1 /**************************************************************************
3 * Copyright 2007 VMware, Inc.
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 VMWARE 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/enums.h"
40 #include "main/ffvertex_prog.h"
41 #include "program/program.h"
42 #include "program/prog_cache.h"
43 #include "program/prog_instruction.h"
44 #include "program/prog_parameter.h"
45 #include "program/prog_print.h"
46 #include "program/prog_statevars.h"
47 #include "util/bitscan.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_distance_mode
:2;
64 unsigned separate_specular
:1;
65 unsigned point_attenuated
:1;
66 unsigned fragprog_inputs_read
:12;
68 GLbitfield varying_vp_inputs
;
71 unsigned light_enabled
:1;
72 unsigned light_eyepos3_is_zero
:1;
73 unsigned light_spotcutoff_is_180
:1;
74 unsigned light_attenuated
:1;
75 unsigned texmat_enabled
:1;
76 unsigned coord_replace
:1;
77 unsigned texgen_enabled
:1;
78 unsigned texgen_mode0
:4;
79 unsigned texgen_mode1
:4;
80 unsigned texgen_mode2
:4;
81 unsigned texgen_mode3
:4;
87 #define TXG_OBJ_LINEAR 1
88 #define TXG_EYE_LINEAR 2
89 #define TXG_SPHERE_MAP 3
90 #define TXG_REFLECTION_MAP 4
91 #define TXG_NORMAL_MAP 5
93 static GLuint
translate_texgen( GLboolean enabled
, GLenum mode
)
99 case GL_OBJECT_LINEAR
: return TXG_OBJ_LINEAR
;
100 case GL_EYE_LINEAR
: return TXG_EYE_LINEAR
;
101 case GL_SPHERE_MAP
: return TXG_SPHERE_MAP
;
102 case GL_REFLECTION_MAP_NV
: return TXG_REFLECTION_MAP
;
103 case GL_NORMAL_MAP_NV
: return TXG_NORMAL_MAP
;
104 default: return TXG_NONE
;
108 #define FDM_EYE_RADIAL 0
109 #define FDM_EYE_PLANE 1
110 #define FDM_EYE_PLANE_ABS 2
111 #define FDM_FROM_ARRAY 3
113 static GLuint
translate_fog_distance_mode(GLenum source
, GLenum mode
)
115 if (source
== GL_FRAGMENT_DEPTH_EXT
) {
117 case GL_EYE_RADIAL_NV
:
118 return FDM_EYE_RADIAL
;
120 return FDM_EYE_PLANE
;
121 default: /* shouldn't happen; fall through to a sensible default */
122 case GL_EYE_PLANE_ABSOLUTE_NV
:
123 return FDM_EYE_PLANE_ABS
;
126 return FDM_FROM_ARRAY
;
130 static GLboolean
check_active_shininess( struct gl_context
*ctx
,
131 const struct state_key
*key
,
134 GLuint attr
= MAT_ATTRIB_FRONT_SHININESS
+ side
;
136 if ((key
->varying_vp_inputs
& VERT_BIT_COLOR0
) &&
137 (key
->light_color_material_mask
& (1 << attr
)))
140 if (key
->varying_vp_inputs
& VERT_BIT_GENERIC(attr
))
143 if (ctx
->Light
.Material
.Attrib
[attr
][0] != 0.0F
)
150 static void make_state_key( struct gl_context
*ctx
, struct state_key
*key
)
152 const struct gl_program
*fp
= ctx
->FragmentProgram
._Current
;
155 memset(key
, 0, sizeof(struct state_key
));
157 /* This now relies on texenvprogram.c being active:
161 key
->need_eye_coords
= ctx
->_NeedEyeCoords
;
163 key
->fragprog_inputs_read
= fp
->info
.inputs_read
;
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
|= (VARYING_BIT_COL0
| VARYING_BIT_TEX0
);
171 if (ctx
->Light
.Enabled
) {
172 key
->light_global_enabled
= 1;
174 if (ctx
->Light
.Model
.LocalViewer
)
175 key
->light_local_viewer
= 1;
177 if (ctx
->Light
.Model
.TwoSide
)
178 key
->light_twoside
= 1;
180 if (ctx
->Light
.Model
.ColorControl
== GL_SEPARATE_SPECULAR_COLOR
)
181 key
->separate_specular
= 1;
183 if (ctx
->Light
.ColorMaterialEnabled
) {
184 key
->light_color_material_mask
= ctx
->Light
._ColorMaterialBitmask
;
187 mask
= ctx
->Light
._EnabledLights
;
189 const int i
= u_bit_scan(&mask
);
190 struct gl_light
*light
= &ctx
->Light
.Light
[i
];
192 key
->unit
[i
].light_enabled
= 1;
194 if (light
->EyePosition
[3] == 0.0F
)
195 key
->unit
[i
].light_eyepos3_is_zero
= 1;
197 if (light
->SpotCutoff
== 180.0F
)
198 key
->unit
[i
].light_spotcutoff_is_180
= 1;
200 if (light
->ConstantAttenuation
!= 1.0F
||
201 light
->LinearAttenuation
!= 0.0F
||
202 light
->QuadraticAttenuation
!= 0.0F
)
203 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 key
->fog_distance_mode
=
225 translate_fog_distance_mode(ctx
->Fog
.FogCoordinateSource
,
226 ctx
->Fog
.FogDistanceMode
);
228 if (ctx
->Point
._Attenuated
)
229 key
->point_attenuated
= 1;
231 mask
= ctx
->Texture
._EnabledCoordUnits
| ctx
->Texture
._TexGenEnabled
232 | ctx
->Texture
._TexMatEnabled
| ctx
->Point
.CoordReplace
;
234 const int i
= u_bit_scan(&mask
);
235 struct gl_texture_unit
*texUnit
= &ctx
->Texture
.Unit
[i
];
237 if (ctx
->Point
.PointSprite
)
238 if (ctx
->Point
.CoordReplace
& (1u << i
))
239 key
->unit
[i
].coord_replace
= 1;
241 if (ctx
->Texture
._TexMatEnabled
& ENABLE_TEXMAT(i
))
242 key
->unit
[i
].texmat_enabled
= 1;
244 if (texUnit
->TexGenEnabled
) {
245 key
->unit
[i
].texgen_enabled
= 1;
247 key
->unit
[i
].texgen_mode0
=
248 translate_texgen( texUnit
->TexGenEnabled
& (1<<0),
249 texUnit
->GenS
.Mode
);
250 key
->unit
[i
].texgen_mode1
=
251 translate_texgen( texUnit
->TexGenEnabled
& (1<<1),
252 texUnit
->GenT
.Mode
);
253 key
->unit
[i
].texgen_mode2
=
254 translate_texgen( texUnit
->TexGenEnabled
& (1<<2),
255 texUnit
->GenR
.Mode
);
256 key
->unit
[i
].texgen_mode3
=
257 translate_texgen( texUnit
->TexGenEnabled
& (1<<3),
258 texUnit
->GenQ
.Mode
);
265 /* Very useful debugging tool - produces annotated listing of
266 * generated program with line/function references for each
267 * instruction back into this file:
272 /* Use uregs to represent registers internally, translate to Mesa's
273 * expected formats on emit.
275 * NOTE: These are passed by value extensively in this file rather
276 * than as usual by pointer reference. If this disturbs you, try
277 * remembering they are just 32bits in size.
279 * GCC is smart enough to deal with these dword-sized structures in
280 * much the same way as if I had defined them as dwords and was using
281 * macros to access and set the fields. This is much nicer and easier
286 GLint idx
:9; /* relative addressing may be negative */
287 /* sizeof(idx) should == sizeof(prog_src_reg::Index) */
295 const struct state_key
*state
;
296 struct gl_program
*program
;
297 GLuint max_inst
; /** number of instructions allocated for program */
298 GLboolean mvp_with_dp4
;
301 GLuint temp_reserved
;
303 struct ureg eye_position
;
304 struct ureg eye_position_z
;
305 struct ureg eye_position_normalized
;
306 struct ureg transformed_normal
;
307 struct ureg identity
;
310 GLuint color_materials
;
314 static const struct ureg undef
= {
332 static struct ureg
make_ureg(GLuint file
, GLint idx
)
338 reg
.swz
= SWIZZLE_NOOP
;
344 static struct ureg
negate( struct ureg reg
)
351 static struct ureg
swizzle( struct ureg reg
, int x
, int y
, int z
, int w
)
353 reg
.swz
= MAKE_SWIZZLE4(GET_SWZ(reg
.swz
, x
),
356 GET_SWZ(reg
.swz
, w
));
361 static struct ureg
swizzle1( struct ureg reg
, int x
)
363 return swizzle(reg
, x
, x
, x
, x
);
367 static struct ureg
get_temp( struct tnl_program
*p
)
369 int bit
= ffs( ~p
->temp_in_use
);
371 _mesa_problem(NULL
, "%s: out of temporaries\n", __FILE__
);
375 if ((GLuint
) bit
> p
->program
->arb
.NumTemporaries
)
376 p
->program
->arb
.NumTemporaries
= bit
;
378 p
->temp_in_use
|= 1<<(bit
-1);
379 return make_ureg(PROGRAM_TEMPORARY
, bit
-1);
383 static struct ureg
reserve_temp( struct tnl_program
*p
)
385 struct ureg temp
= get_temp( p
);
386 p
->temp_reserved
|= 1<<temp
.idx
;
391 static void release_temp( struct tnl_program
*p
, struct ureg reg
)
393 if (reg
.file
== PROGRAM_TEMPORARY
) {
394 p
->temp_in_use
&= ~(1<<reg
.idx
);
395 p
->temp_in_use
|= p
->temp_reserved
; /* can't release reserved temps */
399 static void release_temps( struct tnl_program
*p
)
401 p
->temp_in_use
= p
->temp_reserved
;
405 static struct ureg
register_param5(struct tnl_program
*p
,
412 gl_state_index tokens
[STATE_LENGTH
];
419 idx
= _mesa_add_state_reference(p
->program
->Parameters
, tokens
);
420 return make_ureg(PROGRAM_STATE_VAR
, idx
);
424 #define register_param1(p,s0) register_param5(p,s0,0,0,0,0)
425 #define register_param2(p,s0,s1) register_param5(p,s0,s1,0,0,0)
426 #define register_param3(p,s0,s1,s2) register_param5(p,s0,s1,s2,0,0)
427 #define register_param4(p,s0,s1,s2,s3) register_param5(p,s0,s1,s2,s3,0)
432 * \param input one of VERT_ATTRIB_x tokens.
434 static struct ureg
register_input( struct tnl_program
*p
, GLuint input
)
436 assert(input
< VERT_ATTRIB_MAX
);
438 if (p
->state
->varying_vp_inputs
& VERT_BIT(input
)) {
439 p
->program
->info
.inputs_read
|= VERT_BIT(input
);
440 return make_ureg(PROGRAM_INPUT
, input
);
443 return register_param3( p
, STATE_INTERNAL
, STATE_CURRENT_ATTRIB
, input
);
449 * \param input one of VARYING_SLOT_x tokens.
451 static struct ureg
register_output( struct tnl_program
*p
, GLuint output
)
453 p
->program
->info
.outputs_written
|= BITFIELD64_BIT(output
);
454 return make_ureg(PROGRAM_OUTPUT
, output
);
458 static struct ureg
register_const4f( struct tnl_program
*p
,
464 gl_constant_value values
[4];
471 idx
= _mesa_add_unnamed_constant(p
->program
->Parameters
, values
, 4,
473 assert(swizzle
== SWIZZLE_NOOP
);
474 return make_ureg(PROGRAM_CONSTANT
, idx
);
477 #define register_const1f(p, s0) register_const4f(p, s0, 0, 0, 1)
478 #define register_scalar_const(p, s0) register_const4f(p, s0, s0, s0, s0)
479 #define register_const2f(p, s0, s1) register_const4f(p, s0, s1, 0, 1)
480 #define register_const3f(p, s0, s1, s2) register_const4f(p, s0, s1, s2, 1)
482 static GLboolean
is_undef( struct ureg reg
)
484 return reg
.file
== PROGRAM_UNDEFINED
;
488 static struct ureg
get_identity_param( struct tnl_program
*p
)
490 if (is_undef(p
->identity
))
491 p
->identity
= register_const4f(p
, 0,0,0,1);
496 static void register_matrix_param5( struct tnl_program
*p
,
497 GLint s0
, /* modelview, projection, etc */
498 GLint s1
, /* texture matrix number */
499 GLint s2
, /* first row */
500 GLint s3
, /* last row */
501 GLint s4
, /* inverse, transpose, etc */
502 struct ureg
*matrix
)
506 /* This is a bit sad as the support is there to pull the whole
507 * matrix out in one go:
509 for (i
= 0; i
<= s3
- s2
; i
++)
510 matrix
[i
] = register_param5( p
, s0
, s1
, i
, i
, s4
);
514 static void emit_arg( struct prog_src_register
*src
,
517 src
->File
= reg
.file
;
518 src
->Index
= reg
.idx
;
519 src
->Swizzle
= reg
.swz
;
520 src
->Negate
= reg
.negate
? NEGATE_XYZW
: NEGATE_NONE
;
522 /* Check that bitfield sizes aren't exceeded */
523 assert(src
->Index
== reg
.idx
);
527 static void emit_dst( struct prog_dst_register
*dst
,
528 struct ureg reg
, GLuint mask
)
530 dst
->File
= reg
.file
;
531 dst
->Index
= reg
.idx
;
532 /* allow zero as a shorthand for xyzw */
533 dst
->WriteMask
= mask
? mask
: WRITEMASK_XYZW
;
534 /* Check that bitfield sizes aren't exceeded */
535 assert(dst
->Index
== reg
.idx
);
539 static void debug_insn( struct prog_instruction
*inst
, const char *fn
,
543 static const char *last_fn
;
550 printf("%d:\t", line
);
551 _mesa_print_instruction(inst
);
556 static void emit_op3fn(struct tnl_program
*p
,
567 struct prog_instruction
*inst
;
569 assert(p
->program
->arb
.NumInstructions
<= p
->max_inst
);
571 if (p
->program
->arb
.NumInstructions
== p
->max_inst
) {
572 /* need to extend the program's instruction array */
573 struct prog_instruction
*newInst
;
575 /* double the size */
579 rzalloc_array(p
->program
, struct prog_instruction
, p
->max_inst
);
581 _mesa_error(NULL
, GL_OUT_OF_MEMORY
, "vertex program build");
585 _mesa_copy_instructions(newInst
, p
->program
->arb
.Instructions
,
586 p
->program
->arb
.NumInstructions
);
588 ralloc_free(p
->program
->arb
.Instructions
);
590 p
->program
->arb
.Instructions
= newInst
;
593 nr
= p
->program
->arb
.NumInstructions
++;
595 inst
= &p
->program
->arb
.Instructions
[nr
];
596 inst
->Opcode
= (enum prog_opcode
) op
;
598 emit_arg( &inst
->SrcReg
[0], src0
);
599 emit_arg( &inst
->SrcReg
[1], src1
);
600 emit_arg( &inst
->SrcReg
[2], src2
);
602 emit_dst( &inst
->DstReg
, dest
, mask
);
604 debug_insn(inst
, fn
, line
);
608 #define emit_op3(p, op, dst, mask, src0, src1, src2) \
609 emit_op3fn(p, op, dst, mask, src0, src1, src2, __func__, __LINE__)
611 #define emit_op2(p, op, dst, mask, src0, src1) \
612 emit_op3fn(p, op, dst, mask, src0, src1, undef, __func__, __LINE__)
614 #define emit_op1(p, op, dst, mask, src0) \
615 emit_op3fn(p, op, dst, mask, src0, undef, undef, __func__, __LINE__)
618 static struct ureg
make_temp( struct tnl_program
*p
, struct ureg reg
)
620 if (reg
.file
== PROGRAM_TEMPORARY
&&
621 !(p
->temp_reserved
& (1<<reg
.idx
)))
624 struct ureg temp
= get_temp(p
);
625 emit_op1(p
, OPCODE_MOV
, temp
, 0, reg
);
631 /* Currently no tracking performed of input/output/register size or
632 * active elements. Could be used to reduce these operations, as
633 * could the matrix type.
635 static void emit_matrix_transform_vec4( struct tnl_program
*p
,
637 const struct ureg
*mat
,
640 emit_op2(p
, OPCODE_DP4
, dest
, WRITEMASK_X
, src
, mat
[0]);
641 emit_op2(p
, OPCODE_DP4
, dest
, WRITEMASK_Y
, src
, mat
[1]);
642 emit_op2(p
, OPCODE_DP4
, dest
, WRITEMASK_Z
, src
, mat
[2]);
643 emit_op2(p
, OPCODE_DP4
, dest
, WRITEMASK_W
, src
, mat
[3]);
647 /* This version is much easier to implement if writemasks are not
648 * supported natively on the target or (like SSE), the target doesn't
649 * have a clean/obvious dotproduct implementation.
651 static void emit_transpose_matrix_transform_vec4( struct tnl_program
*p
,
653 const struct ureg
*mat
,
658 if (dest
.file
!= PROGRAM_TEMPORARY
)
663 emit_op2(p
, OPCODE_MUL
, tmp
, 0, swizzle1(src
,X
), mat
[0]);
664 emit_op3(p
, OPCODE_MAD
, tmp
, 0, swizzle1(src
,Y
), mat
[1], tmp
);
665 emit_op3(p
, OPCODE_MAD
, tmp
, 0, swizzle1(src
,Z
), mat
[2], tmp
);
666 emit_op3(p
, OPCODE_MAD
, dest
, 0, swizzle1(src
,W
), mat
[3], tmp
);
668 if (dest
.file
!= PROGRAM_TEMPORARY
)
669 release_temp(p
, tmp
);
673 static void emit_matrix_transform_vec3( struct tnl_program
*p
,
675 const struct ureg
*mat
,
678 emit_op2(p
, OPCODE_DP3
, dest
, WRITEMASK_X
, src
, mat
[0]);
679 emit_op2(p
, OPCODE_DP3
, dest
, WRITEMASK_Y
, src
, mat
[1]);
680 emit_op2(p
, OPCODE_DP3
, dest
, WRITEMASK_Z
, src
, mat
[2]);
684 static void emit_normalize_vec3( struct tnl_program
*p
,
688 struct ureg tmp
= get_temp(p
);
689 emit_op2(p
, OPCODE_DP3
, tmp
, WRITEMASK_X
, src
, src
);
690 emit_op1(p
, OPCODE_RSQ
, tmp
, WRITEMASK_X
, tmp
);
691 emit_op2(p
, OPCODE_MUL
, dest
, 0, src
, swizzle1(tmp
, X
));
692 release_temp(p
, tmp
);
696 static void emit_passthrough( struct tnl_program
*p
,
700 struct ureg out
= register_output(p
, output
);
701 emit_op1(p
, OPCODE_MOV
, out
, 0, register_input(p
, input
));
705 static struct ureg
get_eye_position( struct tnl_program
*p
)
707 if (is_undef(p
->eye_position
)) {
708 struct ureg pos
= register_input( p
, VERT_ATTRIB_POS
);
709 struct ureg modelview
[4];
711 p
->eye_position
= reserve_temp(p
);
713 if (p
->mvp_with_dp4
) {
714 register_matrix_param5( p
, STATE_MODELVIEW_MATRIX
, 0, 0, 3,
717 emit_matrix_transform_vec4(p
, p
->eye_position
, modelview
, pos
);
720 register_matrix_param5( p
, STATE_MODELVIEW_MATRIX
, 0, 0, 3,
721 STATE_MATRIX_TRANSPOSE
, modelview
);
723 emit_transpose_matrix_transform_vec4(p
, p
->eye_position
, modelview
, pos
);
727 return p
->eye_position
;
731 static struct ureg
get_eye_position_z( struct tnl_program
*p
)
733 if (!is_undef(p
->eye_position
))
734 return swizzle1(p
->eye_position
, Z
);
736 if (is_undef(p
->eye_position_z
)) {
737 struct ureg pos
= register_input( p
, VERT_ATTRIB_POS
);
738 struct ureg modelview
[4];
740 p
->eye_position_z
= reserve_temp(p
);
742 register_matrix_param5( p
, STATE_MODELVIEW_MATRIX
, 0, 0, 3,
745 emit_op2(p
, OPCODE_DP4
, p
->eye_position_z
, 0, pos
, modelview
[2]);
748 return p
->eye_position_z
;
752 static struct ureg
get_eye_position_normalized( struct tnl_program
*p
)
754 if (is_undef(p
->eye_position_normalized
)) {
755 struct ureg eye
= get_eye_position(p
);
756 p
->eye_position_normalized
= reserve_temp(p
);
757 emit_normalize_vec3(p
, p
->eye_position_normalized
, eye
);
760 return p
->eye_position_normalized
;
764 static struct ureg
get_transformed_normal( struct tnl_program
*p
)
766 if (is_undef(p
->transformed_normal
) &&
767 !p
->state
->need_eye_coords
&&
768 !p
->state
->normalize
&&
769 !(p
->state
->need_eye_coords
== p
->state
->rescale_normals
))
771 p
->transformed_normal
= register_input(p
, VERT_ATTRIB_NORMAL
);
773 else if (is_undef(p
->transformed_normal
))
775 struct ureg normal
= register_input(p
, VERT_ATTRIB_NORMAL
);
776 struct ureg mvinv
[3];
777 struct ureg transformed_normal
= reserve_temp(p
);
779 if (p
->state
->need_eye_coords
) {
780 register_matrix_param5( p
, STATE_MODELVIEW_MATRIX
, 0, 0, 2,
781 STATE_MATRIX_INVTRANS
, mvinv
);
783 /* Transform to eye space:
785 emit_matrix_transform_vec3( p
, transformed_normal
, mvinv
, normal
);
786 normal
= transformed_normal
;
789 /* Normalize/Rescale:
791 if (p
->state
->normalize
) {
792 emit_normalize_vec3( p
, transformed_normal
, normal
);
793 normal
= transformed_normal
;
795 else if (p
->state
->need_eye_coords
== p
->state
->rescale_normals
) {
796 /* This is already adjusted for eye/non-eye rendering:
798 struct ureg rescale
= register_param2(p
, STATE_INTERNAL
,
801 emit_op2( p
, OPCODE_MUL
, transformed_normal
, 0, normal
, rescale
);
802 normal
= transformed_normal
;
805 assert(normal
.file
== PROGRAM_TEMPORARY
);
806 p
->transformed_normal
= normal
;
809 return p
->transformed_normal
;
813 static void build_hpos( struct tnl_program
*p
)
815 struct ureg pos
= register_input( p
, VERT_ATTRIB_POS
);
816 struct ureg hpos
= register_output( p
, VARYING_SLOT_POS
);
819 if (p
->mvp_with_dp4
) {
820 register_matrix_param5( p
, STATE_MVP_MATRIX
, 0, 0, 3,
822 emit_matrix_transform_vec4( p
, hpos
, mvp
, pos
);
825 register_matrix_param5( p
, STATE_MVP_MATRIX
, 0, 0, 3,
826 STATE_MATRIX_TRANSPOSE
, mvp
);
827 emit_transpose_matrix_transform_vec4( p
, hpos
, mvp
, pos
);
832 static GLuint
material_attrib( GLuint side
, GLuint property
)
834 return (property
- STATE_AMBIENT
) * 2 + side
;
839 * Get a bitmask of which material values vary on a per-vertex basis.
841 static void set_material_flags( struct tnl_program
*p
)
843 p
->color_materials
= 0;
846 if (p
->state
->varying_vp_inputs
& VERT_BIT_COLOR0
) {
848 p
->color_materials
= p
->state
->light_color_material_mask
;
851 p
->materials
|= (p
->state
->varying_vp_inputs
>> VERT_ATTRIB_GENERIC0
);
855 static struct ureg
get_material( struct tnl_program
*p
, GLuint side
,
858 GLuint attrib
= material_attrib(side
, property
);
860 if (p
->color_materials
& (1<<attrib
))
861 return register_input(p
, VERT_ATTRIB_COLOR0
);
862 else if (p
->materials
& (1<<attrib
)) {
863 /* Put material values in the GENERIC slots -- they are not used
864 * for anything in fixed function mode.
866 return register_input( p
, attrib
+ VERT_ATTRIB_GENERIC0
);
869 return register_param3( p
, STATE_MATERIAL
, side
, property
);
872 #define SCENE_COLOR_BITS(side) (( MAT_BIT_FRONT_EMISSION | \
873 MAT_BIT_FRONT_AMBIENT | \
874 MAT_BIT_FRONT_DIFFUSE) << (side))
878 * Either return a precalculated constant value or emit code to
879 * calculate these values dynamically in the case where material calls
880 * are present between begin/end pairs.
882 * Probably want to shift this to the program compilation phase - if
883 * we always emitted the calculation here, a smart compiler could
884 * detect that it was constant (given a certain set of inputs), and
885 * lift it out of the main loop. That way the programs created here
886 * would be independent of the vertex_buffer details.
888 static struct ureg
get_scenecolor( struct tnl_program
*p
, GLuint side
)
890 if (p
->materials
& SCENE_COLOR_BITS(side
)) {
891 struct ureg lm_ambient
= register_param1(p
, STATE_LIGHTMODEL_AMBIENT
);
892 struct ureg material_emission
= get_material(p
, side
, STATE_EMISSION
);
893 struct ureg material_ambient
= get_material(p
, side
, STATE_AMBIENT
);
894 struct ureg material_diffuse
= get_material(p
, side
, STATE_DIFFUSE
);
895 struct ureg tmp
= make_temp(p
, material_diffuse
);
896 emit_op3(p
, OPCODE_MAD
, tmp
, WRITEMASK_XYZ
, lm_ambient
,
897 material_ambient
, material_emission
);
901 return register_param2( p
, STATE_LIGHTMODEL_SCENECOLOR
, side
);
905 static struct ureg
get_lightprod( struct tnl_program
*p
, GLuint light
,
906 GLuint side
, GLuint property
)
908 GLuint attrib
= material_attrib(side
, property
);
909 if (p
->materials
& (1<<attrib
)) {
910 struct ureg light_value
=
911 register_param3(p
, STATE_LIGHT
, light
, property
);
912 struct ureg material_value
= get_material(p
, side
, property
);
913 struct ureg tmp
= get_temp(p
);
914 emit_op2(p
, OPCODE_MUL
, tmp
, 0, light_value
, material_value
);
918 return register_param4(p
, STATE_LIGHTPROD
, light
, side
, property
);
922 static struct ureg
calculate_light_attenuation( struct tnl_program
*p
,
927 struct ureg attenuation
= register_param3(p
, STATE_LIGHT
, i
,
929 struct ureg att
= undef
;
931 /* Calculate spot attenuation:
933 if (!p
->state
->unit
[i
].light_spotcutoff_is_180
) {
934 struct ureg spot_dir_norm
= register_param3(p
, STATE_INTERNAL
,
935 STATE_LIGHT_SPOT_DIR_NORMALIZED
, i
);
936 struct ureg spot
= get_temp(p
);
937 struct ureg slt
= get_temp(p
);
941 emit_op2(p
, OPCODE_DP3
, spot
, 0, negate(VPpli
), spot_dir_norm
);
942 emit_op2(p
, OPCODE_SLT
, slt
, 0, swizzle1(spot_dir_norm
,W
), spot
);
943 emit_op1(p
, OPCODE_ABS
, spot
, 0, spot
);
944 emit_op2(p
, OPCODE_POW
, spot
, 0, spot
, swizzle1(attenuation
, W
));
945 emit_op2(p
, OPCODE_MUL
, att
, 0, slt
, spot
);
947 release_temp(p
, spot
);
948 release_temp(p
, slt
);
951 /* Calculate distance attenuation(See formula (2.4) at glspec 2.1 page 62):
953 * Skip the calucation when _dist_ is undefined(light_eyepos3_is_zero)
955 if (p
->state
->unit
[i
].light_attenuated
&& !is_undef(dist
)) {
959 emit_op1(p
, OPCODE_RCP
, dist
, WRITEMASK_YZ
, dist
);
961 emit_op2(p
, OPCODE_MUL
, dist
, WRITEMASK_XZ
, dist
, swizzle1(dist
,Y
));
963 emit_op2(p
, OPCODE_DP3
, dist
, 0, attenuation
, dist
);
965 if (!p
->state
->unit
[i
].light_spotcutoff_is_180
) {
967 emit_op1(p
, OPCODE_RCP
, dist
, 0, dist
);
968 /* spot-atten * dist-atten */
969 emit_op2(p
, OPCODE_MUL
, att
, 0, dist
, att
);
973 emit_op1(p
, OPCODE_RCP
, att
, 0, dist
);
983 * lit.y = MAX(0, dots.x)
984 * lit.z = SLT(0, dots.x)
986 static void emit_degenerate_lit( struct tnl_program
*p
,
990 struct ureg id
= get_identity_param(p
); /* id = {0,0,0,1} */
992 /* Note that lit.x & lit.w will not be examined. Note also that
993 * dots.xyzw == dots.xxxx.
996 /* MAX lit, id, dots;
998 emit_op2(p
, OPCODE_MAX
, lit
, WRITEMASK_XYZW
, id
, dots
);
1000 /* result[2] = (in > 0 ? 1 : 0)
1001 * SLT lit.z, id.z, dots; # lit.z = (0 < dots.z) ? 1 : 0
1003 emit_op2(p
, OPCODE_SLT
, lit
, WRITEMASK_Z
, swizzle1(id
,Z
), dots
);
1007 /* Need to add some addtional parameters to allow lighting in object
1008 * space - STATE_SPOT_DIRECTION and STATE_HALF_VECTOR implicitly assume eye
1011 static void build_lighting( struct tnl_program
*p
)
1013 const GLboolean twoside
= p
->state
->light_twoside
;
1014 const GLboolean separate
= p
->state
->separate_specular
;
1015 GLuint nr_lights
= 0, count
= 0;
1016 struct ureg normal
= get_transformed_normal(p
);
1017 struct ureg lit
= get_temp(p
);
1018 struct ureg dots
= get_temp(p
);
1019 struct ureg _col0
= undef
, _col1
= undef
;
1020 struct ureg _bfc0
= undef
, _bfc1
= undef
;
1025 * dots.x = dot(normal, VPpli)
1026 * dots.y = dot(normal, halfAngle)
1027 * dots.z = back.shininess
1028 * dots.w = front.shininess
1031 for (i
= 0; i
< MAX_LIGHTS
; i
++)
1032 if (p
->state
->unit
[i
].light_enabled
)
1035 set_material_flags(p
);
1038 if (!p
->state
->material_shininess_is_zero
) {
1039 struct ureg shininess
= get_material(p
, 0, STATE_SHININESS
);
1040 emit_op1(p
, OPCODE_MOV
, dots
, WRITEMASK_W
, swizzle1(shininess
,X
));
1041 release_temp(p
, shininess
);
1044 _col0
= make_temp(p
, get_scenecolor(p
, 0));
1046 _col1
= make_temp(p
, get_identity_param(p
));
1052 if (!p
->state
->material_shininess_is_zero
) {
1053 /* Note that we negate the back-face specular exponent here.
1054 * The negation will be un-done later in the back-face code below.
1056 struct ureg shininess
= get_material(p
, 1, STATE_SHININESS
);
1057 emit_op1(p
, OPCODE_MOV
, dots
, WRITEMASK_Z
,
1058 negate(swizzle1(shininess
,X
)));
1059 release_temp(p
, shininess
);
1062 _bfc0
= make_temp(p
, get_scenecolor(p
, 1));
1064 _bfc1
= make_temp(p
, get_identity_param(p
));
1069 /* If no lights, still need to emit the scenecolor.
1072 struct ureg res0
= register_output( p
, VARYING_SLOT_COL0
);
1073 emit_op1(p
, OPCODE_MOV
, res0
, 0, _col0
);
1077 struct ureg res1
= register_output( p
, VARYING_SLOT_COL1
);
1078 emit_op1(p
, OPCODE_MOV
, res1
, 0, _col1
);
1082 struct ureg res0
= register_output( p
, VARYING_SLOT_BFC0
);
1083 emit_op1(p
, OPCODE_MOV
, res0
, 0, _bfc0
);
1086 if (twoside
&& separate
) {
1087 struct ureg res1
= register_output( p
, VARYING_SLOT_BFC1
);
1088 emit_op1(p
, OPCODE_MOV
, res1
, 0, _bfc1
);
1091 if (nr_lights
== 0) {
1096 for (i
= 0; i
< MAX_LIGHTS
; i
++) {
1097 if (p
->state
->unit
[i
].light_enabled
) {
1098 struct ureg half
= undef
;
1099 struct ureg att
= undef
, VPpli
= undef
;
1100 struct ureg dist
= undef
;
1103 if (p
->state
->unit
[i
].light_eyepos3_is_zero
) {
1104 VPpli
= register_param3(p
, STATE_INTERNAL
,
1105 STATE_LIGHT_POSITION_NORMALIZED
, i
);
1107 struct ureg Ppli
= register_param3(p
, STATE_INTERNAL
,
1108 STATE_LIGHT_POSITION
, i
);
1109 struct ureg V
= get_eye_position(p
);
1111 VPpli
= get_temp(p
);
1114 /* Calculate VPpli vector
1116 emit_op2(p
, OPCODE_SUB
, VPpli
, 0, Ppli
, V
);
1118 /* Normalize VPpli. The dist value also used in
1119 * attenuation below.
1121 emit_op2(p
, OPCODE_DP3
, dist
, 0, VPpli
, VPpli
);
1122 emit_op1(p
, OPCODE_RSQ
, dist
, 0, dist
);
1123 emit_op2(p
, OPCODE_MUL
, VPpli
, 0, VPpli
, dist
);
1126 /* Calculate attenuation:
1128 att
= calculate_light_attenuation(p
, i
, VPpli
, dist
);
1129 release_temp(p
, dist
);
1131 /* Calculate viewer direction, or use infinite viewer:
1133 if (!p
->state
->material_shininess_is_zero
) {
1134 if (p
->state
->light_local_viewer
) {
1135 struct ureg eye_hat
= get_eye_position_normalized(p
);
1137 emit_op2(p
, OPCODE_SUB
, half
, 0, VPpli
, eye_hat
);
1138 emit_normalize_vec3(p
, half
, half
);
1139 } else if (p
->state
->unit
[i
].light_eyepos3_is_zero
) {
1140 half
= register_param3(p
, STATE_INTERNAL
,
1141 STATE_LIGHT_HALF_VECTOR
, i
);
1143 struct ureg z_dir
= swizzle(get_identity_param(p
),X
,Y
,W
,Z
);
1145 emit_op2(p
, OPCODE_ADD
, half
, 0, VPpli
, z_dir
);
1146 emit_normalize_vec3(p
, half
, half
);
1150 /* Calculate dot products:
1152 if (p
->state
->material_shininess_is_zero
) {
1153 emit_op2(p
, OPCODE_DP3
, dots
, 0, normal
, VPpli
);
1156 emit_op2(p
, OPCODE_DP3
, dots
, WRITEMASK_X
, normal
, VPpli
);
1157 emit_op2(p
, OPCODE_DP3
, dots
, WRITEMASK_Y
, normal
, half
);
1160 /* Front face lighting:
1163 struct ureg ambient
= get_lightprod(p
, i
, 0, STATE_AMBIENT
);
1164 struct ureg diffuse
= get_lightprod(p
, i
, 0, STATE_DIFFUSE
);
1165 struct ureg specular
= get_lightprod(p
, i
, 0, STATE_SPECULAR
);
1166 struct ureg res0
, res1
;
1167 GLuint mask0
, mask1
;
1169 if (count
== nr_lights
) {
1171 mask0
= WRITEMASK_XYZ
;
1172 mask1
= WRITEMASK_XYZ
;
1173 res0
= register_output( p
, VARYING_SLOT_COL0
);
1174 res1
= register_output( p
, VARYING_SLOT_COL1
);
1178 mask1
= WRITEMASK_XYZ
;
1180 res1
= register_output( p
, VARYING_SLOT_COL0
);
1190 if (!is_undef(att
)) {
1191 /* light is attenuated by distance */
1192 emit_op1(p
, OPCODE_LIT
, lit
, 0, dots
);
1193 emit_op2(p
, OPCODE_MUL
, lit
, 0, lit
, att
);
1194 emit_op3(p
, OPCODE_MAD
, _col0
, 0, swizzle1(lit
,X
), ambient
, _col0
);
1196 else if (!p
->state
->material_shininess_is_zero
) {
1197 /* there's a non-zero specular term */
1198 emit_op1(p
, OPCODE_LIT
, lit
, 0, dots
);
1199 emit_op2(p
, OPCODE_ADD
, _col0
, 0, ambient
, _col0
);
1202 /* no attenutation, no specular */
1203 emit_degenerate_lit(p
, lit
, dots
);
1204 emit_op2(p
, OPCODE_ADD
, _col0
, 0, ambient
, _col0
);
1207 emit_op3(p
, OPCODE_MAD
, res0
, mask0
, swizzle1(lit
,Y
), diffuse
, _col0
);
1208 emit_op3(p
, OPCODE_MAD
, res1
, mask1
, swizzle1(lit
,Z
), specular
, _col1
);
1210 release_temp(p
, ambient
);
1211 release_temp(p
, diffuse
);
1212 release_temp(p
, specular
);
1215 /* Back face lighting:
1218 struct ureg ambient
= get_lightprod(p
, i
, 1, STATE_AMBIENT
);
1219 struct ureg diffuse
= get_lightprod(p
, i
, 1, STATE_DIFFUSE
);
1220 struct ureg specular
= get_lightprod(p
, i
, 1, STATE_SPECULAR
);
1221 struct ureg res0
, res1
;
1222 GLuint mask0
, mask1
;
1224 if (count
== nr_lights
) {
1226 mask0
= WRITEMASK_XYZ
;
1227 mask1
= WRITEMASK_XYZ
;
1228 res0
= register_output( p
, VARYING_SLOT_BFC0
);
1229 res1
= register_output( p
, VARYING_SLOT_BFC1
);
1233 mask1
= WRITEMASK_XYZ
;
1235 res1
= register_output( p
, VARYING_SLOT_BFC0
);
1245 /* For the back face we need to negate the X and Y component
1246 * dot products. dots.Z has the negated back-face specular
1247 * exponent. We swizzle that into the W position. This
1248 * negation makes the back-face specular term positive again.
1250 dots
= negate(swizzle(dots
,X
,Y
,W
,Z
));
1252 if (!is_undef(att
)) {
1253 emit_op1(p
, OPCODE_LIT
, lit
, 0, dots
);
1254 emit_op2(p
, OPCODE_MUL
, lit
, 0, lit
, att
);
1255 emit_op3(p
, OPCODE_MAD
, _bfc0
, 0, swizzle1(lit
,X
), ambient
, _bfc0
);
1257 else if (!p
->state
->material_shininess_is_zero
) {
1258 emit_op1(p
, OPCODE_LIT
, lit
, 0, dots
);
1259 emit_op2(p
, OPCODE_ADD
, _bfc0
, 0, ambient
, _bfc0
); /**/
1262 emit_degenerate_lit(p
, lit
, dots
);
1263 emit_op2(p
, OPCODE_ADD
, _bfc0
, 0, ambient
, _bfc0
);
1266 emit_op3(p
, OPCODE_MAD
, res0
, mask0
, swizzle1(lit
,Y
), diffuse
, _bfc0
);
1267 emit_op3(p
, OPCODE_MAD
, res1
, mask1
, swizzle1(lit
,Z
), specular
, _bfc1
);
1268 /* restore dots to its original state for subsequent lights
1269 * by negating and swizzling again.
1271 dots
= negate(swizzle(dots
,X
,Y
,W
,Z
));
1273 release_temp(p
, ambient
);
1274 release_temp(p
, diffuse
);
1275 release_temp(p
, specular
);
1278 release_temp(p
, half
);
1279 release_temp(p
, VPpli
);
1280 release_temp(p
, att
);
1288 static void build_fog( struct tnl_program
*p
)
1290 struct ureg fog
= register_output(p
, VARYING_SLOT_FOGC
);
1293 switch (p
->state
->fog_distance_mode
) {
1294 case FDM_EYE_RADIAL
: /* Z = sqrt(Xe*Xe + Ye*Ye + Ze*Ze) */
1295 input
= get_eye_position(p
);
1296 emit_op2(p
, OPCODE_DP3
, fog
, WRITEMASK_X
, input
, input
);
1297 emit_op1(p
, OPCODE_RSQ
, fog
, WRITEMASK_X
, fog
);
1298 emit_op1(p
, OPCODE_RCP
, fog
, WRITEMASK_X
, fog
);
1300 case FDM_EYE_PLANE
: /* Z = Ze */
1301 input
= get_eye_position_z(p
);
1302 emit_op1(p
, OPCODE_MOV
, fog
, WRITEMASK_X
, input
);
1304 case FDM_EYE_PLANE_ABS
: /* Z = abs(Ze) */
1305 input
= get_eye_position_z(p
);
1306 emit_op1(p
, OPCODE_ABS
, fog
, WRITEMASK_X
, input
);
1308 case FDM_FROM_ARRAY
:
1309 input
= swizzle1(register_input(p
, VERT_ATTRIB_FOG
), X
);
1310 emit_op1(p
, OPCODE_ABS
, fog
, WRITEMASK_X
, input
);
1313 assert(!"Bad fog mode in build_fog()");
1317 emit_op1(p
, OPCODE_MOV
, fog
, WRITEMASK_YZW
, get_identity_param(p
));
1321 static void build_reflect_texgen( struct tnl_program
*p
,
1325 struct ureg normal
= get_transformed_normal(p
);
1326 struct ureg eye_hat
= get_eye_position_normalized(p
);
1327 struct ureg tmp
= get_temp(p
);
1330 emit_op2(p
, OPCODE_DP3
, tmp
, 0, normal
, eye_hat
);
1332 emit_op2(p
, OPCODE_ADD
, tmp
, 0, tmp
, tmp
);
1334 emit_op3(p
, OPCODE_MAD
, dest
, writemask
, negate(tmp
), normal
, eye_hat
);
1336 release_temp(p
, tmp
);
1340 static void build_sphere_texgen( struct tnl_program
*p
,
1344 struct ureg normal
= get_transformed_normal(p
);
1345 struct ureg eye_hat
= get_eye_position_normalized(p
);
1346 struct ureg tmp
= get_temp(p
);
1347 struct ureg half
= register_scalar_const(p
, .5);
1348 struct ureg r
= get_temp(p
);
1349 struct ureg inv_m
= get_temp(p
);
1350 struct ureg id
= get_identity_param(p
);
1352 /* Could share the above calculations, but it would be
1353 * a fairly odd state for someone to set (both sphere and
1354 * reflection active for different texture coordinate
1355 * components. Of course - if two texture units enable
1356 * reflect and/or sphere, things start to tilt in favour
1357 * of seperating this out:
1361 emit_op2(p
, OPCODE_DP3
, tmp
, 0, normal
, eye_hat
);
1363 emit_op2(p
, OPCODE_ADD
, tmp
, 0, tmp
, tmp
);
1365 emit_op3(p
, OPCODE_MAD
, r
, 0, negate(tmp
), normal
, eye_hat
);
1367 emit_op2(p
, OPCODE_ADD
, tmp
, 0, r
, swizzle(id
,X
,Y
,W
,Z
));
1368 /* rx^2 + ry^2 + (rz+1)^2 */
1369 emit_op2(p
, OPCODE_DP3
, tmp
, 0, tmp
, tmp
);
1371 emit_op1(p
, OPCODE_RSQ
, tmp
, 0, tmp
);
1373 emit_op2(p
, OPCODE_MUL
, inv_m
, 0, tmp
, half
);
1375 emit_op3(p
, OPCODE_MAD
, dest
, writemask
, r
, inv_m
, half
);
1377 release_temp(p
, tmp
);
1379 release_temp(p
, inv_m
);
1383 static void build_texture_transform( struct tnl_program
*p
)
1387 for (i
= 0; i
< MAX_TEXTURE_COORD_UNITS
; i
++) {
1389 if (!(p
->state
->fragprog_inputs_read
& VARYING_BIT_TEX(i
)))
1392 if (p
->state
->unit
[i
].coord_replace
)
1395 if (p
->state
->unit
[i
].texgen_enabled
||
1396 p
->state
->unit
[i
].texmat_enabled
) {
1398 GLuint texmat_enabled
= p
->state
->unit
[i
].texmat_enabled
;
1399 struct ureg out
= register_output(p
, VARYING_SLOT_TEX0
+ i
);
1400 struct ureg out_texgen
= undef
;
1402 if (p
->state
->unit
[i
].texgen_enabled
) {
1403 GLuint copy_mask
= 0;
1404 GLuint sphere_mask
= 0;
1405 GLuint reflect_mask
= 0;
1406 GLuint normal_mask
= 0;
1410 out_texgen
= get_temp(p
);
1414 modes
[0] = p
->state
->unit
[i
].texgen_mode0
;
1415 modes
[1] = p
->state
->unit
[i
].texgen_mode1
;
1416 modes
[2] = p
->state
->unit
[i
].texgen_mode2
;
1417 modes
[3] = p
->state
->unit
[i
].texgen_mode3
;
1419 for (j
= 0; j
< 4; j
++) {
1421 case TXG_OBJ_LINEAR
: {
1422 struct ureg obj
= register_input(p
, VERT_ATTRIB_POS
);
1424 register_param3(p
, STATE_TEXGEN
, i
,
1425 STATE_TEXGEN_OBJECT_S
+ j
);
1427 emit_op2(p
, OPCODE_DP4
, out_texgen
, WRITEMASK_X
<< j
,
1431 case TXG_EYE_LINEAR
: {
1432 struct ureg eye
= get_eye_position(p
);
1434 register_param3(p
, STATE_TEXGEN
, i
,
1435 STATE_TEXGEN_EYE_S
+ j
);
1437 emit_op2(p
, OPCODE_DP4
, out_texgen
, WRITEMASK_X
<< j
,
1441 case TXG_SPHERE_MAP
:
1442 sphere_mask
|= WRITEMASK_X
<< j
;
1444 case TXG_REFLECTION_MAP
:
1445 reflect_mask
|= WRITEMASK_X
<< j
;
1447 case TXG_NORMAL_MAP
:
1448 normal_mask
|= WRITEMASK_X
<< j
;
1451 copy_mask
|= WRITEMASK_X
<< j
;
1456 build_sphere_texgen(p
, out_texgen
, sphere_mask
);
1460 build_reflect_texgen(p
, out_texgen
, reflect_mask
);
1464 struct ureg normal
= get_transformed_normal(p
);
1465 emit_op1(p
, OPCODE_MOV
, out_texgen
, normal_mask
, normal
);
1469 struct ureg in
= register_input(p
, VERT_ATTRIB_TEX0
+i
);
1470 emit_op1(p
, OPCODE_MOV
, out_texgen
, copy_mask
, in
);
1474 if (texmat_enabled
) {
1475 struct ureg texmat
[4];
1476 struct ureg in
= (!is_undef(out_texgen
) ?
1478 register_input(p
, VERT_ATTRIB_TEX0
+i
));
1479 if (p
->mvp_with_dp4
) {
1480 register_matrix_param5( p
, STATE_TEXTURE_MATRIX
, i
, 0, 3,
1482 emit_matrix_transform_vec4( p
, out
, texmat
, in
);
1485 register_matrix_param5( p
, STATE_TEXTURE_MATRIX
, i
, 0, 3,
1486 STATE_MATRIX_TRANSPOSE
, texmat
);
1487 emit_transpose_matrix_transform_vec4( p
, out
, texmat
, in
);
1494 emit_passthrough(p
, VERT_ATTRIB_TEX0
+i
, VARYING_SLOT_TEX0
+i
);
1501 * Point size attenuation computation.
1503 static void build_atten_pointsize( struct tnl_program
*p
)
1505 struct ureg eye
= get_eye_position_z(p
);
1506 struct ureg state_size
= register_param2(p
, STATE_INTERNAL
, STATE_POINT_SIZE_CLAMPED
);
1507 struct ureg state_attenuation
= register_param1(p
, STATE_POINT_ATTENUATION
);
1508 struct ureg out
= register_output(p
, VARYING_SLOT_PSIZ
);
1509 struct ureg ut
= get_temp(p
);
1512 emit_op1(p
, OPCODE_ABS
, ut
, WRITEMASK_Y
, swizzle1(eye
, Z
));
1513 /* p1 + dist * (p2 + dist * p3); */
1514 emit_op3(p
, OPCODE_MAD
, ut
, WRITEMASK_X
, swizzle1(ut
, Y
),
1515 swizzle1(state_attenuation
, Z
), swizzle1(state_attenuation
, Y
));
1516 emit_op3(p
, OPCODE_MAD
, ut
, WRITEMASK_X
, swizzle1(ut
, Y
),
1517 ut
, swizzle1(state_attenuation
, X
));
1519 /* 1 / sqrt(factor) */
1520 emit_op1(p
, OPCODE_RSQ
, ut
, WRITEMASK_X
, ut
);
1523 /* out = pointSize / sqrt(factor) */
1524 emit_op2(p
, OPCODE_MUL
, out
, WRITEMASK_X
, ut
, state_size
);
1526 /* this is a good place to clamp the point size since there's likely
1527 * no hardware registers to clamp point size at rasterization time.
1529 emit_op2(p
, OPCODE_MUL
, ut
, WRITEMASK_X
, ut
, state_size
);
1530 emit_op2(p
, OPCODE_MAX
, ut
, WRITEMASK_X
, ut
, swizzle1(state_size
, Y
));
1531 emit_op2(p
, OPCODE_MIN
, out
, WRITEMASK_X
, ut
, swizzle1(state_size
, Z
));
1534 release_temp(p
, ut
);
1539 * Pass-though per-vertex point size, from user's point size array.
1541 static void build_array_pointsize( struct tnl_program
*p
)
1543 struct ureg in
= register_input(p
, VERT_ATTRIB_POINT_SIZE
);
1544 struct ureg out
= register_output(p
, VARYING_SLOT_PSIZ
);
1545 emit_op1(p
, OPCODE_MOV
, out
, WRITEMASK_X
, in
);
1549 static void build_tnl_program( struct tnl_program
*p
)
1551 /* Emit the program, starting with the modelview, projection transforms:
1555 /* Lighting calculations:
1557 if (p
->state
->fragprog_inputs_read
& (VARYING_BIT_COL0
|VARYING_BIT_COL1
)) {
1558 if (p
->state
->light_global_enabled
)
1561 if (p
->state
->fragprog_inputs_read
& VARYING_BIT_COL0
)
1562 emit_passthrough(p
, VERT_ATTRIB_COLOR0
, VARYING_SLOT_COL0
);
1564 if (p
->state
->fragprog_inputs_read
& VARYING_BIT_COL1
)
1565 emit_passthrough(p
, VERT_ATTRIB_COLOR1
, VARYING_SLOT_COL1
);
1569 if (p
->state
->fragprog_inputs_read
& VARYING_BIT_FOGC
)
1572 if (p
->state
->fragprog_inputs_read
& VARYING_BITS_TEX_ANY
)
1573 build_texture_transform(p
);
1575 if (p
->state
->point_attenuated
)
1576 build_atten_pointsize(p
);
1577 else if (p
->state
->varying_vp_inputs
& VERT_BIT_POINT_SIZE
)
1578 build_array_pointsize(p
);
1582 emit_op1(p
, OPCODE_END
, undef
, 0, undef
);
1593 create_new_program( const struct state_key
*key
,
1594 struct gl_program
*program
,
1595 GLboolean mvp_with_dp4
,
1598 struct tnl_program p
;
1600 memset(&p
, 0, sizeof(p
));
1602 p
.program
= program
;
1603 p
.eye_position
= undef
;
1604 p
.eye_position_z
= undef
;
1605 p
.eye_position_normalized
= undef
;
1606 p
.transformed_normal
= undef
;
1609 p
.mvp_with_dp4
= mvp_with_dp4
;
1611 if (max_temps
>= sizeof(int) * 8)
1612 p
.temp_reserved
= 0;
1614 p
.temp_reserved
= ~((1<<max_temps
)-1);
1616 /* Start by allocating 32 instructions.
1617 * If we need more, we'll grow the instruction array as needed.
1620 p
.program
->arb
.Instructions
=
1621 rzalloc_array(program
, struct prog_instruction
, p
.max_inst
);
1622 p
.program
->String
= NULL
;
1623 p
.program
->arb
.NumInstructions
=
1624 p
.program
->arb
.NumTemporaries
=
1625 p
.program
->arb
.NumParameters
=
1626 p
.program
->arb
.NumAttributes
= p
.program
->arb
.NumAddressRegs
= 0;
1627 p
.program
->Parameters
= _mesa_new_parameter_list();
1628 p
.program
->info
.inputs_read
= 0;
1629 p
.program
->info
.outputs_written
= 0;
1631 build_tnl_program( &p
);
1636 * Return a vertex program which implements the current fixed-function
1637 * transform/lighting/texgen operations.
1640 _mesa_get_fixed_func_vertex_program(struct gl_context
*ctx
)
1642 struct gl_program
*prog
;
1643 struct state_key key
;
1645 /* Grab all the relevant state and put it in a single structure:
1647 make_state_key(ctx
, &key
);
1649 /* Look for an already-prepared program for this state:
1651 prog
= _mesa_search_program_cache(ctx
->VertexProgram
.Cache
, &key
,
1655 /* OK, we'll have to build a new one */
1657 printf("Build new TNL program\n");
1659 prog
= ctx
->Driver
.NewProgram(ctx
, GL_VERTEX_PROGRAM_ARB
, 0, true);
1663 create_new_program( &key
, prog
,
1664 ctx
->Const
.ShaderCompilerOptions
[MESA_SHADER_VERTEX
].OptimizeForAOS
,
1665 ctx
->Const
.Program
[MESA_SHADER_VERTEX
].MaxTemps
);
1667 if (ctx
->Driver
.ProgramStringNotify
)
1668 ctx
->Driver
.ProgramStringNotify(ctx
, GL_VERTEX_PROGRAM_ARB
, prog
);
1670 _mesa_program_cache_insert(ctx
, ctx
->VertexProgram
.Cache
, &key
,