File: next_hit_device.cpp

package info (click to toggle)
embree 3.13.5%2Bdfsg-2
  • links: PTS, VCS
  • area: main
  • in suites: bookworm
  • size: 27,924 kB
  • sloc: cpp: 180,815; xml: 3,877; ansic: 2,957; python: 1,466; sh: 502; makefile: 229; csh: 42
file content (496 lines) | stat: -rw-r--r-- 15,504 bytes parent folder | download
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
// Copyright 2009-2021 Intel Corporation
// SPDX-License-Identifier: Apache-2.0

#include "../common/math/random_sampler.h"
#include "../common/core/differential_geometry.h"
#include "../common/tutorial/tutorial_device.h"
#include "../common/tutorial/scene_device.h"
#include <algorithm>
#include "../common/tutorial/tutorial.h"

#define SINGLE_PASS 0
#define MULTI_PASS_FIXED_NEXT_HITS 1
#define MULTI_PASS_OPTIMAL_NEXT_HITS 2
#define MULTI_PASS_ESTIMATED_NEXT_HITS 3

namespace embree {

extern "C" ISPCScene* g_ispc_scene;
extern "C" int g_instancing_mode;
extern "C" int g_next_hit_mode;
extern "C" unsigned g_max_next_hits;
extern "C" unsigned g_max_total_hits;
extern "C" bool g_verify;
extern "C" bool g_visualize_errors;
extern "C" bool g_enable_opacity;
extern "C" float g_curve_opacity;

#define MAX_TOTAL_HITS 16*1024

/* number of hits found in previous pass */
int* g_num_prev_hits = nullptr;
unsigned int g_num_prev_hits_width = 0;
unsigned int g_num_prev_hits_height = 0;


/* extended ray structure that gathers all hits along the ray */
struct HitList
{
  HitList ()
    : begin(0), end(0) {}

  /* Hit structure that defines complete order over hits */
  struct Hit
  {
    Hit() {}

    Hit (bool opaque, float t, unsigned int primID = 0xFFFFFFFF, unsigned int geomID = 0xFFFFFFFF, unsigned int instID = 0xFFFFFFFF)
      : opaque(opaque), t(t), primID(primID), geomID(geomID), instID(instID) {}

    /* lexicographical order (t,instID,geomID,primID) */
    __forceinline friend bool operator < (const Hit& a, const Hit& b)
    {
      if (a.t == b.t) {
        if (a.instID == b.instID) {
          if (a.geomID == b.geomID) return a.primID < b.primID;
          else                      return a.geomID < b.geomID;
        }
        else return a.instID < b.instID;
      }
      return a.t < b.t;
    }

    __forceinline friend bool operator == (const Hit& a, const Hit& b) {
      return a.t == b.t && a.primID == b.primID && a.geomID == b.geomID && a.instID == b.instID;
    }

    __forceinline friend bool operator <= (const Hit& a, const Hit& b)
    {
      if (a == b) return true;
      else return a < b;
    }

    __forceinline friend bool operator != (const Hit& a, const Hit& b) {
      return !(a == b);
    }

    friend std::ostream& operator<<(std::ostream& cout, const Hit& hit) {
      return cout << "Hit { opaque = " << hit.opaque << ", t = " << hit.t << ", instID = " << hit.instID << ", geomID = " << hit.geomID << ", primID = " << hit.primID << " }";
    }

  public:
    bool opaque;
    float t;
    unsigned int primID;
    unsigned int geomID;
    unsigned int instID;
  };

  /* return number of gathered hits */
  unsigned int size() const {
    return end-begin;
  }

  /* returns the last hit */
  const Hit& last() const {
    assert(end);
    return hits[end-1];
  }

  /* checks if the last hit is opaque */
  bool last_is_opaque() const {
    return size() && last().opaque;
  }

public:
  unsigned int begin;   // begin of hit list
  unsigned int end;     // end of hit list
  Hit hits[MAX_TOTAL_HITS];   // array to store all found hits to
};

/* we store the Hit list inside the intersection context to access it from the filter functions */
struct IntersectContext
{
  IntersectContext(HitList& hits)
    : hits(hits), max_next_hits(g_max_next_hits) {}

  RTCIntersectContext context;
  HitList& hits;
  unsigned int max_next_hits; // maximal number of hits to collect in a single pass
};

/* scene data */
RTCScene g_scene = nullptr;

RTCScene convertScene(ISPCScene* scene_in)
{
  RTCScene scene_out = ConvertScene(g_device, g_ispc_scene, RTC_BUILD_QUALITY_MEDIUM, RTC_SCENE_FLAG_CONTEXT_FILTER_FUNCTION | RTC_SCENE_FLAG_ROBUST);
  
  /* commit changes to scene */
  return scene_out;
}

/* Filter callback function that gathers all hits */
void gather_all_hits(const struct RTCFilterFunctionNArguments* args)
{
  assert(*args->valid == -1);
  IntersectContext* context = (IntersectContext*) args->context;
  HitList& hits = context->hits;
  RTCRay* ray = (RTCRay*) args->ray;
  RTCHit* hit = (RTCHit*) args->hit;
  assert(args->N == 1);
  args->valid[0] = 0; // ignore all hits

  /* avoid overflow of hits array */
  if (hits.end >= MAX_TOTAL_HITS) return;

  /* check if geometry is opaque */
  ISPCGeometry* geometry = (ISPCGeometry*) args->geometryUserPtr;
  bool opaque = !g_enable_opacity || geometry->type != CURVES;
  
  /* add hit to list */
  hits.hits[hits.end++] = HitList::Hit(opaque,ray->tfar,hit->primID,hit->geomID,hit->instID[0]);
}

/* gathers hits in a single pass */
void single_pass(const Ray& ray_i, HitList& hits_o, RandomSampler& sampler, RayStats& stats)
{
  /* trace ray to gather all hits */
  Ray ray = ray_i;
  IntersectContext context(hits_o);
  rtcInitIntersectContext(&context.context);
  context.context.filter = gather_all_hits;
  rtcIntersect1(g_scene,&context.context,RTCRayHit_(ray));
  RayStats_addRay(stats);

  /* sort hits by extended order */
  std::sort(&context.hits.hits[context.hits.begin],&context.hits.hits[context.hits.end]);

  /* ignore duplicated hits that can occur for tessellated primitives */
  if (hits_o.size())
  {
    unsigned int i=0, j=1;
    for (; j<hits_o.size(); j++) {
      if (hits_o.hits[i] == hits_o.hits[j]) continue;
      hits_o.hits[++i] = hits_o.hits[j];
    }
    hits_o.end = i+1;
  }

  /* drop hits in case we found too many */
  hits_o.end = std::min(hits_o.end, g_max_total_hits);

  /* shade all hits */
  if (g_enable_opacity)
  {
    for (unsigned int i=context.hits.begin; i<context.hits.end; i++)
    {
      /* roussion roulette ray termination */
      bool opaque = context.hits.hits[i].opaque;
      if (RandomSampler_get1D(sampler) < g_curve_opacity)
        opaque = true;
      
      if (opaque) {
        hits_o.end = i+1;
        return;
      }
    }
  }
}

/* Filter callback function that gathers first N hits up to the first opaque surface */
void gather_next_hits(const struct RTCFilterFunctionNArguments* args)
{
  assert(*args->valid == -1);
  IntersectContext* context = (IntersectContext*) args->context;
  HitList& hits = context->hits;
  RTCRay* ray = (RTCRay*) args->ray;
  RTCHit* hit = (RTCHit*) args->hit;
  assert(args->N == 1);
  args->valid[0] = 0; // ignore all hits

  /* avoid overflow of hits array */
  if (hits.end >= MAX_TOTAL_HITS) return;

  /* check if geometry is opaque */
  ISPCGeometry* geometry = (ISPCGeometry*) args->geometryUserPtr;
  bool opaque = !g_enable_opacity || geometry->type != CURVES;
  
  HitList::Hit nhit(opaque, ray->tfar,hit->primID,hit->geomID,hit->instID[0]);

  /* ignore already found hits */
  if (hits.begin > 0 && nhit <= hits.hits[hits.begin-1])
    return;

  /* insert new hit at proper location */
  for (unsigned int i=hits.begin; i<hits.end; i++)
  {
    if (nhit < hits.hits[i]) {
      std::swap(nhit,hits.hits[i]);
      if (hits.hits[i].opaque) {
        hits.end = i+1;
        break;
      }
    }
  }

  /* store farthest hit if place left and last is not opaque */
  if (hits.size() < context->max_next_hits && hits.end < g_max_total_hits && !hits.last_is_opaque())
    hits.hits[hits.end++] = nhit;

  /* shrink tfar when we collected sufficient hits for this pass, or the last hit is opaque */
  if (hits.size() == context->max_next_hits || hits.last_is_opaque())
  {
    ray->tfar = hits.last().t;
    args->valid[0] = -1; // accept hit
  }
}

/* gathers hits in multiple passes */
void multi_pass(const Ray& ray_i, HitList& hits_o, int max_next_hits, RandomSampler& sampler, RayStats& stats)
{
  /* configure intersect context */
  Ray ray = ray_i;
  IntersectContext context(hits_o);
  rtcInitIntersectContext(&context.context);
  context.max_next_hits = max_next_hits;
  context.context.filter = gather_next_hits;

  /* in each pass we collect some hits */
  do {

    /* continue from previous fartherst hit */
    if (context.hits.end) 
      ray.tnear() = context.hits.last().t;

    /* initialize ray */
    ray.tfar = inf;
    ray.geomID = RTC_INVALID_GEOMETRY_ID;
    ray.instID[0] = RTC_INVALID_GEOMETRY_ID;

    /* insert new hits at previous end of hits list */
    context.hits.begin = context.hits.end;
    for (size_t i=0; i<context.max_next_hits; i++)
      if (context.hits.begin+i < g_max_total_hits)
        context.hits.hits[context.hits.begin+i] = HitList::Hit(false,neg_inf);

    rtcIntersect1(g_scene,&context.context,RTCRayHit_(ray));
    RayStats_addRay(stats);

    /* shade all hits */
    if (g_enable_opacity)
    {
      for (unsigned int i=context.hits.begin; i<context.hits.end; i++)
      {
        /* roussion roulette ray termination */
        bool opaque = context.hits.hits[i].opaque;
        if (RandomSampler_get1D(sampler) < g_curve_opacity)
          opaque = true;

        /* remove all farther hits in case we terminate here */
        if (opaque) {
          context.hits.begin = 0;
          context.hits.end = i+1;
          return;
        }
      }
    }
    
  } while (context.hits.size() != 0);
  
  context.hits.begin = 0;
}

/* task that renders a single screen tile */
Vec3ff renderPixelStandard(float x, float y, const ISPCCamera& camera, RayStats& stats)
{
  /* initialize sampler */
  const int ix = (int)x;
  const int iy = (int)y;
  RandomSampler mysampler;
  RandomSampler_init(mysampler, ix, iy, 0);
  
  /* initialize ray */
  Ray ray(Vec3fa(camera.xfm.p), Vec3fa(normalize(x*camera.xfm.l.vx + y*camera.xfm.l.vy + camera.xfm.l.vz)), 0.0f, inf, 0.0f);

  /* either gather hits in single pass or using multiple passes */
  HitList hits;
  switch (g_next_hit_mode)
  {
  case SINGLE_PASS:
    single_pass(ray,hits,mysampler,stats);
    break;
  
  case MULTI_PASS_FIXED_NEXT_HITS:
    multi_pass (ray,hits,g_max_next_hits,mysampler,stats);
    break;

  case MULTI_PASS_OPTIMAL_NEXT_HITS: {
    int num_prev_hits = max(1,g_num_prev_hits[iy*TutorialApplication::instance->width+ix]);
    multi_pass (ray,hits,num_prev_hits,mysampler,stats);
    break;
  }
  case MULTI_PASS_ESTIMATED_NEXT_HITS: {
    int estimated_num_next_hits = (int) min((float)g_max_next_hits, max(1.0f, 0.5f/g_curve_opacity));
    multi_pass (ray,hits,estimated_num_next_hits,mysampler,stats);
    break;
  }
  default:
    assert(false);
  }
   
  /* verify result with gathering all hits */
  bool has_error = false;
  if (g_verify || g_visualize_errors)
  {
    /* repeat using a single pass, which is assumed to produce the correct result */
    HitList verify_hits;
    RandomSampler verify_sampler;
    RandomSampler_init(verify_sampler, ix, iy, 0);
    single_pass(ray,verify_hits,verify_sampler,stats);

    //std::cout << std::hexfloat;
    //for (size_t i=0; i<hits.size(); i++)
    //  PRINT2(i,hits.hits[i]);
     
    //for (size_t i=0; i<verify_hits.size(); i++)
    //  PRINT2(i,verify_hits.hits[i]);
    
    if (verify_hits.size() != hits.size())
      has_error = true;
    
    for (size_t i=verify_hits.begin; i<verify_hits.end; i++)
    {
      if (verify_hits.hits[i] != hits.hits[i])
        has_error = true;
    }

    if (!g_visualize_errors && has_error)
      throw std::runtime_error("hits differ");
  }

  /* calculate random sequence based on hit geomIDs and primIDs */
  RandomSampler sampler = { 0 };
  for (size_t i=hits.begin; i<hits.end; i++) {
    sampler.s = MurmurHash3_mix(sampler.s, hits.hits[i].instID);
    sampler.s = MurmurHash3_mix(sampler.s, hits.hits[i].geomID);
    sampler.s = MurmurHash3_mix(sampler.s, hits.hits[i].primID);
  }
  sampler.s = MurmurHash3_finalize(sampler.s);

  /* map geomID/primID sequence to color */
  Vec3ff color;
  color.x = RandomSampler_getFloat(sampler);
  color.y = RandomSampler_getFloat(sampler);
  color.z = RandomSampler_getFloat(sampler);

  /* mark errors red */
  if (g_visualize_errors)
  {
    color.x = color.y = color.z;
    if (has_error)
      color = Vec3ff(1,0,0);
  }
  
  color.w = (float) hits.size();
  return color;
}

/* renders a single screen tile */
void renderTileStandard(int taskIndex,
                        int threadIndex,
                        int* pixels,
                        const unsigned int width,
                        const unsigned int height,
                        const float time,
                        const ISPCCamera& camera,
                        const int numTilesX,
                        const int numTilesY)
{
  const int t = taskIndex;
  const unsigned int tileY = t / numTilesX;
  const unsigned int tileX = t - tileY * numTilesX;
  const unsigned int x0 = tileX * TILE_SIZE_X;
  const unsigned int x1 = min(x0+TILE_SIZE_X,width);
  const unsigned int y0 = tileY * TILE_SIZE_Y;
  const unsigned int y1 = min(y0+TILE_SIZE_Y,height);

  for (unsigned int y=y0; y<y1; y++) for (unsigned int x=x0; x<x1; x++)
  {
    Vec3ff color = renderPixelStandard((float)x,(float)y,camera,g_stats[threadIndex]);

    /* write color to framebuffer */
    unsigned int r = (unsigned int) (255.0f * clamp(color.x,0.0f,1.0f));
    unsigned int g = (unsigned int) (255.0f * clamp(color.y,0.0f,1.0f));
    unsigned int b = (unsigned int) (255.0f * clamp(color.z,0.0f,1.0f));
    pixels[y*width+x] = (b << 16) + (g << 8) + r;
    g_num_prev_hits[y*width+x] = (int) color.w;
  }
}

/* task that renders a single screen tile */
void renderTileTask (int taskIndex, int threadIndex, int* pixels,
                         const unsigned int width,
                         const unsigned int height,
                         const float time,
                         const ISPCCamera& camera,
                         const int numTilesX,
                         const int numTilesY)
{
  renderTileStandard(taskIndex,threadIndex,pixels,width,height,time,camera,numTilesX,numTilesY);
}

/* called by the C++ code for initialization */
extern "C" void device_init (const char* cfg)
{
}

extern "C" void renderFrameStandard (int* pixels,
                          const unsigned int width,
                          const unsigned int height,
                          const float time,
                          const ISPCCamera& camera)
{
  /* render image */
  const int numTilesX = (width +TILE_SIZE_X-1)/TILE_SIZE_X;
  const int numTilesY = (height+TILE_SIZE_Y-1)/TILE_SIZE_Y;
  parallel_for(size_t(0),size_t(numTilesX*numTilesY),[&](const range<size_t>& range) {
    const int threadIndex = (int)TaskScheduler::threadIndex();
    for (size_t i=range.begin(); i<range.end(); i++)
      renderTileTask((int)i,threadIndex,(int*)pixels,width,height,time,camera,numTilesX,numTilesY);
  });
}

/* called by the C++ code to render */
extern "C" void device_render (unsigned* pixels,
                           const unsigned int width,
                           const unsigned int height,
                           const float time,
                           const ISPCCamera& camera)
{
  /* create scene */
  if (g_scene == nullptr) {
    g_scene = convertScene(g_ispc_scene);
    rtcCommitScene (g_scene);
  }

  /* create buffer to remember previous number of hits found */
  if (!g_num_prev_hits || g_num_prev_hits_width != width || g_num_prev_hits_height != height)
  {
    delete[] g_num_prev_hits;
    g_num_prev_hits = new int[width*height];
    g_num_prev_hits_width = width;
    g_num_prev_hits_height = height;
    for (unsigned int i=0; i<width*height; i++)
      g_num_prev_hits[i] = 1;
  }
}

/* called by the C++ code for cleanup */
extern "C" void device_cleanup ()
{
  rtcReleaseScene (g_scene); g_scene = nullptr;
  delete[] g_num_prev_hits; g_num_prev_hits = nullptr;
}

} // namespace embree