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
|
/*
* 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.
*/
/* This is a template BVH traversal function for subsurface scattering, 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
*
*/
#if BVH_FEATURE(BVH_HAIR)
# define NODE_INTERSECT qbvh_node_intersect
#else
# define NODE_INTERSECT qbvh_aligned_node_intersect
#endif
ccl_device void BVH_FUNCTION_FULL_NAME(QBVH)(KernelGlobals *kg,
const Ray *ray,
SubsurfaceIntersection *ss_isect,
int subsurface_object,
uint *lcg_state,
int max_hits)
{
/* TODO(sergey):
* - Test if pushing distance on the stack helps (for non shadow rays).
* - Separate version for shadow rays.
* - Likely and unlikely for if() statements.
* - SSE for hair.
* - Test restrict attribute for pointers.
*/
/* Traversal stack in CUDA thread-local memory. */
QBVHStackItem traversal_stack[BVH_QSTACK_SIZE];
traversal_stack[0].addr = ENTRYPOINT_SENTINEL;
/* Traversal variables in registers. */
int stack_ptr = 0;
int node_addr = kernel_tex_fetch(__object_node, subsurface_object);
/* 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;
float isect_t = ray->t;
ss_isect->num_hits = 0;
const int object_flag = kernel_tex_fetch(__object_flag, subsurface_object);
if(!(object_flag & SD_TRANSFORM_APPLIED)) {
#if BVH_FEATURE(BVH_MOTION)
Transform ob_itfm;
bvh_instance_motion_push(kg,
subsurface_object,
ray,
&P,
&dir,
&idir,
&isect_t,
&ob_itfm);
#else
bvh_instance_push(kg, subsurface_object, ray, &P, &dir, &idir, &isect_t);
#endif
object = subsurface_object;
}
#ifndef __KERNEL_SSE41__
if(!isfinite(P.x)) {
return;
}
#endif
ssef tnear(0.0f), tfar(isect_t);
#if BVH_FEATURE(BVH_HAIR)
sse3f dir4(ssef(dir.x), ssef(dir.y), ssef(dir.z));
#endif
sse3f idir4(ssef(idir.x), ssef(idir.y), ssef(idir.z));
#ifdef __KERNEL_AVX2__
float3 P_idir = P*idir;
sse3f P_idir4(P_idir.x, P_idir.y, P_idir.z);
#endif
#if BVH_FEATURE(BVH_HAIR) || !defined(__KERNEL_AVX2__)
sse3f org4(ssef(P.x), ssef(P.y), ssef(P.z));
#endif
/* Offsets to select the side that becomes the lower or upper bound. */
int near_x, near_y, near_z;
int far_x, far_y, far_z;
if(idir.x >= 0.0f) { near_x = 0; far_x = 1; } else { near_x = 1; far_x = 0; }
if(idir.y >= 0.0f) { near_y = 2; far_y = 3; } else { near_y = 3; far_y = 2; }
if(idir.z >= 0.0f) { near_z = 4; far_z = 5; } else { near_z = 5; far_z = 4; }
IsectPrecalc isect_precalc;
triangle_intersect_precalc(dir, &isect_precalc);
/* Traversal loop. */
do {
do {
/* Traverse internal nodes. */
while(node_addr >= 0 && node_addr != ENTRYPOINT_SENTINEL) {
ssef dist;
int child_mask = NODE_INTERSECT(kg,
tnear,
tfar,
#ifdef __KERNEL_AVX2__
P_idir4,
#endif
#if BVH_FEATURE(BVH_HAIR) || !defined(__KERNEL_AVX2__)
org4,
#endif
#if BVH_FEATURE(BVH_HAIR)
dir4,
#endif
idir4,
near_x, near_y, near_z,
far_x, far_y, far_z,
node_addr,
&dist);
if(child_mask != 0) {
float4 inodes = kernel_tex_fetch(__bvh_nodes, node_addr+0);
float4 cnodes;
#if BVH_FEATURE(BVH_HAIR)
if(__float_as_uint(inodes.x) & PATH_RAY_NODE_UNALIGNED) {
cnodes = kernel_tex_fetch(__bvh_nodes, node_addr+13);
}
else
#endif
{
cnodes = kernel_tex_fetch(__bvh_nodes, node_addr+7);
}
/* One child is hit, continue with that child. */
int r = __bscf(child_mask);
if(child_mask == 0) {
node_addr = __float_as_int(cnodes[r]);
continue;
}
/* Two children are hit, push far child, and continue with
* closer child.
*/
int c0 = __float_as_int(cnodes[r]);
float d0 = ((float*)&dist)[r];
r = __bscf(child_mask);
int c1 = __float_as_int(cnodes[r]);
float d1 = ((float*)&dist)[r];
if(child_mask == 0) {
if(d1 < d0) {
node_addr = c1;
++stack_ptr;
kernel_assert(stack_ptr < BVH_QSTACK_SIZE);
traversal_stack[stack_ptr].addr = c0;
traversal_stack[stack_ptr].dist = d0;
continue;
}
else {
node_addr = c0;
++stack_ptr;
kernel_assert(stack_ptr < BVH_QSTACK_SIZE);
traversal_stack[stack_ptr].addr = c1;
traversal_stack[stack_ptr].dist = d1;
continue;
}
}
/* Here starts the slow path for 3 or 4 hit children. We push
* all nodes onto the stack to sort them there.
*/
++stack_ptr;
kernel_assert(stack_ptr < BVH_QSTACK_SIZE);
traversal_stack[stack_ptr].addr = c1;
traversal_stack[stack_ptr].dist = d1;
++stack_ptr;
kernel_assert(stack_ptr < BVH_QSTACK_SIZE);
traversal_stack[stack_ptr].addr = c0;
traversal_stack[stack_ptr].dist = d0;
/* Three children are hit, push all onto stack and sort 3
* stack items, continue with closest child.
*/
r = __bscf(child_mask);
int c2 = __float_as_int(cnodes[r]);
float d2 = ((float*)&dist)[r];
if(child_mask == 0) {
++stack_ptr;
kernel_assert(stack_ptr < BVH_QSTACK_SIZE);
traversal_stack[stack_ptr].addr = c2;
traversal_stack[stack_ptr].dist = d2;
qbvh_stack_sort(&traversal_stack[stack_ptr],
&traversal_stack[stack_ptr - 1],
&traversal_stack[stack_ptr - 2]);
node_addr = traversal_stack[stack_ptr].addr;
--stack_ptr;
continue;
}
/* Four children are hit, push all onto stack and sort 4
* stack items, continue with closest child.
*/
r = __bscf(child_mask);
int c3 = __float_as_int(cnodes[r]);
float d3 = ((float*)&dist)[r];
++stack_ptr;
kernel_assert(stack_ptr < BVH_QSTACK_SIZE);
traversal_stack[stack_ptr].addr = c3;
traversal_stack[stack_ptr].dist = d3;
++stack_ptr;
kernel_assert(stack_ptr < BVH_QSTACK_SIZE);
traversal_stack[stack_ptr].addr = c2;
traversal_stack[stack_ptr].dist = d2;
qbvh_stack_sort(&traversal_stack[stack_ptr],
&traversal_stack[stack_ptr - 1],
&traversal_stack[stack_ptr - 2],
&traversal_stack[stack_ptr - 3]);
}
node_addr = traversal_stack[stack_ptr].addr;
--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);
int prim_addr2 = __float_as_int(leaf.y);
const uint type = __float_as_int(leaf.w);
/* Pop. */
node_addr = traversal_stack[stack_ptr].addr;
--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_tex_fetch(__prim_type, prim_addr) == type);
triangle_intersect_subsurface(kg,
&isect_precalc,
ss_isect,
P,
object,
prim_addr,
isect_t,
lcg_state,
max_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_tex_fetch(__prim_type, prim_addr) == type);
motion_triangle_intersect_subsurface(kg,
ss_isect,
P,
dir,
ray->time,
object,
prim_addr,
isect_t,
lcg_state,
max_hits);
}
break;
}
#endif
default:
break;
}
}
} while(node_addr != ENTRYPOINT_SENTINEL);
} while(node_addr != ENTRYPOINT_SENTINEL);
}
#undef NODE_INTERSECT
|