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
;
103 /* Struct to maintain data about a linked GLSL program */
104 struct glsl_shader_prog_link
{
105 struct wine_rb_entry program_lookup_entry
;
106 struct list vshader_entry
;
107 struct list pshader_entry
;
108 GLhandleARB programId
;
109 GLint
*vuniformF_locations
;
110 GLint
*puniformF_locations
;
111 GLint vuniformI_locations
[MAX_CONST_I
];
112 GLint puniformI_locations
[MAX_CONST_I
];
113 GLint posFixup_location
;
114 GLint np2Fixup_location
;
115 GLint bumpenvmat_location
[MAX_TEXTURES
];
116 GLint luminancescale_location
[MAX_TEXTURES
];
117 GLint luminanceoffset_location
[MAX_TEXTURES
];
118 GLint ycorrection_location
;
119 GLenum vertex_color_clamp
;
122 UINT constant_version
;
123 const struct ps_np2fixup_info
*np2Fixup_info
;
126 struct glsl_program_key
132 struct shader_glsl_ctx_priv
{
133 const struct vs_compile_args
*cur_vs_args
;
134 const struct ps_compile_args
*cur_ps_args
;
135 struct ps_np2fixup_info
*cur_np2fixup_info
;
138 struct glsl_ps_compiled_shader
140 struct ps_compile_args args
;
141 struct ps_np2fixup_info np2fixup
;
145 struct glsl_vs_compiled_shader
147 struct vs_compile_args args
;
151 struct glsl_shader_private
155 struct glsl_vs_compiled_shader
*vs
;
156 struct glsl_ps_compiled_shader
*ps
;
158 UINT num_gl_shaders
, shader_array_size
;
161 static const char *debug_gl_shader_type(GLenum type
)
165 #define WINED3D_TO_STR(u) case u: return #u
166 WINED3D_TO_STR(GL_VERTEX_SHADER_ARB
);
167 WINED3D_TO_STR(GL_GEOMETRY_SHADER_ARB
);
168 WINED3D_TO_STR(GL_FRAGMENT_SHADER_ARB
);
169 #undef WINED3D_TO_STR
171 return wine_dbg_sprintf("UNKNOWN(%#x)", type
);
175 static const char *shader_glsl_get_prefix(enum wined3d_shader_type type
)
179 case WINED3D_SHADER_TYPE_VERTEX
:
182 case WINED3D_SHADER_TYPE_GEOMETRY
:
185 case WINED3D_SHADER_TYPE_PIXEL
:
189 FIXME("Unhandled shader type %#x.\n", type
);
194 /* Extract a line from the info log.
195 * Note that this modifies the source string. */
196 static char *get_info_log_line(char **ptr
)
201 if (!(q
= strstr(p
, "\n")))
203 if (!*p
) return NULL
;
213 /** Prints the GLSL info log which will contain error messages if they exist */
214 /* GL locking is done by the caller */
215 static void print_glsl_info_log(const struct wined3d_gl_info
*gl_info
, GLhandleARB obj
)
217 int infologLength
= 0;
220 if (!WARN_ON(d3d_shader
) && !FIXME_ON(d3d_shader
))
223 GL_EXTCALL(glGetObjectParameterivARB(obj
,
224 GL_OBJECT_INFO_LOG_LENGTH_ARB
,
227 /* A size of 1 is just a null-terminated string, so the log should be bigger than
228 * that if there are errors. */
229 if (infologLength
> 1)
233 infoLog
= HeapAlloc(GetProcessHeap(), 0, infologLength
);
234 /* The info log is supposed to be zero-terminated, but at least some
235 * versions of fglrx don't terminate the string properly. The reported
236 * length does include the terminator, so explicitly set it to zero
238 infoLog
[infologLength
- 1] = 0;
239 GL_EXTCALL(glGetInfoLogARB(obj
, infologLength
, NULL
, infoLog
));
242 if (gl_info
->quirks
& WINED3D_QUIRK_INFO_LOG_SPAM
)
244 WARN("Info log received from GLSL shader #%u:\n", obj
);
245 while ((line
= get_info_log_line(&ptr
))) WARN(" %s\n", line
);
249 FIXME("Info log received from GLSL shader #%u:\n", obj
);
250 while ((line
= get_info_log_line(&ptr
))) FIXME(" %s\n", line
);
252 HeapFree(GetProcessHeap(), 0, infoLog
);
256 /* GL locking is done by the caller. */
257 static void shader_glsl_compile(const struct wined3d_gl_info
*gl_info
, GLhandleARB shader
, const char *src
)
259 TRACE("Compiling shader object %u.\n", shader
);
260 GL_EXTCALL(glShaderSourceARB(shader
, 1, &src
, NULL
));
261 checkGLcall("glShaderSourceARB");
262 GL_EXTCALL(glCompileShaderARB(shader
));
263 checkGLcall("glCompileShaderARB");
264 print_glsl_info_log(gl_info
, shader
);
267 /* GL locking is done by the caller. */
268 static void shader_glsl_dump_program_source(const struct wined3d_gl_info
*gl_info
, GLhandleARB program
)
270 GLint i
, object_count
, source_size
= -1;
271 GLhandleARB
*objects
;
274 GL_EXTCALL(glGetObjectParameterivARB(program
, GL_OBJECT_ATTACHED_OBJECTS_ARB
, &object_count
));
275 objects
= HeapAlloc(GetProcessHeap(), 0, object_count
* sizeof(*objects
));
278 ERR("Failed to allocate object array memory.\n");
282 GL_EXTCALL(glGetAttachedObjectsARB(program
, object_count
, NULL
, objects
));
283 for (i
= 0; i
< object_count
; ++i
)
288 GL_EXTCALL(glGetObjectParameterivARB(objects
[i
], GL_OBJECT_SHADER_SOURCE_LENGTH_ARB
, &tmp
));
290 if (source_size
< tmp
)
292 HeapFree(GetProcessHeap(), 0, source
);
294 source
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, tmp
);
297 ERR("Failed to allocate %d bytes for shader source.\n", tmp
);
298 HeapFree(GetProcessHeap(), 0, objects
);
304 FIXME("Object %u:\n", objects
[i
]);
305 GL_EXTCALL(glGetObjectParameterivARB(objects
[i
], GL_OBJECT_SUBTYPE_ARB
, &tmp
));
306 FIXME(" GL_OBJECT_SUBTYPE_ARB: %s.\n", debug_gl_shader_type(tmp
));
307 GL_EXTCALL(glGetObjectParameterivARB(objects
[i
], GL_OBJECT_COMPILE_STATUS_ARB
, &tmp
));
308 FIXME(" GL_OBJECT_COMPILE_STATUS_ARB: %d.\n", tmp
);
312 GL_EXTCALL(glGetShaderSourceARB(objects
[i
], source_size
, NULL
, source
));
313 while ((line
= get_info_log_line(&ptr
))) FIXME(" %s\n", line
);
317 HeapFree(GetProcessHeap(), 0, source
);
318 HeapFree(GetProcessHeap(), 0, objects
);
321 /* GL locking is done by the caller. */
322 static void shader_glsl_validate_link(const struct wined3d_gl_info
*gl_info
, GLhandleARB program
)
326 if (!TRACE_ON(d3d_shader
) && !FIXME_ON(d3d_shader
)) return;
328 GL_EXTCALL(glGetObjectParameterivARB(program
, GL_OBJECT_TYPE_ARB
, &tmp
));
329 if (tmp
== GL_PROGRAM_OBJECT_ARB
)
331 GL_EXTCALL(glGetObjectParameterivARB(program
, GL_OBJECT_LINK_STATUS_ARB
, &tmp
));
334 FIXME("Program %u link status invalid.\n", program
);
335 shader_glsl_dump_program_source(gl_info
, program
);
339 print_glsl_info_log(gl_info
, program
);
343 * Loads (pixel shader) samplers
345 /* GL locking is done by the caller */
346 static void shader_glsl_load_psamplers(const struct wined3d_gl_info
*gl_info
,
347 const DWORD
*tex_unit_map
, GLhandleARB programId
)
350 char sampler_name
[20];
353 for (i
= 0; i
< MAX_FRAGMENT_SAMPLERS
; ++i
)
355 snprintf(sampler_name
, sizeof(sampler_name
), "ps_sampler%u", i
);
356 name_loc
= GL_EXTCALL(glGetUniformLocationARB(programId
, sampler_name
));
357 if (name_loc
!= -1) {
358 DWORD mapped_unit
= tex_unit_map
[i
];
359 if (mapped_unit
!= WINED3D_UNMAPPED_STAGE
&& mapped_unit
< gl_info
->limits
.fragment_samplers
)
361 TRACE("Loading %s for texture %d\n", sampler_name
, mapped_unit
);
362 GL_EXTCALL(glUniform1iARB(name_loc
, mapped_unit
));
363 checkGLcall("glUniform1iARB");
365 ERR("Trying to load sampler %s on unsupported unit %d\n", sampler_name
, mapped_unit
);
371 /* GL locking is done by the caller */
372 static void shader_glsl_load_vsamplers(const struct wined3d_gl_info
*gl_info
,
373 const DWORD
*tex_unit_map
, GLhandleARB programId
)
376 char sampler_name
[20];
379 for (i
= 0; i
< MAX_VERTEX_SAMPLERS
; ++i
)
381 snprintf(sampler_name
, sizeof(sampler_name
), "vs_sampler%u", i
);
382 name_loc
= GL_EXTCALL(glGetUniformLocationARB(programId
, sampler_name
));
383 if (name_loc
!= -1) {
384 DWORD mapped_unit
= tex_unit_map
[MAX_FRAGMENT_SAMPLERS
+ i
];
385 if (mapped_unit
!= WINED3D_UNMAPPED_STAGE
&& mapped_unit
< gl_info
->limits
.combined_samplers
)
387 TRACE("Loading %s for texture %d\n", sampler_name
, mapped_unit
);
388 GL_EXTCALL(glUniform1iARB(name_loc
, mapped_unit
));
389 checkGLcall("glUniform1iARB");
391 ERR("Trying to load sampler %s on unsupported unit %d\n", sampler_name
, mapped_unit
);
397 /* GL locking is done by the caller */
398 static inline void walk_constant_heap(const struct wined3d_gl_info
*gl_info
, const float *constants
,
399 const GLint
*constant_locations
, const struct constant_heap
*heap
, unsigned char *stack
, DWORD version
)
402 unsigned int heap_idx
= 1;
405 if (heap
->entries
[heap_idx
].version
<= version
) return;
407 idx
= heap
->entries
[heap_idx
].idx
;
408 if (constant_locations
[idx
] != -1) GL_EXTCALL(glUniform4fvARB(constant_locations
[idx
], 1, &constants
[idx
* 4]));
409 stack
[stack_idx
] = HEAP_NODE_TRAVERSE_LEFT
;
411 while (stack_idx
>= 0)
413 /* Note that we fall through to the next case statement. */
414 switch(stack
[stack_idx
])
416 case HEAP_NODE_TRAVERSE_LEFT
:
418 unsigned int left_idx
= heap_idx
<< 1;
419 if (left_idx
< heap
->size
&& heap
->entries
[left_idx
].version
> version
)
422 idx
= heap
->entries
[heap_idx
].idx
;
423 if (constant_locations
[idx
] != -1)
424 GL_EXTCALL(glUniform4fvARB(constant_locations
[idx
], 1, &constants
[idx
* 4]));
426 stack
[stack_idx
++] = HEAP_NODE_TRAVERSE_RIGHT
;
427 stack
[stack_idx
] = HEAP_NODE_TRAVERSE_LEFT
;
432 case HEAP_NODE_TRAVERSE_RIGHT
:
434 unsigned int right_idx
= (heap_idx
<< 1) + 1;
435 if (right_idx
< heap
->size
&& heap
->entries
[right_idx
].version
> version
)
437 heap_idx
= right_idx
;
438 idx
= heap
->entries
[heap_idx
].idx
;
439 if (constant_locations
[idx
] != -1)
440 GL_EXTCALL(glUniform4fvARB(constant_locations
[idx
], 1, &constants
[idx
* 4]));
442 stack
[stack_idx
++] = HEAP_NODE_POP
;
443 stack
[stack_idx
] = HEAP_NODE_TRAVERSE_LEFT
;
454 checkGLcall("walk_constant_heap()");
457 /* GL locking is done by the caller */
458 static inline void apply_clamped_constant(const struct wined3d_gl_info
*gl_info
, GLint location
, const GLfloat
*data
)
460 GLfloat clamped_constant
[4];
462 if (location
== -1) return;
464 clamped_constant
[0] = data
[0] < -1.0f
? -1.0f
: data
[0] > 1.0f
? 1.0f
: data
[0];
465 clamped_constant
[1] = data
[1] < -1.0f
? -1.0f
: data
[1] > 1.0f
? 1.0f
: data
[1];
466 clamped_constant
[2] = data
[2] < -1.0f
? -1.0f
: data
[2] > 1.0f
? 1.0f
: data
[2];
467 clamped_constant
[3] = data
[3] < -1.0f
? -1.0f
: data
[3] > 1.0f
? 1.0f
: data
[3];
469 GL_EXTCALL(glUniform4fvARB(location
, 1, clamped_constant
));
472 /* GL locking is done by the caller */
473 static inline void walk_constant_heap_clamped(const struct wined3d_gl_info
*gl_info
, const float *constants
,
474 const GLint
*constant_locations
, const struct constant_heap
*heap
, unsigned char *stack
, DWORD version
)
477 unsigned int heap_idx
= 1;
480 if (heap
->entries
[heap_idx
].version
<= version
) return;
482 idx
= heap
->entries
[heap_idx
].idx
;
483 apply_clamped_constant(gl_info
, constant_locations
[idx
], &constants
[idx
* 4]);
484 stack
[stack_idx
] = HEAP_NODE_TRAVERSE_LEFT
;
486 while (stack_idx
>= 0)
488 /* Note that we fall through to the next case statement. */
489 switch(stack
[stack_idx
])
491 case HEAP_NODE_TRAVERSE_LEFT
:
493 unsigned int left_idx
= heap_idx
<< 1;
494 if (left_idx
< heap
->size
&& heap
->entries
[left_idx
].version
> version
)
497 idx
= heap
->entries
[heap_idx
].idx
;
498 apply_clamped_constant(gl_info
, constant_locations
[idx
], &constants
[idx
* 4]);
500 stack
[stack_idx
++] = HEAP_NODE_TRAVERSE_RIGHT
;
501 stack
[stack_idx
] = HEAP_NODE_TRAVERSE_LEFT
;
506 case HEAP_NODE_TRAVERSE_RIGHT
:
508 unsigned int right_idx
= (heap_idx
<< 1) + 1;
509 if (right_idx
< heap
->size
&& heap
->entries
[right_idx
].version
> version
)
511 heap_idx
= right_idx
;
512 idx
= heap
->entries
[heap_idx
].idx
;
513 apply_clamped_constant(gl_info
, constant_locations
[idx
], &constants
[idx
* 4]);
515 stack
[stack_idx
++] = HEAP_NODE_POP
;
516 stack
[stack_idx
] = HEAP_NODE_TRAVERSE_LEFT
;
527 checkGLcall("walk_constant_heap_clamped()");
530 /* Loads floating point constants (aka uniforms) into the currently set GLSL program. */
531 /* GL locking is done by the caller */
532 static void shader_glsl_load_constantsF(const struct wined3d_shader
*shader
, const struct wined3d_gl_info
*gl_info
,
533 const float *constants
, const GLint
*constant_locations
, const struct constant_heap
*heap
,
534 unsigned char *stack
, UINT version
)
536 const struct wined3d_shader_lconst
*lconst
;
538 /* 1.X pshaders have the constants clamped to [-1;1] implicitly. */
539 if (shader
->reg_maps
.shader_version
.major
== 1
540 && shader
->reg_maps
.shader_version
.type
== WINED3D_SHADER_TYPE_PIXEL
)
541 walk_constant_heap_clamped(gl_info
, constants
, constant_locations
, heap
, stack
, version
);
543 walk_constant_heap(gl_info
, constants
, constant_locations
, heap
, stack
, version
);
545 if (!shader
->load_local_constsF
)
547 TRACE("No need to load local float constants for this shader\n");
551 /* Immediate constants are clamped to [-1;1] at shader creation time if needed */
552 LIST_FOR_EACH_ENTRY(lconst
, &shader
->constantsF
, struct wined3d_shader_lconst
, entry
)
554 GLint location
= constant_locations
[lconst
->idx
];
555 /* We found this uniform name in the program - go ahead and send the data */
556 if (location
!= -1) GL_EXTCALL(glUniform4fvARB(location
, 1, (const GLfloat
*)lconst
->value
));
558 checkGLcall("glUniform4fvARB()");
561 /* Loads integer constants (aka uniforms) into the currently set GLSL program. */
562 /* GL locking is done by the caller */
563 static void shader_glsl_load_constantsI(const struct wined3d_shader
*shader
, const struct wined3d_gl_info
*gl_info
,
564 const GLint locations
[MAX_CONST_I
], const int *constants
, WORD constants_set
)
569 for (i
= 0; constants_set
; constants_set
>>= 1, ++i
)
571 if (!(constants_set
& 1)) continue;
573 TRACE_(d3d_constants
)("Loading constants %u: %i, %i, %i, %i\n",
574 i
, constants
[i
*4], constants
[i
*4+1], constants
[i
*4+2], constants
[i
*4+3]);
576 /* We found this uniform name in the program - go ahead and send the data */
577 GL_EXTCALL(glUniform4ivARB(locations
[i
], 1, &constants
[i
*4]));
578 checkGLcall("glUniform4ivARB");
581 /* Load immediate constants */
582 ptr
= list_head(&shader
->constantsI
);
585 const struct wined3d_shader_lconst
*lconst
= LIST_ENTRY(ptr
, const struct wined3d_shader_lconst
, entry
);
586 unsigned int idx
= lconst
->idx
;
587 const GLint
*values
= (const GLint
*)lconst
->value
;
589 TRACE_(d3d_constants
)("Loading local constants %i: %i, %i, %i, %i\n", idx
,
590 values
[0], values
[1], values
[2], values
[3]);
592 /* We found this uniform name in the program - go ahead and send the data */
593 GL_EXTCALL(glUniform4ivARB(locations
[idx
], 1, values
));
594 checkGLcall("glUniform4ivARB");
595 ptr
= list_next(&shader
->constantsI
, ptr
);
599 /* Loads boolean constants (aka uniforms) into the currently set GLSL program. */
600 /* GL locking is done by the caller */
601 static void shader_glsl_load_constantsB(const struct wined3d_shader
*shader
, const struct wined3d_gl_info
*gl_info
,
602 GLhandleARB programId
, const BOOL
*constants
, WORD constants_set
)
610 prefix
= shader_glsl_get_prefix(shader
->reg_maps
.shader_version
.type
);
612 /* TODO: Benchmark and see if it would be beneficial to store the
613 * locations of the constants to avoid looking up each time */
614 for (i
= 0; constants_set
; constants_set
>>= 1, ++i
)
616 if (!(constants_set
& 1)) continue;
618 TRACE_(d3d_constants
)("Loading constants %i: %i;\n", i
, constants
[i
]);
620 /* TODO: Benchmark and see if it would be beneficial to store the
621 * locations of the constants to avoid looking up each time */
622 snprintf(tmp_name
, sizeof(tmp_name
), "%s_b[%i]", prefix
, i
);
623 tmp_loc
= GL_EXTCALL(glGetUniformLocationARB(programId
, tmp_name
));
626 /* We found this uniform name in the program - go ahead and send the data */
627 GL_EXTCALL(glUniform1ivARB(tmp_loc
, 1, &constants
[i
]));
628 checkGLcall("glUniform1ivARB");
632 /* Load immediate constants */
633 ptr
= list_head(&shader
->constantsB
);
636 const struct wined3d_shader_lconst
*lconst
= LIST_ENTRY(ptr
, const struct wined3d_shader_lconst
, entry
);
637 unsigned int idx
= lconst
->idx
;
638 const GLint
*values
= (const GLint
*)lconst
->value
;
640 TRACE_(d3d_constants
)("Loading local constants %i: %i\n", idx
, values
[0]);
642 snprintf(tmp_name
, sizeof(tmp_name
), "%s_b[%i]", prefix
, idx
);
643 tmp_loc
= GL_EXTCALL(glGetUniformLocationARB(programId
, tmp_name
));
645 /* We found this uniform name in the program - go ahead and send the data */
646 GL_EXTCALL(glUniform1ivARB(tmp_loc
, 1, values
));
647 checkGLcall("glUniform1ivARB");
649 ptr
= list_next(&shader
->constantsB
, ptr
);
653 static void reset_program_constant_version(struct wine_rb_entry
*entry
, void *context
)
655 WINE_RB_ENTRY_VALUE(entry
, struct glsl_shader_prog_link
, program_lookup_entry
)->constant_version
= 0;
659 * Loads the texture dimensions for NP2 fixup into the currently set GLSL program.
661 /* GL locking is done by the caller (state handler) */
662 static void shader_glsl_load_np2fixup_constants(void *shader_priv
,
663 const struct wined3d_gl_info
*gl_info
, const struct wined3d_state
*state
)
665 struct shader_glsl_priv
*glsl_priv
= shader_priv
;
666 const struct glsl_shader_prog_link
*prog
= glsl_priv
->glsl_program
;
668 /* No GLSL program set - nothing to do. */
671 /* NP2 texcoord fixup is (currently) only done for pixelshaders. */
672 if (!use_ps(state
)) return;
674 if (prog
->np2Fixup_info
&& prog
->np2Fixup_location
!= -1)
677 UINT fixup
= prog
->np2Fixup_info
->active
;
678 GLfloat np2fixup_constants
[4 * MAX_FRAGMENT_SAMPLERS
];
680 for (i
= 0; fixup
; fixup
>>= 1, ++i
)
682 const struct wined3d_texture
*tex
= state
->textures
[i
];
683 const unsigned char idx
= prog
->np2Fixup_info
->idx
[i
];
684 GLfloat
*tex_dim
= &np2fixup_constants
[(idx
>> 1) * 4];
688 ERR("Nonexistent texture is flagged for NP2 texcoord fixup.\n");
694 tex_dim
[2] = tex
->pow2_matrix
[0];
695 tex_dim
[3] = tex
->pow2_matrix
[5];
699 tex_dim
[0] = tex
->pow2_matrix
[0];
700 tex_dim
[1] = tex
->pow2_matrix
[5];
704 GL_EXTCALL(glUniform4fvARB(prog
->np2Fixup_location
, prog
->np2Fixup_info
->num_consts
, np2fixup_constants
));
709 * Loads the app-supplied constants into the currently set GLSL program.
711 /* GL locking is done by the caller (state handler) */
712 static void shader_glsl_load_constants(const struct wined3d_context
*context
,
713 BOOL usePixelShader
, BOOL useVertexShader
)
715 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
716 struct wined3d_device
*device
= context
->swapchain
->device
;
717 struct wined3d_stateblock
*stateBlock
= device
->stateBlock
;
718 const struct wined3d_state
*state
= &stateBlock
->state
;
719 struct shader_glsl_priv
*priv
= device
->shader_priv
;
720 float position_fixup
[4];
722 GLhandleARB programId
;
723 struct glsl_shader_prog_link
*prog
= priv
->glsl_program
;
724 UINT constant_version
;
728 /* No GLSL program set - nothing to do. */
731 programId
= prog
->programId
;
732 constant_version
= prog
->constant_version
;
736 const struct wined3d_shader
*vshader
= state
->vertex_shader
;
738 /* Load DirectX 9 float constants/uniforms for vertex shader */
739 shader_glsl_load_constantsF(vshader
, gl_info
, state
->vs_consts_f
,
740 prog
->vuniformF_locations
, &priv
->vconst_heap
, priv
->stack
, constant_version
);
742 /* Load DirectX 9 integer constants/uniforms for vertex shader */
743 shader_glsl_load_constantsI(vshader
, gl_info
, prog
->vuniformI_locations
, state
->vs_consts_i
,
744 stateBlock
->changed
.vertexShaderConstantsI
& vshader
->reg_maps
.integer_constants
);
746 /* Load DirectX 9 boolean constants/uniforms for vertex shader */
747 shader_glsl_load_constantsB(vshader
, gl_info
, programId
, state
->vs_consts_b
,
748 stateBlock
->changed
.vertexShaderConstantsB
& vshader
->reg_maps
.boolean_constants
);
750 /* Upload the position fixup params */
751 shader_get_position_fixup(context
, state
, position_fixup
);
752 GL_EXTCALL(glUniform4fvARB(prog
->posFixup_location
, 1, position_fixup
));
753 checkGLcall("glUniform4fvARB");
758 const struct wined3d_shader
*pshader
= state
->pixel_shader
;
760 /* Load DirectX 9 float constants/uniforms for pixel shader */
761 shader_glsl_load_constantsF(pshader
, gl_info
, state
->ps_consts_f
,
762 prog
->puniformF_locations
, &priv
->pconst_heap
, priv
->stack
, constant_version
);
764 /* Load DirectX 9 integer constants/uniforms for pixel shader */
765 shader_glsl_load_constantsI(pshader
, gl_info
, prog
->puniformI_locations
, state
->ps_consts_i
,
766 stateBlock
->changed
.pixelShaderConstantsI
& pshader
->reg_maps
.integer_constants
);
768 /* Load DirectX 9 boolean constants/uniforms for pixel shader */
769 shader_glsl_load_constantsB(pshader
, gl_info
, programId
, state
->ps_consts_b
,
770 stateBlock
->changed
.pixelShaderConstantsB
& pshader
->reg_maps
.boolean_constants
);
772 /* Upload the environment bump map matrix if needed. The needsbumpmat member specifies the texture stage to load the matrix from.
773 * It can't be 0 for a valid texbem instruction.
775 for(i
= 0; i
< MAX_TEXTURES
; i
++) {
778 if(prog
->bumpenvmat_location
[i
] == -1) continue;
780 data
= (const float *)&state
->texture_states
[i
][WINED3D_TSS_BUMPENV_MAT00
];
781 GL_EXTCALL(glUniformMatrix2fvARB(prog
->bumpenvmat_location
[i
], 1, 0, data
));
782 checkGLcall("glUniformMatrix2fvARB");
784 /* texbeml needs the luminance scale and offset too. If texbeml
785 * is used, needsbumpmat is set too, so we can check that in the
786 * needsbumpmat check. */
787 if (prog
->luminancescale_location
[i
] != -1)
789 const GLfloat
*scale
= (const GLfloat
*)&state
->texture_states
[i
][WINED3D_TSS_BUMPENV_LSCALE
];
790 const GLfloat
*offset
= (const GLfloat
*)&state
->texture_states
[i
][WINED3D_TSS_BUMPENV_LOFFSET
];
792 GL_EXTCALL(glUniform1fvARB(prog
->luminancescale_location
[i
], 1, scale
));
793 checkGLcall("glUniform1fvARB");
794 GL_EXTCALL(glUniform1fvARB(prog
->luminanceoffset_location
[i
], 1, offset
));
795 checkGLcall("glUniform1fvARB");
799 if (prog
->ycorrection_location
!= -1)
801 float correction_params
[4];
803 if (context
->render_offscreen
)
805 correction_params
[0] = 0.0f
;
806 correction_params
[1] = 1.0f
;
808 /* position is window relative, not viewport relative */
809 correction_params
[0] = (float) context
->current_rt
->resource
.height
;
810 correction_params
[1] = -1.0f
;
812 GL_EXTCALL(glUniform4fvARB(prog
->ycorrection_location
, 1, correction_params
));
816 if (priv
->next_constant_version
== UINT_MAX
)
818 TRACE("Max constant version reached, resetting to 0.\n");
819 wine_rb_for_each_entry(&priv
->program_lookup
, reset_program_constant_version
, NULL
);
820 priv
->next_constant_version
= 1;
824 prog
->constant_version
= priv
->next_constant_version
++;
828 static void update_heap_entry(const struct constant_heap
*heap
, unsigned int idx
,
829 unsigned int heap_idx
, DWORD new_version
)
831 struct constant_entry
*entries
= heap
->entries
;
832 unsigned int *positions
= heap
->positions
;
833 unsigned int parent_idx
;
837 parent_idx
= heap_idx
>> 1;
839 if (new_version
<= entries
[parent_idx
].version
) break;
841 entries
[heap_idx
] = entries
[parent_idx
];
842 positions
[entries
[parent_idx
].idx
] = heap_idx
;
843 heap_idx
= parent_idx
;
846 entries
[heap_idx
].version
= new_version
;
847 entries
[heap_idx
].idx
= idx
;
848 positions
[idx
] = heap_idx
;
851 static void shader_glsl_update_float_vertex_constants(struct wined3d_device
*device
, UINT start
, UINT count
)
853 struct shader_glsl_priv
*priv
= device
->shader_priv
;
854 struct constant_heap
*heap
= &priv
->vconst_heap
;
857 for (i
= start
; i
< count
+ start
; ++i
)
859 if (!device
->stateBlock
->changed
.vertexShaderConstantsF
[i
])
860 update_heap_entry(heap
, i
, heap
->size
++, priv
->next_constant_version
);
862 update_heap_entry(heap
, i
, heap
->positions
[i
], priv
->next_constant_version
);
866 static void shader_glsl_update_float_pixel_constants(struct wined3d_device
*device
, UINT start
, UINT count
)
868 struct shader_glsl_priv
*priv
= device
->shader_priv
;
869 struct constant_heap
*heap
= &priv
->pconst_heap
;
872 for (i
= start
; i
< count
+ start
; ++i
)
874 if (!device
->stateBlock
->changed
.pixelShaderConstantsF
[i
])
875 update_heap_entry(heap
, i
, heap
->size
++, priv
->next_constant_version
);
877 update_heap_entry(heap
, i
, heap
->positions
[i
], priv
->next_constant_version
);
881 static unsigned int vec4_varyings(DWORD shader_major
, const struct wined3d_gl_info
*gl_info
)
883 unsigned int ret
= gl_info
->limits
.glsl_varyings
/ 4;
884 /* 4.0 shaders do not write clip coords because d3d10 does not support user clipplanes */
885 if(shader_major
> 3) return ret
;
887 /* 3.0 shaders may need an extra varying for the clip coord on some cards(mostly dx10 ones) */
888 if (gl_info
->quirks
& WINED3D_QUIRK_GLSL_CLIP_VARYING
) ret
-= 1;
892 /** Generate the variable & register declarations for the GLSL output target */
893 static void shader_generate_glsl_declarations(const struct wined3d_context
*context
,
894 struct wined3d_shader_buffer
*buffer
, const struct wined3d_shader
*shader
,
895 const struct wined3d_shader_reg_maps
*reg_maps
, const struct shader_glsl_ctx_priv
*ctx_priv
)
897 const struct wined3d_shader_version
*version
= ®_maps
->shader_version
;
898 const struct wined3d_state
*state
= &shader
->device
->stateBlock
->state
;
899 const struct ps_compile_args
*ps_args
= ctx_priv
->cur_ps_args
;
900 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
901 const struct wined3d_fb_state
*fb
= &shader
->device
->fb
;
902 unsigned int i
, extra_constants_needed
= 0;
903 const struct wined3d_shader_lconst
*lconst
;
907 prefix
= shader_glsl_get_prefix(version
->type
);
909 /* Prototype the subroutines */
910 for (i
= 0, map
= reg_maps
->labels
; map
; map
>>= 1, ++i
)
912 if (map
& 1) shader_addline(buffer
, "void subroutine%u();\n", i
);
915 /* Declare the constants (aka uniforms) */
916 if (shader
->limits
.constant_float
> 0)
918 unsigned max_constantsF
;
920 /* Unless the shader uses indirect addressing, always declare the
921 * maximum array size and ignore that we need some uniforms privately.
922 * E.g. if GL supports 256 uniforms, and we need 2 for the pos fixup
923 * and immediate values, still declare VC[256]. If the shader needs
924 * more uniforms than we have it won't work in any case. If it uses
925 * less, the compiler will figure out which uniforms are really used
926 * and strip them out. This allows a shader to use c255 on a dx9 card,
927 * as long as it doesn't also use all the other constants.
929 * If the shader uses indirect addressing the compiler must assume
930 * that all declared uniforms are used. In this case, declare only the
931 * amount that we're assured to have.
933 * Thus we run into problems in these two cases:
934 * 1) The shader really uses more uniforms than supported.
935 * 2) The shader uses indirect addressing, less constants than
936 * supported, but uses a constant index > #supported consts. */
937 if (version
->type
== WINED3D_SHADER_TYPE_PIXEL
)
939 /* No indirect addressing here. */
940 max_constantsF
= gl_info
->limits
.glsl_ps_float_constants
;
944 if (reg_maps
->usesrelconstF
)
946 /* Subtract the other potential uniforms from the max
947 * available (bools, ints, and 1 row of projection matrix).
948 * Subtract another uniform for immediate values, which have
949 * to be loaded via uniform by the driver as well. The shader
950 * code only uses 0.5, 2.0, 1.0, 128 and -128 in vertex
951 * shader code, so one vec4 should be enough. (Unfortunately
952 * the Nvidia driver doesn't store 128 and -128 in one float).
954 * Writing gl_ClipVertex requires one uniform for each
955 * clipplane as well. */
956 max_constantsF
= gl_info
->limits
.glsl_vs_float_constants
- 3;
957 if(ctx_priv
->cur_vs_args
->clip_enabled
)
959 max_constantsF
-= gl_info
->limits
.clipplanes
;
961 max_constantsF
-= count_bits(reg_maps
->integer_constants
);
962 /* Strictly speaking a bool only uses one scalar, but the nvidia(Linux) compiler doesn't pack them properly,
963 * so each scalar requires a full vec4. We could work around this by packing the booleans ourselves, but
964 * for now take this into account when calculating the number of available constants
966 max_constantsF
-= count_bits(reg_maps
->boolean_constants
);
967 /* Set by driver quirks in directx.c */
968 max_constantsF
-= gl_info
->reserved_glsl_constants
;
970 if (max_constantsF
< shader
->limits
.constant_float
)
972 static unsigned int once
;
975 ERR_(winediag
)("The hardware does not support enough uniform components to run this shader,"
976 " it may not render correctly.\n");
978 WARN("The hardware does not support enough uniform components to run this shader.\n");
983 max_constantsF
= gl_info
->limits
.glsl_vs_float_constants
;
986 max_constantsF
= min(shader
->limits
.constant_float
, max_constantsF
);
987 shader_addline(buffer
, "uniform vec4 %s_c[%u];\n", prefix
, max_constantsF
);
990 /* Always declare the full set of constants, the compiler can remove the
991 * unused ones because d3d doesn't (yet) support indirect int and bool
992 * constant addressing. This avoids problems if the app uses e.g. i0 and i9. */
993 if (shader
->limits
.constant_int
> 0 && reg_maps
->integer_constants
)
994 shader_addline(buffer
, "uniform ivec4 %s_i[%u];\n", prefix
, shader
->limits
.constant_int
);
996 if (shader
->limits
.constant_bool
> 0 && reg_maps
->boolean_constants
)
997 shader_addline(buffer
, "uniform bool %s_b[%u];\n", prefix
, shader
->limits
.constant_bool
);
999 for (i
= 0; i
< WINED3D_MAX_CBS
; ++i
)
1001 if (reg_maps
->cb_sizes
[i
])
1002 shader_addline(buffer
, "uniform vec4 %s_cb%u[%u];\n", prefix
, i
, reg_maps
->cb_sizes
[i
]);
1005 /* Declare texture samplers */
1006 for (i
= 0; i
< shader
->limits
.sampler
; ++i
)
1008 if (reg_maps
->sampler_type
[i
])
1010 BOOL shadow_sampler
= version
->type
== WINED3D_SHADER_TYPE_PIXEL
&& (ps_args
->shadow
& (1 << i
));
1011 const struct wined3d_texture
*texture
;
1013 switch (reg_maps
->sampler_type
[i
])
1017 shader_addline(buffer
, "uniform sampler1DShadow %s_sampler%u;\n", prefix
, i
);
1019 shader_addline(buffer
, "uniform sampler1D %s_sampler%u;\n", prefix
, i
);
1022 texture
= state
->textures
[i
];
1025 if (texture
&& texture
->target
== GL_TEXTURE_RECTANGLE_ARB
)
1026 shader_addline(buffer
, "uniform sampler2DRectShadow %s_sampler%u;\n", prefix
, i
);
1028 shader_addline(buffer
, "uniform sampler2DShadow %s_sampler%u;\n", prefix
, i
);
1032 if (texture
&& texture
->target
== GL_TEXTURE_RECTANGLE_ARB
)
1033 shader_addline(buffer
, "uniform sampler2DRect %s_sampler%u;\n", prefix
, i
);
1035 shader_addline(buffer
, "uniform sampler2D %s_sampler%u;\n", prefix
, i
);
1038 case WINED3DSTT_CUBE
:
1040 FIXME("Unsupported Cube shadow sampler.\n");
1041 shader_addline(buffer
, "uniform samplerCube %s_sampler%u;\n", prefix
, i
);
1043 case WINED3DSTT_VOLUME
:
1045 FIXME("Unsupported 3D shadow sampler.\n");
1046 shader_addline(buffer
, "uniform sampler3D %s_sampler%u;\n", prefix
, i
);
1049 shader_addline(buffer
, "uniform unsupported_sampler %s_sampler%u;\n", prefix
, i
);
1050 FIXME("Unrecognized sampler type: %#x\n", reg_maps
->sampler_type
[i
]);
1056 /* Declare uniforms for NP2 texcoord fixup:
1057 * This is NOT done inside the loop that declares the texture samplers
1058 * since the NP2 fixup code is currently only used for the GeforceFX
1059 * series and when forcing the ARB_npot extension off. Modern cards just
1060 * skip the code anyway, so put it inside a separate loop. */
1061 if (version
->type
== WINED3D_SHADER_TYPE_PIXEL
&& ps_args
->np2_fixup
)
1063 struct ps_np2fixup_info
*fixup
= ctx_priv
->cur_np2fixup_info
;
1066 /* NP2/RECT textures in OpenGL use texcoords in the range [0,width]x[0,height]
1067 * while D3D has them in the (normalized) [0,1]x[0,1] range.
1068 * samplerNP2Fixup stores texture dimensions and is updated through
1069 * shader_glsl_load_np2fixup_constants when the sampler changes. */
1071 for (i
= 0; i
< shader
->limits
.sampler
; ++i
)
1073 if (reg_maps
->sampler_type
[i
])
1075 if (!(ps_args
->np2_fixup
& (1 << i
))) continue;
1077 if (WINED3DSTT_2D
!= reg_maps
->sampler_type
[i
]) {
1078 FIXME("Non-2D texture is flagged for NP2 texcoord fixup.\n");
1082 fixup
->idx
[i
] = cur
++;
1086 fixup
->num_consts
= (cur
+ 1) >> 1;
1087 fixup
->active
= ps_args
->np2_fixup
;
1088 shader_addline(buffer
, "uniform vec4 %s_samplerNP2Fixup[%u];\n", prefix
, fixup
->num_consts
);
1091 /* Declare address variables */
1092 for (i
= 0, map
= reg_maps
->address
; map
; map
>>= 1, ++i
)
1094 if (map
& 1) shader_addline(buffer
, "ivec4 A%u;\n", i
);
1097 /* Declare texture coordinate temporaries and initialize them */
1098 for (i
= 0, map
= reg_maps
->texcoord
; map
; map
>>= 1, ++i
)
1100 if (map
& 1) shader_addline(buffer
, "vec4 T%u = gl_TexCoord[%u];\n", i
, i
);
1103 if (version
->type
== WINED3D_SHADER_TYPE_VERTEX
)
1105 /* Declare attributes. */
1106 for (i
= 0, map
= reg_maps
->input_registers
; map
; map
>>= 1, ++i
)
1109 shader_addline(buffer
, "attribute vec4 %s_in%u;\n", prefix
, i
);
1112 shader_addline(buffer
, "uniform vec4 posFixup;\n");
1113 shader_addline(buffer
, "void order_ps_input(in vec4[%u]);\n", shader
->limits
.packed_output
);
1115 else if (version
->type
== WINED3D_SHADER_TYPE_PIXEL
)
1117 if (version
->major
>= 3)
1119 UINT in_count
= min(vec4_varyings(version
->major
, gl_info
), shader
->limits
.packed_input
);
1122 shader_addline(buffer
, "varying vec4 %s_in[%u];\n", prefix
, in_count
);
1124 /* TODO: Write a replacement shader for the fixed function
1125 * vertex pipeline, so this isn't needed. For fixed function
1126 * vertex processing + 3.0 pixel shader we need a separate
1127 * function in the pixel shader that reads the fixed function
1128 * color into the packed input registers. */
1129 shader_addline(buffer
, "vec4 %s_in[%u];\n", prefix
, in_count
);
1132 for (i
= 0, map
= reg_maps
->bumpmat
; map
; map
>>= 1, ++i
)
1137 shader_addline(buffer
, "uniform mat2 bumpenvmat%d;\n", i
);
1139 if (reg_maps
->luminanceparams
& (1 << i
))
1141 shader_addline(buffer
, "uniform float luminancescale%d;\n", i
);
1142 shader_addline(buffer
, "uniform float luminanceoffset%d;\n", i
);
1143 extra_constants_needed
++;
1146 extra_constants_needed
++;
1149 if (ps_args
->srgb_correction
)
1151 shader_addline(buffer
, "const vec4 srgb_const0 = vec4(%.8e, %.8e, %.8e, %.8e);\n",
1152 srgb_pow
, srgb_mul_high
, srgb_sub_high
, srgb_mul_low
);
1153 shader_addline(buffer
, "const vec4 srgb_const1 = vec4(%.8e, 0.0, 0.0, 0.0);\n",
1156 if (reg_maps
->vpos
|| reg_maps
->usesdsy
)
1158 if (shader
->limits
.constant_float
+ extra_constants_needed
1159 + 1 < gl_info
->limits
.glsl_ps_float_constants
)
1161 shader_addline(buffer
, "uniform vec4 ycorrection;\n");
1162 extra_constants_needed
++;
1166 /* This happens because we do not have proper tracking of the constant registers that are
1167 * actually used, only the max limit of the shader version
1169 FIXME("Cannot find a free uniform for vpos correction params\n");
1170 shader_addline(buffer
, "const vec4 ycorrection = vec4(%f, %f, 0.0, 0.0);\n",
1171 context
->render_offscreen
? 0.0f
: fb
->render_targets
[0]->resource
.height
,
1172 context
->render_offscreen
? 1.0f
: -1.0f
);
1174 shader_addline(buffer
, "vec4 vpos;\n");
1178 /* Declare output register temporaries */
1179 if (shader
->limits
.packed_output
)
1180 shader_addline(buffer
, "vec4 %s_out[%u];\n", prefix
, shader
->limits
.packed_output
);
1182 /* Declare temporary variables */
1183 for (i
= 0, map
= reg_maps
->temporary
; map
; map
>>= 1, ++i
)
1185 if (map
& 1) shader_addline(buffer
, "vec4 R%u;\n", i
);
1188 /* Declare loop registers aLx */
1189 if (version
->major
< 4)
1191 for (i
= 0; i
< reg_maps
->loop_depth
; ++i
)
1193 shader_addline(buffer
, "int aL%u;\n", i
);
1194 shader_addline(buffer
, "int tmpInt%u;\n", i
);
1198 /* Temporary variables for matrix operations */
1199 shader_addline(buffer
, "vec4 tmp0;\n");
1200 shader_addline(buffer
, "vec4 tmp1;\n");
1202 /* Local constants use a different name so they can be loaded once at shader link time
1203 * They can't be hardcoded into the shader text via LC = {x, y, z, w}; because the
1204 * float -> string conversion can cause precision loss.
1206 if (!shader
->load_local_constsF
)
1208 LIST_FOR_EACH_ENTRY(lconst
, &shader
->constantsF
, struct wined3d_shader_lconst
, entry
)
1210 shader_addline(buffer
, "uniform vec4 %s_lc%u;\n", prefix
, lconst
->idx
);
1214 /* Start the main program. */
1215 shader_addline(buffer
, "void main()\n{\n");
1217 /* Direct3D applications expect integer vPos values, while OpenGL drivers
1218 * add approximately 0.5. This causes off-by-one problems as spotted by
1219 * the vPos d3d9 visual test. Unfortunately ATI cards do not add exactly
1220 * 0.5, but rather something like 0.49999999 or 0.50000001, which still
1221 * causes precision troubles when we just subtract 0.5.
1223 * To deal with that, just floor() the position. This will eliminate the
1224 * fraction on all cards.
1226 * TODO: Test how this behaves with multisampling.
1228 * An advantage of floor is that it works even if the driver doesn't add
1229 * 0.5. It is somewhat questionable if 1.5, 2.5, ... are the proper values
1230 * to return in gl_FragCoord, even though coordinates specify the pixel
1231 * centers instead of the pixel corners. This code will behave correctly
1232 * on drivers that returns integer values. */
1233 if (version
->type
== WINED3D_SHADER_TYPE_PIXEL
&& reg_maps
->vpos
)
1234 shader_addline(buffer
,
1235 "vpos = floor(vec4(0, ycorrection[0], 0, 0) + gl_FragCoord * vec4(1, ycorrection[1], 1, 1));\n");
1238 /*****************************************************************************
1239 * Functions to generate GLSL strings from DirectX Shader bytecode begin here.
1241 * For more information, see http://wiki.winehq.org/DirectX-Shaders
1242 ****************************************************************************/
1245 static void shader_glsl_add_src_param(const struct wined3d_shader_instruction
*ins
,
1246 const struct wined3d_shader_src_param
*wined3d_src
, DWORD mask
, struct glsl_src_param
*glsl_src
);
1248 /** Used for opcode modifiers - They multiply the result by the specified amount */
1249 static const char * const shift_glsl_tab
[] = {
1251 "2.0 * ", /* 1 (x2) */
1252 "4.0 * ", /* 2 (x4) */
1253 "8.0 * ", /* 3 (x8) */
1254 "16.0 * ", /* 4 (x16) */
1255 "32.0 * ", /* 5 (x32) */
1262 "0.0625 * ", /* 12 (d16) */
1263 "0.125 * ", /* 13 (d8) */
1264 "0.25 * ", /* 14 (d4) */
1265 "0.5 * " /* 15 (d2) */
1268 /* Generate a GLSL parameter that does the input modifier computation and return the input register/mask to use */
1269 static void shader_glsl_gen_modifier(enum wined3d_shader_src_modifier src_modifier
,
1270 const char *in_reg
, const char *in_regswizzle
, char *out_str
)
1274 switch (src_modifier
)
1276 case WINED3DSPSM_DZ
: /* Need to handle this in the instructions itself (texld & texcrd). */
1277 case WINED3DSPSM_DW
:
1278 case WINED3DSPSM_NONE
:
1279 sprintf(out_str
, "%s%s", in_reg
, in_regswizzle
);
1281 case WINED3DSPSM_NEG
:
1282 sprintf(out_str
, "-%s%s", in_reg
, in_regswizzle
);
1284 case WINED3DSPSM_NOT
:
1285 sprintf(out_str
, "!%s%s", in_reg
, in_regswizzle
);
1287 case WINED3DSPSM_BIAS
:
1288 sprintf(out_str
, "(%s%s - vec4(0.5)%s)", in_reg
, in_regswizzle
, in_regswizzle
);
1290 case WINED3DSPSM_BIASNEG
:
1291 sprintf(out_str
, "-(%s%s - vec4(0.5)%s)", in_reg
, in_regswizzle
, in_regswizzle
);
1293 case WINED3DSPSM_SIGN
:
1294 sprintf(out_str
, "(2.0 * (%s%s - 0.5))", in_reg
, in_regswizzle
);
1296 case WINED3DSPSM_SIGNNEG
:
1297 sprintf(out_str
, "-(2.0 * (%s%s - 0.5))", in_reg
, in_regswizzle
);
1299 case WINED3DSPSM_COMP
:
1300 sprintf(out_str
, "(1.0 - %s%s)", in_reg
, in_regswizzle
);
1302 case WINED3DSPSM_X2
:
1303 sprintf(out_str
, "(2.0 * %s%s)", in_reg
, in_regswizzle
);
1305 case WINED3DSPSM_X2NEG
:
1306 sprintf(out_str
, "-(2.0 * %s%s)", in_reg
, in_regswizzle
);
1308 case WINED3DSPSM_ABS
:
1309 sprintf(out_str
, "abs(%s%s)", in_reg
, in_regswizzle
);
1311 case WINED3DSPSM_ABSNEG
:
1312 sprintf(out_str
, "-abs(%s%s)", in_reg
, in_regswizzle
);
1315 FIXME("Unhandled modifier %u\n", src_modifier
);
1316 sprintf(out_str
, "%s%s", in_reg
, in_regswizzle
);
1320 /** Writes the GLSL variable name that corresponds to the register that the
1321 * DX opcode parameter is trying to access */
1322 static void shader_glsl_get_register_name(const struct wined3d_shader_register
*reg
,
1323 char *register_name
, BOOL
*is_color
, const struct wined3d_shader_instruction
*ins
)
1325 /* oPos, oFog and oPts in D3D */
1326 static const char * const hwrastout_reg_names
[] = {"vs_out[10]", "vs_out[11].x", "vs_out[11].y"};
1328 const struct wined3d_shader
*shader
= ins
->ctx
->shader
;
1329 const struct wined3d_shader_reg_maps
*reg_maps
= ins
->ctx
->reg_maps
;
1330 const struct wined3d_shader_version
*version
= ®_maps
->shader_version
;
1331 const struct wined3d_gl_info
*gl_info
= ins
->ctx
->gl_info
;
1332 const char *prefix
= shader_glsl_get_prefix(version
->type
);
1338 case WINED3DSPR_TEMP
:
1339 sprintf(register_name
, "R%u", reg
->idx
[0].offset
);
1342 case WINED3DSPR_INPUT
:
1343 /* vertex shaders */
1344 if (version
->type
== WINED3D_SHADER_TYPE_VERTEX
)
1346 struct shader_glsl_ctx_priv
*priv
= ins
->ctx
->backend_data
;
1347 if (priv
->cur_vs_args
->swizzle_map
& (1 << reg
->idx
[0].offset
))
1349 sprintf(register_name
, "%s_in%u", prefix
, reg
->idx
[0].offset
);
1353 /* pixel shaders >= 3.0 */
1354 if (version
->major
>= 3)
1356 DWORD idx
= shader
->u
.ps
.input_reg_map
[reg
->idx
[0].offset
];
1357 unsigned int in_count
= vec4_varyings(version
->major
, gl_info
);
1359 if (reg
->idx
[0].rel_addr
)
1361 struct glsl_src_param rel_param
;
1363 shader_glsl_add_src_param(ins
, reg
->idx
[0].rel_addr
, WINED3DSP_WRITEMASK_0
, &rel_param
);
1365 /* Removing a + 0 would be an obvious optimization, but macos doesn't see the NOP
1366 * operation there */
1369 if (shader
->u
.ps
.declared_in_count
> in_count
)
1371 sprintf(register_name
,
1372 "((%s + %u) > %u ? (%s + %u) > %u ? gl_SecondaryColor : gl_Color : %s_in[%s + %u])",
1373 rel_param
.param_str
, idx
, in_count
- 1, rel_param
.param_str
, idx
, in_count
,
1374 prefix
, rel_param
.param_str
, idx
);
1378 sprintf(register_name
, "%s_in[%s + %u]", prefix
, rel_param
.param_str
, idx
);
1383 if (shader
->u
.ps
.declared_in_count
> in_count
)
1385 sprintf(register_name
, "((%s) > %u ? (%s) > %u ? gl_SecondaryColor : gl_Color : %s_in[%s])",
1386 rel_param
.param_str
, in_count
- 1, rel_param
.param_str
, in_count
,
1387 prefix
, rel_param
.param_str
);
1391 sprintf(register_name
, "%s_in[%s]", prefix
, rel_param
.param_str
);
1397 if (idx
== in_count
) sprintf(register_name
, "gl_Color");
1398 else if (idx
== in_count
+ 1) sprintf(register_name
, "gl_SecondaryColor");
1399 else sprintf(register_name
, "%s_in[%u]", prefix
, idx
);
1404 if (!reg
->idx
[0].offset
)
1405 strcpy(register_name
, "gl_Color");
1407 strcpy(register_name
, "gl_SecondaryColor");
1412 case WINED3DSPR_CONST
:
1414 /* Relative addressing */
1415 if (reg
->idx
[0].rel_addr
)
1417 struct glsl_src_param rel_param
;
1418 shader_glsl_add_src_param(ins
, reg
->idx
[0].rel_addr
, WINED3DSP_WRITEMASK_0
, &rel_param
);
1419 if (reg
->idx
[0].offset
)
1420 sprintf(register_name
, "%s_c[%s + %u]", prefix
, rel_param
.param_str
, reg
->idx
[0].offset
);
1422 sprintf(register_name
, "%s_c[%s]", prefix
, rel_param
.param_str
);
1426 if (shader_constant_is_local(shader
, reg
->idx
[0].offset
))
1427 sprintf(register_name
, "%s_lc%u", prefix
, reg
->idx
[0].offset
);
1429 sprintf(register_name
, "%s_c[%u]", prefix
, reg
->idx
[0].offset
);
1434 case WINED3DSPR_CONSTINT
:
1435 sprintf(register_name
, "%s_i[%u]", prefix
, reg
->idx
[0].offset
);
1438 case WINED3DSPR_CONSTBOOL
:
1439 sprintf(register_name
, "%s_b[%u]", prefix
, reg
->idx
[0].offset
);
1442 case WINED3DSPR_TEXTURE
: /* case WINED3DSPR_ADDR: */
1443 if (version
->type
== WINED3D_SHADER_TYPE_PIXEL
)
1444 sprintf(register_name
, "T%u", reg
->idx
[0].offset
);
1446 sprintf(register_name
, "A%u", reg
->idx
[0].offset
);
1449 case WINED3DSPR_LOOP
:
1450 sprintf(register_name
, "aL%u", ins
->ctx
->loop_state
->current_reg
- 1);
1453 case WINED3DSPR_SAMPLER
:
1454 sprintf(register_name
, "%s_sampler%u", prefix
, reg
->idx
[0].offset
);
1457 case WINED3DSPR_COLOROUT
:
1458 if (reg
->idx
[0].offset
>= gl_info
->limits
.buffers
)
1459 WARN("Write to render target %u, only %d supported.\n",
1460 reg
->idx
[0].offset
, gl_info
->limits
.buffers
);
1462 sprintf(register_name
, "gl_FragData[%u]", reg
->idx
[0].offset
);
1465 case WINED3DSPR_RASTOUT
:
1466 sprintf(register_name
, "%s", hwrastout_reg_names
[reg
->idx
[0].offset
]);
1469 case WINED3DSPR_DEPTHOUT
:
1470 sprintf(register_name
, "gl_FragDepth");
1473 case WINED3DSPR_ATTROUT
:
1474 if (!reg
->idx
[0].offset
)
1475 sprintf(register_name
, "%s_out[8]", prefix
);
1477 sprintf(register_name
, "%s_out[9]", prefix
);
1480 case WINED3DSPR_TEXCRDOUT
:
1481 /* Vertex shaders >= 3.0: WINED3DSPR_OUTPUT */
1482 sprintf(register_name
, "%s_out[%u]", prefix
, reg
->idx
[0].offset
);
1485 case WINED3DSPR_MISCTYPE
:
1486 if (!reg
->idx
[0].offset
)
1489 sprintf(register_name
, "vpos");
1491 else if (reg
->idx
[0].offset
== 1)
1493 /* Note that gl_FrontFacing is a bool, while vFace is
1494 * a float for which the sign determines front/back */
1495 sprintf(register_name
, "(gl_FrontFacing ? 1.0 : -1.0)");
1499 FIXME("Unhandled misctype register %u.\n", reg
->idx
[0].offset
);
1500 sprintf(register_name
, "unrecognized_register");
1504 case WINED3DSPR_IMMCONST
:
1505 switch (reg
->immconst_type
)
1507 case WINED3D_IMMCONST_SCALAR
:
1508 switch (reg
->data_type
)
1510 case WINED3D_DATA_FLOAT
:
1511 sprintf(register_name
, "%.8e", *(const float *)reg
->immconst_data
);
1513 case WINED3D_DATA_INT
:
1514 sprintf(register_name
, "%#x", reg
->immconst_data
[0]);
1516 case WINED3D_DATA_RESOURCE
:
1517 case WINED3D_DATA_SAMPLER
:
1518 case WINED3D_DATA_UINT
:
1519 sprintf(register_name
, "%#xu", reg
->immconst_data
[0]);
1522 sprintf(register_name
, "<unhandled data type %#x>", reg
->data_type
);
1527 case WINED3D_IMMCONST_VEC4
:
1528 switch (reg
->data_type
)
1530 case WINED3D_DATA_FLOAT
:
1531 sprintf(register_name
, "vec4(%.8e, %.8e, %.8e, %.8e)",
1532 *(const float *)®
->immconst_data
[0], *(const float *)®
->immconst_data
[1],
1533 *(const float *)®
->immconst_data
[2], *(const float *)®
->immconst_data
[3]);
1535 case WINED3D_DATA_INT
:
1536 sprintf(register_name
, "ivec4(%#x, %#x, %#x, %#x)",
1537 reg
->immconst_data
[0], reg
->immconst_data
[1],
1538 reg
->immconst_data
[2], reg
->immconst_data
[3]);
1540 case WINED3D_DATA_RESOURCE
:
1541 case WINED3D_DATA_SAMPLER
:
1542 case WINED3D_DATA_UINT
:
1543 sprintf(register_name
, "uvec4(%#xu, %#xu, %#xu, %#xu)",
1544 reg
->immconst_data
[0], reg
->immconst_data
[1],
1545 reg
->immconst_data
[2], reg
->immconst_data
[3]);
1548 sprintf(register_name
, "<unhandled data type %#x>", reg
->data_type
);
1554 FIXME("Unhandled immconst type %#x\n", reg
->immconst_type
);
1555 sprintf(register_name
, "<unhandled_immconst_type %#x>", reg
->immconst_type
);
1559 case WINED3DSPR_CONSTBUFFER
:
1560 if (reg
->idx
[1].rel_addr
)
1562 struct glsl_src_param rel_param
;
1563 shader_glsl_add_src_param(ins
, reg
->idx
[1].rel_addr
, WINED3DSP_WRITEMASK_0
, &rel_param
);
1564 sprintf(register_name
, "%s_cb%u[%s + %u]",
1565 prefix
, reg
->idx
[0].offset
, rel_param
.param_str
, reg
->idx
[1].offset
);
1569 sprintf(register_name
, "%s_cb%u[%u]", prefix
, reg
->idx
[0].offset
, reg
->idx
[1].offset
);
1573 case WINED3DSPR_PRIMID
:
1574 sprintf(register_name
, "uint(gl_PrimitiveIDIn)");
1578 FIXME("Unhandled register type %#x.\n", reg
->type
);
1579 sprintf(register_name
, "unrecognized_register");
1584 static void shader_glsl_write_mask_to_str(DWORD write_mask
, char *str
)
1587 if (write_mask
& WINED3DSP_WRITEMASK_0
) *str
++ = 'x';
1588 if (write_mask
& WINED3DSP_WRITEMASK_1
) *str
++ = 'y';
1589 if (write_mask
& WINED3DSP_WRITEMASK_2
) *str
++ = 'z';
1590 if (write_mask
& WINED3DSP_WRITEMASK_3
) *str
++ = 'w';
1594 /* Get the GLSL write mask for the destination register */
1595 static DWORD
shader_glsl_get_write_mask(const struct wined3d_shader_dst_param
*param
, char *write_mask
)
1597 DWORD mask
= param
->write_mask
;
1599 if (shader_is_scalar(¶m
->reg
))
1601 mask
= WINED3DSP_WRITEMASK_0
;
1606 shader_glsl_write_mask_to_str(mask
, write_mask
);
1612 static unsigned int shader_glsl_get_write_mask_size(DWORD write_mask
) {
1613 unsigned int size
= 0;
1615 if (write_mask
& WINED3DSP_WRITEMASK_0
) ++size
;
1616 if (write_mask
& WINED3DSP_WRITEMASK_1
) ++size
;
1617 if (write_mask
& WINED3DSP_WRITEMASK_2
) ++size
;
1618 if (write_mask
& WINED3DSP_WRITEMASK_3
) ++size
;
1623 static void shader_glsl_swizzle_to_str(const DWORD swizzle
, BOOL fixup
, DWORD mask
, char *str
)
1625 /* For registers of type WINED3DDECLTYPE_D3DCOLOR, data is stored as "bgra",
1626 * but addressed as "rgba". To fix this we need to swap the register's x
1627 * and z components. */
1628 const char *swizzle_chars
= fixup
? "zyxw" : "xyzw";
1631 /* swizzle bits fields: wwzzyyxx */
1632 if (mask
& WINED3DSP_WRITEMASK_0
) *str
++ = swizzle_chars
[swizzle
& 0x03];
1633 if (mask
& WINED3DSP_WRITEMASK_1
) *str
++ = swizzle_chars
[(swizzle
>> 2) & 0x03];
1634 if (mask
& WINED3DSP_WRITEMASK_2
) *str
++ = swizzle_chars
[(swizzle
>> 4) & 0x03];
1635 if (mask
& WINED3DSP_WRITEMASK_3
) *str
++ = swizzle_chars
[(swizzle
>> 6) & 0x03];
1639 static void shader_glsl_get_swizzle(const struct wined3d_shader_src_param
*param
,
1640 BOOL fixup
, DWORD mask
, char *swizzle_str
)
1642 if (shader_is_scalar(¶m
->reg
))
1643 *swizzle_str
= '\0';
1645 shader_glsl_swizzle_to_str(param
->swizzle
, fixup
, mask
, swizzle_str
);
1648 /* From a given parameter token, generate the corresponding GLSL string.
1649 * Also, return the actual register name and swizzle in case the
1650 * caller needs this information as well. */
1651 static void shader_glsl_add_src_param(const struct wined3d_shader_instruction
*ins
,
1652 const struct wined3d_shader_src_param
*wined3d_src
, DWORD mask
, struct glsl_src_param
*glsl_src
)
1654 BOOL is_color
= FALSE
;
1655 char swizzle_str
[6];
1657 glsl_src
->reg_name
[0] = '\0';
1658 glsl_src
->param_str
[0] = '\0';
1659 swizzle_str
[0] = '\0';
1661 shader_glsl_get_register_name(&wined3d_src
->reg
, glsl_src
->reg_name
, &is_color
, ins
);
1662 shader_glsl_get_swizzle(wined3d_src
, is_color
, mask
, swizzle_str
);
1664 if (wined3d_src
->reg
.type
== WINED3DSPR_IMMCONST
|| wined3d_src
->reg
.type
== WINED3DSPR_PRIMID
)
1666 shader_glsl_gen_modifier(wined3d_src
->modifiers
, glsl_src
->reg_name
, swizzle_str
, glsl_src
->param_str
);
1670 char param_str
[200];
1672 shader_glsl_gen_modifier(wined3d_src
->modifiers
, glsl_src
->reg_name
, swizzle_str
, param_str
);
1674 switch (wined3d_src
->reg
.data_type
)
1676 case WINED3D_DATA_FLOAT
:
1677 sprintf(glsl_src
->param_str
, "%s", param_str
);
1679 case WINED3D_DATA_INT
:
1680 sprintf(glsl_src
->param_str
, "floatBitsToInt(%s)", param_str
);
1682 case WINED3D_DATA_RESOURCE
:
1683 case WINED3D_DATA_SAMPLER
:
1684 case WINED3D_DATA_UINT
:
1685 sprintf(glsl_src
->param_str
, "floatBitsToUint(%s)", param_str
);
1688 FIXME("Unhandled data type %#x.\n", wined3d_src
->reg
.data_type
);
1689 sprintf(glsl_src
->param_str
, "%s", param_str
);
1695 /* From a given parameter token, generate the corresponding GLSL string.
1696 * Also, return the actual register name and swizzle in case the
1697 * caller needs this information as well. */
1698 static DWORD
shader_glsl_add_dst_param(const struct wined3d_shader_instruction
*ins
,
1699 const struct wined3d_shader_dst_param
*wined3d_dst
, struct glsl_dst_param
*glsl_dst
)
1701 BOOL is_color
= FALSE
;
1703 glsl_dst
->mask_str
[0] = '\0';
1704 glsl_dst
->reg_name
[0] = '\0';
1706 shader_glsl_get_register_name(&wined3d_dst
->reg
, glsl_dst
->reg_name
, &is_color
, ins
);
1707 return shader_glsl_get_write_mask(wined3d_dst
, glsl_dst
->mask_str
);
1710 /* Append the destination part of the instruction to the buffer, return the effective write mask */
1711 static DWORD
shader_glsl_append_dst_ext(struct wined3d_shader_buffer
*buffer
,
1712 const struct wined3d_shader_instruction
*ins
, const struct wined3d_shader_dst_param
*dst
)
1714 struct glsl_dst_param glsl_dst
;
1717 if ((mask
= shader_glsl_add_dst_param(ins
, dst
, &glsl_dst
)))
1719 switch (dst
->reg
.data_type
)
1721 case WINED3D_DATA_FLOAT
:
1722 shader_addline(buffer
, "%s%s = %s(",
1723 glsl_dst
.reg_name
, glsl_dst
.mask_str
, shift_glsl_tab
[dst
->shift
]);
1725 case WINED3D_DATA_INT
:
1726 shader_addline(buffer
, "%s%s = %sintBitsToFloat(",
1727 glsl_dst
.reg_name
, glsl_dst
.mask_str
, shift_glsl_tab
[dst
->shift
]);
1729 case WINED3D_DATA_RESOURCE
:
1730 case WINED3D_DATA_SAMPLER
:
1731 case WINED3D_DATA_UINT
:
1732 shader_addline(buffer
, "%s%s = %suintBitsToFloat(",
1733 glsl_dst
.reg_name
, glsl_dst
.mask_str
, shift_glsl_tab
[dst
->shift
]);
1736 FIXME("Unhandled data type %#x.\n", dst
->reg
.data_type
);
1737 shader_addline(buffer
, "%s%s = %s(",
1738 glsl_dst
.reg_name
, glsl_dst
.mask_str
, shift_glsl_tab
[dst
->shift
]);
1746 /* Append the destination part of the instruction to the buffer, return the effective write mask */
1747 static DWORD
shader_glsl_append_dst(struct wined3d_shader_buffer
*buffer
, const struct wined3d_shader_instruction
*ins
)
1749 return shader_glsl_append_dst_ext(buffer
, ins
, &ins
->dst
[0]);
1752 /** Process GLSL instruction modifiers */
1753 static void shader_glsl_add_instruction_modifiers(const struct wined3d_shader_instruction
*ins
)
1755 struct glsl_dst_param dst_param
;
1758 if (!ins
->dst_count
) return;
1760 modifiers
= ins
->dst
[0].modifiers
;
1761 if (!modifiers
) return;
1763 shader_glsl_add_dst_param(ins
, &ins
->dst
[0], &dst_param
);
1765 if (modifiers
& WINED3DSPDM_SATURATE
)
1767 /* _SAT means to clamp the value of the register to between 0 and 1 */
1768 shader_addline(ins
->ctx
->buffer
, "%s%s = clamp(%s%s, 0.0, 1.0);\n", dst_param
.reg_name
,
1769 dst_param
.mask_str
, dst_param
.reg_name
, dst_param
.mask_str
);
1772 if (modifiers
& WINED3DSPDM_MSAMPCENTROID
)
1774 FIXME("_centroid modifier not handled\n");
1777 if (modifiers
& WINED3DSPDM_PARTIALPRECISION
)
1779 /* MSDN says this modifier can be safely ignored, so that's what we'll do. */
1783 static const char *shader_glsl_get_rel_op(enum wined3d_shader_rel_op op
)
1787 case WINED3D_SHADER_REL_OP_GT
: return ">";
1788 case WINED3D_SHADER_REL_OP_EQ
: return "==";
1789 case WINED3D_SHADER_REL_OP_GE
: return ">=";
1790 case WINED3D_SHADER_REL_OP_LT
: return "<";
1791 case WINED3D_SHADER_REL_OP_NE
: return "!=";
1792 case WINED3D_SHADER_REL_OP_LE
: return "<=";
1794 FIXME("Unrecognized operator %#x.\n", op
);
1799 static void shader_glsl_get_sample_function(const struct wined3d_shader_context
*ctx
,
1800 DWORD sampler_idx
, DWORD flags
, struct glsl_sample_function
*sample_function
)
1802 enum wined3d_sampler_texture_type sampler_type
= ctx
->reg_maps
->sampler_type
[sampler_idx
];
1803 const struct wined3d_gl_info
*gl_info
= ctx
->gl_info
;
1804 BOOL shadow
= ctx
->reg_maps
->shader_version
.type
== WINED3D_SHADER_TYPE_PIXEL
1805 && (((const struct shader_glsl_ctx_priv
*)ctx
->backend_data
)->cur_ps_args
->shadow
& (1 << sampler_idx
));
1806 BOOL projected
= flags
& WINED3D_GLSL_SAMPLE_PROJECTED
;
1807 BOOL texrect
= flags
& WINED3D_GLSL_SAMPLE_RECT
;
1808 BOOL lod
= flags
& WINED3D_GLSL_SAMPLE_LOD
;
1809 BOOL grad
= flags
& WINED3D_GLSL_SAMPLE_GRAD
;
1811 /* Note that there's no such thing as a projected cube texture. */
1812 switch(sampler_type
) {
1818 sample_function
->name
= projected
? "shadow1DProjLod" : "shadow1DLod";
1822 if (gl_info
->supported
[EXT_GPU_SHADER4
])
1823 sample_function
->name
= projected
? "shadow1DProjGrad" : "shadow1DGrad";
1824 else if (gl_info
->supported
[ARB_SHADER_TEXTURE_LOD
])
1825 sample_function
->name
= projected
? "shadow1DProjGradARB" : "shadow1DGradARB";
1828 FIXME("Unsupported 1D shadow grad function.\n");
1829 sample_function
->name
= "unsupported1DGrad";
1834 sample_function
->name
= projected
? "shadow1DProj" : "shadow1D";
1836 sample_function
->coord_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
;
1842 sample_function
->name
= projected
? "texture1DProjLod" : "texture1DLod";
1846 if (gl_info
->supported
[EXT_GPU_SHADER4
])
1847 sample_function
->name
= projected
? "texture1DProjGrad" : "texture1DGrad";
1848 else if (gl_info
->supported
[ARB_SHADER_TEXTURE_LOD
])
1849 sample_function
->name
= projected
? "texture1DProjGradARB" : "texture1DGradARB";
1852 FIXME("Unsupported 1D grad function.\n");
1853 sample_function
->name
= "unsupported1DGrad";
1858 sample_function
->name
= projected
? "texture1DProj" : "texture1D";
1860 sample_function
->coord_mask
= WINED3DSP_WRITEMASK_0
;
1871 sample_function
->name
= projected
? "shadow2DRectProjLod" : "shadow2DRectLod";
1875 if (gl_info
->supported
[EXT_GPU_SHADER4
])
1876 sample_function
->name
= projected
? "shadow2DRectProjGrad" : "shadow2DRectGrad";
1877 else if (gl_info
->supported
[ARB_SHADER_TEXTURE_LOD
])
1878 sample_function
->name
= projected
? "shadow2DRectProjGradARB" : "shadow2DRectGradARB";
1881 FIXME("Unsupported RECT shadow grad function.\n");
1882 sample_function
->name
= "unsupported2DRectGrad";
1887 sample_function
->name
= projected
? "shadow2DRectProj" : "shadow2DRect";
1894 sample_function
->name
= projected
? "shadow2DProjLod" : "shadow2DLod";
1898 if (gl_info
->supported
[EXT_GPU_SHADER4
])
1899 sample_function
->name
= projected
? "shadow2DProjGrad" : "shadow2DGrad";
1900 else if (gl_info
->supported
[ARB_SHADER_TEXTURE_LOD
])
1901 sample_function
->name
= projected
? "shadow2DProjGradARB" : "shadow2DGradARB";
1904 FIXME("Unsupported 2D shadow grad function.\n");
1905 sample_function
->name
= "unsupported2DGrad";
1910 sample_function
->name
= projected
? "shadow2DProj" : "shadow2D";
1913 sample_function
->coord_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
1921 sample_function
->name
= projected
? "texture2DRectProjLod" : "texture2DRectLod";
1925 if (gl_info
->supported
[EXT_GPU_SHADER4
])
1926 sample_function
->name
= projected
? "texture2DRectProjGrad" : "texture2DRectGrad";
1927 else if (gl_info
->supported
[ARB_SHADER_TEXTURE_LOD
])
1928 sample_function
->name
= projected
? "texture2DRectProjGradARB" : "texture2DRectGradARB";
1931 FIXME("Unsupported RECT grad function.\n");
1932 sample_function
->name
= "unsupported2DRectGrad";
1937 sample_function
->name
= projected
? "texture2DRectProj" : "texture2DRect";
1944 sample_function
->name
= projected
? "texture2DProjLod" : "texture2DLod";
1948 if (gl_info
->supported
[EXT_GPU_SHADER4
])
1949 sample_function
->name
= projected
? "texture2DProjGrad" : "texture2DGrad";
1950 else if (gl_info
->supported
[ARB_SHADER_TEXTURE_LOD
])
1951 sample_function
->name
= projected
? "texture2DProjGradARB" : "texture2DGradARB";
1954 FIXME("Unsupported 2D grad function.\n");
1955 sample_function
->name
= "unsupported2DGrad";
1960 sample_function
->name
= projected
? "texture2DProj" : "texture2D";
1963 sample_function
->coord_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
;
1967 case WINED3DSTT_CUBE
:
1970 FIXME("Unsupported Cube shadow function.\n");
1971 sample_function
->name
= "unsupportedCubeShadow";
1972 sample_function
->coord_mask
= 0;
1978 sample_function
->name
= "textureCubeLod";
1982 if (gl_info
->supported
[EXT_GPU_SHADER4
])
1983 sample_function
->name
= "textureCubeGrad";
1984 else if (gl_info
->supported
[ARB_SHADER_TEXTURE_LOD
])
1985 sample_function
->name
= "textureCubeGradARB";
1988 FIXME("Unsupported Cube grad function.\n");
1989 sample_function
->name
= "unsupportedCubeGrad";
1994 sample_function
->name
= "textureCube";
1996 sample_function
->coord_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2000 case WINED3DSTT_VOLUME
:
2003 FIXME("Unsupported 3D shadow function.\n");
2004 sample_function
->name
= "unsupported3DShadow";
2005 sample_function
->coord_mask
= 0;
2011 sample_function
->name
= projected
? "texture3DProjLod" : "texture3DLod";
2015 if (gl_info
->supported
[EXT_GPU_SHADER4
])
2016 sample_function
->name
= projected
? "texture3DProjGrad" : "texture3DGrad";
2017 else if (gl_info
->supported
[ARB_SHADER_TEXTURE_LOD
])
2018 sample_function
->name
= projected
? "texture3DProjGradARB" : "texture3DGradARB";
2021 FIXME("Unsupported 3D grad function.\n");
2022 sample_function
->name
= "unsupported3DGrad";
2027 sample_function
->name
= projected
? "texture3DProj" : "texture3D";
2029 sample_function
->coord_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2034 sample_function
->name
= "";
2035 sample_function
->coord_mask
= 0;
2036 FIXME("Unrecognized sampler type: %#x;\n", sampler_type
);
2041 static void shader_glsl_append_fixup_arg(char *arguments
, const char *reg_name
,
2042 BOOL sign_fixup
, enum fixup_channel_source channel_source
)
2044 switch(channel_source
)
2046 case CHANNEL_SOURCE_ZERO
:
2047 strcat(arguments
, "0.0");
2050 case CHANNEL_SOURCE_ONE
:
2051 strcat(arguments
, "1.0");
2054 case CHANNEL_SOURCE_X
:
2055 strcat(arguments
, reg_name
);
2056 strcat(arguments
, ".x");
2059 case CHANNEL_SOURCE_Y
:
2060 strcat(arguments
, reg_name
);
2061 strcat(arguments
, ".y");
2064 case CHANNEL_SOURCE_Z
:
2065 strcat(arguments
, reg_name
);
2066 strcat(arguments
, ".z");
2069 case CHANNEL_SOURCE_W
:
2070 strcat(arguments
, reg_name
);
2071 strcat(arguments
, ".w");
2075 FIXME("Unhandled channel source %#x\n", channel_source
);
2076 strcat(arguments
, "undefined");
2080 if (sign_fixup
) strcat(arguments
, " * 2.0 - 1.0");
2083 static void shader_glsl_color_correction(const struct wined3d_shader_instruction
*ins
, struct color_fixup_desc fixup
)
2085 struct wined3d_shader_dst_param dst
;
2086 unsigned int mask_size
, remaining
;
2087 struct glsl_dst_param dst_param
;
2088 char arguments
[256];
2092 if (fixup
.x_sign_fixup
|| fixup
.x_source
!= CHANNEL_SOURCE_X
) mask
|= WINED3DSP_WRITEMASK_0
;
2093 if (fixup
.y_sign_fixup
|| fixup
.y_source
!= CHANNEL_SOURCE_Y
) mask
|= WINED3DSP_WRITEMASK_1
;
2094 if (fixup
.z_sign_fixup
|| fixup
.z_source
!= CHANNEL_SOURCE_Z
) mask
|= WINED3DSP_WRITEMASK_2
;
2095 if (fixup
.w_sign_fixup
|| fixup
.w_source
!= CHANNEL_SOURCE_W
) mask
|= WINED3DSP_WRITEMASK_3
;
2096 mask
&= ins
->dst
[0].write_mask
;
2098 if (!mask
) return; /* Nothing to do */
2100 if (is_complex_fixup(fixup
))
2102 enum complex_fixup complex_fixup
= get_complex_fixup(fixup
);
2103 FIXME("Complex fixup (%#x) not supported\n",complex_fixup
);
2107 mask_size
= shader_glsl_get_write_mask_size(mask
);
2110 dst
.write_mask
= mask
;
2111 shader_glsl_add_dst_param(ins
, &dst
, &dst_param
);
2113 arguments
[0] = '\0';
2114 remaining
= mask_size
;
2115 if (mask
& WINED3DSP_WRITEMASK_0
)
2117 shader_glsl_append_fixup_arg(arguments
, dst_param
.reg_name
, fixup
.x_sign_fixup
, fixup
.x_source
);
2118 if (--remaining
) strcat(arguments
, ", ");
2120 if (mask
& WINED3DSP_WRITEMASK_1
)
2122 shader_glsl_append_fixup_arg(arguments
, dst_param
.reg_name
, fixup
.y_sign_fixup
, fixup
.y_source
);
2123 if (--remaining
) strcat(arguments
, ", ");
2125 if (mask
& WINED3DSP_WRITEMASK_2
)
2127 shader_glsl_append_fixup_arg(arguments
, dst_param
.reg_name
, fixup
.z_sign_fixup
, fixup
.z_source
);
2128 if (--remaining
) strcat(arguments
, ", ");
2130 if (mask
& WINED3DSP_WRITEMASK_3
)
2132 shader_glsl_append_fixup_arg(arguments
, dst_param
.reg_name
, fixup
.w_sign_fixup
, fixup
.w_source
);
2133 if (--remaining
) strcat(arguments
, ", ");
2138 shader_addline(ins
->ctx
->buffer
, "%s%s = vec%u(%s);\n",
2139 dst_param
.reg_name
, dst_param
.mask_str
, mask_size
, arguments
);
2143 shader_addline(ins
->ctx
->buffer
, "%s%s = %s;\n", dst_param
.reg_name
, dst_param
.mask_str
, arguments
);
2147 static void PRINTF_ATTR(8, 9) shader_glsl_gen_sample_code(const struct wined3d_shader_instruction
*ins
,
2148 DWORD sampler
, const struct glsl_sample_function
*sample_function
, DWORD swizzle
,
2149 const char *dx
, const char *dy
, const char *bias
, const char *coord_reg_fmt
, ...)
2151 const struct wined3d_shader_version
*version
= &ins
->ctx
->reg_maps
->shader_version
;
2152 char dst_swizzle
[6];
2153 struct color_fixup_desc fixup
;
2154 BOOL np2_fixup
= FALSE
;
2157 shader_glsl_swizzle_to_str(swizzle
, FALSE
, ins
->dst
[0].write_mask
, dst_swizzle
);
2159 if (version
->type
== WINED3D_SHADER_TYPE_PIXEL
)
2161 const struct shader_glsl_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2162 fixup
= priv
->cur_ps_args
->color_fixup
[sampler
];
2164 if(priv
->cur_ps_args
->np2_fixup
& (1 << sampler
)) {
2166 FIXME("Biased sampling from NP2 textures is unsupported\n");
2174 fixup
= COLOR_FIXUP_IDENTITY
; /* FIXME: Vshader color fixup */
2177 shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2179 shader_addline(ins
->ctx
->buffer
, "%s(%s_sampler%u, ",
2180 sample_function
->name
, shader_glsl_get_prefix(version
->type
), sampler
);
2182 va_start(args
, coord_reg_fmt
);
2183 shader_vaddline(ins
->ctx
->buffer
, coord_reg_fmt
, args
);
2187 shader_addline(ins
->ctx
->buffer
, ", %s)%s);\n", bias
, dst_swizzle
);
2190 const struct shader_glsl_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2191 const unsigned char idx
= priv
->cur_np2fixup_info
->idx
[sampler
];
2193 shader_addline(ins
->ctx
->buffer
, " * ps_samplerNP2Fixup[%u].%s)%s);\n", idx
>> 1,
2194 (idx
% 2) ? "zw" : "xy", dst_swizzle
);
2195 } else if(dx
&& dy
) {
2196 shader_addline(ins
->ctx
->buffer
, ", %s, %s)%s);\n", dx
, dy
, dst_swizzle
);
2198 shader_addline(ins
->ctx
->buffer
, ")%s);\n", dst_swizzle
);
2202 if(!is_identity_fixup(fixup
)) {
2203 shader_glsl_color_correction(ins
, fixup
);
2207 /*****************************************************************************
2208 * Begin processing individual instruction opcodes
2209 ****************************************************************************/
2211 static void shader_glsl_binop(const struct wined3d_shader_instruction
*ins
)
2213 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2214 struct glsl_src_param src0_param
;
2215 struct glsl_src_param src1_param
;
2219 /* Determine the GLSL operator to use based on the opcode */
2220 switch (ins
->handler_idx
)
2222 case WINED3DSIH_ADD
: op
= "+"; break;
2223 case WINED3DSIH_AND
: op
= "&"; break;
2224 case WINED3DSIH_DIV
: op
= "/"; break;
2225 case WINED3DSIH_IADD
: op
= "+"; break;
2226 case WINED3DSIH_MUL
: op
= "*"; break;
2227 case WINED3DSIH_SUB
: op
= "-"; break;
2228 case WINED3DSIH_USHR
: op
= ">>"; break;
2229 case WINED3DSIH_XOR
: op
= "^"; break;
2231 op
= "<unhandled operator>";
2232 FIXME("Opcode %#x not yet handled in GLSL\n", ins
->handler_idx
);
2236 write_mask
= shader_glsl_append_dst(buffer
, ins
);
2237 shader_glsl_add_src_param(ins
, &ins
->src
[0], write_mask
, &src0_param
);
2238 shader_glsl_add_src_param(ins
, &ins
->src
[1], write_mask
, &src1_param
);
2239 shader_addline(buffer
, "%s %s %s);\n", src0_param
.param_str
, op
, src1_param
.param_str
);
2242 static void shader_glsl_relop(const struct wined3d_shader_instruction
*ins
)
2244 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2245 struct glsl_src_param src0_param
;
2246 struct glsl_src_param src1_param
;
2247 unsigned int mask_size
;
2251 write_mask
= shader_glsl_append_dst(buffer
, ins
);
2252 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
2253 shader_glsl_add_src_param(ins
, &ins
->src
[0], write_mask
, &src0_param
);
2254 shader_glsl_add_src_param(ins
, &ins
->src
[1], write_mask
, &src1_param
);
2258 switch (ins
->handler_idx
)
2260 case WINED3DSIH_EQ
: op
= "equal"; break;
2261 case WINED3DSIH_GE
: op
= "greaterThanEqual"; break;
2262 case WINED3DSIH_IGE
: op
= "greaterThanEqual"; break;
2263 case WINED3DSIH_LT
: op
= "lessThan"; break;
2265 op
= "<unhandled operator>";
2266 ERR("Unhandled opcode %#x.\n", ins
->handler_idx
);
2270 shader_addline(buffer
, "uvec%u(%s(%s, %s)) * 0xffffffffu);\n",
2271 mask_size
, op
, src0_param
.param_str
, src1_param
.param_str
);
2275 switch (ins
->handler_idx
)
2277 case WINED3DSIH_EQ
: op
= "=="; break;
2278 case WINED3DSIH_GE
: op
= ">="; break;
2279 case WINED3DSIH_IGE
: op
= ">="; break;
2280 case WINED3DSIH_LT
: op
= "<"; break;
2282 op
= "<unhandled operator>";
2283 ERR("Unhandled opcode %#x.\n", ins
->handler_idx
);
2287 shader_addline(buffer
, "%s %s %s ? 0xffffffffu : 0u);\n",
2288 src0_param
.param_str
, op
, src1_param
.param_str
);
2292 static void shader_glsl_imul(const struct wined3d_shader_instruction
*ins
)
2294 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2295 struct glsl_src_param src0_param
;
2296 struct glsl_src_param src1_param
;
2299 /* If we have ARB_gpu_shader5 or GLSL 4.0, we can use imulExtended(). If
2300 * not, we can emulate it. */
2301 if (ins
->dst
[0].reg
.type
!= WINED3DSPR_NULL
)
2302 FIXME("64-bit integer multiplies not implemented.\n");
2304 if (ins
->dst
[1].reg
.type
!= WINED3DSPR_NULL
)
2306 write_mask
= shader_glsl_append_dst_ext(buffer
, ins
, &ins
->dst
[1]);
2307 shader_glsl_add_src_param(ins
, &ins
->src
[0], write_mask
, &src0_param
);
2308 shader_glsl_add_src_param(ins
, &ins
->src
[1], write_mask
, &src1_param
);
2310 shader_addline(ins
->ctx
->buffer
, "%s * %s);\n",
2311 src0_param
.param_str
, src1_param
.param_str
);
2315 static void shader_glsl_udiv(const struct wined3d_shader_instruction
*ins
)
2317 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2318 struct glsl_src_param src0_param
, src1_param
;
2321 if (ins
->dst
[0].reg
.type
!= WINED3DSPR_NULL
)
2324 if (ins
->dst
[1].reg
.type
!= WINED3DSPR_NULL
)
2328 write_mask
= shader_glsl_get_write_mask(&ins
->dst
[0], dst_mask
);
2329 shader_glsl_add_src_param(ins
, &ins
->src
[0], write_mask
, &src0_param
);
2330 shader_glsl_add_src_param(ins
, &ins
->src
[1], write_mask
, &src1_param
);
2331 shader_addline(buffer
, "tmp0%s = %s / %s;\n",
2332 dst_mask
, src0_param
.param_str
, src1_param
.param_str
);
2334 write_mask
= shader_glsl_append_dst_ext(buffer
, ins
, &ins
->dst
[1]);
2335 shader_glsl_add_src_param(ins
, &ins
->src
[0], write_mask
, &src0_param
);
2336 shader_glsl_add_src_param(ins
, &ins
->src
[1], write_mask
, &src1_param
);
2337 shader_addline(buffer
, "%s %% %s));\n", src0_param
.param_str
, src1_param
.param_str
);
2339 shader_glsl_append_dst_ext(buffer
, ins
, &ins
->dst
[0]);
2340 shader_addline(buffer
, "tmp0%s);\n", dst_mask
);
2344 write_mask
= shader_glsl_append_dst_ext(buffer
, ins
, &ins
->dst
[0]);
2345 shader_glsl_add_src_param(ins
, &ins
->src
[0], write_mask
, &src0_param
);
2346 shader_glsl_add_src_param(ins
, &ins
->src
[1], write_mask
, &src1_param
);
2347 shader_addline(buffer
, "%s / %s);\n", src0_param
.param_str
, src1_param
.param_str
);
2350 else if (ins
->dst
[1].reg
.type
!= WINED3DSPR_NULL
)
2352 write_mask
= shader_glsl_append_dst_ext(buffer
, ins
, &ins
->dst
[1]);
2353 shader_glsl_add_src_param(ins
, &ins
->src
[0], write_mask
, &src0_param
);
2354 shader_glsl_add_src_param(ins
, &ins
->src
[1], write_mask
, &src1_param
);
2355 shader_addline(buffer
, "%s %% %s);\n", src0_param
.param_str
, src1_param
.param_str
);
2359 /* Process the WINED3DSIO_MOV opcode using GLSL (dst = src) */
2360 static void shader_glsl_mov(const struct wined3d_shader_instruction
*ins
)
2362 const struct wined3d_gl_info
*gl_info
= ins
->ctx
->gl_info
;
2363 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2364 struct glsl_src_param src0_param
;
2367 write_mask
= shader_glsl_append_dst(buffer
, ins
);
2368 shader_glsl_add_src_param(ins
, &ins
->src
[0], write_mask
, &src0_param
);
2370 /* In vs_1_1 WINED3DSIO_MOV can write to the address register. In later
2371 * shader versions WINED3DSIO_MOVA is used for this. */
2372 if (ins
->ctx
->reg_maps
->shader_version
.major
== 1
2373 && ins
->ctx
->reg_maps
->shader_version
.type
== WINED3D_SHADER_TYPE_VERTEX
2374 && ins
->dst
[0].reg
.type
== WINED3DSPR_ADDR
)
2376 /* This is a simple floor() */
2377 unsigned int mask_size
= shader_glsl_get_write_mask_size(write_mask
);
2378 if (mask_size
> 1) {
2379 shader_addline(buffer
, "ivec%d(floor(%s)));\n", mask_size
, src0_param
.param_str
);
2381 shader_addline(buffer
, "int(floor(%s)));\n", src0_param
.param_str
);
2384 else if(ins
->handler_idx
== WINED3DSIH_MOVA
)
2386 /* We need to *round* to the nearest int here. */
2387 unsigned int mask_size
= shader_glsl_get_write_mask_size(write_mask
);
2389 if (gl_info
->supported
[EXT_GPU_SHADER4
])
2392 shader_addline(buffer
, "ivec%d(round(%s)));\n", mask_size
, src0_param
.param_str
);
2394 shader_addline(buffer
, "int(round(%s)));\n", src0_param
.param_str
);
2399 shader_addline(buffer
, "ivec%d(floor(abs(%s) + vec%d(0.5)) * sign(%s)));\n",
2400 mask_size
, src0_param
.param_str
, mask_size
, src0_param
.param_str
);
2402 shader_addline(buffer
, "int(floor(abs(%s) + 0.5) * sign(%s)));\n",
2403 src0_param
.param_str
, src0_param
.param_str
);
2408 shader_addline(buffer
, "%s);\n", src0_param
.param_str
);
2412 /* Process the dot product operators DP3 and DP4 in GLSL (dst = dot(src0, src1)) */
2413 static void shader_glsl_dot(const struct wined3d_shader_instruction
*ins
)
2415 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2416 struct glsl_src_param src0_param
;
2417 struct glsl_src_param src1_param
;
2418 DWORD dst_write_mask
, src_write_mask
;
2419 unsigned int dst_size
= 0;
2421 dst_write_mask
= shader_glsl_append_dst(buffer
, ins
);
2422 dst_size
= shader_glsl_get_write_mask_size(dst_write_mask
);
2424 /* dp3 works on vec3, dp4 on vec4 */
2425 if (ins
->handler_idx
== WINED3DSIH_DP4
)
2427 src_write_mask
= WINED3DSP_WRITEMASK_ALL
;
2429 src_write_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2432 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_write_mask
, &src0_param
);
2433 shader_glsl_add_src_param(ins
, &ins
->src
[1], src_write_mask
, &src1_param
);
2436 shader_addline(buffer
, "vec%d(dot(%s, %s)));\n", dst_size
, src0_param
.param_str
, src1_param
.param_str
);
2438 shader_addline(buffer
, "dot(%s, %s));\n", src0_param
.param_str
, src1_param
.param_str
);
2442 /* Note that this instruction has some restrictions. The destination write mask
2443 * can't contain the w component, and the source swizzles have to be .xyzw */
2444 static void shader_glsl_cross(const struct wined3d_shader_instruction
*ins
)
2446 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2447 struct glsl_src_param src0_param
;
2448 struct glsl_src_param src1_param
;
2451 shader_glsl_get_write_mask(&ins
->dst
[0], dst_mask
);
2452 shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2453 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
2454 shader_glsl_add_src_param(ins
, &ins
->src
[1], src_mask
, &src1_param
);
2455 shader_addline(ins
->ctx
->buffer
, "cross(%s, %s)%s);\n", src0_param
.param_str
, src1_param
.param_str
, dst_mask
);
2458 static void shader_glsl_cut(const struct wined3d_shader_instruction
*ins
)
2460 shader_addline(ins
->ctx
->buffer
, "EndPrimitive();\n");
2463 /* Process the WINED3DSIO_POW instruction in GLSL (dst = |src0|^src1)
2464 * Src0 and src1 are scalars. Note that D3D uses the absolute of src0, while
2465 * GLSL uses the value as-is. */
2466 static void shader_glsl_pow(const struct wined3d_shader_instruction
*ins
)
2468 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2469 struct glsl_src_param src0_param
;
2470 struct glsl_src_param src1_param
;
2471 DWORD dst_write_mask
;
2472 unsigned int dst_size
;
2474 dst_write_mask
= shader_glsl_append_dst(buffer
, ins
);
2475 dst_size
= shader_glsl_get_write_mask_size(dst_write_mask
);
2477 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2478 shader_glsl_add_src_param(ins
, &ins
->src
[1], WINED3DSP_WRITEMASK_0
, &src1_param
);
2482 shader_addline(buffer
, "vec%u(%s == 0.0 ? 1.0 : pow(abs(%s), %s)));\n",
2483 dst_size
, src1_param
.param_str
, src0_param
.param_str
, src1_param
.param_str
);
2487 shader_addline(buffer
, "%s == 0.0 ? 1.0 : pow(abs(%s), %s));\n",
2488 src1_param
.param_str
, src0_param
.param_str
, src1_param
.param_str
);
2492 /* Process the WINED3DSIO_LOG instruction in GLSL (dst = log2(|src0|))
2493 * Src0 is a scalar. Note that D3D uses the absolute of src0, while
2494 * GLSL uses the value as-is. */
2495 static void shader_glsl_log(const struct wined3d_shader_instruction
*ins
)
2497 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2498 struct glsl_src_param src0_param
;
2499 DWORD dst_write_mask
;
2500 unsigned int dst_size
;
2502 dst_write_mask
= shader_glsl_append_dst(buffer
, ins
);
2503 dst_size
= shader_glsl_get_write_mask_size(dst_write_mask
);
2505 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2509 shader_addline(buffer
, "vec%u(log2(abs(%s))));\n",
2510 dst_size
, src0_param
.param_str
);
2514 shader_addline(buffer
, "log2(abs(%s)));\n",
2515 src0_param
.param_str
);
2519 /* Map the opcode 1-to-1 to the GL code (arg->dst = instruction(src0, src1, ...) */
2520 static void shader_glsl_map2gl(const struct wined3d_shader_instruction
*ins
)
2522 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2523 struct glsl_src_param src_param
;
2524 const char *instruction
;
2528 /* Determine the GLSL function to use based on the opcode */
2529 /* TODO: Possibly make this a table for faster lookups */
2530 switch (ins
->handler_idx
)
2532 case WINED3DSIH_MIN
: instruction
= "min"; break;
2533 case WINED3DSIH_MAX
: instruction
= "max"; break;
2534 case WINED3DSIH_ABS
: instruction
= "abs"; break;
2535 case WINED3DSIH_FRC
: instruction
= "fract"; break;
2536 case WINED3DSIH_EXP
: instruction
= "exp2"; break;
2537 case WINED3DSIH_DSX
: instruction
= "dFdx"; break;
2538 case WINED3DSIH_DSY
: instruction
= "ycorrection.y * dFdy"; break;
2539 case WINED3DSIH_ROUND_NI
: instruction
= "floor"; break;
2540 default: instruction
= "";
2541 FIXME("Opcode %#x not yet handled in GLSL\n", ins
->handler_idx
);
2545 write_mask
= shader_glsl_append_dst(buffer
, ins
);
2547 shader_addline(buffer
, "%s(", instruction
);
2551 shader_glsl_add_src_param(ins
, &ins
->src
[0], write_mask
, &src_param
);
2552 shader_addline(buffer
, "%s", src_param
.param_str
);
2553 for (i
= 1; i
< ins
->src_count
; ++i
)
2555 shader_glsl_add_src_param(ins
, &ins
->src
[i
], write_mask
, &src_param
);
2556 shader_addline(buffer
, ", %s", src_param
.param_str
);
2560 shader_addline(buffer
, "));\n");
2563 static void shader_glsl_nop(const struct wined3d_shader_instruction
*ins
) {}
2565 static void shader_glsl_nrm(const struct wined3d_shader_instruction
*ins
)
2567 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2568 struct glsl_src_param src_param
;
2569 unsigned int mask_size
;
2573 write_mask
= shader_glsl_get_write_mask(ins
->dst
, dst_mask
);
2574 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
2575 shader_glsl_add_src_param(ins
, &ins
->src
[0], write_mask
, &src_param
);
2577 shader_addline(buffer
, "tmp0.x = dot(%s, %s);\n",
2578 src_param
.param_str
, src_param
.param_str
);
2579 shader_glsl_append_dst(buffer
, ins
);
2583 shader_addline(buffer
, "tmp0.x == 0.0 ? vec%u(0.0) : (%s * inversesqrt(tmp0.x)));\n",
2584 mask_size
, src_param
.param_str
);
2588 shader_addline(buffer
, "tmp0.x == 0.0 ? 0.0 : (%s * inversesqrt(tmp0.x)));\n",
2589 src_param
.param_str
);
2593 /** Process the WINED3DSIO_EXPP instruction in GLSL:
2594 * For shader model 1.x, do the following (and honor the writemask, so use a temporary variable):
2595 * dst.x = 2^(floor(src))
2596 * dst.y = src - floor(src)
2597 * dst.z = 2^src (partial precision is allowed, but optional)
2599 * For 2.0 shaders, just do this (honoring writemask and swizzle):
2600 * dst = 2^src; (partial precision is allowed, but optional)
2602 static void shader_glsl_expp(const struct wined3d_shader_instruction
*ins
)
2604 struct glsl_src_param src_param
;
2606 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
, &src_param
);
2608 if (ins
->ctx
->reg_maps
->shader_version
.major
< 2)
2612 shader_addline(ins
->ctx
->buffer
, "tmp0.x = exp2(floor(%s));\n", src_param
.param_str
);
2613 shader_addline(ins
->ctx
->buffer
, "tmp0.y = %s - floor(%s);\n", src_param
.param_str
, src_param
.param_str
);
2614 shader_addline(ins
->ctx
->buffer
, "tmp0.z = exp2(%s);\n", src_param
.param_str
);
2615 shader_addline(ins
->ctx
->buffer
, "tmp0.w = 1.0;\n");
2617 shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2618 shader_glsl_get_write_mask(&ins
->dst
[0], dst_mask
);
2619 shader_addline(ins
->ctx
->buffer
, "tmp0%s);\n", dst_mask
);
2622 unsigned int mask_size
;
2624 write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2625 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
2627 if (mask_size
> 1) {
2628 shader_addline(ins
->ctx
->buffer
, "vec%d(exp2(%s)));\n", mask_size
, src_param
.param_str
);
2630 shader_addline(ins
->ctx
->buffer
, "exp2(%s));\n", src_param
.param_str
);
2635 static void shader_glsl_to_int(const struct wined3d_shader_instruction
*ins
)
2637 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2638 struct glsl_src_param src_param
;
2639 unsigned int mask_size
;
2642 write_mask
= shader_glsl_append_dst(buffer
, ins
);
2643 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
2644 shader_glsl_add_src_param(ins
, &ins
->src
[0], write_mask
, &src_param
);
2647 shader_addline(buffer
, "ivec%u(%s));\n", mask_size
, src_param
.param_str
);
2649 shader_addline(buffer
, "int(%s));\n", src_param
.param_str
);
2652 static void shader_glsl_to_float(const struct wined3d_shader_instruction
*ins
)
2654 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2655 struct glsl_src_param src_param
;
2656 unsigned int mask_size
;
2659 write_mask
= shader_glsl_append_dst(buffer
, ins
);
2660 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
2661 shader_glsl_add_src_param(ins
, &ins
->src
[0], write_mask
, &src_param
);
2664 shader_addline(buffer
, "vec%u(%s));\n", mask_size
, src_param
.param_str
);
2666 shader_addline(buffer
, "float(%s));\n", src_param
.param_str
);
2669 /** Process the RCP (reciprocal or inverse) opcode in GLSL (dst = 1 / src) */
2670 static void shader_glsl_rcp(const struct wined3d_shader_instruction
*ins
)
2672 struct glsl_src_param src_param
;
2674 unsigned int mask_size
;
2676 write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2677 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
2678 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_3
, &src_param
);
2682 shader_addline(ins
->ctx
->buffer
, "vec%u(1.0 / %s));\n",
2683 mask_size
, src_param
.param_str
);
2687 shader_addline(ins
->ctx
->buffer
, "1.0 / %s);\n",
2688 src_param
.param_str
);
2692 static void shader_glsl_rsq(const struct wined3d_shader_instruction
*ins
)
2694 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2695 struct glsl_src_param src_param
;
2697 unsigned int mask_size
;
2699 write_mask
= shader_glsl_append_dst(buffer
, ins
);
2700 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
2702 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_3
, &src_param
);
2706 shader_addline(buffer
, "vec%u(inversesqrt(abs(%s))));\n",
2707 mask_size
, src_param
.param_str
);
2711 shader_addline(buffer
, "inversesqrt(abs(%s)));\n",
2712 src_param
.param_str
);
2716 /** Process signed comparison opcodes in GLSL. */
2717 static void shader_glsl_compare(const struct wined3d_shader_instruction
*ins
)
2719 struct glsl_src_param src0_param
;
2720 struct glsl_src_param src1_param
;
2722 unsigned int mask_size
;
2724 write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2725 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
2726 shader_glsl_add_src_param(ins
, &ins
->src
[0], write_mask
, &src0_param
);
2727 shader_glsl_add_src_param(ins
, &ins
->src
[1], write_mask
, &src1_param
);
2729 if (mask_size
> 1) {
2730 const char *compare
;
2732 switch(ins
->handler_idx
)
2734 case WINED3DSIH_SLT
: compare
= "lessThan"; break;
2735 case WINED3DSIH_SGE
: compare
= "greaterThanEqual"; break;
2736 default: compare
= "";
2737 FIXME("Can't handle opcode %#x\n", ins
->handler_idx
);
2740 shader_addline(ins
->ctx
->buffer
, "vec%d(%s(%s, %s)));\n", mask_size
, compare
,
2741 src0_param
.param_str
, src1_param
.param_str
);
2743 switch(ins
->handler_idx
)
2745 case WINED3DSIH_SLT
:
2746 /* Step(src0, src1) is not suitable here because if src0 == src1 SLT is supposed,
2747 * to return 0.0 but step returns 1.0 because step is not < x
2748 * An alternative is a bvec compare padded with an unused second component.
2749 * step(src1 * -1.0, src0 * -1.0) is not an option because it suffers from the same
2750 * issue. Playing with not() is not possible either because not() does not accept
2753 shader_addline(ins
->ctx
->buffer
, "(%s < %s) ? 1.0 : 0.0);\n",
2754 src0_param
.param_str
, src1_param
.param_str
);
2756 case WINED3DSIH_SGE
:
2757 /* Here we can use the step() function and safe a conditional */
2758 shader_addline(ins
->ctx
->buffer
, "step(%s, %s));\n", src1_param
.param_str
, src0_param
.param_str
);
2761 FIXME("Can't handle opcode %#x\n", ins
->handler_idx
);
2767 static void shader_glsl_conditional_move(const struct wined3d_shader_instruction
*ins
)
2769 const char *condition_prefix
, *condition_suffix
;
2770 struct wined3d_shader_dst_param dst
;
2771 struct glsl_src_param src0_param
;
2772 struct glsl_src_param src1_param
;
2773 struct glsl_src_param src2_param
;
2774 BOOL temp_destination
= FALSE
;
2775 DWORD cmp_channel
= 0;
2780 switch (ins
->handler_idx
)
2782 case WINED3DSIH_CMP
:
2783 condition_prefix
= "";
2784 condition_suffix
= " >= 0.0";
2787 case WINED3DSIH_CND
:
2788 condition_prefix
= "";
2789 condition_suffix
= " > 0.5";
2792 case WINED3DSIH_MOVC
:
2793 condition_prefix
= "bool(";
2794 condition_suffix
= ")";
2798 FIXME("Unhandled instruction %#x.\n", ins
->handler_idx
);
2799 condition_prefix
= "<unhandled prefix>";
2800 condition_suffix
= "<unhandled suffix>";
2804 if (shader_is_scalar(&ins
->src
[0].reg
))
2806 write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2807 shader_glsl_add_src_param(ins
, &ins
->src
[0], write_mask
, &src0_param
);
2808 shader_glsl_add_src_param(ins
, &ins
->src
[1], write_mask
, &src1_param
);
2809 shader_glsl_add_src_param(ins
, &ins
->src
[2], write_mask
, &src2_param
);
2811 shader_addline(ins
->ctx
->buffer
, "%s%s%s ? %s : %s);\n",
2812 condition_prefix
, src0_param
.param_str
, condition_suffix
,
2813 src1_param
.param_str
, src2_param
.param_str
);
2819 /* Splitting the instruction up in multiple lines imposes a problem:
2820 * The first lines may overwrite source parameters of the following lines.
2821 * Deal with that by using a temporary destination register if needed. */
2822 if ((ins
->src
[0].reg
.idx
[0].offset
== dst
.reg
.idx
[0].offset
2823 && ins
->src
[0].reg
.type
== dst
.reg
.type
)
2824 || (ins
->src
[1].reg
.idx
[0].offset
== dst
.reg
.idx
[0].offset
2825 && ins
->src
[1].reg
.type
== dst
.reg
.type
)
2826 || (ins
->src
[2].reg
.idx
[0].offset
== dst
.reg
.idx
[0].offset
2827 && ins
->src
[2].reg
.type
== dst
.reg
.type
))
2828 temp_destination
= TRUE
;
2830 /* Cycle through all source0 channels. */
2831 for (i
= 0; i
< 4; ++i
)
2834 /* Find the destination channels which use the current source0 channel. */
2835 for (j
= 0; j
< 4; ++j
)
2837 if (((ins
->src
[0].swizzle
>> (2 * j
)) & 0x3) == i
)
2839 write_mask
|= WINED3DSP_WRITEMASK_0
<< j
;
2840 cmp_channel
= WINED3DSP_WRITEMASK_0
<< j
;
2843 dst
.write_mask
= ins
->dst
[0].write_mask
& write_mask
;
2845 if (temp_destination
)
2847 if (!(write_mask
= shader_glsl_get_write_mask(&dst
, mask_char
)))
2849 shader_addline(ins
->ctx
->buffer
, "tmp0%s = (", mask_char
);
2851 else if (!(write_mask
= shader_glsl_append_dst_ext(ins
->ctx
->buffer
, ins
, &dst
)))
2854 shader_glsl_add_src_param(ins
, &ins
->src
[0], cmp_channel
, &src0_param
);
2855 shader_glsl_add_src_param(ins
, &ins
->src
[1], write_mask
, &src1_param
);
2856 shader_glsl_add_src_param(ins
, &ins
->src
[2], write_mask
, &src2_param
);
2858 shader_addline(ins
->ctx
->buffer
, "%s%s%s ? %s : %s);\n",
2859 condition_prefix
, src0_param
.param_str
, condition_suffix
,
2860 src1_param
.param_str
, src2_param
.param_str
);
2863 if (temp_destination
)
2865 shader_glsl_get_write_mask(&ins
->dst
[0], mask_char
);
2866 shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2867 shader_addline(ins
->ctx
->buffer
, "tmp0%s);\n", mask_char
);
2871 /** Process the CND opcode in GLSL (dst = (src0 > 0.5) ? src1 : src2) */
2872 /* For ps 1.1-1.3, only a single component of src0 is used. For ps 1.4
2873 * the compare is done per component of src0. */
2874 static void shader_glsl_cnd(const struct wined3d_shader_instruction
*ins
)
2876 struct glsl_src_param src0_param
;
2877 struct glsl_src_param src1_param
;
2878 struct glsl_src_param src2_param
;
2880 DWORD shader_version
= WINED3D_SHADER_VERSION(ins
->ctx
->reg_maps
->shader_version
.major
,
2881 ins
->ctx
->reg_maps
->shader_version
.minor
);
2883 if (shader_version
< WINED3D_SHADER_VERSION(1, 4))
2885 write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2886 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2887 shader_glsl_add_src_param(ins
, &ins
->src
[1], write_mask
, &src1_param
);
2888 shader_glsl_add_src_param(ins
, &ins
->src
[2], write_mask
, &src2_param
);
2890 /* Fun: The D3DSI_COISSUE flag changes the semantic of the cnd instruction for < 1.4 shaders */
2893 shader_addline(ins
->ctx
->buffer
, "%s /* COISSUE! */);\n", src1_param
.param_str
);
2895 shader_addline(ins
->ctx
->buffer
, "%s > 0.5 ? %s : %s);\n",
2896 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
2901 shader_glsl_conditional_move(ins
);
2904 /** GLSL code generation for WINED3DSIO_MAD: Multiply the first 2 opcodes, then add the last */
2905 static void shader_glsl_mad(const struct wined3d_shader_instruction
*ins
)
2907 struct glsl_src_param src0_param
;
2908 struct glsl_src_param src1_param
;
2909 struct glsl_src_param src2_param
;
2912 write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2913 shader_glsl_add_src_param(ins
, &ins
->src
[0], write_mask
, &src0_param
);
2914 shader_glsl_add_src_param(ins
, &ins
->src
[1], write_mask
, &src1_param
);
2915 shader_glsl_add_src_param(ins
, &ins
->src
[2], write_mask
, &src2_param
);
2916 shader_addline(ins
->ctx
->buffer
, "(%s * %s) + %s);\n",
2917 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
2920 /* Handles transforming all WINED3DSIO_M?x? opcodes for
2921 Vertex shaders to GLSL codes */
2922 static void shader_glsl_mnxn(const struct wined3d_shader_instruction
*ins
)
2925 int nComponents
= 0;
2926 struct wined3d_shader_dst_param tmp_dst
= {{0}};
2927 struct wined3d_shader_src_param tmp_src
[2] = {{{0}}};
2928 struct wined3d_shader_instruction tmp_ins
;
2930 memset(&tmp_ins
, 0, sizeof(tmp_ins
));
2932 /* Set constants for the temporary argument */
2933 tmp_ins
.ctx
= ins
->ctx
;
2934 tmp_ins
.dst_count
= 1;
2935 tmp_ins
.dst
= &tmp_dst
;
2936 tmp_ins
.src_count
= 2;
2937 tmp_ins
.src
= tmp_src
;
2939 switch(ins
->handler_idx
)
2941 case WINED3DSIH_M4x4
:
2943 tmp_ins
.handler_idx
= WINED3DSIH_DP4
;
2945 case WINED3DSIH_M4x3
:
2947 tmp_ins
.handler_idx
= WINED3DSIH_DP4
;
2949 case WINED3DSIH_M3x4
:
2951 tmp_ins
.handler_idx
= WINED3DSIH_DP3
;
2953 case WINED3DSIH_M3x3
:
2955 tmp_ins
.handler_idx
= WINED3DSIH_DP3
;
2957 case WINED3DSIH_M3x2
:
2959 tmp_ins
.handler_idx
= WINED3DSIH_DP3
;
2965 tmp_dst
= ins
->dst
[0];
2966 tmp_src
[0] = ins
->src
[0];
2967 tmp_src
[1] = ins
->src
[1];
2968 for (i
= 0; i
< nComponents
; ++i
)
2970 tmp_dst
.write_mask
= WINED3DSP_WRITEMASK_0
<< i
;
2971 shader_glsl_dot(&tmp_ins
);
2972 ++tmp_src
[1].reg
.idx
[0].offset
;
2977 The LRP instruction performs a component-wise linear interpolation
2978 between the second and third operands using the first operand as the
2979 blend factor. Equation: (dst = src2 + src0 * (src1 - src2))
2980 This is equivalent to mix(src2, src1, src0);
2982 static void shader_glsl_lrp(const struct wined3d_shader_instruction
*ins
)
2984 struct glsl_src_param src0_param
;
2985 struct glsl_src_param src1_param
;
2986 struct glsl_src_param src2_param
;
2989 write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2991 shader_glsl_add_src_param(ins
, &ins
->src
[0], write_mask
, &src0_param
);
2992 shader_glsl_add_src_param(ins
, &ins
->src
[1], write_mask
, &src1_param
);
2993 shader_glsl_add_src_param(ins
, &ins
->src
[2], write_mask
, &src2_param
);
2995 shader_addline(ins
->ctx
->buffer
, "mix(%s, %s, %s));\n",
2996 src2_param
.param_str
, src1_param
.param_str
, src0_param
.param_str
);
2999 /** Process the WINED3DSIO_LIT instruction in GLSL:
3000 * dst.x = dst.w = 1.0
3001 * dst.y = (src0.x > 0) ? src0.x
3002 * dst.z = (src0.x > 0) ? ((src0.y > 0) ? pow(src0.y, src.w) : 0) : 0
3003 * where src.w is clamped at +- 128
3005 static void shader_glsl_lit(const struct wined3d_shader_instruction
*ins
)
3007 struct glsl_src_param src0_param
;
3008 struct glsl_src_param src1_param
;
3009 struct glsl_src_param src3_param
;
3012 shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
3013 shader_glsl_get_write_mask(&ins
->dst
[0], dst_mask
);
3015 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
3016 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_1
, &src1_param
);
3017 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_3
, &src3_param
);
3019 /* The sdk specifies the instruction like this
3021 * if(src.x > 0.0) dst.y = src.x
3023 * if(src.x > 0.0 && src.y > 0.0) dst.z = pow(src.y, power);
3026 * (where power = src.w clamped between -128 and 128)
3028 * Obviously that has quite a few conditionals in it which we don't like. So the first step is this:
3029 * dst.x = 1.0 ... No further explanation needed
3030 * dst.y = max(src.y, 0.0); ... If x < 0.0, use 0.0, otherwise x. Same as the conditional
3031 * dst.z = x > 0.0 ? pow(max(y, 0.0), p) : 0; ... 0 ^ power is 0, and otherwise we use y anyway
3032 * dst.w = 1.0. ... Nothing fancy.
3034 * So we still have one conditional in there. So do this:
3035 * dst.z = pow(max(0.0, src.y) * step(0.0, src.x), power);
3037 * 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),
3038 * 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.
3039 * if both x and y are > 0, we get pow(y * 1.0, power), as it is supposed to.
3041 * Unfortunately pow(0.0 ^ 0.0) returns NaN on most GPUs, but lit with src.y = 0 and src.w = 0 returns
3042 * a non-NaN value in dst.z. What we return doesn't matter, as long as it is not NaN. Return 0, which is
3043 * what all Windows HW drivers and GL_ARB_vertex_program's LIT do.
3045 shader_addline(ins
->ctx
->buffer
,
3046 "vec4(1.0, max(%s, 0.0), %s == 0.0 ? 0.0 : "
3047 "pow(max(0.0, %s) * step(0.0, %s), clamp(%s, -128.0, 128.0)), 1.0)%s);\n",
3048 src0_param
.param_str
, src3_param
.param_str
, src1_param
.param_str
,
3049 src0_param
.param_str
, src3_param
.param_str
, dst_mask
);
3052 /** Process the WINED3DSIO_DST instruction in GLSL:
3054 * dst.y = src0.x * src0.y
3058 static void shader_glsl_dst(const struct wined3d_shader_instruction
*ins
)
3060 struct glsl_src_param src0y_param
;
3061 struct glsl_src_param src0z_param
;
3062 struct glsl_src_param src1y_param
;
3063 struct glsl_src_param src1w_param
;
3066 shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
3067 shader_glsl_get_write_mask(&ins
->dst
[0], dst_mask
);
3069 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_1
, &src0y_param
);
3070 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_2
, &src0z_param
);
3071 shader_glsl_add_src_param(ins
, &ins
->src
[1], WINED3DSP_WRITEMASK_1
, &src1y_param
);
3072 shader_glsl_add_src_param(ins
, &ins
->src
[1], WINED3DSP_WRITEMASK_3
, &src1w_param
);
3074 shader_addline(ins
->ctx
->buffer
, "vec4(1.0, %s * %s, %s, %s))%s;\n",
3075 src0y_param
.param_str
, src1y_param
.param_str
, src0z_param
.param_str
, src1w_param
.param_str
, dst_mask
);
3078 /** Process the WINED3DSIO_SINCOS instruction in GLSL:
3079 * VS 2.0 requires that specific cosine and sine constants be passed to this instruction so the hardware
3080 * can handle it. But, these functions are built-in for GLSL, so we can just ignore the last 2 params.
3082 * dst.x = cos(src0.?)
3083 * dst.y = sin(src0.?)
3087 static void shader_glsl_sincos(const struct wined3d_shader_instruction
*ins
)
3089 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
3090 struct glsl_src_param src0_param
;
3093 if (ins
->ctx
->reg_maps
->shader_version
.major
< 4)
3095 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
3097 write_mask
= shader_glsl_append_dst(buffer
, ins
);
3100 case WINED3DSP_WRITEMASK_0
:
3101 shader_addline(buffer
, "cos(%s));\n", src0_param
.param_str
);
3104 case WINED3DSP_WRITEMASK_1
:
3105 shader_addline(buffer
, "sin(%s));\n", src0_param
.param_str
);
3108 case (WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
):
3109 shader_addline(buffer
, "vec2(cos(%s), sin(%s)));\n",
3110 src0_param
.param_str
, src0_param
.param_str
);
3114 ERR("Write mask should be .x, .y or .xy\n");
3121 if (ins
->dst
[0].reg
.type
!= WINED3DSPR_NULL
)
3124 if (ins
->dst
[1].reg
.type
!= WINED3DSPR_NULL
)
3128 write_mask
= shader_glsl_get_write_mask(&ins
->dst
[0], dst_mask
);
3129 shader_glsl_add_src_param(ins
, &ins
->src
[0], write_mask
, &src0_param
);
3130 shader_addline(buffer
, "tmp0%s = sin(%s);\n", dst_mask
, src0_param
.param_str
);
3132 write_mask
= shader_glsl_append_dst_ext(buffer
, ins
, &ins
->dst
[1]);
3133 shader_glsl_add_src_param(ins
, &ins
->src
[0], write_mask
, &src0_param
);
3134 shader_addline(buffer
, "cos(%s));\n", src0_param
.param_str
);
3136 shader_glsl_append_dst_ext(buffer
, ins
, &ins
->dst
[0]);
3137 shader_addline(buffer
, "tmp0%s);\n", dst_mask
);
3141 write_mask
= shader_glsl_append_dst_ext(buffer
, ins
, &ins
->dst
[0]);
3142 shader_glsl_add_src_param(ins
, &ins
->src
[0], write_mask
, &src0_param
);
3143 shader_addline(buffer
, "sin(%s));\n", src0_param
.param_str
);
3146 else if (ins
->dst
[1].reg
.type
!= WINED3DSPR_NULL
)
3148 write_mask
= shader_glsl_append_dst_ext(buffer
, ins
, &ins
->dst
[1]);
3149 shader_glsl_add_src_param(ins
, &ins
->src
[0], write_mask
, &src0_param
);
3150 shader_addline(buffer
, "cos(%s));\n", src0_param
.param_str
);
3154 /* sgn in vs_2_0 has 2 extra parameters(registers for temporary storage) which we don't use
3155 * here. But those extra parameters require a dedicated function for sgn, since map2gl would
3156 * generate invalid code
3158 static void shader_glsl_sgn(const struct wined3d_shader_instruction
*ins
)
3160 struct glsl_src_param src0_param
;
3163 write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
3164 shader_glsl_add_src_param(ins
, &ins
->src
[0], write_mask
, &src0_param
);
3166 shader_addline(ins
->ctx
->buffer
, "sign(%s));\n", src0_param
.param_str
);
3169 /** Process the WINED3DSIO_LOOP instruction in GLSL:
3170 * Start a for() loop where src1.y is the initial value of aL,
3171 * increment aL by src1.z for a total of src1.x iterations.
3172 * Need to use a temporary variable for this operation.
3174 /* FIXME: I don't think nested loops will work correctly this way. */
3175 static void shader_glsl_loop(const struct wined3d_shader_instruction
*ins
)
3177 struct wined3d_shader_loop_state
*loop_state
= ins
->ctx
->loop_state
;
3178 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
3179 const struct wined3d_shader
*shader
= ins
->ctx
->shader
;
3180 const struct wined3d_shader_lconst
*constant
;
3181 struct glsl_src_param src1_param
;
3182 const DWORD
*control_values
= NULL
;
3184 if (ins
->ctx
->reg_maps
->shader_version
.major
< 4)
3186 shader_glsl_add_src_param(ins
, &ins
->src
[1], WINED3DSP_WRITEMASK_ALL
, &src1_param
);
3188 /* Try to hardcode the loop control parameters if possible. Direct3D 9
3189 * class hardware doesn't support real varying indexing, but Microsoft
3190 * designed this feature for Shader model 2.x+. If the loop control is
3191 * known at compile time, the GLSL compiler can unroll the loop, and
3192 * replace indirect addressing with direct addressing. */
3193 if (ins
->src
[1].reg
.type
== WINED3DSPR_CONSTINT
)
3195 LIST_FOR_EACH_ENTRY(constant
, &shader
->constantsI
, struct wined3d_shader_lconst
, entry
)
3197 if (constant
->idx
== ins
->src
[1].reg
.idx
[0].offset
)
3199 control_values
= constant
->value
;
3207 struct wined3d_shader_loop_control loop_control
;
3208 loop_control
.count
= control_values
[0];
3209 loop_control
.start
= control_values
[1];
3210 loop_control
.step
= (int)control_values
[2];
3212 if (loop_control
.step
> 0)
3214 shader_addline(buffer
, "for (aL%u = %u; aL%u < (%u * %d + %u); aL%u += %d)\n{\n",
3215 loop_state
->current_depth
, loop_control
.start
,
3216 loop_state
->current_depth
, loop_control
.count
, loop_control
.step
, loop_control
.start
,
3217 loop_state
->current_depth
, loop_control
.step
);
3219 else if (loop_control
.step
< 0)
3221 shader_addline(buffer
, "for (aL%u = %u; aL%u > (%u * %d + %u); aL%u += %d)\n{\n",
3222 loop_state
->current_depth
, loop_control
.start
,
3223 loop_state
->current_depth
, loop_control
.count
, loop_control
.step
, loop_control
.start
,
3224 loop_state
->current_depth
, loop_control
.step
);
3228 shader_addline(buffer
, "for (aL%u = %u, tmpInt%u = 0; tmpInt%u < %u; tmpInt%u++)\n{\n",
3229 loop_state
->current_depth
, loop_control
.start
, loop_state
->current_depth
,
3230 loop_state
->current_depth
, loop_control
.count
,
3231 loop_state
->current_depth
);
3236 shader_addline(buffer
, "for (tmpInt%u = 0, aL%u = %s.y; tmpInt%u < %s.x; tmpInt%u++, aL%u += %s.z)\n{\n",
3237 loop_state
->current_depth
, loop_state
->current_reg
,
3238 src1_param
.reg_name
, loop_state
->current_depth
, src1_param
.reg_name
,
3239 loop_state
->current_depth
, loop_state
->current_reg
, src1_param
.reg_name
);
3242 ++loop_state
->current_reg
;
3246 shader_addline(buffer
, "for (;;)\n{\n");
3249 ++loop_state
->current_depth
;
3252 static void shader_glsl_end(const struct wined3d_shader_instruction
*ins
)
3254 struct wined3d_shader_loop_state
*loop_state
= ins
->ctx
->loop_state
;
3256 shader_addline(ins
->ctx
->buffer
, "}\n");
3258 if (ins
->handler_idx
== WINED3DSIH_ENDLOOP
)
3260 --loop_state
->current_depth
;
3261 --loop_state
->current_reg
;
3264 if (ins
->handler_idx
== WINED3DSIH_ENDREP
)
3266 --loop_state
->current_depth
;
3270 static void shader_glsl_rep(const struct wined3d_shader_instruction
*ins
)
3272 const struct wined3d_shader
*shader
= ins
->ctx
->shader
;
3273 struct wined3d_shader_loop_state
*loop_state
= ins
->ctx
->loop_state
;
3274 const struct wined3d_shader_lconst
*constant
;
3275 struct glsl_src_param src0_param
;
3276 const DWORD
*control_values
= NULL
;
3278 /* Try to hardcode local values to help the GLSL compiler to unroll and optimize the loop */
3279 if (ins
->src
[0].reg
.type
== WINED3DSPR_CONSTINT
)
3281 LIST_FOR_EACH_ENTRY(constant
, &shader
->constantsI
, struct wined3d_shader_lconst
, entry
)
3283 if (constant
->idx
== ins
->src
[0].reg
.idx
[0].offset
)
3285 control_values
= constant
->value
;
3293 shader_addline(ins
->ctx
->buffer
, "for (tmpInt%d = 0; tmpInt%d < %d; tmpInt%d++) {\n",
3294 loop_state
->current_depth
, loop_state
->current_depth
,
3295 control_values
[0], loop_state
->current_depth
);
3299 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
3300 shader_addline(ins
->ctx
->buffer
, "for (tmpInt%d = 0; tmpInt%d < %s; tmpInt%d++) {\n",
3301 loop_state
->current_depth
, loop_state
->current_depth
,
3302 src0_param
.param_str
, loop_state
->current_depth
);
3305 ++loop_state
->current_depth
;
3308 static void shader_glsl_if(const struct wined3d_shader_instruction
*ins
)
3310 struct glsl_src_param src0_param
;
3312 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
3313 shader_addline(ins
->ctx
->buffer
, "if (%s) {\n", src0_param
.param_str
);
3316 static void shader_glsl_ifc(const struct wined3d_shader_instruction
*ins
)
3318 struct glsl_src_param src0_param
;
3319 struct glsl_src_param src1_param
;
3321 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
3322 shader_glsl_add_src_param(ins
, &ins
->src
[1], WINED3DSP_WRITEMASK_0
, &src1_param
);
3324 shader_addline(ins
->ctx
->buffer
, "if (%s %s %s) {\n",
3325 src0_param
.param_str
, shader_glsl_get_rel_op(ins
->flags
), src1_param
.param_str
);
3328 static void shader_glsl_else(const struct wined3d_shader_instruction
*ins
)
3330 shader_addline(ins
->ctx
->buffer
, "} else {\n");
3333 static void shader_glsl_emit(const struct wined3d_shader_instruction
*ins
)
3335 shader_addline(ins
->ctx
->buffer
, "EmitVertex();\n");
3338 static void shader_glsl_break(const struct wined3d_shader_instruction
*ins
)
3340 shader_addline(ins
->ctx
->buffer
, "break;\n");
3343 /* FIXME: According to MSDN the compare is done per component. */
3344 static void shader_glsl_breakc(const struct wined3d_shader_instruction
*ins
)
3346 struct glsl_src_param src0_param
;
3347 struct glsl_src_param src1_param
;
3349 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
3350 shader_glsl_add_src_param(ins
, &ins
->src
[1], WINED3DSP_WRITEMASK_0
, &src1_param
);
3352 shader_addline(ins
->ctx
->buffer
, "if (%s %s %s) break;\n",
3353 src0_param
.param_str
, shader_glsl_get_rel_op(ins
->flags
), src1_param
.param_str
);
3356 static void shader_glsl_breakp(const struct wined3d_shader_instruction
*ins
)
3358 struct glsl_src_param src_param
;
3360 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
, &src_param
);
3361 shader_addline(ins
->ctx
->buffer
, "if (bool(%s)) break;\n", src_param
.param_str
);
3364 static void shader_glsl_label(const struct wined3d_shader_instruction
*ins
)
3366 shader_addline(ins
->ctx
->buffer
, "}\n");
3367 shader_addline(ins
->ctx
->buffer
, "void subroutine%u()\n{\n", ins
->src
[0].reg
.idx
[0].offset
);
3370 static void shader_glsl_call(const struct wined3d_shader_instruction
*ins
)
3372 shader_addline(ins
->ctx
->buffer
, "subroutine%u();\n", ins
->src
[0].reg
.idx
[0].offset
);
3375 static void shader_glsl_callnz(const struct wined3d_shader_instruction
*ins
)
3377 struct glsl_src_param src1_param
;
3379 shader_glsl_add_src_param(ins
, &ins
->src
[1], WINED3DSP_WRITEMASK_0
, &src1_param
);
3380 shader_addline(ins
->ctx
->buffer
, "if (%s) subroutine%u();\n",
3381 src1_param
.param_str
, ins
->src
[0].reg
.idx
[0].offset
);
3384 static void shader_glsl_ret(const struct wined3d_shader_instruction
*ins
)
3386 /* No-op. The closing } is written when a new function is started, and at the end of the shader. This
3387 * function only suppresses the unhandled instruction warning
3391 /*********************************************
3392 * Pixel Shader Specific Code begins here
3393 ********************************************/
3394 static void shader_glsl_tex(const struct wined3d_shader_instruction
*ins
)
3396 const struct wined3d_shader
*shader
= ins
->ctx
->shader
;
3397 struct wined3d_device
*device
= shader
->device
;
3398 DWORD shader_version
= WINED3D_SHADER_VERSION(ins
->ctx
->reg_maps
->shader_version
.major
,
3399 ins
->ctx
->reg_maps
->shader_version
.minor
);
3400 struct glsl_sample_function sample_function
;
3401 const struct wined3d_texture
*texture
;
3402 DWORD sample_flags
= 0;
3404 DWORD mask
= 0, swizzle
;
3406 /* 1.0-1.4: Use destination register as sampler source.
3407 * 2.0+: Use provided sampler source. */
3408 if (shader_version
< WINED3D_SHADER_VERSION(2,0))
3409 sampler_idx
= ins
->dst
[0].reg
.idx
[0].offset
;
3411 sampler_idx
= ins
->src
[1].reg
.idx
[0].offset
;
3412 texture
= device
->stateBlock
->state
.textures
[sampler_idx
];
3414 if (shader_version
< WINED3D_SHADER_VERSION(1,4))
3416 const struct shader_glsl_ctx_priv
*priv
= ins
->ctx
->backend_data
;
3417 DWORD flags
= (priv
->cur_ps_args
->tex_transform
>> sampler_idx
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
)
3418 & WINED3D_PSARGS_TEXTRANSFORM_MASK
;
3419 enum wined3d_sampler_texture_type sampler_type
= ins
->ctx
->reg_maps
->sampler_type
[sampler_idx
];
3421 /* Projected cube textures don't make a lot of sense, the resulting coordinates stay the same. */
3422 if (flags
& WINED3D_PSARGS_PROJECTED
&& sampler_type
!= WINED3DSTT_CUBE
)
3424 sample_flags
|= WINED3D_GLSL_SAMPLE_PROJECTED
;
3425 switch (flags
& ~WINED3D_PSARGS_PROJECTED
)
3427 case WINED3D_TTFF_COUNT1
:
3428 FIXME("WINED3D_TTFF_PROJECTED with WINED3D_TTFF_COUNT1?\n");
3430 case WINED3D_TTFF_COUNT2
:
3431 mask
= WINED3DSP_WRITEMASK_1
;
3433 case WINED3D_TTFF_COUNT3
:
3434 mask
= WINED3DSP_WRITEMASK_2
;
3436 case WINED3D_TTFF_COUNT4
:
3437 case WINED3D_TTFF_DISABLE
:
3438 mask
= WINED3DSP_WRITEMASK_3
;
3443 else if (shader_version
< WINED3D_SHADER_VERSION(2,0))
3445 enum wined3d_shader_src_modifier src_mod
= ins
->src
[0].modifiers
;
3447 if (src_mod
== WINED3DSPSM_DZ
) {
3448 sample_flags
|= WINED3D_GLSL_SAMPLE_PROJECTED
;
3449 mask
= WINED3DSP_WRITEMASK_2
;
3450 } else if (src_mod
== WINED3DSPSM_DW
) {
3451 sample_flags
|= WINED3D_GLSL_SAMPLE_PROJECTED
;
3452 mask
= WINED3DSP_WRITEMASK_3
;
3455 if (ins
->flags
& WINED3DSI_TEXLD_PROJECT
)
3457 /* ps 2.0 texldp instruction always divides by the fourth component. */
3458 sample_flags
|= WINED3D_GLSL_SAMPLE_PROJECTED
;
3459 mask
= WINED3DSP_WRITEMASK_3
;
3463 if (texture
&& texture
->target
== GL_TEXTURE_RECTANGLE_ARB
)
3464 sample_flags
|= WINED3D_GLSL_SAMPLE_RECT
;
3466 shader_glsl_get_sample_function(ins
->ctx
, sampler_idx
, sample_flags
, &sample_function
);
3467 mask
|= sample_function
.coord_mask
;
3469 if (shader_version
< WINED3D_SHADER_VERSION(2,0)) swizzle
= WINED3DSP_NOSWIZZLE
;
3470 else swizzle
= ins
->src
[1].swizzle
;
3472 /* 1.0-1.3: Use destination register as coordinate source.
3473 1.4+: Use provided coordinate source register. */
3474 if (shader_version
< WINED3D_SHADER_VERSION(1,4))
3477 shader_glsl_write_mask_to_str(mask
, coord_mask
);
3478 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, swizzle
, NULL
, NULL
, NULL
,
3479 "T%u%s", sampler_idx
, coord_mask
);
3483 struct glsl_src_param coord_param
;
3484 shader_glsl_add_src_param(ins
, &ins
->src
[0], mask
, &coord_param
);
3485 if (ins
->flags
& WINED3DSI_TEXLD_BIAS
)
3487 struct glsl_src_param bias
;
3488 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_3
, &bias
);
3489 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, swizzle
, NULL
, NULL
, bias
.param_str
,
3490 "%s", coord_param
.param_str
);
3492 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, swizzle
, NULL
, NULL
, NULL
,
3493 "%s", coord_param
.param_str
);
3498 static void shader_glsl_texldd(const struct wined3d_shader_instruction
*ins
)
3500 const struct wined3d_shader
*shader
= ins
->ctx
->shader
;
3501 struct wined3d_device
*device
= shader
->device
;
3502 const struct wined3d_gl_info
*gl_info
= ins
->ctx
->gl_info
;
3503 struct glsl_src_param coord_param
, dx_param
, dy_param
;
3504 DWORD sample_flags
= WINED3D_GLSL_SAMPLE_GRAD
;
3505 struct glsl_sample_function sample_function
;
3507 DWORD swizzle
= ins
->src
[1].swizzle
;
3508 const struct wined3d_texture
*texture
;
3510 if (!gl_info
->supported
[ARB_SHADER_TEXTURE_LOD
] && !gl_info
->supported
[EXT_GPU_SHADER4
])
3512 FIXME("texldd used, but not supported by hardware. Falling back to regular tex\n");
3513 shader_glsl_tex(ins
);
3517 sampler_idx
= ins
->src
[1].reg
.idx
[0].offset
;
3518 texture
= device
->stateBlock
->state
.textures
[sampler_idx
];
3519 if (texture
&& texture
->target
== GL_TEXTURE_RECTANGLE_ARB
)
3520 sample_flags
|= WINED3D_GLSL_SAMPLE_RECT
;
3522 shader_glsl_get_sample_function(ins
->ctx
, sampler_idx
, sample_flags
, &sample_function
);
3523 shader_glsl_add_src_param(ins
, &ins
->src
[0], sample_function
.coord_mask
, &coord_param
);
3524 shader_glsl_add_src_param(ins
, &ins
->src
[2], sample_function
.coord_mask
, &dx_param
);
3525 shader_glsl_add_src_param(ins
, &ins
->src
[3], sample_function
.coord_mask
, &dy_param
);
3527 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, swizzle
, dx_param
.param_str
, dy_param
.param_str
, NULL
,
3528 "%s", coord_param
.param_str
);
3531 static void shader_glsl_texldl(const struct wined3d_shader_instruction
*ins
)
3533 const struct wined3d_shader
*shader
= ins
->ctx
->shader
;
3534 struct wined3d_device
*device
= shader
->device
;
3535 const struct wined3d_gl_info
*gl_info
= ins
->ctx
->gl_info
;
3536 struct glsl_src_param coord_param
, lod_param
;
3537 DWORD sample_flags
= WINED3D_GLSL_SAMPLE_LOD
;
3538 struct glsl_sample_function sample_function
;
3540 DWORD swizzle
= ins
->src
[1].swizzle
;
3541 const struct wined3d_texture
*texture
;
3543 sampler_idx
= ins
->src
[1].reg
.idx
[0].offset
;
3544 texture
= device
->stateBlock
->state
.textures
[sampler_idx
];
3545 if (texture
&& texture
->target
== GL_TEXTURE_RECTANGLE_ARB
)
3546 sample_flags
|= WINED3D_GLSL_SAMPLE_RECT
;
3548 shader_glsl_get_sample_function(ins
->ctx
, sampler_idx
, sample_flags
, &sample_function
);
3549 shader_glsl_add_src_param(ins
, &ins
->src
[0], sample_function
.coord_mask
, &coord_param
);
3551 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_3
, &lod_param
);
3553 if (!gl_info
->supported
[ARB_SHADER_TEXTURE_LOD
] && !gl_info
->supported
[EXT_GPU_SHADER4
]
3554 && ins
->ctx
->reg_maps
->shader_version
.type
== WINED3D_SHADER_TYPE_PIXEL
)
3556 /* Plain GLSL only supports Lod sampling functions in vertex shaders.
3557 * However, the NVIDIA drivers allow them in fragment shaders as well,
3558 * even without the appropriate extension. */
3559 WARN("Using %s in fragment shader.\n", sample_function
.name
);
3561 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, swizzle
, NULL
, NULL
, lod_param
.param_str
,
3562 "%s", coord_param
.param_str
);
3565 static void shader_glsl_texcoord(const struct wined3d_shader_instruction
*ins
)
3567 /* FIXME: Make this work for more than just 2D textures */
3568 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
3569 DWORD write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
3571 if (!(ins
->ctx
->reg_maps
->shader_version
.major
== 1 && ins
->ctx
->reg_maps
->shader_version
.minor
== 4))
3575 shader_glsl_get_write_mask(&ins
->dst
[0], dst_mask
);
3576 shader_addline(buffer
, "clamp(gl_TexCoord[%u], 0.0, 1.0)%s);\n",
3577 ins
->dst
[0].reg
.idx
[0].offset
, dst_mask
);
3581 enum wined3d_shader_src_modifier src_mod
= ins
->src
[0].modifiers
;
3582 DWORD reg
= ins
->src
[0].reg
.idx
[0].offset
;
3583 char dst_swizzle
[6];
3585 shader_glsl_get_swizzle(&ins
->src
[0], FALSE
, write_mask
, dst_swizzle
);
3587 if (src_mod
== WINED3DSPSM_DZ
)
3589 unsigned int mask_size
= shader_glsl_get_write_mask_size(write_mask
);
3590 struct glsl_src_param div_param
;
3592 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_2
, &div_param
);
3594 if (mask_size
> 1) {
3595 shader_addline(buffer
, "gl_TexCoord[%u]%s / vec%d(%s));\n", reg
, dst_swizzle
, mask_size
, div_param
.param_str
);
3597 shader_addline(buffer
, "gl_TexCoord[%u]%s / %s);\n", reg
, dst_swizzle
, div_param
.param_str
);
3600 else if (src_mod
== WINED3DSPSM_DW
)
3602 unsigned int mask_size
= shader_glsl_get_write_mask_size(write_mask
);
3603 struct glsl_src_param div_param
;
3605 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_3
, &div_param
);
3607 if (mask_size
> 1) {
3608 shader_addline(buffer
, "gl_TexCoord[%u]%s / vec%d(%s));\n", reg
, dst_swizzle
, mask_size
, div_param
.param_str
);
3610 shader_addline(buffer
, "gl_TexCoord[%u]%s / %s);\n", reg
, dst_swizzle
, div_param
.param_str
);
3613 shader_addline(buffer
, "gl_TexCoord[%u]%s);\n", reg
, dst_swizzle
);
3618 /** Process the WINED3DSIO_TEXDP3TEX instruction in GLSL:
3619 * Take a 3-component dot product of the TexCoord[dstreg] and src,
3620 * then perform a 1D texture lookup from stage dstregnum, place into dst. */
3621 static void shader_glsl_texdp3tex(const struct wined3d_shader_instruction
*ins
)
3623 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
3624 DWORD sampler_idx
= ins
->dst
[0].reg
.idx
[0].offset
;
3625 struct glsl_sample_function sample_function
;
3626 struct glsl_src_param src0_param
;
3629 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
3631 /* Do I have to take care about the projected bit? I don't think so, since the dp3 returns only one
3632 * scalar, and projected sampling would require 4.
3634 * It is a dependent read - not valid with conditional NP2 textures
3636 shader_glsl_get_sample_function(ins
->ctx
, sampler_idx
, 0, &sample_function
);
3637 mask_size
= shader_glsl_get_write_mask_size(sample_function
.coord_mask
);
3642 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, WINED3DSP_NOSWIZZLE
, NULL
, NULL
, NULL
,
3643 "dot(gl_TexCoord[%u].xyz, %s)", sampler_idx
, src0_param
.param_str
);
3647 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, WINED3DSP_NOSWIZZLE
, NULL
, NULL
, NULL
,
3648 "vec2(dot(gl_TexCoord[%u].xyz, %s), 0.0)", sampler_idx
, src0_param
.param_str
);
3652 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, WINED3DSP_NOSWIZZLE
, NULL
, NULL
, NULL
,
3653 "vec3(dot(gl_TexCoord[%u].xyz, %s), 0.0, 0.0)", sampler_idx
, src0_param
.param_str
);
3657 FIXME("Unexpected mask size %u\n", mask_size
);
3662 /** Process the WINED3DSIO_TEXDP3 instruction in GLSL:
3663 * Take a 3-component dot product of the TexCoord[dstreg] and src. */
3664 static void shader_glsl_texdp3(const struct wined3d_shader_instruction
*ins
)
3666 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
3667 DWORD dstreg
= ins
->dst
[0].reg
.idx
[0].offset
;
3668 struct glsl_src_param src0_param
;
3670 unsigned int mask_size
;
3672 dst_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
3673 mask_size
= shader_glsl_get_write_mask_size(dst_mask
);
3674 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
3676 if (mask_size
> 1) {
3677 shader_addline(ins
->ctx
->buffer
, "vec%d(dot(T%u.xyz, %s)));\n", mask_size
, dstreg
, src0_param
.param_str
);
3679 shader_addline(ins
->ctx
->buffer
, "dot(T%u.xyz, %s));\n", dstreg
, src0_param
.param_str
);
3683 /** Process the WINED3DSIO_TEXDEPTH instruction in GLSL:
3684 * Calculate the depth as dst.x / dst.y */
3685 static void shader_glsl_texdepth(const struct wined3d_shader_instruction
*ins
)
3687 struct glsl_dst_param dst_param
;
3689 shader_glsl_add_dst_param(ins
, &ins
->dst
[0], &dst_param
);
3691 /* Tests show that texdepth never returns anything below 0.0, and that r5.y is clamped to 1.0.
3692 * Negative input is accepted, -0.25 / -0.5 returns 0.5. GL should clamp gl_FragDepth to [0;1], but
3693 * this doesn't always work, so clamp the results manually. Whether or not the x value is clamped at 1
3694 * too is irrelevant, since if x = 0, any y value < 1.0 (and > 1.0 is not allowed) results in a result
3697 shader_addline(ins
->ctx
->buffer
, "gl_FragDepth = clamp((%s.x / min(%s.y, 1.0)), 0.0, 1.0);\n",
3698 dst_param
.reg_name
, dst_param
.reg_name
);
3701 /** Process the WINED3DSIO_TEXM3X2DEPTH instruction in GLSL:
3702 * Last row of a 3x2 matrix multiply, use the result to calculate the depth:
3703 * Calculate tmp0.y = TexCoord[dstreg] . src.xyz; (tmp0.x has already been calculated)
3704 * depth = (tmp0.y == 0.0) ? 1.0 : tmp0.x / tmp0.y
3706 static void shader_glsl_texm3x2depth(const struct wined3d_shader_instruction
*ins
)
3708 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
3709 DWORD dstreg
= ins
->dst
[0].reg
.idx
[0].offset
;
3710 struct glsl_src_param src0_param
;
3712 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
3714 shader_addline(ins
->ctx
->buffer
, "tmp0.y = dot(T%u.xyz, %s);\n", dstreg
, src0_param
.param_str
);
3715 shader_addline(ins
->ctx
->buffer
, "gl_FragDepth = (tmp0.y == 0.0) ? 1.0 : clamp(tmp0.x / tmp0.y, 0.0, 1.0);\n");
3718 /** Process the WINED3DSIO_TEXM3X2PAD instruction in GLSL
3719 * Calculate the 1st of a 2-row matrix multiplication. */
3720 static void shader_glsl_texm3x2pad(const struct wined3d_shader_instruction
*ins
)
3722 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
3723 DWORD reg
= ins
->dst
[0].reg
.idx
[0].offset
;
3724 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
3725 struct glsl_src_param src0_param
;
3727 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
3728 shader_addline(buffer
, "tmp0.x = dot(T%u.xyz, %s);\n", reg
, src0_param
.param_str
);
3731 /** Process the WINED3DSIO_TEXM3X3PAD instruction in GLSL
3732 * Calculate the 1st or 2nd row of a 3-row matrix multiplication. */
3733 static void shader_glsl_texm3x3pad(const struct wined3d_shader_instruction
*ins
)
3735 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
3736 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
3737 struct wined3d_shader_tex_mx
*tex_mx
= ins
->ctx
->tex_mx
;
3738 DWORD reg
= ins
->dst
[0].reg
.idx
[0].offset
;
3739 struct glsl_src_param src0_param
;
3741 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
3742 shader_addline(buffer
, "tmp0.%c = dot(T%u.xyz, %s);\n", 'x' + tex_mx
->current_row
, reg
, src0_param
.param_str
);
3743 tex_mx
->texcoord_w
[tex_mx
->current_row
++] = reg
;
3746 static void shader_glsl_texm3x2tex(const struct wined3d_shader_instruction
*ins
)
3748 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
3749 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
3750 struct glsl_sample_function sample_function
;
3751 DWORD reg
= ins
->dst
[0].reg
.idx
[0].offset
;
3752 struct glsl_src_param src0_param
;
3754 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
3755 shader_addline(buffer
, "tmp0.y = dot(T%u.xyz, %s);\n", reg
, src0_param
.param_str
);
3757 shader_glsl_get_sample_function(ins
->ctx
, reg
, 0, &sample_function
);
3759 /* Sample the texture using the calculated coordinates */
3760 shader_glsl_gen_sample_code(ins
, reg
, &sample_function
, WINED3DSP_NOSWIZZLE
, NULL
, NULL
, NULL
, "tmp0.xy");
3763 /** Process the WINED3DSIO_TEXM3X3TEX instruction in GLSL
3764 * Perform the 3rd row of a 3x3 matrix multiply, then sample the texture using the calculated coordinates */
3765 static void shader_glsl_texm3x3tex(const struct wined3d_shader_instruction
*ins
)
3767 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
3768 struct wined3d_shader_tex_mx
*tex_mx
= ins
->ctx
->tex_mx
;
3769 struct glsl_sample_function sample_function
;
3770 DWORD reg
= 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
);
3774 shader_addline(ins
->ctx
->buffer
, "tmp0.z = dot(T%u.xyz, %s);\n", reg
, src0_param
.param_str
);
3776 /* Dependent read, not valid with conditional NP2 */
3777 shader_glsl_get_sample_function(ins
->ctx
, reg
, 0, &sample_function
);
3779 /* Sample the texture using the calculated coordinates */
3780 shader_glsl_gen_sample_code(ins
, reg
, &sample_function
, WINED3DSP_NOSWIZZLE
, NULL
, NULL
, NULL
, "tmp0.xyz");
3782 tex_mx
->current_row
= 0;
3785 /** Process the WINED3DSIO_TEXM3X3 instruction in GLSL
3786 * Perform the 3rd row of a 3x3 matrix multiply */
3787 static void shader_glsl_texm3x3(const struct wined3d_shader_instruction
*ins
)
3789 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
3790 struct wined3d_shader_tex_mx
*tex_mx
= ins
->ctx
->tex_mx
;
3791 DWORD reg
= ins
->dst
[0].reg
.idx
[0].offset
;
3792 struct glsl_src_param src0_param
;
3795 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
3797 shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
3798 shader_glsl_get_write_mask(&ins
->dst
[0], dst_mask
);
3799 shader_addline(ins
->ctx
->buffer
, "vec4(tmp0.xy, dot(T%u.xyz, %s), 1.0)%s);\n", reg
, src0_param
.param_str
, dst_mask
);
3801 tex_mx
->current_row
= 0;
3804 /* Process the WINED3DSIO_TEXM3X3SPEC instruction in GLSL
3805 * Perform the final texture lookup based on the previous 2 3x3 matrix multiplies */
3806 static void shader_glsl_texm3x3spec(const struct wined3d_shader_instruction
*ins
)
3808 struct glsl_src_param src0_param
;
3809 struct glsl_src_param src1_param
;
3810 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
3811 struct wined3d_shader_tex_mx
*tex_mx
= ins
->ctx
->tex_mx
;
3812 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
3813 struct glsl_sample_function sample_function
;
3814 DWORD reg
= ins
->dst
[0].reg
.idx
[0].offset
;
3817 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
3818 shader_glsl_add_src_param(ins
, &ins
->src
[1], src_mask
, &src1_param
);
3820 /* Perform the last matrix multiply operation */
3821 shader_addline(buffer
, "tmp0.z = dot(T%u.xyz, %s);\n", reg
, src0_param
.param_str
);
3822 /* Reflection calculation */
3823 shader_addline(buffer
, "tmp0.xyz = -reflect((%s), normalize(tmp0.xyz));\n", src1_param
.param_str
);
3825 /* Dependent read, not valid with conditional NP2 */
3826 shader_glsl_get_sample_function(ins
->ctx
, reg
, 0, &sample_function
);
3827 shader_glsl_write_mask_to_str(sample_function
.coord_mask
, coord_mask
);
3829 /* Sample the texture */
3830 shader_glsl_gen_sample_code(ins
, reg
, &sample_function
, WINED3DSP_NOSWIZZLE
,
3831 NULL
, NULL
, NULL
, "tmp0%s", coord_mask
);
3833 tex_mx
->current_row
= 0;
3836 /* Process the WINED3DSIO_TEXM3X3VSPEC instruction in GLSL
3837 * Perform the final texture lookup based on the previous 2 3x3 matrix multiplies */
3838 static void shader_glsl_texm3x3vspec(const struct wined3d_shader_instruction
*ins
)
3840 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
3841 struct wined3d_shader_tex_mx
*tex_mx
= ins
->ctx
->tex_mx
;
3842 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
3843 struct glsl_sample_function sample_function
;
3844 DWORD reg
= ins
->dst
[0].reg
.idx
[0].offset
;
3845 struct glsl_src_param src0_param
;
3848 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
3850 /* Perform the last matrix multiply operation */
3851 shader_addline(buffer
, "tmp0.z = dot(vec3(T%u), vec3(%s));\n", reg
, src0_param
.param_str
);
3853 /* Construct the eye-ray vector from w coordinates */
3854 shader_addline(buffer
, "tmp1.xyz = normalize(vec3(gl_TexCoord[%u].w, gl_TexCoord[%u].w, gl_TexCoord[%u].w));\n",
3855 tex_mx
->texcoord_w
[0], tex_mx
->texcoord_w
[1], reg
);
3856 shader_addline(buffer
, "tmp0.xyz = -reflect(tmp1.xyz, normalize(tmp0.xyz));\n");
3858 /* Dependent read, not valid with conditional NP2 */
3859 shader_glsl_get_sample_function(ins
->ctx
, reg
, 0, &sample_function
);
3860 shader_glsl_write_mask_to_str(sample_function
.coord_mask
, coord_mask
);
3862 /* Sample the texture using the calculated coordinates */
3863 shader_glsl_gen_sample_code(ins
, reg
, &sample_function
, WINED3DSP_NOSWIZZLE
,
3864 NULL
, NULL
, NULL
, "tmp0%s", coord_mask
);
3866 tex_mx
->current_row
= 0;
3869 /** Process the WINED3DSIO_TEXBEM instruction in GLSL.
3870 * Apply a fake bump map transform.
3871 * texbem is pshader <= 1.3 only, this saves a few version checks
3873 static void shader_glsl_texbem(const struct wined3d_shader_instruction
*ins
)
3875 const struct shader_glsl_ctx_priv
*priv
= ins
->ctx
->backend_data
;
3876 struct glsl_sample_function sample_function
;
3877 struct glsl_src_param coord_param
;
3883 sampler_idx
= ins
->dst
[0].reg
.idx
[0].offset
;
3884 flags
= (priv
->cur_ps_args
->tex_transform
>> sampler_idx
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
)
3885 & WINED3D_PSARGS_TEXTRANSFORM_MASK
;
3887 /* Dependent read, not valid with conditional NP2 */
3888 shader_glsl_get_sample_function(ins
->ctx
, sampler_idx
, 0, &sample_function
);
3889 mask
= sample_function
.coord_mask
;
3891 shader_glsl_write_mask_to_str(mask
, coord_mask
);
3893 /* With projected textures, texbem only divides the static texture coord,
3894 * not the displacement, so we can't let GL handle this. */
3895 if (flags
& WINED3D_PSARGS_PROJECTED
)
3898 char coord_div_mask
[3];
3899 switch (flags
& ~WINED3D_PSARGS_PROJECTED
)
3901 case WINED3D_TTFF_COUNT1
:
3902 FIXME("WINED3D_TTFF_PROJECTED with WINED3D_TTFF_COUNT1?\n");
3904 case WINED3D_TTFF_COUNT2
:
3905 div_mask
= WINED3DSP_WRITEMASK_1
;
3907 case WINED3D_TTFF_COUNT3
:
3908 div_mask
= WINED3DSP_WRITEMASK_2
;
3910 case WINED3D_TTFF_COUNT4
:
3911 case WINED3D_TTFF_DISABLE
:
3912 div_mask
= WINED3DSP_WRITEMASK_3
;
3915 shader_glsl_write_mask_to_str(div_mask
, coord_div_mask
);
3916 shader_addline(ins
->ctx
->buffer
, "T%u%s /= T%u%s;\n", sampler_idx
, coord_mask
, sampler_idx
, coord_div_mask
);
3919 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
, &coord_param
);
3921 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, WINED3DSP_NOSWIZZLE
, NULL
, NULL
, NULL
,
3922 "T%u%s + vec4(bumpenvmat%d * %s, 0.0, 0.0)%s", sampler_idx
, coord_mask
, sampler_idx
,
3923 coord_param
.param_str
, coord_mask
);
3925 if (ins
->handler_idx
== WINED3DSIH_TEXBEML
)
3927 struct glsl_src_param luminance_param
;
3928 struct glsl_dst_param dst_param
;
3930 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_2
, &luminance_param
);
3931 shader_glsl_add_dst_param(ins
, &ins
->dst
[0], &dst_param
);
3933 shader_addline(ins
->ctx
->buffer
, "%s%s *= (%s * luminancescale%d + luminanceoffset%d);\n",
3934 dst_param
.reg_name
, dst_param
.mask_str
,
3935 luminance_param
.param_str
, sampler_idx
, sampler_idx
);
3939 static void shader_glsl_bem(const struct wined3d_shader_instruction
*ins
)
3941 DWORD sampler_idx
= ins
->dst
[0].reg
.idx
[0].offset
;
3942 struct glsl_src_param src0_param
, src1_param
;
3944 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
, &src0_param
);
3945 shader_glsl_add_src_param(ins
, &ins
->src
[1], WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
, &src1_param
);
3947 shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
3948 shader_addline(ins
->ctx
->buffer
, "%s + bumpenvmat%d * %s);\n",
3949 src0_param
.param_str
, sampler_idx
, src1_param
.param_str
);
3952 /** Process the WINED3DSIO_TEXREG2AR instruction in GLSL
3953 * Sample 2D texture at dst using the alpha & red (wx) components of src as texture coordinates */
3954 static void shader_glsl_texreg2ar(const struct wined3d_shader_instruction
*ins
)
3956 DWORD sampler_idx
= ins
->dst
[0].reg
.idx
[0].offset
;
3957 struct glsl_sample_function sample_function
;
3958 struct glsl_src_param src0_param
;
3960 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_ALL
, &src0_param
);
3962 shader_glsl_get_sample_function(ins
->ctx
, sampler_idx
, 0, &sample_function
);
3963 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, WINED3DSP_NOSWIZZLE
, NULL
, NULL
, NULL
,
3964 "%s.wx", src0_param
.reg_name
);
3967 /** Process the WINED3DSIO_TEXREG2GB instruction in GLSL
3968 * Sample 2D texture at dst using the green & blue (yz) components of src as texture coordinates */
3969 static void shader_glsl_texreg2gb(const struct wined3d_shader_instruction
*ins
)
3971 DWORD sampler_idx
= ins
->dst
[0].reg
.idx
[0].offset
;
3972 struct glsl_sample_function sample_function
;
3973 struct glsl_src_param src0_param
;
3975 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_ALL
, &src0_param
);
3977 shader_glsl_get_sample_function(ins
->ctx
, sampler_idx
, 0, &sample_function
);
3978 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, WINED3DSP_NOSWIZZLE
, NULL
, NULL
, NULL
,
3979 "%s.yz", src0_param
.reg_name
);
3982 /** Process the WINED3DSIO_TEXREG2RGB instruction in GLSL
3983 * Sample texture at dst using the rgb (xyz) components of src as texture coordinates */
3984 static void shader_glsl_texreg2rgb(const struct wined3d_shader_instruction
*ins
)
3986 DWORD sampler_idx
= ins
->dst
[0].reg
.idx
[0].offset
;
3987 struct glsl_sample_function sample_function
;
3988 struct glsl_src_param src0_param
;
3990 /* Dependent read, not valid with conditional NP2 */
3991 shader_glsl_get_sample_function(ins
->ctx
, sampler_idx
, 0, &sample_function
);
3992 shader_glsl_add_src_param(ins
, &ins
->src
[0], sample_function
.coord_mask
, &src0_param
);
3994 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, WINED3DSP_NOSWIZZLE
, NULL
, NULL
, NULL
,
3995 "%s", src0_param
.param_str
);
3998 /** Process the WINED3DSIO_TEXKILL instruction in GLSL.
3999 * If any of the first 3 components are < 0, discard this pixel */
4000 static void shader_glsl_texkill(const struct wined3d_shader_instruction
*ins
)
4002 struct glsl_dst_param dst_param
;
4004 /* The argument is a destination parameter, and no writemasks are allowed */
4005 shader_glsl_add_dst_param(ins
, &ins
->dst
[0], &dst_param
);
4006 if (ins
->ctx
->reg_maps
->shader_version
.major
>= 2)
4008 /* 2.0 shaders compare all 4 components in texkill */
4009 shader_addline(ins
->ctx
->buffer
, "if (any(lessThan(%s.xyzw, vec4(0.0)))) discard;\n", dst_param
.reg_name
);
4011 /* 1.X shaders only compare the first 3 components, probably due to the nature of the texkill
4012 * instruction as a tex* instruction, and phase, which kills all a / w components. Even if all
4013 * 4 components are defined, only the first 3 are used
4015 shader_addline(ins
->ctx
->buffer
, "if (any(lessThan(%s.xyz, vec3(0.0)))) discard;\n", dst_param
.reg_name
);
4019 /** Process the WINED3DSIO_DP2ADD instruction in GLSL.
4020 * dst = dot2(src0, src1) + src2 */
4021 static void shader_glsl_dp2add(const struct wined3d_shader_instruction
*ins
)
4023 struct glsl_src_param src0_param
;
4024 struct glsl_src_param src1_param
;
4025 struct glsl_src_param src2_param
;
4027 unsigned int mask_size
;
4029 write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
4030 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
4032 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
, &src0_param
);
4033 shader_glsl_add_src_param(ins
, &ins
->src
[1], WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
, &src1_param
);
4034 shader_glsl_add_src_param(ins
, &ins
->src
[2], WINED3DSP_WRITEMASK_0
, &src2_param
);
4036 if (mask_size
> 1) {
4037 shader_addline(ins
->ctx
->buffer
, "vec%d(dot(%s, %s) + %s));\n",
4038 mask_size
, src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
4040 shader_addline(ins
->ctx
->buffer
, "dot(%s, %s) + %s);\n",
4041 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
4045 static void shader_glsl_input_pack(const struct wined3d_shader
*shader
, struct wined3d_shader_buffer
*buffer
,
4046 const struct wined3d_shader_signature_element
*input_signature
,
4047 const struct wined3d_shader_reg_maps
*reg_maps
,
4048 enum vertexprocessing_mode vertexprocessing
)
4050 WORD map
= reg_maps
->input_registers
;
4053 for (i
= 0; map
; map
>>= 1, ++i
)
4055 const char *semantic_name
;
4060 if (!(map
& 1)) continue;
4062 semantic_name
= input_signature
[i
].semantic_name
;
4063 semantic_idx
= input_signature
[i
].semantic_idx
;
4064 shader_glsl_write_mask_to_str(input_signature
[i
].mask
, reg_mask
);
4066 if (shader_match_semantic(semantic_name
, WINED3D_DECL_USAGE_TEXCOORD
))
4068 if (semantic_idx
< 8 && vertexprocessing
== pretransformed
)
4069 shader_addline(buffer
, "ps_in[%u]%s = gl_TexCoord[%u]%s;\n",
4070 shader
->u
.ps
.input_reg_map
[i
], reg_mask
, semantic_idx
, reg_mask
);
4072 shader_addline(buffer
, "ps_in[%u]%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
4073 shader
->u
.ps
.input_reg_map
[i
], reg_mask
, reg_mask
);
4075 else if (shader_match_semantic(semantic_name
, WINED3D_DECL_USAGE_COLOR
))
4078 shader_addline(buffer
, "ps_in[%u]%s = vec4(gl_Color)%s;\n",
4079 shader
->u
.ps
.input_reg_map
[i
], reg_mask
, reg_mask
);
4080 else if (semantic_idx
== 1)
4081 shader_addline(buffer
, "ps_in[%u]%s = vec4(gl_SecondaryColor)%s;\n",
4082 shader
->u
.ps
.input_reg_map
[i
], reg_mask
, reg_mask
);
4084 shader_addline(buffer
, "ps_in[%u]%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
4085 shader
->u
.ps
.input_reg_map
[i
], reg_mask
, reg_mask
);
4089 shader_addline(buffer
, "ps_in[%u]%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
4090 shader
->u
.ps
.input_reg_map
[i
], reg_mask
, reg_mask
);
4095 /*********************************************
4096 * Vertex Shader Specific Code begins here
4097 ********************************************/
4099 static void add_glsl_program_entry(struct shader_glsl_priv
*priv
, struct glsl_shader_prog_link
*entry
)
4101 struct glsl_program_key key
;
4103 key
.vs_id
= entry
->vs_id
;
4104 key
.ps_id
= entry
->ps_id
;
4106 if (wine_rb_put(&priv
->program_lookup
, &key
, &entry
->program_lookup_entry
) == -1)
4108 ERR("Failed to insert program entry.\n");
4112 static struct glsl_shader_prog_link
*get_glsl_program_entry(const struct shader_glsl_priv
*priv
,
4113 GLhandleARB vs_id
, GLhandleARB ps_id
)
4115 struct wine_rb_entry
*entry
;
4116 struct glsl_program_key key
;
4121 entry
= wine_rb_get(&priv
->program_lookup
, &key
);
4122 return entry
? WINE_RB_ENTRY_VALUE(entry
, struct glsl_shader_prog_link
, program_lookup_entry
) : NULL
;
4125 /* GL locking is done by the caller */
4126 static void delete_glsl_program_entry(struct shader_glsl_priv
*priv
, const struct wined3d_gl_info
*gl_info
,
4127 struct glsl_shader_prog_link
*entry
)
4129 struct glsl_program_key key
;
4131 key
.vs_id
= entry
->vs_id
;
4132 key
.ps_id
= entry
->ps_id
;
4133 wine_rb_remove(&priv
->program_lookup
, &key
);
4135 GL_EXTCALL(glDeleteObjectARB(entry
->programId
));
4137 list_remove(&entry
->vshader_entry
);
4139 list_remove(&entry
->pshader_entry
);
4140 HeapFree(GetProcessHeap(), 0, entry
->vuniformF_locations
);
4141 HeapFree(GetProcessHeap(), 0, entry
->puniformF_locations
);
4142 HeapFree(GetProcessHeap(), 0, entry
);
4145 static void handle_ps3_input(struct wined3d_shader_buffer
*buffer
,
4146 const struct wined3d_gl_info
*gl_info
, const DWORD
*map
,
4147 const struct wined3d_shader_signature_element
*input_signature
,
4148 const struct wined3d_shader_reg_maps
*reg_maps_in
,
4149 const struct wined3d_shader_signature_element
*output_signature
,
4150 const struct wined3d_shader_reg_maps
*reg_maps_out
)
4153 const char *semantic_name_in
;
4154 UINT semantic_idx_in
;
4157 unsigned int in_count
= vec4_varyings(3, gl_info
);
4159 char destination
[50];
4160 WORD input_map
, output_map
;
4162 set
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*set
) * (in_count
+ 2));
4164 input_map
= reg_maps_in
->input_registers
;
4165 for (i
= 0; input_map
; input_map
>>= 1, ++i
)
4167 if (!(input_map
& 1)) continue;
4170 /* Declared, but not read register */
4171 if (in_idx
== ~0U) continue;
4172 if (in_idx
>= (in_count
+ 2))
4174 FIXME("More input varyings declared than supported, expect issues.\n");
4178 if (in_idx
== in_count
)
4179 sprintf(destination
, "gl_FrontColor");
4180 else if (in_idx
== in_count
+ 1)
4181 sprintf(destination
, "gl_FrontSecondaryColor");
4183 sprintf(destination
, "ps_in[%u]", in_idx
);
4185 semantic_name_in
= input_signature
[i
].semantic_name
;
4186 semantic_idx_in
= input_signature
[i
].semantic_idx
;
4189 output_map
= reg_maps_out
->output_registers
;
4190 for (j
= 0; output_map
; output_map
>>= 1, ++j
)
4194 if (!(output_map
& 1)
4195 || semantic_idx_in
!= output_signature
[j
].semantic_idx
4196 || strcmp(semantic_name_in
, output_signature
[j
].semantic_name
)
4197 || !(mask
= input_signature
[i
].mask
& output_signature
[j
].mask
))
4201 shader_glsl_write_mask_to_str(mask
, reg_mask
);
4203 shader_addline(buffer
, "%s%s = vs_out[%u]%s;\n",
4204 destination
, reg_mask
, j
, reg_mask
);
4208 for (i
= 0; i
< in_count
+ 2; ++i
)
4212 if (!set
[i
] || set
[i
] == WINED3DSP_WRITEMASK_ALL
)
4215 if (set
[i
] == ~0U) set
[i
] = 0;
4218 if (!(set
[i
] & WINED3DSP_WRITEMASK_0
)) reg_mask
[size
++] = 'x';
4219 if (!(set
[i
] & WINED3DSP_WRITEMASK_1
)) reg_mask
[size
++] = 'y';
4220 if (!(set
[i
] & WINED3DSP_WRITEMASK_2
)) reg_mask
[size
++] = 'z';
4221 if (!(set
[i
] & WINED3DSP_WRITEMASK_3
)) reg_mask
[size
++] = 'w';
4222 reg_mask
[size
] = '\0';
4225 sprintf(destination
, "gl_FrontColor");
4226 else if (i
== in_count
+ 1)
4227 sprintf(destination
, "gl_FrontSecondaryColor");
4229 sprintf(destination
, "ps_in[%u]", i
);
4231 if (size
== 1) shader_addline(buffer
, "%s.%s = 0.0;\n", destination
, reg_mask
);
4232 else shader_addline(buffer
, "%s.%s = vec%u(0.0);\n", destination
, reg_mask
, size
);
4235 HeapFree(GetProcessHeap(), 0, set
);
4238 /* GL locking is done by the caller */
4239 static GLhandleARB
generate_param_reorder_function(struct wined3d_shader_buffer
*buffer
,
4240 const struct wined3d_shader
*vs
, const struct wined3d_shader
*ps
,
4241 const struct wined3d_gl_info
*gl_info
)
4243 GLhandleARB ret
= 0;
4244 DWORD ps_major
= ps
? ps
->reg_maps
.shader_version
.major
: 0;
4246 const char *semantic_name
;
4249 const struct wined3d_shader_signature_element
*output_signature
= vs
->output_signature
;
4250 WORD map
= vs
->reg_maps
.output_registers
;
4252 shader_buffer_clear(buffer
);
4254 shader_addline(buffer
, "#version 120\n");
4258 shader_addline(buffer
, "void order_ps_input(in vec4 vs_out[%u])\n{\n", vs
->limits
.packed_output
);
4260 for (i
= 0; map
; map
>>= 1, ++i
)
4264 if (!(map
& 1)) continue;
4266 semantic_name
= output_signature
[i
].semantic_name
;
4267 semantic_idx
= output_signature
[i
].semantic_idx
;
4268 write_mask
= output_signature
[i
].mask
;
4269 shader_glsl_write_mask_to_str(write_mask
, reg_mask
);
4271 if (shader_match_semantic(semantic_name
, WINED3D_DECL_USAGE_COLOR
))
4274 shader_addline(buffer
, "gl_FrontColor%s = vs_out[%u]%s;\n",
4275 reg_mask
, i
, reg_mask
);
4276 else if (semantic_idx
== 1)
4277 shader_addline(buffer
, "gl_FrontSecondaryColor%s = vs_out[%u]%s;\n",
4278 reg_mask
, i
, reg_mask
);
4280 else if (shader_match_semantic(semantic_name
, WINED3D_DECL_USAGE_POSITION
))
4282 shader_addline(buffer
, "gl_Position%s = vs_out[%u]%s;\n",
4283 reg_mask
, i
, reg_mask
);
4285 else if (shader_match_semantic(semantic_name
, WINED3D_DECL_USAGE_TEXCOORD
))
4287 if (semantic_idx
< 8)
4289 if (!(gl_info
->quirks
& WINED3D_QUIRK_SET_TEXCOORD_W
) || ps_major
> 0)
4290 write_mask
|= WINED3DSP_WRITEMASK_3
;
4292 shader_addline(buffer
, "gl_TexCoord[%u]%s = vs_out[%u]%s;\n",
4293 semantic_idx
, reg_mask
, i
, reg_mask
);
4294 if (!(write_mask
& WINED3DSP_WRITEMASK_3
))
4295 shader_addline(buffer
, "gl_TexCoord[%u].w = 1.0;\n", semantic_idx
);
4298 else if (shader_match_semantic(semantic_name
, WINED3D_DECL_USAGE_PSIZE
))
4300 shader_addline(buffer
, "gl_PointSize = vs_out[%u].%c;\n", i
, reg_mask
[1]);
4302 else if (shader_match_semantic(semantic_name
, WINED3D_DECL_USAGE_FOG
))
4304 shader_addline(buffer
, "gl_FogFragCoord = clamp(vs_out[%u].%c, 0.0, 1.0);\n", i
, reg_mask
[1]);
4307 shader_addline(buffer
, "}\n");
4311 UINT in_count
= min(vec4_varyings(ps_major
, gl_info
), ps
->limits
.packed_input
);
4312 /* This one is tricky: a 3.0 pixel shader reads from a 3.0 vertex shader */
4313 shader_addline(buffer
, "varying vec4 ps_in[%u];\n", in_count
);
4314 shader_addline(buffer
, "void order_ps_input(in vec4 vs_out[%u])\n{\n", vs
->limits
.packed_output
);
4316 /* First, sort out position and point size. Those are not passed to the pixel shader */
4317 for (i
= 0; map
; map
>>= 1, ++i
)
4319 if (!(map
& 1)) continue;
4321 semantic_name
= output_signature
[i
].semantic_name
;
4322 shader_glsl_write_mask_to_str(output_signature
[i
].mask
, reg_mask
);
4324 if (shader_match_semantic(semantic_name
, WINED3D_DECL_USAGE_POSITION
))
4326 shader_addline(buffer
, "gl_Position%s = vs_out[%u]%s;\n",
4327 reg_mask
, i
, reg_mask
);
4329 else if (shader_match_semantic(semantic_name
, WINED3D_DECL_USAGE_PSIZE
))
4331 shader_addline(buffer
, "gl_PointSize = vs_out[%u].%c;\n", i
, reg_mask
[1]);
4335 /* Then, fix the pixel shader input */
4336 handle_ps3_input(buffer
, gl_info
, ps
->u
.ps
.input_reg_map
, ps
->input_signature
,
4337 &ps
->reg_maps
, output_signature
, &vs
->reg_maps
);
4339 shader_addline(buffer
, "}\n");
4342 ret
= GL_EXTCALL(glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB
));
4343 checkGLcall("glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB)");
4344 shader_glsl_compile(gl_info
, ret
, buffer
->buffer
);
4349 /* GL locking is done by the caller */
4350 static void hardcode_local_constants(const struct wined3d_shader
*shader
,
4351 const struct wined3d_gl_info
*gl_info
, GLhandleARB programId
, const char *prefix
)
4353 const struct wined3d_shader_lconst
*lconst
;
4358 LIST_FOR_EACH_ENTRY(lconst
, &shader
->constantsF
, struct wined3d_shader_lconst
, entry
)
4360 value
= (const float *)lconst
->value
;
4361 snprintf(glsl_name
, sizeof(glsl_name
), "%s_lc%u", prefix
, lconst
->idx
);
4362 tmp_loc
= GL_EXTCALL(glGetUniformLocationARB(programId
, glsl_name
));
4363 GL_EXTCALL(glUniform4fvARB(tmp_loc
, 1, value
));
4365 checkGLcall("Hardcoding local constants");
4368 /* GL locking is done by the caller */
4369 static GLuint
shader_glsl_generate_pshader(const struct wined3d_context
*context
,
4370 struct wined3d_shader_buffer
*buffer
, const struct wined3d_shader
*shader
,
4371 const struct ps_compile_args
*args
, struct ps_np2fixup_info
*np2fixup_info
)
4373 const struct wined3d_shader_reg_maps
*reg_maps
= &shader
->reg_maps
;
4374 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
4375 const DWORD
*function
= shader
->function
;
4376 struct shader_glsl_ctx_priv priv_ctx
;
4378 /* Create the hw GLSL shader object and assign it as the shader->prgId */
4379 GLhandleARB shader_obj
= GL_EXTCALL(glCreateShaderObjectARB(GL_FRAGMENT_SHADER_ARB
));
4381 memset(&priv_ctx
, 0, sizeof(priv_ctx
));
4382 priv_ctx
.cur_ps_args
= args
;
4383 priv_ctx
.cur_np2fixup_info
= np2fixup_info
;
4385 shader_addline(buffer
, "#version 120\n");
4387 if (gl_info
->supported
[ARB_SHADER_BIT_ENCODING
])
4388 shader_addline(buffer
, "#extension GL_ARB_shader_bit_encoding : enable\n");
4389 if (gl_info
->supported
[ARB_SHADER_TEXTURE_LOD
])
4390 shader_addline(buffer
, "#extension GL_ARB_shader_texture_lod : enable\n");
4391 /* The spec says that it doesn't have to be explicitly enabled, but the
4392 * nvidia drivers write a warning if we don't do so. */
4393 if (gl_info
->supported
[ARB_TEXTURE_RECTANGLE
])
4394 shader_addline(buffer
, "#extension GL_ARB_texture_rectangle : enable\n");
4395 if (gl_info
->supported
[EXT_GPU_SHADER4
])
4396 shader_addline(buffer
, "#extension GL_EXT_gpu_shader4 : enable\n");
4398 /* Base Declarations */
4399 shader_generate_glsl_declarations(context
, buffer
, shader
, reg_maps
, &priv_ctx
);
4401 /* Pack 3.0 inputs */
4402 if (reg_maps
->shader_version
.major
>= 3 && args
->vp_mode
!= vertexshader
)
4403 shader_glsl_input_pack(shader
, buffer
, shader
->input_signature
, reg_maps
, args
->vp_mode
);
4405 /* Base Shader Body */
4406 shader_generate_main(shader
, buffer
, reg_maps
, function
, &priv_ctx
);
4408 /* Pixel shaders < 2.0 place the resulting color in R0 implicitly */
4409 if (reg_maps
->shader_version
.major
< 2)
4411 /* Some older cards like GeforceFX ones don't support multiple buffers, so also not gl_FragData */
4412 shader_addline(buffer
, "gl_FragData[0] = R0;\n");
4415 if (args
->srgb_correction
)
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");
4424 /* Pixel shader < 3.0 do not replace the fog stage.
4425 * This implements linear fog computation and blending.
4426 * TODO: non linear fog
4427 * NOTE: gl_Fog.start and gl_Fog.end don't hold fog start s and end e but
4428 * -1/(e-s) and e/(e-s) respectively.
4430 if (reg_maps
->shader_version
.major
< 3)
4433 case FOG_OFF
: break;
4435 shader_addline(buffer
, "float fogstart = -1.0 / (gl_Fog.end - gl_Fog.start);\n");
4436 shader_addline(buffer
, "float fogend = gl_Fog.end * -fogstart;\n");
4437 shader_addline(buffer
, "float Fog = clamp(gl_FogFragCoord * fogstart + fogend, 0.0, 1.0);\n");
4438 shader_addline(buffer
, "gl_FragData[0].xyz = mix(gl_Fog.color.xyz, gl_FragData[0].xyz, Fog);\n");
4441 /* Fog = e^(-gl_Fog.density * gl_FogFragCoord) */
4442 shader_addline(buffer
, "float Fog = exp(-gl_Fog.density * gl_FogFragCoord);\n");
4443 shader_addline(buffer
, "Fog = clamp(Fog, 0.0, 1.0);\n");
4444 shader_addline(buffer
, "gl_FragData[0].xyz = mix(gl_Fog.color.xyz, gl_FragData[0].xyz, Fog);\n");
4447 /* Fog = e^(-(gl_Fog.density * gl_FogFragCoord)^2) */
4448 shader_addline(buffer
, "float Fog = exp(-gl_Fog.density * gl_Fog.density * gl_FogFragCoord * gl_FogFragCoord);\n");
4449 shader_addline(buffer
, "Fog = clamp(Fog, 0.0, 1.0);\n");
4450 shader_addline(buffer
, "gl_FragData[0].xyz = mix(gl_Fog.color.xyz, gl_FragData[0].xyz, Fog);\n");
4455 shader_addline(buffer
, "}\n");
4457 TRACE("Compiling shader object %u\n", shader_obj
);
4458 shader_glsl_compile(gl_info
, shader_obj
, buffer
->buffer
);
4460 /* Store the shader object */
4464 /* GL locking is done by the caller */
4465 static GLuint
shader_glsl_generate_vshader(const struct wined3d_context
*context
,
4466 struct wined3d_shader_buffer
*buffer
, const struct wined3d_shader
*shader
,
4467 const struct vs_compile_args
*args
)
4469 const struct wined3d_shader_reg_maps
*reg_maps
= &shader
->reg_maps
;
4470 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
4471 const DWORD
*function
= shader
->function
;
4472 struct shader_glsl_ctx_priv priv_ctx
;
4474 /* Create the hw GLSL shader program and assign it as the shader->prgId */
4475 GLhandleARB shader_obj
= GL_EXTCALL(glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB
));
4477 shader_addline(buffer
, "#version 120\n");
4479 if (gl_info
->supported
[ARB_SHADER_BIT_ENCODING
])
4480 shader_addline(buffer
, "#extension GL_ARB_shader_bit_encoding : enable\n");
4481 if (gl_info
->supported
[EXT_GPU_SHADER4
])
4482 shader_addline(buffer
, "#extension GL_EXT_gpu_shader4 : enable\n");
4484 memset(&priv_ctx
, 0, sizeof(priv_ctx
));
4485 priv_ctx
.cur_vs_args
= args
;
4487 /* Base Declarations */
4488 shader_generate_glsl_declarations(context
, buffer
, shader
, reg_maps
, &priv_ctx
);
4490 /* Base Shader Body */
4491 shader_generate_main(shader
, buffer
, reg_maps
, function
, &priv_ctx
);
4493 /* Unpack outputs */
4494 shader_addline(buffer
, "order_ps_input(vs_out);\n");
4496 /* The D3DRS_FOGTABLEMODE render state defines if the shader-generated fog coord is used
4497 * or if the fragment depth is used. If the fragment depth is used(FOGTABLEMODE != NONE),
4498 * the fog frag coord is thrown away. If the fog frag coord is used, but not written by
4499 * the shader, it is set to 0.0(fully fogged, since start = 1.0, end = 0.0)
4501 if (args
->fog_src
== VS_FOG_Z
)
4502 shader_addline(buffer
, "gl_FogFragCoord = gl_Position.z;\n");
4503 else if (!reg_maps
->fog
)
4504 shader_addline(buffer
, "gl_FogFragCoord = 0.0;\n");
4506 /* We always store the clipplanes without y inversion */
4507 if (args
->clip_enabled
)
4508 shader_addline(buffer
, "gl_ClipVertex = gl_Position;\n");
4510 /* Write the final position.
4512 * OpenGL coordinates specify the center of the pixel while d3d coords specify
4513 * the corner. The offsets are stored in z and w in posFixup. posFixup.y contains
4514 * 1.0 or -1.0 to turn the rendering upside down for offscreen rendering. PosFixup.x
4515 * contains 1.0 to allow a mad.
4517 shader_addline(buffer
, "gl_Position.y = gl_Position.y * posFixup.y;\n");
4518 shader_addline(buffer
, "gl_Position.xy += posFixup.zw * gl_Position.ww;\n");
4520 /* Z coord [0;1]->[-1;1] mapping, see comment in transform_projection in state.c
4522 * Basically we want (in homogeneous coordinates) z = z * 2 - 1. However, shaders are run
4523 * before the homogeneous divide, so we have to take the w into account: z = ((z / w) * 2 - 1) * w,
4524 * which is the same as z = z * 2 - w.
4526 shader_addline(buffer
, "gl_Position.z = gl_Position.z * 2.0 - gl_Position.w;\n");
4528 shader_addline(buffer
, "}\n");
4530 TRACE("Compiling shader object %u\n", shader_obj
);
4531 shader_glsl_compile(gl_info
, shader_obj
, buffer
->buffer
);
4536 static GLhandleARB
find_glsl_pshader(const struct wined3d_context
*context
,
4537 struct wined3d_shader_buffer
*buffer
, struct wined3d_shader
*shader
,
4538 const struct ps_compile_args
*args
, const struct ps_np2fixup_info
**np2fixup_info
)
4540 struct wined3d_state
*state
= &shader
->device
->stateBlock
->state
;
4541 struct glsl_ps_compiled_shader
*gl_shaders
, *new_array
;
4542 struct glsl_shader_private
*shader_data
;
4543 struct ps_np2fixup_info
*np2fixup
;
4548 if (!shader
->backend_data
)
4550 shader
->backend_data
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*shader_data
));
4551 if (!shader
->backend_data
)
4553 ERR("Failed to allocate backend data.\n");
4557 shader_data
= shader
->backend_data
;
4558 gl_shaders
= shader_data
->gl_shaders
.ps
;
4560 /* Usually we have very few GL shaders for each d3d shader(just 1 or maybe 2),
4561 * so a linear search is more performant than a hashmap or a binary search
4562 * (cache coherency etc)
4564 for (i
= 0; i
< shader_data
->num_gl_shaders
; ++i
)
4566 if (!memcmp(&gl_shaders
[i
].args
, args
, sizeof(*args
)))
4568 if (args
->np2_fixup
)
4569 *np2fixup_info
= &gl_shaders
[i
].np2fixup
;
4570 return gl_shaders
[i
].prgId
;
4574 TRACE("No matching GL shader found for shader %p, compiling a new shader.\n", shader
);
4575 if(shader_data
->shader_array_size
== shader_data
->num_gl_shaders
) {
4576 if (shader_data
->num_gl_shaders
)
4578 new_size
= shader_data
->shader_array_size
+ max(1, shader_data
->shader_array_size
/ 2);
4579 new_array
= HeapReAlloc(GetProcessHeap(), 0, shader_data
->gl_shaders
.ps
,
4580 new_size
* sizeof(*gl_shaders
));
4584 new_array
= HeapAlloc(GetProcessHeap(), 0, sizeof(*gl_shaders
));
4589 ERR("Out of memory\n");
4592 shader_data
->gl_shaders
.ps
= new_array
;
4593 shader_data
->shader_array_size
= new_size
;
4594 gl_shaders
= new_array
;
4597 gl_shaders
[shader_data
->num_gl_shaders
].args
= *args
;
4599 np2fixup
= &gl_shaders
[shader_data
->num_gl_shaders
].np2fixup
;
4600 memset(np2fixup
, 0, sizeof(*np2fixup
));
4601 *np2fixup_info
= args
->np2_fixup
? np2fixup
: NULL
;
4603 pixelshader_update_samplers(&shader
->reg_maps
, state
->textures
);
4605 shader_buffer_clear(buffer
);
4606 ret
= shader_glsl_generate_pshader(context
, buffer
, shader
, args
, np2fixup
);
4607 gl_shaders
[shader_data
->num_gl_shaders
++].prgId
= ret
;
4612 static inline BOOL
vs_args_equal(const struct vs_compile_args
*stored
, const struct vs_compile_args
*new,
4613 const DWORD use_map
) {
4614 if((stored
->swizzle_map
& use_map
) != new->swizzle_map
) return FALSE
;
4615 if((stored
->clip_enabled
) != new->clip_enabled
) return FALSE
;
4616 return stored
->fog_src
== new->fog_src
;
4619 static GLhandleARB
find_glsl_vshader(const struct wined3d_context
*context
,
4620 struct wined3d_shader_buffer
*buffer
, struct wined3d_shader
*shader
,
4621 const struct vs_compile_args
*args
)
4625 DWORD use_map
= shader
->device
->strided_streams
.use_map
;
4626 struct glsl_vs_compiled_shader
*gl_shaders
, *new_array
;
4627 struct glsl_shader_private
*shader_data
;
4630 if (!shader
->backend_data
)
4632 shader
->backend_data
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*shader_data
));
4633 if (!shader
->backend_data
)
4635 ERR("Failed to allocate backend data.\n");
4639 shader_data
= shader
->backend_data
;
4640 gl_shaders
= shader_data
->gl_shaders
.vs
;
4642 /* Usually we have very few GL shaders for each d3d shader(just 1 or maybe 2),
4643 * so a linear search is more performant than a hashmap or a binary search
4644 * (cache coherency etc)
4646 for (i
= 0; i
< shader_data
->num_gl_shaders
; ++i
)
4648 if (vs_args_equal(&gl_shaders
[i
].args
, args
, use_map
))
4649 return gl_shaders
[i
].prgId
;
4652 TRACE("No matching GL shader found for shader %p, compiling a new shader.\n", shader
);
4654 if(shader_data
->shader_array_size
== shader_data
->num_gl_shaders
) {
4655 if (shader_data
->num_gl_shaders
)
4657 new_size
= shader_data
->shader_array_size
+ max(1, shader_data
->shader_array_size
/ 2);
4658 new_array
= HeapReAlloc(GetProcessHeap(), 0, shader_data
->gl_shaders
.vs
,
4659 new_size
* sizeof(*gl_shaders
));
4663 new_array
= HeapAlloc(GetProcessHeap(), 0, sizeof(*gl_shaders
));
4668 ERR("Out of memory\n");
4671 shader_data
->gl_shaders
.vs
= new_array
;
4672 shader_data
->shader_array_size
= new_size
;
4673 gl_shaders
= new_array
;
4676 gl_shaders
[shader_data
->num_gl_shaders
].args
= *args
;
4678 shader_buffer_clear(buffer
);
4679 ret
= shader_glsl_generate_vshader(context
, buffer
, shader
, args
);
4680 gl_shaders
[shader_data
->num_gl_shaders
++].prgId
= ret
;
4685 /** Sets the GLSL program ID for the given pixel and vertex shader combination.
4686 * It sets the programId on the current StateBlock (because it should be called
4687 * inside of the DrawPrimitive() part of the render loop).
4689 * If a program for the given combination does not exist, create one, and store
4690 * the program in the hash table. If it creates a program, it will link the
4691 * given objects, too.
4694 /* GL locking is done by the caller */
4695 static void set_glsl_shader_program(const struct wined3d_context
*context
,
4696 struct wined3d_device
*device
, BOOL use_ps
, BOOL use_vs
)
4698 const struct wined3d_state
*state
= &device
->stateBlock
->state
;
4699 struct wined3d_shader
*vshader
= use_vs
? state
->vertex_shader
: NULL
;
4700 struct wined3d_shader
*pshader
= use_ps
? state
->pixel_shader
: NULL
;
4701 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
4702 const struct ps_np2fixup_info
*np2fixup_info
= NULL
;
4703 struct shader_glsl_priv
*priv
= device
->shader_priv
;
4704 struct glsl_shader_prog_link
*entry
= NULL
;
4705 GLhandleARB programId
= 0;
4706 GLhandleARB reorder_shader_id
= 0;
4709 struct ps_compile_args ps_compile_args
;
4710 struct vs_compile_args vs_compile_args
;
4711 GLhandleARB vs_id
, ps_id
;
4715 find_vs_compile_args(state
, vshader
, &vs_compile_args
);
4716 vs_id
= find_glsl_vshader(context
, &priv
->shader_buffer
, vshader
, &vs_compile_args
);
4725 find_ps_compile_args(state
, pshader
, &ps_compile_args
);
4726 ps_id
= find_glsl_pshader(context
, &priv
->shader_buffer
,
4727 pshader
, &ps_compile_args
, &np2fixup_info
);
4734 if ((entry
= get_glsl_program_entry(priv
, vs_id
, ps_id
)))
4736 priv
->glsl_program
= entry
;
4740 /* If we get to this point, then no matching program exists, so we create one */
4741 programId
= GL_EXTCALL(glCreateProgramObjectARB());
4742 TRACE("Created new GLSL shader program %u\n", programId
);
4744 /* Create the entry */
4745 entry
= HeapAlloc(GetProcessHeap(), 0, sizeof(struct glsl_shader_prog_link
));
4746 entry
->programId
= programId
;
4747 entry
->vs_id
= vs_id
;
4748 entry
->ps_id
= ps_id
;
4749 entry
->constant_version
= 0;
4750 entry
->np2Fixup_info
= np2fixup_info
;
4751 /* Add the hash table entry */
4752 add_glsl_program_entry(priv
, entry
);
4754 /* Set the current program */
4755 priv
->glsl_program
= entry
;
4757 /* Attach GLSL vshader */
4760 WORD map
= vshader
->reg_maps
.input_registers
;
4763 reorder_shader_id
= generate_param_reorder_function(&priv
->shader_buffer
, vshader
, pshader
, gl_info
);
4764 TRACE("Attaching GLSL shader object %u to program %u\n", reorder_shader_id
, programId
);
4765 GL_EXTCALL(glAttachObjectARB(programId
, reorder_shader_id
));
4766 checkGLcall("glAttachObjectARB");
4767 /* Flag the reorder function for deletion, then it will be freed automatically when the program
4770 GL_EXTCALL(glDeleteObjectARB(reorder_shader_id
));
4772 TRACE("Attaching GLSL shader object %u to program %u.\n", vs_id
, programId
);
4773 GL_EXTCALL(glAttachObjectARB(programId
, vs_id
));
4774 checkGLcall("glAttachObjectARB");
4776 /* Bind vertex attributes to a corresponding index number to match
4777 * the same index numbers as ARB_vertex_programs (makes loading
4778 * vertex attributes simpler). With this method, we can use the
4779 * exact same code to load the attributes later for both ARB and
4782 * We have to do this here because we need to know the Program ID
4783 * in order to make the bindings work, and it has to be done prior
4784 * to linking the GLSL program. */
4785 for (i
= 0; map
; map
>>= 1, ++i
)
4787 if (!(map
& 1)) continue;
4789 snprintf(tmp_name
, sizeof(tmp_name
), "vs_in%u", i
);
4790 GL_EXTCALL(glBindAttribLocationARB(programId
, i
, tmp_name
));
4792 checkGLcall("glBindAttribLocationARB");
4794 list_add_head(&vshader
->linked_programs
, &entry
->vshader_entry
);
4797 /* Attach GLSL pshader */
4800 TRACE("Attaching GLSL shader object %u to program %u.\n", ps_id
, programId
);
4801 GL_EXTCALL(glAttachObjectARB(programId
, ps_id
));
4802 checkGLcall("glAttachObjectARB");
4804 list_add_head(&pshader
->linked_programs
, &entry
->pshader_entry
);
4807 /* Link the program */
4808 TRACE("Linking GLSL shader program %u\n", programId
);
4809 GL_EXTCALL(glLinkProgramARB(programId
));
4810 shader_glsl_validate_link(gl_info
, programId
);
4812 entry
->vuniformF_locations
= HeapAlloc(GetProcessHeap(), 0,
4813 sizeof(GLhandleARB
) * gl_info
->limits
.glsl_vs_float_constants
);
4814 for (i
= 0; i
< gl_info
->limits
.glsl_vs_float_constants
; ++i
)
4816 snprintf(glsl_name
, sizeof(glsl_name
), "vs_c[%u]", i
);
4817 entry
->vuniformF_locations
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, glsl_name
));
4819 for (i
= 0; i
< MAX_CONST_I
; ++i
)
4821 snprintf(glsl_name
, sizeof(glsl_name
), "vs_i[%u]", i
);
4822 entry
->vuniformI_locations
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, glsl_name
));
4824 entry
->puniformF_locations
= HeapAlloc(GetProcessHeap(), 0,
4825 sizeof(GLhandleARB
) * gl_info
->limits
.glsl_ps_float_constants
);
4826 for (i
= 0; i
< gl_info
->limits
.glsl_ps_float_constants
; ++i
)
4828 snprintf(glsl_name
, sizeof(glsl_name
), "ps_c[%u]", i
);
4829 entry
->puniformF_locations
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, glsl_name
));
4831 for (i
= 0; i
< MAX_CONST_I
; ++i
)
4833 snprintf(glsl_name
, sizeof(glsl_name
), "ps_i[%u]", i
);
4834 entry
->puniformI_locations
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, glsl_name
));
4840 for(i
= 0; i
< MAX_TEXTURES
; i
++) {
4841 sprintf(name
, "bumpenvmat%u", i
);
4842 entry
->bumpenvmat_location
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, name
));
4843 sprintf(name
, "luminancescale%u", i
);
4844 entry
->luminancescale_location
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, name
));
4845 sprintf(name
, "luminanceoffset%u", i
);
4846 entry
->luminanceoffset_location
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, name
));
4849 if (ps_compile_args
.np2_fixup
)
4851 if (entry
->np2Fixup_info
)
4852 entry
->np2Fixup_location
= GL_EXTCALL(glGetUniformLocationARB(programId
, "ps_samplerNP2Fixup"));
4854 FIXME("NP2 texcoord fixup needed for this pixelshader, but no fixup uniform found.\n");
4858 entry
->posFixup_location
= GL_EXTCALL(glGetUniformLocationARB(programId
, "posFixup"));
4859 entry
->ycorrection_location
= GL_EXTCALL(glGetUniformLocationARB(programId
, "ycorrection"));
4860 checkGLcall("Find glsl program uniform locations");
4862 if (pshader
&& pshader
->reg_maps
.shader_version
.major
>= 3
4863 && pshader
->u
.ps
.declared_in_count
> vec4_varyings(3, gl_info
))
4865 TRACE("Shader %d needs vertex color clamping disabled\n", programId
);
4866 entry
->vertex_color_clamp
= GL_FALSE
;
4868 entry
->vertex_color_clamp
= GL_FIXED_ONLY_ARB
;
4871 /* Set the shader to allow uniform loading on it */
4872 GL_EXTCALL(glUseProgramObjectARB(programId
));
4873 checkGLcall("glUseProgramObjectARB(programId)");
4875 /* Load the vertex and pixel samplers now. The function that finds the mappings makes sure
4876 * that it stays the same for each vertexshader-pixelshader pair(=linked glsl program). If
4877 * a pshader with fixed function pipeline is used there are no vertex samplers, and if a
4878 * vertex shader with fixed function pixel processing is used we make sure that the card
4879 * supports enough samplers to allow the max number of vertex samplers with all possible
4880 * fixed function fragment processing setups. So once the program is linked these samplers
4883 if (vshader
) shader_glsl_load_vsamplers(gl_info
, device
->texUnitMap
, programId
);
4884 if (pshader
) shader_glsl_load_psamplers(gl_info
, device
->texUnitMap
, programId
);
4886 /* If the local constants do not have to be loaded with the environment constants,
4887 * load them now to have them hardcoded in the GLSL program. This saves some CPU cycles
4890 if (pshader
&& !pshader
->load_local_constsF
)
4891 hardcode_local_constants(pshader
, gl_info
, programId
, "ps");
4892 if (vshader
&& !vshader
->load_local_constsF
)
4893 hardcode_local_constants(vshader
, gl_info
, programId
, "vs");
4896 /* GL locking is done by the caller */
4897 static GLhandleARB
create_glsl_blt_shader(const struct wined3d_gl_info
*gl_info
, enum tex_types tex_type
, BOOL masked
)
4899 GLhandleARB program_id
;
4900 GLhandleARB vshader_id
, pshader_id
;
4901 const char *blt_pshader
;
4903 static const char *blt_vshader
=
4907 " gl_Position = gl_Vertex;\n"
4908 " gl_FrontColor = vec4(1.0);\n"
4909 " gl_TexCoord[0] = gl_MultiTexCoord0;\n"
4912 static const char * const blt_pshaders_full
[tex_type_count
] =
4918 "uniform sampler2D sampler;\n"
4921 " gl_FragDepth = texture2D(sampler, gl_TexCoord[0].xy).x;\n"
4927 "uniform samplerCube sampler;\n"
4930 " gl_FragDepth = textureCube(sampler, gl_TexCoord[0].xyz).x;\n"
4934 "#extension GL_ARB_texture_rectangle : enable\n"
4935 "uniform sampler2DRect sampler;\n"
4938 " gl_FragDepth = texture2DRect(sampler, gl_TexCoord[0].xy).x;\n"
4942 static const char * const blt_pshaders_masked
[tex_type_count
] =
4948 "uniform sampler2D sampler;\n"
4949 "uniform vec4 mask;\n"
4952 " if (all(lessThan(gl_FragCoord.xy, mask.zw))) discard;\n"
4953 " gl_FragDepth = texture2D(sampler, gl_TexCoord[0].xy).x;\n"
4959 "uniform samplerCube sampler;\n"
4960 "uniform vec4 mask;\n"
4963 " if (all(lessThan(gl_FragCoord.xy, mask.zw))) discard;\n"
4964 " gl_FragDepth = textureCube(sampler, gl_TexCoord[0].xyz).x;\n"
4968 "#extension GL_ARB_texture_rectangle : enable\n"
4969 "uniform sampler2DRect sampler;\n"
4970 "uniform vec4 mask;\n"
4973 " if (all(lessThan(gl_FragCoord.xy, mask.zw))) discard;\n"
4974 " gl_FragDepth = texture2DRect(sampler, gl_TexCoord[0].xy).x;\n"
4978 blt_pshader
= masked
? blt_pshaders_masked
[tex_type
] : blt_pshaders_full
[tex_type
];
4981 FIXME("tex_type %#x not supported\n", tex_type
);
4985 vshader_id
= GL_EXTCALL(glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB
));
4986 shader_glsl_compile(gl_info
, vshader_id
, blt_vshader
);
4988 pshader_id
= GL_EXTCALL(glCreateShaderObjectARB(GL_FRAGMENT_SHADER_ARB
));
4989 shader_glsl_compile(gl_info
, pshader_id
, blt_pshader
);
4991 program_id
= GL_EXTCALL(glCreateProgramObjectARB());
4992 GL_EXTCALL(glAttachObjectARB(program_id
, vshader_id
));
4993 GL_EXTCALL(glAttachObjectARB(program_id
, pshader_id
));
4994 GL_EXTCALL(glLinkProgramARB(program_id
));
4996 shader_glsl_validate_link(gl_info
, program_id
);
4998 /* Once linked we can mark the shaders for deletion. They will be deleted once the program
5001 GL_EXTCALL(glDeleteObjectARB(vshader_id
));
5002 GL_EXTCALL(glDeleteObjectARB(pshader_id
));
5006 /* GL locking is done by the caller */
5007 static void shader_glsl_select(const struct wined3d_context
*context
, enum wined3d_shader_mode vertex_mode
,
5008 enum wined3d_shader_mode fragment_mode
)
5010 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
5011 struct wined3d_device
*device
= context
->swapchain
->device
;
5012 struct shader_glsl_priv
*priv
= device
->shader_priv
;
5013 GLhandleARB program_id
= 0;
5014 GLenum old_vertex_color_clamp
, current_vertex_color_clamp
;
5015 BOOL useVS
= vertex_mode
== WINED3D_SHADER_MODE_SHADER
;
5016 BOOL usePS
= fragment_mode
== WINED3D_SHADER_MODE_SHADER
;
5018 old_vertex_color_clamp
= priv
->glsl_program
? priv
->glsl_program
->vertex_color_clamp
: GL_FIXED_ONLY_ARB
;
5020 priv
->fragment_pipe
->enable_extension(gl_info
, fragment_mode
== WINED3D_SHADER_MODE_FFP
);
5021 if (useVS
|| usePS
) set_glsl_shader_program(context
, device
, usePS
, useVS
);
5022 else priv
->glsl_program
= NULL
;
5024 current_vertex_color_clamp
= priv
->glsl_program
? priv
->glsl_program
->vertex_color_clamp
: GL_FIXED_ONLY_ARB
;
5026 if (old_vertex_color_clamp
!= current_vertex_color_clamp
)
5028 if (gl_info
->supported
[ARB_COLOR_BUFFER_FLOAT
])
5030 GL_EXTCALL(glClampColorARB(GL_CLAMP_VERTEX_COLOR_ARB
, current_vertex_color_clamp
));
5031 checkGLcall("glClampColorARB");
5035 FIXME("vertex color clamp needs to be changed, but extension not supported.\n");
5039 program_id
= priv
->glsl_program
? priv
->glsl_program
->programId
: 0;
5040 if (program_id
) TRACE("Using GLSL program %u\n", program_id
);
5041 GL_EXTCALL(glUseProgramObjectARB(program_id
));
5042 checkGLcall("glUseProgramObjectARB");
5044 /* In case that NP2 texcoord fixup data is found for the selected program, trigger a reload of the
5045 * constants. This has to be done because it can't be guaranteed that sampler() (from state.c) is
5046 * called between selecting the shader and using it, which results in wrong fixup for some frames. */
5047 if (priv
->glsl_program
&& priv
->glsl_program
->np2Fixup_info
)
5049 shader_glsl_load_np2fixup_constants(priv
, gl_info
, &device
->stateBlock
->state
);
5053 /* GL locking is done by the caller */
5054 static void shader_glsl_select_depth_blt(void *shader_priv
, const struct wined3d_gl_info
*gl_info
,
5055 enum tex_types tex_type
, const SIZE
*ds_mask_size
)
5057 BOOL masked
= ds_mask_size
->cx
&& ds_mask_size
->cy
;
5058 struct shader_glsl_priv
*priv
= shader_priv
;
5059 GLhandleARB
*blt_program
;
5062 blt_program
= masked
? &priv
->depth_blt_program_masked
[tex_type
] : &priv
->depth_blt_program_full
[tex_type
];
5065 *blt_program
= create_glsl_blt_shader(gl_info
, tex_type
, masked
);
5066 loc
= GL_EXTCALL(glGetUniformLocationARB(*blt_program
, "sampler"));
5067 GL_EXTCALL(glUseProgramObjectARB(*blt_program
));
5068 GL_EXTCALL(glUniform1iARB(loc
, 0));
5072 GL_EXTCALL(glUseProgramObjectARB(*blt_program
));
5077 loc
= GL_EXTCALL(glGetUniformLocationARB(*blt_program
, "mask"));
5078 GL_EXTCALL(glUniform4fARB(loc
, 0.0f
, 0.0f
, (float)ds_mask_size
->cx
, (float)ds_mask_size
->cy
));
5082 /* GL locking is done by the caller */
5083 static void shader_glsl_deselect_depth_blt(void *shader_priv
, const struct wined3d_gl_info
*gl_info
)
5085 struct shader_glsl_priv
*priv
= shader_priv
;
5086 GLhandleARB program_id
;
5088 program_id
= priv
->glsl_program
? priv
->glsl_program
->programId
: 0;
5089 if (program_id
) TRACE("Using GLSL program %u\n", program_id
);
5091 GL_EXTCALL(glUseProgramObjectARB(program_id
));
5092 checkGLcall("glUseProgramObjectARB");
5095 static void shader_glsl_destroy(struct wined3d_shader
*shader
)
5097 struct glsl_shader_private
*shader_data
= shader
->backend_data
;
5098 struct wined3d_device
*device
= shader
->device
;
5099 struct shader_glsl_priv
*priv
= device
->shader_priv
;
5100 const struct wined3d_gl_info
*gl_info
;
5101 const struct list
*linked_programs
;
5102 struct wined3d_context
*context
;
5104 if (!shader_data
|| !shader_data
->num_gl_shaders
)
5106 HeapFree(GetProcessHeap(), 0, shader_data
);
5107 shader
->backend_data
= NULL
;
5111 context
= context_acquire(device
, NULL
);
5112 gl_info
= context
->gl_info
;
5114 TRACE("Deleting linked programs.\n");
5115 linked_programs
= &shader
->linked_programs
;
5116 if (linked_programs
->next
)
5118 struct glsl_shader_prog_link
*entry
, *entry2
;
5123 switch (shader
->reg_maps
.shader_version
.type
)
5125 case WINED3D_SHADER_TYPE_PIXEL
:
5127 struct glsl_ps_compiled_shader
*gl_shaders
= shader_data
->gl_shaders
.ps
;
5129 LIST_FOR_EACH_ENTRY_SAFE(entry
, entry2
, linked_programs
,
5130 struct glsl_shader_prog_link
, pshader_entry
)
5132 delete_glsl_program_entry(priv
, gl_info
, entry
);
5135 for (i
= 0; i
< shader_data
->num_gl_shaders
; ++i
)
5137 TRACE("Deleting pixel shader %u.\n", gl_shaders
[i
].prgId
);
5138 if (priv
->glsl_program
&& priv
->glsl_program
->ps_id
== gl_shaders
[i
].prgId
)
5139 shader_glsl_select(context
, WINED3D_SHADER_MODE_NONE
, WINED3D_SHADER_MODE_NONE
);
5140 GL_EXTCALL(glDeleteObjectARB(gl_shaders
[i
].prgId
));
5141 checkGLcall("glDeleteObjectARB");
5143 HeapFree(GetProcessHeap(), 0, shader_data
->gl_shaders
.ps
);
5148 case WINED3D_SHADER_TYPE_VERTEX
:
5150 struct glsl_vs_compiled_shader
*gl_shaders
= shader_data
->gl_shaders
.vs
;
5152 LIST_FOR_EACH_ENTRY_SAFE(entry
, entry2
, linked_programs
,
5153 struct glsl_shader_prog_link
, vshader_entry
)
5155 delete_glsl_program_entry(priv
, gl_info
, entry
);
5158 for (i
= 0; i
< shader_data
->num_gl_shaders
; ++i
)
5160 TRACE("Deleting vertex shader %u.\n", gl_shaders
[i
].prgId
);
5161 if (priv
->glsl_program
&& priv
->glsl_program
->vs_id
== gl_shaders
[i
].prgId
)
5162 shader_glsl_select(context
, WINED3D_SHADER_MODE_NONE
, WINED3D_SHADER_MODE_NONE
);
5163 GL_EXTCALL(glDeleteObjectARB(gl_shaders
[i
].prgId
));
5164 checkGLcall("glDeleteObjectARB");
5166 HeapFree(GetProcessHeap(), 0, shader_data
->gl_shaders
.vs
);
5172 ERR("Unhandled shader type %#x.\n", shader
->reg_maps
.shader_version
.type
);
5179 HeapFree(GetProcessHeap(), 0, shader
->backend_data
);
5180 shader
->backend_data
= NULL
;
5182 context_release(context
);
5185 static int glsl_program_key_compare(const void *key
, const struct wine_rb_entry
*entry
)
5187 const struct glsl_program_key
*k
= key
;
5188 const struct glsl_shader_prog_link
*prog
= WINE_RB_ENTRY_VALUE(entry
,
5189 const struct glsl_shader_prog_link
, program_lookup_entry
);
5191 if (k
->vs_id
> prog
->vs_id
) return 1;
5192 else if (k
->vs_id
< prog
->vs_id
) return -1;
5194 if (k
->ps_id
> prog
->ps_id
) return 1;
5195 else if (k
->ps_id
< prog
->ps_id
) return -1;
5200 static BOOL
constant_heap_init(struct constant_heap
*heap
, unsigned int constant_count
)
5202 SIZE_T size
= (constant_count
+ 1) * sizeof(*heap
->entries
) + constant_count
* sizeof(*heap
->positions
);
5203 void *mem
= HeapAlloc(GetProcessHeap(), 0, size
);
5207 ERR("Failed to allocate memory\n");
5211 heap
->entries
= mem
;
5212 heap
->entries
[1].version
= 0;
5213 heap
->positions
= (unsigned int *)(heap
->entries
+ constant_count
+ 1);
5219 static void constant_heap_free(struct constant_heap
*heap
)
5221 HeapFree(GetProcessHeap(), 0, heap
->entries
);
5224 static const struct wine_rb_functions wined3d_glsl_program_rb_functions
=
5229 glsl_program_key_compare
,
5232 static HRESULT
shader_glsl_alloc(struct wined3d_device
*device
, const struct fragment_pipeline
*fragment_pipe
)
5234 const struct wined3d_gl_info
*gl_info
= &device
->adapter
->gl_info
;
5235 struct shader_glsl_priv
*priv
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(struct shader_glsl_priv
));
5236 SIZE_T stack_size
= wined3d_log2i(max(gl_info
->limits
.glsl_vs_float_constants
,
5237 gl_info
->limits
.glsl_ps_float_constants
)) + 1;
5238 void *fragment_priv
;
5240 if (!(fragment_priv
= fragment_pipe
->alloc_private(&glsl_shader_backend
, priv
)))
5242 ERR("Failed to initialize fragment pipe.\n");
5243 HeapFree(GetProcessHeap(), 0, priv
);
5247 if (!shader_buffer_init(&priv
->shader_buffer
))
5249 ERR("Failed to initialize shader buffer.\n");
5253 priv
->stack
= HeapAlloc(GetProcessHeap(), 0, stack_size
* sizeof(*priv
->stack
));
5256 ERR("Failed to allocate memory.\n");
5260 if (!constant_heap_init(&priv
->vconst_heap
, gl_info
->limits
.glsl_vs_float_constants
))
5262 ERR("Failed to initialize vertex shader constant heap\n");
5266 if (!constant_heap_init(&priv
->pconst_heap
, gl_info
->limits
.glsl_ps_float_constants
))
5268 ERR("Failed to initialize pixel shader constant heap\n");
5272 if (wine_rb_init(&priv
->program_lookup
, &wined3d_glsl_program_rb_functions
) == -1)
5274 ERR("Failed to initialize rbtree.\n");
5278 priv
->next_constant_version
= 1;
5279 device
->fragment_priv
= fragment_priv
;
5280 priv
->fragment_pipe
= fragment_pipe
;
5282 device
->shader_priv
= priv
;
5286 constant_heap_free(&priv
->pconst_heap
);
5287 constant_heap_free(&priv
->vconst_heap
);
5288 HeapFree(GetProcessHeap(), 0, priv
->stack
);
5289 shader_buffer_free(&priv
->shader_buffer
);
5290 fragment_pipe
->free_private(device
);
5291 HeapFree(GetProcessHeap(), 0, priv
);
5292 return E_OUTOFMEMORY
;
5295 /* Context activation is done by the caller. */
5296 static void shader_glsl_free(struct wined3d_device
*device
)
5298 const struct wined3d_gl_info
*gl_info
= &device
->adapter
->gl_info
;
5299 struct shader_glsl_priv
*priv
= device
->shader_priv
;
5303 for (i
= 0; i
< tex_type_count
; ++i
)
5305 if (priv
->depth_blt_program_full
[i
])
5307 GL_EXTCALL(glDeleteObjectARB(priv
->depth_blt_program_full
[i
]));
5309 if (priv
->depth_blt_program_masked
[i
])
5311 GL_EXTCALL(glDeleteObjectARB(priv
->depth_blt_program_masked
[i
]));
5316 wine_rb_destroy(&priv
->program_lookup
, NULL
, NULL
);
5317 constant_heap_free(&priv
->pconst_heap
);
5318 constant_heap_free(&priv
->vconst_heap
);
5319 HeapFree(GetProcessHeap(), 0, priv
->stack
);
5320 shader_buffer_free(&priv
->shader_buffer
);
5321 priv
->fragment_pipe
->free_private(device
);
5323 HeapFree(GetProcessHeap(), 0, device
->shader_priv
);
5324 device
->shader_priv
= NULL
;
5327 static void shader_glsl_context_destroyed(void *shader_priv
, const struct wined3d_context
*context
) {}
5329 static void shader_glsl_get_caps(const struct wined3d_gl_info
*gl_info
, struct shader_caps
*caps
)
5333 if (gl_info
->supported
[EXT_GPU_SHADER4
] && gl_info
->supported
[ARB_SHADER_BIT_ENCODING
]
5334 && gl_info
->supported
[ARB_GEOMETRY_SHADER4
] && gl_info
->glsl_version
>= MAKEDWORD_VERSION(1, 50))
5336 /* ARB_shader_texture_lod or EXT_gpu_shader4 is required for the SM3
5337 * texldd and texldl instructions. */
5338 else if (gl_info
->supported
[ARB_SHADER_TEXTURE_LOD
] || gl_info
->supported
[EXT_GPU_SHADER4
])
5342 TRACE("Shader model %u.\n", shader_model
);
5344 caps
->vs_version
= shader_model
;
5345 caps
->gs_version
= shader_model
;
5346 caps
->ps_version
= shader_model
;
5348 caps
->vs_uniform_count
= gl_info
->limits
.glsl_vs_float_constants
;
5349 caps
->ps_uniform_count
= gl_info
->limits
.glsl_ps_float_constants
;
5351 /* FIXME: The following line is card dependent. -8.0 to 8.0 is the
5352 * Direct3D minimum requirement.
5354 * Both GL_ARB_fragment_program and GLSL require a "maximum representable magnitude"
5355 * of colors to be 2^10, and 2^32 for other floats. Should we use 1024 here?
5357 * The problem is that the refrast clamps temporary results in the shader to
5358 * [-MaxValue;+MaxValue]. If the card's max value is bigger than the one we advertize here,
5359 * then applications may miss the clamping behavior. On the other hand, if it is smaller,
5360 * the shader will generate incorrect results too. Unfortunately, GL deliberately doesn't
5361 * offer a way to query this.
5363 caps
->ps_1x_max_value
= 8.0;
5365 caps
->vs_clipping
= TRUE
;
5368 static BOOL
shader_glsl_color_fixup_supported(struct color_fixup_desc fixup
)
5370 if (TRACE_ON(d3d_shader
) && TRACE_ON(d3d
))
5372 TRACE("Checking support for fixup:\n");
5373 dump_color_fixup_desc(fixup
);
5376 /* We support everything except YUV conversions. */
5377 if (!is_complex_fixup(fixup
))
5383 TRACE("[FAILED]\n");
5387 static const SHADER_HANDLER shader_glsl_instruction_handler_table
[WINED3DSIH_TABLE_SIZE
] =
5389 /* WINED3DSIH_ABS */ shader_glsl_map2gl
,
5390 /* WINED3DSIH_ADD */ shader_glsl_binop
,
5391 /* WINED3DSIH_AND */ shader_glsl_binop
,
5392 /* WINED3DSIH_BEM */ shader_glsl_bem
,
5393 /* WINED3DSIH_BREAK */ shader_glsl_break
,
5394 /* WINED3DSIH_BREAKC */ shader_glsl_breakc
,
5395 /* WINED3DSIH_BREAKP */ shader_glsl_breakp
,
5396 /* WINED3DSIH_CALL */ shader_glsl_call
,
5397 /* WINED3DSIH_CALLNZ */ shader_glsl_callnz
,
5398 /* WINED3DSIH_CMP */ shader_glsl_conditional_move
,
5399 /* WINED3DSIH_CND */ shader_glsl_cnd
,
5400 /* WINED3DSIH_CRS */ shader_glsl_cross
,
5401 /* WINED3DSIH_CUT */ shader_glsl_cut
,
5402 /* WINED3DSIH_DCL */ shader_glsl_nop
,
5403 /* WINED3DSIH_DCL_CONSTANT_BUFFER */ shader_glsl_nop
,
5404 /* WINED3DSIH_DCL_INPUT_PRIMITIVE */ shader_glsl_nop
,
5405 /* WINED3DSIH_DCL_OUTPUT_TOPOLOGY */ shader_glsl_nop
,
5406 /* WINED3DSIH_DCL_VERTICES_OUT */ shader_glsl_nop
,
5407 /* WINED3DSIH_DEF */ shader_glsl_nop
,
5408 /* WINED3DSIH_DEFB */ shader_glsl_nop
,
5409 /* WINED3DSIH_DEFI */ shader_glsl_nop
,
5410 /* WINED3DSIH_DIV */ shader_glsl_binop
,
5411 /* WINED3DSIH_DP2ADD */ shader_glsl_dp2add
,
5412 /* WINED3DSIH_DP3 */ shader_glsl_dot
,
5413 /* WINED3DSIH_DP4 */ shader_glsl_dot
,
5414 /* WINED3DSIH_DST */ shader_glsl_dst
,
5415 /* WINED3DSIH_DSX */ shader_glsl_map2gl
,
5416 /* WINED3DSIH_DSY */ shader_glsl_map2gl
,
5417 /* WINED3DSIH_ELSE */ shader_glsl_else
,
5418 /* WINED3DSIH_EMIT */ shader_glsl_emit
,
5419 /* WINED3DSIH_ENDIF */ shader_glsl_end
,
5420 /* WINED3DSIH_ENDLOOP */ shader_glsl_end
,
5421 /* WINED3DSIH_ENDREP */ shader_glsl_end
,
5422 /* WINED3DSIH_EQ */ shader_glsl_relop
,
5423 /* WINED3DSIH_EXP */ shader_glsl_map2gl
,
5424 /* WINED3DSIH_EXPP */ shader_glsl_expp
,
5425 /* WINED3DSIH_FRC */ shader_glsl_map2gl
,
5426 /* WINED3DSIH_FTOI */ shader_glsl_to_int
,
5427 /* WINED3DSIH_GE */ shader_glsl_relop
,
5428 /* WINED3DSIH_IADD */ shader_glsl_binop
,
5429 /* WINED3DSIH_IEQ */ NULL
,
5430 /* WINED3DSIH_IF */ shader_glsl_if
,
5431 /* WINED3DSIH_IFC */ shader_glsl_ifc
,
5432 /* WINED3DSIH_IGE */ shader_glsl_relop
,
5433 /* WINED3DSIH_IMUL */ shader_glsl_imul
,
5434 /* WINED3DSIH_ITOF */ shader_glsl_to_float
,
5435 /* WINED3DSIH_LABEL */ shader_glsl_label
,
5436 /* WINED3DSIH_LD */ NULL
,
5437 /* WINED3DSIH_LIT */ shader_glsl_lit
,
5438 /* WINED3DSIH_LOG */ shader_glsl_log
,
5439 /* WINED3DSIH_LOGP */ shader_glsl_log
,
5440 /* WINED3DSIH_LOOP */ shader_glsl_loop
,
5441 /* WINED3DSIH_LRP */ shader_glsl_lrp
,
5442 /* WINED3DSIH_LT */ shader_glsl_relop
,
5443 /* WINED3DSIH_M3x2 */ shader_glsl_mnxn
,
5444 /* WINED3DSIH_M3x3 */ shader_glsl_mnxn
,
5445 /* WINED3DSIH_M3x4 */ shader_glsl_mnxn
,
5446 /* WINED3DSIH_M4x3 */ shader_glsl_mnxn
,
5447 /* WINED3DSIH_M4x4 */ shader_glsl_mnxn
,
5448 /* WINED3DSIH_MAD */ shader_glsl_mad
,
5449 /* WINED3DSIH_MAX */ shader_glsl_map2gl
,
5450 /* WINED3DSIH_MIN */ shader_glsl_map2gl
,
5451 /* WINED3DSIH_MOV */ shader_glsl_mov
,
5452 /* WINED3DSIH_MOVA */ shader_glsl_mov
,
5453 /* WINED3DSIH_MOVC */ shader_glsl_conditional_move
,
5454 /* WINED3DSIH_MUL */ shader_glsl_binop
,
5455 /* WINED3DSIH_NOP */ shader_glsl_nop
,
5456 /* WINED3DSIH_NRM */ shader_glsl_nrm
,
5457 /* WINED3DSIH_PHASE */ shader_glsl_nop
,
5458 /* WINED3DSIH_POW */ shader_glsl_pow
,
5459 /* WINED3DSIH_RCP */ shader_glsl_rcp
,
5460 /* WINED3DSIH_REP */ shader_glsl_rep
,
5461 /* WINED3DSIH_RET */ shader_glsl_ret
,
5462 /* WINED3DSIH_ROUND_NI */ shader_glsl_map2gl
,
5463 /* WINED3DSIH_RSQ */ shader_glsl_rsq
,
5464 /* WINED3DSIH_SAMPLE */ NULL
,
5465 /* WINED3DSIH_SAMPLE_GRAD */ NULL
,
5466 /* WINED3DSIH_SAMPLE_LOD */ NULL
,
5467 /* WINED3DSIH_SETP */ NULL
,
5468 /* WINED3DSIH_SGE */ shader_glsl_compare
,
5469 /* WINED3DSIH_SGN */ shader_glsl_sgn
,
5470 /* WINED3DSIH_SINCOS */ shader_glsl_sincos
,
5471 /* WINED3DSIH_SLT */ shader_glsl_compare
,
5472 /* WINED3DSIH_SQRT */ NULL
,
5473 /* WINED3DSIH_SUB */ shader_glsl_binop
,
5474 /* WINED3DSIH_TEX */ shader_glsl_tex
,
5475 /* WINED3DSIH_TEXBEM */ shader_glsl_texbem
,
5476 /* WINED3DSIH_TEXBEML */ shader_glsl_texbem
,
5477 /* WINED3DSIH_TEXCOORD */ shader_glsl_texcoord
,
5478 /* WINED3DSIH_TEXDEPTH */ shader_glsl_texdepth
,
5479 /* WINED3DSIH_TEXDP3 */ shader_glsl_texdp3
,
5480 /* WINED3DSIH_TEXDP3TEX */ shader_glsl_texdp3tex
,
5481 /* WINED3DSIH_TEXKILL */ shader_glsl_texkill
,
5482 /* WINED3DSIH_TEXLDD */ shader_glsl_texldd
,
5483 /* WINED3DSIH_TEXLDL */ shader_glsl_texldl
,
5484 /* WINED3DSIH_TEXM3x2DEPTH */ shader_glsl_texm3x2depth
,
5485 /* WINED3DSIH_TEXM3x2PAD */ shader_glsl_texm3x2pad
,
5486 /* WINED3DSIH_TEXM3x2TEX */ shader_glsl_texm3x2tex
,
5487 /* WINED3DSIH_TEXM3x3 */ shader_glsl_texm3x3
,
5488 /* WINED3DSIH_TEXM3x3DIFF */ NULL
,
5489 /* WINED3DSIH_TEXM3x3PAD */ shader_glsl_texm3x3pad
,
5490 /* WINED3DSIH_TEXM3x3SPEC */ shader_glsl_texm3x3spec
,
5491 /* WINED3DSIH_TEXM3x3TEX */ shader_glsl_texm3x3tex
,
5492 /* WINED3DSIH_TEXM3x3VSPEC */ shader_glsl_texm3x3vspec
,
5493 /* WINED3DSIH_TEXREG2AR */ shader_glsl_texreg2ar
,
5494 /* WINED3DSIH_TEXREG2GB */ shader_glsl_texreg2gb
,
5495 /* WINED3DSIH_TEXREG2RGB */ shader_glsl_texreg2rgb
,
5496 /* WINED3DSIH_UDIV */ shader_glsl_udiv
,
5497 /* WINED3DSIH_USHR */ shader_glsl_binop
,
5498 /* WINED3DSIH_UTOF */ shader_glsl_to_float
,
5499 /* WINED3DSIH_XOR */ shader_glsl_binop
,
5502 static void shader_glsl_handle_instruction(const struct wined3d_shader_instruction
*ins
) {
5503 SHADER_HANDLER hw_fct
;
5505 /* Select handler */
5506 hw_fct
= shader_glsl_instruction_handler_table
[ins
->handler_idx
];
5508 /* Unhandled opcode */
5511 FIXME("Backend can't handle opcode %#x\n", ins
->handler_idx
);
5516 shader_glsl_add_instruction_modifiers(ins
);
5519 static BOOL
shader_glsl_has_ffp_proj_control(void *shader_priv
)
5521 struct shader_glsl_priv
*priv
= shader_priv
;
5523 return priv
->fragment_pipe
->ffp_proj_control
;
5526 const struct wined3d_shader_backend_ops glsl_shader_backend
=
5528 shader_glsl_handle_instruction
,
5530 shader_glsl_select_depth_blt
,
5531 shader_glsl_deselect_depth_blt
,
5532 shader_glsl_update_float_vertex_constants
,
5533 shader_glsl_update_float_pixel_constants
,
5534 shader_glsl_load_constants
,
5535 shader_glsl_load_np2fixup_constants
,
5536 shader_glsl_destroy
,
5539 shader_glsl_context_destroyed
,
5540 shader_glsl_get_caps
,
5541 shader_glsl_color_fixup_supported
,
5542 shader_glsl_has_ffp_proj_control
,