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.
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 "shader/program.h"
42 #include "shader/prog_cache.h"
43 #include "shader/prog_instruction.h"
44 #include "shader/prog_parameter.h"
45 #include "shader/prog_print.h"
46 #include "shader/prog_statevars.h"
50 unsigned light_global_enabled
:1;
51 unsigned light_local_viewer
:1;
52 unsigned light_twoside
:1;
53 unsigned light_color_material
:1;
54 unsigned light_color_material_mask
:12;
55 unsigned light_material_mask
:12;
56 unsigned material_shininess_is_zero
:1;
58 unsigned need_eye_coords
:1;
60 unsigned rescale_normals
:1;
61 unsigned fog_source_is_depth
:1;
62 unsigned tnl_do_vertex_fog
:1;
63 unsigned separate_specular
:1;
65 unsigned point_attenuated
:1;
66 unsigned texture_enabled_global
:1;
67 unsigned fragprog_inputs_read
:12;
70 unsigned light_enabled
:1;
71 unsigned light_eyepos3_is_zero
:1;
72 unsigned light_spotcutoff_is_180
:1;
73 unsigned light_attenuated
:1;
74 unsigned texunit_really_enabled
:1;
75 unsigned texmat_enabled
:1;
76 unsigned texgen_enabled
:4;
77 unsigned texgen_mode0
:4;
78 unsigned texgen_mode1
:4;
79 unsigned texgen_mode2
:4;
80 unsigned texgen_mode3
:4;
91 static GLuint
translate_fog_mode( GLenum mode
)
94 case GL_LINEAR
: return FOG_LINEAR
;
95 case GL_EXP
: return FOG_EXP
;
96 case GL_EXP2
: return FOG_EXP2
;
97 default: return FOG_NONE
;
102 #define TXG_OBJ_LINEAR 1
103 #define TXG_EYE_LINEAR 2
104 #define TXG_SPHERE_MAP 3
105 #define TXG_REFLECTION_MAP 4
106 #define TXG_NORMAL_MAP 5
108 static GLuint
translate_texgen( GLboolean enabled
, GLenum mode
)
114 case GL_OBJECT_LINEAR
: return TXG_OBJ_LINEAR
;
115 case GL_EYE_LINEAR
: return TXG_EYE_LINEAR
;
116 case GL_SPHERE_MAP
: return TXG_SPHERE_MAP
;
117 case GL_REFLECTION_MAP_NV
: return TXG_REFLECTION_MAP
;
118 case GL_NORMAL_MAP_NV
: return TXG_NORMAL_MAP
;
119 default: return TXG_NONE
;
125 * Returns bitmask of flags indicating which materials are set per-vertex
127 * XXX get these from the VBO...
130 tnl_get_per_vertex_materials(GLcontext
*ctx
)
132 GLbitfield mask
= 0x0;
134 TNLcontext
*tnl
= TNL_CONTEXT(ctx
);
135 struct vertex_buffer
*VB
= &tnl
->vb
;
138 for (i
= _TNL_FIRST_MAT
; i
<= _TNL_LAST_MAT
; i
++)
139 if (VB
->AttribPtr
[i
] && VB
->AttribPtr
[i
]->stride
)
140 mask
|= 1 << (i
- _TNL_FIRST_MAT
);
146 * Should fog be computed per-vertex?
149 tnl_get_per_vertex_fog(GLcontext
*ctx
)
152 TNLcontext
*tnl
= TNL_CONTEXT(ctx
);
153 return tnl
->_DoVertexFog
;
159 static GLboolean
check_active_shininess( GLcontext
*ctx
,
160 const struct state_key
*key
,
163 GLuint bit
= 1 << (MAT_ATTRIB_FRONT_SHININESS
+ side
);
165 if (key
->light_color_material_mask
& bit
)
168 if (key
->light_material_mask
& bit
)
171 if (ctx
->Light
.Material
.Attrib
[MAT_ATTRIB_FRONT_SHININESS
+ side
][0] != 0.0F
)
180 static struct state_key
*make_state_key( GLcontext
*ctx
)
182 const struct gl_fragment_program
*fp
;
183 struct state_key
*key
= CALLOC_STRUCT(state_key
);
186 fp
= ctx
->FragmentProgram
._Current
;
188 /* This now relies on texenvprogram.c being active:
192 key
->need_eye_coords
= ctx
->_NeedEyeCoords
;
194 key
->fragprog_inputs_read
= fp
->Base
.InputsRead
;
196 if (ctx
->RenderMode
== GL_FEEDBACK
) {
197 /* make sure the vertprog emits color and tex0 */
198 key
->fragprog_inputs_read
|= (FRAG_BIT_COL0
| FRAG_BIT_TEX0
);
201 key
->separate_specular
= (ctx
->Light
.Model
.ColorControl
==
202 GL_SEPARATE_SPECULAR_COLOR
);
204 if (ctx
->Light
.Enabled
) {
205 key
->light_global_enabled
= 1;
207 if (ctx
->Light
.Model
.LocalViewer
)
208 key
->light_local_viewer
= 1;
210 if (ctx
->Light
.Model
.TwoSide
)
211 key
->light_twoside
= 1;
213 if (ctx
->Light
.ColorMaterialEnabled
) {
214 key
->light_color_material
= 1;
215 key
->light_color_material_mask
= ctx
->Light
.ColorMaterialBitmask
;
218 key
->light_material_mask
= tnl_get_per_vertex_materials(ctx
);
220 for (i
= 0; i
< MAX_LIGHTS
; i
++) {
221 struct gl_light
*light
= &ctx
->Light
.Light
[i
];
223 if (light
->Enabled
) {
224 key
->unit
[i
].light_enabled
= 1;
226 if (light
->EyePosition
[3] == 0.0)
227 key
->unit
[i
].light_eyepos3_is_zero
= 1;
229 if (light
->SpotCutoff
== 180.0)
230 key
->unit
[i
].light_spotcutoff_is_180
= 1;
232 if (light
->ConstantAttenuation
!= 1.0 ||
233 light
->LinearAttenuation
!= 0.0 ||
234 light
->QuadraticAttenuation
!= 0.0)
235 key
->unit
[i
].light_attenuated
= 1;
239 if (check_active_shininess(ctx
, key
, 0)) {
240 key
->material_shininess_is_zero
= 0;
242 else if (key
->light_twoside
&&
243 check_active_shininess(ctx
, key
, 1)) {
244 key
->material_shininess_is_zero
= 0;
247 key
->material_shininess_is_zero
= 1;
251 if (ctx
->Transform
.Normalize
)
254 if (ctx
->Transform
.RescaleNormals
)
255 key
->rescale_normals
= 1;
257 key
->fog_mode
= translate_fog_mode(fp
->FogOption
);
259 if (ctx
->Fog
.FogCoordinateSource
== GL_FRAGMENT_DEPTH_EXT
)
260 key
->fog_source_is_depth
= 1;
262 key
->tnl_do_vertex_fog
= tnl_get_per_vertex_fog(ctx
);
264 if (ctx
->Point
._Attenuated
)
265 key
->point_attenuated
= 1;
267 if (ctx
->Texture
._TexGenEnabled
||
268 ctx
->Texture
._TexMatEnabled
||
269 ctx
->Texture
._EnabledUnits
)
270 key
->texture_enabled_global
= 1;
272 for (i
= 0; i
< MAX_TEXTURE_UNITS
; i
++) {
273 struct gl_texture_unit
*texUnit
= &ctx
->Texture
.Unit
[i
];
275 if (texUnit
->_ReallyEnabled
)
276 key
->unit
[i
].texunit_really_enabled
= 1;
278 if (ctx
->Texture
._TexMatEnabled
& ENABLE_TEXMAT(i
))
279 key
->unit
[i
].texmat_enabled
= 1;
281 if (texUnit
->TexGenEnabled
) {
282 key
->unit
[i
].texgen_enabled
= 1;
284 key
->unit
[i
].texgen_mode0
=
285 translate_texgen( texUnit
->TexGenEnabled
& (1<<0),
287 key
->unit
[i
].texgen_mode1
=
288 translate_texgen( texUnit
->TexGenEnabled
& (1<<1),
290 key
->unit
[i
].texgen_mode2
=
291 translate_texgen( texUnit
->TexGenEnabled
& (1<<2),
293 key
->unit
[i
].texgen_mode3
=
294 translate_texgen( texUnit
->TexGenEnabled
& (1<<3),
304 /* Very useful debugging tool - produces annotated listing of
305 * generated program with line/function references for each
306 * instruction back into this file:
310 /* Should be tunable by the driver - do we want to do matrix
311 * multiplications with DP4's or with MUL/MAD's? SSE works better
312 * with the latter, drivers may differ.
318 /* Use uregs to represent registers internally, translate to Mesa's
319 * expected formats on emit.
321 * NOTE: These are passed by value extensively in this file rather
322 * than as usual by pointer reference. If this disturbs you, try
323 * remembering they are just 32bits in size.
325 * GCC is smart enough to deal with these dword-sized structures in
326 * much the same way as if I had defined them as dwords and was using
327 * macros to access and set the fields. This is much nicer and easier
332 GLint idx
:8; /* relative addressing may be negative */
340 const struct state_key
*state
;
341 struct gl_vertex_program
*program
;
344 GLuint temp_reserved
;
346 struct ureg eye_position
;
347 struct ureg eye_position_z
;
348 struct ureg eye_position_normalized
;
349 struct ureg transformed_normal
;
350 struct ureg identity
;
353 GLuint color_materials
;
357 static const struct ureg undef
= {
375 static struct ureg
make_ureg(GLuint file
, GLint idx
)
381 reg
.swz
= SWIZZLE_NOOP
;
388 static struct ureg
negate( struct ureg reg
)
395 static struct ureg
swizzle( struct ureg reg
, int x
, int y
, int z
, int w
)
397 reg
.swz
= MAKE_SWIZZLE4(GET_SWZ(reg
.swz
, x
),
400 GET_SWZ(reg
.swz
, w
));
405 static struct ureg
swizzle1( struct ureg reg
, int x
)
407 return swizzle(reg
, x
, x
, x
, x
);
410 static struct ureg
get_temp( struct tnl_program
*p
)
412 int bit
= _mesa_ffs( ~p
->temp_in_use
);
414 _mesa_problem(NULL
, "%s: out of temporaries\n", __FILE__
);
418 if ((GLuint
) bit
> p
->program
->Base
.NumTemporaries
)
419 p
->program
->Base
.NumTemporaries
= bit
;
421 p
->temp_in_use
|= 1<<(bit
-1);
422 return make_ureg(PROGRAM_TEMPORARY
, bit
-1);
425 static struct ureg
reserve_temp( struct tnl_program
*p
)
427 struct ureg temp
= get_temp( p
);
428 p
->temp_reserved
|= 1<<temp
.idx
;
432 static void release_temp( struct tnl_program
*p
, struct ureg reg
)
434 if (reg
.file
== PROGRAM_TEMPORARY
) {
435 p
->temp_in_use
&= ~(1<<reg
.idx
);
436 p
->temp_in_use
|= p
->temp_reserved
; /* can't release reserved temps */
440 static void release_temps( struct tnl_program
*p
)
442 p
->temp_in_use
= p
->temp_reserved
;
447 static struct ureg
register_input( struct tnl_program
*p
, GLuint input
)
449 p
->program
->Base
.InputsRead
|= (1<<input
);
450 return make_ureg(PROGRAM_INPUT
, input
);
453 static struct ureg
register_output( struct tnl_program
*p
, GLuint output
)
455 p
->program
->Base
.OutputsWritten
|= (1<<output
);
456 return make_ureg(PROGRAM_OUTPUT
, output
);
459 static struct ureg
register_const4f( struct tnl_program
*p
,
472 idx
= _mesa_add_unnamed_constant( p
->program
->Base
.Parameters
, values
, 4,
474 ASSERT(swizzle
== SWIZZLE_NOOP
);
475 return make_ureg(PROGRAM_CONSTANT
, idx
);
478 #define register_const1f(p, s0) register_const4f(p, s0, 0, 0, 1)
479 #define register_scalar_const(p, s0) register_const4f(p, s0, s0, s0, s0)
480 #define register_const2f(p, s0, s1) register_const4f(p, s0, s1, 0, 1)
481 #define register_const3f(p, s0, s1, s2) register_const4f(p, s0, s1, s2, 1)
483 static GLboolean
is_undef( struct ureg reg
)
485 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 struct ureg
register_param5(struct tnl_program
*p
,
503 gl_state_index tokens
[STATE_LENGTH
];
510 idx
= _mesa_add_state_reference( p
->program
->Base
.Parameters
, tokens
);
511 return make_ureg(PROGRAM_STATE_VAR
, idx
);
515 #define register_param1(p,s0) register_param5(p,s0,0,0,0,0)
516 #define register_param2(p,s0,s1) register_param5(p,s0,s1,0,0,0)
517 #define register_param3(p,s0,s1,s2) register_param5(p,s0,s1,s2,0,0)
518 #define register_param4(p,s0,s1,s2,s3) register_param5(p,s0,s1,s2,s3,0)
521 static void register_matrix_param5( struct tnl_program
*p
,
522 GLint s0
, /* modelview, projection, etc */
523 GLint s1
, /* texture matrix number */
524 GLint s2
, /* first row */
525 GLint s3
, /* last row */
526 GLint s4
, /* inverse, transpose, etc */
527 struct ureg
*matrix
)
531 /* This is a bit sad as the support is there to pull the whole
532 * matrix out in one go:
534 for (i
= 0; i
<= s3
- s2
; i
++)
535 matrix
[i
] = register_param5( p
, s0
, s1
, i
, i
, s4
);
539 static void emit_arg( struct prog_src_register
*src
,
542 src
->File
= reg
.file
;
543 src
->Index
= reg
.idx
;
544 src
->Swizzle
= reg
.swz
;
545 src
->NegateBase
= reg
.negate
? NEGATE_XYZW
: 0;
551 static void emit_dst( struct prog_dst_register
*dst
,
552 struct ureg reg
, GLuint mask
)
554 dst
->File
= reg
.file
;
555 dst
->Index
= reg
.idx
;
556 /* allow zero as a shorthand for xyzw */
557 dst
->WriteMask
= mask
? mask
: WRITEMASK_XYZW
;
558 dst
->CondMask
= COND_TR
; /* always pass cond test */
559 dst
->CondSwizzle
= SWIZZLE_NOOP
;
564 static void debug_insn( struct prog_instruction
*inst
, const char *fn
,
568 static const char *last_fn
;
572 _mesa_printf("%s:\n", fn
);
575 _mesa_printf("%d:\t", line
);
576 _mesa_print_instruction(inst
);
581 static void emit_op3fn(struct tnl_program
*p
,
591 GLuint nr
= p
->program
->Base
.NumInstructions
++;
592 struct prog_instruction
*inst
= &p
->program
->Base
.Instructions
[nr
];
594 if (p
->program
->Base
.NumInstructions
> MAX_INSN
) {
595 _mesa_problem(0, "Out of instructions in emit_op3fn\n");
599 inst
->Opcode
= (enum prog_opcode
) op
;
603 emit_arg( &inst
->SrcReg
[0], src0
);
604 emit_arg( &inst
->SrcReg
[1], src1
);
605 emit_arg( &inst
->SrcReg
[2], src2
);
607 emit_dst( &inst
->DstReg
, dest
, mask
);
609 debug_insn(inst
, fn
, line
);
613 #define emit_op3(p, op, dst, mask, src0, src1, src2) \
614 emit_op3fn(p, op, dst, mask, src0, src1, src2, __FUNCTION__, __LINE__)
616 #define emit_op2(p, op, dst, mask, src0, src1) \
617 emit_op3fn(p, op, dst, mask, src0, src1, undef, __FUNCTION__, __LINE__)
619 #define emit_op1(p, op, dst, mask, src0) \
620 emit_op3fn(p, op, dst, mask, src0, undef, undef, __FUNCTION__, __LINE__)
623 static struct ureg
make_temp( struct tnl_program
*p
, struct ureg reg
)
625 if (reg
.file
== PROGRAM_TEMPORARY
&&
626 !(p
->temp_reserved
& (1<<reg
.idx
)))
629 struct ureg temp
= get_temp(p
);
630 emit_op1(p
, OPCODE_MOV
, temp
, 0, reg
);
636 /* Currently no tracking performed of input/output/register size or
637 * active elements. Could be used to reduce these operations, as
638 * could the matrix type.
640 static void emit_matrix_transform_vec4( struct tnl_program
*p
,
642 const struct ureg
*mat
,
645 emit_op2(p
, OPCODE_DP4
, dest
, WRITEMASK_X
, src
, mat
[0]);
646 emit_op2(p
, OPCODE_DP4
, dest
, WRITEMASK_Y
, src
, mat
[1]);
647 emit_op2(p
, OPCODE_DP4
, dest
, WRITEMASK_Z
, src
, mat
[2]);
648 emit_op2(p
, OPCODE_DP4
, dest
, WRITEMASK_W
, src
, mat
[3]);
651 /* This version is much easier to implement if writemasks are not
652 * supported natively on the target or (like SSE), the target doesn't
653 * have a clean/obvious dotproduct implementation.
655 static void emit_transpose_matrix_transform_vec4( struct tnl_program
*p
,
657 const struct ureg
*mat
,
662 if (dest
.file
!= PROGRAM_TEMPORARY
)
667 emit_op2(p
, OPCODE_MUL
, tmp
, 0, swizzle1(src
,X
), mat
[0]);
668 emit_op3(p
, OPCODE_MAD
, tmp
, 0, swizzle1(src
,Y
), mat
[1], tmp
);
669 emit_op3(p
, OPCODE_MAD
, tmp
, 0, swizzle1(src
,Z
), mat
[2], tmp
);
670 emit_op3(p
, OPCODE_MAD
, dest
, 0, swizzle1(src
,W
), mat
[3], tmp
);
672 if (dest
.file
!= PROGRAM_TEMPORARY
)
673 release_temp(p
, tmp
);
676 static void emit_matrix_transform_vec3( struct tnl_program
*p
,
678 const struct ureg
*mat
,
681 emit_op2(p
, OPCODE_DP3
, dest
, WRITEMASK_X
, src
, mat
[0]);
682 emit_op2(p
, OPCODE_DP3
, dest
, WRITEMASK_Y
, src
, mat
[1]);
683 emit_op2(p
, OPCODE_DP3
, dest
, WRITEMASK_Z
, src
, mat
[2]);
687 static void emit_normalize_vec3( struct tnl_program
*p
,
691 struct ureg tmp
= get_temp(p
);
692 emit_op2(p
, OPCODE_DP3
, tmp
, WRITEMASK_X
, src
, src
);
693 emit_op1(p
, OPCODE_RSQ
, tmp
, WRITEMASK_X
, tmp
);
694 emit_op2(p
, OPCODE_MUL
, dest
, 0, src
, swizzle1(tmp
, X
));
695 release_temp(p
, tmp
);
698 static void emit_passthrough( struct tnl_program
*p
,
702 struct ureg out
= register_output(p
, output
);
703 emit_op1(p
, OPCODE_MOV
, out
, 0, register_input(p
, input
));
706 static struct ureg
get_eye_position( struct tnl_program
*p
)
708 if (is_undef(p
->eye_position
)) {
709 struct ureg pos
= register_input( p
, VERT_ATTRIB_POS
);
710 struct ureg modelview
[4];
712 p
->eye_position
= reserve_temp(p
);
715 register_matrix_param5( p
, STATE_MODELVIEW_MATRIX
, 0, 0, 3,
718 emit_matrix_transform_vec4(p
, p
->eye_position
, modelview
, pos
);
721 register_matrix_param5( p
, STATE_MODELVIEW_MATRIX
, 0, 0, 3,
722 STATE_MATRIX_TRANSPOSE
, modelview
);
724 emit_transpose_matrix_transform_vec4(p
, p
->eye_position
, modelview
, pos
);
728 return p
->eye_position
;
732 static struct ureg
get_eye_position_z( struct tnl_program
*p
)
734 if (!is_undef(p
->eye_position
))
735 return swizzle1(p
->eye_position
, Z
);
737 if (is_undef(p
->eye_position_z
)) {
738 struct ureg pos
= register_input( p
, VERT_ATTRIB_POS
);
739 struct ureg modelview
[4];
741 p
->eye_position_z
= reserve_temp(p
);
743 register_matrix_param5( p
, STATE_MODELVIEW_MATRIX
, 0, 0, 3,
746 emit_op2(p
, OPCODE_DP4
, p
->eye_position_z
, 0, pos
, modelview
[2]);
749 return p
->eye_position_z
;
754 static struct ureg
get_eye_position_normalized( struct tnl_program
*p
)
756 if (is_undef(p
->eye_position_normalized
)) {
757 struct ureg eye
= get_eye_position(p
);
758 p
->eye_position_normalized
= reserve_temp(p
);
759 emit_normalize_vec3(p
, p
->eye_position_normalized
, eye
);
762 return p
->eye_position_normalized
;
766 static struct ureg
get_transformed_normal( struct tnl_program
*p
)
768 if (is_undef(p
->transformed_normal
) &&
769 !p
->state
->need_eye_coords
&&
770 !p
->state
->normalize
&&
771 !(p
->state
->need_eye_coords
== p
->state
->rescale_normals
))
773 p
->transformed_normal
= register_input(p
, VERT_ATTRIB_NORMAL
);
775 else if (is_undef(p
->transformed_normal
))
777 struct ureg normal
= register_input(p
, VERT_ATTRIB_NORMAL
);
778 struct ureg mvinv
[3];
779 struct ureg transformed_normal
= reserve_temp(p
);
781 if (p
->state
->need_eye_coords
) {
782 register_matrix_param5( p
, STATE_MODELVIEW_MATRIX
, 0, 0, 2,
783 STATE_MATRIX_INVTRANS
, mvinv
);
785 /* Transform to eye space:
787 emit_matrix_transform_vec3( p
, transformed_normal
, mvinv
, normal
);
788 normal
= transformed_normal
;
791 /* Normalize/Rescale:
793 if (p
->state
->normalize
) {
794 emit_normalize_vec3( p
, transformed_normal
, normal
);
795 normal
= transformed_normal
;
797 else if (p
->state
->need_eye_coords
== p
->state
->rescale_normals
) {
798 /* This is already adjusted for eye/non-eye rendering:
800 struct ureg rescale
= register_param2(p
, STATE_INTERNAL
,
803 emit_op2( p
, OPCODE_MUL
, transformed_normal
, 0, normal
, rescale
);
804 normal
= transformed_normal
;
807 assert(normal
.file
== PROGRAM_TEMPORARY
);
808 p
->transformed_normal
= normal
;
811 return p
->transformed_normal
;
816 static void build_hpos( struct tnl_program
*p
)
818 struct ureg pos
= register_input( p
, VERT_ATTRIB_POS
);
819 struct ureg hpos
= register_output( p
, VERT_RESULT_HPOS
);
823 register_matrix_param5( p
, STATE_MVP_MATRIX
, 0, 0, 3,
825 emit_matrix_transform_vec4( p
, hpos
, mvp
, pos
);
828 register_matrix_param5( p
, STATE_MVP_MATRIX
, 0, 0, 3,
829 STATE_MATRIX_TRANSPOSE
, mvp
);
830 emit_transpose_matrix_transform_vec4( p
, hpos
, mvp
, pos
);
835 static GLuint
material_attrib( GLuint side
, GLuint property
)
837 return ((property
- STATE_AMBIENT
) * 2 +
841 /* Get a bitmask of which material values vary on a per-vertex basis.
843 static void set_material_flags( struct tnl_program
*p
)
845 p
->color_materials
= 0;
848 if (p
->state
->light_color_material
) {
850 p
->color_materials
= p
->state
->light_color_material_mask
;
853 p
->materials
|= p
->state
->light_material_mask
;
857 /* XXX temporary!!! */
858 #define _TNL_ATTRIB_MAT_FRONT_AMBIENT 32
860 static struct ureg
get_material( struct tnl_program
*p
, GLuint side
,
863 GLuint attrib
= material_attrib(side
, property
);
865 if (p
->color_materials
& (1<<attrib
))
866 return register_input(p
, VERT_ATTRIB_COLOR0
);
867 else if (p
->materials
& (1<<attrib
))
868 return register_input( p
, attrib
+ _TNL_ATTRIB_MAT_FRONT_AMBIENT
);
870 return register_param3( p
, STATE_MATERIAL
, side
, property
);
873 #define SCENE_COLOR_BITS(side) (( MAT_BIT_FRONT_EMISSION | \
874 MAT_BIT_FRONT_AMBIENT | \
875 MAT_BIT_FRONT_DIFFUSE) << (side))
877 /* Either return a precalculated constant value or emit code to
878 * calculate these values dynamically in the case where material calls
879 * are present between begin/end pairs.
881 * Probably want to shift this to the program compilation phase - if
882 * we always emitted the calculation here, a smart compiler could
883 * detect that it was constant (given a certain set of inputs), and
884 * lift it out of the main loop. That way the programs created here
885 * would be independent of the vertex_buffer details.
887 static struct ureg
get_scenecolor( struct tnl_program
*p
, GLuint side
)
889 if (p
->materials
& SCENE_COLOR_BITS(side
)) {
890 struct ureg lm_ambient
= register_param1(p
, STATE_LIGHTMODEL_AMBIENT
);
891 struct ureg material_emission
= get_material(p
, side
, STATE_EMISSION
);
892 struct ureg material_ambient
= get_material(p
, side
, STATE_AMBIENT
);
893 struct ureg material_diffuse
= get_material(p
, side
, STATE_DIFFUSE
);
894 struct ureg tmp
= make_temp(p
, material_diffuse
);
895 emit_op3(p
, OPCODE_MAD
, tmp
, WRITEMASK_XYZ
, lm_ambient
,
896 material_ambient
, material_emission
);
900 return register_param2( p
, STATE_LIGHTMODEL_SCENECOLOR
, side
);
904 static struct ureg
get_lightprod( struct tnl_program
*p
, GLuint light
,
905 GLuint side
, GLuint property
)
907 GLuint attrib
= material_attrib(side
, property
);
908 if (p
->materials
& (1<<attrib
)) {
909 struct ureg light_value
=
910 register_param3(p
, STATE_LIGHT
, light
, property
);
911 struct ureg material_value
= get_material(p
, side
, property
);
912 struct ureg tmp
= get_temp(p
);
913 emit_op2(p
, OPCODE_MUL
, tmp
, 0, light_value
, material_value
);
917 return register_param4(p
, STATE_LIGHTPROD
, light
, side
, property
);
920 static struct ureg
calculate_light_attenuation( struct tnl_program
*p
,
925 struct ureg attenuation
= register_param3(p
, STATE_LIGHT
, i
,
927 struct ureg att
= get_temp(p
);
929 /* Calculate spot attenuation:
931 if (!p
->state
->unit
[i
].light_spotcutoff_is_180
) {
932 struct ureg spot_dir_norm
= register_param3(p
, STATE_INTERNAL
,
933 STATE_LIGHT_SPOT_DIR_NORMALIZED
, i
);
934 struct ureg spot
= get_temp(p
);
935 struct ureg slt
= get_temp(p
);
937 emit_op2(p
, OPCODE_DP3
, spot
, 0, negate(VPpli
), spot_dir_norm
);
938 emit_op2(p
, OPCODE_SLT
, slt
, 0, swizzle1(spot_dir_norm
,W
), spot
);
939 emit_op2(p
, OPCODE_POW
, spot
, 0, spot
, swizzle1(attenuation
, W
));
940 emit_op2(p
, OPCODE_MUL
, att
, 0, slt
, spot
);
942 release_temp(p
, spot
);
943 release_temp(p
, slt
);
946 /* Calculate distance attenuation:
948 if (p
->state
->unit
[i
].light_attenuated
) {
951 emit_op1(p
, OPCODE_RCP
, dist
, WRITEMASK_YZ
, dist
);
953 emit_op2(p
, OPCODE_MUL
, dist
, WRITEMASK_XZ
, dist
, swizzle1(dist
,Y
));
955 emit_op2(p
, OPCODE_DP3
, dist
, 0, attenuation
, dist
);
957 if (!p
->state
->unit
[i
].light_spotcutoff_is_180
) {
959 emit_op1(p
, OPCODE_RCP
, dist
, 0, dist
);
960 /* spot-atten * dist-atten */
961 emit_op2(p
, OPCODE_MUL
, att
, 0, dist
, att
);
964 emit_op1(p
, OPCODE_RCP
, att
, 0, dist
);
972 static void emit_degenerate_lit( struct tnl_program
*p
,
976 struct ureg id
= get_identity_param(p
);
978 /* Note that result.x & result.w will not be examined. Note also that
979 * dots.xyzw == dots.xxxx.
982 /* result[1] = MAX2(in, 0)
984 emit_op2(p
, OPCODE_MAX
, lit
, 0, id
, dots
);
986 /* result[2] = (in > 0 ? 1 : 0)
988 emit_op2(p
, OPCODE_SLT
, lit
, WRITEMASK_Z
,
994 /* Need to add some addtional parameters to allow lighting in object
995 * space - STATE_SPOT_DIRECTION and STATE_HALF_VECTOR implicitly assume eye
998 static void build_lighting( struct tnl_program
*p
)
1000 const GLboolean twoside
= p
->state
->light_twoside
;
1001 const GLboolean separate
= p
->state
->separate_specular
;
1002 GLuint nr_lights
= 0, count
= 0;
1003 struct ureg normal
= get_transformed_normal(p
);
1004 struct ureg lit
= get_temp(p
);
1005 struct ureg dots
= get_temp(p
);
1006 struct ureg _col0
= undef
, _col1
= undef
;
1007 struct ureg _bfc0
= undef
, _bfc1
= undef
;
1010 for (i
= 0; i
< MAX_LIGHTS
; i
++)
1011 if (p
->state
->unit
[i
].light_enabled
)
1014 set_material_flags(p
);
1017 if (!p
->state
->material_shininess_is_zero
) {
1018 struct ureg shininess
= get_material(p
, 0, STATE_SHININESS
);
1019 emit_op1(p
, OPCODE_MOV
, dots
, WRITEMASK_W
, swizzle1(shininess
,X
));
1020 release_temp(p
, shininess
);
1023 _col0
= make_temp(p
, get_scenecolor(p
, 0));
1025 _col1
= make_temp(p
, get_identity_param(p
));
1032 if (!p
->state
->material_shininess_is_zero
) {
1033 struct ureg shininess
= get_material(p
, 1, STATE_SHININESS
);
1034 emit_op1(p
, OPCODE_MOV
, dots
, WRITEMASK_Z
,
1035 negate(swizzle1(shininess
,X
)));
1036 release_temp(p
, shininess
);
1039 _bfc0
= make_temp(p
, get_scenecolor(p
, 1));
1041 _bfc1
= make_temp(p
, get_identity_param(p
));
1046 /* If no lights, still need to emit the scenecolor.
1049 struct ureg res0
= register_output( p
, VERT_RESULT_COL0
);
1050 emit_op1(p
, OPCODE_MOV
, res0
, 0, _col0
);
1054 struct ureg res1
= register_output( p
, VERT_RESULT_COL1
);
1055 emit_op1(p
, OPCODE_MOV
, res1
, 0, _col1
);
1059 struct ureg res0
= register_output( p
, VERT_RESULT_BFC0
);
1060 emit_op1(p
, OPCODE_MOV
, res0
, 0, _bfc0
);
1063 if (twoside
&& separate
) {
1064 struct ureg res1
= register_output( p
, VERT_RESULT_BFC1
);
1065 emit_op1(p
, OPCODE_MOV
, res1
, 0, _bfc1
);
1068 if (nr_lights
== 0) {
1073 for (i
= 0; i
< MAX_LIGHTS
; i
++) {
1074 if (p
->state
->unit
[i
].light_enabled
) {
1075 struct ureg half
= undef
;
1076 struct ureg att
= undef
, VPpli
= undef
;
1080 if (p
->state
->unit
[i
].light_eyepos3_is_zero
) {
1081 /* Can used precomputed constants in this case.
1082 * Attenuation never applies to infinite lights.
1084 VPpli
= register_param3(p
, STATE_INTERNAL
,
1085 STATE_LIGHT_POSITION_NORMALIZED
, i
);
1087 if (!p
->state
->material_shininess_is_zero
) {
1088 if (p
->state
->light_local_viewer
) {
1089 struct ureg eye_hat
= get_eye_position_normalized(p
);
1091 emit_op2(p
, OPCODE_SUB
, half
, 0, VPpli
, eye_hat
);
1092 emit_normalize_vec3(p
, half
, half
);
1094 half
= register_param3(p
, STATE_INTERNAL
,
1095 STATE_LIGHT_HALF_VECTOR
, i
);
1100 struct ureg Ppli
= register_param3(p
, STATE_INTERNAL
,
1101 STATE_LIGHT_POSITION
, i
);
1102 struct ureg V
= get_eye_position(p
);
1103 struct ureg dist
= get_temp(p
);
1105 VPpli
= get_temp(p
);
1107 /* Calculate VPpli vector
1109 emit_op2(p
, OPCODE_SUB
, VPpli
, 0, Ppli
, V
);
1111 /* Normalize VPpli. The dist value also used in
1112 * attenuation below.
1114 emit_op2(p
, OPCODE_DP3
, dist
, 0, VPpli
, VPpli
);
1115 emit_op1(p
, OPCODE_RSQ
, dist
, 0, dist
);
1116 emit_op2(p
, OPCODE_MUL
, VPpli
, 0, VPpli
, dist
);
1118 /* Calculate attenuation:
1120 if (!p
->state
->unit
[i
].light_spotcutoff_is_180
||
1121 p
->state
->unit
[i
].light_attenuated
) {
1122 att
= calculate_light_attenuation(p
, i
, VPpli
, dist
);
1125 /* Calculate viewer direction, or use infinite viewer:
1127 if (!p
->state
->material_shininess_is_zero
) {
1130 if (p
->state
->light_local_viewer
) {
1131 struct ureg eye_hat
= get_eye_position_normalized(p
);
1132 emit_op2(p
, OPCODE_SUB
, half
, 0, VPpli
, eye_hat
);
1135 struct ureg z_dir
= swizzle(get_identity_param(p
),X
,Y
,W
,Z
);
1136 emit_op2(p
, OPCODE_ADD
, half
, 0, VPpli
, z_dir
);
1139 emit_normalize_vec3(p
, half
, half
);
1142 release_temp(p
, dist
);
1145 /* Calculate dot products:
1147 if (p
->state
->material_shininess_is_zero
) {
1148 emit_op2(p
, OPCODE_DP3
, dots
, 0, normal
, VPpli
);
1151 emit_op2(p
, OPCODE_DP3
, dots
, WRITEMASK_X
, normal
, VPpli
);
1152 emit_op2(p
, OPCODE_DP3
, dots
, WRITEMASK_Y
, normal
, half
);
1155 /* Front face lighting:
1158 struct ureg ambient
= get_lightprod(p
, i
, 0, STATE_AMBIENT
);
1159 struct ureg diffuse
= get_lightprod(p
, i
, 0, STATE_DIFFUSE
);
1160 struct ureg specular
= get_lightprod(p
, i
, 0, STATE_SPECULAR
);
1161 struct ureg res0
, res1
;
1162 GLuint mask0
, mask1
;
1165 if (count
== nr_lights
) {
1167 mask0
= WRITEMASK_XYZ
;
1168 mask1
= WRITEMASK_XYZ
;
1169 res0
= register_output( p
, VERT_RESULT_COL0
);
1170 res1
= register_output( p
, VERT_RESULT_COL1
);
1174 mask1
= WRITEMASK_XYZ
;
1176 res1
= register_output( p
, VERT_RESULT_COL0
);
1186 if (!is_undef(att
)) {
1187 emit_op1(p
, OPCODE_LIT
, lit
, 0, dots
);
1188 emit_op2(p
, OPCODE_MUL
, lit
, 0, lit
, att
);
1189 emit_op3(p
, OPCODE_MAD
, _col0
, 0, swizzle1(lit
,X
), ambient
, _col0
);
1191 else if (!p
->state
->material_shininess_is_zero
) {
1192 emit_op1(p
, OPCODE_LIT
, lit
, 0, dots
);
1193 emit_op2(p
, OPCODE_ADD
, _col0
, 0, ambient
, _col0
);
1196 emit_degenerate_lit(p
, lit
, dots
);
1197 emit_op2(p
, OPCODE_ADD
, _col0
, 0, ambient
, _col0
);
1200 emit_op3(p
, OPCODE_MAD
, res0
, mask0
, swizzle1(lit
,Y
), diffuse
, _col0
);
1201 emit_op3(p
, OPCODE_MAD
, res1
, mask1
, swizzle1(lit
,Z
), specular
, _col1
);
1203 release_temp(p
, ambient
);
1204 release_temp(p
, diffuse
);
1205 release_temp(p
, specular
);
1208 /* Back face lighting:
1211 struct ureg ambient
= get_lightprod(p
, i
, 1, STATE_AMBIENT
);
1212 struct ureg diffuse
= get_lightprod(p
, i
, 1, STATE_DIFFUSE
);
1213 struct ureg specular
= get_lightprod(p
, i
, 1, STATE_SPECULAR
);
1214 struct ureg res0
, res1
;
1215 GLuint mask0
, mask1
;
1217 if (count
== nr_lights
) {
1219 mask0
= WRITEMASK_XYZ
;
1220 mask1
= WRITEMASK_XYZ
;
1221 res0
= register_output( p
, VERT_RESULT_BFC0
);
1222 res1
= register_output( p
, VERT_RESULT_BFC1
);
1226 mask1
= WRITEMASK_XYZ
;
1228 res1
= register_output( p
, VERT_RESULT_BFC0
);
1237 dots
= negate(swizzle(dots
,X
,Y
,W
,Z
));
1239 if (!is_undef(att
)) {
1240 emit_op1(p
, OPCODE_LIT
, lit
, 0, dots
);
1241 emit_op2(p
, OPCODE_MUL
, lit
, 0, lit
, att
);
1242 emit_op3(p
, OPCODE_MAD
, _bfc0
, 0, swizzle1(lit
,X
), ambient
, _bfc0
);
1244 else if (!p
->state
->material_shininess_is_zero
) {
1245 emit_op1(p
, OPCODE_LIT
, lit
, 0, dots
);
1246 emit_op2(p
, OPCODE_ADD
, _col0
, 0, ambient
, _col0
);
1249 emit_degenerate_lit(p
, lit
, dots
);
1250 emit_op2(p
, OPCODE_ADD
, _col0
, 0, ambient
, _col0
);
1253 emit_op2(p
, OPCODE_ADD
, _bfc0
, 0, ambient
, _bfc0
);
1254 emit_op3(p
, OPCODE_MAD
, res0
, mask0
, swizzle1(lit
,Y
), diffuse
, _bfc0
);
1255 emit_op3(p
, OPCODE_MAD
, res1
, mask1
, swizzle1(lit
,Z
), specular
, _bfc1
);
1257 release_temp(p
, ambient
);
1258 release_temp(p
, diffuse
);
1259 release_temp(p
, specular
);
1262 release_temp(p
, half
);
1263 release_temp(p
, VPpli
);
1264 release_temp(p
, att
);
1272 static void build_fog( struct tnl_program
*p
)
1274 struct ureg fog
= register_output(p
, VERT_RESULT_FOGC
);
1277 if (p
->state
->fog_source_is_depth
) {
1278 input
= get_eye_position_z(p
);
1281 input
= swizzle1(register_input(p
, VERT_ATTRIB_FOG
), X
);
1284 if (p
->state
->fog_mode
&& p
->state
->tnl_do_vertex_fog
) {
1285 struct ureg params
= register_param2(p
, STATE_INTERNAL
,
1286 STATE_FOG_PARAMS_OPTIMIZED
);
1287 struct ureg tmp
= get_temp(p
);
1288 GLboolean useabs
= (p
->state
->fog_mode
!= FOG_EXP2
);
1291 emit_op1(p
, OPCODE_ABS
, tmp
, 0, input
);
1294 switch (p
->state
->fog_mode
) {
1296 struct ureg id
= get_identity_param(p
);
1297 emit_op3(p
, OPCODE_MAD
, tmp
, 0, useabs
? tmp
: input
,
1298 swizzle1(params
,X
), swizzle1(params
,Y
));
1299 emit_op2(p
, OPCODE_MAX
, tmp
, 0, tmp
, swizzle1(id
,X
)); /* saturate */
1300 emit_op2(p
, OPCODE_MIN
, fog
, WRITEMASK_X
, tmp
, swizzle1(id
,W
));
1304 emit_op2(p
, OPCODE_MUL
, tmp
, 0, useabs
? tmp
: input
,
1305 swizzle1(params
,Z
));
1306 emit_op1(p
, OPCODE_EX2
, fog
, WRITEMASK_X
, negate(tmp
));
1309 emit_op2(p
, OPCODE_MUL
, tmp
, 0, input
, swizzle1(params
,W
));
1310 emit_op2(p
, OPCODE_MUL
, tmp
, 0, tmp
, tmp
);
1311 emit_op1(p
, OPCODE_EX2
, fog
, WRITEMASK_X
, negate(tmp
));
1315 release_temp(p
, tmp
);
1318 /* results = incoming fog coords (compute fog per-fragment later)
1320 * KW: Is it really necessary to do anything in this case?
1321 * BP: Yes, we always need to compute the absolute value, unless
1322 * we want to push that down into the fragment program...
1324 GLboolean useabs
= GL_TRUE
;
1325 emit_op1(p
, useabs
? OPCODE_ABS
: OPCODE_MOV
, fog
, WRITEMASK_X
, input
);
1329 static void build_reflect_texgen( struct tnl_program
*p
,
1333 struct ureg normal
= get_transformed_normal(p
);
1334 struct ureg eye_hat
= get_eye_position_normalized(p
);
1335 struct ureg tmp
= get_temp(p
);
1338 emit_op2(p
, OPCODE_DP3
, tmp
, 0, normal
, eye_hat
);
1340 emit_op2(p
, OPCODE_ADD
, tmp
, 0, tmp
, tmp
);
1342 emit_op3(p
, OPCODE_MAD
, dest
, writemask
, negate(tmp
), normal
, eye_hat
);
1344 release_temp(p
, tmp
);
1347 static void build_sphere_texgen( struct tnl_program
*p
,
1351 struct ureg normal
= get_transformed_normal(p
);
1352 struct ureg eye_hat
= get_eye_position_normalized(p
);
1353 struct ureg tmp
= get_temp(p
);
1354 struct ureg half
= register_scalar_const(p
, .5);
1355 struct ureg r
= get_temp(p
);
1356 struct ureg inv_m
= get_temp(p
);
1357 struct ureg id
= get_identity_param(p
);
1359 /* Could share the above calculations, but it would be
1360 * a fairly odd state for someone to set (both sphere and
1361 * reflection active for different texture coordinate
1362 * components. Of course - if two texture units enable
1363 * reflect and/or sphere, things start to tilt in favour
1364 * of seperating this out:
1368 emit_op2(p
, OPCODE_DP3
, tmp
, 0, normal
, eye_hat
);
1370 emit_op2(p
, OPCODE_ADD
, tmp
, 0, tmp
, tmp
);
1372 emit_op3(p
, OPCODE_MAD
, r
, 0, negate(tmp
), normal
, eye_hat
);
1374 emit_op2(p
, OPCODE_ADD
, tmp
, 0, r
, swizzle(id
,X
,Y
,W
,Z
));
1375 /* rx^2 + ry^2 + (rz+1)^2 */
1376 emit_op2(p
, OPCODE_DP3
, tmp
, 0, tmp
, tmp
);
1378 emit_op1(p
, OPCODE_RSQ
, tmp
, 0, tmp
);
1380 emit_op2(p
, OPCODE_MUL
, inv_m
, 0, tmp
, half
);
1382 emit_op3(p
, OPCODE_MAD
, dest
, writemask
, r
, inv_m
, half
);
1384 release_temp(p
, tmp
);
1386 release_temp(p
, inv_m
);
1390 static void build_texture_transform( struct tnl_program
*p
)
1394 for (i
= 0; i
< MAX_TEXTURE_UNITS
; i
++) {
1396 if (!(p
->state
->fragprog_inputs_read
& FRAG_BIT_TEX(i
)))
1399 if (p
->state
->unit
[i
].texgen_enabled
||
1400 p
->state
->unit
[i
].texmat_enabled
) {
1402 GLuint texmat_enabled
= p
->state
->unit
[i
].texmat_enabled
;
1403 struct ureg out
= register_output(p
, VERT_RESULT_TEX0
+ i
);
1404 struct ureg out_texgen
= undef
;
1406 if (p
->state
->unit
[i
].texgen_enabled
) {
1407 GLuint copy_mask
= 0;
1408 GLuint sphere_mask
= 0;
1409 GLuint reflect_mask
= 0;
1410 GLuint normal_mask
= 0;
1414 out_texgen
= get_temp(p
);
1418 modes
[0] = p
->state
->unit
[i
].texgen_mode0
;
1419 modes
[1] = p
->state
->unit
[i
].texgen_mode1
;
1420 modes
[2] = p
->state
->unit
[i
].texgen_mode2
;
1421 modes
[3] = p
->state
->unit
[i
].texgen_mode3
;
1423 for (j
= 0; j
< 4; j
++) {
1425 case TXG_OBJ_LINEAR
: {
1426 struct ureg obj
= register_input(p
, VERT_ATTRIB_POS
);
1428 register_param3(p
, STATE_TEXGEN
, i
,
1429 STATE_TEXGEN_OBJECT_S
+ j
);
1431 emit_op2(p
, OPCODE_DP4
, out_texgen
, WRITEMASK_X
<< j
,
1435 case TXG_EYE_LINEAR
: {
1436 struct ureg eye
= get_eye_position(p
);
1438 register_param3(p
, STATE_TEXGEN
, i
,
1439 STATE_TEXGEN_EYE_S
+ j
);
1441 emit_op2(p
, OPCODE_DP4
, out_texgen
, WRITEMASK_X
<< j
,
1445 case TXG_SPHERE_MAP
:
1446 sphere_mask
|= WRITEMASK_X
<< j
;
1448 case TXG_REFLECTION_MAP
:
1449 reflect_mask
|= WRITEMASK_X
<< j
;
1451 case TXG_NORMAL_MAP
:
1452 normal_mask
|= WRITEMASK_X
<< j
;
1455 copy_mask
|= WRITEMASK_X
<< j
;
1462 build_sphere_texgen(p
, out_texgen
, sphere_mask
);
1466 build_reflect_texgen(p
, out_texgen
, reflect_mask
);
1470 struct ureg normal
= get_transformed_normal(p
);
1471 emit_op1(p
, OPCODE_MOV
, out_texgen
, normal_mask
, normal
);
1475 struct ureg in
= register_input(p
, VERT_ATTRIB_TEX0
+i
);
1476 emit_op1(p
, OPCODE_MOV
, out_texgen
, copy_mask
, in
);
1480 if (texmat_enabled
) {
1481 struct ureg texmat
[4];
1482 struct ureg in
= (!is_undef(out_texgen
) ?
1484 register_input(p
, VERT_ATTRIB_TEX0
+i
));
1486 register_matrix_param5( p
, STATE_TEXTURE_MATRIX
, i
, 0, 3,
1488 emit_matrix_transform_vec4( p
, out
, texmat
, in
);
1491 register_matrix_param5( p
, STATE_TEXTURE_MATRIX
, i
, 0, 3,
1492 STATE_MATRIX_TRANSPOSE
, texmat
);
1493 emit_transpose_matrix_transform_vec4( p
, out
, texmat
, in
);
1500 emit_passthrough(p
, VERT_ATTRIB_TEX0
+i
, VERT_RESULT_TEX0
+i
);
1506 static void build_pointsize( struct tnl_program
*p
)
1508 struct ureg eye
= get_eye_position_z(p
);
1509 struct ureg state_size
= register_param1(p
, STATE_POINT_SIZE
);
1510 struct ureg state_attenuation
= register_param1(p
, STATE_POINT_ATTENUATION
);
1511 struct ureg out
= register_output(p
, VERT_RESULT_PSIZ
);
1512 struct ureg ut
= get_temp(p
);
1515 emit_op1(p
, OPCODE_ABS
, ut
, WRITEMASK_Y
, swizzle1(eye
, Z
));
1516 /* p1 + dist * (p2 + dist * p3); */
1517 emit_op3(p
, OPCODE_MAD
, ut
, WRITEMASK_X
, swizzle1(ut
, Y
),
1518 swizzle1(state_attenuation
, Z
), swizzle1(state_attenuation
, Y
));
1519 emit_op3(p
, OPCODE_MAD
, ut
, WRITEMASK_X
, swizzle1(ut
, Y
),
1520 ut
, swizzle1(state_attenuation
, X
));
1522 /* 1 / sqrt(factor) */
1523 emit_op1(p
, OPCODE_RSQ
, ut
, WRITEMASK_X
, ut
);
1526 /* out = pointSize / sqrt(factor) */
1527 emit_op2(p
, OPCODE_MUL
, out
, WRITEMASK_X
, ut
, state_size
);
1529 /* this is a good place to clamp the point size since there's likely
1530 * no hardware registers to clamp point size at rasterization time.
1532 emit_op2(p
, OPCODE_MUL
, ut
, WRITEMASK_X
, ut
, state_size
);
1533 emit_op2(p
, OPCODE_MAX
, ut
, WRITEMASK_X
, ut
, swizzle1(state_size
, Y
));
1534 emit_op2(p
, OPCODE_MIN
, out
, WRITEMASK_X
, ut
, swizzle1(state_size
, Z
));
1537 release_temp(p
, ut
);
1541 * Emit constant point size.
1543 static void constant_pointsize( struct tnl_program
*p
)
1545 struct ureg state_size
= register_param1(p
, STATE_POINT_SIZE
);
1546 struct ureg out
= register_output(p
, VERT_RESULT_PSIZ
);
1547 emit_op1(p
, OPCODE_MOV
, out
, WRITEMASK_X
, state_size
);
1550 static void build_tnl_program( struct tnl_program
*p
)
1551 { /* Emit the program, starting with modelviewproject:
1555 /* Lighting calculations:
1557 if (p
->state
->fragprog_inputs_read
& (FRAG_BIT_COL0
|FRAG_BIT_COL1
)) {
1558 if (p
->state
->light_global_enabled
)
1561 if (p
->state
->fragprog_inputs_read
& FRAG_BIT_COL0
)
1562 emit_passthrough(p
, VERT_ATTRIB_COLOR0
, VERT_RESULT_COL0
);
1564 if (p
->state
->fragprog_inputs_read
& FRAG_BIT_COL1
)
1565 emit_passthrough(p
, VERT_ATTRIB_COLOR1
, VERT_RESULT_COL1
);
1569 if ((p
->state
->fragprog_inputs_read
& FRAG_BIT_FOGC
) ||
1570 p
->state
->fog_mode
!= FOG_NONE
)
1573 if (p
->state
->fragprog_inputs_read
& FRAG_BITS_TEX_ANY
)
1574 build_texture_transform(p
);
1576 if (p
->state
->point_attenuated
)
1580 constant_pointsize(p
);
1585 emit_op1(p
, OPCODE_END
, undef
, 0, undef
);
1590 _mesa_printf ("\n");
1596 create_new_program( const struct state_key
*key
,
1597 struct gl_vertex_program
*program
,
1600 struct tnl_program p
;
1602 _mesa_memset(&p
, 0, sizeof(p
));
1604 p
.program
= program
;
1605 p
.eye_position
= undef
;
1606 p
.eye_position_z
= undef
;
1607 p
.eye_position_normalized
= undef
;
1608 p
.transformed_normal
= undef
;
1612 if (max_temps
>= sizeof(int) * 8)
1613 p
.temp_reserved
= 0;
1615 p
.temp_reserved
= ~((1<<max_temps
)-1);
1617 p
.program
->Base
.Instructions
= _mesa_alloc_instructions(MAX_INSN
);
1618 p
.program
->Base
.String
= NULL
;
1619 p
.program
->Base
.NumInstructions
=
1620 p
.program
->Base
.NumTemporaries
=
1621 p
.program
->Base
.NumParameters
=
1622 p
.program
->Base
.NumAttributes
= p
.program
->Base
.NumAddressRegs
= 0;
1623 p
.program
->Base
.Parameters
= _mesa_new_parameter_list();
1624 p
.program
->Base
.InputsRead
= 0;
1625 p
.program
->Base
.OutputsWritten
= 0;
1627 build_tnl_program( &p
);
1632 * Return a vertex program which implements the current fixed-function
1633 * transform/lighting/texgen operations.
1634 * XXX move this into core mesa (main/)
1636 struct gl_vertex_program
*
1637 _mesa_get_fixed_func_vertex_program(GLcontext
*ctx
)
1639 struct gl_vertex_program
*prog
;
1640 struct state_key
*key
;
1642 /* Grab all the relevent state and put it in a single structure:
1644 key
= make_state_key(ctx
);
1646 /* Look for an already-prepared program for this state:
1648 prog
= (struct gl_vertex_program
*)
1649 _mesa_search_program_cache(ctx
->VertexProgram
.Cache
, key
, sizeof(*key
));
1652 /* OK, we'll have to build a new one */
1654 _mesa_printf("Build new TNL program\n");
1656 prog
= (struct gl_vertex_program
*)
1657 ctx
->Driver
.NewProgram(ctx
, GL_VERTEX_PROGRAM_ARB
, 0);
1661 create_new_program( key
, prog
,
1662 ctx
->Const
.VertexProgram
.MaxTemps
);
1665 if (ctx
->Driver
.ProgramStringNotify
)
1666 ctx
->Driver
.ProgramStringNotify( ctx
, GL_VERTEX_PROGRAM_ARB
,
1669 _mesa_program_cache_insert(ctx
, ctx
->VertexProgram
.Cache
,
1670 key
, sizeof(*key
), &prog
->Base
);