6ea54fe9ff
The intention of this commit it to address issues mentioned in the
reports T43865,T50164 and T50452.
The code is based on Embree code with some extra vectorization
to speed up single ray to single triangle intersection.
Unfortunately, such a fix is not coming for free. There is some
slowdown for AVX2 processors, mainly due to different vectorization
code, which caused different number of instructions to be executed
and different instructions-per-cycle counters. But on another hand
this commit makes pre-AVX2 platforms such as AVX and SSE4.1 a bit
faster. The prerformance goes as following:
2.78c AVX2 2.78c AVX Patch AVX2 Patch AVX
BMW 05:21.09 06:05.34 05:32.97 (+3.5%) 05:34.97 (-8.5%)
Classroom 16:55.36 18:24.51 17:10.41 (+1.4%) 17:15.87 (-6.3%)
Fishy Cat 08:08.49 08:36.26 08:09.19 (+0.2%) 08:12.25 (-4.7%
Koro 11:22.54 11:45.24 11:13.25 (-1.5%) 11:43.81 (-0.3%)
Barcelone 14:18.32 16:09.46 14:15.20 (-0.4%) 14:25.15 (-10.8%)
On GPU the performance is about 1.5-2% slower in my tests on GTX1080
but afraid we can't do much as a part of this chaneg here and
consider it a price to pay for more proper intersection check.
Made in collaboration with Maxym Dmytrychenko, big thanks to him!
Reviewers: brecht, juicyfruit, lukasstockner97, dingto
Differential Revision: https://developer.blender.org/D1574
268 lines
8.5 KiB
C
268 lines
8.5 KiB
C
/*
|
|
* 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.
|
|
*/
|
|
|
|
#ifdef __QBVH__
|
|
# include "qbvh_subsurface.h"
|
|
#endif
|
|
|
|
#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 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
|
|
*
|
|
*/
|
|
|
|
#ifndef __KERNEL_GPU__
|
|
ccl_device
|
|
#else
|
|
ccl_device_inline
|
|
#endif
|
|
void BVH_FUNCTION_FULL_NAME(BVH)(KernelGlobals *kg,
|
|
const Ray *ray,
|
|
SubsurfaceIntersection *ss_isect,
|
|
int subsurface_object,
|
|
uint *lcg_state,
|
|
int max_hits)
|
|
{
|
|
/* 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_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_OBJECT_TRANSFORM_APPLIED)) {
|
|
#if BVH_FEATURE(BVH_MOTION)
|
|
Transform ob_itfm;
|
|
isect_t = bvh_instance_motion_push(kg,
|
|
subsurface_object,
|
|
ray,
|
|
&P,
|
|
&dir,
|
|
&idir,
|
|
isect_t,
|
|
&ob_itfm);
|
|
#else
|
|
isect_t = bvh_instance_push(kg, subsurface_object, ray, &P, &dir, &idir, isect_t);
|
|
#endif
|
|
object = subsurface_object;
|
|
}
|
|
|
|
#if defined(__KERNEL_SSE2__)
|
|
const shuffle_swap_t shuf_identity = shuffle_swap_identity();
|
|
const shuffle_swap_t shuf_swap = shuffle_swap_swap();
|
|
|
|
const ssef pn = cast(ssei(0, 0, 0x80000000, 0x80000000));
|
|
ssef Psplat[3], idirsplat[3];
|
|
# if BVH_FEATURE(BVH_HAIR)
|
|
ssef tnear(0.0f), tfar(isect_t);
|
|
# endif
|
|
shuffle_swap_t shufflexyz[3];
|
|
|
|
Psplat[0] = ssef(P.x);
|
|
Psplat[1] = ssef(P.y);
|
|
Psplat[2] = ssef(P.z);
|
|
|
|
ssef tsplat(0.0f, 0.0f, -isect_t, -isect_t);
|
|
|
|
gen_idirsplat_swap(pn, shuf_identity, shuf_swap, idir, idirsplat, shufflexyz);
|
|
#endif
|
|
|
|
/* 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);
|
|
|
|
#if !defined(__KERNEL_SSE2__)
|
|
traverse_mask = NODE_INTERSECT(kg,
|
|
P,
|
|
# if BVH_FEATURE(BVH_HAIR)
|
|
dir,
|
|
# endif
|
|
idir,
|
|
isect_t,
|
|
node_addr,
|
|
PATH_RAY_ALL_VISIBILITY,
|
|
dist);
|
|
#else // __KERNEL_SSE2__
|
|
traverse_mask = NODE_INTERSECT(kg,
|
|
P,
|
|
dir,
|
|
# if BVH_FEATURE(BVH_HAIR)
|
|
tnear,
|
|
tfar,
|
|
# endif
|
|
tsplat,
|
|
Psplat,
|
|
idirsplat,
|
|
shufflexyz,
|
|
node_addr,
|
|
PATH_RAY_ALL_VISIBILITY,
|
|
dist);
|
|
#endif // __KERNEL_SSE2__
|
|
|
|
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);
|
|
|
|
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 */
|
|
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,
|
|
ss_isect,
|
|
P,
|
|
dir,
|
|
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);
|
|
}
|
|
|
|
ccl_device_inline void BVH_FUNCTION_NAME(KernelGlobals *kg,
|
|
const Ray *ray,
|
|
SubsurfaceIntersection *ss_isect,
|
|
int subsurface_object,
|
|
uint *lcg_state,
|
|
int max_hits)
|
|
{
|
|
#ifdef __QBVH__
|
|
if(kernel_data.bvh.use_qbvh) {
|
|
return BVH_FUNCTION_FULL_NAME(QBVH)(kg,
|
|
ray,
|
|
ss_isect,
|
|
subsurface_object,
|
|
lcg_state,
|
|
max_hits);
|
|
}
|
|
else
|
|
#endif
|
|
{
|
|
kernel_assert(kernel_data.bvh.use_qbvh == false);
|
|
return BVH_FUNCTION_FULL_NAME(BVH)(kg,
|
|
ray,
|
|
ss_isect,
|
|
subsurface_object,
|
|
lcg_state,
|
|
max_hits);
|
|
}
|
|
}
|
|
|
|
#undef BVH_FUNCTION_NAME
|
|
#undef BVH_FUNCTION_FEATURES
|
|
#undef NODE_INTERSECT
|