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 2009-2011 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.
33 #include "wine/port.h"
38 #include "wined3d_private.h"
40 WINE_DEFAULT_DEBUG_CHANNEL(d3d_shader
);
41 WINE_DECLARE_DEBUG_CHANNEL(d3d_constants
);
42 WINE_DECLARE_DEBUG_CHANNEL(d3d
);
43 WINE_DECLARE_DEBUG_CHANNEL(winediag
);
45 #define WINED3D_GLSL_SAMPLE_PROJECTED 0x1
46 #define WINED3D_GLSL_SAMPLE_RECT 0x2
47 #define WINED3D_GLSL_SAMPLE_LOD 0x4
48 #define WINED3D_GLSL_SAMPLE_GRAD 0x8
62 struct glsl_sample_function
70 HEAP_NODE_TRAVERSE_LEFT
,
71 HEAP_NODE_TRAVERSE_RIGHT
,
83 struct constant_entry
*entries
;
84 unsigned int *positions
;
88 /* GLSL shader private data */
89 struct shader_glsl_priv
{
90 struct wined3d_shader_buffer shader_buffer
;
91 struct wine_rb_tree program_lookup
;
92 struct glsl_shader_prog_link
*glsl_program
;
93 struct constant_heap vconst_heap
;
94 struct constant_heap pconst_heap
;
96 GLhandleARB depth_blt_program_full
[tex_type_count
];
97 GLhandleARB depth_blt_program_masked
[tex_type_count
];
98 UINT next_constant_version
;
100 const struct fragment_pipeline
*fragment_pipe
;
101 struct wine_rb_tree ffp_fragment_shaders
;
102 BOOL ffp_proj_control
;
105 struct glsl_vs_program
107 struct list shader_entry
;
109 GLenum vertex_color_clamp
;
110 GLint
*uniform_f_locations
;
111 GLint uniform_i_locations
[MAX_CONST_I
];
112 GLint pos_fixup_location
;
115 struct glsl_gs_program
117 struct list shader_entry
;
121 struct glsl_ps_program
123 struct list shader_entry
;
125 GLint
*uniform_f_locations
;
126 GLint uniform_i_locations
[MAX_CONST_I
];
127 GLint bumpenv_mat_location
[MAX_TEXTURES
];
128 GLint bumpenv_lum_scale_location
[MAX_TEXTURES
];
129 GLint bumpenv_lum_offset_location
[MAX_TEXTURES
];
130 GLint tex_factor_location
;
131 GLint specular_enable_location
;
132 GLint ycorrection_location
;
133 GLint np2_fixup_location
;
134 const struct ps_np2fixup_info
*np2_fixup_info
;
137 /* Struct to maintain data about a linked GLSL program */
138 struct glsl_shader_prog_link
140 struct wine_rb_entry program_lookup_entry
;
141 struct glsl_vs_program vs
;
142 struct glsl_gs_program gs
;
143 struct glsl_ps_program ps
;
144 GLhandleARB programId
;
145 UINT constant_version
;
148 struct glsl_program_key
155 struct shader_glsl_ctx_priv
{
156 const struct vs_compile_args
*cur_vs_args
;
157 const struct ps_compile_args
*cur_ps_args
;
158 struct ps_np2fixup_info
*cur_np2fixup_info
;
161 struct glsl_ps_compiled_shader
163 struct ps_compile_args args
;
164 struct ps_np2fixup_info np2fixup
;
168 struct glsl_vs_compiled_shader
170 struct vs_compile_args args
;
174 struct glsl_gs_compiled_shader
179 struct glsl_shader_private
183 struct glsl_vs_compiled_shader
*vs
;
184 struct glsl_gs_compiled_shader
*gs
;
185 struct glsl_ps_compiled_shader
*ps
;
187 UINT num_gl_shaders
, shader_array_size
;
190 struct glsl_ffp_fragment_shader
192 struct ffp_frag_desc entry
;
194 struct list linked_programs
;
197 static const char *debug_gl_shader_type(GLenum type
)
201 #define WINED3D_TO_STR(u) case u: return #u
202 WINED3D_TO_STR(GL_VERTEX_SHADER_ARB
);
203 WINED3D_TO_STR(GL_GEOMETRY_SHADER_ARB
);
204 WINED3D_TO_STR(GL_FRAGMENT_SHADER_ARB
);
205 #undef WINED3D_TO_STR
207 return wine_dbg_sprintf("UNKNOWN(%#x)", type
);
211 static const char *shader_glsl_get_prefix(enum wined3d_shader_type type
)
215 case WINED3D_SHADER_TYPE_VERTEX
:
218 case WINED3D_SHADER_TYPE_GEOMETRY
:
221 case WINED3D_SHADER_TYPE_PIXEL
:
225 FIXME("Unhandled shader type %#x.\n", type
);
230 /* Extract a line from the info log.
231 * Note that this modifies the source string. */
232 static char *get_info_log_line(char **ptr
)
237 if (!(q
= strstr(p
, "\n")))
239 if (!*p
) return NULL
;
249 /* Context activation is done by the caller. */
250 static void print_glsl_info_log(const struct wined3d_gl_info
*gl_info
, GLhandleARB obj
)
252 int infologLength
= 0;
255 if (!WARN_ON(d3d_shader
) && !FIXME_ON(d3d_shader
))
258 GL_EXTCALL(glGetObjectParameterivARB(obj
,
259 GL_OBJECT_INFO_LOG_LENGTH_ARB
,
262 /* A size of 1 is just a null-terminated string, so the log should be bigger than
263 * that if there are errors. */
264 if (infologLength
> 1)
268 infoLog
= HeapAlloc(GetProcessHeap(), 0, infologLength
);
269 /* The info log is supposed to be zero-terminated, but at least some
270 * versions of fglrx don't terminate the string properly. The reported
271 * length does include the terminator, so explicitly set it to zero
273 infoLog
[infologLength
- 1] = 0;
274 GL_EXTCALL(glGetInfoLogARB(obj
, infologLength
, NULL
, infoLog
));
277 if (gl_info
->quirks
& WINED3D_QUIRK_INFO_LOG_SPAM
)
279 WARN("Info log received from GLSL shader #%u:\n", obj
);
280 while ((line
= get_info_log_line(&ptr
))) WARN(" %s\n", line
);
284 FIXME("Info log received from GLSL shader #%u:\n", obj
);
285 while ((line
= get_info_log_line(&ptr
))) FIXME(" %s\n", line
);
287 HeapFree(GetProcessHeap(), 0, infoLog
);
291 /* Context activation is done by the caller. */
292 static void shader_glsl_compile(const struct wined3d_gl_info
*gl_info
, GLhandleARB shader
, const char *src
)
294 TRACE("Compiling shader object %u.\n", shader
);
295 GL_EXTCALL(glShaderSourceARB(shader
, 1, &src
, NULL
));
296 checkGLcall("glShaderSourceARB");
297 GL_EXTCALL(glCompileShaderARB(shader
));
298 checkGLcall("glCompileShaderARB");
299 print_glsl_info_log(gl_info
, shader
);
302 /* Context activation is done by the caller. */
303 static void shader_glsl_dump_program_source(const struct wined3d_gl_info
*gl_info
, GLhandleARB program
)
305 GLint i
, object_count
, source_size
= -1;
306 GLhandleARB
*objects
;
309 GL_EXTCALL(glGetObjectParameterivARB(program
, GL_OBJECT_ATTACHED_OBJECTS_ARB
, &object_count
));
310 objects
= HeapAlloc(GetProcessHeap(), 0, object_count
* sizeof(*objects
));
313 ERR("Failed to allocate object array memory.\n");
317 GL_EXTCALL(glGetAttachedObjectsARB(program
, object_count
, NULL
, objects
));
318 for (i
= 0; i
< object_count
; ++i
)
323 GL_EXTCALL(glGetObjectParameterivARB(objects
[i
], GL_OBJECT_SHADER_SOURCE_LENGTH_ARB
, &tmp
));
325 if (source_size
< tmp
)
327 HeapFree(GetProcessHeap(), 0, source
);
329 source
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, tmp
);
332 ERR("Failed to allocate %d bytes for shader source.\n", tmp
);
333 HeapFree(GetProcessHeap(), 0, objects
);
339 FIXME("Object %u:\n", objects
[i
]);
340 GL_EXTCALL(glGetObjectParameterivARB(objects
[i
], GL_OBJECT_SUBTYPE_ARB
, &tmp
));
341 FIXME(" GL_OBJECT_SUBTYPE_ARB: %s.\n", debug_gl_shader_type(tmp
));
342 GL_EXTCALL(glGetObjectParameterivARB(objects
[i
], GL_OBJECT_COMPILE_STATUS_ARB
, &tmp
));
343 FIXME(" GL_OBJECT_COMPILE_STATUS_ARB: %d.\n", tmp
);
347 GL_EXTCALL(glGetShaderSourceARB(objects
[i
], source_size
, NULL
, source
));
348 while ((line
= get_info_log_line(&ptr
))) FIXME(" %s\n", line
);
352 HeapFree(GetProcessHeap(), 0, source
);
353 HeapFree(GetProcessHeap(), 0, objects
);
356 /* Context activation is done by the caller. */
357 static void shader_glsl_validate_link(const struct wined3d_gl_info
*gl_info
, GLhandleARB program
)
361 if (!TRACE_ON(d3d_shader
) && !FIXME_ON(d3d_shader
)) return;
363 GL_EXTCALL(glGetObjectParameterivARB(program
, GL_OBJECT_TYPE_ARB
, &tmp
));
364 if (tmp
== GL_PROGRAM_OBJECT_ARB
)
366 GL_EXTCALL(glGetObjectParameterivARB(program
, GL_OBJECT_LINK_STATUS_ARB
, &tmp
));
369 FIXME("Program %u link status invalid.\n", program
);
370 shader_glsl_dump_program_source(gl_info
, program
);
374 print_glsl_info_log(gl_info
, program
);
377 /* Context activation is done by the caller. */
378 static void shader_glsl_load_psamplers(const struct wined3d_gl_info
*gl_info
,
379 const DWORD
*tex_unit_map
, GLhandleARB programId
)
382 char sampler_name
[20];
385 for (i
= 0; i
< MAX_FRAGMENT_SAMPLERS
; ++i
)
387 snprintf(sampler_name
, sizeof(sampler_name
), "ps_sampler%u", i
);
388 name_loc
= GL_EXTCALL(glGetUniformLocationARB(programId
, sampler_name
));
389 if (name_loc
!= -1) {
390 DWORD mapped_unit
= tex_unit_map
[i
];
391 if (mapped_unit
!= WINED3D_UNMAPPED_STAGE
&& mapped_unit
< gl_info
->limits
.fragment_samplers
)
393 TRACE("Loading %s for texture %d\n", sampler_name
, mapped_unit
);
394 GL_EXTCALL(glUniform1iARB(name_loc
, mapped_unit
));
395 checkGLcall("glUniform1iARB");
397 ERR("Trying to load sampler %s on unsupported unit %d\n", sampler_name
, mapped_unit
);
403 /* Context activation is done by the caller. */
404 static void shader_glsl_load_vsamplers(const struct wined3d_gl_info
*gl_info
,
405 const DWORD
*tex_unit_map
, GLhandleARB programId
)
408 char sampler_name
[20];
411 for (i
= 0; i
< MAX_VERTEX_SAMPLERS
; ++i
)
413 snprintf(sampler_name
, sizeof(sampler_name
), "vs_sampler%u", i
);
414 name_loc
= GL_EXTCALL(glGetUniformLocationARB(programId
, sampler_name
));
415 if (name_loc
!= -1) {
416 DWORD mapped_unit
= tex_unit_map
[MAX_FRAGMENT_SAMPLERS
+ i
];
417 if (mapped_unit
!= WINED3D_UNMAPPED_STAGE
&& mapped_unit
< gl_info
->limits
.combined_samplers
)
419 TRACE("Loading %s for texture %d\n", sampler_name
, mapped_unit
);
420 GL_EXTCALL(glUniform1iARB(name_loc
, mapped_unit
));
421 checkGLcall("glUniform1iARB");
423 ERR("Trying to load sampler %s on unsupported unit %d\n", sampler_name
, mapped_unit
);
429 /* Context activation is done by the caller. */
430 static inline void walk_constant_heap(const struct wined3d_gl_info
*gl_info
, const float *constants
,
431 const GLint
*constant_locations
, const struct constant_heap
*heap
, unsigned char *stack
, DWORD version
)
434 unsigned int heap_idx
= 1;
437 if (heap
->entries
[heap_idx
].version
<= version
) return;
439 idx
= heap
->entries
[heap_idx
].idx
;
440 if (constant_locations
[idx
] != -1) GL_EXTCALL(glUniform4fvARB(constant_locations
[idx
], 1, &constants
[idx
* 4]));
441 stack
[stack_idx
] = HEAP_NODE_TRAVERSE_LEFT
;
443 while (stack_idx
>= 0)
445 /* Note that we fall through to the next case statement. */
446 switch(stack
[stack_idx
])
448 case HEAP_NODE_TRAVERSE_LEFT
:
450 unsigned int left_idx
= heap_idx
<< 1;
451 if (left_idx
< heap
->size
&& heap
->entries
[left_idx
].version
> version
)
454 idx
= heap
->entries
[heap_idx
].idx
;
455 if (constant_locations
[idx
] != -1)
456 GL_EXTCALL(glUniform4fvARB(constant_locations
[idx
], 1, &constants
[idx
* 4]));
458 stack
[stack_idx
++] = HEAP_NODE_TRAVERSE_RIGHT
;
459 stack
[stack_idx
] = HEAP_NODE_TRAVERSE_LEFT
;
464 case HEAP_NODE_TRAVERSE_RIGHT
:
466 unsigned int right_idx
= (heap_idx
<< 1) + 1;
467 if (right_idx
< heap
->size
&& heap
->entries
[right_idx
].version
> version
)
469 heap_idx
= right_idx
;
470 idx
= heap
->entries
[heap_idx
].idx
;
471 if (constant_locations
[idx
] != -1)
472 GL_EXTCALL(glUniform4fvARB(constant_locations
[idx
], 1, &constants
[idx
* 4]));
474 stack
[stack_idx
++] = HEAP_NODE_POP
;
475 stack
[stack_idx
] = HEAP_NODE_TRAVERSE_LEFT
;
486 checkGLcall("walk_constant_heap()");
489 /* Context activation is done by the caller. */
490 static inline void apply_clamped_constant(const struct wined3d_gl_info
*gl_info
, GLint location
, const GLfloat
*data
)
492 GLfloat clamped_constant
[4];
494 if (location
== -1) return;
496 clamped_constant
[0] = data
[0] < -1.0f
? -1.0f
: data
[0] > 1.0f
? 1.0f
: data
[0];
497 clamped_constant
[1] = data
[1] < -1.0f
? -1.0f
: data
[1] > 1.0f
? 1.0f
: data
[1];
498 clamped_constant
[2] = data
[2] < -1.0f
? -1.0f
: data
[2] > 1.0f
? 1.0f
: data
[2];
499 clamped_constant
[3] = data
[3] < -1.0f
? -1.0f
: data
[3] > 1.0f
? 1.0f
: data
[3];
501 GL_EXTCALL(glUniform4fvARB(location
, 1, clamped_constant
));
504 /* Context activation is done by the caller. */
505 static inline void walk_constant_heap_clamped(const struct wined3d_gl_info
*gl_info
, const float *constants
,
506 const GLint
*constant_locations
, const struct constant_heap
*heap
, unsigned char *stack
, DWORD version
)
509 unsigned int heap_idx
= 1;
512 if (heap
->entries
[heap_idx
].version
<= version
) return;
514 idx
= heap
->entries
[heap_idx
].idx
;
515 apply_clamped_constant(gl_info
, constant_locations
[idx
], &constants
[idx
* 4]);
516 stack
[stack_idx
] = HEAP_NODE_TRAVERSE_LEFT
;
518 while (stack_idx
>= 0)
520 /* Note that we fall through to the next case statement. */
521 switch(stack
[stack_idx
])
523 case HEAP_NODE_TRAVERSE_LEFT
:
525 unsigned int left_idx
= heap_idx
<< 1;
526 if (left_idx
< heap
->size
&& heap
->entries
[left_idx
].version
> version
)
529 idx
= heap
->entries
[heap_idx
].idx
;
530 apply_clamped_constant(gl_info
, constant_locations
[idx
], &constants
[idx
* 4]);
532 stack
[stack_idx
++] = HEAP_NODE_TRAVERSE_RIGHT
;
533 stack
[stack_idx
] = HEAP_NODE_TRAVERSE_LEFT
;
538 case HEAP_NODE_TRAVERSE_RIGHT
:
540 unsigned int right_idx
= (heap_idx
<< 1) + 1;
541 if (right_idx
< heap
->size
&& heap
->entries
[right_idx
].version
> version
)
543 heap_idx
= right_idx
;
544 idx
= heap
->entries
[heap_idx
].idx
;
545 apply_clamped_constant(gl_info
, constant_locations
[idx
], &constants
[idx
* 4]);
547 stack
[stack_idx
++] = HEAP_NODE_POP
;
548 stack
[stack_idx
] = HEAP_NODE_TRAVERSE_LEFT
;
559 checkGLcall("walk_constant_heap_clamped()");
562 /* Context activation is done by the caller. */
563 static void shader_glsl_load_constantsF(const struct wined3d_shader
*shader
, const struct wined3d_gl_info
*gl_info
,
564 const float *constants
, const GLint
*constant_locations
, const struct constant_heap
*heap
,
565 unsigned char *stack
, UINT version
)
567 const struct wined3d_shader_lconst
*lconst
;
569 /* 1.X pshaders have the constants clamped to [-1;1] implicitly. */
570 if (shader
->reg_maps
.shader_version
.major
== 1
571 && shader
->reg_maps
.shader_version
.type
== WINED3D_SHADER_TYPE_PIXEL
)
572 walk_constant_heap_clamped(gl_info
, constants
, constant_locations
, heap
, stack
, version
);
574 walk_constant_heap(gl_info
, constants
, constant_locations
, heap
, stack
, version
);
576 if (!shader
->load_local_constsF
)
578 TRACE("No need to load local float constants for this shader\n");
582 /* Immediate constants are clamped to [-1;1] at shader creation time if needed */
583 LIST_FOR_EACH_ENTRY(lconst
, &shader
->constantsF
, struct wined3d_shader_lconst
, entry
)
585 GLint location
= constant_locations
[lconst
->idx
];
586 /* We found this uniform name in the program - go ahead and send the data */
587 if (location
!= -1) GL_EXTCALL(glUniform4fvARB(location
, 1, (const GLfloat
*)lconst
->value
));
589 checkGLcall("glUniform4fvARB()");
592 /* Context activation is done by the caller. */
593 static void shader_glsl_load_constantsI(const struct wined3d_shader
*shader
, const struct wined3d_gl_info
*gl_info
,
594 const GLint locations
[MAX_CONST_I
], const int *constants
, WORD constants_set
)
599 for (i
= 0; constants_set
; constants_set
>>= 1, ++i
)
601 if (!(constants_set
& 1)) continue;
603 TRACE_(d3d_constants
)("Loading constants %u: %i, %i, %i, %i\n",
604 i
, constants
[i
*4], constants
[i
*4+1], constants
[i
*4+2], constants
[i
*4+3]);
606 /* We found this uniform name in the program - go ahead and send the data */
607 GL_EXTCALL(glUniform4ivARB(locations
[i
], 1, &constants
[i
*4]));
608 checkGLcall("glUniform4ivARB");
611 /* Load immediate constants */
612 ptr
= list_head(&shader
->constantsI
);
615 const struct wined3d_shader_lconst
*lconst
= LIST_ENTRY(ptr
, const struct wined3d_shader_lconst
, entry
);
616 unsigned int idx
= lconst
->idx
;
617 const GLint
*values
= (const GLint
*)lconst
->value
;
619 TRACE_(d3d_constants
)("Loading local constants %i: %i, %i, %i, %i\n", idx
,
620 values
[0], values
[1], values
[2], values
[3]);
622 /* We found this uniform name in the program - go ahead and send the data */
623 GL_EXTCALL(glUniform4ivARB(locations
[idx
], 1, values
));
624 checkGLcall("glUniform4ivARB");
625 ptr
= list_next(&shader
->constantsI
, ptr
);
629 /* Context activation is done by the caller. */
630 static void shader_glsl_load_constantsB(const struct wined3d_shader
*shader
, const struct wined3d_gl_info
*gl_info
,
631 GLhandleARB programId
, const BOOL
*constants
, WORD constants_set
)
639 prefix
= shader_glsl_get_prefix(shader
->reg_maps
.shader_version
.type
);
641 /* TODO: Benchmark and see if it would be beneficial to store the
642 * locations of the constants to avoid looking up each time */
643 for (i
= 0; constants_set
; constants_set
>>= 1, ++i
)
645 if (!(constants_set
& 1)) continue;
647 TRACE_(d3d_constants
)("Loading constants %i: %i;\n", i
, constants
[i
]);
649 /* TODO: Benchmark and see if it would be beneficial to store the
650 * locations of the constants to avoid looking up each time */
651 snprintf(tmp_name
, sizeof(tmp_name
), "%s_b[%i]", prefix
, i
);
652 tmp_loc
= GL_EXTCALL(glGetUniformLocationARB(programId
, tmp_name
));
653 GL_EXTCALL(glUniform1ivARB(tmp_loc
, 1, &constants
[i
]));
654 checkGLcall("glUniform1ivARB");
657 /* Load immediate constants */
658 ptr
= list_head(&shader
->constantsB
);
661 const struct wined3d_shader_lconst
*lconst
= LIST_ENTRY(ptr
, const struct wined3d_shader_lconst
, entry
);
662 unsigned int idx
= lconst
->idx
;
663 const GLint
*values
= (const GLint
*)lconst
->value
;
665 TRACE_(d3d_constants
)("Loading local constants %i: %i\n", idx
, values
[0]);
667 snprintf(tmp_name
, sizeof(tmp_name
), "%s_b[%i]", prefix
, idx
);
668 tmp_loc
= GL_EXTCALL(glGetUniformLocationARB(programId
, tmp_name
));
670 /* We found this uniform name in the program - go ahead and send the data */
671 GL_EXTCALL(glUniform1ivARB(tmp_loc
, 1, values
));
672 checkGLcall("glUniform1ivARB");
674 ptr
= list_next(&shader
->constantsB
, ptr
);
678 static void reset_program_constant_version(struct wine_rb_entry
*entry
, void *context
)
680 WINE_RB_ENTRY_VALUE(entry
, struct glsl_shader_prog_link
, program_lookup_entry
)->constant_version
= 0;
683 /* Context activation is done by the caller (state handler). */
684 static void shader_glsl_load_np2fixup_constants(void *shader_priv
,
685 const struct wined3d_gl_info
*gl_info
, const struct wined3d_state
*state
)
687 struct shader_glsl_priv
*glsl_priv
= shader_priv
;
688 const struct glsl_shader_prog_link
*prog
= glsl_priv
->glsl_program
;
690 /* No GLSL program set - nothing to do. */
693 /* NP2 texcoord fixup is (currently) only done for pixelshaders. */
694 if (!use_ps(state
)) return;
696 if (prog
->ps
.np2_fixup_info
&& prog
->ps
.np2_fixup_location
!= -1)
699 UINT fixup
= prog
->ps
.np2_fixup_info
->active
;
700 GLfloat np2fixup_constants
[4 * MAX_FRAGMENT_SAMPLERS
];
702 for (i
= 0; fixup
; fixup
>>= 1, ++i
)
704 const struct wined3d_texture
*tex
= state
->textures
[i
];
705 const unsigned char idx
= prog
->ps
.np2_fixup_info
->idx
[i
];
706 GLfloat
*tex_dim
= &np2fixup_constants
[(idx
>> 1) * 4];
710 ERR("Nonexistent texture is flagged for NP2 texcoord fixup.\n");
716 tex_dim
[2] = tex
->pow2_matrix
[0];
717 tex_dim
[3] = tex
->pow2_matrix
[5];
721 tex_dim
[0] = tex
->pow2_matrix
[0];
722 tex_dim
[1] = tex
->pow2_matrix
[5];
726 GL_EXTCALL(glUniform4fvARB(prog
->ps
.np2_fixup_location
,
727 prog
->ps
.np2_fixup_info
->num_consts
, np2fixup_constants
));
731 /* Context activation is done by the caller (state handler). */
732 static void shader_glsl_load_constants(const struct wined3d_context
*context
,
733 BOOL usePixelShader
, BOOL useVertexShader
)
735 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
736 struct wined3d_device
*device
= context
->swapchain
->device
;
737 struct wined3d_stateblock
*stateBlock
= device
->stateBlock
;
738 const struct wined3d_state
*state
= &stateBlock
->state
;
739 struct shader_glsl_priv
*priv
= device
->shader_priv
;
740 float position_fixup
[4];
742 GLhandleARB programId
;
743 struct glsl_shader_prog_link
*prog
= priv
->glsl_program
;
744 UINT constant_version
;
748 /* No GLSL program set - nothing to do. */
751 programId
= prog
->programId
;
752 constant_version
= prog
->constant_version
;
756 const struct wined3d_shader
*vshader
= state
->vertex_shader
;
758 /* Load DirectX 9 float constants/uniforms for vertex shader */
759 shader_glsl_load_constantsF(vshader
, gl_info
, state
->vs_consts_f
,
760 prog
->vs
.uniform_f_locations
, &priv
->vconst_heap
, priv
->stack
, constant_version
);
762 /* Load DirectX 9 integer constants/uniforms for vertex shader */
763 shader_glsl_load_constantsI(vshader
, gl_info
, prog
->vs
.uniform_i_locations
, state
->vs_consts_i
,
764 stateBlock
->changed
.vertexShaderConstantsI
& vshader
->reg_maps
.integer_constants
);
766 /* Load DirectX 9 boolean constants/uniforms for vertex shader */
767 shader_glsl_load_constantsB(vshader
, gl_info
, programId
, state
->vs_consts_b
,
768 stateBlock
->changed
.vertexShaderConstantsB
& vshader
->reg_maps
.boolean_constants
);
770 /* Upload the position fixup params */
771 shader_get_position_fixup(context
, state
, position_fixup
);
772 GL_EXTCALL(glUniform4fvARB(prog
->vs
.pos_fixup_location
, 1, position_fixup
));
773 checkGLcall("glUniform4fvARB");
778 const struct wined3d_shader
*pshader
= state
->pixel_shader
;
780 /* Load DirectX 9 float constants/uniforms for pixel shader */
781 shader_glsl_load_constantsF(pshader
, gl_info
, state
->ps_consts_f
,
782 prog
->ps
.uniform_f_locations
, &priv
->pconst_heap
, priv
->stack
, constant_version
);
784 /* Load DirectX 9 integer constants/uniforms for pixel shader */
785 shader_glsl_load_constantsI(pshader
, gl_info
, prog
->ps
.uniform_i_locations
, state
->ps_consts_i
,
786 stateBlock
->changed
.pixelShaderConstantsI
& pshader
->reg_maps
.integer_constants
);
788 /* Load DirectX 9 boolean constants/uniforms for pixel shader */
789 shader_glsl_load_constantsB(pshader
, gl_info
, programId
, state
->ps_consts_b
,
790 stateBlock
->changed
.pixelShaderConstantsB
& pshader
->reg_maps
.boolean_constants
);
792 /* Upload the environment bump map matrix if needed. The needsbumpmat
793 * member specifies the texture stage to load the matrix from. It
794 * can't be 0 for a valid texbem instruction. */
795 for (i
= 0; i
< MAX_TEXTURES
; ++i
)
799 if (prog
->ps
.bumpenv_mat_location
[i
] == -1)
802 data
= (const float *)&state
->texture_states
[i
][WINED3D_TSS_BUMPENV_MAT00
];
803 GL_EXTCALL(glUniformMatrix2fvARB(prog
->ps
.bumpenv_mat_location
[i
], 1, 0, data
));
804 checkGLcall("glUniformMatrix2fvARB");
806 /* texbeml needs the luminance scale and offset too. If texbeml
807 * is used, needsbumpmat is set too, so we can check that in the
808 * needsbumpmat check. */
809 if (prog
->ps
.bumpenv_lum_scale_location
[i
] != -1)
811 const GLfloat
*scale
= (const GLfloat
*)&state
->texture_states
[i
][WINED3D_TSS_BUMPENV_LSCALE
];
812 const GLfloat
*offset
= (const GLfloat
*)&state
->texture_states
[i
][WINED3D_TSS_BUMPENV_LOFFSET
];
814 GL_EXTCALL(glUniform1fvARB(prog
->ps
.bumpenv_lum_scale_location
[i
], 1, scale
));
815 checkGLcall("glUniform1fvARB");
816 GL_EXTCALL(glUniform1fvARB(prog
->ps
.bumpenv_lum_offset_location
[i
], 1, offset
));
817 checkGLcall("glUniform1fvARB");
821 if (prog
->ps
.ycorrection_location
!= -1)
823 float correction_params
[4];
825 if (context
->render_offscreen
)
827 correction_params
[0] = 0.0f
;
828 correction_params
[1] = 1.0f
;
830 /* position is window relative, not viewport relative */
831 correction_params
[0] = (float) context
->current_rt
->resource
.height
;
832 correction_params
[1] = -1.0f
;
834 GL_EXTCALL(glUniform4fvARB(prog
->ps
.ycorrection_location
, 1, correction_params
));
837 else if (priv
->fragment_pipe
== &glsl_fragment_pipe
)
841 for (i
= 0; i
< MAX_TEXTURES
; ++i
)
843 GL_EXTCALL(glUniformMatrix2fvARB(prog
->ps
.bumpenv_mat_location
[i
], 1, 0,
844 (const float *)&state
->texture_states
[i
][WINED3D_TSS_BUMPENV_MAT00
]));
845 GL_EXTCALL(glUniform1fARB(prog
->ps
.bumpenv_lum_scale_location
[i
],
846 *(const float *)&state
->texture_states
[i
][WINED3D_TSS_BUMPENV_LSCALE
]));
847 GL_EXTCALL(glUniform1fARB(prog
->ps
.bumpenv_lum_offset_location
[i
],
848 *(const float *)&state
->texture_states
[i
][WINED3D_TSS_BUMPENV_LOFFSET
]));
851 D3DCOLORTOGLFLOAT4(state
->render_states
[WINED3D_RS_TEXTUREFACTOR
], col
);
852 GL_EXTCALL(glUniform4fARB(prog
->ps
.tex_factor_location
, col
[0], col
[1], col
[2], col
[3]));
854 if (state
->render_states
[WINED3D_RS_SPECULARENABLE
])
855 GL_EXTCALL(glUniform4fARB(prog
->ps
.specular_enable_location
, 1.0f
, 1.0f
, 1.0f
, 0.0f
));
857 GL_EXTCALL(glUniform4fARB(prog
->ps
.specular_enable_location
, 0.0f
, 0.0f
, 0.0f
, 0.0f
));
859 checkGLcall("fixed function uniforms");
862 if (priv
->next_constant_version
== UINT_MAX
)
864 TRACE("Max constant version reached, resetting to 0.\n");
865 wine_rb_for_each_entry(&priv
->program_lookup
, reset_program_constant_version
, NULL
);
866 priv
->next_constant_version
= 1;
870 prog
->constant_version
= priv
->next_constant_version
++;
874 static void update_heap_entry(const struct constant_heap
*heap
, unsigned int idx
,
875 unsigned int heap_idx
, DWORD new_version
)
877 struct constant_entry
*entries
= heap
->entries
;
878 unsigned int *positions
= heap
->positions
;
879 unsigned int parent_idx
;
883 parent_idx
= heap_idx
>> 1;
885 if (new_version
<= entries
[parent_idx
].version
) break;
887 entries
[heap_idx
] = entries
[parent_idx
];
888 positions
[entries
[parent_idx
].idx
] = heap_idx
;
889 heap_idx
= parent_idx
;
892 entries
[heap_idx
].version
= new_version
;
893 entries
[heap_idx
].idx
= idx
;
894 positions
[idx
] = heap_idx
;
897 static void shader_glsl_update_float_vertex_constants(struct wined3d_device
*device
, UINT start
, UINT count
)
899 struct shader_glsl_priv
*priv
= device
->shader_priv
;
900 struct constant_heap
*heap
= &priv
->vconst_heap
;
903 for (i
= start
; i
< count
+ start
; ++i
)
905 if (!device
->stateBlock
->changed
.vertexShaderConstantsF
[i
])
906 update_heap_entry(heap
, i
, heap
->size
++, priv
->next_constant_version
);
908 update_heap_entry(heap
, i
, heap
->positions
[i
], priv
->next_constant_version
);
912 static void shader_glsl_update_float_pixel_constants(struct wined3d_device
*device
, UINT start
, UINT count
)
914 struct shader_glsl_priv
*priv
= device
->shader_priv
;
915 struct constant_heap
*heap
= &priv
->pconst_heap
;
918 for (i
= start
; i
< count
+ start
; ++i
)
920 if (!device
->stateBlock
->changed
.pixelShaderConstantsF
[i
])
921 update_heap_entry(heap
, i
, heap
->size
++, priv
->next_constant_version
);
923 update_heap_entry(heap
, i
, heap
->positions
[i
], priv
->next_constant_version
);
927 static unsigned int vec4_varyings(DWORD shader_major
, const struct wined3d_gl_info
*gl_info
)
929 unsigned int ret
= gl_info
->limits
.glsl_varyings
/ 4;
930 /* 4.0 shaders do not write clip coords because d3d10 does not support user clipplanes */
931 if(shader_major
> 3) return ret
;
933 /* 3.0 shaders may need an extra varying for the clip coord on some cards(mostly dx10 ones) */
934 if (gl_info
->quirks
& WINED3D_QUIRK_GLSL_CLIP_VARYING
) ret
-= 1;
938 /** Generate the variable & register declarations for the GLSL output target */
939 static void shader_generate_glsl_declarations(const struct wined3d_context
*context
,
940 struct wined3d_shader_buffer
*buffer
, const struct wined3d_shader
*shader
,
941 const struct wined3d_shader_reg_maps
*reg_maps
, const struct shader_glsl_ctx_priv
*ctx_priv
)
943 const struct wined3d_shader_version
*version
= ®_maps
->shader_version
;
944 const struct wined3d_state
*state
= &shader
->device
->stateBlock
->state
;
945 const struct ps_compile_args
*ps_args
= ctx_priv
->cur_ps_args
;
946 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
947 const struct wined3d_fb_state
*fb
= &shader
->device
->fb
;
948 unsigned int i
, extra_constants_needed
= 0;
949 const struct wined3d_shader_lconst
*lconst
;
953 prefix
= shader_glsl_get_prefix(version
->type
);
955 /* Prototype the subroutines */
956 for (i
= 0, map
= reg_maps
->labels
; map
; map
>>= 1, ++i
)
958 if (map
& 1) shader_addline(buffer
, "void subroutine%u();\n", i
);
961 /* Declare the constants (aka uniforms) */
962 if (shader
->limits
.constant_float
> 0)
964 unsigned max_constantsF
;
966 /* Unless the shader uses indirect addressing, always declare the
967 * maximum array size and ignore that we need some uniforms privately.
968 * E.g. if GL supports 256 uniforms, and we need 2 for the pos fixup
969 * and immediate values, still declare VC[256]. If the shader needs
970 * more uniforms than we have it won't work in any case. If it uses
971 * less, the compiler will figure out which uniforms are really used
972 * and strip them out. This allows a shader to use c255 on a dx9 card,
973 * as long as it doesn't also use all the other constants.
975 * If the shader uses indirect addressing the compiler must assume
976 * that all declared uniforms are used. In this case, declare only the
977 * amount that we're assured to have.
979 * Thus we run into problems in these two cases:
980 * 1) The shader really uses more uniforms than supported.
981 * 2) The shader uses indirect addressing, less constants than
982 * supported, but uses a constant index > #supported consts. */
983 if (version
->type
== WINED3D_SHADER_TYPE_PIXEL
)
985 /* No indirect addressing here. */
986 max_constantsF
= gl_info
->limits
.glsl_ps_float_constants
;
990 if (reg_maps
->usesrelconstF
)
992 /* Subtract the other potential uniforms from the max
993 * available (bools, ints, and 1 row of projection matrix).
994 * Subtract another uniform for immediate values, which have
995 * to be loaded via uniform by the driver as well. The shader
996 * code only uses 0.5, 2.0, 1.0, 128 and -128 in vertex
997 * shader code, so one vec4 should be enough. (Unfortunately
998 * the Nvidia driver doesn't store 128 and -128 in one float).
1000 * Writing gl_ClipVertex requires one uniform for each
1001 * clipplane as well. */
1002 max_constantsF
= gl_info
->limits
.glsl_vs_float_constants
- 3;
1003 if(ctx_priv
->cur_vs_args
->clip_enabled
)
1005 max_constantsF
-= gl_info
->limits
.clipplanes
;
1007 max_constantsF
-= count_bits(reg_maps
->integer_constants
);
1008 /* Strictly speaking a bool only uses one scalar, but the nvidia(Linux) compiler doesn't pack them properly,
1009 * so each scalar requires a full vec4. We could work around this by packing the booleans ourselves, but
1010 * for now take this into account when calculating the number of available constants
1012 max_constantsF
-= count_bits(reg_maps
->boolean_constants
);
1013 /* Set by driver quirks in directx.c */
1014 max_constantsF
-= gl_info
->reserved_glsl_constants
;
1016 if (max_constantsF
< shader
->limits
.constant_float
)
1018 static unsigned int once
;
1021 ERR_(winediag
)("The hardware does not support enough uniform components to run this shader,"
1022 " it may not render correctly.\n");
1024 WARN("The hardware does not support enough uniform components to run this shader.\n");
1029 max_constantsF
= gl_info
->limits
.glsl_vs_float_constants
;
1032 max_constantsF
= min(shader
->limits
.constant_float
, max_constantsF
);
1033 shader_addline(buffer
, "uniform vec4 %s_c[%u];\n", prefix
, max_constantsF
);
1036 /* Always declare the full set of constants, the compiler can remove the
1037 * unused ones because d3d doesn't (yet) support indirect int and bool
1038 * constant addressing. This avoids problems if the app uses e.g. i0 and i9. */
1039 if (shader
->limits
.constant_int
> 0 && reg_maps
->integer_constants
)
1040 shader_addline(buffer
, "uniform ivec4 %s_i[%u];\n", prefix
, shader
->limits
.constant_int
);
1042 if (shader
->limits
.constant_bool
> 0 && reg_maps
->boolean_constants
)
1043 shader_addline(buffer
, "uniform bool %s_b[%u];\n", prefix
, shader
->limits
.constant_bool
);
1045 for (i
= 0; i
< WINED3D_MAX_CBS
; ++i
)
1047 if (reg_maps
->cb_sizes
[i
])
1048 shader_addline(buffer
, "uniform vec4 %s_cb%u[%u];\n", prefix
, i
, reg_maps
->cb_sizes
[i
]);
1051 /* Declare texture samplers */
1052 for (i
= 0; i
< shader
->limits
.sampler
; ++i
)
1054 if (reg_maps
->sampler_type
[i
])
1056 BOOL shadow_sampler
= version
->type
== WINED3D_SHADER_TYPE_PIXEL
&& (ps_args
->shadow
& (1 << i
));
1059 switch (reg_maps
->sampler_type
[i
])
1063 shader_addline(buffer
, "uniform sampler1DShadow %s_sampler%u;\n", prefix
, i
);
1065 shader_addline(buffer
, "uniform sampler1D %s_sampler%u;\n", prefix
, i
);
1068 tex_rect
= version
->type
== WINED3D_SHADER_TYPE_PIXEL
&& (ps_args
->np2_fixup
& (1 << i
));
1069 tex_rect
= tex_rect
&& gl_info
->supported
[ARB_TEXTURE_RECTANGLE
];
1073 shader_addline(buffer
, "uniform sampler2DRectShadow %s_sampler%u;\n", prefix
, i
);
1075 shader_addline(buffer
, "uniform sampler2DShadow %s_sampler%u;\n", prefix
, i
);
1080 shader_addline(buffer
, "uniform sampler2DRect %s_sampler%u;\n", prefix
, i
);
1082 shader_addline(buffer
, "uniform sampler2D %s_sampler%u;\n", prefix
, i
);
1085 case WINED3DSTT_CUBE
:
1087 FIXME("Unsupported Cube shadow sampler.\n");
1088 shader_addline(buffer
, "uniform samplerCube %s_sampler%u;\n", prefix
, i
);
1090 case WINED3DSTT_VOLUME
:
1092 FIXME("Unsupported 3D shadow sampler.\n");
1093 shader_addline(buffer
, "uniform sampler3D %s_sampler%u;\n", prefix
, i
);
1096 shader_addline(buffer
, "uniform unsupported_sampler %s_sampler%u;\n", prefix
, i
);
1097 FIXME("Unrecognized sampler type: %#x\n", reg_maps
->sampler_type
[i
]);
1103 /* Declare uniforms for NP2 texcoord fixup:
1104 * This is NOT done inside the loop that declares the texture samplers
1105 * since the NP2 fixup code is currently only used for the GeforceFX
1106 * series and when forcing the ARB_npot extension off. Modern cards just
1107 * skip the code anyway, so put it inside a separate loop. */
1108 if (version
->type
== WINED3D_SHADER_TYPE_PIXEL
&& ps_args
->np2_fixup
)
1110 struct ps_np2fixup_info
*fixup
= ctx_priv
->cur_np2fixup_info
;
1113 /* NP2/RECT textures in OpenGL use texcoords in the range [0,width]x[0,height]
1114 * while D3D has them in the (normalized) [0,1]x[0,1] range.
1115 * samplerNP2Fixup stores texture dimensions and is updated through
1116 * shader_glsl_load_np2fixup_constants when the sampler changes. */
1118 for (i
= 0; i
< shader
->limits
.sampler
; ++i
)
1120 if (reg_maps
->sampler_type
[i
])
1122 if (!(ps_args
->np2_fixup
& (1 << i
))) continue;
1124 if (WINED3DSTT_2D
!= reg_maps
->sampler_type
[i
]) {
1125 FIXME("Non-2D texture is flagged for NP2 texcoord fixup.\n");
1129 fixup
->idx
[i
] = cur
++;
1133 fixup
->num_consts
= (cur
+ 1) >> 1;
1134 fixup
->active
= ps_args
->np2_fixup
;
1135 shader_addline(buffer
, "uniform vec4 %s_samplerNP2Fixup[%u];\n", prefix
, fixup
->num_consts
);
1138 /* Declare address variables */
1139 for (i
= 0, map
= reg_maps
->address
; map
; map
>>= 1, ++i
)
1141 if (map
& 1) shader_addline(buffer
, "ivec4 A%u;\n", i
);
1144 /* Declare texture coordinate temporaries and initialize them */
1145 for (i
= 0, map
= reg_maps
->texcoord
; map
; map
>>= 1, ++i
)
1147 if (map
& 1) shader_addline(buffer
, "vec4 T%u = gl_TexCoord[%u];\n", i
, i
);
1150 if (version
->type
== WINED3D_SHADER_TYPE_VERTEX
)
1152 /* Declare attributes. */
1153 for (i
= 0, map
= reg_maps
->input_registers
; map
; map
>>= 1, ++i
)
1156 shader_addline(buffer
, "attribute vec4 %s_in%u;\n", prefix
, i
);
1159 shader_addline(buffer
, "uniform vec4 posFixup;\n");
1160 shader_addline(buffer
, "void order_ps_input(in vec4[%u]);\n", shader
->limits
.packed_output
);
1162 else if (version
->type
== WINED3D_SHADER_TYPE_GEOMETRY
)
1164 shader_addline(buffer
, "varying in vec4 gs_in[][%u];\n", shader
->limits
.packed_input
);
1166 else if (version
->type
== WINED3D_SHADER_TYPE_PIXEL
)
1168 if (version
->major
>= 3)
1170 UINT in_count
= min(vec4_varyings(version
->major
, gl_info
), shader
->limits
.packed_input
);
1173 shader_addline(buffer
, "varying vec4 %s_in[%u];\n", prefix
, in_count
);
1175 /* TODO: Write a replacement shader for the fixed function
1176 * vertex pipeline, so this isn't needed. For fixed function
1177 * vertex processing + 3.0 pixel shader we need a separate
1178 * function in the pixel shader that reads the fixed function
1179 * color into the packed input registers. */
1180 shader_addline(buffer
, "vec4 %s_in[%u];\n", prefix
, in_count
);
1183 for (i
= 0, map
= reg_maps
->bumpmat
; map
; map
>>= 1, ++i
)
1188 shader_addline(buffer
, "uniform mat2 bumpenv_mat%u;\n", i
);
1190 if (reg_maps
->luminanceparams
& (1 << i
))
1192 shader_addline(buffer
, "uniform float bumpenv_lum_scale%u;\n", i
);
1193 shader_addline(buffer
, "uniform float bumpenv_lum_offset%u;\n", i
);
1194 extra_constants_needed
++;
1197 extra_constants_needed
++;
1200 if (ps_args
->srgb_correction
)
1202 shader_addline(buffer
, "const vec4 srgb_const0 = vec4(%.8e, %.8e, %.8e, %.8e);\n",
1203 srgb_pow
, srgb_mul_high
, srgb_sub_high
, srgb_mul_low
);
1204 shader_addline(buffer
, "const vec4 srgb_const1 = vec4(%.8e, 0.0, 0.0, 0.0);\n",
1207 if (reg_maps
->vpos
|| reg_maps
->usesdsy
)
1209 if (shader
->limits
.constant_float
+ extra_constants_needed
1210 + 1 < gl_info
->limits
.glsl_ps_float_constants
)
1212 shader_addline(buffer
, "uniform vec4 ycorrection;\n");
1213 extra_constants_needed
++;
1217 /* This happens because we do not have proper tracking of the constant registers that are
1218 * actually used, only the max limit of the shader version
1220 FIXME("Cannot find a free uniform for vpos correction params\n");
1221 shader_addline(buffer
, "const vec4 ycorrection = vec4(%f, %f, 0.0, 0.0);\n",
1222 context
->render_offscreen
? 0.0f
: fb
->render_targets
[0]->resource
.height
,
1223 context
->render_offscreen
? 1.0f
: -1.0f
);
1225 shader_addline(buffer
, "vec4 vpos;\n");
1229 /* Declare output register temporaries */
1230 if (shader
->limits
.packed_output
)
1231 shader_addline(buffer
, "vec4 %s_out[%u];\n", prefix
, shader
->limits
.packed_output
);
1233 /* Declare temporary variables */
1234 for (i
= 0, map
= reg_maps
->temporary
; map
; map
>>= 1, ++i
)
1236 if (map
& 1) shader_addline(buffer
, "vec4 R%u;\n", i
);
1239 /* Declare loop registers aLx */
1240 if (version
->major
< 4)
1242 for (i
= 0; i
< reg_maps
->loop_depth
; ++i
)
1244 shader_addline(buffer
, "int aL%u;\n", i
);
1245 shader_addline(buffer
, "int tmpInt%u;\n", i
);
1249 /* Temporary variables for matrix operations */
1250 shader_addline(buffer
, "vec4 tmp0;\n");
1251 shader_addline(buffer
, "vec4 tmp1;\n");
1253 if (!shader
->load_local_constsF
)
1255 LIST_FOR_EACH_ENTRY(lconst
, &shader
->constantsF
, struct wined3d_shader_lconst
, entry
)
1258 value
= (const float *)lconst
->value
;
1259 shader_addline(buffer
, "const vec4 %s_lc%u = vec4(%.8e, %.8e, %.8e, %.8e);\n",
1260 prefix
, lconst
->idx
, value
[0], value
[1], value
[2], value
[3]);
1264 /* Start the main program. */
1265 shader_addline(buffer
, "void main()\n{\n");
1267 /* Direct3D applications expect integer vPos values, while OpenGL drivers
1268 * add approximately 0.5. This causes off-by-one problems as spotted by
1269 * the vPos d3d9 visual test. Unfortunately ATI cards do not add exactly
1270 * 0.5, but rather something like 0.49999999 or 0.50000001, which still
1271 * causes precision troubles when we just subtract 0.5.
1273 * To deal with that, just floor() the position. This will eliminate the
1274 * fraction on all cards.
1276 * TODO: Test how this behaves with multisampling.
1278 * An advantage of floor is that it works even if the driver doesn't add
1279 * 0.5. It is somewhat questionable if 1.5, 2.5, ... are the proper values
1280 * to return in gl_FragCoord, even though coordinates specify the pixel
1281 * centers instead of the pixel corners. This code will behave correctly
1282 * on drivers that returns integer values. */
1283 if (version
->type
== WINED3D_SHADER_TYPE_PIXEL
&& reg_maps
->vpos
)
1284 shader_addline(buffer
,
1285 "vpos = floor(vec4(0, ycorrection[0], 0, 0) + gl_FragCoord * vec4(1, ycorrection[1], 1, 1));\n");
1288 /*****************************************************************************
1289 * Functions to generate GLSL strings from DirectX Shader bytecode begin here.
1291 * For more information, see http://wiki.winehq.org/DirectX-Shaders
1292 ****************************************************************************/
1295 static void shader_glsl_add_src_param(const struct wined3d_shader_instruction
*ins
,
1296 const struct wined3d_shader_src_param
*wined3d_src
, DWORD mask
, struct glsl_src_param
*glsl_src
);
1298 /** Used for opcode modifiers - They multiply the result by the specified amount */
1299 static const char * const shift_glsl_tab
[] = {
1301 "2.0 * ", /* 1 (x2) */
1302 "4.0 * ", /* 2 (x4) */
1303 "8.0 * ", /* 3 (x8) */
1304 "16.0 * ", /* 4 (x16) */
1305 "32.0 * ", /* 5 (x32) */
1312 "0.0625 * ", /* 12 (d16) */
1313 "0.125 * ", /* 13 (d8) */
1314 "0.25 * ", /* 14 (d4) */
1315 "0.5 * " /* 15 (d2) */
1318 /* Generate a GLSL parameter that does the input modifier computation and return the input register/mask to use */
1319 static void shader_glsl_gen_modifier(enum wined3d_shader_src_modifier src_modifier
,
1320 const char *in_reg
, const char *in_regswizzle
, char *out_str
)
1324 switch (src_modifier
)
1326 case WINED3DSPSM_DZ
: /* Need to handle this in the instructions itself (texld & texcrd). */
1327 case WINED3DSPSM_DW
:
1328 case WINED3DSPSM_NONE
:
1329 sprintf(out_str
, "%s%s", in_reg
, in_regswizzle
);
1331 case WINED3DSPSM_NEG
:
1332 sprintf(out_str
, "-%s%s", in_reg
, in_regswizzle
);
1334 case WINED3DSPSM_NOT
:
1335 sprintf(out_str
, "!%s%s", in_reg
, in_regswizzle
);
1337 case WINED3DSPSM_BIAS
:
1338 sprintf(out_str
, "(%s%s - vec4(0.5)%s)", in_reg
, in_regswizzle
, in_regswizzle
);
1340 case WINED3DSPSM_BIASNEG
:
1341 sprintf(out_str
, "-(%s%s - vec4(0.5)%s)", in_reg
, in_regswizzle
, in_regswizzle
);
1343 case WINED3DSPSM_SIGN
:
1344 sprintf(out_str
, "(2.0 * (%s%s - 0.5))", in_reg
, in_regswizzle
);
1346 case WINED3DSPSM_SIGNNEG
:
1347 sprintf(out_str
, "-(2.0 * (%s%s - 0.5))", in_reg
, in_regswizzle
);
1349 case WINED3DSPSM_COMP
:
1350 sprintf(out_str
, "(1.0 - %s%s)", in_reg
, in_regswizzle
);
1352 case WINED3DSPSM_X2
:
1353 sprintf(out_str
, "(2.0 * %s%s)", in_reg
, in_regswizzle
);
1355 case WINED3DSPSM_X2NEG
:
1356 sprintf(out_str
, "-(2.0 * %s%s)", in_reg
, in_regswizzle
);
1358 case WINED3DSPSM_ABS
:
1359 sprintf(out_str
, "abs(%s%s)", in_reg
, in_regswizzle
);
1361 case WINED3DSPSM_ABSNEG
:
1362 sprintf(out_str
, "-abs(%s%s)", in_reg
, in_regswizzle
);
1365 FIXME("Unhandled modifier %u\n", src_modifier
);
1366 sprintf(out_str
, "%s%s", in_reg
, in_regswizzle
);
1370 /** Writes the GLSL variable name that corresponds to the register that the
1371 * DX opcode parameter is trying to access */
1372 static void shader_glsl_get_register_name(const struct wined3d_shader_register
*reg
,
1373 char *register_name
, BOOL
*is_color
, const struct wined3d_shader_instruction
*ins
)
1375 /* oPos, oFog and oPts in D3D */
1376 static const char * const hwrastout_reg_names
[] = {"vs_out[10]", "vs_out[11].x", "vs_out[11].y"};
1378 const struct wined3d_shader
*shader
= ins
->ctx
->shader
;
1379 const struct wined3d_shader_reg_maps
*reg_maps
= ins
->ctx
->reg_maps
;
1380 const struct wined3d_shader_version
*version
= ®_maps
->shader_version
;
1381 const struct wined3d_gl_info
*gl_info
= ins
->ctx
->gl_info
;
1382 const char *prefix
= shader_glsl_get_prefix(version
->type
);
1383 struct glsl_src_param rel_param0
, rel_param1
;
1385 if (reg
->idx
[0].offset
!= ~0U && reg
->idx
[0].rel_addr
)
1386 shader_glsl_add_src_param(ins
, reg
->idx
[0].rel_addr
, WINED3DSP_WRITEMASK_0
, &rel_param0
);
1387 if (reg
->idx
[1].offset
!= ~0U && reg
->idx
[1].rel_addr
)
1388 shader_glsl_add_src_param(ins
, reg
->idx
[1].rel_addr
, WINED3DSP_WRITEMASK_0
, &rel_param1
);
1393 case WINED3DSPR_TEMP
:
1394 sprintf(register_name
, "R%u", reg
->idx
[0].offset
);
1397 case WINED3DSPR_INPUT
:
1398 /* vertex shaders */
1399 if (version
->type
== WINED3D_SHADER_TYPE_VERTEX
)
1401 struct shader_glsl_ctx_priv
*priv
= ins
->ctx
->backend_data
;
1402 if (priv
->cur_vs_args
->swizzle_map
& (1 << reg
->idx
[0].offset
))
1404 sprintf(register_name
, "%s_in%u", prefix
, reg
->idx
[0].offset
);
1408 if (version
->type
== WINED3D_SHADER_TYPE_GEOMETRY
)
1410 if (reg
->idx
[0].rel_addr
)
1412 if (reg
->idx
[1].rel_addr
)
1413 sprintf(register_name
, "gs_in[%s + %u][%s + %u]",
1414 rel_param0
.param_str
, reg
->idx
[0].offset
, rel_param1
.param_str
, reg
->idx
[1].offset
);
1416 sprintf(register_name
, "gs_in[%s + %u][%u]",
1417 rel_param0
.param_str
, reg
->idx
[0].offset
, reg
->idx
[1].offset
);
1419 else if (reg
->idx
[1].rel_addr
)
1420 sprintf(register_name
, "gs_in[%u][%s + %u]",
1421 reg
->idx
[0].offset
, rel_param1
.param_str
, reg
->idx
[1].offset
);
1423 sprintf(register_name
, "gs_in[%u][%u]", reg
->idx
[0].offset
, reg
->idx
[1].offset
);
1427 /* pixel shaders >= 3.0 */
1428 if (version
->major
>= 3)
1430 DWORD idx
= shader
->u
.ps
.input_reg_map
[reg
->idx
[0].offset
];
1431 unsigned int in_count
= vec4_varyings(version
->major
, gl_info
);
1433 if (reg
->idx
[0].rel_addr
)
1435 /* Removing a + 0 would be an obvious optimization, but
1436 * OS X doesn't see the NOP operation there. */
1439 if (shader
->u
.ps
.declared_in_count
> in_count
)
1441 sprintf(register_name
,
1442 "((%s + %u) > %u ? (%s + %u) > %u ? gl_SecondaryColor : gl_Color : %s_in[%s + %u])",
1443 rel_param0
.param_str
, idx
, in_count
- 1, rel_param0
.param_str
, idx
, in_count
,
1444 prefix
, rel_param0
.param_str
, idx
);
1448 sprintf(register_name
, "%s_in[%s + %u]", prefix
, rel_param0
.param_str
, idx
);
1453 if (shader
->u
.ps
.declared_in_count
> in_count
)
1455 sprintf(register_name
, "((%s) > %u ? (%s) > %u ? gl_SecondaryColor : gl_Color : %s_in[%s])",
1456 rel_param0
.param_str
, in_count
- 1, rel_param0
.param_str
, in_count
,
1457 prefix
, rel_param0
.param_str
);
1461 sprintf(register_name
, "%s_in[%s]", prefix
, rel_param0
.param_str
);
1467 if (idx
== in_count
) sprintf(register_name
, "gl_Color");
1468 else if (idx
== in_count
+ 1) sprintf(register_name
, "gl_SecondaryColor");
1469 else sprintf(register_name
, "%s_in[%u]", prefix
, idx
);
1474 if (!reg
->idx
[0].offset
)
1475 strcpy(register_name
, "gl_Color");
1477 strcpy(register_name
, "gl_SecondaryColor");
1482 case WINED3DSPR_CONST
:
1484 /* Relative addressing */
1485 if (reg
->idx
[0].rel_addr
)
1487 if (reg
->idx
[0].offset
)
1488 sprintf(register_name
, "%s_c[%s + %u]", prefix
, rel_param0
.param_str
, reg
->idx
[0].offset
);
1490 sprintf(register_name
, "%s_c[%s]", prefix
, rel_param0
.param_str
);
1494 if (shader_constant_is_local(shader
, reg
->idx
[0].offset
))
1495 sprintf(register_name
, "%s_lc%u", prefix
, reg
->idx
[0].offset
);
1497 sprintf(register_name
, "%s_c[%u]", prefix
, reg
->idx
[0].offset
);
1502 case WINED3DSPR_CONSTINT
:
1503 sprintf(register_name
, "%s_i[%u]", prefix
, reg
->idx
[0].offset
);
1506 case WINED3DSPR_CONSTBOOL
:
1507 sprintf(register_name
, "%s_b[%u]", prefix
, reg
->idx
[0].offset
);
1510 case WINED3DSPR_TEXTURE
: /* case WINED3DSPR_ADDR: */
1511 if (version
->type
== WINED3D_SHADER_TYPE_PIXEL
)
1512 sprintf(register_name
, "T%u", reg
->idx
[0].offset
);
1514 sprintf(register_name
, "A%u", reg
->idx
[0].offset
);
1517 case WINED3DSPR_LOOP
:
1518 sprintf(register_name
, "aL%u", ins
->ctx
->loop_state
->current_reg
- 1);
1521 case WINED3DSPR_SAMPLER
:
1522 sprintf(register_name
, "%s_sampler%u", prefix
, reg
->idx
[0].offset
);
1525 case WINED3DSPR_COLOROUT
:
1526 if (reg
->idx
[0].offset
>= gl_info
->limits
.buffers
)
1527 WARN("Write to render target %u, only %d supported.\n",
1528 reg
->idx
[0].offset
, gl_info
->limits
.buffers
);
1530 sprintf(register_name
, "gl_FragData[%u]", reg
->idx
[0].offset
);
1533 case WINED3DSPR_RASTOUT
:
1534 sprintf(register_name
, "%s", hwrastout_reg_names
[reg
->idx
[0].offset
]);
1537 case WINED3DSPR_DEPTHOUT
:
1538 sprintf(register_name
, "gl_FragDepth");
1541 case WINED3DSPR_ATTROUT
:
1542 if (!reg
->idx
[0].offset
)
1543 sprintf(register_name
, "%s_out[8]", prefix
);
1545 sprintf(register_name
, "%s_out[9]", prefix
);
1548 case WINED3DSPR_TEXCRDOUT
:
1549 /* Vertex shaders >= 3.0: WINED3DSPR_OUTPUT */
1550 sprintf(register_name
, "%s_out[%u]", prefix
, reg
->idx
[0].offset
);
1553 case WINED3DSPR_MISCTYPE
:
1554 if (!reg
->idx
[0].offset
)
1557 sprintf(register_name
, "vpos");
1559 else if (reg
->idx
[0].offset
== 1)
1561 /* Note that gl_FrontFacing is a bool, while vFace is
1562 * a float for which the sign determines front/back */
1563 sprintf(register_name
, "(gl_FrontFacing ? 1.0 : -1.0)");
1567 FIXME("Unhandled misctype register %u.\n", reg
->idx
[0].offset
);
1568 sprintf(register_name
, "unrecognized_register");
1572 case WINED3DSPR_IMMCONST
:
1573 switch (reg
->immconst_type
)
1575 case WINED3D_IMMCONST_SCALAR
:
1576 switch (reg
->data_type
)
1578 case WINED3D_DATA_FLOAT
:
1579 sprintf(register_name
, "%.8e", *(const float *)reg
->immconst_data
);
1581 case WINED3D_DATA_INT
:
1582 sprintf(register_name
, "%#x", reg
->immconst_data
[0]);
1584 case WINED3D_DATA_RESOURCE
:
1585 case WINED3D_DATA_SAMPLER
:
1586 case WINED3D_DATA_UINT
:
1587 sprintf(register_name
, "%#xu", reg
->immconst_data
[0]);
1590 sprintf(register_name
, "<unhandled data type %#x>", reg
->data_type
);
1595 case WINED3D_IMMCONST_VEC4
:
1596 switch (reg
->data_type
)
1598 case WINED3D_DATA_FLOAT
:
1599 sprintf(register_name
, "vec4(%.8e, %.8e, %.8e, %.8e)",
1600 *(const float *)®
->immconst_data
[0], *(const float *)®
->immconst_data
[1],
1601 *(const float *)®
->immconst_data
[2], *(const float *)®
->immconst_data
[3]);
1603 case WINED3D_DATA_INT
:
1604 sprintf(register_name
, "ivec4(%#x, %#x, %#x, %#x)",
1605 reg
->immconst_data
[0], reg
->immconst_data
[1],
1606 reg
->immconst_data
[2], reg
->immconst_data
[3]);
1608 case WINED3D_DATA_RESOURCE
:
1609 case WINED3D_DATA_SAMPLER
:
1610 case WINED3D_DATA_UINT
:
1611 sprintf(register_name
, "uvec4(%#xu, %#xu, %#xu, %#xu)",
1612 reg
->immconst_data
[0], reg
->immconst_data
[1],
1613 reg
->immconst_data
[2], reg
->immconst_data
[3]);
1616 sprintf(register_name
, "<unhandled data type %#x>", reg
->data_type
);
1622 FIXME("Unhandled immconst type %#x\n", reg
->immconst_type
);
1623 sprintf(register_name
, "<unhandled_immconst_type %#x>", reg
->immconst_type
);
1627 case WINED3DSPR_CONSTBUFFER
:
1628 if (reg
->idx
[1].rel_addr
)
1629 sprintf(register_name
, "%s_cb%u[%s + %u]",
1630 prefix
, reg
->idx
[0].offset
, rel_param1
.param_str
, reg
->idx
[1].offset
);
1632 sprintf(register_name
, "%s_cb%u[%u]", prefix
, reg
->idx
[0].offset
, reg
->idx
[1].offset
);
1635 case WINED3DSPR_PRIMID
:
1636 sprintf(register_name
, "uint(gl_PrimitiveIDIn)");
1640 FIXME("Unhandled register type %#x.\n", reg
->type
);
1641 sprintf(register_name
, "unrecognized_register");
1646 static void shader_glsl_write_mask_to_str(DWORD write_mask
, char *str
)
1649 if (write_mask
& WINED3DSP_WRITEMASK_0
) *str
++ = 'x';
1650 if (write_mask
& WINED3DSP_WRITEMASK_1
) *str
++ = 'y';
1651 if (write_mask
& WINED3DSP_WRITEMASK_2
) *str
++ = 'z';
1652 if (write_mask
& WINED3DSP_WRITEMASK_3
) *str
++ = 'w';
1656 /* Get the GLSL write mask for the destination register */
1657 static DWORD
shader_glsl_get_write_mask(const struct wined3d_shader_dst_param
*param
, char *write_mask
)
1659 DWORD mask
= param
->write_mask
;
1661 if (shader_is_scalar(¶m
->reg
))
1663 mask
= WINED3DSP_WRITEMASK_0
;
1668 shader_glsl_write_mask_to_str(mask
, write_mask
);
1674 static unsigned int shader_glsl_get_write_mask_size(DWORD write_mask
) {
1675 unsigned int size
= 0;
1677 if (write_mask
& WINED3DSP_WRITEMASK_0
) ++size
;
1678 if (write_mask
& WINED3DSP_WRITEMASK_1
) ++size
;
1679 if (write_mask
& WINED3DSP_WRITEMASK_2
) ++size
;
1680 if (write_mask
& WINED3DSP_WRITEMASK_3
) ++size
;
1685 static void shader_glsl_swizzle_to_str(const DWORD swizzle
, BOOL fixup
, DWORD mask
, char *str
)
1687 /* For registers of type WINED3DDECLTYPE_D3DCOLOR, data is stored as "bgra",
1688 * but addressed as "rgba". To fix this we need to swap the register's x
1689 * and z components. */
1690 const char *swizzle_chars
= fixup
? "zyxw" : "xyzw";
1693 /* swizzle bits fields: wwzzyyxx */
1694 if (mask
& WINED3DSP_WRITEMASK_0
) *str
++ = swizzle_chars
[swizzle
& 0x03];
1695 if (mask
& WINED3DSP_WRITEMASK_1
) *str
++ = swizzle_chars
[(swizzle
>> 2) & 0x03];
1696 if (mask
& WINED3DSP_WRITEMASK_2
) *str
++ = swizzle_chars
[(swizzle
>> 4) & 0x03];
1697 if (mask
& WINED3DSP_WRITEMASK_3
) *str
++ = swizzle_chars
[(swizzle
>> 6) & 0x03];
1701 static void shader_glsl_get_swizzle(const struct wined3d_shader_src_param
*param
,
1702 BOOL fixup
, DWORD mask
, char *swizzle_str
)
1704 if (shader_is_scalar(¶m
->reg
))
1705 *swizzle_str
= '\0';
1707 shader_glsl_swizzle_to_str(param
->swizzle
, fixup
, mask
, swizzle_str
);
1710 /* From a given parameter token, generate the corresponding GLSL string.
1711 * Also, return the actual register name and swizzle in case the
1712 * caller needs this information as well. */
1713 static void shader_glsl_add_src_param(const struct wined3d_shader_instruction
*ins
,
1714 const struct wined3d_shader_src_param
*wined3d_src
, DWORD mask
, struct glsl_src_param
*glsl_src
)
1716 BOOL is_color
= FALSE
;
1717 char swizzle_str
[6];
1719 glsl_src
->reg_name
[0] = '\0';
1720 glsl_src
->param_str
[0] = '\0';
1721 swizzle_str
[0] = '\0';
1723 shader_glsl_get_register_name(&wined3d_src
->reg
, glsl_src
->reg_name
, &is_color
, ins
);
1724 shader_glsl_get_swizzle(wined3d_src
, is_color
, mask
, swizzle_str
);
1726 if (wined3d_src
->reg
.type
== WINED3DSPR_IMMCONST
|| wined3d_src
->reg
.type
== WINED3DSPR_PRIMID
)
1728 shader_glsl_gen_modifier(wined3d_src
->modifiers
, glsl_src
->reg_name
, swizzle_str
, glsl_src
->param_str
);
1732 char param_str
[200];
1734 shader_glsl_gen_modifier(wined3d_src
->modifiers
, glsl_src
->reg_name
, swizzle_str
, param_str
);
1736 switch (wined3d_src
->reg
.data_type
)
1738 case WINED3D_DATA_FLOAT
:
1739 sprintf(glsl_src
->param_str
, "%s", param_str
);
1741 case WINED3D_DATA_INT
:
1742 sprintf(glsl_src
->param_str
, "floatBitsToInt(%s)", param_str
);
1744 case WINED3D_DATA_RESOURCE
:
1745 case WINED3D_DATA_SAMPLER
:
1746 case WINED3D_DATA_UINT
:
1747 sprintf(glsl_src
->param_str
, "floatBitsToUint(%s)", param_str
);
1750 FIXME("Unhandled data type %#x.\n", wined3d_src
->reg
.data_type
);
1751 sprintf(glsl_src
->param_str
, "%s", param_str
);
1757 /* From a given parameter token, generate the corresponding GLSL string.
1758 * Also, return the actual register name and swizzle in case the
1759 * caller needs this information as well. */
1760 static DWORD
shader_glsl_add_dst_param(const struct wined3d_shader_instruction
*ins
,
1761 const struct wined3d_shader_dst_param
*wined3d_dst
, struct glsl_dst_param
*glsl_dst
)
1763 BOOL is_color
= FALSE
;
1765 glsl_dst
->mask_str
[0] = '\0';
1766 glsl_dst
->reg_name
[0] = '\0';
1768 shader_glsl_get_register_name(&wined3d_dst
->reg
, glsl_dst
->reg_name
, &is_color
, ins
);
1769 return shader_glsl_get_write_mask(wined3d_dst
, glsl_dst
->mask_str
);
1772 /* Append the destination part of the instruction to the buffer, return the effective write mask */
1773 static DWORD
shader_glsl_append_dst_ext(struct wined3d_shader_buffer
*buffer
,
1774 const struct wined3d_shader_instruction
*ins
, const struct wined3d_shader_dst_param
*dst
)
1776 struct glsl_dst_param glsl_dst
;
1779 if ((mask
= shader_glsl_add_dst_param(ins
, dst
, &glsl_dst
)))
1781 switch (dst
->reg
.data_type
)
1783 case WINED3D_DATA_FLOAT
:
1784 shader_addline(buffer
, "%s%s = %s(",
1785 glsl_dst
.reg_name
, glsl_dst
.mask_str
, shift_glsl_tab
[dst
->shift
]);
1787 case WINED3D_DATA_INT
:
1788 shader_addline(buffer
, "%s%s = %sintBitsToFloat(",
1789 glsl_dst
.reg_name
, glsl_dst
.mask_str
, shift_glsl_tab
[dst
->shift
]);
1791 case WINED3D_DATA_RESOURCE
:
1792 case WINED3D_DATA_SAMPLER
:
1793 case WINED3D_DATA_UINT
:
1794 shader_addline(buffer
, "%s%s = %suintBitsToFloat(",
1795 glsl_dst
.reg_name
, glsl_dst
.mask_str
, shift_glsl_tab
[dst
->shift
]);
1798 FIXME("Unhandled data type %#x.\n", dst
->reg
.data_type
);
1799 shader_addline(buffer
, "%s%s = %s(",
1800 glsl_dst
.reg_name
, glsl_dst
.mask_str
, shift_glsl_tab
[dst
->shift
]);
1808 /* Append the destination part of the instruction to the buffer, return the effective write mask */
1809 static DWORD
shader_glsl_append_dst(struct wined3d_shader_buffer
*buffer
, const struct wined3d_shader_instruction
*ins
)
1811 return shader_glsl_append_dst_ext(buffer
, ins
, &ins
->dst
[0]);
1814 /** Process GLSL instruction modifiers */
1815 static void shader_glsl_add_instruction_modifiers(const struct wined3d_shader_instruction
*ins
)
1817 struct glsl_dst_param dst_param
;
1820 if (!ins
->dst_count
) return;
1822 modifiers
= ins
->dst
[0].modifiers
;
1823 if (!modifiers
) return;
1825 shader_glsl_add_dst_param(ins
, &ins
->dst
[0], &dst_param
);
1827 if (modifiers
& WINED3DSPDM_SATURATE
)
1829 /* _SAT means to clamp the value of the register to between 0 and 1 */
1830 shader_addline(ins
->ctx
->buffer
, "%s%s = clamp(%s%s, 0.0, 1.0);\n", dst_param
.reg_name
,
1831 dst_param
.mask_str
, dst_param
.reg_name
, dst_param
.mask_str
);
1834 if (modifiers
& WINED3DSPDM_MSAMPCENTROID
)
1836 FIXME("_centroid modifier not handled\n");
1839 if (modifiers
& WINED3DSPDM_PARTIALPRECISION
)
1841 /* MSDN says this modifier can be safely ignored, so that's what we'll do. */
1845 static const char *shader_glsl_get_rel_op(enum wined3d_shader_rel_op op
)
1849 case WINED3D_SHADER_REL_OP_GT
: return ">";
1850 case WINED3D_SHADER_REL_OP_EQ
: return "==";
1851 case WINED3D_SHADER_REL_OP_GE
: return ">=";
1852 case WINED3D_SHADER_REL_OP_LT
: return "<";
1853 case WINED3D_SHADER_REL_OP_NE
: return "!=";
1854 case WINED3D_SHADER_REL_OP_LE
: return "<=";
1856 FIXME("Unrecognized operator %#x.\n", op
);
1861 static void shader_glsl_get_sample_function(const struct wined3d_shader_context
*ctx
,
1862 DWORD sampler_idx
, DWORD flags
, struct glsl_sample_function
*sample_function
)
1864 enum wined3d_sampler_texture_type sampler_type
= ctx
->reg_maps
->sampler_type
[sampler_idx
];
1865 const struct wined3d_gl_info
*gl_info
= ctx
->gl_info
;
1866 BOOL shadow
= ctx
->reg_maps
->shader_version
.type
== WINED3D_SHADER_TYPE_PIXEL
1867 && (((const struct shader_glsl_ctx_priv
*)ctx
->backend_data
)->cur_ps_args
->shadow
& (1 << sampler_idx
));
1868 BOOL projected
= flags
& WINED3D_GLSL_SAMPLE_PROJECTED
;
1869 BOOL texrect
= flags
& WINED3D_GLSL_SAMPLE_RECT
;
1870 BOOL lod
= flags
& WINED3D_GLSL_SAMPLE_LOD
;
1871 BOOL grad
= flags
& WINED3D_GLSL_SAMPLE_GRAD
;
1873 /* Note that there's no such thing as a projected cube texture. */
1874 switch(sampler_type
) {
1880 sample_function
->name
= projected
? "shadow1DProjLod" : "shadow1DLod";
1884 if (gl_info
->supported
[EXT_GPU_SHADER4
])
1885 sample_function
->name
= projected
? "shadow1DProjGrad" : "shadow1DGrad";
1886 else if (gl_info
->supported
[ARB_SHADER_TEXTURE_LOD
])
1887 sample_function
->name
= projected
? "shadow1DProjGradARB" : "shadow1DGradARB";
1890 FIXME("Unsupported 1D shadow grad function.\n");
1891 sample_function
->name
= "unsupported1DGrad";
1896 sample_function
->name
= projected
? "shadow1DProj" : "shadow1D";
1898 sample_function
->coord_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
;
1904 sample_function
->name
= projected
? "texture1DProjLod" : "texture1DLod";
1908 if (gl_info
->supported
[EXT_GPU_SHADER4
])
1909 sample_function
->name
= projected
? "texture1DProjGrad" : "texture1DGrad";
1910 else if (gl_info
->supported
[ARB_SHADER_TEXTURE_LOD
])
1911 sample_function
->name
= projected
? "texture1DProjGradARB" : "texture1DGradARB";
1914 FIXME("Unsupported 1D grad function.\n");
1915 sample_function
->name
= "unsupported1DGrad";
1920 sample_function
->name
= projected
? "texture1DProj" : "texture1D";
1922 sample_function
->coord_mask
= WINED3DSP_WRITEMASK_0
;
1933 sample_function
->name
= projected
? "shadow2DRectProjLod" : "shadow2DRectLod";
1937 if (gl_info
->supported
[EXT_GPU_SHADER4
])
1938 sample_function
->name
= projected
? "shadow2DRectProjGrad" : "shadow2DRectGrad";
1939 else if (gl_info
->supported
[ARB_SHADER_TEXTURE_LOD
])
1940 sample_function
->name
= projected
? "shadow2DRectProjGradARB" : "shadow2DRectGradARB";
1943 FIXME("Unsupported RECT shadow grad function.\n");
1944 sample_function
->name
= "unsupported2DRectGrad";
1949 sample_function
->name
= projected
? "shadow2DRectProj" : "shadow2DRect";
1956 sample_function
->name
= projected
? "shadow2DProjLod" : "shadow2DLod";
1960 if (gl_info
->supported
[EXT_GPU_SHADER4
])
1961 sample_function
->name
= projected
? "shadow2DProjGrad" : "shadow2DGrad";
1962 else if (gl_info
->supported
[ARB_SHADER_TEXTURE_LOD
])
1963 sample_function
->name
= projected
? "shadow2DProjGradARB" : "shadow2DGradARB";
1966 FIXME("Unsupported 2D shadow grad function.\n");
1967 sample_function
->name
= "unsupported2DGrad";
1972 sample_function
->name
= projected
? "shadow2DProj" : "shadow2D";
1975 sample_function
->coord_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
1983 sample_function
->name
= projected
? "texture2DRectProjLod" : "texture2DRectLod";
1987 if (gl_info
->supported
[EXT_GPU_SHADER4
])
1988 sample_function
->name
= projected
? "texture2DRectProjGrad" : "texture2DRectGrad";
1989 else if (gl_info
->supported
[ARB_SHADER_TEXTURE_LOD
])
1990 sample_function
->name
= projected
? "texture2DRectProjGradARB" : "texture2DRectGradARB";
1993 FIXME("Unsupported RECT grad function.\n");
1994 sample_function
->name
= "unsupported2DRectGrad";
1999 sample_function
->name
= projected
? "texture2DRectProj" : "texture2DRect";
2006 sample_function
->name
= projected
? "texture2DProjLod" : "texture2DLod";
2010 if (gl_info
->supported
[EXT_GPU_SHADER4
])
2011 sample_function
->name
= projected
? "texture2DProjGrad" : "texture2DGrad";
2012 else if (gl_info
->supported
[ARB_SHADER_TEXTURE_LOD
])
2013 sample_function
->name
= projected
? "texture2DProjGradARB" : "texture2DGradARB";
2016 FIXME("Unsupported 2D grad function.\n");
2017 sample_function
->name
= "unsupported2DGrad";
2022 sample_function
->name
= projected
? "texture2DProj" : "texture2D";
2025 sample_function
->coord_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
;
2029 case WINED3DSTT_CUBE
:
2032 FIXME("Unsupported Cube shadow function.\n");
2033 sample_function
->name
= "unsupportedCubeShadow";
2034 sample_function
->coord_mask
= 0;
2040 sample_function
->name
= "textureCubeLod";
2044 if (gl_info
->supported
[EXT_GPU_SHADER4
])
2045 sample_function
->name
= "textureCubeGrad";
2046 else if (gl_info
->supported
[ARB_SHADER_TEXTURE_LOD
])
2047 sample_function
->name
= "textureCubeGradARB";
2050 FIXME("Unsupported Cube grad function.\n");
2051 sample_function
->name
= "unsupportedCubeGrad";
2056 sample_function
->name
= "textureCube";
2058 sample_function
->coord_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2062 case WINED3DSTT_VOLUME
:
2065 FIXME("Unsupported 3D shadow function.\n");
2066 sample_function
->name
= "unsupported3DShadow";
2067 sample_function
->coord_mask
= 0;
2073 sample_function
->name
= projected
? "texture3DProjLod" : "texture3DLod";
2077 if (gl_info
->supported
[EXT_GPU_SHADER4
])
2078 sample_function
->name
= projected
? "texture3DProjGrad" : "texture3DGrad";
2079 else if (gl_info
->supported
[ARB_SHADER_TEXTURE_LOD
])
2080 sample_function
->name
= projected
? "texture3DProjGradARB" : "texture3DGradARB";
2083 FIXME("Unsupported 3D grad function.\n");
2084 sample_function
->name
= "unsupported3DGrad";
2089 sample_function
->name
= projected
? "texture3DProj" : "texture3D";
2091 sample_function
->coord_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2096 sample_function
->name
= "";
2097 sample_function
->coord_mask
= 0;
2098 FIXME("Unrecognized sampler type: %#x;\n", sampler_type
);
2103 static void shader_glsl_append_fixup_arg(char *arguments
, const char *reg_name
,
2104 BOOL sign_fixup
, enum fixup_channel_source channel_source
)
2106 switch(channel_source
)
2108 case CHANNEL_SOURCE_ZERO
:
2109 strcat(arguments
, "0.0");
2112 case CHANNEL_SOURCE_ONE
:
2113 strcat(arguments
, "1.0");
2116 case CHANNEL_SOURCE_X
:
2117 strcat(arguments
, reg_name
);
2118 strcat(arguments
, ".x");
2121 case CHANNEL_SOURCE_Y
:
2122 strcat(arguments
, reg_name
);
2123 strcat(arguments
, ".y");
2126 case CHANNEL_SOURCE_Z
:
2127 strcat(arguments
, reg_name
);
2128 strcat(arguments
, ".z");
2131 case CHANNEL_SOURCE_W
:
2132 strcat(arguments
, reg_name
);
2133 strcat(arguments
, ".w");
2137 FIXME("Unhandled channel source %#x\n", channel_source
);
2138 strcat(arguments
, "undefined");
2142 if (sign_fixup
) strcat(arguments
, " * 2.0 - 1.0");
2145 static void shader_glsl_color_correction_ext(struct wined3d_shader_buffer
*buffer
,
2146 const char *reg_name
, DWORD mask
, struct color_fixup_desc fixup
)
2148 unsigned int mask_size
, remaining
;
2149 DWORD fixup_mask
= 0;
2150 char arguments
[256];
2153 if (fixup
.x_sign_fixup
|| fixup
.x_source
!= CHANNEL_SOURCE_X
) fixup_mask
|= WINED3DSP_WRITEMASK_0
;
2154 if (fixup
.y_sign_fixup
|| fixup
.y_source
!= CHANNEL_SOURCE_Y
) fixup_mask
|= WINED3DSP_WRITEMASK_1
;
2155 if (fixup
.z_sign_fixup
|| fixup
.z_source
!= CHANNEL_SOURCE_Z
) fixup_mask
|= WINED3DSP_WRITEMASK_2
;
2156 if (fixup
.w_sign_fixup
|| fixup
.w_source
!= CHANNEL_SOURCE_W
) fixup_mask
|= WINED3DSP_WRITEMASK_3
;
2157 if (!(mask
&= fixup_mask
))
2160 if (is_complex_fixup(fixup
))
2162 enum complex_fixup complex_fixup
= get_complex_fixup(fixup
);
2163 FIXME("Complex fixup (%#x) not supported\n",complex_fixup
);
2167 shader_glsl_write_mask_to_str(mask
, mask_str
);
2168 mask_size
= shader_glsl_get_write_mask_size(mask
);
2170 arguments
[0] = '\0';
2171 remaining
= mask_size
;
2172 if (mask
& WINED3DSP_WRITEMASK_0
)
2174 shader_glsl_append_fixup_arg(arguments
, reg_name
, fixup
.x_sign_fixup
, fixup
.x_source
);
2175 if (--remaining
) strcat(arguments
, ", ");
2177 if (mask
& WINED3DSP_WRITEMASK_1
)
2179 shader_glsl_append_fixup_arg(arguments
, reg_name
, fixup
.y_sign_fixup
, fixup
.y_source
);
2180 if (--remaining
) strcat(arguments
, ", ");
2182 if (mask
& WINED3DSP_WRITEMASK_2
)
2184 shader_glsl_append_fixup_arg(arguments
, reg_name
, fixup
.z_sign_fixup
, fixup
.z_source
);
2185 if (--remaining
) strcat(arguments
, ", ");
2187 if (mask
& WINED3DSP_WRITEMASK_3
)
2189 shader_glsl_append_fixup_arg(arguments
, reg_name
, fixup
.w_sign_fixup
, fixup
.w_source
);
2190 if (--remaining
) strcat(arguments
, ", ");
2194 shader_addline(buffer
, "%s%s = vec%u(%s);\n", reg_name
, mask_str
, mask_size
, arguments
);
2196 shader_addline(buffer
, "%s%s = %s;\n", reg_name
, mask_str
, arguments
);
2199 static void shader_glsl_color_correction(const struct wined3d_shader_instruction
*ins
, struct color_fixup_desc fixup
)
2204 shader_glsl_get_register_name(&ins
->dst
[0].reg
, reg_name
, &is_color
, ins
);
2205 shader_glsl_color_correction_ext(ins
->ctx
->buffer
, reg_name
, ins
->dst
[0].write_mask
, fixup
);
2208 static void PRINTF_ATTR(8, 9) shader_glsl_gen_sample_code(const struct wined3d_shader_instruction
*ins
,
2209 DWORD sampler
, const struct glsl_sample_function
*sample_function
, DWORD swizzle
,
2210 const char *dx
, const char *dy
, const char *bias
, const char *coord_reg_fmt
, ...)
2212 const struct wined3d_shader_version
*version
= &ins
->ctx
->reg_maps
->shader_version
;
2213 char dst_swizzle
[6];
2214 struct color_fixup_desc fixup
;
2215 BOOL np2_fixup
= FALSE
;
2218 shader_glsl_swizzle_to_str(swizzle
, FALSE
, ins
->dst
[0].write_mask
, dst_swizzle
);
2220 if (version
->type
== WINED3D_SHADER_TYPE_PIXEL
)
2222 const struct shader_glsl_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2223 fixup
= priv
->cur_ps_args
->color_fixup
[sampler
];
2225 if(priv
->cur_ps_args
->np2_fixup
& (1 << sampler
)) {
2227 FIXME("Biased sampling from NP2 textures is unsupported\n");
2235 fixup
= COLOR_FIXUP_IDENTITY
; /* FIXME: Vshader color fixup */
2238 shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2240 shader_addline(ins
->ctx
->buffer
, "%s(%s_sampler%u, ",
2241 sample_function
->name
, shader_glsl_get_prefix(version
->type
), sampler
);
2243 va_start(args
, coord_reg_fmt
);
2244 shader_vaddline(ins
->ctx
->buffer
, coord_reg_fmt
, args
);
2248 shader_addline(ins
->ctx
->buffer
, ", %s)%s);\n", bias
, dst_swizzle
);
2251 const struct shader_glsl_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2252 const unsigned char idx
= priv
->cur_np2fixup_info
->idx
[sampler
];
2254 shader_addline(ins
->ctx
->buffer
, " * ps_samplerNP2Fixup[%u].%s)%s);\n", idx
>> 1,
2255 (idx
% 2) ? "zw" : "xy", dst_swizzle
);
2256 } else if(dx
&& dy
) {
2257 shader_addline(ins
->ctx
->buffer
, ", %s, %s)%s);\n", dx
, dy
, dst_swizzle
);
2259 shader_addline(ins
->ctx
->buffer
, ")%s);\n", dst_swizzle
);
2263 if(!is_identity_fixup(fixup
)) {
2264 shader_glsl_color_correction(ins
, fixup
);
2268 /*****************************************************************************
2269 * Begin processing individual instruction opcodes
2270 ****************************************************************************/
2272 static void shader_glsl_binop(const struct wined3d_shader_instruction
*ins
)
2274 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2275 struct glsl_src_param src0_param
;
2276 struct glsl_src_param src1_param
;
2280 /* Determine the GLSL operator to use based on the opcode */
2281 switch (ins
->handler_idx
)
2283 case WINED3DSIH_ADD
: op
= "+"; break;
2284 case WINED3DSIH_AND
: op
= "&"; break;
2285 case WINED3DSIH_DIV
: op
= "/"; break;
2286 case WINED3DSIH_IADD
: op
= "+"; break;
2287 case WINED3DSIH_MUL
: op
= "*"; break;
2288 case WINED3DSIH_SUB
: op
= "-"; break;
2289 case WINED3DSIH_USHR
: op
= ">>"; break;
2290 case WINED3DSIH_XOR
: op
= "^"; break;
2292 op
= "<unhandled operator>";
2293 FIXME("Opcode %#x not yet handled in GLSL\n", ins
->handler_idx
);
2297 write_mask
= shader_glsl_append_dst(buffer
, ins
);
2298 shader_glsl_add_src_param(ins
, &ins
->src
[0], write_mask
, &src0_param
);
2299 shader_glsl_add_src_param(ins
, &ins
->src
[1], write_mask
, &src1_param
);
2300 shader_addline(buffer
, "%s %s %s);\n", src0_param
.param_str
, op
, src1_param
.param_str
);
2303 static void shader_glsl_relop(const struct wined3d_shader_instruction
*ins
)
2305 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2306 struct glsl_src_param src0_param
;
2307 struct glsl_src_param src1_param
;
2308 unsigned int mask_size
;
2312 write_mask
= shader_glsl_append_dst(buffer
, ins
);
2313 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
2314 shader_glsl_add_src_param(ins
, &ins
->src
[0], write_mask
, &src0_param
);
2315 shader_glsl_add_src_param(ins
, &ins
->src
[1], write_mask
, &src1_param
);
2319 switch (ins
->handler_idx
)
2321 case WINED3DSIH_EQ
: op
= "equal"; break;
2322 case WINED3DSIH_GE
: op
= "greaterThanEqual"; break;
2323 case WINED3DSIH_IGE
: op
= "greaterThanEqual"; break;
2324 case WINED3DSIH_LT
: op
= "lessThan"; break;
2326 op
= "<unhandled operator>";
2327 ERR("Unhandled opcode %#x.\n", ins
->handler_idx
);
2331 shader_addline(buffer
, "uvec%u(%s(%s, %s)) * 0xffffffffu);\n",
2332 mask_size
, op
, src0_param
.param_str
, src1_param
.param_str
);
2336 switch (ins
->handler_idx
)
2338 case WINED3DSIH_EQ
: op
= "=="; break;
2339 case WINED3DSIH_GE
: op
= ">="; break;
2340 case WINED3DSIH_IGE
: op
= ">="; break;
2341 case WINED3DSIH_LT
: op
= "<"; break;
2343 op
= "<unhandled operator>";
2344 ERR("Unhandled opcode %#x.\n", ins
->handler_idx
);
2348 shader_addline(buffer
, "%s %s %s ? 0xffffffffu : 0u);\n",
2349 src0_param
.param_str
, op
, src1_param
.param_str
);
2353 static void shader_glsl_imul(const struct wined3d_shader_instruction
*ins
)
2355 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2356 struct glsl_src_param src0_param
;
2357 struct glsl_src_param src1_param
;
2360 /* If we have ARB_gpu_shader5 or GLSL 4.0, we can use imulExtended(). If
2361 * not, we can emulate it. */
2362 if (ins
->dst
[0].reg
.type
!= WINED3DSPR_NULL
)
2363 FIXME("64-bit integer multiplies not implemented.\n");
2365 if (ins
->dst
[1].reg
.type
!= WINED3DSPR_NULL
)
2367 write_mask
= shader_glsl_append_dst_ext(buffer
, ins
, &ins
->dst
[1]);
2368 shader_glsl_add_src_param(ins
, &ins
->src
[0], write_mask
, &src0_param
);
2369 shader_glsl_add_src_param(ins
, &ins
->src
[1], write_mask
, &src1_param
);
2371 shader_addline(ins
->ctx
->buffer
, "%s * %s);\n",
2372 src0_param
.param_str
, src1_param
.param_str
);
2376 static void shader_glsl_udiv(const struct wined3d_shader_instruction
*ins
)
2378 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2379 struct glsl_src_param src0_param
, src1_param
;
2382 if (ins
->dst
[0].reg
.type
!= WINED3DSPR_NULL
)
2385 if (ins
->dst
[1].reg
.type
!= WINED3DSPR_NULL
)
2389 write_mask
= shader_glsl_get_write_mask(&ins
->dst
[0], dst_mask
);
2390 shader_glsl_add_src_param(ins
, &ins
->src
[0], write_mask
, &src0_param
);
2391 shader_glsl_add_src_param(ins
, &ins
->src
[1], write_mask
, &src1_param
);
2392 shader_addline(buffer
, "tmp0%s = %s / %s;\n",
2393 dst_mask
, src0_param
.param_str
, src1_param
.param_str
);
2395 write_mask
= shader_glsl_append_dst_ext(buffer
, ins
, &ins
->dst
[1]);
2396 shader_glsl_add_src_param(ins
, &ins
->src
[0], write_mask
, &src0_param
);
2397 shader_glsl_add_src_param(ins
, &ins
->src
[1], write_mask
, &src1_param
);
2398 shader_addline(buffer
, "%s %% %s));\n", src0_param
.param_str
, src1_param
.param_str
);
2400 shader_glsl_append_dst_ext(buffer
, ins
, &ins
->dst
[0]);
2401 shader_addline(buffer
, "tmp0%s);\n", dst_mask
);
2405 write_mask
= shader_glsl_append_dst_ext(buffer
, ins
, &ins
->dst
[0]);
2406 shader_glsl_add_src_param(ins
, &ins
->src
[0], write_mask
, &src0_param
);
2407 shader_glsl_add_src_param(ins
, &ins
->src
[1], write_mask
, &src1_param
);
2408 shader_addline(buffer
, "%s / %s);\n", src0_param
.param_str
, src1_param
.param_str
);
2411 else if (ins
->dst
[1].reg
.type
!= WINED3DSPR_NULL
)
2413 write_mask
= shader_glsl_append_dst_ext(buffer
, ins
, &ins
->dst
[1]);
2414 shader_glsl_add_src_param(ins
, &ins
->src
[0], write_mask
, &src0_param
);
2415 shader_glsl_add_src_param(ins
, &ins
->src
[1], write_mask
, &src1_param
);
2416 shader_addline(buffer
, "%s %% %s);\n", src0_param
.param_str
, src1_param
.param_str
);
2420 /* Process the WINED3DSIO_MOV opcode using GLSL (dst = src) */
2421 static void shader_glsl_mov(const struct wined3d_shader_instruction
*ins
)
2423 const struct wined3d_gl_info
*gl_info
= ins
->ctx
->gl_info
;
2424 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2425 struct glsl_src_param src0_param
;
2428 write_mask
= shader_glsl_append_dst(buffer
, ins
);
2429 shader_glsl_add_src_param(ins
, &ins
->src
[0], write_mask
, &src0_param
);
2431 /* In vs_1_1 WINED3DSIO_MOV can write to the address register. In later
2432 * shader versions WINED3DSIO_MOVA is used for this. */
2433 if (ins
->ctx
->reg_maps
->shader_version
.major
== 1
2434 && ins
->ctx
->reg_maps
->shader_version
.type
== WINED3D_SHADER_TYPE_VERTEX
2435 && ins
->dst
[0].reg
.type
== WINED3DSPR_ADDR
)
2437 /* This is a simple floor() */
2438 unsigned int mask_size
= shader_glsl_get_write_mask_size(write_mask
);
2439 if (mask_size
> 1) {
2440 shader_addline(buffer
, "ivec%d(floor(%s)));\n", mask_size
, src0_param
.param_str
);
2442 shader_addline(buffer
, "int(floor(%s)));\n", src0_param
.param_str
);
2445 else if(ins
->handler_idx
== WINED3DSIH_MOVA
)
2447 /* We need to *round* to the nearest int here. */
2448 unsigned int mask_size
= shader_glsl_get_write_mask_size(write_mask
);
2450 if (gl_info
->supported
[EXT_GPU_SHADER4
])
2453 shader_addline(buffer
, "ivec%d(round(%s)));\n", mask_size
, src0_param
.param_str
);
2455 shader_addline(buffer
, "int(round(%s)));\n", src0_param
.param_str
);
2460 shader_addline(buffer
, "ivec%d(floor(abs(%s) + vec%d(0.5)) * sign(%s)));\n",
2461 mask_size
, src0_param
.param_str
, mask_size
, src0_param
.param_str
);
2463 shader_addline(buffer
, "int(floor(abs(%s) + 0.5) * sign(%s)));\n",
2464 src0_param
.param_str
, src0_param
.param_str
);
2469 shader_addline(buffer
, "%s);\n", src0_param
.param_str
);
2473 /* Process the dot product operators DP3 and DP4 in GLSL (dst = dot(src0, src1)) */
2474 static void shader_glsl_dot(const struct wined3d_shader_instruction
*ins
)
2476 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2477 struct glsl_src_param src0_param
;
2478 struct glsl_src_param src1_param
;
2479 DWORD dst_write_mask
, src_write_mask
;
2480 unsigned int dst_size
= 0;
2482 dst_write_mask
= shader_glsl_append_dst(buffer
, ins
);
2483 dst_size
= shader_glsl_get_write_mask_size(dst_write_mask
);
2485 /* dp3 works on vec3, dp4 on vec4 */
2486 if (ins
->handler_idx
== WINED3DSIH_DP4
)
2488 src_write_mask
= WINED3DSP_WRITEMASK_ALL
;
2490 src_write_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2493 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_write_mask
, &src0_param
);
2494 shader_glsl_add_src_param(ins
, &ins
->src
[1], src_write_mask
, &src1_param
);
2497 shader_addline(buffer
, "vec%d(dot(%s, %s)));\n", dst_size
, src0_param
.param_str
, src1_param
.param_str
);
2499 shader_addline(buffer
, "dot(%s, %s));\n", src0_param
.param_str
, src1_param
.param_str
);
2503 /* Note that this instruction has some restrictions. The destination write mask
2504 * can't contain the w component, and the source swizzles have to be .xyzw */
2505 static void shader_glsl_cross(const struct wined3d_shader_instruction
*ins
)
2507 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2508 struct glsl_src_param src0_param
;
2509 struct glsl_src_param src1_param
;
2512 shader_glsl_get_write_mask(&ins
->dst
[0], dst_mask
);
2513 shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2514 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
2515 shader_glsl_add_src_param(ins
, &ins
->src
[1], src_mask
, &src1_param
);
2516 shader_addline(ins
->ctx
->buffer
, "cross(%s, %s)%s);\n", src0_param
.param_str
, src1_param
.param_str
, dst_mask
);
2519 static void shader_glsl_cut(const struct wined3d_shader_instruction
*ins
)
2521 shader_addline(ins
->ctx
->buffer
, "EndPrimitive();\n");
2524 /* Process the WINED3DSIO_POW instruction in GLSL (dst = |src0|^src1)
2525 * Src0 and src1 are scalars. Note that D3D uses the absolute of src0, while
2526 * GLSL uses the value as-is. */
2527 static void shader_glsl_pow(const struct wined3d_shader_instruction
*ins
)
2529 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2530 struct glsl_src_param src0_param
;
2531 struct glsl_src_param src1_param
;
2532 DWORD dst_write_mask
;
2533 unsigned int dst_size
;
2535 dst_write_mask
= shader_glsl_append_dst(buffer
, ins
);
2536 dst_size
= shader_glsl_get_write_mask_size(dst_write_mask
);
2538 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2539 shader_glsl_add_src_param(ins
, &ins
->src
[1], WINED3DSP_WRITEMASK_0
, &src1_param
);
2543 shader_addline(buffer
, "vec%u(%s == 0.0 ? 1.0 : pow(abs(%s), %s)));\n",
2544 dst_size
, src1_param
.param_str
, src0_param
.param_str
, src1_param
.param_str
);
2548 shader_addline(buffer
, "%s == 0.0 ? 1.0 : pow(abs(%s), %s));\n",
2549 src1_param
.param_str
, src0_param
.param_str
, src1_param
.param_str
);
2553 /* Process the WINED3DSIO_LOG instruction in GLSL (dst = log2(|src0|))
2554 * Src0 is a scalar. Note that D3D uses the absolute of src0, while
2555 * GLSL uses the value as-is. */
2556 static void shader_glsl_log(const struct wined3d_shader_instruction
*ins
)
2558 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2559 struct glsl_src_param src0_param
;
2560 DWORD dst_write_mask
;
2561 unsigned int dst_size
;
2563 dst_write_mask
= shader_glsl_append_dst(buffer
, ins
);
2564 dst_size
= shader_glsl_get_write_mask_size(dst_write_mask
);
2566 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2570 shader_addline(buffer
, "vec%u(log2(abs(%s))));\n",
2571 dst_size
, src0_param
.param_str
);
2575 shader_addline(buffer
, "log2(abs(%s)));\n",
2576 src0_param
.param_str
);
2580 /* Map the opcode 1-to-1 to the GL code (arg->dst = instruction(src0, src1, ...) */
2581 static void shader_glsl_map2gl(const struct wined3d_shader_instruction
*ins
)
2583 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2584 struct glsl_src_param src_param
;
2585 const char *instruction
;
2589 /* Determine the GLSL function to use based on the opcode */
2590 /* TODO: Possibly make this a table for faster lookups */
2591 switch (ins
->handler_idx
)
2593 case WINED3DSIH_MIN
: instruction
= "min"; break;
2594 case WINED3DSIH_MAX
: instruction
= "max"; break;
2595 case WINED3DSIH_ABS
: instruction
= "abs"; break;
2596 case WINED3DSIH_FRC
: instruction
= "fract"; break;
2597 case WINED3DSIH_EXP
: instruction
= "exp2"; break;
2598 case WINED3DSIH_DSX
: instruction
= "dFdx"; break;
2599 case WINED3DSIH_DSY
: instruction
= "ycorrection.y * dFdy"; break;
2600 case WINED3DSIH_ROUND_NI
: instruction
= "floor"; break;
2601 default: instruction
= "";
2602 FIXME("Opcode %#x not yet handled in GLSL\n", ins
->handler_idx
);
2606 write_mask
= shader_glsl_append_dst(buffer
, ins
);
2608 shader_addline(buffer
, "%s(", instruction
);
2612 shader_glsl_add_src_param(ins
, &ins
->src
[0], write_mask
, &src_param
);
2613 shader_addline(buffer
, "%s", src_param
.param_str
);
2614 for (i
= 1; i
< ins
->src_count
; ++i
)
2616 shader_glsl_add_src_param(ins
, &ins
->src
[i
], write_mask
, &src_param
);
2617 shader_addline(buffer
, ", %s", src_param
.param_str
);
2621 shader_addline(buffer
, "));\n");
2624 static void shader_glsl_nop(const struct wined3d_shader_instruction
*ins
) {}
2626 static void shader_glsl_nrm(const struct wined3d_shader_instruction
*ins
)
2628 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2629 struct glsl_src_param src_param
;
2630 unsigned int mask_size
;
2634 write_mask
= shader_glsl_get_write_mask(ins
->dst
, dst_mask
);
2635 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
2636 shader_glsl_add_src_param(ins
, &ins
->src
[0], write_mask
, &src_param
);
2638 shader_addline(buffer
, "tmp0.x = dot(%s, %s);\n",
2639 src_param
.param_str
, src_param
.param_str
);
2640 shader_glsl_append_dst(buffer
, ins
);
2644 shader_addline(buffer
, "tmp0.x == 0.0 ? vec%u(0.0) : (%s * inversesqrt(tmp0.x)));\n",
2645 mask_size
, src_param
.param_str
);
2649 shader_addline(buffer
, "tmp0.x == 0.0 ? 0.0 : (%s * inversesqrt(tmp0.x)));\n",
2650 src_param
.param_str
);
2654 /** Process the WINED3DSIO_EXPP instruction in GLSL:
2655 * For shader model 1.x, do the following (and honor the writemask, so use a temporary variable):
2656 * dst.x = 2^(floor(src))
2657 * dst.y = src - floor(src)
2658 * dst.z = 2^src (partial precision is allowed, but optional)
2660 * For 2.0 shaders, just do this (honoring writemask and swizzle):
2661 * dst = 2^src; (partial precision is allowed, but optional)
2663 static void shader_glsl_expp(const struct wined3d_shader_instruction
*ins
)
2665 struct glsl_src_param src_param
;
2667 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
, &src_param
);
2669 if (ins
->ctx
->reg_maps
->shader_version
.major
< 2)
2673 shader_addline(ins
->ctx
->buffer
, "tmp0.x = exp2(floor(%s));\n", src_param
.param_str
);
2674 shader_addline(ins
->ctx
->buffer
, "tmp0.y = %s - floor(%s);\n", src_param
.param_str
, src_param
.param_str
);
2675 shader_addline(ins
->ctx
->buffer
, "tmp0.z = exp2(%s);\n", src_param
.param_str
);
2676 shader_addline(ins
->ctx
->buffer
, "tmp0.w = 1.0;\n");
2678 shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2679 shader_glsl_get_write_mask(&ins
->dst
[0], dst_mask
);
2680 shader_addline(ins
->ctx
->buffer
, "tmp0%s);\n", dst_mask
);
2683 unsigned int mask_size
;
2685 write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2686 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
2688 if (mask_size
> 1) {
2689 shader_addline(ins
->ctx
->buffer
, "vec%d(exp2(%s)));\n", mask_size
, src_param
.param_str
);
2691 shader_addline(ins
->ctx
->buffer
, "exp2(%s));\n", src_param
.param_str
);
2696 static void shader_glsl_to_int(const struct wined3d_shader_instruction
*ins
)
2698 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2699 struct glsl_src_param src_param
;
2700 unsigned int mask_size
;
2703 write_mask
= shader_glsl_append_dst(buffer
, ins
);
2704 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
2705 shader_glsl_add_src_param(ins
, &ins
->src
[0], write_mask
, &src_param
);
2708 shader_addline(buffer
, "ivec%u(%s));\n", mask_size
, src_param
.param_str
);
2710 shader_addline(buffer
, "int(%s));\n", src_param
.param_str
);
2713 static void shader_glsl_to_float(const struct wined3d_shader_instruction
*ins
)
2715 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2716 struct glsl_src_param src_param
;
2717 unsigned int mask_size
;
2720 write_mask
= shader_glsl_append_dst(buffer
, ins
);
2721 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
2722 shader_glsl_add_src_param(ins
, &ins
->src
[0], write_mask
, &src_param
);
2725 shader_addline(buffer
, "vec%u(%s));\n", mask_size
, src_param
.param_str
);
2727 shader_addline(buffer
, "float(%s));\n", src_param
.param_str
);
2730 /** Process the RCP (reciprocal or inverse) opcode in GLSL (dst = 1 / src) */
2731 static void shader_glsl_rcp(const struct wined3d_shader_instruction
*ins
)
2733 struct glsl_src_param src_param
;
2735 unsigned int mask_size
;
2737 write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2738 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
2739 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_3
, &src_param
);
2743 shader_addline(ins
->ctx
->buffer
, "vec%u(1.0 / %s));\n",
2744 mask_size
, src_param
.param_str
);
2748 shader_addline(ins
->ctx
->buffer
, "1.0 / %s);\n",
2749 src_param
.param_str
);
2753 static void shader_glsl_rsq(const struct wined3d_shader_instruction
*ins
)
2755 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2756 struct glsl_src_param src_param
;
2758 unsigned int mask_size
;
2760 write_mask
= shader_glsl_append_dst(buffer
, ins
);
2761 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
2763 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_3
, &src_param
);
2767 shader_addline(buffer
, "vec%u(inversesqrt(abs(%s))));\n",
2768 mask_size
, src_param
.param_str
);
2772 shader_addline(buffer
, "inversesqrt(abs(%s)));\n",
2773 src_param
.param_str
);
2777 /** Process signed comparison opcodes in GLSL. */
2778 static void shader_glsl_compare(const struct wined3d_shader_instruction
*ins
)
2780 struct glsl_src_param src0_param
;
2781 struct glsl_src_param src1_param
;
2783 unsigned int mask_size
;
2785 write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2786 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
2787 shader_glsl_add_src_param(ins
, &ins
->src
[0], write_mask
, &src0_param
);
2788 shader_glsl_add_src_param(ins
, &ins
->src
[1], write_mask
, &src1_param
);
2790 if (mask_size
> 1) {
2791 const char *compare
;
2793 switch(ins
->handler_idx
)
2795 case WINED3DSIH_SLT
: compare
= "lessThan"; break;
2796 case WINED3DSIH_SGE
: compare
= "greaterThanEqual"; break;
2797 default: compare
= "";
2798 FIXME("Can't handle opcode %#x\n", ins
->handler_idx
);
2801 shader_addline(ins
->ctx
->buffer
, "vec%d(%s(%s, %s)));\n", mask_size
, compare
,
2802 src0_param
.param_str
, src1_param
.param_str
);
2804 switch(ins
->handler_idx
)
2806 case WINED3DSIH_SLT
:
2807 /* Step(src0, src1) is not suitable here because if src0 == src1 SLT is supposed,
2808 * to return 0.0 but step returns 1.0 because step is not < x
2809 * An alternative is a bvec compare padded with an unused second component.
2810 * step(src1 * -1.0, src0 * -1.0) is not an option because it suffers from the same
2811 * issue. Playing with not() is not possible either because not() does not accept
2814 shader_addline(ins
->ctx
->buffer
, "(%s < %s) ? 1.0 : 0.0);\n",
2815 src0_param
.param_str
, src1_param
.param_str
);
2817 case WINED3DSIH_SGE
:
2818 /* Here we can use the step() function and safe a conditional */
2819 shader_addline(ins
->ctx
->buffer
, "step(%s, %s));\n", src1_param
.param_str
, src0_param
.param_str
);
2822 FIXME("Can't handle opcode %#x\n", ins
->handler_idx
);
2828 static void shader_glsl_conditional_move(const struct wined3d_shader_instruction
*ins
)
2830 const char *condition_prefix
, *condition_suffix
;
2831 struct wined3d_shader_dst_param dst
;
2832 struct glsl_src_param src0_param
;
2833 struct glsl_src_param src1_param
;
2834 struct glsl_src_param src2_param
;
2835 BOOL temp_destination
= FALSE
;
2836 DWORD cmp_channel
= 0;
2841 switch (ins
->handler_idx
)
2843 case WINED3DSIH_CMP
:
2844 condition_prefix
= "";
2845 condition_suffix
= " >= 0.0";
2848 case WINED3DSIH_CND
:
2849 condition_prefix
= "";
2850 condition_suffix
= " > 0.5";
2853 case WINED3DSIH_MOVC
:
2854 condition_prefix
= "bool(";
2855 condition_suffix
= ")";
2859 FIXME("Unhandled instruction %#x.\n", ins
->handler_idx
);
2860 condition_prefix
= "<unhandled prefix>";
2861 condition_suffix
= "<unhandled suffix>";
2865 if (shader_is_scalar(&ins
->dst
[0].reg
) || shader_is_scalar(&ins
->src
[0].reg
))
2867 write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2868 shader_glsl_add_src_param(ins
, &ins
->src
[0], write_mask
, &src0_param
);
2869 shader_glsl_add_src_param(ins
, &ins
->src
[1], write_mask
, &src1_param
);
2870 shader_glsl_add_src_param(ins
, &ins
->src
[2], write_mask
, &src2_param
);
2872 shader_addline(ins
->ctx
->buffer
, "%s%s%s ? %s : %s);\n",
2873 condition_prefix
, src0_param
.param_str
, condition_suffix
,
2874 src1_param
.param_str
, src2_param
.param_str
);
2880 /* Splitting the instruction up in multiple lines imposes a problem:
2881 * The first lines may overwrite source parameters of the following lines.
2882 * Deal with that by using a temporary destination register if needed. */
2883 if ((ins
->src
[0].reg
.idx
[0].offset
== dst
.reg
.idx
[0].offset
2884 && ins
->src
[0].reg
.type
== dst
.reg
.type
)
2885 || (ins
->src
[1].reg
.idx
[0].offset
== dst
.reg
.idx
[0].offset
2886 && ins
->src
[1].reg
.type
== dst
.reg
.type
)
2887 || (ins
->src
[2].reg
.idx
[0].offset
== dst
.reg
.idx
[0].offset
2888 && ins
->src
[2].reg
.type
== dst
.reg
.type
))
2889 temp_destination
= TRUE
;
2891 /* Cycle through all source0 channels. */
2892 for (i
= 0; i
< 4; ++i
)
2895 /* Find the destination channels which use the current source0 channel. */
2896 for (j
= 0; j
< 4; ++j
)
2898 if (((ins
->src
[0].swizzle
>> (2 * j
)) & 0x3) == i
)
2900 write_mask
|= WINED3DSP_WRITEMASK_0
<< j
;
2901 cmp_channel
= WINED3DSP_WRITEMASK_0
<< j
;
2904 dst
.write_mask
= ins
->dst
[0].write_mask
& write_mask
;
2906 if (temp_destination
)
2908 if (!(write_mask
= shader_glsl_get_write_mask(&dst
, mask_char
)))
2910 shader_addline(ins
->ctx
->buffer
, "tmp0%s = (", mask_char
);
2912 else if (!(write_mask
= shader_glsl_append_dst_ext(ins
->ctx
->buffer
, ins
, &dst
)))
2915 shader_glsl_add_src_param(ins
, &ins
->src
[0], cmp_channel
, &src0_param
);
2916 shader_glsl_add_src_param(ins
, &ins
->src
[1], write_mask
, &src1_param
);
2917 shader_glsl_add_src_param(ins
, &ins
->src
[2], write_mask
, &src2_param
);
2919 shader_addline(ins
->ctx
->buffer
, "%s%s%s ? %s : %s);\n",
2920 condition_prefix
, src0_param
.param_str
, condition_suffix
,
2921 src1_param
.param_str
, src2_param
.param_str
);
2924 if (temp_destination
)
2926 shader_glsl_get_write_mask(&ins
->dst
[0], mask_char
);
2927 shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2928 shader_addline(ins
->ctx
->buffer
, "tmp0%s);\n", mask_char
);
2932 /** Process the CND opcode in GLSL (dst = (src0 > 0.5) ? src1 : src2) */
2933 /* For ps 1.1-1.3, only a single component of src0 is used. For ps 1.4
2934 * the compare is done per component of src0. */
2935 static void shader_glsl_cnd(const struct wined3d_shader_instruction
*ins
)
2937 struct glsl_src_param src0_param
;
2938 struct glsl_src_param src1_param
;
2939 struct glsl_src_param src2_param
;
2941 DWORD shader_version
= WINED3D_SHADER_VERSION(ins
->ctx
->reg_maps
->shader_version
.major
,
2942 ins
->ctx
->reg_maps
->shader_version
.minor
);
2944 if (shader_version
< WINED3D_SHADER_VERSION(1, 4))
2946 write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2947 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2948 shader_glsl_add_src_param(ins
, &ins
->src
[1], write_mask
, &src1_param
);
2949 shader_glsl_add_src_param(ins
, &ins
->src
[2], write_mask
, &src2_param
);
2951 /* Fun: The D3DSI_COISSUE flag changes the semantic of the cnd instruction for < 1.4 shaders */
2954 shader_addline(ins
->ctx
->buffer
, "%s /* COISSUE! */);\n", src1_param
.param_str
);
2956 shader_addline(ins
->ctx
->buffer
, "%s > 0.5 ? %s : %s);\n",
2957 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
2962 shader_glsl_conditional_move(ins
);
2965 /** GLSL code generation for WINED3DSIO_MAD: Multiply the first 2 opcodes, then add the last */
2966 static void shader_glsl_mad(const struct wined3d_shader_instruction
*ins
)
2968 struct glsl_src_param src0_param
;
2969 struct glsl_src_param src1_param
;
2970 struct glsl_src_param src2_param
;
2973 write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2974 shader_glsl_add_src_param(ins
, &ins
->src
[0], write_mask
, &src0_param
);
2975 shader_glsl_add_src_param(ins
, &ins
->src
[1], write_mask
, &src1_param
);
2976 shader_glsl_add_src_param(ins
, &ins
->src
[2], write_mask
, &src2_param
);
2977 shader_addline(ins
->ctx
->buffer
, "(%s * %s) + %s);\n",
2978 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
2981 /* Handles transforming all WINED3DSIO_M?x? opcodes for
2982 Vertex shaders to GLSL codes */
2983 static void shader_glsl_mnxn(const struct wined3d_shader_instruction
*ins
)
2986 int nComponents
= 0;
2987 struct wined3d_shader_dst_param tmp_dst
= {{0}};
2988 struct wined3d_shader_src_param tmp_src
[2] = {{{0}}};
2989 struct wined3d_shader_instruction tmp_ins
;
2991 memset(&tmp_ins
, 0, sizeof(tmp_ins
));
2993 /* Set constants for the temporary argument */
2994 tmp_ins
.ctx
= ins
->ctx
;
2995 tmp_ins
.dst_count
= 1;
2996 tmp_ins
.dst
= &tmp_dst
;
2997 tmp_ins
.src_count
= 2;
2998 tmp_ins
.src
= tmp_src
;
3000 switch(ins
->handler_idx
)
3002 case WINED3DSIH_M4x4
:
3004 tmp_ins
.handler_idx
= WINED3DSIH_DP4
;
3006 case WINED3DSIH_M4x3
:
3008 tmp_ins
.handler_idx
= WINED3DSIH_DP4
;
3010 case WINED3DSIH_M3x4
:
3012 tmp_ins
.handler_idx
= WINED3DSIH_DP3
;
3014 case WINED3DSIH_M3x3
:
3016 tmp_ins
.handler_idx
= WINED3DSIH_DP3
;
3018 case WINED3DSIH_M3x2
:
3020 tmp_ins
.handler_idx
= WINED3DSIH_DP3
;
3026 tmp_dst
= ins
->dst
[0];
3027 tmp_src
[0] = ins
->src
[0];
3028 tmp_src
[1] = ins
->src
[1];
3029 for (i
= 0; i
< nComponents
; ++i
)
3031 tmp_dst
.write_mask
= WINED3DSP_WRITEMASK_0
<< i
;
3032 shader_glsl_dot(&tmp_ins
);
3033 ++tmp_src
[1].reg
.idx
[0].offset
;
3038 The LRP instruction performs a component-wise linear interpolation
3039 between the second and third operands using the first operand as the
3040 blend factor. Equation: (dst = src2 + src0 * (src1 - src2))
3041 This is equivalent to mix(src2, src1, src0);
3043 static void shader_glsl_lrp(const struct wined3d_shader_instruction
*ins
)
3045 struct glsl_src_param src0_param
;
3046 struct glsl_src_param src1_param
;
3047 struct glsl_src_param src2_param
;
3050 write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
3052 shader_glsl_add_src_param(ins
, &ins
->src
[0], write_mask
, &src0_param
);
3053 shader_glsl_add_src_param(ins
, &ins
->src
[1], write_mask
, &src1_param
);
3054 shader_glsl_add_src_param(ins
, &ins
->src
[2], write_mask
, &src2_param
);
3056 shader_addline(ins
->ctx
->buffer
, "mix(%s, %s, %s));\n",
3057 src2_param
.param_str
, src1_param
.param_str
, src0_param
.param_str
);
3060 /** Process the WINED3DSIO_LIT instruction in GLSL:
3061 * dst.x = dst.w = 1.0
3062 * dst.y = (src0.x > 0) ? src0.x
3063 * dst.z = (src0.x > 0) ? ((src0.y > 0) ? pow(src0.y, src.w) : 0) : 0
3064 * where src.w is clamped at +- 128
3066 static void shader_glsl_lit(const struct wined3d_shader_instruction
*ins
)
3068 struct glsl_src_param src0_param
;
3069 struct glsl_src_param src1_param
;
3070 struct glsl_src_param src3_param
;
3073 shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
3074 shader_glsl_get_write_mask(&ins
->dst
[0], dst_mask
);
3076 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
3077 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_1
, &src1_param
);
3078 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_3
, &src3_param
);
3080 /* The sdk specifies the instruction like this
3082 * if(src.x > 0.0) dst.y = src.x
3084 * if(src.x > 0.0 && src.y > 0.0) dst.z = pow(src.y, power);
3087 * (where power = src.w clamped between -128 and 128)
3089 * Obviously that has quite a few conditionals in it which we don't like. So the first step is this:
3090 * dst.x = 1.0 ... No further explanation needed
3091 * dst.y = max(src.y, 0.0); ... If x < 0.0, use 0.0, otherwise x. Same as the conditional
3092 * dst.z = x > 0.0 ? pow(max(y, 0.0), p) : 0; ... 0 ^ power is 0, and otherwise we use y anyway
3093 * dst.w = 1.0. ... Nothing fancy.
3095 * So we still have one conditional in there. So do this:
3096 * dst.z = pow(max(0.0, src.y) * step(0.0, src.x), power);
3098 * 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),
3099 * 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.
3100 * if both x and y are > 0, we get pow(y * 1.0, power), as it is supposed to.
3102 * Unfortunately pow(0.0 ^ 0.0) returns NaN on most GPUs, but lit with src.y = 0 and src.w = 0 returns
3103 * a non-NaN value in dst.z. What we return doesn't matter, as long as it is not NaN. Return 0, which is
3104 * what all Windows HW drivers and GL_ARB_vertex_program's LIT do.
3106 shader_addline(ins
->ctx
->buffer
,
3107 "vec4(1.0, max(%s, 0.0), %s == 0.0 ? 0.0 : "
3108 "pow(max(0.0, %s) * step(0.0, %s), clamp(%s, -128.0, 128.0)), 1.0)%s);\n",
3109 src0_param
.param_str
, src3_param
.param_str
, src1_param
.param_str
,
3110 src0_param
.param_str
, src3_param
.param_str
, dst_mask
);
3113 /** Process the WINED3DSIO_DST instruction in GLSL:
3115 * dst.y = src0.x * src0.y
3119 static void shader_glsl_dst(const struct wined3d_shader_instruction
*ins
)
3121 struct glsl_src_param src0y_param
;
3122 struct glsl_src_param src0z_param
;
3123 struct glsl_src_param src1y_param
;
3124 struct glsl_src_param src1w_param
;
3127 shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
3128 shader_glsl_get_write_mask(&ins
->dst
[0], dst_mask
);
3130 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_1
, &src0y_param
);
3131 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_2
, &src0z_param
);
3132 shader_glsl_add_src_param(ins
, &ins
->src
[1], WINED3DSP_WRITEMASK_1
, &src1y_param
);
3133 shader_glsl_add_src_param(ins
, &ins
->src
[1], WINED3DSP_WRITEMASK_3
, &src1w_param
);
3135 shader_addline(ins
->ctx
->buffer
, "vec4(1.0, %s * %s, %s, %s))%s;\n",
3136 src0y_param
.param_str
, src1y_param
.param_str
, src0z_param
.param_str
, src1w_param
.param_str
, dst_mask
);
3139 /** Process the WINED3DSIO_SINCOS instruction in GLSL:
3140 * VS 2.0 requires that specific cosine and sine constants be passed to this instruction so the hardware
3141 * can handle it. But, these functions are built-in for GLSL, so we can just ignore the last 2 params.
3143 * dst.x = cos(src0.?)
3144 * dst.y = sin(src0.?)
3148 static void shader_glsl_sincos(const struct wined3d_shader_instruction
*ins
)
3150 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
3151 struct glsl_src_param src0_param
;
3154 if (ins
->ctx
->reg_maps
->shader_version
.major
< 4)
3156 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
3158 write_mask
= shader_glsl_append_dst(buffer
, ins
);
3161 case WINED3DSP_WRITEMASK_0
:
3162 shader_addline(buffer
, "cos(%s));\n", src0_param
.param_str
);
3165 case WINED3DSP_WRITEMASK_1
:
3166 shader_addline(buffer
, "sin(%s));\n", src0_param
.param_str
);
3169 case (WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
):
3170 shader_addline(buffer
, "vec2(cos(%s), sin(%s)));\n",
3171 src0_param
.param_str
, src0_param
.param_str
);
3175 ERR("Write mask should be .x, .y or .xy\n");
3182 if (ins
->dst
[0].reg
.type
!= WINED3DSPR_NULL
)
3185 if (ins
->dst
[1].reg
.type
!= WINED3DSPR_NULL
)
3189 write_mask
= shader_glsl_get_write_mask(&ins
->dst
[0], dst_mask
);
3190 shader_glsl_add_src_param(ins
, &ins
->src
[0], write_mask
, &src0_param
);
3191 shader_addline(buffer
, "tmp0%s = sin(%s);\n", dst_mask
, src0_param
.param_str
);
3193 write_mask
= shader_glsl_append_dst_ext(buffer
, ins
, &ins
->dst
[1]);
3194 shader_glsl_add_src_param(ins
, &ins
->src
[0], write_mask
, &src0_param
);
3195 shader_addline(buffer
, "cos(%s));\n", src0_param
.param_str
);
3197 shader_glsl_append_dst_ext(buffer
, ins
, &ins
->dst
[0]);
3198 shader_addline(buffer
, "tmp0%s);\n", dst_mask
);
3202 write_mask
= shader_glsl_append_dst_ext(buffer
, ins
, &ins
->dst
[0]);
3203 shader_glsl_add_src_param(ins
, &ins
->src
[0], write_mask
, &src0_param
);
3204 shader_addline(buffer
, "sin(%s));\n", src0_param
.param_str
);
3207 else if (ins
->dst
[1].reg
.type
!= WINED3DSPR_NULL
)
3209 write_mask
= shader_glsl_append_dst_ext(buffer
, ins
, &ins
->dst
[1]);
3210 shader_glsl_add_src_param(ins
, &ins
->src
[0], write_mask
, &src0_param
);
3211 shader_addline(buffer
, "cos(%s));\n", src0_param
.param_str
);
3215 /* sgn in vs_2_0 has 2 extra parameters(registers for temporary storage) which we don't use
3216 * here. But those extra parameters require a dedicated function for sgn, since map2gl would
3217 * generate invalid code
3219 static void shader_glsl_sgn(const struct wined3d_shader_instruction
*ins
)
3221 struct glsl_src_param src0_param
;
3224 write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
3225 shader_glsl_add_src_param(ins
, &ins
->src
[0], write_mask
, &src0_param
);
3227 shader_addline(ins
->ctx
->buffer
, "sign(%s));\n", src0_param
.param_str
);
3230 /** Process the WINED3DSIO_LOOP instruction in GLSL:
3231 * Start a for() loop where src1.y is the initial value of aL,
3232 * increment aL by src1.z for a total of src1.x iterations.
3233 * Need to use a temporary variable for this operation.
3235 /* FIXME: I don't think nested loops will work correctly this way. */
3236 static void shader_glsl_loop(const struct wined3d_shader_instruction
*ins
)
3238 struct wined3d_shader_loop_state
*loop_state
= ins
->ctx
->loop_state
;
3239 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
3240 const struct wined3d_shader
*shader
= ins
->ctx
->shader
;
3241 const struct wined3d_shader_lconst
*constant
;
3242 struct glsl_src_param src1_param
;
3243 const DWORD
*control_values
= NULL
;
3245 if (ins
->ctx
->reg_maps
->shader_version
.major
< 4)
3247 shader_glsl_add_src_param(ins
, &ins
->src
[1], WINED3DSP_WRITEMASK_ALL
, &src1_param
);
3249 /* Try to hardcode the loop control parameters if possible. Direct3D 9
3250 * class hardware doesn't support real varying indexing, but Microsoft
3251 * designed this feature for Shader model 2.x+. If the loop control is
3252 * known at compile time, the GLSL compiler can unroll the loop, and
3253 * replace indirect addressing with direct addressing. */
3254 if (ins
->src
[1].reg
.type
== WINED3DSPR_CONSTINT
)
3256 LIST_FOR_EACH_ENTRY(constant
, &shader
->constantsI
, struct wined3d_shader_lconst
, entry
)
3258 if (constant
->idx
== ins
->src
[1].reg
.idx
[0].offset
)
3260 control_values
= constant
->value
;
3268 struct wined3d_shader_loop_control loop_control
;
3269 loop_control
.count
= control_values
[0];
3270 loop_control
.start
= control_values
[1];
3271 loop_control
.step
= (int)control_values
[2];
3273 if (loop_control
.step
> 0)
3275 shader_addline(buffer
, "for (aL%u = %u; aL%u < (%u * %d + %u); aL%u += %d)\n{\n",
3276 loop_state
->current_depth
, loop_control
.start
,
3277 loop_state
->current_depth
, loop_control
.count
, loop_control
.step
, loop_control
.start
,
3278 loop_state
->current_depth
, loop_control
.step
);
3280 else if (loop_control
.step
< 0)
3282 shader_addline(buffer
, "for (aL%u = %u; aL%u > (%u * %d + %u); aL%u += %d)\n{\n",
3283 loop_state
->current_depth
, loop_control
.start
,
3284 loop_state
->current_depth
, loop_control
.count
, loop_control
.step
, loop_control
.start
,
3285 loop_state
->current_depth
, loop_control
.step
);
3289 shader_addline(buffer
, "for (aL%u = %u, tmpInt%u = 0; tmpInt%u < %u; tmpInt%u++)\n{\n",
3290 loop_state
->current_depth
, loop_control
.start
, loop_state
->current_depth
,
3291 loop_state
->current_depth
, loop_control
.count
,
3292 loop_state
->current_depth
);
3297 shader_addline(buffer
, "for (tmpInt%u = 0, aL%u = %s.y; tmpInt%u < %s.x; tmpInt%u++, aL%u += %s.z)\n{\n",
3298 loop_state
->current_depth
, loop_state
->current_reg
,
3299 src1_param
.reg_name
, loop_state
->current_depth
, src1_param
.reg_name
,
3300 loop_state
->current_depth
, loop_state
->current_reg
, src1_param
.reg_name
);
3303 ++loop_state
->current_reg
;
3307 shader_addline(buffer
, "for (;;)\n{\n");
3310 ++loop_state
->current_depth
;
3313 static void shader_glsl_end(const struct wined3d_shader_instruction
*ins
)
3315 struct wined3d_shader_loop_state
*loop_state
= ins
->ctx
->loop_state
;
3317 shader_addline(ins
->ctx
->buffer
, "}\n");
3319 if (ins
->handler_idx
== WINED3DSIH_ENDLOOP
)
3321 --loop_state
->current_depth
;
3322 --loop_state
->current_reg
;
3325 if (ins
->handler_idx
== WINED3DSIH_ENDREP
)
3327 --loop_state
->current_depth
;
3331 static void shader_glsl_rep(const struct wined3d_shader_instruction
*ins
)
3333 const struct wined3d_shader
*shader
= ins
->ctx
->shader
;
3334 struct wined3d_shader_loop_state
*loop_state
= ins
->ctx
->loop_state
;
3335 const struct wined3d_shader_lconst
*constant
;
3336 struct glsl_src_param src0_param
;
3337 const DWORD
*control_values
= NULL
;
3339 /* Try to hardcode local values to help the GLSL compiler to unroll and optimize the loop */
3340 if (ins
->src
[0].reg
.type
== WINED3DSPR_CONSTINT
)
3342 LIST_FOR_EACH_ENTRY(constant
, &shader
->constantsI
, struct wined3d_shader_lconst
, entry
)
3344 if (constant
->idx
== ins
->src
[0].reg
.idx
[0].offset
)
3346 control_values
= constant
->value
;
3354 shader_addline(ins
->ctx
->buffer
, "for (tmpInt%d = 0; tmpInt%d < %d; tmpInt%d++) {\n",
3355 loop_state
->current_depth
, loop_state
->current_depth
,
3356 control_values
[0], loop_state
->current_depth
);
3360 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
3361 shader_addline(ins
->ctx
->buffer
, "for (tmpInt%d = 0; tmpInt%d < %s; tmpInt%d++) {\n",
3362 loop_state
->current_depth
, loop_state
->current_depth
,
3363 src0_param
.param_str
, loop_state
->current_depth
);
3366 ++loop_state
->current_depth
;
3369 static void shader_glsl_if(const struct wined3d_shader_instruction
*ins
)
3371 struct glsl_src_param src0_param
;
3373 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
3374 shader_addline(ins
->ctx
->buffer
, "if (%s) {\n", src0_param
.param_str
);
3377 static void shader_glsl_ifc(const struct wined3d_shader_instruction
*ins
)
3379 struct glsl_src_param src0_param
;
3380 struct glsl_src_param src1_param
;
3382 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
3383 shader_glsl_add_src_param(ins
, &ins
->src
[1], WINED3DSP_WRITEMASK_0
, &src1_param
);
3385 shader_addline(ins
->ctx
->buffer
, "if (%s %s %s) {\n",
3386 src0_param
.param_str
, shader_glsl_get_rel_op(ins
->flags
), src1_param
.param_str
);
3389 static void shader_glsl_else(const struct wined3d_shader_instruction
*ins
)
3391 shader_addline(ins
->ctx
->buffer
, "} else {\n");
3394 static void shader_glsl_emit(const struct wined3d_shader_instruction
*ins
)
3396 shader_addline(ins
->ctx
->buffer
, "EmitVertex();\n");
3399 static void shader_glsl_break(const struct wined3d_shader_instruction
*ins
)
3401 shader_addline(ins
->ctx
->buffer
, "break;\n");
3404 /* FIXME: According to MSDN the compare is done per component. */
3405 static void shader_glsl_breakc(const struct wined3d_shader_instruction
*ins
)
3407 struct glsl_src_param src0_param
;
3408 struct glsl_src_param src1_param
;
3410 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
3411 shader_glsl_add_src_param(ins
, &ins
->src
[1], WINED3DSP_WRITEMASK_0
, &src1_param
);
3413 shader_addline(ins
->ctx
->buffer
, "if (%s %s %s) break;\n",
3414 src0_param
.param_str
, shader_glsl_get_rel_op(ins
->flags
), src1_param
.param_str
);
3417 static void shader_glsl_breakp(const struct wined3d_shader_instruction
*ins
)
3419 struct glsl_src_param src_param
;
3421 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
, &src_param
);
3422 shader_addline(ins
->ctx
->buffer
, "if (bool(%s)) break;\n", src_param
.param_str
);
3425 static void shader_glsl_label(const struct wined3d_shader_instruction
*ins
)
3427 shader_addline(ins
->ctx
->buffer
, "}\n");
3428 shader_addline(ins
->ctx
->buffer
, "void subroutine%u()\n{\n", ins
->src
[0].reg
.idx
[0].offset
);
3431 static void shader_glsl_call(const struct wined3d_shader_instruction
*ins
)
3433 shader_addline(ins
->ctx
->buffer
, "subroutine%u();\n", ins
->src
[0].reg
.idx
[0].offset
);
3436 static void shader_glsl_callnz(const struct wined3d_shader_instruction
*ins
)
3438 struct glsl_src_param src1_param
;
3440 shader_glsl_add_src_param(ins
, &ins
->src
[1], WINED3DSP_WRITEMASK_0
, &src1_param
);
3441 shader_addline(ins
->ctx
->buffer
, "if (%s) subroutine%u();\n",
3442 src1_param
.param_str
, ins
->src
[0].reg
.idx
[0].offset
);
3445 static void shader_glsl_ret(const struct wined3d_shader_instruction
*ins
)
3447 /* No-op. The closing } is written when a new function is started, and at the end of the shader. This
3448 * function only suppresses the unhandled instruction warning
3452 /*********************************************
3453 * Pixel Shader Specific Code begins here
3454 ********************************************/
3455 static void shader_glsl_tex(const struct wined3d_shader_instruction
*ins
)
3457 const struct wined3d_shader
*shader
= ins
->ctx
->shader
;
3458 struct wined3d_device
*device
= shader
->device
;
3459 DWORD shader_version
= WINED3D_SHADER_VERSION(ins
->ctx
->reg_maps
->shader_version
.major
,
3460 ins
->ctx
->reg_maps
->shader_version
.minor
);
3461 struct glsl_sample_function sample_function
;
3462 const struct wined3d_texture
*texture
;
3463 DWORD sample_flags
= 0;
3465 DWORD mask
= 0, swizzle
;
3467 /* 1.0-1.4: Use destination register as sampler source.
3468 * 2.0+: Use provided sampler source. */
3469 if (shader_version
< WINED3D_SHADER_VERSION(2,0))
3470 sampler_idx
= ins
->dst
[0].reg
.idx
[0].offset
;
3472 sampler_idx
= ins
->src
[1].reg
.idx
[0].offset
;
3473 texture
= device
->stateBlock
->state
.textures
[sampler_idx
];
3475 if (shader_version
< WINED3D_SHADER_VERSION(1,4))
3477 const struct shader_glsl_ctx_priv
*priv
= ins
->ctx
->backend_data
;
3478 DWORD flags
= (priv
->cur_ps_args
->tex_transform
>> sampler_idx
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
)
3479 & WINED3D_PSARGS_TEXTRANSFORM_MASK
;
3480 enum wined3d_sampler_texture_type sampler_type
= ins
->ctx
->reg_maps
->sampler_type
[sampler_idx
];
3482 /* Projected cube textures don't make a lot of sense, the resulting coordinates stay the same. */
3483 if (flags
& WINED3D_PSARGS_PROJECTED
&& sampler_type
!= WINED3DSTT_CUBE
)
3485 sample_flags
|= WINED3D_GLSL_SAMPLE_PROJECTED
;
3486 switch (flags
& ~WINED3D_PSARGS_PROJECTED
)
3488 case WINED3D_TTFF_COUNT1
:
3489 FIXME("WINED3D_TTFF_PROJECTED with WINED3D_TTFF_COUNT1?\n");
3491 case WINED3D_TTFF_COUNT2
:
3492 mask
= WINED3DSP_WRITEMASK_1
;
3494 case WINED3D_TTFF_COUNT3
:
3495 mask
= WINED3DSP_WRITEMASK_2
;
3497 case WINED3D_TTFF_COUNT4
:
3498 case WINED3D_TTFF_DISABLE
:
3499 mask
= WINED3DSP_WRITEMASK_3
;
3504 else if (shader_version
< WINED3D_SHADER_VERSION(2,0))
3506 enum wined3d_shader_src_modifier src_mod
= ins
->src
[0].modifiers
;
3508 if (src_mod
== WINED3DSPSM_DZ
) {
3509 sample_flags
|= WINED3D_GLSL_SAMPLE_PROJECTED
;
3510 mask
= WINED3DSP_WRITEMASK_2
;
3511 } else if (src_mod
== WINED3DSPSM_DW
) {
3512 sample_flags
|= WINED3D_GLSL_SAMPLE_PROJECTED
;
3513 mask
= WINED3DSP_WRITEMASK_3
;
3516 if (ins
->flags
& WINED3DSI_TEXLD_PROJECT
)
3518 /* ps 2.0 texldp instruction always divides by the fourth component. */
3519 sample_flags
|= WINED3D_GLSL_SAMPLE_PROJECTED
;
3520 mask
= WINED3DSP_WRITEMASK_3
;
3524 if (texture
&& texture
->target
== GL_TEXTURE_RECTANGLE_ARB
)
3525 sample_flags
|= WINED3D_GLSL_SAMPLE_RECT
;
3527 shader_glsl_get_sample_function(ins
->ctx
, sampler_idx
, sample_flags
, &sample_function
);
3528 mask
|= sample_function
.coord_mask
;
3530 if (shader_version
< WINED3D_SHADER_VERSION(2,0)) swizzle
= WINED3DSP_NOSWIZZLE
;
3531 else swizzle
= ins
->src
[1].swizzle
;
3533 /* 1.0-1.3: Use destination register as coordinate source.
3534 1.4+: Use provided coordinate source register. */
3535 if (shader_version
< WINED3D_SHADER_VERSION(1,4))
3538 shader_glsl_write_mask_to_str(mask
, coord_mask
);
3539 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, swizzle
, NULL
, NULL
, NULL
,
3540 "T%u%s", sampler_idx
, coord_mask
);
3544 struct glsl_src_param coord_param
;
3545 shader_glsl_add_src_param(ins
, &ins
->src
[0], mask
, &coord_param
);
3546 if (ins
->flags
& WINED3DSI_TEXLD_BIAS
)
3548 struct glsl_src_param bias
;
3549 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_3
, &bias
);
3550 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, swizzle
, NULL
, NULL
, bias
.param_str
,
3551 "%s", coord_param
.param_str
);
3553 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, swizzle
, NULL
, NULL
, NULL
,
3554 "%s", coord_param
.param_str
);
3559 static void shader_glsl_texldd(const struct wined3d_shader_instruction
*ins
)
3561 const struct wined3d_shader
*shader
= ins
->ctx
->shader
;
3562 struct wined3d_device
*device
= shader
->device
;
3563 const struct wined3d_gl_info
*gl_info
= ins
->ctx
->gl_info
;
3564 struct glsl_src_param coord_param
, dx_param
, dy_param
;
3565 DWORD sample_flags
= WINED3D_GLSL_SAMPLE_GRAD
;
3566 struct glsl_sample_function sample_function
;
3568 DWORD swizzle
= ins
->src
[1].swizzle
;
3569 const struct wined3d_texture
*texture
;
3571 if (!gl_info
->supported
[ARB_SHADER_TEXTURE_LOD
] && !gl_info
->supported
[EXT_GPU_SHADER4
])
3573 FIXME("texldd used, but not supported by hardware. Falling back to regular tex\n");
3574 shader_glsl_tex(ins
);
3578 sampler_idx
= ins
->src
[1].reg
.idx
[0].offset
;
3579 texture
= device
->stateBlock
->state
.textures
[sampler_idx
];
3580 if (texture
&& texture
->target
== GL_TEXTURE_RECTANGLE_ARB
)
3581 sample_flags
|= WINED3D_GLSL_SAMPLE_RECT
;
3583 shader_glsl_get_sample_function(ins
->ctx
, sampler_idx
, sample_flags
, &sample_function
);
3584 shader_glsl_add_src_param(ins
, &ins
->src
[0], sample_function
.coord_mask
, &coord_param
);
3585 shader_glsl_add_src_param(ins
, &ins
->src
[2], sample_function
.coord_mask
, &dx_param
);
3586 shader_glsl_add_src_param(ins
, &ins
->src
[3], sample_function
.coord_mask
, &dy_param
);
3588 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, swizzle
, dx_param
.param_str
, dy_param
.param_str
, NULL
,
3589 "%s", coord_param
.param_str
);
3592 static void shader_glsl_texldl(const struct wined3d_shader_instruction
*ins
)
3594 const struct wined3d_shader
*shader
= ins
->ctx
->shader
;
3595 struct wined3d_device
*device
= shader
->device
;
3596 const struct wined3d_gl_info
*gl_info
= ins
->ctx
->gl_info
;
3597 struct glsl_src_param coord_param
, lod_param
;
3598 DWORD sample_flags
= WINED3D_GLSL_SAMPLE_LOD
;
3599 struct glsl_sample_function sample_function
;
3601 DWORD swizzle
= ins
->src
[1].swizzle
;
3602 const struct wined3d_texture
*texture
;
3604 sampler_idx
= ins
->src
[1].reg
.idx
[0].offset
;
3605 texture
= device
->stateBlock
->state
.textures
[sampler_idx
];
3606 if (texture
&& texture
->target
== GL_TEXTURE_RECTANGLE_ARB
)
3607 sample_flags
|= WINED3D_GLSL_SAMPLE_RECT
;
3609 shader_glsl_get_sample_function(ins
->ctx
, sampler_idx
, sample_flags
, &sample_function
);
3610 shader_glsl_add_src_param(ins
, &ins
->src
[0], sample_function
.coord_mask
, &coord_param
);
3612 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_3
, &lod_param
);
3614 if (!gl_info
->supported
[ARB_SHADER_TEXTURE_LOD
] && !gl_info
->supported
[EXT_GPU_SHADER4
]
3615 && ins
->ctx
->reg_maps
->shader_version
.type
== WINED3D_SHADER_TYPE_PIXEL
)
3617 /* Plain GLSL only supports Lod sampling functions in vertex shaders.
3618 * However, the NVIDIA drivers allow them in fragment shaders as well,
3619 * even without the appropriate extension. */
3620 WARN("Using %s in fragment shader.\n", sample_function
.name
);
3622 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, swizzle
, NULL
, NULL
, lod_param
.param_str
,
3623 "%s", coord_param
.param_str
);
3626 static void shader_glsl_texcoord(const struct wined3d_shader_instruction
*ins
)
3628 /* FIXME: Make this work for more than just 2D textures */
3629 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
3630 DWORD write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
3632 if (!(ins
->ctx
->reg_maps
->shader_version
.major
== 1 && ins
->ctx
->reg_maps
->shader_version
.minor
== 4))
3636 shader_glsl_get_write_mask(&ins
->dst
[0], dst_mask
);
3637 shader_addline(buffer
, "clamp(gl_TexCoord[%u], 0.0, 1.0)%s);\n",
3638 ins
->dst
[0].reg
.idx
[0].offset
, dst_mask
);
3642 enum wined3d_shader_src_modifier src_mod
= ins
->src
[0].modifiers
;
3643 DWORD reg
= ins
->src
[0].reg
.idx
[0].offset
;
3644 char dst_swizzle
[6];
3646 shader_glsl_get_swizzle(&ins
->src
[0], FALSE
, write_mask
, dst_swizzle
);
3648 if (src_mod
== WINED3DSPSM_DZ
)
3650 unsigned int mask_size
= shader_glsl_get_write_mask_size(write_mask
);
3651 struct glsl_src_param div_param
;
3653 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_2
, &div_param
);
3655 if (mask_size
> 1) {
3656 shader_addline(buffer
, "gl_TexCoord[%u]%s / vec%d(%s));\n", reg
, dst_swizzle
, mask_size
, div_param
.param_str
);
3658 shader_addline(buffer
, "gl_TexCoord[%u]%s / %s);\n", reg
, dst_swizzle
, div_param
.param_str
);
3661 else if (src_mod
== WINED3DSPSM_DW
)
3663 unsigned int mask_size
= shader_glsl_get_write_mask_size(write_mask
);
3664 struct glsl_src_param div_param
;
3666 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_3
, &div_param
);
3668 if (mask_size
> 1) {
3669 shader_addline(buffer
, "gl_TexCoord[%u]%s / vec%d(%s));\n", reg
, dst_swizzle
, mask_size
, div_param
.param_str
);
3671 shader_addline(buffer
, "gl_TexCoord[%u]%s / %s);\n", reg
, dst_swizzle
, div_param
.param_str
);
3674 shader_addline(buffer
, "gl_TexCoord[%u]%s);\n", reg
, dst_swizzle
);
3679 /** Process the WINED3DSIO_TEXDP3TEX instruction in GLSL:
3680 * Take a 3-component dot product of the TexCoord[dstreg] and src,
3681 * then perform a 1D texture lookup from stage dstregnum, place into dst. */
3682 static void shader_glsl_texdp3tex(const struct wined3d_shader_instruction
*ins
)
3684 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
3685 DWORD sampler_idx
= ins
->dst
[0].reg
.idx
[0].offset
;
3686 struct glsl_sample_function sample_function
;
3687 struct glsl_src_param src0_param
;
3690 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
3692 /* Do I have to take care about the projected bit? I don't think so, since the dp3 returns only one
3693 * scalar, and projected sampling would require 4.
3695 * It is a dependent read - not valid with conditional NP2 textures
3697 shader_glsl_get_sample_function(ins
->ctx
, sampler_idx
, 0, &sample_function
);
3698 mask_size
= shader_glsl_get_write_mask_size(sample_function
.coord_mask
);
3703 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, WINED3DSP_NOSWIZZLE
, NULL
, NULL
, NULL
,
3704 "dot(gl_TexCoord[%u].xyz, %s)", sampler_idx
, src0_param
.param_str
);
3708 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, WINED3DSP_NOSWIZZLE
, NULL
, NULL
, NULL
,
3709 "vec2(dot(gl_TexCoord[%u].xyz, %s), 0.0)", sampler_idx
, src0_param
.param_str
);
3713 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, WINED3DSP_NOSWIZZLE
, NULL
, NULL
, NULL
,
3714 "vec3(dot(gl_TexCoord[%u].xyz, %s), 0.0, 0.0)", sampler_idx
, src0_param
.param_str
);
3718 FIXME("Unexpected mask size %u\n", mask_size
);
3723 /** Process the WINED3DSIO_TEXDP3 instruction in GLSL:
3724 * Take a 3-component dot product of the TexCoord[dstreg] and src. */
3725 static void shader_glsl_texdp3(const struct wined3d_shader_instruction
*ins
)
3727 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
3728 DWORD dstreg
= ins
->dst
[0].reg
.idx
[0].offset
;
3729 struct glsl_src_param src0_param
;
3731 unsigned int mask_size
;
3733 dst_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
3734 mask_size
= shader_glsl_get_write_mask_size(dst_mask
);
3735 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
3737 if (mask_size
> 1) {
3738 shader_addline(ins
->ctx
->buffer
, "vec%d(dot(T%u.xyz, %s)));\n", mask_size
, dstreg
, src0_param
.param_str
);
3740 shader_addline(ins
->ctx
->buffer
, "dot(T%u.xyz, %s));\n", dstreg
, src0_param
.param_str
);
3744 /** Process the WINED3DSIO_TEXDEPTH instruction in GLSL:
3745 * Calculate the depth as dst.x / dst.y */
3746 static void shader_glsl_texdepth(const struct wined3d_shader_instruction
*ins
)
3748 struct glsl_dst_param dst_param
;
3750 shader_glsl_add_dst_param(ins
, &ins
->dst
[0], &dst_param
);
3752 /* Tests show that texdepth never returns anything below 0.0, and that r5.y is clamped to 1.0.
3753 * Negative input is accepted, -0.25 / -0.5 returns 0.5. GL should clamp gl_FragDepth to [0;1], but
3754 * this doesn't always work, so clamp the results manually. Whether or not the x value is clamped at 1
3755 * too is irrelevant, since if x = 0, any y value < 1.0 (and > 1.0 is not allowed) results in a result
3758 shader_addline(ins
->ctx
->buffer
, "gl_FragDepth = clamp((%s.x / min(%s.y, 1.0)), 0.0, 1.0);\n",
3759 dst_param
.reg_name
, dst_param
.reg_name
);
3762 /** Process the WINED3DSIO_TEXM3X2DEPTH instruction in GLSL:
3763 * Last row of a 3x2 matrix multiply, use the result to calculate the depth:
3764 * Calculate tmp0.y = TexCoord[dstreg] . src.xyz; (tmp0.x has already been calculated)
3765 * depth = (tmp0.y == 0.0) ? 1.0 : tmp0.x / tmp0.y
3767 static void shader_glsl_texm3x2depth(const struct wined3d_shader_instruction
*ins
)
3769 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
3770 DWORD dstreg
= ins
->dst
[0].reg
.idx
[0].offset
;
3771 struct glsl_src_param src0_param
;
3773 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
3775 shader_addline(ins
->ctx
->buffer
, "tmp0.y = dot(T%u.xyz, %s);\n", dstreg
, src0_param
.param_str
);
3776 shader_addline(ins
->ctx
->buffer
, "gl_FragDepth = (tmp0.y == 0.0) ? 1.0 : clamp(tmp0.x / tmp0.y, 0.0, 1.0);\n");
3779 /** Process the WINED3DSIO_TEXM3X2PAD instruction in GLSL
3780 * Calculate the 1st of a 2-row matrix multiplication. */
3781 static void shader_glsl_texm3x2pad(const struct wined3d_shader_instruction
*ins
)
3783 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
3784 DWORD reg
= ins
->dst
[0].reg
.idx
[0].offset
;
3785 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
3786 struct glsl_src_param src0_param
;
3788 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
3789 shader_addline(buffer
, "tmp0.x = dot(T%u.xyz, %s);\n", reg
, src0_param
.param_str
);
3792 /** Process the WINED3DSIO_TEXM3X3PAD instruction in GLSL
3793 * Calculate the 1st or 2nd row of a 3-row matrix multiplication. */
3794 static void shader_glsl_texm3x3pad(const struct wined3d_shader_instruction
*ins
)
3796 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
3797 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
3798 struct wined3d_shader_tex_mx
*tex_mx
= ins
->ctx
->tex_mx
;
3799 DWORD reg
= ins
->dst
[0].reg
.idx
[0].offset
;
3800 struct glsl_src_param src0_param
;
3802 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
3803 shader_addline(buffer
, "tmp0.%c = dot(T%u.xyz, %s);\n", 'x' + tex_mx
->current_row
, reg
, src0_param
.param_str
);
3804 tex_mx
->texcoord_w
[tex_mx
->current_row
++] = reg
;
3807 static void shader_glsl_texm3x2tex(const struct wined3d_shader_instruction
*ins
)
3809 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
3810 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
3811 struct glsl_sample_function sample_function
;
3812 DWORD reg
= ins
->dst
[0].reg
.idx
[0].offset
;
3813 struct glsl_src_param src0_param
;
3815 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
3816 shader_addline(buffer
, "tmp0.y = dot(T%u.xyz, %s);\n", reg
, src0_param
.param_str
);
3818 shader_glsl_get_sample_function(ins
->ctx
, reg
, 0, &sample_function
);
3820 /* Sample the texture using the calculated coordinates */
3821 shader_glsl_gen_sample_code(ins
, reg
, &sample_function
, WINED3DSP_NOSWIZZLE
, NULL
, NULL
, NULL
, "tmp0.xy");
3824 /** Process the WINED3DSIO_TEXM3X3TEX instruction in GLSL
3825 * Perform the 3rd row of a 3x3 matrix multiply, then sample the texture using the calculated coordinates */
3826 static void shader_glsl_texm3x3tex(const struct wined3d_shader_instruction
*ins
)
3828 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
3829 struct wined3d_shader_tex_mx
*tex_mx
= ins
->ctx
->tex_mx
;
3830 struct glsl_sample_function sample_function
;
3831 DWORD reg
= ins
->dst
[0].reg
.idx
[0].offset
;
3832 struct glsl_src_param src0_param
;
3834 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
3835 shader_addline(ins
->ctx
->buffer
, "tmp0.z = dot(T%u.xyz, %s);\n", reg
, src0_param
.param_str
);
3837 /* Dependent read, not valid with conditional NP2 */
3838 shader_glsl_get_sample_function(ins
->ctx
, reg
, 0, &sample_function
);
3840 /* Sample the texture using the calculated coordinates */
3841 shader_glsl_gen_sample_code(ins
, reg
, &sample_function
, WINED3DSP_NOSWIZZLE
, NULL
, NULL
, NULL
, "tmp0.xyz");
3843 tex_mx
->current_row
= 0;
3846 /** Process the WINED3DSIO_TEXM3X3 instruction in GLSL
3847 * Perform the 3rd row of a 3x3 matrix multiply */
3848 static void shader_glsl_texm3x3(const struct wined3d_shader_instruction
*ins
)
3850 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
3851 struct wined3d_shader_tex_mx
*tex_mx
= ins
->ctx
->tex_mx
;
3852 DWORD reg
= ins
->dst
[0].reg
.idx
[0].offset
;
3853 struct glsl_src_param src0_param
;
3856 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
3858 shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
3859 shader_glsl_get_write_mask(&ins
->dst
[0], dst_mask
);
3860 shader_addline(ins
->ctx
->buffer
, "vec4(tmp0.xy, dot(T%u.xyz, %s), 1.0)%s);\n", reg
, src0_param
.param_str
, dst_mask
);
3862 tex_mx
->current_row
= 0;
3865 /* Process the WINED3DSIO_TEXM3X3SPEC instruction in GLSL
3866 * Perform the final texture lookup based on the previous 2 3x3 matrix multiplies */
3867 static void shader_glsl_texm3x3spec(const struct wined3d_shader_instruction
*ins
)
3869 struct glsl_src_param src0_param
;
3870 struct glsl_src_param src1_param
;
3871 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
3872 struct wined3d_shader_tex_mx
*tex_mx
= ins
->ctx
->tex_mx
;
3873 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
3874 struct glsl_sample_function sample_function
;
3875 DWORD reg
= ins
->dst
[0].reg
.idx
[0].offset
;
3878 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
3879 shader_glsl_add_src_param(ins
, &ins
->src
[1], src_mask
, &src1_param
);
3881 /* Perform the last matrix multiply operation */
3882 shader_addline(buffer
, "tmp0.z = dot(T%u.xyz, %s);\n", reg
, src0_param
.param_str
);
3883 /* Reflection calculation */
3884 shader_addline(buffer
, "tmp0.xyz = -reflect((%s), normalize(tmp0.xyz));\n", src1_param
.param_str
);
3886 /* Dependent read, not valid with conditional NP2 */
3887 shader_glsl_get_sample_function(ins
->ctx
, reg
, 0, &sample_function
);
3888 shader_glsl_write_mask_to_str(sample_function
.coord_mask
, coord_mask
);
3890 /* Sample the texture */
3891 shader_glsl_gen_sample_code(ins
, reg
, &sample_function
, WINED3DSP_NOSWIZZLE
,
3892 NULL
, NULL
, NULL
, "tmp0%s", coord_mask
);
3894 tex_mx
->current_row
= 0;
3897 /* Process the WINED3DSIO_TEXM3X3VSPEC instruction in GLSL
3898 * Perform the final texture lookup based on the previous 2 3x3 matrix multiplies */
3899 static void shader_glsl_texm3x3vspec(const struct wined3d_shader_instruction
*ins
)
3901 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
3902 struct wined3d_shader_tex_mx
*tex_mx
= ins
->ctx
->tex_mx
;
3903 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
3904 struct glsl_sample_function sample_function
;
3905 DWORD reg
= ins
->dst
[0].reg
.idx
[0].offset
;
3906 struct glsl_src_param src0_param
;
3909 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
3911 /* Perform the last matrix multiply operation */
3912 shader_addline(buffer
, "tmp0.z = dot(vec3(T%u), vec3(%s));\n", reg
, src0_param
.param_str
);
3914 /* Construct the eye-ray vector from w coordinates */
3915 shader_addline(buffer
, "tmp1.xyz = normalize(vec3(gl_TexCoord[%u].w, gl_TexCoord[%u].w, gl_TexCoord[%u].w));\n",
3916 tex_mx
->texcoord_w
[0], tex_mx
->texcoord_w
[1], reg
);
3917 shader_addline(buffer
, "tmp0.xyz = -reflect(tmp1.xyz, normalize(tmp0.xyz));\n");
3919 /* Dependent read, not valid with conditional NP2 */
3920 shader_glsl_get_sample_function(ins
->ctx
, reg
, 0, &sample_function
);
3921 shader_glsl_write_mask_to_str(sample_function
.coord_mask
, coord_mask
);
3923 /* Sample the texture using the calculated coordinates */
3924 shader_glsl_gen_sample_code(ins
, reg
, &sample_function
, WINED3DSP_NOSWIZZLE
,
3925 NULL
, NULL
, NULL
, "tmp0%s", coord_mask
);
3927 tex_mx
->current_row
= 0;
3930 /** Process the WINED3DSIO_TEXBEM instruction in GLSL.
3931 * Apply a fake bump map transform.
3932 * texbem is pshader <= 1.3 only, this saves a few version checks
3934 static void shader_glsl_texbem(const struct wined3d_shader_instruction
*ins
)
3936 const struct shader_glsl_ctx_priv
*priv
= ins
->ctx
->backend_data
;
3937 struct glsl_sample_function sample_function
;
3938 struct glsl_src_param coord_param
;
3944 sampler_idx
= ins
->dst
[0].reg
.idx
[0].offset
;
3945 flags
= (priv
->cur_ps_args
->tex_transform
>> sampler_idx
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
)
3946 & WINED3D_PSARGS_TEXTRANSFORM_MASK
;
3948 /* Dependent read, not valid with conditional NP2 */
3949 shader_glsl_get_sample_function(ins
->ctx
, sampler_idx
, 0, &sample_function
);
3950 mask
= sample_function
.coord_mask
;
3952 shader_glsl_write_mask_to_str(mask
, coord_mask
);
3954 /* With projected textures, texbem only divides the static texture coord,
3955 * not the displacement, so we can't let GL handle this. */
3956 if (flags
& WINED3D_PSARGS_PROJECTED
)
3959 char coord_div_mask
[3];
3960 switch (flags
& ~WINED3D_PSARGS_PROJECTED
)
3962 case WINED3D_TTFF_COUNT1
:
3963 FIXME("WINED3D_TTFF_PROJECTED with WINED3D_TTFF_COUNT1?\n");
3965 case WINED3D_TTFF_COUNT2
:
3966 div_mask
= WINED3DSP_WRITEMASK_1
;
3968 case WINED3D_TTFF_COUNT3
:
3969 div_mask
= WINED3DSP_WRITEMASK_2
;
3971 case WINED3D_TTFF_COUNT4
:
3972 case WINED3D_TTFF_DISABLE
:
3973 div_mask
= WINED3DSP_WRITEMASK_3
;
3976 shader_glsl_write_mask_to_str(div_mask
, coord_div_mask
);
3977 shader_addline(ins
->ctx
->buffer
, "T%u%s /= T%u%s;\n", sampler_idx
, coord_mask
, sampler_idx
, coord_div_mask
);
3980 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
, &coord_param
);
3982 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, WINED3DSP_NOSWIZZLE
, NULL
, NULL
, NULL
,
3983 "T%u%s + vec4(bumpenv_mat%u * %s, 0.0, 0.0)%s", sampler_idx
, coord_mask
, sampler_idx
,
3984 coord_param
.param_str
, coord_mask
);
3986 if (ins
->handler_idx
== WINED3DSIH_TEXBEML
)
3988 struct glsl_src_param luminance_param
;
3989 struct glsl_dst_param dst_param
;
3991 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_2
, &luminance_param
);
3992 shader_glsl_add_dst_param(ins
, &ins
->dst
[0], &dst_param
);
3994 shader_addline(ins
->ctx
->buffer
, "%s%s *= (%s * bumpenv_lum_scale%u + bumpenv_lum_offset%u);\n",
3995 dst_param
.reg_name
, dst_param
.mask_str
,
3996 luminance_param
.param_str
, sampler_idx
, sampler_idx
);
4000 static void shader_glsl_bem(const struct wined3d_shader_instruction
*ins
)
4002 DWORD sampler_idx
= ins
->dst
[0].reg
.idx
[0].offset
;
4003 struct glsl_src_param src0_param
, src1_param
;
4005 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
, &src0_param
);
4006 shader_glsl_add_src_param(ins
, &ins
->src
[1], WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
, &src1_param
);
4008 shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
4009 shader_addline(ins
->ctx
->buffer
, "%s + bumpenv_mat%u * %s);\n",
4010 src0_param
.param_str
, sampler_idx
, src1_param
.param_str
);
4013 /** Process the WINED3DSIO_TEXREG2AR instruction in GLSL
4014 * Sample 2D texture at dst using the alpha & red (wx) components of src as texture coordinates */
4015 static void shader_glsl_texreg2ar(const struct wined3d_shader_instruction
*ins
)
4017 DWORD sampler_idx
= ins
->dst
[0].reg
.idx
[0].offset
;
4018 struct glsl_sample_function sample_function
;
4019 struct glsl_src_param src0_param
;
4021 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_ALL
, &src0_param
);
4023 shader_glsl_get_sample_function(ins
->ctx
, sampler_idx
, 0, &sample_function
);
4024 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, WINED3DSP_NOSWIZZLE
, NULL
, NULL
, NULL
,
4025 "%s.wx", src0_param
.reg_name
);
4028 /** Process the WINED3DSIO_TEXREG2GB instruction in GLSL
4029 * Sample 2D texture at dst using the green & blue (yz) components of src as texture coordinates */
4030 static void shader_glsl_texreg2gb(const struct wined3d_shader_instruction
*ins
)
4032 DWORD sampler_idx
= ins
->dst
[0].reg
.idx
[0].offset
;
4033 struct glsl_sample_function sample_function
;
4034 struct glsl_src_param src0_param
;
4036 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_ALL
, &src0_param
);
4038 shader_glsl_get_sample_function(ins
->ctx
, sampler_idx
, 0, &sample_function
);
4039 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, WINED3DSP_NOSWIZZLE
, NULL
, NULL
, NULL
,
4040 "%s.yz", src0_param
.reg_name
);
4043 /** Process the WINED3DSIO_TEXREG2RGB instruction in GLSL
4044 * Sample texture at dst using the rgb (xyz) components of src as texture coordinates */
4045 static void shader_glsl_texreg2rgb(const struct wined3d_shader_instruction
*ins
)
4047 DWORD sampler_idx
= ins
->dst
[0].reg
.idx
[0].offset
;
4048 struct glsl_sample_function sample_function
;
4049 struct glsl_src_param src0_param
;
4051 /* Dependent read, not valid with conditional NP2 */
4052 shader_glsl_get_sample_function(ins
->ctx
, sampler_idx
, 0, &sample_function
);
4053 shader_glsl_add_src_param(ins
, &ins
->src
[0], sample_function
.coord_mask
, &src0_param
);
4055 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, WINED3DSP_NOSWIZZLE
, NULL
, NULL
, NULL
,
4056 "%s", src0_param
.param_str
);
4059 /** Process the WINED3DSIO_TEXKILL instruction in GLSL.
4060 * If any of the first 3 components are < 0, discard this pixel */
4061 static void shader_glsl_texkill(const struct wined3d_shader_instruction
*ins
)
4063 struct glsl_dst_param dst_param
;
4065 /* The argument is a destination parameter, and no writemasks are allowed */
4066 shader_glsl_add_dst_param(ins
, &ins
->dst
[0], &dst_param
);
4067 if (ins
->ctx
->reg_maps
->shader_version
.major
>= 2)
4069 /* 2.0 shaders compare all 4 components in texkill */
4070 shader_addline(ins
->ctx
->buffer
, "if (any(lessThan(%s.xyzw, vec4(0.0)))) discard;\n", dst_param
.reg_name
);
4072 /* 1.X shaders only compare the first 3 components, probably due to the nature of the texkill
4073 * instruction as a tex* instruction, and phase, which kills all a / w components. Even if all
4074 * 4 components are defined, only the first 3 are used
4076 shader_addline(ins
->ctx
->buffer
, "if (any(lessThan(%s.xyz, vec3(0.0)))) discard;\n", dst_param
.reg_name
);
4080 /** Process the WINED3DSIO_DP2ADD instruction in GLSL.
4081 * dst = dot2(src0, src1) + src2 */
4082 static void shader_glsl_dp2add(const struct wined3d_shader_instruction
*ins
)
4084 struct glsl_src_param src0_param
;
4085 struct glsl_src_param src1_param
;
4086 struct glsl_src_param src2_param
;
4088 unsigned int mask_size
;
4090 write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
4091 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
4093 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
, &src0_param
);
4094 shader_glsl_add_src_param(ins
, &ins
->src
[1], WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
, &src1_param
);
4095 shader_glsl_add_src_param(ins
, &ins
->src
[2], WINED3DSP_WRITEMASK_0
, &src2_param
);
4097 if (mask_size
> 1) {
4098 shader_addline(ins
->ctx
->buffer
, "vec%d(dot(%s, %s) + %s));\n",
4099 mask_size
, src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
4101 shader_addline(ins
->ctx
->buffer
, "dot(%s, %s) + %s);\n",
4102 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
4106 static void shader_glsl_input_pack(const struct wined3d_shader
*shader
, struct wined3d_shader_buffer
*buffer
,
4107 const struct wined3d_shader_signature_element
*input_signature
,
4108 const struct wined3d_shader_reg_maps
*reg_maps
,
4109 enum vertexprocessing_mode vertexprocessing
)
4111 WORD map
= reg_maps
->input_registers
;
4114 for (i
= 0; map
; map
>>= 1, ++i
)
4116 const char *semantic_name
;
4121 if (!(map
& 1)) continue;
4123 semantic_name
= input_signature
[i
].semantic_name
;
4124 semantic_idx
= input_signature
[i
].semantic_idx
;
4125 shader_glsl_write_mask_to_str(input_signature
[i
].mask
, reg_mask
);
4127 if (shader_match_semantic(semantic_name
, WINED3D_DECL_USAGE_TEXCOORD
))
4129 if (semantic_idx
< 8 && vertexprocessing
== pretransformed
)
4130 shader_addline(buffer
, "ps_in[%u]%s = gl_TexCoord[%u]%s;\n",
4131 shader
->u
.ps
.input_reg_map
[i
], reg_mask
, semantic_idx
, reg_mask
);
4133 shader_addline(buffer
, "ps_in[%u]%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
4134 shader
->u
.ps
.input_reg_map
[i
], reg_mask
, reg_mask
);
4136 else if (shader_match_semantic(semantic_name
, WINED3D_DECL_USAGE_COLOR
))
4139 shader_addline(buffer
, "ps_in[%u]%s = vec4(gl_Color)%s;\n",
4140 shader
->u
.ps
.input_reg_map
[i
], reg_mask
, reg_mask
);
4141 else if (semantic_idx
== 1)
4142 shader_addline(buffer
, "ps_in[%u]%s = vec4(gl_SecondaryColor)%s;\n",
4143 shader
->u
.ps
.input_reg_map
[i
], reg_mask
, reg_mask
);
4145 shader_addline(buffer
, "ps_in[%u]%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
4146 shader
->u
.ps
.input_reg_map
[i
], reg_mask
, reg_mask
);
4150 shader_addline(buffer
, "ps_in[%u]%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
4151 shader
->u
.ps
.input_reg_map
[i
], reg_mask
, reg_mask
);
4156 /*********************************************
4157 * Vertex Shader Specific Code begins here
4158 ********************************************/
4160 static void add_glsl_program_entry(struct shader_glsl_priv
*priv
, struct glsl_shader_prog_link
*entry
)
4162 struct glsl_program_key key
;
4164 key
.vs_id
= entry
->vs
.id
;
4165 key
.gs_id
= entry
->gs
.id
;
4166 key
.ps_id
= entry
->ps
.id
;
4168 if (wine_rb_put(&priv
->program_lookup
, &key
, &entry
->program_lookup_entry
) == -1)
4170 ERR("Failed to insert program entry.\n");
4174 static struct glsl_shader_prog_link
*get_glsl_program_entry(const struct shader_glsl_priv
*priv
,
4175 GLhandleARB vs_id
, GLhandleARB gs_id
, GLhandleARB ps_id
)
4177 struct wine_rb_entry
*entry
;
4178 struct glsl_program_key key
;
4184 entry
= wine_rb_get(&priv
->program_lookup
, &key
);
4185 return entry
? WINE_RB_ENTRY_VALUE(entry
, struct glsl_shader_prog_link
, program_lookup_entry
) : NULL
;
4188 /* Context activation is done by the caller. */
4189 static void delete_glsl_program_entry(struct shader_glsl_priv
*priv
, const struct wined3d_gl_info
*gl_info
,
4190 struct glsl_shader_prog_link
*entry
)
4192 struct glsl_program_key key
;
4194 key
.vs_id
= entry
->vs
.id
;
4195 key
.gs_id
= entry
->gs
.id
;
4196 key
.ps_id
= entry
->ps
.id
;
4197 wine_rb_remove(&priv
->program_lookup
, &key
);
4199 GL_EXTCALL(glDeleteObjectARB(entry
->programId
));
4201 list_remove(&entry
->vs
.shader_entry
);
4203 list_remove(&entry
->gs
.shader_entry
);
4205 list_remove(&entry
->ps
.shader_entry
);
4206 HeapFree(GetProcessHeap(), 0, entry
->vs
.uniform_f_locations
);
4207 HeapFree(GetProcessHeap(), 0, entry
->ps
.uniform_f_locations
);
4208 HeapFree(GetProcessHeap(), 0, entry
);
4211 static void handle_ps3_input(struct wined3d_shader_buffer
*buffer
,
4212 const struct wined3d_gl_info
*gl_info
, const DWORD
*map
,
4213 const struct wined3d_shader_signature_element
*input_signature
,
4214 const struct wined3d_shader_reg_maps
*reg_maps_in
,
4215 const struct wined3d_shader_signature_element
*output_signature
,
4216 const struct wined3d_shader_reg_maps
*reg_maps_out
)
4219 const char *semantic_name_in
;
4220 UINT semantic_idx_in
;
4223 unsigned int in_count
= vec4_varyings(3, gl_info
);
4225 char destination
[50];
4226 WORD input_map
, output_map
;
4228 set
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*set
) * (in_count
+ 2));
4230 input_map
= reg_maps_in
->input_registers
;
4231 for (i
= 0; input_map
; input_map
>>= 1, ++i
)
4233 if (!(input_map
& 1)) continue;
4236 /* Declared, but not read register */
4237 if (in_idx
== ~0U) continue;
4238 if (in_idx
>= (in_count
+ 2))
4240 FIXME("More input varyings declared than supported, expect issues.\n");
4244 if (in_idx
== in_count
)
4245 sprintf(destination
, "gl_FrontColor");
4246 else if (in_idx
== in_count
+ 1)
4247 sprintf(destination
, "gl_FrontSecondaryColor");
4249 sprintf(destination
, "ps_in[%u]", in_idx
);
4251 semantic_name_in
= input_signature
[i
].semantic_name
;
4252 semantic_idx_in
= input_signature
[i
].semantic_idx
;
4255 output_map
= reg_maps_out
->output_registers
;
4256 for (j
= 0; output_map
; output_map
>>= 1, ++j
)
4260 if (!(output_map
& 1)
4261 || semantic_idx_in
!= output_signature
[j
].semantic_idx
4262 || strcmp(semantic_name_in
, output_signature
[j
].semantic_name
)
4263 || !(mask
= input_signature
[i
].mask
& output_signature
[j
].mask
))
4267 shader_glsl_write_mask_to_str(mask
, reg_mask
);
4269 shader_addline(buffer
, "%s%s = vs_out[%u]%s;\n",
4270 destination
, reg_mask
, j
, reg_mask
);
4274 for (i
= 0; i
< in_count
+ 2; ++i
)
4278 if (!set
[i
] || set
[i
] == WINED3DSP_WRITEMASK_ALL
)
4281 if (set
[i
] == ~0U) set
[i
] = 0;
4284 if (!(set
[i
] & WINED3DSP_WRITEMASK_0
)) reg_mask
[size
++] = 'x';
4285 if (!(set
[i
] & WINED3DSP_WRITEMASK_1
)) reg_mask
[size
++] = 'y';
4286 if (!(set
[i
] & WINED3DSP_WRITEMASK_2
)) reg_mask
[size
++] = 'z';
4287 if (!(set
[i
] & WINED3DSP_WRITEMASK_3
)) reg_mask
[size
++] = 'w';
4288 reg_mask
[size
] = '\0';
4291 sprintf(destination
, "gl_FrontColor");
4292 else if (i
== in_count
+ 1)
4293 sprintf(destination
, "gl_FrontSecondaryColor");
4295 sprintf(destination
, "ps_in[%u]", i
);
4297 if (size
== 1) shader_addline(buffer
, "%s.%s = 0.0;\n", destination
, reg_mask
);
4298 else shader_addline(buffer
, "%s.%s = vec%u(0.0);\n", destination
, reg_mask
, size
);
4301 HeapFree(GetProcessHeap(), 0, set
);
4304 /* Context activation is done by the caller. */
4305 static GLhandleARB
generate_param_reorder_function(struct wined3d_shader_buffer
*buffer
,
4306 const struct wined3d_shader
*vs
, const struct wined3d_shader
*ps
,
4307 const struct wined3d_gl_info
*gl_info
)
4309 GLhandleARB ret
= 0;
4310 DWORD ps_major
= ps
? ps
->reg_maps
.shader_version
.major
: 0;
4312 const char *semantic_name
;
4315 const struct wined3d_shader_signature_element
*output_signature
= vs
->output_signature
;
4316 WORD map
= vs
->reg_maps
.output_registers
;
4318 shader_buffer_clear(buffer
);
4320 shader_addline(buffer
, "#version 120\n");
4324 shader_addline(buffer
, "void order_ps_input(in vec4 vs_out[%u])\n{\n", vs
->limits
.packed_output
);
4326 for (i
= 0; map
; map
>>= 1, ++i
)
4330 if (!(map
& 1)) continue;
4332 semantic_name
= output_signature
[i
].semantic_name
;
4333 semantic_idx
= output_signature
[i
].semantic_idx
;
4334 write_mask
= output_signature
[i
].mask
;
4335 shader_glsl_write_mask_to_str(write_mask
, reg_mask
);
4337 if (shader_match_semantic(semantic_name
, WINED3D_DECL_USAGE_COLOR
))
4340 shader_addline(buffer
, "gl_FrontColor%s = vs_out[%u]%s;\n",
4341 reg_mask
, i
, reg_mask
);
4342 else if (semantic_idx
== 1)
4343 shader_addline(buffer
, "gl_FrontSecondaryColor%s = vs_out[%u]%s;\n",
4344 reg_mask
, i
, reg_mask
);
4346 else if (shader_match_semantic(semantic_name
, WINED3D_DECL_USAGE_POSITION
))
4348 shader_addline(buffer
, "gl_Position%s = vs_out[%u]%s;\n",
4349 reg_mask
, i
, reg_mask
);
4351 else if (shader_match_semantic(semantic_name
, WINED3D_DECL_USAGE_TEXCOORD
))
4353 if (semantic_idx
< 8)
4355 if (!(gl_info
->quirks
& WINED3D_QUIRK_SET_TEXCOORD_W
) || ps_major
> 0)
4356 write_mask
|= WINED3DSP_WRITEMASK_3
;
4358 shader_addline(buffer
, "gl_TexCoord[%u]%s = vs_out[%u]%s;\n",
4359 semantic_idx
, reg_mask
, i
, reg_mask
);
4360 if (!(write_mask
& WINED3DSP_WRITEMASK_3
))
4361 shader_addline(buffer
, "gl_TexCoord[%u].w = 1.0;\n", semantic_idx
);
4364 else if (shader_match_semantic(semantic_name
, WINED3D_DECL_USAGE_PSIZE
))
4366 shader_addline(buffer
, "gl_PointSize = vs_out[%u].%c;\n", i
, reg_mask
[1]);
4368 else if (shader_match_semantic(semantic_name
, WINED3D_DECL_USAGE_FOG
))
4370 shader_addline(buffer
, "gl_FogFragCoord = clamp(vs_out[%u].%c, 0.0, 1.0);\n", i
, reg_mask
[1]);
4373 shader_addline(buffer
, "}\n");
4377 UINT in_count
= min(vec4_varyings(ps_major
, gl_info
), ps
->limits
.packed_input
);
4378 /* This one is tricky: a 3.0 pixel shader reads from a 3.0 vertex shader */
4379 shader_addline(buffer
, "varying vec4 ps_in[%u];\n", in_count
);
4380 shader_addline(buffer
, "void order_ps_input(in vec4 vs_out[%u])\n{\n", vs
->limits
.packed_output
);
4382 /* First, sort out position and point size. Those are not passed to the pixel shader */
4383 for (i
= 0; map
; map
>>= 1, ++i
)
4385 if (!(map
& 1)) continue;
4387 semantic_name
= output_signature
[i
].semantic_name
;
4388 shader_glsl_write_mask_to_str(output_signature
[i
].mask
, reg_mask
);
4390 if (shader_match_semantic(semantic_name
, WINED3D_DECL_USAGE_POSITION
))
4392 shader_addline(buffer
, "gl_Position%s = vs_out[%u]%s;\n",
4393 reg_mask
, i
, reg_mask
);
4395 else if (shader_match_semantic(semantic_name
, WINED3D_DECL_USAGE_PSIZE
))
4397 shader_addline(buffer
, "gl_PointSize = vs_out[%u].%c;\n", i
, reg_mask
[1]);
4401 /* Then, fix the pixel shader input */
4402 handle_ps3_input(buffer
, gl_info
, ps
->u
.ps
.input_reg_map
, ps
->input_signature
,
4403 &ps
->reg_maps
, output_signature
, &vs
->reg_maps
);
4405 shader_addline(buffer
, "}\n");
4408 ret
= GL_EXTCALL(glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB
));
4409 checkGLcall("glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB)");
4410 shader_glsl_compile(gl_info
, ret
, buffer
->buffer
);
4415 static void shader_glsl_generate_srgb_write_correction(struct wined3d_shader_buffer
*buffer
)
4417 shader_addline(buffer
, "tmp0.xyz = pow(gl_FragData[0].xyz, vec3(srgb_const0.x));\n");
4418 shader_addline(buffer
, "tmp0.xyz = tmp0.xyz * vec3(srgb_const0.y) - vec3(srgb_const0.z);\n");
4419 shader_addline(buffer
, "tmp1.xyz = gl_FragData[0].xyz * vec3(srgb_const0.w);\n");
4420 shader_addline(buffer
, "bvec3 srgb_compare = lessThan(gl_FragData[0].xyz, vec3(srgb_const1.x));\n");
4421 shader_addline(buffer
, "gl_FragData[0].xyz = mix(tmp0.xyz, tmp1.xyz, vec3(srgb_compare));\n");
4422 shader_addline(buffer
, "gl_FragData[0] = clamp(gl_FragData[0], 0.0, 1.0);\n");
4425 static void shader_glsl_generate_fog_code(struct wined3d_shader_buffer
*buffer
, enum fogmode mode
)
4433 /* Fog = (gl_Fog.end - gl_FogFragCoord) / (gl_Fog.end - gl_Fog.start) */
4434 shader_addline(buffer
, "float Fog = (gl_Fog.end - gl_FogFragCoord) / (gl_Fog.end - gl_Fog.start);\n");
4438 /* Fog = e^-(gl_Fog.density * gl_FogFragCoord) */
4439 shader_addline(buffer
, "float Fog = exp(-gl_Fog.density * gl_FogFragCoord);\n");
4443 /* Fog = e^-((gl_Fog.density * gl_FogFragCoord)^2) */
4444 shader_addline(buffer
, "float Fog = exp(-gl_Fog.density * gl_Fog.density * gl_FogFragCoord * gl_FogFragCoord);\n");
4448 ERR("Invalid fog mode %#x.\n", mode
);
4452 shader_addline(buffer
, "gl_FragData[0].xyz = mix(gl_Fog.color.xyz, gl_FragData[0].xyz, clamp(Fog, 0.0, 1.0));\n");
4455 /* Context activation is done by the caller. */
4456 static GLuint
shader_glsl_generate_pshader(const struct wined3d_context
*context
,
4457 struct wined3d_shader_buffer
*buffer
, const struct wined3d_shader
*shader
,
4458 const struct ps_compile_args
*args
, struct ps_np2fixup_info
*np2fixup_info
)
4460 const struct wined3d_shader_reg_maps
*reg_maps
= &shader
->reg_maps
;
4461 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
4462 const DWORD
*function
= shader
->function
;
4463 struct shader_glsl_ctx_priv priv_ctx
;
4465 /* Create the hw GLSL shader object and assign it as the shader->prgId */
4466 GLhandleARB shader_obj
= GL_EXTCALL(glCreateShaderObjectARB(GL_FRAGMENT_SHADER_ARB
));
4468 memset(&priv_ctx
, 0, sizeof(priv_ctx
));
4469 priv_ctx
.cur_ps_args
= args
;
4470 priv_ctx
.cur_np2fixup_info
= np2fixup_info
;
4472 shader_addline(buffer
, "#version 120\n");
4474 if (gl_info
->supported
[ARB_SHADER_BIT_ENCODING
])
4475 shader_addline(buffer
, "#extension GL_ARB_shader_bit_encoding : enable\n");
4476 if (gl_info
->supported
[ARB_SHADER_TEXTURE_LOD
])
4477 shader_addline(buffer
, "#extension GL_ARB_shader_texture_lod : enable\n");
4478 /* The spec says that it doesn't have to be explicitly enabled, but the
4479 * nvidia drivers write a warning if we don't do so. */
4480 if (gl_info
->supported
[ARB_TEXTURE_RECTANGLE
])
4481 shader_addline(buffer
, "#extension GL_ARB_texture_rectangle : enable\n");
4482 if (gl_info
->supported
[EXT_GPU_SHADER4
])
4483 shader_addline(buffer
, "#extension GL_EXT_gpu_shader4 : enable\n");
4485 /* Base Declarations */
4486 shader_generate_glsl_declarations(context
, buffer
, shader
, reg_maps
, &priv_ctx
);
4488 /* Pack 3.0 inputs */
4489 if (reg_maps
->shader_version
.major
>= 3 && args
->vp_mode
!= vertexshader
)
4490 shader_glsl_input_pack(shader
, buffer
, shader
->input_signature
, reg_maps
, args
->vp_mode
);
4492 /* Base Shader Body */
4493 shader_generate_main(shader
, buffer
, reg_maps
, function
, &priv_ctx
);
4495 /* Pixel shaders < 2.0 place the resulting color in R0 implicitly */
4496 if (reg_maps
->shader_version
.major
< 2)
4498 /* Some older cards like GeforceFX ones don't support multiple buffers, so also not gl_FragData */
4499 shader_addline(buffer
, "gl_FragData[0] = R0;\n");
4502 if (args
->srgb_correction
)
4503 shader_glsl_generate_srgb_write_correction(buffer
);
4505 /* SM < 3 does not replace the fog stage. */
4506 if (reg_maps
->shader_version
.major
< 3)
4507 shader_glsl_generate_fog_code(buffer
, args
->fog
);
4509 shader_addline(buffer
, "}\n");
4511 TRACE("Compiling shader object %u\n", shader_obj
);
4512 shader_glsl_compile(gl_info
, shader_obj
, buffer
->buffer
);
4514 /* Store the shader object */
4518 /* Context activation is done by the caller. */
4519 static GLuint
shader_glsl_generate_vshader(const struct wined3d_context
*context
,
4520 struct wined3d_shader_buffer
*buffer
, const struct wined3d_shader
*shader
,
4521 const struct vs_compile_args
*args
)
4523 const struct wined3d_shader_reg_maps
*reg_maps
= &shader
->reg_maps
;
4524 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
4525 const DWORD
*function
= shader
->function
;
4526 struct shader_glsl_ctx_priv priv_ctx
;
4528 /* Create the hw GLSL shader program and assign it as the shader->prgId */
4529 GLhandleARB shader_obj
= GL_EXTCALL(glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB
));
4531 shader_addline(buffer
, "#version 120\n");
4533 if (gl_info
->supported
[ARB_SHADER_BIT_ENCODING
])
4534 shader_addline(buffer
, "#extension GL_ARB_shader_bit_encoding : enable\n");
4535 if (gl_info
->supported
[EXT_GPU_SHADER4
])
4536 shader_addline(buffer
, "#extension GL_EXT_gpu_shader4 : enable\n");
4538 memset(&priv_ctx
, 0, sizeof(priv_ctx
));
4539 priv_ctx
.cur_vs_args
= args
;
4541 /* Base Declarations */
4542 shader_generate_glsl_declarations(context
, buffer
, shader
, reg_maps
, &priv_ctx
);
4544 /* Base Shader Body */
4545 shader_generate_main(shader
, buffer
, reg_maps
, function
, &priv_ctx
);
4547 /* Unpack outputs */
4548 shader_addline(buffer
, "order_ps_input(vs_out);\n");
4550 /* The D3DRS_FOGTABLEMODE render state defines if the shader-generated fog coord is used
4551 * or if the fragment depth is used. If the fragment depth is used(FOGTABLEMODE != NONE),
4552 * the fog frag coord is thrown away. If the fog frag coord is used, but not written by
4553 * the shader, it is set to 0.0(fully fogged, since start = 1.0, end = 0.0)
4555 if (args
->fog_src
== VS_FOG_Z
)
4556 shader_addline(buffer
, "gl_FogFragCoord = gl_Position.z;\n");
4557 else if (!reg_maps
->fog
)
4558 shader_addline(buffer
, "gl_FogFragCoord = 0.0;\n");
4560 /* We always store the clipplanes without y inversion */
4561 if (args
->clip_enabled
)
4562 shader_addline(buffer
, "gl_ClipVertex = gl_Position;\n");
4564 /* Write the final position.
4566 * OpenGL coordinates specify the center of the pixel while d3d coords specify
4567 * the corner. The offsets are stored in z and w in posFixup. posFixup.y contains
4568 * 1.0 or -1.0 to turn the rendering upside down for offscreen rendering. PosFixup.x
4569 * contains 1.0 to allow a mad.
4571 shader_addline(buffer
, "gl_Position.y = gl_Position.y * posFixup.y;\n");
4572 shader_addline(buffer
, "gl_Position.xy += posFixup.zw * gl_Position.ww;\n");
4574 /* Z coord [0;1]->[-1;1] mapping, see comment in transform_projection in state.c
4576 * Basically we want (in homogeneous coordinates) z = z * 2 - 1. However, shaders are run
4577 * before the homogeneous divide, so we have to take the w into account: z = ((z / w) * 2 - 1) * w,
4578 * which is the same as z = z * 2 - w.
4580 shader_addline(buffer
, "gl_Position.z = gl_Position.z * 2.0 - gl_Position.w;\n");
4582 shader_addline(buffer
, "}\n");
4584 TRACE("Compiling shader object %u\n", shader_obj
);
4585 shader_glsl_compile(gl_info
, shader_obj
, buffer
->buffer
);
4590 /* Context activation is done by the caller. */
4591 static GLhandleARB
shader_glsl_generate_geometry_shader(const struct wined3d_context
*context
,
4592 struct wined3d_shader_buffer
*buffer
, const struct wined3d_shader
*shader
)
4594 const struct wined3d_shader_reg_maps
*reg_maps
= &shader
->reg_maps
;
4595 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
4596 const DWORD
*function
= shader
->function
;
4597 struct shader_glsl_ctx_priv priv_ctx
;
4598 GLhandleARB shader_id
;
4600 shader_id
= GL_EXTCALL(glCreateShaderObjectARB(GL_GEOMETRY_SHADER_ARB
));
4602 shader_addline(buffer
, "#version 120\n");
4604 if (gl_info
->supported
[ARB_GEOMETRY_SHADER4
])
4605 shader_addline(buffer
, "#extension GL_ARB_geometry_shader4 : enable\n");
4606 if (gl_info
->supported
[ARB_SHADER_BIT_ENCODING
])
4607 shader_addline(buffer
, "#extension GL_ARB_shader_bit_encoding : enable\n");
4608 if (gl_info
->supported
[EXT_GPU_SHADER4
])
4609 shader_addline(buffer
, "#extension GL_EXT_gpu_shader4 : enable\n");
4611 memset(&priv_ctx
, 0, sizeof(priv_ctx
));
4612 shader_generate_glsl_declarations(context
, buffer
, shader
, reg_maps
, &priv_ctx
);
4613 shader_generate_main(shader
, buffer
, reg_maps
, function
, &priv_ctx
);
4614 shader_addline(buffer
, "}\n");
4616 TRACE("Compiling shader object %u.\n", shader_id
);
4617 shader_glsl_compile(gl_info
, shader_id
, buffer
->buffer
);
4622 static GLhandleARB
find_glsl_pshader(const struct wined3d_context
*context
,
4623 struct wined3d_shader_buffer
*buffer
, struct wined3d_shader
*shader
,
4624 const struct ps_compile_args
*args
, const struct ps_np2fixup_info
**np2fixup_info
)
4626 struct glsl_ps_compiled_shader
*gl_shaders
, *new_array
;
4627 struct glsl_shader_private
*shader_data
;
4628 struct ps_np2fixup_info
*np2fixup
;
4633 if (!shader
->backend_data
)
4635 shader
->backend_data
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*shader_data
));
4636 if (!shader
->backend_data
)
4638 ERR("Failed to allocate backend data.\n");
4642 shader_data
= shader
->backend_data
;
4643 gl_shaders
= shader_data
->gl_shaders
.ps
;
4645 /* Usually we have very few GL shaders for each d3d shader(just 1 or maybe 2),
4646 * so a linear search is more performant than a hashmap or a binary search
4647 * (cache coherency etc)
4649 for (i
= 0; i
< shader_data
->num_gl_shaders
; ++i
)
4651 if (!memcmp(&gl_shaders
[i
].args
, args
, sizeof(*args
)))
4653 if (args
->np2_fixup
)
4654 *np2fixup_info
= &gl_shaders
[i
].np2fixup
;
4655 return gl_shaders
[i
].prgId
;
4659 TRACE("No matching GL shader found for shader %p, compiling a new shader.\n", shader
);
4660 if(shader_data
->shader_array_size
== shader_data
->num_gl_shaders
) {
4661 if (shader_data
->num_gl_shaders
)
4663 new_size
= shader_data
->shader_array_size
+ max(1, shader_data
->shader_array_size
/ 2);
4664 new_array
= HeapReAlloc(GetProcessHeap(), 0, shader_data
->gl_shaders
.ps
,
4665 new_size
* sizeof(*gl_shaders
));
4669 new_array
= HeapAlloc(GetProcessHeap(), 0, sizeof(*gl_shaders
));
4674 ERR("Out of memory\n");
4677 shader_data
->gl_shaders
.ps
= new_array
;
4678 shader_data
->shader_array_size
= new_size
;
4679 gl_shaders
= new_array
;
4682 gl_shaders
[shader_data
->num_gl_shaders
].args
= *args
;
4684 np2fixup
= &gl_shaders
[shader_data
->num_gl_shaders
].np2fixup
;
4685 memset(np2fixup
, 0, sizeof(*np2fixup
));
4686 *np2fixup_info
= args
->np2_fixup
? np2fixup
: NULL
;
4688 pixelshader_update_samplers(shader
, args
->tex_types
);
4690 shader_buffer_clear(buffer
);
4691 ret
= shader_glsl_generate_pshader(context
, buffer
, shader
, args
, np2fixup
);
4692 gl_shaders
[shader_data
->num_gl_shaders
++].prgId
= ret
;
4697 static inline BOOL
vs_args_equal(const struct vs_compile_args
*stored
, const struct vs_compile_args
*new,
4698 const DWORD use_map
) {
4699 if((stored
->swizzle_map
& use_map
) != new->swizzle_map
) return FALSE
;
4700 if((stored
->clip_enabled
) != new->clip_enabled
) return FALSE
;
4701 return stored
->fog_src
== new->fog_src
;
4704 static GLhandleARB
find_glsl_vshader(const struct wined3d_context
*context
,
4705 struct wined3d_shader_buffer
*buffer
, struct wined3d_shader
*shader
,
4706 const struct vs_compile_args
*args
)
4710 DWORD use_map
= shader
->device
->strided_streams
.use_map
;
4711 struct glsl_vs_compiled_shader
*gl_shaders
, *new_array
;
4712 struct glsl_shader_private
*shader_data
;
4715 if (!shader
->backend_data
)
4717 shader
->backend_data
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*shader_data
));
4718 if (!shader
->backend_data
)
4720 ERR("Failed to allocate backend data.\n");
4724 shader_data
= shader
->backend_data
;
4725 gl_shaders
= shader_data
->gl_shaders
.vs
;
4727 /* Usually we have very few GL shaders for each d3d shader(just 1 or maybe 2),
4728 * so a linear search is more performant than a hashmap or a binary search
4729 * (cache coherency etc)
4731 for (i
= 0; i
< shader_data
->num_gl_shaders
; ++i
)
4733 if (vs_args_equal(&gl_shaders
[i
].args
, args
, use_map
))
4734 return gl_shaders
[i
].prgId
;
4737 TRACE("No matching GL shader found for shader %p, compiling a new shader.\n", shader
);
4739 if(shader_data
->shader_array_size
== shader_data
->num_gl_shaders
) {
4740 if (shader_data
->num_gl_shaders
)
4742 new_size
= shader_data
->shader_array_size
+ max(1, shader_data
->shader_array_size
/ 2);
4743 new_array
= HeapReAlloc(GetProcessHeap(), 0, shader_data
->gl_shaders
.vs
,
4744 new_size
* sizeof(*gl_shaders
));
4748 new_array
= HeapAlloc(GetProcessHeap(), 0, sizeof(*gl_shaders
));
4753 ERR("Out of memory\n");
4756 shader_data
->gl_shaders
.vs
= new_array
;
4757 shader_data
->shader_array_size
= new_size
;
4758 gl_shaders
= new_array
;
4761 gl_shaders
[shader_data
->num_gl_shaders
].args
= *args
;
4763 shader_buffer_clear(buffer
);
4764 ret
= shader_glsl_generate_vshader(context
, buffer
, shader
, args
);
4765 gl_shaders
[shader_data
->num_gl_shaders
++].prgId
= ret
;
4770 static GLhandleARB
find_glsl_geometry_shader(const struct wined3d_context
*context
,
4771 struct wined3d_shader_buffer
*buffer
, struct wined3d_shader
*shader
)
4773 struct glsl_gs_compiled_shader
*gl_shaders
;
4774 struct glsl_shader_private
*shader_data
;
4777 if (!shader
->backend_data
)
4779 if (!(shader
->backend_data
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*shader_data
))))
4781 ERR("Failed to allocate backend data.\n");
4785 shader_data
= shader
->backend_data
;
4786 gl_shaders
= shader_data
->gl_shaders
.gs
;
4788 if (shader_data
->num_gl_shaders
)
4789 return gl_shaders
[0].id
;
4791 TRACE("No matching GL shader found for shader %p, compiling a new shader.\n", shader
);
4793 if (!(shader_data
->gl_shaders
.gs
= HeapAlloc(GetProcessHeap(), 0, sizeof(*gl_shaders
))))
4795 ERR("Failed to allocate GL shader array.\n");
4798 shader_data
->shader_array_size
= 1;
4799 gl_shaders
= shader_data
->gl_shaders
.gs
;
4801 shader_buffer_clear(buffer
);
4802 ret
= shader_glsl_generate_geometry_shader(context
, buffer
, shader
);
4803 gl_shaders
[shader_data
->num_gl_shaders
++].id
= ret
;
4808 static const char *shader_glsl_get_ffp_fragment_op_arg(struct wined3d_shader_buffer
*buffer
,
4809 DWORD argnum
, unsigned int stage
, DWORD arg
)
4813 if (arg
== ARG_UNUSED
)
4814 return "<unused arg>";
4816 switch (arg
& WINED3DTA_SELECTMASK
)
4818 case WINED3DTA_DIFFUSE
:
4822 case WINED3DTA_CURRENT
:
4829 case WINED3DTA_TEXTURE
:
4832 case 0: ret
= "tex0"; break;
4833 case 1: ret
= "tex1"; break;
4834 case 2: ret
= "tex2"; break;
4835 case 3: ret
= "tex3"; break;
4836 case 4: ret
= "tex4"; break;
4837 case 5: ret
= "tex5"; break;
4838 case 6: ret
= "tex6"; break;
4839 case 7: ret
= "tex7"; break;
4841 ret
= "<invalid texture>";
4846 case WINED3DTA_TFACTOR
:
4850 case WINED3DTA_SPECULAR
:
4851 ret
= "gl_SecondaryColor";
4854 case WINED3DTA_TEMP
:
4858 case WINED3DTA_CONSTANT
:
4859 FIXME("Per-stage constants not implemented.\n");
4862 case 0: ret
= "const0"; break;
4863 case 1: ret
= "const1"; break;
4864 case 2: ret
= "const2"; break;
4865 case 3: ret
= "const3"; break;
4866 case 4: ret
= "const4"; break;
4867 case 5: ret
= "const5"; break;
4868 case 6: ret
= "const6"; break;
4869 case 7: ret
= "const7"; break;
4871 ret
= "<invalid constant>";
4877 return "<unhandled arg>";
4880 if (arg
& WINED3DTA_COMPLEMENT
)
4882 shader_addline(buffer
, "arg%u = vec4(1.0) - %s;\n", argnum
, ret
);
4885 else if (argnum
== 1)
4887 else if (argnum
== 2)
4891 if (arg
& WINED3DTA_ALPHAREPLICATE
)
4893 shader_addline(buffer
, "arg%u = vec4(%s.w);\n", argnum
, ret
);
4896 else if (argnum
== 1)
4898 else if (argnum
== 2)
4905 static void shader_glsl_ffp_fragment_op(struct wined3d_shader_buffer
*buffer
, unsigned int stage
, BOOL color
,
4906 BOOL alpha
, DWORD dst
, DWORD op
, DWORD dw_arg0
, DWORD dw_arg1
, DWORD dw_arg2
)
4908 const char *dstmask
, *dstreg
, *arg0
, *arg1
, *arg2
;
4918 dstreg
= "temp_reg";
4922 arg0
= shader_glsl_get_ffp_fragment_op_arg(buffer
, 0, stage
, dw_arg0
);
4923 arg1
= shader_glsl_get_ffp_fragment_op_arg(buffer
, 1, stage
, dw_arg1
);
4924 arg2
= shader_glsl_get_ffp_fragment_op_arg(buffer
, 2, stage
, dw_arg2
);
4928 case WINED3D_TOP_DISABLE
:
4930 shader_addline(buffer
, "%s%s = gl_Color%s;\n", dstreg
, dstmask
, dstmask
);
4933 case WINED3D_TOP_SELECT_ARG1
:
4934 shader_addline(buffer
, "%s%s = %s%s;\n", dstreg
, dstmask
, arg1
, dstmask
);
4937 case WINED3D_TOP_SELECT_ARG2
:
4938 shader_addline(buffer
, "%s%s = %s%s;\n", dstreg
, dstmask
, arg2
, dstmask
);
4941 case WINED3D_TOP_MODULATE
:
4942 shader_addline(buffer
, "%s%s = %s%s * %s%s;\n", dstreg
, dstmask
, arg1
, dstmask
, arg2
, dstmask
);
4945 case WINED3D_TOP_MODULATE_4X
:
4946 shader_addline(buffer
, "%s%s = clamp(%s%s * %s%s * 4.0, 0.0, 1.0);\n",
4947 dstreg
, dstmask
, arg1
, dstmask
, arg2
, dstmask
);
4950 case WINED3D_TOP_MODULATE_2X
:
4951 shader_addline(buffer
, "%s%s = clamp(%s%s * %s%s * 2.0, 0.0, 1.0);\n",
4952 dstreg
, dstmask
, arg1
, dstmask
, arg2
, dstmask
);
4955 case WINED3D_TOP_ADD
:
4956 shader_addline(buffer
, "%s%s = clamp(%s%s + %s%s, 0.0, 1.0);\n",
4957 dstreg
, dstmask
, arg1
, dstmask
, arg2
, dstmask
);
4960 case WINED3D_TOP_ADD_SIGNED
:
4961 shader_addline(buffer
, "%s%s = clamp(%s%s + (%s - vec4(0.5))%s, 0.0, 1.0);\n",
4962 dstreg
, dstmask
, arg1
, dstmask
, arg2
, dstmask
);
4965 case WINED3D_TOP_ADD_SIGNED_2X
:
4966 shader_addline(buffer
, "%s%s = clamp((%s%s + (%s - vec4(0.5))%s) * 2.0, 0.0, 1.0);\n",
4967 dstreg
, dstmask
, arg1
, dstmask
, arg2
, dstmask
);
4970 case WINED3D_TOP_SUBTRACT
:
4971 shader_addline(buffer
, "%s%s = clamp(%s%s - %s%s, 0.0, 1.0);\n",
4972 dstreg
, dstmask
, arg1
, dstmask
, arg2
, dstmask
);
4975 case WINED3D_TOP_ADD_SMOOTH
:
4976 shader_addline(buffer
, "%s%s = clamp((vec4(1.0) - %s)%s * %s%s + %s%s, 0.0, 1.0);\n",
4977 dstreg
, dstmask
, arg1
, dstmask
, arg2
, dstmask
, arg1
, dstmask
);
4980 case WINED3D_TOP_BLEND_DIFFUSE_ALPHA
:
4981 arg0
= shader_glsl_get_ffp_fragment_op_arg(buffer
, 0, stage
, WINED3DTA_DIFFUSE
);
4982 shader_addline(buffer
, "%s%s = mix(%s%s, %s%s, %s.w);\n",
4983 dstreg
, dstmask
, arg2
, dstmask
, arg1
, dstmask
, arg0
);
4986 case WINED3D_TOP_BLEND_TEXTURE_ALPHA
:
4987 arg0
= shader_glsl_get_ffp_fragment_op_arg(buffer
, 0, stage
, WINED3DTA_TEXTURE
);
4988 shader_addline(buffer
, "%s%s = mix(%s%s, %s%s, %s.w);\n",
4989 dstreg
, dstmask
, arg2
, dstmask
, arg1
, dstmask
, arg0
);
4992 case WINED3D_TOP_BLEND_FACTOR_ALPHA
:
4993 arg0
= shader_glsl_get_ffp_fragment_op_arg(buffer
, 0, stage
, WINED3DTA_TFACTOR
);
4994 shader_addline(buffer
, "%s%s = mix(%s%s, %s%s, %s.w);\n",
4995 dstreg
, dstmask
, arg2
, dstmask
, arg1
, dstmask
, arg0
);
4998 case WINED3D_TOP_BLEND_TEXTURE_ALPHA_PM
:
4999 arg0
= shader_glsl_get_ffp_fragment_op_arg(buffer
, 0, stage
, WINED3DTA_TEXTURE
);
5000 shader_addline(buffer
, "%s%s = clamp(%s%s * (1.0 - %s.w) + %s%s, 0.0, 1.0);\n",
5001 dstreg
, dstmask
, arg2
, dstmask
, arg0
, arg1
, dstmask
);
5004 case WINED3D_TOP_BLEND_CURRENT_ALPHA
:
5005 arg0
= shader_glsl_get_ffp_fragment_op_arg(buffer
, 0, stage
, WINED3DTA_CURRENT
);
5006 shader_addline(buffer
, "%s%s = mix(%s%s, %s%s, %s.w);\n",
5007 dstreg
, dstmask
, arg2
, dstmask
, arg1
, dstmask
, arg0
);
5010 case WINED3D_TOP_MODULATE_ALPHA_ADD_COLOR
:
5011 shader_addline(buffer
, "%s%s = clamp(%s%s * %s.w + %s%s, 0.0, 1.0);\n",
5012 dstreg
, dstmask
, arg2
, dstmask
, arg1
, arg1
, dstmask
);
5015 case WINED3D_TOP_MODULATE_COLOR_ADD_ALPHA
:
5016 shader_addline(buffer
, "%s%s = clamp(%s%s * %s%s + %s.w, 0.0, 1.0);\n",
5017 dstreg
, dstmask
, arg1
, dstmask
, arg2
, dstmask
, arg1
);
5020 case WINED3D_TOP_MODULATE_INVALPHA_ADD_COLOR
:
5021 shader_addline(buffer
, "%s%s = clamp(%s%s * (1.0 - %s.w) + %s%s, 0.0, 1.0);\n",
5022 dstreg
, dstmask
, arg2
, dstmask
, arg1
, arg1
, dstmask
);
5024 case WINED3D_TOP_MODULATE_INVCOLOR_ADD_ALPHA
:
5025 shader_addline(buffer
, "%s%s = clamp((vec4(1.0) - %s)%s * %s%s + %s.w, 0.0, 1.0);\n",
5026 dstreg
, dstmask
, arg1
, dstmask
, arg2
, dstmask
, arg1
);
5029 case WINED3D_TOP_BUMPENVMAP
:
5030 case WINED3D_TOP_BUMPENVMAP_LUMINANCE
:
5031 /* These are handled in the first pass, nothing to do. */
5034 case WINED3D_TOP_DOTPRODUCT3
:
5035 shader_addline(buffer
, "%s%s = vec4(clamp(dot(%s.xyz - 0.5, %s.xyz - 0.5) * 4.0, 0.0, 1.0))%s;\n",
5036 dstreg
, dstmask
, arg1
, arg2
, dstmask
);
5039 case WINED3D_TOP_MULTIPLY_ADD
:
5040 shader_addline(buffer
, "%s%s = clamp(%s%s * %s%s + %s%s, 0.0, 1.0);\n",
5041 dstreg
, dstmask
, arg1
, dstmask
, arg2
, dstmask
, arg0
, dstmask
);
5044 case WINED3D_TOP_LERP
:
5045 /* MSDN isn't quite right here. */
5046 shader_addline(buffer
, "%s%s = mix(%s%s, %s%s, %s%s);\n",
5047 dstreg
, dstmask
, arg2
, dstmask
, arg1
, dstmask
, arg0
, dstmask
);
5051 FIXME("Unhandled operation %#x.\n", op
);
5056 /* Context activation is done by the caller. */
5057 static GLuint
shader_glsl_generate_ffp_fragment_shader(struct wined3d_shader_buffer
*buffer
,
5058 const struct ffp_frag_settings
*settings
, const struct wined3d_gl_info
*gl_info
)
5060 BOOL tempreg_used
= FALSE
, tfactor_used
= FALSE
;
5061 BYTE lum_map
= 0, bump_map
= 0, tex_map
= 0;
5062 const char *final_combiner_src
= "ret";
5063 UINT lowest_disabled_stage
;
5064 GLhandleARB shader_obj
;
5065 DWORD arg0
, arg1
, arg2
;
5068 shader_buffer_clear(buffer
);
5070 /* Find out which textures are read */
5071 for (stage
= 0; stage
< MAX_TEXTURES
; ++stage
)
5073 if (settings
->op
[stage
].cop
== WINED3D_TOP_DISABLE
)
5076 arg0
= settings
->op
[stage
].carg0
& WINED3DTA_SELECTMASK
;
5077 arg1
= settings
->op
[stage
].carg1
& WINED3DTA_SELECTMASK
;
5078 arg2
= settings
->op
[stage
].carg2
& WINED3DTA_SELECTMASK
;
5080 if (arg0
== WINED3DTA_TEXTURE
|| arg1
== WINED3DTA_TEXTURE
|| arg2
== WINED3DTA_TEXTURE
)
5081 tex_map
|= 1 << stage
;
5082 if (arg0
== WINED3DTA_TFACTOR
|| arg1
== WINED3DTA_TFACTOR
|| arg2
== WINED3DTA_TFACTOR
)
5083 tfactor_used
= TRUE
;
5084 if (arg0
== WINED3DTA_TEMP
|| arg1
== WINED3DTA_TEMP
|| arg2
== WINED3DTA_TEMP
)
5085 tempreg_used
= TRUE
;
5086 if (settings
->op
[stage
].dst
== tempreg
)
5087 tempreg_used
= TRUE
;
5089 switch (settings
->op
[stage
].cop
)
5091 case WINED3D_TOP_BUMPENVMAP_LUMINANCE
:
5092 lum_map
|= 1 << stage
;
5094 case WINED3D_TOP_BUMPENVMAP
:
5095 bump_map
|= 1 << stage
;
5097 case WINED3D_TOP_BLEND_TEXTURE_ALPHA
:
5098 case WINED3D_TOP_BLEND_TEXTURE_ALPHA_PM
:
5099 tex_map
|= 1 << stage
;
5102 case WINED3D_TOP_BLEND_FACTOR_ALPHA
:
5103 tfactor_used
= TRUE
;
5110 if (settings
->op
[stage
].aop
== WINED3D_TOP_DISABLE
)
5113 arg0
= settings
->op
[stage
].aarg0
& WINED3DTA_SELECTMASK
;
5114 arg1
= settings
->op
[stage
].aarg1
& WINED3DTA_SELECTMASK
;
5115 arg2
= settings
->op
[stage
].aarg2
& WINED3DTA_SELECTMASK
;
5117 if (arg0
== WINED3DTA_TEXTURE
|| arg1
== WINED3DTA_TEXTURE
|| arg2
== WINED3DTA_TEXTURE
)
5118 tex_map
|= 1 << stage
;
5119 if (arg0
== WINED3DTA_TFACTOR
|| arg1
== WINED3DTA_TFACTOR
|| arg2
== WINED3DTA_TFACTOR
)
5120 tfactor_used
= TRUE
;
5121 if (arg0
== WINED3DTA_TEMP
|| arg1
== WINED3DTA_TEMP
|| arg2
== WINED3DTA_TEMP
)
5122 tempreg_used
= TRUE
;
5124 lowest_disabled_stage
= stage
;
5126 shader_addline(buffer
, "#version 120\n");
5128 shader_addline(buffer
, "vec4 tmp0, tmp1;\n");
5129 shader_addline(buffer
, "vec4 ret;\n");
5130 if (tempreg_used
|| settings
->sRGB_write
)
5131 shader_addline(buffer
, "vec4 temp_reg;\n");
5132 shader_addline(buffer
, "vec4 arg0, arg1, arg2;\n");
5134 for (stage
= 0; stage
< MAX_TEXTURES
; ++stage
)
5136 if (!(tex_map
& (1 << stage
)))
5139 switch (settings
->op
[stage
].tex_type
)
5142 shader_addline(buffer
, "uniform sampler1D ps_sampler%u;\n", stage
);
5145 shader_addline(buffer
, "uniform sampler2D ps_sampler%u;\n", stage
);
5148 shader_addline(buffer
, "uniform sampler3D ps_sampler%u;\n", stage
);
5151 shader_addline(buffer
, "uniform samplerCube ps_sampler%u;\n", stage
);
5154 shader_addline(buffer
, "uniform sampler2DRect ps_sampler%u;\n", stage
);
5157 FIXME("Unhandled sampler type %#x.\n", settings
->op
[stage
].tex_type
);
5161 shader_addline(buffer
, "vec4 tex%u;\n", stage
);
5163 if (!(bump_map
& (1 << stage
)))
5165 shader_addline(buffer
, "uniform mat2 bumpenv_mat%u;\n", stage
);
5167 if (!(lum_map
& (1 << stage
)))
5169 shader_addline(buffer
, "uniform float bumpenv_lum_scale%u;\n", stage
);
5170 shader_addline(buffer
, "uniform float bumpenv_lum_offset%u;\n", stage
);
5173 shader_addline(buffer
, "uniform vec4 tex_factor;\n");
5174 shader_addline(buffer
, "uniform vec4 specular_enable;\n");
5176 if (settings
->sRGB_write
)
5178 shader_addline(buffer
, "const vec4 srgb_const0 = vec4(%.8e, %.8e, %.8e, %.8e);\n",
5179 srgb_pow
, srgb_mul_high
, srgb_sub_high
, srgb_mul_low
);
5180 shader_addline(buffer
, "const vec4 srgb_const1 = vec4(%.8e, 0.0, 0.0, 0.0);\n",
5184 shader_addline(buffer
, "void main()\n{\n");
5186 if (lowest_disabled_stage
< 7 && settings
->emul_clipplanes
)
5187 shader_addline(buffer
, "if (any(lessThan(gl_texCoord[7], vec4(0.0)))) discard;\n");
5189 /* Generate texture sampling instructions) */
5190 for (stage
= 0; stage
< MAX_TEXTURES
&& settings
->op
[stage
].cop
!= WINED3D_TOP_DISABLE
; ++stage
)
5192 const char *texture_function
, *coord_mask
;
5193 char tex_reg_name
[8];
5196 if (!(tex_map
& (1 << stage
)))
5199 if (settings
->op
[stage
].projected
== proj_none
)
5203 else if (settings
->op
[stage
].projected
== proj_count4
5204 || settings
->op
[stage
].projected
== proj_count3
)
5210 FIXME("Unexpected projection mode %d\n", settings
->op
[stage
].projected
);
5214 if (settings
->op
[stage
].cop
== WINED3D_TOP_BUMPENVMAP
5215 || settings
->op
[stage
].cop
== WINED3D_TOP_BUMPENVMAP_LUMINANCE
)
5220 switch (settings
->op
[stage
].tex_type
)
5225 texture_function
= "texture1DProj";
5230 texture_function
= "texture1D";
5237 texture_function
= "texture2DProj";
5242 texture_function
= "texture2D";
5249 texture_function
= "texture3DProj";
5250 coord_mask
= "xyzw";
5254 texture_function
= "texture3D";
5259 texture_function
= "textureCube";
5265 texture_function
= "texture2DRectProj";
5270 texture_function
= "texture2DRect";
5275 FIXME("Unhandled texture type %#x.\n", settings
->op
[stage
].tex_type
);
5276 texture_function
= "";
5277 coord_mask
= "xyzw";
5282 && (settings
->op
[stage
- 1].cop
== WINED3D_TOP_BUMPENVMAP
5283 || settings
->op
[stage
- 1].cop
== WINED3D_TOP_BUMPENVMAP_LUMINANCE
))
5285 shader_addline(buffer
, "ret.xy = bumpenv_mat%u * tex%u.xy;\n", stage
- 1, stage
- 1);
5287 /* With projective textures, texbem only divides the static
5288 * texture coord, not the displacement, so multiply the
5289 * displacement with the dividing parameter before passing it to
5291 if (settings
->op
[stage
].projected
!= proj_none
)
5293 if (settings
->op
[stage
].projected
== proj_count4
)
5295 shader_addline(buffer
, "ret.xy = (ret.xy * gl_TexCoord[%u].w) + gl_TexCoord[%u].xy;\n",
5297 shader_addline(buffer
, "ret.zw = gl_TexCoord[%u].ww;\n", stage
);
5301 shader_addline(buffer
, "ret.xy = (ret.xy * gl_TexCoord[%u].z) + gl_TexCoord[%u].xy;\n",
5303 shader_addline(buffer
, "ret.zw = gl_TexCoord[%u].zz;\n", stage
);
5308 shader_addline(buffer
, "ret = gl_TexCoord[%u] + ret.xyxy;\n", stage
);
5312 shader_addline(buffer
, "tex%u = clamp(%s(ps_sampler%u, ret.%s), 0.0, 1.0);\n",
5313 stage
, texture_function
, stage
, coord_mask
);
5315 shader_addline(buffer
, "tex%u = %s(ps_sampler%u, ret.%s);\n",
5316 stage
, texture_function
, stage
, coord_mask
);
5318 if (settings
->op
[stage
- 1].cop
== WINED3D_TOP_BUMPENVMAP_LUMINANCE
)
5319 shader_addline(buffer
, "tex%u *= clamp(tex%u.z * bumpenv_lum_scale%u + bumpenv_lum_offset%u, 0.0, 1.0);\n",
5320 stage
, stage
- 1, stage
- 1, stage
- 1);
5322 else if (settings
->op
[stage
].projected
== proj_count3
)
5325 shader_addline(buffer
, "tex%u = clamp(%s(ps_sampler%u, gl_TexCoord[%u].xyz), 0.0, 1.0);\n",
5326 stage
, texture_function
, stage
, stage
);
5328 shader_addline(buffer
, "tex%u = %s(ps_sampler%u, gl_TexCoord[%u].xyz);\n",
5329 stage
, texture_function
, stage
, stage
);
5334 shader_addline(buffer
, "tex%u = clamp(%s(ps_sampler%u, gl_TexCoord[%u].%s), 0.0, 1.0);\n",
5335 stage
, texture_function
, stage
, stage
, coord_mask
);
5337 shader_addline(buffer
, "tex%u = %s(ps_sampler%u, gl_TexCoord[%u].%s);\n",
5338 stage
, texture_function
, stage
, stage
, coord_mask
);
5341 sprintf(tex_reg_name
, "tex%u", stage
);
5342 shader_glsl_color_correction_ext(buffer
, tex_reg_name
, WINED3DSP_WRITEMASK_ALL
,
5343 settings
->op
[stage
].color_fixup
);
5346 /* Generate the main shader */
5347 for (stage
= 0; stage
< MAX_TEXTURES
; ++stage
)
5351 if (settings
->op
[stage
].cop
== WINED3D_TOP_DISABLE
)
5354 final_combiner_src
= "gl_Color";
5358 if (settings
->op
[stage
].cop
== WINED3D_TOP_SELECT_ARG1
5359 && settings
->op
[stage
].aop
== WINED3D_TOP_SELECT_ARG1
)
5360 op_equal
= settings
->op
[stage
].carg1
== settings
->op
[stage
].aarg1
;
5361 else if (settings
->op
[stage
].cop
== WINED3D_TOP_SELECT_ARG1
5362 && settings
->op
[stage
].aop
== WINED3D_TOP_SELECT_ARG2
)
5363 op_equal
= settings
->op
[stage
].carg1
== settings
->op
[stage
].aarg2
;
5364 else if (settings
->op
[stage
].cop
== WINED3D_TOP_SELECT_ARG2
5365 && settings
->op
[stage
].aop
== WINED3D_TOP_SELECT_ARG1
)
5366 op_equal
= settings
->op
[stage
].carg2
== settings
->op
[stage
].aarg1
;
5367 else if (settings
->op
[stage
].cop
== WINED3D_TOP_SELECT_ARG2
5368 && settings
->op
[stage
].aop
== WINED3D_TOP_SELECT_ARG2
)
5369 op_equal
= settings
->op
[stage
].carg2
== settings
->op
[stage
].aarg2
;
5371 op_equal
= settings
->op
[stage
].aop
== settings
->op
[stage
].cop
5372 && settings
->op
[stage
].carg0
== settings
->op
[stage
].aarg0
5373 && settings
->op
[stage
].carg1
== settings
->op
[stage
].aarg1
5374 && settings
->op
[stage
].carg2
== settings
->op
[stage
].aarg2
;
5376 if (settings
->op
[stage
].aop
== WINED3D_TOP_DISABLE
)
5378 shader_glsl_ffp_fragment_op(buffer
, stage
, TRUE
, FALSE
, settings
->op
[stage
].dst
,
5379 settings
->op
[stage
].cop
, settings
->op
[stage
].carg0
,
5380 settings
->op
[stage
].carg1
, settings
->op
[stage
].carg2
);
5382 shader_addline(buffer
, "ret.w = gl_Color.w;\n");
5386 shader_glsl_ffp_fragment_op(buffer
, stage
, TRUE
, TRUE
, settings
->op
[stage
].dst
,
5387 settings
->op
[stage
].cop
, settings
->op
[stage
].carg0
,
5388 settings
->op
[stage
].carg1
, settings
->op
[stage
].carg2
);
5392 shader_glsl_ffp_fragment_op(buffer
, stage
, TRUE
, FALSE
, settings
->op
[stage
].dst
,
5393 settings
->op
[stage
].cop
, settings
->op
[stage
].carg0
,
5394 settings
->op
[stage
].carg1
, settings
->op
[stage
].carg2
);
5395 shader_glsl_ffp_fragment_op(buffer
, stage
, FALSE
, TRUE
, settings
->op
[stage
].dst
,
5396 settings
->op
[stage
].aop
, settings
->op
[stage
].aarg0
,
5397 settings
->op
[stage
].aarg1
, settings
->op
[stage
].aarg2
);
5401 shader_addline(buffer
, "gl_FragData[0] = gl_SecondaryColor * specular_enable + %s;\n", final_combiner_src
);
5403 if (settings
->sRGB_write
)
5404 shader_glsl_generate_srgb_write_correction(buffer
);
5406 shader_glsl_generate_fog_code(buffer
, settings
->fog
);
5408 shader_addline(buffer
, "}\n");
5410 shader_obj
= GL_EXTCALL(glCreateShaderObjectARB(GL_FRAGMENT_SHADER_ARB
));
5411 shader_glsl_compile(gl_info
, shader_obj
, buffer
->buffer
);
5415 static struct glsl_ffp_fragment_shader
*shader_glsl_find_ffp_fragment_shader(struct shader_glsl_priv
*priv
,
5416 const struct wined3d_gl_info
*gl_info
, const struct ffp_frag_settings
*args
)
5418 struct glsl_ffp_fragment_shader
*glsl_desc
;
5419 const struct ffp_frag_desc
*desc
;
5421 if ((desc
= find_ffp_frag_shader(&priv
->ffp_fragment_shaders
, args
)))
5422 return CONTAINING_RECORD(desc
, struct glsl_ffp_fragment_shader
, entry
);
5424 if (!(glsl_desc
= HeapAlloc(GetProcessHeap(), 0, sizeof(*glsl_desc
))))
5427 glsl_desc
->entry
.settings
= *args
;
5428 glsl_desc
->id
= shader_glsl_generate_ffp_fragment_shader(&priv
->shader_buffer
, args
, gl_info
);
5429 list_init(&glsl_desc
->linked_programs
);
5430 add_ffp_frag_shader(&priv
->ffp_fragment_shaders
, &glsl_desc
->entry
);
5436 static void shader_glsl_init_vs_uniform_locations(const struct wined3d_gl_info
*gl_info
,
5437 GLhandleARB program_id
, struct glsl_vs_program
*vs
)
5442 vs
->uniform_f_locations
= HeapAlloc(GetProcessHeap(), 0,
5443 sizeof(GLhandleARB
) * gl_info
->limits
.glsl_vs_float_constants
);
5444 for (i
= 0; i
< gl_info
->limits
.glsl_vs_float_constants
; ++i
)
5446 snprintf(name
, sizeof(name
), "vs_c[%u]", i
);
5447 vs
->uniform_f_locations
[i
] = GL_EXTCALL(glGetUniformLocationARB(program_id
, name
));
5450 for (i
= 0; i
< MAX_CONST_I
; ++i
)
5452 snprintf(name
, sizeof(name
), "vs_i[%u]", i
);
5453 vs
->uniform_i_locations
[i
] = GL_EXTCALL(glGetUniformLocationARB(program_id
, name
));
5456 vs
->pos_fixup_location
= GL_EXTCALL(glGetUniformLocationARB(program_id
, "posFixup"));
5459 static void shader_glsl_init_ps_uniform_locations(const struct wined3d_gl_info
*gl_info
,
5460 GLhandleARB program_id
, struct glsl_ps_program
*ps
)
5465 ps
->uniform_f_locations
= HeapAlloc(GetProcessHeap(), 0,
5466 sizeof(GLhandleARB
) * gl_info
->limits
.glsl_ps_float_constants
);
5467 for (i
= 0; i
< gl_info
->limits
.glsl_ps_float_constants
; ++i
)
5469 snprintf(name
, sizeof(name
), "ps_c[%u]", i
);
5470 ps
->uniform_f_locations
[i
] = GL_EXTCALL(glGetUniformLocationARB(program_id
, name
));
5473 for (i
= 0; i
< MAX_CONST_I
; ++i
)
5475 snprintf(name
, sizeof(name
), "ps_i[%u]", i
);
5476 ps
->uniform_i_locations
[i
] = GL_EXTCALL(glGetUniformLocationARB(program_id
, name
));
5479 for (i
= 0; i
< MAX_TEXTURES
; ++i
)
5481 snprintf(name
, sizeof(name
), "bumpenv_mat%u", i
);
5482 ps
->bumpenv_mat_location
[i
] = GL_EXTCALL(glGetUniformLocationARB(program_id
, name
));
5483 snprintf(name
, sizeof(name
), "bumpenv_lum_scale%u", i
);
5484 ps
->bumpenv_lum_scale_location
[i
] = GL_EXTCALL(glGetUniformLocationARB(program_id
, name
));
5485 snprintf(name
, sizeof(name
), "bumpenv_lum_offset%u", i
);
5486 ps
->bumpenv_lum_offset_location
[i
] = GL_EXTCALL(glGetUniformLocationARB(program_id
, name
));
5489 ps
->tex_factor_location
= GL_EXTCALL(glGetUniformLocationARB(program_id
, "tex_factor"));
5490 ps
->specular_enable_location
= GL_EXTCALL(glGetUniformLocationARB(program_id
, "specular_enable"));
5491 ps
->np2_fixup_location
= GL_EXTCALL(glGetUniformLocationARB(program_id
, "ps_samplerNP2Fixup"));
5492 ps
->ycorrection_location
= GL_EXTCALL(glGetUniformLocationARB(program_id
, "ycorrection"));
5495 /* Context activation is done by the caller. */
5496 static void set_glsl_shader_program(const struct wined3d_context
*context
, struct wined3d_device
*device
,
5497 enum wined3d_shader_mode vertex_mode
, enum wined3d_shader_mode fragment_mode
)
5499 const struct wined3d_state
*state
= &device
->stateBlock
->state
;
5500 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
5501 const struct ps_np2fixup_info
*np2fixup_info
= NULL
;
5502 struct shader_glsl_priv
*priv
= device
->shader_priv
;
5503 struct glsl_shader_prog_link
*entry
= NULL
;
5504 struct wined3d_shader
*vshader
= NULL
;
5505 struct wined3d_shader
*gshader
= NULL
;
5506 struct wined3d_shader
*pshader
= NULL
;
5507 GLhandleARB programId
= 0;
5508 GLhandleARB reorder_shader_id
= 0;
5510 struct ps_compile_args ps_compile_args
;
5511 struct vs_compile_args vs_compile_args
;
5512 GLhandleARB vs_id
, gs_id
, ps_id
;
5513 struct list
*ps_list
;
5515 if (vertex_mode
== WINED3D_SHADER_MODE_SHADER
)
5517 vshader
= state
->vertex_shader
;
5518 find_vs_compile_args(state
, vshader
, &vs_compile_args
);
5519 vs_id
= find_glsl_vshader(context
, &priv
->shader_buffer
, vshader
, &vs_compile_args
);
5521 if ((gshader
= state
->geometry_shader
))
5522 gs_id
= find_glsl_geometry_shader(context
, &priv
->shader_buffer
, gshader
);
5532 if (fragment_mode
== WINED3D_SHADER_MODE_SHADER
)
5534 pshader
= state
->pixel_shader
;
5535 find_ps_compile_args(state
, pshader
, &ps_compile_args
);
5536 ps_id
= find_glsl_pshader(context
, &priv
->shader_buffer
,
5537 pshader
, &ps_compile_args
, &np2fixup_info
);
5538 ps_list
= &pshader
->linked_programs
;
5540 else if (fragment_mode
== WINED3D_SHADER_MODE_FFP
&& priv
->fragment_pipe
== &glsl_fragment_pipe
)
5542 struct glsl_ffp_fragment_shader
*ffp_shader
;
5543 struct ffp_frag_settings settings
;
5545 gen_ffp_frag_op(device
, state
, &settings
, FALSE
);
5546 ffp_shader
= shader_glsl_find_ffp_fragment_shader(priv
, gl_info
, &settings
);
5547 ps_id
= ffp_shader
->id
;
5548 ps_list
= &ffp_shader
->linked_programs
;
5555 if ((!vs_id
&& !gs_id
&& !ps_id
) || (entry
= get_glsl_program_entry(priv
, vs_id
, gs_id
, ps_id
)))
5557 priv
->glsl_program
= entry
;
5561 /* If we get to this point, then no matching program exists, so we create one */
5562 programId
= GL_EXTCALL(glCreateProgramObjectARB());
5563 TRACE("Created new GLSL shader program %u\n", programId
);
5565 /* Create the entry */
5566 entry
= HeapAlloc(GetProcessHeap(), 0, sizeof(struct glsl_shader_prog_link
));
5567 entry
->programId
= programId
;
5568 entry
->vs
.id
= vs_id
;
5569 entry
->gs
.id
= gs_id
;
5570 entry
->ps
.id
= ps_id
;
5571 entry
->constant_version
= 0;
5572 entry
->ps
.np2_fixup_info
= np2fixup_info
;
5573 /* Add the hash table entry */
5574 add_glsl_program_entry(priv
, entry
);
5576 /* Set the current program */
5577 priv
->glsl_program
= entry
;
5579 /* Attach GLSL vshader */
5582 WORD map
= vshader
->reg_maps
.input_registers
;
5585 reorder_shader_id
= generate_param_reorder_function(&priv
->shader_buffer
, vshader
, pshader
, gl_info
);
5586 TRACE("Attaching GLSL shader object %u to program %u\n", reorder_shader_id
, programId
);
5587 GL_EXTCALL(glAttachObjectARB(programId
, reorder_shader_id
));
5588 checkGLcall("glAttachObjectARB");
5589 /* Flag the reorder function for deletion, then it will be freed automatically when the program
5592 GL_EXTCALL(glDeleteObjectARB(reorder_shader_id
));
5594 TRACE("Attaching GLSL shader object %u to program %u.\n", vs_id
, programId
);
5595 GL_EXTCALL(glAttachObjectARB(programId
, vs_id
));
5596 checkGLcall("glAttachObjectARB");
5598 /* Bind vertex attributes to a corresponding index number to match
5599 * the same index numbers as ARB_vertex_programs (makes loading
5600 * vertex attributes simpler). With this method, we can use the
5601 * exact same code to load the attributes later for both ARB and
5604 * We have to do this here because we need to know the Program ID
5605 * in order to make the bindings work, and it has to be done prior
5606 * to linking the GLSL program. */
5607 for (i
= 0; map
; map
>>= 1, ++i
)
5609 if (!(map
& 1)) continue;
5611 snprintf(tmp_name
, sizeof(tmp_name
), "vs_in%u", i
);
5612 GL_EXTCALL(glBindAttribLocationARB(programId
, i
, tmp_name
));
5614 checkGLcall("glBindAttribLocationARB");
5616 list_add_head(&vshader
->linked_programs
, &entry
->vs
.shader_entry
);
5621 TRACE("Attaching GLSL geometry shader object %u to program %u.\n", gs_id
, programId
);
5622 GL_EXTCALL(glAttachObjectARB(programId
, gs_id
));
5623 checkGLcall("glAttachObjectARB");
5625 TRACE("input type %s, output type %s, vertices out %u.\n",
5626 debug_d3dprimitivetype(gshader
->u
.gs
.input_type
),
5627 debug_d3dprimitivetype(gshader
->u
.gs
.output_type
),
5628 gshader
->u
.gs
.vertices_out
);
5629 GL_EXTCALL(glProgramParameteriARB(programId
, GL_GEOMETRY_INPUT_TYPE_ARB
,
5630 gl_primitive_type_from_d3d(gshader
->u
.gs
.input_type
)));
5631 GL_EXTCALL(glProgramParameteriARB(programId
, GL_GEOMETRY_OUTPUT_TYPE_ARB
,
5632 gl_primitive_type_from_d3d(gshader
->u
.gs
.output_type
)));
5633 GL_EXTCALL(glProgramParameteriARB(programId
, GL_GEOMETRY_VERTICES_OUT_ARB
,
5634 gshader
->u
.gs
.vertices_out
));
5635 checkGLcall("glProgramParameteriARB");
5637 list_add_head(&gshader
->linked_programs
, &entry
->gs
.shader_entry
);
5640 /* Attach GLSL pshader */
5643 TRACE("Attaching GLSL shader object %u to program %u.\n", ps_id
, programId
);
5644 GL_EXTCALL(glAttachObjectARB(programId
, ps_id
));
5645 checkGLcall("glAttachObjectARB");
5647 list_add_head(ps_list
, &entry
->ps
.shader_entry
);
5650 /* Link the program */
5651 TRACE("Linking GLSL shader program %u\n", programId
);
5652 GL_EXTCALL(glLinkProgramARB(programId
));
5653 shader_glsl_validate_link(gl_info
, programId
);
5655 shader_glsl_init_vs_uniform_locations(gl_info
, programId
, &entry
->vs
);
5656 shader_glsl_init_ps_uniform_locations(gl_info
, programId
, &entry
->ps
);
5657 checkGLcall("Find glsl program uniform locations");
5659 if (pshader
&& pshader
->reg_maps
.shader_version
.major
>= 3
5660 && pshader
->u
.ps
.declared_in_count
> vec4_varyings(3, gl_info
))
5662 TRACE("Shader %d needs vertex color clamping disabled\n", programId
);
5663 entry
->vs
.vertex_color_clamp
= GL_FALSE
;
5667 entry
->vs
.vertex_color_clamp
= GL_FIXED_ONLY_ARB
;
5670 /* Set the shader to allow uniform loading on it */
5671 GL_EXTCALL(glUseProgramObjectARB(programId
));
5672 checkGLcall("glUseProgramObjectARB(programId)");
5674 /* Load the vertex and pixel samplers now. The function that finds the mappings makes sure
5675 * that it stays the same for each vertexshader-pixelshader pair(=linked glsl program). If
5676 * a pshader with fixed function pipeline is used there are no vertex samplers, and if a
5677 * vertex shader with fixed function pixel processing is used we make sure that the card
5678 * supports enough samplers to allow the max number of vertex samplers with all possible
5679 * fixed function fragment processing setups. So once the program is linked these samplers
5682 shader_glsl_load_vsamplers(gl_info
, device
->texUnitMap
, programId
);
5683 shader_glsl_load_psamplers(gl_info
, device
->texUnitMap
, programId
);
5686 /* Context activation is done by the caller. */
5687 static GLhandleARB
create_glsl_blt_shader(const struct wined3d_gl_info
*gl_info
, enum tex_types tex_type
, BOOL masked
)
5689 GLhandleARB program_id
;
5690 GLhandleARB vshader_id
, pshader_id
;
5691 const char *blt_pshader
;
5693 static const char *blt_vshader
=
5697 " gl_Position = gl_Vertex;\n"
5698 " gl_FrontColor = vec4(1.0);\n"
5699 " gl_TexCoord[0] = gl_MultiTexCoord0;\n"
5702 static const char * const blt_pshaders_full
[tex_type_count
] =
5708 "uniform sampler2D sampler;\n"
5711 " gl_FragDepth = texture2D(sampler, gl_TexCoord[0].xy).x;\n"
5717 "uniform samplerCube sampler;\n"
5720 " gl_FragDepth = textureCube(sampler, gl_TexCoord[0].xyz).x;\n"
5724 "#extension GL_ARB_texture_rectangle : enable\n"
5725 "uniform sampler2DRect sampler;\n"
5728 " gl_FragDepth = texture2DRect(sampler, gl_TexCoord[0].xy).x;\n"
5732 static const char * const blt_pshaders_masked
[tex_type_count
] =
5738 "uniform sampler2D sampler;\n"
5739 "uniform vec4 mask;\n"
5742 " if (all(lessThan(gl_FragCoord.xy, mask.zw))) discard;\n"
5743 " gl_FragDepth = texture2D(sampler, gl_TexCoord[0].xy).x;\n"
5749 "uniform samplerCube sampler;\n"
5750 "uniform vec4 mask;\n"
5753 " if (all(lessThan(gl_FragCoord.xy, mask.zw))) discard;\n"
5754 " gl_FragDepth = textureCube(sampler, gl_TexCoord[0].xyz).x;\n"
5758 "#extension GL_ARB_texture_rectangle : enable\n"
5759 "uniform sampler2DRect sampler;\n"
5760 "uniform vec4 mask;\n"
5763 " if (all(lessThan(gl_FragCoord.xy, mask.zw))) discard;\n"
5764 " gl_FragDepth = texture2DRect(sampler, gl_TexCoord[0].xy).x;\n"
5768 blt_pshader
= masked
? blt_pshaders_masked
[tex_type
] : blt_pshaders_full
[tex_type
];
5771 FIXME("tex_type %#x not supported\n", tex_type
);
5775 vshader_id
= GL_EXTCALL(glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB
));
5776 shader_glsl_compile(gl_info
, vshader_id
, blt_vshader
);
5778 pshader_id
= GL_EXTCALL(glCreateShaderObjectARB(GL_FRAGMENT_SHADER_ARB
));
5779 shader_glsl_compile(gl_info
, pshader_id
, blt_pshader
);
5781 program_id
= GL_EXTCALL(glCreateProgramObjectARB());
5782 GL_EXTCALL(glAttachObjectARB(program_id
, vshader_id
));
5783 GL_EXTCALL(glAttachObjectARB(program_id
, pshader_id
));
5784 GL_EXTCALL(glLinkProgramARB(program_id
));
5786 shader_glsl_validate_link(gl_info
, program_id
);
5788 /* Once linked we can mark the shaders for deletion. They will be deleted once the program
5791 GL_EXTCALL(glDeleteObjectARB(vshader_id
));
5792 GL_EXTCALL(glDeleteObjectARB(pshader_id
));
5796 /* Context activation is done by the caller. */
5797 static void shader_glsl_select(const struct wined3d_context
*context
, enum wined3d_shader_mode vertex_mode
,
5798 enum wined3d_shader_mode fragment_mode
)
5800 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
5801 struct wined3d_device
*device
= context
->swapchain
->device
;
5802 struct shader_glsl_priv
*priv
= device
->shader_priv
;
5803 GLhandleARB program_id
= 0;
5804 GLenum old_vertex_color_clamp
, current_vertex_color_clamp
;
5806 old_vertex_color_clamp
= priv
->glsl_program
? priv
->glsl_program
->vs
.vertex_color_clamp
: GL_FIXED_ONLY_ARB
;
5807 set_glsl_shader_program(context
, device
, vertex_mode
, fragment_mode
);
5808 current_vertex_color_clamp
= priv
->glsl_program
? priv
->glsl_program
->vs
.vertex_color_clamp
: GL_FIXED_ONLY_ARB
;
5809 if (old_vertex_color_clamp
!= current_vertex_color_clamp
)
5811 if (gl_info
->supported
[ARB_COLOR_BUFFER_FLOAT
])
5813 GL_EXTCALL(glClampColorARB(GL_CLAMP_VERTEX_COLOR_ARB
, current_vertex_color_clamp
));
5814 checkGLcall("glClampColorARB");
5818 FIXME("vertex color clamp needs to be changed, but extension not supported.\n");
5822 program_id
= priv
->glsl_program
? priv
->glsl_program
->programId
: 0;
5823 if (program_id
) TRACE("Using GLSL program %u\n", program_id
);
5824 GL_EXTCALL(glUseProgramObjectARB(program_id
));
5825 checkGLcall("glUseProgramObjectARB");
5827 /* In case that NP2 texcoord fixup data is found for the selected program, trigger a reload of the
5828 * constants. This has to be done because it can't be guaranteed that sampler() (from state.c) is
5829 * called between selecting the shader and using it, which results in wrong fixup for some frames. */
5830 if (priv
->glsl_program
&& priv
->glsl_program
->ps
.np2_fixup_info
)
5832 shader_glsl_load_np2fixup_constants(priv
, gl_info
, &device
->stateBlock
->state
);
5836 /* Context activation is done by the caller. */
5837 static void shader_glsl_select_depth_blt(void *shader_priv
, const struct wined3d_gl_info
*gl_info
,
5838 enum tex_types tex_type
, const SIZE
*ds_mask_size
)
5840 BOOL masked
= ds_mask_size
->cx
&& ds_mask_size
->cy
;
5841 struct shader_glsl_priv
*priv
= shader_priv
;
5842 GLhandleARB
*blt_program
;
5845 blt_program
= masked
? &priv
->depth_blt_program_masked
[tex_type
] : &priv
->depth_blt_program_full
[tex_type
];
5848 *blt_program
= create_glsl_blt_shader(gl_info
, tex_type
, masked
);
5849 loc
= GL_EXTCALL(glGetUniformLocationARB(*blt_program
, "sampler"));
5850 GL_EXTCALL(glUseProgramObjectARB(*blt_program
));
5851 GL_EXTCALL(glUniform1iARB(loc
, 0));
5855 GL_EXTCALL(glUseProgramObjectARB(*blt_program
));
5860 loc
= GL_EXTCALL(glGetUniformLocationARB(*blt_program
, "mask"));
5861 GL_EXTCALL(glUniform4fARB(loc
, 0.0f
, 0.0f
, (float)ds_mask_size
->cx
, (float)ds_mask_size
->cy
));
5865 /* Context activation is done by the caller. */
5866 static void shader_glsl_deselect_depth_blt(void *shader_priv
, const struct wined3d_gl_info
*gl_info
)
5868 struct shader_glsl_priv
*priv
= shader_priv
;
5869 GLhandleARB program_id
;
5871 program_id
= priv
->glsl_program
? priv
->glsl_program
->programId
: 0;
5872 if (program_id
) TRACE("Using GLSL program %u\n", program_id
);
5874 GL_EXTCALL(glUseProgramObjectARB(program_id
));
5875 checkGLcall("glUseProgramObjectARB");
5878 static void shader_glsl_destroy(struct wined3d_shader
*shader
)
5880 struct glsl_shader_private
*shader_data
= shader
->backend_data
;
5881 struct wined3d_device
*device
= shader
->device
;
5882 struct shader_glsl_priv
*priv
= device
->shader_priv
;
5883 const struct wined3d_gl_info
*gl_info
;
5884 const struct list
*linked_programs
;
5885 struct wined3d_context
*context
;
5887 if (!shader_data
|| !shader_data
->num_gl_shaders
)
5889 HeapFree(GetProcessHeap(), 0, shader_data
);
5890 shader
->backend_data
= NULL
;
5894 context
= context_acquire(device
, NULL
);
5895 gl_info
= context
->gl_info
;
5897 TRACE("Deleting linked programs.\n");
5898 linked_programs
= &shader
->linked_programs
;
5899 if (linked_programs
->next
)
5901 struct glsl_shader_prog_link
*entry
, *entry2
;
5904 switch (shader
->reg_maps
.shader_version
.type
)
5906 case WINED3D_SHADER_TYPE_PIXEL
:
5908 struct glsl_ps_compiled_shader
*gl_shaders
= shader_data
->gl_shaders
.ps
;
5910 LIST_FOR_EACH_ENTRY_SAFE(entry
, entry2
, linked_programs
,
5911 struct glsl_shader_prog_link
, ps
.shader_entry
)
5913 delete_glsl_program_entry(priv
, gl_info
, entry
);
5916 for (i
= 0; i
< shader_data
->num_gl_shaders
; ++i
)
5918 TRACE("Deleting pixel shader %u.\n", gl_shaders
[i
].prgId
);
5919 if (priv
->glsl_program
&& priv
->glsl_program
->ps
.id
== gl_shaders
[i
].prgId
)
5920 shader_glsl_select(context
, WINED3D_SHADER_MODE_NONE
, WINED3D_SHADER_MODE_NONE
);
5921 GL_EXTCALL(glDeleteObjectARB(gl_shaders
[i
].prgId
));
5922 checkGLcall("glDeleteObjectARB");
5924 HeapFree(GetProcessHeap(), 0, shader_data
->gl_shaders
.ps
);
5929 case WINED3D_SHADER_TYPE_VERTEX
:
5931 struct glsl_vs_compiled_shader
*gl_shaders
= shader_data
->gl_shaders
.vs
;
5933 LIST_FOR_EACH_ENTRY_SAFE(entry
, entry2
, linked_programs
,
5934 struct glsl_shader_prog_link
, vs
.shader_entry
)
5936 delete_glsl_program_entry(priv
, gl_info
, entry
);
5939 for (i
= 0; i
< shader_data
->num_gl_shaders
; ++i
)
5941 TRACE("Deleting vertex shader %u.\n", gl_shaders
[i
].prgId
);
5942 if (priv
->glsl_program
&& priv
->glsl_program
->vs
.id
== gl_shaders
[i
].prgId
)
5943 shader_glsl_select(context
, WINED3D_SHADER_MODE_NONE
, WINED3D_SHADER_MODE_NONE
);
5944 GL_EXTCALL(glDeleteObjectARB(gl_shaders
[i
].prgId
));
5945 checkGLcall("glDeleteObjectARB");
5947 HeapFree(GetProcessHeap(), 0, shader_data
->gl_shaders
.vs
);
5952 case WINED3D_SHADER_TYPE_GEOMETRY
:
5954 struct glsl_gs_compiled_shader
*gl_shaders
= shader_data
->gl_shaders
.gs
;
5956 LIST_FOR_EACH_ENTRY_SAFE(entry
, entry2
, linked_programs
,
5957 struct glsl_shader_prog_link
, gs
.shader_entry
)
5959 delete_glsl_program_entry(priv
, gl_info
, entry
);
5962 for (i
= 0; i
< shader_data
->num_gl_shaders
; ++i
)
5964 TRACE("Deleting geometry shader %u.\n", gl_shaders
[i
].id
);
5965 if (priv
->glsl_program
&& priv
->glsl_program
->gs
.id
== gl_shaders
[i
].id
)
5966 shader_glsl_select(context
, WINED3D_SHADER_MODE_NONE
, WINED3D_SHADER_MODE_NONE
);
5967 GL_EXTCALL(glDeleteObjectARB(gl_shaders
[i
].id
));
5968 checkGLcall("glDeleteObjectARB");
5970 HeapFree(GetProcessHeap(), 0, shader_data
->gl_shaders
.gs
);
5976 ERR("Unhandled shader type %#x.\n", shader
->reg_maps
.shader_version
.type
);
5981 HeapFree(GetProcessHeap(), 0, shader
->backend_data
);
5982 shader
->backend_data
= NULL
;
5984 context_release(context
);
5987 static int glsl_program_key_compare(const void *key
, const struct wine_rb_entry
*entry
)
5989 const struct glsl_program_key
*k
= key
;
5990 const struct glsl_shader_prog_link
*prog
= WINE_RB_ENTRY_VALUE(entry
,
5991 const struct glsl_shader_prog_link
, program_lookup_entry
);
5993 if (k
->vs_id
> prog
->vs
.id
) return 1;
5994 else if (k
->vs_id
< prog
->vs
.id
) return -1;
5996 if (k
->gs_id
> prog
->gs
.id
) return 1;
5997 else if (k
->gs_id
< prog
->gs
.id
) return -1;
5999 if (k
->ps_id
> prog
->ps
.id
) return 1;
6000 else if (k
->ps_id
< prog
->ps
.id
) return -1;
6005 static BOOL
constant_heap_init(struct constant_heap
*heap
, unsigned int constant_count
)
6007 SIZE_T size
= (constant_count
+ 1) * sizeof(*heap
->entries
) + constant_count
* sizeof(*heap
->positions
);
6008 void *mem
= HeapAlloc(GetProcessHeap(), 0, size
);
6012 ERR("Failed to allocate memory\n");
6016 heap
->entries
= mem
;
6017 heap
->entries
[1].version
= 0;
6018 heap
->positions
= (unsigned int *)(heap
->entries
+ constant_count
+ 1);
6024 static void constant_heap_free(struct constant_heap
*heap
)
6026 HeapFree(GetProcessHeap(), 0, heap
->entries
);
6029 static const struct wine_rb_functions wined3d_glsl_program_rb_functions
=
6034 glsl_program_key_compare
,
6037 static HRESULT
shader_glsl_alloc(struct wined3d_device
*device
, const struct fragment_pipeline
*fragment_pipe
)
6039 const struct wined3d_gl_info
*gl_info
= &device
->adapter
->gl_info
;
6040 struct shader_glsl_priv
*priv
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(struct shader_glsl_priv
));
6041 SIZE_T stack_size
= wined3d_log2i(max(gl_info
->limits
.glsl_vs_float_constants
,
6042 gl_info
->limits
.glsl_ps_float_constants
)) + 1;
6043 struct fragment_caps fragment_caps
;
6044 void *fragment_priv
;
6046 if (!(fragment_priv
= fragment_pipe
->alloc_private(&glsl_shader_backend
, priv
)))
6048 ERR("Failed to initialize fragment pipe.\n");
6049 HeapFree(GetProcessHeap(), 0, priv
);
6053 if (!shader_buffer_init(&priv
->shader_buffer
))
6055 ERR("Failed to initialize shader buffer.\n");
6059 priv
->stack
= HeapAlloc(GetProcessHeap(), 0, stack_size
* sizeof(*priv
->stack
));
6062 ERR("Failed to allocate memory.\n");
6066 if (!constant_heap_init(&priv
->vconst_heap
, gl_info
->limits
.glsl_vs_float_constants
))
6068 ERR("Failed to initialize vertex shader constant heap\n");
6072 if (!constant_heap_init(&priv
->pconst_heap
, gl_info
->limits
.glsl_ps_float_constants
))
6074 ERR("Failed to initialize pixel shader constant heap\n");
6078 if (wine_rb_init(&priv
->program_lookup
, &wined3d_glsl_program_rb_functions
) == -1)
6080 ERR("Failed to initialize rbtree.\n");
6084 priv
->next_constant_version
= 1;
6085 fragment_pipe
->get_caps(gl_info
, &fragment_caps
);
6086 priv
->ffp_proj_control
= fragment_caps
.wined3d_caps
& WINED3D_FRAGMENT_CAP_PROJ_CONTROL
;
6087 device
->fragment_priv
= fragment_priv
;
6088 priv
->fragment_pipe
= fragment_pipe
;
6090 device
->shader_priv
= priv
;
6094 constant_heap_free(&priv
->pconst_heap
);
6095 constant_heap_free(&priv
->vconst_heap
);
6096 HeapFree(GetProcessHeap(), 0, priv
->stack
);
6097 shader_buffer_free(&priv
->shader_buffer
);
6098 fragment_pipe
->free_private(device
);
6099 HeapFree(GetProcessHeap(), 0, priv
);
6100 return E_OUTOFMEMORY
;
6103 /* Context activation is done by the caller. */
6104 static void shader_glsl_free(struct wined3d_device
*device
)
6106 const struct wined3d_gl_info
*gl_info
= &device
->adapter
->gl_info
;
6107 struct shader_glsl_priv
*priv
= device
->shader_priv
;
6110 for (i
= 0; i
< tex_type_count
; ++i
)
6112 if (priv
->depth_blt_program_full
[i
])
6114 GL_EXTCALL(glDeleteObjectARB(priv
->depth_blt_program_full
[i
]));
6116 if (priv
->depth_blt_program_masked
[i
])
6118 GL_EXTCALL(glDeleteObjectARB(priv
->depth_blt_program_masked
[i
]));
6122 wine_rb_destroy(&priv
->program_lookup
, NULL
, NULL
);
6123 constant_heap_free(&priv
->pconst_heap
);
6124 constant_heap_free(&priv
->vconst_heap
);
6125 HeapFree(GetProcessHeap(), 0, priv
->stack
);
6126 shader_buffer_free(&priv
->shader_buffer
);
6127 priv
->fragment_pipe
->free_private(device
);
6129 HeapFree(GetProcessHeap(), 0, device
->shader_priv
);
6130 device
->shader_priv
= NULL
;
6133 static void shader_glsl_context_destroyed(void *shader_priv
, const struct wined3d_context
*context
) {}
6135 static void shader_glsl_get_caps(const struct wined3d_gl_info
*gl_info
, struct shader_caps
*caps
)
6139 if (gl_info
->supported
[EXT_GPU_SHADER4
] && gl_info
->supported
[ARB_SHADER_BIT_ENCODING
]
6140 && gl_info
->supported
[ARB_GEOMETRY_SHADER4
] && gl_info
->glsl_version
>= MAKEDWORD_VERSION(1, 50)
6141 && gl_info
->supported
[ARB_DRAW_ELEMENTS_BASE_VERTEX
] && gl_info
->supported
[ARB_DRAW_INSTANCED
])
6143 /* ARB_shader_texture_lod or EXT_gpu_shader4 is required for the SM3
6144 * texldd and texldl instructions. */
6145 else if (gl_info
->supported
[ARB_SHADER_TEXTURE_LOD
] || gl_info
->supported
[EXT_GPU_SHADER4
])
6149 TRACE("Shader model %u.\n", shader_model
);
6151 caps
->vs_version
= min(wined3d_settings
.max_sm_vs
, shader_model
);
6152 caps
->gs_version
= min(wined3d_settings
.max_sm_gs
, shader_model
);
6153 caps
->ps_version
= min(wined3d_settings
.max_sm_ps
, shader_model
);
6155 caps
->vs_uniform_count
= gl_info
->limits
.glsl_vs_float_constants
;
6156 caps
->ps_uniform_count
= gl_info
->limits
.glsl_ps_float_constants
;
6158 /* FIXME: The following line is card dependent. -8.0 to 8.0 is the
6159 * Direct3D minimum requirement.
6161 * Both GL_ARB_fragment_program and GLSL require a "maximum representable magnitude"
6162 * of colors to be 2^10, and 2^32 for other floats. Should we use 1024 here?
6164 * The problem is that the refrast clamps temporary results in the shader to
6165 * [-MaxValue;+MaxValue]. If the card's max value is bigger than the one we advertize here,
6166 * then applications may miss the clamping behavior. On the other hand, if it is smaller,
6167 * the shader will generate incorrect results too. Unfortunately, GL deliberately doesn't
6168 * offer a way to query this.
6170 caps
->ps_1x_max_value
= 8.0;
6172 /* Ideally we'd only set caps like sRGB writes here if supported by both
6173 * the shader backend and the fragment pipe, but we can get called before
6174 * shader_glsl_alloc(). */
6175 caps
->wined3d_caps
= WINED3D_SHADER_CAP_VS_CLIPPING
6176 | WINED3D_SHADER_CAP_SRGB_WRITE
;
6179 static BOOL
shader_glsl_color_fixup_supported(struct color_fixup_desc fixup
)
6181 if (TRACE_ON(d3d_shader
) && TRACE_ON(d3d
))
6183 TRACE("Checking support for fixup:\n");
6184 dump_color_fixup_desc(fixup
);
6187 /* We support everything except YUV conversions. */
6188 if (!is_complex_fixup(fixup
))
6194 TRACE("[FAILED]\n");
6198 static const SHADER_HANDLER shader_glsl_instruction_handler_table
[WINED3DSIH_TABLE_SIZE
] =
6200 /* WINED3DSIH_ABS */ shader_glsl_map2gl
,
6201 /* WINED3DSIH_ADD */ shader_glsl_binop
,
6202 /* WINED3DSIH_AND */ shader_glsl_binop
,
6203 /* WINED3DSIH_BEM */ shader_glsl_bem
,
6204 /* WINED3DSIH_BREAK */ shader_glsl_break
,
6205 /* WINED3DSIH_BREAKC */ shader_glsl_breakc
,
6206 /* WINED3DSIH_BREAKP */ shader_glsl_breakp
,
6207 /* WINED3DSIH_CALL */ shader_glsl_call
,
6208 /* WINED3DSIH_CALLNZ */ shader_glsl_callnz
,
6209 /* WINED3DSIH_CMP */ shader_glsl_conditional_move
,
6210 /* WINED3DSIH_CND */ shader_glsl_cnd
,
6211 /* WINED3DSIH_CRS */ shader_glsl_cross
,
6212 /* WINED3DSIH_CUT */ shader_glsl_cut
,
6213 /* WINED3DSIH_DCL */ shader_glsl_nop
,
6214 /* WINED3DSIH_DCL_CONSTANT_BUFFER */ shader_glsl_nop
,
6215 /* WINED3DSIH_DCL_INPUT_PRIMITIVE */ shader_glsl_nop
,
6216 /* WINED3DSIH_DCL_OUTPUT_TOPOLOGY */ shader_glsl_nop
,
6217 /* WINED3DSIH_DCL_VERTICES_OUT */ shader_glsl_nop
,
6218 /* WINED3DSIH_DEF */ shader_glsl_nop
,
6219 /* WINED3DSIH_DEFB */ shader_glsl_nop
,
6220 /* WINED3DSIH_DEFI */ shader_glsl_nop
,
6221 /* WINED3DSIH_DIV */ shader_glsl_binop
,
6222 /* WINED3DSIH_DP2ADD */ shader_glsl_dp2add
,
6223 /* WINED3DSIH_DP3 */ shader_glsl_dot
,
6224 /* WINED3DSIH_DP4 */ shader_glsl_dot
,
6225 /* WINED3DSIH_DST */ shader_glsl_dst
,
6226 /* WINED3DSIH_DSX */ shader_glsl_map2gl
,
6227 /* WINED3DSIH_DSY */ shader_glsl_map2gl
,
6228 /* WINED3DSIH_ELSE */ shader_glsl_else
,
6229 /* WINED3DSIH_EMIT */ shader_glsl_emit
,
6230 /* WINED3DSIH_ENDIF */ shader_glsl_end
,
6231 /* WINED3DSIH_ENDLOOP */ shader_glsl_end
,
6232 /* WINED3DSIH_ENDREP */ shader_glsl_end
,
6233 /* WINED3DSIH_EQ */ shader_glsl_relop
,
6234 /* WINED3DSIH_EXP */ shader_glsl_map2gl
,
6235 /* WINED3DSIH_EXPP */ shader_glsl_expp
,
6236 /* WINED3DSIH_FRC */ shader_glsl_map2gl
,
6237 /* WINED3DSIH_FTOI */ shader_glsl_to_int
,
6238 /* WINED3DSIH_GE */ shader_glsl_relop
,
6239 /* WINED3DSIH_IADD */ shader_glsl_binop
,
6240 /* WINED3DSIH_IEQ */ NULL
,
6241 /* WINED3DSIH_IF */ shader_glsl_if
,
6242 /* WINED3DSIH_IFC */ shader_glsl_ifc
,
6243 /* WINED3DSIH_IGE */ shader_glsl_relop
,
6244 /* WINED3DSIH_IMUL */ shader_glsl_imul
,
6245 /* WINED3DSIH_ITOF */ shader_glsl_to_float
,
6246 /* WINED3DSIH_LABEL */ shader_glsl_label
,
6247 /* WINED3DSIH_LD */ NULL
,
6248 /* WINED3DSIH_LIT */ shader_glsl_lit
,
6249 /* WINED3DSIH_LOG */ shader_glsl_log
,
6250 /* WINED3DSIH_LOGP */ shader_glsl_log
,
6251 /* WINED3DSIH_LOOP */ shader_glsl_loop
,
6252 /* WINED3DSIH_LRP */ shader_glsl_lrp
,
6253 /* WINED3DSIH_LT */ shader_glsl_relop
,
6254 /* WINED3DSIH_M3x2 */ shader_glsl_mnxn
,
6255 /* WINED3DSIH_M3x3 */ shader_glsl_mnxn
,
6256 /* WINED3DSIH_M3x4 */ shader_glsl_mnxn
,
6257 /* WINED3DSIH_M4x3 */ shader_glsl_mnxn
,
6258 /* WINED3DSIH_M4x4 */ shader_glsl_mnxn
,
6259 /* WINED3DSIH_MAD */ shader_glsl_mad
,
6260 /* WINED3DSIH_MAX */ shader_glsl_map2gl
,
6261 /* WINED3DSIH_MIN */ shader_glsl_map2gl
,
6262 /* WINED3DSIH_MOV */ shader_glsl_mov
,
6263 /* WINED3DSIH_MOVA */ shader_glsl_mov
,
6264 /* WINED3DSIH_MOVC */ shader_glsl_conditional_move
,
6265 /* WINED3DSIH_MUL */ shader_glsl_binop
,
6266 /* WINED3DSIH_NOP */ shader_glsl_nop
,
6267 /* WINED3DSIH_NRM */ shader_glsl_nrm
,
6268 /* WINED3DSIH_PHASE */ shader_glsl_nop
,
6269 /* WINED3DSIH_POW */ shader_glsl_pow
,
6270 /* WINED3DSIH_RCP */ shader_glsl_rcp
,
6271 /* WINED3DSIH_REP */ shader_glsl_rep
,
6272 /* WINED3DSIH_RET */ shader_glsl_ret
,
6273 /* WINED3DSIH_ROUND_NI */ shader_glsl_map2gl
,
6274 /* WINED3DSIH_RSQ */ shader_glsl_rsq
,
6275 /* WINED3DSIH_SAMPLE */ NULL
,
6276 /* WINED3DSIH_SAMPLE_GRAD */ NULL
,
6277 /* WINED3DSIH_SAMPLE_LOD */ NULL
,
6278 /* WINED3DSIH_SETP */ NULL
,
6279 /* WINED3DSIH_SGE */ shader_glsl_compare
,
6280 /* WINED3DSIH_SGN */ shader_glsl_sgn
,
6281 /* WINED3DSIH_SINCOS */ shader_glsl_sincos
,
6282 /* WINED3DSIH_SLT */ shader_glsl_compare
,
6283 /* WINED3DSIH_SQRT */ NULL
,
6284 /* WINED3DSIH_SUB */ shader_glsl_binop
,
6285 /* WINED3DSIH_TEX */ shader_glsl_tex
,
6286 /* WINED3DSIH_TEXBEM */ shader_glsl_texbem
,
6287 /* WINED3DSIH_TEXBEML */ shader_glsl_texbem
,
6288 /* WINED3DSIH_TEXCOORD */ shader_glsl_texcoord
,
6289 /* WINED3DSIH_TEXDEPTH */ shader_glsl_texdepth
,
6290 /* WINED3DSIH_TEXDP3 */ shader_glsl_texdp3
,
6291 /* WINED3DSIH_TEXDP3TEX */ shader_glsl_texdp3tex
,
6292 /* WINED3DSIH_TEXKILL */ shader_glsl_texkill
,
6293 /* WINED3DSIH_TEXLDD */ shader_glsl_texldd
,
6294 /* WINED3DSIH_TEXLDL */ shader_glsl_texldl
,
6295 /* WINED3DSIH_TEXM3x2DEPTH */ shader_glsl_texm3x2depth
,
6296 /* WINED3DSIH_TEXM3x2PAD */ shader_glsl_texm3x2pad
,
6297 /* WINED3DSIH_TEXM3x2TEX */ shader_glsl_texm3x2tex
,
6298 /* WINED3DSIH_TEXM3x3 */ shader_glsl_texm3x3
,
6299 /* WINED3DSIH_TEXM3x3DIFF */ NULL
,
6300 /* WINED3DSIH_TEXM3x3PAD */ shader_glsl_texm3x3pad
,
6301 /* WINED3DSIH_TEXM3x3SPEC */ shader_glsl_texm3x3spec
,
6302 /* WINED3DSIH_TEXM3x3TEX */ shader_glsl_texm3x3tex
,
6303 /* WINED3DSIH_TEXM3x3VSPEC */ shader_glsl_texm3x3vspec
,
6304 /* WINED3DSIH_TEXREG2AR */ shader_glsl_texreg2ar
,
6305 /* WINED3DSIH_TEXREG2GB */ shader_glsl_texreg2gb
,
6306 /* WINED3DSIH_TEXREG2RGB */ shader_glsl_texreg2rgb
,
6307 /* WINED3DSIH_UDIV */ shader_glsl_udiv
,
6308 /* WINED3DSIH_USHR */ shader_glsl_binop
,
6309 /* WINED3DSIH_UTOF */ shader_glsl_to_float
,
6310 /* WINED3DSIH_XOR */ shader_glsl_binop
,
6313 static void shader_glsl_handle_instruction(const struct wined3d_shader_instruction
*ins
) {
6314 SHADER_HANDLER hw_fct
;
6316 /* Select handler */
6317 hw_fct
= shader_glsl_instruction_handler_table
[ins
->handler_idx
];
6319 /* Unhandled opcode */
6322 FIXME("Backend can't handle opcode %#x\n", ins
->handler_idx
);
6327 shader_glsl_add_instruction_modifiers(ins
);
6330 static BOOL
shader_glsl_has_ffp_proj_control(void *shader_priv
)
6332 struct shader_glsl_priv
*priv
= shader_priv
;
6334 return priv
->ffp_proj_control
;
6337 const struct wined3d_shader_backend_ops glsl_shader_backend
=
6339 shader_glsl_handle_instruction
,
6341 shader_glsl_select_depth_blt
,
6342 shader_glsl_deselect_depth_blt
,
6343 shader_glsl_update_float_vertex_constants
,
6344 shader_glsl_update_float_pixel_constants
,
6345 shader_glsl_load_constants
,
6346 shader_glsl_load_np2fixup_constants
,
6347 shader_glsl_destroy
,
6350 shader_glsl_context_destroyed
,
6351 shader_glsl_get_caps
,
6352 shader_glsl_color_fixup_supported
,
6353 shader_glsl_has_ffp_proj_control
,
6356 static void glsl_fragment_pipe_enable(const struct wined3d_gl_info
*gl_info
, BOOL enable
)
6358 /* Nothing to do. */
6361 static void glsl_fragment_pipe_get_caps(const struct wined3d_gl_info
*gl_info
, struct fragment_caps
*caps
)
6363 caps
->wined3d_caps
= WINED3D_FRAGMENT_CAP_PROJ_CONTROL
6364 | WINED3D_FRAGMENT_CAP_SRGB_WRITE
;
6365 caps
->PrimitiveMiscCaps
= WINED3DPMISCCAPS_TSSARGTEMP
;
6366 caps
->TextureOpCaps
= WINED3DTEXOPCAPS_DISABLE
6367 | WINED3DTEXOPCAPS_SELECTARG1
6368 | WINED3DTEXOPCAPS_SELECTARG2
6369 | WINED3DTEXOPCAPS_MODULATE4X
6370 | WINED3DTEXOPCAPS_MODULATE2X
6371 | WINED3DTEXOPCAPS_MODULATE
6372 | WINED3DTEXOPCAPS_ADDSIGNED2X
6373 | WINED3DTEXOPCAPS_ADDSIGNED
6374 | WINED3DTEXOPCAPS_ADD
6375 | WINED3DTEXOPCAPS_SUBTRACT
6376 | WINED3DTEXOPCAPS_ADDSMOOTH
6377 | WINED3DTEXOPCAPS_BLENDCURRENTALPHA
6378 | WINED3DTEXOPCAPS_BLENDFACTORALPHA
6379 | WINED3DTEXOPCAPS_BLENDTEXTUREALPHA
6380 | WINED3DTEXOPCAPS_BLENDDIFFUSEALPHA
6381 | WINED3DTEXOPCAPS_BLENDTEXTUREALPHAPM
6382 | WINED3DTEXOPCAPS_MODULATEALPHA_ADDCOLOR
6383 | WINED3DTEXOPCAPS_MODULATECOLOR_ADDALPHA
6384 | WINED3DTEXOPCAPS_MODULATEINVCOLOR_ADDALPHA
6385 | WINED3DTEXOPCAPS_MODULATEINVALPHA_ADDCOLOR
6386 | WINED3DTEXOPCAPS_DOTPRODUCT3
6387 | WINED3DTEXOPCAPS_MULTIPLYADD
6388 | WINED3DTEXOPCAPS_LERP
6389 | WINED3DTEXOPCAPS_BUMPENVMAP
6390 | WINED3DTEXOPCAPS_BUMPENVMAPLUMINANCE
;
6391 caps
->MaxTextureBlendStages
= 8;
6392 caps
->MaxSimultaneousTextures
= min(gl_info
->limits
.fragment_samplers
, 8);
6395 static void *glsl_fragment_pipe_alloc(const struct wined3d_shader_backend_ops
*shader_backend
, void *shader_priv
)
6397 struct shader_glsl_priv
*priv
;
6399 if (shader_backend
== &glsl_shader_backend
)
6403 if (wine_rb_init(&priv
->ffp_fragment_shaders
, &wined3d_ffp_frag_program_rb_functions
) == -1)
6405 ERR("Failed to initialize rbtree.\n");
6412 FIXME("GLSL fragment pipe without GLSL shader backend not implemented.\n");
6417 struct glsl_ffp_destroy_ctx
6419 struct shader_glsl_priv
*priv
;
6420 const struct wined3d_gl_info
*gl_info
;
6423 static void shader_glsl_free_ffp_fragment_shader(struct wine_rb_entry
*entry
, void *context
)
6425 struct glsl_ffp_fragment_shader
*shader
= WINE_RB_ENTRY_VALUE(entry
,
6426 struct glsl_ffp_fragment_shader
, entry
.entry
);
6427 struct glsl_shader_prog_link
*program
, *program2
;
6428 struct glsl_ffp_destroy_ctx
*ctx
= context
;
6430 LIST_FOR_EACH_ENTRY_SAFE(program
, program2
, &shader
->linked_programs
,
6431 struct glsl_shader_prog_link
, ps
.shader_entry
)
6433 delete_glsl_program_entry(ctx
->priv
, ctx
->gl_info
, program
);
6435 ctx
->gl_info
->gl_ops
.ext
.p_glDeleteObjectARB(shader
->id
);
6436 HeapFree(GetProcessHeap(), 0, shader
);
6439 /* Context activation is done by the caller. */
6440 static void glsl_fragment_pipe_free(struct wined3d_device
*device
)
6442 struct shader_glsl_priv
*priv
= device
->fragment_priv
;
6443 struct glsl_ffp_destroy_ctx ctx
;
6446 ctx
.gl_info
= &device
->adapter
->gl_info
;
6447 wine_rb_destroy(&priv
->ffp_fragment_shaders
, shader_glsl_free_ffp_fragment_shader
, &ctx
);
6450 static void glsl_fragment_pipe_shader(struct wined3d_context
*context
,
6451 const struct wined3d_state
*state
, DWORD state_id
)
6453 context
->last_was_pshader
= use_ps(state
);
6455 context
->select_shader
= 1;
6456 context
->load_constants
= 1;
6459 static void glsl_fragment_pipe_fog(struct wined3d_context
*context
,
6460 const struct wined3d_state
*state
, DWORD state_id
)
6462 BOOL use_vshader
= use_vs(state
);
6463 enum fogsource new_source
;
6465 context
->select_shader
= 1;
6466 context
->load_constants
= 1;
6468 if (!state
->render_states
[WINED3D_RS_FOGENABLE
])
6471 if (state
->render_states
[WINED3D_RS_FOGTABLEMODE
] == WINED3D_FOG_NONE
)
6474 new_source
= FOGSOURCE_VS
;
6475 else if (state
->render_states
[WINED3D_RS_FOGVERTEXMODE
] == WINED3D_FOG_NONE
|| context
->last_was_rhw
)
6476 new_source
= FOGSOURCE_COORD
;
6478 new_source
= FOGSOURCE_FFP
;
6482 new_source
= FOGSOURCE_FFP
;
6485 if (new_source
!= context
->fog_source
)
6487 context
->fog_source
= new_source
;
6488 state_fogstartend(context
, state
, STATE_RENDER(WINED3D_RS_FOGSTART
));
6492 static void glsl_fragment_pipe_tex_transform(struct wined3d_context
*context
,
6493 const struct wined3d_state
*state
, DWORD state_id
)
6495 context
->select_shader
= 1;
6496 context
->load_constants
= 1;
6499 static void glsl_fragment_pipe_invalidate_constants(struct wined3d_context
*context
,
6500 const struct wined3d_state
*state
, DWORD state_id
)
6502 context
->load_constants
= 1;
6505 static const struct StateEntryTemplate glsl_fragment_pipe_state_template
[] =
6507 {STATE_RENDER(WINED3D_RS_TEXTUREFACTOR
), {STATE_RENDER(WINED3D_RS_TEXTUREFACTOR
), glsl_fragment_pipe_invalidate_constants
}, WINED3D_GL_EXT_NONE
},
6508 {STATE_TEXTURESTAGE(0, WINED3D_TSS_COLOR_OP
), {STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6509 {STATE_TEXTURESTAGE(0, WINED3D_TSS_COLOR_ARG1
), {STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6510 {STATE_TEXTURESTAGE(0, WINED3D_TSS_COLOR_ARG2
), {STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6511 {STATE_TEXTURESTAGE(0, WINED3D_TSS_COLOR_ARG0
), {STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6512 {STATE_TEXTURESTAGE(0, WINED3D_TSS_ALPHA_OP
), {STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6513 {STATE_TEXTURESTAGE(0, WINED3D_TSS_ALPHA_ARG1
), {STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6514 {STATE_TEXTURESTAGE(0, WINED3D_TSS_ALPHA_ARG2
), {STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6515 {STATE_TEXTURESTAGE(0, WINED3D_TSS_ALPHA_ARG0
), {STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6516 {STATE_TEXTURESTAGE(0, WINED3D_TSS_RESULT_ARG
), {STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6517 {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT00
), {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT00
), glsl_fragment_pipe_invalidate_constants
}, WINED3D_GL_EXT_NONE
},
6518 {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT01
), {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6519 {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT10
), {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6520 {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT11
), {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6521 {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_LSCALE
), {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_LSCALE
), glsl_fragment_pipe_invalidate_constants
}, WINED3D_GL_EXT_NONE
},
6522 {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_LOFFSET
), {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6523 {STATE_TEXTURESTAGE(1, WINED3D_TSS_COLOR_OP
), {STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6524 {STATE_TEXTURESTAGE(1, WINED3D_TSS_COLOR_ARG1
), {STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6525 {STATE_TEXTURESTAGE(1, WINED3D_TSS_COLOR_ARG2
), {STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6526 {STATE_TEXTURESTAGE(1, WINED3D_TSS_COLOR_ARG0
), {STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6527 {STATE_TEXTURESTAGE(1, WINED3D_TSS_ALPHA_OP
), {STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6528 {STATE_TEXTURESTAGE(1, WINED3D_TSS_ALPHA_ARG1
), {STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6529 {STATE_TEXTURESTAGE(1, WINED3D_TSS_ALPHA_ARG2
), {STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6530 {STATE_TEXTURESTAGE(1, WINED3D_TSS_ALPHA_ARG0
), {STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6531 {STATE_TEXTURESTAGE(1, WINED3D_TSS_RESULT_ARG
), {STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6532 {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT00
), {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT00
), glsl_fragment_pipe_invalidate_constants
}, WINED3D_GL_EXT_NONE
},
6533 {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT01
), {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6534 {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT10
), {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6535 {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT11
), {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6536 {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_LSCALE
), {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_LSCALE
), glsl_fragment_pipe_invalidate_constants
}, WINED3D_GL_EXT_NONE
},
6537 {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_LOFFSET
), {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6538 {STATE_TEXTURESTAGE(2, WINED3D_TSS_COLOR_OP
), {STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6539 {STATE_TEXTURESTAGE(2, WINED3D_TSS_COLOR_ARG1
), {STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6540 {STATE_TEXTURESTAGE(2, WINED3D_TSS_COLOR_ARG2
), {STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6541 {STATE_TEXTURESTAGE(2, WINED3D_TSS_COLOR_ARG0
), {STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6542 {STATE_TEXTURESTAGE(2, WINED3D_TSS_ALPHA_OP
), {STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6543 {STATE_TEXTURESTAGE(2, WINED3D_TSS_ALPHA_ARG1
), {STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6544 {STATE_TEXTURESTAGE(2, WINED3D_TSS_ALPHA_ARG2
), {STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6545 {STATE_TEXTURESTAGE(2, WINED3D_TSS_ALPHA_ARG0
), {STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6546 {STATE_TEXTURESTAGE(2, WINED3D_TSS_RESULT_ARG
), {STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6547 {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT00
), {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT00
), glsl_fragment_pipe_invalidate_constants
}, WINED3D_GL_EXT_NONE
},
6548 {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT01
), {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6549 {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT10
), {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6550 {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT11
), {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6551 {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_LSCALE
), {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_LSCALE
), glsl_fragment_pipe_invalidate_constants
}, WINED3D_GL_EXT_NONE
},
6552 {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_LOFFSET
), {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6553 {STATE_TEXTURESTAGE(3, WINED3D_TSS_COLOR_OP
), {STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6554 {STATE_TEXTURESTAGE(3, WINED3D_TSS_COLOR_ARG1
), {STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6555 {STATE_TEXTURESTAGE(3, WINED3D_TSS_COLOR_ARG2
), {STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6556 {STATE_TEXTURESTAGE(3, WINED3D_TSS_COLOR_ARG0
), {STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6557 {STATE_TEXTURESTAGE(3, WINED3D_TSS_ALPHA_OP
), {STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6558 {STATE_TEXTURESTAGE(3, WINED3D_TSS_ALPHA_ARG1
), {STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6559 {STATE_TEXTURESTAGE(3, WINED3D_TSS_ALPHA_ARG2
), {STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6560 {STATE_TEXTURESTAGE(3, WINED3D_TSS_ALPHA_ARG0
), {STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6561 {STATE_TEXTURESTAGE(3, WINED3D_TSS_RESULT_ARG
), {STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6562 {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT00
), {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT00
), glsl_fragment_pipe_invalidate_constants
}, WINED3D_GL_EXT_NONE
},
6563 {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT01
), {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6564 {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT10
), {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6565 {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT11
), {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6566 {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_LSCALE
), {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_LSCALE
), glsl_fragment_pipe_invalidate_constants
}, WINED3D_GL_EXT_NONE
},
6567 {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_LOFFSET
), {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6568 {STATE_TEXTURESTAGE(4, WINED3D_TSS_COLOR_OP
), {STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6569 {STATE_TEXTURESTAGE(4, WINED3D_TSS_COLOR_ARG1
), {STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6570 {STATE_TEXTURESTAGE(4, WINED3D_TSS_COLOR_ARG2
), {STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6571 {STATE_TEXTURESTAGE(4, WINED3D_TSS_COLOR_ARG0
), {STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6572 {STATE_TEXTURESTAGE(4, WINED3D_TSS_ALPHA_OP
), {STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6573 {STATE_TEXTURESTAGE(4, WINED3D_TSS_ALPHA_ARG1
), {STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6574 {STATE_TEXTURESTAGE(4, WINED3D_TSS_ALPHA_ARG2
), {STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6575 {STATE_TEXTURESTAGE(4, WINED3D_TSS_ALPHA_ARG0
), {STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6576 {STATE_TEXTURESTAGE(4, WINED3D_TSS_RESULT_ARG
), {STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6577 {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT00
), {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT00
), glsl_fragment_pipe_invalidate_constants
}, WINED3D_GL_EXT_NONE
},
6578 {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT01
), {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6579 {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT10
), {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6580 {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT11
), {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6581 {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_LSCALE
), {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_LSCALE
), glsl_fragment_pipe_invalidate_constants
}, WINED3D_GL_EXT_NONE
},
6582 {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_LOFFSET
), {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6583 {STATE_TEXTURESTAGE(5, WINED3D_TSS_COLOR_OP
), {STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6584 {STATE_TEXTURESTAGE(5, WINED3D_TSS_COLOR_ARG1
), {STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6585 {STATE_TEXTURESTAGE(5, WINED3D_TSS_COLOR_ARG2
), {STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6586 {STATE_TEXTURESTAGE(5, WINED3D_TSS_COLOR_ARG0
), {STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6587 {STATE_TEXTURESTAGE(5, WINED3D_TSS_ALPHA_OP
), {STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6588 {STATE_TEXTURESTAGE(5, WINED3D_TSS_ALPHA_ARG1
), {STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6589 {STATE_TEXTURESTAGE(5, WINED3D_TSS_ALPHA_ARG2
), {STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6590 {STATE_TEXTURESTAGE(5, WINED3D_TSS_ALPHA_ARG0
), {STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6591 {STATE_TEXTURESTAGE(5, WINED3D_TSS_RESULT_ARG
), {STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6592 {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT00
), {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT00
), glsl_fragment_pipe_invalidate_constants
}, WINED3D_GL_EXT_NONE
},
6593 {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT01
), {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6594 {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT10
), {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6595 {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT11
), {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6596 {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_LSCALE
), {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_LSCALE
), glsl_fragment_pipe_invalidate_constants
}, WINED3D_GL_EXT_NONE
},
6597 {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_LOFFSET
), {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6598 {STATE_TEXTURESTAGE(6, WINED3D_TSS_COLOR_OP
), {STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6599 {STATE_TEXTURESTAGE(6, WINED3D_TSS_COLOR_ARG1
), {STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6600 {STATE_TEXTURESTAGE(6, WINED3D_TSS_COLOR_ARG2
), {STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6601 {STATE_TEXTURESTAGE(6, WINED3D_TSS_COLOR_ARG0
), {STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6602 {STATE_TEXTURESTAGE(6, WINED3D_TSS_ALPHA_OP
), {STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6603 {STATE_TEXTURESTAGE(6, WINED3D_TSS_ALPHA_ARG1
), {STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6604 {STATE_TEXTURESTAGE(6, WINED3D_TSS_ALPHA_ARG2
), {STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6605 {STATE_TEXTURESTAGE(6, WINED3D_TSS_ALPHA_ARG0
), {STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6606 {STATE_TEXTURESTAGE(6, WINED3D_TSS_RESULT_ARG
), {STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6607 {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT00
), {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT00
), glsl_fragment_pipe_invalidate_constants
}, WINED3D_GL_EXT_NONE
},
6608 {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT01
), {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6609 {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT10
), {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6610 {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT11
), {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6611 {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_LSCALE
), {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_LSCALE
), glsl_fragment_pipe_invalidate_constants
}, WINED3D_GL_EXT_NONE
},
6612 {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_LOFFSET
), {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6613 {STATE_TEXTURESTAGE(7, WINED3D_TSS_COLOR_OP
), {STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6614 {STATE_TEXTURESTAGE(7, WINED3D_TSS_COLOR_ARG1
), {STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6615 {STATE_TEXTURESTAGE(7, WINED3D_TSS_COLOR_ARG2
), {STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6616 {STATE_TEXTURESTAGE(7, WINED3D_TSS_COLOR_ARG0
), {STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6617 {STATE_TEXTURESTAGE(7, WINED3D_TSS_ALPHA_OP
), {STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6618 {STATE_TEXTURESTAGE(7, WINED3D_TSS_ALPHA_ARG1
), {STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6619 {STATE_TEXTURESTAGE(7, WINED3D_TSS_ALPHA_ARG2
), {STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6620 {STATE_TEXTURESTAGE(7, WINED3D_TSS_ALPHA_ARG0
), {STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6621 {STATE_TEXTURESTAGE(7, WINED3D_TSS_RESULT_ARG
), {STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6622 {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT00
), {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT00
), glsl_fragment_pipe_invalidate_constants
}, WINED3D_GL_EXT_NONE
},
6623 {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT01
), {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6624 {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT10
), {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6625 {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT11
), {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6626 {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_LSCALE
), {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_LSCALE
), glsl_fragment_pipe_invalidate_constants
}, WINED3D_GL_EXT_NONE
},
6627 {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_LOFFSET
), {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6628 {STATE_PIXELSHADER
, {STATE_PIXELSHADER
, glsl_fragment_pipe_shader
}, WINED3D_GL_EXT_NONE
},
6629 {STATE_RENDER(WINED3D_RS_FOGENABLE
), {STATE_RENDER(WINED3D_RS_FOGENABLE
), glsl_fragment_pipe_fog
}, WINED3D_GL_EXT_NONE
},
6630 {STATE_RENDER(WINED3D_RS_FOGTABLEMODE
), {STATE_RENDER(WINED3D_RS_FOGENABLE
), NULL
}, WINED3D_GL_EXT_NONE
},
6631 {STATE_RENDER(WINED3D_RS_FOGVERTEXMODE
), {STATE_RENDER(WINED3D_RS_FOGENABLE
), NULL
}, WINED3D_GL_EXT_NONE
},
6632 {STATE_RENDER(WINED3D_RS_FOGSTART
), {STATE_RENDER(WINED3D_RS_FOGSTART
), state_fogstartend
}, WINED3D_GL_EXT_NONE
},
6633 {STATE_RENDER(WINED3D_RS_FOGEND
), {STATE_RENDER(WINED3D_RS_FOGSTART
), NULL
}, WINED3D_GL_EXT_NONE
},
6634 {STATE_RENDER(WINED3D_RS_SRGBWRITEENABLE
), {STATE_RENDER(WINED3D_RS_SRGBWRITEENABLE
), state_srgbwrite
}, ARB_FRAMEBUFFER_SRGB
},
6635 {STATE_RENDER(WINED3D_RS_SRGBWRITEENABLE
), {STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6636 {STATE_RENDER(WINED3D_RS_FOGCOLOR
), {STATE_RENDER(WINED3D_RS_FOGCOLOR
), state_fogcolor
}, WINED3D_GL_EXT_NONE
},
6637 {STATE_RENDER(WINED3D_RS_FOGDENSITY
), {STATE_RENDER(WINED3D_RS_FOGDENSITY
), state_fogdensity
}, WINED3D_GL_EXT_NONE
},
6638 {STATE_TEXTURESTAGE(0,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(0, WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), glsl_fragment_pipe_tex_transform
}, WINED3D_GL_EXT_NONE
},
6639 {STATE_TEXTURESTAGE(1,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(1, WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), glsl_fragment_pipe_tex_transform
}, WINED3D_GL_EXT_NONE
},
6640 {STATE_TEXTURESTAGE(2,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(2, WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), glsl_fragment_pipe_tex_transform
}, WINED3D_GL_EXT_NONE
},
6641 {STATE_TEXTURESTAGE(3,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(3, WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), glsl_fragment_pipe_tex_transform
}, WINED3D_GL_EXT_NONE
},
6642 {STATE_TEXTURESTAGE(4,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(4, WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), glsl_fragment_pipe_tex_transform
}, WINED3D_GL_EXT_NONE
},
6643 {STATE_TEXTURESTAGE(5,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(5, WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), glsl_fragment_pipe_tex_transform
}, WINED3D_GL_EXT_NONE
},
6644 {STATE_TEXTURESTAGE(6,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(6, WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), glsl_fragment_pipe_tex_transform
}, WINED3D_GL_EXT_NONE
},
6645 {STATE_TEXTURESTAGE(7,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(7, WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), glsl_fragment_pipe_tex_transform
}, WINED3D_GL_EXT_NONE
},
6646 {STATE_RENDER(WINED3D_RS_SPECULARENABLE
), {STATE_RENDER(WINED3D_RS_SPECULARENABLE
), glsl_fragment_pipe_invalidate_constants
}, WINED3D_GL_EXT_NONE
},
6647 {0 /* Terminate */, {0, 0 }, WINED3D_GL_EXT_NONE
},
6650 const struct fragment_pipeline glsl_fragment_pipe
=
6652 glsl_fragment_pipe_enable
,
6653 glsl_fragment_pipe_get_caps
,
6654 glsl_fragment_pipe_alloc
,
6655 glsl_fragment_pipe_free
,
6656 shader_glsl_color_fixup_supported
,
6657 glsl_fragment_pipe_state_template
,