1 /**************************************************************************
3 * Copyright 2007 VMware, Inc.
6 * Permission is hereby granted, free of charge, to any person obtaining a
7 * copy of this software and associated documentation files (the
8 * "Software"), to deal in the Software without restriction, including
9 * without limitation the rights to use, copy, modify, merge, publish,
10 * distribute, sub license, and/or sell copies of the Software, and to
11 * permit persons to whom the Software is furnished to do so, subject to
12 * the following conditions:
14 * The above copyright notice and this permission notice (including the
15 * next paragraph) shall be included in all copies or substantial portions
18 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
19 * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
20 * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
21 * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
22 * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
23 * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
24 * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
26 **************************************************************************/
29 * \file ffvertex_prog.c
31 * Create a vertex program to execute the current fixed function T&L pipeline.
32 * \author Keith Whitwell
36 #include "main/glheader.h"
37 #include "main/mtypes.h"
38 #include "main/macros.h"
39 #include "main/enums.h"
40 #include "main/ffvertex_prog.h"
41 #include "program/program.h"
42 #include "program/prog_cache.h"
43 #include "program/prog_instruction.h"
44 #include "program/prog_parameter.h"
45 #include "program/prog_print.h"
46 #include "program/prog_statevars.h"
49 /** Max of number of lights and texture coord units */
50 #define NUM_UNITS MAX2(MAX_TEXTURE_COORD_UNITS, MAX_LIGHTS)
53 unsigned light_color_material_mask:12;
54 unsigned light_global_enabled:1;
55 unsigned light_local_viewer:1;
56 unsigned light_twoside:1;
57 unsigned material_shininess_is_zero:1;
58 unsigned need_eye_coords:1;
60 unsigned rescale_normals:1;
62 unsigned fog_source_is_depth:1;
63 unsigned fog_distance_mode:2;
64 unsigned separate_specular:1;
65 unsigned point_attenuated:1;
66 unsigned point_array:1;
67 unsigned texture_enabled_global:1;
68 unsigned fragprog_inputs_read:12;
70 GLbitfield64 varying_vp_inputs;
73 unsigned light_enabled:1;
74 unsigned light_eyepos3_is_zero:1;
75 unsigned light_spotcutoff_is_180:1;
76 unsigned light_attenuated:1;
77 unsigned texunit_really_enabled:1;
78 unsigned texmat_enabled:1;
79 unsigned coord_replace:1;
80 unsigned texgen_enabled:4;
81 unsigned texgen_mode0:4;
82 unsigned texgen_mode1:4;
83 unsigned texgen_mode2:4;
84 unsigned texgen_mode3:4;
90 #define TXG_OBJ_LINEAR 1
91 #define TXG_EYE_LINEAR 2
92 #define TXG_SPHERE_MAP 3
93 #define TXG_REFLECTION_MAP 4
94 #define TXG_NORMAL_MAP 5
96 static GLuint translate_texgen( GLboolean enabled, GLenum mode )
102 case GL_OBJECT_LINEAR: return TXG_OBJ_LINEAR;
103 case GL_EYE_LINEAR: return TXG_EYE_LINEAR;
104 case GL_SPHERE_MAP: return TXG_SPHERE_MAP;
105 case GL_REFLECTION_MAP_NV: return TXG_REFLECTION_MAP;
106 case GL_NORMAL_MAP_NV: return TXG_NORMAL_MAP;
107 default: return TXG_NONE;
111 #define FDM_EYE_RADIAL 0
112 #define FDM_EYE_PLANE 1
113 #define FDM_EYE_PLANE_ABS 2
115 static GLuint translate_fog_distance_mode( GLenum mode )
118 case GL_EYE_RADIAL_NV:
119 return FDM_EYE_RADIAL;
121 return FDM_EYE_PLANE;
122 default: /* shouldn't happen; fall through to a sensible default */
123 case GL_EYE_PLANE_ABSOLUTE_NV:
124 return FDM_EYE_PLANE_ABS;
128 static GLboolean check_active_shininess( struct gl_context *ctx,
129 const struct state_key *key,
132 GLuint attr = MAT_ATTRIB_FRONT_SHININESS + side;
134 if ((key->varying_vp_inputs & VERT_BIT_COLOR0) &&
135 (key->light_color_material_mask & (1 << attr)))
138 if (key->varying_vp_inputs & VERT_BIT_GENERIC(attr))
141 if (ctx->Light.Material.Attrib[attr][0] != 0.0F)
148 static void make_state_key( struct gl_context *ctx, struct state_key *key )
150 const struct gl_fragment_program *fp;
153 memset(key, 0, sizeof(struct state_key));
154 fp = ctx->FragmentProgram._Current;
156 /* This now relies on texenvprogram.c being active:
160 key->need_eye_coords = ctx->_NeedEyeCoords;
162 key->fragprog_inputs_read = fp->Base.InputsRead;
163 key->varying_vp_inputs = ctx->varying_vp_inputs;
165 if (ctx->RenderMode == GL_FEEDBACK) {
166 /* make sure the vertprog emits color and tex0 */
167 key->fragprog_inputs_read |= (VARYING_BIT_COL0 | VARYING_BIT_TEX0);
170 key->separate_specular = (ctx->Light.Model.ColorControl ==
171 GL_SEPARATE_SPECULAR_COLOR);
173 if (ctx->Light.Enabled) {
174 key->light_global_enabled = 1;
176 if (ctx->Light.Model.LocalViewer)
177 key->light_local_viewer = 1;
179 if (ctx->Light.Model.TwoSide)
180 key->light_twoside = 1;
182 if (ctx->Light.ColorMaterialEnabled) {
183 key->light_color_material_mask = ctx->Light._ColorMaterialBitmask;
186 for (i = 0; i < MAX_LIGHTS; i++) {
187 struct gl_light *light = &ctx->Light.Light[i];
189 if (light->Enabled) {
190 key->unit[i].light_enabled = 1;
192 if (light->EyePosition[3] == 0.0F)
193 key->unit[i].light_eyepos3_is_zero = 1;
195 if (light->SpotCutoff == 180.0F)
196 key->unit[i].light_spotcutoff_is_180 = 1;
198 if (light->ConstantAttenuation != 1.0F ||
199 light->LinearAttenuation != 0.0F ||
200 light->QuadraticAttenuation != 0.0F)
201 key->unit[i].light_attenuated = 1;
205 if (check_active_shininess(ctx, key, 0)) {
206 key->material_shininess_is_zero = 0;
208 else if (key->light_twoside &&
209 check_active_shininess(ctx, key, 1)) {
210 key->material_shininess_is_zero = 0;
213 key->material_shininess_is_zero = 1;
217 if (ctx->Transform.Normalize)
220 if (ctx->Transform.RescaleNormals)
221 key->rescale_normals = 1;
223 if (ctx->Fog.FogCoordinateSource == GL_FRAGMENT_DEPTH_EXT) {
224 key->fog_source_is_depth = 1;
225 key->fog_distance_mode = translate_fog_distance_mode(ctx->Fog.FogDistanceMode);
228 if (ctx->Point._Attenuated)
229 key->point_attenuated = 1;
231 if (ctx->Array.VAO->VertexAttrib[VERT_ATTRIB_POINT_SIZE].Enabled)
232 key->point_array = 1;
234 if (ctx->Texture._TexGenEnabled ||
235 ctx->Texture._TexMatEnabled ||
236 ctx->Texture._MaxEnabledTexImageUnit != -1)
237 key->texture_enabled_global = 1;
239 for (i = 0; i < MAX_TEXTURE_COORD_UNITS; i++) {
240 struct gl_texture_unit *texUnit = &ctx->Texture.Unit[i];
242 if (texUnit->_Current)
243 key->unit[i].texunit_really_enabled = 1;
245 if (ctx->Point.PointSprite)
246 if (ctx->Point.CoordReplace[i])
247 key->unit[i].coord_replace = 1;
249 if (ctx->Texture._TexMatEnabled & ENABLE_TEXMAT(i))
250 key->unit[i].texmat_enabled = 1;
252 if (texUnit->TexGenEnabled) {
253 key->unit[i].texgen_enabled = 1;
255 key->unit[i].texgen_mode0 =
256 translate_texgen( texUnit->TexGenEnabled & (1<<0),
257 texUnit->GenS.Mode );
258 key->unit[i].texgen_mode1 =
259 translate_texgen( texUnit->TexGenEnabled & (1<<1),
260 texUnit->GenT.Mode );
261 key->unit[i].texgen_mode2 =
262 translate_texgen( texUnit->TexGenEnabled & (1<<2),
263 texUnit->GenR.Mode );
264 key->unit[i].texgen_mode3 =
265 translate_texgen( texUnit->TexGenEnabled & (1<<3),
266 texUnit->GenQ.Mode );
273 /* Very useful debugging tool - produces annotated listing of
274 * generated program with line/function references for each
275 * instruction back into this file:
280 /* Use uregs to represent registers internally, translate to Mesa's
281 * expected formats on emit.
283 * NOTE: These are passed by value extensively in this file rather
284 * than as usual by pointer reference. If this disturbs you, try
285 * remembering they are just 32bits in size.
287 * GCC is smart enough to deal with these dword-sized structures in
288 * much the same way as if I had defined them as dwords and was using
289 * macros to access and set the fields. This is much nicer and easier
294 GLint idx:9; /* relative addressing may be negative */
295 /* sizeof(idx) should == sizeof(prog_src_reg::Index) */
303 const struct state_key *state;
304 struct gl_vertex_program *program;
305 GLuint max_inst; /** number of instructions allocated for program */
306 GLboolean mvp_with_dp4;
309 GLuint temp_reserved;
311 struct ureg eye_position;
312 struct ureg eye_position_z;
313 struct ureg eye_position_normalized;
314 struct ureg transformed_normal;
315 struct ureg identity;
318 GLuint color_materials;
322 static const struct ureg undef = {
340 static struct ureg make_ureg(GLuint file, GLint idx)
346 reg.swz = SWIZZLE_NOOP;
352 static struct ureg negate( struct ureg reg )
359 static struct ureg swizzle( struct ureg reg, int x, int y, int z, int w )
361 reg.swz = MAKE_SWIZZLE4(GET_SWZ(reg.swz, x),
364 GET_SWZ(reg.swz, w));
369 static struct ureg swizzle1( struct ureg reg, int x )
371 return swizzle(reg, x, x, x, x);
375 static struct ureg get_temp( struct tnl_program *p )
377 int bit = ffs( ~p->temp_in_use );
379 _mesa_problem(NULL, "%s: out of temporaries\n", __FILE__);
383 if ((GLuint) bit > p->program->Base.NumTemporaries)
384 p->program->Base.NumTemporaries = bit;
386 p->temp_in_use |= 1<<(bit-1);
387 return make_ureg(PROGRAM_TEMPORARY, bit-1);
391 static struct ureg reserve_temp( struct tnl_program *p )
393 struct ureg temp = get_temp( p );
394 p->temp_reserved |= 1<<temp.idx;
399 static void release_temp( struct tnl_program *p, struct ureg reg )
401 if (reg.file == PROGRAM_TEMPORARY) {
402 p->temp_in_use &= ~(1<<reg.idx);
403 p->temp_in_use |= p->temp_reserved; /* can't release reserved temps */
407 static void release_temps( struct tnl_program *p )
409 p->temp_in_use = p->temp_reserved;
413 static struct ureg register_param5(struct tnl_program *p,
420 gl_state_index tokens[STATE_LENGTH];
427 idx = _mesa_add_state_reference( p->program->Base.Parameters, tokens );
428 return make_ureg(PROGRAM_STATE_VAR, idx);
432 #define register_param1(p,s0) register_param5(p,s0,0,0,0,0)
433 #define register_param2(p,s0,s1) register_param5(p,s0,s1,0,0,0)
434 #define register_param3(p,s0,s1,s2) register_param5(p,s0,s1,s2,0,0)
435 #define register_param4(p,s0,s1,s2,s3) register_param5(p,s0,s1,s2,s3,0)
440 * \param input one of VERT_ATTRIB_x tokens.
442 static struct ureg register_input( struct tnl_program *p, GLuint input )
444 assert(input < VERT_ATTRIB_MAX);
446 if (p->state->varying_vp_inputs & VERT_BIT(input)) {
447 p->program->Base.InputsRead |= VERT_BIT(input);
448 return make_ureg(PROGRAM_INPUT, input);
451 return register_param3( p, STATE_INTERNAL, STATE_CURRENT_ATTRIB, input );
457 * \param input one of VARYING_SLOT_x tokens.
459 static struct ureg register_output( struct tnl_program *p, GLuint output )
461 p->program->Base.OutputsWritten |= BITFIELD64_BIT(output);
462 return make_ureg(PROGRAM_OUTPUT, output);
466 static struct ureg register_const4f( struct tnl_program *p,
472 gl_constant_value values[4];
479 idx = _mesa_add_unnamed_constant( p->program->Base.Parameters, values, 4,
481 assert(swizzle == SWIZZLE_NOOP);
482 return make_ureg(PROGRAM_CONSTANT, idx);
485 #define register_const1f(p, s0) register_const4f(p, s0, 0, 0, 1)
486 #define register_scalar_const(p, s0) register_const4f(p, s0, s0, s0, s0)
487 #define register_const2f(p, s0, s1) register_const4f(p, s0, s1, 0, 1)
488 #define register_const3f(p, s0, s1, s2) register_const4f(p, s0, s1, s2, 1)
490 static GLboolean is_undef( struct ureg reg )
492 return reg.file == PROGRAM_UNDEFINED;
496 static struct ureg get_identity_param( struct tnl_program *p )
498 if (is_undef(p->identity))
499 p->identity = register_const4f(p, 0,0,0,1);
504 static void register_matrix_param5( struct tnl_program *p,
505 GLint s0, /* modelview, projection, etc */
506 GLint s1, /* texture matrix number */
507 GLint s2, /* first row */
508 GLint s3, /* last row */
509 GLint s4, /* inverse, transpose, etc */
510 struct ureg *matrix )
514 /* This is a bit sad as the support is there to pull the whole
515 * matrix out in one go:
517 for (i = 0; i <= s3 - s2; i++)
518 matrix[i] = register_param5( p, s0, s1, i, i, s4 );
522 static void emit_arg( struct prog_src_register *src,
525 src->File = reg.file;
526 src->Index = reg.idx;
527 src->Swizzle = reg.swz;
528 src->Negate = reg.negate ? NEGATE_XYZW : NEGATE_NONE;
530 /* Check that bitfield sizes aren't exceeded */
531 assert(src->Index == reg.idx);
535 static void emit_dst( struct prog_dst_register *dst,
536 struct ureg reg, GLuint mask )
538 dst->File = reg.file;
539 dst->Index = reg.idx;
540 /* allow zero as a shorthand for xyzw */
541 dst->WriteMask = mask ? mask : WRITEMASK_XYZW;
542 /* Check that bitfield sizes aren't exceeded */
543 assert(dst->Index == reg.idx);
547 static void debug_insn( struct prog_instruction *inst, const char *fn,
551 static const char *last_fn;
558 printf("%d:\t", line);
559 _mesa_print_instruction(inst);
564 static void emit_op3fn(struct tnl_program *p,
575 struct prog_instruction *inst;
577 assert(p->program->Base.NumInstructions <= p->max_inst);
579 if (p->program->Base.NumInstructions == p->max_inst) {
580 /* need to extend the program's instruction array */
581 struct prog_instruction *newInst;
583 /* double the size */
586 newInst = _mesa_alloc_instructions(p->max_inst);
588 _mesa_error(NULL, GL_OUT_OF_MEMORY, "vertex program build");
592 _mesa_copy_instructions(newInst,
593 p->program->Base.Instructions,
594 p->program->Base.NumInstructions);
596 _mesa_free_instructions(p->program->Base.Instructions,
597 p->program->Base.NumInstructions);
599 p->program->Base.Instructions = newInst;
602 nr = p->program->Base.NumInstructions++;
604 inst = &p->program->Base.Instructions[nr];
605 inst->Opcode = (enum prog_opcode) op;
607 emit_arg( &inst->SrcReg[0], src0 );
608 emit_arg( &inst->SrcReg[1], src1 );
609 emit_arg( &inst->SrcReg[2], src2 );
611 emit_dst( &inst->DstReg, dest, mask );
613 debug_insn(inst, fn, line);
617 #define emit_op3(p, op, dst, mask, src0, src1, src2) \
618 emit_op3fn(p, op, dst, mask, src0, src1, src2, __func__, __LINE__)
620 #define emit_op2(p, op, dst, mask, src0, src1) \
621 emit_op3fn(p, op, dst, mask, src0, src1, undef, __func__, __LINE__)
623 #define emit_op1(p, op, dst, mask, src0) \
624 emit_op3fn(p, op, dst, mask, src0, undef, undef, __func__, __LINE__)
627 static struct ureg make_temp( struct tnl_program *p, struct ureg reg )
629 if (reg.file == PROGRAM_TEMPORARY &&
630 !(p->temp_reserved & (1<<reg.idx)))
633 struct ureg temp = get_temp(p);
634 emit_op1(p, OPCODE_MOV, temp, 0, reg);
640 /* Currently no tracking performed of input/output/register size or
641 * active elements. Could be used to reduce these operations, as
642 * could the matrix type.
644 static void emit_matrix_transform_vec4( struct tnl_program *p,
646 const struct ureg *mat,
649 emit_op2(p, OPCODE_DP4, dest, WRITEMASK_X, src, mat[0]);
650 emit_op2(p, OPCODE_DP4, dest, WRITEMASK_Y, src, mat[1]);
651 emit_op2(p, OPCODE_DP4, dest, WRITEMASK_Z, src, mat[2]);
652 emit_op2(p, OPCODE_DP4, dest, WRITEMASK_W, src, mat[3]);
656 /* This version is much easier to implement if writemasks are not
657 * supported natively on the target or (like SSE), the target doesn't
658 * have a clean/obvious dotproduct implementation.
660 static void emit_transpose_matrix_transform_vec4( struct tnl_program *p,
662 const struct ureg *mat,
667 if (dest.file != PROGRAM_TEMPORARY)
672 emit_op2(p, OPCODE_MUL, tmp, 0, swizzle1(src,X), mat[0]);
673 emit_op3(p, OPCODE_MAD, tmp, 0, swizzle1(src,Y), mat[1], tmp);
674 emit_op3(p, OPCODE_MAD, tmp, 0, swizzle1(src,Z), mat[2], tmp);
675 emit_op3(p, OPCODE_MAD, dest, 0, swizzle1(src,W), mat[3], tmp);
677 if (dest.file != PROGRAM_TEMPORARY)
678 release_temp(p, tmp);
682 static void emit_matrix_transform_vec3( struct tnl_program *p,
684 const struct ureg *mat,
687 emit_op2(p, OPCODE_DP3, dest, WRITEMASK_X, src, mat[0]);
688 emit_op2(p, OPCODE_DP3, dest, WRITEMASK_Y, src, mat[1]);
689 emit_op2(p, OPCODE_DP3, dest, WRITEMASK_Z, src, mat[2]);
693 static void emit_normalize_vec3( struct tnl_program *p,
697 struct ureg tmp = get_temp(p);
698 emit_op2(p, OPCODE_DP3, tmp, WRITEMASK_X, src, src);
699 emit_op1(p, OPCODE_RSQ, tmp, WRITEMASK_X, tmp);
700 emit_op2(p, OPCODE_MUL, dest, 0, src, swizzle1(tmp, X));
701 release_temp(p, tmp);
705 static void emit_passthrough( struct tnl_program *p,
709 struct ureg out = register_output(p, output);
710 emit_op1(p, OPCODE_MOV, out, 0, register_input(p, input));
714 static struct ureg get_eye_position( struct tnl_program *p )
716 if (is_undef(p->eye_position)) {
717 struct ureg pos = register_input( p, VERT_ATTRIB_POS );
718 struct ureg modelview[4];
720 p->eye_position = reserve_temp(p);
722 if (p->mvp_with_dp4) {
723 register_matrix_param5( p, STATE_MODELVIEW_MATRIX, 0, 0, 3,
726 emit_matrix_transform_vec4(p, p->eye_position, modelview, pos);
729 register_matrix_param5( p, STATE_MODELVIEW_MATRIX, 0, 0, 3,
730 STATE_MATRIX_TRANSPOSE, modelview );
732 emit_transpose_matrix_transform_vec4(p, p->eye_position, modelview, pos);
736 return p->eye_position;
740 static struct ureg get_eye_position_z( struct tnl_program *p )
742 if (!is_undef(p->eye_position))
743 return swizzle1(p->eye_position, Z);
745 if (is_undef(p->eye_position_z)) {
746 struct ureg pos = register_input( p, VERT_ATTRIB_POS );
747 struct ureg modelview[4];
749 p->eye_position_z = reserve_temp(p);
751 register_matrix_param5( p, STATE_MODELVIEW_MATRIX, 0, 0, 3,
754 emit_op2(p, OPCODE_DP4, p->eye_position_z, 0, pos, modelview[2]);
757 return p->eye_position_z;
761 static struct ureg get_eye_position_normalized( struct tnl_program *p )
763 if (is_undef(p->eye_position_normalized)) {
764 struct ureg eye = get_eye_position(p);
765 p->eye_position_normalized = reserve_temp(p);
766 emit_normalize_vec3(p, p->eye_position_normalized, eye);
769 return p->eye_position_normalized;
773 static struct ureg get_transformed_normal( struct tnl_program *p )
775 if (is_undef(p->transformed_normal) &&
776 !p->state->need_eye_coords &&
777 !p->state->normalize &&
778 !(p->state->need_eye_coords == p->state->rescale_normals))
780 p->transformed_normal = register_input(p, VERT_ATTRIB_NORMAL );
782 else if (is_undef(p->transformed_normal))
784 struct ureg normal = register_input(p, VERT_ATTRIB_NORMAL );
785 struct ureg mvinv[3];
786 struct ureg transformed_normal = reserve_temp(p);
788 if (p->state->need_eye_coords) {
789 register_matrix_param5( p, STATE_MODELVIEW_MATRIX, 0, 0, 2,
790 STATE_MATRIX_INVTRANS, mvinv );
792 /* Transform to eye space:
794 emit_matrix_transform_vec3( p, transformed_normal, mvinv, normal );
795 normal = transformed_normal;
798 /* Normalize/Rescale:
800 if (p->state->normalize) {
801 emit_normalize_vec3( p, transformed_normal, normal );
802 normal = transformed_normal;
804 else if (p->state->need_eye_coords == p->state->rescale_normals) {
805 /* This is already adjusted for eye/non-eye rendering:
807 struct ureg rescale = register_param2(p, STATE_INTERNAL,
810 emit_op2( p, OPCODE_MUL, transformed_normal, 0, normal, rescale );
811 normal = transformed_normal;
814 assert(normal.file == PROGRAM_TEMPORARY);
815 p->transformed_normal = normal;
818 return p->transformed_normal;
822 static void build_hpos( struct tnl_program *p )
824 struct ureg pos = register_input( p, VERT_ATTRIB_POS );
825 struct ureg hpos = register_output( p, VARYING_SLOT_POS );
828 if (p->mvp_with_dp4) {
829 register_matrix_param5( p, STATE_MVP_MATRIX, 0, 0, 3,
831 emit_matrix_transform_vec4( p, hpos, mvp, pos );
834 register_matrix_param5( p, STATE_MVP_MATRIX, 0, 0, 3,
835 STATE_MATRIX_TRANSPOSE, mvp );
836 emit_transpose_matrix_transform_vec4( p, hpos, mvp, pos );
841 static GLuint material_attrib( GLuint side, GLuint property )
843 return (property - STATE_AMBIENT) * 2 + side;
848 * Get a bitmask of which material values vary on a per-vertex basis.
850 static void set_material_flags( struct tnl_program *p )
852 p->color_materials = 0;
855 if (p->state->varying_vp_inputs & VERT_BIT_COLOR0) {
857 p->color_materials = p->state->light_color_material_mask;
860 p->materials |= (p->state->varying_vp_inputs >> VERT_ATTRIB_GENERIC0);
864 static struct ureg get_material( struct tnl_program *p, GLuint side,
867 GLuint attrib = material_attrib(side, property);
869 if (p->color_materials & (1<<attrib))
870 return register_input(p, VERT_ATTRIB_COLOR0);
871 else if (p->materials & (1<<attrib)) {
872 /* Put material values in the GENERIC slots -- they are not used
873 * for anything in fixed function mode.
875 return register_input( p, attrib + VERT_ATTRIB_GENERIC0 );
878 return register_param3( p, STATE_MATERIAL, side, property );
881 #define SCENE_COLOR_BITS(side) (( MAT_BIT_FRONT_EMISSION | \
882 MAT_BIT_FRONT_AMBIENT | \
883 MAT_BIT_FRONT_DIFFUSE) << (side))
887 * Either return a precalculated constant value or emit code to
888 * calculate these values dynamically in the case where material calls
889 * are present between begin/end pairs.
891 * Probably want to shift this to the program compilation phase - if
892 * we always emitted the calculation here, a smart compiler could
893 * detect that it was constant (given a certain set of inputs), and
894 * lift it out of the main loop. That way the programs created here
895 * would be independent of the vertex_buffer details.
897 static struct ureg get_scenecolor( struct tnl_program *p, GLuint side )
899 if (p->materials & SCENE_COLOR_BITS(side)) {
900 struct ureg lm_ambient = register_param1(p, STATE_LIGHTMODEL_AMBIENT);
901 struct ureg material_emission = get_material(p, side, STATE_EMISSION);
902 struct ureg material_ambient = get_material(p, side, STATE_AMBIENT);
903 struct ureg material_diffuse = get_material(p, side, STATE_DIFFUSE);
904 struct ureg tmp = make_temp(p, material_diffuse);
905 emit_op3(p, OPCODE_MAD, tmp, WRITEMASK_XYZ, lm_ambient,
906 material_ambient, material_emission);
910 return register_param2( p, STATE_LIGHTMODEL_SCENECOLOR, side );
914 static struct ureg get_lightprod( struct tnl_program *p, GLuint light,
915 GLuint side, GLuint property )
917 GLuint attrib = material_attrib(side, property);
918 if (p->materials & (1<<attrib)) {
919 struct ureg light_value =
920 register_param3(p, STATE_LIGHT, light, property);
921 struct ureg material_value = get_material(p, side, property);
922 struct ureg tmp = get_temp(p);
923 emit_op2(p, OPCODE_MUL, tmp, 0, light_value, material_value);
927 return register_param4(p, STATE_LIGHTPROD, light, side, property);
931 static struct ureg calculate_light_attenuation( struct tnl_program *p,
936 struct ureg attenuation = register_param3(p, STATE_LIGHT, i,
938 struct ureg att = undef;
940 /* Calculate spot attenuation:
942 if (!p->state->unit[i].light_spotcutoff_is_180) {
943 struct ureg spot_dir_norm = register_param3(p, STATE_INTERNAL,
944 STATE_LIGHT_SPOT_DIR_NORMALIZED, i);
945 struct ureg spot = get_temp(p);
946 struct ureg slt = get_temp(p);
950 emit_op2(p, OPCODE_DP3, spot, 0, negate(VPpli), spot_dir_norm);
951 emit_op2(p, OPCODE_SLT, slt, 0, swizzle1(spot_dir_norm,W), spot);
952 emit_op1(p, OPCODE_ABS, spot, 0, spot);
953 emit_op2(p, OPCODE_POW, spot, 0, spot, swizzle1(attenuation, W));
954 emit_op2(p, OPCODE_MUL, att, 0, slt, spot);
956 release_temp(p, spot);
957 release_temp(p, slt);
960 /* Calculate distance attenuation(See formula (2.4) at glspec 2.1 page 62):
962 * Skip the calucation when _dist_ is undefined(light_eyepos3_is_zero)
964 if (p->state->unit[i].light_attenuated && !is_undef(dist)) {
968 emit_op1(p, OPCODE_RCP, dist, WRITEMASK_YZ, dist);
970 emit_op2(p, OPCODE_MUL, dist, WRITEMASK_XZ, dist, swizzle1(dist,Y));
972 emit_op2(p, OPCODE_DP3, dist, 0, attenuation, dist);
974 if (!p->state->unit[i].light_spotcutoff_is_180) {
976 emit_op1(p, OPCODE_RCP, dist, 0, dist);
977 /* spot-atten * dist-atten */
978 emit_op2(p, OPCODE_MUL, att, 0, dist, att);
982 emit_op1(p, OPCODE_RCP, att, 0, dist);
992 * lit.y = MAX(0, dots.x)
993 * lit.z = SLT(0, dots.x)
995 static void emit_degenerate_lit( struct tnl_program *p,
999 struct ureg id = get_identity_param(p); /* id = {0,0,0,1} */
1001 /* Note that lit.x & lit.w will not be examined. Note also that
1002 * dots.xyzw == dots.xxxx.
1005 /* MAX lit, id, dots;
1007 emit_op2(p, OPCODE_MAX, lit, WRITEMASK_XYZW, id, dots);
1009 /* result[2] = (in > 0 ? 1 : 0)
1010 * SLT lit.z, id.z, dots; # lit.z = (0 < dots.z) ? 1 : 0
1012 emit_op2(p, OPCODE_SLT, lit, WRITEMASK_Z, swizzle1(id,Z), dots);
1016 /* Need to add some addtional parameters to allow lighting in object
1017 * space - STATE_SPOT_DIRECTION and STATE_HALF_VECTOR implicitly assume eye
1020 static void build_lighting( struct tnl_program *p )
1022 const GLboolean twoside = p->state->light_twoside;
1023 const GLboolean separate = p->state->separate_specular;
1024 GLuint nr_lights = 0, count = 0;
1025 struct ureg normal = get_transformed_normal(p);
1026 struct ureg lit = get_temp(p);
1027 struct ureg dots = get_temp(p);
1028 struct ureg _col0 = undef, _col1 = undef;
1029 struct ureg _bfc0 = undef, _bfc1 = undef;
1034 * dots.x = dot(normal, VPpli)
1035 * dots.y = dot(normal, halfAngle)
1036 * dots.z = back.shininess
1037 * dots.w = front.shininess
1040 for (i = 0; i < MAX_LIGHTS; i++)
1041 if (p->state->unit[i].light_enabled)
1044 set_material_flags(p);
1047 if (!p->state->material_shininess_is_zero) {
1048 struct ureg shininess = get_material(p, 0, STATE_SHININESS);
1049 emit_op1(p, OPCODE_MOV, dots, WRITEMASK_W, swizzle1(shininess,X));
1050 release_temp(p, shininess);
1053 _col0 = make_temp(p, get_scenecolor(p, 0));
1055 _col1 = make_temp(p, get_identity_param(p));
1061 if (!p->state->material_shininess_is_zero) {
1062 /* Note that we negate the back-face specular exponent here.
1063 * The negation will be un-done later in the back-face code below.
1065 struct ureg shininess = get_material(p, 1, STATE_SHININESS);
1066 emit_op1(p, OPCODE_MOV, dots, WRITEMASK_Z,
1067 negate(swizzle1(shininess,X)));
1068 release_temp(p, shininess);
1071 _bfc0 = make_temp(p, get_scenecolor(p, 1));
1073 _bfc1 = make_temp(p, get_identity_param(p));
1078 /* If no lights, still need to emit the scenecolor.
1081 struct ureg res0 = register_output( p, VARYING_SLOT_COL0 );
1082 emit_op1(p, OPCODE_MOV, res0, 0, _col0);
1086 struct ureg res1 = register_output( p, VARYING_SLOT_COL1 );
1087 emit_op1(p, OPCODE_MOV, res1, 0, _col1);
1091 struct ureg res0 = register_output( p, VARYING_SLOT_BFC0 );
1092 emit_op1(p, OPCODE_MOV, res0, 0, _bfc0);
1095 if (twoside && separate) {
1096 struct ureg res1 = register_output( p, VARYING_SLOT_BFC1 );
1097 emit_op1(p, OPCODE_MOV, res1, 0, _bfc1);
1100 if (nr_lights == 0) {
1105 for (i = 0; i < MAX_LIGHTS; i++) {
1106 if (p->state->unit[i].light_enabled) {
1107 struct ureg half = undef;
1108 struct ureg att = undef, VPpli = undef;
1109 struct ureg dist = undef;
1112 if (p->state->unit[i].light_eyepos3_is_zero) {
1113 VPpli = register_param3(p, STATE_INTERNAL,
1114 STATE_LIGHT_POSITION_NORMALIZED, i);
1116 struct ureg Ppli = register_param3(p, STATE_INTERNAL,
1117 STATE_LIGHT_POSITION, i);
1118 struct ureg V = get_eye_position(p);
1120 VPpli = get_temp(p);
1123 /* Calculate VPpli vector
1125 emit_op2(p, OPCODE_SUB, VPpli, 0, Ppli, V);
1127 /* Normalize VPpli. The dist value also used in
1128 * attenuation below.
1130 emit_op2(p, OPCODE_DP3, dist, 0, VPpli, VPpli);
1131 emit_op1(p, OPCODE_RSQ, dist, 0, dist);
1132 emit_op2(p, OPCODE_MUL, VPpli, 0, VPpli, dist);
1135 /* Calculate attenuation:
1137 att = calculate_light_attenuation(p, i, VPpli, dist);
1138 release_temp(p, dist);
1140 /* Calculate viewer direction, or use infinite viewer:
1142 if (!p->state->material_shininess_is_zero) {
1143 if (p->state->light_local_viewer) {
1144 struct ureg eye_hat = get_eye_position_normalized(p);
1146 emit_op2(p, OPCODE_SUB, half, 0, VPpli, eye_hat);
1147 emit_normalize_vec3(p, half, half);
1148 } else if (p->state->unit[i].light_eyepos3_is_zero) {
1149 half = register_param3(p, STATE_INTERNAL,
1150 STATE_LIGHT_HALF_VECTOR, i);
1152 struct ureg z_dir = swizzle(get_identity_param(p),X,Y,W,Z);
1154 emit_op2(p, OPCODE_ADD, half, 0, VPpli, z_dir);
1155 emit_normalize_vec3(p, half, half);
1159 /* Calculate dot products:
1161 if (p->state->material_shininess_is_zero) {
1162 emit_op2(p, OPCODE_DP3, dots, 0, normal, VPpli);
1165 emit_op2(p, OPCODE_DP3, dots, WRITEMASK_X, normal, VPpli);
1166 emit_op2(p, OPCODE_DP3, dots, WRITEMASK_Y, normal, half);
1169 /* Front face lighting:
1172 struct ureg ambient = get_lightprod(p, i, 0, STATE_AMBIENT);
1173 struct ureg diffuse = get_lightprod(p, i, 0, STATE_DIFFUSE);
1174 struct ureg specular = get_lightprod(p, i, 0, STATE_SPECULAR);
1175 struct ureg res0, res1;
1176 GLuint mask0, mask1;
1178 if (count == nr_lights) {
1180 mask0 = WRITEMASK_XYZ;
1181 mask1 = WRITEMASK_XYZ;
1182 res0 = register_output( p, VARYING_SLOT_COL0 );
1183 res1 = register_output( p, VARYING_SLOT_COL1 );
1187 mask1 = WRITEMASK_XYZ;
1189 res1 = register_output( p, VARYING_SLOT_COL0 );
1199 if (!is_undef(att)) {
1200 /* light is attenuated by distance */
1201 emit_op1(p, OPCODE_LIT, lit, 0, dots);
1202 emit_op2(p, OPCODE_MUL, lit, 0, lit, att);
1203 emit_op3(p, OPCODE_MAD, _col0, 0, swizzle1(lit,X), ambient, _col0);
1205 else if (!p->state->material_shininess_is_zero) {
1206 /* there's a non-zero specular term */
1207 emit_op1(p, OPCODE_LIT, lit, 0, dots);
1208 emit_op2(p, OPCODE_ADD, _col0, 0, ambient, _col0);
1211 /* no attenutation, no specular */
1212 emit_degenerate_lit(p, lit, dots);
1213 emit_op2(p, OPCODE_ADD, _col0, 0, ambient, _col0);
1216 emit_op3(p, OPCODE_MAD, res0, mask0, swizzle1(lit,Y), diffuse, _col0);
1217 emit_op3(p, OPCODE_MAD, res1, mask1, swizzle1(lit,Z), specular, _col1);
1219 release_temp(p, ambient);
1220 release_temp(p, diffuse);
1221 release_temp(p, specular);
1224 /* Back face lighting:
1227 struct ureg ambient = get_lightprod(p, i, 1, STATE_AMBIENT);
1228 struct ureg diffuse = get_lightprod(p, i, 1, STATE_DIFFUSE);
1229 struct ureg specular = get_lightprod(p, i, 1, STATE_SPECULAR);
1230 struct ureg res0, res1;
1231 GLuint mask0, mask1;
1233 if (count == nr_lights) {
1235 mask0 = WRITEMASK_XYZ;
1236 mask1 = WRITEMASK_XYZ;
1237 res0 = register_output( p, VARYING_SLOT_BFC0 );
1238 res1 = register_output( p, VARYING_SLOT_BFC1 );
1242 mask1 = WRITEMASK_XYZ;
1244 res1 = register_output( p, VARYING_SLOT_BFC0 );
1254 /* For the back face we need to negate the X and Y component
1255 * dot products. dots.Z has the negated back-face specular
1256 * exponent. We swizzle that into the W position. This
1257 * negation makes the back-face specular term positive again.
1259 dots = negate(swizzle(dots,X,Y,W,Z));
1261 if (!is_undef(att)) {
1262 emit_op1(p, OPCODE_LIT, lit, 0, dots);
1263 emit_op2(p, OPCODE_MUL, lit, 0, lit, att);
1264 emit_op3(p, OPCODE_MAD, _bfc0, 0, swizzle1(lit,X), ambient, _bfc0);
1266 else if (!p->state->material_shininess_is_zero) {
1267 emit_op1(p, OPCODE_LIT, lit, 0, dots);
1268 emit_op2(p, OPCODE_ADD, _bfc0, 0, ambient, _bfc0); /**/
1271 emit_degenerate_lit(p, lit, dots);
1272 emit_op2(p, OPCODE_ADD, _bfc0, 0, ambient, _bfc0);
1275 emit_op3(p, OPCODE_MAD, res0, mask0, swizzle1(lit,Y), diffuse, _bfc0);
1276 emit_op3(p, OPCODE_MAD, res1, mask1, swizzle1(lit,Z), specular, _bfc1);
1277 /* restore dots to its original state for subsequent lights
1278 * by negating and swizzling again.
1280 dots = negate(swizzle(dots,X,Y,W,Z));
1282 release_temp(p, ambient);
1283 release_temp(p, diffuse);
1284 release_temp(p, specular);
1287 release_temp(p, half);
1288 release_temp(p, VPpli);
1289 release_temp(p, att);
1297 static void build_fog( struct tnl_program *p )
1299 struct ureg fog = register_output(p, VARYING_SLOT_FOGC);
1302 if (p->state->fog_source_is_depth) {
1304 switch (p->state->fog_distance_mode) {
1305 case FDM_EYE_RADIAL: /* Z = sqrt(Xe*Xe + Ye*Ye + Ze*Ze) */
1306 input = get_eye_position(p);
1307 emit_op2(p, OPCODE_DP3, fog, WRITEMASK_X, input, input);
1308 emit_op1(p, OPCODE_RSQ, fog, WRITEMASK_X, fog);
1309 emit_op1(p, OPCODE_RCP, fog, WRITEMASK_X, fog);
1311 case FDM_EYE_PLANE: /* Z = Ze */
1312 input = get_eye_position_z(p);
1313 emit_op1(p, OPCODE_MOV, fog, WRITEMASK_X, input);
1315 case FDM_EYE_PLANE_ABS: /* Z = abs(Ze) */
1316 input = get_eye_position_z(p);
1317 emit_op1(p, OPCODE_ABS, fog, WRITEMASK_X, input);
1320 assert(!"Bad fog mode in build_fog()");
1326 input = swizzle1(register_input(p, VERT_ATTRIB_FOG), X);
1327 emit_op1(p, OPCODE_ABS, fog, WRITEMASK_X, input);
1330 emit_op1(p, OPCODE_MOV, fog, WRITEMASK_YZW, get_identity_param(p));
1334 static void build_reflect_texgen( struct tnl_program *p,
1338 struct ureg normal = get_transformed_normal(p);
1339 struct ureg eye_hat = get_eye_position_normalized(p);
1340 struct ureg tmp = get_temp(p);
1343 emit_op2(p, OPCODE_DP3, tmp, 0, normal, eye_hat);
1345 emit_op2(p, OPCODE_ADD, tmp, 0, tmp, tmp);
1347 emit_op3(p, OPCODE_MAD, dest, writemask, negate(tmp), normal, eye_hat);
1349 release_temp(p, tmp);
1353 static void build_sphere_texgen( struct tnl_program *p,
1357 struct ureg normal = get_transformed_normal(p);
1358 struct ureg eye_hat = get_eye_position_normalized(p);
1359 struct ureg tmp = get_temp(p);
1360 struct ureg half = register_scalar_const(p, .5);
1361 struct ureg r = get_temp(p);
1362 struct ureg inv_m = get_temp(p);
1363 struct ureg id = get_identity_param(p);
1365 /* Could share the above calculations, but it would be
1366 * a fairly odd state for someone to set (both sphere and
1367 * reflection active for different texture coordinate
1368 * components. Of course - if two texture units enable
1369 * reflect and/or sphere, things start to tilt in favour
1370 * of seperating this out:
1374 emit_op2(p, OPCODE_DP3, tmp, 0, normal, eye_hat);
1376 emit_op2(p, OPCODE_ADD, tmp, 0, tmp, tmp);
1378 emit_op3(p, OPCODE_MAD, r, 0, negate(tmp), normal, eye_hat);
1380 emit_op2(p, OPCODE_ADD, tmp, 0, r, swizzle(id,X,Y,W,Z));
1381 /* rx^2 + ry^2 + (rz+1)^2 */
1382 emit_op2(p, OPCODE_DP3, tmp, 0, tmp, tmp);
1384 emit_op1(p, OPCODE_RSQ, tmp, 0, tmp);
1386 emit_op2(p, OPCODE_MUL, inv_m, 0, tmp, half);
1388 emit_op3(p, OPCODE_MAD, dest, writemask, r, inv_m, half);
1390 release_temp(p, tmp);
1392 release_temp(p, inv_m);
1396 static void build_texture_transform( struct tnl_program *p )
1400 for (i = 0; i < MAX_TEXTURE_COORD_UNITS; i++) {
1402 if (!(p->state->fragprog_inputs_read & VARYING_BIT_TEX(i)))
1405 if (p->state->unit[i].coord_replace)
1408 if (p->state->unit[i].texgen_enabled ||
1409 p->state->unit[i].texmat_enabled) {
1411 GLuint texmat_enabled = p->state->unit[i].texmat_enabled;
1412 struct ureg out = register_output(p, VARYING_SLOT_TEX0 + i);
1413 struct ureg out_texgen = undef;
1415 if (p->state->unit[i].texgen_enabled) {
1416 GLuint copy_mask = 0;
1417 GLuint sphere_mask = 0;
1418 GLuint reflect_mask = 0;
1419 GLuint normal_mask = 0;
1423 out_texgen = get_temp(p);
1427 modes[0] = p->state->unit[i].texgen_mode0;
1428 modes[1] = p->state->unit[i].texgen_mode1;
1429 modes[2] = p->state->unit[i].texgen_mode2;
1430 modes[3] = p->state->unit[i].texgen_mode3;
1432 for (j = 0; j < 4; j++) {
1434 case TXG_OBJ_LINEAR: {
1435 struct ureg obj = register_input(p, VERT_ATTRIB_POS);
1437 register_param3(p, STATE_TEXGEN, i,
1438 STATE_TEXGEN_OBJECT_S + j);
1440 emit_op2(p, OPCODE_DP4, out_texgen, WRITEMASK_X << j,
1444 case TXG_EYE_LINEAR: {
1445 struct ureg eye = get_eye_position(p);
1447 register_param3(p, STATE_TEXGEN, i,
1448 STATE_TEXGEN_EYE_S + j);
1450 emit_op2(p, OPCODE_DP4, out_texgen, WRITEMASK_X << j,
1454 case TXG_SPHERE_MAP:
1455 sphere_mask |= WRITEMASK_X << j;
1457 case TXG_REFLECTION_MAP:
1458 reflect_mask |= WRITEMASK_X << j;
1460 case TXG_NORMAL_MAP:
1461 normal_mask |= WRITEMASK_X << j;
1464 copy_mask |= WRITEMASK_X << j;
1469 build_sphere_texgen(p, out_texgen, sphere_mask);
1473 build_reflect_texgen(p, out_texgen, reflect_mask);
1477 struct ureg normal = get_transformed_normal(p);
1478 emit_op1(p, OPCODE_MOV, out_texgen, normal_mask, normal );
1482 struct ureg in = register_input(p, VERT_ATTRIB_TEX0+i);
1483 emit_op1(p, OPCODE_MOV, out_texgen, copy_mask, in );
1487 if (texmat_enabled) {
1488 struct ureg texmat[4];
1489 struct ureg in = (!is_undef(out_texgen) ?
1491 register_input(p, VERT_ATTRIB_TEX0+i));
1492 if (p->mvp_with_dp4) {
1493 register_matrix_param5( p, STATE_TEXTURE_MATRIX, i, 0, 3,
1495 emit_matrix_transform_vec4( p, out, texmat, in );
1498 register_matrix_param5( p, STATE_TEXTURE_MATRIX, i, 0, 3,
1499 STATE_MATRIX_TRANSPOSE, texmat );
1500 emit_transpose_matrix_transform_vec4( p, out, texmat, in );
1507 emit_passthrough(p, VERT_ATTRIB_TEX0+i, VARYING_SLOT_TEX0+i);
1514 * Point size attenuation computation.
1516 static void build_atten_pointsize( struct tnl_program *p )
1518 struct ureg eye = get_eye_position_z(p);
1519 struct ureg state_size = register_param2(p, STATE_INTERNAL, STATE_POINT_SIZE_CLAMPED);
1520 struct ureg state_attenuation = register_param1(p, STATE_POINT_ATTENUATION);
1521 struct ureg out = register_output(p, VARYING_SLOT_PSIZ);
1522 struct ureg ut = get_temp(p);
1525 emit_op1(p, OPCODE_ABS, ut, WRITEMASK_Y, swizzle1(eye, Z));
1526 /* p1 + dist * (p2 + dist * p3); */
1527 emit_op3(p, OPCODE_MAD, ut, WRITEMASK_X, swizzle1(ut, Y),
1528 swizzle1(state_attenuation, Z), swizzle1(state_attenuation, Y));
1529 emit_op3(p, OPCODE_MAD, ut, WRITEMASK_X, swizzle1(ut, Y),
1530 ut, swizzle1(state_attenuation, X));
1532 /* 1 / sqrt(factor) */
1533 emit_op1(p, OPCODE_RSQ, ut, WRITEMASK_X, ut );
1536 /* out = pointSize / sqrt(factor) */
1537 emit_op2(p, OPCODE_MUL, out, WRITEMASK_X, ut, state_size);
1539 /* this is a good place to clamp the point size since there's likely
1540 * no hardware registers to clamp point size at rasterization time.
1542 emit_op2(p, OPCODE_MUL, ut, WRITEMASK_X, ut, state_size);
1543 emit_op2(p, OPCODE_MAX, ut, WRITEMASK_X, ut, swizzle1(state_size, Y));
1544 emit_op2(p, OPCODE_MIN, out, WRITEMASK_X, ut, swizzle1(state_size, Z));
1547 release_temp(p, ut);
1552 * Pass-though per-vertex point size, from user's point size array.
1554 static void build_array_pointsize( struct tnl_program *p )
1556 struct ureg in = register_input(p, VERT_ATTRIB_POINT_SIZE);
1557 struct ureg out = register_output(p, VARYING_SLOT_PSIZ);
1558 emit_op1(p, OPCODE_MOV, out, WRITEMASK_X, in);
1562 static void build_tnl_program( struct tnl_program *p )
1564 /* Emit the program, starting with the modelview, projection transforms:
1568 /* Lighting calculations:
1570 if (p->state->fragprog_inputs_read & (VARYING_BIT_COL0|VARYING_BIT_COL1)) {
1571 if (p->state->light_global_enabled)
1574 if (p->state->fragprog_inputs_read & VARYING_BIT_COL0)
1575 emit_passthrough(p, VERT_ATTRIB_COLOR0, VARYING_SLOT_COL0);
1577 if (p->state->fragprog_inputs_read & VARYING_BIT_COL1)
1578 emit_passthrough(p, VERT_ATTRIB_COLOR1, VARYING_SLOT_COL1);
1582 if (p->state->fragprog_inputs_read & VARYING_BIT_FOGC)
1585 if (p->state->fragprog_inputs_read & VARYING_BITS_TEX_ANY)
1586 build_texture_transform(p);
1588 if (p->state->point_attenuated)
1589 build_atten_pointsize(p);
1590 else if (p->state->point_array)
1591 build_array_pointsize(p);
1595 emit_op1(p, OPCODE_END, undef, 0, undef);
1606 create_new_program( const struct state_key *key,
1607 struct gl_vertex_program *program,
1608 GLboolean mvp_with_dp4,
1611 struct tnl_program p;
1613 memset(&p, 0, sizeof(p));
1615 p.program = program;
1616 p.eye_position = undef;
1617 p.eye_position_z = undef;
1618 p.eye_position_normalized = undef;
1619 p.transformed_normal = undef;
1622 p.mvp_with_dp4 = mvp_with_dp4;
1624 if (max_temps >= sizeof(int) * 8)
1625 p.temp_reserved = 0;
1627 p.temp_reserved = ~((1<<max_temps)-1);
1629 /* Start by allocating 32 instructions.
1630 * If we need more, we'll grow the instruction array as needed.
1633 p.program->Base.Instructions = _mesa_alloc_instructions(p.max_inst);
1634 p.program->Base.String = NULL;
1635 p.program->Base.NumInstructions =
1636 p.program->Base.NumTemporaries =
1637 p.program->Base.NumParameters =
1638 p.program->Base.NumAttributes = p.program->Base.NumAddressRegs = 0;
1639 p.program->Base.Parameters = _mesa_new_parameter_list();
1640 p.program->Base.InputsRead = 0;
1641 p.program->Base.OutputsWritten = 0;
1643 build_tnl_program( &p );
1648 * Return a vertex program which implements the current fixed-function
1649 * transform/lighting/texgen operations.
1651 struct gl_vertex_program *
1652 _mesa_get_fixed_func_vertex_program(struct gl_context *ctx)
1654 struct gl_vertex_program *prog;
1655 struct state_key key;
1657 /* Grab all the relevant state and put it in a single structure:
1659 make_state_key(ctx, &key);
1661 /* Look for an already-prepared program for this state:
1663 prog = gl_vertex_program(
1664 _mesa_search_program_cache(ctx->VertexProgram.Cache, &key, sizeof(key)));
1667 /* OK, we'll have to build a new one */
1669 printf("Build new TNL program\n");
1671 prog = gl_vertex_program(ctx->Driver.NewProgram(ctx, GL_VERTEX_PROGRAM_ARB, 0));
1675 create_new_program( &key, prog,
1676 ctx->Const.ShaderCompilerOptions[MESA_SHADER_VERTEX].OptimizeForAOS,
1677 ctx->Const.Program[MESA_SHADER_VERTEX].MaxTemps );
1679 if (ctx->Driver.ProgramStringNotify)
1680 ctx->Driver.ProgramStringNotify( ctx, GL_VERTEX_PROGRAM_ARB,
1683 _mesa_program_cache_insert(ctx, ctx->VertexProgram.Cache,
1684 &key, sizeof(key), &prog->Base);