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
|
/* SPDX-License-Identifier: Apache-2.0
* Adapted from code Copyright 2009-2010 NVIDIA Corporation,
* and code copyright 2009-2012 Intel Corporation
*
* Modifications Copyright 2011-2022 Blender Foundation. */
#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 for volumes, where
* various features can be enabled/disabled. This way we can compile optimized
* versions for each case without new features slowing things down.
*
* BVH_MOTION: motion blur rendering
*/
#ifndef __KERNEL_GPU__
ccl_device
#else
ccl_device_inline
#endif
uint BVH_FUNCTION_FULL_NAME(BVH)(KernelGlobals kg,
ccl_private const Ray *ray,
Intersection *isect_array,
const uint max_hits,
const uint visibility)
{
/* todo:
* - test if pushing distance on the stack helps (for non shadow rays)
* - separate version for shadow rays
* - 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);
const float tmin = ray->tmin;
int object = OBJECT_NONE;
float isect_t = ray->tmax;
int num_hits_in_instance = 0;
uint num_hits = 0;
isect_array->t = ray->tmax;
/* 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_data_fetch(bvh_nodes, node_addr + 0);
traverse_mask = NODE_INTERSECT(kg,
P,
#if BVH_FEATURE(BVH_HAIR)
dir,
#endif
idir,
tmin,
isect_t,
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_data_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);
bool hit;
/* pop */
node_addr = traversal_stack[stack_ptr];
--stack_ptr;
/* primitive intersection */
switch (type & PRIMITIVE_ALL) {
case PRIMITIVE_TRIANGLE: {
/* intersect ray against primitive */
for (; prim_addr < prim_addr2; prim_addr++) {
kernel_assert(kernel_data_fetch(prim_type, prim_addr) == type);
/* only primitives from volume object */
const int prim_object = (object == OBJECT_NONE) ?
kernel_data_fetch(prim_object, prim_addr) :
object;
const int prim = kernel_data_fetch(prim_index, prim_addr);
if (intersection_skip_self(ray->self, prim_object, prim)) {
continue;
}
int object_flag = kernel_data_fetch(object_flag, prim_object);
if ((object_flag & SD_OBJECT_HAS_VOLUME) == 0) {
continue;
}
hit = triangle_intersect(kg,
isect_array,
P,
dir,
tmin,
isect_t,
visibility,
prim_object,
prim,
prim_addr);
if (hit) {
/* Move on to next entry in intersections array. */
isect_array++;
num_hits++;
num_hits_in_instance++;
isect_array->t = isect_t;
if (num_hits == max_hits) {
return num_hits;
}
}
}
break;
}
#if BVH_FEATURE(BVH_MOTION)
case PRIMITIVE_MOTION_TRIANGLE: {
/* intersect ray against primitive */
for (; prim_addr < prim_addr2; prim_addr++) {
kernel_assert(kernel_data_fetch(prim_type, prim_addr) == type);
/* only primitives from volume object */
const int prim_object = (object == OBJECT_NONE) ?
kernel_data_fetch(prim_object, prim_addr) :
object;
const int prim = kernel_data_fetch(prim_index, prim_addr);
if (intersection_skip_self(ray->self, prim_object, prim)) {
continue;
}
int object_flag = kernel_data_fetch(object_flag, prim_object);
if ((object_flag & SD_OBJECT_HAS_VOLUME) == 0) {
continue;
}
hit = motion_triangle_intersect(kg,
isect_array,
P,
dir,
tmin,
isect_t,
ray->time,
visibility,
prim_object,
prim,
prim_addr);
if (hit) {
/* Move on to next entry in intersections array. */
isect_array++;
num_hits++;
num_hits_in_instance++;
isect_array->t = isect_t;
if (num_hits == max_hits) {
return num_hits;
}
}
}
break;
}
#endif /* BVH_MOTION */
default: {
break;
}
}
}
else {
/* instance push */
object = kernel_data_fetch(prim_object, -prim_addr - 1);
int object_flag = kernel_data_fetch(object_flag, object);
if (object_flag & SD_OBJECT_HAS_VOLUME) {
#if BVH_FEATURE(BVH_MOTION)
bvh_instance_motion_push(kg, object, ray, &P, &dir, &idir);
#else
bvh_instance_push(kg, object, ray, &P, &dir, &idir);
#endif
num_hits_in_instance = 0;
isect_array->t = isect_t;
++stack_ptr;
kernel_assert(stack_ptr < BVH_STACK_SIZE);
traversal_stack[stack_ptr] = ENTRYPOINT_SENTINEL;
node_addr = kernel_data_fetch(object_node, object);
}
else {
/* pop */
object = OBJECT_NONE;
node_addr = traversal_stack[stack_ptr];
--stack_ptr;
}
}
}
} 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);
#else
bvh_instance_pop(kg, object, ray, &P, &dir, &idir);
#endif
object = OBJECT_NONE;
node_addr = traversal_stack[stack_ptr];
--stack_ptr;
}
} while (node_addr != ENTRYPOINT_SENTINEL);
return num_hits;
}
ccl_device_inline uint BVH_FUNCTION_NAME(KernelGlobals kg,
ccl_private const Ray *ray,
Intersection *isect_array,
const uint max_hits,
const uint visibility)
{
return BVH_FUNCTION_FULL_NAME(BVH)(kg, ray, isect_array, max_hits, visibility);
}
#undef BVH_FUNCTION_NAME
#undef BVH_FUNCTION_FEATURES
#undef NODE_INTERSECT
|
