network release 9
[carveJwlIkooP6JGAAIwe30JlM.git] / scene.h
diff --git a/scene.h b/scene.h
index d5fa2758b314b0bc19a4bb104df6c736a0efdbcc..d8fac1ad524a05672feabee916237ddfbeac8acf 100644 (file)
--- a/scene.h
+++ b/scene.h
-#include "vg/vg.h"
+#ifndef SCENE_H
+#define SCENE_H
+
+#include "common.h"
 #include "model.h"
+#include "bvh.h"
 
-GLuint tex_dual_noise;
+typedef struct scene_context scene_context;
+typedef struct scene_vert scene_vert;
 
-struct scene
-{
-   glmesh mesh;
+#pragma pack(push,1)
 
-   model_vert *verts;
-   u32 *indices;
+/* 32 byte vertexs, we don't care about the normals too much,
+ *    maybe possible to bring down uv to i16s too */
+struct scene_vert
+{
+   v3f co;        /* 3*32 */
+   v2f uv;        /* 2*32 */
+   i8  norm[4];   /* 4*8 */
+   u16 flags; /* only for the cpu. its junk on the gpu */
+   u16 unused[3];
+};
 
-   u32 vertex_count,
-       indice_count,
-       vertex_cap,
-       indice_cap;
+#pragma pack(pop)
 
-   boxf bbx;
+/* 
+ * 1. this should probably be a CONTEXT based approach unlike this mess.
+ * take a bit of the mdl_context ideas and redo this header. its messed up
+ * pretty bad right now.
+ */
 
-   struct shadower
-   {
-      sdf_primative sdf;
-      esdf_type sdf_type;
-   }
-   *shadowers;
+struct scene_context
+{
+   scene_vert *arrvertices;
+   u32 *arrindices;
 
-   u32 shadower_count,
-       shadower_cap;
+   u32 vertex_count, indice_count,
+       max_vertices, max_indices;
 
-   submodel submesh;
+   boxf bbx;
+   mdl_submesh submesh;
 };
 
-static void scene_init( scene *pscene )
+static u32 scene_mem_required( scene_context *ctx )
 {
-   pscene->verts = NULL;
-   pscene->indices = NULL;
-   pscene->vertex_count = 0;
-   pscene->indice_count = 0;
-   pscene->shadowers = NULL;
-   pscene->shadower_count = 0;
-   pscene->shadower_cap = 0;
-   pscene->submesh.indice_start = 0;
-   pscene->submesh.indice_count = 0;
-
-   v3_fill( pscene->bbx[0],  999999.9f );
-   v3_fill( pscene->bbx[1], -999999.9f );
-
-   static int noise_ready = 0;
-   if( !noise_ready )
-   {
-      noise_ready = 1;
-
-      u8 *buf = malloc( 256*256*2 );
-
-      for( int i=0; i<256*256; i++ )
-      {
-         u8 val = rand()&0xff;
-         buf[i*2] = val;
-      }
+   u32 vertex_length = vg_align8(ctx->max_vertices * sizeof(scene_vert)),
+       index_length  = vg_align8(ctx->max_indices  * sizeof(u32));
 
-      for( int y=0; y<256; y++ )
-      {
-         for( int x=0; x<256; x++ )
-         {
-            u8 *pr = &buf[(y*256+x)*2],
-               *pg = &buf[(((y+17)&0xff)*256+((x+37)&0xff))*2+1];
-            *pg = *pr;
-         }
-      }
-      
-      /* TODO: This texture should be delted somewhere */
-      glGenTextures( 1, &tex_dual_noise );
-      glBindTexture( GL_TEXTURE_2D, tex_dual_noise );
-      glTexImage2D( GL_TEXTURE_2D, 0, GL_RG, 256, 256, 0, GL_RG,
-            GL_UNSIGNED_BYTE, buf );
+   return vertex_length + index_length;
+}
 
-      vg_tex2d_linear();
-      vg_tex2d_repeat();
+static 
+void scene_init( scene_context *ctx, u32 max_vertices, u32 max_indices )
+{
+   ctx->vertex_count = 0;
+   ctx->indice_count = 0;
+   ctx->max_vertices = max_vertices;
+   ctx->max_indices = max_indices;
+   ctx->arrindices = NULL; /* must be filled out by user */
+   ctx->arrvertices = NULL;
 
-      free( buf );
-   }
+   memset( &ctx->submesh, 0, sizeof(mdl_submesh) );
+
+   v3_fill( ctx->bbx[0],  999999.9f );
+   v3_fill( ctx->bbx[1], -999999.9f );
 }
 
-/* https://www.shadertoy.com/view/4sfGzS */
-#define SHADER_VALUE_NOISE_3D                         \
-"uniform sampler2D uTexNoise;"                        \
-""                                                    \
-"float noise( vec3 x )"                               \
-"{"                                                   \
-   "vec3 i = floor(x);"                               \
-   "vec3 f = fract(x);"                               \
-   "f = f*f*(3.0-2.0*f);"                             \
-   "vec2 uv = (i.xy+vec2(37.0,17.0)*i.z) + f.xy;"     \
-   "vec2 rg = texture( uTexNoise, (uv+0.5)/256.0).yx;"\
-   "return mix( rg.x, rg.y, f.z );"                   \
-"}"                                                   \
-""                                                    \
-"const mat3 m = mat3( 0.00,  0.80,  0.60,"            \
-                    "-0.80,  0.36, -0.48,"            \
-                    "-0.60, -0.48,  0.64 );"          \
-""                                                    \
-"float fractalNoise( vec3 x )"                        \
-"{"                                                   \
-   "vec3 q = 8.0*x;"                                  \
-   "float f;"                                         \
-   "f  = 0.5000*noise( q ); q = m*q*2.01;"            \
-   "f += 0.2500*noise( q ); q = m*q*2.02;"            \
-   "f += 0.1250*noise( q ); q = m*q*2.03;"            \
-   "f += 0.0625*noise( q ); q = m*q*2.01;"            \
-   "return f;"                                        \
-"}"
-
-#define VERTEX_STANDARD_ATTRIBUTES                    \
-       "layout (location=0) in vec3 a_co;"                \
-   "layout (location=1) in vec3 a_norm;"              \
-   "layout (location=2) in vec4 a_colour;"            \
-   "layout (location=3) in vec2 a_uv;"
-
-SHADER_DEFINE( shader_debug_vcol,
-
-   /*Include*/ VERTEX_STANDARD_ATTRIBUTES
-
-       "uniform mat4 uPv;"
-   "uniform mat4 uMdl;"
-   "uniform float uTime;"
-   "uniform float uSwayAmt;"
-   ""
-   "out vec4 aColour;"
-   "out vec2 aUv;"
-   "out vec3 aNorm;"
-   "out vec3 aCo;"
-   ""
-   "vec3 compute_sway( vec3 pos )"
-   "{"
-      "vec4 sines = vec4( sin(uTime + pos.x)*1.0," 
-                         "sin(uTime*1.2 + pos.z*2.0)*1.1,"
-                         "sin(uTime*2.33)*0.5,"
-                         "sin(uTime*0.6 + pos.x*0.3)*1.3 );"
-
-      "vec3 offset = vec3( sines.x+sines.y*sines.w, 0.0, sines.x+sines.z );"
-      "return pos + offset*a_colour.r*uSwayAmt;"
-   "}"
-       ""
-       "void main()"
-       "{"
-      "vec3 swaypos = compute_sway( a_co );"
-               "gl_Position = uPv * uMdl * vec4( swaypos, 1.0 );"
-      "aColour = a_colour;"
-      "aUv = a_uv;"
-      "aNorm = normalize(mat3(uMdl) * a_norm);"
-      "aCo = a_co;"
-       "}",
-   /* Fragment */
-       "out vec4 FragColor;"
-       ""
-   "uniform int uMode;"
-   "uniform sampler2D uTexMain;"
-   "uniform sampler2D uTexGradients;"
-   ""
-   /*Include*/ SHADER_VALUE_NOISE_3D
-   ""
-   "in vec4 aColour;"
-   "in vec2 aUv;"
-   "in vec3 aNorm;"
-   "in vec3 aCo;"
-   ""
-       "void main()"
-       "{"
-      "vec4 colour = vec4(1.0,0.0,0.5,1.0);"
-      "vec4 diffuse = texture( uTexMain, aUv );"
-
-      "if( uMode == 1 )"
-      "{"
-         "colour = vec4(aNorm * 0.5 + 0.5, 1.0);"
-      "}"
-      "if( uMode == 2 )"
-      "{"
-         "colour = aColour;"
-      "}"
-      "if( uMode == 3 )"
-      "{"
-         "float light = dot(aNorm, vec3(0.2,0.8,0.1));"
-         "vec3 grid3 = fract(aCo);"
-         
-         "colour = vec4(vec3(light)*(1.0-grid3*0.3),1.0);"
-      "}"
-      "if( uMode == 4 )"
-      "{"
-         "colour = vec4( aUv, 0.0, 1.0 );"
-      "}"
-      "if( uMode == 5 )"
-      "{"
-         "if( diffuse.a < 0.45 ) discard;"
-         "colour = diffuse;"
-      "}"
-      "if( uMode == 6 )"
-      "{"
-         "float r1 = fractalNoise(aCo);"
-         "colour = vec4( vec3(r1), 1.0 );"
-      "}"
-      "if( uMode == 7 )"
-      "{"
-         "if( diffuse.a < 0.2 ) discard;"
-         "float lighting = 1.0 - aColour.g*0.8;"
-   
-         "float light1 = max(0.0,dot(-vec3(0.5,-0.8,0.25), aNorm));"
-         "float light2 = max(0.0,dot(-vec3(-0.8,0.5,-0.25), aNorm));"
-         "vec3 lt = vec3(0.2,0.2,0.2 ) + " 
-                   "vec3(1.0,1.0,0.9)*light1 + "
-                   "vec3(0.1,0.3,0.4 )*light2;"
-
-
-         "colour = vec4(vec3(pow(lighting,1.6)*(diffuse.r*0.7+0.5)),1.0);"
-         "colour = vec4(colour.rgb*lt,1.0);"
-
-         "vec2 gradUV = vec2(lighting*1.9,aColour.b*0.8);"
-         "vec4 gradient_sample = texture( uTexGradients, gradUV );"
-         "colour = colour*gradient_sample;"
-      "}"
-      "if( uMode == 8 )"
-      "{"
-         "if( diffuse.a < 0.45 ) discard;"
-         "float light = 1.0 - aColour.g;"
-         "light = pow(light,1.6)*(diffuse.r*0.7+0.5);"
-         "float r1 = fractalNoise(aCo*0.01);"
-         
-         "vec2 gradUV = vec2(light*1.9,r1+aColour.b);"
-         "vec4 gradient_sample = texture( uTexGradients, gradUV );"
-         "colour = gradient_sample*light;"
-      "}"
-
-               "FragColor = colour;"
-       "}"
-       ,
-       UNIFORMS({ "uPv", "uMode", "uTexMain", "uTexGradients", "uTexNoise", \
-              "uTime", "uSwayAmt", "uMdl" })
-)
-
-SHADER_DEFINE( shader_standard_lit,
-
-   /*Include*/ VERTEX_STANDARD_ATTRIBUTES
-
-       "uniform mat4 uPv;"
-   "uniform mat4 uMdl;"
-   ""
-   "out vec4 aColour;"
-   "out vec2 aUv;"
-   "out vec3 aNorm;"
-   "out vec3 aCo;"
-   ""
-       "void main()"
-       "{"
-               "gl_Position = uPv * uMdl * vec4( a_co, 1.0 );"
-      "aColour = a_colour;"
-      "aUv = a_uv;"
-      "aNorm = mat3(uMdl) * a_norm;"
-      "aCo = a_co;"
-       "}",
-   /* Fragment */
-       "out vec4 FragColor;"
-       ""
-   "uniform sampler2D uTexMain;"
-   "uniform vec4 uColour;"
-   ""
-   "in vec4 aColour;"
-   "in vec2 aUv;"
-   "in vec3 aNorm;"
-   "in vec3 aCo;"
-   ""
-       "void main()"
-       "{"
-      "vec3 diffuse = texture( uTexMain, aUv ).rgb;"
-
-      "float light1 = max(0.0,dot(-vec3(0.5,-0.8,0.25), aNorm));"
-      "float light2 = max(0.0,dot(-vec3(-0.8,0.5,-0.25), aNorm));"
-      "diffuse += vec3(0.2,0.2,0.2 ) + " 
-                 "vec3(1.0,1.0,0.9)*light1 + "
-                 "vec3(0.1,0.3,0.4 )*light2;"
-
-               "FragColor = vec4((diffuse*uColour.rgb),"
-                        "aColour.a*uColour.a);"
-       "}"
-       ,
-       UNIFORMS({ "uColour","uTexMain","uPv","uMdl" })
-)
-
-SHADER_DEFINE( shader_unlit,
-
-   /*Include*/ VERTEX_STANDARD_ATTRIBUTES
-
-       "uniform mat4 uPv;"
-   "uniform mat4 uMdl;"
-   ""
-   "out vec4 aColour;"
-   "out vec2 aUv;"
-   "out vec3 aNorm;"
-   "out vec3 aCo;"
-   ""
-       "void main()"
-       "{"
-               "gl_Position = uPv * uMdl * vec4( a_co, 1.0 );"
-      "aColour = a_colour;"
-      "aUv = a_uv;"
-      "aNorm = mat3(uMdl) * a_norm;"
-      "aCo = a_co;"
-       "}",
-   /* Fragment */
-       "out vec4 FragColor;"
-       ""
-   "uniform sampler2D uTexMain;"
-   "uniform vec4 uColour;"
-   ""
-   "in vec4 aColour;"
-   "in vec2 aUv;"
-   "in vec3 aNorm;"
-   "in vec3 aCo;"
-   ""
-       "void main()"
-       "{"
-      "vec3 diffuse = texture( uTexMain, aUv ).rgb;"
-      "FragColor = vec4(pow(diffuse,vec3(1.0)),1.0);"
-       "}"
-       ,
-       UNIFORMS({ "uTexMain", "uPv", "uMdl" })
-)
-
-static void *buffer_reserve( void *buffer, u32 count, u32 *cap, u32 amount, 
-      size_t emsize )
+void scene_supply_buffer( scene_context *ctx, void *buffer )
 {
-   if( count+amount > *cap )
-   {
-      *cap = VG_MAX( (*cap)*2, (*cap)+amount );
-      
-      return realloc( buffer, (*cap) * emsize );
-   }
+   u32 vertex_length = vg_align8( ctx->max_vertices * sizeof(scene_vert) );
+
+   ctx->arrvertices = buffer;
+   ctx->arrindices  = (u32*)(((u8*)buffer) + vertex_length);
+}
 
-   return buffer;
+static void scene_vert_pack_norm( scene_vert *vert, v3f norm, f32 blend ){
+   v3f n;
+   v3_muls( norm, 127.0f, n );
+   v3_minv( n, (v3f){  127.0f,  127.0f,  127.0f }, n );
+   v3_maxv( n, (v3f){ -127.0f, -127.0f, -127.0f }, n );
+   vert->norm[0] = n[0];
+   vert->norm[1] = n[1];
+   vert->norm[2] = n[2];
+   vert->norm[3] = blend * 127.0f;
 }
 
 /* 
  * Append a model into the scene with a given transform
  */
-static void scene_add_model( scene *pscene, model *mdl, submodel *submodel,
-      v3f pos, float yaw, float scale )
+static void scene_add_mdl_submesh( scene_context *ctx, mdl_context *mdl, 
+                                      mdl_submesh *sm, m4x3f transform )
 {
-   pscene->verts = buffer_reserve( pscene->verts, pscene->vertex_count, 
-         &pscene->vertex_cap, submodel->vertex_count, sizeof(model_vert) );
-   pscene->indices = buffer_reserve( pscene->indices, pscene->indice_count,
-         &pscene->indice_cap, submodel->indice_count, sizeof(u32) );
-
-   if( submodel->sdf_type )
-   {
-      pscene->shadowers = buffer_reserve( pscene->shadowers, 
-            pscene->shadower_count, &pscene->shadower_cap, 1,
-            sizeof( struct shadower ));
-      
-      struct shadower *shadower = 
-         &pscene->shadowers[ pscene->shadower_count ++ ];
+   if( ctx->vertex_count + sm->vertex_count > ctx->max_vertices ){
+      vg_fatal_error( "Scene vertex buffer overflow (%u exceeds %u)\n",
+                        ctx->vertex_count + sm->vertex_count, 
+                        ctx->max_vertices );
+   }
 
-      shadower->sdf = submodel->sdf;
-      shadower->sdf_type = submodel->sdf_type;
-      
-      v2_muls( shadower->sdf.info, scale, shadower->sdf.info );
-      v3_muls( shadower->sdf.origin, scale, shadower->sdf.origin );
-      v3_add( pos, shadower->sdf.origin, shadower->sdf.origin );
+   if( ctx->indice_count + sm->indice_count > ctx->max_indices ){
+      vg_fatal_error( "Scene index buffer overflow (%u exceeds %u)\n",
+                        ctx->indice_count + sm->indice_count,
+                        ctx->max_indices );
    }
+
+   mdl_vert   *src_verts = mdl_arritm( &mdl->verts, sm->vertex_start );
+   scene_vert *dst_verts = &ctx->arrvertices[ ctx->vertex_count ];
+
+   u32 *src_indices    =  mdl_arritm( &mdl->indices, sm->indice_start ),
+       *dst_indices    = &ctx->arrindices[ ctx->indice_count ];
    
    /* Transform and place vertices */
-   model_vert *src_verts = submodel_vert_data( mdl, submodel );
-   u32 *src_indices = submodel_indice_data( mdl, submodel );
-   
-   m4x3f mtx;
-   m4x3_identity( mtx );
-   m4x3_translate( mtx, pos );
-   m4x3_rotate_y( mtx, yaw );
-   m4x3_scale( mtx, scale );
-
    boxf bbxnew;
-   box_copy( submodel->bbx, bbxnew );
-   m4x3_transform_aabb( mtx, bbxnew );
-   box_concat( pscene->bbx, bbxnew );
+   box_init_inf( bbxnew );
+   m4x3_expand_aabb_aabb( transform, bbxnew, sm->bbx );
+   box_concat( ctx->bbx, bbxnew );
+
+   m3x3f normal_matrix;
+   m3x3_copy( transform, normal_matrix );
+   v3_normalize( normal_matrix[0] );
+   v3_normalize( normal_matrix[1] );
+   v3_normalize( normal_matrix[2] );
    
-   m3x3f rotation;
-   m4x3_to_3x3( mtx, rotation );
-
-   float rand_hue = vg_randf();
+   for( u32 i=0; i<sm->vertex_count; i++ ){
+      mdl_vert   *src   = &src_verts[i];
+      scene_vert *pvert = &dst_verts[i];
 
-   for( u32 i=0; i<submodel->vertex_count; i++ )
-   {
-      model_vert *pvert = &pscene->verts[ pscene->vertex_count+i ],
-                 *src = &src_verts[ i ];
+      m4x3_mulv( transform, src->co, pvert->co );
 
-      m4x3_mulv( mtx, src->co, pvert->co );
-      m3x3_mulv( rotation, src->norm, pvert->norm );
-
-      v4_copy( src->colour, pvert->colour );
-      v2_copy( src->uv, pvert->uv );
+      v3f normal;
+      m3x3_mulv( normal_matrix, src->norm, normal );
+      scene_vert_pack_norm( pvert, normal, src->colour[0]*(1.0f/255.0f) );
       
-      float rel_y = src->co[1] / submodel->bbx[1][1];
-      pvert->colour[0] = rel_y;
-      pvert->colour[2] = rand_hue;
+      v2_copy( src->uv, pvert->uv );
    }
 
-   for( u32 i=0; i<submodel->indice_count; i++ )
-   {
-      u32 *pidx = &pscene->indices[ pscene->indice_count+i ];
-      *pidx = src_indices[i] + pscene->vertex_count;
-   }
+   u32 real_indices = 0;
+   for( u32 i=0; i<sm->indice_count/3; i++ ){
+      u32 *src = &src_indices[i*3],
+          *dst = &dst_indices[real_indices];
 
-   pscene->vertex_count += submodel->vertex_count;
-   pscene->indice_count += submodel->indice_count;
-}
+      v3f ab, ac, tn;
+      v3_sub( src_verts[src[2]].co, src_verts[src[0]].co, ab );
+      v3_sub( src_verts[src[1]].co, src_verts[src[0]].co, ac );
+      v3_cross( ac, ab, tn );
 
-static void scene_copy_slice( scene *pscene, submodel *sm )
-{
-   sm->indice_start = pscene->submesh.indice_start;
-   sm->indice_count = pscene->indice_count - sm->indice_start;
-   
-   pscene->submesh.indice_start = pscene->indice_count;
-}
-
-static void scene_shadow_sphere( scene *pscene, v3f sphere, 
-      v4f params, v3f lightdir )
-{
-   for( int i=0; i<pscene->vertex_count; i++ )
-   {
-      model_vert *vert = &pscene->verts[i];
+#if 0
+      if( v3_length2( tn ) <= 0.00001f )
+         continue;
+#endif
 
-      v3f delta;
-      v3_sub( sphere, vert->co, delta );
+      dst[0] = src[0] + ctx->vertex_count;
+      dst[1] = src[1] + ctx->vertex_count;
+      dst[2] = src[2] + ctx->vertex_count;
 
-      float d = v3_dot( lightdir, delta );
-      v3f closest;
+      real_indices += 3;
+   }
 
-      v3_muls( lightdir, d, closest );
-      float dist = v3_dist( closest, delta ),
-            shading = vg_maxf( dist - params[0], 0.0f );
+   if( real_indices != sm->indice_count )
+      vg_warn( "Zero area triangles in model\n" );
 
-      shading = vg_minf( shading * params[1], 1.0f );
-      vert->colour[1] *= shading;
-   }
+   ctx->vertex_count += sm->vertex_count;
+   ctx->indice_count += real_indices;
 }
 
-static void scene_shadow_gradient( scene *pscene, int comp,
-      float start, float length )
+/*
+ * One by one adders for simplified access (mostly procedural stuff)
+ */
+static void scene_push_tri( scene_context *ctx, u32 tri[3] )
 {
-   float scale = 1.0f / length;
+   if( ctx->indice_count + 3 > ctx->max_indices )
+      vg_fatal_error( "Scene indice buffer overflow (%u exceeds %u)\n",
+                        ctx->indice_count+3, ctx->max_indices );
 
-   for( int i=0; i<pscene->vertex_count; i++ )
-   {
-      model_vert *vert = &pscene->verts[i];
-      float shading = start + vert->co[comp] * scale;
+   u32 *dst = &ctx->arrindices[ ctx->indice_count ];
 
-      vert->colour[1] = shading;
-   }
+   dst[0] = tri[0];
+   dst[1] = tri[1];
+   dst[2] = tri[2];
+
+   ctx->indice_count += 3;
 }
 
-/* Temporary */
-static int sample_scene_height( scene *pscene, v3f pos, v3f norm )
+static void scene_push_vert( scene_context *ctx, scene_vert *v )
 {
-   for( int i=0; i<pscene->indice_count/3; i++ )
-   {
-      u32 *tri = &pscene->indices[i*3];
+   if( ctx->vertex_count + 1 > ctx->max_vertices )
+      vg_fatal_error( "Scene vertex buffer overflow (%u exceeds %u)\n",
+                        ctx->vertex_count+1, ctx->max_vertices );
 
-      float *pA = pscene->verts[tri[0]].co,
-            *pB = pscene->verts[tri[1]].co,
-            *pC = pscene->verts[tri[2]].co;
-      
-      float height;
-      if( triangle_raycast( pA, pB, pC, pos, &height ))
-      {
-         pos[1] = height;
-         
-         if( norm )
-         {
-            v3f v0, v1;
-            v3_sub( pA, pB, v0 );
-            v3_sub( pC, pB, v1 );
-            v3_cross( v1, v0, norm );
-            v3_normalize( norm );
-         }
+   scene_vert *dst = &ctx->arrvertices[ ctx->vertex_count ];
+   *dst = *v;
 
-         return 1;
-      }
-   }
-   return 0;
+   ctx->vertex_count ++;
 }
 
-static void sample_scene_normal( scene *pscene, v3f pos, v3f normal )
+static void scene_copy_slice( scene_context *ctx, mdl_submesh *sm )
 {
-   for( int i=0; i<pscene->indice_count/3; i++ )
-   {
-      u32 *tri = &pscene->indices[i*3];
-      
-      float height;
-      if( triangle_raycast( 
-            pscene->verts[ tri[0] ].co,
-            pscene->verts[ tri[1] ].co,
-            pscene->verts[ tri[2] ].co, pos, &height ))
-      {
-         v3f v0, v1;
-
-         v3_sub( pscene->verts[ tri[1] ].co,
-                 pscene->verts[ tri[0] ].co,
-                 v0 );
-
-         v3_sub( pscene->verts[ tri[2] ].co,
-                 pscene->verts[ tri[0] ].co,
-                 v1 );
-
-         v3_cross( v0, v1, normal );
-         v3_normalize( normal );
-         return;
-      }
-   }
+   sm->indice_start = ctx->submesh.indice_start;
+   sm->indice_count = ctx->indice_count - sm->indice_start;
 
-   normal[0] = 0.0f;
-   normal[1] = 1.0f;
-   normal[2] = 0.0f;
+   sm->vertex_start = ctx->submesh.vertex_start;
+   sm->vertex_count = ctx->vertex_count - sm->vertex_start;
+   
+   ctx->submesh.indice_start = ctx->indice_count;
+   ctx->submesh.vertex_start = ctx->vertex_count;
 }
 
-/* 
- * Experimental SDF based shadows
- *
- * https://iquilezles.org/articles/distfunctions/
- */
-static float sd_cone( v3f co, sdf_primative *prim )
-{
-   float bound = prim->info[1]*1.75f;
-   if( v3_dist2( prim->origin, co ) > bound*bound )
-      return 999999.9f;
+static void scene_set_vertex_flags( scene_context *ctx, 
+                                       u32 start, u32 count, u16 flags ){
+   for( u32 i=0; i<count; i++ )
+      ctx->arrvertices[ start + i ].flags = flags;
+}
 
-   v3f p;
-   v3_sub( co, prim->origin, p );
-   
-   float h = prim->info[1];
-   v2f c = { prim->info[2], prim->info[3] };
+struct scene_upload_info{
+   scene_context *ctx;
+   glmesh *mesh;
+};
 
-   v2f q, w, a, b;
-   v2_muls( (v2f){ c[0]/c[1], -1.0f }, h, q );
-   
-   w[0] = v2_length( (v2f){ p[0], p[2] } );
-   w[1] = p[1];
+static void async_scene_upload( void *payload, u32 size )
+{
+   struct scene_upload_info *info = payload;
 
-   v2_muladds( w, q, -vg_clampf( v2_dot(w,q)/v2_dot(q,q), 0.0f, 1.0f ), a );
-   v2_muladd( w, q, (v2f){ vg_clampf( w[0]/q[0], 0.0f, 1.0f ), 1.0f }, b );
+   //assert( mesh->loaded == 0 );
+   
+   glmesh *mesh = info->mesh;
+   scene_context *ctx = info->ctx;
 
-   float k = vg_signf( q[1] ),
-         d = vg_minf( v2_dot( a,a ), v2_dot( b,b ) ),
-         s = vg_maxf( k*(w[0]*q[1]-w[1]*q[0]), k*(w[1]-q[1]) );
+   glGenVertexArrays( 1, &mesh->vao );
+   glGenBuffers( 1, &mesh->vbo );
+   glGenBuffers( 1, &mesh->ebo );
+   glBindVertexArray( mesh->vao );
 
-   return sqrtf(d)*vg_signf(s);
-}
+   size_t stride = sizeof(scene_vert);
 
-#define CACHE_AMBIENT_SHAPES
+   glBindBuffer( GL_ARRAY_BUFFER, mesh->vbo );
+   glBufferData( GL_ARRAY_BUFFER, ctx->vertex_count*stride, 
+                 ctx->arrvertices, GL_STATIC_DRAW );
 
-static float scene_ambient_sample( scene *pscene, v3f pos, v3f dir )
-{
-   float accum = 0.0f;
-
-#ifdef CACHE_AMBIENT_SHAPES
-   static struct shadower *local_shadowers[32];
-   static int local_shadower_count = 0;
-   static v3f local_shadower_last = { -99999.9f, -999999.9f, -9999999.9f };
-
-   if( v3_dist2( pos, local_shadower_last ) > 10.0f*10.0f )
-   {
-      local_shadower_count = 0;
-      v3_copy( pos, local_shadower_last );
-
-      for( int k=0; k<pscene->shadower_count; k++ )
-      {
-         struct shadower *shadower = &pscene->shadowers[k];
-
-         if( sd_cone( pos, &shadower->sdf ) <= 20.0f )
-         {
-            local_shadowers[ local_shadower_count ++ ] = shadower;
-            if( local_shadower_count == vg_list_size( local_shadowers ) )
-               break;
-         }
-      }
-   }
-#endif
+   glBindVertexArray( mesh->vao );
+   glBindBuffer( GL_ELEMENT_ARRAY_BUFFER, mesh->ebo );
+   glBufferData( GL_ELEMENT_ARRAY_BUFFER, ctx->indice_count*sizeof(u32),
+                 ctx->arrindices, GL_STATIC_DRAW );
+   
+   /* 0: coordinates */
+   glVertexAttribPointer( 0, 3, GL_FLOAT, GL_FALSE, stride, (void*)0 );
+   glEnableVertexAttribArray( 0 );
 
-   for( int j=0; j<5; j++ )
-   {
-      v3f tracepos;
-      v3_muladds( pos, dir, 1.5f*(float)j, tracepos );
+   /* 1: normal */
+   glVertexAttribPointer( 1, 4, GL_BYTE, GL_TRUE,
+         stride, (void *)offsetof(scene_vert, norm) );
+   glEnableVertexAttribArray( 1 );
 
-      float mindist = 99999.9f;
+   /* 2: uv */
+   glVertexAttribPointer( 2, 2, GL_FLOAT, GL_FALSE, 
+         stride, (void *)offsetof(scene_vert, uv) );
+   glEnableVertexAttribArray( 2 );
 
-#ifndef CACHE_AMBIENT_SHAPES
+   VG_CHECK_GL_ERR();
 
-      for( int k=0; k<pscene->shadower_count; k++ ){
-         struct shadower *shadower = &pscene->shadowers[k];
-#else
+   mesh->indice_count = ctx->indice_count;
+   mesh->loaded = 1;
 
-      for( int k=0; k<local_shadower_count; k++ ){
-         struct shadower *shadower = local_shadowers[k];
-#endif
+   vg_info( "Scene upload ( XYZ_f32 UV_f32 XYZW_i8 )[ u32 ]\n" );
+   vg_info( "   indices:%u\n", ctx->indice_count );
+   vg_info( "   verts:%u\n",   ctx->vertex_count );
+}
 
-         float dist = vg_maxf( 0.0f, sd_cone( tracepos, &shadower->sdf ));
-         mindist = vg_minf( mindist, dist );
-      }
+static void scene_upload_async( scene_context *ctx, glmesh *mesh )
+{
+   vg_async_item *call = vg_async_alloc( sizeof(struct scene_upload_info) );
 
+   struct scene_upload_info *info = call->payload;
+   info->mesh = mesh;
+   info->ctx = ctx;
 
-      accum += vg_clampf( 1.0f - mindist, 0.0f, 1.0f )*0.2f;
-   }
-   
-   return accum;
+   vg_async_dispatch( call, async_scene_upload );
 }
 
-#define DYNAMIC_GRID
-#define JUST_DO_EVERY_VERT
-
-static void scene_compute_occlusion( scene *pscene )
+static
+vg_async_item *scene_alloc_async( scene_context *scene, glmesh *mesh,
+                                  u32 max_vertices, u32 max_indices )
 {
-   v3f sundir = { 0.2f, 0.9f, 0.2f };
-   v3_normalize( sundir );
-
-   /* TODO: Make this sample grid be dynamically required.
-    *
-    *   1. Only resample the light grid (1x1x1), when a vertex is outside the 
-    *      current cube
-    *
-    *   2. Reorder all vertices so that each group of vertices that fit in a 
-    *      cube are next to eachother in the buffer. This will save cache
-    *      misses.
-    *
-    *      for the sorting algorithm, i think we can already assume that *most
-    *      vertices will be quite close to eachother. so instead of doing an
-    *      exhaustive search we can reorder 1k chunks at a time.
-    */
-
-   v3f sample_area;
-   v3_sub( pscene->bbx[1], pscene->bbx[0], sample_area );
-   v3_ceil( sample_area, sample_area );
-   int ax = sample_area[0],
-       ay = sample_area[1],
-       az = sample_area[2];
-
-#ifndef DYNAMIC_GRID
-   float *samplegrid = malloc( ax*ay*az* sizeof(float) );
-
-   for( int x=0; x<ax; x++ ){
-   for( int y=0; y<ay; y++ ){
-   for( int z=0; z<az; z++ )
-   {
-      v3f sample_pos = { x,y,z };
-      v3_add( pscene->bbx[0], sample_pos, sample_pos );
-      float accum = scene_ambient_sample( pscene, sample_pos, sundir );
-
-      samplegrid[x + y*ax + z*ax*ay] = accum;
-   }}}
-#else
-   v3i cube_pos = { -999999, -999999, -999999 };
-   int cube_resamples = 0, hits = 0, misses = 0;
-
-   float s0=0.0f,s1=0.0f,s2=0.0f,s3=0.0f,s4=0.0f,s5=0.0f,s6=0.0f,s7=0.0f;
-#endif
+   scene_init( scene, max_vertices, max_indices );
+   u32 buf_size = scene_mem_required( scene );
 
-   for( int i=0; i<pscene->vertex_count; i++ )
-   {
-      model_vert *vert = &pscene->verts[i];
-      v3f rel, q;
-      
-#ifndef DYNAMIC_GRID
-      v3_sub( vert->co, pscene->bbx[0], q );
-#else
-      v3_copy( vert->co, q );
-#endif
+   u32 hdr_size = vg_align8(sizeof(struct scene_upload_info));
+   vg_async_item *call = vg_async_alloc( hdr_size + buf_size );
 
-      v3_floor( q, rel );
-      v3_sub( q, rel, q );
-      
-      int x=rel[0],
-          y=rel[1],
-          z=rel[2];
-
-#ifndef JUST_DO_EVERY_VERT
-#ifndef DYNAMIC_GRID
-      x = VG_MIN(x,ax-2);
-      y = VG_MIN(y,ay-2);
-      z = VG_MIN(z,az-2);
-      x = VG_MAX(x,0);
-      y = VG_MAX(y,0);
-      z = VG_MAX(z,0);
-
-      float 
-         s0 = samplegrid[ x    +  y*ax    + z*ax*ay],
-         s1 = samplegrid[(x+1) +  y*ax    + z*ax*ay],
-         s2 = samplegrid[ x    + (y+1)*ax + z*ax*ay],
-         s3 = samplegrid[(x+1) + (y+1)*ax + z*ax*ay],
-         s4 = samplegrid[ x    +  y*ax    + (z+1)*ax*ay],
-         s5 = samplegrid[(x+1) +  y*ax    + (z+1)*ax*ay],
-         s6 = samplegrid[ x    + (y+1)*ax + (z+1)*ax*ay],
-         s7 = samplegrid[(x+1) + (y+1)*ax + (z+1)*ax*ay],
-#else
-      if( x!=cube_pos[0] || y!=cube_pos[1] || z!=cube_pos[2] )
-      {
-         cube_pos[0] = x;
-         cube_pos[1] = y;
-         cube_pos[2] = z;
-
-         s0 = scene_ambient_sample( pscene, (v3f){ x,y,z }, sundir );
-         s1 = scene_ambient_sample( pscene, (v3f){ x+1,y,z }, sundir );
-         s2 = scene_ambient_sample( pscene, (v3f){ x,y+1,z }, sundir );
-         s3 = scene_ambient_sample( pscene, (v3f){ x+1,y+1,z }, sundir );
-         s4 = scene_ambient_sample( pscene, (v3f){ x,y,z+1 }, sundir );
-         s5 = scene_ambient_sample( pscene, (v3f){ x+1,y,z+1 }, sundir );
-         s6 = scene_ambient_sample( pscene, (v3f){ x,y+1,z+1 }, sundir );
-         s7 = scene_ambient_sample( pscene, (v3f){ x+1,y+1,z+1 }, sundir );
-
-         cube_resamples += 8;
-         misses ++;
-      }
-      else
-         hits ++;
+   struct scene_upload_info *info = call->payload;
 
-      float
-#endif
+   info->mesh = mesh;
+   info->ctx = scene;
 
-         s0_s1 = vg_lerpf( s0, s1, q[0] ),
-         s2_s3 = vg_lerpf( s2, s3, q[0] ),
-         s4_s5 = vg_lerpf( s4, s5, q[0] ),
-         s6_s7 = vg_lerpf( s6, s7, q[0] ),
+   void *buffer = ((u8*)call->payload)+hdr_size;
+   scene_supply_buffer( scene, buffer );
 
-         s0s1_s2s3 = vg_lerpf( s0_s1, s2_s3, q[1] ),
-         s4s5_s6s7 = vg_lerpf( s4_s5, s6_s7, q[1] ),
-         s0s1s2s3_s4s5s6s7 = vg_lerpf( s0s1_s2s3, s4s5_s6s7, q[2] );
+   return call;
+}
 
-      vert->colour[1] = s0s1s2s3_s4s5s6s7;
-#else
-      vert->colour[1] = scene_ambient_sample( pscene, vert->co, sundir );
-#endif
-   }
 
-#ifndef DYNAMIC_GRID
-   int cube_resamples = -1, misses = 0, hits = 0;
-#endif
+/*
+ * BVH implementation
+ */
+
+static void scene_bh_expand_bound( void *user, boxf bound, u32 item_index )
+{
+   scene_context *s = user;
+   scene_vert *pa = &s->arrvertices[ s->arrindices[item_index*3+0] ],
+              *pb = &s->arrvertices[ s->arrindices[item_index*3+1] ],
+              *pc = &s->arrvertices[ s->arrindices[item_index*3+2] ];
+   
+  box_addpt( bound, pa->co );
+  box_addpt( bound, pb->co );
+  box_addpt( bound, pc->co );
+}
 
-   int static_samples = ax*ay*az,
-       vertex_samples = pscene->vertex_count;
+static float scene_bh_centroid( void *user, u32 item_index, int axis )
+{
+   scene_context *s = user;
+   scene_vert *pa = &s->arrvertices[ s->arrindices[item_index*3+0] ],
+              *pb = &s->arrvertices[ s->arrindices[item_index*3+1] ],
+              *pc = &s->arrvertices[ s->arrindices[item_index*3+2] ];
 
-   if( cube_resamples < static_samples )
-      vg_success( "Walking cube beat static grid (%d<%d. %d)!\n",
-         cube_resamples, static_samples, vertex_samples );
-   else
-      vg_warn( "Walking cube was worse than static grid (%d<%d. %d).\n",
-         cube_resamples, static_samples, vertex_samples );
+   #if 0
 
-   vg_info( "Hits; %d, misses: %d\n", hits, misses );
+   float min, max;
 
-#ifndef DYNAMIC_GRID
-   free( samplegrid );
-#endif
+   min = vg_minf( pa->co[axis], pb->co[axis] );
+   max = vg_maxf( pa->co[axis], pb->co[axis] );
+   min = vg_minf( min, pc->co[axis] );
+   max = vg_maxf( max, pc->co[axis] );
 
-   return;
+   return (min+max) * 0.5f;
 
-   for( int i=0; i<pscene->vertex_count; i++ )
-   {
-      model_vert *vert = &pscene->verts[i];
-      float accum = 0.0f;
+   #else
+   return (pa->co[axis] + pb->co[axis] + pc->co[axis]) * (1.0f/3.0f);
+   #endif
+}
 
-      for( int j=0; j<5; j++ )
-      {
-         v3f tracepos;
-         v3_copy( vert->co, tracepos );
-         v3_muladds( tracepos, sundir, 1.5f*(float)j, tracepos );
+static void scene_bh_swap( void *user, u32 ia, u32 ib )
+{
+   scene_context *s = user;
 
-         float mindist = 99999.9f;
+   u32 *ti = &s->arrindices[ia*3];
+   u32 *tj = &s->arrindices[ib*3];
 
-         for( int k=0; k<pscene->shadower_count; k++ )
-         {
-            struct shadower *shadower = &pscene->shadowers[k];
-            float dist = vg_maxf( 0.0f, sd_cone( tracepos, &shadower->sdf ));
-            mindist = vg_minf( mindist, dist );
-         }
+   u32 temp[3];
+   temp[0] = ti[0];
+   temp[1] = ti[1];
+   temp[2] = ti[2];
 
-         accum += vg_clampf( 1.0f - mindist, 0.0f, 1.0f )*0.2f;
-      }
+   ti[0] = tj[0];
+   ti[1] = tj[1];
+   ti[2] = tj[2];
 
-      vert->colour[1] = vg_minf( accum, 1.0f );
-   }
+   tj[0] = temp[0];
+   tj[1] = temp[1];
+   tj[2] = temp[2];
 }
 
-static void scene_upload( scene *pscene )
+static void scene_bh_debug( void *user, u32 item_index )
 {
-   mesh_upload( &pscene->mesh,
-         pscene->verts, pscene->vertex_count,
-         pscene->indices, pscene->indice_count );
-
-   vg_info( "Scene upload\n" );
-   vg_info( "   indices:%u\n", pscene->indice_count );
-   vg_info( "   verts:%u\n", pscene->vertex_count );
+   scene_context *s = user;
+   u32 idx = item_index*3;
+   scene_vert *pa = &s->arrvertices[ s->arrindices[ idx+0 ] ],
+              *pb = &s->arrvertices[ s->arrindices[ idx+1 ] ],
+              *pc = &s->arrvertices[ s->arrindices[ idx+2 ] ];
+
+   vg_line( pa->co, pb->co, 0xff0000ff );
+   vg_line( pb->co, pc->co, 0xff0000ff );
+   vg_line( pc->co, pa->co, 0xff0000ff );
 }
 
-float scene_tree_sway = 0.1f;
-
-static void scene_foliage_shader_use(void)
+static void scene_bh_closest( void *user, u32 index, v3f point, v3f closest )
 {
-   SHADER_USE( shader_debug_vcol );
+   scene_context *s = user;
 
-       glUniformMatrix4fv( SHADER_UNIFORM( shader_debug_vcol, "uPv" ), 
-         1, GL_FALSE, (float *)vg_pv );
+   v3f positions[3];
+   u32 *tri = &s->arrindices[ index*3 ];
+   for( int i=0; i<3; i++ )
+      v3_copy( s->arrvertices[tri[i]].co, positions[i] );
 
-   glUniform1i( SHADER_UNIFORM( shader_debug_vcol, "uMode" ), debugview );
-   glUniform1i( SHADER_UNIFORM( shader_debug_vcol, "uTexMain" ), 0 );
-   
-   glUniform1i( SHADER_UNIFORM( shader_debug_vcol, "uTexGradients" ), 1 );
-   vg_tex2d_bind( &tex_gradients, 1 );
-
-   glUniform1i( SHADER_UNIFORM( shader_debug_vcol, "uTexNoise" ), 2 );
-   glActiveTexture( GL_TEXTURE2 );
-   glBindTexture( GL_TEXTURE_2D, tex_dual_noise );
-
-   glUniform1f( SHADER_UNIFORM( shader_debug_vcol, "uTime" ), vg_time );
-   glUniform1f( SHADER_UNIFORM( shader_debug_vcol, "uSwayAmt" ), 
-         scene_tree_sway );
+   closest_on_triangle_1( point, positions, closest );
 }
 
-static void scene_bind( scene *pscene )
+static bh_system bh_system_scene = 
 {
-   mesh_bind( &pscene->mesh );
-}
+   .expand_bound = scene_bh_expand_bound,
+   .item_centroid = scene_bh_centroid,
+   .item_closest = scene_bh_closest,
+   .item_swap = scene_bh_swap,
+   .item_debug = scene_bh_debug,
+   .system_type = 0x1
+};
 
-static void scene_draw( scene *pscene )
+/*
+ * An extra step is added onto the end to calculate the hit normal
+ */
+static int scene_raycast( scene_context *s, bh_tree *bh, 
+                             v3f co, v3f dir, ray_hit *hit, u16 ignore )
 {
-   mesh_drawn( 0, pscene->indice_count );
-}
+   hit->tri = NULL;
 
-static void scene_debugsdf( scene *pscene )
-{
-   for( int i=0; i<pscene->shadower_count; i++ )
-   {
-      struct shadower *shadower = &pscene->shadowers[i];
-
-      v3f base, side;
-      v3_copy( shadower->sdf.origin, base );
-      base[1] -= shadower->sdf.info[1];
-      v3_copy( base, side );
-      side[0] += shadower->sdf.info[0];
-
-      vg_line2( shadower->sdf.origin, base, 0xff00ff00, 0xff0000ff );
-      vg_line2( side, base, 0xff00ff00, 0xff0000ff );
-      vg_line( side, shadower->sdf.origin, 0xff00ff00 );
+   bh_iter it;
+   bh_iter_init_ray( 0, &it, co, dir, hit->dist );
+   i32 idx;
+
+   while( bh_next( bh, &it, &idx ) ){
+      u32 *tri = &s->arrindices[ idx*3 ];
+
+      if( s->arrvertices[tri[0]].flags & ignore )  continue;
+
+      v3f vs[3];
+      for( u32 i=0; i<3; i++ )
+         v3_copy( s->arrvertices[tri[i]].co, vs[i] );
+      
+      f32 t;
+      if( ray_tri( vs, co, dir, &t, 0 ) ){
+         if( t < hit->dist ){
+            hit->dist = t;
+            hit->tri = tri;
+         }
+      }
    }
 
-   v3f p0 = { pscene->bbx[0][0], pscene->bbx[0][1], pscene->bbx[0][2] },
-       p1 = { pscene->bbx[0][0], pscene->bbx[1][1], pscene->bbx[0][2] }, 
-       p2 = { pscene->bbx[1][0], pscene->bbx[1][1], pscene->bbx[0][2] },
-       p3 = { pscene->bbx[1][0], pscene->bbx[0][1], pscene->bbx[0][2] },
+   if( hit->tri ){
+      v3f v0, v1;
+      
+      float *pa = s->arrvertices[hit->tri[0]].co,
+            *pb = s->arrvertices[hit->tri[1]].co,
+            *pc = s->arrvertices[hit->tri[2]].co;
+
+      v3_sub( pa, pb, v0 );
+      v3_sub( pc, pb, v1 );
+      v3_cross( v1, v0, hit->normal );
+      v3_normalize( hit->normal );
+      v3_muladds( co, dir, hit->dist, hit->pos );
+   }
 
-       p4 = { pscene->bbx[0][0], pscene->bbx[0][1], pscene->bbx[1][2] },
-       p5 = { pscene->bbx[0][0], pscene->bbx[1][1], pscene->bbx[1][2] }, 
-       p6 = { pscene->bbx[1][0], pscene->bbx[1][1], pscene->bbx[1][2] },
-       p7 = { pscene->bbx[1][0], pscene->bbx[0][1], pscene->bbx[1][2] };
-   
-   u32 col = 0xffff00c8;
-   vg_line( p0, p1, col );
-   vg_line( p1, p2, col );
-   vg_line( p2, p3, col );
-   vg_line( p3, p0, col );
-
-   vg_line( p4, p5, col );
-   vg_line( p5, p6, col );
-   vg_line( p6, p7, col );
-   vg_line( p7, p4, col );
-
-   vg_line( p0, p4, col );
-   vg_line( p1, p5, col );
-   vg_line( p2, p6, col );
-   vg_line( p3, p7, col );
+   return hit->tri?1:0;
 }
 
-static void scene_register(void)
+static bh_tree *scene_bh_create( void *lin_alloc, scene_context *s )
 {
-   SHADER_INIT( shader_debug_vcol );
-   SHADER_INIT( shader_standard_lit );
-   SHADER_INIT( shader_unlit );
+   u32 triangle_count = s->indice_count / 3;
+   return bh_create( lin_alloc, &bh_system_scene, s, triangle_count, 2 );
 }
+
+#endif