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
|
/*
* Adapted from code Copyright 2009-2010 NVIDIA Corporation
* Modifications Copyright 2011, Blender Foundation.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
/* Motion Triangle Primitive
*
* These are stored as regular triangles, plus extra positions and normals at
* times other than the frame center. Computing the triangle vertex positions
* or normals at a given ray time is a matter of interpolation of the two steps
* between which the ray time lies.
*
* The extra positions and normals are stored as ATTR_STD_MOTION_VERTEX_POSITION
* and ATTR_STD_MOTION_VERTEX_NORMAL mesh attributes.
*/
CCL_NAMESPACE_BEGIN
/* Time interpolation of vertex positions and normals */
ccl_device_inline int find_attribute_motion(KernelGlobals *kg, int object, uint id, AttributeElement *elem)
{
/* todo: find a better (faster) solution for this, maybe store offset per object */
uint attr_offset = object*kernel_data.bvh.attributes_map_stride;
uint4 attr_map = kernel_tex_fetch(__attributes_map, attr_offset);
while(attr_map.x != id) {
attr_offset += ATTR_PRIM_TYPES;
attr_map = kernel_tex_fetch(__attributes_map, attr_offset);
}
*elem = (AttributeElement)attr_map.y;
/* return result */
return (attr_map.y == ATTR_ELEMENT_NONE) ? (int)ATTR_STD_NOT_FOUND : (int)attr_map.z;
}
ccl_device_inline void motion_triangle_verts_for_step(KernelGlobals *kg, uint4 tri_vindex, int offset, int numverts, int numsteps, int step, float3 verts[3])
{
if(step == numsteps) {
/* center step: regular vertex location */
verts[0] = float4_to_float3(kernel_tex_fetch(__prim_tri_verts, tri_vindex.w+0));
verts[1] = float4_to_float3(kernel_tex_fetch(__prim_tri_verts, tri_vindex.w+1));
verts[2] = float4_to_float3(kernel_tex_fetch(__prim_tri_verts, tri_vindex.w+2));
}
else {
/* center step not store in this array */
if(step > numsteps)
step--;
offset += step*numverts;
verts[0] = float4_to_float3(kernel_tex_fetch(__attributes_float3, offset + tri_vindex.x));
verts[1] = float4_to_float3(kernel_tex_fetch(__attributes_float3, offset + tri_vindex.y));
verts[2] = float4_to_float3(kernel_tex_fetch(__attributes_float3, offset + tri_vindex.z));
}
}
ccl_device_inline void motion_triangle_normals_for_step(KernelGlobals *kg, uint4 tri_vindex, int offset, int numverts, int numsteps, int step, float3 normals[3])
{
if(step == numsteps) {
/* center step: regular vertex location */
normals[0] = float4_to_float3(kernel_tex_fetch(__tri_vnormal, tri_vindex.x));
normals[1] = float4_to_float3(kernel_tex_fetch(__tri_vnormal, tri_vindex.y));
normals[2] = float4_to_float3(kernel_tex_fetch(__tri_vnormal, tri_vindex.z));
}
else {
/* center step not stored in this array */
if(step > numsteps)
step--;
offset += step*numverts;
normals[0] = float4_to_float3(kernel_tex_fetch(__attributes_float3, offset + tri_vindex.x));
normals[1] = float4_to_float3(kernel_tex_fetch(__attributes_float3, offset + tri_vindex.y));
normals[2] = float4_to_float3(kernel_tex_fetch(__attributes_float3, offset + tri_vindex.z));
}
}
ccl_device_inline void motion_triangle_vertices(KernelGlobals *kg, int object, int prim, float time, float3 verts[3])
{
/* get motion info */
int numsteps, numverts;
object_motion_info(kg, object, &numsteps, &numverts, NULL);
/* figure out which steps we need to fetch and their interpolation factor */
int maxstep = numsteps*2;
int step = min((int)(time*maxstep), maxstep-1);
float t = time*maxstep - step;
/* find attribute */
AttributeElement elem;
int offset = find_attribute_motion(kg, object, ATTR_STD_MOTION_VERTEX_POSITION, &elem);
kernel_assert(offset != ATTR_STD_NOT_FOUND);
/* fetch vertex coordinates */
float3 next_verts[3];
uint4 tri_vindex = kernel_tex_fetch(__tri_vindex, prim);
motion_triangle_verts_for_step(kg, tri_vindex, offset, numverts, numsteps, step, verts);
motion_triangle_verts_for_step(kg, tri_vindex, offset, numverts, numsteps, step+1, next_verts);
/* interpolate between steps */
verts[0] = (1.0f - t)*verts[0] + t*next_verts[0];
verts[1] = (1.0f - t)*verts[1] + t*next_verts[1];
verts[2] = (1.0f - t)*verts[2] + t*next_verts[2];
}
/* Refine triangle intersection to more precise hit point. For rays that travel
* far the precision is often not so good, this reintersects the primitive from
* a closer distance. */
ccl_device_inline float3 motion_triangle_refine(KernelGlobals *kg, ShaderData *sd, const Intersection *isect, const Ray *ray, float3 verts[3])
{
float3 P = ray->P;
float3 D = ray->D;
float t = isect->t;
#ifdef __INTERSECTION_REFINE__
if(isect->object != OBJECT_NONE) {
if(UNLIKELY(t == 0.0f)) {
return P;
}
# ifdef __OBJECT_MOTION__
Transform tfm = ccl_fetch(sd, ob_itfm);
# else
Transform tfm = object_fetch_transform(kg, isect->object, OBJECT_INVERSE_TRANSFORM);
# endif
P = transform_point(&tfm, P);
D = transform_direction(&tfm, D*t);
D = normalize_len(D, &t);
}
P = P + D*t;
/* compute refined intersection distance */
const float3 e1 = verts[0] - verts[2];
const float3 e2 = verts[1] - verts[2];
const float3 s1 = cross(D, e2);
const float invdivisor = 1.0f/dot(s1, e1);
const float3 d = P - verts[2];
const float3 s2 = cross(d, e1);
float rt = dot(e2, s2)*invdivisor;
/* compute refined position */
P = P + D*rt;
if(isect->object != OBJECT_NONE) {
# ifdef __OBJECT_MOTION__
Transform tfm = ccl_fetch(sd, ob_tfm);
# else
Transform tfm = object_fetch_transform(kg, isect->object, OBJECT_TRANSFORM);
# endif
P = transform_point(&tfm, P);
}
return P;
#else
return P + D*t;
#endif
}
/* Same as above, except that isect->t is assumed to be in object space for instancing */
#ifdef __SUBSURFACE__
# if defined(__KERNEL_CUDA__) && (defined(i386) || defined(_M_IX86))
ccl_device_noinline
# else
ccl_device_inline
# endif
float3 motion_triangle_refine_subsurface(KernelGlobals *kg, ShaderData *sd, const Intersection *isect, const Ray *ray, float3 verts[3])
{
float3 P = ray->P;
float3 D = ray->D;
float t = isect->t;
# ifdef __INTERSECTION_REFINE__
if(isect->object != OBJECT_NONE) {
# ifdef __OBJECT_MOTION__
Transform tfm = ccl_fetch(sd, ob_itfm);
# else
Transform tfm = object_fetch_transform(kg, isect->object, OBJECT_INVERSE_TRANSFORM);
# endif
P = transform_point(&tfm, P);
D = transform_direction(&tfm, D);
D = normalize(D);
}
P = P + D*t;
/* compute refined intersection distance */
const float3 e1 = verts[0] - verts[2];
const float3 e2 = verts[1] - verts[2];
const float3 s1 = cross(D, e2);
const float invdivisor = 1.0f/dot(s1, e1);
const float3 d = P - verts[2];
const float3 s2 = cross(d, e1);
float rt = dot(e2, s2)*invdivisor;
P = P + D*rt;
if(isect->object != OBJECT_NONE) {
# ifdef __OBJECT_MOTION__
Transform tfm = ccl_fetch(sd, ob_tfm);
# else
Transform tfm = object_fetch_transform(kg, isect->object, OBJECT_TRANSFORM);
# endif
P = transform_point(&tfm, P);
}
return P;
# else
return P + D*t;
# endif
}
#endif
/* Setup of motion triangle specific parts of ShaderData, moved into this one
* function to more easily share computation of interpolated positions and
* normals */
/* return 3 triangle vertex normals */
ccl_device_noinline void motion_triangle_shader_setup(KernelGlobals *kg, ShaderData *sd, const Intersection *isect, const Ray *ray, bool subsurface)
{
/* get shader */
ccl_fetch(sd, shader) = kernel_tex_fetch(__tri_shader, ccl_fetch(sd, prim));
/* get motion info */
int numsteps, numverts;
object_motion_info(kg, ccl_fetch(sd, object), &numsteps, &numverts, NULL);
/* figure out which steps we need to fetch and their interpolation factor */
int maxstep = numsteps*2;
int step = min((int)(ccl_fetch(sd, time)*maxstep), maxstep-1);
float t = ccl_fetch(sd, time)*maxstep - step;
/* find attribute */
AttributeElement elem;
int offset = find_attribute_motion(kg, ccl_fetch(sd, object), ATTR_STD_MOTION_VERTEX_POSITION, &elem);
kernel_assert(offset != ATTR_STD_NOT_FOUND);
/* fetch vertex coordinates */
float3 verts[3], next_verts[3];
uint4 tri_vindex = kernel_tex_fetch(__tri_vindex, ccl_fetch(sd, prim));
motion_triangle_verts_for_step(kg, tri_vindex, offset, numverts, numsteps, step, verts);
motion_triangle_verts_for_step(kg, tri_vindex, offset, numverts, numsteps, step+1, next_verts);
/* interpolate between steps */
verts[0] = (1.0f - t)*verts[0] + t*next_verts[0];
verts[1] = (1.0f - t)*verts[1] + t*next_verts[1];
verts[2] = (1.0f - t)*verts[2] + t*next_verts[2];
/* compute refined position */
#ifdef __SUBSURFACE__
if(!subsurface)
#endif
ccl_fetch(sd, P) = motion_triangle_refine(kg, sd, isect, ray, verts);
#ifdef __SUBSURFACE__
else
ccl_fetch(sd, P) = motion_triangle_refine_subsurface(kg, sd, isect, ray, verts);
#endif
/* compute face normal */
float3 Ng;
if(ccl_fetch(sd, flag) & SD_NEGATIVE_SCALE_APPLIED)
Ng = normalize(cross(verts[2] - verts[0], verts[1] - verts[0]));
else
Ng = normalize(cross(verts[1] - verts[0], verts[2] - verts[0]));
ccl_fetch(sd, Ng) = Ng;
ccl_fetch(sd, N) = Ng;
/* compute derivatives of P w.r.t. uv */
#ifdef __DPDU__
ccl_fetch(sd, dPdu) = (verts[0] - verts[2]);
ccl_fetch(sd, dPdv) = (verts[1] - verts[2]);
#endif
/* compute smooth normal */
if(ccl_fetch(sd, shader) & SHADER_SMOOTH_NORMAL) {
/* find attribute */
AttributeElement elem;
int offset = find_attribute_motion(kg, ccl_fetch(sd, object), ATTR_STD_MOTION_VERTEX_NORMAL, &elem);
kernel_assert(offset != ATTR_STD_NOT_FOUND);
/* fetch vertex coordinates */
float3 normals[3], next_normals[3];
motion_triangle_normals_for_step(kg, tri_vindex, offset, numverts, numsteps, step, normals);
motion_triangle_normals_for_step(kg, tri_vindex, offset, numverts, numsteps, step+1, next_normals);
/* interpolate between steps */
normals[0] = (1.0f - t)*normals[0] + t*next_normals[0];
normals[1] = (1.0f - t)*normals[1] + t*next_normals[1];
normals[2] = (1.0f - t)*normals[2] + t*next_normals[2];
/* interpolate between vertices */
float u = ccl_fetch(sd, u);
float v = ccl_fetch(sd, v);
float w = 1.0f - u - v;
ccl_fetch(sd, N) = (u*normals[0] + v*normals[1] + w*normals[2]);
}
}
/* Ray intersection. We simply compute the vertex positions at the given ray
* time and do a ray intersection with the resulting triangle */
ccl_device_inline bool motion_triangle_intersect(KernelGlobals *kg, Intersection *isect,
float3 P, float3 dir, float time, uint visibility, int object, int triAddr)
{
/* primitive index for vertex location lookup */
int prim = kernel_tex_fetch(__prim_index, triAddr);
int fobject = (object == OBJECT_NONE)? kernel_tex_fetch(__prim_object, triAddr): object;
/* get vertex locations for intersection */
float3 verts[3];
motion_triangle_vertices(kg, fobject, prim, time, verts);
/* ray-triangle intersection, unoptimized */
float t, u, v;
if(ray_triangle_intersect_uv(P, dir, isect->t, verts[2], verts[0], verts[1], &u, &v, &t)) {
#ifdef __VISIBILITY_FLAG__
/* visibility flag test. we do it here under the assumption
* that most triangles are culled by node flags */
if(kernel_tex_fetch(__prim_visibility, triAddr) & visibility)
#endif
{
isect->t = t;
isect->u = u;
isect->v = v;
isect->prim = triAddr;
isect->object = object;
isect->type = PRIMITIVE_MOTION_TRIANGLE;
return true;
}
}
return false;
}
/* Special ray intersection routines for subsurface scattering. In that case we
* only want to intersect with primitives in the same object, and if case of
* multiple hits we pick a single random primitive as the intersection point. */
#ifdef __SUBSURFACE__
ccl_device_inline void motion_triangle_intersect_subsurface(
KernelGlobals *kg,
SubsurfaceIntersection *ss_isect,
float3 P,
float3 dir,
float time,
int object,
int triAddr,
float tmax,
uint *lcg_state,
int max_hits)
{
/* primitive index for vertex location lookup */
int prim = kernel_tex_fetch(__prim_index, triAddr);
int fobject = (object == OBJECT_NONE)? kernel_tex_fetch(__prim_object, triAddr): object;
/* get vertex locations for intersection */
float3 verts[3];
motion_triangle_vertices(kg, fobject, prim, time, verts);
/* ray-triangle intersection, unoptimized */
float t, u, v;
if(ray_triangle_intersect_uv(P, dir, tmax, verts[2], verts[0], verts[1], &u, &v, &t)) {
for(int i = min(max_hits, ss_isect->num_hits) - 1; i >= 0; --i) {
if(ss_isect->hits[i].t == t) {
return;
}
}
ss_isect->num_hits++;
int hit;
if(ss_isect->num_hits <= max_hits) {
hit = ss_isect->num_hits - 1;
}
else {
/* reservoir sampling: if we are at the maximum number of
* hits, randomly replace element or skip it */
hit = lcg_step_uint(lcg_state) % ss_isect->num_hits;
if(hit >= max_hits)
return;
}
/* record intersection */
Intersection *isect = &ss_isect->hits[hit];
isect->t = t;
isect->u = u;
isect->v = v;
isect->prim = triAddr;
isect->object = object;
isect->type = PRIMITIVE_MOTION_TRIANGLE;
/* Record geometric normal. */
ss_isect->Ng[hit] = normalize(cross(verts[1] - verts[0],
verts[2] - verts[0]));
}
}
#endif
CCL_NAMESPACE_END
|