Cycles: Reduce memory usage by de-duplicating triangle storage
There are several internal changes for this: First idea is to make __tri_verts to behave similar to __tri_storage, meaning, __tri_verts array now contains all vertices of all triangles instead of just mesh vertices. This saves some lookup when reading triangle coordinates in functions like triangle_normal(). In order to make it efficient needed to store global triangle offset somewhere. So no __tri_vindex.w contains a global triangle index which can be used to read triangle vertices. Additionally, the order of vertices in that array is aligned with primitives from BVH. This is needed to keep cache as much coherent as possible for BVH traversal. This causes some extra tricks needed to fill the array in and deal with True Displacement but those trickery is fully required to prevent noticeable slowdown. Next idea was to use this __tri_verts instead of __tri_storage in intersection code. Unfortunately, this is quite tricky to do without noticeable speed loss. Mainly this loss is caused by extra lookup happening to access vertex coordinate. Fortunately, tricks here and there (i,e, some types changes to avoid casts which are not really coming for free) reduces those losses to an acceptable level. So now they are within couple of percent only, On a positive site we've achieved: - Few percent of memory save with triangle-only scenes. Actual save in this case is close to size of all vertices. On a more fine-subdivided scenes this benefit might become more obvious. - Huge memory save of hairy scenes. For example, on koro.blend there is about 20% memory save. Similar figure for bunny.blend. This memory save was the main goal of this commit to move forward with Hair BVH which required more memory per BVH node. So while this sounds exciting, this memory optimization will become invisible by upcoming Hair BVH work. But again on a positive side, we can add an option to NOT use Hair BVH and then we'll have same-ish render times as we've got currently but will have this 20% memory benefit on hairy scenes.
This commit is contained in:
parent
1eacbf47e3
commit
17e7454263
|
|
@ -121,7 +121,7 @@ void BVH::refit(Progress& progress)
|
|||
|
||||
/* Triangles */
|
||||
|
||||
void BVH::pack_triangle(int idx, float4 storage[3])
|
||||
void BVH::pack_triangle(int idx, float4 tri_verts[3])
|
||||
{
|
||||
int tob = pack.prim_object[idx];
|
||||
assert(tob >= 0 && tob < objects.size());
|
||||
|
|
@ -129,49 +129,58 @@ void BVH::pack_triangle(int idx, float4 storage[3])
|
|||
|
||||
int tidx = pack.prim_index[idx];
|
||||
Mesh::Triangle t = mesh->get_triangle(tidx);
|
||||
const float3* vpos = &mesh->verts[0];
|
||||
const float3 *vpos = &mesh->verts[0];
|
||||
float3 v0 = vpos[t.v[0]];
|
||||
float3 v1 = vpos[t.v[1]];
|
||||
float3 v2 = vpos[t.v[2]];
|
||||
|
||||
storage[0] = float3_to_float4(v0);
|
||||
storage[1] = float3_to_float4(v1);
|
||||
storage[2] = float3_to_float4(v2);
|
||||
tri_verts[0] = float3_to_float4(v0);
|
||||
tri_verts[1] = float3_to_float4(v1);
|
||||
tri_verts[2] = float3_to_float4(v2);
|
||||
}
|
||||
|
||||
void BVH::pack_primitives()
|
||||
{
|
||||
int nsize = TRI_NODE_SIZE;
|
||||
size_t tidx_size = pack.prim_index.size();
|
||||
|
||||
pack.tri_storage.clear();
|
||||
pack.tri_storage.resize(tidx_size * nsize);
|
||||
const size_t tidx_size = pack.prim_index.size();
|
||||
size_t num_prim_triangles = 0;
|
||||
/* Count number of triangles primitives in BVH. */
|
||||
for(unsigned int i = 0; i < tidx_size; i++) {
|
||||
if((pack.prim_index[i] != -1)) {
|
||||
if ((pack.prim_type[i] & PRIMITIVE_ALL_TRIANGLE) != 0) {
|
||||
++num_prim_triangles;
|
||||
}
|
||||
}
|
||||
}
|
||||
/* Reserve size for arrays. */
|
||||
pack.prim_tri_index.clear();
|
||||
pack.prim_tri_index.resize(tidx_size);
|
||||
pack.prim_tri_verts.clear();
|
||||
pack.prim_tri_verts.resize(num_prim_triangles * 3);
|
||||
pack.prim_visibility.clear();
|
||||
pack.prim_visibility.resize(tidx_size);
|
||||
|
||||
/* Fill in all the arrays. */
|
||||
size_t prim_triangle_index = 0;
|
||||
for(unsigned int i = 0; i < tidx_size; i++) {
|
||||
if(pack.prim_index[i] != -1) {
|
||||
float4 storage[3];
|
||||
|
||||
if(pack.prim_type[i] & PRIMITIVE_TRIANGLE) {
|
||||
pack_triangle(i, storage);
|
||||
}
|
||||
else {
|
||||
/* Avoid use of uninitialized memory. */
|
||||
memset(&storage, 0, sizeof(storage));
|
||||
}
|
||||
|
||||
memcpy(&pack.tri_storage[i * nsize], storage, sizeof(float4)*3);
|
||||
|
||||
int tob = pack.prim_object[i];
|
||||
Object *ob = objects[tob];
|
||||
|
||||
if((pack.prim_type[i] & PRIMITIVE_ALL_TRIANGLE) != 0) {
|
||||
pack_triangle(i, (float4*)&pack.prim_tri_verts[3 * prim_triangle_index]);
|
||||
pack.prim_tri_index[i] = 3 * prim_triangle_index;
|
||||
++prim_triangle_index;
|
||||
}
|
||||
else {
|
||||
pack.prim_tri_index[i] = -1;
|
||||
}
|
||||
|
||||
pack.prim_visibility[i] = ob->visibility;
|
||||
|
||||
if(pack.prim_type[i] & PRIMITIVE_ALL_CURVE)
|
||||
pack.prim_visibility[i] |= PATH_RAY_CURVE;
|
||||
}
|
||||
else {
|
||||
memset(&pack.tri_storage[i * nsize], 0, sizeof(float4)*3);
|
||||
pack.prim_tri_index[i] = -1;
|
||||
pack.prim_visibility[i] = 0;
|
||||
}
|
||||
}
|
||||
|
|
@ -208,10 +217,10 @@ void BVH::pack_instances(size_t nodes_size, size_t leaf_nodes_size)
|
|||
|
||||
/* reserve */
|
||||
size_t prim_index_size = pack.prim_index.size();
|
||||
size_t tri_storage_size = pack.tri_storage.size();
|
||||
size_t prim_tri_verts_size = pack.prim_tri_verts.size();
|
||||
|
||||
size_t pack_prim_index_offset = prim_index_size;
|
||||
size_t pack_tri_storage_offset = tri_storage_size;
|
||||
size_t pack_prim_tri_verts_offset = prim_tri_verts_size;
|
||||
size_t pack_nodes_offset = nodes_size;
|
||||
size_t pack_leaf_nodes_offset = leaf_nodes_size;
|
||||
size_t object_offset = 0;
|
||||
|
|
@ -225,7 +234,7 @@ void BVH::pack_instances(size_t nodes_size, size_t leaf_nodes_size)
|
|||
if(mesh->need_build_bvh()) {
|
||||
if(mesh_map.find(mesh) == mesh_map.end()) {
|
||||
prim_index_size += bvh->pack.prim_index.size();
|
||||
tri_storage_size += bvh->pack.tri_storage.size();
|
||||
prim_tri_verts_size += bvh->pack.prim_tri_verts.size();
|
||||
nodes_size += bvh->pack.nodes.size();
|
||||
leaf_nodes_size += bvh->pack.leaf_nodes.size();
|
||||
|
||||
|
|
@ -240,7 +249,8 @@ void BVH::pack_instances(size_t nodes_size, size_t leaf_nodes_size)
|
|||
pack.prim_type.resize(prim_index_size);
|
||||
pack.prim_object.resize(prim_index_size);
|
||||
pack.prim_visibility.resize(prim_index_size);
|
||||
pack.tri_storage.resize(tri_storage_size);
|
||||
pack.prim_tri_verts.resize(prim_tri_verts_size);
|
||||
pack.prim_tri_index.resize(prim_index_size);
|
||||
pack.nodes.resize(nodes_size);
|
||||
pack.leaf_nodes.resize(leaf_nodes_size);
|
||||
pack.object_node.resize(objects.size());
|
||||
|
|
@ -249,7 +259,8 @@ void BVH::pack_instances(size_t nodes_size, size_t leaf_nodes_size)
|
|||
int *pack_prim_type = (pack.prim_type.size())? &pack.prim_type[0]: NULL;
|
||||
int *pack_prim_object = (pack.prim_object.size())? &pack.prim_object[0]: NULL;
|
||||
uint *pack_prim_visibility = (pack.prim_visibility.size())? &pack.prim_visibility[0]: NULL;
|
||||
float4 *pack_tri_storage = (pack.tri_storage.size())? &pack.tri_storage[0]: NULL;
|
||||
float4 *pack_prim_tri_verts = (pack.prim_tri_verts.size())? &pack.prim_tri_verts[0]: NULL;
|
||||
uint *pack_prim_tri_index = (pack.prim_tri_index.size())? &pack.prim_tri_index[0]: NULL;
|
||||
int4 *pack_nodes = (pack.nodes.size())? &pack.nodes[0]: NULL;
|
||||
int4 *pack_leaf_nodes = (pack.leaf_nodes.size())? &pack.leaf_nodes[0]: NULL;
|
||||
|
||||
|
|
@ -290,18 +301,24 @@ void BVH::pack_instances(size_t nodes_size, size_t leaf_nodes_size)
|
|||
|
||||
mesh_map[mesh] = pack.object_node[object_offset-1];
|
||||
|
||||
/* merge primitive and object indexes */
|
||||
/* merge primitive, object and triangle indexes */
|
||||
if(bvh->pack.prim_index.size()) {
|
||||
size_t bvh_prim_index_size = bvh->pack.prim_index.size();
|
||||
int *bvh_prim_index = &bvh->pack.prim_index[0];
|
||||
int *bvh_prim_type = &bvh->pack.prim_type[0];
|
||||
uint *bvh_prim_visibility = &bvh->pack.prim_visibility[0];
|
||||
uint *bvh_prim_tri_index = &bvh->pack.prim_tri_index[0];
|
||||
|
||||
for(size_t i = 0; i < bvh_prim_index_size; i++) {
|
||||
if(bvh->pack.prim_type[i] & PRIMITIVE_ALL_CURVE)
|
||||
if(bvh->pack.prim_type[i] & PRIMITIVE_ALL_CURVE) {
|
||||
pack_prim_index[pack_prim_index_offset] = bvh_prim_index[i] + mesh_curve_offset;
|
||||
else
|
||||
pack_prim_tri_index[pack_prim_index_offset] = -1;
|
||||
}
|
||||
else {
|
||||
pack_prim_index[pack_prim_index_offset] = bvh_prim_index[i] + mesh_tri_offset;
|
||||
pack_prim_tri_index[pack_prim_index_offset] =
|
||||
bvh_prim_tri_index[i] + pack_prim_tri_verts_offset;
|
||||
}
|
||||
|
||||
pack_prim_type[pack_prim_index_offset] = bvh_prim_type[i];
|
||||
pack_prim_visibility[pack_prim_index_offset] = bvh_prim_visibility[i];
|
||||
|
|
@ -310,12 +327,13 @@ void BVH::pack_instances(size_t nodes_size, size_t leaf_nodes_size)
|
|||
}
|
||||
}
|
||||
|
||||
/* merge triangle intersection data */
|
||||
if(bvh->pack.tri_storage.size()) {
|
||||
memcpy(pack_tri_storage + pack_tri_storage_offset,
|
||||
&bvh->pack.tri_storage[0],
|
||||
bvh->pack.tri_storage.size()*sizeof(float4));
|
||||
pack_tri_storage_offset += bvh->pack.tri_storage.size();
|
||||
/* Merge triangle vertices data. */
|
||||
if(bvh->pack.prim_tri_verts.size()) {
|
||||
const size_t prim_tri_size = bvh->pack.prim_tri_verts.size();
|
||||
memcpy(pack_prim_tri_verts + pack_prim_tri_verts_offset,
|
||||
&bvh->pack.prim_tri_verts[0],
|
||||
prim_tri_size*sizeof(float4));
|
||||
pack_prim_tri_verts_offset += prim_tri_size;
|
||||
}
|
||||
|
||||
/* merge nodes */
|
||||
|
|
|
|||
|
|
@ -52,8 +52,10 @@ struct PackedBVH {
|
|||
array<int4> leaf_nodes;
|
||||
/* object index to BVH node index mapping for instances */
|
||||
array<int> object_node;
|
||||
/* Aligned triangle storage for fatser lookup in the kernel. */
|
||||
array<float4> tri_storage;
|
||||
/* Mapping from primitive index to index in triangle array. */
|
||||
array<uint> prim_tri_index;
|
||||
/* Continuous storage of triangle vertices. */
|
||||
array<float4> prim_tri_verts;
|
||||
/* primitive type - triangle or strand */
|
||||
array<int> prim_type;
|
||||
/* visibility visibilitys for primitives */
|
||||
|
|
@ -91,7 +93,7 @@ public:
|
|||
protected:
|
||||
BVH(const BVHParams& params, const vector<Object*>& objects);
|
||||
|
||||
/* triangles and strands*/
|
||||
/* triangles and strands */
|
||||
void pack_primitives();
|
||||
void pack_triangle(int idx, float4 storage[3]);
|
||||
|
||||
|
|
|
|||
|
|
@ -47,13 +47,13 @@ ccl_device_inline int find_attribute_motion(KernelGlobals *kg, int object, uint
|
|||
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, float3 tri_vindex, int offset, int numverts, int numsteps, int step, float3 verts[3])
|
||||
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(__tri_verts, __float_as_int(tri_vindex.x)));
|
||||
verts[1] = float4_to_float3(kernel_tex_fetch(__tri_verts, __float_as_int(tri_vindex.y)));
|
||||
verts[2] = float4_to_float3(kernel_tex_fetch(__tri_verts, __float_as_int(tri_vindex.z)));
|
||||
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 */
|
||||
|
|
@ -62,19 +62,19 @@ ccl_device_inline void motion_triangle_verts_for_step(KernelGlobals *kg, float3
|
|||
|
||||
offset += step*numverts;
|
||||
|
||||
verts[0] = float4_to_float3(kernel_tex_fetch(__attributes_float3, offset + __float_as_int(tri_vindex.x)));
|
||||
verts[1] = float4_to_float3(kernel_tex_fetch(__attributes_float3, offset + __float_as_int(tri_vindex.y)));
|
||||
verts[2] = float4_to_float3(kernel_tex_fetch(__attributes_float3, offset + __float_as_int(tri_vindex.z)));
|
||||
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, float3 tri_vindex, int offset, int numverts, int numsteps, int step, float3 normals[3])
|
||||
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, __float_as_int(tri_vindex.x)));
|
||||
normals[1] = float4_to_float3(kernel_tex_fetch(__tri_vnormal, __float_as_int(tri_vindex.y)));
|
||||
normals[2] = float4_to_float3(kernel_tex_fetch(__tri_vnormal, __float_as_int(tri_vindex.z)));
|
||||
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 */
|
||||
|
|
@ -83,9 +83,9 @@ ccl_device_inline void motion_triangle_normals_for_step(KernelGlobals *kg, float
|
|||
|
||||
offset += step*numverts;
|
||||
|
||||
normals[0] = float4_to_float3(kernel_tex_fetch(__attributes_float3, offset + __float_as_int(tri_vindex.x)));
|
||||
normals[1] = float4_to_float3(kernel_tex_fetch(__attributes_float3, offset + __float_as_int(tri_vindex.y)));
|
||||
normals[2] = float4_to_float3(kernel_tex_fetch(__attributes_float3, offset + __float_as_int(tri_vindex.z)));
|
||||
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));
|
||||
}
|
||||
}
|
||||
|
||||
|
|
@ -107,7 +107,7 @@ ccl_device_inline void motion_triangle_vertices(KernelGlobals *kg, int object, i
|
|||
|
||||
/* fetch vertex coordinates */
|
||||
float3 next_verts[3];
|
||||
float3 tri_vindex = float4_to_float3(kernel_tex_fetch(__tri_vindex, prim));
|
||||
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);
|
||||
|
|
@ -259,7 +259,7 @@ ccl_device_noinline void motion_triangle_shader_setup(KernelGlobals *kg, ShaderD
|
|||
|
||||
/* fetch vertex coordinates */
|
||||
float3 verts[3], next_verts[3];
|
||||
float3 tri_vindex = float4_to_float3(kernel_tex_fetch(__tri_vindex, ccl_fetch(sd, prim)));
|
||||
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);
|
||||
|
|
|
|||
|
|
@ -27,12 +27,11 @@ CCL_NAMESPACE_BEGIN
|
|||
ccl_device_inline float3 triangle_normal(KernelGlobals *kg, ShaderData *sd)
|
||||
{
|
||||
/* load triangle vertices */
|
||||
float4 tri_vindex = kernel_tex_fetch(__tri_vindex, ccl_fetch(sd, prim));
|
||||
const uint4 tri_vindex = kernel_tex_fetch(__tri_vindex, ccl_fetch(sd, prim));
|
||||
const float3 v0 = float4_to_float3(kernel_tex_fetch(__prim_tri_verts, tri_vindex.w+0));
|
||||
const float3 v1 = float4_to_float3(kernel_tex_fetch(__prim_tri_verts, tri_vindex.w+1));
|
||||
const float3 v2 = float4_to_float3(kernel_tex_fetch(__prim_tri_verts, tri_vindex.w+2));
|
||||
|
||||
float3 v0 = float4_to_float3(kernel_tex_fetch(__tri_verts, __float_as_int(tri_vindex.x)));
|
||||
float3 v1 = float4_to_float3(kernel_tex_fetch(__tri_verts, __float_as_int(tri_vindex.y)));
|
||||
float3 v2 = float4_to_float3(kernel_tex_fetch(__tri_verts, __float_as_int(tri_vindex.z)));
|
||||
|
||||
/* return normal */
|
||||
if(ccl_fetch(sd, flag) & SD_NEGATIVE_SCALE_APPLIED)
|
||||
return normalize(cross(v2 - v0, v1 - v0));
|
||||
|
|
@ -44,11 +43,10 @@ ccl_device_inline float3 triangle_normal(KernelGlobals *kg, ShaderData *sd)
|
|||
ccl_device_inline void triangle_point_normal(KernelGlobals *kg, int object, int prim, float u, float v, float3 *P, float3 *Ng, int *shader)
|
||||
{
|
||||
/* load triangle vertices */
|
||||
float4 tri_vindex = kernel_tex_fetch(__tri_vindex, prim);
|
||||
|
||||
float3 v0 = float4_to_float3(kernel_tex_fetch(__tri_verts, __float_as_int(tri_vindex.x)));
|
||||
float3 v1 = float4_to_float3(kernel_tex_fetch(__tri_verts, __float_as_int(tri_vindex.y)));
|
||||
float3 v2 = float4_to_float3(kernel_tex_fetch(__tri_verts, __float_as_int(tri_vindex.z)));
|
||||
const uint4 tri_vindex = kernel_tex_fetch(__tri_vindex, prim);
|
||||
float3 v0 = float4_to_float3(kernel_tex_fetch(__prim_tri_verts, tri_vindex.w+0));
|
||||
float3 v1 = float4_to_float3(kernel_tex_fetch(__prim_tri_verts, tri_vindex.w+1));
|
||||
float3 v2 = float4_to_float3(kernel_tex_fetch(__prim_tri_verts, tri_vindex.w+2));
|
||||
|
||||
/* compute point */
|
||||
float t = 1.0f - u - v;
|
||||
|
|
@ -71,11 +69,10 @@ ccl_device_inline void triangle_point_normal(KernelGlobals *kg, int object, int
|
|||
|
||||
ccl_device_inline void triangle_vertices(KernelGlobals *kg, int prim, float3 P[3])
|
||||
{
|
||||
float4 tri_vindex = kernel_tex_fetch(__tri_vindex, prim);
|
||||
|
||||
P[0] = float4_to_float3(kernel_tex_fetch(__tri_verts, __float_as_int(tri_vindex.x)));
|
||||
P[1] = float4_to_float3(kernel_tex_fetch(__tri_verts, __float_as_int(tri_vindex.y)));
|
||||
P[2] = float4_to_float3(kernel_tex_fetch(__tri_verts, __float_as_int(tri_vindex.z)));
|
||||
const uint4 tri_vindex = kernel_tex_fetch(__tri_vindex, prim);
|
||||
P[0] = float4_to_float3(kernel_tex_fetch(__prim_tri_verts, tri_vindex.w+0));
|
||||
P[1] = float4_to_float3(kernel_tex_fetch(__prim_tri_verts, tri_vindex.w+1));
|
||||
P[2] = float4_to_float3(kernel_tex_fetch(__prim_tri_verts, tri_vindex.w+2));
|
||||
}
|
||||
|
||||
/* Interpolate smooth vertex normal from vertices */
|
||||
|
|
@ -83,11 +80,10 @@ ccl_device_inline void triangle_vertices(KernelGlobals *kg, int prim, float3 P[3
|
|||
ccl_device_inline float3 triangle_smooth_normal(KernelGlobals *kg, int prim, float u, float v)
|
||||
{
|
||||
/* load triangle vertices */
|
||||
float4 tri_vindex = kernel_tex_fetch(__tri_vindex, prim);
|
||||
|
||||
float3 n0 = float4_to_float3(kernel_tex_fetch(__tri_vnormal, __float_as_int(tri_vindex.x)));
|
||||
float3 n1 = float4_to_float3(kernel_tex_fetch(__tri_vnormal, __float_as_int(tri_vindex.y)));
|
||||
float3 n2 = float4_to_float3(kernel_tex_fetch(__tri_vnormal, __float_as_int(tri_vindex.z)));
|
||||
const uint4 tri_vindex = kernel_tex_fetch(__tri_vindex, prim);
|
||||
float3 n0 = float4_to_float3(kernel_tex_fetch(__tri_vnormal, tri_vindex.x));
|
||||
float3 n1 = float4_to_float3(kernel_tex_fetch(__tri_vnormal, tri_vindex.y));
|
||||
float3 n2 = float4_to_float3(kernel_tex_fetch(__tri_vnormal, tri_vindex.z));
|
||||
|
||||
return normalize((1.0f - u - v)*n2 + u*n0 + v*n1);
|
||||
}
|
||||
|
|
@ -97,11 +93,10 @@ ccl_device_inline float3 triangle_smooth_normal(KernelGlobals *kg, int prim, flo
|
|||
ccl_device_inline void triangle_dPdudv(KernelGlobals *kg, int prim, ccl_addr_space float3 *dPdu, ccl_addr_space float3 *dPdv)
|
||||
{
|
||||
/* fetch triangle vertex coordinates */
|
||||
float4 tri_vindex = kernel_tex_fetch(__tri_vindex, prim);
|
||||
|
||||
float3 p0 = float4_to_float3(kernel_tex_fetch(__tri_verts, __float_as_int(tri_vindex.x)));
|
||||
float3 p1 = float4_to_float3(kernel_tex_fetch(__tri_verts, __float_as_int(tri_vindex.y)));
|
||||
float3 p2 = float4_to_float3(kernel_tex_fetch(__tri_verts, __float_as_int(tri_vindex.z)));
|
||||
const uint4 tri_vindex = kernel_tex_fetch(__tri_vindex, prim);
|
||||
const float3 p0 = float4_to_float3(kernel_tex_fetch(__prim_tri_verts, tri_vindex.w+0));
|
||||
const float3 p1 = float4_to_float3(kernel_tex_fetch(__prim_tri_verts, tri_vindex.w+1));
|
||||
const float3 p2 = float4_to_float3(kernel_tex_fetch(__prim_tri_verts, tri_vindex.w+2));
|
||||
|
||||
/* compute derivatives of P w.r.t. uv */
|
||||
*dPdu = (p0 - p2);
|
||||
|
|
@ -119,11 +114,11 @@ ccl_device float triangle_attribute_float(KernelGlobals *kg, const ShaderData *s
|
|||
return kernel_tex_fetch(__attributes_float, offset + ccl_fetch(sd, prim));
|
||||
}
|
||||
else if(elem == ATTR_ELEMENT_VERTEX || elem == ATTR_ELEMENT_VERTEX_MOTION) {
|
||||
float4 tri_vindex = kernel_tex_fetch(__tri_vindex, ccl_fetch(sd, prim));
|
||||
uint4 tri_vindex = kernel_tex_fetch(__tri_vindex, ccl_fetch(sd, prim));
|
||||
|
||||
float f0 = kernel_tex_fetch(__attributes_float, offset + __float_as_int(tri_vindex.x));
|
||||
float f1 = kernel_tex_fetch(__attributes_float, offset + __float_as_int(tri_vindex.y));
|
||||
float f2 = kernel_tex_fetch(__attributes_float, offset + __float_as_int(tri_vindex.z));
|
||||
float f0 = kernel_tex_fetch(__attributes_float, offset + tri_vindex.x);
|
||||
float f1 = kernel_tex_fetch(__attributes_float, offset + tri_vindex.y);
|
||||
float f2 = kernel_tex_fetch(__attributes_float, offset + tri_vindex.z);
|
||||
|
||||
#ifdef __RAY_DIFFERENTIALS__
|
||||
if(dx) *dx = ccl_fetch(sd, du).dx*f0 + ccl_fetch(sd, dv).dx*f1 - (ccl_fetch(sd, du).dx + ccl_fetch(sd, dv).dx)*f2;
|
||||
|
|
@ -162,11 +157,11 @@ ccl_device float3 triangle_attribute_float3(KernelGlobals *kg, const ShaderData
|
|||
return float4_to_float3(kernel_tex_fetch(__attributes_float3, offset + ccl_fetch(sd, prim)));
|
||||
}
|
||||
else if(elem == ATTR_ELEMENT_VERTEX || elem == ATTR_ELEMENT_VERTEX_MOTION) {
|
||||
float4 tri_vindex = kernel_tex_fetch(__tri_vindex, ccl_fetch(sd, prim));
|
||||
uint4 tri_vindex = kernel_tex_fetch(__tri_vindex, ccl_fetch(sd, prim));
|
||||
|
||||
float3 f0 = float4_to_float3(kernel_tex_fetch(__attributes_float3, offset + __float_as_int(tri_vindex.x)));
|
||||
float3 f1 = float4_to_float3(kernel_tex_fetch(__attributes_float3, offset + __float_as_int(tri_vindex.y)));
|
||||
float3 f2 = float4_to_float3(kernel_tex_fetch(__attributes_float3, offset + __float_as_int(tri_vindex.z)));
|
||||
float3 f0 = float4_to_float3(kernel_tex_fetch(__attributes_float3, offset + tri_vindex.x));
|
||||
float3 f1 = float4_to_float3(kernel_tex_fetch(__attributes_float3, offset + tri_vindex.y));
|
||||
float3 f2 = float4_to_float3(kernel_tex_fetch(__attributes_float3, offset + tri_vindex.z));
|
||||
|
||||
#ifdef __RAY_DIFFERENTIALS__
|
||||
if(dx) *dx = ccl_fetch(sd, du).dx*f0 + ccl_fetch(sd, dv).dx*f1 - (ccl_fetch(sd, du).dx + ccl_fetch(sd, dv).dx)*f2;
|
||||
|
|
|
|||
|
|
@ -106,9 +106,10 @@ ccl_device_inline bool triangle_intersect(KernelGlobals *kg,
|
|||
const float Sz = isect_precalc->Sz;
|
||||
|
||||
/* Calculate vertices relative to ray origin. */
|
||||
const float4 tri_a = kernel_tex_fetch(__tri_storage, triAddr*TRI_NODE_SIZE+0),
|
||||
tri_b = kernel_tex_fetch(__tri_storage, triAddr*TRI_NODE_SIZE+1),
|
||||
tri_c = kernel_tex_fetch(__tri_storage, triAddr*TRI_NODE_SIZE+2);
|
||||
const uint tri_vindex = kernel_tex_fetch(__prim_tri_index, triAddr);
|
||||
const float4 tri_a = kernel_tex_fetch(__prim_tri_verts, tri_vindex+0),
|
||||
tri_b = kernel_tex_fetch(__prim_tri_verts, tri_vindex+1),
|
||||
tri_c = kernel_tex_fetch(__prim_tri_verts, tri_vindex+2);
|
||||
const float3 A = make_float3(tri_a.x - P.x, tri_a.y - P.y, tri_a.z - P.z);
|
||||
const float3 B = make_float3(tri_b.x - P.x, tri_b.y - P.y, tri_b.z - P.z);
|
||||
const float3 C = make_float3(tri_c.x - P.x, tri_c.y - P.y, tri_c.z - P.z);
|
||||
|
|
@ -202,9 +203,10 @@ ccl_device_inline void triangle_intersect_subsurface(
|
|||
const float Sz = isect_precalc->Sz;
|
||||
|
||||
/* Calculate vertices relative to ray origin. */
|
||||
const float4 tri_a = kernel_tex_fetch(__tri_storage, triAddr*TRI_NODE_SIZE+0),
|
||||
tri_b = kernel_tex_fetch(__tri_storage, triAddr*TRI_NODE_SIZE+1),
|
||||
tri_c = kernel_tex_fetch(__tri_storage, triAddr*TRI_NODE_SIZE+2);
|
||||
const uint tri_vindex = kernel_tex_fetch(__prim_tri_index, triAddr);
|
||||
const float4 tri_a = kernel_tex_fetch(__prim_tri_verts, tri_vindex+0),
|
||||
tri_b = kernel_tex_fetch(__prim_tri_verts, tri_vindex+1),
|
||||
tri_c = kernel_tex_fetch(__prim_tri_verts, tri_vindex+2);
|
||||
const float3 A = make_float3(tri_a.x - P.x, tri_a.y - P.y, tri_a.z - P.z);
|
||||
const float3 B = make_float3(tri_b.x - P.x, tri_b.y - P.y, tri_b.z - P.z);
|
||||
const float3 C = make_float3(tri_c.x - P.x, tri_c.y - P.y, tri_c.z - P.z);
|
||||
|
|
@ -324,9 +326,10 @@ ccl_device_inline float3 triangle_refine(KernelGlobals *kg,
|
|||
|
||||
P = P + D*t;
|
||||
|
||||
const float4 tri_a = kernel_tex_fetch(__tri_storage, isect->prim*TRI_NODE_SIZE+0),
|
||||
tri_b = kernel_tex_fetch(__tri_storage, isect->prim*TRI_NODE_SIZE+1),
|
||||
tri_c = kernel_tex_fetch(__tri_storage, isect->prim*TRI_NODE_SIZE+2);
|
||||
const uint tri_vindex = kernel_tex_fetch(__prim_tri_index, isect->prim);
|
||||
const float4 tri_a = kernel_tex_fetch(__prim_tri_verts, tri_vindex+0),
|
||||
tri_b = kernel_tex_fetch(__prim_tri_verts, tri_vindex+1),
|
||||
tri_c = kernel_tex_fetch(__prim_tri_verts, tri_vindex+2);
|
||||
float3 edge1 = make_float3(tri_a.x - tri_c.x, tri_a.y - tri_c.y, tri_a.z - tri_c.z);
|
||||
float3 edge2 = make_float3(tri_b.x - tri_c.x, tri_b.y - tri_c.y, tri_b.z - tri_c.z);
|
||||
float3 tvec = make_float3(P.x - tri_c.x, P.y - tri_c.y, P.z - tri_c.z);
|
||||
|
|
@ -381,9 +384,10 @@ ccl_device_inline float3 triangle_refine_subsurface(KernelGlobals *kg,
|
|||
P = P + D*t;
|
||||
|
||||
#ifdef __INTERSECTION_REFINE__
|
||||
const float4 tri_a = kernel_tex_fetch(__tri_storage, isect->prim*TRI_NODE_SIZE+0),
|
||||
tri_b = kernel_tex_fetch(__tri_storage, isect->prim*TRI_NODE_SIZE+1),
|
||||
tri_c = kernel_tex_fetch(__tri_storage, isect->prim*TRI_NODE_SIZE+2);
|
||||
const uint tri_vindex = kernel_tex_fetch(__prim_tri_index, isect->prim);
|
||||
const float4 tri_a = kernel_tex_fetch(__prim_tri_verts, tri_vindex+0),
|
||||
tri_b = kernel_tex_fetch(__prim_tri_verts, tri_vindex+1),
|
||||
tri_c = kernel_tex_fetch(__prim_tri_verts, tri_vindex+2);
|
||||
float3 edge1 = make_float3(tri_a.x - tri_c.x, tri_a.y - tri_c.y, tri_a.z - tri_c.z);
|
||||
float3 edge2 = make_float3(tri_b.x - tri_c.x, tri_b.y - tri_c.y, tri_b.z - tri_c.z);
|
||||
float3 tvec = make_float3(P.x - tri_c.x, P.y - tri_c.y, P.z - tri_c.z);
|
||||
|
|
|
|||
|
|
@ -25,7 +25,8 @@
|
|||
/* bvh */
|
||||
KERNEL_TEX(float4, texture_float4, __bvh_nodes)
|
||||
KERNEL_TEX(float4, texture_float4, __bvh_leaf_nodes)
|
||||
KERNEL_TEX(float4, texture_float4, __tri_storage)
|
||||
KERNEL_TEX(float4, texture_float4, __prim_tri_verts)
|
||||
KERNEL_TEX(uint, texture_uint, __prim_tri_index)
|
||||
KERNEL_TEX(uint, texture_uint, __prim_type)
|
||||
KERNEL_TEX(uint, texture_uint, __prim_visibility)
|
||||
KERNEL_TEX(uint, texture_uint, __prim_index)
|
||||
|
|
@ -39,8 +40,7 @@ KERNEL_TEX(float4, texture_float4, __objects_vector)
|
|||
/* triangles */
|
||||
KERNEL_TEX(uint, texture_uint, __tri_shader)
|
||||
KERNEL_TEX(float4, texture_float4, __tri_vnormal)
|
||||
KERNEL_TEX(float4, texture_float4, __tri_vindex)
|
||||
KERNEL_TEX(float4, texture_float4, __tri_verts)
|
||||
KERNEL_TEX(uint4, texture_uint4, __tri_vindex)
|
||||
|
||||
/* curves */
|
||||
KERNEL_TEX(float4, texture_float4, __curves)
|
||||
|
|
|
|||
|
|
@ -472,30 +472,19 @@ void Mesh::pack_normals(Scene *scene, uint *tri_shader, float4 *vnormal)
|
|||
}
|
||||
}
|
||||
|
||||
void Mesh::pack_verts(float4 *tri_verts, float4 *tri_vindex, size_t vert_offset)
|
||||
void Mesh::pack_verts(const vector<uint>& tri_prim_index,
|
||||
uint4 *tri_vindex,
|
||||
size_t vert_offset,
|
||||
size_t tri_offset)
|
||||
{
|
||||
size_t verts_size = verts.size();
|
||||
|
||||
if(verts_size) {
|
||||
float3 *verts_ptr = &verts[0];
|
||||
|
||||
for(size_t i = 0; i < verts_size; i++) {
|
||||
float3 p = verts_ptr[i];
|
||||
tri_verts[i] = make_float4(p.x, p.y, p.z, 0.0f);
|
||||
}
|
||||
}
|
||||
|
||||
size_t triangles_size = num_triangles();
|
||||
|
||||
const size_t triangles_size = num_triangles();
|
||||
if(triangles_size) {
|
||||
for(size_t i = 0; i < triangles_size; i++) {
|
||||
Triangle t = get_triangle(i);
|
||||
|
||||
tri_vindex[i] = make_float4(
|
||||
__int_as_float(t.v[0] + vert_offset),
|
||||
__int_as_float(t.v[1] + vert_offset),
|
||||
__int_as_float(t.v[2] + vert_offset),
|
||||
0);
|
||||
tri_vindex[i] = make_uint4(t.v[0] + vert_offset,
|
||||
t.v[1] + vert_offset,
|
||||
t.v[2] + vert_offset,
|
||||
tri_prim_index[i + tri_offset]);
|
||||
}
|
||||
}
|
||||
}
|
||||
|
|
@ -1070,42 +1059,82 @@ void MeshManager::device_update_attributes(Device *device, DeviceScene *dscene,
|
|||
}
|
||||
}
|
||||
|
||||
void MeshManager::device_update_mesh(Device *device, DeviceScene *dscene, Scene *scene, Progress& progress)
|
||||
void MeshManager::mesh_calc_offset(Scene *scene)
|
||||
{
|
||||
/* count and update offsets */
|
||||
size_t vert_size = 0;
|
||||
size_t tri_size = 0;
|
||||
|
||||
size_t curve_key_size = 0;
|
||||
size_t curve_size = 0;
|
||||
|
||||
foreach(Mesh *mesh, scene->meshes) {
|
||||
mesh->vert_offset = vert_size;
|
||||
mesh->tri_offset = tri_size;
|
||||
|
||||
mesh->curvekey_offset = curve_key_size;
|
||||
mesh->curve_offset = curve_size;
|
||||
|
||||
vert_size += mesh->verts.size();
|
||||
tri_size += mesh->num_triangles();
|
||||
|
||||
curve_key_size += mesh->curve_keys.size();
|
||||
curve_size += mesh->num_curves();
|
||||
}
|
||||
}
|
||||
|
||||
void MeshManager::device_update_mesh(Device *device,
|
||||
DeviceScene *dscene,
|
||||
Scene *scene,
|
||||
bool for_displacement,
|
||||
Progress& progress)
|
||||
{
|
||||
/* Count. */
|
||||
size_t vert_size = 0;
|
||||
size_t tri_size = 0;
|
||||
size_t curve_key_size = 0;
|
||||
size_t curve_size = 0;
|
||||
foreach(Mesh *mesh, scene->meshes) {
|
||||
vert_size += mesh->verts.size();
|
||||
tri_size += mesh->num_triangles();
|
||||
curve_key_size += mesh->curve_keys.size();
|
||||
curve_size += mesh->num_curves();
|
||||
}
|
||||
/* Create mapping from triangle to primitive triangle array. */
|
||||
vector<uint> tri_prim_index(tri_size);
|
||||
if(for_displacement) {
|
||||
/* For displacement kernels we do some trickery to make them believe
|
||||
* we've got all required data ready. However, that data is different
|
||||
* from final render kernels since we don't have BVH yet, so can't
|
||||
* really use same semantic of arrays.
|
||||
*/
|
||||
foreach(Mesh *mesh, scene->meshes) {
|
||||
for(size_t i = 0; i < mesh->num_triangles(); ++i) {
|
||||
tri_prim_index[i + mesh->tri_offset] = 3 * (i + mesh->tri_offset);
|
||||
}
|
||||
}
|
||||
}
|
||||
else {
|
||||
PackedBVH& pack = bvh->pack;
|
||||
for(size_t i = 0; i < pack.prim_index.size(); ++i) {
|
||||
if ((pack.prim_type[i] & PRIMITIVE_ALL_TRIANGLE) != 0) {
|
||||
tri_prim_index[pack.prim_index[i]] = pack.prim_tri_index[i];
|
||||
}
|
||||
}
|
||||
}
|
||||
/* Fill in all the arrays. */
|
||||
if(tri_size != 0) {
|
||||
/* normals */
|
||||
progress.set_status("Updating Mesh", "Computing normals");
|
||||
|
||||
uint *tri_shader = dscene->tri_shader.resize(tri_size);
|
||||
float4 *vnormal = dscene->tri_vnormal.resize(vert_size);
|
||||
float4 *tri_verts = dscene->tri_verts.resize(vert_size);
|
||||
float4 *tri_vindex = dscene->tri_vindex.resize(tri_size);
|
||||
uint4 *tri_vindex = dscene->tri_vindex.resize(tri_size);
|
||||
|
||||
foreach(Mesh *mesh, scene->meshes) {
|
||||
mesh->pack_normals(scene, &tri_shader[mesh->tri_offset], &vnormal[mesh->vert_offset]);
|
||||
mesh->pack_verts(&tri_verts[mesh->vert_offset], &tri_vindex[mesh->tri_offset], mesh->vert_offset);
|
||||
|
||||
mesh->pack_normals(scene,
|
||||
&tri_shader[mesh->tri_offset],
|
||||
&vnormal[mesh->vert_offset]);
|
||||
mesh->pack_verts(tri_prim_index,
|
||||
&tri_vindex[mesh->tri_offset],
|
||||
mesh->vert_offset,
|
||||
mesh->tri_offset);
|
||||
if(progress.get_cancel()) return;
|
||||
}
|
||||
|
||||
|
|
@ -1114,10 +1143,8 @@ void MeshManager::device_update_mesh(Device *device, DeviceScene *dscene, Scene
|
|||
|
||||
device->tex_alloc("__tri_shader", dscene->tri_shader);
|
||||
device->tex_alloc("__tri_vnormal", dscene->tri_vnormal);
|
||||
device->tex_alloc("__tri_verts", dscene->tri_verts);
|
||||
device->tex_alloc("__tri_vindex", dscene->tri_vindex);
|
||||
}
|
||||
|
||||
if(curve_size != 0) {
|
||||
progress.set_status("Updating Mesh", "Copying Strands to device");
|
||||
|
||||
|
|
@ -1132,6 +1159,19 @@ void MeshManager::device_update_mesh(Device *device, DeviceScene *dscene, Scene
|
|||
device->tex_alloc("__curve_keys", dscene->curve_keys);
|
||||
device->tex_alloc("__curves", dscene->curves);
|
||||
}
|
||||
if(for_displacement) {
|
||||
float4 *prim_tri_verts = dscene->prim_tri_verts.resize(tri_size * 3);
|
||||
foreach(Mesh *mesh, scene->meshes) {
|
||||
for(size_t i = 0; i < mesh->num_triangles(); ++i) {
|
||||
Mesh::Triangle t = mesh->get_triangle(i);
|
||||
size_t offset = 3 * (i + mesh->tri_offset);
|
||||
prim_tri_verts[offset + 0] = float3_to_float4(mesh->verts[t.v[0]]);
|
||||
prim_tri_verts[offset + 1] = float3_to_float4(mesh->verts[t.v[1]]);
|
||||
prim_tri_verts[offset + 2] = float3_to_float4(mesh->verts[t.v[2]]);
|
||||
}
|
||||
}
|
||||
device->tex_alloc("__prim_tri_verts", dscene->prim_tri_verts);
|
||||
}
|
||||
}
|
||||
|
||||
void MeshManager::device_update_bvh(Device *device, DeviceScene *dscene, Scene *scene, Progress& progress)
|
||||
|
|
@ -1170,9 +1210,13 @@ void MeshManager::device_update_bvh(Device *device, DeviceScene *dscene, Scene *
|
|||
dscene->object_node.reference((uint*)&pack.object_node[0], pack.object_node.size());
|
||||
device->tex_alloc("__object_node", dscene->object_node);
|
||||
}
|
||||
if(pack.tri_storage.size()) {
|
||||
dscene->tri_storage.reference(&pack.tri_storage[0], pack.tri_storage.size());
|
||||
device->tex_alloc("__tri_storage", dscene->tri_storage);
|
||||
if(pack.prim_tri_index.size()) {
|
||||
dscene->prim_tri_index.reference((uint*)&pack.prim_tri_index[0], pack.prim_tri_index.size());
|
||||
device->tex_alloc("__prim_tri_index", dscene->prim_tri_index);
|
||||
}
|
||||
if(pack.prim_tri_verts.size()) {
|
||||
dscene->prim_tri_verts.reference((float4*)&pack.prim_tri_verts[0], pack.prim_tri_verts.size());
|
||||
device->tex_alloc("__prim_tri_verts", dscene->prim_tri_verts);
|
||||
}
|
||||
if(pack.prim_type.size()) {
|
||||
dscene->prim_type.reference((uint*)&pack.prim_type[0], pack.prim_type.size());
|
||||
|
|
@ -1273,7 +1317,7 @@ void MeshManager::device_update(Device *device, DeviceScene *dscene, Scene *scen
|
|||
|
||||
VLOG(1) << "Total " << scene->meshes.size() << " meshes.";
|
||||
|
||||
/* update normals */
|
||||
/* Update normals. */
|
||||
foreach(Mesh *mesh, scene->meshes) {
|
||||
foreach(Shader *shader, mesh->used_shaders) {
|
||||
if(shader->need_update_attributes)
|
||||
|
|
@ -1289,17 +1333,17 @@ void MeshManager::device_update(Device *device, DeviceScene *dscene, Scene *scen
|
|||
}
|
||||
|
||||
/* Update images needed for true displacement. */
|
||||
bool need_displacement_images = false;
|
||||
bool true_displacement_used = false;
|
||||
bool old_need_object_flags_update = false;
|
||||
foreach(Mesh *mesh, scene->meshes) {
|
||||
if(mesh->need_update &&
|
||||
mesh->displacement_method != Mesh::DISPLACE_BUMP)
|
||||
{
|
||||
need_displacement_images = true;
|
||||
true_displacement_used = true;
|
||||
break;
|
||||
}
|
||||
}
|
||||
if(need_displacement_images) {
|
||||
if(true_displacement_used) {
|
||||
VLOG(1) << "Updating images used for true displacement.";
|
||||
device_update_displacement_images(device, dscene, scene, progress);
|
||||
old_need_object_flags_update = scene->object_manager->need_flags_update;
|
||||
|
|
@ -1310,45 +1354,47 @@ void MeshManager::device_update(Device *device, DeviceScene *dscene, Scene *scen
|
|||
false);
|
||||
}
|
||||
|
||||
/* device update */
|
||||
/* Device update. */
|
||||
device_free(device, dscene);
|
||||
|
||||
device_update_mesh(device, dscene, scene, progress);
|
||||
mesh_calc_offset(scene);
|
||||
if(true_displacement_used) {
|
||||
device_update_mesh(device, dscene, scene, true, progress);
|
||||
}
|
||||
if(progress.get_cancel()) return;
|
||||
|
||||
device_update_attributes(device, dscene, scene, progress);
|
||||
if(progress.get_cancel()) return;
|
||||
|
||||
/* update displacement */
|
||||
/* Update displacement. */
|
||||
bool displacement_done = false;
|
||||
|
||||
foreach(Mesh *mesh, scene->meshes)
|
||||
if(mesh->need_update && displace(device, dscene, scene, mesh, progress))
|
||||
foreach(Mesh *mesh, scene->meshes) {
|
||||
if(mesh->need_update &&
|
||||
displace(device, dscene, scene, mesh, progress))
|
||||
{
|
||||
displacement_done = true;
|
||||
}
|
||||
}
|
||||
|
||||
/* todo: properly handle cancel halfway displacement */
|
||||
/* TODO: properly handle cancel halfway displacement */
|
||||
if(progress.get_cancel()) return;
|
||||
|
||||
/* device re-update after displacement */
|
||||
/* Device re-update after displacement. */
|
||||
if(displacement_done) {
|
||||
device_free(device, dscene);
|
||||
|
||||
device_update_mesh(device, dscene, scene, progress);
|
||||
if(progress.get_cancel()) return;
|
||||
|
||||
device_update_attributes(device, dscene, scene, progress);
|
||||
if(progress.get_cancel()) return;
|
||||
}
|
||||
|
||||
/* update bvh */
|
||||
/* Update bvh. */
|
||||
size_t i = 0, num_bvh = 0;
|
||||
|
||||
foreach(Mesh *mesh, scene->meshes)
|
||||
if(mesh->need_update && mesh->need_build_bvh())
|
||||
foreach(Mesh *mesh, scene->meshes) {
|
||||
if(mesh->need_update && mesh->need_build_bvh()) {
|
||||
num_bvh++;
|
||||
|
||||
}
|
||||
}
|
||||
TaskPool pool;
|
||||
|
||||
foreach(Mesh *mesh, scene->meshes) {
|
||||
if(mesh->need_update) {
|
||||
pool.push(function_bind(&Mesh::compute_bvh,
|
||||
|
|
@ -1362,14 +1408,14 @@ void MeshManager::device_update(Device *device, DeviceScene *dscene, Scene *scen
|
|||
}
|
||||
}
|
||||
}
|
||||
|
||||
TaskPool::Summary summary;
|
||||
pool.wait_work(&summary);
|
||||
VLOG(2) << "Objects BVH build pool statistics:\n"
|
||||
<< summary.full_report();
|
||||
|
||||
foreach(Shader *shader, scene->shaders)
|
||||
foreach(Shader *shader, scene->shaders) {
|
||||
shader->need_update_attributes = false;
|
||||
}
|
||||
|
||||
#ifdef __OBJECT_MOTION__
|
||||
Scene::MotionType need_motion = scene->need_motion(device->info.advanced_shading);
|
||||
|
|
@ -1378,18 +1424,23 @@ void MeshManager::device_update(Device *device, DeviceScene *dscene, Scene *scen
|
|||
bool motion_blur = false;
|
||||
#endif
|
||||
|
||||
/* update obejcts */
|
||||
/* Update objects. */
|
||||
vector<Object *> volume_objects;
|
||||
foreach(Object *object, scene->objects)
|
||||
foreach(Object *object, scene->objects) {
|
||||
object->compute_bounds(motion_blur);
|
||||
}
|
||||
|
||||
if(progress.get_cancel()) return;
|
||||
|
||||
device_update_bvh(device, dscene, scene, progress);
|
||||
if(progress.get_cancel()) return;
|
||||
|
||||
device_update_mesh(device, dscene, scene, false, progress);
|
||||
if(progress.get_cancel()) return;
|
||||
|
||||
need_update = false;
|
||||
|
||||
if(need_displacement_images) {
|
||||
if(true_displacement_used) {
|
||||
/* Re-tag flags for update, so they're re-evaluated
|
||||
* for meshes with correct bounding boxes.
|
||||
*
|
||||
|
|
@ -1405,7 +1456,8 @@ void MeshManager::device_free(Device *device, DeviceScene *dscene)
|
|||
device->tex_free(dscene->bvh_nodes);
|
||||
device->tex_free(dscene->bvh_leaf_nodes);
|
||||
device->tex_free(dscene->object_node);
|
||||
device->tex_free(dscene->tri_storage);
|
||||
device->tex_free(dscene->prim_tri_verts);
|
||||
device->tex_free(dscene->prim_tri_index);
|
||||
device->tex_free(dscene->prim_type);
|
||||
device->tex_free(dscene->prim_visibility);
|
||||
device->tex_free(dscene->prim_index);
|
||||
|
|
@ -1413,7 +1465,6 @@ void MeshManager::device_free(Device *device, DeviceScene *dscene)
|
|||
device->tex_free(dscene->tri_shader);
|
||||
device->tex_free(dscene->tri_vnormal);
|
||||
device->tex_free(dscene->tri_vindex);
|
||||
device->tex_free(dscene->tri_verts);
|
||||
device->tex_free(dscene->curves);
|
||||
device->tex_free(dscene->curve_keys);
|
||||
device->tex_free(dscene->attributes_map);
|
||||
|
|
@ -1423,7 +1474,8 @@ void MeshManager::device_free(Device *device, DeviceScene *dscene)
|
|||
|
||||
dscene->bvh_nodes.clear();
|
||||
dscene->object_node.clear();
|
||||
dscene->tri_storage.clear();
|
||||
dscene->prim_tri_verts.clear();
|
||||
dscene->prim_tri_index.clear();
|
||||
dscene->prim_type.clear();
|
||||
dscene->prim_visibility.clear();
|
||||
dscene->prim_index.clear();
|
||||
|
|
@ -1431,7 +1483,6 @@ void MeshManager::device_free(Device *device, DeviceScene *dscene)
|
|||
dscene->tri_shader.clear();
|
||||
dscene->tri_vnormal.clear();
|
||||
dscene->tri_vindex.clear();
|
||||
dscene->tri_verts.clear();
|
||||
dscene->curves.clear();
|
||||
dscene->curve_keys.clear();
|
||||
dscene->attributes_map.clear();
|
||||
|
|
|
|||
|
|
@ -167,7 +167,10 @@ public:
|
|||
void add_vertex_normals();
|
||||
|
||||
void pack_normals(Scene *scene, uint *shader, float4 *vnormal);
|
||||
void pack_verts(float4 *tri_verts, float4 *tri_vindex, size_t vert_offset);
|
||||
void pack_verts(const vector<uint>& tri_prim_index,
|
||||
uint4 *tri_vindex,
|
||||
size_t vert_offset,
|
||||
size_t tri_offset);
|
||||
void pack_curves(Scene *scene, float4 *curve_key_co, float4 *curve_data, size_t curvekey_offset);
|
||||
void compute_bvh(SceneParams *params, Progress *progress, int n, int total);
|
||||
|
||||
|
|
@ -213,15 +216,41 @@ public:
|
|||
void update_svm_attributes(Device *device, DeviceScene *dscene, Scene *scene, vector<AttributeRequestSet>& mesh_attributes);
|
||||
|
||||
void device_update(Device *device, DeviceScene *dscene, Scene *scene, Progress& progress);
|
||||
void device_update_object(Device *device, DeviceScene *dscene, Scene *scene, Progress& progress);
|
||||
void device_update_mesh(Device *device, DeviceScene *dscene, Scene *scene, Progress& progress);
|
||||
void device_update_attributes(Device *device, DeviceScene *dscene, Scene *scene, Progress& progress);
|
||||
void device_update_bvh(Device *device, DeviceScene *dscene, Scene *scene, Progress& progress);
|
||||
void device_update_flags(Device *device, DeviceScene *dscene, Scene *scene, Progress& progress);
|
||||
void device_update_displacement_images(Device *device, DeviceScene *dscene, Scene *scene, Progress& progress);
|
||||
|
||||
void device_free(Device *device, DeviceScene *dscene);
|
||||
|
||||
void tag_update(Scene *scene);
|
||||
|
||||
protected:
|
||||
/* Calculate verts/triangles/curves offsets in global arrays. */
|
||||
void mesh_calc_offset(Scene *scene);
|
||||
|
||||
void device_update_object(Device *device,
|
||||
DeviceScene *dscene,
|
||||
Scene *scene,
|
||||
Progress& progress);
|
||||
|
||||
void device_update_mesh(Device *device,
|
||||
DeviceScene *dscene,
|
||||
Scene *scene,
|
||||
bool for_displacement,
|
||||
Progress& progress);
|
||||
|
||||
void device_update_attributes(Device *device,
|
||||
DeviceScene *dscene,
|
||||
Scene *scene,
|
||||
Progress& progress);
|
||||
|
||||
void device_update_bvh(Device *device,
|
||||
DeviceScene *dscene,
|
||||
Scene *scene,
|
||||
Progress& progress);
|
||||
|
||||
void device_update_displacement_images(Device *device,
|
||||
DeviceScene *dscene,
|
||||
Scene *scene,
|
||||
Progress& progress);
|
||||
};
|
||||
|
||||
CCL_NAMESPACE_END
|
||||
|
|
|
|||
|
|
@ -63,7 +63,8 @@ public:
|
|||
device_vector<float4> bvh_nodes;
|
||||
device_vector<float4> bvh_leaf_nodes;
|
||||
device_vector<uint> object_node;
|
||||
device_vector<float4> tri_storage;
|
||||
device_vector<uint> prim_tri_index;
|
||||
device_vector<float4> prim_tri_verts;
|
||||
device_vector<uint> prim_type;
|
||||
device_vector<uint> prim_visibility;
|
||||
device_vector<uint> prim_index;
|
||||
|
|
@ -72,8 +73,7 @@ public:
|
|||
/* mesh */
|
||||
device_vector<uint> tri_shader;
|
||||
device_vector<float4> tri_vnormal;
|
||||
device_vector<float4> tri_vindex;
|
||||
device_vector<float4> tri_verts;
|
||||
device_vector<uint4> tri_vindex;
|
||||
|
||||
device_vector<float4> curves;
|
||||
device_vector<float4> curve_keys;
|
||||
|
|
|
|||
Loading…
Reference in New Issue