1 /**************************************************************************
3 * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
5 * Copyright 2009 VMware, Inc. All Rights Reserved.
7 * Permission is hereby granted, free of charge, to any person obtaining a
8 * copy of this software and associated documentation files (the
9 * "Software"), to deal in the Software without restriction, including
10 * without limitation the rights to use, copy, modify, merge, publish,
11 * distribute, sub license, and/or sell copies of the Software, and to
12 * permit persons to whom the Software is furnished to do so, subject to
13 * the following conditions:
15 * The above copyright notice and this permission notice (including the
16 * next paragraph) shall be included in all copies or substantial portions
19 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
20 * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
21 * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
22 * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
23 * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
24 * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
25 * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
27 **************************************************************************/
31 #include "shader/program.h"
32 #include "shader/prog_parameter.h"
33 #include "shader/prog_cache.h"
34 #include "shader/prog_instruction.h"
35 #include "shader/prog_print.h"
36 #include "shader/prog_statevars.h"
37 #include "shader/programopt.h"
38 #include "texenvprogram.h"
42 * Note on texture units:
44 * The number of texture units supported by fixed-function fragment
45 * processing is MAX_TEXTURE_COORD_UNITS, not MAX_TEXTURE_IMAGE_UNITS.
46 * That's because there's a one-to-one correspondence between texture
47 * coordinates and samplers in fixed-function processing.
49 * Since fixed-function vertex processing is limited to MAX_TEXTURE_COORD_UNITS
50 * sets of texcoords, so is fixed-function fragment processing.
52 * We can safely use ctx->Const.MaxTextureUnits for loop bounds.
56 struct texenvprog_cache_item
60 struct gl_fragment_program
*data
;
61 struct texenvprog_cache_item
*next
;
65 texenv_doing_secondary_color(GLcontext
*ctx
)
67 if (ctx
->Light
.Enabled
&&
68 (ctx
->Light
.Model
.ColorControl
== GL_SEPARATE_SPECULAR_COLOR
))
71 if (ctx
->Fog
.ColorSumEnabled
)
78 * Up to nine instructions per tex unit, plus fog, specular color.
80 #define MAX_INSTRUCTIONS ((MAX_TEXTURE_COORD_UNITS * 9) + 12)
82 #define DISASSEM (MESA_VERBOSE & VERBOSE_DISASSEM)
86 __extension__ GLubyte Source
:4; /**< SRC_x */
87 __extension__ GLubyte Operand
:3; /**< OPR_x */
89 GLubyte Source
; /**< SRC_x */
90 GLubyte Operand
; /**< OPR_x */
95 GLuint nr_enabled_units
:8;
96 GLuint enabled_units
:8;
97 GLuint separate_specular
:1;
99 GLuint fog_mode
:2; /**< FOG_x */
100 GLuint inputs_available
:12;
102 /* NOTE: This array of structs must be last! (see "keySize" below) */
105 GLuint source_index
:3; /**< TEXTURE_x_INDEX */
107 GLuint ScaleShiftRGB
:2;
108 GLuint ScaleShiftA
:2;
110 GLuint NumArgsRGB
:3; /**< up to MAX_COMBINER_TERMS */
111 GLuint ModeRGB
:5; /**< MODE_x */
113 GLuint NumArgsA
:3; /**< up to MAX_COMBINER_TERMS */
114 GLuint ModeA
:5; /**< MODE_x */
116 GLuint texture_cyl_wrap
:1; /**< For gallium test/debug only */
118 struct mode_opt OptRGB
[MAX_COMBINER_TERMS
];
119 struct mode_opt OptA
[MAX_COMBINER_TERMS
];
120 } unit
[MAX_TEXTURE_UNITS
];
126 #define FOG_UNKNOWN 3
128 static GLuint
translate_fog_mode( GLenum mode
)
131 case GL_LINEAR
: return FOG_LINEAR
;
132 case GL_EXP
: return FOG_EXP
;
133 case GL_EXP2
: return FOG_EXP2
;
134 default: return FOG_UNKNOWN
;
138 #define OPR_SRC_COLOR 0
139 #define OPR_ONE_MINUS_SRC_COLOR 1
140 #define OPR_SRC_ALPHA 2
141 #define OPR_ONE_MINUS_SRC_ALPHA 3
144 #define OPR_UNKNOWN 7
146 static GLuint
translate_operand( GLenum operand
)
149 case GL_SRC_COLOR
: return OPR_SRC_COLOR
;
150 case GL_ONE_MINUS_SRC_COLOR
: return OPR_ONE_MINUS_SRC_COLOR
;
151 case GL_SRC_ALPHA
: return OPR_SRC_ALPHA
;
152 case GL_ONE_MINUS_SRC_ALPHA
: return OPR_ONE_MINUS_SRC_ALPHA
;
153 case GL_ZERO
: return OPR_ZERO
;
154 case GL_ONE
: return OPR_ONE
;
161 #define SRC_TEXTURE 0
162 #define SRC_TEXTURE0 1
163 #define SRC_TEXTURE1 2
164 #define SRC_TEXTURE2 3
165 #define SRC_TEXTURE3 4
166 #define SRC_TEXTURE4 5
167 #define SRC_TEXTURE5 6
168 #define SRC_TEXTURE6 7
169 #define SRC_TEXTURE7 8
170 #define SRC_CONSTANT 9
171 #define SRC_PRIMARY_COLOR 10
172 #define SRC_PREVIOUS 11
174 #define SRC_UNKNOWN 15
176 static GLuint
translate_source( GLenum src
)
179 case GL_TEXTURE
: return SRC_TEXTURE
;
187 case GL_TEXTURE7
: return SRC_TEXTURE0
+ (src
- GL_TEXTURE0
);
188 case GL_CONSTANT
: return SRC_CONSTANT
;
189 case GL_PRIMARY_COLOR
: return SRC_PRIMARY_COLOR
;
190 case GL_PREVIOUS
: return SRC_PREVIOUS
;
199 #define MODE_REPLACE 0 /* r = a0 */
200 #define MODE_MODULATE 1 /* r = a0 * a1 */
201 #define MODE_ADD 2 /* r = a0 + a1 */
202 #define MODE_ADD_SIGNED 3 /* r = a0 + a1 - 0.5 */
203 #define MODE_INTERPOLATE 4 /* r = a0 * a2 + a1 * (1 - a2) */
204 #define MODE_SUBTRACT 5 /* r = a0 - a1 */
205 #define MODE_DOT3_RGB 6 /* r = a0 . a1 */
206 #define MODE_DOT3_RGB_EXT 7 /* r = a0 . a1 */
207 #define MODE_DOT3_RGBA 8 /* r = a0 . a1 */
208 #define MODE_DOT3_RGBA_EXT 9 /* r = a0 . a1 */
209 #define MODE_MODULATE_ADD_ATI 10 /* r = a0 * a2 + a1 */
210 #define MODE_MODULATE_SIGNED_ADD_ATI 11 /* r = a0 * a2 + a1 - 0.5 */
211 #define MODE_MODULATE_SUBTRACT_ATI 12 /* r = a0 * a2 - a1 */
212 #define MODE_ADD_PRODUCTS 13 /* r = a0 * a1 + a2 * a3 */
213 #define MODE_ADD_PRODUCTS_SIGNED 14 /* r = a0 * a1 + a2 * a3 - 0.5 */
214 #define MODE_BUMP_ENVMAP_ATI 15 /* special */
215 #define MODE_UNKNOWN 16
218 * Translate GL combiner state into a MODE_x value
220 static GLuint
translate_mode( GLenum envMode
, GLenum mode
)
223 case GL_REPLACE
: return MODE_REPLACE
;
224 case GL_MODULATE
: return MODE_MODULATE
;
226 if (envMode
== GL_COMBINE4_NV
)
227 return MODE_ADD_PRODUCTS
;
231 if (envMode
== GL_COMBINE4_NV
)
232 return MODE_ADD_PRODUCTS_SIGNED
;
234 return MODE_ADD_SIGNED
;
235 case GL_INTERPOLATE
: return MODE_INTERPOLATE
;
236 case GL_SUBTRACT
: return MODE_SUBTRACT
;
237 case GL_DOT3_RGB
: return MODE_DOT3_RGB
;
238 case GL_DOT3_RGB_EXT
: return MODE_DOT3_RGB_EXT
;
239 case GL_DOT3_RGBA
: return MODE_DOT3_RGBA
;
240 case GL_DOT3_RGBA_EXT
: return MODE_DOT3_RGBA_EXT
;
241 case GL_MODULATE_ADD_ATI
: return MODE_MODULATE_ADD_ATI
;
242 case GL_MODULATE_SIGNED_ADD_ATI
: return MODE_MODULATE_SIGNED_ADD_ATI
;
243 case GL_MODULATE_SUBTRACT_ATI
: return MODE_MODULATE_SUBTRACT_ATI
;
244 case GL_BUMP_ENVMAP_ATI
: return MODE_BUMP_ENVMAP_ATI
;
253 * Do we need to clamp the results of the given texture env/combine mode?
254 * If the inputs to the mode are in [0,1] we don't always have to clamp
258 need_saturate( GLuint mode
)
263 case MODE_INTERPOLATE
:
266 case MODE_ADD_SIGNED
:
269 case MODE_DOT3_RGB_EXT
:
271 case MODE_DOT3_RGBA_EXT
:
272 case MODE_MODULATE_ADD_ATI
:
273 case MODE_MODULATE_SIGNED_ADD_ATI
:
274 case MODE_MODULATE_SUBTRACT_ATI
:
275 case MODE_ADD_PRODUCTS
:
276 case MODE_ADD_PRODUCTS_SIGNED
:
277 case MODE_BUMP_ENVMAP_ATI
:
288 * Translate TEXTURE_x_BIT to TEXTURE_x_INDEX.
290 static GLuint
translate_tex_src_bit( GLbitfield bit
)
293 return _mesa_ffs(bit
) - 1;
297 #define VERT_BIT_TEX_ANY (0xff << VERT_ATTRIB_TEX0)
298 #define VERT_RESULT_TEX_ANY (0xff << VERT_RESULT_TEX0)
301 * Identify all possible varying inputs. The fragment program will
302 * never reference non-varying inputs, but will track them via state
305 * This function figures out all the inputs that the fragment program
306 * has access to. The bitmask is later reduced to just those which
307 * are actually referenced.
309 static GLbitfield
get_fp_input_mask( GLcontext
*ctx
)
312 const GLboolean vertexShader
= (ctx
->Shader
.CurrentProgram
&&
313 ctx
->Shader
.CurrentProgram
->LinkStatus
&&
314 ctx
->Shader
.CurrentProgram
->VertexProgram
);
315 const GLboolean vertexProgram
= ctx
->VertexProgram
._Enabled
;
316 GLbitfield fp_inputs
= 0x0;
318 if (ctx
->VertexProgram
._Overriden
) {
319 /* Somebody's messing with the vertex program and we don't have
320 * a clue what's happening. Assume that it could be producing
321 * all possible outputs.
325 else if (ctx
->RenderMode
== GL_FEEDBACK
) {
326 /* _NEW_RENDERMODE */
327 fp_inputs
= (FRAG_BIT_COL0
| FRAG_BIT_TEX0
);
329 else if (!(vertexProgram
|| vertexShader
) ||
330 !ctx
->VertexProgram
._Current
) {
331 /* Fixed function vertex logic */
333 GLbitfield varying_inputs
= ctx
->varying_vp_inputs
;
335 /* These get generated in the setup routine regardless of the
339 if (ctx
->Point
.PointSprite
)
340 varying_inputs
|= FRAG_BITS_TEX_ANY
;
342 /* First look at what values may be computed by the generated
346 if (ctx
->Light
.Enabled
) {
347 fp_inputs
|= FRAG_BIT_COL0
;
349 if (texenv_doing_secondary_color(ctx
))
350 fp_inputs
|= FRAG_BIT_COL1
;
354 fp_inputs
|= (ctx
->Texture
._TexGenEnabled
|
355 ctx
->Texture
._TexMatEnabled
) << FRAG_ATTRIB_TEX0
;
357 /* Then look at what might be varying as a result of enabled
360 if (varying_inputs
& VERT_BIT_COLOR0
)
361 fp_inputs
|= FRAG_BIT_COL0
;
362 if (varying_inputs
& VERT_BIT_COLOR1
)
363 fp_inputs
|= FRAG_BIT_COL1
;
365 fp_inputs
|= (((varying_inputs
& VERT_BIT_TEX_ANY
) >> VERT_ATTRIB_TEX0
)
366 << FRAG_ATTRIB_TEX0
);
370 /* calculate from vp->outputs */
371 struct gl_vertex_program
*vprog
;
372 GLbitfield64 vp_outputs
;
374 /* Choose GLSL vertex shader over ARB vertex program. Need this
375 * since vertex shader state validation comes after fragment state
376 * validation (see additional comments in state.c).
379 vprog
= ctx
->Shader
.CurrentProgram
->VertexProgram
;
381 vprog
= ctx
->VertexProgram
.Current
;
383 vp_outputs
= vprog
->Base
.OutputsWritten
;
385 /* These get generated in the setup routine regardless of the
389 if (ctx
->Point
.PointSprite
)
390 vp_outputs
|= FRAG_BITS_TEX_ANY
;
392 if (vp_outputs
& (1 << VERT_RESULT_COL0
))
393 fp_inputs
|= FRAG_BIT_COL0
;
394 if (vp_outputs
& (1 << VERT_RESULT_COL1
))
395 fp_inputs
|= FRAG_BIT_COL1
;
397 fp_inputs
|= (((vp_outputs
& VERT_RESULT_TEX_ANY
) >> VERT_RESULT_TEX0
)
398 << FRAG_ATTRIB_TEX0
);
406 * Examine current texture environment state and generate a unique
407 * key to identify it.
409 static GLuint
make_state_key( GLcontext
*ctx
, struct state_key
*key
)
412 GLbitfield inputs_referenced
= FRAG_BIT_COL0
;
413 const GLbitfield inputs_available
= get_fp_input_mask( ctx
);
416 memset(key
, 0, sizeof(*key
));
419 for (i
= 0; i
< ctx
->Const
.MaxTextureUnits
; i
++) {
420 const struct gl_texture_unit
*texUnit
= &ctx
->Texture
.Unit
[i
];
421 const struct gl_texture_object
*texObj
= texUnit
->_Current
;
422 const struct gl_tex_env_combine_state
*comb
= texUnit
->_CurrentCombine
;
425 if (!texUnit
->_ReallyEnabled
|| !texUnit
->Enabled
)
428 format
= texObj
->Image
[0][texObj
->BaseLevel
]->_BaseFormat
;
430 key
->unit
[i
].enabled
= 1;
431 key
->enabled_units
|= (1<<i
);
432 key
->nr_enabled_units
= i
+ 1;
433 inputs_referenced
|= FRAG_BIT_TEX(i
);
435 key
->unit
[i
].source_index
=
436 translate_tex_src_bit(texUnit
->_ReallyEnabled
);
438 key
->unit
[i
].shadow
= ((texObj
->CompareMode
== GL_COMPARE_R_TO_TEXTURE
) &&
439 ((format
== GL_DEPTH_COMPONENT
) ||
440 (format
== GL_DEPTH_STENCIL_EXT
)));
442 key
->unit
[i
].NumArgsRGB
= comb
->_NumArgsRGB
;
443 key
->unit
[i
].NumArgsA
= comb
->_NumArgsA
;
445 key
->unit
[i
].ModeRGB
=
446 translate_mode(texUnit
->EnvMode
, comb
->ModeRGB
);
448 translate_mode(texUnit
->EnvMode
, comb
->ModeA
);
450 key
->unit
[i
].ScaleShiftRGB
= comb
->ScaleShiftRGB
;
451 key
->unit
[i
].ScaleShiftA
= comb
->ScaleShiftA
;
453 for (j
= 0; j
< MAX_COMBINER_TERMS
; j
++) {
454 key
->unit
[i
].OptRGB
[j
].Operand
= translate_operand(comb
->OperandRGB
[j
]);
455 key
->unit
[i
].OptA
[j
].Operand
= translate_operand(comb
->OperandA
[j
]);
456 key
->unit
[i
].OptRGB
[j
].Source
= translate_source(comb
->SourceRGB
[j
]);
457 key
->unit
[i
].OptA
[j
].Source
= translate_source(comb
->SourceA
[j
]);
460 if (key
->unit
[i
].ModeRGB
== MODE_BUMP_ENVMAP_ATI
) {
461 /* requires some special translation */
462 key
->unit
[i
].NumArgsRGB
= 2;
463 key
->unit
[i
].ScaleShiftRGB
= 0;
464 key
->unit
[i
].OptRGB
[0].Operand
= OPR_SRC_COLOR
;
465 key
->unit
[i
].OptRGB
[0].Source
= SRC_TEXTURE
;
466 key
->unit
[i
].OptRGB
[1].Operand
= OPR_SRC_COLOR
;
467 key
->unit
[i
].OptRGB
[1].Source
= texUnit
->BumpTarget
- GL_TEXTURE0
+ SRC_TEXTURE0
;
470 /* this is a back-door for enabling cylindrical texture wrap mode */
471 if (texObj
->Priority
== 0.125)
472 key
->unit
[i
].texture_cyl_wrap
= 1;
475 /* _NEW_LIGHT | _NEW_FOG */
476 if (texenv_doing_secondary_color(ctx
)) {
477 key
->separate_specular
= 1;
478 inputs_referenced
|= FRAG_BIT_COL1
;
482 if (ctx
->Fog
.Enabled
) {
483 key
->fog_enabled
= 1;
484 key
->fog_mode
= translate_fog_mode(ctx
->Fog
.Mode
);
485 inputs_referenced
|= FRAG_BIT_FOGC
; /* maybe */
488 key
->inputs_available
= (inputs_available
& inputs_referenced
);
490 /* compute size of state key, ignoring unused texture units */
491 keySize
= sizeof(*key
) - sizeof(key
->unit
)
492 + key
->nr_enabled_units
* sizeof(key
->unit
[0]);
499 * Use uregs to represent registers internally, translate to Mesa's
500 * expected formats on emit.
502 * NOTE: These are passed by value extensively in this file rather
503 * than as usual by pointer reference. If this disturbs you, try
504 * remembering they are just 32bits in size.
506 * GCC is smart enough to deal with these dword-sized structures in
507 * much the same way as if I had defined them as dwords and was using
508 * macros to access and set the fields. This is much nicer and easier
519 static const struct ureg undef
= {
528 /** State used to build the fragment program:
530 struct texenv_fragment_program
{
531 struct gl_fragment_program
*program
;
532 struct state_key
*state
;
534 GLbitfield alu_temps
; /**< Track texture indirections, see spec. */
535 GLbitfield temps_output
; /**< Track texture indirections, see spec. */
536 GLbitfield temp_in_use
; /**< Tracks temporary regs which are in use. */
539 struct ureg src_texture
[MAX_TEXTURE_COORD_UNITS
];
540 /* Reg containing each texture unit's sampled texture color,
544 struct ureg texcoord_tex
[MAX_TEXTURE_COORD_UNITS
];
545 /* Reg containing texcoord for a texture unit,
546 * needed for bump mapping, else undef.
549 struct ureg src_previous
; /**< Reg containing color from previous
550 * stage. May need to be decl'd.
553 GLuint last_tex_stage
; /**< Number of last enabled texture unit */
562 static struct ureg
make_ureg(GLuint file
, GLuint idx
)
568 reg
.swz
= SWIZZLE_NOOP
;
573 static struct ureg
swizzle( struct ureg reg
, int x
, int y
, int z
, int w
)
575 reg
.swz
= MAKE_SWIZZLE4(GET_SWZ(reg
.swz
, x
),
578 GET_SWZ(reg
.swz
, w
));
583 static struct ureg
swizzle1( struct ureg reg
, int x
)
585 return swizzle(reg
, x
, x
, x
, x
);
588 static struct ureg
negate( struct ureg reg
)
594 static GLboolean
is_undef( struct ureg reg
)
596 return reg
.file
== PROGRAM_UNDEFINED
;
600 static struct ureg
get_temp( struct texenv_fragment_program
*p
)
604 /* First try and reuse temps which have been used already:
606 bit
= _mesa_ffs( ~p
->temp_in_use
& p
->alu_temps
);
608 /* Then any unused temporary:
611 bit
= _mesa_ffs( ~p
->temp_in_use
);
614 _mesa_problem(NULL
, "%s: out of temporaries\n", __FILE__
);
618 if ((GLuint
) bit
> p
->program
->Base
.NumTemporaries
)
619 p
->program
->Base
.NumTemporaries
= bit
;
621 p
->temp_in_use
|= 1<<(bit
-1);
622 return make_ureg(PROGRAM_TEMPORARY
, (bit
-1));
625 static struct ureg
get_tex_temp( struct texenv_fragment_program
*p
)
629 /* First try to find available temp not previously used (to avoid
630 * starting a new texture indirection). According to the spec, the
631 * ~p->temps_output isn't necessary, but will keep it there for
634 bit
= _mesa_ffs( ~p
->temp_in_use
& ~p
->alu_temps
& ~p
->temps_output
);
636 /* Then any unused temporary:
639 bit
= _mesa_ffs( ~p
->temp_in_use
);
642 _mesa_problem(NULL
, "%s: out of temporaries\n", __FILE__
);
646 if ((GLuint
) bit
> p
->program
->Base
.NumTemporaries
)
647 p
->program
->Base
.NumTemporaries
= bit
;
649 p
->temp_in_use
|= 1<<(bit
-1);
650 return make_ureg(PROGRAM_TEMPORARY
, (bit
-1));
654 /** Mark a temp reg as being no longer allocatable. */
655 static void reserve_temp( struct texenv_fragment_program
*p
, struct ureg r
)
657 if (r
.file
== PROGRAM_TEMPORARY
)
658 p
->temps_output
|= (1 << r
.idx
);
662 static void release_temps(GLcontext
*ctx
, struct texenv_fragment_program
*p
)
664 GLuint max_temp
= ctx
->Const
.FragmentProgram
.MaxTemps
;
666 /* KW: To support tex_env_crossbar, don't release the registers in
669 if (max_temp
>= sizeof(int) * 8)
670 p
->temp_in_use
= p
->temps_output
;
672 p
->temp_in_use
= ~((1<<max_temp
)-1) | p
->temps_output
;
676 static struct ureg
register_param5( struct texenv_fragment_program
*p
,
683 gl_state_index tokens
[STATE_LENGTH
];
690 idx
= _mesa_add_state_reference( p
->program
->Base
.Parameters
, tokens
);
691 return make_ureg(PROGRAM_STATE_VAR
, idx
);
695 #define register_param1(p,s0) register_param5(p,s0,0,0,0,0)
696 #define register_param2(p,s0,s1) register_param5(p,s0,s1,0,0,0)
697 #define register_param3(p,s0,s1,s2) register_param5(p,s0,s1,s2,0,0)
698 #define register_param4(p,s0,s1,s2,s3) register_param5(p,s0,s1,s2,s3,0)
700 static GLuint
frag_to_vert_attrib( GLuint attrib
)
703 case FRAG_ATTRIB_COL0
: return VERT_ATTRIB_COLOR0
;
704 case FRAG_ATTRIB_COL1
: return VERT_ATTRIB_COLOR1
;
706 assert(attrib
>= FRAG_ATTRIB_TEX0
);
707 assert(attrib
<= FRAG_ATTRIB_TEX7
);
708 return attrib
- FRAG_ATTRIB_TEX0
+ VERT_ATTRIB_TEX0
;
713 static struct ureg
register_input( struct texenv_fragment_program
*p
, GLuint input
)
715 if (p
->state
->inputs_available
& (1<<input
)) {
716 p
->program
->Base
.InputsRead
|= (1 << input
);
717 return make_ureg(PROGRAM_INPUT
, input
);
720 GLuint idx
= frag_to_vert_attrib( input
);
721 return register_param3( p
, STATE_INTERNAL
, STATE_CURRENT_ATTRIB
, idx
);
726 static void emit_arg( struct prog_src_register
*reg
,
729 reg
->File
= ureg
.file
;
730 reg
->Index
= ureg
.idx
;
731 reg
->Swizzle
= ureg
.swz
;
732 reg
->Negate
= ureg
.negatebase
? NEGATE_XYZW
: NEGATE_NONE
;
736 static void emit_dst( struct prog_dst_register
*dst
,
737 struct ureg ureg
, GLuint mask
)
739 dst
->File
= ureg
.file
;
740 dst
->Index
= ureg
.idx
;
741 dst
->WriteMask
= mask
;
742 dst
->CondMask
= COND_TR
; /* always pass cond test */
743 dst
->CondSwizzle
= SWIZZLE_NOOP
;
746 static struct prog_instruction
*
747 emit_op(struct texenv_fragment_program
*p
,
756 const GLuint nr
= p
->program
->Base
.NumInstructions
++;
757 struct prog_instruction
*inst
= &p
->program
->Base
.Instructions
[nr
];
759 assert(nr
< MAX_INSTRUCTIONS
);
761 _mesa_init_instructions(inst
, 1);
764 emit_arg( &inst
->SrcReg
[0], src0
);
765 emit_arg( &inst
->SrcReg
[1], src1
);
766 emit_arg( &inst
->SrcReg
[2], src2
);
768 inst
->SaturateMode
= saturate
? SATURATE_ZERO_ONE
: SATURATE_OFF
;
770 emit_dst( &inst
->DstReg
, dest
, mask
);
773 /* Accounting for indirection tracking:
775 if (dest
.file
== PROGRAM_TEMPORARY
)
776 p
->temps_output
|= 1 << dest
.idx
;
783 static struct ureg
emit_arith( struct texenv_fragment_program
*p
,
792 emit_op(p
, op
, dest
, mask
, saturate
, src0
, src1
, src2
);
794 /* Accounting for indirection tracking:
796 if (src0
.file
== PROGRAM_TEMPORARY
)
797 p
->alu_temps
|= 1 << src0
.idx
;
799 if (!is_undef(src1
) && src1
.file
== PROGRAM_TEMPORARY
)
800 p
->alu_temps
|= 1 << src1
.idx
;
802 if (!is_undef(src2
) && src2
.file
== PROGRAM_TEMPORARY
)
803 p
->alu_temps
|= 1 << src2
.idx
;
805 if (dest
.file
== PROGRAM_TEMPORARY
)
806 p
->alu_temps
|= 1 << dest
.idx
;
808 p
->program
->Base
.NumAluInstructions
++;
812 static struct ureg
emit_texld( struct texenv_fragment_program
*p
,
821 struct prog_instruction
*inst
= emit_op( p
, op
,
823 GL_FALSE
, /* don't saturate? */
828 inst
->TexSrcTarget
= tex_idx
;
829 inst
->TexSrcUnit
= tex_unit
;
830 inst
->TexShadow
= tex_shadow
;
832 p
->program
->Base
.NumTexInstructions
++;
834 /* Accounting for indirection tracking:
836 reserve_temp(p
, dest
);
839 /* Is this a texture indirection?
841 if ((coord
.file
== PROGRAM_TEMPORARY
&&
842 (p
->temps_output
& (1<<coord
.idx
))) ||
843 (dest
.file
== PROGRAM_TEMPORARY
&&
844 (p
->alu_temps
& (1<<dest
.idx
)))) {
845 p
->program
->Base
.NumTexIndirections
++;
846 p
->temps_output
= 1<<coord
.idx
;
848 assert(0); /* KW: texture env crossbar */
856 static struct ureg
register_const4f( struct texenv_fragment_program
*p
,
869 idx
= _mesa_add_unnamed_constant( p
->program
->Base
.Parameters
, values
, 4,
871 r
= make_ureg(PROGRAM_CONSTANT
, idx
);
876 #define register_scalar_const(p, s0) register_const4f(p, s0, s0, s0, s0)
877 #define register_const1f(p, s0) register_const4f(p, s0, 0, 0, 1)
878 #define register_const2f(p, s0, s1) register_const4f(p, s0, s1, 0, 1)
879 #define register_const3f(p, s0, s1, s2) register_const4f(p, s0, s1, s2, 1)
882 static struct ureg
get_one( struct texenv_fragment_program
*p
)
884 if (is_undef(p
->one
))
885 p
->one
= register_scalar_const(p
, 1.0);
889 static struct ureg
get_half( struct texenv_fragment_program
*p
)
891 if (is_undef(p
->half
))
892 p
->half
= register_scalar_const(p
, 0.5);
896 static struct ureg
get_zero( struct texenv_fragment_program
*p
)
898 if (is_undef(p
->zero
))
899 p
->zero
= register_scalar_const(p
, 0.0);
904 static void program_error( struct texenv_fragment_program
*p
, const char *msg
)
906 _mesa_problem(NULL
, msg
);
910 static struct ureg
get_source( struct texenv_fragment_program
*p
,
911 GLuint src
, GLuint unit
)
915 assert(!is_undef(p
->src_texture
[unit
]));
916 return p
->src_texture
[unit
];
926 assert(!is_undef(p
->src_texture
[src
- SRC_TEXTURE0
]));
927 return p
->src_texture
[src
- SRC_TEXTURE0
];
930 return register_param2(p
, STATE_TEXENV_COLOR
, unit
);
932 case SRC_PRIMARY_COLOR
:
933 return register_input(p
, FRAG_ATTRIB_COL0
);
939 if (is_undef(p
->src_previous
))
940 return register_input(p
, FRAG_ATTRIB_COL0
);
942 return p
->src_previous
;
950 static struct ureg
emit_combine_source( struct texenv_fragment_program
*p
,
956 struct ureg arg
, src
, one
;
958 src
= get_source(p
, source
, unit
);
961 case OPR_ONE_MINUS_SRC_COLOR
:
963 * Emit tmp = 1.0 - arg.xyzw
967 return emit_arith( p
, OPCODE_SUB
, arg
, mask
, 0, one
, src
, undef
);
970 if (mask
== WRITEMASK_W
)
973 return swizzle1( src
, SWIZZLE_W
);
974 case OPR_ONE_MINUS_SRC_ALPHA
:
976 * Emit tmp = 1.0 - arg.wwww
980 return emit_arith(p
, OPCODE_SUB
, arg
, mask
, 0,
981 one
, swizzle1(src
, SWIZZLE_W
), undef
);
995 * Check if the RGB and Alpha sources and operands match for the given
996 * texture unit's combinder state. When the RGB and A sources and
997 * operands match, we can emit fewer instructions.
999 static GLboolean
args_match( const struct state_key
*key
, GLuint unit
)
1001 GLuint i
, numArgs
= key
->unit
[unit
].NumArgsRGB
;
1003 for (i
= 0; i
< numArgs
; i
++) {
1004 if (key
->unit
[unit
].OptA
[i
].Source
!= key
->unit
[unit
].OptRGB
[i
].Source
)
1007 switch (key
->unit
[unit
].OptA
[i
].Operand
) {
1009 switch (key
->unit
[unit
].OptRGB
[i
].Operand
) {
1017 case OPR_ONE_MINUS_SRC_ALPHA
:
1018 switch (key
->unit
[unit
].OptRGB
[i
].Operand
) {
1019 case OPR_ONE_MINUS_SRC_COLOR
:
1020 case OPR_ONE_MINUS_SRC_ALPHA
:
1027 return GL_FALSE
; /* impossible */
1034 static struct ureg
emit_combine( struct texenv_fragment_program
*p
,
1041 const struct mode_opt
*opt
)
1043 struct ureg src
[MAX_COMBINER_TERMS
];
1044 struct ureg tmp
, half
;
1047 assert(nr
<= MAX_COMBINER_TERMS
);
1049 for (i
= 0; i
< nr
; i
++)
1050 src
[i
] = emit_combine_source( p
, mask
, unit
, opt
[i
].Source
, opt
[i
].Operand
);
1054 if (mask
== WRITEMASK_XYZW
&& !saturate
)
1057 return emit_arith( p
, OPCODE_MOV
, dest
, mask
, saturate
, src
[0], undef
, undef
);
1059 return emit_arith( p
, OPCODE_MUL
, dest
, mask
, saturate
,
1060 src
[0], src
[1], undef
);
1062 return emit_arith( p
, OPCODE_ADD
, dest
, mask
, saturate
,
1063 src
[0], src
[1], undef
);
1064 case MODE_ADD_SIGNED
:
1065 /* tmp = arg0 + arg1
1069 tmp
= get_temp( p
);
1070 emit_arith( p
, OPCODE_ADD
, tmp
, mask
, 0, src
[0], src
[1], undef
);
1071 emit_arith( p
, OPCODE_SUB
, dest
, mask
, saturate
, tmp
, half
, undef
);
1073 case MODE_INTERPOLATE
:
1074 /* Arg0 * (Arg2) + Arg1 * (1-Arg2) -- note arguments are reordered:
1076 return emit_arith( p
, OPCODE_LRP
, dest
, mask
, saturate
, src
[2], src
[0], src
[1] );
1079 return emit_arith( p
, OPCODE_SUB
, dest
, mask
, saturate
, src
[0], src
[1], undef
);
1081 case MODE_DOT3_RGBA
:
1082 case MODE_DOT3_RGBA_EXT
:
1083 case MODE_DOT3_RGB_EXT
:
1084 case MODE_DOT3_RGB
: {
1085 struct ureg tmp0
= get_temp( p
);
1086 struct ureg tmp1
= get_temp( p
);
1087 struct ureg neg1
= register_scalar_const(p
, -1);
1088 struct ureg two
= register_scalar_const(p
, 2);
1090 /* tmp0 = 2*src0 - 1
1093 * dst = tmp0 dot3 tmp1
1095 emit_arith( p
, OPCODE_MAD
, tmp0
, WRITEMASK_XYZW
, 0,
1098 if (_mesa_memcmp(&src
[0], &src
[1], sizeof(struct ureg
)) == 0)
1101 emit_arith( p
, OPCODE_MAD
, tmp1
, WRITEMASK_XYZW
, 0,
1103 emit_arith( p
, OPCODE_DP3
, dest
, mask
, saturate
, tmp0
, tmp1
, undef
);
1106 case MODE_MODULATE_ADD_ATI
:
1107 /* Arg0 * Arg2 + Arg1 */
1108 return emit_arith( p
, OPCODE_MAD
, dest
, mask
, saturate
,
1109 src
[0], src
[2], src
[1] );
1110 case MODE_MODULATE_SIGNED_ADD_ATI
: {
1111 /* Arg0 * Arg2 + Arg1 - 0.5 */
1112 struct ureg tmp0
= get_temp(p
);
1114 emit_arith( p
, OPCODE_MAD
, tmp0
, mask
, 0, src
[0], src
[2], src
[1] );
1115 emit_arith( p
, OPCODE_SUB
, dest
, mask
, saturate
, tmp0
, half
, undef
);
1118 case MODE_MODULATE_SUBTRACT_ATI
:
1119 /* Arg0 * Arg2 - Arg1 */
1120 emit_arith( p
, OPCODE_MAD
, dest
, mask
, 0, src
[0], src
[2], negate(src
[1]) );
1122 case MODE_ADD_PRODUCTS
:
1123 /* Arg0 * Arg1 + Arg2 * Arg3 */
1125 struct ureg tmp0
= get_temp(p
);
1126 emit_arith( p
, OPCODE_MUL
, tmp0
, mask
, 0, src
[0], src
[1], undef
);
1127 emit_arith( p
, OPCODE_MAD
, dest
, mask
, saturate
, src
[2], src
[3], tmp0
);
1130 case MODE_ADD_PRODUCTS_SIGNED
:
1131 /* Arg0 * Arg1 + Arg2 * Arg3 - 0.5 */
1133 struct ureg tmp0
= get_temp(p
);
1135 emit_arith( p
, OPCODE_MUL
, tmp0
, mask
, 0, src
[0], src
[1], undef
);
1136 emit_arith( p
, OPCODE_MAD
, tmp0
, mask
, 0, src
[2], src
[3], tmp0
);
1137 emit_arith( p
, OPCODE_SUB
, dest
, mask
, saturate
, tmp0
, half
, undef
);
1140 case MODE_BUMP_ENVMAP_ATI
:
1141 /* special - not handled here */
1152 * Generate instructions for one texture unit's env/combiner mode.
1155 emit_texenv(struct texenv_fragment_program
*p
, GLuint unit
)
1157 const struct state_key
*key
= p
->state
;
1158 GLboolean rgb_saturate
, alpha_saturate
;
1159 GLuint rgb_shift
, alpha_shift
;
1160 struct ureg out
, dest
;
1162 if (!key
->unit
[unit
].enabled
) {
1163 return get_source(p
, SRC_PREVIOUS
, 0);
1165 if (key
->unit
[unit
].ModeRGB
== MODE_BUMP_ENVMAP_ATI
) {
1166 /* this isn't really a env stage delivering a color and handled elsewhere */
1167 return get_source(p
, SRC_PREVIOUS
, 0);
1170 switch (key
->unit
[unit
].ModeRGB
) {
1171 case MODE_DOT3_RGB_EXT
:
1172 alpha_shift
= key
->unit
[unit
].ScaleShiftA
;
1175 case MODE_DOT3_RGBA_EXT
:
1180 rgb_shift
= key
->unit
[unit
].ScaleShiftRGB
;
1181 alpha_shift
= key
->unit
[unit
].ScaleShiftA
;
1185 /* If we'll do rgb/alpha shifting don't saturate in emit_combine().
1186 * We don't want to clamp twice.
1189 rgb_saturate
= GL_FALSE
; /* saturate after rgb shift */
1190 else if (need_saturate(key
->unit
[unit
].ModeRGB
))
1191 rgb_saturate
= GL_TRUE
;
1193 rgb_saturate
= GL_FALSE
;
1196 alpha_saturate
= GL_FALSE
; /* saturate after alpha shift */
1197 else if (need_saturate(key
->unit
[unit
].ModeA
))
1198 alpha_saturate
= GL_TRUE
;
1200 alpha_saturate
= GL_FALSE
;
1202 /* If this is the very last calculation, emit direct to output reg:
1204 if (key
->separate_specular
||
1205 unit
!= p
->last_tex_stage
||
1208 dest
= get_temp( p
);
1210 dest
= make_ureg(PROGRAM_OUTPUT
, FRAG_RESULT_COLOR
);
1212 /* Emit the RGB and A combine ops
1214 if (key
->unit
[unit
].ModeRGB
== key
->unit
[unit
].ModeA
&&
1215 args_match(key
, unit
)) {
1216 out
= emit_combine( p
, dest
, WRITEMASK_XYZW
, rgb_saturate
,
1218 key
->unit
[unit
].NumArgsRGB
,
1219 key
->unit
[unit
].ModeRGB
,
1220 key
->unit
[unit
].OptRGB
);
1222 else if (key
->unit
[unit
].ModeRGB
== MODE_DOT3_RGBA_EXT
||
1223 key
->unit
[unit
].ModeRGB
== MODE_DOT3_RGBA
) {
1224 out
= emit_combine( p
, dest
, WRITEMASK_XYZW
, rgb_saturate
,
1226 key
->unit
[unit
].NumArgsRGB
,
1227 key
->unit
[unit
].ModeRGB
,
1228 key
->unit
[unit
].OptRGB
);
1231 /* Need to do something to stop from re-emitting identical
1232 * argument calculations here:
1234 out
= emit_combine( p
, dest
, WRITEMASK_XYZ
, rgb_saturate
,
1236 key
->unit
[unit
].NumArgsRGB
,
1237 key
->unit
[unit
].ModeRGB
,
1238 key
->unit
[unit
].OptRGB
);
1239 out
= emit_combine( p
, dest
, WRITEMASK_W
, alpha_saturate
,
1241 key
->unit
[unit
].NumArgsA
,
1242 key
->unit
[unit
].ModeA
,
1243 key
->unit
[unit
].OptA
);
1246 /* Deal with the final shift:
1248 if (alpha_shift
|| rgb_shift
) {
1250 GLboolean saturate
= GL_TRUE
; /* always saturate at this point */
1252 if (rgb_shift
== alpha_shift
) {
1253 shift
= register_scalar_const(p
, (GLfloat
)(1<<rgb_shift
));
1256 shift
= register_const4f(p
,
1257 (GLfloat
)(1<<rgb_shift
),
1258 (GLfloat
)(1<<rgb_shift
),
1259 (GLfloat
)(1<<rgb_shift
),
1260 (GLfloat
)(1<<alpha_shift
));
1262 return emit_arith( p
, OPCODE_MUL
, dest
, WRITEMASK_XYZW
,
1263 saturate
, out
, shift
, undef
);
1271 * Generate instruction for getting a texture source term.
1273 static void load_texture( struct texenv_fragment_program
*p
, GLuint unit
)
1275 if (is_undef(p
->src_texture
[unit
])) {
1276 const GLuint texTarget
= p
->state
->unit
[unit
].source_index
;
1277 struct ureg texcoord
;
1278 struct ureg tmp
= get_tex_temp( p
);
1280 if (is_undef(p
->texcoord_tex
[unit
])) {
1281 texcoord
= register_input(p
, FRAG_ATTRIB_TEX0
+unit
);
1284 /* might want to reuse this reg for tex output actually */
1285 texcoord
= p
->texcoord_tex
[unit
];
1288 /* TODO: Use D0_MASK_XY where possible.
1290 if (p
->state
->unit
[unit
].enabled
) {
1291 GLboolean shadow
= GL_FALSE
;
1293 if (p
->state
->unit
[unit
].shadow
) {
1294 p
->program
->Base
.ShadowSamplers
|= 1 << unit
;
1298 p
->src_texture
[unit
] = emit_texld( p
, OPCODE_TXP
,
1299 tmp
, WRITEMASK_XYZW
,
1300 unit
, texTarget
, shadow
,
1303 p
->program
->Base
.SamplersUsed
|= (1 << unit
);
1304 /* This identity mapping should already be in place
1305 * (see _mesa_init_program_struct()) but let's be safe.
1307 p
->program
->Base
.SamplerUnits
[unit
] = unit
;
1310 p
->src_texture
[unit
] = get_zero(p
);
1312 if (p
->state
->unit
[unit
].texture_cyl_wrap
) {
1313 /* set flag which is checked by Mesa->Gallium program translation */
1314 p
->program
->Base
.InputFlags
[0] |= PROG_PARAM_BIT_CYL_WRAP
;
1320 static GLboolean
load_texenv_source( struct texenv_fragment_program
*p
,
1321 GLuint src
, GLuint unit
)
1325 load_texture(p
, unit
);
1336 load_texture(p
, src
- SRC_TEXTURE0
);
1340 /* not a texture src - do nothing */
1349 * Generate instructions for loading all texture source terms.
1352 load_texunit_sources( struct texenv_fragment_program
*p
, GLuint unit
)
1354 const struct state_key
*key
= p
->state
;
1357 for (i
= 0; i
< key
->unit
[unit
].NumArgsRGB
; i
++) {
1358 load_texenv_source( p
, key
->unit
[unit
].OptRGB
[i
].Source
, unit
);
1361 for (i
= 0; i
< key
->unit
[unit
].NumArgsA
; i
++) {
1362 load_texenv_source( p
, key
->unit
[unit
].OptA
[i
].Source
, unit
);
1369 * Generate instructions for loading bump map textures.
1372 load_texunit_bumpmap( struct texenv_fragment_program
*p
, GLuint unit
)
1374 const struct state_key
*key
= p
->state
;
1375 GLuint bumpedUnitNr
= key
->unit
[unit
].OptRGB
[1].Source
- SRC_TEXTURE0
;
1376 struct ureg texcDst
, bumpMapRes
;
1377 struct ureg constdudvcolor
= register_const4f(p
, 0.0, 0.0, 0.0, 1.0);
1378 struct ureg texcSrc
= register_input(p
, FRAG_ATTRIB_TEX0
+ bumpedUnitNr
);
1379 struct ureg rotMat0
= register_param3( p
, STATE_INTERNAL
, STATE_ROT_MATRIX_0
, unit
);
1380 struct ureg rotMat1
= register_param3( p
, STATE_INTERNAL
, STATE_ROT_MATRIX_1
, unit
);
1382 load_texenv_source( p
, unit
+ SRC_TEXTURE0
, unit
);
1384 bumpMapRes
= get_source(p
, key
->unit
[unit
].OptRGB
[0].Source
, unit
);
1385 texcDst
= get_tex_temp( p
);
1386 p
->texcoord_tex
[bumpedUnitNr
] = texcDst
;
1388 /* Apply rot matrix and add coords to be available in next phase.
1389 * dest = (Arg0.xxxx * rotMat0 + Arg1) + (Arg0.yyyy * rotMat1)
1390 * note only 2 coords are affected the rest are left unchanged (mul by 0)
1392 emit_arith( p
, OPCODE_MAD
, texcDst
, WRITEMASK_XYZW
, 0,
1393 swizzle1(bumpMapRes
, SWIZZLE_X
), rotMat0
, texcSrc
);
1394 emit_arith( p
, OPCODE_MAD
, texcDst
, WRITEMASK_XYZW
, 0,
1395 swizzle1(bumpMapRes
, SWIZZLE_Y
), rotMat1
, texcDst
);
1397 /* Move 0,0,0,1 into bumpmap src if someone (crossbar) is foolish
1398 * enough to access this later, should optimize away.
1400 emit_arith( p
, OPCODE_MOV
, bumpMapRes
, WRITEMASK_XYZW
, 0,
1401 constdudvcolor
, undef
, undef
);
1407 * Generate a new fragment program which implements the context's
1408 * current texture env/combine mode.
1411 create_new_program(GLcontext
*ctx
, struct state_key
*key
,
1412 struct gl_fragment_program
*program
)
1414 struct prog_instruction instBuffer
[MAX_INSTRUCTIONS
];
1415 struct texenv_fragment_program p
;
1417 struct ureg cf
, out
;
1419 memset(&p
, 0, sizeof(p
));
1421 p
.program
= program
;
1423 /* During code generation, use locally-allocated instruction buffer,
1424 * then alloc dynamic storage below.
1426 p
.program
->Base
.Instructions
= instBuffer
;
1427 p
.program
->Base
.Target
= GL_FRAGMENT_PROGRAM_ARB
;
1428 p
.program
->Base
.String
= NULL
;
1429 p
.program
->Base
.NumTexIndirections
= 1; /* is this right? */
1430 p
.program
->Base
.NumTexInstructions
= 0;
1431 p
.program
->Base
.NumAluInstructions
= 0;
1432 p
.program
->Base
.NumInstructions
= 0;
1433 p
.program
->Base
.NumTemporaries
= 0;
1434 p
.program
->Base
.NumParameters
= 0;
1435 p
.program
->Base
.NumAttributes
= 0;
1436 p
.program
->Base
.NumAddressRegs
= 0;
1437 p
.program
->Base
.Parameters
= _mesa_new_parameter_list();
1438 p
.program
->Base
.InputsRead
= 0x0;
1439 p
.program
->Base
.OutputsWritten
= 1 << FRAG_RESULT_COLOR
;
1441 for (unit
= 0; unit
< ctx
->Const
.MaxTextureUnits
; unit
++) {
1442 p
.src_texture
[unit
] = undef
;
1443 p
.texcoord_tex
[unit
] = undef
;
1446 p
.src_previous
= undef
;
1451 p
.last_tex_stage
= 0;
1452 release_temps(ctx
, &p
);
1454 if (key
->enabled_units
) {
1455 GLboolean needbumpstage
= GL_FALSE
;
1457 /* Zeroth pass - bump map textures first */
1458 for (unit
= 0; unit
< key
->nr_enabled_units
; unit
++)
1459 if (key
->unit
[unit
].enabled
&&
1460 key
->unit
[unit
].ModeRGB
== MODE_BUMP_ENVMAP_ATI
) {
1461 needbumpstage
= GL_TRUE
;
1462 load_texunit_bumpmap( &p
, unit
);
1465 p
.program
->Base
.NumTexIndirections
++;
1467 /* First pass - to support texture_env_crossbar, first identify
1468 * all referenced texture sources and emit texld instructions
1471 for (unit
= 0; unit
< key
->nr_enabled_units
; unit
++)
1472 if (key
->unit
[unit
].enabled
) {
1473 load_texunit_sources( &p
, unit
);
1474 p
.last_tex_stage
= unit
;
1477 /* Second pass - emit combine instructions to build final color:
1479 for (unit
= 0; unit
< key
->nr_enabled_units
; unit
++)
1480 if (key
->unit
[unit
].enabled
) {
1481 p
.src_previous
= emit_texenv( &p
, unit
);
1482 reserve_temp(&p
, p
.src_previous
); /* don't re-use this temp reg */
1483 release_temps(ctx
, &p
); /* release all temps */
1487 cf
= get_source( &p
, SRC_PREVIOUS
, 0 );
1488 out
= make_ureg( PROGRAM_OUTPUT
, FRAG_RESULT_COLOR
);
1490 if (key
->separate_specular
) {
1491 /* Emit specular add.
1493 struct ureg s
= register_input(&p
, FRAG_ATTRIB_COL1
);
1494 emit_arith( &p
, OPCODE_ADD
, out
, WRITEMASK_XYZ
, 0, cf
, s
, undef
);
1495 emit_arith( &p
, OPCODE_MOV
, out
, WRITEMASK_W
, 0, cf
, undef
, undef
);
1497 else if (_mesa_memcmp(&cf
, &out
, sizeof(cf
)) != 0) {
1498 /* Will wind up in here if no texture enabled or a couple of
1499 * other scenarios (GL_REPLACE for instance).
1501 emit_arith( &p
, OPCODE_MOV
, out
, WRITEMASK_XYZW
, 0, cf
, undef
, undef
);
1506 emit_arith( &p
, OPCODE_END
, undef
, WRITEMASK_XYZW
, 0, undef
, undef
, undef
);
1508 if (key
->fog_enabled
) {
1509 /* Pull fog mode from GLcontext, the value in the state key is
1510 * a reduced value and not what is expected in FogOption
1512 p
.program
->FogOption
= ctx
->Fog
.Mode
;
1513 p
.program
->Base
.InputsRead
|= FRAG_BIT_FOGC
;
1516 p
.program
->FogOption
= GL_NONE
;
1519 if (p
.program
->Base
.NumTexIndirections
> ctx
->Const
.FragmentProgram
.MaxTexIndirections
)
1520 program_error(&p
, "Exceeded max nr indirect texture lookups");
1522 if (p
.program
->Base
.NumTexInstructions
> ctx
->Const
.FragmentProgram
.MaxTexInstructions
)
1523 program_error(&p
, "Exceeded max TEX instructions");
1525 if (p
.program
->Base
.NumAluInstructions
> ctx
->Const
.FragmentProgram
.MaxAluInstructions
)
1526 program_error(&p
, "Exceeded max ALU instructions");
1528 ASSERT(p
.program
->Base
.NumInstructions
<= MAX_INSTRUCTIONS
);
1530 /* Allocate final instruction array */
1531 p
.program
->Base
.Instructions
1532 = _mesa_alloc_instructions(p
.program
->Base
.NumInstructions
);
1533 if (!p
.program
->Base
.Instructions
) {
1534 _mesa_error(ctx
, GL_OUT_OF_MEMORY
,
1535 "generating tex env program");
1538 _mesa_copy_instructions(p
.program
->Base
.Instructions
, instBuffer
,
1539 p
.program
->Base
.NumInstructions
);
1541 if (p
.program
->FogOption
) {
1542 _mesa_append_fog_code(ctx
, p
.program
);
1543 p
.program
->FogOption
= GL_NONE
;
1547 /* Notify driver the fragment program has (actually) changed.
1549 if (ctx
->Driver
.ProgramStringNotify
) {
1550 GLboolean ok
= ctx
->Driver
.ProgramStringNotify(ctx
,
1551 GL_FRAGMENT_PROGRAM_ARB
,
1553 /* Driver should be able to handle any texenv programs as long as
1554 * the driver correctly reported max number of texture units correctly,
1558 (void) ok
; /* silence unused var warning */
1562 _mesa_print_program(&p
.program
->Base
);
1569 * Return a fragment program which implements the current
1570 * fixed-function texture, fog and color-sum operations.
1572 struct gl_fragment_program
*
1573 _mesa_get_fixed_func_fragment_program(GLcontext
*ctx
)
1575 struct gl_fragment_program
*prog
;
1576 struct state_key key
;
1579 keySize
= make_state_key(ctx
, &key
);
1581 prog
= (struct gl_fragment_program
*)
1582 _mesa_search_program_cache(ctx
->FragmentProgram
.Cache
,
1586 prog
= (struct gl_fragment_program
*)
1587 ctx
->Driver
.NewProgram(ctx
, GL_FRAGMENT_PROGRAM_ARB
, 0);
1589 create_new_program(ctx
, &key
, prog
);
1591 _mesa_program_cache_insert(ctx
, ctx
->FragmentProgram
.Cache
,
1592 &key
, keySize
, &prog
->Base
);