File: multi_instanced_geometry_device.cpp

package info (click to toggle)
embree 4.3.3%2Bdfsg-1
  • links: PTS, VCS
  • area: main
  • in suites: forky, sid, trixie
  • size: 100,656 kB
  • sloc: cpp: 228,918; xml: 40,944; ansic: 2,685; python: 812; sh: 635; makefile: 228; csh: 42
file content (402 lines) | stat: -rw-r--r-- 12,923 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
// Copyright 2009-2021 Intel Corporation
// SPDX-License-Identifier: Apache-2.0

#include "multi_instanced_geometry_device.h"

namespace embree {

/* all features required by this tutorial */
#define FEATURE_MASK \
  RTC_FEATURE_FLAG_TRIANGLE | \
  RTC_FEATURE_FLAG_INSTANCE
  
RTCScene g_scene = nullptr;
TutorialData g_data;
extern "C" bool g_changed;

/*
 * There is an issue in ISPC where foreach_tiled can generate
 * empty gangs. This is problematic when scalar operations
 * (e.g. increment) run inside the loop.
 */
#define FOREACH_TILED_MITIGATION if (1 == 0) { continue; }

/* 
 * Accumulate an instance transformation given an instance stack. 
 * We use this only for normal transformations in this example.
 */
LinearSpace3fa accumulateNormalTransform(const TutorialData& data,
                                        const Ray& ray,
                                        float time)
{
  LinearSpace3fa transform = LinearSpace3fa(one);
  for (unsigned int level = 0; level < RTC_MAX_INSTANCE_LEVEL_COUNT && ray.instID[level] != RTC_INVALID_GEOMETRY_ID; ++level)
  {
    assert(level < data.g_instanceLevels.numLevels);
    const unsigned int instId = ray.instID[level];
    assert(instId < data.g_instanceLevels.numInstancesOnLevel[level]);
    LinearSpace3fa M = data.g_instanceLevels.normalTransforms[level][instId];
    transform = transform * LinearSpace3fa(M);
  }
  return transform;
}


/*
 * A simplistic sky model consisting of a directional sun
 * and a constant sky.
 */
void sampleLightDirection(const Vec3fa& xi,
                          Vec3fa& dir,
                          Vec3fa& emission)
{
  const Vec3fa sunDir = normalize(Vec3fa(-1.f, 1.f, 1.f));
  const Vec3fa sunEmission = Vec3fa(1.f);
  const Vec3fa skyEmission = Vec3fa(.2f);
  const float skyPdf = 1.f/4.f/float(M_PI);
  const float sunWeight = .1f; // Put most samples into the sky, the sun will converge instantly.

  if (xi.z < sunWeight)
    dir = sunDir;
  else
  {
    // Uniform sphere sampling around +Y axis.
    const float theta = acos(1.f - 2.f * xi.x);
    const float phi = 2.f * float(M_PI) * xi.y;
    const float st = sin(theta);
    dir = Vec3fa(st * cos(phi), cos(theta), -st * sin(phi));
  }

  emission = skyEmission;
  float pdf = (1.f-sunWeight) * skyPdf;
  if (sunDir.x == dir.x && sunDir.y == dir.y && sunDir.z == dir.z)
  {
    emission = emission + sunEmission;
    pdf = pdf + sunWeight;
  }

  emission = emission * rcp(pdf);
}

/*
 * Pixel filter importance sampling.
 * This uses the Box-Mueller transform to obtain Gaussian samples.
 */
Vec2f sampleGaussianPixelFilter(RandomSampler& sampler)
{
  const float phi = 2.f * float(M_PI) * RandomSampler_get1D(sampler);
  const float threeSigma = 1.f;
  const float sigma = threeSigma / 3.f;

  // Rejection sampling: we don't want any samples outside 3 sigma.
  float radius = (float)inf;
  while (radius <= 0.f || radius > threeSigma)
  {
    const float xi = RandomSampler_get1D(sampler);
    radius = sqrtf(sigma * (-2.f) * log(xi));
  }

  return Vec2f(radius * cos(phi), radius * sin(phi));
}

/*
 * Sample a primary ray.
 */
Ray samplePrimaryRay(const TutorialData& data,
                     unsigned int x,
                     unsigned int x0,
                     unsigned int y,
                     unsigned int y0,
                     const ISPCCamera& camera,
                     RandomSampler& sampler,
                     RayStats& stats)
{
  RandomSampler_init(sampler, (int)x, (int)y, data.g_accu_count);

  const unsigned int id = (y-y0) * TILE_SIZE_X + (x-x0);
  const Vec2f offset = sampleGaussianPixelFilter(sampler);
  const float fx = (float)x + 0.5f + offset.x;
  const float fy = (float)y + 0.5f + offset.y;
  const Vec3fa o = Vec3fa(camera.xfm.p);
  const Vec3fa w = Vec3fa(normalize(fx*camera.xfm.l.vx 
                                     + fy*camera.xfm.l.vy 
                                     + camera.xfm.l.vz));

  Ray ray;
  init_Ray(ray, o, w, 0.f, (float)inf);
  ray.id = id;
  RayStats_addRay(stats);
  return ray;
}

/* 
 * Make a new shadow ray.
 */
inline Ray makeShadowRay(const Ray& primary,
                         const Vec3fa& lightDir,
                         RayStats& stats)
{
  const Vec3fa o = primary.org + primary.tfar * primary.dir;

  Ray ray;
  init_Ray(ray, o, lightDir, 0.001f, (float)inf);
  ray.id = -1;
  RayStats_addShadowRay(stats);
  return ray;
} 

/*
 * Our shader for this scene: Lambertian shading with normal display.
 */
Vec3fa shade(const TutorialData& data,
             const Ray& primaryRay, 
             const Ray& shadowRay,
             const Vec3fa& lightDir,
             const Vec3fa& emission)
{
  if (primaryRay.geomID == RTC_INVALID_GEOMETRY_ID || shadowRay.tfar < 0.f)
    return Vec3fa(0.f);

  const LinearSpace3fa xfm = accumulateNormalTransform(data,primaryRay, 0.f);
  Vec3fa Ns = normalize(xfmVector(xfm, Vec3fa(primaryRay.Ng)));

  const float cosThetaOut = dot(Ns, lightDir);
  const float cosThetaIn = -dot(Ns, primaryRay.dir);

  // Block transmission.
  if ((cosThetaOut >= 0) != (cosThetaIn >= 0))
    return Vec3fa(0.f);

  // Make sure backfaces shade correctly.
  if (cosThetaIn < 0.f)
    Ns = -1.f * Ns;

  return emission
       * clamp(abs(cosThetaOut), 0.f, 1.f) 
       * (Vec3fa(0.5f) + 0.5f * Ns);
}

/*
 * Convert a floating-point value in [0, 1] to 8 bit.
 */
inline unsigned int floatToByte(float channel)
{
  channel = 255.1f * clamp(channel, 0.f, 1.f);
  return 0xff & (unsigned int) channel;
}

/*
 * Pack an RGB8 color value from three floats.
 */
inline unsigned int packRGB8(const Vec3fa& color)
{
  const unsigned int r = floatToByte(color.x);
  const unsigned int g = floatToByte(color.y);
  const unsigned int b = floatToByte(color.z);
  return (b << 16) + (g << 8) + r;
}

/*
 * Splat a color into the framebuffer.
 */
void splat(const TutorialData& data,
           int* pixels,
           const unsigned int width,
           unsigned int x,
           unsigned int y,
           const Vec3fa& color)
{
  const unsigned int pixIdx = y * width + x;
  const Vec3ff accu_color = data.g_accu[pixIdx] + Vec3ff(color.x,color.y,color.z,1.0f); 
  data.g_accu[pixIdx] = accu_color;
  if (accu_color.w > 0)
  {
    float f = rcp(accu_color.w);
    pixels[pixIdx] = packRGB8(Vec3fa(accu_color * f));
  }
}

void renderPixelStandard(const TutorialData& data, int x, int y,
                         int* pixels,
                         const unsigned int width,
                         const unsigned int height,
                         const float time,
                         const ISPCCamera& camera, RayStats& stats)
{
  RTCIntersectArguments iargs;
  rtcInitIntersectArguments(&iargs);
  iargs.feature_mask = (RTCFeatureFlags) (FEATURE_MASK);

  RTCOccludedArguments sargs;
  rtcInitOccludedArguments(&sargs);
  sargs.feature_mask = (RTCFeatureFlags) (FEATURE_MASK);

  RandomSampler sampler;
  Ray primaryRay = samplePrimaryRay(data, x, 0, y, 0, camera, sampler, stats);
  rtcIntersect1(data.g_scene, RTCRayHit_(primaryRay), &iargs);
  
  Vec3fa color = Vec3fa(0.f);
  if (primaryRay.geomID != RTC_INVALID_GEOMETRY_ID)
  {
    Vec3fa lightDir;
    Vec3fa emission;
    sampleLightDirection(RandomSampler_get3D(sampler), lightDir, emission);
    Ray shadowRay = makeShadowRay(primaryRay, lightDir, stats);
    rtcOccluded1(data.g_scene, RTCRay_(shadowRay), &sargs);
    color = shade(data, primaryRay, shadowRay, lightDir, emission);
  }
  
  splat(data, pixels, width, x, y, color);
}

/*
 * Render a single tile.
 */
void renderTileNormal(int* pixels,
                      const unsigned int width,
                      const unsigned int height,
                      const float time,
                      const ISPCCamera& camera,
                      const unsigned int x0,
                      const unsigned int x1,
                      const unsigned int y0,
                      const unsigned int y1,
                      RayStats& stats)
{
  for (unsigned int y=y0; y<y1; y++) for (unsigned int x=x0; x<x1; x++)
  {
    FOREACH_TILED_MITIGATION;
    renderPixelStandard(g_data,x,y,pixels,width,height,time,camera,stats);
  }
}

// ======================================================================== //
// TUTORIAL API.
// ======================================================================== //

/*
 * A 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)
{
  const unsigned int tileY = taskIndex / numTilesX;
  const unsigned int tileX = taskIndex - 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);

  renderTileNormal(pixels, width, height,
                   time, camera,
                   x0, x1, y0, y1,
                   g_stats[threadIndex]);
}

/* 
 * Called by the C++ code for initialization. 
 */
extern "C" void device_init(char* cfg)
{
  TutorialData_Constructor(&g_data);
  g_scene = g_data.g_scene = initializeScene(g_data, g_device);
}


extern "C" void renderFrameStandard(int* pixels,
                                const unsigned int width,
                                const unsigned int height,
                                const float time,
                                const ISPCCamera& camera)
{
#if defined(EMBREE_SYCL_TUTORIAL) && !defined(EMBREE_SYCL_RT_SIMULATION)
  TutorialData ldata = g_data;
  sycl::event event = global_gpu_queue->submit([=](sycl::handler& cgh){
    const sycl::nd_range<2> nd_range = make_nd_range(height,width);
    cgh.parallel_for(nd_range,[=](sycl::nd_item<2> item) {
      const unsigned int x = item.get_global_id(1); if (x >= width ) return;
      const unsigned int y = item.get_global_id(0); if (y >= height) return;
      RayStats stats;
      renderPixelStandard(ldata,x,y,pixels,width,height,time,camera,stats);
    });
  });
  global_gpu_queue->wait_and_throw();

  const auto t0 = event.template get_profiling_info<sycl::info::event_profiling::command_start>();
  const auto t1 = event.template get_profiling_info<sycl::info::event_profiling::command_end>();
  const double dt = (t1-t0)*1E-9;
  ((ISPCCamera*)&camera)->render_time = dt;
  
#else
  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,pixels,width,height,time,camera,numTilesX,numTilesY);
  }); 
#endif
}

/* 
 * Called by the C++ code to render. 
 */
extern "C" void device_render(int* pixels,
                          const unsigned int width,
                          const unsigned int height,
                          const float time,
                          const ISPCCamera& camera)
{
  if (g_data.g_accu_width != width || g_data.g_accu_height != height) {
    alignedUSMFree(g_data.g_accu);
    g_data.g_accu = (Vec3ff*) alignedUSMMalloc((width*height)*sizeof(Vec3ff),16,EMBREE_USM_SHARED_DEVICE_READ_WRITE);
    g_data.g_accu_width = width;
    g_data.g_accu_height = height;
    for (unsigned int i=0; i<width*height; i++)
      g_data.g_accu[i] = Vec3ff(0.0f);
  }
  
  bool camera_changed = g_changed; 
  g_changed = false;
  camera_changed |= ne(g_data.g_accu_vx,camera.xfm.l.vx); g_data.g_accu_vx = camera.xfm.l.vx;
  camera_changed |= ne(g_data.g_accu_vy,camera.xfm.l.vy); g_data.g_accu_vy = camera.xfm.l.vy;
  camera_changed |= ne(g_data.g_accu_vz,camera.xfm.l.vz); g_data.g_accu_vz = camera.xfm.l.vz;
  camera_changed |= ne(g_data.g_accu_p, camera.xfm.p);    g_data.g_accu_p  = camera.xfm.p;

  if (camera_changed)
  {
    g_data.g_accu_count=0;
    for (unsigned int i=0; i<width*height; i++)
      g_data.g_accu[i] = Vec3ff(0.0f);
  }
  else
    g_data.g_accu_count++;
}

/*
 * Called by the C++ code for cleanup.
 */
extern "C" void device_cleanup ()
{
  TutorialData_Destructor(&g_data);
}

/*
 * This must be here for the linker to find, but we will not use it.
 */
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) { }

} // namespace embree