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
  48 void csr_rt_save_tga( csr_target *rt, const char *path, u32 offset, u32 nc );
  49
  50 // Obsolete
  51 void csr_rt_save_buffers( csr_target *rt, const char *basename, const char *subname );
  52
  53 // Implementation
  54 //=======================================================================================================================
  55
  56 struct csr_shader
  57 {
  58         u32 stride;
  59         csr_frag_program frag;
  60         csr_frag_clear clear;
  61 };
  62
  63 struct csr_target
  64 {
  65         void  *colour;
  66         float *depth;
  67
  68         u32 x, y;
  69         boxf bounds;
  70         float scale;
  71
  72         v2f subsamples[ 8 ];
  73         int num_samples;
  74         v2f *sample_src;
  75
  76         csr_shader *shader;
  77 };
  78
  79 struct csr_filter
  80 {
  81         const char *visgroup;           // Limit to this visgroup only
  82         const char *classname;          // Limit to this exact classname. will not draw world
  83
  84         int compute_bounds_only;
  85 };
  86
  87 enum EMSAA
  88 {
  89         k_EMSAA_none,
  90         k_EMSAA_2x2,
  91         k_EMSAA_RGSS,
  92         k_EMSAA_8R
  93 };
  94
  95 #ifdef CSR_EXECUTABLE
  96
  97 // MSAA patterns
  98 v2f csr_msaa_1[] =
  99 {
 100         {0.f, 0.f}
 101 };
 102
 103 // XX
 104 // XX
 105 v2f csr_msaa_2x2[] =
 106 {
 107         {  0x0.4p0f,  0x0.4p0f },
 108         {  0x0.4p0f, -0x0.4p0f },
 109         { -0x0.4p0f, -0x0.4p0f },
 110         { -0x0.4p0f,  0x0.4p0f }
 111 };
 112
 113 //   X
 114 // X
 115 //    X
 116 // X
 117 v2f csr_msaa_2x2rgss[] =
 118 {
 119         {  0x0.2p0f,  0x0.6p0f },
 120         { -0x0.6p0f,  0x0.2p0f },
 121         { -0x0.2p0f, -0x0.6p0f },
 122         {  0x0.6p0f, -0x0.2p0f }
 123 };
 124
 125 //     X
 126 //   X
 127 // X
 128 //       X
 129 //  X
 130 //        X
 131 //      X
 132 //    X
 133 v2f csr_msaa_8rook[] =
 134 {
 135         {  0x0.1p0f,  0x0.7p0f },
 136         {  0x0.5p0f,  0x0.1p0f },
 137         {  0x0.7p0f, -0x0.3p0f },
 138         {  0x0.3p0f, -0x0.5p0f },
 139         { -0x0.1p0f, -0x0.7p0f },
 140         { -0x0.5p0f, -0x0.1p0f },
 141         { -0x0.7p0f,  0x0.3p0f },
 142         { -0x0.3p0f,  0x0.5p0f }
 143 };
 144
 145
 146 void csr_create_target( csr_target *rt, u32 x, u32 y, EMSAA aa, csr_shader *shader )
 147 {
 148         rt->x = x;
 149         rt->y = y;
 150
 151         switch( aa )
 152         {
 153                 default:
 154                 case k_EMSAA_none:
 155                         rt->num_samples = 1;
 156                         rt->sample_src = csr_msaa_1;
 157                         break;
 158
 159                 case k_EMSAA_2x2:
 160                         rt->num_samples = 4;
 161                         rt->sample_src = csr_msaa_2x2;
 162                         break;
 163
 164                 case k_EMSAA_RGSS:
 165                         rt->num_samples = 4;
 166                         rt->sample_src = csr_msaa_2x2rgss;
 167                         break;
 168
 169                 case k_EMSAA_8R:
 170                         rt->num_samples = 8;
 171                         rt->sample_src = csr_msaa_8rook;
 172                         break;
 173         }
 174
 175         rt->shader = shader;
 176         rt->depth = (float *)csr_malloc( x*y*rt->num_samples * sizeof(float) );
 177         rt->colour = csr_malloc( x * y * rt->shader->stride * rt->num_samples );
 178
 179         v3_fill( rt->bounds[0], INFINITY );
 180         v3_fill( rt->bounds[1], -INFINITY );
 181 }
 182
 183 void csr_update_subsamples( csr_target *rt )
 184 {
 185         float range_x = (rt->bounds[1][0]-rt->bounds[0][0]);
 186         float range_y = (rt->bounds[1][1]-rt->bounds[0][1]);
 187
 188         v2f pixel_size = { range_x/(float)rt->x, range_y/(float)rt->y };
 189
 190         for( int i = 0; i < rt->num_samples; i ++ )
 191         {
 192                 v2_mul( rt->sample_src[i], pixel_size, rt->subsamples[i] );
 193         }
 194 }
 195
 196 void csr_rt_free( csr_target *rt )
 197 {
 198         free( rt->depth );
 199         free( rt->colour );
 200 }
 201
 202 void csr_rt_clear( csr_target *rt )
 203 {
 204         for( u32 i = 0; i < rt->x*rt->y*rt->num_samples; i ++ )
 205         {
 206                 rt->shader->clear( rt->colour + i * rt->shader->stride );
 207                 rt->depth[i] = 0.f;
 208         }
 209 }
 210
 211 void csr_auto_fit( csr_target *rt, float padding )
 212 {
 213         // Correct aspect ratio to be square
 214         float dx, dy, l, cx, cy;
 215
 216         dx = rt->bounds[1][0] - rt->bounds[0][0];
 217         dy = rt->bounds[1][1] - rt->bounds[0][1];
 218
 219         l = fmaxf( dx, dy ) * .5f;
 220
 221         cx = (rt->bounds[1][0] + rt->bounds[0][0]) * .5f;
 222         cy = (rt->bounds[1][1] + rt->bounds[0][1]) * .5f;
 223
 224         rt->bounds[0][0] = cx - l - padding;
 225         rt->bounds[1][0] = cx + l + padding;
 226         rt->bounds[0][1] = cy - l - padding;
 227         rt->bounds[1][1] = cy + l + padding;
 228
 229         rt->scale = l + padding;
 230
 231         csr_update_subsamples( rt );
 232 }
 233
 234 void csr_write_txt( char const *path, const char *name, csr_target *rt  )
 235 {
 236         FILE *write_ptr;
 237
 238         write_ptr = fopen( path, "w" );
 239
 240         fprintf( write_ptr, "\"%s\"\n\{\n", name );
 241         fprintf( write_ptr, "\t\"material\" \"overviews/%s\"\n", name );
 242         fprintf( write_ptr, "\t\"pos_x\" \"%.8f\"\n", rt->bounds[0][0] );
 243         fprintf( write_ptr, "\t\"pos_y\" \"%.8f\"\n", rt->bounds[0][1] );
 244         fprintf( write_ptr, "\t\"scale\" \"%.8f\"\n", rt->scale / (float)rt->x );
 245         fprintf( write_ptr, "}\n" );
 246
 247         fclose( write_ptr );
 248 }
 249
 250 void simple_raster( csr_target *rt, vmf_vert tri[3] )
 251 {
 252         // Very very simplified rasterizing algorithm
 253         v2f bmin = { 0.f, 0.f };
 254         v2f bmax = { rt->x, rt->y };
 255
 256         v2_minv( tri[0].co, tri[1].co, bmin );
 257         v2_minv( tri[2].co, bmin, bmin );
 258
 259         v2_maxv( tri[0].co, tri[1].co, bmax );
 260         v2_maxv( tri[2].co, bmax, bmax );
 261
 262         float range_x = (rt->bounds[1][0]-rt->bounds[0][0])/(float)rt->x;
 263         float range_y = (rt->bounds[1][1]-rt->bounds[0][1])/(float)rt->y;
 264
 265         int start_x = csr_min( rt->x-1, csr_max( 0,       floorf( (bmin[0]-rt->bounds[0][0])/range_x)));
 266         int end_x =   csr_max( 0,       csr_min( rt->x-1,  ceilf( (bmax[0]-rt->bounds[0][0])/range_x)));
 267         int start_y = csr_min( rt->y-1, csr_max( 0,       floorf( (bmin[1]-rt->bounds[0][1])/range_y)));
 268         int end_y =   csr_max( 0,       csr_min( rt->y-1,  ceilf( (bmax[1]-rt->bounds[0][1])/range_y)));
 269
 270         v2f v0, v1, v2, vp;
 271         float d, bca = 0.f, bcb = 0.f, bcc = 0.f;
 272
 273         v2_sub( tri[1].co, tri[0].co, v0 );
 274         v2_sub( tri[2].co, tri[0].co, v1 );
 275         v2_sub( tri[1].co, tri[2].co, v2 );
 276         d = 1.f / (v0[0]*v1[1] - v1[0]*v0[1]);
 277
 278         // Backface culling
 279         if( v2_cross( v0, v1 ) > 0.f )
 280                 return;
 281
 282         v2f trace_origin;
 283
 284         for( u32 py = start_y; py <= end_y; py ++ )
 285         {
 286                 trace_origin[1] = csr_lerpf( rt->bounds[0][1], rt->bounds[1][1], (float)py/(float)rt->y );
 287
 288                 for( u32 px = start_x; px <= end_x; px ++ )
 289                 {
 290                         u32 sample_index = ((rt->y-py-1)*rt->x+px) * rt->num_samples;
 291
 292                         void *frag = rt->colour + sample_index*rt->shader->stride;
 293                         float *depth = &rt->depth[ sample_index ];
 294
 295                         trace_origin[0] = csr_lerpf( rt->bounds[0][0], rt->bounds[1][0], (float)px/(float)rt->x );
 296
 297                         // Determine coverage
 298                         for( int i = 0; i < rt->num_samples; i ++ )
 299                         {
 300                                 v3f sample_origin;
 301
 302                                 v2_add( rt->subsamples[ i ], trace_origin, sample_origin );
 303                                 v2_sub( sample_origin, tri[0].co, vp );
 304
 305                                 if( v2_cross( v0, vp ) > 0.f )
 306                                         continue;
 307                                 if( v2_cross( vp, v1 ) > 0.f )
 308                                         continue;
 309
 310                                 v2f vp2;
 311                                 v2_sub( sample_origin, tri[2].co, vp2 );
 312
 313                                 if( v2_cross( vp2, v2 ) > 0.f )
 314                                         continue;
 315
 316                                 bcb = (vp[0]*v1[1] - v1[0]*vp[1]) * d;
 317                                 bcc = (v0[0]*vp[1] - vp[0]*v0[1]) * d;
 318                                 bca = 1.f - bcb - bcc;
 319
 320                                 float hit = (tri[0].co[2] * bca + tri[1].co[2] * bcb + tri[2].co[2] * bcc) +16385.f;
 321
 322                                 if( hit > depth[i] )
 323                                 {
 324                                         depth[i] = hit;
 325                                         rt->shader->frag( frag+i*rt->shader->stride, tri, bca, bcb, bcc );
 326                                 }
 327                         }
 328                 }
 329         }
 330 }
 331
 332 void csr_draw( csr_target *rt, vmf_vert *triangles, u32 triangle_count, m4x3f transform )
 333 {
 334         m3x3f normal;
 335         vmf_vert new_tri[3];
 336
 337         // Derive normal matrix
 338         m4x3_to_3x3( transform, normal );
 339         m3x3_inv_transpose( normal, normal );
 340
 341         for( u32 i = 0; i < triangle_count; i ++ )
 342         {
 343                 vmf_vert *triangle = triangles + i*3;
 344
 345                 m4x3_mulv( transform, triangle[0].co, new_tri[0].co );
 346                 m4x3_mulv( transform, triangle[1].co, new_tri[1].co );
 347                 m4x3_mulv( transform, triangle[2].co, new_tri[2].co );
 348
 349                 m3x3_mulv( normal, triangle[0].nrm, new_tri[0].nrm );
 350                 m3x3_mulv( normal, triangle[1].nrm, new_tri[1].nrm );
 351                 m3x3_mulv( normal, triangle[2].nrm, new_tri[2].nrm );
 352
 353                 v3_normalize( new_tri[0].nrm );
 354                 v3_normalize( new_tri[1].nrm );
 355                 v3_normalize( new_tri[2].nrm );
 356
 357                 m4x3_mulv( transform, triangles[0].origin, new_tri[0].origin );
 358
 359                 simple_raster( rt, new_tri );
 360         }
 361 }
 362
 363 void csr_vmf_render( csr_target *rt, vmf_map *map, vdf_node *root, csr_filter *filter, m4x3f prev, m4x3f inst )
 364 {
 365         m4x3f transform = M4X3_IDENTITY;
 366         vmf_solid solid;
 367         vmf_vert tri[3];
 368         boxf trf_bounds;
 369
 370         u32 group_id = 0;
 371         int filter_visgroups = 0, filter_classname = 0, compute_bounds_only = 0;
 372
 373         if( filter )
 374         {
 375                 if( filter->visgroup )
 376                 {
 377                         filter_visgroups = 1;
 378                         group_id = vmf_visgroup_id( root, filter->visgroup );
 379                 }
 380
 381                 if( filter->classname )
 382                 {
 383                         filter_classname = 1;
 384                 }
 385
 386                 compute_bounds_only = filter->compute_bounds_only;
 387         }
 388
 389         // Multiply previous transform with instance transform to create basis
 390         if( prev )
 391         {
 392                 m4x3_mul( prev, inst, transform );
 393         }
 394
 395         // Gather world brushes
 396         solidgen_ctx_init( &solid );
 397
 398         if( !filter_classname )
 399         {
 400                 vdf_node *world = vdf_next( root, "world", NULL );
 401
 402                 vdf_foreach( world, "solid", brush )
 403                 {
 404                         if( filter_visgroups && !vmf_visgroup_match( brush, group_id ) )
 405                                 continue;
 406
 407                         // TODO: heap-use-after-free
 408                         solidgen_push( &solid, brush );
 409                 }
 410         }
 411
 412         // Actual entity loop
 413         m4x3f model;
 414
 415         vdf_foreach( root, "entity", ent )
 416         {
 417                 if( filter_visgroups && !vmf_visgroup_match( ent, group_id ) )
 418                         continue;
 419
 420                 if( filter_classname )
 421                         if( strcmp( kv_get( ent, "classname", "" ), filter->classname ) )
 422                                 continue;
 423
 424                 if( ent->user & VMF_FLAG_IS_PROP )
 425                 {
 426                         // Create model transform
 427                         m4x3_identity( model );
 428
 429                         vmf_entity_transform( ent, model );
 430                         m4x3_mul( transform, model, model );
 431
 432                         // Draw model
 433                         mdl_mesh_t *mdl = &map->models[ ent->user1 ].mdl;
 434
 435                         if( compute_bounds_only )
 436                         {
 437                                 map->models[ ent->user1 ].need_load = 1;
 438                                 m4x3_expand_aabb_point( model, rt->bounds, (v3f){0.f,0.f,0.f} );
 439                         }
 440                         else
 441                         {
 442                                 for( int i = 0; i < mdl->num_indices/3; i ++ )
 443                                 {
 444                                         for( int j = 0; j < 3; j ++ )
 445                                         {
 446                                                 v3_copy( &mdl->vertices[ mdl->indices[ i*3+j ] *8 ],   tri[j].co );
 447                                                 v3_copy( &mdl->vertices[ mdl->indices[ i*3+j ] *8+3 ], tri[j].nrm );
 448                                                 v3_zero( tri[j].origin );
 449                                         }
 450
 451                                         csr_draw( rt, tri, 1, model );
 452                                 }
 453                         }
 454                 }
 455                 else if( ent->user & VMF_FLAG_IS_INSTANCE )
 456                 {
 457                         m4x3_identity( model );
 458                         vmf_entity_transform( ent, model );
 459
 460                         csr_vmf_render( rt, map, map->cache[ ent->user1 ].root, filter, transform, model );
 461                 }
 462                 else
 463                 {
 464                         // Brush entity
 465                         vdf_foreach( ent, "solid", ent_solid )
 466                         {
 467                                 solidgen_push( &solid, ent_solid );
 468                         }
 469                 }
 470         }
 471
 472         if( compute_bounds_only )
 473         {
 474                 solidgen_bounds( &solid, trf_bounds );
 475                 m4x3_transform_aabb( transform, trf_bounds );
 476                 box_concat( rt->bounds, trf_bounds );
 477         }
 478         else
 479         {
 480                 // Draw brushes
 481                 for( int i = 0; i < csr_sb_count( solid.indices )/3; i ++ )
 482                 {
 483                         u32 * base = solid.indices + i*3;
 484
 485                         tri[0] = solid.verts[ base[0] ];
 486                         tri[1] = solid.verts[ base[1] ];
 487                         tri[2] = solid.verts[ base[2] ];
 488
 489                         csr_draw( rt, tri, 1, transform );
 490                 }
 491         }
 492
 493         solidgen_ctx_reset( &solid );
 494         solidgen_ctx_free( &solid );
 495 }
 496
 497 // Obsolete
 498 void csr_rt_save_buffers( csr_target *rt, const char *basename, const char *subname )
 499 {
 500         char output[ 512 ];
 501
 502         float *image = (float *)csr_malloc( 1024*1024*sizeof(float)*3 );
 503
 504         float contrib = 1.f/(float)rt->num_samples;
 505
 506         for( int l = 0; l < rt->x; l ++ )
 507         {
 508                 for( int x = 0; x < rt->y; x ++ )
 509                 {
 510                         float *dst = &image[ (l*1024+x)*3 ];
 511                         void *src = rt->colour + ((1023-l)*1024+x) * rt->num_samples * rt->shader->stride;
 512
 513                         v3_muls( (float *)src, contrib, dst );
 514
 515                         for( int j = 1; j < rt->num_samples; j ++ )
 516                         {
 517                                 v3_muladds( dst, (float *)(src + j*rt->shader->stride), contrib, dst );
 518                         }
 519                 }
 520         }
 521
 522         // Save position buffer
 523         strcpy( output, basename );
 524         strcat( output, "." );
 525         strcat( output, subname );
 526         strcat( output, "_position.pfm" );
 527         csr_32f_write( output, rt->x, rt->y, image );
 528
 529         free( image );
 530 }
 531
 532 // Save floating point buffer to tga. Must be in range (0-1)
 533 // Offset and stride are in bytes
 534 void csr_rt_save_tga( csr_target *rt, const char *path, u32 offset, u32 nc )
 535 {
 536         u8 *image = (u8 *)csr_malloc( rt->x*rt->y * 4 );
 537
 538         float contrib = 255.f/(float)rt->num_samples;
 539
 540         for( int y = 0; y < rt->y; y ++ )
 541         {
 542                 for( int x = 0; x < rt->x; x ++ )
 543                 {
 544                         u32 pixel_index = (y*rt->x + x);
 545
 546                         void *src = rt->colour + offset + pixel_index * rt->num_samples * rt->shader->stride;
 547                         u8 *dst = image + pixel_index*4;
 548
 549                         v4f accum = { 0.f, 0.f, 0.f, 0.f };
 550
 551                         for( int k = 0; k < rt->num_samples; k ++ )
 552                         {
 553                                 float *src_sample = (float *)(src + k*rt->shader->stride);
 554
 555                                 for( int j = 0; j < nc; j ++ )
 556                                 {
 557                                         accum[ j ] += src_sample[ j ] * contrib;
 558                                 }
 559                         }
 560
 561                         // TODO: Clamp this value
 562                         dst[0] = accum[0];
 563                         dst[1] = accum[1];
 564                         dst[2] = accum[2];
 565                         dst[3] = accum[3];
 566                 }
 567         }
 568
 569         csr_tga_write( path, rt->x, rt->y, nc, image );
 570         free( image );
 571 }
 572
 573 #endif