/* * Mesa 3-D graphics library * Version: 6.5 * * Copyright (C) 1999-2006 Brian Paul All Rights Reserved. * * Permission is hereby granted, free of charge, to any person obtaining a * copy of this software and associated documentation files (the "Software"), * to deal in the Software without restriction, including without limitation * the rights to use, copy, modify, merge, publish, distribute, sublicense, * and/or sell copies of the Software, and to permit persons to whom the * Software is furnished to do so, subject to the following conditions: * * The above copyright notice and this permission notice shall be included * in all copies or substantial portions of the Software. * * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL * THE AUTHORS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ #include "main/glheader.h" #include "main/colormac.h" #include "main/light.h" #include "main/macros.h" #include "main/imports.h" #include "main/simple_list.h" #include "main/mtypes.h" #include "math/m_translate.h" #include "t_context.h" #include "t_pipeline.h" #include "tnl.h" #define LIGHT_TWOSIDE 0x1 #define LIGHT_MATERIAL 0x2 #define MAX_LIGHT_FUNC 0x4 typedef void (*light_func)( struct gl_context *ctx, struct vertex_buffer *VB, struct tnl_pipeline_stage *stage, GLvector4f *input ); /** * Information for updating current material attributes from vertex color, * for GL_COLOR_MATERIAL. */ struct material_cursor { const GLfloat *ptr; /* points to src vertex color (in VB array) */ GLuint stride; /* stride to next vertex color (bytes) */ GLfloat *current; /* points to material attribute to update */ GLuint size; /* vertex/color size: 1, 2, 3 or 4 */ }; /** * Data private to this pipeline stage. */ struct light_stage_data { GLvector4f Input; GLvector4f LitColor[2]; GLvector4f LitSecondary[2]; light_func *light_func_tab; struct material_cursor mat[MAT_ATTRIB_MAX]; GLuint mat_count; GLuint mat_bitmask; }; #define LIGHT_STAGE_DATA(stage) ((struct light_stage_data *)(stage->privatePtr)) /**********************************************************************/ /***** Lighting computation *****/ /**********************************************************************/ /* * Notes: * When two-sided lighting is enabled we compute the color (or index) * for both the front and back side of the primitive. Then, when the * orientation of the facet is later learned, we can determine which * color (or index) to use for rendering. * * KW: We now know orientation in advance and only shade for * the side or sides which are actually required. * * Variables: * n = normal vector * V = vertex position * P = light source position * Pe = (0,0,0,1) * * Precomputed: * IF P[3]==0 THEN * // light at infinity * IF local_viewer THEN * _VP_inf_norm = unit vector from V to P // Precompute * ELSE * // eye at infinity * _h_inf_norm = Normalize( VP + <0,0,1> ) // Precompute * ENDIF * ENDIF * * Functions: * Normalize( v ) = normalized vector v * Magnitude( v ) = length of vector v */ static void validate_shine_table( struct gl_context *ctx, GLuint side, GLfloat shininess ) { TNLcontext *tnl = TNL_CONTEXT(ctx); struct tnl_shine_tab *list = tnl->_ShineTabList; struct tnl_shine_tab *s; ASSERT(side < 2); foreach(s, list) if ( s->shininess == shininess ) break; if (s == list) { GLint j; GLfloat *m; foreach(s, list) if (s->refcount == 0) break; m = s->tab; m[0] = 0.0; if (shininess == 0.0) { for (j = 1 ; j <= SHINE_TABLE_SIZE ; j++) m[j] = 1.0; } else { for (j = 1 ; j < SHINE_TABLE_SIZE ; j++) { GLdouble t, x = j / (GLfloat) (SHINE_TABLE_SIZE - 1); if (x < 0.005) /* underflow check */ x = 0.005; t = pow(x, shininess); if (t > 1e-20) m[j] = (GLfloat) t; else m[j] = 0.0; } m[SHINE_TABLE_SIZE] = 1.0; } s->shininess = shininess; } if (tnl->_ShineTable[side]) tnl->_ShineTable[side]->refcount--; tnl->_ShineTable[side] = s; move_to_tail( list, s ); s->refcount++; } void _tnl_validate_shine_tables( struct gl_context *ctx ) { TNLcontext *tnl = TNL_CONTEXT(ctx); GLfloat shininess; shininess = ctx->Light.Material.Attrib[MAT_ATTRIB_FRONT_SHININESS][0]; if (!tnl->_ShineTable[0] || tnl->_ShineTable[0]->shininess != shininess) validate_shine_table( ctx, 0, shininess ); shininess = ctx->Light.Material.Attrib[MAT_ATTRIB_BACK_SHININESS][0]; if (!tnl->_ShineTable[1] || tnl->_ShineTable[1]->shininess != shininess) validate_shine_table( ctx, 1, shininess ); } /** * In the case of colormaterial, the effected material attributes * should already have been bound to point to the incoming color data, * prior to running the pipeline. * This function copies the vertex's color to the material attributes * which are tracking glColor. * It's called per-vertex in the lighting loop. */ static void update_materials(struct gl_context *ctx, struct light_stage_data *store) { GLuint i; for (i = 0 ; i < store->mat_count ; i++) { /* update the material */ COPY_CLEAN_4V(store->mat[i].current, store->mat[i].size, store->mat[i].ptr); /* increment src vertex color pointer */ STRIDE_F(store->mat[i].ptr, store->mat[i].stride); } /* recompute derived light/material values */ _mesa_update_material( ctx, store->mat_bitmask ); /* XXX we should only call this if we're tracking/changing the specular * exponent. */ _tnl_validate_shine_tables( ctx ); } /** * Prepare things prior to running the lighting stage. * Return number of material attributes which will track vertex color. */ static GLuint prepare_materials(struct gl_context *ctx, struct vertex_buffer *VB, struct light_stage_data *store) { GLuint i; store->mat_count = 0; store->mat_bitmask = 0; /* Examine the _ColorMaterialBitmask to determine which materials * track vertex color. Override the material attribute's pointer * with the color pointer for each one. */ if (ctx->Light.ColorMaterialEnabled) { const GLuint bitmask = ctx->Light._ColorMaterialBitmask; for (i = 0 ; i < MAT_ATTRIB_MAX ; i++) if (bitmask & (1<AttribPtr[_TNL_ATTRIB_MAT_FRONT_AMBIENT + i] = VB->AttribPtr[_TNL_ATTRIB_COLOR0]; } /* Now, for each material attribute that's tracking vertex color, save * some values (ptr, stride, size, current) that we'll need in * update_materials(), above, that'll actually copy the vertex color to * the material attribute(s). */ for (i = _TNL_FIRST_MAT; i <= _TNL_LAST_MAT; i++) { if (VB->AttribPtr[i]->stride) { const GLuint j = store->mat_count++; const GLuint attr = i - _TNL_ATTRIB_MAT_FRONT_AMBIENT; store->mat[j].ptr = VB->AttribPtr[i]->start; store->mat[j].stride = VB->AttribPtr[i]->stride; store->mat[j].size = VB->AttribPtr[i]->size; store->mat[j].current = ctx->Light.Material.Attrib[attr]; store->mat_bitmask |= (1<mat_count; } /* * Compute dp ^ SpecularExponent. * Lerp between adjacent values in the f(x) lookup table, giving a * continuous function, with adequate overall accuracy. (Though still * pretty good compared to a straight lookup). */ static inline GLfloat lookup_shininess(const struct gl_context *ctx, GLuint face, GLfloat dp) { TNLcontext *tnl = TNL_CONTEXT(ctx); const struct tnl_shine_tab *tab = tnl->_ShineTable[face]; float f = dp * (SHINE_TABLE_SIZE - 1); int k = (int) f; if (k < 0 /* gcc may cast an overflow float value to negative int value */ || k > SHINE_TABLE_SIZE - 2) return powf(dp, tab->shininess); else return tab->tab[k] + (f - k) * (tab->tab[k+1] - tab->tab[k]); } /* Tables for all the shading functions. */ static light_func _tnl_light_tab[MAX_LIGHT_FUNC]; static light_func _tnl_light_fast_tab[MAX_LIGHT_FUNC]; static light_func _tnl_light_fast_single_tab[MAX_LIGHT_FUNC]; static light_func _tnl_light_spec_tab[MAX_LIGHT_FUNC]; #define TAG(x) x #define IDX (0) #include "t_vb_lighttmp.h" #define TAG(x) x##_twoside #define IDX (LIGHT_TWOSIDE) #include "t_vb_lighttmp.h" #define TAG(x) x##_material #define IDX (LIGHT_MATERIAL) #include "t_vb_lighttmp.h" #define TAG(x) x##_twoside_material #define IDX (LIGHT_TWOSIDE|LIGHT_MATERIAL) #include "t_vb_lighttmp.h" static void init_lighting_tables( void ) { static int done; if (!done) { init_light_tab(); init_light_tab_twoside(); init_light_tab_material(); init_light_tab_twoside_material(); done = 1; } } static GLboolean run_lighting( struct gl_context *ctx, struct tnl_pipeline_stage *stage ) { struct light_stage_data *store = LIGHT_STAGE_DATA(stage); TNLcontext *tnl = TNL_CONTEXT(ctx); struct vertex_buffer *VB = &tnl->vb; GLvector4f *input = ctx->_NeedEyeCoords ? VB->EyePtr : VB->AttribPtr[_TNL_ATTRIB_POS]; GLuint idx; if (!ctx->Light.Enabled || ctx->VertexProgram._Current) return GL_TRUE; /* Make sure we can talk about position x,y and z: */ if (input->size <= 2 && input == VB->AttribPtr[_TNL_ATTRIB_POS]) { _math_trans_4f( store->Input.data, VB->AttribPtr[_TNL_ATTRIB_POS]->data, VB->AttribPtr[_TNL_ATTRIB_POS]->stride, GL_FLOAT, VB->AttribPtr[_TNL_ATTRIB_POS]->size, 0, VB->Count ); if (input->size <= 2) { /* Clean z. */ _mesa_vector4f_clean_elem(&store->Input, VB->Count, 2); } if (input->size <= 1) { /* Clean y. */ _mesa_vector4f_clean_elem(&store->Input, VB->Count, 1); } input = &store->Input; } idx = 0; if (prepare_materials( ctx, VB, store )) idx |= LIGHT_MATERIAL; if (ctx->Light.Model.TwoSide) idx |= LIGHT_TWOSIDE; /* The individual functions know about replaying side-effects * vs. full re-execution. */ store->light_func_tab[idx]( ctx, VB, stage, input ); return GL_TRUE; } /* Called in place of do_lighting when the light table may have changed. */ static void validate_lighting( struct gl_context *ctx, struct tnl_pipeline_stage *stage ) { light_func *tab; if (!ctx->Light.Enabled || ctx->VertexProgram._Current) return; if (ctx->Light._NeedVertices) { if (ctx->Light.Model.ColorControl == GL_SEPARATE_SPECULAR_COLOR) tab = _tnl_light_spec_tab; else tab = _tnl_light_tab; } else { if (ctx->Light.EnabledList.next == ctx->Light.EnabledList.prev) tab = _tnl_light_fast_single_tab; else tab = _tnl_light_fast_tab; } LIGHT_STAGE_DATA(stage)->light_func_tab = tab; /* This and the above should only be done on _NEW_LIGHT: */ TNL_CONTEXT(ctx)->Driver.NotifyMaterialChange( ctx ); } /* Called the first time stage->run is called. In effect, don't * allocate data until the first time the stage is run. */ static GLboolean init_lighting( struct gl_context *ctx, struct tnl_pipeline_stage *stage ) { TNLcontext *tnl = TNL_CONTEXT(ctx); struct light_stage_data *store; GLuint size = tnl->vb.Size; stage->privatePtr = malloc(sizeof(*store)); store = LIGHT_STAGE_DATA(stage); if (!store) return GL_FALSE; /* Do onetime init. */ init_lighting_tables(); _mesa_vector4f_alloc( &store->Input, 0, size, 32 ); _mesa_vector4f_alloc( &store->LitColor[0], 0, size, 32 ); _mesa_vector4f_alloc( &store->LitColor[1], 0, size, 32 ); _mesa_vector4f_alloc( &store->LitSecondary[0], 0, size, 32 ); _mesa_vector4f_alloc( &store->LitSecondary[1], 0, size, 32 ); store->LitColor[0].size = 4; store->LitColor[1].size = 4; store->LitSecondary[0].size = 3; store->LitSecondary[1].size = 3; return GL_TRUE; } static void dtr( struct tnl_pipeline_stage *stage ) { struct light_stage_data *store = LIGHT_STAGE_DATA(stage); if (store) { _mesa_vector4f_free( &store->Input ); _mesa_vector4f_free( &store->LitColor[0] ); _mesa_vector4f_free( &store->LitColor[1] ); _mesa_vector4f_free( &store->LitSecondary[0] ); _mesa_vector4f_free( &store->LitSecondary[1] ); free( store ); stage->privatePtr = NULL; } } const struct tnl_pipeline_stage _tnl_lighting_stage = { "lighting", /* name */ NULL, /* private_data */ init_lighting, dtr, /* destroy */ validate_lighting, run_lighting };