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 point_array
:1;
67 unsigned texture_enabled_global
:1;
68 unsigned fragprog_inputs_read
:12;
71 unsigned light_enabled
:1;
72 unsigned light_eyepos3_is_zero
:1;
73 unsigned light_spotcutoff_is_180
:1;
74 unsigned light_attenuated
:1;
75 unsigned texunit_really_enabled
:1;
76 unsigned texmat_enabled
:1;
77 unsigned texgen_enabled
:4;
78 unsigned texgen_mode0
:4;
79 unsigned texgen_mode1
:4;
80 unsigned texgen_mode2
:4;
81 unsigned texgen_mode3
:4;
92 static GLuint
translate_fog_mode( GLenum mode
)
95 case GL_LINEAR
: return FOG_LINEAR
;
96 case GL_EXP
: return FOG_EXP
;
97 case GL_EXP2
: return FOG_EXP2
;
98 default: return FOG_NONE
;
103 #define TXG_OBJ_LINEAR 1
104 #define TXG_EYE_LINEAR 2
105 #define TXG_SPHERE_MAP 3
106 #define TXG_REFLECTION_MAP 4
107 #define TXG_NORMAL_MAP 5
109 static GLuint
translate_texgen( GLboolean enabled
, GLenum mode
)
115 case GL_OBJECT_LINEAR
: return TXG_OBJ_LINEAR
;
116 case GL_EYE_LINEAR
: return TXG_EYE_LINEAR
;
117 case GL_SPHERE_MAP
: return TXG_SPHERE_MAP
;
118 case GL_REFLECTION_MAP_NV
: return TXG_REFLECTION_MAP
;
119 case GL_NORMAL_MAP_NV
: return TXG_NORMAL_MAP
;
120 default: return TXG_NONE
;
126 * Returns bitmask of flags indicating which materials are set per-vertex
128 * XXX get these from the VBO...
131 tnl_get_per_vertex_materials(GLcontext
*ctx
)
133 GLbitfield mask
= 0x0;
135 TNLcontext
*tnl
= TNL_CONTEXT(ctx
);
136 struct vertex_buffer
*VB
= &tnl
->vb
;
139 for (i
= _TNL_FIRST_MAT
; i
<= _TNL_LAST_MAT
; i
++)
140 if (VB
->AttribPtr
[i
] && VB
->AttribPtr
[i
]->stride
)
141 mask
|= 1 << (i
- _TNL_FIRST_MAT
);
147 * Should fog be computed per-vertex?
150 tnl_get_per_vertex_fog(GLcontext
*ctx
)
153 TNLcontext
*tnl
= TNL_CONTEXT(ctx
);
154 return tnl
->_DoVertexFog
;
160 static GLboolean
check_active_shininess( GLcontext
*ctx
,
161 const struct state_key
*key
,
164 GLuint bit
= 1 << (MAT_ATTRIB_FRONT_SHININESS
+ side
);
166 if (key
->light_color_material_mask
& bit
)
169 if (key
->light_material_mask
& bit
)
172 if (ctx
->Light
.Material
.Attrib
[MAT_ATTRIB_FRONT_SHININESS
+ side
][0] != 0.0F
)
181 static struct state_key
*make_state_key( GLcontext
*ctx
)
183 const struct gl_fragment_program
*fp
;
184 struct state_key
*key
= CALLOC_STRUCT(state_key
);
187 fp
= ctx
->FragmentProgram
._Current
;
189 /* This now relies on texenvprogram.c being active:
193 key
->need_eye_coords
= ctx
->_NeedEyeCoords
;
195 key
->fragprog_inputs_read
= fp
->Base
.InputsRead
;
197 if (ctx
->RenderMode
== GL_FEEDBACK
) {
198 /* make sure the vertprog emits color and tex0 */
199 key
->fragprog_inputs_read
|= (FRAG_BIT_COL0
| FRAG_BIT_TEX0
);
202 key
->separate_specular
= (ctx
->Light
.Model
.ColorControl
==
203 GL_SEPARATE_SPECULAR_COLOR
);
205 if (ctx
->Light
.Enabled
) {
206 key
->light_global_enabled
= 1;
208 if (ctx
->Light
.Model
.LocalViewer
)
209 key
->light_local_viewer
= 1;
211 if (ctx
->Light
.Model
.TwoSide
)
212 key
->light_twoside
= 1;
214 if (ctx
->Light
.ColorMaterialEnabled
) {
215 key
->light_color_material
= 1;
216 key
->light_color_material_mask
= ctx
->Light
.ColorMaterialBitmask
;
219 key
->light_material_mask
= tnl_get_per_vertex_materials(ctx
);
221 for (i
= 0; i
< MAX_LIGHTS
; i
++) {
222 struct gl_light
*light
= &ctx
->Light
.Light
[i
];
224 if (light
->Enabled
) {
225 key
->unit
[i
].light_enabled
= 1;
227 if (light
->EyePosition
[3] == 0.0)
228 key
->unit
[i
].light_eyepos3_is_zero
= 1;
230 if (light
->SpotCutoff
== 180.0)
231 key
->unit
[i
].light_spotcutoff_is_180
= 1;
233 if (light
->ConstantAttenuation
!= 1.0 ||
234 light
->LinearAttenuation
!= 0.0 ||
235 light
->QuadraticAttenuation
!= 0.0)
236 key
->unit
[i
].light_attenuated
= 1;
240 if (check_active_shininess(ctx
, key
, 0)) {
241 key
->material_shininess_is_zero
= 0;
243 else if (key
->light_twoside
&&
244 check_active_shininess(ctx
, key
, 1)) {
245 key
->material_shininess_is_zero
= 0;
248 key
->material_shininess_is_zero
= 1;
252 if (ctx
->Transform
.Normalize
)
255 if (ctx
->Transform
.RescaleNormals
)
256 key
->rescale_normals
= 1;
258 key
->fog_mode
= translate_fog_mode(fp
->FogOption
);
260 if (ctx
->Fog
.FogCoordinateSource
== GL_FRAGMENT_DEPTH_EXT
)
261 key
->fog_source_is_depth
= 1;
263 key
->tnl_do_vertex_fog
= tnl_get_per_vertex_fog(ctx
);
265 if (ctx
->Point
._Attenuated
)
266 key
->point_attenuated
= 1;
268 #if FEATURE_point_size_array
269 if (ctx
->Array
.ArrayObj
->PointSize
.Enabled
)
270 key
->point_array
= 1;
273 if (ctx
->Texture
._TexGenEnabled
||
274 ctx
->Texture
._TexMatEnabled
||
275 ctx
->Texture
._EnabledUnits
)
276 key
->texture_enabled_global
= 1;
278 for (i
= 0; i
< MAX_TEXTURE_UNITS
; i
++) {
279 struct gl_texture_unit
*texUnit
= &ctx
->Texture
.Unit
[i
];
281 if (texUnit
->_ReallyEnabled
)
282 key
->unit
[i
].texunit_really_enabled
= 1;
284 if (ctx
->Texture
._TexMatEnabled
& ENABLE_TEXMAT(i
))
285 key
->unit
[i
].texmat_enabled
= 1;
287 if (texUnit
->TexGenEnabled
) {
288 key
->unit
[i
].texgen_enabled
= 1;
290 key
->unit
[i
].texgen_mode0
=
291 translate_texgen( texUnit
->TexGenEnabled
& (1<<0),
293 key
->unit
[i
].texgen_mode1
=
294 translate_texgen( texUnit
->TexGenEnabled
& (1<<1),
296 key
->unit
[i
].texgen_mode2
=
297 translate_texgen( texUnit
->TexGenEnabled
& (1<<2),
299 key
->unit
[i
].texgen_mode3
=
300 translate_texgen( texUnit
->TexGenEnabled
& (1<<3),
310 /* Very useful debugging tool - produces annotated listing of
311 * generated program with line/function references for each
312 * instruction back into this file:
316 /* Should be tunable by the driver - do we want to do matrix
317 * multiplications with DP4's or with MUL/MAD's? SSE works better
318 * with the latter, drivers may differ.
324 /* Use uregs to represent registers internally, translate to Mesa's
325 * expected formats on emit.
327 * NOTE: These are passed by value extensively in this file rather
328 * than as usual by pointer reference. If this disturbs you, try
329 * remembering they are just 32bits in size.
331 * GCC is smart enough to deal with these dword-sized structures in
332 * much the same way as if I had defined them as dwords and was using
333 * macros to access and set the fields. This is much nicer and easier
338 GLint idx
:8; /* relative addressing may be negative */
346 const struct state_key
*state
;
347 struct gl_vertex_program
*program
;
350 GLuint temp_reserved
;
352 struct ureg eye_position
;
353 struct ureg eye_position_z
;
354 struct ureg eye_position_normalized
;
355 struct ureg transformed_normal
;
356 struct ureg identity
;
359 GLuint color_materials
;
363 static const struct ureg undef
= {
381 static struct ureg
make_ureg(GLuint file
, GLint idx
)
387 reg
.swz
= SWIZZLE_NOOP
;
394 static struct ureg
negate( struct ureg reg
)
401 static struct ureg
swizzle( struct ureg reg
, int x
, int y
, int z
, int w
)
403 reg
.swz
= MAKE_SWIZZLE4(GET_SWZ(reg
.swz
, x
),
406 GET_SWZ(reg
.swz
, w
));
411 static struct ureg
swizzle1( struct ureg reg
, int x
)
413 return swizzle(reg
, x
, x
, x
, x
);
416 static struct ureg
get_temp( struct tnl_program
*p
)
418 int bit
= _mesa_ffs( ~p
->temp_in_use
);
420 _mesa_problem(NULL
, "%s: out of temporaries\n", __FILE__
);
424 if ((GLuint
) bit
> p
->program
->Base
.NumTemporaries
)
425 p
->program
->Base
.NumTemporaries
= bit
;
427 p
->temp_in_use
|= 1<<(bit
-1);
428 return make_ureg(PROGRAM_TEMPORARY
, bit
-1);
431 static struct ureg
reserve_temp( struct tnl_program
*p
)
433 struct ureg temp
= get_temp( p
);
434 p
->temp_reserved
|= 1<<temp
.idx
;
438 static void release_temp( struct tnl_program
*p
, struct ureg reg
)
440 if (reg
.file
== PROGRAM_TEMPORARY
) {
441 p
->temp_in_use
&= ~(1<<reg
.idx
);
442 p
->temp_in_use
|= p
->temp_reserved
; /* can't release reserved temps */
446 static void release_temps( struct tnl_program
*p
)
448 p
->temp_in_use
= p
->temp_reserved
;
454 * \param input one of VERT_ATTRIB_x tokens.
456 static struct ureg
register_input( struct tnl_program
*p
, GLuint input
)
458 p
->program
->Base
.InputsRead
|= (1<<input
);
459 return make_ureg(PROGRAM_INPUT
, input
);
463 * \param input one of VERT_RESULT_x tokens.
465 static struct ureg
register_output( struct tnl_program
*p
, GLuint output
)
467 p
->program
->Base
.OutputsWritten
|= (1<<output
);
468 return make_ureg(PROGRAM_OUTPUT
, output
);
471 static struct ureg
register_const4f( struct tnl_program
*p
,
484 idx
= _mesa_add_unnamed_constant( p
->program
->Base
.Parameters
, values
, 4,
486 ASSERT(swizzle
== SWIZZLE_NOOP
);
487 return make_ureg(PROGRAM_CONSTANT
, idx
);
490 #define register_const1f(p, s0) register_const4f(p, s0, 0, 0, 1)
491 #define register_scalar_const(p, s0) register_const4f(p, s0, s0, s0, s0)
492 #define register_const2f(p, s0, s1) register_const4f(p, s0, s1, 0, 1)
493 #define register_const3f(p, s0, s1, s2) register_const4f(p, s0, s1, s2, 1)
495 static GLboolean
is_undef( struct ureg reg
)
497 return reg
.file
== PROGRAM_UNDEFINED
;
500 static struct ureg
get_identity_param( struct tnl_program
*p
)
502 if (is_undef(p
->identity
))
503 p
->identity
= register_const4f(p
, 0,0,0,1);
508 static struct ureg
register_param5(struct tnl_program
*p
,
515 gl_state_index tokens
[STATE_LENGTH
];
522 idx
= _mesa_add_state_reference( p
->program
->Base
.Parameters
, tokens
);
523 return make_ureg(PROGRAM_STATE_VAR
, idx
);
527 #define register_param1(p,s0) register_param5(p,s0,0,0,0,0)
528 #define register_param2(p,s0,s1) register_param5(p,s0,s1,0,0,0)
529 #define register_param3(p,s0,s1,s2) register_param5(p,s0,s1,s2,0,0)
530 #define register_param4(p,s0,s1,s2,s3) register_param5(p,s0,s1,s2,s3,0)
533 static void register_matrix_param5( struct tnl_program
*p
,
534 GLint s0
, /* modelview, projection, etc */
535 GLint s1
, /* texture matrix number */
536 GLint s2
, /* first row */
537 GLint s3
, /* last row */
538 GLint s4
, /* inverse, transpose, etc */
539 struct ureg
*matrix
)
543 /* This is a bit sad as the support is there to pull the whole
544 * matrix out in one go:
546 for (i
= 0; i
<= s3
- s2
; i
++)
547 matrix
[i
] = register_param5( p
, s0
, s1
, i
, i
, s4
);
551 static void emit_arg( struct prog_src_register
*src
,
554 src
->File
= reg
.file
;
555 src
->Index
= reg
.idx
;
556 src
->Swizzle
= reg
.swz
;
557 src
->NegateBase
= reg
.negate
? NEGATE_XYZW
: 0;
563 static void emit_dst( struct prog_dst_register
*dst
,
564 struct ureg reg
, GLuint mask
)
566 dst
->File
= reg
.file
;
567 dst
->Index
= reg
.idx
;
568 /* allow zero as a shorthand for xyzw */
569 dst
->WriteMask
= mask
? mask
: WRITEMASK_XYZW
;
570 dst
->CondMask
= COND_TR
; /* always pass cond test */
571 dst
->CondSwizzle
= SWIZZLE_NOOP
;
576 static void debug_insn( struct prog_instruction
*inst
, const char *fn
,
580 static const char *last_fn
;
584 _mesa_printf("%s:\n", fn
);
587 _mesa_printf("%d:\t", line
);
588 _mesa_print_instruction(inst
);
593 static void emit_op3fn(struct tnl_program
*p
,
603 GLuint nr
= p
->program
->Base
.NumInstructions
++;
604 struct prog_instruction
*inst
= &p
->program
->Base
.Instructions
[nr
];
606 if (p
->program
->Base
.NumInstructions
> MAX_INSN
) {
607 _mesa_problem(0, "Out of instructions in emit_op3fn\n");
611 inst
->Opcode
= (enum prog_opcode
) op
;
615 emit_arg( &inst
->SrcReg
[0], src0
);
616 emit_arg( &inst
->SrcReg
[1], src1
);
617 emit_arg( &inst
->SrcReg
[2], src2
);
619 emit_dst( &inst
->DstReg
, dest
, mask
);
621 debug_insn(inst
, fn
, line
);
625 #define emit_op3(p, op, dst, mask, src0, src1, src2) \
626 emit_op3fn(p, op, dst, mask, src0, src1, src2, __FUNCTION__, __LINE__)
628 #define emit_op2(p, op, dst, mask, src0, src1) \
629 emit_op3fn(p, op, dst, mask, src0, src1, undef, __FUNCTION__, __LINE__)
631 #define emit_op1(p, op, dst, mask, src0) \
632 emit_op3fn(p, op, dst, mask, src0, undef, undef, __FUNCTION__, __LINE__)
635 static struct ureg
make_temp( struct tnl_program
*p
, struct ureg reg
)
637 if (reg
.file
== PROGRAM_TEMPORARY
&&
638 !(p
->temp_reserved
& (1<<reg
.idx
)))
641 struct ureg temp
= get_temp(p
);
642 emit_op1(p
, OPCODE_MOV
, temp
, 0, reg
);
648 /* Currently no tracking performed of input/output/register size or
649 * active elements. Could be used to reduce these operations, as
650 * could the matrix type.
652 static void emit_matrix_transform_vec4( struct tnl_program
*p
,
654 const struct ureg
*mat
,
657 emit_op2(p
, OPCODE_DP4
, dest
, WRITEMASK_X
, src
, mat
[0]);
658 emit_op2(p
, OPCODE_DP4
, dest
, WRITEMASK_Y
, src
, mat
[1]);
659 emit_op2(p
, OPCODE_DP4
, dest
, WRITEMASK_Z
, src
, mat
[2]);
660 emit_op2(p
, OPCODE_DP4
, dest
, WRITEMASK_W
, src
, mat
[3]);
663 /* This version is much easier to implement if writemasks are not
664 * supported natively on the target or (like SSE), the target doesn't
665 * have a clean/obvious dotproduct implementation.
667 static void emit_transpose_matrix_transform_vec4( struct tnl_program
*p
,
669 const struct ureg
*mat
,
674 if (dest
.file
!= PROGRAM_TEMPORARY
)
679 emit_op2(p
, OPCODE_MUL
, tmp
, 0, swizzle1(src
,X
), mat
[0]);
680 emit_op3(p
, OPCODE_MAD
, tmp
, 0, swizzle1(src
,Y
), mat
[1], tmp
);
681 emit_op3(p
, OPCODE_MAD
, tmp
, 0, swizzle1(src
,Z
), mat
[2], tmp
);
682 emit_op3(p
, OPCODE_MAD
, dest
, 0, swizzle1(src
,W
), mat
[3], tmp
);
684 if (dest
.file
!= PROGRAM_TEMPORARY
)
685 release_temp(p
, tmp
);
688 static void emit_matrix_transform_vec3( struct tnl_program
*p
,
690 const struct ureg
*mat
,
693 emit_op2(p
, OPCODE_DP3
, dest
, WRITEMASK_X
, src
, mat
[0]);
694 emit_op2(p
, OPCODE_DP3
, dest
, WRITEMASK_Y
, src
, mat
[1]);
695 emit_op2(p
, OPCODE_DP3
, dest
, WRITEMASK_Z
, src
, mat
[2]);
699 static void emit_normalize_vec3( struct tnl_program
*p
,
703 struct ureg tmp
= get_temp(p
);
704 emit_op2(p
, OPCODE_DP3
, tmp
, WRITEMASK_X
, src
, src
);
705 emit_op1(p
, OPCODE_RSQ
, tmp
, WRITEMASK_X
, tmp
);
706 emit_op2(p
, OPCODE_MUL
, dest
, 0, src
, swizzle1(tmp
, X
));
707 release_temp(p
, tmp
);
710 static void emit_passthrough( struct tnl_program
*p
,
714 struct ureg out
= register_output(p
, output
);
715 emit_op1(p
, OPCODE_MOV
, out
, 0, register_input(p
, input
));
718 static struct ureg
get_eye_position( struct tnl_program
*p
)
720 if (is_undef(p
->eye_position
)) {
721 struct ureg pos
= register_input( p
, VERT_ATTRIB_POS
);
722 struct ureg modelview
[4];
724 p
->eye_position
= reserve_temp(p
);
727 register_matrix_param5( p
, STATE_MODELVIEW_MATRIX
, 0, 0, 3,
730 emit_matrix_transform_vec4(p
, p
->eye_position
, modelview
, pos
);
733 register_matrix_param5( p
, STATE_MODELVIEW_MATRIX
, 0, 0, 3,
734 STATE_MATRIX_TRANSPOSE
, modelview
);
736 emit_transpose_matrix_transform_vec4(p
, p
->eye_position
, modelview
, pos
);
740 return p
->eye_position
;
744 static struct ureg
get_eye_position_z( struct tnl_program
*p
)
746 if (!is_undef(p
->eye_position
))
747 return swizzle1(p
->eye_position
, Z
);
749 if (is_undef(p
->eye_position_z
)) {
750 struct ureg pos
= register_input( p
, VERT_ATTRIB_POS
);
751 struct ureg modelview
[4];
753 p
->eye_position_z
= reserve_temp(p
);
755 register_matrix_param5( p
, STATE_MODELVIEW_MATRIX
, 0, 0, 3,
758 emit_op2(p
, OPCODE_DP4
, p
->eye_position_z
, 0, pos
, modelview
[2]);
761 return p
->eye_position_z
;
766 static struct ureg
get_eye_position_normalized( struct tnl_program
*p
)
768 if (is_undef(p
->eye_position_normalized
)) {
769 struct ureg eye
= get_eye_position(p
);
770 p
->eye_position_normalized
= reserve_temp(p
);
771 emit_normalize_vec3(p
, p
->eye_position_normalized
, eye
);
774 return p
->eye_position_normalized
;
778 static struct ureg
get_transformed_normal( struct tnl_program
*p
)
780 if (is_undef(p
->transformed_normal
) &&
781 !p
->state
->need_eye_coords
&&
782 !p
->state
->normalize
&&
783 !(p
->state
->need_eye_coords
== p
->state
->rescale_normals
))
785 p
->transformed_normal
= register_input(p
, VERT_ATTRIB_NORMAL
);
787 else if (is_undef(p
->transformed_normal
))
789 struct ureg normal
= register_input(p
, VERT_ATTRIB_NORMAL
);
790 struct ureg mvinv
[3];
791 struct ureg transformed_normal
= reserve_temp(p
);
793 if (p
->state
->need_eye_coords
) {
794 register_matrix_param5( p
, STATE_MODELVIEW_MATRIX
, 0, 0, 2,
795 STATE_MATRIX_INVTRANS
, mvinv
);
797 /* Transform to eye space:
799 emit_matrix_transform_vec3( p
, transformed_normal
, mvinv
, normal
);
800 normal
= transformed_normal
;
803 /* Normalize/Rescale:
805 if (p
->state
->normalize
) {
806 emit_normalize_vec3( p
, transformed_normal
, normal
);
807 normal
= transformed_normal
;
809 else if (p
->state
->need_eye_coords
== p
->state
->rescale_normals
) {
810 /* This is already adjusted for eye/non-eye rendering:
812 struct ureg rescale
= register_param2(p
, STATE_INTERNAL
,
815 emit_op2( p
, OPCODE_MUL
, transformed_normal
, 0, normal
, rescale
);
816 normal
= transformed_normal
;
819 assert(normal
.file
== PROGRAM_TEMPORARY
);
820 p
->transformed_normal
= normal
;
823 return p
->transformed_normal
;
828 static void build_hpos( struct tnl_program
*p
)
830 struct ureg pos
= register_input( p
, VERT_ATTRIB_POS
);
831 struct ureg hpos
= register_output( p
, VERT_RESULT_HPOS
);
835 register_matrix_param5( p
, STATE_MVP_MATRIX
, 0, 0, 3,
837 emit_matrix_transform_vec4( p
, hpos
, mvp
, pos
);
840 register_matrix_param5( p
, STATE_MVP_MATRIX
, 0, 0, 3,
841 STATE_MATRIX_TRANSPOSE
, mvp
);
842 emit_transpose_matrix_transform_vec4( p
, hpos
, mvp
, pos
);
847 static GLuint
material_attrib( GLuint side
, GLuint property
)
849 return ((property
- STATE_AMBIENT
) * 2 +
853 /* Get a bitmask of which material values vary on a per-vertex basis.
855 static void set_material_flags( struct tnl_program
*p
)
857 p
->color_materials
= 0;
860 if (p
->state
->light_color_material
) {
862 p
->color_materials
= p
->state
->light_color_material_mask
;
865 p
->materials
|= p
->state
->light_material_mask
;
869 /* XXX temporary!!! */
870 #define _TNL_ATTRIB_MAT_FRONT_AMBIENT 32
872 static struct ureg
get_material( struct tnl_program
*p
, GLuint side
,
875 GLuint attrib
= material_attrib(side
, property
);
877 if (p
->color_materials
& (1<<attrib
))
878 return register_input(p
, VERT_ATTRIB_COLOR0
);
879 else if (p
->materials
& (1<<attrib
))
880 return register_input( p
, attrib
+ _TNL_ATTRIB_MAT_FRONT_AMBIENT
);
882 return register_param3( p
, STATE_MATERIAL
, side
, property
);
885 #define SCENE_COLOR_BITS(side) (( MAT_BIT_FRONT_EMISSION | \
886 MAT_BIT_FRONT_AMBIENT | \
887 MAT_BIT_FRONT_DIFFUSE) << (side))
889 /* Either return a precalculated constant value or emit code to
890 * calculate these values dynamically in the case where material calls
891 * are present between begin/end pairs.
893 * Probably want to shift this to the program compilation phase - if
894 * we always emitted the calculation here, a smart compiler could
895 * detect that it was constant (given a certain set of inputs), and
896 * lift it out of the main loop. That way the programs created here
897 * would be independent of the vertex_buffer details.
899 static struct ureg
get_scenecolor( struct tnl_program
*p
, GLuint side
)
901 if (p
->materials
& SCENE_COLOR_BITS(side
)) {
902 struct ureg lm_ambient
= register_param1(p
, STATE_LIGHTMODEL_AMBIENT
);
903 struct ureg material_emission
= get_material(p
, side
, STATE_EMISSION
);
904 struct ureg material_ambient
= get_material(p
, side
, STATE_AMBIENT
);
905 struct ureg material_diffuse
= get_material(p
, side
, STATE_DIFFUSE
);
906 struct ureg tmp
= make_temp(p
, material_diffuse
);
907 emit_op3(p
, OPCODE_MAD
, tmp
, WRITEMASK_XYZ
, lm_ambient
,
908 material_ambient
, material_emission
);
912 return register_param2( p
, STATE_LIGHTMODEL_SCENECOLOR
, side
);
916 static struct ureg
get_lightprod( struct tnl_program
*p
, GLuint light
,
917 GLuint side
, GLuint property
)
919 GLuint attrib
= material_attrib(side
, property
);
920 if (p
->materials
& (1<<attrib
)) {
921 struct ureg light_value
=
922 register_param3(p
, STATE_LIGHT
, light
, property
);
923 struct ureg material_value
= get_material(p
, side
, property
);
924 struct ureg tmp
= get_temp(p
);
925 emit_op2(p
, OPCODE_MUL
, tmp
, 0, light_value
, material_value
);
929 return register_param4(p
, STATE_LIGHTPROD
, light
, side
, property
);
932 static struct ureg
calculate_light_attenuation( struct tnl_program
*p
,
937 struct ureg attenuation
= register_param3(p
, STATE_LIGHT
, i
,
939 struct ureg att
= get_temp(p
);
941 /* Calculate spot attenuation:
943 if (!p
->state
->unit
[i
].light_spotcutoff_is_180
) {
944 struct ureg spot_dir_norm
= register_param3(p
, STATE_INTERNAL
,
945 STATE_LIGHT_SPOT_DIR_NORMALIZED
, i
);
946 struct ureg spot
= get_temp(p
);
947 struct ureg slt
= get_temp(p
);
949 emit_op2(p
, OPCODE_DP3
, spot
, 0, negate(VPpli
), spot_dir_norm
);
950 emit_op2(p
, OPCODE_SLT
, slt
, 0, swizzle1(spot_dir_norm
,W
), spot
);
951 emit_op2(p
, OPCODE_POW
, spot
, 0, spot
, swizzle1(attenuation
, W
));
952 emit_op2(p
, OPCODE_MUL
, att
, 0, slt
, spot
);
954 release_temp(p
, spot
);
955 release_temp(p
, slt
);
958 /* Calculate distance attenuation:
960 if (p
->state
->unit
[i
].light_attenuated
) {
963 emit_op1(p
, OPCODE_RCP
, dist
, WRITEMASK_YZ
, dist
);
965 emit_op2(p
, OPCODE_MUL
, dist
, WRITEMASK_XZ
, dist
, swizzle1(dist
,Y
));
967 emit_op2(p
, OPCODE_DP3
, dist
, 0, attenuation
, dist
);
969 if (!p
->state
->unit
[i
].light_spotcutoff_is_180
) {
971 emit_op1(p
, OPCODE_RCP
, dist
, 0, dist
);
972 /* spot-atten * dist-atten */
973 emit_op2(p
, OPCODE_MUL
, att
, 0, dist
, att
);
976 emit_op1(p
, OPCODE_RCP
, att
, 0, dist
);
986 * lit.y = MAX(0, dots.x)
987 * lit.z = SLT(0, dots.x)
989 static void emit_degenerate_lit( struct tnl_program
*p
,
993 struct ureg id
= get_identity_param(p
); /* id = {0,0,0,1} */
995 /* Note that lit.x & lit.w will not be examined. Note also that
996 * dots.xyzw == dots.xxxx.
999 /* MAX lit, id, dots;
1001 emit_op2(p
, OPCODE_MAX
, lit
, WRITEMASK_XYZW
, id
, dots
);
1003 /* result[2] = (in > 0 ? 1 : 0)
1004 * SLT lit.z, id.z, dots; # lit.z = (0 < dots.z) ? 1 : 0
1006 emit_op2(p
, OPCODE_SLT
, lit
, WRITEMASK_Z
, swizzle1(id
,Z
), dots
);
1010 /* Need to add some addtional parameters to allow lighting in object
1011 * space - STATE_SPOT_DIRECTION and STATE_HALF_VECTOR implicitly assume eye
1014 static void build_lighting( struct tnl_program
*p
)
1016 const GLboolean twoside
= p
->state
->light_twoside
;
1017 const GLboolean separate
= p
->state
->separate_specular
;
1018 GLuint nr_lights
= 0, count
= 0;
1019 struct ureg normal
= get_transformed_normal(p
);
1020 struct ureg lit
= get_temp(p
);
1021 struct ureg dots
= get_temp(p
);
1022 struct ureg _col0
= undef
, _col1
= undef
;
1023 struct ureg _bfc0
= undef
, _bfc1
= undef
;
1028 * dot.x = dot(normal, VPpli)
1029 * dot.y = dot(normal, halfAngle)
1030 * dot.z = back.shininess
1031 * dot.w = front.shininess
1034 for (i
= 0; i
< MAX_LIGHTS
; i
++)
1035 if (p
->state
->unit
[i
].light_enabled
)
1038 set_material_flags(p
);
1041 if (!p
->state
->material_shininess_is_zero
) {
1042 struct ureg shininess
= get_material(p
, 0, STATE_SHININESS
);
1043 emit_op1(p
, OPCODE_MOV
, dots
, WRITEMASK_W
, swizzle1(shininess
,X
));
1044 release_temp(p
, shininess
);
1047 _col0
= make_temp(p
, get_scenecolor(p
, 0));
1049 _col1
= make_temp(p
, get_identity_param(p
));
1056 if (!p
->state
->material_shininess_is_zero
) {
1057 struct ureg shininess
= get_material(p
, 1, STATE_SHININESS
);
1058 emit_op1(p
, OPCODE_MOV
, dots
, WRITEMASK_Z
,
1059 negate(swizzle1(shininess
,X
)));
1060 release_temp(p
, shininess
);
1063 _bfc0
= make_temp(p
, get_scenecolor(p
, 1));
1065 _bfc1
= make_temp(p
, get_identity_param(p
));
1070 /* If no lights, still need to emit the scenecolor.
1073 struct ureg res0
= register_output( p
, VERT_RESULT_COL0
);
1074 emit_op1(p
, OPCODE_MOV
, res0
, 0, _col0
);
1078 struct ureg res1
= register_output( p
, VERT_RESULT_COL1
);
1079 emit_op1(p
, OPCODE_MOV
, res1
, 0, _col1
);
1083 struct ureg res0
= register_output( p
, VERT_RESULT_BFC0
);
1084 emit_op1(p
, OPCODE_MOV
, res0
, 0, _bfc0
);
1087 if (twoside
&& separate
) {
1088 struct ureg res1
= register_output( p
, VERT_RESULT_BFC1
);
1089 emit_op1(p
, OPCODE_MOV
, res1
, 0, _bfc1
);
1092 if (nr_lights
== 0) {
1097 for (i
= 0; i
< MAX_LIGHTS
; i
++) {
1098 if (p
->state
->unit
[i
].light_enabled
) {
1099 struct ureg half
= undef
;
1100 struct ureg att
= undef
, VPpli
= undef
;
1104 if (p
->state
->unit
[i
].light_eyepos3_is_zero
) {
1105 /* Can used precomputed constants in this case.
1106 * Attenuation never applies to infinite lights.
1108 VPpli
= register_param3(p
, STATE_INTERNAL
,
1109 STATE_LIGHT_POSITION_NORMALIZED
, i
);
1111 if (!p
->state
->material_shininess_is_zero
) {
1112 if (p
->state
->light_local_viewer
) {
1113 struct ureg eye_hat
= get_eye_position_normalized(p
);
1115 emit_op2(p
, OPCODE_SUB
, half
, 0, VPpli
, eye_hat
);
1116 emit_normalize_vec3(p
, half
, half
);
1118 half
= register_param3(p
, STATE_INTERNAL
,
1119 STATE_LIGHT_HALF_VECTOR
, i
);
1124 struct ureg Ppli
= register_param3(p
, STATE_INTERNAL
,
1125 STATE_LIGHT_POSITION
, i
);
1126 struct ureg V
= get_eye_position(p
);
1127 struct ureg dist
= get_temp(p
);
1129 VPpli
= get_temp(p
);
1131 /* Calculate VPpli vector
1133 emit_op2(p
, OPCODE_SUB
, VPpli
, 0, Ppli
, V
);
1135 /* Normalize VPpli. The dist value also used in
1136 * attenuation below.
1138 emit_op2(p
, OPCODE_DP3
, dist
, 0, VPpli
, VPpli
);
1139 emit_op1(p
, OPCODE_RSQ
, dist
, 0, dist
);
1140 emit_op2(p
, OPCODE_MUL
, VPpli
, 0, VPpli
, dist
);
1142 /* Calculate attenuation:
1144 if (!p
->state
->unit
[i
].light_spotcutoff_is_180
||
1145 p
->state
->unit
[i
].light_attenuated
) {
1146 att
= calculate_light_attenuation(p
, i
, VPpli
, dist
);
1149 /* Calculate viewer direction, or use infinite viewer:
1151 if (!p
->state
->material_shininess_is_zero
) {
1154 if (p
->state
->light_local_viewer
) {
1155 struct ureg eye_hat
= get_eye_position_normalized(p
);
1156 emit_op2(p
, OPCODE_SUB
, half
, 0, VPpli
, eye_hat
);
1159 struct ureg z_dir
= swizzle(get_identity_param(p
),X
,Y
,W
,Z
);
1160 emit_op2(p
, OPCODE_ADD
, half
, 0, VPpli
, z_dir
);
1163 emit_normalize_vec3(p
, half
, half
);
1166 release_temp(p
, dist
);
1169 /* Calculate dot products:
1171 if (p
->state
->material_shininess_is_zero
) {
1172 emit_op2(p
, OPCODE_DP3
, dots
, 0, normal
, VPpli
);
1175 emit_op2(p
, OPCODE_DP3
, dots
, WRITEMASK_X
, normal
, VPpli
);
1176 emit_op2(p
, OPCODE_DP3
, dots
, WRITEMASK_Y
, normal
, half
);
1179 /* Front face lighting:
1182 struct ureg ambient
= get_lightprod(p
, i
, 0, STATE_AMBIENT
);
1183 struct ureg diffuse
= get_lightprod(p
, i
, 0, STATE_DIFFUSE
);
1184 struct ureg specular
= get_lightprod(p
, i
, 0, STATE_SPECULAR
);
1185 struct ureg res0
, res1
;
1186 GLuint mask0
, mask1
;
1189 if (count
== nr_lights
) {
1191 mask0
= WRITEMASK_XYZ
;
1192 mask1
= WRITEMASK_XYZ
;
1193 res0
= register_output( p
, VERT_RESULT_COL0
);
1194 res1
= register_output( p
, VERT_RESULT_COL1
);
1198 mask1
= WRITEMASK_XYZ
;
1200 res1
= register_output( p
, VERT_RESULT_COL0
);
1210 if (!is_undef(att
)) {
1211 /* light is attenuated by distance */
1212 emit_op1(p
, OPCODE_LIT
, lit
, 0, dots
);
1213 emit_op2(p
, OPCODE_MUL
, lit
, 0, lit
, att
);
1214 emit_op3(p
, OPCODE_MAD
, _col0
, 0, swizzle1(lit
,X
), ambient
, _col0
);
1216 else if (!p
->state
->material_shininess_is_zero
) {
1217 /* there's a non-zero specular term */
1218 emit_op1(p
, OPCODE_LIT
, lit
, 0, dots
);
1219 emit_op2(p
, OPCODE_ADD
, _col0
, 0, ambient
, _col0
);
1222 /* no attenutation, no specular */
1223 emit_degenerate_lit(p
, lit
, dots
);
1224 emit_op2(p
, OPCODE_ADD
, _col0
, 0, ambient
, _col0
);
1227 emit_op3(p
, OPCODE_MAD
, res0
, mask0
, swizzle1(lit
,Y
), diffuse
, _col0
);
1228 emit_op3(p
, OPCODE_MAD
, res1
, mask1
, swizzle1(lit
,Z
), specular
, _col1
);
1230 release_temp(p
, ambient
);
1231 release_temp(p
, diffuse
);
1232 release_temp(p
, specular
);
1235 /* Back face lighting:
1238 struct ureg ambient
= get_lightprod(p
, i
, 1, STATE_AMBIENT
);
1239 struct ureg diffuse
= get_lightprod(p
, i
, 1, STATE_DIFFUSE
);
1240 struct ureg specular
= get_lightprod(p
, i
, 1, STATE_SPECULAR
);
1241 struct ureg res0
, res1
;
1242 GLuint mask0
, mask1
;
1244 if (count
== nr_lights
) {
1246 mask0
= WRITEMASK_XYZ
;
1247 mask1
= WRITEMASK_XYZ
;
1248 res0
= register_output( p
, VERT_RESULT_BFC0
);
1249 res1
= register_output( p
, VERT_RESULT_BFC1
);
1253 mask1
= WRITEMASK_XYZ
;
1255 res1
= register_output( p
, VERT_RESULT_BFC0
);
1264 dots
= negate(swizzle(dots
,X
,Y
,W
,Z
));
1266 if (!is_undef(att
)) {
1267 emit_op1(p
, OPCODE_LIT
, lit
, 0, dots
);
1268 emit_op2(p
, OPCODE_MUL
, lit
, 0, lit
, att
);
1269 emit_op3(p
, OPCODE_MAD
, _bfc0
, 0, swizzle1(lit
,X
), ambient
, _bfc0
);
1271 else if (!p
->state
->material_shininess_is_zero
) {
1272 emit_op1(p
, OPCODE_LIT
, lit
, 0, dots
);
1273 emit_op2(p
, OPCODE_ADD
, _col0
, 0, ambient
, _col0
);
1276 emit_degenerate_lit(p
, lit
, dots
);
1277 emit_op2(p
, OPCODE_ADD
, _col0
, 0, ambient
, _col0
);
1280 emit_op2(p
, OPCODE_ADD
, _bfc0
, 0, ambient
, _bfc0
);
1281 emit_op3(p
, OPCODE_MAD
, res0
, mask0
, swizzle1(lit
,Y
), diffuse
, _bfc0
);
1282 emit_op3(p
, OPCODE_MAD
, res1
, mask1
, swizzle1(lit
,Z
), specular
, _bfc1
);
1284 release_temp(p
, ambient
);
1285 release_temp(p
, diffuse
);
1286 release_temp(p
, specular
);
1289 release_temp(p
, half
);
1290 release_temp(p
, VPpli
);
1291 release_temp(p
, att
);
1299 static void build_fog( struct tnl_program
*p
)
1301 struct ureg fog
= register_output(p
, VERT_RESULT_FOGC
);
1304 if (p
->state
->fog_source_is_depth
) {
1305 input
= get_eye_position_z(p
);
1308 input
= swizzle1(register_input(p
, VERT_ATTRIB_FOG
), X
);
1311 if (p
->state
->fog_mode
&& p
->state
->tnl_do_vertex_fog
) {
1312 struct ureg params
= register_param2(p
, STATE_INTERNAL
,
1313 STATE_FOG_PARAMS_OPTIMIZED
);
1314 struct ureg tmp
= get_temp(p
);
1315 GLboolean useabs
= (p
->state
->fog_mode
!= FOG_EXP2
);
1318 emit_op1(p
, OPCODE_ABS
, tmp
, 0, input
);
1321 switch (p
->state
->fog_mode
) {
1323 struct ureg id
= get_identity_param(p
);
1324 emit_op3(p
, OPCODE_MAD
, tmp
, 0, useabs
? tmp
: input
,
1325 swizzle1(params
,X
), swizzle1(params
,Y
));
1326 emit_op2(p
, OPCODE_MAX
, tmp
, 0, tmp
, swizzle1(id
,X
)); /* saturate */
1327 emit_op2(p
, OPCODE_MIN
, fog
, WRITEMASK_X
, tmp
, swizzle1(id
,W
));
1331 emit_op2(p
, OPCODE_MUL
, tmp
, 0, useabs
? tmp
: input
,
1332 swizzle1(params
,Z
));
1333 emit_op1(p
, OPCODE_EX2
, fog
, WRITEMASK_X
, negate(tmp
));
1336 emit_op2(p
, OPCODE_MUL
, tmp
, 0, input
, swizzle1(params
,W
));
1337 emit_op2(p
, OPCODE_MUL
, tmp
, 0, tmp
, tmp
);
1338 emit_op1(p
, OPCODE_EX2
, fog
, WRITEMASK_X
, negate(tmp
));
1342 release_temp(p
, tmp
);
1345 /* results = incoming fog coords (compute fog per-fragment later)
1347 * KW: Is it really necessary to do anything in this case?
1348 * BP: Yes, we always need to compute the absolute value, unless
1349 * we want to push that down into the fragment program...
1351 GLboolean useabs
= GL_TRUE
;
1352 emit_op1(p
, useabs
? OPCODE_ABS
: OPCODE_MOV
, fog
, WRITEMASK_X
, input
);
1356 static void build_reflect_texgen( struct tnl_program
*p
,
1360 struct ureg normal
= get_transformed_normal(p
);
1361 struct ureg eye_hat
= get_eye_position_normalized(p
);
1362 struct ureg tmp
= get_temp(p
);
1365 emit_op2(p
, OPCODE_DP3
, tmp
, 0, normal
, eye_hat
);
1367 emit_op2(p
, OPCODE_ADD
, tmp
, 0, tmp
, tmp
);
1369 emit_op3(p
, OPCODE_MAD
, dest
, writemask
, negate(tmp
), normal
, eye_hat
);
1371 release_temp(p
, tmp
);
1374 static void build_sphere_texgen( struct tnl_program
*p
,
1378 struct ureg normal
= get_transformed_normal(p
);
1379 struct ureg eye_hat
= get_eye_position_normalized(p
);
1380 struct ureg tmp
= get_temp(p
);
1381 struct ureg half
= register_scalar_const(p
, .5);
1382 struct ureg r
= get_temp(p
);
1383 struct ureg inv_m
= get_temp(p
);
1384 struct ureg id
= get_identity_param(p
);
1386 /* Could share the above calculations, but it would be
1387 * a fairly odd state for someone to set (both sphere and
1388 * reflection active for different texture coordinate
1389 * components. Of course - if two texture units enable
1390 * reflect and/or sphere, things start to tilt in favour
1391 * of seperating this out:
1395 emit_op2(p
, OPCODE_DP3
, tmp
, 0, normal
, eye_hat
);
1397 emit_op2(p
, OPCODE_ADD
, tmp
, 0, tmp
, tmp
);
1399 emit_op3(p
, OPCODE_MAD
, r
, 0, negate(tmp
), normal
, eye_hat
);
1401 emit_op2(p
, OPCODE_ADD
, tmp
, 0, r
, swizzle(id
,X
,Y
,W
,Z
));
1402 /* rx^2 + ry^2 + (rz+1)^2 */
1403 emit_op2(p
, OPCODE_DP3
, tmp
, 0, tmp
, tmp
);
1405 emit_op1(p
, OPCODE_RSQ
, tmp
, 0, tmp
);
1407 emit_op2(p
, OPCODE_MUL
, inv_m
, 0, tmp
, half
);
1409 emit_op3(p
, OPCODE_MAD
, dest
, writemask
, r
, inv_m
, half
);
1411 release_temp(p
, tmp
);
1413 release_temp(p
, inv_m
);
1417 static void build_texture_transform( struct tnl_program
*p
)
1421 for (i
= 0; i
< MAX_TEXTURE_UNITS
; i
++) {
1423 if (!(p
->state
->fragprog_inputs_read
& FRAG_BIT_TEX(i
)))
1426 if (p
->state
->unit
[i
].texgen_enabled
||
1427 p
->state
->unit
[i
].texmat_enabled
) {
1429 GLuint texmat_enabled
= p
->state
->unit
[i
].texmat_enabled
;
1430 struct ureg out
= register_output(p
, VERT_RESULT_TEX0
+ i
);
1431 struct ureg out_texgen
= undef
;
1433 if (p
->state
->unit
[i
].texgen_enabled
) {
1434 GLuint copy_mask
= 0;
1435 GLuint sphere_mask
= 0;
1436 GLuint reflect_mask
= 0;
1437 GLuint normal_mask
= 0;
1441 out_texgen
= get_temp(p
);
1445 modes
[0] = p
->state
->unit
[i
].texgen_mode0
;
1446 modes
[1] = p
->state
->unit
[i
].texgen_mode1
;
1447 modes
[2] = p
->state
->unit
[i
].texgen_mode2
;
1448 modes
[3] = p
->state
->unit
[i
].texgen_mode3
;
1450 for (j
= 0; j
< 4; j
++) {
1452 case TXG_OBJ_LINEAR
: {
1453 struct ureg obj
= register_input(p
, VERT_ATTRIB_POS
);
1455 register_param3(p
, STATE_TEXGEN
, i
,
1456 STATE_TEXGEN_OBJECT_S
+ j
);
1458 emit_op2(p
, OPCODE_DP4
, out_texgen
, WRITEMASK_X
<< j
,
1462 case TXG_EYE_LINEAR
: {
1463 struct ureg eye
= get_eye_position(p
);
1465 register_param3(p
, STATE_TEXGEN
, i
,
1466 STATE_TEXGEN_EYE_S
+ j
);
1468 emit_op2(p
, OPCODE_DP4
, out_texgen
, WRITEMASK_X
<< j
,
1472 case TXG_SPHERE_MAP
:
1473 sphere_mask
|= WRITEMASK_X
<< j
;
1475 case TXG_REFLECTION_MAP
:
1476 reflect_mask
|= WRITEMASK_X
<< j
;
1478 case TXG_NORMAL_MAP
:
1479 normal_mask
|= WRITEMASK_X
<< j
;
1482 copy_mask
|= WRITEMASK_X
<< j
;
1489 build_sphere_texgen(p
, out_texgen
, sphere_mask
);
1493 build_reflect_texgen(p
, out_texgen
, reflect_mask
);
1497 struct ureg normal
= get_transformed_normal(p
);
1498 emit_op1(p
, OPCODE_MOV
, out_texgen
, normal_mask
, normal
);
1502 struct ureg in
= register_input(p
, VERT_ATTRIB_TEX0
+i
);
1503 emit_op1(p
, OPCODE_MOV
, out_texgen
, copy_mask
, in
);
1507 if (texmat_enabled
) {
1508 struct ureg texmat
[4];
1509 struct ureg in
= (!is_undef(out_texgen
) ?
1511 register_input(p
, VERT_ATTRIB_TEX0
+i
));
1513 register_matrix_param5( p
, STATE_TEXTURE_MATRIX
, i
, 0, 3,
1515 emit_matrix_transform_vec4( p
, out
, texmat
, in
);
1518 register_matrix_param5( p
, STATE_TEXTURE_MATRIX
, i
, 0, 3,
1519 STATE_MATRIX_TRANSPOSE
, texmat
);
1520 emit_transpose_matrix_transform_vec4( p
, out
, texmat
, in
);
1527 emit_passthrough(p
, VERT_ATTRIB_TEX0
+i
, VERT_RESULT_TEX0
+i
);
1534 * Point size attenuation computation.
1536 static void build_atten_pointsize( struct tnl_program
*p
)
1538 struct ureg eye
= get_eye_position_z(p
);
1539 struct ureg state_size
= register_param1(p
, STATE_POINT_SIZE
);
1540 struct ureg state_attenuation
= register_param1(p
, STATE_POINT_ATTENUATION
);
1541 struct ureg out
= register_output(p
, VERT_RESULT_PSIZ
);
1542 struct ureg ut
= get_temp(p
);
1545 emit_op1(p
, OPCODE_ABS
, ut
, WRITEMASK_Y
, swizzle1(eye
, Z
));
1546 /* p1 + dist * (p2 + dist * p3); */
1547 emit_op3(p
, OPCODE_MAD
, ut
, WRITEMASK_X
, swizzle1(ut
, Y
),
1548 swizzle1(state_attenuation
, Z
), swizzle1(state_attenuation
, Y
));
1549 emit_op3(p
, OPCODE_MAD
, ut
, WRITEMASK_X
, swizzle1(ut
, Y
),
1550 ut
, swizzle1(state_attenuation
, X
));
1552 /* 1 / sqrt(factor) */
1553 emit_op1(p
, OPCODE_RSQ
, ut
, WRITEMASK_X
, ut
);
1556 /* out = pointSize / sqrt(factor) */
1557 emit_op2(p
, OPCODE_MUL
, out
, WRITEMASK_X
, ut
, state_size
);
1559 /* this is a good place to clamp the point size since there's likely
1560 * no hardware registers to clamp point size at rasterization time.
1562 emit_op2(p
, OPCODE_MUL
, ut
, WRITEMASK_X
, ut
, state_size
);
1563 emit_op2(p
, OPCODE_MAX
, ut
, WRITEMASK_X
, ut
, swizzle1(state_size
, Y
));
1564 emit_op2(p
, OPCODE_MIN
, out
, WRITEMASK_X
, ut
, swizzle1(state_size
, Z
));
1567 release_temp(p
, ut
);
1571 * Emit constant point size.
1573 static void build_constant_pointsize( struct tnl_program
*p
)
1575 struct ureg state_size
= register_param1(p
, STATE_POINT_SIZE
);
1576 struct ureg out
= register_output(p
, VERT_RESULT_PSIZ
);
1577 emit_op1(p
, OPCODE_MOV
, out
, WRITEMASK_X
, state_size
);
1581 * Pass-though per-vertex point size, from user's point size array.
1583 static void build_array_pointsize( struct tnl_program
*p
)
1585 struct ureg in
= register_input(p
, VERT_ATTRIB_POINT_SIZE
);
1586 struct ureg out
= register_output(p
, VERT_RESULT_PSIZ
);
1587 emit_op1(p
, OPCODE_MOV
, out
, WRITEMASK_X
, in
);
1591 static void build_tnl_program( struct tnl_program
*p
)
1592 { /* Emit the program, starting with modelviewproject:
1596 /* Lighting calculations:
1598 if (p
->state
->fragprog_inputs_read
& (FRAG_BIT_COL0
|FRAG_BIT_COL1
)) {
1599 if (p
->state
->light_global_enabled
)
1602 if (p
->state
->fragprog_inputs_read
& FRAG_BIT_COL0
)
1603 emit_passthrough(p
, VERT_ATTRIB_COLOR0
, VERT_RESULT_COL0
);
1605 if (p
->state
->fragprog_inputs_read
& FRAG_BIT_COL1
)
1606 emit_passthrough(p
, VERT_ATTRIB_COLOR1
, VERT_RESULT_COL1
);
1610 if ((p
->state
->fragprog_inputs_read
& FRAG_BIT_FOGC
) ||
1611 p
->state
->fog_mode
!= FOG_NONE
)
1614 if (p
->state
->fragprog_inputs_read
& FRAG_BITS_TEX_ANY
)
1615 build_texture_transform(p
);
1617 if (p
->state
->point_attenuated
)
1618 build_atten_pointsize(p
);
1619 else if (p
->state
->point_array
)
1620 build_array_pointsize(p
);
1623 build_constant_pointsize(p
);
1628 emit_op1(p
, OPCODE_END
, undef
, 0, undef
);
1633 _mesa_printf ("\n");
1639 create_new_program( const struct state_key
*key
,
1640 struct gl_vertex_program
*program
,
1643 struct tnl_program p
;
1645 _mesa_memset(&p
, 0, sizeof(p
));
1647 p
.program
= program
;
1648 p
.eye_position
= undef
;
1649 p
.eye_position_z
= undef
;
1650 p
.eye_position_normalized
= undef
;
1651 p
.transformed_normal
= undef
;
1655 if (max_temps
>= sizeof(int) * 8)
1656 p
.temp_reserved
= 0;
1658 p
.temp_reserved
= ~((1<<max_temps
)-1);
1660 p
.program
->Base
.Instructions
= _mesa_alloc_instructions(MAX_INSN
);
1661 p
.program
->Base
.String
= NULL
;
1662 p
.program
->Base
.NumInstructions
=
1663 p
.program
->Base
.NumTemporaries
=
1664 p
.program
->Base
.NumParameters
=
1665 p
.program
->Base
.NumAttributes
= p
.program
->Base
.NumAddressRegs
= 0;
1666 p
.program
->Base
.Parameters
= _mesa_new_parameter_list();
1667 p
.program
->Base
.InputsRead
= 0;
1668 p
.program
->Base
.OutputsWritten
= 0;
1670 build_tnl_program( &p
);
1675 * Return a vertex program which implements the current fixed-function
1676 * transform/lighting/texgen operations.
1677 * XXX move this into core mesa (main/)
1679 struct gl_vertex_program
*
1680 _mesa_get_fixed_func_vertex_program(GLcontext
*ctx
)
1682 struct gl_vertex_program
*prog
;
1683 struct state_key
*key
;
1685 /* Grab all the relevent state and put it in a single structure:
1687 key
= make_state_key(ctx
);
1689 /* Look for an already-prepared program for this state:
1691 prog
= (struct gl_vertex_program
*)
1692 _mesa_search_program_cache(ctx
->VertexProgram
.Cache
, key
, sizeof(*key
));
1695 /* OK, we'll have to build a new one */
1697 _mesa_printf("Build new TNL program\n");
1699 prog
= (struct gl_vertex_program
*)
1700 ctx
->Driver
.NewProgram(ctx
, GL_VERTEX_PROGRAM_ARB
, 0);
1704 create_new_program( key
, prog
,
1705 ctx
->Const
.VertexProgram
.MaxTemps
);
1708 if (ctx
->Driver
.ProgramStringNotify
)
1709 ctx
->Driver
.ProgramStringNotify( ctx
, GL_VERTEX_PROGRAM_ARB
,
1712 _mesa_program_cache_insert(ctx
, ctx
->VertexProgram
.Cache
,
1713 key
, sizeof(*key
), &prog
->Base
);