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
|
/*
* Adapted from code Copyright 2009-2010 NVIDIA Corporation,
* and code copyright 2009-2012 Intel Corporation
*
* Modifications Copyright 2011-2013, 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.
*/
#if BVH_FEATURE(BVH_HAIR)
# define NODE_INTERSECT bvh_node_intersect
#else
# define NODE_INTERSECT bvh_aligned_node_intersect
#endif
/* This is a template BVH traversal function, where various features can be
* enabled/disabled. This way we can compile optimized versions for each case
* without new features slowing things down.
*
* BVH_HAIR: hair curve rendering
* BVH_POINTCLOUD: point cloud rendering
* BVH_MOTION: motion blur rendering
*/
#ifndef __KERNEL_GPU__
ccl_device
#else
ccl_device_inline
#endif
bool BVH_FUNCTION_FULL_NAME(BVH)(KernelGlobals kg,
ccl_private const Ray *ray,
IntegratorShadowState state,
const uint visibility,
const uint max_hits,
ccl_private uint *num_recorded_hits,
ccl_private float *throughput)
{
/* todo:
* - likely and unlikely for if() statements
* - test restrict attribute for pointers
*/
/* traversal stack in CUDA thread-local memory */
int traversal_stack[BVH_STACK_SIZE];
traversal_stack[0] = ENTRYPOINT_SENTINEL;
/* traversal variables in registers */
int stack_ptr = 0;
int node_addr = kernel_data.bvh.root;
/* ray parameters in registers */
float3 P = ray->P;
float3 dir = bvh_clamp_direction(ray->D);
float3 idir = bvh_inverse_direction(dir);
int object = OBJECT_NONE;
uint num_hits = 0;
#if BVH_FEATURE(BVH_MOTION)
Transform ob_itfm;
#endif
/* Max distance in world space. May be dynamically reduced when max number of
* recorded hits is exceeded and we no longer need to find hits beyond the max
* distance found. */
float t_max_world = ray->t;
/* Equal to t_max_world when traversing top level BVH, transformed into local
* space when entering instances. */
float t_max_current = t_max_world;
/* Conversion from world to local space for the current instance if any, 1.0
* otherwise. */
float t_world_to_instance = 1.0f;
*num_recorded_hits = 0;
*throughput = 1.0f;
/* traversal loop */
do {
do {
/* traverse internal nodes */
while (node_addr >= 0 && node_addr != ENTRYPOINT_SENTINEL) {
int node_addr_child1, traverse_mask;
float dist[2];
float4 cnodes = kernel_tex_fetch(__bvh_nodes, node_addr + 0);
traverse_mask = NODE_INTERSECT(kg,
P,
#if BVH_FEATURE(BVH_HAIR)
dir,
#endif
idir,
t_max_current,
node_addr,
visibility,
dist);
node_addr = __float_as_int(cnodes.z);
node_addr_child1 = __float_as_int(cnodes.w);
if (traverse_mask == 3) {
/* Both children were intersected, push the farther one. */
bool is_closest_child1 = (dist[1] < dist[0]);
if (is_closest_child1) {
int tmp = node_addr;
node_addr = node_addr_child1;
node_addr_child1 = tmp;
}
++stack_ptr;
kernel_assert(stack_ptr < BVH_STACK_SIZE);
traversal_stack[stack_ptr] = node_addr_child1;
}
else {
/* One child was intersected. */
if (traverse_mask == 2) {
node_addr = node_addr_child1;
}
else if (traverse_mask == 0) {
/* Neither child was intersected. */
node_addr = traversal_stack[stack_ptr];
--stack_ptr;
}
}
}
/* if node is leaf, fetch triangle list */
if (node_addr < 0) {
float4 leaf = kernel_tex_fetch(__bvh_leaf_nodes, (-node_addr - 1));
int prim_addr = __float_as_int(leaf.x);
if (prim_addr >= 0) {
const int prim_addr2 = __float_as_int(leaf.y);
const uint type = __float_as_int(leaf.w);
/* pop */
node_addr = traversal_stack[stack_ptr];
--stack_ptr;
/* primitive intersection */
while (prim_addr < prim_addr2) {
kernel_assert((kernel_tex_fetch(__prim_type, prim_addr) & PRIMITIVE_ALL) ==
(type & PRIMITIVE_ALL));
bool hit;
/* todo: specialized intersect functions which don't fill in
* isect unless needed and check SD_HAS_TRANSPARENT_SHADOW?
* might give a few % performance improvement */
Intersection isect ccl_optional_struct_init;
switch (type & PRIMITIVE_ALL) {
case PRIMITIVE_TRIANGLE: {
hit = triangle_intersect(
kg, &isect, P, dir, t_max_current, visibility, object, prim_addr);
break;
}
#if BVH_FEATURE(BVH_MOTION)
case PRIMITIVE_MOTION_TRIANGLE: {
hit = motion_triangle_intersect(
kg, &isect, P, dir, t_max_current, ray->time, visibility, object, prim_addr);
break;
}
#endif
#if BVH_FEATURE(BVH_HAIR)
case PRIMITIVE_CURVE_THICK:
case PRIMITIVE_MOTION_CURVE_THICK:
case PRIMITIVE_CURVE_RIBBON:
case PRIMITIVE_MOTION_CURVE_RIBBON: {
if ((type & PRIMITIVE_ALL_MOTION) && kernel_data.bvh.use_bvh_steps) {
const float2 prim_time = kernel_tex_fetch(__prim_time, prim_addr);
if (ray->time < prim_time.x || ray->time > prim_time.y) {
hit = false;
break;
}
}
const int curve_object = (object == OBJECT_NONE) ?
kernel_tex_fetch(__prim_object, prim_addr) :
object;
const int curve_type = kernel_tex_fetch(__prim_type, prim_addr);
const int curve_prim = kernel_tex_fetch(__prim_index, prim_addr);
hit = curve_intersect(kg,
&isect,
P,
dir,
t_max_current,
curve_object,
curve_prim,
ray->time,
curve_type);
break;
}
#endif
#if BVH_FEATURE(BVH_POINTCLOUD)
case PRIMITIVE_POINT:
case PRIMITIVE_MOTION_POINT: {
if ((type & PRIMITIVE_ALL_MOTION) && kernel_data.bvh.use_bvh_steps) {
const float2 prim_time = kernel_tex_fetch(__prim_time, prim_addr);
if (ray->time < prim_time.x || ray->time > prim_time.y) {
hit = false;
break;
}
}
const int point_object = (object == OBJECT_NONE) ?
kernel_tex_fetch(__prim_object, prim_addr) :
object;
const int point_prim = kernel_tex_fetch(__prim_index, prim_addr);
const int point_type = kernel_tex_fetch(__prim_type, prim_addr);
hit = point_intersect(kg,
&isect,
P,
dir,
t_max_current,
point_object,
point_prim,
ray->time,
point_type);
break;
}
#endif /* BVH_FEATURE(BVH_POINTCLOUD) */
default: {
hit = false;
break;
}
}
/* shadow ray early termination */
if (hit) {
/* Convert intersection distance to world space. */
isect.t /= t_world_to_instance;
/* detect if this surface has a shader with transparent shadows */
/* todo: optimize so primitive visibility flag indicates if
* the primitive has a transparent shadow shader? */
const int flags = intersection_get_shader_flags(kg, isect.prim, isect.type);
if (!(flags & SD_HAS_TRANSPARENT_SHADOW) || num_hits >= max_hits) {
/* If no transparent shadows, all light is blocked and we can
* stop immediately. */
return true;
}
num_hits++;
bool record_intersection = true;
/* Always use baked shadow transparency for curves. */
if (isect.type & PRIMITIVE_ALL_CURVE) {
*throughput *= intersection_curve_shadow_transparency(
kg, isect.object, isect.prim, isect.u);
if (*throughput < CURVE_SHADOW_TRANSPARENCY_CUTOFF) {
return true;
}
else {
record_intersection = false;
}
}
if (record_intersection) {
/* Increase the number of hits, possibly beyond max_hits, we will
* simply not record those and only keep the max_hits closest. */
uint record_index = (*num_recorded_hits)++;
const uint max_record_hits = min(max_hits, INTEGRATOR_SHADOW_ISECT_SIZE);
if (record_index >= max_record_hits - 1) {
/* If maximum number of hits reached, find the intersection with
* the largest distance to potentially replace when another hit
* is found. */
const int num_recorded_hits = min(max_record_hits, record_index);
float max_recorded_t = INTEGRATOR_STATE_ARRAY(state, shadow_isect, 0, t);
int max_recorded_hit = 0;
for (int i = 1; i < num_recorded_hits; i++) {
const float isect_t = INTEGRATOR_STATE_ARRAY(state, shadow_isect, i, t);
if (isect_t > max_recorded_t) {
max_recorded_t = isect_t;
max_recorded_hit = i;
}
}
if (record_index >= max_record_hits) {
record_index = max_recorded_hit;
}
/* Limit the ray distance and stop counting hits beyond this. */
t_max_world = max(max_recorded_t, isect.t);
t_max_current = t_max_world * t_world_to_instance;
}
integrator_state_write_shadow_isect(state, &isect, record_index);
}
}
prim_addr++;
}
}
else {
/* instance push */
object = kernel_tex_fetch(__prim_object, -prim_addr - 1);
#if BVH_FEATURE(BVH_MOTION)
t_world_to_instance = bvh_instance_motion_push(
kg, object, ray, &P, &dir, &idir, &ob_itfm);
#else
t_world_to_instance = bvh_instance_push(kg, object, ray, &P, &dir, &idir);
#endif
/* Convert intersection to object space. */
t_max_current *= t_world_to_instance;
++stack_ptr;
kernel_assert(stack_ptr < BVH_STACK_SIZE);
traversal_stack[stack_ptr] = ENTRYPOINT_SENTINEL;
node_addr = kernel_tex_fetch(__object_node, object);
}
}
} while (node_addr != ENTRYPOINT_SENTINEL);
if (stack_ptr >= 0) {
kernel_assert(object != OBJECT_NONE);
/* Instance pop. */
#if BVH_FEATURE(BVH_MOTION)
bvh_instance_motion_pop(kg, object, ray, &P, &dir, &idir, FLT_MAX, &ob_itfm);
#else
bvh_instance_pop(kg, object, ray, &P, &dir, &idir, FLT_MAX);
#endif
/* Restore world space ray length. */
t_max_current = t_max_world;
object = OBJECT_NONE;
t_world_to_instance = 1.0f;
node_addr = traversal_stack[stack_ptr];
--stack_ptr;
}
} while (node_addr != ENTRYPOINT_SENTINEL);
return false;
}
ccl_device_inline bool BVH_FUNCTION_NAME(KernelGlobals kg,
ccl_private const Ray *ray,
IntegratorShadowState state,
const uint visibility,
const uint max_hits,
ccl_private uint *num_recorded_hits,
ccl_private float *throughput)
{
return BVH_FUNCTION_FULL_NAME(BVH)(
kg, ray, state, visibility, max_hits, num_recorded_hits, throughput);
}
#undef BVH_FUNCTION_NAME
#undef BVH_FUNCTION_FEATURES
#undef NODE_INTERSECT
|