1 /**************************************************************************
3 * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
6 * Permission is hereby granted, free of charge, to any person obtaining a
7 * copy of this software and associated documentation files (the
8 * "Software"), to deal in the Software without restriction, including
9 * without limitation the rights to use, copy, modify, merge, publish,
10 * distribute, sub license, and/or sell copies of the Software, and to
11 * permit persons to whom the Software is furnished to do so, subject to
12 * the following conditions:
14 * The above copyright notice and this permission notice (including the
15 * next paragraph) shall be included in all copies or substantial portions
18 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
19 * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
20 * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
21 * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
22 * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
23 * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
24 * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
26 **************************************************************************/
29 * \file ffvertex_prog.c
31 * Create a vertex program to execute the current fixed function T&L pipeline.
32 * \author Keith Whitwell
36 #include "main/glheader.h"
37 #include "main/mtypes.h"
38 #include "main/macros.h"
39 #include "main/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_color_material_mask
:12;
51 unsigned light_material_mask
:12;
52 unsigned light_global_enabled
:1;
53 unsigned light_local_viewer
:1;
54 unsigned light_twoside
:1;
55 unsigned light_color_material
:1;
56 unsigned material_shininess_is_zero
:1;
57 unsigned need_eye_coords
:1;
59 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 point_array
:1;
67 unsigned texture_enabled_global
:1;
68 unsigned fragprog_inputs_read
:12;
70 unsigned 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 texgen_enabled
:4;
80 unsigned texgen_mode0
:4;
81 unsigned texgen_mode1
:4;
82 unsigned texgen_mode2
:4;
83 unsigned texgen_mode3
:4;
94 static GLuint
translate_fog_mode( GLenum mode
)
97 case GL_LINEAR
: return FOG_LINEAR
;
98 case GL_EXP
: return FOG_EXP
;
99 case GL_EXP2
: return FOG_EXP2
;
100 default: return FOG_NONE
;
106 #define TXG_OBJ_LINEAR 1
107 #define TXG_EYE_LINEAR 2
108 #define TXG_SPHERE_MAP 3
109 #define TXG_REFLECTION_MAP 4
110 #define TXG_NORMAL_MAP 5
112 static GLuint
translate_texgen( GLboolean enabled
, GLenum mode
)
118 case GL_OBJECT_LINEAR
: return TXG_OBJ_LINEAR
;
119 case GL_EYE_LINEAR
: return TXG_EYE_LINEAR
;
120 case GL_SPHERE_MAP
: return TXG_SPHERE_MAP
;
121 case GL_REFLECTION_MAP_NV
: return TXG_REFLECTION_MAP
;
122 case GL_NORMAL_MAP_NV
: return TXG_NORMAL_MAP
;
123 default: return TXG_NONE
;
129 * Returns bitmask of flags indicating which materials are set per-vertex
131 * XXX get these from the VBO...
134 tnl_get_per_vertex_materials(GLcontext
*ctx
)
136 GLbitfield mask
= 0x0;
138 TNLcontext
*tnl
= TNL_CONTEXT(ctx
);
139 struct vertex_buffer
*VB
= &tnl
->vb
;
142 for (i
= _TNL_FIRST_MAT
; i
<= _TNL_LAST_MAT
; i
++)
143 if (VB
->AttribPtr
[i
] && VB
->AttribPtr
[i
]->stride
)
144 mask
|= 1 << (i
- _TNL_FIRST_MAT
);
151 * Should fog be computed per-vertex?
154 tnl_get_per_vertex_fog(GLcontext
*ctx
)
157 TNLcontext
*tnl
= TNL_CONTEXT(ctx
);
158 return tnl
->_DoVertexFog
;
165 static GLboolean
check_active_shininess( GLcontext
*ctx
,
166 const struct state_key
*key
,
169 GLuint bit
= 1 << (MAT_ATTRIB_FRONT_SHININESS
+ side
);
171 if (key
->light_color_material_mask
& bit
)
174 if (key
->light_material_mask
& bit
)
177 if (ctx
->Light
.Material
.Attrib
[MAT_ATTRIB_FRONT_SHININESS
+ side
][0] != 0.0F
)
184 static void make_state_key( GLcontext
*ctx
, struct state_key
*key
)
186 const struct gl_fragment_program
*fp
;
189 memset(key
, 0, sizeof(struct state_key
));
190 fp
= ctx
->FragmentProgram
._Current
;
192 /* This now relies on texenvprogram.c being active:
196 key
->need_eye_coords
= ctx
->_NeedEyeCoords
;
198 key
->fragprog_inputs_read
= fp
->Base
.InputsRead
;
199 key
->varying_vp_inputs
= ctx
->varying_vp_inputs
;
201 if (ctx
->RenderMode
== GL_FEEDBACK
) {
202 /* make sure the vertprog emits color and tex0 */
203 key
->fragprog_inputs_read
|= (FRAG_BIT_COL0
| FRAG_BIT_TEX0
);
206 key
->separate_specular
= (ctx
->Light
.Model
.ColorControl
==
207 GL_SEPARATE_SPECULAR_COLOR
);
209 if (ctx
->Light
.Enabled
) {
210 key
->light_global_enabled
= 1;
212 if (ctx
->Light
.Model
.LocalViewer
)
213 key
->light_local_viewer
= 1;
215 if (ctx
->Light
.Model
.TwoSide
)
216 key
->light_twoside
= 1;
218 if (ctx
->Light
.ColorMaterialEnabled
) {
219 key
->light_color_material
= 1;
220 key
->light_color_material_mask
= ctx
->Light
.ColorMaterialBitmask
;
223 key
->light_material_mask
= tnl_get_per_vertex_materials(ctx
);
225 for (i
= 0; i
< MAX_LIGHTS
; i
++) {
226 struct gl_light
*light
= &ctx
->Light
.Light
[i
];
228 if (light
->Enabled
) {
229 key
->unit
[i
].light_enabled
= 1;
231 if (light
->EyePosition
[3] == 0.0)
232 key
->unit
[i
].light_eyepos3_is_zero
= 1;
234 if (light
->SpotCutoff
== 180.0)
235 key
->unit
[i
].light_spotcutoff_is_180
= 1;
237 if (light
->ConstantAttenuation
!= 1.0 ||
238 light
->LinearAttenuation
!= 0.0 ||
239 light
->QuadraticAttenuation
!= 0.0)
240 key
->unit
[i
].light_attenuated
= 1;
244 if (check_active_shininess(ctx
, key
, 0)) {
245 key
->material_shininess_is_zero
= 0;
247 else if (key
->light_twoside
&&
248 check_active_shininess(ctx
, key
, 1)) {
249 key
->material_shininess_is_zero
= 0;
252 key
->material_shininess_is_zero
= 1;
256 if (ctx
->Transform
.Normalize
)
259 if (ctx
->Transform
.RescaleNormals
)
260 key
->rescale_normals
= 1;
262 key
->fog_mode
= translate_fog_mode(fp
->FogOption
);
264 if (ctx
->Fog
.FogCoordinateSource
== GL_FRAGMENT_DEPTH_EXT
)
265 key
->fog_source_is_depth
= 1;
267 key
->tnl_do_vertex_fog
= tnl_get_per_vertex_fog(ctx
);
269 if (ctx
->Point
._Attenuated
)
270 key
->point_attenuated
= 1;
272 #if FEATURE_point_size_array
273 if (ctx
->Array
.ArrayObj
->PointSize
.Enabled
)
274 key
->point_array
= 1;
277 if (ctx
->Texture
._TexGenEnabled
||
278 ctx
->Texture
._TexMatEnabled
||
279 ctx
->Texture
._EnabledUnits
)
280 key
->texture_enabled_global
= 1;
282 for (i
= 0; i
< MAX_TEXTURE_COORD_UNITS
; i
++) {
283 struct gl_texture_unit
*texUnit
= &ctx
->Texture
.Unit
[i
];
285 if (texUnit
->_ReallyEnabled
)
286 key
->unit
[i
].texunit_really_enabled
= 1;
288 if (ctx
->Texture
._TexMatEnabled
& ENABLE_TEXMAT(i
))
289 key
->unit
[i
].texmat_enabled
= 1;
291 if (texUnit
->TexGenEnabled
) {
292 key
->unit
[i
].texgen_enabled
= 1;
294 key
->unit
[i
].texgen_mode0
=
295 translate_texgen( texUnit
->TexGenEnabled
& (1<<0),
296 texUnit
->GenS
.Mode
);
297 key
->unit
[i
].texgen_mode1
=
298 translate_texgen( texUnit
->TexGenEnabled
& (1<<1),
299 texUnit
->GenT
.Mode
);
300 key
->unit
[i
].texgen_mode2
=
301 translate_texgen( texUnit
->TexGenEnabled
& (1<<2),
302 texUnit
->GenR
.Mode
);
303 key
->unit
[i
].texgen_mode3
=
304 translate_texgen( texUnit
->TexGenEnabled
& (1<<3),
305 texUnit
->GenQ
.Mode
);
312 /* Very useful debugging tool - produces annotated listing of
313 * generated program with line/function references for each
314 * instruction back into this file:
318 /* Should be tunable by the driver - do we want to do matrix
319 * multiplications with DP4's or with MUL/MAD's? SSE works better
320 * with the latter, drivers may differ.
325 /* Use uregs to represent registers internally, translate to Mesa's
326 * expected formats on emit.
328 * NOTE: These are passed by value extensively in this file rather
329 * than as usual by pointer reference. If this disturbs you, try
330 * remembering they are just 32bits in size.
332 * GCC is smart enough to deal with these dword-sized structures in
333 * much the same way as if I had defined them as dwords and was using
334 * macros to access and set the fields. This is much nicer and easier
339 GLint idx
:9; /* relative addressing may be negative */
340 /* sizeof(idx) should == sizeof(prog_src_reg::Index) */
348 const struct state_key
*state
;
349 struct gl_vertex_program
*program
;
350 GLint max_inst
; /** number of instructions allocated for program */
353 GLuint temp_reserved
;
355 struct ureg eye_position
;
356 struct ureg eye_position_z
;
357 struct ureg eye_position_normalized
;
358 struct ureg transformed_normal
;
359 struct ureg identity
;
362 GLuint color_materials
;
366 static const struct ureg undef
= {
384 static struct ureg
make_ureg(GLuint file
, GLint idx
)
390 reg
.swz
= SWIZZLE_NOOP
;
397 static struct ureg
negate( struct ureg reg
)
404 static struct ureg
swizzle( struct ureg reg
, int x
, int y
, int z
, int w
)
406 reg
.swz
= MAKE_SWIZZLE4(GET_SWZ(reg
.swz
, x
),
409 GET_SWZ(reg
.swz
, w
));
414 static struct ureg
swizzle1( struct ureg reg
, int x
)
416 return swizzle(reg
, x
, x
, x
, x
);
420 static struct ureg
get_temp( struct tnl_program
*p
)
422 int bit
= _mesa_ffs( ~p
->temp_in_use
);
424 _mesa_problem(NULL
, "%s: out of temporaries\n", __FILE__
);
428 if ((GLuint
) bit
> p
->program
->Base
.NumTemporaries
)
429 p
->program
->Base
.NumTemporaries
= bit
;
431 p
->temp_in_use
|= 1<<(bit
-1);
432 return make_ureg(PROGRAM_TEMPORARY
, bit
-1);
436 static struct ureg
reserve_temp( struct tnl_program
*p
)
438 struct ureg temp
= get_temp( p
);
439 p
->temp_reserved
|= 1<<temp
.idx
;
444 static void release_temp( struct tnl_program
*p
, struct ureg reg
)
446 if (reg
.file
== PROGRAM_TEMPORARY
) {
447 p
->temp_in_use
&= ~(1<<reg
.idx
);
448 p
->temp_in_use
|= p
->temp_reserved
; /* can't release reserved temps */
452 static void release_temps( struct tnl_program
*p
)
454 p
->temp_in_use
= p
->temp_reserved
;
458 static struct ureg
register_param5(struct tnl_program
*p
,
465 gl_state_index tokens
[STATE_LENGTH
];
472 idx
= _mesa_add_state_reference( p
->program
->Base
.Parameters
, tokens
);
473 return make_ureg(PROGRAM_STATE_VAR
, idx
);
477 #define register_param1(p,s0) register_param5(p,s0,0,0,0,0)
478 #define register_param2(p,s0,s1) register_param5(p,s0,s1,0,0,0)
479 #define register_param3(p,s0,s1,s2) register_param5(p,s0,s1,s2,0,0)
480 #define register_param4(p,s0,s1,s2,s3) register_param5(p,s0,s1,s2,s3,0)
485 * \param input one of VERT_ATTRIB_x tokens.
487 static struct ureg
register_input( struct tnl_program
*p
, GLuint input
)
489 /* Material attribs are passed here as inputs >= 32
491 if (input
>= 32 || (p
->state
->varying_vp_inputs
& (1<<input
))) {
492 p
->program
->Base
.InputsRead
|= (1<<input
);
493 return make_ureg(PROGRAM_INPUT
, input
);
496 return register_param3( p
, STATE_INTERNAL
, STATE_CURRENT_ATTRIB
, input
);
502 * \param input one of VERT_RESULT_x tokens.
504 static struct ureg
register_output( struct tnl_program
*p
, GLuint output
)
506 p
->program
->Base
.OutputsWritten
|= (1<<output
);
507 return make_ureg(PROGRAM_OUTPUT
, output
);
511 static struct ureg
register_const4f( struct tnl_program
*p
,
524 idx
= _mesa_add_unnamed_constant( p
->program
->Base
.Parameters
, values
, 4,
526 ASSERT(swizzle
== SWIZZLE_NOOP
);
527 return make_ureg(PROGRAM_CONSTANT
, idx
);
530 #define register_const1f(p, s0) register_const4f(p, s0, 0, 0, 1)
531 #define register_scalar_const(p, s0) register_const4f(p, s0, s0, s0, s0)
532 #define register_const2f(p, s0, s1) register_const4f(p, s0, s1, 0, 1)
533 #define register_const3f(p, s0, s1, s2) register_const4f(p, s0, s1, s2, 1)
535 static GLboolean
is_undef( struct ureg reg
)
537 return reg
.file
== PROGRAM_UNDEFINED
;
541 static struct ureg
get_identity_param( struct tnl_program
*p
)
543 if (is_undef(p
->identity
))
544 p
->identity
= register_const4f(p
, 0,0,0,1);
549 static void register_matrix_param5( struct tnl_program
*p
,
550 GLint s0
, /* modelview, projection, etc */
551 GLint s1
, /* texture matrix number */
552 GLint s2
, /* first row */
553 GLint s3
, /* last row */
554 GLint s4
, /* inverse, transpose, etc */
555 struct ureg
*matrix
)
559 /* This is a bit sad as the support is there to pull the whole
560 * matrix out in one go:
562 for (i
= 0; i
<= s3
- s2
; i
++)
563 matrix
[i
] = register_param5( p
, s0
, s1
, i
, i
, s4
);
567 static void emit_arg( struct prog_src_register
*src
,
570 src
->File
= reg
.file
;
571 src
->Index
= reg
.idx
;
572 src
->Swizzle
= reg
.swz
;
573 src
->NegateBase
= reg
.negate
? NEGATE_XYZW
: 0;
577 /* Check that bitfield sizes aren't exceeded */
578 ASSERT(src
->Index
== reg
.idx
);
582 static void emit_dst( struct prog_dst_register
*dst
,
583 struct ureg reg
, GLuint mask
)
585 dst
->File
= reg
.file
;
586 dst
->Index
= reg
.idx
;
587 /* allow zero as a shorthand for xyzw */
588 dst
->WriteMask
= mask
? mask
: WRITEMASK_XYZW
;
589 dst
->CondMask
= COND_TR
; /* always pass cond test */
590 dst
->CondSwizzle
= SWIZZLE_NOOP
;
593 /* Check that bitfield sizes aren't exceeded */
594 ASSERT(dst
->Index
== reg
.idx
);
598 static void debug_insn( struct prog_instruction
*inst
, const char *fn
,
602 static const char *last_fn
;
606 _mesa_printf("%s:\n", fn
);
609 _mesa_printf("%d:\t", line
);
610 _mesa_print_instruction(inst
);
615 static void emit_op3fn(struct tnl_program
*p
,
626 struct prog_instruction
*inst
;
628 assert((GLint
) p
->program
->Base
.NumInstructions
<= p
->max_inst
);
630 if (p
->program
->Base
.NumInstructions
== p
->max_inst
) {
631 /* need to extend the program's instruction array */
632 struct prog_instruction
*newInst
;
634 /* double the size */
637 newInst
= _mesa_alloc_instructions(p
->max_inst
);
639 _mesa_error(NULL
, GL_OUT_OF_MEMORY
, "vertex program build");
643 _mesa_copy_instructions(newInst
,
644 p
->program
->Base
.Instructions
,
645 p
->program
->Base
.NumInstructions
);
647 _mesa_free_instructions(p
->program
->Base
.Instructions
,
648 p
->program
->Base
.NumInstructions
);
650 p
->program
->Base
.Instructions
= newInst
;
653 nr
= p
->program
->Base
.NumInstructions
++;
655 inst
= &p
->program
->Base
.Instructions
[nr
];
656 inst
->Opcode
= (enum prog_opcode
) op
;
659 emit_arg( &inst
->SrcReg
[0], src0
);
660 emit_arg( &inst
->SrcReg
[1], src1
);
661 emit_arg( &inst
->SrcReg
[2], src2
);
663 emit_dst( &inst
->DstReg
, dest
, mask
);
665 debug_insn(inst
, fn
, line
);
669 #define emit_op3(p, op, dst, mask, src0, src1, src2) \
670 emit_op3fn(p, op, dst, mask, src0, src1, src2, __FUNCTION__, __LINE__)
672 #define emit_op2(p, op, dst, mask, src0, src1) \
673 emit_op3fn(p, op, dst, mask, src0, src1, undef, __FUNCTION__, __LINE__)
675 #define emit_op1(p, op, dst, mask, src0) \
676 emit_op3fn(p, op, dst, mask, src0, undef, undef, __FUNCTION__, __LINE__)
679 static struct ureg
make_temp( struct tnl_program
*p
, struct ureg reg
)
681 if (reg
.file
== PROGRAM_TEMPORARY
&&
682 !(p
->temp_reserved
& (1<<reg
.idx
)))
685 struct ureg temp
= get_temp(p
);
686 emit_op1(p
, OPCODE_MOV
, temp
, 0, reg
);
692 /* Currently no tracking performed of input/output/register size or
693 * active elements. Could be used to reduce these operations, as
694 * could the matrix type.
696 static void emit_matrix_transform_vec4( struct tnl_program
*p
,
698 const struct ureg
*mat
,
701 emit_op2(p
, OPCODE_DP4
, dest
, WRITEMASK_X
, src
, mat
[0]);
702 emit_op2(p
, OPCODE_DP4
, dest
, WRITEMASK_Y
, src
, mat
[1]);
703 emit_op2(p
, OPCODE_DP4
, dest
, WRITEMASK_Z
, src
, mat
[2]);
704 emit_op2(p
, OPCODE_DP4
, dest
, WRITEMASK_W
, src
, mat
[3]);
708 /* This version is much easier to implement if writemasks are not
709 * supported natively on the target or (like SSE), the target doesn't
710 * have a clean/obvious dotproduct implementation.
712 static void emit_transpose_matrix_transform_vec4( struct tnl_program
*p
,
714 const struct ureg
*mat
,
719 if (dest
.file
!= PROGRAM_TEMPORARY
)
724 emit_op2(p
, OPCODE_MUL
, tmp
, 0, swizzle1(src
,X
), mat
[0]);
725 emit_op3(p
, OPCODE_MAD
, tmp
, 0, swizzle1(src
,Y
), mat
[1], tmp
);
726 emit_op3(p
, OPCODE_MAD
, tmp
, 0, swizzle1(src
,Z
), mat
[2], tmp
);
727 emit_op3(p
, OPCODE_MAD
, dest
, 0, swizzle1(src
,W
), mat
[3], tmp
);
729 if (dest
.file
!= PROGRAM_TEMPORARY
)
730 release_temp(p
, tmp
);
734 static void emit_matrix_transform_vec3( struct tnl_program
*p
,
736 const struct ureg
*mat
,
739 emit_op2(p
, OPCODE_DP3
, dest
, WRITEMASK_X
, src
, mat
[0]);
740 emit_op2(p
, OPCODE_DP3
, dest
, WRITEMASK_Y
, src
, mat
[1]);
741 emit_op2(p
, OPCODE_DP3
, dest
, WRITEMASK_Z
, src
, mat
[2]);
745 static void emit_normalize_vec3( struct tnl_program
*p
,
750 /* XXX use this when drivers are ready for NRM3 */
751 emit_op1(p
, OPCODE_NRM3
, dest
, WRITEMASK_XYZ
, src
);
753 struct ureg tmp
= get_temp(p
);
754 emit_op2(p
, OPCODE_DP3
, tmp
, WRITEMASK_X
, src
, src
);
755 emit_op1(p
, OPCODE_RSQ
, tmp
, WRITEMASK_X
, tmp
);
756 emit_op2(p
, OPCODE_MUL
, dest
, 0, src
, swizzle1(tmp
, X
));
757 release_temp(p
, tmp
);
762 static void emit_passthrough( struct tnl_program
*p
,
766 struct ureg out
= register_output(p
, output
);
767 emit_op1(p
, OPCODE_MOV
, out
, 0, register_input(p
, input
));
771 static struct ureg
get_eye_position( struct tnl_program
*p
)
773 if (is_undef(p
->eye_position
)) {
774 struct ureg pos
= register_input( p
, VERT_ATTRIB_POS
);
775 struct ureg modelview
[4];
777 p
->eye_position
= reserve_temp(p
);
780 register_matrix_param5( p
, STATE_MODELVIEW_MATRIX
, 0, 0, 3,
783 emit_matrix_transform_vec4(p
, p
->eye_position
, modelview
, pos
);
786 register_matrix_param5( p
, STATE_MODELVIEW_MATRIX
, 0, 0, 3,
787 STATE_MATRIX_TRANSPOSE
, modelview
);
789 emit_transpose_matrix_transform_vec4(p
, p
->eye_position
, modelview
, pos
);
793 return p
->eye_position
;
797 static struct ureg
get_eye_position_z( struct tnl_program
*p
)
799 if (!is_undef(p
->eye_position
))
800 return swizzle1(p
->eye_position
, Z
);
802 if (is_undef(p
->eye_position_z
)) {
803 struct ureg pos
= register_input( p
, VERT_ATTRIB_POS
);
804 struct ureg modelview
[4];
806 p
->eye_position_z
= reserve_temp(p
);
808 register_matrix_param5( p
, STATE_MODELVIEW_MATRIX
, 0, 0, 3,
811 emit_op2(p
, OPCODE_DP4
, p
->eye_position_z
, 0, pos
, modelview
[2]);
814 return p
->eye_position_z
;
818 static struct ureg
get_eye_position_normalized( struct tnl_program
*p
)
820 if (is_undef(p
->eye_position_normalized
)) {
821 struct ureg eye
= get_eye_position(p
);
822 p
->eye_position_normalized
= reserve_temp(p
);
823 emit_normalize_vec3(p
, p
->eye_position_normalized
, eye
);
826 return p
->eye_position_normalized
;
830 static struct ureg
get_transformed_normal( struct tnl_program
*p
)
832 if (is_undef(p
->transformed_normal
) &&
833 !p
->state
->need_eye_coords
&&
834 !p
->state
->normalize
&&
835 !(p
->state
->need_eye_coords
== p
->state
->rescale_normals
))
837 p
->transformed_normal
= register_input(p
, VERT_ATTRIB_NORMAL
);
839 else if (is_undef(p
->transformed_normal
))
841 struct ureg normal
= register_input(p
, VERT_ATTRIB_NORMAL
);
842 struct ureg mvinv
[3];
843 struct ureg transformed_normal
= reserve_temp(p
);
845 if (p
->state
->need_eye_coords
) {
846 register_matrix_param5( p
, STATE_MODELVIEW_MATRIX
, 0, 0, 2,
847 STATE_MATRIX_INVTRANS
, mvinv
);
849 /* Transform to eye space:
851 emit_matrix_transform_vec3( p
, transformed_normal
, mvinv
, normal
);
852 normal
= transformed_normal
;
855 /* Normalize/Rescale:
857 if (p
->state
->normalize
) {
858 emit_normalize_vec3( p
, transformed_normal
, normal
);
859 normal
= transformed_normal
;
861 else if (p
->state
->need_eye_coords
== p
->state
->rescale_normals
) {
862 /* This is already adjusted for eye/non-eye rendering:
864 struct ureg rescale
= register_param2(p
, STATE_INTERNAL
,
867 emit_op2( p
, OPCODE_MUL
, transformed_normal
, 0, normal
, rescale
);
868 normal
= transformed_normal
;
871 assert(normal
.file
== PROGRAM_TEMPORARY
);
872 p
->transformed_normal
= normal
;
875 return p
->transformed_normal
;
879 static void build_hpos( struct tnl_program
*p
)
881 struct ureg pos
= register_input( p
, VERT_ATTRIB_POS
);
882 struct ureg hpos
= register_output( p
, VERT_RESULT_HPOS
);
886 register_matrix_param5( p
, STATE_MVP_MATRIX
, 0, 0, 3,
888 emit_matrix_transform_vec4( p
, hpos
, mvp
, pos
);
891 register_matrix_param5( p
, STATE_MVP_MATRIX
, 0, 0, 3,
892 STATE_MATRIX_TRANSPOSE
, mvp
);
893 emit_transpose_matrix_transform_vec4( p
, hpos
, mvp
, pos
);
898 static GLuint
material_attrib( GLuint side
, GLuint property
)
900 return (property
- STATE_AMBIENT
) * 2 + side
;
905 * Get a bitmask of which material values vary on a per-vertex basis.
907 static void set_material_flags( struct tnl_program
*p
)
909 p
->color_materials
= 0;
912 if (p
->state
->light_color_material
) {
914 p
->color_materials
= p
->state
->light_color_material_mask
;
917 p
->materials
|= p
->state
->light_material_mask
;
921 /* XXX temporary!!! */
922 #define _TNL_ATTRIB_MAT_FRONT_AMBIENT 32
924 static struct ureg
get_material( struct tnl_program
*p
, GLuint side
,
927 GLuint attrib
= material_attrib(side
, property
);
929 if (p
->color_materials
& (1<<attrib
))
930 return register_input(p
, VERT_ATTRIB_COLOR0
);
931 else if (p
->materials
& (1<<attrib
))
932 return register_input( p
, attrib
+ _TNL_ATTRIB_MAT_FRONT_AMBIENT
);
934 return register_param3( p
, STATE_MATERIAL
, side
, property
);
937 #define SCENE_COLOR_BITS(side) (( MAT_BIT_FRONT_EMISSION | \
938 MAT_BIT_FRONT_AMBIENT | \
939 MAT_BIT_FRONT_DIFFUSE) << (side))
943 * Either return a precalculated constant value or emit code to
944 * calculate these values dynamically in the case where material calls
945 * are present between begin/end pairs.
947 * Probably want to shift this to the program compilation phase - if
948 * we always emitted the calculation here, a smart compiler could
949 * detect that it was constant (given a certain set of inputs), and
950 * lift it out of the main loop. That way the programs created here
951 * would be independent of the vertex_buffer details.
953 static struct ureg
get_scenecolor( struct tnl_program
*p
, GLuint side
)
955 if (p
->materials
& SCENE_COLOR_BITS(side
)) {
956 struct ureg lm_ambient
= register_param1(p
, STATE_LIGHTMODEL_AMBIENT
);
957 struct ureg material_emission
= get_material(p
, side
, STATE_EMISSION
);
958 struct ureg material_ambient
= get_material(p
, side
, STATE_AMBIENT
);
959 struct ureg material_diffuse
= get_material(p
, side
, STATE_DIFFUSE
);
960 struct ureg tmp
= make_temp(p
, material_diffuse
);
961 emit_op3(p
, OPCODE_MAD
, tmp
, WRITEMASK_XYZ
, lm_ambient
,
962 material_ambient
, material_emission
);
966 return register_param2( p
, STATE_LIGHTMODEL_SCENECOLOR
, side
);
970 static struct ureg
get_lightprod( struct tnl_program
*p
, GLuint light
,
971 GLuint side
, GLuint property
)
973 GLuint attrib
= material_attrib(side
, property
);
974 if (p
->materials
& (1<<attrib
)) {
975 struct ureg light_value
=
976 register_param3(p
, STATE_LIGHT
, light
, property
);
977 struct ureg material_value
= get_material(p
, side
, property
);
978 struct ureg tmp
= get_temp(p
);
979 emit_op2(p
, OPCODE_MUL
, tmp
, 0, light_value
, material_value
);
983 return register_param4(p
, STATE_LIGHTPROD
, light
, side
, property
);
987 static struct ureg
calculate_light_attenuation( struct tnl_program
*p
,
992 struct ureg attenuation
= register_param3(p
, STATE_LIGHT
, i
,
994 struct ureg att
= get_temp(p
);
996 /* Calculate spot attenuation:
998 if (!p
->state
->unit
[i
].light_spotcutoff_is_180
) {
999 struct ureg spot_dir_norm
= register_param3(p
, STATE_INTERNAL
,
1000 STATE_LIGHT_SPOT_DIR_NORMALIZED
, i
);
1001 struct ureg spot
= get_temp(p
);
1002 struct ureg slt
= get_temp(p
);
1004 emit_op2(p
, OPCODE_DP3
, spot
, 0, negate(VPpli
), spot_dir_norm
);
1005 emit_op2(p
, OPCODE_SLT
, slt
, 0, swizzle1(spot_dir_norm
,W
), spot
);
1006 emit_op2(p
, OPCODE_POW
, spot
, 0, spot
, swizzle1(attenuation
, W
));
1007 emit_op2(p
, OPCODE_MUL
, att
, 0, slt
, spot
);
1009 release_temp(p
, spot
);
1010 release_temp(p
, slt
);
1013 /* Calculate distance attenuation:
1015 if (p
->state
->unit
[i
].light_attenuated
) {
1017 emit_op1(p
, OPCODE_RCP
, dist
, WRITEMASK_YZ
, dist
);
1019 emit_op2(p
, OPCODE_MUL
, dist
, WRITEMASK_XZ
, dist
, swizzle1(dist
,Y
));
1021 emit_op2(p
, OPCODE_DP3
, dist
, 0, attenuation
, dist
);
1023 if (!p
->state
->unit
[i
].light_spotcutoff_is_180
) {
1025 emit_op1(p
, OPCODE_RCP
, dist
, 0, dist
);
1026 /* spot-atten * dist-atten */
1027 emit_op2(p
, OPCODE_MUL
, att
, 0, dist
, att
);
1031 emit_op1(p
, OPCODE_RCP
, att
, 0, dist
);
1041 * lit.y = MAX(0, dots.x)
1042 * lit.z = SLT(0, dots.x)
1044 static void emit_degenerate_lit( struct tnl_program
*p
,
1048 struct ureg id
= get_identity_param(p
); /* id = {0,0,0,1} */
1050 /* Note that lit.x & lit.w will not be examined. Note also that
1051 * dots.xyzw == dots.xxxx.
1054 /* MAX lit, id, dots;
1056 emit_op2(p
, OPCODE_MAX
, lit
, WRITEMASK_XYZW
, id
, dots
);
1058 /* result[2] = (in > 0 ? 1 : 0)
1059 * SLT lit.z, id.z, dots; # lit.z = (0 < dots.z) ? 1 : 0
1061 emit_op2(p
, OPCODE_SLT
, lit
, WRITEMASK_Z
, swizzle1(id
,Z
), dots
);
1065 /* Need to add some addtional parameters to allow lighting in object
1066 * space - STATE_SPOT_DIRECTION and STATE_HALF_VECTOR implicitly assume eye
1069 static void build_lighting( struct tnl_program
*p
)
1071 const GLboolean twoside
= p
->state
->light_twoside
;
1072 const GLboolean separate
= p
->state
->separate_specular
;
1073 GLuint nr_lights
= 0, count
= 0;
1074 struct ureg normal
= get_transformed_normal(p
);
1075 struct ureg lit
= get_temp(p
);
1076 struct ureg dots
= get_temp(p
);
1077 struct ureg _col0
= undef
, _col1
= undef
;
1078 struct ureg _bfc0
= undef
, _bfc1
= undef
;
1083 * dots.x = dot(normal, VPpli)
1084 * dots.y = dot(normal, halfAngle)
1085 * dots.z = back.shininess
1086 * dots.w = front.shininess
1089 for (i
= 0; i
< MAX_LIGHTS
; i
++)
1090 if (p
->state
->unit
[i
].light_enabled
)
1093 set_material_flags(p
);
1096 if (!p
->state
->material_shininess_is_zero
) {
1097 struct ureg shininess
= get_material(p
, 0, STATE_SHININESS
);
1098 emit_op1(p
, OPCODE_MOV
, dots
, WRITEMASK_W
, swizzle1(shininess
,X
));
1099 release_temp(p
, shininess
);
1102 _col0
= make_temp(p
, get_scenecolor(p
, 0));
1104 _col1
= make_temp(p
, get_identity_param(p
));
1110 if (!p
->state
->material_shininess_is_zero
) {
1111 /* Note that we negate the back-face specular exponent here.
1112 * The negation will be un-done later in the back-face code below.
1114 struct ureg shininess
= get_material(p
, 1, STATE_SHININESS
);
1115 emit_op1(p
, OPCODE_MOV
, dots
, WRITEMASK_Z
,
1116 negate(swizzle1(shininess
,X
)));
1117 release_temp(p
, shininess
);
1120 _bfc0
= make_temp(p
, get_scenecolor(p
, 1));
1122 _bfc1
= make_temp(p
, get_identity_param(p
));
1127 /* If no lights, still need to emit the scenecolor.
1130 struct ureg res0
= register_output( p
, VERT_RESULT_COL0
);
1131 emit_op1(p
, OPCODE_MOV
, res0
, 0, _col0
);
1135 struct ureg res1
= register_output( p
, VERT_RESULT_COL1
);
1136 emit_op1(p
, OPCODE_MOV
, res1
, 0, _col1
);
1140 struct ureg res0
= register_output( p
, VERT_RESULT_BFC0
);
1141 emit_op1(p
, OPCODE_MOV
, res0
, 0, _bfc0
);
1144 if (twoside
&& separate
) {
1145 struct ureg res1
= register_output( p
, VERT_RESULT_BFC1
);
1146 emit_op1(p
, OPCODE_MOV
, res1
, 0, _bfc1
);
1149 if (nr_lights
== 0) {
1154 for (i
= 0; i
< MAX_LIGHTS
; i
++) {
1155 if (p
->state
->unit
[i
].light_enabled
) {
1156 struct ureg half
= undef
;
1157 struct ureg att
= undef
, VPpli
= undef
;
1161 if (p
->state
->unit
[i
].light_eyepos3_is_zero
) {
1162 /* Can used precomputed constants in this case.
1163 * Attenuation never applies to infinite lights.
1165 VPpli
= register_param3(p
, STATE_INTERNAL
,
1166 STATE_LIGHT_POSITION_NORMALIZED
, i
);
1168 if (!p
->state
->material_shininess_is_zero
) {
1169 if (p
->state
->light_local_viewer
) {
1170 struct ureg eye_hat
= get_eye_position_normalized(p
);
1172 emit_op2(p
, OPCODE_SUB
, half
, 0, VPpli
, eye_hat
);
1173 emit_normalize_vec3(p
, half
, half
);
1176 half
= register_param3(p
, STATE_INTERNAL
,
1177 STATE_LIGHT_HALF_VECTOR
, i
);
1182 struct ureg Ppli
= register_param3(p
, STATE_INTERNAL
,
1183 STATE_LIGHT_POSITION
, i
);
1184 struct ureg V
= get_eye_position(p
);
1185 struct ureg dist
= get_temp(p
);
1187 VPpli
= get_temp(p
);
1189 /* Calculate VPpli vector
1191 emit_op2(p
, OPCODE_SUB
, VPpli
, 0, Ppli
, V
);
1193 /* Normalize VPpli. The dist value also used in
1194 * attenuation below.
1196 emit_op2(p
, OPCODE_DP3
, dist
, 0, VPpli
, VPpli
);
1197 emit_op1(p
, OPCODE_RSQ
, dist
, 0, dist
);
1198 emit_op2(p
, OPCODE_MUL
, VPpli
, 0, VPpli
, dist
);
1200 /* Calculate attenuation:
1202 if (!p
->state
->unit
[i
].light_spotcutoff_is_180
||
1203 p
->state
->unit
[i
].light_attenuated
) {
1204 att
= calculate_light_attenuation(p
, i
, VPpli
, dist
);
1207 /* Calculate viewer direction, or use infinite viewer:
1209 if (!p
->state
->material_shininess_is_zero
) {
1212 if (p
->state
->light_local_viewer
) {
1213 struct ureg eye_hat
= get_eye_position_normalized(p
);
1214 emit_op2(p
, OPCODE_SUB
, half
, 0, VPpli
, eye_hat
);
1217 struct ureg z_dir
= swizzle(get_identity_param(p
),X
,Y
,W
,Z
);
1218 emit_op2(p
, OPCODE_ADD
, half
, 0, VPpli
, z_dir
);
1221 emit_normalize_vec3(p
, half
, half
);
1224 release_temp(p
, dist
);
1227 /* Calculate dot products:
1229 if (p
->state
->material_shininess_is_zero
) {
1230 emit_op2(p
, OPCODE_DP3
, dots
, 0, normal
, VPpli
);
1233 emit_op2(p
, OPCODE_DP3
, dots
, WRITEMASK_X
, normal
, VPpli
);
1234 emit_op2(p
, OPCODE_DP3
, dots
, WRITEMASK_Y
, normal
, half
);
1237 /* Front face lighting:
1240 struct ureg ambient
= get_lightprod(p
, i
, 0, STATE_AMBIENT
);
1241 struct ureg diffuse
= get_lightprod(p
, i
, 0, STATE_DIFFUSE
);
1242 struct ureg specular
= get_lightprod(p
, i
, 0, STATE_SPECULAR
);
1243 struct ureg res0
, res1
;
1244 GLuint mask0
, mask1
;
1246 if (count
== nr_lights
) {
1248 mask0
= WRITEMASK_XYZ
;
1249 mask1
= WRITEMASK_XYZ
;
1250 res0
= register_output( p
, VERT_RESULT_COL0
);
1251 res1
= register_output( p
, VERT_RESULT_COL1
);
1255 mask1
= WRITEMASK_XYZ
;
1257 res1
= register_output( p
, VERT_RESULT_COL0
);
1267 if (!is_undef(att
)) {
1268 /* light is attenuated by distance */
1269 emit_op1(p
, OPCODE_LIT
, lit
, 0, dots
);
1270 emit_op2(p
, OPCODE_MUL
, lit
, 0, lit
, att
);
1271 emit_op3(p
, OPCODE_MAD
, _col0
, 0, swizzle1(lit
,X
), ambient
, _col0
);
1273 else if (!p
->state
->material_shininess_is_zero
) {
1274 /* there's a non-zero specular term */
1275 emit_op1(p
, OPCODE_LIT
, lit
, 0, dots
);
1276 emit_op2(p
, OPCODE_ADD
, _col0
, 0, ambient
, _col0
);
1279 /* no attenutation, no specular */
1280 emit_degenerate_lit(p
, lit
, dots
);
1281 emit_op2(p
, OPCODE_ADD
, _col0
, 0, ambient
, _col0
);
1284 emit_op3(p
, OPCODE_MAD
, res0
, mask0
, swizzle1(lit
,Y
), diffuse
, _col0
);
1285 emit_op3(p
, OPCODE_MAD
, res1
, mask1
, swizzle1(lit
,Z
), specular
, _col1
);
1287 release_temp(p
, ambient
);
1288 release_temp(p
, diffuse
);
1289 release_temp(p
, specular
);
1292 /* Back face lighting:
1295 struct ureg ambient
= get_lightprod(p
, i
, 1, STATE_AMBIENT
);
1296 struct ureg diffuse
= get_lightprod(p
, i
, 1, STATE_DIFFUSE
);
1297 struct ureg specular
= get_lightprod(p
, i
, 1, STATE_SPECULAR
);
1298 struct ureg res0
, res1
;
1299 GLuint mask0
, mask1
;
1301 if (count
== nr_lights
) {
1303 mask0
= WRITEMASK_XYZ
;
1304 mask1
= WRITEMASK_XYZ
;
1305 res0
= register_output( p
, VERT_RESULT_BFC0
);
1306 res1
= register_output( p
, VERT_RESULT_BFC1
);
1310 mask1
= WRITEMASK_XYZ
;
1312 res1
= register_output( p
, VERT_RESULT_BFC0
);
1322 /* For the back face we need to negate the X and Y component
1323 * dot products. dots.Z has the negated back-face specular
1324 * exponent. We swizzle that into the W position. This
1325 * negation makes the back-face specular term positive again.
1327 dots
= negate(swizzle(dots
,X
,Y
,W
,Z
));
1329 if (!is_undef(att
)) {
1330 emit_op1(p
, OPCODE_LIT
, lit
, 0, dots
);
1331 emit_op2(p
, OPCODE_MUL
, lit
, 0, lit
, att
);
1332 emit_op3(p
, OPCODE_MAD
, _bfc0
, 0, swizzle1(lit
,X
), ambient
, _bfc0
);
1334 else if (!p
->state
->material_shininess_is_zero
) {
1335 emit_op1(p
, OPCODE_LIT
, lit
, 0, dots
);
1336 emit_op2(p
, OPCODE_ADD
, _bfc0
, 0, ambient
, _bfc0
); /**/
1339 emit_degenerate_lit(p
, lit
, dots
);
1340 emit_op2(p
, OPCODE_ADD
, _bfc0
, 0, ambient
, _bfc0
);
1343 emit_op3(p
, OPCODE_MAD
, res0
, mask0
, swizzle1(lit
,Y
), diffuse
, _bfc0
);
1344 emit_op3(p
, OPCODE_MAD
, res1
, mask1
, swizzle1(lit
,Z
), specular
, _bfc1
);
1345 /* restore dots to its original state for subsequent lights
1346 * by negating and swizzling again.
1348 dots
= negate(swizzle(dots
,X
,Y
,W
,Z
));
1350 release_temp(p
, ambient
);
1351 release_temp(p
, diffuse
);
1352 release_temp(p
, specular
);
1355 release_temp(p
, half
);
1356 release_temp(p
, VPpli
);
1357 release_temp(p
, att
);
1365 static void build_fog( struct tnl_program
*p
)
1367 struct ureg fog
= register_output(p
, VERT_RESULT_FOGC
);
1370 if (p
->state
->fog_source_is_depth
) {
1371 input
= get_eye_position_z(p
);
1374 input
= swizzle1(register_input(p
, VERT_ATTRIB_FOG
), X
);
1377 if (p
->state
->fog_mode
&& p
->state
->tnl_do_vertex_fog
) {
1378 struct ureg params
= register_param2(p
, STATE_INTERNAL
,
1379 STATE_FOG_PARAMS_OPTIMIZED
);
1380 struct ureg tmp
= get_temp(p
);
1381 GLboolean useabs
= (p
->state
->fog_mode
!= FOG_EXP2
);
1384 emit_op1(p
, OPCODE_ABS
, tmp
, 0, input
);
1387 switch (p
->state
->fog_mode
) {
1389 struct ureg id
= get_identity_param(p
);
1390 emit_op3(p
, OPCODE_MAD
, tmp
, 0, useabs
? tmp
: input
,
1391 swizzle1(params
,X
), swizzle1(params
,Y
));
1392 emit_op2(p
, OPCODE_MAX
, tmp
, 0, tmp
, swizzle1(id
,X
)); /* saturate */
1393 emit_op2(p
, OPCODE_MIN
, fog
, WRITEMASK_X
, tmp
, swizzle1(id
,W
));
1397 emit_op2(p
, OPCODE_MUL
, tmp
, 0, useabs
? tmp
: input
,
1398 swizzle1(params
,Z
));
1399 emit_op1(p
, OPCODE_EX2
, fog
, WRITEMASK_X
, negate(tmp
));
1402 emit_op2(p
, OPCODE_MUL
, tmp
, 0, input
, swizzle1(params
,W
));
1403 emit_op2(p
, OPCODE_MUL
, tmp
, 0, tmp
, tmp
);
1404 emit_op1(p
, OPCODE_EX2
, fog
, WRITEMASK_X
, negate(tmp
));
1408 release_temp(p
, tmp
);
1411 /* results = incoming fog coords (compute fog per-fragment later)
1413 * KW: Is it really necessary to do anything in this case?
1414 * BP: Yes, we always need to compute the absolute value, unless
1415 * we want to push that down into the fragment program...
1417 GLboolean useabs
= GL_TRUE
;
1418 emit_op1(p
, useabs
? OPCODE_ABS
: OPCODE_MOV
, fog
, WRITEMASK_X
, input
);
1423 static void build_reflect_texgen( struct tnl_program
*p
,
1427 struct ureg normal
= get_transformed_normal(p
);
1428 struct ureg eye_hat
= get_eye_position_normalized(p
);
1429 struct ureg tmp
= get_temp(p
);
1432 emit_op2(p
, OPCODE_DP3
, tmp
, 0, normal
, eye_hat
);
1434 emit_op2(p
, OPCODE_ADD
, tmp
, 0, tmp
, tmp
);
1436 emit_op3(p
, OPCODE_MAD
, dest
, writemask
, negate(tmp
), normal
, eye_hat
);
1438 release_temp(p
, tmp
);
1442 static void build_sphere_texgen( struct tnl_program
*p
,
1446 struct ureg normal
= get_transformed_normal(p
);
1447 struct ureg eye_hat
= get_eye_position_normalized(p
);
1448 struct ureg tmp
= get_temp(p
);
1449 struct ureg half
= register_scalar_const(p
, .5);
1450 struct ureg r
= get_temp(p
);
1451 struct ureg inv_m
= get_temp(p
);
1452 struct ureg id
= get_identity_param(p
);
1454 /* Could share the above calculations, but it would be
1455 * a fairly odd state for someone to set (both sphere and
1456 * reflection active for different texture coordinate
1457 * components. Of course - if two texture units enable
1458 * reflect and/or sphere, things start to tilt in favour
1459 * of seperating this out:
1463 emit_op2(p
, OPCODE_DP3
, tmp
, 0, normal
, eye_hat
);
1465 emit_op2(p
, OPCODE_ADD
, tmp
, 0, tmp
, tmp
);
1467 emit_op3(p
, OPCODE_MAD
, r
, 0, negate(tmp
), normal
, eye_hat
);
1469 emit_op2(p
, OPCODE_ADD
, tmp
, 0, r
, swizzle(id
,X
,Y
,W
,Z
));
1470 /* rx^2 + ry^2 + (rz+1)^2 */
1471 emit_op2(p
, OPCODE_DP3
, tmp
, 0, tmp
, tmp
);
1473 emit_op1(p
, OPCODE_RSQ
, tmp
, 0, tmp
);
1475 emit_op2(p
, OPCODE_MUL
, inv_m
, 0, tmp
, half
);
1477 emit_op3(p
, OPCODE_MAD
, dest
, writemask
, r
, inv_m
, half
);
1479 release_temp(p
, tmp
);
1481 release_temp(p
, inv_m
);
1485 static void build_texture_transform( struct tnl_program
*p
)
1489 for (i
= 0; i
< MAX_TEXTURE_COORD_UNITS
; i
++) {
1491 if (!(p
->state
->fragprog_inputs_read
& FRAG_BIT_TEX(i
)))
1494 if (p
->state
->unit
[i
].texgen_enabled
||
1495 p
->state
->unit
[i
].texmat_enabled
) {
1497 GLuint texmat_enabled
= p
->state
->unit
[i
].texmat_enabled
;
1498 struct ureg out
= register_output(p
, VERT_RESULT_TEX0
+ i
);
1499 struct ureg out_texgen
= undef
;
1501 if (p
->state
->unit
[i
].texgen_enabled
) {
1502 GLuint copy_mask
= 0;
1503 GLuint sphere_mask
= 0;
1504 GLuint reflect_mask
= 0;
1505 GLuint normal_mask
= 0;
1509 out_texgen
= get_temp(p
);
1513 modes
[0] = p
->state
->unit
[i
].texgen_mode0
;
1514 modes
[1] = p
->state
->unit
[i
].texgen_mode1
;
1515 modes
[2] = p
->state
->unit
[i
].texgen_mode2
;
1516 modes
[3] = p
->state
->unit
[i
].texgen_mode3
;
1518 for (j
= 0; j
< 4; j
++) {
1520 case TXG_OBJ_LINEAR
: {
1521 struct ureg obj
= register_input(p
, VERT_ATTRIB_POS
);
1523 register_param3(p
, STATE_TEXGEN
, i
,
1524 STATE_TEXGEN_OBJECT_S
+ j
);
1526 emit_op2(p
, OPCODE_DP4
, out_texgen
, WRITEMASK_X
<< j
,
1530 case TXG_EYE_LINEAR
: {
1531 struct ureg eye
= get_eye_position(p
);
1533 register_param3(p
, STATE_TEXGEN
, i
,
1534 STATE_TEXGEN_EYE_S
+ j
);
1536 emit_op2(p
, OPCODE_DP4
, out_texgen
, WRITEMASK_X
<< j
,
1540 case TXG_SPHERE_MAP
:
1541 sphere_mask
|= WRITEMASK_X
<< j
;
1543 case TXG_REFLECTION_MAP
:
1544 reflect_mask
|= WRITEMASK_X
<< j
;
1546 case TXG_NORMAL_MAP
:
1547 normal_mask
|= WRITEMASK_X
<< j
;
1550 copy_mask
|= WRITEMASK_X
<< j
;
1555 build_sphere_texgen(p
, out_texgen
, sphere_mask
);
1559 build_reflect_texgen(p
, out_texgen
, reflect_mask
);
1563 struct ureg normal
= get_transformed_normal(p
);
1564 emit_op1(p
, OPCODE_MOV
, out_texgen
, normal_mask
, normal
);
1568 struct ureg in
= register_input(p
, VERT_ATTRIB_TEX0
+i
);
1569 emit_op1(p
, OPCODE_MOV
, out_texgen
, copy_mask
, in
);
1573 if (texmat_enabled
) {
1574 struct ureg texmat
[4];
1575 struct ureg in
= (!is_undef(out_texgen
) ?
1577 register_input(p
, VERT_ATTRIB_TEX0
+i
));
1579 register_matrix_param5( p
, STATE_TEXTURE_MATRIX
, i
, 0, 3,
1581 emit_matrix_transform_vec4( p
, out
, texmat
, in
);
1584 register_matrix_param5( p
, STATE_TEXTURE_MATRIX
, i
, 0, 3,
1585 STATE_MATRIX_TRANSPOSE
, texmat
);
1586 emit_transpose_matrix_transform_vec4( p
, out
, texmat
, in
);
1593 emit_passthrough(p
, VERT_ATTRIB_TEX0
+i
, VERT_RESULT_TEX0
+i
);
1600 * Point size attenuation computation.
1602 static void build_atten_pointsize( struct tnl_program
*p
)
1604 struct ureg eye
= get_eye_position_z(p
);
1605 struct ureg state_size
= register_param1(p
, STATE_POINT_SIZE
);
1606 struct ureg state_attenuation
= register_param1(p
, STATE_POINT_ATTENUATION
);
1607 struct ureg out
= register_output(p
, VERT_RESULT_PSIZ
);
1608 struct ureg ut
= get_temp(p
);
1611 emit_op1(p
, OPCODE_ABS
, ut
, WRITEMASK_Y
, swizzle1(eye
, Z
));
1612 /* p1 + dist * (p2 + dist * p3); */
1613 emit_op3(p
, OPCODE_MAD
, ut
, WRITEMASK_X
, swizzle1(ut
, Y
),
1614 swizzle1(state_attenuation
, Z
), swizzle1(state_attenuation
, Y
));
1615 emit_op3(p
, OPCODE_MAD
, ut
, WRITEMASK_X
, swizzle1(ut
, Y
),
1616 ut
, swizzle1(state_attenuation
, X
));
1618 /* 1 / sqrt(factor) */
1619 emit_op1(p
, OPCODE_RSQ
, ut
, WRITEMASK_X
, ut
);
1622 /* out = pointSize / sqrt(factor) */
1623 emit_op2(p
, OPCODE_MUL
, out
, WRITEMASK_X
, ut
, state_size
);
1625 /* this is a good place to clamp the point size since there's likely
1626 * no hardware registers to clamp point size at rasterization time.
1628 emit_op2(p
, OPCODE_MUL
, ut
, WRITEMASK_X
, ut
, state_size
);
1629 emit_op2(p
, OPCODE_MAX
, ut
, WRITEMASK_X
, ut
, swizzle1(state_size
, Y
));
1630 emit_op2(p
, OPCODE_MIN
, out
, WRITEMASK_X
, ut
, swizzle1(state_size
, Z
));
1633 release_temp(p
, ut
);
1638 * Emit constant point size.
1640 static void build_constant_pointsize( struct tnl_program
*p
)
1642 struct ureg state_size
= register_param1(p
, STATE_POINT_SIZE
);
1643 struct ureg out
= register_output(p
, VERT_RESULT_PSIZ
);
1644 emit_op1(p
, OPCODE_MOV
, out
, WRITEMASK_X
, state_size
);
1649 * Pass-though per-vertex point size, from user's point size array.
1651 static void build_array_pointsize( struct tnl_program
*p
)
1653 struct ureg in
= register_input(p
, VERT_ATTRIB_POINT_SIZE
);
1654 struct ureg out
= register_output(p
, VERT_RESULT_PSIZ
);
1655 emit_op1(p
, OPCODE_MOV
, out
, WRITEMASK_X
, in
);
1659 static void build_tnl_program( struct tnl_program
*p
)
1661 /* Emit the program, starting with modelviewproject:
1665 /* Lighting calculations:
1667 if (p
->state
->fragprog_inputs_read
& (FRAG_BIT_COL0
|FRAG_BIT_COL1
)) {
1668 if (p
->state
->light_global_enabled
)
1671 if (p
->state
->fragprog_inputs_read
& FRAG_BIT_COL0
)
1672 emit_passthrough(p
, VERT_ATTRIB_COLOR0
, VERT_RESULT_COL0
);
1674 if (p
->state
->fragprog_inputs_read
& FRAG_BIT_COL1
)
1675 emit_passthrough(p
, VERT_ATTRIB_COLOR1
, VERT_RESULT_COL1
);
1679 if ((p
->state
->fragprog_inputs_read
& FRAG_BIT_FOGC
) ||
1680 p
->state
->fog_mode
!= FOG_NONE
)
1683 if (p
->state
->fragprog_inputs_read
& FRAG_BITS_TEX_ANY
)
1684 build_texture_transform(p
);
1686 if (p
->state
->point_attenuated
)
1687 build_atten_pointsize(p
);
1688 else if (p
->state
->point_array
)
1689 build_array_pointsize(p
);
1692 build_constant_pointsize(p
);
1694 (void) build_constant_pointsize
;
1699 emit_op1(p
, OPCODE_END
, undef
, 0, undef
);
1704 _mesa_printf ("\n");
1710 create_new_program( const struct state_key
*key
,
1711 struct gl_vertex_program
*program
,
1714 struct tnl_program p
;
1716 _mesa_memset(&p
, 0, sizeof(p
));
1718 p
.program
= program
;
1719 p
.eye_position
= undef
;
1720 p
.eye_position_z
= undef
;
1721 p
.eye_position_normalized
= undef
;
1722 p
.transformed_normal
= undef
;
1726 if (max_temps
>= sizeof(int) * 8)
1727 p
.temp_reserved
= 0;
1729 p
.temp_reserved
= ~((1<<max_temps
)-1);
1731 /* Start by allocating 32 instructions.
1732 * If we need more, we'll grow the instruction array as needed.
1735 p
.program
->Base
.Instructions
= _mesa_alloc_instructions(p
.max_inst
);
1736 p
.program
->Base
.String
= NULL
;
1737 p
.program
->Base
.NumInstructions
=
1738 p
.program
->Base
.NumTemporaries
=
1739 p
.program
->Base
.NumParameters
=
1740 p
.program
->Base
.NumAttributes
= p
.program
->Base
.NumAddressRegs
= 0;
1741 p
.program
->Base
.Parameters
= _mesa_new_parameter_list();
1742 p
.program
->Base
.InputsRead
= 0;
1743 p
.program
->Base
.OutputsWritten
= 0;
1745 build_tnl_program( &p
);
1750 * Return a vertex program which implements the current fixed-function
1751 * transform/lighting/texgen operations.
1752 * XXX move this into core mesa (main/)
1754 struct gl_vertex_program
*
1755 _mesa_get_fixed_func_vertex_program(GLcontext
*ctx
)
1757 struct gl_vertex_program
*prog
;
1758 struct state_key key
;
1760 /* Grab all the relevent state and put it in a single structure:
1762 make_state_key(ctx
, &key
);
1764 /* Look for an already-prepared program for this state:
1766 prog
= (struct gl_vertex_program
*)
1767 _mesa_search_program_cache(ctx
->VertexProgram
.Cache
, &key
, sizeof(key
));
1770 /* OK, we'll have to build a new one */
1772 _mesa_printf("Build new TNL program\n");
1774 prog
= (struct gl_vertex_program
*)
1775 ctx
->Driver
.NewProgram(ctx
, GL_VERTEX_PROGRAM_ARB
, 0);
1779 create_new_program( &key
, prog
,
1780 ctx
->Const
.VertexProgram
.MaxTemps
);
1783 if (ctx
->Driver
.ProgramStringNotify
)
1784 ctx
->Driver
.ProgramStringNotify( ctx
, GL_VERTEX_PROGRAM_ARB
,
1787 _mesa_program_cache_insert(ctx
, ctx
->VertexProgram
.Cache
,
1788 &key
, sizeof(key
), &prog
->Base
);