Files
test2/source/blender/blenkernel/intern/mesh_mirror.c
Hans Goudey cfa53e0fbe Refactor: Move normals out of MVert, lazy calculation
As described in T91186, this commit moves mesh vertex normals into a
contiguous array of float vectors in a custom data layer, how face
normals are currently stored.

The main interface is documented in `BKE_mesh.h`. Vertex and face
normals are now calculated on-demand and cached, retrieved with an
"ensure" function. Since the logical state of a mesh is now "has
normals when necessary", they can be retrieved from a `const` mesh.

The goal is to use on-demand calculation for all derived data, but
leave room for eager calculation for performance purposes (modifier
evaluation is threaded, but viewport data generation is not).

**Benefits**
This moves us closer to a SoA approach rather than the current AoS
paradigm. Accessing a contiguous `float3` is much more efficient than
retrieving data from a larger struct. The memory requirements for
accessing only normals or vertex locations are smaller, and at the
cost of more memory usage for just normals, they now don't have to
be converted between float and short, which also simplifies code

In the future, the remaining items can be removed from `MVert`,
leaving only `float3`, which has similar benefits (see T93602).

Removing the combination of derived and original data makes it
conceptually simpler to only calculate normals when necessary.
This is especially important now that we have more opportunities
for temporary meshes in geometry nodes.

**Performance**
In addition to the theoretical future performance improvements by
making `MVert == float3`, I've done some basic performance testing
on this patch directly. The data is fairly rough, but it gives an idea
about where things stand generally.
 - Mesh line primitive 4m Verts: 1.16x faster (36 -> 31 ms),
   showing that accessing just `MVert` is now more efficient.
 - Spring Splash Screen: 1.03-1.06 -> 1.06-1.11 FPS, a very slight
   change that at least shows there is no regression.
 - Sprite Fright Snail Smoosh: 3.30-3.40 -> 3.42-3.50 FPS, a small
   but observable speedup.
 - Set Position Node with Scaled Normal: 1.36x faster (53 -> 39 ms),
   shows that using normals in geometry nodes is faster.
 - Normal Calculation 1.6m Vert Cube: 1.19x faster (25 -> 21 ms),
   shows that calculating normals is slightly faster now.
 - File Size of 1.6m Vert Cube: 1.03x smaller (214.7 -> 208.4 MB),
   Normals are not saved in files, which can help with large meshes.

As for memory usage, it may be slightly more in some cases, but
I didn't observe any difference in the production files I tested.

**Tests**
Some modifiers and cycles test results need to be updated with this
commit, for two reasons:
 - The subdivision surface modifier is not responsible for calculating
   normals anymore. In master, the modifier creates different normals
   than the result of the `Mesh` normal calculation, so this is a bug
   fix.
 - There are small differences in the results of some modifiers that
   use normals because they are not converted to and from `short`
   anymore.

**Future improvements**
 - Remove `ModifierTypeInfo::dependsOnNormals`. Code in each modifier
   already retrieves normals if they are needed anyway.
 - Copy normals as part of a better CoW system for attributes.
 - Make more areas use lazy instead of eager normal calculation.
 - Remove `BKE_mesh_normals_tag_dirty` in more places since that is
   now the default state of a new mesh.
 - Possibly apply a similar change to derived face corner normals.

Differential Revision: https://developer.blender.org/D12770
2022-01-13 14:38:25 -06:00

500 lines
17 KiB
C

/*
* This program is free software; you can redistribute it and/or
* modify it under the terms of the GNU General Public License
* as published by the Free Software Foundation; either version 2
* of the License, or (at your option) any later version.
*
* This program is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU General Public License for more details.
*
* You should have received a copy of the GNU General Public License
* along with this program; if not, write to the Free Software Foundation,
* Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
*
* The Original Code is Copyright (C) Blender Foundation.
* All rights reserved.
*/
/** \file
* \ingroup bke
*/
#include "BLI_math.h"
#include "DNA_mesh_types.h"
#include "DNA_meshdata_types.h"
#include "DNA_object_types.h"
#include "BKE_deform.h"
#include "BKE_lib_id.h"
#include "BKE_lib_query.h"
#include "BKE_mesh.h"
#include "BKE_mesh_mirror.h"
#include "BKE_modifier.h"
#include "bmesh.h"
#include "bmesh_tools.h"
#include "MEM_guardedalloc.h"
#include "MOD_modifiertypes.h"
Mesh *BKE_mesh_mirror_bisect_on_mirror_plane_for_modifier(MirrorModifierData *mmd,
const Mesh *mesh,
int axis,
const float plane_co[3],
float plane_no[3])
{
bool do_bisect_flip_axis = ((axis == 0 && mmd->flag & MOD_MIR_BISECT_FLIP_AXIS_X) ||
(axis == 1 && mmd->flag & MOD_MIR_BISECT_FLIP_AXIS_Y) ||
(axis == 2 && mmd->flag & MOD_MIR_BISECT_FLIP_AXIS_Z));
const float bisect_distance = mmd->bisect_threshold;
Mesh *result;
BMesh *bm;
BMIter viter;
BMVert *v, *v_next;
bm = BKE_mesh_to_bmesh_ex(mesh,
&(struct BMeshCreateParams){0},
&(struct BMeshFromMeshParams){
.calc_face_normal = true,
.cd_mask_extra = {.vmask = CD_MASK_ORIGINDEX,
.emask = CD_MASK_ORIGINDEX,
.pmask = CD_MASK_ORIGINDEX},
});
/* Define bisecting plane (aka mirror plane). */
float plane[4];
if (!do_bisect_flip_axis) {
/* That reversed condition is a little weird, but for some reason that's how you keep
* the part of the mesh which is on the non-mirrored side when flip option is disabled.
* I think this is the expected behavior. */
negate_v3(plane_no);
}
plane_from_point_normal_v3(plane, plane_co, plane_no);
BM_mesh_bisect_plane(bm, plane, true, false, 0, 0, bisect_distance);
/* Plane definitions for vert killing. */
float plane_offset[4];
copy_v3_v3(plane_offset, plane);
plane_offset[3] = plane[3] - bisect_distance;
/* Delete verts across the mirror plane. */
BM_ITER_MESH_MUTABLE (v, v_next, &viter, bm, BM_VERTS_OF_MESH) {
if (plane_point_side_v3(plane_offset, v->co) > 0.0f) {
BM_vert_kill(bm, v);
}
}
result = BKE_mesh_from_bmesh_for_eval_nomain(bm, NULL, mesh);
BM_mesh_free(bm);
return result;
}
void BKE_mesh_mirror_apply_mirror_on_axis(struct Main *bmain,
Mesh *mesh,
const int axis,
const float dist)
{
BMesh *bm = BKE_mesh_to_bmesh_ex(mesh,
&(struct BMeshCreateParams){
.use_toolflags = 1,
},
&(struct BMeshFromMeshParams){
.calc_face_normal = true,
.cd_mask_extra =
{
.vmask = CD_MASK_SHAPEKEY,
},
});
BMO_op_callf(bm,
(BMO_FLAG_DEFAULTS & ~BMO_FLAG_RESPECT_HIDE),
"symmetrize input=%avef direction=%i dist=%f use_shapekey=%b",
axis,
dist,
true);
BM_mesh_bm_to_me(bmain,
bm,
mesh,
(&(struct BMeshToMeshParams){
.calc_object_remap = true,
}));
BM_mesh_free(bm);
}
Mesh *BKE_mesh_mirror_apply_mirror_on_axis_for_modifier(MirrorModifierData *mmd,
Object *ob,
const Mesh *mesh,
const int axis,
const bool use_correct_order_on_merge)
{
const float tolerance_sq = mmd->tolerance * mmd->tolerance;
const bool do_vtargetmap = (mmd->flag & MOD_MIR_NO_MERGE) == 0;
int tot_vtargetmap = 0; /* total merge vertices */
const bool do_bisect = ((axis == 0 && mmd->flag & MOD_MIR_BISECT_AXIS_X) ||
(axis == 1 && mmd->flag & MOD_MIR_BISECT_AXIS_Y) ||
(axis == 2 && mmd->flag & MOD_MIR_BISECT_AXIS_Z));
Mesh *result;
MVert *mv, *mv_prev;
MEdge *me;
MLoop *ml;
MPoly *mp;
float mtx[4][4];
float plane_co[3], plane_no[3];
int i;
int a, totshape;
int *vtargetmap = NULL, *vtmap_a = NULL, *vtmap_b = NULL;
/* mtx is the mirror transformation */
unit_m4(mtx);
mtx[axis][axis] = -1.0f;
Object *mirror_ob = mmd->mirror_ob;
if (mirror_ob != NULL) {
float tmp[4][4];
float itmp[4][4];
/* tmp is a transform from coords relative to the object's own origin,
* to coords relative to the mirror object origin */
invert_m4_m4(tmp, mirror_ob->obmat);
mul_m4_m4m4(tmp, tmp, ob->obmat);
/* itmp is the reverse transform back to origin-relative coordinates */
invert_m4_m4(itmp, tmp);
/* combine matrices to get a single matrix that translates coordinates into
* mirror-object-relative space, does the mirror, and translates back to
* origin-relative space */
mul_m4_series(mtx, itmp, mtx, tmp);
if (do_bisect) {
copy_v3_v3(plane_co, itmp[3]);
copy_v3_v3(plane_no, itmp[axis]);
/* Account for non-uniform scale in `ob`, see: T87592. */
float ob_scale[3] = {
len_squared_v3(ob->obmat[0]),
len_squared_v3(ob->obmat[1]),
len_squared_v3(ob->obmat[2]),
};
/* Scale to avoid precision loss with extreme values. */
const float ob_scale_max = max_fff(UNPACK3(ob_scale));
if (LIKELY(ob_scale_max != 0.0f)) {
mul_v3_fl(ob_scale, 1.0f / ob_scale_max);
mul_v3_v3(plane_no, ob_scale);
}
}
}
else if (do_bisect) {
copy_v3_v3(plane_co, mtx[3]);
/* Need to negate here, since that axis is inverted (for mirror transform). */
negate_v3_v3(plane_no, mtx[axis]);
}
Mesh *mesh_bisect = NULL;
if (do_bisect) {
mesh_bisect = BKE_mesh_mirror_bisect_on_mirror_plane_for_modifier(
mmd, mesh, axis, plane_co, plane_no);
mesh = mesh_bisect;
}
const int maxVerts = mesh->totvert;
const int maxEdges = mesh->totedge;
const int maxLoops = mesh->totloop;
const int maxPolys = mesh->totpoly;
result = BKE_mesh_new_nomain_from_template(
mesh, maxVerts * 2, maxEdges * 2, 0, maxLoops * 2, maxPolys * 2);
/* Copy custom-data to original geometry. */
CustomData_copy_data(&mesh->vdata, &result->vdata, 0, 0, maxVerts);
CustomData_copy_data(&mesh->edata, &result->edata, 0, 0, maxEdges);
CustomData_copy_data(&mesh->ldata, &result->ldata, 0, 0, maxLoops);
CustomData_copy_data(&mesh->pdata, &result->pdata, 0, 0, maxPolys);
/* Subsurf for eg won't have mesh data in the custom-data arrays.
* now add mvert/medge/mpoly layers. */
if (!CustomData_has_layer(&mesh->vdata, CD_MVERT)) {
memcpy(result->mvert, mesh->mvert, sizeof(*result->mvert) * mesh->totvert);
}
if (!CustomData_has_layer(&mesh->edata, CD_MEDGE)) {
memcpy(result->medge, mesh->medge, sizeof(*result->medge) * mesh->totedge);
}
if (!CustomData_has_layer(&mesh->pdata, CD_MPOLY)) {
memcpy(result->mloop, mesh->mloop, sizeof(*result->mloop) * mesh->totloop);
memcpy(result->mpoly, mesh->mpoly, sizeof(*result->mpoly) * mesh->totpoly);
}
/* Copy custom-data to new geometry,
* copy from its self because this data may have been created in the checks above. */
CustomData_copy_data(&result->vdata, &result->vdata, 0, maxVerts, maxVerts);
CustomData_copy_data(&result->edata, &result->edata, 0, maxEdges, maxEdges);
/* loops are copied later */
CustomData_copy_data(&result->pdata, &result->pdata, 0, maxPolys, maxPolys);
if (do_vtargetmap) {
/* second half is filled with -1 */
vtargetmap = MEM_malloc_arrayN(maxVerts, sizeof(int[2]), "MOD_mirror tarmap");
vtmap_a = vtargetmap;
vtmap_b = vtargetmap + maxVerts;
}
/* mirror vertex coordinates */
mv_prev = result->mvert;
mv = mv_prev + maxVerts;
for (i = 0; i < maxVerts; i++, mv++, mv_prev++) {
mul_m4_v3(mtx, mv->co);
if (do_vtargetmap) {
/* Compare location of the original and mirrored vertex,
* to see if they should be mapped for merging.
*
* Always merge from the copied into the original vertices so it's possible to
* generate a 1:1 mapping by scanning vertices from the beginning of the array
* as is done in #BKE_editmesh_vert_coords_when_deformed. Without this,
* the coordinates returned will sometimes point to the copied vertex locations, see:
* T91444.
*
* However, such a change also affects non-versionable things like some modifiers binding, so
* we cannot enforce that behavior on existing modifiers, in which case we keep using the
* old, incorrect behavior of merging the source vertex into its copy.
*/
if (use_correct_order_on_merge) {
if (UNLIKELY(len_squared_v3v3(mv_prev->co, mv->co) < tolerance_sq)) {
*vtmap_b = i;
tot_vtargetmap++;
/* average location */
mid_v3_v3v3(mv->co, mv_prev->co, mv->co);
copy_v3_v3(mv_prev->co, mv->co);
}
else {
*vtmap_b = -1;
}
/* Fill here to avoid 2x loops. */
*vtmap_a = -1;
}
else {
if (UNLIKELY(len_squared_v3v3(mv_prev->co, mv->co) < tolerance_sq)) {
*vtmap_a = maxVerts + i;
tot_vtargetmap++;
/* average location */
mid_v3_v3v3(mv->co, mv_prev->co, mv->co);
copy_v3_v3(mv_prev->co, mv->co);
}
else {
*vtmap_a = -1;
}
/* Fill here to avoid 2x loops. */
*vtmap_b = -1;
}
vtmap_a++;
vtmap_b++;
}
}
/* handle shape keys */
totshape = CustomData_number_of_layers(&result->vdata, CD_SHAPEKEY);
for (a = 0; a < totshape; a++) {
float(*cos)[3] = CustomData_get_layer_n(&result->vdata, CD_SHAPEKEY, a);
for (i = maxVerts; i < result->totvert; i++) {
mul_m4_v3(mtx, cos[i]);
}
}
/* adjust mirrored edge vertex indices */
me = result->medge + maxEdges;
for (i = 0; i < maxEdges; i++, me++) {
me->v1 += maxVerts;
me->v2 += maxVerts;
}
/* adjust mirrored poly loopstart indices, and reverse loop order (normals) */
mp = result->mpoly + maxPolys;
ml = result->mloop;
for (i = 0; i < maxPolys; i++, mp++) {
MLoop *ml2;
int j, e;
/* reverse the loop, but we keep the first vertex in the face the same,
* to ensure that quads are split the same way as on the other side */
CustomData_copy_data(
&result->ldata, &result->ldata, mp->loopstart, mp->loopstart + maxLoops, 1);
for (j = 1; j < mp->totloop; j++) {
CustomData_copy_data(&result->ldata,
&result->ldata,
mp->loopstart + j,
mp->loopstart + maxLoops + mp->totloop - j,
1);
}
ml2 = ml + mp->loopstart + maxLoops;
e = ml2[0].e;
for (j = 0; j < mp->totloop - 1; j++) {
ml2[j].e = ml2[j + 1].e;
}
ml2[mp->totloop - 1].e = e;
mp->loopstart += maxLoops;
}
/* adjust mirrored loop vertex and edge indices */
ml = result->mloop + maxLoops;
for (i = 0; i < maxLoops; i++, ml++) {
ml->v += maxVerts;
ml->e += maxEdges;
}
/* handle uvs,
* let tessface recalc handle updating the MTFace data */
if (mmd->flag & (MOD_MIR_MIRROR_U | MOD_MIR_MIRROR_V) ||
(is_zero_v2(mmd->uv_offset_copy) == false)) {
const bool do_mirr_u = (mmd->flag & MOD_MIR_MIRROR_U) != 0;
const bool do_mirr_v = (mmd->flag & MOD_MIR_MIRROR_V) != 0;
/* If set, flip around center of each tile. */
const bool do_mirr_udim = (mmd->flag & MOD_MIR_MIRROR_UDIM) != 0;
const int totuv = CustomData_number_of_layers(&result->ldata, CD_MLOOPUV);
for (a = 0; a < totuv; a++) {
MLoopUV *dmloopuv = CustomData_get_layer_n(&result->ldata, CD_MLOOPUV, a);
int j = maxLoops;
dmloopuv += j; /* second set of loops only */
for (; j-- > 0; dmloopuv++) {
if (do_mirr_u) {
float u = dmloopuv->uv[0];
if (do_mirr_udim) {
dmloopuv->uv[0] = ceilf(u) - fmodf(u, 1.0f) + mmd->uv_offset[0];
}
else {
dmloopuv->uv[0] = 1.0f - u + mmd->uv_offset[0];
}
}
if (do_mirr_v) {
float v = dmloopuv->uv[1];
if (do_mirr_udim) {
dmloopuv->uv[1] = ceilf(v) - fmodf(v, 1.0f) + mmd->uv_offset[1];
}
else {
dmloopuv->uv[1] = 1.0f - v + mmd->uv_offset[1];
}
}
dmloopuv->uv[0] += mmd->uv_offset_copy[0];
dmloopuv->uv[1] += mmd->uv_offset_copy[1];
}
}
}
/* handle custom split normals */
if (ob->type == OB_MESH && (((Mesh *)ob->data)->flag & ME_AUTOSMOOTH) &&
CustomData_has_layer(&result->ldata, CD_CUSTOMLOOPNORMAL)) {
const int totloop = result->totloop;
const int totpoly = result->totpoly;
float(*loop_normals)[3] = MEM_calloc_arrayN((size_t)totloop, sizeof(*loop_normals), __func__);
CustomData *ldata = &result->ldata;
short(*clnors)[2] = CustomData_get_layer(ldata, CD_CUSTOMLOOPNORMAL);
MLoopNorSpaceArray lnors_spacearr = {NULL};
/* The transform matrix of a normal must be
* the transpose of inverse of transform matrix of the geometry... */
float mtx_nor[4][4];
invert_m4_m4(mtx_nor, mtx);
transpose_m4(mtx_nor);
/* calculate custom normals into loop_normals, then mirror first half into second half */
BKE_mesh_normals_loop_split(result->mvert,
BKE_mesh_vertex_normals_ensure(mesh),
result->totvert,
result->medge,
result->totedge,
result->mloop,
loop_normals,
totloop,
result->mpoly,
BKE_mesh_poly_normals_ensure(mesh),
totpoly,
true,
mesh->smoothresh,
&lnors_spacearr,
clnors,
NULL);
/* mirroring has to account for loops being reversed in polys in second half */
mp = result->mpoly;
for (i = 0; i < maxPolys; i++, mp++) {
MPoly *mpmirror = result->mpoly + maxPolys + i;
int j;
for (j = mp->loopstart; j < mp->loopstart + mp->totloop; j++) {
int mirrorj = mpmirror->loopstart;
if (j > mp->loopstart) {
mirrorj += mpmirror->totloop - (j - mp->loopstart);
}
copy_v3_v3(loop_normals[mirrorj], loop_normals[j]);
mul_m4_v3(mtx_nor, loop_normals[mirrorj]);
BKE_lnor_space_custom_normal_to_data(
lnors_spacearr.lspacearr[mirrorj], loop_normals[mirrorj], clnors[mirrorj]);
}
}
MEM_freeN(loop_normals);
BKE_lnor_spacearr_free(&lnors_spacearr);
}
/* handle vgroup stuff */
if ((mmd->flag & MOD_MIR_VGROUP) && CustomData_has_layer(&result->vdata, CD_MDEFORMVERT)) {
MDeformVert *dvert = (MDeformVert *)CustomData_get_layer(&result->vdata, CD_MDEFORMVERT) +
maxVerts;
int *flip_map = NULL, flip_map_len = 0;
flip_map = BKE_object_defgroup_flip_map(ob, &flip_map_len, false);
if (flip_map) {
for (i = 0; i < maxVerts; dvert++, i++) {
/* merged vertices get both groups, others get flipped */
if (do_vtargetmap && (vtargetmap[i] != -1)) {
BKE_defvert_flip_merged(dvert, flip_map, flip_map_len);
}
else {
BKE_defvert_flip(dvert, flip_map, flip_map_len);
}
}
MEM_freeN(flip_map);
}
}
if (do_vtargetmap) {
/* slow - so only call if one or more merge verts are found,
* users may leave this on and not realize there is nothing to merge - campbell */
if (tot_vtargetmap) {
result = BKE_mesh_merge_verts(
result, vtargetmap, tot_vtargetmap, MESH_MERGE_VERTS_DUMP_IF_MAPPED);
}
MEM_freeN(vtargetmap);
}
if (mesh_bisect != NULL) {
BKE_id_free(NULL, mesh_bisect);
}
return result;
}