[Bf-blender-cvs] [17e7454] master: Cycles: Reduce memory usage by de-duplicating triangle storage
Sergey Sharybin
noreply at git.blender.org
Thu Jul 7 17:29:07 CEST 2016
Commit: 17e745426375fe76f5011992d582c01b682d348b
Author: Sergey Sharybin
Date: Fri Jun 10 16:13:50 2016 +0200
Branches: master
https://developer.blender.org/rB17e745426375fe76f5011992d582c01b682d348b
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.
===================================================================
M intern/cycles/bvh/bvh.cpp
M intern/cycles/bvh/bvh.h
M intern/cycles/kernel/geom/geom_motion_triangle.h
M intern/cycles/kernel/geom/geom_triangle.h
M intern/cycles/kernel/geom/geom_triangle_intersect.h
M intern/cycles/kernel/kernel_textures.h
M intern/cycles/render/mesh.cpp
M intern/cycles/render/mesh.h
M intern/cycles/render/scene.h
===================================================================
diff --git a/intern/cycles/bvh/bvh.cpp b/intern/cycles/bvh/bvh.cpp
index 311d729..5669bab 100644
--- a/intern/cycles/bvh/bvh.cpp
+++ b/intern/cycles/bvh/bvh.cpp
@@ -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];
+ int tob = pack.prim_object[i];
+ Object *ob = objects[tob];
- if(pack.prim_type[i] & PRIMITIVE_TRIANGLE) {
- pack_triangle(i, storage);
+ 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 {
- /* Avoid use of uninitialized memory. */
- memset(&storage, 0, sizeof(storage));
+ pack.prim_tri_index[i] = -1;
}
- memcpy(&pack.tri_storage[i * nsize], storage, sizeof(float4)*3);
-
- int tob = pack.prim_object[i];
- Object *ob = objects[tob];
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 */
diff --git a/intern/cycles/bvh/bvh.h b/intern/cycles/bvh/bvh.h
index b1655f0..da7c9f3 100644
--- a/intern/cycles/bvh/bvh.h
+++ b/intern/cycles/bvh/bvh.h
@@ -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]);
diff --git a/intern/cycles/kernel/geom/geom_motion_triangle.h b/intern/cycles/kernel/geom/geom_motion_triangle.h
index ffe5552..2fb8e21 100644
--- a/intern/cycles/kernel/geom/geom_motion_triangle.h
+++ b/intern/cycles/kernel/geom/geom_motion_triangle.h
@@ -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));
}
}
@@ Diff output truncated at 10240 characters. @@
More information about the Bf-blender-cvs
mailing list