command line, multisampling, optimisations
[csRadar.git] / csrDraw.h
index b6e17caf8d2f1377469056c5064f695f16eaab82..29189a1e9cfd16cc21e10dc6d5367cdcc9627845 100644 (file)
--- a/csrDraw.h
+++ b/csrDraw.h
@@ -3,33 +3,60 @@ typedef struct csr_target csr_target;
 typedef struct csr_filter csr_filter;
 
 struct csr_frag
-{
-       u32 id; // Triangle index
-       float depth;    // 'depth testing'
-       
+{      
        v3f co;
        v3f nrm;
+       
+       float depth;
 };
 
 struct csr_target
 {
        csr_frag *fragments;
+       
        u32 x, y; 
-       v4f bounds;
+       boxf bounds;
+       float scale;
+       
+       v2f subsamples[ 16 ];
+       int num_samples;
 };
 
 struct csr_filter
 {
        const char *visgroup;           // Limit to this visgroup only
        const char *classname;          // Limit to this exact classname. will not draw world
+       
+       int compute_bounds_only;
 };
 
-void csr_create_target( csr_target *rt, u32 x, u32 y, v4f bounds )
+void csr_create_target( csr_target *rt, u32 x, u32 y )
 {
        rt->x = x;
        rt->y = y;
-       rt->fragments = (csr_frag *)csr_malloc( x*y*sizeof(csr_frag) );
-       v4_copy( bounds, rt->bounds );
+       rt->num_samples = 4;
+       
+       rt->fragments = (csr_frag *)csr_malloc( x*y*sizeof(csr_frag)*rt->num_samples );
+       
+       v3_fill( rt->bounds[0], INFINITY );
+       v3_fill( rt->bounds[1], -INFINITY );
+}
+
+void csr_update_subsamples( csr_target *rt )
+{
+       float range_x = (rt->bounds[1][0]-rt->bounds[0][0]);
+       float range_y = (rt->bounds[1][1]-rt->bounds[0][1]);
+
+       v2f pixel_size = { range_x/(float)rt->x, range_y/(float)rt->y };
+
+       rt->subsamples[0][0] = pixel_size[0] * -0.25f;
+       rt->subsamples[0][1] = 0.f;
+       rt->subsamples[1][0] = pixel_size[0] * 0.75f;
+       rt->subsamples[1][1] = pixel_size[1] * 0.25f;
+       rt->subsamples[2][0] = 0.f;
+       rt->subsamples[2][1] = pixel_size[1] * 0.5f;
+       rt->subsamples[3][0] = pixel_size[0] * 0.5f;
+       rt->subsamples[3][1] = pixel_size[1] * 0.75f;
 }
 
 void csr_rt_free( csr_target *rt )
@@ -39,17 +66,56 @@ void csr_rt_free( csr_target *rt )
 
 void csr_rt_clear( csr_target *rt )
 {
-       for( u32 i = 0; i < rt->x*rt->y; i ++ )
+       for( u32 i = 0; i < rt->x*rt->y*rt->num_samples; i ++ )
        {
-               rt->fragments[ i ].depth = 0.f;
+               v3_zero( rt->fragments[ i ].co );
+               v3_zero( rt->fragments[ i ].nrm );
+               rt->fragments[i].depth = 0.f;
        }
 }
 
-void simple_raster( csr_target *rt, vmf_vert tri[3], int id )
+void csr_auto_fit( csr_target *rt, float padding )
 {
-       // Very simplified tracing algorithm
-       float tqa = 0.f, tqb = 0.f;
+       // Correct aspect ratio to be square
+       float dx, dy, d, l, cx, cy;
+       dx = rt->bounds[1][0] - rt->bounds[0][0];
+       dy = rt->bounds[1][1] - rt->bounds[0][1];
+       
+       l = fmaxf( dx, dy );
+       d = l * ( l / dx ) * .5f;
+               
+       cx = (rt->bounds[1][0] + rt->bounds[0][0]) * .5f;
+       cy = (rt->bounds[1][1] + rt->bounds[0][1]) * .5f;
+
+       rt->bounds[0][0] = cx - d - padding;
+       rt->bounds[1][0] = cx + d + padding;
+       rt->bounds[0][1] = cy - d - padding;
+       rt->bounds[1][1] = cy + d + padding;
        
+       rt->scale = d + padding;
+       
+       csr_update_subsamples( rt );
+}
+
+void csr_write_txt( char const *path, const char *name, csr_target *rt  )
+{
+       FILE *write_ptr;
+       
+       write_ptr = fopen( path, "w" );
+       
+       fprintf( write_ptr, "\"%s\"\n\{\n", name );
+       fprintf( write_ptr, "\t\"material\" \"overviews/%s\"\n", name );
+       fprintf( write_ptr, "\t\"pos_x\" \"%.8f\"\n", rt->bounds[0][0] );
+       fprintf( write_ptr, "\t\"pos_y\" \"%.8f\"\n", rt->bounds[0][1] );
+       fprintf( write_ptr, "\t\"scale\" \"%.8f\"\n", rt->scale / (float)rt->x );
+       fprintf( write_ptr, "}\n" );
+       
+       fclose( write_ptr );
+}
+
+void simple_raster( csr_target *rt, vmf_vert tri[3] )
+{
+       // Very very simplified rasterizing algorithm
        v2f bmin = { 0.f, 0.f };
        v2f bmax = { rt->x, rt->y };
        
@@ -59,35 +125,72 @@ void simple_raster( csr_target *rt, vmf_vert tri[3], int id )
        v2_maxv( tri[0].co, tri[1].co, bmax );
        v2_maxv( tri[2].co, bmax, bmax );
 
-       float range_x = (rt->bounds[2]-rt->bounds[0])/(float)rt->x;
-       float range_y = (rt->bounds[3]-rt->bounds[1])/(float)rt->y;
+       float range_x = (rt->bounds[1][0]-rt->bounds[0][0])/(float)rt->x;
+       float range_y = (rt->bounds[1][1]-rt->bounds[0][1])/(float)rt->y;
        
-       int start_x = csr_min( rt->x-1, csr_max( 0,     floorf( (bmin[0]-rt->bounds[0])/range_x)));
-       int end_x =   csr_max( 0,       csr_min( rt->x-1, floorf( (bmax[0]-rt->bounds[0])/range_x )));
-       int start_y = csr_min( rt->y-1, csr_max( 0,      ceilf( (bmin[1]-rt->bounds[1])/range_y )));
-       int end_y =   csr_max( 0,       csr_min( rt->y-1,  ceilf( (bmax[1]-rt->bounds[1])/range_y )));
+       int start_x = csr_min( rt->x-1, csr_max( 0,       floorf( (bmin[0]-rt->bounds[0][0])/range_x)));
+       int end_x =   csr_max( 0,       csr_min( rt->x-1,  ceilf( (bmax[0]-rt->bounds[0][0])/range_x)));
+       int start_y = csr_min( rt->y-1, csr_max( 0,       floorf( (bmin[1]-rt->bounds[0][1])/range_y)));
+       int end_y =   csr_max( 0,       csr_min( rt->y-1,  ceilf( (bmax[1]-rt->bounds[0][1])/range_y)));
        
-       v3f trace_dir = { 0.f, 0.f, 1.f };
-       v3f trace_origin = { 0.f, 0.f, -16385.f };
+       v2f v0, v1, v2, vp;
+       float d, bca = 0.f, bcb = 0.f, bcc = 0.f;
+       
+       v2_sub( tri[1].co, tri[0].co, v0 );
+       v2_sub( tri[2].co, tri[0].co, v1 );
+       v2_sub( tri[1].co, tri[2].co, v2 );
+       d = 1.f / (v0[0]*v1[1] - v1[0]*v0[1]);
+
+       // Backface culling
+       if( v2_cross( v0, v1 ) > 0.f )
+               return;
+
+       v2f trace_origin;
 
        for( u32 py = start_y; py <= end_y; py ++ )
        {
-               trace_origin[1] = csr_lerpf( rt->bounds[1], rt->bounds[3], (float)py/(float)rt->y );
+               trace_origin[1] = csr_lerpf( rt->bounds[0][1], rt->bounds[1][1], (float)py/(float)rt->y );
                
                for( u32 px = start_x; px <= end_x; px ++ )
                {
-                       csr_frag *frag = &rt->fragments[ py * rt->y + px ];
+                       csr_frag *frag = &rt->fragments[ (py * rt->y + px) * rt->num_samples ];
 
-                       trace_origin[0] = csr_lerpf( rt->bounds[0], rt->bounds[2], (float)px/(float)rt->x );
-                       float tdepth = csr_ray_tri( trace_origin, trace_dir, tri[0].co, tri[1].co, tri[2].co, &tqa, &tqb );
+                       trace_origin[0] = csr_lerpf( rt->bounds[0][0], rt->bounds[1][0], (float)px/(float)rt->x );
                        
-                       if( tdepth > frag->depth )
+                       // Determine coverage
+                       for( int i = 0; i < rt->num_samples; i ++ )
                        {
-                               frag->depth = tdepth;
+                               v3f sample_origin;
+                               
+                               v2_add( rt->subsamples[ i ], trace_origin, sample_origin );
+                               v2_sub( sample_origin, tri[0].co, vp );
+                               
+                               if( v2_cross( v0, vp ) > 0.f )
+                                       continue;
+                               if( v2_cross( vp, v1 ) > 0.f )
+                                       continue;
                                
-                               v3_muls( tri[1].co, tqa, frag->co );
-                               v3_muladds( frag->co, tri[2].co, tqb, frag->co );
-                               v3_muladds( frag->co, tri[0].co, 1.f - tqa - tqb, frag->co );
+                               v2f vp2;
+                               v2_sub( sample_origin, tri[2].co, vp2 );
+                               
+                               if( v2_cross( vp2, v2 ) > 0.f )
+                                       continue;
+
+                               bcb = (vp[0]*v1[1] - v1[0]*vp[1]) * d;
+                               bcc = (v0[0]*vp[1] - vp[0]*v0[1]) * d;
+                               bca = 1.f - bcb - bcc;
+                               
+                               float hit = (tri[0].co[2] * bca + tri[1].co[2] * bcb + tri[2].co[2] * bcc) +16385.f;
+                               
+                               if( hit > frag[i].depth )
+                               {
+                                       frag[i].depth = hit;
+                                       v3_muls( tri[0].co, bca, frag[i].co );
+                                       v3_muladds( frag[i].co, tri[1].co, bcb, frag[i].co );
+                                       v3_muladds( frag[i].co, tri[2].co, bcc, frag[i].co );
+                                       
+                                       // TODO: Same for normal map
+                               }
                        }
                }
        }
@@ -113,7 +216,7 @@ void csr_draw( csr_target *rt, vmf_vert *triangles, u32 triangle_count, m4x3f tr
                m3x3_mulv( normal, triangle[1].nrm, new_tri[1].nrm );
                m3x3_mulv( normal, triangle[2].nrm, new_tri[2].nrm );
                
-               simple_raster( rt, new_tri, 0 );
+               simple_raster( rt, new_tri );
        }
 }
 
@@ -123,9 +226,10 @@ void draw_vmf_group( csr_target *rt, vmf_map *map, vdf_node *root, csr_filter *f
        vmf_solid solid;
        vmf_vert tri[3];
        vdf_node *ent_solid;
+       boxf trf_bounds;
 
        u32 group_id = 0;
-       int filter_visgroups = 0, filter_classname = 0;
+       int filter_visgroups = 0, filter_classname = 0, compute_bounds_only = 0;
        
        if( filter )
        {
@@ -139,6 +243,8 @@ void draw_vmf_group( csr_target *rt, vmf_map *map, vdf_node *root, csr_filter *f
                {
                        filter_classname = 1;
                }
+               
+               compute_bounds_only = filter->compute_bounds_only;
        }
        
        // Multiply previous transform with instance transform to create basis
@@ -185,17 +291,29 @@ void draw_vmf_group( csr_target *rt, vmf_map *map, vdf_node *root, csr_filter *f
                        
                        // Draw model
                        mdl_mesh_t *mdl = &map->models[ ent->user1 ].mdl;
-                       for( int i = 0; i < mdl->num_indices/3; i ++ )
+                       
+                       if( compute_bounds_only )
                        {
-                               for( int j = 0; j < 3; j ++ )
+                               box_copy( mdl->bounds, trf_bounds );
+                               m4x3_transform_aabb( model, trf_bounds );
+                               
+                               // Join
+                               box_concat( rt->bounds, trf_bounds );
+                       }
+                       else
+                       {
+                               for( int i = 0; i < mdl->num_indices/3; i ++ )
                                {
-                                       v3_copy( &mdl->vertices[ mdl->indices[ i*3+j ] *8 ],   tri[j].co );
-                                       v3_copy( &mdl->vertices[ mdl->indices[ i*3+j ] *8+3 ], tri[j].nrm );
-                                       tri[j].xy[0] = 0.f;
-                                       tri[j].xy[1] = 0.f;
+                                       for( int j = 0; j < 3; j ++ )
+                                       {
+                                               v3_copy( &mdl->vertices[ mdl->indices[ i*3+j ] *8 ],   tri[j].co );
+                                               v3_copy( &mdl->vertices[ mdl->indices[ i*3+j ] *8+3 ], tri[j].nrm );
+                                               tri[j].xy[0] = 0.f;
+                                               tri[j].xy[1] = 0.f;
+                                       }
+                                       
+                                       csr_draw( rt, tri, 1, model );
                                }
-                               
-                               csr_draw( rt, tri, 1, model );
                        }
                } 
                else if( ent->user & VMF_FLAG_IS_INSTANCE )
@@ -215,18 +333,58 @@ void draw_vmf_group( csr_target *rt, vmf_map *map, vdf_node *root, csr_filter *f
                }
        }
        
-       // Draw brushes
-       for( int i = 0; i < csr_sb_count( solid.indices )/3; i ++ )
+       if( compute_bounds_only )
        {
-               u32 * base = solid.indices + i*3;
-               
-               tri[0] = solid.verts[ base[0] ];
-               tri[1] = solid.verts[ base[1] ];
-               tri[2] = solid.verts[ base[2] ];
-               
-               csr_draw( rt, tri, 1, transform );
+               solidgen_bounds( &solid, trf_bounds );
+               m4x3_transform_aabb( transform, trf_bounds );
+               box_concat( rt->bounds, trf_bounds );
+       }
+       else
+       {
+               // Draw brushes
+               for( int i = 0; i < csr_sb_count( solid.indices )/3; i ++ )
+               {
+                       u32 * base = solid.indices + i*3;
+                       
+                       tri[0] = solid.verts[ base[0] ];
+                       tri[1] = solid.verts[ base[1] ];
+                       tri[2] = solid.verts[ base[2] ];
+                       
+                       csr_draw( rt, tri, 1, transform );
+               }       
        }
        
        solidgen_ctx_reset( &solid );
        solidgen_ctx_free( &solid );
 }
+
+void csr_rt_save_buffers( csr_target *rt, const char *basename, const char *subname )
+{
+       char output[ 512 ];
+       
+       float *image = (float *)csr_malloc( 1024*1024*sizeof(float)*3 );
+       
+       for( int l = 0; l < rt->x; l ++ )
+       {
+               for( int x = 0; x < rt->y; x ++ )
+               {
+                       float *dst = &image[ (l*1024+x)*3 ];
+                       csr_frag *src = &rt->fragments[ ((1023-l)*1024+x)*rt->num_samples ];
+                       
+                       v3_zero( dst );
+                       v3_muls( src[0].co, 1.f/(float)rt->num_samples, dst );
+                       v3_muladds( dst, src[1].co, 1.f/(float)rt->num_samples, dst );
+                       v3_muladds( dst, src[2].co, 1.f/(float)rt->num_samples, dst );
+                       v3_muladds( dst, src[3].co, 1.f/(float)rt->num_samples, dst );
+               }
+       }
+       
+       // Save position buffer
+       strcpy( output, basename );
+       strcat( output, "." );
+       strcat( output, subname );
+       strcat( output, "_position.pfm" );
+       csr_32f_write( output, rt->x, rt->y, image );   
+       
+       free( image );
+}