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
->Negate
= reg
.negate
? NEGATE_XYZW
: NEGATE_NONE
;
576 /* Check that bitfield sizes aren't exceeded */
577 ASSERT(src
->Index
== reg
.idx
);
581 static void emit_dst( struct prog_dst_register
*dst
,
582 struct ureg reg
, GLuint mask
)
584 dst
->File
= reg
.file
;
585 dst
->Index
= reg
.idx
;
586 /* allow zero as a shorthand for xyzw */
587 dst
->WriteMask
= mask
? mask
: WRITEMASK_XYZW
;
588 dst
->CondMask
= COND_TR
; /* always pass cond test */
589 dst
->CondSwizzle
= SWIZZLE_NOOP
;
592 /* Check that bitfield sizes aren't exceeded */
593 ASSERT(dst
->Index
== reg
.idx
);
597 static void debug_insn( struct prog_instruction
*inst
, const char *fn
,
601 static const char *last_fn
;
605 _mesa_printf("%s:\n", fn
);
608 _mesa_printf("%d:\t", line
);
609 _mesa_print_instruction(inst
);
614 static void emit_op3fn(struct tnl_program
*p
,
625 struct prog_instruction
*inst
;
627 assert((GLint
) p
->program
->Base
.NumInstructions
<= p
->max_inst
);
629 if (p
->program
->Base
.NumInstructions
== p
->max_inst
) {
630 /* need to extend the program's instruction array */
631 struct prog_instruction
*newInst
;
633 /* double the size */
636 newInst
= _mesa_alloc_instructions(p
->max_inst
);
638 _mesa_error(NULL
, GL_OUT_OF_MEMORY
, "vertex program build");
642 _mesa_copy_instructions(newInst
,
643 p
->program
->Base
.Instructions
,
644 p
->program
->Base
.NumInstructions
);
646 _mesa_free_instructions(p
->program
->Base
.Instructions
,
647 p
->program
->Base
.NumInstructions
);
649 p
->program
->Base
.Instructions
= newInst
;
652 nr
= p
->program
->Base
.NumInstructions
++;
654 inst
= &p
->program
->Base
.Instructions
[nr
];
655 inst
->Opcode
= (enum prog_opcode
) op
;
658 emit_arg( &inst
->SrcReg
[0], src0
);
659 emit_arg( &inst
->SrcReg
[1], src1
);
660 emit_arg( &inst
->SrcReg
[2], src2
);
662 emit_dst( &inst
->DstReg
, dest
, mask
);
664 debug_insn(inst
, fn
, line
);
668 #define emit_op3(p, op, dst, mask, src0, src1, src2) \
669 emit_op3fn(p, op, dst, mask, src0, src1, src2, __FUNCTION__, __LINE__)
671 #define emit_op2(p, op, dst, mask, src0, src1) \
672 emit_op3fn(p, op, dst, mask, src0, src1, undef, __FUNCTION__, __LINE__)
674 #define emit_op1(p, op, dst, mask, src0) \
675 emit_op3fn(p, op, dst, mask, src0, undef, undef, __FUNCTION__, __LINE__)
678 static struct ureg
make_temp( struct tnl_program
*p
, struct ureg reg
)
680 if (reg
.file
== PROGRAM_TEMPORARY
&&
681 !(p
->temp_reserved
& (1<<reg
.idx
)))
684 struct ureg temp
= get_temp(p
);
685 emit_op1(p
, OPCODE_MOV
, temp
, 0, reg
);
691 /* Currently no tracking performed of input/output/register size or
692 * active elements. Could be used to reduce these operations, as
693 * could the matrix type.
695 static void emit_matrix_transform_vec4( struct tnl_program
*p
,
697 const struct ureg
*mat
,
700 emit_op2(p
, OPCODE_DP4
, dest
, WRITEMASK_X
, src
, mat
[0]);
701 emit_op2(p
, OPCODE_DP4
, dest
, WRITEMASK_Y
, src
, mat
[1]);
702 emit_op2(p
, OPCODE_DP4
, dest
, WRITEMASK_Z
, src
, mat
[2]);
703 emit_op2(p
, OPCODE_DP4
, dest
, WRITEMASK_W
, src
, mat
[3]);
707 /* This version is much easier to implement if writemasks are not
708 * supported natively on the target or (like SSE), the target doesn't
709 * have a clean/obvious dotproduct implementation.
711 static void emit_transpose_matrix_transform_vec4( struct tnl_program
*p
,
713 const struct ureg
*mat
,
718 if (dest
.file
!= PROGRAM_TEMPORARY
)
723 emit_op2(p
, OPCODE_MUL
, tmp
, 0, swizzle1(src
,X
), mat
[0]);
724 emit_op3(p
, OPCODE_MAD
, tmp
, 0, swizzle1(src
,Y
), mat
[1], tmp
);
725 emit_op3(p
, OPCODE_MAD
, tmp
, 0, swizzle1(src
,Z
), mat
[2], tmp
);
726 emit_op3(p
, OPCODE_MAD
, dest
, 0, swizzle1(src
,W
), mat
[3], tmp
);
728 if (dest
.file
!= PROGRAM_TEMPORARY
)
729 release_temp(p
, tmp
);
733 static void emit_matrix_transform_vec3( struct tnl_program
*p
,
735 const struct ureg
*mat
,
738 emit_op2(p
, OPCODE_DP3
, dest
, WRITEMASK_X
, src
, mat
[0]);
739 emit_op2(p
, OPCODE_DP3
, dest
, WRITEMASK_Y
, src
, mat
[1]);
740 emit_op2(p
, OPCODE_DP3
, dest
, WRITEMASK_Z
, src
, mat
[2]);
744 static void emit_normalize_vec3( struct tnl_program
*p
,
749 /* XXX use this when drivers are ready for NRM3 */
750 emit_op1(p
, OPCODE_NRM3
, dest
, WRITEMASK_XYZ
, src
);
752 struct ureg tmp
= get_temp(p
);
753 emit_op2(p
, OPCODE_DP3
, tmp
, WRITEMASK_X
, src
, src
);
754 emit_op1(p
, OPCODE_RSQ
, tmp
, WRITEMASK_X
, tmp
);
755 emit_op2(p
, OPCODE_MUL
, dest
, 0, src
, swizzle1(tmp
, X
));
756 release_temp(p
, tmp
);
761 static void emit_passthrough( struct tnl_program
*p
,
765 struct ureg out
= register_output(p
, output
);
766 emit_op1(p
, OPCODE_MOV
, out
, 0, register_input(p
, input
));
770 static struct ureg
get_eye_position( struct tnl_program
*p
)
772 if (is_undef(p
->eye_position
)) {
773 struct ureg pos
= register_input( p
, VERT_ATTRIB_POS
);
774 struct ureg modelview
[4];
776 p
->eye_position
= reserve_temp(p
);
779 register_matrix_param5( p
, STATE_MODELVIEW_MATRIX
, 0, 0, 3,
782 emit_matrix_transform_vec4(p
, p
->eye_position
, modelview
, pos
);
785 register_matrix_param5( p
, STATE_MODELVIEW_MATRIX
, 0, 0, 3,
786 STATE_MATRIX_TRANSPOSE
, modelview
);
788 emit_transpose_matrix_transform_vec4(p
, p
->eye_position
, modelview
, pos
);
792 return p
->eye_position
;
796 static struct ureg
get_eye_position_z( struct tnl_program
*p
)
798 if (!is_undef(p
->eye_position
))
799 return swizzle1(p
->eye_position
, Z
);
801 if (is_undef(p
->eye_position_z
)) {
802 struct ureg pos
= register_input( p
, VERT_ATTRIB_POS
);
803 struct ureg modelview
[4];
805 p
->eye_position_z
= reserve_temp(p
);
807 register_matrix_param5( p
, STATE_MODELVIEW_MATRIX
, 0, 0, 3,
810 emit_op2(p
, OPCODE_DP4
, p
->eye_position_z
, 0, pos
, modelview
[2]);
813 return p
->eye_position_z
;
817 static struct ureg
get_eye_position_normalized( struct tnl_program
*p
)
819 if (is_undef(p
->eye_position_normalized
)) {
820 struct ureg eye
= get_eye_position(p
);
821 p
->eye_position_normalized
= reserve_temp(p
);
822 emit_normalize_vec3(p
, p
->eye_position_normalized
, eye
);
825 return p
->eye_position_normalized
;
829 static struct ureg
get_transformed_normal( struct tnl_program
*p
)
831 if (is_undef(p
->transformed_normal
) &&
832 !p
->state
->need_eye_coords
&&
833 !p
->state
->normalize
&&
834 !(p
->state
->need_eye_coords
== p
->state
->rescale_normals
))
836 p
->transformed_normal
= register_input(p
, VERT_ATTRIB_NORMAL
);
838 else if (is_undef(p
->transformed_normal
))
840 struct ureg normal
= register_input(p
, VERT_ATTRIB_NORMAL
);
841 struct ureg mvinv
[3];
842 struct ureg transformed_normal
= reserve_temp(p
);
844 if (p
->state
->need_eye_coords
) {
845 register_matrix_param5( p
, STATE_MODELVIEW_MATRIX
, 0, 0, 2,
846 STATE_MATRIX_INVTRANS
, mvinv
);
848 /* Transform to eye space:
850 emit_matrix_transform_vec3( p
, transformed_normal
, mvinv
, normal
);
851 normal
= transformed_normal
;
854 /* Normalize/Rescale:
856 if (p
->state
->normalize
) {
857 emit_normalize_vec3( p
, transformed_normal
, normal
);
858 normal
= transformed_normal
;
860 else if (p
->state
->need_eye_coords
== p
->state
->rescale_normals
) {
861 /* This is already adjusted for eye/non-eye rendering:
863 struct ureg rescale
= register_param2(p
, STATE_INTERNAL
,
866 emit_op2( p
, OPCODE_MUL
, transformed_normal
, 0, normal
, rescale
);
867 normal
= transformed_normal
;
870 assert(normal
.file
== PROGRAM_TEMPORARY
);
871 p
->transformed_normal
= normal
;
874 return p
->transformed_normal
;
878 static void build_hpos( struct tnl_program
*p
)
880 struct ureg pos
= register_input( p
, VERT_ATTRIB_POS
);
881 struct ureg hpos
= register_output( p
, VERT_RESULT_HPOS
);
885 register_matrix_param5( p
, STATE_MVP_MATRIX
, 0, 0, 3,
887 emit_matrix_transform_vec4( p
, hpos
, mvp
, pos
);
890 register_matrix_param5( p
, STATE_MVP_MATRIX
, 0, 0, 3,
891 STATE_MATRIX_TRANSPOSE
, mvp
);
892 emit_transpose_matrix_transform_vec4( p
, hpos
, mvp
, pos
);
897 static GLuint
material_attrib( GLuint side
, GLuint property
)
899 return (property
- STATE_AMBIENT
) * 2 + side
;
904 * Get a bitmask of which material values vary on a per-vertex basis.
906 static void set_material_flags( struct tnl_program
*p
)
908 p
->color_materials
= 0;
911 if (p
->state
->light_color_material
) {
913 p
->color_materials
= p
->state
->light_color_material_mask
;
916 p
->materials
|= p
->state
->light_material_mask
;
920 /* XXX temporary!!! */
921 #define _TNL_ATTRIB_MAT_FRONT_AMBIENT 32
923 static struct ureg
get_material( struct tnl_program
*p
, GLuint side
,
926 GLuint attrib
= material_attrib(side
, property
);
928 if (p
->color_materials
& (1<<attrib
))
929 return register_input(p
, VERT_ATTRIB_COLOR0
);
930 else if (p
->materials
& (1<<attrib
))
931 return register_input( p
, attrib
+ _TNL_ATTRIB_MAT_FRONT_AMBIENT
);
933 return register_param3( p
, STATE_MATERIAL
, side
, property
);
936 #define SCENE_COLOR_BITS(side) (( MAT_BIT_FRONT_EMISSION | \
937 MAT_BIT_FRONT_AMBIENT | \
938 MAT_BIT_FRONT_DIFFUSE) << (side))
942 * Either return a precalculated constant value or emit code to
943 * calculate these values dynamically in the case where material calls
944 * are present between begin/end pairs.
946 * Probably want to shift this to the program compilation phase - if
947 * we always emitted the calculation here, a smart compiler could
948 * detect that it was constant (given a certain set of inputs), and
949 * lift it out of the main loop. That way the programs created here
950 * would be independent of the vertex_buffer details.
952 static struct ureg
get_scenecolor( struct tnl_program
*p
, GLuint side
)
954 if (p
->materials
& SCENE_COLOR_BITS(side
)) {
955 struct ureg lm_ambient
= register_param1(p
, STATE_LIGHTMODEL_AMBIENT
);
956 struct ureg material_emission
= get_material(p
, side
, STATE_EMISSION
);
957 struct ureg material_ambient
= get_material(p
, side
, STATE_AMBIENT
);
958 struct ureg material_diffuse
= get_material(p
, side
, STATE_DIFFUSE
);
959 struct ureg tmp
= make_temp(p
, material_diffuse
);
960 emit_op3(p
, OPCODE_MAD
, tmp
, WRITEMASK_XYZ
, lm_ambient
,
961 material_ambient
, material_emission
);
965 return register_param2( p
, STATE_LIGHTMODEL_SCENECOLOR
, side
);
969 static struct ureg
get_lightprod( struct tnl_program
*p
, GLuint light
,
970 GLuint side
, GLuint property
)
972 GLuint attrib
= material_attrib(side
, property
);
973 if (p
->materials
& (1<<attrib
)) {
974 struct ureg light_value
=
975 register_param3(p
, STATE_LIGHT
, light
, property
);
976 struct ureg material_value
= get_material(p
, side
, property
);
977 struct ureg tmp
= get_temp(p
);
978 emit_op2(p
, OPCODE_MUL
, tmp
, 0, light_value
, material_value
);
982 return register_param4(p
, STATE_LIGHTPROD
, light
, side
, property
);
986 static struct ureg
calculate_light_attenuation( struct tnl_program
*p
,
991 struct ureg attenuation
= register_param3(p
, STATE_LIGHT
, i
,
993 struct ureg att
= get_temp(p
);
995 /* Calculate spot attenuation:
997 if (!p
->state
->unit
[i
].light_spotcutoff_is_180
) {
998 struct ureg spot_dir_norm
= register_param3(p
, STATE_INTERNAL
,
999 STATE_LIGHT_SPOT_DIR_NORMALIZED
, i
);
1000 struct ureg spot
= get_temp(p
);
1001 struct ureg slt
= get_temp(p
);
1003 emit_op2(p
, OPCODE_DP3
, spot
, 0, negate(VPpli
), spot_dir_norm
);
1004 emit_op2(p
, OPCODE_SLT
, slt
, 0, swizzle1(spot_dir_norm
,W
), spot
);
1005 emit_op2(p
, OPCODE_POW
, spot
, 0, spot
, swizzle1(attenuation
, W
));
1006 emit_op2(p
, OPCODE_MUL
, att
, 0, slt
, spot
);
1008 release_temp(p
, spot
);
1009 release_temp(p
, slt
);
1012 /* Calculate distance attenuation:
1014 if (p
->state
->unit
[i
].light_attenuated
) {
1016 emit_op1(p
, OPCODE_RCP
, dist
, WRITEMASK_YZ
, dist
);
1018 emit_op2(p
, OPCODE_MUL
, dist
, WRITEMASK_XZ
, dist
, swizzle1(dist
,Y
));
1020 emit_op2(p
, OPCODE_DP3
, dist
, 0, attenuation
, dist
);
1022 if (!p
->state
->unit
[i
].light_spotcutoff_is_180
) {
1024 emit_op1(p
, OPCODE_RCP
, dist
, 0, dist
);
1025 /* spot-atten * dist-atten */
1026 emit_op2(p
, OPCODE_MUL
, att
, 0, dist
, att
);
1030 emit_op1(p
, OPCODE_RCP
, att
, 0, dist
);
1040 * lit.y = MAX(0, dots.x)
1041 * lit.z = SLT(0, dots.x)
1043 static void emit_degenerate_lit( struct tnl_program
*p
,
1047 struct ureg id
= get_identity_param(p
); /* id = {0,0,0,1} */
1049 /* Note that lit.x & lit.w will not be examined. Note also that
1050 * dots.xyzw == dots.xxxx.
1053 /* MAX lit, id, dots;
1055 emit_op2(p
, OPCODE_MAX
, lit
, WRITEMASK_XYZW
, id
, dots
);
1057 /* result[2] = (in > 0 ? 1 : 0)
1058 * SLT lit.z, id.z, dots; # lit.z = (0 < dots.z) ? 1 : 0
1060 emit_op2(p
, OPCODE_SLT
, lit
, WRITEMASK_Z
, swizzle1(id
,Z
), dots
);
1064 /* Need to add some addtional parameters to allow lighting in object
1065 * space - STATE_SPOT_DIRECTION and STATE_HALF_VECTOR implicitly assume eye
1068 static void build_lighting( struct tnl_program
*p
)
1070 const GLboolean twoside
= p
->state
->light_twoside
;
1071 const GLboolean separate
= p
->state
->separate_specular
;
1072 GLuint nr_lights
= 0, count
= 0;
1073 struct ureg normal
= get_transformed_normal(p
);
1074 struct ureg lit
= get_temp(p
);
1075 struct ureg dots
= get_temp(p
);
1076 struct ureg _col0
= undef
, _col1
= undef
;
1077 struct ureg _bfc0
= undef
, _bfc1
= undef
;
1082 * dots.x = dot(normal, VPpli)
1083 * dots.y = dot(normal, halfAngle)
1084 * dots.z = back.shininess
1085 * dots.w = front.shininess
1088 for (i
= 0; i
< MAX_LIGHTS
; i
++)
1089 if (p
->state
->unit
[i
].light_enabled
)
1092 set_material_flags(p
);
1095 if (!p
->state
->material_shininess_is_zero
) {
1096 struct ureg shininess
= get_material(p
, 0, STATE_SHININESS
);
1097 emit_op1(p
, OPCODE_MOV
, dots
, WRITEMASK_W
, swizzle1(shininess
,X
));
1098 release_temp(p
, shininess
);
1101 _col0
= make_temp(p
, get_scenecolor(p
, 0));
1103 _col1
= make_temp(p
, get_identity_param(p
));
1109 if (!p
->state
->material_shininess_is_zero
) {
1110 /* Note that we negate the back-face specular exponent here.
1111 * The negation will be un-done later in the back-face code below.
1113 struct ureg shininess
= get_material(p
, 1, STATE_SHININESS
);
1114 emit_op1(p
, OPCODE_MOV
, dots
, WRITEMASK_Z
,
1115 negate(swizzle1(shininess
,X
)));
1116 release_temp(p
, shininess
);
1119 _bfc0
= make_temp(p
, get_scenecolor(p
, 1));
1121 _bfc1
= make_temp(p
, get_identity_param(p
));
1126 /* If no lights, still need to emit the scenecolor.
1129 struct ureg res0
= register_output( p
, VERT_RESULT_COL0
);
1130 emit_op1(p
, OPCODE_MOV
, res0
, 0, _col0
);
1134 struct ureg res1
= register_output( p
, VERT_RESULT_COL1
);
1135 emit_op1(p
, OPCODE_MOV
, res1
, 0, _col1
);
1139 struct ureg res0
= register_output( p
, VERT_RESULT_BFC0
);
1140 emit_op1(p
, OPCODE_MOV
, res0
, 0, _bfc0
);
1143 if (twoside
&& separate
) {
1144 struct ureg res1
= register_output( p
, VERT_RESULT_BFC1
);
1145 emit_op1(p
, OPCODE_MOV
, res1
, 0, _bfc1
);
1148 if (nr_lights
== 0) {
1153 for (i
= 0; i
< MAX_LIGHTS
; i
++) {
1154 if (p
->state
->unit
[i
].light_enabled
) {
1155 struct ureg half
= undef
;
1156 struct ureg att
= undef
, VPpli
= undef
;
1160 if (p
->state
->unit
[i
].light_eyepos3_is_zero
) {
1161 /* Can used precomputed constants in this case.
1162 * Attenuation never applies to infinite lights.
1164 VPpli
= register_param3(p
, STATE_INTERNAL
,
1165 STATE_LIGHT_POSITION_NORMALIZED
, i
);
1167 if (!p
->state
->material_shininess_is_zero
) {
1168 if (p
->state
->light_local_viewer
) {
1169 struct ureg eye_hat
= get_eye_position_normalized(p
);
1171 emit_op2(p
, OPCODE_SUB
, half
, 0, VPpli
, eye_hat
);
1172 emit_normalize_vec3(p
, half
, half
);
1175 half
= register_param3(p
, STATE_INTERNAL
,
1176 STATE_LIGHT_HALF_VECTOR
, i
);
1181 struct ureg Ppli
= register_param3(p
, STATE_INTERNAL
,
1182 STATE_LIGHT_POSITION
, i
);
1183 struct ureg V
= get_eye_position(p
);
1184 struct ureg dist
= get_temp(p
);
1186 VPpli
= get_temp(p
);
1188 /* Calculate VPpli vector
1190 emit_op2(p
, OPCODE_SUB
, VPpli
, 0, Ppli
, V
);
1192 /* Normalize VPpli. The dist value also used in
1193 * attenuation below.
1195 emit_op2(p
, OPCODE_DP3
, dist
, 0, VPpli
, VPpli
);
1196 emit_op1(p
, OPCODE_RSQ
, dist
, 0, dist
);
1197 emit_op2(p
, OPCODE_MUL
, VPpli
, 0, VPpli
, dist
);
1199 /* Calculate attenuation:
1201 if (!p
->state
->unit
[i
].light_spotcutoff_is_180
||
1202 p
->state
->unit
[i
].light_attenuated
) {
1203 att
= calculate_light_attenuation(p
, i
, VPpli
, dist
);
1206 /* Calculate viewer direction, or use infinite viewer:
1208 if (!p
->state
->material_shininess_is_zero
) {
1211 if (p
->state
->light_local_viewer
) {
1212 struct ureg eye_hat
= get_eye_position_normalized(p
);
1213 emit_op2(p
, OPCODE_SUB
, half
, 0, VPpli
, eye_hat
);
1216 struct ureg z_dir
= swizzle(get_identity_param(p
),X
,Y
,W
,Z
);
1217 emit_op2(p
, OPCODE_ADD
, half
, 0, VPpli
, z_dir
);
1220 emit_normalize_vec3(p
, half
, half
);
1223 release_temp(p
, dist
);
1226 /* Calculate dot products:
1228 if (p
->state
->material_shininess_is_zero
) {
1229 emit_op2(p
, OPCODE_DP3
, dots
, 0, normal
, VPpli
);
1232 emit_op2(p
, OPCODE_DP3
, dots
, WRITEMASK_X
, normal
, VPpli
);
1233 emit_op2(p
, OPCODE_DP3
, dots
, WRITEMASK_Y
, normal
, half
);
1236 /* Front face lighting:
1239 struct ureg ambient
= get_lightprod(p
, i
, 0, STATE_AMBIENT
);
1240 struct ureg diffuse
= get_lightprod(p
, i
, 0, STATE_DIFFUSE
);
1241 struct ureg specular
= get_lightprod(p
, i
, 0, STATE_SPECULAR
);
1242 struct ureg res0
, res1
;
1243 GLuint mask0
, mask1
;
1245 if (count
== nr_lights
) {
1247 mask0
= WRITEMASK_XYZ
;
1248 mask1
= WRITEMASK_XYZ
;
1249 res0
= register_output( p
, VERT_RESULT_COL0
);
1250 res1
= register_output( p
, VERT_RESULT_COL1
);
1254 mask1
= WRITEMASK_XYZ
;
1256 res1
= register_output( p
, VERT_RESULT_COL0
);
1266 if (!is_undef(att
)) {
1267 /* light is attenuated by distance */
1268 emit_op1(p
, OPCODE_LIT
, lit
, 0, dots
);
1269 emit_op2(p
, OPCODE_MUL
, lit
, 0, lit
, att
);
1270 emit_op3(p
, OPCODE_MAD
, _col0
, 0, swizzle1(lit
,X
), ambient
, _col0
);
1272 else if (!p
->state
->material_shininess_is_zero
) {
1273 /* there's a non-zero specular term */
1274 emit_op1(p
, OPCODE_LIT
, lit
, 0, dots
);
1275 emit_op2(p
, OPCODE_ADD
, _col0
, 0, ambient
, _col0
);
1278 /* no attenutation, no specular */
1279 emit_degenerate_lit(p
, lit
, dots
);
1280 emit_op2(p
, OPCODE_ADD
, _col0
, 0, ambient
, _col0
);
1283 emit_op3(p
, OPCODE_MAD
, res0
, mask0
, swizzle1(lit
,Y
), diffuse
, _col0
);
1284 emit_op3(p
, OPCODE_MAD
, res1
, mask1
, swizzle1(lit
,Z
), specular
, _col1
);
1286 release_temp(p
, ambient
);
1287 release_temp(p
, diffuse
);
1288 release_temp(p
, specular
);
1291 /* Back face lighting:
1294 struct ureg ambient
= get_lightprod(p
, i
, 1, STATE_AMBIENT
);
1295 struct ureg diffuse
= get_lightprod(p
, i
, 1, STATE_DIFFUSE
);
1296 struct ureg specular
= get_lightprod(p
, i
, 1, STATE_SPECULAR
);
1297 struct ureg res0
, res1
;
1298 GLuint mask0
, mask1
;
1300 if (count
== nr_lights
) {
1302 mask0
= WRITEMASK_XYZ
;
1303 mask1
= WRITEMASK_XYZ
;
1304 res0
= register_output( p
, VERT_RESULT_BFC0
);
1305 res1
= register_output( p
, VERT_RESULT_BFC1
);
1309 mask1
= WRITEMASK_XYZ
;
1311 res1
= register_output( p
, VERT_RESULT_BFC0
);
1321 /* For the back face we need to negate the X and Y component
1322 * dot products. dots.Z has the negated back-face specular
1323 * exponent. We swizzle that into the W position. This
1324 * negation makes the back-face specular term positive again.
1326 dots
= negate(swizzle(dots
,X
,Y
,W
,Z
));
1328 if (!is_undef(att
)) {
1329 emit_op1(p
, OPCODE_LIT
, lit
, 0, dots
);
1330 emit_op2(p
, OPCODE_MUL
, lit
, 0, lit
, att
);
1331 emit_op3(p
, OPCODE_MAD
, _bfc0
, 0, swizzle1(lit
,X
), ambient
, _bfc0
);
1333 else if (!p
->state
->material_shininess_is_zero
) {
1334 emit_op1(p
, OPCODE_LIT
, lit
, 0, dots
);
1335 emit_op2(p
, OPCODE_ADD
, _bfc0
, 0, ambient
, _bfc0
); /**/
1338 emit_degenerate_lit(p
, lit
, dots
);
1339 emit_op2(p
, OPCODE_ADD
, _bfc0
, 0, ambient
, _bfc0
);
1342 emit_op3(p
, OPCODE_MAD
, res0
, mask0
, swizzle1(lit
,Y
), diffuse
, _bfc0
);
1343 emit_op3(p
, OPCODE_MAD
, res1
, mask1
, swizzle1(lit
,Z
), specular
, _bfc1
);
1344 /* restore dots to its original state for subsequent lights
1345 * by negating and swizzling again.
1347 dots
= negate(swizzle(dots
,X
,Y
,W
,Z
));
1349 release_temp(p
, ambient
);
1350 release_temp(p
, diffuse
);
1351 release_temp(p
, specular
);
1354 release_temp(p
, half
);
1355 release_temp(p
, VPpli
);
1356 release_temp(p
, att
);
1364 static void build_fog( struct tnl_program
*p
)
1366 struct ureg fog
= register_output(p
, VERT_RESULT_FOGC
);
1369 if (p
->state
->fog_source_is_depth
) {
1370 input
= get_eye_position_z(p
);
1373 input
= swizzle1(register_input(p
, VERT_ATTRIB_FOG
), X
);
1376 if (p
->state
->fog_mode
&& p
->state
->tnl_do_vertex_fog
) {
1377 struct ureg params
= register_param2(p
, STATE_INTERNAL
,
1378 STATE_FOG_PARAMS_OPTIMIZED
);
1379 struct ureg tmp
= get_temp(p
);
1380 GLboolean useabs
= (p
->state
->fog_mode
!= FOG_EXP2
);
1383 emit_op1(p
, OPCODE_ABS
, tmp
, 0, input
);
1386 switch (p
->state
->fog_mode
) {
1388 struct ureg id
= get_identity_param(p
);
1389 emit_op3(p
, OPCODE_MAD
, tmp
, 0, useabs
? tmp
: input
,
1390 swizzle1(params
,X
), swizzle1(params
,Y
));
1391 emit_op2(p
, OPCODE_MAX
, tmp
, 0, tmp
, swizzle1(id
,X
)); /* saturate */
1392 emit_op2(p
, OPCODE_MIN
, fog
, WRITEMASK_X
, tmp
, swizzle1(id
,W
));
1396 emit_op2(p
, OPCODE_MUL
, tmp
, 0, useabs
? tmp
: input
,
1397 swizzle1(params
,Z
));
1398 emit_op1(p
, OPCODE_EX2
, fog
, WRITEMASK_X
, negate(tmp
));
1401 emit_op2(p
, OPCODE_MUL
, tmp
, 0, input
, swizzle1(params
,W
));
1402 emit_op2(p
, OPCODE_MUL
, tmp
, 0, tmp
, tmp
);
1403 emit_op1(p
, OPCODE_EX2
, fog
, WRITEMASK_X
, negate(tmp
));
1407 release_temp(p
, tmp
);
1410 /* results = incoming fog coords (compute fog per-fragment later)
1412 * KW: Is it really necessary to do anything in this case?
1413 * BP: Yes, we always need to compute the absolute value, unless
1414 * we want to push that down into the fragment program...
1416 GLboolean useabs
= GL_TRUE
;
1417 emit_op1(p
, useabs
? OPCODE_ABS
: OPCODE_MOV
, fog
, WRITEMASK_X
, input
);
1422 static void build_reflect_texgen( struct tnl_program
*p
,
1426 struct ureg normal
= get_transformed_normal(p
);
1427 struct ureg eye_hat
= get_eye_position_normalized(p
);
1428 struct ureg tmp
= get_temp(p
);
1431 emit_op2(p
, OPCODE_DP3
, tmp
, 0, normal
, eye_hat
);
1433 emit_op2(p
, OPCODE_ADD
, tmp
, 0, tmp
, tmp
);
1435 emit_op3(p
, OPCODE_MAD
, dest
, writemask
, negate(tmp
), normal
, eye_hat
);
1437 release_temp(p
, tmp
);
1441 static void build_sphere_texgen( struct tnl_program
*p
,
1445 struct ureg normal
= get_transformed_normal(p
);
1446 struct ureg eye_hat
= get_eye_position_normalized(p
);
1447 struct ureg tmp
= get_temp(p
);
1448 struct ureg half
= register_scalar_const(p
, .5);
1449 struct ureg r
= get_temp(p
);
1450 struct ureg inv_m
= get_temp(p
);
1451 struct ureg id
= get_identity_param(p
);
1453 /* Could share the above calculations, but it would be
1454 * a fairly odd state for someone to set (both sphere and
1455 * reflection active for different texture coordinate
1456 * components. Of course - if two texture units enable
1457 * reflect and/or sphere, things start to tilt in favour
1458 * of seperating this out:
1462 emit_op2(p
, OPCODE_DP3
, tmp
, 0, normal
, eye_hat
);
1464 emit_op2(p
, OPCODE_ADD
, tmp
, 0, tmp
, tmp
);
1466 emit_op3(p
, OPCODE_MAD
, r
, 0, negate(tmp
), normal
, eye_hat
);
1468 emit_op2(p
, OPCODE_ADD
, tmp
, 0, r
, swizzle(id
,X
,Y
,W
,Z
));
1469 /* rx^2 + ry^2 + (rz+1)^2 */
1470 emit_op2(p
, OPCODE_DP3
, tmp
, 0, tmp
, tmp
);
1472 emit_op1(p
, OPCODE_RSQ
, tmp
, 0, tmp
);
1474 emit_op2(p
, OPCODE_MUL
, inv_m
, 0, tmp
, half
);
1476 emit_op3(p
, OPCODE_MAD
, dest
, writemask
, r
, inv_m
, half
);
1478 release_temp(p
, tmp
);
1480 release_temp(p
, inv_m
);
1484 static void build_texture_transform( struct tnl_program
*p
)
1488 for (i
= 0; i
< MAX_TEXTURE_COORD_UNITS
; i
++) {
1490 if (!(p
->state
->fragprog_inputs_read
& FRAG_BIT_TEX(i
)))
1493 if (p
->state
->unit
[i
].texgen_enabled
||
1494 p
->state
->unit
[i
].texmat_enabled
) {
1496 GLuint texmat_enabled
= p
->state
->unit
[i
].texmat_enabled
;
1497 struct ureg out
= register_output(p
, VERT_RESULT_TEX0
+ i
);
1498 struct ureg out_texgen
= undef
;
1500 if (p
->state
->unit
[i
].texgen_enabled
) {
1501 GLuint copy_mask
= 0;
1502 GLuint sphere_mask
= 0;
1503 GLuint reflect_mask
= 0;
1504 GLuint normal_mask
= 0;
1508 out_texgen
= get_temp(p
);
1512 modes
[0] = p
->state
->unit
[i
].texgen_mode0
;
1513 modes
[1] = p
->state
->unit
[i
].texgen_mode1
;
1514 modes
[2] = p
->state
->unit
[i
].texgen_mode2
;
1515 modes
[3] = p
->state
->unit
[i
].texgen_mode3
;
1517 for (j
= 0; j
< 4; j
++) {
1519 case TXG_OBJ_LINEAR
: {
1520 struct ureg obj
= register_input(p
, VERT_ATTRIB_POS
);
1522 register_param3(p
, STATE_TEXGEN
, i
,
1523 STATE_TEXGEN_OBJECT_S
+ j
);
1525 emit_op2(p
, OPCODE_DP4
, out_texgen
, WRITEMASK_X
<< j
,
1529 case TXG_EYE_LINEAR
: {
1530 struct ureg eye
= get_eye_position(p
);
1532 register_param3(p
, STATE_TEXGEN
, i
,
1533 STATE_TEXGEN_EYE_S
+ j
);
1535 emit_op2(p
, OPCODE_DP4
, out_texgen
, WRITEMASK_X
<< j
,
1539 case TXG_SPHERE_MAP
:
1540 sphere_mask
|= WRITEMASK_X
<< j
;
1542 case TXG_REFLECTION_MAP
:
1543 reflect_mask
|= WRITEMASK_X
<< j
;
1545 case TXG_NORMAL_MAP
:
1546 normal_mask
|= WRITEMASK_X
<< j
;
1549 copy_mask
|= WRITEMASK_X
<< j
;
1554 build_sphere_texgen(p
, out_texgen
, sphere_mask
);
1558 build_reflect_texgen(p
, out_texgen
, reflect_mask
);
1562 struct ureg normal
= get_transformed_normal(p
);
1563 emit_op1(p
, OPCODE_MOV
, out_texgen
, normal_mask
, normal
);
1567 struct ureg in
= register_input(p
, VERT_ATTRIB_TEX0
+i
);
1568 emit_op1(p
, OPCODE_MOV
, out_texgen
, copy_mask
, in
);
1572 if (texmat_enabled
) {
1573 struct ureg texmat
[4];
1574 struct ureg in
= (!is_undef(out_texgen
) ?
1576 register_input(p
, VERT_ATTRIB_TEX0
+i
));
1578 register_matrix_param5( p
, STATE_TEXTURE_MATRIX
, i
, 0, 3,
1580 emit_matrix_transform_vec4( p
, out
, texmat
, in
);
1583 register_matrix_param5( p
, STATE_TEXTURE_MATRIX
, i
, 0, 3,
1584 STATE_MATRIX_TRANSPOSE
, texmat
);
1585 emit_transpose_matrix_transform_vec4( p
, out
, texmat
, in
);
1592 emit_passthrough(p
, VERT_ATTRIB_TEX0
+i
, VERT_RESULT_TEX0
+i
);
1599 * Point size attenuation computation.
1601 static void build_atten_pointsize( struct tnl_program
*p
)
1603 struct ureg eye
= get_eye_position_z(p
);
1604 struct ureg state_size
= register_param1(p
, STATE_POINT_SIZE
);
1605 struct ureg state_attenuation
= register_param1(p
, STATE_POINT_ATTENUATION
);
1606 struct ureg out
= register_output(p
, VERT_RESULT_PSIZ
);
1607 struct ureg ut
= get_temp(p
);
1610 emit_op1(p
, OPCODE_ABS
, ut
, WRITEMASK_Y
, swizzle1(eye
, Z
));
1611 /* p1 + dist * (p2 + dist * p3); */
1612 emit_op3(p
, OPCODE_MAD
, ut
, WRITEMASK_X
, swizzle1(ut
, Y
),
1613 swizzle1(state_attenuation
, Z
), swizzle1(state_attenuation
, Y
));
1614 emit_op3(p
, OPCODE_MAD
, ut
, WRITEMASK_X
, swizzle1(ut
, Y
),
1615 ut
, swizzle1(state_attenuation
, X
));
1617 /* 1 / sqrt(factor) */
1618 emit_op1(p
, OPCODE_RSQ
, ut
, WRITEMASK_X
, ut
);
1621 /* out = pointSize / sqrt(factor) */
1622 emit_op2(p
, OPCODE_MUL
, out
, WRITEMASK_X
, ut
, state_size
);
1624 /* this is a good place to clamp the point size since there's likely
1625 * no hardware registers to clamp point size at rasterization time.
1627 emit_op2(p
, OPCODE_MUL
, ut
, WRITEMASK_X
, ut
, state_size
);
1628 emit_op2(p
, OPCODE_MAX
, ut
, WRITEMASK_X
, ut
, swizzle1(state_size
, Y
));
1629 emit_op2(p
, OPCODE_MIN
, out
, WRITEMASK_X
, ut
, swizzle1(state_size
, Z
));
1632 release_temp(p
, ut
);
1637 * Emit constant point size.
1639 static void build_constant_pointsize( struct tnl_program
*p
)
1641 struct ureg state_size
= register_param1(p
, STATE_POINT_SIZE
);
1642 struct ureg out
= register_output(p
, VERT_RESULT_PSIZ
);
1643 emit_op1(p
, OPCODE_MOV
, out
, WRITEMASK_X
, state_size
);
1648 * Pass-though per-vertex point size, from user's point size array.
1650 static void build_array_pointsize( struct tnl_program
*p
)
1652 struct ureg in
= register_input(p
, VERT_ATTRIB_POINT_SIZE
);
1653 struct ureg out
= register_output(p
, VERT_RESULT_PSIZ
);
1654 emit_op1(p
, OPCODE_MOV
, out
, WRITEMASK_X
, in
);
1658 static void build_tnl_program( struct tnl_program
*p
)
1660 /* Emit the program, starting with modelviewproject:
1664 /* Lighting calculations:
1666 if (p
->state
->fragprog_inputs_read
& (FRAG_BIT_COL0
|FRAG_BIT_COL1
)) {
1667 if (p
->state
->light_global_enabled
)
1670 if (p
->state
->fragprog_inputs_read
& FRAG_BIT_COL0
)
1671 emit_passthrough(p
, VERT_ATTRIB_COLOR0
, VERT_RESULT_COL0
);
1673 if (p
->state
->fragprog_inputs_read
& FRAG_BIT_COL1
)
1674 emit_passthrough(p
, VERT_ATTRIB_COLOR1
, VERT_RESULT_COL1
);
1678 if ((p
->state
->fragprog_inputs_read
& FRAG_BIT_FOGC
) ||
1679 p
->state
->fog_mode
!= FOG_NONE
)
1682 if (p
->state
->fragprog_inputs_read
& FRAG_BITS_TEX_ANY
)
1683 build_texture_transform(p
);
1685 if (p
->state
->point_attenuated
)
1686 build_atten_pointsize(p
);
1687 else if (p
->state
->point_array
)
1688 build_array_pointsize(p
);
1691 build_constant_pointsize(p
);
1693 (void) build_constant_pointsize
;
1698 emit_op1(p
, OPCODE_END
, undef
, 0, undef
);
1703 _mesa_printf ("\n");
1709 create_new_program( const struct state_key
*key
,
1710 struct gl_vertex_program
*program
,
1713 struct tnl_program p
;
1715 _mesa_memset(&p
, 0, sizeof(p
));
1717 p
.program
= program
;
1718 p
.eye_position
= undef
;
1719 p
.eye_position_z
= undef
;
1720 p
.eye_position_normalized
= undef
;
1721 p
.transformed_normal
= undef
;
1725 if (max_temps
>= sizeof(int) * 8)
1726 p
.temp_reserved
= 0;
1728 p
.temp_reserved
= ~((1<<max_temps
)-1);
1730 /* Start by allocating 32 instructions.
1731 * If we need more, we'll grow the instruction array as needed.
1734 p
.program
->Base
.Instructions
= _mesa_alloc_instructions(p
.max_inst
);
1735 p
.program
->Base
.String
= NULL
;
1736 p
.program
->Base
.NumInstructions
=
1737 p
.program
->Base
.NumTemporaries
=
1738 p
.program
->Base
.NumParameters
=
1739 p
.program
->Base
.NumAttributes
= p
.program
->Base
.NumAddressRegs
= 0;
1740 p
.program
->Base
.Parameters
= _mesa_new_parameter_list();
1741 p
.program
->Base
.InputsRead
= 0;
1742 p
.program
->Base
.OutputsWritten
= 0;
1744 build_tnl_program( &p
);
1749 * Return a vertex program which implements the current fixed-function
1750 * transform/lighting/texgen operations.
1751 * XXX move this into core mesa (main/)
1753 struct gl_vertex_program
*
1754 _mesa_get_fixed_func_vertex_program(GLcontext
*ctx
)
1756 struct gl_vertex_program
*prog
;
1757 struct state_key key
;
1759 /* Grab all the relevent state and put it in a single structure:
1761 make_state_key(ctx
, &key
);
1763 /* Look for an already-prepared program for this state:
1765 prog
= (struct gl_vertex_program
*)
1766 _mesa_search_program_cache(ctx
->VertexProgram
.Cache
, &key
, sizeof(key
));
1769 /* OK, we'll have to build a new one */
1771 _mesa_printf("Build new TNL program\n");
1773 prog
= (struct gl_vertex_program
*)
1774 ctx
->Driver
.NewProgram(ctx
, GL_VERTEX_PROGRAM_ARB
, 0);
1778 create_new_program( &key
, prog
,
1779 ctx
->Const
.VertexProgram
.MaxTemps
);
1782 if (ctx
->Driver
.ProgramStringNotify
)
1783 ctx
->Driver
.ProgramStringNotify( ctx
, GL_VERTEX_PROGRAM_ARB
,
1786 _mesa_program_cache_insert(ctx
, ctx
->VertexProgram
.Cache
,
1787 &key
, sizeof(key
), &prog
->Base
);