1 /**************************************************************************
3 * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
6 * Permission is hereby granted, free of charge, to any person obtaining a
7 * copy of this software and associated documentation files (the
8 * "Software"), to deal in the Software without restriction, including
9 * without limitation the rights to use, copy, modify, merge, publish,
10 * distribute, sub license, and/or sell copies of the Software, and to
11 * permit persons to whom the Software is furnished to do so, subject to
12 * the following conditions:
14 * The above copyright notice and this permission notice (including the
15 * next paragraph) shall be included in all copies or substantial portions
18 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
19 * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
20 * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
21 * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
22 * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
23 * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
24 * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
26 **************************************************************************/
29 * \file ffvertex_prog.c
31 * Create a vertex program to execute the current fixed function T&L pipeline.
32 * \author Keith Whitwell
36 #include "main/glheader.h"
37 #include "main/mtypes.h"
38 #include "main/macros.h"
39 #include "main/mfeatures.h"
40 #include "main/enums.h"
41 #include "main/ffvertex_prog.h"
42 #include "program/program.h"
43 #include "program/prog_cache.h"
44 #include "program/prog_instruction.h"
45 #include "program/prog_parameter.h"
46 #include "program/prog_print.h"
47 #include "program/prog_statevars.h"
50 /** Max of number of lights and texture coord units */
51 #define NUM_UNITS MAX2(MAX_TEXTURE_COORD_UNITS, MAX_LIGHTS)
54 unsigned light_color_material_mask
:12;
55 unsigned light_global_enabled
:1;
56 unsigned light_local_viewer
:1;
57 unsigned light_twoside
:1;
58 unsigned material_shininess_is_zero
:1;
59 unsigned need_eye_coords
:1;
61 unsigned rescale_normals
:1;
63 unsigned fog_source_is_depth
:1;
64 unsigned fog_distance_mode
:2;
65 unsigned separate_specular
:1;
66 unsigned point_attenuated
:1;
67 unsigned point_array
:1;
68 unsigned texture_enabled_global
:1;
69 unsigned fragprog_inputs_read
:12;
71 GLbitfield64 varying_vp_inputs
;
74 unsigned light_enabled
:1;
75 unsigned light_eyepos3_is_zero
:1;
76 unsigned light_spotcutoff_is_180
:1;
77 unsigned light_attenuated
:1;
78 unsigned texunit_really_enabled
:1;
79 unsigned texmat_enabled
:1;
80 unsigned coord_replace
:1;
81 unsigned texgen_enabled
:4;
82 unsigned texgen_mode0
:4;
83 unsigned texgen_mode1
:4;
84 unsigned texgen_mode2
:4;
85 unsigned texgen_mode3
:4;
91 #define TXG_OBJ_LINEAR 1
92 #define TXG_EYE_LINEAR 2
93 #define TXG_SPHERE_MAP 3
94 #define TXG_REFLECTION_MAP 4
95 #define TXG_NORMAL_MAP 5
97 static GLuint
translate_texgen( GLboolean enabled
, GLenum mode
)
103 case GL_OBJECT_LINEAR
: return TXG_OBJ_LINEAR
;
104 case GL_EYE_LINEAR
: return TXG_EYE_LINEAR
;
105 case GL_SPHERE_MAP
: return TXG_SPHERE_MAP
;
106 case GL_REFLECTION_MAP_NV
: return TXG_REFLECTION_MAP
;
107 case GL_NORMAL_MAP_NV
: return TXG_NORMAL_MAP
;
108 default: return TXG_NONE
;
112 #define FDM_EYE_RADIAL 0
113 #define FDM_EYE_PLANE 1
114 #define FDM_EYE_PLANE_ABS 2
116 static GLuint
translate_fog_distance_mode( GLenum mode
)
119 case GL_EYE_RADIAL_NV
:
120 return FDM_EYE_RADIAL
;
122 return FDM_EYE_PLANE
;
123 default: /* shouldn't happen; fall through to a sensible default */
124 case GL_EYE_PLANE_ABSOLUTE_NV
:
125 return FDM_EYE_PLANE_ABS
;
129 static GLboolean
check_active_shininess( struct gl_context
*ctx
,
130 const struct state_key
*key
,
133 GLuint attr
= MAT_ATTRIB_FRONT_SHININESS
+ side
;
135 if ((key
->varying_vp_inputs
& VERT_BIT_COLOR0
) &&
136 (key
->light_color_material_mask
& (1 << attr
)))
139 if (key
->varying_vp_inputs
& VERT_ATTRIB_GENERIC(attr
))
142 if (ctx
->Light
.Material
.Attrib
[attr
][0] != 0.0F
)
149 static void make_state_key( struct gl_context
*ctx
, struct state_key
*key
)
151 const struct gl_fragment_program
*fp
;
154 memset(key
, 0, sizeof(struct state_key
));
155 fp
= ctx
->FragmentProgram
._Current
;
157 /* This now relies on texenvprogram.c being active:
161 key
->need_eye_coords
= ctx
->_NeedEyeCoords
;
163 key
->fragprog_inputs_read
= fp
->Base
.InputsRead
;
164 key
->varying_vp_inputs
= ctx
->varying_vp_inputs
;
166 if (ctx
->RenderMode
== GL_FEEDBACK
) {
167 /* make sure the vertprog emits color and tex0 */
168 key
->fragprog_inputs_read
|= (FRAG_BIT_COL0
| FRAG_BIT_TEX0
);
171 key
->separate_specular
= (ctx
->Light
.Model
.ColorControl
==
172 GL_SEPARATE_SPECULAR_COLOR
);
174 if (ctx
->Light
.Enabled
) {
175 key
->light_global_enabled
= 1;
177 if (ctx
->Light
.Model
.LocalViewer
)
178 key
->light_local_viewer
= 1;
180 if (ctx
->Light
.Model
.TwoSide
)
181 key
->light_twoside
= 1;
183 if (ctx
->Light
.ColorMaterialEnabled
) {
184 key
->light_color_material_mask
= ctx
->Light
._ColorMaterialBitmask
;
187 for (i
= 0; i
< MAX_LIGHTS
; i
++) {
188 struct gl_light
*light
= &ctx
->Light
.Light
[i
];
190 if (light
->Enabled
) {
191 key
->unit
[i
].light_enabled
= 1;
193 if (light
->EyePosition
[3] == 0.0)
194 key
->unit
[i
].light_eyepos3_is_zero
= 1;
196 if (light
->SpotCutoff
== 180.0)
197 key
->unit
[i
].light_spotcutoff_is_180
= 1;
199 if (light
->ConstantAttenuation
!= 1.0 ||
200 light
->LinearAttenuation
!= 0.0 ||
201 light
->QuadraticAttenuation
!= 0.0)
202 key
->unit
[i
].light_attenuated
= 1;
206 if (check_active_shininess(ctx
, key
, 0)) {
207 key
->material_shininess_is_zero
= 0;
209 else if (key
->light_twoside
&&
210 check_active_shininess(ctx
, key
, 1)) {
211 key
->material_shininess_is_zero
= 0;
214 key
->material_shininess_is_zero
= 1;
218 if (ctx
->Transform
.Normalize
)
221 if (ctx
->Transform
.RescaleNormals
)
222 key
->rescale_normals
= 1;
224 if (ctx
->Fog
.FogCoordinateSource
== GL_FRAGMENT_DEPTH_EXT
) {
225 key
->fog_source_is_depth
= 1;
226 key
->fog_distance_mode
= translate_fog_distance_mode(ctx
->Fog
.FogDistanceMode
);
229 if (ctx
->Point
._Attenuated
)
230 key
->point_attenuated
= 1;
232 if (ctx
->Array
.ArrayObj
->VertexAttrib
[VERT_ATTRIB_POINT_SIZE
].Enabled
)
233 key
->point_array
= 1;
235 if (ctx
->Texture
._TexGenEnabled
||
236 ctx
->Texture
._TexMatEnabled
||
237 ctx
->Texture
._EnabledUnits
)
238 key
->texture_enabled_global
= 1;
240 for (i
= 0; i
< MAX_TEXTURE_COORD_UNITS
; i
++) {
241 struct gl_texture_unit
*texUnit
= &ctx
->Texture
.Unit
[i
];
243 if (texUnit
->_ReallyEnabled
)
244 key
->unit
[i
].texunit_really_enabled
= 1;
246 if (ctx
->Point
.PointSprite
)
247 if (ctx
->Point
.CoordReplace
[i
])
248 key
->unit
[i
].coord_replace
= 1;
250 if (ctx
->Texture
._TexMatEnabled
& ENABLE_TEXMAT(i
))
251 key
->unit
[i
].texmat_enabled
= 1;
253 if (texUnit
->TexGenEnabled
) {
254 key
->unit
[i
].texgen_enabled
= 1;
256 key
->unit
[i
].texgen_mode0
=
257 translate_texgen( texUnit
->TexGenEnabled
& (1<<0),
258 texUnit
->GenS
.Mode
);
259 key
->unit
[i
].texgen_mode1
=
260 translate_texgen( texUnit
->TexGenEnabled
& (1<<1),
261 texUnit
->GenT
.Mode
);
262 key
->unit
[i
].texgen_mode2
=
263 translate_texgen( texUnit
->TexGenEnabled
& (1<<2),
264 texUnit
->GenR
.Mode
);
265 key
->unit
[i
].texgen_mode3
=
266 translate_texgen( texUnit
->TexGenEnabled
& (1<<3),
267 texUnit
->GenQ
.Mode
);
274 /* Very useful debugging tool - produces annotated listing of
275 * generated program with line/function references for each
276 * instruction back into this file:
281 /* Use uregs to represent registers internally, translate to Mesa's
282 * expected formats on emit.
284 * NOTE: These are passed by value extensively in this file rather
285 * than as usual by pointer reference. If this disturbs you, try
286 * remembering they are just 32bits in size.
288 * GCC is smart enough to deal with these dword-sized structures in
289 * much the same way as if I had defined them as dwords and was using
290 * macros to access and set the fields. This is much nicer and easier
295 GLint idx
:9; /* relative addressing may be negative */
296 /* sizeof(idx) should == sizeof(prog_src_reg::Index) */
304 const struct state_key
*state
;
305 struct gl_vertex_program
*program
;
306 GLint max_inst
; /** number of instructions allocated for program */
307 GLboolean mvp_with_dp4
;
310 GLuint temp_reserved
;
312 struct ureg eye_position
;
313 struct ureg eye_position_z
;
314 struct ureg eye_position_normalized
;
315 struct ureg transformed_normal
;
316 struct ureg identity
;
319 GLuint color_materials
;
323 static const struct ureg undef
= {
341 static struct ureg
make_ureg(GLuint file
, GLint idx
)
347 reg
.swz
= SWIZZLE_NOOP
;
354 static struct ureg
negate( struct ureg reg
)
361 static struct ureg
swizzle( struct ureg reg
, int x
, int y
, int z
, int w
)
363 reg
.swz
= MAKE_SWIZZLE4(GET_SWZ(reg
.swz
, x
),
366 GET_SWZ(reg
.swz
, w
));
371 static struct ureg
swizzle1( struct ureg reg
, int x
)
373 return swizzle(reg
, x
, x
, x
, x
);
377 static struct ureg
get_temp( struct tnl_program
*p
)
379 int bit
= ffs( ~p
->temp_in_use
);
381 _mesa_problem(NULL
, "%s: out of temporaries\n", __FILE__
);
385 if ((GLuint
) bit
> p
->program
->Base
.NumTemporaries
)
386 p
->program
->Base
.NumTemporaries
= bit
;
388 p
->temp_in_use
|= 1<<(bit
-1);
389 return make_ureg(PROGRAM_TEMPORARY
, bit
-1);
393 static struct ureg
reserve_temp( struct tnl_program
*p
)
395 struct ureg temp
= get_temp( p
);
396 p
->temp_reserved
|= 1<<temp
.idx
;
401 static void release_temp( struct tnl_program
*p
, struct ureg reg
)
403 if (reg
.file
== PROGRAM_TEMPORARY
) {
404 p
->temp_in_use
&= ~(1<<reg
.idx
);
405 p
->temp_in_use
|= p
->temp_reserved
; /* can't release reserved temps */
409 static void release_temps( struct tnl_program
*p
)
411 p
->temp_in_use
= p
->temp_reserved
;
415 static struct ureg
register_param5(struct tnl_program
*p
,
422 gl_state_index tokens
[STATE_LENGTH
];
429 idx
= _mesa_add_state_reference( p
->program
->Base
.Parameters
, tokens
);
430 return make_ureg(PROGRAM_STATE_VAR
, idx
);
434 #define register_param1(p,s0) register_param5(p,s0,0,0,0,0)
435 #define register_param2(p,s0,s1) register_param5(p,s0,s1,0,0,0)
436 #define register_param3(p,s0,s1,s2) register_param5(p,s0,s1,s2,0,0)
437 #define register_param4(p,s0,s1,s2,s3) register_param5(p,s0,s1,s2,s3,0)
442 * \param input one of VERT_ATTRIB_x tokens.
444 static struct ureg
register_input( struct tnl_program
*p
, GLuint input
)
446 assert(input
< VERT_ATTRIB_MAX
);
448 if (p
->state
->varying_vp_inputs
& VERT_BIT(input
)) {
449 p
->program
->Base
.InputsRead
|= VERT_BIT(input
);
450 return make_ureg(PROGRAM_INPUT
, input
);
453 return register_param3( p
, STATE_INTERNAL
, STATE_CURRENT_ATTRIB
, input
);
459 * \param input one of VERT_RESULT_x tokens.
461 static struct ureg
register_output( struct tnl_program
*p
, GLuint output
)
463 p
->program
->Base
.OutputsWritten
|= BITFIELD64_BIT(output
);
464 return make_ureg(PROGRAM_OUTPUT
, output
);
468 static struct ureg
register_const4f( struct tnl_program
*p
,
474 gl_constant_value values
[4];
481 idx
= _mesa_add_unnamed_constant( p
->program
->Base
.Parameters
, values
, 4,
483 ASSERT(swizzle
== SWIZZLE_NOOP
);
484 return make_ureg(PROGRAM_CONSTANT
, idx
);
487 #define register_const1f(p, s0) register_const4f(p, s0, 0, 0, 1)
488 #define register_scalar_const(p, s0) register_const4f(p, s0, s0, s0, s0)
489 #define register_const2f(p, s0, s1) register_const4f(p, s0, s1, 0, 1)
490 #define register_const3f(p, s0, s1, s2) register_const4f(p, s0, s1, s2, 1)
492 static GLboolean
is_undef( struct ureg reg
)
494 return reg
.file
== PROGRAM_UNDEFINED
;
498 static struct ureg
get_identity_param( struct tnl_program
*p
)
500 if (is_undef(p
->identity
))
501 p
->identity
= register_const4f(p
, 0,0,0,1);
506 static void register_matrix_param5( struct tnl_program
*p
,
507 GLint s0
, /* modelview, projection, etc */
508 GLint s1
, /* texture matrix number */
509 GLint s2
, /* first row */
510 GLint s3
, /* last row */
511 GLint s4
, /* inverse, transpose, etc */
512 struct ureg
*matrix
)
516 /* This is a bit sad as the support is there to pull the whole
517 * matrix out in one go:
519 for (i
= 0; i
<= s3
- s2
; i
++)
520 matrix
[i
] = register_param5( p
, s0
, s1
, i
, i
, s4
);
524 static void emit_arg( struct prog_src_register
*src
,
527 src
->File
= reg
.file
;
528 src
->Index
= reg
.idx
;
529 src
->Swizzle
= reg
.swz
;
530 src
->Negate
= reg
.negate
? NEGATE_XYZW
: NEGATE_NONE
;
533 /* Check that bitfield sizes aren't exceeded */
534 ASSERT(src
->Index
== reg
.idx
);
538 static void emit_dst( struct prog_dst_register
*dst
,
539 struct ureg reg
, GLuint mask
)
541 dst
->File
= reg
.file
;
542 dst
->Index
= reg
.idx
;
543 /* allow zero as a shorthand for xyzw */
544 dst
->WriteMask
= mask
? mask
: WRITEMASK_XYZW
;
545 dst
->CondMask
= COND_TR
; /* always pass cond test */
546 dst
->CondSwizzle
= SWIZZLE_NOOP
;
547 /* Check that bitfield sizes aren't exceeded */
548 ASSERT(dst
->Index
== reg
.idx
);
552 static void debug_insn( struct prog_instruction
*inst
, const char *fn
,
556 static const char *last_fn
;
563 printf("%d:\t", line
);
564 _mesa_print_instruction(inst
);
569 static void emit_op3fn(struct tnl_program
*p
,
580 struct prog_instruction
*inst
;
582 assert((GLint
) p
->program
->Base
.NumInstructions
<= p
->max_inst
);
584 if (p
->program
->Base
.NumInstructions
== p
->max_inst
) {
585 /* need to extend the program's instruction array */
586 struct prog_instruction
*newInst
;
588 /* double the size */
591 newInst
= _mesa_alloc_instructions(p
->max_inst
);
593 _mesa_error(NULL
, GL_OUT_OF_MEMORY
, "vertex program build");
597 _mesa_copy_instructions(newInst
,
598 p
->program
->Base
.Instructions
,
599 p
->program
->Base
.NumInstructions
);
601 _mesa_free_instructions(p
->program
->Base
.Instructions
,
602 p
->program
->Base
.NumInstructions
);
604 p
->program
->Base
.Instructions
= newInst
;
607 nr
= p
->program
->Base
.NumInstructions
++;
609 inst
= &p
->program
->Base
.Instructions
[nr
];
610 inst
->Opcode
= (enum prog_opcode
) op
;
613 emit_arg( &inst
->SrcReg
[0], src0
);
614 emit_arg( &inst
->SrcReg
[1], src1
);
615 emit_arg( &inst
->SrcReg
[2], src2
);
617 emit_dst( &inst
->DstReg
, dest
, mask
);
619 debug_insn(inst
, fn
, line
);
623 #define emit_op3(p, op, dst, mask, src0, src1, src2) \
624 emit_op3fn(p, op, dst, mask, src0, src1, src2, __FUNCTION__, __LINE__)
626 #define emit_op2(p, op, dst, mask, src0, src1) \
627 emit_op3fn(p, op, dst, mask, src0, src1, undef, __FUNCTION__, __LINE__)
629 #define emit_op1(p, op, dst, mask, src0) \
630 emit_op3fn(p, op, dst, mask, src0, undef, undef, __FUNCTION__, __LINE__)
633 static struct ureg
make_temp( struct tnl_program
*p
, struct ureg reg
)
635 if (reg
.file
== PROGRAM_TEMPORARY
&&
636 !(p
->temp_reserved
& (1<<reg
.idx
)))
639 struct ureg temp
= get_temp(p
);
640 emit_op1(p
, OPCODE_MOV
, temp
, 0, reg
);
646 /* Currently no tracking performed of input/output/register size or
647 * active elements. Could be used to reduce these operations, as
648 * could the matrix type.
650 static void emit_matrix_transform_vec4( struct tnl_program
*p
,
652 const struct ureg
*mat
,
655 emit_op2(p
, OPCODE_DP4
, dest
, WRITEMASK_X
, src
, mat
[0]);
656 emit_op2(p
, OPCODE_DP4
, dest
, WRITEMASK_Y
, src
, mat
[1]);
657 emit_op2(p
, OPCODE_DP4
, dest
, WRITEMASK_Z
, src
, mat
[2]);
658 emit_op2(p
, OPCODE_DP4
, dest
, WRITEMASK_W
, src
, mat
[3]);
662 /* This version is much easier to implement if writemasks are not
663 * supported natively on the target or (like SSE), the target doesn't
664 * have a clean/obvious dotproduct implementation.
666 static void emit_transpose_matrix_transform_vec4( struct tnl_program
*p
,
668 const struct ureg
*mat
,
673 if (dest
.file
!= PROGRAM_TEMPORARY
)
678 emit_op2(p
, OPCODE_MUL
, tmp
, 0, swizzle1(src
,X
), mat
[0]);
679 emit_op3(p
, OPCODE_MAD
, tmp
, 0, swizzle1(src
,Y
), mat
[1], tmp
);
680 emit_op3(p
, OPCODE_MAD
, tmp
, 0, swizzle1(src
,Z
), mat
[2], tmp
);
681 emit_op3(p
, OPCODE_MAD
, dest
, 0, swizzle1(src
,W
), mat
[3], tmp
);
683 if (dest
.file
!= PROGRAM_TEMPORARY
)
684 release_temp(p
, tmp
);
688 static void emit_matrix_transform_vec3( struct tnl_program
*p
,
690 const struct ureg
*mat
,
693 emit_op2(p
, OPCODE_DP3
, dest
, WRITEMASK_X
, src
, mat
[0]);
694 emit_op2(p
, OPCODE_DP3
, dest
, WRITEMASK_Y
, src
, mat
[1]);
695 emit_op2(p
, OPCODE_DP3
, dest
, WRITEMASK_Z
, src
, mat
[2]);
699 static void emit_normalize_vec3( struct tnl_program
*p
,
704 /* XXX use this when drivers are ready for NRM3 */
705 emit_op1(p
, OPCODE_NRM3
, dest
, WRITEMASK_XYZ
, src
);
707 struct ureg tmp
= get_temp(p
);
708 emit_op2(p
, OPCODE_DP3
, tmp
, WRITEMASK_X
, src
, src
);
709 emit_op1(p
, OPCODE_RSQ
, tmp
, WRITEMASK_X
, tmp
);
710 emit_op2(p
, OPCODE_MUL
, dest
, 0, src
, swizzle1(tmp
, X
));
711 release_temp(p
, tmp
);
716 static void emit_passthrough( struct tnl_program
*p
,
720 struct ureg out
= register_output(p
, output
);
721 emit_op1(p
, OPCODE_MOV
, out
, 0, register_input(p
, input
));
725 static struct ureg
get_eye_position( struct tnl_program
*p
)
727 if (is_undef(p
->eye_position
)) {
728 struct ureg pos
= register_input( p
, VERT_ATTRIB_POS
);
729 struct ureg modelview
[4];
731 p
->eye_position
= reserve_temp(p
);
733 if (p
->mvp_with_dp4
) {
734 register_matrix_param5( p
, STATE_MODELVIEW_MATRIX
, 0, 0, 3,
737 emit_matrix_transform_vec4(p
, p
->eye_position
, modelview
, pos
);
740 register_matrix_param5( p
, STATE_MODELVIEW_MATRIX
, 0, 0, 3,
741 STATE_MATRIX_TRANSPOSE
, modelview
);
743 emit_transpose_matrix_transform_vec4(p
, p
->eye_position
, modelview
, pos
);
747 return p
->eye_position
;
751 static struct ureg
get_eye_position_z( struct tnl_program
*p
)
753 if (!is_undef(p
->eye_position
))
754 return swizzle1(p
->eye_position
, Z
);
756 if (is_undef(p
->eye_position_z
)) {
757 struct ureg pos
= register_input( p
, VERT_ATTRIB_POS
);
758 struct ureg modelview
[4];
760 p
->eye_position_z
= reserve_temp(p
);
762 register_matrix_param5( p
, STATE_MODELVIEW_MATRIX
, 0, 0, 3,
765 emit_op2(p
, OPCODE_DP4
, p
->eye_position_z
, 0, pos
, modelview
[2]);
768 return p
->eye_position_z
;
772 static struct ureg
get_eye_position_normalized( struct tnl_program
*p
)
774 if (is_undef(p
->eye_position_normalized
)) {
775 struct ureg eye
= get_eye_position(p
);
776 p
->eye_position_normalized
= reserve_temp(p
);
777 emit_normalize_vec3(p
, p
->eye_position_normalized
, eye
);
780 return p
->eye_position_normalized
;
784 static struct ureg
get_transformed_normal( struct tnl_program
*p
)
786 if (is_undef(p
->transformed_normal
) &&
787 !p
->state
->need_eye_coords
&&
788 !p
->state
->normalize
&&
789 !(p
->state
->need_eye_coords
== p
->state
->rescale_normals
))
791 p
->transformed_normal
= register_input(p
, VERT_ATTRIB_NORMAL
);
793 else if (is_undef(p
->transformed_normal
))
795 struct ureg normal
= register_input(p
, VERT_ATTRIB_NORMAL
);
796 struct ureg mvinv
[3];
797 struct ureg transformed_normal
= reserve_temp(p
);
799 if (p
->state
->need_eye_coords
) {
800 register_matrix_param5( p
, STATE_MODELVIEW_MATRIX
, 0, 0, 2,
801 STATE_MATRIX_INVTRANS
, mvinv
);
803 /* Transform to eye space:
805 emit_matrix_transform_vec3( p
, transformed_normal
, mvinv
, normal
);
806 normal
= transformed_normal
;
809 /* Normalize/Rescale:
811 if (p
->state
->normalize
) {
812 emit_normalize_vec3( p
, transformed_normal
, normal
);
813 normal
= transformed_normal
;
815 else if (p
->state
->need_eye_coords
== p
->state
->rescale_normals
) {
816 /* This is already adjusted for eye/non-eye rendering:
818 struct ureg rescale
= register_param2(p
, STATE_INTERNAL
,
821 emit_op2( p
, OPCODE_MUL
, transformed_normal
, 0, normal
, rescale
);
822 normal
= transformed_normal
;
825 assert(normal
.file
== PROGRAM_TEMPORARY
);
826 p
->transformed_normal
= normal
;
829 return p
->transformed_normal
;
833 static void build_hpos( struct tnl_program
*p
)
835 struct ureg pos
= register_input( p
, VERT_ATTRIB_POS
);
836 struct ureg hpos
= register_output( p
, VERT_RESULT_HPOS
);
839 if (p
->mvp_with_dp4
) {
840 register_matrix_param5( p
, STATE_MVP_MATRIX
, 0, 0, 3,
842 emit_matrix_transform_vec4( p
, hpos
, mvp
, pos
);
845 register_matrix_param5( p
, STATE_MVP_MATRIX
, 0, 0, 3,
846 STATE_MATRIX_TRANSPOSE
, mvp
);
847 emit_transpose_matrix_transform_vec4( p
, hpos
, mvp
, pos
);
852 static GLuint
material_attrib( GLuint side
, GLuint property
)
854 return (property
- STATE_AMBIENT
) * 2 + side
;
859 * Get a bitmask of which material values vary on a per-vertex basis.
861 static void set_material_flags( struct tnl_program
*p
)
863 p
->color_materials
= 0;
866 if (p
->state
->varying_vp_inputs
& VERT_BIT_COLOR0
) {
868 p
->color_materials
= p
->state
->light_color_material_mask
;
871 p
->materials
|= (p
->state
->varying_vp_inputs
>> VERT_ATTRIB_GENERIC0
);
875 static struct ureg
get_material( struct tnl_program
*p
, GLuint side
,
878 GLuint attrib
= material_attrib(side
, property
);
880 if (p
->color_materials
& (1<<attrib
))
881 return register_input(p
, VERT_ATTRIB_COLOR0
);
882 else if (p
->materials
& (1<<attrib
)) {
883 /* Put material values in the GENERIC slots -- they are not used
884 * for anything in fixed function mode.
886 return register_input( p
, attrib
+ VERT_ATTRIB_GENERIC0
);
889 return register_param3( p
, STATE_MATERIAL
, side
, property
);
892 #define SCENE_COLOR_BITS(side) (( MAT_BIT_FRONT_EMISSION | \
893 MAT_BIT_FRONT_AMBIENT | \
894 MAT_BIT_FRONT_DIFFUSE) << (side))
898 * Either return a precalculated constant value or emit code to
899 * calculate these values dynamically in the case where material calls
900 * are present between begin/end pairs.
902 * Probably want to shift this to the program compilation phase - if
903 * we always emitted the calculation here, a smart compiler could
904 * detect that it was constant (given a certain set of inputs), and
905 * lift it out of the main loop. That way the programs created here
906 * would be independent of the vertex_buffer details.
908 static struct ureg
get_scenecolor( struct tnl_program
*p
, GLuint side
)
910 if (p
->materials
& SCENE_COLOR_BITS(side
)) {
911 struct ureg lm_ambient
= register_param1(p
, STATE_LIGHTMODEL_AMBIENT
);
912 struct ureg material_emission
= get_material(p
, side
, STATE_EMISSION
);
913 struct ureg material_ambient
= get_material(p
, side
, STATE_AMBIENT
);
914 struct ureg material_diffuse
= get_material(p
, side
, STATE_DIFFUSE
);
915 struct ureg tmp
= make_temp(p
, material_diffuse
);
916 emit_op3(p
, OPCODE_MAD
, tmp
, WRITEMASK_XYZ
, lm_ambient
,
917 material_ambient
, material_emission
);
921 return register_param2( p
, STATE_LIGHTMODEL_SCENECOLOR
, side
);
925 static struct ureg
get_lightprod( struct tnl_program
*p
, GLuint light
,
926 GLuint side
, GLuint property
)
928 GLuint attrib
= material_attrib(side
, property
);
929 if (p
->materials
& (1<<attrib
)) {
930 struct ureg light_value
=
931 register_param3(p
, STATE_LIGHT
, light
, property
);
932 struct ureg material_value
= get_material(p
, side
, property
);
933 struct ureg tmp
= get_temp(p
);
934 emit_op2(p
, OPCODE_MUL
, tmp
, 0, light_value
, material_value
);
938 return register_param4(p
, STATE_LIGHTPROD
, light
, side
, property
);
942 static struct ureg
calculate_light_attenuation( struct tnl_program
*p
,
947 struct ureg attenuation
= register_param3(p
, STATE_LIGHT
, i
,
949 struct ureg att
= undef
;
951 /* Calculate spot attenuation:
953 if (!p
->state
->unit
[i
].light_spotcutoff_is_180
) {
954 struct ureg spot_dir_norm
= register_param3(p
, STATE_INTERNAL
,
955 STATE_LIGHT_SPOT_DIR_NORMALIZED
, i
);
956 struct ureg spot
= get_temp(p
);
957 struct ureg slt
= get_temp(p
);
961 emit_op2(p
, OPCODE_DP3
, spot
, 0, negate(VPpli
), spot_dir_norm
);
962 emit_op2(p
, OPCODE_SLT
, slt
, 0, swizzle1(spot_dir_norm
,W
), spot
);
963 emit_op2(p
, OPCODE_POW
, spot
, 0, spot
, swizzle1(attenuation
, W
));
964 emit_op2(p
, OPCODE_MUL
, att
, 0, slt
, spot
);
966 release_temp(p
, spot
);
967 release_temp(p
, slt
);
970 /* Calculate distance attenuation(See formula (2.4) at glspec 2.1 page 62):
972 * Skip the calucation when _dist_ is undefined(light_eyepos3_is_zero)
974 if (p
->state
->unit
[i
].light_attenuated
&& !is_undef(dist
)) {
978 emit_op1(p
, OPCODE_RCP
, dist
, WRITEMASK_YZ
, dist
);
980 emit_op2(p
, OPCODE_MUL
, dist
, WRITEMASK_XZ
, dist
, swizzle1(dist
,Y
));
982 emit_op2(p
, OPCODE_DP3
, dist
, 0, attenuation
, dist
);
984 if (!p
->state
->unit
[i
].light_spotcutoff_is_180
) {
986 emit_op1(p
, OPCODE_RCP
, dist
, 0, dist
);
987 /* spot-atten * dist-atten */
988 emit_op2(p
, OPCODE_MUL
, att
, 0, dist
, att
);
992 emit_op1(p
, OPCODE_RCP
, att
, 0, dist
);
1002 * lit.y = MAX(0, dots.x)
1003 * lit.z = SLT(0, dots.x)
1005 static void emit_degenerate_lit( struct tnl_program
*p
,
1009 struct ureg id
= get_identity_param(p
); /* id = {0,0,0,1} */
1011 /* Note that lit.x & lit.w will not be examined. Note also that
1012 * dots.xyzw == dots.xxxx.
1015 /* MAX lit, id, dots;
1017 emit_op2(p
, OPCODE_MAX
, lit
, WRITEMASK_XYZW
, id
, dots
);
1019 /* result[2] = (in > 0 ? 1 : 0)
1020 * SLT lit.z, id.z, dots; # lit.z = (0 < dots.z) ? 1 : 0
1022 emit_op2(p
, OPCODE_SLT
, lit
, WRITEMASK_Z
, swizzle1(id
,Z
), dots
);
1026 /* Need to add some addtional parameters to allow lighting in object
1027 * space - STATE_SPOT_DIRECTION and STATE_HALF_VECTOR implicitly assume eye
1030 static void build_lighting( struct tnl_program
*p
)
1032 const GLboolean twoside
= p
->state
->light_twoside
;
1033 const GLboolean separate
= p
->state
->separate_specular
;
1034 GLuint nr_lights
= 0, count
= 0;
1035 struct ureg normal
= get_transformed_normal(p
);
1036 struct ureg lit
= get_temp(p
);
1037 struct ureg dots
= get_temp(p
);
1038 struct ureg _col0
= undef
, _col1
= undef
;
1039 struct ureg _bfc0
= undef
, _bfc1
= undef
;
1044 * dots.x = dot(normal, VPpli)
1045 * dots.y = dot(normal, halfAngle)
1046 * dots.z = back.shininess
1047 * dots.w = front.shininess
1050 for (i
= 0; i
< MAX_LIGHTS
; i
++)
1051 if (p
->state
->unit
[i
].light_enabled
)
1054 set_material_flags(p
);
1057 if (!p
->state
->material_shininess_is_zero
) {
1058 struct ureg shininess
= get_material(p
, 0, STATE_SHININESS
);
1059 emit_op1(p
, OPCODE_MOV
, dots
, WRITEMASK_W
, swizzle1(shininess
,X
));
1060 release_temp(p
, shininess
);
1063 _col0
= make_temp(p
, get_scenecolor(p
, 0));
1065 _col1
= make_temp(p
, get_identity_param(p
));
1071 if (!p
->state
->material_shininess_is_zero
) {
1072 /* Note that we negate the back-face specular exponent here.
1073 * The negation will be un-done later in the back-face code below.
1075 struct ureg shininess
= get_material(p
, 1, STATE_SHININESS
);
1076 emit_op1(p
, OPCODE_MOV
, dots
, WRITEMASK_Z
,
1077 negate(swizzle1(shininess
,X
)));
1078 release_temp(p
, shininess
);
1081 _bfc0
= make_temp(p
, get_scenecolor(p
, 1));
1083 _bfc1
= make_temp(p
, get_identity_param(p
));
1088 /* If no lights, still need to emit the scenecolor.
1091 struct ureg res0
= register_output( p
, VERT_RESULT_COL0
);
1092 emit_op1(p
, OPCODE_MOV
, res0
, 0, _col0
);
1096 struct ureg res1
= register_output( p
, VERT_RESULT_COL1
);
1097 emit_op1(p
, OPCODE_MOV
, res1
, 0, _col1
);
1101 struct ureg res0
= register_output( p
, VERT_RESULT_BFC0
);
1102 emit_op1(p
, OPCODE_MOV
, res0
, 0, _bfc0
);
1105 if (twoside
&& separate
) {
1106 struct ureg res1
= register_output( p
, VERT_RESULT_BFC1
);
1107 emit_op1(p
, OPCODE_MOV
, res1
, 0, _bfc1
);
1110 if (nr_lights
== 0) {
1115 for (i
= 0; i
< MAX_LIGHTS
; i
++) {
1116 if (p
->state
->unit
[i
].light_enabled
) {
1117 struct ureg half
= undef
;
1118 struct ureg att
= undef
, VPpli
= undef
;
1119 struct ureg dist
= undef
;
1122 if (p
->state
->unit
[i
].light_eyepos3_is_zero
) {
1123 VPpli
= register_param3(p
, STATE_INTERNAL
,
1124 STATE_LIGHT_POSITION_NORMALIZED
, i
);
1126 struct ureg Ppli
= register_param3(p
, STATE_INTERNAL
,
1127 STATE_LIGHT_POSITION
, i
);
1128 struct ureg V
= get_eye_position(p
);
1130 VPpli
= get_temp(p
);
1133 /* Calculate VPpli vector
1135 emit_op2(p
, OPCODE_SUB
, VPpli
, 0, Ppli
, V
);
1137 /* Normalize VPpli. The dist value also used in
1138 * attenuation below.
1140 emit_op2(p
, OPCODE_DP3
, dist
, 0, VPpli
, VPpli
);
1141 emit_op1(p
, OPCODE_RSQ
, dist
, 0, dist
);
1142 emit_op2(p
, OPCODE_MUL
, VPpli
, 0, VPpli
, dist
);
1145 /* Calculate attenuation:
1147 att
= calculate_light_attenuation(p
, i
, VPpli
, dist
);
1148 release_temp(p
, dist
);
1150 /* Calculate viewer direction, or use infinite viewer:
1152 if (!p
->state
->material_shininess_is_zero
) {
1153 if (p
->state
->light_local_viewer
) {
1154 struct ureg eye_hat
= get_eye_position_normalized(p
);
1156 emit_op2(p
, OPCODE_SUB
, half
, 0, VPpli
, eye_hat
);
1157 emit_normalize_vec3(p
, half
, half
);
1158 } else if (p
->state
->unit
[i
].light_eyepos3_is_zero
) {
1159 half
= register_param3(p
, STATE_INTERNAL
,
1160 STATE_LIGHT_HALF_VECTOR
, i
);
1162 struct ureg z_dir
= swizzle(get_identity_param(p
),X
,Y
,W
,Z
);
1164 emit_op2(p
, OPCODE_ADD
, half
, 0, VPpli
, z_dir
);
1165 emit_normalize_vec3(p
, half
, half
);
1169 /* Calculate dot products:
1171 if (p
->state
->material_shininess_is_zero
) {
1172 emit_op2(p
, OPCODE_DP3
, dots
, 0, normal
, VPpli
);
1175 emit_op2(p
, OPCODE_DP3
, dots
, WRITEMASK_X
, normal
, VPpli
);
1176 emit_op2(p
, OPCODE_DP3
, dots
, WRITEMASK_Y
, normal
, half
);
1179 /* Front face lighting:
1182 struct ureg ambient
= get_lightprod(p
, i
, 0, STATE_AMBIENT
);
1183 struct ureg diffuse
= get_lightprod(p
, i
, 0, STATE_DIFFUSE
);
1184 struct ureg specular
= get_lightprod(p
, i
, 0, STATE_SPECULAR
);
1185 struct ureg res0
, res1
;
1186 GLuint mask0
, mask1
;
1188 if (count
== nr_lights
) {
1190 mask0
= WRITEMASK_XYZ
;
1191 mask1
= WRITEMASK_XYZ
;
1192 res0
= register_output( p
, VERT_RESULT_COL0
);
1193 res1
= register_output( p
, VERT_RESULT_COL1
);
1197 mask1
= WRITEMASK_XYZ
;
1199 res1
= register_output( p
, VERT_RESULT_COL0
);
1209 if (!is_undef(att
)) {
1210 /* light is attenuated by distance */
1211 emit_op1(p
, OPCODE_LIT
, lit
, 0, dots
);
1212 emit_op2(p
, OPCODE_MUL
, lit
, 0, lit
, att
);
1213 emit_op3(p
, OPCODE_MAD
, _col0
, 0, swizzle1(lit
,X
), ambient
, _col0
);
1215 else if (!p
->state
->material_shininess_is_zero
) {
1216 /* there's a non-zero specular term */
1217 emit_op1(p
, OPCODE_LIT
, lit
, 0, dots
);
1218 emit_op2(p
, OPCODE_ADD
, _col0
, 0, ambient
, _col0
);
1221 /* no attenutation, no specular */
1222 emit_degenerate_lit(p
, lit
, dots
);
1223 emit_op2(p
, OPCODE_ADD
, _col0
, 0, ambient
, _col0
);
1226 emit_op3(p
, OPCODE_MAD
, res0
, mask0
, swizzle1(lit
,Y
), diffuse
, _col0
);
1227 emit_op3(p
, OPCODE_MAD
, res1
, mask1
, swizzle1(lit
,Z
), specular
, _col1
);
1229 release_temp(p
, ambient
);
1230 release_temp(p
, diffuse
);
1231 release_temp(p
, specular
);
1234 /* Back face lighting:
1237 struct ureg ambient
= get_lightprod(p
, i
, 1, STATE_AMBIENT
);
1238 struct ureg diffuse
= get_lightprod(p
, i
, 1, STATE_DIFFUSE
);
1239 struct ureg specular
= get_lightprod(p
, i
, 1, STATE_SPECULAR
);
1240 struct ureg res0
, res1
;
1241 GLuint mask0
, mask1
;
1243 if (count
== nr_lights
) {
1245 mask0
= WRITEMASK_XYZ
;
1246 mask1
= WRITEMASK_XYZ
;
1247 res0
= register_output( p
, VERT_RESULT_BFC0
);
1248 res1
= register_output( p
, VERT_RESULT_BFC1
);
1252 mask1
= WRITEMASK_XYZ
;
1254 res1
= register_output( p
, VERT_RESULT_BFC0
);
1264 /* For the back face we need to negate the X and Y component
1265 * dot products. dots.Z has the negated back-face specular
1266 * exponent. We swizzle that into the W position. This
1267 * negation makes the back-face specular term positive again.
1269 dots
= negate(swizzle(dots
,X
,Y
,W
,Z
));
1271 if (!is_undef(att
)) {
1272 emit_op1(p
, OPCODE_LIT
, lit
, 0, dots
);
1273 emit_op2(p
, OPCODE_MUL
, lit
, 0, lit
, att
);
1274 emit_op3(p
, OPCODE_MAD
, _bfc0
, 0, swizzle1(lit
,X
), ambient
, _bfc0
);
1276 else if (!p
->state
->material_shininess_is_zero
) {
1277 emit_op1(p
, OPCODE_LIT
, lit
, 0, dots
);
1278 emit_op2(p
, OPCODE_ADD
, _bfc0
, 0, ambient
, _bfc0
); /**/
1281 emit_degenerate_lit(p
, lit
, dots
);
1282 emit_op2(p
, OPCODE_ADD
, _bfc0
, 0, ambient
, _bfc0
);
1285 emit_op3(p
, OPCODE_MAD
, res0
, mask0
, swizzle1(lit
,Y
), diffuse
, _bfc0
);
1286 emit_op3(p
, OPCODE_MAD
, res1
, mask1
, swizzle1(lit
,Z
), specular
, _bfc1
);
1287 /* restore dots to its original state for subsequent lights
1288 * by negating and swizzling again.
1290 dots
= negate(swizzle(dots
,X
,Y
,W
,Z
));
1292 release_temp(p
, ambient
);
1293 release_temp(p
, diffuse
);
1294 release_temp(p
, specular
);
1297 release_temp(p
, half
);
1298 release_temp(p
, VPpli
);
1299 release_temp(p
, att
);
1307 static void build_fog( struct tnl_program
*p
)
1309 struct ureg fog
= register_output(p
, VERT_RESULT_FOGC
);
1312 if (p
->state
->fog_source_is_depth
) {
1314 switch (p
->state
->fog_distance_mode
) {
1315 case FDM_EYE_RADIAL
: /* Z = sqrt(Xe*Xe + Ye*Ye + Ze*Ze) */
1316 input
= get_eye_position(p
);
1317 emit_op2(p
, OPCODE_DP3
, fog
, WRITEMASK_X
, input
, input
);
1318 emit_op1(p
, OPCODE_RSQ
, fog
, WRITEMASK_X
, fog
);
1319 emit_op1(p
, OPCODE_RCP
, fog
, WRITEMASK_X
, fog
);
1321 case FDM_EYE_PLANE
: /* Z = Ze */
1322 input
= get_eye_position_z(p
);
1323 emit_op1(p
, OPCODE_MOV
, fog
, WRITEMASK_X
, input
);
1325 case FDM_EYE_PLANE_ABS
: /* Z = abs(Ze) */
1326 input
= get_eye_position_z(p
);
1327 emit_op1(p
, OPCODE_ABS
, fog
, WRITEMASK_X
, input
);
1329 default: assert(0); break; /* can't happen */
1334 input
= swizzle1(register_input(p
, VERT_ATTRIB_FOG
), X
);
1335 emit_op1(p
, OPCODE_ABS
, fog
, WRITEMASK_X
, input
);
1338 emit_op1(p
, OPCODE_MOV
, fog
, WRITEMASK_YZW
, get_identity_param(p
));
1342 static void build_reflect_texgen( struct tnl_program
*p
,
1346 struct ureg normal
= get_transformed_normal(p
);
1347 struct ureg eye_hat
= get_eye_position_normalized(p
);
1348 struct ureg tmp
= get_temp(p
);
1351 emit_op2(p
, OPCODE_DP3
, tmp
, 0, normal
, eye_hat
);
1353 emit_op2(p
, OPCODE_ADD
, tmp
, 0, tmp
, tmp
);
1355 emit_op3(p
, OPCODE_MAD
, dest
, writemask
, negate(tmp
), normal
, eye_hat
);
1357 release_temp(p
, tmp
);
1361 static void build_sphere_texgen( struct tnl_program
*p
,
1365 struct ureg normal
= get_transformed_normal(p
);
1366 struct ureg eye_hat
= get_eye_position_normalized(p
);
1367 struct ureg tmp
= get_temp(p
);
1368 struct ureg half
= register_scalar_const(p
, .5);
1369 struct ureg r
= get_temp(p
);
1370 struct ureg inv_m
= get_temp(p
);
1371 struct ureg id
= get_identity_param(p
);
1373 /* Could share the above calculations, but it would be
1374 * a fairly odd state for someone to set (both sphere and
1375 * reflection active for different texture coordinate
1376 * components. Of course - if two texture units enable
1377 * reflect and/or sphere, things start to tilt in favour
1378 * of seperating this out:
1382 emit_op2(p
, OPCODE_DP3
, tmp
, 0, normal
, eye_hat
);
1384 emit_op2(p
, OPCODE_ADD
, tmp
, 0, tmp
, tmp
);
1386 emit_op3(p
, OPCODE_MAD
, r
, 0, negate(tmp
), normal
, eye_hat
);
1388 emit_op2(p
, OPCODE_ADD
, tmp
, 0, r
, swizzle(id
,X
,Y
,W
,Z
));
1389 /* rx^2 + ry^2 + (rz+1)^2 */
1390 emit_op2(p
, OPCODE_DP3
, tmp
, 0, tmp
, tmp
);
1392 emit_op1(p
, OPCODE_RSQ
, tmp
, 0, tmp
);
1394 emit_op2(p
, OPCODE_MUL
, inv_m
, 0, tmp
, half
);
1396 emit_op3(p
, OPCODE_MAD
, dest
, writemask
, r
, inv_m
, half
);
1398 release_temp(p
, tmp
);
1400 release_temp(p
, inv_m
);
1404 static void build_texture_transform( struct tnl_program
*p
)
1408 for (i
= 0; i
< MAX_TEXTURE_COORD_UNITS
; i
++) {
1410 if (!(p
->state
->fragprog_inputs_read
& FRAG_BIT_TEX(i
)))
1413 if (p
->state
->unit
[i
].coord_replace
)
1416 if (p
->state
->unit
[i
].texgen_enabled
||
1417 p
->state
->unit
[i
].texmat_enabled
) {
1419 GLuint texmat_enabled
= p
->state
->unit
[i
].texmat_enabled
;
1420 struct ureg out
= register_output(p
, VERT_RESULT_TEX0
+ i
);
1421 struct ureg out_texgen
= undef
;
1423 if (p
->state
->unit
[i
].texgen_enabled
) {
1424 GLuint copy_mask
= 0;
1425 GLuint sphere_mask
= 0;
1426 GLuint reflect_mask
= 0;
1427 GLuint normal_mask
= 0;
1431 out_texgen
= get_temp(p
);
1435 modes
[0] = p
->state
->unit
[i
].texgen_mode0
;
1436 modes
[1] = p
->state
->unit
[i
].texgen_mode1
;
1437 modes
[2] = p
->state
->unit
[i
].texgen_mode2
;
1438 modes
[3] = p
->state
->unit
[i
].texgen_mode3
;
1440 for (j
= 0; j
< 4; j
++) {
1442 case TXG_OBJ_LINEAR
: {
1443 struct ureg obj
= register_input(p
, VERT_ATTRIB_POS
);
1445 register_param3(p
, STATE_TEXGEN
, i
,
1446 STATE_TEXGEN_OBJECT_S
+ j
);
1448 emit_op2(p
, OPCODE_DP4
, out_texgen
, WRITEMASK_X
<< j
,
1452 case TXG_EYE_LINEAR
: {
1453 struct ureg eye
= get_eye_position(p
);
1455 register_param3(p
, STATE_TEXGEN
, i
,
1456 STATE_TEXGEN_EYE_S
+ j
);
1458 emit_op2(p
, OPCODE_DP4
, out_texgen
, WRITEMASK_X
<< j
,
1462 case TXG_SPHERE_MAP
:
1463 sphere_mask
|= WRITEMASK_X
<< j
;
1465 case TXG_REFLECTION_MAP
:
1466 reflect_mask
|= WRITEMASK_X
<< j
;
1468 case TXG_NORMAL_MAP
:
1469 normal_mask
|= WRITEMASK_X
<< j
;
1472 copy_mask
|= WRITEMASK_X
<< j
;
1477 build_sphere_texgen(p
, out_texgen
, sphere_mask
);
1481 build_reflect_texgen(p
, out_texgen
, reflect_mask
);
1485 struct ureg normal
= get_transformed_normal(p
);
1486 emit_op1(p
, OPCODE_MOV
, out_texgen
, normal_mask
, normal
);
1490 struct ureg in
= register_input(p
, VERT_ATTRIB_TEX0
+i
);
1491 emit_op1(p
, OPCODE_MOV
, out_texgen
, copy_mask
, in
);
1495 if (texmat_enabled
) {
1496 struct ureg texmat
[4];
1497 struct ureg in
= (!is_undef(out_texgen
) ?
1499 register_input(p
, VERT_ATTRIB_TEX0
+i
));
1500 if (p
->mvp_with_dp4
) {
1501 register_matrix_param5( p
, STATE_TEXTURE_MATRIX
, i
, 0, 3,
1503 emit_matrix_transform_vec4( p
, out
, texmat
, in
);
1506 register_matrix_param5( p
, STATE_TEXTURE_MATRIX
, i
, 0, 3,
1507 STATE_MATRIX_TRANSPOSE
, texmat
);
1508 emit_transpose_matrix_transform_vec4( p
, out
, texmat
, in
);
1515 emit_passthrough(p
, VERT_ATTRIB_TEX0
+i
, VERT_RESULT_TEX0
+i
);
1522 * Point size attenuation computation.
1524 static void build_atten_pointsize( struct tnl_program
*p
)
1526 struct ureg eye
= get_eye_position_z(p
);
1527 struct ureg state_size
= register_param2(p
, STATE_INTERNAL
, STATE_POINT_SIZE_CLAMPED
);
1528 struct ureg state_attenuation
= register_param1(p
, STATE_POINT_ATTENUATION
);
1529 struct ureg out
= register_output(p
, VERT_RESULT_PSIZ
);
1530 struct ureg ut
= get_temp(p
);
1533 emit_op1(p
, OPCODE_ABS
, ut
, WRITEMASK_Y
, swizzle1(eye
, Z
));
1534 /* p1 + dist * (p2 + dist * p3); */
1535 emit_op3(p
, OPCODE_MAD
, ut
, WRITEMASK_X
, swizzle1(ut
, Y
),
1536 swizzle1(state_attenuation
, Z
), swizzle1(state_attenuation
, Y
));
1537 emit_op3(p
, OPCODE_MAD
, ut
, WRITEMASK_X
, swizzle1(ut
, Y
),
1538 ut
, swizzle1(state_attenuation
, X
));
1540 /* 1 / sqrt(factor) */
1541 emit_op1(p
, OPCODE_RSQ
, ut
, WRITEMASK_X
, ut
);
1544 /* out = pointSize / sqrt(factor) */
1545 emit_op2(p
, OPCODE_MUL
, out
, WRITEMASK_X
, ut
, state_size
);
1547 /* this is a good place to clamp the point size since there's likely
1548 * no hardware registers to clamp point size at rasterization time.
1550 emit_op2(p
, OPCODE_MUL
, ut
, WRITEMASK_X
, ut
, state_size
);
1551 emit_op2(p
, OPCODE_MAX
, ut
, WRITEMASK_X
, ut
, swizzle1(state_size
, Y
));
1552 emit_op2(p
, OPCODE_MIN
, out
, WRITEMASK_X
, ut
, swizzle1(state_size
, Z
));
1555 release_temp(p
, ut
);
1560 * Pass-though per-vertex point size, from user's point size array.
1562 static void build_array_pointsize( struct tnl_program
*p
)
1564 struct ureg in
= register_input(p
, VERT_ATTRIB_POINT_SIZE
);
1565 struct ureg out
= register_output(p
, VERT_RESULT_PSIZ
);
1566 emit_op1(p
, OPCODE_MOV
, out
, WRITEMASK_X
, in
);
1570 static void build_tnl_program( struct tnl_program
*p
)
1572 /* Emit the program, starting with the modelview, projection transforms:
1576 /* Lighting calculations:
1578 if (p
->state
->fragprog_inputs_read
& (FRAG_BIT_COL0
|FRAG_BIT_COL1
)) {
1579 if (p
->state
->light_global_enabled
)
1582 if (p
->state
->fragprog_inputs_read
& FRAG_BIT_COL0
)
1583 emit_passthrough(p
, VERT_ATTRIB_COLOR0
, VERT_RESULT_COL0
);
1585 if (p
->state
->fragprog_inputs_read
& FRAG_BIT_COL1
)
1586 emit_passthrough(p
, VERT_ATTRIB_COLOR1
, VERT_RESULT_COL1
);
1590 if (p
->state
->fragprog_inputs_read
& FRAG_BIT_FOGC
)
1593 if (p
->state
->fragprog_inputs_read
& FRAG_BITS_TEX_ANY
)
1594 build_texture_transform(p
);
1596 if (p
->state
->point_attenuated
)
1597 build_atten_pointsize(p
);
1598 else if (p
->state
->point_array
)
1599 build_array_pointsize(p
);
1603 emit_op1(p
, OPCODE_END
, undef
, 0, undef
);
1614 create_new_program( const struct state_key
*key
,
1615 struct gl_vertex_program
*program
,
1616 GLboolean mvp_with_dp4
,
1619 struct tnl_program p
;
1621 memset(&p
, 0, sizeof(p
));
1623 p
.program
= program
;
1624 p
.eye_position
= undef
;
1625 p
.eye_position_z
= undef
;
1626 p
.eye_position_normalized
= undef
;
1627 p
.transformed_normal
= undef
;
1630 p
.mvp_with_dp4
= mvp_with_dp4
;
1632 if (max_temps
>= sizeof(int) * 8)
1633 p
.temp_reserved
= 0;
1635 p
.temp_reserved
= ~((1<<max_temps
)-1);
1637 /* Start by allocating 32 instructions.
1638 * If we need more, we'll grow the instruction array as needed.
1641 p
.program
->Base
.Instructions
= _mesa_alloc_instructions(p
.max_inst
);
1642 p
.program
->Base
.String
= NULL
;
1643 p
.program
->Base
.NumInstructions
=
1644 p
.program
->Base
.NumTemporaries
=
1645 p
.program
->Base
.NumParameters
=
1646 p
.program
->Base
.NumAttributes
= p
.program
->Base
.NumAddressRegs
= 0;
1647 p
.program
->Base
.Parameters
= _mesa_new_parameter_list();
1648 p
.program
->Base
.InputsRead
= 0;
1649 p
.program
->Base
.OutputsWritten
= 0;
1651 build_tnl_program( &p
);
1656 * Return a vertex program which implements the current fixed-function
1657 * transform/lighting/texgen operations.
1659 struct gl_vertex_program
*
1660 _mesa_get_fixed_func_vertex_program(struct gl_context
*ctx
)
1662 struct gl_vertex_program
*prog
;
1663 struct state_key key
;
1665 /* Grab all the relevent state and put it in a single structure:
1667 make_state_key(ctx
, &key
);
1669 /* Look for an already-prepared program for this state:
1671 prog
= gl_vertex_program(
1672 _mesa_search_program_cache(ctx
->VertexProgram
.Cache
, &key
, sizeof(key
)));
1675 /* OK, we'll have to build a new one */
1677 printf("Build new TNL program\n");
1679 prog
= gl_vertex_program(ctx
->Driver
.NewProgram(ctx
, GL_VERTEX_PROGRAM_ARB
, 0));
1683 create_new_program( &key
, prog
,
1685 ctx
->Const
.VertexProgram
.MaxTemps
);
1688 if (ctx
->Driver
.ProgramStringNotify
)
1689 ctx
->Driver
.ProgramStringNotify( ctx
, GL_VERTEX_PROGRAM_ARB
,
1692 _mesa_program_cache_insert(ctx
, ctx
->VertexProgram
.Cache
,
1693 &key
, sizeof(key
), &prog
->Base
);