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(
685 IWineD3DDevice
* device
,
687 char useVertexShader
) {
689 const IWineD3DDeviceImpl
* deviceImpl
= (const IWineD3DDeviceImpl
*) device
;
690 const struct glsl_shader_prog_link
* prog
= ((struct shader_glsl_priv
*)(deviceImpl
->shader_priv
))->glsl_program
;
693 /* No GLSL program set - nothing to do. */
697 if (!usePixelShader
) {
698 /* NP2 texcoord fixup is (currently) only done for pixelshaders. */
702 if (prog
->ps_args
.np2_fixup
&& -1 != prog
->np2Fixup_location
) {
703 const struct wined3d_gl_info
*gl_info
= &deviceImpl
->adapter
->gl_info
;
704 const IWineD3DStateBlockImpl
* stateBlock
= (const IWineD3DStateBlockImpl
*) deviceImpl
->stateBlock
;
706 UINT fixup
= prog
->ps_args
.np2_fixup
;
707 GLfloat np2fixup_constants
[4 * MAX_FRAGMENT_SAMPLERS
];
709 for (i
= 0; fixup
; fixup
>>= 1, ++i
) {
710 const unsigned char idx
= prog
->np2Fixup_info
->idx
[i
];
711 const IWineD3DBaseTextureImpl
*tex
= stateBlock
->state
.textures
[i
];
712 GLfloat
*tex_dim
= &np2fixup_constants
[(idx
>> 1) * 4];
716 ERR("Nonexistent texture is flagged for NP2 texcoord fixup.\n");
721 tex_dim
[2] = tex
->baseTexture
.pow2Matrix
[0]; tex_dim
[3] = tex
->baseTexture
.pow2Matrix
[5];
723 tex_dim
[0] = tex
->baseTexture
.pow2Matrix
[0]; tex_dim
[1] = tex
->baseTexture
.pow2Matrix
[5];
727 GL_EXTCALL(glUniform4fvARB(prog
->np2Fixup_location
, prog
->np2Fixup_info
->num_consts
, np2fixup_constants
));
732 * Loads the app-supplied constants into the currently set GLSL program.
734 /* GL locking is done by the caller (state handler) */
735 static void shader_glsl_load_constants(const struct wined3d_context
*context
,
736 char usePixelShader
, char useVertexShader
)
738 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
739 IWineD3DDeviceImpl
*device
= context
->swapchain
->device
;
740 IWineD3DStateBlockImpl
* stateBlock
= device
->stateBlock
;
741 struct shader_glsl_priv
*priv
= device
->shader_priv
;
743 GLhandleARB programId
;
744 struct glsl_shader_prog_link
*prog
= priv
->glsl_program
;
745 UINT constant_version
;
749 /* No GLSL program set - nothing to do. */
752 programId
= prog
->programId
;
753 constant_version
= prog
->constant_version
;
757 IWineD3DBaseShaderImpl
*vshader
= (IWineD3DBaseShaderImpl
*)stateBlock
->state
.vertex_shader
;
759 /* Load DirectX 9 float constants/uniforms for vertex shader */
760 shader_glsl_load_constantsF(vshader
, gl_info
, stateBlock
->state
.vs_consts_f
,
761 prog
->vuniformF_locations
, &priv
->vconst_heap
, priv
->stack
, constant_version
);
763 /* Load DirectX 9 integer constants/uniforms for vertex shader */
764 shader_glsl_load_constantsI(vshader
, gl_info
, prog
->vuniformI_locations
, stateBlock
->state
.vs_consts_i
,
765 stateBlock
->changed
.vertexShaderConstantsI
& vshader
->baseShader
.reg_maps
.integer_constants
);
767 /* Load DirectX 9 boolean constants/uniforms for vertex shader */
768 shader_glsl_load_constantsB(vshader
, gl_info
, programId
, stateBlock
->state
.vs_consts_b
,
769 stateBlock
->changed
.vertexShaderConstantsB
& vshader
->baseShader
.reg_maps
.boolean_constants
);
771 /* Upload the position fixup params */
772 GL_EXTCALL(glUniform4fvARB(prog
->posFixup_location
, 1, &device
->posFixup
[0]));
773 checkGLcall("glUniform4fvARB");
778 IWineD3DBaseShaderImpl
*pshader
= (IWineD3DBaseShaderImpl
*)stateBlock
->state
.pixel_shader
;
780 /* Load DirectX 9 float constants/uniforms for pixel shader */
781 shader_glsl_load_constantsF(pshader
, gl_info
, stateBlock
->state
.ps_consts_f
,
782 prog
->puniformF_locations
, &priv
->pconst_heap
, priv
->stack
, constant_version
);
784 /* Load DirectX 9 integer constants/uniforms for pixel shader */
785 shader_glsl_load_constantsI(pshader
, gl_info
, prog
->puniformI_locations
, stateBlock
->state
.ps_consts_i
,
786 stateBlock
->changed
.pixelShaderConstantsI
& pshader
->baseShader
.reg_maps
.integer_constants
);
788 /* Load DirectX 9 boolean constants/uniforms for pixel shader */
789 shader_glsl_load_constantsB(pshader
, gl_info
, programId
, stateBlock
->state
.ps_consts_b
,
790 stateBlock
->changed
.pixelShaderConstantsB
& pshader
->baseShader
.reg_maps
.boolean_constants
);
792 /* Upload the environment bump map matrix if needed. The needsbumpmat member specifies the texture stage to load the matrix from.
793 * It can't be 0 for a valid texbem instruction.
795 for(i
= 0; i
< MAX_TEXTURES
; i
++) {
798 if(prog
->bumpenvmat_location
[i
] == -1) continue;
800 data
= (const float *)&stateBlock
->state
.texture_states
[i
][WINED3DTSS_BUMPENVMAT00
];
801 GL_EXTCALL(glUniformMatrix2fvARB(prog
->bumpenvmat_location
[i
], 1, 0, data
));
802 checkGLcall("glUniformMatrix2fvARB");
804 /* texbeml needs the luminance scale and offset too. If texbeml
805 * is used, needsbumpmat is set too, so we can check that in the
806 * needsbumpmat check. */
807 if (prog
->luminancescale_location
[i
] != -1)
809 const GLfloat
*scale
= (const GLfloat
*)&stateBlock
->state
.texture_states
[i
][WINED3DTSS_BUMPENVLSCALE
];
810 const GLfloat
*offset
= (const GLfloat
*)&stateBlock
->state
.texture_states
[i
][WINED3DTSS_BUMPENVLOFFSET
];
812 GL_EXTCALL(glUniform1fvARB(prog
->luminancescale_location
[i
], 1, scale
));
813 checkGLcall("glUniform1fvARB");
814 GL_EXTCALL(glUniform1fvARB(prog
->luminanceoffset_location
[i
], 1, offset
));
815 checkGLcall("glUniform1fvARB");
819 if(((IWineD3DPixelShaderImpl
*) pshader
)->vpos_uniform
) {
820 float correction_params
[4];
822 if (context
->render_offscreen
)
824 correction_params
[0] = 0.0f
;
825 correction_params
[1] = 1.0f
;
827 /* position is window relative, not viewport relative */
828 correction_params
[0] = context
->current_rt
->currentDesc
.Height
;
829 correction_params
[1] = -1.0f
;
831 GL_EXTCALL(glUniform4fvARB(prog
->ycorrection_location
, 1, correction_params
));
835 if (priv
->next_constant_version
== UINT_MAX
)
837 TRACE("Max constant version reached, resetting to 0.\n");
838 wine_rb_for_each_entry(&priv
->program_lookup
, reset_program_constant_version
, NULL
);
839 priv
->next_constant_version
= 1;
843 prog
->constant_version
= priv
->next_constant_version
++;
847 static inline void update_heap_entry(struct constant_heap
*heap
, unsigned int idx
,
848 unsigned int heap_idx
, DWORD new_version
)
850 struct constant_entry
*entries
= heap
->entries
;
851 unsigned int *positions
= heap
->positions
;
852 unsigned int parent_idx
;
856 parent_idx
= heap_idx
>> 1;
858 if (new_version
<= entries
[parent_idx
].version
) break;
860 entries
[heap_idx
] = entries
[parent_idx
];
861 positions
[entries
[parent_idx
].idx
] = heap_idx
;
862 heap_idx
= parent_idx
;
865 entries
[heap_idx
].version
= new_version
;
866 entries
[heap_idx
].idx
= idx
;
867 positions
[idx
] = heap_idx
;
870 static void shader_glsl_update_float_vertex_constants(IWineD3DDevice
*iface
, UINT start
, UINT count
)
872 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
873 struct shader_glsl_priv
*priv
= This
->shader_priv
;
874 struct constant_heap
*heap
= &priv
->vconst_heap
;
877 for (i
= start
; i
< count
+ start
; ++i
)
879 if (!This
->stateBlock
->changed
.vertexShaderConstantsF
[i
])
880 update_heap_entry(heap
, i
, heap
->size
++, priv
->next_constant_version
);
882 update_heap_entry(heap
, i
, heap
->positions
[i
], priv
->next_constant_version
);
886 static void shader_glsl_update_float_pixel_constants(IWineD3DDevice
*iface
, UINT start
, UINT count
)
888 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
889 struct shader_glsl_priv
*priv
= This
->shader_priv
;
890 struct constant_heap
*heap
= &priv
->pconst_heap
;
893 for (i
= start
; i
< count
+ start
; ++i
)
895 if (!This
->stateBlock
->changed
.pixelShaderConstantsF
[i
])
896 update_heap_entry(heap
, i
, heap
->size
++, priv
->next_constant_version
);
898 update_heap_entry(heap
, i
, heap
->positions
[i
], priv
->next_constant_version
);
902 static unsigned int vec4_varyings(DWORD shader_major
, const struct wined3d_gl_info
*gl_info
)
904 unsigned int ret
= gl_info
->limits
.glsl_varyings
/ 4;
905 /* 4.0 shaders do not write clip coords because d3d10 does not support user clipplanes */
906 if(shader_major
> 3) return ret
;
908 /* 3.0 shaders may need an extra varying for the clip coord on some cards(mostly dx10 ones) */
909 if (gl_info
->quirks
& WINED3D_QUIRK_GLSL_CLIP_VARYING
) ret
-= 1;
913 /** Generate the variable & register declarations for the GLSL output target */
914 static void shader_generate_glsl_declarations(const struct wined3d_context
*context
,
915 struct wined3d_shader_buffer
*buffer
, IWineD3DBaseShader
*iface
,
916 const shader_reg_maps
*reg_maps
, struct shader_glsl_ctx_priv
*ctx_priv
)
918 IWineD3DBaseShaderImpl
* This
= (IWineD3DBaseShaderImpl
*) iface
;
919 IWineD3DDeviceImpl
*device
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
920 const struct wined3d_state
*state
= &device
->stateBlock
->state
;
921 const struct ps_compile_args
*ps_args
= ctx_priv
->cur_ps_args
;
922 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
923 unsigned int i
, extra_constants_needed
= 0;
924 const local_constant
*lconst
;
927 /* There are some minor differences between pixel and vertex shaders */
928 char pshader
= shader_is_pshader_version(reg_maps
->shader_version
.type
);
929 char prefix
= pshader
? 'P' : 'V';
931 /* Prototype the subroutines */
932 for (i
= 0, map
= reg_maps
->labels
; map
; map
>>= 1, ++i
)
934 if (map
& 1) shader_addline(buffer
, "void subroutine%u();\n", i
);
937 /* Declare the constants (aka uniforms) */
938 if (This
->baseShader
.limits
.constant_float
> 0) {
939 unsigned max_constantsF
;
940 /* Unless the shader uses indirect addressing, always declare the maximum array size and ignore that we need some
941 * uniforms privately. E.g. if GL supports 256 uniforms, and we need 2 for the pos fixup and immediate values, still
942 * declare VC[256]. If the shader needs more uniforms than we have it won't work in any case. If it uses less, the
943 * compiler will figure out which uniforms are really used and strip them out. This allows a shader to use c255 on
944 * a dx9 card, as long as it doesn't also use all the other constants.
946 * If the shader uses indirect addressing the compiler must assume that all declared uniforms are used. In this case,
947 * declare only the amount that we're assured to have.
949 * Thus we run into problems in these two cases:
950 * 1) The shader really uses more uniforms than supported
951 * 2) The shader uses indirect addressing, less constants than supported, but uses a constant index > #supported consts
955 /* No indirect addressing here. */
956 max_constantsF
= gl_info
->limits
.glsl_ps_float_constants
;
960 if(This
->baseShader
.reg_maps
.usesrelconstF
) {
961 /* Subtract the other potential uniforms from the max available (bools, ints, and 1 row of projection matrix).
962 * Subtract another uniform for immediate values, which have to be loaded via uniform by the driver as well.
963 * The shader code only uses 0.5, 2.0, 1.0, 128 and -128 in vertex shader code, so one vec4 should be enough
964 * (Unfortunately the Nvidia driver doesn't store 128 and -128 in one float).
966 * Writing gl_ClipVertex requires one uniform for each clipplane as well.
968 max_constantsF
= gl_info
->limits
.glsl_vs_float_constants
- 3;
969 if(ctx_priv
->cur_vs_args
->clip_enabled
)
971 max_constantsF
-= gl_info
->limits
.clipplanes
;
973 max_constantsF
-= count_bits(This
->baseShader
.reg_maps
.integer_constants
);
974 /* Strictly speaking a bool only uses one scalar, but the nvidia(Linux) compiler doesn't pack them properly,
975 * so each scalar requires a full vec4. We could work around this by packing the booleans ourselves, but
976 * for now take this into account when calculating the number of available constants
978 max_constantsF
-= count_bits(This
->baseShader
.reg_maps
.boolean_constants
);
979 /* Set by driver quirks in directx.c */
980 max_constantsF
-= gl_info
->reserved_glsl_constants
;
984 max_constantsF
= gl_info
->limits
.glsl_vs_float_constants
;
987 max_constantsF
= min(This
->baseShader
.limits
.constant_float
, max_constantsF
);
988 shader_addline(buffer
, "uniform vec4 %cC[%u];\n", prefix
, max_constantsF
);
991 /* Always declare the full set of constants, the compiler can remove the unused ones because d3d doesn't(yet)
992 * support indirect int and bool constant addressing. This avoids problems if the app uses e.g. i0 and i9.
994 if (This
->baseShader
.limits
.constant_int
> 0 && This
->baseShader
.reg_maps
.integer_constants
)
995 shader_addline(buffer
, "uniform ivec4 %cI[%u];\n", prefix
, This
->baseShader
.limits
.constant_int
);
997 if (This
->baseShader
.limits
.constant_bool
> 0 && This
->baseShader
.reg_maps
.boolean_constants
)
998 shader_addline(buffer
, "uniform bool %cB[%u];\n", prefix
, This
->baseShader
.limits
.constant_bool
);
1002 shader_addline(buffer
, "uniform vec4 posFixup;\n");
1003 shader_addline(buffer
, "void order_ps_input(in vec4[%u]);\n", MAX_REG_OUTPUT
);
1007 for (i
= 0, map
= reg_maps
->bumpmat
; map
; map
>>= 1, ++i
)
1009 if (!(map
& 1)) continue;
1011 shader_addline(buffer
, "uniform mat2 bumpenvmat%d;\n", i
);
1013 if (reg_maps
->luminanceparams
& (1 << i
))
1015 shader_addline(buffer
, "uniform float luminancescale%d;\n", i
);
1016 shader_addline(buffer
, "uniform float luminanceoffset%d;\n", i
);
1017 extra_constants_needed
++;
1020 extra_constants_needed
++;
1023 if (ps_args
->srgb_correction
)
1025 shader_addline(buffer
, "const vec4 srgb_const0 = vec4(%.8e, %.8e, %.8e, %.8e);\n",
1026 srgb_pow
, srgb_mul_high
, srgb_sub_high
, srgb_mul_low
);
1027 shader_addline(buffer
, "const vec4 srgb_const1 = vec4(%.8e, 0.0, 0.0, 0.0);\n",
1030 if (reg_maps
->vpos
|| reg_maps
->usesdsy
)
1032 if (This
->baseShader
.limits
.constant_float
+ extra_constants_needed
1033 + 1 < gl_info
->limits
.glsl_ps_float_constants
)
1035 shader_addline(buffer
, "uniform vec4 ycorrection;\n");
1036 ((IWineD3DPixelShaderImpl
*) This
)->vpos_uniform
= 1;
1037 extra_constants_needed
++;
1039 /* This happens because we do not have proper tracking of the constant registers that are
1040 * actually used, only the max limit of the shader version
1042 FIXME("Cannot find a free uniform for vpos correction params\n");
1043 shader_addline(buffer
, "const vec4 ycorrection = vec4(%f, %f, 0.0, 0.0);\n",
1044 context
->render_offscreen
? 0.0f
: device
->render_targets
[0]->currentDesc
.Height
,
1045 context
->render_offscreen
? 1.0f
: -1.0f
);
1047 shader_addline(buffer
, "vec4 vpos;\n");
1051 /* Declare texture samplers */
1052 for (i
= 0; i
< This
->baseShader
.limits
.sampler
; i
++) {
1053 if (reg_maps
->sampler_type
[i
])
1055 IWineD3DBaseTextureImpl
*texture
;
1057 switch (reg_maps
->sampler_type
[i
])
1060 if (pshader
&& ps_args
->shadow
& (1 << i
))
1061 shader_addline(buffer
, "uniform sampler1DShadow %csampler%u;\n", prefix
, i
);
1063 shader_addline(buffer
, "uniform sampler1D %csampler%u;\n", prefix
, i
);
1066 texture
= state
->textures
[i
];
1067 if (pshader
&& ps_args
->shadow
& (1 << i
))
1069 if (texture
&& texture
->baseTexture
.target
== GL_TEXTURE_RECTANGLE_ARB
)
1070 shader_addline(buffer
, "uniform sampler2DRectShadow %csampler%u;\n", prefix
, i
);
1072 shader_addline(buffer
, "uniform sampler2DShadow %csampler%u;\n", prefix
, i
);
1076 if (texture
&& texture
->baseTexture
.target
== GL_TEXTURE_RECTANGLE_ARB
)
1077 shader_addline(buffer
, "uniform sampler2DRect %csampler%u;\n", prefix
, i
);
1079 shader_addline(buffer
, "uniform sampler2D %csampler%u;\n", prefix
, i
);
1082 case WINED3DSTT_CUBE
:
1083 if (pshader
&& ps_args
->shadow
& (1 << i
)) FIXME("Unsupported Cube shadow sampler.\n");
1084 shader_addline(buffer
, "uniform samplerCube %csampler%u;\n", prefix
, i
);
1086 case WINED3DSTT_VOLUME
:
1087 if (pshader
&& ps_args
->shadow
& (1 << i
)) FIXME("Unsupported 3D shadow sampler.\n");
1088 shader_addline(buffer
, "uniform sampler3D %csampler%u;\n", prefix
, i
);
1091 shader_addline(buffer
, "uniform unsupported_sampler %csampler%u;\n", prefix
, i
);
1092 FIXME("Unrecognized sampler type: %#x\n", reg_maps
->sampler_type
[i
]);
1098 /* Declare uniforms for NP2 texcoord fixup:
1099 * This is NOT done inside the loop that declares the texture samplers since the NP2 fixup code
1100 * is currently only used for the GeforceFX series and when forcing the ARB_npot extension off.
1101 * Modern cards just skip the code anyway, so put it inside a separate loop. */
1102 if (pshader
&& ps_args
->np2_fixup
) {
1104 struct ps_np2fixup_info
* const fixup
= ctx_priv
->cur_np2fixup_info
;
1107 /* NP2/RECT textures in OpenGL use texcoords in the range [0,width]x[0,height]
1108 * while D3D has them in the (normalized) [0,1]x[0,1] range.
1109 * samplerNP2Fixup stores texture dimensions and is updated through
1110 * shader_glsl_load_np2fixup_constants when the sampler changes. */
1112 for (i
= 0; i
< This
->baseShader
.limits
.sampler
; ++i
) {
1113 if (reg_maps
->sampler_type
[i
]) {
1114 if (!(ps_args
->np2_fixup
& (1 << i
))) continue;
1116 if (WINED3DSTT_2D
!= reg_maps
->sampler_type
[i
]) {
1117 FIXME("Non-2D texture is flagged for NP2 texcoord fixup.\n");
1121 fixup
->idx
[i
] = cur
++;
1125 fixup
->num_consts
= (cur
+ 1) >> 1;
1126 shader_addline(buffer
, "uniform vec4 %csamplerNP2Fixup[%u];\n", prefix
, fixup
->num_consts
);
1129 /* Declare address variables */
1130 for (i
= 0, map
= reg_maps
->address
; map
; map
>>= 1, ++i
)
1132 if (map
& 1) shader_addline(buffer
, "ivec4 A%u;\n", i
);
1135 /* Declare texture coordinate temporaries and initialize them */
1136 for (i
= 0, map
= reg_maps
->texcoord
; map
; map
>>= 1, ++i
)
1138 if (map
& 1) shader_addline(buffer
, "vec4 T%u = gl_TexCoord[%u];\n", i
, i
);
1141 /* Declare input register varyings. Only pixel shader, vertex shaders have that declared in the
1142 * helper function shader that is linked in at link time
1144 if (pshader
&& reg_maps
->shader_version
.major
>= 3)
1148 shader_addline(buffer
, "varying vec4 IN[%u];\n", vec4_varyings(reg_maps
->shader_version
.major
, gl_info
));
1150 /* TODO: Write a replacement shader for the fixed function vertex pipeline, so this isn't needed.
1151 * For fixed function vertex processing + 3.0 pixel shader we need a separate function in the
1152 * pixel shader that reads the fixed function color into the packed input registers.
1154 shader_addline(buffer
, "vec4 IN[%u];\n", vec4_varyings(reg_maps
->shader_version
.major
, gl_info
));
1158 /* Declare output register temporaries */
1159 if(This
->baseShader
.limits
.packed_output
) {
1160 shader_addline(buffer
, "vec4 OUT[%u];\n", This
->baseShader
.limits
.packed_output
);
1163 /* Declare temporary variables */
1164 for (i
= 0, map
= reg_maps
->temporary
; map
; map
>>= 1, ++i
)
1166 if (map
& 1) shader_addline(buffer
, "vec4 R%u;\n", i
);
1169 /* Declare attributes */
1170 if (reg_maps
->shader_version
.type
== WINED3D_SHADER_TYPE_VERTEX
)
1172 for (i
= 0, map
= reg_maps
->input_registers
; map
; map
>>= 1, ++i
)
1174 if (map
& 1) shader_addline(buffer
, "attribute vec4 attrib%i;\n", i
);
1178 /* Declare loop registers aLx */
1179 for (i
= 0; i
< reg_maps
->loop_depth
; i
++) {
1180 shader_addline(buffer
, "int aL%u;\n", i
);
1181 shader_addline(buffer
, "int tmpInt%u;\n", i
);
1184 /* Temporary variables for matrix operations */
1185 shader_addline(buffer
, "vec4 tmp0;\n");
1186 shader_addline(buffer
, "vec4 tmp1;\n");
1188 /* Local constants use a different name so they can be loaded once at shader link time
1189 * They can't be hardcoded into the shader text via LC = {x, y, z, w}; because the
1190 * float -> string conversion can cause precision loss.
1192 if(!This
->baseShader
.load_local_constsF
) {
1193 LIST_FOR_EACH_ENTRY(lconst
, &This
->baseShader
.constantsF
, local_constant
, entry
) {
1194 shader_addline(buffer
, "uniform vec4 %cLC%u;\n", prefix
, lconst
->idx
);
1198 /* Start the main program */
1199 shader_addline(buffer
, "void main() {\n");
1200 if(pshader
&& reg_maps
->vpos
) {
1201 /* DirectX apps expect integer values, while OpenGL drivers add approximately 0.5. This causes
1202 * off-by-one problems as spotted by the vPos d3d9 visual test. Unfortunately the ATI cards do
1203 * not add exactly 0.5, but rather something like 0.49999999 or 0.50000001, which still causes
1204 * precision troubles when we just substract 0.5.
1206 * To deal with that just floor() the position. This will eliminate the fraction on all cards.
1208 * TODO: Test how that behaves with multisampling once we can enable multisampling in winex11.
1210 * An advantage of floor is that it works even if the driver doesn't add 1/2. It is somewhat
1211 * questionable if 1.5, 2.5, ... are the proper values to return in gl_FragCoord, even though
1212 * coordinates specify the pixel centers instead of the pixel corners. This code will behave
1213 * correctly on drivers that returns integer values.
1215 shader_addline(buffer
, "vpos = floor(vec4(0, ycorrection[0], 0, 0) + gl_FragCoord * vec4(1, ycorrection[1], 1, 1));\n");
1219 /*****************************************************************************
1220 * Functions to generate GLSL strings from DirectX Shader bytecode begin here.
1222 * For more information, see http://wiki.winehq.org/DirectX-Shaders
1223 ****************************************************************************/
1226 static void shader_glsl_add_src_param(const struct wined3d_shader_instruction
*ins
,
1227 const struct wined3d_shader_src_param
*wined3d_src
, DWORD mask
, glsl_src_param_t
*glsl_src
);
1229 /** Used for opcode modifiers - They multiply the result by the specified amount */
1230 static const char * const shift_glsl_tab
[] = {
1232 "2.0 * ", /* 1 (x2) */
1233 "4.0 * ", /* 2 (x4) */
1234 "8.0 * ", /* 3 (x8) */
1235 "16.0 * ", /* 4 (x16) */
1236 "32.0 * ", /* 5 (x32) */
1243 "0.0625 * ", /* 12 (d16) */
1244 "0.125 * ", /* 13 (d8) */
1245 "0.25 * ", /* 14 (d4) */
1246 "0.5 * " /* 15 (d2) */
1249 /* Generate a GLSL parameter that does the input modifier computation and return the input register/mask to use */
1250 static void shader_glsl_gen_modifier(DWORD src_modifier
, const char *in_reg
, const char *in_regswizzle
, char *out_str
)
1254 switch (src_modifier
)
1256 case WINED3DSPSM_DZ
: /* Need to handle this in the instructions itself (texld & texcrd). */
1257 case WINED3DSPSM_DW
:
1258 case WINED3DSPSM_NONE
:
1259 sprintf(out_str
, "%s%s", in_reg
, in_regswizzle
);
1261 case WINED3DSPSM_NEG
:
1262 sprintf(out_str
, "-%s%s", in_reg
, in_regswizzle
);
1264 case WINED3DSPSM_NOT
:
1265 sprintf(out_str
, "!%s%s", in_reg
, in_regswizzle
);
1267 case WINED3DSPSM_BIAS
:
1268 sprintf(out_str
, "(%s%s - vec4(0.5)%s)", in_reg
, in_regswizzle
, in_regswizzle
);
1270 case WINED3DSPSM_BIASNEG
:
1271 sprintf(out_str
, "-(%s%s - vec4(0.5)%s)", in_reg
, in_regswizzle
, in_regswizzle
);
1273 case WINED3DSPSM_SIGN
:
1274 sprintf(out_str
, "(2.0 * (%s%s - 0.5))", in_reg
, in_regswizzle
);
1276 case WINED3DSPSM_SIGNNEG
:
1277 sprintf(out_str
, "-(2.0 * (%s%s - 0.5))", in_reg
, in_regswizzle
);
1279 case WINED3DSPSM_COMP
:
1280 sprintf(out_str
, "(1.0 - %s%s)", in_reg
, in_regswizzle
);
1282 case WINED3DSPSM_X2
:
1283 sprintf(out_str
, "(2.0 * %s%s)", in_reg
, in_regswizzle
);
1285 case WINED3DSPSM_X2NEG
:
1286 sprintf(out_str
, "-(2.0 * %s%s)", in_reg
, in_regswizzle
);
1288 case WINED3DSPSM_ABS
:
1289 sprintf(out_str
, "abs(%s%s)", in_reg
, in_regswizzle
);
1291 case WINED3DSPSM_ABSNEG
:
1292 sprintf(out_str
, "-abs(%s%s)", in_reg
, in_regswizzle
);
1295 FIXME("Unhandled modifier %u\n", src_modifier
);
1296 sprintf(out_str
, "%s%s", in_reg
, in_regswizzle
);
1300 /** Writes the GLSL variable name that corresponds to the register that the
1301 * DX opcode parameter is trying to access */
1302 static void shader_glsl_get_register_name(const struct wined3d_shader_register
*reg
,
1303 char *register_name
, BOOL
*is_color
, const struct wined3d_shader_instruction
*ins
)
1305 /* oPos, oFog and oPts in D3D */
1306 static const char * const hwrastout_reg_names
[] = {"OUT[10]", "OUT[11].x", "OUT[11].y"};
1308 IWineD3DBaseShaderImpl
*This
= (IWineD3DBaseShaderImpl
*)ins
->ctx
->shader
;
1309 const struct wined3d_gl_info
*gl_info
= ins
->ctx
->gl_info
;
1310 char pshader
= shader_is_pshader_version(This
->baseShader
.reg_maps
.shader_version
.type
);
1316 case WINED3DSPR_TEMP
:
1317 sprintf(register_name
, "R%u", reg
->idx
);
1320 case WINED3DSPR_INPUT
:
1321 /* vertex shaders */
1324 struct shader_glsl_ctx_priv
*priv
= ins
->ctx
->backend_data
;
1325 if (priv
->cur_vs_args
->swizzle_map
& (1 << reg
->idx
)) *is_color
= TRUE
;
1326 sprintf(register_name
, "attrib%u", reg
->idx
);
1330 /* pixel shaders >= 3.0 */
1331 if (This
->baseShader
.reg_maps
.shader_version
.major
>= 3)
1333 DWORD idx
= ((IWineD3DPixelShaderImpl
*)This
)->input_reg_map
[reg
->idx
];
1334 unsigned int in_count
= vec4_varyings(This
->baseShader
.reg_maps
.shader_version
.major
, gl_info
);
1338 glsl_src_param_t rel_param
;
1340 shader_glsl_add_src_param(ins
, reg
->rel_addr
, WINED3DSP_WRITEMASK_0
, &rel_param
);
1342 /* Removing a + 0 would be an obvious optimization, but macos doesn't see the NOP
1343 * operation there */
1346 if (((IWineD3DPixelShaderImpl
*)This
)->declared_in_count
> in_count
)
1348 sprintf(register_name
,
1349 "((%s + %u) > %d ? (%s + %u) > %d ? gl_SecondaryColor : gl_Color : IN[%s + %u])",
1350 rel_param
.param_str
, idx
, in_count
- 1, rel_param
.param_str
, idx
, in_count
,
1351 rel_param
.param_str
, idx
);
1355 sprintf(register_name
, "IN[%s + %u]", rel_param
.param_str
, idx
);
1360 if (((IWineD3DPixelShaderImpl
*)This
)->declared_in_count
> in_count
)
1362 sprintf(register_name
, "((%s) > %d ? (%s) > %d ? gl_SecondaryColor : gl_Color : IN[%s])",
1363 rel_param
.param_str
, in_count
- 1, rel_param
.param_str
, in_count
,
1364 rel_param
.param_str
);
1368 sprintf(register_name
, "IN[%s]", rel_param
.param_str
);
1374 if (idx
== in_count
) sprintf(register_name
, "gl_Color");
1375 else if (idx
== in_count
+ 1) sprintf(register_name
, "gl_SecondaryColor");
1376 else sprintf(register_name
, "IN[%u]", idx
);
1381 if (!reg
->idx
) strcpy(register_name
, "gl_Color");
1382 else strcpy(register_name
, "gl_SecondaryColor");
1387 case WINED3DSPR_CONST
:
1389 const char prefix
= pshader
? 'P' : 'V';
1391 /* Relative addressing */
1394 glsl_src_param_t rel_param
;
1395 shader_glsl_add_src_param(ins
, reg
->rel_addr
, WINED3DSP_WRITEMASK_0
, &rel_param
);
1396 if (reg
->idx
) sprintf(register_name
, "%cC[%s + %u]", prefix
, rel_param
.param_str
, reg
->idx
);
1397 else sprintf(register_name
, "%cC[%s]", prefix
, rel_param
.param_str
);
1401 if (shader_constant_is_local(This
, reg
->idx
))
1402 sprintf(register_name
, "%cLC%u", prefix
, reg
->idx
);
1404 sprintf(register_name
, "%cC[%u]", prefix
, reg
->idx
);
1409 case WINED3DSPR_CONSTINT
:
1410 if (pshader
) sprintf(register_name
, "PI[%u]", reg
->idx
);
1411 else sprintf(register_name
, "VI[%u]", reg
->idx
);
1414 case WINED3DSPR_CONSTBOOL
:
1415 if (pshader
) sprintf(register_name
, "PB[%u]", reg
->idx
);
1416 else sprintf(register_name
, "VB[%u]", reg
->idx
);
1419 case WINED3DSPR_TEXTURE
: /* case WINED3DSPR_ADDR: */
1420 if (pshader
) sprintf(register_name
, "T%u", reg
->idx
);
1421 else sprintf(register_name
, "A%u", reg
->idx
);
1424 case WINED3DSPR_LOOP
:
1425 sprintf(register_name
, "aL%u", This
->baseShader
.cur_loop_regno
- 1);
1428 case WINED3DSPR_SAMPLER
:
1429 if (pshader
) sprintf(register_name
, "Psampler%u", reg
->idx
);
1430 else sprintf(register_name
, "Vsampler%u", reg
->idx
);
1433 case WINED3DSPR_COLOROUT
:
1434 if (reg
->idx
>= gl_info
->limits
.buffers
)
1435 WARN("Write to render target %u, only %d supported.\n", reg
->idx
, gl_info
->limits
.buffers
);
1437 sprintf(register_name
, "gl_FragData[%u]", reg
->idx
);
1440 case WINED3DSPR_RASTOUT
:
1441 sprintf(register_name
, "%s", hwrastout_reg_names
[reg
->idx
]);
1444 case WINED3DSPR_DEPTHOUT
:
1445 sprintf(register_name
, "gl_FragDepth");
1448 case WINED3DSPR_ATTROUT
:
1449 if (!reg
->idx
) sprintf(register_name
, "OUT[8]");
1450 else sprintf(register_name
, "OUT[9]");
1453 case WINED3DSPR_TEXCRDOUT
:
1454 /* Vertex shaders >= 3.0: WINED3DSPR_OUTPUT */
1455 sprintf(register_name
, "OUT[%u]", reg
->idx
);
1458 case WINED3DSPR_MISCTYPE
:
1462 sprintf(register_name
, "vpos");
1464 else if (reg
->idx
== 1)
1466 /* Note that gl_FrontFacing is a bool, while vFace is
1467 * a float for which the sign determines front/back */
1468 sprintf(register_name
, "(gl_FrontFacing ? 1.0 : -1.0)");
1472 FIXME("Unhandled misctype register %d\n", reg
->idx
);
1473 sprintf(register_name
, "unrecognized_register");
1477 case WINED3DSPR_IMMCONST
:
1478 switch (reg
->immconst_type
)
1480 case WINED3D_IMMCONST_FLOAT
:
1481 sprintf(register_name
, "%.8e", *(const float *)reg
->immconst_data
);
1484 case WINED3D_IMMCONST_FLOAT4
:
1485 sprintf(register_name
, "vec4(%.8e, %.8e, %.8e, %.8e)",
1486 *(const float *)®
->immconst_data
[0], *(const float *)®
->immconst_data
[1],
1487 *(const float *)®
->immconst_data
[2], *(const float *)®
->immconst_data
[3]);
1491 FIXME("Unhandled immconst type %#x\n", reg
->immconst_type
);
1492 sprintf(register_name
, "<unhandled_immconst_type %#x>", reg
->immconst_type
);
1497 FIXME("Unhandled register name Type(%d)\n", reg
->type
);
1498 sprintf(register_name
, "unrecognized_register");
1503 static void shader_glsl_write_mask_to_str(DWORD write_mask
, char *str
)
1506 if (write_mask
& WINED3DSP_WRITEMASK_0
) *str
++ = 'x';
1507 if (write_mask
& WINED3DSP_WRITEMASK_1
) *str
++ = 'y';
1508 if (write_mask
& WINED3DSP_WRITEMASK_2
) *str
++ = 'z';
1509 if (write_mask
& WINED3DSP_WRITEMASK_3
) *str
++ = 'w';
1513 /* Get the GLSL write mask for the destination register */
1514 static DWORD
shader_glsl_get_write_mask(const struct wined3d_shader_dst_param
*param
, char *write_mask
)
1516 DWORD mask
= param
->write_mask
;
1518 if (shader_is_scalar(¶m
->reg
))
1520 mask
= WINED3DSP_WRITEMASK_0
;
1525 shader_glsl_write_mask_to_str(mask
, write_mask
);
1531 static unsigned int shader_glsl_get_write_mask_size(DWORD write_mask
) {
1532 unsigned int size
= 0;
1534 if (write_mask
& WINED3DSP_WRITEMASK_0
) ++size
;
1535 if (write_mask
& WINED3DSP_WRITEMASK_1
) ++size
;
1536 if (write_mask
& WINED3DSP_WRITEMASK_2
) ++size
;
1537 if (write_mask
& WINED3DSP_WRITEMASK_3
) ++size
;
1542 static void shader_glsl_swizzle_to_str(const DWORD swizzle
, BOOL fixup
, DWORD mask
, char *str
)
1544 /* For registers of type WINED3DDECLTYPE_D3DCOLOR, data is stored as "bgra",
1545 * but addressed as "rgba". To fix this we need to swap the register's x
1546 * and z components. */
1547 const char *swizzle_chars
= fixup
? "zyxw" : "xyzw";
1550 /* swizzle bits fields: wwzzyyxx */
1551 if (mask
& WINED3DSP_WRITEMASK_0
) *str
++ = swizzle_chars
[swizzle
& 0x03];
1552 if (mask
& WINED3DSP_WRITEMASK_1
) *str
++ = swizzle_chars
[(swizzle
>> 2) & 0x03];
1553 if (mask
& WINED3DSP_WRITEMASK_2
) *str
++ = swizzle_chars
[(swizzle
>> 4) & 0x03];
1554 if (mask
& WINED3DSP_WRITEMASK_3
) *str
++ = swizzle_chars
[(swizzle
>> 6) & 0x03];
1558 static void shader_glsl_get_swizzle(const struct wined3d_shader_src_param
*param
,
1559 BOOL fixup
, DWORD mask
, char *swizzle_str
)
1561 if (shader_is_scalar(¶m
->reg
))
1562 *swizzle_str
= '\0';
1564 shader_glsl_swizzle_to_str(param
->swizzle
, fixup
, mask
, swizzle_str
);
1567 /* From a given parameter token, generate the corresponding GLSL string.
1568 * Also, return the actual register name and swizzle in case the
1569 * caller needs this information as well. */
1570 static void shader_glsl_add_src_param(const struct wined3d_shader_instruction
*ins
,
1571 const struct wined3d_shader_src_param
*wined3d_src
, DWORD mask
, glsl_src_param_t
*glsl_src
)
1573 BOOL is_color
= FALSE
;
1574 char swizzle_str
[6];
1576 glsl_src
->reg_name
[0] = '\0';
1577 glsl_src
->param_str
[0] = '\0';
1578 swizzle_str
[0] = '\0';
1580 shader_glsl_get_register_name(&wined3d_src
->reg
, glsl_src
->reg_name
, &is_color
, ins
);
1581 shader_glsl_get_swizzle(wined3d_src
, is_color
, mask
, swizzle_str
);
1582 shader_glsl_gen_modifier(wined3d_src
->modifiers
, glsl_src
->reg_name
, swizzle_str
, glsl_src
->param_str
);
1585 /* From a given parameter token, generate the corresponding GLSL string.
1586 * Also, return the actual register name and swizzle in case the
1587 * caller needs this information as well. */
1588 static DWORD
shader_glsl_add_dst_param(const struct wined3d_shader_instruction
*ins
,
1589 const struct wined3d_shader_dst_param
*wined3d_dst
, glsl_dst_param_t
*glsl_dst
)
1591 BOOL is_color
= FALSE
;
1593 glsl_dst
->mask_str
[0] = '\0';
1594 glsl_dst
->reg_name
[0] = '\0';
1596 shader_glsl_get_register_name(&wined3d_dst
->reg
, glsl_dst
->reg_name
, &is_color
, ins
);
1597 return shader_glsl_get_write_mask(wined3d_dst
, glsl_dst
->mask_str
);
1600 /* Append the destination part of the instruction to the buffer, return the effective write mask */
1601 static DWORD
shader_glsl_append_dst_ext(struct wined3d_shader_buffer
*buffer
,
1602 const struct wined3d_shader_instruction
*ins
, const struct wined3d_shader_dst_param
*dst
)
1604 glsl_dst_param_t glsl_dst
;
1607 mask
= shader_glsl_add_dst_param(ins
, dst
, &glsl_dst
);
1608 if (mask
) shader_addline(buffer
, "%s%s = %s(", glsl_dst
.reg_name
, glsl_dst
.mask_str
, shift_glsl_tab
[dst
->shift
]);
1613 /* Append the destination part of the instruction to the buffer, return the effective write mask */
1614 static DWORD
shader_glsl_append_dst(struct wined3d_shader_buffer
*buffer
, const struct wined3d_shader_instruction
*ins
)
1616 return shader_glsl_append_dst_ext(buffer
, ins
, &ins
->dst
[0]);
1619 /** Process GLSL instruction modifiers */
1620 static void shader_glsl_add_instruction_modifiers(const struct wined3d_shader_instruction
*ins
)
1622 glsl_dst_param_t dst_param
;
1625 if (!ins
->dst_count
) return;
1627 modifiers
= ins
->dst
[0].modifiers
;
1628 if (!modifiers
) return;
1630 shader_glsl_add_dst_param(ins
, &ins
->dst
[0], &dst_param
);
1632 if (modifiers
& WINED3DSPDM_SATURATE
)
1634 /* _SAT means to clamp the value of the register to between 0 and 1 */
1635 shader_addline(ins
->ctx
->buffer
, "%s%s = clamp(%s%s, 0.0, 1.0);\n", dst_param
.reg_name
,
1636 dst_param
.mask_str
, dst_param
.reg_name
, dst_param
.mask_str
);
1639 if (modifiers
& WINED3DSPDM_MSAMPCENTROID
)
1641 FIXME("_centroid modifier not handled\n");
1644 if (modifiers
& WINED3DSPDM_PARTIALPRECISION
)
1646 /* MSDN says this modifier can be safely ignored, so that's what we'll do. */
1650 static inline const char *shader_get_comp_op(DWORD op
)
1653 case COMPARISON_GT
: return ">";
1654 case COMPARISON_EQ
: return "==";
1655 case COMPARISON_GE
: return ">=";
1656 case COMPARISON_LT
: return "<";
1657 case COMPARISON_NE
: return "!=";
1658 case COMPARISON_LE
: return "<=";
1660 FIXME("Unrecognized comparison value: %u\n", op
);
1665 static void shader_glsl_get_sample_function(const struct wined3d_shader_context
*ctx
,
1666 DWORD sampler_idx
, DWORD flags
, glsl_sample_function_t
*sample_function
)
1668 WINED3DSAMPLER_TEXTURE_TYPE sampler_type
= ctx
->reg_maps
->sampler_type
[sampler_idx
];
1669 const struct wined3d_gl_info
*gl_info
= ctx
->gl_info
;
1670 BOOL shadow
= shader_is_pshader_version(ctx
->reg_maps
->shader_version
.type
)
1671 && (((const struct shader_glsl_ctx_priv
*)ctx
->backend_data
)->cur_ps_args
->shadow
& (1 << sampler_idx
));
1672 BOOL projected
= flags
& WINED3D_GLSL_SAMPLE_PROJECTED
;
1673 BOOL texrect
= flags
& WINED3D_GLSL_SAMPLE_RECT
;
1674 BOOL lod
= flags
& WINED3D_GLSL_SAMPLE_LOD
;
1675 BOOL grad
= flags
& WINED3D_GLSL_SAMPLE_GRAD
;
1677 /* Note that there's no such thing as a projected cube texture. */
1678 switch(sampler_type
) {
1684 sample_function
->name
= projected
? "shadow1DProjLod" : "shadow1DLod";
1688 if (gl_info
->supported
[EXT_GPU_SHADER4
])
1689 sample_function
->name
= projected
? "shadow1DProjGrad" : "shadow1DGrad";
1690 else if (gl_info
->supported
[ARB_SHADER_TEXTURE_LOD
])
1691 sample_function
->name
= projected
? "shadow1DProjGradARB" : "shadow1DGradARB";
1694 FIXME("Unsupported 1D shadow grad function.\n");
1695 sample_function
->name
= "unsupported1DGrad";
1700 sample_function
->name
= projected
? "shadow1DProj" : "shadow1D";
1702 sample_function
->coord_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
;
1708 sample_function
->name
= projected
? "texture1DProjLod" : "texture1DLod";
1712 if (gl_info
->supported
[EXT_GPU_SHADER4
])
1713 sample_function
->name
= projected
? "texture1DProjGrad" : "texture1DGrad";
1714 else if (gl_info
->supported
[ARB_SHADER_TEXTURE_LOD
])
1715 sample_function
->name
= projected
? "texture1DProjGradARB" : "texture1DGradARB";
1718 FIXME("Unsupported 1D grad function.\n");
1719 sample_function
->name
= "unsupported1DGrad";
1724 sample_function
->name
= projected
? "texture1DProj" : "texture1D";
1726 sample_function
->coord_mask
= WINED3DSP_WRITEMASK_0
;
1737 sample_function
->name
= projected
? "shadow2DRectProjLod" : "shadow2DRectLod";
1741 if (gl_info
->supported
[EXT_GPU_SHADER4
])
1742 sample_function
->name
= projected
? "shadow2DRectProjGrad" : "shadow2DRectGrad";
1743 else if (gl_info
->supported
[ARB_SHADER_TEXTURE_LOD
])
1744 sample_function
->name
= projected
? "shadow2DRectProjGradARB" : "shadow2DRectGradARB";
1747 FIXME("Unsupported RECT shadow grad function.\n");
1748 sample_function
->name
= "unsupported2DRectGrad";
1753 sample_function
->name
= projected
? "shadow2DRectProj" : "shadow2DRect";
1760 sample_function
->name
= projected
? "shadow2DProjLod" : "shadow2DLod";
1764 if (gl_info
->supported
[EXT_GPU_SHADER4
])
1765 sample_function
->name
= projected
? "shadow2DProjGrad" : "shadow2DGrad";
1766 else if (gl_info
->supported
[ARB_SHADER_TEXTURE_LOD
])
1767 sample_function
->name
= projected
? "shadow2DProjGradARB" : "shadow2DGradARB";
1770 FIXME("Unsupported 2D shadow grad function.\n");
1771 sample_function
->name
= "unsupported2DGrad";
1776 sample_function
->name
= projected
? "shadow2DProj" : "shadow2D";
1779 sample_function
->coord_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
1787 sample_function
->name
= projected
? "texture2DRectProjLod" : "texture2DRectLod";
1791 if (gl_info
->supported
[EXT_GPU_SHADER4
])
1792 sample_function
->name
= projected
? "texture2DRectProjGrad" : "texture2DRectGrad";
1793 else if (gl_info
->supported
[ARB_SHADER_TEXTURE_LOD
])
1794 sample_function
->name
= projected
? "texture2DRectProjGradARB" : "texture2DRectGradARB";
1797 FIXME("Unsupported RECT grad function.\n");
1798 sample_function
->name
= "unsupported2DRectGrad";
1803 sample_function
->name
= projected
? "texture2DRectProj" : "texture2DRect";
1810 sample_function
->name
= projected
? "texture2DProjLod" : "texture2DLod";
1814 if (gl_info
->supported
[EXT_GPU_SHADER4
])
1815 sample_function
->name
= projected
? "texture2DProjGrad" : "texture2DGrad";
1816 else if (gl_info
->supported
[ARB_SHADER_TEXTURE_LOD
])
1817 sample_function
->name
= projected
? "texture2DProjGradARB" : "texture2DGradARB";
1820 FIXME("Unsupported 2D grad function.\n");
1821 sample_function
->name
= "unsupported2DGrad";
1826 sample_function
->name
= projected
? "texture2DProj" : "texture2D";
1829 sample_function
->coord_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
;
1833 case WINED3DSTT_CUBE
:
1836 FIXME("Unsupported Cube shadow function.\n");
1837 sample_function
->name
= "unsupportedCubeShadow";
1838 sample_function
->coord_mask
= 0;
1844 sample_function
->name
= "textureCubeLod";
1848 if (gl_info
->supported
[EXT_GPU_SHADER4
])
1849 sample_function
->name
= "textureCubeGrad";
1850 else if (gl_info
->supported
[ARB_SHADER_TEXTURE_LOD
])
1851 sample_function
->name
= "textureCubeGradARB";
1854 FIXME("Unsupported Cube grad function.\n");
1855 sample_function
->name
= "unsupportedCubeGrad";
1860 sample_function
->name
= "textureCube";
1862 sample_function
->coord_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
1866 case WINED3DSTT_VOLUME
:
1869 FIXME("Unsupported 3D shadow function.\n");
1870 sample_function
->name
= "unsupported3DShadow";
1871 sample_function
->coord_mask
= 0;
1877 sample_function
->name
= projected
? "texture3DProjLod" : "texture3DLod";
1881 if (gl_info
->supported
[EXT_GPU_SHADER4
])
1882 sample_function
->name
= projected
? "texture3DProjGrad" : "texture3DGrad";
1883 else if (gl_info
->supported
[ARB_SHADER_TEXTURE_LOD
])
1884 sample_function
->name
= projected
? "texture3DProjGradARB" : "texture3DGradARB";
1887 FIXME("Unsupported 3D grad function.\n");
1888 sample_function
->name
= "unsupported3DGrad";
1893 sample_function
->name
= projected
? "texture3DProj" : "texture3D";
1895 sample_function
->coord_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
1900 sample_function
->name
= "";
1901 sample_function
->coord_mask
= 0;
1902 FIXME("Unrecognized sampler type: %#x;\n", sampler_type
);
1907 static void shader_glsl_append_fixup_arg(char *arguments
, const char *reg_name
,
1908 BOOL sign_fixup
, enum fixup_channel_source channel_source
)
1910 switch(channel_source
)
1912 case CHANNEL_SOURCE_ZERO
:
1913 strcat(arguments
, "0.0");
1916 case CHANNEL_SOURCE_ONE
:
1917 strcat(arguments
, "1.0");
1920 case CHANNEL_SOURCE_X
:
1921 strcat(arguments
, reg_name
);
1922 strcat(arguments
, ".x");
1925 case CHANNEL_SOURCE_Y
:
1926 strcat(arguments
, reg_name
);
1927 strcat(arguments
, ".y");
1930 case CHANNEL_SOURCE_Z
:
1931 strcat(arguments
, reg_name
);
1932 strcat(arguments
, ".z");
1935 case CHANNEL_SOURCE_W
:
1936 strcat(arguments
, reg_name
);
1937 strcat(arguments
, ".w");
1941 FIXME("Unhandled channel source %#x\n", channel_source
);
1942 strcat(arguments
, "undefined");
1946 if (sign_fixup
) strcat(arguments
, " * 2.0 - 1.0");
1949 static void shader_glsl_color_correction(const struct wined3d_shader_instruction
*ins
, struct color_fixup_desc fixup
)
1951 struct wined3d_shader_dst_param dst
;
1952 unsigned int mask_size
, remaining
;
1953 glsl_dst_param_t dst_param
;
1954 char arguments
[256];
1958 if (fixup
.x_sign_fixup
|| fixup
.x_source
!= CHANNEL_SOURCE_X
) mask
|= WINED3DSP_WRITEMASK_0
;
1959 if (fixup
.y_sign_fixup
|| fixup
.y_source
!= CHANNEL_SOURCE_Y
) mask
|= WINED3DSP_WRITEMASK_1
;
1960 if (fixup
.z_sign_fixup
|| fixup
.z_source
!= CHANNEL_SOURCE_Z
) mask
|= WINED3DSP_WRITEMASK_2
;
1961 if (fixup
.w_sign_fixup
|| fixup
.w_source
!= CHANNEL_SOURCE_W
) mask
|= WINED3DSP_WRITEMASK_3
;
1962 mask
&= ins
->dst
[0].write_mask
;
1964 if (!mask
) return; /* Nothing to do */
1966 if (is_complex_fixup(fixup
))
1968 enum complex_fixup complex_fixup
= get_complex_fixup(fixup
);
1969 FIXME("Complex fixup (%#x) not supported\n",complex_fixup
);
1973 mask_size
= shader_glsl_get_write_mask_size(mask
);
1976 dst
.write_mask
= mask
;
1977 shader_glsl_add_dst_param(ins
, &dst
, &dst_param
);
1979 arguments
[0] = '\0';
1980 remaining
= mask_size
;
1981 if (mask
& WINED3DSP_WRITEMASK_0
)
1983 shader_glsl_append_fixup_arg(arguments
, dst_param
.reg_name
, fixup
.x_sign_fixup
, fixup
.x_source
);
1984 if (--remaining
) strcat(arguments
, ", ");
1986 if (mask
& WINED3DSP_WRITEMASK_1
)
1988 shader_glsl_append_fixup_arg(arguments
, dst_param
.reg_name
, fixup
.y_sign_fixup
, fixup
.y_source
);
1989 if (--remaining
) strcat(arguments
, ", ");
1991 if (mask
& WINED3DSP_WRITEMASK_2
)
1993 shader_glsl_append_fixup_arg(arguments
, dst_param
.reg_name
, fixup
.z_sign_fixup
, fixup
.z_source
);
1994 if (--remaining
) strcat(arguments
, ", ");
1996 if (mask
& WINED3DSP_WRITEMASK_3
)
1998 shader_glsl_append_fixup_arg(arguments
, dst_param
.reg_name
, fixup
.w_sign_fixup
, fixup
.w_source
);
1999 if (--remaining
) strcat(arguments
, ", ");
2004 shader_addline(ins
->ctx
->buffer
, "%s%s = vec%u(%s);\n",
2005 dst_param
.reg_name
, dst_param
.mask_str
, mask_size
, arguments
);
2009 shader_addline(ins
->ctx
->buffer
, "%s%s = %s;\n", dst_param
.reg_name
, dst_param
.mask_str
, arguments
);
2013 static void PRINTF_ATTR(8, 9) shader_glsl_gen_sample_code(const struct wined3d_shader_instruction
*ins
,
2014 DWORD sampler
, const glsl_sample_function_t
*sample_function
, DWORD swizzle
,
2015 const char *dx
, const char *dy
,
2016 const char *bias
, const char *coord_reg_fmt
, ...)
2018 const char *sampler_base
;
2019 char dst_swizzle
[6];
2020 struct color_fixup_desc fixup
;
2021 BOOL np2_fixup
= FALSE
;
2024 shader_glsl_swizzle_to_str(swizzle
, FALSE
, ins
->dst
[0].write_mask
, dst_swizzle
);
2026 if (shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
.type
))
2028 const struct shader_glsl_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2029 fixup
= priv
->cur_ps_args
->color_fixup
[sampler
];
2030 sampler_base
= "Psampler";
2032 if(priv
->cur_ps_args
->np2_fixup
& (1 << sampler
)) {
2034 FIXME("Biased sampling from NP2 textures is unsupported\n");
2040 sampler_base
= "Vsampler";
2041 fixup
= COLOR_FIXUP_IDENTITY
; /* FIXME: Vshader color fixup */
2044 shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2046 shader_addline(ins
->ctx
->buffer
, "%s(%s%u, ", sample_function
->name
, sampler_base
, sampler
);
2048 va_start(args
, coord_reg_fmt
);
2049 shader_vaddline(ins
->ctx
->buffer
, coord_reg_fmt
, args
);
2053 shader_addline(ins
->ctx
->buffer
, ", %s)%s);\n", bias
, dst_swizzle
);
2056 const struct shader_glsl_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2057 const unsigned char idx
= priv
->cur_np2fixup_info
->idx
[sampler
];
2059 shader_addline(ins
->ctx
->buffer
, " * PsamplerNP2Fixup[%u].%s)%s);\n", idx
>> 1,
2060 (idx
% 2) ? "zw" : "xy", dst_swizzle
);
2061 } else if(dx
&& dy
) {
2062 shader_addline(ins
->ctx
->buffer
, ", %s, %s)%s);\n", dx
, dy
, dst_swizzle
);
2064 shader_addline(ins
->ctx
->buffer
, ")%s);\n", dst_swizzle
);
2068 if(!is_identity_fixup(fixup
)) {
2069 shader_glsl_color_correction(ins
, fixup
);
2073 /*****************************************************************************
2074 * Begin processing individual instruction opcodes
2075 ****************************************************************************/
2077 /* Generate GLSL arithmetic functions (dst = src1 + src2) */
2078 static void shader_glsl_arith(const struct wined3d_shader_instruction
*ins
)
2080 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2081 glsl_src_param_t src0_param
;
2082 glsl_src_param_t src1_param
;
2086 /* Determine the GLSL operator to use based on the opcode */
2087 switch (ins
->handler_idx
)
2089 case WINED3DSIH_MUL
: op
= '*'; break;
2090 case WINED3DSIH_ADD
: op
= '+'; break;
2091 case WINED3DSIH_SUB
: op
= '-'; break;
2094 FIXME("Opcode %#x not yet handled in GLSL\n", ins
->handler_idx
);
2098 write_mask
= shader_glsl_append_dst(buffer
, ins
);
2099 shader_glsl_add_src_param(ins
, &ins
->src
[0], write_mask
, &src0_param
);
2100 shader_glsl_add_src_param(ins
, &ins
->src
[1], write_mask
, &src1_param
);
2101 shader_addline(buffer
, "%s %c %s);\n", src0_param
.param_str
, op
, src1_param
.param_str
);
2104 /* Process the WINED3DSIO_MOV opcode using GLSL (dst = src) */
2105 static void shader_glsl_mov(const struct wined3d_shader_instruction
*ins
)
2107 const struct wined3d_gl_info
*gl_info
= ins
->ctx
->gl_info
;
2108 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2109 glsl_src_param_t src0_param
;
2112 write_mask
= shader_glsl_append_dst(buffer
, ins
);
2113 shader_glsl_add_src_param(ins
, &ins
->src
[0], write_mask
, &src0_param
);
2115 /* In vs_1_1 WINED3DSIO_MOV can write to the address register. In later
2116 * shader versions WINED3DSIO_MOVA is used for this. */
2117 if (ins
->ctx
->reg_maps
->shader_version
.major
== 1
2118 && !shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
.type
)
2119 && ins
->dst
[0].reg
.type
== WINED3DSPR_ADDR
)
2121 /* This is a simple floor() */
2122 unsigned int mask_size
= shader_glsl_get_write_mask_size(write_mask
);
2123 if (mask_size
> 1) {
2124 shader_addline(buffer
, "ivec%d(floor(%s)));\n", mask_size
, src0_param
.param_str
);
2126 shader_addline(buffer
, "int(floor(%s)));\n", src0_param
.param_str
);
2129 else if(ins
->handler_idx
== WINED3DSIH_MOVA
)
2131 /* We need to *round* to the nearest int here. */
2132 unsigned int mask_size
= shader_glsl_get_write_mask_size(write_mask
);
2134 if (gl_info
->supported
[EXT_GPU_SHADER4
])
2137 shader_addline(buffer
, "ivec%d(round(%s)));\n", mask_size
, src0_param
.param_str
);
2139 shader_addline(buffer
, "int(round(%s)));\n", src0_param
.param_str
);
2144 shader_addline(buffer
, "ivec%d(floor(abs(%s) + vec%d(0.5)) * sign(%s)));\n",
2145 mask_size
, src0_param
.param_str
, mask_size
, src0_param
.param_str
);
2147 shader_addline(buffer
, "int(floor(abs(%s) + 0.5) * sign(%s)));\n",
2148 src0_param
.param_str
, src0_param
.param_str
);
2153 shader_addline(buffer
, "%s);\n", src0_param
.param_str
);
2157 /* Process the dot product operators DP3 and DP4 in GLSL (dst = dot(src0, src1)) */
2158 static void shader_glsl_dot(const struct wined3d_shader_instruction
*ins
)
2160 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2161 glsl_src_param_t src0_param
;
2162 glsl_src_param_t src1_param
;
2163 DWORD dst_write_mask
, src_write_mask
;
2164 unsigned int dst_size
= 0;
2166 dst_write_mask
= shader_glsl_append_dst(buffer
, ins
);
2167 dst_size
= shader_glsl_get_write_mask_size(dst_write_mask
);
2169 /* dp3 works on vec3, dp4 on vec4 */
2170 if (ins
->handler_idx
== WINED3DSIH_DP4
)
2172 src_write_mask
= WINED3DSP_WRITEMASK_ALL
;
2174 src_write_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2177 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_write_mask
, &src0_param
);
2178 shader_glsl_add_src_param(ins
, &ins
->src
[1], src_write_mask
, &src1_param
);
2181 shader_addline(buffer
, "vec%d(dot(%s, %s)));\n", dst_size
, src0_param
.param_str
, src1_param
.param_str
);
2183 shader_addline(buffer
, "dot(%s, %s));\n", src0_param
.param_str
, src1_param
.param_str
);
2187 /* Note that this instruction has some restrictions. The destination write mask
2188 * can't contain the w component, and the source swizzles have to be .xyzw */
2189 static void shader_glsl_cross(const struct wined3d_shader_instruction
*ins
)
2191 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2192 glsl_src_param_t src0_param
;
2193 glsl_src_param_t src1_param
;
2196 shader_glsl_get_write_mask(&ins
->dst
[0], dst_mask
);
2197 shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2198 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
2199 shader_glsl_add_src_param(ins
, &ins
->src
[1], src_mask
, &src1_param
);
2200 shader_addline(ins
->ctx
->buffer
, "cross(%s, %s)%s);\n", src0_param
.param_str
, src1_param
.param_str
, dst_mask
);
2203 /* Process the WINED3DSIO_POW instruction in GLSL (dst = |src0|^src1)
2204 * Src0 and src1 are scalars. Note that D3D uses the absolute of src0, while
2205 * GLSL uses the value as-is. */
2206 static void shader_glsl_pow(const struct wined3d_shader_instruction
*ins
)
2208 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2209 glsl_src_param_t src0_param
;
2210 glsl_src_param_t src1_param
;
2211 DWORD dst_write_mask
;
2212 unsigned int dst_size
;
2214 dst_write_mask
= shader_glsl_append_dst(buffer
, ins
);
2215 dst_size
= shader_glsl_get_write_mask_size(dst_write_mask
);
2217 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2218 shader_glsl_add_src_param(ins
, &ins
->src
[1], WINED3DSP_WRITEMASK_0
, &src1_param
);
2222 shader_addline(buffer
, "vec%u(pow(abs(%s), %s)));\n",
2223 dst_size
, src0_param
.param_str
, src1_param
.param_str
);
2227 shader_addline(buffer
, "pow(abs(%s), %s));\n",
2228 src0_param
.param_str
, src1_param
.param_str
);
2232 /* Process the WINED3DSIO_LOG instruction in GLSL (dst = log2(|src0|))
2233 * Src0 is a scalar. Note that D3D uses the absolute of src0, while
2234 * GLSL uses the value as-is. */
2235 static void shader_glsl_log(const struct wined3d_shader_instruction
*ins
)
2237 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2238 glsl_src_param_t src0_param
;
2239 DWORD dst_write_mask
;
2240 unsigned int dst_size
;
2242 dst_write_mask
= shader_glsl_append_dst(buffer
, ins
);
2243 dst_size
= shader_glsl_get_write_mask_size(dst_write_mask
);
2245 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2249 shader_addline(buffer
, "vec%u(log2(abs(%s))));\n",
2250 dst_size
, src0_param
.param_str
);
2254 shader_addline(buffer
, "log2(abs(%s)));\n",
2255 src0_param
.param_str
);
2259 /* Map the opcode 1-to-1 to the GL code (arg->dst = instruction(src0, src1, ...) */
2260 static void shader_glsl_map2gl(const struct wined3d_shader_instruction
*ins
)
2262 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2263 glsl_src_param_t src_param
;
2264 const char *instruction
;
2268 /* Determine the GLSL function to use based on the opcode */
2269 /* TODO: Possibly make this a table for faster lookups */
2270 switch (ins
->handler_idx
)
2272 case WINED3DSIH_MIN
: instruction
= "min"; break;
2273 case WINED3DSIH_MAX
: instruction
= "max"; break;
2274 case WINED3DSIH_ABS
: instruction
= "abs"; break;
2275 case WINED3DSIH_FRC
: instruction
= "fract"; break;
2276 case WINED3DSIH_EXP
: instruction
= "exp2"; break;
2277 case WINED3DSIH_DSX
: instruction
= "dFdx"; break;
2278 case WINED3DSIH_DSY
: instruction
= "ycorrection.y * dFdy"; break;
2279 default: instruction
= "";
2280 FIXME("Opcode %#x not yet handled in GLSL\n", ins
->handler_idx
);
2284 write_mask
= shader_glsl_append_dst(buffer
, ins
);
2286 shader_addline(buffer
, "%s(", instruction
);
2290 shader_glsl_add_src_param(ins
, &ins
->src
[0], write_mask
, &src_param
);
2291 shader_addline(buffer
, "%s", src_param
.param_str
);
2292 for (i
= 1; i
< ins
->src_count
; ++i
)
2294 shader_glsl_add_src_param(ins
, &ins
->src
[i
], write_mask
, &src_param
);
2295 shader_addline(buffer
, ", %s", src_param
.param_str
);
2299 shader_addline(buffer
, "));\n");
2302 static void shader_glsl_nrm(const struct wined3d_shader_instruction
*ins
)
2304 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2305 glsl_src_param_t src_param
;
2306 unsigned int mask_size
;
2310 write_mask
= shader_glsl_get_write_mask(ins
->dst
, dst_mask
);
2311 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
2312 shader_glsl_add_src_param(ins
, &ins
->src
[0], write_mask
, &src_param
);
2314 shader_addline(buffer
, "tmp0.x = length(%s);\n", src_param
.param_str
);
2315 shader_glsl_append_dst(buffer
, ins
);
2318 shader_addline(buffer
, "tmp0.x == 0.0 ? vec%u(0.0) : (%s / tmp0.x));\n",
2319 mask_size
, src_param
.param_str
);
2323 shader_addline(buffer
, "tmp0.x == 0.0 ? 0.0 : (%s / tmp0.x));\n",
2324 src_param
.param_str
);
2328 /** Process the WINED3DSIO_EXPP instruction in GLSL:
2329 * For shader model 1.x, do the following (and honor the writemask, so use a temporary variable):
2330 * dst.x = 2^(floor(src))
2331 * dst.y = src - floor(src)
2332 * dst.z = 2^src (partial precision is allowed, but optional)
2334 * For 2.0 shaders, just do this (honoring writemask and swizzle):
2335 * dst = 2^src; (partial precision is allowed, but optional)
2337 static void shader_glsl_expp(const struct wined3d_shader_instruction
*ins
)
2339 glsl_src_param_t src_param
;
2341 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
, &src_param
);
2343 if (ins
->ctx
->reg_maps
->shader_version
.major
< 2)
2347 shader_addline(ins
->ctx
->buffer
, "tmp0.x = exp2(floor(%s));\n", src_param
.param_str
);
2348 shader_addline(ins
->ctx
->buffer
, "tmp0.y = %s - floor(%s);\n", src_param
.param_str
, src_param
.param_str
);
2349 shader_addline(ins
->ctx
->buffer
, "tmp0.z = exp2(%s);\n", src_param
.param_str
);
2350 shader_addline(ins
->ctx
->buffer
, "tmp0.w = 1.0;\n");
2352 shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2353 shader_glsl_get_write_mask(&ins
->dst
[0], dst_mask
);
2354 shader_addline(ins
->ctx
->buffer
, "tmp0%s);\n", dst_mask
);
2357 unsigned int mask_size
;
2359 write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2360 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
2362 if (mask_size
> 1) {
2363 shader_addline(ins
->ctx
->buffer
, "vec%d(exp2(%s)));\n", mask_size
, src_param
.param_str
);
2365 shader_addline(ins
->ctx
->buffer
, "exp2(%s));\n", src_param
.param_str
);
2370 /** Process the RCP (reciprocal or inverse) opcode in GLSL (dst = 1 / src) */
2371 static void shader_glsl_rcp(const struct wined3d_shader_instruction
*ins
)
2373 glsl_src_param_t src_param
;
2375 unsigned int mask_size
;
2377 write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2378 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
2379 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_3
, &src_param
);
2383 shader_addline(ins
->ctx
->buffer
, "vec%u(1.0 / %s));\n",
2384 mask_size
, src_param
.param_str
);
2388 shader_addline(ins
->ctx
->buffer
, "1.0 / %s);\n",
2389 src_param
.param_str
);
2393 static void shader_glsl_rsq(const struct wined3d_shader_instruction
*ins
)
2395 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2396 glsl_src_param_t src_param
;
2398 unsigned int mask_size
;
2400 write_mask
= shader_glsl_append_dst(buffer
, ins
);
2401 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
2403 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_3
, &src_param
);
2407 shader_addline(buffer
, "vec%u(inversesqrt(abs(%s))));\n",
2408 mask_size
, src_param
.param_str
);
2412 shader_addline(buffer
, "inversesqrt(abs(%s)));\n",
2413 src_param
.param_str
);
2417 /** Process signed comparison opcodes in GLSL. */
2418 static void shader_glsl_compare(const struct wined3d_shader_instruction
*ins
)
2420 glsl_src_param_t src0_param
;
2421 glsl_src_param_t src1_param
;
2423 unsigned int mask_size
;
2425 write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2426 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
2427 shader_glsl_add_src_param(ins
, &ins
->src
[0], write_mask
, &src0_param
);
2428 shader_glsl_add_src_param(ins
, &ins
->src
[1], write_mask
, &src1_param
);
2430 if (mask_size
> 1) {
2431 const char *compare
;
2433 switch(ins
->handler_idx
)
2435 case WINED3DSIH_SLT
: compare
= "lessThan"; break;
2436 case WINED3DSIH_SGE
: compare
= "greaterThanEqual"; break;
2437 default: compare
= "";
2438 FIXME("Can't handle opcode %#x\n", ins
->handler_idx
);
2441 shader_addline(ins
->ctx
->buffer
, "vec%d(%s(%s, %s)));\n", mask_size
, compare
,
2442 src0_param
.param_str
, src1_param
.param_str
);
2444 switch(ins
->handler_idx
)
2446 case WINED3DSIH_SLT
:
2447 /* Step(src0, src1) is not suitable here because if src0 == src1 SLT is supposed,
2448 * to return 0.0 but step returns 1.0 because step is not < x
2449 * An alternative is a bvec compare padded with an unused second component.
2450 * step(src1 * -1.0, src0 * -1.0) is not an option because it suffers from the same
2451 * issue. Playing with not() is not possible either because not() does not accept
2454 shader_addline(ins
->ctx
->buffer
, "(%s < %s) ? 1.0 : 0.0);\n",
2455 src0_param
.param_str
, src1_param
.param_str
);
2457 case WINED3DSIH_SGE
:
2458 /* Here we can use the step() function and safe a conditional */
2459 shader_addline(ins
->ctx
->buffer
, "step(%s, %s));\n", src1_param
.param_str
, src0_param
.param_str
);
2462 FIXME("Can't handle opcode %#x\n", ins
->handler_idx
);
2468 /** Process CMP instruction in GLSL (dst = src0 >= 0.0 ? src1 : src2), per channel */
2469 static void shader_glsl_cmp(const struct wined3d_shader_instruction
*ins
)
2471 glsl_src_param_t src0_param
;
2472 glsl_src_param_t src1_param
;
2473 glsl_src_param_t src2_param
;
2474 DWORD write_mask
, cmp_channel
= 0;
2477 BOOL temp_destination
= FALSE
;
2479 if (shader_is_scalar(&ins
->src
[0].reg
))
2481 write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2483 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_ALL
, &src0_param
);
2484 shader_glsl_add_src_param(ins
, &ins
->src
[1], write_mask
, &src1_param
);
2485 shader_glsl_add_src_param(ins
, &ins
->src
[2], write_mask
, &src2_param
);
2487 shader_addline(ins
->ctx
->buffer
, "%s >= 0.0 ? %s : %s);\n",
2488 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
2490 DWORD dst_mask
= ins
->dst
[0].write_mask
;
2491 struct wined3d_shader_dst_param dst
= ins
->dst
[0];
2493 /* Cycle through all source0 channels */
2494 for (i
=0; i
<4; i
++) {
2496 /* Find the destination channels which use the current source0 channel */
2497 for (j
=0; j
<4; j
++) {
2498 if (((ins
->src
[0].swizzle
>> (2 * j
)) & 0x3) == i
)
2500 write_mask
|= WINED3DSP_WRITEMASK_0
<< j
;
2501 cmp_channel
= WINED3DSP_WRITEMASK_0
<< j
;
2504 dst
.write_mask
= dst_mask
& write_mask
;
2506 /* Splitting the cmp instruction up in multiple lines imposes a problem:
2507 * The first lines may overwrite source parameters of the following lines.
2508 * Deal with that by using a temporary destination register if needed
2510 if ((ins
->src
[0].reg
.idx
== ins
->dst
[0].reg
.idx
2511 && ins
->src
[0].reg
.type
== ins
->dst
[0].reg
.type
)
2512 || (ins
->src
[1].reg
.idx
== ins
->dst
[0].reg
.idx
2513 && ins
->src
[1].reg
.type
== ins
->dst
[0].reg
.type
)
2514 || (ins
->src
[2].reg
.idx
== ins
->dst
[0].reg
.idx
2515 && ins
->src
[2].reg
.type
== ins
->dst
[0].reg
.type
))
2517 write_mask
= shader_glsl_get_write_mask(&dst
, mask_char
);
2518 if (!write_mask
) continue;
2519 shader_addline(ins
->ctx
->buffer
, "tmp0%s = (", mask_char
);
2520 temp_destination
= TRUE
;
2522 write_mask
= shader_glsl_append_dst_ext(ins
->ctx
->buffer
, ins
, &dst
);
2523 if (!write_mask
) continue;
2526 shader_glsl_add_src_param(ins
, &ins
->src
[0], cmp_channel
, &src0_param
);
2527 shader_glsl_add_src_param(ins
, &ins
->src
[1], write_mask
, &src1_param
);
2528 shader_glsl_add_src_param(ins
, &ins
->src
[2], write_mask
, &src2_param
);
2530 shader_addline(ins
->ctx
->buffer
, "%s >= 0.0 ? %s : %s);\n",
2531 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
2534 if(temp_destination
) {
2535 shader_glsl_get_write_mask(&ins
->dst
[0], mask_char
);
2536 shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2537 shader_addline(ins
->ctx
->buffer
, "tmp0%s);\n", mask_char
);
2543 /** Process the CND opcode in GLSL (dst = (src0 > 0.5) ? src1 : src2) */
2544 /* For ps 1.1-1.3, only a single component of src0 is used. For ps 1.4
2545 * the compare is done per component of src0. */
2546 static void shader_glsl_cnd(const struct wined3d_shader_instruction
*ins
)
2548 struct wined3d_shader_dst_param dst
;
2549 glsl_src_param_t src0_param
;
2550 glsl_src_param_t src1_param
;
2551 glsl_src_param_t src2_param
;
2552 DWORD write_mask
, cmp_channel
= 0;
2555 DWORD shader_version
= WINED3D_SHADER_VERSION(ins
->ctx
->reg_maps
->shader_version
.major
,
2556 ins
->ctx
->reg_maps
->shader_version
.minor
);
2558 if (shader_version
< WINED3D_SHADER_VERSION(1, 4))
2560 write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2561 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2562 shader_glsl_add_src_param(ins
, &ins
->src
[1], write_mask
, &src1_param
);
2563 shader_glsl_add_src_param(ins
, &ins
->src
[2], write_mask
, &src2_param
);
2565 /* Fun: The D3DSI_COISSUE flag changes the semantic of the cnd instruction for < 1.4 shaders */
2568 shader_addline(ins
->ctx
->buffer
, "%s /* COISSUE! */);\n", src1_param
.param_str
);
2570 shader_addline(ins
->ctx
->buffer
, "%s > 0.5 ? %s : %s);\n",
2571 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
2575 /* Cycle through all source0 channels */
2576 dst_mask
= ins
->dst
[0].write_mask
;
2578 for (i
=0; i
<4; i
++) {
2580 /* Find the destination channels which use the current source0 channel */
2581 for (j
=0; j
<4; j
++) {
2582 if (((ins
->src
[0].swizzle
>> (2 * j
)) & 0x3) == i
)
2584 write_mask
|= WINED3DSP_WRITEMASK_0
<< j
;
2585 cmp_channel
= WINED3DSP_WRITEMASK_0
<< j
;
2589 dst
.write_mask
= dst_mask
& write_mask
;
2590 write_mask
= shader_glsl_append_dst_ext(ins
->ctx
->buffer
, ins
, &dst
);
2591 if (!write_mask
) continue;
2593 shader_glsl_add_src_param(ins
, &ins
->src
[0], cmp_channel
, &src0_param
);
2594 shader_glsl_add_src_param(ins
, &ins
->src
[1], write_mask
, &src1_param
);
2595 shader_glsl_add_src_param(ins
, &ins
->src
[2], write_mask
, &src2_param
);
2597 shader_addline(ins
->ctx
->buffer
, "%s > 0.5 ? %s : %s);\n",
2598 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
2602 /** GLSL code generation for WINED3DSIO_MAD: Multiply the first 2 opcodes, then add the last */
2603 static void shader_glsl_mad(const struct wined3d_shader_instruction
*ins
)
2605 glsl_src_param_t src0_param
;
2606 glsl_src_param_t src1_param
;
2607 glsl_src_param_t src2_param
;
2610 write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2611 shader_glsl_add_src_param(ins
, &ins
->src
[0], write_mask
, &src0_param
);
2612 shader_glsl_add_src_param(ins
, &ins
->src
[1], write_mask
, &src1_param
);
2613 shader_glsl_add_src_param(ins
, &ins
->src
[2], write_mask
, &src2_param
);
2614 shader_addline(ins
->ctx
->buffer
, "(%s * %s) + %s);\n",
2615 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
2618 /* Handles transforming all WINED3DSIO_M?x? opcodes for
2619 Vertex shaders to GLSL codes */
2620 static void shader_glsl_mnxn(const struct wined3d_shader_instruction
*ins
)
2623 int nComponents
= 0;
2624 struct wined3d_shader_dst_param tmp_dst
= {{0}};
2625 struct wined3d_shader_src_param tmp_src
[2] = {{{0}}};
2626 struct wined3d_shader_instruction tmp_ins
;
2628 memset(&tmp_ins
, 0, sizeof(tmp_ins
));
2630 /* Set constants for the temporary argument */
2631 tmp_ins
.ctx
= ins
->ctx
;
2632 tmp_ins
.dst_count
= 1;
2633 tmp_ins
.dst
= &tmp_dst
;
2634 tmp_ins
.src_count
= 2;
2635 tmp_ins
.src
= tmp_src
;
2637 switch(ins
->handler_idx
)
2639 case WINED3DSIH_M4x4
:
2641 tmp_ins
.handler_idx
= WINED3DSIH_DP4
;
2643 case WINED3DSIH_M4x3
:
2645 tmp_ins
.handler_idx
= WINED3DSIH_DP4
;
2647 case WINED3DSIH_M3x4
:
2649 tmp_ins
.handler_idx
= WINED3DSIH_DP3
;
2651 case WINED3DSIH_M3x3
:
2653 tmp_ins
.handler_idx
= WINED3DSIH_DP3
;
2655 case WINED3DSIH_M3x2
:
2657 tmp_ins
.handler_idx
= WINED3DSIH_DP3
;
2663 tmp_dst
= ins
->dst
[0];
2664 tmp_src
[0] = ins
->src
[0];
2665 tmp_src
[1] = ins
->src
[1];
2666 for (i
= 0; i
< nComponents
; ++i
)
2668 tmp_dst
.write_mask
= WINED3DSP_WRITEMASK_0
<< i
;
2669 shader_glsl_dot(&tmp_ins
);
2670 ++tmp_src
[1].reg
.idx
;
2675 The LRP instruction performs a component-wise linear interpolation
2676 between the second and third operands using the first operand as the
2677 blend factor. Equation: (dst = src2 + src0 * (src1 - src2))
2678 This is equivalent to mix(src2, src1, src0);
2680 static void shader_glsl_lrp(const struct wined3d_shader_instruction
*ins
)
2682 glsl_src_param_t src0_param
;
2683 glsl_src_param_t src1_param
;
2684 glsl_src_param_t src2_param
;
2687 write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2689 shader_glsl_add_src_param(ins
, &ins
->src
[0], write_mask
, &src0_param
);
2690 shader_glsl_add_src_param(ins
, &ins
->src
[1], write_mask
, &src1_param
);
2691 shader_glsl_add_src_param(ins
, &ins
->src
[2], write_mask
, &src2_param
);
2693 shader_addline(ins
->ctx
->buffer
, "mix(%s, %s, %s));\n",
2694 src2_param
.param_str
, src1_param
.param_str
, src0_param
.param_str
);
2697 /** Process the WINED3DSIO_LIT instruction in GLSL:
2698 * dst.x = dst.w = 1.0
2699 * dst.y = (src0.x > 0) ? src0.x
2700 * dst.z = (src0.x > 0) ? ((src0.y > 0) ? pow(src0.y, src.w) : 0) : 0
2701 * where src.w is clamped at +- 128
2703 static void shader_glsl_lit(const struct wined3d_shader_instruction
*ins
)
2705 glsl_src_param_t src0_param
;
2706 glsl_src_param_t src1_param
;
2707 glsl_src_param_t src3_param
;
2710 shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2711 shader_glsl_get_write_mask(&ins
->dst
[0], dst_mask
);
2713 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2714 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_1
, &src1_param
);
2715 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_3
, &src3_param
);
2717 /* The sdk specifies the instruction like this
2719 * if(src.x > 0.0) dst.y = src.x
2721 * if(src.x > 0.0 && src.y > 0.0) dst.z = pow(src.y, power);
2725 * Obviously that has quite a few conditionals in it which we don't like. So the first step is this:
2726 * dst.x = 1.0 ... No further explanation needed
2727 * dst.y = max(src.y, 0.0); ... If x < 0.0, use 0.0, otherwise x. Same as the conditional
2728 * dst.z = x > 0.0 ? pow(max(y, 0.0), p) : 0; ... 0 ^ power is 0, and otherwise we use y anyway
2729 * dst.w = 1.0. ... Nothing fancy.
2731 * So we still have one conditional in there. So do this:
2732 * dst.z = pow(max(0.0, src.y) * step(0.0, src.x), power);
2734 * 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),
2735 * 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.
2736 * if both x and y are > 0, we get pow(y * 1.0, power), as it is supposed to
2738 shader_addline(ins
->ctx
->buffer
,
2739 "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",
2740 src0_param
.param_str
, src1_param
.param_str
, src0_param
.param_str
, src3_param
.param_str
, dst_mask
);
2743 /** Process the WINED3DSIO_DST instruction in GLSL:
2745 * dst.y = src0.x * src0.y
2749 static void shader_glsl_dst(const struct wined3d_shader_instruction
*ins
)
2751 glsl_src_param_t src0y_param
;
2752 glsl_src_param_t src0z_param
;
2753 glsl_src_param_t src1y_param
;
2754 glsl_src_param_t src1w_param
;
2757 shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2758 shader_glsl_get_write_mask(&ins
->dst
[0], dst_mask
);
2760 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_1
, &src0y_param
);
2761 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_2
, &src0z_param
);
2762 shader_glsl_add_src_param(ins
, &ins
->src
[1], WINED3DSP_WRITEMASK_1
, &src1y_param
);
2763 shader_glsl_add_src_param(ins
, &ins
->src
[1], WINED3DSP_WRITEMASK_3
, &src1w_param
);
2765 shader_addline(ins
->ctx
->buffer
, "vec4(1.0, %s * %s, %s, %s))%s;\n",
2766 src0y_param
.param_str
, src1y_param
.param_str
, src0z_param
.param_str
, src1w_param
.param_str
, dst_mask
);
2769 /** Process the WINED3DSIO_SINCOS instruction in GLSL:
2770 * VS 2.0 requires that specific cosine and sine constants be passed to this instruction so the hardware
2771 * can handle it. But, these functions are built-in for GLSL, so we can just ignore the last 2 params.
2773 * dst.x = cos(src0.?)
2774 * dst.y = sin(src0.?)
2778 static void shader_glsl_sincos(const struct wined3d_shader_instruction
*ins
)
2780 glsl_src_param_t src0_param
;
2783 write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2784 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2786 switch (write_mask
) {
2787 case WINED3DSP_WRITEMASK_0
:
2788 shader_addline(ins
->ctx
->buffer
, "cos(%s));\n", src0_param
.param_str
);
2791 case WINED3DSP_WRITEMASK_1
:
2792 shader_addline(ins
->ctx
->buffer
, "sin(%s));\n", src0_param
.param_str
);
2795 case (WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
):
2796 shader_addline(ins
->ctx
->buffer
, "vec2(cos(%s), sin(%s)));\n", src0_param
.param_str
, src0_param
.param_str
);
2800 ERR("Write mask should be .x, .y or .xy\n");
2805 /* sgn in vs_2_0 has 2 extra parameters(registers for temporary storage) which we don't use
2806 * here. But those extra parameters require a dedicated function for sgn, since map2gl would
2807 * generate invalid code
2809 static void shader_glsl_sgn(const struct wined3d_shader_instruction
*ins
)
2811 glsl_src_param_t src0_param
;
2814 write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2815 shader_glsl_add_src_param(ins
, &ins
->src
[0], write_mask
, &src0_param
);
2817 shader_addline(ins
->ctx
->buffer
, "sign(%s));\n", src0_param
.param_str
);
2820 /** Process the WINED3DSIO_LOOP instruction in GLSL:
2821 * Start a for() loop where src1.y is the initial value of aL,
2822 * increment aL by src1.z for a total of src1.x iterations.
2823 * Need to use a temporary variable for this operation.
2825 /* FIXME: I don't think nested loops will work correctly this way. */
2826 static void shader_glsl_loop(const struct wined3d_shader_instruction
*ins
)
2828 glsl_src_param_t src1_param
;
2829 IWineD3DBaseShaderImpl
*shader
= (IWineD3DBaseShaderImpl
*)ins
->ctx
->shader
;
2830 const DWORD
*control_values
= NULL
;
2831 const local_constant
*constant
;
2833 shader_glsl_add_src_param(ins
, &ins
->src
[1], WINED3DSP_WRITEMASK_ALL
, &src1_param
);
2835 /* Try to hardcode the loop control parameters if possible. Direct3D 9 class hardware doesn't support real
2836 * varying indexing, but Microsoft designed this feature for Shader model 2.x+. If the loop control is
2837 * known at compile time, the GLSL compiler can unroll the loop, and replace indirect addressing with direct
2840 if (ins
->src
[1].reg
.type
== WINED3DSPR_CONSTINT
)
2842 LIST_FOR_EACH_ENTRY(constant
, &shader
->baseShader
.constantsI
, local_constant
, entry
) {
2843 if (constant
->idx
== ins
->src
[1].reg
.idx
)
2845 control_values
= constant
->value
;
2853 struct wined3d_shader_loop_control loop_control
;
2854 loop_control
.count
= control_values
[0];
2855 loop_control
.start
= control_values
[1];
2856 loop_control
.step
= (int)control_values
[2];
2858 if (loop_control
.step
> 0)
2860 shader_addline(ins
->ctx
->buffer
, "for (aL%u = %u; aL%u < (%u * %d + %u); aL%u += %d) {\n",
2861 shader
->baseShader
.cur_loop_depth
, loop_control
.start
,
2862 shader
->baseShader
.cur_loop_depth
, loop_control
.count
, loop_control
.step
, loop_control
.start
,
2863 shader
->baseShader
.cur_loop_depth
, loop_control
.step
);
2865 else if (loop_control
.step
< 0)
2867 shader_addline(ins
->ctx
->buffer
, "for (aL%u = %u; aL%u > (%u * %d + %u); aL%u += %d) {\n",
2868 shader
->baseShader
.cur_loop_depth
, loop_control
.start
,
2869 shader
->baseShader
.cur_loop_depth
, loop_control
.count
, loop_control
.step
, loop_control
.start
,
2870 shader
->baseShader
.cur_loop_depth
, loop_control
.step
);
2874 shader_addline(ins
->ctx
->buffer
, "for (aL%u = %u, tmpInt%u = 0; tmpInt%u < %u; tmpInt%u++) {\n",
2875 shader
->baseShader
.cur_loop_depth
, loop_control
.start
, shader
->baseShader
.cur_loop_depth
,
2876 shader
->baseShader
.cur_loop_depth
, loop_control
.count
,
2877 shader
->baseShader
.cur_loop_depth
);
2880 shader_addline(ins
->ctx
->buffer
,
2881 "for (tmpInt%u = 0, aL%u = %s.y; tmpInt%u < %s.x; tmpInt%u++, aL%u += %s.z) {\n",
2882 shader
->baseShader
.cur_loop_depth
, shader
->baseShader
.cur_loop_regno
,
2883 src1_param
.reg_name
, shader
->baseShader
.cur_loop_depth
, src1_param
.reg_name
,
2884 shader
->baseShader
.cur_loop_depth
, shader
->baseShader
.cur_loop_regno
, src1_param
.reg_name
);
2887 shader
->baseShader
.cur_loop_depth
++;
2888 shader
->baseShader
.cur_loop_regno
++;
2891 static void shader_glsl_end(const struct wined3d_shader_instruction
*ins
)
2893 IWineD3DBaseShaderImpl
*shader
= (IWineD3DBaseShaderImpl
*)ins
->ctx
->shader
;
2895 shader_addline(ins
->ctx
->buffer
, "}\n");
2897 if (ins
->handler_idx
== WINED3DSIH_ENDLOOP
)
2899 shader
->baseShader
.cur_loop_depth
--;
2900 shader
->baseShader
.cur_loop_regno
--;
2903 if (ins
->handler_idx
== WINED3DSIH_ENDREP
)
2905 shader
->baseShader
.cur_loop_depth
--;
2909 static void shader_glsl_rep(const struct wined3d_shader_instruction
*ins
)
2911 IWineD3DBaseShaderImpl
*shader
= (IWineD3DBaseShaderImpl
*)ins
->ctx
->shader
;
2912 glsl_src_param_t src0_param
;
2913 const DWORD
*control_values
= NULL
;
2914 const local_constant
*constant
;
2916 /* Try to hardcode local values to help the GLSL compiler to unroll and optimize the loop */
2917 if (ins
->src
[0].reg
.type
== WINED3DSPR_CONSTINT
)
2919 LIST_FOR_EACH_ENTRY(constant
, &shader
->baseShader
.constantsI
, local_constant
, entry
)
2921 if (constant
->idx
== ins
->src
[0].reg
.idx
)
2923 control_values
= constant
->value
;
2929 if(control_values
) {
2930 shader_addline(ins
->ctx
->buffer
, "for (tmpInt%d = 0; tmpInt%d < %d; tmpInt%d++) {\n",
2931 shader
->baseShader
.cur_loop_depth
, shader
->baseShader
.cur_loop_depth
,
2932 control_values
[0], shader
->baseShader
.cur_loop_depth
);
2934 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2935 shader_addline(ins
->ctx
->buffer
, "for (tmpInt%d = 0; tmpInt%d < %s; tmpInt%d++) {\n",
2936 shader
->baseShader
.cur_loop_depth
, shader
->baseShader
.cur_loop_depth
,
2937 src0_param
.param_str
, shader
->baseShader
.cur_loop_depth
);
2939 shader
->baseShader
.cur_loop_depth
++;
2942 static void shader_glsl_if(const struct wined3d_shader_instruction
*ins
)
2944 glsl_src_param_t src0_param
;
2946 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2947 shader_addline(ins
->ctx
->buffer
, "if (%s) {\n", src0_param
.param_str
);
2950 static void shader_glsl_ifc(const struct wined3d_shader_instruction
*ins
)
2952 glsl_src_param_t src0_param
;
2953 glsl_src_param_t src1_param
;
2955 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2956 shader_glsl_add_src_param(ins
, &ins
->src
[1], WINED3DSP_WRITEMASK_0
, &src1_param
);
2958 shader_addline(ins
->ctx
->buffer
, "if (%s %s %s) {\n",
2959 src0_param
.param_str
, shader_get_comp_op(ins
->flags
), src1_param
.param_str
);
2962 static void shader_glsl_else(const struct wined3d_shader_instruction
*ins
)
2964 shader_addline(ins
->ctx
->buffer
, "} else {\n");
2967 static void shader_glsl_break(const struct wined3d_shader_instruction
*ins
)
2969 shader_addline(ins
->ctx
->buffer
, "break;\n");
2972 /* FIXME: According to MSDN the compare is done per component. */
2973 static void shader_glsl_breakc(const struct wined3d_shader_instruction
*ins
)
2975 glsl_src_param_t src0_param
;
2976 glsl_src_param_t src1_param
;
2978 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2979 shader_glsl_add_src_param(ins
, &ins
->src
[1], WINED3DSP_WRITEMASK_0
, &src1_param
);
2981 shader_addline(ins
->ctx
->buffer
, "if (%s %s %s) break;\n",
2982 src0_param
.param_str
, shader_get_comp_op(ins
->flags
), src1_param
.param_str
);
2985 static void shader_glsl_label(const struct wined3d_shader_instruction
*ins
)
2987 shader_addline(ins
->ctx
->buffer
, "}\n");
2988 shader_addline(ins
->ctx
->buffer
, "void subroutine%u () {\n", ins
->src
[0].reg
.idx
);
2991 static void shader_glsl_call(const struct wined3d_shader_instruction
*ins
)
2993 shader_addline(ins
->ctx
->buffer
, "subroutine%u();\n", ins
->src
[0].reg
.idx
);
2996 static void shader_glsl_callnz(const struct wined3d_shader_instruction
*ins
)
2998 glsl_src_param_t src1_param
;
3000 shader_glsl_add_src_param(ins
, &ins
->src
[1], WINED3DSP_WRITEMASK_0
, &src1_param
);
3001 shader_addline(ins
->ctx
->buffer
, "if (%s) subroutine%u();\n", src1_param
.param_str
, ins
->src
[0].reg
.idx
);
3004 static void shader_glsl_ret(const struct wined3d_shader_instruction
*ins
)
3006 /* No-op. The closing } is written when a new function is started, and at the end of the shader. This
3007 * function only suppresses the unhandled instruction warning
3011 /*********************************************
3012 * Pixel Shader Specific Code begins here
3013 ********************************************/
3014 static void shader_glsl_tex(const struct wined3d_shader_instruction
*ins
)
3016 IWineD3DBaseShaderImpl
*shader
= (IWineD3DBaseShaderImpl
*)ins
->ctx
->shader
;
3017 IWineD3DDeviceImpl
*deviceImpl
= (IWineD3DDeviceImpl
*)shader
->baseShader
.device
;
3018 DWORD shader_version
= WINED3D_SHADER_VERSION(ins
->ctx
->reg_maps
->shader_version
.major
,
3019 ins
->ctx
->reg_maps
->shader_version
.minor
);
3020 glsl_sample_function_t sample_function
;
3021 IWineD3DBaseTextureImpl
*texture
;
3022 DWORD sample_flags
= 0;
3024 DWORD mask
= 0, swizzle
;
3026 /* 1.0-1.4: Use destination register as sampler source.
3027 * 2.0+: Use provided sampler source. */
3028 if (shader_version
< WINED3D_SHADER_VERSION(2,0)) sampler_idx
= ins
->dst
[0].reg
.idx
;
3029 else sampler_idx
= ins
->src
[1].reg
.idx
;
3030 texture
= deviceImpl
->stateBlock
->state
.textures
[sampler_idx
];
3032 if (shader_version
< WINED3D_SHADER_VERSION(1,4))
3034 DWORD flags
= deviceImpl
->stateBlock
->state
.texture_states
[sampler_idx
][WINED3DTSS_TEXTURETRANSFORMFLAGS
];
3035 WINED3DSAMPLER_TEXTURE_TYPE sampler_type
= ins
->ctx
->reg_maps
->sampler_type
[sampler_idx
];
3037 /* Projected cube textures don't make a lot of sense, the resulting coordinates stay the same. */
3038 if (flags
& WINED3DTTFF_PROJECTED
&& sampler_type
!= WINED3DSTT_CUBE
) {
3039 sample_flags
|= WINED3D_GLSL_SAMPLE_PROJECTED
;
3040 switch (flags
& ~WINED3DTTFF_PROJECTED
) {
3041 case WINED3DTTFF_COUNT1
: FIXME("WINED3DTTFF_PROJECTED with WINED3DTTFF_COUNT1?\n"); break;
3042 case WINED3DTTFF_COUNT2
: mask
= WINED3DSP_WRITEMASK_1
; break;
3043 case WINED3DTTFF_COUNT3
: mask
= WINED3DSP_WRITEMASK_2
; break;
3044 case WINED3DTTFF_COUNT4
:
3045 case WINED3DTTFF_DISABLE
: mask
= WINED3DSP_WRITEMASK_3
; break;
3049 else if (shader_version
< WINED3D_SHADER_VERSION(2,0))
3051 DWORD src_mod
= ins
->src
[0].modifiers
;
3053 if (src_mod
== WINED3DSPSM_DZ
) {
3054 sample_flags
|= WINED3D_GLSL_SAMPLE_PROJECTED
;
3055 mask
= WINED3DSP_WRITEMASK_2
;
3056 } else if (src_mod
== WINED3DSPSM_DW
) {
3057 sample_flags
|= WINED3D_GLSL_SAMPLE_PROJECTED
;
3058 mask
= WINED3DSP_WRITEMASK_3
;
3061 if (ins
->flags
& WINED3DSI_TEXLD_PROJECT
)
3063 /* ps 2.0 texldp instruction always divides by the fourth component. */
3064 sample_flags
|= WINED3D_GLSL_SAMPLE_PROJECTED
;
3065 mask
= WINED3DSP_WRITEMASK_3
;
3069 if (texture
&& texture
->baseTexture
.target
== GL_TEXTURE_RECTANGLE_ARB
)
3070 sample_flags
|= WINED3D_GLSL_SAMPLE_RECT
;
3072 shader_glsl_get_sample_function(ins
->ctx
, sampler_idx
, sample_flags
, &sample_function
);
3073 mask
|= sample_function
.coord_mask
;
3075 if (shader_version
< WINED3D_SHADER_VERSION(2,0)) swizzle
= WINED3DSP_NOSWIZZLE
;
3076 else swizzle
= ins
->src
[1].swizzle
;
3078 /* 1.0-1.3: Use destination register as coordinate source.
3079 1.4+: Use provided coordinate source register. */
3080 if (shader_version
< WINED3D_SHADER_VERSION(1,4))
3083 shader_glsl_write_mask_to_str(mask
, coord_mask
);
3084 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, swizzle
, NULL
, NULL
, NULL
,
3085 "T%u%s", sampler_idx
, coord_mask
);
3087 glsl_src_param_t coord_param
;
3088 shader_glsl_add_src_param(ins
, &ins
->src
[0], mask
, &coord_param
);
3089 if (ins
->flags
& WINED3DSI_TEXLD_BIAS
)
3091 glsl_src_param_t bias
;
3092 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_3
, &bias
);
3093 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, swizzle
, NULL
, NULL
, bias
.param_str
,
3094 "%s", coord_param
.param_str
);
3096 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, swizzle
, NULL
, NULL
, NULL
,
3097 "%s", coord_param
.param_str
);
3102 static void shader_glsl_texldd(const struct wined3d_shader_instruction
*ins
)
3104 IWineD3DBaseShaderImpl
*This
= (IWineD3DBaseShaderImpl
*)ins
->ctx
->shader
;
3105 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
3106 const struct wined3d_gl_info
*gl_info
= ins
->ctx
->gl_info
;
3107 glsl_sample_function_t sample_function
;
3108 glsl_src_param_t coord_param
, dx_param
, dy_param
;
3109 DWORD sample_flags
= WINED3D_GLSL_SAMPLE_GRAD
;
3111 DWORD swizzle
= ins
->src
[1].swizzle
;
3112 IWineD3DBaseTextureImpl
*texture
;
3114 if (!gl_info
->supported
[ARB_SHADER_TEXTURE_LOD
] && !gl_info
->supported
[EXT_GPU_SHADER4
])
3116 FIXME("texldd used, but not supported by hardware. Falling back to regular tex\n");
3117 shader_glsl_tex(ins
);
3121 sampler_idx
= ins
->src
[1].reg
.idx
;
3122 texture
= deviceImpl
->stateBlock
->state
.textures
[sampler_idx
];
3123 if (texture
&& texture
->baseTexture
.target
== GL_TEXTURE_RECTANGLE_ARB
)
3124 sample_flags
|= WINED3D_GLSL_SAMPLE_RECT
;
3126 shader_glsl_get_sample_function(ins
->ctx
, sampler_idx
, sample_flags
, &sample_function
);
3127 shader_glsl_add_src_param(ins
, &ins
->src
[0], sample_function
.coord_mask
, &coord_param
);
3128 shader_glsl_add_src_param(ins
, &ins
->src
[2], sample_function
.coord_mask
, &dx_param
);
3129 shader_glsl_add_src_param(ins
, &ins
->src
[3], sample_function
.coord_mask
, &dy_param
);
3131 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, swizzle
, dx_param
.param_str
, dy_param
.param_str
, NULL
,
3132 "%s", coord_param
.param_str
);
3135 static void shader_glsl_texldl(const struct wined3d_shader_instruction
*ins
)
3137 IWineD3DBaseShaderImpl
*This
= (IWineD3DBaseShaderImpl
*)ins
->ctx
->shader
;
3138 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
3139 const struct wined3d_gl_info
*gl_info
= ins
->ctx
->gl_info
;
3140 glsl_sample_function_t sample_function
;
3141 glsl_src_param_t coord_param
, lod_param
;
3142 DWORD sample_flags
= WINED3D_GLSL_SAMPLE_LOD
;
3144 DWORD swizzle
= ins
->src
[1].swizzle
;
3145 IWineD3DBaseTextureImpl
*texture
;
3147 sampler_idx
= ins
->src
[1].reg
.idx
;
3148 texture
= deviceImpl
->stateBlock
->state
.textures
[sampler_idx
];
3149 if (texture
&& texture
->baseTexture
.target
== GL_TEXTURE_RECTANGLE_ARB
)
3150 sample_flags
|= WINED3D_GLSL_SAMPLE_RECT
;
3152 shader_glsl_get_sample_function(ins
->ctx
, sampler_idx
, sample_flags
, &sample_function
);
3153 shader_glsl_add_src_param(ins
, &ins
->src
[0], sample_function
.coord_mask
, &coord_param
);
3155 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_3
, &lod_param
);
3157 if (!gl_info
->supported
[ARB_SHADER_TEXTURE_LOD
] && !gl_info
->supported
[EXT_GPU_SHADER4
]
3158 && shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
.type
))
3160 /* The GLSL spec claims the Lod sampling functions are only supported in vertex shaders.
3161 * However, they seem to work just fine in fragment shaders as well. */
3162 WARN("Using %s in fragment shader.\n", sample_function
.name
);
3164 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, swizzle
, NULL
, NULL
, lod_param
.param_str
,
3165 "%s", coord_param
.param_str
);
3168 static void shader_glsl_texcoord(const struct wined3d_shader_instruction
*ins
)
3170 /* FIXME: Make this work for more than just 2D textures */
3171 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
3172 DWORD write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
3174 if (!(ins
->ctx
->reg_maps
->shader_version
.major
== 1 && ins
->ctx
->reg_maps
->shader_version
.minor
== 4))
3178 shader_glsl_get_write_mask(&ins
->dst
[0], dst_mask
);
3179 shader_addline(buffer
, "clamp(gl_TexCoord[%u], 0.0, 1.0)%s);\n",
3180 ins
->dst
[0].reg
.idx
, dst_mask
);
3182 DWORD reg
= ins
->src
[0].reg
.idx
;
3183 DWORD src_mod
= ins
->src
[0].modifiers
;
3184 char dst_swizzle
[6];
3186 shader_glsl_get_swizzle(&ins
->src
[0], FALSE
, write_mask
, dst_swizzle
);
3188 if (src_mod
== WINED3DSPSM_DZ
) {
3189 glsl_src_param_t div_param
;
3190 unsigned int mask_size
= shader_glsl_get_write_mask_size(write_mask
);
3191 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_2
, &div_param
);
3193 if (mask_size
> 1) {
3194 shader_addline(buffer
, "gl_TexCoord[%u]%s / vec%d(%s));\n", reg
, dst_swizzle
, mask_size
, div_param
.param_str
);
3196 shader_addline(buffer
, "gl_TexCoord[%u]%s / %s);\n", reg
, dst_swizzle
, div_param
.param_str
);
3198 } else if (src_mod
== WINED3DSPSM_DW
) {
3199 glsl_src_param_t div_param
;
3200 unsigned int mask_size
= shader_glsl_get_write_mask_size(write_mask
);
3201 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_3
, &div_param
);
3203 if (mask_size
> 1) {
3204 shader_addline(buffer
, "gl_TexCoord[%u]%s / vec%d(%s));\n", reg
, dst_swizzle
, mask_size
, div_param
.param_str
);
3206 shader_addline(buffer
, "gl_TexCoord[%u]%s / %s);\n", reg
, dst_swizzle
, div_param
.param_str
);
3209 shader_addline(buffer
, "gl_TexCoord[%u]%s);\n", reg
, dst_swizzle
);
3214 /** Process the WINED3DSIO_TEXDP3TEX instruction in GLSL:
3215 * Take a 3-component dot product of the TexCoord[dstreg] and src,
3216 * then perform a 1D texture lookup from stage dstregnum, place into dst. */
3217 static void shader_glsl_texdp3tex(const struct wined3d_shader_instruction
*ins
)
3219 glsl_src_param_t src0_param
;
3220 glsl_sample_function_t sample_function
;
3221 DWORD sampler_idx
= ins
->dst
[0].reg
.idx
;
3222 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
3225 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
3227 /* Do I have to take care about the projected bit? I don't think so, since the dp3 returns only one
3228 * scalar, and projected sampling would require 4.
3230 * It is a dependent read - not valid with conditional NP2 textures
3232 shader_glsl_get_sample_function(ins
->ctx
, sampler_idx
, 0, &sample_function
);
3233 mask_size
= shader_glsl_get_write_mask_size(sample_function
.coord_mask
);
3238 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, WINED3DSP_NOSWIZZLE
, NULL
, NULL
, NULL
,
3239 "dot(gl_TexCoord[%u].xyz, %s)", sampler_idx
, src0_param
.param_str
);
3243 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, WINED3DSP_NOSWIZZLE
, NULL
, NULL
, NULL
,
3244 "vec2(dot(gl_TexCoord[%u].xyz, %s), 0.0)", sampler_idx
, src0_param
.param_str
);
3248 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, WINED3DSP_NOSWIZZLE
, NULL
, NULL
, NULL
,
3249 "vec3(dot(gl_TexCoord[%u].xyz, %s), 0.0, 0.0)", sampler_idx
, src0_param
.param_str
);
3253 FIXME("Unexpected mask size %u\n", mask_size
);
3258 /** Process the WINED3DSIO_TEXDP3 instruction in GLSL:
3259 * Take a 3-component dot product of the TexCoord[dstreg] and src. */
3260 static void shader_glsl_texdp3(const struct wined3d_shader_instruction
*ins
)
3262 glsl_src_param_t src0_param
;
3263 DWORD dstreg
= ins
->dst
[0].reg
.idx
;
3264 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
3266 unsigned int mask_size
;
3268 dst_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
3269 mask_size
= shader_glsl_get_write_mask_size(dst_mask
);
3270 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
3272 if (mask_size
> 1) {
3273 shader_addline(ins
->ctx
->buffer
, "vec%d(dot(T%u.xyz, %s)));\n", mask_size
, dstreg
, src0_param
.param_str
);
3275 shader_addline(ins
->ctx
->buffer
, "dot(T%u.xyz, %s));\n", dstreg
, src0_param
.param_str
);
3279 /** Process the WINED3DSIO_TEXDEPTH instruction in GLSL:
3280 * Calculate the depth as dst.x / dst.y */
3281 static void shader_glsl_texdepth(const struct wined3d_shader_instruction
*ins
)
3283 glsl_dst_param_t dst_param
;
3285 shader_glsl_add_dst_param(ins
, &ins
->dst
[0], &dst_param
);
3287 /* Tests show that texdepth never returns anything below 0.0, and that r5.y is clamped to 1.0.
3288 * Negative input is accepted, -0.25 / -0.5 returns 0.5. GL should clamp gl_FragDepth to [0;1], but
3289 * this doesn't always work, so clamp the results manually. Whether or not the x value is clamped at 1
3290 * too is irrelevant, since if x = 0, any y value < 1.0 (and > 1.0 is not allowed) results in a result
3293 shader_addline(ins
->ctx
->buffer
, "gl_FragDepth = clamp((%s.x / min(%s.y, 1.0)), 0.0, 1.0);\n",
3294 dst_param
.reg_name
, dst_param
.reg_name
);
3297 /** Process the WINED3DSIO_TEXM3X2DEPTH instruction in GLSL:
3298 * Last row of a 3x2 matrix multiply, use the result to calculate the depth:
3299 * Calculate tmp0.y = TexCoord[dstreg] . src.xyz; (tmp0.x has already been calculated)
3300 * depth = (tmp0.y == 0.0) ? 1.0 : tmp0.x / tmp0.y
3302 static void shader_glsl_texm3x2depth(const struct wined3d_shader_instruction
*ins
)
3304 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
3305 DWORD dstreg
= ins
->dst
[0].reg
.idx
;
3306 glsl_src_param_t src0_param
;
3308 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
3310 shader_addline(ins
->ctx
->buffer
, "tmp0.y = dot(T%u.xyz, %s);\n", dstreg
, src0_param
.param_str
);
3311 shader_addline(ins
->ctx
->buffer
, "gl_FragDepth = (tmp0.y == 0.0) ? 1.0 : clamp(tmp0.x / tmp0.y, 0.0, 1.0);\n");
3314 /** Process the WINED3DSIO_TEXM3X2PAD instruction in GLSL
3315 * Calculate the 1st of a 2-row matrix multiplication. */
3316 static void shader_glsl_texm3x2pad(const struct wined3d_shader_instruction
*ins
)
3318 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
3319 DWORD reg
= ins
->dst
[0].reg
.idx
;
3320 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
3321 glsl_src_param_t src0_param
;
3323 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
3324 shader_addline(buffer
, "tmp0.x = dot(T%u.xyz, %s);\n", reg
, src0_param
.param_str
);
3327 /** Process the WINED3DSIO_TEXM3X3PAD instruction in GLSL
3328 * Calculate the 1st or 2nd row of a 3-row matrix multiplication. */
3329 static void shader_glsl_texm3x3pad(const struct wined3d_shader_instruction
*ins
)
3331 IWineD3DBaseShaderImpl
*shader
= (IWineD3DBaseShaderImpl
*)ins
->ctx
->shader
;
3332 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
3333 DWORD reg
= ins
->dst
[0].reg
.idx
;
3334 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
3335 SHADER_PARSE_STATE
* current_state
= &shader
->baseShader
.parse_state
;
3336 glsl_src_param_t src0_param
;
3338 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
3339 shader_addline(buffer
, "tmp0.%c = dot(T%u.xyz, %s);\n", 'x' + current_state
->current_row
, reg
, src0_param
.param_str
);
3340 current_state
->texcoord_w
[current_state
->current_row
++] = reg
;
3343 static void shader_glsl_texm3x2tex(const struct wined3d_shader_instruction
*ins
)
3345 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
3346 DWORD reg
= ins
->dst
[0].reg
.idx
;
3347 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
3348 glsl_src_param_t src0_param
;
3349 glsl_sample_function_t sample_function
;
3351 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
3352 shader_addline(buffer
, "tmp0.y = dot(T%u.xyz, %s);\n", reg
, src0_param
.param_str
);
3354 shader_glsl_get_sample_function(ins
->ctx
, reg
, 0, &sample_function
);
3356 /* Sample the texture using the calculated coordinates */
3357 shader_glsl_gen_sample_code(ins
, reg
, &sample_function
, WINED3DSP_NOSWIZZLE
, NULL
, NULL
, NULL
, "tmp0.xy");
3360 /** Process the WINED3DSIO_TEXM3X3TEX instruction in GLSL
3361 * Perform the 3rd row of a 3x3 matrix multiply, then sample the texture using the calculated coordinates */
3362 static void shader_glsl_texm3x3tex(const struct wined3d_shader_instruction
*ins
)
3364 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
3365 IWineD3DBaseShaderImpl
*shader
= (IWineD3DBaseShaderImpl
*)ins
->ctx
->shader
;
3366 SHADER_PARSE_STATE
*current_state
= &shader
->baseShader
.parse_state
;
3367 glsl_src_param_t src0_param
;
3368 DWORD reg
= ins
->dst
[0].reg
.idx
;
3369 glsl_sample_function_t sample_function
;
3371 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
3372 shader_addline(ins
->ctx
->buffer
, "tmp0.z = dot(T%u.xyz, %s);\n", reg
, src0_param
.param_str
);
3374 /* Dependent read, not valid with conditional NP2 */
3375 shader_glsl_get_sample_function(ins
->ctx
, reg
, 0, &sample_function
);
3377 /* Sample the texture using the calculated coordinates */
3378 shader_glsl_gen_sample_code(ins
, reg
, &sample_function
, WINED3DSP_NOSWIZZLE
, NULL
, NULL
, NULL
, "tmp0.xyz");
3380 current_state
->current_row
= 0;
3383 /** Process the WINED3DSIO_TEXM3X3 instruction in GLSL
3384 * Perform the 3rd row of a 3x3 matrix multiply */
3385 static void shader_glsl_texm3x3(const struct wined3d_shader_instruction
*ins
)
3387 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
3388 IWineD3DBaseShaderImpl
*shader
= (IWineD3DBaseShaderImpl
*)ins
->ctx
->shader
;
3389 SHADER_PARSE_STATE
*current_state
= &shader
->baseShader
.parse_state
;
3390 glsl_src_param_t src0_param
;
3392 DWORD reg
= ins
->dst
[0].reg
.idx
;
3394 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
3396 shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
3397 shader_glsl_get_write_mask(&ins
->dst
[0], dst_mask
);
3398 shader_addline(ins
->ctx
->buffer
, "vec4(tmp0.xy, dot(T%u.xyz, %s), 1.0)%s);\n", reg
, src0_param
.param_str
, dst_mask
);
3400 current_state
->current_row
= 0;
3403 /* Process the WINED3DSIO_TEXM3X3SPEC instruction in GLSL
3404 * Perform the final texture lookup based on the previous 2 3x3 matrix multiplies */
3405 static void shader_glsl_texm3x3spec(const struct wined3d_shader_instruction
*ins
)
3407 IWineD3DBaseShaderImpl
*shader
= (IWineD3DBaseShaderImpl
*)ins
->ctx
->shader
;
3408 DWORD reg
= ins
->dst
[0].reg
.idx
;
3409 glsl_src_param_t src0_param
;
3410 glsl_src_param_t src1_param
;
3411 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
3412 SHADER_PARSE_STATE
* current_state
= &shader
->baseShader
.parse_state
;
3413 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
3414 glsl_sample_function_t sample_function
;
3416 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
3417 shader_glsl_add_src_param(ins
, &ins
->src
[1], src_mask
, &src1_param
);
3419 /* Perform the last matrix multiply operation */
3420 shader_addline(buffer
, "tmp0.z = dot(T%u.xyz, %s);\n", reg
, src0_param
.param_str
);
3421 /* Reflection calculation */
3422 shader_addline(buffer
, "tmp0.xyz = -reflect((%s), normalize(tmp0.xyz));\n", src1_param
.param_str
);
3424 /* Dependent read, not valid with conditional NP2 */
3425 shader_glsl_get_sample_function(ins
->ctx
, reg
, 0, &sample_function
);
3427 /* Sample the texture */
3428 shader_glsl_gen_sample_code(ins
, reg
, &sample_function
, WINED3DSP_NOSWIZZLE
, NULL
, NULL
, NULL
, "tmp0.xyz");
3430 current_state
->current_row
= 0;
3433 /* Process the WINED3DSIO_TEXM3X3VSPEC instruction in GLSL
3434 * Perform the final texture lookup based on the previous 2 3x3 matrix multiplies */
3435 static void shader_glsl_texm3x3vspec(const struct wined3d_shader_instruction
*ins
)
3437 IWineD3DBaseShaderImpl
*shader
= (IWineD3DBaseShaderImpl
*)ins
->ctx
->shader
;
3438 DWORD reg
= ins
->dst
[0].reg
.idx
;
3439 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
3440 SHADER_PARSE_STATE
* current_state
= &shader
->baseShader
.parse_state
;
3441 glsl_src_param_t src0_param
;
3442 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
3443 glsl_sample_function_t sample_function
;
3445 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
3447 /* Perform the last matrix multiply operation */
3448 shader_addline(buffer
, "tmp0.z = dot(vec3(T%u), vec3(%s));\n", reg
, src0_param
.param_str
);
3450 /* Construct the eye-ray vector from w coordinates */
3451 shader_addline(buffer
, "tmp1.xyz = normalize(vec3(gl_TexCoord[%u].w, gl_TexCoord[%u].w, gl_TexCoord[%u].w));\n",
3452 current_state
->texcoord_w
[0], current_state
->texcoord_w
[1], reg
);
3453 shader_addline(buffer
, "tmp0.xyz = -reflect(tmp1.xyz, normalize(tmp0.xyz));\n");
3455 /* Dependent read, not valid with conditional NP2 */
3456 shader_glsl_get_sample_function(ins
->ctx
, reg
, 0, &sample_function
);
3458 /* Sample the texture using the calculated coordinates */
3459 shader_glsl_gen_sample_code(ins
, reg
, &sample_function
, WINED3DSP_NOSWIZZLE
, NULL
, NULL
, NULL
, "tmp0.xyz");
3461 current_state
->current_row
= 0;
3464 /** Process the WINED3DSIO_TEXBEM instruction in GLSL.
3465 * Apply a fake bump map transform.
3466 * texbem is pshader <= 1.3 only, this saves a few version checks
3468 static void shader_glsl_texbem(const struct wined3d_shader_instruction
*ins
)
3470 IWineD3DBaseShaderImpl
*shader
= (IWineD3DBaseShaderImpl
*)ins
->ctx
->shader
;
3471 IWineD3DDeviceImpl
*deviceImpl
= (IWineD3DDeviceImpl
*)shader
->baseShader
.device
;
3472 glsl_sample_function_t sample_function
;
3473 glsl_src_param_t coord_param
;
3479 sampler_idx
= ins
->dst
[0].reg
.idx
;
3480 flags
= deviceImpl
->stateBlock
->state
.texture_states
[sampler_idx
][WINED3DTSS_TEXTURETRANSFORMFLAGS
];
3482 /* Dependent read, not valid with conditional NP2 */
3483 shader_glsl_get_sample_function(ins
->ctx
, sampler_idx
, 0, &sample_function
);
3484 mask
= sample_function
.coord_mask
;
3486 shader_glsl_write_mask_to_str(mask
, coord_mask
);
3488 /* with projective textures, texbem only divides the static texture coord, not the displacement,
3489 * so we can't let the GL handle this.
3491 if (flags
& WINED3DTTFF_PROJECTED
) {
3493 char coord_div_mask
[3];
3494 switch (flags
& ~WINED3DTTFF_PROJECTED
) {
3495 case WINED3DTTFF_COUNT1
: FIXME("WINED3DTTFF_PROJECTED with WINED3DTTFF_COUNT1?\n"); break;
3496 case WINED3DTTFF_COUNT2
: div_mask
= WINED3DSP_WRITEMASK_1
; break;
3497 case WINED3DTTFF_COUNT3
: div_mask
= WINED3DSP_WRITEMASK_2
; break;
3498 case WINED3DTTFF_COUNT4
:
3499 case WINED3DTTFF_DISABLE
: div_mask
= WINED3DSP_WRITEMASK_3
; break;
3501 shader_glsl_write_mask_to_str(div_mask
, coord_div_mask
);
3502 shader_addline(ins
->ctx
->buffer
, "T%u%s /= T%u%s;\n", sampler_idx
, coord_mask
, sampler_idx
, coord_div_mask
);
3505 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
, &coord_param
);
3507 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, WINED3DSP_NOSWIZZLE
, NULL
, NULL
, NULL
,
3508 "T%u%s + vec4(bumpenvmat%d * %s, 0.0, 0.0)%s", sampler_idx
, coord_mask
, sampler_idx
,
3509 coord_param
.param_str
, coord_mask
);
3511 if (ins
->handler_idx
== WINED3DSIH_TEXBEML
)
3513 glsl_src_param_t luminance_param
;
3514 glsl_dst_param_t dst_param
;
3516 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_2
, &luminance_param
);
3517 shader_glsl_add_dst_param(ins
, &ins
->dst
[0], &dst_param
);
3519 shader_addline(ins
->ctx
->buffer
, "%s%s *= (%s * luminancescale%d + luminanceoffset%d);\n",
3520 dst_param
.reg_name
, dst_param
.mask_str
,
3521 luminance_param
.param_str
, sampler_idx
, sampler_idx
);
3525 static void shader_glsl_bem(const struct wined3d_shader_instruction
*ins
)
3527 glsl_src_param_t src0_param
, src1_param
;
3528 DWORD sampler_idx
= ins
->dst
[0].reg
.idx
;
3530 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
, &src0_param
);
3531 shader_glsl_add_src_param(ins
, &ins
->src
[1], WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
, &src1_param
);
3533 shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
3534 shader_addline(ins
->ctx
->buffer
, "%s + bumpenvmat%d * %s);\n",
3535 src0_param
.param_str
, sampler_idx
, src1_param
.param_str
);
3538 /** Process the WINED3DSIO_TEXREG2AR instruction in GLSL
3539 * Sample 2D texture at dst using the alpha & red (wx) components of src as texture coordinates */
3540 static void shader_glsl_texreg2ar(const struct wined3d_shader_instruction
*ins
)
3542 glsl_src_param_t src0_param
;
3543 DWORD sampler_idx
= ins
->dst
[0].reg
.idx
;
3544 glsl_sample_function_t sample_function
;
3546 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_ALL
, &src0_param
);
3548 shader_glsl_get_sample_function(ins
->ctx
, sampler_idx
, 0, &sample_function
);
3549 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, WINED3DSP_NOSWIZZLE
, NULL
, NULL
, NULL
,
3550 "%s.wx", src0_param
.reg_name
);
3553 /** Process the WINED3DSIO_TEXREG2GB instruction in GLSL
3554 * Sample 2D texture at dst using the green & blue (yz) components of src as texture coordinates */
3555 static void shader_glsl_texreg2gb(const struct wined3d_shader_instruction
*ins
)
3557 glsl_src_param_t src0_param
;
3558 DWORD sampler_idx
= ins
->dst
[0].reg
.idx
;
3559 glsl_sample_function_t sample_function
;
3561 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_ALL
, &src0_param
);
3563 shader_glsl_get_sample_function(ins
->ctx
, sampler_idx
, 0, &sample_function
);
3564 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, WINED3DSP_NOSWIZZLE
, NULL
, NULL
, NULL
,
3565 "%s.yz", src0_param
.reg_name
);
3568 /** Process the WINED3DSIO_TEXREG2RGB instruction in GLSL
3569 * Sample texture at dst using the rgb (xyz) components of src as texture coordinates */
3570 static void shader_glsl_texreg2rgb(const struct wined3d_shader_instruction
*ins
)
3572 glsl_src_param_t src0_param
;
3573 DWORD sampler_idx
= ins
->dst
[0].reg
.idx
;
3574 glsl_sample_function_t sample_function
;
3576 /* Dependent read, not valid with conditional NP2 */
3577 shader_glsl_get_sample_function(ins
->ctx
, sampler_idx
, 0, &sample_function
);
3578 shader_glsl_add_src_param(ins
, &ins
->src
[0], sample_function
.coord_mask
, &src0_param
);
3580 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, WINED3DSP_NOSWIZZLE
, NULL
, NULL
, NULL
,
3581 "%s", src0_param
.param_str
);
3584 /** Process the WINED3DSIO_TEXKILL instruction in GLSL.
3585 * If any of the first 3 components are < 0, discard this pixel */
3586 static void shader_glsl_texkill(const struct wined3d_shader_instruction
*ins
)
3588 glsl_dst_param_t dst_param
;
3590 /* The argument is a destination parameter, and no writemasks are allowed */
3591 shader_glsl_add_dst_param(ins
, &ins
->dst
[0], &dst_param
);
3592 if (ins
->ctx
->reg_maps
->shader_version
.major
>= 2)
3594 /* 2.0 shaders compare all 4 components in texkill */
3595 shader_addline(ins
->ctx
->buffer
, "if (any(lessThan(%s.xyzw, vec4(0.0)))) discard;\n", dst_param
.reg_name
);
3597 /* 1.X shaders only compare the first 3 components, probably due to the nature of the texkill
3598 * instruction as a tex* instruction, and phase, which kills all a / w components. Even if all
3599 * 4 components are defined, only the first 3 are used
3601 shader_addline(ins
->ctx
->buffer
, "if (any(lessThan(%s.xyz, vec3(0.0)))) discard;\n", dst_param
.reg_name
);
3605 /** Process the WINED3DSIO_DP2ADD instruction in GLSL.
3606 * dst = dot2(src0, src1) + src2 */
3607 static void shader_glsl_dp2add(const struct wined3d_shader_instruction
*ins
)
3609 glsl_src_param_t src0_param
;
3610 glsl_src_param_t src1_param
;
3611 glsl_src_param_t src2_param
;
3613 unsigned int mask_size
;
3615 write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
3616 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
3618 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
, &src0_param
);
3619 shader_glsl_add_src_param(ins
, &ins
->src
[1], WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
, &src1_param
);
3620 shader_glsl_add_src_param(ins
, &ins
->src
[2], WINED3DSP_WRITEMASK_0
, &src2_param
);
3622 if (mask_size
> 1) {
3623 shader_addline(ins
->ctx
->buffer
, "vec%d(dot(%s, %s) + %s));\n",
3624 mask_size
, src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
3626 shader_addline(ins
->ctx
->buffer
, "dot(%s, %s) + %s);\n",
3627 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
3631 static void shader_glsl_input_pack(IWineD3DPixelShader
*iface
, struct wined3d_shader_buffer
*buffer
,
3632 const struct wined3d_shader_signature_element
*input_signature
, const struct shader_reg_maps
*reg_maps
,
3633 enum vertexprocessing_mode vertexprocessing
)
3636 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*)iface
;
3637 WORD map
= reg_maps
->input_registers
;
3639 for (i
= 0; map
; map
>>= 1, ++i
)
3641 const char *semantic_name
;
3646 if (!(map
& 1)) continue;
3648 semantic_name
= input_signature
[i
].semantic_name
;
3649 semantic_idx
= input_signature
[i
].semantic_idx
;
3650 shader_glsl_write_mask_to_str(input_signature
[i
].mask
, reg_mask
);
3652 if (shader_match_semantic(semantic_name
, WINED3DDECLUSAGE_TEXCOORD
))
3654 if (semantic_idx
< 8 && vertexprocessing
== pretransformed
)
3655 shader_addline(buffer
, "IN[%u]%s = gl_TexCoord[%u]%s;\n",
3656 This
->input_reg_map
[i
], reg_mask
, semantic_idx
, reg_mask
);
3658 shader_addline(buffer
, "IN[%u]%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
3659 This
->input_reg_map
[i
], reg_mask
, reg_mask
);
3661 else if (shader_match_semantic(semantic_name
, WINED3DDECLUSAGE_COLOR
))
3664 shader_addline(buffer
, "IN[%u]%s = vec4(gl_Color)%s;\n",
3665 This
->input_reg_map
[i
], reg_mask
, reg_mask
);
3666 else if (semantic_idx
== 1)
3667 shader_addline(buffer
, "IN[%u]%s = vec4(gl_SecondaryColor)%s;\n",
3668 This
->input_reg_map
[i
], reg_mask
, reg_mask
);
3670 shader_addline(buffer
, "IN[%u]%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
3671 This
->input_reg_map
[i
], reg_mask
, reg_mask
);
3675 shader_addline(buffer
, "IN[%u]%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
3676 This
->input_reg_map
[i
], reg_mask
, reg_mask
);
3681 /*********************************************
3682 * Vertex Shader Specific Code begins here
3683 ********************************************/
3685 static void add_glsl_program_entry(struct shader_glsl_priv
*priv
, struct glsl_shader_prog_link
*entry
) {
3686 glsl_program_key_t key
;
3688 key
.vshader
= entry
->vshader
;
3689 key
.pshader
= entry
->pshader
;
3690 key
.vs_args
= entry
->vs_args
;
3691 key
.ps_args
= entry
->ps_args
;
3693 if (wine_rb_put(&priv
->program_lookup
, &key
, &entry
->program_lookup_entry
) == -1)
3695 ERR("Failed to insert program entry.\n");
3699 static struct glsl_shader_prog_link
*get_glsl_program_entry(struct shader_glsl_priv
*priv
,
3700 IWineD3DVertexShader
*vshader
, IWineD3DPixelShader
*pshader
, struct vs_compile_args
*vs_args
,
3701 struct ps_compile_args
*ps_args
) {
3702 struct wine_rb_entry
*entry
;
3703 glsl_program_key_t key
;
3705 key
.vshader
= vshader
;
3706 key
.pshader
= pshader
;
3707 key
.vs_args
= *vs_args
;
3708 key
.ps_args
= *ps_args
;
3710 entry
= wine_rb_get(&priv
->program_lookup
, &key
);
3711 return entry
? WINE_RB_ENTRY_VALUE(entry
, struct glsl_shader_prog_link
, program_lookup_entry
) : NULL
;
3714 /* GL locking is done by the caller */
3715 static void delete_glsl_program_entry(struct shader_glsl_priv
*priv
, const struct wined3d_gl_info
*gl_info
,
3716 struct glsl_shader_prog_link
*entry
)
3718 glsl_program_key_t key
;
3720 key
.vshader
= entry
->vshader
;
3721 key
.pshader
= entry
->pshader
;
3722 key
.vs_args
= entry
->vs_args
;
3723 key
.ps_args
= entry
->ps_args
;
3724 wine_rb_remove(&priv
->program_lookup
, &key
);
3726 GL_EXTCALL(glDeleteObjectARB(entry
->programId
));
3727 if (entry
->vshader
) list_remove(&entry
->vshader_entry
);
3728 if (entry
->pshader
) list_remove(&entry
->pshader_entry
);
3729 HeapFree(GetProcessHeap(), 0, entry
->vuniformF_locations
);
3730 HeapFree(GetProcessHeap(), 0, entry
->puniformF_locations
);
3731 HeapFree(GetProcessHeap(), 0, entry
);
3734 static void handle_ps3_input(struct wined3d_shader_buffer
*buffer
, const struct wined3d_gl_info
*gl_info
, const DWORD
*map
,
3735 const struct wined3d_shader_signature_element
*input_signature
, const struct shader_reg_maps
*reg_maps_in
,
3736 const struct wined3d_shader_signature_element
*output_signature
, const struct shader_reg_maps
*reg_maps_out
)
3739 const char *semantic_name_in
;
3740 UINT semantic_idx_in
;
3743 unsigned int in_count
= vec4_varyings(3, gl_info
);
3745 char destination
[50];
3746 WORD input_map
, output_map
;
3748 set
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*set
) * (in_count
+ 2));
3750 input_map
= reg_maps_in
->input_registers
;
3751 for (i
= 0; input_map
; input_map
>>= 1, ++i
)
3753 if (!(input_map
& 1)) continue;
3756 /* Declared, but not read register */
3757 if (in_idx
== ~0U) continue;
3758 if (in_idx
>= (in_count
+ 2))
3760 FIXME("More input varyings declared than supported, expect issues.\n");
3764 if (in_idx
== in_count
) {
3765 sprintf(destination
, "gl_FrontColor");
3766 } else if (in_idx
== in_count
+ 1) {
3767 sprintf(destination
, "gl_FrontSecondaryColor");
3769 sprintf(destination
, "IN[%u]", in_idx
);
3772 semantic_name_in
= input_signature
[i
].semantic_name
;
3773 semantic_idx_in
= input_signature
[i
].semantic_idx
;
3776 output_map
= reg_maps_out
->output_registers
;
3777 for (j
= 0; output_map
; output_map
>>= 1, ++j
)
3781 if (!(output_map
& 1)
3782 || semantic_idx_in
!= output_signature
[j
].semantic_idx
3783 || strcmp(semantic_name_in
, output_signature
[j
].semantic_name
)
3784 || !(mask
= input_signature
[i
].mask
& output_signature
[j
].mask
))
3788 shader_glsl_write_mask_to_str(mask
, reg_mask
);
3790 shader_addline(buffer
, "%s%s = clamp(OUT[%u]%s, -FLT_MAX, FLT_MAX);\n",
3791 destination
, reg_mask
, j
, reg_mask
);
3795 for (i
= 0; i
< in_count
+ 2; ++i
)
3799 if (!set
[i
] || set
[i
] == WINED3DSP_WRITEMASK_ALL
)
3802 if (set
[i
] == ~0U) set
[i
] = 0;
3805 if (!(set
[i
] & WINED3DSP_WRITEMASK_0
)) reg_mask
[size
++] = 'x';
3806 if (!(set
[i
] & WINED3DSP_WRITEMASK_1
)) reg_mask
[size
++] = 'y';
3807 if (!(set
[i
] & WINED3DSP_WRITEMASK_2
)) reg_mask
[size
++] = 'z';
3808 if (!(set
[i
] & WINED3DSP_WRITEMASK_3
)) reg_mask
[size
++] = 'w';
3809 reg_mask
[size
] = '\0';
3811 if (i
== in_count
) sprintf(destination
, "gl_FrontColor");
3812 else if (i
== in_count
+ 1) sprintf(destination
, "gl_FrontSecondaryColor");
3813 else sprintf(destination
, "IN[%u]", i
);
3815 if (size
== 1) shader_addline(buffer
, "%s.%s = 0.0;\n", destination
, reg_mask
);
3816 else shader_addline(buffer
, "%s.%s = vec%u(0.0);\n", destination
, reg_mask
, size
);
3819 HeapFree(GetProcessHeap(), 0, set
);
3822 /* GL locking is done by the caller */
3823 static GLhandleARB
generate_param_reorder_function(struct wined3d_shader_buffer
*buffer
,
3824 IWineD3DVertexShader
*vertexshader
, IWineD3DPixelShader
*pixelshader
, const struct wined3d_gl_info
*gl_info
)
3826 GLhandleARB ret
= 0;
3827 IWineD3DVertexShaderImpl
*vs
= (IWineD3DVertexShaderImpl
*) vertexshader
;
3828 IWineD3DPixelShaderImpl
*ps
= (IWineD3DPixelShaderImpl
*) pixelshader
;
3829 DWORD ps_major
= ps
? ps
->baseShader
.reg_maps
.shader_version
.major
: 0;
3831 const char *semantic_name
;
3834 const struct wined3d_shader_signature_element
*output_signature
= vs
->baseShader
.output_signature
;
3835 WORD map
= vs
->baseShader
.reg_maps
.output_registers
;
3837 shader_buffer_clear(buffer
);
3839 shader_addline(buffer
, "#version 120\n");
3840 shader_addline(buffer
, "const float FLT_MAX = 1e38;\n");
3844 shader_addline(buffer
, "void order_ps_input(in vec4 OUT[%u]) {\n", MAX_REG_OUTPUT
);
3846 for (i
= 0; map
; map
>>= 1, ++i
)
3850 if (!(map
& 1)) continue;
3852 semantic_name
= output_signature
[i
].semantic_name
;
3853 semantic_idx
= output_signature
[i
].semantic_idx
;
3854 write_mask
= output_signature
[i
].mask
;
3855 shader_glsl_write_mask_to_str(write_mask
, reg_mask
);
3857 if (shader_match_semantic(semantic_name
, WINED3DDECLUSAGE_COLOR
))
3860 shader_addline(buffer
, "gl_FrontColor%s = clamp(OUT[%u]%s, -FLT_MAX, FLT_MAX);\n",
3861 reg_mask
, i
, reg_mask
);
3862 else if (semantic_idx
== 1)
3863 shader_addline(buffer
, "gl_FrontSecondaryColor%s = clamp(OUT[%u]%s, -FLT_MAX, FLT_MAX);\n",
3864 reg_mask
, i
, reg_mask
);
3866 else if (shader_match_semantic(semantic_name
, WINED3DDECLUSAGE_POSITION
))
3868 shader_addline(buffer
, "gl_Position%s = clamp(OUT[%u]%s, -FLT_MAX, FLT_MAX);\n",
3869 reg_mask
, i
, reg_mask
);
3871 else if (shader_match_semantic(semantic_name
, WINED3DDECLUSAGE_TEXCOORD
))
3873 if (semantic_idx
< 8)
3875 if (!(gl_info
->quirks
& WINED3D_QUIRK_SET_TEXCOORD_W
) || ps_major
> 0)
3876 write_mask
|= WINED3DSP_WRITEMASK_3
;
3878 shader_addline(buffer
, "gl_TexCoord[%u]%s = clamp(OUT[%u]%s, -FLT_MAX, FLT_MAX);\n",
3879 semantic_idx
, reg_mask
, i
, reg_mask
);
3880 if (!(write_mask
& WINED3DSP_WRITEMASK_3
))
3881 shader_addline(buffer
, "gl_TexCoord[%u].w = 1.0;\n", semantic_idx
);
3884 else if (shader_match_semantic(semantic_name
, WINED3DDECLUSAGE_PSIZE
))
3886 shader_addline(buffer
, "gl_PointSize = clamp(OUT[%u].%c, -FLT_MAX, FLT_MAX);\n", i
, reg_mask
[1]);
3888 else if (shader_match_semantic(semantic_name
, WINED3DDECLUSAGE_FOG
))
3890 shader_addline(buffer
, "gl_FogFragCoord = clamp(OUT[%u].%c, -FLT_MAX, FLT_MAX);\n", i
, reg_mask
[1]);
3893 shader_addline(buffer
, "}\n");
3898 /* This one is tricky: a 3.0 pixel shader reads from a 3.0 vertex shader */
3899 shader_addline(buffer
, "varying vec4 IN[%u];\n", vec4_varyings(3, gl_info
));
3900 shader_addline(buffer
, "void order_ps_input(in vec4 OUT[%u]) {\n", MAX_REG_OUTPUT
);
3902 /* First, sort out position and point size. Those are not passed to the pixel shader */
3903 for (i
= 0; map
; map
>>= 1, ++i
)
3905 if (!(map
& 1)) continue;
3907 semantic_name
= output_signature
[i
].semantic_name
;
3908 shader_glsl_write_mask_to_str(output_signature
[i
].mask
, reg_mask
);
3910 if (shader_match_semantic(semantic_name
, WINED3DDECLUSAGE_POSITION
))
3912 shader_addline(buffer
, "gl_Position%s = clamp(OUT[%u]%s, -FLT_MAX, FLT_MAX);\n",
3913 reg_mask
, i
, reg_mask
);
3915 else if (shader_match_semantic(semantic_name
, WINED3DDECLUSAGE_PSIZE
))
3917 shader_addline(buffer
, "gl_PointSize = clamp(OUT[%u].%c, -FLT_MAX, FLT_MAX);\n", i
, reg_mask
[1]);
3921 /* Then, fix the pixel shader input */
3922 handle_ps3_input(buffer
, gl_info
, ps
->input_reg_map
, ps
->baseShader
.input_signature
,
3923 &ps
->baseShader
.reg_maps
, output_signature
, &vs
->baseShader
.reg_maps
);
3925 shader_addline(buffer
, "}\n");
3928 ret
= GL_EXTCALL(glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB
));
3929 checkGLcall("glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB)");
3930 shader_glsl_compile(gl_info
, ret
, buffer
->buffer
);
3935 /* GL locking is done by the caller */
3936 static void hardcode_local_constants(IWineD3DBaseShaderImpl
*shader
, const struct wined3d_gl_info
*gl_info
,
3937 GLhandleARB programId
, char prefix
)
3939 const local_constant
*lconst
;
3944 LIST_FOR_EACH_ENTRY(lconst
, &shader
->baseShader
.constantsF
, local_constant
, entry
) {
3945 value
= (const float *)lconst
->value
;
3946 snprintf(glsl_name
, sizeof(glsl_name
), "%cLC%u", prefix
, lconst
->idx
);
3947 tmp_loc
= GL_EXTCALL(glGetUniformLocationARB(programId
, glsl_name
));
3948 GL_EXTCALL(glUniform4fvARB(tmp_loc
, 1, value
));
3950 checkGLcall("Hardcoding local constants");
3953 /* GL locking is done by the caller */
3954 static GLuint
shader_glsl_generate_pshader(const struct wined3d_context
*context
,
3955 struct wined3d_shader_buffer
*buffer
, IWineD3DPixelShaderImpl
*This
,
3956 const struct ps_compile_args
*args
, struct ps_np2fixup_info
*np2fixup_info
)
3958 const struct shader_reg_maps
*reg_maps
= &This
->baseShader
.reg_maps
;
3959 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
3960 CONST DWORD
*function
= This
->baseShader
.function
;
3961 struct shader_glsl_ctx_priv priv_ctx
;
3963 /* Create the hw GLSL shader object and assign it as the shader->prgId */
3964 GLhandleARB shader_obj
= GL_EXTCALL(glCreateShaderObjectARB(GL_FRAGMENT_SHADER_ARB
));
3966 memset(&priv_ctx
, 0, sizeof(priv_ctx
));
3967 priv_ctx
.cur_ps_args
= args
;
3968 priv_ctx
.cur_np2fixup_info
= np2fixup_info
;
3970 shader_addline(buffer
, "#version 120\n");
3972 if (gl_info
->supported
[ARB_SHADER_TEXTURE_LOD
] && reg_maps
->usestexldd
)
3974 shader_addline(buffer
, "#extension GL_ARB_shader_texture_lod : enable\n");
3976 if (gl_info
->supported
[ARB_TEXTURE_RECTANGLE
])
3978 /* The spec says that it doesn't have to be explicitly enabled, but the nvidia
3979 * drivers write a warning if we don't do so
3981 shader_addline(buffer
, "#extension GL_ARB_texture_rectangle : enable\n");
3983 if (gl_info
->supported
[EXT_GPU_SHADER4
])
3985 shader_addline(buffer
, "#extension GL_EXT_gpu_shader4 : enable\n");
3988 /* Base Declarations */
3989 shader_generate_glsl_declarations(context
, buffer
, (IWineD3DBaseShader
*)This
, reg_maps
, &priv_ctx
);
3991 /* Pack 3.0 inputs */
3992 if (reg_maps
->shader_version
.major
>= 3 && args
->vp_mode
!= vertexshader
)
3994 shader_glsl_input_pack((IWineD3DPixelShader
*) This
, buffer
,
3995 This
->baseShader
.input_signature
, reg_maps
, args
->vp_mode
);
3998 /* Base Shader Body */
3999 shader_generate_main((IWineD3DBaseShader
*)This
, buffer
, reg_maps
, function
, &priv_ctx
);
4001 /* Pixel shaders < 2.0 place the resulting color in R0 implicitly */
4002 if (reg_maps
->shader_version
.major
< 2)
4004 /* Some older cards like GeforceFX ones don't support multiple buffers, so also not gl_FragData */
4005 shader_addline(buffer
, "gl_FragData[0] = R0;\n");
4008 if (args
->srgb_correction
)
4010 shader_addline(buffer
, "tmp0.xyz = pow(gl_FragData[0].xyz, vec3(srgb_const0.x));\n");
4011 shader_addline(buffer
, "tmp0.xyz = tmp0.xyz * vec3(srgb_const0.y) - vec3(srgb_const0.z);\n");
4012 shader_addline(buffer
, "tmp1.xyz = gl_FragData[0].xyz * vec3(srgb_const0.w);\n");
4013 shader_addline(buffer
, "bvec3 srgb_compare = lessThan(gl_FragData[0].xyz, vec3(srgb_const1.x));\n");
4014 shader_addline(buffer
, "gl_FragData[0].xyz = mix(tmp0.xyz, tmp1.xyz, vec3(srgb_compare));\n");
4015 shader_addline(buffer
, "gl_FragData[0] = clamp(gl_FragData[0], 0.0, 1.0);\n");
4017 /* Pixel shader < 3.0 do not replace the fog stage.
4018 * This implements linear fog computation and blending.
4019 * TODO: non linear fog
4020 * NOTE: gl_Fog.start and gl_Fog.end don't hold fog start s and end e but
4021 * -1/(e-s) and e/(e-s) respectively.
4023 if (reg_maps
->shader_version
.major
< 3)
4026 case FOG_OFF
: break;
4028 shader_addline(buffer
, "float fogstart = -1.0 / (gl_Fog.end - gl_Fog.start);\n");
4029 shader_addline(buffer
, "float fogend = gl_Fog.end * -fogstart;\n");
4030 shader_addline(buffer
, "float Fog = clamp(gl_FogFragCoord * fogstart + fogend, 0.0, 1.0);\n");
4031 shader_addline(buffer
, "gl_FragData[0].xyz = mix(gl_Fog.color.xyz, gl_FragData[0].xyz, Fog);\n");
4034 /* Fog = e^(-gl_Fog.density * gl_FogFragCoord) */
4035 shader_addline(buffer
, "float Fog = exp(-gl_Fog.density * gl_FogFragCoord);\n");
4036 shader_addline(buffer
, "Fog = clamp(Fog, 0.0, 1.0);\n");
4037 shader_addline(buffer
, "gl_FragData[0].xyz = mix(gl_Fog.color.xyz, gl_FragData[0].xyz, Fog);\n");
4040 /* Fog = e^(-(gl_Fog.density * gl_FogFragCoord)^2) */
4041 shader_addline(buffer
, "float Fog = exp(-gl_Fog.density * gl_Fog.density * gl_FogFragCoord * gl_FogFragCoord);\n");
4042 shader_addline(buffer
, "Fog = clamp(Fog, 0.0, 1.0);\n");
4043 shader_addline(buffer
, "gl_FragData[0].xyz = mix(gl_Fog.color.xyz, gl_FragData[0].xyz, Fog);\n");
4048 shader_addline(buffer
, "}\n");
4050 TRACE("Compiling shader object %u\n", shader_obj
);
4051 shader_glsl_compile(gl_info
, shader_obj
, buffer
->buffer
);
4053 /* Store the shader object */
4057 /* GL locking is done by the caller */
4058 static GLuint
shader_glsl_generate_vshader(const struct wined3d_context
*context
,
4059 struct wined3d_shader_buffer
*buffer
, IWineD3DVertexShaderImpl
*This
,
4060 const struct vs_compile_args
*args
)
4062 const struct shader_reg_maps
*reg_maps
= &This
->baseShader
.reg_maps
;
4063 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
4064 CONST DWORD
*function
= This
->baseShader
.function
;
4065 struct shader_glsl_ctx_priv priv_ctx
;
4067 /* Create the hw GLSL shader program and assign it as the shader->prgId */
4068 GLhandleARB shader_obj
= GL_EXTCALL(glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB
));
4070 shader_addline(buffer
, "#version 120\n");
4072 if (gl_info
->supported
[EXT_GPU_SHADER4
])
4074 shader_addline(buffer
, "#extension GL_EXT_gpu_shader4 : enable\n");
4077 memset(&priv_ctx
, 0, sizeof(priv_ctx
));
4078 priv_ctx
.cur_vs_args
= args
;
4080 /* Base Declarations */
4081 shader_generate_glsl_declarations(context
, buffer
, (IWineD3DBaseShader
*)This
, reg_maps
, &priv_ctx
);
4083 /* Base Shader Body */
4084 shader_generate_main((IWineD3DBaseShader
*)This
, buffer
, reg_maps
, function
, &priv_ctx
);
4086 /* Unpack outputs */
4087 shader_addline(buffer
, "order_ps_input(OUT);\n");
4089 /* The D3DRS_FOGTABLEMODE render state defines if the shader-generated fog coord is used
4090 * or if the fragment depth is used. If the fragment depth is used(FOGTABLEMODE != NONE),
4091 * the fog frag coord is thrown away. If the fog frag coord is used, but not written by
4092 * the shader, it is set to 0.0(fully fogged, since start = 1.0, end = 0.0)
4094 if(args
->fog_src
== VS_FOG_Z
) {
4095 shader_addline(buffer
, "gl_FogFragCoord = gl_Position.z;\n");
4096 } else if (!reg_maps
->fog
) {
4097 shader_addline(buffer
, "gl_FogFragCoord = 0.0;\n");
4100 /* Write the final position.
4102 * OpenGL coordinates specify the center of the pixel while d3d coords specify
4103 * the corner. The offsets are stored in z and w in posFixup. posFixup.y contains
4104 * 1.0 or -1.0 to turn the rendering upside down for offscreen rendering. PosFixup.x
4105 * contains 1.0 to allow a mad.
4107 shader_addline(buffer
, "gl_Position.y = gl_Position.y * posFixup.y;\n");
4108 shader_addline(buffer
, "gl_Position.xy += posFixup.zw * gl_Position.ww;\n");
4109 if(args
->clip_enabled
) {
4110 shader_addline(buffer
, "gl_ClipVertex = gl_Position;\n");
4113 /* Z coord [0;1]->[-1;1] mapping, see comment in transform_projection in state.c
4115 * Basically we want (in homogeneous coordinates) z = z * 2 - 1. However, shaders are run
4116 * before the homogeneous divide, so we have to take the w into account: z = ((z / w) * 2 - 1) * w,
4117 * which is the same as z = z * 2 - w.
4119 shader_addline(buffer
, "gl_Position.z = gl_Position.z * 2.0 - gl_Position.w;\n");
4121 shader_addline(buffer
, "}\n");
4123 TRACE("Compiling shader object %u\n", shader_obj
);
4124 shader_glsl_compile(gl_info
, shader_obj
, buffer
->buffer
);
4129 static GLhandleARB
find_glsl_pshader(const struct wined3d_context
*context
,
4130 struct wined3d_shader_buffer
*buffer
, IWineD3DPixelShaderImpl
*shader
,
4131 const struct ps_compile_args
*args
, const struct ps_np2fixup_info
**np2fixup_info
)
4133 struct wined3d_state
*state
= &((IWineD3DDeviceImpl
*)shader
->baseShader
.device
)->stateBlock
->state
;
4136 struct glsl_ps_compiled_shader
*new_array
;
4137 struct glsl_pshader_private
*shader_data
;
4138 struct ps_np2fixup_info
*np2fixup
= NULL
;
4141 if (!shader
->baseShader
.backend_data
)
4143 shader
->baseShader
.backend_data
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*shader_data
));
4144 if (!shader
->baseShader
.backend_data
)
4146 ERR("Failed to allocate backend data.\n");
4150 shader_data
= shader
->baseShader
.backend_data
;
4152 /* Usually we have very few GL shaders for each d3d shader(just 1 or maybe 2),
4153 * so a linear search is more performant than a hashmap or a binary search
4154 * (cache coherency etc)
4156 for (i
= 0; i
< shader_data
->num_gl_shaders
; ++i
)
4158 if (!memcmp(&shader_data
->gl_shaders
[i
].args
, args
, sizeof(*args
)))
4160 if (args
->np2_fixup
) *np2fixup_info
= &shader_data
->gl_shaders
[i
].np2fixup
;
4161 return shader_data
->gl_shaders
[i
].prgId
;
4165 TRACE("No matching GL shader found for shader %p, compiling a new shader.\n", shader
);
4166 if(shader_data
->shader_array_size
== shader_data
->num_gl_shaders
) {
4167 if (shader_data
->num_gl_shaders
)
4169 new_size
= shader_data
->shader_array_size
+ max(1, shader_data
->shader_array_size
/ 2);
4170 new_array
= HeapReAlloc(GetProcessHeap(), 0, shader_data
->gl_shaders
,
4171 new_size
* sizeof(*shader_data
->gl_shaders
));
4173 new_array
= HeapAlloc(GetProcessHeap(), 0, sizeof(*shader_data
->gl_shaders
));
4178 ERR("Out of memory\n");
4181 shader_data
->gl_shaders
= new_array
;
4182 shader_data
->shader_array_size
= new_size
;
4185 shader_data
->gl_shaders
[shader_data
->num_gl_shaders
].args
= *args
;
4187 memset(&shader_data
->gl_shaders
[shader_data
->num_gl_shaders
].np2fixup
, 0, sizeof(struct ps_np2fixup_info
));
4188 if (args
->np2_fixup
) np2fixup
= &shader_data
->gl_shaders
[shader_data
->num_gl_shaders
].np2fixup
;
4190 pixelshader_update_samplers(&shader
->baseShader
.reg_maps
, (IWineD3DBaseTexture
**)state
->textures
);
4192 shader_buffer_clear(buffer
);
4193 ret
= shader_glsl_generate_pshader(context
, buffer
, shader
, args
, np2fixup
);
4194 shader_data
->gl_shaders
[shader_data
->num_gl_shaders
++].prgId
= ret
;
4195 *np2fixup_info
= np2fixup
;
4200 static inline BOOL
vs_args_equal(const struct vs_compile_args
*stored
, const struct vs_compile_args
*new,
4201 const DWORD use_map
) {
4202 if((stored
->swizzle_map
& use_map
) != new->swizzle_map
) return FALSE
;
4203 if((stored
->clip_enabled
) != new->clip_enabled
) return FALSE
;
4204 return stored
->fog_src
== new->fog_src
;
4207 static GLhandleARB
find_glsl_vshader(const struct wined3d_context
*context
,
4208 struct wined3d_shader_buffer
*buffer
, IWineD3DVertexShaderImpl
*shader
,
4209 const struct vs_compile_args
*args
)
4213 struct glsl_vs_compiled_shader
*new_array
;
4214 DWORD use_map
= ((IWineD3DDeviceImpl
*)shader
->baseShader
.device
)->strided_streams
.use_map
;
4215 struct glsl_vshader_private
*shader_data
;
4218 if (!shader
->baseShader
.backend_data
)
4220 shader
->baseShader
.backend_data
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*shader_data
));
4221 if (!shader
->baseShader
.backend_data
)
4223 ERR("Failed to allocate backend data.\n");
4227 shader_data
= shader
->baseShader
.backend_data
;
4229 /* Usually we have very few GL shaders for each d3d shader(just 1 or maybe 2),
4230 * so a linear search is more performant than a hashmap or a binary search
4231 * (cache coherency etc)
4233 for(i
= 0; i
< shader_data
->num_gl_shaders
; i
++) {
4234 if(vs_args_equal(&shader_data
->gl_shaders
[i
].args
, args
, use_map
)) {
4235 return shader_data
->gl_shaders
[i
].prgId
;
4239 TRACE("No matching GL shader found for shader %p, compiling a new shader.\n", shader
);
4241 if(shader_data
->shader_array_size
== shader_data
->num_gl_shaders
) {
4242 if (shader_data
->num_gl_shaders
)
4244 new_size
= shader_data
->shader_array_size
+ max(1, shader_data
->shader_array_size
/ 2);
4245 new_array
= HeapReAlloc(GetProcessHeap(), 0, shader_data
->gl_shaders
,
4246 new_size
* sizeof(*shader_data
->gl_shaders
));
4248 new_array
= HeapAlloc(GetProcessHeap(), 0, sizeof(*shader_data
->gl_shaders
));
4253 ERR("Out of memory\n");
4256 shader_data
->gl_shaders
= new_array
;
4257 shader_data
->shader_array_size
= new_size
;
4260 shader_data
->gl_shaders
[shader_data
->num_gl_shaders
].args
= *args
;
4262 shader_buffer_clear(buffer
);
4263 ret
= shader_glsl_generate_vshader(context
, buffer
, shader
, args
);
4264 shader_data
->gl_shaders
[shader_data
->num_gl_shaders
++].prgId
= ret
;
4269 /** Sets the GLSL program ID for the given pixel and vertex shader combination.
4270 * It sets the programId on the current StateBlock (because it should be called
4271 * inside of the DrawPrimitive() part of the render loop).
4273 * If a program for the given combination does not exist, create one, and store
4274 * the program in the hash table. If it creates a program, it will link the
4275 * given objects, too.
4278 /* GL locking is done by the caller */
4279 static void set_glsl_shader_program(const struct wined3d_context
*context
,
4280 IWineD3DDeviceImpl
*device
, BOOL use_ps
, BOOL use_vs
)
4282 const struct wined3d_state
*state
= &device
->stateBlock
->state
;
4283 IWineD3DVertexShader
*vshader
= use_vs
? (IWineD3DVertexShader
*)state
->vertex_shader
: NULL
;
4284 IWineD3DPixelShader
*pshader
= use_ps
? (IWineD3DPixelShader
*)state
->pixel_shader
: NULL
;
4285 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
4286 struct shader_glsl_priv
*priv
= device
->shader_priv
;
4287 struct glsl_shader_prog_link
*entry
= NULL
;
4288 GLhandleARB programId
= 0;
4289 GLhandleARB reorder_shader_id
= 0;
4292 struct ps_compile_args ps_compile_args
;
4293 struct vs_compile_args vs_compile_args
;
4295 if (vshader
) find_vs_compile_args(state
, (IWineD3DVertexShaderImpl
*)vshader
, &vs_compile_args
);
4296 if (pshader
) find_ps_compile_args(state
, (IWineD3DPixelShaderImpl
*)pshader
, &ps_compile_args
);
4298 entry
= get_glsl_program_entry(priv
, vshader
, pshader
, &vs_compile_args
, &ps_compile_args
);
4300 priv
->glsl_program
= entry
;
4304 /* If we get to this point, then no matching program exists, so we create one */
4305 programId
= GL_EXTCALL(glCreateProgramObjectARB());
4306 TRACE("Created new GLSL shader program %u\n", programId
);
4308 /* Create the entry */
4309 entry
= HeapAlloc(GetProcessHeap(), 0, sizeof(struct glsl_shader_prog_link
));
4310 entry
->programId
= programId
;
4311 entry
->vshader
= vshader
;
4312 entry
->pshader
= pshader
;
4313 entry
->vs_args
= vs_compile_args
;
4314 entry
->ps_args
= ps_compile_args
;
4315 entry
->constant_version
= 0;
4316 entry
->np2Fixup_info
= NULL
;
4317 /* Add the hash table entry */
4318 add_glsl_program_entry(priv
, entry
);
4320 /* Set the current program */
4321 priv
->glsl_program
= entry
;
4323 /* Attach GLSL vshader */
4326 GLhandleARB vshader_id
= find_glsl_vshader(context
, &priv
->shader_buffer
,
4327 (IWineD3DVertexShaderImpl
*)vshader
, &vs_compile_args
);
4328 WORD map
= ((IWineD3DBaseShaderImpl
*)vshader
)->baseShader
.reg_maps
.input_registers
;
4331 reorder_shader_id
= generate_param_reorder_function(&priv
->shader_buffer
, vshader
, pshader
, gl_info
);
4332 TRACE("Attaching GLSL shader object %u to program %u\n", reorder_shader_id
, programId
);
4333 GL_EXTCALL(glAttachObjectARB(programId
, reorder_shader_id
));
4334 checkGLcall("glAttachObjectARB");
4335 /* Flag the reorder function for deletion, then it will be freed automatically when the program
4338 GL_EXTCALL(glDeleteObjectARB(reorder_shader_id
));
4340 TRACE("Attaching GLSL shader object %u to program %u\n", vshader_id
, programId
);
4341 GL_EXTCALL(glAttachObjectARB(programId
, vshader_id
));
4342 checkGLcall("glAttachObjectARB");
4344 /* Bind vertex attributes to a corresponding index number to match
4345 * the same index numbers as ARB_vertex_programs (makes loading
4346 * vertex attributes simpler). With this method, we can use the
4347 * exact same code to load the attributes later for both ARB and
4350 * We have to do this here because we need to know the Program ID
4351 * in order to make the bindings work, and it has to be done prior
4352 * to linking the GLSL program. */
4353 for (i
= 0; map
; map
>>= 1, ++i
)
4355 if (!(map
& 1)) continue;
4357 snprintf(tmp_name
, sizeof(tmp_name
), "attrib%u", i
);
4358 GL_EXTCALL(glBindAttribLocationARB(programId
, i
, tmp_name
));
4360 checkGLcall("glBindAttribLocationARB");
4362 list_add_head(&((IWineD3DBaseShaderImpl
*)vshader
)->baseShader
.linked_programs
, &entry
->vshader_entry
);
4365 /* Attach GLSL pshader */
4368 GLhandleARB pshader_id
= find_glsl_pshader(context
, &priv
->shader_buffer
,
4369 (IWineD3DPixelShaderImpl
*)pshader
, &ps_compile_args
, &entry
->np2Fixup_info
);
4370 TRACE("Attaching GLSL shader object %u to program %u\n", pshader_id
, programId
);
4371 GL_EXTCALL(glAttachObjectARB(programId
, pshader_id
));
4372 checkGLcall("glAttachObjectARB");
4374 list_add_head(&((IWineD3DBaseShaderImpl
*)pshader
)->baseShader
.linked_programs
, &entry
->pshader_entry
);
4377 /* Link the program */
4378 TRACE("Linking GLSL shader program %u\n", programId
);
4379 GL_EXTCALL(glLinkProgramARB(programId
));
4380 shader_glsl_validate_link(gl_info
, programId
);
4382 entry
->vuniformF_locations
= HeapAlloc(GetProcessHeap(), 0,
4383 sizeof(GLhandleARB
) * gl_info
->limits
.glsl_vs_float_constants
);
4384 for (i
= 0; i
< gl_info
->limits
.glsl_vs_float_constants
; ++i
)
4386 snprintf(glsl_name
, sizeof(glsl_name
), "VC[%i]", i
);
4387 entry
->vuniformF_locations
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, glsl_name
));
4389 for (i
= 0; i
< MAX_CONST_I
; ++i
)
4391 snprintf(glsl_name
, sizeof(glsl_name
), "VI[%i]", i
);
4392 entry
->vuniformI_locations
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, glsl_name
));
4394 entry
->puniformF_locations
= HeapAlloc(GetProcessHeap(), 0,
4395 sizeof(GLhandleARB
) * gl_info
->limits
.glsl_ps_float_constants
);
4396 for (i
= 0; i
< gl_info
->limits
.glsl_ps_float_constants
; ++i
)
4398 snprintf(glsl_name
, sizeof(glsl_name
), "PC[%i]", i
);
4399 entry
->puniformF_locations
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, glsl_name
));
4401 for (i
= 0; i
< MAX_CONST_I
; ++i
)
4403 snprintf(glsl_name
, sizeof(glsl_name
), "PI[%i]", i
);
4404 entry
->puniformI_locations
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, glsl_name
));
4410 for(i
= 0; i
< MAX_TEXTURES
; i
++) {
4411 sprintf(name
, "bumpenvmat%u", i
);
4412 entry
->bumpenvmat_location
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, name
));
4413 sprintf(name
, "luminancescale%u", i
);
4414 entry
->luminancescale_location
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, name
));
4415 sprintf(name
, "luminanceoffset%u", i
);
4416 entry
->luminanceoffset_location
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, name
));
4419 if (ps_compile_args
.np2_fixup
) {
4420 if (entry
->np2Fixup_info
) {
4421 entry
->np2Fixup_location
= GL_EXTCALL(glGetUniformLocationARB(programId
, "PsamplerNP2Fixup"));
4423 FIXME("NP2 texcoord fixup needed for this pixelshader, but no fixup uniform found.\n");
4428 entry
->posFixup_location
= GL_EXTCALL(glGetUniformLocationARB(programId
, "posFixup"));
4429 entry
->ycorrection_location
= GL_EXTCALL(glGetUniformLocationARB(programId
, "ycorrection"));
4430 checkGLcall("Find glsl program uniform locations");
4433 && ((IWineD3DPixelShaderImpl
*)pshader
)->baseShader
.reg_maps
.shader_version
.major
>= 3
4434 && ((IWineD3DPixelShaderImpl
*)pshader
)->declared_in_count
> vec4_varyings(3, gl_info
))
4436 TRACE("Shader %d needs vertex color clamping disabled\n", programId
);
4437 entry
->vertex_color_clamp
= GL_FALSE
;
4439 entry
->vertex_color_clamp
= GL_FIXED_ONLY_ARB
;
4442 /* Set the shader to allow uniform loading on it */
4443 GL_EXTCALL(glUseProgramObjectARB(programId
));
4444 checkGLcall("glUseProgramObjectARB(programId)");
4446 /* Load the vertex and pixel samplers now. The function that finds the mappings makes sure
4447 * that it stays the same for each vertexshader-pixelshader pair(=linked glsl program). If
4448 * a pshader with fixed function pipeline is used there are no vertex samplers, and if a
4449 * vertex shader with fixed function pixel processing is used we make sure that the card
4450 * supports enough samplers to allow the max number of vertex samplers with all possible
4451 * fixed function fragment processing setups. So once the program is linked these samplers
4454 if (vshader
) shader_glsl_load_vsamplers(gl_info
, device
->texUnitMap
, programId
);
4455 if (pshader
) shader_glsl_load_psamplers(gl_info
, device
->texUnitMap
, programId
);
4457 /* If the local constants do not have to be loaded with the environment constants,
4458 * load them now to have them hardcoded in the GLSL program. This saves some CPU cycles
4461 if (pshader
&& !((IWineD3DBaseShaderImpl
*)pshader
)->baseShader
.load_local_constsF
)
4463 hardcode_local_constants((IWineD3DBaseShaderImpl
*) pshader
, gl_info
, programId
, 'P');
4465 if (vshader
&& !((IWineD3DBaseShaderImpl
*)vshader
)->baseShader
.load_local_constsF
)
4467 hardcode_local_constants((IWineD3DBaseShaderImpl
*) vshader
, gl_info
, programId
, 'V');
4471 /* GL locking is done by the caller */
4472 static GLhandleARB
create_glsl_blt_shader(const struct wined3d_gl_info
*gl_info
, enum tex_types tex_type
, BOOL masked
)
4474 GLhandleARB program_id
;
4475 GLhandleARB vshader_id
, pshader_id
;
4476 const char *blt_pshader
;
4478 static const char *blt_vshader
=
4482 " gl_Position = gl_Vertex;\n"
4483 " gl_FrontColor = vec4(1.0);\n"
4484 " gl_TexCoord[0] = gl_MultiTexCoord0;\n"
4487 static const char * const blt_pshaders_full
[tex_type_count
] =
4493 "uniform sampler2D sampler;\n"
4496 " gl_FragDepth = texture2D(sampler, gl_TexCoord[0].xy).x;\n"
4502 "uniform samplerCube sampler;\n"
4505 " gl_FragDepth = textureCube(sampler, gl_TexCoord[0].xyz).x;\n"
4509 "#extension GL_ARB_texture_rectangle : enable\n"
4510 "uniform sampler2DRect sampler;\n"
4513 " gl_FragDepth = texture2DRect(sampler, gl_TexCoord[0].xy).x;\n"
4517 static const char * const blt_pshaders_masked
[tex_type_count
] =
4523 "uniform sampler2D sampler;\n"
4524 "uniform vec4 mask;\n"
4527 " if (all(lessThan(gl_FragCoord.xy, mask.zw))) discard;\n"
4528 " gl_FragDepth = texture2D(sampler, gl_TexCoord[0].xy).x;\n"
4534 "uniform samplerCube sampler;\n"
4535 "uniform vec4 mask;\n"
4538 " if (all(lessThan(gl_FragCoord.xy, mask.zw))) discard;\n"
4539 " gl_FragDepth = textureCube(sampler, gl_TexCoord[0].xyz).x;\n"
4543 "#extension GL_ARB_texture_rectangle : enable\n"
4544 "uniform sampler2DRect sampler;\n"
4545 "uniform vec4 mask;\n"
4548 " if (all(lessThan(gl_FragCoord.xy, mask.zw))) discard;\n"
4549 " gl_FragDepth = texture2DRect(sampler, gl_TexCoord[0].xy).x;\n"
4553 blt_pshader
= masked
? blt_pshaders_masked
[tex_type
] : blt_pshaders_full
[tex_type
];
4556 FIXME("tex_type %#x not supported\n", tex_type
);
4560 vshader_id
= GL_EXTCALL(glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB
));
4561 shader_glsl_compile(gl_info
, vshader_id
, blt_vshader
);
4563 pshader_id
= GL_EXTCALL(glCreateShaderObjectARB(GL_FRAGMENT_SHADER_ARB
));
4564 shader_glsl_compile(gl_info
, pshader_id
, blt_pshader
);
4566 program_id
= GL_EXTCALL(glCreateProgramObjectARB());
4567 GL_EXTCALL(glAttachObjectARB(program_id
, vshader_id
));
4568 GL_EXTCALL(glAttachObjectARB(program_id
, pshader_id
));
4569 GL_EXTCALL(glLinkProgramARB(program_id
));
4571 shader_glsl_validate_link(gl_info
, program_id
);
4573 /* Once linked we can mark the shaders for deletion. They will be deleted once the program
4576 GL_EXTCALL(glDeleteObjectARB(vshader_id
));
4577 GL_EXTCALL(glDeleteObjectARB(pshader_id
));
4581 /* GL locking is done by the caller */
4582 static void shader_glsl_select(const struct wined3d_context
*context
, BOOL usePS
, BOOL useVS
)
4584 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
4585 IWineD3DDeviceImpl
*device
= context
->swapchain
->device
;
4586 struct shader_glsl_priv
*priv
= device
->shader_priv
;
4587 GLhandleARB program_id
= 0;
4588 GLenum old_vertex_color_clamp
, current_vertex_color_clamp
;
4590 old_vertex_color_clamp
= priv
->glsl_program
? priv
->glsl_program
->vertex_color_clamp
: GL_FIXED_ONLY_ARB
;
4592 if (useVS
|| usePS
) set_glsl_shader_program(context
, device
, usePS
, useVS
);
4593 else priv
->glsl_program
= NULL
;
4595 current_vertex_color_clamp
= priv
->glsl_program
? priv
->glsl_program
->vertex_color_clamp
: GL_FIXED_ONLY_ARB
;
4597 if (old_vertex_color_clamp
!= current_vertex_color_clamp
)
4599 if (gl_info
->supported
[ARB_COLOR_BUFFER_FLOAT
])
4601 GL_EXTCALL(glClampColorARB(GL_CLAMP_VERTEX_COLOR_ARB
, current_vertex_color_clamp
));
4602 checkGLcall("glClampColorARB");
4606 FIXME("vertex color clamp needs to be changed, but extension not supported.\n");
4610 program_id
= priv
->glsl_program
? priv
->glsl_program
->programId
: 0;
4611 if (program_id
) TRACE("Using GLSL program %u\n", program_id
);
4612 GL_EXTCALL(glUseProgramObjectARB(program_id
));
4613 checkGLcall("glUseProgramObjectARB");
4615 /* In case that NP2 texcoord fixup data is found for the selected program, trigger a reload of the
4616 * constants. This has to be done because it can't be guaranteed that sampler() (from state.c) is
4617 * called between selecting the shader and using it, which results in wrong fixup for some frames. */
4618 if (priv
->glsl_program
&& priv
->glsl_program
->np2Fixup_info
)
4620 shader_glsl_load_np2fixup_constants((IWineD3DDevice
*)device
, usePS
, useVS
);
4624 /* GL locking is done by the caller */
4625 static void shader_glsl_select_depth_blt(IWineD3DDevice
*iface
,
4626 enum tex_types tex_type
, const SIZE
*ds_mask_size
)
4628 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
4629 const struct wined3d_gl_info
*gl_info
= &This
->adapter
->gl_info
;
4630 BOOL masked
= ds_mask_size
->cx
&& ds_mask_size
->cy
;
4631 struct shader_glsl_priv
*priv
= This
->shader_priv
;
4632 GLhandleARB
*blt_program
;
4635 blt_program
= masked
? &priv
->depth_blt_program_masked
[tex_type
] : &priv
->depth_blt_program_full
[tex_type
];
4638 *blt_program
= create_glsl_blt_shader(gl_info
, tex_type
, masked
);
4639 loc
= GL_EXTCALL(glGetUniformLocationARB(*blt_program
, "sampler"));
4640 GL_EXTCALL(glUseProgramObjectARB(*blt_program
));
4641 GL_EXTCALL(glUniform1iARB(loc
, 0));
4645 GL_EXTCALL(glUseProgramObjectARB(*blt_program
));
4650 loc
= GL_EXTCALL(glGetUniformLocationARB(*blt_program
, "mask"));
4651 GL_EXTCALL(glUniform4fARB(loc
, 0.0f
, 0.0f
, (float)ds_mask_size
->cx
, (float)ds_mask_size
->cy
));
4655 /* GL locking is done by the caller */
4656 static void shader_glsl_deselect_depth_blt(IWineD3DDevice
*iface
) {
4657 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
4658 const struct wined3d_gl_info
*gl_info
= &This
->adapter
->gl_info
;
4659 struct shader_glsl_priv
*priv
= This
->shader_priv
;
4660 GLhandleARB program_id
;
4662 program_id
= priv
->glsl_program
? priv
->glsl_program
->programId
: 0;
4663 if (program_id
) TRACE("Using GLSL program %u\n", program_id
);
4665 GL_EXTCALL(glUseProgramObjectARB(program_id
));
4666 checkGLcall("glUseProgramObjectARB");
4669 static void shader_glsl_destroy(IWineD3DBaseShader
*iface
) {
4670 const struct list
*linked_programs
;
4671 IWineD3DBaseShaderImpl
*This
= (IWineD3DBaseShaderImpl
*) iface
;
4672 IWineD3DDeviceImpl
*device
= (IWineD3DDeviceImpl
*)This
->baseShader
.device
;
4673 struct shader_glsl_priv
*priv
= device
->shader_priv
;
4674 const struct wined3d_gl_info
*gl_info
;
4675 struct wined3d_context
*context
;
4677 /* Note: Do not use QueryInterface here to find out which shader type this is because this code
4678 * can be called from IWineD3DBaseShader::Release
4680 char pshader
= shader_is_pshader_version(This
->baseShader
.reg_maps
.shader_version
.type
);
4683 struct glsl_pshader_private
*shader_data
;
4684 shader_data
= This
->baseShader
.backend_data
;
4685 if (!shader_data
|| !shader_data
->num_gl_shaders
)
4687 HeapFree(GetProcessHeap(), 0, shader_data
);
4688 This
->baseShader
.backend_data
= NULL
;
4692 context
= context_acquire(device
, NULL
);
4693 gl_info
= context
->gl_info
;
4695 if (priv
->glsl_program
&& (IWineD3DBaseShader
*)priv
->glsl_program
->pshader
== iface
)
4698 shader_glsl_select(context
, FALSE
, FALSE
);
4702 struct glsl_vshader_private
*shader_data
;
4703 shader_data
= This
->baseShader
.backend_data
;
4704 if (!shader_data
|| !shader_data
->num_gl_shaders
)
4706 HeapFree(GetProcessHeap(), 0, shader_data
);
4707 This
->baseShader
.backend_data
= NULL
;
4711 context
= context_acquire(device
, NULL
);
4712 gl_info
= context
->gl_info
;
4714 if (priv
->glsl_program
&& (IWineD3DBaseShader
*)priv
->glsl_program
->vshader
== iface
)
4717 shader_glsl_select(context
, FALSE
, FALSE
);
4722 linked_programs
= &This
->baseShader
.linked_programs
;
4724 TRACE("Deleting linked programs\n");
4725 if (linked_programs
->next
) {
4726 struct glsl_shader_prog_link
*entry
, *entry2
;
4730 LIST_FOR_EACH_ENTRY_SAFE(entry
, entry2
, linked_programs
, struct glsl_shader_prog_link
, pshader_entry
) {
4731 delete_glsl_program_entry(priv
, gl_info
, entry
);
4734 LIST_FOR_EACH_ENTRY_SAFE(entry
, entry2
, linked_programs
, struct glsl_shader_prog_link
, vshader_entry
) {
4735 delete_glsl_program_entry(priv
, gl_info
, entry
);
4743 struct glsl_pshader_private
*shader_data
= This
->baseShader
.backend_data
;
4746 for(i
= 0; i
< shader_data
->num_gl_shaders
; i
++) {
4747 TRACE("deleting pshader %u\n", shader_data
->gl_shaders
[i
].prgId
);
4748 GL_EXTCALL(glDeleteObjectARB(shader_data
->gl_shaders
[i
].prgId
));
4749 checkGLcall("glDeleteObjectARB");
4752 HeapFree(GetProcessHeap(), 0, shader_data
->gl_shaders
);
4757 struct glsl_vshader_private
*shader_data
= This
->baseShader
.backend_data
;
4760 for(i
= 0; i
< shader_data
->num_gl_shaders
; i
++) {
4761 TRACE("deleting vshader %u\n", shader_data
->gl_shaders
[i
].prgId
);
4762 GL_EXTCALL(glDeleteObjectARB(shader_data
->gl_shaders
[i
].prgId
));
4763 checkGLcall("glDeleteObjectARB");
4766 HeapFree(GetProcessHeap(), 0, shader_data
->gl_shaders
);
4769 HeapFree(GetProcessHeap(), 0, This
->baseShader
.backend_data
);
4770 This
->baseShader
.backend_data
= NULL
;
4772 context_release(context
);
4775 static int glsl_program_key_compare(const void *key
, const struct wine_rb_entry
*entry
)
4777 const glsl_program_key_t
*k
= key
;
4778 const struct glsl_shader_prog_link
*prog
= WINE_RB_ENTRY_VALUE(entry
,
4779 const struct glsl_shader_prog_link
, program_lookup_entry
);
4782 if (k
->vshader
> prog
->vshader
) return 1;
4783 else if (k
->vshader
< prog
->vshader
) return -1;
4785 if (k
->pshader
> prog
->pshader
) return 1;
4786 else if (k
->pshader
< prog
->pshader
) return -1;
4788 if (k
->vshader
&& (cmp
= memcmp(&k
->vs_args
, &prog
->vs_args
, sizeof(prog
->vs_args
)))) return cmp
;
4789 if (k
->pshader
&& (cmp
= memcmp(&k
->ps_args
, &prog
->ps_args
, sizeof(prog
->ps_args
)))) return cmp
;
4794 static BOOL
constant_heap_init(struct constant_heap
*heap
, unsigned int constant_count
)
4796 SIZE_T size
= (constant_count
+ 1) * sizeof(*heap
->entries
) + constant_count
* sizeof(*heap
->positions
);
4797 void *mem
= HeapAlloc(GetProcessHeap(), 0, size
);
4801 ERR("Failed to allocate memory\n");
4805 heap
->entries
= mem
;
4806 heap
->entries
[1].version
= 0;
4807 heap
->positions
= (unsigned int *)(heap
->entries
+ constant_count
+ 1);
4813 static void constant_heap_free(struct constant_heap
*heap
)
4815 HeapFree(GetProcessHeap(), 0, heap
->entries
);
4818 static const struct wine_rb_functions wined3d_glsl_program_rb_functions
=
4823 glsl_program_key_compare
,
4826 static HRESULT
shader_glsl_alloc(IWineD3DDevice
*iface
) {
4827 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
4828 const struct wined3d_gl_info
*gl_info
= &This
->adapter
->gl_info
;
4829 struct shader_glsl_priv
*priv
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(struct shader_glsl_priv
));
4830 SIZE_T stack_size
= wined3d_log2i(max(gl_info
->limits
.glsl_vs_float_constants
,
4831 gl_info
->limits
.glsl_ps_float_constants
)) + 1;
4833 if (!shader_buffer_init(&priv
->shader_buffer
))
4835 ERR("Failed to initialize shader buffer.\n");
4839 priv
->stack
= HeapAlloc(GetProcessHeap(), 0, stack_size
* sizeof(*priv
->stack
));
4842 ERR("Failed to allocate memory.\n");
4846 if (!constant_heap_init(&priv
->vconst_heap
, gl_info
->limits
.glsl_vs_float_constants
))
4848 ERR("Failed to initialize vertex shader constant heap\n");
4852 if (!constant_heap_init(&priv
->pconst_heap
, gl_info
->limits
.glsl_ps_float_constants
))
4854 ERR("Failed to initialize pixel shader constant heap\n");
4858 if (wine_rb_init(&priv
->program_lookup
, &wined3d_glsl_program_rb_functions
) == -1)
4860 ERR("Failed to initialize rbtree.\n");
4864 priv
->next_constant_version
= 1;
4866 This
->shader_priv
= priv
;
4870 constant_heap_free(&priv
->pconst_heap
);
4871 constant_heap_free(&priv
->vconst_heap
);
4872 HeapFree(GetProcessHeap(), 0, priv
->stack
);
4873 shader_buffer_free(&priv
->shader_buffer
);
4874 HeapFree(GetProcessHeap(), 0, priv
);
4875 return E_OUTOFMEMORY
;
4878 /* Context activation is done by the caller. */
4879 static void shader_glsl_free(IWineD3DDevice
*iface
) {
4880 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
4881 const struct wined3d_gl_info
*gl_info
= &This
->adapter
->gl_info
;
4882 struct shader_glsl_priv
*priv
= This
->shader_priv
;
4886 for (i
= 0; i
< tex_type_count
; ++i
)
4888 if (priv
->depth_blt_program_full
[i
])
4890 GL_EXTCALL(glDeleteObjectARB(priv
->depth_blt_program_full
[i
]));
4892 if (priv
->depth_blt_program_masked
[i
])
4894 GL_EXTCALL(glDeleteObjectARB(priv
->depth_blt_program_masked
[i
]));
4899 wine_rb_destroy(&priv
->program_lookup
, NULL
, NULL
);
4900 constant_heap_free(&priv
->pconst_heap
);
4901 constant_heap_free(&priv
->vconst_heap
);
4902 HeapFree(GetProcessHeap(), 0, priv
->stack
);
4903 shader_buffer_free(&priv
->shader_buffer
);
4905 HeapFree(GetProcessHeap(), 0, This
->shader_priv
);
4906 This
->shader_priv
= NULL
;
4909 static BOOL
shader_glsl_dirty_const(IWineD3DDevice
*iface
) {
4910 /* TODO: GL_EXT_bindable_uniform can be used to share constants across shaders */
4914 static void shader_glsl_get_caps(const struct wined3d_gl_info
*gl_info
, struct shader_caps
*pCaps
)
4916 /* Nvidia Geforce6/7 or Ati R4xx/R5xx cards with GLSL support, support VS 3.0 but older Nvidia/Ati
4917 * models with GLSL support only support 2.0. In case of nvidia we can detect VS 2.0 support based
4918 * on the version of NV_vertex_program.
4919 * For Ati cards there's no way using glsl (it abstracts the lowlevel info away) and also not
4920 * using ARB_vertex_program. It is safe to assume that when a card supports pixel shader 2.0 it
4921 * supports vertex shader 2.0 too and the way around. We can detect ps2.0 using the maximum number
4922 * of native instructions, so use that here. For more info see the pixel shader versioning code below.
4924 if ((gl_info
->supported
[NV_VERTEX_PROGRAM2
] && !gl_info
->supported
[NV_VERTEX_PROGRAM3
])
4925 || gl_info
->limits
.arb_ps_instructions
<= 512)
4926 pCaps
->VertexShaderVersion
= WINED3DVS_VERSION(2,0);
4928 pCaps
->VertexShaderVersion
= WINED3DVS_VERSION(3,0);
4929 TRACE_(d3d_caps
)("Hardware vertex shader version %d.%d enabled (GLSL)\n", (pCaps
->VertexShaderVersion
>> 8) & 0xff, pCaps
->VertexShaderVersion
& 0xff);
4930 pCaps
->MaxVertexShaderConst
= gl_info
->limits
.glsl_vs_float_constants
;
4932 /* Older DX9-class videocards (GeforceFX / Radeon >9500/X*00) only support pixel shader 2.0/2.0a/2.0b.
4933 * In OpenGL the extensions related to GLSL abstract lowlevel GL info away which is needed
4934 * to distinguish between 2.0 and 3.0 (and 2.0a/2.0b). In case of Nvidia we use their fragment
4935 * program extensions. On other hardware including ATI GL_ARB_fragment_program offers the info
4936 * in max native instructions. Intel and others also offer the info in this extension but they
4937 * don't support GLSL (at least on Windows).
4939 * PS2.0 requires at least 96 instructions, 2.0a/2.0b go up to 512. Assume that if the number
4940 * of instructions is 512 or less we have to do with ps2.0 hardware.
4941 * 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.
4943 if ((gl_info
->supported
[NV_FRAGMENT_PROGRAM
] && !gl_info
->supported
[NV_FRAGMENT_PROGRAM2
])
4944 || gl_info
->limits
.arb_ps_instructions
<= 512)
4945 pCaps
->PixelShaderVersion
= WINED3DPS_VERSION(2,0);
4947 pCaps
->PixelShaderVersion
= WINED3DPS_VERSION(3,0);
4949 pCaps
->MaxPixelShaderConst
= gl_info
->limits
.glsl_ps_float_constants
;
4951 /* FIXME: The following line is card dependent. -8.0 to 8.0 is the
4952 * Direct3D minimum requirement.
4954 * Both GL_ARB_fragment_program and GLSL require a "maximum representable magnitude"
4955 * of colors to be 2^10, and 2^32 for other floats. Should we use 1024 here?
4957 * The problem is that the refrast clamps temporary results in the shader to
4958 * [-MaxValue;+MaxValue]. If the card's max value is bigger than the one we advertize here,
4959 * then applications may miss the clamping behavior. On the other hand, if it is smaller,
4960 * the shader will generate incorrect results too. Unfortunately, GL deliberately doesn't
4961 * offer a way to query this.
4963 pCaps
->PixelShader1xMaxValue
= 8.0;
4964 TRACE_(d3d_caps
)("Hardware pixel shader version %d.%d enabled (GLSL)\n", (pCaps
->PixelShaderVersion
>> 8) & 0xff, pCaps
->PixelShaderVersion
& 0xff);
4966 pCaps
->VSClipping
= TRUE
;
4969 static BOOL
shader_glsl_color_fixup_supported(struct color_fixup_desc fixup
)
4971 if (TRACE_ON(d3d_shader
) && TRACE_ON(d3d
))
4973 TRACE("Checking support for fixup:\n");
4974 dump_color_fixup_desc(fixup
);
4977 /* We support everything except YUV conversions. */
4978 if (!is_complex_fixup(fixup
))
4984 TRACE("[FAILED]\n");
4988 static const SHADER_HANDLER shader_glsl_instruction_handler_table
[WINED3DSIH_TABLE_SIZE
] =
4990 /* WINED3DSIH_ABS */ shader_glsl_map2gl
,
4991 /* WINED3DSIH_ADD */ shader_glsl_arith
,
4992 /* WINED3DSIH_BEM */ shader_glsl_bem
,
4993 /* WINED3DSIH_BREAK */ shader_glsl_break
,
4994 /* WINED3DSIH_BREAKC */ shader_glsl_breakc
,
4995 /* WINED3DSIH_BREAKP */ NULL
,
4996 /* WINED3DSIH_CALL */ shader_glsl_call
,
4997 /* WINED3DSIH_CALLNZ */ shader_glsl_callnz
,
4998 /* WINED3DSIH_CMP */ shader_glsl_cmp
,
4999 /* WINED3DSIH_CND */ shader_glsl_cnd
,
5000 /* WINED3DSIH_CRS */ shader_glsl_cross
,
5001 /* WINED3DSIH_CUT */ NULL
,
5002 /* WINED3DSIH_DCL */ NULL
,
5003 /* WINED3DSIH_DEF */ NULL
,
5004 /* WINED3DSIH_DEFB */ NULL
,
5005 /* WINED3DSIH_DEFI */ 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_IADD */ NULL
,
5021 /* WINED3DSIH_IF */ shader_glsl_if
,
5022 /* WINED3DSIH_IFC */ shader_glsl_ifc
,
5023 /* WINED3DSIH_IGE */ NULL
,
5024 /* WINED3DSIH_LABEL */ shader_glsl_label
,
5025 /* WINED3DSIH_LIT */ shader_glsl_lit
,
5026 /* WINED3DSIH_LOG */ shader_glsl_log
,
5027 /* WINED3DSIH_LOGP */ shader_glsl_log
,
5028 /* WINED3DSIH_LOOP */ shader_glsl_loop
,
5029 /* WINED3DSIH_LRP */ shader_glsl_lrp
,
5030 /* WINED3DSIH_LT */ NULL
,
5031 /* WINED3DSIH_M3x2 */ shader_glsl_mnxn
,
5032 /* WINED3DSIH_M3x3 */ shader_glsl_mnxn
,
5033 /* WINED3DSIH_M3x4 */ shader_glsl_mnxn
,
5034 /* WINED3DSIH_M4x3 */ shader_glsl_mnxn
,
5035 /* WINED3DSIH_M4x4 */ shader_glsl_mnxn
,
5036 /* WINED3DSIH_MAD */ shader_glsl_mad
,
5037 /* WINED3DSIH_MAX */ shader_glsl_map2gl
,
5038 /* WINED3DSIH_MIN */ shader_glsl_map2gl
,
5039 /* WINED3DSIH_MOV */ shader_glsl_mov
,
5040 /* WINED3DSIH_MOVA */ shader_glsl_mov
,
5041 /* WINED3DSIH_MUL */ shader_glsl_arith
,
5042 /* WINED3DSIH_NOP */ NULL
,
5043 /* WINED3DSIH_NRM */ shader_glsl_nrm
,
5044 /* WINED3DSIH_PHASE */ NULL
,
5045 /* WINED3DSIH_POW */ shader_glsl_pow
,
5046 /* WINED3DSIH_RCP */ shader_glsl_rcp
,
5047 /* WINED3DSIH_REP */ shader_glsl_rep
,
5048 /* WINED3DSIH_RET */ shader_glsl_ret
,
5049 /* WINED3DSIH_RSQ */ shader_glsl_rsq
,
5050 /* WINED3DSIH_SETP */ NULL
,
5051 /* WINED3DSIH_SGE */ shader_glsl_compare
,
5052 /* WINED3DSIH_SGN */ shader_glsl_sgn
,
5053 /* WINED3DSIH_SINCOS */ shader_glsl_sincos
,
5054 /* WINED3DSIH_SLT */ shader_glsl_compare
,
5055 /* WINED3DSIH_SUB */ shader_glsl_arith
,
5056 /* WINED3DSIH_TEX */ shader_glsl_tex
,
5057 /* WINED3DSIH_TEXBEM */ shader_glsl_texbem
,
5058 /* WINED3DSIH_TEXBEML */ shader_glsl_texbem
,
5059 /* WINED3DSIH_TEXCOORD */ shader_glsl_texcoord
,
5060 /* WINED3DSIH_TEXDEPTH */ shader_glsl_texdepth
,
5061 /* WINED3DSIH_TEXDP3 */ shader_glsl_texdp3
,
5062 /* WINED3DSIH_TEXDP3TEX */ shader_glsl_texdp3tex
,
5063 /* WINED3DSIH_TEXKILL */ shader_glsl_texkill
,
5064 /* WINED3DSIH_TEXLDD */ shader_glsl_texldd
,
5065 /* WINED3DSIH_TEXLDL */ shader_glsl_texldl
,
5066 /* WINED3DSIH_TEXM3x2DEPTH */ shader_glsl_texm3x2depth
,
5067 /* WINED3DSIH_TEXM3x2PAD */ shader_glsl_texm3x2pad
,
5068 /* WINED3DSIH_TEXM3x2TEX */ shader_glsl_texm3x2tex
,
5069 /* WINED3DSIH_TEXM3x3 */ shader_glsl_texm3x3
,
5070 /* WINED3DSIH_TEXM3x3DIFF */ NULL
,
5071 /* WINED3DSIH_TEXM3x3PAD */ shader_glsl_texm3x3pad
,
5072 /* WINED3DSIH_TEXM3x3SPEC */ shader_glsl_texm3x3spec
,
5073 /* WINED3DSIH_TEXM3x3TEX */ shader_glsl_texm3x3tex
,
5074 /* WINED3DSIH_TEXM3x3VSPEC */ shader_glsl_texm3x3vspec
,
5075 /* WINED3DSIH_TEXREG2AR */ shader_glsl_texreg2ar
,
5076 /* WINED3DSIH_TEXREG2GB */ shader_glsl_texreg2gb
,
5077 /* WINED3DSIH_TEXREG2RGB */ shader_glsl_texreg2rgb
,
5080 static void shader_glsl_handle_instruction(const struct wined3d_shader_instruction
*ins
) {
5081 SHADER_HANDLER hw_fct
;
5083 /* Select handler */
5084 hw_fct
= shader_glsl_instruction_handler_table
[ins
->handler_idx
];
5086 /* Unhandled opcode */
5089 FIXME("Backend can't handle opcode %#x\n", ins
->handler_idx
);
5094 shader_glsl_add_instruction_modifiers(ins
);
5097 const shader_backend_t glsl_shader_backend
= {
5098 shader_glsl_handle_instruction
,
5100 shader_glsl_select_depth_blt
,
5101 shader_glsl_deselect_depth_blt
,
5102 shader_glsl_update_float_vertex_constants
,
5103 shader_glsl_update_float_pixel_constants
,
5104 shader_glsl_load_constants
,
5105 shader_glsl_load_np2fixup_constants
,
5106 shader_glsl_destroy
,
5109 shader_glsl_dirty_const
,
5110 shader_glsl_get_caps
,
5111 shader_glsl_color_fixup_supported
,