tornavis/source/blender/blenkernel/intern/mesh_iterators.cc

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

390 lines
13 KiB
C++
Raw Normal View History

/* SPDX-FileCopyrightText: 2023 Blender Authors
*
* SPDX-License-Identifier: GPL-2.0-or-later */
/** \file
* \ingroup bke
*
* Functions for iterating mesh features.
*/
#include "DNA_mesh_types.h"
#include "DNA_meshdata_types.h"
#include "BKE_customdata.hh"
#include "BKE_editmesh.hh"
#include "BKE_editmesh_cache.hh"
#include "BKE_mesh.hh"
#include "BKE_mesh_iterators.hh"
#include "BLI_bitmap.h"
#include "BLI_math_vector.h"
#include "MEM_guardedalloc.h"
/* General note on iterating verts/loops/edges/faces and end mode.
*
* The edit mesh pointer is set for both final and cage meshes in both cases when there are
* modifiers applied and not. This helps consistency of checks in the draw manager, where the
* existence of the edit mesh pointer does not depend on object configuration.
*
* For the iterating, however, we need to follow the `CD_ORIGINDEX` code paths when there are
* modifiers applied on the cage. In the code terms it means that the check for the edit mode code
* path needs to consist of both edit mesh and edit data checks. */
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 21:37:58 +01:00
void BKE_mesh_foreach_mapped_vert(
const Mesh *mesh,
void (*func)(void *user_data, int index, const float co[3], const float no[3]),
void *user_data,
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 21:37:58 +01:00
MeshForeachFlag flag)
{
if (mesh->edit_mesh != nullptr && mesh->runtime->edit_data != nullptr) {
BMEditMesh *em = mesh->edit_mesh;
BMesh *bm = em->bm;
BMIter iter;
BMVert *eve;
int i;
if (!mesh->runtime->edit_data->vertexCos.is_empty()) {
const blender::Span<blender::float3> positions = mesh->runtime->edit_data->vertexCos;
blender::Span<blender::float3> vert_normals;
if (flag & MESH_FOREACH_USE_NORMAL) {
BKE_editmesh_cache_ensure_vert_normals(em, mesh->runtime->edit_data);
vert_normals = mesh->runtime->edit_data->vertexNos;
}
BM_ITER_MESH_INDEX (eve, &iter, bm, BM_VERTS_OF_MESH, i) {
const float *no = (flag & MESH_FOREACH_USE_NORMAL) ? &vert_normals[i].x : nullptr;
func(user_data, i, positions[i], no);
}
}
else {
BM_ITER_MESH_INDEX (eve, &iter, bm, BM_VERTS_OF_MESH, i) {
const float *no = (flag & MESH_FOREACH_USE_NORMAL) ? eve->no : nullptr;
func(user_data, i, eve->co, no);
}
}
}
else {
const blender::Span<blender::float3> positions = mesh->vert_positions();
const int *index = static_cast<const int *>(
CustomData_get_layer(&mesh->vert_data, CD_ORIGINDEX));
blender::Span<blender::float3> vert_normals;
if (flag & MESH_FOREACH_USE_NORMAL) {
vert_normals = mesh->vert_normals();
}
if (index) {
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 06:10:43 +01:00
for (int i = 0; i < mesh->totvert; i++) {
const float *no = (flag & MESH_FOREACH_USE_NORMAL) ? &vert_normals[i].x : nullptr;
const int orig = *index++;
if (orig == ORIGINDEX_NONE) {
continue;
}
func(user_data, orig, positions[i], no);
}
}
else {
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 06:10:43 +01:00
for (int i = 0; i < mesh->totvert; i++) {
const float *no = (flag & MESH_FOREACH_USE_NORMAL) ? &vert_normals[i].x : nullptr;
func(user_data, i, positions[i], no);
}
}
}
}
void BKE_mesh_foreach_mapped_edge(
Mesh *mesh,
const int tot_edges,
void (*func)(void *user_data, int index, const float v0co[3], const float v1co[3]),
void *user_data)
{
if (mesh->edit_mesh != nullptr && mesh->runtime->edit_data) {
BMEditMesh *em = mesh->edit_mesh;
BMesh *bm = em->bm;
BMIter iter;
BMEdge *eed;
int i;
if (!mesh->runtime->edit_data->vertexCos.is_empty()) {
const blender::Span<blender::float3> positions = mesh->runtime->edit_data->vertexCos;
BM_mesh_elem_index_ensure(bm, BM_VERT);
BM_ITER_MESH_INDEX (eed, &iter, bm, BM_EDGES_OF_MESH, i) {
func(user_data,
i,
positions[BM_elem_index_get(eed->v1)],
positions[BM_elem_index_get(eed->v2)]);
}
}
else {
BM_ITER_MESH_INDEX (eed, &iter, bm, BM_EDGES_OF_MESH, i) {
func(user_data, i, eed->v1->co, eed->v2->co);
}
}
}
else {
const blender::Span<blender::float3> positions = mesh->vert_positions();
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 blender::Span<blender::int2> edges = mesh->edges();
const int *index = static_cast<const int *>(
CustomData_get_layer(&mesh->edge_data, CD_ORIGINDEX));
if (index) {
for (const int i : edges.index_range()) {
const int orig = *index++;
if (orig == ORIGINDEX_NONE) {
continue;
}
func(user_data, orig, positions[edges[i][0]], positions[edges[i][1]]);
}
}
else if (mesh->totedge == tot_edges) {
for (const int i : edges.index_range()) {
func(user_data, i, positions[edges[i][0]], positions[edges[i][1]]);
}
}
}
}
void BKE_mesh_foreach_mapped_loop(Mesh *mesh,
void (*func)(void *user_data,
int vertex_index,
int face_index,
const float co[3],
const float no[3]),
void *user_data,
MeshForeachFlag flag)
{
/* We can't use `dm->getLoopDataLayout(dm)` here,
* we want to always access `dm->loopData`, `EditDerivedBMesh` would
* return loop data from BMesh itself. */
if (mesh->edit_mesh != nullptr && mesh->runtime->edit_data) {
BMEditMesh *em = mesh->edit_mesh;
BMesh *bm = em->bm;
BMIter iter;
BMFace *efa;
const blender::Span<blender::float3> positions = mesh->runtime->edit_data->vertexCos;
/* XXX: investigate using EditMesh data. */
blender::Span<blender::float3> corner_normals;
if (flag & MESH_FOREACH_USE_NORMAL) {
Mesh: Replace auto smooth with node group Design task: #93551 This PR replaces the auto smooth option with a geometry nodes modifier that sets the sharp edge attribute. This solves a fair number of long- standing problems related to auto smooth, simplifies the process of normal computation, and allows Blender to automatically choose between face, vertex, and face corner normals based on the sharp edge and face attributes. Versioning adds a geometry node group to objects with meshes that had auto-smooth enabled. The modifier can be applied, which also improves performance. Auto smooth is now unnecessary to get a combination of sharp and smooth edges. In general workflows are changed a bit. Separate procedural and destructive workflows are available. Custom normals can be used immediately without turning on the removed auto smooth option. **Procedural** The node group asset "Smooth by Angle" is the main way to set sharp normals based on the edge angle. It can be accessed directly in the add modifier menu. Of course the modifier can be reordered, muted, or applied like any other, or changed internally like any geometry nodes modifier. **Destructive** Often the sharp edges don't need to be dynamic. This can give better performance since edge angles don't need to be recalculated. In edit mode the two operators "Select Sharp Edges" and "Mark Sharp" can be used. In other modes, the "Shade Smooth by Angle" controls the edge sharpness directly. ### Breaking API Changes - `use_auto_smooth` is removed. Face corner normals are now used automatically if there are mixed smooth vs. not smooth tags. Meshes now always use custom normals if they exist. - In Cycles, the lack of the separate auto smooth state makes normals look triangulated when all faces are shaded smooth. - `auto_smooth_angle` is removed. Replaced by a modifier (or operator) controlling the sharp edge attribute. This means the mesh itself (without an object) doesn't know anything about automatically smoothing by angle anymore. - `create_normals_split`, `calc_normals_split`, and `free_normals_split` are removed, and are replaced by the simpler `Mesh.corner_normals` collection property. Since it gives access to the normals cache, it is automatically updated when relevant data changes. Addons are updated here: https://projects.blender.org/blender/blender-addons/pulls/104609 ### Tests - `geo_node_curves_test_deform_curves_on_surface` has slightly different results because face corner normals are used instead of interpolated vertex normals. - `bf_wavefront_obj_tests` has different export results for one file which mixed sharp and smooth faces without turning on auto smooth. - `cycles_mesh_cpu` has one object which is completely flat shaded. Previously every edge was split before rendering, now it looks triangulated. Pull Request: https://projects.blender.org/blender/blender/pulls/108014
2023-10-20 16:54:08 +02:00
corner_normals = mesh->corner_normals();
}
int f_idx;
BM_mesh_elem_index_ensure(bm, BM_VERT);
BM_ITER_MESH_INDEX (efa, &iter, bm, BM_FACES_OF_MESH, f_idx) {
BMLoop *l_iter, *l_first;
l_iter = l_first = BM_FACE_FIRST_LOOP(efa);
do {
const BMVert *eve = l_iter->v;
const int v_idx = BM_elem_index_get(eve);
func(user_data,
v_idx,
f_idx,
positions.is_empty() ? positions[v_idx] : blender::float3(eve->co),
corner_normals.is_empty() ? nullptr : &corner_normals[BM_elem_index_get(l_iter)].x);
} while ((l_iter = l_iter->next) != l_first);
}
}
else {
blender::Span<blender::float3> corner_normals;
if (flag & MESH_FOREACH_USE_NORMAL) {
Mesh: Replace auto smooth with node group Design task: #93551 This PR replaces the auto smooth option with a geometry nodes modifier that sets the sharp edge attribute. This solves a fair number of long- standing problems related to auto smooth, simplifies the process of normal computation, and allows Blender to automatically choose between face, vertex, and face corner normals based on the sharp edge and face attributes. Versioning adds a geometry node group to objects with meshes that had auto-smooth enabled. The modifier can be applied, which also improves performance. Auto smooth is now unnecessary to get a combination of sharp and smooth edges. In general workflows are changed a bit. Separate procedural and destructive workflows are available. Custom normals can be used immediately without turning on the removed auto smooth option. **Procedural** The node group asset "Smooth by Angle" is the main way to set sharp normals based on the edge angle. It can be accessed directly in the add modifier menu. Of course the modifier can be reordered, muted, or applied like any other, or changed internally like any geometry nodes modifier. **Destructive** Often the sharp edges don't need to be dynamic. This can give better performance since edge angles don't need to be recalculated. In edit mode the two operators "Select Sharp Edges" and "Mark Sharp" can be used. In other modes, the "Shade Smooth by Angle" controls the edge sharpness directly. ### Breaking API Changes - `use_auto_smooth` is removed. Face corner normals are now used automatically if there are mixed smooth vs. not smooth tags. Meshes now always use custom normals if they exist. - In Cycles, the lack of the separate auto smooth state makes normals look triangulated when all faces are shaded smooth. - `auto_smooth_angle` is removed. Replaced by a modifier (or operator) controlling the sharp edge attribute. This means the mesh itself (without an object) doesn't know anything about automatically smoothing by angle anymore. - `create_normals_split`, `calc_normals_split`, and `free_normals_split` are removed, and are replaced by the simpler `Mesh.corner_normals` collection property. Since it gives access to the normals cache, it is automatically updated when relevant data changes. Addons are updated here: https://projects.blender.org/blender/blender-addons/pulls/104609 ### Tests - `geo_node_curves_test_deform_curves_on_surface` has slightly different results because face corner normals are used instead of interpolated vertex normals. - `bf_wavefront_obj_tests` has different export results for one file which mixed sharp and smooth faces without turning on auto smooth. - `cycles_mesh_cpu` has one object which is completely flat shaded. Previously every edge was split before rendering, now it looks triangulated. Pull Request: https://projects.blender.org/blender/blender/pulls/108014
2023-10-20 16:54:08 +02:00
corner_normals = mesh->corner_normals();
}
const blender::Span<blender::float3> positions = mesh->vert_positions();
const blender::OffsetIndices faces = mesh->faces();
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 blender::Span<int> corner_verts = mesh->corner_verts();
const int *v_index = static_cast<const int *>(
CustomData_get_layer(&mesh->vert_data, CD_ORIGINDEX));
const int *f_index = static_cast<const int *>(
CustomData_get_layer(&mesh->face_data, CD_ORIGINDEX));
if (v_index || f_index) {
for (const int face_i : faces.index_range()) {
for (const int corner : faces[face_i]) {
const int vert = 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
const int v_idx = v_index ? v_index[vert] : vert;
const int f_idx = f_index ? f_index[face_i] : face_i;
const float *no = corner_normals.is_empty() ? nullptr : &corner_normals[corner].x;
if (ELEM(ORIGINDEX_NONE, v_idx, f_idx)) {
continue;
}
func(user_data, v_idx, f_idx, positions[vert], no);
}
}
}
else {
for (const int face_i : faces.index_range()) {
for (const int corner : faces[face_i]) {
const int vert = corner_verts[corner];
const float *no = corner_normals.is_empty() ? nullptr : &corner_normals[corner].x;
func(user_data, vert, face_i, positions[vert], no);
}
}
}
}
}
void BKE_mesh_foreach_mapped_face_center(
Mesh *mesh,
void (*func)(void *user_data, int index, const float cent[3], const float no[3]),
void *user_data,
MeshForeachFlag flag)
{
using namespace blender;
if (mesh->edit_mesh != nullptr && mesh->runtime->edit_data != nullptr) {
BMEditMesh *em = mesh->edit_mesh;
BMesh *bm = em->bm;
blender::Span<blender::float3> face_centers;
blender::Span<blender::float3> face_normals;
BMFace *efa;
BMIter iter;
int i;
BKE_editmesh_cache_ensure_face_centers(em, mesh->runtime->edit_data);
face_centers = mesh->runtime->edit_data->faceCos; /* always set */
if (flag & MESH_FOREACH_USE_NORMAL) {
BKE_editmesh_cache_ensure_face_normals(em, mesh->runtime->edit_data);
face_normals = mesh->runtime->edit_data->faceNos; /* maybe nullptr */
}
if (!face_normals.is_empty()) {
BM_ITER_MESH_INDEX (efa, &iter, bm, BM_FACES_OF_MESH, i) {
const float *no = face_normals[i];
func(user_data, i, face_centers[i], no);
}
}
else {
BM_ITER_MESH_INDEX (efa, &iter, bm, BM_FACES_OF_MESH, i) {
const float *no = (flag & MESH_FOREACH_USE_NORMAL) ? efa->no : nullptr;
func(user_data, i, face_centers[i], no);
}
}
}
else {
const blender::Span<float3> positions = mesh->vert_positions();
const blender::OffsetIndices faces = mesh->faces();
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 blender::Span<int> corner_verts = mesh->corner_verts();
const int *index = static_cast<const int *>(
CustomData_get_layer(&mesh->face_data, CD_ORIGINDEX));
if (index) {
for (const int i : faces.index_range()) {
const int orig = *index++;
if (orig == ORIGINDEX_NONE) {
continue;
}
const Span<int> face_verts = corner_verts.slice(faces[i]);
const float3 center = bke::mesh::face_center_calc(positions, face_verts);
if (flag & MESH_FOREACH_USE_NORMAL) {
const float3 normal = bke::mesh::face_normal_calc(positions, face_verts);
func(user_data, orig, center, normal);
}
else {
func(user_data, orig, center, nullptr);
}
}
}
else {
for (const int i : faces.index_range()) {
const Span<int> face_verts = corner_verts.slice(faces[i]);
const float3 center = bke::mesh::face_center_calc(positions, face_verts);
if (flag & MESH_FOREACH_USE_NORMAL) {
const float3 normal = bke::mesh::face_normal_calc(positions, face_verts);
func(user_data, i, center, normal);
}
else {
func(user_data, i, center, nullptr);
}
}
}
}
}
void BKE_mesh_foreach_mapped_subdiv_face_center(
Mesh *mesh,
void (*func)(void *user_data, int index, const float cent[3], const float no[3]),
void *user_data,
MeshForeachFlag flag)
{
const blender::Span<blender::float3> positions = mesh->vert_positions();
const blender::OffsetIndices faces = mesh->faces();
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 blender::Span<int> corner_verts = mesh->corner_verts();
blender::Span<blender::float3> vert_normals;
if (flag & MESH_FOREACH_USE_NORMAL) {
vert_normals = mesh->vert_normals();
}
const int *index = static_cast<const int *>(
CustomData_get_layer(&mesh->face_data, CD_ORIGINDEX));
const blender::BitSpan facedot_tags = mesh->runtime->subsurf_face_dot_tags;
if (index) {
for (const int i : faces.index_range()) {
const int orig = *index++;
if (orig == ORIGINDEX_NONE) {
continue;
}
for (const int vert : corner_verts.slice(faces[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
if (facedot_tags[vert]) {
func(user_data,
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 21:37:58 +01:00
orig,
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
positions[vert],
(flag & MESH_FOREACH_USE_NORMAL) ? &vert_normals[vert].x : nullptr);
}
}
}
}
else {
for (const int i : faces.index_range()) {
for (const int vert : corner_verts.slice(faces[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
if (facedot_tags[vert]) {
func(user_data,
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
positions[vert],
(flag & MESH_FOREACH_USE_NORMAL) ? &vert_normals[vert].x : nullptr);
}
}
}
}
}
/* Helpers based on above foreach loopers> */
struct MappedVCosData {
float (*vertexcos)[3];
BLI_bitmap *vertex_visit;
};
static void get_vertexcos__mapFunc(void *user_data,
int index,
const float co[3],
const float /*no*/[3])
{
MappedVCosData *mapped_vcos_data = (MappedVCosData *)user_data;
if (BLI_BITMAP_TEST(mapped_vcos_data->vertex_visit, index) == 0) {
/* We need coord from prototype vertex, not from copies,
* we assume they stored in the beginning of vertex array stored in evaluated mesh
* (mirror modifier for eg does this). */
copy_v3_v3(mapped_vcos_data->vertexcos[index], co);
BLI_BITMAP_ENABLE(mapped_vcos_data->vertex_visit, index);
}
}
void BKE_mesh_foreach_mapped_vert_coords_get(const Mesh *me_eval,
float (*r_cos)[3],
const int totcos)
{
MappedVCosData user_data;
memset(r_cos, 0, sizeof(*r_cos) * totcos);
user_data.vertexcos = r_cos;
user_data.vertex_visit = BLI_BITMAP_NEW(totcos, __func__);
BKE_mesh_foreach_mapped_vert(me_eval, get_vertexcos__mapFunc, &user_data, MESH_FOREACH_NOP);
MEM_freeN(user_data.vertex_visit);
}