1 /**************************************************************************
3 * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
5 * Copyright 2009 VMware, Inc. All Rights Reserved.
6 * Copyright © 2010-2011 Intel Corporation
8 * Permission is hereby granted, free of charge, to any person obtaining a
9 * copy of this software and associated documentation files (the
10 * "Software"), to deal in the Software without restriction, including
11 * without limitation the rights to use, copy, modify, merge, publish,
12 * distribute, sub license, and/or sell copies of the Software, and to
13 * permit persons to whom the Software is furnished to do so, subject to
14 * the following conditions:
16 * The above copyright notice and this permission notice (including the
17 * next paragraph) shall be included in all copies or substantial portions
20 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
21 * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
22 * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
23 * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
24 * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
25 * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
26 * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
28 **************************************************************************/
34 #include "main/uniforms.h"
35 #include "main/macros.h"
36 #include "program/program.h"
37 #include "program/prog_parameter.h"
38 #include "program/prog_cache.h"
39 #include "program/prog_instruction.h"
40 #include "program/prog_print.h"
41 #include "program/prog_statevars.h"
42 #include "program/programopt.h"
43 #include "texenvprogram.h"
45 #include "../glsl/glsl_types.h"
46 #include "../glsl/ir.h"
47 #include "../glsl/glsl_symbol_table.h"
48 #include "../glsl/glsl_parser_extras.h"
49 #include "../glsl/ir_optimization.h"
50 #include "../glsl/ir_print_visitor.h"
51 #include "../program/ir_to_mesa.h"
54 * Note on texture units:
56 * The number of texture units supported by fixed-function fragment
57 * processing is MAX_TEXTURE_COORD_UNITS, not MAX_TEXTURE_IMAGE_UNITS.
58 * That's because there's a one-to-one correspondence between texture
59 * coordinates and samplers in fixed-function processing.
61 * Since fixed-function vertex processing is limited to MAX_TEXTURE_COORD_UNITS
62 * sets of texcoords, so is fixed-function fragment processing.
64 * We can safely use ctx->Const.MaxTextureUnits for loop bounds.
68 struct texenvprog_cache_item
72 struct gl_shader_program
*data
;
73 struct texenvprog_cache_item
*next
;
77 texenv_doing_secondary_color(struct gl_context
*ctx
)
79 if (ctx
->Light
.Enabled
&&
80 (ctx
->Light
.Model
.ColorControl
== GL_SEPARATE_SPECULAR_COLOR
))
83 if (ctx
->Fog
.ColorSumEnabled
)
91 __extension__ GLubyte Source
:4; /**< SRC_x */
92 __extension__ GLubyte Operand
:3; /**< OPR_x */
94 GLubyte Source
; /**< SRC_x */
95 GLubyte Operand
; /**< OPR_x */
100 GLuint nr_enabled_units
:8;
101 GLuint enabled_units
:8;
102 GLuint separate_specular
:1;
103 GLuint fog_enabled
:1;
104 GLuint fog_mode
:2; /**< FOG_x */
105 GLuint inputs_available
:12;
106 GLuint num_draw_buffers
:4;
108 /* NOTE: This array of structs must be last! (see "keySize" below) */
111 GLuint source_index
:3; /**< TEXTURE_x_INDEX */
113 GLuint ScaleShiftRGB
:2;
114 GLuint ScaleShiftA
:2;
116 GLuint NumArgsRGB
:3; /**< up to MAX_COMBINER_TERMS */
117 GLuint ModeRGB
:5; /**< MODE_x */
119 GLuint NumArgsA
:3; /**< up to MAX_COMBINER_TERMS */
120 GLuint ModeA
:5; /**< MODE_x */
122 struct mode_opt OptRGB
[MAX_COMBINER_TERMS
];
123 struct mode_opt OptA
[MAX_COMBINER_TERMS
];
124 } unit
[MAX_TEXTURE_UNITS
];
130 #define FOG_UNKNOWN 3
132 static GLuint
translate_fog_mode( GLenum mode
)
135 case GL_LINEAR
: return FOG_LINEAR
;
136 case GL_EXP
: return FOG_EXP
;
137 case GL_EXP2
: return FOG_EXP2
;
138 default: return FOG_UNKNOWN
;
142 #define OPR_SRC_COLOR 0
143 #define OPR_ONE_MINUS_SRC_COLOR 1
144 #define OPR_SRC_ALPHA 2
145 #define OPR_ONE_MINUS_SRC_ALPHA 3
148 #define OPR_UNKNOWN 7
150 static GLuint
translate_operand( GLenum operand
)
153 case GL_SRC_COLOR
: return OPR_SRC_COLOR
;
154 case GL_ONE_MINUS_SRC_COLOR
: return OPR_ONE_MINUS_SRC_COLOR
;
155 case GL_SRC_ALPHA
: return OPR_SRC_ALPHA
;
156 case GL_ONE_MINUS_SRC_ALPHA
: return OPR_ONE_MINUS_SRC_ALPHA
;
157 case GL_ZERO
: return OPR_ZERO
;
158 case GL_ONE
: return OPR_ONE
;
165 #define SRC_TEXTURE 0
166 #define SRC_TEXTURE0 1
167 #define SRC_TEXTURE1 2
168 #define SRC_TEXTURE2 3
169 #define SRC_TEXTURE3 4
170 #define SRC_TEXTURE4 5
171 #define SRC_TEXTURE5 6
172 #define SRC_TEXTURE6 7
173 #define SRC_TEXTURE7 8
174 #define SRC_CONSTANT 9
175 #define SRC_PRIMARY_COLOR 10
176 #define SRC_PREVIOUS 11
178 #define SRC_UNKNOWN 15
180 static GLuint
translate_source( GLenum src
)
183 case GL_TEXTURE
: return SRC_TEXTURE
;
191 case GL_TEXTURE7
: return SRC_TEXTURE0
+ (src
- GL_TEXTURE0
);
192 case GL_CONSTANT
: return SRC_CONSTANT
;
193 case GL_PRIMARY_COLOR
: return SRC_PRIMARY_COLOR
;
194 case GL_PREVIOUS
: return SRC_PREVIOUS
;
203 #define MODE_REPLACE 0 /* r = a0 */
204 #define MODE_MODULATE 1 /* r = a0 * a1 */
205 #define MODE_ADD 2 /* r = a0 + a1 */
206 #define MODE_ADD_SIGNED 3 /* r = a0 + a1 - 0.5 */
207 #define MODE_INTERPOLATE 4 /* r = a0 * a2 + a1 * (1 - a2) */
208 #define MODE_SUBTRACT 5 /* r = a0 - a1 */
209 #define MODE_DOT3_RGB 6 /* r = a0 . a1 */
210 #define MODE_DOT3_RGB_EXT 7 /* r = a0 . a1 */
211 #define MODE_DOT3_RGBA 8 /* r = a0 . a1 */
212 #define MODE_DOT3_RGBA_EXT 9 /* r = a0 . a1 */
213 #define MODE_MODULATE_ADD_ATI 10 /* r = a0 * a2 + a1 */
214 #define MODE_MODULATE_SIGNED_ADD_ATI 11 /* r = a0 * a2 + a1 - 0.5 */
215 #define MODE_MODULATE_SUBTRACT_ATI 12 /* r = a0 * a2 - a1 */
216 #define MODE_ADD_PRODUCTS 13 /* r = a0 * a1 + a2 * a3 */
217 #define MODE_ADD_PRODUCTS_SIGNED 14 /* r = a0 * a1 + a2 * a3 - 0.5 */
218 #define MODE_BUMP_ENVMAP_ATI 15 /* special */
219 #define MODE_UNKNOWN 16
222 * Translate GL combiner state into a MODE_x value
224 static GLuint
translate_mode( GLenum envMode
, GLenum mode
)
227 case GL_REPLACE
: return MODE_REPLACE
;
228 case GL_MODULATE
: return MODE_MODULATE
;
230 if (envMode
== GL_COMBINE4_NV
)
231 return MODE_ADD_PRODUCTS
;
235 if (envMode
== GL_COMBINE4_NV
)
236 return MODE_ADD_PRODUCTS_SIGNED
;
238 return MODE_ADD_SIGNED
;
239 case GL_INTERPOLATE
: return MODE_INTERPOLATE
;
240 case GL_SUBTRACT
: return MODE_SUBTRACT
;
241 case GL_DOT3_RGB
: return MODE_DOT3_RGB
;
242 case GL_DOT3_RGB_EXT
: return MODE_DOT3_RGB_EXT
;
243 case GL_DOT3_RGBA
: return MODE_DOT3_RGBA
;
244 case GL_DOT3_RGBA_EXT
: return MODE_DOT3_RGBA_EXT
;
245 case GL_MODULATE_ADD_ATI
: return MODE_MODULATE_ADD_ATI
;
246 case GL_MODULATE_SIGNED_ADD_ATI
: return MODE_MODULATE_SIGNED_ADD_ATI
;
247 case GL_MODULATE_SUBTRACT_ATI
: return MODE_MODULATE_SUBTRACT_ATI
;
248 case GL_BUMP_ENVMAP_ATI
: return MODE_BUMP_ENVMAP_ATI
;
257 * Do we need to clamp the results of the given texture env/combine mode?
258 * If the inputs to the mode are in [0,1] we don't always have to clamp
262 need_saturate( GLuint mode
)
267 case MODE_INTERPOLATE
:
270 case MODE_ADD_SIGNED
:
273 case MODE_DOT3_RGB_EXT
:
275 case MODE_DOT3_RGBA_EXT
:
276 case MODE_MODULATE_ADD_ATI
:
277 case MODE_MODULATE_SIGNED_ADD_ATI
:
278 case MODE_MODULATE_SUBTRACT_ATI
:
279 case MODE_ADD_PRODUCTS
:
280 case MODE_ADD_PRODUCTS_SIGNED
:
281 case MODE_BUMP_ENVMAP_ATI
:
292 * Translate TEXTURE_x_BIT to TEXTURE_x_INDEX.
294 static GLuint
translate_tex_src_bit( GLbitfield bit
)
297 return _mesa_ffs(bit
) - 1;
301 #define VERT_BIT_TEX_ANY (0xff << VERT_ATTRIB_TEX0)
302 #define VERT_RESULT_TEX_ANY (0xff << VERT_RESULT_TEX0)
305 * Identify all possible varying inputs. The fragment program will
306 * never reference non-varying inputs, but will track them via state
309 * This function figures out all the inputs that the fragment program
310 * has access to. The bitmask is later reduced to just those which
311 * are actually referenced.
313 static GLbitfield
get_fp_input_mask( struct gl_context
*ctx
)
316 const GLboolean vertexShader
=
317 (ctx
->Shader
.CurrentVertexProgram
&&
318 ctx
->Shader
.CurrentVertexProgram
->LinkStatus
&&
319 ctx
->Shader
.CurrentVertexProgram
->_LinkedShaders
[MESA_SHADER_VERTEX
]);
320 const GLboolean vertexProgram
= ctx
->VertexProgram
._Enabled
;
321 GLbitfield fp_inputs
= 0x0;
323 if (ctx
->VertexProgram
._Overriden
) {
324 /* Somebody's messing with the vertex program and we don't have
325 * a clue what's happening. Assume that it could be producing
326 * all possible outputs.
330 else if (ctx
->RenderMode
== GL_FEEDBACK
) {
331 /* _NEW_RENDERMODE */
332 fp_inputs
= (FRAG_BIT_COL0
| FRAG_BIT_TEX0
);
334 else if (!(vertexProgram
|| vertexShader
)) {
335 /* Fixed function vertex logic */
337 GLbitfield varying_inputs
= ctx
->varying_vp_inputs
;
339 /* These get generated in the setup routine regardless of the
343 if (ctx
->Point
.PointSprite
)
344 varying_inputs
|= FRAG_BITS_TEX_ANY
;
346 /* First look at what values may be computed by the generated
350 if (ctx
->Light
.Enabled
) {
351 fp_inputs
|= FRAG_BIT_COL0
;
353 if (texenv_doing_secondary_color(ctx
))
354 fp_inputs
|= FRAG_BIT_COL1
;
358 fp_inputs
|= (ctx
->Texture
._TexGenEnabled
|
359 ctx
->Texture
._TexMatEnabled
) << FRAG_ATTRIB_TEX0
;
361 /* Then look at what might be varying as a result of enabled
364 if (varying_inputs
& VERT_BIT_COLOR0
)
365 fp_inputs
|= FRAG_BIT_COL0
;
366 if (varying_inputs
& VERT_BIT_COLOR1
)
367 fp_inputs
|= FRAG_BIT_COL1
;
369 fp_inputs
|= (((varying_inputs
& VERT_BIT_TEX_ANY
) >> VERT_ATTRIB_TEX0
)
370 << FRAG_ATTRIB_TEX0
);
374 /* calculate from vp->outputs */
375 struct gl_program
*vprog
;
376 GLbitfield64 vp_outputs
;
378 /* Choose GLSL vertex shader over ARB vertex program. Need this
379 * since vertex shader state validation comes after fragment state
380 * validation (see additional comments in state.c).
383 vprog
= ctx
->Shader
.CurrentVertexProgram
->_LinkedShaders
[MESA_SHADER_VERTEX
]->Program
;
385 vprog
= &ctx
->VertexProgram
.Current
->Base
;
387 vp_outputs
= vprog
->OutputsWritten
;
389 /* These get generated in the setup routine regardless of the
393 if (ctx
->Point
.PointSprite
)
394 vp_outputs
|= FRAG_BITS_TEX_ANY
;
396 if (vp_outputs
& (1 << VERT_RESULT_COL0
))
397 fp_inputs
|= FRAG_BIT_COL0
;
398 if (vp_outputs
& (1 << VERT_RESULT_COL1
))
399 fp_inputs
|= FRAG_BIT_COL1
;
401 fp_inputs
|= (((vp_outputs
& VERT_RESULT_TEX_ANY
) >> VERT_RESULT_TEX0
)
402 << FRAG_ATTRIB_TEX0
);
410 * Examine current texture environment state and generate a unique
411 * key to identify it.
413 static GLuint
make_state_key( struct gl_context
*ctx
, struct state_key
*key
)
416 GLbitfield inputs_referenced
= FRAG_BIT_COL0
;
417 const GLbitfield inputs_available
= get_fp_input_mask( ctx
);
420 memset(key
, 0, sizeof(*key
));
423 for (i
= 0; i
< ctx
->Const
.MaxTextureUnits
; i
++) {
424 const struct gl_texture_unit
*texUnit
= &ctx
->Texture
.Unit
[i
];
425 const struct gl_texture_object
*texObj
= texUnit
->_Current
;
426 const struct gl_tex_env_combine_state
*comb
= texUnit
->_CurrentCombine
;
429 if (!texUnit
->_ReallyEnabled
|| !texUnit
->Enabled
)
432 format
= texObj
->Image
[0][texObj
->BaseLevel
]->_BaseFormat
;
434 key
->unit
[i
].enabled
= 1;
435 key
->enabled_units
|= (1<<i
);
436 key
->nr_enabled_units
= i
+ 1;
437 inputs_referenced
|= FRAG_BIT_TEX(i
);
439 key
->unit
[i
].source_index
=
440 translate_tex_src_bit(texUnit
->_ReallyEnabled
);
442 key
->unit
[i
].shadow
=
443 ((texObj
->Sampler
.CompareMode
== GL_COMPARE_R_TO_TEXTURE
) &&
444 ((format
== GL_DEPTH_COMPONENT
) ||
445 (format
== GL_DEPTH_STENCIL_EXT
)));
447 key
->unit
[i
].NumArgsRGB
= comb
->_NumArgsRGB
;
448 key
->unit
[i
].NumArgsA
= comb
->_NumArgsA
;
450 key
->unit
[i
].ModeRGB
=
451 translate_mode(texUnit
->EnvMode
, comb
->ModeRGB
);
453 translate_mode(texUnit
->EnvMode
, comb
->ModeA
);
455 key
->unit
[i
].ScaleShiftRGB
= comb
->ScaleShiftRGB
;
456 key
->unit
[i
].ScaleShiftA
= comb
->ScaleShiftA
;
458 for (j
= 0; j
< MAX_COMBINER_TERMS
; j
++) {
459 key
->unit
[i
].OptRGB
[j
].Operand
= translate_operand(comb
->OperandRGB
[j
]);
460 key
->unit
[i
].OptA
[j
].Operand
= translate_operand(comb
->OperandA
[j
]);
461 key
->unit
[i
].OptRGB
[j
].Source
= translate_source(comb
->SourceRGB
[j
]);
462 key
->unit
[i
].OptA
[j
].Source
= translate_source(comb
->SourceA
[j
]);
465 if (key
->unit
[i
].ModeRGB
== MODE_BUMP_ENVMAP_ATI
) {
466 /* requires some special translation */
467 key
->unit
[i
].NumArgsRGB
= 2;
468 key
->unit
[i
].ScaleShiftRGB
= 0;
469 key
->unit
[i
].OptRGB
[0].Operand
= OPR_SRC_COLOR
;
470 key
->unit
[i
].OptRGB
[0].Source
= SRC_TEXTURE
;
471 key
->unit
[i
].OptRGB
[1].Operand
= OPR_SRC_COLOR
;
472 key
->unit
[i
].OptRGB
[1].Source
= texUnit
->BumpTarget
- GL_TEXTURE0
+ SRC_TEXTURE0
;
476 /* _NEW_LIGHT | _NEW_FOG */
477 if (texenv_doing_secondary_color(ctx
)) {
478 key
->separate_specular
= 1;
479 inputs_referenced
|= FRAG_BIT_COL1
;
483 if (ctx
->Fog
.Enabled
) {
484 key
->fog_enabled
= 1;
485 key
->fog_mode
= translate_fog_mode(ctx
->Fog
.Mode
);
486 inputs_referenced
|= FRAG_BIT_FOGC
; /* maybe */
490 key
->num_draw_buffers
= ctx
->DrawBuffer
->_NumColorDrawBuffers
;
492 key
->inputs_available
= (inputs_available
& inputs_referenced
);
494 /* compute size of state key, ignoring unused texture units */
495 keySize
= sizeof(*key
) - sizeof(key
->unit
)
496 + key
->nr_enabled_units
* sizeof(key
->unit
[0]);
502 /** State used to build the fragment program:
504 struct texenv_fragment_program
{
505 struct gl_shader_program
*shader_program
;
506 struct gl_shader
*shader
;
507 struct gl_fragment_program
*program
;
508 exec_list
*instructions
;
509 exec_list
*top_instructions
;
511 struct state_key
*state
;
513 GLbitfield alu_temps
; /**< Track texture indirections, see spec. */
514 GLbitfield temps_output
; /**< Track texture indirections, see spec. */
515 GLbitfield temp_in_use
; /**< Tracks temporary regs which are in use. */
518 ir_variable
*src_texture
[MAX_TEXTURE_COORD_UNITS
];
519 /* Reg containing each texture unit's sampled texture color,
523 /* Texcoord override from bumpmapping. */
524 struct ir_variable
*texcoord_tex
[MAX_TEXTURE_COORD_UNITS
];
526 /* Reg containing texcoord for a texture unit,
527 * needed for bump mapping, else undef.
530 ir_rvalue
*src_previous
; /**< Reg containing color from previous
531 * stage. May need to be decl'd.
534 GLuint last_tex_stage
; /**< Number of last enabled texture unit */
538 get_current_attrib(struct texenv_fragment_program
*p
, GLuint attrib
)
540 ir_variable
*current
;
543 current
= p
->shader
->symbols
->get_variable("gl_CurrentAttribFragMESA");
544 current
->max_array_access
= MAX2(current
->max_array_access
, attrib
);
545 val
= new(p
->mem_ctx
) ir_dereference_variable(current
);
546 ir_rvalue
*index
= new(p
->mem_ctx
) ir_constant(attrib
);
547 return new(p
->mem_ctx
) ir_dereference_array(val
, index
);
551 get_gl_Color(struct texenv_fragment_program
*p
)
553 if (p
->state
->inputs_available
& FRAG_BIT_COL0
) {
554 ir_variable
*var
= p
->shader
->symbols
->get_variable("gl_Color");
556 return new(p
->mem_ctx
) ir_dereference_variable(var
);
558 return get_current_attrib(p
, VERT_ATTRIB_COLOR0
);
563 get_source(struct texenv_fragment_program
*p
,
564 GLuint src
, GLuint unit
)
567 ir_dereference
*deref
;
571 return new(p
->mem_ctx
) ir_dereference_variable(p
->src_texture
[unit
]);
581 return new(p
->mem_ctx
)
582 ir_dereference_variable(p
->src_texture
[src
- SRC_TEXTURE0
]);
585 var
= p
->shader
->symbols
->get_variable("gl_TextureEnvColor");
587 deref
= new(p
->mem_ctx
) ir_dereference_variable(var
);
588 var
->max_array_access
= MAX2(var
->max_array_access
, unit
);
589 return new(p
->mem_ctx
) ir_dereference_array(deref
,
590 new(p
->mem_ctx
) ir_constant(unit
));
592 case SRC_PRIMARY_COLOR
:
593 var
= p
->shader
->symbols
->get_variable("gl_Color");
595 return new(p
->mem_ctx
) ir_dereference_variable(var
);
598 return new(p
->mem_ctx
) ir_constant(0.0f
);
601 if (!p
->src_previous
) {
602 return get_gl_Color(p
);
604 return p
->src_previous
->clone(p
->mem_ctx
, NULL
);
614 emit_combine_source(struct texenv_fragment_program
*p
,
621 src
= get_source(p
, source
, unit
);
624 case OPR_ONE_MINUS_SRC_COLOR
:
625 return new(p
->mem_ctx
) ir_expression(ir_binop_sub
,
626 new(p
->mem_ctx
) ir_constant(1.0f
),
630 return new(p
->mem_ctx
) ir_swizzle(src
, 3, 3, 3, 3, 1);
632 case OPR_ONE_MINUS_SRC_ALPHA
:
633 return new(p
->mem_ctx
) ir_expression(ir_binop_sub
,
634 new(p
->mem_ctx
) ir_constant(1.0f
),
635 new(p
->mem_ctx
) ir_swizzle(src
,
639 return new(p
->mem_ctx
) ir_constant(0.0f
);
641 return new(p
->mem_ctx
) ir_constant(1.0f
);
651 * Check if the RGB and Alpha sources and operands match for the given
652 * texture unit's combinder state. When the RGB and A sources and
653 * operands match, we can emit fewer instructions.
655 static GLboolean
args_match( const struct state_key
*key
, GLuint unit
)
657 GLuint i
, numArgs
= key
->unit
[unit
].NumArgsRGB
;
659 for (i
= 0; i
< numArgs
; i
++) {
660 if (key
->unit
[unit
].OptA
[i
].Source
!= key
->unit
[unit
].OptRGB
[i
].Source
)
663 switch (key
->unit
[unit
].OptA
[i
].Operand
) {
665 switch (key
->unit
[unit
].OptRGB
[i
].Operand
) {
673 case OPR_ONE_MINUS_SRC_ALPHA
:
674 switch (key
->unit
[unit
].OptRGB
[i
].Operand
) {
675 case OPR_ONE_MINUS_SRC_COLOR
:
676 case OPR_ONE_MINUS_SRC_ALPHA
:
683 return GL_FALSE
; /* impossible */
691 smear(struct texenv_fragment_program
*p
, ir_rvalue
*val
)
693 if (!val
->type
->is_scalar())
696 return new(p
->mem_ctx
) ir_swizzle(val
, 0, 0, 0, 0, 4);
700 emit_combine(struct texenv_fragment_program
*p
,
704 const struct mode_opt
*opt
)
706 ir_rvalue
*src
[MAX_COMBINER_TERMS
];
707 ir_rvalue
*tmp0
, *tmp1
;
710 assert(nr
<= MAX_COMBINER_TERMS
);
712 for (i
= 0; i
< nr
; i
++)
713 src
[i
] = emit_combine_source( p
, unit
, opt
[i
].Source
, opt
[i
].Operand
);
720 return new(p
->mem_ctx
) ir_expression(ir_binop_mul
, src
[0], src
[1]);
723 return new(p
->mem_ctx
) ir_expression(ir_binop_add
, src
[0], src
[1]);
725 case MODE_ADD_SIGNED
:
726 tmp0
= new(p
->mem_ctx
) ir_expression(ir_binop_add
, src
[0], src
[1]);
727 return new(p
->mem_ctx
) ir_expression(ir_binop_add
, tmp0
,
728 new(p
->mem_ctx
) ir_constant(-0.5f
));
730 case MODE_INTERPOLATE
:
731 /* Arg0 * (Arg2) + Arg1 * (1-Arg2) */
732 tmp0
= new(p
->mem_ctx
) ir_expression(ir_binop_mul
, src
[0], src
[2]);
734 tmp1
= new(p
->mem_ctx
) ir_expression(ir_binop_sub
,
735 new(p
->mem_ctx
) ir_constant(1.0f
),
736 src
[2]->clone(p
->mem_ctx
, NULL
));
737 tmp1
= new(p
->mem_ctx
) ir_expression(ir_binop_mul
, src
[1], tmp1
);
739 return new(p
->mem_ctx
) ir_expression(ir_binop_add
, tmp0
, tmp1
);
742 return new(p
->mem_ctx
) ir_expression(ir_binop_sub
, src
[0], src
[1]);
745 case MODE_DOT3_RGBA_EXT
:
746 case MODE_DOT3_RGB_EXT
:
747 case MODE_DOT3_RGB
: {
748 tmp0
= new(p
->mem_ctx
) ir_expression(ir_binop_mul
, src
[0],
749 new(p
->mem_ctx
) ir_constant(2.0f
));
750 tmp0
= new(p
->mem_ctx
) ir_expression(ir_binop_add
, tmp0
,
751 new(p
->mem_ctx
) ir_constant(-1.0f
));
752 tmp0
= new(p
->mem_ctx
) ir_swizzle(smear(p
, tmp0
), 0, 1, 2, 3, 3);
754 tmp1
= new(p
->mem_ctx
) ir_expression(ir_binop_mul
, src
[1],
755 new(p
->mem_ctx
) ir_constant(2.0f
));
756 tmp1
= new(p
->mem_ctx
) ir_expression(ir_binop_add
, tmp1
,
757 new(p
->mem_ctx
) ir_constant(-1.0f
));
758 tmp1
= new(p
->mem_ctx
) ir_swizzle(smear(p
, tmp1
), 0, 1, 2, 3, 3);
760 return new(p
->mem_ctx
) ir_expression(ir_binop_dot
, tmp0
, tmp1
);
762 case MODE_MODULATE_ADD_ATI
:
763 tmp0
= new(p
->mem_ctx
) ir_expression(ir_binop_mul
, src
[0], src
[2]);
764 return new(p
->mem_ctx
) ir_expression(ir_binop_add
, tmp0
, src
[1]);
766 case MODE_MODULATE_SIGNED_ADD_ATI
:
767 tmp0
= new(p
->mem_ctx
) ir_expression(ir_binop_mul
, src
[0], src
[2]);
768 tmp0
= new(p
->mem_ctx
) ir_expression(ir_binop_add
, tmp0
, src
[1]);
769 return new(p
->mem_ctx
) ir_expression(ir_binop_add
, tmp0
,
770 new(p
->mem_ctx
) ir_constant(-0.5f
));
772 case MODE_MODULATE_SUBTRACT_ATI
:
773 tmp0
= new(p
->mem_ctx
) ir_expression(ir_binop_mul
, src
[0], src
[2]);
774 return new(p
->mem_ctx
) ir_expression(ir_binop_sub
, tmp0
, src
[1]);
776 case MODE_ADD_PRODUCTS
:
777 tmp0
= new(p
->mem_ctx
) ir_expression(ir_binop_mul
, src
[0], src
[1]);
778 tmp1
= new(p
->mem_ctx
) ir_expression(ir_binop_mul
, src
[2], src
[3]);
779 return new(p
->mem_ctx
) ir_expression(ir_binop_add
, tmp0
, tmp1
);
781 case MODE_ADD_PRODUCTS_SIGNED
:
782 tmp0
= new(p
->mem_ctx
) ir_expression(ir_binop_mul
, src
[0], src
[1]);
783 tmp1
= new(p
->mem_ctx
) ir_expression(ir_binop_mul
, src
[2], src
[3]);
784 tmp0
= new(p
->mem_ctx
) ir_expression(ir_binop_add
, tmp0
, tmp1
);
785 return new(p
->mem_ctx
) ir_expression(ir_binop_add
, tmp0
,
786 new(p
->mem_ctx
) ir_constant(-0.5f
));
788 case MODE_BUMP_ENVMAP_ATI
:
789 /* special - not handled here */
799 saturate(struct texenv_fragment_program
*p
, ir_rvalue
*val
)
801 val
= new(p
->mem_ctx
) ir_expression(ir_binop_min
, val
,
802 new(p
->mem_ctx
) ir_constant(1.0f
));
803 return new(p
->mem_ctx
) ir_expression(ir_binop_max
, val
,
804 new(p
->mem_ctx
) ir_constant(0.0f
));
808 * Generate instructions for one texture unit's env/combiner mode.
811 emit_texenv(struct texenv_fragment_program
*p
, GLuint unit
)
813 const struct state_key
*key
= p
->state
;
814 GLboolean rgb_saturate
, alpha_saturate
;
815 GLuint rgb_shift
, alpha_shift
;
817 if (!key
->unit
[unit
].enabled
) {
818 return get_source(p
, SRC_PREVIOUS
, 0);
820 if (key
->unit
[unit
].ModeRGB
== MODE_BUMP_ENVMAP_ATI
) {
821 /* this isn't really a env stage delivering a color and handled elsewhere */
822 return get_source(p
, SRC_PREVIOUS
, 0);
825 switch (key
->unit
[unit
].ModeRGB
) {
826 case MODE_DOT3_RGB_EXT
:
827 alpha_shift
= key
->unit
[unit
].ScaleShiftA
;
830 case MODE_DOT3_RGBA_EXT
:
835 rgb_shift
= key
->unit
[unit
].ScaleShiftRGB
;
836 alpha_shift
= key
->unit
[unit
].ScaleShiftA
;
840 /* If we'll do rgb/alpha shifting don't saturate in emit_combine().
841 * We don't want to clamp twice.
844 rgb_saturate
= GL_FALSE
; /* saturate after rgb shift */
845 else if (need_saturate(key
->unit
[unit
].ModeRGB
))
846 rgb_saturate
= GL_TRUE
;
848 rgb_saturate
= GL_FALSE
;
851 alpha_saturate
= GL_FALSE
; /* saturate after alpha shift */
852 else if (need_saturate(key
->unit
[unit
].ModeA
))
853 alpha_saturate
= GL_TRUE
;
855 alpha_saturate
= GL_FALSE
;
857 ir_variable
*temp_var
= new(p
->mem_ctx
) ir_variable(glsl_type::vec4_type
,
860 p
->instructions
->push_tail(temp_var
);
862 ir_dereference
*deref
;
863 ir_assignment
*assign
;
866 /* Emit the RGB and A combine ops
868 if (key
->unit
[unit
].ModeRGB
== key
->unit
[unit
].ModeA
&&
869 args_match(key
, unit
)) {
870 val
= emit_combine(p
, unit
,
871 key
->unit
[unit
].NumArgsRGB
,
872 key
->unit
[unit
].ModeRGB
,
873 key
->unit
[unit
].OptRGB
);
876 val
= saturate(p
, val
);
878 deref
= new(p
->mem_ctx
) ir_dereference_variable(temp_var
);
879 assign
= new(p
->mem_ctx
) ir_assignment(deref
, val
);
880 p
->instructions
->push_tail(assign
);
882 else if (key
->unit
[unit
].ModeRGB
== MODE_DOT3_RGBA_EXT
||
883 key
->unit
[unit
].ModeRGB
== MODE_DOT3_RGBA
) {
884 ir_rvalue
*val
= emit_combine(p
, unit
,
885 key
->unit
[unit
].NumArgsRGB
,
886 key
->unit
[unit
].ModeRGB
,
887 key
->unit
[unit
].OptRGB
);
890 val
= saturate(p
, val
);
891 deref
= new(p
->mem_ctx
) ir_dereference_variable(temp_var
);
892 assign
= new(p
->mem_ctx
) ir_assignment(deref
, val
);
893 p
->instructions
->push_tail(assign
);
896 /* Need to do something to stop from re-emitting identical
897 * argument calculations here:
899 val
= emit_combine(p
, unit
,
900 key
->unit
[unit
].NumArgsRGB
,
901 key
->unit
[unit
].ModeRGB
,
902 key
->unit
[unit
].OptRGB
);
904 val
= new(p
->mem_ctx
) ir_swizzle(val
, 0, 1, 2, 3, 3);
906 val
= saturate(p
, val
);
907 deref
= new(p
->mem_ctx
) ir_dereference_variable(temp_var
);
908 assign
= new(p
->mem_ctx
) ir_assignment(deref
, val
, NULL
, WRITEMASK_XYZ
);
909 p
->instructions
->push_tail(assign
);
911 val
= emit_combine(p
, unit
,
912 key
->unit
[unit
].NumArgsA
,
913 key
->unit
[unit
].ModeA
,
914 key
->unit
[unit
].OptA
);
916 val
= new(p
->mem_ctx
) ir_swizzle(val
, 3, 3, 3, 3, 1);
918 val
= saturate(p
, val
);
919 deref
= new(p
->mem_ctx
) ir_dereference_variable(temp_var
);
920 assign
= new(p
->mem_ctx
) ir_assignment(deref
, val
, NULL
, WRITEMASK_W
);
921 p
->instructions
->push_tail(assign
);
924 deref
= new(p
->mem_ctx
) ir_dereference_variable(temp_var
);
926 /* Deal with the final shift:
928 if (alpha_shift
|| rgb_shift
) {
931 if (rgb_shift
== alpha_shift
) {
932 shift
= new(p
->mem_ctx
) ir_constant((float)(1 << rgb_shift
));
935 float const_data
[4] = {
941 shift
= new(p
->mem_ctx
) ir_constant(glsl_type::vec4_type
,
942 (ir_constant_data
*)const_data
);
945 return saturate(p
, new(p
->mem_ctx
) ir_expression(ir_binop_mul
,
954 * Generate instruction for getting a texture source term.
956 static void load_texture( struct texenv_fragment_program
*p
, GLuint unit
)
958 ir_dereference
*deref
;
959 ir_assignment
*assign
;
961 if (p
->src_texture
[unit
])
964 const GLuint texTarget
= p
->state
->unit
[unit
].source_index
;
967 if (!(p
->state
->inputs_available
& (FRAG_BIT_TEX0
<< unit
))) {
968 texcoord
= get_current_attrib(p
, VERT_ATTRIB_TEX0
+ unit
);
969 } else if (p
->texcoord_tex
[unit
]) {
970 texcoord
= new(p
->mem_ctx
) ir_dereference_variable(p
->texcoord_tex
[unit
]);
972 ir_variable
*tc_array
= p
->shader
->symbols
->get_variable("gl_TexCoord");
974 texcoord
= new(p
->mem_ctx
) ir_dereference_variable(tc_array
);
975 ir_rvalue
*index
= new(p
->mem_ctx
) ir_constant(unit
);
976 texcoord
= new(p
->mem_ctx
) ir_dereference_array(texcoord
, index
);
977 tc_array
->max_array_access
= MAX2(tc_array
->max_array_access
, unit
);
980 if (!p
->state
->unit
[unit
].enabled
) {
981 p
->src_texture
[unit
] = new(p
->mem_ctx
) ir_variable(glsl_type::vec4_type
,
984 p
->instructions
->push_tail(p
->src_texture
[unit
]);
986 deref
= new(p
->mem_ctx
) ir_dereference_variable(p
->src_texture
[unit
]);
987 assign
= new(p
->mem_ctx
) ir_assignment(deref
,
988 new(p
->mem_ctx
) ir_constant(0.0f
));
989 p
->instructions
->push_tail(assign
);
993 const glsl_type
*sampler_type
= NULL
;
997 case TEXTURE_1D_INDEX
:
998 if (p
->state
->unit
[unit
].shadow
)
999 sampler_type
= p
->shader
->symbols
->get_type("sampler1DShadow");
1001 sampler_type
= p
->shader
->symbols
->get_type("sampler1D");
1004 case TEXTURE_1D_ARRAY_INDEX
:
1005 if (p
->state
->unit
[unit
].shadow
)
1006 sampler_type
= p
->shader
->symbols
->get_type("sampler1DArrayShadow");
1008 sampler_type
= p
->shader
->symbols
->get_type("sampler1DArray");
1011 case TEXTURE_2D_INDEX
:
1012 if (p
->state
->unit
[unit
].shadow
)
1013 sampler_type
= p
->shader
->symbols
->get_type("sampler2DShadow");
1015 sampler_type
= p
->shader
->symbols
->get_type("sampler2D");
1018 case TEXTURE_2D_ARRAY_INDEX
:
1019 if (p
->state
->unit
[unit
].shadow
)
1020 sampler_type
= p
->shader
->symbols
->get_type("sampler2DArrayShadow");
1022 sampler_type
= p
->shader
->symbols
->get_type("sampler2DArray");
1025 case TEXTURE_RECT_INDEX
:
1026 if (p
->state
->unit
[unit
].shadow
)
1027 sampler_type
= p
->shader
->symbols
->get_type("sampler2DRectShadow");
1029 sampler_type
= p
->shader
->symbols
->get_type("sampler2DRect");
1032 case TEXTURE_3D_INDEX
:
1033 assert(!p
->state
->unit
[unit
].shadow
);
1034 sampler_type
= p
->shader
->symbols
->get_type("sampler3D");
1037 case TEXTURE_CUBE_INDEX
:
1038 if (p
->state
->unit
[unit
].shadow
)
1039 sampler_type
= p
->shader
->symbols
->get_type("samplerCubeShadow");
1041 sampler_type
= p
->shader
->symbols
->get_type("samplerCube");
1046 p
->src_texture
[unit
] = new(p
->mem_ctx
) ir_variable(glsl_type::vec4_type
,
1049 p
->instructions
->push_tail(p
->src_texture
[unit
]);
1051 ir_texture
*tex
= new(p
->mem_ctx
) ir_texture(ir_tex
);
1054 char *sampler_name
= ralloc_asprintf(p
->mem_ctx
, "sampler_%d", unit
);
1055 ir_variable
*sampler
= new(p
->mem_ctx
) ir_variable(sampler_type
,
1058 p
->top_instructions
->push_head(sampler
);
1059 deref
= new(p
->mem_ctx
) ir_dereference_variable(sampler
);
1060 tex
->set_sampler(deref
, glsl_type::vec4_type
);
1062 tex
->coordinate
= new(p
->mem_ctx
) ir_swizzle(texcoord
, 0, 1, 2, 3, coords
);
1064 if (p
->state
->unit
[unit
].shadow
) {
1065 texcoord
= texcoord
->clone(p
->mem_ctx
, NULL
);
1066 tex
->shadow_comparitor
= new(p
->mem_ctx
) ir_swizzle(texcoord
,
1072 texcoord
= texcoord
->clone(p
->mem_ctx
, NULL
);
1073 tex
->projector
= new(p
->mem_ctx
) ir_swizzle(texcoord
, 3, 0, 0, 0, 1);
1075 deref
= new(p
->mem_ctx
) ir_dereference_variable(p
->src_texture
[unit
]);
1076 assign
= new(p
->mem_ctx
) ir_assignment(deref
, tex
);
1077 p
->instructions
->push_tail(assign
);
1081 load_texenv_source(struct texenv_fragment_program
*p
,
1082 GLuint src
, GLuint unit
)
1086 load_texture(p
, unit
);
1097 load_texture(p
, src
- SRC_TEXTURE0
);
1101 /* not a texture src - do nothing */
1108 * Generate instructions for loading all texture source terms.
1111 load_texunit_sources( struct texenv_fragment_program
*p
, GLuint unit
)
1113 const struct state_key
*key
= p
->state
;
1116 for (i
= 0; i
< key
->unit
[unit
].NumArgsRGB
; i
++) {
1117 load_texenv_source( p
, key
->unit
[unit
].OptRGB
[i
].Source
, unit
);
1120 for (i
= 0; i
< key
->unit
[unit
].NumArgsA
; i
++) {
1121 load_texenv_source( p
, key
->unit
[unit
].OptA
[i
].Source
, unit
);
1128 * Generate instructions for loading bump map textures.
1131 load_texunit_bumpmap( struct texenv_fragment_program
*p
, GLuint unit
)
1133 const struct state_key
*key
= p
->state
;
1134 GLuint bumpedUnitNr
= key
->unit
[unit
].OptRGB
[1].Source
- SRC_TEXTURE0
;
1136 ir_rvalue
*texcoord
;
1137 ir_variable
*rot_mat_0_var
, *rot_mat_1_var
;
1138 ir_dereference_variable
*rot_mat_0
, *rot_mat_1
;
1140 rot_mat_0_var
= p
->shader
->symbols
->get_variable("gl_MESABumpRotMatrix0");
1141 rot_mat_1_var
= p
->shader
->symbols
->get_variable("gl_MESABumpRotMatrix1");
1142 rot_mat_0
= new(p
->mem_ctx
) ir_dereference_variable(rot_mat_0_var
);
1143 rot_mat_1
= new(p
->mem_ctx
) ir_dereference_variable(rot_mat_1_var
);
1145 ir_variable
*tc_array
= p
->shader
->symbols
->get_variable("gl_TexCoord");
1147 texcoord
= new(p
->mem_ctx
) ir_dereference_variable(tc_array
);
1148 ir_rvalue
*index
= new(p
->mem_ctx
) ir_constant(bumpedUnitNr
);
1149 texcoord
= new(p
->mem_ctx
) ir_dereference_array(texcoord
, index
);
1150 tc_array
->max_array_access
= MAX2(tc_array
->max_array_access
, unit
);
1152 load_texenv_source( p
, unit
+ SRC_TEXTURE0
, unit
);
1154 /* Apply rot matrix and add coords to be available in next phase.
1155 * dest = Arg1 + (Arg0.xx * rotMat0) + (Arg0.yy * rotMat1)
1156 * note only 2 coords are affected the rest are left unchanged (mul by 0)
1158 ir_dereference
*deref
;
1159 ir_assignment
*assign
;
1160 ir_rvalue
*bump_x
, *bump_y
;
1162 texcoord
= smear(p
, texcoord
);
1164 /* bump_texcoord = texcoord */
1165 ir_variable
*bumped
= new(p
->mem_ctx
) ir_variable(texcoord
->type
,
1168 p
->instructions
->push_tail(bumped
);
1170 deref
= new(p
->mem_ctx
) ir_dereference_variable(bumped
);
1171 assign
= new(p
->mem_ctx
) ir_assignment(deref
, texcoord
);
1172 p
->instructions
->push_tail(assign
);
1174 /* bump_texcoord.xy += arg0.x * rotmat0 + arg0.y * rotmat1 */
1175 bump
= get_source(p
, key
->unit
[unit
].OptRGB
[0].Source
, unit
);
1176 bump_x
= new(p
->mem_ctx
) ir_swizzle(bump
, 0, 0, 0, 0, 1);
1177 bump
= bump
->clone(p
->mem_ctx
, NULL
);
1178 bump_y
= new(p
->mem_ctx
) ir_swizzle(bump
, 1, 0, 0, 0, 1);
1180 bump_x
= new(p
->mem_ctx
) ir_expression(ir_binop_mul
, bump_x
, rot_mat_0
);
1181 bump_y
= new(p
->mem_ctx
) ir_expression(ir_binop_mul
, bump_y
, rot_mat_1
);
1183 ir_expression
*expr
;
1184 expr
= new(p
->mem_ctx
) ir_expression(ir_binop_add
, bump_x
, bump_y
);
1186 deref
= new(p
->mem_ctx
) ir_dereference_variable(bumped
);
1187 expr
= new(p
->mem_ctx
) ir_expression(ir_binop_add
,
1188 new(p
->mem_ctx
) ir_swizzle(deref
,
1193 deref
= new(p
->mem_ctx
) ir_dereference_variable(bumped
);
1194 assign
= new(p
->mem_ctx
) ir_assignment(deref
, expr
, NULL
, WRITEMASK_XY
);
1195 p
->instructions
->push_tail(assign
);
1197 p
->texcoord_tex
[bumpedUnitNr
] = bumped
;
1201 * Applies the fog calculations.
1203 * This is basically like the ARB_fragment_prorgam fog options. Note
1204 * that ffvertex_prog.c produces fogcoord for us when
1205 * GL_FOG_COORDINATE_EXT is set to GL_FRAGMENT_DEPTH_EXT.
1208 emit_fog_instructions(struct texenv_fragment_program
*p
,
1209 ir_rvalue
*fragcolor
)
1211 struct state_key
*key
= p
->state
;
1212 ir_rvalue
*f
, *temp
;
1213 ir_variable
*params
, *oparams
;
1214 ir_variable
*fogcoord
;
1215 ir_assignment
*assign
;
1217 /* Temporary storage for the whole fog result. Fog calculations
1218 * only affect rgb so we're hanging on to the .a value of fragcolor
1221 ir_variable
*fog_result
= new(p
->mem_ctx
) ir_variable(glsl_type::vec4_type
,
1224 p
->instructions
->push_tail(fog_result
);
1225 temp
= new(p
->mem_ctx
) ir_dereference_variable(fog_result
);
1226 assign
= new(p
->mem_ctx
) ir_assignment(temp
, fragcolor
);
1227 p
->instructions
->push_tail(assign
);
1229 temp
= new(p
->mem_ctx
) ir_dereference_variable(fog_result
);
1230 fragcolor
= new(p
->mem_ctx
) ir_swizzle(temp
, 0, 1, 2, 3, 3);
1232 oparams
= p
->shader
->symbols
->get_variable("gl_MESAFogParamsOptimized");
1233 fogcoord
= p
->shader
->symbols
->get_variable("gl_FogFragCoord");
1234 params
= p
->shader
->symbols
->get_variable("gl_Fog");
1235 f
= new(p
->mem_ctx
) ir_dereference_variable(fogcoord
);
1237 ir_variable
*f_var
= new(p
->mem_ctx
) ir_variable(glsl_type::float_type
,
1238 "fog_factor", ir_var_auto
);
1239 p
->instructions
->push_tail(f_var
);
1241 switch (key
->fog_mode
) {
1243 /* f = (end - z) / (end - start)
1245 * gl_MesaFogParamsOptimized gives us (-1 / (end - start)) and
1246 * (end / (end - start)) so we can generate a single MAD.
1248 temp
= new(p
->mem_ctx
) ir_dereference_variable(oparams
);
1249 temp
= new(p
->mem_ctx
) ir_swizzle(temp
, 0, 0, 0, 0, 1);
1250 f
= new(p
->mem_ctx
) ir_expression(ir_binop_mul
, f
, temp
);
1252 temp
= new(p
->mem_ctx
) ir_dereference_variable(oparams
);
1253 temp
= new(p
->mem_ctx
) ir_swizzle(temp
, 1, 0, 0, 0, 1);
1254 f
= new(p
->mem_ctx
) ir_expression(ir_binop_add
, f
, temp
);
1257 /* f = e^(-(density * fogcoord))
1259 * gl_MesaFogParamsOptimized gives us density/ln(2) so we can
1260 * use EXP2 which is generally the native instruction without
1261 * having to do any further math on the fog density uniform.
1263 temp
= new(p
->mem_ctx
) ir_dereference_variable(oparams
);
1264 temp
= new(p
->mem_ctx
) ir_swizzle(temp
, 2, 0, 0, 0, 1);
1265 f
= new(p
->mem_ctx
) ir_expression(ir_binop_mul
, f
, temp
);
1266 f
= new(p
->mem_ctx
) ir_expression(ir_unop_neg
, f
);
1267 f
= new(p
->mem_ctx
) ir_expression(ir_unop_exp2
, f
);
1270 /* f = e^(-(density * fogcoord)^2)
1272 * gl_MesaFogParamsOptimized gives us density/sqrt(ln(2)) so we
1273 * can do this like FOG_EXP but with a squaring after the
1274 * multiply by density.
1276 ir_variable
*temp_var
= new(p
->mem_ctx
) ir_variable(glsl_type::float_type
,
1279 p
->instructions
->push_tail(temp_var
);
1281 temp
= new(p
->mem_ctx
) ir_dereference_variable(oparams
);
1282 temp
= new(p
->mem_ctx
) ir_swizzle(temp
, 3, 0, 0, 0, 1);
1283 f
= new(p
->mem_ctx
) ir_expression(ir_binop_mul
,
1286 temp
= new(p
->mem_ctx
) ir_dereference_variable(temp_var
);
1287 ir_assignment
*assign
= new(p
->mem_ctx
) ir_assignment(temp
, f
);
1288 p
->instructions
->push_tail(assign
);
1290 f
= new(p
->mem_ctx
) ir_dereference_variable(temp_var
);
1291 temp
= new(p
->mem_ctx
) ir_dereference_variable(temp_var
);
1292 f
= new(p
->mem_ctx
) ir_expression(ir_binop_mul
, f
, temp
);
1293 f
= new(p
->mem_ctx
) ir_expression(ir_unop_neg
, f
);
1294 f
= new(p
->mem_ctx
) ir_expression(ir_unop_exp2
, f
);
1300 temp
= new(p
->mem_ctx
) ir_dereference_variable(f_var
);
1301 assign
= new(p
->mem_ctx
) ir_assignment(temp
, f
);
1302 p
->instructions
->push_tail(assign
);
1304 f
= new(p
->mem_ctx
) ir_dereference_variable(f_var
);
1305 f
= new(p
->mem_ctx
) ir_expression(ir_binop_sub
,
1306 new(p
->mem_ctx
) ir_constant(1.0f
),
1308 temp
= new(p
->mem_ctx
) ir_dereference_variable(params
);
1309 temp
= new(p
->mem_ctx
) ir_dereference_record(temp
, "color");
1310 temp
= new(p
->mem_ctx
) ir_swizzle(temp
, 0, 1, 2, 3, 3);
1311 temp
= new(p
->mem_ctx
) ir_expression(ir_binop_mul
, temp
, f
);
1313 f
= new(p
->mem_ctx
) ir_dereference_variable(f_var
);
1314 f
= new(p
->mem_ctx
) ir_expression(ir_binop_mul
, fragcolor
, f
);
1315 f
= new(p
->mem_ctx
) ir_expression(ir_binop_add
, temp
, f
);
1317 ir_dereference
*deref
= new(p
->mem_ctx
) ir_dereference_variable(fog_result
);
1318 assign
= new(p
->mem_ctx
) ir_assignment(deref
, f
, NULL
, WRITEMASK_XYZ
);
1319 p
->instructions
->push_tail(assign
);
1321 return new(p
->mem_ctx
) ir_dereference_variable(fog_result
);
1325 emit_instructions(struct texenv_fragment_program
*p
)
1327 struct state_key
*key
= p
->state
;
1330 if (key
->enabled_units
) {
1331 /* Zeroth pass - bump map textures first */
1332 for (unit
= 0; unit
< key
->nr_enabled_units
; unit
++) {
1333 if (key
->unit
[unit
].enabled
&&
1334 key
->unit
[unit
].ModeRGB
== MODE_BUMP_ENVMAP_ATI
) {
1335 load_texunit_bumpmap(p
, unit
);
1339 /* First pass - to support texture_env_crossbar, first identify
1340 * all referenced texture sources and emit texld instructions
1343 for (unit
= 0; unit
< key
->nr_enabled_units
; unit
++)
1344 if (key
->unit
[unit
].enabled
) {
1345 load_texunit_sources(p
, unit
);
1346 p
->last_tex_stage
= unit
;
1349 /* Second pass - emit combine instructions to build final color:
1351 for (unit
= 0; unit
< key
->nr_enabled_units
; unit
++) {
1352 if (key
->unit
[unit
].enabled
) {
1353 p
->src_previous
= emit_texenv(p
, unit
);
1358 ir_rvalue
*cf
= get_source(p
, SRC_PREVIOUS
, 0);
1359 ir_dereference_variable
*deref
;
1360 ir_assignment
*assign
;
1362 if (key
->separate_specular
) {
1364 ir_variable
*spec_result
= new(p
->mem_ctx
) ir_variable(glsl_type::vec4_type
,
1368 p
->instructions
->push_tail(spec_result
);
1370 deref
= new(p
->mem_ctx
) ir_dereference_variable(spec_result
);
1371 assign
= new(p
->mem_ctx
) ir_assignment(deref
, cf
);
1372 p
->instructions
->push_tail(assign
);
1374 deref
= new(p
->mem_ctx
) ir_dereference_variable(spec_result
);
1375 tmp0
= new(p
->mem_ctx
) ir_swizzle(deref
, 0, 1, 2, 3, 3);
1377 ir_rvalue
*secondary
;
1378 if (p
->state
->inputs_available
& FRAG_BIT_COL1
) {
1380 p
->shader
->symbols
->get_variable("gl_SecondaryColor");
1382 secondary
= new(p
->mem_ctx
) ir_dereference_variable(var
);
1384 secondary
= get_current_attrib(p
, VERT_ATTRIB_COLOR1
);
1386 secondary
= new(p
->mem_ctx
) ir_swizzle(secondary
, 0, 1, 2, 3, 3);
1388 tmp0
= new(p
->mem_ctx
) ir_expression(ir_binop_add
, tmp0
, secondary
);
1390 deref
= new(p
->mem_ctx
) ir_dereference_variable(spec_result
);
1391 assign
= new(p
->mem_ctx
) ir_assignment(deref
, tmp0
, NULL
, WRITEMASK_XYZ
);
1392 p
->instructions
->push_tail(assign
);
1394 cf
= new(p
->mem_ctx
) ir_dereference_variable(spec_result
);
1397 if (key
->fog_enabled
) {
1398 cf
= emit_fog_instructions(p
, cf
);
1401 ir_variable
*frag_color
= p
->shader
->symbols
->get_variable("gl_FragColor");
1403 deref
= new(p
->mem_ctx
) ir_dereference_variable(frag_color
);
1404 assign
= new(p
->mem_ctx
) ir_assignment(deref
, cf
);
1405 p
->instructions
->push_tail(assign
);
1409 * Generate a new fragment program which implements the context's
1410 * current texture env/combine mode.
1412 static struct gl_shader_program
*
1413 create_new_program(struct gl_context
*ctx
, struct state_key
*key
)
1415 struct texenv_fragment_program p
;
1417 _mesa_glsl_parse_state
*state
;
1419 memset(&p
, 0, sizeof(p
));
1420 p
.mem_ctx
= ralloc_context(NULL
);
1421 p
.shader
= ctx
->Driver
.NewShader(ctx
, 0, GL_FRAGMENT_SHADER
);
1422 p
.shader
->ir
= new(p
.shader
) exec_list
;
1423 state
= new(p
.shader
) _mesa_glsl_parse_state(ctx
, GL_FRAGMENT_SHADER
,
1425 p
.shader
->symbols
= state
->symbols
;
1426 p
.top_instructions
= p
.shader
->ir
;
1427 p
.instructions
= p
.shader
->ir
;
1429 p
.shader_program
= ctx
->Driver
.NewShaderProgram(ctx
, 0);
1431 /* Tell the linker to ignore the fact that we're building a
1432 * separate shader, in case we're in a GLES2 context that would
1433 * normally reject that. The real problem is that we're building a
1434 * fixed function program in a GLES2 context at all, but that's a
1435 * big mess to clean up.
1437 p
.shader_program
->InternalSeparateShader
= GL_TRUE
;
1439 state
->language_version
= 130;
1440 _mesa_glsl_initialize_types(state
);
1441 _mesa_glsl_initialize_variables(p
.instructions
, state
);
1443 for (unit
= 0; unit
< ctx
->Const
.MaxTextureUnits
; unit
++) {
1444 p
.src_texture
[unit
] = NULL
;
1445 p
.texcoord_tex
[unit
] = NULL
;
1448 p
.src_previous
= NULL
;
1450 p
.last_tex_stage
= 0;
1452 ir_function
*main_f
= new(p
.mem_ctx
) ir_function("main");
1453 p
.instructions
->push_tail(main_f
);
1454 state
->symbols
->add_function(main_f
);
1456 ir_function_signature
*main_sig
=
1457 new(p
.mem_ctx
) ir_function_signature(p
.shader
->symbols
->get_type("void"));
1458 main_sig
->is_defined
= true;
1459 main_f
->add_signature(main_sig
);
1461 p
.instructions
= &main_sig
->body
;
1462 if (key
->num_draw_buffers
)
1463 emit_instructions(&p
);
1465 validate_ir_tree(p
.shader
->ir
);
1467 while (do_common_optimization(p
.shader
->ir
, false, false, 32))
1469 reparent_ir(p
.shader
->ir
, p
.shader
->ir
);
1471 p
.shader
->CompileStatus
= true;
1472 p
.shader
->Version
= state
->language_version
;
1473 p
.shader
->num_builtins_to_link
= state
->num_builtins_to_link
;
1474 p
.shader_program
->Shaders
=
1475 (gl_shader
**)malloc(sizeof(*p
.shader_program
->Shaders
));
1476 p
.shader_program
->Shaders
[0] = p
.shader
;
1477 p
.shader_program
->NumShaders
= 1;
1479 _mesa_glsl_link_shader(ctx
, p
.shader_program
);
1481 /* Set the sampler uniforms, and relink to get them into the linked
1484 struct gl_program
*fp
;
1485 fp
= p
.shader_program
->_LinkedShaders
[MESA_SHADER_FRAGMENT
]->Program
;
1487 for (unsigned int i
= 0; i
< MAX_TEXTURE_UNITS
; i
++) {
1488 char *name
= ralloc_asprintf(p
.mem_ctx
, "sampler_%d", i
);
1489 int loc
= _mesa_get_uniform_location(ctx
, p
.shader_program
, name
);
1491 /* Avoid using _mesa_uniform() because it flags state
1492 * updates, so if we're generating this shader_program in a
1493 * state update, we end up recursing. Instead, just set the
1494 * value, which is picked up at re-link.
1496 loc
= (loc
& 0xffff) + (loc
>> 16);
1497 int sampler
= fp
->Parameters
->ParameterValues
[loc
][0].f
;
1499 fp
->SamplerUnits
[sampler
] = i
;
1502 _mesa_update_shader_textures_used(fp
);
1503 (void) ctx
->Driver
.ProgramStringNotify(ctx
, fp
->Target
, fp
);
1505 if (!p
.shader_program
->LinkStatus
)
1506 _mesa_problem(ctx
, "Failed to link fixed function fragment shader: %s\n",
1507 p
.shader_program
->InfoLog
);
1509 ralloc_free(p
.mem_ctx
);
1510 return p
.shader_program
;
1516 * Return a fragment program which implements the current
1517 * fixed-function texture, fog and color-sum operations.
1519 struct gl_shader_program
*
1520 _mesa_get_fixed_func_fragment_program(struct gl_context
*ctx
)
1522 struct gl_shader_program
*shader_program
;
1523 struct state_key key
;
1526 keySize
= make_state_key(ctx
, &key
);
1528 shader_program
= (struct gl_shader_program
*)
1529 _mesa_search_program_cache(ctx
->FragmentProgram
.Cache
,
1532 if (!shader_program
) {
1533 shader_program
= create_new_program(ctx
, &key
);
1535 _mesa_shader_cache_insert(ctx
, ctx
->FragmentProgram
.Cache
,
1536 &key
, keySize
, shader_program
);
1539 return shader_program
;