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 wined3d_shader_buffer shader_buffer
;
87 struct wine_rb_tree program_lookup
;
88 struct glsl_shader_prog_link
*glsl_program
;
89 struct constant_heap vconst_heap
;
90 struct constant_heap pconst_heap
;
92 GLhandleARB depth_blt_program
[tex_type_count
];
93 UINT next_constant_version
;
96 /* Struct to maintain data about a linked GLSL program */
97 struct glsl_shader_prog_link
{
98 struct wine_rb_entry program_lookup_entry
;
99 struct list vshader_entry
;
100 struct list pshader_entry
;
101 GLhandleARB programId
;
102 GLint
*vuniformF_locations
;
103 GLint
*puniformF_locations
;
104 GLint vuniformI_locations
[MAX_CONST_I
];
105 GLint puniformI_locations
[MAX_CONST_I
];
106 GLint posFixup_location
;
107 GLint np2Fixup_location
;
108 GLint bumpenvmat_location
[MAX_TEXTURES
];
109 GLint luminancescale_location
[MAX_TEXTURES
];
110 GLint luminanceoffset_location
[MAX_TEXTURES
];
111 GLint ycorrection_location
;
112 GLenum vertex_color_clamp
;
113 IWineD3DVertexShader
*vshader
;
114 IWineD3DPixelShader
*pshader
;
115 struct vs_compile_args vs_args
;
116 struct ps_compile_args ps_args
;
117 UINT constant_version
;
118 const struct ps_np2fixup_info
*np2Fixup_info
;
122 IWineD3DVertexShader
*vshader
;
123 IWineD3DPixelShader
*pshader
;
124 struct ps_compile_args ps_args
;
125 struct vs_compile_args vs_args
;
126 } glsl_program_key_t
;
128 struct shader_glsl_ctx_priv
{
129 const struct vs_compile_args
*cur_vs_args
;
130 const struct ps_compile_args
*cur_ps_args
;
131 struct ps_np2fixup_info
*cur_np2fixup_info
;
134 struct glsl_ps_compiled_shader
136 struct ps_compile_args args
;
137 struct ps_np2fixup_info np2fixup
;
141 struct glsl_pshader_private
143 struct glsl_ps_compiled_shader
*gl_shaders
;
144 UINT num_gl_shaders
, shader_array_size
;
147 struct glsl_vs_compiled_shader
149 struct vs_compile_args args
;
153 struct glsl_vshader_private
155 struct glsl_vs_compiled_shader
*gl_shaders
;
156 UINT num_gl_shaders
, shader_array_size
;
159 /* Extract a line from the info log.
160 * Note that this modifies the source string. */
161 static char *get_info_log_line(char **ptr
)
166 if (!(q
= strstr(p
, "\n")))
168 if (!*p
) return NULL
;
178 /** Prints the GLSL info log which will contain error messages if they exist */
179 /* GL locking is done by the caller */
180 static void print_glsl_info_log(const struct wined3d_gl_info
*gl_info
, GLhandleARB obj
)
182 int infologLength
= 0;
187 static const char * const spam
[] =
189 "Vertex shader was successfully compiled to run on hardware.\n", /* fglrx */
190 "Fragment shader was successfully compiled to run on hardware.\n", /* fglrx */
191 "Fragment shader(s) linked, vertex shader(s) linked. \n ", /* fglrx, with \n */
192 "Fragment shader(s) linked, vertex shader(s) linked.", /* fglrx, no \n */
193 "Vertex shader(s) linked, no fragment shader(s) defined. \n ", /* fglrx, with \n */
194 "Vertex shader(s) linked, no fragment shader(s) defined.", /* fglrx, no \n */
195 "Fragment shader was successfully compiled to run on hardware.\n"
196 "Fragment shader(s) linked, no vertex shader(s) defined.", /* fglrx, no \n */
197 "Fragment shader(s) linked, no vertex shader(s) defined. \n ", /* fglrx, with \n */
200 if (!TRACE_ON(d3d_shader
) && !FIXME_ON(d3d_shader
)) return;
202 GL_EXTCALL(glGetObjectParameterivARB(obj
,
203 GL_OBJECT_INFO_LOG_LENGTH_ARB
,
206 /* A size of 1 is just a null-terminated string, so the log should be bigger than
207 * that if there are errors. */
208 if (infologLength
> 1)
212 /* Fglrx doesn't terminate the string properly, but it tells us the proper length.
213 * So use HEAP_ZERO_MEMORY to avoid uninitialized bytes
215 infoLog
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, infologLength
);
216 GL_EXTCALL(glGetInfoLogARB(obj
, infologLength
, NULL
, infoLog
));
219 for(i
= 0; i
< sizeof(spam
) / sizeof(spam
[0]); i
++) {
220 if(strcmp(infoLog
, spam
[i
]) == 0) {
229 TRACE("Spam received from GLSL shader #%u:\n", obj
);
230 while ((line
= get_info_log_line(&ptr
))) TRACE(" %s\n", line
);
234 FIXME("Error received from GLSL shader #%u:\n", obj
);
235 while ((line
= get_info_log_line(&ptr
))) FIXME(" %s\n", line
);
237 HeapFree(GetProcessHeap(), 0, infoLog
);
242 * Loads (pixel shader) samplers
244 /* GL locking is done by the caller */
245 static void shader_glsl_load_psamplers(const struct wined3d_gl_info
*gl_info
,
246 DWORD
*tex_unit_map
, GLhandleARB programId
)
250 char sampler_name
[20];
252 for (i
= 0; i
< MAX_FRAGMENT_SAMPLERS
; ++i
) {
253 snprintf(sampler_name
, sizeof(sampler_name
), "Psampler%d", i
);
254 name_loc
= GL_EXTCALL(glGetUniformLocationARB(programId
, sampler_name
));
255 if (name_loc
!= -1) {
256 DWORD mapped_unit
= tex_unit_map
[i
];
257 if (mapped_unit
!= WINED3D_UNMAPPED_STAGE
&& mapped_unit
< GL_LIMITS(fragment_samplers
))
259 TRACE("Loading %s for texture %d\n", sampler_name
, mapped_unit
);
260 GL_EXTCALL(glUniform1iARB(name_loc
, mapped_unit
));
261 checkGLcall("glUniform1iARB");
263 ERR("Trying to load sampler %s on unsupported unit %d\n", sampler_name
, mapped_unit
);
269 /* GL locking is done by the caller */
270 static void shader_glsl_load_vsamplers(const struct wined3d_gl_info
*gl_info
,
271 DWORD
*tex_unit_map
, GLhandleARB programId
)
274 char sampler_name
[20];
277 for (i
= 0; i
< MAX_VERTEX_SAMPLERS
; ++i
) {
278 snprintf(sampler_name
, sizeof(sampler_name
), "Vsampler%d", i
);
279 name_loc
= GL_EXTCALL(glGetUniformLocationARB(programId
, sampler_name
));
280 if (name_loc
!= -1) {
281 DWORD mapped_unit
= tex_unit_map
[MAX_FRAGMENT_SAMPLERS
+ i
];
282 if (mapped_unit
!= WINED3D_UNMAPPED_STAGE
&& mapped_unit
< GL_LIMITS(combined_samplers
))
284 TRACE("Loading %s for texture %d\n", sampler_name
, mapped_unit
);
285 GL_EXTCALL(glUniform1iARB(name_loc
, mapped_unit
));
286 checkGLcall("glUniform1iARB");
288 ERR("Trying to load sampler %s on unsupported unit %d\n", sampler_name
, mapped_unit
);
294 /* GL locking is done by the caller */
295 static inline void walk_constant_heap(const struct wined3d_gl_info
*gl_info
, const float *constants
,
296 const GLint
*constant_locations
, const struct constant_heap
*heap
, unsigned char *stack
, DWORD version
)
299 unsigned int heap_idx
= 1;
302 if (heap
->entries
[heap_idx
].version
<= version
) return;
304 idx
= heap
->entries
[heap_idx
].idx
;
305 if (constant_locations
[idx
] != -1) GL_EXTCALL(glUniform4fvARB(constant_locations
[idx
], 1, &constants
[idx
* 4]));
306 stack
[stack_idx
] = HEAP_NODE_TRAVERSE_LEFT
;
308 while (stack_idx
>= 0)
310 /* Note that we fall through to the next case statement. */
311 switch(stack
[stack_idx
])
313 case HEAP_NODE_TRAVERSE_LEFT
:
315 unsigned int left_idx
= heap_idx
<< 1;
316 if (left_idx
< heap
->size
&& heap
->entries
[left_idx
].version
> version
)
319 idx
= heap
->entries
[heap_idx
].idx
;
320 if (constant_locations
[idx
] != -1)
321 GL_EXTCALL(glUniform4fvARB(constant_locations
[idx
], 1, &constants
[idx
* 4]));
323 stack
[stack_idx
++] = HEAP_NODE_TRAVERSE_RIGHT
;
324 stack
[stack_idx
] = HEAP_NODE_TRAVERSE_LEFT
;
329 case HEAP_NODE_TRAVERSE_RIGHT
:
331 unsigned int right_idx
= (heap_idx
<< 1) + 1;
332 if (right_idx
< heap
->size
&& heap
->entries
[right_idx
].version
> version
)
334 heap_idx
= right_idx
;
335 idx
= heap
->entries
[heap_idx
].idx
;
336 if (constant_locations
[idx
] != -1)
337 GL_EXTCALL(glUniform4fvARB(constant_locations
[idx
], 1, &constants
[idx
* 4]));
339 stack
[stack_idx
++] = HEAP_NODE_POP
;
340 stack
[stack_idx
] = HEAP_NODE_TRAVERSE_LEFT
;
353 checkGLcall("walk_constant_heap()");
356 /* GL locking is done by the caller */
357 static inline void apply_clamped_constant(const struct wined3d_gl_info
*gl_info
, GLint location
, const GLfloat
*data
)
359 GLfloat clamped_constant
[4];
361 if (location
== -1) return;
363 clamped_constant
[0] = data
[0] < -1.0f
? -1.0f
: data
[0] > 1.0f
? 1.0f
: data
[0];
364 clamped_constant
[1] = data
[1] < -1.0f
? -1.0f
: data
[1] > 1.0f
? 1.0f
: data
[1];
365 clamped_constant
[2] = data
[2] < -1.0f
? -1.0f
: data
[2] > 1.0f
? 1.0f
: data
[2];
366 clamped_constant
[3] = data
[3] < -1.0f
? -1.0f
: data
[3] > 1.0f
? 1.0f
: data
[3];
368 GL_EXTCALL(glUniform4fvARB(location
, 1, clamped_constant
));
371 /* GL locking is done by the caller */
372 static inline void walk_constant_heap_clamped(const struct wined3d_gl_info
*gl_info
, const float *constants
,
373 const GLint
*constant_locations
, const struct constant_heap
*heap
, unsigned char *stack
, DWORD version
)
376 unsigned int heap_idx
= 1;
379 if (heap
->entries
[heap_idx
].version
<= version
) return;
381 idx
= heap
->entries
[heap_idx
].idx
;
382 apply_clamped_constant(gl_info
, constant_locations
[idx
], &constants
[idx
* 4]);
383 stack
[stack_idx
] = HEAP_NODE_TRAVERSE_LEFT
;
385 while (stack_idx
>= 0)
387 /* Note that we fall through to the next case statement. */
388 switch(stack
[stack_idx
])
390 case HEAP_NODE_TRAVERSE_LEFT
:
392 unsigned int left_idx
= heap_idx
<< 1;
393 if (left_idx
< heap
->size
&& heap
->entries
[left_idx
].version
> version
)
396 idx
= heap
->entries
[heap_idx
].idx
;
397 apply_clamped_constant(gl_info
, constant_locations
[idx
], &constants
[idx
* 4]);
399 stack
[stack_idx
++] = HEAP_NODE_TRAVERSE_RIGHT
;
400 stack
[stack_idx
] = HEAP_NODE_TRAVERSE_LEFT
;
405 case HEAP_NODE_TRAVERSE_RIGHT
:
407 unsigned int right_idx
= (heap_idx
<< 1) + 1;
408 if (right_idx
< heap
->size
&& heap
->entries
[right_idx
].version
> version
)
410 heap_idx
= right_idx
;
411 idx
= heap
->entries
[heap_idx
].idx
;
412 apply_clamped_constant(gl_info
, constant_locations
[idx
], &constants
[idx
* 4]);
414 stack
[stack_idx
++] = HEAP_NODE_POP
;
415 stack
[stack_idx
] = HEAP_NODE_TRAVERSE_LEFT
;
428 checkGLcall("walk_constant_heap_clamped()");
431 /* Loads floating point constants (aka uniforms) into the currently set GLSL program. */
432 /* GL locking is done by the caller */
433 static void shader_glsl_load_constantsF(IWineD3DBaseShaderImpl
*This
, const struct wined3d_gl_info
*gl_info
,
434 const float *constants
, const GLint
*constant_locations
, const struct constant_heap
*heap
,
435 unsigned char *stack
, UINT version
)
437 const local_constant
*lconst
;
439 /* 1.X pshaders have the constants clamped to [-1;1] implicitly. */
440 if (This
->baseShader
.reg_maps
.shader_version
.major
== 1
441 && shader_is_pshader_version(This
->baseShader
.reg_maps
.shader_version
.type
))
442 walk_constant_heap_clamped(gl_info
, constants
, constant_locations
, heap
, stack
, version
);
444 walk_constant_heap(gl_info
, constants
, constant_locations
, heap
, stack
, version
);
446 if (!This
->baseShader
.load_local_constsF
)
448 TRACE("No need to load local float constants for this shader\n");
452 /* Immediate constants are clamped to [-1;1] at shader creation time if needed */
453 LIST_FOR_EACH_ENTRY(lconst
, &This
->baseShader
.constantsF
, local_constant
, entry
)
455 GLint location
= constant_locations
[lconst
->idx
];
456 /* We found this uniform name in the program - go ahead and send the data */
457 if (location
!= -1) GL_EXTCALL(glUniform4fvARB(location
, 1, (const GLfloat
*)lconst
->value
));
459 checkGLcall("glUniform4fvARB()");
462 /* Loads integer constants (aka uniforms) into the currently set GLSL program. */
463 /* GL locking is done by the caller */
464 static void shader_glsl_load_constantsI(IWineD3DBaseShaderImpl
*This
, const struct wined3d_gl_info
*gl_info
,
465 const GLint locations
[MAX_CONST_I
], const int *constants
, WORD constants_set
)
470 for (i
= 0; constants_set
; constants_set
>>= 1, ++i
)
472 if (!(constants_set
& 1)) continue;
474 TRACE_(d3d_constants
)("Loading constants %u: %i, %i, %i, %i\n",
475 i
, constants
[i
*4], constants
[i
*4+1], constants
[i
*4+2], constants
[i
*4+3]);
477 /* We found this uniform name in the program - go ahead and send the data */
478 GL_EXTCALL(glUniform4ivARB(locations
[i
], 1, &constants
[i
*4]));
479 checkGLcall("glUniform4ivARB");
482 /* Load immediate constants */
483 ptr
= list_head(&This
->baseShader
.constantsI
);
485 const struct local_constant
*lconst
= LIST_ENTRY(ptr
, const struct local_constant
, entry
);
486 unsigned int idx
= lconst
->idx
;
487 const GLint
*values
= (const GLint
*)lconst
->value
;
489 TRACE_(d3d_constants
)("Loading local constants %i: %i, %i, %i, %i\n", idx
,
490 values
[0], values
[1], values
[2], values
[3]);
492 /* We found this uniform name in the program - go ahead and send the data */
493 GL_EXTCALL(glUniform4ivARB(locations
[idx
], 1, values
));
494 checkGLcall("glUniform4ivARB");
495 ptr
= list_next(&This
->baseShader
.constantsI
, ptr
);
499 /* Loads boolean constants (aka uniforms) into the currently set GLSL program. */
500 /* GL locking is done by the caller */
501 static void shader_glsl_load_constantsB(IWineD3DBaseShaderImpl
*This
, const struct wined3d_gl_info
*gl_info
,
502 GLhandleARB programId
, const BOOL
*constants
, WORD constants_set
)
510 switch (This
->baseShader
.reg_maps
.shader_version
.type
)
512 case WINED3D_SHADER_TYPE_VERTEX
:
516 case WINED3D_SHADER_TYPE_GEOMETRY
:
520 case WINED3D_SHADER_TYPE_PIXEL
:
525 FIXME("Unknown shader type %#x.\n",
526 This
->baseShader
.reg_maps
.shader_version
.type
);
531 /* TODO: Benchmark and see if it would be beneficial to store the
532 * locations of the constants to avoid looking up each time */
533 for (i
= 0; constants_set
; constants_set
>>= 1, ++i
)
535 if (!(constants_set
& 1)) continue;
537 TRACE_(d3d_constants
)("Loading constants %i: %i;\n", i
, constants
[i
]);
539 /* TODO: Benchmark and see if it would be beneficial to store the
540 * locations of the constants to avoid looking up each time */
541 snprintf(tmp_name
, sizeof(tmp_name
), "%s[%i]", prefix
, i
);
542 tmp_loc
= GL_EXTCALL(glGetUniformLocationARB(programId
, tmp_name
));
545 /* We found this uniform name in the program - go ahead and send the data */
546 GL_EXTCALL(glUniform1ivARB(tmp_loc
, 1, &constants
[i
]));
547 checkGLcall("glUniform1ivARB");
551 /* Load immediate constants */
552 ptr
= list_head(&This
->baseShader
.constantsB
);
554 const struct local_constant
*lconst
= LIST_ENTRY(ptr
, const struct local_constant
, entry
);
555 unsigned int idx
= lconst
->idx
;
556 const GLint
*values
= (const GLint
*)lconst
->value
;
558 TRACE_(d3d_constants
)("Loading local constants %i: %i\n", idx
, values
[0]);
560 snprintf(tmp_name
, sizeof(tmp_name
), "%s[%i]", prefix
, idx
);
561 tmp_loc
= GL_EXTCALL(glGetUniformLocationARB(programId
, tmp_name
));
563 /* We found this uniform name in the program - go ahead and send the data */
564 GL_EXTCALL(glUniform1ivARB(tmp_loc
, 1, values
));
565 checkGLcall("glUniform1ivARB");
567 ptr
= list_next(&This
->baseShader
.constantsB
, ptr
);
571 static void reset_program_constant_version(struct wine_rb_entry
*entry
, void *context
)
573 WINE_RB_ENTRY_VALUE(entry
, struct glsl_shader_prog_link
, program_lookup_entry
)->constant_version
= 0;
577 * Loads the texture dimensions for NP2 fixup into the currently set GLSL program.
579 /* GL locking is done by the caller (state handler) */
580 static void shader_glsl_load_np2fixup_constants(
581 IWineD3DDevice
* device
,
583 char useVertexShader
) {
585 const IWineD3DDeviceImpl
* deviceImpl
= (const IWineD3DDeviceImpl
*) device
;
586 const struct glsl_shader_prog_link
* prog
= ((struct shader_glsl_priv
*)(deviceImpl
->shader_priv
))->glsl_program
;
589 /* No GLSL program set - nothing to do. */
593 if (!usePixelShader
) {
594 /* NP2 texcoord fixup is (currently) only done for pixelshaders. */
598 if (prog
->ps_args
.np2_fixup
&& -1 != prog
->np2Fixup_location
) {
599 const struct wined3d_gl_info
*gl_info
= &deviceImpl
->adapter
->gl_info
;
600 const IWineD3DStateBlockImpl
* stateBlock
= (const IWineD3DStateBlockImpl
*) deviceImpl
->stateBlock
;
602 UINT fixup
= prog
->ps_args
.np2_fixup
;
603 GLfloat np2fixup_constants
[4 * MAX_FRAGMENT_SAMPLERS
];
605 for (i
= 0; fixup
; fixup
>>= 1, ++i
) {
606 const unsigned char idx
= prog
->np2Fixup_info
->idx
[i
];
607 const IWineD3DBaseTextureImpl
* const tex
= (const IWineD3DBaseTextureImpl
*) stateBlock
->textures
[i
];
608 GLfloat
* tex_dim
= &np2fixup_constants
[(idx
>> 1) * 4];
611 FIXME("Nonexistent texture is flagged for NP2 texcoord fixup\n");
616 tex_dim
[2] = tex
->baseTexture
.pow2Matrix
[0]; tex_dim
[3] = tex
->baseTexture
.pow2Matrix
[5];
618 tex_dim
[0] = tex
->baseTexture
.pow2Matrix
[0]; tex_dim
[1] = tex
->baseTexture
.pow2Matrix
[5];
622 GL_EXTCALL(glUniform4fvARB(prog
->np2Fixup_location
, prog
->np2Fixup_info
->num_consts
, np2fixup_constants
));
627 * Loads the app-supplied constants into the currently set GLSL program.
629 /* GL locking is done by the caller (state handler) */
630 static void shader_glsl_load_constants(const struct wined3d_context
*context
,
631 char usePixelShader
, char useVertexShader
)
633 IWineD3DDeviceImpl
*device
= ((IWineD3DSurfaceImpl
*)context
->surface
)->resource
.wineD3DDevice
;
634 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
635 IWineD3DStateBlockImpl
* stateBlock
= device
->stateBlock
;
636 struct shader_glsl_priv
*priv
= device
->shader_priv
;
638 GLhandleARB programId
;
639 struct glsl_shader_prog_link
*prog
= priv
->glsl_program
;
640 UINT constant_version
;
644 /* No GLSL program set - nothing to do. */
647 programId
= prog
->programId
;
648 constant_version
= prog
->constant_version
;
650 if (useVertexShader
) {
651 IWineD3DBaseShaderImpl
* vshader
= (IWineD3DBaseShaderImpl
*) stateBlock
->vertexShader
;
653 /* Load DirectX 9 float constants/uniforms for vertex shader */
654 shader_glsl_load_constantsF(vshader
, gl_info
, stateBlock
->vertexShaderConstantF
,
655 prog
->vuniformF_locations
, &priv
->vconst_heap
, priv
->stack
, constant_version
);
657 /* Load DirectX 9 integer constants/uniforms for vertex shader */
658 shader_glsl_load_constantsI(vshader
, gl_info
, prog
->vuniformI_locations
, stateBlock
->vertexShaderConstantI
,
659 stateBlock
->changed
.vertexShaderConstantsI
& vshader
->baseShader
.reg_maps
.integer_constants
);
661 /* Load DirectX 9 boolean constants/uniforms for vertex shader */
662 shader_glsl_load_constantsB(vshader
, gl_info
, programId
, stateBlock
->vertexShaderConstantB
,
663 stateBlock
->changed
.vertexShaderConstantsB
& vshader
->baseShader
.reg_maps
.boolean_constants
);
665 /* Upload the position fixup params */
666 GL_EXTCALL(glUniform4fvARB(prog
->posFixup_location
, 1, &device
->posFixup
[0]));
667 checkGLcall("glUniform4fvARB");
670 if (usePixelShader
) {
672 IWineD3DBaseShaderImpl
* pshader
= (IWineD3DBaseShaderImpl
*) stateBlock
->pixelShader
;
674 /* Load DirectX 9 float constants/uniforms for pixel shader */
675 shader_glsl_load_constantsF(pshader
, gl_info
, stateBlock
->pixelShaderConstantF
,
676 prog
->puniformF_locations
, &priv
->pconst_heap
, priv
->stack
, constant_version
);
678 /* Load DirectX 9 integer constants/uniforms for pixel shader */
679 shader_glsl_load_constantsI(pshader
, gl_info
, prog
->puniformI_locations
, stateBlock
->pixelShaderConstantI
,
680 stateBlock
->changed
.pixelShaderConstantsI
& pshader
->baseShader
.reg_maps
.integer_constants
);
682 /* Load DirectX 9 boolean constants/uniforms for pixel shader */
683 shader_glsl_load_constantsB(pshader
, gl_info
, programId
, stateBlock
->pixelShaderConstantB
,
684 stateBlock
->changed
.pixelShaderConstantsB
& pshader
->baseShader
.reg_maps
.boolean_constants
);
686 /* Upload the environment bump map matrix if needed. The needsbumpmat member specifies the texture stage to load the matrix from.
687 * It can't be 0 for a valid texbem instruction.
689 for(i
= 0; i
< MAX_TEXTURES
; i
++) {
692 if(prog
->bumpenvmat_location
[i
] == -1) continue;
694 data
= (const float *)&stateBlock
->textureState
[i
][WINED3DTSS_BUMPENVMAT00
];
695 GL_EXTCALL(glUniformMatrix2fvARB(prog
->bumpenvmat_location
[i
], 1, 0, data
));
696 checkGLcall("glUniformMatrix2fvARB");
698 /* texbeml needs the luminance scale and offset too. If texbeml is used, needsbumpmat
699 * is set too, so we can check that in the needsbumpmat check
701 if(prog
->luminancescale_location
[i
] != -1) {
702 const GLfloat
*scale
= (const GLfloat
*)&stateBlock
->textureState
[i
][WINED3DTSS_BUMPENVLSCALE
];
703 const GLfloat
*offset
= (const GLfloat
*)&stateBlock
->textureState
[i
][WINED3DTSS_BUMPENVLOFFSET
];
705 GL_EXTCALL(glUniform1fvARB(prog
->luminancescale_location
[i
], 1, scale
));
706 checkGLcall("glUniform1fvARB");
707 GL_EXTCALL(glUniform1fvARB(prog
->luminanceoffset_location
[i
], 1, offset
));
708 checkGLcall("glUniform1fvARB");
712 if(((IWineD3DPixelShaderImpl
*) pshader
)->vpos_uniform
) {
713 float correction_params
[4];
715 if (context
->render_offscreen
)
717 correction_params
[0] = 0.0f
;
718 correction_params
[1] = 1.0f
;
720 /* position is window relative, not viewport relative */
721 correction_params
[0] = ((IWineD3DSurfaceImpl
*)context
->current_rt
)->currentDesc
.Height
;
722 correction_params
[1] = -1.0f
;
724 GL_EXTCALL(glUniform4fvARB(prog
->ycorrection_location
, 1, correction_params
));
728 if (priv
->next_constant_version
== UINT_MAX
)
730 TRACE("Max constant version reached, resetting to 0.\n");
731 wine_rb_for_each_entry(&priv
->program_lookup
, reset_program_constant_version
, NULL
);
732 priv
->next_constant_version
= 1;
736 prog
->constant_version
= priv
->next_constant_version
++;
740 static inline void update_heap_entry(struct constant_heap
*heap
, unsigned int idx
,
741 unsigned int heap_idx
, DWORD new_version
)
743 struct constant_entry
*entries
= heap
->entries
;
744 unsigned int *positions
= heap
->positions
;
745 unsigned int parent_idx
;
749 parent_idx
= heap_idx
>> 1;
751 if (new_version
<= entries
[parent_idx
].version
) break;
753 entries
[heap_idx
] = entries
[parent_idx
];
754 positions
[entries
[parent_idx
].idx
] = heap_idx
;
755 heap_idx
= parent_idx
;
758 entries
[heap_idx
].version
= new_version
;
759 entries
[heap_idx
].idx
= idx
;
760 positions
[idx
] = heap_idx
;
763 static void shader_glsl_update_float_vertex_constants(IWineD3DDevice
*iface
, UINT start
, UINT count
)
765 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
766 struct shader_glsl_priv
*priv
= This
->shader_priv
;
767 struct constant_heap
*heap
= &priv
->vconst_heap
;
770 for (i
= start
; i
< count
+ start
; ++i
)
772 if (!This
->stateBlock
->changed
.vertexShaderConstantsF
[i
])
773 update_heap_entry(heap
, i
, heap
->size
++, priv
->next_constant_version
);
775 update_heap_entry(heap
, i
, heap
->positions
[i
], priv
->next_constant_version
);
779 static void shader_glsl_update_float_pixel_constants(IWineD3DDevice
*iface
, UINT start
, UINT count
)
781 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
782 struct shader_glsl_priv
*priv
= This
->shader_priv
;
783 struct constant_heap
*heap
= &priv
->pconst_heap
;
786 for (i
= start
; i
< count
+ start
; ++i
)
788 if (!This
->stateBlock
->changed
.pixelShaderConstantsF
[i
])
789 update_heap_entry(heap
, i
, heap
->size
++, priv
->next_constant_version
);
791 update_heap_entry(heap
, i
, heap
->positions
[i
], priv
->next_constant_version
);
795 static unsigned int vec4_varyings(DWORD shader_major
, const struct wined3d_gl_info
*gl_info
)
797 unsigned int ret
= GL_LIMITS(glsl_varyings
) / 4;
798 /* 4.0 shaders do not write clip coords because d3d10 does not support user clipplanes */
799 if(shader_major
> 3) return ret
;
801 /* 3.0 shaders may need an extra varying for the clip coord on some cards(mostly dx10 ones) */
802 if (gl_info
->quirks
& WINED3D_QUIRK_GLSL_CLIP_VARYING
) ret
-= 1;
806 /** Generate the variable & register declarations for the GLSL output target */
807 static void shader_generate_glsl_declarations(const struct wined3d_context
*context
,
808 struct wined3d_shader_buffer
*buffer
, IWineD3DBaseShader
*iface
,
809 const shader_reg_maps
*reg_maps
, struct shader_glsl_ctx_priv
*ctx_priv
)
811 IWineD3DBaseShaderImpl
* This
= (IWineD3DBaseShaderImpl
*) iface
;
812 IWineD3DDeviceImpl
*device
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
813 const struct ps_compile_args
*ps_args
= ctx_priv
->cur_ps_args
;
814 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
815 unsigned int i
, extra_constants_needed
= 0;
816 const local_constant
*lconst
;
819 /* There are some minor differences between pixel and vertex shaders */
820 char pshader
= shader_is_pshader_version(reg_maps
->shader_version
.type
);
821 char prefix
= pshader
? 'P' : 'V';
823 /* Prototype the subroutines */
824 for (i
= 0, map
= reg_maps
->labels
; map
; map
>>= 1, ++i
)
826 if (map
& 1) shader_addline(buffer
, "void subroutine%u();\n", i
);
829 /* Declare the constants (aka uniforms) */
830 if (This
->baseShader
.limits
.constant_float
> 0) {
831 unsigned max_constantsF
;
832 /* Unless the shader uses indirect addressing, always declare the maximum array size and ignore that we need some
833 * uniforms privately. E.g. if GL supports 256 uniforms, and we need 2 for the pos fixup and immediate values, still
834 * declare VC[256]. If the shader needs more uniforms than we have it won't work in any case. If it uses less, the
835 * compiler will figure out which uniforms are really used and strip them out. This allows a shader to use c255 on
836 * a dx9 card, as long as it doesn't also use all the other constants.
838 * If the shader uses indirect addressing the compiler must assume that all declared uniforms are used. In this case,
839 * declare only the amount that we're assured to have.
841 * Thus we run into problems in these two cases:
842 * 1) The shader really uses more uniforms than supported
843 * 2) The shader uses indirect addressing, less constants than supported, but uses a constant index > #supported consts
846 /* No indirect addressing here */
847 max_constantsF
= GL_LIMITS(pshader_constantsF
);
849 if(This
->baseShader
.reg_maps
.usesrelconstF
) {
850 /* Subtract the other potential uniforms from the max available (bools, ints, and 1 row of projection matrix).
851 * Subtract another uniform for immediate values, which have to be loaded via uniform by the driver as well.
852 * The shader code only uses 0.5, 2.0, 1.0, 128 and -128 in vertex shader code, so one vec4 should be enough
853 * (Unfortunately the Nvidia driver doesn't store 128 and -128 in one float).
855 * Writing gl_ClipPos requires one uniform for each clipplane as well.
857 max_constantsF
= GL_LIMITS(vshader_constantsF
) - 3 - GL_LIMITS(clipplanes
);
858 max_constantsF
-= count_bits(This
->baseShader
.reg_maps
.integer_constants
);
859 /* Strictly speaking a bool only uses one scalar, but the nvidia(Linux) compiler doesn't pack them properly,
860 * so each scalar requires a full vec4. We could work around this by packing the booleans ourselves, but
861 * for now take this into account when calculating the number of available constants
863 max_constantsF
-= count_bits(This
->baseShader
.reg_maps
.boolean_constants
);
864 /* Set by driver quirks in directx.c */
865 max_constantsF
-= GLINFO_LOCATION
.reserved_glsl_constants
;
867 max_constantsF
= GL_LIMITS(vshader_constantsF
);
870 max_constantsF
= min(This
->baseShader
.limits
.constant_float
, max_constantsF
);
871 shader_addline(buffer
, "uniform vec4 %cC[%u];\n", prefix
, max_constantsF
);
874 /* Always declare the full set of constants, the compiler can remove the unused ones because d3d doesn't(yet)
875 * support indirect int and bool constant addressing. This avoids problems if the app uses e.g. i0 and i9.
877 if (This
->baseShader
.limits
.constant_int
> 0 && This
->baseShader
.reg_maps
.integer_constants
)
878 shader_addline(buffer
, "uniform ivec4 %cI[%u];\n", prefix
, This
->baseShader
.limits
.constant_int
);
880 if (This
->baseShader
.limits
.constant_bool
> 0 && This
->baseShader
.reg_maps
.boolean_constants
)
881 shader_addline(buffer
, "uniform bool %cB[%u];\n", prefix
, This
->baseShader
.limits
.constant_bool
);
884 shader_addline(buffer
, "uniform vec4 posFixup;\n");
885 /* Predeclaration; This function is added at link time based on the pixel shader.
886 * VS 3.0 shaders have an array OUT[] the shader writes to, earlier versions don't have
887 * that. We know the input to the reorder function at vertex shader compile time, so
888 * we can deal with that. The reorder function for a 1.x and 2.x vertex shader can just
889 * read gl_FrontColor. The output depends on the pixel shader. The reorder function for a
890 * 1.x and 2.x pshader or for fixed function will write gl_FrontColor, and for a 3.0 shader
891 * it will write to the varying array. Here we depend on the shader optimizer on sorting that
892 * out. The nvidia driver only does that if the parameter is inout instead of out, hence the
895 if (reg_maps
->shader_version
.major
>= 3)
897 shader_addline(buffer
, "void order_ps_input(in vec4[%u]);\n", MAX_REG_OUTPUT
);
899 shader_addline(buffer
, "void order_ps_input();\n");
902 for (i
= 0, map
= reg_maps
->bumpmat
; map
; map
>>= 1, ++i
)
904 if (!(map
& 1)) continue;
906 shader_addline(buffer
, "uniform mat2 bumpenvmat%d;\n", i
);
908 if (reg_maps
->luminanceparams
& (1 << i
))
910 shader_addline(buffer
, "uniform float luminancescale%d;\n", i
);
911 shader_addline(buffer
, "uniform float luminanceoffset%d;\n", i
);
912 extra_constants_needed
++;
915 extra_constants_needed
++;
918 if (ps_args
->srgb_correction
)
920 shader_addline(buffer
, "const vec4 srgb_const0 = vec4(%.8e, %.8e, %.8e, %.8e);\n",
921 srgb_pow
, srgb_mul_high
, srgb_sub_high
, srgb_mul_low
);
922 shader_addline(buffer
, "const vec4 srgb_const1 = vec4(%.8e, 0.0, 0.0, 0.0);\n",
925 if(reg_maps
->vpos
|| reg_maps
->usesdsy
) {
926 if(This
->baseShader
.limits
.constant_float
+ extra_constants_needed
+ 1 < GL_LIMITS(pshader_constantsF
)) {
927 shader_addline(buffer
, "uniform vec4 ycorrection;\n");
928 ((IWineD3DPixelShaderImpl
*) This
)->vpos_uniform
= 1;
929 extra_constants_needed
++;
931 /* This happens because we do not have proper tracking of the constant registers that are
932 * actually used, only the max limit of the shader version
934 FIXME("Cannot find a free uniform for vpos correction params\n");
935 shader_addline(buffer
, "const vec4 ycorrection = vec4(%f, %f, 0.0, 0.0);\n",
936 context
->render_offscreen
? 0.0f
: ((IWineD3DSurfaceImpl
*)device
->render_targets
[0])->currentDesc
.Height
,
937 context
->render_offscreen
? 1.0f
: -1.0f
);
939 shader_addline(buffer
, "vec4 vpos;\n");
943 /* Declare texture samplers */
944 for (i
= 0; i
< This
->baseShader
.limits
.sampler
; i
++) {
945 if (reg_maps
->sampler_type
[i
])
947 switch (reg_maps
->sampler_type
[i
])
950 shader_addline(buffer
, "uniform sampler1D %csampler%u;\n", prefix
, i
);
953 if(device
->stateBlock
->textures
[i
] &&
954 IWineD3DBaseTexture_GetTextureDimensions(device
->stateBlock
->textures
[i
]) == GL_TEXTURE_RECTANGLE_ARB
) {
955 shader_addline(buffer
, "uniform sampler2DRect %csampler%u;\n", prefix
, i
);
957 shader_addline(buffer
, "uniform sampler2D %csampler%u;\n", prefix
, i
);
960 case WINED3DSTT_CUBE
:
961 shader_addline(buffer
, "uniform samplerCube %csampler%u;\n", prefix
, i
);
963 case WINED3DSTT_VOLUME
:
964 shader_addline(buffer
, "uniform sampler3D %csampler%u;\n", prefix
, i
);
967 shader_addline(buffer
, "uniform unsupported_sampler %csampler%u;\n", prefix
, i
);
968 FIXME("Unrecognized sampler type: %#x\n", reg_maps
->sampler_type
[i
]);
974 /* Declare uniforms for NP2 texcoord fixup:
975 * This is NOT done inside the loop that declares the texture samplers since the NP2 fixup code
976 * is currently only used for the GeforceFX series and when forcing the ARB_npot extension off.
977 * Modern cards just skip the code anyway, so put it inside a separate loop. */
978 if (pshader
&& ps_args
->np2_fixup
) {
980 struct ps_np2fixup_info
* const fixup
= ctx_priv
->cur_np2fixup_info
;
983 /* NP2/RECT textures in OpenGL use texcoords in the range [0,width]x[0,height]
984 * while D3D has them in the (normalized) [0,1]x[0,1] range.
985 * samplerNP2Fixup stores texture dimensions and is updated through
986 * shader_glsl_load_np2fixup_constants when the sampler changes. */
988 for (i
= 0; i
< This
->baseShader
.limits
.sampler
; ++i
) {
989 if (reg_maps
->sampler_type
[i
]) {
990 if (!(ps_args
->np2_fixup
& (1 << i
))) continue;
992 if (WINED3DSTT_2D
!= reg_maps
->sampler_type
[i
]) {
993 FIXME("Non-2D texture is flagged for NP2 texcoord fixup.\n");
997 fixup
->idx
[i
] = cur
++;
1001 fixup
->num_consts
= (cur
+ 1) >> 1;
1002 shader_addline(buffer
, "uniform vec4 %csamplerNP2Fixup[%u];\n", prefix
, fixup
->num_consts
);
1005 /* Declare address variables */
1006 for (i
= 0, map
= reg_maps
->address
; map
; map
>>= 1, ++i
)
1008 if (map
& 1) shader_addline(buffer
, "ivec4 A%u;\n", i
);
1011 /* Declare texture coordinate temporaries and initialize them */
1012 for (i
= 0, map
= reg_maps
->texcoord
; map
; map
>>= 1, ++i
)
1014 if (map
& 1) shader_addline(buffer
, "vec4 T%u = gl_TexCoord[%u];\n", i
, i
);
1017 /* Declare input register varyings. Only pixel shader, vertex shaders have that declared in the
1018 * helper function shader that is linked in at link time
1020 if (pshader
&& reg_maps
->shader_version
.major
>= 3)
1022 if (use_vs(device
->stateBlock
))
1024 shader_addline(buffer
, "varying vec4 IN[%u];\n", vec4_varyings(reg_maps
->shader_version
.major
, gl_info
));
1026 /* TODO: Write a replacement shader for the fixed function vertex pipeline, so this isn't needed.
1027 * For fixed function vertex processing + 3.0 pixel shader we need a separate function in the
1028 * pixel shader that reads the fixed function color into the packed input registers.
1030 shader_addline(buffer
, "vec4 IN[%u];\n", vec4_varyings(reg_maps
->shader_version
.major
, gl_info
));
1034 /* Declare output register temporaries */
1035 if(This
->baseShader
.limits
.packed_output
) {
1036 shader_addline(buffer
, "vec4 OUT[%u];\n", This
->baseShader
.limits
.packed_output
);
1039 /* Declare temporary variables */
1040 for (i
= 0, map
= reg_maps
->temporary
; map
; map
>>= 1, ++i
)
1042 if (map
& 1) shader_addline(buffer
, "vec4 R%u;\n", i
);
1045 /* Declare attributes */
1046 if (reg_maps
->shader_version
.type
== WINED3D_SHADER_TYPE_VERTEX
)
1048 for (i
= 0, map
= reg_maps
->input_registers
; map
; map
>>= 1, ++i
)
1050 if (map
& 1) shader_addline(buffer
, "attribute vec4 attrib%i;\n", i
);
1054 /* Declare loop registers aLx */
1055 for (i
= 0; i
< reg_maps
->loop_depth
; i
++) {
1056 shader_addline(buffer
, "int aL%u;\n", i
);
1057 shader_addline(buffer
, "int tmpInt%u;\n", i
);
1060 /* Temporary variables for matrix operations */
1061 shader_addline(buffer
, "vec4 tmp0;\n");
1062 shader_addline(buffer
, "vec4 tmp1;\n");
1064 /* Local constants use a different name so they can be loaded once at shader link time
1065 * They can't be hardcoded into the shader text via LC = {x, y, z, w}; because the
1066 * float -> string conversion can cause precision loss.
1068 if(!This
->baseShader
.load_local_constsF
) {
1069 LIST_FOR_EACH_ENTRY(lconst
, &This
->baseShader
.constantsF
, local_constant
, entry
) {
1070 shader_addline(buffer
, "uniform vec4 %cLC%u;\n", prefix
, lconst
->idx
);
1074 /* Start the main program */
1075 shader_addline(buffer
, "void main() {\n");
1076 if(pshader
&& reg_maps
->vpos
) {
1077 /* DirectX apps expect integer values, while OpenGL drivers add approximately 0.5. This causes
1078 * off-by-one problems as spotted by the vPos d3d9 visual test. Unfortunately the ATI cards do
1079 * not add exactly 0.5, but rather something like 0.49999999 or 0.50000001, which still causes
1080 * precision troubles when we just substract 0.5.
1082 * To deal with that just floor() the position. This will eliminate the fraction on all cards.
1084 * TODO: Test how that behaves with multisampling once we can enable multisampling in winex11.
1086 * An advantage of floor is that it works even if the driver doesn't add 1/2. It is somewhat
1087 * questionable if 1.5, 2.5, ... are the proper values to return in gl_FragCoord, even though
1088 * coordinates specify the pixel centers instead of the pixel corners. This code will behave
1089 * correctly on drivers that returns integer values.
1091 shader_addline(buffer
, "vpos = floor(vec4(0, ycorrection[0], 0, 0) + gl_FragCoord * vec4(1, ycorrection[1], 1, 1));\n");
1095 /*****************************************************************************
1096 * Functions to generate GLSL strings from DirectX Shader bytecode begin here.
1098 * For more information, see http://wiki.winehq.org/DirectX-Shaders
1099 ****************************************************************************/
1102 static void shader_glsl_add_src_param(const struct wined3d_shader_instruction
*ins
,
1103 const struct wined3d_shader_src_param
*wined3d_src
, DWORD mask
, glsl_src_param_t
*glsl_src
);
1105 /** Used for opcode modifiers - They multiply the result by the specified amount */
1106 static const char * const shift_glsl_tab
[] = {
1108 "2.0 * ", /* 1 (x2) */
1109 "4.0 * ", /* 2 (x4) */
1110 "8.0 * ", /* 3 (x8) */
1111 "16.0 * ", /* 4 (x16) */
1112 "32.0 * ", /* 5 (x32) */
1119 "0.0625 * ", /* 12 (d16) */
1120 "0.125 * ", /* 13 (d8) */
1121 "0.25 * ", /* 14 (d4) */
1122 "0.5 * " /* 15 (d2) */
1125 /* Generate a GLSL parameter that does the input modifier computation and return the input register/mask to use */
1126 static void shader_glsl_gen_modifier(DWORD src_modifier
, const char *in_reg
, const char *in_regswizzle
, char *out_str
)
1130 switch (src_modifier
)
1132 case WINED3DSPSM_DZ
: /* Need to handle this in the instructions itself (texld & texcrd). */
1133 case WINED3DSPSM_DW
:
1134 case WINED3DSPSM_NONE
:
1135 sprintf(out_str
, "%s%s", in_reg
, in_regswizzle
);
1137 case WINED3DSPSM_NEG
:
1138 sprintf(out_str
, "-%s%s", in_reg
, in_regswizzle
);
1140 case WINED3DSPSM_NOT
:
1141 sprintf(out_str
, "!%s%s", in_reg
, in_regswizzle
);
1143 case WINED3DSPSM_BIAS
:
1144 sprintf(out_str
, "(%s%s - vec4(0.5)%s)", in_reg
, in_regswizzle
, in_regswizzle
);
1146 case WINED3DSPSM_BIASNEG
:
1147 sprintf(out_str
, "-(%s%s - vec4(0.5)%s)", in_reg
, in_regswizzle
, in_regswizzle
);
1149 case WINED3DSPSM_SIGN
:
1150 sprintf(out_str
, "(2.0 * (%s%s - 0.5))", in_reg
, in_regswizzle
);
1152 case WINED3DSPSM_SIGNNEG
:
1153 sprintf(out_str
, "-(2.0 * (%s%s - 0.5))", in_reg
, in_regswizzle
);
1155 case WINED3DSPSM_COMP
:
1156 sprintf(out_str
, "(1.0 - %s%s)", in_reg
, in_regswizzle
);
1158 case WINED3DSPSM_X2
:
1159 sprintf(out_str
, "(2.0 * %s%s)", in_reg
, in_regswizzle
);
1161 case WINED3DSPSM_X2NEG
:
1162 sprintf(out_str
, "-(2.0 * %s%s)", in_reg
, in_regswizzle
);
1164 case WINED3DSPSM_ABS
:
1165 sprintf(out_str
, "abs(%s%s)", in_reg
, in_regswizzle
);
1167 case WINED3DSPSM_ABSNEG
:
1168 sprintf(out_str
, "-abs(%s%s)", in_reg
, in_regswizzle
);
1171 FIXME("Unhandled modifier %u\n", src_modifier
);
1172 sprintf(out_str
, "%s%s", in_reg
, in_regswizzle
);
1176 /** Writes the GLSL variable name that corresponds to the register that the
1177 * DX opcode parameter is trying to access */
1178 static void shader_glsl_get_register_name(const struct wined3d_shader_register
*reg
,
1179 char *register_name
, BOOL
*is_color
, const struct wined3d_shader_instruction
*ins
)
1181 /* oPos, oFog and oPts in D3D */
1182 static const char * const hwrastout_reg_names
[] = { "gl_Position", "gl_FogFragCoord", "gl_PointSize" };
1184 IWineD3DBaseShaderImpl
*This
= (IWineD3DBaseShaderImpl
*)ins
->ctx
->shader
;
1185 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
1186 const struct wined3d_gl_info
*gl_info
= &deviceImpl
->adapter
->gl_info
;
1187 char pshader
= shader_is_pshader_version(This
->baseShader
.reg_maps
.shader_version
.type
);
1193 case WINED3DSPR_TEMP
:
1194 sprintf(register_name
, "R%u", reg
->idx
);
1197 case WINED3DSPR_INPUT
:
1198 /* vertex shaders */
1201 struct shader_glsl_ctx_priv
*priv
= ins
->ctx
->backend_data
;
1202 if (priv
->cur_vs_args
->swizzle_map
& (1 << reg
->idx
)) *is_color
= TRUE
;
1203 sprintf(register_name
, "attrib%u", reg
->idx
);
1207 /* pixel shaders >= 3.0 */
1208 if (This
->baseShader
.reg_maps
.shader_version
.major
>= 3)
1210 DWORD idx
= ((IWineD3DPixelShaderImpl
*)This
)->input_reg_map
[reg
->idx
];
1211 unsigned int in_count
= vec4_varyings(This
->baseShader
.reg_maps
.shader_version
.major
, gl_info
);
1215 glsl_src_param_t rel_param
;
1217 shader_glsl_add_src_param(ins
, reg
->rel_addr
, WINED3DSP_WRITEMASK_0
, &rel_param
);
1219 /* Removing a + 0 would be an obvious optimization, but macos doesn't see the NOP
1220 * operation there */
1223 if (((IWineD3DPixelShaderImpl
*)This
)->declared_in_count
> in_count
)
1225 sprintf(register_name
,
1226 "((%s + %u) > %d ? (%s + %u) > %d ? gl_SecondaryColor : gl_Color : IN[%s + %u])",
1227 rel_param
.param_str
, idx
, in_count
- 1, rel_param
.param_str
, idx
, in_count
,
1228 rel_param
.param_str
, idx
);
1232 sprintf(register_name
, "IN[%s + %u]", rel_param
.param_str
, idx
);
1237 if (((IWineD3DPixelShaderImpl
*)This
)->declared_in_count
> in_count
)
1239 sprintf(register_name
, "((%s) > %d ? (%s) > %d ? gl_SecondaryColor : gl_Color : IN[%s])",
1240 rel_param
.param_str
, in_count
- 1, rel_param
.param_str
, in_count
,
1241 rel_param
.param_str
);
1245 sprintf(register_name
, "IN[%s]", rel_param
.param_str
);
1251 if (idx
== in_count
) sprintf(register_name
, "gl_Color");
1252 else if (idx
== in_count
+ 1) sprintf(register_name
, "gl_SecondaryColor");
1253 else sprintf(register_name
, "IN[%u]", idx
);
1258 if (reg
->idx
== 0) strcpy(register_name
, "gl_Color");
1259 else strcpy(register_name
, "gl_SecondaryColor");
1264 case WINED3DSPR_CONST
:
1266 const char prefix
= pshader
? 'P' : 'V';
1268 /* Relative addressing */
1271 glsl_src_param_t rel_param
;
1272 shader_glsl_add_src_param(ins
, reg
->rel_addr
, WINED3DSP_WRITEMASK_0
, &rel_param
);
1273 if (reg
->idx
) sprintf(register_name
, "%cC[%s + %u]", prefix
, rel_param
.param_str
, reg
->idx
);
1274 else sprintf(register_name
, "%cC[%s]", prefix
, rel_param
.param_str
);
1278 if (shader_constant_is_local(This
, reg
->idx
))
1279 sprintf(register_name
, "%cLC%u", prefix
, reg
->idx
);
1281 sprintf(register_name
, "%cC[%u]", prefix
, reg
->idx
);
1286 case WINED3DSPR_CONSTINT
:
1287 if (pshader
) sprintf(register_name
, "PI[%u]", reg
->idx
);
1288 else sprintf(register_name
, "VI[%u]", reg
->idx
);
1291 case WINED3DSPR_CONSTBOOL
:
1292 if (pshader
) sprintf(register_name
, "PB[%u]", reg
->idx
);
1293 else sprintf(register_name
, "VB[%u]", reg
->idx
);
1296 case WINED3DSPR_TEXTURE
: /* case WINED3DSPR_ADDR: */
1297 if (pshader
) sprintf(register_name
, "T%u", reg
->idx
);
1298 else sprintf(register_name
, "A%u", reg
->idx
);
1301 case WINED3DSPR_LOOP
:
1302 sprintf(register_name
, "aL%u", This
->baseShader
.cur_loop_regno
- 1);
1305 case WINED3DSPR_SAMPLER
:
1306 if (pshader
) sprintf(register_name
, "Psampler%u", reg
->idx
);
1307 else sprintf(register_name
, "Vsampler%u", reg
->idx
);
1310 case WINED3DSPR_COLOROUT
:
1311 if (reg
->idx
>= GL_LIMITS(buffers
))
1312 WARN("Write to render target %u, only %d supported\n", reg
->idx
, GL_LIMITS(buffers
));
1314 sprintf(register_name
, "gl_FragData[%u]", reg
->idx
);
1317 case WINED3DSPR_RASTOUT
:
1318 sprintf(register_name
, "%s", hwrastout_reg_names
[reg
->idx
]);
1321 case WINED3DSPR_DEPTHOUT
:
1322 sprintf(register_name
, "gl_FragDepth");
1325 case WINED3DSPR_ATTROUT
:
1326 if (reg
->idx
== 0) sprintf(register_name
, "gl_FrontColor");
1327 else sprintf(register_name
, "gl_FrontSecondaryColor");
1330 case WINED3DSPR_TEXCRDOUT
:
1331 /* Vertex shaders >= 3.0: WINED3DSPR_OUTPUT */
1332 if (This
->baseShader
.reg_maps
.shader_version
.major
>= 3) sprintf(register_name
, "OUT[%u]", reg
->idx
);
1333 else sprintf(register_name
, "gl_TexCoord[%u]", reg
->idx
);
1336 case WINED3DSPR_MISCTYPE
:
1340 sprintf(register_name
, "vpos");
1342 else if (reg
->idx
== 1)
1344 /* Note that gl_FrontFacing is a bool, while vFace is
1345 * a float for which the sign determines front/back */
1346 sprintf(register_name
, "(gl_FrontFacing ? 1.0 : -1.0)");
1350 FIXME("Unhandled misctype register %d\n", reg
->idx
);
1351 sprintf(register_name
, "unrecognized_register");
1355 case WINED3DSPR_IMMCONST
:
1356 switch (reg
->immconst_type
)
1358 case WINED3D_IMMCONST_FLOAT
:
1359 sprintf(register_name
, "%.8e", *(const float *)reg
->immconst_data
);
1362 case WINED3D_IMMCONST_FLOAT4
:
1363 sprintf(register_name
, "vec4(%.8e, %.8e, %.8e, %.8e)",
1364 *(const float *)®
->immconst_data
[0], *(const float *)®
->immconst_data
[1],
1365 *(const float *)®
->immconst_data
[2], *(const float *)®
->immconst_data
[3]);
1369 FIXME("Unhandled immconst type %#x\n", reg
->immconst_type
);
1370 sprintf(register_name
, "<unhandled_immconst_type %#x>", reg
->immconst_type
);
1375 FIXME("Unhandled register name Type(%d)\n", reg
->type
);
1376 sprintf(register_name
, "unrecognized_register");
1381 static void shader_glsl_write_mask_to_str(DWORD write_mask
, char *str
)
1384 if (write_mask
& WINED3DSP_WRITEMASK_0
) *str
++ = 'x';
1385 if (write_mask
& WINED3DSP_WRITEMASK_1
) *str
++ = 'y';
1386 if (write_mask
& WINED3DSP_WRITEMASK_2
) *str
++ = 'z';
1387 if (write_mask
& WINED3DSP_WRITEMASK_3
) *str
++ = 'w';
1391 /* Get the GLSL write mask for the destination register */
1392 static DWORD
shader_glsl_get_write_mask(const struct wined3d_shader_dst_param
*param
, char *write_mask
)
1394 DWORD mask
= param
->write_mask
;
1396 if (shader_is_scalar(¶m
->reg
))
1398 mask
= WINED3DSP_WRITEMASK_0
;
1403 shader_glsl_write_mask_to_str(mask
, write_mask
);
1409 static unsigned int shader_glsl_get_write_mask_size(DWORD write_mask
) {
1410 unsigned int size
= 0;
1412 if (write_mask
& WINED3DSP_WRITEMASK_0
) ++size
;
1413 if (write_mask
& WINED3DSP_WRITEMASK_1
) ++size
;
1414 if (write_mask
& WINED3DSP_WRITEMASK_2
) ++size
;
1415 if (write_mask
& WINED3DSP_WRITEMASK_3
) ++size
;
1420 static void shader_glsl_swizzle_to_str(const DWORD swizzle
, BOOL fixup
, DWORD mask
, char *str
)
1422 /* For registers of type WINED3DDECLTYPE_D3DCOLOR, data is stored as "bgra",
1423 * but addressed as "rgba". To fix this we need to swap the register's x
1424 * and z components. */
1425 const char *swizzle_chars
= fixup
? "zyxw" : "xyzw";
1428 /* swizzle bits fields: wwzzyyxx */
1429 if (mask
& WINED3DSP_WRITEMASK_0
) *str
++ = swizzle_chars
[swizzle
& 0x03];
1430 if (mask
& WINED3DSP_WRITEMASK_1
) *str
++ = swizzle_chars
[(swizzle
>> 2) & 0x03];
1431 if (mask
& WINED3DSP_WRITEMASK_2
) *str
++ = swizzle_chars
[(swizzle
>> 4) & 0x03];
1432 if (mask
& WINED3DSP_WRITEMASK_3
) *str
++ = swizzle_chars
[(swizzle
>> 6) & 0x03];
1436 static void shader_glsl_get_swizzle(const struct wined3d_shader_src_param
*param
,
1437 BOOL fixup
, DWORD mask
, char *swizzle_str
)
1439 if (shader_is_scalar(¶m
->reg
))
1440 *swizzle_str
= '\0';
1442 shader_glsl_swizzle_to_str(param
->swizzle
, fixup
, mask
, swizzle_str
);
1445 /* From a given parameter token, generate the corresponding GLSL string.
1446 * Also, return the actual register name and swizzle in case the
1447 * caller needs this information as well. */
1448 static void shader_glsl_add_src_param(const struct wined3d_shader_instruction
*ins
,
1449 const struct wined3d_shader_src_param
*wined3d_src
, DWORD mask
, glsl_src_param_t
*glsl_src
)
1451 BOOL is_color
= FALSE
;
1452 char swizzle_str
[6];
1454 glsl_src
->reg_name
[0] = '\0';
1455 glsl_src
->param_str
[0] = '\0';
1456 swizzle_str
[0] = '\0';
1458 shader_glsl_get_register_name(&wined3d_src
->reg
, glsl_src
->reg_name
, &is_color
, ins
);
1459 shader_glsl_get_swizzle(wined3d_src
, is_color
, mask
, swizzle_str
);
1460 shader_glsl_gen_modifier(wined3d_src
->modifiers
, glsl_src
->reg_name
, swizzle_str
, glsl_src
->param_str
);
1463 /* From a given parameter token, generate the corresponding GLSL string.
1464 * Also, return the actual register name and swizzle in case the
1465 * caller needs this information as well. */
1466 static DWORD
shader_glsl_add_dst_param(const struct wined3d_shader_instruction
*ins
,
1467 const struct wined3d_shader_dst_param
*wined3d_dst
, glsl_dst_param_t
*glsl_dst
)
1469 BOOL is_color
= FALSE
;
1471 glsl_dst
->mask_str
[0] = '\0';
1472 glsl_dst
->reg_name
[0] = '\0';
1474 shader_glsl_get_register_name(&wined3d_dst
->reg
, glsl_dst
->reg_name
, &is_color
, ins
);
1475 return shader_glsl_get_write_mask(wined3d_dst
, glsl_dst
->mask_str
);
1478 /* Append the destination part of the instruction to the buffer, return the effective write mask */
1479 static DWORD
shader_glsl_append_dst_ext(struct wined3d_shader_buffer
*buffer
,
1480 const struct wined3d_shader_instruction
*ins
, const struct wined3d_shader_dst_param
*dst
)
1482 glsl_dst_param_t glsl_dst
;
1485 mask
= shader_glsl_add_dst_param(ins
, dst
, &glsl_dst
);
1486 if (mask
) shader_addline(buffer
, "%s%s = %s(", glsl_dst
.reg_name
, glsl_dst
.mask_str
, shift_glsl_tab
[dst
->shift
]);
1491 /* Append the destination part of the instruction to the buffer, return the effective write mask */
1492 static DWORD
shader_glsl_append_dst(struct wined3d_shader_buffer
*buffer
, const struct wined3d_shader_instruction
*ins
)
1494 return shader_glsl_append_dst_ext(buffer
, ins
, &ins
->dst
[0]);
1497 /** Process GLSL instruction modifiers */
1498 static void shader_glsl_add_instruction_modifiers(const struct wined3d_shader_instruction
*ins
)
1500 glsl_dst_param_t dst_param
;
1503 if (!ins
->dst_count
) return;
1505 modifiers
= ins
->dst
[0].modifiers
;
1506 if (!modifiers
) return;
1508 shader_glsl_add_dst_param(ins
, &ins
->dst
[0], &dst_param
);
1510 if (modifiers
& WINED3DSPDM_SATURATE
)
1512 /* _SAT means to clamp the value of the register to between 0 and 1 */
1513 shader_addline(ins
->ctx
->buffer
, "%s%s = clamp(%s%s, 0.0, 1.0);\n", dst_param
.reg_name
,
1514 dst_param
.mask_str
, dst_param
.reg_name
, dst_param
.mask_str
);
1517 if (modifiers
& WINED3DSPDM_MSAMPCENTROID
)
1519 FIXME("_centroid modifier not handled\n");
1522 if (modifiers
& WINED3DSPDM_PARTIALPRECISION
)
1524 /* MSDN says this modifier can be safely ignored, so that's what we'll do. */
1528 static inline const char *shader_get_comp_op(DWORD op
)
1531 case COMPARISON_GT
: return ">";
1532 case COMPARISON_EQ
: return "==";
1533 case COMPARISON_GE
: return ">=";
1534 case COMPARISON_LT
: return "<";
1535 case COMPARISON_NE
: return "!=";
1536 case COMPARISON_LE
: return "<=";
1538 FIXME("Unrecognized comparison value: %u\n", op
);
1543 static void shader_glsl_get_sample_function(DWORD sampler_type
, DWORD flags
, glsl_sample_function_t
*sample_function
)
1545 BOOL projected
= flags
& WINED3D_GLSL_SAMPLE_PROJECTED
;
1546 BOOL texrect
= flags
& WINED3D_GLSL_SAMPLE_RECT
;
1547 BOOL lod
= flags
& WINED3D_GLSL_SAMPLE_LOD
;
1548 BOOL grad
= flags
& WINED3D_GLSL_SAMPLE_GRAD
;
1550 /* Note that there's no such thing as a projected cube texture. */
1551 switch(sampler_type
) {
1554 sample_function
->name
= projected
? "texture1DProjLod" : "texture1DLod";
1556 sample_function
->name
= projected
? "texture1DProjGradARB" : "texture1DGradARB";
1558 sample_function
->name
= projected
? "texture1DProj" : "texture1D";
1560 sample_function
->coord_mask
= WINED3DSP_WRITEMASK_0
;
1565 sample_function
->name
= projected
? "texture2DRectProjLod" : "texture2DRectLod";
1567 /* What good are texrect grad functions? I don't know, but GL_EXT_gpu_shader4 defines them.
1568 * There is no GL_ARB_shader_texture_lod spec yet, so I don't know if they're defined there
1570 sample_function
->name
= projected
? "shadow2DRectProjGradARB" : "shadow2DRectGradARB";
1572 sample_function
->name
= projected
? "texture2DRectProj" : "texture2DRect";
1576 sample_function
->name
= projected
? "texture2DProjLod" : "texture2DLod";
1578 sample_function
->name
= projected
? "texture2DProjGradARB" : "texture2DGradARB";
1580 sample_function
->name
= projected
? "texture2DProj" : "texture2D";
1583 sample_function
->coord_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
;
1585 case WINED3DSTT_CUBE
:
1587 sample_function
->name
= "textureCubeLod";
1589 sample_function
->name
= "textureCubeGradARB";
1591 sample_function
->name
= "textureCube";
1593 sample_function
->coord_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
1595 case WINED3DSTT_VOLUME
:
1597 sample_function
->name
= projected
? "texture3DProjLod" : "texture3DLod";
1599 sample_function
->name
= projected
? "texture3DProjGradARB" : "texture3DGradARB";
1601 sample_function
->name
= projected
? "texture3DProj" : "texture3D";
1603 sample_function
->coord_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
1606 sample_function
->name
= "";
1607 sample_function
->coord_mask
= 0;
1608 FIXME("Unrecognized sampler type: %#x;\n", sampler_type
);
1613 static void shader_glsl_append_fixup_arg(char *arguments
, const char *reg_name
,
1614 BOOL sign_fixup
, enum fixup_channel_source channel_source
)
1616 switch(channel_source
)
1618 case CHANNEL_SOURCE_ZERO
:
1619 strcat(arguments
, "0.0");
1622 case CHANNEL_SOURCE_ONE
:
1623 strcat(arguments
, "1.0");
1626 case CHANNEL_SOURCE_X
:
1627 strcat(arguments
, reg_name
);
1628 strcat(arguments
, ".x");
1631 case CHANNEL_SOURCE_Y
:
1632 strcat(arguments
, reg_name
);
1633 strcat(arguments
, ".y");
1636 case CHANNEL_SOURCE_Z
:
1637 strcat(arguments
, reg_name
);
1638 strcat(arguments
, ".z");
1641 case CHANNEL_SOURCE_W
:
1642 strcat(arguments
, reg_name
);
1643 strcat(arguments
, ".w");
1647 FIXME("Unhandled channel source %#x\n", channel_source
);
1648 strcat(arguments
, "undefined");
1652 if (sign_fixup
) strcat(arguments
, " * 2.0 - 1.0");
1655 static void shader_glsl_color_correction(const struct wined3d_shader_instruction
*ins
, struct color_fixup_desc fixup
)
1657 struct wined3d_shader_dst_param dst
;
1658 unsigned int mask_size
, remaining
;
1659 glsl_dst_param_t dst_param
;
1660 char arguments
[256];
1664 if (fixup
.x_sign_fixup
|| fixup
.x_source
!= CHANNEL_SOURCE_X
) mask
|= WINED3DSP_WRITEMASK_0
;
1665 if (fixup
.y_sign_fixup
|| fixup
.y_source
!= CHANNEL_SOURCE_Y
) mask
|= WINED3DSP_WRITEMASK_1
;
1666 if (fixup
.z_sign_fixup
|| fixup
.z_source
!= CHANNEL_SOURCE_Z
) mask
|= WINED3DSP_WRITEMASK_2
;
1667 if (fixup
.w_sign_fixup
|| fixup
.w_source
!= CHANNEL_SOURCE_W
) mask
|= WINED3DSP_WRITEMASK_3
;
1668 mask
&= ins
->dst
[0].write_mask
;
1670 if (!mask
) return; /* Nothing to do */
1672 if (is_yuv_fixup(fixup
))
1674 enum yuv_fixup yuv_fixup
= get_yuv_fixup(fixup
);
1675 FIXME("YUV fixup (%#x) not supported\n", yuv_fixup
);
1679 mask_size
= shader_glsl_get_write_mask_size(mask
);
1682 dst
.write_mask
= mask
;
1683 shader_glsl_add_dst_param(ins
, &dst
, &dst_param
);
1685 arguments
[0] = '\0';
1686 remaining
= mask_size
;
1687 if (mask
& WINED3DSP_WRITEMASK_0
)
1689 shader_glsl_append_fixup_arg(arguments
, dst_param
.reg_name
, fixup
.x_sign_fixup
, fixup
.x_source
);
1690 if (--remaining
) strcat(arguments
, ", ");
1692 if (mask
& WINED3DSP_WRITEMASK_1
)
1694 shader_glsl_append_fixup_arg(arguments
, dst_param
.reg_name
, fixup
.y_sign_fixup
, fixup
.y_source
);
1695 if (--remaining
) strcat(arguments
, ", ");
1697 if (mask
& WINED3DSP_WRITEMASK_2
)
1699 shader_glsl_append_fixup_arg(arguments
, dst_param
.reg_name
, fixup
.z_sign_fixup
, fixup
.z_source
);
1700 if (--remaining
) strcat(arguments
, ", ");
1702 if (mask
& WINED3DSP_WRITEMASK_3
)
1704 shader_glsl_append_fixup_arg(arguments
, dst_param
.reg_name
, fixup
.w_sign_fixup
, fixup
.w_source
);
1705 if (--remaining
) strcat(arguments
, ", ");
1710 shader_addline(ins
->ctx
->buffer
, "%s%s = vec%u(%s);\n",
1711 dst_param
.reg_name
, dst_param
.mask_str
, mask_size
, arguments
);
1715 shader_addline(ins
->ctx
->buffer
, "%s%s = %s;\n", dst_param
.reg_name
, dst_param
.mask_str
, arguments
);
1719 static void PRINTF_ATTR(8, 9) shader_glsl_gen_sample_code(const struct wined3d_shader_instruction
*ins
,
1720 DWORD sampler
, const glsl_sample_function_t
*sample_function
, DWORD swizzle
,
1721 const char *dx
, const char *dy
,
1722 const char *bias
, const char *coord_reg_fmt
, ...)
1724 const char *sampler_base
;
1725 char dst_swizzle
[6];
1726 struct color_fixup_desc fixup
;
1727 BOOL np2_fixup
= FALSE
;
1730 shader_glsl_swizzle_to_str(swizzle
, FALSE
, ins
->dst
[0].write_mask
, dst_swizzle
);
1732 if (shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
.type
))
1734 const struct shader_glsl_ctx_priv
*priv
= ins
->ctx
->backend_data
;
1735 fixup
= priv
->cur_ps_args
->color_fixup
[sampler
];
1736 sampler_base
= "Psampler";
1738 if(priv
->cur_ps_args
->np2_fixup
& (1 << sampler
)) {
1740 FIXME("Biased sampling from NP2 textures is unsupported\n");
1746 sampler_base
= "Vsampler";
1747 fixup
= COLOR_FIXUP_IDENTITY
; /* FIXME: Vshader color fixup */
1750 shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
1752 shader_addline(ins
->ctx
->buffer
, "%s(%s%u, ", sample_function
->name
, sampler_base
, sampler
);
1754 va_start(args
, coord_reg_fmt
);
1755 shader_vaddline(ins
->ctx
->buffer
, coord_reg_fmt
, args
);
1759 shader_addline(ins
->ctx
->buffer
, ", %s)%s);\n", bias
, dst_swizzle
);
1762 const struct shader_glsl_ctx_priv
*priv
= ins
->ctx
->backend_data
;
1763 const unsigned char idx
= priv
->cur_np2fixup_info
->idx
[sampler
];
1765 shader_addline(ins
->ctx
->buffer
, " * PsamplerNP2Fixup[%u].%s)%s);\n", idx
>> 1,
1766 (idx
% 2) ? "zw" : "xy", dst_swizzle
);
1767 } else if(dx
&& dy
) {
1768 shader_addline(ins
->ctx
->buffer
, ", %s, %s)%s);\n", dx
, dy
, dst_swizzle
);
1770 shader_addline(ins
->ctx
->buffer
, ")%s);\n", dst_swizzle
);
1774 if(!is_identity_fixup(fixup
)) {
1775 shader_glsl_color_correction(ins
, fixup
);
1779 /*****************************************************************************
1780 * Begin processing individual instruction opcodes
1781 ****************************************************************************/
1783 /* Generate GLSL arithmetic functions (dst = src1 + src2) */
1784 static void shader_glsl_arith(const struct wined3d_shader_instruction
*ins
)
1786 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
1787 glsl_src_param_t src0_param
;
1788 glsl_src_param_t src1_param
;
1792 /* Determine the GLSL operator to use based on the opcode */
1793 switch (ins
->handler_idx
)
1795 case WINED3DSIH_MUL
: op
= '*'; break;
1796 case WINED3DSIH_ADD
: op
= '+'; break;
1797 case WINED3DSIH_SUB
: op
= '-'; break;
1800 FIXME("Opcode %#x not yet handled in GLSL\n", ins
->handler_idx
);
1804 write_mask
= shader_glsl_append_dst(buffer
, ins
);
1805 shader_glsl_add_src_param(ins
, &ins
->src
[0], write_mask
, &src0_param
);
1806 shader_glsl_add_src_param(ins
, &ins
->src
[1], write_mask
, &src1_param
);
1807 shader_addline(buffer
, "%s %c %s);\n", src0_param
.param_str
, op
, src1_param
.param_str
);
1810 /* Process the WINED3DSIO_MOV opcode using GLSL (dst = src) */
1811 static void shader_glsl_mov(const struct wined3d_shader_instruction
*ins
)
1813 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
1814 glsl_src_param_t src0_param
;
1817 write_mask
= shader_glsl_append_dst(buffer
, ins
);
1818 shader_glsl_add_src_param(ins
, &ins
->src
[0], write_mask
, &src0_param
);
1820 /* In vs_1_1 WINED3DSIO_MOV can write to the address register. In later
1821 * shader versions WINED3DSIO_MOVA is used for this. */
1822 if (ins
->ctx
->reg_maps
->shader_version
.major
== 1
1823 && !shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
.type
)
1824 && ins
->dst
[0].reg
.type
== WINED3DSPR_ADDR
)
1826 /* This is a simple floor() */
1827 unsigned int mask_size
= shader_glsl_get_write_mask_size(write_mask
);
1828 if (mask_size
> 1) {
1829 shader_addline(buffer
, "ivec%d(floor(%s)));\n", mask_size
, src0_param
.param_str
);
1831 shader_addline(buffer
, "int(floor(%s)));\n", src0_param
.param_str
);
1834 else if(ins
->handler_idx
== WINED3DSIH_MOVA
)
1836 /* We need to *round* to the nearest int here. */
1837 unsigned int mask_size
= shader_glsl_get_write_mask_size(write_mask
);
1838 if (mask_size
> 1) {
1839 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
);
1841 shader_addline(buffer
, "int(floor(abs(%s) + 0.5) * sign(%s)));\n", src0_param
.param_str
, src0_param
.param_str
);
1844 shader_addline(buffer
, "%s);\n", src0_param
.param_str
);
1848 /* Process the dot product operators DP3 and DP4 in GLSL (dst = dot(src0, src1)) */
1849 static void shader_glsl_dot(const struct wined3d_shader_instruction
*ins
)
1851 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
1852 glsl_src_param_t src0_param
;
1853 glsl_src_param_t src1_param
;
1854 DWORD dst_write_mask
, src_write_mask
;
1855 unsigned int dst_size
= 0;
1857 dst_write_mask
= shader_glsl_append_dst(buffer
, ins
);
1858 dst_size
= shader_glsl_get_write_mask_size(dst_write_mask
);
1860 /* dp3 works on vec3, dp4 on vec4 */
1861 if (ins
->handler_idx
== WINED3DSIH_DP4
)
1863 src_write_mask
= WINED3DSP_WRITEMASK_ALL
;
1865 src_write_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
1868 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_write_mask
, &src0_param
);
1869 shader_glsl_add_src_param(ins
, &ins
->src
[1], src_write_mask
, &src1_param
);
1872 shader_addline(buffer
, "vec%d(dot(%s, %s)));\n", dst_size
, src0_param
.param_str
, src1_param
.param_str
);
1874 shader_addline(buffer
, "dot(%s, %s));\n", src0_param
.param_str
, src1_param
.param_str
);
1878 /* Note that this instruction has some restrictions. The destination write mask
1879 * can't contain the w component, and the source swizzles have to be .xyzw */
1880 static void shader_glsl_cross(const struct wined3d_shader_instruction
*ins
)
1882 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
1883 glsl_src_param_t src0_param
;
1884 glsl_src_param_t src1_param
;
1887 shader_glsl_get_write_mask(&ins
->dst
[0], dst_mask
);
1888 shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
1889 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
1890 shader_glsl_add_src_param(ins
, &ins
->src
[1], src_mask
, &src1_param
);
1891 shader_addline(ins
->ctx
->buffer
, "cross(%s, %s)%s);\n", src0_param
.param_str
, src1_param
.param_str
, dst_mask
);
1894 /* Process the WINED3DSIO_POW instruction in GLSL (dst = |src0|^src1)
1895 * Src0 and src1 are scalars. Note that D3D uses the absolute of src0, while
1896 * GLSL uses the value as-is. */
1897 static void shader_glsl_pow(const struct wined3d_shader_instruction
*ins
)
1899 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
1900 glsl_src_param_t src0_param
;
1901 glsl_src_param_t src1_param
;
1902 DWORD dst_write_mask
;
1903 unsigned int dst_size
;
1905 dst_write_mask
= shader_glsl_append_dst(buffer
, ins
);
1906 dst_size
= shader_glsl_get_write_mask_size(dst_write_mask
);
1908 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
1909 shader_glsl_add_src_param(ins
, &ins
->src
[1], WINED3DSP_WRITEMASK_0
, &src1_param
);
1912 shader_addline(buffer
, "vec%d(pow(abs(%s), %s)));\n", dst_size
, src0_param
.param_str
, src1_param
.param_str
);
1914 shader_addline(buffer
, "pow(abs(%s), %s));\n", src0_param
.param_str
, src1_param
.param_str
);
1918 /* Process the WINED3DSIO_LOG instruction in GLSL (dst = log2(|src0|))
1919 * Src0 is a scalar. Note that D3D uses the absolute of src0, while
1920 * GLSL uses the value as-is. */
1921 static void shader_glsl_log(const struct wined3d_shader_instruction
*ins
)
1923 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
1924 glsl_src_param_t src0_param
;
1925 DWORD dst_write_mask
;
1926 unsigned int dst_size
;
1928 dst_write_mask
= shader_glsl_append_dst(buffer
, ins
);
1929 dst_size
= shader_glsl_get_write_mask_size(dst_write_mask
);
1931 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
1934 shader_addline(buffer
, "vec%d(log2(abs(%s))));\n", dst_size
, src0_param
.param_str
);
1936 shader_addline(buffer
, "log2(abs(%s)));\n", src0_param
.param_str
);
1940 /* Map the opcode 1-to-1 to the GL code (arg->dst = instruction(src0, src1, ...) */
1941 static void shader_glsl_map2gl(const struct wined3d_shader_instruction
*ins
)
1943 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
1944 glsl_src_param_t src_param
;
1945 const char *instruction
;
1949 /* Determine the GLSL function to use based on the opcode */
1950 /* TODO: Possibly make this a table for faster lookups */
1951 switch (ins
->handler_idx
)
1953 case WINED3DSIH_MIN
: instruction
= "min"; break;
1954 case WINED3DSIH_MAX
: instruction
= "max"; break;
1955 case WINED3DSIH_ABS
: instruction
= "abs"; break;
1956 case WINED3DSIH_FRC
: instruction
= "fract"; break;
1957 case WINED3DSIH_NRM
: instruction
= "normalize"; break;
1958 case WINED3DSIH_EXP
: instruction
= "exp2"; break;
1959 case WINED3DSIH_SGN
: instruction
= "sign"; break;
1960 case WINED3DSIH_DSX
: instruction
= "dFdx"; break;
1961 case WINED3DSIH_DSY
: instruction
= "ycorrection.y * dFdy"; break;
1962 default: instruction
= "";
1963 FIXME("Opcode %#x not yet handled in GLSL\n", ins
->handler_idx
);
1967 write_mask
= shader_glsl_append_dst(buffer
, ins
);
1969 shader_addline(buffer
, "%s(", instruction
);
1973 shader_glsl_add_src_param(ins
, &ins
->src
[0], write_mask
, &src_param
);
1974 shader_addline(buffer
, "%s", src_param
.param_str
);
1975 for (i
= 1; i
< ins
->src_count
; ++i
)
1977 shader_glsl_add_src_param(ins
, &ins
->src
[i
], write_mask
, &src_param
);
1978 shader_addline(buffer
, ", %s", src_param
.param_str
);
1982 shader_addline(buffer
, "));\n");
1985 /** Process the WINED3DSIO_EXPP instruction in GLSL:
1986 * For shader model 1.x, do the following (and honor the writemask, so use a temporary variable):
1987 * dst.x = 2^(floor(src))
1988 * dst.y = src - floor(src)
1989 * dst.z = 2^src (partial precision is allowed, but optional)
1991 * For 2.0 shaders, just do this (honoring writemask and swizzle):
1992 * dst = 2^src; (partial precision is allowed, but optional)
1994 static void shader_glsl_expp(const struct wined3d_shader_instruction
*ins
)
1996 glsl_src_param_t src_param
;
1998 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
, &src_param
);
2000 if (ins
->ctx
->reg_maps
->shader_version
.major
< 2)
2004 shader_addline(ins
->ctx
->buffer
, "tmp0.x = exp2(floor(%s));\n", src_param
.param_str
);
2005 shader_addline(ins
->ctx
->buffer
, "tmp0.y = %s - floor(%s);\n", src_param
.param_str
, src_param
.param_str
);
2006 shader_addline(ins
->ctx
->buffer
, "tmp0.z = exp2(%s);\n", src_param
.param_str
);
2007 shader_addline(ins
->ctx
->buffer
, "tmp0.w = 1.0;\n");
2009 shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2010 shader_glsl_get_write_mask(&ins
->dst
[0], dst_mask
);
2011 shader_addline(ins
->ctx
->buffer
, "tmp0%s);\n", dst_mask
);
2014 unsigned int mask_size
;
2016 write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2017 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
2019 if (mask_size
> 1) {
2020 shader_addline(ins
->ctx
->buffer
, "vec%d(exp2(%s)));\n", mask_size
, src_param
.param_str
);
2022 shader_addline(ins
->ctx
->buffer
, "exp2(%s));\n", src_param
.param_str
);
2027 /** Process the RCP (reciprocal or inverse) opcode in GLSL (dst = 1 / src) */
2028 static void shader_glsl_rcp(const struct wined3d_shader_instruction
*ins
)
2030 glsl_src_param_t src_param
;
2032 unsigned int mask_size
;
2034 write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2035 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
2036 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_3
, &src_param
);
2038 if (mask_size
> 1) {
2039 shader_addline(ins
->ctx
->buffer
, "vec%d(1.0 / %s));\n", mask_size
, src_param
.param_str
);
2041 shader_addline(ins
->ctx
->buffer
, "1.0 / %s);\n", src_param
.param_str
);
2045 static void shader_glsl_rsq(const struct wined3d_shader_instruction
*ins
)
2047 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2048 glsl_src_param_t src_param
;
2050 unsigned int mask_size
;
2052 write_mask
= shader_glsl_append_dst(buffer
, ins
);
2053 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
2055 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_3
, &src_param
);
2057 if (mask_size
> 1) {
2058 shader_addline(buffer
, "vec%d(inversesqrt(%s)));\n", mask_size
, src_param
.param_str
);
2060 shader_addline(buffer
, "inversesqrt(%s));\n", src_param
.param_str
);
2064 /** Process signed comparison opcodes in GLSL. */
2065 static void shader_glsl_compare(const struct wined3d_shader_instruction
*ins
)
2067 glsl_src_param_t src0_param
;
2068 glsl_src_param_t src1_param
;
2070 unsigned int mask_size
;
2072 write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2073 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
2074 shader_glsl_add_src_param(ins
, &ins
->src
[0], write_mask
, &src0_param
);
2075 shader_glsl_add_src_param(ins
, &ins
->src
[1], write_mask
, &src1_param
);
2077 if (mask_size
> 1) {
2078 const char *compare
;
2080 switch(ins
->handler_idx
)
2082 case WINED3DSIH_SLT
: compare
= "lessThan"; break;
2083 case WINED3DSIH_SGE
: compare
= "greaterThanEqual"; break;
2084 default: compare
= "";
2085 FIXME("Can't handle opcode %#x\n", ins
->handler_idx
);
2088 shader_addline(ins
->ctx
->buffer
, "vec%d(%s(%s, %s)));\n", mask_size
, compare
,
2089 src0_param
.param_str
, src1_param
.param_str
);
2091 switch(ins
->handler_idx
)
2093 case WINED3DSIH_SLT
:
2094 /* Step(src0, src1) is not suitable here because if src0 == src1 SLT is supposed,
2095 * to return 0.0 but step returns 1.0 because step is not < x
2096 * An alternative is a bvec compare padded with an unused second component.
2097 * step(src1 * -1.0, src0 * -1.0) is not an option because it suffers from the same
2098 * issue. Playing with not() is not possible either because not() does not accept
2101 shader_addline(ins
->ctx
->buffer
, "(%s < %s) ? 1.0 : 0.0);\n",
2102 src0_param
.param_str
, src1_param
.param_str
);
2104 case WINED3DSIH_SGE
:
2105 /* Here we can use the step() function and safe a conditional */
2106 shader_addline(ins
->ctx
->buffer
, "step(%s, %s));\n", src1_param
.param_str
, src0_param
.param_str
);
2109 FIXME("Can't handle opcode %#x\n", ins
->handler_idx
);
2115 /** Process CMP instruction in GLSL (dst = src0 >= 0.0 ? src1 : src2), per channel */
2116 static void shader_glsl_cmp(const struct wined3d_shader_instruction
*ins
)
2118 glsl_src_param_t src0_param
;
2119 glsl_src_param_t src1_param
;
2120 glsl_src_param_t src2_param
;
2121 DWORD write_mask
, cmp_channel
= 0;
2124 BOOL temp_destination
= FALSE
;
2126 if (shader_is_scalar(&ins
->src
[0].reg
))
2128 write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2130 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_ALL
, &src0_param
);
2131 shader_glsl_add_src_param(ins
, &ins
->src
[1], write_mask
, &src1_param
);
2132 shader_glsl_add_src_param(ins
, &ins
->src
[2], write_mask
, &src2_param
);
2134 shader_addline(ins
->ctx
->buffer
, "%s >= 0.0 ? %s : %s);\n",
2135 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
2137 DWORD dst_mask
= ins
->dst
[0].write_mask
;
2138 struct wined3d_shader_dst_param dst
= ins
->dst
[0];
2140 /* Cycle through all source0 channels */
2141 for (i
=0; i
<4; i
++) {
2143 /* Find the destination channels which use the current source0 channel */
2144 for (j
=0; j
<4; j
++) {
2145 if (((ins
->src
[0].swizzle
>> (2 * j
)) & 0x3) == i
)
2147 write_mask
|= WINED3DSP_WRITEMASK_0
<< j
;
2148 cmp_channel
= WINED3DSP_WRITEMASK_0
<< j
;
2151 dst
.write_mask
= dst_mask
& write_mask
;
2153 /* Splitting the cmp instruction up in multiple lines imposes a problem:
2154 * The first lines may overwrite source parameters of the following lines.
2155 * Deal with that by using a temporary destination register if needed
2157 if ((ins
->src
[0].reg
.idx
== ins
->dst
[0].reg
.idx
2158 && ins
->src
[0].reg
.type
== ins
->dst
[0].reg
.type
)
2159 || (ins
->src
[1].reg
.idx
== ins
->dst
[0].reg
.idx
2160 && ins
->src
[1].reg
.type
== ins
->dst
[0].reg
.type
)
2161 || (ins
->src
[2].reg
.idx
== ins
->dst
[0].reg
.idx
2162 && ins
->src
[2].reg
.type
== ins
->dst
[0].reg
.type
))
2164 write_mask
= shader_glsl_get_write_mask(&dst
, mask_char
);
2165 if (!write_mask
) continue;
2166 shader_addline(ins
->ctx
->buffer
, "tmp0%s = (", mask_char
);
2167 temp_destination
= TRUE
;
2169 write_mask
= shader_glsl_append_dst_ext(ins
->ctx
->buffer
, ins
, &dst
);
2170 if (!write_mask
) continue;
2173 shader_glsl_add_src_param(ins
, &ins
->src
[0], cmp_channel
, &src0_param
);
2174 shader_glsl_add_src_param(ins
, &ins
->src
[1], write_mask
, &src1_param
);
2175 shader_glsl_add_src_param(ins
, &ins
->src
[2], write_mask
, &src2_param
);
2177 shader_addline(ins
->ctx
->buffer
, "%s >= 0.0 ? %s : %s);\n",
2178 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
2181 if(temp_destination
) {
2182 shader_glsl_get_write_mask(&ins
->dst
[0], mask_char
);
2183 shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2184 shader_addline(ins
->ctx
->buffer
, "tmp0%s);\n", mask_char
);
2190 /** Process the CND opcode in GLSL (dst = (src0 > 0.5) ? src1 : src2) */
2191 /* For ps 1.1-1.3, only a single component of src0 is used. For ps 1.4
2192 * the compare is done per component of src0. */
2193 static void shader_glsl_cnd(const struct wined3d_shader_instruction
*ins
)
2195 struct wined3d_shader_dst_param dst
;
2196 glsl_src_param_t src0_param
;
2197 glsl_src_param_t src1_param
;
2198 glsl_src_param_t src2_param
;
2199 DWORD write_mask
, cmp_channel
= 0;
2202 DWORD shader_version
= WINED3D_SHADER_VERSION(ins
->ctx
->reg_maps
->shader_version
.major
,
2203 ins
->ctx
->reg_maps
->shader_version
.minor
);
2205 if (shader_version
< WINED3D_SHADER_VERSION(1, 4))
2207 write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2208 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2209 shader_glsl_add_src_param(ins
, &ins
->src
[1], write_mask
, &src1_param
);
2210 shader_glsl_add_src_param(ins
, &ins
->src
[2], write_mask
, &src2_param
);
2212 /* Fun: The D3DSI_COISSUE flag changes the semantic of the cnd instruction for < 1.4 shaders */
2215 shader_addline(ins
->ctx
->buffer
, "%s /* COISSUE! */);\n", src1_param
.param_str
);
2217 shader_addline(ins
->ctx
->buffer
, "%s > 0.5 ? %s : %s);\n",
2218 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
2222 /* Cycle through all source0 channels */
2223 dst_mask
= ins
->dst
[0].write_mask
;
2225 for (i
=0; i
<4; i
++) {
2227 /* Find the destination channels which use the current source0 channel */
2228 for (j
=0; j
<4; j
++) {
2229 if (((ins
->src
[0].swizzle
>> (2 * j
)) & 0x3) == i
)
2231 write_mask
|= WINED3DSP_WRITEMASK_0
<< j
;
2232 cmp_channel
= WINED3DSP_WRITEMASK_0
<< j
;
2236 dst
.write_mask
= dst_mask
& write_mask
;
2237 write_mask
= shader_glsl_append_dst_ext(ins
->ctx
->buffer
, ins
, &dst
);
2238 if (!write_mask
) continue;
2240 shader_glsl_add_src_param(ins
, &ins
->src
[0], cmp_channel
, &src0_param
);
2241 shader_glsl_add_src_param(ins
, &ins
->src
[1], write_mask
, &src1_param
);
2242 shader_glsl_add_src_param(ins
, &ins
->src
[2], write_mask
, &src2_param
);
2244 shader_addline(ins
->ctx
->buffer
, "%s > 0.5 ? %s : %s);\n",
2245 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
2249 /** GLSL code generation for WINED3DSIO_MAD: Multiply the first 2 opcodes, then add the last */
2250 static void shader_glsl_mad(const struct wined3d_shader_instruction
*ins
)
2252 glsl_src_param_t src0_param
;
2253 glsl_src_param_t src1_param
;
2254 glsl_src_param_t src2_param
;
2257 write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2258 shader_glsl_add_src_param(ins
, &ins
->src
[0], write_mask
, &src0_param
);
2259 shader_glsl_add_src_param(ins
, &ins
->src
[1], write_mask
, &src1_param
);
2260 shader_glsl_add_src_param(ins
, &ins
->src
[2], write_mask
, &src2_param
);
2261 shader_addline(ins
->ctx
->buffer
, "(%s * %s) + %s);\n",
2262 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
2265 /* Handles transforming all WINED3DSIO_M?x? opcodes for
2266 Vertex shaders to GLSL codes */
2267 static void shader_glsl_mnxn(const struct wined3d_shader_instruction
*ins
)
2270 int nComponents
= 0;
2271 struct wined3d_shader_dst_param tmp_dst
= {{0}};
2272 struct wined3d_shader_src_param tmp_src
[2] = {{{0}}};
2273 struct wined3d_shader_instruction tmp_ins
;
2275 memset(&tmp_ins
, 0, sizeof(tmp_ins
));
2277 /* Set constants for the temporary argument */
2278 tmp_ins
.ctx
= ins
->ctx
;
2279 tmp_ins
.dst_count
= 1;
2280 tmp_ins
.dst
= &tmp_dst
;
2281 tmp_ins
.src_count
= 2;
2282 tmp_ins
.src
= tmp_src
;
2284 switch(ins
->handler_idx
)
2286 case WINED3DSIH_M4x4
:
2288 tmp_ins
.handler_idx
= WINED3DSIH_DP4
;
2290 case WINED3DSIH_M4x3
:
2292 tmp_ins
.handler_idx
= WINED3DSIH_DP4
;
2294 case WINED3DSIH_M3x4
:
2296 tmp_ins
.handler_idx
= WINED3DSIH_DP3
;
2298 case WINED3DSIH_M3x3
:
2300 tmp_ins
.handler_idx
= WINED3DSIH_DP3
;
2302 case WINED3DSIH_M3x2
:
2304 tmp_ins
.handler_idx
= WINED3DSIH_DP3
;
2310 tmp_dst
= ins
->dst
[0];
2311 tmp_src
[0] = ins
->src
[0];
2312 tmp_src
[1] = ins
->src
[1];
2313 for (i
= 0; i
< nComponents
; ++i
)
2315 tmp_dst
.write_mask
= WINED3DSP_WRITEMASK_0
<< i
;
2316 shader_glsl_dot(&tmp_ins
);
2317 ++tmp_src
[1].reg
.idx
;
2322 The LRP instruction performs a component-wise linear interpolation
2323 between the second and third operands using the first operand as the
2324 blend factor. Equation: (dst = src2 + src0 * (src1 - src2))
2325 This is equivalent to mix(src2, src1, src0);
2327 static void shader_glsl_lrp(const struct wined3d_shader_instruction
*ins
)
2329 glsl_src_param_t src0_param
;
2330 glsl_src_param_t src1_param
;
2331 glsl_src_param_t src2_param
;
2334 write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2336 shader_glsl_add_src_param(ins
, &ins
->src
[0], write_mask
, &src0_param
);
2337 shader_glsl_add_src_param(ins
, &ins
->src
[1], write_mask
, &src1_param
);
2338 shader_glsl_add_src_param(ins
, &ins
->src
[2], write_mask
, &src2_param
);
2340 shader_addline(ins
->ctx
->buffer
, "mix(%s, %s, %s));\n",
2341 src2_param
.param_str
, src1_param
.param_str
, src0_param
.param_str
);
2344 /** Process the WINED3DSIO_LIT instruction in GLSL:
2345 * dst.x = dst.w = 1.0
2346 * dst.y = (src0.x > 0) ? src0.x
2347 * dst.z = (src0.x > 0) ? ((src0.y > 0) ? pow(src0.y, src.w) : 0) : 0
2348 * where src.w is clamped at +- 128
2350 static void shader_glsl_lit(const struct wined3d_shader_instruction
*ins
)
2352 glsl_src_param_t src0_param
;
2353 glsl_src_param_t src1_param
;
2354 glsl_src_param_t src3_param
;
2357 shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2358 shader_glsl_get_write_mask(&ins
->dst
[0], dst_mask
);
2360 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2361 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_1
, &src1_param
);
2362 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_3
, &src3_param
);
2364 /* The sdk specifies the instruction like this
2366 * if(src.x > 0.0) dst.y = src.x
2368 * if(src.x > 0.0 && src.y > 0.0) dst.z = pow(src.y, power);
2372 * Obviously that has quite a few conditionals in it which we don't like. So the first step is this:
2373 * dst.x = 1.0 ... No further explanation needed
2374 * dst.y = max(src.y, 0.0); ... If x < 0.0, use 0.0, otherwise x. Same as the conditional
2375 * dst.z = x > 0.0 ? pow(max(y, 0.0), p) : 0; ... 0 ^ power is 0, and otherwise we use y anyway
2376 * dst.w = 1.0. ... Nothing fancy.
2378 * So we still have one conditional in there. So do this:
2379 * dst.z = pow(max(0.0, src.y) * step(0.0, src.x), power);
2381 * 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),
2382 * 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.
2383 * if both x and y are > 0, we get pow(y * 1.0, power), as it is supposed to
2385 shader_addline(ins
->ctx
->buffer
,
2386 "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",
2387 src0_param
.param_str
, src1_param
.param_str
, src0_param
.param_str
, src3_param
.param_str
, dst_mask
);
2390 /** Process the WINED3DSIO_DST instruction in GLSL:
2392 * dst.y = src0.x * src0.y
2396 static void shader_glsl_dst(const struct wined3d_shader_instruction
*ins
)
2398 glsl_src_param_t src0y_param
;
2399 glsl_src_param_t src0z_param
;
2400 glsl_src_param_t src1y_param
;
2401 glsl_src_param_t src1w_param
;
2404 shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2405 shader_glsl_get_write_mask(&ins
->dst
[0], dst_mask
);
2407 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_1
, &src0y_param
);
2408 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_2
, &src0z_param
);
2409 shader_glsl_add_src_param(ins
, &ins
->src
[1], WINED3DSP_WRITEMASK_1
, &src1y_param
);
2410 shader_glsl_add_src_param(ins
, &ins
->src
[1], WINED3DSP_WRITEMASK_3
, &src1w_param
);
2412 shader_addline(ins
->ctx
->buffer
, "vec4(1.0, %s * %s, %s, %s))%s;\n",
2413 src0y_param
.param_str
, src1y_param
.param_str
, src0z_param
.param_str
, src1w_param
.param_str
, dst_mask
);
2416 /** Process the WINED3DSIO_SINCOS instruction in GLSL:
2417 * VS 2.0 requires that specific cosine and sine constants be passed to this instruction so the hardware
2418 * can handle it. But, these functions are built-in for GLSL, so we can just ignore the last 2 params.
2420 * dst.x = cos(src0.?)
2421 * dst.y = sin(src0.?)
2425 static void shader_glsl_sincos(const struct wined3d_shader_instruction
*ins
)
2427 glsl_src_param_t src0_param
;
2430 write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2431 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2433 switch (write_mask
) {
2434 case WINED3DSP_WRITEMASK_0
:
2435 shader_addline(ins
->ctx
->buffer
, "cos(%s));\n", src0_param
.param_str
);
2438 case WINED3DSP_WRITEMASK_1
:
2439 shader_addline(ins
->ctx
->buffer
, "sin(%s));\n", src0_param
.param_str
);
2442 case (WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
):
2443 shader_addline(ins
->ctx
->buffer
, "vec2(cos(%s), sin(%s)));\n", src0_param
.param_str
, src0_param
.param_str
);
2447 ERR("Write mask should be .x, .y or .xy\n");
2452 /** Process the WINED3DSIO_LOOP instruction in GLSL:
2453 * Start a for() loop where src1.y is the initial value of aL,
2454 * increment aL by src1.z for a total of src1.x iterations.
2455 * Need to use a temporary variable for this operation.
2457 /* FIXME: I don't think nested loops will work correctly this way. */
2458 static void shader_glsl_loop(const struct wined3d_shader_instruction
*ins
)
2460 glsl_src_param_t src1_param
;
2461 IWineD3DBaseShaderImpl
*shader
= (IWineD3DBaseShaderImpl
*)ins
->ctx
->shader
;
2462 const DWORD
*control_values
= NULL
;
2463 const local_constant
*constant
;
2465 shader_glsl_add_src_param(ins
, &ins
->src
[1], WINED3DSP_WRITEMASK_ALL
, &src1_param
);
2467 /* Try to hardcode the loop control parameters if possible. Direct3D 9 class hardware doesn't support real
2468 * varying indexing, but Microsoft designed this feature for Shader model 2.x+. If the loop control is
2469 * known at compile time, the GLSL compiler can unroll the loop, and replace indirect addressing with direct
2472 if (ins
->src
[1].reg
.type
== WINED3DSPR_CONSTINT
)
2474 LIST_FOR_EACH_ENTRY(constant
, &shader
->baseShader
.constantsI
, local_constant
, entry
) {
2475 if (constant
->idx
== ins
->src
[1].reg
.idx
)
2477 control_values
= constant
->value
;
2485 struct wined3d_shader_loop_control loop_control
;
2486 loop_control
.count
= control_values
[0];
2487 loop_control
.start
= control_values
[1];
2488 loop_control
.step
= (int)control_values
[2];
2490 if (loop_control
.step
> 0)
2492 shader_addline(ins
->ctx
->buffer
, "for (aL%u = %u; aL%u < (%u * %d + %u); aL%u += %d) {\n",
2493 shader
->baseShader
.cur_loop_depth
, loop_control
.start
,
2494 shader
->baseShader
.cur_loop_depth
, loop_control
.count
, loop_control
.step
, loop_control
.start
,
2495 shader
->baseShader
.cur_loop_depth
, loop_control
.step
);
2497 else if (loop_control
.step
< 0)
2499 shader_addline(ins
->ctx
->buffer
, "for (aL%u = %u; aL%u > (%u * %d + %u); aL%u += %d) {\n",
2500 shader
->baseShader
.cur_loop_depth
, loop_control
.start
,
2501 shader
->baseShader
.cur_loop_depth
, loop_control
.count
, loop_control
.step
, loop_control
.start
,
2502 shader
->baseShader
.cur_loop_depth
, loop_control
.step
);
2506 shader_addline(ins
->ctx
->buffer
, "for (aL%u = %u, tmpInt%u = 0; tmpInt%u < %u; tmpInt%u++) {\n",
2507 shader
->baseShader
.cur_loop_depth
, loop_control
.start
, shader
->baseShader
.cur_loop_depth
,
2508 shader
->baseShader
.cur_loop_depth
, loop_control
.count
,
2509 shader
->baseShader
.cur_loop_depth
);
2512 shader_addline(ins
->ctx
->buffer
,
2513 "for (tmpInt%u = 0, aL%u = %s.y; tmpInt%u < %s.x; tmpInt%u++, aL%u += %s.z) {\n",
2514 shader
->baseShader
.cur_loop_depth
, shader
->baseShader
.cur_loop_regno
,
2515 src1_param
.reg_name
, shader
->baseShader
.cur_loop_depth
, src1_param
.reg_name
,
2516 shader
->baseShader
.cur_loop_depth
, shader
->baseShader
.cur_loop_regno
, src1_param
.reg_name
);
2519 shader
->baseShader
.cur_loop_depth
++;
2520 shader
->baseShader
.cur_loop_regno
++;
2523 static void shader_glsl_end(const struct wined3d_shader_instruction
*ins
)
2525 IWineD3DBaseShaderImpl
*shader
= (IWineD3DBaseShaderImpl
*)ins
->ctx
->shader
;
2527 shader_addline(ins
->ctx
->buffer
, "}\n");
2529 if (ins
->handler_idx
== WINED3DSIH_ENDLOOP
)
2531 shader
->baseShader
.cur_loop_depth
--;
2532 shader
->baseShader
.cur_loop_regno
--;
2535 if (ins
->handler_idx
== WINED3DSIH_ENDREP
)
2537 shader
->baseShader
.cur_loop_depth
--;
2541 static void shader_glsl_rep(const struct wined3d_shader_instruction
*ins
)
2543 IWineD3DBaseShaderImpl
*shader
= (IWineD3DBaseShaderImpl
*)ins
->ctx
->shader
;
2544 glsl_src_param_t src0_param
;
2545 const DWORD
*control_values
= NULL
;
2546 const local_constant
*constant
;
2548 /* Try to hardcode local values to help the GLSL compiler to unroll and optimize the loop */
2549 if (ins
->src
[0].reg
.type
== WINED3DSPR_CONSTINT
)
2551 LIST_FOR_EACH_ENTRY(constant
, &shader
->baseShader
.constantsI
, local_constant
, entry
)
2553 if (constant
->idx
== ins
->src
[0].reg
.idx
)
2555 control_values
= constant
->value
;
2561 if(control_values
) {
2562 shader_addline(ins
->ctx
->buffer
, "for (tmpInt%d = 0; tmpInt%d < %d; tmpInt%d++) {\n",
2563 shader
->baseShader
.cur_loop_depth
, shader
->baseShader
.cur_loop_depth
,
2564 control_values
[0], shader
->baseShader
.cur_loop_depth
);
2566 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2567 shader_addline(ins
->ctx
->buffer
, "for (tmpInt%d = 0; tmpInt%d < %s; tmpInt%d++) {\n",
2568 shader
->baseShader
.cur_loop_depth
, shader
->baseShader
.cur_loop_depth
,
2569 src0_param
.param_str
, shader
->baseShader
.cur_loop_depth
);
2571 shader
->baseShader
.cur_loop_depth
++;
2574 static void shader_glsl_if(const struct wined3d_shader_instruction
*ins
)
2576 glsl_src_param_t src0_param
;
2578 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2579 shader_addline(ins
->ctx
->buffer
, "if (%s) {\n", src0_param
.param_str
);
2582 static void shader_glsl_ifc(const struct wined3d_shader_instruction
*ins
)
2584 glsl_src_param_t src0_param
;
2585 glsl_src_param_t src1_param
;
2587 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2588 shader_glsl_add_src_param(ins
, &ins
->src
[1], WINED3DSP_WRITEMASK_0
, &src1_param
);
2590 shader_addline(ins
->ctx
->buffer
, "if (%s %s %s) {\n",
2591 src0_param
.param_str
, shader_get_comp_op(ins
->flags
), src1_param
.param_str
);
2594 static void shader_glsl_else(const struct wined3d_shader_instruction
*ins
)
2596 shader_addline(ins
->ctx
->buffer
, "} else {\n");
2599 static void shader_glsl_break(const struct wined3d_shader_instruction
*ins
)
2601 shader_addline(ins
->ctx
->buffer
, "break;\n");
2604 /* FIXME: According to MSDN the compare is done per component. */
2605 static void shader_glsl_breakc(const struct wined3d_shader_instruction
*ins
)
2607 glsl_src_param_t src0_param
;
2608 glsl_src_param_t src1_param
;
2610 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2611 shader_glsl_add_src_param(ins
, &ins
->src
[1], WINED3DSP_WRITEMASK_0
, &src1_param
);
2613 shader_addline(ins
->ctx
->buffer
, "if (%s %s %s) break;\n",
2614 src0_param
.param_str
, shader_get_comp_op(ins
->flags
), src1_param
.param_str
);
2617 static void shader_glsl_label(const struct wined3d_shader_instruction
*ins
)
2619 shader_addline(ins
->ctx
->buffer
, "}\n");
2620 shader_addline(ins
->ctx
->buffer
, "void subroutine%u () {\n", ins
->src
[0].reg
.idx
);
2623 static void shader_glsl_call(const struct wined3d_shader_instruction
*ins
)
2625 shader_addline(ins
->ctx
->buffer
, "subroutine%u();\n", ins
->src
[0].reg
.idx
);
2628 static void shader_glsl_callnz(const struct wined3d_shader_instruction
*ins
)
2630 glsl_src_param_t src1_param
;
2632 shader_glsl_add_src_param(ins
, &ins
->src
[1], WINED3DSP_WRITEMASK_0
, &src1_param
);
2633 shader_addline(ins
->ctx
->buffer
, "if (%s) subroutine%u();\n", src1_param
.param_str
, ins
->src
[0].reg
.idx
);
2636 static void shader_glsl_ret(const struct wined3d_shader_instruction
*ins
)
2638 /* No-op. The closing } is written when a new function is started, and at the end of the shader. This
2639 * function only suppresses the unhandled instruction warning
2643 /*********************************************
2644 * Pixel Shader Specific Code begins here
2645 ********************************************/
2646 static void shader_glsl_tex(const struct wined3d_shader_instruction
*ins
)
2648 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*)ins
->ctx
->shader
;
2649 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
2650 DWORD shader_version
= WINED3D_SHADER_VERSION(ins
->ctx
->reg_maps
->shader_version
.major
,
2651 ins
->ctx
->reg_maps
->shader_version
.minor
);
2652 glsl_sample_function_t sample_function
;
2653 DWORD sample_flags
= 0;
2654 WINED3DSAMPLER_TEXTURE_TYPE sampler_type
;
2656 DWORD mask
= 0, swizzle
;
2658 /* 1.0-1.4: Use destination register as sampler source.
2659 * 2.0+: Use provided sampler source. */
2660 if (shader_version
< WINED3D_SHADER_VERSION(2,0)) sampler_idx
= ins
->dst
[0].reg
.idx
;
2661 else sampler_idx
= ins
->src
[1].reg
.idx
;
2662 sampler_type
= ins
->ctx
->reg_maps
->sampler_type
[sampler_idx
];
2664 if (shader_version
< WINED3D_SHADER_VERSION(1,4))
2666 DWORD flags
= deviceImpl
->stateBlock
->textureState
[sampler_idx
][WINED3DTSS_TEXTURETRANSFORMFLAGS
];
2668 /* Projected cube textures don't make a lot of sense, the resulting coordinates stay the same. */
2669 if (flags
& WINED3DTTFF_PROJECTED
&& sampler_type
!= WINED3DSTT_CUBE
) {
2670 sample_flags
|= WINED3D_GLSL_SAMPLE_PROJECTED
;
2671 switch (flags
& ~WINED3DTTFF_PROJECTED
) {
2672 case WINED3DTTFF_COUNT1
: FIXME("WINED3DTTFF_PROJECTED with WINED3DTTFF_COUNT1?\n"); break;
2673 case WINED3DTTFF_COUNT2
: mask
= WINED3DSP_WRITEMASK_1
; break;
2674 case WINED3DTTFF_COUNT3
: mask
= WINED3DSP_WRITEMASK_2
; break;
2675 case WINED3DTTFF_COUNT4
:
2676 case WINED3DTTFF_DISABLE
: mask
= WINED3DSP_WRITEMASK_3
; break;
2680 else if (shader_version
< WINED3D_SHADER_VERSION(2,0))
2682 DWORD src_mod
= ins
->src
[0].modifiers
;
2684 if (src_mod
== WINED3DSPSM_DZ
) {
2685 sample_flags
|= WINED3D_GLSL_SAMPLE_PROJECTED
;
2686 mask
= WINED3DSP_WRITEMASK_2
;
2687 } else if (src_mod
== WINED3DSPSM_DW
) {
2688 sample_flags
|= WINED3D_GLSL_SAMPLE_PROJECTED
;
2689 mask
= WINED3DSP_WRITEMASK_3
;
2692 if (ins
->flags
& WINED3DSI_TEXLD_PROJECT
)
2694 /* ps 2.0 texldp instruction always divides by the fourth component. */
2695 sample_flags
|= WINED3D_GLSL_SAMPLE_PROJECTED
;
2696 mask
= WINED3DSP_WRITEMASK_3
;
2700 if(deviceImpl
->stateBlock
->textures
[sampler_idx
] &&
2701 IWineD3DBaseTexture_GetTextureDimensions(deviceImpl
->stateBlock
->textures
[sampler_idx
]) == GL_TEXTURE_RECTANGLE_ARB
) {
2702 sample_flags
|= WINED3D_GLSL_SAMPLE_RECT
;
2705 shader_glsl_get_sample_function(sampler_type
, sample_flags
, &sample_function
);
2706 mask
|= sample_function
.coord_mask
;
2708 if (shader_version
< WINED3D_SHADER_VERSION(2,0)) swizzle
= WINED3DSP_NOSWIZZLE
;
2709 else swizzle
= ins
->src
[1].swizzle
;
2711 /* 1.0-1.3: Use destination register as coordinate source.
2712 1.4+: Use provided coordinate source register. */
2713 if (shader_version
< WINED3D_SHADER_VERSION(1,4))
2716 shader_glsl_write_mask_to_str(mask
, coord_mask
);
2717 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, swizzle
, NULL
, NULL
, NULL
,
2718 "T%u%s", sampler_idx
, coord_mask
);
2720 glsl_src_param_t coord_param
;
2721 shader_glsl_add_src_param(ins
, &ins
->src
[0], mask
, &coord_param
);
2722 if (ins
->flags
& WINED3DSI_TEXLD_BIAS
)
2724 glsl_src_param_t bias
;
2725 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_3
, &bias
);
2726 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, swizzle
, NULL
, NULL
, bias
.param_str
,
2727 "%s", coord_param
.param_str
);
2729 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, swizzle
, NULL
, NULL
, NULL
,
2730 "%s", coord_param
.param_str
);
2735 static void shader_glsl_texldd(const struct wined3d_shader_instruction
*ins
)
2737 IWineD3DBaseShaderImpl
*This
= (IWineD3DBaseShaderImpl
*)ins
->ctx
->shader
;
2738 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
2739 const struct wined3d_gl_info
*gl_info
= &deviceImpl
->adapter
->gl_info
;
2740 glsl_sample_function_t sample_function
;
2741 glsl_src_param_t coord_param
, dx_param
, dy_param
;
2742 DWORD sample_flags
= WINED3D_GLSL_SAMPLE_GRAD
;
2745 DWORD swizzle
= ins
->src
[1].swizzle
;
2747 if(!GL_SUPPORT(ARB_SHADER_TEXTURE_LOD
)) {
2748 FIXME("texldd used, but not supported by hardware. Falling back to regular tex\n");
2749 return shader_glsl_tex(ins
);
2752 sampler_idx
= ins
->src
[1].reg
.idx
;
2753 sampler_type
= ins
->ctx
->reg_maps
->sampler_type
[sampler_idx
];
2754 if(deviceImpl
->stateBlock
->textures
[sampler_idx
] &&
2755 IWineD3DBaseTexture_GetTextureDimensions(deviceImpl
->stateBlock
->textures
[sampler_idx
]) == GL_TEXTURE_RECTANGLE_ARB
) {
2756 sample_flags
|= WINED3D_GLSL_SAMPLE_RECT
;
2759 shader_glsl_get_sample_function(sampler_type
, sample_flags
, &sample_function
);
2760 shader_glsl_add_src_param(ins
, &ins
->src
[0], sample_function
.coord_mask
, &coord_param
);
2761 shader_glsl_add_src_param(ins
, &ins
->src
[2], sample_function
.coord_mask
, &dx_param
);
2762 shader_glsl_add_src_param(ins
, &ins
->src
[3], sample_function
.coord_mask
, &dy_param
);
2764 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, swizzle
, dx_param
.param_str
, dy_param
.param_str
, NULL
,
2765 "%s", coord_param
.param_str
);
2768 static void shader_glsl_texldl(const struct wined3d_shader_instruction
*ins
)
2770 IWineD3DBaseShaderImpl
*This
= (IWineD3DBaseShaderImpl
*)ins
->ctx
->shader
;
2771 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
2772 glsl_sample_function_t sample_function
;
2773 glsl_src_param_t coord_param
, lod_param
;
2774 DWORD sample_flags
= WINED3D_GLSL_SAMPLE_LOD
;
2777 DWORD swizzle
= ins
->src
[1].swizzle
;
2779 sampler_idx
= ins
->src
[1].reg
.idx
;
2780 sampler_type
= ins
->ctx
->reg_maps
->sampler_type
[sampler_idx
];
2781 if(deviceImpl
->stateBlock
->textures
[sampler_idx
] &&
2782 IWineD3DBaseTexture_GetTextureDimensions(deviceImpl
->stateBlock
->textures
[sampler_idx
]) == GL_TEXTURE_RECTANGLE_ARB
) {
2783 sample_flags
|= WINED3D_GLSL_SAMPLE_RECT
;
2785 shader_glsl_get_sample_function(sampler_type
, sample_flags
, &sample_function
);
2786 shader_glsl_add_src_param(ins
, &ins
->src
[0], sample_function
.coord_mask
, &coord_param
);
2788 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_3
, &lod_param
);
2790 if (shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
.type
))
2792 /* The GLSL spec claims the Lod sampling functions are only supported in vertex shaders.
2793 * However, they seem to work just fine in fragment shaders as well. */
2794 WARN("Using %s in fragment shader.\n", sample_function
.name
);
2796 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, swizzle
, NULL
, NULL
, lod_param
.param_str
,
2797 "%s", coord_param
.param_str
);
2800 static void shader_glsl_texcoord(const struct wined3d_shader_instruction
*ins
)
2802 /* FIXME: Make this work for more than just 2D textures */
2803 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2804 DWORD write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2806 if (!(ins
->ctx
->reg_maps
->shader_version
.major
== 1 && ins
->ctx
->reg_maps
->shader_version
.minor
== 4))
2810 shader_glsl_get_write_mask(&ins
->dst
[0], dst_mask
);
2811 shader_addline(buffer
, "clamp(gl_TexCoord[%u], 0.0, 1.0)%s);\n",
2812 ins
->dst
[0].reg
.idx
, dst_mask
);
2814 DWORD reg
= ins
->src
[0].reg
.idx
;
2815 DWORD src_mod
= ins
->src
[0].modifiers
;
2816 char dst_swizzle
[6];
2818 shader_glsl_get_swizzle(&ins
->src
[0], FALSE
, write_mask
, dst_swizzle
);
2820 if (src_mod
== WINED3DSPSM_DZ
) {
2821 glsl_src_param_t div_param
;
2822 unsigned int mask_size
= shader_glsl_get_write_mask_size(write_mask
);
2823 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_2
, &div_param
);
2825 if (mask_size
> 1) {
2826 shader_addline(buffer
, "gl_TexCoord[%u]%s / vec%d(%s));\n", reg
, dst_swizzle
, mask_size
, div_param
.param_str
);
2828 shader_addline(buffer
, "gl_TexCoord[%u]%s / %s);\n", reg
, dst_swizzle
, div_param
.param_str
);
2830 } else if (src_mod
== WINED3DSPSM_DW
) {
2831 glsl_src_param_t div_param
;
2832 unsigned int mask_size
= shader_glsl_get_write_mask_size(write_mask
);
2833 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_3
, &div_param
);
2835 if (mask_size
> 1) {
2836 shader_addline(buffer
, "gl_TexCoord[%u]%s / vec%d(%s));\n", reg
, dst_swizzle
, mask_size
, div_param
.param_str
);
2838 shader_addline(buffer
, "gl_TexCoord[%u]%s / %s);\n", reg
, dst_swizzle
, div_param
.param_str
);
2841 shader_addline(buffer
, "gl_TexCoord[%u]%s);\n", reg
, dst_swizzle
);
2846 /** Process the WINED3DSIO_TEXDP3TEX instruction in GLSL:
2847 * Take a 3-component dot product of the TexCoord[dstreg] and src,
2848 * then perform a 1D texture lookup from stage dstregnum, place into dst. */
2849 static void shader_glsl_texdp3tex(const struct wined3d_shader_instruction
*ins
)
2851 glsl_src_param_t src0_param
;
2852 glsl_sample_function_t sample_function
;
2853 DWORD sampler_idx
= ins
->dst
[0].reg
.idx
;
2854 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2855 WINED3DSAMPLER_TEXTURE_TYPE sampler_type
= ins
->ctx
->reg_maps
->sampler_type
[sampler_idx
];
2858 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
2860 /* Do I have to take care about the projected bit? I don't think so, since the dp3 returns only one
2861 * scalar, and projected sampling would require 4.
2863 * It is a dependent read - not valid with conditional NP2 textures
2865 shader_glsl_get_sample_function(sampler_type
, 0, &sample_function
);
2866 mask_size
= shader_glsl_get_write_mask_size(sample_function
.coord_mask
);
2871 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, WINED3DSP_NOSWIZZLE
, NULL
, NULL
, NULL
,
2872 "dot(gl_TexCoord[%u].xyz, %s)", sampler_idx
, src0_param
.param_str
);
2876 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, WINED3DSP_NOSWIZZLE
, NULL
, NULL
, NULL
,
2877 "vec2(dot(gl_TexCoord[%u].xyz, %s), 0.0)", sampler_idx
, src0_param
.param_str
);
2881 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, WINED3DSP_NOSWIZZLE
, NULL
, NULL
, NULL
,
2882 "vec3(dot(gl_TexCoord[%u].xyz, %s), 0.0, 0.0)", sampler_idx
, src0_param
.param_str
);
2886 FIXME("Unexpected mask size %u\n", mask_size
);
2891 /** Process the WINED3DSIO_TEXDP3 instruction in GLSL:
2892 * Take a 3-component dot product of the TexCoord[dstreg] and src. */
2893 static void shader_glsl_texdp3(const struct wined3d_shader_instruction
*ins
)
2895 glsl_src_param_t src0_param
;
2896 DWORD dstreg
= ins
->dst
[0].reg
.idx
;
2897 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2899 unsigned int mask_size
;
2901 dst_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2902 mask_size
= shader_glsl_get_write_mask_size(dst_mask
);
2903 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
2905 if (mask_size
> 1) {
2906 shader_addline(ins
->ctx
->buffer
, "vec%d(dot(T%u.xyz, %s)));\n", mask_size
, dstreg
, src0_param
.param_str
);
2908 shader_addline(ins
->ctx
->buffer
, "dot(T%u.xyz, %s));\n", dstreg
, src0_param
.param_str
);
2912 /** Process the WINED3DSIO_TEXDEPTH instruction in GLSL:
2913 * Calculate the depth as dst.x / dst.y */
2914 static void shader_glsl_texdepth(const struct wined3d_shader_instruction
*ins
)
2916 glsl_dst_param_t dst_param
;
2918 shader_glsl_add_dst_param(ins
, &ins
->dst
[0], &dst_param
);
2920 /* Tests show that texdepth never returns anything below 0.0, and that r5.y is clamped to 1.0.
2921 * Negative input is accepted, -0.25 / -0.5 returns 0.5. GL should clamp gl_FragDepth to [0;1], but
2922 * this doesn't always work, so clamp the results manually. Whether or not the x value is clamped at 1
2923 * too is irrelevant, since if x = 0, any y value < 1.0 (and > 1.0 is not allowed) results in a result
2926 shader_addline(ins
->ctx
->buffer
, "gl_FragDepth = clamp((%s.x / min(%s.y, 1.0)), 0.0, 1.0);\n",
2927 dst_param
.reg_name
, dst_param
.reg_name
);
2930 /** Process the WINED3DSIO_TEXM3X2DEPTH instruction in GLSL:
2931 * Last row of a 3x2 matrix multiply, use the result to calculate the depth:
2932 * Calculate tmp0.y = TexCoord[dstreg] . src.xyz; (tmp0.x has already been calculated)
2933 * depth = (tmp0.y == 0.0) ? 1.0 : tmp0.x / tmp0.y
2935 static void shader_glsl_texm3x2depth(const struct wined3d_shader_instruction
*ins
)
2937 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2938 DWORD dstreg
= ins
->dst
[0].reg
.idx
;
2939 glsl_src_param_t src0_param
;
2941 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
2943 shader_addline(ins
->ctx
->buffer
, "tmp0.y = dot(T%u.xyz, %s);\n", dstreg
, src0_param
.param_str
);
2944 shader_addline(ins
->ctx
->buffer
, "gl_FragDepth = (tmp0.y == 0.0) ? 1.0 : clamp(tmp0.x / tmp0.y, 0.0, 1.0);\n");
2947 /** Process the WINED3DSIO_TEXM3X2PAD instruction in GLSL
2948 * Calculate the 1st of a 2-row matrix multiplication. */
2949 static void shader_glsl_texm3x2pad(const struct wined3d_shader_instruction
*ins
)
2951 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2952 DWORD reg
= ins
->dst
[0].reg
.idx
;
2953 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2954 glsl_src_param_t src0_param
;
2956 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
2957 shader_addline(buffer
, "tmp0.x = dot(T%u.xyz, %s);\n", reg
, src0_param
.param_str
);
2960 /** Process the WINED3DSIO_TEXM3X3PAD instruction in GLSL
2961 * Calculate the 1st or 2nd row of a 3-row matrix multiplication. */
2962 static void shader_glsl_texm3x3pad(const struct wined3d_shader_instruction
*ins
)
2964 IWineD3DPixelShaderImpl
*shader
= (IWineD3DPixelShaderImpl
*)ins
->ctx
->shader
;
2965 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2966 DWORD reg
= ins
->dst
[0].reg
.idx
;
2967 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2968 SHADER_PARSE_STATE
* current_state
= &shader
->baseShader
.parse_state
;
2969 glsl_src_param_t src0_param
;
2971 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
2972 shader_addline(buffer
, "tmp0.%c = dot(T%u.xyz, %s);\n", 'x' + current_state
->current_row
, reg
, src0_param
.param_str
);
2973 current_state
->texcoord_w
[current_state
->current_row
++] = reg
;
2976 static void shader_glsl_texm3x2tex(const struct wined3d_shader_instruction
*ins
)
2978 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2979 DWORD reg
= ins
->dst
[0].reg
.idx
;
2980 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2981 glsl_src_param_t src0_param
;
2982 WINED3DSAMPLER_TEXTURE_TYPE sampler_type
= ins
->ctx
->reg_maps
->sampler_type
[reg
];
2983 glsl_sample_function_t sample_function
;
2985 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
2986 shader_addline(buffer
, "tmp0.y = dot(T%u.xyz, %s);\n", reg
, src0_param
.param_str
);
2988 shader_glsl_get_sample_function(sampler_type
, 0, &sample_function
);
2990 /* Sample the texture using the calculated coordinates */
2991 shader_glsl_gen_sample_code(ins
, reg
, &sample_function
, WINED3DSP_NOSWIZZLE
, NULL
, NULL
, NULL
, "tmp0.xy");
2994 /** Process the WINED3DSIO_TEXM3X3TEX instruction in GLSL
2995 * Perform the 3rd row of a 3x3 matrix multiply, then sample the texture using the calculated coordinates */
2996 static void shader_glsl_texm3x3tex(const struct wined3d_shader_instruction
*ins
)
2998 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2999 glsl_src_param_t src0_param
;
3000 DWORD reg
= ins
->dst
[0].reg
.idx
;
3001 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*)ins
->ctx
->shader
;
3002 SHADER_PARSE_STATE
* current_state
= &This
->baseShader
.parse_state
;
3003 WINED3DSAMPLER_TEXTURE_TYPE sampler_type
= ins
->ctx
->reg_maps
->sampler_type
[reg
];
3004 glsl_sample_function_t sample_function
;
3006 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
3007 shader_addline(ins
->ctx
->buffer
, "tmp0.z = dot(T%u.xyz, %s);\n", reg
, src0_param
.param_str
);
3009 /* Dependent read, not valid with conditional NP2 */
3010 shader_glsl_get_sample_function(sampler_type
, 0, &sample_function
);
3012 /* Sample the texture using the calculated coordinates */
3013 shader_glsl_gen_sample_code(ins
, reg
, &sample_function
, WINED3DSP_NOSWIZZLE
, NULL
, NULL
, NULL
, "tmp0.xyz");
3015 current_state
->current_row
= 0;
3018 /** Process the WINED3DSIO_TEXM3X3 instruction in GLSL
3019 * Perform the 3rd row of a 3x3 matrix multiply */
3020 static void shader_glsl_texm3x3(const struct wined3d_shader_instruction
*ins
)
3022 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
3023 glsl_src_param_t src0_param
;
3025 DWORD reg
= ins
->dst
[0].reg
.idx
;
3026 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*)ins
->ctx
->shader
;
3027 SHADER_PARSE_STATE
* current_state
= &This
->baseShader
.parse_state
;
3029 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
3031 shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
3032 shader_glsl_get_write_mask(&ins
->dst
[0], dst_mask
);
3033 shader_addline(ins
->ctx
->buffer
, "vec4(tmp0.xy, dot(T%u.xyz, %s), 1.0)%s);\n", reg
, src0_param
.param_str
, dst_mask
);
3035 current_state
->current_row
= 0;
3038 /* Process the WINED3DSIO_TEXM3X3SPEC instruction in GLSL
3039 * Perform the final texture lookup based on the previous 2 3x3 matrix multiplies */
3040 static void shader_glsl_texm3x3spec(const struct wined3d_shader_instruction
*ins
)
3042 IWineD3DPixelShaderImpl
*shader
= (IWineD3DPixelShaderImpl
*)ins
->ctx
->shader
;
3043 DWORD reg
= ins
->dst
[0].reg
.idx
;
3044 glsl_src_param_t src0_param
;
3045 glsl_src_param_t src1_param
;
3046 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
3047 SHADER_PARSE_STATE
* current_state
= &shader
->baseShader
.parse_state
;
3048 WINED3DSAMPLER_TEXTURE_TYPE stype
= ins
->ctx
->reg_maps
->sampler_type
[reg
];
3049 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
3050 glsl_sample_function_t sample_function
;
3052 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
3053 shader_glsl_add_src_param(ins
, &ins
->src
[1], src_mask
, &src1_param
);
3055 /* Perform the last matrix multiply operation */
3056 shader_addline(buffer
, "tmp0.z = dot(T%u.xyz, %s);\n", reg
, src0_param
.param_str
);
3057 /* Reflection calculation */
3058 shader_addline(buffer
, "tmp0.xyz = -reflect((%s), normalize(tmp0.xyz));\n", src1_param
.param_str
);
3060 /* Dependent read, not valid with conditional NP2 */
3061 shader_glsl_get_sample_function(stype
, 0, &sample_function
);
3063 /* Sample the texture */
3064 shader_glsl_gen_sample_code(ins
, reg
, &sample_function
, WINED3DSP_NOSWIZZLE
, NULL
, NULL
, NULL
, "tmp0.xyz");
3066 current_state
->current_row
= 0;
3069 /* Process the WINED3DSIO_TEXM3X3VSPEC instruction in GLSL
3070 * Perform the final texture lookup based on the previous 2 3x3 matrix multiplies */
3071 static void shader_glsl_texm3x3vspec(const struct wined3d_shader_instruction
*ins
)
3073 IWineD3DPixelShaderImpl
*shader
= (IWineD3DPixelShaderImpl
*)ins
->ctx
->shader
;
3074 DWORD reg
= ins
->dst
[0].reg
.idx
;
3075 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
3076 SHADER_PARSE_STATE
* current_state
= &shader
->baseShader
.parse_state
;
3077 glsl_src_param_t src0_param
;
3078 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
3079 WINED3DSAMPLER_TEXTURE_TYPE sampler_type
= ins
->ctx
->reg_maps
->sampler_type
[reg
];
3080 glsl_sample_function_t sample_function
;
3082 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
3084 /* Perform the last matrix multiply operation */
3085 shader_addline(buffer
, "tmp0.z = dot(vec3(T%u), vec3(%s));\n", reg
, src0_param
.param_str
);
3087 /* Construct the eye-ray vector from w coordinates */
3088 shader_addline(buffer
, "tmp1.xyz = normalize(vec3(gl_TexCoord[%u].w, gl_TexCoord[%u].w, gl_TexCoord[%u].w));\n",
3089 current_state
->texcoord_w
[0], current_state
->texcoord_w
[1], reg
);
3090 shader_addline(buffer
, "tmp0.xyz = -reflect(tmp1.xyz, normalize(tmp0.xyz));\n");
3092 /* Dependent read, not valid with conditional NP2 */
3093 shader_glsl_get_sample_function(sampler_type
, 0, &sample_function
);
3095 /* Sample the texture using the calculated coordinates */
3096 shader_glsl_gen_sample_code(ins
, reg
, &sample_function
, WINED3DSP_NOSWIZZLE
, NULL
, NULL
, NULL
, "tmp0.xyz");
3098 current_state
->current_row
= 0;
3101 /** Process the WINED3DSIO_TEXBEM instruction in GLSL.
3102 * Apply a fake bump map transform.
3103 * texbem is pshader <= 1.3 only, this saves a few version checks
3105 static void shader_glsl_texbem(const struct wined3d_shader_instruction
*ins
)
3107 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*)ins
->ctx
->shader
;
3108 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
3109 glsl_sample_function_t sample_function
;
3110 glsl_src_param_t coord_param
;
3111 WINED3DSAMPLER_TEXTURE_TYPE sampler_type
;
3117 sampler_idx
= ins
->dst
[0].reg
.idx
;
3118 flags
= deviceImpl
->stateBlock
->textureState
[sampler_idx
][WINED3DTSS_TEXTURETRANSFORMFLAGS
];
3120 sampler_type
= ins
->ctx
->reg_maps
->sampler_type
[sampler_idx
];
3121 /* Dependent read, not valid with conditional NP2 */
3122 shader_glsl_get_sample_function(sampler_type
, 0, &sample_function
);
3123 mask
= sample_function
.coord_mask
;
3125 shader_glsl_write_mask_to_str(mask
, coord_mask
);
3127 /* with projective textures, texbem only divides the static texture coord, not the displacement,
3128 * so we can't let the GL handle this.
3130 if (flags
& WINED3DTTFF_PROJECTED
) {
3132 char coord_div_mask
[3];
3133 switch (flags
& ~WINED3DTTFF_PROJECTED
) {
3134 case WINED3DTTFF_COUNT1
: FIXME("WINED3DTTFF_PROJECTED with WINED3DTTFF_COUNT1?\n"); break;
3135 case WINED3DTTFF_COUNT2
: div_mask
= WINED3DSP_WRITEMASK_1
; break;
3136 case WINED3DTTFF_COUNT3
: div_mask
= WINED3DSP_WRITEMASK_2
; break;
3137 case WINED3DTTFF_COUNT4
:
3138 case WINED3DTTFF_DISABLE
: div_mask
= WINED3DSP_WRITEMASK_3
; break;
3140 shader_glsl_write_mask_to_str(div_mask
, coord_div_mask
);
3141 shader_addline(ins
->ctx
->buffer
, "T%u%s /= T%u%s;\n", sampler_idx
, coord_mask
, sampler_idx
, coord_div_mask
);
3144 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
, &coord_param
);
3146 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, WINED3DSP_NOSWIZZLE
, NULL
, NULL
, NULL
,
3147 "T%u%s + vec4(bumpenvmat%d * %s, 0.0, 0.0)%s", sampler_idx
, coord_mask
, sampler_idx
,
3148 coord_param
.param_str
, coord_mask
);
3150 if (ins
->handler_idx
== WINED3DSIH_TEXBEML
)
3152 glsl_src_param_t luminance_param
;
3153 glsl_dst_param_t dst_param
;
3155 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_2
, &luminance_param
);
3156 shader_glsl_add_dst_param(ins
, &ins
->dst
[0], &dst_param
);
3158 shader_addline(ins
->ctx
->buffer
, "%s%s *= (%s * luminancescale%d + luminanceoffset%d);\n",
3159 dst_param
.reg_name
, dst_param
.mask_str
,
3160 luminance_param
.param_str
, sampler_idx
, sampler_idx
);
3164 static void shader_glsl_bem(const struct wined3d_shader_instruction
*ins
)
3166 glsl_src_param_t src0_param
, src1_param
;
3167 DWORD sampler_idx
= ins
->dst
[0].reg
.idx
;
3169 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
, &src0_param
);
3170 shader_glsl_add_src_param(ins
, &ins
->src
[1], WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
, &src1_param
);
3172 shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
3173 shader_addline(ins
->ctx
->buffer
, "%s + bumpenvmat%d * %s);\n",
3174 src0_param
.param_str
, sampler_idx
, src1_param
.param_str
);
3177 /** Process the WINED3DSIO_TEXREG2AR instruction in GLSL
3178 * Sample 2D texture at dst using the alpha & red (wx) components of src as texture coordinates */
3179 static void shader_glsl_texreg2ar(const struct wined3d_shader_instruction
*ins
)
3181 glsl_src_param_t src0_param
;
3182 DWORD sampler_idx
= ins
->dst
[0].reg
.idx
;
3183 WINED3DSAMPLER_TEXTURE_TYPE sampler_type
= ins
->ctx
->reg_maps
->sampler_type
[sampler_idx
];
3184 glsl_sample_function_t sample_function
;
3186 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_ALL
, &src0_param
);
3188 shader_glsl_get_sample_function(sampler_type
, 0, &sample_function
);
3189 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, WINED3DSP_NOSWIZZLE
, NULL
, NULL
, NULL
,
3190 "%s.wx", src0_param
.reg_name
);
3193 /** Process the WINED3DSIO_TEXREG2GB instruction in GLSL
3194 * Sample 2D texture at dst using the green & blue (yz) components of src as texture coordinates */
3195 static void shader_glsl_texreg2gb(const struct wined3d_shader_instruction
*ins
)
3197 glsl_src_param_t src0_param
;
3198 DWORD sampler_idx
= ins
->dst
[0].reg
.idx
;
3199 WINED3DSAMPLER_TEXTURE_TYPE sampler_type
= ins
->ctx
->reg_maps
->sampler_type
[sampler_idx
];
3200 glsl_sample_function_t sample_function
;
3202 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_ALL
, &src0_param
);
3204 shader_glsl_get_sample_function(sampler_type
, 0, &sample_function
);
3205 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, WINED3DSP_NOSWIZZLE
, NULL
, NULL
, NULL
,
3206 "%s.yz", src0_param
.reg_name
);
3209 /** Process the WINED3DSIO_TEXREG2RGB instruction in GLSL
3210 * Sample texture at dst using the rgb (xyz) components of src as texture coordinates */
3211 static void shader_glsl_texreg2rgb(const struct wined3d_shader_instruction
*ins
)
3213 glsl_src_param_t src0_param
;
3214 DWORD sampler_idx
= ins
->dst
[0].reg
.idx
;
3215 WINED3DSAMPLER_TEXTURE_TYPE sampler_type
= ins
->ctx
->reg_maps
->sampler_type
[sampler_idx
];
3216 glsl_sample_function_t sample_function
;
3218 /* Dependent read, not valid with conditional NP2 */
3219 shader_glsl_get_sample_function(sampler_type
, 0, &sample_function
);
3220 shader_glsl_add_src_param(ins
, &ins
->src
[0], sample_function
.coord_mask
, &src0_param
);
3222 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, WINED3DSP_NOSWIZZLE
, NULL
, NULL
, NULL
,
3223 "%s", src0_param
.param_str
);
3226 /** Process the WINED3DSIO_TEXKILL instruction in GLSL.
3227 * If any of the first 3 components are < 0, discard this pixel */
3228 static void shader_glsl_texkill(const struct wined3d_shader_instruction
*ins
)
3230 glsl_dst_param_t dst_param
;
3232 /* The argument is a destination parameter, and no writemasks are allowed */
3233 shader_glsl_add_dst_param(ins
, &ins
->dst
[0], &dst_param
);
3234 if (ins
->ctx
->reg_maps
->shader_version
.major
>= 2)
3236 /* 2.0 shaders compare all 4 components in texkill */
3237 shader_addline(ins
->ctx
->buffer
, "if (any(lessThan(%s.xyzw, vec4(0.0)))) discard;\n", dst_param
.reg_name
);
3239 /* 1.X shaders only compare the first 3 components, probably due to the nature of the texkill
3240 * instruction as a tex* instruction, and phase, which kills all a / w components. Even if all
3241 * 4 components are defined, only the first 3 are used
3243 shader_addline(ins
->ctx
->buffer
, "if (any(lessThan(%s.xyz, vec3(0.0)))) discard;\n", dst_param
.reg_name
);
3247 /** Process the WINED3DSIO_DP2ADD instruction in GLSL.
3248 * dst = dot2(src0, src1) + src2 */
3249 static void shader_glsl_dp2add(const struct wined3d_shader_instruction
*ins
)
3251 glsl_src_param_t src0_param
;
3252 glsl_src_param_t src1_param
;
3253 glsl_src_param_t src2_param
;
3255 unsigned int mask_size
;
3257 write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
3258 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
3260 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
, &src0_param
);
3261 shader_glsl_add_src_param(ins
, &ins
->src
[1], WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
, &src1_param
);
3262 shader_glsl_add_src_param(ins
, &ins
->src
[2], WINED3DSP_WRITEMASK_0
, &src2_param
);
3264 if (mask_size
> 1) {
3265 shader_addline(ins
->ctx
->buffer
, "vec%d(dot(%s, %s) + %s));\n",
3266 mask_size
, src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
3268 shader_addline(ins
->ctx
->buffer
, "dot(%s, %s) + %s);\n",
3269 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
3273 static void shader_glsl_input_pack(IWineD3DPixelShader
*iface
, struct wined3d_shader_buffer
*buffer
,
3274 const struct wined3d_shader_signature_element
*input_signature
, const struct shader_reg_maps
*reg_maps
,
3275 enum vertexprocessing_mode vertexprocessing
)
3278 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*)iface
;
3279 WORD map
= reg_maps
->input_registers
;
3281 for (i
= 0; map
; map
>>= 1, ++i
)
3283 const char *semantic_name
;
3288 if (!(map
& 1)) continue;
3290 semantic_name
= input_signature
[i
].semantic_name
;
3291 semantic_idx
= input_signature
[i
].semantic_idx
;
3292 shader_glsl_write_mask_to_str(input_signature
[i
].mask
, reg_mask
);
3294 if (shader_match_semantic(semantic_name
, WINED3DDECLUSAGE_TEXCOORD
))
3296 if (semantic_idx
< 8 && vertexprocessing
== pretransformed
)
3297 shader_addline(buffer
, "IN[%u]%s = gl_TexCoord[%u]%s;\n",
3298 This
->input_reg_map
[i
], reg_mask
, semantic_idx
, reg_mask
);
3300 shader_addline(buffer
, "IN[%u]%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
3301 This
->input_reg_map
[i
], reg_mask
, reg_mask
);
3303 else if (shader_match_semantic(semantic_name
, WINED3DDECLUSAGE_COLOR
))
3305 if (semantic_idx
== 0)
3306 shader_addline(buffer
, "IN[%u]%s = vec4(gl_Color)%s;\n",
3307 This
->input_reg_map
[i
], reg_mask
, reg_mask
);
3308 else if (semantic_idx
== 1)
3309 shader_addline(buffer
, "IN[%u]%s = vec4(gl_SecondaryColor)%s;\n",
3310 This
->input_reg_map
[i
], reg_mask
, reg_mask
);
3312 shader_addline(buffer
, "IN[%u]%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
3313 This
->input_reg_map
[i
], reg_mask
, reg_mask
);
3317 shader_addline(buffer
, "IN[%u]%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
3318 This
->input_reg_map
[i
], reg_mask
, reg_mask
);
3323 /*********************************************
3324 * Vertex Shader Specific Code begins here
3325 ********************************************/
3327 static void add_glsl_program_entry(struct shader_glsl_priv
*priv
, struct glsl_shader_prog_link
*entry
) {
3328 glsl_program_key_t key
;
3330 key
.vshader
= entry
->vshader
;
3331 key
.pshader
= entry
->pshader
;
3332 key
.vs_args
= entry
->vs_args
;
3333 key
.ps_args
= entry
->ps_args
;
3335 if (wine_rb_put(&priv
->program_lookup
, &key
, &entry
->program_lookup_entry
) == -1)
3337 ERR("Failed to insert program entry.\n");
3341 static struct glsl_shader_prog_link
*get_glsl_program_entry(struct shader_glsl_priv
*priv
,
3342 IWineD3DVertexShader
*vshader
, IWineD3DPixelShader
*pshader
, struct vs_compile_args
*vs_args
,
3343 struct ps_compile_args
*ps_args
) {
3344 struct wine_rb_entry
*entry
;
3345 glsl_program_key_t key
;
3347 key
.vshader
= vshader
;
3348 key
.pshader
= pshader
;
3349 key
.vs_args
= *vs_args
;
3350 key
.ps_args
= *ps_args
;
3352 entry
= wine_rb_get(&priv
->program_lookup
, &key
);
3353 return entry
? WINE_RB_ENTRY_VALUE(entry
, struct glsl_shader_prog_link
, program_lookup_entry
) : NULL
;
3356 /* GL locking is done by the caller */
3357 static void delete_glsl_program_entry(struct shader_glsl_priv
*priv
, const struct wined3d_gl_info
*gl_info
,
3358 struct glsl_shader_prog_link
*entry
)
3360 glsl_program_key_t key
;
3362 key
.vshader
= entry
->vshader
;
3363 key
.pshader
= entry
->pshader
;
3364 key
.vs_args
= entry
->vs_args
;
3365 key
.ps_args
= entry
->ps_args
;
3366 wine_rb_remove(&priv
->program_lookup
, &key
);
3368 GL_EXTCALL(glDeleteObjectARB(entry
->programId
));
3369 if (entry
->vshader
) list_remove(&entry
->vshader_entry
);
3370 if (entry
->pshader
) list_remove(&entry
->pshader_entry
);
3371 HeapFree(GetProcessHeap(), 0, entry
->vuniformF_locations
);
3372 HeapFree(GetProcessHeap(), 0, entry
->puniformF_locations
);
3373 HeapFree(GetProcessHeap(), 0, entry
);
3376 static void handle_ps3_input(struct wined3d_shader_buffer
*buffer
, const struct wined3d_gl_info
*gl_info
, const DWORD
*map
,
3377 const struct wined3d_shader_signature_element
*input_signature
, const struct shader_reg_maps
*reg_maps_in
,
3378 const struct wined3d_shader_signature_element
*output_signature
, const struct shader_reg_maps
*reg_maps_out
)
3381 const char *semantic_name_in
, *semantic_name_out
;
3382 UINT semantic_idx_in
, semantic_idx_out
;
3385 unsigned int in_count
= vec4_varyings(3, gl_info
);
3386 char reg_mask
[6], reg_mask_out
[6];
3387 char destination
[50];
3388 WORD input_map
, output_map
;
3390 set
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*set
) * (in_count
+ 2));
3392 if (!output_signature
)
3394 /* Save gl_FrontColor & gl_FrontSecondaryColor before overwriting them. */
3395 shader_addline(buffer
, "vec4 front_color = gl_FrontColor;\n");
3396 shader_addline(buffer
, "vec4 front_secondary_color = gl_FrontSecondaryColor;\n");
3399 input_map
= reg_maps_in
->input_registers
;
3400 for (i
= 0; input_map
; input_map
>>= 1, ++i
)
3402 if (!(input_map
& 1)) continue;
3405 if (in_idx
>= (in_count
+ 2)) {
3406 FIXME("More input varyings declared than supported, expect issues\n");
3409 else if (map
[i
] == ~0U)
3411 /* Declared, but not read register */
3415 if (in_idx
== in_count
) {
3416 sprintf(destination
, "gl_FrontColor");
3417 } else if (in_idx
== in_count
+ 1) {
3418 sprintf(destination
, "gl_FrontSecondaryColor");
3420 sprintf(destination
, "IN[%u]", in_idx
);
3423 semantic_name_in
= input_signature
[i
].semantic_name
;
3424 semantic_idx_in
= input_signature
[i
].semantic_idx
;
3425 set
[map
[i
]] = input_signature
[i
].mask
;
3426 shader_glsl_write_mask_to_str(input_signature
[i
].mask
, reg_mask
);
3428 if (!output_signature
)
3430 if (shader_match_semantic(semantic_name_in
, WINED3DDECLUSAGE_COLOR
))
3432 if (semantic_idx_in
== 0)
3433 shader_addline(buffer
, "%s%s = front_color%s;\n",
3434 destination
, reg_mask
, reg_mask
);
3435 else if (semantic_idx_in
== 1)
3436 shader_addline(buffer
, "%s%s = front_secondary_color%s;\n",
3437 destination
, reg_mask
, reg_mask
);
3439 shader_addline(buffer
, "%s%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
3440 destination
, reg_mask
, reg_mask
);
3442 else if (shader_match_semantic(semantic_name_in
, WINED3DDECLUSAGE_TEXCOORD
))
3444 if (semantic_idx_in
< 8)
3446 shader_addline(buffer
, "%s%s = gl_TexCoord[%u]%s;\n",
3447 destination
, reg_mask
, semantic_idx_in
, reg_mask
);
3451 shader_addline(buffer
, "%s%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
3452 destination
, reg_mask
, reg_mask
);
3455 else if (shader_match_semantic(semantic_name_in
, WINED3DDECLUSAGE_FOG
))
3457 shader_addline(buffer
, "%s%s = vec4(gl_FogFragCoord, 0.0, 0.0, 0.0)%s;\n",
3458 destination
, reg_mask
, reg_mask
);
3462 shader_addline(buffer
, "%s%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
3463 destination
, reg_mask
, reg_mask
);
3468 output_map
= reg_maps_out
->output_registers
;
3469 for (j
= 0; output_map
; output_map
>>= 1, ++j
)
3471 if (!(output_map
& 1)) continue;
3473 semantic_name_out
= output_signature
[j
].semantic_name
;
3474 semantic_idx_out
= output_signature
[j
].semantic_idx
;
3475 shader_glsl_write_mask_to_str(output_signature
[j
].mask
, reg_mask_out
);
3477 if (semantic_idx_in
== semantic_idx_out
3478 && !strcmp(semantic_name_in
, semantic_name_out
))
3480 shader_addline(buffer
, "%s%s = OUT[%u]%s;\n",
3481 destination
, reg_mask
, j
, reg_mask
);
3486 shader_addline(buffer
, "%s%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
3487 destination
, reg_mask
, reg_mask
);
3492 /* This is solely to make the compiler / linker happy and avoid warning about undefined
3493 * varyings. It shouldn't result in any real code executed on the GPU, since all read
3494 * input varyings are assigned above, if the optimizer works properly.
3496 for(i
= 0; i
< in_count
+ 2; i
++) {
3497 if (set
[i
] && set
[i
] != WINED3DSP_WRITEMASK_ALL
)
3499 unsigned int size
= 0;
3500 memset(reg_mask
, 0, sizeof(reg_mask
));
3501 if(!(set
[i
] & WINED3DSP_WRITEMASK_0
)) {
3502 reg_mask
[size
] = 'x';
3505 if(!(set
[i
] & WINED3DSP_WRITEMASK_1
)) {
3506 reg_mask
[size
] = 'y';
3509 if(!(set
[i
] & WINED3DSP_WRITEMASK_2
)) {
3510 reg_mask
[size
] = 'z';
3513 if(!(set
[i
] & WINED3DSP_WRITEMASK_3
)) {
3514 reg_mask
[size
] = 'w';
3518 if (i
== in_count
) {
3519 sprintf(destination
, "gl_FrontColor");
3520 } else if (i
== in_count
+ 1) {
3521 sprintf(destination
, "gl_FrontSecondaryColor");
3523 sprintf(destination
, "IN[%u]", i
);
3527 shader_addline(buffer
, "%s.%s = 0.0;\n", destination
, reg_mask
);
3529 shader_addline(buffer
, "%s.%s = vec%u(0.0);\n", destination
, reg_mask
, size
);
3534 HeapFree(GetProcessHeap(), 0, set
);
3537 /* GL locking is done by the caller */
3538 static GLhandleARB
generate_param_reorder_function(struct wined3d_shader_buffer
*buffer
,
3539 IWineD3DVertexShader
*vertexshader
, IWineD3DPixelShader
*pixelshader
, const struct wined3d_gl_info
*gl_info
)
3541 GLhandleARB ret
= 0;
3542 IWineD3DVertexShaderImpl
*vs
= (IWineD3DVertexShaderImpl
*) vertexshader
;
3543 IWineD3DPixelShaderImpl
*ps
= (IWineD3DPixelShaderImpl
*) pixelshader
;
3544 IWineD3DDeviceImpl
*device
;
3545 DWORD vs_major
= vs
->baseShader
.reg_maps
.shader_version
.major
;
3546 DWORD ps_major
= ps
? ps
->baseShader
.reg_maps
.shader_version
.major
: 0;
3548 const char *semantic_name
;
3551 const struct wined3d_shader_signature_element
*output_signature
;
3553 shader_buffer_clear(buffer
);
3555 shader_addline(buffer
, "#version 120\n");
3557 if(vs_major
< 3 && ps_major
< 3) {
3558 /* That one is easy: The vertex shader writes to the builtin varyings, the pixel shader reads from them.
3559 * Take care about the texcoord .w fixup though if we're using the fixed function fragment pipeline
3561 device
= (IWineD3DDeviceImpl
*) vs
->baseShader
.device
;
3562 if (((GLINFO_LOCATION
).quirks
& WINED3D_QUIRK_SET_TEXCOORD_W
)
3563 && ps_major
== 0 && vs_major
> 0 && !device
->frag_pipe
->ffp_proj_control
)
3565 shader_addline(buffer
, "void order_ps_input() {\n");
3566 for(i
= 0; i
< min(8, MAX_REG_TEXCRD
); i
++) {
3567 if(vs
->baseShader
.reg_maps
.texcoord_mask
[i
] != 0 &&
3568 vs
->baseShader
.reg_maps
.texcoord_mask
[i
] != WINED3DSP_WRITEMASK_ALL
) {
3569 shader_addline(buffer
, "gl_TexCoord[%u].w = 1.0;\n", i
);
3572 shader_addline(buffer
, "}\n");
3574 shader_addline(buffer
, "void order_ps_input() { /* do nothing */ }\n");
3576 } else if(ps_major
< 3 && vs_major
>= 3) {
3577 WORD map
= vs
->baseShader
.reg_maps
.output_registers
;
3579 /* The vertex shader writes to its own varyings, the pixel shader needs them in the builtin ones */
3580 output_signature
= vs
->output_signature
;
3582 shader_addline(buffer
, "void order_ps_input(in vec4 OUT[%u]) {\n", MAX_REG_OUTPUT
);
3583 for (i
= 0; map
; map
>>= 1, ++i
)
3587 if (!(map
& 1)) continue;
3589 semantic_name
= output_signature
[i
].semantic_name
;
3590 semantic_idx
= output_signature
[i
].semantic_idx
;
3591 write_mask
= output_signature
[i
].mask
;
3592 shader_glsl_write_mask_to_str(write_mask
, reg_mask
);
3594 if (shader_match_semantic(semantic_name
, WINED3DDECLUSAGE_COLOR
))
3596 if (semantic_idx
== 0)
3597 shader_addline(buffer
, "gl_FrontColor%s = OUT[%u]%s;\n", reg_mask
, i
, reg_mask
);
3598 else if (semantic_idx
== 1)
3599 shader_addline(buffer
, "gl_FrontSecondaryColor%s = OUT[%u]%s;\n", reg_mask
, i
, reg_mask
);
3601 else if (shader_match_semantic(semantic_name
, WINED3DDECLUSAGE_POSITION
))
3603 shader_addline(buffer
, "gl_Position%s = OUT[%u]%s;\n", reg_mask
, i
, reg_mask
);
3605 else if (shader_match_semantic(semantic_name
, WINED3DDECLUSAGE_TEXCOORD
))
3607 if (semantic_idx
< 8)
3609 if (!((GLINFO_LOCATION
).quirks
& WINED3D_QUIRK_SET_TEXCOORD_W
) || ps_major
> 0)
3610 write_mask
|= WINED3DSP_WRITEMASK_3
;
3612 shader_addline(buffer
, "gl_TexCoord[%u]%s = OUT[%u]%s;\n",
3613 semantic_idx
, reg_mask
, i
, reg_mask
);
3614 if (!(write_mask
& WINED3DSP_WRITEMASK_3
))
3615 shader_addline(buffer
, "gl_TexCoord[%u].w = 1.0;\n", semantic_idx
);
3618 else if (shader_match_semantic(semantic_name
, WINED3DDECLUSAGE_PSIZE
))
3620 shader_addline(buffer
, "gl_PointSize = OUT[%u].x;\n", i
);
3622 else if (shader_match_semantic(semantic_name
, WINED3DDECLUSAGE_FOG
))
3624 shader_addline(buffer
, "gl_FogFragCoord = OUT[%u].%c;\n", i
, reg_mask
[1]);
3627 shader_addline(buffer
, "}\n");
3629 } else if(ps_major
>= 3 && vs_major
>= 3) {
3630 WORD map
= vs
->baseShader
.reg_maps
.output_registers
;
3632 output_signature
= vs
->output_signature
;
3634 /* This one is tricky: a 3.0 pixel shader reads from a 3.0 vertex shader */
3635 shader_addline(buffer
, "varying vec4 IN[%u];\n", vec4_varyings(3, gl_info
));
3636 shader_addline(buffer
, "void order_ps_input(in vec4 OUT[%u]) {\n", MAX_REG_OUTPUT
);
3638 /* First, sort out position and point size. Those are not passed to the pixel shader */
3639 for (i
= 0; map
; map
>>= 1, ++i
)
3641 if (!(map
& 1)) continue;
3643 semantic_name
= output_signature
[i
].semantic_name
;
3644 shader_glsl_write_mask_to_str(output_signature
[i
].mask
, reg_mask
);
3646 if (shader_match_semantic(semantic_name
, WINED3DDECLUSAGE_POSITION
))
3648 shader_addline(buffer
, "gl_Position%s = OUT[%u]%s;\n", reg_mask
, i
, reg_mask
);
3650 else if (shader_match_semantic(semantic_name
, WINED3DDECLUSAGE_PSIZE
))
3652 shader_addline(buffer
, "gl_PointSize = OUT[%u].x;\n", i
);
3656 /* Then, fix the pixel shader input */
3657 handle_ps3_input(buffer
, gl_info
, ps
->input_reg_map
, ps
->input_signature
,
3658 &ps
->baseShader
.reg_maps
, output_signature
, &vs
->baseShader
.reg_maps
);
3660 shader_addline(buffer
, "}\n");
3661 } else if(ps_major
>= 3 && vs_major
< 3) {
3662 shader_addline(buffer
, "varying vec4 IN[%u];\n", vec4_varyings(3, gl_info
));
3663 shader_addline(buffer
, "void order_ps_input() {\n");
3664 /* The vertex shader wrote to the builtin varyings. There is no need to figure out position and
3665 * point size, but we depend on the optimizers kindness to find out that the pixel shader doesn't
3666 * read gl_TexCoord and gl_ColorX, otherwise we'll run out of varyings
3668 handle_ps3_input(buffer
, gl_info
, ps
->input_reg_map
, ps
->input_signature
,
3669 &ps
->baseShader
.reg_maps
, NULL
, NULL
);
3670 shader_addline(buffer
, "}\n");
3672 ERR("Unexpected vertex and pixel shader version condition: vs: %d, ps: %d\n", vs_major
, ps_major
);
3675 ret
= GL_EXTCALL(glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB
));
3676 checkGLcall("glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB)");
3677 GL_EXTCALL(glShaderSourceARB(ret
, 1, (const char**)&buffer
->buffer
, NULL
));
3678 checkGLcall("glShaderSourceARB(ret, 1, &buffer->buffer, NULL)");
3679 GL_EXTCALL(glCompileShaderARB(ret
));
3680 checkGLcall("glCompileShaderARB(ret)");
3685 /* GL locking is done by the caller */
3686 static void hardcode_local_constants(IWineD3DBaseShaderImpl
*shader
, const struct wined3d_gl_info
*gl_info
,
3687 GLhandleARB programId
, char prefix
)
3689 const local_constant
*lconst
;
3694 LIST_FOR_EACH_ENTRY(lconst
, &shader
->baseShader
.constantsF
, local_constant
, entry
) {
3695 value
= (const float *)lconst
->value
;
3696 snprintf(glsl_name
, sizeof(glsl_name
), "%cLC%u", prefix
, lconst
->idx
);
3697 tmp_loc
= GL_EXTCALL(glGetUniformLocationARB(programId
, glsl_name
));
3698 GL_EXTCALL(glUniform4fvARB(tmp_loc
, 1, value
));
3700 checkGLcall("Hardcoding local constants");
3703 /* GL locking is done by the caller */
3704 static GLuint
shader_glsl_generate_pshader(const struct wined3d_context
*context
,
3705 struct wined3d_shader_buffer
*buffer
, IWineD3DPixelShaderImpl
*This
,
3706 const struct ps_compile_args
*args
, struct ps_np2fixup_info
*np2fixup_info
)
3708 const struct shader_reg_maps
*reg_maps
= &This
->baseShader
.reg_maps
;
3709 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
3710 CONST DWORD
*function
= This
->baseShader
.function
;
3711 struct shader_glsl_ctx_priv priv_ctx
;
3713 /* Create the hw GLSL shader object and assign it as the shader->prgId */
3714 GLhandleARB shader_obj
= GL_EXTCALL(glCreateShaderObjectARB(GL_FRAGMENT_SHADER_ARB
));
3716 memset(&priv_ctx
, 0, sizeof(priv_ctx
));
3717 priv_ctx
.cur_ps_args
= args
;
3718 priv_ctx
.cur_np2fixup_info
= np2fixup_info
;
3720 shader_addline(buffer
, "#version 120\n");
3722 if(GL_SUPPORT(ARB_SHADER_TEXTURE_LOD
) && reg_maps
->usestexldd
) {
3723 shader_addline(buffer
, "#extension GL_ARB_shader_texture_lod : enable\n");
3725 if (GL_SUPPORT(ARB_TEXTURE_RECTANGLE
)) {
3726 /* The spec says that it doesn't have to be explicitly enabled, but the nvidia
3727 * drivers write a warning if we don't do so
3729 shader_addline(buffer
, "#extension GL_ARB_texture_rectangle : enable\n");
3732 /* Base Declarations */
3733 shader_generate_glsl_declarations(context
, buffer
, (IWineD3DBaseShader
*)This
, reg_maps
, &priv_ctx
);
3735 /* Pack 3.0 inputs */
3736 if (reg_maps
->shader_version
.major
>= 3 && args
->vp_mode
!= vertexshader
)
3738 shader_glsl_input_pack((IWineD3DPixelShader
*) This
, buffer
, This
->input_signature
, reg_maps
, args
->vp_mode
);
3741 /* Base Shader Body */
3742 shader_generate_main((IWineD3DBaseShader
*)This
, buffer
, reg_maps
, function
, &priv_ctx
);
3744 /* Pixel shaders < 2.0 place the resulting color in R0 implicitly */
3745 if (reg_maps
->shader_version
.major
< 2)
3747 /* Some older cards like GeforceFX ones don't support multiple buffers, so also not gl_FragData */
3748 shader_addline(buffer
, "gl_FragData[0] = R0;\n");
3751 if (args
->srgb_correction
)
3753 shader_addline(buffer
, "tmp0.xyz = pow(gl_FragData[0].xyz, vec3(srgb_const0.x));\n");
3754 shader_addline(buffer
, "tmp0.xyz = tmp0.xyz * vec3(srgb_const0.y) - vec3(srgb_const0.z);\n");
3755 shader_addline(buffer
, "tmp1.xyz = gl_FragData[0].xyz * vec3(srgb_const0.w);\n");
3756 shader_addline(buffer
, "bvec3 srgb_compare = lessThan(gl_FragData[0].xyz, vec3(srgb_const1.x));\n");
3757 shader_addline(buffer
, "gl_FragData[0].xyz = mix(tmp0.xyz, tmp1.xyz, vec3(srgb_compare));\n");
3758 shader_addline(buffer
, "gl_FragData[0] = clamp(gl_FragData[0], 0.0, 1.0);\n");
3760 /* Pixel shader < 3.0 do not replace the fog stage.
3761 * This implements linear fog computation and blending.
3762 * TODO: non linear fog
3763 * NOTE: gl_Fog.start and gl_Fog.end don't hold fog start s and end e but
3764 * -1/(e-s) and e/(e-s) respectively.
3766 if (reg_maps
->shader_version
.major
< 3)
3769 case FOG_OFF
: break;
3771 shader_addline(buffer
, "float fogstart = -1.0 / (gl_Fog.end - gl_Fog.start);\n");
3772 shader_addline(buffer
, "float fogend = gl_Fog.end * -fogstart;\n");
3773 shader_addline(buffer
, "float Fog = clamp(gl_FogFragCoord * fogstart + fogend, 0.0, 1.0);\n");
3774 shader_addline(buffer
, "gl_FragData[0].xyz = mix(gl_Fog.color.xyz, gl_FragData[0].xyz, Fog);\n");
3777 /* Fog = e^(-gl_Fog.density * gl_FogFragCoord) */
3778 shader_addline(buffer
, "float Fog = exp(-gl_Fog.density * gl_FogFragCoord);\n");
3779 shader_addline(buffer
, "Fog = clamp(Fog, 0.0, 1.0);\n");
3780 shader_addline(buffer
, "gl_FragData[0].xyz = mix(gl_Fog.color.xyz, gl_FragData[0].xyz, Fog);\n");
3783 /* Fog = e^(-(gl_Fog.density * gl_FogFragCoord)^2) */
3784 shader_addline(buffer
, "float Fog = exp(-gl_Fog.density * gl_Fog.density * gl_FogFragCoord * gl_FogFragCoord);\n");
3785 shader_addline(buffer
, "Fog = clamp(Fog, 0.0, 1.0);\n");
3786 shader_addline(buffer
, "gl_FragData[0].xyz = mix(gl_Fog.color.xyz, gl_FragData[0].xyz, Fog);\n");
3791 shader_addline(buffer
, "}\n");
3793 TRACE("Compiling shader object %u\n", shader_obj
);
3794 GL_EXTCALL(glShaderSourceARB(shader_obj
, 1, (const char**)&buffer
->buffer
, NULL
));
3795 GL_EXTCALL(glCompileShaderARB(shader_obj
));
3796 print_glsl_info_log(&GLINFO_LOCATION
, shader_obj
);
3798 /* Store the shader object */
3802 /* GL locking is done by the caller */
3803 static GLuint
shader_glsl_generate_vshader(const struct wined3d_context
*context
,
3804 struct wined3d_shader_buffer
*buffer
, IWineD3DVertexShaderImpl
*This
,
3805 const struct vs_compile_args
*args
)
3807 const struct shader_reg_maps
*reg_maps
= &This
->baseShader
.reg_maps
;
3808 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
3809 CONST DWORD
*function
= This
->baseShader
.function
;
3810 struct shader_glsl_ctx_priv priv_ctx
;
3812 /* Create the hw GLSL shader program and assign it as the shader->prgId */
3813 GLhandleARB shader_obj
= GL_EXTCALL(glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB
));
3815 shader_addline(buffer
, "#version 120\n");
3817 memset(&priv_ctx
, 0, sizeof(priv_ctx
));
3818 priv_ctx
.cur_vs_args
= args
;
3820 /* Base Declarations */
3821 shader_generate_glsl_declarations(context
, buffer
, (IWineD3DBaseShader
*)This
, reg_maps
, &priv_ctx
);
3823 /* Base Shader Body */
3824 shader_generate_main((IWineD3DBaseShader
*)This
, buffer
, reg_maps
, function
, &priv_ctx
);
3826 /* Unpack 3.0 outputs */
3827 if (reg_maps
->shader_version
.major
>= 3) shader_addline(buffer
, "order_ps_input(OUT);\n");
3828 else shader_addline(buffer
, "order_ps_input();\n");
3830 /* The D3DRS_FOGTABLEMODE render state defines if the shader-generated fog coord is used
3831 * or if the fragment depth is used. If the fragment depth is used(FOGTABLEMODE != NONE),
3832 * the fog frag coord is thrown away. If the fog frag coord is used, but not written by
3833 * the shader, it is set to 0.0(fully fogged, since start = 1.0, end = 0.0)
3835 if(args
->fog_src
== VS_FOG_Z
) {
3836 shader_addline(buffer
, "gl_FogFragCoord = gl_Position.z;\n");
3837 } else if (!reg_maps
->fog
) {
3838 shader_addline(buffer
, "gl_FogFragCoord = 0.0;\n");
3841 /* Write the final position.
3843 * OpenGL coordinates specify the center of the pixel while d3d coords specify
3844 * the corner. The offsets are stored in z and w in posFixup. posFixup.y contains
3845 * 1.0 or -1.0 to turn the rendering upside down for offscreen rendering. PosFixup.x
3846 * contains 1.0 to allow a mad.
3848 shader_addline(buffer
, "gl_Position.y = gl_Position.y * posFixup.y;\n");
3849 shader_addline(buffer
, "gl_Position.xy += posFixup.zw * gl_Position.ww;\n");
3850 shader_addline(buffer
, "gl_ClipVertex = gl_Position;\n");
3852 /* Z coord [0;1]->[-1;1] mapping, see comment in transform_projection in state.c
3854 * Basically we want (in homogeneous coordinates) z = z * 2 - 1. However, shaders are run
3855 * before the homogeneous divide, so we have to take the w into account: z = ((z / w) * 2 - 1) * w,
3856 * which is the same as z = z * 2 - w.
3858 shader_addline(buffer
, "gl_Position.z = gl_Position.z * 2.0 - gl_Position.w;\n");
3860 shader_addline(buffer
, "}\n");
3862 TRACE("Compiling shader object %u\n", shader_obj
);
3863 GL_EXTCALL(glShaderSourceARB(shader_obj
, 1, (const char**)&buffer
->buffer
, NULL
));
3864 GL_EXTCALL(glCompileShaderARB(shader_obj
));
3865 print_glsl_info_log(&GLINFO_LOCATION
, shader_obj
);
3870 static GLhandleARB
find_glsl_pshader(const struct wined3d_context
*context
,
3871 struct wined3d_shader_buffer
*buffer
, IWineD3DPixelShaderImpl
*shader
,
3872 const struct ps_compile_args
*args
, const struct ps_np2fixup_info
**np2fixup_info
)
3876 struct glsl_ps_compiled_shader
*new_array
;
3877 struct glsl_pshader_private
*shader_data
;
3878 struct ps_np2fixup_info
*np2fixup
= NULL
;
3881 if(!shader
->backend_priv
) {
3882 shader
->backend_priv
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*shader_data
));
3884 shader_data
= shader
->backend_priv
;
3886 /* Usually we have very few GL shaders for each d3d shader(just 1 or maybe 2),
3887 * so a linear search is more performant than a hashmap or a binary search
3888 * (cache coherency etc)
3890 for(i
= 0; i
< shader_data
->num_gl_shaders
; i
++) {
3891 if(memcmp(&shader_data
->gl_shaders
[i
].args
, args
, sizeof(*args
)) == 0) {
3892 if(args
->np2_fixup
) *np2fixup_info
= &shader_data
->gl_shaders
[i
].np2fixup
;
3893 return shader_data
->gl_shaders
[i
].prgId
;
3897 TRACE("No matching GL shader found for shader %p, compiling a new shader.\n", shader
);
3898 if(shader_data
->shader_array_size
== shader_data
->num_gl_shaders
) {
3899 if (shader_data
->num_gl_shaders
)
3901 new_size
= shader_data
->shader_array_size
+ max(1, shader_data
->shader_array_size
/ 2);
3902 new_array
= HeapReAlloc(GetProcessHeap(), 0, shader_data
->gl_shaders
,
3903 new_size
* sizeof(*shader_data
->gl_shaders
));
3905 new_array
= HeapAlloc(GetProcessHeap(), 0, sizeof(*shader_data
->gl_shaders
));
3910 ERR("Out of memory\n");
3913 shader_data
->gl_shaders
= new_array
;
3914 shader_data
->shader_array_size
= new_size
;
3917 shader_data
->gl_shaders
[shader_data
->num_gl_shaders
].args
= *args
;
3919 memset(&shader_data
->gl_shaders
[shader_data
->num_gl_shaders
].np2fixup
, 0, sizeof(struct ps_np2fixup_info
));
3920 if (args
->np2_fixup
) np2fixup
= &shader_data
->gl_shaders
[shader_data
->num_gl_shaders
].np2fixup
;
3922 pixelshader_update_samplers(&shader
->baseShader
.reg_maps
,
3923 ((IWineD3DDeviceImpl
*)shader
->baseShader
.device
)->stateBlock
->textures
);
3925 shader_buffer_clear(buffer
);
3926 ret
= shader_glsl_generate_pshader(context
, buffer
, shader
, args
, np2fixup
);
3927 shader_data
->gl_shaders
[shader_data
->num_gl_shaders
++].prgId
= ret
;
3928 *np2fixup_info
= np2fixup
;
3933 static inline BOOL
vs_args_equal(const struct vs_compile_args
*stored
, const struct vs_compile_args
*new,
3934 const DWORD use_map
) {
3935 if((stored
->swizzle_map
& use_map
) != new->swizzle_map
) return FALSE
;
3936 return stored
->fog_src
== new->fog_src
;
3939 static GLhandleARB
find_glsl_vshader(const struct wined3d_context
*context
,
3940 struct wined3d_shader_buffer
*buffer
, IWineD3DVertexShaderImpl
*shader
,
3941 const struct vs_compile_args
*args
)
3945 struct glsl_vs_compiled_shader
*new_array
;
3946 DWORD use_map
= ((IWineD3DDeviceImpl
*)shader
->baseShader
.device
)->strided_streams
.use_map
;
3947 struct glsl_vshader_private
*shader_data
;
3950 if(!shader
->backend_priv
) {
3951 shader
->backend_priv
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*shader_data
));
3953 shader_data
= shader
->backend_priv
;
3955 /* Usually we have very few GL shaders for each d3d shader(just 1 or maybe 2),
3956 * so a linear search is more performant than a hashmap or a binary search
3957 * (cache coherency etc)
3959 for(i
= 0; i
< shader_data
->num_gl_shaders
; i
++) {
3960 if(vs_args_equal(&shader_data
->gl_shaders
[i
].args
, args
, use_map
)) {
3961 return shader_data
->gl_shaders
[i
].prgId
;
3965 TRACE("No matching GL shader found for shader %p, compiling a new shader.\n", shader
);
3967 if(shader_data
->shader_array_size
== shader_data
->num_gl_shaders
) {
3968 if (shader_data
->num_gl_shaders
)
3970 new_size
= shader_data
->shader_array_size
+ max(1, shader_data
->shader_array_size
/ 2);
3971 new_array
= HeapReAlloc(GetProcessHeap(), 0, shader_data
->gl_shaders
,
3972 new_size
* sizeof(*shader_data
->gl_shaders
));
3974 new_array
= HeapAlloc(GetProcessHeap(), 0, sizeof(*shader_data
->gl_shaders
));
3979 ERR("Out of memory\n");
3982 shader_data
->gl_shaders
= new_array
;
3983 shader_data
->shader_array_size
= new_size
;
3986 shader_data
->gl_shaders
[shader_data
->num_gl_shaders
].args
= *args
;
3988 shader_buffer_clear(buffer
);
3989 ret
= shader_glsl_generate_vshader(context
, buffer
, shader
, args
);
3990 shader_data
->gl_shaders
[shader_data
->num_gl_shaders
++].prgId
= ret
;
3995 /** Sets the GLSL program ID for the given pixel and vertex shader combination.
3996 * It sets the programId on the current StateBlock (because it should be called
3997 * inside of the DrawPrimitive() part of the render loop).
3999 * If a program for the given combination does not exist, create one, and store
4000 * the program in the hash table. If it creates a program, it will link the
4001 * given objects, too.
4004 /* GL locking is done by the caller */
4005 static void set_glsl_shader_program(const struct wined3d_context
*context
,
4006 IWineD3DDeviceImpl
*device
, BOOL use_ps
, BOOL use_vs
)
4008 IWineD3DVertexShader
*vshader
= use_vs
? device
->stateBlock
->vertexShader
: NULL
;
4009 IWineD3DPixelShader
*pshader
= use_ps
? device
->stateBlock
->pixelShader
: NULL
;
4010 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
4011 struct shader_glsl_priv
*priv
= device
->shader_priv
;
4012 struct glsl_shader_prog_link
*entry
= NULL
;
4013 GLhandleARB programId
= 0;
4014 GLhandleARB reorder_shader_id
= 0;
4017 struct ps_compile_args ps_compile_args
;
4018 struct vs_compile_args vs_compile_args
;
4020 if (vshader
) find_vs_compile_args((IWineD3DVertexShaderImpl
*)vshader
, device
->stateBlock
, &vs_compile_args
);
4021 if (pshader
) find_ps_compile_args((IWineD3DPixelShaderImpl
*)pshader
, device
->stateBlock
, &ps_compile_args
);
4023 entry
= get_glsl_program_entry(priv
, vshader
, pshader
, &vs_compile_args
, &ps_compile_args
);
4025 priv
->glsl_program
= entry
;
4029 /* If we get to this point, then no matching program exists, so we create one */
4030 programId
= GL_EXTCALL(glCreateProgramObjectARB());
4031 TRACE("Created new GLSL shader program %u\n", programId
);
4033 /* Create the entry */
4034 entry
= HeapAlloc(GetProcessHeap(), 0, sizeof(struct glsl_shader_prog_link
));
4035 entry
->programId
= programId
;
4036 entry
->vshader
= vshader
;
4037 entry
->pshader
= pshader
;
4038 entry
->vs_args
= vs_compile_args
;
4039 entry
->ps_args
= ps_compile_args
;
4040 entry
->constant_version
= 0;
4041 entry
->np2Fixup_info
= NULL
;
4042 /* Add the hash table entry */
4043 add_glsl_program_entry(priv
, entry
);
4045 /* Set the current program */
4046 priv
->glsl_program
= entry
;
4048 /* Attach GLSL vshader */
4051 GLhandleARB vshader_id
= find_glsl_vshader(context
, &priv
->shader_buffer
,
4052 (IWineD3DVertexShaderImpl
*)vshader
, &vs_compile_args
);
4053 WORD map
= ((IWineD3DBaseShaderImpl
*)vshader
)->baseShader
.reg_maps
.input_registers
;
4056 reorder_shader_id
= generate_param_reorder_function(&priv
->shader_buffer
, vshader
, pshader
, gl_info
);
4057 TRACE("Attaching GLSL shader object %u to program %u\n", reorder_shader_id
, programId
);
4058 GL_EXTCALL(glAttachObjectARB(programId
, reorder_shader_id
));
4059 checkGLcall("glAttachObjectARB");
4060 /* Flag the reorder function for deletion, then it will be freed automatically when the program
4063 GL_EXTCALL(glDeleteObjectARB(reorder_shader_id
));
4065 TRACE("Attaching GLSL shader object %u to program %u\n", vshader_id
, programId
);
4066 GL_EXTCALL(glAttachObjectARB(programId
, vshader_id
));
4067 checkGLcall("glAttachObjectARB");
4069 /* Bind vertex attributes to a corresponding index number to match
4070 * the same index numbers as ARB_vertex_programs (makes loading
4071 * vertex attributes simpler). With this method, we can use the
4072 * exact same code to load the attributes later for both ARB and
4075 * We have to do this here because we need to know the Program ID
4076 * in order to make the bindings work, and it has to be done prior
4077 * to linking the GLSL program. */
4078 for (i
= 0; map
; map
>>= 1, ++i
)
4080 if (!(map
& 1)) continue;
4082 snprintf(tmp_name
, sizeof(tmp_name
), "attrib%u", i
);
4083 GL_EXTCALL(glBindAttribLocationARB(programId
, i
, tmp_name
));
4085 checkGLcall("glBindAttribLocationARB");
4087 list_add_head(&((IWineD3DBaseShaderImpl
*)vshader
)->baseShader
.linked_programs
, &entry
->vshader_entry
);
4090 /* Attach GLSL pshader */
4093 GLhandleARB pshader_id
= find_glsl_pshader(context
, &priv
->shader_buffer
,
4094 (IWineD3DPixelShaderImpl
*)pshader
, &ps_compile_args
, &entry
->np2Fixup_info
);
4095 TRACE("Attaching GLSL shader object %u to program %u\n", pshader_id
, programId
);
4096 GL_EXTCALL(glAttachObjectARB(programId
, pshader_id
));
4097 checkGLcall("glAttachObjectARB");
4099 list_add_head(&((IWineD3DBaseShaderImpl
*)pshader
)->baseShader
.linked_programs
, &entry
->pshader_entry
);
4102 /* Link the program */
4103 TRACE("Linking GLSL shader program %u\n", programId
);
4104 GL_EXTCALL(glLinkProgramARB(programId
));
4105 print_glsl_info_log(&GLINFO_LOCATION
, programId
);
4107 entry
->vuniformF_locations
= HeapAlloc(GetProcessHeap(), 0, sizeof(GLhandleARB
) * GL_LIMITS(vshader_constantsF
));
4108 for (i
= 0; i
< GL_LIMITS(vshader_constantsF
); ++i
) {
4109 snprintf(glsl_name
, sizeof(glsl_name
), "VC[%i]", i
);
4110 entry
->vuniformF_locations
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, glsl_name
));
4112 for (i
= 0; i
< MAX_CONST_I
; ++i
) {
4113 snprintf(glsl_name
, sizeof(glsl_name
), "VI[%i]", i
);
4114 entry
->vuniformI_locations
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, glsl_name
));
4116 entry
->puniformF_locations
= HeapAlloc(GetProcessHeap(), 0, sizeof(GLhandleARB
) * GL_LIMITS(pshader_constantsF
));
4117 for (i
= 0; i
< GL_LIMITS(pshader_constantsF
); ++i
) {
4118 snprintf(glsl_name
, sizeof(glsl_name
), "PC[%i]", i
);
4119 entry
->puniformF_locations
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, glsl_name
));
4121 for (i
= 0; i
< MAX_CONST_I
; ++i
) {
4122 snprintf(glsl_name
, sizeof(glsl_name
), "PI[%i]", i
);
4123 entry
->puniformI_locations
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, glsl_name
));
4129 for(i
= 0; i
< MAX_TEXTURES
; i
++) {
4130 sprintf(name
, "bumpenvmat%u", i
);
4131 entry
->bumpenvmat_location
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, name
));
4132 sprintf(name
, "luminancescale%u", i
);
4133 entry
->luminancescale_location
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, name
));
4134 sprintf(name
, "luminanceoffset%u", i
);
4135 entry
->luminanceoffset_location
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, name
));
4138 if (ps_compile_args
.np2_fixup
) {
4139 if (entry
->np2Fixup_info
) {
4140 entry
->np2Fixup_location
= GL_EXTCALL(glGetUniformLocationARB(programId
, "PsamplerNP2Fixup"));
4142 FIXME("NP2 texcoord fixup needed for this pixelshader, but no fixup uniform found.\n");
4147 entry
->posFixup_location
= GL_EXTCALL(glGetUniformLocationARB(programId
, "posFixup"));
4148 entry
->ycorrection_location
= GL_EXTCALL(glGetUniformLocationARB(programId
, "ycorrection"));
4149 checkGLcall("Find glsl program uniform locations");
4152 && ((IWineD3DPixelShaderImpl
*)pshader
)->baseShader
.reg_maps
.shader_version
.major
>= 3
4153 && ((IWineD3DPixelShaderImpl
*)pshader
)->declared_in_count
> vec4_varyings(3, gl_info
))
4155 TRACE("Shader %d needs vertex color clamping disabled\n", programId
);
4156 entry
->vertex_color_clamp
= GL_FALSE
;
4158 entry
->vertex_color_clamp
= GL_FIXED_ONLY_ARB
;
4161 /* Set the shader to allow uniform loading on it */
4162 GL_EXTCALL(glUseProgramObjectARB(programId
));
4163 checkGLcall("glUseProgramObjectARB(programId)");
4165 /* Load the vertex and pixel samplers now. The function that finds the mappings makes sure
4166 * that it stays the same for each vertexshader-pixelshader pair(=linked glsl program). If
4167 * a pshader with fixed function pipeline is used there are no vertex samplers, and if a
4168 * vertex shader with fixed function pixel processing is used we make sure that the card
4169 * supports enough samplers to allow the max number of vertex samplers with all possible
4170 * fixed function fragment processing setups. So once the program is linked these samplers
4173 if (vshader
) shader_glsl_load_vsamplers(gl_info
, device
->texUnitMap
, programId
);
4174 if (pshader
) shader_glsl_load_psamplers(gl_info
, device
->texUnitMap
, programId
);
4176 /* If the local constants do not have to be loaded with the environment constants,
4177 * load them now to have them hardcoded in the GLSL program. This saves some CPU cycles
4180 if(pshader
&& !((IWineD3DPixelShaderImpl
*)pshader
)->baseShader
.load_local_constsF
) {
4181 hardcode_local_constants((IWineD3DBaseShaderImpl
*) pshader
, gl_info
, programId
, 'P');
4183 if(vshader
&& !((IWineD3DVertexShaderImpl
*)vshader
)->baseShader
.load_local_constsF
) {
4184 hardcode_local_constants((IWineD3DBaseShaderImpl
*) vshader
, gl_info
, programId
, 'V');
4188 /* GL locking is done by the caller */
4189 static GLhandleARB
create_glsl_blt_shader(const struct wined3d_gl_info
*gl_info
, enum tex_types tex_type
)
4191 GLhandleARB program_id
;
4192 GLhandleARB vshader_id
, pshader_id
;
4193 static const char *blt_vshader
[] =
4198 " gl_Position = gl_Vertex;\n"
4199 " gl_FrontColor = vec4(1.0);\n"
4200 " gl_TexCoord[0] = gl_MultiTexCoord0;\n"
4204 static const char *blt_pshaders
[tex_type_count
] =
4210 "uniform sampler2D sampler;\n"
4213 " gl_FragDepth = texture2D(sampler, gl_TexCoord[0].xy).x;\n"
4219 "uniform samplerCube sampler;\n"
4222 " gl_FragDepth = textureCube(sampler, gl_TexCoord[0].xyz).x;\n"
4226 "#extension GL_ARB_texture_rectangle : enable\n"
4227 "uniform sampler2DRect sampler;\n"
4230 " gl_FragDepth = texture2DRect(sampler, gl_TexCoord[0].xy).x;\n"
4234 if (!blt_pshaders
[tex_type
])
4236 FIXME("tex_type %#x not supported\n", tex_type
);
4240 vshader_id
= GL_EXTCALL(glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB
));
4241 GL_EXTCALL(glShaderSourceARB(vshader_id
, 1, blt_vshader
, NULL
));
4242 GL_EXTCALL(glCompileShaderARB(vshader_id
));
4244 pshader_id
= GL_EXTCALL(glCreateShaderObjectARB(GL_FRAGMENT_SHADER_ARB
));
4245 GL_EXTCALL(glShaderSourceARB(pshader_id
, 1, &blt_pshaders
[tex_type
], NULL
));
4246 GL_EXTCALL(glCompileShaderARB(pshader_id
));
4248 program_id
= GL_EXTCALL(glCreateProgramObjectARB());
4249 GL_EXTCALL(glAttachObjectARB(program_id
, vshader_id
));
4250 GL_EXTCALL(glAttachObjectARB(program_id
, pshader_id
));
4251 GL_EXTCALL(glLinkProgramARB(program_id
));
4253 print_glsl_info_log(&GLINFO_LOCATION
, program_id
);
4255 /* Once linked we can mark the shaders for deletion. They will be deleted once the program
4258 GL_EXTCALL(glDeleteObjectARB(vshader_id
));
4259 GL_EXTCALL(glDeleteObjectARB(pshader_id
));
4263 /* GL locking is done by the caller */
4264 static void shader_glsl_select(const struct wined3d_context
*context
, BOOL usePS
, BOOL useVS
)
4266 IWineD3DDeviceImpl
*device
= ((IWineD3DSurfaceImpl
*)context
->surface
)->resource
.wineD3DDevice
;
4267 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
4268 struct shader_glsl_priv
*priv
= device
->shader_priv
;
4269 GLhandleARB program_id
= 0;
4270 GLenum old_vertex_color_clamp
, current_vertex_color_clamp
;
4272 old_vertex_color_clamp
= priv
->glsl_program
? priv
->glsl_program
->vertex_color_clamp
: GL_FIXED_ONLY_ARB
;
4274 if (useVS
|| usePS
) set_glsl_shader_program(context
, device
, usePS
, useVS
);
4275 else priv
->glsl_program
= NULL
;
4277 current_vertex_color_clamp
= priv
->glsl_program
? priv
->glsl_program
->vertex_color_clamp
: GL_FIXED_ONLY_ARB
;
4279 if (old_vertex_color_clamp
!= current_vertex_color_clamp
) {
4280 if (GL_SUPPORT(ARB_COLOR_BUFFER_FLOAT
)) {
4281 GL_EXTCALL(glClampColorARB(GL_CLAMP_VERTEX_COLOR_ARB
, current_vertex_color_clamp
));
4282 checkGLcall("glClampColorARB");
4284 FIXME("vertex color clamp needs to be changed, but extension not supported.\n");
4288 program_id
= priv
->glsl_program
? priv
->glsl_program
->programId
: 0;
4289 if (program_id
) TRACE("Using GLSL program %u\n", program_id
);
4290 GL_EXTCALL(glUseProgramObjectARB(program_id
));
4291 checkGLcall("glUseProgramObjectARB");
4293 /* In case that NP2 texcoord fixup data is found for the selected program, trigger a reload of the
4294 * constants. This has to be done because it can't be guaranteed that sampler() (from state.c) is
4295 * called between selecting the shader and using it, which results in wrong fixup for some frames. */
4296 if (priv
->glsl_program
&& priv
->glsl_program
->np2Fixup_info
)
4298 shader_glsl_load_np2fixup_constants((IWineD3DDevice
*)device
, usePS
, useVS
);
4302 /* GL locking is done by the caller */
4303 static void shader_glsl_select_depth_blt(IWineD3DDevice
*iface
, enum tex_types tex_type
) {
4304 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
4305 const struct wined3d_gl_info
*gl_info
= &This
->adapter
->gl_info
;
4306 struct shader_glsl_priv
*priv
= This
->shader_priv
;
4307 GLhandleARB
*blt_program
= &priv
->depth_blt_program
[tex_type
];
4309 if (!*blt_program
) {
4311 *blt_program
= create_glsl_blt_shader(gl_info
, tex_type
);
4312 loc
= GL_EXTCALL(glGetUniformLocationARB(*blt_program
, "sampler"));
4313 GL_EXTCALL(glUseProgramObjectARB(*blt_program
));
4314 GL_EXTCALL(glUniform1iARB(loc
, 0));
4316 GL_EXTCALL(glUseProgramObjectARB(*blt_program
));
4320 /* GL locking is done by the caller */
4321 static void shader_glsl_deselect_depth_blt(IWineD3DDevice
*iface
) {
4322 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
4323 const struct wined3d_gl_info
*gl_info
= &This
->adapter
->gl_info
;
4324 struct shader_glsl_priv
*priv
= This
->shader_priv
;
4325 GLhandleARB program_id
;
4327 program_id
= priv
->glsl_program
? priv
->glsl_program
->programId
: 0;
4328 if (program_id
) TRACE("Using GLSL program %u\n", program_id
);
4330 GL_EXTCALL(glUseProgramObjectARB(program_id
));
4331 checkGLcall("glUseProgramObjectARB");
4334 static void shader_glsl_destroy(IWineD3DBaseShader
*iface
) {
4335 const struct list
*linked_programs
;
4336 IWineD3DBaseShaderImpl
*This
= (IWineD3DBaseShaderImpl
*) iface
;
4337 IWineD3DDeviceImpl
*device
= (IWineD3DDeviceImpl
*)This
->baseShader
.device
;
4338 struct shader_glsl_priv
*priv
= device
->shader_priv
;
4339 const struct wined3d_context
*context
;
4340 const struct wined3d_gl_info
*gl_info
;
4341 IWineD3DPixelShaderImpl
*ps
= NULL
;
4342 IWineD3DVertexShaderImpl
*vs
= NULL
;
4344 /* Note: Do not use QueryInterface here to find out which shader type this is because this code
4345 * can be called from IWineD3DBaseShader::Release
4347 char pshader
= shader_is_pshader_version(This
->baseShader
.reg_maps
.shader_version
.type
);
4350 struct glsl_pshader_private
*shader_data
;
4351 ps
= (IWineD3DPixelShaderImpl
*) This
;
4352 shader_data
= ps
->backend_priv
;
4353 if(!shader_data
|| shader_data
->num_gl_shaders
== 0)
4355 HeapFree(GetProcessHeap(), 0, shader_data
);
4356 ps
->backend_priv
= NULL
;
4360 context
= ActivateContext(device
, NULL
, CTXUSAGE_RESOURCELOAD
);
4361 gl_info
= context
->gl_info
;
4363 if (priv
->glsl_program
&& (IWineD3DBaseShader
*)priv
->glsl_program
->pshader
== iface
)
4366 shader_glsl_select(context
, FALSE
, FALSE
);
4370 struct glsl_vshader_private
*shader_data
;
4371 vs
= (IWineD3DVertexShaderImpl
*) This
;
4372 shader_data
= vs
->backend_priv
;
4373 if(!shader_data
|| shader_data
->num_gl_shaders
== 0)
4375 HeapFree(GetProcessHeap(), 0, shader_data
);
4376 vs
->backend_priv
= NULL
;
4380 context
= ActivateContext(device
, NULL
, CTXUSAGE_RESOURCELOAD
);
4381 gl_info
= context
->gl_info
;
4383 if (priv
->glsl_program
&& (IWineD3DBaseShader
*)priv
->glsl_program
->vshader
== iface
)
4386 shader_glsl_select(context
, FALSE
, FALSE
);
4391 linked_programs
= &This
->baseShader
.linked_programs
;
4393 TRACE("Deleting linked programs\n");
4394 if (linked_programs
->next
) {
4395 struct glsl_shader_prog_link
*entry
, *entry2
;
4399 LIST_FOR_EACH_ENTRY_SAFE(entry
, entry2
, linked_programs
, struct glsl_shader_prog_link
, pshader_entry
) {
4400 delete_glsl_program_entry(priv
, gl_info
, entry
);
4403 LIST_FOR_EACH_ENTRY_SAFE(entry
, entry2
, linked_programs
, struct glsl_shader_prog_link
, vshader_entry
) {
4404 delete_glsl_program_entry(priv
, gl_info
, entry
);
4412 struct glsl_pshader_private
*shader_data
= ps
->backend_priv
;
4415 for(i
= 0; i
< shader_data
->num_gl_shaders
; i
++) {
4416 TRACE("deleting pshader %u\n", shader_data
->gl_shaders
[i
].prgId
);
4417 GL_EXTCALL(glDeleteObjectARB(shader_data
->gl_shaders
[i
].prgId
));
4418 checkGLcall("glDeleteObjectARB");
4421 HeapFree(GetProcessHeap(), 0, shader_data
->gl_shaders
);
4422 HeapFree(GetProcessHeap(), 0, shader_data
);
4423 ps
->backend_priv
= NULL
;
4426 struct glsl_vshader_private
*shader_data
= vs
->backend_priv
;
4429 for(i
= 0; i
< shader_data
->num_gl_shaders
; i
++) {
4430 TRACE("deleting vshader %u\n", shader_data
->gl_shaders
[i
].prgId
);
4431 GL_EXTCALL(glDeleteObjectARB(shader_data
->gl_shaders
[i
].prgId
));
4432 checkGLcall("glDeleteObjectARB");
4435 HeapFree(GetProcessHeap(), 0, shader_data
->gl_shaders
);
4436 HeapFree(GetProcessHeap(), 0, shader_data
);
4437 vs
->backend_priv
= NULL
;
4441 static int glsl_program_key_compare(const void *key
, const struct wine_rb_entry
*entry
)
4443 const glsl_program_key_t
*k
= key
;
4444 const struct glsl_shader_prog_link
*prog
= WINE_RB_ENTRY_VALUE(entry
,
4445 const struct glsl_shader_prog_link
, program_lookup_entry
);
4448 if (k
->vshader
> prog
->vshader
) return 1;
4449 else if (k
->vshader
< prog
->vshader
) return -1;
4451 if (k
->pshader
> prog
->pshader
) return 1;
4452 else if (k
->pshader
< prog
->pshader
) return -1;
4454 if (k
->vshader
&& (cmp
= memcmp(&k
->vs_args
, &prog
->vs_args
, sizeof(prog
->vs_args
)))) return cmp
;
4455 if (k
->pshader
&& (cmp
= memcmp(&k
->ps_args
, &prog
->ps_args
, sizeof(prog
->ps_args
)))) return cmp
;
4460 static BOOL
constant_heap_init(struct constant_heap
*heap
, unsigned int constant_count
)
4462 SIZE_T size
= (constant_count
+ 1) * sizeof(*heap
->entries
) + constant_count
* sizeof(*heap
->positions
);
4463 void *mem
= HeapAlloc(GetProcessHeap(), 0, size
);
4467 ERR("Failed to allocate memory\n");
4471 heap
->entries
= mem
;
4472 heap
->entries
[1].version
= 0;
4473 heap
->positions
= (unsigned int *)(heap
->entries
+ constant_count
+ 1);
4479 static void constant_heap_free(struct constant_heap
*heap
)
4481 HeapFree(GetProcessHeap(), 0, heap
->entries
);
4484 static const struct wine_rb_functions wined3d_glsl_program_rb_functions
=
4489 glsl_program_key_compare
,
4492 static HRESULT
shader_glsl_alloc(IWineD3DDevice
*iface
) {
4493 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
4494 const struct wined3d_gl_info
*gl_info
= &This
->adapter
->gl_info
;
4495 struct shader_glsl_priv
*priv
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(struct shader_glsl_priv
));
4496 SIZE_T stack_size
= wined3d_log2i(max(GL_LIMITS(vshader_constantsF
), GL_LIMITS(pshader_constantsF
))) + 1;
4498 if (!shader_buffer_init(&priv
->shader_buffer
))
4500 ERR("Failed to initialize shader buffer.\n");
4504 priv
->stack
= HeapAlloc(GetProcessHeap(), 0, stack_size
* sizeof(*priv
->stack
));
4507 ERR("Failed to allocate memory.\n");
4511 if (!constant_heap_init(&priv
->vconst_heap
, GL_LIMITS(vshader_constantsF
)))
4513 ERR("Failed to initialize vertex shader constant heap\n");
4517 if (!constant_heap_init(&priv
->pconst_heap
, GL_LIMITS(pshader_constantsF
)))
4519 ERR("Failed to initialize pixel shader constant heap\n");
4523 if (wine_rb_init(&priv
->program_lookup
, &wined3d_glsl_program_rb_functions
) == -1)
4525 ERR("Failed to initialize rbtree.\n");
4529 priv
->next_constant_version
= 1;
4531 This
->shader_priv
= priv
;
4535 constant_heap_free(&priv
->pconst_heap
);
4536 constant_heap_free(&priv
->vconst_heap
);
4537 HeapFree(GetProcessHeap(), 0, priv
->stack
);
4538 shader_buffer_free(&priv
->shader_buffer
);
4539 HeapFree(GetProcessHeap(), 0, priv
);
4540 return E_OUTOFMEMORY
;
4543 /* Context activation is done by the caller. */
4544 static void shader_glsl_free(IWineD3DDevice
*iface
) {
4545 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
4546 const struct wined3d_gl_info
*gl_info
= &This
->adapter
->gl_info
;
4547 struct shader_glsl_priv
*priv
= This
->shader_priv
;
4551 for (i
= 0; i
< tex_type_count
; ++i
)
4553 if (priv
->depth_blt_program
[i
])
4555 GL_EXTCALL(glDeleteObjectARB(priv
->depth_blt_program
[i
]));
4560 wine_rb_destroy(&priv
->program_lookup
, NULL
, NULL
);
4561 constant_heap_free(&priv
->pconst_heap
);
4562 constant_heap_free(&priv
->vconst_heap
);
4563 HeapFree(GetProcessHeap(), 0, priv
->stack
);
4565 HeapFree(GetProcessHeap(), 0, This
->shader_priv
);
4566 This
->shader_priv
= NULL
;
4569 static BOOL
shader_glsl_dirty_const(IWineD3DDevice
*iface
) {
4570 /* TODO: GL_EXT_bindable_uniform can be used to share constants across shaders */
4574 static void shader_glsl_get_caps(WINED3DDEVTYPE devtype
,
4575 const struct wined3d_gl_info
*gl_info
, struct shader_caps
*pCaps
)
4577 /* Nvidia Geforce6/7 or Ati R4xx/R5xx cards with GLSL support, support VS 3.0 but older Nvidia/Ati
4578 * models with GLSL support only support 2.0. In case of nvidia we can detect VS 2.0 support based
4579 * on the version of NV_vertex_program.
4580 * For Ati cards there's no way using glsl (it abstracts the lowlevel info away) and also not
4581 * using ARB_vertex_program. It is safe to assume that when a card supports pixel shader 2.0 it
4582 * supports vertex shader 2.0 too and the way around. We can detect ps2.0 using the maximum number
4583 * of native instructions, so use that here. For more info see the pixel shader versioning code below.
4585 if ((gl_info
->supported
[NV_VERTEX_PROGRAM2
] && !gl_info
->supported
[NV_VERTEX_PROGRAM3
])
4586 || gl_info
->ps_arb_max_instructions
<= 512)
4587 pCaps
->VertexShaderVersion
= WINED3DVS_VERSION(2,0);
4589 pCaps
->VertexShaderVersion
= WINED3DVS_VERSION(3,0);
4590 TRACE_(d3d_caps
)("Hardware vertex shader version %d.%d enabled (GLSL)\n", (pCaps
->VertexShaderVersion
>> 8) & 0xff, pCaps
->VertexShaderVersion
& 0xff);
4591 pCaps
->MaxVertexShaderConst
= GL_LIMITS(vshader_constantsF
);
4593 /* Older DX9-class videocards (GeforceFX / Radeon >9500/X*00) only support pixel shader 2.0/2.0a/2.0b.
4594 * In OpenGL the extensions related to GLSL abstract lowlevel GL info away which is needed
4595 * to distinguish between 2.0 and 3.0 (and 2.0a/2.0b). In case of Nvidia we use their fragment
4596 * program extensions. On other hardware including ATI GL_ARB_fragment_program offers the info
4597 * in max native instructions. Intel and others also offer the info in this extension but they
4598 * don't support GLSL (at least on Windows).
4600 * PS2.0 requires at least 96 instructions, 2.0a/2.0b go up to 512. Assume that if the number
4601 * of instructions is 512 or less we have to do with ps2.0 hardware.
4602 * 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.
4604 if ((gl_info
->supported
[NV_FRAGMENT_PROGRAM
] && !gl_info
->supported
[NV_FRAGMENT_PROGRAM2
])
4605 || (gl_info
->ps_arb_max_instructions
<= 512))
4606 pCaps
->PixelShaderVersion
= WINED3DPS_VERSION(2,0);
4608 pCaps
->PixelShaderVersion
= WINED3DPS_VERSION(3,0);
4610 pCaps
->MaxPixelShaderConst
= GL_LIMITS(pshader_constantsF
);
4612 /* FIXME: The following line is card dependent. -8.0 to 8.0 is the
4613 * Direct3D minimum requirement.
4615 * Both GL_ARB_fragment_program and GLSL require a "maximum representable magnitude"
4616 * of colors to be 2^10, and 2^32 for other floats. Should we use 1024 here?
4618 * The problem is that the refrast clamps temporary results in the shader to
4619 * [-MaxValue;+MaxValue]. If the card's max value is bigger than the one we advertize here,
4620 * then applications may miss the clamping behavior. On the other hand, if it is smaller,
4621 * the shader will generate incorrect results too. Unfortunately, GL deliberately doesn't
4622 * offer a way to query this.
4624 pCaps
->PixelShader1xMaxValue
= 8.0;
4625 TRACE_(d3d_caps
)("Hardware pixel shader version %d.%d enabled (GLSL)\n", (pCaps
->PixelShaderVersion
>> 8) & 0xff, pCaps
->PixelShaderVersion
& 0xff);
4627 pCaps
->VSClipping
= TRUE
;
4630 static BOOL
shader_glsl_color_fixup_supported(struct color_fixup_desc fixup
)
4632 if (TRACE_ON(d3d_shader
) && TRACE_ON(d3d
))
4634 TRACE("Checking support for fixup:\n");
4635 dump_color_fixup_desc(fixup
);
4638 /* We support everything except YUV conversions. */
4639 if (!is_yuv_fixup(fixup
))
4645 TRACE("[FAILED]\n");
4649 static const SHADER_HANDLER shader_glsl_instruction_handler_table
[WINED3DSIH_TABLE_SIZE
] =
4651 /* WINED3DSIH_ABS */ shader_glsl_map2gl
,
4652 /* WINED3DSIH_ADD */ shader_glsl_arith
,
4653 /* WINED3DSIH_BEM */ shader_glsl_bem
,
4654 /* WINED3DSIH_BREAK */ shader_glsl_break
,
4655 /* WINED3DSIH_BREAKC */ shader_glsl_breakc
,
4656 /* WINED3DSIH_BREAKP */ NULL
,
4657 /* WINED3DSIH_CALL */ shader_glsl_call
,
4658 /* WINED3DSIH_CALLNZ */ shader_glsl_callnz
,
4659 /* WINED3DSIH_CMP */ shader_glsl_cmp
,
4660 /* WINED3DSIH_CND */ shader_glsl_cnd
,
4661 /* WINED3DSIH_CRS */ shader_glsl_cross
,
4662 /* WINED3DSIH_DCL */ NULL
,
4663 /* WINED3DSIH_DEF */ NULL
,
4664 /* WINED3DSIH_DEFB */ NULL
,
4665 /* WINED3DSIH_DEFI */ NULL
,
4666 /* WINED3DSIH_DP2ADD */ shader_glsl_dp2add
,
4667 /* WINED3DSIH_DP3 */ shader_glsl_dot
,
4668 /* WINED3DSIH_DP4 */ shader_glsl_dot
,
4669 /* WINED3DSIH_DST */ shader_glsl_dst
,
4670 /* WINED3DSIH_DSX */ shader_glsl_map2gl
,
4671 /* WINED3DSIH_DSY */ shader_glsl_map2gl
,
4672 /* WINED3DSIH_ELSE */ shader_glsl_else
,
4673 /* WINED3DSIH_ENDIF */ shader_glsl_end
,
4674 /* WINED3DSIH_ENDLOOP */ shader_glsl_end
,
4675 /* WINED3DSIH_ENDREP */ shader_glsl_end
,
4676 /* WINED3DSIH_EXP */ shader_glsl_map2gl
,
4677 /* WINED3DSIH_EXPP */ shader_glsl_expp
,
4678 /* WINED3DSIH_FRC */ shader_glsl_map2gl
,
4679 /* WINED3DSIH_IF */ shader_glsl_if
,
4680 /* WINED3DSIH_IFC */ shader_glsl_ifc
,
4681 /* WINED3DSIH_LABEL */ shader_glsl_label
,
4682 /* WINED3DSIH_LIT */ shader_glsl_lit
,
4683 /* WINED3DSIH_LOG */ shader_glsl_log
,
4684 /* WINED3DSIH_LOGP */ shader_glsl_log
,
4685 /* WINED3DSIH_LOOP */ shader_glsl_loop
,
4686 /* WINED3DSIH_LRP */ shader_glsl_lrp
,
4687 /* WINED3DSIH_M3x2 */ shader_glsl_mnxn
,
4688 /* WINED3DSIH_M3x3 */ shader_glsl_mnxn
,
4689 /* WINED3DSIH_M3x4 */ shader_glsl_mnxn
,
4690 /* WINED3DSIH_M4x3 */ shader_glsl_mnxn
,
4691 /* WINED3DSIH_M4x4 */ shader_glsl_mnxn
,
4692 /* WINED3DSIH_MAD */ shader_glsl_mad
,
4693 /* WINED3DSIH_MAX */ shader_glsl_map2gl
,
4694 /* WINED3DSIH_MIN */ shader_glsl_map2gl
,
4695 /* WINED3DSIH_MOV */ shader_glsl_mov
,
4696 /* WINED3DSIH_MOVA */ shader_glsl_mov
,
4697 /* WINED3DSIH_MUL */ shader_glsl_arith
,
4698 /* WINED3DSIH_NOP */ NULL
,
4699 /* WINED3DSIH_NRM */ shader_glsl_map2gl
,
4700 /* WINED3DSIH_PHASE */ NULL
,
4701 /* WINED3DSIH_POW */ shader_glsl_pow
,
4702 /* WINED3DSIH_RCP */ shader_glsl_rcp
,
4703 /* WINED3DSIH_REP */ shader_glsl_rep
,
4704 /* WINED3DSIH_RET */ shader_glsl_ret
,
4705 /* WINED3DSIH_RSQ */ shader_glsl_rsq
,
4706 /* WINED3DSIH_SETP */ NULL
,
4707 /* WINED3DSIH_SGE */ shader_glsl_compare
,
4708 /* WINED3DSIH_SGN */ shader_glsl_map2gl
,
4709 /* WINED3DSIH_SINCOS */ shader_glsl_sincos
,
4710 /* WINED3DSIH_SLT */ shader_glsl_compare
,
4711 /* WINED3DSIH_SUB */ shader_glsl_arith
,
4712 /* WINED3DSIH_TEX */ shader_glsl_tex
,
4713 /* WINED3DSIH_TEXBEM */ shader_glsl_texbem
,
4714 /* WINED3DSIH_TEXBEML */ shader_glsl_texbem
,
4715 /* WINED3DSIH_TEXCOORD */ shader_glsl_texcoord
,
4716 /* WINED3DSIH_TEXDEPTH */ shader_glsl_texdepth
,
4717 /* WINED3DSIH_TEXDP3 */ shader_glsl_texdp3
,
4718 /* WINED3DSIH_TEXDP3TEX */ shader_glsl_texdp3tex
,
4719 /* WINED3DSIH_TEXKILL */ shader_glsl_texkill
,
4720 /* WINED3DSIH_TEXLDD */ shader_glsl_texldd
,
4721 /* WINED3DSIH_TEXLDL */ shader_glsl_texldl
,
4722 /* WINED3DSIH_TEXM3x2DEPTH */ shader_glsl_texm3x2depth
,
4723 /* WINED3DSIH_TEXM3x2PAD */ shader_glsl_texm3x2pad
,
4724 /* WINED3DSIH_TEXM3x2TEX */ shader_glsl_texm3x2tex
,
4725 /* WINED3DSIH_TEXM3x3 */ shader_glsl_texm3x3
,
4726 /* WINED3DSIH_TEXM3x3DIFF */ NULL
,
4727 /* WINED3DSIH_TEXM3x3PAD */ shader_glsl_texm3x3pad
,
4728 /* WINED3DSIH_TEXM3x3SPEC */ shader_glsl_texm3x3spec
,
4729 /* WINED3DSIH_TEXM3x3TEX */ shader_glsl_texm3x3tex
,
4730 /* WINED3DSIH_TEXM3x3VSPEC */ shader_glsl_texm3x3vspec
,
4731 /* WINED3DSIH_TEXREG2AR */ shader_glsl_texreg2ar
,
4732 /* WINED3DSIH_TEXREG2GB */ shader_glsl_texreg2gb
,
4733 /* WINED3DSIH_TEXREG2RGB */ shader_glsl_texreg2rgb
,
4736 static void shader_glsl_handle_instruction(const struct wined3d_shader_instruction
*ins
) {
4737 SHADER_HANDLER hw_fct
;
4739 /* Select handler */
4740 hw_fct
= shader_glsl_instruction_handler_table
[ins
->handler_idx
];
4742 /* Unhandled opcode */
4745 FIXME("Backend can't handle opcode %#x\n", ins
->handler_idx
);
4750 shader_glsl_add_instruction_modifiers(ins
);
4753 const shader_backend_t glsl_shader_backend
= {
4754 shader_glsl_handle_instruction
,
4756 shader_glsl_select_depth_blt
,
4757 shader_glsl_deselect_depth_blt
,
4758 shader_glsl_update_float_vertex_constants
,
4759 shader_glsl_update_float_pixel_constants
,
4760 shader_glsl_load_constants
,
4761 shader_glsl_load_np2fixup_constants
,
4762 shader_glsl_destroy
,
4765 shader_glsl_dirty_const
,
4766 shader_glsl_get_caps
,
4767 shader_glsl_color_fixup_supported
,