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
[0]); i
++) {
236 if(strcmp(infoLog
, spam
[i
]) == 0) {
245 TRACE("Spam received from GLSL shader #%u:\n", obj
);
246 while ((line
= get_info_log_line(&ptr
))) TRACE(" %s\n", line
);
250 FIXME("Error received from GLSL shader #%u:\n", obj
);
251 while ((line
= get_info_log_line(&ptr
))) FIXME(" %s\n", line
);
253 HeapFree(GetProcessHeap(), 0, infoLog
);
257 /* GL locking is done by the caller. */
258 static void shader_glsl_dump_program_source(const struct wined3d_gl_info
*gl_info
, GLhandleARB program
)
260 GLint i
, object_count
, source_size
= -1;
261 GLhandleARB
*objects
;
264 GL_EXTCALL(glGetObjectParameterivARB(program
, GL_OBJECT_ATTACHED_OBJECTS_ARB
, &object_count
));
265 objects
= HeapAlloc(GetProcessHeap(), 0, object_count
* sizeof(*objects
));
268 ERR("Failed to allocate object array memory.\n");
272 GL_EXTCALL(glGetAttachedObjectsARB(program
, object_count
, NULL
, objects
));
273 for (i
= 0; i
< object_count
; ++i
)
278 GL_EXTCALL(glGetObjectParameterivARB(objects
[i
], GL_OBJECT_SHADER_SOURCE_LENGTH_ARB
, &tmp
));
280 if (source_size
< tmp
)
282 HeapFree(GetProcessHeap(), 0, source
);
284 source
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, tmp
);
287 ERR("Failed to allocate %d bytes for shader source.\n", tmp
);
288 HeapFree(GetProcessHeap(), 0, objects
);
294 FIXME("Object %u:\n", objects
[i
]);
295 GL_EXTCALL(glGetObjectParameterivARB(objects
[i
], GL_OBJECT_SUBTYPE_ARB
, &tmp
));
296 FIXME(" GL_OBJECT_SUBTYPE_ARB: %s.\n", debug_gl_shader_type(tmp
));
297 GL_EXTCALL(glGetObjectParameterivARB(objects
[i
], GL_OBJECT_COMPILE_STATUS_ARB
, &tmp
));
298 FIXME(" GL_OBJECT_COMPILE_STATUS_ARB: %d.\n", tmp
);
302 GL_EXTCALL(glGetShaderSourceARB(objects
[i
], source_size
, NULL
, source
));
303 while ((line
= get_info_log_line(&ptr
))) FIXME(" %s\n", line
);
307 HeapFree(GetProcessHeap(), 0, source
);
308 HeapFree(GetProcessHeap(), 0, objects
);
311 /* GL locking is done by the caller. */
312 static void shader_glsl_validate_link(const struct wined3d_gl_info
*gl_info
, GLhandleARB program
)
316 if (!TRACE_ON(d3d_shader
) && !FIXME_ON(d3d_shader
)) return;
318 GL_EXTCALL(glGetObjectParameterivARB(program
, GL_OBJECT_TYPE_ARB
, &tmp
));
319 if (tmp
== GL_PROGRAM_OBJECT_ARB
)
321 GL_EXTCALL(glGetObjectParameterivARB(program
, GL_OBJECT_LINK_STATUS_ARB
, &tmp
));
324 FIXME("Program %u link status invalid.\n", program
);
325 shader_glsl_dump_program_source(gl_info
, program
);
329 print_glsl_info_log(gl_info
, program
);
333 * Loads (pixel shader) samplers
335 /* GL locking is done by the caller */
336 static void shader_glsl_load_psamplers(const struct wined3d_gl_info
*gl_info
,
337 DWORD
*tex_unit_map
, GLhandleARB programId
)
341 char sampler_name
[20];
343 for (i
= 0; i
< MAX_FRAGMENT_SAMPLERS
; ++i
) {
344 snprintf(sampler_name
, sizeof(sampler_name
), "Psampler%d", i
);
345 name_loc
= GL_EXTCALL(glGetUniformLocationARB(programId
, sampler_name
));
346 if (name_loc
!= -1) {
347 DWORD mapped_unit
= tex_unit_map
[i
];
348 if (mapped_unit
!= WINED3D_UNMAPPED_STAGE
&& mapped_unit
< gl_info
->limits
.fragment_samplers
)
350 TRACE("Loading %s for texture %d\n", sampler_name
, mapped_unit
);
351 GL_EXTCALL(glUniform1iARB(name_loc
, mapped_unit
));
352 checkGLcall("glUniform1iARB");
354 ERR("Trying to load sampler %s on unsupported unit %d\n", sampler_name
, mapped_unit
);
360 /* GL locking is done by the caller */
361 static void shader_glsl_load_vsamplers(const struct wined3d_gl_info
*gl_info
,
362 DWORD
*tex_unit_map
, GLhandleARB programId
)
365 char sampler_name
[20];
368 for (i
= 0; i
< MAX_VERTEX_SAMPLERS
; ++i
) {
369 snprintf(sampler_name
, sizeof(sampler_name
), "Vsampler%d", i
);
370 name_loc
= GL_EXTCALL(glGetUniformLocationARB(programId
, sampler_name
));
371 if (name_loc
!= -1) {
372 DWORD mapped_unit
= tex_unit_map
[MAX_FRAGMENT_SAMPLERS
+ i
];
373 if (mapped_unit
!= WINED3D_UNMAPPED_STAGE
&& mapped_unit
< gl_info
->limits
.combined_samplers
)
375 TRACE("Loading %s for texture %d\n", sampler_name
, mapped_unit
);
376 GL_EXTCALL(glUniform1iARB(name_loc
, mapped_unit
));
377 checkGLcall("glUniform1iARB");
379 ERR("Trying to load sampler %s on unsupported unit %d\n", sampler_name
, mapped_unit
);
385 /* GL locking is done by the caller */
386 static inline void walk_constant_heap(const struct wined3d_gl_info
*gl_info
, const float *constants
,
387 const GLint
*constant_locations
, const struct constant_heap
*heap
, unsigned char *stack
, DWORD version
)
390 unsigned int heap_idx
= 1;
393 if (heap
->entries
[heap_idx
].version
<= version
) return;
395 idx
= heap
->entries
[heap_idx
].idx
;
396 if (constant_locations
[idx
] != -1) GL_EXTCALL(glUniform4fvARB(constant_locations
[idx
], 1, &constants
[idx
* 4]));
397 stack
[stack_idx
] = HEAP_NODE_TRAVERSE_LEFT
;
399 while (stack_idx
>= 0)
401 /* Note that we fall through to the next case statement. */
402 switch(stack
[stack_idx
])
404 case HEAP_NODE_TRAVERSE_LEFT
:
406 unsigned int left_idx
= heap_idx
<< 1;
407 if (left_idx
< heap
->size
&& heap
->entries
[left_idx
].version
> version
)
410 idx
= heap
->entries
[heap_idx
].idx
;
411 if (constant_locations
[idx
] != -1)
412 GL_EXTCALL(glUniform4fvARB(constant_locations
[idx
], 1, &constants
[idx
* 4]));
414 stack
[stack_idx
++] = HEAP_NODE_TRAVERSE_RIGHT
;
415 stack
[stack_idx
] = HEAP_NODE_TRAVERSE_LEFT
;
420 case HEAP_NODE_TRAVERSE_RIGHT
:
422 unsigned int right_idx
= (heap_idx
<< 1) + 1;
423 if (right_idx
< heap
->size
&& heap
->entries
[right_idx
].version
> version
)
425 heap_idx
= right_idx
;
426 idx
= heap
->entries
[heap_idx
].idx
;
427 if (constant_locations
[idx
] != -1)
428 GL_EXTCALL(glUniform4fvARB(constant_locations
[idx
], 1, &constants
[idx
* 4]));
430 stack
[stack_idx
++] = HEAP_NODE_POP
;
431 stack
[stack_idx
] = HEAP_NODE_TRAVERSE_LEFT
;
444 checkGLcall("walk_constant_heap()");
447 /* GL locking is done by the caller */
448 static inline void apply_clamped_constant(const struct wined3d_gl_info
*gl_info
, GLint location
, const GLfloat
*data
)
450 GLfloat clamped_constant
[4];
452 if (location
== -1) return;
454 clamped_constant
[0] = data
[0] < -1.0f
? -1.0f
: data
[0] > 1.0f
? 1.0f
: data
[0];
455 clamped_constant
[1] = data
[1] < -1.0f
? -1.0f
: data
[1] > 1.0f
? 1.0f
: data
[1];
456 clamped_constant
[2] = data
[2] < -1.0f
? -1.0f
: data
[2] > 1.0f
? 1.0f
: data
[2];
457 clamped_constant
[3] = data
[3] < -1.0f
? -1.0f
: data
[3] > 1.0f
? 1.0f
: data
[3];
459 GL_EXTCALL(glUniform4fvARB(location
, 1, clamped_constant
));
462 /* GL locking is done by the caller */
463 static inline void walk_constant_heap_clamped(const struct wined3d_gl_info
*gl_info
, const float *constants
,
464 const GLint
*constant_locations
, const struct constant_heap
*heap
, unsigned char *stack
, DWORD version
)
467 unsigned int heap_idx
= 1;
470 if (heap
->entries
[heap_idx
].version
<= version
) return;
472 idx
= heap
->entries
[heap_idx
].idx
;
473 apply_clamped_constant(gl_info
, constant_locations
[idx
], &constants
[idx
* 4]);
474 stack
[stack_idx
] = HEAP_NODE_TRAVERSE_LEFT
;
476 while (stack_idx
>= 0)
478 /* Note that we fall through to the next case statement. */
479 switch(stack
[stack_idx
])
481 case HEAP_NODE_TRAVERSE_LEFT
:
483 unsigned int left_idx
= heap_idx
<< 1;
484 if (left_idx
< heap
->size
&& heap
->entries
[left_idx
].version
> version
)
487 idx
= heap
->entries
[heap_idx
].idx
;
488 apply_clamped_constant(gl_info
, constant_locations
[idx
], &constants
[idx
* 4]);
490 stack
[stack_idx
++] = HEAP_NODE_TRAVERSE_RIGHT
;
491 stack
[stack_idx
] = HEAP_NODE_TRAVERSE_LEFT
;
496 case HEAP_NODE_TRAVERSE_RIGHT
:
498 unsigned int right_idx
= (heap_idx
<< 1) + 1;
499 if (right_idx
< heap
->size
&& heap
->entries
[right_idx
].version
> version
)
501 heap_idx
= right_idx
;
502 idx
= heap
->entries
[heap_idx
].idx
;
503 apply_clamped_constant(gl_info
, constant_locations
[idx
], &constants
[idx
* 4]);
505 stack
[stack_idx
++] = HEAP_NODE_POP
;
506 stack
[stack_idx
] = HEAP_NODE_TRAVERSE_LEFT
;
519 checkGLcall("walk_constant_heap_clamped()");
522 /* Loads floating point constants (aka uniforms) into the currently set GLSL program. */
523 /* GL locking is done by the caller */
524 static void shader_glsl_load_constantsF(IWineD3DBaseShaderImpl
*This
, const struct wined3d_gl_info
*gl_info
,
525 const float *constants
, const GLint
*constant_locations
, const struct constant_heap
*heap
,
526 unsigned char *stack
, UINT version
)
528 const local_constant
*lconst
;
530 /* 1.X pshaders have the constants clamped to [-1;1] implicitly. */
531 if (This
->baseShader
.reg_maps
.shader_version
.major
== 1
532 && shader_is_pshader_version(This
->baseShader
.reg_maps
.shader_version
.type
))
533 walk_constant_heap_clamped(gl_info
, constants
, constant_locations
, heap
, stack
, version
);
535 walk_constant_heap(gl_info
, constants
, constant_locations
, heap
, stack
, version
);
537 if (!This
->baseShader
.load_local_constsF
)
539 TRACE("No need to load local float constants for this shader\n");
543 /* Immediate constants are clamped to [-1;1] at shader creation time if needed */
544 LIST_FOR_EACH_ENTRY(lconst
, &This
->baseShader
.constantsF
, local_constant
, entry
)
546 GLint location
= constant_locations
[lconst
->idx
];
547 /* We found this uniform name in the program - go ahead and send the data */
548 if (location
!= -1) GL_EXTCALL(glUniform4fvARB(location
, 1, (const GLfloat
*)lconst
->value
));
550 checkGLcall("glUniform4fvARB()");
553 /* Loads integer constants (aka uniforms) into the currently set GLSL program. */
554 /* GL locking is done by the caller */
555 static void shader_glsl_load_constantsI(IWineD3DBaseShaderImpl
*This
, const struct wined3d_gl_info
*gl_info
,
556 const GLint locations
[MAX_CONST_I
], const int *constants
, WORD constants_set
)
561 for (i
= 0; constants_set
; constants_set
>>= 1, ++i
)
563 if (!(constants_set
& 1)) continue;
565 TRACE_(d3d_constants
)("Loading constants %u: %i, %i, %i, %i\n",
566 i
, constants
[i
*4], constants
[i
*4+1], constants
[i
*4+2], constants
[i
*4+3]);
568 /* We found this uniform name in the program - go ahead and send the data */
569 GL_EXTCALL(glUniform4ivARB(locations
[i
], 1, &constants
[i
*4]));
570 checkGLcall("glUniform4ivARB");
573 /* Load immediate constants */
574 ptr
= list_head(&This
->baseShader
.constantsI
);
576 const struct local_constant
*lconst
= LIST_ENTRY(ptr
, const struct local_constant
, entry
);
577 unsigned int idx
= lconst
->idx
;
578 const GLint
*values
= (const GLint
*)lconst
->value
;
580 TRACE_(d3d_constants
)("Loading local constants %i: %i, %i, %i, %i\n", idx
,
581 values
[0], values
[1], values
[2], values
[3]);
583 /* We found this uniform name in the program - go ahead and send the data */
584 GL_EXTCALL(glUniform4ivARB(locations
[idx
], 1, values
));
585 checkGLcall("glUniform4ivARB");
586 ptr
= list_next(&This
->baseShader
.constantsI
, ptr
);
590 /* Loads boolean constants (aka uniforms) into the currently set GLSL program. */
591 /* GL locking is done by the caller */
592 static void shader_glsl_load_constantsB(IWineD3DBaseShaderImpl
*This
, const struct wined3d_gl_info
*gl_info
,
593 GLhandleARB programId
, const BOOL
*constants
, WORD constants_set
)
601 switch (This
->baseShader
.reg_maps
.shader_version
.type
)
603 case WINED3D_SHADER_TYPE_VERTEX
:
607 case WINED3D_SHADER_TYPE_GEOMETRY
:
611 case WINED3D_SHADER_TYPE_PIXEL
:
616 FIXME("Unknown shader type %#x.\n",
617 This
->baseShader
.reg_maps
.shader_version
.type
);
622 /* TODO: Benchmark and see if it would be beneficial to store the
623 * locations of the constants to avoid looking up each time */
624 for (i
= 0; constants_set
; constants_set
>>= 1, ++i
)
626 if (!(constants_set
& 1)) continue;
628 TRACE_(d3d_constants
)("Loading constants %i: %i;\n", i
, constants
[i
]);
630 /* TODO: Benchmark and see if it would be beneficial to store the
631 * locations of the constants to avoid looking up each time */
632 snprintf(tmp_name
, sizeof(tmp_name
), "%s[%i]", prefix
, i
);
633 tmp_loc
= GL_EXTCALL(glGetUniformLocationARB(programId
, tmp_name
));
636 /* We found this uniform name in the program - go ahead and send the data */
637 GL_EXTCALL(glUniform1ivARB(tmp_loc
, 1, &constants
[i
]));
638 checkGLcall("glUniform1ivARB");
642 /* Load immediate constants */
643 ptr
= list_head(&This
->baseShader
.constantsB
);
645 const struct local_constant
*lconst
= LIST_ENTRY(ptr
, const struct local_constant
, entry
);
646 unsigned int idx
= lconst
->idx
;
647 const GLint
*values
= (const GLint
*)lconst
->value
;
649 TRACE_(d3d_constants
)("Loading local constants %i: %i\n", idx
, values
[0]);
651 snprintf(tmp_name
, sizeof(tmp_name
), "%s[%i]", prefix
, idx
);
652 tmp_loc
= GL_EXTCALL(glGetUniformLocationARB(programId
, tmp_name
));
654 /* We found this uniform name in the program - go ahead and send the data */
655 GL_EXTCALL(glUniform1ivARB(tmp_loc
, 1, values
));
656 checkGLcall("glUniform1ivARB");
658 ptr
= list_next(&This
->baseShader
.constantsB
, ptr
);
662 static void reset_program_constant_version(struct wine_rb_entry
*entry
, void *context
)
664 WINE_RB_ENTRY_VALUE(entry
, struct glsl_shader_prog_link
, program_lookup_entry
)->constant_version
= 0;
668 * Loads the texture dimensions for NP2 fixup into the currently set GLSL program.
670 /* GL locking is done by the caller (state handler) */
671 static void shader_glsl_load_np2fixup_constants(
672 IWineD3DDevice
* device
,
674 char useVertexShader
) {
676 const IWineD3DDeviceImpl
* deviceImpl
= (const IWineD3DDeviceImpl
*) device
;
677 const struct glsl_shader_prog_link
* prog
= ((struct shader_glsl_priv
*)(deviceImpl
->shader_priv
))->glsl_program
;
680 /* No GLSL program set - nothing to do. */
684 if (!usePixelShader
) {
685 /* NP2 texcoord fixup is (currently) only done for pixelshaders. */
689 if (prog
->ps_args
.np2_fixup
&& -1 != prog
->np2Fixup_location
) {
690 const struct wined3d_gl_info
*gl_info
= &deviceImpl
->adapter
->gl_info
;
691 const IWineD3DStateBlockImpl
* stateBlock
= (const IWineD3DStateBlockImpl
*) deviceImpl
->stateBlock
;
693 UINT fixup
= prog
->ps_args
.np2_fixup
;
694 GLfloat np2fixup_constants
[4 * MAX_FRAGMENT_SAMPLERS
];
696 for (i
= 0; fixup
; fixup
>>= 1, ++i
) {
697 const unsigned char idx
= prog
->np2Fixup_info
->idx
[i
];
698 const IWineD3DBaseTextureImpl
* const tex
= (const IWineD3DBaseTextureImpl
*) stateBlock
->textures
[i
];
699 GLfloat
* tex_dim
= &np2fixup_constants
[(idx
>> 1) * 4];
703 ERR("Nonexistent texture is flagged for NP2 texcoord fixup.\n");
708 tex_dim
[2] = tex
->baseTexture
.pow2Matrix
[0]; tex_dim
[3] = tex
->baseTexture
.pow2Matrix
[5];
710 tex_dim
[0] = tex
->baseTexture
.pow2Matrix
[0]; tex_dim
[1] = tex
->baseTexture
.pow2Matrix
[5];
714 GL_EXTCALL(glUniform4fvARB(prog
->np2Fixup_location
, prog
->np2Fixup_info
->num_consts
, np2fixup_constants
));
719 * Loads the app-supplied constants into the currently set GLSL program.
721 /* GL locking is done by the caller (state handler) */
722 static void shader_glsl_load_constants(const struct wined3d_context
*context
,
723 char usePixelShader
, char useVertexShader
)
725 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
726 IWineD3DDeviceImpl
*device
= context
->swapchain
->device
;
727 IWineD3DStateBlockImpl
* stateBlock
= device
->stateBlock
;
728 struct shader_glsl_priv
*priv
= device
->shader_priv
;
730 GLhandleARB programId
;
731 struct glsl_shader_prog_link
*prog
= priv
->glsl_program
;
732 UINT constant_version
;
736 /* No GLSL program set - nothing to do. */
739 programId
= prog
->programId
;
740 constant_version
= prog
->constant_version
;
742 if (useVertexShader
) {
743 IWineD3DBaseShaderImpl
* vshader
= (IWineD3DBaseShaderImpl
*) stateBlock
->vertexShader
;
745 /* Load DirectX 9 float constants/uniforms for vertex shader */
746 shader_glsl_load_constantsF(vshader
, gl_info
, stateBlock
->vertexShaderConstantF
,
747 prog
->vuniformF_locations
, &priv
->vconst_heap
, priv
->stack
, constant_version
);
749 /* Load DirectX 9 integer constants/uniforms for vertex shader */
750 shader_glsl_load_constantsI(vshader
, gl_info
, prog
->vuniformI_locations
, stateBlock
->vertexShaderConstantI
,
751 stateBlock
->changed
.vertexShaderConstantsI
& vshader
->baseShader
.reg_maps
.integer_constants
);
753 /* Load DirectX 9 boolean constants/uniforms for vertex shader */
754 shader_glsl_load_constantsB(vshader
, gl_info
, programId
, stateBlock
->vertexShaderConstantB
,
755 stateBlock
->changed
.vertexShaderConstantsB
& vshader
->baseShader
.reg_maps
.boolean_constants
);
757 /* Upload the position fixup params */
758 GL_EXTCALL(glUniform4fvARB(prog
->posFixup_location
, 1, &device
->posFixup
[0]));
759 checkGLcall("glUniform4fvARB");
762 if (usePixelShader
) {
764 IWineD3DBaseShaderImpl
* pshader
= (IWineD3DBaseShaderImpl
*) stateBlock
->pixelShader
;
766 /* Load DirectX 9 float constants/uniforms for pixel shader */
767 shader_glsl_load_constantsF(pshader
, gl_info
, stateBlock
->pixelShaderConstantF
,
768 prog
->puniformF_locations
, &priv
->pconst_heap
, priv
->stack
, constant_version
);
770 /* Load DirectX 9 integer constants/uniforms for pixel shader */
771 shader_glsl_load_constantsI(pshader
, gl_info
, prog
->puniformI_locations
, stateBlock
->pixelShaderConstantI
,
772 stateBlock
->changed
.pixelShaderConstantsI
& pshader
->baseShader
.reg_maps
.integer_constants
);
774 /* Load DirectX 9 boolean constants/uniforms for pixel shader */
775 shader_glsl_load_constantsB(pshader
, gl_info
, programId
, stateBlock
->pixelShaderConstantB
,
776 stateBlock
->changed
.pixelShaderConstantsB
& pshader
->baseShader
.reg_maps
.boolean_constants
);
778 /* Upload the environment bump map matrix if needed. The needsbumpmat member specifies the texture stage to load the matrix from.
779 * It can't be 0 for a valid texbem instruction.
781 for(i
= 0; i
< MAX_TEXTURES
; i
++) {
784 if(prog
->bumpenvmat_location
[i
] == -1) continue;
786 data
= (const float *)&stateBlock
->textureState
[i
][WINED3DTSS_BUMPENVMAT00
];
787 GL_EXTCALL(glUniformMatrix2fvARB(prog
->bumpenvmat_location
[i
], 1, 0, data
));
788 checkGLcall("glUniformMatrix2fvARB");
790 /* texbeml needs the luminance scale and offset too. If texbeml is used, needsbumpmat
791 * is set too, so we can check that in the needsbumpmat check
793 if(prog
->luminancescale_location
[i
] != -1) {
794 const GLfloat
*scale
= (const GLfloat
*)&stateBlock
->textureState
[i
][WINED3DTSS_BUMPENVLSCALE
];
795 const GLfloat
*offset
= (const GLfloat
*)&stateBlock
->textureState
[i
][WINED3DTSS_BUMPENVLOFFSET
];
797 GL_EXTCALL(glUniform1fvARB(prog
->luminancescale_location
[i
], 1, scale
));
798 checkGLcall("glUniform1fvARB");
799 GL_EXTCALL(glUniform1fvARB(prog
->luminanceoffset_location
[i
], 1, offset
));
800 checkGLcall("glUniform1fvARB");
804 if(((IWineD3DPixelShaderImpl
*) pshader
)->vpos_uniform
) {
805 float correction_params
[4];
807 if (context
->render_offscreen
)
809 correction_params
[0] = 0.0f
;
810 correction_params
[1] = 1.0f
;
812 /* position is window relative, not viewport relative */
813 correction_params
[0] = context
->current_rt
->currentDesc
.Height
;
814 correction_params
[1] = -1.0f
;
816 GL_EXTCALL(glUniform4fvARB(prog
->ycorrection_location
, 1, correction_params
));
820 if (priv
->next_constant_version
== UINT_MAX
)
822 TRACE("Max constant version reached, resetting to 0.\n");
823 wine_rb_for_each_entry(&priv
->program_lookup
, reset_program_constant_version
, NULL
);
824 priv
->next_constant_version
= 1;
828 prog
->constant_version
= priv
->next_constant_version
++;
832 static inline void update_heap_entry(struct constant_heap
*heap
, unsigned int idx
,
833 unsigned int heap_idx
, DWORD new_version
)
835 struct constant_entry
*entries
= heap
->entries
;
836 unsigned int *positions
= heap
->positions
;
837 unsigned int parent_idx
;
841 parent_idx
= heap_idx
>> 1;
843 if (new_version
<= entries
[parent_idx
].version
) break;
845 entries
[heap_idx
] = entries
[parent_idx
];
846 positions
[entries
[parent_idx
].idx
] = heap_idx
;
847 heap_idx
= parent_idx
;
850 entries
[heap_idx
].version
= new_version
;
851 entries
[heap_idx
].idx
= idx
;
852 positions
[idx
] = heap_idx
;
855 static void shader_glsl_update_float_vertex_constants(IWineD3DDevice
*iface
, UINT start
, UINT count
)
857 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
858 struct shader_glsl_priv
*priv
= This
->shader_priv
;
859 struct constant_heap
*heap
= &priv
->vconst_heap
;
862 for (i
= start
; i
< count
+ start
; ++i
)
864 if (!This
->stateBlock
->changed
.vertexShaderConstantsF
[i
])
865 update_heap_entry(heap
, i
, heap
->size
++, priv
->next_constant_version
);
867 update_heap_entry(heap
, i
, heap
->positions
[i
], priv
->next_constant_version
);
871 static void shader_glsl_update_float_pixel_constants(IWineD3DDevice
*iface
, UINT start
, UINT count
)
873 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
874 struct shader_glsl_priv
*priv
= This
->shader_priv
;
875 struct constant_heap
*heap
= &priv
->pconst_heap
;
878 for (i
= start
; i
< count
+ start
; ++i
)
880 if (!This
->stateBlock
->changed
.pixelShaderConstantsF
[i
])
881 update_heap_entry(heap
, i
, heap
->size
++, priv
->next_constant_version
);
883 update_heap_entry(heap
, i
, heap
->positions
[i
], priv
->next_constant_version
);
887 static unsigned int vec4_varyings(DWORD shader_major
, const struct wined3d_gl_info
*gl_info
)
889 unsigned int ret
= gl_info
->limits
.glsl_varyings
/ 4;
890 /* 4.0 shaders do not write clip coords because d3d10 does not support user clipplanes */
891 if(shader_major
> 3) return ret
;
893 /* 3.0 shaders may need an extra varying for the clip coord on some cards(mostly dx10 ones) */
894 if (gl_info
->quirks
& WINED3D_QUIRK_GLSL_CLIP_VARYING
) ret
-= 1;
898 /** Generate the variable & register declarations for the GLSL output target */
899 static void shader_generate_glsl_declarations(const struct wined3d_context
*context
,
900 struct wined3d_shader_buffer
*buffer
, IWineD3DBaseShader
*iface
,
901 const shader_reg_maps
*reg_maps
, struct shader_glsl_ctx_priv
*ctx_priv
)
903 IWineD3DBaseShaderImpl
* This
= (IWineD3DBaseShaderImpl
*) iface
;
904 IWineD3DDeviceImpl
*device
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
905 const struct ps_compile_args
*ps_args
= ctx_priv
->cur_ps_args
;
906 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
907 unsigned int i
, extra_constants_needed
= 0;
908 const local_constant
*lconst
;
911 /* There are some minor differences between pixel and vertex shaders */
912 char pshader
= shader_is_pshader_version(reg_maps
->shader_version
.type
);
913 char prefix
= pshader
? 'P' : 'V';
915 /* Prototype the subroutines */
916 for (i
= 0, map
= reg_maps
->labels
; map
; map
>>= 1, ++i
)
918 if (map
& 1) shader_addline(buffer
, "void subroutine%u();\n", i
);
921 /* Declare the constants (aka uniforms) */
922 if (This
->baseShader
.limits
.constant_float
> 0) {
923 unsigned max_constantsF
;
924 /* Unless the shader uses indirect addressing, always declare the maximum array size and ignore that we need some
925 * uniforms privately. E.g. if GL supports 256 uniforms, and we need 2 for the pos fixup and immediate values, still
926 * declare VC[256]. If the shader needs more uniforms than we have it won't work in any case. If it uses less, the
927 * compiler will figure out which uniforms are really used and strip them out. This allows a shader to use c255 on
928 * a dx9 card, as long as it doesn't also use all the other constants.
930 * If the shader uses indirect addressing the compiler must assume that all declared uniforms are used. In this case,
931 * declare only the amount that we're assured to have.
933 * Thus we run into problems in these two cases:
934 * 1) The shader really uses more uniforms than supported
935 * 2) The shader uses indirect addressing, less constants than supported, but uses a constant index > #supported consts
939 /* No indirect addressing here. */
940 max_constantsF
= gl_info
->limits
.glsl_ps_float_constants
;
944 if(This
->baseShader
.reg_maps
.usesrelconstF
) {
945 /* Subtract the other potential uniforms from the max available (bools, ints, and 1 row of projection matrix).
946 * Subtract another uniform for immediate values, which have to be loaded via uniform by the driver as well.
947 * The shader code only uses 0.5, 2.0, 1.0, 128 and -128 in vertex shader code, so one vec4 should be enough
948 * (Unfortunately the Nvidia driver doesn't store 128 and -128 in one float).
950 * Writing gl_ClipVertex requires one uniform for each clipplane as well.
952 max_constantsF
= gl_info
->limits
.glsl_vs_float_constants
- 3;
953 if(ctx_priv
->cur_vs_args
->clip_enabled
)
955 max_constantsF
-= gl_info
->limits
.clipplanes
;
957 max_constantsF
-= count_bits(This
->baseShader
.reg_maps
.integer_constants
);
958 /* Strictly speaking a bool only uses one scalar, but the nvidia(Linux) compiler doesn't pack them properly,
959 * so each scalar requires a full vec4. We could work around this by packing the booleans ourselves, but
960 * for now take this into account when calculating the number of available constants
962 max_constantsF
-= count_bits(This
->baseShader
.reg_maps
.boolean_constants
);
963 /* Set by driver quirks in directx.c */
964 max_constantsF
-= gl_info
->reserved_glsl_constants
;
968 max_constantsF
= gl_info
->limits
.glsl_vs_float_constants
;
971 max_constantsF
= min(This
->baseShader
.limits
.constant_float
, max_constantsF
);
972 shader_addline(buffer
, "uniform vec4 %cC[%u];\n", prefix
, max_constantsF
);
975 /* Always declare the full set of constants, the compiler can remove the unused ones because d3d doesn't(yet)
976 * support indirect int and bool constant addressing. This avoids problems if the app uses e.g. i0 and i9.
978 if (This
->baseShader
.limits
.constant_int
> 0 && This
->baseShader
.reg_maps
.integer_constants
)
979 shader_addline(buffer
, "uniform ivec4 %cI[%u];\n", prefix
, This
->baseShader
.limits
.constant_int
);
981 if (This
->baseShader
.limits
.constant_bool
> 0 && This
->baseShader
.reg_maps
.boolean_constants
)
982 shader_addline(buffer
, "uniform bool %cB[%u];\n", prefix
, This
->baseShader
.limits
.constant_bool
);
986 shader_addline(buffer
, "uniform vec4 posFixup;\n");
987 shader_addline(buffer
, "void order_ps_input(in vec4[%u]);\n", MAX_REG_OUTPUT
);
991 for (i
= 0, map
= reg_maps
->bumpmat
; map
; map
>>= 1, ++i
)
993 if (!(map
& 1)) continue;
995 shader_addline(buffer
, "uniform mat2 bumpenvmat%d;\n", i
);
997 if (reg_maps
->luminanceparams
& (1 << i
))
999 shader_addline(buffer
, "uniform float luminancescale%d;\n", i
);
1000 shader_addline(buffer
, "uniform float luminanceoffset%d;\n", i
);
1001 extra_constants_needed
++;
1004 extra_constants_needed
++;
1007 if (ps_args
->srgb_correction
)
1009 shader_addline(buffer
, "const vec4 srgb_const0 = vec4(%.8e, %.8e, %.8e, %.8e);\n",
1010 srgb_pow
, srgb_mul_high
, srgb_sub_high
, srgb_mul_low
);
1011 shader_addline(buffer
, "const vec4 srgb_const1 = vec4(%.8e, 0.0, 0.0, 0.0);\n",
1014 if (reg_maps
->vpos
|| reg_maps
->usesdsy
)
1016 if (This
->baseShader
.limits
.constant_float
+ extra_constants_needed
1017 + 1 < gl_info
->limits
.glsl_ps_float_constants
)
1019 shader_addline(buffer
, "uniform vec4 ycorrection;\n");
1020 ((IWineD3DPixelShaderImpl
*) This
)->vpos_uniform
= 1;
1021 extra_constants_needed
++;
1023 /* This happens because we do not have proper tracking of the constant registers that are
1024 * actually used, only the max limit of the shader version
1026 FIXME("Cannot find a free uniform for vpos correction params\n");
1027 shader_addline(buffer
, "const vec4 ycorrection = vec4(%f, %f, 0.0, 0.0);\n",
1028 context
->render_offscreen
? 0.0f
: device
->render_targets
[0]->currentDesc
.Height
,
1029 context
->render_offscreen
? 1.0f
: -1.0f
);
1031 shader_addline(buffer
, "vec4 vpos;\n");
1035 /* Declare texture samplers */
1036 for (i
= 0; i
< This
->baseShader
.limits
.sampler
; i
++) {
1037 if (reg_maps
->sampler_type
[i
])
1039 switch (reg_maps
->sampler_type
[i
])
1042 if (pshader
&& ps_args
->shadow
& (1 << i
))
1043 shader_addline(buffer
, "uniform sampler1DShadow %csampler%u;\n", prefix
, i
);
1045 shader_addline(buffer
, "uniform sampler1D %csampler%u;\n", prefix
, i
);
1048 if (pshader
&& ps_args
->shadow
& (1 << i
))
1050 if (device
->stateBlock
->textures
[i
]
1051 && IWineD3DBaseTexture_GetTextureDimensions(device
->stateBlock
->textures
[i
])
1052 == GL_TEXTURE_RECTANGLE_ARB
)
1053 shader_addline(buffer
, "uniform sampler2DRectShadow %csampler%u;\n", prefix
, i
);
1055 shader_addline(buffer
, "uniform sampler2DShadow %csampler%u;\n", prefix
, i
);
1059 if (device
->stateBlock
->textures
[i
]
1060 && IWineD3DBaseTexture_GetTextureDimensions(device
->stateBlock
->textures
[i
])
1061 == GL_TEXTURE_RECTANGLE_ARB
)
1062 shader_addline(buffer
, "uniform sampler2DRect %csampler%u;\n", prefix
, i
);
1064 shader_addline(buffer
, "uniform sampler2D %csampler%u;\n", prefix
, i
);
1067 case WINED3DSTT_CUBE
:
1068 if (pshader
&& ps_args
->shadow
& (1 << i
)) FIXME("Unsupported Cube shadow sampler.\n");
1069 shader_addline(buffer
, "uniform samplerCube %csampler%u;\n", prefix
, i
);
1071 case WINED3DSTT_VOLUME
:
1072 if (pshader
&& ps_args
->shadow
& (1 << i
)) FIXME("Unsupported 3D shadow sampler.\n");
1073 shader_addline(buffer
, "uniform sampler3D %csampler%u;\n", prefix
, i
);
1076 shader_addline(buffer
, "uniform unsupported_sampler %csampler%u;\n", prefix
, i
);
1077 FIXME("Unrecognized sampler type: %#x\n", reg_maps
->sampler_type
[i
]);
1083 /* Declare uniforms for NP2 texcoord fixup:
1084 * This is NOT done inside the loop that declares the texture samplers since the NP2 fixup code
1085 * is currently only used for the GeforceFX series and when forcing the ARB_npot extension off.
1086 * Modern cards just skip the code anyway, so put it inside a separate loop. */
1087 if (pshader
&& ps_args
->np2_fixup
) {
1089 struct ps_np2fixup_info
* const fixup
= ctx_priv
->cur_np2fixup_info
;
1092 /* NP2/RECT textures in OpenGL use texcoords in the range [0,width]x[0,height]
1093 * while D3D has them in the (normalized) [0,1]x[0,1] range.
1094 * samplerNP2Fixup stores texture dimensions and is updated through
1095 * shader_glsl_load_np2fixup_constants when the sampler changes. */
1097 for (i
= 0; i
< This
->baseShader
.limits
.sampler
; ++i
) {
1098 if (reg_maps
->sampler_type
[i
]) {
1099 if (!(ps_args
->np2_fixup
& (1 << i
))) continue;
1101 if (WINED3DSTT_2D
!= reg_maps
->sampler_type
[i
]) {
1102 FIXME("Non-2D texture is flagged for NP2 texcoord fixup.\n");
1106 fixup
->idx
[i
] = cur
++;
1110 fixup
->num_consts
= (cur
+ 1) >> 1;
1111 shader_addline(buffer
, "uniform vec4 %csamplerNP2Fixup[%u];\n", prefix
, fixup
->num_consts
);
1114 /* Declare address variables */
1115 for (i
= 0, map
= reg_maps
->address
; map
; map
>>= 1, ++i
)
1117 if (map
& 1) shader_addline(buffer
, "ivec4 A%u;\n", i
);
1120 /* Declare texture coordinate temporaries and initialize them */
1121 for (i
= 0, map
= reg_maps
->texcoord
; map
; map
>>= 1, ++i
)
1123 if (map
& 1) shader_addline(buffer
, "vec4 T%u = gl_TexCoord[%u];\n", i
, i
);
1126 /* Declare input register varyings. Only pixel shader, vertex shaders have that declared in the
1127 * helper function shader that is linked in at link time
1129 if (pshader
&& reg_maps
->shader_version
.major
>= 3)
1131 if (use_vs(device
->stateBlock
))
1133 shader_addline(buffer
, "varying vec4 IN[%u];\n", vec4_varyings(reg_maps
->shader_version
.major
, gl_info
));
1135 /* TODO: Write a replacement shader for the fixed function vertex pipeline, so this isn't needed.
1136 * For fixed function vertex processing + 3.0 pixel shader we need a separate function in the
1137 * pixel shader that reads the fixed function color into the packed input registers.
1139 shader_addline(buffer
, "vec4 IN[%u];\n", vec4_varyings(reg_maps
->shader_version
.major
, gl_info
));
1143 /* Declare output register temporaries */
1144 if(This
->baseShader
.limits
.packed_output
) {
1145 shader_addline(buffer
, "vec4 OUT[%u];\n", This
->baseShader
.limits
.packed_output
);
1148 /* Declare temporary variables */
1149 for (i
= 0, map
= reg_maps
->temporary
; map
; map
>>= 1, ++i
)
1151 if (map
& 1) shader_addline(buffer
, "vec4 R%u;\n", i
);
1154 /* Declare attributes */
1155 if (reg_maps
->shader_version
.type
== WINED3D_SHADER_TYPE_VERTEX
)
1157 for (i
= 0, map
= reg_maps
->input_registers
; map
; map
>>= 1, ++i
)
1159 if (map
& 1) shader_addline(buffer
, "attribute vec4 attrib%i;\n", i
);
1163 /* Declare loop registers aLx */
1164 for (i
= 0; i
< reg_maps
->loop_depth
; i
++) {
1165 shader_addline(buffer
, "int aL%u;\n", i
);
1166 shader_addline(buffer
, "int tmpInt%u;\n", i
);
1169 /* Temporary variables for matrix operations */
1170 shader_addline(buffer
, "vec4 tmp0;\n");
1171 shader_addline(buffer
, "vec4 tmp1;\n");
1173 /* Local constants use a different name so they can be loaded once at shader link time
1174 * They can't be hardcoded into the shader text via LC = {x, y, z, w}; because the
1175 * float -> string conversion can cause precision loss.
1177 if(!This
->baseShader
.load_local_constsF
) {
1178 LIST_FOR_EACH_ENTRY(lconst
, &This
->baseShader
.constantsF
, local_constant
, entry
) {
1179 shader_addline(buffer
, "uniform vec4 %cLC%u;\n", prefix
, lconst
->idx
);
1183 /* Start the main program */
1184 shader_addline(buffer
, "void main() {\n");
1185 if(pshader
&& reg_maps
->vpos
) {
1186 /* DirectX apps expect integer values, while OpenGL drivers add approximately 0.5. This causes
1187 * off-by-one problems as spotted by the vPos d3d9 visual test. Unfortunately the ATI cards do
1188 * not add exactly 0.5, but rather something like 0.49999999 or 0.50000001, which still causes
1189 * precision troubles when we just substract 0.5.
1191 * To deal with that just floor() the position. This will eliminate the fraction on all cards.
1193 * TODO: Test how that behaves with multisampling once we can enable multisampling in winex11.
1195 * An advantage of floor is that it works even if the driver doesn't add 1/2. It is somewhat
1196 * questionable if 1.5, 2.5, ... are the proper values to return in gl_FragCoord, even though
1197 * coordinates specify the pixel centers instead of the pixel corners. This code will behave
1198 * correctly on drivers that returns integer values.
1200 shader_addline(buffer
, "vpos = floor(vec4(0, ycorrection[0], 0, 0) + gl_FragCoord * vec4(1, ycorrection[1], 1, 1));\n");
1204 /*****************************************************************************
1205 * Functions to generate GLSL strings from DirectX Shader bytecode begin here.
1207 * For more information, see http://wiki.winehq.org/DirectX-Shaders
1208 ****************************************************************************/
1211 static void shader_glsl_add_src_param(const struct wined3d_shader_instruction
*ins
,
1212 const struct wined3d_shader_src_param
*wined3d_src
, DWORD mask
, glsl_src_param_t
*glsl_src
);
1214 /** Used for opcode modifiers - They multiply the result by the specified amount */
1215 static const char * const shift_glsl_tab
[] = {
1217 "2.0 * ", /* 1 (x2) */
1218 "4.0 * ", /* 2 (x4) */
1219 "8.0 * ", /* 3 (x8) */
1220 "16.0 * ", /* 4 (x16) */
1221 "32.0 * ", /* 5 (x32) */
1228 "0.0625 * ", /* 12 (d16) */
1229 "0.125 * ", /* 13 (d8) */
1230 "0.25 * ", /* 14 (d4) */
1231 "0.5 * " /* 15 (d2) */
1234 /* Generate a GLSL parameter that does the input modifier computation and return the input register/mask to use */
1235 static void shader_glsl_gen_modifier(DWORD src_modifier
, const char *in_reg
, const char *in_regswizzle
, char *out_str
)
1239 switch (src_modifier
)
1241 case WINED3DSPSM_DZ
: /* Need to handle this in the instructions itself (texld & texcrd). */
1242 case WINED3DSPSM_DW
:
1243 case WINED3DSPSM_NONE
:
1244 sprintf(out_str
, "%s%s", in_reg
, in_regswizzle
);
1246 case WINED3DSPSM_NEG
:
1247 sprintf(out_str
, "-%s%s", in_reg
, in_regswizzle
);
1249 case WINED3DSPSM_NOT
:
1250 sprintf(out_str
, "!%s%s", in_reg
, in_regswizzle
);
1252 case WINED3DSPSM_BIAS
:
1253 sprintf(out_str
, "(%s%s - vec4(0.5)%s)", in_reg
, in_regswizzle
, in_regswizzle
);
1255 case WINED3DSPSM_BIASNEG
:
1256 sprintf(out_str
, "-(%s%s - vec4(0.5)%s)", in_reg
, in_regswizzle
, in_regswizzle
);
1258 case WINED3DSPSM_SIGN
:
1259 sprintf(out_str
, "(2.0 * (%s%s - 0.5))", in_reg
, in_regswizzle
);
1261 case WINED3DSPSM_SIGNNEG
:
1262 sprintf(out_str
, "-(2.0 * (%s%s - 0.5))", in_reg
, in_regswizzle
);
1264 case WINED3DSPSM_COMP
:
1265 sprintf(out_str
, "(1.0 - %s%s)", in_reg
, in_regswizzle
);
1267 case WINED3DSPSM_X2
:
1268 sprintf(out_str
, "(2.0 * %s%s)", in_reg
, in_regswizzle
);
1270 case WINED3DSPSM_X2NEG
:
1271 sprintf(out_str
, "-(2.0 * %s%s)", in_reg
, in_regswizzle
);
1273 case WINED3DSPSM_ABS
:
1274 sprintf(out_str
, "abs(%s%s)", in_reg
, in_regswizzle
);
1276 case WINED3DSPSM_ABSNEG
:
1277 sprintf(out_str
, "-abs(%s%s)", in_reg
, in_regswizzle
);
1280 FIXME("Unhandled modifier %u\n", src_modifier
);
1281 sprintf(out_str
, "%s%s", in_reg
, in_regswizzle
);
1285 /** Writes the GLSL variable name that corresponds to the register that the
1286 * DX opcode parameter is trying to access */
1287 static void shader_glsl_get_register_name(const struct wined3d_shader_register
*reg
,
1288 char *register_name
, BOOL
*is_color
, const struct wined3d_shader_instruction
*ins
)
1290 /* oPos, oFog and oPts in D3D */
1291 static const char * const hwrastout_reg_names
[] = {"OUT[10]", "OUT[11].x", "OUT[11].y"};
1293 IWineD3DBaseShaderImpl
*This
= (IWineD3DBaseShaderImpl
*)ins
->ctx
->shader
;
1294 const struct wined3d_gl_info
*gl_info
= ins
->ctx
->gl_info
;
1295 char pshader
= shader_is_pshader_version(This
->baseShader
.reg_maps
.shader_version
.type
);
1301 case WINED3DSPR_TEMP
:
1302 sprintf(register_name
, "R%u", reg
->idx
);
1305 case WINED3DSPR_INPUT
:
1306 /* vertex shaders */
1309 struct shader_glsl_ctx_priv
*priv
= ins
->ctx
->backend_data
;
1310 if (priv
->cur_vs_args
->swizzle_map
& (1 << reg
->idx
)) *is_color
= TRUE
;
1311 sprintf(register_name
, "attrib%u", reg
->idx
);
1315 /* pixel shaders >= 3.0 */
1316 if (This
->baseShader
.reg_maps
.shader_version
.major
>= 3)
1318 DWORD idx
= ((IWineD3DPixelShaderImpl
*)This
)->input_reg_map
[reg
->idx
];
1319 unsigned int in_count
= vec4_varyings(This
->baseShader
.reg_maps
.shader_version
.major
, gl_info
);
1323 glsl_src_param_t rel_param
;
1325 shader_glsl_add_src_param(ins
, reg
->rel_addr
, WINED3DSP_WRITEMASK_0
, &rel_param
);
1327 /* Removing a + 0 would be an obvious optimization, but macos doesn't see the NOP
1328 * operation there */
1331 if (((IWineD3DPixelShaderImpl
*)This
)->declared_in_count
> in_count
)
1333 sprintf(register_name
,
1334 "((%s + %u) > %d ? (%s + %u) > %d ? gl_SecondaryColor : gl_Color : IN[%s + %u])",
1335 rel_param
.param_str
, idx
, in_count
- 1, rel_param
.param_str
, idx
, in_count
,
1336 rel_param
.param_str
, idx
);
1340 sprintf(register_name
, "IN[%s + %u]", rel_param
.param_str
, idx
);
1345 if (((IWineD3DPixelShaderImpl
*)This
)->declared_in_count
> in_count
)
1347 sprintf(register_name
, "((%s) > %d ? (%s) > %d ? gl_SecondaryColor : gl_Color : IN[%s])",
1348 rel_param
.param_str
, in_count
- 1, rel_param
.param_str
, in_count
,
1349 rel_param
.param_str
);
1353 sprintf(register_name
, "IN[%s]", rel_param
.param_str
);
1359 if (idx
== in_count
) sprintf(register_name
, "gl_Color");
1360 else if (idx
== in_count
+ 1) sprintf(register_name
, "gl_SecondaryColor");
1361 else sprintf(register_name
, "IN[%u]", idx
);
1366 if (reg
->idx
== 0) strcpy(register_name
, "gl_Color");
1367 else strcpy(register_name
, "gl_SecondaryColor");
1372 case WINED3DSPR_CONST
:
1374 const char prefix
= pshader
? 'P' : 'V';
1376 /* Relative addressing */
1379 glsl_src_param_t rel_param
;
1380 shader_glsl_add_src_param(ins
, reg
->rel_addr
, WINED3DSP_WRITEMASK_0
, &rel_param
);
1381 if (reg
->idx
) sprintf(register_name
, "%cC[%s + %u]", prefix
, rel_param
.param_str
, reg
->idx
);
1382 else sprintf(register_name
, "%cC[%s]", prefix
, rel_param
.param_str
);
1386 if (shader_constant_is_local(This
, reg
->idx
))
1387 sprintf(register_name
, "%cLC%u", prefix
, reg
->idx
);
1389 sprintf(register_name
, "%cC[%u]", prefix
, reg
->idx
);
1394 case WINED3DSPR_CONSTINT
:
1395 if (pshader
) sprintf(register_name
, "PI[%u]", reg
->idx
);
1396 else sprintf(register_name
, "VI[%u]", reg
->idx
);
1399 case WINED3DSPR_CONSTBOOL
:
1400 if (pshader
) sprintf(register_name
, "PB[%u]", reg
->idx
);
1401 else sprintf(register_name
, "VB[%u]", reg
->idx
);
1404 case WINED3DSPR_TEXTURE
: /* case WINED3DSPR_ADDR: */
1405 if (pshader
) sprintf(register_name
, "T%u", reg
->idx
);
1406 else sprintf(register_name
, "A%u", reg
->idx
);
1409 case WINED3DSPR_LOOP
:
1410 sprintf(register_name
, "aL%u", This
->baseShader
.cur_loop_regno
- 1);
1413 case WINED3DSPR_SAMPLER
:
1414 if (pshader
) sprintf(register_name
, "Psampler%u", reg
->idx
);
1415 else sprintf(register_name
, "Vsampler%u", reg
->idx
);
1418 case WINED3DSPR_COLOROUT
:
1419 if (reg
->idx
>= gl_info
->limits
.buffers
)
1420 WARN("Write to render target %u, only %d supported.\n", reg
->idx
, gl_info
->limits
.buffers
);
1422 sprintf(register_name
, "gl_FragData[%u]", reg
->idx
);
1425 case WINED3DSPR_RASTOUT
:
1426 sprintf(register_name
, "%s", hwrastout_reg_names
[reg
->idx
]);
1429 case WINED3DSPR_DEPTHOUT
:
1430 sprintf(register_name
, "gl_FragDepth");
1433 case WINED3DSPR_ATTROUT
:
1434 if (reg
->idx
== 0) sprintf(register_name
, "OUT[8]");
1435 else sprintf(register_name
, "OUT[9]");
1438 case WINED3DSPR_TEXCRDOUT
:
1439 /* Vertex shaders >= 3.0: WINED3DSPR_OUTPUT */
1440 sprintf(register_name
, "OUT[%u]", reg
->idx
);
1443 case WINED3DSPR_MISCTYPE
:
1447 sprintf(register_name
, "vpos");
1449 else if (reg
->idx
== 1)
1451 /* Note that gl_FrontFacing is a bool, while vFace is
1452 * a float for which the sign determines front/back */
1453 sprintf(register_name
, "(gl_FrontFacing ? 1.0 : -1.0)");
1457 FIXME("Unhandled misctype register %d\n", reg
->idx
);
1458 sprintf(register_name
, "unrecognized_register");
1462 case WINED3DSPR_IMMCONST
:
1463 switch (reg
->immconst_type
)
1465 case WINED3D_IMMCONST_FLOAT
:
1466 sprintf(register_name
, "%.8e", *(const float *)reg
->immconst_data
);
1469 case WINED3D_IMMCONST_FLOAT4
:
1470 sprintf(register_name
, "vec4(%.8e, %.8e, %.8e, %.8e)",
1471 *(const float *)®
->immconst_data
[0], *(const float *)®
->immconst_data
[1],
1472 *(const float *)®
->immconst_data
[2], *(const float *)®
->immconst_data
[3]);
1476 FIXME("Unhandled immconst type %#x\n", reg
->immconst_type
);
1477 sprintf(register_name
, "<unhandled_immconst_type %#x>", reg
->immconst_type
);
1482 FIXME("Unhandled register name Type(%d)\n", reg
->type
);
1483 sprintf(register_name
, "unrecognized_register");
1488 static void shader_glsl_write_mask_to_str(DWORD write_mask
, char *str
)
1491 if (write_mask
& WINED3DSP_WRITEMASK_0
) *str
++ = 'x';
1492 if (write_mask
& WINED3DSP_WRITEMASK_1
) *str
++ = 'y';
1493 if (write_mask
& WINED3DSP_WRITEMASK_2
) *str
++ = 'z';
1494 if (write_mask
& WINED3DSP_WRITEMASK_3
) *str
++ = 'w';
1498 /* Get the GLSL write mask for the destination register */
1499 static DWORD
shader_glsl_get_write_mask(const struct wined3d_shader_dst_param
*param
, char *write_mask
)
1501 DWORD mask
= param
->write_mask
;
1503 if (shader_is_scalar(¶m
->reg
))
1505 mask
= WINED3DSP_WRITEMASK_0
;
1510 shader_glsl_write_mask_to_str(mask
, write_mask
);
1516 static unsigned int shader_glsl_get_write_mask_size(DWORD write_mask
) {
1517 unsigned int size
= 0;
1519 if (write_mask
& WINED3DSP_WRITEMASK_0
) ++size
;
1520 if (write_mask
& WINED3DSP_WRITEMASK_1
) ++size
;
1521 if (write_mask
& WINED3DSP_WRITEMASK_2
) ++size
;
1522 if (write_mask
& WINED3DSP_WRITEMASK_3
) ++size
;
1527 static void shader_glsl_swizzle_to_str(const DWORD swizzle
, BOOL fixup
, DWORD mask
, char *str
)
1529 /* For registers of type WINED3DDECLTYPE_D3DCOLOR, data is stored as "bgra",
1530 * but addressed as "rgba". To fix this we need to swap the register's x
1531 * and z components. */
1532 const char *swizzle_chars
= fixup
? "zyxw" : "xyzw";
1535 /* swizzle bits fields: wwzzyyxx */
1536 if (mask
& WINED3DSP_WRITEMASK_0
) *str
++ = swizzle_chars
[swizzle
& 0x03];
1537 if (mask
& WINED3DSP_WRITEMASK_1
) *str
++ = swizzle_chars
[(swizzle
>> 2) & 0x03];
1538 if (mask
& WINED3DSP_WRITEMASK_2
) *str
++ = swizzle_chars
[(swizzle
>> 4) & 0x03];
1539 if (mask
& WINED3DSP_WRITEMASK_3
) *str
++ = swizzle_chars
[(swizzle
>> 6) & 0x03];
1543 static void shader_glsl_get_swizzle(const struct wined3d_shader_src_param
*param
,
1544 BOOL fixup
, DWORD mask
, char *swizzle_str
)
1546 if (shader_is_scalar(¶m
->reg
))
1547 *swizzle_str
= '\0';
1549 shader_glsl_swizzle_to_str(param
->swizzle
, fixup
, mask
, swizzle_str
);
1552 /* From a given parameter token, generate the corresponding GLSL string.
1553 * Also, return the actual register name and swizzle in case the
1554 * caller needs this information as well. */
1555 static void shader_glsl_add_src_param(const struct wined3d_shader_instruction
*ins
,
1556 const struct wined3d_shader_src_param
*wined3d_src
, DWORD mask
, glsl_src_param_t
*glsl_src
)
1558 BOOL is_color
= FALSE
;
1559 char swizzle_str
[6];
1561 glsl_src
->reg_name
[0] = '\0';
1562 glsl_src
->param_str
[0] = '\0';
1563 swizzle_str
[0] = '\0';
1565 shader_glsl_get_register_name(&wined3d_src
->reg
, glsl_src
->reg_name
, &is_color
, ins
);
1566 shader_glsl_get_swizzle(wined3d_src
, is_color
, mask
, swizzle_str
);
1567 shader_glsl_gen_modifier(wined3d_src
->modifiers
, glsl_src
->reg_name
, swizzle_str
, glsl_src
->param_str
);
1570 /* From a given parameter token, generate the corresponding GLSL string.
1571 * Also, return the actual register name and swizzle in case the
1572 * caller needs this information as well. */
1573 static DWORD
shader_glsl_add_dst_param(const struct wined3d_shader_instruction
*ins
,
1574 const struct wined3d_shader_dst_param
*wined3d_dst
, glsl_dst_param_t
*glsl_dst
)
1576 BOOL is_color
= FALSE
;
1578 glsl_dst
->mask_str
[0] = '\0';
1579 glsl_dst
->reg_name
[0] = '\0';
1581 shader_glsl_get_register_name(&wined3d_dst
->reg
, glsl_dst
->reg_name
, &is_color
, ins
);
1582 return shader_glsl_get_write_mask(wined3d_dst
, glsl_dst
->mask_str
);
1585 /* Append the destination part of the instruction to the buffer, return the effective write mask */
1586 static DWORD
shader_glsl_append_dst_ext(struct wined3d_shader_buffer
*buffer
,
1587 const struct wined3d_shader_instruction
*ins
, const struct wined3d_shader_dst_param
*dst
)
1589 glsl_dst_param_t glsl_dst
;
1592 mask
= shader_glsl_add_dst_param(ins
, dst
, &glsl_dst
);
1593 if (mask
) shader_addline(buffer
, "%s%s = %s(", glsl_dst
.reg_name
, glsl_dst
.mask_str
, shift_glsl_tab
[dst
->shift
]);
1598 /* Append the destination part of the instruction to the buffer, return the effective write mask */
1599 static DWORD
shader_glsl_append_dst(struct wined3d_shader_buffer
*buffer
, const struct wined3d_shader_instruction
*ins
)
1601 return shader_glsl_append_dst_ext(buffer
, ins
, &ins
->dst
[0]);
1604 /** Process GLSL instruction modifiers */
1605 static void shader_glsl_add_instruction_modifiers(const struct wined3d_shader_instruction
*ins
)
1607 glsl_dst_param_t dst_param
;
1610 if (!ins
->dst_count
) return;
1612 modifiers
= ins
->dst
[0].modifiers
;
1613 if (!modifiers
) return;
1615 shader_glsl_add_dst_param(ins
, &ins
->dst
[0], &dst_param
);
1617 if (modifiers
& WINED3DSPDM_SATURATE
)
1619 /* _SAT means to clamp the value of the register to between 0 and 1 */
1620 shader_addline(ins
->ctx
->buffer
, "%s%s = clamp(%s%s, 0.0, 1.0);\n", dst_param
.reg_name
,
1621 dst_param
.mask_str
, dst_param
.reg_name
, dst_param
.mask_str
);
1624 if (modifiers
& WINED3DSPDM_MSAMPCENTROID
)
1626 FIXME("_centroid modifier not handled\n");
1629 if (modifiers
& WINED3DSPDM_PARTIALPRECISION
)
1631 /* MSDN says this modifier can be safely ignored, so that's what we'll do. */
1635 static inline const char *shader_get_comp_op(DWORD op
)
1638 case COMPARISON_GT
: return ">";
1639 case COMPARISON_EQ
: return "==";
1640 case COMPARISON_GE
: return ">=";
1641 case COMPARISON_LT
: return "<";
1642 case COMPARISON_NE
: return "!=";
1643 case COMPARISON_LE
: return "<=";
1645 FIXME("Unrecognized comparison value: %u\n", op
);
1650 static void shader_glsl_get_sample_function(const struct wined3d_shader_context
*ctx
,
1651 DWORD sampler_idx
, DWORD flags
, glsl_sample_function_t
*sample_function
)
1653 WINED3DSAMPLER_TEXTURE_TYPE sampler_type
= ctx
->reg_maps
->sampler_type
[sampler_idx
];
1654 const struct wined3d_gl_info
*gl_info
= ctx
->gl_info
;
1655 BOOL shadow
= shader_is_pshader_version(ctx
->reg_maps
->shader_version
.type
)
1656 && (((const struct shader_glsl_ctx_priv
*)ctx
->backend_data
)->cur_ps_args
->shadow
& (1 << sampler_idx
));
1657 BOOL projected
= flags
& WINED3D_GLSL_SAMPLE_PROJECTED
;
1658 BOOL texrect
= flags
& WINED3D_GLSL_SAMPLE_RECT
;
1659 BOOL lod
= flags
& WINED3D_GLSL_SAMPLE_LOD
;
1660 BOOL grad
= flags
& WINED3D_GLSL_SAMPLE_GRAD
;
1662 /* Note that there's no such thing as a projected cube texture. */
1663 switch(sampler_type
) {
1669 sample_function
->name
= projected
? "shadow1DProjLod" : "shadow1DLod";
1673 if (gl_info
->supported
[EXT_GPU_SHADER4
])
1674 sample_function
->name
= projected
? "shadow1DProjGrad" : "shadow1DGrad";
1675 else if (gl_info
->supported
[ARB_SHADER_TEXTURE_LOD
])
1676 sample_function
->name
= projected
? "shadow1DProjGradARB" : "shadow1DGradARB";
1679 FIXME("Unsupported 1D shadow grad function.\n");
1680 sample_function
->name
= "unsupported1DGrad";
1685 sample_function
->name
= projected
? "shadow1DProj" : "shadow1D";
1687 sample_function
->coord_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
;
1693 sample_function
->name
= projected
? "texture1DProjLod" : "texture1DLod";
1697 if (gl_info
->supported
[EXT_GPU_SHADER4
])
1698 sample_function
->name
= projected
? "texture1DProjGrad" : "texture1DGrad";
1699 else if (gl_info
->supported
[ARB_SHADER_TEXTURE_LOD
])
1700 sample_function
->name
= projected
? "texture1DProjGradARB" : "texture1DGradARB";
1703 FIXME("Unsupported 1D grad function.\n");
1704 sample_function
->name
= "unsupported1DGrad";
1709 sample_function
->name
= projected
? "texture1DProj" : "texture1D";
1711 sample_function
->coord_mask
= WINED3DSP_WRITEMASK_0
;
1722 sample_function
->name
= projected
? "shadow2DRectProjLod" : "shadow2DRectLod";
1726 if (gl_info
->supported
[EXT_GPU_SHADER4
])
1727 sample_function
->name
= projected
? "shadow2DRectProjGrad" : "shadow2DRectGrad";
1728 else if (gl_info
->supported
[ARB_SHADER_TEXTURE_LOD
])
1729 sample_function
->name
= projected
? "shadow2DRectProjGradARB" : "shadow2DRectGradARB";
1732 FIXME("Unsupported RECT shadow grad function.\n");
1733 sample_function
->name
= "unsupported2DRectGrad";
1738 sample_function
->name
= projected
? "shadow2DRectProj" : "shadow2DRect";
1745 sample_function
->name
= projected
? "shadow2DProjLod" : "shadow2DLod";
1749 if (gl_info
->supported
[EXT_GPU_SHADER4
])
1750 sample_function
->name
= projected
? "shadow2DProjGrad" : "shadow2DGrad";
1751 else if (gl_info
->supported
[ARB_SHADER_TEXTURE_LOD
])
1752 sample_function
->name
= projected
? "shadow2DProjGradARB" : "shadow2DGradARB";
1755 FIXME("Unsupported 2D shadow grad function.\n");
1756 sample_function
->name
= "unsupported2DGrad";
1761 sample_function
->name
= projected
? "shadow2DProj" : "shadow2D";
1764 sample_function
->coord_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
1772 sample_function
->name
= projected
? "texture2DRectProjLod" : "texture2DRectLod";
1776 if (gl_info
->supported
[EXT_GPU_SHADER4
])
1777 sample_function
->name
= projected
? "texture2DRectProjGrad" : "texture2DRectGrad";
1778 else if (gl_info
->supported
[ARB_SHADER_TEXTURE_LOD
])
1779 sample_function
->name
= projected
? "texture2DRectProjGradARB" : "texture2DRectGradARB";
1782 FIXME("Unsupported RECT grad function.\n");
1783 sample_function
->name
= "unsupported2DRectGrad";
1788 sample_function
->name
= projected
? "texture2DRectProj" : "texture2DRect";
1795 sample_function
->name
= projected
? "texture2DProjLod" : "texture2DLod";
1799 if (gl_info
->supported
[EXT_GPU_SHADER4
])
1800 sample_function
->name
= projected
? "texture2DProjGrad" : "texture2DGrad";
1801 else if (gl_info
->supported
[ARB_SHADER_TEXTURE_LOD
])
1802 sample_function
->name
= projected
? "texture2DProjGradARB" : "texture2DGradARB";
1805 FIXME("Unsupported 2D grad function.\n");
1806 sample_function
->name
= "unsupported2DGrad";
1811 sample_function
->name
= projected
? "texture2DProj" : "texture2D";
1814 sample_function
->coord_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
;
1818 case WINED3DSTT_CUBE
:
1821 FIXME("Unsupported Cube shadow function.\n");
1822 sample_function
->name
= "unsupportedCubeShadow";
1823 sample_function
->coord_mask
= 0;
1829 sample_function
->name
= "textureCubeLod";
1833 if (gl_info
->supported
[EXT_GPU_SHADER4
])
1834 sample_function
->name
= "textureCubeGrad";
1835 else if (gl_info
->supported
[ARB_SHADER_TEXTURE_LOD
])
1836 sample_function
->name
= "textureCubeGradARB";
1839 FIXME("Unsupported Cube grad function.\n");
1840 sample_function
->name
= "unsupportedCubeGrad";
1845 sample_function
->name
= "textureCube";
1847 sample_function
->coord_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
1851 case WINED3DSTT_VOLUME
:
1854 FIXME("Unsupported 3D shadow function.\n");
1855 sample_function
->name
= "unsupported3DShadow";
1856 sample_function
->coord_mask
= 0;
1862 sample_function
->name
= projected
? "texture3DProjLod" : "texture3DLod";
1866 if (gl_info
->supported
[EXT_GPU_SHADER4
])
1867 sample_function
->name
= projected
? "texture3DProjGrad" : "texture3DGrad";
1868 else if (gl_info
->supported
[ARB_SHADER_TEXTURE_LOD
])
1869 sample_function
->name
= projected
? "texture3DProjGradARB" : "texture3DGradARB";
1872 FIXME("Unsupported 3D grad function.\n");
1873 sample_function
->name
= "unsupported3DGrad";
1878 sample_function
->name
= projected
? "texture3DProj" : "texture3D";
1880 sample_function
->coord_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
1885 sample_function
->name
= "";
1886 sample_function
->coord_mask
= 0;
1887 FIXME("Unrecognized sampler type: %#x;\n", sampler_type
);
1892 static void shader_glsl_append_fixup_arg(char *arguments
, const char *reg_name
,
1893 BOOL sign_fixup
, enum fixup_channel_source channel_source
)
1895 switch(channel_source
)
1897 case CHANNEL_SOURCE_ZERO
:
1898 strcat(arguments
, "0.0");
1901 case CHANNEL_SOURCE_ONE
:
1902 strcat(arguments
, "1.0");
1905 case CHANNEL_SOURCE_X
:
1906 strcat(arguments
, reg_name
);
1907 strcat(arguments
, ".x");
1910 case CHANNEL_SOURCE_Y
:
1911 strcat(arguments
, reg_name
);
1912 strcat(arguments
, ".y");
1915 case CHANNEL_SOURCE_Z
:
1916 strcat(arguments
, reg_name
);
1917 strcat(arguments
, ".z");
1920 case CHANNEL_SOURCE_W
:
1921 strcat(arguments
, reg_name
);
1922 strcat(arguments
, ".w");
1926 FIXME("Unhandled channel source %#x\n", channel_source
);
1927 strcat(arguments
, "undefined");
1931 if (sign_fixup
) strcat(arguments
, " * 2.0 - 1.0");
1934 static void shader_glsl_color_correction(const struct wined3d_shader_instruction
*ins
, struct color_fixup_desc fixup
)
1936 struct wined3d_shader_dst_param dst
;
1937 unsigned int mask_size
, remaining
;
1938 glsl_dst_param_t dst_param
;
1939 char arguments
[256];
1943 if (fixup
.x_sign_fixup
|| fixup
.x_source
!= CHANNEL_SOURCE_X
) mask
|= WINED3DSP_WRITEMASK_0
;
1944 if (fixup
.y_sign_fixup
|| fixup
.y_source
!= CHANNEL_SOURCE_Y
) mask
|= WINED3DSP_WRITEMASK_1
;
1945 if (fixup
.z_sign_fixup
|| fixup
.z_source
!= CHANNEL_SOURCE_Z
) mask
|= WINED3DSP_WRITEMASK_2
;
1946 if (fixup
.w_sign_fixup
|| fixup
.w_source
!= CHANNEL_SOURCE_W
) mask
|= WINED3DSP_WRITEMASK_3
;
1947 mask
&= ins
->dst
[0].write_mask
;
1949 if (!mask
) return; /* Nothing to do */
1951 if (is_complex_fixup(fixup
))
1953 enum complex_fixup complex_fixup
= get_complex_fixup(fixup
);
1954 FIXME("Complex fixup (%#x) not supported\n",complex_fixup
);
1958 mask_size
= shader_glsl_get_write_mask_size(mask
);
1961 dst
.write_mask
= mask
;
1962 shader_glsl_add_dst_param(ins
, &dst
, &dst_param
);
1964 arguments
[0] = '\0';
1965 remaining
= mask_size
;
1966 if (mask
& WINED3DSP_WRITEMASK_0
)
1968 shader_glsl_append_fixup_arg(arguments
, dst_param
.reg_name
, fixup
.x_sign_fixup
, fixup
.x_source
);
1969 if (--remaining
) strcat(arguments
, ", ");
1971 if (mask
& WINED3DSP_WRITEMASK_1
)
1973 shader_glsl_append_fixup_arg(arguments
, dst_param
.reg_name
, fixup
.y_sign_fixup
, fixup
.y_source
);
1974 if (--remaining
) strcat(arguments
, ", ");
1976 if (mask
& WINED3DSP_WRITEMASK_2
)
1978 shader_glsl_append_fixup_arg(arguments
, dst_param
.reg_name
, fixup
.z_sign_fixup
, fixup
.z_source
);
1979 if (--remaining
) strcat(arguments
, ", ");
1981 if (mask
& WINED3DSP_WRITEMASK_3
)
1983 shader_glsl_append_fixup_arg(arguments
, dst_param
.reg_name
, fixup
.w_sign_fixup
, fixup
.w_source
);
1984 if (--remaining
) strcat(arguments
, ", ");
1989 shader_addline(ins
->ctx
->buffer
, "%s%s = vec%u(%s);\n",
1990 dst_param
.reg_name
, dst_param
.mask_str
, mask_size
, arguments
);
1994 shader_addline(ins
->ctx
->buffer
, "%s%s = %s;\n", dst_param
.reg_name
, dst_param
.mask_str
, arguments
);
1998 static void PRINTF_ATTR(8, 9) shader_glsl_gen_sample_code(const struct wined3d_shader_instruction
*ins
,
1999 DWORD sampler
, const glsl_sample_function_t
*sample_function
, DWORD swizzle
,
2000 const char *dx
, const char *dy
,
2001 const char *bias
, const char *coord_reg_fmt
, ...)
2003 const char *sampler_base
;
2004 char dst_swizzle
[6];
2005 struct color_fixup_desc fixup
;
2006 BOOL np2_fixup
= FALSE
;
2009 shader_glsl_swizzle_to_str(swizzle
, FALSE
, ins
->dst
[0].write_mask
, dst_swizzle
);
2011 if (shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
.type
))
2013 const struct shader_glsl_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2014 fixup
= priv
->cur_ps_args
->color_fixup
[sampler
];
2015 sampler_base
= "Psampler";
2017 if(priv
->cur_ps_args
->np2_fixup
& (1 << sampler
)) {
2019 FIXME("Biased sampling from NP2 textures is unsupported\n");
2025 sampler_base
= "Vsampler";
2026 fixup
= COLOR_FIXUP_IDENTITY
; /* FIXME: Vshader color fixup */
2029 shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2031 shader_addline(ins
->ctx
->buffer
, "%s(%s%u, ", sample_function
->name
, sampler_base
, sampler
);
2033 va_start(args
, coord_reg_fmt
);
2034 shader_vaddline(ins
->ctx
->buffer
, coord_reg_fmt
, args
);
2038 shader_addline(ins
->ctx
->buffer
, ", %s)%s);\n", bias
, dst_swizzle
);
2041 const struct shader_glsl_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2042 const unsigned char idx
= priv
->cur_np2fixup_info
->idx
[sampler
];
2044 shader_addline(ins
->ctx
->buffer
, " * PsamplerNP2Fixup[%u].%s)%s);\n", idx
>> 1,
2045 (idx
% 2) ? "zw" : "xy", dst_swizzle
);
2046 } else if(dx
&& dy
) {
2047 shader_addline(ins
->ctx
->buffer
, ", %s, %s)%s);\n", dx
, dy
, dst_swizzle
);
2049 shader_addline(ins
->ctx
->buffer
, ")%s);\n", dst_swizzle
);
2053 if(!is_identity_fixup(fixup
)) {
2054 shader_glsl_color_correction(ins
, fixup
);
2058 /*****************************************************************************
2059 * Begin processing individual instruction opcodes
2060 ****************************************************************************/
2062 /* Generate GLSL arithmetic functions (dst = src1 + src2) */
2063 static void shader_glsl_arith(const struct wined3d_shader_instruction
*ins
)
2065 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2066 glsl_src_param_t src0_param
;
2067 glsl_src_param_t src1_param
;
2071 /* Determine the GLSL operator to use based on the opcode */
2072 switch (ins
->handler_idx
)
2074 case WINED3DSIH_MUL
: op
= '*'; break;
2075 case WINED3DSIH_ADD
: op
= '+'; break;
2076 case WINED3DSIH_SUB
: op
= '-'; break;
2079 FIXME("Opcode %#x not yet handled in GLSL\n", ins
->handler_idx
);
2083 write_mask
= shader_glsl_append_dst(buffer
, ins
);
2084 shader_glsl_add_src_param(ins
, &ins
->src
[0], write_mask
, &src0_param
);
2085 shader_glsl_add_src_param(ins
, &ins
->src
[1], write_mask
, &src1_param
);
2086 shader_addline(buffer
, "%s %c %s);\n", src0_param
.param_str
, op
, src1_param
.param_str
);
2089 /* Process the WINED3DSIO_MOV opcode using GLSL (dst = src) */
2090 static void shader_glsl_mov(const struct wined3d_shader_instruction
*ins
)
2092 const struct wined3d_gl_info
*gl_info
= ins
->ctx
->gl_info
;
2093 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2094 glsl_src_param_t src0_param
;
2097 write_mask
= shader_glsl_append_dst(buffer
, ins
);
2098 shader_glsl_add_src_param(ins
, &ins
->src
[0], write_mask
, &src0_param
);
2100 /* In vs_1_1 WINED3DSIO_MOV can write to the address register. In later
2101 * shader versions WINED3DSIO_MOVA is used for this. */
2102 if (ins
->ctx
->reg_maps
->shader_version
.major
== 1
2103 && !shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
.type
)
2104 && ins
->dst
[0].reg
.type
== WINED3DSPR_ADDR
)
2106 /* This is a simple floor() */
2107 unsigned int mask_size
= shader_glsl_get_write_mask_size(write_mask
);
2108 if (mask_size
> 1) {
2109 shader_addline(buffer
, "ivec%d(floor(%s)));\n", mask_size
, src0_param
.param_str
);
2111 shader_addline(buffer
, "int(floor(%s)));\n", src0_param
.param_str
);
2114 else if(ins
->handler_idx
== WINED3DSIH_MOVA
)
2116 /* We need to *round* to the nearest int here. */
2117 unsigned int mask_size
= shader_glsl_get_write_mask_size(write_mask
);
2119 if (gl_info
->supported
[EXT_GPU_SHADER4
])
2122 shader_addline(buffer
, "ivec%d(round(%s)));\n", mask_size
, src0_param
.param_str
);
2124 shader_addline(buffer
, "int(round(%s)));\n", src0_param
.param_str
);
2129 shader_addline(buffer
, "ivec%d(floor(abs(%s) + vec%d(0.5)) * sign(%s)));\n",
2130 mask_size
, src0_param
.param_str
, mask_size
, src0_param
.param_str
);
2132 shader_addline(buffer
, "int(floor(abs(%s) + 0.5) * sign(%s)));\n",
2133 src0_param
.param_str
, src0_param
.param_str
);
2138 shader_addline(buffer
, "%s);\n", src0_param
.param_str
);
2142 /* Process the dot product operators DP3 and DP4 in GLSL (dst = dot(src0, src1)) */
2143 static void shader_glsl_dot(const struct wined3d_shader_instruction
*ins
)
2145 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2146 glsl_src_param_t src0_param
;
2147 glsl_src_param_t src1_param
;
2148 DWORD dst_write_mask
, src_write_mask
;
2149 unsigned int dst_size
= 0;
2151 dst_write_mask
= shader_glsl_append_dst(buffer
, ins
);
2152 dst_size
= shader_glsl_get_write_mask_size(dst_write_mask
);
2154 /* dp3 works on vec3, dp4 on vec4 */
2155 if (ins
->handler_idx
== WINED3DSIH_DP4
)
2157 src_write_mask
= WINED3DSP_WRITEMASK_ALL
;
2159 src_write_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2162 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_write_mask
, &src0_param
);
2163 shader_glsl_add_src_param(ins
, &ins
->src
[1], src_write_mask
, &src1_param
);
2166 shader_addline(buffer
, "vec%d(dot(%s, %s)));\n", dst_size
, src0_param
.param_str
, src1_param
.param_str
);
2168 shader_addline(buffer
, "dot(%s, %s));\n", src0_param
.param_str
, src1_param
.param_str
);
2172 /* Note that this instruction has some restrictions. The destination write mask
2173 * can't contain the w component, and the source swizzles have to be .xyzw */
2174 static void shader_glsl_cross(const struct wined3d_shader_instruction
*ins
)
2176 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
2177 glsl_src_param_t src0_param
;
2178 glsl_src_param_t src1_param
;
2181 shader_glsl_get_write_mask(&ins
->dst
[0], dst_mask
);
2182 shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2183 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
2184 shader_glsl_add_src_param(ins
, &ins
->src
[1], src_mask
, &src1_param
);
2185 shader_addline(ins
->ctx
->buffer
, "cross(%s, %s)%s);\n", src0_param
.param_str
, src1_param
.param_str
, dst_mask
);
2188 /* Process the WINED3DSIO_POW instruction in GLSL (dst = |src0|^src1)
2189 * Src0 and src1 are scalars. Note that D3D uses the absolute of src0, while
2190 * GLSL uses the value as-is. */
2191 static void shader_glsl_pow(const struct wined3d_shader_instruction
*ins
)
2193 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2194 glsl_src_param_t src0_param
;
2195 glsl_src_param_t src1_param
;
2196 DWORD dst_write_mask
;
2197 unsigned int dst_size
;
2199 dst_write_mask
= shader_glsl_append_dst(buffer
, ins
);
2200 dst_size
= shader_glsl_get_write_mask_size(dst_write_mask
);
2202 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2203 shader_glsl_add_src_param(ins
, &ins
->src
[1], WINED3DSP_WRITEMASK_0
, &src1_param
);
2207 shader_addline(buffer
, "vec%u(pow(abs(%s), %s)));\n",
2208 dst_size
, src0_param
.param_str
, src1_param
.param_str
);
2212 shader_addline(buffer
, "pow(abs(%s), %s));\n",
2213 src0_param
.param_str
, src1_param
.param_str
);
2217 /* Process the WINED3DSIO_LOG instruction in GLSL (dst = log2(|src0|))
2218 * Src0 is a scalar. Note that D3D uses the absolute of src0, while
2219 * GLSL uses the value as-is. */
2220 static void shader_glsl_log(const struct wined3d_shader_instruction
*ins
)
2222 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2223 glsl_src_param_t src0_param
;
2224 DWORD dst_write_mask
;
2225 unsigned int dst_size
;
2227 dst_write_mask
= shader_glsl_append_dst(buffer
, ins
);
2228 dst_size
= shader_glsl_get_write_mask_size(dst_write_mask
);
2230 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2234 shader_addline(buffer
, "vec%u(log2(abs(%s))));\n",
2235 dst_size
, src0_param
.param_str
);
2239 shader_addline(buffer
, "log2(abs(%s)));\n",
2240 src0_param
.param_str
);
2244 /* Map the opcode 1-to-1 to the GL code (arg->dst = instruction(src0, src1, ...) */
2245 static void shader_glsl_map2gl(const struct wined3d_shader_instruction
*ins
)
2247 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2248 glsl_src_param_t src_param
;
2249 const char *instruction
;
2253 /* Determine the GLSL function to use based on the opcode */
2254 /* TODO: Possibly make this a table for faster lookups */
2255 switch (ins
->handler_idx
)
2257 case WINED3DSIH_MIN
: instruction
= "min"; break;
2258 case WINED3DSIH_MAX
: instruction
= "max"; break;
2259 case WINED3DSIH_ABS
: instruction
= "abs"; break;
2260 case WINED3DSIH_FRC
: instruction
= "fract"; break;
2261 case WINED3DSIH_EXP
: instruction
= "exp2"; break;
2262 case WINED3DSIH_DSX
: instruction
= "dFdx"; break;
2263 case WINED3DSIH_DSY
: instruction
= "ycorrection.y * dFdy"; break;
2264 default: instruction
= "";
2265 FIXME("Opcode %#x not yet handled in GLSL\n", ins
->handler_idx
);
2269 write_mask
= shader_glsl_append_dst(buffer
, ins
);
2271 shader_addline(buffer
, "%s(", instruction
);
2275 shader_glsl_add_src_param(ins
, &ins
->src
[0], write_mask
, &src_param
);
2276 shader_addline(buffer
, "%s", src_param
.param_str
);
2277 for (i
= 1; i
< ins
->src_count
; ++i
)
2279 shader_glsl_add_src_param(ins
, &ins
->src
[i
], write_mask
, &src_param
);
2280 shader_addline(buffer
, ", %s", src_param
.param_str
);
2284 shader_addline(buffer
, "));\n");
2287 static void shader_glsl_nrm(const struct wined3d_shader_instruction
*ins
)
2289 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2290 glsl_src_param_t src_param
;
2291 unsigned int mask_size
;
2295 write_mask
= shader_glsl_get_write_mask(ins
->dst
, dst_mask
);
2296 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
2297 shader_glsl_add_src_param(ins
, &ins
->src
[0], write_mask
, &src_param
);
2299 shader_addline(buffer
, "tmp0.x = length(%s);\n", src_param
.param_str
);
2300 shader_glsl_append_dst(buffer
, ins
);
2303 shader_addline(buffer
, "tmp0.x == 0.0 ? vec%u(0.0) : (%s / tmp0.x));\n",
2304 mask_size
, src_param
.param_str
);
2308 shader_addline(buffer
, "tmp0.x == 0.0 ? 0.0 : (%s / tmp0.x));\n",
2309 src_param
.param_str
);
2313 /** Process the WINED3DSIO_EXPP instruction in GLSL:
2314 * For shader model 1.x, do the following (and honor the writemask, so use a temporary variable):
2315 * dst.x = 2^(floor(src))
2316 * dst.y = src - floor(src)
2317 * dst.z = 2^src (partial precision is allowed, but optional)
2319 * For 2.0 shaders, just do this (honoring writemask and swizzle):
2320 * dst = 2^src; (partial precision is allowed, but optional)
2322 static void shader_glsl_expp(const struct wined3d_shader_instruction
*ins
)
2324 glsl_src_param_t src_param
;
2326 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
, &src_param
);
2328 if (ins
->ctx
->reg_maps
->shader_version
.major
< 2)
2332 shader_addline(ins
->ctx
->buffer
, "tmp0.x = exp2(floor(%s));\n", src_param
.param_str
);
2333 shader_addline(ins
->ctx
->buffer
, "tmp0.y = %s - floor(%s);\n", src_param
.param_str
, src_param
.param_str
);
2334 shader_addline(ins
->ctx
->buffer
, "tmp0.z = exp2(%s);\n", src_param
.param_str
);
2335 shader_addline(ins
->ctx
->buffer
, "tmp0.w = 1.0;\n");
2337 shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2338 shader_glsl_get_write_mask(&ins
->dst
[0], dst_mask
);
2339 shader_addline(ins
->ctx
->buffer
, "tmp0%s);\n", dst_mask
);
2342 unsigned int mask_size
;
2344 write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2345 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
2347 if (mask_size
> 1) {
2348 shader_addline(ins
->ctx
->buffer
, "vec%d(exp2(%s)));\n", mask_size
, src_param
.param_str
);
2350 shader_addline(ins
->ctx
->buffer
, "exp2(%s));\n", src_param
.param_str
);
2355 /** Process the RCP (reciprocal or inverse) opcode in GLSL (dst = 1 / src) */
2356 static void shader_glsl_rcp(const struct wined3d_shader_instruction
*ins
)
2358 glsl_src_param_t src_param
;
2360 unsigned int mask_size
;
2362 write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2363 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
2364 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_3
, &src_param
);
2368 shader_addline(ins
->ctx
->buffer
, "vec%u(1.0 / %s));\n",
2369 mask_size
, src_param
.param_str
);
2373 shader_addline(ins
->ctx
->buffer
, "1.0 / %s);\n",
2374 src_param
.param_str
);
2378 static void shader_glsl_rsq(const struct wined3d_shader_instruction
*ins
)
2380 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2381 glsl_src_param_t src_param
;
2383 unsigned int mask_size
;
2385 write_mask
= shader_glsl_append_dst(buffer
, ins
);
2386 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
2388 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_3
, &src_param
);
2392 shader_addline(buffer
, "vec%u(inversesqrt(abs(%s))));\n",
2393 mask_size
, src_param
.param_str
);
2397 shader_addline(buffer
, "inversesqrt(abs(%s)));\n",
2398 src_param
.param_str
);
2402 /** Process signed comparison opcodes in GLSL. */
2403 static void shader_glsl_compare(const struct wined3d_shader_instruction
*ins
)
2405 glsl_src_param_t src0_param
;
2406 glsl_src_param_t src1_param
;
2408 unsigned int mask_size
;
2410 write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2411 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
2412 shader_glsl_add_src_param(ins
, &ins
->src
[0], write_mask
, &src0_param
);
2413 shader_glsl_add_src_param(ins
, &ins
->src
[1], write_mask
, &src1_param
);
2415 if (mask_size
> 1) {
2416 const char *compare
;
2418 switch(ins
->handler_idx
)
2420 case WINED3DSIH_SLT
: compare
= "lessThan"; break;
2421 case WINED3DSIH_SGE
: compare
= "greaterThanEqual"; break;
2422 default: compare
= "";
2423 FIXME("Can't handle opcode %#x\n", ins
->handler_idx
);
2426 shader_addline(ins
->ctx
->buffer
, "vec%d(%s(%s, %s)));\n", mask_size
, compare
,
2427 src0_param
.param_str
, src1_param
.param_str
);
2429 switch(ins
->handler_idx
)
2431 case WINED3DSIH_SLT
:
2432 /* Step(src0, src1) is not suitable here because if src0 == src1 SLT is supposed,
2433 * to return 0.0 but step returns 1.0 because step is not < x
2434 * An alternative is a bvec compare padded with an unused second component.
2435 * step(src1 * -1.0, src0 * -1.0) is not an option because it suffers from the same
2436 * issue. Playing with not() is not possible either because not() does not accept
2439 shader_addline(ins
->ctx
->buffer
, "(%s < %s) ? 1.0 : 0.0);\n",
2440 src0_param
.param_str
, src1_param
.param_str
);
2442 case WINED3DSIH_SGE
:
2443 /* Here we can use the step() function and safe a conditional */
2444 shader_addline(ins
->ctx
->buffer
, "step(%s, %s));\n", src1_param
.param_str
, src0_param
.param_str
);
2447 FIXME("Can't handle opcode %#x\n", ins
->handler_idx
);
2453 /** Process CMP instruction in GLSL (dst = src0 >= 0.0 ? src1 : src2), per channel */
2454 static void shader_glsl_cmp(const struct wined3d_shader_instruction
*ins
)
2456 glsl_src_param_t src0_param
;
2457 glsl_src_param_t src1_param
;
2458 glsl_src_param_t src2_param
;
2459 DWORD write_mask
, cmp_channel
= 0;
2462 BOOL temp_destination
= FALSE
;
2464 if (shader_is_scalar(&ins
->src
[0].reg
))
2466 write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2468 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_ALL
, &src0_param
);
2469 shader_glsl_add_src_param(ins
, &ins
->src
[1], write_mask
, &src1_param
);
2470 shader_glsl_add_src_param(ins
, &ins
->src
[2], write_mask
, &src2_param
);
2472 shader_addline(ins
->ctx
->buffer
, "%s >= 0.0 ? %s : %s);\n",
2473 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
2475 DWORD dst_mask
= ins
->dst
[0].write_mask
;
2476 struct wined3d_shader_dst_param dst
= ins
->dst
[0];
2478 /* Cycle through all source0 channels */
2479 for (i
=0; i
<4; i
++) {
2481 /* Find the destination channels which use the current source0 channel */
2482 for (j
=0; j
<4; j
++) {
2483 if (((ins
->src
[0].swizzle
>> (2 * j
)) & 0x3) == i
)
2485 write_mask
|= WINED3DSP_WRITEMASK_0
<< j
;
2486 cmp_channel
= WINED3DSP_WRITEMASK_0
<< j
;
2489 dst
.write_mask
= dst_mask
& write_mask
;
2491 /* Splitting the cmp instruction up in multiple lines imposes a problem:
2492 * The first lines may overwrite source parameters of the following lines.
2493 * Deal with that by using a temporary destination register if needed
2495 if ((ins
->src
[0].reg
.idx
== ins
->dst
[0].reg
.idx
2496 && ins
->src
[0].reg
.type
== ins
->dst
[0].reg
.type
)
2497 || (ins
->src
[1].reg
.idx
== ins
->dst
[0].reg
.idx
2498 && ins
->src
[1].reg
.type
== ins
->dst
[0].reg
.type
)
2499 || (ins
->src
[2].reg
.idx
== ins
->dst
[0].reg
.idx
2500 && ins
->src
[2].reg
.type
== ins
->dst
[0].reg
.type
))
2502 write_mask
= shader_glsl_get_write_mask(&dst
, mask_char
);
2503 if (!write_mask
) continue;
2504 shader_addline(ins
->ctx
->buffer
, "tmp0%s = (", mask_char
);
2505 temp_destination
= TRUE
;
2507 write_mask
= shader_glsl_append_dst_ext(ins
->ctx
->buffer
, ins
, &dst
);
2508 if (!write_mask
) continue;
2511 shader_glsl_add_src_param(ins
, &ins
->src
[0], cmp_channel
, &src0_param
);
2512 shader_glsl_add_src_param(ins
, &ins
->src
[1], write_mask
, &src1_param
);
2513 shader_glsl_add_src_param(ins
, &ins
->src
[2], write_mask
, &src2_param
);
2515 shader_addline(ins
->ctx
->buffer
, "%s >= 0.0 ? %s : %s);\n",
2516 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
2519 if(temp_destination
) {
2520 shader_glsl_get_write_mask(&ins
->dst
[0], mask_char
);
2521 shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2522 shader_addline(ins
->ctx
->buffer
, "tmp0%s);\n", mask_char
);
2528 /** Process the CND opcode in GLSL (dst = (src0 > 0.5) ? src1 : src2) */
2529 /* For ps 1.1-1.3, only a single component of src0 is used. For ps 1.4
2530 * the compare is done per component of src0. */
2531 static void shader_glsl_cnd(const struct wined3d_shader_instruction
*ins
)
2533 struct wined3d_shader_dst_param dst
;
2534 glsl_src_param_t src0_param
;
2535 glsl_src_param_t src1_param
;
2536 glsl_src_param_t src2_param
;
2537 DWORD write_mask
, cmp_channel
= 0;
2540 DWORD shader_version
= WINED3D_SHADER_VERSION(ins
->ctx
->reg_maps
->shader_version
.major
,
2541 ins
->ctx
->reg_maps
->shader_version
.minor
);
2543 if (shader_version
< WINED3D_SHADER_VERSION(1, 4))
2545 write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2546 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2547 shader_glsl_add_src_param(ins
, &ins
->src
[1], write_mask
, &src1_param
);
2548 shader_glsl_add_src_param(ins
, &ins
->src
[2], write_mask
, &src2_param
);
2550 /* Fun: The D3DSI_COISSUE flag changes the semantic of the cnd instruction for < 1.4 shaders */
2553 shader_addline(ins
->ctx
->buffer
, "%s /* COISSUE! */);\n", src1_param
.param_str
);
2555 shader_addline(ins
->ctx
->buffer
, "%s > 0.5 ? %s : %s);\n",
2556 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
2560 /* Cycle through all source0 channels */
2561 dst_mask
= ins
->dst
[0].write_mask
;
2563 for (i
=0; i
<4; i
++) {
2565 /* Find the destination channels which use the current source0 channel */
2566 for (j
=0; j
<4; j
++) {
2567 if (((ins
->src
[0].swizzle
>> (2 * j
)) & 0x3) == i
)
2569 write_mask
|= WINED3DSP_WRITEMASK_0
<< j
;
2570 cmp_channel
= WINED3DSP_WRITEMASK_0
<< j
;
2574 dst
.write_mask
= dst_mask
& write_mask
;
2575 write_mask
= shader_glsl_append_dst_ext(ins
->ctx
->buffer
, ins
, &dst
);
2576 if (!write_mask
) continue;
2578 shader_glsl_add_src_param(ins
, &ins
->src
[0], cmp_channel
, &src0_param
);
2579 shader_glsl_add_src_param(ins
, &ins
->src
[1], write_mask
, &src1_param
);
2580 shader_glsl_add_src_param(ins
, &ins
->src
[2], write_mask
, &src2_param
);
2582 shader_addline(ins
->ctx
->buffer
, "%s > 0.5 ? %s : %s);\n",
2583 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
2587 /** GLSL code generation for WINED3DSIO_MAD: Multiply the first 2 opcodes, then add the last */
2588 static void shader_glsl_mad(const struct wined3d_shader_instruction
*ins
)
2590 glsl_src_param_t src0_param
;
2591 glsl_src_param_t src1_param
;
2592 glsl_src_param_t src2_param
;
2595 write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2596 shader_glsl_add_src_param(ins
, &ins
->src
[0], write_mask
, &src0_param
);
2597 shader_glsl_add_src_param(ins
, &ins
->src
[1], write_mask
, &src1_param
);
2598 shader_glsl_add_src_param(ins
, &ins
->src
[2], write_mask
, &src2_param
);
2599 shader_addline(ins
->ctx
->buffer
, "(%s * %s) + %s);\n",
2600 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
2603 /* Handles transforming all WINED3DSIO_M?x? opcodes for
2604 Vertex shaders to GLSL codes */
2605 static void shader_glsl_mnxn(const struct wined3d_shader_instruction
*ins
)
2608 int nComponents
= 0;
2609 struct wined3d_shader_dst_param tmp_dst
= {{0}};
2610 struct wined3d_shader_src_param tmp_src
[2] = {{{0}}};
2611 struct wined3d_shader_instruction tmp_ins
;
2613 memset(&tmp_ins
, 0, sizeof(tmp_ins
));
2615 /* Set constants for the temporary argument */
2616 tmp_ins
.ctx
= ins
->ctx
;
2617 tmp_ins
.dst_count
= 1;
2618 tmp_ins
.dst
= &tmp_dst
;
2619 tmp_ins
.src_count
= 2;
2620 tmp_ins
.src
= tmp_src
;
2622 switch(ins
->handler_idx
)
2624 case WINED3DSIH_M4x4
:
2626 tmp_ins
.handler_idx
= WINED3DSIH_DP4
;
2628 case WINED3DSIH_M4x3
:
2630 tmp_ins
.handler_idx
= WINED3DSIH_DP4
;
2632 case WINED3DSIH_M3x4
:
2634 tmp_ins
.handler_idx
= WINED3DSIH_DP3
;
2636 case WINED3DSIH_M3x3
:
2638 tmp_ins
.handler_idx
= WINED3DSIH_DP3
;
2640 case WINED3DSIH_M3x2
:
2642 tmp_ins
.handler_idx
= WINED3DSIH_DP3
;
2648 tmp_dst
= ins
->dst
[0];
2649 tmp_src
[0] = ins
->src
[0];
2650 tmp_src
[1] = ins
->src
[1];
2651 for (i
= 0; i
< nComponents
; ++i
)
2653 tmp_dst
.write_mask
= WINED3DSP_WRITEMASK_0
<< i
;
2654 shader_glsl_dot(&tmp_ins
);
2655 ++tmp_src
[1].reg
.idx
;
2660 The LRP instruction performs a component-wise linear interpolation
2661 between the second and third operands using the first operand as the
2662 blend factor. Equation: (dst = src2 + src0 * (src1 - src2))
2663 This is equivalent to mix(src2, src1, src0);
2665 static void shader_glsl_lrp(const struct wined3d_shader_instruction
*ins
)
2667 glsl_src_param_t src0_param
;
2668 glsl_src_param_t src1_param
;
2669 glsl_src_param_t src2_param
;
2672 write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2674 shader_glsl_add_src_param(ins
, &ins
->src
[0], write_mask
, &src0_param
);
2675 shader_glsl_add_src_param(ins
, &ins
->src
[1], write_mask
, &src1_param
);
2676 shader_glsl_add_src_param(ins
, &ins
->src
[2], write_mask
, &src2_param
);
2678 shader_addline(ins
->ctx
->buffer
, "mix(%s, %s, %s));\n",
2679 src2_param
.param_str
, src1_param
.param_str
, src0_param
.param_str
);
2682 /** Process the WINED3DSIO_LIT instruction in GLSL:
2683 * dst.x = dst.w = 1.0
2684 * dst.y = (src0.x > 0) ? src0.x
2685 * dst.z = (src0.x > 0) ? ((src0.y > 0) ? pow(src0.y, src.w) : 0) : 0
2686 * where src.w is clamped at +- 128
2688 static void shader_glsl_lit(const struct wined3d_shader_instruction
*ins
)
2690 glsl_src_param_t src0_param
;
2691 glsl_src_param_t src1_param
;
2692 glsl_src_param_t src3_param
;
2695 shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2696 shader_glsl_get_write_mask(&ins
->dst
[0], dst_mask
);
2698 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2699 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_1
, &src1_param
);
2700 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_3
, &src3_param
);
2702 /* The sdk specifies the instruction like this
2704 * if(src.x > 0.0) dst.y = src.x
2706 * if(src.x > 0.0 && src.y > 0.0) dst.z = pow(src.y, power);
2710 * Obviously that has quite a few conditionals in it which we don't like. So the first step is this:
2711 * dst.x = 1.0 ... No further explanation needed
2712 * dst.y = max(src.y, 0.0); ... If x < 0.0, use 0.0, otherwise x. Same as the conditional
2713 * dst.z = x > 0.0 ? pow(max(y, 0.0), p) : 0; ... 0 ^ power is 0, and otherwise we use y anyway
2714 * dst.w = 1.0. ... Nothing fancy.
2716 * So we still have one conditional in there. So do this:
2717 * dst.z = pow(max(0.0, src.y) * step(0.0, src.x), power);
2719 * 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),
2720 * 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.
2721 * if both x and y are > 0, we get pow(y * 1.0, power), as it is supposed to
2723 shader_addline(ins
->ctx
->buffer
,
2724 "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",
2725 src0_param
.param_str
, src1_param
.param_str
, src0_param
.param_str
, src3_param
.param_str
, dst_mask
);
2728 /** Process the WINED3DSIO_DST instruction in GLSL:
2730 * dst.y = src0.x * src0.y
2734 static void shader_glsl_dst(const struct wined3d_shader_instruction
*ins
)
2736 glsl_src_param_t src0y_param
;
2737 glsl_src_param_t src0z_param
;
2738 glsl_src_param_t src1y_param
;
2739 glsl_src_param_t src1w_param
;
2742 shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2743 shader_glsl_get_write_mask(&ins
->dst
[0], dst_mask
);
2745 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_1
, &src0y_param
);
2746 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_2
, &src0z_param
);
2747 shader_glsl_add_src_param(ins
, &ins
->src
[1], WINED3DSP_WRITEMASK_1
, &src1y_param
);
2748 shader_glsl_add_src_param(ins
, &ins
->src
[1], WINED3DSP_WRITEMASK_3
, &src1w_param
);
2750 shader_addline(ins
->ctx
->buffer
, "vec4(1.0, %s * %s, %s, %s))%s;\n",
2751 src0y_param
.param_str
, src1y_param
.param_str
, src0z_param
.param_str
, src1w_param
.param_str
, dst_mask
);
2754 /** Process the WINED3DSIO_SINCOS instruction in GLSL:
2755 * VS 2.0 requires that specific cosine and sine constants be passed to this instruction so the hardware
2756 * can handle it. But, these functions are built-in for GLSL, so we can just ignore the last 2 params.
2758 * dst.x = cos(src0.?)
2759 * dst.y = sin(src0.?)
2763 static void shader_glsl_sincos(const struct wined3d_shader_instruction
*ins
)
2765 glsl_src_param_t src0_param
;
2768 write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2769 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2771 switch (write_mask
) {
2772 case WINED3DSP_WRITEMASK_0
:
2773 shader_addline(ins
->ctx
->buffer
, "cos(%s));\n", src0_param
.param_str
);
2776 case WINED3DSP_WRITEMASK_1
:
2777 shader_addline(ins
->ctx
->buffer
, "sin(%s));\n", src0_param
.param_str
);
2780 case (WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
):
2781 shader_addline(ins
->ctx
->buffer
, "vec2(cos(%s), sin(%s)));\n", src0_param
.param_str
, src0_param
.param_str
);
2785 ERR("Write mask should be .x, .y or .xy\n");
2790 /* sgn in vs_2_0 has 2 extra parameters(registers for temporary storage) which we don't use
2791 * here. But those extra parameters require a dedicated function for sgn, since map2gl would
2792 * generate invalid code
2794 static void shader_glsl_sgn(const struct wined3d_shader_instruction
*ins
)
2796 glsl_src_param_t src0_param
;
2799 write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
2800 shader_glsl_add_src_param(ins
, &ins
->src
[0], write_mask
, &src0_param
);
2802 shader_addline(ins
->ctx
->buffer
, "sign(%s));\n", src0_param
.param_str
);
2805 /** Process the WINED3DSIO_LOOP instruction in GLSL:
2806 * Start a for() loop where src1.y is the initial value of aL,
2807 * increment aL by src1.z for a total of src1.x iterations.
2808 * Need to use a temporary variable for this operation.
2810 /* FIXME: I don't think nested loops will work correctly this way. */
2811 static void shader_glsl_loop(const struct wined3d_shader_instruction
*ins
)
2813 glsl_src_param_t src1_param
;
2814 IWineD3DBaseShaderImpl
*shader
= (IWineD3DBaseShaderImpl
*)ins
->ctx
->shader
;
2815 const DWORD
*control_values
= NULL
;
2816 const local_constant
*constant
;
2818 shader_glsl_add_src_param(ins
, &ins
->src
[1], WINED3DSP_WRITEMASK_ALL
, &src1_param
);
2820 /* Try to hardcode the loop control parameters if possible. Direct3D 9 class hardware doesn't support real
2821 * varying indexing, but Microsoft designed this feature for Shader model 2.x+. If the loop control is
2822 * known at compile time, the GLSL compiler can unroll the loop, and replace indirect addressing with direct
2825 if (ins
->src
[1].reg
.type
== WINED3DSPR_CONSTINT
)
2827 LIST_FOR_EACH_ENTRY(constant
, &shader
->baseShader
.constantsI
, local_constant
, entry
) {
2828 if (constant
->idx
== ins
->src
[1].reg
.idx
)
2830 control_values
= constant
->value
;
2838 struct wined3d_shader_loop_control loop_control
;
2839 loop_control
.count
= control_values
[0];
2840 loop_control
.start
= control_values
[1];
2841 loop_control
.step
= (int)control_values
[2];
2843 if (loop_control
.step
> 0)
2845 shader_addline(ins
->ctx
->buffer
, "for (aL%u = %u; aL%u < (%u * %d + %u); aL%u += %d) {\n",
2846 shader
->baseShader
.cur_loop_depth
, loop_control
.start
,
2847 shader
->baseShader
.cur_loop_depth
, loop_control
.count
, loop_control
.step
, loop_control
.start
,
2848 shader
->baseShader
.cur_loop_depth
, loop_control
.step
);
2850 else if (loop_control
.step
< 0)
2852 shader_addline(ins
->ctx
->buffer
, "for (aL%u = %u; aL%u > (%u * %d + %u); aL%u += %d) {\n",
2853 shader
->baseShader
.cur_loop_depth
, loop_control
.start
,
2854 shader
->baseShader
.cur_loop_depth
, loop_control
.count
, loop_control
.step
, loop_control
.start
,
2855 shader
->baseShader
.cur_loop_depth
, loop_control
.step
);
2859 shader_addline(ins
->ctx
->buffer
, "for (aL%u = %u, tmpInt%u = 0; tmpInt%u < %u; tmpInt%u++) {\n",
2860 shader
->baseShader
.cur_loop_depth
, loop_control
.start
, shader
->baseShader
.cur_loop_depth
,
2861 shader
->baseShader
.cur_loop_depth
, loop_control
.count
,
2862 shader
->baseShader
.cur_loop_depth
);
2865 shader_addline(ins
->ctx
->buffer
,
2866 "for (tmpInt%u = 0, aL%u = %s.y; tmpInt%u < %s.x; tmpInt%u++, aL%u += %s.z) {\n",
2867 shader
->baseShader
.cur_loop_depth
, shader
->baseShader
.cur_loop_regno
,
2868 src1_param
.reg_name
, shader
->baseShader
.cur_loop_depth
, src1_param
.reg_name
,
2869 shader
->baseShader
.cur_loop_depth
, shader
->baseShader
.cur_loop_regno
, src1_param
.reg_name
);
2872 shader
->baseShader
.cur_loop_depth
++;
2873 shader
->baseShader
.cur_loop_regno
++;
2876 static void shader_glsl_end(const struct wined3d_shader_instruction
*ins
)
2878 IWineD3DBaseShaderImpl
*shader
= (IWineD3DBaseShaderImpl
*)ins
->ctx
->shader
;
2880 shader_addline(ins
->ctx
->buffer
, "}\n");
2882 if (ins
->handler_idx
== WINED3DSIH_ENDLOOP
)
2884 shader
->baseShader
.cur_loop_depth
--;
2885 shader
->baseShader
.cur_loop_regno
--;
2888 if (ins
->handler_idx
== WINED3DSIH_ENDREP
)
2890 shader
->baseShader
.cur_loop_depth
--;
2894 static void shader_glsl_rep(const struct wined3d_shader_instruction
*ins
)
2896 IWineD3DBaseShaderImpl
*shader
= (IWineD3DBaseShaderImpl
*)ins
->ctx
->shader
;
2897 glsl_src_param_t src0_param
;
2898 const DWORD
*control_values
= NULL
;
2899 const local_constant
*constant
;
2901 /* Try to hardcode local values to help the GLSL compiler to unroll and optimize the loop */
2902 if (ins
->src
[0].reg
.type
== WINED3DSPR_CONSTINT
)
2904 LIST_FOR_EACH_ENTRY(constant
, &shader
->baseShader
.constantsI
, local_constant
, entry
)
2906 if (constant
->idx
== ins
->src
[0].reg
.idx
)
2908 control_values
= constant
->value
;
2914 if(control_values
) {
2915 shader_addline(ins
->ctx
->buffer
, "for (tmpInt%d = 0; tmpInt%d < %d; tmpInt%d++) {\n",
2916 shader
->baseShader
.cur_loop_depth
, shader
->baseShader
.cur_loop_depth
,
2917 control_values
[0], shader
->baseShader
.cur_loop_depth
);
2919 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2920 shader_addline(ins
->ctx
->buffer
, "for (tmpInt%d = 0; tmpInt%d < %s; tmpInt%d++) {\n",
2921 shader
->baseShader
.cur_loop_depth
, shader
->baseShader
.cur_loop_depth
,
2922 src0_param
.param_str
, shader
->baseShader
.cur_loop_depth
);
2924 shader
->baseShader
.cur_loop_depth
++;
2927 static void shader_glsl_if(const struct wined3d_shader_instruction
*ins
)
2929 glsl_src_param_t src0_param
;
2931 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2932 shader_addline(ins
->ctx
->buffer
, "if (%s) {\n", src0_param
.param_str
);
2935 static void shader_glsl_ifc(const struct wined3d_shader_instruction
*ins
)
2937 glsl_src_param_t src0_param
;
2938 glsl_src_param_t src1_param
;
2940 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2941 shader_glsl_add_src_param(ins
, &ins
->src
[1], WINED3DSP_WRITEMASK_0
, &src1_param
);
2943 shader_addline(ins
->ctx
->buffer
, "if (%s %s %s) {\n",
2944 src0_param
.param_str
, shader_get_comp_op(ins
->flags
), src1_param
.param_str
);
2947 static void shader_glsl_else(const struct wined3d_shader_instruction
*ins
)
2949 shader_addline(ins
->ctx
->buffer
, "} else {\n");
2952 static void shader_glsl_break(const struct wined3d_shader_instruction
*ins
)
2954 shader_addline(ins
->ctx
->buffer
, "break;\n");
2957 /* FIXME: According to MSDN the compare is done per component. */
2958 static void shader_glsl_breakc(const struct wined3d_shader_instruction
*ins
)
2960 glsl_src_param_t src0_param
;
2961 glsl_src_param_t src1_param
;
2963 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
, &src0_param
);
2964 shader_glsl_add_src_param(ins
, &ins
->src
[1], WINED3DSP_WRITEMASK_0
, &src1_param
);
2966 shader_addline(ins
->ctx
->buffer
, "if (%s %s %s) break;\n",
2967 src0_param
.param_str
, shader_get_comp_op(ins
->flags
), src1_param
.param_str
);
2970 static void shader_glsl_label(const struct wined3d_shader_instruction
*ins
)
2972 shader_addline(ins
->ctx
->buffer
, "}\n");
2973 shader_addline(ins
->ctx
->buffer
, "void subroutine%u () {\n", ins
->src
[0].reg
.idx
);
2976 static void shader_glsl_call(const struct wined3d_shader_instruction
*ins
)
2978 shader_addline(ins
->ctx
->buffer
, "subroutine%u();\n", ins
->src
[0].reg
.idx
);
2981 static void shader_glsl_callnz(const struct wined3d_shader_instruction
*ins
)
2983 glsl_src_param_t src1_param
;
2985 shader_glsl_add_src_param(ins
, &ins
->src
[1], WINED3DSP_WRITEMASK_0
, &src1_param
);
2986 shader_addline(ins
->ctx
->buffer
, "if (%s) subroutine%u();\n", src1_param
.param_str
, ins
->src
[0].reg
.idx
);
2989 static void shader_glsl_ret(const struct wined3d_shader_instruction
*ins
)
2991 /* No-op. The closing } is written when a new function is started, and at the end of the shader. This
2992 * function only suppresses the unhandled instruction warning
2996 /*********************************************
2997 * Pixel Shader Specific Code begins here
2998 ********************************************/
2999 static void shader_glsl_tex(const struct wined3d_shader_instruction
*ins
)
3001 IWineD3DBaseShaderImpl
*shader
= (IWineD3DBaseShaderImpl
*)ins
->ctx
->shader
;
3002 IWineD3DDeviceImpl
*deviceImpl
= (IWineD3DDeviceImpl
*)shader
->baseShader
.device
;
3003 DWORD shader_version
= WINED3D_SHADER_VERSION(ins
->ctx
->reg_maps
->shader_version
.major
,
3004 ins
->ctx
->reg_maps
->shader_version
.minor
);
3005 glsl_sample_function_t sample_function
;
3006 DWORD sample_flags
= 0;
3008 DWORD mask
= 0, swizzle
;
3010 /* 1.0-1.4: Use destination register as sampler source.
3011 * 2.0+: Use provided sampler source. */
3012 if (shader_version
< WINED3D_SHADER_VERSION(2,0)) sampler_idx
= ins
->dst
[0].reg
.idx
;
3013 else sampler_idx
= ins
->src
[1].reg
.idx
;
3015 if (shader_version
< WINED3D_SHADER_VERSION(1,4))
3017 DWORD flags
= deviceImpl
->stateBlock
->textureState
[sampler_idx
][WINED3DTSS_TEXTURETRANSFORMFLAGS
];
3018 WINED3DSAMPLER_TEXTURE_TYPE sampler_type
= ins
->ctx
->reg_maps
->sampler_type
[sampler_idx
];
3020 /* Projected cube textures don't make a lot of sense, the resulting coordinates stay the same. */
3021 if (flags
& WINED3DTTFF_PROJECTED
&& sampler_type
!= WINED3DSTT_CUBE
) {
3022 sample_flags
|= WINED3D_GLSL_SAMPLE_PROJECTED
;
3023 switch (flags
& ~WINED3DTTFF_PROJECTED
) {
3024 case WINED3DTTFF_COUNT1
: FIXME("WINED3DTTFF_PROJECTED with WINED3DTTFF_COUNT1?\n"); break;
3025 case WINED3DTTFF_COUNT2
: mask
= WINED3DSP_WRITEMASK_1
; break;
3026 case WINED3DTTFF_COUNT3
: mask
= WINED3DSP_WRITEMASK_2
; break;
3027 case WINED3DTTFF_COUNT4
:
3028 case WINED3DTTFF_DISABLE
: mask
= WINED3DSP_WRITEMASK_3
; break;
3032 else if (shader_version
< WINED3D_SHADER_VERSION(2,0))
3034 DWORD src_mod
= ins
->src
[0].modifiers
;
3036 if (src_mod
== WINED3DSPSM_DZ
) {
3037 sample_flags
|= WINED3D_GLSL_SAMPLE_PROJECTED
;
3038 mask
= WINED3DSP_WRITEMASK_2
;
3039 } else if (src_mod
== WINED3DSPSM_DW
) {
3040 sample_flags
|= WINED3D_GLSL_SAMPLE_PROJECTED
;
3041 mask
= WINED3DSP_WRITEMASK_3
;
3044 if (ins
->flags
& WINED3DSI_TEXLD_PROJECT
)
3046 /* ps 2.0 texldp instruction always divides by the fourth component. */
3047 sample_flags
|= WINED3D_GLSL_SAMPLE_PROJECTED
;
3048 mask
= WINED3DSP_WRITEMASK_3
;
3052 if(deviceImpl
->stateBlock
->textures
[sampler_idx
] &&
3053 IWineD3DBaseTexture_GetTextureDimensions(deviceImpl
->stateBlock
->textures
[sampler_idx
]) == GL_TEXTURE_RECTANGLE_ARB
) {
3054 sample_flags
|= WINED3D_GLSL_SAMPLE_RECT
;
3057 shader_glsl_get_sample_function(ins
->ctx
, sampler_idx
, sample_flags
, &sample_function
);
3058 mask
|= sample_function
.coord_mask
;
3060 if (shader_version
< WINED3D_SHADER_VERSION(2,0)) swizzle
= WINED3DSP_NOSWIZZLE
;
3061 else swizzle
= ins
->src
[1].swizzle
;
3063 /* 1.0-1.3: Use destination register as coordinate source.
3064 1.4+: Use provided coordinate source register. */
3065 if (shader_version
< WINED3D_SHADER_VERSION(1,4))
3068 shader_glsl_write_mask_to_str(mask
, coord_mask
);
3069 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, swizzle
, NULL
, NULL
, NULL
,
3070 "T%u%s", sampler_idx
, coord_mask
);
3072 glsl_src_param_t coord_param
;
3073 shader_glsl_add_src_param(ins
, &ins
->src
[0], mask
, &coord_param
);
3074 if (ins
->flags
& WINED3DSI_TEXLD_BIAS
)
3076 glsl_src_param_t bias
;
3077 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_3
, &bias
);
3078 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, swizzle
, NULL
, NULL
, bias
.param_str
,
3079 "%s", coord_param
.param_str
);
3081 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, swizzle
, NULL
, NULL
, NULL
,
3082 "%s", coord_param
.param_str
);
3087 static void shader_glsl_texldd(const struct wined3d_shader_instruction
*ins
)
3089 IWineD3DBaseShaderImpl
*This
= (IWineD3DBaseShaderImpl
*)ins
->ctx
->shader
;
3090 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
3091 const struct wined3d_gl_info
*gl_info
= ins
->ctx
->gl_info
;
3092 glsl_sample_function_t sample_function
;
3093 glsl_src_param_t coord_param
, dx_param
, dy_param
;
3094 DWORD sample_flags
= WINED3D_GLSL_SAMPLE_GRAD
;
3096 DWORD swizzle
= ins
->src
[1].swizzle
;
3098 if (!gl_info
->supported
[ARB_SHADER_TEXTURE_LOD
] && !gl_info
->supported
[EXT_GPU_SHADER4
])
3100 FIXME("texldd used, but not supported by hardware. Falling back to regular tex\n");
3101 return shader_glsl_tex(ins
);
3104 sampler_idx
= ins
->src
[1].reg
.idx
;
3105 if(deviceImpl
->stateBlock
->textures
[sampler_idx
] &&
3106 IWineD3DBaseTexture_GetTextureDimensions(deviceImpl
->stateBlock
->textures
[sampler_idx
]) == GL_TEXTURE_RECTANGLE_ARB
) {
3107 sample_flags
|= WINED3D_GLSL_SAMPLE_RECT
;
3110 shader_glsl_get_sample_function(ins
->ctx
, sampler_idx
, sample_flags
, &sample_function
);
3111 shader_glsl_add_src_param(ins
, &ins
->src
[0], sample_function
.coord_mask
, &coord_param
);
3112 shader_glsl_add_src_param(ins
, &ins
->src
[2], sample_function
.coord_mask
, &dx_param
);
3113 shader_glsl_add_src_param(ins
, &ins
->src
[3], sample_function
.coord_mask
, &dy_param
);
3115 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, swizzle
, dx_param
.param_str
, dy_param
.param_str
, NULL
,
3116 "%s", coord_param
.param_str
);
3119 static void shader_glsl_texldl(const struct wined3d_shader_instruction
*ins
)
3121 IWineD3DBaseShaderImpl
*This
= (IWineD3DBaseShaderImpl
*)ins
->ctx
->shader
;
3122 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
3123 const struct wined3d_gl_info
*gl_info
= ins
->ctx
->gl_info
;
3124 glsl_sample_function_t sample_function
;
3125 glsl_src_param_t coord_param
, lod_param
;
3126 DWORD sample_flags
= WINED3D_GLSL_SAMPLE_LOD
;
3128 DWORD swizzle
= ins
->src
[1].swizzle
;
3130 sampler_idx
= ins
->src
[1].reg
.idx
;
3131 if(deviceImpl
->stateBlock
->textures
[sampler_idx
] &&
3132 IWineD3DBaseTexture_GetTextureDimensions(deviceImpl
->stateBlock
->textures
[sampler_idx
]) == GL_TEXTURE_RECTANGLE_ARB
) {
3133 sample_flags
|= WINED3D_GLSL_SAMPLE_RECT
;
3135 shader_glsl_get_sample_function(ins
->ctx
, sampler_idx
, sample_flags
, &sample_function
);
3136 shader_glsl_add_src_param(ins
, &ins
->src
[0], sample_function
.coord_mask
, &coord_param
);
3138 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_3
, &lod_param
);
3140 if (!gl_info
->supported
[ARB_SHADER_TEXTURE_LOD
] && !gl_info
->supported
[EXT_GPU_SHADER4
]
3141 && shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
.type
))
3143 /* The GLSL spec claims the Lod sampling functions are only supported in vertex shaders.
3144 * However, they seem to work just fine in fragment shaders as well. */
3145 WARN("Using %s in fragment shader.\n", sample_function
.name
);
3147 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, swizzle
, NULL
, NULL
, lod_param
.param_str
,
3148 "%s", coord_param
.param_str
);
3151 static void shader_glsl_texcoord(const struct wined3d_shader_instruction
*ins
)
3153 /* FIXME: Make this work for more than just 2D textures */
3154 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
3155 DWORD write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
3157 if (!(ins
->ctx
->reg_maps
->shader_version
.major
== 1 && ins
->ctx
->reg_maps
->shader_version
.minor
== 4))
3161 shader_glsl_get_write_mask(&ins
->dst
[0], dst_mask
);
3162 shader_addline(buffer
, "clamp(gl_TexCoord[%u], 0.0, 1.0)%s);\n",
3163 ins
->dst
[0].reg
.idx
, dst_mask
);
3165 DWORD reg
= ins
->src
[0].reg
.idx
;
3166 DWORD src_mod
= ins
->src
[0].modifiers
;
3167 char dst_swizzle
[6];
3169 shader_glsl_get_swizzle(&ins
->src
[0], FALSE
, write_mask
, dst_swizzle
);
3171 if (src_mod
== WINED3DSPSM_DZ
) {
3172 glsl_src_param_t div_param
;
3173 unsigned int mask_size
= shader_glsl_get_write_mask_size(write_mask
);
3174 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_2
, &div_param
);
3176 if (mask_size
> 1) {
3177 shader_addline(buffer
, "gl_TexCoord[%u]%s / vec%d(%s));\n", reg
, dst_swizzle
, mask_size
, div_param
.param_str
);
3179 shader_addline(buffer
, "gl_TexCoord[%u]%s / %s);\n", reg
, dst_swizzle
, div_param
.param_str
);
3181 } else if (src_mod
== WINED3DSPSM_DW
) {
3182 glsl_src_param_t div_param
;
3183 unsigned int mask_size
= shader_glsl_get_write_mask_size(write_mask
);
3184 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_3
, &div_param
);
3186 if (mask_size
> 1) {
3187 shader_addline(buffer
, "gl_TexCoord[%u]%s / vec%d(%s));\n", reg
, dst_swizzle
, mask_size
, div_param
.param_str
);
3189 shader_addline(buffer
, "gl_TexCoord[%u]%s / %s);\n", reg
, dst_swizzle
, div_param
.param_str
);
3192 shader_addline(buffer
, "gl_TexCoord[%u]%s);\n", reg
, dst_swizzle
);
3197 /** Process the WINED3DSIO_TEXDP3TEX instruction in GLSL:
3198 * Take a 3-component dot product of the TexCoord[dstreg] and src,
3199 * then perform a 1D texture lookup from stage dstregnum, place into dst. */
3200 static void shader_glsl_texdp3tex(const struct wined3d_shader_instruction
*ins
)
3202 glsl_src_param_t src0_param
;
3203 glsl_sample_function_t sample_function
;
3204 DWORD sampler_idx
= ins
->dst
[0].reg
.idx
;
3205 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
3208 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
3210 /* Do I have to take care about the projected bit? I don't think so, since the dp3 returns only one
3211 * scalar, and projected sampling would require 4.
3213 * It is a dependent read - not valid with conditional NP2 textures
3215 shader_glsl_get_sample_function(ins
->ctx
, sampler_idx
, 0, &sample_function
);
3216 mask_size
= shader_glsl_get_write_mask_size(sample_function
.coord_mask
);
3221 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, WINED3DSP_NOSWIZZLE
, NULL
, NULL
, NULL
,
3222 "dot(gl_TexCoord[%u].xyz, %s)", sampler_idx
, src0_param
.param_str
);
3226 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, WINED3DSP_NOSWIZZLE
, NULL
, NULL
, NULL
,
3227 "vec2(dot(gl_TexCoord[%u].xyz, %s), 0.0)", sampler_idx
, src0_param
.param_str
);
3231 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, WINED3DSP_NOSWIZZLE
, NULL
, NULL
, NULL
,
3232 "vec3(dot(gl_TexCoord[%u].xyz, %s), 0.0, 0.0)", sampler_idx
, src0_param
.param_str
);
3236 FIXME("Unexpected mask size %u\n", mask_size
);
3241 /** Process the WINED3DSIO_TEXDP3 instruction in GLSL:
3242 * Take a 3-component dot product of the TexCoord[dstreg] and src. */
3243 static void shader_glsl_texdp3(const struct wined3d_shader_instruction
*ins
)
3245 glsl_src_param_t src0_param
;
3246 DWORD dstreg
= ins
->dst
[0].reg
.idx
;
3247 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
3249 unsigned int mask_size
;
3251 dst_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
3252 mask_size
= shader_glsl_get_write_mask_size(dst_mask
);
3253 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
3255 if (mask_size
> 1) {
3256 shader_addline(ins
->ctx
->buffer
, "vec%d(dot(T%u.xyz, %s)));\n", mask_size
, dstreg
, src0_param
.param_str
);
3258 shader_addline(ins
->ctx
->buffer
, "dot(T%u.xyz, %s));\n", dstreg
, src0_param
.param_str
);
3262 /** Process the WINED3DSIO_TEXDEPTH instruction in GLSL:
3263 * Calculate the depth as dst.x / dst.y */
3264 static void shader_glsl_texdepth(const struct wined3d_shader_instruction
*ins
)
3266 glsl_dst_param_t dst_param
;
3268 shader_glsl_add_dst_param(ins
, &ins
->dst
[0], &dst_param
);
3270 /* Tests show that texdepth never returns anything below 0.0, and that r5.y is clamped to 1.0.
3271 * Negative input is accepted, -0.25 / -0.5 returns 0.5. GL should clamp gl_FragDepth to [0;1], but
3272 * this doesn't always work, so clamp the results manually. Whether or not the x value is clamped at 1
3273 * too is irrelevant, since if x = 0, any y value < 1.0 (and > 1.0 is not allowed) results in a result
3276 shader_addline(ins
->ctx
->buffer
, "gl_FragDepth = clamp((%s.x / min(%s.y, 1.0)), 0.0, 1.0);\n",
3277 dst_param
.reg_name
, dst_param
.reg_name
);
3280 /** Process the WINED3DSIO_TEXM3X2DEPTH instruction in GLSL:
3281 * Last row of a 3x2 matrix multiply, use the result to calculate the depth:
3282 * Calculate tmp0.y = TexCoord[dstreg] . src.xyz; (tmp0.x has already been calculated)
3283 * depth = (tmp0.y == 0.0) ? 1.0 : tmp0.x / tmp0.y
3285 static void shader_glsl_texm3x2depth(const struct wined3d_shader_instruction
*ins
)
3287 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
3288 DWORD dstreg
= ins
->dst
[0].reg
.idx
;
3289 glsl_src_param_t src0_param
;
3291 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
3293 shader_addline(ins
->ctx
->buffer
, "tmp0.y = dot(T%u.xyz, %s);\n", dstreg
, src0_param
.param_str
);
3294 shader_addline(ins
->ctx
->buffer
, "gl_FragDepth = (tmp0.y == 0.0) ? 1.0 : clamp(tmp0.x / tmp0.y, 0.0, 1.0);\n");
3297 /** Process the WINED3DSIO_TEXM3X2PAD instruction in GLSL
3298 * Calculate the 1st of a 2-row matrix multiplication. */
3299 static void shader_glsl_texm3x2pad(const struct wined3d_shader_instruction
*ins
)
3301 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
3302 DWORD reg
= ins
->dst
[0].reg
.idx
;
3303 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
3304 glsl_src_param_t src0_param
;
3306 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
3307 shader_addline(buffer
, "tmp0.x = dot(T%u.xyz, %s);\n", reg
, src0_param
.param_str
);
3310 /** Process the WINED3DSIO_TEXM3X3PAD instruction in GLSL
3311 * Calculate the 1st or 2nd row of a 3-row matrix multiplication. */
3312 static void shader_glsl_texm3x3pad(const struct wined3d_shader_instruction
*ins
)
3314 IWineD3DBaseShaderImpl
*shader
= (IWineD3DBaseShaderImpl
*)ins
->ctx
->shader
;
3315 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
3316 DWORD reg
= ins
->dst
[0].reg
.idx
;
3317 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
3318 SHADER_PARSE_STATE
* current_state
= &shader
->baseShader
.parse_state
;
3319 glsl_src_param_t src0_param
;
3321 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
3322 shader_addline(buffer
, "tmp0.%c = dot(T%u.xyz, %s);\n", 'x' + current_state
->current_row
, reg
, src0_param
.param_str
);
3323 current_state
->texcoord_w
[current_state
->current_row
++] = reg
;
3326 static void shader_glsl_texm3x2tex(const struct wined3d_shader_instruction
*ins
)
3328 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
3329 DWORD reg
= ins
->dst
[0].reg
.idx
;
3330 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
3331 glsl_src_param_t src0_param
;
3332 glsl_sample_function_t sample_function
;
3334 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
3335 shader_addline(buffer
, "tmp0.y = dot(T%u.xyz, %s);\n", reg
, src0_param
.param_str
);
3337 shader_glsl_get_sample_function(ins
->ctx
, reg
, 0, &sample_function
);
3339 /* Sample the texture using the calculated coordinates */
3340 shader_glsl_gen_sample_code(ins
, reg
, &sample_function
, WINED3DSP_NOSWIZZLE
, NULL
, NULL
, NULL
, "tmp0.xy");
3343 /** Process the WINED3DSIO_TEXM3X3TEX instruction in GLSL
3344 * Perform the 3rd row of a 3x3 matrix multiply, then sample the texture using the calculated coordinates */
3345 static void shader_glsl_texm3x3tex(const struct wined3d_shader_instruction
*ins
)
3347 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
3348 IWineD3DBaseShaderImpl
*shader
= (IWineD3DBaseShaderImpl
*)ins
->ctx
->shader
;
3349 SHADER_PARSE_STATE
*current_state
= &shader
->baseShader
.parse_state
;
3350 glsl_src_param_t src0_param
;
3351 DWORD reg
= ins
->dst
[0].reg
.idx
;
3352 glsl_sample_function_t sample_function
;
3354 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
3355 shader_addline(ins
->ctx
->buffer
, "tmp0.z = dot(T%u.xyz, %s);\n", reg
, src0_param
.param_str
);
3357 /* Dependent read, not valid with conditional NP2 */
3358 shader_glsl_get_sample_function(ins
->ctx
, reg
, 0, &sample_function
);
3360 /* Sample the texture using the calculated coordinates */
3361 shader_glsl_gen_sample_code(ins
, reg
, &sample_function
, WINED3DSP_NOSWIZZLE
, NULL
, NULL
, NULL
, "tmp0.xyz");
3363 current_state
->current_row
= 0;
3366 /** Process the WINED3DSIO_TEXM3X3 instruction in GLSL
3367 * Perform the 3rd row of a 3x3 matrix multiply */
3368 static void shader_glsl_texm3x3(const struct wined3d_shader_instruction
*ins
)
3370 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
3371 IWineD3DBaseShaderImpl
*shader
= (IWineD3DBaseShaderImpl
*)ins
->ctx
->shader
;
3372 SHADER_PARSE_STATE
*current_state
= &shader
->baseShader
.parse_state
;
3373 glsl_src_param_t src0_param
;
3375 DWORD reg
= ins
->dst
[0].reg
.idx
;
3377 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
3379 shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
3380 shader_glsl_get_write_mask(&ins
->dst
[0], dst_mask
);
3381 shader_addline(ins
->ctx
->buffer
, "vec4(tmp0.xy, dot(T%u.xyz, %s), 1.0)%s);\n", reg
, src0_param
.param_str
, dst_mask
);
3383 current_state
->current_row
= 0;
3386 /* Process the WINED3DSIO_TEXM3X3SPEC instruction in GLSL
3387 * Perform the final texture lookup based on the previous 2 3x3 matrix multiplies */
3388 static void shader_glsl_texm3x3spec(const struct wined3d_shader_instruction
*ins
)
3390 IWineD3DBaseShaderImpl
*shader
= (IWineD3DBaseShaderImpl
*)ins
->ctx
->shader
;
3391 DWORD reg
= ins
->dst
[0].reg
.idx
;
3392 glsl_src_param_t src0_param
;
3393 glsl_src_param_t src1_param
;
3394 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
3395 SHADER_PARSE_STATE
* current_state
= &shader
->baseShader
.parse_state
;
3396 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
3397 glsl_sample_function_t sample_function
;
3399 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
3400 shader_glsl_add_src_param(ins
, &ins
->src
[1], src_mask
, &src1_param
);
3402 /* Perform the last matrix multiply operation */
3403 shader_addline(buffer
, "tmp0.z = dot(T%u.xyz, %s);\n", reg
, src0_param
.param_str
);
3404 /* Reflection calculation */
3405 shader_addline(buffer
, "tmp0.xyz = -reflect((%s), normalize(tmp0.xyz));\n", src1_param
.param_str
);
3407 /* Dependent read, not valid with conditional NP2 */
3408 shader_glsl_get_sample_function(ins
->ctx
, reg
, 0, &sample_function
);
3410 /* Sample the texture */
3411 shader_glsl_gen_sample_code(ins
, reg
, &sample_function
, WINED3DSP_NOSWIZZLE
, NULL
, NULL
, NULL
, "tmp0.xyz");
3413 current_state
->current_row
= 0;
3416 /* Process the WINED3DSIO_TEXM3X3VSPEC instruction in GLSL
3417 * Perform the final texture lookup based on the previous 2 3x3 matrix multiplies */
3418 static void shader_glsl_texm3x3vspec(const struct wined3d_shader_instruction
*ins
)
3420 IWineD3DBaseShaderImpl
*shader
= (IWineD3DBaseShaderImpl
*)ins
->ctx
->shader
;
3421 DWORD reg
= ins
->dst
[0].reg
.idx
;
3422 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
3423 SHADER_PARSE_STATE
* current_state
= &shader
->baseShader
.parse_state
;
3424 glsl_src_param_t src0_param
;
3425 DWORD src_mask
= WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
| WINED3DSP_WRITEMASK_2
;
3426 glsl_sample_function_t sample_function
;
3428 shader_glsl_add_src_param(ins
, &ins
->src
[0], src_mask
, &src0_param
);
3430 /* Perform the last matrix multiply operation */
3431 shader_addline(buffer
, "tmp0.z = dot(vec3(T%u), vec3(%s));\n", reg
, src0_param
.param_str
);
3433 /* Construct the eye-ray vector from w coordinates */
3434 shader_addline(buffer
, "tmp1.xyz = normalize(vec3(gl_TexCoord[%u].w, gl_TexCoord[%u].w, gl_TexCoord[%u].w));\n",
3435 current_state
->texcoord_w
[0], current_state
->texcoord_w
[1], reg
);
3436 shader_addline(buffer
, "tmp0.xyz = -reflect(tmp1.xyz, normalize(tmp0.xyz));\n");
3438 /* Dependent read, not valid with conditional NP2 */
3439 shader_glsl_get_sample_function(ins
->ctx
, reg
, 0, &sample_function
);
3441 /* Sample the texture using the calculated coordinates */
3442 shader_glsl_gen_sample_code(ins
, reg
, &sample_function
, WINED3DSP_NOSWIZZLE
, NULL
, NULL
, NULL
, "tmp0.xyz");
3444 current_state
->current_row
= 0;
3447 /** Process the WINED3DSIO_TEXBEM instruction in GLSL.
3448 * Apply a fake bump map transform.
3449 * texbem is pshader <= 1.3 only, this saves a few version checks
3451 static void shader_glsl_texbem(const struct wined3d_shader_instruction
*ins
)
3453 IWineD3DBaseShaderImpl
*shader
= (IWineD3DBaseShaderImpl
*)ins
->ctx
->shader
;
3454 IWineD3DDeviceImpl
*deviceImpl
= (IWineD3DDeviceImpl
*)shader
->baseShader
.device
;
3455 glsl_sample_function_t sample_function
;
3456 glsl_src_param_t coord_param
;
3462 sampler_idx
= ins
->dst
[0].reg
.idx
;
3463 flags
= deviceImpl
->stateBlock
->textureState
[sampler_idx
][WINED3DTSS_TEXTURETRANSFORMFLAGS
];
3465 /* Dependent read, not valid with conditional NP2 */
3466 shader_glsl_get_sample_function(ins
->ctx
, sampler_idx
, 0, &sample_function
);
3467 mask
= sample_function
.coord_mask
;
3469 shader_glsl_write_mask_to_str(mask
, coord_mask
);
3471 /* with projective textures, texbem only divides the static texture coord, not the displacement,
3472 * so we can't let the GL handle this.
3474 if (flags
& WINED3DTTFF_PROJECTED
) {
3476 char coord_div_mask
[3];
3477 switch (flags
& ~WINED3DTTFF_PROJECTED
) {
3478 case WINED3DTTFF_COUNT1
: FIXME("WINED3DTTFF_PROJECTED with WINED3DTTFF_COUNT1?\n"); break;
3479 case WINED3DTTFF_COUNT2
: div_mask
= WINED3DSP_WRITEMASK_1
; break;
3480 case WINED3DTTFF_COUNT3
: div_mask
= WINED3DSP_WRITEMASK_2
; break;
3481 case WINED3DTTFF_COUNT4
:
3482 case WINED3DTTFF_DISABLE
: div_mask
= WINED3DSP_WRITEMASK_3
; break;
3484 shader_glsl_write_mask_to_str(div_mask
, coord_div_mask
);
3485 shader_addline(ins
->ctx
->buffer
, "T%u%s /= T%u%s;\n", sampler_idx
, coord_mask
, sampler_idx
, coord_div_mask
);
3488 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
, &coord_param
);
3490 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, WINED3DSP_NOSWIZZLE
, NULL
, NULL
, NULL
,
3491 "T%u%s + vec4(bumpenvmat%d * %s, 0.0, 0.0)%s", sampler_idx
, coord_mask
, sampler_idx
,
3492 coord_param
.param_str
, coord_mask
);
3494 if (ins
->handler_idx
== WINED3DSIH_TEXBEML
)
3496 glsl_src_param_t luminance_param
;
3497 glsl_dst_param_t dst_param
;
3499 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_2
, &luminance_param
);
3500 shader_glsl_add_dst_param(ins
, &ins
->dst
[0], &dst_param
);
3502 shader_addline(ins
->ctx
->buffer
, "%s%s *= (%s * luminancescale%d + luminanceoffset%d);\n",
3503 dst_param
.reg_name
, dst_param
.mask_str
,
3504 luminance_param
.param_str
, sampler_idx
, sampler_idx
);
3508 static void shader_glsl_bem(const struct wined3d_shader_instruction
*ins
)
3510 glsl_src_param_t src0_param
, src1_param
;
3511 DWORD sampler_idx
= ins
->dst
[0].reg
.idx
;
3513 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
, &src0_param
);
3514 shader_glsl_add_src_param(ins
, &ins
->src
[1], WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
, &src1_param
);
3516 shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
3517 shader_addline(ins
->ctx
->buffer
, "%s + bumpenvmat%d * %s);\n",
3518 src0_param
.param_str
, sampler_idx
, src1_param
.param_str
);
3521 /** Process the WINED3DSIO_TEXREG2AR instruction in GLSL
3522 * Sample 2D texture at dst using the alpha & red (wx) components of src as texture coordinates */
3523 static void shader_glsl_texreg2ar(const struct wined3d_shader_instruction
*ins
)
3525 glsl_src_param_t src0_param
;
3526 DWORD sampler_idx
= ins
->dst
[0].reg
.idx
;
3527 glsl_sample_function_t sample_function
;
3529 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_ALL
, &src0_param
);
3531 shader_glsl_get_sample_function(ins
->ctx
, sampler_idx
, 0, &sample_function
);
3532 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, WINED3DSP_NOSWIZZLE
, NULL
, NULL
, NULL
,
3533 "%s.wx", src0_param
.reg_name
);
3536 /** Process the WINED3DSIO_TEXREG2GB instruction in GLSL
3537 * Sample 2D texture at dst using the green & blue (yz) components of src as texture coordinates */
3538 static void shader_glsl_texreg2gb(const struct wined3d_shader_instruction
*ins
)
3540 glsl_src_param_t src0_param
;
3541 DWORD sampler_idx
= ins
->dst
[0].reg
.idx
;
3542 glsl_sample_function_t sample_function
;
3544 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_ALL
, &src0_param
);
3546 shader_glsl_get_sample_function(ins
->ctx
, sampler_idx
, 0, &sample_function
);
3547 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, WINED3DSP_NOSWIZZLE
, NULL
, NULL
, NULL
,
3548 "%s.yz", src0_param
.reg_name
);
3551 /** Process the WINED3DSIO_TEXREG2RGB instruction in GLSL
3552 * Sample texture at dst using the rgb (xyz) components of src as texture coordinates */
3553 static void shader_glsl_texreg2rgb(const struct wined3d_shader_instruction
*ins
)
3555 glsl_src_param_t src0_param
;
3556 DWORD sampler_idx
= ins
->dst
[0].reg
.idx
;
3557 glsl_sample_function_t sample_function
;
3559 /* Dependent read, not valid with conditional NP2 */
3560 shader_glsl_get_sample_function(ins
->ctx
, sampler_idx
, 0, &sample_function
);
3561 shader_glsl_add_src_param(ins
, &ins
->src
[0], sample_function
.coord_mask
, &src0_param
);
3563 shader_glsl_gen_sample_code(ins
, sampler_idx
, &sample_function
, WINED3DSP_NOSWIZZLE
, NULL
, NULL
, NULL
,
3564 "%s", src0_param
.param_str
);
3567 /** Process the WINED3DSIO_TEXKILL instruction in GLSL.
3568 * If any of the first 3 components are < 0, discard this pixel */
3569 static void shader_glsl_texkill(const struct wined3d_shader_instruction
*ins
)
3571 glsl_dst_param_t dst_param
;
3573 /* The argument is a destination parameter, and no writemasks are allowed */
3574 shader_glsl_add_dst_param(ins
, &ins
->dst
[0], &dst_param
);
3575 if (ins
->ctx
->reg_maps
->shader_version
.major
>= 2)
3577 /* 2.0 shaders compare all 4 components in texkill */
3578 shader_addline(ins
->ctx
->buffer
, "if (any(lessThan(%s.xyzw, vec4(0.0)))) discard;\n", dst_param
.reg_name
);
3580 /* 1.X shaders only compare the first 3 components, probably due to the nature of the texkill
3581 * instruction as a tex* instruction, and phase, which kills all a / w components. Even if all
3582 * 4 components are defined, only the first 3 are used
3584 shader_addline(ins
->ctx
->buffer
, "if (any(lessThan(%s.xyz, vec3(0.0)))) discard;\n", dst_param
.reg_name
);
3588 /** Process the WINED3DSIO_DP2ADD instruction in GLSL.
3589 * dst = dot2(src0, src1) + src2 */
3590 static void shader_glsl_dp2add(const struct wined3d_shader_instruction
*ins
)
3592 glsl_src_param_t src0_param
;
3593 glsl_src_param_t src1_param
;
3594 glsl_src_param_t src2_param
;
3596 unsigned int mask_size
;
3598 write_mask
= shader_glsl_append_dst(ins
->ctx
->buffer
, ins
);
3599 mask_size
= shader_glsl_get_write_mask_size(write_mask
);
3601 shader_glsl_add_src_param(ins
, &ins
->src
[0], WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
, &src0_param
);
3602 shader_glsl_add_src_param(ins
, &ins
->src
[1], WINED3DSP_WRITEMASK_0
| WINED3DSP_WRITEMASK_1
, &src1_param
);
3603 shader_glsl_add_src_param(ins
, &ins
->src
[2], WINED3DSP_WRITEMASK_0
, &src2_param
);
3605 if (mask_size
> 1) {
3606 shader_addline(ins
->ctx
->buffer
, "vec%d(dot(%s, %s) + %s));\n",
3607 mask_size
, src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
3609 shader_addline(ins
->ctx
->buffer
, "dot(%s, %s) + %s);\n",
3610 src0_param
.param_str
, src1_param
.param_str
, src2_param
.param_str
);
3614 static void shader_glsl_input_pack(IWineD3DPixelShader
*iface
, struct wined3d_shader_buffer
*buffer
,
3615 const struct wined3d_shader_signature_element
*input_signature
, const struct shader_reg_maps
*reg_maps
,
3616 enum vertexprocessing_mode vertexprocessing
)
3619 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*)iface
;
3620 WORD map
= reg_maps
->input_registers
;
3622 for (i
= 0; map
; map
>>= 1, ++i
)
3624 const char *semantic_name
;
3629 if (!(map
& 1)) continue;
3631 semantic_name
= input_signature
[i
].semantic_name
;
3632 semantic_idx
= input_signature
[i
].semantic_idx
;
3633 shader_glsl_write_mask_to_str(input_signature
[i
].mask
, reg_mask
);
3635 if (shader_match_semantic(semantic_name
, WINED3DDECLUSAGE_TEXCOORD
))
3637 if (semantic_idx
< 8 && vertexprocessing
== pretransformed
)
3638 shader_addline(buffer
, "IN[%u]%s = gl_TexCoord[%u]%s;\n",
3639 This
->input_reg_map
[i
], reg_mask
, semantic_idx
, reg_mask
);
3641 shader_addline(buffer
, "IN[%u]%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
3642 This
->input_reg_map
[i
], reg_mask
, reg_mask
);
3644 else if (shader_match_semantic(semantic_name
, WINED3DDECLUSAGE_COLOR
))
3646 if (semantic_idx
== 0)
3647 shader_addline(buffer
, "IN[%u]%s = vec4(gl_Color)%s;\n",
3648 This
->input_reg_map
[i
], reg_mask
, reg_mask
);
3649 else if (semantic_idx
== 1)
3650 shader_addline(buffer
, "IN[%u]%s = vec4(gl_SecondaryColor)%s;\n",
3651 This
->input_reg_map
[i
], reg_mask
, reg_mask
);
3653 shader_addline(buffer
, "IN[%u]%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
3654 This
->input_reg_map
[i
], reg_mask
, 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
);
3664 /*********************************************
3665 * Vertex Shader Specific Code begins here
3666 ********************************************/
3668 static void add_glsl_program_entry(struct shader_glsl_priv
*priv
, struct glsl_shader_prog_link
*entry
) {
3669 glsl_program_key_t key
;
3671 key
.vshader
= entry
->vshader
;
3672 key
.pshader
= entry
->pshader
;
3673 key
.vs_args
= entry
->vs_args
;
3674 key
.ps_args
= entry
->ps_args
;
3676 if (wine_rb_put(&priv
->program_lookup
, &key
, &entry
->program_lookup_entry
) == -1)
3678 ERR("Failed to insert program entry.\n");
3682 static struct glsl_shader_prog_link
*get_glsl_program_entry(struct shader_glsl_priv
*priv
,
3683 IWineD3DVertexShader
*vshader
, IWineD3DPixelShader
*pshader
, struct vs_compile_args
*vs_args
,
3684 struct ps_compile_args
*ps_args
) {
3685 struct wine_rb_entry
*entry
;
3686 glsl_program_key_t key
;
3688 key
.vshader
= vshader
;
3689 key
.pshader
= pshader
;
3690 key
.vs_args
= *vs_args
;
3691 key
.ps_args
= *ps_args
;
3693 entry
= wine_rb_get(&priv
->program_lookup
, &key
);
3694 return entry
? WINE_RB_ENTRY_VALUE(entry
, struct glsl_shader_prog_link
, program_lookup_entry
) : NULL
;
3697 /* GL locking is done by the caller */
3698 static void delete_glsl_program_entry(struct shader_glsl_priv
*priv
, const struct wined3d_gl_info
*gl_info
,
3699 struct glsl_shader_prog_link
*entry
)
3701 glsl_program_key_t key
;
3703 key
.vshader
= entry
->vshader
;
3704 key
.pshader
= entry
->pshader
;
3705 key
.vs_args
= entry
->vs_args
;
3706 key
.ps_args
= entry
->ps_args
;
3707 wine_rb_remove(&priv
->program_lookup
, &key
);
3709 GL_EXTCALL(glDeleteObjectARB(entry
->programId
));
3710 if (entry
->vshader
) list_remove(&entry
->vshader_entry
);
3711 if (entry
->pshader
) list_remove(&entry
->pshader_entry
);
3712 HeapFree(GetProcessHeap(), 0, entry
->vuniformF_locations
);
3713 HeapFree(GetProcessHeap(), 0, entry
->puniformF_locations
);
3714 HeapFree(GetProcessHeap(), 0, entry
);
3717 static void handle_ps3_input(struct wined3d_shader_buffer
*buffer
, const struct wined3d_gl_info
*gl_info
, const DWORD
*map
,
3718 const struct wined3d_shader_signature_element
*input_signature
, const struct shader_reg_maps
*reg_maps_in
,
3719 const struct wined3d_shader_signature_element
*output_signature
, const struct shader_reg_maps
*reg_maps_out
)
3722 const char *semantic_name_in
;
3723 UINT semantic_idx_in
;
3726 unsigned int in_count
= vec4_varyings(3, gl_info
);
3728 char destination
[50];
3729 WORD input_map
, output_map
;
3731 set
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*set
) * (in_count
+ 2));
3733 input_map
= reg_maps_in
->input_registers
;
3734 for (i
= 0; input_map
; input_map
>>= 1, ++i
)
3736 if (!(input_map
& 1)) continue;
3739 /* Declared, but not read register */
3740 if (in_idx
== ~0U) continue;
3741 if (in_idx
>= (in_count
+ 2))
3743 FIXME("More input varyings declared than supported, expect issues.\n");
3747 if (in_idx
== in_count
) {
3748 sprintf(destination
, "gl_FrontColor");
3749 } else if (in_idx
== in_count
+ 1) {
3750 sprintf(destination
, "gl_FrontSecondaryColor");
3752 sprintf(destination
, "IN[%u]", in_idx
);
3755 semantic_name_in
= input_signature
[i
].semantic_name
;
3756 semantic_idx_in
= input_signature
[i
].semantic_idx
;
3759 output_map
= reg_maps_out
->output_registers
;
3760 for (j
= 0; output_map
; output_map
>>= 1, ++j
)
3764 if (!(output_map
& 1)
3765 || semantic_idx_in
!= output_signature
[j
].semantic_idx
3766 || strcmp(semantic_name_in
, output_signature
[j
].semantic_name
)
3767 || !(mask
= input_signature
[i
].mask
& output_signature
[j
].mask
))
3771 shader_glsl_write_mask_to_str(mask
, reg_mask
);
3773 shader_addline(buffer
, "%s%s = clamp(OUT[%u]%s, -FLT_MAX, FLT_MAX);\n",
3774 destination
, reg_mask
, j
, reg_mask
);
3778 for (i
= 0; i
< in_count
+ 2; ++i
)
3782 if (!set
[i
] || set
[i
] == WINED3DSP_WRITEMASK_ALL
)
3785 if (set
[i
] == ~0U) set
[i
] = 0;
3788 if (!(set
[i
] & WINED3DSP_WRITEMASK_0
)) reg_mask
[size
++] = 'x';
3789 if (!(set
[i
] & WINED3DSP_WRITEMASK_1
)) reg_mask
[size
++] = 'y';
3790 if (!(set
[i
] & WINED3DSP_WRITEMASK_2
)) reg_mask
[size
++] = 'z';
3791 if (!(set
[i
] & WINED3DSP_WRITEMASK_3
)) reg_mask
[size
++] = 'w';
3792 reg_mask
[size
] = '\0';
3794 if (i
== in_count
) sprintf(destination
, "gl_FrontColor");
3795 else if (i
== in_count
+ 1) sprintf(destination
, "gl_FrontSecondaryColor");
3796 else sprintf(destination
, "IN[%u]", i
);
3798 if (size
== 1) shader_addline(buffer
, "%s.%s = 0.0;\n", destination
, reg_mask
);
3799 else shader_addline(buffer
, "%s.%s = vec%u(0.0);\n", destination
, reg_mask
, size
);
3802 HeapFree(GetProcessHeap(), 0, set
);
3805 /* GL locking is done by the caller */
3806 static GLhandleARB
generate_param_reorder_function(struct wined3d_shader_buffer
*buffer
,
3807 IWineD3DVertexShader
*vertexshader
, IWineD3DPixelShader
*pixelshader
, const struct wined3d_gl_info
*gl_info
)
3809 GLhandleARB ret
= 0;
3810 IWineD3DVertexShaderImpl
*vs
= (IWineD3DVertexShaderImpl
*) vertexshader
;
3811 IWineD3DPixelShaderImpl
*ps
= (IWineD3DPixelShaderImpl
*) pixelshader
;
3812 DWORD ps_major
= ps
? ps
->baseShader
.reg_maps
.shader_version
.major
: 0;
3814 const char *semantic_name
;
3817 const struct wined3d_shader_signature_element
*output_signature
= vs
->baseShader
.output_signature
;
3818 WORD map
= vs
->baseShader
.reg_maps
.output_registers
;
3820 shader_buffer_clear(buffer
);
3822 shader_addline(buffer
, "#version 120\n");
3823 shader_addline(buffer
, "const float FLT_MAX = 1e38;\n");
3827 shader_addline(buffer
, "void order_ps_input(in vec4 OUT[%u]) {\n", MAX_REG_OUTPUT
);
3829 for (i
= 0; map
; map
>>= 1, ++i
)
3833 if (!(map
& 1)) continue;
3835 semantic_name
= output_signature
[i
].semantic_name
;
3836 semantic_idx
= output_signature
[i
].semantic_idx
;
3837 write_mask
= output_signature
[i
].mask
;
3838 shader_glsl_write_mask_to_str(write_mask
, reg_mask
);
3840 if (shader_match_semantic(semantic_name
, WINED3DDECLUSAGE_COLOR
))
3842 if (semantic_idx
== 0)
3843 shader_addline(buffer
, "gl_FrontColor%s = clamp(OUT[%u]%s, -FLT_MAX, FLT_MAX);\n",
3844 reg_mask
, i
, reg_mask
);
3845 else if (semantic_idx
== 1)
3846 shader_addline(buffer
, "gl_FrontSecondaryColor%s = clamp(OUT[%u]%s, -FLT_MAX, FLT_MAX);\n",
3847 reg_mask
, i
, reg_mask
);
3849 else if (shader_match_semantic(semantic_name
, WINED3DDECLUSAGE_POSITION
))
3851 shader_addline(buffer
, "gl_Position%s = clamp(OUT[%u]%s, -FLT_MAX, FLT_MAX);\n",
3852 reg_mask
, i
, reg_mask
);
3854 else if (shader_match_semantic(semantic_name
, WINED3DDECLUSAGE_TEXCOORD
))
3856 if (semantic_idx
< 8)
3858 if (!(gl_info
->quirks
& WINED3D_QUIRK_SET_TEXCOORD_W
) || ps_major
> 0)
3859 write_mask
|= WINED3DSP_WRITEMASK_3
;
3861 shader_addline(buffer
, "gl_TexCoord[%u]%s = clamp(OUT[%u]%s, -FLT_MAX, FLT_MAX);\n",
3862 semantic_idx
, reg_mask
, i
, reg_mask
);
3863 if (!(write_mask
& WINED3DSP_WRITEMASK_3
))
3864 shader_addline(buffer
, "gl_TexCoord[%u].w = 1.0;\n", semantic_idx
);
3867 else if (shader_match_semantic(semantic_name
, WINED3DDECLUSAGE_PSIZE
))
3869 shader_addline(buffer
, "gl_PointSize = clamp(OUT[%u].%c, -FLT_MAX, FLT_MAX);\n", i
, reg_mask
[1]);
3871 else if (shader_match_semantic(semantic_name
, WINED3DDECLUSAGE_FOG
))
3873 shader_addline(buffer
, "gl_FogFragCoord = clamp(OUT[%u].%c, -FLT_MAX, FLT_MAX);\n", i
, reg_mask
[1]);
3876 shader_addline(buffer
, "}\n");
3881 /* This one is tricky: a 3.0 pixel shader reads from a 3.0 vertex shader */
3882 shader_addline(buffer
, "varying vec4 IN[%u];\n", vec4_varyings(3, gl_info
));
3883 shader_addline(buffer
, "void order_ps_input(in vec4 OUT[%u]) {\n", MAX_REG_OUTPUT
);
3885 /* First, sort out position and point size. Those are not passed to the pixel shader */
3886 for (i
= 0; map
; map
>>= 1, ++i
)
3888 if (!(map
& 1)) continue;
3890 semantic_name
= output_signature
[i
].semantic_name
;
3891 shader_glsl_write_mask_to_str(output_signature
[i
].mask
, reg_mask
);
3893 if (shader_match_semantic(semantic_name
, WINED3DDECLUSAGE_POSITION
))
3895 shader_addline(buffer
, "gl_Position%s = clamp(OUT[%u]%s, -FLT_MAX, FLT_MAX);\n",
3896 reg_mask
, i
, reg_mask
);
3898 else if (shader_match_semantic(semantic_name
, WINED3DDECLUSAGE_PSIZE
))
3900 shader_addline(buffer
, "gl_PointSize = clamp(OUT[%u].%c, -FLT_MAX, FLT_MAX);\n", i
, reg_mask
[1]);
3904 /* Then, fix the pixel shader input */
3905 handle_ps3_input(buffer
, gl_info
, ps
->input_reg_map
, ps
->baseShader
.input_signature
,
3906 &ps
->baseShader
.reg_maps
, output_signature
, &vs
->baseShader
.reg_maps
);
3908 shader_addline(buffer
, "}\n");
3911 ret
= GL_EXTCALL(glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB
));
3912 checkGLcall("glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB)");
3913 GL_EXTCALL(glShaderSourceARB(ret
, 1, (const char**)&buffer
->buffer
, NULL
));
3914 checkGLcall("glShaderSourceARB(ret, 1, &buffer->buffer, NULL)");
3915 GL_EXTCALL(glCompileShaderARB(ret
));
3916 checkGLcall("glCompileShaderARB(ret)");
3921 /* GL locking is done by the caller */
3922 static void hardcode_local_constants(IWineD3DBaseShaderImpl
*shader
, const struct wined3d_gl_info
*gl_info
,
3923 GLhandleARB programId
, char prefix
)
3925 const local_constant
*lconst
;
3930 LIST_FOR_EACH_ENTRY(lconst
, &shader
->baseShader
.constantsF
, local_constant
, entry
) {
3931 value
= (const float *)lconst
->value
;
3932 snprintf(glsl_name
, sizeof(glsl_name
), "%cLC%u", prefix
, lconst
->idx
);
3933 tmp_loc
= GL_EXTCALL(glGetUniformLocationARB(programId
, glsl_name
));
3934 GL_EXTCALL(glUniform4fvARB(tmp_loc
, 1, value
));
3936 checkGLcall("Hardcoding local constants");
3939 /* GL locking is done by the caller */
3940 static GLuint
shader_glsl_generate_pshader(const struct wined3d_context
*context
,
3941 struct wined3d_shader_buffer
*buffer
, IWineD3DPixelShaderImpl
*This
,
3942 const struct ps_compile_args
*args
, struct ps_np2fixup_info
*np2fixup_info
)
3944 const struct shader_reg_maps
*reg_maps
= &This
->baseShader
.reg_maps
;
3945 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
3946 CONST DWORD
*function
= This
->baseShader
.function
;
3947 struct shader_glsl_ctx_priv priv_ctx
;
3949 /* Create the hw GLSL shader object and assign it as the shader->prgId */
3950 GLhandleARB shader_obj
= GL_EXTCALL(glCreateShaderObjectARB(GL_FRAGMENT_SHADER_ARB
));
3952 memset(&priv_ctx
, 0, sizeof(priv_ctx
));
3953 priv_ctx
.cur_ps_args
= args
;
3954 priv_ctx
.cur_np2fixup_info
= np2fixup_info
;
3956 shader_addline(buffer
, "#version 120\n");
3958 if (gl_info
->supported
[ARB_SHADER_TEXTURE_LOD
] && reg_maps
->usestexldd
)
3960 shader_addline(buffer
, "#extension GL_ARB_shader_texture_lod : enable\n");
3962 if (gl_info
->supported
[ARB_TEXTURE_RECTANGLE
])
3964 /* The spec says that it doesn't have to be explicitly enabled, but the nvidia
3965 * drivers write a warning if we don't do so
3967 shader_addline(buffer
, "#extension GL_ARB_texture_rectangle : enable\n");
3969 if (gl_info
->supported
[EXT_GPU_SHADER4
])
3971 shader_addline(buffer
, "#extension GL_EXT_gpu_shader4 : enable\n");
3974 /* Base Declarations */
3975 shader_generate_glsl_declarations(context
, buffer
, (IWineD3DBaseShader
*)This
, reg_maps
, &priv_ctx
);
3977 /* Pack 3.0 inputs */
3978 if (reg_maps
->shader_version
.major
>= 3 && args
->vp_mode
!= vertexshader
)
3980 shader_glsl_input_pack((IWineD3DPixelShader
*) This
, buffer
,
3981 This
->baseShader
.input_signature
, reg_maps
, args
->vp_mode
);
3984 /* Base Shader Body */
3985 shader_generate_main((IWineD3DBaseShader
*)This
, buffer
, reg_maps
, function
, &priv_ctx
);
3987 /* Pixel shaders < 2.0 place the resulting color in R0 implicitly */
3988 if (reg_maps
->shader_version
.major
< 2)
3990 /* Some older cards like GeforceFX ones don't support multiple buffers, so also not gl_FragData */
3991 shader_addline(buffer
, "gl_FragData[0] = R0;\n");
3994 if (args
->srgb_correction
)
3996 shader_addline(buffer
, "tmp0.xyz = pow(gl_FragData[0].xyz, vec3(srgb_const0.x));\n");
3997 shader_addline(buffer
, "tmp0.xyz = tmp0.xyz * vec3(srgb_const0.y) - vec3(srgb_const0.z);\n");
3998 shader_addline(buffer
, "tmp1.xyz = gl_FragData[0].xyz * vec3(srgb_const0.w);\n");
3999 shader_addline(buffer
, "bvec3 srgb_compare = lessThan(gl_FragData[0].xyz, vec3(srgb_const1.x));\n");
4000 shader_addline(buffer
, "gl_FragData[0].xyz = mix(tmp0.xyz, tmp1.xyz, vec3(srgb_compare));\n");
4001 shader_addline(buffer
, "gl_FragData[0] = clamp(gl_FragData[0], 0.0, 1.0);\n");
4003 /* Pixel shader < 3.0 do not replace the fog stage.
4004 * This implements linear fog computation and blending.
4005 * TODO: non linear fog
4006 * NOTE: gl_Fog.start and gl_Fog.end don't hold fog start s and end e but
4007 * -1/(e-s) and e/(e-s) respectively.
4009 if (reg_maps
->shader_version
.major
< 3)
4012 case FOG_OFF
: break;
4014 shader_addline(buffer
, "float fogstart = -1.0 / (gl_Fog.end - gl_Fog.start);\n");
4015 shader_addline(buffer
, "float fogend = gl_Fog.end * -fogstart;\n");
4016 shader_addline(buffer
, "float Fog = clamp(gl_FogFragCoord * fogstart + fogend, 0.0, 1.0);\n");
4017 shader_addline(buffer
, "gl_FragData[0].xyz = mix(gl_Fog.color.xyz, gl_FragData[0].xyz, Fog);\n");
4020 /* Fog = e^(-gl_Fog.density * gl_FogFragCoord) */
4021 shader_addline(buffer
, "float Fog = exp(-gl_Fog.density * gl_FogFragCoord);\n");
4022 shader_addline(buffer
, "Fog = clamp(Fog, 0.0, 1.0);\n");
4023 shader_addline(buffer
, "gl_FragData[0].xyz = mix(gl_Fog.color.xyz, gl_FragData[0].xyz, Fog);\n");
4026 /* Fog = e^(-(gl_Fog.density * gl_FogFragCoord)^2) */
4027 shader_addline(buffer
, "float Fog = exp(-gl_Fog.density * gl_Fog.density * gl_FogFragCoord * gl_FogFragCoord);\n");
4028 shader_addline(buffer
, "Fog = clamp(Fog, 0.0, 1.0);\n");
4029 shader_addline(buffer
, "gl_FragData[0].xyz = mix(gl_Fog.color.xyz, gl_FragData[0].xyz, Fog);\n");
4034 shader_addline(buffer
, "}\n");
4036 TRACE("Compiling shader object %u\n", shader_obj
);
4037 GL_EXTCALL(glShaderSourceARB(shader_obj
, 1, (const char**)&buffer
->buffer
, NULL
));
4038 GL_EXTCALL(glCompileShaderARB(shader_obj
));
4039 print_glsl_info_log(gl_info
, shader_obj
);
4041 /* Store the shader object */
4045 /* GL locking is done by the caller */
4046 static GLuint
shader_glsl_generate_vshader(const struct wined3d_context
*context
,
4047 struct wined3d_shader_buffer
*buffer
, IWineD3DVertexShaderImpl
*This
,
4048 const struct vs_compile_args
*args
)
4050 const struct shader_reg_maps
*reg_maps
= &This
->baseShader
.reg_maps
;
4051 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
4052 CONST DWORD
*function
= This
->baseShader
.function
;
4053 struct shader_glsl_ctx_priv priv_ctx
;
4055 /* Create the hw GLSL shader program and assign it as the shader->prgId */
4056 GLhandleARB shader_obj
= GL_EXTCALL(glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB
));
4058 shader_addline(buffer
, "#version 120\n");
4060 if (gl_info
->supported
[EXT_GPU_SHADER4
])
4062 shader_addline(buffer
, "#extension GL_EXT_gpu_shader4 : enable\n");
4065 memset(&priv_ctx
, 0, sizeof(priv_ctx
));
4066 priv_ctx
.cur_vs_args
= args
;
4068 /* Base Declarations */
4069 shader_generate_glsl_declarations(context
, buffer
, (IWineD3DBaseShader
*)This
, reg_maps
, &priv_ctx
);
4071 /* Base Shader Body */
4072 shader_generate_main((IWineD3DBaseShader
*)This
, buffer
, reg_maps
, function
, &priv_ctx
);
4074 /* Unpack outputs */
4075 shader_addline(buffer
, "order_ps_input(OUT);\n");
4077 /* The D3DRS_FOGTABLEMODE render state defines if the shader-generated fog coord is used
4078 * or if the fragment depth is used. If the fragment depth is used(FOGTABLEMODE != NONE),
4079 * the fog frag coord is thrown away. If the fog frag coord is used, but not written by
4080 * the shader, it is set to 0.0(fully fogged, since start = 1.0, end = 0.0)
4082 if(args
->fog_src
== VS_FOG_Z
) {
4083 shader_addline(buffer
, "gl_FogFragCoord = gl_Position.z;\n");
4084 } else if (!reg_maps
->fog
) {
4085 shader_addline(buffer
, "gl_FogFragCoord = 0.0;\n");
4088 /* Write the final position.
4090 * OpenGL coordinates specify the center of the pixel while d3d coords specify
4091 * the corner. The offsets are stored in z and w in posFixup. posFixup.y contains
4092 * 1.0 or -1.0 to turn the rendering upside down for offscreen rendering. PosFixup.x
4093 * contains 1.0 to allow a mad.
4095 shader_addline(buffer
, "gl_Position.y = gl_Position.y * posFixup.y;\n");
4096 shader_addline(buffer
, "gl_Position.xy += posFixup.zw * gl_Position.ww;\n");
4097 if(args
->clip_enabled
) {
4098 shader_addline(buffer
, "gl_ClipVertex = gl_Position;\n");
4101 /* Z coord [0;1]->[-1;1] mapping, see comment in transform_projection in state.c
4103 * Basically we want (in homogeneous coordinates) z = z * 2 - 1. However, shaders are run
4104 * before the homogeneous divide, so we have to take the w into account: z = ((z / w) * 2 - 1) * w,
4105 * which is the same as z = z * 2 - w.
4107 shader_addline(buffer
, "gl_Position.z = gl_Position.z * 2.0 - gl_Position.w;\n");
4109 shader_addline(buffer
, "}\n");
4111 TRACE("Compiling shader object %u\n", shader_obj
);
4112 GL_EXTCALL(glShaderSourceARB(shader_obj
, 1, (const char**)&buffer
->buffer
, NULL
));
4113 GL_EXTCALL(glCompileShaderARB(shader_obj
));
4114 print_glsl_info_log(gl_info
, shader_obj
);
4119 static GLhandleARB
find_glsl_pshader(const struct wined3d_context
*context
,
4120 struct wined3d_shader_buffer
*buffer
, IWineD3DPixelShaderImpl
*shader
,
4121 const struct ps_compile_args
*args
, const struct ps_np2fixup_info
**np2fixup_info
)
4125 struct glsl_ps_compiled_shader
*new_array
;
4126 struct glsl_pshader_private
*shader_data
;
4127 struct ps_np2fixup_info
*np2fixup
= NULL
;
4130 if (!shader
->baseShader
.backend_data
)
4132 shader
->baseShader
.backend_data
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*shader_data
));
4133 if (!shader
->baseShader
.backend_data
)
4135 ERR("Failed to allocate backend data.\n");
4139 shader_data
= shader
->baseShader
.backend_data
;
4141 /* Usually we have very few GL shaders for each d3d shader(just 1 or maybe 2),
4142 * so a linear search is more performant than a hashmap or a binary search
4143 * (cache coherency etc)
4145 for(i
= 0; i
< shader_data
->num_gl_shaders
; i
++) {
4146 if(memcmp(&shader_data
->gl_shaders
[i
].args
, args
, sizeof(*args
)) == 0) {
4147 if(args
->np2_fixup
) *np2fixup_info
= &shader_data
->gl_shaders
[i
].np2fixup
;
4148 return shader_data
->gl_shaders
[i
].prgId
;
4152 TRACE("No matching GL shader found for shader %p, compiling a new shader.\n", shader
);
4153 if(shader_data
->shader_array_size
== shader_data
->num_gl_shaders
) {
4154 if (shader_data
->num_gl_shaders
)
4156 new_size
= shader_data
->shader_array_size
+ max(1, shader_data
->shader_array_size
/ 2);
4157 new_array
= HeapReAlloc(GetProcessHeap(), 0, shader_data
->gl_shaders
,
4158 new_size
* sizeof(*shader_data
->gl_shaders
));
4160 new_array
= HeapAlloc(GetProcessHeap(), 0, sizeof(*shader_data
->gl_shaders
));
4165 ERR("Out of memory\n");
4168 shader_data
->gl_shaders
= new_array
;
4169 shader_data
->shader_array_size
= new_size
;
4172 shader_data
->gl_shaders
[shader_data
->num_gl_shaders
].args
= *args
;
4174 memset(&shader_data
->gl_shaders
[shader_data
->num_gl_shaders
].np2fixup
, 0, sizeof(struct ps_np2fixup_info
));
4175 if (args
->np2_fixup
) np2fixup
= &shader_data
->gl_shaders
[shader_data
->num_gl_shaders
].np2fixup
;
4177 pixelshader_update_samplers(&shader
->baseShader
.reg_maps
,
4178 ((IWineD3DDeviceImpl
*)shader
->baseShader
.device
)->stateBlock
->textures
);
4180 shader_buffer_clear(buffer
);
4181 ret
= shader_glsl_generate_pshader(context
, buffer
, shader
, args
, np2fixup
);
4182 shader_data
->gl_shaders
[shader_data
->num_gl_shaders
++].prgId
= ret
;
4183 *np2fixup_info
= np2fixup
;
4188 static inline BOOL
vs_args_equal(const struct vs_compile_args
*stored
, const struct vs_compile_args
*new,
4189 const DWORD use_map
) {
4190 if((stored
->swizzle_map
& use_map
) != new->swizzle_map
) return FALSE
;
4191 if((stored
->clip_enabled
) != new->clip_enabled
) return FALSE
;
4192 return stored
->fog_src
== new->fog_src
;
4195 static GLhandleARB
find_glsl_vshader(const struct wined3d_context
*context
,
4196 struct wined3d_shader_buffer
*buffer
, IWineD3DVertexShaderImpl
*shader
,
4197 const struct vs_compile_args
*args
)
4201 struct glsl_vs_compiled_shader
*new_array
;
4202 DWORD use_map
= ((IWineD3DDeviceImpl
*)shader
->baseShader
.device
)->strided_streams
.use_map
;
4203 struct glsl_vshader_private
*shader_data
;
4206 if (!shader
->baseShader
.backend_data
)
4208 shader
->baseShader
.backend_data
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*shader_data
));
4209 if (!shader
->baseShader
.backend_data
)
4211 ERR("Failed to allocate backend data.\n");
4215 shader_data
= shader
->baseShader
.backend_data
;
4217 /* Usually we have very few GL shaders for each d3d shader(just 1 or maybe 2),
4218 * so a linear search is more performant than a hashmap or a binary search
4219 * (cache coherency etc)
4221 for(i
= 0; i
< shader_data
->num_gl_shaders
; i
++) {
4222 if(vs_args_equal(&shader_data
->gl_shaders
[i
].args
, args
, use_map
)) {
4223 return shader_data
->gl_shaders
[i
].prgId
;
4227 TRACE("No matching GL shader found for shader %p, compiling a new shader.\n", shader
);
4229 if(shader_data
->shader_array_size
== shader_data
->num_gl_shaders
) {
4230 if (shader_data
->num_gl_shaders
)
4232 new_size
= shader_data
->shader_array_size
+ max(1, shader_data
->shader_array_size
/ 2);
4233 new_array
= HeapReAlloc(GetProcessHeap(), 0, shader_data
->gl_shaders
,
4234 new_size
* sizeof(*shader_data
->gl_shaders
));
4236 new_array
= HeapAlloc(GetProcessHeap(), 0, sizeof(*shader_data
->gl_shaders
));
4241 ERR("Out of memory\n");
4244 shader_data
->gl_shaders
= new_array
;
4245 shader_data
->shader_array_size
= new_size
;
4248 shader_data
->gl_shaders
[shader_data
->num_gl_shaders
].args
= *args
;
4250 shader_buffer_clear(buffer
);
4251 ret
= shader_glsl_generate_vshader(context
, buffer
, shader
, args
);
4252 shader_data
->gl_shaders
[shader_data
->num_gl_shaders
++].prgId
= ret
;
4257 /** Sets the GLSL program ID for the given pixel and vertex shader combination.
4258 * It sets the programId on the current StateBlock (because it should be called
4259 * inside of the DrawPrimitive() part of the render loop).
4261 * If a program for the given combination does not exist, create one, and store
4262 * the program in the hash table. If it creates a program, it will link the
4263 * given objects, too.
4266 /* GL locking is done by the caller */
4267 static void set_glsl_shader_program(const struct wined3d_context
*context
,
4268 IWineD3DDeviceImpl
*device
, BOOL use_ps
, BOOL use_vs
)
4270 IWineD3DVertexShader
*vshader
= use_vs
? device
->stateBlock
->vertexShader
: NULL
;
4271 IWineD3DPixelShader
*pshader
= use_ps
? device
->stateBlock
->pixelShader
: NULL
;
4272 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
4273 struct shader_glsl_priv
*priv
= device
->shader_priv
;
4274 struct glsl_shader_prog_link
*entry
= NULL
;
4275 GLhandleARB programId
= 0;
4276 GLhandleARB reorder_shader_id
= 0;
4279 struct ps_compile_args ps_compile_args
;
4280 struct vs_compile_args vs_compile_args
;
4282 if (vshader
) find_vs_compile_args((IWineD3DVertexShaderImpl
*)vshader
, device
->stateBlock
, &vs_compile_args
);
4283 if (pshader
) find_ps_compile_args((IWineD3DPixelShaderImpl
*)pshader
, device
->stateBlock
, &ps_compile_args
);
4285 entry
= get_glsl_program_entry(priv
, vshader
, pshader
, &vs_compile_args
, &ps_compile_args
);
4287 priv
->glsl_program
= entry
;
4291 /* If we get to this point, then no matching program exists, so we create one */
4292 programId
= GL_EXTCALL(glCreateProgramObjectARB());
4293 TRACE("Created new GLSL shader program %u\n", programId
);
4295 /* Create the entry */
4296 entry
= HeapAlloc(GetProcessHeap(), 0, sizeof(struct glsl_shader_prog_link
));
4297 entry
->programId
= programId
;
4298 entry
->vshader
= vshader
;
4299 entry
->pshader
= pshader
;
4300 entry
->vs_args
= vs_compile_args
;
4301 entry
->ps_args
= ps_compile_args
;
4302 entry
->constant_version
= 0;
4303 entry
->np2Fixup_info
= NULL
;
4304 /* Add the hash table entry */
4305 add_glsl_program_entry(priv
, entry
);
4307 /* Set the current program */
4308 priv
->glsl_program
= entry
;
4310 /* Attach GLSL vshader */
4313 GLhandleARB vshader_id
= find_glsl_vshader(context
, &priv
->shader_buffer
,
4314 (IWineD3DVertexShaderImpl
*)vshader
, &vs_compile_args
);
4315 WORD map
= ((IWineD3DBaseShaderImpl
*)vshader
)->baseShader
.reg_maps
.input_registers
;
4318 reorder_shader_id
= generate_param_reorder_function(&priv
->shader_buffer
, vshader
, pshader
, gl_info
);
4319 TRACE("Attaching GLSL shader object %u to program %u\n", reorder_shader_id
, programId
);
4320 GL_EXTCALL(glAttachObjectARB(programId
, reorder_shader_id
));
4321 checkGLcall("glAttachObjectARB");
4322 /* Flag the reorder function for deletion, then it will be freed automatically when the program
4325 GL_EXTCALL(glDeleteObjectARB(reorder_shader_id
));
4327 TRACE("Attaching GLSL shader object %u to program %u\n", vshader_id
, programId
);
4328 GL_EXTCALL(glAttachObjectARB(programId
, vshader_id
));
4329 checkGLcall("glAttachObjectARB");
4331 /* Bind vertex attributes to a corresponding index number to match
4332 * the same index numbers as ARB_vertex_programs (makes loading
4333 * vertex attributes simpler). With this method, we can use the
4334 * exact same code to load the attributes later for both ARB and
4337 * We have to do this here because we need to know the Program ID
4338 * in order to make the bindings work, and it has to be done prior
4339 * to linking the GLSL program. */
4340 for (i
= 0; map
; map
>>= 1, ++i
)
4342 if (!(map
& 1)) continue;
4344 snprintf(tmp_name
, sizeof(tmp_name
), "attrib%u", i
);
4345 GL_EXTCALL(glBindAttribLocationARB(programId
, i
, tmp_name
));
4347 checkGLcall("glBindAttribLocationARB");
4349 list_add_head(&((IWineD3DBaseShaderImpl
*)vshader
)->baseShader
.linked_programs
, &entry
->vshader_entry
);
4352 /* Attach GLSL pshader */
4355 GLhandleARB pshader_id
= find_glsl_pshader(context
, &priv
->shader_buffer
,
4356 (IWineD3DPixelShaderImpl
*)pshader
, &ps_compile_args
, &entry
->np2Fixup_info
);
4357 TRACE("Attaching GLSL shader object %u to program %u\n", pshader_id
, programId
);
4358 GL_EXTCALL(glAttachObjectARB(programId
, pshader_id
));
4359 checkGLcall("glAttachObjectARB");
4361 list_add_head(&((IWineD3DBaseShaderImpl
*)pshader
)->baseShader
.linked_programs
, &entry
->pshader_entry
);
4364 /* Link the program */
4365 TRACE("Linking GLSL shader program %u\n", programId
);
4366 GL_EXTCALL(glLinkProgramARB(programId
));
4367 shader_glsl_validate_link(gl_info
, programId
);
4369 entry
->vuniformF_locations
= HeapAlloc(GetProcessHeap(), 0,
4370 sizeof(GLhandleARB
) * gl_info
->limits
.glsl_vs_float_constants
);
4371 for (i
= 0; i
< gl_info
->limits
.glsl_vs_float_constants
; ++i
)
4373 snprintf(glsl_name
, sizeof(glsl_name
), "VC[%i]", i
);
4374 entry
->vuniformF_locations
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, glsl_name
));
4376 for (i
= 0; i
< MAX_CONST_I
; ++i
)
4378 snprintf(glsl_name
, sizeof(glsl_name
), "VI[%i]", i
);
4379 entry
->vuniformI_locations
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, glsl_name
));
4381 entry
->puniformF_locations
= HeapAlloc(GetProcessHeap(), 0,
4382 sizeof(GLhandleARB
) * gl_info
->limits
.glsl_ps_float_constants
);
4383 for (i
= 0; i
< gl_info
->limits
.glsl_ps_float_constants
; ++i
)
4385 snprintf(glsl_name
, sizeof(glsl_name
), "PC[%i]", i
);
4386 entry
->puniformF_locations
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, glsl_name
));
4388 for (i
= 0; i
< MAX_CONST_I
; ++i
)
4390 snprintf(glsl_name
, sizeof(glsl_name
), "PI[%i]", i
);
4391 entry
->puniformI_locations
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, glsl_name
));
4397 for(i
= 0; i
< MAX_TEXTURES
; i
++) {
4398 sprintf(name
, "bumpenvmat%u", i
);
4399 entry
->bumpenvmat_location
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, name
));
4400 sprintf(name
, "luminancescale%u", i
);
4401 entry
->luminancescale_location
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, name
));
4402 sprintf(name
, "luminanceoffset%u", i
);
4403 entry
->luminanceoffset_location
[i
] = GL_EXTCALL(glGetUniformLocationARB(programId
, name
));
4406 if (ps_compile_args
.np2_fixup
) {
4407 if (entry
->np2Fixup_info
) {
4408 entry
->np2Fixup_location
= GL_EXTCALL(glGetUniformLocationARB(programId
, "PsamplerNP2Fixup"));
4410 FIXME("NP2 texcoord fixup needed for this pixelshader, but no fixup uniform found.\n");
4415 entry
->posFixup_location
= GL_EXTCALL(glGetUniformLocationARB(programId
, "posFixup"));
4416 entry
->ycorrection_location
= GL_EXTCALL(glGetUniformLocationARB(programId
, "ycorrection"));
4417 checkGLcall("Find glsl program uniform locations");
4420 && ((IWineD3DPixelShaderImpl
*)pshader
)->baseShader
.reg_maps
.shader_version
.major
>= 3
4421 && ((IWineD3DPixelShaderImpl
*)pshader
)->declared_in_count
> vec4_varyings(3, gl_info
))
4423 TRACE("Shader %d needs vertex color clamping disabled\n", programId
);
4424 entry
->vertex_color_clamp
= GL_FALSE
;
4426 entry
->vertex_color_clamp
= GL_FIXED_ONLY_ARB
;
4429 /* Set the shader to allow uniform loading on it */
4430 GL_EXTCALL(glUseProgramObjectARB(programId
));
4431 checkGLcall("glUseProgramObjectARB(programId)");
4433 /* Load the vertex and pixel samplers now. The function that finds the mappings makes sure
4434 * that it stays the same for each vertexshader-pixelshader pair(=linked glsl program). If
4435 * a pshader with fixed function pipeline is used there are no vertex samplers, and if a
4436 * vertex shader with fixed function pixel processing is used we make sure that the card
4437 * supports enough samplers to allow the max number of vertex samplers with all possible
4438 * fixed function fragment processing setups. So once the program is linked these samplers
4441 if (vshader
) shader_glsl_load_vsamplers(gl_info
, device
->texUnitMap
, programId
);
4442 if (pshader
) shader_glsl_load_psamplers(gl_info
, device
->texUnitMap
, programId
);
4444 /* If the local constants do not have to be loaded with the environment constants,
4445 * load them now to have them hardcoded in the GLSL program. This saves some CPU cycles
4448 if (pshader
&& !((IWineD3DBaseShaderImpl
*)pshader
)->baseShader
.load_local_constsF
)
4450 hardcode_local_constants((IWineD3DBaseShaderImpl
*) pshader
, gl_info
, programId
, 'P');
4452 if (vshader
&& !((IWineD3DBaseShaderImpl
*)vshader
)->baseShader
.load_local_constsF
)
4454 hardcode_local_constants((IWineD3DBaseShaderImpl
*) vshader
, gl_info
, programId
, 'V');
4458 /* GL locking is done by the caller */
4459 static GLhandleARB
create_glsl_blt_shader(const struct wined3d_gl_info
*gl_info
, enum tex_types tex_type
, BOOL masked
)
4461 GLhandleARB program_id
;
4462 GLhandleARB vshader_id
, pshader_id
;
4463 const char *blt_pshader
;
4465 static const char *blt_vshader
=
4469 " gl_Position = gl_Vertex;\n"
4470 " gl_FrontColor = vec4(1.0);\n"
4471 " gl_TexCoord[0] = gl_MultiTexCoord0;\n"
4474 static const char * const blt_pshaders_full
[tex_type_count
] =
4480 "uniform sampler2D sampler;\n"
4483 " gl_FragDepth = texture2D(sampler, gl_TexCoord[0].xy).x;\n"
4489 "uniform samplerCube sampler;\n"
4492 " gl_FragDepth = textureCube(sampler, gl_TexCoord[0].xyz).x;\n"
4496 "#extension GL_ARB_texture_rectangle : enable\n"
4497 "uniform sampler2DRect sampler;\n"
4500 " gl_FragDepth = texture2DRect(sampler, gl_TexCoord[0].xy).x;\n"
4504 static const char * const blt_pshaders_masked
[tex_type_count
] =
4510 "uniform sampler2D sampler;\n"
4511 "uniform vec4 mask;\n"
4514 " if (all(lessThan(gl_FragCoord.xy, mask.zw))) discard;\n"
4515 " gl_FragDepth = texture2D(sampler, gl_TexCoord[0].xy).x;\n"
4521 "uniform samplerCube sampler;\n"
4522 "uniform vec4 mask;\n"
4525 " if (all(lessThan(gl_FragCoord.xy, mask.zw))) discard;\n"
4526 " gl_FragDepth = textureCube(sampler, gl_TexCoord[0].xyz).x;\n"
4530 "#extension GL_ARB_texture_rectangle : enable\n"
4531 "uniform sampler2DRect sampler;\n"
4532 "uniform vec4 mask;\n"
4535 " if (all(lessThan(gl_FragCoord.xy, mask.zw))) discard;\n"
4536 " gl_FragDepth = texture2DRect(sampler, gl_TexCoord[0].xy).x;\n"
4540 blt_pshader
= masked
? blt_pshaders_masked
[tex_type
] : blt_pshaders_full
[tex_type
];
4543 FIXME("tex_type %#x not supported\n", tex_type
);
4547 vshader_id
= GL_EXTCALL(glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB
));
4548 GL_EXTCALL(glShaderSourceARB(vshader_id
, 1, &blt_vshader
, NULL
));
4549 GL_EXTCALL(glCompileShaderARB(vshader_id
));
4551 pshader_id
= GL_EXTCALL(glCreateShaderObjectARB(GL_FRAGMENT_SHADER_ARB
));
4552 GL_EXTCALL(glShaderSourceARB(pshader_id
, 1, &blt_pshader
, NULL
));
4553 GL_EXTCALL(glCompileShaderARB(pshader_id
));
4555 program_id
= GL_EXTCALL(glCreateProgramObjectARB());
4556 GL_EXTCALL(glAttachObjectARB(program_id
, vshader_id
));
4557 GL_EXTCALL(glAttachObjectARB(program_id
, pshader_id
));
4558 GL_EXTCALL(glLinkProgramARB(program_id
));
4560 shader_glsl_validate_link(gl_info
, program_id
);
4562 /* Once linked we can mark the shaders for deletion. They will be deleted once the program
4565 GL_EXTCALL(glDeleteObjectARB(vshader_id
));
4566 GL_EXTCALL(glDeleteObjectARB(pshader_id
));
4570 /* GL locking is done by the caller */
4571 static void shader_glsl_select(const struct wined3d_context
*context
, BOOL usePS
, BOOL useVS
)
4573 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
4574 IWineD3DDeviceImpl
*device
= context
->swapchain
->device
;
4575 struct shader_glsl_priv
*priv
= device
->shader_priv
;
4576 GLhandleARB program_id
= 0;
4577 GLenum old_vertex_color_clamp
, current_vertex_color_clamp
;
4579 old_vertex_color_clamp
= priv
->glsl_program
? priv
->glsl_program
->vertex_color_clamp
: GL_FIXED_ONLY_ARB
;
4581 if (useVS
|| usePS
) set_glsl_shader_program(context
, device
, usePS
, useVS
);
4582 else priv
->glsl_program
= NULL
;
4584 current_vertex_color_clamp
= priv
->glsl_program
? priv
->glsl_program
->vertex_color_clamp
: GL_FIXED_ONLY_ARB
;
4586 if (old_vertex_color_clamp
!= current_vertex_color_clamp
)
4588 if (gl_info
->supported
[ARB_COLOR_BUFFER_FLOAT
])
4590 GL_EXTCALL(glClampColorARB(GL_CLAMP_VERTEX_COLOR_ARB
, current_vertex_color_clamp
));
4591 checkGLcall("glClampColorARB");
4595 FIXME("vertex color clamp needs to be changed, but extension not supported.\n");
4599 program_id
= priv
->glsl_program
? priv
->glsl_program
->programId
: 0;
4600 if (program_id
) TRACE("Using GLSL program %u\n", program_id
);
4601 GL_EXTCALL(glUseProgramObjectARB(program_id
));
4602 checkGLcall("glUseProgramObjectARB");
4604 /* In case that NP2 texcoord fixup data is found for the selected program, trigger a reload of the
4605 * constants. This has to be done because it can't be guaranteed that sampler() (from state.c) is
4606 * called between selecting the shader and using it, which results in wrong fixup for some frames. */
4607 if (priv
->glsl_program
&& priv
->glsl_program
->np2Fixup_info
)
4609 shader_glsl_load_np2fixup_constants((IWineD3DDevice
*)device
, usePS
, useVS
);
4613 /* GL locking is done by the caller */
4614 static void shader_glsl_select_depth_blt(IWineD3DDevice
*iface
,
4615 enum tex_types tex_type
, const SIZE
*ds_mask_size
)
4617 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
4618 const struct wined3d_gl_info
*gl_info
= &This
->adapter
->gl_info
;
4619 BOOL masked
= ds_mask_size
->cx
&& ds_mask_size
->cy
;
4620 struct shader_glsl_priv
*priv
= This
->shader_priv
;
4621 GLhandleARB
*blt_program
;
4624 blt_program
= masked
? &priv
->depth_blt_program_masked
[tex_type
] : &priv
->depth_blt_program_full
[tex_type
];
4627 *blt_program
= create_glsl_blt_shader(gl_info
, tex_type
, masked
);
4628 loc
= GL_EXTCALL(glGetUniformLocationARB(*blt_program
, "sampler"));
4629 GL_EXTCALL(glUseProgramObjectARB(*blt_program
));
4630 GL_EXTCALL(glUniform1iARB(loc
, 0));
4634 GL_EXTCALL(glUseProgramObjectARB(*blt_program
));
4639 loc
= GL_EXTCALL(glGetUniformLocationARB(*blt_program
, "mask"));
4640 GL_EXTCALL(glUniform4fARB(loc
, 0.0f
, 0.0f
, (float)ds_mask_size
->cx
, (float)ds_mask_size
->cy
));
4644 /* GL locking is done by the caller */
4645 static void shader_glsl_deselect_depth_blt(IWineD3DDevice
*iface
) {
4646 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
4647 const struct wined3d_gl_info
*gl_info
= &This
->adapter
->gl_info
;
4648 struct shader_glsl_priv
*priv
= This
->shader_priv
;
4649 GLhandleARB program_id
;
4651 program_id
= priv
->glsl_program
? priv
->glsl_program
->programId
: 0;
4652 if (program_id
) TRACE("Using GLSL program %u\n", program_id
);
4654 GL_EXTCALL(glUseProgramObjectARB(program_id
));
4655 checkGLcall("glUseProgramObjectARB");
4658 static void shader_glsl_destroy(IWineD3DBaseShader
*iface
) {
4659 const struct list
*linked_programs
;
4660 IWineD3DBaseShaderImpl
*This
= (IWineD3DBaseShaderImpl
*) iface
;
4661 IWineD3DDeviceImpl
*device
= (IWineD3DDeviceImpl
*)This
->baseShader
.device
;
4662 struct shader_glsl_priv
*priv
= device
->shader_priv
;
4663 const struct wined3d_gl_info
*gl_info
;
4664 struct wined3d_context
*context
;
4666 /* Note: Do not use QueryInterface here to find out which shader type this is because this code
4667 * can be called from IWineD3DBaseShader::Release
4669 char pshader
= shader_is_pshader_version(This
->baseShader
.reg_maps
.shader_version
.type
);
4672 struct glsl_pshader_private
*shader_data
;
4673 shader_data
= This
->baseShader
.backend_data
;
4674 if(!shader_data
|| shader_data
->num_gl_shaders
== 0)
4676 HeapFree(GetProcessHeap(), 0, shader_data
);
4677 This
->baseShader
.backend_data
= NULL
;
4681 context
= context_acquire(device
, NULL
);
4682 gl_info
= context
->gl_info
;
4684 if (priv
->glsl_program
&& (IWineD3DBaseShader
*)priv
->glsl_program
->pshader
== iface
)
4687 shader_glsl_select(context
, FALSE
, FALSE
);
4691 struct glsl_vshader_private
*shader_data
;
4692 shader_data
= This
->baseShader
.backend_data
;
4693 if(!shader_data
|| shader_data
->num_gl_shaders
== 0)
4695 HeapFree(GetProcessHeap(), 0, shader_data
);
4696 This
->baseShader
.backend_data
= NULL
;
4700 context
= context_acquire(device
, NULL
);
4701 gl_info
= context
->gl_info
;
4703 if (priv
->glsl_program
&& (IWineD3DBaseShader
*)priv
->glsl_program
->vshader
== iface
)
4706 shader_glsl_select(context
, FALSE
, FALSE
);
4711 linked_programs
= &This
->baseShader
.linked_programs
;
4713 TRACE("Deleting linked programs\n");
4714 if (linked_programs
->next
) {
4715 struct glsl_shader_prog_link
*entry
, *entry2
;
4719 LIST_FOR_EACH_ENTRY_SAFE(entry
, entry2
, linked_programs
, struct glsl_shader_prog_link
, pshader_entry
) {
4720 delete_glsl_program_entry(priv
, gl_info
, entry
);
4723 LIST_FOR_EACH_ENTRY_SAFE(entry
, entry2
, linked_programs
, struct glsl_shader_prog_link
, vshader_entry
) {
4724 delete_glsl_program_entry(priv
, gl_info
, entry
);
4732 struct glsl_pshader_private
*shader_data
= This
->baseShader
.backend_data
;
4735 for(i
= 0; i
< shader_data
->num_gl_shaders
; i
++) {
4736 TRACE("deleting pshader %u\n", shader_data
->gl_shaders
[i
].prgId
);
4737 GL_EXTCALL(glDeleteObjectARB(shader_data
->gl_shaders
[i
].prgId
));
4738 checkGLcall("glDeleteObjectARB");
4741 HeapFree(GetProcessHeap(), 0, shader_data
->gl_shaders
);
4746 struct glsl_vshader_private
*shader_data
= This
->baseShader
.backend_data
;
4749 for(i
= 0; i
< shader_data
->num_gl_shaders
; i
++) {
4750 TRACE("deleting vshader %u\n", shader_data
->gl_shaders
[i
].prgId
);
4751 GL_EXTCALL(glDeleteObjectARB(shader_data
->gl_shaders
[i
].prgId
));
4752 checkGLcall("glDeleteObjectARB");
4755 HeapFree(GetProcessHeap(), 0, shader_data
->gl_shaders
);
4758 HeapFree(GetProcessHeap(), 0, This
->baseShader
.backend_data
);
4759 This
->baseShader
.backend_data
= NULL
;
4761 context_release(context
);
4764 static int glsl_program_key_compare(const void *key
, const struct wine_rb_entry
*entry
)
4766 const glsl_program_key_t
*k
= key
;
4767 const struct glsl_shader_prog_link
*prog
= WINE_RB_ENTRY_VALUE(entry
,
4768 const struct glsl_shader_prog_link
, program_lookup_entry
);
4771 if (k
->vshader
> prog
->vshader
) return 1;
4772 else if (k
->vshader
< prog
->vshader
) return -1;
4774 if (k
->pshader
> prog
->pshader
) return 1;
4775 else if (k
->pshader
< prog
->pshader
) return -1;
4777 if (k
->vshader
&& (cmp
= memcmp(&k
->vs_args
, &prog
->vs_args
, sizeof(prog
->vs_args
)))) return cmp
;
4778 if (k
->pshader
&& (cmp
= memcmp(&k
->ps_args
, &prog
->ps_args
, sizeof(prog
->ps_args
)))) return cmp
;
4783 static BOOL
constant_heap_init(struct constant_heap
*heap
, unsigned int constant_count
)
4785 SIZE_T size
= (constant_count
+ 1) * sizeof(*heap
->entries
) + constant_count
* sizeof(*heap
->positions
);
4786 void *mem
= HeapAlloc(GetProcessHeap(), 0, size
);
4790 ERR("Failed to allocate memory\n");
4794 heap
->entries
= mem
;
4795 heap
->entries
[1].version
= 0;
4796 heap
->positions
= (unsigned int *)(heap
->entries
+ constant_count
+ 1);
4802 static void constant_heap_free(struct constant_heap
*heap
)
4804 HeapFree(GetProcessHeap(), 0, heap
->entries
);
4807 static const struct wine_rb_functions wined3d_glsl_program_rb_functions
=
4812 glsl_program_key_compare
,
4815 static HRESULT
shader_glsl_alloc(IWineD3DDevice
*iface
) {
4816 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
4817 const struct wined3d_gl_info
*gl_info
= &This
->adapter
->gl_info
;
4818 struct shader_glsl_priv
*priv
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(struct shader_glsl_priv
));
4819 SIZE_T stack_size
= wined3d_log2i(max(gl_info
->limits
.glsl_vs_float_constants
,
4820 gl_info
->limits
.glsl_ps_float_constants
)) + 1;
4822 if (!shader_buffer_init(&priv
->shader_buffer
))
4824 ERR("Failed to initialize shader buffer.\n");
4828 priv
->stack
= HeapAlloc(GetProcessHeap(), 0, stack_size
* sizeof(*priv
->stack
));
4831 ERR("Failed to allocate memory.\n");
4835 if (!constant_heap_init(&priv
->vconst_heap
, gl_info
->limits
.glsl_vs_float_constants
))
4837 ERR("Failed to initialize vertex shader constant heap\n");
4841 if (!constant_heap_init(&priv
->pconst_heap
, gl_info
->limits
.glsl_ps_float_constants
))
4843 ERR("Failed to initialize pixel shader constant heap\n");
4847 if (wine_rb_init(&priv
->program_lookup
, &wined3d_glsl_program_rb_functions
) == -1)
4849 ERR("Failed to initialize rbtree.\n");
4853 priv
->next_constant_version
= 1;
4855 This
->shader_priv
= priv
;
4859 constant_heap_free(&priv
->pconst_heap
);
4860 constant_heap_free(&priv
->vconst_heap
);
4861 HeapFree(GetProcessHeap(), 0, priv
->stack
);
4862 shader_buffer_free(&priv
->shader_buffer
);
4863 HeapFree(GetProcessHeap(), 0, priv
);
4864 return E_OUTOFMEMORY
;
4867 /* Context activation is done by the caller. */
4868 static void shader_glsl_free(IWineD3DDevice
*iface
) {
4869 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
4870 const struct wined3d_gl_info
*gl_info
= &This
->adapter
->gl_info
;
4871 struct shader_glsl_priv
*priv
= This
->shader_priv
;
4875 for (i
= 0; i
< tex_type_count
; ++i
)
4877 if (priv
->depth_blt_program_full
[i
])
4879 GL_EXTCALL(glDeleteObjectARB(priv
->depth_blt_program_full
[i
]));
4881 if (priv
->depth_blt_program_masked
[i
])
4883 GL_EXTCALL(glDeleteObjectARB(priv
->depth_blt_program_masked
[i
]));
4888 wine_rb_destroy(&priv
->program_lookup
, NULL
, NULL
);
4889 constant_heap_free(&priv
->pconst_heap
);
4890 constant_heap_free(&priv
->vconst_heap
);
4891 HeapFree(GetProcessHeap(), 0, priv
->stack
);
4892 shader_buffer_free(&priv
->shader_buffer
);
4894 HeapFree(GetProcessHeap(), 0, This
->shader_priv
);
4895 This
->shader_priv
= NULL
;
4898 static BOOL
shader_glsl_dirty_const(IWineD3DDevice
*iface
) {
4899 /* TODO: GL_EXT_bindable_uniform can be used to share constants across shaders */
4903 static void shader_glsl_get_caps(const struct wined3d_gl_info
*gl_info
, struct shader_caps
*pCaps
)
4905 /* Nvidia Geforce6/7 or Ati R4xx/R5xx cards with GLSL support, support VS 3.0 but older Nvidia/Ati
4906 * models with GLSL support only support 2.0. In case of nvidia we can detect VS 2.0 support based
4907 * on the version of NV_vertex_program.
4908 * For Ati cards there's no way using glsl (it abstracts the lowlevel info away) and also not
4909 * using ARB_vertex_program. It is safe to assume that when a card supports pixel shader 2.0 it
4910 * supports vertex shader 2.0 too and the way around. We can detect ps2.0 using the maximum number
4911 * of native instructions, so use that here. For more info see the pixel shader versioning code below.
4913 if ((gl_info
->supported
[NV_VERTEX_PROGRAM2
] && !gl_info
->supported
[NV_VERTEX_PROGRAM3
])
4914 || gl_info
->limits
.arb_ps_instructions
<= 512)
4915 pCaps
->VertexShaderVersion
= WINED3DVS_VERSION(2,0);
4917 pCaps
->VertexShaderVersion
= WINED3DVS_VERSION(3,0);
4918 TRACE_(d3d_caps
)("Hardware vertex shader version %d.%d enabled (GLSL)\n", (pCaps
->VertexShaderVersion
>> 8) & 0xff, pCaps
->VertexShaderVersion
& 0xff);
4919 pCaps
->MaxVertexShaderConst
= gl_info
->limits
.glsl_vs_float_constants
;
4921 /* Older DX9-class videocards (GeforceFX / Radeon >9500/X*00) only support pixel shader 2.0/2.0a/2.0b.
4922 * In OpenGL the extensions related to GLSL abstract lowlevel GL info away which is needed
4923 * to distinguish between 2.0 and 3.0 (and 2.0a/2.0b). In case of Nvidia we use their fragment
4924 * program extensions. On other hardware including ATI GL_ARB_fragment_program offers the info
4925 * in max native instructions. Intel and others also offer the info in this extension but they
4926 * don't support GLSL (at least on Windows).
4928 * PS2.0 requires at least 96 instructions, 2.0a/2.0b go up to 512. Assume that if the number
4929 * of instructions is 512 or less we have to do with ps2.0 hardware.
4930 * 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.
4932 if ((gl_info
->supported
[NV_FRAGMENT_PROGRAM
] && !gl_info
->supported
[NV_FRAGMENT_PROGRAM2
])
4933 || gl_info
->limits
.arb_ps_instructions
<= 512)
4934 pCaps
->PixelShaderVersion
= WINED3DPS_VERSION(2,0);
4936 pCaps
->PixelShaderVersion
= WINED3DPS_VERSION(3,0);
4938 pCaps
->MaxPixelShaderConst
= gl_info
->limits
.glsl_ps_float_constants
;
4940 /* FIXME: The following line is card dependent. -8.0 to 8.0 is the
4941 * Direct3D minimum requirement.
4943 * Both GL_ARB_fragment_program and GLSL require a "maximum representable magnitude"
4944 * of colors to be 2^10, and 2^32 for other floats. Should we use 1024 here?
4946 * The problem is that the refrast clamps temporary results in the shader to
4947 * [-MaxValue;+MaxValue]. If the card's max value is bigger than the one we advertize here,
4948 * then applications may miss the clamping behavior. On the other hand, if it is smaller,
4949 * the shader will generate incorrect results too. Unfortunately, GL deliberately doesn't
4950 * offer a way to query this.
4952 pCaps
->PixelShader1xMaxValue
= 8.0;
4953 TRACE_(d3d_caps
)("Hardware pixel shader version %d.%d enabled (GLSL)\n", (pCaps
->PixelShaderVersion
>> 8) & 0xff, pCaps
->PixelShaderVersion
& 0xff);
4955 pCaps
->VSClipping
= TRUE
;
4958 static BOOL
shader_glsl_color_fixup_supported(struct color_fixup_desc fixup
)
4960 if (TRACE_ON(d3d_shader
) && TRACE_ON(d3d
))
4962 TRACE("Checking support for fixup:\n");
4963 dump_color_fixup_desc(fixup
);
4966 /* We support everything except YUV conversions. */
4967 if (!is_complex_fixup(fixup
))
4973 TRACE("[FAILED]\n");
4977 static const SHADER_HANDLER shader_glsl_instruction_handler_table
[WINED3DSIH_TABLE_SIZE
] =
4979 /* WINED3DSIH_ABS */ shader_glsl_map2gl
,
4980 /* WINED3DSIH_ADD */ shader_glsl_arith
,
4981 /* WINED3DSIH_BEM */ shader_glsl_bem
,
4982 /* WINED3DSIH_BREAK */ shader_glsl_break
,
4983 /* WINED3DSIH_BREAKC */ shader_glsl_breakc
,
4984 /* WINED3DSIH_BREAKP */ NULL
,
4985 /* WINED3DSIH_CALL */ shader_glsl_call
,
4986 /* WINED3DSIH_CALLNZ */ shader_glsl_callnz
,
4987 /* WINED3DSIH_CMP */ shader_glsl_cmp
,
4988 /* WINED3DSIH_CND */ shader_glsl_cnd
,
4989 /* WINED3DSIH_CRS */ shader_glsl_cross
,
4990 /* WINED3DSIH_CUT */ NULL
,
4991 /* WINED3DSIH_DCL */ NULL
,
4992 /* WINED3DSIH_DEF */ NULL
,
4993 /* WINED3DSIH_DEFB */ NULL
,
4994 /* WINED3DSIH_DEFI */ NULL
,
4995 /* WINED3DSIH_DP2ADD */ shader_glsl_dp2add
,
4996 /* WINED3DSIH_DP3 */ shader_glsl_dot
,
4997 /* WINED3DSIH_DP4 */ shader_glsl_dot
,
4998 /* WINED3DSIH_DST */ shader_glsl_dst
,
4999 /* WINED3DSIH_DSX */ shader_glsl_map2gl
,
5000 /* WINED3DSIH_DSY */ shader_glsl_map2gl
,
5001 /* WINED3DSIH_ELSE */ shader_glsl_else
,
5002 /* WINED3DSIH_EMIT */ NULL
,
5003 /* WINED3DSIH_ENDIF */ shader_glsl_end
,
5004 /* WINED3DSIH_ENDLOOP */ shader_glsl_end
,
5005 /* WINED3DSIH_ENDREP */ shader_glsl_end
,
5006 /* WINED3DSIH_EXP */ shader_glsl_map2gl
,
5007 /* WINED3DSIH_EXPP */ shader_glsl_expp
,
5008 /* WINED3DSIH_FRC */ shader_glsl_map2gl
,
5009 /* WINED3DSIH_IADD */ NULL
,
5010 /* WINED3DSIH_IF */ shader_glsl_if
,
5011 /* WINED3DSIH_IFC */ shader_glsl_ifc
,
5012 /* WINED3DSIH_IGE */ NULL
,
5013 /* WINED3DSIH_LABEL */ shader_glsl_label
,
5014 /* WINED3DSIH_LIT */ shader_glsl_lit
,
5015 /* WINED3DSIH_LOG */ shader_glsl_log
,
5016 /* WINED3DSIH_LOGP */ shader_glsl_log
,
5017 /* WINED3DSIH_LOOP */ shader_glsl_loop
,
5018 /* WINED3DSIH_LRP */ shader_glsl_lrp
,
5019 /* WINED3DSIH_LT */ NULL
,
5020 /* WINED3DSIH_M3x2 */ shader_glsl_mnxn
,
5021 /* WINED3DSIH_M3x3 */ shader_glsl_mnxn
,
5022 /* WINED3DSIH_M3x4 */ shader_glsl_mnxn
,
5023 /* WINED3DSIH_M4x3 */ shader_glsl_mnxn
,
5024 /* WINED3DSIH_M4x4 */ shader_glsl_mnxn
,
5025 /* WINED3DSIH_MAD */ shader_glsl_mad
,
5026 /* WINED3DSIH_MAX */ shader_glsl_map2gl
,
5027 /* WINED3DSIH_MIN */ shader_glsl_map2gl
,
5028 /* WINED3DSIH_MOV */ shader_glsl_mov
,
5029 /* WINED3DSIH_MOVA */ shader_glsl_mov
,
5030 /* WINED3DSIH_MUL */ shader_glsl_arith
,
5031 /* WINED3DSIH_NOP */ NULL
,
5032 /* WINED3DSIH_NRM */ shader_glsl_nrm
,
5033 /* WINED3DSIH_PHASE */ NULL
,
5034 /* WINED3DSIH_POW */ shader_glsl_pow
,
5035 /* WINED3DSIH_RCP */ shader_glsl_rcp
,
5036 /* WINED3DSIH_REP */ shader_glsl_rep
,
5037 /* WINED3DSIH_RET */ shader_glsl_ret
,
5038 /* WINED3DSIH_RSQ */ shader_glsl_rsq
,
5039 /* WINED3DSIH_SETP */ NULL
,
5040 /* WINED3DSIH_SGE */ shader_glsl_compare
,
5041 /* WINED3DSIH_SGN */ shader_glsl_sgn
,
5042 /* WINED3DSIH_SINCOS */ shader_glsl_sincos
,
5043 /* WINED3DSIH_SLT */ shader_glsl_compare
,
5044 /* WINED3DSIH_SUB */ shader_glsl_arith
,
5045 /* WINED3DSIH_TEX */ shader_glsl_tex
,
5046 /* WINED3DSIH_TEXBEM */ shader_glsl_texbem
,
5047 /* WINED3DSIH_TEXBEML */ shader_glsl_texbem
,
5048 /* WINED3DSIH_TEXCOORD */ shader_glsl_texcoord
,
5049 /* WINED3DSIH_TEXDEPTH */ shader_glsl_texdepth
,
5050 /* WINED3DSIH_TEXDP3 */ shader_glsl_texdp3
,
5051 /* WINED3DSIH_TEXDP3TEX */ shader_glsl_texdp3tex
,
5052 /* WINED3DSIH_TEXKILL */ shader_glsl_texkill
,
5053 /* WINED3DSIH_TEXLDD */ shader_glsl_texldd
,
5054 /* WINED3DSIH_TEXLDL */ shader_glsl_texldl
,
5055 /* WINED3DSIH_TEXM3x2DEPTH */ shader_glsl_texm3x2depth
,
5056 /* WINED3DSIH_TEXM3x2PAD */ shader_glsl_texm3x2pad
,
5057 /* WINED3DSIH_TEXM3x2TEX */ shader_glsl_texm3x2tex
,
5058 /* WINED3DSIH_TEXM3x3 */ shader_glsl_texm3x3
,
5059 /* WINED3DSIH_TEXM3x3DIFF */ NULL
,
5060 /* WINED3DSIH_TEXM3x3PAD */ shader_glsl_texm3x3pad
,
5061 /* WINED3DSIH_TEXM3x3SPEC */ shader_glsl_texm3x3spec
,
5062 /* WINED3DSIH_TEXM3x3TEX */ shader_glsl_texm3x3tex
,
5063 /* WINED3DSIH_TEXM3x3VSPEC */ shader_glsl_texm3x3vspec
,
5064 /* WINED3DSIH_TEXREG2AR */ shader_glsl_texreg2ar
,
5065 /* WINED3DSIH_TEXREG2GB */ shader_glsl_texreg2gb
,
5066 /* WINED3DSIH_TEXREG2RGB */ shader_glsl_texreg2rgb
,
5069 static void shader_glsl_handle_instruction(const struct wined3d_shader_instruction
*ins
) {
5070 SHADER_HANDLER hw_fct
;
5072 /* Select handler */
5073 hw_fct
= shader_glsl_instruction_handler_table
[ins
->handler_idx
];
5075 /* Unhandled opcode */
5078 FIXME("Backend can't handle opcode %#x\n", ins
->handler_idx
);
5083 shader_glsl_add_instruction_modifiers(ins
);
5086 const shader_backend_t glsl_shader_backend
= {
5087 shader_glsl_handle_instruction
,
5089 shader_glsl_select_depth_blt
,
5090 shader_glsl_deselect_depth_blt
,
5091 shader_glsl_update_float_vertex_constants
,
5092 shader_glsl_update_float_pixel_constants
,
5093 shader_glsl_load_constants
,
5094 shader_glsl_load_np2fixup_constants
,
5095 shader_glsl_destroy
,
5098 shader_glsl_dirty_const
,
5099 shader_glsl_get_caps
,
5100 shader_glsl_color_fixup_supported
,