Files
test/source/blender/modifiers/intern/MOD_particleinstance.c

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

668 lines
20 KiB
C
Raw Normal View History

/* SPDX-License-Identifier: GPL-2.0-or-later
* Copyright 2005 Blender Foundation. All rights reserved. */
/** \file
* \ingroup modifiers
*/
#include "MEM_guardedalloc.h"
#include "BLI_utildefines.h"
#include "BLI_listbase.h"
#include "BLI_math.h"
#include "BLI_rand.h"
#include "BLI_string.h"
#include "BLT_translation.h"
#include "DNA_defaults.h"
#include "DNA_mesh_types.h"
#include "DNA_meshdata_types.h"
#include "DNA_screen_types.h"
#include "BKE_context.h"
#include "BKE_effect.h"
#include "BKE_lattice.h"
#include "BKE_lib_query.h"
#include "BKE_mesh.h"
#include "BKE_modifier.h"
#include "BKE_particle.h"
#include "BKE_pointcache.h"
#include "BKE_screen.h"
#include "UI_interface.h"
#include "UI_resources.h"
#include "RNA_access.h"
#include "RNA_prototypes.h"
#include "DEG_depsgraph_build.h"
#include "DEG_depsgraph_query.h"
2018-12-30 15:14:00 +11:00
#include "MOD_modifiertypes.h"
#include "MOD_ui_common.h"
2018-12-30 15:14:00 +11:00
static void initData(ModifierData *md)
{
ParticleInstanceModifierData *pimd = (ParticleInstanceModifierData *)md;
BLI_assert(MEMCMP_STRUCT_AFTER_IS_ZERO(pimd, modifier));
MEMCPY_STRUCT_AFTER(pimd, DNA_struct_default_get(ParticleInstanceModifierData), modifier);
}
static void requiredDataMask(ModifierData *md, CustomData_MeshMasks *r_cddata_masks)
{
ParticleInstanceModifierData *pimd = (ParticleInstanceModifierData *)md;
if (pimd->index_layer_name[0] != '\0' || pimd->value_layer_name[0] != '\0') {
r_cddata_masks->lmask |= CD_MASK_PROP_BYTE_COLOR;
}
}
static bool isDisabled(const struct Scene *scene, ModifierData *md, bool useRenderParams)
{
ParticleInstanceModifierData *pimd = (ParticleInstanceModifierData *)md;
ParticleSystem *psys;
ModifierData *ob_md;
/* The object type check is only needed here in case we have a placeholder
* object assigned (because the library containing the mesh is missing).
*
2019-08-31 01:19:22 +10:00
* In other cases it should be impossible to have a type mismatch.
*/
if (!pimd->ob || pimd->ob->type != OB_MESH) {
return true;
}
psys = BLI_findlink(&pimd->ob->particlesystem, pimd->psys - 1);
if (psys == NULL) {
return true;
}
/* If the psys modifier is disabled we cannot use its data.
* First look up the psys modifier from the object, then check if it is enabled.
*/
for (ob_md = pimd->ob->modifiers.first; ob_md; ob_md = ob_md->next) {
if (ob_md->type == eModifierType_ParticleSystem) {
ParticleSystemModifierData *psmd = (ParticleSystemModifierData *)ob_md;
if (psmd->psys == psys) {
int required_mode;
if (useRenderParams) {
required_mode = eModifierMode_Render;
}
else {
required_mode = eModifierMode_Realtime;
}
if (!BKE_modifier_is_enabled(scene, ob_md, required_mode)) {
return true;
}
break;
}
}
}
return false;
}
static void updateDepsgraph(ModifierData *md, const ModifierUpdateDepsgraphContext *ctx)
{
ParticleInstanceModifierData *pimd = (ParticleInstanceModifierData *)md;
if (pimd->ob != NULL) {
DEG_add_object_relation(
ctx->node, pimd->ob, DEG_OB_COMP_TRANSFORM, "Particle Instance Modifier");
DEG_add_object_relation(
ctx->node, pimd->ob, DEG_OB_COMP_GEOMETRY, "Particle Instance Modifier");
}
}
static void foreachIDLink(ModifierData *md, Object *ob, IDWalkFunc walk, void *userData)
{
ParticleInstanceModifierData *pimd = (ParticleInstanceModifierData *)md;
walk(userData, ob, (ID **)&pimd->ob, IDWALK_CB_NOP);
}
static bool particle_skip(ParticleInstanceModifierData *pimd, ParticleSystem *psys, int p)
{
const bool between = (psys->part->childtype == PART_CHILD_FACES);
ParticleData *pa;
int totpart, randp, minp, maxp;
if (p >= psys->totpart) {
ChildParticle *cpa = psys->child + (p - psys->totpart);
2018-06-06 22:17:06 +02:00
pa = psys->particles + (between ? cpa->pa[0] : cpa->parent);
}
else {
pa = psys->particles + p;
}
if (pa) {
if (pa->alive == PARS_UNBORN && (pimd->flag & eParticleInstanceFlag_Unborn) == 0) {
return true;
}
if (pa->alive == PARS_ALIVE && (pimd->flag & eParticleInstanceFlag_Alive) == 0) {
return true;
}
if (pa->alive == PARS_DEAD && (pimd->flag & eParticleInstanceFlag_Dead) == 0) {
return true;
}
if (pa->flag & (PARS_UNEXIST | PARS_NO_DISP)) {
return true;
}
}
if (pimd->particle_amount == 1.0f) {
/* Early output, all particles are to be instanced. */
return false;
}
/* Randomly skip particles based on desired amount of visible particles. */
totpart = psys->totpart + psys->totchild;
/* TODO: make randomization optional? */
randp = (int)(psys_frand(psys, 3578 + p) * totpart) % totpart;
2018-06-06 22:17:06 +02:00
minp = (int)(totpart * pimd->particle_offset) % (totpart + 1);
maxp = (int)(totpart * (pimd->particle_offset + pimd->particle_amount)) % (totpart + 1);
if (maxp > minp) {
return randp < minp || randp >= maxp;
}
if (maxp < minp) {
return randp < minp && randp >= maxp;
}
return true;
return false;
}
static void store_float_in_vcol(MLoopCol *vcol, float float_value)
{
const uchar value = unit_float_to_uchar_clamp(float_value);
vcol->r = vcol->g = vcol->b = value;
vcol->a = 1.0f;
}
static Mesh *modifyMesh(ModifierData *md, const ModifierEvalContext *ctx, Mesh *mesh)
{
Mesh *result;
ParticleInstanceModifierData *pimd = (ParticleInstanceModifierData *)md;
struct Scene *scene = DEG_get_evaluated_scene(ctx->depsgraph);
ParticleSimulationData sim;
ParticleSystem *psys = NULL;
ParticleData *pa = NULL;
int totvert, totpoly, totloop, totedge;
int maxvert, maxpoly, maxloop, maxedge, part_end = 0, part_start;
int k, p, p_skip;
short track = ctx->object->trackflag % 3, trackneg, axis = pimd->axis;
float max_co = 0.0, min_co = 0.0, temp_co[3];
float *size = NULL;
float spacemat[4][4];
const bool use_parents = pimd->flag & eParticleInstanceFlag_Parents;
const bool use_children = pimd->flag & eParticleInstanceFlag_Children;
bool between;
trackneg = ((ctx->object->trackflag > 2) ? 1 : 0);
if (pimd->ob == ctx->object) {
pimd->ob = NULL;
return mesh;
}
if (pimd->ob) {
psys = BLI_findlink(&pimd->ob->particlesystem, pimd->psys - 1);
if (psys == NULL || psys->totpart == 0) {
return mesh;
}
}
else {
return mesh;
}
part_start = use_parents ? 0 : psys->totpart;
part_end = 0;
if (use_parents) {
part_end += psys->totpart;
}
if (use_children) {
part_end += psys->totchild;
}
if (part_end == 0) {
return mesh;
}
sim.depsgraph = ctx->depsgraph;
sim.scene = scene;
sim.ob = pimd->ob;
sim.psys = psys;
sim.psmd = psys_get_modifier(pimd->ob, psys);
between = (psys->part->childtype == PART_CHILD_FACES);
if (pimd->flag & eParticleInstanceFlag_UseSize) {
float *si;
si = size = MEM_calloc_arrayN(part_end, sizeof(float), "particle size array");
if (pimd->flag & eParticleInstanceFlag_Parents) {
for (p = 0, pa = psys->particles; p < psys->totpart; p++, pa++, si++) {
*si = pa->size;
}
}
if (pimd->flag & eParticleInstanceFlag_Children) {
ChildParticle *cpa = psys->child;
for (p = 0; p < psys->totchild; p++, cpa++, si++) {
*si = psys_get_child_size(psys, cpa, 0.0f, NULL);
}
}
}
switch (pimd->space) {
case eParticleInstanceSpace_World:
/* particle states are in world space already */
unit_m4(spacemat);
break;
case eParticleInstanceSpace_Local:
/* get particle states in the particle object's local space */
invert_m4_m4(spacemat, pimd->ob->object_to_world);
break;
default:
/* should not happen */
BLI_assert(false);
break;
}
totvert = mesh->totvert;
totpoly = mesh->totpoly;
totloop = mesh->totloop;
totedge = mesh->totedge;
/* count particles */
maxvert = 0;
maxpoly = 0;
maxloop = 0;
maxedge = 0;
for (p = part_start; p < part_end; p++) {
if (particle_skip(pimd, psys, p)) {
continue;
}
maxvert += totvert;
maxpoly += totpoly;
maxloop += totloop;
maxedge += totedge;
}
psys_sim_data_init(&sim);
if (psys->flag & (PSYS_HAIR_DONE | PSYS_KEYED) || psys->pointcache->flag & PTCACHE_BAKED) {
float min[3], max[3];
INIT_MINMAX(min, max);
BKE_mesh_minmax(mesh, min, max);
min_co = min[track];
max_co = max[track];
}
result = BKE_mesh_new_nomain_from_template(mesh, maxvert, maxedge, 0, maxloop, maxpoly);
const MPoly *orig_mpoly = BKE_mesh_polys(mesh);
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 MLoop *orig_mloop = BKE_mesh_loops(mesh);
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
float(*positions)[3] = BKE_mesh_vert_positions_for_write(result);
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
MEdge *edges = BKE_mesh_edges_for_write(result);
MPoly *mpoly = BKE_mesh_polys_for_write(result);
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
MLoop *mloop = BKE_mesh_loops_for_write(result);
MLoopCol *mloopcols_index = CustomData_get_layer_named(
&result->ldata, CD_PROP_BYTE_COLOR, pimd->index_layer_name);
MLoopCol *mloopcols_value = CustomData_get_layer_named(
&result->ldata, CD_PROP_BYTE_COLOR, pimd->value_layer_name);
int *vert_part_index = NULL;
float *vert_part_value = NULL;
if (mloopcols_index != NULL) {
vert_part_index = MEM_calloc_arrayN(maxvert, sizeof(int), "vertex part index array");
}
if (mloopcols_value) {
vert_part_value = MEM_calloc_arrayN(maxvert, sizeof(float), "vertex part value array");
}
for (p = part_start, p_skip = 0; p < part_end; p++) {
float prev_dir[3];
float frame[4]; /* frame orientation quaternion */
2018-06-06 22:17:06 +02:00
float p_random = psys_frand(psys, 77091 + 283 * p);
/* skip particle? */
if (particle_skip(pimd, psys, p)) {
continue;
}
/* set vertices coordinates */
for (k = 0; k < totvert; k++) {
ParticleKey state;
int vindex = p_skip * totvert + k;
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
CustomData_copy_data(&mesh->vdata, &result->vdata, k, vindex, 1);
if (vert_part_index != NULL) {
vert_part_index[vindex] = p;
}
if (vert_part_value != NULL) {
vert_part_value[vindex] = p_random;
}
/* Change orientation based on object trackflag. */
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
copy_v3_v3(temp_co, positions[vindex]);
positions[vindex][axis] = temp_co[track];
positions[vindex][(axis + 1) % 3] = temp_co[(track + 1) % 3];
positions[vindex][(axis + 2) % 3] = temp_co[(track + 2) % 3];
/* get particle state */
if ((psys->flag & (PSYS_HAIR_DONE | PSYS_KEYED) || psys->pointcache->flag & PTCACHE_BAKED) &&
(pimd->flag & eParticleInstanceFlag_Path)) {
float ran = 0.0f;
if (pimd->random_position != 0.0f) {
ran = pimd->random_position * BLI_hash_frand(psys->seed + p);
}
if (pimd->flag & eParticleInstanceFlag_KeepShape) {
state.time = pimd->position * (1.0f - ran);
}
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 00:10:43 -05:00
state.time = (positions[vindex][axis] - min_co) / (max_co - min_co) * pimd->position *
(1.0f - ran);
if (trackneg) {
state.time = 1.0f - state.time;
}
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
positions[vindex][axis] = 0.0;
}
psys_get_particle_on_path(&sim, p, &state, 1);
normalize_v3(state.vel);
/* Incrementally Rotating Frame (Bishop Frame) */
if (k == 0) {
float hairmat[4][4];
float mat[3][3];
if (p < psys->totpart) {
pa = psys->particles + p;
}
else {
ChildParticle *cpa = psys->child + (p - psys->totpart);
2018-06-06 22:17:06 +02:00
pa = psys->particles + (between ? cpa->pa[0] : cpa->parent);
}
psys_mat_hair_to_global(sim.ob, sim.psmd->mesh_final, sim.psys->part->from, pa, hairmat);
copy_m3_m4(mat, hairmat);
/* to quaternion */
mat3_to_quat(frame, mat);
if (pimd->rotation > 0.0f || pimd->random_rotation > 0.0f) {
2018-06-06 22:17:06 +02:00
float angle = 2.0f * M_PI *
(pimd->rotation +
pimd->random_rotation * (psys_frand(psys, 19957323 + p) - 0.5f));
const float eul[3] = {0.0f, 0.0f, angle};
float rot[4];
eul_to_quat(rot, eul);
mul_qt_qtqt(frame, frame, rot);
}
/* NOTE: direction is same as normal vector currently,
* but best to keep this separate so the frame can be
* rotated later if necessary
*/
copy_v3_v3(prev_dir, state.vel);
}
else {
float rot[4];
/* incrementally rotate along bend direction */
rotation_between_vecs_to_quat(rot, prev_dir, state.vel);
mul_qt_qtqt(frame, rot, frame);
copy_v3_v3(prev_dir, state.vel);
}
copy_qt_qt(state.rot, frame);
#if 0
/* Absolute Frame (Frenet Frame) */
if (state.vel[axis] < -0.9999f || state.vel[axis] > 0.9999f) {
unit_qt(state.rot);
}
else {
float cross[3];
float temp[3] = {0.0f, 0.0f, 0.0f};
temp[axis] = 1.0f;
cross_v3_v3v3(cross, temp, state.vel);
/* state.vel[axis] is the only component surviving from a dot product with the axis */
axis_angle_to_quat(state.rot, cross, saacos(state.vel[axis]));
}
#endif
}
else {
state.time = -1.0;
psys_get_particle_state(&sim, p, &state, 1);
}
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
mul_qt_v3(state.rot, positions[vindex]);
if (pimd->flag & eParticleInstanceFlag_UseSize) {
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
mul_v3_fl(positions[vindex], size[p]);
}
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
add_v3_v3(positions[vindex], state.co);
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
mul_m4_v3(spacemat, positions[vindex]);
}
/* Create edges and adjust edge vertex indices. */
CustomData_copy_data(&mesh->edata, &result->edata, 0, p_skip * totedge, totedge);
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
MEdge *me = &edges[p_skip * totedge];
for (k = 0; k < totedge; k++, me++) {
me->v1 += p_skip * totvert;
me->v2 += p_skip * totvert;
}
/* create polys and loops */
for (k = 0; k < totpoly; k++) {
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 MPoly *inMP = orig_mpoly + k;
MPoly *mp = mpoly + p_skip * totpoly + k;
CustomData_copy_data(&mesh->pdata, &result->pdata, k, p_skip * totpoly + k, 1);
*mp = *inMP;
mp->loopstart += p_skip * totloop;
{
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 MLoop *inML = orig_mloop + inMP->loopstart;
MLoop *ml = mloop + mp->loopstart;
int j = mp->totloop;
CustomData_copy_data(&mesh->ldata, &result->ldata, inMP->loopstart, mp->loopstart, j);
for (; j; j--, ml++, inML++) {
ml->v = inML->v + (p_skip * totvert);
ml->e = inML->e + (p_skip * totedge);
const int ml_index = (ml - mloop);
if (mloopcols_index != NULL) {
const int part_index = vert_part_index[ml->v];
2018-12-08 15:01:47 +11:00
store_float_in_vcol(&mloopcols_index[ml_index],
(float)part_index / (float)(psys->totpart - 1));
}
if (mloopcols_value != NULL) {
const float part_value = vert_part_value[ml->v];
store_float_in_vcol(&mloopcols_value[ml_index], part_value);
}
}
}
}
p_skip++;
}
psys_sim_data_free(&sim);
if (size) {
MEM_freeN(size);
}
MEM_SAFE_FREE(vert_part_index);
MEM_SAFE_FREE(vert_part_value);
return result;
}
static void panel_draw(const bContext *UNUSED(C), Panel *panel)
{
uiLayout *row;
uiLayout *layout = panel->layout;
int toggles_flag = UI_ITEM_R_TOGGLE | UI_ITEM_R_FORCE_BLANK_DECORATE;
PointerRNA ob_ptr;
PointerRNA *ptr = modifier_panel_get_property_pointers(panel, &ob_ptr);
PointerRNA particle_obj_ptr = RNA_pointer_get(ptr, "object");
uiLayoutSetPropSep(layout, true);
uiItemR(layout, ptr, "object", 0, NULL, ICON_NONE);
if (!RNA_pointer_is_null(&particle_obj_ptr)) {
uiItemPointerR(layout,
ptr,
"particle_system",
&particle_obj_ptr,
"particle_systems",
IFACE_("Particle System"),
ICON_NONE);
}
else {
uiItemR(layout, ptr, "particle_system_index", 0, IFACE_("Particle System"), ICON_NONE);
}
uiItemS(layout);
row = uiLayoutRowWithHeading(layout, true, IFACE_("Create Instances"));
uiItemR(row, ptr, "use_normal", toggles_flag, NULL, ICON_NONE);
uiItemR(row, ptr, "use_children", toggles_flag, NULL, ICON_NONE);
uiItemR(row, ptr, "use_size", toggles_flag, NULL, ICON_NONE);
row = uiLayoutRowWithHeading(layout, true, IFACE_("Show"));
uiItemR(row, ptr, "show_alive", toggles_flag, NULL, ICON_NONE);
uiItemR(row, ptr, "show_dead", toggles_flag, NULL, ICON_NONE);
uiItemR(row, ptr, "show_unborn", toggles_flag, NULL, ICON_NONE);
uiItemR(layout, ptr, "particle_amount", 0, IFACE_("Amount"), ICON_NONE);
uiItemR(layout, ptr, "particle_offset", 0, IFACE_("Offset"), ICON_NONE);
uiItemS(layout);
uiItemR(layout, ptr, "space", 0, IFACE_("Coordinate Space"), ICON_NONE);
row = uiLayoutRow(layout, true);
uiItemR(row, ptr, "axis", UI_ITEM_R_EXPAND, NULL, ICON_NONE);
modifier_panel_end(layout, ptr);
}
static void path_panel_draw_header(const bContext *UNUSED(C), Panel *panel)
{
uiLayout *layout = panel->layout;
PointerRNA *ptr = modifier_panel_get_property_pointers(panel, NULL);
uiItemR(layout, ptr, "use_path", 0, IFACE_("Create Along Paths"), ICON_NONE);
}
static void path_panel_draw(const bContext *UNUSED(C), Panel *panel)
{
uiLayout *col;
uiLayout *layout = panel->layout;
PointerRNA ob_ptr;
PointerRNA *ptr = modifier_panel_get_property_pointers(panel, &ob_ptr);
uiLayoutSetPropSep(layout, true);
uiLayoutSetActive(layout, RNA_boolean_get(ptr, "use_path"));
col = uiLayoutColumn(layout, true);
uiItemR(col, ptr, "position", UI_ITEM_R_SLIDER, NULL, ICON_NONE);
uiItemR(col, ptr, "random_position", UI_ITEM_R_SLIDER, IFACE_("Random"), ICON_NONE);
col = uiLayoutColumn(layout, true);
uiItemR(col, ptr, "rotation", UI_ITEM_R_SLIDER, NULL, ICON_NONE);
uiItemR(col, ptr, "random_rotation", UI_ITEM_R_SLIDER, IFACE_("Random"), ICON_NONE);
uiItemR(layout, ptr, "use_preserve_shape", 0, NULL, ICON_NONE);
}
static void layers_panel_draw(const bContext *UNUSED(C), Panel *panel)
{
uiLayout *col;
uiLayout *layout = panel->layout;
PointerRNA ob_ptr;
PointerRNA *ptr = modifier_panel_get_property_pointers(panel, &ob_ptr);
PointerRNA obj_data_ptr = RNA_pointer_get(&ob_ptr, "data");
uiLayoutSetPropSep(layout, true);
col = uiLayoutColumn(layout, false);
uiItemPointerR(
col, ptr, "index_layer_name", &obj_data_ptr, "vertex_colors", IFACE_("Index"), ICON_NONE);
uiItemPointerR(
col, ptr, "value_layer_name", &obj_data_ptr, "vertex_colors", IFACE_("Value"), ICON_NONE);
}
static void panelRegister(ARegionType *region_type)
{
PanelType *panel_type = modifier_panel_register(
region_type, eModifierType_ParticleInstance, panel_draw);
modifier_subpanel_register(
region_type, "paths", "", path_panel_draw_header, path_panel_draw, panel_type);
modifier_subpanel_register(region_type, "layers", "Layers", NULL, layers_panel_draw, panel_type);
}
ModifierTypeInfo modifierType_ParticleInstance = {
/* name */ N_("ParticleInstance"),
/* structName */ "ParticleInstanceModifierData",
/* structSize */ sizeof(ParticleInstanceModifierData),
/* srna */ &RNA_ParticleInstanceModifier,
/* type */ eModifierTypeType_Constructive,
/* flags */ eModifierTypeFlag_AcceptsMesh | eModifierTypeFlag_SupportsMapping |
eModifierTypeFlag_SupportsEditmode | eModifierTypeFlag_EnableInEditmode,
/* icon */ ICON_MOD_PARTICLE_INSTANCE,
/* copyData */ BKE_modifier_copydata_generic,
/* deformVerts */ NULL,
/* deformMatrices */ NULL,
/* deformVertsEM */ NULL,
/* deformMatricesEM */ NULL,
/* modifyMesh */ modifyMesh,
/* modifyGeometrySet */ NULL,
/* initData */ initData,
/* requiredDataMask */ requiredDataMask,
/* freeData */ NULL,
/* isDisabled */ isDisabled,
/* updateDepsgraph */ updateDepsgraph,
/* dependsOnTime */ NULL,
/* dependsOnNormals */ NULL,
/* foreachIDLink */ foreachIDLink,
/* foreachTexLink */ NULL,
/* freeRuntimeData */ NULL,
/* panelRegister */ panelRegister,
/* blendWrite */ NULL,
/* blendRead */ NULL,
};