2 * GLSL pixel and vertex shader implementation
4 * Copyright 2006 Jason Green
5 * Copyright 2006-2007 Henri Verbeet
6 * Copyright 2007-2008 Stefan Dösinger for CodeWeavers
7 * Copyright 2009 Henri Verbeet for CodeWeavers
9 * This library is free software; you can redistribute it and/or
10 * modify it under the terms of the GNU Lesser General Public
11 * License as published by the Free Software Foundation; either
12 * version 2.1 of the License, or (at your option) any later version.
14 * This library is distributed in the hope that it will be useful,
15 * but WITHOUT ANY WARRANTY; without even the implied warranty of
16 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
17 * Lesser General Public License for more details.
19 * You should have received a copy of the GNU Lesser General Public
20 * License along with this library; if not, write to the Free Software
21 * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301, USA
25 * D3D shader asm has swizzles on source parameters, and write masks for
26 * destination parameters. GLSL uses swizzles for both. The result of this is
27 * that for example "mov dst.xw, src.zyxw" becomes "dst.xw = src.zw" in GLSL.
28 * Ie, to generate a proper GLSL source swizzle, we need to take the D3D write
29 * mask for the destination parameter into account.
35 #include "wined3d_private.h"
37 WINE_DEFAULT_DEBUG_CHANNEL(d3d_shader
);
38 WINE_DECLARE_DEBUG_CHANNEL(d3d_constants
);
39 WINE_DECLARE_DEBUG_CHANNEL(d3d_caps
);
40 WINE_DECLARE_DEBUG_CHANNEL(d3d
);
42 #define WINED3D_GLSL_SAMPLE_PROJECTED 0x1
43 #define WINED3D_GLSL_SAMPLE_RECT 0x2
44 #define WINED3D_GLSL_SAMPLE_LOD 0x4
45 #define WINED3D_GLSL_SAMPLE_GRAD 0x8
60 } glsl_sample_function_t
;
64 HEAP_NODE_TRAVERSE_LEFT
,
65 HEAP_NODE_TRAVERSE_RIGHT
,
77 struct constant_entry
*entries
;
78 unsigned int *positions
;
82 /* GLSL shader private data */
83 struct shader_glsl_priv
{
84 struct wined3d_shader_buffer shader_buffer
;
85 struct wine_rb_tree program_lookup
;
86 struct glsl_shader_prog_link
*glsl_program
;
87 struct constant_heap vconst_heap
;
88 struct constant_heap pconst_heap
;
90 GLhandleARB depth_blt_program_full
[tex_type_count
];
91 GLhandleARB depth_blt_program_masked
[tex_type_count
];
92 UINT next_constant_version
;
95 /* Struct to maintain data about a linked GLSL program */
96 struct glsl_shader_prog_link
{
97 struct wine_rb_entry program_lookup_entry
;
98 struct list vshader_entry
;
99 struct list pshader_entry
;
100 GLhandleARB programId
;
101 GLint
*vuniformF_locations
;
102 GLint
*puniformF_locations
;
103 GLint vuniformI_locations
[MAX_CONST_I
];
104 GLint puniformI_locations
[MAX_CONST_I
];
105 GLint posFixup_location
;
106 GLint np2Fixup_location
;
107 GLint bumpenvmat_location
[MAX_TEXTURES
];
108 GLint luminancescale_location
[MAX_TEXTURES
];
109 GLint luminanceoffset_location
[MAX_TEXTURES
];
110 GLint ycorrection_location
;
111 GLenum vertex_color_clamp
;
112 IWineD3DVertexShader
*vshader
;
113 IWineD3DPixelShader
*pshader
;
114 struct vs_compile_args vs_args
;
115 struct ps_compile_args ps_args
;
116 UINT constant_version
;
117 const struct ps_np2fixup_info
*np2Fixup_info
;
121 IWineD3DVertexShader
*vshader
;
122 IWineD3DPixelShader
*pshader
;
123 struct ps_compile_args ps_args
;
124 struct vs_compile_args vs_args
;
125 } glsl_program_key_t
;
127 struct shader_glsl_ctx_priv
{
128 const struct vs_compile_args
*cur_vs_args
;
129 const struct ps_compile_args
*cur_ps_args
;
130 struct ps_np2fixup_info
*cur_np2fixup_info
;
133 struct glsl_ps_compiled_shader
135 struct ps_compile_args args
;
136 struct ps_np2fixup_info np2fixup
;
140 struct glsl_pshader_private
142 struct glsl_ps_compiled_shader
*gl_shaders
;
143 UINT num_gl_shaders
, shader_array_size
;
146 struct glsl_vs_compiled_shader
148 struct vs_compile_args args
;
152 struct glsl_vshader_private
154 struct glsl_vs_compiled_shader
*gl_shaders
;
155 UINT num_gl_shaders
, shader_array_size
;
158 static const char *debug_gl_shader_type(GLenum type
)
162 #define WINED3D_TO_STR(u) case u: return #u
163 WINED3D_TO_STR(GL_VERTEX_SHADER_ARB
);
164 WINED3D_TO_STR(GL_GEOMETRY_SHADER_ARB
);
165 WINED3D_TO_STR(GL_FRAGMENT_SHADER_ARB
);
166 #undef WINED3D_TO_STR
168 return wine_dbg_sprintf("UNKNOWN(%#x)", type
);
172 /* Extract a line from the info log.
173 * Note that this modifies the source string. */
174 static char *get_info_log_line(char **ptr
)
179 if (!(q
= strstr(p
, "\n")))
181 if (!*p
) return NULL
;
191 /** Prints the GLSL info log which will contain error messages if they exist */
192 /* GL locking is done by the caller */
193 static void print_glsl_info_log(const struct wined3d_gl_info
*gl_info
, GLhandleARB obj
)
195 int infologLength
= 0;
200 static const char * const spam
[] =
202 "Vertex shader was successfully compiled to run on hardware.\n", /* fglrx */
203 "Fragment shader was successfully compiled to run on hardware.\n", /* fglrx, with \n */
204 "Fragment shader was successfully compiled to run on hardware.", /* fglrx, no \n */
205 "Fragment shader(s) linked, vertex shader(s) linked. \n ", /* fglrx, with \n */
206 "Fragment shader(s) linked, vertex shader(s) linked. \n", /* fglrx, with \n */
207 "Fragment shader(s) linked, vertex shader(s) linked.", /* fglrx, no \n */
208 "Vertex shader(s) linked, no fragment shader(s) defined. \n ", /* fglrx, with \n */
209 "Vertex shader(s) linked, no fragment shader(s) defined. \n", /* fglrx, with \n */
210 "Vertex shader(s) linked, no fragment shader(s) defined.", /* fglrx, no \n */
211 "Fragment shader(s) linked, no vertex shader(s) defined. \n ", /* fglrx, with \n */
212 "Fragment shader(s) linked, no vertex shader(s) defined. \n", /* fglrx, with \n */
213 "Fragment shader(s) linked, no vertex shader(s) defined.", /* fglrx, no \n */
216 if (!TRACE_ON(d3d_shader
) && !FIXME_ON(d3d_shader
)) return;
218 GL_EXTCALL(glGetObjectParameterivARB(obj
,
219 GL_OBJECT_INFO_LOG_LENGTH_ARB
,
222 /* A size of 1 is just a null-terminated string, so the log should be bigger than
223 * that if there are errors. */
224 if (infologLength
> 1)
228 /* Fglrx doesn't terminate the string properly, but it tells us the proper length.
229 * So use HEAP_ZERO_MEMORY to avoid uninitialized bytes
231 infoLog
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, infologLength
);
232 GL_EXTCALL(glGetInfoLogARB(obj
, infologLength
, NULL
, infoLog
));
235 for (i
= 0; i
< sizeof(spam
) / sizeof(*spam
); ++i
)
237 if (!strcmp(infoLog
, spam
[i
]))
247 TRACE("Spam received from GLSL shader #%u:\n", obj
);
248 while ((line
= get_info_log_line(&ptr
))) TRACE(" %s\n", line
);
252 FIXME("Error received from GLSL shader #%u:\n", obj
);
253 while ((line
= get_info_log_line(&ptr
))) FIXME(" %s\n", line
);
255 HeapFree(GetProcessHeap(), 0, infoLog
);
259 /* GL locking is done by the caller. */
260 static void shader_glsl_compile(const struct wined3d_gl_info
*gl_info
, GLhandleARB shader
, const char *src
)
262 TRACE("Compiling shader object %u.\n", shader
);
263 GL_EXTCALL(glShaderSourceARB(shader
, 1, &src
, NULL
));
264 checkGLcall("glShaderSourceARB");
265 GL_EXTCALL(glCompileShaderARB(shader
));
266 checkGLcall("glCompileShaderARB");
267 print_glsl_info_log(gl_info
, shader
);
270 /* GL locking is done by the caller. */
271 static void shader_glsl_dump_program_source(const struct wined3d_gl_info
*gl_info
, GLhandleARB program
)
273 GLint i
, object_count
, source_size
= -1;
274 GLhandleARB
*objects
;
277 GL_EXTCALL(glGetObjectParameterivARB(program
, GL_OBJECT_ATTACHED_OBJECTS_ARB
, &object_count
));
278 objects
= HeapAlloc(GetProcessHeap(), 0, object_count
* sizeof(*objects
));
281 ERR("Failed to allocate object array memory.\n");
285 GL_EXTCALL(glGetAttachedObjectsARB(program
, object_count
, NULL
, objects
));
286 for (i
= 0; i
< object_count
; ++i
)
291 GL_EXTCALL(glGetObjectParameterivARB(objects
[i
], GL_OBJECT_SHADER_SOURCE_LENGTH_ARB
, &tmp
));
293 if (source_size
< tmp
)
295 HeapFree(GetProcessHeap(), 0, source
);
297 source
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, tmp
);
300 ERR("Failed to allocate %d bytes for shader source.\n", tmp
);
301 HeapFree(GetProcessHeap(), 0, objects
);
307 FIXME("Object %u:\n", objects
[i
]);
308 GL_EXTCALL(glGetObjectParameterivARB(objects
[i
], GL_OBJECT_SUBTYPE_ARB
, &tmp
));
309 FIXME(" GL_OBJECT_SUBTYPE_ARB: %s.\n", debug_gl_shader_type(tmp
));
310 GL_EXTCALL(glGetObjectParameterivARB(objects
[i
], GL_OBJECT_COMPILE_STATUS_ARB
, &tmp
));
311 FIXME(" GL_OBJECT_COMPILE_STATUS_ARB: %d.\n", tmp
);
315 GL_EXTCALL(glGetShaderSourceARB(objects
[i
], source_size
, NULL
, source
));
316 while ((line
= get_info_log_line(&ptr
))) FIXME(" %s\n", line
);
320 HeapFree(GetProcessHeap(), 0, source
);
321 HeapFree(GetProcessHeap(), 0, objects
);
324 /* GL locking is done by the caller. */
325 static void shader_glsl_validate_link(const struct wined3d_gl_info
*gl_info
, GLhandleARB program
)
329 if (!TRACE_ON(d3d_shader
) && !FIXME_ON(d3d_shader
)) return;
331 GL_EXTCALL(glGetObjectParameterivARB(program
, GL_OBJECT_TYPE_ARB
, &tmp
));
332 if (tmp
== GL_PROGRAM_OBJECT_ARB
)
334 GL_EXTCALL(glGetObjectParameterivARB(program
, GL_OBJECT_LINK_STATUS_ARB
, &tmp
));
337 FIXME("Program %u link status invalid.\n", program
);
338 shader_glsl_dump_program_source(gl_info
, program
);
342 print_glsl_info_log(gl_info
, program
);
346 * Loads (pixel shader) samplers
348 /* GL locking is done by the caller */
349 static void shader_glsl_load_psamplers(const struct wined3d_gl_info
*gl_info
,
350 DWORD
*tex_unit_map
, GLhandleARB programId
)
354 char sampler_name
[20];
356 for (i
= 0; i
< MAX_FRAGMENT_SAMPLERS
; ++i
) {
357 snprintf(sampler_name
, sizeof(sampler_name
), "Psampler%d", i
);
358 name_loc
= GL_EXTCALL(glGetUniformLocationARB(programId
, sampler_name
));
359 if (name_loc
!= -1) {
360 DWORD mapped_unit
= tex_unit_map
[i
];
361 if (mapped_unit
!= WINED3D_UNMAPPED_STAGE
&& mapped_unit
< gl_info
->limits
.fragment_samplers
)
363 TRACE("Loading %s for texture %d\n", sampler_name
, mapped_unit
);
364 GL_EXTCALL(glUniform1iARB(name_loc
, mapped_unit
));
365 checkGLcall("glUniform1iARB");
367 ERR("Trying to load sampler %s on unsupported unit %d\n", sampler_name
, mapped_unit
);
373 /* GL locking is done by the caller */
374 static void shader_glsl_load_vsamplers(const struct wined3d_gl_info
*gl_info
,
375 DWORD
*tex_unit_map
, GLhandleARB programId
)
378 char sampler_name
[20];
381 for (i
= 0; i
< MAX_VERTEX_SAMPLERS
; ++i
) {
382 snprintf(sampler_name
, sizeof(sampler_name
), "Vsampler%d", i
);
383 name_loc
= GL_EXTCALL(glGetUniformLocationARB(programId
, sampler_name
));
384 if (name_loc
!= -1) {
385 DWORD mapped_unit
= tex_unit_map
[MAX_FRAGMENT_SAMPLERS
+ i
];
386 if (mapped_unit
!= WINED3D_UNMAPPED_STAGE
&& mapped_unit
< gl_info
->limits
.combined_samplers
)
388 TRACE("Loading %s for texture %d\n", sampler_name
, mapped_unit
);
389 GL_EXTCALL(glUniform1iARB(name_loc
, mapped_unit
));
390 checkGLcall("glUniform1iARB");
392 ERR("Trying to load sampler %s on unsupported unit %d\n", sampler_name
, mapped_unit
);
398 /* GL locking is done by the caller */
399 static inline void walk_constant_heap(const struct wined3d_gl_info
*gl_info
, const float *constants
,
400 const GLint
*constant_locations
, const struct constant_heap
*heap
, unsigned char *stack
, DWORD version
)
403 unsigned int heap_idx
= 1;
406 if (heap
->entries
[heap_idx
].version
<= version
) return;
408 idx
= heap
->entries
[heap_idx
].idx
;
409 if (constant_locations
[idx
] != -1) GL_EXTCALL(glUniform4fvARB(constant_locations
[idx
], 1, &constants
[idx
* 4]));
410 stack
[stack_idx
] = HEAP_NODE_TRAVERSE_LEFT
;
412 while (stack_idx
>= 0)
414 /* Note that we fall through to the next case statement. */
415 switch(stack
[stack_idx
])
417 case HEAP_NODE_TRAVERSE_LEFT
:
419 unsigned int left_idx
= heap_idx
<< 1;
420 if (left_idx
< heap
->size
&& heap
->entries
[left_idx
].version
> version
)
423 idx
= heap
->entries
[heap_idx
].idx
;
424 if (constant_locations
[idx
] != -1)
425 GL_EXTCALL(glUniform4fvARB(constant_locations
[idx
], 1, &constants
[idx
* 4]));
427 stack
[stack_idx
++] = HEAP_NODE_TRAVERSE_RIGHT
;
428 stack
[stack_idx
] = HEAP_NODE_TRAVERSE_LEFT
;
433 case HEAP_NODE_TRAVERSE_RIGHT
:
435 unsigned int right_idx
= (heap_idx
<< 1) + 1;
436 if (right_idx
< heap
->size
&& heap
->entries
[right_idx
].version
> version
)
438 heap_idx
= right_idx
;
439 idx
= heap
->entries
[heap_idx
].idx
;
440 if (constant_locations
[idx
] != -1)
441 GL_EXTCALL(glUniform4fvARB(constant_locations
[idx
], 1, &constants
[idx
* 4]));
443 stack
[stack_idx
++] = HEAP_NODE_POP
;
444 stack
[stack_idx
] = HEAP_NODE_TRAVERSE_LEFT
;
457 checkGLcall("walk_constant_heap()");
460 /* GL locking is done by the caller */
461 static inline void apply_clamped_constant(const struct wined3d_gl_info
*gl_info
, GLint location
, const GLfloat
*data
)
463 GLfloat clamped_constant
[4];
465 if (location
== -1) return;
467 clamped_constant
[0] = data
[0] < -1.0f
? -1.0f
: data
[0] > 1.0f
? 1.0f
: data
[0];
468 clamped_constant
[1] = data
[1] < -1.0f
? -1.0f
: data
[1] > 1.0f
? 1.0f
: data
[1];
469 clamped_constant
[2] = data
[2] < -1.0f
? -1.0f
: data
[2] > 1.0f
? 1.0f
: data
[2];
470 clamped_constant
[3] = data
[3] < -1.0f
? -1.0f
: data
[3] > 1.0f
? 1.0f
: data
[3];
472 GL_EXTCALL(glUniform4fvARB(location
, 1, clamped_constant
));
475 /* GL locking is done by the caller */
476 static inline void walk_constant_heap_clamped(const struct wined3d_gl_info
*gl_info
, const float *constants
,
477 const GLint
*constant_locations
, const struct constant_heap
*heap
, unsigned char *stack
, DWORD version
)
480 unsigned int heap_idx
= 1;
483 if (heap
->entries
[heap_idx
].version
<= version
) return;
485 idx
= heap
->entries
[heap_idx
].idx
;
486 apply_clamped_constant(gl_info
, constant_locations
[idx
], &constants
[idx
* 4]);
487 stack
[stack_idx
] = HEAP_NODE_TRAVERSE_LEFT
;
489 while (stack_idx
>= 0)
491 /* Note that we fall through to the next case statement. */
492 switch(stack
[stack_idx
])
494 case HEAP_NODE_TRAVERSE_LEFT
:
496 unsigned int left_idx
= heap_idx
<< 1;
497 if (left_idx
< heap
->size
&& heap
->entries
[left_idx
].version
> version
)
500 idx
= heap
->entries
[heap_idx
].idx
;
501 apply_clamped_constant(gl_info
, constant_locations
[idx
], &constants
[idx
* 4]);
503 stack
[stack_idx
++] = HEAP_NODE_TRAVERSE_RIGHT
;
504 stack
[stack_idx
] = HEAP_NODE_TRAVERSE_LEFT
;
509 case HEAP_NODE_TRAVERSE_RIGHT
:
511 unsigned int right_idx
= (heap_idx
<< 1) + 1;
512 if (right_idx
< heap
->size
&& heap
->entries
[right_idx
].version
> version
)
514 heap_idx
= right_idx
;
515 idx
= heap
->entries
[heap_idx
].idx
;
516 apply_clamped_constant(gl_info
, constant_locations
[idx
], &constants
[idx
* 4]);
518 stack
[stack_idx
++] = HEAP_NODE_POP
;
519 stack
[stack_idx
] = HEAP_NODE_TRAVERSE_LEFT
;
532 checkGLcall("walk_constant_heap_clamped()");
535 /* Loads floating point constants (aka uniforms) into the currently set GLSL program. */
536 /* GL locking is done by the caller */
537 static void shader_glsl_load_constantsF(IWineD3DBaseShaderImpl
*This
, const struct wined3d_gl_info
*gl_info
,
538 const float *constants
, const GLint
*constant_locations
, const struct constant_heap
*heap
,
539 unsigned char *stack
, UINT version
)
541 const local_constant
*lconst
;
543 /* 1.X pshaders have the constants clamped to [-1;1] implicitly. */
544 if (This
->baseShader
.reg_maps
.shader_version
.major
== 1
545 && shader_is_pshader_version(This
->baseShader
.reg_maps
.shader_version
.type
))
546 walk_constant_heap_clamped(gl_info
, constants
, constant_locations
, heap
, stack
, version
);
548 walk_constant_heap(gl_info
, constants
, constant_locations
, heap
, stack
, version
);
550 if (!This
->baseShader
.load_local_constsF
)
552 TRACE("No need to load local float constants for this shader\n");
556 /* Immediate constants are clamped to [-1;1] at shader creation time if needed */
557 LIST_FOR_EACH_ENTRY(lconst
, &This
->baseShader
.constantsF
, local_constant
, entry
)
559 GLint location
= constant_locations
[lconst
->idx
];
560 /* We found this uniform name in the program - go ahead and send the data */
561 if (location
!= -1) GL_EXTCALL(glUniform4fvARB(location
, 1, (const GLfloat
*)lconst
->value
));
563 checkGLcall("glUniform4fvARB()");
566 /* Loads integer constants (aka uniforms) into the currently set GLSL program. */
567 /* GL locking is done by the caller */
568 static void shader_glsl_load_constantsI(IWineD3DBaseShaderImpl
*This
, const struct wined3d_gl_info
*gl_info
,
569 const GLint locations
[MAX_CONST_I
], const int *constants
, WORD constants_set
)
574 for (i
= 0; constants_set
; constants_set
>>= 1, ++i
)
576 if (!(constants_set
& 1)) continue;
578 TRACE_(d3d_constants
)("Loading constants %u: %i, %i, %i, %i\n",
579 i
, constants
[i
*4], constants
[i
*4+1], constants
[i
*4+2], constants
[i
*4+3]);
581 /* We found this uniform name in the program - go ahead and send the data */
582 GL_EXTCALL(glUniform4ivARB(locations
[i
], 1, &constants
[i
*4]));
583 checkGLcall("glUniform4ivARB");
586 /* Load immediate constants */
587 ptr
= list_head(&This
->baseShader
.constantsI
);
589 const struct local_constant
*lconst
= LIST_ENTRY(ptr
, const struct local_constant
, entry
);
590 unsigned int idx
= lconst
->idx
;
591 const GLint
*values
= (const GLint
*)lconst
->value
;
593 TRACE_(d3d_constants
)("Loading local constants %i: %i, %i, %i, %i\n", idx
,
594 values
[0], values
[1], values
[2], values
[3]);
596 /* We found this uniform name in the program - go ahead and send the data */
597 GL_EXTCALL(glUniform4ivARB(locations
[idx
], 1, values
));
598 checkGLcall("glUniform4ivARB");
599 ptr
= list_next(&This
->baseShader
.constantsI
, ptr
);
603 /* Loads boolean constants (aka uniforms) into the currently set GLSL program. */
604 /* GL locking is done by the caller */
605 static void shader_glsl_load_constantsB(IWineD3DBaseShaderImpl
*This
, const struct wined3d_gl_info
*gl_info
,
606 GLhandleARB programId
, const BOOL
*constants
, WORD constants_set
)
614 switch (This
->baseShader
.reg_maps
.shader_version
.type
)
616 case WINED3D_SHADER_TYPE_VERTEX
:
620 case WINED3D_SHADER_TYPE_GEOMETRY
:
624 case WINED3D_SHADER_TYPE_PIXEL
:
629 FIXME("Unknown shader type %#x.\n",
630 This
->baseShader
.reg_maps
.shader_version
.type
);
635 /* TODO: Benchmark and see if it would be beneficial to store the
636 * locations of the constants to avoid looking up each time */
637 for (i
= 0; constants_set
; constants_set
>>= 1, ++i
)
639 if (!(constants_set
& 1)) continue;
641 TRACE_(d3d_constants
)("Loading constants %i: %i;\n", i
, constants
[i
]);
643 /* TODO: Benchmark and see if it would be beneficial to store the
644 * locations of the constants to avoid looking up each time */
645 snprintf(tmp_name
, sizeof(tmp_name
), "%s[%i]", prefix
, i
);
646 tmp_loc
= GL_EXTCALL(glGetUniformLocationARB(programId
, tmp_name
));
649 /* We found this uniform name in the program - go ahead and send the data */
650 GL_EXTCALL(glUniform1ivARB(tmp_loc
, 1, &constants
[i
]));
651 checkGLcall("glUniform1ivARB");
655 /* Load immediate constants */
656 ptr
= list_head(&This
->baseShader
.constantsB
);
658 const struct local_constant
*lconst
= LIST_ENTRY(ptr
, const struct local_constant
, entry
);
659 unsigned int idx
= lconst
->idx
;
660 const GLint
*values
= (const GLint
*)lconst
->value
;
662 TRACE_(d3d_constants
)("Loading local constants %i: %i\n", idx
, values
[0]);
664 snprintf(tmp_name
, sizeof(tmp_name
), "%s[%i]", prefix
, idx
);
665 tmp_loc
= GL_EXTCALL(glGetUniformLocationARB(programId
, tmp_name
));
667 /* We found this uniform name in the program - go ahead and send the data */
668 GL_EXTCALL(glUniform1ivARB(tmp_loc
, 1, values
));
669 checkGLcall("glUniform1ivARB");
671 ptr
= list_next(&This
->baseShader
.constantsB
, ptr
);
675 static void reset_program_constant_version(struct wine_rb_entry
*entry
, void *context
)
677 WINE_RB_ENTRY_VALUE(entry
, struct glsl_shader_prog_link
, program_lookup_entry
)->constant_version
= 0;
681 * Loads the texture dimensions for NP2 fixup into the currently set GLSL program.
683 /* GL locking is done by the caller (state handler) */
684 static void shader_glsl_load_np2fixup_constants(void *shader_priv
,
685 const struct wined3d_gl_info
*gl_info
, const struct wined3d_state
*state
)
687 struct shader_glsl_priv
*glsl_priv
= shader_priv
;
688 const struct glsl_shader_prog_link
*prog
= glsl_priv
->glsl_program
;
690 /* No GLSL program set - nothing to do. */
693 /* NP2 texcoord fixup is (currently) only done for pixelshaders. */
694 if (!use_ps(state
)) return;
696 if (prog
->ps_args
.np2_fixup
&& prog
->np2Fixup_location
!= -1)
699 UINT fixup
= prog
->ps_args
.np2_fixup
;
700 GLfloat np2fixup_constants
[4 * MAX_FRAGMENT_SAMPLERS
];
702 for (i
= 0; fixup
; fixup
>>= 1, ++i
)
704 const IWineD3DBaseTextureImpl
*tex
= state
->textures
[i
];
705 const unsigned char idx
= prog
->np2Fixup_info
->idx
[i
];
706 GLfloat
*tex_dim
= &np2fixup_constants
[(idx
>> 1) * 4];
710 ERR("Nonexistent texture is flagged for NP2 texcoord fixup.\n");
715 tex_dim
[2] = tex
->baseTexture
.pow2Matrix
[0]; tex_dim
[3] = tex
->baseTexture
.pow2Matrix
[5];
717 tex_dim
[0] = tex
->baseTexture
.pow2Matrix
[0]; tex_dim
[1] = tex
->baseTexture
.pow2Matrix
[5];
721 GL_EXTCALL(glUniform4fvARB(prog
->np2Fixup_location
, prog
->np2Fixup_info
->num_consts
, np2fixup_constants
));
726 * Loads the app-supplied constants into the currently set GLSL program.
728 /* GL locking is done by the caller (state handler) */
729 static void shader_glsl_load_constants(const struct wined3d_context
*context
,
730 char usePixelShader
, char useVertexShader
)
732 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
733 IWineD3DDeviceImpl
*device
= context
->swapchain
->device
;
734 IWineD3DStateBlockImpl
* stateBlock
= device
->stateBlock
;
735 struct shader_glsl_priv
*priv
= device
->shader_priv
;
736 float position_fixup
[4];
738 GLhandleARB programId
;
739 struct glsl_shader_prog_link
*prog
= priv
->glsl_program
;
740 UINT constant_version
;
744 /* No GLSL program set - nothing to do. */
747 programId
= prog
->programId
;
748 constant_version
= prog
->constant_version
;
752 IWineD3DBaseShaderImpl
*vshader
= (IWineD3DBaseShaderImpl
*)stateBlock
->state
.vertex_shader
;
754 /* Load DirectX 9 float constants/uniforms for vertex shader */
755 shader_glsl_load_constantsF(vshader
, gl_info
, stateBlock
->state
.vs_consts_f
,
756 prog
->vuniformF_locations
, &priv
->vconst_heap
, priv
->stack
, constant_version
);
758 /* Load DirectX 9 integer constants/uniforms for vertex shader */
759 shader_glsl_load_constantsI(vshader
, gl_info
, prog
->vuniformI_locations
, stateBlock
->state
.vs_consts_i
,
760 stateBlock
->changed
.vertexShaderConstantsI
& vshader
->baseShader
.reg_maps
.integer_constants
);
762 /* Load DirectX 9 boolean constants/uniforms for vertex shader */
763 shader_glsl_load_constantsB(vshader
, gl_info
, programId
, stateBlock
->state
.vs_consts_b
,
764 stateBlock
->changed
.vertexShaderConstantsB
& vshader
->baseShader
.reg_maps
.boolean_constants
);
766 /* Upload the position fixup params */
767 shader_get_position_fixup(context
, &stateBlock
->state
, position_fixup
);
768 GL_EXTCALL(glUniform4fvARB(prog
->posFixup_location
, 1, position_fixup
));
769 checkGLcall("glUniform4fvARB");
774 IWineD3DBaseShaderImpl
*pshader
= (IWineD3DBaseShaderImpl
*)stateBlock
->state
.pixel_shader
;
776 /* Load DirectX 9 float constants/uniforms for pixel shader */
777 shader_glsl_load_constantsF(pshader
, gl_info
, stateBlock
->state
.ps_consts_f
,
778 prog
->puniformF_locations
, &priv
->pconst_heap
, priv
->stack
, constant_version
);
780 /* Load DirectX 9 integer constants/uniforms for pixel shader */
781 shader_glsl_load_constantsI(pshader
, gl_info
, prog
->puniformI_locations
, stateBlock
->state
.ps_consts_i
,
782 stateBlock
->changed
.pixelShaderConstantsI
& pshader
->baseShader
.reg_maps
.integer_constants
);
784 /* Load DirectX 9 boolean constants/uniforms for pixel shader */
785 shader_glsl_load_constantsB(pshader
, gl_info
, programId
, stateBlock
->state
.ps_consts_b
,
786 stateBlock
->changed
.pixelShaderConstantsB
& pshader
->baseShader
.reg_maps
.boolean_constants
);
788 /* Upload the environment bump map matrix if needed. The needsbumpmat member specifies the texture stage to load the matrix from.
789 * It can't be 0 for a valid texbem instruction.
791 for(i
= 0; i
< MAX_TEXTURES
; i
++) {
794 if(prog
->bumpenvmat_location
[i
] == -1) continue;
796 data
= (const float *)&stateBlock
->state
.texture_states
[i
][WINED3DTSS_BUMPENVMAT00
];
797 GL_EXTCALL(glUniformMatrix2fvARB(prog
->bumpenvmat_location
[i
], 1, 0, data
));
798 checkGLcall("glUniformMatrix2fvARB");
800 /* texbeml needs the luminance scale and offset too. If texbeml
801 * is used, needsbumpmat is set too, so we can check that in the
802 * needsbumpmat check. */
803 if (prog
->luminancescale_location
[i
] != -1)
805 const GLfloat
*scale
= (const GLfloat
*)&stateBlock
->state
.texture_states
[i
][WINED3DTSS_BUMPENVLSCALE
];
806 const GLfloat
*offset
= (const GLfloat
*)&stateBlock
->state
.texture_states
[i
][WINED3DTSS_BUMPENVLOFFSET
];
808 GL_EXTCALL(glUniform1fvARB(prog
->luminancescale_location
[i
], 1, scale
));
809 checkGLcall("glUniform1fvARB");
810 GL_EXTCALL(glUniform1fvARB(prog
->luminanceoffset_location
[i
], 1, offset
));
811 checkGLcall("glUniform1fvARB");
815 if(((IWineD3DPixelShaderImpl
*) pshader
)->vpos_uniform
) {
816 float correction_params
[4];
818 if (context
->render_offscreen
)
820 correction_params
[0] = 0.0f
;
821 correction_params
[1] = 1.0f
;
823 /* position is window relative, not viewport relative */
824 correction_params
[0] = context
->current_rt
->currentDesc
.Height
;
825 correction_params
[1] = -1.0f
;
827 GL_EXTCALL(glUniform4fvARB(prog
->ycorrection_location
, 1, correction_params
));
831 if (priv
->next_constant_version
== UINT_MAX
)
833 TRACE("Max constant version reached, resetting to 0.\n");
834 wine_rb_for_each_entry(&priv
->program_lookup
, reset_program_constant_version
, NULL
);
835 priv
->next_constant_version
= 1;
839 prog
->constant_version
= priv
->next_constant_version
++;
843 static inline void update_heap_entry(struct constant_heap
*heap
, unsigned int idx
,
844 unsigned int heap_idx
, DWORD new_version
)
846 struct constant_entry
*entries
= heap
->entries
;
847 unsigned int *positions
= heap
->positions
;
848 unsigned int parent_idx
;
852 parent_idx
= heap_idx
>> 1;
854 if (new_version
<= entries
[parent_idx
].version
) break;
856 entries
[heap_idx
] = entries
[parent_idx
];
857 positions
[entries
[parent_idx
].idx
] = heap_idx
;
858 heap_idx
= parent_idx
;
861 entries
[heap_idx
].version
= new_version
;
862 entries
[heap_idx
].idx
= idx
;
863 positions
[idx
] = heap_idx
;
866 static void shader_glsl_update_float_vertex_constants(IWineD3DDevice
*iface
, UINT start
, UINT count
)
868 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
869 struct shader_glsl_priv
*priv
= This
->shader_priv
;
870 struct constant_heap
*heap
= &priv
->vconst_heap
;
873 for (i
= start
; i
< count
+ start
; ++i
)
875 if (!This
->stateBlock
->changed
.vertexShaderConstantsF
[i
])
876 update_heap_entry(heap
, i
, heap
->size
++, priv
->next_constant_version
);
878 update_heap_entry(heap
, i
, heap
->positions
[i
], priv
->next_constant_version
);
882 static void shader_glsl_update_float_pixel_constants(IWineD3DDevice
*iface
, UINT start
, UINT count
)
884 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
885 struct shader_glsl_priv
*priv
= This
->shader_priv
;
886 struct constant_heap
*heap
= &priv
->pconst_heap
;
889 for (i
= start
; i
< count
+ start
; ++i
)
891 if (!This
->stateBlock
->changed
.pixelShaderConstantsF
[i
])
892 update_heap_entry(heap
, i
, heap
->size
++, priv
->next_constant_version
);
894 update_heap_entry(heap
, i
, heap
->positions
[i
], priv
->next_constant_version
);
898 static unsigned int vec4_varyings(DWORD shader_major
, const struct wined3d_gl_info
*gl_info
)
900 unsigned int ret
= gl_info
->limits
.glsl_varyings
/ 4;
901 /* 4.0 shaders do not write clip coords because d3d10 does not support user clipplanes */
902 if(shader_major
> 3) return ret
;
904 /* 3.0 shaders may need an extra varying for the clip coord on some cards(mostly dx10 ones) */
905 if (gl_info
->quirks
& WINED3D_QUIRK_GLSL_CLIP_VARYING
) ret
-= 1;
909 /** Generate the variable & register declarations for the GLSL output target */
910 static void shader_generate_glsl_declarations(const struct wined3d_context
*context
,
911 struct wined3d_shader_buffer
*buffer
, IWineD3DBaseShader
*iface
,
912 const shader_reg_maps
*reg_maps
, struct shader_glsl_ctx_priv
*ctx_priv
)
914 IWineD3DBaseShaderImpl
* This
= (IWineD3DBaseShaderImpl
*) iface
;
915 IWineD3DDeviceImpl
*device
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
916 const struct wined3d_state
*state
= &device
->stateBlock
->state
;
917 const struct ps_compile_args
*ps_args
= ctx_priv
->cur_ps_args
;
918 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
919 unsigned int i
, extra_constants_needed
= 0;
920 const local_constant
*lconst
;
923 /* There are some minor differences between pixel and vertex shaders */
924 char pshader
= shader_is_pshader_version(reg_maps
->shader_version
.type
);
925 char prefix
= pshader
? 'P' : 'V';
927 /* Prototype the subroutines */
928 for (i
= 0, map
= reg_maps
->labels
; map
; map
>>= 1, ++i
)
930 if (map
& 1) shader_addline(buffer
, "void subroutine%u();\n", i
);
933 /* Declare the constants (aka uniforms) */
934 if (This
->baseShader
.limits
.constant_float
> 0) {
935 unsigned max_constantsF
;
936 /* Unless the shader uses indirect addressing, always declare the maximum array size and ignore that we need some
937 * uniforms privately. E.g. if GL supports 256 uniforms, and we need 2 for the pos fixup and immediate values, still
938 * declare VC[256]. If the shader needs more uniforms than we have it won't work in any case. If it uses less, the
939 * compiler will figure out which uniforms are really used and strip them out. This allows a shader to use c255 on
940 * a dx9 card, as long as it doesn't also use all the other constants.
942 * If the shader uses indirect addressing the compiler must assume that all declared uniforms are used. In this case,
943 * declare only the amount that we're assured to have.
945 * Thus we run into problems in these two cases:
946 * 1) The shader really uses more uniforms than supported
947 * 2) The shader uses indirect addressing, less constants than supported, but uses a constant index > #supported consts
951 /* No indirect addressing here. */
952 max_constantsF
= gl_info
->limits
.glsl_ps_float_constants
;
956 if(This
->baseShader
.reg_maps
.usesrelconstF
) {
957 /* Subtract the other potential uniforms from the max available (bools, ints, and 1 row of projection matrix).
958 * Subtract another uniform for immediate values, which have to be loaded via uniform by the driver as well.
959 * The shader code only uses 0.5, 2.0, 1.0, 128 and -128 in vertex shader code, so one vec4 should be enough
960 * (Unfortunately the Nvidia driver doesn't store 128 and -128 in one float).
962 * Writing gl_ClipVertex requires one uniform for each clipplane as well.
964 max_constantsF
= gl_info
->limits
.glsl_vs_float_constants
- 3;
965 if(ctx_priv
->cur_vs_args
->clip_enabled
)
967 max_constantsF
-= gl_info
->limits
.clipplanes
;
969 max_constantsF
-= count_bits(This
->baseShader
.reg_maps
.integer_constants
);
970 /* Strictly speaking a bool only uses one scalar, but the nvidia(Linux) compiler doesn't pack them properly,
971 * so each scalar requires a full vec4. We could work around this by packing the booleans ourselves, but
972 * for now take this into account when calculating the number of available constants
974 max_constantsF
-= count_bits(This
->baseShader
.reg_maps
.boolean_constants
);
975 /* Set by driver quirks in directx.c */
976 max_constantsF
-= gl_info
->reserved_glsl_constants
;
980 max_constantsF
= gl_info
->limits
.glsl_vs_float_constants
;
983 max_constantsF
= min(This
->baseShader
.limits
.constant_float
, max_constantsF
);
984 shader_addline(buffer
, "uniform vec4 %cC[%u];\n", prefix
, max_constantsF
);
987 /* Always declare the full set of constants, the compiler can remove the unused ones because d3d doesn't(yet)
988 * support indirect int and bool constant addressing. This avoids problems if the app uses e.g. i0 and i9.
990 if (This
->baseShader
.limits
.constant_int
> 0 && This
->baseShader
.reg_maps
.integer_constants
)
991 shader_addline(buffer
, "uniform ivec4 %cI[%u];\n", prefix
, This
->baseShader
.limits
.constant_int
);
993 if (This
->baseShader
.limits
.constant_bool
> 0 && This
->baseShader
.reg_maps
.boolean_constants
)
994 shader_addline(buffer
, "uniform bool %cB[%u];\n", prefix
, This
->baseShader
.limits
.constant_bool
);
998 shader_addline(buffer
, "uniform vec4 posFixup;\n");
999 shader_addline(buffer
, "void order_ps_input(in vec4[%u]);\n", MAX_REG_OUTPUT
);
1003 for (i
= 0, map
= reg_maps
->bumpmat
; map
; map
>>= 1, ++i
)
1005 if (!(map
& 1)) continue;
1007 shader_addline(buffer
, "uniform mat2 bumpenvmat%d;\n", i
);
1009 if (reg_maps
->luminanceparams
& (1 << i
))
1011 shader_addline(buffer
, "uniform float luminancescale%d;\n", i
);
1012 shader_addline(buffer
, "uniform float luminanceoffset%d;\n", i
);
1013 extra_constants_needed
++;
1016 extra_constants_needed
++;
1019 if (ps_args
->srgb_correction
)
1021 shader_addline(buffer
, "const vec4 srgb_const0 = vec4(%.8e, %.8e, %.8e, %.8e);\n",
1022 srgb_pow
, srgb_mul_high
, srgb_sub_high
, srgb_mul_low
);
1023 shader_addline(buffer
, "const vec4 srgb_const1 = vec4(%.8e, 0.0, 0.0, 0.0);\n",
1026 if (reg_maps
->vpos
|| reg_maps
->usesdsy
)
1028 if (This
->baseShader
.limits
.constant_float
+ extra_constants_needed
1029 + 1 < gl_info
->limits
.glsl_ps_float_constants
)
1031 shader_addline(buffer
, "uniform vec4 ycorrection;\n");
1032 ((IWineD3DPixelShaderImpl
*) This
)->vpos_uniform
= 1;
1033 extra_constants_needed
++;
1035 /* This happens because we do not have proper tracking of the constant registers that are
1036 * actually used, only the max limit of the shader version
1038 FIXME("Cannot find a free uniform for vpos correction params\n");
1039 shader_addline(buffer
, "const vec4 ycorrection = vec4(%f, %f, 0.0, 0.0);\n",
1040 context
->render_offscreen
? 0.0f
: device
->render_targets
[0]->currentDesc
.Height
,
1041 context
->render_offscreen
? 1.0f
: -1.0f
);
1043 shader_addline(buffer
, "vec4 vpos;\n");
1047 /* Declare texture samplers */
1048 for (i
= 0; i
< This
->baseShader
.limits
.sampler
; i
++) {
1049 if (reg_maps
->sampler_type
[i
])
1051 IWineD3DBaseTextureImpl
*texture
;
1053 switch (reg_maps
->sampler_type
[i
])
1056 if (pshader
&& ps_args
->shadow
& (1 << i
))
1057 shader_addline(buffer
, "uniform sampler1DShadow %csampler%u;\n", prefix
, i
);
1059 shader_addline(buffer
, "uniform sampler1D %csampler%u;\n", prefix
, i
);
1062 texture
= state
->textures
[i
];
1063 if (pshader
&& ps_args
->shadow
& (1 << i
))
1065 if (texture
&& texture
->baseTexture
.target
== GL_TEXTURE_RECTANGLE_ARB
)
1066 shader_addline(buffer
, "uniform sampler2DRectShadow %csampler%u;\n", prefix
, i
);
1068 shader_addline(buffer
, "uniform sampler2DShadow %csampler%u;\n", prefix
, i
);
1072 if (texture
&& texture
->baseTexture
.target
== GL_TEXTURE_RECTANGLE_ARB
)
1073 shader_addline(buffer
, "uniform sampler2DRect %csampler%u;\n", prefix
, i
);
1075 shader_addline(buffer
, "uniform sampler2D %csampler%u;\n", prefix
, i
);
1078 case WINED3DSTT_CUBE
:
1079 if (pshader
&& ps_args
->shadow
& (1 << i
)) FIXME("Unsupported Cube shadow sampler.\n");
1080 shader_addline(buffer
, "uniform samplerCube %csampler%u;\n", prefix
, i
);
1082 case WINED3DSTT_VOLUME
:
1083 if (pshader
&& ps_args
->shadow
& (1 << i
)) FIXME("Unsupported 3D shadow sampler.\n");
1084 shader_addline(buffer
, "uniform sampler3D %csampler%u;\n", prefix
, i
);
1087 shader_addline(buffer
, "uniform unsupported_sampler %csampler%u;\n", prefix
, i
);
1088 FIXME("Unrecognized sampler type: %#x\n", reg_maps
->sampler_type
[i
]);
1094 /* Declare uniforms for NP2 texcoord fixup:
1095 * This is NOT done inside the loop that declares the texture samplers since the NP2 fixup code
1096 * is currently only used for the GeforceFX series and when forcing the ARB_npot extension off.
1097 * Modern cards just skip the code anyway, so put it inside a separate loop. */
1098 if (pshader
&& ps_args
->np2_fixup
) {
1100 struct ps_np2fixup_info
* const fixup
= ctx_priv
->cur_np2fixup_info
;
1103 /* NP2/RECT textures in OpenGL use texcoords in the range [0,width]x[0,height]
1104 * while D3D has them in the (normalized) [0,1]x[0,1] range.
1105 * samplerNP2Fixup stores texture dimensions and is updated through
1106 * shader_glsl_load_np2fixup_constants when the sampler changes. */
1108 for (i
= 0; i
< This
->baseShader
.limits
.sampler
; ++i
) {
1109 if (reg_maps
->sampler_type
[i
]) {
1110 if (!(ps_args
->np2_fixup
& (1 << i
))) continue;
1112 if (WINED3DSTT_2D
!= reg_maps
->sampler_type
[i
]) {
1113 FIXME("Non-2D texture is flagged for NP2 texcoord fixup.\n");
1117 fixup
->idx
[i
] = cur
++;
1121 fixup
->num_consts
= (cur
+ 1) >> 1;
1122 shader_addline(buffer
, "uniform vec4 %csamplerNP2Fixup[%u];\n", prefix
, fixup
->num_consts
);
1125 /* Declare address variables */
1126 for (i
= 0, map
= reg_maps
->address
; map
; map
>>= 1, ++i
)
1128 if (map
& 1) shader_addline(buffer
, "ivec4 A%u;\n", i
);
1131 /* Declare texture coordinate temporaries and initialize them */
1132 for (i
= 0, map
= reg_maps
->texcoord
; map
; map
>>= 1, ++i
)
1134 if (map
& 1) shader_addline(buffer
, "vec4 T%u = gl_TexCoord[%u];\n", i
, i
);
1137 /* Declare input register varyings. Only pixel shader, vertex shaders have that declared in the
1138 * helper function shader that is linked in at link time
1140 if (pshader
&& reg_maps
->shader_version
.major
>= 3)
1144 shader_addline(buffer
, "varying vec4 IN[%u];\n", vec4_varyings(reg_maps
->shader_version
.major
, gl_info
));
1146 /* TODO: Write a replacement shader for the fixed function vertex pipeline, so this isn't needed.
1147 * For fixed function vertex processing + 3.0 pixel shader we need a separate function in the
1148 * pixel shader that reads the fixed function color into the packed input registers.
1150 shader_addline(buffer
, "vec4 IN[%u];\n", vec4_varyings(reg_maps
->shader_version
.major
, gl_info
));
1154 /* Declare output register temporaries */
1155 if(This
->baseShader
.limits
.packed_output
) {
1156 shader_addline(buffer
, "vec4 OUT[%u];\n", This
->baseShader
.limits
.packed_output
);
1159 /* Declare temporary variables */
1160 for (i
= 0, map
= reg_maps
->temporary
; map
; map
>>= 1, ++i
)
1162 if (map
& 1) shader_addline(buffer
, "vec4 R%u;\n", i
);
1165 /* Declare attributes */
1166 if (reg_maps
->shader_version
.type
== WINED3D_SHADER_TYPE_VERTEX
)
1168 for (i
= 0, map
= reg_maps
->input_registers
; map
; map
>>= 1, ++i
)
1170 if (map
& 1) shader_addline(buffer
, "attribute vec4 attrib%i;\n", i
);
1174 /* Declare loop registers aLx */
1175 for (i
= 0; i
< reg_maps
->loop_depth
; i
++) {
1176 shader_addline(buffer
, "int aL%u;\n", i
);
1177 shader_addline(buffer
, "int tmpInt%u;\n", i
);
1180 /* Temporary variables for matrix operations */
1181 shader_addline(buffer
, "vec4 tmp0;\n");
1182 shader_addline(buffer
, "vec4 tmp1;\n");
1184 /* Local constants use a different name so they can be loaded once at shader link time
1185 * They can't be hardcoded into the shader text via LC = {x, y, z, w}; because the
1186 * float -> string conversion can cause precision loss.
1188 if(!This
->baseShader
.load_local_constsF
) {
1189 LIST_FOR_EACH_ENTRY(lconst
, &This
->baseShader
.constantsF
, local_constant
, entry
) {
1190 shader_addline(buffer
, "uniform vec4 %cLC%u;\n", prefix
, lconst
->idx
);
1194 /* Start the main program */
1195 shader_addline(buffer
, "void main() {\n");
1196 if(pshader
&& reg_maps
->vpos
) {
1197 /* DirectX apps expect integer values, while OpenGL drivers add approximately 0.5. This causes
1198 * off-by-one problems as spotted by the vPos d3d9 visual test. Unfortunately the ATI cards do
1199 * not add exactly 0.5, but rather something like 0.49999999 or 0.50000001, which still causes
1200 * precision troubles when we just substract 0.5.
1202 * To deal with that just floor() the position. This will eliminate the fraction on all cards.
1204 * TODO: Test how that behaves with multisampling once we can enable multisampling in winex11.
1206 * An advantage of floor is that it works even if the driver doesn't add 1/2. It is somewhat
1207 * questionable if 1.5, 2.5, ... are the proper values to return in gl_FragCoord, even though
1208 * coordinates specify the pixel centers instead of the pixel corners. This code will behave
1209 * correctly on drivers that returns integer values.
1211 shader_addline(buffer
, "vpos = floor(vec4(0, ycorrection[0], 0, 0) + gl_FragCoord * vec4(1, ycorrection[1], 1, 1));\n");
1215 /*****************************************************************************
1216 * Functions to generate GLSL strings from DirectX Shader bytecode begin here.
1218 * For more information, see http://wiki.winehq.org/DirectX-Shaders
1219 ****************************************************************************/
1222 static void shader_glsl_add_src_param(const struct wined3d_shader_instruction
*ins
,
1223 const struct wined3d_shader_src_param
*wined3d_src
, DWORD mask
, glsl_src_param_t
*glsl_src
);
1225 /** Used for opcode modifiers - They multiply the result by the specified amount */
1226 static const char * const shift_glsl_tab
[] = {
1228 "2.0 * ", /* 1 (x2) */
1229 "4.0 * ", /* 2 (x4) */
1230 "8.0 * ", /* 3 (x8) */
1231 "16.0 * ", /* 4 (x16) */
1232 "32.0 * ", /* 5 (x32) */
1239 "0.0625 * ", /* 12 (d16) */
1240 "0.125 * ", /* 13 (d8) */
1241 "0.25 * ", /* 14 (d4) */
1242 "0.5 * " /* 15 (d2) */
1245 /* Generate a GLSL parameter that does the input modifier computation and return the input register/mask to use */
1246 static void shader_glsl_gen_modifier(DWORD src_modifier
, const char *in_reg
, const char *in_regswizzle
, char *out_str
)
1250 switch (src_modifier
)
1252 case WINED3DSPSM_DZ
: /* Need to handle this in the instructions itself (texld & texcrd). */
1253 case WINED3DSPSM_DW
:
1254 case WINED3DSPSM_NONE
:
1255 sprintf(out_str
, "%s%s", in_reg
, in_regswizzle
);
1257 case WINED3DSPSM_NEG
:
1258 sprintf(out_str
, "-%s%s", in_reg
, in_regswizzle
);
1260 case WINED3DSPSM_NOT
:
1261 sprintf(out_str
, "!%s%s", in_reg
, in_regswizzle
);
1263 case WINED3DSPSM_BIAS
:
1264 sprintf(out_str
, "(%s%s - vec4(0.5)%s)", in_reg
, in_regswizzle
, in_regswizzle
);
1266 case WINED3DSPSM_BIASNEG
:
1267 sprintf(out_str
, "-(%s%s - vec4(0.5)%s)", in_reg
, in_regswizzle
, in_regswizzle
);
1269 case WINED3DSPSM_SIGN
:
1270 sprintf(out_str
, "(2.0 * (%s%s - 0.5))", in_reg
, in_regswizzle
);
1272 case WINED3DSPSM_SIGNNEG
:
1273 sprintf(out_str
, "-(2.0 * (%s%s - 0.5))", in_reg
, in_regswizzle
);
1275 case WINED3DSPSM_COMP
:
1276 sprintf(out_str
, "(1.0 - %s%s)", in_reg
, in_regswizzle
);
1278 case WINED3DSPSM_X2
:
1279 sprintf(out_str
, "(2.0 * %s%s)", in_reg
, in_regswizzle
);
1281 case WINED3DSPSM_X2NEG
:
1282 sprintf(out_str
, "-(2.0 * %s%s)", in_reg
, in_regswizzle
);
1284 case WINED3DSPSM_ABS
:
1285 sprintf(out_str
, "abs(%s%s)", in_reg
, in_regswizzle
);
1287 case WINED3DSPSM_ABSNEG
:
1288 sprintf(out_str
, "-abs(%s%s)", in_reg
, in_regswizzle
);
1291 FIXME("Unhandled modifier %u\n", src_modifier
);
1292 sprintf(out_str
, "%s%s", in_reg
, in_regswizzle
);
1296 /** Writes the GLSL variable name that corresponds to the register that the
1297 * DX opcode parameter is trying to access */
1298 static void shader_glsl_get_register_name(const struct wined3d_shader_register
*reg
,
1299 char *register_name
, BOOL
*is_color
, const struct wined3d_shader_instruction
*ins
)
1301 /* oPos, oFog and oPts in D3D */
1302 static const char * const hwrastout_reg_names
[] = {"OUT[10]", "OUT[11].x", "OUT[11].y"};
1304 IWineD3DBaseShaderImpl
*This
= (IWineD3DBaseShaderImpl
*)ins
->ctx
->shader
;
1305 const struct wined3d_gl_info
*gl_info
= ins
->ctx
->gl_info
;
1306 char pshader
= shader_is_pshader_version(This
->baseShader
.reg_maps
.shader_version
.type
);
1312 case WINED3DSPR_TEMP
:
1313 sprintf(register_name
, "R%u", reg
->idx
);
1316 case WINED3DSPR_INPUT
:
1317 /* vertex shaders */
1320 struct shader_glsl_ctx_priv
*priv
= ins
->ctx
->backend_data
;
1321 if (priv
->cur_vs_args
->swizzle_map
& (1 << reg
->idx
)) *is_color
= TRUE
;
1322 sprintf(register_name
, "attrib%u", reg
->idx
);
1326 /* pixel shaders >= 3.0 */
1327 if (This
->baseShader
.reg_maps
.shader_version
.major
>= 3)
1329 DWORD idx
= ((IWineD3DPixelShaderImpl
*)This
)->input_reg_map
[reg
->idx
];
1330 unsigned int in_count
= vec4_varyings(This
->baseShader
.reg_maps
.shader_version
.major
, gl_info
);
1334 glsl_src_param_t rel_param
;
1336 shader_glsl_add_src_param(ins
, reg
->rel_addr
, WINED3DSP_WRITEMASK_0
, &rel_param
);
1338 /* Removing a + 0 would be an obvious optimization, but macos doesn't see the NOP
1339 * operation there */
1342 if (((IWineD3DPixelShaderImpl
*)This
)->declared_in_count
> in_count
)
1344 sprintf(register_name
,
1345 "((%s + %u) > %d ? (%s + %u) > %d ? gl_SecondaryColor : gl_Color : IN[%s + %u])",
1346 rel_param
.param_str
, idx
, in_count
- 1, rel_param
.param_str
, idx
, in_count
,
1347 rel_param
.param_str
, idx
);
1351 sprintf(register_name
, "IN[%s + %u]", rel_param
.param_str
, idx
);
1356 if (((IWineD3DPixelShaderImpl
*)This
)->declared_in_count
> in_count
)
1358 sprintf(register_name
, "((%s) > %d ? (%s) > %d ? gl_SecondaryColor : gl_Color : IN[%s])",
1359 rel_param
.param_str
, in_count
- 1, rel_param
.param_str
, in_count
,
1360 rel_param
.param_str
);
1364 sprintf(register_name
, "IN[%s]", rel_param
.param_str
);
1370 if (idx
== in_count
) sprintf(register_name
, "gl_Color");
1371 else if (idx
== in_count
+ 1) sprintf(register_name
, "gl_SecondaryColor");
1372 else sprintf(register_name
, "IN[%u]", idx
);
1377 if (!reg
->idx
) strcpy(register_name
, "gl_Color");
1378 else strcpy(register_name
, "gl_SecondaryColor");
1383 case WINED3DSPR_CONST
:
1385 const char prefix
= pshader
? 'P' : 'V';
1387 /* Relative addressing */
1390 glsl_src_param_t rel_param
;
1391 shader_glsl_add_src_param(ins
, reg
->rel_addr
, WINED3DSP_WRITEMASK_0
, &rel_param
);
1392 if (reg
->idx
) sprintf(register_name
, "%cC[%s + %u]", prefix
, rel_param
.param_str
, reg
->idx
);
1393 else sprintf(register_name
, "%cC[%s]", prefix
, rel_param
.param_str
);
1397 if (shader_constant_is_local(This
, reg
->idx
))
1398 sprintf(register_name
, "%cLC%u", prefix
, reg
->idx
);
1400 sprintf(register_name
, "%cC[%u]", prefix
, reg
->idx
);
1405 case WINED3DSPR_CONSTINT
:
1406 if (pshader
) sprintf(register_name
, "PI[%u]", reg
->idx
);
1407 else sprintf(register_name
, "VI[%u]", reg
->idx
);
1410 case WINED3DSPR_CONSTBOOL
:
1411 if (pshader
) sprintf(register_name
, "PB[%u]", reg
->idx
);
1412 else sprintf(register_name
, "VB[%u]", reg
->idx
);
1415 case WINED3DSPR_TEXTURE
: /* case WINED3DSPR_ADDR: */
1416 if (pshader
) sprintf(register_name
, "T%u", reg
->idx
);
1417 else sprintf(register_name
, "A%u", reg
->idx
);
1420 case WINED3DSPR_LOOP
:
1421 sprintf(register_name
, "aL%u", This
->baseShader
.cur_loop_regno
- 1);
1424 case WINED3DSPR_SAMPLER
:
1425 if (pshader
) sprintf(register_name
, "Psampler%u", reg
->idx
);
1426 else sprintf(register_name
, "Vsampler%u", reg
->idx
);
1429 case WINED3DSPR_COLOROUT
:
1430 if (reg
->idx
>= gl_info
->limits
.buffers
)
1431 WARN("Write to render target %u, only %d supported.\n", reg
->idx
, gl_info
->limits
.buffers
);
1433 sprintf(register_name
, "gl_FragData[%u]", reg
->idx
);
1436 case WINED3DSPR_RASTOUT
:
1437 sprintf(register_name
, "%s", hwrastout_reg_names
[reg
->idx
]);
1440 case WINED3DSPR_DEPTHOUT
:
1441 sprintf(register_name
, "gl_FragDepth");
1444 case WINED3DSPR_ATTROUT
:
1445 if (!reg
->idx
) sprintf(register_name
, "OUT[8]");
1446 else sprintf(register_name
, "OUT[9]");
1449 case WINED3DSPR_TEXCRDOUT
:
1450 /* Vertex shaders >= 3.0: WINED3DSPR_OUTPUT */
1451 sprintf(register_name
, "OUT[%u]", reg
->idx
);
1454 case WINED3DSPR_MISCTYPE
:
1458 sprintf(register_name
, "vpos");
1460 else if (reg
->idx
== 1)
1462 /* Note that gl_FrontFacing is a bool, while vFace is
1463 * a float for which the sign determines front/back */
1464 sprintf(register_name
, "(gl_FrontFacing ? 1.0 : -1.0)");
1468 FIXME("Unhandled misctype register %d\n", reg
->idx
);
1469 sprintf(register_name
, "unrecognized_register");
1473 case WINED3DSPR_IMMCONST
:
1474 switch (reg
->immconst_type
)
1476 case WINED3D_IMMCONST_SCALAR
:
1477 sprintf(register_name
, "%.8e", *(const float *)reg
->immconst_data
);
1480 case WINED3D_IMMCONST_VEC4
:
1481 sprintf(register_name
, "vec4(%.8e, %.8e, %.8e, %.8e)",
1482 *(const float *)®
->immconst_data
[0], *(const float *)®
->immconst_data
[1],
1483 *(const float *)®
->immconst_data
[2], *(const float *)®
->immconst_data
[3]);
1487 FIXME("Unhandled immconst type %#x\n", reg
->immconst_type
);
1488 sprintf(register_name
, "<unhandled_immconst_type %#x>", reg
->immconst_type
);
1493 FIXME("Unhandled register name Type(%d)\n", reg
->type
);
1494 sprintf(register_name
, "unrecognized_register");
1499 static void shader_glsl_write_mask_to_str(DWORD write_mask
, char *str
)
1502 if (write_mask
& WINED3DSP_WRITEMASK_0
) *str
++ = 'x';
1503 if (write_mask
& WINED3DSP_WRITEMASK_1
) *str
++ = 'y';
1504 if (write_mask
& WINED3DSP_WRITEMASK_2
) *str
++ = 'z';
1505 if (write_mask
& WINED3DSP_WRITEMASK_3
) *str
++ = 'w';
1509 /* Get the GLSL write mask for the destination register */
1510 static DWORD
shader_glsl_get_write_mask(const struct wined3d_shader_dst_param
*param
, char *write_mask
)
1512 DWORD mask
= param
->write_mask
;
1514 if (shader_is_scalar(¶m
->reg
))
1516 mask
= WINED3DSP_WRITEMASK_0
;
1521 shader_glsl_write_mask_to_str(mask
, write_mask
);
1527 static unsigned int shader_glsl_get_write_mask_size(DWORD write_mask
) {
1528 unsigned int size
= 0;
1530 if (write_mask
& WINED3DSP_WRITEMASK_0
) ++size
;
1531 if (write_mask
& WINED3DSP_WRITEMASK_1
) ++size
;
1532 if (write_mask
& WINED3DSP_WRITEMASK_2
) ++size
;
1533 if (write_mask
& WINED3DSP_WRITEMASK_3
) ++size
;
1538 static void shader_glsl_swizzle_to_str(const DWORD swizzle
, BOOL fixup
, DWORD mask
, char *str
)
1540 /* For registers of type WINED3DDECLTYPE_D3DCOLOR, data is stored as "bgra",
1541 * but addressed as "rgba". To fix this we need to swap the register's x
1542 * and z components. */
1543 const char *swizzle_chars
= fixup
? "zyxw" : "xyzw";
1546 /* swizzle bits fields: wwzzyyxx */
1547 if (mask
& WINED3DSP_WRITEMASK_0
) *str
++ = swizzle_chars
[swizzle
& 0x03];
1548 if (mask
& WINED3DSP_WRITEMASK_1
) *str
++ = swizzle_chars
[(swizzle
>> 2) & 0x03];
1549 if (mask
& WINED3DSP_WRITEMASK_2
) *str
++ = swizzle_chars
[(swizzle
>> 4) & 0x03];
1550 if (mask
& WINED3DSP_WRITEMASK_3
) *str
++ = swizzle_chars
[(swizzle
>> 6) & 0x03];
1554 static void shader_glsl_get_swizzle(const struct wined3d_shader_src_param
*param
,
1555 BOOL fixup
, DWORD mask
, char *swizzle_str
)
1557 if (shader_is_scalar(¶m
->reg
))
1558 *swizzle_str
= '\0';
1560 shader_glsl_swizzle_to_str(param
->swizzle
, fixup
, mask
, swizzle_str
);
1563 /* From a given parameter token, generate the corresponding GLSL string.
1564 * Also, return the actual register name and swizzle in case the
1565 * caller needs this information as well. */
1566 static void shader_glsl_add_src_param(const struct wined3d_shader_instruction
*ins
,
1567 const struct wined3d_shader_src_param
*wined3d_src
, DWORD mask
, glsl_src_param_t
*glsl_src
)
1569 BOOL is_color
= FALSE
;
1570 char swizzle_str
[6];
1572 glsl_src
->reg_name
[0] = '\0';
1573 glsl_src
->param_str
[0] = '\0';
1574 swizzle_str
[0] = '\0';
1576 shader_glsl_get_register_name(&wined3d_src
->reg
, glsl_src
->reg_name
, &is_color
, ins
);
1577 shader_glsl_get_swizzle(wined3d_src
, is_color
, mask
, swizzle_str
);
1578 shader_glsl_gen_modifier(wined3d_src
->modifiers
, glsl_src
->reg_name
, swizzle_str
, glsl_src
->param_str
);
1581 /* From a given parameter token, generate the corresponding GLSL string.
1582 * Also, return the actual register name and swizzle in case the
1583 * caller needs this information as well. */
1584 static DWORD
shader_glsl_add_dst_param(const struct wined3d_shader_instruction
*ins
,
1585 const struct wined3d_shader_dst_param
*wined3d_dst
, glsl_dst_param_t
*glsl_dst
)
1587 BOOL is_color
= FALSE
;
1589 glsl_dst
->mask_str
[0] = '\0';
1590 glsl_dst
->reg_name
[0] = '\0';
1592 shader_glsl_get_register_name(&wined3d_dst
->reg
, glsl_dst
->reg_name
, &is_color
, ins
);
1593 return shader_glsl_get_write_mask(wined3d_dst
, glsl_dst
->mask_str
);
1596 /* Append the destination part of the instruction to the buffer, return the effective write mask */
1597 static DWORD
shader_glsl_append_dst_ext(struct wined3d_shader_buffer
*buffer
,
1598 const struct wined3d_shader_instruction
*ins
, const struct wined3d_shader_dst_param
*dst
)
1600 glsl_dst_param_t glsl_dst
;
1603 mask
= shader_glsl_add_dst_param(ins
, dst
, &glsl_dst
);
1604 if (mask
) shader_addline(buffer
, "%s%s = %s(", glsl_dst
.reg_name
, glsl_dst
.mask_str
, shift_glsl_tab
[dst
->shift
]);
1609 /* Append the destination part of the instruction to the buffer, return the effective write mask */
1610 static DWORD
shader_glsl_append_dst(struct wined3d_shader_buffer
*buffer
, const struct wined3d_shader_instruction
*ins
)
1612 return shader_glsl_append_dst_ext(buffer
, ins
, &ins
->dst
[0]);
1615 /** Process GLSL instruction modifiers */
1616 static void shader_glsl_add_instruction_modifiers(const struct wined3d_shader_instruction
*ins
)
1618 glsl_dst_param_t dst_param
;
1621 if (!ins
->dst_count
) return;
1623 modifiers
= ins
->dst
[0].modifiers
;
1624 if (!modifiers
) return;
1626 shader_glsl_add_dst_param(ins
, &ins
->dst
[0], &dst_param
);
1628 if (modifiers
& WINED3DSPDM_SATURATE
)
1630 /* _SAT means to clamp the value of the register to between 0 and 1 */
1631 shader_addline(ins
->ctx
->buffer
, "%s%s = clamp(%s%s, 0.0, 1.0);\n", dst_param
.reg_name
,
1632 dst_param
.mask_str
, dst_param
.reg_name
, dst_param
.mask_str
);
1635 if (modifiers
& WINED3DSPDM_MSAMPCENTROID
)
1637 FIXME("_centroid modifier not handled\n");
1640 if (modifiers
& WINED3DSPDM_PARTIALPRECISION
)
1642 /* MSDN says this modifier can be safely ignored, so that's what we'll do. */
1646 static inline const char *shader_get_comp_op(DWORD op
)
1649 case COMPARISON_GT
: return ">";
1650 case COMPARISON_EQ
: return "==";
1651 case COMPARISON_GE
: return ">=";
1652 case COMPARISON_LT
: return "<";
1653 case COMPARISON_NE
: return "!=";
1654 case COMPARISON_LE
: return "<=";
1656 FIXME("Unrecognized comparison value: %u\n", op
);
1661 static void shader_glsl_get_sample_function(const struct wined3d_shader_context
*ctx
,
1662 DWORD sampler_idx
, DWORD flags
, glsl_sample_function_t
*sample_function
)
1664 WINED3DSAMPLER_TEXTURE_TYPE sampler_type
= ctx
->reg_maps
->sampler_type
[sampler_idx
];
1665 const struct wined3d_gl_info
*gl_info
= ctx
->gl_info
;
1666 BOOL shadow
= shader_is_pshader_version(ctx
->reg_maps
->shader_version
.type
)
1667 && (((const struct shader_glsl_ctx_priv
*)ctx
->backend_data
)->cur_ps_args
->shadow
& (1 << sampler_idx
));
1668 BOOL projected
= flags
& WINED3D_GLSL_SAMPLE_PROJECTED
;
1669 BOOL texrect
= flags
& WINED3D_GLSL_SAMPLE_RECT
;
1670 BOOL lod
= flags
& WINED3D_GLSL_SAMPLE_LOD
;
1671 BOOL grad
= flags
& WINED3D_GLSL_SAMPLE_GRAD
;
1673 /* Note that there's no such thing as a projected cube texture. */
1674 switch(sampler_type
) {
1680 sample_function
->name
= projected
? "shadow1DProjLod" : "shadow1DLod";
1684 if (gl_info
->supported
[EXT_GPU_SHADER4
])
1685 sample_function
->name
= projected
? "shadow1DProjGrad" : "shadow1DGrad";
1686 else if (gl_info
->supported
[ARB_SHADER_TEXTURE_LOD
])
1687 sample_function
->name
= projected
? "shadow1DProjGradARB" : "shadow1DGradARB";
1690 FIXME("Unsupported 1D shadow grad function.\n");
1691 sample_function
->name
= "unsupported1DGrad";
1696 sample_function
->name
= projected
? "shadow1DProj" : "shadow1D";
1698 sample_function
->coord_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
;
1704 sample_function
->name
= projected
? "texture1DProjLod" : "texture1DLod";
1708 if (gl_info
->supported
[EXT_GPU_SHADER4
])
1709 sample_function
->name
= projected
? "texture1DProjGrad" : "texture1DGrad";
1710 else if (gl_info
->supported
[ARB_SHADER_TEXTURE_LOD
])
1711 sample_function
->name
= projected
? "texture1DProjGradARB" : "texture1DGradARB";
1714 FIXME("Unsupported 1D grad function.\n");
1715 sample_function
->name
= "unsupported1DGrad";
1720 sample_function
->name
= projected
? "texture1DProj" : "texture1D";
1722 sample_function
->coord_mask
= WINED3DSP_WRITEMASK_0
;
1733 sample_function
->name
= projected
? "shadow2DRectProjLod" : "shadow2DRectLod";
1737 if (gl_info
->supported
[EXT_GPU_SHADER4
])
1738 sample_function
->name
= projected
? "shadow2DRectProjGrad" : "shadow2DRectGrad";
1739 else if (gl_info
->supported
[ARB_SHADER_TEXTURE_LOD
])
1740 sample_function
->name
= projected
? "shadow2DRectProjGradARB" : "shadow2DRectGradARB";
1743 FIXME("Unsupported RECT shadow grad function.\n");
1744 sample_function
->name
= "unsupported2DRectGrad";
1749 sample_function
->name
= projected
? "shadow2DRectProj" : "shadow2DRect";
1756 sample_function
->name
= projected
? "shadow2DProjLod" : "shadow2DLod";
1760 if (gl_info
->supported
[EXT_GPU_SHADER4
])
1761 sample_function
->name
= projected
? "shadow2DProjGrad" : "shadow2DGrad";
1762 else if (gl_info
->supported
[ARB_SHADER_TEXTURE_LOD
])
1763 sample_function
->name
= projected
? "shadow2DProjGradARB" : "shadow2DGradARB";
1766 FIXME("Unsupported 2D shadow grad function.\n");
1767 sample_function
->name
= "unsupported2DGrad";
1772 sample_function
->name
= projected
? "shadow2DProj" : "shadow2D";
1775 sample_function
->coord_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
1783 sample_function
->name
= projected
? "texture2DRectProjLod" : "texture2DRectLod";
1787 if (gl_info
->supported
[EXT_GPU_SHADER4
])
1788 sample_function
->name
= projected
? "texture2DRectProjGrad" : "texture2DRectGrad";
1789 else if (gl_info
->supported
[ARB_SHADER_TEXTURE_LOD
])
1790 sample_function
->name
= projected
? "texture2DRectProjGradARB" : "texture2DRectGradARB";
1793 FIXME("Unsupported RECT grad function.\n");
1794 sample_function
->name
= "unsupported2DRectGrad";
1799 sample_function
->name
= projected
? "texture2DRectProj" : "texture2DRect";
1806 sample_function
->name
= projected
? "texture2DProjLod" : "texture2DLod";
1810 if (gl_info
->supported
[EXT_GPU_SHADER4
])
1811 sample_function
->name
= projected
? "texture2DProjGrad" : "texture2DGrad";
1812 else if (gl_info
->supported
[ARB_SHADER_TEXTURE_LOD
])
1813 sample_function
->name
= projected
? "texture2DProjGradARB" : "texture2DGradARB";
1816 FIXME("Unsupported 2D grad function.\n");
1817 sample_function
->name
= "unsupported2DGrad";
1822 sample_function
->name
= projected
? "texture2DProj" : "texture2D";
1825 sample_function
->coord_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
;
1829 case WINED3DSTT_CUBE
:
1832 FIXME("Unsupported Cube shadow function.\n");
1833 sample_function
->name
= "unsupportedCubeShadow";
1834 sample_function
->coord_mask
= 0;
1840 sample_function
->name
= "textureCubeLod";
1844 if (gl_info
->supported
[EXT_GPU_SHADER4
])
1845 sample_function
->name
= "textureCubeGrad";
1846 else if (gl_info
->supported
[ARB_SHADER_TEXTURE_LOD
])
1847 sample_function
->name
= "textureCubeGradARB";
1850 FIXME("Unsupported Cube grad function.\n");
1851 sample_function
->name
= "unsupportedCubeGrad";
1856 sample_function
->name
= "textureCube";
1858 sample_function
->coord_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
1862 case WINED3DSTT_VOLUME
:
1865 FIXME("Unsupported 3D shadow function.\n");
1866 sample_function
->name
= "unsupported3DShadow";
1867 sample_function
->coord_mask
= 0;
1873 sample_function
->name
= projected
? "texture3DProjLod" : "texture3DLod";
1877 if (gl_info
->supported
[EXT_GPU_SHADER4
])
1878 sample_function
->name
= projected
? "texture3DProjGrad" : "texture3DGrad";
1879 else if (gl_info
->supported
[ARB_SHADER_TEXTURE_LOD
])
1880 sample_function
->name
= projected
? "texture3DProjGradARB" : "texture3DGradARB";
1883 FIXME("Unsupported 3D grad function.\n");
1884 sample_function
->name
= "unsupported3DGrad";
1889 sample_function
->name
= projected
? "texture3DProj" : "texture3D";
1891 sample_function
->coord_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
1896 sample_function
->name
= "";
1897 sample_function
->coord_mask
= 0;
1898 FIXME("Unrecognized sampler type: %#x;\n", sampler_type
);
1903 static void shader_glsl_append_fixup_arg(char *arguments
, const char *reg_name
,
1904 BOOL sign_fixup
, enum fixup_channel_source channel_source
)
1906 switch(channel_source
)
1908 case CHANNEL_SOURCE_ZERO
:
1909 strcat(arguments
, "0.0");
1912 case CHANNEL_SOURCE_ONE
:
1913 strcat(arguments
, "1.0");
1916 case CHANNEL_SOURCE_X
:
1917 strcat(arguments
, reg_name
);
1918 strcat(arguments
, ".x");
1921 case CHANNEL_SOURCE_Y
:
1922 strcat(arguments
, reg_name
);
1923 strcat(arguments
, ".y");
1926 case CHANNEL_SOURCE_Z
:
1927 strcat(arguments
, reg_name
);
1928 strcat(arguments
, ".z");
1931 case CHANNEL_SOURCE_W
:
1932 strcat(arguments
, reg_name
);
1933 strcat(arguments
, ".w");
1937 FIXME("Unhandled channel source %#x\n", channel_source
);
1938 strcat(arguments
, "undefined");
1942 if (sign_fixup
) strcat(arguments
, " * 2.0 - 1.0");
1945 static void shader_glsl_color_correction(const struct wined3d_shader_instruction
*ins
, struct color_fixup_desc fixup
)
1947 struct wined3d_shader_dst_param dst
;
1948 unsigned int mask_size
, remaining
;
1949 glsl_dst_param_t dst_param
;
1950 char arguments
[256];
1954 if (fixup
.x_sign_fixup
|| fixup
.x_source
!= CHANNEL_SOURCE_X
) mask
|= WINED3DSP_WRITEMASK_0
;
1955 if (fixup
.y_sign_fixup
|| fixup
.y_source
!= CHANNEL_SOURCE_Y
) mask
|= WINED3DSP_WRITEMASK_1
;
1956 if (fixup
.z_sign_fixup
|| fixup
.z_source
!= CHANNEL_SOURCE_Z
) mask
|= WINED3DSP_WRITEMASK_2
;
1957 if (fixup
.w_sign_fixup
|| fixup
.w_source
!= CHANNEL_SOURCE_W
) mask
|= WINED3DSP_WRITEMASK_3
;
1958 mask
&= ins
->dst
[0].write_mask
;
1960 if (!mask
) return; /* Nothing to do */
1962 if (is_complex_fixup(fixup
))
1964 enum complex_fixup complex_fixup
= get_complex_fixup(fixup
);
1965 FIXME("Complex fixup (%#x) not supported\n",complex_fixup
);
1969 mask_size
= shader_glsl_get_write_mask_size(mask
);
1972 dst
.write_mask
= mask
;
1973 shader_glsl_add_dst_param(ins
, &dst
, &dst_param
);
1975 arguments
[0] = '\0';
1976 remaining
= mask_size
;
1977 if (mask
& WINED3DSP_WRITEMASK_0
)
1979 shader_glsl_append_fixup_arg(arguments
, dst_param
.reg_name
, fixup
.x_sign_fixup
, fixup
.x_source
);
1980 if (--remaining
) strcat(arguments
, ", ");
1982 if (mask
& WINED3DSP_WRITEMASK_1
)
1984 shader_glsl_append_fixup_arg(arguments
, dst_param
.reg_name
, fixup
.y_sign_fixup
, fixup
.y_source
);
1985 if (--remaining
) strcat(arguments
, ", ");
1987 if (mask
& WINED3DSP_WRITEMASK_2
)
1989 shader_glsl_append_fixup_arg(arguments
, dst_param
.reg_name
, fixup
.z_sign_fixup
, fixup
.z_source
);
1990 if (--remaining
) strcat(arguments
, ", ");
1992 if (mask
& WINED3DSP_WRITEMASK_3
)
1994 shader_glsl_append_fixup_arg(arguments
, dst_param
.reg_name
, fixup
.w_sign_fixup
, fixup
.w_source
);
1995 if (--remaining
) strcat(arguments
, ", ");
2000 shader_addline(ins
->ctx
->buffer
, "%s%s = vec%u(%s);\n",
2001 dst_param
.reg_name
, dst_param
.mask_str
, mask_size
, arguments
);
2005 shader_addline(ins
->ctx
->buffer
, "%s%s = %s;\n", dst_param
.reg_name
, dst_param
.mask_str
, arguments
);
2009 static void PRINTF_ATTR(8, 9) shader_glsl_gen_sample_code(const struct wined3d_shader_instruction
*ins
,
2010 DWORD sampler
, const glsl_sample_function_t
*sample_function
, DWORD swizzle
,
2011 const char *dx
, const char *dy
,
2012 const char *bias
, const char *coord_reg_fmt
, ...)
2014 const char *sampler_base
;
2015 char dst_swizzle
[6];
2016 struct color_fixup_desc fixup
;
2017 BOOL np2_fixup
= FALSE
;
2020 shader_glsl_swizzle_to_str(swizzle
, FALSE
, ins
->dst
[0].write_mask
, dst_swizzle
);
2022 if (shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
.type
))
2024 const struct shader_glsl_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2025 fixup
= priv
->cur_ps_args
->color_fixup
[sampler
];
2026 sampler_base
= "Psampler";
2028 if(priv
->cur_ps_args
->np2_fixup
& (1 << sampler
)) {
2030 FIXME("Biased sampling from NP2 textures is unsupported\n");
2036 sampler_base
= "Vsampler";
2037 fixup
= COLOR_FIXUP_IDENTITY
; /* FIXME: Vshader color fixup */
2040 shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2042 shader_addline(ins
->ctx
->buffer
, "%s(%s%u, ", sample_function
->name
, sampler_base
, sampler
);
2044 va_start(args
, coord_reg_fmt
);
2045 shader_vaddline(ins
->ctx
->buffer
, coord_reg_fmt
, args
);
2049 shader_addline(ins
->ctx
->buffer
, ", %s)%s);\n", bias
, dst_swizzle
);
2052 const struct shader_glsl_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2053 const unsigned char idx
= priv
->cur_np2fixup_info
->idx
[sampler
];
2055 shader_addline(ins
->ctx
->buffer
, " * PsamplerNP2Fixup[%u].%s)%s);\n", idx
>> 1,
2056 (idx
% 2) ? "zw" : "xy", dst_swizzle
);
2057 } else if(dx
&& dy
) {
2058 shader_addline(ins
->ctx
->buffer
, ", %s, %s)%s);\n", dx
, dy
, dst_swizzle
);
2060 shader_addline(ins
->ctx
->buffer
, ")%s);\n", dst_swizzle
);
2064 if(!is_identity_fixup(fixup
)) {
2065 shader_glsl_color_correction(ins
, fixup
);
2069 /*****************************************************************************
2070 * Begin processing individual instruction opcodes
2071 ****************************************************************************/
2073 /* Generate GLSL arithmetic functions (dst = src1 + src2) */
2074 static void shader_glsl_arith(const struct wined3d_shader_instruction
*ins
)
2076 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2077 glsl_src_param_t src0_param
;
2078 glsl_src_param_t src1_param
;
2082 /* Determine the GLSL operator to use based on the opcode */
2083 switch (ins
->handler_idx
)
2085 case WINED3DSIH_MUL
: op
= '*'; break;
2086 case WINED3DSIH_ADD
: op
= '+'; break;
2087 case WINED3DSIH_SUB
: op
= '-'; break;
2090 FIXME("Opcode %#x not yet handled in GLSL\n", ins
->handler_idx
);
2094 write_mask
= shader_glsl_append_dst(buffer
, ins
);
2095 shader_glsl_add_src_param(ins
, &ins
->src
[0], write_mask
, &src0_param
);
2096 shader_glsl_add_src_param(ins
, &ins
->src
[1], write_mask
, &src1_param
);
2097 shader_addline(buffer
, "%s %c %s);\n", src0_param
.param_str
, op
, src1_param
.param_str
);
2100 /* Process the WINED3DSIO_MOV opcode using GLSL (dst = src) */
2101 static void shader_glsl_mov(const struct wined3d_shader_instruction
*ins
)
2103 const struct wined3d_gl_info
*gl_info
= ins
->ctx
->gl_info
;
2104 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2105 glsl_src_param_t src0_param
;
2108 write_mask
= shader_glsl_append_dst(buffer
, ins
);
2109 shader_glsl_add_src_param(ins
, &ins
->src
[0], write_mask
, &src0_param
);
2111 /* In vs_1_1 WINED3DSIO_MOV can write to the address register. In later
2112 * shader versions WINED3DSIO_MOVA is used for this. */
2113 if (ins
->ctx
->reg_maps
->shader_version
.major
== 1
2114 && !shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
.type
)
2115 && ins
->dst
[0].reg
.type
== WINED3DSPR_ADDR
)
2117 /* This is a simple floor() */
2118 unsigned int mask_size
= shader_glsl_get_write_mask_size(write_mask
);
2119 if (mask_size
> 1) {
2120 shader_addline(buffer
, "ivec%d(floor(%s)));\n", mask_size
, src0_param
.param_str
);
2122 shader_addline(buffer
, "int(floor(%s)));\n", src0_param
.param_str
);
2125 else if(ins
->handler_idx
== WINED3DSIH_MOVA
)
2127 /* We need to *round* to the nearest int here. */
2128 unsigned int mask_size
= shader_glsl_get_write_mask_size(write_mask
);
2130 if (gl_info
->supported
[EXT_GPU_SHADER4
])
2133 shader_addline(buffer
, "ivec%d(round(%s)));\n", mask_size
, src0_param
.param_str
);
2135 shader_addline(buffer
, "int(round(%s)));\n", src0_param
.param_str
);
2140 shader_addline(buffer
, "ivec%d(floor(abs(%s) + vec%d(0.5)) * sign(%s)));\n",
2141 mask_size
, src0_param
.param_str
, mask_size
, src0_param
.param_str
);
2143 shader_addline(buffer
, "int(floor(abs(%s) + 0.5) * sign(%s)));\n",
2144 src0_param
.param_str
, src0_param
.param_str
);
2149 shader_addline(buffer
, "%s);\n", src0_param
.param_str
);
2153 /* Process the dot product operators DP3 and DP4 in GLSL (dst = dot(src0, src1)) */
2154 static void shader_glsl_dot(const struct wined3d_shader_instruction
*ins
)
2156 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2157 glsl_src_param_t src0_param
;
2158 glsl_src_param_t src1_param
;
2159 DWORD dst_write_mask
, src_write_mask
;
2160 unsigned int dst_size
= 0;
2162 dst_write_mask
= shader_glsl_append_dst(buffer
, ins
);
2163 dst_size
= shader_glsl_get_write_mask_size(dst_write_mask
);
2165 /* dp3 works on vec3, dp4 on vec4 */
2166 if (ins
->handler_idx
== WINED3DSIH_DP4
)
2168 src_write_mask
= WINED3DSP_WRITEMASK_ALL
;
2170 src_write_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2173 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_write_mask
, &src0_param
);
2174 shader_glsl_add_src_param(ins
, &ins
->src
[1], src_write_mask
, &src1_param
);
2177 shader_addline(buffer
, "vec%d(dot(%s, %s)));\n", dst_size
, src0_param
.param_str
, src1_param
.param_str
);
2179 shader_addline(buffer
, "dot(%s, %s));\n", src0_param
.param_str
, src1_param
.param_str
);
2183 /* Note that this instruction has some restrictions. The destination write mask
2184 * can't contain the w component, and the source swizzles have to be .xyzw */
2185 static void shader_glsl_cross(const struct wined3d_shader_instruction
*ins
)
2187 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2188 glsl_src_param_t src0_param
;
2189 glsl_src_param_t src1_param
;
2192 shader_glsl_get_write_mask(&ins
->dst
[0], dst_mask
);
2193 shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2194 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
2195 shader_glsl_add_src_param(ins
, &ins
->src
[1], src_mask
, &src1_param
);
2196 shader_addline(ins
->ctx
->buffer
, "cross(%s, %s)%s);\n", src0_param
.param_str
, src1_param
.param_str
, dst_mask
);
2199 /* Process the WINED3DSIO_POW instruction in GLSL (dst = |src0|^src1)
2200 * Src0 and src1 are scalars. Note that D3D uses the absolute of src0, while
2201 * GLSL uses the value as-is. */
2202 static void shader_glsl_pow(const struct wined3d_shader_instruction
*ins
)
2204 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2205 glsl_src_param_t src0_param
;
2206 glsl_src_param_t src1_param
;
2207 DWORD dst_write_mask
;
2208 unsigned int dst_size
;
2210 dst_write_mask
= shader_glsl_append_dst(buffer
, ins
);
2211 dst_size
= shader_glsl_get_write_mask_size(dst_write_mask
);
2213 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2214 shader_glsl_add_src_param(ins
, &ins
->src
[1], WINED3DSP_WRITEMASK_0
, &src1_param
);
2218 shader_addline(buffer
, "vec%u(pow(abs(%s), %s)));\n",
2219 dst_size
, src0_param
.param_str
, src1_param
.param_str
);
2223 shader_addline(buffer
, "pow(abs(%s), %s));\n",
2224 src0_param
.param_str
, src1_param
.param_str
);
2228 /* Process the WINED3DSIO_LOG instruction in GLSL (dst = log2(|src0|))
2229 * Src0 is a scalar. Note that D3D uses the absolute of src0, while
2230 * GLSL uses the value as-is. */
2231 static void shader_glsl_log(const struct wined3d_shader_instruction
*ins
)
2233 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2234 glsl_src_param_t src0_param
;
2235 DWORD dst_write_mask
;
2236 unsigned int dst_size
;
2238 dst_write_mask
= shader_glsl_append_dst(buffer
, ins
);
2239 dst_size
= shader_glsl_get_write_mask_size(dst_write_mask
);
2241 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2245 shader_addline(buffer
, "vec%u(log2(abs(%s))));\n",
2246 dst_size
, src0_param
.param_str
);
2250 shader_addline(buffer
, "log2(abs(%s)));\n",
2251 src0_param
.param_str
);
2255 /* Map the opcode 1-to-1 to the GL code (arg->dst = instruction(src0, src1, ...) */
2256 static void shader_glsl_map2gl(const struct wined3d_shader_instruction
*ins
)
2258 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2259 glsl_src_param_t src_param
;
2260 const char *instruction
;
2264 /* Determine the GLSL function to use based on the opcode */
2265 /* TODO: Possibly make this a table for faster lookups */
2266 switch (ins
->handler_idx
)
2268 case WINED3DSIH_MIN
: instruction
= "min"; break;
2269 case WINED3DSIH_MAX
: instruction
= "max"; break;
2270 case WINED3DSIH_ABS
: instruction
= "abs"; break;
2271 case WINED3DSIH_FRC
: instruction
= "fract"; break;
2272 case WINED3DSIH_EXP
: instruction
= "exp2"; break;
2273 case WINED3DSIH_DSX
: instruction
= "dFdx"; break;
2274 case WINED3DSIH_DSY
: instruction
= "ycorrection.y * dFdy"; break;
2275 default: instruction
= "";
2276 FIXME("Opcode %#x not yet handled in GLSL\n", ins
->handler_idx
);
2280 write_mask
= shader_glsl_append_dst(buffer
, ins
);
2282 shader_addline(buffer
, "%s(", instruction
);
2286 shader_glsl_add_src_param(ins
, &ins
->src
[0], write_mask
, &src_param
);
2287 shader_addline(buffer
, "%s", src_param
.param_str
);
2288 for (i
= 1; i
< ins
->src_count
; ++i
)
2290 shader_glsl_add_src_param(ins
, &ins
->src
[i
], write_mask
, &src_param
);
2291 shader_addline(buffer
, ", %s", src_param
.param_str
);
2295 shader_addline(buffer
, "));\n");
2298 static void shader_glsl_nrm(const struct wined3d_shader_instruction
*ins
)
2300 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2301 glsl_src_param_t src_param
;
2302 unsigned int mask_size
;
2306 write_mask
= shader_glsl_get_write_mask(ins
->dst
, dst_mask
);
2307 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
2308 shader_glsl_add_src_param(ins
, &ins
->src
[0], write_mask
, &src_param
);
2310 shader_addline(buffer
, "tmp0.x = length(%s);\n", src_param
.param_str
);
2311 shader_glsl_append_dst(buffer
, ins
);
2314 shader_addline(buffer
, "tmp0.x == 0.0 ? vec%u(0.0) : (%s / tmp0.x));\n",
2315 mask_size
, src_param
.param_str
);
2319 shader_addline(buffer
, "tmp0.x == 0.0 ? 0.0 : (%s / tmp0.x));\n",
2320 src_param
.param_str
);
2324 /** Process the WINED3DSIO_EXPP instruction in GLSL:
2325 * For shader model 1.x, do the following (and honor the writemask, so use a temporary variable):
2326 * dst.x = 2^(floor(src))
2327 * dst.y = src - floor(src)
2328 * dst.z = 2^src (partial precision is allowed, but optional)
2330 * For 2.0 shaders, just do this (honoring writemask and swizzle):
2331 * dst = 2^src; (partial precision is allowed, but optional)
2333 static void shader_glsl_expp(const struct wined3d_shader_instruction
*ins
)
2335 glsl_src_param_t src_param
;
2337 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
, &src_param
);
2339 if (ins
->ctx
->reg_maps
->shader_version
.major
< 2)
2343 shader_addline(ins
->ctx
->buffer
, "tmp0.x = exp2(floor(%s));\n", src_param
.param_str
);
2344 shader_addline(ins
->ctx
->buffer
, "tmp0.y = %s - floor(%s);\n", src_param
.param_str
, src_param
.param_str
);
2345 shader_addline(ins
->ctx
->buffer
, "tmp0.z = exp2(%s);\n", src_param
.param_str
);
2346 shader_addline(ins
->ctx
->buffer
, "tmp0.w = 1.0;\n");
2348 shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2349 shader_glsl_get_write_mask(&ins
->dst
[0], dst_mask
);
2350 shader_addline(ins
->ctx
->buffer
, "tmp0%s);\n", dst_mask
);
2353 unsigned int mask_size
;
2355 write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2356 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
2358 if (mask_size
> 1) {
2359 shader_addline(ins
->ctx
->buffer
, "vec%d(exp2(%s)));\n", mask_size
, src_param
.param_str
);
2361 shader_addline(ins
->ctx
->buffer
, "exp2(%s));\n", src_param
.param_str
);
2366 /** Process the RCP (reciprocal or inverse) opcode in GLSL (dst = 1 / src) */
2367 static void shader_glsl_rcp(const struct wined3d_shader_instruction
*ins
)
2369 glsl_src_param_t src_param
;
2371 unsigned int mask_size
;
2373 write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2374 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
2375 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_3
, &src_param
);
2379 shader_addline(ins
->ctx
->buffer
, "vec%u(1.0 / %s));\n",
2380 mask_size
, src_param
.param_str
);
2384 shader_addline(ins
->ctx
->buffer
, "1.0 / %s);\n",
2385 src_param
.param_str
);
2389 static void shader_glsl_rsq(const struct wined3d_shader_instruction
*ins
)
2391 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2392 glsl_src_param_t src_param
;
2394 unsigned int mask_size
;
2396 write_mask
= shader_glsl_append_dst(buffer
, ins
);
2397 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
2399 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_3
, &src_param
);
2403 shader_addline(buffer
, "vec%u(inversesqrt(abs(%s))));\n",
2404 mask_size
, src_param
.param_str
);
2408 shader_addline(buffer
, "inversesqrt(abs(%s)));\n",
2409 src_param
.param_str
);
2413 /** Process signed comparison opcodes in GLSL. */
2414 static void shader_glsl_compare(const struct wined3d_shader_instruction
*ins
)
2416 glsl_src_param_t src0_param
;
2417 glsl_src_param_t src1_param
;
2419 unsigned int mask_size
;
2421 write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2422 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
2423 shader_glsl_add_src_param(ins
, &ins
->src
[0], write_mask
, &src0_param
);
2424 shader_glsl_add_src_param(ins
, &ins
->src
[1], write_mask
, &src1_param
);
2426 if (mask_size
> 1) {
2427 const char *compare
;
2429 switch(ins
->handler_idx
)
2431 case WINED3DSIH_SLT
: compare
= "lessThan"; break;
2432 case WINED3DSIH_SGE
: compare
= "greaterThanEqual"; break;
2433 default: compare
= "";
2434 FIXME("Can't handle opcode %#x\n", ins
->handler_idx
);
2437 shader_addline(ins
->ctx
->buffer
, "vec%d(%s(%s, %s)));\n", mask_size
, compare
,
2438 src0_param
.param_str
, src1_param
.param_str
);
2440 switch(ins
->handler_idx
)
2442 case WINED3DSIH_SLT
:
2443 /* Step(src0, src1) is not suitable here because if src0 == src1 SLT is supposed,
2444 * to return 0.0 but step returns 1.0 because step is not < x
2445 * An alternative is a bvec compare padded with an unused second component.
2446 * step(src1 * -1.0, src0 * -1.0) is not an option because it suffers from the same
2447 * issue. Playing with not() is not possible either because not() does not accept
2450 shader_addline(ins
->ctx
->buffer
, "(%s < %s) ? 1.0 : 0.0);\n",
2451 src0_param
.param_str
, src1_param
.param_str
);
2453 case WINED3DSIH_SGE
:
2454 /* Here we can use the step() function and safe a conditional */
2455 shader_addline(ins
->ctx
->buffer
, "step(%s, %s));\n", src1_param
.param_str
, src0_param
.param_str
);
2458 FIXME("Can't handle opcode %#x\n", ins
->handler_idx
);
2464 /** Process CMP instruction in GLSL (dst = src0 >= 0.0 ? src1 : src2), per channel */
2465 static void shader_glsl_cmp(const struct wined3d_shader_instruction
*ins
)
2467 glsl_src_param_t src0_param
;
2468 glsl_src_param_t src1_param
;
2469 glsl_src_param_t src2_param
;
2470 DWORD write_mask
, cmp_channel
= 0;
2473 BOOL temp_destination
= FALSE
;
2475 if (shader_is_scalar(&ins
->src
[0].reg
))
2477 write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2479 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_ALL
, &src0_param
);
2480 shader_glsl_add_src_param(ins
, &ins
->src
[1], write_mask
, &src1_param
);
2481 shader_glsl_add_src_param(ins
, &ins
->src
[2], write_mask
, &src2_param
);
2483 shader_addline(ins
->ctx
->buffer
, "%s >= 0.0 ? %s : %s);\n",
2484 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
2486 DWORD dst_mask
= ins
->dst
[0].write_mask
;
2487 struct wined3d_shader_dst_param dst
= ins
->dst
[0];
2489 /* Cycle through all source0 channels */
2490 for (i
=0; i
<4; i
++) {
2492 /* Find the destination channels which use the current source0 channel */
2493 for (j
=0; j
<4; j
++) {
2494 if (((ins
->src
[0].swizzle
>> (2 * j
)) & 0x3) == i
)
2496 write_mask
|= WINED3DSP_WRITEMASK_0
<< j
;
2497 cmp_channel
= WINED3DSP_WRITEMASK_0
<< j
;
2500 dst
.write_mask
= dst_mask
& write_mask
;
2502 /* Splitting the cmp instruction up in multiple lines imposes a problem:
2503 * The first lines may overwrite source parameters of the following lines.
2504 * Deal with that by using a temporary destination register if needed
2506 if ((ins
->src
[0].reg
.idx
== ins
->dst
[0].reg
.idx
2507 && ins
->src
[0].reg
.type
== ins
->dst
[0].reg
.type
)
2508 || (ins
->src
[1].reg
.idx
== ins
->dst
[0].reg
.idx
2509 && ins
->src
[1].reg
.type
== ins
->dst
[0].reg
.type
)
2510 || (ins
->src
[2].reg
.idx
== ins
->dst
[0].reg
.idx
2511 && ins
->src
[2].reg
.type
== ins
->dst
[0].reg
.type
))
2513 write_mask
= shader_glsl_get_write_mask(&dst
, mask_char
);
2514 if (!write_mask
) continue;
2515 shader_addline(ins
->ctx
->buffer
, "tmp0%s = (", mask_char
);
2516 temp_destination
= TRUE
;
2518 write_mask
= shader_glsl_append_dst_ext(ins
->ctx
->buffer
, ins
, &dst
);
2519 if (!write_mask
) continue;
2522 shader_glsl_add_src_param(ins
, &ins
->src
[0], cmp_channel
, &src0_param
);
2523 shader_glsl_add_src_param(ins
, &ins
->src
[1], write_mask
, &src1_param
);
2524 shader_glsl_add_src_param(ins
, &ins
->src
[2], write_mask
, &src2_param
);
2526 shader_addline(ins
->ctx
->buffer
, "%s >= 0.0 ? %s : %s);\n",
2527 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
2530 if(temp_destination
) {
2531 shader_glsl_get_write_mask(&ins
->dst
[0], mask_char
);
2532 shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2533 shader_addline(ins
->ctx
->buffer
, "tmp0%s);\n", mask_char
);
2539 /** Process the CND opcode in GLSL (dst = (src0 > 0.5) ? src1 : src2) */
2540 /* For ps 1.1-1.3, only a single component of src0 is used. For ps 1.4
2541 * the compare is done per component of src0. */
2542 static void shader_glsl_cnd(const struct wined3d_shader_instruction
*ins
)
2544 struct wined3d_shader_dst_param dst
;
2545 glsl_src_param_t src0_param
;
2546 glsl_src_param_t src1_param
;
2547 glsl_src_param_t src2_param
;
2548 DWORD write_mask
, cmp_channel
= 0;
2551 DWORD shader_version
= WINED3D_SHADER_VERSION(ins
->ctx
->reg_maps
->shader_version
.major
,
2552 ins
->ctx
->reg_maps
->shader_version
.minor
);
2554 if (shader_version
< WINED3D_SHADER_VERSION(1, 4))
2556 write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2557 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2558 shader_glsl_add_src_param(ins
, &ins
->src
[1], write_mask
, &src1_param
);
2559 shader_glsl_add_src_param(ins
, &ins
->src
[2], write_mask
, &src2_param
);
2561 /* Fun: The D3DSI_COISSUE flag changes the semantic of the cnd instruction for < 1.4 shaders */
2564 shader_addline(ins
->ctx
->buffer
, "%s /* COISSUE! */);\n", src1_param
.param_str
);
2566 shader_addline(ins
->ctx
->buffer
, "%s > 0.5 ? %s : %s);\n",
2567 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
2571 /* Cycle through all source0 channels */
2572 dst_mask
= ins
->dst
[0].write_mask
;
2574 for (i
=0; i
<4; i
++) {
2576 /* Find the destination channels which use the current source0 channel */
2577 for (j
=0; j
<4; j
++) {
2578 if (((ins
->src
[0].swizzle
>> (2 * j
)) & 0x3) == i
)
2580 write_mask
|= WINED3DSP_WRITEMASK_0
<< j
;
2581 cmp_channel
= WINED3DSP_WRITEMASK_0
<< j
;
2585 dst
.write_mask
= dst_mask
& write_mask
;
2586 write_mask
= shader_glsl_append_dst_ext(ins
->ctx
->buffer
, ins
, &dst
);
2587 if (!write_mask
) continue;
2589 shader_glsl_add_src_param(ins
, &ins
->src
[0], cmp_channel
, &src0_param
);
2590 shader_glsl_add_src_param(ins
, &ins
->src
[1], write_mask
, &src1_param
);
2591 shader_glsl_add_src_param(ins
, &ins
->src
[2], write_mask
, &src2_param
);
2593 shader_addline(ins
->ctx
->buffer
, "%s > 0.5 ? %s : %s);\n",
2594 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
2598 /** GLSL code generation for WINED3DSIO_MAD: Multiply the first 2 opcodes, then add the last */
2599 static void shader_glsl_mad(const struct wined3d_shader_instruction
*ins
)
2601 glsl_src_param_t src0_param
;
2602 glsl_src_param_t src1_param
;
2603 glsl_src_param_t src2_param
;
2606 write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2607 shader_glsl_add_src_param(ins
, &ins
->src
[0], write_mask
, &src0_param
);
2608 shader_glsl_add_src_param(ins
, &ins
->src
[1], write_mask
, &src1_param
);
2609 shader_glsl_add_src_param(ins
, &ins
->src
[2], write_mask
, &src2_param
);
2610 shader_addline(ins
->ctx
->buffer
, "(%s * %s) + %s);\n",
2611 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
2614 /* Handles transforming all WINED3DSIO_M?x? opcodes for
2615 Vertex shaders to GLSL codes */
2616 static void shader_glsl_mnxn(const struct wined3d_shader_instruction
*ins
)
2619 int nComponents
= 0;
2620 struct wined3d_shader_dst_param tmp_dst
= {{0}};
2621 struct wined3d_shader_src_param tmp_src
[2] = {{{0}}};
2622 struct wined3d_shader_instruction tmp_ins
;
2624 memset(&tmp_ins
, 0, sizeof(tmp_ins
));
2626 /* Set constants for the temporary argument */
2627 tmp_ins
.ctx
= ins
->ctx
;
2628 tmp_ins
.dst_count
= 1;
2629 tmp_ins
.dst
= &tmp_dst
;
2630 tmp_ins
.src_count
= 2;
2631 tmp_ins
.src
= tmp_src
;
2633 switch(ins
->handler_idx
)
2635 case WINED3DSIH_M4x4
:
2637 tmp_ins
.handler_idx
= WINED3DSIH_DP4
;
2639 case WINED3DSIH_M4x3
:
2641 tmp_ins
.handler_idx
= WINED3DSIH_DP4
;
2643 case WINED3DSIH_M3x4
:
2645 tmp_ins
.handler_idx
= WINED3DSIH_DP3
;
2647 case WINED3DSIH_M3x3
:
2649 tmp_ins
.handler_idx
= WINED3DSIH_DP3
;
2651 case WINED3DSIH_M3x2
:
2653 tmp_ins
.handler_idx
= WINED3DSIH_DP3
;
2659 tmp_dst
= ins
->dst
[0];
2660 tmp_src
[0] = ins
->src
[0];
2661 tmp_src
[1] = ins
->src
[1];
2662 for (i
= 0; i
< nComponents
; ++i
)
2664 tmp_dst
.write_mask
= WINED3DSP_WRITEMASK_0
<< i
;
2665 shader_glsl_dot(&tmp_ins
);
2666 ++tmp_src
[1].reg
.idx
;
2671 The LRP instruction performs a component-wise linear interpolation
2672 between the second and third operands using the first operand as the
2673 blend factor. Equation: (dst = src2 + src0 * (src1 - src2))
2674 This is equivalent to mix(src2, src1, src0);
2676 static void shader_glsl_lrp(const struct wined3d_shader_instruction
*ins
)
2678 glsl_src_param_t src0_param
;
2679 glsl_src_param_t src1_param
;
2680 glsl_src_param_t src2_param
;
2683 write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2685 shader_glsl_add_src_param(ins
, &ins
->src
[0], write_mask
, &src0_param
);
2686 shader_glsl_add_src_param(ins
, &ins
->src
[1], write_mask
, &src1_param
);
2687 shader_glsl_add_src_param(ins
, &ins
->src
[2], write_mask
, &src2_param
);
2689 shader_addline(ins
->ctx
->buffer
, "mix(%s, %s, %s));\n",
2690 src2_param
.param_str
, src1_param
.param_str
, src0_param
.param_str
);
2693 /** Process the WINED3DSIO_LIT instruction in GLSL:
2694 * dst.x = dst.w = 1.0
2695 * dst.y = (src0.x > 0) ? src0.x
2696 * dst.z = (src0.x > 0) ? ((src0.y > 0) ? pow(src0.y, src.w) : 0) : 0
2697 * where src.w is clamped at +- 128
2699 static void shader_glsl_lit(const struct wined3d_shader_instruction
*ins
)
2701 glsl_src_param_t src0_param
;
2702 glsl_src_param_t src1_param
;
2703 glsl_src_param_t src3_param
;
2706 shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2707 shader_glsl_get_write_mask(&ins
->dst
[0], dst_mask
);
2709 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2710 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_1
, &src1_param
);
2711 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_3
, &src3_param
);
2713 /* The sdk specifies the instruction like this
2715 * if(src.x > 0.0) dst.y = src.x
2717 * if(src.x > 0.0 && src.y > 0.0) dst.z = pow(src.y, power);
2721 * Obviously that has quite a few conditionals in it which we don't like. So the first step is this:
2722 * dst.x = 1.0 ... No further explanation needed
2723 * dst.y = max(src.y, 0.0); ... If x < 0.0, use 0.0, otherwise x. Same as the conditional
2724 * dst.z = x > 0.0 ? pow(max(y, 0.0), p) : 0; ... 0 ^ power is 0, and otherwise we use y anyway
2725 * dst.w = 1.0. ... Nothing fancy.
2727 * So we still have one conditional in there. So do this:
2728 * dst.z = pow(max(0.0, src.y) * step(0.0, src.x), power);
2730 * 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),
2731 * 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.
2732 * if both x and y are > 0, we get pow(y * 1.0, power), as it is supposed to
2734 shader_addline(ins
->ctx
->buffer
,
2735 "vec4(1.0, max(%s, 0.0), pow(max(0.0, %s) * step(0.0, %s), clamp(%s, -128.0, 128.0)), 1.0)%s);\n",
2736 src0_param
.param_str
, src1_param
.param_str
, src0_param
.param_str
, src3_param
.param_str
, dst_mask
);
2739 /** Process the WINED3DSIO_DST instruction in GLSL:
2741 * dst.y = src0.x * src0.y
2745 static void shader_glsl_dst(const struct wined3d_shader_instruction
*ins
)
2747 glsl_src_param_t src0y_param
;
2748 glsl_src_param_t src0z_param
;
2749 glsl_src_param_t src1y_param
;
2750 glsl_src_param_t src1w_param
;
2753 shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2754 shader_glsl_get_write_mask(&ins
->dst
[0], dst_mask
);
2756 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_1
, &src0y_param
);
2757 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_2
, &src0z_param
);
2758 shader_glsl_add_src_param(ins
, &ins
->src
[1], WINED3DSP_WRITEMASK_1
, &src1y_param
);
2759 shader_glsl_add_src_param(ins
, &ins
->src
[1], WINED3DSP_WRITEMASK_3
, &src1w_param
);
2761 shader_addline(ins
->ctx
->buffer
, "vec4(1.0, %s * %s, %s, %s))%s;\n",
2762 src0y_param
.param_str
, src1y_param
.param_str
, src0z_param
.param_str
, src1w_param
.param_str
, dst_mask
);
2765 /** Process the WINED3DSIO_SINCOS instruction in GLSL:
2766 * VS 2.0 requires that specific cosine and sine constants be passed to this instruction so the hardware
2767 * can handle it. But, these functions are built-in for GLSL, so we can just ignore the last 2 params.
2769 * dst.x = cos(src0.?)
2770 * dst.y = sin(src0.?)
2774 static void shader_glsl_sincos(const struct wined3d_shader_instruction
*ins
)
2776 glsl_src_param_t src0_param
;
2779 write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2780 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2782 switch (write_mask
) {
2783 case WINED3DSP_WRITEMASK_0
:
2784 shader_addline(ins
->ctx
->buffer
, "cos(%s));\n", src0_param
.param_str
);
2787 case WINED3DSP_WRITEMASK_1
:
2788 shader_addline(ins
->ctx
->buffer
, "sin(%s));\n", src0_param
.param_str
);
2791 case (WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
):
2792 shader_addline(ins
->ctx
->buffer
, "vec2(cos(%s), sin(%s)));\n", src0_param
.param_str
, src0_param
.param_str
);
2796 ERR("Write mask should be .x, .y or .xy\n");
2801 /* sgn in vs_2_0 has 2 extra parameters(registers for temporary storage) which we don't use
2802 * here. But those extra parameters require a dedicated function for sgn, since map2gl would
2803 * generate invalid code
2805 static void shader_glsl_sgn(const struct wined3d_shader_instruction
*ins
)
2807 glsl_src_param_t src0_param
;
2810 write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2811 shader_glsl_add_src_param(ins
, &ins
->src
[0], write_mask
, &src0_param
);
2813 shader_addline(ins
->ctx
->buffer
, "sign(%s));\n", src0_param
.param_str
);
2816 /** Process the WINED3DSIO_LOOP instruction in GLSL:
2817 * Start a for() loop where src1.y is the initial value of aL,
2818 * increment aL by src1.z for a total of src1.x iterations.
2819 * Need to use a temporary variable for this operation.
2821 /* FIXME: I don't think nested loops will work correctly this way. */
2822 static void shader_glsl_loop(const struct wined3d_shader_instruction
*ins
)
2824 glsl_src_param_t src1_param
;
2825 IWineD3DBaseShaderImpl
*shader
= (IWineD3DBaseShaderImpl
*)ins
->ctx
->shader
;
2826 const DWORD
*control_values
= NULL
;
2827 const local_constant
*constant
;
2829 shader_glsl_add_src_param(ins
, &ins
->src
[1], WINED3DSP_WRITEMASK_ALL
, &src1_param
);
2831 /* Try to hardcode the loop control parameters if possible. Direct3D 9 class hardware doesn't support real
2832 * varying indexing, but Microsoft designed this feature for Shader model 2.x+. If the loop control is
2833 * known at compile time, the GLSL compiler can unroll the loop, and replace indirect addressing with direct
2836 if (ins
->src
[1].reg
.type
== WINED3DSPR_CONSTINT
)
2838 LIST_FOR_EACH_ENTRY(constant
, &shader
->baseShader
.constantsI
, local_constant
, entry
) {
2839 if (constant
->idx
== ins
->src
[1].reg
.idx
)
2841 control_values
= constant
->value
;
2849 struct wined3d_shader_loop_control loop_control
;
2850 loop_control
.count
= control_values
[0];
2851 loop_control
.start
= control_values
[1];
2852 loop_control
.step
= (int)control_values
[2];
2854 if (loop_control
.step
> 0)
2856 shader_addline(ins
->ctx
->buffer
, "for (aL%u = %u; aL%u < (%u * %d + %u); aL%u += %d) {\n",
2857 shader
->baseShader
.cur_loop_depth
, loop_control
.start
,
2858 shader
->baseShader
.cur_loop_depth
, loop_control
.count
, loop_control
.step
, loop_control
.start
,
2859 shader
->baseShader
.cur_loop_depth
, loop_control
.step
);
2861 else if (loop_control
.step
< 0)
2863 shader_addline(ins
->ctx
->buffer
, "for (aL%u = %u; aL%u > (%u * %d + %u); aL%u += %d) {\n",
2864 shader
->baseShader
.cur_loop_depth
, loop_control
.start
,
2865 shader
->baseShader
.cur_loop_depth
, loop_control
.count
, loop_control
.step
, loop_control
.start
,
2866 shader
->baseShader
.cur_loop_depth
, loop_control
.step
);
2870 shader_addline(ins
->ctx
->buffer
, "for (aL%u = %u, tmpInt%u = 0; tmpInt%u < %u; tmpInt%u++) {\n",
2871 shader
->baseShader
.cur_loop_depth
, loop_control
.start
, shader
->baseShader
.cur_loop_depth
,
2872 shader
->baseShader
.cur_loop_depth
, loop_control
.count
,
2873 shader
->baseShader
.cur_loop_depth
);
2876 shader_addline(ins
->ctx
->buffer
,
2877 "for (tmpInt%u = 0, aL%u = %s.y; tmpInt%u < %s.x; tmpInt%u++, aL%u += %s.z) {\n",
2878 shader
->baseShader
.cur_loop_depth
, shader
->baseShader
.cur_loop_regno
,
2879 src1_param
.reg_name
, shader
->baseShader
.cur_loop_depth
, src1_param
.reg_name
,
2880 shader
->baseShader
.cur_loop_depth
, shader
->baseShader
.cur_loop_regno
, src1_param
.reg_name
);
2883 shader
->baseShader
.cur_loop_depth
++;
2884 shader
->baseShader
.cur_loop_regno
++;
2887 static void shader_glsl_end(const struct wined3d_shader_instruction
*ins
)
2889 IWineD3DBaseShaderImpl
*shader
= (IWineD3DBaseShaderImpl
*)ins
->ctx
->shader
;
2891 shader_addline(ins
->ctx
->buffer
, "}\n");
2893 if (ins
->handler_idx
== WINED3DSIH_ENDLOOP
)
2895 shader
->baseShader
.cur_loop_depth
--;
2896 shader
->baseShader
.cur_loop_regno
--;
2899 if (ins
->handler_idx
== WINED3DSIH_ENDREP
)
2901 shader
->baseShader
.cur_loop_depth
--;
2905 static void shader_glsl_rep(const struct wined3d_shader_instruction
*ins
)
2907 IWineD3DBaseShaderImpl
*shader
= (IWineD3DBaseShaderImpl
*)ins
->ctx
->shader
;
2908 glsl_src_param_t src0_param
;
2909 const DWORD
*control_values
= NULL
;
2910 const local_constant
*constant
;
2912 /* Try to hardcode local values to help the GLSL compiler to unroll and optimize the loop */
2913 if (ins
->src
[0].reg
.type
== WINED3DSPR_CONSTINT
)
2915 LIST_FOR_EACH_ENTRY(constant
, &shader
->baseShader
.constantsI
, local_constant
, entry
)
2917 if (constant
->idx
== ins
->src
[0].reg
.idx
)
2919 control_values
= constant
->value
;
2925 if(control_values
) {
2926 shader_addline(ins
->ctx
->buffer
, "for (tmpInt%d = 0; tmpInt%d < %d; tmpInt%d++) {\n",
2927 shader
->baseShader
.cur_loop_depth
, shader
->baseShader
.cur_loop_depth
,
2928 control_values
[0], shader
->baseShader
.cur_loop_depth
);
2930 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2931 shader_addline(ins
->ctx
->buffer
, "for (tmpInt%d = 0; tmpInt%d < %s; tmpInt%d++) {\n",
2932 shader
->baseShader
.cur_loop_depth
, shader
->baseShader
.cur_loop_depth
,
2933 src0_param
.param_str
, shader
->baseShader
.cur_loop_depth
);
2935 shader
->baseShader
.cur_loop_depth
++;
2938 static void shader_glsl_if(const struct wined3d_shader_instruction
*ins
)
2940 glsl_src_param_t src0_param
;
2942 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2943 shader_addline(ins
->ctx
->buffer
, "if (%s) {\n", src0_param
.param_str
);
2946 static void shader_glsl_ifc(const struct wined3d_shader_instruction
*ins
)
2948 glsl_src_param_t src0_param
;
2949 glsl_src_param_t src1_param
;
2951 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2952 shader_glsl_add_src_param(ins
, &ins
->src
[1], WINED3DSP_WRITEMASK_0
, &src1_param
);
2954 shader_addline(ins
->ctx
->buffer
, "if (%s %s %s) {\n",
2955 src0_param
.param_str
, shader_get_comp_op(ins
->flags
), src1_param
.param_str
);
2958 static void shader_glsl_else(const struct wined3d_shader_instruction
*ins
)
2960 shader_addline(ins
->ctx
->buffer
, "} else {\n");
2963 static void shader_glsl_break(const struct wined3d_shader_instruction
*ins
)
2965 shader_addline(ins
->ctx
->buffer
, "break;\n");
2968 /* FIXME: According to MSDN the compare is done per component. */
2969 static void shader_glsl_breakc(const struct wined3d_shader_instruction
*ins
)
2971 glsl_src_param_t src0_param
;
2972 glsl_src_param_t src1_param
;
2974 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2975 shader_glsl_add_src_param(ins
, &ins
->src
[1], WINED3DSP_WRITEMASK_0
, &src1_param
);
2977 shader_addline(ins
->ctx
->buffer
, "if (%s %s %s) break;\n",
2978 src0_param
.param_str
, shader_get_comp_op(ins
->flags
), src1_param
.param_str
);
2981 static void shader_glsl_label(const struct wined3d_shader_instruction
*ins
)
2983 shader_addline(ins
->ctx
->buffer
, "}\n");
2984 shader_addline(ins
->ctx
->buffer
, "void subroutine%u () {\n", ins
->src
[0].reg
.idx
);
2987 static void shader_glsl_call(const struct wined3d_shader_instruction
*ins
)
2989 shader_addline(ins
->ctx
->buffer
, "subroutine%u();\n", ins
->src
[0].reg
.idx
);
2992 static void shader_glsl_callnz(const struct wined3d_shader_instruction
*ins
)
2994 glsl_src_param_t src1_param
;
2996 shader_glsl_add_src_param(ins
, &ins
->src
[1], WINED3DSP_WRITEMASK_0
, &src1_param
);
2997 shader_addline(ins
->ctx
->buffer
, "if (%s) subroutine%u();\n", src1_param
.param_str
, ins
->src
[0].reg
.idx
);
3000 static void shader_glsl_ret(const struct wined3d_shader_instruction
*ins
)
3002 /* No-op. The closing } is written when a new function is started, and at the end of the shader. This
3003 * function only suppresses the unhandled instruction warning
3007 /*********************************************
3008 * Pixel Shader Specific Code begins here
3009 ********************************************/
3010 static void shader_glsl_tex(const struct wined3d_shader_instruction
*ins
)
3012 IWineD3DBaseShaderImpl
*shader
= (IWineD3DBaseShaderImpl
*)ins
->ctx
->shader
;
3013 IWineD3DDeviceImpl
*deviceImpl
= (IWineD3DDeviceImpl
*)shader
->baseShader
.device
;
3014 DWORD shader_version
= WINED3D_SHADER_VERSION(ins
->ctx
->reg_maps
->shader_version
.major
,
3015 ins
->ctx
->reg_maps
->shader_version
.minor
);
3016 glsl_sample_function_t sample_function
;
3017 IWineD3DBaseTextureImpl
*texture
;
3018 DWORD sample_flags
= 0;
3020 DWORD mask
= 0, swizzle
;
3022 /* 1.0-1.4: Use destination register as sampler source.
3023 * 2.0+: Use provided sampler source. */
3024 if (shader_version
< WINED3D_SHADER_VERSION(2,0)) sampler_idx
= ins
->dst
[0].reg
.idx
;
3025 else sampler_idx
= ins
->src
[1].reg
.idx
;
3026 texture
= deviceImpl
->stateBlock
->state
.textures
[sampler_idx
];
3028 if (shader_version
< WINED3D_SHADER_VERSION(1,4))
3030 const struct shader_glsl_ctx_priv
*priv
= ins
->ctx
->backend_data
;
3031 DWORD flags
= (priv
->cur_ps_args
->tex_transform
>> sampler_idx
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
)
3032 & WINED3D_PSARGS_TEXTRANSFORM_MASK
;
3033 WINED3DSAMPLER_TEXTURE_TYPE sampler_type
= ins
->ctx
->reg_maps
->sampler_type
[sampler_idx
];
3035 /* Projected cube textures don't make a lot of sense, the resulting coordinates stay the same. */
3036 if (flags
& WINED3D_PSARGS_PROJECTED
&& sampler_type
!= WINED3DSTT_CUBE
) {
3037 sample_flags
|= WINED3D_GLSL_SAMPLE_PROJECTED
;
3038 switch (flags
& ~WINED3D_PSARGS_PROJECTED
) {
3039 case WINED3DTTFF_COUNT1
: FIXME("WINED3DTTFF_PROJECTED with WINED3DTTFF_COUNT1?\n"); break;
3040 case WINED3DTTFF_COUNT2
: mask
= WINED3DSP_WRITEMASK_1
; break;
3041 case WINED3DTTFF_COUNT3
: mask
= WINED3DSP_WRITEMASK_2
; break;
3042 case WINED3DTTFF_COUNT4
:
3043 case WINED3DTTFF_DISABLE
: mask
= WINED3DSP_WRITEMASK_3
; break;
3047 else if (shader_version
< WINED3D_SHADER_VERSION(2,0))
3049 DWORD src_mod
= ins
->src
[0].modifiers
;
3051 if (src_mod
== WINED3DSPSM_DZ
) {
3052 sample_flags
|= WINED3D_GLSL_SAMPLE_PROJECTED
;
3053 mask
= WINED3DSP_WRITEMASK_2
;
3054 } else if (src_mod
== WINED3DSPSM_DW
) {
3055 sample_flags
|= WINED3D_GLSL_SAMPLE_PROJECTED
;
3056 mask
= WINED3DSP_WRITEMASK_3
;
3059 if (ins
->flags
& WINED3DSI_TEXLD_PROJECT
)
3061 /* ps 2.0 texldp instruction always divides by the fourth component. */
3062 sample_flags
|= WINED3D_GLSL_SAMPLE_PROJECTED
;
3063 mask
= WINED3DSP_WRITEMASK_3
;
3067 if (texture
&& texture
->baseTexture
.target
== GL_TEXTURE_RECTANGLE_ARB
)
3068 sample_flags
|= WINED3D_GLSL_SAMPLE_RECT
;
3070 shader_glsl_get_sample_function(ins
->ctx
, sampler_idx
, sample_flags
, &sample_function
);
3071 mask
|= sample_function
.coord_mask
;
3073 if (shader_version
< WINED3D_SHADER_VERSION(2,0)) swizzle
= WINED3DSP_NOSWIZZLE
;
3074 else swizzle
= ins
->src
[1].swizzle
;
3076 /* 1.0-1.3: Use destination register as coordinate source.
3077 1.4+: Use provided coordinate source register. */
3078 if (shader_version
< WINED3D_SHADER_VERSION(1,4))
3081 shader_glsl_write_mask_to_str(mask
, coord_mask
);
3082 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, swizzle
, NULL
, NULL
, NULL
,
3083 "T%u%s", sampler_idx
, coord_mask
);
3085 glsl_src_param_t coord_param
;
3086 shader_glsl_add_src_param(ins
, &ins
->src
[0], mask
, &coord_param
);
3087 if (ins
->flags
& WINED3DSI_TEXLD_BIAS
)
3089 glsl_src_param_t bias
;
3090 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_3
, &bias
);
3091 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, swizzle
, NULL
, NULL
, bias
.param_str
,
3092 "%s", coord_param
.param_str
);
3094 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, swizzle
, NULL
, NULL
, NULL
,
3095 "%s", coord_param
.param_str
);
3100 static void shader_glsl_texldd(const struct wined3d_shader_instruction
*ins
)
3102 IWineD3DBaseShaderImpl
*This
= (IWineD3DBaseShaderImpl
*)ins
->ctx
->shader
;
3103 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
3104 const struct wined3d_gl_info
*gl_info
= ins
->ctx
->gl_info
;
3105 glsl_sample_function_t sample_function
;
3106 glsl_src_param_t coord_param
, dx_param
, dy_param
;
3107 DWORD sample_flags
= WINED3D_GLSL_SAMPLE_GRAD
;
3109 DWORD swizzle
= ins
->src
[1].swizzle
;
3110 IWineD3DBaseTextureImpl
*texture
;
3112 if (!gl_info
->supported
[ARB_SHADER_TEXTURE_LOD
] && !gl_info
->supported
[EXT_GPU_SHADER4
])
3114 FIXME("texldd used, but not supported by hardware. Falling back to regular tex\n");
3115 shader_glsl_tex(ins
);
3119 sampler_idx
= ins
->src
[1].reg
.idx
;
3120 texture
= deviceImpl
->stateBlock
->state
.textures
[sampler_idx
];
3121 if (texture
&& texture
->baseTexture
.target
== GL_TEXTURE_RECTANGLE_ARB
)
3122 sample_flags
|= WINED3D_GLSL_SAMPLE_RECT
;
3124 shader_glsl_get_sample_function(ins
->ctx
, sampler_idx
, sample_flags
, &sample_function
);
3125 shader_glsl_add_src_param(ins
, &ins
->src
[0], sample_function
.coord_mask
, &coord_param
);
3126 shader_glsl_add_src_param(ins
, &ins
->src
[2], sample_function
.coord_mask
, &dx_param
);
3127 shader_glsl_add_src_param(ins
, &ins
->src
[3], sample_function
.coord_mask
, &dy_param
);
3129 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, swizzle
, dx_param
.param_str
, dy_param
.param_str
, NULL
,
3130 "%s", coord_param
.param_str
);
3133 static void shader_glsl_texldl(const struct wined3d_shader_instruction
*ins
)
3135 IWineD3DBaseShaderImpl
*This
= (IWineD3DBaseShaderImpl
*)ins
->ctx
->shader
;
3136 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
3137 const struct wined3d_gl_info
*gl_info
= ins
->ctx
->gl_info
;
3138 glsl_sample_function_t sample_function
;
3139 glsl_src_param_t coord_param
, lod_param
;
3140 DWORD sample_flags
= WINED3D_GLSL_SAMPLE_LOD
;
3142 DWORD swizzle
= ins
->src
[1].swizzle
;
3143 IWineD3DBaseTextureImpl
*texture
;
3145 sampler_idx
= ins
->src
[1].reg
.idx
;
3146 texture
= deviceImpl
->stateBlock
->state
.textures
[sampler_idx
];
3147 if (texture
&& texture
->baseTexture
.target
== GL_TEXTURE_RECTANGLE_ARB
)
3148 sample_flags
|= WINED3D_GLSL_SAMPLE_RECT
;
3150 shader_glsl_get_sample_function(ins
->ctx
, sampler_idx
, sample_flags
, &sample_function
);
3151 shader_glsl_add_src_param(ins
, &ins
->src
[0], sample_function
.coord_mask
, &coord_param
);
3153 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_3
, &lod_param
);
3155 if (!gl_info
->supported
[ARB_SHADER_TEXTURE_LOD
] && !gl_info
->supported
[EXT_GPU_SHADER4
]
3156 && shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
.type
))
3158 /* The GLSL spec claims the Lod sampling functions are only supported in vertex shaders.
3159 * However, they seem to work just fine in fragment shaders as well. */
3160 WARN("Using %s in fragment shader.\n", sample_function
.name
);
3162 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, swizzle
, NULL
, NULL
, lod_param
.param_str
,
3163 "%s", coord_param
.param_str
);
3166 static void shader_glsl_texcoord(const struct wined3d_shader_instruction
*ins
)
3168 /* FIXME: Make this work for more than just 2D textures */
3169 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
3170 DWORD write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
3172 if (!(ins
->ctx
->reg_maps
->shader_version
.major
== 1 && ins
->ctx
->reg_maps
->shader_version
.minor
== 4))
3176 shader_glsl_get_write_mask(&ins
->dst
[0], dst_mask
);
3177 shader_addline(buffer
, "clamp(gl_TexCoord[%u], 0.0, 1.0)%s);\n",
3178 ins
->dst
[0].reg
.idx
, dst_mask
);
3180 DWORD reg
= ins
->src
[0].reg
.idx
;
3181 DWORD src_mod
= ins
->src
[0].modifiers
;
3182 char dst_swizzle
[6];
3184 shader_glsl_get_swizzle(&ins
->src
[0], FALSE
, write_mask
, dst_swizzle
);
3186 if (src_mod
== WINED3DSPSM_DZ
) {
3187 glsl_src_param_t div_param
;
3188 unsigned int mask_size
= shader_glsl_get_write_mask_size(write_mask
);
3189 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_2
, &div_param
);
3191 if (mask_size
> 1) {
3192 shader_addline(buffer
, "gl_TexCoord[%u]%s / vec%d(%s));\n", reg
, dst_swizzle
, mask_size
, div_param
.param_str
);
3194 shader_addline(buffer
, "gl_TexCoord[%u]%s / %s);\n", reg
, dst_swizzle
, div_param
.param_str
);
3196 } else if (src_mod
== WINED3DSPSM_DW
) {
3197 glsl_src_param_t div_param
;
3198 unsigned int mask_size
= shader_glsl_get_write_mask_size(write_mask
);
3199 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_3
, &div_param
);
3201 if (mask_size
> 1) {
3202 shader_addline(buffer
, "gl_TexCoord[%u]%s / vec%d(%s));\n", reg
, dst_swizzle
, mask_size
, div_param
.param_str
);
3204 shader_addline(buffer
, "gl_TexCoord[%u]%s / %s);\n", reg
, dst_swizzle
, div_param
.param_str
);
3207 shader_addline(buffer
, "gl_TexCoord[%u]%s);\n", reg
, dst_swizzle
);
3212 /** Process the WINED3DSIO_TEXDP3TEX instruction in GLSL:
3213 * Take a 3-component dot product of the TexCoord[dstreg] and src,
3214 * then perform a 1D texture lookup from stage dstregnum, place into dst. */
3215 static void shader_glsl_texdp3tex(const struct wined3d_shader_instruction
*ins
)
3217 glsl_src_param_t src0_param
;
3218 glsl_sample_function_t sample_function
;
3219 DWORD sampler_idx
= ins
->dst
[0].reg
.idx
;
3220 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
3223 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
3225 /* Do I have to take care about the projected bit? I don't think so, since the dp3 returns only one
3226 * scalar, and projected sampling would require 4.
3228 * It is a dependent read - not valid with conditional NP2 textures
3230 shader_glsl_get_sample_function(ins
->ctx
, sampler_idx
, 0, &sample_function
);
3231 mask_size
= shader_glsl_get_write_mask_size(sample_function
.coord_mask
);
3236 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, WINED3DSP_NOSWIZZLE
, NULL
, NULL
, NULL
,
3237 "dot(gl_TexCoord[%u].xyz, %s)", sampler_idx
, src0_param
.param_str
);
3241 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, WINED3DSP_NOSWIZZLE
, NULL
, NULL
, NULL
,
3242 "vec2(dot(gl_TexCoord[%u].xyz, %s), 0.0)", sampler_idx
, src0_param
.param_str
);
3246 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, WINED3DSP_NOSWIZZLE
, NULL
, NULL
, NULL
,
3247 "vec3(dot(gl_TexCoord[%u].xyz, %s), 0.0, 0.0)", sampler_idx
, src0_param
.param_str
);
3251 FIXME("Unexpected mask size %u\n", mask_size
);
3256 /** Process the WINED3DSIO_TEXDP3 instruction in GLSL:
3257 * Take a 3-component dot product of the TexCoord[dstreg] and src. */
3258 static void shader_glsl_texdp3(const struct wined3d_shader_instruction
*ins
)
3260 glsl_src_param_t src0_param
;
3261 DWORD dstreg
= ins
->dst
[0].reg
.idx
;
3262 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
3264 unsigned int mask_size
;
3266 dst_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
3267 mask_size
= shader_glsl_get_write_mask_size(dst_mask
);
3268 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
3270 if (mask_size
> 1) {
3271 shader_addline(ins
->ctx
->buffer
, "vec%d(dot(T%u.xyz, %s)));\n", mask_size
, dstreg
, src0_param
.param_str
);
3273 shader_addline(ins
->ctx
->buffer
, "dot(T%u.xyz, %s));\n", dstreg
, src0_param
.param_str
);
3277 /** Process the WINED3DSIO_TEXDEPTH instruction in GLSL:
3278 * Calculate the depth as dst.x / dst.y */
3279 static void shader_glsl_texdepth(const struct wined3d_shader_instruction
*ins
)
3281 glsl_dst_param_t dst_param
;
3283 shader_glsl_add_dst_param(ins
, &ins
->dst
[0], &dst_param
);
3285 /* Tests show that texdepth never returns anything below 0.0, and that r5.y is clamped to 1.0.
3286 * Negative input is accepted, -0.25 / -0.5 returns 0.5. GL should clamp gl_FragDepth to [0;1], but
3287 * this doesn't always work, so clamp the results manually. Whether or not the x value is clamped at 1
3288 * too is irrelevant, since if x = 0, any y value < 1.0 (and > 1.0 is not allowed) results in a result
3291 shader_addline(ins
->ctx
->buffer
, "gl_FragDepth = clamp((%s.x / min(%s.y, 1.0)), 0.0, 1.0);\n",
3292 dst_param
.reg_name
, dst_param
.reg_name
);
3295 /** Process the WINED3DSIO_TEXM3X2DEPTH instruction in GLSL:
3296 * Last row of a 3x2 matrix multiply, use the result to calculate the depth:
3297 * Calculate tmp0.y = TexCoord[dstreg] . src.xyz; (tmp0.x has already been calculated)
3298 * depth = (tmp0.y == 0.0) ? 1.0 : tmp0.x / tmp0.y
3300 static void shader_glsl_texm3x2depth(const struct wined3d_shader_instruction
*ins
)
3302 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
3303 DWORD dstreg
= ins
->dst
[0].reg
.idx
;
3304 glsl_src_param_t src0_param
;
3306 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
3308 shader_addline(ins
->ctx
->buffer
, "tmp0.y = dot(T%u.xyz, %s);\n", dstreg
, src0_param
.param_str
);
3309 shader_addline(ins
->ctx
->buffer
, "gl_FragDepth = (tmp0.y == 0.0) ? 1.0 : clamp(tmp0.x / tmp0.y, 0.0, 1.0);\n");
3312 /** Process the WINED3DSIO_TEXM3X2PAD instruction in GLSL
3313 * Calculate the 1st of a 2-row matrix multiplication. */
3314 static void shader_glsl_texm3x2pad(const struct wined3d_shader_instruction
*ins
)
3316 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
3317 DWORD reg
= ins
->dst
[0].reg
.idx
;
3318 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
3319 glsl_src_param_t src0_param
;
3321 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
3322 shader_addline(buffer
, "tmp0.x = dot(T%u.xyz, %s);\n", reg
, src0_param
.param_str
);
3325 /** Process the WINED3DSIO_TEXM3X3PAD instruction in GLSL
3326 * Calculate the 1st or 2nd row of a 3-row matrix multiplication. */
3327 static void shader_glsl_texm3x3pad(const struct wined3d_shader_instruction
*ins
)
3329 IWineD3DBaseShaderImpl
*shader
= (IWineD3DBaseShaderImpl
*)ins
->ctx
->shader
;
3330 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
3331 DWORD reg
= ins
->dst
[0].reg
.idx
;
3332 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
3333 SHADER_PARSE_STATE
* current_state
= &shader
->baseShader
.parse_state
;
3334 glsl_src_param_t src0_param
;
3336 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
3337 shader_addline(buffer
, "tmp0.%c = dot(T%u.xyz, %s);\n", 'x' + current_state
->current_row
, reg
, src0_param
.param_str
);
3338 current_state
->texcoord_w
[current_state
->current_row
++] = reg
;
3341 static void shader_glsl_texm3x2tex(const struct wined3d_shader_instruction
*ins
)
3343 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
3344 DWORD reg
= ins
->dst
[0].reg
.idx
;
3345 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
3346 glsl_src_param_t src0_param
;
3347 glsl_sample_function_t sample_function
;
3349 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
3350 shader_addline(buffer
, "tmp0.y = dot(T%u.xyz, %s);\n", reg
, src0_param
.param_str
);
3352 shader_glsl_get_sample_function(ins
->ctx
, reg
, 0, &sample_function
);
3354 /* Sample the texture using the calculated coordinates */
3355 shader_glsl_gen_sample_code(ins
, reg
, &sample_function
, WINED3DSP_NOSWIZZLE
, NULL
, NULL
, NULL
, "tmp0.xy");
3358 /** Process the WINED3DSIO_TEXM3X3TEX instruction in GLSL
3359 * Perform the 3rd row of a 3x3 matrix multiply, then sample the texture using the calculated coordinates */
3360 static void shader_glsl_texm3x3tex(const struct wined3d_shader_instruction
*ins
)
3362 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
3363 IWineD3DBaseShaderImpl
*shader
= (IWineD3DBaseShaderImpl
*)ins
->ctx
->shader
;
3364 SHADER_PARSE_STATE
*current_state
= &shader
->baseShader
.parse_state
;
3365 glsl_src_param_t src0_param
;
3366 DWORD reg
= ins
->dst
[0].reg
.idx
;
3367 glsl_sample_function_t sample_function
;
3369 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
3370 shader_addline(ins
->ctx
->buffer
, "tmp0.z = dot(T%u.xyz, %s);\n", reg
, src0_param
.param_str
);
3372 /* Dependent read, not valid with conditional NP2 */
3373 shader_glsl_get_sample_function(ins
->ctx
, reg
, 0, &sample_function
);
3375 /* Sample the texture using the calculated coordinates */
3376 shader_glsl_gen_sample_code(ins
, reg
, &sample_function
, WINED3DSP_NOSWIZZLE
, NULL
, NULL
, NULL
, "tmp0.xyz");
3378 current_state
->current_row
= 0;
3381 /** Process the WINED3DSIO_TEXM3X3 instruction in GLSL
3382 * Perform the 3rd row of a 3x3 matrix multiply */
3383 static void shader_glsl_texm3x3(const struct wined3d_shader_instruction
*ins
)
3385 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
3386 IWineD3DBaseShaderImpl
*shader
= (IWineD3DBaseShaderImpl
*)ins
->ctx
->shader
;
3387 SHADER_PARSE_STATE
*current_state
= &shader
->baseShader
.parse_state
;
3388 glsl_src_param_t src0_param
;
3390 DWORD reg
= ins
->dst
[0].reg
.idx
;
3392 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
3394 shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
3395 shader_glsl_get_write_mask(&ins
->dst
[0], dst_mask
);
3396 shader_addline(ins
->ctx
->buffer
, "vec4(tmp0.xy, dot(T%u.xyz, %s), 1.0)%s);\n", reg
, src0_param
.param_str
, dst_mask
);
3398 current_state
->current_row
= 0;
3401 /* Process the WINED3DSIO_TEXM3X3SPEC instruction in GLSL
3402 * Perform the final texture lookup based on the previous 2 3x3 matrix multiplies */
3403 static void shader_glsl_texm3x3spec(const struct wined3d_shader_instruction
*ins
)
3405 IWineD3DBaseShaderImpl
*shader
= (IWineD3DBaseShaderImpl
*)ins
->ctx
->shader
;
3406 DWORD reg
= ins
->dst
[0].reg
.idx
;
3407 glsl_src_param_t src0_param
;
3408 glsl_src_param_t src1_param
;
3409 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
3410 SHADER_PARSE_STATE
* current_state
= &shader
->baseShader
.parse_state
;
3411 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
3412 glsl_sample_function_t sample_function
;
3414 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
3415 shader_glsl_add_src_param(ins
, &ins
->src
[1], src_mask
, &src1_param
);
3417 /* Perform the last matrix multiply operation */
3418 shader_addline(buffer
, "tmp0.z = dot(T%u.xyz, %s);\n", reg
, src0_param
.param_str
);
3419 /* Reflection calculation */
3420 shader_addline(buffer
, "tmp0.xyz = -reflect((%s), normalize(tmp0.xyz));\n", src1_param
.param_str
);
3422 /* Dependent read, not valid with conditional NP2 */
3423 shader_glsl_get_sample_function(ins
->ctx
, reg
, 0, &sample_function
);
3425 /* Sample the texture */
3426 shader_glsl_gen_sample_code(ins
, reg
, &sample_function
, WINED3DSP_NOSWIZZLE
, NULL
, NULL
, NULL
, "tmp0.xyz");
3428 current_state
->current_row
= 0;
3431 /* Process the WINED3DSIO_TEXM3X3VSPEC instruction in GLSL
3432 * Perform the final texture lookup based on the previous 2 3x3 matrix multiplies */
3433 static void shader_glsl_texm3x3vspec(const struct wined3d_shader_instruction
*ins
)
3435 IWineD3DBaseShaderImpl
*shader
= (IWineD3DBaseShaderImpl
*)ins
->ctx
->shader
;
3436 DWORD reg
= ins
->dst
[0].reg
.idx
;
3437 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
3438 SHADER_PARSE_STATE
* current_state
= &shader
->baseShader
.parse_state
;
3439 glsl_src_param_t src0_param
;
3440 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
3441 glsl_sample_function_t sample_function
;
3443 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
3445 /* Perform the last matrix multiply operation */
3446 shader_addline(buffer
, "tmp0.z = dot(vec3(T%u), vec3(%s));\n", reg
, src0_param
.param_str
);
3448 /* Construct the eye-ray vector from w coordinates */
3449 shader_addline(buffer
, "tmp1.xyz = normalize(vec3(gl_TexCoord[%u].w, gl_TexCoord[%u].w, gl_TexCoord[%u].w));\n",
3450 current_state
->texcoord_w
[0], current_state
->texcoord_w
[1], reg
);
3451 shader_addline(buffer
, "tmp0.xyz = -reflect(tmp1.xyz, normalize(tmp0.xyz));\n");
3453 /* Dependent read, not valid with conditional NP2 */
3454 shader_glsl_get_sample_function(ins
->ctx
, reg
, 0, &sample_function
);
3456 /* Sample the texture using the calculated coordinates */
3457 shader_glsl_gen_sample_code(ins
, reg
, &sample_function
, WINED3DSP_NOSWIZZLE
, NULL
, NULL
, NULL
, "tmp0.xyz");
3459 current_state
->current_row
= 0;
3462 /** Process the WINED3DSIO_TEXBEM instruction in GLSL.
3463 * Apply a fake bump map transform.
3464 * texbem is pshader <= 1.3 only, this saves a few version checks
3466 static void shader_glsl_texbem(const struct wined3d_shader_instruction
*ins
)
3468 const struct shader_glsl_ctx_priv
*priv
= ins
->ctx
->backend_data
;
3469 glsl_sample_function_t sample_function
;
3470 glsl_src_param_t coord_param
;
3476 sampler_idx
= ins
->dst
[0].reg
.idx
;
3477 flags
= (priv
->cur_ps_args
->tex_transform
>> sampler_idx
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
)
3478 & WINED3D_PSARGS_TEXTRANSFORM_MASK
;
3480 /* Dependent read, not valid with conditional NP2 */
3481 shader_glsl_get_sample_function(ins
->ctx
, sampler_idx
, 0, &sample_function
);
3482 mask
= sample_function
.coord_mask
;
3484 shader_glsl_write_mask_to_str(mask
, coord_mask
);
3486 /* with projective textures, texbem only divides the static texture coord, not the displacement,
3487 * so we can't let the GL handle this.
3489 if (flags
& WINED3D_PSARGS_PROJECTED
) {
3491 char coord_div_mask
[3];
3492 switch (flags
& ~WINED3D_PSARGS_PROJECTED
) {
3493 case WINED3DTTFF_COUNT1
: FIXME("WINED3DTTFF_PROJECTED with WINED3DTTFF_COUNT1?\n"); break;
3494 case WINED3DTTFF_COUNT2
: div_mask
= WINED3DSP_WRITEMASK_1
; break;
3495 case WINED3DTTFF_COUNT3
: div_mask
= WINED3DSP_WRITEMASK_2
; break;
3496 case WINED3DTTFF_COUNT4
:
3497 case WINED3DTTFF_DISABLE
: div_mask
= WINED3DSP_WRITEMASK_3
; break;
3499 shader_glsl_write_mask_to_str(div_mask
, coord_div_mask
);
3500 shader_addline(ins
->ctx
->buffer
, "T%u%s /= T%u%s;\n", sampler_idx
, coord_mask
, sampler_idx
, coord_div_mask
);
3503 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
, &coord_param
);
3505 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, WINED3DSP_NOSWIZZLE
, NULL
, NULL
, NULL
,
3506 "T%u%s + vec4(bumpenvmat%d * %s, 0.0, 0.0)%s", sampler_idx
, coord_mask
, sampler_idx
,
3507 coord_param
.param_str
, coord_mask
);
3509 if (ins
->handler_idx
== WINED3DSIH_TEXBEML
)
3511 glsl_src_param_t luminance_param
;
3512 glsl_dst_param_t dst_param
;
3514 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_2
, &luminance_param
);
3515 shader_glsl_add_dst_param(ins
, &ins
->dst
[0], &dst_param
);
3517 shader_addline(ins
->ctx
->buffer
, "%s%s *= (%s * luminancescale%d + luminanceoffset%d);\n",
3518 dst_param
.reg_name
, dst_param
.mask_str
,
3519 luminance_param
.param_str
, sampler_idx
, sampler_idx
);
3523 static void shader_glsl_bem(const struct wined3d_shader_instruction
*ins
)
3525 glsl_src_param_t src0_param
, src1_param
;
3526 DWORD sampler_idx
= ins
->dst
[0].reg
.idx
;
3528 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
, &src0_param
);
3529 shader_glsl_add_src_param(ins
, &ins
->src
[1], WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
, &src1_param
);
3531 shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
3532 shader_addline(ins
->ctx
->buffer
, "%s + bumpenvmat%d * %s);\n",
3533 src0_param
.param_str
, sampler_idx
, src1_param
.param_str
);
3536 /** Process the WINED3DSIO_TEXREG2AR instruction in GLSL
3537 * Sample 2D texture at dst using the alpha & red (wx) components of src as texture coordinates */
3538 static void shader_glsl_texreg2ar(const struct wined3d_shader_instruction
*ins
)
3540 glsl_src_param_t src0_param
;
3541 DWORD sampler_idx
= ins
->dst
[0].reg
.idx
;
3542 glsl_sample_function_t sample_function
;
3544 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_ALL
, &src0_param
);
3546 shader_glsl_get_sample_function(ins
->ctx
, sampler_idx
, 0, &sample_function
);
3547 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, WINED3DSP_NOSWIZZLE
, NULL
, NULL
, NULL
,
3548 "%s.wx", src0_param
.reg_name
);
3551 /** Process the WINED3DSIO_TEXREG2GB instruction in GLSL
3552 * Sample 2D texture at dst using the green & blue (yz) components of src as texture coordinates */
3553 static void shader_glsl_texreg2gb(const struct wined3d_shader_instruction
*ins
)
3555 glsl_src_param_t src0_param
;
3556 DWORD sampler_idx
= ins
->dst
[0].reg
.idx
;
3557 glsl_sample_function_t sample_function
;
3559 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_ALL
, &src0_param
);
3561 shader_glsl_get_sample_function(ins
->ctx
, sampler_idx
, 0, &sample_function
);
3562 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, WINED3DSP_NOSWIZZLE
, NULL
, NULL
, NULL
,
3563 "%s.yz", src0_param
.reg_name
);
3566 /** Process the WINED3DSIO_TEXREG2RGB instruction in GLSL
3567 * Sample texture at dst using the rgb (xyz) components of src as texture coordinates */
3568 static void shader_glsl_texreg2rgb(const struct wined3d_shader_instruction
*ins
)
3570 glsl_src_param_t src0_param
;
3571 DWORD sampler_idx
= ins
->dst
[0].reg
.idx
;
3572 glsl_sample_function_t sample_function
;
3574 /* Dependent read, not valid with conditional NP2 */
3575 shader_glsl_get_sample_function(ins
->ctx
, sampler_idx
, 0, &sample_function
);
3576 shader_glsl_add_src_param(ins
, &ins
->src
[0], sample_function
.coord_mask
, &src0_param
);
3578 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, WINED3DSP_NOSWIZZLE
, NULL
, NULL
, NULL
,
3579 "%s", src0_param
.param_str
);
3582 /** Process the WINED3DSIO_TEXKILL instruction in GLSL.
3583 * If any of the first 3 components are < 0, discard this pixel */
3584 static void shader_glsl_texkill(const struct wined3d_shader_instruction
*ins
)
3586 glsl_dst_param_t dst_param
;
3588 /* The argument is a destination parameter, and no writemasks are allowed */
3589 shader_glsl_add_dst_param(ins
, &ins
->dst
[0], &dst_param
);
3590 if (ins
->ctx
->reg_maps
->shader_version
.major
>= 2)
3592 /* 2.0 shaders compare all 4 components in texkill */
3593 shader_addline(ins
->ctx
->buffer
, "if (any(lessThan(%s.xyzw, vec4(0.0)))) discard;\n", dst_param
.reg_name
);
3595 /* 1.X shaders only compare the first 3 components, probably due to the nature of the texkill
3596 * instruction as a tex* instruction, and phase, which kills all a / w components. Even if all
3597 * 4 components are defined, only the first 3 are used
3599 shader_addline(ins
->ctx
->buffer
, "if (any(lessThan(%s.xyz, vec3(0.0)))) discard;\n", dst_param
.reg_name
);
3603 /** Process the WINED3DSIO_DP2ADD instruction in GLSL.
3604 * dst = dot2(src0, src1) + src2 */
3605 static void shader_glsl_dp2add(const struct wined3d_shader_instruction
*ins
)
3607 glsl_src_param_t src0_param
;
3608 glsl_src_param_t src1_param
;
3609 glsl_src_param_t src2_param
;
3611 unsigned int mask_size
;
3613 write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
3614 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
3616 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
, &src0_param
);
3617 shader_glsl_add_src_param(ins
, &ins
->src
[1], WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
, &src1_param
);
3618 shader_glsl_add_src_param(ins
, &ins
->src
[2], WINED3DSP_WRITEMASK_0
, &src2_param
);
3620 if (mask_size
> 1) {
3621 shader_addline(ins
->ctx
->buffer
, "vec%d(dot(%s, %s) + %s));\n",
3622 mask_size
, src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
3624 shader_addline(ins
->ctx
->buffer
, "dot(%s, %s) + %s);\n",
3625 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
3629 static void shader_glsl_input_pack(IWineD3DPixelShader
*iface
, struct wined3d_shader_buffer
*buffer
,
3630 const struct wined3d_shader_signature_element
*input_signature
, const struct shader_reg_maps
*reg_maps
,
3631 enum vertexprocessing_mode vertexprocessing
)
3634 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*)iface
;
3635 WORD map
= reg_maps
->input_registers
;
3637 for (i
= 0; map
; map
>>= 1, ++i
)
3639 const char *semantic_name
;
3644 if (!(map
& 1)) continue;
3646 semantic_name
= input_signature
[i
].semantic_name
;
3647 semantic_idx
= input_signature
[i
].semantic_idx
;
3648 shader_glsl_write_mask_to_str(input_signature
[i
].mask
, reg_mask
);
3650 if (shader_match_semantic(semantic_name
, WINED3DDECLUSAGE_TEXCOORD
))
3652 if (semantic_idx
< 8 && vertexprocessing
== pretransformed
)
3653 shader_addline(buffer
, "IN[%u]%s = gl_TexCoord[%u]%s;\n",
3654 This
->input_reg_map
[i
], reg_mask
, semantic_idx
, reg_mask
);
3656 shader_addline(buffer
, "IN[%u]%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
3657 This
->input_reg_map
[i
], reg_mask
, reg_mask
);
3659 else if (shader_match_semantic(semantic_name
, WINED3DDECLUSAGE_COLOR
))
3662 shader_addline(buffer
, "IN[%u]%s = vec4(gl_Color)%s;\n",
3663 This
->input_reg_map
[i
], reg_mask
, reg_mask
);
3664 else if (semantic_idx
== 1)
3665 shader_addline(buffer
, "IN[%u]%s = vec4(gl_SecondaryColor)%s;\n",
3666 This
->input_reg_map
[i
], reg_mask
, reg_mask
);
3668 shader_addline(buffer
, "IN[%u]%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
3669 This
->input_reg_map
[i
], reg_mask
, reg_mask
);
3673 shader_addline(buffer
, "IN[%u]%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
3674 This
->input_reg_map
[i
], reg_mask
, reg_mask
);
3679 /*********************************************
3680 * Vertex Shader Specific Code begins here
3681 ********************************************/
3683 static void add_glsl_program_entry(struct shader_glsl_priv
*priv
, struct glsl_shader_prog_link
*entry
) {
3684 glsl_program_key_t key
;
3686 key
.vshader
= entry
->vshader
;
3687 key
.pshader
= entry
->pshader
;
3688 key
.vs_args
= entry
->vs_args
;
3689 key
.ps_args
= entry
->ps_args
;
3691 if (wine_rb_put(&priv
->program_lookup
, &key
, &entry
->program_lookup_entry
) == -1)
3693 ERR("Failed to insert program entry.\n");
3697 static struct glsl_shader_prog_link
*get_glsl_program_entry(struct shader_glsl_priv
*priv
,
3698 IWineD3DVertexShader
*vshader
, IWineD3DPixelShader
*pshader
, struct vs_compile_args
*vs_args
,
3699 struct ps_compile_args
*ps_args
) {
3700 struct wine_rb_entry
*entry
;
3701 glsl_program_key_t key
;
3703 key
.vshader
= vshader
;
3704 key
.pshader
= pshader
;
3705 key
.vs_args
= *vs_args
;
3706 key
.ps_args
= *ps_args
;
3708 entry
= wine_rb_get(&priv
->program_lookup
, &key
);
3709 return entry
? WINE_RB_ENTRY_VALUE(entry
, struct glsl_shader_prog_link
, program_lookup_entry
) : NULL
;
3712 /* GL locking is done by the caller */
3713 static void delete_glsl_program_entry(struct shader_glsl_priv
*priv
, const struct wined3d_gl_info
*gl_info
,
3714 struct glsl_shader_prog_link
*entry
)
3716 glsl_program_key_t key
;
3718 key
.vshader
= entry
->vshader
;
3719 key
.pshader
= entry
->pshader
;
3720 key
.vs_args
= entry
->vs_args
;
3721 key
.ps_args
= entry
->ps_args
;
3722 wine_rb_remove(&priv
->program_lookup
, &key
);
3724 GL_EXTCALL(glDeleteObjectARB(entry
->programId
));
3725 if (entry
->vshader
) list_remove(&entry
->vshader_entry
);
3726 if (entry
->pshader
) list_remove(&entry
->pshader_entry
);
3727 HeapFree(GetProcessHeap(), 0, entry
->vuniformF_locations
);
3728 HeapFree(GetProcessHeap(), 0, entry
->puniformF_locations
);
3729 HeapFree(GetProcessHeap(), 0, entry
);
3732 static void handle_ps3_input(struct wined3d_shader_buffer
*buffer
, const struct wined3d_gl_info
*gl_info
, const DWORD
*map
,
3733 const struct wined3d_shader_signature_element
*input_signature
, const struct shader_reg_maps
*reg_maps_in
,
3734 const struct wined3d_shader_signature_element
*output_signature
, const struct shader_reg_maps
*reg_maps_out
)
3737 const char *semantic_name_in
;
3738 UINT semantic_idx_in
;
3741 unsigned int in_count
= vec4_varyings(3, gl_info
);
3743 char destination
[50];
3744 WORD input_map
, output_map
;
3746 set
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*set
) * (in_count
+ 2));
3748 input_map
= reg_maps_in
->input_registers
;
3749 for (i
= 0; input_map
; input_map
>>= 1, ++i
)
3751 if (!(input_map
& 1)) continue;
3754 /* Declared, but not read register */
3755 if (in_idx
== ~0U) continue;
3756 if (in_idx
>= (in_count
+ 2))
3758 FIXME("More input varyings declared than supported, expect issues.\n");
3762 if (in_idx
== in_count
) {
3763 sprintf(destination
, "gl_FrontColor");
3764 } else if (in_idx
== in_count
+ 1) {
3765 sprintf(destination
, "gl_FrontSecondaryColor");
3767 sprintf(destination
, "IN[%u]", in_idx
);
3770 semantic_name_in
= input_signature
[i
].semantic_name
;
3771 semantic_idx_in
= input_signature
[i
].semantic_idx
;
3774 output_map
= reg_maps_out
->output_registers
;
3775 for (j
= 0; output_map
; output_map
>>= 1, ++j
)
3779 if (!(output_map
& 1)
3780 || semantic_idx_in
!= output_signature
[j
].semantic_idx
3781 || strcmp(semantic_name_in
, output_signature
[j
].semantic_name
)
3782 || !(mask
= input_signature
[i
].mask
& output_signature
[j
].mask
))
3786 shader_glsl_write_mask_to_str(mask
, reg_mask
);
3788 shader_addline(buffer
, "%s%s = clamp(OUT[%u]%s, -FLT_MAX, FLT_MAX);\n",
3789 destination
, reg_mask
, j
, reg_mask
);
3793 for (i
= 0; i
< in_count
+ 2; ++i
)
3797 if (!set
[i
] || set
[i
] == WINED3DSP_WRITEMASK_ALL
)
3800 if (set
[i
] == ~0U) set
[i
] = 0;
3803 if (!(set
[i
] & WINED3DSP_WRITEMASK_0
)) reg_mask
[size
++] = 'x';
3804 if (!(set
[i
] & WINED3DSP_WRITEMASK_1
)) reg_mask
[size
++] = 'y';
3805 if (!(set
[i
] & WINED3DSP_WRITEMASK_2
)) reg_mask
[size
++] = 'z';
3806 if (!(set
[i
] & WINED3DSP_WRITEMASK_3
)) reg_mask
[size
++] = 'w';
3807 reg_mask
[size
] = '\0';
3809 if (i
== in_count
) sprintf(destination
, "gl_FrontColor");
3810 else if (i
== in_count
+ 1) sprintf(destination
, "gl_FrontSecondaryColor");
3811 else sprintf(destination
, "IN[%u]", i
);
3813 if (size
== 1) shader_addline(buffer
, "%s.%s = 0.0;\n", destination
, reg_mask
);
3814 else shader_addline(buffer
, "%s.%s = vec%u(0.0);\n", destination
, reg_mask
, size
);
3817 HeapFree(GetProcessHeap(), 0, set
);
3820 /* GL locking is done by the caller */
3821 static GLhandleARB
generate_param_reorder_function(struct wined3d_shader_buffer
*buffer
,
3822 IWineD3DVertexShader
*vertexshader
, IWineD3DPixelShader
*pixelshader
, const struct wined3d_gl_info
*gl_info
)
3824 GLhandleARB ret
= 0;
3825 IWineD3DVertexShaderImpl
*vs
= (IWineD3DVertexShaderImpl
*) vertexshader
;
3826 IWineD3DPixelShaderImpl
*ps
= (IWineD3DPixelShaderImpl
*) pixelshader
;
3827 DWORD ps_major
= ps
? ps
->baseShader
.reg_maps
.shader_version
.major
: 0;
3829 const char *semantic_name
;
3832 const struct wined3d_shader_signature_element
*output_signature
= vs
->baseShader
.output_signature
;
3833 WORD map
= vs
->baseShader
.reg_maps
.output_registers
;
3835 shader_buffer_clear(buffer
);
3837 shader_addline(buffer
, "#version 120\n");
3838 shader_addline(buffer
, "const float FLT_MAX = 1e38;\n");
3842 shader_addline(buffer
, "void order_ps_input(in vec4 OUT[%u]) {\n", MAX_REG_OUTPUT
);
3844 for (i
= 0; map
; map
>>= 1, ++i
)
3848 if (!(map
& 1)) continue;
3850 semantic_name
= output_signature
[i
].semantic_name
;
3851 semantic_idx
= output_signature
[i
].semantic_idx
;
3852 write_mask
= output_signature
[i
].mask
;
3853 shader_glsl_write_mask_to_str(write_mask
, reg_mask
);
3855 if (shader_match_semantic(semantic_name
, WINED3DDECLUSAGE_COLOR
))
3858 shader_addline(buffer
, "gl_FrontColor%s = clamp(OUT[%u]%s, -FLT_MAX, FLT_MAX);\n",
3859 reg_mask
, i
, reg_mask
);
3860 else if (semantic_idx
== 1)
3861 shader_addline(buffer
, "gl_FrontSecondaryColor%s = clamp(OUT[%u]%s, -FLT_MAX, FLT_MAX);\n",
3862 reg_mask
, i
, reg_mask
);
3864 else if (shader_match_semantic(semantic_name
, WINED3DDECLUSAGE_POSITION
))
3866 shader_addline(buffer
, "gl_Position%s = clamp(OUT[%u]%s, -FLT_MAX, FLT_MAX);\n",
3867 reg_mask
, i
, reg_mask
);
3869 else if (shader_match_semantic(semantic_name
, WINED3DDECLUSAGE_TEXCOORD
))
3871 if (semantic_idx
< 8)
3873 if (!(gl_info
->quirks
& WINED3D_QUIRK_SET_TEXCOORD_W
) || ps_major
> 0)
3874 write_mask
|= WINED3DSP_WRITEMASK_3
;
3876 shader_addline(buffer
, "gl_TexCoord[%u]%s = clamp(OUT[%u]%s, -FLT_MAX, FLT_MAX);\n",
3877 semantic_idx
, reg_mask
, i
, reg_mask
);
3878 if (!(write_mask
& WINED3DSP_WRITEMASK_3
))
3879 shader_addline(buffer
, "gl_TexCoord[%u].w = 1.0;\n", semantic_idx
);
3882 else if (shader_match_semantic(semantic_name
, WINED3DDECLUSAGE_PSIZE
))
3884 shader_addline(buffer
, "gl_PointSize = clamp(OUT[%u].%c, -FLT_MAX, FLT_MAX);\n", i
, reg_mask
[1]);
3886 else if (shader_match_semantic(semantic_name
, WINED3DDECLUSAGE_FOG
))
3888 shader_addline(buffer
, "gl_FogFragCoord = clamp(OUT[%u].%c, -FLT_MAX, FLT_MAX);\n", i
, reg_mask
[1]);
3891 shader_addline(buffer
, "}\n");
3896 /* This one is tricky: a 3.0 pixel shader reads from a 3.0 vertex shader */
3897 shader_addline(buffer
, "varying vec4 IN[%u];\n", vec4_varyings(3, gl_info
));
3898 shader_addline(buffer
, "void order_ps_input(in vec4 OUT[%u]) {\n", MAX_REG_OUTPUT
);
3900 /* First, sort out position and point size. Those are not passed to the pixel shader */
3901 for (i
= 0; map
; map
>>= 1, ++i
)
3903 if (!(map
& 1)) continue;
3905 semantic_name
= output_signature
[i
].semantic_name
;
3906 shader_glsl_write_mask_to_str(output_signature
[i
].mask
, reg_mask
);
3908 if (shader_match_semantic(semantic_name
, WINED3DDECLUSAGE_POSITION
))
3910 shader_addline(buffer
, "gl_Position%s = clamp(OUT[%u]%s, -FLT_MAX, FLT_MAX);\n",
3911 reg_mask
, i
, reg_mask
);
3913 else if (shader_match_semantic(semantic_name
, WINED3DDECLUSAGE_PSIZE
))
3915 shader_addline(buffer
, "gl_PointSize = clamp(OUT[%u].%c, -FLT_MAX, FLT_MAX);\n", i
, reg_mask
[1]);
3919 /* Then, fix the pixel shader input */
3920 handle_ps3_input(buffer
, gl_info
, ps
->input_reg_map
, ps
->baseShader
.input_signature
,
3921 &ps
->baseShader
.reg_maps
, output_signature
, &vs
->baseShader
.reg_maps
);
3923 shader_addline(buffer
, "}\n");
3926 ret
= GL_EXTCALL(glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB
));
3927 checkGLcall("glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB)");
3928 shader_glsl_compile(gl_info
, ret
, buffer
->buffer
);
3933 /* GL locking is done by the caller */
3934 static void hardcode_local_constants(IWineD3DBaseShaderImpl
*shader
, const struct wined3d_gl_info
*gl_info
,
3935 GLhandleARB programId
, char prefix
)
3937 const local_constant
*lconst
;
3942 LIST_FOR_EACH_ENTRY(lconst
, &shader
->baseShader
.constantsF
, local_constant
, entry
) {
3943 value
= (const float *)lconst
->value
;
3944 snprintf(glsl_name
, sizeof(glsl_name
), "%cLC%u", prefix
, lconst
->idx
);
3945 tmp_loc
= GL_EXTCALL(glGetUniformLocationARB(programId
, glsl_name
));
3946 GL_EXTCALL(glUniform4fvARB(tmp_loc
, 1, value
));
3948 checkGLcall("Hardcoding local constants");
3951 /* GL locking is done by the caller */
3952 static GLuint
shader_glsl_generate_pshader(const struct wined3d_context
*context
,
3953 struct wined3d_shader_buffer
*buffer
, IWineD3DPixelShaderImpl
*This
,
3954 const struct ps_compile_args
*args
, struct ps_np2fixup_info
*np2fixup_info
)
3956 const struct shader_reg_maps
*reg_maps
= &This
->baseShader
.reg_maps
;
3957 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
3958 CONST DWORD
*function
= This
->baseShader
.function
;
3959 struct shader_glsl_ctx_priv priv_ctx
;
3961 /* Create the hw GLSL shader object and assign it as the shader->prgId */
3962 GLhandleARB shader_obj
= GL_EXTCALL(glCreateShaderObjectARB(GL_FRAGMENT_SHADER_ARB
));
3964 memset(&priv_ctx
, 0, sizeof(priv_ctx
));
3965 priv_ctx
.cur_ps_args
= args
;
3966 priv_ctx
.cur_np2fixup_info
= np2fixup_info
;
3968 shader_addline(buffer
, "#version 120\n");
3970 if (gl_info
->supported
[ARB_SHADER_TEXTURE_LOD
] && reg_maps
->usestexldd
)
3972 shader_addline(buffer
, "#extension GL_ARB_shader_texture_lod : enable\n");
3974 if (gl_info
->supported
[ARB_TEXTURE_RECTANGLE
])
3976 /* The spec says that it doesn't have to be explicitly enabled, but the nvidia
3977 * drivers write a warning if we don't do so
3979 shader_addline(buffer
, "#extension GL_ARB_texture_rectangle : enable\n");
3981 if (gl_info
->supported
[EXT_GPU_SHADER4
])
3983 shader_addline(buffer
, "#extension GL_EXT_gpu_shader4 : enable\n");
3986 /* Base Declarations */
3987 shader_generate_glsl_declarations(context
, buffer
, (IWineD3DBaseShader
*)This
, reg_maps
, &priv_ctx
);
3989 /* Pack 3.0 inputs */
3990 if (reg_maps
->shader_version
.major
>= 3 && args
->vp_mode
!= vertexshader
)
3992 shader_glsl_input_pack((IWineD3DPixelShader
*) This
, buffer
,
3993 This
->baseShader
.input_signature
, reg_maps
, args
->vp_mode
);
3996 /* Base Shader Body */
3997 shader_generate_main((IWineD3DBaseShader
*)This
, buffer
, reg_maps
, function
, &priv_ctx
);
3999 /* Pixel shaders < 2.0 place the resulting color in R0 implicitly */
4000 if (reg_maps
->shader_version
.major
< 2)
4002 /* Some older cards like GeforceFX ones don't support multiple buffers, so also not gl_FragData */
4003 shader_addline(buffer
, "gl_FragData[0] = R0;\n");
4006 if (args
->srgb_correction
)
4008 shader_addline(buffer
, "tmp0.xyz = pow(gl_FragData[0].xyz, vec3(srgb_const0.x));\n");
4009 shader_addline(buffer
, "tmp0.xyz = tmp0.xyz * vec3(srgb_const0.y) - vec3(srgb_const0.z);\n");
4010 shader_addline(buffer
, "tmp1.xyz = gl_FragData[0].xyz * vec3(srgb_const0.w);\n");
4011 shader_addline(buffer
, "bvec3 srgb_compare = lessThan(gl_FragData[0].xyz, vec3(srgb_const1.x));\n");
4012 shader_addline(buffer
, "gl_FragData[0].xyz = mix(tmp0.xyz, tmp1.xyz, vec3(srgb_compare));\n");
4013 shader_addline(buffer
, "gl_FragData[0] = clamp(gl_FragData[0], 0.0, 1.0);\n");
4015 /* Pixel shader < 3.0 do not replace the fog stage.
4016 * This implements linear fog computation and blending.
4017 * TODO: non linear fog
4018 * NOTE: gl_Fog.start and gl_Fog.end don't hold fog start s and end e but
4019 * -1/(e-s) and e/(e-s) respectively.
4021 if (reg_maps
->shader_version
.major
< 3)
4024 case FOG_OFF
: break;
4026 shader_addline(buffer
, "float fogstart = -1.0 / (gl_Fog.end - gl_Fog.start);\n");
4027 shader_addline(buffer
, "float fogend = gl_Fog.end * -fogstart;\n");
4028 shader_addline(buffer
, "float Fog = clamp(gl_FogFragCoord * fogstart + fogend, 0.0, 1.0);\n");
4029 shader_addline(buffer
, "gl_FragData[0].xyz = mix(gl_Fog.color.xyz, gl_FragData[0].xyz, Fog);\n");
4032 /* Fog = e^(-gl_Fog.density * gl_FogFragCoord) */
4033 shader_addline(buffer
, "float Fog = exp(-gl_Fog.density * gl_FogFragCoord);\n");
4034 shader_addline(buffer
, "Fog = clamp(Fog, 0.0, 1.0);\n");
4035 shader_addline(buffer
, "gl_FragData[0].xyz = mix(gl_Fog.color.xyz, gl_FragData[0].xyz, Fog);\n");
4038 /* Fog = e^(-(gl_Fog.density * gl_FogFragCoord)^2) */
4039 shader_addline(buffer
, "float Fog = exp(-gl_Fog.density * gl_Fog.density * gl_FogFragCoord * gl_FogFragCoord);\n");
4040 shader_addline(buffer
, "Fog = clamp(Fog, 0.0, 1.0);\n");
4041 shader_addline(buffer
, "gl_FragData[0].xyz = mix(gl_Fog.color.xyz, gl_FragData[0].xyz, Fog);\n");
4046 shader_addline(buffer
, "}\n");
4048 TRACE("Compiling shader object %u\n", shader_obj
);
4049 shader_glsl_compile(gl_info
, shader_obj
, buffer
->buffer
);
4051 /* Store the shader object */
4055 /* GL locking is done by the caller */
4056 static GLuint
shader_glsl_generate_vshader(const struct wined3d_context
*context
,
4057 struct wined3d_shader_buffer
*buffer
, IWineD3DVertexShaderImpl
*This
,
4058 const struct vs_compile_args
*args
)
4060 const struct shader_reg_maps
*reg_maps
= &This
->baseShader
.reg_maps
;
4061 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
4062 CONST DWORD
*function
= This
->baseShader
.function
;
4063 struct shader_glsl_ctx_priv priv_ctx
;
4065 /* Create the hw GLSL shader program and assign it as the shader->prgId */
4066 GLhandleARB shader_obj
= GL_EXTCALL(glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB
));
4068 shader_addline(buffer
, "#version 120\n");
4070 if (gl_info
->supported
[EXT_GPU_SHADER4
])
4072 shader_addline(buffer
, "#extension GL_EXT_gpu_shader4 : enable\n");
4075 memset(&priv_ctx
, 0, sizeof(priv_ctx
));
4076 priv_ctx
.cur_vs_args
= args
;
4078 /* Base Declarations */
4079 shader_generate_glsl_declarations(context
, buffer
, (IWineD3DBaseShader
*)This
, reg_maps
, &priv_ctx
);
4081 /* Base Shader Body */
4082 shader_generate_main((IWineD3DBaseShader
*)This
, buffer
, reg_maps
, function
, &priv_ctx
);
4084 /* Unpack outputs */
4085 shader_addline(buffer
, "order_ps_input(OUT);\n");
4087 /* The D3DRS_FOGTABLEMODE render state defines if the shader-generated fog coord is used
4088 * or if the fragment depth is used. If the fragment depth is used(FOGTABLEMODE != NONE),
4089 * the fog frag coord is thrown away. If the fog frag coord is used, but not written by
4090 * the shader, it is set to 0.0(fully fogged, since start = 1.0, end = 0.0)
4092 if(args
->fog_src
== VS_FOG_Z
) {
4093 shader_addline(buffer
, "gl_FogFragCoord = gl_Position.z;\n");
4094 } else if (!reg_maps
->fog
) {
4095 shader_addline(buffer
, "gl_FogFragCoord = 0.0;\n");
4098 /* Write the final position.
4100 * OpenGL coordinates specify the center of the pixel while d3d coords specify
4101 * the corner. The offsets are stored in z and w in posFixup. posFixup.y contains
4102 * 1.0 or -1.0 to turn the rendering upside down for offscreen rendering. PosFixup.x
4103 * contains 1.0 to allow a mad.
4105 shader_addline(buffer
, "gl_Position.y = gl_Position.y * posFixup.y;\n");
4106 shader_addline(buffer
, "gl_Position.xy += posFixup.zw * gl_Position.ww;\n");
4107 if(args
->clip_enabled
) {
4108 shader_addline(buffer
, "gl_ClipVertex = gl_Position;\n");
4111 /* Z coord [0;1]->[-1;1] mapping, see comment in transform_projection in state.c
4113 * Basically we want (in homogeneous coordinates) z = z * 2 - 1. However, shaders are run
4114 * before the homogeneous divide, so we have to take the w into account: z = ((z / w) * 2 - 1) * w,
4115 * which is the same as z = z * 2 - w.
4117 shader_addline(buffer
, "gl_Position.z = gl_Position.z * 2.0 - gl_Position.w;\n");
4119 shader_addline(buffer
, "}\n");
4121 TRACE("Compiling shader object %u\n", shader_obj
);
4122 shader_glsl_compile(gl_info
, shader_obj
, buffer
->buffer
);
4127 static GLhandleARB
find_glsl_pshader(const struct wined3d_context
*context
,
4128 struct wined3d_shader_buffer
*buffer
, IWineD3DPixelShaderImpl
*shader
,
4129 const struct ps_compile_args
*args
, const struct ps_np2fixup_info
**np2fixup_info
)
4131 struct wined3d_state
*state
= &((IWineD3DDeviceImpl
*)shader
->baseShader
.device
)->stateBlock
->state
;
4134 struct glsl_ps_compiled_shader
*new_array
;
4135 struct glsl_pshader_private
*shader_data
;
4136 struct ps_np2fixup_info
*np2fixup
= NULL
;
4139 if (!shader
->baseShader
.backend_data
)
4141 shader
->baseShader
.backend_data
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*shader_data
));
4142 if (!shader
->baseShader
.backend_data
)
4144 ERR("Failed to allocate backend data.\n");
4148 shader_data
= shader
->baseShader
.backend_data
;
4150 /* Usually we have very few GL shaders for each d3d shader(just 1 or maybe 2),
4151 * so a linear search is more performant than a hashmap or a binary search
4152 * (cache coherency etc)
4154 for (i
= 0; i
< shader_data
->num_gl_shaders
; ++i
)
4156 if (!memcmp(&shader_data
->gl_shaders
[i
].args
, args
, sizeof(*args
)))
4158 if (args
->np2_fixup
) *np2fixup_info
= &shader_data
->gl_shaders
[i
].np2fixup
;
4159 return shader_data
->gl_shaders
[i
].prgId
;
4163 TRACE("No matching GL shader found for shader %p, compiling a new shader.\n", shader
);
4164 if(shader_data
->shader_array_size
== shader_data
->num_gl_shaders
) {
4165 if (shader_data
->num_gl_shaders
)
4167 new_size
= shader_data
->shader_array_size
+ max(1, shader_data
->shader_array_size
/ 2);
4168 new_array
= HeapReAlloc(GetProcessHeap(), 0, shader_data
->gl_shaders
,
4169 new_size
* sizeof(*shader_data
->gl_shaders
));
4171 new_array
= HeapAlloc(GetProcessHeap(), 0, sizeof(*shader_data
->gl_shaders
));
4176 ERR("Out of memory\n");
4179 shader_data
->gl_shaders
= new_array
;
4180 shader_data
->shader_array_size
= new_size
;
4183 shader_data
->gl_shaders
[shader_data
->num_gl_shaders
].args
= *args
;
4185 memset(&shader_data
->gl_shaders
[shader_data
->num_gl_shaders
].np2fixup
, 0, sizeof(struct ps_np2fixup_info
));
4186 if (args
->np2_fixup
) np2fixup
= &shader_data
->gl_shaders
[shader_data
->num_gl_shaders
].np2fixup
;
4188 pixelshader_update_samplers(&shader
->baseShader
.reg_maps
, (IWineD3DBaseTexture
**)state
->textures
);
4190 shader_buffer_clear(buffer
);
4191 ret
= shader_glsl_generate_pshader(context
, buffer
, shader
, args
, np2fixup
);
4192 shader_data
->gl_shaders
[shader_data
->num_gl_shaders
++].prgId
= ret
;
4193 *np2fixup_info
= np2fixup
;
4198 static inline BOOL
vs_args_equal(const struct vs_compile_args
*stored
, const struct vs_compile_args
*new,
4199 const DWORD use_map
) {
4200 if((stored
->swizzle_map
& use_map
) != new->swizzle_map
) return FALSE
;
4201 if((stored
->clip_enabled
) != new->clip_enabled
) return FALSE
;
4202 return stored
->fog_src
== new->fog_src
;
4205 static GLhandleARB
find_glsl_vshader(const struct wined3d_context
*context
,
4206 struct wined3d_shader_buffer
*buffer
, IWineD3DVertexShaderImpl
*shader
,
4207 const struct vs_compile_args
*args
)
4211 struct glsl_vs_compiled_shader
*new_array
;
4212 DWORD use_map
= ((IWineD3DDeviceImpl
*)shader
->baseShader
.device
)->strided_streams
.use_map
;
4213 struct glsl_vshader_private
*shader_data
;
4216 if (!shader
->baseShader
.backend_data
)
4218 shader
->baseShader
.backend_data
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*shader_data
));
4219 if (!shader
->baseShader
.backend_data
)
4221 ERR("Failed to allocate backend data.\n");
4225 shader_data
= shader
->baseShader
.backend_data
;
4227 /* Usually we have very few GL shaders for each d3d shader(just 1 or maybe 2),
4228 * so a linear search is more performant than a hashmap or a binary search
4229 * (cache coherency etc)
4231 for(i
= 0; i
< shader_data
->num_gl_shaders
; i
++) {
4232 if(vs_args_equal(&shader_data
->gl_shaders
[i
].args
, args
, use_map
)) {
4233 return shader_data
->gl_shaders
[i
].prgId
;
4237 TRACE("No matching GL shader found for shader %p, compiling a new shader.\n", shader
);
4239 if(shader_data
->shader_array_size
== shader_data
->num_gl_shaders
) {
4240 if (shader_data
->num_gl_shaders
)
4242 new_size
= shader_data
->shader_array_size
+ max(1, shader_data
->shader_array_size
/ 2);
4243 new_array
= HeapReAlloc(GetProcessHeap(), 0, shader_data
->gl_shaders
,
4244 new_size
* sizeof(*shader_data
->gl_shaders
));
4246 new_array
= HeapAlloc(GetProcessHeap(), 0, sizeof(*shader_data
->gl_shaders
));
4251 ERR("Out of memory\n");
4254 shader_data
->gl_shaders
= new_array
;
4255 shader_data
->shader_array_size
= new_size
;
4258 shader_data
->gl_shaders
[shader_data
->num_gl_shaders
].args
= *args
;
4260 shader_buffer_clear(buffer
);
4261 ret
= shader_glsl_generate_vshader(context
, buffer
, shader
, args
);
4262 shader_data
->gl_shaders
[shader_data
->num_gl_shaders
++].prgId
= ret
;
4267 /** Sets the GLSL program ID for the given pixel and vertex shader combination.
4268 * It sets the programId on the current StateBlock (because it should be called
4269 * inside of the DrawPrimitive() part of the render loop).
4271 * If a program for the given combination does not exist, create one, and store
4272 * the program in the hash table. If it creates a program, it will link the
4273 * given objects, too.
4276 /* GL locking is done by the caller */
4277 static void set_glsl_shader_program(const struct wined3d_context
*context
,
4278 IWineD3DDeviceImpl
*device
, BOOL use_ps
, BOOL use_vs
)
4280 const struct wined3d_state
*state
= &device
->stateBlock
->state
;
4281 IWineD3DVertexShader
*vshader
= use_vs
? (IWineD3DVertexShader
*)state
->vertex_shader
: NULL
;
4282 IWineD3DPixelShader
*pshader
= use_ps
? (IWineD3DPixelShader
*)state
->pixel_shader
: NULL
;
4283 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
4284 struct shader_glsl_priv
*priv
= device
->shader_priv
;
4285 struct glsl_shader_prog_link
*entry
= NULL
;
4286 GLhandleARB programId
= 0;
4287 GLhandleARB reorder_shader_id
= 0;
4290 struct ps_compile_args ps_compile_args
;
4291 struct vs_compile_args vs_compile_args
;
4293 if (vshader
) find_vs_compile_args(state
, (IWineD3DVertexShaderImpl
*)vshader
, &vs_compile_args
);
4294 if (pshader
) find_ps_compile_args(state
, (IWineD3DPixelShaderImpl
*)pshader
, &ps_compile_args
);
4296 entry
= get_glsl_program_entry(priv
, vshader
, pshader
, &vs_compile_args
, &ps_compile_args
);
4298 priv
->glsl_program
= entry
;
4302 /* If we get to this point, then no matching program exists, so we create one */
4303 programId
= GL_EXTCALL(glCreateProgramObjectARB());
4304 TRACE("Created new GLSL shader program %u\n", programId
);
4306 /* Create the entry */
4307 entry
= HeapAlloc(GetProcessHeap(), 0, sizeof(struct glsl_shader_prog_link
));
4308 entry
->programId
= programId
;
4309 entry
->vshader
= vshader
;
4310 entry
->pshader
= pshader
;
4311 entry
->vs_args
= vs_compile_args
;
4312 entry
->ps_args
= ps_compile_args
;
4313 entry
->constant_version
= 0;
4314 entry
->np2Fixup_info
= NULL
;
4315 /* Add the hash table entry */
4316 add_glsl_program_entry(priv
, entry
);
4318 /* Set the current program */
4319 priv
->glsl_program
= entry
;
4321 /* Attach GLSL vshader */
4324 GLhandleARB vshader_id
= find_glsl_vshader(context
, &priv
->shader_buffer
,
4325 (IWineD3DVertexShaderImpl
*)vshader
, &vs_compile_args
);
4326 WORD map
= ((IWineD3DBaseShaderImpl
*)vshader
)->baseShader
.reg_maps
.input_registers
;
4329 reorder_shader_id
= generate_param_reorder_function(&priv
->shader_buffer
, vshader
, pshader
, gl_info
);
4330 TRACE("Attaching GLSL shader object %u to program %u\n", reorder_shader_id
, programId
);
4331 GL_EXTCALL(glAttachObjectARB(programId
, reorder_shader_id
));
4332 checkGLcall("glAttachObjectARB");
4333 /* Flag the reorder function for deletion, then it will be freed automatically when the program
4336 GL_EXTCALL(glDeleteObjectARB(reorder_shader_id
));
4338 TRACE("Attaching GLSL shader object %u to program %u\n", vshader_id
, programId
);
4339 GL_EXTCALL(glAttachObjectARB(programId
, vshader_id
));
4340 checkGLcall("glAttachObjectARB");
4342 /* Bind vertex attributes to a corresponding index number to match
4343 * the same index numbers as ARB_vertex_programs (makes loading
4344 * vertex attributes simpler). With this method, we can use the
4345 * exact same code to load the attributes later for both ARB and
4348 * We have to do this here because we need to know the Program ID
4349 * in order to make the bindings work, and it has to be done prior
4350 * to linking the GLSL program. */
4351 for (i
= 0; map
; map
>>= 1, ++i
)
4353 if (!(map
& 1)) continue;
4355 snprintf(tmp_name
, sizeof(tmp_name
), "attrib%u", i
);
4356 GL_EXTCALL(glBindAttribLocationARB(programId
, i
, tmp_name
));
4358 checkGLcall("glBindAttribLocationARB");
4360 list_add_head(&((IWineD3DBaseShaderImpl
*)vshader
)->baseShader
.linked_programs
, &entry
->vshader_entry
);
4363 /* Attach GLSL pshader */
4366 GLhandleARB pshader_id
= find_glsl_pshader(context
, &priv
->shader_buffer
,
4367 (IWineD3DPixelShaderImpl
*)pshader
, &ps_compile_args
, &entry
->np2Fixup_info
);
4368 TRACE("Attaching GLSL shader object %u to program %u\n", pshader_id
, programId
);
4369 GL_EXTCALL(glAttachObjectARB(programId
, pshader_id
));
4370 checkGLcall("glAttachObjectARB");
4372 list_add_head(&((IWineD3DBaseShaderImpl
*)pshader
)->baseShader
.linked_programs
, &entry
->pshader_entry
);
4375 /* Link the program */
4376 TRACE("Linking GLSL shader program %u\n", programId
);
4377 GL_EXTCALL(glLinkProgramARB(programId
));
4378 shader_glsl_validate_link(gl_info
, programId
);
4380 entry
->vuniformF_locations
= HeapAlloc(GetProcessHeap(), 0,
4381 sizeof(GLhandleARB
) * gl_info
->limits
.glsl_vs_float_constants
);
4382 for (i
= 0; i
< gl_info
->limits
.glsl_vs_float_constants
; ++i
)
4384 snprintf(glsl_name
, sizeof(glsl_name
), "VC[%i]", i
);
4385 entry
->vuniformF_locations
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, glsl_name
));
4387 for (i
= 0; i
< MAX_CONST_I
; ++i
)
4389 snprintf(glsl_name
, sizeof(glsl_name
), "VI[%i]", i
);
4390 entry
->vuniformI_locations
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, glsl_name
));
4392 entry
->puniformF_locations
= HeapAlloc(GetProcessHeap(), 0,
4393 sizeof(GLhandleARB
) * gl_info
->limits
.glsl_ps_float_constants
);
4394 for (i
= 0; i
< gl_info
->limits
.glsl_ps_float_constants
; ++i
)
4396 snprintf(glsl_name
, sizeof(glsl_name
), "PC[%i]", i
);
4397 entry
->puniformF_locations
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, glsl_name
));
4399 for (i
= 0; i
< MAX_CONST_I
; ++i
)
4401 snprintf(glsl_name
, sizeof(glsl_name
), "PI[%i]", i
);
4402 entry
->puniformI_locations
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, glsl_name
));
4408 for(i
= 0; i
< MAX_TEXTURES
; i
++) {
4409 sprintf(name
, "bumpenvmat%u", i
);
4410 entry
->bumpenvmat_location
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, name
));
4411 sprintf(name
, "luminancescale%u", i
);
4412 entry
->luminancescale_location
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, name
));
4413 sprintf(name
, "luminanceoffset%u", i
);
4414 entry
->luminanceoffset_location
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, name
));
4417 if (ps_compile_args
.np2_fixup
) {
4418 if (entry
->np2Fixup_info
) {
4419 entry
->np2Fixup_location
= GL_EXTCALL(glGetUniformLocationARB(programId
, "PsamplerNP2Fixup"));
4421 FIXME("NP2 texcoord fixup needed for this pixelshader, but no fixup uniform found.\n");
4426 entry
->posFixup_location
= GL_EXTCALL(glGetUniformLocationARB(programId
, "posFixup"));
4427 entry
->ycorrection_location
= GL_EXTCALL(glGetUniformLocationARB(programId
, "ycorrection"));
4428 checkGLcall("Find glsl program uniform locations");
4431 && ((IWineD3DPixelShaderImpl
*)pshader
)->baseShader
.reg_maps
.shader_version
.major
>= 3
4432 && ((IWineD3DPixelShaderImpl
*)pshader
)->declared_in_count
> vec4_varyings(3, gl_info
))
4434 TRACE("Shader %d needs vertex color clamping disabled\n", programId
);
4435 entry
->vertex_color_clamp
= GL_FALSE
;
4437 entry
->vertex_color_clamp
= GL_FIXED_ONLY_ARB
;
4440 /* Set the shader to allow uniform loading on it */
4441 GL_EXTCALL(glUseProgramObjectARB(programId
));
4442 checkGLcall("glUseProgramObjectARB(programId)");
4444 /* Load the vertex and pixel samplers now. The function that finds the mappings makes sure
4445 * that it stays the same for each vertexshader-pixelshader pair(=linked glsl program). If
4446 * a pshader with fixed function pipeline is used there are no vertex samplers, and if a
4447 * vertex shader with fixed function pixel processing is used we make sure that the card
4448 * supports enough samplers to allow the max number of vertex samplers with all possible
4449 * fixed function fragment processing setups. So once the program is linked these samplers
4452 if (vshader
) shader_glsl_load_vsamplers(gl_info
, device
->texUnitMap
, programId
);
4453 if (pshader
) shader_glsl_load_psamplers(gl_info
, device
->texUnitMap
, programId
);
4455 /* If the local constants do not have to be loaded with the environment constants,
4456 * load them now to have them hardcoded in the GLSL program. This saves some CPU cycles
4459 if (pshader
&& !((IWineD3DBaseShaderImpl
*)pshader
)->baseShader
.load_local_constsF
)
4461 hardcode_local_constants((IWineD3DBaseShaderImpl
*) pshader
, gl_info
, programId
, 'P');
4463 if (vshader
&& !((IWineD3DBaseShaderImpl
*)vshader
)->baseShader
.load_local_constsF
)
4465 hardcode_local_constants((IWineD3DBaseShaderImpl
*) vshader
, gl_info
, programId
, 'V');
4469 /* GL locking is done by the caller */
4470 static GLhandleARB
create_glsl_blt_shader(const struct wined3d_gl_info
*gl_info
, enum tex_types tex_type
, BOOL masked
)
4472 GLhandleARB program_id
;
4473 GLhandleARB vshader_id
, pshader_id
;
4474 const char *blt_pshader
;
4476 static const char *blt_vshader
=
4480 " gl_Position = gl_Vertex;\n"
4481 " gl_FrontColor = vec4(1.0);\n"
4482 " gl_TexCoord[0] = gl_MultiTexCoord0;\n"
4485 static const char * const blt_pshaders_full
[tex_type_count
] =
4491 "uniform sampler2D sampler;\n"
4494 " gl_FragDepth = texture2D(sampler, gl_TexCoord[0].xy).x;\n"
4500 "uniform samplerCube sampler;\n"
4503 " gl_FragDepth = textureCube(sampler, gl_TexCoord[0].xyz).x;\n"
4507 "#extension GL_ARB_texture_rectangle : enable\n"
4508 "uniform sampler2DRect sampler;\n"
4511 " gl_FragDepth = texture2DRect(sampler, gl_TexCoord[0].xy).x;\n"
4515 static const char * const blt_pshaders_masked
[tex_type_count
] =
4521 "uniform sampler2D sampler;\n"
4522 "uniform vec4 mask;\n"
4525 " if (all(lessThan(gl_FragCoord.xy, mask.zw))) discard;\n"
4526 " gl_FragDepth = texture2D(sampler, gl_TexCoord[0].xy).x;\n"
4532 "uniform samplerCube sampler;\n"
4533 "uniform vec4 mask;\n"
4536 " if (all(lessThan(gl_FragCoord.xy, mask.zw))) discard;\n"
4537 " gl_FragDepth = textureCube(sampler, gl_TexCoord[0].xyz).x;\n"
4541 "#extension GL_ARB_texture_rectangle : enable\n"
4542 "uniform sampler2DRect sampler;\n"
4543 "uniform vec4 mask;\n"
4546 " if (all(lessThan(gl_FragCoord.xy, mask.zw))) discard;\n"
4547 " gl_FragDepth = texture2DRect(sampler, gl_TexCoord[0].xy).x;\n"
4551 blt_pshader
= masked
? blt_pshaders_masked
[tex_type
] : blt_pshaders_full
[tex_type
];
4554 FIXME("tex_type %#x not supported\n", tex_type
);
4558 vshader_id
= GL_EXTCALL(glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB
));
4559 shader_glsl_compile(gl_info
, vshader_id
, blt_vshader
);
4561 pshader_id
= GL_EXTCALL(glCreateShaderObjectARB(GL_FRAGMENT_SHADER_ARB
));
4562 shader_glsl_compile(gl_info
, pshader_id
, blt_pshader
);
4564 program_id
= GL_EXTCALL(glCreateProgramObjectARB());
4565 GL_EXTCALL(glAttachObjectARB(program_id
, vshader_id
));
4566 GL_EXTCALL(glAttachObjectARB(program_id
, pshader_id
));
4567 GL_EXTCALL(glLinkProgramARB(program_id
));
4569 shader_glsl_validate_link(gl_info
, program_id
);
4571 /* Once linked we can mark the shaders for deletion. They will be deleted once the program
4574 GL_EXTCALL(glDeleteObjectARB(vshader_id
));
4575 GL_EXTCALL(glDeleteObjectARB(pshader_id
));
4579 /* GL locking is done by the caller */
4580 static void shader_glsl_select(const struct wined3d_context
*context
, BOOL usePS
, BOOL useVS
)
4582 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
4583 IWineD3DDeviceImpl
*device
= context
->swapchain
->device
;
4584 struct shader_glsl_priv
*priv
= device
->shader_priv
;
4585 GLhandleARB program_id
= 0;
4586 GLenum old_vertex_color_clamp
, current_vertex_color_clamp
;
4588 old_vertex_color_clamp
= priv
->glsl_program
? priv
->glsl_program
->vertex_color_clamp
: GL_FIXED_ONLY_ARB
;
4590 if (useVS
|| usePS
) set_glsl_shader_program(context
, device
, usePS
, useVS
);
4591 else priv
->glsl_program
= NULL
;
4593 current_vertex_color_clamp
= priv
->glsl_program
? priv
->glsl_program
->vertex_color_clamp
: GL_FIXED_ONLY_ARB
;
4595 if (old_vertex_color_clamp
!= current_vertex_color_clamp
)
4597 if (gl_info
->supported
[ARB_COLOR_BUFFER_FLOAT
])
4599 GL_EXTCALL(glClampColorARB(GL_CLAMP_VERTEX_COLOR_ARB
, current_vertex_color_clamp
));
4600 checkGLcall("glClampColorARB");
4604 FIXME("vertex color clamp needs to be changed, but extension not supported.\n");
4608 program_id
= priv
->glsl_program
? priv
->glsl_program
->programId
: 0;
4609 if (program_id
) TRACE("Using GLSL program %u\n", program_id
);
4610 GL_EXTCALL(glUseProgramObjectARB(program_id
));
4611 checkGLcall("glUseProgramObjectARB");
4613 /* In case that NP2 texcoord fixup data is found for the selected program, trigger a reload of the
4614 * constants. This has to be done because it can't be guaranteed that sampler() (from state.c) is
4615 * called between selecting the shader and using it, which results in wrong fixup for some frames. */
4616 if (priv
->glsl_program
&& priv
->glsl_program
->np2Fixup_info
)
4618 shader_glsl_load_np2fixup_constants(priv
, gl_info
, &device
->stateBlock
->state
);
4622 /* GL locking is done by the caller */
4623 static void shader_glsl_select_depth_blt(IWineD3DDevice
*iface
,
4624 enum tex_types tex_type
, const SIZE
*ds_mask_size
)
4626 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
4627 const struct wined3d_gl_info
*gl_info
= &This
->adapter
->gl_info
;
4628 BOOL masked
= ds_mask_size
->cx
&& ds_mask_size
->cy
;
4629 struct shader_glsl_priv
*priv
= This
->shader_priv
;
4630 GLhandleARB
*blt_program
;
4633 blt_program
= masked
? &priv
->depth_blt_program_masked
[tex_type
] : &priv
->depth_blt_program_full
[tex_type
];
4636 *blt_program
= create_glsl_blt_shader(gl_info
, tex_type
, masked
);
4637 loc
= GL_EXTCALL(glGetUniformLocationARB(*blt_program
, "sampler"));
4638 GL_EXTCALL(glUseProgramObjectARB(*blt_program
));
4639 GL_EXTCALL(glUniform1iARB(loc
, 0));
4643 GL_EXTCALL(glUseProgramObjectARB(*blt_program
));
4648 loc
= GL_EXTCALL(glGetUniformLocationARB(*blt_program
, "mask"));
4649 GL_EXTCALL(glUniform4fARB(loc
, 0.0f
, 0.0f
, (float)ds_mask_size
->cx
, (float)ds_mask_size
->cy
));
4653 /* GL locking is done by the caller */
4654 static void shader_glsl_deselect_depth_blt(IWineD3DDevice
*iface
) {
4655 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
4656 const struct wined3d_gl_info
*gl_info
= &This
->adapter
->gl_info
;
4657 struct shader_glsl_priv
*priv
= This
->shader_priv
;
4658 GLhandleARB program_id
;
4660 program_id
= priv
->glsl_program
? priv
->glsl_program
->programId
: 0;
4661 if (program_id
) TRACE("Using GLSL program %u\n", program_id
);
4663 GL_EXTCALL(glUseProgramObjectARB(program_id
));
4664 checkGLcall("glUseProgramObjectARB");
4667 static void shader_glsl_destroy(IWineD3DBaseShader
*iface
) {
4668 const struct list
*linked_programs
;
4669 IWineD3DBaseShaderImpl
*This
= (IWineD3DBaseShaderImpl
*) iface
;
4670 IWineD3DDeviceImpl
*device
= (IWineD3DDeviceImpl
*)This
->baseShader
.device
;
4671 struct shader_glsl_priv
*priv
= device
->shader_priv
;
4672 const struct wined3d_gl_info
*gl_info
;
4673 struct wined3d_context
*context
;
4675 /* Note: Do not use QueryInterface here to find out which shader type this is because this code
4676 * can be called from IWineD3DBaseShader::Release
4678 char pshader
= shader_is_pshader_version(This
->baseShader
.reg_maps
.shader_version
.type
);
4681 struct glsl_pshader_private
*shader_data
;
4682 shader_data
= This
->baseShader
.backend_data
;
4683 if (!shader_data
|| !shader_data
->num_gl_shaders
)
4685 HeapFree(GetProcessHeap(), 0, shader_data
);
4686 This
->baseShader
.backend_data
= NULL
;
4690 context
= context_acquire(device
, NULL
);
4691 gl_info
= context
->gl_info
;
4693 if (priv
->glsl_program
&& (IWineD3DBaseShader
*)priv
->glsl_program
->pshader
== iface
)
4696 shader_glsl_select(context
, FALSE
, FALSE
);
4700 struct glsl_vshader_private
*shader_data
;
4701 shader_data
= This
->baseShader
.backend_data
;
4702 if (!shader_data
|| !shader_data
->num_gl_shaders
)
4704 HeapFree(GetProcessHeap(), 0, shader_data
);
4705 This
->baseShader
.backend_data
= NULL
;
4709 context
= context_acquire(device
, NULL
);
4710 gl_info
= context
->gl_info
;
4712 if (priv
->glsl_program
&& (IWineD3DBaseShader
*)priv
->glsl_program
->vshader
== iface
)
4715 shader_glsl_select(context
, FALSE
, FALSE
);
4720 linked_programs
= &This
->baseShader
.linked_programs
;
4722 TRACE("Deleting linked programs\n");
4723 if (linked_programs
->next
) {
4724 struct glsl_shader_prog_link
*entry
, *entry2
;
4728 LIST_FOR_EACH_ENTRY_SAFE(entry
, entry2
, linked_programs
, struct glsl_shader_prog_link
, pshader_entry
) {
4729 delete_glsl_program_entry(priv
, gl_info
, entry
);
4732 LIST_FOR_EACH_ENTRY_SAFE(entry
, entry2
, linked_programs
, struct glsl_shader_prog_link
, vshader_entry
) {
4733 delete_glsl_program_entry(priv
, gl_info
, entry
);
4741 struct glsl_pshader_private
*shader_data
= This
->baseShader
.backend_data
;
4744 for(i
= 0; i
< shader_data
->num_gl_shaders
; i
++) {
4745 TRACE("deleting pshader %u\n", shader_data
->gl_shaders
[i
].prgId
);
4746 GL_EXTCALL(glDeleteObjectARB(shader_data
->gl_shaders
[i
].prgId
));
4747 checkGLcall("glDeleteObjectARB");
4750 HeapFree(GetProcessHeap(), 0, shader_data
->gl_shaders
);
4755 struct glsl_vshader_private
*shader_data
= This
->baseShader
.backend_data
;
4758 for(i
= 0; i
< shader_data
->num_gl_shaders
; i
++) {
4759 TRACE("deleting vshader %u\n", shader_data
->gl_shaders
[i
].prgId
);
4760 GL_EXTCALL(glDeleteObjectARB(shader_data
->gl_shaders
[i
].prgId
));
4761 checkGLcall("glDeleteObjectARB");
4764 HeapFree(GetProcessHeap(), 0, shader_data
->gl_shaders
);
4767 HeapFree(GetProcessHeap(), 0, This
->baseShader
.backend_data
);
4768 This
->baseShader
.backend_data
= NULL
;
4770 context_release(context
);
4773 static int glsl_program_key_compare(const void *key
, const struct wine_rb_entry
*entry
)
4775 const glsl_program_key_t
*k
= key
;
4776 const struct glsl_shader_prog_link
*prog
= WINE_RB_ENTRY_VALUE(entry
,
4777 const struct glsl_shader_prog_link
, program_lookup_entry
);
4780 if (k
->vshader
> prog
->vshader
) return 1;
4781 else if (k
->vshader
< prog
->vshader
) return -1;
4783 if (k
->pshader
> prog
->pshader
) return 1;
4784 else if (k
->pshader
< prog
->pshader
) return -1;
4786 if (k
->vshader
&& (cmp
= memcmp(&k
->vs_args
, &prog
->vs_args
, sizeof(prog
->vs_args
)))) return cmp
;
4787 if (k
->pshader
&& (cmp
= memcmp(&k
->ps_args
, &prog
->ps_args
, sizeof(prog
->ps_args
)))) return cmp
;
4792 static BOOL
constant_heap_init(struct constant_heap
*heap
, unsigned int constant_count
)
4794 SIZE_T size
= (constant_count
+ 1) * sizeof(*heap
->entries
) + constant_count
* sizeof(*heap
->positions
);
4795 void *mem
= HeapAlloc(GetProcessHeap(), 0, size
);
4799 ERR("Failed to allocate memory\n");
4803 heap
->entries
= mem
;
4804 heap
->entries
[1].version
= 0;
4805 heap
->positions
= (unsigned int *)(heap
->entries
+ constant_count
+ 1);
4811 static void constant_heap_free(struct constant_heap
*heap
)
4813 HeapFree(GetProcessHeap(), 0, heap
->entries
);
4816 static const struct wine_rb_functions wined3d_glsl_program_rb_functions
=
4821 glsl_program_key_compare
,
4824 static HRESULT
shader_glsl_alloc(IWineD3DDevice
*iface
) {
4825 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
4826 const struct wined3d_gl_info
*gl_info
= &This
->adapter
->gl_info
;
4827 struct shader_glsl_priv
*priv
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(struct shader_glsl_priv
));
4828 SIZE_T stack_size
= wined3d_log2i(max(gl_info
->limits
.glsl_vs_float_constants
,
4829 gl_info
->limits
.glsl_ps_float_constants
)) + 1;
4831 if (!shader_buffer_init(&priv
->shader_buffer
))
4833 ERR("Failed to initialize shader buffer.\n");
4837 priv
->stack
= HeapAlloc(GetProcessHeap(), 0, stack_size
* sizeof(*priv
->stack
));
4840 ERR("Failed to allocate memory.\n");
4844 if (!constant_heap_init(&priv
->vconst_heap
, gl_info
->limits
.glsl_vs_float_constants
))
4846 ERR("Failed to initialize vertex shader constant heap\n");
4850 if (!constant_heap_init(&priv
->pconst_heap
, gl_info
->limits
.glsl_ps_float_constants
))
4852 ERR("Failed to initialize pixel shader constant heap\n");
4856 if (wine_rb_init(&priv
->program_lookup
, &wined3d_glsl_program_rb_functions
) == -1)
4858 ERR("Failed to initialize rbtree.\n");
4862 priv
->next_constant_version
= 1;
4864 This
->shader_priv
= priv
;
4868 constant_heap_free(&priv
->pconst_heap
);
4869 constant_heap_free(&priv
->vconst_heap
);
4870 HeapFree(GetProcessHeap(), 0, priv
->stack
);
4871 shader_buffer_free(&priv
->shader_buffer
);
4872 HeapFree(GetProcessHeap(), 0, priv
);
4873 return E_OUTOFMEMORY
;
4876 /* Context activation is done by the caller. */
4877 static void shader_glsl_free(IWineD3DDevice
*iface
) {
4878 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
4879 const struct wined3d_gl_info
*gl_info
= &This
->adapter
->gl_info
;
4880 struct shader_glsl_priv
*priv
= This
->shader_priv
;
4884 for (i
= 0; i
< tex_type_count
; ++i
)
4886 if (priv
->depth_blt_program_full
[i
])
4888 GL_EXTCALL(glDeleteObjectARB(priv
->depth_blt_program_full
[i
]));
4890 if (priv
->depth_blt_program_masked
[i
])
4892 GL_EXTCALL(glDeleteObjectARB(priv
->depth_blt_program_masked
[i
]));
4897 wine_rb_destroy(&priv
->program_lookup
, NULL
, NULL
);
4898 constant_heap_free(&priv
->pconst_heap
);
4899 constant_heap_free(&priv
->vconst_heap
);
4900 HeapFree(GetProcessHeap(), 0, priv
->stack
);
4901 shader_buffer_free(&priv
->shader_buffer
);
4903 HeapFree(GetProcessHeap(), 0, This
->shader_priv
);
4904 This
->shader_priv
= NULL
;
4907 static BOOL
shader_glsl_dirty_const(IWineD3DDevice
*iface
) {
4908 /* TODO: GL_EXT_bindable_uniform can be used to share constants across shaders */
4912 static void shader_glsl_get_caps(const struct wined3d_gl_info
*gl_info
, struct shader_caps
*pCaps
)
4914 /* Nvidia Geforce6/7 or Ati R4xx/R5xx cards with GLSL support, support VS 3.0 but older Nvidia/Ati
4915 * models with GLSL support only support 2.0. In case of nvidia we can detect VS 2.0 support based
4916 * on the version of NV_vertex_program.
4917 * For Ati cards there's no way using glsl (it abstracts the lowlevel info away) and also not
4918 * using ARB_vertex_program. It is safe to assume that when a card supports pixel shader 2.0 it
4919 * supports vertex shader 2.0 too and the way around. We can detect ps2.0 using the maximum number
4920 * of native instructions, so use that here. For more info see the pixel shader versioning code below.
4922 if ((gl_info
->supported
[NV_VERTEX_PROGRAM2
] && !gl_info
->supported
[NV_VERTEX_PROGRAM3
])
4923 || gl_info
->limits
.arb_ps_instructions
<= 512)
4924 pCaps
->VertexShaderVersion
= WINED3DVS_VERSION(2,0);
4926 pCaps
->VertexShaderVersion
= WINED3DVS_VERSION(3,0);
4927 TRACE_(d3d_caps
)("Hardware vertex shader version %d.%d enabled (GLSL)\n", (pCaps
->VertexShaderVersion
>> 8) & 0xff, pCaps
->VertexShaderVersion
& 0xff);
4928 pCaps
->MaxVertexShaderConst
= gl_info
->limits
.glsl_vs_float_constants
;
4930 /* Older DX9-class videocards (GeforceFX / Radeon >9500/X*00) only support pixel shader 2.0/2.0a/2.0b.
4931 * In OpenGL the extensions related to GLSL abstract lowlevel GL info away which is needed
4932 * to distinguish between 2.0 and 3.0 (and 2.0a/2.0b). In case of Nvidia we use their fragment
4933 * program extensions. On other hardware including ATI GL_ARB_fragment_program offers the info
4934 * in max native instructions. Intel and others also offer the info in this extension but they
4935 * don't support GLSL (at least on Windows).
4937 * PS2.0 requires at least 96 instructions, 2.0a/2.0b go up to 512. Assume that if the number
4938 * of instructions is 512 or less we have to do with ps2.0 hardware.
4939 * NOTE: ps3.0 hardware requires 512 or more instructions but ati and nvidia offer 'enough' (1024 vs 4096) on their most basic ps3.0 hardware.
4941 if ((gl_info
->supported
[NV_FRAGMENT_PROGRAM
] && !gl_info
->supported
[NV_FRAGMENT_PROGRAM2
])
4942 || gl_info
->limits
.arb_ps_instructions
<= 512)
4943 pCaps
->PixelShaderVersion
= WINED3DPS_VERSION(2,0);
4945 pCaps
->PixelShaderVersion
= WINED3DPS_VERSION(3,0);
4947 pCaps
->MaxPixelShaderConst
= gl_info
->limits
.glsl_ps_float_constants
;
4949 /* FIXME: The following line is card dependent. -8.0 to 8.0 is the
4950 * Direct3D minimum requirement.
4952 * Both GL_ARB_fragment_program and GLSL require a "maximum representable magnitude"
4953 * of colors to be 2^10, and 2^32 for other floats. Should we use 1024 here?
4955 * The problem is that the refrast clamps temporary results in the shader to
4956 * [-MaxValue;+MaxValue]. If the card's max value is bigger than the one we advertize here,
4957 * then applications may miss the clamping behavior. On the other hand, if it is smaller,
4958 * the shader will generate incorrect results too. Unfortunately, GL deliberately doesn't
4959 * offer a way to query this.
4961 pCaps
->PixelShader1xMaxValue
= 8.0;
4962 TRACE_(d3d_caps
)("Hardware pixel shader version %d.%d enabled (GLSL)\n", (pCaps
->PixelShaderVersion
>> 8) & 0xff, pCaps
->PixelShaderVersion
& 0xff);
4964 pCaps
->VSClipping
= TRUE
;
4967 static BOOL
shader_glsl_color_fixup_supported(struct color_fixup_desc fixup
)
4969 if (TRACE_ON(d3d_shader
) && TRACE_ON(d3d
))
4971 TRACE("Checking support for fixup:\n");
4972 dump_color_fixup_desc(fixup
);
4975 /* We support everything except YUV conversions. */
4976 if (!is_complex_fixup(fixup
))
4982 TRACE("[FAILED]\n");
4986 static const SHADER_HANDLER shader_glsl_instruction_handler_table
[WINED3DSIH_TABLE_SIZE
] =
4988 /* WINED3DSIH_ABS */ shader_glsl_map2gl
,
4989 /* WINED3DSIH_ADD */ shader_glsl_arith
,
4990 /* WINED3DSIH_AND */ NULL
,
4991 /* WINED3DSIH_BEM */ shader_glsl_bem
,
4992 /* WINED3DSIH_BREAK */ shader_glsl_break
,
4993 /* WINED3DSIH_BREAKC */ shader_glsl_breakc
,
4994 /* WINED3DSIH_BREAKP */ NULL
,
4995 /* WINED3DSIH_CALL */ shader_glsl_call
,
4996 /* WINED3DSIH_CALLNZ */ shader_glsl_callnz
,
4997 /* WINED3DSIH_CMP */ shader_glsl_cmp
,
4998 /* WINED3DSIH_CND */ shader_glsl_cnd
,
4999 /* WINED3DSIH_CRS */ shader_glsl_cross
,
5000 /* WINED3DSIH_CUT */ NULL
,
5001 /* WINED3DSIH_DCL */ NULL
,
5002 /* WINED3DSIH_DEF */ NULL
,
5003 /* WINED3DSIH_DEFB */ NULL
,
5004 /* WINED3DSIH_DEFI */ NULL
,
5005 /* WINED3DSIH_DIV */ NULL
,
5006 /* WINED3DSIH_DP2ADD */ shader_glsl_dp2add
,
5007 /* WINED3DSIH_DP3 */ shader_glsl_dot
,
5008 /* WINED3DSIH_DP4 */ shader_glsl_dot
,
5009 /* WINED3DSIH_DST */ shader_glsl_dst
,
5010 /* WINED3DSIH_DSX */ shader_glsl_map2gl
,
5011 /* WINED3DSIH_DSY */ shader_glsl_map2gl
,
5012 /* WINED3DSIH_ELSE */ shader_glsl_else
,
5013 /* WINED3DSIH_EMIT */ NULL
,
5014 /* WINED3DSIH_ENDIF */ shader_glsl_end
,
5015 /* WINED3DSIH_ENDLOOP */ shader_glsl_end
,
5016 /* WINED3DSIH_ENDREP */ shader_glsl_end
,
5017 /* WINED3DSIH_EXP */ shader_glsl_map2gl
,
5018 /* WINED3DSIH_EXPP */ shader_glsl_expp
,
5019 /* WINED3DSIH_FRC */ shader_glsl_map2gl
,
5020 /* WINED3DSIH_FTOI */ NULL
,
5021 /* WINED3DSIH_IADD */ NULL
,
5022 /* WINED3DSIH_IEQ */ NULL
,
5023 /* WINED3DSIH_IF */ shader_glsl_if
,
5024 /* WINED3DSIH_IFC */ shader_glsl_ifc
,
5025 /* WINED3DSIH_IGE */ NULL
,
5026 /* WINED3DSIH_IMUL */ NULL
,
5027 /* WINED3DSIH_ITOF */ NULL
,
5028 /* WINED3DSIH_LABEL */ shader_glsl_label
,
5029 /* WINED3DSIH_LD */ NULL
,
5030 /* WINED3DSIH_LIT */ shader_glsl_lit
,
5031 /* WINED3DSIH_LOG */ shader_glsl_log
,
5032 /* WINED3DSIH_LOGP */ shader_glsl_log
,
5033 /* WINED3DSIH_LOOP */ shader_glsl_loop
,
5034 /* WINED3DSIH_LRP */ shader_glsl_lrp
,
5035 /* WINED3DSIH_LT */ NULL
,
5036 /* WINED3DSIH_M3x2 */ shader_glsl_mnxn
,
5037 /* WINED3DSIH_M3x3 */ shader_glsl_mnxn
,
5038 /* WINED3DSIH_M3x4 */ shader_glsl_mnxn
,
5039 /* WINED3DSIH_M4x3 */ shader_glsl_mnxn
,
5040 /* WINED3DSIH_M4x4 */ shader_glsl_mnxn
,
5041 /* WINED3DSIH_MAD */ shader_glsl_mad
,
5042 /* WINED3DSIH_MAX */ shader_glsl_map2gl
,
5043 /* WINED3DSIH_MIN */ shader_glsl_map2gl
,
5044 /* WINED3DSIH_MOV */ shader_glsl_mov
,
5045 /* WINED3DSIH_MOVA */ shader_glsl_mov
,
5046 /* WINED3DSIH_MOVC */ NULL
,
5047 /* WINED3DSIH_MUL */ shader_glsl_arith
,
5048 /* WINED3DSIH_NOP */ NULL
,
5049 /* WINED3DSIH_NRM */ shader_glsl_nrm
,
5050 /* WINED3DSIH_PHASE */ NULL
,
5051 /* WINED3DSIH_POW */ shader_glsl_pow
,
5052 /* WINED3DSIH_RCP */ shader_glsl_rcp
,
5053 /* WINED3DSIH_REP */ shader_glsl_rep
,
5054 /* WINED3DSIH_RET */ shader_glsl_ret
,
5055 /* WINED3DSIH_RSQ */ shader_glsl_rsq
,
5056 /* WINED3DSIH_SAMPLE */ NULL
,
5057 /* WINED3DSIH_SAMPLE_GRAD */ NULL
,
5058 /* WINED3DSIH_SAMPLE_LOD */ NULL
,
5059 /* WINED3DSIH_SETP */ NULL
,
5060 /* WINED3DSIH_SGE */ shader_glsl_compare
,
5061 /* WINED3DSIH_SGN */ shader_glsl_sgn
,
5062 /* WINED3DSIH_SINCOS */ shader_glsl_sincos
,
5063 /* WINED3DSIH_SLT */ shader_glsl_compare
,
5064 /* WINED3DSIH_SQRT */ NULL
,
5065 /* WINED3DSIH_SUB */ shader_glsl_arith
,
5066 /* WINED3DSIH_TEX */ shader_glsl_tex
,
5067 /* WINED3DSIH_TEXBEM */ shader_glsl_texbem
,
5068 /* WINED3DSIH_TEXBEML */ shader_glsl_texbem
,
5069 /* WINED3DSIH_TEXCOORD */ shader_glsl_texcoord
,
5070 /* WINED3DSIH_TEXDEPTH */ shader_glsl_texdepth
,
5071 /* WINED3DSIH_TEXDP3 */ shader_glsl_texdp3
,
5072 /* WINED3DSIH_TEXDP3TEX */ shader_glsl_texdp3tex
,
5073 /* WINED3DSIH_TEXKILL */ shader_glsl_texkill
,
5074 /* WINED3DSIH_TEXLDD */ shader_glsl_texldd
,
5075 /* WINED3DSIH_TEXLDL */ shader_glsl_texldl
,
5076 /* WINED3DSIH_TEXM3x2DEPTH */ shader_glsl_texm3x2depth
,
5077 /* WINED3DSIH_TEXM3x2PAD */ shader_glsl_texm3x2pad
,
5078 /* WINED3DSIH_TEXM3x2TEX */ shader_glsl_texm3x2tex
,
5079 /* WINED3DSIH_TEXM3x3 */ shader_glsl_texm3x3
,
5080 /* WINED3DSIH_TEXM3x3DIFF */ NULL
,
5081 /* WINED3DSIH_TEXM3x3PAD */ shader_glsl_texm3x3pad
,
5082 /* WINED3DSIH_TEXM3x3SPEC */ shader_glsl_texm3x3spec
,
5083 /* WINED3DSIH_TEXM3x3TEX */ shader_glsl_texm3x3tex
,
5084 /* WINED3DSIH_TEXM3x3VSPEC */ shader_glsl_texm3x3vspec
,
5085 /* WINED3DSIH_TEXREG2AR */ shader_glsl_texreg2ar
,
5086 /* WINED3DSIH_TEXREG2GB */ shader_glsl_texreg2gb
,
5087 /* WINED3DSIH_TEXREG2RGB */ shader_glsl_texreg2rgb
,
5088 /* WINED3DSIH_UTOF */ NULL
,
5091 static void shader_glsl_handle_instruction(const struct wined3d_shader_instruction
*ins
) {
5092 SHADER_HANDLER hw_fct
;
5094 /* Select handler */
5095 hw_fct
= shader_glsl_instruction_handler_table
[ins
->handler_idx
];
5097 /* Unhandled opcode */
5100 FIXME("Backend can't handle opcode %#x\n", ins
->handler_idx
);
5105 shader_glsl_add_instruction_modifiers(ins
);
5108 const shader_backend_t glsl_shader_backend
= {
5109 shader_glsl_handle_instruction
,
5111 shader_glsl_select_depth_blt
,
5112 shader_glsl_deselect_depth_blt
,
5113 shader_glsl_update_float_vertex_constants
,
5114 shader_glsl_update_float_pixel_constants
,
5115 shader_glsl_load_constants
,
5116 shader_glsl_load_np2fixup_constants
,
5117 shader_glsl_destroy
,
5120 shader_glsl_dirty_const
,
5121 shader_glsl_get_caps
,
5122 shader_glsl_color_fixup_supported
,