Files
test/source/blender/blenkernel/intern/geometry_component_mesh.cc

Ignoring revisions in .git-blame-ignore-revs. Click here to bypass and see the normal blame view.

1316 lines
46 KiB
C++
Raw Normal View History

/* SPDX-License-Identifier: GPL-2.0-or-later */
#include "BLI_listbase.h"
#include "BLI_task.hh"
#include "DNA_mesh_types.h"
#include "DNA_meshdata_types.h"
#include "DNA_object_types.h"
#include "BKE_attribute_math.hh"
#include "BKE_deform.h"
#include "BKE_geometry_fields.hh"
#include "BKE_geometry_set.hh"
#include "BKE_lib_id.h"
#include "BKE_mesh.hh"
#include "BKE_mesh_mapping.h"
#include "FN_multi_function_builder.hh"
#include "attribute_access_intern.hh"
extern "C" MDeformVert *BKE_object_defgroup_data_create(ID *id);
/* -------------------------------------------------------------------- */
/** \name Geometry Component Implementation
* \{ */
MeshComponent::MeshComponent() : GeometryComponent(GEO_COMPONENT_TYPE_MESH) {}
MeshComponent::~MeshComponent()
{
this->clear();
}
GeometryComponent *MeshComponent::copy() const
{
MeshComponent *new_component = new MeshComponent();
if (mesh_ != nullptr) {
new_component->mesh_ = BKE_mesh_copy_for_eval(mesh_);
new_component->ownership_ = GeometryOwnershipType::Owned;
}
return new_component;
}
void MeshComponent::clear()
{
BLI: support weak users and version in implicit sharing info The main goal of these changes is to support checking if some data has been changed over time. This is used by the WIP simulation nodes during baking to detect which attributes have to be stored in every frame because they have changed. By using a combination of a weak user count and a version counter, it is possible to detect that an attribute (or any data controlled by implicit sharing) has not been changed with O(1) memory and time. It's still possible that the data has been changed multiple times and is the same in the end and beginning of course. That wouldn't be detected using this mechanism. The `ImplicitSharingInfo` struct has a new weak user count. A weak reference is one that does not keep the referenced data alive, but makes sure that the `ImplicitSharingInfo` itself is not deleted. If some piece of data has one strong and multiple weak users, it is still mutable. If the strong user count goes down to zero, the referenced data is freed. Remaining weak users can check for this condition using `is_expired`. This is a bit similar to `std::weak_ptr` but there is an important difference: a weak user can not become a strong user while one can create a `shared_ptr` from a `weak_ptr`. This restriction is necessary, because some code might be changing the referenced data assuming that it is the only owner. If another thread suddenly adds a new owner, the data would be shared again and the first thread would not have been allowed to modify the data in the first place. There is also a new integer version counter in `ImplicitSharingInfo`. It is incremented whenever some code wants to modify the referenced data. Obviously, this can only be done when the data is not shared because then it would be immutable. By comparing an old and new version number of the same sharing info, one can check if the data has been modified. One has to keep a weak reference to the sharing info together with the old version number to ensure that the new sharing info is still the same as the old one. Without this, it can happen that the sharing info was freed and a new one was allocated at the same pointer address. Using a strong reference for this purpose does not work, because then the data would never be modified because it's shared.
2023-04-28 12:03:42 +02:00
BLI_assert(this->is_mutable() || this->is_expired());
if (mesh_ != nullptr) {
if (ownership_ == GeometryOwnershipType::Owned) {
BKE_id_free(nullptr, mesh_);
}
mesh_ = nullptr;
}
}
bool MeshComponent::has_mesh() const
{
return mesh_ != nullptr;
}
void MeshComponent::replace(Mesh *mesh, GeometryOwnershipType ownership)
{
BLI_assert(this->is_mutable());
this->clear();
mesh_ = mesh;
ownership_ = ownership;
}
Mesh *MeshComponent::release()
{
BLI_assert(this->is_mutable());
Mesh *mesh = mesh_;
mesh_ = nullptr;
return mesh;
}
const Mesh *MeshComponent::get_for_read() const
{
return mesh_;
}
Mesh *MeshComponent::get_for_write()
{
BLI_assert(this->is_mutable());
if (ownership_ == GeometryOwnershipType::ReadOnly) {
mesh_ = BKE_mesh_copy_for_eval(mesh_);
ownership_ = GeometryOwnershipType::Owned;
}
return mesh_;
}
bool MeshComponent::is_empty() const
{
return mesh_ == nullptr;
}
bool MeshComponent::owns_direct_data() const
{
return ownership_ == GeometryOwnershipType::Owned;
}
void MeshComponent::ensure_owns_direct_data()
{
BLI_assert(this->is_mutable());
if (ownership_ != GeometryOwnershipType::Owned) {
mesh_ = BKE_mesh_copy_for_eval(mesh_);
ownership_ = GeometryOwnershipType::Owned;
}
}
/** \} */
/* -------------------------------------------------------------------- */
/** \name Mesh Normals Field Input
* \{ */
namespace blender::bke {
VArray<float3> mesh_normals_varray(const Mesh &mesh,
const IndexMask mask,
const eAttrDomain domain)
{
switch (domain) {
case ATTR_DOMAIN_FACE: {
return VArray<float3>::ForSpan(mesh.poly_normals());
}
case ATTR_DOMAIN_POINT: {
return VArray<float3>::ForSpan(mesh.vert_normals());
}
case ATTR_DOMAIN_EDGE: {
/* In this case, start with vertex normals and convert to the edge domain, since the
Refactor: Move normals out of MVert, lazy calculation As described in T91186, this commit moves mesh vertex normals into a contiguous array of float vectors in a custom data layer, how face normals are currently stored. The main interface is documented in `BKE_mesh.h`. Vertex and face normals are now calculated on-demand and cached, retrieved with an "ensure" function. Since the logical state of a mesh is now "has normals when necessary", they can be retrieved from a `const` mesh. The goal is to use on-demand calculation for all derived data, but leave room for eager calculation for performance purposes (modifier evaluation is threaded, but viewport data generation is not). **Benefits** This moves us closer to a SoA approach rather than the current AoS paradigm. Accessing a contiguous `float3` is much more efficient than retrieving data from a larger struct. The memory requirements for accessing only normals or vertex locations are smaller, and at the cost of more memory usage for just normals, they now don't have to be converted between float and short, which also simplifies code In the future, the remaining items can be removed from `MVert`, leaving only `float3`, which has similar benefits (see T93602). Removing the combination of derived and original data makes it conceptually simpler to only calculate normals when necessary. This is especially important now that we have more opportunities for temporary meshes in geometry nodes. **Performance** In addition to the theoretical future performance improvements by making `MVert == float3`, I've done some basic performance testing on this patch directly. The data is fairly rough, but it gives an idea about where things stand generally. - Mesh line primitive 4m Verts: 1.16x faster (36 -> 31 ms), showing that accessing just `MVert` is now more efficient. - Spring Splash Screen: 1.03-1.06 -> 1.06-1.11 FPS, a very slight change that at least shows there is no regression. - Sprite Fright Snail Smoosh: 3.30-3.40 -> 3.42-3.50 FPS, a small but observable speedup. - Set Position Node with Scaled Normal: 1.36x faster (53 -> 39 ms), shows that using normals in geometry nodes is faster. - Normal Calculation 1.6m Vert Cube: 1.19x faster (25 -> 21 ms), shows that calculating normals is slightly faster now. - File Size of 1.6m Vert Cube: 1.03x smaller (214.7 -> 208.4 MB), Normals are not saved in files, which can help with large meshes. As for memory usage, it may be slightly more in some cases, but I didn't observe any difference in the production files I tested. **Tests** Some modifiers and cycles test results need to be updated with this commit, for two reasons: - The subdivision surface modifier is not responsible for calculating normals anymore. In master, the modifier creates different normals than the result of the `Mesh` normal calculation, so this is a bug fix. - There are small differences in the results of some modifiers that use normals because they are not converted to and from `short` anymore. **Future improvements** - Remove `ModifierTypeInfo::dependsOnNormals`. Code in each modifier already retrieves normals if they are needed anyway. - Copy normals as part of a better CoW system for attributes. - Make more areas use lazy instead of eager normal calculation. - Remove `BKE_mesh_normals_tag_dirty` in more places since that is now the default state of a new mesh. - Possibly apply a similar change to derived face corner normals. Differential Revision: https://developer.blender.org/D12770
2022-01-13 14:37:58 -06:00
* conversion from edges to vertices is very simple. Use "manual" domain interpolation
* instead of the GeometryComponent API to avoid calculating unnecessary values and to
* allow normalizing the result more simply. */
Span<float3> vert_normals = mesh.vert_normals();
Mesh: Move edges to a generic attribute Implements #95966, as the final step of #95965. This commit changes the storage of mesh edge vertex indices from the `MEdge` type to the generic `int2` attribute type. This follows the general design for geometry and the attribute system, where the data storage type and the usage semantics are separated. The main benefit of the change is reduced memory usage-- the requirements of storing mesh edges is reduced by 1/3. For example, this saves 8MB on a 1 million vertex grid. This also gives performance benefits to any memory-bound mesh processing algorithm that uses edges. Another benefit is that all of the edge's vertex indices are contiguous. In a few cases, it's helpful to process all of them as `Span<int>` rather than `Span<int2>`. Similarly, the type is more likely to match a generic format used by a library, or code that shouldn't know about specific Blender `Mesh` types. Various Notes: - The `.edge_verts` name is used to reflect a mapping between domains, similar to `.corner_verts`, etc. The period means that it the data shouldn't change arbitrarily by the user or procedural operations. - `edge[0]` is now used instead of `edge.v1` - Signed integers are used instead of unsigned to reduce the mixing of signed-ness, which can be error prone. - All of the previously used core mesh data types (`MVert`, `MEdge`, `MLoop`, `MPoly` are now deprecated. Only generic types are used). - The `vec2i` DNA type is used in the few C files where necessary. Pull Request: https://projects.blender.org/blender/blender/pulls/106638
2023-04-17 13:47:41 +02:00
const Span<int2> edges = mesh.edges();
Array<float3> edge_normals(mask.min_array_size());
for (const int i : mask) {
Mesh: Move edges to a generic attribute Implements #95966, as the final step of #95965. This commit changes the storage of mesh edge vertex indices from the `MEdge` type to the generic `int2` attribute type. This follows the general design for geometry and the attribute system, where the data storage type and the usage semantics are separated. The main benefit of the change is reduced memory usage-- the requirements of storing mesh edges is reduced by 1/3. For example, this saves 8MB on a 1 million vertex grid. This also gives performance benefits to any memory-bound mesh processing algorithm that uses edges. Another benefit is that all of the edge's vertex indices are contiguous. In a few cases, it's helpful to process all of them as `Span<int>` rather than `Span<int2>`. Similarly, the type is more likely to match a generic format used by a library, or code that shouldn't know about specific Blender `Mesh` types. Various Notes: - The `.edge_verts` name is used to reflect a mapping between domains, similar to `.corner_verts`, etc. The period means that it the data shouldn't change arbitrarily by the user or procedural operations. - `edge[0]` is now used instead of `edge.v1` - Signed integers are used instead of unsigned to reduce the mixing of signed-ness, which can be error prone. - All of the previously used core mesh data types (`MVert`, `MEdge`, `MLoop`, `MPoly` are now deprecated. Only generic types are used). - The `vec2i` DNA type is used in the few C files where necessary. Pull Request: https://projects.blender.org/blender/blender/pulls/106638
2023-04-17 13:47:41 +02:00
const int2 &edge = edges[i];
BLI: Refactor vector types & functions to use templates This patch implements the vector types (i.e:`float2`) by making heavy usage of templating. All vector functions are now outside of the vector classes (inside the `blender::math` namespace) and are not vector size dependent for the most part. In the ongoing effort to make shaders less GL centric, we are aiming to share more code between GLSL and C++ to avoid code duplication. ####Motivations: - We are aiming to share UBO and SSBO structures between GLSL and C++. This means we will use many of the existing vector types and others we currently don't have (uintX, intX). All these variations were asking for many more code duplication. - Deduplicate existing code which is duplicated for each vector size. - We also want to share small functions. Which means that vector functions should be static and not in the class namespace. - Reduce friction to use these types in new projects due to their incompleteness. - The current state of the `BLI_(float|double|mpq)(2|3|4).hh` is a bit of a let down. Most clases are incomplete, out of sync with each others with different codestyles, and some functions that should be static are not (i.e: `float3::reflect()`). ####Upsides: - Still support `.x, .y, .z, .w` for readability. - Compact, readable and easilly extendable. - All of the vector functions are available for all the vectors types and can be restricted to certain types. Also template specialization let us define exception for special class (like mpq). - With optimization ON, the compiler unroll the loops and performance is the same. ####Downsides: - Might impact debugability. Though I would arge that the bugs are rarelly caused by the vector class itself (since the operations are quite trivial) but by the type conversions. - Might impact compile time. I did not saw a significant impact since the usage is not really widespread. - Functions needs to be rewritten to support arbitrary vector length. For instance, one can't call `len_squared_v3v3` in `math::length_squared()` and call it a day. - Type cast does not work with the template version of the `math::` vector functions. Meaning you need to manually cast `float *` and `(float *)[3]` to `float3` for the function calls. i.e: `math::distance_squared(float3(nearest.co), positions[i]);` - Some parts might loose in readability: `float3::dot(v1.normalized(), v2.normalized())` becoming `math::dot(math::normalize(v1), math::normalize(v2))` But I propose, when appropriate, to use `using namespace blender::math;` on function local or file scope to increase readability. `dot(normalize(v1), normalize(v2))` ####Consideration: - Include back `.length()` method. It is quite handy and is more C++ oriented. - I considered the GLM library as a candidate for replacement. It felt like too much for what we need and would be difficult to extend / modify to our needs. - I used Macros to reduce code in operators declaration and potential copy paste bugs. This could reduce debugability and could be reverted. - This touches `delaunay_2d.cc` and the intersection code. I would like to know @howardt opinion on the matter. - The `noexcept` on the copy constructor of `mpq(2|3)` is being removed. But according to @JacquesLucke it is not a real problem for now. I would like to give a huge thanks to @JacquesLucke who helped during this and pushed me to reduce the duplication further. Reviewed By: brecht, sergey, JacquesLucke Differential Revision: https://developer.blender.org/D13791
2022-01-12 12:57:07 +01:00
edge_normals[i] = math::normalize(
Mesh: Move edges to a generic attribute Implements #95966, as the final step of #95965. This commit changes the storage of mesh edge vertex indices from the `MEdge` type to the generic `int2` attribute type. This follows the general design for geometry and the attribute system, where the data storage type and the usage semantics are separated. The main benefit of the change is reduced memory usage-- the requirements of storing mesh edges is reduced by 1/3. For example, this saves 8MB on a 1 million vertex grid. This also gives performance benefits to any memory-bound mesh processing algorithm that uses edges. Another benefit is that all of the edge's vertex indices are contiguous. In a few cases, it's helpful to process all of them as `Span<int>` rather than `Span<int2>`. Similarly, the type is more likely to match a generic format used by a library, or code that shouldn't know about specific Blender `Mesh` types. Various Notes: - The `.edge_verts` name is used to reflect a mapping between domains, similar to `.corner_verts`, etc. The period means that it the data shouldn't change arbitrarily by the user or procedural operations. - `edge[0]` is now used instead of `edge.v1` - Signed integers are used instead of unsigned to reduce the mixing of signed-ness, which can be error prone. - All of the previously used core mesh data types (`MVert`, `MEdge`, `MLoop`, `MPoly` are now deprecated. Only generic types are used). - The `vec2i` DNA type is used in the few C files where necessary. Pull Request: https://projects.blender.org/blender/blender/pulls/106638
2023-04-17 13:47:41 +02:00
math::interpolate(vert_normals[edge[0]], vert_normals[edge[1]], 0.5f));
}
return VArray<float3>::ForContainer(std::move(edge_normals));
}
case ATTR_DOMAIN_CORNER: {
/* The normals on corners are just the mesh's face normals, so start with the face normal
Refactor: Move normals out of MVert, lazy calculation As described in T91186, this commit moves mesh vertex normals into a contiguous array of float vectors in a custom data layer, how face normals are currently stored. The main interface is documented in `BKE_mesh.h`. Vertex and face normals are now calculated on-demand and cached, retrieved with an "ensure" function. Since the logical state of a mesh is now "has normals when necessary", they can be retrieved from a `const` mesh. The goal is to use on-demand calculation for all derived data, but leave room for eager calculation for performance purposes (modifier evaluation is threaded, but viewport data generation is not). **Benefits** This moves us closer to a SoA approach rather than the current AoS paradigm. Accessing a contiguous `float3` is much more efficient than retrieving data from a larger struct. The memory requirements for accessing only normals or vertex locations are smaller, and at the cost of more memory usage for just normals, they now don't have to be converted between float and short, which also simplifies code In the future, the remaining items can be removed from `MVert`, leaving only `float3`, which has similar benefits (see T93602). Removing the combination of derived and original data makes it conceptually simpler to only calculate normals when necessary. This is especially important now that we have more opportunities for temporary meshes in geometry nodes. **Performance** In addition to the theoretical future performance improvements by making `MVert == float3`, I've done some basic performance testing on this patch directly. The data is fairly rough, but it gives an idea about where things stand generally. - Mesh line primitive 4m Verts: 1.16x faster (36 -> 31 ms), showing that accessing just `MVert` is now more efficient. - Spring Splash Screen: 1.03-1.06 -> 1.06-1.11 FPS, a very slight change that at least shows there is no regression. - Sprite Fright Snail Smoosh: 3.30-3.40 -> 3.42-3.50 FPS, a small but observable speedup. - Set Position Node with Scaled Normal: 1.36x faster (53 -> 39 ms), shows that using normals in geometry nodes is faster. - Normal Calculation 1.6m Vert Cube: 1.19x faster (25 -> 21 ms), shows that calculating normals is slightly faster now. - File Size of 1.6m Vert Cube: 1.03x smaller (214.7 -> 208.4 MB), Normals are not saved in files, which can help with large meshes. As for memory usage, it may be slightly more in some cases, but I didn't observe any difference in the production files I tested. **Tests** Some modifiers and cycles test results need to be updated with this commit, for two reasons: - The subdivision surface modifier is not responsible for calculating normals anymore. In master, the modifier creates different normals than the result of the `Mesh` normal calculation, so this is a bug fix. - There are small differences in the results of some modifiers that use normals because they are not converted to and from `short` anymore. **Future improvements** - Remove `ModifierTypeInfo::dependsOnNormals`. Code in each modifier already retrieves normals if they are needed anyway. - Copy normals as part of a better CoW system for attributes. - Make more areas use lazy instead of eager normal calculation. - Remove `BKE_mesh_normals_tag_dirty` in more places since that is now the default state of a new mesh. - Possibly apply a similar change to derived face corner normals. Differential Revision: https://developer.blender.org/D12770
2022-01-13 14:37:58 -06:00
* array and copy the face normal for each of its corners. In this case using the mesh
* component's generic domain interpolation is fine, the data will still be normalized,
* since the face normal is just copied to every corner. */
return mesh.attributes().adapt_domain(
VArray<float3>::ForSpan(mesh.poly_normals()), ATTR_DOMAIN_FACE, ATTR_DOMAIN_CORNER);
}
default:
return {};
}
}
} // namespace blender::bke
/** \} */
/* -------------------------------------------------------------------- */
/** \name Attribute Access
* \{ */
namespace blender::bke {
template<typename T>
static void adapt_mesh_domain_corner_to_point_impl(const Mesh &mesh,
Geometry Nodes: use virtual arrays in internal attribute api A virtual array is a data structure that is similar to a normal array in that its elements can be accessed by an index. However, a virtual array does not have to be a contiguous array internally. Instead, its elements can be layed out arbitrarily while element access happens through a virtual function call. However, the virtual array data structures are designed so that the virtual function call can be avoided in cases where it could become a bottleneck. Most commonly, a virtual array is backed by an actual array/span or is a single value internally, that is the same for every index. Besides those, there are many more specialized virtual arrays like the ones that provides vertex positions based on the `MVert` struct or vertex group weights. Not all attributes used by geometry nodes are stored in simple contiguous arrays. To provide uniform access to all kinds of attributes, the attribute API has to provide virtual array functionality that hides the implementation details of attributes. Before this refactor, the attribute API provided its own virtual array implementation as part of the `ReadAttribute` and `WriteAttribute` types. That resulted in unnecessary code duplication with the virtual array system. Even worse, it bound many algorithms used by geometry nodes to the specifics of the attribute API, even though they could also use different data sources (such as data from sockets, default values, later results of expressions, ...). This refactor removes the `ReadAttribute` and `WriteAttribute` types and replaces them with `GVArray` and `GVMutableArray` respectively. The `GV` stands for "generic virtual". The "generic" means that the data type contained in those virtual arrays is only known at run-time. There are the corresponding statically typed types `VArray<T>` and `VMutableArray<T>` as well. No regressions are expected from this refactor. It does come with one improvement for users. The attribute API can convert the data type on write now. This is especially useful when writing to builtin attributes like `material_index` with e.g. the Attribute Math node (which usually just writes to float attributes, while `material_index` is an integer attribute). Differential Revision: https://developer.blender.org/D10994
2021-04-17 16:41:03 +02:00
const VArray<T> &old_values,
MutableSpan<T> r_values)
{
BLI_assert(r_values.size() == mesh.totvert);
Mesh: Replace MLoop struct with generic attributes Implements #102359. Split the `MLoop` struct into two separate integer arrays called `corner_verts` and `corner_edges`, referring to the vertex each corner is attached to and the next edge around the face at each corner. These arrays can be sliced to give access to the edges or vertices in a face. Then they are often referred to as "poly_verts" or "poly_edges". The main benefits are halving the necessary memory bandwidth when only one array is used and simplifications from using regular integer indices instead of a special-purpose struct. The commit also starts a renaming from "loop" to "corner" in mesh code. Like the other mesh struct of array refactors, forward compatibility is kept by writing files with the older format. This will be done until 4.0 to ease the transition process. Looking at a small portion of the patch should give a good impression for the rest of the changes. I tried to make the changes as small as possible so it's easy to tell the correctness from the diff. Though I found Blender developers have been very inventive over the last decade when finding different ways to loop over the corners in a face. For performance, nearly every piece of code that deals with `Mesh` is slightly impacted. Any algorithm that is memory bottle-necked should see an improvement. For example, here is a comparison of interpolating a vertex float attribute to face corners (Ryzen 3700x): **Before** (Average: 3.7 ms, Min: 3.4 ms) ``` threading::parallel_for(loops.index_range(), 4096, [&](IndexRange range) { for (const int64_t i : range) { dst[i] = src[loops[i].v]; } }); ``` **After** (Average: 2.9 ms, Min: 2.6 ms) ``` array_utils::gather(src, corner_verts, dst); ``` That's an improvement of 28% to the average timings, and it's also a simplification, since an index-based routine can be used instead. For more examples using the new arrays, see the design task. Pull Request: https://projects.blender.org/blender/blender/pulls/104424
2023-03-20 15:55:13 +01:00
const Span<int> corner_verts = mesh.corner_verts();
Mesh: Remove redundant custom data pointers For copy-on-write, we want to share attribute arrays between meshes where possible. Mutable pointers like `Mesh.mvert` make that difficult by making ownership vague. They also make code more complex by adding redundancy. The simplest solution is just removing them and retrieving layers from `CustomData` as needed. Similar changes have already been applied to curves and point clouds (e9f82d3dc7ee, 410a6efb747f). Removing use of the pointers generally makes code more obvious and more reusable. Mesh data is now accessed with a C++ API (`Mesh::edges()` or `Mesh::edges_for_write()`), and a C API (`BKE_mesh_edges(mesh)`). The CoW changes this commit makes possible are described in T95845 and T95842, and started in D14139 and D14140. The change also simplifies the ongoing mesh struct-of-array refactors from T95965. **RNA/Python Access Performance** Theoretically, accessing mesh elements with the RNA API may become slower, since the layer needs to be found on every random access. However, overhead is already high enough that this doesn't make a noticible differenc, and performance is actually improved in some cases. Random access can be up to 10% faster, but other situations might be a bit slower. Generally using `foreach_get/set` are the best way to improve performance. See the differential revision for more discussion about Python performance. Cycles has been updated to use raw pointers and the internal Blender mesh types, mostly because there is no sense in having this overhead when it's already compiled with Blender. In my tests this roughly halves the Cycles mesh creation time (0.19s to 0.10s for a 1 million face grid). Differential Revision: https://developer.blender.org/D15488
2022-09-05 11:56:34 -05:00
attribute_math::DefaultMixer<T> mixer(r_values);
Mesh: Replace MLoop struct with generic attributes Implements #102359. Split the `MLoop` struct into two separate integer arrays called `corner_verts` and `corner_edges`, referring to the vertex each corner is attached to and the next edge around the face at each corner. These arrays can be sliced to give access to the edges or vertices in a face. Then they are often referred to as "poly_verts" or "poly_edges". The main benefits are halving the necessary memory bandwidth when only one array is used and simplifications from using regular integer indices instead of a special-purpose struct. The commit also starts a renaming from "loop" to "corner" in mesh code. Like the other mesh struct of array refactors, forward compatibility is kept by writing files with the older format. This will be done until 4.0 to ease the transition process. Looking at a small portion of the patch should give a good impression for the rest of the changes. I tried to make the changes as small as possible so it's easy to tell the correctness from the diff. Though I found Blender developers have been very inventive over the last decade when finding different ways to loop over the corners in a face. For performance, nearly every piece of code that deals with `Mesh` is slightly impacted. Any algorithm that is memory bottle-necked should see an improvement. For example, here is a comparison of interpolating a vertex float attribute to face corners (Ryzen 3700x): **Before** (Average: 3.7 ms, Min: 3.4 ms) ``` threading::parallel_for(loops.index_range(), 4096, [&](IndexRange range) { for (const int64_t i : range) { dst[i] = src[loops[i].v]; } }); ``` **After** (Average: 2.9 ms, Min: 2.6 ms) ``` array_utils::gather(src, corner_verts, dst); ``` That's an improvement of 28% to the average timings, and it's also a simplification, since an index-based routine can be used instead. For more examples using the new arrays, see the design task. Pull Request: https://projects.blender.org/blender/blender/pulls/104424
2023-03-20 15:55:13 +01:00
for (const int corner : IndexRange(mesh.totloop)) {
mixer.mix_in(corner_verts[corner], old_values[corner]);
}
mixer.finalize();
}
/* A vertex is selected if all connected face corners were selected and it is not loose. */
template<>
void adapt_mesh_domain_corner_to_point_impl(const Mesh &mesh,
const VArray<bool> &old_values,
MutableSpan<bool> r_values)
{
BLI_assert(r_values.size() == mesh.totvert);
Mesh: Replace MLoop struct with generic attributes Implements #102359. Split the `MLoop` struct into two separate integer arrays called `corner_verts` and `corner_edges`, referring to the vertex each corner is attached to and the next edge around the face at each corner. These arrays can be sliced to give access to the edges or vertices in a face. Then they are often referred to as "poly_verts" or "poly_edges". The main benefits are halving the necessary memory bandwidth when only one array is used and simplifications from using regular integer indices instead of a special-purpose struct. The commit also starts a renaming from "loop" to "corner" in mesh code. Like the other mesh struct of array refactors, forward compatibility is kept by writing files with the older format. This will be done until 4.0 to ease the transition process. Looking at a small portion of the patch should give a good impression for the rest of the changes. I tried to make the changes as small as possible so it's easy to tell the correctness from the diff. Though I found Blender developers have been very inventive over the last decade when finding different ways to loop over the corners in a face. For performance, nearly every piece of code that deals with `Mesh` is slightly impacted. Any algorithm that is memory bottle-necked should see an improvement. For example, here is a comparison of interpolating a vertex float attribute to face corners (Ryzen 3700x): **Before** (Average: 3.7 ms, Min: 3.4 ms) ``` threading::parallel_for(loops.index_range(), 4096, [&](IndexRange range) { for (const int64_t i : range) { dst[i] = src[loops[i].v]; } }); ``` **After** (Average: 2.9 ms, Min: 2.6 ms) ``` array_utils::gather(src, corner_verts, dst); ``` That's an improvement of 28% to the average timings, and it's also a simplification, since an index-based routine can be used instead. For more examples using the new arrays, see the design task. Pull Request: https://projects.blender.org/blender/blender/pulls/104424
2023-03-20 15:55:13 +01:00
const Span<int> corner_verts = mesh.corner_verts();
Mesh: Remove redundant custom data pointers For copy-on-write, we want to share attribute arrays between meshes where possible. Mutable pointers like `Mesh.mvert` make that difficult by making ownership vague. They also make code more complex by adding redundancy. The simplest solution is just removing them and retrieving layers from `CustomData` as needed. Similar changes have already been applied to curves and point clouds (e9f82d3dc7ee, 410a6efb747f). Removing use of the pointers generally makes code more obvious and more reusable. Mesh data is now accessed with a C++ API (`Mesh::edges()` or `Mesh::edges_for_write()`), and a C API (`BKE_mesh_edges(mesh)`). The CoW changes this commit makes possible are described in T95845 and T95842, and started in D14139 and D14140. The change also simplifies the ongoing mesh struct-of-array refactors from T95965. **RNA/Python Access Performance** Theoretically, accessing mesh elements with the RNA API may become slower, since the layer needs to be found on every random access. However, overhead is already high enough that this doesn't make a noticible differenc, and performance is actually improved in some cases. Random access can be up to 10% faster, but other situations might be a bit slower. Generally using `foreach_get/set` are the best way to improve performance. See the differential revision for more discussion about Python performance. Cycles has been updated to use raw pointers and the internal Blender mesh types, mostly because there is no sense in having this overhead when it's already compiled with Blender. In my tests this roughly halves the Cycles mesh creation time (0.19s to 0.10s for a 1 million face grid). Differential Revision: https://developer.blender.org/D15488
2022-09-05 11:56:34 -05:00
r_values.fill(true);
Mesh: Replace MLoop struct with generic attributes Implements #102359. Split the `MLoop` struct into two separate integer arrays called `corner_verts` and `corner_edges`, referring to the vertex each corner is attached to and the next edge around the face at each corner. These arrays can be sliced to give access to the edges or vertices in a face. Then they are often referred to as "poly_verts" or "poly_edges". The main benefits are halving the necessary memory bandwidth when only one array is used and simplifications from using regular integer indices instead of a special-purpose struct. The commit also starts a renaming from "loop" to "corner" in mesh code. Like the other mesh struct of array refactors, forward compatibility is kept by writing files with the older format. This will be done until 4.0 to ease the transition process. Looking at a small portion of the patch should give a good impression for the rest of the changes. I tried to make the changes as small as possible so it's easy to tell the correctness from the diff. Though I found Blender developers have been very inventive over the last decade when finding different ways to loop over the corners in a face. For performance, nearly every piece of code that deals with `Mesh` is slightly impacted. Any algorithm that is memory bottle-necked should see an improvement. For example, here is a comparison of interpolating a vertex float attribute to face corners (Ryzen 3700x): **Before** (Average: 3.7 ms, Min: 3.4 ms) ``` threading::parallel_for(loops.index_range(), 4096, [&](IndexRange range) { for (const int64_t i : range) { dst[i] = src[loops[i].v]; } }); ``` **After** (Average: 2.9 ms, Min: 2.6 ms) ``` array_utils::gather(src, corner_verts, dst); ``` That's an improvement of 28% to the average timings, and it's also a simplification, since an index-based routine can be used instead. For more examples using the new arrays, see the design task. Pull Request: https://projects.blender.org/blender/blender/pulls/104424
2023-03-20 15:55:13 +01:00
for (const int corner : IndexRange(mesh.totloop)) {
const int point_index = corner_verts[corner];
Mesh: Replace MLoop struct with generic attributes Implements #102359. Split the `MLoop` struct into two separate integer arrays called `corner_verts` and `corner_edges`, referring to the vertex each corner is attached to and the next edge around the face at each corner. These arrays can be sliced to give access to the edges or vertices in a face. Then they are often referred to as "poly_verts" or "poly_edges". The main benefits are halving the necessary memory bandwidth when only one array is used and simplifications from using regular integer indices instead of a special-purpose struct. The commit also starts a renaming from "loop" to "corner" in mesh code. Like the other mesh struct of array refactors, forward compatibility is kept by writing files with the older format. This will be done until 4.0 to ease the transition process. Looking at a small portion of the patch should give a good impression for the rest of the changes. I tried to make the changes as small as possible so it's easy to tell the correctness from the diff. Though I found Blender developers have been very inventive over the last decade when finding different ways to loop over the corners in a face. For performance, nearly every piece of code that deals with `Mesh` is slightly impacted. Any algorithm that is memory bottle-necked should see an improvement. For example, here is a comparison of interpolating a vertex float attribute to face corners (Ryzen 3700x): **Before** (Average: 3.7 ms, Min: 3.4 ms) ``` threading::parallel_for(loops.index_range(), 4096, [&](IndexRange range) { for (const int64_t i : range) { dst[i] = src[loops[i].v]; } }); ``` **After** (Average: 2.9 ms, Min: 2.6 ms) ``` array_utils::gather(src, corner_verts, dst); ``` That's an improvement of 28% to the average timings, and it's also a simplification, since an index-based routine can be used instead. For more examples using the new arrays, see the design task. Pull Request: https://projects.blender.org/blender/blender/pulls/104424
2023-03-20 15:55:13 +01:00
if (!old_values[corner]) {
r_values[point_index] = false;
}
}
/* Deselect loose vertices without corners that are still selected from the 'true' default. */
const bke::LooseVertCache &loose_verts = mesh.verts_no_face();
if (loose_verts.count > 0) {
const BitSpan bits = loose_verts.is_loose_bits;
threading::parallel_for(bits.index_range(), 2048, [&](const IndexRange range) {
for (const int vert_index : range) {
if (bits[vert_index]) {
r_values[vert_index] = false;
}
}
});
}
}
Geometry Nodes: refactor virtual array system Goals of this refactor: * Simplify creating virtual arrays. * Simplify passing virtual arrays around. * Simplify converting between typed and generic virtual arrays. * Reduce memory allocations. As a quick reminder, a virtual arrays is a data structure that behaves like an array (i.e. it can be accessed using an index). However, it may not actually be stored as array internally. The two most important implementations of virtual arrays are those that correspond to an actual plain array and those that have the same value for every index. However, many more implementations exist for various reasons (interfacing with legacy attributes, unified iterator over all points in multiple splines, ...). With this refactor the core types (`VArray`, `GVArray`, `VMutableArray` and `GVMutableArray`) can be used like "normal values". They typically live on the stack. Before, they were usually inside a `std::unique_ptr`. This makes passing them around much easier. Creation of new virtual arrays is also much simpler now due to some constructors. Memory allocations are reduced by making use of small object optimization inside the core types. Previously, `VArray` was a class with virtual methods that had to be overridden to change the behavior of a the virtual array. Now,`VArray` has a fixed size and has no virtual methods. Instead it contains a `VArrayImpl` that is similar to the old `VArray`. `VArrayImpl` should rarely ever be used directly, unless a new virtual array implementation is added. To support the small object optimization for many `VArrayImpl` classes, a new `blender::Any` type is added. It is similar to `std::any` with two additional features. It has an adjustable inline buffer size and alignment. The inline buffer size of `std::any` can't be relied on and is usually too small for our use case here. Furthermore, `blender::Any` can store additional user-defined type information without increasing the stack size. Differential Revision: https://developer.blender.org/D12986
2021-11-16 10:15:51 +01:00
static GVArray adapt_mesh_domain_corner_to_point(const Mesh &mesh, const GVArray &varray)
{
GArray<> values(varray.type(), mesh.totvert);
Geometry Nodes: refactor virtual array system Goals of this refactor: * Simplify creating virtual arrays. * Simplify passing virtual arrays around. * Simplify converting between typed and generic virtual arrays. * Reduce memory allocations. As a quick reminder, a virtual arrays is a data structure that behaves like an array (i.e. it can be accessed using an index). However, it may not actually be stored as array internally. The two most important implementations of virtual arrays are those that correspond to an actual plain array and those that have the same value for every index. However, many more implementations exist for various reasons (interfacing with legacy attributes, unified iterator over all points in multiple splines, ...). With this refactor the core types (`VArray`, `GVArray`, `VMutableArray` and `GVMutableArray`) can be used like "normal values". They typically live on the stack. Before, they were usually inside a `std::unique_ptr`. This makes passing them around much easier. Creation of new virtual arrays is also much simpler now due to some constructors. Memory allocations are reduced by making use of small object optimization inside the core types. Previously, `VArray` was a class with virtual methods that had to be overridden to change the behavior of a the virtual array. Now,`VArray` has a fixed size and has no virtual methods. Instead it contains a `VArrayImpl` that is similar to the old `VArray`. `VArrayImpl` should rarely ever be used directly, unless a new virtual array implementation is added. To support the small object optimization for many `VArrayImpl` classes, a new `blender::Any` type is added. It is similar to `std::any` with two additional features. It has an adjustable inline buffer size and alignment. The inline buffer size of `std::any` can't be relied on and is usually too small for our use case here. Furthermore, `blender::Any` can store additional user-defined type information without increasing the stack size. Differential Revision: https://developer.blender.org/D12986
2021-11-16 10:15:51 +01:00
attribute_math::convert_to_static_type(varray.type(), [&](auto dummy) {
using T = decltype(dummy);
if constexpr (!std::is_void_v<attribute_math::DefaultMixer<T>>) {
/* We compute all interpolated values at once, because for this interpolation, one has to
* iterate over all loops anyway. */
adapt_mesh_domain_corner_to_point_impl<T>(
mesh, varray.typed<T>(), values.as_mutable_span().typed<T>());
}
});
return GVArray::ForGArray(std::move(values));
}
/**
* Each corner's value is simply a copy of the value at its vertex.
*/
Geometry Nodes: refactor virtual array system Goals of this refactor: * Simplify creating virtual arrays. * Simplify passing virtual arrays around. * Simplify converting between typed and generic virtual arrays. * Reduce memory allocations. As a quick reminder, a virtual arrays is a data structure that behaves like an array (i.e. it can be accessed using an index). However, it may not actually be stored as array internally. The two most important implementations of virtual arrays are those that correspond to an actual plain array and those that have the same value for every index. However, many more implementations exist for various reasons (interfacing with legacy attributes, unified iterator over all points in multiple splines, ...). With this refactor the core types (`VArray`, `GVArray`, `VMutableArray` and `GVMutableArray`) can be used like "normal values". They typically live on the stack. Before, they were usually inside a `std::unique_ptr`. This makes passing them around much easier. Creation of new virtual arrays is also much simpler now due to some constructors. Memory allocations are reduced by making use of small object optimization inside the core types. Previously, `VArray` was a class with virtual methods that had to be overridden to change the behavior of a the virtual array. Now,`VArray` has a fixed size and has no virtual methods. Instead it contains a `VArrayImpl` that is similar to the old `VArray`. `VArrayImpl` should rarely ever be used directly, unless a new virtual array implementation is added. To support the small object optimization for many `VArrayImpl` classes, a new `blender::Any` type is added. It is similar to `std::any` with two additional features. It has an adjustable inline buffer size and alignment. The inline buffer size of `std::any` can't be relied on and is usually too small for our use case here. Furthermore, `blender::Any` can store additional user-defined type information without increasing the stack size. Differential Revision: https://developer.blender.org/D12986
2021-11-16 10:15:51 +01:00
static GVArray adapt_mesh_domain_point_to_corner(const Mesh &mesh, const GVArray &varray)
{
Mesh: Replace MLoop struct with generic attributes Implements #102359. Split the `MLoop` struct into two separate integer arrays called `corner_verts` and `corner_edges`, referring to the vertex each corner is attached to and the next edge around the face at each corner. These arrays can be sliced to give access to the edges or vertices in a face. Then they are often referred to as "poly_verts" or "poly_edges". The main benefits are halving the necessary memory bandwidth when only one array is used and simplifications from using regular integer indices instead of a special-purpose struct. The commit also starts a renaming from "loop" to "corner" in mesh code. Like the other mesh struct of array refactors, forward compatibility is kept by writing files with the older format. This will be done until 4.0 to ease the transition process. Looking at a small portion of the patch should give a good impression for the rest of the changes. I tried to make the changes as small as possible so it's easy to tell the correctness from the diff. Though I found Blender developers have been very inventive over the last decade when finding different ways to loop over the corners in a face. For performance, nearly every piece of code that deals with `Mesh` is slightly impacted. Any algorithm that is memory bottle-necked should see an improvement. For example, here is a comparison of interpolating a vertex float attribute to face corners (Ryzen 3700x): **Before** (Average: 3.7 ms, Min: 3.4 ms) ``` threading::parallel_for(loops.index_range(), 4096, [&](IndexRange range) { for (const int64_t i : range) { dst[i] = src[loops[i].v]; } }); ``` **After** (Average: 2.9 ms, Min: 2.6 ms) ``` array_utils::gather(src, corner_verts, dst); ``` That's an improvement of 28% to the average timings, and it's also a simplification, since an index-based routine can be used instead. For more examples using the new arrays, see the design task. Pull Request: https://projects.blender.org/blender/blender/pulls/104424
2023-03-20 15:55:13 +01:00
const Span<int> corner_verts = mesh.corner_verts();
Mesh: Remove redundant custom data pointers For copy-on-write, we want to share attribute arrays between meshes where possible. Mutable pointers like `Mesh.mvert` make that difficult by making ownership vague. They also make code more complex by adding redundancy. The simplest solution is just removing them and retrieving layers from `CustomData` as needed. Similar changes have already been applied to curves and point clouds (e9f82d3dc7ee, 410a6efb747f). Removing use of the pointers generally makes code more obvious and more reusable. Mesh data is now accessed with a C++ API (`Mesh::edges()` or `Mesh::edges_for_write()`), and a C API (`BKE_mesh_edges(mesh)`). The CoW changes this commit makes possible are described in T95845 and T95842, and started in D14139 and D14140. The change also simplifies the ongoing mesh struct-of-array refactors from T95965. **RNA/Python Access Performance** Theoretically, accessing mesh elements with the RNA API may become slower, since the layer needs to be found on every random access. However, overhead is already high enough that this doesn't make a noticible differenc, and performance is actually improved in some cases. Random access can be up to 10% faster, but other situations might be a bit slower. Generally using `foreach_get/set` are the best way to improve performance. See the differential revision for more discussion about Python performance. Cycles has been updated to use raw pointers and the internal Blender mesh types, mostly because there is no sense in having this overhead when it's already compiled with Blender. In my tests this roughly halves the Cycles mesh creation time (0.19s to 0.10s for a 1 million face grid). Differential Revision: https://developer.blender.org/D15488
2022-09-05 11:56:34 -05:00
Geometry Nodes: refactor virtual array system Goals of this refactor: * Simplify creating virtual arrays. * Simplify passing virtual arrays around. * Simplify converting between typed and generic virtual arrays. * Reduce memory allocations. As a quick reminder, a virtual arrays is a data structure that behaves like an array (i.e. it can be accessed using an index). However, it may not actually be stored as array internally. The two most important implementations of virtual arrays are those that correspond to an actual plain array and those that have the same value for every index. However, many more implementations exist for various reasons (interfacing with legacy attributes, unified iterator over all points in multiple splines, ...). With this refactor the core types (`VArray`, `GVArray`, `VMutableArray` and `GVMutableArray`) can be used like "normal values". They typically live on the stack. Before, they were usually inside a `std::unique_ptr`. This makes passing them around much easier. Creation of new virtual arrays is also much simpler now due to some constructors. Memory allocations are reduced by making use of small object optimization inside the core types. Previously, `VArray` was a class with virtual methods that had to be overridden to change the behavior of a the virtual array. Now,`VArray` has a fixed size and has no virtual methods. Instead it contains a `VArrayImpl` that is similar to the old `VArray`. `VArrayImpl` should rarely ever be used directly, unless a new virtual array implementation is added. To support the small object optimization for many `VArrayImpl` classes, a new `blender::Any` type is added. It is similar to `std::any` with two additional features. It has an adjustable inline buffer size and alignment. The inline buffer size of `std::any` can't be relied on and is usually too small for our use case here. Furthermore, `blender::Any` can store additional user-defined type information without increasing the stack size. Differential Revision: https://developer.blender.org/D12986
2021-11-16 10:15:51 +01:00
GVArray new_varray;
attribute_math::convert_to_static_type(varray.type(), [&](auto dummy) {
using T = decltype(dummy);
Mesh: Replace MLoop struct with generic attributes Implements #102359. Split the `MLoop` struct into two separate integer arrays called `corner_verts` and `corner_edges`, referring to the vertex each corner is attached to and the next edge around the face at each corner. These arrays can be sliced to give access to the edges or vertices in a face. Then they are often referred to as "poly_verts" or "poly_edges". The main benefits are halving the necessary memory bandwidth when only one array is used and simplifications from using regular integer indices instead of a special-purpose struct. The commit also starts a renaming from "loop" to "corner" in mesh code. Like the other mesh struct of array refactors, forward compatibility is kept by writing files with the older format. This will be done until 4.0 to ease the transition process. Looking at a small portion of the patch should give a good impression for the rest of the changes. I tried to make the changes as small as possible so it's easy to tell the correctness from the diff. Though I found Blender developers have been very inventive over the last decade when finding different ways to loop over the corners in a face. For performance, nearly every piece of code that deals with `Mesh` is slightly impacted. Any algorithm that is memory bottle-necked should see an improvement. For example, here is a comparison of interpolating a vertex float attribute to face corners (Ryzen 3700x): **Before** (Average: 3.7 ms, Min: 3.4 ms) ``` threading::parallel_for(loops.index_range(), 4096, [&](IndexRange range) { for (const int64_t i : range) { dst[i] = src[loops[i].v]; } }); ``` **After** (Average: 2.9 ms, Min: 2.6 ms) ``` array_utils::gather(src, corner_verts, dst); ``` That's an improvement of 28% to the average timings, and it's also a simplification, since an index-based routine can be used instead. For more examples using the new arrays, see the design task. Pull Request: https://projects.blender.org/blender/blender/pulls/104424
2023-03-20 15:55:13 +01:00
new_varray = VArray<T>::ForFunc(
mesh.totloop, [corner_verts, varray = varray.typed<T>()](const int64_t corner) {
return varray[corner_verts[corner]];
});
});
Geometry Nodes: use virtual arrays in internal attribute api A virtual array is a data structure that is similar to a normal array in that its elements can be accessed by an index. However, a virtual array does not have to be a contiguous array internally. Instead, its elements can be layed out arbitrarily while element access happens through a virtual function call. However, the virtual array data structures are designed so that the virtual function call can be avoided in cases where it could become a bottleneck. Most commonly, a virtual array is backed by an actual array/span or is a single value internally, that is the same for every index. Besides those, there are many more specialized virtual arrays like the ones that provides vertex positions based on the `MVert` struct or vertex group weights. Not all attributes used by geometry nodes are stored in simple contiguous arrays. To provide uniform access to all kinds of attributes, the attribute API has to provide virtual array functionality that hides the implementation details of attributes. Before this refactor, the attribute API provided its own virtual array implementation as part of the `ReadAttribute` and `WriteAttribute` types. That resulted in unnecessary code duplication with the virtual array system. Even worse, it bound many algorithms used by geometry nodes to the specifics of the attribute API, even though they could also use different data sources (such as data from sockets, default values, later results of expressions, ...). This refactor removes the `ReadAttribute` and `WriteAttribute` types and replaces them with `GVArray` and `GVMutableArray` respectively. The `GV` stands for "generic virtual". The "generic" means that the data type contained in those virtual arrays is only known at run-time. There are the corresponding statically typed types `VArray<T>` and `VMutableArray<T>` as well. No regressions are expected from this refactor. It does come with one improvement for users. The attribute API can convert the data type on write now. This is especially useful when writing to builtin attributes like `material_index` with e.g. the Attribute Math node (which usually just writes to float attributes, while `material_index` is an integer attribute). Differential Revision: https://developer.blender.org/D10994
2021-04-17 16:41:03 +02:00
return new_varray;
}
Geometry Nodes: refactor virtual array system Goals of this refactor: * Simplify creating virtual arrays. * Simplify passing virtual arrays around. * Simplify converting between typed and generic virtual arrays. * Reduce memory allocations. As a quick reminder, a virtual arrays is a data structure that behaves like an array (i.e. it can be accessed using an index). However, it may not actually be stored as array internally. The two most important implementations of virtual arrays are those that correspond to an actual plain array and those that have the same value for every index. However, many more implementations exist for various reasons (interfacing with legacy attributes, unified iterator over all points in multiple splines, ...). With this refactor the core types (`VArray`, `GVArray`, `VMutableArray` and `GVMutableArray`) can be used like "normal values". They typically live on the stack. Before, they were usually inside a `std::unique_ptr`. This makes passing them around much easier. Creation of new virtual arrays is also much simpler now due to some constructors. Memory allocations are reduced by making use of small object optimization inside the core types. Previously, `VArray` was a class with virtual methods that had to be overridden to change the behavior of a the virtual array. Now,`VArray` has a fixed size and has no virtual methods. Instead it contains a `VArrayImpl` that is similar to the old `VArray`. `VArrayImpl` should rarely ever be used directly, unless a new virtual array implementation is added. To support the small object optimization for many `VArrayImpl` classes, a new `blender::Any` type is added. It is similar to `std::any` with two additional features. It has an adjustable inline buffer size and alignment. The inline buffer size of `std::any` can't be relied on and is usually too small for our use case here. Furthermore, `blender::Any` can store additional user-defined type information without increasing the stack size. Differential Revision: https://developer.blender.org/D12986
2021-11-16 10:15:51 +01:00
static GVArray adapt_mesh_domain_corner_to_face(const Mesh &mesh, const GVArray &varray)
{
Mesh: Replace MPoly struct with offset indices Implements #95967. Currently the `MPoly` struct is 12 bytes, and stores the index of a face's first corner and the number of corners/verts/edges. Polygons and corners are always created in order by Blender, meaning each face's corners will be after the previous face's corners. We can take advantage of this fact and eliminate the redundancy in mesh face storage by only storing a single integer corner offset for each face. The size of the face is then encoded by the offset of the next face. The size of a single integer is 4 bytes, so this reduces memory usage by 3 times. The same method is used for `CurvesGeometry`, so Blender already has an abstraction to simplify using these offsets called `OffsetIndices`. This class is used to easily retrieve a range of corner indices for each face. This also gives the opportunity for sharing some logic with curves. Another benefit of the change is that the offsets and sizes stored in `MPoly` can no longer disagree with each other. Storing faces in the order of their corners can simplify some code too. Face/polygon variables now use the `IndexRange` type, which comes with quite a few utilities that can simplify code. Some: - The offset integer array has to be one longer than the face count to avoid a branch for every face, which means the data is no longer part of the mesh's `CustomData`. - We lose the ability to "reference" an original mesh's offset array until more reusable CoW from #104478 is committed. That will be added in a separate commit. - Since they aren't part of `CustomData`, poly offsets often have to be copied manually. - To simplify using `OffsetIndices` in many places, some functions and structs in headers were moved to only compile in C++. - All meshes created by Blender use the same order for faces and face corners, but just in case, meshes with mismatched order are fixed by versioning code. - `MeshPolygon.totloop` is no longer editable in RNA. This API break is necessary here unfortunately. It should be worth it in 3.6, since that's the best way to allow loading meshes from 4.0, which is important for an LTS version. Pull Request: https://projects.blender.org/blender/blender/pulls/105938
2023-04-04 20:39:28 +02:00
const OffsetIndices polys = mesh.polys();
Mesh: Remove redundant custom data pointers For copy-on-write, we want to share attribute arrays between meshes where possible. Mutable pointers like `Mesh.mvert` make that difficult by making ownership vague. They also make code more complex by adding redundancy. The simplest solution is just removing them and retrieving layers from `CustomData` as needed. Similar changes have already been applied to curves and point clouds (e9f82d3dc7ee, 410a6efb747f). Removing use of the pointers generally makes code more obvious and more reusable. Mesh data is now accessed with a C++ API (`Mesh::edges()` or `Mesh::edges_for_write()`), and a C API (`BKE_mesh_edges(mesh)`). The CoW changes this commit makes possible are described in T95845 and T95842, and started in D14139 and D14140. The change also simplifies the ongoing mesh struct-of-array refactors from T95965. **RNA/Python Access Performance** Theoretically, accessing mesh elements with the RNA API may become slower, since the layer needs to be found on every random access. However, overhead is already high enough that this doesn't make a noticible differenc, and performance is actually improved in some cases. Random access can be up to 10% faster, but other situations might be a bit slower. Generally using `foreach_get/set` are the best way to improve performance. See the differential revision for more discussion about Python performance. Cycles has been updated to use raw pointers and the internal Blender mesh types, mostly because there is no sense in having this overhead when it's already compiled with Blender. In my tests this roughly halves the Cycles mesh creation time (0.19s to 0.10s for a 1 million face grid). Differential Revision: https://developer.blender.org/D15488
2022-09-05 11:56:34 -05:00
Geometry Nodes: refactor virtual array system Goals of this refactor: * Simplify creating virtual arrays. * Simplify passing virtual arrays around. * Simplify converting between typed and generic virtual arrays. * Reduce memory allocations. As a quick reminder, a virtual arrays is a data structure that behaves like an array (i.e. it can be accessed using an index). However, it may not actually be stored as array internally. The two most important implementations of virtual arrays are those that correspond to an actual plain array and those that have the same value for every index. However, many more implementations exist for various reasons (interfacing with legacy attributes, unified iterator over all points in multiple splines, ...). With this refactor the core types (`VArray`, `GVArray`, `VMutableArray` and `GVMutableArray`) can be used like "normal values". They typically live on the stack. Before, they were usually inside a `std::unique_ptr`. This makes passing them around much easier. Creation of new virtual arrays is also much simpler now due to some constructors. Memory allocations are reduced by making use of small object optimization inside the core types. Previously, `VArray` was a class with virtual methods that had to be overridden to change the behavior of a the virtual array. Now,`VArray` has a fixed size and has no virtual methods. Instead it contains a `VArrayImpl` that is similar to the old `VArray`. `VArrayImpl` should rarely ever be used directly, unless a new virtual array implementation is added. To support the small object optimization for many `VArrayImpl` classes, a new `blender::Any` type is added. It is similar to `std::any` with two additional features. It has an adjustable inline buffer size and alignment. The inline buffer size of `std::any` can't be relied on and is usually too small for our use case here. Furthermore, `blender::Any` can store additional user-defined type information without increasing the stack size. Differential Revision: https://developer.blender.org/D12986
2021-11-16 10:15:51 +01:00
GVArray new_varray;
attribute_math::convert_to_static_type(varray.type(), [&](auto dummy) {
using T = decltype(dummy);
if constexpr (!std::is_void_v<attribute_math::DefaultMixer<T>>) {
if constexpr (std::is_same_v<T, bool>) {
new_varray = VArray<T>::ForFunc(
Mesh: Remove redundant custom data pointers For copy-on-write, we want to share attribute arrays between meshes where possible. Mutable pointers like `Mesh.mvert` make that difficult by making ownership vague. They also make code more complex by adding redundancy. The simplest solution is just removing them and retrieving layers from `CustomData` as needed. Similar changes have already been applied to curves and point clouds (e9f82d3dc7ee, 410a6efb747f). Removing use of the pointers generally makes code more obvious and more reusable. Mesh data is now accessed with a C++ API (`Mesh::edges()` or `Mesh::edges_for_write()`), and a C API (`BKE_mesh_edges(mesh)`). The CoW changes this commit makes possible are described in T95845 and T95842, and started in D14139 and D14140. The change also simplifies the ongoing mesh struct-of-array refactors from T95965. **RNA/Python Access Performance** Theoretically, accessing mesh elements with the RNA API may become slower, since the layer needs to be found on every random access. However, overhead is already high enough that this doesn't make a noticible differenc, and performance is actually improved in some cases. Random access can be up to 10% faster, but other situations might be a bit slower. Generally using `foreach_get/set` are the best way to improve performance. See the differential revision for more discussion about Python performance. Cycles has been updated to use raw pointers and the internal Blender mesh types, mostly because there is no sense in having this overhead when it's already compiled with Blender. In my tests this roughly halves the Cycles mesh creation time (0.19s to 0.10s for a 1 million face grid). Differential Revision: https://developer.blender.org/D15488
2022-09-05 11:56:34 -05:00
polys.size(), [polys, varray = varray.typed<bool>()](const int face_index) {
/* A face is selected if all of its corners were selected. */
Mesh: Replace MPoly struct with offset indices Implements #95967. Currently the `MPoly` struct is 12 bytes, and stores the index of a face's first corner and the number of corners/verts/edges. Polygons and corners are always created in order by Blender, meaning each face's corners will be after the previous face's corners. We can take advantage of this fact and eliminate the redundancy in mesh face storage by only storing a single integer corner offset for each face. The size of the face is then encoded by the offset of the next face. The size of a single integer is 4 bytes, so this reduces memory usage by 3 times. The same method is used for `CurvesGeometry`, so Blender already has an abstraction to simplify using these offsets called `OffsetIndices`. This class is used to easily retrieve a range of corner indices for each face. This also gives the opportunity for sharing some logic with curves. Another benefit of the change is that the offsets and sizes stored in `MPoly` can no longer disagree with each other. Storing faces in the order of their corners can simplify some code too. Face/polygon variables now use the `IndexRange` type, which comes with quite a few utilities that can simplify code. Some: - The offset integer array has to be one longer than the face count to avoid a branch for every face, which means the data is no longer part of the mesh's `CustomData`. - We lose the ability to "reference" an original mesh's offset array until more reusable CoW from #104478 is committed. That will be added in a separate commit. - Since they aren't part of `CustomData`, poly offsets often have to be copied manually. - To simplify using `OffsetIndices` in many places, some functions and structs in headers were moved to only compile in C++. - All meshes created by Blender use the same order for faces and face corners, but just in case, meshes with mismatched order are fixed by versioning code. - `MeshPolygon.totloop` is no longer editable in RNA. This API break is necessary here unfortunately. It should be worth it in 3.6, since that's the best way to allow loading meshes from 4.0, which is important for an LTS version. Pull Request: https://projects.blender.org/blender/blender/pulls/105938
2023-04-04 20:39:28 +02:00
for (const int loop_index : polys[face_index]) {
if (!varray[loop_index]) {
return false;
}
}
return true;
});
}
else {
new_varray = VArray<T>::ForFunc(
Mesh: Remove redundant custom data pointers For copy-on-write, we want to share attribute arrays between meshes where possible. Mutable pointers like `Mesh.mvert` make that difficult by making ownership vague. They also make code more complex by adding redundancy. The simplest solution is just removing them and retrieving layers from `CustomData` as needed. Similar changes have already been applied to curves and point clouds (e9f82d3dc7ee, 410a6efb747f). Removing use of the pointers generally makes code more obvious and more reusable. Mesh data is now accessed with a C++ API (`Mesh::edges()` or `Mesh::edges_for_write()`), and a C API (`BKE_mesh_edges(mesh)`). The CoW changes this commit makes possible are described in T95845 and T95842, and started in D14139 and D14140. The change also simplifies the ongoing mesh struct-of-array refactors from T95965. **RNA/Python Access Performance** Theoretically, accessing mesh elements with the RNA API may become slower, since the layer needs to be found on every random access. However, overhead is already high enough that this doesn't make a noticible differenc, and performance is actually improved in some cases. Random access can be up to 10% faster, but other situations might be a bit slower. Generally using `foreach_get/set` are the best way to improve performance. See the differential revision for more discussion about Python performance. Cycles has been updated to use raw pointers and the internal Blender mesh types, mostly because there is no sense in having this overhead when it's already compiled with Blender. In my tests this roughly halves the Cycles mesh creation time (0.19s to 0.10s for a 1 million face grid). Differential Revision: https://developer.blender.org/D15488
2022-09-05 11:56:34 -05:00
polys.size(), [polys, varray = varray.typed<T>()](const int face_index) {
T return_value;
attribute_math::DefaultMixer<T> mixer({&return_value, 1});
Mesh: Replace MPoly struct with offset indices Implements #95967. Currently the `MPoly` struct is 12 bytes, and stores the index of a face's first corner and the number of corners/verts/edges. Polygons and corners are always created in order by Blender, meaning each face's corners will be after the previous face's corners. We can take advantage of this fact and eliminate the redundancy in mesh face storage by only storing a single integer corner offset for each face. The size of the face is then encoded by the offset of the next face. The size of a single integer is 4 bytes, so this reduces memory usage by 3 times. The same method is used for `CurvesGeometry`, so Blender already has an abstraction to simplify using these offsets called `OffsetIndices`. This class is used to easily retrieve a range of corner indices for each face. This also gives the opportunity for sharing some logic with curves. Another benefit of the change is that the offsets and sizes stored in `MPoly` can no longer disagree with each other. Storing faces in the order of their corners can simplify some code too. Face/polygon variables now use the `IndexRange` type, which comes with quite a few utilities that can simplify code. Some: - The offset integer array has to be one longer than the face count to avoid a branch for every face, which means the data is no longer part of the mesh's `CustomData`. - We lose the ability to "reference" an original mesh's offset array until more reusable CoW from #104478 is committed. That will be added in a separate commit. - Since they aren't part of `CustomData`, poly offsets often have to be copied manually. - To simplify using `OffsetIndices` in many places, some functions and structs in headers were moved to only compile in C++. - All meshes created by Blender use the same order for faces and face corners, but just in case, meshes with mismatched order are fixed by versioning code. - `MeshPolygon.totloop` is no longer editable in RNA. This API break is necessary here unfortunately. It should be worth it in 3.6, since that's the best way to allow loading meshes from 4.0, which is important for an LTS version. Pull Request: https://projects.blender.org/blender/blender/pulls/105938
2023-04-04 20:39:28 +02:00
for (const int loop_index : polys[face_index]) {
const T value = varray[loop_index];
mixer.mix_in(0, value);
}
mixer.finalize();
return return_value;
});
}
}
});
Geometry Nodes: use virtual arrays in internal attribute api A virtual array is a data structure that is similar to a normal array in that its elements can be accessed by an index. However, a virtual array does not have to be a contiguous array internally. Instead, its elements can be layed out arbitrarily while element access happens through a virtual function call. However, the virtual array data structures are designed so that the virtual function call can be avoided in cases where it could become a bottleneck. Most commonly, a virtual array is backed by an actual array/span or is a single value internally, that is the same for every index. Besides those, there are many more specialized virtual arrays like the ones that provides vertex positions based on the `MVert` struct or vertex group weights. Not all attributes used by geometry nodes are stored in simple contiguous arrays. To provide uniform access to all kinds of attributes, the attribute API has to provide virtual array functionality that hides the implementation details of attributes. Before this refactor, the attribute API provided its own virtual array implementation as part of the `ReadAttribute` and `WriteAttribute` types. That resulted in unnecessary code duplication with the virtual array system. Even worse, it bound many algorithms used by geometry nodes to the specifics of the attribute API, even though they could also use different data sources (such as data from sockets, default values, later results of expressions, ...). This refactor removes the `ReadAttribute` and `WriteAttribute` types and replaces them with `GVArray` and `GVMutableArray` respectively. The `GV` stands for "generic virtual". The "generic" means that the data type contained in those virtual arrays is only known at run-time. There are the corresponding statically typed types `VArray<T>` and `VMutableArray<T>` as well. No regressions are expected from this refactor. It does come with one improvement for users. The attribute API can convert the data type on write now. This is especially useful when writing to builtin attributes like `material_index` with e.g. the Attribute Math node (which usually just writes to float attributes, while `material_index` is an integer attribute). Differential Revision: https://developer.blender.org/D10994
2021-04-17 16:41:03 +02:00
return new_varray;
}
template<typename T>
static void adapt_mesh_domain_corner_to_edge_impl(const Mesh &mesh,
Geometry Nodes: use virtual arrays in internal attribute api A virtual array is a data structure that is similar to a normal array in that its elements can be accessed by an index. However, a virtual array does not have to be a contiguous array internally. Instead, its elements can be layed out arbitrarily while element access happens through a virtual function call. However, the virtual array data structures are designed so that the virtual function call can be avoided in cases where it could become a bottleneck. Most commonly, a virtual array is backed by an actual array/span or is a single value internally, that is the same for every index. Besides those, there are many more specialized virtual arrays like the ones that provides vertex positions based on the `MVert` struct or vertex group weights. Not all attributes used by geometry nodes are stored in simple contiguous arrays. To provide uniform access to all kinds of attributes, the attribute API has to provide virtual array functionality that hides the implementation details of attributes. Before this refactor, the attribute API provided its own virtual array implementation as part of the `ReadAttribute` and `WriteAttribute` types. That resulted in unnecessary code duplication with the virtual array system. Even worse, it bound many algorithms used by geometry nodes to the specifics of the attribute API, even though they could also use different data sources (such as data from sockets, default values, later results of expressions, ...). This refactor removes the `ReadAttribute` and `WriteAttribute` types and replaces them with `GVArray` and `GVMutableArray` respectively. The `GV` stands for "generic virtual". The "generic" means that the data type contained in those virtual arrays is only known at run-time. There are the corresponding statically typed types `VArray<T>` and `VMutableArray<T>` as well. No regressions are expected from this refactor. It does come with one improvement for users. The attribute API can convert the data type on write now. This is especially useful when writing to builtin attributes like `material_index` with e.g. the Attribute Math node (which usually just writes to float attributes, while `material_index` is an integer attribute). Differential Revision: https://developer.blender.org/D10994
2021-04-17 16:41:03 +02:00
const VArray<T> &old_values,
MutableSpan<T> r_values)
{
BLI_assert(r_values.size() == mesh.totedge);
Mesh: Replace MPoly struct with offset indices Implements #95967. Currently the `MPoly` struct is 12 bytes, and stores the index of a face's first corner and the number of corners/verts/edges. Polygons and corners are always created in order by Blender, meaning each face's corners will be after the previous face's corners. We can take advantage of this fact and eliminate the redundancy in mesh face storage by only storing a single integer corner offset for each face. The size of the face is then encoded by the offset of the next face. The size of a single integer is 4 bytes, so this reduces memory usage by 3 times. The same method is used for `CurvesGeometry`, so Blender already has an abstraction to simplify using these offsets called `OffsetIndices`. This class is used to easily retrieve a range of corner indices for each face. This also gives the opportunity for sharing some logic with curves. Another benefit of the change is that the offsets and sizes stored in `MPoly` can no longer disagree with each other. Storing faces in the order of their corners can simplify some code too. Face/polygon variables now use the `IndexRange` type, which comes with quite a few utilities that can simplify code. Some: - The offset integer array has to be one longer than the face count to avoid a branch for every face, which means the data is no longer part of the mesh's `CustomData`. - We lose the ability to "reference" an original mesh's offset array until more reusable CoW from #104478 is committed. That will be added in a separate commit. - Since they aren't part of `CustomData`, poly offsets often have to be copied manually. - To simplify using `OffsetIndices` in many places, some functions and structs in headers were moved to only compile in C++. - All meshes created by Blender use the same order for faces and face corners, but just in case, meshes with mismatched order are fixed by versioning code. - `MeshPolygon.totloop` is no longer editable in RNA. This API break is necessary here unfortunately. It should be worth it in 3.6, since that's the best way to allow loading meshes from 4.0, which is important for an LTS version. Pull Request: https://projects.blender.org/blender/blender/pulls/105938
2023-04-04 20:39:28 +02:00
const OffsetIndices polys = mesh.polys();
Mesh: Replace MLoop struct with generic attributes Implements #102359. Split the `MLoop` struct into two separate integer arrays called `corner_verts` and `corner_edges`, referring to the vertex each corner is attached to and the next edge around the face at each corner. These arrays can be sliced to give access to the edges or vertices in a face. Then they are often referred to as "poly_verts" or "poly_edges". The main benefits are halving the necessary memory bandwidth when only one array is used and simplifications from using regular integer indices instead of a special-purpose struct. The commit also starts a renaming from "loop" to "corner" in mesh code. Like the other mesh struct of array refactors, forward compatibility is kept by writing files with the older format. This will be done until 4.0 to ease the transition process. Looking at a small portion of the patch should give a good impression for the rest of the changes. I tried to make the changes as small as possible so it's easy to tell the correctness from the diff. Though I found Blender developers have been very inventive over the last decade when finding different ways to loop over the corners in a face. For performance, nearly every piece of code that deals with `Mesh` is slightly impacted. Any algorithm that is memory bottle-necked should see an improvement. For example, here is a comparison of interpolating a vertex float attribute to face corners (Ryzen 3700x): **Before** (Average: 3.7 ms, Min: 3.4 ms) ``` threading::parallel_for(loops.index_range(), 4096, [&](IndexRange range) { for (const int64_t i : range) { dst[i] = src[loops[i].v]; } }); ``` **After** (Average: 2.9 ms, Min: 2.6 ms) ``` array_utils::gather(src, corner_verts, dst); ``` That's an improvement of 28% to the average timings, and it's also a simplification, since an index-based routine can be used instead. For more examples using the new arrays, see the design task. Pull Request: https://projects.blender.org/blender/blender/pulls/104424
2023-03-20 15:55:13 +01:00
const Span<int> corner_edges = mesh.corner_edges();
Mesh: Remove redundant custom data pointers For copy-on-write, we want to share attribute arrays between meshes where possible. Mutable pointers like `Mesh.mvert` make that difficult by making ownership vague. They also make code more complex by adding redundancy. The simplest solution is just removing them and retrieving layers from `CustomData` as needed. Similar changes have already been applied to curves and point clouds (e9f82d3dc7ee, 410a6efb747f). Removing use of the pointers generally makes code more obvious and more reusable. Mesh data is now accessed with a C++ API (`Mesh::edges()` or `Mesh::edges_for_write()`), and a C API (`BKE_mesh_edges(mesh)`). The CoW changes this commit makes possible are described in T95845 and T95842, and started in D14139 and D14140. The change also simplifies the ongoing mesh struct-of-array refactors from T95965. **RNA/Python Access Performance** Theoretically, accessing mesh elements with the RNA API may become slower, since the layer needs to be found on every random access. However, overhead is already high enough that this doesn't make a noticible differenc, and performance is actually improved in some cases. Random access can be up to 10% faster, but other situations might be a bit slower. Generally using `foreach_get/set` are the best way to improve performance. See the differential revision for more discussion about Python performance. Cycles has been updated to use raw pointers and the internal Blender mesh types, mostly because there is no sense in having this overhead when it's already compiled with Blender. In my tests this roughly halves the Cycles mesh creation time (0.19s to 0.10s for a 1 million face grid). Differential Revision: https://developer.blender.org/D15488
2022-09-05 11:56:34 -05:00
attribute_math::DefaultMixer<T> mixer(r_values);
Mesh: Remove redundant custom data pointers For copy-on-write, we want to share attribute arrays between meshes where possible. Mutable pointers like `Mesh.mvert` make that difficult by making ownership vague. They also make code more complex by adding redundancy. The simplest solution is just removing them and retrieving layers from `CustomData` as needed. Similar changes have already been applied to curves and point clouds (e9f82d3dc7ee, 410a6efb747f). Removing use of the pointers generally makes code more obvious and more reusable. Mesh data is now accessed with a C++ API (`Mesh::edges()` or `Mesh::edges_for_write()`), and a C API (`BKE_mesh_edges(mesh)`). The CoW changes this commit makes possible are described in T95845 and T95842, and started in D14139 and D14140. The change also simplifies the ongoing mesh struct-of-array refactors from T95965. **RNA/Python Access Performance** Theoretically, accessing mesh elements with the RNA API may become slower, since the layer needs to be found on every random access. However, overhead is already high enough that this doesn't make a noticible differenc, and performance is actually improved in some cases. Random access can be up to 10% faster, but other situations might be a bit slower. Generally using `foreach_get/set` are the best way to improve performance. See the differential revision for more discussion about Python performance. Cycles has been updated to use raw pointers and the internal Blender mesh types, mostly because there is no sense in having this overhead when it's already compiled with Blender. In my tests this roughly halves the Cycles mesh creation time (0.19s to 0.10s for a 1 million face grid). Differential Revision: https://developer.blender.org/D15488
2022-09-05 11:56:34 -05:00
for (const int poly_index : polys.index_range()) {
Mesh: Replace MPoly struct with offset indices Implements #95967. Currently the `MPoly` struct is 12 bytes, and stores the index of a face's first corner and the number of corners/verts/edges. Polygons and corners are always created in order by Blender, meaning each face's corners will be after the previous face's corners. We can take advantage of this fact and eliminate the redundancy in mesh face storage by only storing a single integer corner offset for each face. The size of the face is then encoded by the offset of the next face. The size of a single integer is 4 bytes, so this reduces memory usage by 3 times. The same method is used for `CurvesGeometry`, so Blender already has an abstraction to simplify using these offsets called `OffsetIndices`. This class is used to easily retrieve a range of corner indices for each face. This also gives the opportunity for sharing some logic with curves. Another benefit of the change is that the offsets and sizes stored in `MPoly` can no longer disagree with each other. Storing faces in the order of their corners can simplify some code too. Face/polygon variables now use the `IndexRange` type, which comes with quite a few utilities that can simplify code. Some: - The offset integer array has to be one longer than the face count to avoid a branch for every face, which means the data is no longer part of the mesh's `CustomData`. - We lose the ability to "reference" an original mesh's offset array until more reusable CoW from #104478 is committed. That will be added in a separate commit. - Since they aren't part of `CustomData`, poly offsets often have to be copied manually. - To simplify using `OffsetIndices` in many places, some functions and structs in headers were moved to only compile in C++. - All meshes created by Blender use the same order for faces and face corners, but just in case, meshes with mismatched order are fixed by versioning code. - `MeshPolygon.totloop` is no longer editable in RNA. This API break is necessary here unfortunately. It should be worth it in 3.6, since that's the best way to allow loading meshes from 4.0, which is important for an LTS version. Pull Request: https://projects.blender.org/blender/blender/pulls/105938
2023-04-04 20:39:28 +02:00
const IndexRange poly = polys[poly_index];
/* For every edge, mix values from the two adjacent corners (the current and next corner). */
Mesh: Replace MPoly struct with offset indices Implements #95967. Currently the `MPoly` struct is 12 bytes, and stores the index of a face's first corner and the number of corners/verts/edges. Polygons and corners are always created in order by Blender, meaning each face's corners will be after the previous face's corners. We can take advantage of this fact and eliminate the redundancy in mesh face storage by only storing a single integer corner offset for each face. The size of the face is then encoded by the offset of the next face. The size of a single integer is 4 bytes, so this reduces memory usage by 3 times. The same method is used for `CurvesGeometry`, so Blender already has an abstraction to simplify using these offsets called `OffsetIndices`. This class is used to easily retrieve a range of corner indices for each face. This also gives the opportunity for sharing some logic with curves. Another benefit of the change is that the offsets and sizes stored in `MPoly` can no longer disagree with each other. Storing faces in the order of their corners can simplify some code too. Face/polygon variables now use the `IndexRange` type, which comes with quite a few utilities that can simplify code. Some: - The offset integer array has to be one longer than the face count to avoid a branch for every face, which means the data is no longer part of the mesh's `CustomData`. - We lose the ability to "reference" an original mesh's offset array until more reusable CoW from #104478 is committed. That will be added in a separate commit. - Since they aren't part of `CustomData`, poly offsets often have to be copied manually. - To simplify using `OffsetIndices` in many places, some functions and structs in headers were moved to only compile in C++. - All meshes created by Blender use the same order for faces and face corners, but just in case, meshes with mismatched order are fixed by versioning code. - `MeshPolygon.totloop` is no longer editable in RNA. This API break is necessary here unfortunately. It should be worth it in 3.6, since that's the best way to allow loading meshes from 4.0, which is important for an LTS version. Pull Request: https://projects.blender.org/blender/blender/pulls/105938
2023-04-04 20:39:28 +02:00
for (const int i : IndexRange(poly.size())) {
const int next_i = (i + 1) % poly.size();
const int loop_i = poly.start() + i;
const int next_loop_i = poly.start() + next_i;
Mesh: Replace MLoop struct with generic attributes Implements #102359. Split the `MLoop` struct into two separate integer arrays called `corner_verts` and `corner_edges`, referring to the vertex each corner is attached to and the next edge around the face at each corner. These arrays can be sliced to give access to the edges or vertices in a face. Then they are often referred to as "poly_verts" or "poly_edges". The main benefits are halving the necessary memory bandwidth when only one array is used and simplifications from using regular integer indices instead of a special-purpose struct. The commit also starts a renaming from "loop" to "corner" in mesh code. Like the other mesh struct of array refactors, forward compatibility is kept by writing files with the older format. This will be done until 4.0 to ease the transition process. Looking at a small portion of the patch should give a good impression for the rest of the changes. I tried to make the changes as small as possible so it's easy to tell the correctness from the diff. Though I found Blender developers have been very inventive over the last decade when finding different ways to loop over the corners in a face. For performance, nearly every piece of code that deals with `Mesh` is slightly impacted. Any algorithm that is memory bottle-necked should see an improvement. For example, here is a comparison of interpolating a vertex float attribute to face corners (Ryzen 3700x): **Before** (Average: 3.7 ms, Min: 3.4 ms) ``` threading::parallel_for(loops.index_range(), 4096, [&](IndexRange range) { for (const int64_t i : range) { dst[i] = src[loops[i].v]; } }); ``` **After** (Average: 2.9 ms, Min: 2.6 ms) ``` array_utils::gather(src, corner_verts, dst); ``` That's an improvement of 28% to the average timings, and it's also a simplification, since an index-based routine can be used instead. For more examples using the new arrays, see the design task. Pull Request: https://projects.blender.org/blender/blender/pulls/104424
2023-03-20 15:55:13 +01:00
const int edge_index = corner_edges[loop_i];
mixer.mix_in(edge_index, old_values[loop_i]);
mixer.mix_in(edge_index, old_values[next_loop_i]);
}
}
mixer.finalize();
}
/* An edge is selected if all corners on adjacent faces were selected. */
template<>
void adapt_mesh_domain_corner_to_edge_impl(const Mesh &mesh,
const VArray<bool> &old_values,
MutableSpan<bool> r_values)
{
BLI_assert(r_values.size() == mesh.totedge);
Mesh: Replace MPoly struct with offset indices Implements #95967. Currently the `MPoly` struct is 12 bytes, and stores the index of a face's first corner and the number of corners/verts/edges. Polygons and corners are always created in order by Blender, meaning each face's corners will be after the previous face's corners. We can take advantage of this fact and eliminate the redundancy in mesh face storage by only storing a single integer corner offset for each face. The size of the face is then encoded by the offset of the next face. The size of a single integer is 4 bytes, so this reduces memory usage by 3 times. The same method is used for `CurvesGeometry`, so Blender already has an abstraction to simplify using these offsets called `OffsetIndices`. This class is used to easily retrieve a range of corner indices for each face. This also gives the opportunity for sharing some logic with curves. Another benefit of the change is that the offsets and sizes stored in `MPoly` can no longer disagree with each other. Storing faces in the order of their corners can simplify some code too. Face/polygon variables now use the `IndexRange` type, which comes with quite a few utilities that can simplify code. Some: - The offset integer array has to be one longer than the face count to avoid a branch for every face, which means the data is no longer part of the mesh's `CustomData`. - We lose the ability to "reference" an original mesh's offset array until more reusable CoW from #104478 is committed. That will be added in a separate commit. - Since they aren't part of `CustomData`, poly offsets often have to be copied manually. - To simplify using `OffsetIndices` in many places, some functions and structs in headers were moved to only compile in C++. - All meshes created by Blender use the same order for faces and face corners, but just in case, meshes with mismatched order are fixed by versioning code. - `MeshPolygon.totloop` is no longer editable in RNA. This API break is necessary here unfortunately. It should be worth it in 3.6, since that's the best way to allow loading meshes from 4.0, which is important for an LTS version. Pull Request: https://projects.blender.org/blender/blender/pulls/105938
2023-04-04 20:39:28 +02:00
const OffsetIndices polys = mesh.polys();
Mesh: Replace MLoop struct with generic attributes Implements #102359. Split the `MLoop` struct into two separate integer arrays called `corner_verts` and `corner_edges`, referring to the vertex each corner is attached to and the next edge around the face at each corner. These arrays can be sliced to give access to the edges or vertices in a face. Then they are often referred to as "poly_verts" or "poly_edges". The main benefits are halving the necessary memory bandwidth when only one array is used and simplifications from using regular integer indices instead of a special-purpose struct. The commit also starts a renaming from "loop" to "corner" in mesh code. Like the other mesh struct of array refactors, forward compatibility is kept by writing files with the older format. This will be done until 4.0 to ease the transition process. Looking at a small portion of the patch should give a good impression for the rest of the changes. I tried to make the changes as small as possible so it's easy to tell the correctness from the diff. Though I found Blender developers have been very inventive over the last decade when finding different ways to loop over the corners in a face. For performance, nearly every piece of code that deals with `Mesh` is slightly impacted. Any algorithm that is memory bottle-necked should see an improvement. For example, here is a comparison of interpolating a vertex float attribute to face corners (Ryzen 3700x): **Before** (Average: 3.7 ms, Min: 3.4 ms) ``` threading::parallel_for(loops.index_range(), 4096, [&](IndexRange range) { for (const int64_t i : range) { dst[i] = src[loops[i].v]; } }); ``` **After** (Average: 2.9 ms, Min: 2.6 ms) ``` array_utils::gather(src, corner_verts, dst); ``` That's an improvement of 28% to the average timings, and it's also a simplification, since an index-based routine can be used instead. For more examples using the new arrays, see the design task. Pull Request: https://projects.blender.org/blender/blender/pulls/104424
2023-03-20 15:55:13 +01:00
const Span<int> corner_edges = mesh.corner_edges();
r_values.fill(true);
Mesh: Remove redundant custom data pointers For copy-on-write, we want to share attribute arrays between meshes where possible. Mutable pointers like `Mesh.mvert` make that difficult by making ownership vague. They also make code more complex by adding redundancy. The simplest solution is just removing them and retrieving layers from `CustomData` as needed. Similar changes have already been applied to curves and point clouds (e9f82d3dc7ee, 410a6efb747f). Removing use of the pointers generally makes code more obvious and more reusable. Mesh data is now accessed with a C++ API (`Mesh::edges()` or `Mesh::edges_for_write()`), and a C API (`BKE_mesh_edges(mesh)`). The CoW changes this commit makes possible are described in T95845 and T95842, and started in D14139 and D14140. The change also simplifies the ongoing mesh struct-of-array refactors from T95965. **RNA/Python Access Performance** Theoretically, accessing mesh elements with the RNA API may become slower, since the layer needs to be found on every random access. However, overhead is already high enough that this doesn't make a noticible differenc, and performance is actually improved in some cases. Random access can be up to 10% faster, but other situations might be a bit slower. Generally using `foreach_get/set` are the best way to improve performance. See the differential revision for more discussion about Python performance. Cycles has been updated to use raw pointers and the internal Blender mesh types, mostly because there is no sense in having this overhead when it's already compiled with Blender. In my tests this roughly halves the Cycles mesh creation time (0.19s to 0.10s for a 1 million face grid). Differential Revision: https://developer.blender.org/D15488
2022-09-05 11:56:34 -05:00
for (const int poly_index : polys.index_range()) {
Mesh: Replace MPoly struct with offset indices Implements #95967. Currently the `MPoly` struct is 12 bytes, and stores the index of a face's first corner and the number of corners/verts/edges. Polygons and corners are always created in order by Blender, meaning each face's corners will be after the previous face's corners. We can take advantage of this fact and eliminate the redundancy in mesh face storage by only storing a single integer corner offset for each face. The size of the face is then encoded by the offset of the next face. The size of a single integer is 4 bytes, so this reduces memory usage by 3 times. The same method is used for `CurvesGeometry`, so Blender already has an abstraction to simplify using these offsets called `OffsetIndices`. This class is used to easily retrieve a range of corner indices for each face. This also gives the opportunity for sharing some logic with curves. Another benefit of the change is that the offsets and sizes stored in `MPoly` can no longer disagree with each other. Storing faces in the order of their corners can simplify some code too. Face/polygon variables now use the `IndexRange` type, which comes with quite a few utilities that can simplify code. Some: - The offset integer array has to be one longer than the face count to avoid a branch for every face, which means the data is no longer part of the mesh's `CustomData`. - We lose the ability to "reference" an original mesh's offset array until more reusable CoW from #104478 is committed. That will be added in a separate commit. - Since they aren't part of `CustomData`, poly offsets often have to be copied manually. - To simplify using `OffsetIndices` in many places, some functions and structs in headers were moved to only compile in C++. - All meshes created by Blender use the same order for faces and face corners, but just in case, meshes with mismatched order are fixed by versioning code. - `MeshPolygon.totloop` is no longer editable in RNA. This API break is necessary here unfortunately. It should be worth it in 3.6, since that's the best way to allow loading meshes from 4.0, which is important for an LTS version. Pull Request: https://projects.blender.org/blender/blender/pulls/105938
2023-04-04 20:39:28 +02:00
const IndexRange poly = polys[poly_index];
Mesh: Replace MPoly struct with offset indices Implements #95967. Currently the `MPoly` struct is 12 bytes, and stores the index of a face's first corner and the number of corners/verts/edges. Polygons and corners are always created in order by Blender, meaning each face's corners will be after the previous face's corners. We can take advantage of this fact and eliminate the redundancy in mesh face storage by only storing a single integer corner offset for each face. The size of the face is then encoded by the offset of the next face. The size of a single integer is 4 bytes, so this reduces memory usage by 3 times. The same method is used for `CurvesGeometry`, so Blender already has an abstraction to simplify using these offsets called `OffsetIndices`. This class is used to easily retrieve a range of corner indices for each face. This also gives the opportunity for sharing some logic with curves. Another benefit of the change is that the offsets and sizes stored in `MPoly` can no longer disagree with each other. Storing faces in the order of their corners can simplify some code too. Face/polygon variables now use the `IndexRange` type, which comes with quite a few utilities that can simplify code. Some: - The offset integer array has to be one longer than the face count to avoid a branch for every face, which means the data is no longer part of the mesh's `CustomData`. - We lose the ability to "reference" an original mesh's offset array until more reusable CoW from #104478 is committed. That will be added in a separate commit. - Since they aren't part of `CustomData`, poly offsets often have to be copied manually. - To simplify using `OffsetIndices` in many places, some functions and structs in headers were moved to only compile in C++. - All meshes created by Blender use the same order for faces and face corners, but just in case, meshes with mismatched order are fixed by versioning code. - `MeshPolygon.totloop` is no longer editable in RNA. This API break is necessary here unfortunately. It should be worth it in 3.6, since that's the best way to allow loading meshes from 4.0, which is important for an LTS version. Pull Request: https://projects.blender.org/blender/blender/pulls/105938
2023-04-04 20:39:28 +02:00
for (const int i : IndexRange(poly.size())) {
const int next_i = (i + 1) % poly.size();
const int loop_i = poly[i];
const int next_loop_i = poly[next_i];
Mesh: Replace MLoop struct with generic attributes Implements #102359. Split the `MLoop` struct into two separate integer arrays called `corner_verts` and `corner_edges`, referring to the vertex each corner is attached to and the next edge around the face at each corner. These arrays can be sliced to give access to the edges or vertices in a face. Then they are often referred to as "poly_verts" or "poly_edges". The main benefits are halving the necessary memory bandwidth when only one array is used and simplifications from using regular integer indices instead of a special-purpose struct. The commit also starts a renaming from "loop" to "corner" in mesh code. Like the other mesh struct of array refactors, forward compatibility is kept by writing files with the older format. This will be done until 4.0 to ease the transition process. Looking at a small portion of the patch should give a good impression for the rest of the changes. I tried to make the changes as small as possible so it's easy to tell the correctness from the diff. Though I found Blender developers have been very inventive over the last decade when finding different ways to loop over the corners in a face. For performance, nearly every piece of code that deals with `Mesh` is slightly impacted. Any algorithm that is memory bottle-necked should see an improvement. For example, here is a comparison of interpolating a vertex float attribute to face corners (Ryzen 3700x): **Before** (Average: 3.7 ms, Min: 3.4 ms) ``` threading::parallel_for(loops.index_range(), 4096, [&](IndexRange range) { for (const int64_t i : range) { dst[i] = src[loops[i].v]; } }); ``` **After** (Average: 2.9 ms, Min: 2.6 ms) ``` array_utils::gather(src, corner_verts, dst); ``` That's an improvement of 28% to the average timings, and it's also a simplification, since an index-based routine can be used instead. For more examples using the new arrays, see the design task. Pull Request: https://projects.blender.org/blender/blender/pulls/104424
2023-03-20 15:55:13 +01:00
const int edge_index = corner_edges[loop_i];
if (!old_values[loop_i] || !old_values[next_loop_i]) {
r_values[edge_index] = false;
}
}
}
const bke::LooseEdgeCache &loose_edges = mesh.loose_edges();
if (loose_edges.count > 0) {
/* Deselect loose edges without corners that are still selected from the 'true' default. */
threading::parallel_for(IndexRange(mesh.totedge), 2048, [&](const IndexRange range) {
for (const int edge_index : range) {
if (loose_edges.is_loose_bits[edge_index]) {
r_values[edge_index] = false;
}
}
});
}
}
Geometry Nodes: refactor virtual array system Goals of this refactor: * Simplify creating virtual arrays. * Simplify passing virtual arrays around. * Simplify converting between typed and generic virtual arrays. * Reduce memory allocations. As a quick reminder, a virtual arrays is a data structure that behaves like an array (i.e. it can be accessed using an index). However, it may not actually be stored as array internally. The two most important implementations of virtual arrays are those that correspond to an actual plain array and those that have the same value for every index. However, many more implementations exist for various reasons (interfacing with legacy attributes, unified iterator over all points in multiple splines, ...). With this refactor the core types (`VArray`, `GVArray`, `VMutableArray` and `GVMutableArray`) can be used like "normal values". They typically live on the stack. Before, they were usually inside a `std::unique_ptr`. This makes passing them around much easier. Creation of new virtual arrays is also much simpler now due to some constructors. Memory allocations are reduced by making use of small object optimization inside the core types. Previously, `VArray` was a class with virtual methods that had to be overridden to change the behavior of a the virtual array. Now,`VArray` has a fixed size and has no virtual methods. Instead it contains a `VArrayImpl` that is similar to the old `VArray`. `VArrayImpl` should rarely ever be used directly, unless a new virtual array implementation is added. To support the small object optimization for many `VArrayImpl` classes, a new `blender::Any` type is added. It is similar to `std::any` with two additional features. It has an adjustable inline buffer size and alignment. The inline buffer size of `std::any` can't be relied on and is usually too small for our use case here. Furthermore, `blender::Any` can store additional user-defined type information without increasing the stack size. Differential Revision: https://developer.blender.org/D12986
2021-11-16 10:15:51 +01:00
static GVArray adapt_mesh_domain_corner_to_edge(const Mesh &mesh, const GVArray &varray)
{
GArray<> values(varray.type(), mesh.totedge);
Geometry Nodes: refactor virtual array system Goals of this refactor: * Simplify creating virtual arrays. * Simplify passing virtual arrays around. * Simplify converting between typed and generic virtual arrays. * Reduce memory allocations. As a quick reminder, a virtual arrays is a data structure that behaves like an array (i.e. it can be accessed using an index). However, it may not actually be stored as array internally. The two most important implementations of virtual arrays are those that correspond to an actual plain array and those that have the same value for every index. However, many more implementations exist for various reasons (interfacing with legacy attributes, unified iterator over all points in multiple splines, ...). With this refactor the core types (`VArray`, `GVArray`, `VMutableArray` and `GVMutableArray`) can be used like "normal values". They typically live on the stack. Before, they were usually inside a `std::unique_ptr`. This makes passing them around much easier. Creation of new virtual arrays is also much simpler now due to some constructors. Memory allocations are reduced by making use of small object optimization inside the core types. Previously, `VArray` was a class with virtual methods that had to be overridden to change the behavior of a the virtual array. Now,`VArray` has a fixed size and has no virtual methods. Instead it contains a `VArrayImpl` that is similar to the old `VArray`. `VArrayImpl` should rarely ever be used directly, unless a new virtual array implementation is added. To support the small object optimization for many `VArrayImpl` classes, a new `blender::Any` type is added. It is similar to `std::any` with two additional features. It has an adjustable inline buffer size and alignment. The inline buffer size of `std::any` can't be relied on and is usually too small for our use case here. Furthermore, `blender::Any` can store additional user-defined type information without increasing the stack size. Differential Revision: https://developer.blender.org/D12986
2021-11-16 10:15:51 +01:00
attribute_math::convert_to_static_type(varray.type(), [&](auto dummy) {
using T = decltype(dummy);
if constexpr (!std::is_void_v<attribute_math::DefaultMixer<T>>) {
adapt_mesh_domain_corner_to_edge_impl<T>(
mesh, varray.typed<T>(), values.as_mutable_span().typed<T>());
}
});
return GVArray::ForGArray(std::move(values));
}
template<typename T>
void adapt_mesh_domain_face_to_point_impl(const Mesh &mesh,
Geometry Nodes: use virtual arrays in internal attribute api A virtual array is a data structure that is similar to a normal array in that its elements can be accessed by an index. However, a virtual array does not have to be a contiguous array internally. Instead, its elements can be layed out arbitrarily while element access happens through a virtual function call. However, the virtual array data structures are designed so that the virtual function call can be avoided in cases where it could become a bottleneck. Most commonly, a virtual array is backed by an actual array/span or is a single value internally, that is the same for every index. Besides those, there are many more specialized virtual arrays like the ones that provides vertex positions based on the `MVert` struct or vertex group weights. Not all attributes used by geometry nodes are stored in simple contiguous arrays. To provide uniform access to all kinds of attributes, the attribute API has to provide virtual array functionality that hides the implementation details of attributes. Before this refactor, the attribute API provided its own virtual array implementation as part of the `ReadAttribute` and `WriteAttribute` types. That resulted in unnecessary code duplication with the virtual array system. Even worse, it bound many algorithms used by geometry nodes to the specifics of the attribute API, even though they could also use different data sources (such as data from sockets, default values, later results of expressions, ...). This refactor removes the `ReadAttribute` and `WriteAttribute` types and replaces them with `GVArray` and `GVMutableArray` respectively. The `GV` stands for "generic virtual". The "generic" means that the data type contained in those virtual arrays is only known at run-time. There are the corresponding statically typed types `VArray<T>` and `VMutableArray<T>` as well. No regressions are expected from this refactor. It does come with one improvement for users. The attribute API can convert the data type on write now. This is especially useful when writing to builtin attributes like `material_index` with e.g. the Attribute Math node (which usually just writes to float attributes, while `material_index` is an integer attribute). Differential Revision: https://developer.blender.org/D10994
2021-04-17 16:41:03 +02:00
const VArray<T> &old_values,
MutableSpan<T> r_values)
{
BLI_assert(r_values.size() == mesh.totvert);
Mesh: Replace MPoly struct with offset indices Implements #95967. Currently the `MPoly` struct is 12 bytes, and stores the index of a face's first corner and the number of corners/verts/edges. Polygons and corners are always created in order by Blender, meaning each face's corners will be after the previous face's corners. We can take advantage of this fact and eliminate the redundancy in mesh face storage by only storing a single integer corner offset for each face. The size of the face is then encoded by the offset of the next face. The size of a single integer is 4 bytes, so this reduces memory usage by 3 times. The same method is used for `CurvesGeometry`, so Blender already has an abstraction to simplify using these offsets called `OffsetIndices`. This class is used to easily retrieve a range of corner indices for each face. This also gives the opportunity for sharing some logic with curves. Another benefit of the change is that the offsets and sizes stored in `MPoly` can no longer disagree with each other. Storing faces in the order of their corners can simplify some code too. Face/polygon variables now use the `IndexRange` type, which comes with quite a few utilities that can simplify code. Some: - The offset integer array has to be one longer than the face count to avoid a branch for every face, which means the data is no longer part of the mesh's `CustomData`. - We lose the ability to "reference" an original mesh's offset array until more reusable CoW from #104478 is committed. That will be added in a separate commit. - Since they aren't part of `CustomData`, poly offsets often have to be copied manually. - To simplify using `OffsetIndices` in many places, some functions and structs in headers were moved to only compile in C++. - All meshes created by Blender use the same order for faces and face corners, but just in case, meshes with mismatched order are fixed by versioning code. - `MeshPolygon.totloop` is no longer editable in RNA. This API break is necessary here unfortunately. It should be worth it in 3.6, since that's the best way to allow loading meshes from 4.0, which is important for an LTS version. Pull Request: https://projects.blender.org/blender/blender/pulls/105938
2023-04-04 20:39:28 +02:00
const OffsetIndices polys = mesh.polys();
Mesh: Replace MLoop struct with generic attributes Implements #102359. Split the `MLoop` struct into two separate integer arrays called `corner_verts` and `corner_edges`, referring to the vertex each corner is attached to and the next edge around the face at each corner. These arrays can be sliced to give access to the edges or vertices in a face. Then they are often referred to as "poly_verts" or "poly_edges". The main benefits are halving the necessary memory bandwidth when only one array is used and simplifications from using regular integer indices instead of a special-purpose struct. The commit also starts a renaming from "loop" to "corner" in mesh code. Like the other mesh struct of array refactors, forward compatibility is kept by writing files with the older format. This will be done until 4.0 to ease the transition process. Looking at a small portion of the patch should give a good impression for the rest of the changes. I tried to make the changes as small as possible so it's easy to tell the correctness from the diff. Though I found Blender developers have been very inventive over the last decade when finding different ways to loop over the corners in a face. For performance, nearly every piece of code that deals with `Mesh` is slightly impacted. Any algorithm that is memory bottle-necked should see an improvement. For example, here is a comparison of interpolating a vertex float attribute to face corners (Ryzen 3700x): **Before** (Average: 3.7 ms, Min: 3.4 ms) ``` threading::parallel_for(loops.index_range(), 4096, [&](IndexRange range) { for (const int64_t i : range) { dst[i] = src[loops[i].v]; } }); ``` **After** (Average: 2.9 ms, Min: 2.6 ms) ``` array_utils::gather(src, corner_verts, dst); ``` That's an improvement of 28% to the average timings, and it's also a simplification, since an index-based routine can be used instead. For more examples using the new arrays, see the design task. Pull Request: https://projects.blender.org/blender/blender/pulls/104424
2023-03-20 15:55:13 +01:00
const Span<int> corner_verts = mesh.corner_verts();
Mesh: Remove redundant custom data pointers For copy-on-write, we want to share attribute arrays between meshes where possible. Mutable pointers like `Mesh.mvert` make that difficult by making ownership vague. They also make code more complex by adding redundancy. The simplest solution is just removing them and retrieving layers from `CustomData` as needed. Similar changes have already been applied to curves and point clouds (e9f82d3dc7ee, 410a6efb747f). Removing use of the pointers generally makes code more obvious and more reusable. Mesh data is now accessed with a C++ API (`Mesh::edges()` or `Mesh::edges_for_write()`), and a C API (`BKE_mesh_edges(mesh)`). The CoW changes this commit makes possible are described in T95845 and T95842, and started in D14139 and D14140. The change also simplifies the ongoing mesh struct-of-array refactors from T95965. **RNA/Python Access Performance** Theoretically, accessing mesh elements with the RNA API may become slower, since the layer needs to be found on every random access. However, overhead is already high enough that this doesn't make a noticible differenc, and performance is actually improved in some cases. Random access can be up to 10% faster, but other situations might be a bit slower. Generally using `foreach_get/set` are the best way to improve performance. See the differential revision for more discussion about Python performance. Cycles has been updated to use raw pointers and the internal Blender mesh types, mostly because there is no sense in having this overhead when it's already compiled with Blender. In my tests this roughly halves the Cycles mesh creation time (0.19s to 0.10s for a 1 million face grid). Differential Revision: https://developer.blender.org/D15488
2022-09-05 11:56:34 -05:00
attribute_math::DefaultMixer<T> mixer(r_values);
Mesh: Remove redundant custom data pointers For copy-on-write, we want to share attribute arrays between meshes where possible. Mutable pointers like `Mesh.mvert` make that difficult by making ownership vague. They also make code more complex by adding redundancy. The simplest solution is just removing them and retrieving layers from `CustomData` as needed. Similar changes have already been applied to curves and point clouds (e9f82d3dc7ee, 410a6efb747f). Removing use of the pointers generally makes code more obvious and more reusable. Mesh data is now accessed with a C++ API (`Mesh::edges()` or `Mesh::edges_for_write()`), and a C API (`BKE_mesh_edges(mesh)`). The CoW changes this commit makes possible are described in T95845 and T95842, and started in D14139 and D14140. The change also simplifies the ongoing mesh struct-of-array refactors from T95965. **RNA/Python Access Performance** Theoretically, accessing mesh elements with the RNA API may become slower, since the layer needs to be found on every random access. However, overhead is already high enough that this doesn't make a noticible differenc, and performance is actually improved in some cases. Random access can be up to 10% faster, but other situations might be a bit slower. Generally using `foreach_get/set` are the best way to improve performance. See the differential revision for more discussion about Python performance. Cycles has been updated to use raw pointers and the internal Blender mesh types, mostly because there is no sense in having this overhead when it's already compiled with Blender. In my tests this roughly halves the Cycles mesh creation time (0.19s to 0.10s for a 1 million face grid). Differential Revision: https://developer.blender.org/D15488
2022-09-05 11:56:34 -05:00
for (const int poly_index : polys.index_range()) {
const T value = old_values[poly_index];
Mesh: Replace MPoly struct with offset indices Implements #95967. Currently the `MPoly` struct is 12 bytes, and stores the index of a face's first corner and the number of corners/verts/edges. Polygons and corners are always created in order by Blender, meaning each face's corners will be after the previous face's corners. We can take advantage of this fact and eliminate the redundancy in mesh face storage by only storing a single integer corner offset for each face. The size of the face is then encoded by the offset of the next face. The size of a single integer is 4 bytes, so this reduces memory usage by 3 times. The same method is used for `CurvesGeometry`, so Blender already has an abstraction to simplify using these offsets called `OffsetIndices`. This class is used to easily retrieve a range of corner indices for each face. This also gives the opportunity for sharing some logic with curves. Another benefit of the change is that the offsets and sizes stored in `MPoly` can no longer disagree with each other. Storing faces in the order of their corners can simplify some code too. Face/polygon variables now use the `IndexRange` type, which comes with quite a few utilities that can simplify code. Some: - The offset integer array has to be one longer than the face count to avoid a branch for every face, which means the data is no longer part of the mesh's `CustomData`. - We lose the ability to "reference" an original mesh's offset array until more reusable CoW from #104478 is committed. That will be added in a separate commit. - Since they aren't part of `CustomData`, poly offsets often have to be copied manually. - To simplify using `OffsetIndices` in many places, some functions and structs in headers were moved to only compile in C++. - All meshes created by Blender use the same order for faces and face corners, but just in case, meshes with mismatched order are fixed by versioning code. - `MeshPolygon.totloop` is no longer editable in RNA. This API break is necessary here unfortunately. It should be worth it in 3.6, since that's the best way to allow loading meshes from 4.0, which is important for an LTS version. Pull Request: https://projects.blender.org/blender/blender/pulls/105938
2023-04-04 20:39:28 +02:00
for (const int vert : corner_verts.slice(polys[poly_index])) {
Mesh: Replace MLoop struct with generic attributes Implements #102359. Split the `MLoop` struct into two separate integer arrays called `corner_verts` and `corner_edges`, referring to the vertex each corner is attached to and the next edge around the face at each corner. These arrays can be sliced to give access to the edges or vertices in a face. Then they are often referred to as "poly_verts" or "poly_edges". The main benefits are halving the necessary memory bandwidth when only one array is used and simplifications from using regular integer indices instead of a special-purpose struct. The commit also starts a renaming from "loop" to "corner" in mesh code. Like the other mesh struct of array refactors, forward compatibility is kept by writing files with the older format. This will be done until 4.0 to ease the transition process. Looking at a small portion of the patch should give a good impression for the rest of the changes. I tried to make the changes as small as possible so it's easy to tell the correctness from the diff. Though I found Blender developers have been very inventive over the last decade when finding different ways to loop over the corners in a face. For performance, nearly every piece of code that deals with `Mesh` is slightly impacted. Any algorithm that is memory bottle-necked should see an improvement. For example, here is a comparison of interpolating a vertex float attribute to face corners (Ryzen 3700x): **Before** (Average: 3.7 ms, Min: 3.4 ms) ``` threading::parallel_for(loops.index_range(), 4096, [&](IndexRange range) { for (const int64_t i : range) { dst[i] = src[loops[i].v]; } }); ``` **After** (Average: 2.9 ms, Min: 2.6 ms) ``` array_utils::gather(src, corner_verts, dst); ``` That's an improvement of 28% to the average timings, and it's also a simplification, since an index-based routine can be used instead. For more examples using the new arrays, see the design task. Pull Request: https://projects.blender.org/blender/blender/pulls/104424
2023-03-20 15:55:13 +01:00
mixer.mix_in(vert, value);
}
}
mixer.finalize();
}
/* A vertex is selected if any of the connected faces were selected. */
template<>
void adapt_mesh_domain_face_to_point_impl(const Mesh &mesh,
const VArray<bool> &old_values,
MutableSpan<bool> r_values)
{
BLI_assert(r_values.size() == mesh.totvert);
Mesh: Replace MPoly struct with offset indices Implements #95967. Currently the `MPoly` struct is 12 bytes, and stores the index of a face's first corner and the number of corners/verts/edges. Polygons and corners are always created in order by Blender, meaning each face's corners will be after the previous face's corners. We can take advantage of this fact and eliminate the redundancy in mesh face storage by only storing a single integer corner offset for each face. The size of the face is then encoded by the offset of the next face. The size of a single integer is 4 bytes, so this reduces memory usage by 3 times. The same method is used for `CurvesGeometry`, so Blender already has an abstraction to simplify using these offsets called `OffsetIndices`. This class is used to easily retrieve a range of corner indices for each face. This also gives the opportunity for sharing some logic with curves. Another benefit of the change is that the offsets and sizes stored in `MPoly` can no longer disagree with each other. Storing faces in the order of their corners can simplify some code too. Face/polygon variables now use the `IndexRange` type, which comes with quite a few utilities that can simplify code. Some: - The offset integer array has to be one longer than the face count to avoid a branch for every face, which means the data is no longer part of the mesh's `CustomData`. - We lose the ability to "reference" an original mesh's offset array until more reusable CoW from #104478 is committed. That will be added in a separate commit. - Since they aren't part of `CustomData`, poly offsets often have to be copied manually. - To simplify using `OffsetIndices` in many places, some functions and structs in headers were moved to only compile in C++. - All meshes created by Blender use the same order for faces and face corners, but just in case, meshes with mismatched order are fixed by versioning code. - `MeshPolygon.totloop` is no longer editable in RNA. This API break is necessary here unfortunately. It should be worth it in 3.6, since that's the best way to allow loading meshes from 4.0, which is important for an LTS version. Pull Request: https://projects.blender.org/blender/blender/pulls/105938
2023-04-04 20:39:28 +02:00
const OffsetIndices polys = mesh.polys();
Mesh: Replace MLoop struct with generic attributes Implements #102359. Split the `MLoop` struct into two separate integer arrays called `corner_verts` and `corner_edges`, referring to the vertex each corner is attached to and the next edge around the face at each corner. These arrays can be sliced to give access to the edges or vertices in a face. Then they are often referred to as "poly_verts" or "poly_edges". The main benefits are halving the necessary memory bandwidth when only one array is used and simplifications from using regular integer indices instead of a special-purpose struct. The commit also starts a renaming from "loop" to "corner" in mesh code. Like the other mesh struct of array refactors, forward compatibility is kept by writing files with the older format. This will be done until 4.0 to ease the transition process. Looking at a small portion of the patch should give a good impression for the rest of the changes. I tried to make the changes as small as possible so it's easy to tell the correctness from the diff. Though I found Blender developers have been very inventive over the last decade when finding different ways to loop over the corners in a face. For performance, nearly every piece of code that deals with `Mesh` is slightly impacted. Any algorithm that is memory bottle-necked should see an improvement. For example, here is a comparison of interpolating a vertex float attribute to face corners (Ryzen 3700x): **Before** (Average: 3.7 ms, Min: 3.4 ms) ``` threading::parallel_for(loops.index_range(), 4096, [&](IndexRange range) { for (const int64_t i : range) { dst[i] = src[loops[i].v]; } }); ``` **After** (Average: 2.9 ms, Min: 2.6 ms) ``` array_utils::gather(src, corner_verts, dst); ``` That's an improvement of 28% to the average timings, and it's also a simplification, since an index-based routine can be used instead. For more examples using the new arrays, see the design task. Pull Request: https://projects.blender.org/blender/blender/pulls/104424
2023-03-20 15:55:13 +01:00
const Span<int> corner_verts = mesh.corner_verts();
r_values.fill(false);
threading::parallel_for(polys.index_range(), 2048, [&](const IndexRange range) {
for (const int poly_index : range) {
if (old_values[poly_index]) {
Mesh: Replace MPoly struct with offset indices Implements #95967. Currently the `MPoly` struct is 12 bytes, and stores the index of a face's first corner and the number of corners/verts/edges. Polygons and corners are always created in order by Blender, meaning each face's corners will be after the previous face's corners. We can take advantage of this fact and eliminate the redundancy in mesh face storage by only storing a single integer corner offset for each face. The size of the face is then encoded by the offset of the next face. The size of a single integer is 4 bytes, so this reduces memory usage by 3 times. The same method is used for `CurvesGeometry`, so Blender already has an abstraction to simplify using these offsets called `OffsetIndices`. This class is used to easily retrieve a range of corner indices for each face. This also gives the opportunity for sharing some logic with curves. Another benefit of the change is that the offsets and sizes stored in `MPoly` can no longer disagree with each other. Storing faces in the order of their corners can simplify some code too. Face/polygon variables now use the `IndexRange` type, which comes with quite a few utilities that can simplify code. Some: - The offset integer array has to be one longer than the face count to avoid a branch for every face, which means the data is no longer part of the mesh's `CustomData`. - We lose the ability to "reference" an original mesh's offset array until more reusable CoW from #104478 is committed. That will be added in a separate commit. - Since they aren't part of `CustomData`, poly offsets often have to be copied manually. - To simplify using `OffsetIndices` in many places, some functions and structs in headers were moved to only compile in C++. - All meshes created by Blender use the same order for faces and face corners, but just in case, meshes with mismatched order are fixed by versioning code. - `MeshPolygon.totloop` is no longer editable in RNA. This API break is necessary here unfortunately. It should be worth it in 3.6, since that's the best way to allow loading meshes from 4.0, which is important for an LTS version. Pull Request: https://projects.blender.org/blender/blender/pulls/105938
2023-04-04 20:39:28 +02:00
for (const int vert : corner_verts.slice(polys[poly_index])) {
Mesh: Replace MLoop struct with generic attributes Implements #102359. Split the `MLoop` struct into two separate integer arrays called `corner_verts` and `corner_edges`, referring to the vertex each corner is attached to and the next edge around the face at each corner. These arrays can be sliced to give access to the edges or vertices in a face. Then they are often referred to as "poly_verts" or "poly_edges". The main benefits are halving the necessary memory bandwidth when only one array is used and simplifications from using regular integer indices instead of a special-purpose struct. The commit also starts a renaming from "loop" to "corner" in mesh code. Like the other mesh struct of array refactors, forward compatibility is kept by writing files with the older format. This will be done until 4.0 to ease the transition process. Looking at a small portion of the patch should give a good impression for the rest of the changes. I tried to make the changes as small as possible so it's easy to tell the correctness from the diff. Though I found Blender developers have been very inventive over the last decade when finding different ways to loop over the corners in a face. For performance, nearly every piece of code that deals with `Mesh` is slightly impacted. Any algorithm that is memory bottle-necked should see an improvement. For example, here is a comparison of interpolating a vertex float attribute to face corners (Ryzen 3700x): **Before** (Average: 3.7 ms, Min: 3.4 ms) ``` threading::parallel_for(loops.index_range(), 4096, [&](IndexRange range) { for (const int64_t i : range) { dst[i] = src[loops[i].v]; } }); ``` **After** (Average: 2.9 ms, Min: 2.6 ms) ``` array_utils::gather(src, corner_verts, dst); ``` That's an improvement of 28% to the average timings, and it's also a simplification, since an index-based routine can be used instead. For more examples using the new arrays, see the design task. Pull Request: https://projects.blender.org/blender/blender/pulls/104424
2023-03-20 15:55:13 +01:00
r_values[vert] = true;
}
}
}
});
}
Geometry Nodes: refactor virtual array system Goals of this refactor: * Simplify creating virtual arrays. * Simplify passing virtual arrays around. * Simplify converting between typed and generic virtual arrays. * Reduce memory allocations. As a quick reminder, a virtual arrays is a data structure that behaves like an array (i.e. it can be accessed using an index). However, it may not actually be stored as array internally. The two most important implementations of virtual arrays are those that correspond to an actual plain array and those that have the same value for every index. However, many more implementations exist for various reasons (interfacing with legacy attributes, unified iterator over all points in multiple splines, ...). With this refactor the core types (`VArray`, `GVArray`, `VMutableArray` and `GVMutableArray`) can be used like "normal values". They typically live on the stack. Before, they were usually inside a `std::unique_ptr`. This makes passing them around much easier. Creation of new virtual arrays is also much simpler now due to some constructors. Memory allocations are reduced by making use of small object optimization inside the core types. Previously, `VArray` was a class with virtual methods that had to be overridden to change the behavior of a the virtual array. Now,`VArray` has a fixed size and has no virtual methods. Instead it contains a `VArrayImpl` that is similar to the old `VArray`. `VArrayImpl` should rarely ever be used directly, unless a new virtual array implementation is added. To support the small object optimization for many `VArrayImpl` classes, a new `blender::Any` type is added. It is similar to `std::any` with two additional features. It has an adjustable inline buffer size and alignment. The inline buffer size of `std::any` can't be relied on and is usually too small for our use case here. Furthermore, `blender::Any` can store additional user-defined type information without increasing the stack size. Differential Revision: https://developer.blender.org/D12986
2021-11-16 10:15:51 +01:00
static GVArray adapt_mesh_domain_face_to_point(const Mesh &mesh, const GVArray &varray)
{
GArray<> values(varray.type(), mesh.totvert);
Geometry Nodes: refactor virtual array system Goals of this refactor: * Simplify creating virtual arrays. * Simplify passing virtual arrays around. * Simplify converting between typed and generic virtual arrays. * Reduce memory allocations. As a quick reminder, a virtual arrays is a data structure that behaves like an array (i.e. it can be accessed using an index). However, it may not actually be stored as array internally. The two most important implementations of virtual arrays are those that correspond to an actual plain array and those that have the same value for every index. However, many more implementations exist for various reasons (interfacing with legacy attributes, unified iterator over all points in multiple splines, ...). With this refactor the core types (`VArray`, `GVArray`, `VMutableArray` and `GVMutableArray`) can be used like "normal values". They typically live on the stack. Before, they were usually inside a `std::unique_ptr`. This makes passing them around much easier. Creation of new virtual arrays is also much simpler now due to some constructors. Memory allocations are reduced by making use of small object optimization inside the core types. Previously, `VArray` was a class with virtual methods that had to be overridden to change the behavior of a the virtual array. Now,`VArray` has a fixed size and has no virtual methods. Instead it contains a `VArrayImpl` that is similar to the old `VArray`. `VArrayImpl` should rarely ever be used directly, unless a new virtual array implementation is added. To support the small object optimization for many `VArrayImpl` classes, a new `blender::Any` type is added. It is similar to `std::any` with two additional features. It has an adjustable inline buffer size and alignment. The inline buffer size of `std::any` can't be relied on and is usually too small for our use case here. Furthermore, `blender::Any` can store additional user-defined type information without increasing the stack size. Differential Revision: https://developer.blender.org/D12986
2021-11-16 10:15:51 +01:00
attribute_math::convert_to_static_type(varray.type(), [&](auto dummy) {
using T = decltype(dummy);
if constexpr (!std::is_void_v<attribute_math::DefaultMixer<T>>) {
adapt_mesh_domain_face_to_point_impl<T>(
mesh, varray.typed<T>(), values.as_mutable_span().typed<T>());
}
});
return GVArray::ForGArray(std::move(values));
}
/* Each corner's value is simply a copy of the value at its face. */
template<typename T>
void adapt_mesh_domain_face_to_corner_impl(const Mesh &mesh,
Geometry Nodes: use virtual arrays in internal attribute api A virtual array is a data structure that is similar to a normal array in that its elements can be accessed by an index. However, a virtual array does not have to be a contiguous array internally. Instead, its elements can be layed out arbitrarily while element access happens through a virtual function call. However, the virtual array data structures are designed so that the virtual function call can be avoided in cases where it could become a bottleneck. Most commonly, a virtual array is backed by an actual array/span or is a single value internally, that is the same for every index. Besides those, there are many more specialized virtual arrays like the ones that provides vertex positions based on the `MVert` struct or vertex group weights. Not all attributes used by geometry nodes are stored in simple contiguous arrays. To provide uniform access to all kinds of attributes, the attribute API has to provide virtual array functionality that hides the implementation details of attributes. Before this refactor, the attribute API provided its own virtual array implementation as part of the `ReadAttribute` and `WriteAttribute` types. That resulted in unnecessary code duplication with the virtual array system. Even worse, it bound many algorithms used by geometry nodes to the specifics of the attribute API, even though they could also use different data sources (such as data from sockets, default values, later results of expressions, ...). This refactor removes the `ReadAttribute` and `WriteAttribute` types and replaces them with `GVArray` and `GVMutableArray` respectively. The `GV` stands for "generic virtual". The "generic" means that the data type contained in those virtual arrays is only known at run-time. There are the corresponding statically typed types `VArray<T>` and `VMutableArray<T>` as well. No regressions are expected from this refactor. It does come with one improvement for users. The attribute API can convert the data type on write now. This is especially useful when writing to builtin attributes like `material_index` with e.g. the Attribute Math node (which usually just writes to float attributes, while `material_index` is an integer attribute). Differential Revision: https://developer.blender.org/D10994
2021-04-17 16:41:03 +02:00
const VArray<T> &old_values,
MutableSpan<T> r_values)
{
BLI_assert(r_values.size() == mesh.totloop);
Mesh: Replace MPoly struct with offset indices Implements #95967. Currently the `MPoly` struct is 12 bytes, and stores the index of a face's first corner and the number of corners/verts/edges. Polygons and corners are always created in order by Blender, meaning each face's corners will be after the previous face's corners. We can take advantage of this fact and eliminate the redundancy in mesh face storage by only storing a single integer corner offset for each face. The size of the face is then encoded by the offset of the next face. The size of a single integer is 4 bytes, so this reduces memory usage by 3 times. The same method is used for `CurvesGeometry`, so Blender already has an abstraction to simplify using these offsets called `OffsetIndices`. This class is used to easily retrieve a range of corner indices for each face. This also gives the opportunity for sharing some logic with curves. Another benefit of the change is that the offsets and sizes stored in `MPoly` can no longer disagree with each other. Storing faces in the order of their corners can simplify some code too. Face/polygon variables now use the `IndexRange` type, which comes with quite a few utilities that can simplify code. Some: - The offset integer array has to be one longer than the face count to avoid a branch for every face, which means the data is no longer part of the mesh's `CustomData`. - We lose the ability to "reference" an original mesh's offset array until more reusable CoW from #104478 is committed. That will be added in a separate commit. - Since they aren't part of `CustomData`, poly offsets often have to be copied manually. - To simplify using `OffsetIndices` in many places, some functions and structs in headers were moved to only compile in C++. - All meshes created by Blender use the same order for faces and face corners, but just in case, meshes with mismatched order are fixed by versioning code. - `MeshPolygon.totloop` is no longer editable in RNA. This API break is necessary here unfortunately. It should be worth it in 3.6, since that's the best way to allow loading meshes from 4.0, which is important for an LTS version. Pull Request: https://projects.blender.org/blender/blender/pulls/105938
2023-04-04 20:39:28 +02:00
const OffsetIndices polys = mesh.polys();
Mesh: Remove redundant custom data pointers For copy-on-write, we want to share attribute arrays between meshes where possible. Mutable pointers like `Mesh.mvert` make that difficult by making ownership vague. They also make code more complex by adding redundancy. The simplest solution is just removing them and retrieving layers from `CustomData` as needed. Similar changes have already been applied to curves and point clouds (e9f82d3dc7ee, 410a6efb747f). Removing use of the pointers generally makes code more obvious and more reusable. Mesh data is now accessed with a C++ API (`Mesh::edges()` or `Mesh::edges_for_write()`), and a C API (`BKE_mesh_edges(mesh)`). The CoW changes this commit makes possible are described in T95845 and T95842, and started in D14139 and D14140. The change also simplifies the ongoing mesh struct-of-array refactors from T95965. **RNA/Python Access Performance** Theoretically, accessing mesh elements with the RNA API may become slower, since the layer needs to be found on every random access. However, overhead is already high enough that this doesn't make a noticible differenc, and performance is actually improved in some cases. Random access can be up to 10% faster, but other situations might be a bit slower. Generally using `foreach_get/set` are the best way to improve performance. See the differential revision for more discussion about Python performance. Cycles has been updated to use raw pointers and the internal Blender mesh types, mostly because there is no sense in having this overhead when it's already compiled with Blender. In my tests this roughly halves the Cycles mesh creation time (0.19s to 0.10s for a 1 million face grid). Differential Revision: https://developer.blender.org/D15488
2022-09-05 11:56:34 -05:00
threading::parallel_for(polys.index_range(), 1024, [&](const IndexRange range) {
for (const int poly_index : range) {
Mesh: Replace MPoly struct with offset indices Implements #95967. Currently the `MPoly` struct is 12 bytes, and stores the index of a face's first corner and the number of corners/verts/edges. Polygons and corners are always created in order by Blender, meaning each face's corners will be after the previous face's corners. We can take advantage of this fact and eliminate the redundancy in mesh face storage by only storing a single integer corner offset for each face. The size of the face is then encoded by the offset of the next face. The size of a single integer is 4 bytes, so this reduces memory usage by 3 times. The same method is used for `CurvesGeometry`, so Blender already has an abstraction to simplify using these offsets called `OffsetIndices`. This class is used to easily retrieve a range of corner indices for each face. This also gives the opportunity for sharing some logic with curves. Another benefit of the change is that the offsets and sizes stored in `MPoly` can no longer disagree with each other. Storing faces in the order of their corners can simplify some code too. Face/polygon variables now use the `IndexRange` type, which comes with quite a few utilities that can simplify code. Some: - The offset integer array has to be one longer than the face count to avoid a branch for every face, which means the data is no longer part of the mesh's `CustomData`. - We lose the ability to "reference" an original mesh's offset array until more reusable CoW from #104478 is committed. That will be added in a separate commit. - Since they aren't part of `CustomData`, poly offsets often have to be copied manually. - To simplify using `OffsetIndices` in many places, some functions and structs in headers were moved to only compile in C++. - All meshes created by Blender use the same order for faces and face corners, but just in case, meshes with mismatched order are fixed by versioning code. - `MeshPolygon.totloop` is no longer editable in RNA. This API break is necessary here unfortunately. It should be worth it in 3.6, since that's the best way to allow loading meshes from 4.0, which is important for an LTS version. Pull Request: https://projects.blender.org/blender/blender/pulls/105938
2023-04-04 20:39:28 +02:00
MutableSpan<T> poly_corner_values = r_values.slice(polys[poly_index]);
poly_corner_values.fill(old_values[poly_index]);
}
});
}
Geometry Nodes: refactor virtual array system Goals of this refactor: * Simplify creating virtual arrays. * Simplify passing virtual arrays around. * Simplify converting between typed and generic virtual arrays. * Reduce memory allocations. As a quick reminder, a virtual arrays is a data structure that behaves like an array (i.e. it can be accessed using an index). However, it may not actually be stored as array internally. The two most important implementations of virtual arrays are those that correspond to an actual plain array and those that have the same value for every index. However, many more implementations exist for various reasons (interfacing with legacy attributes, unified iterator over all points in multiple splines, ...). With this refactor the core types (`VArray`, `GVArray`, `VMutableArray` and `GVMutableArray`) can be used like "normal values". They typically live on the stack. Before, they were usually inside a `std::unique_ptr`. This makes passing them around much easier. Creation of new virtual arrays is also much simpler now due to some constructors. Memory allocations are reduced by making use of small object optimization inside the core types. Previously, `VArray` was a class with virtual methods that had to be overridden to change the behavior of a the virtual array. Now,`VArray` has a fixed size and has no virtual methods. Instead it contains a `VArrayImpl` that is similar to the old `VArray`. `VArrayImpl` should rarely ever be used directly, unless a new virtual array implementation is added. To support the small object optimization for many `VArrayImpl` classes, a new `blender::Any` type is added. It is similar to `std::any` with two additional features. It has an adjustable inline buffer size and alignment. The inline buffer size of `std::any` can't be relied on and is usually too small for our use case here. Furthermore, `blender::Any` can store additional user-defined type information without increasing the stack size. Differential Revision: https://developer.blender.org/D12986
2021-11-16 10:15:51 +01:00
static GVArray adapt_mesh_domain_face_to_corner(const Mesh &mesh, const GVArray &varray)
{
GArray<> values(varray.type(), mesh.totloop);
Geometry Nodes: refactor virtual array system Goals of this refactor: * Simplify creating virtual arrays. * Simplify passing virtual arrays around. * Simplify converting between typed and generic virtual arrays. * Reduce memory allocations. As a quick reminder, a virtual arrays is a data structure that behaves like an array (i.e. it can be accessed using an index). However, it may not actually be stored as array internally. The two most important implementations of virtual arrays are those that correspond to an actual plain array and those that have the same value for every index. However, many more implementations exist for various reasons (interfacing with legacy attributes, unified iterator over all points in multiple splines, ...). With this refactor the core types (`VArray`, `GVArray`, `VMutableArray` and `GVMutableArray`) can be used like "normal values". They typically live on the stack. Before, they were usually inside a `std::unique_ptr`. This makes passing them around much easier. Creation of new virtual arrays is also much simpler now due to some constructors. Memory allocations are reduced by making use of small object optimization inside the core types. Previously, `VArray` was a class with virtual methods that had to be overridden to change the behavior of a the virtual array. Now,`VArray` has a fixed size and has no virtual methods. Instead it contains a `VArrayImpl` that is similar to the old `VArray`. `VArrayImpl` should rarely ever be used directly, unless a new virtual array implementation is added. To support the small object optimization for many `VArrayImpl` classes, a new `blender::Any` type is added. It is similar to `std::any` with two additional features. It has an adjustable inline buffer size and alignment. The inline buffer size of `std::any` can't be relied on and is usually too small for our use case here. Furthermore, `blender::Any` can store additional user-defined type information without increasing the stack size. Differential Revision: https://developer.blender.org/D12986
2021-11-16 10:15:51 +01:00
attribute_math::convert_to_static_type(varray.type(), [&](auto dummy) {
using T = decltype(dummy);
if constexpr (!std::is_void_v<attribute_math::DefaultMixer<T>>) {
adapt_mesh_domain_face_to_corner_impl<T>(
mesh, varray.typed<T>(), values.as_mutable_span().typed<T>());
}
});
return GVArray::ForGArray(std::move(values));
}
template<typename T>
void adapt_mesh_domain_face_to_edge_impl(const Mesh &mesh,
Geometry Nodes: use virtual arrays in internal attribute api A virtual array is a data structure that is similar to a normal array in that its elements can be accessed by an index. However, a virtual array does not have to be a contiguous array internally. Instead, its elements can be layed out arbitrarily while element access happens through a virtual function call. However, the virtual array data structures are designed so that the virtual function call can be avoided in cases where it could become a bottleneck. Most commonly, a virtual array is backed by an actual array/span or is a single value internally, that is the same for every index. Besides those, there are many more specialized virtual arrays like the ones that provides vertex positions based on the `MVert` struct or vertex group weights. Not all attributes used by geometry nodes are stored in simple contiguous arrays. To provide uniform access to all kinds of attributes, the attribute API has to provide virtual array functionality that hides the implementation details of attributes. Before this refactor, the attribute API provided its own virtual array implementation as part of the `ReadAttribute` and `WriteAttribute` types. That resulted in unnecessary code duplication with the virtual array system. Even worse, it bound many algorithms used by geometry nodes to the specifics of the attribute API, even though they could also use different data sources (such as data from sockets, default values, later results of expressions, ...). This refactor removes the `ReadAttribute` and `WriteAttribute` types and replaces them with `GVArray` and `GVMutableArray` respectively. The `GV` stands for "generic virtual". The "generic" means that the data type contained in those virtual arrays is only known at run-time. There are the corresponding statically typed types `VArray<T>` and `VMutableArray<T>` as well. No regressions are expected from this refactor. It does come with one improvement for users. The attribute API can convert the data type on write now. This is especially useful when writing to builtin attributes like `material_index` with e.g. the Attribute Math node (which usually just writes to float attributes, while `material_index` is an integer attribute). Differential Revision: https://developer.blender.org/D10994
2021-04-17 16:41:03 +02:00
const VArray<T> &old_values,
MutableSpan<T> r_values)
{
BLI_assert(r_values.size() == mesh.totedge);
Mesh: Replace MPoly struct with offset indices Implements #95967. Currently the `MPoly` struct is 12 bytes, and stores the index of a face's first corner and the number of corners/verts/edges. Polygons and corners are always created in order by Blender, meaning each face's corners will be after the previous face's corners. We can take advantage of this fact and eliminate the redundancy in mesh face storage by only storing a single integer corner offset for each face. The size of the face is then encoded by the offset of the next face. The size of a single integer is 4 bytes, so this reduces memory usage by 3 times. The same method is used for `CurvesGeometry`, so Blender already has an abstraction to simplify using these offsets called `OffsetIndices`. This class is used to easily retrieve a range of corner indices for each face. This also gives the opportunity for sharing some logic with curves. Another benefit of the change is that the offsets and sizes stored in `MPoly` can no longer disagree with each other. Storing faces in the order of their corners can simplify some code too. Face/polygon variables now use the `IndexRange` type, which comes with quite a few utilities that can simplify code. Some: - The offset integer array has to be one longer than the face count to avoid a branch for every face, which means the data is no longer part of the mesh's `CustomData`. - We lose the ability to "reference" an original mesh's offset array until more reusable CoW from #104478 is committed. That will be added in a separate commit. - Since they aren't part of `CustomData`, poly offsets often have to be copied manually. - To simplify using `OffsetIndices` in many places, some functions and structs in headers were moved to only compile in C++. - All meshes created by Blender use the same order for faces and face corners, but just in case, meshes with mismatched order are fixed by versioning code. - `MeshPolygon.totloop` is no longer editable in RNA. This API break is necessary here unfortunately. It should be worth it in 3.6, since that's the best way to allow loading meshes from 4.0, which is important for an LTS version. Pull Request: https://projects.blender.org/blender/blender/pulls/105938
2023-04-04 20:39:28 +02:00
const OffsetIndices polys = mesh.polys();
Mesh: Replace MLoop struct with generic attributes Implements #102359. Split the `MLoop` struct into two separate integer arrays called `corner_verts` and `corner_edges`, referring to the vertex each corner is attached to and the next edge around the face at each corner. These arrays can be sliced to give access to the edges or vertices in a face. Then they are often referred to as "poly_verts" or "poly_edges". The main benefits are halving the necessary memory bandwidth when only one array is used and simplifications from using regular integer indices instead of a special-purpose struct. The commit also starts a renaming from "loop" to "corner" in mesh code. Like the other mesh struct of array refactors, forward compatibility is kept by writing files with the older format. This will be done until 4.0 to ease the transition process. Looking at a small portion of the patch should give a good impression for the rest of the changes. I tried to make the changes as small as possible so it's easy to tell the correctness from the diff. Though I found Blender developers have been very inventive over the last decade when finding different ways to loop over the corners in a face. For performance, nearly every piece of code that deals with `Mesh` is slightly impacted. Any algorithm that is memory bottle-necked should see an improvement. For example, here is a comparison of interpolating a vertex float attribute to face corners (Ryzen 3700x): **Before** (Average: 3.7 ms, Min: 3.4 ms) ``` threading::parallel_for(loops.index_range(), 4096, [&](IndexRange range) { for (const int64_t i : range) { dst[i] = src[loops[i].v]; } }); ``` **After** (Average: 2.9 ms, Min: 2.6 ms) ``` array_utils::gather(src, corner_verts, dst); ``` That's an improvement of 28% to the average timings, and it's also a simplification, since an index-based routine can be used instead. For more examples using the new arrays, see the design task. Pull Request: https://projects.blender.org/blender/blender/pulls/104424
2023-03-20 15:55:13 +01:00
const Span<int> corner_edges = mesh.corner_edges();
Mesh: Remove redundant custom data pointers For copy-on-write, we want to share attribute arrays between meshes where possible. Mutable pointers like `Mesh.mvert` make that difficult by making ownership vague. They also make code more complex by adding redundancy. The simplest solution is just removing them and retrieving layers from `CustomData` as needed. Similar changes have already been applied to curves and point clouds (e9f82d3dc7ee, 410a6efb747f). Removing use of the pointers generally makes code more obvious and more reusable. Mesh data is now accessed with a C++ API (`Mesh::edges()` or `Mesh::edges_for_write()`), and a C API (`BKE_mesh_edges(mesh)`). The CoW changes this commit makes possible are described in T95845 and T95842, and started in D14139 and D14140. The change also simplifies the ongoing mesh struct-of-array refactors from T95965. **RNA/Python Access Performance** Theoretically, accessing mesh elements with the RNA API may become slower, since the layer needs to be found on every random access. However, overhead is already high enough that this doesn't make a noticible differenc, and performance is actually improved in some cases. Random access can be up to 10% faster, but other situations might be a bit slower. Generally using `foreach_get/set` are the best way to improve performance. See the differential revision for more discussion about Python performance. Cycles has been updated to use raw pointers and the internal Blender mesh types, mostly because there is no sense in having this overhead when it's already compiled with Blender. In my tests this roughly halves the Cycles mesh creation time (0.19s to 0.10s for a 1 million face grid). Differential Revision: https://developer.blender.org/D15488
2022-09-05 11:56:34 -05:00
attribute_math::DefaultMixer<T> mixer(r_values);
Mesh: Remove redundant custom data pointers For copy-on-write, we want to share attribute arrays between meshes where possible. Mutable pointers like `Mesh.mvert` make that difficult by making ownership vague. They also make code more complex by adding redundancy. The simplest solution is just removing them and retrieving layers from `CustomData` as needed. Similar changes have already been applied to curves and point clouds (e9f82d3dc7ee, 410a6efb747f). Removing use of the pointers generally makes code more obvious and more reusable. Mesh data is now accessed with a C++ API (`Mesh::edges()` or `Mesh::edges_for_write()`), and a C API (`BKE_mesh_edges(mesh)`). The CoW changes this commit makes possible are described in T95845 and T95842, and started in D14139 and D14140. The change also simplifies the ongoing mesh struct-of-array refactors from T95965. **RNA/Python Access Performance** Theoretically, accessing mesh elements with the RNA API may become slower, since the layer needs to be found on every random access. However, overhead is already high enough that this doesn't make a noticible differenc, and performance is actually improved in some cases. Random access can be up to 10% faster, but other situations might be a bit slower. Generally using `foreach_get/set` are the best way to improve performance. See the differential revision for more discussion about Python performance. Cycles has been updated to use raw pointers and the internal Blender mesh types, mostly because there is no sense in having this overhead when it's already compiled with Blender. In my tests this roughly halves the Cycles mesh creation time (0.19s to 0.10s for a 1 million face grid). Differential Revision: https://developer.blender.org/D15488
2022-09-05 11:56:34 -05:00
for (const int poly_index : polys.index_range()) {
const T value = old_values[poly_index];
Mesh: Replace MPoly struct with offset indices Implements #95967. Currently the `MPoly` struct is 12 bytes, and stores the index of a face's first corner and the number of corners/verts/edges. Polygons and corners are always created in order by Blender, meaning each face's corners will be after the previous face's corners. We can take advantage of this fact and eliminate the redundancy in mesh face storage by only storing a single integer corner offset for each face. The size of the face is then encoded by the offset of the next face. The size of a single integer is 4 bytes, so this reduces memory usage by 3 times. The same method is used for `CurvesGeometry`, so Blender already has an abstraction to simplify using these offsets called `OffsetIndices`. This class is used to easily retrieve a range of corner indices for each face. This also gives the opportunity for sharing some logic with curves. Another benefit of the change is that the offsets and sizes stored in `MPoly` can no longer disagree with each other. Storing faces in the order of their corners can simplify some code too. Face/polygon variables now use the `IndexRange` type, which comes with quite a few utilities that can simplify code. Some: - The offset integer array has to be one longer than the face count to avoid a branch for every face, which means the data is no longer part of the mesh's `CustomData`. - We lose the ability to "reference" an original mesh's offset array until more reusable CoW from #104478 is committed. That will be added in a separate commit. - Since they aren't part of `CustomData`, poly offsets often have to be copied manually. - To simplify using `OffsetIndices` in many places, some functions and structs in headers were moved to only compile in C++. - All meshes created by Blender use the same order for faces and face corners, but just in case, meshes with mismatched order are fixed by versioning code. - `MeshPolygon.totloop` is no longer editable in RNA. This API break is necessary here unfortunately. It should be worth it in 3.6, since that's the best way to allow loading meshes from 4.0, which is important for an LTS version. Pull Request: https://projects.blender.org/blender/blender/pulls/105938
2023-04-04 20:39:28 +02:00
for (const int edge : corner_edges.slice(polys[poly_index])) {
Mesh: Replace MLoop struct with generic attributes Implements #102359. Split the `MLoop` struct into two separate integer arrays called `corner_verts` and `corner_edges`, referring to the vertex each corner is attached to and the next edge around the face at each corner. These arrays can be sliced to give access to the edges or vertices in a face. Then they are often referred to as "poly_verts" or "poly_edges". The main benefits are halving the necessary memory bandwidth when only one array is used and simplifications from using regular integer indices instead of a special-purpose struct. The commit also starts a renaming from "loop" to "corner" in mesh code. Like the other mesh struct of array refactors, forward compatibility is kept by writing files with the older format. This will be done until 4.0 to ease the transition process. Looking at a small portion of the patch should give a good impression for the rest of the changes. I tried to make the changes as small as possible so it's easy to tell the correctness from the diff. Though I found Blender developers have been very inventive over the last decade when finding different ways to loop over the corners in a face. For performance, nearly every piece of code that deals with `Mesh` is slightly impacted. Any algorithm that is memory bottle-necked should see an improvement. For example, here is a comparison of interpolating a vertex float attribute to face corners (Ryzen 3700x): **Before** (Average: 3.7 ms, Min: 3.4 ms) ``` threading::parallel_for(loops.index_range(), 4096, [&](IndexRange range) { for (const int64_t i : range) { dst[i] = src[loops[i].v]; } }); ``` **After** (Average: 2.9 ms, Min: 2.6 ms) ``` array_utils::gather(src, corner_verts, dst); ``` That's an improvement of 28% to the average timings, and it's also a simplification, since an index-based routine can be used instead. For more examples using the new arrays, see the design task. Pull Request: https://projects.blender.org/blender/blender/pulls/104424
2023-03-20 15:55:13 +01:00
mixer.mix_in(edge, value);
}
}
mixer.finalize();
}
/* An edge is selected if any connected face was selected. */
template<>
void adapt_mesh_domain_face_to_edge_impl(const Mesh &mesh,
const VArray<bool> &old_values,
MutableSpan<bool> r_values)
{
BLI_assert(r_values.size() == mesh.totedge);
Mesh: Replace MPoly struct with offset indices Implements #95967. Currently the `MPoly` struct is 12 bytes, and stores the index of a face's first corner and the number of corners/verts/edges. Polygons and corners are always created in order by Blender, meaning each face's corners will be after the previous face's corners. We can take advantage of this fact and eliminate the redundancy in mesh face storage by only storing a single integer corner offset for each face. The size of the face is then encoded by the offset of the next face. The size of a single integer is 4 bytes, so this reduces memory usage by 3 times. The same method is used for `CurvesGeometry`, so Blender already has an abstraction to simplify using these offsets called `OffsetIndices`. This class is used to easily retrieve a range of corner indices for each face. This also gives the opportunity for sharing some logic with curves. Another benefit of the change is that the offsets and sizes stored in `MPoly` can no longer disagree with each other. Storing faces in the order of their corners can simplify some code too. Face/polygon variables now use the `IndexRange` type, which comes with quite a few utilities that can simplify code. Some: - The offset integer array has to be one longer than the face count to avoid a branch for every face, which means the data is no longer part of the mesh's `CustomData`. - We lose the ability to "reference" an original mesh's offset array until more reusable CoW from #104478 is committed. That will be added in a separate commit. - Since they aren't part of `CustomData`, poly offsets often have to be copied manually. - To simplify using `OffsetIndices` in many places, some functions and structs in headers were moved to only compile in C++. - All meshes created by Blender use the same order for faces and face corners, but just in case, meshes with mismatched order are fixed by versioning code. - `MeshPolygon.totloop` is no longer editable in RNA. This API break is necessary here unfortunately. It should be worth it in 3.6, since that's the best way to allow loading meshes from 4.0, which is important for an LTS version. Pull Request: https://projects.blender.org/blender/blender/pulls/105938
2023-04-04 20:39:28 +02:00
const OffsetIndices polys = mesh.polys();
Mesh: Replace MLoop struct with generic attributes Implements #102359. Split the `MLoop` struct into two separate integer arrays called `corner_verts` and `corner_edges`, referring to the vertex each corner is attached to and the next edge around the face at each corner. These arrays can be sliced to give access to the edges or vertices in a face. Then they are often referred to as "poly_verts" or "poly_edges". The main benefits are halving the necessary memory bandwidth when only one array is used and simplifications from using regular integer indices instead of a special-purpose struct. The commit also starts a renaming from "loop" to "corner" in mesh code. Like the other mesh struct of array refactors, forward compatibility is kept by writing files with the older format. This will be done until 4.0 to ease the transition process. Looking at a small portion of the patch should give a good impression for the rest of the changes. I tried to make the changes as small as possible so it's easy to tell the correctness from the diff. Though I found Blender developers have been very inventive over the last decade when finding different ways to loop over the corners in a face. For performance, nearly every piece of code that deals with `Mesh` is slightly impacted. Any algorithm that is memory bottle-necked should see an improvement. For example, here is a comparison of interpolating a vertex float attribute to face corners (Ryzen 3700x): **Before** (Average: 3.7 ms, Min: 3.4 ms) ``` threading::parallel_for(loops.index_range(), 4096, [&](IndexRange range) { for (const int64_t i : range) { dst[i] = src[loops[i].v]; } }); ``` **After** (Average: 2.9 ms, Min: 2.6 ms) ``` array_utils::gather(src, corner_verts, dst); ``` That's an improvement of 28% to the average timings, and it's also a simplification, since an index-based routine can be used instead. For more examples using the new arrays, see the design task. Pull Request: https://projects.blender.org/blender/blender/pulls/104424
2023-03-20 15:55:13 +01:00
const Span<int> corner_edges = mesh.corner_edges();
r_values.fill(false);
threading::parallel_for(polys.index_range(), 2048, [&](const IndexRange range) {
for (const int poly_index : range) {
if (old_values[poly_index]) {
Mesh: Replace MPoly struct with offset indices Implements #95967. Currently the `MPoly` struct is 12 bytes, and stores the index of a face's first corner and the number of corners/verts/edges. Polygons and corners are always created in order by Blender, meaning each face's corners will be after the previous face's corners. We can take advantage of this fact and eliminate the redundancy in mesh face storage by only storing a single integer corner offset for each face. The size of the face is then encoded by the offset of the next face. The size of a single integer is 4 bytes, so this reduces memory usage by 3 times. The same method is used for `CurvesGeometry`, so Blender already has an abstraction to simplify using these offsets called `OffsetIndices`. This class is used to easily retrieve a range of corner indices for each face. This also gives the opportunity for sharing some logic with curves. Another benefit of the change is that the offsets and sizes stored in `MPoly` can no longer disagree with each other. Storing faces in the order of their corners can simplify some code too. Face/polygon variables now use the `IndexRange` type, which comes with quite a few utilities that can simplify code. Some: - The offset integer array has to be one longer than the face count to avoid a branch for every face, which means the data is no longer part of the mesh's `CustomData`. - We lose the ability to "reference" an original mesh's offset array until more reusable CoW from #104478 is committed. That will be added in a separate commit. - Since they aren't part of `CustomData`, poly offsets often have to be copied manually. - To simplify using `OffsetIndices` in many places, some functions and structs in headers were moved to only compile in C++. - All meshes created by Blender use the same order for faces and face corners, but just in case, meshes with mismatched order are fixed by versioning code. - `MeshPolygon.totloop` is no longer editable in RNA. This API break is necessary here unfortunately. It should be worth it in 3.6, since that's the best way to allow loading meshes from 4.0, which is important for an LTS version. Pull Request: https://projects.blender.org/blender/blender/pulls/105938
2023-04-04 20:39:28 +02:00
for (const int edge : corner_edges.slice(polys[poly_index])) {
Mesh: Replace MLoop struct with generic attributes Implements #102359. Split the `MLoop` struct into two separate integer arrays called `corner_verts` and `corner_edges`, referring to the vertex each corner is attached to and the next edge around the face at each corner. These arrays can be sliced to give access to the edges or vertices in a face. Then they are often referred to as "poly_verts" or "poly_edges". The main benefits are halving the necessary memory bandwidth when only one array is used and simplifications from using regular integer indices instead of a special-purpose struct. The commit also starts a renaming from "loop" to "corner" in mesh code. Like the other mesh struct of array refactors, forward compatibility is kept by writing files with the older format. This will be done until 4.0 to ease the transition process. Looking at a small portion of the patch should give a good impression for the rest of the changes. I tried to make the changes as small as possible so it's easy to tell the correctness from the diff. Though I found Blender developers have been very inventive over the last decade when finding different ways to loop over the corners in a face. For performance, nearly every piece of code that deals with `Mesh` is slightly impacted. Any algorithm that is memory bottle-necked should see an improvement. For example, here is a comparison of interpolating a vertex float attribute to face corners (Ryzen 3700x): **Before** (Average: 3.7 ms, Min: 3.4 ms) ``` threading::parallel_for(loops.index_range(), 4096, [&](IndexRange range) { for (const int64_t i : range) { dst[i] = src[loops[i].v]; } }); ``` **After** (Average: 2.9 ms, Min: 2.6 ms) ``` array_utils::gather(src, corner_verts, dst); ``` That's an improvement of 28% to the average timings, and it's also a simplification, since an index-based routine can be used instead. For more examples using the new arrays, see the design task. Pull Request: https://projects.blender.org/blender/blender/pulls/104424
2023-03-20 15:55:13 +01:00
r_values[edge] = true;
}
}
}
});
}
Geometry Nodes: refactor virtual array system Goals of this refactor: * Simplify creating virtual arrays. * Simplify passing virtual arrays around. * Simplify converting between typed and generic virtual arrays. * Reduce memory allocations. As a quick reminder, a virtual arrays is a data structure that behaves like an array (i.e. it can be accessed using an index). However, it may not actually be stored as array internally. The two most important implementations of virtual arrays are those that correspond to an actual plain array and those that have the same value for every index. However, many more implementations exist for various reasons (interfacing with legacy attributes, unified iterator over all points in multiple splines, ...). With this refactor the core types (`VArray`, `GVArray`, `VMutableArray` and `GVMutableArray`) can be used like "normal values". They typically live on the stack. Before, they were usually inside a `std::unique_ptr`. This makes passing them around much easier. Creation of new virtual arrays is also much simpler now due to some constructors. Memory allocations are reduced by making use of small object optimization inside the core types. Previously, `VArray` was a class with virtual methods that had to be overridden to change the behavior of a the virtual array. Now,`VArray` has a fixed size and has no virtual methods. Instead it contains a `VArrayImpl` that is similar to the old `VArray`. `VArrayImpl` should rarely ever be used directly, unless a new virtual array implementation is added. To support the small object optimization for many `VArrayImpl` classes, a new `blender::Any` type is added. It is similar to `std::any` with two additional features. It has an adjustable inline buffer size and alignment. The inline buffer size of `std::any` can't be relied on and is usually too small for our use case here. Furthermore, `blender::Any` can store additional user-defined type information without increasing the stack size. Differential Revision: https://developer.blender.org/D12986
2021-11-16 10:15:51 +01:00
static GVArray adapt_mesh_domain_face_to_edge(const Mesh &mesh, const GVArray &varray)
{
GArray<> values(varray.type(), mesh.totedge);
Geometry Nodes: refactor virtual array system Goals of this refactor: * Simplify creating virtual arrays. * Simplify passing virtual arrays around. * Simplify converting between typed and generic virtual arrays. * Reduce memory allocations. As a quick reminder, a virtual arrays is a data structure that behaves like an array (i.e. it can be accessed using an index). However, it may not actually be stored as array internally. The two most important implementations of virtual arrays are those that correspond to an actual plain array and those that have the same value for every index. However, many more implementations exist for various reasons (interfacing with legacy attributes, unified iterator over all points in multiple splines, ...). With this refactor the core types (`VArray`, `GVArray`, `VMutableArray` and `GVMutableArray`) can be used like "normal values". They typically live on the stack. Before, they were usually inside a `std::unique_ptr`. This makes passing them around much easier. Creation of new virtual arrays is also much simpler now due to some constructors. Memory allocations are reduced by making use of small object optimization inside the core types. Previously, `VArray` was a class with virtual methods that had to be overridden to change the behavior of a the virtual array. Now,`VArray` has a fixed size and has no virtual methods. Instead it contains a `VArrayImpl` that is similar to the old `VArray`. `VArrayImpl` should rarely ever be used directly, unless a new virtual array implementation is added. To support the small object optimization for many `VArrayImpl` classes, a new `blender::Any` type is added. It is similar to `std::any` with two additional features. It has an adjustable inline buffer size and alignment. The inline buffer size of `std::any` can't be relied on and is usually too small for our use case here. Furthermore, `blender::Any` can store additional user-defined type information without increasing the stack size. Differential Revision: https://developer.blender.org/D12986
2021-11-16 10:15:51 +01:00
attribute_math::convert_to_static_type(varray.type(), [&](auto dummy) {
using T = decltype(dummy);
if constexpr (!std::is_void_v<attribute_math::DefaultMixer<T>>) {
adapt_mesh_domain_face_to_edge_impl<T>(
mesh, varray.typed<T>(), values.as_mutable_span().typed<T>());
}
});
return GVArray::ForGArray(std::move(values));
}
Geometry Nodes: refactor virtual array system Goals of this refactor: * Simplify creating virtual arrays. * Simplify passing virtual arrays around. * Simplify converting between typed and generic virtual arrays. * Reduce memory allocations. As a quick reminder, a virtual arrays is a data structure that behaves like an array (i.e. it can be accessed using an index). However, it may not actually be stored as array internally. The two most important implementations of virtual arrays are those that correspond to an actual plain array and those that have the same value for every index. However, many more implementations exist for various reasons (interfacing with legacy attributes, unified iterator over all points in multiple splines, ...). With this refactor the core types (`VArray`, `GVArray`, `VMutableArray` and `GVMutableArray`) can be used like "normal values". They typically live on the stack. Before, they were usually inside a `std::unique_ptr`. This makes passing them around much easier. Creation of new virtual arrays is also much simpler now due to some constructors. Memory allocations are reduced by making use of small object optimization inside the core types. Previously, `VArray` was a class with virtual methods that had to be overridden to change the behavior of a the virtual array. Now,`VArray` has a fixed size and has no virtual methods. Instead it contains a `VArrayImpl` that is similar to the old `VArray`. `VArrayImpl` should rarely ever be used directly, unless a new virtual array implementation is added. To support the small object optimization for many `VArrayImpl` classes, a new `blender::Any` type is added. It is similar to `std::any` with two additional features. It has an adjustable inline buffer size and alignment. The inline buffer size of `std::any` can't be relied on and is usually too small for our use case here. Furthermore, `blender::Any` can store additional user-defined type information without increasing the stack size. Differential Revision: https://developer.blender.org/D12986
2021-11-16 10:15:51 +01:00
static GVArray adapt_mesh_domain_point_to_face(const Mesh &mesh, const GVArray &varray)
{
Mesh: Replace MPoly struct with offset indices Implements #95967. Currently the `MPoly` struct is 12 bytes, and stores the index of a face's first corner and the number of corners/verts/edges. Polygons and corners are always created in order by Blender, meaning each face's corners will be after the previous face's corners. We can take advantage of this fact and eliminate the redundancy in mesh face storage by only storing a single integer corner offset for each face. The size of the face is then encoded by the offset of the next face. The size of a single integer is 4 bytes, so this reduces memory usage by 3 times. The same method is used for `CurvesGeometry`, so Blender already has an abstraction to simplify using these offsets called `OffsetIndices`. This class is used to easily retrieve a range of corner indices for each face. This also gives the opportunity for sharing some logic with curves. Another benefit of the change is that the offsets and sizes stored in `MPoly` can no longer disagree with each other. Storing faces in the order of their corners can simplify some code too. Face/polygon variables now use the `IndexRange` type, which comes with quite a few utilities that can simplify code. Some: - The offset integer array has to be one longer than the face count to avoid a branch for every face, which means the data is no longer part of the mesh's `CustomData`. - We lose the ability to "reference" an original mesh's offset array until more reusable CoW from #104478 is committed. That will be added in a separate commit. - Since they aren't part of `CustomData`, poly offsets often have to be copied manually. - To simplify using `OffsetIndices` in many places, some functions and structs in headers were moved to only compile in C++. - All meshes created by Blender use the same order for faces and face corners, but just in case, meshes with mismatched order are fixed by versioning code. - `MeshPolygon.totloop` is no longer editable in RNA. This API break is necessary here unfortunately. It should be worth it in 3.6, since that's the best way to allow loading meshes from 4.0, which is important for an LTS version. Pull Request: https://projects.blender.org/blender/blender/pulls/105938
2023-04-04 20:39:28 +02:00
const OffsetIndices polys = mesh.polys();
Mesh: Replace MLoop struct with generic attributes Implements #102359. Split the `MLoop` struct into two separate integer arrays called `corner_verts` and `corner_edges`, referring to the vertex each corner is attached to and the next edge around the face at each corner. These arrays can be sliced to give access to the edges or vertices in a face. Then they are often referred to as "poly_verts" or "poly_edges". The main benefits are halving the necessary memory bandwidth when only one array is used and simplifications from using regular integer indices instead of a special-purpose struct. The commit also starts a renaming from "loop" to "corner" in mesh code. Like the other mesh struct of array refactors, forward compatibility is kept by writing files with the older format. This will be done until 4.0 to ease the transition process. Looking at a small portion of the patch should give a good impression for the rest of the changes. I tried to make the changes as small as possible so it's easy to tell the correctness from the diff. Though I found Blender developers have been very inventive over the last decade when finding different ways to loop over the corners in a face. For performance, nearly every piece of code that deals with `Mesh` is slightly impacted. Any algorithm that is memory bottle-necked should see an improvement. For example, here is a comparison of interpolating a vertex float attribute to face corners (Ryzen 3700x): **Before** (Average: 3.7 ms, Min: 3.4 ms) ``` threading::parallel_for(loops.index_range(), 4096, [&](IndexRange range) { for (const int64_t i : range) { dst[i] = src[loops[i].v]; } }); ``` **After** (Average: 2.9 ms, Min: 2.6 ms) ``` array_utils::gather(src, corner_verts, dst); ``` That's an improvement of 28% to the average timings, and it's also a simplification, since an index-based routine can be used instead. For more examples using the new arrays, see the design task. Pull Request: https://projects.blender.org/blender/blender/pulls/104424
2023-03-20 15:55:13 +01:00
const Span<int> corner_verts = mesh.corner_verts();
Mesh: Remove redundant custom data pointers For copy-on-write, we want to share attribute arrays between meshes where possible. Mutable pointers like `Mesh.mvert` make that difficult by making ownership vague. They also make code more complex by adding redundancy. The simplest solution is just removing them and retrieving layers from `CustomData` as needed. Similar changes have already been applied to curves and point clouds (e9f82d3dc7ee, 410a6efb747f). Removing use of the pointers generally makes code more obvious and more reusable. Mesh data is now accessed with a C++ API (`Mesh::edges()` or `Mesh::edges_for_write()`), and a C API (`BKE_mesh_edges(mesh)`). The CoW changes this commit makes possible are described in T95845 and T95842, and started in D14139 and D14140. The change also simplifies the ongoing mesh struct-of-array refactors from T95965. **RNA/Python Access Performance** Theoretically, accessing mesh elements with the RNA API may become slower, since the layer needs to be found on every random access. However, overhead is already high enough that this doesn't make a noticible differenc, and performance is actually improved in some cases. Random access can be up to 10% faster, but other situations might be a bit slower. Generally using `foreach_get/set` are the best way to improve performance. See the differential revision for more discussion about Python performance. Cycles has been updated to use raw pointers and the internal Blender mesh types, mostly because there is no sense in having this overhead when it's already compiled with Blender. In my tests this roughly halves the Cycles mesh creation time (0.19s to 0.10s for a 1 million face grid). Differential Revision: https://developer.blender.org/D15488
2022-09-05 11:56:34 -05:00
Geometry Nodes: refactor virtual array system Goals of this refactor: * Simplify creating virtual arrays. * Simplify passing virtual arrays around. * Simplify converting between typed and generic virtual arrays. * Reduce memory allocations. As a quick reminder, a virtual arrays is a data structure that behaves like an array (i.e. it can be accessed using an index). However, it may not actually be stored as array internally. The two most important implementations of virtual arrays are those that correspond to an actual plain array and those that have the same value for every index. However, many more implementations exist for various reasons (interfacing with legacy attributes, unified iterator over all points in multiple splines, ...). With this refactor the core types (`VArray`, `GVArray`, `VMutableArray` and `GVMutableArray`) can be used like "normal values". They typically live on the stack. Before, they were usually inside a `std::unique_ptr`. This makes passing them around much easier. Creation of new virtual arrays is also much simpler now due to some constructors. Memory allocations are reduced by making use of small object optimization inside the core types. Previously, `VArray` was a class with virtual methods that had to be overridden to change the behavior of a the virtual array. Now,`VArray` has a fixed size and has no virtual methods. Instead it contains a `VArrayImpl` that is similar to the old `VArray`. `VArrayImpl` should rarely ever be used directly, unless a new virtual array implementation is added. To support the small object optimization for many `VArrayImpl` classes, a new `blender::Any` type is added. It is similar to `std::any` with two additional features. It has an adjustable inline buffer size and alignment. The inline buffer size of `std::any` can't be relied on and is usually too small for our use case here. Furthermore, `blender::Any` can store additional user-defined type information without increasing the stack size. Differential Revision: https://developer.blender.org/D12986
2021-11-16 10:15:51 +01:00
GVArray new_varray;
attribute_math::convert_to_static_type(varray.type(), [&](auto dummy) {
using T = decltype(dummy);
if constexpr (!std::is_void_v<attribute_math::DefaultMixer<T>>) {
if constexpr (std::is_same_v<T, bool>) {
new_varray = VArray<T>::ForFunc(
Mesh: Replace MLoop struct with generic attributes Implements #102359. Split the `MLoop` struct into two separate integer arrays called `corner_verts` and `corner_edges`, referring to the vertex each corner is attached to and the next edge around the face at each corner. These arrays can be sliced to give access to the edges or vertices in a face. Then they are often referred to as "poly_verts" or "poly_edges". The main benefits are halving the necessary memory bandwidth when only one array is used and simplifications from using regular integer indices instead of a special-purpose struct. The commit also starts a renaming from "loop" to "corner" in mesh code. Like the other mesh struct of array refactors, forward compatibility is kept by writing files with the older format. This will be done until 4.0 to ease the transition process. Looking at a small portion of the patch should give a good impression for the rest of the changes. I tried to make the changes as small as possible so it's easy to tell the correctness from the diff. Though I found Blender developers have been very inventive over the last decade when finding different ways to loop over the corners in a face. For performance, nearly every piece of code that deals with `Mesh` is slightly impacted. Any algorithm that is memory bottle-necked should see an improvement. For example, here is a comparison of interpolating a vertex float attribute to face corners (Ryzen 3700x): **Before** (Average: 3.7 ms, Min: 3.4 ms) ``` threading::parallel_for(loops.index_range(), 4096, [&](IndexRange range) { for (const int64_t i : range) { dst[i] = src[loops[i].v]; } }); ``` **After** (Average: 2.9 ms, Min: 2.6 ms) ``` array_utils::gather(src, corner_verts, dst); ``` That's an improvement of 28% to the average timings, and it's also a simplification, since an index-based routine can be used instead. For more examples using the new arrays, see the design task. Pull Request: https://projects.blender.org/blender/blender/pulls/104424
2023-03-20 15:55:13 +01:00
mesh.totpoly,
[corner_verts, polys, varray = varray.typed<bool>()](const int face_index) {
/* A face is selected if all of its vertices were selected. */
Mesh: Replace MPoly struct with offset indices Implements #95967. Currently the `MPoly` struct is 12 bytes, and stores the index of a face's first corner and the number of corners/verts/edges. Polygons and corners are always created in order by Blender, meaning each face's corners will be after the previous face's corners. We can take advantage of this fact and eliminate the redundancy in mesh face storage by only storing a single integer corner offset for each face. The size of the face is then encoded by the offset of the next face. The size of a single integer is 4 bytes, so this reduces memory usage by 3 times. The same method is used for `CurvesGeometry`, so Blender already has an abstraction to simplify using these offsets called `OffsetIndices`. This class is used to easily retrieve a range of corner indices for each face. This also gives the opportunity for sharing some logic with curves. Another benefit of the change is that the offsets and sizes stored in `MPoly` can no longer disagree with each other. Storing faces in the order of their corners can simplify some code too. Face/polygon variables now use the `IndexRange` type, which comes with quite a few utilities that can simplify code. Some: - The offset integer array has to be one longer than the face count to avoid a branch for every face, which means the data is no longer part of the mesh's `CustomData`. - We lose the ability to "reference" an original mesh's offset array until more reusable CoW from #104478 is committed. That will be added in a separate commit. - Since they aren't part of `CustomData`, poly offsets often have to be copied manually. - To simplify using `OffsetIndices` in many places, some functions and structs in headers were moved to only compile in C++. - All meshes created by Blender use the same order for faces and face corners, but just in case, meshes with mismatched order are fixed by versioning code. - `MeshPolygon.totloop` is no longer editable in RNA. This API break is necessary here unfortunately. It should be worth it in 3.6, since that's the best way to allow loading meshes from 4.0, which is important for an LTS version. Pull Request: https://projects.blender.org/blender/blender/pulls/105938
2023-04-04 20:39:28 +02:00
for (const int vert : corner_verts.slice(polys[face_index])) {
Mesh: Replace MLoop struct with generic attributes Implements #102359. Split the `MLoop` struct into two separate integer arrays called `corner_verts` and `corner_edges`, referring to the vertex each corner is attached to and the next edge around the face at each corner. These arrays can be sliced to give access to the edges or vertices in a face. Then they are often referred to as "poly_verts" or "poly_edges". The main benefits are halving the necessary memory bandwidth when only one array is used and simplifications from using regular integer indices instead of a special-purpose struct. The commit also starts a renaming from "loop" to "corner" in mesh code. Like the other mesh struct of array refactors, forward compatibility is kept by writing files with the older format. This will be done until 4.0 to ease the transition process. Looking at a small portion of the patch should give a good impression for the rest of the changes. I tried to make the changes as small as possible so it's easy to tell the correctness from the diff. Though I found Blender developers have been very inventive over the last decade when finding different ways to loop over the corners in a face. For performance, nearly every piece of code that deals with `Mesh` is slightly impacted. Any algorithm that is memory bottle-necked should see an improvement. For example, here is a comparison of interpolating a vertex float attribute to face corners (Ryzen 3700x): **Before** (Average: 3.7 ms, Min: 3.4 ms) ``` threading::parallel_for(loops.index_range(), 4096, [&](IndexRange range) { for (const int64_t i : range) { dst[i] = src[loops[i].v]; } }); ``` **After** (Average: 2.9 ms, Min: 2.6 ms) ``` array_utils::gather(src, corner_verts, dst); ``` That's an improvement of 28% to the average timings, and it's also a simplification, since an index-based routine can be used instead. For more examples using the new arrays, see the design task. Pull Request: https://projects.blender.org/blender/blender/pulls/104424
2023-03-20 15:55:13 +01:00
if (!varray[vert]) {
return false;
}
}
return true;
});
}
else {
new_varray = VArray<T>::ForFunc(
Mesh: Replace MLoop struct with generic attributes Implements #102359. Split the `MLoop` struct into two separate integer arrays called `corner_verts` and `corner_edges`, referring to the vertex each corner is attached to and the next edge around the face at each corner. These arrays can be sliced to give access to the edges or vertices in a face. Then they are often referred to as "poly_verts" or "poly_edges". The main benefits are halving the necessary memory bandwidth when only one array is used and simplifications from using regular integer indices instead of a special-purpose struct. The commit also starts a renaming from "loop" to "corner" in mesh code. Like the other mesh struct of array refactors, forward compatibility is kept by writing files with the older format. This will be done until 4.0 to ease the transition process. Looking at a small portion of the patch should give a good impression for the rest of the changes. I tried to make the changes as small as possible so it's easy to tell the correctness from the diff. Though I found Blender developers have been very inventive over the last decade when finding different ways to loop over the corners in a face. For performance, nearly every piece of code that deals with `Mesh` is slightly impacted. Any algorithm that is memory bottle-necked should see an improvement. For example, here is a comparison of interpolating a vertex float attribute to face corners (Ryzen 3700x): **Before** (Average: 3.7 ms, Min: 3.4 ms) ``` threading::parallel_for(loops.index_range(), 4096, [&](IndexRange range) { for (const int64_t i : range) { dst[i] = src[loops[i].v]; } }); ``` **After** (Average: 2.9 ms, Min: 2.6 ms) ``` array_utils::gather(src, corner_verts, dst); ``` That's an improvement of 28% to the average timings, and it's also a simplification, since an index-based routine can be used instead. For more examples using the new arrays, see the design task. Pull Request: https://projects.blender.org/blender/blender/pulls/104424
2023-03-20 15:55:13 +01:00
mesh.totpoly, [corner_verts, polys, varray = varray.typed<T>()](const int face_index) {
T return_value;
attribute_math::DefaultMixer<T> mixer({&return_value, 1});
Mesh: Replace MPoly struct with offset indices Implements #95967. Currently the `MPoly` struct is 12 bytes, and stores the index of a face's first corner and the number of corners/verts/edges. Polygons and corners are always created in order by Blender, meaning each face's corners will be after the previous face's corners. We can take advantage of this fact and eliminate the redundancy in mesh face storage by only storing a single integer corner offset for each face. The size of the face is then encoded by the offset of the next face. The size of a single integer is 4 bytes, so this reduces memory usage by 3 times. The same method is used for `CurvesGeometry`, so Blender already has an abstraction to simplify using these offsets called `OffsetIndices`. This class is used to easily retrieve a range of corner indices for each face. This also gives the opportunity for sharing some logic with curves. Another benefit of the change is that the offsets and sizes stored in `MPoly` can no longer disagree with each other. Storing faces in the order of their corners can simplify some code too. Face/polygon variables now use the `IndexRange` type, which comes with quite a few utilities that can simplify code. Some: - The offset integer array has to be one longer than the face count to avoid a branch for every face, which means the data is no longer part of the mesh's `CustomData`. - We lose the ability to "reference" an original mesh's offset array until more reusable CoW from #104478 is committed. That will be added in a separate commit. - Since they aren't part of `CustomData`, poly offsets often have to be copied manually. - To simplify using `OffsetIndices` in many places, some functions and structs in headers were moved to only compile in C++. - All meshes created by Blender use the same order for faces and face corners, but just in case, meshes with mismatched order are fixed by versioning code. - `MeshPolygon.totloop` is no longer editable in RNA. This API break is necessary here unfortunately. It should be worth it in 3.6, since that's the best way to allow loading meshes from 4.0, which is important for an LTS version. Pull Request: https://projects.blender.org/blender/blender/pulls/105938
2023-04-04 20:39:28 +02:00
for (const int vert : corner_verts.slice(polys[face_index])) {
Mesh: Replace MLoop struct with generic attributes Implements #102359. Split the `MLoop` struct into two separate integer arrays called `corner_verts` and `corner_edges`, referring to the vertex each corner is attached to and the next edge around the face at each corner. These arrays can be sliced to give access to the edges or vertices in a face. Then they are often referred to as "poly_verts" or "poly_edges". The main benefits are halving the necessary memory bandwidth when only one array is used and simplifications from using regular integer indices instead of a special-purpose struct. The commit also starts a renaming from "loop" to "corner" in mesh code. Like the other mesh struct of array refactors, forward compatibility is kept by writing files with the older format. This will be done until 4.0 to ease the transition process. Looking at a small portion of the patch should give a good impression for the rest of the changes. I tried to make the changes as small as possible so it's easy to tell the correctness from the diff. Though I found Blender developers have been very inventive over the last decade when finding different ways to loop over the corners in a face. For performance, nearly every piece of code that deals with `Mesh` is slightly impacted. Any algorithm that is memory bottle-necked should see an improvement. For example, here is a comparison of interpolating a vertex float attribute to face corners (Ryzen 3700x): **Before** (Average: 3.7 ms, Min: 3.4 ms) ``` threading::parallel_for(loops.index_range(), 4096, [&](IndexRange range) { for (const int64_t i : range) { dst[i] = src[loops[i].v]; } }); ``` **After** (Average: 2.9 ms, Min: 2.6 ms) ``` array_utils::gather(src, corner_verts, dst); ``` That's an improvement of 28% to the average timings, and it's also a simplification, since an index-based routine can be used instead. For more examples using the new arrays, see the design task. Pull Request: https://projects.blender.org/blender/blender/pulls/104424
2023-03-20 15:55:13 +01:00
mixer.mix_in(0, varray[vert]);
}
mixer.finalize();
return return_value;
});
}
}
});
Geometry Nodes: use virtual arrays in internal attribute api A virtual array is a data structure that is similar to a normal array in that its elements can be accessed by an index. However, a virtual array does not have to be a contiguous array internally. Instead, its elements can be layed out arbitrarily while element access happens through a virtual function call. However, the virtual array data structures are designed so that the virtual function call can be avoided in cases where it could become a bottleneck. Most commonly, a virtual array is backed by an actual array/span or is a single value internally, that is the same for every index. Besides those, there are many more specialized virtual arrays like the ones that provides vertex positions based on the `MVert` struct or vertex group weights. Not all attributes used by geometry nodes are stored in simple contiguous arrays. To provide uniform access to all kinds of attributes, the attribute API has to provide virtual array functionality that hides the implementation details of attributes. Before this refactor, the attribute API provided its own virtual array implementation as part of the `ReadAttribute` and `WriteAttribute` types. That resulted in unnecessary code duplication with the virtual array system. Even worse, it bound many algorithms used by geometry nodes to the specifics of the attribute API, even though they could also use different data sources (such as data from sockets, default values, later results of expressions, ...). This refactor removes the `ReadAttribute` and `WriteAttribute` types and replaces them with `GVArray` and `GVMutableArray` respectively. The `GV` stands for "generic virtual". The "generic" means that the data type contained in those virtual arrays is only known at run-time. There are the corresponding statically typed types `VArray<T>` and `VMutableArray<T>` as well. No regressions are expected from this refactor. It does come with one improvement for users. The attribute API can convert the data type on write now. This is especially useful when writing to builtin attributes like `material_index` with e.g. the Attribute Math node (which usually just writes to float attributes, while `material_index` is an integer attribute). Differential Revision: https://developer.blender.org/D10994
2021-04-17 16:41:03 +02:00
return new_varray;
}
Geometry Nodes: refactor virtual array system Goals of this refactor: * Simplify creating virtual arrays. * Simplify passing virtual arrays around. * Simplify converting between typed and generic virtual arrays. * Reduce memory allocations. As a quick reminder, a virtual arrays is a data structure that behaves like an array (i.e. it can be accessed using an index). However, it may not actually be stored as array internally. The two most important implementations of virtual arrays are those that correspond to an actual plain array and those that have the same value for every index. However, many more implementations exist for various reasons (interfacing with legacy attributes, unified iterator over all points in multiple splines, ...). With this refactor the core types (`VArray`, `GVArray`, `VMutableArray` and `GVMutableArray`) can be used like "normal values". They typically live on the stack. Before, they were usually inside a `std::unique_ptr`. This makes passing them around much easier. Creation of new virtual arrays is also much simpler now due to some constructors. Memory allocations are reduced by making use of small object optimization inside the core types. Previously, `VArray` was a class with virtual methods that had to be overridden to change the behavior of a the virtual array. Now,`VArray` has a fixed size and has no virtual methods. Instead it contains a `VArrayImpl` that is similar to the old `VArray`. `VArrayImpl` should rarely ever be used directly, unless a new virtual array implementation is added. To support the small object optimization for many `VArrayImpl` classes, a new `blender::Any` type is added. It is similar to `std::any` with two additional features. It has an adjustable inline buffer size and alignment. The inline buffer size of `std::any` can't be relied on and is usually too small for our use case here. Furthermore, `blender::Any` can store additional user-defined type information without increasing the stack size. Differential Revision: https://developer.blender.org/D12986
2021-11-16 10:15:51 +01:00
static GVArray adapt_mesh_domain_point_to_edge(const Mesh &mesh, const GVArray &varray)
{
Mesh: Move edges to a generic attribute Implements #95966, as the final step of #95965. This commit changes the storage of mesh edge vertex indices from the `MEdge` type to the generic `int2` attribute type. This follows the general design for geometry and the attribute system, where the data storage type and the usage semantics are separated. The main benefit of the change is reduced memory usage-- the requirements of storing mesh edges is reduced by 1/3. For example, this saves 8MB on a 1 million vertex grid. This also gives performance benefits to any memory-bound mesh processing algorithm that uses edges. Another benefit is that all of the edge's vertex indices are contiguous. In a few cases, it's helpful to process all of them as `Span<int>` rather than `Span<int2>`. Similarly, the type is more likely to match a generic format used by a library, or code that shouldn't know about specific Blender `Mesh` types. Various Notes: - The `.edge_verts` name is used to reflect a mapping between domains, similar to `.corner_verts`, etc. The period means that it the data shouldn't change arbitrarily by the user or procedural operations. - `edge[0]` is now used instead of `edge.v1` - Signed integers are used instead of unsigned to reduce the mixing of signed-ness, which can be error prone. - All of the previously used core mesh data types (`MVert`, `MEdge`, `MLoop`, `MPoly` are now deprecated. Only generic types are used). - The `vec2i` DNA type is used in the few C files where necessary. Pull Request: https://projects.blender.org/blender/blender/pulls/106638
2023-04-17 13:47:41 +02:00
const Span<int2> edges = mesh.edges();
Mesh: Remove redundant custom data pointers For copy-on-write, we want to share attribute arrays between meshes where possible. Mutable pointers like `Mesh.mvert` make that difficult by making ownership vague. They also make code more complex by adding redundancy. The simplest solution is just removing them and retrieving layers from `CustomData` as needed. Similar changes have already been applied to curves and point clouds (e9f82d3dc7ee, 410a6efb747f). Removing use of the pointers generally makes code more obvious and more reusable. Mesh data is now accessed with a C++ API (`Mesh::edges()` or `Mesh::edges_for_write()`), and a C API (`BKE_mesh_edges(mesh)`). The CoW changes this commit makes possible are described in T95845 and T95842, and started in D14139 and D14140. The change also simplifies the ongoing mesh struct-of-array refactors from T95965. **RNA/Python Access Performance** Theoretically, accessing mesh elements with the RNA API may become slower, since the layer needs to be found on every random access. However, overhead is already high enough that this doesn't make a noticible differenc, and performance is actually improved in some cases. Random access can be up to 10% faster, but other situations might be a bit slower. Generally using `foreach_get/set` are the best way to improve performance. See the differential revision for more discussion about Python performance. Cycles has been updated to use raw pointers and the internal Blender mesh types, mostly because there is no sense in having this overhead when it's already compiled with Blender. In my tests this roughly halves the Cycles mesh creation time (0.19s to 0.10s for a 1 million face grid). Differential Revision: https://developer.blender.org/D15488
2022-09-05 11:56:34 -05:00
Geometry Nodes: refactor virtual array system Goals of this refactor: * Simplify creating virtual arrays. * Simplify passing virtual arrays around. * Simplify converting between typed and generic virtual arrays. * Reduce memory allocations. As a quick reminder, a virtual arrays is a data structure that behaves like an array (i.e. it can be accessed using an index). However, it may not actually be stored as array internally. The two most important implementations of virtual arrays are those that correspond to an actual plain array and those that have the same value for every index. However, many more implementations exist for various reasons (interfacing with legacy attributes, unified iterator over all points in multiple splines, ...). With this refactor the core types (`VArray`, `GVArray`, `VMutableArray` and `GVMutableArray`) can be used like "normal values". They typically live on the stack. Before, they were usually inside a `std::unique_ptr`. This makes passing them around much easier. Creation of new virtual arrays is also much simpler now due to some constructors. Memory allocations are reduced by making use of small object optimization inside the core types. Previously, `VArray` was a class with virtual methods that had to be overridden to change the behavior of a the virtual array. Now,`VArray` has a fixed size and has no virtual methods. Instead it contains a `VArrayImpl` that is similar to the old `VArray`. `VArrayImpl` should rarely ever be used directly, unless a new virtual array implementation is added. To support the small object optimization for many `VArrayImpl` classes, a new `blender::Any` type is added. It is similar to `std::any` with two additional features. It has an adjustable inline buffer size and alignment. The inline buffer size of `std::any` can't be relied on and is usually too small for our use case here. Furthermore, `blender::Any` can store additional user-defined type information without increasing the stack size. Differential Revision: https://developer.blender.org/D12986
2021-11-16 10:15:51 +01:00
GVArray new_varray;
attribute_math::convert_to_static_type(varray.type(), [&](auto dummy) {
using T = decltype(dummy);
if constexpr (!std::is_void_v<attribute_math::DefaultMixer<T>>) {
if constexpr (std::is_same_v<T, bool>) {
/* An edge is selected if both of its vertices were selected. */
new_varray = VArray<bool>::ForFunc(
Mesh: Remove redundant custom data pointers For copy-on-write, we want to share attribute arrays between meshes where possible. Mutable pointers like `Mesh.mvert` make that difficult by making ownership vague. They also make code more complex by adding redundancy. The simplest solution is just removing them and retrieving layers from `CustomData` as needed. Similar changes have already been applied to curves and point clouds (e9f82d3dc7ee, 410a6efb747f). Removing use of the pointers generally makes code more obvious and more reusable. Mesh data is now accessed with a C++ API (`Mesh::edges()` or `Mesh::edges_for_write()`), and a C API (`BKE_mesh_edges(mesh)`). The CoW changes this commit makes possible are described in T95845 and T95842, and started in D14139 and D14140. The change also simplifies the ongoing mesh struct-of-array refactors from T95965. **RNA/Python Access Performance** Theoretically, accessing mesh elements with the RNA API may become slower, since the layer needs to be found on every random access. However, overhead is already high enough that this doesn't make a noticible differenc, and performance is actually improved in some cases. Random access can be up to 10% faster, but other situations might be a bit slower. Generally using `foreach_get/set` are the best way to improve performance. See the differential revision for more discussion about Python performance. Cycles has been updated to use raw pointers and the internal Blender mesh types, mostly because there is no sense in having this overhead when it's already compiled with Blender. In my tests this roughly halves the Cycles mesh creation time (0.19s to 0.10s for a 1 million face grid). Differential Revision: https://developer.blender.org/D15488
2022-09-05 11:56:34 -05:00
edges.size(), [edges, varray = varray.typed<bool>()](const int edge_index) {
Mesh: Move edges to a generic attribute Implements #95966, as the final step of #95965. This commit changes the storage of mesh edge vertex indices from the `MEdge` type to the generic `int2` attribute type. This follows the general design for geometry and the attribute system, where the data storage type and the usage semantics are separated. The main benefit of the change is reduced memory usage-- the requirements of storing mesh edges is reduced by 1/3. For example, this saves 8MB on a 1 million vertex grid. This also gives performance benefits to any memory-bound mesh processing algorithm that uses edges. Another benefit is that all of the edge's vertex indices are contiguous. In a few cases, it's helpful to process all of them as `Span<int>` rather than `Span<int2>`. Similarly, the type is more likely to match a generic format used by a library, or code that shouldn't know about specific Blender `Mesh` types. Various Notes: - The `.edge_verts` name is used to reflect a mapping between domains, similar to `.corner_verts`, etc. The period means that it the data shouldn't change arbitrarily by the user or procedural operations. - `edge[0]` is now used instead of `edge.v1` - Signed integers are used instead of unsigned to reduce the mixing of signed-ness, which can be error prone. - All of the previously used core mesh data types (`MVert`, `MEdge`, `MLoop`, `MPoly` are now deprecated. Only generic types are used). - The `vec2i` DNA type is used in the few C files where necessary. Pull Request: https://projects.blender.org/blender/blender/pulls/106638
2023-04-17 13:47:41 +02:00
const int2 &edge = edges[edge_index];
return varray[edge[0]] && varray[edge[1]];
});
}
else {
new_varray = VArray<T>::ForFunc(
Mesh: Remove redundant custom data pointers For copy-on-write, we want to share attribute arrays between meshes where possible. Mutable pointers like `Mesh.mvert` make that difficult by making ownership vague. They also make code more complex by adding redundancy. The simplest solution is just removing them and retrieving layers from `CustomData` as needed. Similar changes have already been applied to curves and point clouds (e9f82d3dc7ee, 410a6efb747f). Removing use of the pointers generally makes code more obvious and more reusable. Mesh data is now accessed with a C++ API (`Mesh::edges()` or `Mesh::edges_for_write()`), and a C API (`BKE_mesh_edges(mesh)`). The CoW changes this commit makes possible are described in T95845 and T95842, and started in D14139 and D14140. The change also simplifies the ongoing mesh struct-of-array refactors from T95965. **RNA/Python Access Performance** Theoretically, accessing mesh elements with the RNA API may become slower, since the layer needs to be found on every random access. However, overhead is already high enough that this doesn't make a noticible differenc, and performance is actually improved in some cases. Random access can be up to 10% faster, but other situations might be a bit slower. Generally using `foreach_get/set` are the best way to improve performance. See the differential revision for more discussion about Python performance. Cycles has been updated to use raw pointers and the internal Blender mesh types, mostly because there is no sense in having this overhead when it's already compiled with Blender. In my tests this roughly halves the Cycles mesh creation time (0.19s to 0.10s for a 1 million face grid). Differential Revision: https://developer.blender.org/D15488
2022-09-05 11:56:34 -05:00
edges.size(), [edges, varray = varray.typed<T>()](const int edge_index) {
T return_value;
attribute_math::DefaultMixer<T> mixer({&return_value, 1});
Mesh: Move edges to a generic attribute Implements #95966, as the final step of #95965. This commit changes the storage of mesh edge vertex indices from the `MEdge` type to the generic `int2` attribute type. This follows the general design for geometry and the attribute system, where the data storage type and the usage semantics are separated. The main benefit of the change is reduced memory usage-- the requirements of storing mesh edges is reduced by 1/3. For example, this saves 8MB on a 1 million vertex grid. This also gives performance benefits to any memory-bound mesh processing algorithm that uses edges. Another benefit is that all of the edge's vertex indices are contiguous. In a few cases, it's helpful to process all of them as `Span<int>` rather than `Span<int2>`. Similarly, the type is more likely to match a generic format used by a library, or code that shouldn't know about specific Blender `Mesh` types. Various Notes: - The `.edge_verts` name is used to reflect a mapping between domains, similar to `.corner_verts`, etc. The period means that it the data shouldn't change arbitrarily by the user or procedural operations. - `edge[0]` is now used instead of `edge.v1` - Signed integers are used instead of unsigned to reduce the mixing of signed-ness, which can be error prone. - All of the previously used core mesh data types (`MVert`, `MEdge`, `MLoop`, `MPoly` are now deprecated. Only generic types are used). - The `vec2i` DNA type is used in the few C files where necessary. Pull Request: https://projects.blender.org/blender/blender/pulls/106638
2023-04-17 13:47:41 +02:00
const int2 &edge = edges[edge_index];
mixer.mix_in(0, varray[edge[0]]);
mixer.mix_in(0, varray[edge[1]]);
mixer.finalize();
return return_value;
});
}
}
});
Geometry Nodes: use virtual arrays in internal attribute api A virtual array is a data structure that is similar to a normal array in that its elements can be accessed by an index. However, a virtual array does not have to be a contiguous array internally. Instead, its elements can be layed out arbitrarily while element access happens through a virtual function call. However, the virtual array data structures are designed so that the virtual function call can be avoided in cases where it could become a bottleneck. Most commonly, a virtual array is backed by an actual array/span or is a single value internally, that is the same for every index. Besides those, there are many more specialized virtual arrays like the ones that provides vertex positions based on the `MVert` struct or vertex group weights. Not all attributes used by geometry nodes are stored in simple contiguous arrays. To provide uniform access to all kinds of attributes, the attribute API has to provide virtual array functionality that hides the implementation details of attributes. Before this refactor, the attribute API provided its own virtual array implementation as part of the `ReadAttribute` and `WriteAttribute` types. That resulted in unnecessary code duplication with the virtual array system. Even worse, it bound many algorithms used by geometry nodes to the specifics of the attribute API, even though they could also use different data sources (such as data from sockets, default values, later results of expressions, ...). This refactor removes the `ReadAttribute` and `WriteAttribute` types and replaces them with `GVArray` and `GVMutableArray` respectively. The `GV` stands for "generic virtual". The "generic" means that the data type contained in those virtual arrays is only known at run-time. There are the corresponding statically typed types `VArray<T>` and `VMutableArray<T>` as well. No regressions are expected from this refactor. It does come with one improvement for users. The attribute API can convert the data type on write now. This is especially useful when writing to builtin attributes like `material_index` with e.g. the Attribute Math node (which usually just writes to float attributes, while `material_index` is an integer attribute). Differential Revision: https://developer.blender.org/D10994
2021-04-17 16:41:03 +02:00
return new_varray;
}
template<typename T>
void adapt_mesh_domain_edge_to_corner_impl(const Mesh &mesh,
Geometry Nodes: use virtual arrays in internal attribute api A virtual array is a data structure that is similar to a normal array in that its elements can be accessed by an index. However, a virtual array does not have to be a contiguous array internally. Instead, its elements can be layed out arbitrarily while element access happens through a virtual function call. However, the virtual array data structures are designed so that the virtual function call can be avoided in cases where it could become a bottleneck. Most commonly, a virtual array is backed by an actual array/span or is a single value internally, that is the same for every index. Besides those, there are many more specialized virtual arrays like the ones that provides vertex positions based on the `MVert` struct or vertex group weights. Not all attributes used by geometry nodes are stored in simple contiguous arrays. To provide uniform access to all kinds of attributes, the attribute API has to provide virtual array functionality that hides the implementation details of attributes. Before this refactor, the attribute API provided its own virtual array implementation as part of the `ReadAttribute` and `WriteAttribute` types. That resulted in unnecessary code duplication with the virtual array system. Even worse, it bound many algorithms used by geometry nodes to the specifics of the attribute API, even though they could also use different data sources (such as data from sockets, default values, later results of expressions, ...). This refactor removes the `ReadAttribute` and `WriteAttribute` types and replaces them with `GVArray` and `GVMutableArray` respectively. The `GV` stands for "generic virtual". The "generic" means that the data type contained in those virtual arrays is only known at run-time. There are the corresponding statically typed types `VArray<T>` and `VMutableArray<T>` as well. No regressions are expected from this refactor. It does come with one improvement for users. The attribute API can convert the data type on write now. This is especially useful when writing to builtin attributes like `material_index` with e.g. the Attribute Math node (which usually just writes to float attributes, while `material_index` is an integer attribute). Differential Revision: https://developer.blender.org/D10994
2021-04-17 16:41:03 +02:00
const VArray<T> &old_values,
MutableSpan<T> r_values)
{
BLI_assert(r_values.size() == mesh.totloop);
Mesh: Replace MPoly struct with offset indices Implements #95967. Currently the `MPoly` struct is 12 bytes, and stores the index of a face's first corner and the number of corners/verts/edges. Polygons and corners are always created in order by Blender, meaning each face's corners will be after the previous face's corners. We can take advantage of this fact and eliminate the redundancy in mesh face storage by only storing a single integer corner offset for each face. The size of the face is then encoded by the offset of the next face. The size of a single integer is 4 bytes, so this reduces memory usage by 3 times. The same method is used for `CurvesGeometry`, so Blender already has an abstraction to simplify using these offsets called `OffsetIndices`. This class is used to easily retrieve a range of corner indices for each face. This also gives the opportunity for sharing some logic with curves. Another benefit of the change is that the offsets and sizes stored in `MPoly` can no longer disagree with each other. Storing faces in the order of their corners can simplify some code too. Face/polygon variables now use the `IndexRange` type, which comes with quite a few utilities that can simplify code. Some: - The offset integer array has to be one longer than the face count to avoid a branch for every face, which means the data is no longer part of the mesh's `CustomData`. - We lose the ability to "reference" an original mesh's offset array until more reusable CoW from #104478 is committed. That will be added in a separate commit. - Since they aren't part of `CustomData`, poly offsets often have to be copied manually. - To simplify using `OffsetIndices` in many places, some functions and structs in headers were moved to only compile in C++. - All meshes created by Blender use the same order for faces and face corners, but just in case, meshes with mismatched order are fixed by versioning code. - `MeshPolygon.totloop` is no longer editable in RNA. This API break is necessary here unfortunately. It should be worth it in 3.6, since that's the best way to allow loading meshes from 4.0, which is important for an LTS version. Pull Request: https://projects.blender.org/blender/blender/pulls/105938
2023-04-04 20:39:28 +02:00
const OffsetIndices polys = mesh.polys();
Mesh: Replace MLoop struct with generic attributes Implements #102359. Split the `MLoop` struct into two separate integer arrays called `corner_verts` and `corner_edges`, referring to the vertex each corner is attached to and the next edge around the face at each corner. These arrays can be sliced to give access to the edges or vertices in a face. Then they are often referred to as "poly_verts" or "poly_edges". The main benefits are halving the necessary memory bandwidth when only one array is used and simplifications from using regular integer indices instead of a special-purpose struct. The commit also starts a renaming from "loop" to "corner" in mesh code. Like the other mesh struct of array refactors, forward compatibility is kept by writing files with the older format. This will be done until 4.0 to ease the transition process. Looking at a small portion of the patch should give a good impression for the rest of the changes. I tried to make the changes as small as possible so it's easy to tell the correctness from the diff. Though I found Blender developers have been very inventive over the last decade when finding different ways to loop over the corners in a face. For performance, nearly every piece of code that deals with `Mesh` is slightly impacted. Any algorithm that is memory bottle-necked should see an improvement. For example, here is a comparison of interpolating a vertex float attribute to face corners (Ryzen 3700x): **Before** (Average: 3.7 ms, Min: 3.4 ms) ``` threading::parallel_for(loops.index_range(), 4096, [&](IndexRange range) { for (const int64_t i : range) { dst[i] = src[loops[i].v]; } }); ``` **After** (Average: 2.9 ms, Min: 2.6 ms) ``` array_utils::gather(src, corner_verts, dst); ``` That's an improvement of 28% to the average timings, and it's also a simplification, since an index-based routine can be used instead. For more examples using the new arrays, see the design task. Pull Request: https://projects.blender.org/blender/blender/pulls/104424
2023-03-20 15:55:13 +01:00
const Span<int> corner_edges = mesh.corner_edges();
Mesh: Remove redundant custom data pointers For copy-on-write, we want to share attribute arrays between meshes where possible. Mutable pointers like `Mesh.mvert` make that difficult by making ownership vague. They also make code more complex by adding redundancy. The simplest solution is just removing them and retrieving layers from `CustomData` as needed. Similar changes have already been applied to curves and point clouds (e9f82d3dc7ee, 410a6efb747f). Removing use of the pointers generally makes code more obvious and more reusable. Mesh data is now accessed with a C++ API (`Mesh::edges()` or `Mesh::edges_for_write()`), and a C API (`BKE_mesh_edges(mesh)`). The CoW changes this commit makes possible are described in T95845 and T95842, and started in D14139 and D14140. The change also simplifies the ongoing mesh struct-of-array refactors from T95965. **RNA/Python Access Performance** Theoretically, accessing mesh elements with the RNA API may become slower, since the layer needs to be found on every random access. However, overhead is already high enough that this doesn't make a noticible differenc, and performance is actually improved in some cases. Random access can be up to 10% faster, but other situations might be a bit slower. Generally using `foreach_get/set` are the best way to improve performance. See the differential revision for more discussion about Python performance. Cycles has been updated to use raw pointers and the internal Blender mesh types, mostly because there is no sense in having this overhead when it's already compiled with Blender. In my tests this roughly halves the Cycles mesh creation time (0.19s to 0.10s for a 1 million face grid). Differential Revision: https://developer.blender.org/D15488
2022-09-05 11:56:34 -05:00
attribute_math::DefaultMixer<T> mixer(r_values);
Mesh: Remove redundant custom data pointers For copy-on-write, we want to share attribute arrays between meshes where possible. Mutable pointers like `Mesh.mvert` make that difficult by making ownership vague. They also make code more complex by adding redundancy. The simplest solution is just removing them and retrieving layers from `CustomData` as needed. Similar changes have already been applied to curves and point clouds (e9f82d3dc7ee, 410a6efb747f). Removing use of the pointers generally makes code more obvious and more reusable. Mesh data is now accessed with a C++ API (`Mesh::edges()` or `Mesh::edges_for_write()`), and a C API (`BKE_mesh_edges(mesh)`). The CoW changes this commit makes possible are described in T95845 and T95842, and started in D14139 and D14140. The change also simplifies the ongoing mesh struct-of-array refactors from T95965. **RNA/Python Access Performance** Theoretically, accessing mesh elements with the RNA API may become slower, since the layer needs to be found on every random access. However, overhead is already high enough that this doesn't make a noticible differenc, and performance is actually improved in some cases. Random access can be up to 10% faster, but other situations might be a bit slower. Generally using `foreach_get/set` are the best way to improve performance. See the differential revision for more discussion about Python performance. Cycles has been updated to use raw pointers and the internal Blender mesh types, mostly because there is no sense in having this overhead when it's already compiled with Blender. In my tests this roughly halves the Cycles mesh creation time (0.19s to 0.10s for a 1 million face grid). Differential Revision: https://developer.blender.org/D15488
2022-09-05 11:56:34 -05:00
for (const int poly_index : polys.index_range()) {
Mesh: Replace MPoly struct with offset indices Implements #95967. Currently the `MPoly` struct is 12 bytes, and stores the index of a face's first corner and the number of corners/verts/edges. Polygons and corners are always created in order by Blender, meaning each face's corners will be after the previous face's corners. We can take advantage of this fact and eliminate the redundancy in mesh face storage by only storing a single integer corner offset for each face. The size of the face is then encoded by the offset of the next face. The size of a single integer is 4 bytes, so this reduces memory usage by 3 times. The same method is used for `CurvesGeometry`, so Blender already has an abstraction to simplify using these offsets called `OffsetIndices`. This class is used to easily retrieve a range of corner indices for each face. This also gives the opportunity for sharing some logic with curves. Another benefit of the change is that the offsets and sizes stored in `MPoly` can no longer disagree with each other. Storing faces in the order of their corners can simplify some code too. Face/polygon variables now use the `IndexRange` type, which comes with quite a few utilities that can simplify code. Some: - The offset integer array has to be one longer than the face count to avoid a branch for every face, which means the data is no longer part of the mesh's `CustomData`. - We lose the ability to "reference" an original mesh's offset array until more reusable CoW from #104478 is committed. That will be added in a separate commit. - Since they aren't part of `CustomData`, poly offsets often have to be copied manually. - To simplify using `OffsetIndices` in many places, some functions and structs in headers were moved to only compile in C++. - All meshes created by Blender use the same order for faces and face corners, but just in case, meshes with mismatched order are fixed by versioning code. - `MeshPolygon.totloop` is no longer editable in RNA. This API break is necessary here unfortunately. It should be worth it in 3.6, since that's the best way to allow loading meshes from 4.0, which is important for an LTS version. Pull Request: https://projects.blender.org/blender/blender/pulls/105938
2023-04-04 20:39:28 +02:00
const IndexRange poly = polys[poly_index];
/* For every corner, mix the values from the adjacent edges on the face. */
Mesh: Replace MPoly struct with offset indices Implements #95967. Currently the `MPoly` struct is 12 bytes, and stores the index of a face's first corner and the number of corners/verts/edges. Polygons and corners are always created in order by Blender, meaning each face's corners will be after the previous face's corners. We can take advantage of this fact and eliminate the redundancy in mesh face storage by only storing a single integer corner offset for each face. The size of the face is then encoded by the offset of the next face. The size of a single integer is 4 bytes, so this reduces memory usage by 3 times. The same method is used for `CurvesGeometry`, so Blender already has an abstraction to simplify using these offsets called `OffsetIndices`. This class is used to easily retrieve a range of corner indices for each face. This also gives the opportunity for sharing some logic with curves. Another benefit of the change is that the offsets and sizes stored in `MPoly` can no longer disagree with each other. Storing faces in the order of their corners can simplify some code too. Face/polygon variables now use the `IndexRange` type, which comes with quite a few utilities that can simplify code. Some: - The offset integer array has to be one longer than the face count to avoid a branch for every face, which means the data is no longer part of the mesh's `CustomData`. - We lose the ability to "reference" an original mesh's offset array until more reusable CoW from #104478 is committed. That will be added in a separate commit. - Since they aren't part of `CustomData`, poly offsets often have to be copied manually. - To simplify using `OffsetIndices` in many places, some functions and structs in headers were moved to only compile in C++. - All meshes created by Blender use the same order for faces and face corners, but just in case, meshes with mismatched order are fixed by versioning code. - `MeshPolygon.totloop` is no longer editable in RNA. This API break is necessary here unfortunately. It should be worth it in 3.6, since that's the best way to allow loading meshes from 4.0, which is important for an LTS version. Pull Request: https://projects.blender.org/blender/blender/pulls/105938
2023-04-04 20:39:28 +02:00
for (const int loop_index : poly) {
const int loop_index_prev = mesh::poly_corner_prev(poly, loop_index);
Mesh: Replace MLoop struct with generic attributes Implements #102359. Split the `MLoop` struct into two separate integer arrays called `corner_verts` and `corner_edges`, referring to the vertex each corner is attached to and the next edge around the face at each corner. These arrays can be sliced to give access to the edges or vertices in a face. Then they are often referred to as "poly_verts" or "poly_edges". The main benefits are halving the necessary memory bandwidth when only one array is used and simplifications from using regular integer indices instead of a special-purpose struct. The commit also starts a renaming from "loop" to "corner" in mesh code. Like the other mesh struct of array refactors, forward compatibility is kept by writing files with the older format. This will be done until 4.0 to ease the transition process. Looking at a small portion of the patch should give a good impression for the rest of the changes. I tried to make the changes as small as possible so it's easy to tell the correctness from the diff. Though I found Blender developers have been very inventive over the last decade when finding different ways to loop over the corners in a face. For performance, nearly every piece of code that deals with `Mesh` is slightly impacted. Any algorithm that is memory bottle-necked should see an improvement. For example, here is a comparison of interpolating a vertex float attribute to face corners (Ryzen 3700x): **Before** (Average: 3.7 ms, Min: 3.4 ms) ``` threading::parallel_for(loops.index_range(), 4096, [&](IndexRange range) { for (const int64_t i : range) { dst[i] = src[loops[i].v]; } }); ``` **After** (Average: 2.9 ms, Min: 2.6 ms) ``` array_utils::gather(src, corner_verts, dst); ``` That's an improvement of 28% to the average timings, and it's also a simplification, since an index-based routine can be used instead. For more examples using the new arrays, see the design task. Pull Request: https://projects.blender.org/blender/blender/pulls/104424
2023-03-20 15:55:13 +01:00
const int edge = corner_edges[loop_index];
const int edge_prev = corner_edges[loop_index_prev];
mixer.mix_in(loop_index, old_values[edge]);
mixer.mix_in(loop_index, old_values[edge_prev]);
}
}
mixer.finalize();
}
/* A corner is selected if its two adjacent edges were selected. */
template<>
void adapt_mesh_domain_edge_to_corner_impl(const Mesh &mesh,
const VArray<bool> &old_values,
MutableSpan<bool> r_values)
{
BLI_assert(r_values.size() == mesh.totloop);
Mesh: Replace MPoly struct with offset indices Implements #95967. Currently the `MPoly` struct is 12 bytes, and stores the index of a face's first corner and the number of corners/verts/edges. Polygons and corners are always created in order by Blender, meaning each face's corners will be after the previous face's corners. We can take advantage of this fact and eliminate the redundancy in mesh face storage by only storing a single integer corner offset for each face. The size of the face is then encoded by the offset of the next face. The size of a single integer is 4 bytes, so this reduces memory usage by 3 times. The same method is used for `CurvesGeometry`, so Blender already has an abstraction to simplify using these offsets called `OffsetIndices`. This class is used to easily retrieve a range of corner indices for each face. This also gives the opportunity for sharing some logic with curves. Another benefit of the change is that the offsets and sizes stored in `MPoly` can no longer disagree with each other. Storing faces in the order of their corners can simplify some code too. Face/polygon variables now use the `IndexRange` type, which comes with quite a few utilities that can simplify code. Some: - The offset integer array has to be one longer than the face count to avoid a branch for every face, which means the data is no longer part of the mesh's `CustomData`. - We lose the ability to "reference" an original mesh's offset array until more reusable CoW from #104478 is committed. That will be added in a separate commit. - Since they aren't part of `CustomData`, poly offsets often have to be copied manually. - To simplify using `OffsetIndices` in many places, some functions and structs in headers were moved to only compile in C++. - All meshes created by Blender use the same order for faces and face corners, but just in case, meshes with mismatched order are fixed by versioning code. - `MeshPolygon.totloop` is no longer editable in RNA. This API break is necessary here unfortunately. It should be worth it in 3.6, since that's the best way to allow loading meshes from 4.0, which is important for an LTS version. Pull Request: https://projects.blender.org/blender/blender/pulls/105938
2023-04-04 20:39:28 +02:00
const OffsetIndices polys = mesh.polys();
Mesh: Replace MLoop struct with generic attributes Implements #102359. Split the `MLoop` struct into two separate integer arrays called `corner_verts` and `corner_edges`, referring to the vertex each corner is attached to and the next edge around the face at each corner. These arrays can be sliced to give access to the edges or vertices in a face. Then they are often referred to as "poly_verts" or "poly_edges". The main benefits are halving the necessary memory bandwidth when only one array is used and simplifications from using regular integer indices instead of a special-purpose struct. The commit also starts a renaming from "loop" to "corner" in mesh code. Like the other mesh struct of array refactors, forward compatibility is kept by writing files with the older format. This will be done until 4.0 to ease the transition process. Looking at a small portion of the patch should give a good impression for the rest of the changes. I tried to make the changes as small as possible so it's easy to tell the correctness from the diff. Though I found Blender developers have been very inventive over the last decade when finding different ways to loop over the corners in a face. For performance, nearly every piece of code that deals with `Mesh` is slightly impacted. Any algorithm that is memory bottle-necked should see an improvement. For example, here is a comparison of interpolating a vertex float attribute to face corners (Ryzen 3700x): **Before** (Average: 3.7 ms, Min: 3.4 ms) ``` threading::parallel_for(loops.index_range(), 4096, [&](IndexRange range) { for (const int64_t i : range) { dst[i] = src[loops[i].v]; } }); ``` **After** (Average: 2.9 ms, Min: 2.6 ms) ``` array_utils::gather(src, corner_verts, dst); ``` That's an improvement of 28% to the average timings, and it's also a simplification, since an index-based routine can be used instead. For more examples using the new arrays, see the design task. Pull Request: https://projects.blender.org/blender/blender/pulls/104424
2023-03-20 15:55:13 +01:00
const Span<int> corner_edges = mesh.corner_edges();
r_values.fill(false);
threading::parallel_for(polys.index_range(), 2048, [&](const IndexRange range) {
for (const int poly_index : range) {
Mesh: Replace MPoly struct with offset indices Implements #95967. Currently the `MPoly` struct is 12 bytes, and stores the index of a face's first corner and the number of corners/verts/edges. Polygons and corners are always created in order by Blender, meaning each face's corners will be after the previous face's corners. We can take advantage of this fact and eliminate the redundancy in mesh face storage by only storing a single integer corner offset for each face. The size of the face is then encoded by the offset of the next face. The size of a single integer is 4 bytes, so this reduces memory usage by 3 times. The same method is used for `CurvesGeometry`, so Blender already has an abstraction to simplify using these offsets called `OffsetIndices`. This class is used to easily retrieve a range of corner indices for each face. This also gives the opportunity for sharing some logic with curves. Another benefit of the change is that the offsets and sizes stored in `MPoly` can no longer disagree with each other. Storing faces in the order of their corners can simplify some code too. Face/polygon variables now use the `IndexRange` type, which comes with quite a few utilities that can simplify code. Some: - The offset integer array has to be one longer than the face count to avoid a branch for every face, which means the data is no longer part of the mesh's `CustomData`. - We lose the ability to "reference" an original mesh's offset array until more reusable CoW from #104478 is committed. That will be added in a separate commit. - Since they aren't part of `CustomData`, poly offsets often have to be copied manually. - To simplify using `OffsetIndices` in many places, some functions and structs in headers were moved to only compile in C++. - All meshes created by Blender use the same order for faces and face corners, but just in case, meshes with mismatched order are fixed by versioning code. - `MeshPolygon.totloop` is no longer editable in RNA. This API break is necessary here unfortunately. It should be worth it in 3.6, since that's the best way to allow loading meshes from 4.0, which is important for an LTS version. Pull Request: https://projects.blender.org/blender/blender/pulls/105938
2023-04-04 20:39:28 +02:00
const IndexRange poly = polys[poly_index];
for (const int loop_index : poly) {
const int loop_index_prev = mesh::poly_corner_prev(poly, loop_index);
Mesh: Replace MLoop struct with generic attributes Implements #102359. Split the `MLoop` struct into two separate integer arrays called `corner_verts` and `corner_edges`, referring to the vertex each corner is attached to and the next edge around the face at each corner. These arrays can be sliced to give access to the edges or vertices in a face. Then they are often referred to as "poly_verts" or "poly_edges". The main benefits are halving the necessary memory bandwidth when only one array is used and simplifications from using regular integer indices instead of a special-purpose struct. The commit also starts a renaming from "loop" to "corner" in mesh code. Like the other mesh struct of array refactors, forward compatibility is kept by writing files with the older format. This will be done until 4.0 to ease the transition process. Looking at a small portion of the patch should give a good impression for the rest of the changes. I tried to make the changes as small as possible so it's easy to tell the correctness from the diff. Though I found Blender developers have been very inventive over the last decade when finding different ways to loop over the corners in a face. For performance, nearly every piece of code that deals with `Mesh` is slightly impacted. Any algorithm that is memory bottle-necked should see an improvement. For example, here is a comparison of interpolating a vertex float attribute to face corners (Ryzen 3700x): **Before** (Average: 3.7 ms, Min: 3.4 ms) ``` threading::parallel_for(loops.index_range(), 4096, [&](IndexRange range) { for (const int64_t i : range) { dst[i] = src[loops[i].v]; } }); ``` **After** (Average: 2.9 ms, Min: 2.6 ms) ``` array_utils::gather(src, corner_verts, dst); ``` That's an improvement of 28% to the average timings, and it's also a simplification, since an index-based routine can be used instead. For more examples using the new arrays, see the design task. Pull Request: https://projects.blender.org/blender/blender/pulls/104424
2023-03-20 15:55:13 +01:00
const int edge = corner_edges[loop_index];
const int edge_prev = corner_edges[loop_index_prev];
if (old_values[edge] && old_values[edge_prev]) {
r_values[loop_index] = true;
}
}
}
});
}
Geometry Nodes: refactor virtual array system Goals of this refactor: * Simplify creating virtual arrays. * Simplify passing virtual arrays around. * Simplify converting between typed and generic virtual arrays. * Reduce memory allocations. As a quick reminder, a virtual arrays is a data structure that behaves like an array (i.e. it can be accessed using an index). However, it may not actually be stored as array internally. The two most important implementations of virtual arrays are those that correspond to an actual plain array and those that have the same value for every index. However, many more implementations exist for various reasons (interfacing with legacy attributes, unified iterator over all points in multiple splines, ...). With this refactor the core types (`VArray`, `GVArray`, `VMutableArray` and `GVMutableArray`) can be used like "normal values". They typically live on the stack. Before, they were usually inside a `std::unique_ptr`. This makes passing them around much easier. Creation of new virtual arrays is also much simpler now due to some constructors. Memory allocations are reduced by making use of small object optimization inside the core types. Previously, `VArray` was a class with virtual methods that had to be overridden to change the behavior of a the virtual array. Now,`VArray` has a fixed size and has no virtual methods. Instead it contains a `VArrayImpl` that is similar to the old `VArray`. `VArrayImpl` should rarely ever be used directly, unless a new virtual array implementation is added. To support the small object optimization for many `VArrayImpl` classes, a new `blender::Any` type is added. It is similar to `std::any` with two additional features. It has an adjustable inline buffer size and alignment. The inline buffer size of `std::any` can't be relied on and is usually too small for our use case here. Furthermore, `blender::Any` can store additional user-defined type information without increasing the stack size. Differential Revision: https://developer.blender.org/D12986
2021-11-16 10:15:51 +01:00
static GVArray adapt_mesh_domain_edge_to_corner(const Mesh &mesh, const GVArray &varray)
{
GArray<> values(varray.type(), mesh.totloop);
Geometry Nodes: refactor virtual array system Goals of this refactor: * Simplify creating virtual arrays. * Simplify passing virtual arrays around. * Simplify converting between typed and generic virtual arrays. * Reduce memory allocations. As a quick reminder, a virtual arrays is a data structure that behaves like an array (i.e. it can be accessed using an index). However, it may not actually be stored as array internally. The two most important implementations of virtual arrays are those that correspond to an actual plain array and those that have the same value for every index. However, many more implementations exist for various reasons (interfacing with legacy attributes, unified iterator over all points in multiple splines, ...). With this refactor the core types (`VArray`, `GVArray`, `VMutableArray` and `GVMutableArray`) can be used like "normal values". They typically live on the stack. Before, they were usually inside a `std::unique_ptr`. This makes passing them around much easier. Creation of new virtual arrays is also much simpler now due to some constructors. Memory allocations are reduced by making use of small object optimization inside the core types. Previously, `VArray` was a class with virtual methods that had to be overridden to change the behavior of a the virtual array. Now,`VArray` has a fixed size and has no virtual methods. Instead it contains a `VArrayImpl` that is similar to the old `VArray`. `VArrayImpl` should rarely ever be used directly, unless a new virtual array implementation is added. To support the small object optimization for many `VArrayImpl` classes, a new `blender::Any` type is added. It is similar to `std::any` with two additional features. It has an adjustable inline buffer size and alignment. The inline buffer size of `std::any` can't be relied on and is usually too small for our use case here. Furthermore, `blender::Any` can store additional user-defined type information without increasing the stack size. Differential Revision: https://developer.blender.org/D12986
2021-11-16 10:15:51 +01:00
attribute_math::convert_to_static_type(varray.type(), [&](auto dummy) {
using T = decltype(dummy);
if constexpr (!std::is_void_v<attribute_math::DefaultMixer<T>>) {
adapt_mesh_domain_edge_to_corner_impl<T>(
mesh, varray.typed<T>(), values.as_mutable_span().typed<T>());
}
});
return GVArray::ForGArray(std::move(values));
}
template<typename T>
static void adapt_mesh_domain_edge_to_point_impl(const Mesh &mesh,
Geometry Nodes: use virtual arrays in internal attribute api A virtual array is a data structure that is similar to a normal array in that its elements can be accessed by an index. However, a virtual array does not have to be a contiguous array internally. Instead, its elements can be layed out arbitrarily while element access happens through a virtual function call. However, the virtual array data structures are designed so that the virtual function call can be avoided in cases where it could become a bottleneck. Most commonly, a virtual array is backed by an actual array/span or is a single value internally, that is the same for every index. Besides those, there are many more specialized virtual arrays like the ones that provides vertex positions based on the `MVert` struct or vertex group weights. Not all attributes used by geometry nodes are stored in simple contiguous arrays. To provide uniform access to all kinds of attributes, the attribute API has to provide virtual array functionality that hides the implementation details of attributes. Before this refactor, the attribute API provided its own virtual array implementation as part of the `ReadAttribute` and `WriteAttribute` types. That resulted in unnecessary code duplication with the virtual array system. Even worse, it bound many algorithms used by geometry nodes to the specifics of the attribute API, even though they could also use different data sources (such as data from sockets, default values, later results of expressions, ...). This refactor removes the `ReadAttribute` and `WriteAttribute` types and replaces them with `GVArray` and `GVMutableArray` respectively. The `GV` stands for "generic virtual". The "generic" means that the data type contained in those virtual arrays is only known at run-time. There are the corresponding statically typed types `VArray<T>` and `VMutableArray<T>` as well. No regressions are expected from this refactor. It does come with one improvement for users. The attribute API can convert the data type on write now. This is especially useful when writing to builtin attributes like `material_index` with e.g. the Attribute Math node (which usually just writes to float attributes, while `material_index` is an integer attribute). Differential Revision: https://developer.blender.org/D10994
2021-04-17 16:41:03 +02:00
const VArray<T> &old_values,
MutableSpan<T> r_values)
{
BLI_assert(r_values.size() == mesh.totvert);
Mesh: Move edges to a generic attribute Implements #95966, as the final step of #95965. This commit changes the storage of mesh edge vertex indices from the `MEdge` type to the generic `int2` attribute type. This follows the general design for geometry and the attribute system, where the data storage type and the usage semantics are separated. The main benefit of the change is reduced memory usage-- the requirements of storing mesh edges is reduced by 1/3. For example, this saves 8MB on a 1 million vertex grid. This also gives performance benefits to any memory-bound mesh processing algorithm that uses edges. Another benefit is that all of the edge's vertex indices are contiguous. In a few cases, it's helpful to process all of them as `Span<int>` rather than `Span<int2>`. Similarly, the type is more likely to match a generic format used by a library, or code that shouldn't know about specific Blender `Mesh` types. Various Notes: - The `.edge_verts` name is used to reflect a mapping between domains, similar to `.corner_verts`, etc. The period means that it the data shouldn't change arbitrarily by the user or procedural operations. - `edge[0]` is now used instead of `edge.v1` - Signed integers are used instead of unsigned to reduce the mixing of signed-ness, which can be error prone. - All of the previously used core mesh data types (`MVert`, `MEdge`, `MLoop`, `MPoly` are now deprecated. Only generic types are used). - The `vec2i` DNA type is used in the few C files where necessary. Pull Request: https://projects.blender.org/blender/blender/pulls/106638
2023-04-17 13:47:41 +02:00
const Span<int2> edges = mesh.edges();
Mesh: Remove redundant custom data pointers For copy-on-write, we want to share attribute arrays between meshes where possible. Mutable pointers like `Mesh.mvert` make that difficult by making ownership vague. They also make code more complex by adding redundancy. The simplest solution is just removing them and retrieving layers from `CustomData` as needed. Similar changes have already been applied to curves and point clouds (e9f82d3dc7ee, 410a6efb747f). Removing use of the pointers generally makes code more obvious and more reusable. Mesh data is now accessed with a C++ API (`Mesh::edges()` or `Mesh::edges_for_write()`), and a C API (`BKE_mesh_edges(mesh)`). The CoW changes this commit makes possible are described in T95845 and T95842, and started in D14139 and D14140. The change also simplifies the ongoing mesh struct-of-array refactors from T95965. **RNA/Python Access Performance** Theoretically, accessing mesh elements with the RNA API may become slower, since the layer needs to be found on every random access. However, overhead is already high enough that this doesn't make a noticible differenc, and performance is actually improved in some cases. Random access can be up to 10% faster, but other situations might be a bit slower. Generally using `foreach_get/set` are the best way to improve performance. See the differential revision for more discussion about Python performance. Cycles has been updated to use raw pointers and the internal Blender mesh types, mostly because there is no sense in having this overhead when it's already compiled with Blender. In my tests this roughly halves the Cycles mesh creation time (0.19s to 0.10s for a 1 million face grid). Differential Revision: https://developer.blender.org/D15488
2022-09-05 11:56:34 -05:00
attribute_math::DefaultMixer<T> mixer(r_values);
for (const int edge_index : IndexRange(mesh.totedge)) {
Mesh: Move edges to a generic attribute Implements #95966, as the final step of #95965. This commit changes the storage of mesh edge vertex indices from the `MEdge` type to the generic `int2` attribute type. This follows the general design for geometry and the attribute system, where the data storage type and the usage semantics are separated. The main benefit of the change is reduced memory usage-- the requirements of storing mesh edges is reduced by 1/3. For example, this saves 8MB on a 1 million vertex grid. This also gives performance benefits to any memory-bound mesh processing algorithm that uses edges. Another benefit is that all of the edge's vertex indices are contiguous. In a few cases, it's helpful to process all of them as `Span<int>` rather than `Span<int2>`. Similarly, the type is more likely to match a generic format used by a library, or code that shouldn't know about specific Blender `Mesh` types. Various Notes: - The `.edge_verts` name is used to reflect a mapping between domains, similar to `.corner_verts`, etc. The period means that it the data shouldn't change arbitrarily by the user or procedural operations. - `edge[0]` is now used instead of `edge.v1` - Signed integers are used instead of unsigned to reduce the mixing of signed-ness, which can be error prone. - All of the previously used core mesh data types (`MVert`, `MEdge`, `MLoop`, `MPoly` are now deprecated. Only generic types are used). - The `vec2i` DNA type is used in the few C files where necessary. Pull Request: https://projects.blender.org/blender/blender/pulls/106638
2023-04-17 13:47:41 +02:00
const int2 &edge = edges[edge_index];
const T value = old_values[edge_index];
Mesh: Move edges to a generic attribute Implements #95966, as the final step of #95965. This commit changes the storage of mesh edge vertex indices from the `MEdge` type to the generic `int2` attribute type. This follows the general design for geometry and the attribute system, where the data storage type and the usage semantics are separated. The main benefit of the change is reduced memory usage-- the requirements of storing mesh edges is reduced by 1/3. For example, this saves 8MB on a 1 million vertex grid. This also gives performance benefits to any memory-bound mesh processing algorithm that uses edges. Another benefit is that all of the edge's vertex indices are contiguous. In a few cases, it's helpful to process all of them as `Span<int>` rather than `Span<int2>`. Similarly, the type is more likely to match a generic format used by a library, or code that shouldn't know about specific Blender `Mesh` types. Various Notes: - The `.edge_verts` name is used to reflect a mapping between domains, similar to `.corner_verts`, etc. The period means that it the data shouldn't change arbitrarily by the user or procedural operations. - `edge[0]` is now used instead of `edge.v1` - Signed integers are used instead of unsigned to reduce the mixing of signed-ness, which can be error prone. - All of the previously used core mesh data types (`MVert`, `MEdge`, `MLoop`, `MPoly` are now deprecated. Only generic types are used). - The `vec2i` DNA type is used in the few C files where necessary. Pull Request: https://projects.blender.org/blender/blender/pulls/106638
2023-04-17 13:47:41 +02:00
mixer.mix_in(edge[0], value);
mixer.mix_in(edge[1], value);
}
mixer.finalize();
}
/* A vertex is selected if any connected edge was selected. */
template<>
void adapt_mesh_domain_edge_to_point_impl(const Mesh &mesh,
const VArray<bool> &old_values,
MutableSpan<bool> r_values)
{
BLI_assert(r_values.size() == mesh.totvert);
Mesh: Move edges to a generic attribute Implements #95966, as the final step of #95965. This commit changes the storage of mesh edge vertex indices from the `MEdge` type to the generic `int2` attribute type. This follows the general design for geometry and the attribute system, where the data storage type and the usage semantics are separated. The main benefit of the change is reduced memory usage-- the requirements of storing mesh edges is reduced by 1/3. For example, this saves 8MB on a 1 million vertex grid. This also gives performance benefits to any memory-bound mesh processing algorithm that uses edges. Another benefit is that all of the edge's vertex indices are contiguous. In a few cases, it's helpful to process all of them as `Span<int>` rather than `Span<int2>`. Similarly, the type is more likely to match a generic format used by a library, or code that shouldn't know about specific Blender `Mesh` types. Various Notes: - The `.edge_verts` name is used to reflect a mapping between domains, similar to `.corner_verts`, etc. The period means that it the data shouldn't change arbitrarily by the user or procedural operations. - `edge[0]` is now used instead of `edge.v1` - Signed integers are used instead of unsigned to reduce the mixing of signed-ness, which can be error prone. - All of the previously used core mesh data types (`MVert`, `MEdge`, `MLoop`, `MPoly` are now deprecated. Only generic types are used). - The `vec2i` DNA type is used in the few C files where necessary. Pull Request: https://projects.blender.org/blender/blender/pulls/106638
2023-04-17 13:47:41 +02:00
const Span<int2> edges = mesh.edges();
/* Multiple threads can write to the same index here, but they are only
* writing true, and writing to single bytes is expected to be threadsafe. */
r_values.fill(false);
threading::parallel_for(edges.index_range(), 4096, [&](const IndexRange range) {
for (const int edge_index : range) {
if (old_values[edge_index]) {
Mesh: Move edges to a generic attribute Implements #95966, as the final step of #95965. This commit changes the storage of mesh edge vertex indices from the `MEdge` type to the generic `int2` attribute type. This follows the general design for geometry and the attribute system, where the data storage type and the usage semantics are separated. The main benefit of the change is reduced memory usage-- the requirements of storing mesh edges is reduced by 1/3. For example, this saves 8MB on a 1 million vertex grid. This also gives performance benefits to any memory-bound mesh processing algorithm that uses edges. Another benefit is that all of the edge's vertex indices are contiguous. In a few cases, it's helpful to process all of them as `Span<int>` rather than `Span<int2>`. Similarly, the type is more likely to match a generic format used by a library, or code that shouldn't know about specific Blender `Mesh` types. Various Notes: - The `.edge_verts` name is used to reflect a mapping between domains, similar to `.corner_verts`, etc. The period means that it the data shouldn't change arbitrarily by the user or procedural operations. - `edge[0]` is now used instead of `edge.v1` - Signed integers are used instead of unsigned to reduce the mixing of signed-ness, which can be error prone. - All of the previously used core mesh data types (`MVert`, `MEdge`, `MLoop`, `MPoly` are now deprecated. Only generic types are used). - The `vec2i` DNA type is used in the few C files where necessary. Pull Request: https://projects.blender.org/blender/blender/pulls/106638
2023-04-17 13:47:41 +02:00
const int2 &edge = edges[edge_index];
r_values[edge[0]] = true;
r_values[edge[1]] = true;
}
}
});
}
Geometry Nodes: refactor virtual array system Goals of this refactor: * Simplify creating virtual arrays. * Simplify passing virtual arrays around. * Simplify converting between typed and generic virtual arrays. * Reduce memory allocations. As a quick reminder, a virtual arrays is a data structure that behaves like an array (i.e. it can be accessed using an index). However, it may not actually be stored as array internally. The two most important implementations of virtual arrays are those that correspond to an actual plain array and those that have the same value for every index. However, many more implementations exist for various reasons (interfacing with legacy attributes, unified iterator over all points in multiple splines, ...). With this refactor the core types (`VArray`, `GVArray`, `VMutableArray` and `GVMutableArray`) can be used like "normal values". They typically live on the stack. Before, they were usually inside a `std::unique_ptr`. This makes passing them around much easier. Creation of new virtual arrays is also much simpler now due to some constructors. Memory allocations are reduced by making use of small object optimization inside the core types. Previously, `VArray` was a class with virtual methods that had to be overridden to change the behavior of a the virtual array. Now,`VArray` has a fixed size and has no virtual methods. Instead it contains a `VArrayImpl` that is similar to the old `VArray`. `VArrayImpl` should rarely ever be used directly, unless a new virtual array implementation is added. To support the small object optimization for many `VArrayImpl` classes, a new `blender::Any` type is added. It is similar to `std::any` with two additional features. It has an adjustable inline buffer size and alignment. The inline buffer size of `std::any` can't be relied on and is usually too small for our use case here. Furthermore, `blender::Any` can store additional user-defined type information without increasing the stack size. Differential Revision: https://developer.blender.org/D12986
2021-11-16 10:15:51 +01:00
static GVArray adapt_mesh_domain_edge_to_point(const Mesh &mesh, const GVArray &varray)
{
GArray<> values(varray.type(), mesh.totvert);
Geometry Nodes: refactor virtual array system Goals of this refactor: * Simplify creating virtual arrays. * Simplify passing virtual arrays around. * Simplify converting between typed and generic virtual arrays. * Reduce memory allocations. As a quick reminder, a virtual arrays is a data structure that behaves like an array (i.e. it can be accessed using an index). However, it may not actually be stored as array internally. The two most important implementations of virtual arrays are those that correspond to an actual plain array and those that have the same value for every index. However, many more implementations exist for various reasons (interfacing with legacy attributes, unified iterator over all points in multiple splines, ...). With this refactor the core types (`VArray`, `GVArray`, `VMutableArray` and `GVMutableArray`) can be used like "normal values". They typically live on the stack. Before, they were usually inside a `std::unique_ptr`. This makes passing them around much easier. Creation of new virtual arrays is also much simpler now due to some constructors. Memory allocations are reduced by making use of small object optimization inside the core types. Previously, `VArray` was a class with virtual methods that had to be overridden to change the behavior of a the virtual array. Now,`VArray` has a fixed size and has no virtual methods. Instead it contains a `VArrayImpl` that is similar to the old `VArray`. `VArrayImpl` should rarely ever be used directly, unless a new virtual array implementation is added. To support the small object optimization for many `VArrayImpl` classes, a new `blender::Any` type is added. It is similar to `std::any` with two additional features. It has an adjustable inline buffer size and alignment. The inline buffer size of `std::any` can't be relied on and is usually too small for our use case here. Furthermore, `blender::Any` can store additional user-defined type information without increasing the stack size. Differential Revision: https://developer.blender.org/D12986
2021-11-16 10:15:51 +01:00
attribute_math::convert_to_static_type(varray.type(), [&](auto dummy) {
using T = decltype(dummy);
if constexpr (!std::is_void_v<attribute_math::DefaultMixer<T>>) {
adapt_mesh_domain_edge_to_point_impl<T>(
mesh, varray.typed<T>(), values.as_mutable_span().typed<T>());
}
});
return GVArray::ForGArray(std::move(values));
}
Geometry Nodes: refactor virtual array system Goals of this refactor: * Simplify creating virtual arrays. * Simplify passing virtual arrays around. * Simplify converting between typed and generic virtual arrays. * Reduce memory allocations. As a quick reminder, a virtual arrays is a data structure that behaves like an array (i.e. it can be accessed using an index). However, it may not actually be stored as array internally. The two most important implementations of virtual arrays are those that correspond to an actual plain array and those that have the same value for every index. However, many more implementations exist for various reasons (interfacing with legacy attributes, unified iterator over all points in multiple splines, ...). With this refactor the core types (`VArray`, `GVArray`, `VMutableArray` and `GVMutableArray`) can be used like "normal values". They typically live on the stack. Before, they were usually inside a `std::unique_ptr`. This makes passing them around much easier. Creation of new virtual arrays is also much simpler now due to some constructors. Memory allocations are reduced by making use of small object optimization inside the core types. Previously, `VArray` was a class with virtual methods that had to be overridden to change the behavior of a the virtual array. Now,`VArray` has a fixed size and has no virtual methods. Instead it contains a `VArrayImpl` that is similar to the old `VArray`. `VArrayImpl` should rarely ever be used directly, unless a new virtual array implementation is added. To support the small object optimization for many `VArrayImpl` classes, a new `blender::Any` type is added. It is similar to `std::any` with two additional features. It has an adjustable inline buffer size and alignment. The inline buffer size of `std::any` can't be relied on and is usually too small for our use case here. Furthermore, `blender::Any` can store additional user-defined type information without increasing the stack size. Differential Revision: https://developer.blender.org/D12986
2021-11-16 10:15:51 +01:00
static GVArray adapt_mesh_domain_edge_to_face(const Mesh &mesh, const GVArray &varray)
{
Mesh: Replace MPoly struct with offset indices Implements #95967. Currently the `MPoly` struct is 12 bytes, and stores the index of a face's first corner and the number of corners/verts/edges. Polygons and corners are always created in order by Blender, meaning each face's corners will be after the previous face's corners. We can take advantage of this fact and eliminate the redundancy in mesh face storage by only storing a single integer corner offset for each face. The size of the face is then encoded by the offset of the next face. The size of a single integer is 4 bytes, so this reduces memory usage by 3 times. The same method is used for `CurvesGeometry`, so Blender already has an abstraction to simplify using these offsets called `OffsetIndices`. This class is used to easily retrieve a range of corner indices for each face. This also gives the opportunity for sharing some logic with curves. Another benefit of the change is that the offsets and sizes stored in `MPoly` can no longer disagree with each other. Storing faces in the order of their corners can simplify some code too. Face/polygon variables now use the `IndexRange` type, which comes with quite a few utilities that can simplify code. Some: - The offset integer array has to be one longer than the face count to avoid a branch for every face, which means the data is no longer part of the mesh's `CustomData`. - We lose the ability to "reference" an original mesh's offset array until more reusable CoW from #104478 is committed. That will be added in a separate commit. - Since they aren't part of `CustomData`, poly offsets often have to be copied manually. - To simplify using `OffsetIndices` in many places, some functions and structs in headers were moved to only compile in C++. - All meshes created by Blender use the same order for faces and face corners, but just in case, meshes with mismatched order are fixed by versioning code. - `MeshPolygon.totloop` is no longer editable in RNA. This API break is necessary here unfortunately. It should be worth it in 3.6, since that's the best way to allow loading meshes from 4.0, which is important for an LTS version. Pull Request: https://projects.blender.org/blender/blender/pulls/105938
2023-04-04 20:39:28 +02:00
const OffsetIndices polys = mesh.polys();
Mesh: Replace MLoop struct with generic attributes Implements #102359. Split the `MLoop` struct into two separate integer arrays called `corner_verts` and `corner_edges`, referring to the vertex each corner is attached to and the next edge around the face at each corner. These arrays can be sliced to give access to the edges or vertices in a face. Then they are often referred to as "poly_verts" or "poly_edges". The main benefits are halving the necessary memory bandwidth when only one array is used and simplifications from using regular integer indices instead of a special-purpose struct. The commit also starts a renaming from "loop" to "corner" in mesh code. Like the other mesh struct of array refactors, forward compatibility is kept by writing files with the older format. This will be done until 4.0 to ease the transition process. Looking at a small portion of the patch should give a good impression for the rest of the changes. I tried to make the changes as small as possible so it's easy to tell the correctness from the diff. Though I found Blender developers have been very inventive over the last decade when finding different ways to loop over the corners in a face. For performance, nearly every piece of code that deals with `Mesh` is slightly impacted. Any algorithm that is memory bottle-necked should see an improvement. For example, here is a comparison of interpolating a vertex float attribute to face corners (Ryzen 3700x): **Before** (Average: 3.7 ms, Min: 3.4 ms) ``` threading::parallel_for(loops.index_range(), 4096, [&](IndexRange range) { for (const int64_t i : range) { dst[i] = src[loops[i].v]; } }); ``` **After** (Average: 2.9 ms, Min: 2.6 ms) ``` array_utils::gather(src, corner_verts, dst); ``` That's an improvement of 28% to the average timings, and it's also a simplification, since an index-based routine can be used instead. For more examples using the new arrays, see the design task. Pull Request: https://projects.blender.org/blender/blender/pulls/104424
2023-03-20 15:55:13 +01:00
const Span<int> corner_edges = mesh.corner_edges();
Mesh: Remove redundant custom data pointers For copy-on-write, we want to share attribute arrays between meshes where possible. Mutable pointers like `Mesh.mvert` make that difficult by making ownership vague. They also make code more complex by adding redundancy. The simplest solution is just removing them and retrieving layers from `CustomData` as needed. Similar changes have already been applied to curves and point clouds (e9f82d3dc7ee, 410a6efb747f). Removing use of the pointers generally makes code more obvious and more reusable. Mesh data is now accessed with a C++ API (`Mesh::edges()` or `Mesh::edges_for_write()`), and a C API (`BKE_mesh_edges(mesh)`). The CoW changes this commit makes possible are described in T95845 and T95842, and started in D14139 and D14140. The change also simplifies the ongoing mesh struct-of-array refactors from T95965. **RNA/Python Access Performance** Theoretically, accessing mesh elements with the RNA API may become slower, since the layer needs to be found on every random access. However, overhead is already high enough that this doesn't make a noticible differenc, and performance is actually improved in some cases. Random access can be up to 10% faster, but other situations might be a bit slower. Generally using `foreach_get/set` are the best way to improve performance. See the differential revision for more discussion about Python performance. Cycles has been updated to use raw pointers and the internal Blender mesh types, mostly because there is no sense in having this overhead when it's already compiled with Blender. In my tests this roughly halves the Cycles mesh creation time (0.19s to 0.10s for a 1 million face grid). Differential Revision: https://developer.blender.org/D15488
2022-09-05 11:56:34 -05:00
Geometry Nodes: refactor virtual array system Goals of this refactor: * Simplify creating virtual arrays. * Simplify passing virtual arrays around. * Simplify converting between typed and generic virtual arrays. * Reduce memory allocations. As a quick reminder, a virtual arrays is a data structure that behaves like an array (i.e. it can be accessed using an index). However, it may not actually be stored as array internally. The two most important implementations of virtual arrays are those that correspond to an actual plain array and those that have the same value for every index. However, many more implementations exist for various reasons (interfacing with legacy attributes, unified iterator over all points in multiple splines, ...). With this refactor the core types (`VArray`, `GVArray`, `VMutableArray` and `GVMutableArray`) can be used like "normal values". They typically live on the stack. Before, they were usually inside a `std::unique_ptr`. This makes passing them around much easier. Creation of new virtual arrays is also much simpler now due to some constructors. Memory allocations are reduced by making use of small object optimization inside the core types. Previously, `VArray` was a class with virtual methods that had to be overridden to change the behavior of a the virtual array. Now,`VArray` has a fixed size and has no virtual methods. Instead it contains a `VArrayImpl` that is similar to the old `VArray`. `VArrayImpl` should rarely ever be used directly, unless a new virtual array implementation is added. To support the small object optimization for many `VArrayImpl` classes, a new `blender::Any` type is added. It is similar to `std::any` with two additional features. It has an adjustable inline buffer size and alignment. The inline buffer size of `std::any` can't be relied on and is usually too small for our use case here. Furthermore, `blender::Any` can store additional user-defined type information without increasing the stack size. Differential Revision: https://developer.blender.org/D12986
2021-11-16 10:15:51 +01:00
GVArray new_varray;
attribute_math::convert_to_static_type(varray.type(), [&](auto dummy) {
using T = decltype(dummy);
if constexpr (!std::is_void_v<attribute_math::DefaultMixer<T>>) {
if constexpr (std::is_same_v<T, bool>) {
/* A face is selected if all of its edges are selected. */
new_varray = VArray<bool>::ForFunc(
Mesh: Replace MLoop struct with generic attributes Implements #102359. Split the `MLoop` struct into two separate integer arrays called `corner_verts` and `corner_edges`, referring to the vertex each corner is attached to and the next edge around the face at each corner. These arrays can be sliced to give access to the edges or vertices in a face. Then they are often referred to as "poly_verts" or "poly_edges". The main benefits are halving the necessary memory bandwidth when only one array is used and simplifications from using regular integer indices instead of a special-purpose struct. The commit also starts a renaming from "loop" to "corner" in mesh code. Like the other mesh struct of array refactors, forward compatibility is kept by writing files with the older format. This will be done until 4.0 to ease the transition process. Looking at a small portion of the patch should give a good impression for the rest of the changes. I tried to make the changes as small as possible so it's easy to tell the correctness from the diff. Though I found Blender developers have been very inventive over the last decade when finding different ways to loop over the corners in a face. For performance, nearly every piece of code that deals with `Mesh` is slightly impacted. Any algorithm that is memory bottle-necked should see an improvement. For example, here is a comparison of interpolating a vertex float attribute to face corners (Ryzen 3700x): **Before** (Average: 3.7 ms, Min: 3.4 ms) ``` threading::parallel_for(loops.index_range(), 4096, [&](IndexRange range) { for (const int64_t i : range) { dst[i] = src[loops[i].v]; } }); ``` **After** (Average: 2.9 ms, Min: 2.6 ms) ``` array_utils::gather(src, corner_verts, dst); ``` That's an improvement of 28% to the average timings, and it's also a simplification, since an index-based routine can be used instead. For more examples using the new arrays, see the design task. Pull Request: https://projects.blender.org/blender/blender/pulls/104424
2023-03-20 15:55:13 +01:00
polys.size(), [corner_edges, polys, varray = varray.typed<T>()](const int face_index) {
Mesh: Replace MPoly struct with offset indices Implements #95967. Currently the `MPoly` struct is 12 bytes, and stores the index of a face's first corner and the number of corners/verts/edges. Polygons and corners are always created in order by Blender, meaning each face's corners will be after the previous face's corners. We can take advantage of this fact and eliminate the redundancy in mesh face storage by only storing a single integer corner offset for each face. The size of the face is then encoded by the offset of the next face. The size of a single integer is 4 bytes, so this reduces memory usage by 3 times. The same method is used for `CurvesGeometry`, so Blender already has an abstraction to simplify using these offsets called `OffsetIndices`. This class is used to easily retrieve a range of corner indices for each face. This also gives the opportunity for sharing some logic with curves. Another benefit of the change is that the offsets and sizes stored in `MPoly` can no longer disagree with each other. Storing faces in the order of their corners can simplify some code too. Face/polygon variables now use the `IndexRange` type, which comes with quite a few utilities that can simplify code. Some: - The offset integer array has to be one longer than the face count to avoid a branch for every face, which means the data is no longer part of the mesh's `CustomData`. - We lose the ability to "reference" an original mesh's offset array until more reusable CoW from #104478 is committed. That will be added in a separate commit. - Since they aren't part of `CustomData`, poly offsets often have to be copied manually. - To simplify using `OffsetIndices` in many places, some functions and structs in headers were moved to only compile in C++. - All meshes created by Blender use the same order for faces and face corners, but just in case, meshes with mismatched order are fixed by versioning code. - `MeshPolygon.totloop` is no longer editable in RNA. This API break is necessary here unfortunately. It should be worth it in 3.6, since that's the best way to allow loading meshes from 4.0, which is important for an LTS version. Pull Request: https://projects.blender.org/blender/blender/pulls/105938
2023-04-04 20:39:28 +02:00
for (const int edge : corner_edges.slice(polys[face_index])) {
Mesh: Replace MLoop struct with generic attributes Implements #102359. Split the `MLoop` struct into two separate integer arrays called `corner_verts` and `corner_edges`, referring to the vertex each corner is attached to and the next edge around the face at each corner. These arrays can be sliced to give access to the edges or vertices in a face. Then they are often referred to as "poly_verts" or "poly_edges". The main benefits are halving the necessary memory bandwidth when only one array is used and simplifications from using regular integer indices instead of a special-purpose struct. The commit also starts a renaming from "loop" to "corner" in mesh code. Like the other mesh struct of array refactors, forward compatibility is kept by writing files with the older format. This will be done until 4.0 to ease the transition process. Looking at a small portion of the patch should give a good impression for the rest of the changes. I tried to make the changes as small as possible so it's easy to tell the correctness from the diff. Though I found Blender developers have been very inventive over the last decade when finding different ways to loop over the corners in a face. For performance, nearly every piece of code that deals with `Mesh` is slightly impacted. Any algorithm that is memory bottle-necked should see an improvement. For example, here is a comparison of interpolating a vertex float attribute to face corners (Ryzen 3700x): **Before** (Average: 3.7 ms, Min: 3.4 ms) ``` threading::parallel_for(loops.index_range(), 4096, [&](IndexRange range) { for (const int64_t i : range) { dst[i] = src[loops[i].v]; } }); ``` **After** (Average: 2.9 ms, Min: 2.6 ms) ``` array_utils::gather(src, corner_verts, dst); ``` That's an improvement of 28% to the average timings, and it's also a simplification, since an index-based routine can be used instead. For more examples using the new arrays, see the design task. Pull Request: https://projects.blender.org/blender/blender/pulls/104424
2023-03-20 15:55:13 +01:00
if (!varray[edge]) {
return false;
}
}
return true;
});
}
else {
new_varray = VArray<T>::ForFunc(
Mesh: Replace MLoop struct with generic attributes Implements #102359. Split the `MLoop` struct into two separate integer arrays called `corner_verts` and `corner_edges`, referring to the vertex each corner is attached to and the next edge around the face at each corner. These arrays can be sliced to give access to the edges or vertices in a face. Then they are often referred to as "poly_verts" or "poly_edges". The main benefits are halving the necessary memory bandwidth when only one array is used and simplifications from using regular integer indices instead of a special-purpose struct. The commit also starts a renaming from "loop" to "corner" in mesh code. Like the other mesh struct of array refactors, forward compatibility is kept by writing files with the older format. This will be done until 4.0 to ease the transition process. Looking at a small portion of the patch should give a good impression for the rest of the changes. I tried to make the changes as small as possible so it's easy to tell the correctness from the diff. Though I found Blender developers have been very inventive over the last decade when finding different ways to loop over the corners in a face. For performance, nearly every piece of code that deals with `Mesh` is slightly impacted. Any algorithm that is memory bottle-necked should see an improvement. For example, here is a comparison of interpolating a vertex float attribute to face corners (Ryzen 3700x): **Before** (Average: 3.7 ms, Min: 3.4 ms) ``` threading::parallel_for(loops.index_range(), 4096, [&](IndexRange range) { for (const int64_t i : range) { dst[i] = src[loops[i].v]; } }); ``` **After** (Average: 2.9 ms, Min: 2.6 ms) ``` array_utils::gather(src, corner_verts, dst); ``` That's an improvement of 28% to the average timings, and it's also a simplification, since an index-based routine can be used instead. For more examples using the new arrays, see the design task. Pull Request: https://projects.blender.org/blender/blender/pulls/104424
2023-03-20 15:55:13 +01:00
polys.size(), [corner_edges, polys, varray = varray.typed<T>()](const int face_index) {
T return_value;
attribute_math::DefaultMixer<T> mixer({&return_value, 1});
Mesh: Replace MPoly struct with offset indices Implements #95967. Currently the `MPoly` struct is 12 bytes, and stores the index of a face's first corner and the number of corners/verts/edges. Polygons and corners are always created in order by Blender, meaning each face's corners will be after the previous face's corners. We can take advantage of this fact and eliminate the redundancy in mesh face storage by only storing a single integer corner offset for each face. The size of the face is then encoded by the offset of the next face. The size of a single integer is 4 bytes, so this reduces memory usage by 3 times. The same method is used for `CurvesGeometry`, so Blender already has an abstraction to simplify using these offsets called `OffsetIndices`. This class is used to easily retrieve a range of corner indices for each face. This also gives the opportunity for sharing some logic with curves. Another benefit of the change is that the offsets and sizes stored in `MPoly` can no longer disagree with each other. Storing faces in the order of their corners can simplify some code too. Face/polygon variables now use the `IndexRange` type, which comes with quite a few utilities that can simplify code. Some: - The offset integer array has to be one longer than the face count to avoid a branch for every face, which means the data is no longer part of the mesh's `CustomData`. - We lose the ability to "reference" an original mesh's offset array until more reusable CoW from #104478 is committed. That will be added in a separate commit. - Since they aren't part of `CustomData`, poly offsets often have to be copied manually. - To simplify using `OffsetIndices` in many places, some functions and structs in headers were moved to only compile in C++. - All meshes created by Blender use the same order for faces and face corners, but just in case, meshes with mismatched order are fixed by versioning code. - `MeshPolygon.totloop` is no longer editable in RNA. This API break is necessary here unfortunately. It should be worth it in 3.6, since that's the best way to allow loading meshes from 4.0, which is important for an LTS version. Pull Request: https://projects.blender.org/blender/blender/pulls/105938
2023-04-04 20:39:28 +02:00
for (const int edge : corner_edges.slice(polys[face_index])) {
Mesh: Replace MLoop struct with generic attributes Implements #102359. Split the `MLoop` struct into two separate integer arrays called `corner_verts` and `corner_edges`, referring to the vertex each corner is attached to and the next edge around the face at each corner. These arrays can be sliced to give access to the edges or vertices in a face. Then they are often referred to as "poly_verts" or "poly_edges". The main benefits are halving the necessary memory bandwidth when only one array is used and simplifications from using regular integer indices instead of a special-purpose struct. The commit also starts a renaming from "loop" to "corner" in mesh code. Like the other mesh struct of array refactors, forward compatibility is kept by writing files with the older format. This will be done until 4.0 to ease the transition process. Looking at a small portion of the patch should give a good impression for the rest of the changes. I tried to make the changes as small as possible so it's easy to tell the correctness from the diff. Though I found Blender developers have been very inventive over the last decade when finding different ways to loop over the corners in a face. For performance, nearly every piece of code that deals with `Mesh` is slightly impacted. Any algorithm that is memory bottle-necked should see an improvement. For example, here is a comparison of interpolating a vertex float attribute to face corners (Ryzen 3700x): **Before** (Average: 3.7 ms, Min: 3.4 ms) ``` threading::parallel_for(loops.index_range(), 4096, [&](IndexRange range) { for (const int64_t i : range) { dst[i] = src[loops[i].v]; } }); ``` **After** (Average: 2.9 ms, Min: 2.6 ms) ``` array_utils::gather(src, corner_verts, dst); ``` That's an improvement of 28% to the average timings, and it's also a simplification, since an index-based routine can be used instead. For more examples using the new arrays, see the design task. Pull Request: https://projects.blender.org/blender/blender/pulls/104424
2023-03-20 15:55:13 +01:00
mixer.mix_in(0, varray[edge]);
}
mixer.finalize();
return return_value;
});
}
}
});
Geometry Nodes: use virtual arrays in internal attribute api A virtual array is a data structure that is similar to a normal array in that its elements can be accessed by an index. However, a virtual array does not have to be a contiguous array internally. Instead, its elements can be layed out arbitrarily while element access happens through a virtual function call. However, the virtual array data structures are designed so that the virtual function call can be avoided in cases where it could become a bottleneck. Most commonly, a virtual array is backed by an actual array/span or is a single value internally, that is the same for every index. Besides those, there are many more specialized virtual arrays like the ones that provides vertex positions based on the `MVert` struct or vertex group weights. Not all attributes used by geometry nodes are stored in simple contiguous arrays. To provide uniform access to all kinds of attributes, the attribute API has to provide virtual array functionality that hides the implementation details of attributes. Before this refactor, the attribute API provided its own virtual array implementation as part of the `ReadAttribute` and `WriteAttribute` types. That resulted in unnecessary code duplication with the virtual array system. Even worse, it bound many algorithms used by geometry nodes to the specifics of the attribute API, even though they could also use different data sources (such as data from sockets, default values, later results of expressions, ...). This refactor removes the `ReadAttribute` and `WriteAttribute` types and replaces them with `GVArray` and `GVMutableArray` respectively. The `GV` stands for "generic virtual". The "generic" means that the data type contained in those virtual arrays is only known at run-time. There are the corresponding statically typed types `VArray<T>` and `VMutableArray<T>` as well. No regressions are expected from this refactor. It does come with one improvement for users. The attribute API can convert the data type on write now. This is especially useful when writing to builtin attributes like `material_index` with e.g. the Attribute Math node (which usually just writes to float attributes, while `material_index` is an integer attribute). Differential Revision: https://developer.blender.org/D10994
2021-04-17 16:41:03 +02:00
return new_varray;
}
} // namespace blender::bke
static bool can_simple_adapt_for_single(const Mesh &mesh,
const eAttrDomain from_domain,
const eAttrDomain to_domain)
{
/* For some domain combinations, a single value will always map directly. For others, there may
* be loose elements on the result domain that should have the default value rather than the
* single value from the source. */
switch (from_domain) {
case ATTR_DOMAIN_POINT:
/* All other domains are always connected to points. */
return true;
case ATTR_DOMAIN_EDGE:
if (to_domain == ATTR_DOMAIN_POINT) {
return mesh.loose_verts().count == 0;
}
return true;
case ATTR_DOMAIN_FACE:
if (to_domain == ATTR_DOMAIN_POINT) {
return mesh.verts_no_face().count == 0;
}
if (to_domain == ATTR_DOMAIN_EDGE) {
return mesh.loose_edges().count == 0;
}
return true;
case ATTR_DOMAIN_CORNER:
if (to_domain == ATTR_DOMAIN_POINT) {
return mesh.verts_no_face().count == 0;
}
if (to_domain == ATTR_DOMAIN_EDGE) {
return mesh.loose_edges().count == 0;
}
return true;
default:
BLI_assert_unreachable();
return false;
}
}
Geometry Nodes: new geometry attribute API Currently, there are two attribute API. The first, defined in `BKE_attribute.h` is accessible from RNA and C code. The second is implemented with `GeometryComponent` and is only accessible in C++ code. The second is widely used, but only being accessible through the `GeometrySet` API makes it awkward to use, and even impossible for types that don't correspond directly to a geometry component like `CurvesGeometry`. This patch adds a new attribute API, designed to replace the `GeometryComponent` attribute API now, and to eventually replace or be the basis of the other one. The basic idea is that there is an `AttributeAccessor` class that allows code to interact with a set of attributes owned by some geometry. The accessor itself has no ownership. `AttributeAccessor` is a simple type that can be passed around by value. That makes it easy to return it from functions and to store it in containers. For const-correctness, there is also a `MutableAttributeAccessor` that allows changing individual and can add or remove attributes. Currently, `AttributeAccessor` is composed of two pointers. The first is a pointer to the owner of the attribute data. The second is a pointer to a struct with function pointers, that is similar to a virtual function table. The functions know how to access attributes on the owner. The actual attribute access for geometries is still implemented with the `AttributeProvider` pattern, which makes it easy to support different sources of attributes on a geometry and simplifies dealing with built-in attributes. There are different ways to get an attribute accessor for a geometry: * `GeometryComponent.attributes()` * `CurvesGeometry.attributes()` * `bke::mesh_attributes(const Mesh &)` * `bke::pointcloud_attributes(const PointCloud &)` All of these also have a `_for_write` variant that returns a `MutabelAttributeAccessor`. Differential Revision: https://developer.blender.org/D15280
2022-07-08 16:16:56 +02:00
static blender::GVArray adapt_mesh_attribute_domain(const Mesh &mesh,
const blender::GVArray &varray,
const eAttrDomain from_domain,
const eAttrDomain to_domain)
{
Geometry Nodes: use virtual arrays in internal attribute api A virtual array is a data structure that is similar to a normal array in that its elements can be accessed by an index. However, a virtual array does not have to be a contiguous array internally. Instead, its elements can be layed out arbitrarily while element access happens through a virtual function call. However, the virtual array data structures are designed so that the virtual function call can be avoided in cases where it could become a bottleneck. Most commonly, a virtual array is backed by an actual array/span or is a single value internally, that is the same for every index. Besides those, there are many more specialized virtual arrays like the ones that provides vertex positions based on the `MVert` struct or vertex group weights. Not all attributes used by geometry nodes are stored in simple contiguous arrays. To provide uniform access to all kinds of attributes, the attribute API has to provide virtual array functionality that hides the implementation details of attributes. Before this refactor, the attribute API provided its own virtual array implementation as part of the `ReadAttribute` and `WriteAttribute` types. That resulted in unnecessary code duplication with the virtual array system. Even worse, it bound many algorithms used by geometry nodes to the specifics of the attribute API, even though they could also use different data sources (such as data from sockets, default values, later results of expressions, ...). This refactor removes the `ReadAttribute` and `WriteAttribute` types and replaces them with `GVArray` and `GVMutableArray` respectively. The `GV` stands for "generic virtual". The "generic" means that the data type contained in those virtual arrays is only known at run-time. There are the corresponding statically typed types `VArray<T>` and `VMutableArray<T>` as well. No regressions are expected from this refactor. It does come with one improvement for users. The attribute API can convert the data type on write now. This is especially useful when writing to builtin attributes like `material_index` with e.g. the Attribute Math node (which usually just writes to float attributes, while `material_index` is an integer attribute). Differential Revision: https://developer.blender.org/D10994
2021-04-17 16:41:03 +02:00
if (!varray) {
return {};
}
Geometry Nodes: refactor virtual array system Goals of this refactor: * Simplify creating virtual arrays. * Simplify passing virtual arrays around. * Simplify converting between typed and generic virtual arrays. * Reduce memory allocations. As a quick reminder, a virtual arrays is a data structure that behaves like an array (i.e. it can be accessed using an index). However, it may not actually be stored as array internally. The two most important implementations of virtual arrays are those that correspond to an actual plain array and those that have the same value for every index. However, many more implementations exist for various reasons (interfacing with legacy attributes, unified iterator over all points in multiple splines, ...). With this refactor the core types (`VArray`, `GVArray`, `VMutableArray` and `GVMutableArray`) can be used like "normal values". They typically live on the stack. Before, they were usually inside a `std::unique_ptr`. This makes passing them around much easier. Creation of new virtual arrays is also much simpler now due to some constructors. Memory allocations are reduced by making use of small object optimization inside the core types. Previously, `VArray` was a class with virtual methods that had to be overridden to change the behavior of a the virtual array. Now,`VArray` has a fixed size and has no virtual methods. Instead it contains a `VArrayImpl` that is similar to the old `VArray`. `VArrayImpl` should rarely ever be used directly, unless a new virtual array implementation is added. To support the small object optimization for many `VArrayImpl` classes, a new `blender::Any` type is added. It is similar to `std::any` with two additional features. It has an adjustable inline buffer size and alignment. The inline buffer size of `std::any` can't be relied on and is usually too small for our use case here. Furthermore, `blender::Any` can store additional user-defined type information without increasing the stack size. Differential Revision: https://developer.blender.org/D12986
2021-11-16 10:15:51 +01:00
if (varray.size() == 0) {
return {};
}
Geometry Nodes: use virtual arrays in internal attribute api A virtual array is a data structure that is similar to a normal array in that its elements can be accessed by an index. However, a virtual array does not have to be a contiguous array internally. Instead, its elements can be layed out arbitrarily while element access happens through a virtual function call. However, the virtual array data structures are designed so that the virtual function call can be avoided in cases where it could become a bottleneck. Most commonly, a virtual array is backed by an actual array/span or is a single value internally, that is the same for every index. Besides those, there are many more specialized virtual arrays like the ones that provides vertex positions based on the `MVert` struct or vertex group weights. Not all attributes used by geometry nodes are stored in simple contiguous arrays. To provide uniform access to all kinds of attributes, the attribute API has to provide virtual array functionality that hides the implementation details of attributes. Before this refactor, the attribute API provided its own virtual array implementation as part of the `ReadAttribute` and `WriteAttribute` types. That resulted in unnecessary code duplication with the virtual array system. Even worse, it bound many algorithms used by geometry nodes to the specifics of the attribute API, even though they could also use different data sources (such as data from sockets, default values, later results of expressions, ...). This refactor removes the `ReadAttribute` and `WriteAttribute` types and replaces them with `GVArray` and `GVMutableArray` respectively. The `GV` stands for "generic virtual". The "generic" means that the data type contained in those virtual arrays is only known at run-time. There are the corresponding statically typed types `VArray<T>` and `VMutableArray<T>` as well. No regressions are expected from this refactor. It does come with one improvement for users. The attribute API can convert the data type on write now. This is especially useful when writing to builtin attributes like `material_index` with e.g. the Attribute Math node (which usually just writes to float attributes, while `material_index` is an integer attribute). Differential Revision: https://developer.blender.org/D10994
2021-04-17 16:41:03 +02:00
if (from_domain == to_domain) {
return varray;
}
if (varray.is_single()) {
if (can_simple_adapt_for_single(mesh, from_domain, to_domain)) {
BUFFER_FOR_CPP_TYPE_VALUE(varray.type(), value);
varray.get_internal_single(value);
return blender::GVArray::ForSingle(
varray.type(), mesh.attributes().domain_size(to_domain), value);
}
}
Geometry Nodes: use virtual arrays in internal attribute api A virtual array is a data structure that is similar to a normal array in that its elements can be accessed by an index. However, a virtual array does not have to be a contiguous array internally. Instead, its elements can be layed out arbitrarily while element access happens through a virtual function call. However, the virtual array data structures are designed so that the virtual function call can be avoided in cases where it could become a bottleneck. Most commonly, a virtual array is backed by an actual array/span or is a single value internally, that is the same for every index. Besides those, there are many more specialized virtual arrays like the ones that provides vertex positions based on the `MVert` struct or vertex group weights. Not all attributes used by geometry nodes are stored in simple contiguous arrays. To provide uniform access to all kinds of attributes, the attribute API has to provide virtual array functionality that hides the implementation details of attributes. Before this refactor, the attribute API provided its own virtual array implementation as part of the `ReadAttribute` and `WriteAttribute` types. That resulted in unnecessary code duplication with the virtual array system. Even worse, it bound many algorithms used by geometry nodes to the specifics of the attribute API, even though they could also use different data sources (such as data from sockets, default values, later results of expressions, ...). This refactor removes the `ReadAttribute` and `WriteAttribute` types and replaces them with `GVArray` and `GVMutableArray` respectively. The `GV` stands for "generic virtual". The "generic" means that the data type contained in those virtual arrays is only known at run-time. There are the corresponding statically typed types `VArray<T>` and `VMutableArray<T>` as well. No regressions are expected from this refactor. It does come with one improvement for users. The attribute API can convert the data type on write now. This is especially useful when writing to builtin attributes like `material_index` with e.g. the Attribute Math node (which usually just writes to float attributes, while `material_index` is an integer attribute). Differential Revision: https://developer.blender.org/D10994
2021-04-17 16:41:03 +02:00
switch (from_domain) {
case ATTR_DOMAIN_CORNER: {
Geometry Nodes: use virtual arrays in internal attribute api A virtual array is a data structure that is similar to a normal array in that its elements can be accessed by an index. However, a virtual array does not have to be a contiguous array internally. Instead, its elements can be layed out arbitrarily while element access happens through a virtual function call. However, the virtual array data structures are designed so that the virtual function call can be avoided in cases where it could become a bottleneck. Most commonly, a virtual array is backed by an actual array/span or is a single value internally, that is the same for every index. Besides those, there are many more specialized virtual arrays like the ones that provides vertex positions based on the `MVert` struct or vertex group weights. Not all attributes used by geometry nodes are stored in simple contiguous arrays. To provide uniform access to all kinds of attributes, the attribute API has to provide virtual array functionality that hides the implementation details of attributes. Before this refactor, the attribute API provided its own virtual array implementation as part of the `ReadAttribute` and `WriteAttribute` types. That resulted in unnecessary code duplication with the virtual array system. Even worse, it bound many algorithms used by geometry nodes to the specifics of the attribute API, even though they could also use different data sources (such as data from sockets, default values, later results of expressions, ...). This refactor removes the `ReadAttribute` and `WriteAttribute` types and replaces them with `GVArray` and `GVMutableArray` respectively. The `GV` stands for "generic virtual". The "generic" means that the data type contained in those virtual arrays is only known at run-time. There are the corresponding statically typed types `VArray<T>` and `VMutableArray<T>` as well. No regressions are expected from this refactor. It does come with one improvement for users. The attribute API can convert the data type on write now. This is especially useful when writing to builtin attributes like `material_index` with e.g. the Attribute Math node (which usually just writes to float attributes, while `material_index` is an integer attribute). Differential Revision: https://developer.blender.org/D10994
2021-04-17 16:41:03 +02:00
switch (to_domain) {
case ATTR_DOMAIN_POINT:
Geometry Nodes: new geometry attribute API Currently, there are two attribute API. The first, defined in `BKE_attribute.h` is accessible from RNA and C code. The second is implemented with `GeometryComponent` and is only accessible in C++ code. The second is widely used, but only being accessible through the `GeometrySet` API makes it awkward to use, and even impossible for types that don't correspond directly to a geometry component like `CurvesGeometry`. This patch adds a new attribute API, designed to replace the `GeometryComponent` attribute API now, and to eventually replace or be the basis of the other one. The basic idea is that there is an `AttributeAccessor` class that allows code to interact with a set of attributes owned by some geometry. The accessor itself has no ownership. `AttributeAccessor` is a simple type that can be passed around by value. That makes it easy to return it from functions and to store it in containers. For const-correctness, there is also a `MutableAttributeAccessor` that allows changing individual and can add or remove attributes. Currently, `AttributeAccessor` is composed of two pointers. The first is a pointer to the owner of the attribute data. The second is a pointer to a struct with function pointers, that is similar to a virtual function table. The functions know how to access attributes on the owner. The actual attribute access for geometries is still implemented with the `AttributeProvider` pattern, which makes it easy to support different sources of attributes on a geometry and simplifies dealing with built-in attributes. There are different ways to get an attribute accessor for a geometry: * `GeometryComponent.attributes()` * `CurvesGeometry.attributes()` * `bke::mesh_attributes(const Mesh &)` * `bke::pointcloud_attributes(const PointCloud &)` All of these also have a `_for_write` variant that returns a `MutabelAttributeAccessor`. Differential Revision: https://developer.blender.org/D15280
2022-07-08 16:16:56 +02:00
return blender::bke::adapt_mesh_domain_corner_to_point(mesh, varray);
case ATTR_DOMAIN_FACE:
Geometry Nodes: new geometry attribute API Currently, there are two attribute API. The first, defined in `BKE_attribute.h` is accessible from RNA and C code. The second is implemented with `GeometryComponent` and is only accessible in C++ code. The second is widely used, but only being accessible through the `GeometrySet` API makes it awkward to use, and even impossible for types that don't correspond directly to a geometry component like `CurvesGeometry`. This patch adds a new attribute API, designed to replace the `GeometryComponent` attribute API now, and to eventually replace or be the basis of the other one. The basic idea is that there is an `AttributeAccessor` class that allows code to interact with a set of attributes owned by some geometry. The accessor itself has no ownership. `AttributeAccessor` is a simple type that can be passed around by value. That makes it easy to return it from functions and to store it in containers. For const-correctness, there is also a `MutableAttributeAccessor` that allows changing individual and can add or remove attributes. Currently, `AttributeAccessor` is composed of two pointers. The first is a pointer to the owner of the attribute data. The second is a pointer to a struct with function pointers, that is similar to a virtual function table. The functions know how to access attributes on the owner. The actual attribute access for geometries is still implemented with the `AttributeProvider` pattern, which makes it easy to support different sources of attributes on a geometry and simplifies dealing with built-in attributes. There are different ways to get an attribute accessor for a geometry: * `GeometryComponent.attributes()` * `CurvesGeometry.attributes()` * `bke::mesh_attributes(const Mesh &)` * `bke::pointcloud_attributes(const PointCloud &)` All of these also have a `_for_write` variant that returns a `MutabelAttributeAccessor`. Differential Revision: https://developer.blender.org/D15280
2022-07-08 16:16:56 +02:00
return blender::bke::adapt_mesh_domain_corner_to_face(mesh, varray);
case ATTR_DOMAIN_EDGE:
Geometry Nodes: new geometry attribute API Currently, there are two attribute API. The first, defined in `BKE_attribute.h` is accessible from RNA and C code. The second is implemented with `GeometryComponent` and is only accessible in C++ code. The second is widely used, but only being accessible through the `GeometrySet` API makes it awkward to use, and even impossible for types that don't correspond directly to a geometry component like `CurvesGeometry`. This patch adds a new attribute API, designed to replace the `GeometryComponent` attribute API now, and to eventually replace or be the basis of the other one. The basic idea is that there is an `AttributeAccessor` class that allows code to interact with a set of attributes owned by some geometry. The accessor itself has no ownership. `AttributeAccessor` is a simple type that can be passed around by value. That makes it easy to return it from functions and to store it in containers. For const-correctness, there is also a `MutableAttributeAccessor` that allows changing individual and can add or remove attributes. Currently, `AttributeAccessor` is composed of two pointers. The first is a pointer to the owner of the attribute data. The second is a pointer to a struct with function pointers, that is similar to a virtual function table. The functions know how to access attributes on the owner. The actual attribute access for geometries is still implemented with the `AttributeProvider` pattern, which makes it easy to support different sources of attributes on a geometry and simplifies dealing with built-in attributes. There are different ways to get an attribute accessor for a geometry: * `GeometryComponent.attributes()` * `CurvesGeometry.attributes()` * `bke::mesh_attributes(const Mesh &)` * `bke::pointcloud_attributes(const PointCloud &)` All of these also have a `_for_write` variant that returns a `MutabelAttributeAccessor`. Differential Revision: https://developer.blender.org/D15280
2022-07-08 16:16:56 +02:00
return blender::bke::adapt_mesh_domain_corner_to_edge(mesh, varray);
default:
break;
}
break;
}
case ATTR_DOMAIN_POINT: {
Geometry Nodes: use virtual arrays in internal attribute api A virtual array is a data structure that is similar to a normal array in that its elements can be accessed by an index. However, a virtual array does not have to be a contiguous array internally. Instead, its elements can be layed out arbitrarily while element access happens through a virtual function call. However, the virtual array data structures are designed so that the virtual function call can be avoided in cases where it could become a bottleneck. Most commonly, a virtual array is backed by an actual array/span or is a single value internally, that is the same for every index. Besides those, there are many more specialized virtual arrays like the ones that provides vertex positions based on the `MVert` struct or vertex group weights. Not all attributes used by geometry nodes are stored in simple contiguous arrays. To provide uniform access to all kinds of attributes, the attribute API has to provide virtual array functionality that hides the implementation details of attributes. Before this refactor, the attribute API provided its own virtual array implementation as part of the `ReadAttribute` and `WriteAttribute` types. That resulted in unnecessary code duplication with the virtual array system. Even worse, it bound many algorithms used by geometry nodes to the specifics of the attribute API, even though they could also use different data sources (such as data from sockets, default values, later results of expressions, ...). This refactor removes the `ReadAttribute` and `WriteAttribute` types and replaces them with `GVArray` and `GVMutableArray` respectively. The `GV` stands for "generic virtual". The "generic" means that the data type contained in those virtual arrays is only known at run-time. There are the corresponding statically typed types `VArray<T>` and `VMutableArray<T>` as well. No regressions are expected from this refactor. It does come with one improvement for users. The attribute API can convert the data type on write now. This is especially useful when writing to builtin attributes like `material_index` with e.g. the Attribute Math node (which usually just writes to float attributes, while `material_index` is an integer attribute). Differential Revision: https://developer.blender.org/D10994
2021-04-17 16:41:03 +02:00
switch (to_domain) {
case ATTR_DOMAIN_CORNER:
Geometry Nodes: new geometry attribute API Currently, there are two attribute API. The first, defined in `BKE_attribute.h` is accessible from RNA and C code. The second is implemented with `GeometryComponent` and is only accessible in C++ code. The second is widely used, but only being accessible through the `GeometrySet` API makes it awkward to use, and even impossible for types that don't correspond directly to a geometry component like `CurvesGeometry`. This patch adds a new attribute API, designed to replace the `GeometryComponent` attribute API now, and to eventually replace or be the basis of the other one. The basic idea is that there is an `AttributeAccessor` class that allows code to interact with a set of attributes owned by some geometry. The accessor itself has no ownership. `AttributeAccessor` is a simple type that can be passed around by value. That makes it easy to return it from functions and to store it in containers. For const-correctness, there is also a `MutableAttributeAccessor` that allows changing individual and can add or remove attributes. Currently, `AttributeAccessor` is composed of two pointers. The first is a pointer to the owner of the attribute data. The second is a pointer to a struct with function pointers, that is similar to a virtual function table. The functions know how to access attributes on the owner. The actual attribute access for geometries is still implemented with the `AttributeProvider` pattern, which makes it easy to support different sources of attributes on a geometry and simplifies dealing with built-in attributes. There are different ways to get an attribute accessor for a geometry: * `GeometryComponent.attributes()` * `CurvesGeometry.attributes()` * `bke::mesh_attributes(const Mesh &)` * `bke::pointcloud_attributes(const PointCloud &)` All of these also have a `_for_write` variant that returns a `MutabelAttributeAccessor`. Differential Revision: https://developer.blender.org/D15280
2022-07-08 16:16:56 +02:00
return blender::bke::adapt_mesh_domain_point_to_corner(mesh, varray);
case ATTR_DOMAIN_FACE:
Geometry Nodes: new geometry attribute API Currently, there are two attribute API. The first, defined in `BKE_attribute.h` is accessible from RNA and C code. The second is implemented with `GeometryComponent` and is only accessible in C++ code. The second is widely used, but only being accessible through the `GeometrySet` API makes it awkward to use, and even impossible for types that don't correspond directly to a geometry component like `CurvesGeometry`. This patch adds a new attribute API, designed to replace the `GeometryComponent` attribute API now, and to eventually replace or be the basis of the other one. The basic idea is that there is an `AttributeAccessor` class that allows code to interact with a set of attributes owned by some geometry. The accessor itself has no ownership. `AttributeAccessor` is a simple type that can be passed around by value. That makes it easy to return it from functions and to store it in containers. For const-correctness, there is also a `MutableAttributeAccessor` that allows changing individual and can add or remove attributes. Currently, `AttributeAccessor` is composed of two pointers. The first is a pointer to the owner of the attribute data. The second is a pointer to a struct with function pointers, that is similar to a virtual function table. The functions know how to access attributes on the owner. The actual attribute access for geometries is still implemented with the `AttributeProvider` pattern, which makes it easy to support different sources of attributes on a geometry and simplifies dealing with built-in attributes. There are different ways to get an attribute accessor for a geometry: * `GeometryComponent.attributes()` * `CurvesGeometry.attributes()` * `bke::mesh_attributes(const Mesh &)` * `bke::pointcloud_attributes(const PointCloud &)` All of these also have a `_for_write` variant that returns a `MutabelAttributeAccessor`. Differential Revision: https://developer.blender.org/D15280
2022-07-08 16:16:56 +02:00
return blender::bke::adapt_mesh_domain_point_to_face(mesh, varray);
case ATTR_DOMAIN_EDGE:
Geometry Nodes: new geometry attribute API Currently, there are two attribute API. The first, defined in `BKE_attribute.h` is accessible from RNA and C code. The second is implemented with `GeometryComponent` and is only accessible in C++ code. The second is widely used, but only being accessible through the `GeometrySet` API makes it awkward to use, and even impossible for types that don't correspond directly to a geometry component like `CurvesGeometry`. This patch adds a new attribute API, designed to replace the `GeometryComponent` attribute API now, and to eventually replace or be the basis of the other one. The basic idea is that there is an `AttributeAccessor` class that allows code to interact with a set of attributes owned by some geometry. The accessor itself has no ownership. `AttributeAccessor` is a simple type that can be passed around by value. That makes it easy to return it from functions and to store it in containers. For const-correctness, there is also a `MutableAttributeAccessor` that allows changing individual and can add or remove attributes. Currently, `AttributeAccessor` is composed of two pointers. The first is a pointer to the owner of the attribute data. The second is a pointer to a struct with function pointers, that is similar to a virtual function table. The functions know how to access attributes on the owner. The actual attribute access for geometries is still implemented with the `AttributeProvider` pattern, which makes it easy to support different sources of attributes on a geometry and simplifies dealing with built-in attributes. There are different ways to get an attribute accessor for a geometry: * `GeometryComponent.attributes()` * `CurvesGeometry.attributes()` * `bke::mesh_attributes(const Mesh &)` * `bke::pointcloud_attributes(const PointCloud &)` All of these also have a `_for_write` variant that returns a `MutabelAttributeAccessor`. Differential Revision: https://developer.blender.org/D15280
2022-07-08 16:16:56 +02:00
return blender::bke::adapt_mesh_domain_point_to_edge(mesh, varray);
default:
break;
}
break;
}
case ATTR_DOMAIN_FACE: {
Geometry Nodes: use virtual arrays in internal attribute api A virtual array is a data structure that is similar to a normal array in that its elements can be accessed by an index. However, a virtual array does not have to be a contiguous array internally. Instead, its elements can be layed out arbitrarily while element access happens through a virtual function call. However, the virtual array data structures are designed so that the virtual function call can be avoided in cases where it could become a bottleneck. Most commonly, a virtual array is backed by an actual array/span or is a single value internally, that is the same for every index. Besides those, there are many more specialized virtual arrays like the ones that provides vertex positions based on the `MVert` struct or vertex group weights. Not all attributes used by geometry nodes are stored in simple contiguous arrays. To provide uniform access to all kinds of attributes, the attribute API has to provide virtual array functionality that hides the implementation details of attributes. Before this refactor, the attribute API provided its own virtual array implementation as part of the `ReadAttribute` and `WriteAttribute` types. That resulted in unnecessary code duplication with the virtual array system. Even worse, it bound many algorithms used by geometry nodes to the specifics of the attribute API, even though they could also use different data sources (such as data from sockets, default values, later results of expressions, ...). This refactor removes the `ReadAttribute` and `WriteAttribute` types and replaces them with `GVArray` and `GVMutableArray` respectively. The `GV` stands for "generic virtual". The "generic" means that the data type contained in those virtual arrays is only known at run-time. There are the corresponding statically typed types `VArray<T>` and `VMutableArray<T>` as well. No regressions are expected from this refactor. It does come with one improvement for users. The attribute API can convert the data type on write now. This is especially useful when writing to builtin attributes like `material_index` with e.g. the Attribute Math node (which usually just writes to float attributes, while `material_index` is an integer attribute). Differential Revision: https://developer.blender.org/D10994
2021-04-17 16:41:03 +02:00
switch (to_domain) {
case ATTR_DOMAIN_POINT:
Geometry Nodes: new geometry attribute API Currently, there are two attribute API. The first, defined in `BKE_attribute.h` is accessible from RNA and C code. The second is implemented with `GeometryComponent` and is only accessible in C++ code. The second is widely used, but only being accessible through the `GeometrySet` API makes it awkward to use, and even impossible for types that don't correspond directly to a geometry component like `CurvesGeometry`. This patch adds a new attribute API, designed to replace the `GeometryComponent` attribute API now, and to eventually replace or be the basis of the other one. The basic idea is that there is an `AttributeAccessor` class that allows code to interact with a set of attributes owned by some geometry. The accessor itself has no ownership. `AttributeAccessor` is a simple type that can be passed around by value. That makes it easy to return it from functions and to store it in containers. For const-correctness, there is also a `MutableAttributeAccessor` that allows changing individual and can add or remove attributes. Currently, `AttributeAccessor` is composed of two pointers. The first is a pointer to the owner of the attribute data. The second is a pointer to a struct with function pointers, that is similar to a virtual function table. The functions know how to access attributes on the owner. The actual attribute access for geometries is still implemented with the `AttributeProvider` pattern, which makes it easy to support different sources of attributes on a geometry and simplifies dealing with built-in attributes. There are different ways to get an attribute accessor for a geometry: * `GeometryComponent.attributes()` * `CurvesGeometry.attributes()` * `bke::mesh_attributes(const Mesh &)` * `bke::pointcloud_attributes(const PointCloud &)` All of these also have a `_for_write` variant that returns a `MutabelAttributeAccessor`. Differential Revision: https://developer.blender.org/D15280
2022-07-08 16:16:56 +02:00
return blender::bke::adapt_mesh_domain_face_to_point(mesh, varray);
case ATTR_DOMAIN_CORNER:
Geometry Nodes: new geometry attribute API Currently, there are two attribute API. The first, defined in `BKE_attribute.h` is accessible from RNA and C code. The second is implemented with `GeometryComponent` and is only accessible in C++ code. The second is widely used, but only being accessible through the `GeometrySet` API makes it awkward to use, and even impossible for types that don't correspond directly to a geometry component like `CurvesGeometry`. This patch adds a new attribute API, designed to replace the `GeometryComponent` attribute API now, and to eventually replace or be the basis of the other one. The basic idea is that there is an `AttributeAccessor` class that allows code to interact with a set of attributes owned by some geometry. The accessor itself has no ownership. `AttributeAccessor` is a simple type that can be passed around by value. That makes it easy to return it from functions and to store it in containers. For const-correctness, there is also a `MutableAttributeAccessor` that allows changing individual and can add or remove attributes. Currently, `AttributeAccessor` is composed of two pointers. The first is a pointer to the owner of the attribute data. The second is a pointer to a struct with function pointers, that is similar to a virtual function table. The functions know how to access attributes on the owner. The actual attribute access for geometries is still implemented with the `AttributeProvider` pattern, which makes it easy to support different sources of attributes on a geometry and simplifies dealing with built-in attributes. There are different ways to get an attribute accessor for a geometry: * `GeometryComponent.attributes()` * `CurvesGeometry.attributes()` * `bke::mesh_attributes(const Mesh &)` * `bke::pointcloud_attributes(const PointCloud &)` All of these also have a `_for_write` variant that returns a `MutabelAttributeAccessor`. Differential Revision: https://developer.blender.org/D15280
2022-07-08 16:16:56 +02:00
return blender::bke::adapt_mesh_domain_face_to_corner(mesh, varray);
case ATTR_DOMAIN_EDGE:
Geometry Nodes: new geometry attribute API Currently, there are two attribute API. The first, defined in `BKE_attribute.h` is accessible from RNA and C code. The second is implemented with `GeometryComponent` and is only accessible in C++ code. The second is widely used, but only being accessible through the `GeometrySet` API makes it awkward to use, and even impossible for types that don't correspond directly to a geometry component like `CurvesGeometry`. This patch adds a new attribute API, designed to replace the `GeometryComponent` attribute API now, and to eventually replace or be the basis of the other one. The basic idea is that there is an `AttributeAccessor` class that allows code to interact with a set of attributes owned by some geometry. The accessor itself has no ownership. `AttributeAccessor` is a simple type that can be passed around by value. That makes it easy to return it from functions and to store it in containers. For const-correctness, there is also a `MutableAttributeAccessor` that allows changing individual and can add or remove attributes. Currently, `AttributeAccessor` is composed of two pointers. The first is a pointer to the owner of the attribute data. The second is a pointer to a struct with function pointers, that is similar to a virtual function table. The functions know how to access attributes on the owner. The actual attribute access for geometries is still implemented with the `AttributeProvider` pattern, which makes it easy to support different sources of attributes on a geometry and simplifies dealing with built-in attributes. There are different ways to get an attribute accessor for a geometry: * `GeometryComponent.attributes()` * `CurvesGeometry.attributes()` * `bke::mesh_attributes(const Mesh &)` * `bke::pointcloud_attributes(const PointCloud &)` All of these also have a `_for_write` variant that returns a `MutabelAttributeAccessor`. Differential Revision: https://developer.blender.org/D15280
2022-07-08 16:16:56 +02:00
return blender::bke::adapt_mesh_domain_face_to_edge(mesh, varray);
default:
break;
}
break;
}
case ATTR_DOMAIN_EDGE: {
Geometry Nodes: use virtual arrays in internal attribute api A virtual array is a data structure that is similar to a normal array in that its elements can be accessed by an index. However, a virtual array does not have to be a contiguous array internally. Instead, its elements can be layed out arbitrarily while element access happens through a virtual function call. However, the virtual array data structures are designed so that the virtual function call can be avoided in cases where it could become a bottleneck. Most commonly, a virtual array is backed by an actual array/span or is a single value internally, that is the same for every index. Besides those, there are many more specialized virtual arrays like the ones that provides vertex positions based on the `MVert` struct or vertex group weights. Not all attributes used by geometry nodes are stored in simple contiguous arrays. To provide uniform access to all kinds of attributes, the attribute API has to provide virtual array functionality that hides the implementation details of attributes. Before this refactor, the attribute API provided its own virtual array implementation as part of the `ReadAttribute` and `WriteAttribute` types. That resulted in unnecessary code duplication with the virtual array system. Even worse, it bound many algorithms used by geometry nodes to the specifics of the attribute API, even though they could also use different data sources (such as data from sockets, default values, later results of expressions, ...). This refactor removes the `ReadAttribute` and `WriteAttribute` types and replaces them with `GVArray` and `GVMutableArray` respectively. The `GV` stands for "generic virtual". The "generic" means that the data type contained in those virtual arrays is only known at run-time. There are the corresponding statically typed types `VArray<T>` and `VMutableArray<T>` as well. No regressions are expected from this refactor. It does come with one improvement for users. The attribute API can convert the data type on write now. This is especially useful when writing to builtin attributes like `material_index` with e.g. the Attribute Math node (which usually just writes to float attributes, while `material_index` is an integer attribute). Differential Revision: https://developer.blender.org/D10994
2021-04-17 16:41:03 +02:00
switch (to_domain) {
case ATTR_DOMAIN_CORNER:
Geometry Nodes: new geometry attribute API Currently, there are two attribute API. The first, defined in `BKE_attribute.h` is accessible from RNA and C code. The second is implemented with `GeometryComponent` and is only accessible in C++ code. The second is widely used, but only being accessible through the `GeometrySet` API makes it awkward to use, and even impossible for types that don't correspond directly to a geometry component like `CurvesGeometry`. This patch adds a new attribute API, designed to replace the `GeometryComponent` attribute API now, and to eventually replace or be the basis of the other one. The basic idea is that there is an `AttributeAccessor` class that allows code to interact with a set of attributes owned by some geometry. The accessor itself has no ownership. `AttributeAccessor` is a simple type that can be passed around by value. That makes it easy to return it from functions and to store it in containers. For const-correctness, there is also a `MutableAttributeAccessor` that allows changing individual and can add or remove attributes. Currently, `AttributeAccessor` is composed of two pointers. The first is a pointer to the owner of the attribute data. The second is a pointer to a struct with function pointers, that is similar to a virtual function table. The functions know how to access attributes on the owner. The actual attribute access for geometries is still implemented with the `AttributeProvider` pattern, which makes it easy to support different sources of attributes on a geometry and simplifies dealing with built-in attributes. There are different ways to get an attribute accessor for a geometry: * `GeometryComponent.attributes()` * `CurvesGeometry.attributes()` * `bke::mesh_attributes(const Mesh &)` * `bke::pointcloud_attributes(const PointCloud &)` All of these also have a `_for_write` variant that returns a `MutabelAttributeAccessor`. Differential Revision: https://developer.blender.org/D15280
2022-07-08 16:16:56 +02:00
return blender::bke::adapt_mesh_domain_edge_to_corner(mesh, varray);
case ATTR_DOMAIN_POINT:
Geometry Nodes: new geometry attribute API Currently, there are two attribute API. The first, defined in `BKE_attribute.h` is accessible from RNA and C code. The second is implemented with `GeometryComponent` and is only accessible in C++ code. The second is widely used, but only being accessible through the `GeometrySet` API makes it awkward to use, and even impossible for types that don't correspond directly to a geometry component like `CurvesGeometry`. This patch adds a new attribute API, designed to replace the `GeometryComponent` attribute API now, and to eventually replace or be the basis of the other one. The basic idea is that there is an `AttributeAccessor` class that allows code to interact with a set of attributes owned by some geometry. The accessor itself has no ownership. `AttributeAccessor` is a simple type that can be passed around by value. That makes it easy to return it from functions and to store it in containers. For const-correctness, there is also a `MutableAttributeAccessor` that allows changing individual and can add or remove attributes. Currently, `AttributeAccessor` is composed of two pointers. The first is a pointer to the owner of the attribute data. The second is a pointer to a struct with function pointers, that is similar to a virtual function table. The functions know how to access attributes on the owner. The actual attribute access for geometries is still implemented with the `AttributeProvider` pattern, which makes it easy to support different sources of attributes on a geometry and simplifies dealing with built-in attributes. There are different ways to get an attribute accessor for a geometry: * `GeometryComponent.attributes()` * `CurvesGeometry.attributes()` * `bke::mesh_attributes(const Mesh &)` * `bke::pointcloud_attributes(const PointCloud &)` All of these also have a `_for_write` variant that returns a `MutabelAttributeAccessor`. Differential Revision: https://developer.blender.org/D15280
2022-07-08 16:16:56 +02:00
return blender::bke::adapt_mesh_domain_edge_to_point(mesh, varray);
case ATTR_DOMAIN_FACE:
Geometry Nodes: new geometry attribute API Currently, there are two attribute API. The first, defined in `BKE_attribute.h` is accessible from RNA and C code. The second is implemented with `GeometryComponent` and is only accessible in C++ code. The second is widely used, but only being accessible through the `GeometrySet` API makes it awkward to use, and even impossible for types that don't correspond directly to a geometry component like `CurvesGeometry`. This patch adds a new attribute API, designed to replace the `GeometryComponent` attribute API now, and to eventually replace or be the basis of the other one. The basic idea is that there is an `AttributeAccessor` class that allows code to interact with a set of attributes owned by some geometry. The accessor itself has no ownership. `AttributeAccessor` is a simple type that can be passed around by value. That makes it easy to return it from functions and to store it in containers. For const-correctness, there is also a `MutableAttributeAccessor` that allows changing individual and can add or remove attributes. Currently, `AttributeAccessor` is composed of two pointers. The first is a pointer to the owner of the attribute data. The second is a pointer to a struct with function pointers, that is similar to a virtual function table. The functions know how to access attributes on the owner. The actual attribute access for geometries is still implemented with the `AttributeProvider` pattern, which makes it easy to support different sources of attributes on a geometry and simplifies dealing with built-in attributes. There are different ways to get an attribute accessor for a geometry: * `GeometryComponent.attributes()` * `CurvesGeometry.attributes()` * `bke::mesh_attributes(const Mesh &)` * `bke::pointcloud_attributes(const PointCloud &)` All of these also have a `_for_write` variant that returns a `MutabelAttributeAccessor`. Differential Revision: https://developer.blender.org/D15280
2022-07-08 16:16:56 +02:00
return blender::bke::adapt_mesh_domain_edge_to_face(mesh, varray);
default:
break;
}
break;
}
default:
break;
}
return {};
}
namespace blender::bke {
Geometry Nodes: new geometry attribute API Currently, there are two attribute API. The first, defined in `BKE_attribute.h` is accessible from RNA and C code. The second is implemented with `GeometryComponent` and is only accessible in C++ code. The second is widely used, but only being accessible through the `GeometrySet` API makes it awkward to use, and even impossible for types that don't correspond directly to a geometry component like `CurvesGeometry`. This patch adds a new attribute API, designed to replace the `GeometryComponent` attribute API now, and to eventually replace or be the basis of the other one. The basic idea is that there is an `AttributeAccessor` class that allows code to interact with a set of attributes owned by some geometry. The accessor itself has no ownership. `AttributeAccessor` is a simple type that can be passed around by value. That makes it easy to return it from functions and to store it in containers. For const-correctness, there is also a `MutableAttributeAccessor` that allows changing individual and can add or remove attributes. Currently, `AttributeAccessor` is composed of two pointers. The first is a pointer to the owner of the attribute data. The second is a pointer to a struct with function pointers, that is similar to a virtual function table. The functions know how to access attributes on the owner. The actual attribute access for geometries is still implemented with the `AttributeProvider` pattern, which makes it easy to support different sources of attributes on a geometry and simplifies dealing with built-in attributes. There are different ways to get an attribute accessor for a geometry: * `GeometryComponent.attributes()` * `CurvesGeometry.attributes()` * `bke::mesh_attributes(const Mesh &)` * `bke::pointcloud_attributes(const PointCloud &)` All of these also have a `_for_write` variant that returns a `MutabelAttributeAccessor`. Differential Revision: https://developer.blender.org/D15280
2022-07-08 16:16:56 +02:00
static void tag_component_positions_changed(void *owner)
{
Geometry Nodes: new geometry attribute API Currently, there are two attribute API. The first, defined in `BKE_attribute.h` is accessible from RNA and C code. The second is implemented with `GeometryComponent` and is only accessible in C++ code. The second is widely used, but only being accessible through the `GeometrySet` API makes it awkward to use, and even impossible for types that don't correspond directly to a geometry component like `CurvesGeometry`. This patch adds a new attribute API, designed to replace the `GeometryComponent` attribute API now, and to eventually replace or be the basis of the other one. The basic idea is that there is an `AttributeAccessor` class that allows code to interact with a set of attributes owned by some geometry. The accessor itself has no ownership. `AttributeAccessor` is a simple type that can be passed around by value. That makes it easy to return it from functions and to store it in containers. For const-correctness, there is also a `MutableAttributeAccessor` that allows changing individual and can add or remove attributes. Currently, `AttributeAccessor` is composed of two pointers. The first is a pointer to the owner of the attribute data. The second is a pointer to a struct with function pointers, that is similar to a virtual function table. The functions know how to access attributes on the owner. The actual attribute access for geometries is still implemented with the `AttributeProvider` pattern, which makes it easy to support different sources of attributes on a geometry and simplifies dealing with built-in attributes. There are different ways to get an attribute accessor for a geometry: * `GeometryComponent.attributes()` * `CurvesGeometry.attributes()` * `bke::mesh_attributes(const Mesh &)` * `bke::pointcloud_attributes(const PointCloud &)` All of these also have a `_for_write` variant that returns a `MutabelAttributeAccessor`. Differential Revision: https://developer.blender.org/D15280
2022-07-08 16:16:56 +02:00
Mesh *mesh = static_cast<Mesh *>(owner);
if (mesh != nullptr) {
BKE_mesh_tag_positions_changed(mesh);
}
}
class VArrayImpl_For_VertexWeights final : public VMutableArrayImpl<float> {
private:
MDeformVert *dverts_;
const int dvert_index_;
public:
Mesh: Remove redundant custom data pointers For copy-on-write, we want to share attribute arrays between meshes where possible. Mutable pointers like `Mesh.mvert` make that difficult by making ownership vague. They also make code more complex by adding redundancy. The simplest solution is just removing them and retrieving layers from `CustomData` as needed. Similar changes have already been applied to curves and point clouds (e9f82d3dc7ee, 410a6efb747f). Removing use of the pointers generally makes code more obvious and more reusable. Mesh data is now accessed with a C++ API (`Mesh::edges()` or `Mesh::edges_for_write()`), and a C API (`BKE_mesh_edges(mesh)`). The CoW changes this commit makes possible are described in T95845 and T95842, and started in D14139 and D14140. The change also simplifies the ongoing mesh struct-of-array refactors from T95965. **RNA/Python Access Performance** Theoretically, accessing mesh elements with the RNA API may become slower, since the layer needs to be found on every random access. However, overhead is already high enough that this doesn't make a noticible differenc, and performance is actually improved in some cases. Random access can be up to 10% faster, but other situations might be a bit slower. Generally using `foreach_get/set` are the best way to improve performance. See the differential revision for more discussion about Python performance. Cycles has been updated to use raw pointers and the internal Blender mesh types, mostly because there is no sense in having this overhead when it's already compiled with Blender. In my tests this roughly halves the Cycles mesh creation time (0.19s to 0.10s for a 1 million face grid). Differential Revision: https://developer.blender.org/D15488
2022-09-05 11:56:34 -05:00
VArrayImpl_For_VertexWeights(MutableSpan<MDeformVert> dverts, const int dvert_index)
: VMutableArrayImpl<float>(dverts.size()), dverts_(dverts.data()), dvert_index_(dvert_index)
{
}
VArrayImpl_For_VertexWeights(Span<MDeformVert> dverts, const int dvert_index)
: VMutableArrayImpl<float>(dverts.size()),
dverts_(const_cast<MDeformVert *>(dverts.data())),
dvert_index_(dvert_index)
{
}
Geometry Nodes: refactor virtual array system Goals of this refactor: * Simplify creating virtual arrays. * Simplify passing virtual arrays around. * Simplify converting between typed and generic virtual arrays. * Reduce memory allocations. As a quick reminder, a virtual arrays is a data structure that behaves like an array (i.e. it can be accessed using an index). However, it may not actually be stored as array internally. The two most important implementations of virtual arrays are those that correspond to an actual plain array and those that have the same value for every index. However, many more implementations exist for various reasons (interfacing with legacy attributes, unified iterator over all points in multiple splines, ...). With this refactor the core types (`VArray`, `GVArray`, `VMutableArray` and `GVMutableArray`) can be used like "normal values". They typically live on the stack. Before, they were usually inside a `std::unique_ptr`. This makes passing them around much easier. Creation of new virtual arrays is also much simpler now due to some constructors. Memory allocations are reduced by making use of small object optimization inside the core types. Previously, `VArray` was a class with virtual methods that had to be overridden to change the behavior of a the virtual array. Now,`VArray` has a fixed size and has no virtual methods. Instead it contains a `VArrayImpl` that is similar to the old `VArray`. `VArrayImpl` should rarely ever be used directly, unless a new virtual array implementation is added. To support the small object optimization for many `VArrayImpl` classes, a new `blender::Any` type is added. It is similar to `std::any` with two additional features. It has an adjustable inline buffer size and alignment. The inline buffer size of `std::any` can't be relied on and is usually too small for our use case here. Furthermore, `blender::Any` can store additional user-defined type information without increasing the stack size. Differential Revision: https://developer.blender.org/D12986
2021-11-16 10:15:51 +01:00
float get(const int64_t index) const override
{
if (dverts_ == nullptr) {
Geometry Nodes: use virtual arrays in internal attribute api A virtual array is a data structure that is similar to a normal array in that its elements can be accessed by an index. However, a virtual array does not have to be a contiguous array internally. Instead, its elements can be layed out arbitrarily while element access happens through a virtual function call. However, the virtual array data structures are designed so that the virtual function call can be avoided in cases where it could become a bottleneck. Most commonly, a virtual array is backed by an actual array/span or is a single value internally, that is the same for every index. Besides those, there are many more specialized virtual arrays like the ones that provides vertex positions based on the `MVert` struct or vertex group weights. Not all attributes used by geometry nodes are stored in simple contiguous arrays. To provide uniform access to all kinds of attributes, the attribute API has to provide virtual array functionality that hides the implementation details of attributes. Before this refactor, the attribute API provided its own virtual array implementation as part of the `ReadAttribute` and `WriteAttribute` types. That resulted in unnecessary code duplication with the virtual array system. Even worse, it bound many algorithms used by geometry nodes to the specifics of the attribute API, even though they could also use different data sources (such as data from sockets, default values, later results of expressions, ...). This refactor removes the `ReadAttribute` and `WriteAttribute` types and replaces them with `GVArray` and `GVMutableArray` respectively. The `GV` stands for "generic virtual". The "generic" means that the data type contained in those virtual arrays is only known at run-time. There are the corresponding statically typed types `VArray<T>` and `VMutableArray<T>` as well. No regressions are expected from this refactor. It does come with one improvement for users. The attribute API can convert the data type on write now. This is especially useful when writing to builtin attributes like `material_index` with e.g. the Attribute Math node (which usually just writes to float attributes, while `material_index` is an integer attribute). Differential Revision: https://developer.blender.org/D10994
2021-04-17 16:41:03 +02:00
return 0.0f;
}
if (const MDeformWeight *weight = this->find_weight_at_index(index)) {
return weight->weight;
}
Geometry Nodes: use virtual arrays in internal attribute api A virtual array is a data structure that is similar to a normal array in that its elements can be accessed by an index. However, a virtual array does not have to be a contiguous array internally. Instead, its elements can be layed out arbitrarily while element access happens through a virtual function call. However, the virtual array data structures are designed so that the virtual function call can be avoided in cases where it could become a bottleneck. Most commonly, a virtual array is backed by an actual array/span or is a single value internally, that is the same for every index. Besides those, there are many more specialized virtual arrays like the ones that provides vertex positions based on the `MVert` struct or vertex group weights. Not all attributes used by geometry nodes are stored in simple contiguous arrays. To provide uniform access to all kinds of attributes, the attribute API has to provide virtual array functionality that hides the implementation details of attributes. Before this refactor, the attribute API provided its own virtual array implementation as part of the `ReadAttribute` and `WriteAttribute` types. That resulted in unnecessary code duplication with the virtual array system. Even worse, it bound many algorithms used by geometry nodes to the specifics of the attribute API, even though they could also use different data sources (such as data from sockets, default values, later results of expressions, ...). This refactor removes the `ReadAttribute` and `WriteAttribute` types and replaces them with `GVArray` and `GVMutableArray` respectively. The `GV` stands for "generic virtual". The "generic" means that the data type contained in those virtual arrays is only known at run-time. There are the corresponding statically typed types `VArray<T>` and `VMutableArray<T>` as well. No regressions are expected from this refactor. It does come with one improvement for users. The attribute API can convert the data type on write now. This is especially useful when writing to builtin attributes like `material_index` with e.g. the Attribute Math node (which usually just writes to float attributes, while `material_index` is an integer attribute). Differential Revision: https://developer.blender.org/D10994
2021-04-17 16:41:03 +02:00
return 0.0f;
}
void set(const int64_t index, const float value) override
{
MDeformVert &dvert = dverts_[index];
if (value == 0.0f) {
if (MDeformWeight *weight = this->find_weight_at_index(index)) {
weight->weight = 0.0f;
}
}
else {
MDeformWeight *weight = BKE_defvert_ensure_index(&dvert, dvert_index_);
weight->weight = value;
}
}
void set_all(Span<float> src) override
{
threading::parallel_for(src.index_range(), 4096, [&](const IndexRange range) {
for (const int64_t i : range) {
this->set(i, src[i]);
}
});
}
void materialize(IndexMask mask, float *dst) const override
{
if (dverts_ == nullptr) {
mask.foreach_index([&](const int i) { dst[i] = 0.0f; });
}
threading::parallel_for(mask.index_range(), 4096, [&](const IndexRange range) {
for (const int64_t i : mask.slice(range)) {
if (const MDeformWeight *weight = this->find_weight_at_index(i)) {
dst[i] = weight->weight;
}
else {
dst[i] = 0.0f;
}
}
});
}
void materialize_to_uninitialized(IndexMask mask, float *dst) const override
{
this->materialize(mask, dst);
}
private:
MDeformWeight *find_weight_at_index(const int64_t index)
{
for (MDeformWeight &weight : MutableSpan(dverts_[index].dw, dverts_[index].totweight)) {
if (weight.def_nr == dvert_index_) {
return &weight;
}
}
return nullptr;
}
const MDeformWeight *find_weight_at_index(const int64_t index) const
{
for (const MDeformWeight &weight : Span(dverts_[index].dw, dverts_[index].totweight)) {
if (weight.def_nr == dvert_index_) {
return &weight;
}
}
return nullptr;
}
};
/**
* This provider makes vertex groups available as float attributes.
*/
class VertexGroupsAttributeProvider final : public DynamicAttributesProvider {
public:
Geometry Nodes: new geometry attribute API Currently, there are two attribute API. The first, defined in `BKE_attribute.h` is accessible from RNA and C code. The second is implemented with `GeometryComponent` and is only accessible in C++ code. The second is widely used, but only being accessible through the `GeometrySet` API makes it awkward to use, and even impossible for types that don't correspond directly to a geometry component like `CurvesGeometry`. This patch adds a new attribute API, designed to replace the `GeometryComponent` attribute API now, and to eventually replace or be the basis of the other one. The basic idea is that there is an `AttributeAccessor` class that allows code to interact with a set of attributes owned by some geometry. The accessor itself has no ownership. `AttributeAccessor` is a simple type that can be passed around by value. That makes it easy to return it from functions and to store it in containers. For const-correctness, there is also a `MutableAttributeAccessor` that allows changing individual and can add or remove attributes. Currently, `AttributeAccessor` is composed of two pointers. The first is a pointer to the owner of the attribute data. The second is a pointer to a struct with function pointers, that is similar to a virtual function table. The functions know how to access attributes on the owner. The actual attribute access for geometries is still implemented with the `AttributeProvider` pattern, which makes it easy to support different sources of attributes on a geometry and simplifies dealing with built-in attributes. There are different ways to get an attribute accessor for a geometry: * `GeometryComponent.attributes()` * `CurvesGeometry.attributes()` * `bke::mesh_attributes(const Mesh &)` * `bke::pointcloud_attributes(const PointCloud &)` All of these also have a `_for_write` variant that returns a `MutabelAttributeAccessor`. Differential Revision: https://developer.blender.org/D15280
2022-07-08 16:16:56 +02:00
GAttributeReader try_get_for_read(const void *owner,
const AttributeIDRef &attribute_id) const final
{
if (attribute_id.is_anonymous()) {
return {};
}
Geometry Nodes: new geometry attribute API Currently, there are two attribute API. The first, defined in `BKE_attribute.h` is accessible from RNA and C code. The second is implemented with `GeometryComponent` and is only accessible in C++ code. The second is widely used, but only being accessible through the `GeometrySet` API makes it awkward to use, and even impossible for types that don't correspond directly to a geometry component like `CurvesGeometry`. This patch adds a new attribute API, designed to replace the `GeometryComponent` attribute API now, and to eventually replace or be the basis of the other one. The basic idea is that there is an `AttributeAccessor` class that allows code to interact with a set of attributes owned by some geometry. The accessor itself has no ownership. `AttributeAccessor` is a simple type that can be passed around by value. That makes it easy to return it from functions and to store it in containers. For const-correctness, there is also a `MutableAttributeAccessor` that allows changing individual and can add or remove attributes. Currently, `AttributeAccessor` is composed of two pointers. The first is a pointer to the owner of the attribute data. The second is a pointer to a struct with function pointers, that is similar to a virtual function table. The functions know how to access attributes on the owner. The actual attribute access for geometries is still implemented with the `AttributeProvider` pattern, which makes it easy to support different sources of attributes on a geometry and simplifies dealing with built-in attributes. There are different ways to get an attribute accessor for a geometry: * `GeometryComponent.attributes()` * `CurvesGeometry.attributes()` * `bke::mesh_attributes(const Mesh &)` * `bke::pointcloud_attributes(const PointCloud &)` All of these also have a `_for_write` variant that returns a `MutabelAttributeAccessor`. Differential Revision: https://developer.blender.org/D15280
2022-07-08 16:16:56 +02:00
const Mesh *mesh = static_cast<const Mesh *>(owner);
if (mesh == nullptr) {
return {};
}
const std::string name = attribute_id.name();
const int vertex_group_index = BLI_findstringindex(
&mesh->vertex_group_names, name.c_str(), offsetof(bDeformGroup, name));
if (vertex_group_index < 0) {
return {};
}
Mesh: Remove redundant custom data pointers For copy-on-write, we want to share attribute arrays between meshes where possible. Mutable pointers like `Mesh.mvert` make that difficult by making ownership vague. They also make code more complex by adding redundancy. The simplest solution is just removing them and retrieving layers from `CustomData` as needed. Similar changes have already been applied to curves and point clouds (e9f82d3dc7ee, 410a6efb747f). Removing use of the pointers generally makes code more obvious and more reusable. Mesh data is now accessed with a C++ API (`Mesh::edges()` or `Mesh::edges_for_write()`), and a C API (`BKE_mesh_edges(mesh)`). The CoW changes this commit makes possible are described in T95845 and T95842, and started in D14139 and D14140. The change also simplifies the ongoing mesh struct-of-array refactors from T95965. **RNA/Python Access Performance** Theoretically, accessing mesh elements with the RNA API may become slower, since the layer needs to be found on every random access. However, overhead is already high enough that this doesn't make a noticible differenc, and performance is actually improved in some cases. Random access can be up to 10% faster, but other situations might be a bit slower. Generally using `foreach_get/set` are the best way to improve performance. See the differential revision for more discussion about Python performance. Cycles has been updated to use raw pointers and the internal Blender mesh types, mostly because there is no sense in having this overhead when it's already compiled with Blender. In my tests this roughly halves the Cycles mesh creation time (0.19s to 0.10s for a 1 million face grid). Differential Revision: https://developer.blender.org/D15488
2022-09-05 11:56:34 -05:00
const Span<MDeformVert> dverts = mesh->deform_verts();
if (dverts.is_empty()) {
static const float default_value = 0.0f;
Geometry Nodes: refactor virtual array system Goals of this refactor: * Simplify creating virtual arrays. * Simplify passing virtual arrays around. * Simplify converting between typed and generic virtual arrays. * Reduce memory allocations. As a quick reminder, a virtual arrays is a data structure that behaves like an array (i.e. it can be accessed using an index). However, it may not actually be stored as array internally. The two most important implementations of virtual arrays are those that correspond to an actual plain array and those that have the same value for every index. However, many more implementations exist for various reasons (interfacing with legacy attributes, unified iterator over all points in multiple splines, ...). With this refactor the core types (`VArray`, `GVArray`, `VMutableArray` and `GVMutableArray`) can be used like "normal values". They typically live on the stack. Before, they were usually inside a `std::unique_ptr`. This makes passing them around much easier. Creation of new virtual arrays is also much simpler now due to some constructors. Memory allocations are reduced by making use of small object optimization inside the core types. Previously, `VArray` was a class with virtual methods that had to be overridden to change the behavior of a the virtual array. Now,`VArray` has a fixed size and has no virtual methods. Instead it contains a `VArrayImpl` that is similar to the old `VArray`. `VArrayImpl` should rarely ever be used directly, unless a new virtual array implementation is added. To support the small object optimization for many `VArrayImpl` classes, a new `blender::Any` type is added. It is similar to `std::any` with two additional features. It has an adjustable inline buffer size and alignment. The inline buffer size of `std::any` can't be relied on and is usually too small for our use case here. Furthermore, `blender::Any` can store additional user-defined type information without increasing the stack size. Differential Revision: https://developer.blender.org/D12986
2021-11-16 10:15:51 +01:00
return {VArray<float>::ForSingle(default_value, mesh->totvert), ATTR_DOMAIN_POINT};
}
Mesh: Remove redundant custom data pointers For copy-on-write, we want to share attribute arrays between meshes where possible. Mutable pointers like `Mesh.mvert` make that difficult by making ownership vague. They also make code more complex by adding redundancy. The simplest solution is just removing them and retrieving layers from `CustomData` as needed. Similar changes have already been applied to curves and point clouds (e9f82d3dc7ee, 410a6efb747f). Removing use of the pointers generally makes code more obvious and more reusable. Mesh data is now accessed with a C++ API (`Mesh::edges()` or `Mesh::edges_for_write()`), and a C API (`BKE_mesh_edges(mesh)`). The CoW changes this commit makes possible are described in T95845 and T95842, and started in D14139 and D14140. The change also simplifies the ongoing mesh struct-of-array refactors from T95965. **RNA/Python Access Performance** Theoretically, accessing mesh elements with the RNA API may become slower, since the layer needs to be found on every random access. However, overhead is already high enough that this doesn't make a noticible differenc, and performance is actually improved in some cases. Random access can be up to 10% faster, but other situations might be a bit slower. Generally using `foreach_get/set` are the best way to improve performance. See the differential revision for more discussion about Python performance. Cycles has been updated to use raw pointers and the internal Blender mesh types, mostly because there is no sense in having this overhead when it's already compiled with Blender. In my tests this roughly halves the Cycles mesh creation time (0.19s to 0.10s for a 1 million face grid). Differential Revision: https://developer.blender.org/D15488
2022-09-05 11:56:34 -05:00
return {VArray<float>::For<VArrayImpl_For_VertexWeights>(dverts, vertex_group_index),
Geometry Nodes: use virtual arrays in internal attribute api A virtual array is a data structure that is similar to a normal array in that its elements can be accessed by an index. However, a virtual array does not have to be a contiguous array internally. Instead, its elements can be layed out arbitrarily while element access happens through a virtual function call. However, the virtual array data structures are designed so that the virtual function call can be avoided in cases where it could become a bottleneck. Most commonly, a virtual array is backed by an actual array/span or is a single value internally, that is the same for every index. Besides those, there are many more specialized virtual arrays like the ones that provides vertex positions based on the `MVert` struct or vertex group weights. Not all attributes used by geometry nodes are stored in simple contiguous arrays. To provide uniform access to all kinds of attributes, the attribute API has to provide virtual array functionality that hides the implementation details of attributes. Before this refactor, the attribute API provided its own virtual array implementation as part of the `ReadAttribute` and `WriteAttribute` types. That resulted in unnecessary code duplication with the virtual array system. Even worse, it bound many algorithms used by geometry nodes to the specifics of the attribute API, even though they could also use different data sources (such as data from sockets, default values, later results of expressions, ...). This refactor removes the `ReadAttribute` and `WriteAttribute` types and replaces them with `GVArray` and `GVMutableArray` respectively. The `GV` stands for "generic virtual". The "generic" means that the data type contained in those virtual arrays is only known at run-time. There are the corresponding statically typed types `VArray<T>` and `VMutableArray<T>` as well. No regressions are expected from this refactor. It does come with one improvement for users. The attribute API can convert the data type on write now. This is especially useful when writing to builtin attributes like `material_index` with e.g. the Attribute Math node (which usually just writes to float attributes, while `material_index` is an integer attribute). Differential Revision: https://developer.blender.org/D10994
2021-04-17 16:41:03 +02:00
ATTR_DOMAIN_POINT};
}
Geometry Nodes: new geometry attribute API Currently, there are two attribute API. The first, defined in `BKE_attribute.h` is accessible from RNA and C code. The second is implemented with `GeometryComponent` and is only accessible in C++ code. The second is widely used, but only being accessible through the `GeometrySet` API makes it awkward to use, and even impossible for types that don't correspond directly to a geometry component like `CurvesGeometry`. This patch adds a new attribute API, designed to replace the `GeometryComponent` attribute API now, and to eventually replace or be the basis of the other one. The basic idea is that there is an `AttributeAccessor` class that allows code to interact with a set of attributes owned by some geometry. The accessor itself has no ownership. `AttributeAccessor` is a simple type that can be passed around by value. That makes it easy to return it from functions and to store it in containers. For const-correctness, there is also a `MutableAttributeAccessor` that allows changing individual and can add or remove attributes. Currently, `AttributeAccessor` is composed of two pointers. The first is a pointer to the owner of the attribute data. The second is a pointer to a struct with function pointers, that is similar to a virtual function table. The functions know how to access attributes on the owner. The actual attribute access for geometries is still implemented with the `AttributeProvider` pattern, which makes it easy to support different sources of attributes on a geometry and simplifies dealing with built-in attributes. There are different ways to get an attribute accessor for a geometry: * `GeometryComponent.attributes()` * `CurvesGeometry.attributes()` * `bke::mesh_attributes(const Mesh &)` * `bke::pointcloud_attributes(const PointCloud &)` All of these also have a `_for_write` variant that returns a `MutabelAttributeAccessor`. Differential Revision: https://developer.blender.org/D15280
2022-07-08 16:16:56 +02:00
GAttributeWriter try_get_for_write(void *owner, const AttributeIDRef &attribute_id) const final
{
if (attribute_id.is_anonymous()) {
return {};
}
Geometry Nodes: new geometry attribute API Currently, there are two attribute API. The first, defined in `BKE_attribute.h` is accessible from RNA and C code. The second is implemented with `GeometryComponent` and is only accessible in C++ code. The second is widely used, but only being accessible through the `GeometrySet` API makes it awkward to use, and even impossible for types that don't correspond directly to a geometry component like `CurvesGeometry`. This patch adds a new attribute API, designed to replace the `GeometryComponent` attribute API now, and to eventually replace or be the basis of the other one. The basic idea is that there is an `AttributeAccessor` class that allows code to interact with a set of attributes owned by some geometry. The accessor itself has no ownership. `AttributeAccessor` is a simple type that can be passed around by value. That makes it easy to return it from functions and to store it in containers. For const-correctness, there is also a `MutableAttributeAccessor` that allows changing individual and can add or remove attributes. Currently, `AttributeAccessor` is composed of two pointers. The first is a pointer to the owner of the attribute data. The second is a pointer to a struct with function pointers, that is similar to a virtual function table. The functions know how to access attributes on the owner. The actual attribute access for geometries is still implemented with the `AttributeProvider` pattern, which makes it easy to support different sources of attributes on a geometry and simplifies dealing with built-in attributes. There are different ways to get an attribute accessor for a geometry: * `GeometryComponent.attributes()` * `CurvesGeometry.attributes()` * `bke::mesh_attributes(const Mesh &)` * `bke::pointcloud_attributes(const PointCloud &)` All of these also have a `_for_write` variant that returns a `MutabelAttributeAccessor`. Differential Revision: https://developer.blender.org/D15280
2022-07-08 16:16:56 +02:00
Mesh *mesh = static_cast<Mesh *>(owner);
if (mesh == nullptr) {
return {};
}
const std::string name = attribute_id.name();
const int vertex_group_index = BLI_findstringindex(
&mesh->vertex_group_names, name.c_str(), offsetof(bDeformGroup, name));
if (vertex_group_index < 0) {
return {};
}
Mesh: Remove redundant custom data pointers For copy-on-write, we want to share attribute arrays between meshes where possible. Mutable pointers like `Mesh.mvert` make that difficult by making ownership vague. They also make code more complex by adding redundancy. The simplest solution is just removing them and retrieving layers from `CustomData` as needed. Similar changes have already been applied to curves and point clouds (e9f82d3dc7ee, 410a6efb747f). Removing use of the pointers generally makes code more obvious and more reusable. Mesh data is now accessed with a C++ API (`Mesh::edges()` or `Mesh::edges_for_write()`), and a C API (`BKE_mesh_edges(mesh)`). The CoW changes this commit makes possible are described in T95845 and T95842, and started in D14139 and D14140. The change also simplifies the ongoing mesh struct-of-array refactors from T95965. **RNA/Python Access Performance** Theoretically, accessing mesh elements with the RNA API may become slower, since the layer needs to be found on every random access. However, overhead is already high enough that this doesn't make a noticible differenc, and performance is actually improved in some cases. Random access can be up to 10% faster, but other situations might be a bit slower. Generally using `foreach_get/set` are the best way to improve performance. See the differential revision for more discussion about Python performance. Cycles has been updated to use raw pointers and the internal Blender mesh types, mostly because there is no sense in having this overhead when it's already compiled with Blender. In my tests this roughly halves the Cycles mesh creation time (0.19s to 0.10s for a 1 million face grid). Differential Revision: https://developer.blender.org/D15488
2022-09-05 11:56:34 -05:00
MutableSpan<MDeformVert> dverts = mesh->deform_verts_for_write();
return {VMutableArray<float>::For<VArrayImpl_For_VertexWeights>(dverts, vertex_group_index),
Geometry Nodes: refactor virtual array system Goals of this refactor: * Simplify creating virtual arrays. * Simplify passing virtual arrays around. * Simplify converting between typed and generic virtual arrays. * Reduce memory allocations. As a quick reminder, a virtual arrays is a data structure that behaves like an array (i.e. it can be accessed using an index). However, it may not actually be stored as array internally. The two most important implementations of virtual arrays are those that correspond to an actual plain array and those that have the same value for every index. However, many more implementations exist for various reasons (interfacing with legacy attributes, unified iterator over all points in multiple splines, ...). With this refactor the core types (`VArray`, `GVArray`, `VMutableArray` and `GVMutableArray`) can be used like "normal values". They typically live on the stack. Before, they were usually inside a `std::unique_ptr`. This makes passing them around much easier. Creation of new virtual arrays is also much simpler now due to some constructors. Memory allocations are reduced by making use of small object optimization inside the core types. Previously, `VArray` was a class with virtual methods that had to be overridden to change the behavior of a the virtual array. Now,`VArray` has a fixed size and has no virtual methods. Instead it contains a `VArrayImpl` that is similar to the old `VArray`. `VArrayImpl` should rarely ever be used directly, unless a new virtual array implementation is added. To support the small object optimization for many `VArrayImpl` classes, a new `blender::Any` type is added. It is similar to `std::any` with two additional features. It has an adjustable inline buffer size and alignment. The inline buffer size of `std::any` can't be relied on and is usually too small for our use case here. Furthermore, `blender::Any` can store additional user-defined type information without increasing the stack size. Differential Revision: https://developer.blender.org/D12986
2021-11-16 10:15:51 +01:00
ATTR_DOMAIN_POINT};
}
Geometry Nodes: new geometry attribute API Currently, there are two attribute API. The first, defined in `BKE_attribute.h` is accessible from RNA and C code. The second is implemented with `GeometryComponent` and is only accessible in C++ code. The second is widely used, but only being accessible through the `GeometrySet` API makes it awkward to use, and even impossible for types that don't correspond directly to a geometry component like `CurvesGeometry`. This patch adds a new attribute API, designed to replace the `GeometryComponent` attribute API now, and to eventually replace or be the basis of the other one. The basic idea is that there is an `AttributeAccessor` class that allows code to interact with a set of attributes owned by some geometry. The accessor itself has no ownership. `AttributeAccessor` is a simple type that can be passed around by value. That makes it easy to return it from functions and to store it in containers. For const-correctness, there is also a `MutableAttributeAccessor` that allows changing individual and can add or remove attributes. Currently, `AttributeAccessor` is composed of two pointers. The first is a pointer to the owner of the attribute data. The second is a pointer to a struct with function pointers, that is similar to a virtual function table. The functions know how to access attributes on the owner. The actual attribute access for geometries is still implemented with the `AttributeProvider` pattern, which makes it easy to support different sources of attributes on a geometry and simplifies dealing with built-in attributes. There are different ways to get an attribute accessor for a geometry: * `GeometryComponent.attributes()` * `CurvesGeometry.attributes()` * `bke::mesh_attributes(const Mesh &)` * `bke::pointcloud_attributes(const PointCloud &)` All of these also have a `_for_write` variant that returns a `MutabelAttributeAccessor`. Differential Revision: https://developer.blender.org/D15280
2022-07-08 16:16:56 +02:00
bool try_delete(void *owner, const AttributeIDRef &attribute_id) const final
{
if (attribute_id.is_anonymous()) {
return false;
}
Geometry Nodes: new geometry attribute API Currently, there are two attribute API. The first, defined in `BKE_attribute.h` is accessible from RNA and C code. The second is implemented with `GeometryComponent` and is only accessible in C++ code. The second is widely used, but only being accessible through the `GeometrySet` API makes it awkward to use, and even impossible for types that don't correspond directly to a geometry component like `CurvesGeometry`. This patch adds a new attribute API, designed to replace the `GeometryComponent` attribute API now, and to eventually replace or be the basis of the other one. The basic idea is that there is an `AttributeAccessor` class that allows code to interact with a set of attributes owned by some geometry. The accessor itself has no ownership. `AttributeAccessor` is a simple type that can be passed around by value. That makes it easy to return it from functions and to store it in containers. For const-correctness, there is also a `MutableAttributeAccessor` that allows changing individual and can add or remove attributes. Currently, `AttributeAccessor` is composed of two pointers. The first is a pointer to the owner of the attribute data. The second is a pointer to a struct with function pointers, that is similar to a virtual function table. The functions know how to access attributes on the owner. The actual attribute access for geometries is still implemented with the `AttributeProvider` pattern, which makes it easy to support different sources of attributes on a geometry and simplifies dealing with built-in attributes. There are different ways to get an attribute accessor for a geometry: * `GeometryComponent.attributes()` * `CurvesGeometry.attributes()` * `bke::mesh_attributes(const Mesh &)` * `bke::pointcloud_attributes(const PointCloud &)` All of these also have a `_for_write` variant that returns a `MutabelAttributeAccessor`. Differential Revision: https://developer.blender.org/D15280
2022-07-08 16:16:56 +02:00
Mesh *mesh = static_cast<Mesh *>(owner);
if (mesh == nullptr) {
return true;
}
const std::string name = attribute_id.name();
int index;
bDeformGroup *group;
if (!BKE_id_defgroup_name_find(&mesh->id, name.c_str(), &index, &group)) {
return false;
}
BLI_remlink(&mesh->vertex_group_names, group);
MEM_freeN(group);
Mesh: Remove redundant custom data pointers For copy-on-write, we want to share attribute arrays between meshes where possible. Mutable pointers like `Mesh.mvert` make that difficult by making ownership vague. They also make code more complex by adding redundancy. The simplest solution is just removing them and retrieving layers from `CustomData` as needed. Similar changes have already been applied to curves and point clouds (e9f82d3dc7ee, 410a6efb747f). Removing use of the pointers generally makes code more obvious and more reusable. Mesh data is now accessed with a C++ API (`Mesh::edges()` or `Mesh::edges_for_write()`), and a C API (`BKE_mesh_edges(mesh)`). The CoW changes this commit makes possible are described in T95845 and T95842, and started in D14139 and D14140. The change also simplifies the ongoing mesh struct-of-array refactors from T95965. **RNA/Python Access Performance** Theoretically, accessing mesh elements with the RNA API may become slower, since the layer needs to be found on every random access. However, overhead is already high enough that this doesn't make a noticible differenc, and performance is actually improved in some cases. Random access can be up to 10% faster, but other situations might be a bit slower. Generally using `foreach_get/set` are the best way to improve performance. See the differential revision for more discussion about Python performance. Cycles has been updated to use raw pointers and the internal Blender mesh types, mostly because there is no sense in having this overhead when it's already compiled with Blender. In my tests this roughly halves the Cycles mesh creation time (0.19s to 0.10s for a 1 million face grid). Differential Revision: https://developer.blender.org/D15488
2022-09-05 11:56:34 -05:00
if (mesh->deform_verts().is_empty()) {
return true;
}
MutableSpan<MDeformVert> dverts = mesh->deform_verts_for_write();
threading::parallel_for(dverts.index_range(), 1024, [&](IndexRange range) {
for (MDeformVert &dvert : dverts.slice(range)) {
MDeformWeight *weight = BKE_defvert_find_index(&dvert, index);
BKE_defvert_remove_group(&dvert, weight);
for (MDeformWeight &weight : MutableSpan(dvert.dw, dvert.totweight)) {
if (weight.def_nr > index) {
weight.def_nr--;
}
}
}
});
return true;
}
Geometry Nodes: new geometry attribute API Currently, there are two attribute API. The first, defined in `BKE_attribute.h` is accessible from RNA and C code. The second is implemented with `GeometryComponent` and is only accessible in C++ code. The second is widely used, but only being accessible through the `GeometrySet` API makes it awkward to use, and even impossible for types that don't correspond directly to a geometry component like `CurvesGeometry`. This patch adds a new attribute API, designed to replace the `GeometryComponent` attribute API now, and to eventually replace or be the basis of the other one. The basic idea is that there is an `AttributeAccessor` class that allows code to interact with a set of attributes owned by some geometry. The accessor itself has no ownership. `AttributeAccessor` is a simple type that can be passed around by value. That makes it easy to return it from functions and to store it in containers. For const-correctness, there is also a `MutableAttributeAccessor` that allows changing individual and can add or remove attributes. Currently, `AttributeAccessor` is composed of two pointers. The first is a pointer to the owner of the attribute data. The second is a pointer to a struct with function pointers, that is similar to a virtual function table. The functions know how to access attributes on the owner. The actual attribute access for geometries is still implemented with the `AttributeProvider` pattern, which makes it easy to support different sources of attributes on a geometry and simplifies dealing with built-in attributes. There are different ways to get an attribute accessor for a geometry: * `GeometryComponent.attributes()` * `CurvesGeometry.attributes()` * `bke::mesh_attributes(const Mesh &)` * `bke::pointcloud_attributes(const PointCloud &)` All of these also have a `_for_write` variant that returns a `MutabelAttributeAccessor`. Differential Revision: https://developer.blender.org/D15280
2022-07-08 16:16:56 +02:00
bool foreach_attribute(const void *owner, const AttributeForeachCallback callback) const final
{
Geometry Nodes: new geometry attribute API Currently, there are two attribute API. The first, defined in `BKE_attribute.h` is accessible from RNA and C code. The second is implemented with `GeometryComponent` and is only accessible in C++ code. The second is widely used, but only being accessible through the `GeometrySet` API makes it awkward to use, and even impossible for types that don't correspond directly to a geometry component like `CurvesGeometry`. This patch adds a new attribute API, designed to replace the `GeometryComponent` attribute API now, and to eventually replace or be the basis of the other one. The basic idea is that there is an `AttributeAccessor` class that allows code to interact with a set of attributes owned by some geometry. The accessor itself has no ownership. `AttributeAccessor` is a simple type that can be passed around by value. That makes it easy to return it from functions and to store it in containers. For const-correctness, there is also a `MutableAttributeAccessor` that allows changing individual and can add or remove attributes. Currently, `AttributeAccessor` is composed of two pointers. The first is a pointer to the owner of the attribute data. The second is a pointer to a struct with function pointers, that is similar to a virtual function table. The functions know how to access attributes on the owner. The actual attribute access for geometries is still implemented with the `AttributeProvider` pattern, which makes it easy to support different sources of attributes on a geometry and simplifies dealing with built-in attributes. There are different ways to get an attribute accessor for a geometry: * `GeometryComponent.attributes()` * `CurvesGeometry.attributes()` * `bke::mesh_attributes(const Mesh &)` * `bke::pointcloud_attributes(const PointCloud &)` All of these also have a `_for_write` variant that returns a `MutabelAttributeAccessor`. Differential Revision: https://developer.blender.org/D15280
2022-07-08 16:16:56 +02:00
const Mesh *mesh = static_cast<const Mesh *>(owner);
if (mesh == nullptr) {
return true;
}
LISTBASE_FOREACH (const bDeformGroup *, group, &mesh->vertex_group_names) {
if (!callback(group->name, {ATTR_DOMAIN_POINT, CD_PROP_FLOAT})) {
return false;
}
}
return true;
}
void foreach_domain(const FunctionRef<void(eAttrDomain)> callback) const final
{
callback(ATTR_DOMAIN_POINT);
}
};
/**
* In this function all the attribute providers for a mesh component are created. Most data in this
* function is statically allocated, because it does not change over time.
*/
static ComponentAttributeProviders create_attribute_providers_for_mesh()
{
#define MAKE_MUTABLE_CUSTOM_DATA_GETTER(NAME) \
Geometry Nodes: new geometry attribute API Currently, there are two attribute API. The first, defined in `BKE_attribute.h` is accessible from RNA and C code. The second is implemented with `GeometryComponent` and is only accessible in C++ code. The second is widely used, but only being accessible through the `GeometrySet` API makes it awkward to use, and even impossible for types that don't correspond directly to a geometry component like `CurvesGeometry`. This patch adds a new attribute API, designed to replace the `GeometryComponent` attribute API now, and to eventually replace or be the basis of the other one. The basic idea is that there is an `AttributeAccessor` class that allows code to interact with a set of attributes owned by some geometry. The accessor itself has no ownership. `AttributeAccessor` is a simple type that can be passed around by value. That makes it easy to return it from functions and to store it in containers. For const-correctness, there is also a `MutableAttributeAccessor` that allows changing individual and can add or remove attributes. Currently, `AttributeAccessor` is composed of two pointers. The first is a pointer to the owner of the attribute data. The second is a pointer to a struct with function pointers, that is similar to a virtual function table. The functions know how to access attributes on the owner. The actual attribute access for geometries is still implemented with the `AttributeProvider` pattern, which makes it easy to support different sources of attributes on a geometry and simplifies dealing with built-in attributes. There are different ways to get an attribute accessor for a geometry: * `GeometryComponent.attributes()` * `CurvesGeometry.attributes()` * `bke::mesh_attributes(const Mesh &)` * `bke::pointcloud_attributes(const PointCloud &)` All of these also have a `_for_write` variant that returns a `MutabelAttributeAccessor`. Differential Revision: https://developer.blender.org/D15280
2022-07-08 16:16:56 +02:00
[](void *owner) -> CustomData * { \
Mesh *mesh = static_cast<Mesh *>(owner); \
return &mesh->NAME; \
}
#define MAKE_CONST_CUSTOM_DATA_GETTER(NAME) \
Geometry Nodes: new geometry attribute API Currently, there are two attribute API. The first, defined in `BKE_attribute.h` is accessible from RNA and C code. The second is implemented with `GeometryComponent` and is only accessible in C++ code. The second is widely used, but only being accessible through the `GeometrySet` API makes it awkward to use, and even impossible for types that don't correspond directly to a geometry component like `CurvesGeometry`. This patch adds a new attribute API, designed to replace the `GeometryComponent` attribute API now, and to eventually replace or be the basis of the other one. The basic idea is that there is an `AttributeAccessor` class that allows code to interact with a set of attributes owned by some geometry. The accessor itself has no ownership. `AttributeAccessor` is a simple type that can be passed around by value. That makes it easy to return it from functions and to store it in containers. For const-correctness, there is also a `MutableAttributeAccessor` that allows changing individual and can add or remove attributes. Currently, `AttributeAccessor` is composed of two pointers. The first is a pointer to the owner of the attribute data. The second is a pointer to a struct with function pointers, that is similar to a virtual function table. The functions know how to access attributes on the owner. The actual attribute access for geometries is still implemented with the `AttributeProvider` pattern, which makes it easy to support different sources of attributes on a geometry and simplifies dealing with built-in attributes. There are different ways to get an attribute accessor for a geometry: * `GeometryComponent.attributes()` * `CurvesGeometry.attributes()` * `bke::mesh_attributes(const Mesh &)` * `bke::pointcloud_attributes(const PointCloud &)` All of these also have a `_for_write` variant that returns a `MutabelAttributeAccessor`. Differential Revision: https://developer.blender.org/D15280
2022-07-08 16:16:56 +02:00
[](const void *owner) -> const CustomData * { \
const Mesh *mesh = static_cast<const Mesh *>(owner); \
return &mesh->NAME; \
}
#define MAKE_GET_ELEMENT_NUM_GETTER(NAME) \
[](const void *owner) -> int { \
const Mesh *mesh = static_cast<const Mesh *>(owner); \
return mesh->NAME; \
}
static CustomDataAccessInfo corner_access = {MAKE_MUTABLE_CUSTOM_DATA_GETTER(ldata),
MAKE_CONST_CUSTOM_DATA_GETTER(ldata),
MAKE_GET_ELEMENT_NUM_GETTER(totloop)};
static CustomDataAccessInfo point_access = {MAKE_MUTABLE_CUSTOM_DATA_GETTER(vdata),
MAKE_CONST_CUSTOM_DATA_GETTER(vdata),
MAKE_GET_ELEMENT_NUM_GETTER(totvert)};
static CustomDataAccessInfo edge_access = {MAKE_MUTABLE_CUSTOM_DATA_GETTER(edata),
MAKE_CONST_CUSTOM_DATA_GETTER(edata),
MAKE_GET_ELEMENT_NUM_GETTER(totedge)};
static CustomDataAccessInfo face_access = {MAKE_MUTABLE_CUSTOM_DATA_GETTER(pdata),
MAKE_CONST_CUSTOM_DATA_GETTER(pdata),
MAKE_GET_ELEMENT_NUM_GETTER(totpoly)};
#undef MAKE_CONST_CUSTOM_DATA_GETTER
#undef MAKE_MUTABLE_CUSTOM_DATA_GETTER
Mesh: Move positions to a generic attribute **Changes** As described in T93602, this patch removes all use of the `MVert` struct, replacing it with a generic named attribute with the name `"position"`, consistent with other geometry types. Variable names have been changed from `verts` to `positions`, to align with the attribute name and the more generic design (positions are not vertices, they are just an attribute stored on the point domain). This change is made possible by previous commits that moved all other data out of `MVert` to runtime data or other generic attributes. What remains is mostly a simple type change. Though, the type still shows up 859 times, so the patch is quite large. One compromise is that now `CD_MASK_BAREMESH` now contains `CD_PROP_FLOAT3`. With the general move towards generic attributes over custom data types, we are removing use of these type masks anyway. **Benefits** The most obvious benefit is reduced memory usage and the benefits that brings in memory-bound situations. `float3` is only 3 bytes, in comparison to `MVert` which was 4. When there are millions of vertices this starts to matter more. The other benefits come from using a more generic type. Instead of writing algorithms specifically for `MVert`, code can just use arrays of vectors. This will allow eliminating many temporary arrays or wrappers used to extract positions. Many possible improvements aren't implemented in this patch, though I did switch simplify or remove the process of creating temporary position arrays in a few places. The design clarity that "positions are just another attribute" brings allows removing explicit copying of vertices in some procedural operations-- they are just processed like most other attributes. **Performance** This touches so many areas that it's hard to benchmark exhaustively, but I observed some areas as examples. * The mesh line node with 4 million count was 1.5x (8ms to 12ms) faster. * The Spring splash screen went from ~4.3 to ~4.5 fps. * The subdivision surface modifier/node was slightly faster RNA access through Python may be slightly slower, since now we need a name lookup instead of just a custom data type lookup for each index. **Future Improvements** * Remove uses of "vert_coords" functions: * `BKE_mesh_vert_coords_alloc` * `BKE_mesh_vert_coords_get` * `BKE_mesh_vert_coords_apply{_with_mat4}` * Remove more hidden copying of positions * General simplification now possible in many areas * Convert more code to C++ to use `float3` instead of `float[3]` * Currently `reinterpret_cast` is used for those C-API functions Differential Revision: https://developer.blender.org/D15982
2023-01-10 00:10:43 -05:00
static BuiltinCustomDataLayerProvider position("position",
ATTR_DOMAIN_POINT,
CD_PROP_FLOAT3,
CD_PROP_FLOAT3,
BuiltinAttributeProvider::Creatable,
Mesh: Move positions to a generic attribute **Changes** As described in T93602, this patch removes all use of the `MVert` struct, replacing it with a generic named attribute with the name `"position"`, consistent with other geometry types. Variable names have been changed from `verts` to `positions`, to align with the attribute name and the more generic design (positions are not vertices, they are just an attribute stored on the point domain). This change is made possible by previous commits that moved all other data out of `MVert` to runtime data or other generic attributes. What remains is mostly a simple type change. Though, the type still shows up 859 times, so the patch is quite large. One compromise is that now `CD_MASK_BAREMESH` now contains `CD_PROP_FLOAT3`. With the general move towards generic attributes over custom data types, we are removing use of these type masks anyway. **Benefits** The most obvious benefit is reduced memory usage and the benefits that brings in memory-bound situations. `float3` is only 3 bytes, in comparison to `MVert` which was 4. When there are millions of vertices this starts to matter more. The other benefits come from using a more generic type. Instead of writing algorithms specifically for `MVert`, code can just use arrays of vectors. This will allow eliminating many temporary arrays or wrappers used to extract positions. Many possible improvements aren't implemented in this patch, though I did switch simplify or remove the process of creating temporary position arrays in a few places. The design clarity that "positions are just another attribute" brings allows removing explicit copying of vertices in some procedural operations-- they are just processed like most other attributes. **Performance** This touches so many areas that it's hard to benchmark exhaustively, but I observed some areas as examples. * The mesh line node with 4 million count was 1.5x (8ms to 12ms) faster. * The Spring splash screen went from ~4.3 to ~4.5 fps. * The subdivision surface modifier/node was slightly faster RNA access through Python may be slightly slower, since now we need a name lookup instead of just a custom data type lookup for each index. **Future Improvements** * Remove uses of "vert_coords" functions: * `BKE_mesh_vert_coords_alloc` * `BKE_mesh_vert_coords_get` * `BKE_mesh_vert_coords_apply{_with_mat4}` * Remove more hidden copying of positions * General simplification now possible in many areas * Convert more code to C++ to use `float3` instead of `float[3]` * Currently `reinterpret_cast` is used for those C-API functions Differential Revision: https://developer.blender.org/D15982
2023-01-10 00:10:43 -05:00
BuiltinAttributeProvider::NonDeletable,
point_access,
tag_component_positions_changed);
static BuiltinCustomDataLayerProvider id("id",
ATTR_DOMAIN_POINT,
CD_PROP_INT32,
CD_PROP_INT32,
BuiltinAttributeProvider::Creatable,
BuiltinAttributeProvider::Deletable,
point_access,
nullptr);
static const auto material_index_clamp = mf::build::SI1_SO<int, int>(
"Material Index Validate",
[](int value) {
/* Use #short for the maximum since many areas still use that type for indices. */
return std::clamp<int>(value, 0, std::numeric_limits<short>::max());
},
mf::build::exec_presets::AllSpanOrSingle());
static BuiltinCustomDataLayerProvider material_index("material_index",
ATTR_DOMAIN_FACE,
CD_PROP_INT32,
CD_PROP_INT32,
BuiltinAttributeProvider::Creatable,
BuiltinAttributeProvider::Deletable,
face_access,
nullptr,
AttributeValidator{&material_index_clamp});
Mesh: Move edges to a generic attribute Implements #95966, as the final step of #95965. This commit changes the storage of mesh edge vertex indices from the `MEdge` type to the generic `int2` attribute type. This follows the general design for geometry and the attribute system, where the data storage type and the usage semantics are separated. The main benefit of the change is reduced memory usage-- the requirements of storing mesh edges is reduced by 1/3. For example, this saves 8MB on a 1 million vertex grid. This also gives performance benefits to any memory-bound mesh processing algorithm that uses edges. Another benefit is that all of the edge's vertex indices are contiguous. In a few cases, it's helpful to process all of them as `Span<int>` rather than `Span<int2>`. Similarly, the type is more likely to match a generic format used by a library, or code that shouldn't know about specific Blender `Mesh` types. Various Notes: - The `.edge_verts` name is used to reflect a mapping between domains, similar to `.corner_verts`, etc. The period means that it the data shouldn't change arbitrarily by the user or procedural operations. - `edge[0]` is now used instead of `edge.v1` - Signed integers are used instead of unsigned to reduce the mixing of signed-ness, which can be error prone. - All of the previously used core mesh data types (`MVert`, `MEdge`, `MLoop`, `MPoly` are now deprecated. Only generic types are used). - The `vec2i` DNA type is used in the few C files where necessary. Pull Request: https://projects.blender.org/blender/blender/pulls/106638
2023-04-17 13:47:41 +02:00
static const auto int2_index_clamp = mf::build::SI1_SO<int2, int2>(
"Index Validate",
[](int2 value) { return math::max(value, int2(0)); },
mf::build::exec_presets::AllSpanOrSingle());
static BuiltinCustomDataLayerProvider edge_verts(".edge_verts",
ATTR_DOMAIN_EDGE,
CD_PROP_INT32_2D,
CD_PROP_INT32_2D,
BuiltinAttributeProvider::Creatable,
Mesh: Move edges to a generic attribute Implements #95966, as the final step of #95965. This commit changes the storage of mesh edge vertex indices from the `MEdge` type to the generic `int2` attribute type. This follows the general design for geometry and the attribute system, where the data storage type and the usage semantics are separated. The main benefit of the change is reduced memory usage-- the requirements of storing mesh edges is reduced by 1/3. For example, this saves 8MB on a 1 million vertex grid. This also gives performance benefits to any memory-bound mesh processing algorithm that uses edges. Another benefit is that all of the edge's vertex indices are contiguous. In a few cases, it's helpful to process all of them as `Span<int>` rather than `Span<int2>`. Similarly, the type is more likely to match a generic format used by a library, or code that shouldn't know about specific Blender `Mesh` types. Various Notes: - The `.edge_verts` name is used to reflect a mapping between domains, similar to `.corner_verts`, etc. The period means that it the data shouldn't change arbitrarily by the user or procedural operations. - `edge[0]` is now used instead of `edge.v1` - Signed integers are used instead of unsigned to reduce the mixing of signed-ness, which can be error prone. - All of the previously used core mesh data types (`MVert`, `MEdge`, `MLoop`, `MPoly` are now deprecated. Only generic types are used). - The `vec2i` DNA type is used in the few C files where necessary. Pull Request: https://projects.blender.org/blender/blender/pulls/106638
2023-04-17 13:47:41 +02:00
BuiltinAttributeProvider::NonDeletable,
edge_access,
nullptr,
AttributeValidator{&int2_index_clamp});
Mesh: Replace MLoop struct with generic attributes Implements #102359. Split the `MLoop` struct into two separate integer arrays called `corner_verts` and `corner_edges`, referring to the vertex each corner is attached to and the next edge around the face at each corner. These arrays can be sliced to give access to the edges or vertices in a face. Then they are often referred to as "poly_verts" or "poly_edges". The main benefits are halving the necessary memory bandwidth when only one array is used and simplifications from using regular integer indices instead of a special-purpose struct. The commit also starts a renaming from "loop" to "corner" in mesh code. Like the other mesh struct of array refactors, forward compatibility is kept by writing files with the older format. This will be done until 4.0 to ease the transition process. Looking at a small portion of the patch should give a good impression for the rest of the changes. I tried to make the changes as small as possible so it's easy to tell the correctness from the diff. Though I found Blender developers have been very inventive over the last decade when finding different ways to loop over the corners in a face. For performance, nearly every piece of code that deals with `Mesh` is slightly impacted. Any algorithm that is memory bottle-necked should see an improvement. For example, here is a comparison of interpolating a vertex float attribute to face corners (Ryzen 3700x): **Before** (Average: 3.7 ms, Min: 3.4 ms) ``` threading::parallel_for(loops.index_range(), 4096, [&](IndexRange range) { for (const int64_t i : range) { dst[i] = src[loops[i].v]; } }); ``` **After** (Average: 2.9 ms, Min: 2.6 ms) ``` array_utils::gather(src, corner_verts, dst); ``` That's an improvement of 28% to the average timings, and it's also a simplification, since an index-based routine can be used instead. For more examples using the new arrays, see the design task. Pull Request: https://projects.blender.org/blender/blender/pulls/104424
2023-03-20 15:55:13 +01:00
/* Note: This clamping is more of a last resort, since it's quite easy to make an
* invalid mesh that will crash Blender by arbitrarily editing this attribute. */
static const auto int_index_clamp = mf::build::SI1_SO<int, int>(
"Index Validate",
[](int value) { return std::max(value, 0); },
mf::build::exec_presets::AllSpanOrSingle());
static BuiltinCustomDataLayerProvider corner_vert(".corner_vert",
ATTR_DOMAIN_CORNER,
CD_PROP_INT32,
CD_PROP_INT32,
BuiltinAttributeProvider::Creatable,
Mesh: Replace MLoop struct with generic attributes Implements #102359. Split the `MLoop` struct into two separate integer arrays called `corner_verts` and `corner_edges`, referring to the vertex each corner is attached to and the next edge around the face at each corner. These arrays can be sliced to give access to the edges or vertices in a face. Then they are often referred to as "poly_verts" or "poly_edges". The main benefits are halving the necessary memory bandwidth when only one array is used and simplifications from using regular integer indices instead of a special-purpose struct. The commit also starts a renaming from "loop" to "corner" in mesh code. Like the other mesh struct of array refactors, forward compatibility is kept by writing files with the older format. This will be done until 4.0 to ease the transition process. Looking at a small portion of the patch should give a good impression for the rest of the changes. I tried to make the changes as small as possible so it's easy to tell the correctness from the diff. Though I found Blender developers have been very inventive over the last decade when finding different ways to loop over the corners in a face. For performance, nearly every piece of code that deals with `Mesh` is slightly impacted. Any algorithm that is memory bottle-necked should see an improvement. For example, here is a comparison of interpolating a vertex float attribute to face corners (Ryzen 3700x): **Before** (Average: 3.7 ms, Min: 3.4 ms) ``` threading::parallel_for(loops.index_range(), 4096, [&](IndexRange range) { for (const int64_t i : range) { dst[i] = src[loops[i].v]; } }); ``` **After** (Average: 2.9 ms, Min: 2.6 ms) ``` array_utils::gather(src, corner_verts, dst); ``` That's an improvement of 28% to the average timings, and it's also a simplification, since an index-based routine can be used instead. For more examples using the new arrays, see the design task. Pull Request: https://projects.blender.org/blender/blender/pulls/104424
2023-03-20 15:55:13 +01:00
BuiltinAttributeProvider::NonDeletable,
corner_access,
nullptr,
AttributeValidator{&int_index_clamp});
static BuiltinCustomDataLayerProvider corner_edge(".corner_edge",
ATTR_DOMAIN_CORNER,
CD_PROP_INT32,
CD_PROP_INT32,
BuiltinAttributeProvider::Creatable,
Mesh: Replace MLoop struct with generic attributes Implements #102359. Split the `MLoop` struct into two separate integer arrays called `corner_verts` and `corner_edges`, referring to the vertex each corner is attached to and the next edge around the face at each corner. These arrays can be sliced to give access to the edges or vertices in a face. Then they are often referred to as "poly_verts" or "poly_edges". The main benefits are halving the necessary memory bandwidth when only one array is used and simplifications from using regular integer indices instead of a special-purpose struct. The commit also starts a renaming from "loop" to "corner" in mesh code. Like the other mesh struct of array refactors, forward compatibility is kept by writing files with the older format. This will be done until 4.0 to ease the transition process. Looking at a small portion of the patch should give a good impression for the rest of the changes. I tried to make the changes as small as possible so it's easy to tell the correctness from the diff. Though I found Blender developers have been very inventive over the last decade when finding different ways to loop over the corners in a face. For performance, nearly every piece of code that deals with `Mesh` is slightly impacted. Any algorithm that is memory bottle-necked should see an improvement. For example, here is a comparison of interpolating a vertex float attribute to face corners (Ryzen 3700x): **Before** (Average: 3.7 ms, Min: 3.4 ms) ``` threading::parallel_for(loops.index_range(), 4096, [&](IndexRange range) { for (const int64_t i : range) { dst[i] = src[loops[i].v]; } }); ``` **After** (Average: 2.9 ms, Min: 2.6 ms) ``` array_utils::gather(src, corner_verts, dst); ``` That's an improvement of 28% to the average timings, and it's also a simplification, since an index-based routine can be used instead. For more examples using the new arrays, see the design task. Pull Request: https://projects.blender.org/blender/blender/pulls/104424
2023-03-20 15:55:13 +01:00
BuiltinAttributeProvider::NonDeletable,
corner_access,
nullptr,
AttributeValidator{&int_index_clamp});
Mesh: Move face shade smooth flag to a generic attribute Currently the shade smooth status for mesh faces is stored as part of `MPoly::flag`. As described in #95967, this moves that information to a separate boolean attribute. It also flips its status, so the attribute is now called `sharp_face`, which mirrors the existing `sharp_edge` attribute. The attribute doesn't need to be allocated when all faces are smooth. Forward compatibility is kept until 4.0 like the other mesh refactors. This will reduce memory bandwidth requirements for some operations, since the array of booleans uses 12 times less memory than `MPoly`. It also allows faces to be stored more efficiently in the future, since the flag is now unused. It's also possible to use generic functions to process the values. For example, finding whether there is a sharp face is just `sharp_faces.contains(true)`. The `shade_smooth` attribute is no longer accessible with geometry nodes. Since there were dedicated accessor nodes for that data, that shouldn't be a problem. That's difficult to version automatically since the named attribute nodes could be used in arbitrary combinations. **Implementation notes:** - The attribute and array variables in the code use the `sharp_faces` term, to be consistent with the user-facing "sharp faces" wording, and to avoid requiring many renames when #101689 is implemented. - Cycles now accesses smooth face status with the generic attribute, to avoid overhead. - Changing the zero-value from "smooth" to "flat" takes some care to make sure defaults are the same. - Versioning for the edge mode extrude node is particularly complex. New nodes are added by versioning to propagate the attribute in its old inverted state. - A lot of access is still done through the `CustomData` API rather than the attribute API because of a few functions. That can be cleaned up easily in the future. - In the future we would benefit from a way to store attributes as a single value for when all faces are sharp. Pull Request: https://projects.blender.org/blender/blender/pulls/104422
2023-03-08 15:36:18 +01:00
static BuiltinCustomDataLayerProvider sharp_face("sharp_face",
ATTR_DOMAIN_FACE,
CD_PROP_BOOL,
CD_PROP_BOOL,
BuiltinAttributeProvider::Creatable,
BuiltinAttributeProvider::Deletable,
face_access,
nullptr);
static BuiltinCustomDataLayerProvider sharp_edge("sharp_edge",
ATTR_DOMAIN_EDGE,
CD_PROP_BOOL,
CD_PROP_BOOL,
BuiltinAttributeProvider::Creatable,
BuiltinAttributeProvider::Deletable,
edge_access,
nullptr);
static const auto crease_clamp = mf::build::SI1_SO<float, float>(
"Crease Clamp",
[](float value) { return std::clamp(value, 0.0f, 1.0f); },
mf::build::exec_presets::AllSpanOrSingle());
static BuiltinCustomDataLayerProvider crease("crease",
ATTR_DOMAIN_EDGE,
CD_PROP_FLOAT,
CD_CREASE,
BuiltinAttributeProvider::Creatable,
BuiltinAttributeProvider::Deletable,
edge_access,
nullptr,
AttributeValidator{&crease_clamp});
static VertexGroupsAttributeProvider vertex_groups;
static CustomDataAttributeProvider corner_custom_data(ATTR_DOMAIN_CORNER, corner_access);
static CustomDataAttributeProvider point_custom_data(ATTR_DOMAIN_POINT, point_access);
static CustomDataAttributeProvider edge_custom_data(ATTR_DOMAIN_EDGE, edge_access);
static CustomDataAttributeProvider face_custom_data(ATTR_DOMAIN_FACE, face_access);
Mesh: Replace MLoop struct with generic attributes Implements #102359. Split the `MLoop` struct into two separate integer arrays called `corner_verts` and `corner_edges`, referring to the vertex each corner is attached to and the next edge around the face at each corner. These arrays can be sliced to give access to the edges or vertices in a face. Then they are often referred to as "poly_verts" or "poly_edges". The main benefits are halving the necessary memory bandwidth when only one array is used and simplifications from using regular integer indices instead of a special-purpose struct. The commit also starts a renaming from "loop" to "corner" in mesh code. Like the other mesh struct of array refactors, forward compatibility is kept by writing files with the older format. This will be done until 4.0 to ease the transition process. Looking at a small portion of the patch should give a good impression for the rest of the changes. I tried to make the changes as small as possible so it's easy to tell the correctness from the diff. Though I found Blender developers have been very inventive over the last decade when finding different ways to loop over the corners in a face. For performance, nearly every piece of code that deals with `Mesh` is slightly impacted. Any algorithm that is memory bottle-necked should see an improvement. For example, here is a comparison of interpolating a vertex float attribute to face corners (Ryzen 3700x): **Before** (Average: 3.7 ms, Min: 3.4 ms) ``` threading::parallel_for(loops.index_range(), 4096, [&](IndexRange range) { for (const int64_t i : range) { dst[i] = src[loops[i].v]; } }); ``` **After** (Average: 2.9 ms, Min: 2.6 ms) ``` array_utils::gather(src, corner_verts, dst); ``` That's an improvement of 28% to the average timings, and it's also a simplification, since an index-based routine can be used instead. For more examples using the new arrays, see the design task. Pull Request: https://projects.blender.org/blender/blender/pulls/104424
2023-03-20 15:55:13 +01:00
return ComponentAttributeProviders({&position,
Mesh: Move edges to a generic attribute Implements #95966, as the final step of #95965. This commit changes the storage of mesh edge vertex indices from the `MEdge` type to the generic `int2` attribute type. This follows the general design for geometry and the attribute system, where the data storage type and the usage semantics are separated. The main benefit of the change is reduced memory usage-- the requirements of storing mesh edges is reduced by 1/3. For example, this saves 8MB on a 1 million vertex grid. This also gives performance benefits to any memory-bound mesh processing algorithm that uses edges. Another benefit is that all of the edge's vertex indices are contiguous. In a few cases, it's helpful to process all of them as `Span<int>` rather than `Span<int2>`. Similarly, the type is more likely to match a generic format used by a library, or code that shouldn't know about specific Blender `Mesh` types. Various Notes: - The `.edge_verts` name is used to reflect a mapping between domains, similar to `.corner_verts`, etc. The period means that it the data shouldn't change arbitrarily by the user or procedural operations. - `edge[0]` is now used instead of `edge.v1` - Signed integers are used instead of unsigned to reduce the mixing of signed-ness, which can be error prone. - All of the previously used core mesh data types (`MVert`, `MEdge`, `MLoop`, `MPoly` are now deprecated. Only generic types are used). - The `vec2i` DNA type is used in the few C files where necessary. Pull Request: https://projects.blender.org/blender/blender/pulls/106638
2023-04-17 13:47:41 +02:00
&edge_verts,
Mesh: Replace MLoop struct with generic attributes Implements #102359. Split the `MLoop` struct into two separate integer arrays called `corner_verts` and `corner_edges`, referring to the vertex each corner is attached to and the next edge around the face at each corner. These arrays can be sliced to give access to the edges or vertices in a face. Then they are often referred to as "poly_verts" or "poly_edges". The main benefits are halving the necessary memory bandwidth when only one array is used and simplifications from using regular integer indices instead of a special-purpose struct. The commit also starts a renaming from "loop" to "corner" in mesh code. Like the other mesh struct of array refactors, forward compatibility is kept by writing files with the older format. This will be done until 4.0 to ease the transition process. Looking at a small portion of the patch should give a good impression for the rest of the changes. I tried to make the changes as small as possible so it's easy to tell the correctness from the diff. Though I found Blender developers have been very inventive over the last decade when finding different ways to loop over the corners in a face. For performance, nearly every piece of code that deals with `Mesh` is slightly impacted. Any algorithm that is memory bottle-necked should see an improvement. For example, here is a comparison of interpolating a vertex float attribute to face corners (Ryzen 3700x): **Before** (Average: 3.7 ms, Min: 3.4 ms) ``` threading::parallel_for(loops.index_range(), 4096, [&](IndexRange range) { for (const int64_t i : range) { dst[i] = src[loops[i].v]; } }); ``` **After** (Average: 2.9 ms, Min: 2.6 ms) ``` array_utils::gather(src, corner_verts, dst); ``` That's an improvement of 28% to the average timings, and it's also a simplification, since an index-based routine can be used instead. For more examples using the new arrays, see the design task. Pull Request: https://projects.blender.org/blender/blender/pulls/104424
2023-03-20 15:55:13 +01:00
&corner_vert,
&corner_edge,
&id,
&material_index,
&sharp_face,
&sharp_edge,
&crease},
{&corner_custom_data,
&vertex_groups,
&point_custom_data,
&edge_custom_data,
&face_custom_data});
}
Geometry Nodes: new geometry attribute API Currently, there are two attribute API. The first, defined in `BKE_attribute.h` is accessible from RNA and C code. The second is implemented with `GeometryComponent` and is only accessible in C++ code. The second is widely used, but only being accessible through the `GeometrySet` API makes it awkward to use, and even impossible for types that don't correspond directly to a geometry component like `CurvesGeometry`. This patch adds a new attribute API, designed to replace the `GeometryComponent` attribute API now, and to eventually replace or be the basis of the other one. The basic idea is that there is an `AttributeAccessor` class that allows code to interact with a set of attributes owned by some geometry. The accessor itself has no ownership. `AttributeAccessor` is a simple type that can be passed around by value. That makes it easy to return it from functions and to store it in containers. For const-correctness, there is also a `MutableAttributeAccessor` that allows changing individual and can add or remove attributes. Currently, `AttributeAccessor` is composed of two pointers. The first is a pointer to the owner of the attribute data. The second is a pointer to a struct with function pointers, that is similar to a virtual function table. The functions know how to access attributes on the owner. The actual attribute access for geometries is still implemented with the `AttributeProvider` pattern, which makes it easy to support different sources of attributes on a geometry and simplifies dealing with built-in attributes. There are different ways to get an attribute accessor for a geometry: * `GeometryComponent.attributes()` * `CurvesGeometry.attributes()` * `bke::mesh_attributes(const Mesh &)` * `bke::pointcloud_attributes(const PointCloud &)` All of these also have a `_for_write` variant that returns a `MutabelAttributeAccessor`. Differential Revision: https://developer.blender.org/D15280
2022-07-08 16:16:56 +02:00
static AttributeAccessorFunctions get_mesh_accessor_functions()
{
static const ComponentAttributeProviders providers = create_attribute_providers_for_mesh();
AttributeAccessorFunctions fn =
attribute_accessor_functions::accessor_functions_for_providers<providers>();
fn.domain_size = [](const void *owner, const eAttrDomain domain) {
if (owner == nullptr) {
return 0;
}
const Mesh &mesh = *static_cast<const Mesh *>(owner);
switch (domain) {
case ATTR_DOMAIN_POINT:
return mesh.totvert;
case ATTR_DOMAIN_EDGE:
return mesh.totedge;
case ATTR_DOMAIN_FACE:
return mesh.totpoly;
case ATTR_DOMAIN_CORNER:
return mesh.totloop;
default:
return 0;
}
};
fn.domain_supported = [](const void * /*owner*/, const eAttrDomain domain) {
Geometry Nodes: new geometry attribute API Currently, there are two attribute API. The first, defined in `BKE_attribute.h` is accessible from RNA and C code. The second is implemented with `GeometryComponent` and is only accessible in C++ code. The second is widely used, but only being accessible through the `GeometrySet` API makes it awkward to use, and even impossible for types that don't correspond directly to a geometry component like `CurvesGeometry`. This patch adds a new attribute API, designed to replace the `GeometryComponent` attribute API now, and to eventually replace or be the basis of the other one. The basic idea is that there is an `AttributeAccessor` class that allows code to interact with a set of attributes owned by some geometry. The accessor itself has no ownership. `AttributeAccessor` is a simple type that can be passed around by value. That makes it easy to return it from functions and to store it in containers. For const-correctness, there is also a `MutableAttributeAccessor` that allows changing individual and can add or remove attributes. Currently, `AttributeAccessor` is composed of two pointers. The first is a pointer to the owner of the attribute data. The second is a pointer to a struct with function pointers, that is similar to a virtual function table. The functions know how to access attributes on the owner. The actual attribute access for geometries is still implemented with the `AttributeProvider` pattern, which makes it easy to support different sources of attributes on a geometry and simplifies dealing with built-in attributes. There are different ways to get an attribute accessor for a geometry: * `GeometryComponent.attributes()` * `CurvesGeometry.attributes()` * `bke::mesh_attributes(const Mesh &)` * `bke::pointcloud_attributes(const PointCloud &)` All of these also have a `_for_write` variant that returns a `MutabelAttributeAccessor`. Differential Revision: https://developer.blender.org/D15280
2022-07-08 16:16:56 +02:00
return ELEM(domain, ATTR_DOMAIN_POINT, ATTR_DOMAIN_EDGE, ATTR_DOMAIN_FACE, ATTR_DOMAIN_CORNER);
};
fn.adapt_domain = [](const void *owner,
const blender::GVArray &varray,
const eAttrDomain from_domain,
const eAttrDomain to_domain) -> blender::GVArray {
if (owner == nullptr) {
return {};
}
const Mesh &mesh = *static_cast<const Mesh *>(owner);
return adapt_mesh_attribute_domain(mesh, varray, from_domain, to_domain);
};
return fn;
}
static const AttributeAccessorFunctions &get_mesh_accessor_functions_ref()
{
static const AttributeAccessorFunctions fn = get_mesh_accessor_functions();
return fn;
}
} // namespace blender::bke
blender::bke::AttributeAccessor Mesh::attributes() const
Geometry Nodes: new geometry attribute API Currently, there are two attribute API. The first, defined in `BKE_attribute.h` is accessible from RNA and C code. The second is implemented with `GeometryComponent` and is only accessible in C++ code. The second is widely used, but only being accessible through the `GeometrySet` API makes it awkward to use, and even impossible for types that don't correspond directly to a geometry component like `CurvesGeometry`. This patch adds a new attribute API, designed to replace the `GeometryComponent` attribute API now, and to eventually replace or be the basis of the other one. The basic idea is that there is an `AttributeAccessor` class that allows code to interact with a set of attributes owned by some geometry. The accessor itself has no ownership. `AttributeAccessor` is a simple type that can be passed around by value. That makes it easy to return it from functions and to store it in containers. For const-correctness, there is also a `MutableAttributeAccessor` that allows changing individual and can add or remove attributes. Currently, `AttributeAccessor` is composed of two pointers. The first is a pointer to the owner of the attribute data. The second is a pointer to a struct with function pointers, that is similar to a virtual function table. The functions know how to access attributes on the owner. The actual attribute access for geometries is still implemented with the `AttributeProvider` pattern, which makes it easy to support different sources of attributes on a geometry and simplifies dealing with built-in attributes. There are different ways to get an attribute accessor for a geometry: * `GeometryComponent.attributes()` * `CurvesGeometry.attributes()` * `bke::mesh_attributes(const Mesh &)` * `bke::pointcloud_attributes(const PointCloud &)` All of these also have a `_for_write` variant that returns a `MutabelAttributeAccessor`. Differential Revision: https://developer.blender.org/D15280
2022-07-08 16:16:56 +02:00
{
return blender::bke::AttributeAccessor(this, blender::bke::get_mesh_accessor_functions_ref());
Geometry Nodes: new geometry attribute API Currently, there are two attribute API. The first, defined in `BKE_attribute.h` is accessible from RNA and C code. The second is implemented with `GeometryComponent` and is only accessible in C++ code. The second is widely used, but only being accessible through the `GeometrySet` API makes it awkward to use, and even impossible for types that don't correspond directly to a geometry component like `CurvesGeometry`. This patch adds a new attribute API, designed to replace the `GeometryComponent` attribute API now, and to eventually replace or be the basis of the other one. The basic idea is that there is an `AttributeAccessor` class that allows code to interact with a set of attributes owned by some geometry. The accessor itself has no ownership. `AttributeAccessor` is a simple type that can be passed around by value. That makes it easy to return it from functions and to store it in containers. For const-correctness, there is also a `MutableAttributeAccessor` that allows changing individual and can add or remove attributes. Currently, `AttributeAccessor` is composed of two pointers. The first is a pointer to the owner of the attribute data. The second is a pointer to a struct with function pointers, that is similar to a virtual function table. The functions know how to access attributes on the owner. The actual attribute access for geometries is still implemented with the `AttributeProvider` pattern, which makes it easy to support different sources of attributes on a geometry and simplifies dealing with built-in attributes. There are different ways to get an attribute accessor for a geometry: * `GeometryComponent.attributes()` * `CurvesGeometry.attributes()` * `bke::mesh_attributes(const Mesh &)` * `bke::pointcloud_attributes(const PointCloud &)` All of these also have a `_for_write` variant that returns a `MutabelAttributeAccessor`. Differential Revision: https://developer.blender.org/D15280
2022-07-08 16:16:56 +02:00
}
blender::bke::MutableAttributeAccessor Mesh::attributes_for_write()
Geometry Nodes: new geometry attribute API Currently, there are two attribute API. The first, defined in `BKE_attribute.h` is accessible from RNA and C code. The second is implemented with `GeometryComponent` and is only accessible in C++ code. The second is widely used, but only being accessible through the `GeometrySet` API makes it awkward to use, and even impossible for types that don't correspond directly to a geometry component like `CurvesGeometry`. This patch adds a new attribute API, designed to replace the `GeometryComponent` attribute API now, and to eventually replace or be the basis of the other one. The basic idea is that there is an `AttributeAccessor` class that allows code to interact with a set of attributes owned by some geometry. The accessor itself has no ownership. `AttributeAccessor` is a simple type that can be passed around by value. That makes it easy to return it from functions and to store it in containers. For const-correctness, there is also a `MutableAttributeAccessor` that allows changing individual and can add or remove attributes. Currently, `AttributeAccessor` is composed of two pointers. The first is a pointer to the owner of the attribute data. The second is a pointer to a struct with function pointers, that is similar to a virtual function table. The functions know how to access attributes on the owner. The actual attribute access for geometries is still implemented with the `AttributeProvider` pattern, which makes it easy to support different sources of attributes on a geometry and simplifies dealing with built-in attributes. There are different ways to get an attribute accessor for a geometry: * `GeometryComponent.attributes()` * `CurvesGeometry.attributes()` * `bke::mesh_attributes(const Mesh &)` * `bke::pointcloud_attributes(const PointCloud &)` All of these also have a `_for_write` variant that returns a `MutabelAttributeAccessor`. Differential Revision: https://developer.blender.org/D15280
2022-07-08 16:16:56 +02:00
{
return blender::bke::MutableAttributeAccessor(this,
blender::bke::get_mesh_accessor_functions_ref());
Geometry Nodes: new geometry attribute API Currently, there are two attribute API. The first, defined in `BKE_attribute.h` is accessible from RNA and C code. The second is implemented with `GeometryComponent` and is only accessible in C++ code. The second is widely used, but only being accessible through the `GeometrySet` API makes it awkward to use, and even impossible for types that don't correspond directly to a geometry component like `CurvesGeometry`. This patch adds a new attribute API, designed to replace the `GeometryComponent` attribute API now, and to eventually replace or be the basis of the other one. The basic idea is that there is an `AttributeAccessor` class that allows code to interact with a set of attributes owned by some geometry. The accessor itself has no ownership. `AttributeAccessor` is a simple type that can be passed around by value. That makes it easy to return it from functions and to store it in containers. For const-correctness, there is also a `MutableAttributeAccessor` that allows changing individual and can add or remove attributes. Currently, `AttributeAccessor` is composed of two pointers. The first is a pointer to the owner of the attribute data. The second is a pointer to a struct with function pointers, that is similar to a virtual function table. The functions know how to access attributes on the owner. The actual attribute access for geometries is still implemented with the `AttributeProvider` pattern, which makes it easy to support different sources of attributes on a geometry and simplifies dealing with built-in attributes. There are different ways to get an attribute accessor for a geometry: * `GeometryComponent.attributes()` * `CurvesGeometry.attributes()` * `bke::mesh_attributes(const Mesh &)` * `bke::pointcloud_attributes(const PointCloud &)` All of these also have a `_for_write` variant that returns a `MutabelAttributeAccessor`. Differential Revision: https://developer.blender.org/D15280
2022-07-08 16:16:56 +02:00
}
std::optional<blender::bke::AttributeAccessor> MeshComponent::attributes() const
{
return blender::bke::AttributeAccessor(mesh_, blender::bke::get_mesh_accessor_functions_ref());
}
std::optional<blender::bke::MutableAttributeAccessor> MeshComponent::attributes_for_write()
{
Mesh *mesh = this->get_for_write();
return blender::bke::MutableAttributeAccessor(mesh,
Geometry Nodes: new geometry attribute API Currently, there are two attribute API. The first, defined in `BKE_attribute.h` is accessible from RNA and C code. The second is implemented with `GeometryComponent` and is only accessible in C++ code. The second is widely used, but only being accessible through the `GeometrySet` API makes it awkward to use, and even impossible for types that don't correspond directly to a geometry component like `CurvesGeometry`. This patch adds a new attribute API, designed to replace the `GeometryComponent` attribute API now, and to eventually replace or be the basis of the other one. The basic idea is that there is an `AttributeAccessor` class that allows code to interact with a set of attributes owned by some geometry. The accessor itself has no ownership. `AttributeAccessor` is a simple type that can be passed around by value. That makes it easy to return it from functions and to store it in containers. For const-correctness, there is also a `MutableAttributeAccessor` that allows changing individual and can add or remove attributes. Currently, `AttributeAccessor` is composed of two pointers. The first is a pointer to the owner of the attribute data. The second is a pointer to a struct with function pointers, that is similar to a virtual function table. The functions know how to access attributes on the owner. The actual attribute access for geometries is still implemented with the `AttributeProvider` pattern, which makes it easy to support different sources of attributes on a geometry and simplifies dealing with built-in attributes. There are different ways to get an attribute accessor for a geometry: * `GeometryComponent.attributes()` * `CurvesGeometry.attributes()` * `bke::mesh_attributes(const Mesh &)` * `bke::pointcloud_attributes(const PointCloud &)` All of these also have a `_for_write` variant that returns a `MutabelAttributeAccessor`. Differential Revision: https://developer.blender.org/D15280
2022-07-08 16:16:56 +02:00
blender::bke::get_mesh_accessor_functions_ref());
}
/** \} */