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"
49 /** Max of number of lights and texture coord units */
50 #define NUM_UNITS MAX2(MAX_TEXTURE_COORD_UNITS, MAX_LIGHTS)
53 unsigned light_color_material_mask
:12;
54 unsigned light_global_enabled
:1;
55 unsigned light_local_viewer
:1;
56 unsigned light_twoside
:1;
57 unsigned material_shininess_is_zero
:1;
58 unsigned need_eye_coords
:1;
60 unsigned rescale_normals
:1;
62 unsigned fog_source_is_depth
:1;
63 unsigned fog_distance_mode
:2;
64 unsigned separate_specular
:1;
65 unsigned point_attenuated
:1;
66 unsigned point_array
:1;
67 unsigned texture_enabled_global
:1;
68 unsigned fragprog_inputs_read
:12;
70 GLbitfield64 varying_vp_inputs
;
73 unsigned light_enabled
:1;
74 unsigned light_eyepos3_is_zero
:1;
75 unsigned light_spotcutoff_is_180
:1;
76 unsigned light_attenuated
:1;
77 unsigned texunit_really_enabled
:1;
78 unsigned texmat_enabled
:1;
79 unsigned coord_replace
:1;
80 unsigned texgen_enabled
:4;
81 unsigned texgen_mode0
:4;
82 unsigned texgen_mode1
:4;
83 unsigned texgen_mode2
:4;
84 unsigned texgen_mode3
:4;
90 #define TXG_OBJ_LINEAR 1
91 #define TXG_EYE_LINEAR 2
92 #define TXG_SPHERE_MAP 3
93 #define TXG_REFLECTION_MAP 4
94 #define TXG_NORMAL_MAP 5
96 static GLuint
translate_texgen( GLboolean enabled
, GLenum mode
)
102 case GL_OBJECT_LINEAR
: return TXG_OBJ_LINEAR
;
103 case GL_EYE_LINEAR
: return TXG_EYE_LINEAR
;
104 case GL_SPHERE_MAP
: return TXG_SPHERE_MAP
;
105 case GL_REFLECTION_MAP_NV
: return TXG_REFLECTION_MAP
;
106 case GL_NORMAL_MAP_NV
: return TXG_NORMAL_MAP
;
107 default: return TXG_NONE
;
111 #define FDM_EYE_RADIAL 0
112 #define FDM_EYE_PLANE 1
113 #define FDM_EYE_PLANE_ABS 2
115 static GLuint
translate_fog_distance_mode( GLenum mode
)
118 case GL_EYE_RADIAL_NV
:
119 return FDM_EYE_RADIAL
;
121 return FDM_EYE_PLANE
;
122 default: /* shouldn't happen; fall through to a sensible default */
123 case GL_EYE_PLANE_ABSOLUTE_NV
:
124 return FDM_EYE_PLANE_ABS
;
128 static GLboolean
check_active_shininess( struct gl_context
*ctx
,
129 const struct state_key
*key
,
132 GLuint attr
= MAT_ATTRIB_FRONT_SHININESS
+ side
;
134 if ((key
->varying_vp_inputs
& VERT_BIT_COLOR0
) &&
135 (key
->light_color_material_mask
& (1 << attr
)))
138 if (key
->varying_vp_inputs
& VERT_BIT_GENERIC(attr
))
141 if (ctx
->Light
.Material
.Attrib
[attr
][0] != 0.0F
)
148 static void make_state_key( struct gl_context
*ctx
, struct state_key
*key
)
150 const struct gl_fragment_program
*fp
;
153 memset(key
, 0, sizeof(struct state_key
));
154 fp
= ctx
->FragmentProgram
._Current
;
156 /* This now relies on texenvprogram.c being active:
160 key
->need_eye_coords
= ctx
->_NeedEyeCoords
;
162 key
->fragprog_inputs_read
= fp
->Base
.InputsRead
;
163 key
->varying_vp_inputs
= ctx
->varying_vp_inputs
;
165 if (ctx
->RenderMode
== GL_FEEDBACK
) {
166 /* make sure the vertprog emits color and tex0 */
167 key
->fragprog_inputs_read
|= (VARYING_BIT_COL0
| VARYING_BIT_TEX0
);
170 key
->separate_specular
= (ctx
->Light
.Model
.ColorControl
==
171 GL_SEPARATE_SPECULAR_COLOR
);
173 if (ctx
->Light
.Enabled
) {
174 key
->light_global_enabled
= 1;
176 if (ctx
->Light
.Model
.LocalViewer
)
177 key
->light_local_viewer
= 1;
179 if (ctx
->Light
.Model
.TwoSide
)
180 key
->light_twoside
= 1;
182 if (ctx
->Light
.ColorMaterialEnabled
) {
183 key
->light_color_material_mask
= ctx
->Light
._ColorMaterialBitmask
;
186 for (i
= 0; i
< MAX_LIGHTS
; i
++) {
187 struct gl_light
*light
= &ctx
->Light
.Light
[i
];
189 if (light
->Enabled
) {
190 key
->unit
[i
].light_enabled
= 1;
192 if (light
->EyePosition
[3] == 0.0F
)
193 key
->unit
[i
].light_eyepos3_is_zero
= 1;
195 if (light
->SpotCutoff
== 180.0F
)
196 key
->unit
[i
].light_spotcutoff_is_180
= 1;
198 if (light
->ConstantAttenuation
!= 1.0F
||
199 light
->LinearAttenuation
!= 0.0F
||
200 light
->QuadraticAttenuation
!= 0.0F
)
201 key
->unit
[i
].light_attenuated
= 1;
205 if (check_active_shininess(ctx
, key
, 0)) {
206 key
->material_shininess_is_zero
= 0;
208 else if (key
->light_twoside
&&
209 check_active_shininess(ctx
, key
, 1)) {
210 key
->material_shininess_is_zero
= 0;
213 key
->material_shininess_is_zero
= 1;
217 if (ctx
->Transform
.Normalize
)
220 if (ctx
->Transform
.RescaleNormals
)
221 key
->rescale_normals
= 1;
223 if (ctx
->Fog
.FogCoordinateSource
== GL_FRAGMENT_DEPTH_EXT
) {
224 key
->fog_source_is_depth
= 1;
225 key
->fog_distance_mode
= translate_fog_distance_mode(ctx
->Fog
.FogDistanceMode
);
228 if (ctx
->Point
._Attenuated
)
229 key
->point_attenuated
= 1;
231 if (ctx
->Array
.VAO
->VertexAttrib
[VERT_ATTRIB_POINT_SIZE
].Enabled
)
232 key
->point_array
= 1;
234 if (ctx
->Texture
._TexGenEnabled
||
235 ctx
->Texture
._TexMatEnabled
||
236 ctx
->Texture
._MaxEnabledTexImageUnit
!= -1)
237 key
->texture_enabled_global
= 1;
239 for (i
= 0; i
< MAX_TEXTURE_COORD_UNITS
; i
++) {
240 struct gl_texture_unit
*texUnit
= &ctx
->Texture
.Unit
[i
];
242 if (texUnit
->_Current
)
243 key
->unit
[i
].texunit_really_enabled
= 1;
245 if (ctx
->Point
.PointSprite
)
246 if (ctx
->Point
.CoordReplace
[i
])
247 key
->unit
[i
].coord_replace
= 1;
249 if (ctx
->Texture
._TexMatEnabled
& ENABLE_TEXMAT(i
))
250 key
->unit
[i
].texmat_enabled
= 1;
252 if (texUnit
->TexGenEnabled
) {
253 key
->unit
[i
].texgen_enabled
= 1;
255 key
->unit
[i
].texgen_mode0
=
256 translate_texgen( texUnit
->TexGenEnabled
& (1<<0),
257 texUnit
->GenS
.Mode
);
258 key
->unit
[i
].texgen_mode1
=
259 translate_texgen( texUnit
->TexGenEnabled
& (1<<1),
260 texUnit
->GenT
.Mode
);
261 key
->unit
[i
].texgen_mode2
=
262 translate_texgen( texUnit
->TexGenEnabled
& (1<<2),
263 texUnit
->GenR
.Mode
);
264 key
->unit
[i
].texgen_mode3
=
265 translate_texgen( texUnit
->TexGenEnabled
& (1<<3),
266 texUnit
->GenQ
.Mode
);
273 /* Very useful debugging tool - produces annotated listing of
274 * generated program with line/function references for each
275 * instruction back into this file:
280 /* Use uregs to represent registers internally, translate to Mesa's
281 * expected formats on emit.
283 * NOTE: These are passed by value extensively in this file rather
284 * than as usual by pointer reference. If this disturbs you, try
285 * remembering they are just 32bits in size.
287 * GCC is smart enough to deal with these dword-sized structures in
288 * much the same way as if I had defined them as dwords and was using
289 * macros to access and set the fields. This is much nicer and easier
294 GLint idx
:9; /* relative addressing may be negative */
295 /* sizeof(idx) should == sizeof(prog_src_reg::Index) */
304 const struct state_key
*state
;
305 struct gl_vertex_program
*program
;
306 GLuint 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
= {
342 static struct ureg
make_ureg(GLuint file
, GLint idx
)
349 reg
.swz
= SWIZZLE_NOOP
;
356 static struct ureg
absolute( struct ureg reg
)
364 static struct ureg
negate( struct ureg reg
)
371 static struct ureg
swizzle( struct ureg reg
, int x
, int y
, int z
, int w
)
373 reg
.swz
= MAKE_SWIZZLE4(GET_SWZ(reg
.swz
, x
),
376 GET_SWZ(reg
.swz
, w
));
381 static struct ureg
swizzle1( struct ureg reg
, int x
)
383 return swizzle(reg
, x
, x
, x
, x
);
387 static struct ureg
get_temp( struct tnl_program
*p
)
389 int bit
= ffs( ~p
->temp_in_use
);
391 _mesa_problem(NULL
, "%s: out of temporaries\n", __FILE__
);
395 if ((GLuint
) bit
> p
->program
->Base
.NumTemporaries
)
396 p
->program
->Base
.NumTemporaries
= bit
;
398 p
->temp_in_use
|= 1<<(bit
-1);
399 return make_ureg(PROGRAM_TEMPORARY
, bit
-1);
403 static struct ureg
reserve_temp( struct tnl_program
*p
)
405 struct ureg temp
= get_temp( p
);
406 p
->temp_reserved
|= 1<<temp
.idx
;
411 static void release_temp( struct tnl_program
*p
, struct ureg reg
)
413 if (reg
.file
== PROGRAM_TEMPORARY
) {
414 p
->temp_in_use
&= ~(1<<reg
.idx
);
415 p
->temp_in_use
|= p
->temp_reserved
; /* can't release reserved temps */
419 static void release_temps( struct tnl_program
*p
)
421 p
->temp_in_use
= p
->temp_reserved
;
425 static struct ureg
register_param5(struct tnl_program
*p
,
432 gl_state_index tokens
[STATE_LENGTH
];
439 idx
= _mesa_add_state_reference( p
->program
->Base
.Parameters
, tokens
);
440 return make_ureg(PROGRAM_STATE_VAR
, idx
);
444 #define register_param1(p,s0) register_param5(p,s0,0,0,0,0)
445 #define register_param2(p,s0,s1) register_param5(p,s0,s1,0,0,0)
446 #define register_param3(p,s0,s1,s2) register_param5(p,s0,s1,s2,0,0)
447 #define register_param4(p,s0,s1,s2,s3) register_param5(p,s0,s1,s2,s3,0)
452 * \param input one of VERT_ATTRIB_x tokens.
454 static struct ureg
register_input( struct tnl_program
*p
, GLuint input
)
456 assert(input
< VERT_ATTRIB_MAX
);
458 if (p
->state
->varying_vp_inputs
& VERT_BIT(input
)) {
459 p
->program
->Base
.InputsRead
|= VERT_BIT(input
);
460 return make_ureg(PROGRAM_INPUT
, input
);
463 return register_param3( p
, STATE_INTERNAL
, STATE_CURRENT_ATTRIB
, input
);
469 * \param input one of VARYING_SLOT_x tokens.
471 static struct ureg
register_output( struct tnl_program
*p
, GLuint output
)
473 p
->program
->Base
.OutputsWritten
|= BITFIELD64_BIT(output
);
474 return make_ureg(PROGRAM_OUTPUT
, output
);
478 static struct ureg
register_const4f( struct tnl_program
*p
,
484 gl_constant_value values
[4];
491 idx
= _mesa_add_unnamed_constant( p
->program
->Base
.Parameters
, values
, 4,
493 assert(swizzle
== SWIZZLE_NOOP
);
494 return make_ureg(PROGRAM_CONSTANT
, idx
);
497 #define register_const1f(p, s0) register_const4f(p, s0, 0, 0, 1)
498 #define register_scalar_const(p, s0) register_const4f(p, s0, s0, s0, s0)
499 #define register_const2f(p, s0, s1) register_const4f(p, s0, s1, 0, 1)
500 #define register_const3f(p, s0, s1, s2) register_const4f(p, s0, s1, s2, 1)
502 static GLboolean
is_undef( struct ureg reg
)
504 return reg
.file
== PROGRAM_UNDEFINED
;
508 static struct ureg
get_identity_param( struct tnl_program
*p
)
510 if (is_undef(p
->identity
))
511 p
->identity
= register_const4f(p
, 0,0,0,1);
516 static void register_matrix_param5( struct tnl_program
*p
,
517 GLint s0
, /* modelview, projection, etc */
518 GLint s1
, /* texture matrix number */
519 GLint s2
, /* first row */
520 GLint s3
, /* last row */
521 GLint s4
, /* inverse, transpose, etc */
522 struct ureg
*matrix
)
526 /* This is a bit sad as the support is there to pull the whole
527 * matrix out in one go:
529 for (i
= 0; i
<= s3
- s2
; i
++)
530 matrix
[i
] = register_param5( p
, s0
, s1
, i
, i
, s4
);
534 static void emit_arg( struct prog_src_register
*src
,
537 src
->File
= reg
.file
;
538 src
->Index
= reg
.idx
;
539 src
->Swizzle
= reg
.swz
;
540 src
->Negate
= reg
.negate
? NEGATE_XYZW
: NEGATE_NONE
;
542 /* Check that bitfield sizes aren't exceeded */
543 assert(src
->Index
== reg
.idx
);
547 static void emit_dst( struct prog_dst_register
*dst
,
548 struct ureg reg
, GLuint mask
)
550 dst
->File
= reg
.file
;
551 dst
->Index
= reg
.idx
;
552 /* allow zero as a shorthand for xyzw */
553 dst
->WriteMask
= mask
? mask
: WRITEMASK_XYZW
;
554 /* Check that bitfield sizes aren't exceeded */
555 assert(dst
->Index
== reg
.idx
);
559 static void debug_insn( struct prog_instruction
*inst
, const char *fn
,
563 static const char *last_fn
;
570 printf("%d:\t", line
);
571 _mesa_print_instruction(inst
);
576 static void emit_op3fn(struct tnl_program
*p
,
587 struct prog_instruction
*inst
;
589 assert(p
->program
->Base
.NumInstructions
<= p
->max_inst
);
591 if (p
->program
->Base
.NumInstructions
== p
->max_inst
) {
592 /* need to extend the program's instruction array */
593 struct prog_instruction
*newInst
;
595 /* double the size */
598 newInst
= _mesa_alloc_instructions(p
->max_inst
);
600 _mesa_error(NULL
, GL_OUT_OF_MEMORY
, "vertex program build");
604 _mesa_copy_instructions(newInst
,
605 p
->program
->Base
.Instructions
,
606 p
->program
->Base
.NumInstructions
);
608 _mesa_free_instructions(p
->program
->Base
.Instructions
,
609 p
->program
->Base
.NumInstructions
);
611 p
->program
->Base
.Instructions
= newInst
;
614 nr
= p
->program
->Base
.NumInstructions
++;
616 inst
= &p
->program
->Base
.Instructions
[nr
];
617 inst
->Opcode
= (enum prog_opcode
) op
;
619 emit_arg( &inst
->SrcReg
[0], src0
);
620 emit_arg( &inst
->SrcReg
[1], src1
);
621 emit_arg( &inst
->SrcReg
[2], src2
);
623 emit_dst( &inst
->DstReg
, dest
, mask
);
625 debug_insn(inst
, fn
, line
);
629 #define emit_op3(p, op, dst, mask, src0, src1, src2) \
630 emit_op3fn(p, op, dst, mask, src0, src1, src2, __func__, __LINE__)
632 #define emit_op2(p, op, dst, mask, src0, src1) \
633 emit_op3fn(p, op, dst, mask, src0, src1, undef, __func__, __LINE__)
635 #define emit_op1(p, op, dst, mask, src0) \
636 emit_op3fn(p, op, dst, mask, src0, undef, undef, __func__, __LINE__)
639 static struct ureg
make_temp( struct tnl_program
*p
, struct ureg reg
)
641 if (reg
.file
== PROGRAM_TEMPORARY
&&
642 !(p
->temp_reserved
& (1<<reg
.idx
)))
645 struct ureg temp
= get_temp(p
);
646 emit_op1(p
, OPCODE_MOV
, temp
, 0, reg
);
652 /* Currently no tracking performed of input/output/register size or
653 * active elements. Could be used to reduce these operations, as
654 * could the matrix type.
656 static void emit_matrix_transform_vec4( struct tnl_program
*p
,
658 const struct ureg
*mat
,
661 emit_op2(p
, OPCODE_DP4
, dest
, WRITEMASK_X
, src
, mat
[0]);
662 emit_op2(p
, OPCODE_DP4
, dest
, WRITEMASK_Y
, src
, mat
[1]);
663 emit_op2(p
, OPCODE_DP4
, dest
, WRITEMASK_Z
, src
, mat
[2]);
664 emit_op2(p
, OPCODE_DP4
, dest
, WRITEMASK_W
, src
, mat
[3]);
668 /* This version is much easier to implement if writemasks are not
669 * supported natively on the target or (like SSE), the target doesn't
670 * have a clean/obvious dotproduct implementation.
672 static void emit_transpose_matrix_transform_vec4( struct tnl_program
*p
,
674 const struct ureg
*mat
,
679 if (dest
.file
!= PROGRAM_TEMPORARY
)
684 emit_op2(p
, OPCODE_MUL
, tmp
, 0, swizzle1(src
,X
), mat
[0]);
685 emit_op3(p
, OPCODE_MAD
, tmp
, 0, swizzle1(src
,Y
), mat
[1], tmp
);
686 emit_op3(p
, OPCODE_MAD
, tmp
, 0, swizzle1(src
,Z
), mat
[2], tmp
);
687 emit_op3(p
, OPCODE_MAD
, dest
, 0, swizzle1(src
,W
), mat
[3], tmp
);
689 if (dest
.file
!= PROGRAM_TEMPORARY
)
690 release_temp(p
, tmp
);
694 static void emit_matrix_transform_vec3( struct tnl_program
*p
,
696 const struct ureg
*mat
,
699 emit_op2(p
, OPCODE_DP3
, dest
, WRITEMASK_X
, src
, mat
[0]);
700 emit_op2(p
, OPCODE_DP3
, dest
, WRITEMASK_Y
, src
, mat
[1]);
701 emit_op2(p
, OPCODE_DP3
, dest
, WRITEMASK_Z
, src
, mat
[2]);
705 static void emit_normalize_vec3( struct tnl_program
*p
,
709 struct ureg tmp
= get_temp(p
);
710 emit_op2(p
, OPCODE_DP3
, tmp
, WRITEMASK_X
, src
, src
);
711 emit_op1(p
, OPCODE_RSQ
, tmp
, WRITEMASK_X
, tmp
);
712 emit_op2(p
, OPCODE_MUL
, dest
, 0, src
, swizzle1(tmp
, X
));
713 release_temp(p
, tmp
);
717 static void emit_passthrough( struct tnl_program
*p
,
721 struct ureg out
= register_output(p
, output
);
722 emit_op1(p
, OPCODE_MOV
, out
, 0, register_input(p
, input
));
726 static struct ureg
get_eye_position( struct tnl_program
*p
)
728 if (is_undef(p
->eye_position
)) {
729 struct ureg pos
= register_input( p
, VERT_ATTRIB_POS
);
730 struct ureg modelview
[4];
732 p
->eye_position
= reserve_temp(p
);
734 if (p
->mvp_with_dp4
) {
735 register_matrix_param5( p
, STATE_MODELVIEW_MATRIX
, 0, 0, 3,
738 emit_matrix_transform_vec4(p
, p
->eye_position
, modelview
, pos
);
741 register_matrix_param5( p
, STATE_MODELVIEW_MATRIX
, 0, 0, 3,
742 STATE_MATRIX_TRANSPOSE
, modelview
);
744 emit_transpose_matrix_transform_vec4(p
, p
->eye_position
, modelview
, pos
);
748 return p
->eye_position
;
752 static struct ureg
get_eye_position_z( struct tnl_program
*p
)
754 if (!is_undef(p
->eye_position
))
755 return swizzle1(p
->eye_position
, Z
);
757 if (is_undef(p
->eye_position_z
)) {
758 struct ureg pos
= register_input( p
, VERT_ATTRIB_POS
);
759 struct ureg modelview
[4];
761 p
->eye_position_z
= reserve_temp(p
);
763 register_matrix_param5( p
, STATE_MODELVIEW_MATRIX
, 0, 0, 3,
766 emit_op2(p
, OPCODE_DP4
, p
->eye_position_z
, 0, pos
, modelview
[2]);
769 return p
->eye_position_z
;
773 static struct ureg
get_eye_position_normalized( struct tnl_program
*p
)
775 if (is_undef(p
->eye_position_normalized
)) {
776 struct ureg eye
= get_eye_position(p
);
777 p
->eye_position_normalized
= reserve_temp(p
);
778 emit_normalize_vec3(p
, p
->eye_position_normalized
, eye
);
781 return p
->eye_position_normalized
;
785 static struct ureg
get_transformed_normal( struct tnl_program
*p
)
787 if (is_undef(p
->transformed_normal
) &&
788 !p
->state
->need_eye_coords
&&
789 !p
->state
->normalize
&&
790 !(p
->state
->need_eye_coords
== p
->state
->rescale_normals
))
792 p
->transformed_normal
= register_input(p
, VERT_ATTRIB_NORMAL
);
794 else if (is_undef(p
->transformed_normal
))
796 struct ureg normal
= register_input(p
, VERT_ATTRIB_NORMAL
);
797 struct ureg mvinv
[3];
798 struct ureg transformed_normal
= reserve_temp(p
);
800 if (p
->state
->need_eye_coords
) {
801 register_matrix_param5( p
, STATE_MODELVIEW_MATRIX
, 0, 0, 2,
802 STATE_MATRIX_INVTRANS
, mvinv
);
804 /* Transform to eye space:
806 emit_matrix_transform_vec3( p
, transformed_normal
, mvinv
, normal
);
807 normal
= transformed_normal
;
810 /* Normalize/Rescale:
812 if (p
->state
->normalize
) {
813 emit_normalize_vec3( p
, transformed_normal
, normal
);
814 normal
= transformed_normal
;
816 else if (p
->state
->need_eye_coords
== p
->state
->rescale_normals
) {
817 /* This is already adjusted for eye/non-eye rendering:
819 struct ureg rescale
= register_param2(p
, STATE_INTERNAL
,
822 emit_op2( p
, OPCODE_MUL
, transformed_normal
, 0, normal
, rescale
);
823 normal
= transformed_normal
;
826 assert(normal
.file
== PROGRAM_TEMPORARY
);
827 p
->transformed_normal
= normal
;
830 return p
->transformed_normal
;
834 static void build_hpos( struct tnl_program
*p
)
836 struct ureg pos
= register_input( p
, VERT_ATTRIB_POS
);
837 struct ureg hpos
= register_output( p
, VARYING_SLOT_POS
);
840 if (p
->mvp_with_dp4
) {
841 register_matrix_param5( p
, STATE_MVP_MATRIX
, 0, 0, 3,
843 emit_matrix_transform_vec4( p
, hpos
, mvp
, pos
);
846 register_matrix_param5( p
, STATE_MVP_MATRIX
, 0, 0, 3,
847 STATE_MATRIX_TRANSPOSE
, mvp
);
848 emit_transpose_matrix_transform_vec4( p
, hpos
, mvp
, pos
);
853 static GLuint
material_attrib( GLuint side
, GLuint property
)
855 return (property
- STATE_AMBIENT
) * 2 + side
;
860 * Get a bitmask of which material values vary on a per-vertex basis.
862 static void set_material_flags( struct tnl_program
*p
)
864 p
->color_materials
= 0;
867 if (p
->state
->varying_vp_inputs
& VERT_BIT_COLOR0
) {
869 p
->color_materials
= p
->state
->light_color_material_mask
;
872 p
->materials
|= (p
->state
->varying_vp_inputs
>> VERT_ATTRIB_GENERIC0
);
876 static struct ureg
get_material( struct tnl_program
*p
, GLuint side
,
879 GLuint attrib
= material_attrib(side
, property
);
881 if (p
->color_materials
& (1<<attrib
))
882 return register_input(p
, VERT_ATTRIB_COLOR0
);
883 else if (p
->materials
& (1<<attrib
)) {
884 /* Put material values in the GENERIC slots -- they are not used
885 * for anything in fixed function mode.
887 return register_input( p
, attrib
+ VERT_ATTRIB_GENERIC0
);
890 return register_param3( p
, STATE_MATERIAL
, side
, property
);
893 #define SCENE_COLOR_BITS(side) (( MAT_BIT_FRONT_EMISSION | \
894 MAT_BIT_FRONT_AMBIENT | \
895 MAT_BIT_FRONT_DIFFUSE) << (side))
899 * Either return a precalculated constant value or emit code to
900 * calculate these values dynamically in the case where material calls
901 * are present between begin/end pairs.
903 * Probably want to shift this to the program compilation phase - if
904 * we always emitted the calculation here, a smart compiler could
905 * detect that it was constant (given a certain set of inputs), and
906 * lift it out of the main loop. That way the programs created here
907 * would be independent of the vertex_buffer details.
909 static struct ureg
get_scenecolor( struct tnl_program
*p
, GLuint side
)
911 if (p
->materials
& SCENE_COLOR_BITS(side
)) {
912 struct ureg lm_ambient
= register_param1(p
, STATE_LIGHTMODEL_AMBIENT
);
913 struct ureg material_emission
= get_material(p
, side
, STATE_EMISSION
);
914 struct ureg material_ambient
= get_material(p
, side
, STATE_AMBIENT
);
915 struct ureg material_diffuse
= get_material(p
, side
, STATE_DIFFUSE
);
916 struct ureg tmp
= make_temp(p
, material_diffuse
);
917 emit_op3(p
, OPCODE_MAD
, tmp
, WRITEMASK_XYZ
, lm_ambient
,
918 material_ambient
, material_emission
);
922 return register_param2( p
, STATE_LIGHTMODEL_SCENECOLOR
, side
);
926 static struct ureg
get_lightprod( struct tnl_program
*p
, GLuint light
,
927 GLuint side
, GLuint property
)
929 GLuint attrib
= material_attrib(side
, property
);
930 if (p
->materials
& (1<<attrib
)) {
931 struct ureg light_value
=
932 register_param3(p
, STATE_LIGHT
, light
, property
);
933 struct ureg material_value
= get_material(p
, side
, property
);
934 struct ureg tmp
= get_temp(p
);
935 emit_op2(p
, OPCODE_MUL
, tmp
, 0, light_value
, material_value
);
939 return register_param4(p
, STATE_LIGHTPROD
, light
, side
, property
);
943 static struct ureg
calculate_light_attenuation( struct tnl_program
*p
,
948 struct ureg attenuation
= register_param3(p
, STATE_LIGHT
, i
,
950 struct ureg att
= undef
;
952 /* Calculate spot attenuation:
954 if (!p
->state
->unit
[i
].light_spotcutoff_is_180
) {
955 struct ureg spot_dir_norm
= register_param3(p
, STATE_INTERNAL
,
956 STATE_LIGHT_SPOT_DIR_NORMALIZED
, i
);
957 struct ureg spot
= get_temp(p
);
958 struct ureg slt
= get_temp(p
);
962 emit_op2(p
, OPCODE_DP3
, spot
, 0, negate(VPpli
), spot_dir_norm
);
963 emit_op2(p
, OPCODE_SLT
, slt
, 0, swizzle1(spot_dir_norm
,W
), spot
);
964 emit_op2(p
, OPCODE_POW
, spot
, 0, absolute(spot
), swizzle1(attenuation
, W
));
965 emit_op2(p
, OPCODE_MUL
, att
, 0, slt
, spot
);
967 release_temp(p
, spot
);
968 release_temp(p
, slt
);
971 /* Calculate distance attenuation(See formula (2.4) at glspec 2.1 page 62):
973 * Skip the calucation when _dist_ is undefined(light_eyepos3_is_zero)
975 if (p
->state
->unit
[i
].light_attenuated
&& !is_undef(dist
)) {
979 emit_op1(p
, OPCODE_RCP
, dist
, WRITEMASK_YZ
, dist
);
981 emit_op2(p
, OPCODE_MUL
, dist
, WRITEMASK_XZ
, dist
, swizzle1(dist
,Y
));
983 emit_op2(p
, OPCODE_DP3
, dist
, 0, attenuation
, dist
);
985 if (!p
->state
->unit
[i
].light_spotcutoff_is_180
) {
987 emit_op1(p
, OPCODE_RCP
, dist
, 0, dist
);
988 /* spot-atten * dist-atten */
989 emit_op2(p
, OPCODE_MUL
, att
, 0, dist
, att
);
993 emit_op1(p
, OPCODE_RCP
, att
, 0, dist
);
1003 * lit.y = MAX(0, dots.x)
1004 * lit.z = SLT(0, dots.x)
1006 static void emit_degenerate_lit( struct tnl_program
*p
,
1010 struct ureg id
= get_identity_param(p
); /* id = {0,0,0,1} */
1012 /* Note that lit.x & lit.w will not be examined. Note also that
1013 * dots.xyzw == dots.xxxx.
1016 /* MAX lit, id, dots;
1018 emit_op2(p
, OPCODE_MAX
, lit
, WRITEMASK_XYZW
, id
, dots
);
1020 /* result[2] = (in > 0 ? 1 : 0)
1021 * SLT lit.z, id.z, dots; # lit.z = (0 < dots.z) ? 1 : 0
1023 emit_op2(p
, OPCODE_SLT
, lit
, WRITEMASK_Z
, swizzle1(id
,Z
), dots
);
1027 /* Need to add some addtional parameters to allow lighting in object
1028 * space - STATE_SPOT_DIRECTION and STATE_HALF_VECTOR implicitly assume eye
1031 static void build_lighting( struct tnl_program
*p
)
1033 const GLboolean twoside
= p
->state
->light_twoside
;
1034 const GLboolean separate
= p
->state
->separate_specular
;
1035 GLuint nr_lights
= 0, count
= 0;
1036 struct ureg normal
= get_transformed_normal(p
);
1037 struct ureg lit
= get_temp(p
);
1038 struct ureg dots
= get_temp(p
);
1039 struct ureg _col0
= undef
, _col1
= undef
;
1040 struct ureg _bfc0
= undef
, _bfc1
= undef
;
1045 * dots.x = dot(normal, VPpli)
1046 * dots.y = dot(normal, halfAngle)
1047 * dots.z = back.shininess
1048 * dots.w = front.shininess
1051 for (i
= 0; i
< MAX_LIGHTS
; i
++)
1052 if (p
->state
->unit
[i
].light_enabled
)
1055 set_material_flags(p
);
1058 if (!p
->state
->material_shininess_is_zero
) {
1059 struct ureg shininess
= get_material(p
, 0, STATE_SHININESS
);
1060 emit_op1(p
, OPCODE_MOV
, dots
, WRITEMASK_W
, swizzle1(shininess
,X
));
1061 release_temp(p
, shininess
);
1064 _col0
= make_temp(p
, get_scenecolor(p
, 0));
1066 _col1
= make_temp(p
, get_identity_param(p
));
1072 if (!p
->state
->material_shininess_is_zero
) {
1073 /* Note that we negate the back-face specular exponent here.
1074 * The negation will be un-done later in the back-face code below.
1076 struct ureg shininess
= get_material(p
, 1, STATE_SHININESS
);
1077 emit_op1(p
, OPCODE_MOV
, dots
, WRITEMASK_Z
,
1078 negate(swizzle1(shininess
,X
)));
1079 release_temp(p
, shininess
);
1082 _bfc0
= make_temp(p
, get_scenecolor(p
, 1));
1084 _bfc1
= make_temp(p
, get_identity_param(p
));
1089 /* If no lights, still need to emit the scenecolor.
1092 struct ureg res0
= register_output( p
, VARYING_SLOT_COL0
);
1093 emit_op1(p
, OPCODE_MOV
, res0
, 0, _col0
);
1097 struct ureg res1
= register_output( p
, VARYING_SLOT_COL1
);
1098 emit_op1(p
, OPCODE_MOV
, res1
, 0, _col1
);
1102 struct ureg res0
= register_output( p
, VARYING_SLOT_BFC0
);
1103 emit_op1(p
, OPCODE_MOV
, res0
, 0, _bfc0
);
1106 if (twoside
&& separate
) {
1107 struct ureg res1
= register_output( p
, VARYING_SLOT_BFC1
);
1108 emit_op1(p
, OPCODE_MOV
, res1
, 0, _bfc1
);
1111 if (nr_lights
== 0) {
1116 for (i
= 0; i
< MAX_LIGHTS
; i
++) {
1117 if (p
->state
->unit
[i
].light_enabled
) {
1118 struct ureg half
= undef
;
1119 struct ureg att
= undef
, VPpli
= undef
;
1120 struct ureg dist
= undef
;
1123 if (p
->state
->unit
[i
].light_eyepos3_is_zero
) {
1124 VPpli
= register_param3(p
, STATE_INTERNAL
,
1125 STATE_LIGHT_POSITION_NORMALIZED
, i
);
1127 struct ureg Ppli
= register_param3(p
, STATE_INTERNAL
,
1128 STATE_LIGHT_POSITION
, i
);
1129 struct ureg V
= get_eye_position(p
);
1131 VPpli
= get_temp(p
);
1134 /* Calculate VPpli vector
1136 emit_op2(p
, OPCODE_SUB
, VPpli
, 0, Ppli
, V
);
1138 /* Normalize VPpli. The dist value also used in
1139 * attenuation below.
1141 emit_op2(p
, OPCODE_DP3
, dist
, 0, VPpli
, VPpli
);
1142 emit_op1(p
, OPCODE_RSQ
, dist
, 0, dist
);
1143 emit_op2(p
, OPCODE_MUL
, VPpli
, 0, VPpli
, dist
);
1146 /* Calculate attenuation:
1148 att
= calculate_light_attenuation(p
, i
, VPpli
, dist
);
1149 release_temp(p
, dist
);
1151 /* Calculate viewer direction, or use infinite viewer:
1153 if (!p
->state
->material_shininess_is_zero
) {
1154 if (p
->state
->light_local_viewer
) {
1155 struct ureg eye_hat
= get_eye_position_normalized(p
);
1157 emit_op2(p
, OPCODE_SUB
, half
, 0, VPpli
, eye_hat
);
1158 emit_normalize_vec3(p
, half
, half
);
1159 } else if (p
->state
->unit
[i
].light_eyepos3_is_zero
) {
1160 half
= register_param3(p
, STATE_INTERNAL
,
1161 STATE_LIGHT_HALF_VECTOR
, i
);
1163 struct ureg z_dir
= swizzle(get_identity_param(p
),X
,Y
,W
,Z
);
1165 emit_op2(p
, OPCODE_ADD
, half
, 0, VPpli
, z_dir
);
1166 emit_normalize_vec3(p
, half
, half
);
1170 /* Calculate dot products:
1172 if (p
->state
->material_shininess_is_zero
) {
1173 emit_op2(p
, OPCODE_DP3
, dots
, 0, normal
, VPpli
);
1176 emit_op2(p
, OPCODE_DP3
, dots
, WRITEMASK_X
, normal
, VPpli
);
1177 emit_op2(p
, OPCODE_DP3
, dots
, WRITEMASK_Y
, normal
, half
);
1180 /* Front face lighting:
1183 struct ureg ambient
= get_lightprod(p
, i
, 0, STATE_AMBIENT
);
1184 struct ureg diffuse
= get_lightprod(p
, i
, 0, STATE_DIFFUSE
);
1185 struct ureg specular
= get_lightprod(p
, i
, 0, STATE_SPECULAR
);
1186 struct ureg res0
, res1
;
1187 GLuint mask0
, mask1
;
1189 if (count
== nr_lights
) {
1191 mask0
= WRITEMASK_XYZ
;
1192 mask1
= WRITEMASK_XYZ
;
1193 res0
= register_output( p
, VARYING_SLOT_COL0
);
1194 res1
= register_output( p
, VARYING_SLOT_COL1
);
1198 mask1
= WRITEMASK_XYZ
;
1200 res1
= register_output( p
, VARYING_SLOT_COL0
);
1210 if (!is_undef(att
)) {
1211 /* light is attenuated by distance */
1212 emit_op1(p
, OPCODE_LIT
, lit
, 0, dots
);
1213 emit_op2(p
, OPCODE_MUL
, lit
, 0, lit
, att
);
1214 emit_op3(p
, OPCODE_MAD
, _col0
, 0, swizzle1(lit
,X
), ambient
, _col0
);
1216 else if (!p
->state
->material_shininess_is_zero
) {
1217 /* there's a non-zero specular term */
1218 emit_op1(p
, OPCODE_LIT
, lit
, 0, dots
);
1219 emit_op2(p
, OPCODE_ADD
, _col0
, 0, ambient
, _col0
);
1222 /* no attenutation, no specular */
1223 emit_degenerate_lit(p
, lit
, dots
);
1224 emit_op2(p
, OPCODE_ADD
, _col0
, 0, ambient
, _col0
);
1227 emit_op3(p
, OPCODE_MAD
, res0
, mask0
, swizzle1(lit
,Y
), diffuse
, _col0
);
1228 emit_op3(p
, OPCODE_MAD
, res1
, mask1
, swizzle1(lit
,Z
), specular
, _col1
);
1230 release_temp(p
, ambient
);
1231 release_temp(p
, diffuse
);
1232 release_temp(p
, specular
);
1235 /* Back face lighting:
1238 struct ureg ambient
= get_lightprod(p
, i
, 1, STATE_AMBIENT
);
1239 struct ureg diffuse
= get_lightprod(p
, i
, 1, STATE_DIFFUSE
);
1240 struct ureg specular
= get_lightprod(p
, i
, 1, STATE_SPECULAR
);
1241 struct ureg res0
, res1
;
1242 GLuint mask0
, mask1
;
1244 if (count
== nr_lights
) {
1246 mask0
= WRITEMASK_XYZ
;
1247 mask1
= WRITEMASK_XYZ
;
1248 res0
= register_output( p
, VARYING_SLOT_BFC0
);
1249 res1
= register_output( p
, VARYING_SLOT_BFC1
);
1253 mask1
= WRITEMASK_XYZ
;
1255 res1
= register_output( p
, VARYING_SLOT_BFC0
);
1265 /* For the back face we need to negate the X and Y component
1266 * dot products. dots.Z has the negated back-face specular
1267 * exponent. We swizzle that into the W position. This
1268 * negation makes the back-face specular term positive again.
1270 dots
= negate(swizzle(dots
,X
,Y
,W
,Z
));
1272 if (!is_undef(att
)) {
1273 emit_op1(p
, OPCODE_LIT
, lit
, 0, dots
);
1274 emit_op2(p
, OPCODE_MUL
, lit
, 0, lit
, att
);
1275 emit_op3(p
, OPCODE_MAD
, _bfc0
, 0, swizzle1(lit
,X
), ambient
, _bfc0
);
1277 else if (!p
->state
->material_shininess_is_zero
) {
1278 emit_op1(p
, OPCODE_LIT
, lit
, 0, dots
);
1279 emit_op2(p
, OPCODE_ADD
, _bfc0
, 0, ambient
, _bfc0
); /**/
1282 emit_degenerate_lit(p
, lit
, dots
);
1283 emit_op2(p
, OPCODE_ADD
, _bfc0
, 0, ambient
, _bfc0
);
1286 emit_op3(p
, OPCODE_MAD
, res0
, mask0
, swizzle1(lit
,Y
), diffuse
, _bfc0
);
1287 emit_op3(p
, OPCODE_MAD
, res1
, mask1
, swizzle1(lit
,Z
), specular
, _bfc1
);
1288 /* restore dots to its original state for subsequent lights
1289 * by negating and swizzling again.
1291 dots
= negate(swizzle(dots
,X
,Y
,W
,Z
));
1293 release_temp(p
, ambient
);
1294 release_temp(p
, diffuse
);
1295 release_temp(p
, specular
);
1298 release_temp(p
, half
);
1299 release_temp(p
, VPpli
);
1300 release_temp(p
, att
);
1308 static void build_fog( struct tnl_program
*p
)
1310 struct ureg fog
= register_output(p
, VARYING_SLOT_FOGC
);
1313 if (p
->state
->fog_source_is_depth
) {
1315 switch (p
->state
->fog_distance_mode
) {
1316 case FDM_EYE_RADIAL
: /* Z = sqrt(Xe*Xe + Ye*Ye + Ze*Ze) */
1317 input
= get_eye_position(p
);
1318 emit_op2(p
, OPCODE_DP3
, fog
, WRITEMASK_X
, input
, input
);
1319 emit_op1(p
, OPCODE_RSQ
, fog
, WRITEMASK_X
, fog
);
1320 emit_op1(p
, OPCODE_RCP
, fog
, WRITEMASK_X
, fog
);
1322 case FDM_EYE_PLANE
: /* Z = Ze */
1323 input
= get_eye_position_z(p
);
1324 emit_op1(p
, OPCODE_MOV
, fog
, WRITEMASK_X
, input
);
1326 case FDM_EYE_PLANE_ABS
: /* Z = abs(Ze) */
1327 input
= get_eye_position_z(p
);
1328 emit_op1(p
, OPCODE_ABS
, fog
, WRITEMASK_X
, input
);
1331 assert(!"Bad fog mode in build_fog()");
1337 input
= swizzle1(register_input(p
, VERT_ATTRIB_FOG
), X
);
1338 emit_op1(p
, OPCODE_ABS
, fog
, WRITEMASK_X
, input
);
1341 emit_op1(p
, OPCODE_MOV
, fog
, WRITEMASK_YZW
, get_identity_param(p
));
1345 static void build_reflect_texgen( struct tnl_program
*p
,
1349 struct ureg normal
= get_transformed_normal(p
);
1350 struct ureg eye_hat
= get_eye_position_normalized(p
);
1351 struct ureg tmp
= get_temp(p
);
1354 emit_op2(p
, OPCODE_DP3
, tmp
, 0, normal
, eye_hat
);
1356 emit_op2(p
, OPCODE_ADD
, tmp
, 0, tmp
, tmp
);
1358 emit_op3(p
, OPCODE_MAD
, dest
, writemask
, negate(tmp
), normal
, eye_hat
);
1360 release_temp(p
, tmp
);
1364 static void build_sphere_texgen( struct tnl_program
*p
,
1368 struct ureg normal
= get_transformed_normal(p
);
1369 struct ureg eye_hat
= get_eye_position_normalized(p
);
1370 struct ureg tmp
= get_temp(p
);
1371 struct ureg half
= register_scalar_const(p
, .5);
1372 struct ureg r
= get_temp(p
);
1373 struct ureg inv_m
= get_temp(p
);
1374 struct ureg id
= get_identity_param(p
);
1376 /* Could share the above calculations, but it would be
1377 * a fairly odd state for someone to set (both sphere and
1378 * reflection active for different texture coordinate
1379 * components. Of course - if two texture units enable
1380 * reflect and/or sphere, things start to tilt in favour
1381 * of seperating this out:
1385 emit_op2(p
, OPCODE_DP3
, tmp
, 0, normal
, eye_hat
);
1387 emit_op2(p
, OPCODE_ADD
, tmp
, 0, tmp
, tmp
);
1389 emit_op3(p
, OPCODE_MAD
, r
, 0, negate(tmp
), normal
, eye_hat
);
1391 emit_op2(p
, OPCODE_ADD
, tmp
, 0, r
, swizzle(id
,X
,Y
,W
,Z
));
1392 /* rx^2 + ry^2 + (rz+1)^2 */
1393 emit_op2(p
, OPCODE_DP3
, tmp
, 0, tmp
, tmp
);
1395 emit_op1(p
, OPCODE_RSQ
, tmp
, 0, tmp
);
1397 emit_op2(p
, OPCODE_MUL
, inv_m
, 0, tmp
, half
);
1399 emit_op3(p
, OPCODE_MAD
, dest
, writemask
, r
, inv_m
, half
);
1401 release_temp(p
, tmp
);
1403 release_temp(p
, inv_m
);
1407 static void build_texture_transform( struct tnl_program
*p
)
1411 for (i
= 0; i
< MAX_TEXTURE_COORD_UNITS
; i
++) {
1413 if (!(p
->state
->fragprog_inputs_read
& VARYING_BIT_TEX(i
)))
1416 if (p
->state
->unit
[i
].coord_replace
)
1419 if (p
->state
->unit
[i
].texgen_enabled
||
1420 p
->state
->unit
[i
].texmat_enabled
) {
1422 GLuint texmat_enabled
= p
->state
->unit
[i
].texmat_enabled
;
1423 struct ureg out
= register_output(p
, VARYING_SLOT_TEX0
+ i
);
1424 struct ureg out_texgen
= undef
;
1426 if (p
->state
->unit
[i
].texgen_enabled
) {
1427 GLuint copy_mask
= 0;
1428 GLuint sphere_mask
= 0;
1429 GLuint reflect_mask
= 0;
1430 GLuint normal_mask
= 0;
1434 out_texgen
= get_temp(p
);
1438 modes
[0] = p
->state
->unit
[i
].texgen_mode0
;
1439 modes
[1] = p
->state
->unit
[i
].texgen_mode1
;
1440 modes
[2] = p
->state
->unit
[i
].texgen_mode2
;
1441 modes
[3] = p
->state
->unit
[i
].texgen_mode3
;
1443 for (j
= 0; j
< 4; j
++) {
1445 case TXG_OBJ_LINEAR
: {
1446 struct ureg obj
= register_input(p
, VERT_ATTRIB_POS
);
1448 register_param3(p
, STATE_TEXGEN
, i
,
1449 STATE_TEXGEN_OBJECT_S
+ j
);
1451 emit_op2(p
, OPCODE_DP4
, out_texgen
, WRITEMASK_X
<< j
,
1455 case TXG_EYE_LINEAR
: {
1456 struct ureg eye
= get_eye_position(p
);
1458 register_param3(p
, STATE_TEXGEN
, i
,
1459 STATE_TEXGEN_EYE_S
+ j
);
1461 emit_op2(p
, OPCODE_DP4
, out_texgen
, WRITEMASK_X
<< j
,
1465 case TXG_SPHERE_MAP
:
1466 sphere_mask
|= WRITEMASK_X
<< j
;
1468 case TXG_REFLECTION_MAP
:
1469 reflect_mask
|= WRITEMASK_X
<< j
;
1471 case TXG_NORMAL_MAP
:
1472 normal_mask
|= WRITEMASK_X
<< j
;
1475 copy_mask
|= WRITEMASK_X
<< j
;
1480 build_sphere_texgen(p
, out_texgen
, sphere_mask
);
1484 build_reflect_texgen(p
, out_texgen
, reflect_mask
);
1488 struct ureg normal
= get_transformed_normal(p
);
1489 emit_op1(p
, OPCODE_MOV
, out_texgen
, normal_mask
, normal
);
1493 struct ureg in
= register_input(p
, VERT_ATTRIB_TEX0
+i
);
1494 emit_op1(p
, OPCODE_MOV
, out_texgen
, copy_mask
, in
);
1498 if (texmat_enabled
) {
1499 struct ureg texmat
[4];
1500 struct ureg in
= (!is_undef(out_texgen
) ?
1502 register_input(p
, VERT_ATTRIB_TEX0
+i
));
1503 if (p
->mvp_with_dp4
) {
1504 register_matrix_param5( p
, STATE_TEXTURE_MATRIX
, i
, 0, 3,
1506 emit_matrix_transform_vec4( p
, out
, texmat
, in
);
1509 register_matrix_param5( p
, STATE_TEXTURE_MATRIX
, i
, 0, 3,
1510 STATE_MATRIX_TRANSPOSE
, texmat
);
1511 emit_transpose_matrix_transform_vec4( p
, out
, texmat
, in
);
1518 emit_passthrough(p
, VERT_ATTRIB_TEX0
+i
, VARYING_SLOT_TEX0
+i
);
1525 * Point size attenuation computation.
1527 static void build_atten_pointsize( struct tnl_program
*p
)
1529 struct ureg eye
= get_eye_position_z(p
);
1530 struct ureg state_size
= register_param2(p
, STATE_INTERNAL
, STATE_POINT_SIZE_CLAMPED
);
1531 struct ureg state_attenuation
= register_param1(p
, STATE_POINT_ATTENUATION
);
1532 struct ureg out
= register_output(p
, VARYING_SLOT_PSIZ
);
1533 struct ureg ut
= get_temp(p
);
1536 emit_op1(p
, OPCODE_ABS
, ut
, WRITEMASK_Y
, swizzle1(eye
, Z
));
1537 /* p1 + dist * (p2 + dist * p3); */
1538 emit_op3(p
, OPCODE_MAD
, ut
, WRITEMASK_X
, swizzle1(ut
, Y
),
1539 swizzle1(state_attenuation
, Z
), swizzle1(state_attenuation
, Y
));
1540 emit_op3(p
, OPCODE_MAD
, ut
, WRITEMASK_X
, swizzle1(ut
, Y
),
1541 ut
, swizzle1(state_attenuation
, X
));
1543 /* 1 / sqrt(factor) */
1544 emit_op1(p
, OPCODE_RSQ
, ut
, WRITEMASK_X
, ut
);
1547 /* out = pointSize / sqrt(factor) */
1548 emit_op2(p
, OPCODE_MUL
, out
, WRITEMASK_X
, ut
, state_size
);
1550 /* this is a good place to clamp the point size since there's likely
1551 * no hardware registers to clamp point size at rasterization time.
1553 emit_op2(p
, OPCODE_MUL
, ut
, WRITEMASK_X
, ut
, state_size
);
1554 emit_op2(p
, OPCODE_MAX
, ut
, WRITEMASK_X
, ut
, swizzle1(state_size
, Y
));
1555 emit_op2(p
, OPCODE_MIN
, out
, WRITEMASK_X
, ut
, swizzle1(state_size
, Z
));
1558 release_temp(p
, ut
);
1563 * Pass-though per-vertex point size, from user's point size array.
1565 static void build_array_pointsize( struct tnl_program
*p
)
1567 struct ureg in
= register_input(p
, VERT_ATTRIB_POINT_SIZE
);
1568 struct ureg out
= register_output(p
, VARYING_SLOT_PSIZ
);
1569 emit_op1(p
, OPCODE_MOV
, out
, WRITEMASK_X
, in
);
1573 static void build_tnl_program( struct tnl_program
*p
)
1575 /* Emit the program, starting with the modelview, projection transforms:
1579 /* Lighting calculations:
1581 if (p
->state
->fragprog_inputs_read
& (VARYING_BIT_COL0
|VARYING_BIT_COL1
)) {
1582 if (p
->state
->light_global_enabled
)
1585 if (p
->state
->fragprog_inputs_read
& VARYING_BIT_COL0
)
1586 emit_passthrough(p
, VERT_ATTRIB_COLOR0
, VARYING_SLOT_COL0
);
1588 if (p
->state
->fragprog_inputs_read
& VARYING_BIT_COL1
)
1589 emit_passthrough(p
, VERT_ATTRIB_COLOR1
, VARYING_SLOT_COL1
);
1593 if (p
->state
->fragprog_inputs_read
& VARYING_BIT_FOGC
)
1596 if (p
->state
->fragprog_inputs_read
& VARYING_BITS_TEX_ANY
)
1597 build_texture_transform(p
);
1599 if (p
->state
->point_attenuated
)
1600 build_atten_pointsize(p
);
1601 else if (p
->state
->point_array
)
1602 build_array_pointsize(p
);
1606 emit_op1(p
, OPCODE_END
, undef
, 0, undef
);
1617 create_new_program( const struct state_key
*key
,
1618 struct gl_vertex_program
*program
,
1619 GLboolean mvp_with_dp4
,
1622 struct tnl_program p
;
1624 memset(&p
, 0, sizeof(p
));
1626 p
.program
= program
;
1627 p
.eye_position
= undef
;
1628 p
.eye_position_z
= undef
;
1629 p
.eye_position_normalized
= undef
;
1630 p
.transformed_normal
= undef
;
1633 p
.mvp_with_dp4
= mvp_with_dp4
;
1635 if (max_temps
>= sizeof(int) * 8)
1636 p
.temp_reserved
= 0;
1638 p
.temp_reserved
= ~((1<<max_temps
)-1);
1640 /* Start by allocating 32 instructions.
1641 * If we need more, we'll grow the instruction array as needed.
1644 p
.program
->Base
.Instructions
= _mesa_alloc_instructions(p
.max_inst
);
1645 p
.program
->Base
.String
= NULL
;
1646 p
.program
->Base
.NumInstructions
=
1647 p
.program
->Base
.NumTemporaries
=
1648 p
.program
->Base
.NumParameters
=
1649 p
.program
->Base
.NumAttributes
= p
.program
->Base
.NumAddressRegs
= 0;
1650 p
.program
->Base
.Parameters
= _mesa_new_parameter_list();
1651 p
.program
->Base
.InputsRead
= 0;
1652 p
.program
->Base
.OutputsWritten
= 0;
1654 build_tnl_program( &p
);
1659 * Return a vertex program which implements the current fixed-function
1660 * transform/lighting/texgen operations.
1662 struct gl_vertex_program
*
1663 _mesa_get_fixed_func_vertex_program(struct gl_context
*ctx
)
1665 struct gl_vertex_program
*prog
;
1666 struct state_key key
;
1668 /* Grab all the relevant state and put it in a single structure:
1670 make_state_key(ctx
, &key
);
1672 /* Look for an already-prepared program for this state:
1674 prog
= gl_vertex_program(
1675 _mesa_search_program_cache(ctx
->VertexProgram
.Cache
, &key
, sizeof(key
)));
1678 /* OK, we'll have to build a new one */
1680 printf("Build new TNL program\n");
1682 prog
= gl_vertex_program(ctx
->Driver
.NewProgram(ctx
, GL_VERTEX_PROGRAM_ARB
, 0));
1686 create_new_program( &key
, prog
,
1687 ctx
->Const
.ShaderCompilerOptions
[MESA_SHADER_VERTEX
].OptimizeForAOS
,
1688 ctx
->Const
.Program
[MESA_SHADER_VERTEX
].MaxTemps
);
1690 if (ctx
->Driver
.ProgramStringNotify
)
1691 ctx
->Driver
.ProgramStringNotify( ctx
, GL_VERTEX_PROGRAM_ARB
,
1694 _mesa_program_cache_insert(ctx
, ctx
->VertexProgram
.Cache
,
1695 &key
, sizeof(key
), &prog
->Base
);