_size
;
1774 dst_write_mask
= shader_glsl_append_dst(buffer
, ins
);
1775 dst_size
= shader_glsl_get_write_mask_size(dst_write_mask
);
1777 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
1780 shader_addline(buffer
, "vec%d(log2(abs(%s))));\n", dst_size
, src0_param
.param_str
);
1782 shader_addline(buffer
, "log2(abs(%s)));\n", src0_param
.param_str
);
1786 /* Map the opcode 1-to-1 to the GL code (arg->dst = instruction(src0, src1, ...) */
1787 static void shader_glsl_map2gl(const struct wined3d_shader_instruction
*ins
)
1789 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
1790 glsl_src_param_t src_param
;
1791 const char *instruction
;
1795 /* Determine the GLSL function to use based on the opcode */
1796 /* TODO: Possibly make this a table for faster lookups */
1797 switch (ins
->handler_idx
)
1799 case WINED3DSIH_MIN
: instruction
= "min"; break;
1800 case WINED3DSIH_MAX
: instruction
= "max"; break;
1801 case WINED3DSIH_ABS
: instruction
= "abs"; break;
1802 case WINED3DSIH_FRC
: instruction
= "fract"; break;
1803 case WINED3DSIH_NRM
: instruction
= "normalize"; break;
1804 case WINED3DSIH_EXP
: instruction
= "exp2"; break;
1805 case WINED3DSIH_SGN
: instruction
= "sign"; break;
1806 case WINED3DSIH_DSX
: instruction
= "dFdx"; break;
1807 case WINED3DSIH_DSY
: instruction
= "ycorrection.y * dFdy"; break;
1808 default: instruction
= "";
1809 FIXME("Opcode %#x not yet handled in GLSL\n", ins
->handler_idx
);
1813 write_mask
= shader_glsl_append_dst(buffer
, ins
);
1815 shader_addline(buffer
, "%s(", instruction
);
1819 shader_glsl_add_src_param(ins
, &ins
->src
[0], write_mask
, &src_param
);
1820 shader_addline(buffer
, "%s", src_param
.param_str
);
1821 for (i
= 1; i
< ins
->src_count
; ++i
)
1823 shader_glsl_add_src_param(ins
, &ins
->src
[i
], write_mask
, &src_param
);
1824 shader_addline(buffer
, ", %s", src_param
.param_str
);
1828 shader_addline(buffer
, "));\n");
1831 /** Process the WINED3DSIO_EXPP instruction in GLSL:
1832 * For shader model 1.x, do the following (and honor the writemask, so use a temporary variable):
1833 * dst.x = 2^(floor(src))
1834 * dst.y = src - floor(src)
1835 * dst.z = 2^src (partial precision is allowed, but optional)
1837 * For 2.0 shaders, just do this (honoring writemask and swizzle):
1838 * dst = 2^src; (partial precision is allowed, but optional)
1840 static void shader_glsl_expp(const struct wined3d_shader_instruction
*ins
)
1842 glsl_src_param_t src_param
;
1844 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
, &src_param
);
1846 if (ins
->ctx
->reg_maps
->shader_version
.major
< 2)
1850 shader_addline(ins
->ctx
->buffer
, "tmp0.x = exp2(floor(%s));\n", src_param
.param_str
);
1851 shader_addline(ins
->ctx
->buffer
, "tmp0.y = %s - floor(%s);\n", src_param
.param_str
, src_param
.param_str
);
1852 shader_addline(ins
->ctx
->buffer
, "tmp0.z = exp2(%s);\n", src_param
.param_str
);
1853 shader_addline(ins
->ctx
->buffer
, "tmp0.w = 1.0;\n");
1855 shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
1856 shader_glsl_get_write_mask(&ins
->dst
[0], dst_mask
);
1857 shader_addline(ins
->ctx
->buffer
, "tmp0%s);\n", dst_mask
);
1860 unsigned int mask_size
;
1862 write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
1863 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
1865 if (mask_size
> 1) {
1866 shader_addline(ins
->ctx
->buffer
, "vec%d(exp2(%s)));\n", mask_size
, src_param
.param_str
);
1868 shader_addline(ins
->ctx
->buffer
, "exp2(%s));\n", src_param
.param_str
);
1873 /** Process the RCP (reciprocal or inverse) opcode in GLSL (dst = 1 / src) */
1874 static void shader_glsl_rcp(const struct wined3d_shader_instruction
*ins
)
1876 glsl_src_param_t src_param
;
1878 unsigned int mask_size
;
1880 write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
1881 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
1882 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_3
, &src_param
);
1884 if (mask_size
> 1) {
1885 shader_addline(ins
->ctx
->buffer
, "vec%d(1.0 / %s));\n", mask_size
, src_param
.param_str
);
1887 shader_addline(ins
->ctx
->buffer
, "1.0 / %s);\n", src_param
.param_str
);
1891 static void shader_glsl_rsq(const struct wined3d_shader_instruction
*ins
)
1893 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
1894 glsl_src_param_t src_param
;
1896 unsigned int mask_size
;
1898 write_mask
= shader_glsl_append_dst(buffer
, ins
);
1899 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
1901 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_3
, &src_param
);
1903 if (mask_size
> 1) {
1904 shader_addline(buffer
, "vec%d(inversesqrt(%s)));\n", mask_size
, src_param
.param_str
);
1906 shader_addline(buffer
, "inversesqrt(%s));\n", src_param
.param_str
);
1910 /** Process signed comparison opcodes in GLSL. */
1911 static void shader_glsl_compare(const struct wined3d_shader_instruction
*ins
)
1913 glsl_src_param_t src0_param
;
1914 glsl_src_param_t src1_param
;
1916 unsigned int mask_size
;
1918 write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
1919 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
1920 shader_glsl_add_src_param(ins
, &ins
->src
[0], write_mask
, &src0_param
);
1921 shader_glsl_add_src_param(ins
, &ins
->src
[1], write_mask
, &src1_param
);
1923 if (mask_size
> 1) {
1924 const char *compare
;
1926 switch(ins
->handler_idx
)
1928 case WINED3DSIH_SLT
: compare
= "lessThan"; break;
1929 case WINED3DSIH_SGE
: compare
= "greaterThanEqual"; break;
1930 default: compare
= "";
1931 FIXME("Can't handle opcode %#x\n", ins
->handler_idx
);
1934 shader_addline(ins
->ctx
->buffer
, "vec%d(%s(%s, %s)));\n", mask_size
, compare
,
1935 src0_param
.param_str
, src1_param
.param_str
);
1937 switch(ins
->handler_idx
)
1939 case WINED3DSIH_SLT
:
1940 /* Step(src0, src1) is not suitable here because if src0 == src1 SLT is supposed,
1941 * to return 0.0 but step returns 1.0 because step is not < x
1942 * An alternative is a bvec compare padded with an unused second component.
1943 * step(src1 * -1.0, src0 * -1.0) is not an option because it suffers from the same
1944 * issue. Playing with not() is not possible either because not() does not accept
1947 shader_addline(ins
->ctx
->buffer
, "(%s < %s) ? 1.0 : 0.0);\n",
1948 src0_param
.param_str
, src1_param
.param_str
);
1950 case WINED3DSIH_SGE
:
1951 /* Here we can use the step() function and safe a conditional */
1952 shader_addline(ins
->ctx
->buffer
, "step(%s, %s));\n", src1_param
.param_str
, src0_param
.param_str
);
1955 FIXME("Can't handle opcode %#x\n", ins
->handler_idx
);
1961 /** Process CMP instruction in GLSL (dst = src0 >= 0.0 ? src1 : src2), per channel */
1962 static void shader_glsl_cmp(const struct wined3d_shader_instruction
*ins
)
1964 glsl_src_param_t src0_param
;
1965 glsl_src_param_t src1_param
;
1966 glsl_src_param_t src2_param
;
1967 DWORD write_mask
, cmp_channel
= 0;
1970 BOOL temp_destination
= FALSE
;
1972 if (shader_is_scalar(ins
->src
[0].register_type
, ins
->src
[0].register_idx
))
1974 write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
1976 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_ALL
, &src0_param
);
1977 shader_glsl_add_src_param(ins
, &ins
->src
[1], write_mask
, &src1_param
);
1978 shader_glsl_add_src_param(ins
, &ins
->src
[2], write_mask
, &src2_param
);
1980 shader_addline(ins
->ctx
->buffer
, "%s >= 0.0 ? %s : %s);\n",
1981 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
1983 DWORD dst_mask
= ins
->dst
[0].write_mask
;
1984 struct wined3d_shader_dst_param dst
= ins
->dst
[0];
1986 /* Cycle through all source0 channels */
1987 for (i
=0; i
<4; i
++) {
1989 /* Find the destination channels which use the current source0 channel */
1990 for (j
=0; j
<4; j
++) {
1991 if (((ins
->src
[0].swizzle
>> (2 * j
)) & 0x3) == i
)
1993 write_mask
|= WINED3DSP_WRITEMASK_0
<< j
;
1994 cmp_channel
= WINED3DSP_WRITEMASK_0
<< j
;
1997 dst
.write_mask
= dst_mask
& write_mask
;
1999 /* Splitting the cmp instruction up in multiple lines imposes a problem:
2000 * The first lines may overwrite source parameters of the following lines.
2001 * Deal with that by using a temporary destination register if needed
2003 if ((ins
->src
[0].register_idx
== ins
->dst
[0].register_idx
2004 && ins
->src
[0].register_type
== ins
->dst
[0].register_type
)
2005 || (ins
->src
[1].register_idx
== ins
->dst
[0].register_idx
2006 && ins
->src
[1].register_type
== ins
->dst
[0].register_type
)
2007 || (ins
->src
[2].register_idx
== ins
->dst
[0].register_idx
2008 && ins
->src
[2].register_type
== ins
->dst
[0].register_type
))
2010 write_mask
= shader_glsl_get_write_mask(&dst
, mask_char
);
2011 if (!write_mask
) continue;
2012 shader_addline(ins
->ctx
->buffer
, "tmp0%s = (", mask_char
);
2013 temp_destination
= TRUE
;
2015 write_mask
= shader_glsl_append_dst_ext(ins
->ctx
->buffer
, ins
, &dst
);
2016 if (!write_mask
) continue;
2019 shader_glsl_add_src_param(ins
, &ins
->src
[0], cmp_channel
, &src0_param
);
2020 shader_glsl_add_src_param(ins
, &ins
->src
[1], write_mask
, &src1_param
);
2021 shader_glsl_add_src_param(ins
, &ins
->src
[2], write_mask
, &src2_param
);
2023 shader_addline(ins
->ctx
->buffer
, "%s >= 0.0 ? %s : %s);\n",
2024 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
2027 if(temp_destination
) {
2028 shader_glsl_get_write_mask(&ins
->dst
[0], mask_char
);
2029 shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2030 shader_addline(ins
->ctx
->buffer
, "tmp0%s);\n", mask_char
);
2036 /** Process the CND opcode in GLSL (dst = (src0 > 0.5) ? src1 : src2) */
2037 /* For ps 1.1-1.3, only a single component of src0 is used. For ps 1.4
2038 * the compare is done per component of src0. */
2039 static void shader_glsl_cnd(const struct wined3d_shader_instruction
*ins
)
2041 struct wined3d_shader_dst_param dst
;
2042 glsl_src_param_t src0_param
;
2043 glsl_src_param_t src1_param
;
2044 glsl_src_param_t src2_param
;
2045 DWORD write_mask
, cmp_channel
= 0;
2048 DWORD shader_version
= WINED3D_SHADER_VERSION(ins
->ctx
->reg_maps
->shader_version
.major
,
2049 ins
->ctx
->reg_maps
->shader_version
.minor
);
2051 if (shader_version
< WINED3D_SHADER_VERSION(1, 4))
2053 write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2054 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2055 shader_glsl_add_src_param(ins
, &ins
->src
[1], write_mask
, &src1_param
);
2056 shader_glsl_add_src_param(ins
, &ins
->src
[2], write_mask
, &src2_param
);
2058 /* Fun: The D3DSI_COISSUE flag changes the semantic of the cnd instruction for < 1.4 shaders */
2061 shader_addline(ins
->ctx
->buffer
, "%s /* COISSUE! */);\n", src1_param
.param_str
);
2063 shader_addline(ins
->ctx
->buffer
, "%s > 0.5 ? %s : %s);\n",
2064 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
2068 /* Cycle through all source0 channels */
2069 dst_mask
= ins
->dst
[0].write_mask
;
2071 for (i
=0; i
<4; i
++) {
2073 /* Find the destination channels which use the current source0 channel */
2074 for (j
=0; j
<4; j
++) {
2075 if (((ins
->src
[0].swizzle
>> (2 * j
)) & 0x3) == i
)
2077 write_mask
|= WINED3DSP_WRITEMASK_0
<< j
;
2078 cmp_channel
= WINED3DSP_WRITEMASK_0
<< j
;
2082 dst
.write_mask
= dst_mask
& write_mask
;
2083 write_mask
= shader_glsl_append_dst_ext(ins
->ctx
->buffer
, ins
, &dst
);
2084 if (!write_mask
) continue;
2086 shader_glsl_add_src_param(ins
, &ins
->src
[0], cmp_channel
, &src0_param
);
2087 shader_glsl_add_src_param(ins
, &ins
->src
[1], write_mask
, &src1_param
);
2088 shader_glsl_add_src_param(ins
, &ins
->src
[2], write_mask
, &src2_param
);
2090 shader_addline(ins
->ctx
->buffer
, "%s > 0.5 ? %s : %s);\n",
2091 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
2095 /** GLSL code generation for WINED3DSIO_MAD: Multiply the first 2 opcodes, then add the last */
2096 static void shader_glsl_mad(const struct wined3d_shader_instruction
*ins
)
2098 glsl_src_param_t src0_param
;
2099 glsl_src_param_t src1_param
;
2100 glsl_src_param_t src2_param
;
2103 write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2104 shader_glsl_add_src_param(ins
, &ins
->src
[0], write_mask
, &src0_param
);
2105 shader_glsl_add_src_param(ins
, &ins
->src
[1], write_mask
, &src1_param
);
2106 shader_glsl_add_src_param(ins
, &ins
->src
[2], write_mask
, &src2_param
);
2107 shader_addline(ins
->ctx
->buffer
, "(%s * %s) + %s);\n",
2108 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
2111 /** Handles transforming all WINED3DSIO_M?x? opcodes for
2112 Vertex shaders to GLSL codes */
2113 static void shader_glsl_mnxn(const struct wined3d_shader_instruction
*ins
)
2116 int nComponents
= 0;
2117 struct wined3d_shader_dst_param tmp_dst
= {0};
2118 struct wined3d_shader_src_param tmp_src
[2] = {{0}};
2119 struct wined3d_shader_instruction tmp_ins
;
2121 memset(&tmp_ins
, 0, sizeof(tmp_ins
));
2123 /* Set constants for the temporary argument */
2124 tmp_ins
.ctx
= ins
->ctx
;
2125 tmp_ins
.dst_count
= 1;
2126 tmp_ins
.dst
= &tmp_dst
;
2127 tmp_ins
.src_count
= 2;
2128 tmp_ins
.src
= tmp_src
;
2130 switch(ins
->handler_idx
)
2132 case WINED3DSIH_M4x4
:
2134 tmp_ins
.handler_idx
= WINED3DSIH_DP4
;
2136 case WINED3DSIH_M4x3
:
2138 tmp_ins
.handler_idx
= WINED3DSIH_DP4
;
2140 case WINED3DSIH_M3x4
:
2142 tmp_ins
.handler_idx
= WINED3DSIH_DP3
;
2144 case WINED3DSIH_M3x3
:
2146 tmp_ins
.handler_idx
= WINED3DSIH_DP3
;
2148 case WINED3DSIH_M3x2
:
2150 tmp_ins
.handler_idx
= WINED3DSIH_DP3
;
2156 tmp_dst
= ins
->dst
[0];
2157 tmp_src
[0] = ins
->src
[0];
2158 tmp_src
[1] = ins
->src
[1];
2159 for (i
= 0; i
< nComponents
; ++i
)
2161 tmp_dst
.write_mask
= WINED3DSP_WRITEMASK_0
<< i
;
2162 shader_glsl_dot(&tmp_ins
);
2163 ++tmp_src
[1].register_idx
;
2168 The LRP instruction performs a component-wise linear interpolation
2169 between the second and third operands using the first operand as the
2170 blend factor. Equation: (dst = src2 + src0 * (src1 - src2))
2171 This is equivalent to mix(src2, src1, src0);
2173 static void shader_glsl_lrp(const struct wined3d_shader_instruction
*ins
)
2175 glsl_src_param_t src0_param
;
2176 glsl_src_param_t src1_param
;
2177 glsl_src_param_t src2_param
;
2180 write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2182 shader_glsl_add_src_param(ins
, &ins
->src
[0], write_mask
, &src0_param
);
2183 shader_glsl_add_src_param(ins
, &ins
->src
[1], write_mask
, &src1_param
);
2184 shader_glsl_add_src_param(ins
, &ins
->src
[2], write_mask
, &src2_param
);
2186 shader_addline(ins
->ctx
->buffer
, "mix(%s, %s, %s));\n",
2187 src2_param
.param_str
, src1_param
.param_str
, src0_param
.param_str
);
2190 /** Process the WINED3DSIO_LIT instruction in GLSL:
2191 * dst.x = dst.w = 1.0
2192 * dst.y = (src0.x > 0) ? src0.x
2193 * dst.z = (src0.x > 0) ? ((src0.y > 0) ? pow(src0.y, src.w) : 0) : 0
2194 * where src.w is clamped at +- 128
2196 static void shader_glsl_lit(const struct wined3d_shader_instruction
*ins
)
2198 glsl_src_param_t src0_param
;
2199 glsl_src_param_t src1_param
;
2200 glsl_src_param_t src3_param
;
2203 shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2204 shader_glsl_get_write_mask(&ins
->dst
[0], dst_mask
);
2206 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2207 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_1
, &src1_param
);
2208 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_3
, &src3_param
);
2210 /* The sdk specifies the instruction like this
2212 * if(src.x > 0.0) dst.y = src.x
2214 * if(src.x > 0.0 && src.y > 0.0) dst.z = pow(src.y, power);
2218 * Obviously that has quite a few conditionals in it which we don't like. So the first step is this:
2219 * dst.x = 1.0 ... No further explanation needed
2220 * dst.y = max(src.y, 0.0); ... If x < 0.0, use 0.0, otherwise x. Same as the conditional
2221 * dst.z = x > 0.0 ? pow(max(y, 0.0), p) : 0; ... 0 ^ power is 0, and otherwise we use y anyway
2222 * dst.w = 1.0. ... Nothing fancy.
2224 * So we still have one conditional in there. So do this:
2225 * dst.z = pow(max(0.0, src.y) * step(0.0, src.x), power);
2227 * 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),
2228 * 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.
2229 * if both x and y are > 0, we get pow(y * 1.0, power), as it is supposed to
2231 shader_addline(ins
->ctx
->buffer
,
2232 "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",
2233 src0_param
.param_str
, src1_param
.param_str
, src0_param
.param_str
, src3_param
.param_str
, dst_mask
);
2236 /** Process the WINED3DSIO_DST instruction in GLSL:
2238 * dst.y = src0.x * src0.y
2242 static void shader_glsl_dst(const struct wined3d_shader_instruction
*ins
)
2244 glsl_src_param_t src0y_param
;
2245 glsl_src_param_t src0z_param
;
2246 glsl_src_param_t src1y_param
;
2247 glsl_src_param_t src1w_param
;
2250 shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2251 shader_glsl_get_write_mask(&ins
->dst
[0], dst_mask
);
2253 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_1
, &src0y_param
);
2254 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_2
, &src0z_param
);
2255 shader_glsl_add_src_param(ins
, &ins
->src
[1], WINED3DSP_WRITEMASK_1
, &src1y_param
);
2256 shader_glsl_add_src_param(ins
, &ins
->src
[1], WINED3DSP_WRITEMASK_3
, &src1w_param
);
2258 shader_addline(ins
->ctx
->buffer
, "vec4(1.0, %s * %s, %s, %s))%s;\n",
2259 src0y_param
.param_str
, src1y_param
.param_str
, src0z_param
.param_str
, src1w_param
.param_str
, dst_mask
);
2262 /** Process the WINED3DSIO_SINCOS instruction in GLSL:
2263 * VS 2.0 requires that specific cosine and sine constants be passed to this instruction so the hardware
2264 * can handle it. But, these functions are built-in for GLSL, so we can just ignore the last 2 params.
2266 * dst.x = cos(src0.?)
2267 * dst.y = sin(src0.?)
2271 static void shader_glsl_sincos(const struct wined3d_shader_instruction
*ins
)
2273 glsl_src_param_t src0_param
;
2276 write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2277 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2279 switch (write_mask
) {
2280 case WINED3DSP_WRITEMASK_0
:
2281 shader_addline(ins
->ctx
->buffer
, "cos(%s));\n", src0_param
.param_str
);
2284 case WINED3DSP_WRITEMASK_1
:
2285 shader_addline(ins
->ctx
->buffer
, "sin(%s));\n", src0_param
.param_str
);
2288 case (WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
):
2289 shader_addline(ins
->ctx
->buffer
, "vec2(cos(%s), sin(%s)));\n", src0_param
.param_str
, src0_param
.param_str
);
2293 ERR("Write mask should be .x, .y or .xy\n");
2298 /** Process the WINED3DSIO_LOOP instruction in GLSL:
2299 * Start a for() loop where src1.y is the initial value of aL,
2300 * increment aL by src1.z for a total of src1.x iterations.
2301 * Need to use a temporary variable for this operation.
2303 /* FIXME: I don't think nested loops will work correctly this way. */
2304 static void shader_glsl_loop(const struct wined3d_shader_instruction
*ins
)
2306 glsl_src_param_t src1_param
;
2307 IWineD3DBaseShaderImpl
*shader
= (IWineD3DBaseShaderImpl
*)ins
->ctx
->shader
;
2308 const DWORD
*control_values
= NULL
;
2309 const local_constant
*constant
;
2311 shader_glsl_add_src_param(ins
, &ins
->src
[1], WINED3DSP_WRITEMASK_ALL
, &src1_param
);
2313 /* Try to hardcode the loop control parameters if possible. Direct3D 9 class hardware doesn't support real
2314 * varying indexing, but Microsoft designed this feature for Shader model 2.x+. If the loop control is
2315 * known at compile time, the GLSL compiler can unroll the loop, and replace indirect addressing with direct
2318 if (ins
->src
[1].register_type
== WINED3DSPR_CONSTINT
)
2320 LIST_FOR_EACH_ENTRY(constant
, &shader
->baseShader
.constantsI
, local_constant
, entry
) {
2321 if (constant
->idx
== ins
->src
[1].register_idx
)
2323 control_values
= constant
->value
;
2329 if(control_values
) {
2330 if(control_values
[2] > 0) {
2331 shader_addline(ins
->ctx
->buffer
, "for (aL%u = %d; aL%u < (%d * %d + %d); aL%u += %d) {\n",
2332 shader
->baseShader
.cur_loop_depth
, control_values
[1],
2333 shader
->baseShader
.cur_loop_depth
, control_values
[0], control_values
[2], control_values
[1],
2334 shader
->baseShader
.cur_loop_depth
, control_values
[2]);
2335 } else if(control_values
[2] == 0) {
2336 shader_addline(ins
->ctx
->buffer
, "for (aL%u = %d, tmpInt%u = 0; tmpInt%u < %d; tmpInt%u++) {\n",
2337 shader
->baseShader
.cur_loop_depth
, control_values
[1], shader
->baseShader
.cur_loop_depth
,
2338 shader
->baseShader
.cur_loop_depth
, control_values
[0],
2339 shader
->baseShader
.cur_loop_depth
);
2341 shader_addline(ins
->ctx
->buffer
, "for (aL%u = %d; aL%u > (%d * %d + %d); aL%u += %d) {\n",
2342 shader
->baseShader
.cur_loop_depth
, control_values
[1],
2343 shader
->baseShader
.cur_loop_depth
, control_values
[0], control_values
[2], control_values
[1],
2344 shader
->baseShader
.cur_loop_depth
, control_values
[2]);
2347 shader_addline(ins
->ctx
->buffer
,
2348 "for (tmpInt%u = 0, aL%u = %s.y; tmpInt%u < %s.x; tmpInt%u++, aL%u += %s.z) {\n",
2349 shader
->baseShader
.cur_loop_depth
, shader
->baseShader
.cur_loop_regno
,
2350 src1_param
.reg_name
, shader
->baseShader
.cur_loop_depth
, src1_param
.reg_name
,
2351 shader
->baseShader
.cur_loop_depth
, shader
->baseShader
.cur_loop_regno
, src1_param
.reg_name
);
2354 shader
->baseShader
.cur_loop_depth
++;
2355 shader
->baseShader
.cur_loop_regno
++;
2358 static void shader_glsl_end(const struct wined3d_shader_instruction
*ins
)
2360 IWineD3DBaseShaderImpl
*shader
= (IWineD3DBaseShaderImpl
*)ins
->ctx
->shader
;
2362 shader_addline(ins
->ctx
->buffer
, "}\n");
2364 if (ins
->handler_idx
== WINED3DSIH_ENDLOOP
)
2366 shader
->baseShader
.cur_loop_depth
--;
2367 shader
->baseShader
.cur_loop_regno
--;
2370 if (ins
->handler_idx
== WINED3DSIH_ENDREP
)
2372 shader
->baseShader
.cur_loop_depth
--;
2376 static void shader_glsl_rep(const struct wined3d_shader_instruction
*ins
)
2378 IWineD3DBaseShaderImpl
*shader
= (IWineD3DBaseShaderImpl
*)ins
->ctx
->shader
;
2379 glsl_src_param_t src0_param
;
2380 const DWORD
*control_values
= NULL
;
2381 const local_constant
*constant
;
2383 /* Try to hardcode local values to help the GLSL compiler to unroll and optimize the loop */
2384 if(ins
->src
[0].register_type
== WINED3DSPR_CONSTINT
) {
2385 LIST_FOR_EACH_ENTRY(constant
, &shader
->baseShader
.constantsI
, local_constant
, entry
) {
2386 if(constant
->idx
== ins
->src
[0].register_idx
) {
2387 control_values
= constant
->value
;
2393 if(control_values
) {
2394 shader_addline(ins
->ctx
->buffer
, "for (tmpInt%d = 0; tmpInt%d < %d; tmpInt%d++) {\n",
2395 shader
->baseShader
.cur_loop_depth
, shader
->baseShader
.cur_loop_depth
,
2396 control_values
[0], shader
->baseShader
.cur_loop_depth
);
2398 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2399 shader_addline(ins
->ctx
->buffer
, "for (tmpInt%d = 0; tmpInt%d < %s; tmpInt%d++) {\n",
2400 shader
->baseShader
.cur_loop_depth
, shader
->baseShader
.cur_loop_depth
,
2401 src0_param
.param_str
, shader
->baseShader
.cur_loop_depth
);
2403 shader
->baseShader
.cur_loop_depth
++;
2406 static void shader_glsl_if(const struct wined3d_shader_instruction
*ins
)
2408 glsl_src_param_t src0_param
;
2410 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2411 shader_addline(ins
->ctx
->buffer
, "if (%s) {\n", src0_param
.param_str
);
2414 static void shader_glsl_ifc(const struct wined3d_shader_instruction
*ins
)
2416 glsl_src_param_t src0_param
;
2417 glsl_src_param_t src1_param
;
2419 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2420 shader_glsl_add_src_param(ins
, &ins
->src
[1], WINED3DSP_WRITEMASK_0
, &src1_param
);
2422 shader_addline(ins
->ctx
->buffer
, "if (%s %s %s) {\n",
2423 src0_param
.param_str
, shader_get_comp_op(ins
->flags
), src1_param
.param_str
);
2426 static void shader_glsl_else(const struct wined3d_shader_instruction
*ins
)
2428 shader_addline(ins
->ctx
->buffer
, "} else {\n");
2431 static void shader_glsl_break(const struct wined3d_shader_instruction
*ins
)
2433 shader_addline(ins
->ctx
->buffer
, "break;\n");
2436 /* FIXME: According to MSDN the compare is done per component. */
2437 static void shader_glsl_breakc(const struct wined3d_shader_instruction
*ins
)
2439 glsl_src_param_t src0_param
;
2440 glsl_src_param_t src1_param
;
2442 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2443 shader_glsl_add_src_param(ins
, &ins
->src
[1], WINED3DSP_WRITEMASK_0
, &src1_param
);
2445 shader_addline(ins
->ctx
->buffer
, "if (%s %s %s) break;\n",
2446 src0_param
.param_str
, shader_get_comp_op(ins
->flags
), src1_param
.param_str
);
2449 static void shader_glsl_label(const struct wined3d_shader_instruction
*ins
)
2451 shader_addline(ins
->ctx
->buffer
, "}\n");
2452 shader_addline(ins
->ctx
->buffer
, "void subroutine%u () {\n", ins
->src
[0].register_idx
);
2455 static void shader_glsl_call(const struct wined3d_shader_instruction
*ins
)
2457 shader_addline(ins
->ctx
->buffer
, "subroutine%u();\n", ins
->src
[0].register_idx
);
2460 static void shader_glsl_callnz(const struct wined3d_shader_instruction
*ins
)
2462 glsl_src_param_t src1_param
;
2464 shader_glsl_add_src_param(ins
, &ins
->src
[1], WINED3DSP_WRITEMASK_0
, &src1_param
);
2465 shader_addline(ins
->ctx
->buffer
, "if (%s) subroutine%u();\n", src1_param
.param_str
, ins
->src
[0].register_idx
);
2468 /*********************************************
2469 * Pixel Shader Specific Code begins here
2470 ********************************************/
2471 static void pshader_glsl_tex(const struct wined3d_shader_instruction
*ins
)
2473 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*)ins
->ctx
->shader
;
2474 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
2475 DWORD shader_version
= WINED3D_SHADER_VERSION(ins
->ctx
->reg_maps
->shader_version
.major
,
2476 ins
->ctx
->reg_maps
->shader_version
.minor
);
2477 glsl_sample_function_t sample_function
;
2478 DWORD sample_flags
= 0;
2479 WINED3DSAMPLER_TEXTURE_TYPE sampler_type
;
2481 DWORD mask
= 0, swizzle
;
2483 /* 1.0-1.4: Use destination register as sampler source.
2484 * 2.0+: Use provided sampler source. */
2485 if (shader_version
< WINED3D_SHADER_VERSION(2,0)) sampler_idx
= ins
->dst
[0].register_idx
;
2486 else sampler_idx
= ins
->src
[1].register_idx
;
2487 sampler_type
= ins
->ctx
->reg_maps
->sampler_type
[sampler_idx
];
2489 if (shader_version
< WINED3D_SHADER_VERSION(1,4))
2491 DWORD flags
= deviceImpl
->stateBlock
->textureState
[sampler_idx
][WINED3DTSS_TEXTURETRANSFORMFLAGS
];
2493 /* Projected cube textures don't make a lot of sense, the resulting coordinates stay the same. */
2494 if (flags
& WINED3DTTFF_PROJECTED
&& sampler_type
!= WINED3DSTT_CUBE
) {
2495 sample_flags
|= WINED3D_GLSL_SAMPLE_PROJECTED
;
2496 switch (flags
& ~WINED3DTTFF_PROJECTED
) {
2497 case WINED3DTTFF_COUNT1
: FIXME("WINED3DTTFF_PROJECTED with WINED3DTTFF_COUNT1?\n"); break;
2498 case WINED3DTTFF_COUNT2
: mask
= WINED3DSP_WRITEMASK_1
; break;
2499 case WINED3DTTFF_COUNT3
: mask
= WINED3DSP_WRITEMASK_2
; break;
2500 case WINED3DTTFF_COUNT4
:
2501 case WINED3DTTFF_DISABLE
: mask
= WINED3DSP_WRITEMASK_3
; break;
2505 else if (shader_version
< WINED3D_SHADER_VERSION(2,0))
2507 DWORD src_mod
= ins
->src
[0].modifiers
;
2509 if (src_mod
== WINED3DSPSM_DZ
) {
2510 sample_flags
|= WINED3D_GLSL_SAMPLE_PROJECTED
;
2511 mask
= WINED3DSP_WRITEMASK_2
;
2512 } else if (src_mod
== WINED3DSPSM_DW
) {
2513 sample_flags
|= WINED3D_GLSL_SAMPLE_PROJECTED
;
2514 mask
= WINED3DSP_WRITEMASK_3
;
2517 if (ins
->flags
& WINED3DSI_TEXLD_PROJECT
)
2519 /* ps 2.0 texldp instruction always divides by the fourth component. */
2520 sample_flags
|= WINED3D_GLSL_SAMPLE_PROJECTED
;
2521 mask
= WINED3DSP_WRITEMASK_3
;
2525 if(deviceImpl
->stateBlock
->textures
[sampler_idx
] &&
2526 IWineD3DBaseTexture_GetTextureDimensions(deviceImpl
->stateBlock
->textures
[sampler_idx
]) == GL_TEXTURE_RECTANGLE_ARB
) {
2527 sample_flags
|= WINED3D_GLSL_SAMPLE_RECT
;
2530 shader_glsl_get_sample_function(sampler_type
, sample_flags
, &sample_function
);
2531 mask
|= sample_function
.coord_mask
;
2533 if (shader_version
< WINED3D_SHADER_VERSION(2,0)) swizzle
= WINED3DSP_NOSWIZZLE
;
2534 else swizzle
= ins
->src
[1].swizzle
;
2536 /* 1.0-1.3: Use destination register as coordinate source.
2537 1.4+: Use provided coordinate source register. */
2538 if (shader_version
< WINED3D_SHADER_VERSION(1,4))
2541 shader_glsl_write_mask_to_str(mask
, coord_mask
);
2542 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, swizzle
, NULL
, NULL
, NULL
,
2543 "T%u%s", sampler_idx
, coord_mask
);
2545 glsl_src_param_t coord_param
;
2546 shader_glsl_add_src_param(ins
, &ins
->src
[0], mask
, &coord_param
);
2547 if (ins
->flags
& WINED3DSI_TEXLD_BIAS
)
2549 glsl_src_param_t bias
;
2550 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_3
, &bias
);
2551 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, swizzle
, NULL
, NULL
, bias
.param_str
,
2552 "%s", coord_param
.param_str
);
2554 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, swizzle
, NULL
, NULL
, NULL
,
2555 "%s", coord_param
.param_str
);
2560 static void shader_glsl_texldd(const struct wined3d_shader_instruction
*ins
)
2562 IWineD3DBaseShaderImpl
*This
= (IWineD3DBaseShaderImpl
*)ins
->ctx
->shader
;
2563 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
2564 const WineD3D_GL_Info
* gl_info
= &deviceImpl
->adapter
->gl_info
;
2565 glsl_sample_function_t sample_function
;
2566 glsl_src_param_t coord_param
, dx_param
, dy_param
;
2567 DWORD sample_flags
= WINED3D_GLSL_SAMPLE_GRAD
;
2570 DWORD swizzle
= ins
->src
[1].swizzle
;
2572 if(!GL_SUPPORT(ARB_SHADER_TEXTURE_LOD
)) {
2573 FIXME("texldd used, but not supported by hardware. Falling back to regular tex\n");
2574 return pshader_glsl_tex(ins
);
2577 sampler_idx
= ins
->src
[1].register_idx
;
2578 sampler_type
= ins
->ctx
->reg_maps
->sampler_type
[sampler_idx
];
2579 if(deviceImpl
->stateBlock
->textures
[sampler_idx
] &&
2580 IWineD3DBaseTexture_GetTextureDimensions(deviceImpl
->stateBlock
->textures
[sampler_idx
]) == GL_TEXTURE_RECTANGLE_ARB
) {
2581 sample_flags
|= WINED3D_GLSL_SAMPLE_RECT
;
2584 shader_glsl_get_sample_function(sampler_type
, sample_flags
, &sample_function
);
2585 shader_glsl_add_src_param(ins
, &ins
->src
[0], sample_function
.coord_mask
, &coord_param
);
2586 shader_glsl_add_src_param(ins
, &ins
->src
[2], sample_function
.coord_mask
, &dx_param
);
2587 shader_glsl_add_src_param(ins
, &ins
->src
[3], sample_function
.coord_mask
, &dy_param
);
2589 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, swizzle
, dx_param
.param_str
, dy_param
.param_str
, NULL
,
2590 "%s", coord_param
.param_str
);
2593 static void shader_glsl_texldl(const struct wined3d_shader_instruction
*ins
)
2595 IWineD3DBaseShaderImpl
*This
= (IWineD3DBaseShaderImpl
*)ins
->ctx
->shader
;
2596 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
2597 glsl_sample_function_t sample_function
;
2598 glsl_src_param_t coord_param
, lod_param
;
2599 DWORD sample_flags
= WINED3D_GLSL_SAMPLE_LOD
;
2602 DWORD swizzle
= ins
->src
[1].swizzle
;
2604 sampler_idx
= ins
->src
[1].register_idx
;
2605 sampler_type
= ins
->ctx
->reg_maps
->sampler_type
[sampler_idx
];
2606 if(deviceImpl
->stateBlock
->textures
[sampler_idx
] &&
2607 IWineD3DBaseTexture_GetTextureDimensions(deviceImpl
->stateBlock
->textures
[sampler_idx
]) == GL_TEXTURE_RECTANGLE_ARB
) {
2608 sample_flags
|= WINED3D_GLSL_SAMPLE_RECT
;
2610 shader_glsl_get_sample_function(sampler_type
, sample_flags
, &sample_function
);
2611 shader_glsl_add_src_param(ins
, &ins
->src
[0], sample_function
.coord_mask
, &coord_param
);
2613 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_3
, &lod_param
);
2615 if (shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
.type
))
2617 /* The GLSL spec claims the Lod sampling functions are only supported in vertex shaders.
2618 * However, they seem to work just fine in fragment shaders as well. */
2619 WARN("Using %s in fragment shader.\n", sample_function
.name
);
2621 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, swizzle
, NULL
, NULL
, lod_param
.param_str
,
2622 "%s", coord_param
.param_str
);
2625 static void pshader_glsl_texcoord(const struct wined3d_shader_instruction
*ins
)
2627 /* FIXME: Make this work for more than just 2D textures */
2628 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
2629 DWORD write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2631 if (!(ins
->ctx
->reg_maps
->shader_version
.major
== 1 && ins
->ctx
->reg_maps
->shader_version
.minor
== 4))
2635 shader_glsl_get_write_mask(&ins
->dst
[0], dst_mask
);
2636 shader_addline(buffer
, "clamp(gl_TexCoord[%u], 0.0, 1.0)%s);\n",
2637 ins
->dst
[0].register_idx
, dst_mask
);
2639 DWORD reg
= ins
->src
[0].register_idx
;
2640 DWORD src_mod
= ins
->src
[0].modifiers
;
2641 char dst_swizzle
[6];
2643 shader_glsl_get_swizzle(&ins
->src
[0], FALSE
, write_mask
, dst_swizzle
);
2645 if (src_mod
== WINED3DSPSM_DZ
) {
2646 glsl_src_param_t div_param
;
2647 unsigned int mask_size
= shader_glsl_get_write_mask_size(write_mask
);
2648 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_2
, &div_param
);
2650 if (mask_size
> 1) {
2651 shader_addline(buffer
, "gl_TexCoord[%u]%s / vec%d(%s));\n", reg
, dst_swizzle
, mask_size
, div_param
.param_str
);
2653 shader_addline(buffer
, "gl_TexCoord[%u]%s / %s);\n", reg
, dst_swizzle
, div_param
.param_str
);
2655 } else if (src_mod
== WINED3DSPSM_DW
) {
2656 glsl_src_param_t div_param
;
2657 unsigned int mask_size
= shader_glsl_get_write_mask_size(write_mask
);
2658 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_3
, &div_param
);
2660 if (mask_size
> 1) {
2661 shader_addline(buffer
, "gl_TexCoord[%u]%s / vec%d(%s));\n", reg
, dst_swizzle
, mask_size
, div_param
.param_str
);
2663 shader_addline(buffer
, "gl_TexCoord[%u]%s / %s);\n", reg
, dst_swizzle
, div_param
.param_str
);
2666 shader_addline(buffer
, "gl_TexCoord[%u]%s);\n", reg
, dst_swizzle
);
2671 /** Process the WINED3DSIO_TEXDP3TEX instruction in GLSL:
2672 * Take a 3-component dot product of the TexCoord[dstreg] and src,
2673 * then perform a 1D texture lookup from stage dstregnum, place into dst. */
2674 static void pshader_glsl_texdp3tex(const struct wined3d_shader_instruction
*ins
)
2676 glsl_src_param_t src0_param
;
2677 glsl_sample_function_t sample_function
;
2678 DWORD sampler_idx
= ins
->dst
[0].register_idx
;
2679 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2680 WINED3DSAMPLER_TEXTURE_TYPE sampler_type
= ins
->ctx
->reg_maps
->sampler_type
[sampler_idx
];
2683 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
2685 /* Do I have to take care about the projected bit? I don't think so, since the dp3 returns only one
2686 * scalar, and projected sampling would require 4.
2688 * It is a dependent read - not valid with conditional NP2 textures
2690 shader_glsl_get_sample_function(sampler_type
, 0, &sample_function
);
2691 mask_size
= shader_glsl_get_write_mask_size(sample_function
.coord_mask
);
2696 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, WINED3DSP_NOSWIZZLE
, NULL
, NULL
, NULL
,
2697 "dot(gl_TexCoord[%u].xyz, %s)", sampler_idx
, src0_param
.param_str
);
2701 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, WINED3DSP_NOSWIZZLE
, NULL
, NULL
, NULL
,
2702 "vec2(dot(gl_TexCoord[%u].xyz, %s), 0.0)", sampler_idx
, src0_param
.param_str
);
2706 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, WINED3DSP_NOSWIZZLE
, NULL
, NULL
, NULL
,
2707 "vec3(dot(gl_TexCoord[%u].xyz, %s), 0.0, 0.0)", sampler_idx
, src0_param
.param_str
);
2711 FIXME("Unexpected mask size %u\n", mask_size
);
2716 /** Process the WINED3DSIO_TEXDP3 instruction in GLSL:
2717 * Take a 3-component dot product of the TexCoord[dstreg] and src. */
2718 static void pshader_glsl_texdp3(const struct wined3d_shader_instruction
*ins
)
2720 glsl_src_param_t src0_param
;
2721 DWORD dstreg
= ins
->dst
[0].register_idx
;
2722 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2724 unsigned int mask_size
;
2726 dst_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2727 mask_size
= shader_glsl_get_write_mask_size(dst_mask
);
2728 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
2730 if (mask_size
> 1) {
2731 shader_addline(ins
->ctx
->buffer
, "vec%d(dot(T%u.xyz, %s)));\n", mask_size
, dstreg
, src0_param
.param_str
);
2733 shader_addline(ins
->ctx
->buffer
, "dot(T%u.xyz, %s));\n", dstreg
, src0_param
.param_str
);
2737 /** Process the WINED3DSIO_TEXDEPTH instruction in GLSL:
2738 * Calculate the depth as dst.x / dst.y */
2739 static void pshader_glsl_texdepth(const struct wined3d_shader_instruction
*ins
)
2741 glsl_dst_param_t dst_param
;
2743 shader_glsl_add_dst_param(ins
, &ins
->dst
[0], &dst_param
);
2745 /* Tests show that texdepth never returns anything below 0.0, and that r5.y is clamped to 1.0.
2746 * Negative input is accepted, -0.25 / -0.5 returns 0.5. GL should clamp gl_FragDepth to [0;1], but
2747 * this doesn't always work, so clamp the results manually. Whether or not the x value is clamped at 1
2748 * too is irrelevant, since if x = 0, any y value < 1.0 (and > 1.0 is not allowed) results in a result
2751 shader_addline(ins
->ctx
->buffer
, "gl_FragDepth = clamp((%s.x / min(%s.y, 1.0)), 0.0, 1.0);\n",
2752 dst_param
.reg_name
, dst_param
.reg_name
);
2755 /** Process the WINED3DSIO_TEXM3X2DEPTH instruction in GLSL:
2756 * Last row of a 3x2 matrix multiply, use the result to calculate the depth:
2757 * Calculate tmp0.y = TexCoord[dstreg] . src.xyz; (tmp0.x has already been calculated)
2758 * depth = (tmp0.y == 0.0) ? 1.0 : tmp0.x / tmp0.y
2760 static void pshader_glsl_texm3x2depth(const struct wined3d_shader_instruction
*ins
)
2762 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2763 DWORD dstreg
= ins
->dst
[0].register_idx
;
2764 glsl_src_param_t src0_param
;
2766 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
2768 shader_addline(ins
->ctx
->buffer
, "tmp0.y = dot(T%u.xyz, %s);\n", dstreg
, src0_param
.param_str
);
2769 shader_addline(ins
->ctx
->buffer
, "gl_FragDepth = (tmp0.y == 0.0) ? 1.0 : clamp(tmp0.x / tmp0.y, 0.0, 1.0);\n");
2772 /** Process the WINED3DSIO_TEXM3X2PAD instruction in GLSL
2773 * Calculate the 1st of a 2-row matrix multiplication. */
2774 static void pshader_glsl_texm3x2pad(const struct wined3d_shader_instruction
*ins
)
2776 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2777 DWORD reg
= ins
->dst
[0].register_idx
;
2778 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
2779 glsl_src_param_t src0_param
;
2781 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
2782 shader_addline(buffer
, "tmp0.x = dot(T%u.xyz, %s);\n", reg
, src0_param
.param_str
);
2785 /** Process the WINED3DSIO_TEXM3X3PAD instruction in GLSL
2786 * Calculate the 1st or 2nd row of a 3-row matrix multiplication. */
2787 static void pshader_glsl_texm3x3pad(const struct wined3d_shader_instruction
*ins
)
2789 IWineD3DPixelShaderImpl
*shader
= (IWineD3DPixelShaderImpl
*)ins
->ctx
->shader
;
2790 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2791 DWORD reg
= ins
->dst
[0].register_idx
;
2792 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
2793 SHADER_PARSE_STATE
* current_state
= &shader
->baseShader
.parse_state
;
2794 glsl_src_param_t src0_param
;
2796 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
2797 shader_addline(buffer
, "tmp0.%c = dot(T%u.xyz, %s);\n", 'x' + current_state
->current_row
, reg
, src0_param
.param_str
);
2798 current_state
->texcoord_w
[current_state
->current_row
++] = reg
;
2801 static void pshader_glsl_texm3x2tex(const struct wined3d_shader_instruction
*ins
)
2803 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2804 DWORD reg
= ins
->dst
[0].register_idx
;
2805 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
2806 glsl_src_param_t src0_param
;
2807 WINED3DSAMPLER_TEXTURE_TYPE sampler_type
= ins
->ctx
->reg_maps
->sampler_type
[reg
];
2808 glsl_sample_function_t sample_function
;
2810 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
2811 shader_addline(buffer
, "tmp0.y = dot(T%u.xyz, %s);\n", reg
, src0_param
.param_str
);
2813 shader_glsl_get_sample_function(sampler_type
, 0, &sample_function
);
2815 /* Sample the texture using the calculated coordinates */
2816 shader_glsl_gen_sample_code(ins
, reg
, &sample_function
, WINED3DSP_NOSWIZZLE
, NULL
, NULL
, NULL
, "tmp0.xy");
2819 /** Process the WINED3DSIO_TEXM3X3TEX instruction in GLSL
2820 * Perform the 3rd row of a 3x3 matrix multiply, then sample the texture using the calculated coordinates */
2821 static void pshader_glsl_texm3x3tex(const struct wined3d_shader_instruction
*ins
)
2823 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2824 glsl_src_param_t src0_param
;
2825 DWORD reg
= ins
->dst
[0].register_idx
;
2826 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*)ins
->ctx
->shader
;
2827 SHADER_PARSE_STATE
* current_state
= &This
->baseShader
.parse_state
;
2828 WINED3DSAMPLER_TEXTURE_TYPE sampler_type
= ins
->ctx
->reg_maps
->sampler_type
[reg
];
2829 glsl_sample_function_t sample_function
;
2831 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
2832 shader_addline(ins
->ctx
->buffer
, "tmp0.z = dot(T%u.xyz, %s);\n", reg
, src0_param
.param_str
);
2834 /* Dependent read, not valid with conditional NP2 */
2835 shader_glsl_get_sample_function(sampler_type
, 0, &sample_function
);
2837 /* Sample the texture using the calculated coordinates */
2838 shader_glsl_gen_sample_code(ins
, reg
, &sample_function
, WINED3DSP_NOSWIZZLE
, NULL
, NULL
, NULL
, "tmp0.xyz");
2840 current_state
->current_row
= 0;
2843 /** Process the WINED3DSIO_TEXM3X3 instruction in GLSL
2844 * Perform the 3rd row of a 3x3 matrix multiply */
2845 static void pshader_glsl_texm3x3(const struct wined3d_shader_instruction
*ins
)
2847 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2848 glsl_src_param_t src0_param
;
2850 DWORD reg
= ins
->dst
[0].register_idx
;
2851 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*)ins
->ctx
->shader
;
2852 SHADER_PARSE_STATE
* current_state
= &This
->baseShader
.parse_state
;
2854 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
2856 shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2857 shader_glsl_get_write_mask(&ins
->dst
[0], dst_mask
);
2858 shader_addline(ins
->ctx
->buffer
, "vec4(tmp0.xy, dot(T%u.xyz, %s), 1.0)%s);\n", reg
, src0_param
.param_str
, dst_mask
);
2860 current_state
->current_row
= 0;
2863 /** Process the WINED3DSIO_TEXM3X3SPEC instruction in GLSL
2864 * Perform the final texture lookup based on the previous 2 3x3 matrix multiplies */
2865 static void pshader_glsl_texm3x3spec(const struct wined3d_shader_instruction
*ins
)
2867 IWineD3DPixelShaderImpl
*shader
= (IWineD3DPixelShaderImpl
*)ins
->ctx
->shader
;
2868 DWORD reg
= ins
->dst
[0].register_idx
;
2869 glsl_src_param_t src0_param
;
2870 glsl_src_param_t src1_param
;
2871 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
2872 SHADER_PARSE_STATE
* current_state
= &shader
->baseShader
.parse_state
;
2873 WINED3DSAMPLER_TEXTURE_TYPE stype
= ins
->ctx
->reg_maps
->sampler_type
[reg
];
2874 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2875 glsl_sample_function_t sample_function
;
2877 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
2878 shader_glsl_add_src_param(ins
, &ins
->src
[1], src_mask
, &src1_param
);
2880 /* Perform the last matrix multiply operation */
2881 shader_addline(buffer
, "tmp0.z = dot(T%u.xyz, %s);\n", reg
, src0_param
.param_str
);
2882 /* Reflection calculation */
2883 shader_addline(buffer
, "tmp0.xyz = -reflect((%s), normalize(tmp0.xyz));\n", src1_param
.param_str
);
2885 /* Dependent read, not valid with conditional NP2 */
2886 shader_glsl_get_sample_function(stype
, 0, &sample_function
);
2888 /* Sample the texture */
2889 shader_glsl_gen_sample_code(ins
, reg
, &sample_function
, WINED3DSP_NOSWIZZLE
, NULL
, NULL
, NULL
, "tmp0.xyz");
2891 current_state
->current_row
= 0;
2894 /** Process the WINED3DSIO_TEXM3X3VSPEC instruction in GLSL
2895 * Perform the final texture lookup based on the previous 2 3x3 matrix multiplies */
2896 static void pshader_glsl_texm3x3vspec(const struct wined3d_shader_instruction
*ins
)
2898 IWineD3DPixelShaderImpl
*shader
= (IWineD3DPixelShaderImpl
*)ins
->ctx
->shader
;
2899 DWORD reg
= ins
->dst
[0].register_idx
;
2900 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
2901 SHADER_PARSE_STATE
* current_state
= &shader
->baseShader
.parse_state
;
2902 glsl_src_param_t src0_param
;
2903 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2904 WINED3DSAMPLER_TEXTURE_TYPE sampler_type
= ins
->ctx
->reg_maps
->sampler_type
[reg
];
2905 glsl_sample_function_t sample_function
;
2907 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
2909 /* Perform the last matrix multiply operation */
2910 shader_addline(buffer
, "tmp0.z = dot(vec3(T%u), vec3(%s));\n", reg
, src0_param
.param_str
);
2912 /* Construct the eye-ray vector from w coordinates */
2913 shader_addline(buffer
, "tmp1.xyz = normalize(vec3(gl_TexCoord[%u].w, gl_TexCoord[%u].w, gl_TexCoord[%u].w));\n",
2914 current_state
->texcoord_w
[0], current_state
->texcoord_w
[1], reg
);
2915 shader_addline(buffer
, "tmp0.xyz = -reflect(tmp1.xyz, normalize(tmp0.xyz));\n");
2917 /* Dependent read, not valid with conditional NP2 */
2918 shader_glsl_get_sample_function(sampler_type
, 0, &sample_function
);
2920 /* Sample the texture using the calculated coordinates */
2921 shader_glsl_gen_sample_code(ins
, reg
, &sample_function
, WINED3DSP_NOSWIZZLE
, NULL
, NULL
, NULL
, "tmp0.xyz");
2923 current_state
->current_row
= 0;
2926 /** Process the WINED3DSIO_TEXBEM instruction in GLSL.
2927 * Apply a fake bump map transform.
2928 * texbem is pshader <= 1.3 only, this saves a few version checks
2930 static void pshader_glsl_texbem(const struct wined3d_shader_instruction
*ins
)
2932 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*)ins
->ctx
->shader
;
2933 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
2934 glsl_sample_function_t sample_function
;
2935 glsl_src_param_t coord_param
;
2936 WINED3DSAMPLER_TEXTURE_TYPE sampler_type
;
2942 sampler_idx
= ins
->dst
[0].register_idx
;
2943 flags
= deviceImpl
->stateBlock
->textureState
[sampler_idx
][WINED3DTSS_TEXTURETRANSFORMFLAGS
];
2945 sampler_type
= ins
->ctx
->reg_maps
->sampler_type
[sampler_idx
];
2946 /* Dependent read, not valid with conditional NP2 */
2947 shader_glsl_get_sample_function(sampler_type
, 0, &sample_function
);
2948 mask
= sample_function
.coord_mask
;
2950 shader_glsl_write_mask_to_str(mask
, coord_mask
);
2952 /* with projective textures, texbem only divides the static texture coord, not the displacement,
2953 * so we can't let the GL handle this.
2955 if (flags
& WINED3DTTFF_PROJECTED
) {
2957 char coord_div_mask
[3];
2958 switch (flags
& ~WINED3DTTFF_PROJECTED
) {
2959 case WINED3DTTFF_COUNT1
: FIXME("WINED3DTTFF_PROJECTED with WINED3DTTFF_COUNT1?\n"); break;
2960 case WINED3DTTFF_COUNT2
: div_mask
= WINED3DSP_WRITEMASK_1
; break;
2961 case WINED3DTTFF_COUNT3
: div_mask
= WINED3DSP_WRITEMASK_2
; break;
2962 case WINED3DTTFF_COUNT4
:
2963 case WINED3DTTFF_DISABLE
: div_mask
= WINED3DSP_WRITEMASK_3
; break;
2965 shader_glsl_write_mask_to_str(div_mask
, coord_div_mask
);
2966 shader_addline(ins
->ctx
->buffer
, "T%u%s /= T%u%s;\n", sampler_idx
, coord_mask
, sampler_idx
, coord_div_mask
);
2969 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
, &coord_param
);
2971 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, WINED3DSP_NOSWIZZLE
, NULL
, NULL
, NULL
,
2972 "T%u%s + vec4(bumpenvmat%d * %s, 0.0, 0.0)%s", sampler_idx
, coord_mask
, sampler_idx
,
2973 coord_param
.param_str
, coord_mask
);
2975 if (ins
->handler_idx
== WINED3DSIH_TEXBEML
)
2977 glsl_src_param_t luminance_param
;
2978 glsl_dst_param_t dst_param
;
2980 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_2
, &luminance_param
);
2981 shader_glsl_add_dst_param(ins
, &ins
->dst
[0], &dst_param
);
2983 shader_addline(ins
->ctx
->buffer
, "%s%s *= (%s * luminancescale%d + luminanceoffset%d);\n",
2984 dst_param
.reg_name
, dst_param
.mask_str
,
2985 luminance_param
.param_str
, sampler_idx
, sampler_idx
);
2989 static void pshader_glsl_bem(const struct wined3d_shader_instruction
*ins
)
2991 glsl_src_param_t src0_param
, src1_param
;
2992 DWORD sampler_idx
= ins
->dst
[0].register_idx
;
2994 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
, &src0_param
);
2995 shader_glsl_add_src_param(ins
, &ins
->src
[1], WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
, &src1_param
);
2997 shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2998 shader_addline(ins
->ctx
->buffer
, "%s + bumpenvmat%d * %s);\n",
2999 src0_param
.param_str
, sampler_idx
, src1_param
.param_str
);
3002 /** Process the WINED3DSIO_TEXREG2AR instruction in GLSL
3003 * Sample 2D texture at dst using the alpha & red (wx) components of src as texture coordinates */
3004 static void pshader_glsl_texreg2ar(const struct wined3d_shader_instruction
*ins
)
3006 glsl_src_param_t src0_param
;
3007 DWORD sampler_idx
= ins
->dst
[0].register_idx
;
3008 WINED3DSAMPLER_TEXTURE_TYPE sampler_type
= ins
->ctx
->reg_maps
->sampler_type
[sampler_idx
];
3009 glsl_sample_function_t sample_function
;
3011 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_ALL
, &src0_param
);
3013 shader_glsl_get_sample_function(sampler_type
, 0, &sample_function
);
3014 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, WINED3DSP_NOSWIZZLE
, NULL
, NULL
, NULL
,
3015 "%s.wx", src0_param
.reg_name
);
3018 /** Process the WINED3DSIO_TEXREG2GB instruction in GLSL
3019 * Sample 2D texture at dst using the green & blue (yz) components of src as texture coordinates */
3020 static void pshader_glsl_texreg2gb(const struct wined3d_shader_instruction
*ins
)
3022 glsl_src_param_t src0_param
;
3023 DWORD sampler_idx
= ins
->dst
[0].register_idx
;
3024 WINED3DSAMPLER_TEXTURE_TYPE sampler_type
= ins
->ctx
->reg_maps
->sampler_type
[sampler_idx
];
3025 glsl_sample_function_t sample_function
;
3027 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_ALL
, &src0_param
);
3029 shader_glsl_get_sample_function(sampler_type
, 0, &sample_function
);
3030 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, WINED3DSP_NOSWIZZLE
, NULL
, NULL
, NULL
,
3031 "%s.yz", src0_param
.reg_name
);
3034 /** Process the WINED3DSIO_TEXREG2RGB instruction in GLSL
3035 * Sample texture at dst using the rgb (xyz) components of src as texture coordinates */
3036 static void pshader_glsl_texreg2rgb(const struct wined3d_shader_instruction
*ins
)
3038 glsl_src_param_t src0_param
;
3039 DWORD sampler_idx
= ins
->dst
[0].register_idx
;
3040 WINED3DSAMPLER_TEXTURE_TYPE sampler_type
= ins
->ctx
->reg_maps
->sampler_type
[sampler_idx
];
3041 glsl_sample_function_t sample_function
;
3043 /* Dependent read, not valid with conditional NP2 */
3044 shader_glsl_get_sample_function(sampler_type
, 0, &sample_function
);
3045 shader_glsl_add_src_param(ins
, &ins
->src
[0], sample_function
.coord_mask
, &src0_param
);
3047 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, WINED3DSP_NOSWIZZLE
, NULL
, NULL
, NULL
,
3048 "%s", src0_param
.param_str
);
3051 /** Process the WINED3DSIO_TEXKILL instruction in GLSL.
3052 * If any of the first 3 components are < 0, discard this pixel */
3053 static void pshader_glsl_texkill(const struct wined3d_shader_instruction
*ins
)
3055 glsl_dst_param_t dst_param
;
3057 /* The argument is a destination parameter, and no writemasks are allowed */
3058 shader_glsl_add_dst_param(ins
, &ins
->dst
[0], &dst_param
);
3059 if (ins
->ctx
->reg_maps
->shader_version
.major
>= 2)
3061 /* 2.0 shaders compare all 4 components in texkill */
3062 shader_addline(ins
->ctx
->buffer
, "if (any(lessThan(%s.xyzw, vec4(0.0)))) discard;\n", dst_param
.reg_name
);
3064 /* 1.X shaders only compare the first 3 components, probably due to the nature of the texkill
3065 * instruction as a tex* instruction, and phase, which kills all a / w components. Even if all
3066 * 4 components are defined, only the first 3 are used
3068 shader_addline(ins
->ctx
->buffer
, "if (any(lessThan(%s.xyz, vec3(0.0)))) discard;\n", dst_param
.reg_name
);
3072 /** Process the WINED3DSIO_DP2ADD instruction in GLSL.
3073 * dst = dot2(src0, src1) + src2 */
3074 static void pshader_glsl_dp2add(const struct wined3d_shader_instruction
*ins
)
3076 glsl_src_param_t src0_param
;
3077 glsl_src_param_t src1_param
;
3078 glsl_src_param_t src2_param
;
3080 unsigned int mask_size
;
3082 write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
3083 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
3085 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
, &src0_param
);
3086 shader_glsl_add_src_param(ins
, &ins
->src
[1], WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
, &src1_param
);
3087 shader_glsl_add_src_param(ins
, &ins
->src
[2], WINED3DSP_WRITEMASK_0
, &src2_param
);
3089 if (mask_size
> 1) {
3090 shader_addline(ins
->ctx
->buffer
, "vec%d(dot(%s, %s) + %s));\n",
3091 mask_size
, src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
3093 shader_addline(ins
->ctx
->buffer
, "dot(%s, %s) + %s);\n",
3094 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
3098 static void pshader_glsl_input_pack(IWineD3DPixelShader
*iface
, SHADER_BUFFER
*buffer
,
3099 const struct wined3d_shader_semantic
*semantics_in
, const struct shader_reg_maps
*reg_maps
,
3100 enum vertexprocessing_mode vertexprocessing
)
3103 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*)iface
;
3105 for (i
= 0; i
< MAX_REG_INPUT
; ++i
)
3107 DWORD usage
, usage_idx
;
3111 if (!reg_maps
->packed_input
[i
]) continue;
3113 usage
= semantics_in
[i
].usage
;
3114 usage_idx
= semantics_in
[i
].usage_idx
;
3115 shader_glsl_get_write_mask(&semantics_in
[i
].reg
, reg_mask
);
3119 case WINED3DDECLUSAGE_TEXCOORD
:
3120 if (usage_idx
< 8 && vertexprocessing
== pretransformed
)
3121 shader_addline(buffer
, "IN[%u]%s = gl_TexCoord[%u]%s;\n",
3122 This
->input_reg_map
[i
], reg_mask
, usage_idx
, reg_mask
);
3124 shader_addline(buffer
, "IN[%u]%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
3125 This
->input_reg_map
[i
], reg_mask
, reg_mask
);
3128 case WINED3DDECLUSAGE_COLOR
:
3130 shader_addline(buffer
, "IN[%u]%s = vec4(gl_Color)%s;\n",
3131 This
->input_reg_map
[i
], reg_mask
, reg_mask
);
3132 else if (usage_idx
== 1)
3133 shader_addline(buffer
, "IN[%u]%s = vec4(gl_SecondaryColor)%s;\n",
3134 This
->input_reg_map
[i
], reg_mask
, reg_mask
);
3136 shader_addline(buffer
, "IN[%u]%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
3137 This
->input_reg_map
[i
], reg_mask
, reg_mask
);
3141 shader_addline(buffer
, "IN[%u]%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
3142 This
->input_reg_map
[i
], reg_mask
, reg_mask
);
3148 /*********************************************
3149 * Vertex Shader Specific Code begins here
3150 ********************************************/
3152 static void add_glsl_program_entry(struct shader_glsl_priv
*priv
, struct glsl_shader_prog_link
*entry
) {
3153 glsl_program_key_t
*key
;
3155 key
= HeapAlloc(GetProcessHeap(), 0, sizeof(glsl_program_key_t
));
3156 key
->vshader
= entry
->vshader
;
3157 key
->pshader
= entry
->pshader
;
3158 key
->vs_args
= entry
->vs_args
;
3159 key
->ps_args
= entry
->ps_args
;
3161 hash_table_put(priv
->glsl_program_lookup
, key
, entry
);
3164 static struct glsl_shader_prog_link
*get_glsl_program_entry(struct shader_glsl_priv
*priv
,
3165 IWineD3DVertexShader
*vshader
, IWineD3DPixelShader
*pshader
, struct vs_compile_args
*vs_args
,
3166 struct ps_compile_args
*ps_args
) {
3167 glsl_program_key_t key
;
3169 key
.vshader
= vshader
;
3170 key
.pshader
= pshader
;
3171 key
.vs_args
= *vs_args
;
3172 key
.ps_args
= *ps_args
;
3174 return hash_table_get(priv
->glsl_program_lookup
, &key
);
3177 static void delete_glsl_program_entry(struct shader_glsl_priv
*priv
, const WineD3D_GL_Info
*gl_info
,
3178 struct glsl_shader_prog_link
*entry
)
3180 glsl_program_key_t
*key
;
3182 key
= HeapAlloc(GetProcessHeap(), 0, sizeof(glsl_program_key_t
));
3183 key
->vshader
= entry
->vshader
;
3184 key
->pshader
= entry
->pshader
;
3185 key
->vs_args
= entry
->vs_args
;
3186 key
->ps_args
= entry
->ps_args
;
3187 hash_table_remove(priv
->glsl_program_lookup
, key
);
3189 GL_EXTCALL(glDeleteObjectARB(entry
->programId
));
3190 if (entry
->vshader
) list_remove(&entry
->vshader_entry
);
3191 if (entry
->pshader
) list_remove(&entry
->pshader_entry
);
3192 HeapFree(GetProcessHeap(), 0, entry
->vuniformF_locations
);
3193 HeapFree(GetProcessHeap(), 0, entry
->puniformF_locations
);
3194 HeapFree(GetProcessHeap(), 0, entry
);
3197 static void handle_ps3_input(SHADER_BUFFER
*buffer
, const WineD3D_GL_Info
*gl_info
, const DWORD
*map
,
3198 const struct wined3d_shader_semantic
*semantics_in
, const struct shader_reg_maps
*reg_maps_in
,
3199 const struct wined3d_shader_semantic
*semantics_out
, const struct shader_reg_maps
*reg_maps_out
)
3202 DWORD usage
, usage_idx
, usage_out
, usage_idx_out
;
3205 DWORD in_count
= GL_LIMITS(glsl_varyings
) / 4;
3206 char reg_mask
[6], reg_mask_out
[6];
3207 char destination
[50];
3209 set
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*set
) * (in_count
+ 2));
3211 if (!semantics_out
) {
3212 /* Save gl_FrontColor & gl_FrontSecondaryColor before overwriting them. */
3213 shader_addline(buffer
, "vec4 front_color = gl_FrontColor;\n");
3214 shader_addline(buffer
, "vec4 front_secondary_color = gl_FrontSecondaryColor;\n");
3217 for(i
= 0; i
< MAX_REG_INPUT
; i
++) {
3218 if (!reg_maps_in
->packed_input
[i
]) continue;
3221 if (in_idx
>= (in_count
+ 2)) {
3222 FIXME("More input varyings declared than supported, expect issues\n");
3225 else if (map
[i
] == ~0U)
3227 /* Declared, but not read register */
3231 if (in_idx
== in_count
) {
3232 sprintf(destination
, "gl_FrontColor");
3233 } else if (in_idx
== in_count
+ 1) {
3234 sprintf(destination
, "gl_FrontSecondaryColor");
3236 sprintf(destination
, "IN[%u]", in_idx
);
3239 usage
= semantics_in
[i
].usage
;
3240 usage_idx
= semantics_in
[i
].usage_idx
;
3241 set
[map
[i
]] = shader_glsl_get_write_mask(&semantics_in
[i
].reg
, reg_mask
);
3243 if(!semantics_out
) {
3245 case WINED3DDECLUSAGE_COLOR
:
3247 shader_addline(buffer
, "%s%s = front_color%s;\n",
3248 destination
, reg_mask
, reg_mask
);
3249 else if (usage_idx
== 1)
3250 shader_addline(buffer
, "%s%s = front_secondary_color%s;\n",
3251 destination
, reg_mask
, reg_mask
);
3253 shader_addline(buffer
, "%s%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
3254 destination
, reg_mask
, reg_mask
);
3257 case WINED3DDECLUSAGE_TEXCOORD
:
3258 if (usage_idx
< 8) {
3259 shader_addline(buffer
, "%s%s = gl_TexCoord[%u]%s;\n",
3260 destination
, reg_mask
, usage_idx
, reg_mask
);
3262 shader_addline(buffer
, "%s%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
3263 destination
, reg_mask
, reg_mask
);
3267 case WINED3DDECLUSAGE_FOG
:
3268 shader_addline(buffer
, "%s%s = vec4(gl_FogFragCoord, 0.0, 0.0, 0.0)%s;\n",
3269 destination
, reg_mask
, reg_mask
);
3273 shader_addline(buffer
, "%s%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
3274 destination
, reg_mask
, reg_mask
);
3278 for(j
= 0; j
< MAX_REG_OUTPUT
; j
++) {
3279 if (!reg_maps_out
->packed_output
[j
]) continue;
3281 usage_out
= semantics_out
[j
].usage
;
3282 usage_idx_out
= semantics_out
[j
].usage_idx
;
3283 shader_glsl_get_write_mask(&semantics_out
[j
].reg
, reg_mask_out
);
3285 if(usage
== usage_out
&&
3286 usage_idx
== usage_idx_out
) {
3287 shader_addline(buffer
, "%s%s = OUT[%u]%s;\n",
3288 destination
, reg_mask
, j
, reg_mask
);
3293 shader_addline(buffer
, "%s%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
3294 destination
, reg_mask
, reg_mask
);
3299 /* This is solely to make the compiler / linker happy and avoid warning about undefined
3300 * varyings. It shouldn't result in any real code executed on the GPU, since all read
3301 * input varyings are assigned above, if the optimizer works properly.
3303 for(i
= 0; i
< in_count
+ 2; i
++) {
3304 if(set
[i
] != WINED3DSP_WRITEMASK_ALL
) {
3305 unsigned int size
= 0;
3306 memset(reg_mask
, 0, sizeof(reg_mask
));
3307 if(!(set
[i
] & WINED3DSP_WRITEMASK_0
)) {
3308 reg_mask
[size
] = 'x';
3311 if(!(set
[i
] & WINED3DSP_WRITEMASK_1
)) {
3312 reg_mask
[size
] = 'y';
3315 if(!(set
[i
] & WINED3DSP_WRITEMASK_2
)) {
3316 reg_mask
[size
] = 'z';
3319 if(!(set
[i
] & WINED3DSP_WRITEMASK_3
)) {
3320 reg_mask
[size
] = 'w';
3324 if (i
== in_count
) {
3325 sprintf(destination
, "gl_FrontColor");
3326 } else if (i
== in_count
+ 1) {
3327 sprintf(destination
, "gl_FrontSecondaryColor");
3329 sprintf(destination
, "IN[%u]", i
);
3333 shader_addline(buffer
, "%s.%s = 0.0;\n", destination
, reg_mask
);
3335 shader_addline(buffer
, "%s.%s = vec%u(0.0);\n", destination
, reg_mask
, size
);
3340 HeapFree(GetProcessHeap(), 0, set
);
3343 static GLhandleARB
generate_param_reorder_function(IWineD3DVertexShader
*vertexshader
,
3344 IWineD3DPixelShader
*pixelshader
, const WineD3D_GL_Info
*gl_info
)
3346 GLhandleARB ret
= 0;
3347 IWineD3DVertexShaderImpl
*vs
= (IWineD3DVertexShaderImpl
*) vertexshader
;
3348 IWineD3DPixelShaderImpl
*ps
= (IWineD3DPixelShaderImpl
*) pixelshader
;
3349 IWineD3DDeviceImpl
*device
;
3350 DWORD vs_major
= vs
->baseShader
.reg_maps
.shader_version
.major
;
3351 DWORD ps_major
= ps
? ps
->baseShader
.reg_maps
.shader_version
.major
: 0;
3353 SHADER_BUFFER buffer
;
3354 DWORD usage
, usage_idx
, writemask
;
3356 const struct wined3d_shader_semantic
*semantics_out
;
3358 shader_buffer_init(&buffer
);
3360 shader_addline(&buffer
, "#version 120\n");
3362 if(vs_major
< 3 && ps_major
< 3) {
3363 /* That one is easy: The vertex shader writes to the builtin varyings, the pixel shader reads from them.
3364 * Take care about the texcoord .w fixup though if we're using the fixed function fragment pipeline
3366 device
= (IWineD3DDeviceImpl
*) vs
->baseShader
.device
;
3367 if((GLINFO_LOCATION
).set_texcoord_w
&& ps_major
== 0 && vs_major
> 0 &&
3368 !device
->frag_pipe
->ffp_proj_control
) {
3369 shader_addline(&buffer
, "void order_ps_input() {\n");
3370 for(i
= 0; i
< min(8, MAX_REG_TEXCRD
); i
++) {
3371 if(vs
->baseShader
.reg_maps
.texcoord_mask
[i
] != 0 &&
3372 vs
->baseShader
.reg_maps
.texcoord_mask
[i
] != WINED3DSP_WRITEMASK_ALL
) {
3373 shader_addline(&buffer
, "gl_TexCoord[%u].w = 1.0;\n", i
);
3376 shader_addline(&buffer
, "}\n");
3378 shader_addline(&buffer
, "void order_ps_input() { /* do nothing */ }\n");
3380 } else if(ps_major
< 3 && vs_major
>= 3) {
3381 /* The vertex shader writes to its own varyings, the pixel shader needs them in the builtin ones */
3382 semantics_out
= vs
->semantics_out
;
3384 shader_addline(&buffer
, "void order_ps_input(in vec4 OUT[%u]) {\n", MAX_REG_OUTPUT
);
3385 for(i
= 0; i
< MAX_REG_OUTPUT
; i
++) {
3386 if (!vs
->baseShader
.reg_maps
.packed_output
[i
]) continue;
3388 usage
= semantics_out
[i
].usage
;
3389 usage_idx
= semantics_out
[i
].usage_idx
;
3390 writemask
= shader_glsl_get_write_mask(&semantics_out
[i
].reg
, reg_mask
);
3393 case WINED3DDECLUSAGE_COLOR
:
3395 shader_addline(&buffer
, "gl_FrontColor%s = OUT[%u]%s;\n", reg_mask
, i
, reg_mask
);
3396 else if (usage_idx
== 1)
3397 shader_addline(&buffer
, "gl_FrontSecondaryColor%s = OUT[%u]%s;\n", reg_mask
, i
, reg_mask
);
3400 case WINED3DDECLUSAGE_POSITION
:
3401 shader_addline(&buffer
, "gl_Position%s = OUT[%u]%s;\n", reg_mask
, i
, reg_mask
);
3404 case WINED3DDECLUSAGE_TEXCOORD
:
3405 if (usage_idx
< 8) {
3406 if(!(GLINFO_LOCATION
).set_texcoord_w
|| ps_major
> 0) writemask
|= WINED3DSP_WRITEMASK_3
;
3408 shader_addline(&buffer
, "gl_TexCoord[%u]%s = OUT[%u]%s;\n",
3409 usage_idx
, reg_mask
, i
, reg_mask
);
3410 if(!(writemask
& WINED3DSP_WRITEMASK_3
)) {
3411 shader_addline(&buffer
, "gl_TexCoord[%u].w = 1.0;\n", usage_idx
);
3416 case WINED3DDECLUSAGE_PSIZE
:
3417 shader_addline(&buffer
, "gl_PointSize = OUT[%u].x;\n", i
);
3420 case WINED3DDECLUSAGE_FOG
:
3421 shader_addline(&buffer
, "gl_FogFragCoord = OUT[%u].%c;\n", i
, reg_mask
[1]);
3428 shader_addline(&buffer
, "}\n");
3430 } else if(ps_major
>= 3 && vs_major
>= 3) {
3431 semantics_out
= vs
->semantics_out
;
3433 /* This one is tricky: a 3.0 pixel shader reads from a 3.0 vertex shader */
3434 shader_addline(&buffer
, "varying vec4 IN[%u];\n", GL_LIMITS(glsl_varyings
) / 4);
3435 shader_addline(&buffer
, "void order_ps_input(in vec4 OUT[%u]) {\n", MAX_REG_OUTPUT
);
3437 /* First, sort out position and point size. Those are not passed to the pixel shader */
3438 for(i
= 0; i
< MAX_REG_OUTPUT
; i
++) {
3439 if (!vs
->baseShader
.reg_maps
.packed_output
[i
]) continue;
3441 usage
= semantics_out
[i
].usage
;
3442 usage_idx
= semantics_out
[i
].usage_idx
;
3443 shader_glsl_get_write_mask(&semantics_out
[i
].reg
, reg_mask
);
3446 case WINED3DDECLUSAGE_POSITION
:
3447 shader_addline(&buffer
, "gl_Position%s = OUT[%u]%s;\n", reg_mask
, i
, reg_mask
);
3450 case WINED3DDECLUSAGE_PSIZE
:
3451 shader_addline(&buffer
, "gl_PointSize = OUT[%u].x;\n", i
);
3459 /* Then, fix the pixel shader input */
3460 handle_ps3_input(&buffer
, gl_info
, ps
->input_reg_map
,
3461 ps
->semantics_in
, &ps
->baseShader
.reg_maps
, semantics_out
, &vs
->baseShader
.reg_maps
);
3463 shader_addline(&buffer
, "}\n");
3464 } else if(ps_major
>= 3 && vs_major
< 3) {
3465 shader_addline(&buffer
, "varying vec4 IN[%u];\n", GL_LIMITS(glsl_varyings
) / 4);
3466 shader_addline(&buffer
, "void order_ps_input() {\n");
3467 /* The vertex shader wrote to the builtin varyings. There is no need to figure out position and
3468 * point size, but we depend on the optimizers kindness to find out that the pixel shader doesn't
3469 * read gl_TexCoord and gl_ColorX, otherwise we'll run out of varyings
3471 handle_ps3_input(&buffer
, gl_info
, ps
->input_reg_map
, ps
->semantics_in
, &ps
->baseShader
.reg_maps
, NULL
, NULL
);
3472 shader_addline(&buffer
, "}\n");
3474 ERR("Unexpected vertex and pixel shader version condition: vs: %d, ps: %d\n", vs_major
, ps_major
);
3477 ret
= GL_EXTCALL(glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB
));
3478 checkGLcall("glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB)");
3479 GL_EXTCALL(glShaderSourceARB(ret
, 1, (const char**)&buffer
.buffer
, NULL
));
3480 checkGLcall("glShaderSourceARB(ret, 1, &buffer.buffer, NULL)");
3481 GL_EXTCALL(glCompileShaderARB(ret
));
3482 checkGLcall("glCompileShaderARB(ret)");
3484 shader_buffer_free(&buffer
);
3488 static void hardcode_local_constants(IWineD3DBaseShaderImpl
*shader
, const WineD3D_GL_Info
*gl_info
,
3489 GLhandleARB programId
, char prefix
)
3491 const local_constant
*lconst
;
3496 LIST_FOR_EACH_ENTRY(lconst
, &shader
->baseShader
.constantsF
, local_constant
, entry
) {
3497 value
= (const float *)lconst
->value
;
3498 snprintf(glsl_name
, sizeof(glsl_name
), "%cLC%u", prefix
, lconst
->idx
);
3499 tmp_loc
= GL_EXTCALL(glGetUniformLocationARB(programId
, glsl_name
));
3500 GL_EXTCALL(glUniform4fvARB(tmp_loc
, 1, value
));
3502 checkGLcall("Hardcoding local constants\n");
3505 /** Sets the GLSL program ID for the given pixel and vertex shader combination.
3506 * It sets the programId on the current StateBlock (because it should be called
3507 * inside of the DrawPrimitive() part of the render loop).
3509 * If a program for the given combination does not exist, create one, and store
3510 * the program in the hash table. If it creates a program, it will link the
3511 * given objects, too.
3513 static void set_glsl_shader_program(IWineD3DDevice
*iface
, BOOL use_ps
, BOOL use_vs
) {
3514 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
3515 struct shader_glsl_priv
*priv
= This
->shader_priv
;
3516 const WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
3517 IWineD3DPixelShader
*pshader
= This
->stateBlock
->pixelShader
;
3518 IWineD3DVertexShader
*vshader
= This
->stateBlock
->vertexShader
;
3519 struct glsl_shader_prog_link
*entry
= NULL
;
3520 GLhandleARB programId
= 0;
3521 GLhandleARB reorder_shader_id
= 0;
3524 GLhandleARB vshader_id
, pshader_id
;
3525 struct ps_compile_args ps_compile_args
;
3526 struct vs_compile_args vs_compile_args
;
3529 find_vs_compile_args((IWineD3DVertexShaderImpl
*)This
->stateBlock
->vertexShader
, This
->stateBlock
, &vs_compile_args
);
3531 /* FIXME: Do we really have to spend CPU cycles to generate a few zeroed bytes? */
3532 memset(&vs_compile_args
, 0, sizeof(vs_compile_args
));
3535 find_ps_compile_args((IWineD3DPixelShaderImpl
*)This
->stateBlock
->pixelShader
, This
->stateBlock
, &ps_compile_args
);
3537 /* FIXME: Do we really have to spend CPU cycles to generate a few zeroed bytes? */
3538 memset(&ps_compile_args
, 0, sizeof(ps_compile_args
));
3540 entry
= get_glsl_program_entry(priv
, vshader
, pshader
, &vs_compile_args
, &ps_compile_args
);
3542 priv
->glsl_program
= entry
;
3546 /* If we get to this point, then no matching program exists, so we create one */
3547 programId
= GL_EXTCALL(glCreateProgramObjectARB());
3548 TRACE("Created new GLSL shader program %u\n", programId
);
3550 /* Create the entry */
3551 entry
= HeapAlloc(GetProcessHeap(), 0, sizeof(struct glsl_shader_prog_link
));
3552 entry
->programId
= programId
;
3553 entry
->vshader
= vshader
;
3554 entry
->pshader
= pshader
;
3555 entry
->vs_args
= vs_compile_args
;
3556 entry
->ps_args
= ps_compile_args
;
3557 entry
->constant_version
= 0;
3558 /* Add the hash table entry */
3559 add_glsl_program_entry(priv
, entry
);
3561 /* Set the current program */
3562 priv
->glsl_program
= entry
;
3565 vshader_id
= find_gl_vshader((IWineD3DVertexShaderImpl
*) vshader
, &vs_compile_args
);
3570 /* Attach GLSL vshader */
3572 const unsigned int max_attribs
= 16; /* TODO: Will this always be the case? It is at the moment... */
3575 reorder_shader_id
= generate_param_reorder_function(vshader
, pshader
, gl_info
);
3576 TRACE("Attaching GLSL shader object %u to program %u\n", reorder_shader_id
, programId
);
3577 GL_EXTCALL(glAttachObjectARB(programId
, reorder_shader_id
));
3578 checkGLcall("glAttachObjectARB");
3579 /* Flag the reorder function for deletion, then it will be freed automatically when the program
3582 GL_EXTCALL(glDeleteObjectARB(reorder_shader_id
));
3584 TRACE("Attaching GLSL shader object %u to program %u\n", vshader_id
, programId
);
3585 GL_EXTCALL(glAttachObjectARB(programId
, vshader_id
));
3586 checkGLcall("glAttachObjectARB");
3588 /* Bind vertex attributes to a corresponding index number to match
3589 * the same index numbers as ARB_vertex_programs (makes loading
3590 * vertex attributes simpler). With this method, we can use the
3591 * exact same code to load the attributes later for both ARB and
3594 * We have to do this here because we need to know the Program ID
3595 * in order to make the bindings work, and it has to be done prior
3596 * to linking the GLSL program. */
3597 for (i
= 0; i
< max_attribs
; ++i
) {
3598 if (((IWineD3DBaseShaderImpl
*)vshader
)->baseShader
.reg_maps
.attributes
[i
]) {
3599 snprintf(tmp_name
, sizeof(tmp_name
), "attrib%i", i
);
3600 GL_EXTCALL(glBindAttribLocationARB(programId
, i
, tmp_name
));
3603 checkGLcall("glBindAttribLocationARB");
3605 list_add_head(&((IWineD3DBaseShaderImpl
*)vshader
)->baseShader
.linked_programs
, &entry
->vshader_entry
);
3609 pshader_id
= find_gl_pshader((IWineD3DPixelShaderImpl
*) pshader
, &ps_compile_args
);
3614 /* Attach GLSL pshader */
3616 TRACE("Attaching GLSL shader object %u to program %u\n", pshader_id
, programId
);
3617 GL_EXTCALL(glAttachObjectARB(programId
, pshader_id
));
3618 checkGLcall("glAttachObjectARB");
3620 list_add_head(&((IWineD3DBaseShaderImpl
*)pshader
)->baseShader
.linked_programs
, &entry
->pshader_entry
);
3623 /* Link the program */
3624 TRACE("Linking GLSL shader program %u\n", programId
);
3625 GL_EXTCALL(glLinkProgramARB(programId
));
3626 print_glsl_info_log(&GLINFO_LOCATION
, programId
);
3628 entry
->vuniformF_locations
= HeapAlloc(GetProcessHeap(), 0, sizeof(GLhandleARB
) * GL_LIMITS(vshader_constantsF
));
3629 for (i
= 0; i
< GL_LIMITS(vshader_constantsF
); ++i
) {
3630 snprintf(glsl_name
, sizeof(glsl_name
), "VC[%i]", i
);
3631 entry
->vuniformF_locations
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, glsl_name
));
3633 for (i
= 0; i
< MAX_CONST_I
; ++i
) {
3634 snprintf(glsl_name
, sizeof(glsl_name
), "VI[%i]", i
);
3635 entry
->vuniformI_locations
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, glsl_name
));
3637 entry
->puniformF_locations
= HeapAlloc(GetProcessHeap(), 0, sizeof(GLhandleARB
) * GL_LIMITS(pshader_constantsF
));
3638 for (i
= 0; i
< GL_LIMITS(pshader_constantsF
); ++i
) {
3639 snprintf(glsl_name
, sizeof(glsl_name
), "PC[%i]", i
);
3640 entry
->puniformF_locations
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, glsl_name
));
3642 for (i
= 0; i
< MAX_CONST_I
; ++i
) {
3643 snprintf(glsl_name
, sizeof(glsl_name
), "PI[%i]", i
);
3644 entry
->puniformI_locations
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, glsl_name
));
3648 for(i
= 0; i
< ((IWineD3DPixelShaderImpl
*)pshader
)->numbumpenvmatconsts
; i
++) {
3650 sprintf(name
, "bumpenvmat%d", ((IWineD3DPixelShaderImpl
*)pshader
)->bumpenvmatconst
[i
].texunit
);
3651 entry
->bumpenvmat_location
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, name
));
3652 sprintf(name
, "luminancescale%d", ((IWineD3DPixelShaderImpl
*)pshader
)->luminanceconst
[i
].texunit
);
3653 entry
->luminancescale_location
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, name
));
3654 sprintf(name
, "luminanceoffset%d", ((IWineD3DPixelShaderImpl
*)pshader
)->luminanceconst
[i
].texunit
);
3655 entry
->luminanceoffset_location
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, name
));
3659 if (use_ps
&& ps_compile_args
.np2_fixup
) {
3661 for (i
= 0; i
< MAX_FRAGMENT_SAMPLERS
; ++i
) {
3662 if (ps_compile_args
.np2_fixup
& (1 << i
)) {
3663 sprintf(name
, "PsamplerNP2Fixup%u", i
);
3664 entry
->np2Fixup_location
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, name
));
3666 entry
->np2Fixup_location
[i
] = -1;
3671 entry
->posFixup_location
= GL_EXTCALL(glGetUniformLocationARB(programId
, "posFixup"));
3672 entry
->ycorrection_location
= GL_EXTCALL(glGetUniformLocationARB(programId
, "ycorrection"));
3673 checkGLcall("Find glsl program uniform locations");
3676 && ((IWineD3DPixelShaderImpl
*)pshader
)->baseShader
.reg_maps
.shader_version
.major
>= 3
3677 && ((IWineD3DPixelShaderImpl
*)pshader
)->declared_in_count
> GL_LIMITS(glsl_varyings
) / 4)
3679 TRACE("Shader %d needs vertex color clamping disabled\n", programId
);
3680 entry
->vertex_color_clamp
= GL_FALSE
;
3682 entry
->vertex_color_clamp
= GL_FIXED_ONLY_ARB
;
3685 /* Set the shader to allow uniform loading on it */
3686 GL_EXTCALL(glUseProgramObjectARB(programId
));
3687 checkGLcall("glUseProgramObjectARB(programId)");
3689 /* Load the vertex and pixel samplers now. The function that finds the mappings makes sure
3690 * that it stays the same for each vertexshader-pixelshader pair(=linked glsl program). If
3691 * a pshader with fixed function pipeline is used there are no vertex samplers, and if a
3692 * vertex shader with fixed function pixel processing is used we make sure that the card
3693 * supports enough samplers to allow the max number of vertex samplers with all possible
3694 * fixed function fragment processing setups. So once the program is linked these samplers
3698 /* Load vertex shader samplers */
3699 shader_glsl_load_vsamplers(gl_info
, This
->texUnitMap
, programId
);
3702 /* Load pixel shader samplers */
3703 shader_glsl_load_psamplers(gl_info
, This
->texUnitMap
, programId
);
3706 /* If the local constants do not have to be loaded with the environment constants,
3707 * load them now to have them hardcoded in the GLSL program. This saves some CPU cycles
3710 if(pshader
&& !((IWineD3DPixelShaderImpl
*)pshader
)->baseShader
.load_local_constsF
) {
3711 hardcode_local_constants((IWineD3DBaseShaderImpl
*) pshader
, gl_info
, programId
, 'P');
3713 if(vshader
&& !((IWineD3DVertexShaderImpl
*)vshader
)->baseShader
.load_local_constsF
) {
3714 hardcode_local_constants((IWineD3DBaseShaderImpl
*) vshader
, gl_info
, programId
, 'V');
3718 static GLhandleARB
create_glsl_blt_shader(const WineD3D_GL_Info
*gl_info
, enum tex_types tex_type
)
3720 GLhandleARB program_id
;
3721 GLhandleARB vshader_id
, pshader_id
;
3722 static const char *blt_vshader
[] =
3727 " gl_Position = gl_Vertex;\n"
3728 " gl_FrontColor = vec4(1.0);\n"
3729 " gl_TexCoord[0] = gl_MultiTexCoord0;\n"
3733 static const char *blt_pshaders
[tex_type_count
] =
3739 "uniform sampler2D sampler;\n"
3742 " gl_FragDepth = texture2D(sampler, gl_TexCoord[0].xy).x;\n"
3748 "uniform samplerCube sampler;\n"
3751 " gl_FragDepth = textureCube(sampler, gl_TexCoord[0].xyz).x;\n"
3755 "#extension GL_ARB_texture_rectangle : enable\n"
3756 "uniform sampler2DRect sampler;\n"
3759 " gl_FragDepth = texture2DRect(sampler, gl_TexCoord[0].xy).x;\n"
3763 if (!blt_pshaders
[tex_type
])
3765 FIXME("tex_type %#x not supported\n", tex_type
);
3769 vshader_id
= GL_EXTCALL(glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB
));
3770 GL_EXTCALL(glShaderSourceARB(vshader_id
, 1, blt_vshader
, NULL
));
3771 GL_EXTCALL(glCompileShaderARB(vshader_id
));
3773 pshader_id
= GL_EXTCALL(glCreateShaderObjectARB(GL_FRAGMENT_SHADER_ARB
));
3774 GL_EXTCALL(glShaderSourceARB(pshader_id
, 1, &blt_pshaders
[tex_type
], NULL
));
3775 GL_EXTCALL(glCompileShaderARB(pshader_id
));
3777 program_id
= GL_EXTCALL(glCreateProgramObjectARB());
3778 GL_EXTCALL(glAttachObjectARB(program_id
, vshader_id
));
3779 GL_EXTCALL(glAttachObjectARB(program_id
, pshader_id
));
3780 GL_EXTCALL(glLinkProgramARB(program_id
));
3782 print_glsl_info_log(&GLINFO_LOCATION
, program_id
);
3784 /* Once linked we can mark the shaders for deletion. They will be deleted once the program
3787 GL_EXTCALL(glDeleteObjectARB(vshader_id
));
3788 GL_EXTCALL(glDeleteObjectARB(pshader_id
));
3792 static void shader_glsl_select(IWineD3DDevice
*iface
, BOOL usePS
, BOOL useVS
) {
3793 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
3794 struct shader_glsl_priv
*priv
= This
->shader_priv
;
3795 const WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
3796 GLhandleARB program_id
= 0;
3797 GLenum old_vertex_color_clamp
, current_vertex_color_clamp
;
3799 old_vertex_color_clamp
= priv
->glsl_program
? priv
->glsl_program
->vertex_color_clamp
: GL_FIXED_ONLY_ARB
;
3801 if (useVS
|| usePS
) set_glsl_shader_program(iface
, usePS
, useVS
);
3802 else priv
->glsl_program
= NULL
;
3804 current_vertex_color_clamp
= priv
->glsl_program
? priv
->glsl_program
->vertex_color_clamp
: GL_FIXED_ONLY_ARB
;
3806 if (old_vertex_color_clamp
!= current_vertex_color_clamp
) {
3807 if (GL_SUPPORT(ARB_COLOR_BUFFER_FLOAT
)) {
3808 GL_EXTCALL(glClampColorARB(GL_CLAMP_VERTEX_COLOR_ARB
, current_vertex_color_clamp
));
3809 checkGLcall("glClampColorARB");
3811 FIXME("vertex color clamp needs to be changed, but extension not supported.\n");
3815 program_id
= priv
->glsl_program
? priv
->glsl_program
->programId
: 0;
3816 if (program_id
) TRACE("Using GLSL program %u\n", program_id
);
3817 GL_EXTCALL(glUseProgramObjectARB(program_id
));
3818 checkGLcall("glUseProgramObjectARB");
3821 static void shader_glsl_select_depth_blt(IWineD3DDevice
*iface
, enum tex_types tex_type
) {
3822 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
3823 const WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
3824 struct shader_glsl_priv
*priv
= This
->shader_priv
;
3825 GLhandleARB
*blt_program
= &priv
->depth_blt_program
[tex_type
];
3827 if (!*blt_program
) {
3829 *blt_program
= create_glsl_blt_shader(gl_info
, tex_type
);
3830 loc
= GL_EXTCALL(glGetUniformLocationARB(*blt_program
, "sampler"));
3831 GL_EXTCALL(glUseProgramObjectARB(*blt_program
));
3832 GL_EXTCALL(glUniform1iARB(loc
, 0));
3834 GL_EXTCALL(glUseProgramObjectARB(*blt_program
));
3838 static void shader_glsl_deselect_depth_blt(IWineD3DDevice
*iface
) {
3839 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
3840 const WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
3841 struct shader_glsl_priv
*priv
= This
->shader_priv
;
3842 GLhandleARB program_id
;
3844 program_id
= priv
->glsl_program
? priv
->glsl_program
->programId
: 0;
3845 if (program_id
) TRACE("Using GLSL program %u\n", program_id
);
3847 GL_EXTCALL(glUseProgramObjectARB(program_id
));
3848 checkGLcall("glUseProgramObjectARB");
3851 static void shader_glsl_destroy(IWineD3DBaseShader
*iface
) {
3852 const struct list
*linked_programs
;
3853 IWineD3DBaseShaderImpl
*This
= (IWineD3DBaseShaderImpl
*) iface
;
3854 IWineD3DDeviceImpl
*device
= (IWineD3DDeviceImpl
*)This
->baseShader
.device
;
3855 struct shader_glsl_priv
*priv
= device
->shader_priv
;
3856 const WineD3D_GL_Info
*gl_info
= &device
->adapter
->gl_info
;
3857 IWineD3DPixelShaderImpl
*ps
= NULL
;
3858 IWineD3DVertexShaderImpl
*vs
= NULL
;
3860 /* Note: Do not use QueryInterface here to find out which shader type this is because this code
3861 * can be called from IWineD3DBaseShader::Release
3863 char pshader
= shader_is_pshader_version(This
->baseShader
.reg_maps
.shader_version
.type
);
3866 ps
= (IWineD3DPixelShaderImpl
*) This
;
3867 if(ps
->num_gl_shaders
== 0) return;
3868 if (priv
->glsl_program
&& (IWineD3DBaseShader
*)priv
->glsl_program
->pshader
== iface
)
3869 shader_glsl_select(This
->baseShader
.device
, FALSE
, FALSE
);
3871 vs
= (IWineD3DVertexShaderImpl
*) This
;
3872 if(vs
->num_gl_shaders
== 0) return;
3873 if (priv
->glsl_program
&& (IWineD3DBaseShader
*)priv
->glsl_program
->vshader
== iface
)
3874 shader_glsl_select(This
->baseShader
.device
, FALSE
, FALSE
);
3877 linked_programs
= &This
->baseShader
.linked_programs
;
3879 TRACE("Deleting linked programs\n");
3880 if (linked_programs
->next
) {
3881 struct glsl_shader_prog_link
*entry
, *entry2
;
3884 LIST_FOR_EACH_ENTRY_SAFE(entry
, entry2
, linked_programs
, struct glsl_shader_prog_link
, pshader_entry
) {
3885 delete_glsl_program_entry(priv
, gl_info
, entry
);
3888 LIST_FOR_EACH_ENTRY_SAFE(entry
, entry2
, linked_programs
, struct glsl_shader_prog_link
, vshader_entry
) {
3889 delete_glsl_program_entry(priv
, gl_info
, entry
);
3898 for(i
= 0; i
< ps
->num_gl_shaders
; i
++) {
3899 TRACE("deleting pshader %u\n", ps
->gl_shaders
[i
].prgId
);
3900 GL_EXTCALL(glDeleteObjectARB(ps
->gl_shaders
[i
].prgId
));
3901 checkGLcall("glDeleteObjectARB");
3904 HeapFree(GetProcessHeap(), 0, ps
->gl_shaders
);
3905 ps
->gl_shaders
= NULL
;
3906 ps
->num_gl_shaders
= 0;
3907 ps
->shader_array_size
= 0;
3912 for(i
= 0; i
< vs
->num_gl_shaders
; i
++) {
3913 TRACE("deleting vshader %u\n", vs
->gl_shaders
[i
].prgId
);
3914 GL_EXTCALL(glDeleteObjectARB(vs
->gl_shaders
[i
].prgId
));
3915 checkGLcall("glDeleteObjectARB");
3918 HeapFree(GetProcessHeap(), 0, vs
->gl_shaders
);
3919 vs
->gl_shaders
= NULL
;
3920 vs
->num_gl_shaders
= 0;
3921 vs
->shader_array_size
= 0;
3925 static unsigned int glsl_program_key_hash(const void *key
)
3927 const glsl_program_key_t
*k
= key
;
3929 unsigned int hash
= ((DWORD_PTR
) k
->vshader
) | ((DWORD_PTR
) k
->pshader
) << 16;
3930 hash
+= ~(hash
<< 15);
3931 hash
^= (hash
>> 10);
3932 hash
+= (hash
<< 3);
3933 hash
^= (hash
>> 6);
3934 hash
+= ~(hash
<< 11);
3935 hash
^= (hash
>> 16);
3940 static BOOL
glsl_program_key_compare(const void *keya
, const void *keyb
)
3942 const glsl_program_key_t
*ka
= keya
;
3943 const glsl_program_key_t
*kb
= keyb
;
3945 return ka
->vshader
== kb
->vshader
&& ka
->pshader
== kb
->pshader
&&
3946 (memcmp(&ka
->ps_args
, &kb
->ps_args
, sizeof(kb
->ps_args
)) == 0) &&
3947 (memcmp(&ka
->vs_args
, &kb
->vs_args
, sizeof(kb
->vs_args
)) == 0);
3950 static BOOL
constant_heap_init(struct constant_heap
*heap
, unsigned int constant_count
)
3952 SIZE_T size
= (constant_count
+ 1) * sizeof(*heap
->entries
) + constant_count
* sizeof(*heap
->positions
);
3953 void *mem
= HeapAlloc(GetProcessHeap(), 0, size
);
3957 ERR("Failed to allocate memory\n");
3961 heap
->entries
= mem
;
3962 heap
->entries
[1].version
= 0;
3963 heap
->positions
= (unsigned int *)(heap
->entries
+ constant_count
+ 1);
3969 static void constant_heap_free(struct constant_heap
*heap
)
3971 HeapFree(GetProcessHeap(), 0, heap
->entries
);
3974 static HRESULT
shader_glsl_alloc(IWineD3DDevice
*iface
) {
3975 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
3976 const WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
3977 struct shader_glsl_priv
*priv
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(struct shader_glsl_priv
));
3978 SIZE_T stack_size
= wined3d_log2i(max(GL_LIMITS(vshader_constantsF
), GL_LIMITS(pshader_constantsF
))) + 1;
3980 priv
->stack
= HeapAlloc(GetProcessHeap(), 0, stack_size
* sizeof(*priv
->stack
));
3983 ERR("Failed to allocate memory.\n");
3984 HeapFree(GetProcessHeap(), 0, priv
);
3985 return E_OUTOFMEMORY
;
3988 if (!constant_heap_init(&priv
->vconst_heap
, GL_LIMITS(vshader_constantsF
)))
3990 ERR("Failed to initialize vertex shader constant heap\n");
3991 HeapFree(GetProcessHeap(), 0, priv
->stack
);
3992 HeapFree(GetProcessHeap(), 0, priv
);
3993 return E_OUTOFMEMORY
;
3996 if (!constant_heap_init(&priv
->pconst_heap
, GL_LIMITS(pshader_constantsF
)))
3998 ERR("Failed to initialize pixel shader constant heap\n");
3999 constant_heap_free(&priv
->vconst_heap
);
4000 HeapFree(GetProcessHeap(), 0, priv
->stack
);
4001 HeapFree(GetProcessHeap(), 0, priv
);
4002 return E_OUTOFMEMORY
;
4005 priv
->glsl_program_lookup
= hash_table_create(glsl_program_key_hash
, glsl_program_key_compare
);
4006 priv
->next_constant_version
= 1;
4008 This
->shader_priv
= priv
;
4012 static void shader_glsl_free(IWineD3DDevice
*iface
) {
4013 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
4014 const WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
4015 struct shader_glsl_priv
*priv
= This
->shader_priv
;
4018 for (i
= 0; i
< tex_type_count
; ++i
)
4020 if (priv
->depth_blt_program
[i
])
4022 GL_EXTCALL(glDeleteObjectARB(priv
->depth_blt_program
[i
]));
4026 hash_table_destroy(priv
->glsl_program_lookup
, NULL
, NULL
);
4027 constant_heap_free(&priv
->pconst_heap
);
4028 constant_heap_free(&priv
->vconst_heap
);
4030 HeapFree(GetProcessHeap(), 0, This
->shader_priv
);
4031 This
->shader_priv
= NULL
;
4034 static BOOL
shader_glsl_dirty_const(IWineD3DDevice
*iface
) {
4035 /* TODO: GL_EXT_bindable_uniform can be used to share constants across shaders */
4039 static GLuint
shader_glsl_generate_pshader(IWineD3DPixelShader
*iface
,
4040 SHADER_BUFFER
*buffer
, const struct ps_compile_args
*args
)
4042 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*)iface
;
4043 const struct shader_reg_maps
*reg_maps
= &This
->baseShader
.reg_maps
;
4044 CONST DWORD
*function
= This
->baseShader
.function
;
4045 const char *fragcolor
;
4046 const WineD3D_GL_Info
*gl_info
= &((IWineD3DDeviceImpl
*)This
->baseShader
.device
)->adapter
->gl_info
;
4048 /* Create the hw GLSL shader object and assign it as the shader->prgId */
4049 GLhandleARB shader_obj
= GL_EXTCALL(glCreateShaderObjectARB(GL_FRAGMENT_SHADER_ARB
));
4051 shader_addline(buffer
, "#version 120\n");
4053 if (GL_SUPPORT(ARB_DRAW_BUFFERS
)) {
4054 shader_addline(buffer
, "#extension GL_ARB_draw_buffers : enable\n");
4056 if(GL_SUPPORT(ARB_SHADER_TEXTURE_LOD
) && reg_maps
->usestexldd
) {
4057 shader_addline(buffer
, "#extension GL_ARB_shader_texture_lod : enable\n");
4059 if (GL_SUPPORT(ARB_TEXTURE_RECTANGLE
)) {
4060 /* The spec says that it doesn't have to be explicitly enabled, but the nvidia
4061 * drivers write a warning if we don't do so
4063 shader_addline(buffer
, "#extension GL_ARB_texture_rectangle : enable\n");
4066 /* Base Declarations */
4067 shader_generate_glsl_declarations( (IWineD3DBaseShader
*) This
, reg_maps
, buffer
, &GLINFO_LOCATION
, args
);
4069 /* Pack 3.0 inputs */
4070 if (reg_maps
->shader_version
.major
>= 3 && args
->vp_mode
!= vertexshader
)
4072 pshader_glsl_input_pack(iface
, buffer
, This
->semantics_in
, reg_maps
, args
->vp_mode
);
4075 /* Base Shader Body */
4076 shader_generate_main((IWineD3DBaseShader
*)This
, buffer
, reg_maps
, function
);
4078 /* Pixel shaders < 2.0 place the resulting color in R0 implicitly */
4079 if (reg_maps
->shader_version
.major
< 2)
4081 /* Some older cards like GeforceFX ones don't support multiple buffers, so also not gl_FragData */
4082 if(GL_SUPPORT(ARB_DRAW_BUFFERS
))
4083 shader_addline(buffer
, "gl_FragData[0] = R0;\n");
4085 shader_addline(buffer
, "gl_FragColor = R0;\n");
4088 if(GL_SUPPORT(ARB_DRAW_BUFFERS
)) {
4089 fragcolor
= "gl_FragData[0]";
4091 fragcolor
= "gl_FragColor";
4093 if(args
->srgb_correction
) {
4094 shader_addline(buffer
, "tmp0.xyz = pow(%s.xyz, vec3(%f, %f, %f)) * vec3(%f, %f, %f) - vec3(%f, %f, %f);\n",
4095 fragcolor
, srgb_pow
, srgb_pow
, srgb_pow
, srgb_mul_high
, srgb_mul_high
, srgb_mul_high
,
4096 srgb_sub_high
, srgb_sub_high
, srgb_sub_high
);
4097 shader_addline(buffer
, "tmp1.xyz = %s.xyz * srgb_mul_low.xyz;\n", fragcolor
);
4098 shader_addline(buffer
, "%s.x = %s.x < srgb_comparison.x ? tmp1.x : tmp0.x;\n", fragcolor
, fragcolor
);
4099 shader_addline(buffer
, "%s.y = %s.y < srgb_comparison.y ? tmp1.y : tmp0.y;\n", fragcolor
, fragcolor
);
4100 shader_addline(buffer
, "%s.z = %s.z < srgb_comparison.z ? tmp1.z : tmp0.z;\n", fragcolor
, fragcolor
);
4101 shader_addline(buffer
, "%s = clamp(%s, 0.0, 1.0);\n", fragcolor
, fragcolor
);
4103 /* Pixel shader < 3.0 do not replace the fog stage.
4104 * This implements linear fog computation and blending.
4105 * TODO: non linear fog
4106 * NOTE: gl_Fog.start and gl_Fog.end don't hold fog start s and end e but
4107 * -1/(e-s) and e/(e-s) respectively.
4109 if (reg_maps
->shader_version
.major
< 3)
4112 case FOG_OFF
: break;
4114 shader_addline(buffer
, "float fogstart = -1.0 / (gl_Fog.end - gl_Fog.start);\n");
4115 shader_addline(buffer
, "float fogend = gl_Fog.end * -fogstart;\n");
4116 shader_addline(buffer
, "float Fog = clamp(gl_FogFragCoord * fogstart + fogend, 0.0, 1.0);\n");
4117 shader_addline(buffer
, "%s.xyz = mix(gl_Fog.color.xyz, %s.xyz, Fog);\n", fragcolor
, fragcolor
);
4120 /* Fog = e^(-gl_Fog.density * gl_FogFragCoord) */
4121 shader_addline(buffer
, "float Fog = exp(-gl_Fog.density * gl_FogFragCoord);\n");
4122 shader_addline(buffer
, "Fog = clamp(Fog, 0.0, 1.0);\n");
4123 shader_addline(buffer
, "%s.xyz = mix(gl_Fog.color.xyz, %s.xyz, Fog);\n", fragcolor
, fragcolor
);
4126 /* Fog = e^(-(gl_Fog.density * gl_FogFragCoord)^2) */
4127 shader_addline(buffer
, "float Fog = exp(-gl_Fog.density * gl_Fog.density * gl_FogFragCoord * gl_FogFragCoord);\n");
4128 shader_addline(buffer
, "Fog = clamp(Fog, 0.0, 1.0);\n");
4129 shader_addline(buffer
, "%s.xyz = mix(gl_Fog.color.xyz, %s.xyz, Fog);\n", fragcolor
, fragcolor
);
4134 shader_addline(buffer
, "}\n");
4136 TRACE("Compiling shader object %u\n", shader_obj
);
4137 GL_EXTCALL(glShaderSourceARB(shader_obj
, 1, (const char**)&buffer
->buffer
, NULL
));
4138 GL_EXTCALL(glCompileShaderARB(shader_obj
));
4139 print_glsl_info_log(&GLINFO_LOCATION
, shader_obj
);
4141 /* Store the shader object */
4145 static GLuint
shader_glsl_generate_vshader(IWineD3DVertexShader
*iface
,
4146 SHADER_BUFFER
*buffer
, const struct vs_compile_args
*args
)
4148 IWineD3DVertexShaderImpl
*This
= (IWineD3DVertexShaderImpl
*)iface
;
4149 const struct shader_reg_maps
*reg_maps
= &This
->baseShader
.reg_maps
;
4150 CONST DWORD
*function
= This
->baseShader
.function
;
4151 const WineD3D_GL_Info
*gl_info
= &((IWineD3DDeviceImpl
*)This
->baseShader
.device
)->adapter
->gl_info
;
4153 /* Create the hw GLSL shader program and assign it as the shader->prgId */
4154 GLhandleARB shader_obj
= GL_EXTCALL(glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB
));
4156 shader_addline(buffer
, "#version 120\n");
4158 /* Base Declarations */
4159 shader_generate_glsl_declarations( (IWineD3DBaseShader
*) This
, reg_maps
, buffer
, &GLINFO_LOCATION
, NULL
);
4161 /* Base Shader Body */
4162 shader_generate_main((IWineD3DBaseShader
*)This
, buffer
, reg_maps
, function
);
4164 /* Unpack 3.0 outputs */
4165 if (reg_maps
->shader_version
.major
>= 3) shader_addline(buffer
, "order_ps_input(OUT);\n");
4166 else shader_addline(buffer
, "order_ps_input();\n");
4168 /* The D3DRS_FOGTABLEMODE render state defines if the shader-generated fog coord is used
4169 * or if the fragment depth is used. If the fragment depth is used(FOGTABLEMODE != NONE),
4170 * the fog frag coord is thrown away. If the fog frag coord is used, but not written by
4171 * the shader, it is set to 0.0(fully fogged, since start = 1.0, end = 0.0)
4173 if(args
->fog_src
== VS_FOG_Z
) {
4174 shader_addline(buffer
, "gl_FogFragCoord = gl_Position.z;\n");
4175 } else if (!reg_maps
->fog
) {
4176 shader_addline(buffer
, "gl_FogFragCoord = 0.0;\n");
4179 /* Write the final position.
4181 * OpenGL coordinates specify the center of the pixel while d3d coords specify
4182 * the corner. The offsets are stored in z and w in posFixup. posFixup.y contains
4183 * 1.0 or -1.0 to turn the rendering upside down for offscreen rendering. PosFixup.x
4184 * contains 1.0 to allow a mad.
4186 shader_addline(buffer
, "gl_Position.y = gl_Position.y * posFixup.y;\n");
4187 shader_addline(buffer
, "gl_Position.xy += posFixup.zw * gl_Position.ww;\n");
4189 /* Z coord [0;1]->[-1;1] mapping, see comment in transform_projection in state.c
4191 * Basically we want (in homogeneous coordinates) z = z * 2 - 1. However, shaders are run
4192 * before the homogeneous divide, so we have to take the w into account: z = ((z / w) * 2 - 1) * w,
4193 * which is the same as z = z * 2 - w.
4195 shader_addline(buffer
, "gl_Position.z = gl_Position.z * 2.0 - gl_Position.w;\n");
4197 shader_addline(buffer
, "}\n");
4199 TRACE("Compiling shader object %u\n", shader_obj
);
4200 GL_EXTCALL(glShaderSourceARB(shader_obj
, 1, (const char**)&buffer
->buffer
, NULL
));
4201 GL_EXTCALL(glCompileShaderARB(shader_obj
));
4202 print_glsl_info_log(&GLINFO_LOCATION
, shader_obj
);
4207 static void shader_glsl_get_caps(WINED3DDEVTYPE devtype
, const WineD3D_GL_Info
*gl_info
, struct shader_caps
*pCaps
)
4209 /* Nvidia Geforce6/7 or Ati R4xx/R5xx cards with GLSL support, support VS 3.0 but older Nvidia/Ati
4210 * models with GLSL support only support 2.0. In case of nvidia we can detect VS 2.0 support using
4211 * vs_nv_version which is based on NV_vertex_program.
4212 * For Ati cards there's no way using glsl (it abstracts the lowlevel info away) and also not
4213 * using ARB_vertex_program. It is safe to assume that when a card supports pixel shader 2.0 it
4214 * supports vertex shader 2.0 too and the way around. We can detect ps2.0 using the maximum number
4215 * of native instructions, so use that here. For more info see the pixel shader versioning code below.
4217 if((GLINFO_LOCATION
.vs_nv_version
== VS_VERSION_20
) || (GLINFO_LOCATION
.ps_arb_max_instructions
<= 512))
4218 pCaps
->VertexShaderVersion
= WINED3DVS_VERSION(2,0);
4220 pCaps
->VertexShaderVersion
= WINED3DVS_VERSION(3,0);
4221 TRACE_(d3d_caps
)("Hardware vertex shader version %d.%d enabled (GLSL)\n", (pCaps
->VertexShaderVersion
>> 8) & 0xff, pCaps
->VertexShaderVersion
& 0xff);
4222 pCaps
->MaxVertexShaderConst
= GL_LIMITS(vshader_constantsF
);
4224 /* Older DX9-class videocards (GeforceFX / Radeon >9500/X*00) only support pixel shader 2.0/2.0a/2.0b.
4225 * In OpenGL the extensions related to GLSL abstract lowlevel GL info away which is needed
4226 * to distinguish between 2.0 and 3.0 (and 2.0a/2.0b). In case of Nvidia we use their fragment
4227 * program extensions. On other hardware including ATI GL_ARB_fragment_program offers the info
4228 * in max native instructions. Intel and others also offer the info in this extension but they
4229 * don't support GLSL (at least on Windows).
4231 * PS2.0 requires at least 96 instructions, 2.0a/2.0b go up to 512. Assume that if the number
4232 * of instructions is 512 or less we have to do with ps2.0 hardware.
4233 * 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.
4235 if((GLINFO_LOCATION
.ps_nv_version
== PS_VERSION_20
) || (GLINFO_LOCATION
.ps_arb_max_instructions
<= 512))
4236 pCaps
->PixelShaderVersion
= WINED3DPS_VERSION(2,0);
4238 pCaps
->PixelShaderVersion
= WINED3DPS_VERSION(3,0);
4240 pCaps
->MaxPixelShaderConst
= GL_LIMITS(pshader_constantsF
);
4242 /* FIXME: The following line is card dependent. -8.0 to 8.0 is the
4243 * Direct3D minimum requirement.
4245 * Both GL_ARB_fragment_program and GLSL require a "maximum representable magnitude"
4246 * of colors to be 2^10, and 2^32 for other floats. Should we use 1024 here?
4248 * The problem is that the refrast clamps temporary results in the shader to
4249 * [-MaxValue;+MaxValue]. If the card's max value is bigger than the one we advertize here,
4250 * then applications may miss the clamping behavior. On the other hand, if it is smaller,
4251 * the shader will generate incorrect results too. Unfortunately, GL deliberately doesn't
4252 * offer a way to query this.
4254 pCaps
->PixelShader1xMaxValue
= 8.0;
4255 TRACE_(d3d_caps
)("Hardware pixel shader version %d.%d enabled (GLSL)\n", (pCaps
->PixelShaderVersion
>> 8) & 0xff, pCaps
->PixelShaderVersion
& 0xff);
4258 static BOOL
shader_glsl_color_fixup_supported(struct color_fixup_desc fixup
)
4260 if (TRACE_ON(d3d_shader
) && TRACE_ON(d3d
))
4262 TRACE("Checking support for fixup:\n");
4263 dump_color_fixup_desc(fixup
);
4266 /* We support everything except YUV conversions. */
4267 if (!is_yuv_fixup(fixup
))
4273 TRACE("[FAILED]\n");
4277 static const SHADER_HANDLER shader_glsl_instruction_handler_table
[WINED3DSIH_TABLE_SIZE
] =
4279 /* WINED3DSIH_ABS */ shader_glsl_map2gl
,
4280 /* WINED3DSIH_ADD */ shader_glsl_arith
,
4281 /* WINED3DSIH_BEM */ pshader_glsl_bem
,
4282 /* WINED3DSIH_BREAK */ shader_glsl_break
,
4283 /* WINED3DSIH_BREAKC */ shader_glsl_breakc
,
4284 /* WINED3DSIH_BREAKP */ NULL
,
4285 /* WINED3DSIH_CALL */ shader_glsl_call
,
4286 /* WINED3DSIH_CALLNZ */ shader_glsl_callnz
,
4287 /* WINED3DSIH_CMP */ shader_glsl_cmp
,
4288 /* WINED3DSIH_CND */ shader_glsl_cnd
,
4289 /* WINED3DSIH_CRS */ shader_glsl_cross
,
4290 /* WINED3DSIH_DCL */ NULL
,
4291 /* WINED3DSIH_DEF */ NULL
,
4292 /* WINED3DSIH_DEFB */ NULL
,
4293 /* WINED3DSIH_DEFI */ NULL
,
4294 /* WINED3DSIH_DP2ADD */ pshader_glsl_dp2add
,
4295 /* WINED3DSIH_DP3 */ shader_glsl_dot
,
4296 /* WINED3DSIH_DP4 */ shader_glsl_dot
,
4297 /* WINED3DSIH_DST */ shader_glsl_dst
,
4298 /* WINED3DSIH_DSX */ shader_glsl_map2gl
,
4299 /* WINED3DSIH_DSY */ shader_glsl_map2gl
,
4300 /* WINED3DSIH_ELSE */ shader_glsl_else
,
4301 /* WINED3DSIH_ENDIF */ shader_glsl_end
,
4302 /* WINED3DSIH_ENDLOOP */ shader_glsl_end
,
4303 /* WINED3DSIH_ENDREP */ shader_glsl_end
,
4304 /* WINED3DSIH_EXP */ shader_glsl_map2gl
,
4305 /* WINED3DSIH_EXPP */ shader_glsl_expp
,
4306 /* WINED3DSIH_FRC */ shader_glsl_map2gl
,
4307 /* WINED3DSIH_IF */ shader_glsl_if
,
4308 /* WINED3DSIH_IFC */ shader_glsl_ifc
,
4309 /* WINED3DSIH_LABEL */ shader_glsl_label
,
4310 /* WINED3DSIH_LIT */ shader_glsl_lit
,
4311 /* WINED3DSIH_LOG */ shader_glsl_log
,
4312 /* WINED3DSIH_LOGP */ shader_glsl_log
,
4313 /* WINED3DSIH_LOOP */ shader_glsl_loop
,
4314 /* WINED3DSIH_LRP */ shader_glsl_lrp
,
4315 /* WINED3DSIH_M3x2 */ shader_glsl_mnxn
,
4316 /* WINED3DSIH_M3x3 */ shader_glsl_mnxn
,
4317 /* WINED3DSIH_M3x4 */ shader_glsl_mnxn
,
4318 /* WINED3DSIH_M4x3 */ shader_glsl_mnxn
,
4319 /* WINED3DSIH_M4x4 */ shader_glsl_mnxn
,
4320 /* WINED3DSIH_MAD */ shader_glsl_mad
,
4321 /* WINED3DSIH_MAX */ shader_glsl_map2gl
,
4322 /* WINED3DSIH_MIN */ shader_glsl_map2gl
,
4323 /* WINED3DSIH_MOV */ shader_glsl_mov
,
4324 /* WINED3DSIH_MOVA */ shader_glsl_mov
,
4325 /* WINED3DSIH_MUL */ shader_glsl_arith
,
4326 /* WINED3DSIH_NOP */ NULL
,
4327 /* WINED3DSIH_NRM */ shader_glsl_map2gl
,
4328 /* WINED3DSIH_PHASE */ NULL
,
4329 /* WINED3DSIH_POW */ shader_glsl_pow
,
4330 /* WINED3DSIH_RCP */ shader_glsl_rcp
,
4331 /* WINED3DSIH_REP */ shader_glsl_rep
,
4332 /* WINED3DSIH_RET */ NULL
,
4333 /* WINED3DSIH_RSQ */ shader_glsl_rsq
,
4334 /* WINED3DSIH_SETP */ NULL
,
4335 /* WINED3DSIH_SGE */ shader_glsl_compare
,
4336 /* WINED3DSIH_SGN */ shader_glsl_map2gl
,
4337 /* WINED3DSIH_SINCOS */ shader_glsl_sincos
,
4338 /* WINED3DSIH_SLT */ shader_glsl_compare
,
4339 /* WINED3DSIH_SUB */ shader_glsl_arith
,
4340 /* WINED3DSIH_TEX */ pshader_glsl_tex
,
4341 /* WINED3DSIH_TEXBEM */ pshader_glsl_texbem
,
4342 /* WINED3DSIH_TEXBEML */ pshader_glsl_texbem
,
4343 /* WINED3DSIH_TEXCOORD */ pshader_glsl_texcoord
,
4344 /* WINED3DSIH_TEXDEPTH */ pshader_glsl_texdepth
,
4345 /* WINED3DSIH_TEXDP3 */ pshader_glsl_texdp3
,
4346 /* WINED3DSIH_TEXDP3TEX */ pshader_glsl_texdp3tex
,
4347 /* WINED3DSIH_TEXKILL */ pshader_glsl_texkill
,
4348 /* WINED3DSIH_TEXLDD */ shader_glsl_texldd
,
4349 /* WINED3DSIH_TEXLDL */ shader_glsl_texldl
,
4350 /* WINED3DSIH_TEXM3x2DEPTH */ pshader_glsl_texm3x2depth
,
4351 /* WINED3DSIH_TEXM3x2PAD */ pshader_glsl_texm3x2pad
,
4352 /* WINED3DSIH_TEXM3x2TEX */ pshader_glsl_texm3x2tex
,
4353 /* WINED3DSIH_TEXM3x3 */ pshader_glsl_texm3x3
,
4354 /* WINED3DSIH_TEXM3x3DIFF */ NULL
,
4355 /* WINED3DSIH_TEXM3x3PAD */ pshader_glsl_texm3x3pad
,
4356 /* WINED3DSIH_TEXM3x3SPEC */ pshader_glsl_texm3x3spec
,
4357 /* WINED3DSIH_TEXM3x3TEX */ pshader_glsl_texm3x3tex
,
4358 /* WINED3DSIH_TEXM3x3VSPEC */ pshader_glsl_texm3x3vspec
,
4359 /* WINED3DSIH_TEXREG2AR */ pshader_glsl_texreg2ar
,
4360 /* WINED3DSIH_TEXREG2GB */ pshader_glsl_texreg2gb
,
4361 /* WINED3DSIH_TEXREG2RGB */ pshader_glsl_texreg2rgb
,
4364 const shader_backend_t glsl_shader_backend
= {
4365 shader_glsl_instruction_handler_table
,
4367 shader_glsl_select_depth_blt
,
4368 shader_glsl_deselect_depth_blt
,
4369 shader_glsl_update_float_vertex_constants
,
4370 shader_glsl_update_float_pixel_constants
,
4371 shader_glsl_load_constants
,
4372 shader_glsl_load_np2fixup_constants
,
4373 shader_glsl_destroy
,
4376 shader_glsl_dirty_const
,
4377 shader_glsl_generate_pshader
,
4378 shader_glsl_generate_vshader
,
4379 shader_glsl_get_caps
,
4380 shader_glsl_color_fixup_supported
,