csrDraw.h

   1 // Copyright (C) 2021 Harry Godden (hgn)
   2
   3 // Extremely simple software renderer. Only supports orthographic
   4 //=======================================================================================================================
   5
   6 typedef struct csr_target csr_target;
   7 typedef struct csr_filter csr_filter;
   8 typedef struct csr_shader csr_shader;
   9 typedef enum EMSAA EMSAA;
  10
  11 typedef void (* csr_frag_program)( void *, vmf_vert[3], float, float, float );
  12 typedef void (* csr_frag_clear)( void * );
  13
  14 // API
  15 //=======================================================================================================================
  16
  17 // Create a render target. Resolution, msaa, and shader must be known at this point!
  18 void csr_create_target( csr_target *rt, u32 x, u32 y, EMSAA aa, csr_shader *shader );
  19 void csr_rt_clear( csr_target *rt );
  20 void csr_rt_free( csr_target *rt );
  21
  22 // Refit bounds so that it is square, and crops to center with padding
  23 void csr_auto_fit( csr_target *rt, float padding );
  24
  25 // Run this after bounds have been adjusted on the RT to update the size of the msaa
  26 // Auto fit will call this.
  27 void csr_update_subsamples( csr_target *rt );
  28
  29 // Write CS:GO radar txt
  30 void csr_write_txt( char const *path, const char *name, csr_target *rt  );
  31
  32 // Render calls
  33 // ------------
  34
  35 // Render a finalzied triangle into target. Coordinates are world space
  36 void simple_raster( csr_target *rt, vmf_vert tri[3] );
  37
  38 // Draw a batch of triangles with an affine world transformation
  39 void csr_draw( csr_target *rt, vmf_vert *triangles, u32 triangle_count, m4x3f transform );
  40
  41 // Draw VMF with filtering options. Will automatically branch into instances
  42 // You should call this with the last two recursive arguments (prev,inst), set to NULL
  43 //
  44 // Filter is optional, it can be st to NULL to just render everything.
  45 void csr_vmf_render( csr_target *rt, vmf_map *map, vdf_node *root, csr_filter *filter, m4x3f prev, m4x3f inst );
  46
  47 void csr_rt_save_tga( csr_target *rt, const char *path, u32 offset, u32 nc );
  48 void csr_rt_save_c32f( csr_target *rt, const char *path, u32 offset );
  49
  50
  51 // Implementation
  52 //=======================================================================================================================
  53
  54 struct csr_shader
  55 {
  56         u32 stride;
  57         csr_frag_program frag;
  58         csr_frag_clear clear;
  59 };
  60
  61 struct csr_target
  62 {
  63         void  *colour;
  64         float *depth;
  65
  66         u32 x, y;
  67         boxf bounds;
  68         float scale;
  69
  70         v2f subsamples[ 8 ];
  71         int num_samples;
  72         v2f *sample_src;
  73
  74         csr_shader *shader;
  75 };
  76
  77 struct csr_filter
  78 {
  79         const char *visgroup;           // Limit to this visgroup only
  80         const char *classname;          // Limit to this exact classname. will not draw world
  81
  82         int compute_bounds_only;
  83 };
  84
  85 enum EMSAA
  86 {
  87         k_EMSAA_none,
  88         k_EMSAA_2x2,
  89         k_EMSAA_RGSS,
  90         k_EMSAA_8R
  91 };
  92
  93 #ifdef CSR_EXECUTABLE
  94
  95 // MSAA patterns
  96 v2f csr_msaa_1[] =
  97 {
  98         {0.f, 0.f}
  99 };
 100
 101 // XX
 102 // XX
 103 v2f csr_msaa_2x2[] =
 104 {
 105         {  0x0.4p0f,  0x0.4p0f },
 106         {  0x0.4p0f, -0x0.4p0f },
 107         { -0x0.4p0f, -0x0.4p0f },
 108         { -0x0.4p0f,  0x0.4p0f }
 109 };
 110
 111 //   X
 112 // X
 113 //    X
 114 // X
 115 v2f csr_msaa_2x2rgss[] =
 116 {
 117         {  0x0.2p0f,  0x0.6p0f },
 118         { -0x0.6p0f,  0x0.2p0f },
 119         { -0x0.2p0f, -0x0.6p0f },
 120         {  0x0.6p0f, -0x0.2p0f }
 121 };
 122
 123 //     X
 124 //   X
 125 // X
 126 //       X
 127 //  X
 128 //        X
 129 //      X
 130 //    X
 131 v2f csr_msaa_8rook[] =
 132 {
 133         {  0x0.1p0f,  0x0.7p0f },
 134         {  0x0.5p0f,  0x0.1p0f },
 135         {  0x0.7p0f, -0x0.3p0f },
 136         {  0x0.3p0f, -0x0.5p0f },
 137         { -0x0.1p0f, -0x0.7p0f },
 138         { -0x0.5p0f, -0x0.1p0f },
 139         { -0x0.7p0f,  0x0.3p0f },
 140         { -0x0.3p0f,  0x0.5p0f }
 141 };
 142
 143
 144 void csr_create_target( csr_target *rt, u32 x, u32 y, EMSAA aa, csr_shader *shader )
 145 {
 146         rt->x = x;
 147         rt->y = y;
 148
 149         switch( aa )
 150         {
 151                 default:
 152                 case k_EMSAA_none:
 153                         rt->num_samples = 1;
 154                         rt->sample_src = csr_msaa_1;
 155                         break;
 156
 157                 case k_EMSAA_2x2:
 158                         rt->num_samples = 4;
 159                         rt->sample_src = csr_msaa_2x2;
 160                         break;
 161
 162                 case k_EMSAA_RGSS:
 163                         rt->num_samples = 4;
 164                         rt->sample_src = csr_msaa_2x2rgss;
 165                         break;
 166
 167                 case k_EMSAA_8R:
 168                         rt->num_samples = 8;
 169                         rt->sample_src = csr_msaa_8rook;
 170                         break;
 171         }
 172
 173         rt->shader = shader;
 174         rt->depth = (float *)csr_malloc( x*y*rt->num_samples * sizeof(float) );
 175         rt->colour = csr_malloc( x * y * rt->shader->stride * rt->num_samples );
 176
 177         v3_fill( rt->bounds[0], INFINITY );
 178         v3_fill( rt->bounds[1], -INFINITY );
 179 }
 180
 181 void csr_update_subsamples( csr_target *rt )
 182 {
 183         float range_x = (rt->bounds[1][0]-rt->bounds[0][0]);
 184         float range_y = (rt->bounds[1][1]-rt->bounds[0][1]);
 185
 186         v2f pixel_size = { range_x/(float)rt->x, range_y/(float)rt->y };
 187
 188         for( int i = 0; i < rt->num_samples; i ++ )
 189         {
 190                 v2_mul( rt->sample_src[i], pixel_size, rt->subsamples[i] );
 191         }
 192 }
 193
 194 void csr_rt_free( csr_target *rt )
 195 {
 196         free( rt->depth );
 197         free( rt->colour );
 198 }
 199
 200 void csr_rt_clear( csr_target *rt )
 201 {
 202         for( u32 i = 0; i < rt->x*rt->y*rt->num_samples; i ++ )
 203         {
 204                 rt->shader->clear( rt->colour + i * rt->shader->stride );
 205                 rt->depth[i] = 0.f;
 206         }
 207 }
 208
 209 void csr_auto_fit( csr_target *rt, float padding )
 210 {
 211         // Correct aspect ratio to be square
 212         float dx, dy, l, cx, cy;
 213
 214         dx = rt->bounds[1][0] - rt->bounds[0][0];
 215         dy = rt->bounds[1][1] - rt->bounds[0][1];
 216
 217         l = fmaxf( dx, dy ) * .5f;
 218
 219         cx = (rt->bounds[1][0] + rt->bounds[0][0]) * .5f;
 220         cy = (rt->bounds[1][1] + rt->bounds[0][1]) * .5f;
 221
 222         rt->bounds[0][0] = cx - l - padding;
 223         rt->bounds[1][0] = cx + l + padding;
 224         rt->bounds[0][1] = cy - l - padding;
 225         rt->bounds[1][1] = cy + l + padding;
 226
 227         rt->scale = l + padding;
 228
 229         csr_update_subsamples( rt );
 230 }
 231
 232 void simple_raster( csr_target *rt, vmf_vert tri[3] )
 233 {
 234         // Very very simplified rasterizing algorithm
 235         v2f bmin = { 0.f, 0.f };
 236         v2f bmax = { rt->x, rt->y };
 237
 238         v2_minv( tri[0].co, tri[1].co, bmin );
 239         v2_minv( tri[2].co, bmin, bmin );
 240
 241         v2_maxv( tri[0].co, tri[1].co, bmax );
 242         v2_maxv( tri[2].co, bmax, bmax );
 243
 244         float range_x = (rt->bounds[1][0]-rt->bounds[0][0])/(float)rt->x;
 245         float range_y = (rt->bounds[1][1]-rt->bounds[0][1])/(float)rt->y;
 246
 247         int start_x = csr_min( rt->x-1, csr_max( 0,       floorf( (bmin[0]-rt->bounds[0][0])/range_x)));
 248         int end_x =   csr_max( 0,       csr_min( rt->x-1,  ceilf( (bmax[0]-rt->bounds[0][0])/range_x)));
 249         int start_y = csr_min( rt->y-1, csr_max( 0,       floorf( (bmin[1]-rt->bounds[0][1])/range_y)));
 250         int end_y =   csr_max( 0,       csr_min( rt->y-1,  ceilf( (bmax[1]-rt->bounds[0][1])/range_y)));
 251
 252         v2f v0, v1, v2, vp;
 253         float d, bca = 0.f, bcb = 0.f, bcc = 0.f;
 254
 255         v2_sub( tri[1].co, tri[0].co, v0 );
 256         v2_sub( tri[2].co, tri[0].co, v1 );
 257         v2_sub( tri[1].co, tri[2].co, v2 );
 258         d = 1.f / (v0[0]*v1[1] - v1[0]*v0[1]);
 259
 260         // Backface culling
 261         if( v2_cross( v0, v1 ) > 0.f )
 262                 return;
 263
 264         v2f trace_origin;
 265
 266         for( u32 py = start_y; py <= end_y; py ++ )
 267         {
 268                 trace_origin[1] = csr_lerpf( rt->bounds[0][1], rt->bounds[1][1], (float)py/(float)rt->y );
 269
 270                 for( u32 px = start_x; px <= end_x; px ++ )
 271                 {
 272                         u32 sample_index = ((rt->y-py-1)*rt->x+px) * rt->num_samples;
 273
 274                         void *frag = rt->colour + sample_index*rt->shader->stride;
 275                         float *depth = &rt->depth[ sample_index ];
 276
 277                         trace_origin[0] = csr_lerpf( rt->bounds[0][0], rt->bounds[1][0], (float)px/(float)rt->x );
 278
 279                         // Determine coverage
 280                         for( int i = 0; i < rt->num_samples; i ++ )
 281                         {
 282                                 v3f sample_origin;
 283
 284                                 v2_add( rt->subsamples[ i ], trace_origin, sample_origin );
 285                                 v2_sub( sample_origin, tri[0].co, vp );
 286
 287                                 if( v2_cross( v0, vp ) > 0.f )
 288                                         continue;
 289                                 if( v2_cross( vp, v1 ) > 0.f )
 290                                         continue;
 291
 292                                 v2f vp2;
 293                                 v2_sub( sample_origin, tri[2].co, vp2 );
 294
 295                                 if( v2_cross( vp2, v2 ) > 0.f )
 296                                         continue;
 297
 298                                 bcb = (vp[0]*v1[1] - v1[0]*vp[1]) * d;
 299                                 bcc = (v0[0]*vp[1] - vp[0]*v0[1]) * d;
 300                                 bca = 1.f - bcb - bcc;
 301
 302                                 float hit = tri[0].co[2] * bca + tri[1].co[2] * bcb + tri[2].co[2] * bcc;
 303                                 float hit_depth = hit + 16385.f;
 304
 305                                 if( hit_depth > depth[i] && hit >= rt->bounds[0][2] && hit <= rt->bounds[1][2] )
 306                                 {
 307                                         depth[i] = hit_depth;
 308                                         rt->shader->frag( frag+i*rt->shader->stride, tri, bca, bcb, bcc );
 309                                 }
 310                         }
 311                 }
 312         }
 313 }
 314
 315 void csr_draw( csr_target *rt, vmf_vert *triangles, u32 triangle_count, m4x3f transform )
 316 {
 317         m3x3f normal;
 318         vmf_vert new_tri[3];
 319
 320         // Derive normal matrix
 321         m4x3_to_3x3( transform, normal );
 322         m3x3_inv_transpose( normal, normal );
 323
 324         for( u32 i = 0; i < triangle_count; i ++ )
 325         {
 326                 vmf_vert *triangle = triangles + i*3;
 327
 328                 m4x3_mulv( transform, triangle[0].co, new_tri[0].co );
 329                 m4x3_mulv( transform, triangle[1].co, new_tri[1].co );
 330                 m4x3_mulv( transform, triangle[2].co, new_tri[2].co );
 331
 332                 m3x3_mulv( normal, triangle[0].nrm, new_tri[0].nrm );
 333                 m3x3_mulv( normal, triangle[1].nrm, new_tri[1].nrm );
 334                 m3x3_mulv( normal, triangle[2].nrm, new_tri[2].nrm );
 335
 336                 v3_normalize( new_tri[0].nrm );
 337                 v3_normalize( new_tri[1].nrm );
 338                 v3_normalize( new_tri[2].nrm );
 339
 340                 m4x3_mulv( transform, triangles[0].origin, new_tri[0].origin );
 341
 342                 simple_raster( rt, new_tri );
 343         }
 344 }
 345
 346 void csr_vmf_render( csr_target *rt, vmf_map *map, vdf_node *root, csr_filter *filter, m4x3f prev, m4x3f inst )
 347 {
 348         m4x3f transform = M4X3_IDENTITY;
 349         vmf_solid solid;
 350         vmf_vert tri[3];
 351         boxf trf_bounds;
 352
 353         u32 group_id = 0;
 354         int filter_visgroups = 0, filter_classname = 0, compute_bounds_only = 0;
 355
 356         if( filter )
 357         {
 358                 if( filter->visgroup )
 359                 {
 360                         filter_visgroups = 1;
 361                         group_id = vmf_visgroup_id( root, filter->visgroup );
 362                 }
 363
 364                 if( filter->classname )
 365                 {
 366                         filter_classname = 1;
 367                 }
 368
 369                 compute_bounds_only = filter->compute_bounds_only;
 370         }
 371
 372         // Multiply previous transform with instance transform to create basis
 373         if( prev )
 374         {
 375                 m4x3_mul( prev, inst, transform );
 376         }
 377
 378         // Gather world brushes
 379         solidgen_ctx_init( &solid );
 380
 381         if( !filter_classname )
 382         {
 383                 vdf_node *world = vdf_next( root, "world", NULL );
 384
 385                 vdf_foreach( world, "solid", brush )
 386                 {
 387                         if( filter_visgroups && !vmf_visgroup_match( brush, group_id ) )
 388                                 continue;
 389
 390                         // TODO: heap-use-after-free
 391                         solidgen_push( &solid, brush );
 392                 }
 393         }
 394
 395         // Actual entity loop
 396         m4x3f model;
 397
 398         vdf_foreach( root, "entity", ent )
 399         {
 400                 if( filter_visgroups && !vmf_visgroup_match( ent, group_id ) )
 401                         continue;
 402
 403                 if( filter_classname )
 404                         if( strcmp( kv_get( ent, "classname", "" ), filter->classname ) )
 405                                 continue;
 406
 407                 if( ent->user & VMF_FLAG_IS_PROP )
 408                 {
 409                         // Create model transform
 410                         m4x3_identity( model );
 411
 412                         vmf_entity_transform( ent, model );
 413                         m4x3_mul( transform, model, model );
 414
 415                         // Draw model
 416                         mdl_mesh_t *mdl = &map->models[ ent->user1 ].mdl;
 417
 418                         if( compute_bounds_only )
 419                         {
 420                                 map->models[ ent->user1 ].need_load = 1;
 421                                 m4x3_expand_aabb_point( model, rt->bounds, (v3f){0.f,0.f,0.f} );
 422                         }
 423                         else
 424                         {
 425                                 for( int i = 0; i < mdl->num_indices/3; i ++ )
 426                                 {
 427                                         for( int j = 0; j < 3; j ++ )
 428                                         {
 429                                                 v3_copy( &mdl->vertices[ mdl->indices[ i*3+j ] *8 ],   tri[j].co );
 430                                                 v3_copy( &mdl->vertices[ mdl->indices[ i*3+j ] *8+3 ], tri[j].nrm );
 431                                                 v3_zero( tri[j].origin );
 432                                         }
 433
 434                                         csr_draw( rt, tri, 1, model );
 435                                 }
 436                         }
 437                 }
 438                 else if( ent->user & VMF_FLAG_IS_INSTANCE )
 439                 {
 440                         m4x3_identity( model );
 441                         vmf_entity_transform( ent, model );
 442
 443                         csr_vmf_render( rt, map, map->cache[ ent->user1 ].root, filter, transform, model );
 444                 }
 445                 else
 446                 {
 447                         // Brush entity
 448                         vdf_foreach( ent, "solid", ent_solid )
 449                         {
 450                                 solidgen_push( &solid, ent_solid );
 451                         }
 452                 }
 453         }
 454
 455         if( compute_bounds_only )
 456         {
 457                 solidgen_bounds( &solid, trf_bounds );
 458                 m4x3_transform_aabb( transform, trf_bounds );
 459                 box_concat( rt->bounds, trf_bounds );
 460         }
 461         else
 462         {
 463                 // Draw brushes
 464                 for( int i = 0; i < csr_sb_count( solid.indices )/3; i ++ )
 465                 {
 466                         u32 * base = solid.indices + i*3;
 467
 468                         tri[0] = solid.verts[ base[0] ];
 469                         tri[1] = solid.verts[ base[1] ];
 470                         tri[2] = solid.verts[ base[2] ];
 471
 472                         csr_draw( rt, tri, 1, transform );
 473                 }
 474         }
 475
 476         solidgen_ctx_reset( &solid );
 477         solidgen_ctx_free( &solid );
 478 }
 479
 480 void csr_write_filerr( const char *path )
 481 {
 482         log_error( "File write error (No such file or directory): '%s'\n", path );
 483 }
 484
 485 void csr_write_txt( char const *path, const char *name, csr_target *rt  )
 486 {
 487         FILE *write_ptr;
 488
 489         write_ptr = fopen( path, "w" );
 490
 491         if( write_ptr )
 492         {
 493                 fprintf( write_ptr, "\"%s\"\n\{\n", name );
 494                 fprintf( write_ptr, "\t\"material\" \"overviews/%s\"\n", name );
 495                 fprintf( write_ptr, "\t\"pos_x\" \"%.8f\"\n", rt->bounds[0][0] );
 496                 fprintf( write_ptr, "\t\"pos_y\" \"%.8f\"\n", rt->bounds[0][1] );
 497                 fprintf( write_ptr, "\t\"scale\" \"%.8f\"\n", rt->scale / (float)rt->x );
 498                 fprintf( write_ptr, "}\n" );
 499
 500                 fclose( write_ptr );
 501         }
 502         else
 503         {
 504                 csr_write_filerr( path );
 505         }
 506 }
 507
 508 // ALWAYS RGB32
 509 void csr_rt_save_c32f( csr_target *rt, const char *path, u32 offset )
 510 {
 511         float *image = (float *)csr_malloc( rt->x*rt->y*3*sizeof(float) );
 512
 513         float contrib = 1.f/(float)rt->num_samples;
 514
 515         for( int i = 0; i < rt->x*rt->y; i ++ )
 516         {
 517                 void *src = rt->colour + offset + i * rt->num_samples * rt->shader->stride;
 518                 float *dst = image + i*3;
 519
 520                 v3_zero( dst );
 521                 for( int k = 0; k < rt->num_samples; k ++ )
 522                 {
 523                         v3_muladds( dst, (float *)(src + k*rt->shader->stride), contrib, dst );
 524                 }
 525         }
 526
 527         if( !csr_32f_write( path, rt->x, rt->y, image ) )
 528                 csr_write_filerr( path );
 529
 530         free( image );
 531 }
 532
 533 // Save floating point buffer to tga. Must be in range (0-1)
 534 // Offset and stride are in bytes
 535 void csr_rt_save_tga( csr_target *rt, const char *path, u32 offset, u32 nc )
 536 {
 537         u8 *image = (u8 *)csr_malloc( rt->x*rt->y * 4 );
 538
 539         float contrib = 255.f/(float)rt->num_samples;
 540
 541         for( int i = 0; i < rt->x*rt->y; i ++ )
 542         {
 543                 void *src = rt->colour + offset + i * rt->num_samples * rt->shader->stride;
 544                 u8 *dst = image + i*4;
 545
 546                 v4f accum = { 0.f, 0.f, 0.f, 0.f };
 547
 548                 for( int k = 0; k < rt->num_samples; k ++ )
 549                 {
 550                         float *src_sample = (float *)(src + k*rt->shader->stride);
 551
 552                         for( int j = 0; j < nc; j ++ )
 553                         {
 554                                 accum[ j ] += src_sample[ j ] * contrib;
 555                         }
 556                 }
 557
 558                 // TODO: Clamp this value
 559                 dst[0] = accum[0];
 560                 dst[1] = accum[1];
 561                 dst[2] = accum[2];
 562                 dst[3] = accum[3];
 563         }
 564
 565         if( !csr_tga_write( path, rt->x, rt->y, nc, image ) )
 566                 csr_write_filerr( path );
 567
 568         free( image );
 569 }
 570
 571 #endif