source/blender/render/intern/source/rendercore.c

   1 /**
   2  * $Id$
   3  *
   4  * ***** BEGIN GPL LICENSE BLOCK *****
   5  *
   6  * This program is free software; you can redistribute it and/or
   7  * modify it under the terms of the GNU General Public License
   8  * as published by the Free Software Foundation; either version 2
   9  * of the License, or (at your option) any later version.
  10  *
  11  * This program is distributed in the hope that it will be useful,
  12  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  13  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  14  * GNU General Public License for more details.
  15  *
  16  * You should have received a copy of the GNU General Public License
  17  * along with this program; if not, write to the Free Software Foundation,
  18  * Inc., 59 Temple Place - Suite 330, Boston, MA  02111-1307, USA.
  19  *
  20  * The Original Code is Copyright (C) 2001-2002 by NaN Holding BV.
  21  * All rights reserved.
  22  *
  23  * Contributors: Hos, Robert Wenzlaff.
  24  * Contributors: 2004/2005/2006 Blender Foundation, full recode
  25  *
  26  * ***** END GPL LICENSE BLOCK *****
  27  */
  28
  29 /* system includes */
  30 #include <stdio.h>
  31 #include <math.h>
  32 #include <float.h>
  33 #include <string.h>
  34
  35 /* External modules: */
  36 #include "MEM_guardedalloc.h"
  37
  38 #include "BLI_arithb.h"
  39 #include "BLI_blenlib.h"
  40 #include "BLI_jitter.h"
  41 #include "BLI_rand.h"
  42 #include "BLI_threads.h"
  43
  44 #include "BKE_utildefines.h"
  45
  46 #include "DNA_image_types.h"
  47 #include "DNA_lamp_types.h"
  48 #include "DNA_material_types.h"
  49 #include "DNA_meshdata_types.h"
  50 #include "DNA_group_types.h"
  51
  52 #include "BKE_global.h"
  53 #include "BKE_image.h"
  54 #include "BKE_main.h"
  55 #include "BKE_node.h"
  56 #include "BKE_texture.h"
  57
  58 #include "IMB_imbuf_types.h"
  59 #include "IMB_imbuf.h"
  60
  61 /* local include */
  62 #include "renderpipeline.h"
  63 #include "render_types.h"
  64 #include "renderdatabase.h"
  65 #include "occlusion.h"
  66 #include "pixelblending.h"
  67 #include "pixelshading.h"
  68 #include "shadbuf.h"
  69 #include "shading.h"
  70 #include "sss.h"
  71 #include "zbuf.h"
  72 #include "RE_raytrace.h"
  73
  74 #include "PIL_time.h"
  75
  76 /* own include */
  77 #include "rendercore.h"
  78
  79
  80 /* ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ */
  81 /* defined in pipeline.c, is hardcopy of active dynamic allocated Render */
  82 /* only to be used here in this file, it's for speed */
  83 extern struct Render R;
  84 /* ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ */
  85
  86 /* x and y are current pixels in rect to be rendered */
  87 /* do not normalize! */
  88 void calc_view_vector(float *view, float x, float y)
  89 {
  90
  91         view[2]= -ABS(R.clipsta);
  92
  93         if(R.r.mode & R_ORTHO) {
  94                 view[0]= view[1]= 0.0f;
  95         }
  96         else {
  97
  98                 if(R.r.mode & R_PANORAMA)
  99                         x-= R.panodxp;
 100
 101                 /* move x and y to real viewplane coords */
 102                 x= (x/(float)R.winx);
 103                 view[0]= R.viewplane.xmin + x*(R.viewplane.xmax - R.viewplane.xmin);
 104
 105                 y= (y/(float)R.winy);
 106                 view[1]= R.viewplane.ymin + y*(R.viewplane.ymax - R.viewplane.ymin);
 107
 108 //              if(R.flag & R_SEC_FIELD) {
 109 //                      if(R.r.mode & R_ODDFIELD) view[1]= (y+R.ystart)*R.ycor;
 110 //                      else view[1]= (y+R.ystart+1.0)*R.ycor;
 111 //              }
 112 //              else view[1]= (y+R.ystart+R.bluroffsy+0.5)*R.ycor;
 113
 114                 if(R.r.mode & R_PANORAMA) {
 115                         float u= view[0] + R.panodxv; float v= view[2];
 116                         view[0]= R.panoco*u + R.panosi*v;
 117                         view[2]= -R.panosi*u + R.panoco*v;
 118                 }
 119         }
 120 }
 121
 122 void calc_renderco_ortho(float *co, float x, float y, int z)
 123 {
 124         /* x and y 3d coordinate can be derived from pixel coord and winmat */
 125         float fx= 2.0f/(R.winx*R.winmat[0][0]);
 126         float fy= 2.0f/(R.winy*R.winmat[1][1]);
 127         float zco;
 128
 129         co[0]= (x - 0.5f*R.winx)*fx - R.winmat[3][0]/R.winmat[0][0];
 130         co[1]= (y - 0.5f*R.winy)*fy - R.winmat[3][1]/R.winmat[1][1];
 131
 132         zco= ((float)z)/2147483647.0f;
 133         co[2]= R.winmat[3][2]/( R.winmat[2][3]*zco - R.winmat[2][2] );
 134 }
 135
 136 void calc_renderco_zbuf(float *co, float *view, int z)
 137 {
 138         float fac, zco;
 139
 140         /* inverse of zbuf calc: zbuf = MAXZ*hoco_z/hoco_w */
 141         zco= ((float)z)/2147483647.0f;
 142         co[2]= R.winmat[3][2]/( R.winmat[2][3]*zco - R.winmat[2][2] );
 143
 144         fac= co[2]/view[2];
 145         co[0]= fac*view[0];
 146         co[1]= fac*view[1];
 147 }
 148
 149 /* also used in zbuf.c and shadbuf.c */
 150 int count_mask(unsigned short mask)
 151 {
 152         if(R.samples)
 153                 return (R.samples->cmask[mask & 255]+R.samples->cmask[mask>>8]);
 154         return 0;
 155 }
 156
 157 static int calchalo_z(HaloRen *har, int zz)
 158 {
 159
 160         if(har->type & HA_ONLYSKY) {
 161                 if(zz < 0x7FFFFFF0) zz= - 0x7FFFFF;     /* edge render messes zvalues */
 162         }
 163         else {
 164                 zz= (zz>>8);
 165         }
 166         return zz;
 167 }
 168
 169
 170
 171 static void halo_pixelstruct(HaloRen *har, RenderLayer **rlpp, int totsample, int od, float dist, float xn, float yn, PixStr *ps)
 172 {
 173         float col[4], accol[4], fac;
 174         int amount, amountm, zz, flarec, sample, fullsample, mask=0;
 175
 176         fullsample= (totsample > 1);
 177         amount= 0;
 178         accol[0]=accol[1]=accol[2]=accol[3]= 0.0f;
 179         flarec= har->flarec;
 180
 181         while(ps) {
 182                 amountm= count_mask(ps->mask);
 183                 amount+= amountm;
 184
 185                 zz= calchalo_z(har, ps->z);
 186                 if((zz> har->zs) || (har->mat && (har->mat->mode & MA_HALO_SOFT))) {
 187                         if(shadeHaloFloat(har, col, zz, dist, xn, yn, flarec)) {
 188                                 flarec= 0;
 189
 190                                 if(fullsample) {
 191                                         for(sample=0; sample<totsample; sample++)
 192                                                 if(ps->mask & (1 << sample))
 193                                                         addalphaAddfacFloat(rlpp[sample]->rectf + od*4, col, har->add);
 194                                 }
 195                                 else {
 196                                         fac= ((float)amountm)/(float)R.osa;
 197                                         accol[0]+= fac*col[0];
 198                                         accol[1]+= fac*col[1];
 199                                         accol[2]+= fac*col[2];
 200                                         accol[3]+= fac*col[3];
 201                                 }
 202                         }
 203                 }
 204
 205                 mask |= ps->mask;
 206                 ps= ps->next;
 207         }
 208
 209         /* now do the sky sub-pixels */
 210         amount= R.osa-amount;
 211         if(amount) {
 212                 if(shadeHaloFloat(har, col, 0x7FFFFF, dist, xn, yn, flarec)) {
 213                         if(!fullsample) {
 214                                 fac= ((float)amount)/(float)R.osa;
 215                                 accol[0]+= fac*col[0];
 216                                 accol[1]+= fac*col[1];
 217                                 accol[2]+= fac*col[2];
 218                                 accol[3]+= fac*col[3];
 219                         }
 220                 }
 221         }
 222
 223         if(fullsample) {
 224                 for(sample=0; sample<totsample; sample++)
 225                         if(!(mask & (1 << sample)))
 226                                 addalphaAddfacFloat(rlpp[sample]->rectf + od*4, col, har->add);
 227         }
 228         else {
 229                 col[0]= accol[0];
 230                 col[1]= accol[1];
 231                 col[2]= accol[2];
 232                 col[3]= accol[3];
 233
 234                 for(sample=0; sample<totsample; sample++)
 235                         addalphaAddfacFloat(rlpp[sample]->rectf + od*4, col, har->add);
 236         }
 237 }
 238
 239 static void halo_tile(RenderPart *pa, RenderLayer *rl)
 240 {
 241         RenderLayer *rlpp[RE_MAX_OSA];
 242         HaloRen *har;
 243         rcti disprect= pa->disprect, testrect= pa->disprect;
 244         float dist, xsq, ysq, xn, yn;
 245         float col[4];
 246         intptr_t *rd= NULL;
 247         int a, *rz, zz, y, sample, totsample, od;
 248         short minx, maxx, miny, maxy, x;
 249         unsigned int lay= rl->lay;
 250
 251         /* we don't render halos in the cropped area, gives errors in flare counter */
 252         if(pa->crop) {
 253                 testrect.xmin+= pa->crop;
 254                 testrect.xmax-= pa->crop;
 255                 testrect.ymin+= pa->crop;
 256                 testrect.ymax-= pa->crop;
 257         }
 258
 259         totsample= get_sample_layers(pa, rl, rlpp);
 260
 261         for(a=0; a<R.tothalo; a++) {
 262                 har= R.sortedhalos[a];
 263
 264                 /* layer test, clip halo with y */
 265                 if((har->lay & lay)==0);
 266                 else if(testrect.ymin > har->maxy);
 267                 else if(testrect.ymax < har->miny);
 268                 else {
 269
 270                         minx= floor(har->xs-har->rad);
 271                         maxx= ceil(har->xs+har->rad);
 272
 273                         if(testrect.xmin > maxx);
 274                         else if(testrect.xmax < minx);
 275                         else {
 276
 277                                 minx= MAX2(minx, testrect.xmin);
 278                                 maxx= MIN2(maxx, testrect.xmax);
 279
 280                                 miny= MAX2(har->miny, testrect.ymin);
 281                                 maxy= MIN2(har->maxy, testrect.ymax);
 282
 283                                 for(y=miny; y<maxy; y++) {
 284                                         int rectofs= (y-disprect.ymin)*pa->rectx + (minx - disprect.xmin);
 285                                         rz= pa->rectz + rectofs;
 286                                         od= rectofs;
 287
 288                                         if(pa->rectdaps)
 289                                                 rd= pa->rectdaps + rectofs;
 290
 291                                         yn= (y-har->ys)*R.ycor;
 292                                         ysq= yn*yn;
 293
 294                                         for(x=minx; x<maxx; x++, rz++, od++) {
 295                                                 xn= x- har->xs;
 296                                                 xsq= xn*xn;
 297                                                 dist= xsq+ysq;
 298                                                 if(dist<har->radsq) {
 299                                                         if(rd && *rd) {
 300                                                                 halo_pixelstruct(har, rlpp, totsample, od, dist, xn, yn, (PixStr *)*rd);
 301                                                         }
 302                                                         else {
 303                                                                 zz= calchalo_z(har, *rz);
 304                                                                 if((zz> har->zs) || (har->mat && (har->mat->mode & MA_HALO_SOFT))) {
 305                                                                         if(shadeHaloFloat(har, col, zz, dist, xn, yn, har->flarec)) {
 306                                                                                 for(sample=0; sample<totsample; sample++)
 307                                                                                         addalphaAddfacFloat(rlpp[sample]->rectf + od*4, col, har->add);
 308                                                                         }
 309                                                                 }
 310                                                         }
 311                                                 }
 312                                                 if(rd) rd++;
 313                                         }
 314                                 }
 315                         }
 316                 }
 317                 if(R.test_break(R.tbh) ) break;
 318         }
 319 }
 320
 321 static void lamphalo_tile(RenderPart *pa, RenderLayer *rl)
 322 {
 323         RenderLayer *rlpp[RE_MAX_OSA];
 324         ShadeInput shi;
 325         float *pass;
 326         float fac, col[4];
 327         intptr_t *rd= pa->rectdaps;
 328         int *rz= pa->rectz;
 329         int x, y, sample, totsample, fullsample, od;
 330
 331         totsample= get_sample_layers(pa, rl, rlpp);
 332         fullsample= (totsample > 1);
 333
 334         shade_input_initialize(&shi, pa, rl, 0); /* this zero's ShadeInput for us */
 335
 336         for(od=0, y=pa->disprect.ymin; y<pa->disprect.ymax; y++) {
 337                 for(x=pa->disprect.xmin; x<pa->disprect.xmax; x++, rz++, od++) {
 338
 339                         calc_view_vector(shi.view, x, y);
 340
 341                         if(rd && *rd) {
 342                                 PixStr *ps= (PixStr *)*rd;
 343                                 int count, totsamp= 0, mask= 0;
 344
 345                                 while(ps) {
 346                                         if(R.r.mode & R_ORTHO)
 347                                                 calc_renderco_ortho(shi.co, (float)x, (float)y, ps->z);
 348                                         else
 349                                                 calc_renderco_zbuf(shi.co, shi.view, ps->z);
 350
 351                                         totsamp+= count= count_mask(ps->mask);
 352                                         mask |= ps->mask;
 353
 354                                         col[0]= col[1]= col[2]= col[3]= 0.0f;
 355                                         renderspothalo(&shi, col, 1.0f);
 356
 357                                         if(fullsample) {
 358                                                 for(sample=0; sample<totsample; sample++) {
 359                                                         if(ps->mask & (1 << sample)) {
 360                                                                 pass= rlpp[sample]->rectf + od*4;
 361                                                                 pass[0]+= col[0];
 362                                                                 pass[1]+= col[1];
 363                                                                 pass[2]+= col[2];
 364                                                                 pass[3]+= col[3];
 365                                                                 if(pass[3]>1.0f) pass[3]= 1.0f;
 366                                                         }
 367                                                 }
 368                                         }
 369                                         else {
 370                                                 fac= ((float)count)/(float)R.osa;
 371                                                 pass= rl->rectf + od*4;
 372                                                 pass[0]+= fac*col[0];
 373                                                 pass[1]+= fac*col[1];
 374                                                 pass[2]+= fac*col[2];
 375                                                 pass[3]+= fac*col[3];
 376                                                 if(pass[3]>1.0f) pass[3]= 1.0f;
 377                                         }
 378
 379                                         ps= ps->next;
 380                                 }
 381
 382                                 if(totsamp<R.osa) {
 383                                         shi.co[2]= 0.0f;
 384
 385                                         col[0]= col[1]= col[2]= col[3]= 0.0f;
 386                                         renderspothalo(&shi, col, 1.0f);
 387
 388                                         if(fullsample) {
 389                                                 for(sample=0; sample<totsample; sample++) {
 390                                                         if(!(mask & (1 << sample))) {
 391                                                                 pass= rlpp[sample]->rectf + od*4;
 392                                                                 pass[0]+= col[0];
 393                                                                 pass[1]+= col[1];
 394                                                                 pass[2]+= col[2];
 395                                                                 pass[3]+= col[3];
 396                                                                 if(pass[3]>1.0f) pass[3]= 1.0f;
 397                                                         }
 398                                                 }
 399                                         }
 400                                         else {
 401                                                 fac= ((float)R.osa-totsamp)/(float)R.osa;
 402                                                 pass= rl->rectf + od*4;
 403                                                 pass[0]+= fac*col[0];
 404                                                 pass[1]+= fac*col[1];
 405                                                 pass[2]+= fac*col[2];
 406                                                 pass[3]+= fac*col[3];
 407                                                 if(pass[3]>1.0f) pass[3]= 1.0f;
 408                                         }
 409                                 }
 410                         }
 411                         else {
 412                                 if(R.r.mode & R_ORTHO)
 413                                         calc_renderco_ortho(shi.co, (float)x, (float)y, *rz);
 414                                 else
 415                                         calc_renderco_zbuf(shi.co, shi.view, *rz);
 416
 417                                 col[0]= col[1]= col[2]= col[3]= 0.0f;
 418                                 renderspothalo(&shi, col, 1.0f);
 419
 420                                 for(sample=0; sample<totsample; sample++) {
 421                                         pass= rlpp[sample]->rectf + od*4;
 422                                         pass[0]+= col[0];
 423                                         pass[1]+= col[1];
 424                                         pass[2]+= col[2];
 425                                         pass[3]+= col[3];
 426                                         if(pass[3]>1.0f) pass[3]= 1.0f;
 427                                 }
 428                         }
 429
 430                         if(rd) rd++;
 431                 }
 432                 if(y&1)
 433                         if(R.test_break(R.tbh)) break;
 434         }
 435 }
 436
 437
 438 /* ********************* MAINLOOPS ******************** */
 439
 440 /* osa version */
 441 static void add_filt_passes(RenderLayer *rl, int curmask, int rectx, int offset, ShadeInput *shi, ShadeResult *shr)
 442 {
 443         RenderPass *rpass;
 444
 445         /* combined rgb */
 446         add_filt_fmask(curmask, shr->combined, rl->rectf + 4*offset, rectx);
 447
 448         for(rpass= rl->passes.first; rpass; rpass= rpass->next) {
 449                 float *fp, *col= NULL;
 450                 int pixsize= 3;
 451
 452                 switch(rpass->passtype) {
 453                         case SCE_PASS_Z:
 454                                 fp= rpass->rect + offset;
 455                                 *fp= shr->z;
 456                                 break;
 457                         case SCE_PASS_RGBA:
 458                                 col= shr->col;
 459                                 pixsize= 4;
 460                                 break;
 461                         case SCE_PASS_DIFFUSE:
 462                                 col= shr->diff;
 463                                 break;
 464                         case SCE_PASS_SPEC:
 465                                 col= shr->spec;
 466                                 break;
 467                         case SCE_PASS_SHADOW:
 468                                 col= shr->shad;
 469                                 break;
 470                         case SCE_PASS_AO:
 471                                 col= shr->ao;
 472                                 break;
 473                         case SCE_PASS_REFLECT:
 474                                 col= shr->refl;
 475                                 break;
 476                         case SCE_PASS_REFRACT:
 477                                 col= shr->refr;
 478                                 break;
 479                         case SCE_PASS_RADIO:
 480                                 col= shr->rad;
 481                                 break;
 482                         case SCE_PASS_NORMAL:
 483                                 col= shr->nor;
 484                                 break;
 485                         case SCE_PASS_UV:
 486                                 /* box filter only, gauss will screwup UV too much */
 487                                 if(shi->totuv) {
 488                                         float mult= (float)count_mask(curmask)/(float)R.osa;
 489                                         fp= rpass->rect + 3*offset;
 490                                         fp[0]+= mult*(0.5f + 0.5f*shi->uv[shi->actuv].uv[0]);
 491                                         fp[1]+= mult*(0.5f + 0.5f*shi->uv[shi->actuv].uv[1]);
 492                                         fp[2]+= mult;
 493                                 }
 494                                 break;
 495                         case SCE_PASS_INDEXOB:
 496                                 /* no filter */
 497                                 if(shi->vlr) {
 498                                         fp= rpass->rect + offset;
 499                                         if(*fp==0.0f)
 500                                                 *fp= (float)shi->obr->ob->index;
 501                                 }
 502                                 break;
 503                         case SCE_PASS_MIST:
 504                                 /*  */
 505                                 col= &shr->mist;
 506                                 pixsize= 1;
 507                                 break;
 508
 509                         case SCE_PASS_VECTOR:
 510                         {
 511                                 /* add minimum speed in pixel, no filter */
 512                                 fp= rpass->rect + 4*offset;
 513                                 if( (ABS(shr->winspeed[0]) + ABS(shr->winspeed[1]))< (ABS(fp[0]) + ABS(fp[1])) ) {
 514                                         fp[0]= shr->winspeed[0];
 515                                         fp[1]= shr->winspeed[1];
 516                                 }
 517                                 if( (ABS(shr->winspeed[2]) + ABS(shr->winspeed[3]))< (ABS(fp[2]) + ABS(fp[3])) ) {
 518                                         fp[2]= shr->winspeed[2];
 519                                         fp[3]= shr->winspeed[3];
 520                                 }
 521                         }
 522                                 break;
 523                 }
 524                 if(col) {
 525                         fp= rpass->rect + pixsize*offset;
 526                         add_filt_fmask_pixsize(curmask, col, fp, rectx, pixsize);
 527                 }
 528         }
 529 }
 530
 531 /* non-osa version */
 532 static void add_passes(RenderLayer *rl, int offset, ShadeInput *shi, ShadeResult *shr)
 533 {
 534         RenderPass *rpass;
 535         float *fp;
 536
 537         fp= rl->rectf + 4*offset;
 538         QUATCOPY(fp, shr->combined);
 539
 540         for(rpass= rl->passes.first; rpass; rpass= rpass->next) {
 541                 float *col= NULL, uvcol[3];
 542                 int a, pixsize= 3;
 543
 544                 switch(rpass->passtype) {
 545                         case SCE_PASS_Z:
 546                                 fp= rpass->rect + offset;
 547                                 *fp= shr->z;
 548                                 break;
 549                         case SCE_PASS_RGBA:
 550                                 col= shr->col;
 551                                 pixsize= 4;
 552                                 break;
 553                         case SCE_PASS_DIFFUSE:
 554                                 col= shr->diff;
 555                                 break;
 556                         case SCE_PASS_SPEC:
 557                                 col= shr->spec;
 558                                 break;
 559                         case SCE_PASS_SHADOW:
 560                                 col= shr->shad;
 561                                 break;
 562                         case SCE_PASS_AO:
 563                                 col= shr->ao;
 564                                 break;
 565                         case SCE_PASS_REFLECT:
 566                                 col= shr->refl;
 567                                 break;
 568                         case SCE_PASS_REFRACT:
 569                                 col= shr->refr;
 570                                 break;
 571                         case SCE_PASS_RADIO:
 572                                 col= shr->rad;
 573                                 break;
 574                         case SCE_PASS_NORMAL:
 575                                 col= shr->nor;
 576                                 break;
 577                         case SCE_PASS_UV:
 578                                 if(shi->totuv) {
 579                                         uvcol[0]= 0.5f + 0.5f*shi->uv[shi->actuv].uv[0];
 580                                         uvcol[1]= 0.5f + 0.5f*shi->uv[shi->actuv].uv[1];
 581                                         uvcol[2]= 1.0f;
 582                                         col= uvcol;
 583                                 }
 584                                 break;
 585                         case SCE_PASS_VECTOR:
 586                                 col= shr->winspeed;
 587                                 pixsize= 4;
 588                                 break;
 589                         case SCE_PASS_INDEXOB:
 590                                 if(shi->vlr) {
 591                                         fp= rpass->rect + offset;
 592                                         *fp= (float)shi->obr->ob->index;
 593                                 }
 594                                 break;
 595                         case SCE_PASS_MIST:
 596                                 fp= rpass->rect + offset;
 597                                 *fp= shr->mist;
 598                                 break;
 599                 }
 600                 if(col) {
 601                         fp= rpass->rect + pixsize*offset;
 602                         for(a=0; a<pixsize; a++)
 603                                 fp[a]= col[a];
 604                 }
 605         }
 606 }
 607
 608 int get_sample_layers(RenderPart *pa, RenderLayer *rl, RenderLayer **rlpp)
 609 {
 610
 611         if(pa->fullresult.first) {
 612                 int sample, nr= BLI_findindex(&pa->result->layers, rl);
 613
 614                 for(sample=0; sample<R.osa; sample++) {
 615                         RenderResult *rr= BLI_findlink(&pa->fullresult, sample);
 616
 617                         rlpp[sample]= BLI_findlink(&rr->layers, nr);
 618                 }
 619                 return R.osa;
 620         }
 621         else {
 622                 rlpp[0]= rl;
 623                 return 1;
 624         }
 625 }
 626
 627
 628 /* only do sky, is default in the solid layer (shade_tile) btw */
 629 static void sky_tile(RenderPart *pa, RenderLayer *rl)
 630 {
 631         RenderLayer *rlpp[RE_MAX_OSA];
 632         int x, y, od=0, totsample;
 633
 634         if(R.r.alphamode!=R_ADDSKY)
 635                 return;
 636
 637         totsample= get_sample_layers(pa, rl, rlpp);
 638
 639         for(y=pa->disprect.ymin; y<pa->disprect.ymax; y++) {
 640                 for(x=pa->disprect.xmin; x<pa->disprect.xmax; x++, od+=4) {
 641                         float col[4];
 642                         int sample, done= 0;
 643
 644                         for(sample= 0; sample<totsample; sample++) {
 645                                 float *pass= rlpp[sample]->rectf + od;
 646
 647                                 if(pass[3]<1.0f) {
 648
 649                                         if(done==0) {
 650                                                 shadeSkyPixel(col, x, y, pa->thread);
 651                                                 done= 1;
 652                                         }
 653
 654                                         if(pass[3]==0.0f) {
 655                                                 QUATCOPY(pass, col);
 656                                         }
 657                                         else {
 658                                                 addAlphaUnderFloat(pass, col);
 659                                         }
 660                                 }
 661                         }
 662                 }
 663
 664                 if(y&1)
 665                         if(R.test_break(R.tbh)) break;
 666         }
 667 }
 668
 669 static void atm_tile(RenderPart *pa, RenderLayer *rl)
 670 {
 671         RenderPass *zpass;
 672         GroupObject *go;
 673         LampRen *lar;
 674         RenderLayer *rlpp[RE_MAX_OSA];
 675         int totsample;
 676         int x, y, od= 0;
 677
 678         totsample= get_sample_layers(pa, rl, rlpp);
 679
 680         /* check that z pass is enabled */
 681         if(pa->rectz==NULL) return;
 682         for(zpass= rl->passes.first; zpass; zpass= zpass->next)
 683                 if(zpass->passtype==SCE_PASS_Z)
 684                         break;
 685
 686         if(zpass==NULL) return;
 687
 688         /* check for at least one sun lamp that its atmosphere flag is is enabled */
 689         for(go=R.lights.first; go; go= go->next) {
 690                 lar= go->lampren;
 691                 if(lar->type==LA_SUN && lar->sunsky && (lar->sunsky->effect_type & LA_SUN_EFFECT_AP))
 692                         break;
 693         }
 694         /* do nothign and return if there is no sun lamp */
 695         if(go==NULL)
 696                 return;
 697
 698         /* for each x,y and each sample, and each sun lamp*/
 699         for(y=pa->disprect.ymin; y<pa->disprect.ymax; y++) {
 700                 for(x=pa->disprect.xmin; x<pa->disprect.xmax; x++, od++) {
 701                         int sample;
 702
 703                         for(sample=0; sample<totsample; sample++) {
 704                                 float *zrect= RE_RenderLayerGetPass(rlpp[sample], SCE_PASS_Z) + od;
 705                                 float *rgbrect = rlpp[sample]->rectf + 4*od;
 706                                 float rgb[3];
 707                                 int done= 0;
 708
 709                                 for(go=R.lights.first; go; go= go->next) {
 710
 711
 712                                         lar= go->lampren;
 713                                         if(lar->type==LA_SUN && lar->sunsky) {
 714
 715                                                 /* if it's sky continue and don't apply atmosphere effect on it */
 716                                                 if(*zrect >= 9.9e10 || rgbrect[3]==0.0f) {
 717                                                         continue;
 718                                                 }
 719
 720                                                 if((lar->sunsky->effect_type & LA_SUN_EFFECT_AP)) {
 721                                                         float tmp_rgb[3];
 722
 723                                                         VECCOPY(tmp_rgb, rgbrect);
 724                                                         if(rgbrect[3]!=1.0f) {  /* de-premul */
 725                                                                 float div= 1.0f/rgbrect[3];
 726                                                                 VECMUL(tmp_rgb, div);
 727                                                         }
 728                                                         shadeAtmPixel(lar->sunsky, tmp_rgb, x, y, *zrect);
 729                                                         if(rgbrect[3]!=1.0f) {  /* premul */
 730                                                                 VECMUL(tmp_rgb, rgbrect[3]);
 731                                                         }
 732
 733                                                         if(done==0) {
 734                                                                 VECCOPY(rgb, tmp_rgb);
 735                                                                 done = 1;
 736                                                         }
 737                                                         else{
 738                                                                 rgb[0] = 0.5f*rgb[0] + 0.5f*tmp_rgb[0];
 739                                                                 rgb[1] = 0.5f*rgb[1] + 0.5f*tmp_rgb[1];
 740                                                                 rgb[2] = 0.5f*rgb[2] + 0.5f*tmp_rgb[2];
 741                                                         }
 742                                                 }
 743                                         }
 744                                 }
 745
 746                                 /* if at least for one sun lamp aerial perspective was applied*/
 747                                 if(done) {
 748                                         VECCOPY(rgbrect, rgb);
 749                                 }
 750                         }
 751                 }
 752         }
 753 }
 754
 755 static void shadeDA_tile(RenderPart *pa, RenderLayer *rl)
 756 {
 757         RenderResult *rr= pa->result;
 758         ShadeSample ssamp;
 759         intptr_t *rd, *rectdaps= pa->rectdaps;
 760         int samp;
 761         int x, y, seed, crop=0, offs=0, od;
 762
 763         if(R.test_break(R.tbh)) return;
 764
 765         /* irregular shadowb buffer creation */
 766         if(R.r.mode & R_SHADOW)
 767                 ISB_create(pa, NULL);
 768
 769         /* we set per pixel a fixed seed, for random AO and shadow samples */
 770         seed= pa->rectx*pa->disprect.ymin;
 771
 772         /* general shader info, passes */
 773         shade_sample_initialize(&ssamp, pa, rl);
 774
 775         /* occlusion caching */
 776         if(R.occlusiontree)
 777                 cache_occ_samples(&R, pa, &ssamp);
 778
 779         /* filtered render, for now we assume only 1 filter size */
 780         if(pa->crop) {
 781                 crop= 1;
 782                 rectdaps+= pa->rectx + 1;
 783                 offs= pa->rectx + 1;
 784         }
 785
 786         /* scanline updates have to be 2 lines behind */
 787         rr->renrect.ymin= 0;
 788         rr->renrect.ymax= -2*crop;
 789         rr->renlay= rl;
 790
 791         for(y=pa->disprect.ymin+crop; y<pa->disprect.ymax-crop; y++, rr->renrect.ymax++) {
 792                 rd= rectdaps;
 793                 od= offs;
 794
 795                 for(x=pa->disprect.xmin+crop; x<pa->disprect.xmax-crop; x++, rd++, od++) {
 796                         BLI_thread_srandom(pa->thread, seed++);
 797
 798                         if(*rd) {
 799                                 if(shade_samples(&ssamp, (PixStr *)(*rd), x, y)) {
 800
 801                                         /* multisample buffers or filtered mask filling? */
 802                                         if(pa->fullresult.first) {
 803                                                 int a;
 804                                                 for(samp=0; samp<ssamp.tot; samp++) {
 805                                                         int smask= ssamp.shi[samp].mask;
 806                                                         for(a=0; a<R.osa; a++) {
 807                                                                 int mask= 1<<a;
 808                                                                 if(smask & mask)
 809                                                                         add_passes(ssamp.rlpp[a], od, &ssamp.shi[samp], &ssamp.shr[samp]);
 810                                                         }
 811                                                 }
 812                                         }
 813                                         else {
 814                                                 for(samp=0; samp<ssamp.tot; samp++)
 815                                                         add_filt_passes(rl, ssamp.shi[samp].mask, pa->rectx, od, &ssamp.shi[samp], &ssamp.shr[samp]);
 816                                         }
 817                                 }
 818                         }
 819                 }
 820
 821                 rectdaps+= pa->rectx;
 822                 offs+= pa->rectx;
 823
 824                 if(y&1) if(R.test_break(R.tbh)) break;
 825         }
 826
 827         /* disable scanline updating */
 828         rr->renlay= NULL;
 829
 830         if(R.r.mode & R_SHADOW)
 831                 ISB_free(pa);
 832
 833         if(R.occlusiontree)
 834                 free_occ_samples(&R, pa);
 835 }
 836
 837 /* ************* pixel struct ******** */
 838
 839
 840 static PixStrMain *addpsmain(ListBase *lb)
 841 {
 842         PixStrMain *psm;
 843
 844         psm= (PixStrMain *)MEM_mallocN(sizeof(PixStrMain),"pixstrMain");
 845         BLI_addtail(lb, psm);
 846
 847         psm->ps= (PixStr *)MEM_mallocN(4096*sizeof(PixStr),"pixstr");
 848         psm->counter= 0;
 849
 850         return psm;
 851 }
 852
 853 static void freeps(ListBase *lb)
 854 {
 855         PixStrMain *psm, *psmnext;
 856
 857         for(psm= lb->first; psm; psm= psmnext) {
 858                 psmnext= psm->next;
 859                 if(psm->ps)
 860                         MEM_freeN(psm->ps);
 861                 MEM_freeN(psm);
 862         }
 863         lb->first= lb->last= NULL;
 864 }
 865
 866 static void addps(ListBase *lb, intptr_t *rd, int obi, int facenr, int z, int maskz, unsigned short mask)
 867 {
 868         PixStrMain *psm;
 869         PixStr *ps, *last= NULL;
 870
 871         if(*rd) {
 872                 ps= (PixStr *)(*rd);
 873
 874                 while(ps) {
 875                         if( ps->obi == obi && ps->facenr == facenr ) {
 876                                 ps->mask |= mask;
 877                                 return;
 878                         }
 879                         last= ps;
 880                         ps= ps->next;
 881                 }
 882         }
 883
 884         /* make new PS (pixel struct) */
 885         psm= lb->last;
 886
 887         if(psm->counter==4095)
 888                 psm= addpsmain(lb);
 889
 890         ps= psm->ps + psm->counter++;
 891
 892         if(last) last->next= ps;
 893         else *rd= (intptr_t)ps;
 894
 895         ps->next= NULL;
 896         ps->obi= obi;
 897         ps->facenr= facenr;
 898         ps->z= z;
 899         ps->maskz= maskz;
 900         ps->mask = mask;
 901         ps->shadfac= 0;
 902 }
 903
 904 static void edge_enhance_add(RenderPart *pa, float *rectf, float *arect)
 905 {
 906         float addcol[4];
 907         int pix;
 908
 909         if(arect==NULL)
 910                 return;
 911
 912         for(pix= pa->rectx*pa->recty; pix>0; pix--, arect++, rectf+=4) {
 913                 if(*arect != 0.0f) {
 914                         addcol[0]= *arect * R.r.edgeR;
 915                         addcol[1]= *arect * R.r.edgeG;
 916                         addcol[2]= *arect * R.r.edgeB;
 917                         addcol[3]= *arect;
 918                         addAlphaOverFloat(rectf, addcol);
 919                 }
 920         }
 921 }
 922
 923 static void convert_to_key_alpha(RenderPart *pa, RenderLayer *rl)
 924 {
 925         RenderLayer *rlpp[RE_MAX_OSA];
 926         int y, sample, totsample;
 927
 928         totsample= get_sample_layers(pa, rl, rlpp);
 929
 930         for(sample= 0; sample<totsample; sample++) {
 931                 float *rectf= rlpp[sample]->rectf;
 932
 933                 for(y= pa->rectx*pa->recty; y>0; y--, rectf+=4) {
 934                         if(rectf[3] >= 1.0f);
 935                         else if(rectf[3] > 0.0f) {
 936                                 rectf[0] /= rectf[3];
 937                                 rectf[1] /= rectf[3];
 938                                 rectf[2] /= rectf[3];
 939                         }
 940                 }
 941         }
 942 }
 943
 944 /* adds only alpha values */
 945 void edge_enhance_tile(RenderPart *pa, float *rectf, int *rectz)
 946 {
 947         /* use zbuffer to define edges, add it to the image */
 948         int y, x, col, *rz, *rz1, *rz2, *rz3;
 949         int zval1, zval2, zval3;
 950         float *rf;
 951
 952         /* shift values in zbuffer 4 to the right (anti overflows), for filter we need multiplying with 12 max */
 953         rz= rectz;
 954         if(rz==NULL) return;
 955
 956         for(y=0; y<pa->recty; y++)
 957                 for(x=0; x<pa->rectx; x++, rz++) (*rz)>>= 4;
 958
 959         rz1= rectz;
 960         rz2= rz1+pa->rectx;
 961         rz3= rz2+pa->rectx;
 962
 963         rf= rectf+pa->rectx+1;
 964
 965         for(y=0; y<pa->recty-2; y++) {
 966                 for(x=0; x<pa->rectx-2; x++, rz1++, rz2++, rz3++, rf++) {
 967
 968                         /* prevent overflow with sky z values */
 969                         zval1=   rz1[0] + 2*rz1[1] +   rz1[2];
 970                         zval2=  2*rz2[0]           + 2*rz2[2];
 971                         zval3=   rz3[0] + 2*rz3[1] +   rz3[2];
 972
 973                         col= ( 4*rz2[1] - (zval1 + zval2 + zval3)/3 );
 974                         if(col<0) col= -col;
 975
 976                         col >>= 5;
 977                         if(col > (1<<16)) col= (1<<16);
 978                         else col= (R.r.edgeint*col)>>8;
 979
 980                         if(col>0) {
 981                                 float fcol;
 982
 983                                 if(col>255) fcol= 1.0f;
 984                                 else fcol= (float)col/255.0f;
 985
 986                                 if(R.osa)
 987                                         *rf+= fcol/(float)R.osa;
 988                                 else
 989                                         *rf= fcol;
 990                         }
 991                 }
 992                 rz1+= 2;
 993                 rz2+= 2;
 994                 rz3+= 2;
 995                 rf+= 2;
 996         }
 997
 998         /* shift back zbuf values, we might need it still */
 999         rz= rectz;
1000         for(y=0; y<pa->recty; y++)
1001                 for(x=0; x<pa->rectx; x++, rz++) (*rz)<<= 4;
1002
1003 }
1004
1005 static void reset_sky_speed(RenderPart *pa, RenderLayer *rl)
1006 {
1007         /* for all pixels with max speed, set to zero */
1008     RenderLayer *rlpp[RE_MAX_OSA];
1009         float *fp;
1010         int a, sample, totsample;
1011
1012         totsample= get_sample_layers(pa, rl, rlpp);
1013
1014         for(sample= 0; sample<totsample; sample++) {
1015                 fp= RE_RenderLayerGetPass(rlpp[sample], SCE_PASS_VECTOR);
1016                 if(fp==NULL) break;
1017
1018                 for(a= 4*pa->rectx*pa->recty - 1; a>=0; a--)
1019                         if(fp[a] == PASS_VECTOR_MAX) fp[a]= 0.0f;
1020         }
1021 }
1022
1023 static unsigned short *make_solid_mask(RenderPart *pa)
1024 {
1025         intptr_t *rd= pa->rectdaps;
1026         unsigned short *solidmask, *sp;
1027         int x;
1028
1029         if(rd==NULL) return NULL;
1030
1031         sp=solidmask= MEM_mallocN(sizeof(short)*pa->rectx*pa->recty, "solidmask");
1032
1033         for(x=pa->rectx*pa->recty; x>0; x--, rd++, sp++) {
1034                 if(*rd) {
1035                         PixStr *ps= (PixStr *)*rd;
1036
1037                         *sp= ps->mask;
1038                         for(ps= ps->next; ps; ps= ps->next)
1039                                 *sp |= ps->mask;
1040                 }
1041                 else
1042                         *sp= 0;
1043         }
1044
1045         return solidmask;
1046 }
1047
1048 static void addAlphaOverFloatMask(float *dest, float *source, unsigned short dmask, unsigned short smask)
1049 {
1050         unsigned short shared= dmask & smask;
1051         float mul= 1.0 - source[3];
1052
1053         if(shared) {    /* overlapping masks */
1054
1055                 /* masks differ, we make a mixture of 'add' and 'over' */
1056                 if(shared!=dmask) {
1057                         float shared_bits= (float)count_mask(shared);           /* alpha over */
1058                         float tot_bits= (float)count_mask(smask|dmask);         /* alpha add */
1059
1060                         float add= (tot_bits - shared_bits)/tot_bits;           /* add level */
1061                         mul= add + (1.0f-add)*mul;
1062                 }
1063         }
1064         else if(dmask && smask) {
1065                 /* works for premul only, of course */
1066                 dest[0]+= source[0];
1067                 dest[1]+= source[1];
1068                 dest[2]+= source[2];
1069                 dest[3]+= source[3];
1070
1071                 return;
1072         }
1073
1074         dest[0]= (mul*dest[0]) + source[0];
1075         dest[1]= (mul*dest[1]) + source[1];
1076         dest[2]= (mul*dest[2]) + source[2];
1077         dest[3]= (mul*dest[3]) + source[3];
1078 }
1079
1080 typedef struct ZbufSolidData {
1081         RenderLayer *rl;
1082         ListBase *psmlist;
1083         float *edgerect;
1084 } ZbufSolidData;
1085
1086 void make_pixelstructs(RenderPart *pa, ZSpan *zspan, int sample, void *data)
1087 {
1088         ZbufSolidData *sdata= (ZbufSolidData*)data;
1089         ListBase *lb= sdata->psmlist;
1090         intptr_t *rd= pa->rectdaps;
1091         int *ro= zspan->recto;
1092         int *rp= zspan->rectp;
1093         int *rz= zspan->rectz;
1094         int *rm= zspan->rectmask;
1095         int x, y;
1096         int mask= 1<<sample;
1097
1098         for(y=0; y<pa->recty; y++) {
1099                 for(x=0; x<pa->rectx; x++, rd++, rp++, ro++, rz++, rm++) {
1100                         if(*rp) {
1101                                 addps(lb, rd, *ro, *rp, *rz, (zspan->rectmask)? *rm: 0, mask);
1102                         }
1103                 }
1104         }
1105
1106         if(sdata->rl->layflag & SCE_LAY_EDGE)
1107                 if(R.r.mode & R_EDGE)
1108                         edge_enhance_tile(pa, sdata->edgerect, zspan->rectz);
1109 }
1110
1111 /* main call for shading Delta Accum, for OSA */
1112 /* supposed to be fully threadable! */
1113 void zbufshadeDA_tile(RenderPart *pa)
1114 {
1115         RenderResult *rr= pa->result;
1116         RenderLayer *rl;
1117         ListBase psmlist= {NULL, NULL};
1118         float *edgerect= NULL;
1119
1120         /* allocate the necessary buffers */
1121                                 /* zbuffer inits these rects */
1122         pa->recto= MEM_mallocN(sizeof(int)*pa->rectx*pa->recty, "recto");
1123         pa->rectp= MEM_mallocN(sizeof(int)*pa->rectx*pa->recty, "rectp");
1124         pa->rectz= MEM_mallocN(sizeof(int)*pa->rectx*pa->recty, "rectz");
1125         for(rl= rr->layers.first; rl; rl= rl->next) {
1126                 if((rl->layflag & SCE_LAY_ZMASK) && (rl->layflag & SCE_LAY_NEG_ZMASK))
1127                         pa->rectmask= MEM_mallocN(sizeof(int)*pa->rectx*pa->recty, "rectmask");
1128
1129                 /* initialize pixelstructs and edge buffer */
1130                 addpsmain(&psmlist);
1131                 pa->rectdaps= MEM_callocN(sizeof(intptr_t)*pa->rectx*pa->recty+4, "zbufDArectd");
1132
1133                 if(rl->layflag & SCE_LAY_EDGE)
1134                         if(R.r.mode & R_EDGE)
1135                                 edgerect= MEM_callocN(sizeof(float)*pa->rectx*pa->recty, "rectedge");
1136
1137                 /* always fill visibility */
1138                 for(pa->sample=0; pa->sample<R.osa; pa->sample+=4) {
1139                         ZbufSolidData sdata;
1140
1141                         sdata.rl= rl;
1142                         sdata.psmlist= &psmlist;
1143                         sdata.edgerect= edgerect;
1144                         zbuffer_solid(pa, rl, make_pixelstructs, &sdata);
1145                         if(R.test_break(R.tbh)) break;
1146                 }
1147
1148                 /* shades solid */
1149                 if(rl->layflag & SCE_LAY_SOLID)
1150                         shadeDA_tile(pa, rl);
1151
1152                 /* lamphalo after solid, before ztra, looks nicest because ztra does own halo */
1153                 if(R.flag & R_LAMPHALO)
1154                         if(rl->layflag & SCE_LAY_HALO)
1155                                 lamphalo_tile(pa, rl);
1156
1157                 /* halo before ztra, because ztra fills in zbuffer now */
1158                 if(R.flag & R_HALO)
1159                         if(rl->layflag & SCE_LAY_HALO)
1160                                 halo_tile(pa, rl);
1161
1162                 /* transp layer */
1163                 if(R.flag & R_ZTRA || R.totstrand) {
1164                         if(rl->layflag & (SCE_LAY_ZTRA|SCE_LAY_STRAND)) {
1165                                 if(pa->fullresult.first) {
1166                                         zbuffer_transp_shade(pa, rl, rl->rectf, &psmlist);
1167                                 }
1168                                 else {
1169                                         unsigned short *ztramask, *solidmask= NULL; /* 16 bits, MAX_OSA */
1170
1171                                         /* allocate, but not free here, for asynchronous display of this rect in main thread */
1172                                         rl->acolrect= MEM_callocN(4*sizeof(float)*pa->rectx*pa->recty, "alpha layer");
1173
1174                                         /* swap for live updates, and it is used in zbuf.c!!! */
1175                                         SWAP(float *, rl->acolrect, rl->rectf);
1176                                         ztramask= zbuffer_transp_shade(pa, rl, rl->rectf, &psmlist);
1177                                         SWAP(float *, rl->acolrect, rl->rectf);
1178
1179                                         /* zbuffer transp only returns ztramask if there's solid rendered */
1180                                         if(ztramask)
1181                                                 solidmask= make_solid_mask(pa);
1182
1183                                         if(ztramask && solidmask) {
1184                                                 unsigned short *sps= solidmask, *spz= ztramask;
1185                                                 unsigned short fullmask= (1<<R.osa)-1;
1186                                                 float *fcol= rl->rectf; float *acol= rl->acolrect;
1187                                                 int x;
1188
1189                                                 for(x=pa->rectx*pa->recty; x>0; x--, acol+=4, fcol+=4, sps++, spz++) {
1190                                                         if(*sps == fullmask)
1191                                                                 addAlphaOverFloat(fcol, acol);
1192                                                         else
1193                                                                 addAlphaOverFloatMask(fcol, acol, *sps, *spz);
1194                                                 }
1195                                         }
1196                                         else {
1197                                                 float *fcol= rl->rectf; float *acol= rl->acolrect;
1198                                                 int x;
1199                                                 for(x=pa->rectx*pa->recty; x>0; x--, acol+=4, fcol+=4) {
1200                                                         addAlphaOverFloat(fcol, acol);
1201                                                 }
1202                                         }
1203                                         if(solidmask) MEM_freeN(solidmask);
1204                                         if(ztramask) MEM_freeN(ztramask);
1205                                 }
1206                         }
1207                 }
1208
1209                 /* sun/sky */
1210                 if(rl->layflag & SCE_LAY_SKY)
1211                         atm_tile(pa, rl);
1212
1213                 /* sky before edge */
1214                 if(rl->layflag & SCE_LAY_SKY)
1215                         sky_tile(pa, rl);
1216
1217                 /* extra layers */
1218                 if(rl->layflag & SCE_LAY_EDGE)
1219                         if(R.r.mode & R_EDGE)
1220                                 edge_enhance_add(pa, rl->rectf, edgerect);
1221
1222                 if(rl->passflag & SCE_PASS_VECTOR)
1223                         reset_sky_speed(pa, rl);
1224
1225                 /* de-premul alpha */
1226                 if(R.r.alphamode & R_ALPHAKEY)
1227                         convert_to_key_alpha(pa, rl);
1228
1229                 /* free stuff within loop! */
1230                 MEM_freeN(pa->rectdaps); pa->rectdaps= NULL;
1231                 freeps(&psmlist);
1232
1233                 if(edgerect) MEM_freeN(edgerect);
1234                 edgerect= NULL;
1235
1236                 if(pa->rectmask) {
1237                         MEM_freeN(pa->rectmask);
1238                         pa->rectmask= NULL;
1239                 }
1240         }
1241
1242         /* free all */
1243         MEM_freeN(pa->recto); pa->recto= NULL;
1244         MEM_freeN(pa->rectp); pa->rectp= NULL;
1245         MEM_freeN(pa->rectz); pa->rectz= NULL;
1246
1247         /* display active layer */
1248         rr->renrect.ymin=rr->renrect.ymax= 0;
1249         rr->renlay= render_get_active_layer(&R, rr);
1250 }
1251
1252
1253 /* ------------------------------------------------------------------------ */
1254
1255 /* non OSA case, full tile render */
1256 /* supposed to be fully threadable! */
1257 void zbufshade_tile(RenderPart *pa)
1258 {
1259         ShadeSample ssamp;
1260         RenderResult *rr= pa->result;
1261         RenderLayer *rl;
1262         PixStr ps;
1263         float *edgerect= NULL;
1264
1265         /* fake pixel struct, to comply to osa render */
1266         ps.next= NULL;
1267         ps.mask= 0xFFFF;
1268
1269         /* zbuffer code clears/inits rects */
1270         pa->recto= MEM_mallocN(sizeof(int)*pa->rectx*pa->recty, "recto");
1271         pa->rectp= MEM_mallocN(sizeof(int)*pa->rectx*pa->recty, "rectp");
1272         pa->rectz= MEM_mallocN(sizeof(int)*pa->rectx*pa->recty, "rectz");
1273
1274         for(rl= rr->layers.first; rl; rl= rl->next) {
1275                 if((rl->layflag & SCE_LAY_ZMASK) && (rl->layflag & SCE_LAY_NEG_ZMASK))
1276                         pa->rectmask= MEM_mallocN(sizeof(int)*pa->rectx*pa->recty, "rectmask");
1277
1278                 /* general shader info, passes */
1279                 shade_sample_initialize(&ssamp, pa, rl);
1280
1281                 zbuffer_solid(pa, rl, NULL, NULL);
1282
1283                 if(!R.test_break(R.tbh)) {      /* NOTE: this if() is not consistant */
1284
1285                         /* edges only for solid part, ztransp doesn't support it yet anti-aliased */
1286                         if(rl->layflag & SCE_LAY_EDGE) {
1287                                 if(R.r.mode & R_EDGE) {
1288                                         edgerect= MEM_callocN(sizeof(float)*pa->rectx*pa->recty, "rectedge");
1289                                         edge_enhance_tile(pa, edgerect, pa->rectz);
1290                                 }
1291                         }
1292
1293                         /* initialize scanline updates for main thread */
1294                         rr->renrect.ymin= 0;
1295                         rr->renlay= rl;
1296
1297                         if(rl->layflag & SCE_LAY_SOLID) {
1298                                 float *fcol= rl->rectf;
1299                                 int *ro= pa->recto, *rp= pa->rectp, *rz= pa->rectz;
1300                                 int x, y, offs=0, seed;
1301
1302                                 /* we set per pixel a fixed seed, for random AO and shadow samples */
1303                                 seed= pa->rectx*pa->disprect.ymin;
1304
1305                                 /* irregular shadowb buffer creation */
1306                                 if(R.r.mode & R_SHADOW)
1307                                         ISB_create(pa, NULL);
1308
1309                                 if(R.occlusiontree)
1310                                         cache_occ_samples(&R, pa, &ssamp);
1311
1312                                 for(y=pa->disprect.ymin; y<pa->disprect.ymax; y++, rr->renrect.ymax++) {
1313                                         for(x=pa->disprect.xmin; x<pa->disprect.xmax; x++, ro++, rz++, rp++, fcol+=4, offs++) {
1314                                                 /* per pixel fixed seed */
1315                                                 BLI_thread_srandom(pa->thread, seed++);
1316
1317                                                 if(*rp) {
1318                                                         ps.obi= *ro;
1319                                                         ps.facenr= *rp;
1320                                                         ps.z= *rz;
1321                                                         if(shade_samples(&ssamp, &ps, x, y)) {
1322                                                                 /* combined and passes */
1323                                                                 add_passes(rl, offs, ssamp.shi, ssamp.shr);
1324                                                         }
1325                                                 }
1326                                         }
1327                                         if(y&1)
1328                                                 if(R.test_break(R.tbh)) break;
1329                                 }
1330
1331                                 if(R.occlusiontree)
1332                                         free_occ_samples(&R, pa);
1333
1334                                 if(R.r.mode & R_SHADOW)
1335                                         ISB_free(pa);
1336                         }
1337
1338                         /* disable scanline updating */
1339                         rr->renlay= NULL;
1340                 }
1341
1342                 /* lamphalo after solid, before ztra, looks nicest because ztra does own halo */
1343                 if(R.flag & R_LAMPHALO)
1344                         if(rl->layflag & SCE_LAY_HALO)
1345                                 lamphalo_tile(pa, rl);
1346
1347                 /* halo before ztra, because ztra fills in zbuffer now */
1348                 if(R.flag & R_HALO)
1349                         if(rl->layflag & SCE_LAY_HALO)
1350                                 halo_tile(pa, rl);
1351
1352                 if(R.flag & R_ZTRA || R.totstrand) {
1353                         if(rl->layflag & (SCE_LAY_ZTRA|SCE_LAY_STRAND)) {
1354                                 float *fcol, *acol;
1355                                 int x;
1356
1357                                 /* allocate, but not free here, for asynchronous display of this rect in main thread */
1358                                 rl->acolrect= MEM_callocN(4*sizeof(float)*pa->rectx*pa->recty, "alpha layer");
1359
1360                                 /* swap for live updates */
1361                                 SWAP(float *, rl->acolrect, rl->rectf);
1362                                 zbuffer_transp_shade(pa, rl, rl->rectf, NULL);
1363                                 SWAP(float *, rl->acolrect, rl->rectf);
1364
1365                                 fcol= rl->rectf; acol= rl->acolrect;
1366                                 for(x=pa->rectx*pa->recty; x>0; x--, acol+=4, fcol+=4) {
1367                                         addAlphaOverFloat(fcol, acol);
1368                                 }
1369                         }
1370                 }
1371
1372                 /* sun/sky */
1373                 if(rl->layflag & SCE_LAY_SKY)
1374                         atm_tile(pa, rl);
1375
1376                 /* sky before edge */
1377                 if(rl->layflag & SCE_LAY_SKY)
1378                         sky_tile(pa, rl);
1379
1380                 if(!R.test_break(R.tbh)) {
1381                         if(rl->layflag & SCE_LAY_EDGE)
1382                                 if(R.r.mode & R_EDGE)
1383                                         edge_enhance_add(pa, rl->rectf, edgerect);
1384                 }
1385
1386                 if(rl->passflag & SCE_PASS_VECTOR)
1387                         reset_sky_speed(pa, rl);
1388
1389                 /* de-premul alpha */
1390                 if(R.r.alphamode & R_ALPHAKEY)
1391                         convert_to_key_alpha(pa, rl);
1392
1393                 if(edgerect) MEM_freeN(edgerect);
1394                 edgerect= NULL;
1395
1396                 if(pa->rectmask) {
1397                         MEM_freeN(pa->rectmask);
1398                         pa->rectmask= NULL;
1399                 }
1400         }
1401
1402         /* display active layer */
1403         rr->renrect.ymin=rr->renrect.ymax= 0;
1404         rr->renlay= render_get_active_layer(&R, rr);
1405
1406         MEM_freeN(pa->recto); pa->recto= NULL;
1407         MEM_freeN(pa->rectp); pa->rectp= NULL;
1408         MEM_freeN(pa->rectz); pa->rectz= NULL;
1409 }
1410
1411 /* SSS preprocess tile render, fully threadable */
1412 typedef struct ZBufSSSHandle {
1413         RenderPart *pa;
1414         ListBase psmlist;
1415         int totps;
1416 } ZBufSSSHandle;
1417
1418 static void addps_sss(void *cb_handle, int obi, int facenr, int x, int y, int z)
1419 {
1420         ZBufSSSHandle *handle = cb_handle;
1421         RenderPart *pa= handle->pa;
1422
1423         /* extra border for filter gives double samples on part edges,
1424            don't use those */
1425         if(x<pa->crop || x>=pa->rectx-pa->crop)
1426                 return;
1427         if(y<pa->crop || y>=pa->recty-pa->crop)
1428                 return;
1429
1430         if(pa->rectall) {
1431                 intptr_t *rs= pa->rectall + pa->rectx*y + x;
1432
1433                 addps(&handle->psmlist, rs, obi, facenr, z, 0, 0);
1434                 handle->totps++;
1435         }
1436         if(pa->rectz) {
1437                 int *rz= pa->rectz + pa->rectx*y + x;
1438                 int *rp= pa->rectp + pa->rectx*y + x;
1439                 int *ro= pa->recto + pa->rectx*y + x;
1440
1441                 if(z < *rz) {
1442                         if(*rp == 0)
1443                                 handle->totps++;
1444                         *rz= z;
1445                         *rp= facenr;
1446                         *ro= obi;
1447                 }
1448         }
1449         if(pa->rectbackz) {
1450                 int *rz= pa->rectbackz + pa->rectx*y + x;
1451                 int *rp= pa->rectbackp + pa->rectx*y + x;
1452                 int *ro= pa->rectbacko + pa->rectx*y + x;
1453
1454                 if(z >= *rz) {
1455                         if(*rp == 0)
1456                                 handle->totps++;
1457                         *rz= z;
1458                         *rp= facenr;
1459                         *ro= obi;
1460                 }
1461         }
1462 }
1463
1464 static void shade_sample_sss(ShadeSample *ssamp, Material *mat, ObjectInstanceRen *obi, VlakRen *vlr, int quad, float x, float y, float z, float *co, float *color, float *area)
1465 {
1466         ShadeInput *shi= ssamp->shi;
1467         ShadeResult shr;
1468         float texfac, orthoarea, nor[3], alpha, sx, sy;
1469
1470         /* cache for shadow */
1471         shi->samplenr= R.shadowsamplenr[shi->thread]++;
1472
1473         if(quad)
1474                 shade_input_set_triangle_i(shi, obi, vlr, 0, 2, 3);
1475         else
1476                 shade_input_set_triangle_i(shi, obi, vlr, 0, 1, 2);
1477
1478         /* center pixel */
1479         sx = x + 0.5f;
1480         sy = y + 0.5f;
1481
1482         /* we estimate the area here using shi->dxco and shi->dyco. we need to
1483            enabled shi->osatex these are filled. we compute two areas, one with
1484            the normal pointed at the camera and one with the original normal, and
1485            then clamp to avoid a too large contribution from a single pixel */
1486         shi->osatex= 1;
1487
1488         VECCOPY(nor, shi->facenor);
1489         calc_view_vector(shi->facenor, sx, sy);
1490         Normalize(shi->facenor);
1491         shade_input_set_viewco(shi, x, y, sx, sy, z);
1492         orthoarea= VecLength(shi->dxco)*VecLength(shi->dyco);
1493
1494         VECCOPY(shi->facenor, nor);
1495         shade_input_set_viewco(shi, x, y, sx, sy, z);
1496         *area= VecLength(shi->dxco)*VecLength(shi->dyco);
1497         *area= MIN2(*area, 2.0f*orthoarea);
1498
1499         shade_input_set_uv(shi);
1500         shade_input_set_normals(shi);
1501
1502         /* we don't want flipped normals, they screw up back scattering */
1503         if(shi->flippednor)
1504                 shade_input_flip_normals(shi);
1505
1506         /* not a pretty solution, but fixes common cases */
1507         if(shi->obr->ob && shi->obr->ob->transflag & OB_NEG_SCALE) {
1508                 VecNegf(shi->vn);
1509                 VecNegf(shi->vno);
1510         }
1511
1512         /* if nodetree, use the material that we are currently preprocessing
1513            instead of the node material */
1514         if(shi->mat->nodetree && shi->mat->use_nodes)
1515                 shi->mat= mat;
1516
1517         /* init material vars */
1518         // note, keep this synced with render_types.h
1519         memcpy(&shi->r, &shi->mat->r, 23*sizeof(float));
1520         shi->har= shi->mat->har;
1521
1522         /* render */
1523         shade_input_set_shade_texco(shi);
1524
1525         shade_samples_do_AO(ssamp);
1526         shade_material_loop(shi, &shr);
1527
1528         VECCOPY(co, shi->co);
1529         VECCOPY(color, shr.combined);
1530
1531         /* texture blending */
1532         texfac= shi->mat->sss_texfac;
1533
1534         alpha= shr.combined[3];
1535         *area *= alpha;
1536 }
1537
1538 static void zbufshade_sss_free(RenderPart *pa)
1539 {
1540 #if 0
1541         MEM_freeN(pa->rectall); pa->rectall= NULL;
1542         freeps(&handle.psmlist);
1543 #else
1544         MEM_freeN(pa->rectz); pa->rectz= NULL;
1545         MEM_freeN(pa->rectp); pa->rectp= NULL;
1546         MEM_freeN(pa->recto); pa->recto= NULL;
1547         MEM_freeN(pa->rectbackz); pa->rectbackz= NULL;
1548         MEM_freeN(pa->rectbackp); pa->rectbackp= NULL;
1549         MEM_freeN(pa->rectbacko); pa->rectbacko= NULL;
1550 #endif
1551 }
1552
1553 void zbufshade_sss_tile(RenderPart *pa)
1554 {
1555         Render *re= &R;
1556         ShadeSample ssamp;
1557         ZBufSSSHandle handle;
1558         RenderResult *rr= pa->result;
1559         RenderLayer *rl;
1560         VlakRen *vlr;
1561         Material *mat= re->sss_mat;
1562         float (*co)[3], (*color)[3], *area, *fcol;
1563         int x, y, seed, quad, totpoint, display = !(re->r.scemode & R_PREVIEWBUTS);
1564         int *ro, *rz, *rp, *rbo, *rbz, *rbp, lay;
1565 #if 0
1566         PixStr *ps;
1567         intptr_t *rs;
1568         int z;
1569 #endif
1570
1571         /* setup pixelstr list and buffer for zbuffering */
1572         handle.pa= pa;
1573         handle.totps= 0;
1574
1575 #if 0
1576         handle.psmlist.first= handle.psmlist.last= NULL;
1577         addpsmain(&handle.psmlist);
1578
1579         pa->rectall= MEM_callocN(sizeof(intptr_t)*pa->rectx*pa->recty+4, "rectall");
1580 #else
1581         pa->recto= MEM_mallocN(sizeof(int)*pa->rectx*pa->recty, "recto");
1582         pa->rectp= MEM_mallocN(sizeof(int)*pa->rectx*pa->recty, "rectp");
1583         pa->rectz= MEM_mallocN(sizeof(int)*pa->rectx*pa->recty, "rectz");
1584         pa->rectbacko= MEM_mallocN(sizeof(int)*pa->rectx*pa->recty, "rectbacko");
1585         pa->rectbackp= MEM_mallocN(sizeof(int)*pa->rectx*pa->recty, "rectbackp");
1586         pa->rectbackz= MEM_mallocN(sizeof(int)*pa->rectx*pa->recty, "rectbackz");
1587 #endif
1588
1589         /* setup shade sample with correct passes */
1590         memset(&ssamp, 0, sizeof(ssamp));
1591         shade_sample_initialize(&ssamp, pa, rr->layers.first);
1592         ssamp.tot= 1;
1593
1594         for(rl=rr->layers.first; rl; rl=rl->next) {
1595                 ssamp.shi[0].lay |= rl->lay;
1596                 ssamp.shi[0].layflag |= rl->layflag;
1597                 ssamp.shi[0].passflag |= rl->passflag;
1598                 ssamp.shi[0].combinedflag |= ~rl->pass_xor;
1599         }
1600
1601         rl= rr->layers.first;
1602         ssamp.shi[0].passflag |= SCE_PASS_RGBA|SCE_PASS_COMBINED;
1603         ssamp.shi[0].combinedflag &= ~(SCE_PASS_SPEC);
1604         ssamp.shi[0].mat_override= NULL;
1605         ssamp.shi[0].light_override= NULL;
1606         lay= ssamp.shi[0].lay;
1607
1608         /* create the pixelstrs to be used later */
1609         zbuffer_sss(pa, lay, &handle, addps_sss);
1610
1611         if(handle.totps==0) {
1612                 zbufshade_sss_free(pa);
1613                 return;
1614         }
1615
1616         fcol= rl->rectf;
1617
1618         co= MEM_mallocN(sizeof(float)*3*handle.totps, "SSSCo");
1619         color= MEM_mallocN(sizeof(float)*3*handle.totps, "SSSColor");
1620         area= MEM_mallocN(sizeof(float)*handle.totps, "SSSArea");
1621
1622 #if 0
1623         /* create ISB (does not work currently!) */
1624         if(re->r.mode & R_SHADOW)
1625                 ISB_create(pa, NULL);
1626 #endif
1627
1628         if(display) {
1629                 /* initialize scanline updates for main thread */
1630                 rr->renrect.ymin= 0;
1631                 rr->renlay= rl;
1632         }
1633
1634         seed= pa->rectx*pa->disprect.ymin;
1635 #if 0
1636         rs= pa->rectall;
1637 #else
1638         rz= pa->rectz;
1639         rp= pa->rectp;
1640         ro= pa->recto;
1641         rbz= pa->rectbackz;
1642         rbp= pa->rectbackp;
1643         rbo= pa->rectbacko;
1644 #endif
1645         totpoint= 0;
1646
1647         for(y=pa->disprect.ymin; y<pa->disprect.ymax; y++, rr->renrect.ymax++) {
1648                 for(x=pa->disprect.xmin; x<pa->disprect.xmax; x++, fcol+=4) {
1649                         /* per pixel fixed seed */
1650                         BLI_thread_srandom(pa->thread, seed++);
1651
1652 #if 0
1653                         if(rs) {
1654                                 /* for each sample in this pixel, shade it */
1655                                 for(ps=(PixStr*)*rs; ps; ps=ps->next) {
1656                                         ObjectInstanceRen *obi= &re->objectinstance[ps->obi];
1657                                         ObjectRen *obr= obi->obr;
1658                                         vlr= RE_findOrAddVlak(obr, (ps->facenr-1) & RE_QUAD_MASK);
1659                                         quad= (ps->facenr & RE_QUAD_OFFS);
1660                                         z= ps->z;
1661
1662                                         shade_sample_sss(&ssamp, mat, obi, vlr, quad, x, y, z,
1663                                                 co[totpoint], color[totpoint], &area[totpoint]);
1664
1665                                         totpoint++;
1666
1667                                         VECADD(fcol, fcol, color);
1668                                         fcol[3]= 1.0f;
1669                                 }
1670
1671                                 rs++;
1672                         }
1673 #else
1674                         if(rp) {
1675                                 if(*rp != 0) {
1676                                         ObjectInstanceRen *obi= &re->objectinstance[*ro];
1677                                         ObjectRen *obr= obi->obr;
1678
1679                                         /* shade front */
1680                                         vlr= RE_findOrAddVlak(obr, (*rp-1) & RE_QUAD_MASK);
1681                                         quad= ((*rp) & RE_QUAD_OFFS);
1682
1683                                         shade_sample_sss(&ssamp, mat, obi, vlr, quad, x, y, *rz,
1684                                                 co[totpoint], color[totpoint], &area[totpoint]);
1685
1686                                         VECADD(fcol, fcol, color[totpoint]);
1687                                         fcol[3]= 1.0f;
1688                                         totpoint++;
1689                                 }
1690
1691                                 rp++; rz++; ro++;
1692                         }
1693
1694                         if(rbp) {
1695                                 if(*rbp != 0 && !(*rbp == *(rp-1) && *rbo == *(ro-1))) {
1696                                         ObjectInstanceRen *obi= &re->objectinstance[*rbo];
1697                                         ObjectRen *obr= obi->obr;
1698
1699                                         /* shade back */
1700                                         vlr= RE_findOrAddVlak(obr, (*rbp-1) & RE_QUAD_MASK);
1701                                         quad= ((*rbp) & RE_QUAD_OFFS);
1702
1703                                         shade_sample_sss(&ssamp, mat, obi, vlr, quad, x, y, *rbz,
1704                                                 co[totpoint], color[totpoint], &area[totpoint]);
1705
1706                                         /* to indicate this is a back sample */
1707                                         area[totpoint]= -area[totpoint];
1708
1709                                         VECADD(fcol, fcol, color[totpoint]);
1710                                         fcol[3]= 1.0f;
1711                                         totpoint++;
1712                                 }
1713
1714                                 rbz++; rbp++; rbo++;
1715                         }
1716 #endif
1717                 }
1718
1719                 if(y&1)
1720                         if(re->test_break(re->tbh)) break;
1721         }
1722
1723         /* note: after adding we do not free these arrays, sss keeps them */
1724         if(totpoint > 0) {
1725                 sss_add_points(re, co, color, area, totpoint);
1726         }
1727         else {
1728                 MEM_freeN(co);
1729                 MEM_freeN(color);
1730                 MEM_freeN(area);
1731         }
1732
1733 #if 0
1734         if(re->r.mode & R_SHADOW)
1735                 ISB_free(pa);
1736 #endif
1737
1738         if(display) {
1739                 /* display active layer */
1740                 rr->renrect.ymin=rr->renrect.ymax= 0;
1741                 rr->renlay= render_get_active_layer(&R, rr);
1742         }
1743
1744         zbufshade_sss_free(pa);
1745 }
1746
1747 /* ------------------------------------------------------------------------ */
1748
1749 static void renderhalo_post(RenderResult *rr, float *rectf, HaloRen *har)       /* postprocess version */
1750 {
1751         float dist, xsq, ysq, xn, yn, colf[4], *rectft, *rtf;
1752         float haloxs, haloys;
1753         int minx, maxx, miny, maxy, x, y;
1754
1755         /* calculate the disprect mapped coordinate for halo. note: rectx is disprect corrected */
1756         haloxs= har->xs - R.disprect.xmin;
1757         haloys= har->ys - R.disprect.ymin;
1758
1759         har->miny= miny= haloys - har->rad/R.ycor;
1760         har->maxy= maxy= haloys + har->rad/R.ycor;
1761
1762         if(maxy<0);
1763         else if(rr->recty<miny);
1764         else {
1765                 minx= floor(haloxs-har->rad);
1766                 maxx= ceil(haloxs+har->rad);
1767
1768                 if(maxx<0);
1769                 else if(rr->rectx<minx);
1770                 else {
1771
1772                         if(minx<0) minx= 0;
1773                         if(maxx>=rr->rectx) maxx= rr->rectx-1;
1774                         if(miny<0) miny= 0;
1775                         if(maxy>rr->recty) maxy= rr->recty;
1776
1777                         rectft= rectf+ 4*rr->rectx*miny;
1778
1779                         for(y=miny; y<maxy; y++) {
1780
1781                                 rtf= rectft+4*minx;
1782
1783                                 yn= (y - haloys)*R.ycor;
1784                                 ysq= yn*yn;
1785
1786                                 for(x=minx; x<=maxx; x++) {
1787                                         xn= x - haloxs;
1788                                         xsq= xn*xn;
1789                                         dist= xsq+ysq;
1790                                         if(dist<har->radsq) {
1791
1792                                                 if(shadeHaloFloat(har, colf, 0x7FFFFF, dist, xn, yn, har->flarec))
1793                                                         addalphaAddfacFloat(rtf, colf, har->add);
1794                                         }
1795                                         rtf+=4;
1796                                 }
1797
1798                                 rectft+= 4*rr->rectx;
1799
1800                                 if(R.test_break(R.tbh)) break;
1801                         }
1802                 }
1803         }
1804 }
1805 /* ------------------------------------------------------------------------ */
1806
1807 static void renderflare(RenderResult *rr, float *rectf, HaloRen *har)
1808 {
1809         extern float hashvectf[];
1810         HaloRen fla;
1811         Material *ma;
1812         float *rc, rad, alfa, visifac, vec[3];
1813         int b, type;
1814
1815         fla= *har;
1816         fla.linec= fla.ringc= fla.flarec= 0;
1817
1818         rad= har->rad;
1819         alfa= har->alfa;
1820
1821         visifac= R.ycor*(har->pixels);
1822         /* all radials added / r^3  == 1.0f! */
1823         visifac /= (har->rad*har->rad*har->rad);
1824         visifac*= visifac;
1825
1826         ma= har->mat;
1827
1828         /* first halo: just do */
1829
1830         har->rad= rad*ma->flaresize*visifac;
1831         har->radsq= har->rad*har->rad;
1832         har->zs= fla.zs= 0;
1833
1834         har->alfa= alfa*visifac;
1835
1836         renderhalo_post(rr, rectf, har);
1837
1838         /* next halo's: the flares */
1839         rc= hashvectf + ma->seed2;
1840
1841         for(b=1; b<har->flarec; b++) {
1842
1843                 fla.r= fabs(rc[0]);
1844                 fla.g= fabs(rc[1]);
1845                 fla.b= fabs(rc[2]);
1846                 fla.alfa= ma->flareboost*fabs(alfa*visifac*rc[3]);
1847                 fla.hard= 20.0f + fabs(70*rc[7]);
1848                 fla.tex= 0;
1849
1850                 type= (int)(fabs(3.9*rc[6]));
1851
1852                 fla.rad= ma->subsize*sqrt(fabs(2.0f*har->rad*rc[4]));
1853
1854                 if(type==3) {
1855                         fla.rad*= 3.0f;
1856                         fla.rad+= R.rectx/10;
1857                 }
1858
1859                 fla.radsq= fla.rad*fla.rad;
1860
1861                 vec[0]= 1.4*rc[5]*(har->xs-R.winx/2);
1862                 vec[1]= 1.4*rc[5]*(har->ys-R.winy/2);
1863                 vec[2]= 32.0f*sqrt(vec[0]*vec[0] + vec[1]*vec[1] + 1.0f);
1864
1865                 fla.xs= R.winx/2 + vec[0] + (1.2+rc[8])*R.rectx*vec[0]/vec[2];
1866                 fla.ys= R.winy/2 + vec[1] + (1.2+rc[8])*R.rectx*vec[1]/vec[2];
1867
1868                 if(R.flag & R_SEC_FIELD) {
1869                         if(R.r.mode & R_ODDFIELD) fla.ys += 0.5;
1870                         else fla.ys -= 0.5;
1871                 }
1872                 if(type & 1) fla.type= HA_FLARECIRC;
1873                 else fla.type= 0;
1874                 renderhalo_post(rr, rectf, &fla);
1875
1876                 fla.alfa*= 0.5;
1877                 if(type & 2) fla.type= HA_FLARECIRC;
1878                 else fla.type= 0;
1879                 renderhalo_post(rr, rectf, &fla);
1880
1881                 rc+= 7;
1882         }
1883 }
1884
1885 /* needs recode... integrate this better! */
1886 void add_halo_flare(Render *re)
1887 {
1888         RenderResult *rr= re->result;
1889         RenderLayer *rl;
1890         HaloRen *har;
1891         int a, mode, do_draw=0;
1892
1893         /* for now, we get the first renderlayer in list with halos set */
1894         for(rl= rr->layers.first; rl; rl= rl->next)
1895                 if(rl->layflag & SCE_LAY_HALO)
1896                         break;
1897
1898         if(rl==NULL || rl->rectf==NULL)
1899                 return;
1900
1901         mode= R.r.mode;
1902         R.r.mode &= ~R_PANORAMA;
1903
1904         project_renderdata(&R, projectverto, 0, 0, 0);
1905
1906         for(a=0; a<R.tothalo; a++) {
1907                 har= R.sortedhalos[a];
1908
1909                 if(har->flarec) {
1910                         do_draw= 1;
1911                         renderflare(rr, rl->rectf, har);
1912                 }
1913         }
1914
1915         if(do_draw) {
1916                 /* weak... the display callback wants an active renderlayer pointer... */
1917                 rr->renlay= rl;
1918                 re->display_draw(re->ddh, rr, NULL);
1919         }
1920
1921         R.r.mode= mode;
1922 }
1923
1924 /* ************************* used for shaded view ************************ */
1925
1926 /* if *re, then initialize, otherwise execute */
1927 void RE_shade_external(Render *re, ShadeInput *shi, ShadeResult *shr)
1928 {
1929         static VlakRen vlr;
1930         static ObjectRen obr;
1931         static ObjectInstanceRen obi;
1932
1933         /* init */
1934         if(re) {
1935                 R= *re;
1936
1937                 /* fake render face */
1938                 memset(&vlr, 0, sizeof(VlakRen));
1939                 memset(&obr, 0, sizeof(ObjectRen));
1940                 memset(&obi, 0, sizeof(ObjectInstanceRen));
1941                 obr.lay= -1;
1942                 obi.obr= &obr;
1943
1944                 return;
1945         }
1946         shi->vlr= &vlr;
1947         shi->obr= &obr;
1948         shi->obi= &obi;
1949
1950         if(shi->mat->nodetree && shi->mat->use_nodes)
1951                 ntreeShaderExecTree(shi->mat->nodetree, shi, shr);
1952         else {
1953                 /* copy all relevant material vars, note, keep this synced with render_types.h */
1954                 memcpy(&shi->r, &shi->mat->r, 23*sizeof(float));
1955                 shi->har= shi->mat->har;
1956
1957                 shade_material_loop(shi, shr);
1958         }
1959 }
1960
1961 /* ************************* bake ************************ */
1962
1963
1964 typedef struct BakeShade {
1965         ShadeSample ssamp;
1966         ObjectInstanceRen *obi;
1967         VlakRen *vlr;
1968
1969         ZSpan *zspan;
1970         Image *ima;
1971         ImBuf *ibuf;
1972
1973         int rectx, recty, quad, type, vdone, ready;
1974
1975         float dir[3];
1976         Object *actob;
1977
1978         unsigned int *rect;
1979         float *rect_float;
1980
1981         int usemask;
1982         char *rect_mask; /* bake pixel mask */
1983
1984         float dxco[3], dyco[3];
1985 } BakeShade;
1986
1987 /* bake uses a char mask to know what has been baked */
1988 #define BAKE_MASK_NULL          0
1989 #define BAKE_MASK_MARGIN        1
1990 #define BAKE_MASK_BAKED         2
1991 static void bake_mask_filter_extend( char *mask, int width, int height )
1992 {
1993         char *row1, *row2, *row3;
1994         int rowlen, x, y;
1995         char *temprect;
1996
1997         rowlen= width;
1998
1999         /* make a copy, to prevent flooding */
2000         temprect= MEM_dupallocN(mask);
2001
2002         for(y=1; y<=height; y++) {
2003                 /* setup rows */
2004                 row1= (char *)(temprect + (y-2)*rowlen);
2005                 row2= row1 + rowlen;
2006                 row3= row2 + rowlen;
2007                 if(y==1)
2008                         row1= row2;
2009                 else if(y==height)
2010                         row3= row2;
2011
2012                 for(x=0; x<rowlen; x++) {
2013                         if (mask[((y-1)*rowlen)+x]==0) {
2014                                 if (*row1 || *row2 || *row3 || *(row1+1) || *(row3+1) ) {
2015                                         mask[((y-1)*rowlen)+x] = BAKE_MASK_MARGIN;
2016                                 } else if((x!=rowlen-1) && (*(row1+2) || *(row2+2) || *(row3+2)) ) {
2017                                         mask[((y-1)*rowlen)+x] = BAKE_MASK_MARGIN;
2018                                 }
2019                         }
2020
2021                         if(x!=0) {
2022                                 row1++; row2++; row3++;
2023                         }
2024                 }
2025         }
2026         MEM_freeN(temprect);
2027 }
2028
2029 static void bake_mask_clear( ImBuf *ibuf, char *mask, char val )
2030 {
2031         int x,y;
2032         if (ibuf->rect_float) {
2033                 for(x=0; x<ibuf->x; x++) {
2034                         for(y=0; y<ibuf->y; y++) {
2035                                 if (mask[ibuf->x*y + x] == val) {
2036                                         float *col= ibuf->rect_float + 4*(ibuf->x*y + x);
2037                                         col[0] = col[1] = col[2] = col[3] = 0.0f;
2038                                 }
2039                         }
2040                 }
2041
2042         } else {
2043                 /* char buffer */
2044                 for(x=0; x<ibuf->x; x++) {
2045                         for(y=0; y<ibuf->y; y++) {
2046                                 if (mask[ibuf->x*y + x] == val) {
2047                                         char *col= (char *)(ibuf->rect + ibuf->x*y + x);
2048                                         col[0] = col[1] = col[2] = col[3] = 0;
2049                                 }
2050                         }
2051                 }
2052         }
2053 }
2054
2055 static void bake_set_shade_input(ObjectInstanceRen *obi, VlakRen *vlr, ShadeInput *shi, int quad, int isect, int x, int y, float u, float v)
2056 {
2057         if(isect) {
2058                 /* raytrace intersection with different u,v than scanconvert */
2059                 if(vlr->v4) {
2060                         if(quad)
2061                                 shade_input_set_triangle_i(shi, obi, vlr, 2, 1, 3);
2062                         else
2063                                 shade_input_set_triangle_i(shi, obi, vlr, 0, 1, 3);
2064                 }
2065                 else
2066                         shade_input_set_triangle_i(shi, obi, vlr, 0, 1, 2);
2067         }
2068         else {
2069                 /* regular scanconvert */
2070                 if(quad)
2071                         shade_input_set_triangle_i(shi, obi, vlr, 0, 2, 3);
2072                 else
2073                         shade_input_set_triangle_i(shi, obi, vlr, 0, 1, 2);
2074         }
2075
2076         /* cache for shadow */
2077         shi->samplenr= R.shadowsamplenr[shi->thread]++;
2078
2079         shi->mask= 0xFFFF; /* all samples */
2080
2081         shi->u= -u;
2082         shi->v= -v;
2083         shi->xs= x;
2084         shi->ys= y;
2085
2086         shade_input_set_uv(shi);
2087         shade_input_set_normals(shi);
2088
2089         /* no normal flip */
2090         if(shi->flippednor)
2091                 shade_input_flip_normals(shi);
2092
2093         /* set up view vector to look right at the surface (note that the normal
2094          * is negated in the renderer so it does not need to be done here) */
2095         shi->view[0]= shi->vn[0];
2096         shi->view[1]= shi->vn[1];
2097         shi->view[2]= shi->vn[2];
2098 }
2099
2100 static void bake_shade(void *handle, Object *ob, ShadeInput *shi, int quad, int x, int y, float u, float v, float *tvn, float *ttang)
2101 {
2102         BakeShade *bs= handle;
2103         ShadeSample *ssamp= &bs->ssamp;
2104         ShadeResult shr;
2105         VlakRen *vlr= shi->vlr;
2106
2107         /* init material vars */
2108         memcpy(&shi->r, &shi->mat->r, 23*sizeof(float));        // note, keep this synced with render_types.h
2109         shi->har= shi->mat->har;
2110
2111         if(bs->type==RE_BAKE_AO) {
2112                 ambient_occlusion(shi);
2113
2114                 if(R.r.bake_flag & R_BAKE_NORMALIZE)
2115                         VECCOPY(shr.combined, shi->ao)
2116                 else
2117                         ambient_occlusion_to_diffuse(shi, shr.combined);
2118         }
2119         else {
2120                 if (bs->type==RE_BAKE_SHADOW) /* Why do shadows set the color anyhow?, ignore material color for baking */
2121                         shi->r = shi->g = shi->b = 1.0f;
2122
2123                 shade_input_set_shade_texco(shi);
2124
2125                 if(!ELEM3(bs->type, RE_BAKE_NORMALS, RE_BAKE_TEXTURE, RE_BAKE_SHADOW))
2126                         shade_samples_do_AO(ssamp);
2127
2128                 if(shi->mat->nodetree && shi->mat->use_nodes) {
2129                         ntreeShaderExecTree(shi->mat->nodetree, shi, &shr);
2130                         shi->mat= vlr->mat;             /* shi->mat is being set in nodetree */
2131                 }
2132                 else
2133                         shade_material_loop(shi, &shr);
2134
2135                 if(bs->type==RE_BAKE_NORMALS) {
2136                         float nor[3];
2137
2138                         VECCOPY(nor, shi->vn);
2139
2140                         if(R.r.bake_normal_space == R_BAKE_SPACE_CAMERA);
2141                         else if(R.r.bake_normal_space == R_BAKE_SPACE_TANGENT) {
2142                                 float mat[3][3], imat[3][3];
2143
2144                                 /* bitangent */
2145                                 if(tvn && ttang) {
2146                                         VECCOPY(mat[0], ttang);
2147                                         Crossf(mat[1], tvn, ttang);
2148                                         VECCOPY(mat[2], tvn);
2149                                 }
2150                                 else {
2151                                         VECCOPY(mat[0], shi->nmaptang);
2152                                         Crossf(mat[1], shi->vn, shi->nmaptang);
2153                                         VECCOPY(mat[2], shi->vn);
2154                                 }
2155
2156                                 Mat3Inv(imat, mat);
2157                                 Mat3MulVecfl(imat, nor);
2158                         }
2159                         else if(R.r.bake_normal_space == R_BAKE_SPACE_OBJECT)
2160                                 Mat4Mul3Vecfl(ob->imat, nor); /* ob->imat includes viewinv! */
2161                         else if(R.r.bake_normal_space == R_BAKE_SPACE_WORLD)
2162                                 Mat4Mul3Vecfl(R.viewinv, nor);
2163
2164                         Normalize(nor); /* in case object has scaling */
2165
2166                         shr.combined[0]= nor[0]/2.0f + 0.5f;
2167                         shr.combined[1]= 0.5f - nor[1]/2.0f;
2168                         shr.combined[2]= nor[2]/2.0f + 0.5f;
2169                 }
2170                 else if(bs->type==RE_BAKE_TEXTURE) {
2171                         shr.combined[0]= shi->r;
2172                         shr.combined[1]= shi->g;
2173                         shr.combined[2]= shi->b;
2174                         shr.alpha = shi->alpha;
2175                 }
2176                 else if(bs->type==RE_BAKE_SHADOW) {
2177                         VECCOPY(shr.combined, shr.shad);
2178                         shr.alpha = shi->alpha;
2179                 }
2180         }
2181
2182         if(bs->rect_float) {
2183                 float *col= bs->rect_float + 4*(bs->rectx*y + x);
2184                 VECCOPY(col, shr.combined);
2185                 if (bs->type==RE_BAKE_ALL || bs->type==RE_BAKE_TEXTURE) {
2186                         col[3]= shr.alpha;
2187                 } else {
2188                         col[3]= 1.0;
2189                 }
2190         }
2191         else {
2192                 char *col= (char *)(bs->rect + bs->rectx*y + x);
2193                 col[0]= FTOCHAR(shr.combined[0]);
2194                 col[1]= FTOCHAR(shr.combined[1]);
2195                 col[2]= FTOCHAR(shr.combined[2]);
2196
2197
2198                 if (bs->type==RE_BAKE_ALL || bs->type==RE_BAKE_TEXTURE) {
2199                         col[3]= FTOCHAR(shr.alpha);
2200                 } else {
2201                         col[3]= 255;
2202                 }
2203         }
2204
2205         if (bs->rect_mask) {
2206                 bs->rect_mask[bs->rectx*y + x] = BAKE_MASK_BAKED;
2207         }
2208 }
2209
2210 static void bake_displacement(void *handle, ShadeInput *shi, float dist, int x, int y)
2211 {
2212         BakeShade *bs= handle;
2213         float disp;
2214
2215         if(R.r.bake_flag & R_BAKE_NORMALIZE && R.r.bake_maxdist) {
2216                 disp = (dist+R.r.bake_maxdist) / (R.r.bake_maxdist*2); /* alter the range from [-bake_maxdist, bake_maxdist] to [0, 1]*/
2217         } else {
2218                 disp = 0.5 + dist; /* alter the range from [-0.5,0.5] to [0,1]*/
2219         }
2220
2221         if(bs->rect_float) {
2222                 float *col= bs->rect_float + 4*(bs->rectx*y + x);
2223                 col[0] = col[1] = col[2] = disp;
2224                 col[3]= 1.0f;
2225         } else {
2226                 char *col= (char *)(bs->rect + bs->rectx*y + x);
2227                 col[0]= FTOCHAR(disp);
2228                 col[1]= FTOCHAR(disp);
2229                 col[2]= FTOCHAR(disp);
2230                 col[3]= 255;
2231         }
2232         if (bs->rect_mask) {
2233                 bs->rect_mask[bs->rectx*y + x] = BAKE_MASK_BAKED;
2234         }
2235 }
2236
2237 static int bake_check_intersect(Isect *is, int ob, RayFace *face)
2238 {
2239         BakeShade *bs = (BakeShade*)is->userdata;
2240
2241         /* no direction checking for now, doesn't always improve the result
2242          * (INPR(shi->facenor, bs->dir) > 0.0f); */
2243
2244         return (R.objectinstance[ob].obr->ob != bs->actob);
2245 }
2246
2247 static int bake_intersect_tree(RayTree* raytree, Isect* isect, float *start, float *dir, float sign, float *hitco, float *dist)
2248 {
2249         float maxdist;
2250         int hit;
2251
2252         /* might be useful to make a user setting for maxsize*/
2253         if(R.r.bake_maxdist > 0.0f)
2254                 maxdist= R.r.bake_maxdist;
2255         else
2256                 maxdist= RE_ray_tree_max_size(R.raytree) + R.r.bake_biasdist;
2257
2258         VECADDFAC(isect->start, start, dir, -R.r.bake_biasdist);
2259
2260         isect->end[0] = isect->start[0] + dir[0]*maxdist*sign;
2261         isect->end[1] = isect->start[1] + dir[1]*maxdist*sign;
2262         isect->end[2] = isect->start[2] + dir[2]*maxdist*sign;
2263
2264         hit = RE_ray_tree_intersect_check(R.raytree, isect, bake_check_intersect);
2265         if(hit) {
2266                 hitco[0] = isect->start[0] + isect->labda*isect->vec[0];
2267                 hitco[1] = isect->start[1] + isect->labda*isect->vec[1];
2268                 hitco[2] = isect->start[2] + isect->labda*isect->vec[2];
2269
2270                 *dist= VecLenf(start, hitco);
2271         }
2272
2273         return hit;
2274 }
2275
2276 static void bake_set_vlr_dxyco(BakeShade *bs, float *uv1, float *uv2, float *uv3)
2277 {
2278         VlakRen *vlr= bs->vlr;
2279         float A, d1, d2, d3, *v1, *v2, *v3;
2280
2281         if(bs->quad) {
2282                 v1= vlr->v1->co;
2283                 v2= vlr->v3->co;
2284                 v3= vlr->v4->co;
2285         }
2286         else {
2287                 v1= vlr->v1->co;
2288                 v2= vlr->v2->co;
2289                 v3= vlr->v3->co;
2290         }
2291
2292         /* formula derived from barycentric coordinates:
2293          * (uvArea1*v1 + uvArea2*v2 + uvArea3*v3)/uvArea
2294          * then taking u and v partial derivatives to get dxco and dyco */
2295         A= (uv2[0] - uv1[0])*(uv3[1] - uv1[1]) - (uv3[0] - uv1[0])*(uv2[1] - uv1[1]);
2296
2297         if(fabs(A) > FLT_EPSILON) {
2298                 A= 0.5f/A;
2299
2300                 d1= uv2[1] - uv3[1];
2301                 d2= uv3[1] - uv1[1];
2302                 d3= uv1[1] - uv2[1];
2303                 bs->dxco[0]= (v1[0]*d1 + v2[0]*d2 + v3[0]*d3)*A;
2304                 bs->dxco[1]= (v1[1]*d1 + v2[1]*d2 + v3[1]*d3)*A;
2305                 bs->dxco[2]= (v1[2]*d1 + v2[2]*d2 + v3[2]*d3)*A;
2306
2307                 d1= uv3[0] - uv2[0];
2308                 d2= uv1[0] - uv3[0];
2309                 d3= uv2[0] - uv1[0];
2310                 bs->dyco[0]= (v1[0]*d1 + v2[0]*d2 + v3[0]*d3)*A;
2311                 bs->dyco[1]= (v1[1]*d1 + v2[1]*d2 + v3[1]*d3)*A;
2312                 bs->dyco[2]= (v1[2]*d1 + v2[2]*d2 + v3[2]*d3)*A;
2313         }
2314         else {
2315                 bs->dxco[0]= bs->dxco[1]= bs->dxco[2]= 0.0f;
2316                 bs->dyco[0]= bs->dyco[1]= bs->dyco[2]= 0.0f;
2317         }
2318
2319         if(bs->obi->flag & R_TRANSFORMED) {
2320                 Mat3MulVecfl(bs->obi->nmat, bs->dxco);
2321                 Mat3MulVecfl(bs->obi->nmat, bs->dyco);
2322         }
2323 }
2324
2325 static void do_bake_shade(void *handle, int x, int y, float u, float v)
2326 {
2327         BakeShade *bs= handle;
2328         VlakRen *vlr= bs->vlr;
2329         ObjectInstanceRen *obi= bs->obi;
2330         Object *ob= obi->obr->ob;
2331         float l, *v1, *v2, *v3, tvn[3], ttang[3];
2332         int quad;
2333         ShadeSample *ssamp= &bs->ssamp;
2334         ShadeInput *shi= ssamp->shi;
2335
2336         /* fast threadsafe break test */
2337         if(R.test_break(R.tbh))
2338                 return;
2339
2340         /* setup render coordinates */
2341         if(bs->quad) {
2342                 v1= vlr->v1->co;
2343                 v2= vlr->v3->co;
2344                 v3= vlr->v4->co;
2345         }
2346         else {
2347                 v1= vlr->v1->co;
2348                 v2= vlr->v2->co;
2349                 v3= vlr->v3->co;
2350         }
2351
2352         /* renderco */
2353         l= 1.0f-u-v;
2354
2355         shi->co[0]= l*v3[0]+u*v1[0]+v*v2[0];
2356         shi->co[1]= l*v3[1]+u*v1[1]+v*v2[1];
2357         shi->co[2]= l*v3[2]+u*v1[2]+v*v2[2];
2358
2359         if(obi->flag & R_TRANSFORMED)
2360                 Mat4MulVecfl(obi->mat, shi->co);
2361
2362         VECCOPY(shi->dxco, bs->dxco);
2363         VECCOPY(shi->dyco, bs->dyco);
2364
2365         quad= bs->quad;
2366         bake_set_shade_input(obi, vlr, shi, quad, 0, x, y, u, v);
2367
2368         if(bs->type==RE_BAKE_NORMALS && R.r.bake_normal_space==R_BAKE_SPACE_TANGENT) {
2369                 shade_input_set_shade_texco(shi);
2370                 VECCOPY(tvn, shi->vn);
2371                 VECCOPY(ttang, shi->nmaptang);
2372         }
2373
2374         /* if we are doing selected to active baking, find point on other face */
2375         if(bs->actob) {
2376                 Isect isec, minisec;
2377                 float co[3], minco[3], dist, mindist=0.0f;
2378                 int hit, sign, dir=1;
2379
2380                 /* intersect with ray going forward and backward*/
2381                 hit= 0;
2382                 memset(&minisec, 0, sizeof(minisec));
2383                 minco[0]= minco[1]= minco[2]= 0.0f;
2384
2385                 VECCOPY(bs->dir, shi->vn);
2386
2387                 for(sign=-1; sign<=1; sign+=2) {
2388                         memset(&isec, 0, sizeof(isec));
2389                         isec.mode= RE_RAY_MIRROR;
2390                         isec.faceorig= (RayFace*)vlr;
2391                         isec.oborig= RAY_OBJECT_SET(&R, obi);
2392                         isec.userdata= bs;
2393
2394                         if(bake_intersect_tree(R.raytree, &isec, shi->co, shi->vn, sign, co, &dist)) {
2395                                 if(!hit || VecLenf(shi->co, co) < VecLenf(shi->co, minco)) {
2396                                         minisec= isec;
2397                                         mindist= dist;
2398                                         VECCOPY(minco, co);
2399                                         hit= 1;
2400                                         dir = sign;
2401                                 }
2402                         }
2403                 }
2404
2405                 if (bs->type==RE_BAKE_DISPLACEMENT) {
2406                         if(hit)
2407                                 bake_displacement(handle, shi, (dir==-1)? mindist:-mindist, x, y);
2408                         else
2409                                 bake_displacement(handle, shi, 0.0f, x, y);
2410                         return;
2411                 }
2412
2413                 /* if hit, we shade from the new point, otherwise from point one starting face */
2414                 if(hit) {
2415                         vlr= (VlakRen*)minisec.face;
2416                         obi= RAY_OBJECT_GET(&R, minisec.ob);
2417                         quad= (minisec.isect == 2);
2418                         VECCOPY(shi->co, minco);
2419
2420                         u= -minisec.u;
2421                         v= -minisec.v;
2422                         bake_set_shade_input(obi, vlr, shi, quad, 1, x, y, u, v);
2423                 }
2424         }
2425
2426         if(bs->type==RE_BAKE_NORMALS && R.r.bake_normal_space==R_BAKE_SPACE_TANGENT)
2427                 bake_shade(handle, ob, shi, quad, x, y, u, v, tvn, ttang);
2428         else
2429                 bake_shade(handle, ob, shi, quad, x, y, u, v, 0, 0);
2430 }
2431
2432 static int get_next_bake_face(BakeShade *bs)
2433 {
2434         ObjectRen *obr;
2435         VlakRen *vlr;
2436         MTFace *tface;
2437         static int v= 0, vdone= 0;
2438         static ObjectInstanceRen *obi= NULL;
2439
2440         if(bs==NULL) {
2441                 vlr= NULL;
2442                 v= vdone= 0;
2443                 obi= R.instancetable.first;
2444                 return 0;
2445         }
2446
2447         BLI_lock_thread(LOCK_CUSTOM1);
2448
2449         for(; obi; obi=obi->next, v=0) {
2450                 obr= obi->obr;
2451
2452                 for(; v<obr->totvlak; v++) {
2453                         vlr= RE_findOrAddVlak(obr, v);
2454
2455                         if((bs->actob && bs->actob == obr->ob) || (!bs->actob && (obr->ob->flag & SELECT))) {
2456                                 tface= RE_vlakren_get_tface(obr, vlr, obr->bakemtface, NULL, 0);
2457
2458                                 if(tface && tface->tpage) {
2459                                         Image *ima= tface->tpage;
2460                                         ImBuf *ibuf= BKE_image_get_ibuf(ima, NULL);
2461                                         float vec[4]= {0.0f, 0.0f, 0.0f, 0.0f};
2462
2463                                         if(ibuf==NULL)
2464                                                 continue;
2465
2466                                         if(ibuf->rect==NULL && ibuf->rect_float==NULL)
2467                                                 continue;
2468
2469                                         if(ibuf->rect_float && !(ibuf->channels==0 || ibuf->channels==4))
2470                                                 continue;
2471
2472                                         /* find the image for the first time? */
2473                                         if(ima->id.flag & LIB_DOIT) {
2474                                                 ima->id.flag &= ~LIB_DOIT;
2475
2476                                                 /* we either fill in float or char, this ensures things go fine */
2477                                                 if(ibuf->rect_float)
2478                                                         imb_freerectImBuf(ibuf);
2479                                                 /* clear image */
2480                                                 if(R.r.bake_flag & R_BAKE_CLEAR)
2481                                                         IMB_rectfill(ibuf, vec);
2482
2483                                                 /* might be read by UI to set active image for display */
2484                                                 R.bakebuf= ima;
2485                                         }
2486
2487                                         bs->obi= obi;
2488                                         bs->vlr= vlr;
2489
2490                                         bs->vdone++;    /* only for error message if nothing was rendered */
2491                                         v++;
2492
2493                                         BLI_unlock_thread(LOCK_CUSTOM1);
2494                                         return 1;
2495                                 }
2496                         }
2497                 }
2498         }
2499
2500         BLI_unlock_thread(LOCK_CUSTOM1);
2501         return 0;
2502 }
2503
2504 /* already have tested for tface and ima and zspan */
2505 static void shade_tface(BakeShade *bs)
2506 {
2507         VlakRen *vlr= bs->vlr;
2508         ObjectInstanceRen *obi= bs->obi;
2509         ObjectRen *obr= obi->obr;
2510         MTFace *tface= RE_vlakren_get_tface(obr, vlr, obr->bakemtface, NULL, 0);
2511         Image *ima= tface->tpage;
2512         float vec[4][2];
2513         int a, i1, i2, i3;
2514
2515         /* check valid zspan */
2516         if(ima!=bs->ima) {
2517                 bs->ima= ima;
2518                 bs->ibuf= BKE_image_get_ibuf(ima, NULL);
2519                 /* note, these calls only free/fill contents of zspan struct, not zspan itself */
2520                 zbuf_free_span(bs->zspan);
2521                 zbuf_alloc_span(bs->zspan, bs->ibuf->x, bs->ibuf->y, R.clipcrop);
2522         }
2523
2524         bs->rectx= bs->ibuf->x;
2525         bs->recty= bs->ibuf->y;
2526         bs->rect= bs->ibuf->rect;
2527         bs->rect_float= bs->ibuf->rect_float;
2528         bs->quad= 0;
2529
2530         if (bs->usemask) {
2531                 if (bs->ibuf->userdata==NULL) {
2532                         BLI_lock_thread(LOCK_CUSTOM1);
2533                         if (bs->ibuf->userdata==NULL) { /* since the thread was locked, its possible another thread alloced the value */
2534                                 bs->ibuf->userdata = (void *)MEM_callocN(sizeof(char)*bs->rectx*bs->recty, "BakeMask");
2535                                 bs->rect_mask= (char *)bs->ibuf->userdata;
2536                         }
2537                         BLI_unlock_thread(LOCK_CUSTOM1);
2538                 } else {
2539                         bs->rect_mask= (char *)bs->ibuf->userdata;
2540                 }
2541         }
2542
2543         /* get pixel level vertex coordinates */
2544         for(a=0; a<4; a++) {
2545                 vec[a][0]= tface->uv[a][0]*(float)bs->rectx - 0.5f;
2546                 vec[a][1]= tface->uv[a][1]*(float)bs->recty - 0.5f;
2547         }
2548
2549         /* UV indices have to be corrected for possible quad->tria splits */
2550         i1= 0; i2= 1; i3= 2;
2551         vlr_set_uv_indices(vlr, &i1, &i2, &i3);
2552         bake_set_vlr_dxyco(bs, vec[i1], vec[i2], vec[i3]);
2553         zspan_scanconvert(bs->zspan, bs, vec[i1], vec[i2], vec[i3], do_bake_shade);
2554
2555         if(vlr->v4) {
2556                 bs->quad= 1;
2557                 bake_set_vlr_dxyco(bs, vec[0], vec[2], vec[3]);
2558                 zspan_scanconvert(bs->zspan, bs, vec[0], vec[2], vec[3], do_bake_shade);
2559         }
2560 }
2561
2562 static void *do_bake_thread(void *bs_v)
2563 {
2564         BakeShade *bs= bs_v;
2565
2566         while(get_next_bake_face(bs)) {
2567                 shade_tface(bs);
2568
2569                 /* fast threadsafe break test */
2570                 if(R.test_break(R.tbh))
2571                         break;
2572         }
2573         bs->ready= 1;
2574
2575         return NULL;
2576 }
2577
2578 /* using object selection tags, the faces with UV maps get baked */
2579 /* render should have been setup */
2580 /* returns 0 if nothing was handled */
2581 int RE_bake_shade_all_selected(Render *re, int type, Object *actob)
2582 {
2583         BakeShade handles[BLENDER_MAX_THREADS];
2584         ListBase threads;
2585         Image *ima;
2586         int a, vdone=0, usemask=0;
2587
2588         /* initialize render global */
2589         R= *re;
2590         R.bakebuf= NULL;
2591
2592         /* initialize static vars */
2593         get_next_bake_face(NULL);
2594
2595         /* do we need a mask? */
2596         if (re->r.bake_filter && (re->r.bake_flag & R_BAKE_CLEAR)==0)
2597                 usemask = 1;
2598
2599         /* baker uses this flag to detect if image was initialized */
2600         for(ima= G.main->image.first; ima; ima= ima->id.next) {
2601                 ImBuf *ibuf= BKE_image_get_ibuf(ima, NULL);
2602                 ima->id.flag |= LIB_DOIT;
2603                 if (ibuf)
2604                         ibuf->userdata = NULL; /* use for masking if needed */
2605         }
2606
2607         BLI_init_threads(&threads, do_bake_thread, re->r.threads);
2608
2609         /* get the threads running */
2610         for(a=0; a<re->r.threads; a++) {
2611                 /* set defaults in handles */
2612                 memset(&handles[a], 0, sizeof(BakeShade));
2613
2614                 handles[a].ssamp.shi[0].lay= re->scene->lay;
2615
2616                 if (type==RE_BAKE_SHADOW) {
2617                         handles[a].ssamp.shi[0].passflag= SCE_PASS_SHADOW;
2618                 } else {
2619                         handles[a].ssamp.shi[0].passflag= SCE_PASS_COMBINED;
2620                 }
2621                 handles[a].ssamp.shi[0].combinedflag= ~(SCE_PASS_SPEC);
2622                 handles[a].ssamp.shi[0].thread= a;
2623                 handles[a].ssamp.tot= 1;
2624
2625                 handles[a].type= type;
2626                 handles[a].actob= actob;
2627                 handles[a].zspan= MEM_callocN(sizeof(ZSpan), "zspan for bake");
2628
2629                 handles[a].usemask = usemask;
2630
2631                 BLI_insert_thread(&threads, &handles[a]);
2632         }
2633
2634         /* wait for everything to be done */
2635         a= 0;
2636         while(a!=re->r.threads) {
2637
2638                 PIL_sleep_ms(50);
2639
2640                 for(a=0; a<re->r.threads; a++)
2641                         if(handles[a].ready==0)
2642                                 break;
2643         }
2644
2645         /* filter and refresh images */
2646         for(ima= G.main->image.first; ima; ima= ima->id.next) {
2647                 if((ima->id.flag & LIB_DOIT)==0) {
2648                         ImBuf *ibuf= BKE_image_get_ibuf(ima, NULL);
2649                         if (re->r.bake_filter) {
2650                                 if (usemask) {
2651                                         /* extend the mask +2 pixels from the image,
2652                                          * this is so colors dont blend in from outside */
2653                                         char *temprect;
2654
2655                                         for(a=0; a<re->r.bake_filter; a++)
2656                                                 bake_mask_filter_extend((char *)ibuf->userdata, ibuf->x, ibuf->y);
2657
2658                                         temprect = MEM_dupallocN(ibuf->userdata);
2659
2660                                         /* expand twice to clear this many pixels, so they blend back in */
2661                                         bake_mask_filter_extend(temprect, ibuf->x, ibuf->y);
2662                                         bake_mask_filter_extend(temprect, ibuf->x, ibuf->y);
2663
2664                                         /* clear all pixels in the margin*/
2665                                         bake_mask_clear(ibuf, temprect, BAKE_MASK_MARGIN);
2666                                         MEM_freeN(temprect);
2667                                 }
2668
2669                                 for(a=0; a<re->r.bake_filter; a++) {
2670                                         /*the mask, ibuf->userdata - can be null, in this case only zero alpha is used */
2671                                         IMB_filter_extend(ibuf, (char *)ibuf->userdata);
2672                                 }
2673
2674                                 if (ibuf->userdata) {
2675                                         MEM_freeN(ibuf->userdata);
2676                                         ibuf->userdata= NULL;
2677                                 }
2678                         }
2679                         ibuf->userflags |= IB_BITMAPDIRTY;
2680                         if (ibuf->rect_float) IMB_rect_from_float(ibuf);
2681                 }
2682         }
2683
2684         /* calculate return value */
2685         for(a=0; a<re->r.threads; a++) {
2686                 vdone+= handles[a].vdone;
2687
2688                 zbuf_free_span(handles[a].zspan);
2689                 MEM_freeN(handles[a].zspan);
2690         }
2691
2692         BLI_end_threads(&threads);
2693         return vdone;
2694 }
2695
2696 struct Image *RE_bake_shade_get_image(void)
2697 {
2698         return R.bakebuf;
2699 }
2700