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 struct mode_opt OptRGB
[MAX_COMBINER_TERMS
];
117 struct mode_opt OptA
[MAX_COMBINER_TERMS
];
118 } unit
[MAX_TEXTURE_UNITS
];
124 #define FOG_UNKNOWN 3
126 static GLuint
translate_fog_mode( GLenum mode
)
129 case GL_LINEAR
: return FOG_LINEAR
;
130 case GL_EXP
: return FOG_EXP
;
131 case GL_EXP2
: return FOG_EXP2
;
132 default: return FOG_UNKNOWN
;
136 #define OPR_SRC_COLOR 0
137 #define OPR_ONE_MINUS_SRC_COLOR 1
138 #define OPR_SRC_ALPHA 2
139 #define OPR_ONE_MINUS_SRC_ALPHA 3
142 #define OPR_UNKNOWN 7
144 static GLuint
translate_operand( GLenum operand
)
147 case GL_SRC_COLOR
: return OPR_SRC_COLOR
;
148 case GL_ONE_MINUS_SRC_COLOR
: return OPR_ONE_MINUS_SRC_COLOR
;
149 case GL_SRC_ALPHA
: return OPR_SRC_ALPHA
;
150 case GL_ONE_MINUS_SRC_ALPHA
: return OPR_ONE_MINUS_SRC_ALPHA
;
151 case GL_ZERO
: return OPR_ZERO
;
152 case GL_ONE
: return OPR_ONE
;
159 #define SRC_TEXTURE 0
160 #define SRC_TEXTURE0 1
161 #define SRC_TEXTURE1 2
162 #define SRC_TEXTURE2 3
163 #define SRC_TEXTURE3 4
164 #define SRC_TEXTURE4 5
165 #define SRC_TEXTURE5 6
166 #define SRC_TEXTURE6 7
167 #define SRC_TEXTURE7 8
168 #define SRC_CONSTANT 9
169 #define SRC_PRIMARY_COLOR 10
170 #define SRC_PREVIOUS 11
172 #define SRC_UNKNOWN 15
174 static GLuint
translate_source( GLenum src
)
177 case GL_TEXTURE
: return SRC_TEXTURE
;
185 case GL_TEXTURE7
: return SRC_TEXTURE0
+ (src
- GL_TEXTURE0
);
186 case GL_CONSTANT
: return SRC_CONSTANT
;
187 case GL_PRIMARY_COLOR
: return SRC_PRIMARY_COLOR
;
188 case GL_PREVIOUS
: return SRC_PREVIOUS
;
197 #define MODE_REPLACE 0 /* r = a0 */
198 #define MODE_MODULATE 1 /* r = a0 * a1 */
199 #define MODE_ADD 2 /* r = a0 + a1 */
200 #define MODE_ADD_SIGNED 3 /* r = a0 + a1 - 0.5 */
201 #define MODE_INTERPOLATE 4 /* r = a0 * a2 + a1 * (1 - a2) */
202 #define MODE_SUBTRACT 5 /* r = a0 - a1 */
203 #define MODE_DOT3_RGB 6 /* r = a0 . a1 */
204 #define MODE_DOT3_RGB_EXT 7 /* r = a0 . a1 */
205 #define MODE_DOT3_RGBA 8 /* r = a0 . a1 */
206 #define MODE_DOT3_RGBA_EXT 9 /* r = a0 . a1 */
207 #define MODE_MODULATE_ADD_ATI 10 /* r = a0 * a2 + a1 */
208 #define MODE_MODULATE_SIGNED_ADD_ATI 11 /* r = a0 * a2 + a1 - 0.5 */
209 #define MODE_MODULATE_SUBTRACT_ATI 12 /* r = a0 * a2 - a1 */
210 #define MODE_ADD_PRODUCTS 13 /* r = a0 * a1 + a2 * a3 */
211 #define MODE_ADD_PRODUCTS_SIGNED 14 /* r = a0 * a1 + a2 * a3 - 0.5 */
212 #define MODE_BUMP_ENVMAP_ATI 15 /* special */
213 #define MODE_UNKNOWN 16
216 * Translate GL combiner state into a MODE_x value
218 static GLuint
translate_mode( GLenum envMode
, GLenum mode
)
221 case GL_REPLACE
: return MODE_REPLACE
;
222 case GL_MODULATE
: return MODE_MODULATE
;
224 if (envMode
== GL_COMBINE4_NV
)
225 return MODE_ADD_PRODUCTS
;
229 if (envMode
== GL_COMBINE4_NV
)
230 return MODE_ADD_PRODUCTS_SIGNED
;
232 return MODE_ADD_SIGNED
;
233 case GL_INTERPOLATE
: return MODE_INTERPOLATE
;
234 case GL_SUBTRACT
: return MODE_SUBTRACT
;
235 case GL_DOT3_RGB
: return MODE_DOT3_RGB
;
236 case GL_DOT3_RGB_EXT
: return MODE_DOT3_RGB_EXT
;
237 case GL_DOT3_RGBA
: return MODE_DOT3_RGBA
;
238 case GL_DOT3_RGBA_EXT
: return MODE_DOT3_RGBA_EXT
;
239 case GL_MODULATE_ADD_ATI
: return MODE_MODULATE_ADD_ATI
;
240 case GL_MODULATE_SIGNED_ADD_ATI
: return MODE_MODULATE_SIGNED_ADD_ATI
;
241 case GL_MODULATE_SUBTRACT_ATI
: return MODE_MODULATE_SUBTRACT_ATI
;
242 case GL_BUMP_ENVMAP_ATI
: return MODE_BUMP_ENVMAP_ATI
;
251 * Do we need to clamp the results of the given texture env/combine mode?
252 * If the inputs to the mode are in [0,1] we don't always have to clamp
256 need_saturate( GLuint mode
)
261 case MODE_INTERPOLATE
:
264 case MODE_ADD_SIGNED
:
267 case MODE_DOT3_RGB_EXT
:
269 case MODE_DOT3_RGBA_EXT
:
270 case MODE_MODULATE_ADD_ATI
:
271 case MODE_MODULATE_SIGNED_ADD_ATI
:
272 case MODE_MODULATE_SUBTRACT_ATI
:
273 case MODE_ADD_PRODUCTS
:
274 case MODE_ADD_PRODUCTS_SIGNED
:
275 case MODE_BUMP_ENVMAP_ATI
:
285 * Translate TEXTURE_x_BIT to TEXTURE_x_INDEX.
287 static GLuint
translate_tex_src_bit( GLbitfield bit
)
290 return _mesa_ffs(bit
) - 1;
294 #define VERT_BIT_TEX_ANY (0xff << VERT_ATTRIB_TEX0)
295 #define VERT_RESULT_TEX_ANY (0xff << VERT_RESULT_TEX0)
298 * Identify all possible varying inputs. The fragment program will
299 * never reference non-varying inputs, but will track them via state
302 * This function figures out all the inputs that the fragment program
303 * has access to. The bitmask is later reduced to just those which
304 * are actually referenced.
306 static GLbitfield
get_fp_input_mask( GLcontext
*ctx
)
309 const GLboolean vertexShader
= (ctx
->Shader
.CurrentProgram
&&
310 ctx
->Shader
.CurrentProgram
->LinkStatus
&&
311 ctx
->Shader
.CurrentProgram
->VertexProgram
);
312 const GLboolean vertexProgram
= ctx
->VertexProgram
._Enabled
;
313 GLbitfield fp_inputs
= 0x0;
315 if (ctx
->VertexProgram
._Overriden
) {
316 /* Somebody's messing with the vertex program and we don't have
317 * a clue what's happening. Assume that it could be producing
318 * all possible outputs.
322 else if (ctx
->RenderMode
== GL_FEEDBACK
) {
323 /* _NEW_RENDERMODE */
324 fp_inputs
= (FRAG_BIT_COL0
| FRAG_BIT_TEX0
);
326 else if (!(vertexProgram
|| vertexShader
) ||
327 !ctx
->VertexProgram
._Current
) {
328 /* Fixed function vertex logic */
330 GLbitfield varying_inputs
= ctx
->varying_vp_inputs
;
332 /* These get generated in the setup routine regardless of the
336 if (ctx
->Point
.PointSprite
)
337 varying_inputs
|= FRAG_BITS_TEX_ANY
;
339 /* First look at what values may be computed by the generated
343 if (ctx
->Light
.Enabled
) {
344 fp_inputs
|= FRAG_BIT_COL0
;
346 if (texenv_doing_secondary_color(ctx
))
347 fp_inputs
|= FRAG_BIT_COL1
;
351 fp_inputs
|= (ctx
->Texture
._TexGenEnabled
|
352 ctx
->Texture
._TexMatEnabled
) << FRAG_ATTRIB_TEX0
;
354 /* Then look at what might be varying as a result of enabled
357 if (varying_inputs
& VERT_BIT_COLOR0
)
358 fp_inputs
|= FRAG_BIT_COL0
;
359 if (varying_inputs
& VERT_BIT_COLOR1
)
360 fp_inputs
|= FRAG_BIT_COL1
;
362 fp_inputs
|= (((varying_inputs
& VERT_BIT_TEX_ANY
) >> VERT_ATTRIB_TEX0
)
363 << FRAG_ATTRIB_TEX0
);
367 /* calculate from vp->outputs */
368 struct gl_vertex_program
*vprog
;
369 GLbitfield vp_outputs
;
371 /* Choose GLSL vertex shader over ARB vertex program. Need this
372 * since vertex shader state validation comes after fragment state
373 * validation (see additional comments in state.c).
376 vprog
= ctx
->Shader
.CurrentProgram
->VertexProgram
;
378 vprog
= ctx
->VertexProgram
.Current
;
380 vp_outputs
= vprog
->Base
.OutputsWritten
;
382 /* These get generated in the setup routine regardless of the
386 if (ctx
->Point
.PointSprite
)
387 vp_outputs
|= FRAG_BITS_TEX_ANY
;
389 if (vp_outputs
& (1 << VERT_RESULT_COL0
))
390 fp_inputs
|= FRAG_BIT_COL0
;
391 if (vp_outputs
& (1 << VERT_RESULT_COL1
))
392 fp_inputs
|= FRAG_BIT_COL1
;
394 fp_inputs
|= (((vp_outputs
& VERT_RESULT_TEX_ANY
) >> VERT_RESULT_TEX0
)
395 << FRAG_ATTRIB_TEX0
);
403 * Examine current texture environment state and generate a unique
404 * key to identify it.
406 static GLuint
make_state_key( GLcontext
*ctx
, struct state_key
*key
)
409 GLbitfield inputs_referenced
= FRAG_BIT_COL0
;
410 const GLbitfield inputs_available
= get_fp_input_mask( ctx
);
413 memset(key
, 0, sizeof(*key
));
416 for (i
= 0; i
< ctx
->Const
.MaxTextureUnits
; i
++) {
417 const struct gl_texture_unit
*texUnit
= &ctx
->Texture
.Unit
[i
];
418 const struct gl_texture_object
*texObj
= texUnit
->_Current
;
419 const struct gl_tex_env_combine_state
*comb
= texUnit
->_CurrentCombine
;
422 if (!texUnit
->_ReallyEnabled
|| !texUnit
->Enabled
)
425 format
= texObj
->Image
[0][texObj
->BaseLevel
]->_BaseFormat
;
427 key
->unit
[i
].enabled
= 1;
428 key
->enabled_units
|= (1<<i
);
429 key
->nr_enabled_units
= i
+ 1;
430 inputs_referenced
|= FRAG_BIT_TEX(i
);
432 key
->unit
[i
].source_index
=
433 translate_tex_src_bit(texUnit
->_ReallyEnabled
);
435 key
->unit
[i
].shadow
= ((texObj
->CompareMode
== GL_COMPARE_R_TO_TEXTURE
) &&
436 ((format
== GL_DEPTH_COMPONENT
) ||
437 (format
== GL_DEPTH_STENCIL_EXT
)));
439 key
->unit
[i
].NumArgsRGB
= comb
->_NumArgsRGB
;
440 key
->unit
[i
].NumArgsA
= comb
->_NumArgsA
;
442 key
->unit
[i
].ModeRGB
=
443 translate_mode(texUnit
->EnvMode
, comb
->ModeRGB
);
445 translate_mode(texUnit
->EnvMode
, comb
->ModeA
);
447 key
->unit
[i
].ScaleShiftRGB
= comb
->ScaleShiftRGB
;
448 key
->unit
[i
].ScaleShiftA
= comb
->ScaleShiftA
;
450 for (j
= 0; j
< MAX_COMBINER_TERMS
; j
++) {
451 key
->unit
[i
].OptRGB
[j
].Operand
= translate_operand(comb
->OperandRGB
[j
]);
452 key
->unit
[i
].OptA
[j
].Operand
= translate_operand(comb
->OperandA
[j
]);
453 key
->unit
[i
].OptRGB
[j
].Source
= translate_source(comb
->SourceRGB
[j
]);
454 key
->unit
[i
].OptA
[j
].Source
= translate_source(comb
->SourceA
[j
]);
457 if (key
->unit
[i
].ModeRGB
== MODE_BUMP_ENVMAP_ATI
) {
458 /* requires some special translation */
459 key
->unit
[i
].NumArgsRGB
= 2;
460 key
->unit
[i
].ScaleShiftRGB
= 0;
461 key
->unit
[i
].OptRGB
[0].Operand
= OPR_SRC_COLOR
;
462 key
->unit
[i
].OptRGB
[0].Source
= SRC_TEXTURE
;
463 key
->unit
[i
].OptRGB
[1].Operand
= OPR_SRC_COLOR
;
464 key
->unit
[i
].OptRGB
[1].Source
= texUnit
->BumpTarget
- GL_TEXTURE0
+ SRC_TEXTURE0
;
468 /* _NEW_LIGHT | _NEW_FOG */
469 if (texenv_doing_secondary_color(ctx
)) {
470 key
->separate_specular
= 1;
471 inputs_referenced
|= FRAG_BIT_COL1
;
475 if (ctx
->Fog
.Enabled
) {
476 key
->fog_enabled
= 1;
477 key
->fog_mode
= translate_fog_mode(ctx
->Fog
.Mode
);
478 inputs_referenced
|= FRAG_BIT_FOGC
; /* maybe */
481 key
->inputs_available
= (inputs_available
& inputs_referenced
);
483 /* compute size of state key, ignoring unused texture units */
484 keySize
= sizeof(*key
) - sizeof(key
->unit
)
485 + key
->nr_enabled_units
* sizeof(key
->unit
[0]);
492 * Use uregs to represent registers internally, translate to Mesa's
493 * expected formats on emit.
495 * NOTE: These are passed by value extensively in this file rather
496 * than as usual by pointer reference. If this disturbs you, try
497 * remembering they are just 32bits in size.
499 * GCC is smart enough to deal with these dword-sized structures in
500 * much the same way as if I had defined them as dwords and was using
501 * macros to access and set the fields. This is much nicer and easier
512 static const struct ureg undef
= {
521 /** State used to build the fragment program:
523 struct texenv_fragment_program
{
524 struct gl_fragment_program
*program
;
525 struct state_key
*state
;
527 GLbitfield alu_temps
; /**< Track texture indirections, see spec. */
528 GLbitfield temps_output
; /**< Track texture indirections, see spec. */
529 GLbitfield temp_in_use
; /**< Tracks temporary regs which are in use. */
532 struct ureg src_texture
[MAX_TEXTURE_COORD_UNITS
];
533 /* Reg containing each texture unit's sampled texture color,
537 struct ureg texcoord_tex
[MAX_TEXTURE_COORD_UNITS
];
538 /* Reg containing texcoord for a texture unit,
539 * needed for bump mapping, else undef.
542 struct ureg src_previous
; /**< Reg containing color from previous
543 * stage. May need to be decl'd.
546 GLuint last_tex_stage
; /**< Number of last enabled texture unit */
555 static struct ureg
make_ureg(GLuint file
, GLuint idx
)
561 reg
.swz
= SWIZZLE_NOOP
;
566 static struct ureg
swizzle( struct ureg reg
, int x
, int y
, int z
, int w
)
568 reg
.swz
= MAKE_SWIZZLE4(GET_SWZ(reg
.swz
, x
),
571 GET_SWZ(reg
.swz
, w
));
576 static struct ureg
swizzle1( struct ureg reg
, int x
)
578 return swizzle(reg
, x
, x
, x
, x
);
581 static struct ureg
negate( struct ureg reg
)
587 static GLboolean
is_undef( struct ureg reg
)
589 return reg
.file
== PROGRAM_UNDEFINED
;
593 static struct ureg
get_temp( struct texenv_fragment_program
*p
)
597 /* First try and reuse temps which have been used already:
599 bit
= _mesa_ffs( ~p
->temp_in_use
& p
->alu_temps
);
601 /* Then any unused temporary:
604 bit
= _mesa_ffs( ~p
->temp_in_use
);
607 _mesa_problem(NULL
, "%s: out of temporaries\n", __FILE__
);
611 if ((GLuint
) bit
> p
->program
->Base
.NumTemporaries
)
612 p
->program
->Base
.NumTemporaries
= bit
;
614 p
->temp_in_use
|= 1<<(bit
-1);
615 return make_ureg(PROGRAM_TEMPORARY
, (bit
-1));
618 static struct ureg
get_tex_temp( struct texenv_fragment_program
*p
)
622 /* First try to find available temp not previously used (to avoid
623 * starting a new texture indirection). According to the spec, the
624 * ~p->temps_output isn't necessary, but will keep it there for
627 bit
= _mesa_ffs( ~p
->temp_in_use
& ~p
->alu_temps
& ~p
->temps_output
);
629 /* Then any unused temporary:
632 bit
= _mesa_ffs( ~p
->temp_in_use
);
635 _mesa_problem(NULL
, "%s: out of temporaries\n", __FILE__
);
639 if ((GLuint
) bit
> p
->program
->Base
.NumTemporaries
)
640 p
->program
->Base
.NumTemporaries
= bit
;
642 p
->temp_in_use
|= 1<<(bit
-1);
643 return make_ureg(PROGRAM_TEMPORARY
, (bit
-1));
647 /** Mark a temp reg as being no longer allocatable. */
648 static void reserve_temp( struct texenv_fragment_program
*p
, struct ureg r
)
650 if (r
.file
== PROGRAM_TEMPORARY
)
651 p
->temps_output
|= (1 << r
.idx
);
655 static void release_temps(GLcontext
*ctx
, struct texenv_fragment_program
*p
)
657 GLuint max_temp
= ctx
->Const
.FragmentProgram
.MaxTemps
;
659 /* KW: To support tex_env_crossbar, don't release the registers in
662 if (max_temp
>= sizeof(int) * 8)
663 p
->temp_in_use
= p
->temps_output
;
665 p
->temp_in_use
= ~((1<<max_temp
)-1) | p
->temps_output
;
669 static struct ureg
register_param5( struct texenv_fragment_program
*p
,
676 gl_state_index tokens
[STATE_LENGTH
];
683 idx
= _mesa_add_state_reference( p
->program
->Base
.Parameters
, tokens
);
684 return make_ureg(PROGRAM_STATE_VAR
, idx
);
688 #define register_param1(p,s0) register_param5(p,s0,0,0,0,0)
689 #define register_param2(p,s0,s1) register_param5(p,s0,s1,0,0,0)
690 #define register_param3(p,s0,s1,s2) register_param5(p,s0,s1,s2,0,0)
691 #define register_param4(p,s0,s1,s2,s3) register_param5(p,s0,s1,s2,s3,0)
693 static GLuint
frag_to_vert_attrib( GLuint attrib
)
696 case FRAG_ATTRIB_COL0
: return VERT_ATTRIB_COLOR0
;
697 case FRAG_ATTRIB_COL1
: return VERT_ATTRIB_COLOR1
;
699 assert(attrib
>= FRAG_ATTRIB_TEX0
);
700 assert(attrib
<= FRAG_ATTRIB_TEX7
);
701 return attrib
- FRAG_ATTRIB_TEX0
+ VERT_ATTRIB_TEX0
;
706 static struct ureg
register_input( struct texenv_fragment_program
*p
, GLuint input
)
708 if (p
->state
->inputs_available
& (1<<input
)) {
709 p
->program
->Base
.InputsRead
|= (1 << input
);
710 return make_ureg(PROGRAM_INPUT
, input
);
713 GLuint idx
= frag_to_vert_attrib( input
);
714 return register_param3( p
, STATE_INTERNAL
, STATE_CURRENT_ATTRIB
, idx
);
719 static void emit_arg( struct prog_src_register
*reg
,
722 reg
->File
= ureg
.file
;
723 reg
->Index
= ureg
.idx
;
724 reg
->Swizzle
= ureg
.swz
;
725 reg
->Negate
= ureg
.negatebase
? NEGATE_XYZW
: NEGATE_NONE
;
729 static void emit_dst( struct prog_dst_register
*dst
,
730 struct ureg ureg
, GLuint mask
)
732 dst
->File
= ureg
.file
;
733 dst
->Index
= ureg
.idx
;
734 dst
->WriteMask
= mask
;
735 dst
->CondMask
= COND_TR
; /* always pass cond test */
736 dst
->CondSwizzle
= SWIZZLE_NOOP
;
739 static struct prog_instruction
*
740 emit_op(struct texenv_fragment_program
*p
,
749 const GLuint nr
= p
->program
->Base
.NumInstructions
++;
750 struct prog_instruction
*inst
= &p
->program
->Base
.Instructions
[nr
];
752 assert(nr
< MAX_INSTRUCTIONS
);
754 _mesa_init_instructions(inst
, 1);
757 emit_arg( &inst
->SrcReg
[0], src0
);
758 emit_arg( &inst
->SrcReg
[1], src1
);
759 emit_arg( &inst
->SrcReg
[2], src2
);
761 inst
->SaturateMode
= saturate
? SATURATE_ZERO_ONE
: SATURATE_OFF
;
763 emit_dst( &inst
->DstReg
, dest
, mask
);
766 /* Accounting for indirection tracking:
768 if (dest
.file
== PROGRAM_TEMPORARY
)
769 p
->temps_output
|= 1 << dest
.idx
;
776 static struct ureg
emit_arith( struct texenv_fragment_program
*p
,
785 emit_op(p
, op
, dest
, mask
, saturate
, src0
, src1
, src2
);
787 /* Accounting for indirection tracking:
789 if (src0
.file
== PROGRAM_TEMPORARY
)
790 p
->alu_temps
|= 1 << src0
.idx
;
792 if (!is_undef(src1
) && src1
.file
== PROGRAM_TEMPORARY
)
793 p
->alu_temps
|= 1 << src1
.idx
;
795 if (!is_undef(src2
) && src2
.file
== PROGRAM_TEMPORARY
)
796 p
->alu_temps
|= 1 << src2
.idx
;
798 if (dest
.file
== PROGRAM_TEMPORARY
)
799 p
->alu_temps
|= 1 << dest
.idx
;
801 p
->program
->Base
.NumAluInstructions
++;
805 static struct ureg
emit_texld( struct texenv_fragment_program
*p
,
814 struct prog_instruction
*inst
= emit_op( p
, op
,
816 GL_FALSE
, /* don't saturate? */
821 inst
->TexSrcTarget
= tex_idx
;
822 inst
->TexSrcUnit
= tex_unit
;
823 inst
->TexShadow
= tex_shadow
;
825 p
->program
->Base
.NumTexInstructions
++;
827 /* Accounting for indirection tracking:
829 reserve_temp(p
, dest
);
832 /* Is this a texture indirection?
834 if ((coord
.file
== PROGRAM_TEMPORARY
&&
835 (p
->temps_output
& (1<<coord
.idx
))) ||
836 (dest
.file
== PROGRAM_TEMPORARY
&&
837 (p
->alu_temps
& (1<<dest
.idx
)))) {
838 p
->program
->Base
.NumTexIndirections
++;
839 p
->temps_output
= 1<<coord
.idx
;
841 assert(0); /* KW: texture env crossbar */
849 static struct ureg
register_const4f( struct texenv_fragment_program
*p
,
862 idx
= _mesa_add_unnamed_constant( p
->program
->Base
.Parameters
, values
, 4,
864 r
= make_ureg(PROGRAM_CONSTANT
, idx
);
869 #define register_scalar_const(p, s0) register_const4f(p, s0, s0, s0, s0)
870 #define register_const1f(p, s0) register_const4f(p, s0, 0, 0, 1)
871 #define register_const2f(p, s0, s1) register_const4f(p, s0, s1, 0, 1)
872 #define register_const3f(p, s0, s1, s2) register_const4f(p, s0, s1, s2, 1)
875 static struct ureg
get_one( struct texenv_fragment_program
*p
)
877 if (is_undef(p
->one
))
878 p
->one
= register_scalar_const(p
, 1.0);
882 static struct ureg
get_half( struct texenv_fragment_program
*p
)
884 if (is_undef(p
->half
))
885 p
->half
= register_scalar_const(p
, 0.5);
889 static struct ureg
get_zero( struct texenv_fragment_program
*p
)
891 if (is_undef(p
->zero
))
892 p
->zero
= register_scalar_const(p
, 0.0);
897 static void program_error( struct texenv_fragment_program
*p
, const char *msg
)
899 _mesa_problem(NULL
, msg
);
903 static struct ureg
get_source( struct texenv_fragment_program
*p
,
904 GLuint src
, GLuint unit
)
908 assert(!is_undef(p
->src_texture
[unit
]));
909 return p
->src_texture
[unit
];
919 assert(!is_undef(p
->src_texture
[src
- SRC_TEXTURE0
]));
920 return p
->src_texture
[src
- SRC_TEXTURE0
];
923 return register_param2(p
, STATE_TEXENV_COLOR
, unit
);
925 case SRC_PRIMARY_COLOR
:
926 return register_input(p
, FRAG_ATTRIB_COL0
);
932 if (is_undef(p
->src_previous
))
933 return register_input(p
, FRAG_ATTRIB_COL0
);
935 return p
->src_previous
;
943 static struct ureg
emit_combine_source( struct texenv_fragment_program
*p
,
949 struct ureg arg
, src
, one
;
951 src
= get_source(p
, source
, unit
);
954 case OPR_ONE_MINUS_SRC_COLOR
:
956 * Emit tmp = 1.0 - arg.xyzw
960 return emit_arith( p
, OPCODE_SUB
, arg
, mask
, 0, one
, src
, undef
);
963 if (mask
== WRITEMASK_W
)
966 return swizzle1( src
, SWIZZLE_W
);
967 case OPR_ONE_MINUS_SRC_ALPHA
:
969 * Emit tmp = 1.0 - arg.wwww
973 return emit_arith(p
, OPCODE_SUB
, arg
, mask
, 0,
974 one
, swizzle1(src
, SWIZZLE_W
), undef
);
988 * Check if the RGB and Alpha sources and operands match for the given
989 * texture unit's combinder state. When the RGB and A sources and
990 * operands match, we can emit fewer instructions.
992 static GLboolean
args_match( const struct state_key
*key
, GLuint unit
)
994 GLuint i
, numArgs
= key
->unit
[unit
].NumArgsRGB
;
996 for (i
= 0; i
< numArgs
; i
++) {
997 if (key
->unit
[unit
].OptA
[i
].Source
!= key
->unit
[unit
].OptRGB
[i
].Source
)
1000 switch (key
->unit
[unit
].OptA
[i
].Operand
) {
1002 switch (key
->unit
[unit
].OptRGB
[i
].Operand
) {
1010 case OPR_ONE_MINUS_SRC_ALPHA
:
1011 switch (key
->unit
[unit
].OptRGB
[i
].Operand
) {
1012 case OPR_ONE_MINUS_SRC_COLOR
:
1013 case OPR_ONE_MINUS_SRC_ALPHA
:
1020 return GL_FALSE
; /* impossible */
1027 static struct ureg
emit_combine( struct texenv_fragment_program
*p
,
1034 const struct mode_opt
*opt
)
1036 struct ureg src
[MAX_COMBINER_TERMS
];
1037 struct ureg tmp
, half
;
1040 assert(nr
<= MAX_COMBINER_TERMS
);
1042 tmp
= undef
; /* silence warning (bug 5318) */
1044 for (i
= 0; i
< nr
; i
++)
1045 src
[i
] = emit_combine_source( p
, mask
, unit
, opt
[i
].Source
, opt
[i
].Operand
);
1049 if (mask
== WRITEMASK_XYZW
&& !saturate
)
1052 return emit_arith( p
, OPCODE_MOV
, dest
, mask
, saturate
, src
[0], undef
, undef
);
1054 return emit_arith( p
, OPCODE_MUL
, dest
, mask
, saturate
,
1055 src
[0], src
[1], undef
);
1057 return emit_arith( p
, OPCODE_ADD
, dest
, mask
, saturate
,
1058 src
[0], src
[1], undef
);
1059 case MODE_ADD_SIGNED
:
1060 /* tmp = arg0 + arg1
1064 tmp
= get_temp( p
);
1065 emit_arith( p
, OPCODE_ADD
, tmp
, mask
, 0, src
[0], src
[1], undef
);
1066 emit_arith( p
, OPCODE_SUB
, dest
, mask
, saturate
, tmp
, half
, undef
);
1068 case MODE_INTERPOLATE
:
1069 /* Arg0 * (Arg2) + Arg1 * (1-Arg2) -- note arguments are reordered:
1071 return emit_arith( p
, OPCODE_LRP
, dest
, mask
, saturate
, src
[2], src
[0], src
[1] );
1074 return emit_arith( p
, OPCODE_SUB
, dest
, mask
, saturate
, src
[0], src
[1], undef
);
1076 case MODE_DOT3_RGBA
:
1077 case MODE_DOT3_RGBA_EXT
:
1078 case MODE_DOT3_RGB_EXT
:
1079 case MODE_DOT3_RGB
: {
1080 struct ureg tmp0
= get_temp( p
);
1081 struct ureg tmp1
= get_temp( p
);
1082 struct ureg neg1
= register_scalar_const(p
, -1);
1083 struct ureg two
= register_scalar_const(p
, 2);
1085 /* tmp0 = 2*src0 - 1
1088 * dst = tmp0 dot3 tmp1
1090 emit_arith( p
, OPCODE_MAD
, tmp0
, WRITEMASK_XYZW
, 0,
1093 if (_mesa_memcmp(&src
[0], &src
[1], sizeof(struct ureg
)) == 0)
1096 emit_arith( p
, OPCODE_MAD
, tmp1
, WRITEMASK_XYZW
, 0,
1098 emit_arith( p
, OPCODE_DP3
, dest
, mask
, saturate
, tmp0
, tmp1
, undef
);
1101 case MODE_MODULATE_ADD_ATI
:
1102 /* Arg0 * Arg2 + Arg1 */
1103 return emit_arith( p
, OPCODE_MAD
, dest
, mask
, saturate
,
1104 src
[0], src
[2], src
[1] );
1105 case MODE_MODULATE_SIGNED_ADD_ATI
: {
1106 /* Arg0 * Arg2 + Arg1 - 0.5 */
1107 struct ureg tmp0
= get_temp(p
);
1109 emit_arith( p
, OPCODE_MAD
, tmp0
, mask
, 0, src
[0], src
[2], src
[1] );
1110 emit_arith( p
, OPCODE_SUB
, dest
, mask
, saturate
, tmp0
, half
, undef
);
1113 case MODE_MODULATE_SUBTRACT_ATI
:
1114 /* Arg0 * Arg2 - Arg1 */
1115 emit_arith( p
, OPCODE_MAD
, dest
, mask
, 0, src
[0], src
[2], negate(src
[1]) );
1117 case MODE_ADD_PRODUCTS
:
1118 /* Arg0 * Arg1 + Arg2 * Arg3 */
1120 struct ureg tmp0
= get_temp(p
);
1121 emit_arith( p
, OPCODE_MUL
, tmp0
, mask
, 0, src
[0], src
[1], undef
);
1122 emit_arith( p
, OPCODE_MAD
, dest
, mask
, saturate
, src
[2], src
[3], tmp0
);
1125 case MODE_ADD_PRODUCTS_SIGNED
:
1126 /* Arg0 * Arg1 + Arg2 * Arg3 - 0.5 */
1128 struct ureg tmp0
= get_temp(p
);
1130 emit_arith( p
, OPCODE_MUL
, tmp0
, mask
, 0, src
[0], src
[1], undef
);
1131 emit_arith( p
, OPCODE_MAD
, tmp0
, mask
, 0, src
[2], src
[3], tmp0
);
1132 emit_arith( p
, OPCODE_SUB
, dest
, mask
, saturate
, tmp0
, half
, undef
);
1135 case MODE_BUMP_ENVMAP_ATI
:
1136 /* special - not handled here */
1147 * Generate instructions for one texture unit's env/combiner mode.
1150 emit_texenv(struct texenv_fragment_program
*p
, GLuint unit
)
1152 const struct state_key
*key
= p
->state
;
1153 GLboolean rgb_saturate
, alpha_saturate
;
1154 GLuint rgb_shift
, alpha_shift
;
1155 struct ureg out
, dest
;
1157 if (!key
->unit
[unit
].enabled
) {
1158 return get_source(p
, SRC_PREVIOUS
, 0);
1160 if (key
->unit
[unit
].ModeRGB
== MODE_BUMP_ENVMAP_ATI
) {
1161 /* this isn't really a env stage delivering a color and handled elsewhere */
1162 return get_source(p
, SRC_PREVIOUS
, 0);
1165 switch (key
->unit
[unit
].ModeRGB
) {
1166 case MODE_DOT3_RGB_EXT
:
1167 alpha_shift
= key
->unit
[unit
].ScaleShiftA
;
1170 case MODE_DOT3_RGBA_EXT
:
1175 rgb_shift
= key
->unit
[unit
].ScaleShiftRGB
;
1176 alpha_shift
= key
->unit
[unit
].ScaleShiftA
;
1180 /* If we'll do rgb/alpha shifting don't saturate in emit_combine().
1181 * We don't want to clamp twice.
1184 rgb_saturate
= GL_FALSE
; /* saturate after rgb shift */
1185 else if (need_saturate(key
->unit
[unit
].ModeRGB
))
1186 rgb_saturate
= GL_TRUE
;
1188 rgb_saturate
= GL_FALSE
;
1191 alpha_saturate
= GL_FALSE
; /* saturate after alpha shift */
1192 else if (need_saturate(key
->unit
[unit
].ModeA
))
1193 alpha_saturate
= GL_TRUE
;
1195 alpha_saturate
= GL_FALSE
;
1197 /* If this is the very last calculation, emit direct to output reg:
1199 if (key
->separate_specular
||
1200 unit
!= p
->last_tex_stage
||
1203 dest
= get_temp( p
);
1205 dest
= make_ureg(PROGRAM_OUTPUT
, FRAG_RESULT_COLOR
);
1207 /* Emit the RGB and A combine ops
1209 if (key
->unit
[unit
].ModeRGB
== key
->unit
[unit
].ModeA
&&
1210 args_match(key
, unit
)) {
1211 out
= emit_combine( p
, dest
, WRITEMASK_XYZW
, rgb_saturate
,
1213 key
->unit
[unit
].NumArgsRGB
,
1214 key
->unit
[unit
].ModeRGB
,
1215 key
->unit
[unit
].OptRGB
);
1217 else if (key
->unit
[unit
].ModeRGB
== MODE_DOT3_RGBA_EXT
||
1218 key
->unit
[unit
].ModeRGB
== MODE_DOT3_RGBA
) {
1219 out
= emit_combine( p
, dest
, WRITEMASK_XYZW
, rgb_saturate
,
1221 key
->unit
[unit
].NumArgsRGB
,
1222 key
->unit
[unit
].ModeRGB
,
1223 key
->unit
[unit
].OptRGB
);
1226 /* Need to do something to stop from re-emitting identical
1227 * argument calculations here:
1229 out
= emit_combine( p
, dest
, WRITEMASK_XYZ
, rgb_saturate
,
1231 key
->unit
[unit
].NumArgsRGB
,
1232 key
->unit
[unit
].ModeRGB
,
1233 key
->unit
[unit
].OptRGB
);
1234 out
= emit_combine( p
, dest
, WRITEMASK_W
, alpha_saturate
,
1236 key
->unit
[unit
].NumArgsA
,
1237 key
->unit
[unit
].ModeA
,
1238 key
->unit
[unit
].OptA
);
1241 /* Deal with the final shift:
1243 if (alpha_shift
|| rgb_shift
) {
1245 GLboolean saturate
= GL_TRUE
; /* always saturate at this point */
1247 if (rgb_shift
== alpha_shift
) {
1248 shift
= register_scalar_const(p
, (GLfloat
)(1<<rgb_shift
));
1251 shift
= register_const4f(p
,
1252 (GLfloat
)(1<<rgb_shift
),
1253 (GLfloat
)(1<<rgb_shift
),
1254 (GLfloat
)(1<<rgb_shift
),
1255 (GLfloat
)(1<<alpha_shift
));
1257 return emit_arith( p
, OPCODE_MUL
, dest
, WRITEMASK_XYZW
,
1258 saturate
, out
, shift
, undef
);
1266 * Generate instruction for getting a texture source term.
1268 static void load_texture( struct texenv_fragment_program
*p
, GLuint unit
)
1270 if (is_undef(p
->src_texture
[unit
])) {
1271 const GLuint texTarget
= p
->state
->unit
[unit
].source_index
;
1272 struct ureg texcoord
;
1273 struct ureg tmp
= get_tex_temp( p
);
1275 if (is_undef(p
->texcoord_tex
[unit
])) {
1276 texcoord
= register_input(p
, FRAG_ATTRIB_TEX0
+unit
);
1279 /* might want to reuse this reg for tex output actually */
1280 texcoord
= p
->texcoord_tex
[unit
];
1283 /* TODO: Use D0_MASK_XY where possible.
1285 if (p
->state
->unit
[unit
].enabled
) {
1286 GLboolean shadow
= GL_FALSE
;
1288 if (p
->state
->unit
[unit
].shadow
) {
1289 p
->program
->Base
.ShadowSamplers
|= 1 << unit
;
1293 p
->src_texture
[unit
] = emit_texld( p
, OPCODE_TXP
,
1294 tmp
, WRITEMASK_XYZW
,
1295 unit
, texTarget
, shadow
,
1298 p
->program
->Base
.SamplersUsed
|= (1 << unit
);
1299 /* This identity mapping should already be in place
1300 * (see _mesa_init_program_struct()) but let's be safe.
1302 p
->program
->Base
.SamplerUnits
[unit
] = unit
;
1305 p
->src_texture
[unit
] = get_zero(p
);
1309 static GLboolean
load_texenv_source( struct texenv_fragment_program
*p
,
1310 GLuint src
, GLuint unit
)
1314 load_texture(p
, unit
);
1325 load_texture(p
, src
- SRC_TEXTURE0
);
1329 /* not a texture src - do nothing */
1338 * Generate instructions for loading all texture source terms.
1341 load_texunit_sources( struct texenv_fragment_program
*p
, GLuint unit
)
1343 const struct state_key
*key
= p
->state
;
1346 for (i
= 0; i
< key
->unit
[unit
].NumArgsRGB
; i
++) {
1347 load_texenv_source( p
, key
->unit
[unit
].OptRGB
[i
].Source
, unit
);
1350 for (i
= 0; i
< key
->unit
[unit
].NumArgsA
; i
++) {
1351 load_texenv_source( p
, key
->unit
[unit
].OptA
[i
].Source
, unit
);
1358 * Generate instructions for loading bump map textures.
1361 load_texunit_bumpmap( struct texenv_fragment_program
*p
, GLuint unit
)
1363 const struct state_key
*key
= p
->state
;
1364 GLuint bumpedUnitNr
= key
->unit
[unit
].OptRGB
[1].Source
- SRC_TEXTURE0
;
1365 struct ureg texcDst
, bumpMapRes
;
1366 struct ureg constdudvcolor
= register_const4f(p
, 0.0, 0.0, 0.0, 1.0);
1367 struct ureg texcSrc
= register_input(p
, FRAG_ATTRIB_TEX0
+ bumpedUnitNr
);
1368 struct ureg rotMat0
= register_param3( p
, STATE_INTERNAL
, STATE_ROT_MATRIX_0
, unit
);
1369 struct ureg rotMat1
= register_param3( p
, STATE_INTERNAL
, STATE_ROT_MATRIX_1
, unit
);
1371 load_texenv_source( p
, unit
+ SRC_TEXTURE0
, unit
);
1373 bumpMapRes
= get_source(p
, key
->unit
[unit
].OptRGB
[0].Source
, unit
);
1374 texcDst
= get_tex_temp( p
);
1375 p
->texcoord_tex
[bumpedUnitNr
] = texcDst
;
1377 /* Apply rot matrix and add coords to be available in next phase.
1378 * dest = (Arg0.xxxx * rotMat0 + Arg1) + (Arg0.yyyy * rotMat1)
1379 * note only 2 coords are affected the rest are left unchanged (mul by 0)
1381 emit_arith( p
, OPCODE_MAD
, texcDst
, WRITEMASK_XYZW
, 0,
1382 swizzle1(bumpMapRes
, SWIZZLE_X
), rotMat0
, texcSrc
);
1383 emit_arith( p
, OPCODE_MAD
, texcDst
, WRITEMASK_XYZW
, 0,
1384 swizzle1(bumpMapRes
, SWIZZLE_Y
), rotMat1
, texcDst
);
1386 /* Move 0,0,0,1 into bumpmap src if someone (crossbar) is foolish
1387 * enough to access this later, should optimize away.
1389 emit_arith( p
, OPCODE_MOV
, bumpMapRes
, WRITEMASK_XYZW
, 0,
1390 constdudvcolor
, undef
, undef
);
1396 * Generate a new fragment program which implements the context's
1397 * current texture env/combine mode.
1400 create_new_program(GLcontext
*ctx
, struct state_key
*key
,
1401 struct gl_fragment_program
*program
)
1403 struct prog_instruction instBuffer
[MAX_INSTRUCTIONS
];
1404 struct texenv_fragment_program p
;
1406 struct ureg cf
, out
;
1408 _mesa_memset(&p
, 0, sizeof(p
));
1410 p
.program
= program
;
1412 /* During code generation, use locally-allocated instruction buffer,
1413 * then alloc dynamic storage below.
1415 p
.program
->Base
.Instructions
= instBuffer
;
1416 p
.program
->Base
.Target
= GL_FRAGMENT_PROGRAM_ARB
;
1417 p
.program
->Base
.String
= NULL
;
1418 p
.program
->Base
.NumTexIndirections
= 1; /* is this right? */
1419 p
.program
->Base
.NumTexInstructions
= 0;
1420 p
.program
->Base
.NumAluInstructions
= 0;
1421 p
.program
->Base
.NumInstructions
= 0;
1422 p
.program
->Base
.NumTemporaries
= 0;
1423 p
.program
->Base
.NumParameters
= 0;
1424 p
.program
->Base
.NumAttributes
= 0;
1425 p
.program
->Base
.NumAddressRegs
= 0;
1426 p
.program
->Base
.Parameters
= _mesa_new_parameter_list();
1427 p
.program
->Base
.InputsRead
= 0x0;
1428 p
.program
->Base
.OutputsWritten
= 1 << FRAG_RESULT_COLOR
;
1430 for (unit
= 0; unit
< ctx
->Const
.MaxTextureUnits
; unit
++) {
1431 p
.src_texture
[unit
] = undef
;
1432 p
.texcoord_tex
[unit
] = undef
;
1435 p
.src_previous
= undef
;
1440 p
.last_tex_stage
= 0;
1441 release_temps(ctx
, &p
);
1443 if (key
->enabled_units
) {
1444 GLboolean needbumpstage
= GL_FALSE
;
1446 /* Zeroth pass - bump map textures first */
1447 for (unit
= 0; unit
< key
->nr_enabled_units
; unit
++)
1448 if (key
->unit
[unit
].enabled
&&
1449 key
->unit
[unit
].ModeRGB
== MODE_BUMP_ENVMAP_ATI
) {
1450 needbumpstage
= GL_TRUE
;
1451 load_texunit_bumpmap( &p
, unit
);
1454 p
.program
->Base
.NumTexIndirections
++;
1456 /* First pass - to support texture_env_crossbar, first identify
1457 * all referenced texture sources and emit texld instructions
1460 for (unit
= 0; unit
< key
->nr_enabled_units
; unit
++)
1461 if (key
->unit
[unit
].enabled
) {
1462 load_texunit_sources( &p
, unit
);
1463 p
.last_tex_stage
= unit
;
1466 /* Second pass - emit combine instructions to build final color:
1468 for (unit
= 0; unit
< key
->nr_enabled_units
; unit
++)
1469 if (key
->unit
[unit
].enabled
) {
1470 p
.src_previous
= emit_texenv( &p
, unit
);
1471 reserve_temp(&p
, p
.src_previous
); /* don't re-use this temp reg */
1472 release_temps(ctx
, &p
); /* release all temps */
1476 cf
= get_source( &p
, SRC_PREVIOUS
, 0 );
1477 out
= make_ureg( PROGRAM_OUTPUT
, FRAG_RESULT_COLOR
);
1479 if (key
->separate_specular
) {
1480 /* Emit specular add.
1482 struct ureg s
= register_input(&p
, FRAG_ATTRIB_COL1
);
1483 emit_arith( &p
, OPCODE_ADD
, out
, WRITEMASK_XYZ
, 0, cf
, s
, undef
);
1484 emit_arith( &p
, OPCODE_MOV
, out
, WRITEMASK_W
, 0, cf
, undef
, undef
);
1486 else if (_mesa_memcmp(&cf
, &out
, sizeof(cf
)) != 0) {
1487 /* Will wind up in here if no texture enabled or a couple of
1488 * other scenarios (GL_REPLACE for instance).
1490 emit_arith( &p
, OPCODE_MOV
, out
, WRITEMASK_XYZW
, 0, cf
, undef
, undef
);
1495 emit_arith( &p
, OPCODE_END
, undef
, WRITEMASK_XYZW
, 0, undef
, undef
, undef
);
1497 if (key
->fog_enabled
) {
1498 /* Pull fog mode from GLcontext, the value in the state key is
1499 * a reduced value and not what is expected in FogOption
1501 p
.program
->FogOption
= ctx
->Fog
.Mode
;
1502 p
.program
->Base
.InputsRead
|= FRAG_BIT_FOGC
;
1505 p
.program
->FogOption
= GL_NONE
;
1508 if (p
.program
->Base
.NumTexIndirections
> ctx
->Const
.FragmentProgram
.MaxTexIndirections
)
1509 program_error(&p
, "Exceeded max nr indirect texture lookups");
1511 if (p
.program
->Base
.NumTexInstructions
> ctx
->Const
.FragmentProgram
.MaxTexInstructions
)
1512 program_error(&p
, "Exceeded max TEX instructions");
1514 if (p
.program
->Base
.NumAluInstructions
> ctx
->Const
.FragmentProgram
.MaxAluInstructions
)
1515 program_error(&p
, "Exceeded max ALU instructions");
1517 ASSERT(p
.program
->Base
.NumInstructions
<= MAX_INSTRUCTIONS
);
1519 /* Allocate final instruction array */
1520 p
.program
->Base
.Instructions
1521 = _mesa_alloc_instructions(p
.program
->Base
.NumInstructions
);
1522 if (!p
.program
->Base
.Instructions
) {
1523 _mesa_error(ctx
, GL_OUT_OF_MEMORY
,
1524 "generating tex env program");
1527 _mesa_copy_instructions(p
.program
->Base
.Instructions
, instBuffer
,
1528 p
.program
->Base
.NumInstructions
);
1530 if (p
.program
->FogOption
) {
1531 _mesa_append_fog_code(ctx
, p
.program
);
1532 p
.program
->FogOption
= GL_NONE
;
1536 /* Notify driver the fragment program has (actually) changed.
1538 if (ctx
->Driver
.ProgramStringNotify
) {
1539 ctx
->Driver
.ProgramStringNotify( ctx
, GL_FRAGMENT_PROGRAM_ARB
,
1544 _mesa_print_program(&p
.program
->Base
);
1551 * Return a fragment program which implements the current
1552 * fixed-function texture, fog and color-sum operations.
1554 struct gl_fragment_program
*
1555 _mesa_get_fixed_func_fragment_program(GLcontext
*ctx
)
1557 struct gl_fragment_program
*prog
;
1558 struct state_key key
;
1561 keySize
= make_state_key(ctx
, &key
);
1563 prog
= (struct gl_fragment_program
*)
1564 _mesa_search_program_cache(ctx
->FragmentProgram
.Cache
,
1568 prog
= (struct gl_fragment_program
*)
1569 ctx
->Driver
.NewProgram(ctx
, GL_FRAGMENT_PROGRAM_ARB
, 0);
1571 create_new_program(ctx
, &key
, prog
);
1573 _mesa_program_cache_insert(ctx
, ctx
->FragmentProgram
.Cache
,
1574 &key
, keySize
, &prog
->Base
);