archive/blender-archive
Archived
1
1
Fork 0
Code Pull Requests Packages Activity
This repository has been archived on 2023-10-09. You can view files and clone it, but cannot push or open issues or pull requests.
Files
b9451c04931d228161c1fb212c6c7a09545a86b7
blender-archive/source/blender/draw/intern/draw_cache_impl_mesh.c

2013 lines
64 KiB
C
Raw Normal View History

Convert MBC_ API to Mesh (instead of derived mesh) and move it to mesh_render This includes a few fixes in the MBC_ api. The idea here is for this to be the only interface the render engines will deal with for the meshes. If we need to expose special options for sculpting engine we refactor this accordingly. But for now we are shaping this in a per-case base. Note: * We still need to hook up to the depsgraph to force clear/update of batch_cache when mesh changes (I'm waiting for Sergey Sharybin's depsgraph update for this though) * Also ideally we could/should use BMesh directly instead of DerivedMesh, but this will do for now. Note 2: In the end I renamed the `BKE_mesh_render` functions to `static mesh_render`. We can re-expose them as BKE_* later once we need it. Reviewers: merwin Subscribers: fclem Differential Revision: https://developer.blender.org/D2476
2017-01-25 09:16:29 +01:00
/*
* ***** BEGIN GPL LICENSE BLOCK *****
*
* 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) 2017 by Blender Foundation.
* All rights reserved.
*
* Contributor(s): Blender Foundation, Mike Erwin, Dalai Felinto
*
* ***** END GPL LICENSE BLOCK *****
*/
Cleanup: move draw-cache creation from BKE to DRW Creating draw-cache should only ever be used by the draw-manager.
2017-04-21 21:14:11 +10:00
/** \file draw_cache_impl_mesh.c
* \ingroup draw
Convert MBC_ API to Mesh (instead of derived mesh) and move it to mesh_render This includes a few fixes in the MBC_ api. The idea here is for this to be the only interface the render engines will deal with for the meshes. If we need to expose special options for sculpting engine we refactor this accordingly. But for now we are shaping this in a per-case base. Note: * We still need to hook up to the depsgraph to force clear/update of batch_cache when mesh changes (I'm waiting for Sergey Sharybin's depsgraph update for this though) * Also ideally we could/should use BMesh directly instead of DerivedMesh, but this will do for now. Note 2: In the end I renamed the `BKE_mesh_render` functions to `static mesh_render`. We can re-expose them as BKE_* later once we need it. Reviewers: merwin Subscribers: fclem Differential Revision: https://developer.blender.org/D2476
2017-01-25 09:16:29 +01:00
*
* \brief Mesh API for render engines
*/
#include "MEM_guardedalloc.h"
Rework of BKE's mesh_render to support BMesh directly. Note that since there is no (efficient) ways to get arrays of MVert/MEdge/etc. out of a BMesh, I refactored quite heavily internals of BKE_mesh_render. Now, when you do need acess to mesh data to generate cached batches, you create an abstract struct from mesh (either Mesh or BMesh if available), and then use advanced helpers to extract needed data, on a per-item basis (no more handling of arrays of verts/edges/... in batches code). This allows to: * Avoid having to create arrays of BMesh elements. * Take advantage of existing advanced BMesh topology and connectivity data. Reviewers: dfelinto, fclem, merwin Differential Revision: https://developer.blender.org/D2521
2017-02-23 11:54:40 +01:00
#include "BLI_utildefines.h"
Clay Engine: Correct Mesh Normals
2017-02-16 16:19:48 +01:00
#include "BLI_math_vector.h"
Mesh Batch Cache: Add support for uv/vcol/tangent data layers. This is a bad implementation waiting for some improvement that are : - Gawain support for multiple names. -> will remove duplication of attribute alloc - Glsl safe name string escaping. -> will remove any risk of hash colision. Old glsl drawing was relying on attribute being choosen by DerivedMesh at drawing time. For this reason, we declare all possible attrib "name" for each data layer inside the batches. In the glsl code, we declare required data by type and name. Then Gawain only bind vbos if names correspond. This is way cleaner as we do not need to access the CD itself when drawing. One other problem is that the hash maybe 11 caracters long and rapidly overload gawain's attrib name buffer.
2017-05-01 17:58:15 +02:00
#include "BLI_string.h"
Clay Engine: Correct Mesh Normals
2017-02-16 16:19:48 +01:00
Convert MBC_ API to Mesh (instead of derived mesh) and move it to mesh_render This includes a few fixes in the MBC_ api. The idea here is for this to be the only interface the render engines will deal with for the meshes. If we need to expose special options for sculpting engine we refactor this accordingly. But for now we are shaping this in a per-case base. Note: * We still need to hook up to the depsgraph to force clear/update of batch_cache when mesh changes (I'm waiting for Sergey Sharybin's depsgraph update for this though) * Also ideally we could/should use BMesh directly instead of DerivedMesh, but this will do for now. Note 2: In the end I renamed the `BKE_mesh_render` functions to `static mesh_render`. We can re-expose them as BKE_* later once we need it. Reviewers: merwin Subscribers: fclem Differential Revision: https://developer.blender.org/D2476
2017-01-25 09:16:29 +01:00
#include "DNA_mesh_types.h"
#include "DNA_meshdata_types.h"
#include "BKE_customdata.h"
#include "BKE_DerivedMesh.h"
#include "BKE_editmesh.h"
#include "BKE_mesh.h"
Rework of BKE's mesh_render to support BMesh directly. Note that since there is no (efficient) ways to get arrays of MVert/MEdge/etc. out of a BMesh, I refactored quite heavily internals of BKE_mesh_render. Now, when you do need acess to mesh data to generate cached batches, you create an abstract struct from mesh (either Mesh or BMesh if available), and then use advanced helpers to extract needed data, on a per-item basis (no more handling of arrays of verts/edges/... in batches code). This allows to: * Avoid having to create arrays of BMesh elements. * Take advantage of existing advanced BMesh topology and connectivity data. Reviewers: dfelinto, fclem, merwin Differential Revision: https://developer.blender.org/D2521
2017-02-23 11:54:40 +01:00
#include "bmesh.h"
Convert MBC_ API to Mesh (instead of derived mesh) and move it to mesh_render This includes a few fixes in the MBC_ api. The idea here is for this to be the only interface the render engines will deal with for the meshes. If we need to expose special options for sculpting engine we refactor this accordingly. But for now we are shaping this in a per-case base. Note: * We still need to hook up to the depsgraph to force clear/update of batch_cache when mesh changes (I'm waiting for Sergey Sharybin's depsgraph update for this though) * Also ideally we could/should use BMesh directly instead of DerivedMesh, but this will do for now. Note 2: In the end I renamed the `BKE_mesh_render` functions to `static mesh_render`. We can re-expose them as BKE_* later once we need it. Reviewers: merwin Subscribers: fclem Differential Revision: https://developer.blender.org/D2476
2017-01-25 09:16:29 +01:00
#include "GPU_batch.h"
Cleanup: move draw-cache creation from BKE to DRW Creating draw-cache should only ever be used by the draw-manager.
2017-04-21 21:14:11 +10:00
#include "draw_cache_impl.h" /* own include */
static void mesh_batch_cache_clear(Mesh *me);
Convert MBC_ API to Mesh (instead of derived mesh) and move it to mesh_render This includes a few fixes in the MBC_ api. The idea here is for this to be the only interface the render engines will deal with for the meshes. If we need to expose special options for sculpting engine we refactor this accordingly. But for now we are shaping this in a per-case base. Note: * We still need to hook up to the depsgraph to force clear/update of batch_cache when mesh changes (I'm waiting for Sergey Sharybin's depsgraph update for this though) * Also ideally we could/should use BMesh directly instead of DerivedMesh, but this will do for now. Note 2: In the end I renamed the `BKE_mesh_render` functions to `static mesh_render`. We can re-expose them as BKE_* later once we need it. Reviewers: merwin Subscribers: fclem Differential Revision: https://developer.blender.org/D2476
2017-01-25 09:16:29 +01:00
/* ---------------------------------------------------------------------- */
Rework of BKE's mesh_render to support BMesh directly. Note that since there is no (efficient) ways to get arrays of MVert/MEdge/etc. out of a BMesh, I refactored quite heavily internals of BKE_mesh_render. Now, when you do need acess to mesh data to generate cached batches, you create an abstract struct from mesh (either Mesh or BMesh if available), and then use advanced helpers to extract needed data, on a per-item basis (no more handling of arrays of verts/edges/... in batches code). This allows to: * Avoid having to create arrays of BMesh elements. * Take advantage of existing advanced BMesh topology and connectivity data. Reviewers: dfelinto, fclem, merwin Differential Revision: https://developer.blender.org/D2521
2017-02-23 11:54:40 +01:00
/* Mesh/BMesh Interface, direct access to basic data. */
Convert MBC_ API to Mesh (instead of derived mesh) and move it to mesh_render This includes a few fixes in the MBC_ api. The idea here is for this to be the only interface the render engines will deal with for the meshes. If we need to expose special options for sculpting engine we refactor this accordingly. But for now we are shaping this in a per-case base. Note: * We still need to hook up to the depsgraph to force clear/update of batch_cache when mesh changes (I'm waiting for Sergey Sharybin's depsgraph update for this though) * Also ideally we could/should use BMesh directly instead of DerivedMesh, but this will do for now. Note 2: In the end I renamed the `BKE_mesh_render` functions to `static mesh_render`. We can re-expose them as BKE_* later once we need it. Reviewers: merwin Subscribers: fclem Differential Revision: https://developer.blender.org/D2476
2017-01-25 09:16:29 +01:00
Cleanup: replace num/ct/tot with 'len'
2017-04-19 00:10:32 +10:00
static int mesh_render_verts_len_get(Mesh *me)
Convert MBC_ API to Mesh (instead of derived mesh) and move it to mesh_render This includes a few fixes in the MBC_ api. The idea here is for this to be the only interface the render engines will deal with for the meshes. If we need to expose special options for sculpting engine we refactor this accordingly. But for now we are shaping this in a per-case base. Note: * We still need to hook up to the depsgraph to force clear/update of batch_cache when mesh changes (I'm waiting for Sergey Sharybin's depsgraph update for this though) * Also ideally we could/should use BMesh directly instead of DerivedMesh, but this will do for now. Note 2: In the end I renamed the `BKE_mesh_render` functions to `static mesh_render`. We can re-expose them as BKE_* later once we need it. Reviewers: merwin Subscribers: fclem Differential Revision: https://developer.blender.org/D2476
2017-01-25 09:16:29 +01:00
{
Rework of BKE's mesh_render to support BMesh directly. Note that since there is no (efficient) ways to get arrays of MVert/MEdge/etc. out of a BMesh, I refactored quite heavily internals of BKE_mesh_render. Now, when you do need acess to mesh data to generate cached batches, you create an abstract struct from mesh (either Mesh or BMesh if available), and then use advanced helpers to extract needed data, on a per-item basis (no more handling of arrays of verts/edges/... in batches code). This allows to: * Avoid having to create arrays of BMesh elements. * Take advantage of existing advanced BMesh topology and connectivity data. Reviewers: dfelinto, fclem, merwin Differential Revision: https://developer.blender.org/D2521
2017-02-23 11:54:40 +01:00
return me->edit_btmesh ? me->edit_btmesh->bm->totvert : me->totvert;
Convert MBC_ API to Mesh (instead of derived mesh) and move it to mesh_render This includes a few fixes in the MBC_ api. The idea here is for this to be the only interface the render engines will deal with for the meshes. If we need to expose special options for sculpting engine we refactor this accordingly. But for now we are shaping this in a per-case base. Note: * We still need to hook up to the depsgraph to force clear/update of batch_cache when mesh changes (I'm waiting for Sergey Sharybin's depsgraph update for this though) * Also ideally we could/should use BMesh directly instead of DerivedMesh, but this will do for now. Note 2: In the end I renamed the `BKE_mesh_render` functions to `static mesh_render`. We can re-expose them as BKE_* later once we need it. Reviewers: merwin Subscribers: fclem Differential Revision: https://developer.blender.org/D2476
2017-01-25 09:16:29 +01:00
}
Cleanup: replace num/ct/tot with 'len'
2017-04-19 00:10:32 +10:00
static int mesh_render_edges_len_get(Mesh *me)
Convert MBC_ API to Mesh (instead of derived mesh) and move it to mesh_render This includes a few fixes in the MBC_ api. The idea here is for this to be the only interface the render engines will deal with for the meshes. If we need to expose special options for sculpting engine we refactor this accordingly. But for now we are shaping this in a per-case base. Note: * We still need to hook up to the depsgraph to force clear/update of batch_cache when mesh changes (I'm waiting for Sergey Sharybin's depsgraph update for this though) * Also ideally we could/should use BMesh directly instead of DerivedMesh, but this will do for now. Note 2: In the end I renamed the `BKE_mesh_render` functions to `static mesh_render`. We can re-expose them as BKE_* later once we need it. Reviewers: merwin Subscribers: fclem Differential Revision: https://developer.blender.org/D2476
2017-01-25 09:16:29 +01:00
{
Rework of BKE's mesh_render to support BMesh directly. Note that since there is no (efficient) ways to get arrays of MVert/MEdge/etc. out of a BMesh, I refactored quite heavily internals of BKE_mesh_render. Now, when you do need acess to mesh data to generate cached batches, you create an abstract struct from mesh (either Mesh or BMesh if available), and then use advanced helpers to extract needed data, on a per-item basis (no more handling of arrays of verts/edges/... in batches code). This allows to: * Avoid having to create arrays of BMesh elements. * Take advantage of existing advanced BMesh topology and connectivity data. Reviewers: dfelinto, fclem, merwin Differential Revision: https://developer.blender.org/D2521
2017-02-23 11:54:40 +01:00
return me->edit_btmesh ? me->edit_btmesh->bm->totedge : me->totedge;
Convert MBC_ API to Mesh (instead of derived mesh) and move it to mesh_render This includes a few fixes in the MBC_ api. The idea here is for this to be the only interface the render engines will deal with for the meshes. If we need to expose special options for sculpting engine we refactor this accordingly. But for now we are shaping this in a per-case base. Note: * We still need to hook up to the depsgraph to force clear/update of batch_cache when mesh changes (I'm waiting for Sergey Sharybin's depsgraph update for this though) * Also ideally we could/should use BMesh directly instead of DerivedMesh, but this will do for now. Note 2: In the end I renamed the `BKE_mesh_render` functions to `static mesh_render`. We can re-expose them as BKE_* later once we need it. Reviewers: merwin Subscribers: fclem Differential Revision: https://developer.blender.org/D2476
2017-01-25 09:16:29 +01:00
}
Cleanup: replace num/ct/tot with 'len'
2017-04-19 00:10:32 +10:00
static int mesh_render_looptri_len_get(Mesh *me)
Convert MBC_ API to Mesh (instead of derived mesh) and move it to mesh_render This includes a few fixes in the MBC_ api. The idea here is for this to be the only interface the render engines will deal with for the meshes. If we need to expose special options for sculpting engine we refactor this accordingly. But for now we are shaping this in a per-case base. Note: * We still need to hook up to the depsgraph to force clear/update of batch_cache when mesh changes (I'm waiting for Sergey Sharybin's depsgraph update for this though) * Also ideally we could/should use BMesh directly instead of DerivedMesh, but this will do for now. Note 2: In the end I renamed the `BKE_mesh_render` functions to `static mesh_render`. We can re-expose them as BKE_* later once we need it. Reviewers: merwin Subscribers: fclem Differential Revision: https://developer.blender.org/D2476
2017-01-25 09:16:29 +01:00
{
Rework of BKE's mesh_render to support BMesh directly. Note that since there is no (efficient) ways to get arrays of MVert/MEdge/etc. out of a BMesh, I refactored quite heavily internals of BKE_mesh_render. Now, when you do need acess to mesh data to generate cached batches, you create an abstract struct from mesh (either Mesh or BMesh if available), and then use advanced helpers to extract needed data, on a per-item basis (no more handling of arrays of verts/edges/... in batches code). This allows to: * Avoid having to create arrays of BMesh elements. * Take advantage of existing advanced BMesh topology and connectivity data. Reviewers: dfelinto, fclem, merwin Differential Revision: https://developer.blender.org/D2521
2017-02-23 11:54:40 +01:00
return me->edit_btmesh ? me->edit_btmesh->tottri : poly_to_tri_count(me->totpoly, me->totloop);
Convert MBC_ API to Mesh (instead of derived mesh) and move it to mesh_render This includes a few fixes in the MBC_ api. The idea here is for this to be the only interface the render engines will deal with for the meshes. If we need to expose special options for sculpting engine we refactor this accordingly. But for now we are shaping this in a per-case base. Note: * We still need to hook up to the depsgraph to force clear/update of batch_cache when mesh changes (I'm waiting for Sergey Sharybin's depsgraph update for this though) * Also ideally we could/should use BMesh directly instead of DerivedMesh, but this will do for now. Note 2: In the end I renamed the `BKE_mesh_render` functions to `static mesh_render`. We can re-expose them as BKE_* later once we need it. Reviewers: merwin Subscribers: fclem Differential Revision: https://developer.blender.org/D2476
2017-01-25 09:16:29 +01:00
}
Cleanup: replace num/ct/tot with 'len'
2017-04-19 00:10:32 +10:00
static int mesh_render_polys_len_get(Mesh *me)
Convert MBC_ API to Mesh (instead of derived mesh) and move it to mesh_render This includes a few fixes in the MBC_ api. The idea here is for this to be the only interface the render engines will deal with for the meshes. If we need to expose special options for sculpting engine we refactor this accordingly. But for now we are shaping this in a per-case base. Note: * We still need to hook up to the depsgraph to force clear/update of batch_cache when mesh changes (I'm waiting for Sergey Sharybin's depsgraph update for this though) * Also ideally we could/should use BMesh directly instead of DerivedMesh, but this will do for now. Note 2: In the end I renamed the `BKE_mesh_render` functions to `static mesh_render`. We can re-expose them as BKE_* later once we need it. Reviewers: merwin Subscribers: fclem Differential Revision: https://developer.blender.org/D2476
2017-01-25 09:16:29 +01:00
{
Rework of BKE's mesh_render to support BMesh directly. Note that since there is no (efficient) ways to get arrays of MVert/MEdge/etc. out of a BMesh, I refactored quite heavily internals of BKE_mesh_render. Now, when you do need acess to mesh data to generate cached batches, you create an abstract struct from mesh (either Mesh or BMesh if available), and then use advanced helpers to extract needed data, on a per-item basis (no more handling of arrays of verts/edges/... in batches code). This allows to: * Avoid having to create arrays of BMesh elements. * Take advantage of existing advanced BMesh topology and connectivity data. Reviewers: dfelinto, fclem, merwin Differential Revision: https://developer.blender.org/D2521
2017-02-23 11:54:40 +01:00
return me->edit_btmesh ? me->edit_btmesh->bm->totface : me->totpoly;
Convert MBC_ API to Mesh (instead of derived mesh) and move it to mesh_render This includes a few fixes in the MBC_ api. The idea here is for this to be the only interface the render engines will deal with for the meshes. If we need to expose special options for sculpting engine we refactor this accordingly. But for now we are shaping this in a per-case base. Note: * We still need to hook up to the depsgraph to force clear/update of batch_cache when mesh changes (I'm waiting for Sergey Sharybin's depsgraph update for this though) * Also ideally we could/should use BMesh directly instead of DerivedMesh, but this will do for now. Note 2: In the end I renamed the `BKE_mesh_render` functions to `static mesh_render`. We can re-expose them as BKE_* later once we need it. Reviewers: merwin Subscribers: fclem Differential Revision: https://developer.blender.org/D2476
2017-01-25 09:16:29 +01:00
}
Eevee: Make default shaders works. - Added New Batch cache routine to get the split mesh for each material. (optimization needed) - Did some shader modifications to create default_frag and support a somwhat correct fresnel for lamps (to finish)
2017-04-25 18:46:59 +02:00
static int mesh_render_mat_ct_get(Mesh *me)
{
return MAX2(1, me->totcol);
}
Cleanup: replace num/ct/tot with 'len'
2017-04-19 00:10:32 +10:00
static int UNUSED_FUNCTION(mesh_render_loops_len_get)(Mesh *me)
Clay Engine: Correct Mesh Normals
2017-02-16 16:19:48 +01:00
{
Rework of BKE's mesh_render to support BMesh directly. Note that since there is no (efficient) ways to get arrays of MVert/MEdge/etc. out of a BMesh, I refactored quite heavily internals of BKE_mesh_render. Now, when you do need acess to mesh data to generate cached batches, you create an abstract struct from mesh (either Mesh or BMesh if available), and then use advanced helpers to extract needed data, on a per-item basis (no more handling of arrays of verts/edges/... in batches code). This allows to: * Avoid having to create arrays of BMesh elements. * Take advantage of existing advanced BMesh topology and connectivity data. Reviewers: dfelinto, fclem, merwin Differential Revision: https://developer.blender.org/D2521
2017-02-23 11:54:40 +01:00
return me->edit_btmesh ? me->edit_btmesh->bm->totloop : me->totloop;
Clay Engine: Correct Mesh Normals
2017-02-16 16:19:48 +01:00
}
Rework of BKE's mesh_render to support BMesh directly. Note that since there is no (efficient) ways to get arrays of MVert/MEdge/etc. out of a BMesh, I refactored quite heavily internals of BKE_mesh_render. Now, when you do need acess to mesh data to generate cached batches, you create an abstract struct from mesh (either Mesh or BMesh if available), and then use advanced helpers to extract needed data, on a per-item basis (no more handling of arrays of verts/edges/... in batches code). This allows to: * Avoid having to create arrays of BMesh elements. * Take advantage of existing advanced BMesh topology and connectivity data. Reviewers: dfelinto, fclem, merwin Differential Revision: https://developer.blender.org/D2521
2017-02-23 11:54:40 +01:00
/* ---------------------------------------------------------------------- */
/* Mesh/BMesh Interface, indirect, partially cached access to complex data. */
typedef struct EdgeAdjacentPolys {
int count;
int face_index[2];
} EdgeAdjacentPolys;
Cleanup: replace char array with struct
2017-04-12 14:08:44 +10:00
typedef struct EdgeDrawAttr {
unsigned char v_flag;
unsigned char e_flag;
unsigned char crease;
unsigned char bweight;
} EdgeDrawAttr;
Rework of BKE's mesh_render to support BMesh directly. Note that since there is no (efficient) ways to get arrays of MVert/MEdge/etc. out of a BMesh, I refactored quite heavily internals of BKE_mesh_render. Now, when you do need acess to mesh data to generate cached batches, you create an abstract struct from mesh (either Mesh or BMesh if available), and then use advanced helpers to extract needed data, on a per-item basis (no more handling of arrays of verts/edges/... in batches code). This allows to: * Avoid having to create arrays of BMesh elements. * Take advantage of existing advanced BMesh topology and connectivity data. Reviewers: dfelinto, fclem, merwin Differential Revision: https://developer.blender.org/D2521
2017-02-23 11:54:40 +01:00
typedef struct MeshRenderData {
int types;
Cleanup: replace num/ct/tot with 'len'
2017-04-19 00:10:32 +10:00
int vert_len;
int edge_len;
int tri_len;
int loop_len;
int poly_len;
Eevee: Make default shaders works. - Added New Batch cache routine to get the split mesh for each material. (optimization needed) - Did some shader modifications to create default_frag and support a somwhat correct fresnel for lamps (to finish)
2017-04-25 18:46:59 +02:00
int mat_ct;
Cleanup: replace num/ct/tot with 'len'
2017-04-19 00:10:32 +10:00
int loose_vert_len;
int loose_edge_len;
Rework of BKE's mesh_render to support BMesh directly. Note that since there is no (efficient) ways to get arrays of MVert/MEdge/etc. out of a BMesh, I refactored quite heavily internals of BKE_mesh_render. Now, when you do need acess to mesh data to generate cached batches, you create an abstract struct from mesh (either Mesh or BMesh if available), and then use advanced helpers to extract needed data, on a per-item basis (no more handling of arrays of verts/edges/... in batches code). This allows to: * Avoid having to create arrays of BMesh elements. * Take advantage of existing advanced BMesh topology and connectivity data. Reviewers: dfelinto, fclem, merwin Differential Revision: https://developer.blender.org/D2521
2017-02-23 11:54:40 +01:00
BMEditMesh *edit_bmesh;
MVert *mvert;
MEdge *medge;
MLoop *mloop;
MPoly *mpoly;
Mesh Batch Cache: Add support for uv/vcol/tangent data layers. This is a bad implementation waiting for some improvement that are : - Gawain support for multiple names. -> will remove duplication of attribute alloc - Glsl safe name string escaping. -> will remove any risk of hash colision. Old glsl drawing was relying on attribute being choosen by DerivedMesh at drawing time. For this reason, we declare all possible attrib "name" for each data layer inside the batches. In the glsl code, we declare required data by type and name. Then Gawain only bind vbos if names correspond. This is way cleaner as we do not need to access the CD itself when drawing. One other problem is that the hash maybe 11 caracters long and rapidly overload gawain's attrib name buffer.
2017-05-01 17:58:15 +02:00
float (*orco)[3];
MLoopUV **mloopuv;
MCol **mcol;
float (**mtangent)[4];
Rework of BKE's mesh_render to support BMesh directly. Note that since there is no (efficient) ways to get arrays of MVert/MEdge/etc. out of a BMesh, I refactored quite heavily internals of BKE_mesh_render. Now, when you do need acess to mesh data to generate cached batches, you create an abstract struct from mesh (either Mesh or BMesh if available), and then use advanced helpers to extract needed data, on a per-item basis (no more handling of arrays of verts/edges/... in batches code). This allows to: * Avoid having to create arrays of BMesh elements. * Take advantage of existing advanced BMesh topology and connectivity data. Reviewers: dfelinto, fclem, merwin Differential Revision: https://developer.blender.org/D2521
2017-02-23 11:54:40 +01:00
Clay Engine: Edit mesh overlays Based on the previous overlay shader from merwin. This shader takes care of clipped vertex cases and do all edit mode face info in one pass (except face centers). As the shading is done one the triangle itself the visual can't go beyond the surface of the mesh. That leads to half displayed edges on the outline of the mesh. This problem can be fixed by a second pass. This is work in progress.
2017-02-26 21:07:37 +01:00
BMVert *eve_act;
BMEdge *eed_act;
BMFace *efa_act;
Mesh Batch Cache: Add support for uv/vcol/tangent data layers. This is a bad implementation waiting for some improvement that are : - Gawain support for multiple names. -> will remove duplication of attribute alloc - Glsl safe name string escaping. -> will remove any risk of hash colision. Old glsl drawing was relying on attribute being choosen by DerivedMesh at drawing time. For this reason, we declare all possible attrib "name" for each data layer inside the batches. In the glsl code, we declare required data by type and name. Then Gawain only bind vbos if names correspond. This is way cleaner as we do not need to access the CD itself when drawing. One other problem is that the hash maybe 11 caracters long and rapidly overload gawain's attrib name buffer.
2017-05-01 17:58:15 +02:00
int uv_ct;
int vcol_ct;
bool *auto_vcol;
int uv_active;
int vcol_active;
int tangent_active;
Clay Engine: Edit mesh overlays Based on the previous overlay shader from merwin. This shader takes care of clipped vertex cases and do all edit mode face info in one pass (except face centers). As the shading is done one the triangle itself the visual can't go beyond the surface of the mesh. That leads to half displayed edges on the outline of the mesh. This problem can be fixed by a second pass. This is work in progress.
2017-02-26 21:07:37 +01:00
int crease_ofs;
int bweight_ofs;
Mesh Batch Cache: Add support for uv/vcol/tangent data layers. This is a bad implementation waiting for some improvement that are : - Gawain support for multiple names. -> will remove duplication of attribute alloc - Glsl safe name string escaping. -> will remove any risk of hash colision. Old glsl drawing was relying on attribute being choosen by DerivedMesh at drawing time. For this reason, we declare all possible attrib "name" for each data layer inside the batches. In the glsl code, we declare required data by type and name. Then Gawain only bind vbos if names correspond. This is way cleaner as we do not need to access the CD itself when drawing. One other problem is that the hash maybe 11 caracters long and rapidly overload gawain's attrib name buffer.
2017-05-01 17:58:15 +02:00
int *uv_ofs;
int *vcol_ofs;
int *tangent_ofs;
char (*auto_names)[32];
char (*uv_names)[32];
char (*vcol_names)[32];
char (*tangent_names)[32];
Clay Engine: Edit mesh overlays Based on the previous overlay shader from merwin. This shader takes care of clipped vertex cases and do all edit mode face info in one pass (except face centers). As the shading is done one the triangle itself the visual can't go beyond the surface of the mesh. That leads to half displayed edges on the outline of the mesh. This problem can be fixed by a second pass. This is work in progress.
2017-02-26 21:07:37 +01:00
Rework of BKE's mesh_render to support BMesh directly. Note that since there is no (efficient) ways to get arrays of MVert/MEdge/etc. out of a BMesh, I refactored quite heavily internals of BKE_mesh_render. Now, when you do need acess to mesh data to generate cached batches, you create an abstract struct from mesh (either Mesh or BMesh if available), and then use advanced helpers to extract needed data, on a per-item basis (no more handling of arrays of verts/edges/... in batches code). This allows to: * Avoid having to create arrays of BMesh elements. * Take advantage of existing advanced BMesh topology and connectivity data. Reviewers: dfelinto, fclem, merwin Differential Revision: https://developer.blender.org/D2521
2017-02-23 11:54:40 +01:00
/* Data created on-demand (usually not for bmesh-based data). */
EdgeAdjacentPolys *edges_adjacent_polys;
MLoopTri *mlooptri;
Edit Mesh overlays: added looses edges and verts as triangle. We loose a bit of memory for (more?) performance.
2017-02-28 14:30:34 +01:00
int *loose_edges;
int *loose_verts;
Rework of BKE's mesh_render to support BMesh directly. Note that since there is no (efficient) ways to get arrays of MVert/MEdge/etc. out of a BMesh, I refactored quite heavily internals of BKE_mesh_render. Now, when you do need acess to mesh data to generate cached batches, you create an abstract struct from mesh (either Mesh or BMesh if available), and then use advanced helpers to extract needed data, on a per-item basis (no more handling of arrays of verts/edges/... in batches code). This allows to: * Avoid having to create arrays of BMesh elements. * Take advantage of existing advanced BMesh topology and connectivity data. Reviewers: dfelinto, fclem, merwin Differential Revision: https://developer.blender.org/D2521
2017-02-23 11:54:40 +01:00
float (*poly_normals)[3];
short (*poly_normals_short)[3];
short (*vert_normals_short)[3];
} MeshRenderData;
enum {
Edit Mesh overlays: added looses edges and verts as triangle. We loose a bit of memory for (more?) performance.
2017-02-28 14:30:34 +01:00
MR_DATATYPE_VERT = 1 << 0,
Edit Mode overlays: separate multiple shaders for loose edges and verts
2017-03-01 14:07:04 +01:00
MR_DATATYPE_EDGE = 1 << 1,
MR_DATATYPE_LOOPTRI = 1 << 2,
MR_DATATYPE_LOOP = 1 << 3,
MR_DATATYPE_POLY = 1 << 4,
MR_DATATYPE_OVERLAY = 1 << 5,
Mesh Batch Cache: Add support for uv/vcol/tangent data layers. This is a bad implementation waiting for some improvement that are : - Gawain support for multiple names. -> will remove duplication of attribute alloc - Glsl safe name string escaping. -> will remove any risk of hash colision. Old glsl drawing was relying on attribute being choosen by DerivedMesh at drawing time. For this reason, we declare all possible attrib "name" for each data layer inside the batches. In the glsl code, we declare required data by type and name. Then Gawain only bind vbos if names correspond. This is way cleaner as we do not need to access the CD itself when drawing. One other problem is that the hash maybe 11 caracters long and rapidly overload gawain's attrib name buffer.
2017-05-01 17:58:15 +02:00
MR_DATATYPE_SHADING = 1 << 6,
Rework of BKE's mesh_render to support BMesh directly. Note that since there is no (efficient) ways to get arrays of MVert/MEdge/etc. out of a BMesh, I refactored quite heavily internals of BKE_mesh_render. Now, when you do need acess to mesh data to generate cached batches, you create an abstract struct from mesh (either Mesh or BMesh if available), and then use advanced helpers to extract needed data, on a per-item basis (no more handling of arrays of verts/edges/... in batches code). This allows to: * Avoid having to create arrays of BMesh elements. * Take advantage of existing advanced BMesh topology and connectivity data. Reviewers: dfelinto, fclem, merwin Differential Revision: https://developer.blender.org/D2521
2017-02-23 11:54:40 +01:00
};
Draw Manager: T51218 Support edit-mesh hide
2017-04-18 23:29:55 +10:00
/**
* These functions look like they would be slow but they will typically return true on the first iteration.
* Only false when all attached elements are hidden.
*/
static bool bm_vert_has_visible_edge(const BMVert *v)
{
const BMEdge *e_iter, *e_first;
e_iter = e_first = v->e;
do {
if (!BM_elem_flag_test(e_iter, BM_ELEM_HIDDEN)) {
return true;
}
} while ((e_iter = BM_DISK_EDGE_NEXT(e_iter, v)) != e_first);
return false;
}
static bool bm_edge_has_visible_face(const BMEdge *e)
{
const BMLoop *l_iter, *l_first;
l_iter = l_first = e->l;
do {
if (!BM_elem_flag_test(l_iter->f, BM_ELEM_HIDDEN)) {
return true;
}
} while ((l_iter = l_iter->radial_next) != l_first);
return false;
}
Rework of BKE's mesh_render to support BMesh directly. Note that since there is no (efficient) ways to get arrays of MVert/MEdge/etc. out of a BMesh, I refactored quite heavily internals of BKE_mesh_render. Now, when you do need acess to mesh data to generate cached batches, you create an abstract struct from mesh (either Mesh or BMesh if available), and then use advanced helpers to extract needed data, on a per-item basis (no more handling of arrays of verts/edges/... in batches code). This allows to: * Avoid having to create arrays of BMesh elements. * Take advantage of existing advanced BMesh topology and connectivity data. Reviewers: dfelinto, fclem, merwin Differential Revision: https://developer.blender.org/D2521
2017-02-23 11:54:40 +01:00
static MeshRenderData *mesh_render_data_create(Mesh *me, const int types)
{
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
MeshRenderData *rdata = MEM_callocN(sizeof(*rdata), __func__);
rdata->types = types;
Eevee: Make default shaders works. - Added New Batch cache routine to get the split mesh for each material. (optimization needed) - Did some shader modifications to create default_frag and support a somwhat correct fresnel for lamps (to finish)
2017-04-25 18:46:59 +02:00
rdata->mat_ct = mesh_render_mat_ct_get(me);
Rework of BKE's mesh_render to support BMesh directly. Note that since there is no (efficient) ways to get arrays of MVert/MEdge/etc. out of a BMesh, I refactored quite heavily internals of BKE_mesh_render. Now, when you do need acess to mesh data to generate cached batches, you create an abstract struct from mesh (either Mesh or BMesh if available), and then use advanced helpers to extract needed data, on a per-item basis (no more handling of arrays of verts/edges/... in batches code). This allows to: * Avoid having to create arrays of BMesh elements. * Take advantage of existing advanced BMesh topology and connectivity data. Reviewers: dfelinto, fclem, merwin Differential Revision: https://developer.blender.org/D2521
2017-02-23 11:54:40 +01:00
if (me->edit_btmesh) {
BMEditMesh *embm = me->edit_btmesh;
BMesh *bm = embm->bm;
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
rdata->edit_bmesh = embm;
Rework of BKE's mesh_render to support BMesh directly. Note that since there is no (efficient) ways to get arrays of MVert/MEdge/etc. out of a BMesh, I refactored quite heavily internals of BKE_mesh_render. Now, when you do need acess to mesh data to generate cached batches, you create an abstract struct from mesh (either Mesh or BMesh if available), and then use advanced helpers to extract needed data, on a per-item basis (no more handling of arrays of verts/edges/... in batches code). This allows to: * Avoid having to create arrays of BMesh elements. * Take advantage of existing advanced BMesh topology and connectivity data. Reviewers: dfelinto, fclem, merwin Differential Revision: https://developer.blender.org/D2521
2017-02-23 11:54:40 +01:00
int bm_ensure_types = 0;
Edit Mode overlays: separate multiple shaders for loose edges and verts
2017-03-01 14:07:04 +01:00
if (types & (MR_DATATYPE_VERT)) {
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
rdata->vert_len = bm->totvert;
Rework of BKE's mesh_render to support BMesh directly. Note that since there is no (efficient) ways to get arrays of MVert/MEdge/etc. out of a BMesh, I refactored quite heavily internals of BKE_mesh_render. Now, when you do need acess to mesh data to generate cached batches, you create an abstract struct from mesh (either Mesh or BMesh if available), and then use advanced helpers to extract needed data, on a per-item basis (no more handling of arrays of verts/edges/... in batches code). This allows to: * Avoid having to create arrays of BMesh elements. * Take advantage of existing advanced BMesh topology and connectivity data. Reviewers: dfelinto, fclem, merwin Differential Revision: https://developer.blender.org/D2521
2017-02-23 11:54:40 +01:00
bm_ensure_types |= BM_VERT;
}
Edit Mode overlays: separate multiple shaders for loose edges and verts
2017-03-01 14:07:04 +01:00
if (types & (MR_DATATYPE_EDGE)) {
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
rdata->edge_len = bm->totedge;
Rework of BKE's mesh_render to support BMesh directly. Note that since there is no (efficient) ways to get arrays of MVert/MEdge/etc. out of a BMesh, I refactored quite heavily internals of BKE_mesh_render. Now, when you do need acess to mesh data to generate cached batches, you create an abstract struct from mesh (either Mesh or BMesh if available), and then use advanced helpers to extract needed data, on a per-item basis (no more handling of arrays of verts/edges/... in batches code). This allows to: * Avoid having to create arrays of BMesh elements. * Take advantage of existing advanced BMesh topology and connectivity data. Reviewers: dfelinto, fclem, merwin Differential Revision: https://developer.blender.org/D2521
2017-02-23 11:54:40 +01:00
bm_ensure_types |= BM_EDGE;
}
if (types & MR_DATATYPE_LOOPTRI) {
BKE_editmesh_tessface_calc(embm);
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
rdata->tri_len = embm->tottri;
Rework of BKE's mesh_render to support BMesh directly. Note that since there is no (efficient) ways to get arrays of MVert/MEdge/etc. out of a BMesh, I refactored quite heavily internals of BKE_mesh_render. Now, when you do need acess to mesh data to generate cached batches, you create an abstract struct from mesh (either Mesh or BMesh if available), and then use advanced helpers to extract needed data, on a per-item basis (no more handling of arrays of verts/edges/... in batches code). This allows to: * Avoid having to create arrays of BMesh elements. * Take advantage of existing advanced BMesh topology and connectivity data. Reviewers: dfelinto, fclem, merwin Differential Revision: https://developer.blender.org/D2521
2017-02-23 11:54:40 +01:00
}
if (types & MR_DATATYPE_LOOP) {
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
rdata->loop_len = bm->totloop;
Rework of BKE's mesh_render to support BMesh directly. Note that since there is no (efficient) ways to get arrays of MVert/MEdge/etc. out of a BMesh, I refactored quite heavily internals of BKE_mesh_render. Now, when you do need acess to mesh data to generate cached batches, you create an abstract struct from mesh (either Mesh or BMesh if available), and then use advanced helpers to extract needed data, on a per-item basis (no more handling of arrays of verts/edges/... in batches code). This allows to: * Avoid having to create arrays of BMesh elements. * Take advantage of existing advanced BMesh topology and connectivity data. Reviewers: dfelinto, fclem, merwin Differential Revision: https://developer.blender.org/D2521
2017-02-23 11:54:40 +01:00
bm_ensure_types |= BM_LOOP;
}
if (types & MR_DATATYPE_POLY) {
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
rdata->poly_len = bm->totface;
Rework of BKE's mesh_render to support BMesh directly. Note that since there is no (efficient) ways to get arrays of MVert/MEdge/etc. out of a BMesh, I refactored quite heavily internals of BKE_mesh_render. Now, when you do need acess to mesh data to generate cached batches, you create an abstract struct from mesh (either Mesh or BMesh if available), and then use advanced helpers to extract needed data, on a per-item basis (no more handling of arrays of verts/edges/... in batches code). This allows to: * Avoid having to create arrays of BMesh elements. * Take advantage of existing advanced BMesh topology and connectivity data. Reviewers: dfelinto, fclem, merwin Differential Revision: https://developer.blender.org/D2521
2017-02-23 11:54:40 +01:00
bm_ensure_types |= BM_FACE;
}
Edit Mode overlays: separate multiple shaders for loose edges and verts
2017-03-01 14:07:04 +01:00
if (types & MR_DATATYPE_OVERLAY) {
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
rdata->efa_act = BM_mesh_active_face_get(bm, false, true);
rdata->eed_act = BM_mesh_active_edge_get(bm);
rdata->eve_act = BM_mesh_active_vert_get(bm);
rdata->crease_ofs = CustomData_get_offset(&bm->edata, CD_CREASE);
rdata->bweight_ofs = CustomData_get_offset(&bm->edata, CD_BWEIGHT);
Clay Engine: Edit mesh overlays Based on the previous overlay shader from merwin. This shader takes care of clipped vertex cases and do all edit mode face info in one pass (except face centers). As the shading is done one the triangle itself the visual can't go beyond the surface of the mesh. That leads to half displayed edges on the outline of the mesh. This problem can be fixed by a second pass. This is work in progress.
2017-02-26 21:07:37 +01:00
}
Rework of BKE's mesh_render to support BMesh directly. Note that since there is no (efficient) ways to get arrays of MVert/MEdge/etc. out of a BMesh, I refactored quite heavily internals of BKE_mesh_render. Now, when you do need acess to mesh data to generate cached batches, you create an abstract struct from mesh (either Mesh or BMesh if available), and then use advanced helpers to extract needed data, on a per-item basis (no more handling of arrays of verts/edges/... in batches code). This allows to: * Avoid having to create arrays of BMesh elements. * Take advantage of existing advanced BMesh topology and connectivity data. Reviewers: dfelinto, fclem, merwin Differential Revision: https://developer.blender.org/D2521
2017-02-23 11:54:40 +01:00
BM_mesh_elem_index_ensure(bm, bm_ensure_types);
BM_mesh_elem_table_ensure(bm, bm_ensure_types & ~BM_LOOP);
Edit Mode overlays: separate multiple shaders for loose edges and verts
2017-03-01 14:07:04 +01:00
if (types & MR_DATATYPE_OVERLAY) {
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
rdata->loose_vert_len = rdata->loose_edge_len = 0;
Edit Mode overlays: separate multiple shaders for loose edges and verts
2017-03-01 14:07:04 +01:00
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
int *lverts = rdata->loose_verts = MEM_mallocN(rdata->vert_len * sizeof(int), "Loose Vert");
int *ledges = rdata->loose_edges = MEM_mallocN(rdata->edge_len * sizeof(int), "Loose Edges");
Edit Mode overlays: separate multiple shaders for loose edges and verts
2017-03-01 14:07:04 +01:00
EditMesh Render: remove edges-of-vert loop Replace topology check which checked edges twice with a loop on the edge array.
2017-04-12 15:42:41 +10:00
{
BLI_assert((bm->elem_table_dirty & BM_VERT) == 0);
BMVert **vtable = bm->vtable;
for (int i = 0; i < bm->totvert; i++) {
const BMVert *v = vtable[i];
Draw Manager: T51218 Support edit-mesh hide
2017-04-18 23:29:55 +10:00
if (!BM_elem_flag_test(v, BM_ELEM_HIDDEN)) {
/* Loose vert */
if (v->e == NULL || !bm_vert_has_visible_edge(v)) {
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
lverts[rdata->loose_vert_len++] = i;
Draw Manager: T51218 Support edit-mesh hide
2017-04-18 23:29:55 +10:00
}
EditMesh Render: remove edges-of-vert loop Replace topology check which checked edges twice with a loop on the edge array.
2017-04-12 15:42:41 +10:00
}
Edit Mesh overlays: added looses edges and verts as triangle. We loose a bit of memory for (more?) performance.
2017-02-28 14:30:34 +01:00
}
EditMesh Render: remove edges-of-vert loop Replace topology check which checked edges twice with a loop on the edge array.
2017-04-12 15:42:41 +10:00
}
Edit Mesh overlays: added looses edges and verts as triangle. We loose a bit of memory for (more?) performance.
2017-02-28 14:30:34 +01:00
EditMesh Render: remove edges-of-vert loop Replace topology check which checked edges twice with a loop on the edge array.
2017-04-12 15:42:41 +10:00
{
BLI_assert((bm->elem_table_dirty & BM_EDGE) == 0);
BMEdge **etable = bm->etable;
for (int i = 0; i < bm->totedge; i++) {
const BMEdge *e = etable[i];
Draw Manager: T51218 Support edit-mesh hide
2017-04-18 23:29:55 +10:00
if (!BM_elem_flag_test(e, BM_ELEM_HIDDEN)) {
/* Loose edge */
if (e->l == NULL || !bm_edge_has_visible_face(e)) {
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
ledges[rdata->loose_edge_len++] = i;
Draw Manager: T51218 Support edit-mesh hide
2017-04-18 23:29:55 +10:00
}
Edit Mesh overlays: added looses edges and verts as triangle. We loose a bit of memory for (more?) performance.
2017-02-28 14:30:34 +01:00
}
EditMesh Render: remove edges-of-vert loop Replace topology check which checked edges twice with a loop on the edge array.
2017-04-12 15:42:41 +10:00
}
Edit Mesh overlays: added looses edges and verts as triangle. We loose a bit of memory for (more?) performance.
2017-02-28 14:30:34 +01:00
}
EditMesh Render: remove edges-of-vert loop Replace topology check which checked edges twice with a loop on the edge array.
2017-04-12 15:42:41 +10:00
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
rdata->loose_verts = MEM_reallocN(rdata->loose_verts, rdata->loose_vert_len * sizeof(int));
rdata->loose_edges = MEM_reallocN(rdata->loose_edges, rdata->loose_edge_len * sizeof(int));
Edit Mesh overlays: added looses edges and verts as triangle. We loose a bit of memory for (more?) performance.
2017-02-28 14:30:34 +01:00
}
Rework of BKE's mesh_render to support BMesh directly. Note that since there is no (efficient) ways to get arrays of MVert/MEdge/etc. out of a BMesh, I refactored quite heavily internals of BKE_mesh_render. Now, when you do need acess to mesh data to generate cached batches, you create an abstract struct from mesh (either Mesh or BMesh if available), and then use advanced helpers to extract needed data, on a per-item basis (no more handling of arrays of verts/edges/... in batches code). This allows to: * Avoid having to create arrays of BMesh elements. * Take advantage of existing advanced BMesh topology and connectivity data. Reviewers: dfelinto, fclem, merwin Differential Revision: https://developer.blender.org/D2521
2017-02-23 11:54:40 +01:00
}
else {
Edit Mode overlays: separate multiple shaders for loose edges and verts
2017-03-01 14:07:04 +01:00
if (types & (MR_DATATYPE_VERT)) {
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
rdata->vert_len = me->totvert;
rdata->mvert = CustomData_get_layer(&me->vdata, CD_MVERT);
Rework of BKE's mesh_render to support BMesh directly. Note that since there is no (efficient) ways to get arrays of MVert/MEdge/etc. out of a BMesh, I refactored quite heavily internals of BKE_mesh_render. Now, when you do need acess to mesh data to generate cached batches, you create an abstract struct from mesh (either Mesh or BMesh if available), and then use advanced helpers to extract needed data, on a per-item basis (no more handling of arrays of verts/edges/... in batches code). This allows to: * Avoid having to create arrays of BMesh elements. * Take advantage of existing advanced BMesh topology and connectivity data. Reviewers: dfelinto, fclem, merwin Differential Revision: https://developer.blender.org/D2521
2017-02-23 11:54:40 +01:00
}
Edit Mode overlays: separate multiple shaders for loose edges and verts
2017-03-01 14:07:04 +01:00
if (types & (MR_DATATYPE_EDGE)) {
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
rdata->edge_len = me->totedge;
rdata->medge = CustomData_get_layer(&me->edata, CD_MEDGE);
Rework of BKE's mesh_render to support BMesh directly. Note that since there is no (efficient) ways to get arrays of MVert/MEdge/etc. out of a BMesh, I refactored quite heavily internals of BKE_mesh_render. Now, when you do need acess to mesh data to generate cached batches, you create an abstract struct from mesh (either Mesh or BMesh if available), and then use advanced helpers to extract needed data, on a per-item basis (no more handling of arrays of verts/edges/... in batches code). This allows to: * Avoid having to create arrays of BMesh elements. * Take advantage of existing advanced BMesh topology and connectivity data. Reviewers: dfelinto, fclem, merwin Differential Revision: https://developer.blender.org/D2521
2017-02-23 11:54:40 +01:00
}
if (types & MR_DATATYPE_LOOPTRI) {
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
const int tri_len = rdata->tri_len = poly_to_tri_count(me->totpoly, me->totloop);
rdata->mlooptri = MEM_mallocN(sizeof(*rdata->mlooptri) * tri_len, __func__);
BKE_mesh_recalc_looptri(me->mloop, me->mpoly, me->mvert, me->totloop, me->totpoly, rdata->mlooptri);
Rework of BKE's mesh_render to support BMesh directly. Note that since there is no (efficient) ways to get arrays of MVert/MEdge/etc. out of a BMesh, I refactored quite heavily internals of BKE_mesh_render. Now, when you do need acess to mesh data to generate cached batches, you create an abstract struct from mesh (either Mesh or BMesh if available), and then use advanced helpers to extract needed data, on a per-item basis (no more handling of arrays of verts/edges/... in batches code). This allows to: * Avoid having to create arrays of BMesh elements. * Take advantage of existing advanced BMesh topology and connectivity data. Reviewers: dfelinto, fclem, merwin Differential Revision: https://developer.blender.org/D2521
2017-02-23 11:54:40 +01:00
}
if (types & MR_DATATYPE_LOOP) {
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
rdata->loop_len = me->totloop;
rdata->mloop = CustomData_get_layer(&me->ldata, CD_MLOOP);
Rework of BKE's mesh_render to support BMesh directly. Note that since there is no (efficient) ways to get arrays of MVert/MEdge/etc. out of a BMesh, I refactored quite heavily internals of BKE_mesh_render. Now, when you do need acess to mesh data to generate cached batches, you create an abstract struct from mesh (either Mesh or BMesh if available), and then use advanced helpers to extract needed data, on a per-item basis (no more handling of arrays of verts/edges/... in batches code). This allows to: * Avoid having to create arrays of BMesh elements. * Take advantage of existing advanced BMesh topology and connectivity data. Reviewers: dfelinto, fclem, merwin Differential Revision: https://developer.blender.org/D2521
2017-02-23 11:54:40 +01:00
}
if (types & MR_DATATYPE_POLY) {
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
rdata->poly_len = me->totpoly;
rdata->mpoly = CustomData_get_layer(&me->pdata, CD_MPOLY);
Rework of BKE's mesh_render to support BMesh directly. Note that since there is no (efficient) ways to get arrays of MVert/MEdge/etc. out of a BMesh, I refactored quite heavily internals of BKE_mesh_render. Now, when you do need acess to mesh data to generate cached batches, you create an abstract struct from mesh (either Mesh or BMesh if available), and then use advanced helpers to extract needed data, on a per-item basis (no more handling of arrays of verts/edges/... in batches code). This allows to: * Avoid having to create arrays of BMesh elements. * Take advantage of existing advanced BMesh topology and connectivity data. Reviewers: dfelinto, fclem, merwin Differential Revision: https://developer.blender.org/D2521
2017-02-23 11:54:40 +01:00
}
}
Convert MBC_ API to Mesh (instead of derived mesh) and move it to mesh_render This includes a few fixes in the MBC_ api. The idea here is for this to be the only interface the render engines will deal with for the meshes. If we need to expose special options for sculpting engine we refactor this accordingly. But for now we are shaping this in a per-case base. Note: * We still need to hook up to the depsgraph to force clear/update of batch_cache when mesh changes (I'm waiting for Sergey Sharybin's depsgraph update for this though) * Also ideally we could/should use BMesh directly instead of DerivedMesh, but this will do for now. Note 2: In the end I renamed the `BKE_mesh_render` functions to `static mesh_render`. We can re-expose them as BKE_* later once we need it. Reviewers: merwin Subscribers: fclem Differential Revision: https://developer.blender.org/D2476
2017-01-25 09:16:29 +01:00
Mesh Batch Cache: Add support for uv/vcol/tangent data layers. This is a bad implementation waiting for some improvement that are : - Gawain support for multiple names. -> will remove duplication of attribute alloc - Glsl safe name string escaping. -> will remove any risk of hash colision. Old glsl drawing was relying on attribute being choosen by DerivedMesh at drawing time. For this reason, we declare all possible attrib "name" for each data layer inside the batches. In the glsl code, we declare required data by type and name. Then Gawain only bind vbos if names correspond. This is way cleaner as we do not need to access the CD itself when drawing. One other problem is that the hash maybe 11 caracters long and rapidly overload gawain's attrib name buffer.
2017-05-01 17:58:15 +02:00
if (types & MR_DATATYPE_SHADING) {
rdata->uv_ct = CustomData_number_of_layers(&me->ldata, CD_MLOOPUV);
rdata->vcol_ct = CustomData_number_of_layers(&me->ldata, CD_MLOOPCOL);
rdata->mloopuv = MEM_mallocN(sizeof(*rdata->mloopuv) * rdata->uv_ct, "rdata->mloopuv");
rdata->mcol = MEM_mallocN(sizeof(*rdata->mcol) * rdata->vcol_ct, "rdata->mcol");
rdata->mtangent = MEM_mallocN(sizeof(*rdata->mtangent) * rdata->uv_ct, "rdata->mtangent");
rdata->uv_names = MEM_mallocN(sizeof(*rdata->uv_names) * rdata->uv_ct, "rdata->uv_names");
rdata->vcol_names = MEM_mallocN(sizeof(*rdata->vcol_names) * rdata->vcol_ct, "rdata->vcol_names");
rdata->tangent_names = MEM_mallocN(sizeof(*rdata->tangent_names) * rdata->uv_ct, "rdata->tangent_names");
rdata->uv_ofs = MEM_mallocN(sizeof(*rdata->uv_ofs) * rdata->uv_ct, "rdata->uv_ofs");
rdata->vcol_ofs = MEM_mallocN(sizeof(*rdata->vcol_ofs) * rdata->vcol_ct, "rdata->vcol_ofs");
rdata->tangent_ofs = MEM_mallocN(sizeof(*rdata->tangent_ofs) * rdata->uv_ct, "rdata->tangent_ofs");
/* Allocate max */
rdata->auto_vcol = MEM_callocN(sizeof(*rdata->auto_vcol) * rdata->vcol_ct, "rdata->auto_vcol");
rdata->auto_names = MEM_mallocN(sizeof(*rdata->auto_names) * (rdata->vcol_ct + rdata->uv_ct), "rdata->auto_names");
/* XXX FIXME XXX */
/* We use a hash to identify each data layer based on its name.
* Gawain then search for this name in the current shader and bind if it exists.
* NOTE : This is prone to hash collision.
* One solution to hash collision would be to format the cd layer name
* to a safe glsl var name, but without name clash.
* NOTE 2 : Replicate changes to code_generate_vertex_new() in gpu_codegen.c */
for (int i = 0; i < rdata->vcol_ct; ++i) {
const char *name = CustomData_get_layer_name(&me->ldata, CD_MLOOPCOL, i);
unsigned int hash = BLI_ghashutil_strhash_p(name);
BLI_snprintf(rdata->vcol_names[i], sizeof(*rdata->vcol_names), "c%u", hash);
rdata->mcol[i] = CustomData_get_layer_n(&me->ldata, CD_MLOOPCOL, i);
if (rdata->edit_bmesh) {
rdata->vcol_ofs[i] = CustomData_get_n_offset(&rdata->edit_bmesh->bm->ldata, CD_MLOOPCOL, i);
}
/* Gather number of auto layers. */
/* We only do vcols that are not overridden by uvs */
if (CustomData_get_named_layer_index(&me->ldata, CD_MLOOPUV, name) == -1) {
BLI_snprintf(rdata->auto_names[rdata->uv_ct + i], sizeof(*rdata->auto_names), "a%u", hash);
rdata->auto_vcol[i] = true;
}
}
/* Start Fresh */
CustomData_free_layers(&me->ldata, CD_MLOOPTANGENT, me->totloop);
for (int i = 0; i < rdata->uv_ct; ++i) {
const char *name = CustomData_get_layer_name(&me->ldata, CD_MLOOPUV, i);
unsigned int hash = BLI_ghashutil_strhash_p(name);
{
/* UVs */
BLI_snprintf(rdata->uv_names[i], sizeof(*rdata->uv_names), "u%u", hash);
rdata->mloopuv[i] = CustomData_get_layer_n(&me->ldata, CD_MLOOPUV, i);
if (rdata->edit_bmesh) {
rdata->uv_ofs[i] = CustomData_get_n_offset(&rdata->edit_bmesh->bm->ldata, CD_MLOOPUV, i);
}
BLI_snprintf(rdata->auto_names[i], sizeof(*rdata->auto_names), "a%u", hash);
}
{
/* Tangents*/
BLI_snprintf(rdata->tangent_names[i], sizeof(*rdata->tangent_names), "t%u", hash);
if (rdata->edit_bmesh) {
#if 0 /* TODO Waiting for the port of emDM_calc_loop_tangents */
BMesh *bm = rdata->edit_bmesh->bm;
float (*loopnors)[3] = CustomData_get_layer(&bm->ldata, CD_NORMAL);
rdata->mtangent[i] = CustomData_add_layer(&bm->ldata, CD_MLOOPTANGENT, CD_CALLOC, NULL, bm->totloop);
CustomData_set_layer_flag(&bm->ldata, CD_MLOOPTANGENT, CD_FLAG_TEMPORARY);
BKE_mesh_loop_tangents_ex(bm->mvert, bm->totvert, bm->mloop, rdata->mtangent[i],
loopnors, rdata->mloopuv[i], bm->totloop, bm->mpoly, bm->totpoly, NULL);
rdata->tangent_ofs[i] = CustomData_get_n_offset(&bm->ldata, CD_MLOOPTANGENT, i);
#else
rdata->tangent_ofs[i] = -1;
#endif
}
else {
if (!CustomData_has_layer(&me->ldata, CD_NORMAL)) {
BKE_mesh_calc_normals_split(me);
}
float (*loopnors)[3] = CustomData_get_layer(&me->ldata, CD_NORMAL);
rdata->mtangent[i] = CustomData_add_layer(&me->ldata, CD_MLOOPTANGENT, CD_CALLOC, NULL, me->totloop);
CustomData_set_layer_flag(&me->ldata, CD_MLOOPTANGENT, CD_FLAG_TEMPORARY);
BKE_mesh_loop_tangents_ex(me->mvert, me->totvert, me->mloop, rdata->mtangent[i],
loopnors, rdata->mloopuv[i], me->totloop, me->mpoly, me->totpoly, NULL);
}
}
}
rdata->uv_active = CustomData_get_active_layer_index(&me->ldata, CD_MLOOPUV) - CustomData_get_layer_index(&me->ldata, CD_MLOOPUV);
rdata->vcol_active = CustomData_get_active_layer_index(&me->ldata, CD_MLOOPCOL) - CustomData_get_layer_index(&me->ldata, CD_MLOOPCOL);
rdata->tangent_active = CustomData_get_active_layer_index(&me->ldata, CD_MLOOPTANGENT) - CustomData_get_layer_index(&me->ldata, CD_MLOOPTANGENT);
rdata->orco = CustomData_get_layer(&me->vdata, CD_ORCO);
/* If orco is not available compute it ourselves */
if (!rdata->orco) {
if (me->edit_btmesh) {
BMesh *bm = me->edit_btmesh->bm;
rdata->orco = MEM_mallocN(sizeof(*rdata->orco) * rdata->vert_len, "orco mesh");
BLI_assert((bm->elem_table_dirty & BM_VERT) == 0);
BMVert **vtable = bm->vtable;
for (int i = 0; i < bm->totvert; i++) {
copy_v3_v3(rdata->orco[i], vtable[i]->co);
}
BKE_mesh_orco_verts_transform(me, rdata->orco, rdata->vert_len, 0);
}
else {
rdata->orco = MEM_mallocN(sizeof(*rdata->orco) * rdata->vert_len, "orco mesh");
MVert *mvert = rdata->mvert;
for (int a = 0; a < rdata->vert_len; a++, mvert++) {
copy_v3_v3(rdata->orco[a], mvert->co);
}
BKE_mesh_orco_verts_transform(me, rdata->orco, rdata->vert_len, 0);
}
}
}
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
return rdata;
Convert MBC_ API to Mesh (instead of derived mesh) and move it to mesh_render This includes a few fixes in the MBC_ api. The idea here is for this to be the only interface the render engines will deal with for the meshes. If we need to expose special options for sculpting engine we refactor this accordingly. But for now we are shaping this in a per-case base. Note: * We still need to hook up to the depsgraph to force clear/update of batch_cache when mesh changes (I'm waiting for Sergey Sharybin's depsgraph update for this though) * Also ideally we could/should use BMesh directly instead of DerivedMesh, but this will do for now. Note 2: In the end I renamed the `BKE_mesh_render` functions to `static mesh_render`. We can re-expose them as BKE_* later once we need it. Reviewers: merwin Subscribers: fclem Differential Revision: https://developer.blender.org/D2476
2017-01-25 09:16:29 +01:00
}
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
static void mesh_render_data_free(MeshRenderData *rdata)
Convert MBC_ API to Mesh (instead of derived mesh) and move it to mesh_render This includes a few fixes in the MBC_ api. The idea here is for this to be the only interface the render engines will deal with for the meshes. If we need to expose special options for sculpting engine we refactor this accordingly. But for now we are shaping this in a per-case base. Note: * We still need to hook up to the depsgraph to force clear/update of batch_cache when mesh changes (I'm waiting for Sergey Sharybin's depsgraph update for this though) * Also ideally we could/should use BMesh directly instead of DerivedMesh, but this will do for now. Note 2: In the end I renamed the `BKE_mesh_render` functions to `static mesh_render`. We can re-expose them as BKE_* later once we need it. Reviewers: merwin Subscribers: fclem Differential Revision: https://developer.blender.org/D2476
2017-01-25 09:16:29 +01:00
{
Mesh Batch Cache: Add support for uv/vcol/tangent data layers. This is a bad implementation waiting for some improvement that are : - Gawain support for multiple names. -> will remove duplication of attribute alloc - Glsl safe name string escaping. -> will remove any risk of hash colision. Old glsl drawing was relying on attribute being choosen by DerivedMesh at drawing time. For this reason, we declare all possible attrib "name" for each data layer inside the batches. In the glsl code, we declare required data by type and name. Then Gawain only bind vbos if names correspond. This is way cleaner as we do not need to access the CD itself when drawing. One other problem is that the hash maybe 11 caracters long and rapidly overload gawain's attrib name buffer.
2017-05-01 17:58:15 +02:00
MEM_SAFE_FREE(rdata->auto_vcol);
MEM_SAFE_FREE(rdata->auto_names);
MEM_SAFE_FREE(rdata->uv_ofs);
MEM_SAFE_FREE(rdata->vcol_ofs);
MEM_SAFE_FREE(rdata->tangent_ofs);
MEM_SAFE_FREE(rdata->orco);
MEM_SAFE_FREE(rdata->mloopuv);
MEM_SAFE_FREE(rdata->mcol);
MEM_SAFE_FREE(rdata->mtangent);
MEM_SAFE_FREE(rdata->loose_verts);
MEM_SAFE_FREE(rdata->loose_edges);
MEM_SAFE_FREE(rdata->edges_adjacent_polys);
MEM_SAFE_FREE(rdata->mlooptri);
MEM_SAFE_FREE(rdata->poly_normals);
MEM_SAFE_FREE(rdata->poly_normals_short);
MEM_SAFE_FREE(rdata->vert_normals_short);
MEM_SAFE_FREE(rdata->uv_names);
MEM_SAFE_FREE(rdata->vcol_names);
MEM_SAFE_FREE(rdata->tangent_names);
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
MEM_freeN(rdata);
Convert MBC_ API to Mesh (instead of derived mesh) and move it to mesh_render This includes a few fixes in the MBC_ api. The idea here is for this to be the only interface the render engines will deal with for the meshes. If we need to expose special options for sculpting engine we refactor this accordingly. But for now we are shaping this in a per-case base. Note: * We still need to hook up to the depsgraph to force clear/update of batch_cache when mesh changes (I'm waiting for Sergey Sharybin's depsgraph update for this though) * Also ideally we could/should use BMesh directly instead of DerivedMesh, but this will do for now. Note 2: In the end I renamed the `BKE_mesh_render` functions to `static mesh_render`. We can re-expose them as BKE_* later once we need it. Reviewers: merwin Subscribers: fclem Differential Revision: https://developer.blender.org/D2476
2017-01-25 09:16:29 +01:00
}
Mesh Batch Cache: Add support for uv/vcol/tangent data layers. This is a bad implementation waiting for some improvement that are : - Gawain support for multiple names. -> will remove duplication of attribute alloc - Glsl safe name string escaping. -> will remove any risk of hash colision. Old glsl drawing was relying on attribute being choosen by DerivedMesh at drawing time. For this reason, we declare all possible attrib "name" for each data layer inside the batches. In the glsl code, we declare required data by type and name. Then Gawain only bind vbos if names correspond. This is way cleaner as we do not need to access the CD itself when drawing. One other problem is that the hash maybe 11 caracters long and rapidly overload gawain's attrib name buffer.
2017-05-01 17:58:15 +02:00
static const char *mesh_render_data_uv_auto_layer_name_get(const MeshRenderData *rdata, int layer)
{
BLI_assert(rdata->types & MR_DATATYPE_SHADING);
return rdata->auto_names[layer];
}
static const char *mesh_render_data_vcol_auto_layer_name_get(const MeshRenderData *rdata, int layer)
{
BLI_assert(rdata->types & MR_DATATYPE_SHADING);
return rdata->auto_names[rdata->uv_ct + layer];
}
static const char *mesh_render_data_uv_layer_name_get(const MeshRenderData *rdata, int layer)
{
BLI_assert(rdata->types & MR_DATATYPE_SHADING);
return rdata->uv_names[layer];
}
static const char *mesh_render_data_vcol_layer_name_get(const MeshRenderData *rdata, int layer)
{
BLI_assert(rdata->types & MR_DATATYPE_SHADING);
return rdata->vcol_names[layer];
}
static const char *mesh_render_data_tangent_layer_name_get(const MeshRenderData *rdata, int layer)
{
BLI_assert(rdata->types & MR_DATATYPE_SHADING);
return rdata->tangent_names[layer];
}
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
static int mesh_render_data_verts_len_get(const MeshRenderData *rdata)
Convert MBC_ API to Mesh (instead of derived mesh) and move it to mesh_render This includes a few fixes in the MBC_ api. The idea here is for this to be the only interface the render engines will deal with for the meshes. If we need to expose special options for sculpting engine we refactor this accordingly. But for now we are shaping this in a per-case base. Note: * We still need to hook up to the depsgraph to force clear/update of batch_cache when mesh changes (I'm waiting for Sergey Sharybin's depsgraph update for this though) * Also ideally we could/should use BMesh directly instead of DerivedMesh, but this will do for now. Note 2: In the end I renamed the `BKE_mesh_render` functions to `static mesh_render`. We can re-expose them as BKE_* later once we need it. Reviewers: merwin Subscribers: fclem Differential Revision: https://developer.blender.org/D2476
2017-01-25 09:16:29 +01:00
{
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
BLI_assert(rdata->types & MR_DATATYPE_VERT);
return rdata->vert_len;
Convert MBC_ API to Mesh (instead of derived mesh) and move it to mesh_render This includes a few fixes in the MBC_ api. The idea here is for this to be the only interface the render engines will deal with for the meshes. If we need to expose special options for sculpting engine we refactor this accordingly. But for now we are shaping this in a per-case base. Note: * We still need to hook up to the depsgraph to force clear/update of batch_cache when mesh changes (I'm waiting for Sergey Sharybin's depsgraph update for this though) * Also ideally we could/should use BMesh directly instead of DerivedMesh, but this will do for now. Note 2: In the end I renamed the `BKE_mesh_render` functions to `static mesh_render`. We can re-expose them as BKE_* later once we need it. Reviewers: merwin Subscribers: fclem Differential Revision: https://developer.blender.org/D2476
2017-01-25 09:16:29 +01:00
}
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
static int mesh_render_data_loose_verts_len_get(const MeshRenderData *rdata)
Edit Mesh overlays: added looses edges and verts as triangle. We loose a bit of memory for (more?) performance.
2017-02-28 14:30:34 +01:00
{
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
BLI_assert(rdata->types & MR_DATATYPE_OVERLAY);
return rdata->loose_vert_len;
Edit Mesh overlays: added looses edges and verts as triangle. We loose a bit of memory for (more?) performance.
2017-02-28 14:30:34 +01:00
}
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
static int mesh_render_data_edges_len_get(const MeshRenderData *rdata)
Convert MBC_ API to Mesh (instead of derived mesh) and move it to mesh_render This includes a few fixes in the MBC_ api. The idea here is for this to be the only interface the render engines will deal with for the meshes. If we need to expose special options for sculpting engine we refactor this accordingly. But for now we are shaping this in a per-case base. Note: * We still need to hook up to the depsgraph to force clear/update of batch_cache when mesh changes (I'm waiting for Sergey Sharybin's depsgraph update for this though) * Also ideally we could/should use BMesh directly instead of DerivedMesh, but this will do for now. Note 2: In the end I renamed the `BKE_mesh_render` functions to `static mesh_render`. We can re-expose them as BKE_* later once we need it. Reviewers: merwin Subscribers: fclem Differential Revision: https://developer.blender.org/D2476
2017-01-25 09:16:29 +01:00
{
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
BLI_assert(rdata->types & MR_DATATYPE_EDGE);
return rdata->edge_len;
Convert MBC_ API to Mesh (instead of derived mesh) and move it to mesh_render This includes a few fixes in the MBC_ api. The idea here is for this to be the only interface the render engines will deal with for the meshes. If we need to expose special options for sculpting engine we refactor this accordingly. But for now we are shaping this in a per-case base. Note: * We still need to hook up to the depsgraph to force clear/update of batch_cache when mesh changes (I'm waiting for Sergey Sharybin's depsgraph update for this though) * Also ideally we could/should use BMesh directly instead of DerivedMesh, but this will do for now. Note 2: In the end I renamed the `BKE_mesh_render` functions to `static mesh_render`. We can re-expose them as BKE_* later once we need it. Reviewers: merwin Subscribers: fclem Differential Revision: https://developer.blender.org/D2476
2017-01-25 09:16:29 +01:00
}
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
static int mesh_render_data_loose_edges_len_get(const MeshRenderData *rdata)
Edit Mesh overlays: added looses edges and verts as triangle. We loose a bit of memory for (more?) performance.
2017-02-28 14:30:34 +01:00
{
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
BLI_assert(rdata->types & MR_DATATYPE_OVERLAY);
return rdata->loose_edge_len;
Edit Mesh overlays: added looses edges and verts as triangle. We loose a bit of memory for (more?) performance.
2017-02-28 14:30:34 +01:00
}
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
static int mesh_render_data_looptri_len_get(const MeshRenderData *rdata)
Convert MBC_ API to Mesh (instead of derived mesh) and move it to mesh_render This includes a few fixes in the MBC_ api. The idea here is for this to be the only interface the render engines will deal with for the meshes. If we need to expose special options for sculpting engine we refactor this accordingly. But for now we are shaping this in a per-case base. Note: * We still need to hook up to the depsgraph to force clear/update of batch_cache when mesh changes (I'm waiting for Sergey Sharybin's depsgraph update for this though) * Also ideally we could/should use BMesh directly instead of DerivedMesh, but this will do for now. Note 2: In the end I renamed the `BKE_mesh_render` functions to `static mesh_render`. We can re-expose them as BKE_* later once we need it. Reviewers: merwin Subscribers: fclem Differential Revision: https://developer.blender.org/D2476
2017-01-25 09:16:29 +01:00
{
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
BLI_assert(rdata->types & MR_DATATYPE_LOOPTRI);
return rdata->tri_len;
Convert MBC_ API to Mesh (instead of derived mesh) and move it to mesh_render This includes a few fixes in the MBC_ api. The idea here is for this to be the only interface the render engines will deal with for the meshes. If we need to expose special options for sculpting engine we refactor this accordingly. But for now we are shaping this in a per-case base. Note: * We still need to hook up to the depsgraph to force clear/update of batch_cache when mesh changes (I'm waiting for Sergey Sharybin's depsgraph update for this though) * Also ideally we could/should use BMesh directly instead of DerivedMesh, but this will do for now. Note 2: In the end I renamed the `BKE_mesh_render` functions to `static mesh_render`. We can re-expose them as BKE_* later once we need it. Reviewers: merwin Subscribers: fclem Differential Revision: https://developer.blender.org/D2476
2017-01-25 09:16:29 +01:00
}
Eevee: Make default shaders works. - Added New Batch cache routine to get the split mesh for each material. (optimization needed) - Did some shader modifications to create default_frag and support a somwhat correct fresnel for lamps (to finish)
2017-04-25 18:46:59 +02:00
static int mesh_render_data_mat_ct_get(const MeshRenderData *rdata)
{
BLI_assert(rdata->types & MR_DATATYPE_POLY);
return rdata->mat_ct;
}
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
static int UNUSED_FUNCTION(mesh_render_data_loops_len_get)(const MeshRenderData *rdata)
Convert MBC_ API to Mesh (instead of derived mesh) and move it to mesh_render This includes a few fixes in the MBC_ api. The idea here is for this to be the only interface the render engines will deal with for the meshes. If we need to expose special options for sculpting engine we refactor this accordingly. But for now we are shaping this in a per-case base. Note: * We still need to hook up to the depsgraph to force clear/update of batch_cache when mesh changes (I'm waiting for Sergey Sharybin's depsgraph update for this though) * Also ideally we could/should use BMesh directly instead of DerivedMesh, but this will do for now. Note 2: In the end I renamed the `BKE_mesh_render` functions to `static mesh_render`. We can re-expose them as BKE_* later once we need it. Reviewers: merwin Subscribers: fclem Differential Revision: https://developer.blender.org/D2476
2017-01-25 09:16:29 +01:00
{
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
BLI_assert(rdata->types & MR_DATATYPE_LOOP);
return rdata->loop_len;
Convert MBC_ API to Mesh (instead of derived mesh) and move it to mesh_render This includes a few fixes in the MBC_ api. The idea here is for this to be the only interface the render engines will deal with for the meshes. If we need to expose special options for sculpting engine we refactor this accordingly. But for now we are shaping this in a per-case base. Note: * We still need to hook up to the depsgraph to force clear/update of batch_cache when mesh changes (I'm waiting for Sergey Sharybin's depsgraph update for this though) * Also ideally we could/should use BMesh directly instead of DerivedMesh, but this will do for now. Note 2: In the end I renamed the `BKE_mesh_render` functions to `static mesh_render`. We can re-expose them as BKE_* later once we need it. Reviewers: merwin Subscribers: fclem Differential Revision: https://developer.blender.org/D2476
2017-01-25 09:16:29 +01:00
}
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
static int mesh_render_data_polys_len_get(const MeshRenderData *rdata)
Convert MBC_ API to Mesh (instead of derived mesh) and move it to mesh_render This includes a few fixes in the MBC_ api. The idea here is for this to be the only interface the render engines will deal with for the meshes. If we need to expose special options for sculpting engine we refactor this accordingly. But for now we are shaping this in a per-case base. Note: * We still need to hook up to the depsgraph to force clear/update of batch_cache when mesh changes (I'm waiting for Sergey Sharybin's depsgraph update for this though) * Also ideally we could/should use BMesh directly instead of DerivedMesh, but this will do for now. Note 2: In the end I renamed the `BKE_mesh_render` functions to `static mesh_render`. We can re-expose them as BKE_* later once we need it. Reviewers: merwin Subscribers: fclem Differential Revision: https://developer.blender.org/D2476
2017-01-25 09:16:29 +01:00
{
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
BLI_assert(rdata->types & MR_DATATYPE_POLY);
return rdata->poly_len;
Convert MBC_ API to Mesh (instead of derived mesh) and move it to mesh_render This includes a few fixes in the MBC_ api. The idea here is for this to be the only interface the render engines will deal with for the meshes. If we need to expose special options for sculpting engine we refactor this accordingly. But for now we are shaping this in a per-case base. Note: * We still need to hook up to the depsgraph to force clear/update of batch_cache when mesh changes (I'm waiting for Sergey Sharybin's depsgraph update for this though) * Also ideally we could/should use BMesh directly instead of DerivedMesh, but this will do for now. Note 2: In the end I renamed the `BKE_mesh_render` functions to `static mesh_render`. We can re-expose them as BKE_* later once we need it. Reviewers: merwin Subscribers: fclem Differential Revision: https://developer.blender.org/D2476
2017-01-25 09:16:29 +01:00
}
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
static float *mesh_render_data_vert_co(const MeshRenderData *rdata, const int vert_idx)
Convert MBC_ API to Mesh (instead of derived mesh) and move it to mesh_render This includes a few fixes in the MBC_ api. The idea here is for this to be the only interface the render engines will deal with for the meshes. If we need to expose special options for sculpting engine we refactor this accordingly. But for now we are shaping this in a per-case base. Note: * We still need to hook up to the depsgraph to force clear/update of batch_cache when mesh changes (I'm waiting for Sergey Sharybin's depsgraph update for this though) * Also ideally we could/should use BMesh directly instead of DerivedMesh, but this will do for now. Note 2: In the end I renamed the `BKE_mesh_render` functions to `static mesh_render`. We can re-expose them as BKE_* later once we need it. Reviewers: merwin Subscribers: fclem Differential Revision: https://developer.blender.org/D2476
2017-01-25 09:16:29 +01:00
{
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
BLI_assert(rdata->types & MR_DATATYPE_VERT);
Convert MBC_ API to Mesh (instead of derived mesh) and move it to mesh_render This includes a few fixes in the MBC_ api. The idea here is for this to be the only interface the render engines will deal with for the meshes. If we need to expose special options for sculpting engine we refactor this accordingly. But for now we are shaping this in a per-case base. Note: * We still need to hook up to the depsgraph to force clear/update of batch_cache when mesh changes (I'm waiting for Sergey Sharybin's depsgraph update for this though) * Also ideally we could/should use BMesh directly instead of DerivedMesh, but this will do for now. Note 2: In the end I renamed the `BKE_mesh_render` functions to `static mesh_render`. We can re-expose them as BKE_* later once we need it. Reviewers: merwin Subscribers: fclem Differential Revision: https://developer.blender.org/D2476
2017-01-25 09:16:29 +01:00
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
if (rdata->edit_bmesh) {
BMesh *bm = rdata->edit_bmesh->bm;
Rework of BKE's mesh_render to support BMesh directly. Note that since there is no (efficient) ways to get arrays of MVert/MEdge/etc. out of a BMesh, I refactored quite heavily internals of BKE_mesh_render. Now, when you do need acess to mesh data to generate cached batches, you create an abstract struct from mesh (either Mesh or BMesh if available), and then use advanced helpers to extract needed data, on a per-item basis (no more handling of arrays of verts/edges/... in batches code). This allows to: * Avoid having to create arrays of BMesh elements. * Take advantage of existing advanced BMesh topology and connectivity data. Reviewers: dfelinto, fclem, merwin Differential Revision: https://developer.blender.org/D2521
2017-02-23 11:54:40 +01:00
BMVert *bv = BM_vert_at_index(bm, vert_idx);
return bv->co;
}
else {
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
return rdata->mvert[vert_idx].co;
Rework of BKE's mesh_render to support BMesh directly. Note that since there is no (efficient) ways to get arrays of MVert/MEdge/etc. out of a BMesh, I refactored quite heavily internals of BKE_mesh_render. Now, when you do need acess to mesh data to generate cached batches, you create an abstract struct from mesh (either Mesh or BMesh if available), and then use advanced helpers to extract needed data, on a per-item basis (no more handling of arrays of verts/edges/... in batches code). This allows to: * Avoid having to create arrays of BMesh elements. * Take advantage of existing advanced BMesh topology and connectivity data. Reviewers: dfelinto, fclem, merwin Differential Revision: https://developer.blender.org/D2521
2017-02-23 11:54:40 +01:00
}
Clay Engine: Correct Mesh Normals
2017-02-16 16:19:48 +01:00
}
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
static short *mesh_render_data_vert_nor(const MeshRenderData *rdata, const int vert_idx)
Edit Mesh overlay: Ported Display Normals option
2017-03-09 01:29:58 +01:00
{
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
BLI_assert(rdata->types & MR_DATATYPE_VERT);
Edit Mesh overlay: Ported Display Normals option
2017-03-09 01:29:58 +01:00
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
if (rdata->edit_bmesh) {
Edit Mesh overlay: Ported Display Normals option
2017-03-09 01:29:58 +01:00
static short fno[3];
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
BMesh *bm = rdata->edit_bmesh->bm;
Edit Mesh overlay: Ported Display Normals option
2017-03-09 01:29:58 +01:00
BMVert *bv = BM_vert_at_index(bm, vert_idx);
normal_float_to_short_v3(fno, bv->no);
return fno;
}
else {
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
return rdata->mvert[vert_idx].no;
Edit Mesh overlay: Ported Display Normals option
2017-03-09 01:29:58 +01:00
}
}
Draw Manager: T51218 Support edit-mesh hide
2017-04-18 23:29:55 +10:00
static bool mesh_render_data_edge_verts_indices_get(
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
const MeshRenderData *rdata, const int edge_idx,
Cleanup: naming & line length
2017-04-12 18:23:17 +10:00
int r_vert_idx[2])
Convert MBC_ API to Mesh (instead of derived mesh) and move it to mesh_render This includes a few fixes in the MBC_ api. The idea here is for this to be the only interface the render engines will deal with for the meshes. If we need to expose special options for sculpting engine we refactor this accordingly. But for now we are shaping this in a per-case base. Note: * We still need to hook up to the depsgraph to force clear/update of batch_cache when mesh changes (I'm waiting for Sergey Sharybin's depsgraph update for this though) * Also ideally we could/should use BMesh directly instead of DerivedMesh, but this will do for now. Note 2: In the end I renamed the `BKE_mesh_render` functions to `static mesh_render`. We can re-expose them as BKE_* later once we need it. Reviewers: merwin Subscribers: fclem Differential Revision: https://developer.blender.org/D2476
2017-01-25 09:16:29 +01:00
{
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
BLI_assert(rdata->types & MR_DATATYPE_EDGE);
Convert MBC_ API to Mesh (instead of derived mesh) and move it to mesh_render This includes a few fixes in the MBC_ api. The idea here is for this to be the only interface the render engines will deal with for the meshes. If we need to expose special options for sculpting engine we refactor this accordingly. But for now we are shaping this in a per-case base. Note: * We still need to hook up to the depsgraph to force clear/update of batch_cache when mesh changes (I'm waiting for Sergey Sharybin's depsgraph update for this though) * Also ideally we could/should use BMesh directly instead of DerivedMesh, but this will do for now. Note 2: In the end I renamed the `BKE_mesh_render` functions to `static mesh_render`. We can re-expose them as BKE_* later once we need it. Reviewers: merwin Subscribers: fclem Differential Revision: https://developer.blender.org/D2476
2017-01-25 09:16:29 +01:00
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
if (rdata->edit_bmesh) {
const BMEdge *bm_edge = BM_edge_at_index(rdata->edit_bmesh->bm, edge_idx);
Draw Manager: T51218 Support edit-mesh hide
2017-04-18 23:29:55 +10:00
if (BM_elem_flag_test(bm_edge, BM_ELEM_HIDDEN)) {
return false;
}
Rework of BKE's mesh_render to support BMesh directly. Note that since there is no (efficient) ways to get arrays of MVert/MEdge/etc. out of a BMesh, I refactored quite heavily internals of BKE_mesh_render. Now, when you do need acess to mesh data to generate cached batches, you create an abstract struct from mesh (either Mesh or BMesh if available), and then use advanced helpers to extract needed data, on a per-item basis (no more handling of arrays of verts/edges/... in batches code). This allows to: * Avoid having to create arrays of BMesh elements. * Take advantage of existing advanced BMesh topology and connectivity data. Reviewers: dfelinto, fclem, merwin Differential Revision: https://developer.blender.org/D2521
2017-02-23 11:54:40 +01:00
r_vert_idx[0] = BM_elem_index_get(bm_edge->v1);
r_vert_idx[1] = BM_elem_index_get(bm_edge->v2);
}
else {
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
const MEdge *me = &rdata->medge[edge_idx];
Rework of BKE's mesh_render to support BMesh directly. Note that since there is no (efficient) ways to get arrays of MVert/MEdge/etc. out of a BMesh, I refactored quite heavily internals of BKE_mesh_render. Now, when you do need acess to mesh data to generate cached batches, you create an abstract struct from mesh (either Mesh or BMesh if available), and then use advanced helpers to extract needed data, on a per-item basis (no more handling of arrays of verts/edges/... in batches code). This allows to: * Avoid having to create arrays of BMesh elements. * Take advantage of existing advanced BMesh topology and connectivity data. Reviewers: dfelinto, fclem, merwin Differential Revision: https://developer.blender.org/D2521
2017-02-23 11:54:40 +01:00
r_vert_idx[0] = me->v1;
r_vert_idx[1] = me->v2;
}
Draw Manager: T51218 Support edit-mesh hide
2017-04-18 23:29:55 +10:00
return true;
Convert MBC_ API to Mesh (instead of derived mesh) and move it to mesh_render This includes a few fixes in the MBC_ api. The idea here is for this to be the only interface the render engines will deal with for the meshes. If we need to expose special options for sculpting engine we refactor this accordingly. But for now we are shaping this in a per-case base. Note: * We still need to hook up to the depsgraph to force clear/update of batch_cache when mesh changes (I'm waiting for Sergey Sharybin's depsgraph update for this though) * Also ideally we could/should use BMesh directly instead of DerivedMesh, but this will do for now. Note 2: In the end I renamed the `BKE_mesh_render` functions to `static mesh_render`. We can re-expose them as BKE_* later once we need it. Reviewers: merwin Subscribers: fclem Differential Revision: https://developer.blender.org/D2476
2017-01-25 09:16:29 +01:00
}
Draw Manager: T51218 Support edit-mesh hide
2017-04-18 23:29:55 +10:00
static bool mesh_render_data_pnors_pcenter_select_get(
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
MeshRenderData *rdata, const int poly,
Cleanup: naming & line length
2017-04-12 18:23:17 +10:00
float r_pnors[3], float r_center[3], bool *r_selected)
Edit Mode overlay: Moved Shaders to draw modules and resolved some draw issue. We don't want to clutter gpu_shader.c with engine specific code Added face's center dot Simplified loose vert shader
2017-03-02 01:07:03 +01:00
{
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
BLI_assert(rdata->types & (MR_DATATYPE_VERT | MR_DATATYPE_LOOP | MR_DATATYPE_POLY));
Edit Mode overlay: Moved Shaders to draw modules and resolved some draw issue. We don't want to clutter gpu_shader.c with engine specific code Added face's center dot Simplified loose vert shader
2017-03-02 01:07:03 +01:00
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
if (rdata->edit_bmesh) {
const BMFace *bf = BM_face_at_index(rdata->edit_bmesh->bm, poly);
Draw Manager: T51218 Support edit-mesh hide
2017-04-18 23:29:55 +10:00
if (BM_elem_flag_test(bf, BM_ELEM_HIDDEN)) {
return false;
}
Cleanup: naming & line length
2017-04-12 18:23:17 +10:00
BM_face_calc_center_mean(bf, r_center);
BM_face_calc_normal(bf, r_pnors);
*r_selected = (BM_elem_flag_test(bf, BM_ELEM_SELECT) != 0) ? true : false;
Edit Mode overlay: Moved Shaders to draw modules and resolved some draw issue. We don't want to clutter gpu_shader.c with engine specific code Added face's center dot Simplified loose vert shader
2017-03-02 01:07:03 +01:00
}
else {
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
MVert *mvert = rdata->mvert;
const MPoly *mpoly = rdata->mpoly + poly;
const MLoop *mloop = rdata->mloop + mpoly->loopstart;
Edit Mode overlay: Moved Shaders to draw modules and resolved some draw issue. We don't want to clutter gpu_shader.c with engine specific code Added face's center dot Simplified loose vert shader
2017-03-02 01:07:03 +01:00
Cleanup: naming & line length
2017-04-12 18:23:17 +10:00
BKE_mesh_calc_poly_center(mpoly, mloop, mvert, r_center);
BKE_mesh_calc_poly_normal(mpoly, mloop, mvert, r_pnors);
Edit Mode overlay: Moved Shaders to draw modules and resolved some draw issue. We don't want to clutter gpu_shader.c with engine specific code Added face's center dot Simplified loose vert shader
2017-03-02 01:07:03 +01:00
Cleanup: naming & line length
2017-04-12 18:23:17 +10:00
*r_selected = false; /* No selection if not in edit mode */
Edit Mode overlay: Moved Shaders to draw modules and resolved some draw issue. We don't want to clutter gpu_shader.c with engine specific code Added face's center dot Simplified loose vert shader
2017-03-02 01:07:03 +01:00
}
Draw Manager: T51218 Support edit-mesh hide
2017-04-18 23:29:55 +10:00
return true;
Edit Mode overlay: Moved Shaders to draw modules and resolved some draw issue. We don't want to clutter gpu_shader.c with engine specific code Added face's center dot Simplified loose vert shader
2017-03-02 01:07:03 +01:00
}
Rework of BKE's mesh_render to support BMesh directly. Note that since there is no (efficient) ways to get arrays of MVert/MEdge/etc. out of a BMesh, I refactored quite heavily internals of BKE_mesh_render. Now, when you do need acess to mesh data to generate cached batches, you create an abstract struct from mesh (either Mesh or BMesh if available), and then use advanced helpers to extract needed data, on a per-item basis (no more handling of arrays of verts/edges/... in batches code). This allows to: * Avoid having to create arrays of BMesh elements. * Take advantage of existing advanced BMesh topology and connectivity data. Reviewers: dfelinto, fclem, merwin Differential Revision: https://developer.blender.org/D2521
2017-02-23 11:54:40 +01:00
static bool mesh_render_data_edge_vcos_manifold_pnors(
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
MeshRenderData *rdata, const int edge_index,
Draw Manager: T51218 Support edit-mesh hide
2017-04-18 23:29:55 +10:00
float **r_vco1, float **r_vco2, float **r_pnor1, float **r_pnor2, bool *r_is_manifold)
Convert MBC_ API to Mesh (instead of derived mesh) and move it to mesh_render This includes a few fixes in the MBC_ api. The idea here is for this to be the only interface the render engines will deal with for the meshes. If we need to expose special options for sculpting engine we refactor this accordingly. But for now we are shaping this in a per-case base. Note: * We still need to hook up to the depsgraph to force clear/update of batch_cache when mesh changes (I'm waiting for Sergey Sharybin's depsgraph update for this though) * Also ideally we could/should use BMesh directly instead of DerivedMesh, but this will do for now. Note 2: In the end I renamed the `BKE_mesh_render` functions to `static mesh_render`. We can re-expose them as BKE_* later once we need it. Reviewers: merwin Subscribers: fclem Differential Revision: https://developer.blender.org/D2476
2017-01-25 09:16:29 +01:00
{
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
BLI_assert(rdata->types & (MR_DATATYPE_VERT | MR_DATATYPE_EDGE | MR_DATATYPE_LOOP | MR_DATATYPE_POLY));
Convert MBC_ API to Mesh (instead of derived mesh) and move it to mesh_render This includes a few fixes in the MBC_ api. The idea here is for this to be the only interface the render engines will deal with for the meshes. If we need to expose special options for sculpting engine we refactor this accordingly. But for now we are shaping this in a per-case base. Note: * We still need to hook up to the depsgraph to force clear/update of batch_cache when mesh changes (I'm waiting for Sergey Sharybin's depsgraph update for this though) * Also ideally we could/should use BMesh directly instead of DerivedMesh, but this will do for now. Note 2: In the end I renamed the `BKE_mesh_render` functions to `static mesh_render`. We can re-expose them as BKE_* later once we need it. Reviewers: merwin Subscribers: fclem Differential Revision: https://developer.blender.org/D2476
2017-01-25 09:16:29 +01:00
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
if (rdata->edit_bmesh) {
BMesh *bm = rdata->edit_bmesh->bm;
Rework of BKE's mesh_render to support BMesh directly. Note that since there is no (efficient) ways to get arrays of MVert/MEdge/etc. out of a BMesh, I refactored quite heavily internals of BKE_mesh_render. Now, when you do need acess to mesh data to generate cached batches, you create an abstract struct from mesh (either Mesh or BMesh if available), and then use advanced helpers to extract needed data, on a per-item basis (no more handling of arrays of verts/edges/... in batches code). This allows to: * Avoid having to create arrays of BMesh elements. * Take advantage of existing advanced BMesh topology and connectivity data. Reviewers: dfelinto, fclem, merwin Differential Revision: https://developer.blender.org/D2521
2017-02-23 11:54:40 +01:00
BMEdge *bm_edge = BM_edge_at_index(bm, edge_index);
Draw Manager: T51218 Support edit-mesh hide
2017-04-18 23:29:55 +10:00
if (BM_elem_flag_test(bm_edge, BM_ELEM_HIDDEN)) {
return false;
}
Rework of BKE's mesh_render to support BMesh directly. Note that since there is no (efficient) ways to get arrays of MVert/MEdge/etc. out of a BMesh, I refactored quite heavily internals of BKE_mesh_render. Now, when you do need acess to mesh data to generate cached batches, you create an abstract struct from mesh (either Mesh or BMesh if available), and then use advanced helpers to extract needed data, on a per-item basis (no more handling of arrays of verts/edges/... in batches code). This allows to: * Avoid having to create arrays of BMesh elements. * Take advantage of existing advanced BMesh topology and connectivity data. Reviewers: dfelinto, fclem, merwin Differential Revision: https://developer.blender.org/D2521
2017-02-23 11:54:40 +01:00
*r_vco1 = bm_edge->v1->co;
*r_vco2 = bm_edge->v2->co;
if (BM_edge_is_manifold(bm_edge)) {
*r_pnor1 = bm_edge->l->f->no;
*r_pnor2 = bm_edge->l->radial_next->f->no;
Draw Manager: T51218 Support edit-mesh hide
2017-04-18 23:29:55 +10:00
*r_is_manifold = true;
}
else {
*r_is_manifold = false;
Rework of BKE's mesh_render to support BMesh directly. Note that since there is no (efficient) ways to get arrays of MVert/MEdge/etc. out of a BMesh, I refactored quite heavily internals of BKE_mesh_render. Now, when you do need acess to mesh data to generate cached batches, you create an abstract struct from mesh (either Mesh or BMesh if available), and then use advanced helpers to extract needed data, on a per-item basis (no more handling of arrays of verts/edges/... in batches code). This allows to: * Avoid having to create arrays of BMesh elements. * Take advantage of existing advanced BMesh topology and connectivity data. Reviewers: dfelinto, fclem, merwin Differential Revision: https://developer.blender.org/D2521
2017-02-23 11:54:40 +01:00
}
}
else {
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
MVert *mvert = rdata->mvert;
MEdge *medge = rdata->medge;
EdgeAdjacentPolys *eap = rdata->edges_adjacent_polys;
float (*pnors)[3] = rdata->poly_normals;
Rework of BKE's mesh_render to support BMesh directly. Note that since there is no (efficient) ways to get arrays of MVert/MEdge/etc. out of a BMesh, I refactored quite heavily internals of BKE_mesh_render. Now, when you do need acess to mesh data to generate cached batches, you create an abstract struct from mesh (either Mesh or BMesh if available), and then use advanced helpers to extract needed data, on a per-item basis (no more handling of arrays of verts/edges/... in batches code). This allows to: * Avoid having to create arrays of BMesh elements. * Take advantage of existing advanced BMesh topology and connectivity data. Reviewers: dfelinto, fclem, merwin Differential Revision: https://developer.blender.org/D2521
2017-02-23 11:54:40 +01:00
if (!eap) {
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
const MLoop *mloop = rdata->mloop;
const MPoly *mpoly = rdata->mpoly;
const int poly_len = rdata->poly_len;
Rework of BKE's mesh_render to support BMesh directly. Note that since there is no (efficient) ways to get arrays of MVert/MEdge/etc. out of a BMesh, I refactored quite heavily internals of BKE_mesh_render. Now, when you do need acess to mesh data to generate cached batches, you create an abstract struct from mesh (either Mesh or BMesh if available), and then use advanced helpers to extract needed data, on a per-item basis (no more handling of arrays of verts/edges/... in batches code). This allows to: * Avoid having to create arrays of BMesh elements. * Take advantage of existing advanced BMesh topology and connectivity data. Reviewers: dfelinto, fclem, merwin Differential Revision: https://developer.blender.org/D2521
2017-02-23 11:54:40 +01:00
const bool do_pnors = (pnors == NULL);
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
eap = rdata->edges_adjacent_polys = MEM_callocN(sizeof(*eap) * rdata->edge_len, __func__);
Rework of BKE's mesh_render to support BMesh directly. Note that since there is no (efficient) ways to get arrays of MVert/MEdge/etc. out of a BMesh, I refactored quite heavily internals of BKE_mesh_render. Now, when you do need acess to mesh data to generate cached batches, you create an abstract struct from mesh (either Mesh or BMesh if available), and then use advanced helpers to extract needed data, on a per-item basis (no more handling of arrays of verts/edges/... in batches code). This allows to: * Avoid having to create arrays of BMesh elements. * Take advantage of existing advanced BMesh topology and connectivity data. Reviewers: dfelinto, fclem, merwin Differential Revision: https://developer.blender.org/D2521
2017-02-23 11:54:40 +01:00
if (do_pnors) {
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
pnors = rdata->poly_normals = MEM_mallocN(sizeof(*pnors) * poly_len, __func__);
Rework of BKE's mesh_render to support BMesh directly. Note that since there is no (efficient) ways to get arrays of MVert/MEdge/etc. out of a BMesh, I refactored quite heavily internals of BKE_mesh_render. Now, when you do need acess to mesh data to generate cached batches, you create an abstract struct from mesh (either Mesh or BMesh if available), and then use advanced helpers to extract needed data, on a per-item basis (no more handling of arrays of verts/edges/... in batches code). This allows to: * Avoid having to create arrays of BMesh elements. * Take advantage of existing advanced BMesh topology and connectivity data. Reviewers: dfelinto, fclem, merwin Differential Revision: https://developer.blender.org/D2521
2017-02-23 11:54:40 +01:00
}
Clay Engine: Correct Mesh Normals
2017-02-16 16:19:48 +01:00
Cleanup: replace num/ct/tot with 'len'
2017-04-19 00:10:32 +10:00
for (int i = 0; i < poly_len; i++, mpoly++) {
Rework of BKE's mesh_render to support BMesh directly. Note that since there is no (efficient) ways to get arrays of MVert/MEdge/etc. out of a BMesh, I refactored quite heavily internals of BKE_mesh_render. Now, when you do need acess to mesh data to generate cached batches, you create an abstract struct from mesh (either Mesh or BMesh if available), and then use advanced helpers to extract needed data, on a per-item basis (no more handling of arrays of verts/edges/... in batches code). This allows to: * Avoid having to create arrays of BMesh elements. * Take advantage of existing advanced BMesh topology and connectivity data. Reviewers: dfelinto, fclem, merwin Differential Revision: https://developer.blender.org/D2521
2017-02-23 11:54:40 +01:00
if (do_pnors) {
BKE_mesh_calc_poly_normal(mpoly, mloop + mpoly->loopstart, mvert, pnors[i]);
}
const int loopend = mpoly->loopstart + mpoly->totloop;
for (int j = mpoly->loopstart; j < loopend; j++) {
const int edge_idx = mloop[j].e;
if (eap[edge_idx].count < 2) {
eap[edge_idx].face_index[eap[edge_idx].count] = i;
}
eap[edge_idx].count++;
}
}
}
BLI_assert(eap && pnors);
Convert MBC_ API to Mesh (instead of derived mesh) and move it to mesh_render This includes a few fixes in the MBC_ api. The idea here is for this to be the only interface the render engines will deal with for the meshes. If we need to expose special options for sculpting engine we refactor this accordingly. But for now we are shaping this in a per-case base. Note: * We still need to hook up to the depsgraph to force clear/update of batch_cache when mesh changes (I'm waiting for Sergey Sharybin's depsgraph update for this though) * Also ideally we could/should use BMesh directly instead of DerivedMesh, but this will do for now. Note 2: In the end I renamed the `BKE_mesh_render` functions to `static mesh_render`. We can re-expose them as BKE_* later once we need it. Reviewers: merwin Subscribers: fclem Differential Revision: https://developer.blender.org/D2476
2017-01-25 09:16:29 +01:00
Rework of BKE's mesh_render to support BMesh directly. Note that since there is no (efficient) ways to get arrays of MVert/MEdge/etc. out of a BMesh, I refactored quite heavily internals of BKE_mesh_render. Now, when you do need acess to mesh data to generate cached batches, you create an abstract struct from mesh (either Mesh or BMesh if available), and then use advanced helpers to extract needed data, on a per-item basis (no more handling of arrays of verts/edges/... in batches code). This allows to: * Avoid having to create arrays of BMesh elements. * Take advantage of existing advanced BMesh topology and connectivity data. Reviewers: dfelinto, fclem, merwin Differential Revision: https://developer.blender.org/D2521
2017-02-23 11:54:40 +01:00
*r_vco1 = mvert[medge[edge_index].v1].co;
*r_vco2 = mvert[medge[edge_index].v2].co;
if (eap[edge_index].count == 2) {
*r_pnor1 = pnors[eap[edge_index].face_index[0]];
*r_pnor2 = pnors[eap[edge_index].face_index[1]];
Draw Manager: T51218 Support edit-mesh hide
2017-04-18 23:29:55 +10:00
*r_is_manifold = true;
}
else {
*r_is_manifold = false;
Rework of BKE's mesh_render to support BMesh directly. Note that since there is no (efficient) ways to get arrays of MVert/MEdge/etc. out of a BMesh, I refactored quite heavily internals of BKE_mesh_render. Now, when you do need acess to mesh data to generate cached batches, you create an abstract struct from mesh (either Mesh or BMesh if available), and then use advanced helpers to extract needed data, on a per-item basis (no more handling of arrays of verts/edges/... in batches code). This allows to: * Avoid having to create arrays of BMesh elements. * Take advantage of existing advanced BMesh topology and connectivity data. Reviewers: dfelinto, fclem, merwin Differential Revision: https://developer.blender.org/D2521
2017-02-23 11:54:40 +01:00
}
}
Convert MBC_ API to Mesh (instead of derived mesh) and move it to mesh_render This includes a few fixes in the MBC_ api. The idea here is for this to be the only interface the render engines will deal with for the meshes. If we need to expose special options for sculpting engine we refactor this accordingly. But for now we are shaping this in a per-case base. Note: * We still need to hook up to the depsgraph to force clear/update of batch_cache when mesh changes (I'm waiting for Sergey Sharybin's depsgraph update for this though) * Also ideally we could/should use BMesh directly instead of DerivedMesh, but this will do for now. Note 2: In the end I renamed the `BKE_mesh_render` functions to `static mesh_render`. We can re-expose them as BKE_* later once we need it. Reviewers: merwin Subscribers: fclem Differential Revision: https://developer.blender.org/D2476
2017-01-25 09:16:29 +01:00
Draw Manager: T51218 Support edit-mesh hide
2017-04-18 23:29:55 +10:00
return true;
Convert MBC_ API to Mesh (instead of derived mesh) and move it to mesh_render This includes a few fixes in the MBC_ api. The idea here is for this to be the only interface the render engines will deal with for the meshes. If we need to expose special options for sculpting engine we refactor this accordingly. But for now we are shaping this in a per-case base. Note: * We still need to hook up to the depsgraph to force clear/update of batch_cache when mesh changes (I'm waiting for Sergey Sharybin's depsgraph update for this though) * Also ideally we could/should use BMesh directly instead of DerivedMesh, but this will do for now. Note 2: In the end I renamed the `BKE_mesh_render` functions to `static mesh_render`. We can re-expose them as BKE_* later once we need it. Reviewers: merwin Subscribers: fclem Differential Revision: https://developer.blender.org/D2476
2017-01-25 09:16:29 +01:00
}
Draw Manager: T51218 Support edit-mesh hide
2017-04-18 23:29:55 +10:00
static bool mesh_render_data_looptri_vert_indices_get(
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
const MeshRenderData *rdata, const int tri_idx,
EditMode: remove edge-hash from MeshRenderData LoopTri's can access this data.
2017-04-12 13:42:27 +10:00
int r_vert_idx[3])
Convert MBC_ API to Mesh (instead of derived mesh) and move it to mesh_render This includes a few fixes in the MBC_ api. The idea here is for this to be the only interface the render engines will deal with for the meshes. If we need to expose special options for sculpting engine we refactor this accordingly. But for now we are shaping this in a per-case base. Note: * We still need to hook up to the depsgraph to force clear/update of batch_cache when mesh changes (I'm waiting for Sergey Sharybin's depsgraph update for this though) * Also ideally we could/should use BMesh directly instead of DerivedMesh, but this will do for now. Note 2: In the end I renamed the `BKE_mesh_render` functions to `static mesh_render`. We can re-expose them as BKE_* later once we need it. Reviewers: merwin Subscribers: fclem Differential Revision: https://developer.blender.org/D2476
2017-01-25 09:16:29 +01:00
{
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
BLI_assert(rdata->types & (MR_DATATYPE_LOOPTRI | MR_DATATYPE_LOOP));
Convert MBC_ API to Mesh (instead of derived mesh) and move it to mesh_render This includes a few fixes in the MBC_ api. The idea here is for this to be the only interface the render engines will deal with for the meshes. If we need to expose special options for sculpting engine we refactor this accordingly. But for now we are shaping this in a per-case base. Note: * We still need to hook up to the depsgraph to force clear/update of batch_cache when mesh changes (I'm waiting for Sergey Sharybin's depsgraph update for this though) * Also ideally we could/should use BMesh directly instead of DerivedMesh, but this will do for now. Note 2: In the end I renamed the `BKE_mesh_render` functions to `static mesh_render`. We can re-expose them as BKE_* later once we need it. Reviewers: merwin Subscribers: fclem Differential Revision: https://developer.blender.org/D2476
2017-01-25 09:16:29 +01:00
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
if (rdata->edit_bmesh) {
const BMLoop **bm_looptri = (const BMLoop **)rdata->edit_bmesh->looptris[tri_idx];
Draw Manager: T51218 Support edit-mesh hide
2017-04-18 23:29:55 +10:00
if (BM_elem_flag_test(bm_looptri[0]->f, BM_ELEM_HIDDEN)) {
return false;
}
Rework of BKE's mesh_render to support BMesh directly. Note that since there is no (efficient) ways to get arrays of MVert/MEdge/etc. out of a BMesh, I refactored quite heavily internals of BKE_mesh_render. Now, when you do need acess to mesh data to generate cached batches, you create an abstract struct from mesh (either Mesh or BMesh if available), and then use advanced helpers to extract needed data, on a per-item basis (no more handling of arrays of verts/edges/... in batches code). This allows to: * Avoid having to create arrays of BMesh elements. * Take advantage of existing advanced BMesh topology and connectivity data. Reviewers: dfelinto, fclem, merwin Differential Revision: https://developer.blender.org/D2521
2017-02-23 11:54:40 +01:00
r_vert_idx[0] = BM_elem_index_get(bm_looptri[0]->v);
r_vert_idx[1] = BM_elem_index_get(bm_looptri[1]->v);
r_vert_idx[2] = BM_elem_index_get(bm_looptri[2]->v);
}
else {
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
const unsigned int *l_idx = rdata->mlooptri[tri_idx].tri;
const MLoop *l_tri[3] = {&rdata->mloop[l_idx[0]], &rdata->mloop[l_idx[1]], &rdata->mloop[l_idx[2]]};
EditMode: remove edge-hash from MeshRenderData LoopTri's can access this data.
2017-04-12 13:42:27 +10:00
r_vert_idx[0] = l_tri[0]->v;
r_vert_idx[1] = l_tri[1]->v;
r_vert_idx[2] = l_tri[2]->v;
}
Draw Manager: T51218 Support edit-mesh hide
2017-04-18 23:29:55 +10:00
return true;
EditMode: remove edge-hash from MeshRenderData LoopTri's can access this data.
2017-04-12 13:42:27 +10:00
}
Eevee: Make default shaders works. - Added New Batch cache routine to get the split mesh for each material. (optimization needed) - Did some shader modifications to create default_frag and support a somwhat correct fresnel for lamps (to finish)
2017-04-25 18:46:59 +02:00
static bool mesh_render_data_looptri_mat_index_get(
const MeshRenderData *rdata, const int tri_idx,
short *r_face_mat)
{
BLI_assert(rdata->types & (MR_DATATYPE_LOOPTRI | MR_DATATYPE_LOOP | MR_DATATYPE_POLY));
if (rdata->edit_bmesh) {
const BMLoop **bm_looptri = (const BMLoop **)rdata->edit_bmesh->looptris[tri_idx];
if (BM_elem_flag_test(bm_looptri[0]->f, BM_ELEM_HIDDEN)) {
return false;
}
*r_face_mat = ((BMFace *)bm_looptri[0]->f)->mat_nr;
}
else {
const int poly_idx = rdata->mlooptri[tri_idx].poly; ;
const MPoly *poly = &rdata->mpoly[poly_idx]; ;
*r_face_mat = poly->mat_nr;
}
return true;
}
EditMode: remove edge-hash from MeshRenderData LoopTri's can access this data.
2017-04-12 13:42:27 +10:00
/**
* Version of #mesh_render_data_looptri_verts_indices_get that assigns
* edge indices too \a r_edges_idx (-1 for non-existant edges).
*/
Draw Manager: T51218 Support edit-mesh hide
2017-04-18 23:29:55 +10:00
static bool mesh_render_data_looptri_vert_edge_indices_get(
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
const MeshRenderData *rdata, const int tri_idx,
EditMode: remove edge-hash from MeshRenderData LoopTri's can access this data.
2017-04-12 13:42:27 +10:00
int r_vert_idx[3], int r_edges_idx[3])
{
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
BLI_assert(rdata->types & (MR_DATATYPE_LOOPTRI | MR_DATATYPE_LOOP));
EditMode: remove edge-hash from MeshRenderData LoopTri's can access this data.
2017-04-12 13:42:27 +10:00
unsigned int e_pair_edge[2];
unsigned int e_pair_loop[2];
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
if (rdata->edit_bmesh) {
const BMLoop **bm_looptri = (const BMLoop **)rdata->edit_bmesh->looptris[tri_idx];
EditMode: remove edge-hash from MeshRenderData LoopTri's can access this data.
2017-04-12 13:42:27 +10:00
Draw Manager: T51218 Support edit-mesh hide
2017-04-18 23:29:55 +10:00
if (BM_elem_flag_test(bm_looptri[0]->f, BM_ELEM_HIDDEN)) {
return false;
}
Cleanup: replace char array with struct
2017-04-12 14:08:44 +10:00
/* assign 'r_edges_idx' & 'r_vert_idx' */
EditMode: remove edge-hash from MeshRenderData LoopTri's can access this data.
2017-04-12 13:42:27 +10:00
int j, j_next;
for (j = 2, j_next = 0; j_next < 3; j = j_next++) {
const BMLoop *l = bm_looptri[j], *l_next = bm_looptri[j_next];
const BMEdge *e = l->e;
ARRAY_SET_ITEMS(e_pair_edge, BM_elem_index_get(e->v1), BM_elem_index_get(e->v2));
ARRAY_SET_ITEMS(e_pair_loop, BM_elem_index_get(l->v), BM_elem_index_get(l_next->v));
if ((e_pair_edge[0] == e_pair_loop[0] && e_pair_edge[1] == e_pair_loop[1]) ||
(e_pair_edge[0] == e_pair_loop[1] && e_pair_edge[1] == e_pair_loop[0]))
{
r_edges_idx[j] = BM_elem_index_get(l->e);
}
else {
r_edges_idx[j] = -1;
}
r_vert_idx[j] = e_pair_loop[0]; /* BM_elem_index_get(l->v) */
}
}
else {
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
const unsigned int *l_idx = rdata->mlooptri[tri_idx].tri;
const MLoop *l_tri[3] = {&rdata->mloop[l_idx[0]], &rdata->mloop[l_idx[1]], &rdata->mloop[l_idx[2]]};
EditMode: remove edge-hash from MeshRenderData LoopTri's can access this data.
2017-04-12 13:42:27 +10:00
Cleanup: replace char array with struct
2017-04-12 14:08:44 +10:00
/* assign 'r_edges_idx' & 'r_vert_idx' */
EditMode: remove edge-hash from MeshRenderData LoopTri's can access this data.
2017-04-12 13:42:27 +10:00
int j, j_next;
for (j = 2, j_next = 0; j_next < 3; j = j_next++) {
const MLoop *l = l_tri[j], *l_next = l_tri[j_next];
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
const MEdge *e = &rdata->medge[l->e]; \
EditMode: remove edge-hash from MeshRenderData LoopTri's can access this data.
2017-04-12 13:42:27 +10:00
ARRAY_SET_ITEMS(e_pair_edge, e->v1, e->v2);
ARRAY_SET_ITEMS(e_pair_loop, l->v, l_next->v);
if ((e_pair_edge[0] == e_pair_loop[0] && e_pair_edge[1] == e_pair_loop[1]) ||
(e_pair_edge[0] == e_pair_loop[1] && e_pair_edge[1] == e_pair_loop[0]))
{
r_edges_idx[j] = l->e;
}
else {
r_edges_idx[j] = -1;
}
r_vert_idx[j] = e_pair_loop[0]; /* l->v */
}
Rework of BKE's mesh_render to support BMesh directly. Note that since there is no (efficient) ways to get arrays of MVert/MEdge/etc. out of a BMesh, I refactored quite heavily internals of BKE_mesh_render. Now, when you do need acess to mesh data to generate cached batches, you create an abstract struct from mesh (either Mesh or BMesh if available), and then use advanced helpers to extract needed data, on a per-item basis (no more handling of arrays of verts/edges/... in batches code). This allows to: * Avoid having to create arrays of BMesh elements. * Take advantage of existing advanced BMesh topology and connectivity data. Reviewers: dfelinto, fclem, merwin Differential Revision: https://developer.blender.org/D2521
2017-02-23 11:54:40 +01:00
}
Draw Manager: T51218 Support edit-mesh hide
2017-04-18 23:29:55 +10:00
return true;
Convert MBC_ API to Mesh (instead of derived mesh) and move it to mesh_render This includes a few fixes in the MBC_ api. The idea here is for this to be the only interface the render engines will deal with for the meshes. If we need to expose special options for sculpting engine we refactor this accordingly. But for now we are shaping this in a per-case base. Note: * We still need to hook up to the depsgraph to force clear/update of batch_cache when mesh changes (I'm waiting for Sergey Sharybin's depsgraph update for this though) * Also ideally we could/should use BMesh directly instead of DerivedMesh, but this will do for now. Note 2: In the end I renamed the `BKE_mesh_render` functions to `static mesh_render`. We can re-expose them as BKE_* later once we need it. Reviewers: merwin Subscribers: fclem Differential Revision: https://developer.blender.org/D2476
2017-01-25 09:16:29 +01:00
}
Mesh Batch Cache: Add support for uv/vcol/tangent data layers. This is a bad implementation waiting for some improvement that are : - Gawain support for multiple names. -> will remove duplication of attribute alloc - Glsl safe name string escaping. -> will remove any risk of hash colision. Old glsl drawing was relying on attribute being choosen by DerivedMesh at drawing time. For this reason, we declare all possible attrib "name" for each data layer inside the batches. In the glsl code, we declare required data by type and name. Then Gawain only bind vbos if names correspond. This is way cleaner as we do not need to access the CD itself when drawing. One other problem is that the hash maybe 11 caracters long and rapidly overload gawain's attrib name buffer.
2017-05-01 17:58:15 +02:00
static void mesh_render_data_looptri_uvs_get(
MeshRenderData *rdata, const int tri_idx, const int uv_layer,
float *(*r_vert_uvs)[3])
{
if (rdata->edit_bmesh) {
const BMLoop **bm_looptri = (const BMLoop **)rdata->edit_bmesh->looptris[tri_idx];
(*r_vert_uvs)[0] = ((MLoopUV *)BM_ELEM_CD_GET_VOID_P(bm_looptri[0], rdata->uv_ofs[uv_layer]))->uv;
(*r_vert_uvs)[1] = ((MLoopUV *)BM_ELEM_CD_GET_VOID_P(bm_looptri[1], rdata->uv_ofs[uv_layer]))->uv;
(*r_vert_uvs)[2] = ((MLoopUV *)BM_ELEM_CD_GET_VOID_P(bm_looptri[2], rdata->uv_ofs[uv_layer]))->uv;
}
else {
const MLoopTri *mlt = &rdata->mlooptri[tri_idx];
(*r_vert_uvs)[0] = rdata->mloopuv[uv_layer][mlt->tri[0]].uv;
(*r_vert_uvs)[1] = rdata->mloopuv[uv_layer][mlt->tri[1]].uv;
(*r_vert_uvs)[2] = rdata->mloopuv[uv_layer][mlt->tri[2]].uv;
}
}
static void mesh_render_data_looptri_cols_get(
MeshRenderData *rdata, const int tri_idx, const int vcol_layer,
unsigned char *(*r_vert_cols)[3])
{
if (rdata->edit_bmesh) {
const BMLoop **bm_looptri = (const BMLoop **)rdata->edit_bmesh->looptris[tri_idx];
(*r_vert_cols)[0] = &((MCol *)BM_ELEM_CD_GET_VOID_P(bm_looptri[0], rdata->vcol_ofs[vcol_layer]))->a;
(*r_vert_cols)[1] = &((MCol *)BM_ELEM_CD_GET_VOID_P(bm_looptri[1], rdata->vcol_ofs[vcol_layer]))->a;
(*r_vert_cols)[2] = &((MCol *)BM_ELEM_CD_GET_VOID_P(bm_looptri[2], rdata->vcol_ofs[vcol_layer]))->a;
}
else {
const MLoopTri *mlt = &rdata->mlooptri[tri_idx];
(*r_vert_cols)[0] = &rdata->mcol[vcol_layer][mlt->tri[0]].a;
(*r_vert_cols)[1] = &rdata->mcol[vcol_layer][mlt->tri[1]].a;
(*r_vert_cols)[2] = &rdata->mcol[vcol_layer][mlt->tri[2]].a;
}
}
static void mesh_render_data_looptri_tans_get(
MeshRenderData *rdata, const int tri_idx, const int tangent_layer,
float *(*r_vert_tans)[3])
{
if (rdata->edit_bmesh) {
#if 0 /* waiting for edit mesh tangent calculation */
const BMLoop **bm_looptri = (const BMLoop **)rdata->edit_bmesh->looptris[tri_idx];
(*r_vert_tans)[0] = ((float *)BM_ELEM_CD_GET_VOID_P(bm_looptri[0], rdata->tangent_ofs[tangent_layer]));
(*r_vert_tans)[1] = ((float *)BM_ELEM_CD_GET_VOID_P(bm_looptri[1], rdata->tangent_ofs[tangent_layer]));
(*r_vert_tans)[2] = ((float *)BM_ELEM_CD_GET_VOID_P(bm_looptri[2], rdata->tangent_ofs[tangent_layer]));
#else
static float tan[4] = {0.0f};
(*r_vert_tans)[0] = tan;
(*r_vert_tans)[1] = tan;
(*r_vert_tans)[2] = tan;
#endif
}
else {
const MLoopTri *mlt = &rdata->mlooptri[tri_idx];
(*r_vert_tans)[0] = rdata->mtangent[tangent_layer][mlt->tri[0]];
(*r_vert_tans)[1] = rdata->mtangent[tangent_layer][mlt->tri[1]];
(*r_vert_tans)[2] = rdata->mtangent[tangent_layer][mlt->tri[2]];
}
}
static void mesh_render_data_looptri_orcos_get(
MeshRenderData *rdata, const int tri_idx,
float *(*r_vert_orcos)[3])
{
BLI_assert(rdata->types & (MR_DATATYPE_VERT | MR_DATATYPE_LOOPTRI | MR_DATATYPE_LOOP | MR_DATATYPE_SHADING));
if (rdata->edit_bmesh) {
const BMLoop **bm_looptri = (const BMLoop **)rdata->edit_bmesh->looptris[tri_idx];
(*r_vert_orcos)[0] = rdata->orco[BM_elem_index_get(bm_looptri[0]->v)];
(*r_vert_orcos)[1] = rdata->orco[BM_elem_index_get(bm_looptri[1]->v)];
(*r_vert_orcos)[2] = rdata->orco[BM_elem_index_get(bm_looptri[2]->v)];
}
else {
const MLoopTri *mlt = &rdata->mlooptri[tri_idx];
(*r_vert_orcos)[0] = rdata->orco[rdata->mloop[mlt->tri[0]].v];
(*r_vert_orcos)[1] = rdata->orco[rdata->mloop[mlt->tri[1]].v];
(*r_vert_orcos)[2] = rdata->orco[rdata->mloop[mlt->tri[2]].v];
}
}
Rework of BKE's mesh_render to support BMesh directly. Note that since there is no (efficient) ways to get arrays of MVert/MEdge/etc. out of a BMesh, I refactored quite heavily internals of BKE_mesh_render. Now, when you do need acess to mesh data to generate cached batches, you create an abstract struct from mesh (either Mesh or BMesh if available), and then use advanced helpers to extract needed data, on a per-item basis (no more handling of arrays of verts/edges/... in batches code). This allows to: * Avoid having to create arrays of BMesh elements. * Take advantage of existing advanced BMesh topology and connectivity data. Reviewers: dfelinto, fclem, merwin Differential Revision: https://developer.blender.org/D2521
2017-02-23 11:54:40 +01:00
static bool mesh_render_data_looptri_cos_nors_smooth_get(
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
MeshRenderData *rdata, const int tri_idx,
Draw Manager: T51218 Support edit-mesh hide
2017-04-18 23:29:55 +10:00
float *(*r_vert_cos)[3], short **r_tri_nor, short *(*r_vert_nors)[3], bool *r_is_smooth)
Convert MBC_ API to Mesh (instead of derived mesh) and move it to mesh_render This includes a few fixes in the MBC_ api. The idea here is for this to be the only interface the render engines will deal with for the meshes. If we need to expose special options for sculpting engine we refactor this accordingly. But for now we are shaping this in a per-case base. Note: * We still need to hook up to the depsgraph to force clear/update of batch_cache when mesh changes (I'm waiting for Sergey Sharybin's depsgraph update for this though) * Also ideally we could/should use BMesh directly instead of DerivedMesh, but this will do for now. Note 2: In the end I renamed the `BKE_mesh_render` functions to `static mesh_render`. We can re-expose them as BKE_* later once we need it. Reviewers: merwin Subscribers: fclem Differential Revision: https://developer.blender.org/D2476
2017-01-25 09:16:29 +01:00
{
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
BLI_assert(rdata->types & (MR_DATATYPE_VERT | MR_DATATYPE_LOOPTRI | MR_DATATYPE_LOOP | MR_DATATYPE_POLY));
Convert MBC_ API to Mesh (instead of derived mesh) and move it to mesh_render This includes a few fixes in the MBC_ api. The idea here is for this to be the only interface the render engines will deal with for the meshes. If we need to expose special options for sculpting engine we refactor this accordingly. But for now we are shaping this in a per-case base. Note: * We still need to hook up to the depsgraph to force clear/update of batch_cache when mesh changes (I'm waiting for Sergey Sharybin's depsgraph update for this though) * Also ideally we could/should use BMesh directly instead of DerivedMesh, but this will do for now. Note 2: In the end I renamed the `BKE_mesh_render` functions to `static mesh_render`. We can re-expose them as BKE_* later once we need it. Reviewers: merwin Subscribers: fclem Differential Revision: https://developer.blender.org/D2476
2017-01-25 09:16:29 +01:00
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
if (rdata->edit_bmesh) {
const BMLoop **bm_looptri = (const BMLoop **)rdata->edit_bmesh->looptris[tri_idx];
Draw Manager: T51218 Support edit-mesh hide
2017-04-18 23:29:55 +10:00
if (BM_elem_flag_test(bm_looptri[0]->f, BM_ELEM_HIDDEN)) {
return false;
}
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
short (*pnors_short)[3] = rdata->poly_normals_short;
short (*vnors_short)[3] = rdata->vert_normals_short;
Rework of BKE's mesh_render to support BMesh directly. Note that since there is no (efficient) ways to get arrays of MVert/MEdge/etc. out of a BMesh, I refactored quite heavily internals of BKE_mesh_render. Now, when you do need acess to mesh data to generate cached batches, you create an abstract struct from mesh (either Mesh or BMesh if available), and then use advanced helpers to extract needed data, on a per-item basis (no more handling of arrays of verts/edges/... in batches code). This allows to: * Avoid having to create arrays of BMesh elements. * Take advantage of existing advanced BMesh topology and connectivity data. Reviewers: dfelinto, fclem, merwin Differential Revision: https://developer.blender.org/D2521
2017-02-23 11:54:40 +01:00
if (!pnors_short) {
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
BMesh *bm = rdata->edit_bmesh->bm;
Rework of BKE's mesh_render to support BMesh directly. Note that since there is no (efficient) ways to get arrays of MVert/MEdge/etc. out of a BMesh, I refactored quite heavily internals of BKE_mesh_render. Now, when you do need acess to mesh data to generate cached batches, you create an abstract struct from mesh (either Mesh or BMesh if available), and then use advanced helpers to extract needed data, on a per-item basis (no more handling of arrays of verts/edges/... in batches code). This allows to: * Avoid having to create arrays of BMesh elements. * Take advantage of existing advanced BMesh topology and connectivity data. Reviewers: dfelinto, fclem, merwin Differential Revision: https://developer.blender.org/D2521
2017-02-23 11:54:40 +01:00
BMIter fiter;
BMFace *face;
int i;
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
pnors_short = rdata->poly_normals_short = MEM_mallocN(sizeof(*pnors_short) * rdata->poly_len, __func__);
Rework of BKE's mesh_render to support BMesh directly. Note that since there is no (efficient) ways to get arrays of MVert/MEdge/etc. out of a BMesh, I refactored quite heavily internals of BKE_mesh_render. Now, when you do need acess to mesh data to generate cached batches, you create an abstract struct from mesh (either Mesh or BMesh if available), and then use advanced helpers to extract needed data, on a per-item basis (no more handling of arrays of verts/edges/... in batches code). This allows to: * Avoid having to create arrays of BMesh elements. * Take advantage of existing advanced BMesh topology and connectivity data. Reviewers: dfelinto, fclem, merwin Differential Revision: https://developer.blender.org/D2521
2017-02-23 11:54:40 +01:00
BM_ITER_MESH_INDEX(face, &fiter, bm, BM_FACES_OF_MESH, i) {
normal_float_to_short_v3(pnors_short[i], face->no);
}
}
if (!vnors_short) {
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
BMesh *bm = rdata->edit_bmesh->bm;
Rework of BKE's mesh_render to support BMesh directly. Note that since there is no (efficient) ways to get arrays of MVert/MEdge/etc. out of a BMesh, I refactored quite heavily internals of BKE_mesh_render. Now, when you do need acess to mesh data to generate cached batches, you create an abstract struct from mesh (either Mesh or BMesh if available), and then use advanced helpers to extract needed data, on a per-item basis (no more handling of arrays of verts/edges/... in batches code). This allows to: * Avoid having to create arrays of BMesh elements. * Take advantage of existing advanced BMesh topology and connectivity data. Reviewers: dfelinto, fclem, merwin Differential Revision: https://developer.blender.org/D2521
2017-02-23 11:54:40 +01:00
BMIter viter;
BMVert *vert;
int i;
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
vnors_short = rdata->vert_normals_short = MEM_mallocN(sizeof(*vnors_short) * rdata->vert_len, __func__);
Rework of BKE's mesh_render to support BMesh directly. Note that since there is no (efficient) ways to get arrays of MVert/MEdge/etc. out of a BMesh, I refactored quite heavily internals of BKE_mesh_render. Now, when you do need acess to mesh data to generate cached batches, you create an abstract struct from mesh (either Mesh or BMesh if available), and then use advanced helpers to extract needed data, on a per-item basis (no more handling of arrays of verts/edges/... in batches code). This allows to: * Avoid having to create arrays of BMesh elements. * Take advantage of existing advanced BMesh topology and connectivity data. Reviewers: dfelinto, fclem, merwin Differential Revision: https://developer.blender.org/D2521
2017-02-23 11:54:40 +01:00
BM_ITER_MESH_INDEX(vert, &viter, bm, BM_VERT, i) {
normal_float_to_short_v3(vnors_short[i], vert->no);
}
}
(*r_vert_cos)[0] = bm_looptri[0]->v->co;
(*r_vert_cos)[1] = bm_looptri[1]->v->co;
(*r_vert_cos)[2] = bm_looptri[2]->v->co;
*r_tri_nor = pnors_short[BM_elem_index_get(bm_looptri[0]->f)];
(*r_vert_nors)[0] = vnors_short[BM_elem_index_get(bm_looptri[0]->v)];
(*r_vert_nors)[1] = vnors_short[BM_elem_index_get(bm_looptri[1]->v)];
(*r_vert_nors)[2] = vnors_short[BM_elem_index_get(bm_looptri[2]->v)];
Draw Manager: T51218 Support edit-mesh hide
2017-04-18 23:29:55 +10:00
*r_is_smooth = BM_elem_flag_test_bool(bm_looptri[0]->f, BM_ELEM_SMOOTH);
Rework of BKE's mesh_render to support BMesh directly. Note that since there is no (efficient) ways to get arrays of MVert/MEdge/etc. out of a BMesh, I refactored quite heavily internals of BKE_mesh_render. Now, when you do need acess to mesh data to generate cached batches, you create an abstract struct from mesh (either Mesh or BMesh if available), and then use advanced helpers to extract needed data, on a per-item basis (no more handling of arrays of verts/edges/... in batches code). This allows to: * Avoid having to create arrays of BMesh elements. * Take advantage of existing advanced BMesh topology and connectivity data. Reviewers: dfelinto, fclem, merwin Differential Revision: https://developer.blender.org/D2521
2017-02-23 11:54:40 +01:00
}
else {
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
const MLoopTri *mlt = &rdata->mlooptri[tri_idx];
short (*pnors_short)[3] = rdata->poly_normals_short;
Rework of BKE's mesh_render to support BMesh directly. Note that since there is no (efficient) ways to get arrays of MVert/MEdge/etc. out of a BMesh, I refactored quite heavily internals of BKE_mesh_render. Now, when you do need acess to mesh data to generate cached batches, you create an abstract struct from mesh (either Mesh or BMesh if available), and then use advanced helpers to extract needed data, on a per-item basis (no more handling of arrays of verts/edges/... in batches code). This allows to: * Avoid having to create arrays of BMesh elements. * Take advantage of existing advanced BMesh topology and connectivity data. Reviewers: dfelinto, fclem, merwin Differential Revision: https://developer.blender.org/D2521
2017-02-23 11:54:40 +01:00
if (!pnors_short) {
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
float (*pnors)[3] = rdata->poly_normals;
Rework of BKE's mesh_render to support BMesh directly. Note that since there is no (efficient) ways to get arrays of MVert/MEdge/etc. out of a BMesh, I refactored quite heavily internals of BKE_mesh_render. Now, when you do need acess to mesh data to generate cached batches, you create an abstract struct from mesh (either Mesh or BMesh if available), and then use advanced helpers to extract needed data, on a per-item basis (no more handling of arrays of verts/edges/... in batches code). This allows to: * Avoid having to create arrays of BMesh elements. * Take advantage of existing advanced BMesh topology and connectivity data. Reviewers: dfelinto, fclem, merwin Differential Revision: https://developer.blender.org/D2521
2017-02-23 11:54:40 +01:00
if (!pnors) {
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
pnors = rdata->poly_normals = MEM_mallocN(sizeof(*pnors) * rdata->poly_len, __func__);
Rework of BKE's mesh_render to support BMesh directly. Note that since there is no (efficient) ways to get arrays of MVert/MEdge/etc. out of a BMesh, I refactored quite heavily internals of BKE_mesh_render. Now, when you do need acess to mesh data to generate cached batches, you create an abstract struct from mesh (either Mesh or BMesh if available), and then use advanced helpers to extract needed data, on a per-item basis (no more handling of arrays of verts/edges/... in batches code). This allows to: * Avoid having to create arrays of BMesh elements. * Take advantage of existing advanced BMesh topology and connectivity data. Reviewers: dfelinto, fclem, merwin Differential Revision: https://developer.blender.org/D2521
2017-02-23 11:54:40 +01:00
BKE_mesh_calc_normals_poly(
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
rdata->mvert, NULL, rdata->vert_len,
rdata->mloop, rdata->mpoly, rdata->loop_len, rdata->poly_len, pnors, true);
Rework of BKE's mesh_render to support BMesh directly. Note that since there is no (efficient) ways to get arrays of MVert/MEdge/etc. out of a BMesh, I refactored quite heavily internals of BKE_mesh_render. Now, when you do need acess to mesh data to generate cached batches, you create an abstract struct from mesh (either Mesh or BMesh if available), and then use advanced helpers to extract needed data, on a per-item basis (no more handling of arrays of verts/edges/... in batches code). This allows to: * Avoid having to create arrays of BMesh elements. * Take advantage of existing advanced BMesh topology and connectivity data. Reviewers: dfelinto, fclem, merwin Differential Revision: https://developer.blender.org/D2521
2017-02-23 11:54:40 +01:00
}
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
pnors_short = rdata->poly_normals_short = MEM_mallocN(sizeof(*pnors_short) * rdata->poly_len, __func__);
for (int i = 0; i < rdata->poly_len; i++) {
Rework of BKE's mesh_render to support BMesh directly. Note that since there is no (efficient) ways to get arrays of MVert/MEdge/etc. out of a BMesh, I refactored quite heavily internals of BKE_mesh_render. Now, when you do need acess to mesh data to generate cached batches, you create an abstract struct from mesh (either Mesh or BMesh if available), and then use advanced helpers to extract needed data, on a per-item basis (no more handling of arrays of verts/edges/... in batches code). This allows to: * Avoid having to create arrays of BMesh elements. * Take advantage of existing advanced BMesh topology and connectivity data. Reviewers: dfelinto, fclem, merwin Differential Revision: https://developer.blender.org/D2521
2017-02-23 11:54:40 +01:00
normal_float_to_short_v3(pnors_short[i], pnors[i]);
}
}
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
(*r_vert_cos)[0] = rdata->mvert[rdata->mloop[mlt->tri[0]].v].co;
(*r_vert_cos)[1] = rdata->mvert[rdata->mloop[mlt->tri[1]].v].co;
(*r_vert_cos)[2] = rdata->mvert[rdata->mloop[mlt->tri[2]].v].co;
Rework of BKE's mesh_render to support BMesh directly. Note that since there is no (efficient) ways to get arrays of MVert/MEdge/etc. out of a BMesh, I refactored quite heavily internals of BKE_mesh_render. Now, when you do need acess to mesh data to generate cached batches, you create an abstract struct from mesh (either Mesh or BMesh if available), and then use advanced helpers to extract needed data, on a per-item basis (no more handling of arrays of verts/edges/... in batches code). This allows to: * Avoid having to create arrays of BMesh elements. * Take advantage of existing advanced BMesh topology and connectivity data. Reviewers: dfelinto, fclem, merwin Differential Revision: https://developer.blender.org/D2521
2017-02-23 11:54:40 +01:00
*r_tri_nor = pnors_short[mlt->poly];
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
(*r_vert_nors)[0] = rdata->mvert[rdata->mloop[mlt->tri[0]].v].no;
(*r_vert_nors)[1] = rdata->mvert[rdata->mloop[mlt->tri[1]].v].no;
(*r_vert_nors)[2] = rdata->mvert[rdata->mloop[mlt->tri[2]].v].no;
Rework of BKE's mesh_render to support BMesh directly. Note that since there is no (efficient) ways to get arrays of MVert/MEdge/etc. out of a BMesh, I refactored quite heavily internals of BKE_mesh_render. Now, when you do need acess to mesh data to generate cached batches, you create an abstract struct from mesh (either Mesh or BMesh if available), and then use advanced helpers to extract needed data, on a per-item basis (no more handling of arrays of verts/edges/... in batches code). This allows to: * Avoid having to create arrays of BMesh elements. * Take advantage of existing advanced BMesh topology and connectivity data. Reviewers: dfelinto, fclem, merwin Differential Revision: https://developer.blender.org/D2521
2017-02-23 11:54:40 +01:00
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
*r_is_smooth = (rdata->mpoly[mlt->poly].flag & ME_SMOOTH) != 0;
Rework of BKE's mesh_render to support BMesh directly. Note that since there is no (efficient) ways to get arrays of MVert/MEdge/etc. out of a BMesh, I refactored quite heavily internals of BKE_mesh_render. Now, when you do need acess to mesh data to generate cached batches, you create an abstract struct from mesh (either Mesh or BMesh if available), and then use advanced helpers to extract needed data, on a per-item basis (no more handling of arrays of verts/edges/... in batches code). This allows to: * Avoid having to create arrays of BMesh elements. * Take advantage of existing advanced BMesh topology and connectivity data. Reviewers: dfelinto, fclem, merwin Differential Revision: https://developer.blender.org/D2521
2017-02-23 11:54:40 +01:00
}
Draw Manager: T51218 Support edit-mesh hide
2017-04-18 23:29:55 +10:00
return true;
Convert MBC_ API to Mesh (instead of derived mesh) and move it to mesh_render This includes a few fixes in the MBC_ api. The idea here is for this to be the only interface the render engines will deal with for the meshes. If we need to expose special options for sculpting engine we refactor this accordingly. But for now we are shaping this in a per-case base. Note: * We still need to hook up to the depsgraph to force clear/update of batch_cache when mesh changes (I'm waiting for Sergey Sharybin's depsgraph update for this though) * Also ideally we could/should use BMesh directly instead of DerivedMesh, but this will do for now. Note 2: In the end I renamed the `BKE_mesh_render` functions to `static mesh_render`. We can re-expose them as BKE_* later once we need it. Reviewers: merwin Subscribers: fclem Differential Revision: https://developer.blender.org/D2476
2017-01-25 09:16:29 +01:00
}
Edit Mode overlays: separate multiple shaders for loose edges and verts
2017-03-01 14:07:04 +01:00
/* First 2 bytes are bit flags
* 3rd is for sharp edges
* 4rd is for creased edges */
enum {
VFLAG_VERTEX_ACTIVE = 1 << 0,
VFLAG_VERTEX_SELECTED = 1 << 1,
VFLAG_FACE_ACTIVE = 1 << 2,
VFLAG_FACE_SELECTED = 1 << 3,
};
enum {
VFLAG_EDGE_EXISTS = 1 << 0,
VFLAG_EDGE_ACTIVE = 1 << 1,
VFLAG_EDGE_SELECTED = 1 << 2,
VFLAG_EDGE_SEAM = 1 << 3,
VFLAG_EDGE_SHARP = 1 << 4,
/* Beware to not go over 1 << 7
GLSL: rename edit_overlay -> edit_mesh_overlay Avoid confusion with overlays with non-mesh types
2017-04-13 14:13:46 +10:00
* (see gpu_shader_edit_mesh_overlay_geom.glsl) */
Edit Mode overlays: separate multiple shaders for loose edges and verts
2017-03-01 14:07:04 +01:00
};
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
static unsigned char mesh_render_data_looptri_flag(MeshRenderData *rdata, const int f)
Edit Mode overlays: separate multiple shaders for loose edges and verts
2017-03-01 14:07:04 +01:00
{
unsigned char fflag = 0;
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
if (rdata->edit_bmesh) {
BMFace *bf = rdata->edit_bmesh->looptris[f][0]->f;
Edit Mode overlays: separate multiple shaders for loose edges and verts
2017-03-01 14:07:04 +01:00
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
if (bf == rdata->efa_act)
Edit Mode overlays: separate multiple shaders for loose edges and verts
2017-03-01 14:07:04 +01:00
fflag |= VFLAG_FACE_ACTIVE;
if (BM_elem_flag_test(bf, BM_ELEM_SELECT))
fflag |= VFLAG_FACE_SELECTED;
}
return fflag;
}
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
static EdgeDrawAttr *mesh_render_data_edge_flag(MeshRenderData *rdata, const int e)
Edit Mode overlays: separate multiple shaders for loose edges and verts
2017-03-01 14:07:04 +01:00
{
Cleanup: replace char array with struct
2017-04-12 14:08:44 +10:00
static EdgeDrawAttr eattr;
memset(&eattr, 0, sizeof(eattr));
Edit Mode overlays: separate multiple shaders for loose edges and verts
2017-03-01 14:07:04 +01:00
EditMode: remove edge-hash from MeshRenderData LoopTri's can access this data.
2017-04-12 13:42:27 +10:00
if (e == -1) {
Cleanup: replace char array with struct
2017-04-12 14:08:44 +10:00
return &eattr;
EditMode: remove edge-hash from MeshRenderData LoopTri's can access this data.
2017-04-12 13:42:27 +10:00
}
Edit Mode overlays: separate multiple shaders for loose edges and verts
2017-03-01 14:07:04 +01:00
/* if edge exists */
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
if (rdata->edit_bmesh) {
BMesh *bm = rdata->edit_bmesh->bm;
Edit Mode overlays: separate multiple shaders for loose edges and verts
2017-03-01 14:07:04 +01:00
BMEdge *be = NULL;
EditMode: remove edge-hash from MeshRenderData LoopTri's can access this data.
2017-04-12 13:42:27 +10:00
be = BM_edge_at_index(bm, e);
Edit Mode overlays: separate multiple shaders for loose edges and verts
2017-03-01 14:07:04 +01:00
Cleanup: replace char array with struct
2017-04-12 14:08:44 +10:00
eattr.e_flag |= VFLAG_EDGE_EXISTS;
Edit Mode overlays: separate multiple shaders for loose edges and verts
2017-03-01 14:07:04 +01:00
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
if (be == rdata->eed_act)
Cleanup: replace char array with struct
2017-04-12 14:08:44 +10:00
eattr.e_flag |= VFLAG_EDGE_ACTIVE;
Edit Mode overlays: separate multiple shaders for loose edges and verts
2017-03-01 14:07:04 +01:00
EditMode: remove edge-hash from MeshRenderData LoopTri's can access this data.
2017-04-12 13:42:27 +10:00
if (BM_elem_flag_test(be, BM_ELEM_SELECT))
Cleanup: replace char array with struct
2017-04-12 14:08:44 +10:00
eattr.e_flag |= VFLAG_EDGE_SELECTED;
Edit Mode overlays: separate multiple shaders for loose edges and verts
2017-03-01 14:07:04 +01:00
EditMode: remove edge-hash from MeshRenderData LoopTri's can access this data.
2017-04-12 13:42:27 +10:00
if (BM_elem_flag_test(be, BM_ELEM_SEAM))
Cleanup: replace char array with struct
2017-04-12 14:08:44 +10:00
eattr.e_flag |= VFLAG_EDGE_SEAM;
Edit Mode overlays: separate multiple shaders for loose edges and verts
2017-03-01 14:07:04 +01:00
EditMode: remove edge-hash from MeshRenderData LoopTri's can access this data.
2017-04-12 13:42:27 +10:00
if (!BM_elem_flag_test(be, BM_ELEM_SMOOTH))
Cleanup: replace char array with struct
2017-04-12 14:08:44 +10:00
eattr.e_flag |= VFLAG_EDGE_SHARP;
Edit Mode overlays: separate multiple shaders for loose edges and verts
2017-03-01 14:07:04 +01:00
EditMode: remove edge-hash from MeshRenderData LoopTri's can access this data.
2017-04-12 13:42:27 +10:00
/* Use a byte for value range */
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
if (rdata->crease_ofs != -1) {
float crease = BM_ELEM_CD_GET_FLOAT(be, rdata->crease_ofs);
EditMode: remove edge-hash from MeshRenderData LoopTri's can access this data.
2017-04-12 13:42:27 +10:00
if (crease > 0) {
Cleanup: replace char array with struct
2017-04-12 14:08:44 +10:00
eattr.crease = (char)(crease * 255.0f);
Edit Mode overlays: separate multiple shaders for loose edges and verts
2017-03-01 14:07:04 +01:00
}
EditMode: remove edge-hash from MeshRenderData LoopTri's can access this data.
2017-04-12 13:42:27 +10:00
}
Edit Mode overlays: separate multiple shaders for loose edges and verts
2017-03-01 14:07:04 +01:00
EditMode: remove edge-hash from MeshRenderData LoopTri's can access this data.
2017-04-12 13:42:27 +10:00
/* Use a byte for value range */
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
if (rdata->bweight_ofs != -1) {
float bweight = BM_ELEM_CD_GET_FLOAT(be, rdata->bweight_ofs);
EditMode: remove edge-hash from MeshRenderData LoopTri's can access this data.
2017-04-12 13:42:27 +10:00
if (bweight > 0) {
Cleanup: replace char array with struct
2017-04-12 14:08:44 +10:00
eattr.bweight = (char)(bweight * 255.0f);
Edit Mode overlays: separate multiple shaders for loose edges and verts
2017-03-01 14:07:04 +01:00
}
}
}
EditMode: remove edge-hash from MeshRenderData LoopTri's can access this data.
2017-04-12 13:42:27 +10:00
else {
Cleanup: replace char array with struct
2017-04-12 14:08:44 +10:00
eattr.e_flag |= VFLAG_EDGE_EXISTS;
Edit Mode overlays: separate multiple shaders for loose edges and verts
2017-03-01 14:07:04 +01:00
}
Cleanup: replace char array with struct
2017-04-12 14:08:44 +10:00
return &eattr;
Edit Mode overlays: separate multiple shaders for loose edges and verts
2017-03-01 14:07:04 +01:00
}
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
static unsigned char mesh_render_data_vertex_flag(MeshRenderData *rdata, const int v)
Edit Mode overlays: separate multiple shaders for loose edges and verts
2017-03-01 14:07:04 +01:00
{
unsigned char vflag = 0;
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
if (rdata->edit_bmesh) {
BMesh *bm = rdata->edit_bmesh->bm;
Edit Mode overlays: separate multiple shaders for loose edges and verts
2017-03-01 14:07:04 +01:00
BMVert *bv = BM_vert_at_index(bm, v);
/* Current vertex */
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
if (bv == rdata->eve_act)
Edit Mode overlays: separate multiple shaders for loose edges and verts
2017-03-01 14:07:04 +01:00
vflag |= VFLAG_VERTEX_ACTIVE;
if (BM_elem_flag_test(bv, BM_ELEM_SELECT))
vflag |= VFLAG_VERTEX_SELECTED;
}
return vflag;
}
static void add_overlay_tri(
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
MeshRenderData *rdata, VertexBuffer *vbo, const unsigned int pos_id, const unsigned int edgeMod_id,
EditMode: remove edge-hash from MeshRenderData LoopTri's can access this data.
2017-04-12 13:42:27 +10:00
const int tri_vert_idx[3], const int tri_edge_idx[3], const int f, const int base_vert_idx)
Edit Mode overlays: separate multiple shaders for loose edges and verts
2017-03-01 14:07:04 +01:00
{
EditMode: remove edge-hash from MeshRenderData LoopTri's can access this data.
2017-04-12 13:42:27 +10:00
const float *pos;
Cleanup: replace char array with struct
2017-04-12 14:08:44 +10:00
EdgeDrawAttr *eattr;
EditMode: remove edge-hash from MeshRenderData LoopTri's can access this data.
2017-04-12 13:42:27 +10:00
unsigned char fflag;
unsigned char vflag;
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
pos = mesh_render_data_vert_co(rdata, tri_vert_idx[0]);
eattr = mesh_render_data_edge_flag(rdata, tri_edge_idx[1]);
fflag = mesh_render_data_looptri_flag(rdata, f);
vflag = mesh_render_data_vertex_flag(rdata, tri_vert_idx[0]);
Cleanup: replace char array with struct
2017-04-12 14:08:44 +10:00
eattr->v_flag = fflag | vflag;
Gawain: add VertexBuffer prefix to functions See intern/gawain for the API change. Other files are updated to use the new names.
2017-04-04 20:45:23 -04:00
VertexBuffer_set_attrib(vbo, pos_id, base_vert_idx + 0, pos);
Cleanup: replace char array with struct
2017-04-12 14:08:44 +10:00
VertexBuffer_set_attrib(vbo, edgeMod_id, base_vert_idx + 0, eattr);
Edit Mode overlays: separate multiple shaders for loose edges and verts
2017-03-01 14:07:04 +01:00
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
pos = mesh_render_data_vert_co(rdata, tri_vert_idx[1]);
eattr = mesh_render_data_edge_flag(rdata, tri_edge_idx[2]);
vflag = mesh_render_data_vertex_flag(rdata, tri_vert_idx[1]);
Cleanup: replace char array with struct
2017-04-12 14:08:44 +10:00
eattr->v_flag = fflag | vflag;
Gawain: add VertexBuffer prefix to functions See intern/gawain for the API change. Other files are updated to use the new names.
2017-04-04 20:45:23 -04:00
VertexBuffer_set_attrib(vbo, pos_id, base_vert_idx + 1, pos);
Cleanup: replace char array with struct
2017-04-12 14:08:44 +10:00
VertexBuffer_set_attrib(vbo, edgeMod_id, base_vert_idx + 1, eattr);
Edit Mode overlays: separate multiple shaders for loose edges and verts
2017-03-01 14:07:04 +01:00
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
pos = mesh_render_data_vert_co(rdata, tri_vert_idx[2]);
eattr = mesh_render_data_edge_flag(rdata, tri_edge_idx[0]);
vflag = mesh_render_data_vertex_flag(rdata, tri_vert_idx[2]);
Cleanup: replace char array with struct
2017-04-12 14:08:44 +10:00
eattr->v_flag = fflag | vflag;
Gawain: add VertexBuffer prefix to functions See intern/gawain for the API change. Other files are updated to use the new names.
2017-04-04 20:45:23 -04:00
VertexBuffer_set_attrib(vbo, pos_id, base_vert_idx + 2, pos);
Cleanup: replace char array with struct
2017-04-12 14:08:44 +10:00
VertexBuffer_set_attrib(vbo, edgeMod_id, base_vert_idx + 2, eattr);
Edit Mode overlays: separate multiple shaders for loose edges and verts
2017-03-01 14:07:04 +01:00
}
static void add_overlay_loose_edge(
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
MeshRenderData *rdata, VertexBuffer *vbo, const unsigned int pos_id, const unsigned int edgeMod_id,
Edit Mode overlays: separate multiple shaders for loose edges and verts
2017-03-01 14:07:04 +01:00
const int v1, const int v2, const int e, const int base_vert_idx)
{
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
EdgeDrawAttr *eattr = mesh_render_data_edge_flag(rdata, e);
const float *pos = mesh_render_data_vert_co(rdata, v1);
eattr->v_flag = mesh_render_data_vertex_flag(rdata, v1);
Gawain: add VertexBuffer prefix to functions See intern/gawain for the API change. Other files are updated to use the new names.
2017-04-04 20:45:23 -04:00
VertexBuffer_set_attrib(vbo, pos_id, base_vert_idx + 0, pos);
Cleanup: replace char array with struct
2017-04-12 14:08:44 +10:00
VertexBuffer_set_attrib(vbo, edgeMod_id, base_vert_idx + 0, eattr);
Edit Mode overlays: separate multiple shaders for loose edges and verts
2017-03-01 14:07:04 +01:00
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
pos = mesh_render_data_vert_co(rdata, v2);
eattr->v_flag = mesh_render_data_vertex_flag(rdata, v2);
Gawain: add VertexBuffer prefix to functions See intern/gawain for the API change. Other files are updated to use the new names.
2017-04-04 20:45:23 -04:00
VertexBuffer_set_attrib(vbo, pos_id, base_vert_idx + 1, pos);
Cleanup: replace char array with struct
2017-04-12 14:08:44 +10:00
VertexBuffer_set_attrib(vbo, edgeMod_id, base_vert_idx + 1, eattr);
Edit Mode overlays: separate multiple shaders for loose edges and verts
2017-03-01 14:07:04 +01:00
}
static void add_overlay_loose_vert(
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
MeshRenderData *rdata, VertexBuffer *vbo, const unsigned int pos_id, const unsigned int edgeMod_id,
Edit Mode overlays: separate multiple shaders for loose edges and verts
2017-03-01 14:07:04 +01:00
const int v, const int base_vert_idx)
{
unsigned char vflag[4] = {0, 0, 0, 0};
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
const float *pos = mesh_render_data_vert_co(rdata, v);
vflag[0] = mesh_render_data_vertex_flag(rdata, v);
Gawain: add VertexBuffer prefix to functions See intern/gawain for the API change. Other files are updated to use the new names.
2017-04-04 20:45:23 -04:00
VertexBuffer_set_attrib(vbo, pos_id, base_vert_idx + 0, pos);
VertexBuffer_set_attrib(vbo, edgeMod_id, base_vert_idx + 0, vflag);
Edit Mode overlays: separate multiple shaders for loose edges and verts
2017-03-01 14:07:04 +01:00
}
Rework of BKE's mesh_render to support BMesh directly. Note that since there is no (efficient) ways to get arrays of MVert/MEdge/etc. out of a BMesh, I refactored quite heavily internals of BKE_mesh_render. Now, when you do need acess to mesh data to generate cached batches, you create an abstract struct from mesh (either Mesh or BMesh if available), and then use advanced helpers to extract needed data, on a per-item basis (no more handling of arrays of verts/edges/... in batches code). This allows to: * Avoid having to create arrays of BMesh elements. * Take advantage of existing advanced BMesh topology and connectivity data. Reviewers: dfelinto, fclem, merwin Differential Revision: https://developer.blender.org/D2521
2017-02-23 11:54:40 +01:00
Convert MBC_ API to Mesh (instead of derived mesh) and move it to mesh_render This includes a few fixes in the MBC_ api. The idea here is for this to be the only interface the render engines will deal with for the meshes. If we need to expose special options for sculpting engine we refactor this accordingly. But for now we are shaping this in a per-case base. Note: * We still need to hook up to the depsgraph to force clear/update of batch_cache when mesh changes (I'm waiting for Sergey Sharybin's depsgraph update for this though) * Also ideally we could/should use BMesh directly instead of DerivedMesh, but this will do for now. Note 2: In the end I renamed the `BKE_mesh_render` functions to `static mesh_render`. We can re-expose them as BKE_* later once we need it. Reviewers: merwin Subscribers: fclem Differential Revision: https://developer.blender.org/D2476
2017-01-25 09:16:29 +01:00
/* ---------------------------------------------------------------------- */
/* Mesh Batch Cache */
typedef struct MeshBatchCache {
VertexBuffer *pos_in_order;
Edit Mesh overlay: Ported Display Normals option
2017-03-09 01:29:58 +01:00
VertexBuffer *nor_in_order;
Convert MBC_ API to Mesh (instead of derived mesh) and move it to mesh_render This includes a few fixes in the MBC_ api. The idea here is for this to be the only interface the render engines will deal with for the meshes. If we need to expose special options for sculpting engine we refactor this accordingly. But for now we are shaping this in a per-case base. Note: * We still need to hook up to the depsgraph to force clear/update of batch_cache when mesh changes (I'm waiting for Sergey Sharybin's depsgraph update for this though) * Also ideally we could/should use BMesh directly instead of DerivedMesh, but this will do for now. Note 2: In the end I renamed the `BKE_mesh_render` functions to `static mesh_render`. We can re-expose them as BKE_* later once we need it. Reviewers: merwin Subscribers: fclem Differential Revision: https://developer.blender.org/D2476
2017-01-25 09:16:29 +01:00
ElementList *edges_in_order;
ElementList *triangles_in_order;
Batch *all_verts;
Batch *all_edges;
Batch *all_triangles;
Edit Mesh overlay: Ported Display Normals option
2017-03-09 01:29:58 +01:00
VertexBuffer *pos_with_normals;
Batch *triangles_with_normals;
Batch *points_with_normals;
Convert MBC_ API to Mesh (instead of derived mesh) and move it to mesh_render This includes a few fixes in the MBC_ api. The idea here is for this to be the only interface the render engines will deal with for the meshes. If we need to expose special options for sculpting engine we refactor this accordingly. But for now we are shaping this in a per-case base. Note: * We still need to hook up to the depsgraph to force clear/update of batch_cache when mesh changes (I'm waiting for Sergey Sharybin's depsgraph update for this though) * Also ideally we could/should use BMesh directly instead of DerivedMesh, but this will do for now. Note 2: In the end I renamed the `BKE_mesh_render` functions to `static mesh_render`. We can re-expose them as BKE_* later once we need it. Reviewers: merwin Subscribers: fclem Differential Revision: https://developer.blender.org/D2476
2017-01-25 09:16:29 +01:00
Batch *fancy_edges; /* owns its vertex buffer (not shared) */
Edit Mode overlays: separate multiple shaders for loose edges and verts
2017-03-01 14:07:04 +01:00
Eevee: Make default shaders works. - Added New Batch cache routine to get the split mesh for each material. (optimization needed) - Did some shader modifications to create default_frag and support a somwhat correct fresnel for lamps (to finish)
2017-04-25 18:46:59 +02:00
/* Maybe have shaded_triangles_data split into pos_nor and uv_tangent
* to minimise data transfer for skinned mesh. */
Mesh Batch Cache: Add support for uv/vcol/tangent data layers. This is a bad implementation waiting for some improvement that are : - Gawain support for multiple names. -> will remove duplication of attribute alloc - Glsl safe name string escaping. -> will remove any risk of hash colision. Old glsl drawing was relying on attribute being choosen by DerivedMesh at drawing time. For this reason, we declare all possible attrib "name" for each data layer inside the batches. In the glsl code, we declare required data by type and name. Then Gawain only bind vbos if names correspond. This is way cleaner as we do not need to access the CD itself when drawing. One other problem is that the hash maybe 11 caracters long and rapidly overload gawain's attrib name buffer.
2017-05-01 17:58:15 +02:00
VertexFormat shaded_triangles_format;
Eevee: Make default shaders works. - Added New Batch cache routine to get the split mesh for each material. (optimization needed) - Did some shader modifications to create default_frag and support a somwhat correct fresnel for lamps (to finish)
2017-04-25 18:46:59 +02:00
VertexBuffer *shaded_triangles_data;
ElementList **shaded_triangles_in_order;
Batch **shaded_triangles;
Edit Mode overlays: separate multiple shaders for loose edges and verts
2017-03-01 14:07:04 +01:00
/* TODO : split in 2 buffers to avoid unnecessary
* data transfer when selecting/deselecting
* and combine into one batch and use offsets to render
* Tri / edges / verts separately */
Batch *overlay_triangles;
Batch *overlay_loose_verts;
Batch *overlay_loose_edges;
Edit Mode overlay: Moved Shaders to draw modules and resolved some draw issue. We don't want to clutter gpu_shader.c with engine specific code Added face's center dot Simplified loose vert shader
2017-03-02 01:07:03 +01:00
Batch *overlay_facedots;
Convert MBC_ API to Mesh (instead of derived mesh) and move it to mesh_render This includes a few fixes in the MBC_ api. The idea here is for this to be the only interface the render engines will deal with for the meshes. If we need to expose special options for sculpting engine we refactor this accordingly. But for now we are shaping this in a per-case base. Note: * We still need to hook up to the depsgraph to force clear/update of batch_cache when mesh changes (I'm waiting for Sergey Sharybin's depsgraph update for this though) * Also ideally we could/should use BMesh directly instead of DerivedMesh, but this will do for now. Note 2: In the end I renamed the `BKE_mesh_render` functions to `static mesh_render`. We can re-expose them as BKE_* later once we need it. Reviewers: merwin Subscribers: fclem Differential Revision: https://developer.blender.org/D2476
2017-01-25 09:16:29 +01:00
Use the ultimate depsgraph callback function Note: when in edit mode this depsgraph update is not being called. We are using DerivedMesh in those cases, so it is fine. I would like to investigate this though
2017-01-26 18:56:52 +01:00
/* settings to determine if cache is invalid */
bool is_dirty;
Cleanup: replace num/ct/tot with 'len'
2017-04-19 00:10:32 +10:00
int edge_len;
int tri_len;
int poly_len;
int vert_len;
Eevee: Make default shaders works. - Added New Batch cache routine to get the split mesh for each material. (optimization needed) - Did some shader modifications to create default_frag and support a somwhat correct fresnel for lamps (to finish)
2017-04-25 18:46:59 +02:00
int mat_ct;
Convert MBC_ API to Mesh (instead of derived mesh) and move it to mesh_render This includes a few fixes in the MBC_ api. The idea here is for this to be the only interface the render engines will deal with for the meshes. If we need to expose special options for sculpting engine we refactor this accordingly. But for now we are shaping this in a per-case base. Note: * We still need to hook up to the depsgraph to force clear/update of batch_cache when mesh changes (I'm waiting for Sergey Sharybin's depsgraph update for this though) * Also ideally we could/should use BMesh directly instead of DerivedMesh, but this will do for now. Note 2: In the end I renamed the `BKE_mesh_render` functions to `static mesh_render`. We can re-expose them as BKE_* later once we need it. Reviewers: merwin Subscribers: fclem Differential Revision: https://developer.blender.org/D2476
2017-01-25 09:16:29 +01:00
bool is_editmode;
} MeshBatchCache;
Rework of BKE's mesh_render to support BMesh directly. Note that since there is no (efficient) ways to get arrays of MVert/MEdge/etc. out of a BMesh, I refactored quite heavily internals of BKE_mesh_render. Now, when you do need acess to mesh data to generate cached batches, you create an abstract struct from mesh (either Mesh or BMesh if available), and then use advanced helpers to extract needed data, on a per-item basis (no more handling of arrays of verts/edges/... in batches code). This allows to: * Avoid having to create arrays of BMesh elements. * Take advantage of existing advanced BMesh topology and connectivity data. Reviewers: dfelinto, fclem, merwin Differential Revision: https://developer.blender.org/D2521
2017-02-23 11:54:40 +01:00
/* Batch cache management. */
Convert MBC_ API to Mesh (instead of derived mesh) and move it to mesh_render This includes a few fixes in the MBC_ api. The idea here is for this to be the only interface the render engines will deal with for the meshes. If we need to expose special options for sculpting engine we refactor this accordingly. But for now we are shaping this in a per-case base. Note: * We still need to hook up to the depsgraph to force clear/update of batch_cache when mesh changes (I'm waiting for Sergey Sharybin's depsgraph update for this though) * Also ideally we could/should use BMesh directly instead of DerivedMesh, but this will do for now. Note 2: In the end I renamed the `BKE_mesh_render` functions to `static mesh_render`. We can re-expose them as BKE_* later once we need it. Reviewers: merwin Subscribers: fclem Differential Revision: https://developer.blender.org/D2476
2017-01-25 09:16:29 +01:00
static bool mesh_batch_cache_valid(Mesh *me)
{
MeshBatchCache *cache = me->batch_cache;
if (cache == NULL) {
return false;
}
Eevee: Make default shaders works. - Added New Batch cache routine to get the split mesh for each material. (optimization needed) - Did some shader modifications to create default_frag and support a somwhat correct fresnel for lamps (to finish)
2017-04-25 18:46:59 +02:00
/* XXX find another place for this */
if (cache->mat_ct != mesh_render_mat_ct_get(me)) {
cache->is_dirty = true;
}
Convert MBC_ API to Mesh (instead of derived mesh) and move it to mesh_render This includes a few fixes in the MBC_ api. The idea here is for this to be the only interface the render engines will deal with for the meshes. If we need to expose special options for sculpting engine we refactor this accordingly. But for now we are shaping this in a per-case base. Note: * We still need to hook up to the depsgraph to force clear/update of batch_cache when mesh changes (I'm waiting for Sergey Sharybin's depsgraph update for this though) * Also ideally we could/should use BMesh directly instead of DerivedMesh, but this will do for now. Note 2: In the end I renamed the `BKE_mesh_render` functions to `static mesh_render`. We can re-expose them as BKE_* later once we need it. Reviewers: merwin Subscribers: fclem Differential Revision: https://developer.blender.org/D2476
2017-01-25 09:16:29 +01:00
if (cache->is_editmode != (me->edit_btmesh != NULL)) {
return false;
}
Use the ultimate depsgraph callback function Note: when in edit mode this depsgraph update is not being called. We are using DerivedMesh in those cases, so it is fine. I would like to investigate this though
2017-01-26 18:56:52 +01:00
if (cache->is_dirty == false) {
return true;
}
else {
Mesh batch caches update from depsgraph We no longer need to rely on Derived mesh flag since the mesh is properly flushed from the depsgraph
2017-01-26 19:08:27 +01:00
if (cache->is_editmode) {
return false;
}
Cleanup: replace num/ct/tot with 'len'
2017-04-19 00:10:32 +10:00
else if ((cache->vert_len != mesh_render_verts_len_get(me)) ||
(cache->edge_len != mesh_render_edges_len_get(me)) ||
(cache->tri_len != mesh_render_looptri_len_get(me)) ||
Eevee: Make default shaders works. - Added New Batch cache routine to get the split mesh for each material. (optimization needed) - Did some shader modifications to create default_frag and support a somwhat correct fresnel for lamps (to finish)
2017-04-25 18:46:59 +02:00
(cache->poly_len != mesh_render_polys_len_get(me)) ||
(cache->mat_ct != mesh_render_mat_ct_get(me)))
Use the ultimate depsgraph callback function Note: when in edit mode this depsgraph update is not being called. We are using DerivedMesh in those cases, so it is fine. I would like to investigate this though
2017-01-26 18:56:52 +01:00
{
return false;
}
Convert MBC_ API to Mesh (instead of derived mesh) and move it to mesh_render This includes a few fixes in the MBC_ api. The idea here is for this to be the only interface the render engines will deal with for the meshes. If we need to expose special options for sculpting engine we refactor this accordingly. But for now we are shaping this in a per-case base. Note: * We still need to hook up to the depsgraph to force clear/update of batch_cache when mesh changes (I'm waiting for Sergey Sharybin's depsgraph update for this though) * Also ideally we could/should use BMesh directly instead of DerivedMesh, but this will do for now. Note 2: In the end I renamed the `BKE_mesh_render` functions to `static mesh_render`. We can re-expose them as BKE_* later once we need it. Reviewers: merwin Subscribers: fclem Differential Revision: https://developer.blender.org/D2476
2017-01-25 09:16:29 +01:00
}
return true;
}
static void mesh_batch_cache_init(Mesh *me)
{
MeshBatchCache *cache = me->batch_cache;
Rework of BKE's mesh_render to support BMesh directly. Note that since there is no (efficient) ways to get arrays of MVert/MEdge/etc. out of a BMesh, I refactored quite heavily internals of BKE_mesh_render. Now, when you do need acess to mesh data to generate cached batches, you create an abstract struct from mesh (either Mesh or BMesh if available), and then use advanced helpers to extract needed data, on a per-item basis (no more handling of arrays of verts/edges/... in batches code). This allows to: * Avoid having to create arrays of BMesh elements. * Take advantage of existing advanced BMesh topology and connectivity data. Reviewers: dfelinto, fclem, merwin Differential Revision: https://developer.blender.org/D2521
2017-02-23 11:54:40 +01:00
if (!cache) {
cache = me->batch_cache = MEM_callocN(sizeof(*cache), __func__);
}
else {
memset(cache, 0, sizeof(*cache));
}
Convert MBC_ API to Mesh (instead of derived mesh) and move it to mesh_render This includes a few fixes in the MBC_ api. The idea here is for this to be the only interface the render engines will deal with for the meshes. If we need to expose special options for sculpting engine we refactor this accordingly. But for now we are shaping this in a per-case base. Note: * We still need to hook up to the depsgraph to force clear/update of batch_cache when mesh changes (I'm waiting for Sergey Sharybin's depsgraph update for this though) * Also ideally we could/should use BMesh directly instead of DerivedMesh, but this will do for now. Note 2: In the end I renamed the `BKE_mesh_render` functions to `static mesh_render`. We can re-expose them as BKE_* later once we need it. Reviewers: merwin Subscribers: fclem Differential Revision: https://developer.blender.org/D2476
2017-01-25 09:16:29 +01:00
cache->is_editmode = me->edit_btmesh != NULL;
Mesh batch caches update from depsgraph We no longer need to rely on Derived mesh flag since the mesh is properly flushed from the depsgraph
2017-01-26 19:08:27 +01:00
if (cache->is_editmode == false) {
Cleanup: replace num/ct/tot with 'len'
2017-04-19 00:10:32 +10:00
cache->edge_len = mesh_render_edges_len_get(me);
cache->tri_len = mesh_render_looptri_len_get(me);
cache->poly_len = mesh_render_polys_len_get(me);
cache->vert_len = mesh_render_verts_len_get(me);
Convert MBC_ API to Mesh (instead of derived mesh) and move it to mesh_render This includes a few fixes in the MBC_ api. The idea here is for this to be the only interface the render engines will deal with for the meshes. If we need to expose special options for sculpting engine we refactor this accordingly. But for now we are shaping this in a per-case base. Note: * We still need to hook up to the depsgraph to force clear/update of batch_cache when mesh changes (I'm waiting for Sergey Sharybin's depsgraph update for this though) * Also ideally we could/should use BMesh directly instead of DerivedMesh, but this will do for now. Note 2: In the end I renamed the `BKE_mesh_render` functions to `static mesh_render`. We can re-expose them as BKE_* later once we need it. Reviewers: merwin Subscribers: fclem Differential Revision: https://developer.blender.org/D2476
2017-01-25 09:16:29 +01:00
}
Use the ultimate depsgraph callback function Note: when in edit mode this depsgraph update is not being called. We are using DerivedMesh in those cases, so it is fine. I would like to investigate this though
2017-01-26 18:56:52 +01:00
Eevee: Make default shaders works. - Added New Batch cache routine to get the split mesh for each material. (optimization needed) - Did some shader modifications to create default_frag and support a somwhat correct fresnel for lamps (to finish)
2017-04-25 18:46:59 +02:00
cache->mat_ct = mesh_render_mat_ct_get(me);
Use the ultimate depsgraph callback function Note: when in edit mode this depsgraph update is not being called. We are using DerivedMesh in those cases, so it is fine. I would like to investigate this though
2017-01-26 18:56:52 +01:00
cache->is_dirty = false;
Convert MBC_ API to Mesh (instead of derived mesh) and move it to mesh_render This includes a few fixes in the MBC_ api. The idea here is for this to be the only interface the render engines will deal with for the meshes. If we need to expose special options for sculpting engine we refactor this accordingly. But for now we are shaping this in a per-case base. Note: * We still need to hook up to the depsgraph to force clear/update of batch_cache when mesh changes (I'm waiting for Sergey Sharybin's depsgraph update for this though) * Also ideally we could/should use BMesh directly instead of DerivedMesh, but this will do for now. Note 2: In the end I renamed the `BKE_mesh_render` functions to `static mesh_render`. We can re-expose them as BKE_* later once we need it. Reviewers: merwin Subscribers: fclem Differential Revision: https://developer.blender.org/D2476
2017-01-25 09:16:29 +01:00
}
static MeshBatchCache *mesh_batch_cache_get(Mesh *me)
{
if (!mesh_batch_cache_valid(me)) {
Cleanup: move draw-cache creation from BKE to DRW Creating draw-cache should only ever be used by the draw-manager.
2017-04-21 21:14:11 +10:00
mesh_batch_cache_clear(me);
Convert MBC_ API to Mesh (instead of derived mesh) and move it to mesh_render This includes a few fixes in the MBC_ api. The idea here is for this to be the only interface the render engines will deal with for the meshes. If we need to expose special options for sculpting engine we refactor this accordingly. But for now we are shaping this in a per-case base. Note: * We still need to hook up to the depsgraph to force clear/update of batch_cache when mesh changes (I'm waiting for Sergey Sharybin's depsgraph update for this though) * Also ideally we could/should use BMesh directly instead of DerivedMesh, but this will do for now. Note 2: In the end I renamed the `BKE_mesh_render` functions to `static mesh_render`. We can re-expose them as BKE_* later once we need it. Reviewers: merwin Subscribers: fclem Differential Revision: https://developer.blender.org/D2476
2017-01-25 09:16:29 +01:00
mesh_batch_cache_init(me);
}
return me->batch_cache;
}
Cleanup: move draw-cache creation from BKE to DRW Creating draw-cache should only ever be used by the draw-manager.
2017-04-21 21:14:11 +10:00
void DRW_mesh_batch_cache_dirty(Mesh *me, int mode)
Convert MBC_ API to Mesh (instead of derived mesh) and move it to mesh_render This includes a few fixes in the MBC_ api. The idea here is for this to be the only interface the render engines will deal with for the meshes. If we need to expose special options for sculpting engine we refactor this accordingly. But for now we are shaping this in a per-case base. Note: * We still need to hook up to the depsgraph to force clear/update of batch_cache when mesh changes (I'm waiting for Sergey Sharybin's depsgraph update for this though) * Also ideally we could/should use BMesh directly instead of DerivedMesh, but this will do for now. Note 2: In the end I renamed the `BKE_mesh_render` functions to `static mesh_render`. We can re-expose them as BKE_* later once we need it. Reviewers: merwin Subscribers: fclem Differential Revision: https://developer.blender.org/D2476
2017-01-25 09:16:29 +01:00
{
Rework of BKE's mesh_render to support BMesh directly. Note that since there is no (efficient) ways to get arrays of MVert/MEdge/etc. out of a BMesh, I refactored quite heavily internals of BKE_mesh_render. Now, when you do need acess to mesh data to generate cached batches, you create an abstract struct from mesh (either Mesh or BMesh if available), and then use advanced helpers to extract needed data, on a per-item basis (no more handling of arrays of verts/edges/... in batches code). This allows to: * Avoid having to create arrays of BMesh elements. * Take advantage of existing advanced BMesh topology and connectivity data. Reviewers: dfelinto, fclem, merwin Differential Revision: https://developer.blender.org/D2521
2017-02-23 11:54:40 +01:00
MeshBatchCache *cache = me->batch_cache;
Cleanup: move draw-cache creation from BKE to DRW Creating draw-cache should only ever be used by the draw-manager.
2017-04-21 21:14:11 +10:00
if (cache == NULL) {
return;
Rework of BKE's mesh_render to support BMesh directly. Note that since there is no (efficient) ways to get arrays of MVert/MEdge/etc. out of a BMesh, I refactored quite heavily internals of BKE_mesh_render. Now, when you do need acess to mesh data to generate cached batches, you create an abstract struct from mesh (either Mesh or BMesh if available), and then use advanced helpers to extract needed data, on a per-item basis (no more handling of arrays of verts/edges/... in batches code). This allows to: * Avoid having to create arrays of BMesh elements. * Take advantage of existing advanced BMesh topology and connectivity data. Reviewers: dfelinto, fclem, merwin Differential Revision: https://developer.blender.org/D2521
2017-02-23 11:54:40 +01:00
}
Cleanup: move draw-cache creation from BKE to DRW Creating draw-cache should only ever be used by the draw-manager.
2017-04-21 21:14:11 +10:00
switch (mode) {
case BKE_MESH_BATCH_DIRTY_ALL:
cache->is_dirty = true;
break;
case BKE_MESH_BATCH_DIRTY_SELECT:
/* TODO Separate Flag vbo */
BATCH_DISCARD_ALL_SAFE(cache->overlay_triangles);
BATCH_DISCARD_ALL_SAFE(cache->overlay_loose_verts);
BATCH_DISCARD_ALL_SAFE(cache->overlay_loose_edges);
BATCH_DISCARD_ALL_SAFE(cache->overlay_facedots);
break;
default:
BLI_assert(0);
Edit Mode overlay: Update cache on selection
2017-03-06 03:58:08 +01:00
}
}
Cleanup: move draw-cache creation from BKE to DRW Creating draw-cache should only ever be used by the draw-manager.
2017-04-21 21:14:11 +10:00
static void mesh_batch_cache_clear(Mesh *me)
Rework of BKE's mesh_render to support BMesh directly. Note that since there is no (efficient) ways to get arrays of MVert/MEdge/etc. out of a BMesh, I refactored quite heavily internals of BKE_mesh_render. Now, when you do need acess to mesh data to generate cached batches, you create an abstract struct from mesh (either Mesh or BMesh if available), and then use advanced helpers to extract needed data, on a per-item basis (no more handling of arrays of verts/edges/... in batches code). This allows to: * Avoid having to create arrays of BMesh elements. * Take advantage of existing advanced BMesh topology and connectivity data. Reviewers: dfelinto, fclem, merwin Differential Revision: https://developer.blender.org/D2521
2017-02-23 11:54:40 +01:00
{
MeshBatchCache *cache = me->batch_cache;
if (!cache) {
return;
}
Use 'safe' macros for common free operation Same as MEM_SAFE_FREE macro, checks for NULL, runs free then sets NULL. Blocks of code that do this many times are noisy and likely errors here wouldn't be noticed immediately. Also NULL's static vars which were being left set.
2017-04-13 13:30:53 +10:00
BATCH_DISCARD_SAFE(cache->all_verts);
BATCH_DISCARD_SAFE(cache->all_edges);
BATCH_DISCARD_SAFE(cache->all_triangles);
Rework of BKE's mesh_render to support BMesh directly. Note that since there is no (efficient) ways to get arrays of MVert/MEdge/etc. out of a BMesh, I refactored quite heavily internals of BKE_mesh_render. Now, when you do need acess to mesh data to generate cached batches, you create an abstract struct from mesh (either Mesh or BMesh if available), and then use advanced helpers to extract needed data, on a per-item basis (no more handling of arrays of verts/edges/... in batches code). This allows to: * Avoid having to create arrays of BMesh elements. * Take advantage of existing advanced BMesh topology and connectivity data. Reviewers: dfelinto, fclem, merwin Differential Revision: https://developer.blender.org/D2521
2017-02-23 11:54:40 +01:00
Use 'safe' macros for common free operation Same as MEM_SAFE_FREE macro, checks for NULL, runs free then sets NULL. Blocks of code that do this many times are noisy and likely errors here wouldn't be noticed immediately. Also NULL's static vars which were being left set.
2017-04-13 13:30:53 +10:00
VERTEXBUFFER_DISCARD_SAFE(cache->pos_in_order);
ELEMENTLIST_DISCARD_SAFE(cache->edges_in_order);
ELEMENTLIST_DISCARD_SAFE(cache->triangles_in_order);
Rework of BKE's mesh_render to support BMesh directly. Note that since there is no (efficient) ways to get arrays of MVert/MEdge/etc. out of a BMesh, I refactored quite heavily internals of BKE_mesh_render. Now, when you do need acess to mesh data to generate cached batches, you create an abstract struct from mesh (either Mesh or BMesh if available), and then use advanced helpers to extract needed data, on a per-item basis (no more handling of arrays of verts/edges/... in batches code). This allows to: * Avoid having to create arrays of BMesh elements. * Take advantage of existing advanced BMesh topology and connectivity data. Reviewers: dfelinto, fclem, merwin Differential Revision: https://developer.blender.org/D2521
2017-02-23 11:54:40 +01:00
Use 'safe' macros for common free operation Same as MEM_SAFE_FREE macro, checks for NULL, runs free then sets NULL. Blocks of code that do this many times are noisy and likely errors here wouldn't be noticed immediately. Also NULL's static vars which were being left set.
2017-04-13 13:30:53 +10:00
BATCH_DISCARD_ALL_SAFE(cache->overlay_triangles);
BATCH_DISCARD_ALL_SAFE(cache->overlay_loose_verts);
BATCH_DISCARD_ALL_SAFE(cache->overlay_loose_edges);
BATCH_DISCARD_ALL_SAFE(cache->overlay_facedots);
Edit Mode overlays: separate multiple shaders for loose edges and verts
2017-03-01 14:07:04 +01:00
Use 'safe' macros for common free operation Same as MEM_SAFE_FREE macro, checks for NULL, runs free then sets NULL. Blocks of code that do this many times are noisy and likely errors here wouldn't be noticed immediately. Also NULL's static vars which were being left set.
2017-04-13 13:30:53 +10:00
BATCH_DISCARD_SAFE(cache->triangles_with_normals);
BATCH_DISCARD_SAFE(cache->points_with_normals);
VERTEXBUFFER_DISCARD_SAFE(cache->pos_with_normals);
Rework of BKE's mesh_render to support BMesh directly. Note that since there is no (efficient) ways to get arrays of MVert/MEdge/etc. out of a BMesh, I refactored quite heavily internals of BKE_mesh_render. Now, when you do need acess to mesh data to generate cached batches, you create an abstract struct from mesh (either Mesh or BMesh if available), and then use advanced helpers to extract needed data, on a per-item basis (no more handling of arrays of verts/edges/... in batches code). This allows to: * Avoid having to create arrays of BMesh elements. * Take advantage of existing advanced BMesh topology and connectivity data. Reviewers: dfelinto, fclem, merwin Differential Revision: https://developer.blender.org/D2521
2017-02-23 11:54:40 +01:00
Use 'safe' macros for common free operation Same as MEM_SAFE_FREE macro, checks for NULL, runs free then sets NULL. Blocks of code that do this many times are noisy and likely errors here wouldn't be noticed immediately. Also NULL's static vars which were being left set.
2017-04-13 13:30:53 +10:00
BATCH_DISCARD_ALL_SAFE(cache->fancy_edges);
Eevee: Make default shaders works. - Added New Batch cache routine to get the split mesh for each material. (optimization needed) - Did some shader modifications to create default_frag and support a somwhat correct fresnel for lamps (to finish)
2017-04-25 18:46:59 +02:00
VERTEXBUFFER_DISCARD_SAFE(cache->shaded_triangles_data);
if (cache->shaded_triangles_in_order) {
for (int i = 0; i < cache->mat_ct; ++i) {
ELEMENTLIST_DISCARD_SAFE(cache->shaded_triangles_in_order[i]);
}
}
if (cache->shaded_triangles) {
for (int i = 0; i < cache->mat_ct; ++i) {
BATCH_DISCARD_SAFE(cache->shaded_triangles[i]);
}
}
MEM_SAFE_FREE(cache->shaded_triangles_in_order);
MEM_SAFE_FREE(cache->shaded_triangles);
Rework of BKE's mesh_render to support BMesh directly. Note that since there is no (efficient) ways to get arrays of MVert/MEdge/etc. out of a BMesh, I refactored quite heavily internals of BKE_mesh_render. Now, when you do need acess to mesh data to generate cached batches, you create an abstract struct from mesh (either Mesh or BMesh if available), and then use advanced helpers to extract needed data, on a per-item basis (no more handling of arrays of verts/edges/... in batches code). This allows to: * Avoid having to create arrays of BMesh elements. * Take advantage of existing advanced BMesh topology and connectivity data. Reviewers: dfelinto, fclem, merwin Differential Revision: https://developer.blender.org/D2521
2017-02-23 11:54:40 +01:00
}
Cleanup: move draw-cache creation from BKE to DRW Creating draw-cache should only ever be used by the draw-manager.
2017-04-21 21:14:11 +10:00
void DRW_mesh_batch_cache_free(Mesh *me)
Rework of BKE's mesh_render to support BMesh directly. Note that since there is no (efficient) ways to get arrays of MVert/MEdge/etc. out of a BMesh, I refactored quite heavily internals of BKE_mesh_render. Now, when you do need acess to mesh data to generate cached batches, you create an abstract struct from mesh (either Mesh or BMesh if available), and then use advanced helpers to extract needed data, on a per-item basis (no more handling of arrays of verts/edges/... in batches code). This allows to: * Avoid having to create arrays of BMesh elements. * Take advantage of existing advanced BMesh topology and connectivity data. Reviewers: dfelinto, fclem, merwin Differential Revision: https://developer.blender.org/D2521
2017-02-23 11:54:40 +01:00
{
Cleanup: move draw-cache creation from BKE to DRW Creating draw-cache should only ever be used by the draw-manager.
2017-04-21 21:14:11 +10:00
mesh_batch_cache_clear(me);
Rework of BKE's mesh_render to support BMesh directly. Note that since there is no (efficient) ways to get arrays of MVert/MEdge/etc. out of a BMesh, I refactored quite heavily internals of BKE_mesh_render. Now, when you do need acess to mesh data to generate cached batches, you create an abstract struct from mesh (either Mesh or BMesh if available), and then use advanced helpers to extract needed data, on a per-item basis (no more handling of arrays of verts/edges/... in batches code). This allows to: * Avoid having to create arrays of BMesh elements. * Take advantage of existing advanced BMesh topology and connectivity data. Reviewers: dfelinto, fclem, merwin Differential Revision: https://developer.blender.org/D2521
2017-02-23 11:54:40 +01:00
MEM_SAFE_FREE(me->batch_cache);
}
/* Batch cache usage. */
Eevee: Make default shaders works. - Added New Batch cache routine to get the split mesh for each material. (optimization needed) - Did some shader modifications to create default_frag and support a somwhat correct fresnel for lamps (to finish)
2017-04-25 18:46:59 +02:00
static VertexBuffer *mesh_batch_cache_get_shading_data(MeshRenderData *rdata, MeshBatchCache *cache)
{
BLI_assert(rdata->types & (MR_DATATYPE_VERT | MR_DATATYPE_LOOPTRI | MR_DATATYPE_LOOP | MR_DATATYPE_POLY));
if (cache->shaded_triangles_data == NULL) {
unsigned int vidx = 0, nidx = 0;
Mesh Batch Cache: Add support for uv/vcol/tangent data layers. This is a bad implementation waiting for some improvement that are : - Gawain support for multiple names. -> will remove duplication of attribute alloc - Glsl safe name string escaping. -> will remove any risk of hash colision. Old glsl drawing was relying on attribute being choosen by DerivedMesh at drawing time. For this reason, we declare all possible attrib "name" for each data layer inside the batches. In the glsl code, we declare required data by type and name. Then Gawain only bind vbos if names correspond. This is way cleaner as we do not need to access the CD itself when drawing. One other problem is that the hash maybe 11 caracters long and rapidly overload gawain's attrib name buffer.
2017-05-01 17:58:15 +02:00
const char *attrib_name;
Eevee: Make default shaders works. - Added New Batch cache routine to get the split mesh for each material. (optimization needed) - Did some shader modifications to create default_frag and support a somwhat correct fresnel for lamps (to finish)
2017-04-25 18:46:59 +02:00
Mesh Batch Cache: Add support for uv/vcol/tangent data layers. This is a bad implementation waiting for some improvement that are : - Gawain support for multiple names. -> will remove duplication of attribute alloc - Glsl safe name string escaping. -> will remove any risk of hash colision. Old glsl drawing was relying on attribute being choosen by DerivedMesh at drawing time. For this reason, we declare all possible attrib "name" for each data layer inside the batches. In the glsl code, we declare required data by type and name. Then Gawain only bind vbos if names correspond. This is way cleaner as we do not need to access the CD itself when drawing. One other problem is that the hash maybe 11 caracters long and rapidly overload gawain's attrib name buffer.
2017-05-01 17:58:15 +02:00
VertexFormat *format = &cache->shaded_triangles_format;
VertexFormat_clear(format);
/* initialize vertex format */
unsigned int pos_id = VertexFormat_add_attrib(format, "pos", COMP_F32, 3, KEEP_FLOAT);
unsigned int nor_id = VertexFormat_add_attrib(format, "nor", COMP_I16, 3, NORMALIZE_INT_TO_FLOAT);
unsigned int orco_id = VertexFormat_add_attrib(format, "orco", COMP_F32, 3, KEEP_FLOAT);
unsigned int *uv_id = MEM_mallocN(sizeof(*uv_id) * rdata->uv_ct, "UV attrib format");
unsigned int *uv_auto_id = MEM_mallocN(sizeof(*uv_id) * rdata->uv_ct, "UV attrib format");
unsigned int *vcol_id = MEM_mallocN(sizeof(*vcol_id) * rdata->vcol_ct, "Vcol attrib format");
unsigned int *vcol_auto_id = MEM_mallocN(sizeof(*vcol_id) * rdata->vcol_ct, "Vcol attrib format");
unsigned int *tangent_id = MEM_mallocN(sizeof(*tangent_id) * rdata->uv_ct, "Tangent attrib format");
/* XXX TODO : We are allocating for the active layers
* but we only need to bind the right layer to the default attrib.
* This is a gawain limitation to solve. */
unsigned int active_uv_id = VertexFormat_add_attrib(format, "u", COMP_F32, 2, KEEP_FLOAT);
unsigned int active_vcol_id = VertexFormat_add_attrib(format, "c", COMP_U8, 3, NORMALIZE_INT_TO_FLOAT);
unsigned int active_tangent_id = VertexFormat_add_attrib(format, "t", COMP_F32, 4, KEEP_FLOAT);
for (int i = 0; i < rdata->uv_ct; i++) {
/* UV */
attrib_name = mesh_render_data_uv_layer_name_get(rdata, i);
uv_id[i] = VertexFormat_add_attrib(format, attrib_name, COMP_F32, 2, KEEP_FLOAT);
/* Auto Name */
/* TODO Remove when when have aliases */
attrib_name = mesh_render_data_uv_auto_layer_name_get(rdata, i);
uv_auto_id[i] = VertexFormat_add_attrib(format, attrib_name, COMP_F32, 3, KEEP_FLOAT);
/* Tangent */
attrib_name = mesh_render_data_tangent_layer_name_get(rdata, i);
tangent_id[i] = VertexFormat_add_attrib(format, attrib_name, COMP_F32, 3, KEEP_FLOAT);
}
for (int i = 0; i < rdata->vcol_ct; i++) {
attrib_name = mesh_render_data_vcol_layer_name_get(rdata, i);
vcol_id[i] = VertexFormat_add_attrib(format, attrib_name, COMP_U8, 3, NORMALIZE_INT_TO_FLOAT);
/* Auto layer */
if (rdata->auto_vcol[i]) {
attrib_name = mesh_render_data_vcol_auto_layer_name_get(rdata, i);
vcol_auto_id[i] = VertexFormat_add_attrib(format, attrib_name, COMP_U8, 3, NORMALIZE_INT_TO_FLOAT);
}
Eevee: Make default shaders works. - Added New Batch cache routine to get the split mesh for each material. (optimization needed) - Did some shader modifications to create default_frag and support a somwhat correct fresnel for lamps (to finish)
2017-04-25 18:46:59 +02:00
}
const int tri_len = mesh_render_data_looptri_len_get(rdata);
Mesh Batch Cache: Add support for uv/vcol/tangent data layers. This is a bad implementation waiting for some improvement that are : - Gawain support for multiple names. -> will remove duplication of attribute alloc - Glsl safe name string escaping. -> will remove any risk of hash colision. Old glsl drawing was relying on attribute being choosen by DerivedMesh at drawing time. For this reason, we declare all possible attrib "name" for each data layer inside the batches. In the glsl code, we declare required data by type and name. Then Gawain only bind vbos if names correspond. This is way cleaner as we do not need to access the CD itself when drawing. One other problem is that the hash maybe 11 caracters long and rapidly overload gawain's attrib name buffer.
2017-05-01 17:58:15 +02:00
VertexBuffer *vbo = cache->shaded_triangles_data = VertexBuffer_create_with_format(format);
Eevee: Make default shaders works. - Added New Batch cache routine to get the split mesh for each material. (optimization needed) - Did some shader modifications to create default_frag and support a somwhat correct fresnel for lamps (to finish)
2017-04-25 18:46:59 +02:00
const int vbo_len_capacity = tri_len * 3;
int vbo_len_used = 0;
VertexBuffer_allocate_data(vbo, vbo_len_capacity);
/* TODO deduplicate all verts and make use of ElementList in mesh_batch_cache_get_shaded_triangles_in_order. */
for (int i = 0; i < tri_len; i++) {
float *tri_vert_cos[3];
Mesh Batch Cache: Add support for uv/vcol/tangent data layers. This is a bad implementation waiting for some improvement that are : - Gawain support for multiple names. -> will remove duplication of attribute alloc - Glsl safe name string escaping. -> will remove any risk of hash colision. Old glsl drawing was relying on attribute being choosen by DerivedMesh at drawing time. For this reason, we declare all possible attrib "name" for each data layer inside the batches. In the glsl code, we declare required data by type and name. Then Gawain only bind vbos if names correspond. This is way cleaner as we do not need to access the CD itself when drawing. One other problem is that the hash maybe 11 caracters long and rapidly overload gawain's attrib name buffer.
2017-05-01 17:58:15 +02:00
float *tri_uvs[3], *tri_tans[3], *tri_orcos[3];
unsigned char *tri_cols[3];
Eevee: Make default shaders works. - Added New Batch cache routine to get the split mesh for each material. (optimization needed) - Did some shader modifications to create default_frag and support a somwhat correct fresnel for lamps (to finish)
2017-04-25 18:46:59 +02:00
short *tri_nor, *tri_vert_nors[3];
bool is_smooth;
if (mesh_render_data_looptri_cos_nors_smooth_get(
rdata, i, &tri_vert_cos, &tri_nor, &tri_vert_nors, &is_smooth))
{
Mesh Batch Cache: Add support for uv/vcol/tangent data layers. This is a bad implementation waiting for some improvement that are : - Gawain support for multiple names. -> will remove duplication of attribute alloc - Glsl safe name string escaping. -> will remove any risk of hash colision. Old glsl drawing was relying on attribute being choosen by DerivedMesh at drawing time. For this reason, we declare all possible attrib "name" for each data layer inside the batches. In the glsl code, we declare required data by type and name. Then Gawain only bind vbos if names correspond. This is way cleaner as we do not need to access the CD itself when drawing. One other problem is that the hash maybe 11 caracters long and rapidly overload gawain's attrib name buffer.
2017-05-01 17:58:15 +02:00
/* NORs */
Eevee: Make default shaders works. - Added New Batch cache routine to get the split mesh for each material. (optimization needed) - Did some shader modifications to create default_frag and support a somwhat correct fresnel for lamps (to finish)
2017-04-25 18:46:59 +02:00
if (is_smooth) {
VertexBuffer_set_attrib(vbo, nor_id, nidx++, tri_vert_nors[0]);
VertexBuffer_set_attrib(vbo, nor_id, nidx++, tri_vert_nors[1]);
VertexBuffer_set_attrib(vbo, nor_id, nidx++, tri_vert_nors[2]);
}
else {
VertexBuffer_set_attrib(vbo, nor_id, nidx++, tri_nor);
VertexBuffer_set_attrib(vbo, nor_id, nidx++, tri_nor);
VertexBuffer_set_attrib(vbo, nor_id, nidx++, tri_nor);
}
Mesh Batch Cache: Add support for uv/vcol/tangent data layers. This is a bad implementation waiting for some improvement that are : - Gawain support for multiple names. -> will remove duplication of attribute alloc - Glsl safe name string escaping. -> will remove any risk of hash colision. Old glsl drawing was relying on attribute being choosen by DerivedMesh at drawing time. For this reason, we declare all possible attrib "name" for each data layer inside the batches. In the glsl code, we declare required data by type and name. Then Gawain only bind vbos if names correspond. This is way cleaner as we do not need to access the CD itself when drawing. One other problem is that the hash maybe 11 caracters long and rapidly overload gawain's attrib name buffer.
2017-05-01 17:58:15 +02:00
/* UVs & TANGENTs */
for (int j = 0; j < rdata->uv_ct; j++) {
mesh_render_data_looptri_uvs_get(rdata, i, j, &tri_uvs);
VertexBuffer_set_attrib(vbo, uv_id[j], vidx + 0, tri_uvs[0]);
VertexBuffer_set_attrib(vbo, uv_id[j], vidx + 1, tri_uvs[1]);
VertexBuffer_set_attrib(vbo, uv_id[j], vidx + 2, tri_uvs[2]);
/* TODO remove this when aliases will be implemented */
VertexBuffer_set_attrib(vbo, uv_auto_id[j], vidx + 0, tri_uvs[0]);
VertexBuffer_set_attrib(vbo, uv_auto_id[j], vidx + 1, tri_uvs[1]);
VertexBuffer_set_attrib(vbo, uv_auto_id[j], vidx + 2, tri_uvs[2]);
mesh_render_data_looptri_tans_get(rdata, i, j, &tri_tans);
VertexBuffer_set_attrib(vbo, tangent_id[j], vidx + 0, tri_tans[0]);
VertexBuffer_set_attrib(vbo, tangent_id[j], vidx + 1, tri_tans[1]);
VertexBuffer_set_attrib(vbo, tangent_id[j], vidx + 2, tri_tans[2]);
}
/* TODO remove this when aliases will be implemented */
if (rdata->uv_ct != 0) {
mesh_render_data_looptri_uvs_get(rdata, i, rdata->uv_active, &tri_uvs);
VertexBuffer_set_attrib(vbo, active_uv_id, vidx + 0, tri_uvs[0]);
VertexBuffer_set_attrib(vbo, active_uv_id, vidx + 1, tri_uvs[1]);
VertexBuffer_set_attrib(vbo, active_uv_id, vidx + 2, tri_uvs[2]);
mesh_render_data_looptri_tans_get(rdata, i, rdata->tangent_active, &tri_tans);
VertexBuffer_set_attrib(vbo, active_tangent_id, vidx + 0, tri_tans[0]);
VertexBuffer_set_attrib(vbo, active_tangent_id, vidx + 1, tri_tans[1]);
VertexBuffer_set_attrib(vbo, active_tangent_id, vidx + 2, tri_tans[2]);
}
/* VCOLs */
for (int j = 0; j < rdata->vcol_ct; j++) {
mesh_render_data_looptri_cols_get(rdata, i, j, &tri_cols);
VertexBuffer_set_attrib(vbo, vcol_id[j], vidx + 0, tri_cols[0]);
VertexBuffer_set_attrib(vbo, vcol_id[j], vidx + 1, tri_cols[1]);
VertexBuffer_set_attrib(vbo, vcol_id[j], vidx + 2, tri_cols[2]);
/* Auto layer */
if (rdata->auto_vcol[j]) {
/* TODO remove this when aliases will be implemented */
VertexBuffer_set_attrib(vbo, vcol_auto_id[j], vidx + 0, tri_cols[0]);
VertexBuffer_set_attrib(vbo, vcol_auto_id[j], vidx + 1, tri_cols[1]);
VertexBuffer_set_attrib(vbo, vcol_auto_id[j], vidx + 2, tri_cols[2]);
}
}
/* TODO remove this when aliases will be implemented */
if (rdata->vcol_ct != 0) {
mesh_render_data_looptri_cols_get(rdata, i, rdata->vcol_active, &tri_cols);
VertexBuffer_set_attrib(vbo, active_vcol_id, vidx + 0, tri_cols[0]);
VertexBuffer_set_attrib(vbo, active_vcol_id, vidx + 1, tri_cols[1]);
VertexBuffer_set_attrib(vbo, active_vcol_id, vidx + 2, tri_cols[2]);
}
/* ORCO */
mesh_render_data_looptri_orcos_get(rdata, i, &tri_orcos);
VertexBuffer_set_attrib(vbo, orco_id, vidx + 0, tri_orcos[0]);
VertexBuffer_set_attrib(vbo, orco_id, vidx + 1, tri_orcos[1]);
VertexBuffer_set_attrib(vbo, orco_id, vidx + 2, tri_orcos[2]);
/* COs */
Eevee: Make default shaders works. - Added New Batch cache routine to get the split mesh for each material. (optimization needed) - Did some shader modifications to create default_frag and support a somwhat correct fresnel for lamps (to finish)
2017-04-25 18:46:59 +02:00
VertexBuffer_set_attrib(vbo, pos_id, vidx++, tri_vert_cos[0]);
VertexBuffer_set_attrib(vbo, pos_id, vidx++, tri_vert_cos[1]);
VertexBuffer_set_attrib(vbo, pos_id, vidx++, tri_vert_cos[2]);
}
}
vbo_len_used = vidx;
if (vbo_len_capacity != vbo_len_used) {
VertexBuffer_resize_data(vbo, vbo_len_used);
}
Mesh Batch Cache: Add support for uv/vcol/tangent data layers. This is a bad implementation waiting for some improvement that are : - Gawain support for multiple names. -> will remove duplication of attribute alloc - Glsl safe name string escaping. -> will remove any risk of hash colision. Old glsl drawing was relying on attribute being choosen by DerivedMesh at drawing time. For this reason, we declare all possible attrib "name" for each data layer inside the batches. In the glsl code, we declare required data by type and name. Then Gawain only bind vbos if names correspond. This is way cleaner as we do not need to access the CD itself when drawing. One other problem is that the hash maybe 11 caracters long and rapidly overload gawain's attrib name buffer.
2017-05-01 17:58:15 +02:00
MEM_freeN(uv_id);
MEM_freeN(uv_auto_id);
MEM_freeN(vcol_id);
MEM_freeN(vcol_auto_id);
MEM_freeN(tangent_id);
Eevee: Make default shaders works. - Added New Batch cache routine to get the split mesh for each material. (optimization needed) - Did some shader modifications to create default_frag and support a somwhat correct fresnel for lamps (to finish)
2017-04-25 18:46:59 +02:00
}
return cache->shaded_triangles_data;
}
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
static VertexBuffer *mesh_batch_cache_get_pos_and_normals(MeshRenderData *rdata, MeshBatchCache *cache)
Edit Mesh overlay: Ported Display Normals option
2017-03-09 01:29:58 +01:00
{
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
BLI_assert(rdata->types & (MR_DATATYPE_VERT | MR_DATATYPE_LOOPTRI | MR_DATATYPE_LOOP | MR_DATATYPE_POLY));
Edit Mesh overlay: Ported Display Normals option
2017-03-09 01:29:58 +01:00
if (cache->pos_with_normals == NULL) {
unsigned int vidx = 0, nidx = 0;
static VertexFormat format = { 0 };
static unsigned int pos_id, nor_id;
if (format.attrib_ct == 0) {
/* initialize vertex format */
Gawain: VertexFormat_add_attrib (function name change) See intern/gawain for the API change. Other files are updated to use the new name. Also updated every call site to the recommended style: unsigned int foo = VertexFormat_add_attrib(format, "foo", COMP_ ... )
2017-04-06 00:59:04 -04:00
pos_id = VertexFormat_add_attrib(&format, "pos", COMP_F32, 3, KEEP_FLOAT);
nor_id = VertexFormat_add_attrib(&format, "nor", COMP_I16, 3, NORMALIZE_INT_TO_FLOAT);
Edit Mesh overlay: Ported Display Normals option
2017-03-09 01:29:58 +01:00
}
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
const int tri_len = mesh_render_data_looptri_len_get(rdata);
Edit Mesh overlay: Ported Display Normals option
2017-03-09 01:29:58 +01:00
Cleanup: replace num/ct/tot with 'len'
2017-04-19 00:10:32 +10:00
VertexBuffer *vbo = cache->pos_with_normals = VertexBuffer_create_with_format(&format);
Edit Mesh overlay: Ported Display Normals option
2017-03-09 01:29:58 +01:00
Cleanup: replace num/ct/tot with 'len'
2017-04-19 00:10:32 +10:00
const int vbo_len_capacity = tri_len * 3;
int vbo_len_used = 0;
VertexBuffer_allocate_data(vbo, vbo_len_capacity);
for (int i = 0; i < tri_len; i++) {
Edit Mesh overlay: Ported Display Normals option
2017-03-09 01:29:58 +01:00
float *tri_vert_cos[3];
short *tri_nor, *tri_vert_nors[3];
Draw Manager: T51218 Support edit-mesh hide
2017-04-18 23:29:55 +10:00
bool is_smooth;
Edit Mesh overlay: Ported Display Normals option
2017-03-09 01:29:58 +01:00
Draw Manager: T51218 Support edit-mesh hide
2017-04-18 23:29:55 +10:00
if (mesh_render_data_looptri_cos_nors_smooth_get(
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
rdata, i, &tri_vert_cos, &tri_nor, &tri_vert_nors, &is_smooth))
Draw Manager: T51218 Support edit-mesh hide
2017-04-18 23:29:55 +10:00
{
if (is_smooth) {
Cleanup: replace num/ct/tot with 'len'
2017-04-19 00:10:32 +10:00
VertexBuffer_set_attrib(vbo, nor_id, nidx++, tri_vert_nors[0]);
VertexBuffer_set_attrib(vbo, nor_id, nidx++, tri_vert_nors[1]);
VertexBuffer_set_attrib(vbo, nor_id, nidx++, tri_vert_nors[2]);
Draw Manager: T51218 Support edit-mesh hide
2017-04-18 23:29:55 +10:00
}
else {
Cleanup: replace num/ct/tot with 'len'
2017-04-19 00:10:32 +10:00
VertexBuffer_set_attrib(vbo, nor_id, nidx++, tri_nor);
VertexBuffer_set_attrib(vbo, nor_id, nidx++, tri_nor);
VertexBuffer_set_attrib(vbo, nor_id, nidx++, tri_nor);
Draw Manager: T51218 Support edit-mesh hide
2017-04-18 23:29:55 +10:00
}
Edit Mesh overlay: Ported Display Normals option
2017-03-09 01:29:58 +01:00
Cleanup: replace num/ct/tot with 'len'
2017-04-19 00:10:32 +10:00
VertexBuffer_set_attrib(vbo, pos_id, vidx++, tri_vert_cos[0]);
VertexBuffer_set_attrib(vbo, pos_id, vidx++, tri_vert_cos[1]);
VertexBuffer_set_attrib(vbo, pos_id, vidx++, tri_vert_cos[2]);
Edit Mesh overlay: Ported Display Normals option
2017-03-09 01:29:58 +01:00
}
Draw Manager: T51218 Support edit-mesh hide
2017-04-18 23:29:55 +10:00
}
Cleanup: replace num/ct/tot with 'len'
2017-04-19 00:10:32 +10:00
vbo_len_used = vidx;
Edit Mesh overlay: Ported Display Normals option
2017-03-09 01:29:58 +01:00
Cleanup: replace num/ct/tot with 'len'
2017-04-19 00:10:32 +10:00
if (vbo_len_capacity != vbo_len_used) {
VertexBuffer_resize_data(vbo, vbo_len_used);
Edit Mesh overlay: Ported Display Normals option
2017-03-09 01:29:58 +01:00
}
}
return cache->pos_with_normals;
}
Eevee: Make default shaders works. - Added New Batch cache routine to get the split mesh for each material. (optimization needed) - Did some shader modifications to create default_frag and support a somwhat correct fresnel for lamps (to finish)
2017-04-25 18:46:59 +02:00
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
static VertexBuffer *mesh_batch_cache_get_pos_and_nor_in_order(MeshRenderData *rdata, MeshBatchCache *cache)
Rework of BKE's mesh_render to support BMesh directly. Note that since there is no (efficient) ways to get arrays of MVert/MEdge/etc. out of a BMesh, I refactored quite heavily internals of BKE_mesh_render. Now, when you do need acess to mesh data to generate cached batches, you create an abstract struct from mesh (either Mesh or BMesh if available), and then use advanced helpers to extract needed data, on a per-item basis (no more handling of arrays of verts/edges/... in batches code). This allows to: * Avoid having to create arrays of BMesh elements. * Take advantage of existing advanced BMesh topology and connectivity data. Reviewers: dfelinto, fclem, merwin Differential Revision: https://developer.blender.org/D2521
2017-02-23 11:54:40 +01:00
{
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
BLI_assert(rdata->types & MR_DATATYPE_VERT);
Convert MBC_ API to Mesh (instead of derived mesh) and move it to mesh_render This includes a few fixes in the MBC_ api. The idea here is for this to be the only interface the render engines will deal with for the meshes. If we need to expose special options for sculpting engine we refactor this accordingly. But for now we are shaping this in a per-case base. Note: * We still need to hook up to the depsgraph to force clear/update of batch_cache when mesh changes (I'm waiting for Sergey Sharybin's depsgraph update for this though) * Also ideally we could/should use BMesh directly instead of DerivedMesh, but this will do for now. Note 2: In the end I renamed the `BKE_mesh_render` functions to `static mesh_render`. We can re-expose them as BKE_* later once we need it. Reviewers: merwin Subscribers: fclem Differential Revision: https://developer.blender.org/D2476
2017-01-25 09:16:29 +01:00
if (cache->pos_in_order == NULL) {
static VertexFormat format = { 0 };
Edit Mesh overlay: Ported Display Normals option
2017-03-09 01:29:58 +01:00
static unsigned pos_id, nor_id;
Convert MBC_ API to Mesh (instead of derived mesh) and move it to mesh_render This includes a few fixes in the MBC_ api. The idea here is for this to be the only interface the render engines will deal with for the meshes. If we need to expose special options for sculpting engine we refactor this accordingly. But for now we are shaping this in a per-case base. Note: * We still need to hook up to the depsgraph to force clear/update of batch_cache when mesh changes (I'm waiting for Sergey Sharybin's depsgraph update for this though) * Also ideally we could/should use BMesh directly instead of DerivedMesh, but this will do for now. Note 2: In the end I renamed the `BKE_mesh_render` functions to `static mesh_render`. We can re-expose them as BKE_* later once we need it. Reviewers: merwin Subscribers: fclem Differential Revision: https://developer.blender.org/D2476
2017-01-25 09:16:29 +01:00
if (format.attrib_ct == 0) {
/* initialize vertex format */
Gawain: VertexFormat_add_attrib (function name change) See intern/gawain for the API change. Other files are updated to use the new name. Also updated every call site to the recommended style: unsigned int foo = VertexFormat_add_attrib(format, "foo", COMP_ ... )
2017-04-06 00:59:04 -04:00
pos_id = VertexFormat_add_attrib(&format, "pos", COMP_F32, 3, KEEP_FLOAT);
nor_id = VertexFormat_add_attrib(&format, "nor", COMP_I16, 3, NORMALIZE_INT_TO_FLOAT);
Convert MBC_ API to Mesh (instead of derived mesh) and move it to mesh_render This includes a few fixes in the MBC_ api. The idea here is for this to be the only interface the render engines will deal with for the meshes. If we need to expose special options for sculpting engine we refactor this accordingly. But for now we are shaping this in a per-case base. Note: * We still need to hook up to the depsgraph to force clear/update of batch_cache when mesh changes (I'm waiting for Sergey Sharybin's depsgraph update for this though) * Also ideally we could/should use BMesh directly instead of DerivedMesh, but this will do for now. Note 2: In the end I renamed the `BKE_mesh_render` functions to `static mesh_render`. We can re-expose them as BKE_* later once we need it. Reviewers: merwin Subscribers: fclem Differential Revision: https://developer.blender.org/D2476
2017-01-25 09:16:29 +01:00
}
Cleanup: replace num/ct/tot with 'len'
2017-04-19 00:10:32 +10:00
VertexBuffer *vbo = cache->pos_in_order = VertexBuffer_create_with_format(&format);
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
const int vbo_len_capacity = mesh_render_data_verts_len_get(rdata);
Cleanup: replace num/ct/tot with 'len'
2017-04-19 00:10:32 +10:00
VertexBuffer_allocate_data(vbo, vbo_len_capacity);
for (int i = 0; i < vbo_len_capacity; ++i) {
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
VertexBuffer_set_attrib(vbo, pos_id, i, mesh_render_data_vert_co(rdata, i));
VertexBuffer_set_attrib(vbo, nor_id, i, mesh_render_data_vert_nor(rdata, i));
Convert MBC_ API to Mesh (instead of derived mesh) and move it to mesh_render This includes a few fixes in the MBC_ api. The idea here is for this to be the only interface the render engines will deal with for the meshes. If we need to expose special options for sculpting engine we refactor this accordingly. But for now we are shaping this in a per-case base. Note: * We still need to hook up to the depsgraph to force clear/update of batch_cache when mesh changes (I'm waiting for Sergey Sharybin's depsgraph update for this though) * Also ideally we could/should use BMesh directly instead of DerivedMesh, but this will do for now. Note 2: In the end I renamed the `BKE_mesh_render` functions to `static mesh_render`. We can re-expose them as BKE_* later once we need it. Reviewers: merwin Subscribers: fclem Differential Revision: https://developer.blender.org/D2476
2017-01-25 09:16:29 +01:00
}
}
return cache->pos_in_order;
}
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
static ElementList *mesh_batch_cache_get_edges_in_order(MeshRenderData *rdata, MeshBatchCache *cache)
Convert MBC_ API to Mesh (instead of derived mesh) and move it to mesh_render This includes a few fixes in the MBC_ api. The idea here is for this to be the only interface the render engines will deal with for the meshes. If we need to expose special options for sculpting engine we refactor this accordingly. But for now we are shaping this in a per-case base. Note: * We still need to hook up to the depsgraph to force clear/update of batch_cache when mesh changes (I'm waiting for Sergey Sharybin's depsgraph update for this though) * Also ideally we could/should use BMesh directly instead of DerivedMesh, but this will do for now. Note 2: In the end I renamed the `BKE_mesh_render` functions to `static mesh_render`. We can re-expose them as BKE_* later once we need it. Reviewers: merwin Subscribers: fclem Differential Revision: https://developer.blender.org/D2476
2017-01-25 09:16:29 +01:00
{
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
BLI_assert(rdata->types & (MR_DATATYPE_VERT | MR_DATATYPE_EDGE));
Convert MBC_ API to Mesh (instead of derived mesh) and move it to mesh_render This includes a few fixes in the MBC_ api. The idea here is for this to be the only interface the render engines will deal with for the meshes. If we need to expose special options for sculpting engine we refactor this accordingly. But for now we are shaping this in a per-case base. Note: * We still need to hook up to the depsgraph to force clear/update of batch_cache when mesh changes (I'm waiting for Sergey Sharybin's depsgraph update for this though) * Also ideally we could/should use BMesh directly instead of DerivedMesh, but this will do for now. Note 2: In the end I renamed the `BKE_mesh_render` functions to `static mesh_render`. We can re-expose them as BKE_* later once we need it. Reviewers: merwin Subscribers: fclem Differential Revision: https://developer.blender.org/D2476
2017-01-25 09:16:29 +01:00
if (cache->edges_in_order == NULL) {
Rework of BKE's mesh_render to support BMesh directly. Note that since there is no (efficient) ways to get arrays of MVert/MEdge/etc. out of a BMesh, I refactored quite heavily internals of BKE_mesh_render. Now, when you do need acess to mesh data to generate cached batches, you create an abstract struct from mesh (either Mesh or BMesh if available), and then use advanced helpers to extract needed data, on a per-item basis (no more handling of arrays of verts/edges/... in batches code). This allows to: * Avoid having to create arrays of BMesh elements. * Take advantage of existing advanced BMesh topology and connectivity data. Reviewers: dfelinto, fclem, merwin Differential Revision: https://developer.blender.org/D2521
2017-02-23 11:54:40 +01:00
printf("Caching edges in order...\n");
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
const int vert_len = mesh_render_data_verts_len_get(rdata);
const int edge_len = mesh_render_data_edges_len_get(rdata);
Convert MBC_ API to Mesh (instead of derived mesh) and move it to mesh_render This includes a few fixes in the MBC_ api. The idea here is for this to be the only interface the render engines will deal with for the meshes. If we need to expose special options for sculpting engine we refactor this accordingly. But for now we are shaping this in a per-case base. Note: * We still need to hook up to the depsgraph to force clear/update of batch_cache when mesh changes (I'm waiting for Sergey Sharybin's depsgraph update for this though) * Also ideally we could/should use BMesh directly instead of DerivedMesh, but this will do for now. Note 2: In the end I renamed the `BKE_mesh_render` functions to `static mesh_render`. We can re-expose them as BKE_* later once we need it. Reviewers: merwin Subscribers: fclem Differential Revision: https://developer.blender.org/D2476
2017-01-25 09:16:29 +01:00
ElementListBuilder elb;
Cleanup: replace num/ct/tot with 'len'
2017-04-19 00:10:32 +10:00
ElementListBuilder_init(&elb, PRIM_LINES, edge_len, vert_len);
for (int i = 0; i < edge_len; ++i) {
Rework of BKE's mesh_render to support BMesh directly. Note that since there is no (efficient) ways to get arrays of MVert/MEdge/etc. out of a BMesh, I refactored quite heavily internals of BKE_mesh_render. Now, when you do need acess to mesh data to generate cached batches, you create an abstract struct from mesh (either Mesh or BMesh if available), and then use advanced helpers to extract needed data, on a per-item basis (no more handling of arrays of verts/edges/... in batches code). This allows to: * Avoid having to create arrays of BMesh elements. * Take advantage of existing advanced BMesh topology and connectivity data. Reviewers: dfelinto, fclem, merwin Differential Revision: https://developer.blender.org/D2521
2017-02-23 11:54:40 +01:00
int vert_idx[2];
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
if (mesh_render_data_edge_verts_indices_get(rdata, i, vert_idx)) {
Draw Manager: T51218 Support edit-mesh hide
2017-04-18 23:29:55 +10:00
add_line_vertices(&elb, vert_idx[0], vert_idx[1]);
}
Convert MBC_ API to Mesh (instead of derived mesh) and move it to mesh_render This includes a few fixes in the MBC_ api. The idea here is for this to be the only interface the render engines will deal with for the meshes. If we need to expose special options for sculpting engine we refactor this accordingly. But for now we are shaping this in a per-case base. Note: * We still need to hook up to the depsgraph to force clear/update of batch_cache when mesh changes (I'm waiting for Sergey Sharybin's depsgraph update for this though) * Also ideally we could/should use BMesh directly instead of DerivedMesh, but this will do for now. Note 2: In the end I renamed the `BKE_mesh_render` functions to `static mesh_render`. We can re-expose them as BKE_* later once we need it. Reviewers: merwin Subscribers: fclem Differential Revision: https://developer.blender.org/D2476
2017-01-25 09:16:29 +01:00
}
cache->edges_in_order = ElementList_build(&elb);
}
return cache->edges_in_order;
}
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
static ElementList *mesh_batch_cache_get_triangles_in_order(MeshRenderData *rdata, MeshBatchCache *cache)
Convert MBC_ API to Mesh (instead of derived mesh) and move it to mesh_render This includes a few fixes in the MBC_ api. The idea here is for this to be the only interface the render engines will deal with for the meshes. If we need to expose special options for sculpting engine we refactor this accordingly. But for now we are shaping this in a per-case base. Note: * We still need to hook up to the depsgraph to force clear/update of batch_cache when mesh changes (I'm waiting for Sergey Sharybin's depsgraph update for this though) * Also ideally we could/should use BMesh directly instead of DerivedMesh, but this will do for now. Note 2: In the end I renamed the `BKE_mesh_render` functions to `static mesh_render`. We can re-expose them as BKE_* later once we need it. Reviewers: merwin Subscribers: fclem Differential Revision: https://developer.blender.org/D2476
2017-01-25 09:16:29 +01:00
{
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
BLI_assert(rdata->types & (MR_DATATYPE_VERT | MR_DATATYPE_LOOPTRI));
Convert MBC_ API to Mesh (instead of derived mesh) and move it to mesh_render This includes a few fixes in the MBC_ api. The idea here is for this to be the only interface the render engines will deal with for the meshes. If we need to expose special options for sculpting engine we refactor this accordingly. But for now we are shaping this in a per-case base. Note: * We still need to hook up to the depsgraph to force clear/update of batch_cache when mesh changes (I'm waiting for Sergey Sharybin's depsgraph update for this though) * Also ideally we could/should use BMesh directly instead of DerivedMesh, but this will do for now. Note 2: In the end I renamed the `BKE_mesh_render` functions to `static mesh_render`. We can re-expose them as BKE_* later once we need it. Reviewers: merwin Subscribers: fclem Differential Revision: https://developer.blender.org/D2476
2017-01-25 09:16:29 +01:00
if (cache->triangles_in_order == NULL) {
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
const int vert_len = mesh_render_data_verts_len_get(rdata);
const int tri_len = mesh_render_data_looptri_len_get(rdata);
Convert MBC_ API to Mesh (instead of derived mesh) and move it to mesh_render This includes a few fixes in the MBC_ api. The idea here is for this to be the only interface the render engines will deal with for the meshes. If we need to expose special options for sculpting engine we refactor this accordingly. But for now we are shaping this in a per-case base. Note: * We still need to hook up to the depsgraph to force clear/update of batch_cache when mesh changes (I'm waiting for Sergey Sharybin's depsgraph update for this though) * Also ideally we could/should use BMesh directly instead of DerivedMesh, but this will do for now. Note 2: In the end I renamed the `BKE_mesh_render` functions to `static mesh_render`. We can re-expose them as BKE_* later once we need it. Reviewers: merwin Subscribers: fclem Differential Revision: https://developer.blender.org/D2476
2017-01-25 09:16:29 +01:00
ElementListBuilder elb;
Cleanup: replace num/ct/tot with 'len'
2017-04-19 00:10:32 +10:00
ElementListBuilder_init(&elb, PRIM_TRIANGLES, tri_len, vert_len);
for (int i = 0; i < tri_len; ++i) {
Rework of BKE's mesh_render to support BMesh directly. Note that since there is no (efficient) ways to get arrays of MVert/MEdge/etc. out of a BMesh, I refactored quite heavily internals of BKE_mesh_render. Now, when you do need acess to mesh data to generate cached batches, you create an abstract struct from mesh (either Mesh or BMesh if available), and then use advanced helpers to extract needed data, on a per-item basis (no more handling of arrays of verts/edges/... in batches code). This allows to: * Avoid having to create arrays of BMesh elements. * Take advantage of existing advanced BMesh topology and connectivity data. Reviewers: dfelinto, fclem, merwin Differential Revision: https://developer.blender.org/D2521
2017-02-23 11:54:40 +01:00
int tri_vert_idx[3];
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
if (mesh_render_data_looptri_vert_indices_get(rdata, i, tri_vert_idx)) {
Draw Manager: T51218 Support edit-mesh hide
2017-04-18 23:29:55 +10:00
add_triangle_vertices(&elb, tri_vert_idx[0], tri_vert_idx[1], tri_vert_idx[2]);
}
Convert MBC_ API to Mesh (instead of derived mesh) and move it to mesh_render This includes a few fixes in the MBC_ api. The idea here is for this to be the only interface the render engines will deal with for the meshes. If we need to expose special options for sculpting engine we refactor this accordingly. But for now we are shaping this in a per-case base. Note: * We still need to hook up to the depsgraph to force clear/update of batch_cache when mesh changes (I'm waiting for Sergey Sharybin's depsgraph update for this though) * Also ideally we could/should use BMesh directly instead of DerivedMesh, but this will do for now. Note 2: In the end I renamed the `BKE_mesh_render` functions to `static mesh_render`. We can re-expose them as BKE_* later once we need it. Reviewers: merwin Subscribers: fclem Differential Revision: https://developer.blender.org/D2476
2017-01-25 09:16:29 +01:00
}
cache->triangles_in_order = ElementList_build(&elb);
}
return cache->triangles_in_order;
}
Eevee: Make default shaders works. - Added New Batch cache routine to get the split mesh for each material. (optimization needed) - Did some shader modifications to create default_frag and support a somwhat correct fresnel for lamps (to finish)
2017-04-25 18:46:59 +02:00
static ElementList **mesh_batch_cache_get_shaded_triangles_in_order(MeshRenderData *rdata, MeshBatchCache *cache)
{
BLI_assert(rdata->types & (MR_DATATYPE_VERT | MR_DATATYPE_LOOPTRI | MR_DATATYPE_POLY));
if (cache->shaded_triangles_in_order == NULL) {
const int tri_len = mesh_render_data_looptri_len_get(rdata);
const int mat_ct = mesh_render_data_mat_ct_get(rdata);
int *mat_tri_len = MEM_callocN(sizeof(*mat_tri_len) * mat_ct, "mat_tri_len");
cache->shaded_triangles_in_order = MEM_callocN(sizeof(*cache->shaded_triangles) * mat_ct, "shaded_triangles_in_order");
ElementListBuilder *elb = MEM_callocN(sizeof(*elb) * mat_ct, "shaded ElementListBuilder");
for (int i = 0; i < tri_len; ++i) {
short ma_id;
if (mesh_render_data_looptri_mat_index_get(rdata, i, &ma_id)) {
mat_tri_len[ma_id] += 1;
}
}
/* Init ELBs. */
for (int i = 0; i < mat_ct; ++i) {
ElementListBuilder_init(&elb[i], PRIM_TRIANGLES, mat_tri_len[i], tri_len * 3);
}
/* Populate ELBs. */
unsigned int nidx = 0;
for (int i = 0; i < tri_len; ++i) {
short ma_id;
/* TODO deduplicate verts see mesh_batch_cache_get_shading_data */
if (mesh_render_data_looptri_mat_index_get(rdata, i, &ma_id)) {
add_triangle_vertices(&elb[ma_id], nidx + 0, nidx + 1, nidx + 2);
nidx += 3;
}
}
/* Build ELBs. */
for (int i = 0; i < mat_ct; ++i) {
cache->shaded_triangles_in_order[i] = ElementList_build(&elb[i]);
}
MEM_freeN(mat_tri_len);
MEM_freeN(elb);
}
return cache->shaded_triangles_in_order;
}
Cleanup: move draw-cache creation from BKE to DRW Creating draw-cache should only ever be used by the draw-manager.
2017-04-21 21:14:11 +10:00
Batch *DRW_mesh_batch_cache_get_all_edges(Mesh *me)
Convert MBC_ API to Mesh (instead of derived mesh) and move it to mesh_render This includes a few fixes in the MBC_ api. The idea here is for this to be the only interface the render engines will deal with for the meshes. If we need to expose special options for sculpting engine we refactor this accordingly. But for now we are shaping this in a per-case base. Note: * We still need to hook up to the depsgraph to force clear/update of batch_cache when mesh changes (I'm waiting for Sergey Sharybin's depsgraph update for this though) * Also ideally we could/should use BMesh directly instead of DerivedMesh, but this will do for now. Note 2: In the end I renamed the `BKE_mesh_render` functions to `static mesh_render`. We can re-expose them as BKE_* later once we need it. Reviewers: merwin Subscribers: fclem Differential Revision: https://developer.blender.org/D2476
2017-01-25 09:16:29 +01:00
{
MeshBatchCache *cache = mesh_batch_cache_get(me);
if (cache->all_edges == NULL) {
/* create batch from Mesh */
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
MeshRenderData *rdata = mesh_render_data_create(me, MR_DATATYPE_VERT | MR_DATATYPE_EDGE);
Rework of BKE's mesh_render to support BMesh directly. Note that since there is no (efficient) ways to get arrays of MVert/MEdge/etc. out of a BMesh, I refactored quite heavily internals of BKE_mesh_render. Now, when you do need acess to mesh data to generate cached batches, you create an abstract struct from mesh (either Mesh or BMesh if available), and then use advanced helpers to extract needed data, on a per-item basis (no more handling of arrays of verts/edges/... in batches code). This allows to: * Avoid having to create arrays of BMesh elements. * Take advantage of existing advanced BMesh topology and connectivity data. Reviewers: dfelinto, fclem, merwin Differential Revision: https://developer.blender.org/D2521
2017-02-23 11:54:40 +01:00
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
cache->all_edges = Batch_create(PRIM_LINES, mesh_batch_cache_get_pos_and_nor_in_order(rdata, cache),
mesh_batch_cache_get_edges_in_order(rdata, cache));
Rework of BKE's mesh_render to support BMesh directly. Note that since there is no (efficient) ways to get arrays of MVert/MEdge/etc. out of a BMesh, I refactored quite heavily internals of BKE_mesh_render. Now, when you do need acess to mesh data to generate cached batches, you create an abstract struct from mesh (either Mesh or BMesh if available), and then use advanced helpers to extract needed data, on a per-item basis (no more handling of arrays of verts/edges/... in batches code). This allows to: * Avoid having to create arrays of BMesh elements. * Take advantage of existing advanced BMesh topology and connectivity data. Reviewers: dfelinto, fclem, merwin Differential Revision: https://developer.blender.org/D2521
2017-02-23 11:54:40 +01:00
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
mesh_render_data_free(rdata);
Convert MBC_ API to Mesh (instead of derived mesh) and move it to mesh_render This includes a few fixes in the MBC_ api. The idea here is for this to be the only interface the render engines will deal with for the meshes. If we need to expose special options for sculpting engine we refactor this accordingly. But for now we are shaping this in a per-case base. Note: * We still need to hook up to the depsgraph to force clear/update of batch_cache when mesh changes (I'm waiting for Sergey Sharybin's depsgraph update for this though) * Also ideally we could/should use BMesh directly instead of DerivedMesh, but this will do for now. Note 2: In the end I renamed the `BKE_mesh_render` functions to `static mesh_render`. We can re-expose them as BKE_* later once we need it. Reviewers: merwin Subscribers: fclem Differential Revision: https://developer.blender.org/D2476
2017-01-25 09:16:29 +01:00
}
return cache->all_edges;
}
Cleanup: move draw-cache creation from BKE to DRW Creating draw-cache should only ever be used by the draw-manager.
2017-04-21 21:14:11 +10:00
Batch *DRW_mesh_batch_cache_get_all_triangles(Mesh *me)
Convert MBC_ API to Mesh (instead of derived mesh) and move it to mesh_render This includes a few fixes in the MBC_ api. The idea here is for this to be the only interface the render engines will deal with for the meshes. If we need to expose special options for sculpting engine we refactor this accordingly. But for now we are shaping this in a per-case base. Note: * We still need to hook up to the depsgraph to force clear/update of batch_cache when mesh changes (I'm waiting for Sergey Sharybin's depsgraph update for this though) * Also ideally we could/should use BMesh directly instead of DerivedMesh, but this will do for now. Note 2: In the end I renamed the `BKE_mesh_render` functions to `static mesh_render`. We can re-expose them as BKE_* later once we need it. Reviewers: merwin Subscribers: fclem Differential Revision: https://developer.blender.org/D2476
2017-01-25 09:16:29 +01:00
{
MeshBatchCache *cache = mesh_batch_cache_get(me);
if (cache->all_triangles == NULL) {
/* create batch from DM */
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
MeshRenderData *rdata = mesh_render_data_create(me, MR_DATATYPE_VERT | MR_DATATYPE_LOOPTRI);
Rework of BKE's mesh_render to support BMesh directly. Note that since there is no (efficient) ways to get arrays of MVert/MEdge/etc. out of a BMesh, I refactored quite heavily internals of BKE_mesh_render. Now, when you do need acess to mesh data to generate cached batches, you create an abstract struct from mesh (either Mesh or BMesh if available), and then use advanced helpers to extract needed data, on a per-item basis (no more handling of arrays of verts/edges/... in batches code). This allows to: * Avoid having to create arrays of BMesh elements. * Take advantage of existing advanced BMesh topology and connectivity data. Reviewers: dfelinto, fclem, merwin Differential Revision: https://developer.blender.org/D2521
2017-02-23 11:54:40 +01:00
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
cache->all_triangles = Batch_create(PRIM_TRIANGLES, mesh_batch_cache_get_pos_and_nor_in_order(rdata, cache),
mesh_batch_cache_get_triangles_in_order(rdata, cache));
Rework of BKE's mesh_render to support BMesh directly. Note that since there is no (efficient) ways to get arrays of MVert/MEdge/etc. out of a BMesh, I refactored quite heavily internals of BKE_mesh_render. Now, when you do need acess to mesh data to generate cached batches, you create an abstract struct from mesh (either Mesh or BMesh if available), and then use advanced helpers to extract needed data, on a per-item basis (no more handling of arrays of verts/edges/... in batches code). This allows to: * Avoid having to create arrays of BMesh elements. * Take advantage of existing advanced BMesh topology and connectivity data. Reviewers: dfelinto, fclem, merwin Differential Revision: https://developer.blender.org/D2521
2017-02-23 11:54:40 +01:00
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
mesh_render_data_free(rdata);
Convert MBC_ API to Mesh (instead of derived mesh) and move it to mesh_render This includes a few fixes in the MBC_ api. The idea here is for this to be the only interface the render engines will deal with for the meshes. If we need to expose special options for sculpting engine we refactor this accordingly. But for now we are shaping this in a per-case base. Note: * We still need to hook up to the depsgraph to force clear/update of batch_cache when mesh changes (I'm waiting for Sergey Sharybin's depsgraph update for this though) * Also ideally we could/should use BMesh directly instead of DerivedMesh, but this will do for now. Note 2: In the end I renamed the `BKE_mesh_render` functions to `static mesh_render`. We can re-expose them as BKE_* later once we need it. Reviewers: merwin Subscribers: fclem Differential Revision: https://developer.blender.org/D2476
2017-01-25 09:16:29 +01:00
}
return cache->all_triangles;
}
Cleanup: move draw-cache creation from BKE to DRW Creating draw-cache should only ever be used by the draw-manager.
2017-04-21 21:14:11 +10:00
Batch *DRW_mesh_batch_cache_get_triangles_with_normals(Mesh *me)
Clay Engine: Support vertex normals (smooth shading), no face normals yet.
2017-02-15 13:32:35 +01:00
{
MeshBatchCache *cache = mesh_batch_cache_get(me);
if (cache->triangles_with_normals == NULL) {
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
MeshRenderData *rdata = mesh_render_data_create(me, MR_DATATYPE_VERT | MR_DATATYPE_LOOPTRI | MR_DATATYPE_LOOP | MR_DATATYPE_POLY);
Clay Engine: Support vertex normals (smooth shading), no face normals yet.
2017-02-15 13:32:35 +01:00
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
cache->triangles_with_normals = Batch_create(PRIM_TRIANGLES, mesh_batch_cache_get_pos_and_normals(rdata, cache), NULL);
Clay Engine: Support vertex normals (smooth shading), no face normals yet.
2017-02-15 13:32:35 +01:00
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
mesh_render_data_free(rdata);
Edit Mesh overlay: Ported Display Normals option
2017-03-09 01:29:58 +01:00
}
Clay Engine: Support vertex normals (smooth shading), no face normals yet.
2017-02-15 13:32:35 +01:00
Edit Mesh overlay: Ported Display Normals option
2017-03-09 01:29:58 +01:00
return cache->triangles_with_normals;
}
Clay Engine: Correct Mesh Normals
2017-02-16 16:19:48 +01:00
Cleanup: move draw-cache creation from BKE to DRW Creating draw-cache should only ever be used by the draw-manager.
2017-04-21 21:14:11 +10:00
Batch *DRW_mesh_batch_cache_get_points_with_normals(Mesh *me)
Edit Mesh overlay: Ported Display Normals option
2017-03-09 01:29:58 +01:00
{
MeshBatchCache *cache = mesh_batch_cache_get(me);
Clay Engine: Support vertex normals (smooth shading), no face normals yet.
2017-02-15 13:32:35 +01:00
Edit Mesh overlay: Ported Display Normals option
2017-03-09 01:29:58 +01:00
if (cache->points_with_normals == NULL) {
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
MeshRenderData *rdata = mesh_render_data_create(me, MR_DATATYPE_VERT | MR_DATATYPE_LOOPTRI | MR_DATATYPE_LOOP | MR_DATATYPE_POLY);
Clay Engine: Support vertex normals (smooth shading), no face normals yet.
2017-02-15 13:32:35 +01:00
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
cache->points_with_normals = Batch_create(PRIM_POINTS, mesh_batch_cache_get_pos_and_normals(rdata, cache), NULL);
Clay Engine: Correct Mesh Normals
2017-02-16 16:19:48 +01:00
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
mesh_render_data_free(rdata);
Clay Engine: Support vertex normals (smooth shading), no face normals yet.
2017-02-15 13:32:35 +01:00
}
Edit Mesh overlay: Ported Display Normals option
2017-03-09 01:29:58 +01:00
return cache->points_with_normals;
Clay Engine: Support vertex normals (smooth shading), no face normals yet.
2017-02-15 13:32:35 +01:00
}
Cleanup: move draw-cache creation from BKE to DRW Creating draw-cache should only ever be used by the draw-manager.
2017-04-21 21:14:11 +10:00
Batch *DRW_mesh_batch_cache_get_all_verts(Mesh *me)
Convert MBC_ API to Mesh (instead of derived mesh) and move it to mesh_render This includes a few fixes in the MBC_ api. The idea here is for this to be the only interface the render engines will deal with for the meshes. If we need to expose special options for sculpting engine we refactor this accordingly. But for now we are shaping this in a per-case base. Note: * We still need to hook up to the depsgraph to force clear/update of batch_cache when mesh changes (I'm waiting for Sergey Sharybin's depsgraph update for this though) * Also ideally we could/should use BMesh directly instead of DerivedMesh, but this will do for now. Note 2: In the end I renamed the `BKE_mesh_render` functions to `static mesh_render`. We can re-expose them as BKE_* later once we need it. Reviewers: merwin Subscribers: fclem Differential Revision: https://developer.blender.org/D2476
2017-01-25 09:16:29 +01:00
{
MeshBatchCache *cache = mesh_batch_cache_get(me);
if (cache->all_verts == NULL) {
/* create batch from DM */
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
MeshRenderData *rdata = mesh_render_data_create(me, MR_DATATYPE_VERT);
Rework of BKE's mesh_render to support BMesh directly. Note that since there is no (efficient) ways to get arrays of MVert/MEdge/etc. out of a BMesh, I refactored quite heavily internals of BKE_mesh_render. Now, when you do need acess to mesh data to generate cached batches, you create an abstract struct from mesh (either Mesh or BMesh if available), and then use advanced helpers to extract needed data, on a per-item basis (no more handling of arrays of verts/edges/... in batches code). This allows to: * Avoid having to create arrays of BMesh elements. * Take advantage of existing advanced BMesh topology and connectivity data. Reviewers: dfelinto, fclem, merwin Differential Revision: https://developer.blender.org/D2521
2017-02-23 11:54:40 +01:00
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
cache->all_verts = Batch_create(PRIM_POINTS, mesh_batch_cache_get_pos_and_nor_in_order(rdata, cache), NULL);
Rework of BKE's mesh_render to support BMesh directly. Note that since there is no (efficient) ways to get arrays of MVert/MEdge/etc. out of a BMesh, I refactored quite heavily internals of BKE_mesh_render. Now, when you do need acess to mesh data to generate cached batches, you create an abstract struct from mesh (either Mesh or BMesh if available), and then use advanced helpers to extract needed data, on a per-item basis (no more handling of arrays of verts/edges/... in batches code). This allows to: * Avoid having to create arrays of BMesh elements. * Take advantage of existing advanced BMesh topology and connectivity data. Reviewers: dfelinto, fclem, merwin Differential Revision: https://developer.blender.org/D2521
2017-02-23 11:54:40 +01:00
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
mesh_render_data_free(rdata);
Convert MBC_ API to Mesh (instead of derived mesh) and move it to mesh_render This includes a few fixes in the MBC_ api. The idea here is for this to be the only interface the render engines will deal with for the meshes. If we need to expose special options for sculpting engine we refactor this accordingly. But for now we are shaping this in a per-case base. Note: * We still need to hook up to the depsgraph to force clear/update of batch_cache when mesh changes (I'm waiting for Sergey Sharybin's depsgraph update for this though) * Also ideally we could/should use BMesh directly instead of DerivedMesh, but this will do for now. Note 2: In the end I renamed the `BKE_mesh_render` functions to `static mesh_render`. We can re-expose them as BKE_* later once we need it. Reviewers: merwin Subscribers: fclem Differential Revision: https://developer.blender.org/D2476
2017-01-25 09:16:29 +01:00
}
return cache->all_verts;
}
Cleanup: move draw-cache creation from BKE to DRW Creating draw-cache should only ever be used by the draw-manager.
2017-04-21 21:14:11 +10:00
Batch *DRW_mesh_batch_cache_get_fancy_edges(Mesh *me)
Convert MBC_ API to Mesh (instead of derived mesh) and move it to mesh_render This includes a few fixes in the MBC_ api. The idea here is for this to be the only interface the render engines will deal with for the meshes. If we need to expose special options for sculpting engine we refactor this accordingly. But for now we are shaping this in a per-case base. Note: * We still need to hook up to the depsgraph to force clear/update of batch_cache when mesh changes (I'm waiting for Sergey Sharybin's depsgraph update for this though) * Also ideally we could/should use BMesh directly instead of DerivedMesh, but this will do for now. Note 2: In the end I renamed the `BKE_mesh_render` functions to `static mesh_render`. We can re-expose them as BKE_* later once we need it. Reviewers: merwin Subscribers: fclem Differential Revision: https://developer.blender.org/D2476
2017-01-25 09:16:29 +01:00
{
MeshBatchCache *cache = mesh_batch_cache_get(me);
if (cache->fancy_edges == NULL) {
/* create batch from DM */
static VertexFormat format = { 0 };
Rework of BKE's mesh_render to support BMesh directly. Note that since there is no (efficient) ways to get arrays of MVert/MEdge/etc. out of a BMesh, I refactored quite heavily internals of BKE_mesh_render. Now, when you do need acess to mesh data to generate cached batches, you create an abstract struct from mesh (either Mesh or BMesh if available), and then use advanced helpers to extract needed data, on a per-item basis (no more handling of arrays of verts/edges/... in batches code). This allows to: * Avoid having to create arrays of BMesh elements. * Take advantage of existing advanced BMesh topology and connectivity data. Reviewers: dfelinto, fclem, merwin Differential Revision: https://developer.blender.org/D2521
2017-02-23 11:54:40 +01:00
static unsigned int pos_id, n1_id, n2_id;
Convert MBC_ API to Mesh (instead of derived mesh) and move it to mesh_render This includes a few fixes in the MBC_ api. The idea here is for this to be the only interface the render engines will deal with for the meshes. If we need to expose special options for sculpting engine we refactor this accordingly. But for now we are shaping this in a per-case base. Note: * We still need to hook up to the depsgraph to force clear/update of batch_cache when mesh changes (I'm waiting for Sergey Sharybin's depsgraph update for this though) * Also ideally we could/should use BMesh directly instead of DerivedMesh, but this will do for now. Note 2: In the end I renamed the `BKE_mesh_render` functions to `static mesh_render`. We can re-expose them as BKE_* later once we need it. Reviewers: merwin Subscribers: fclem Differential Revision: https://developer.blender.org/D2476
2017-01-25 09:16:29 +01:00
if (format.attrib_ct == 0) {
/* initialize vertex format */
Gawain: VertexFormat_add_attrib (function name change) See intern/gawain for the API change. Other files are updated to use the new name. Also updated every call site to the recommended style: unsigned int foo = VertexFormat_add_attrib(format, "foo", COMP_ ... )
2017-04-06 00:59:04 -04:00
pos_id = VertexFormat_add_attrib(&format, "pos", COMP_F32, 3, KEEP_FLOAT);
Convert MBC_ API to Mesh (instead of derived mesh) and move it to mesh_render This includes a few fixes in the MBC_ api. The idea here is for this to be the only interface the render engines will deal with for the meshes. If we need to expose special options for sculpting engine we refactor this accordingly. But for now we are shaping this in a per-case base. Note: * We still need to hook up to the depsgraph to force clear/update of batch_cache when mesh changes (I'm waiting for Sergey Sharybin's depsgraph update for this though) * Also ideally we could/should use BMesh directly instead of DerivedMesh, but this will do for now. Note 2: In the end I renamed the `BKE_mesh_render` functions to `static mesh_render`. We can re-expose them as BKE_* later once we need it. Reviewers: merwin Subscribers: fclem Differential Revision: https://developer.blender.org/D2476
2017-01-25 09:16:29 +01:00
#if USE_10_10_10 /* takes 1/3 the space */
Gawain: VertexFormat_add_attrib (function name change) See intern/gawain for the API change. Other files are updated to use the new name. Also updated every call site to the recommended style: unsigned int foo = VertexFormat_add_attrib(format, "foo", COMP_ ... )
2017-04-06 00:59:04 -04:00
n1_id = VertexFormat_add_attrib(&format, "N1", COMP_I10, 3, NORMALIZE_INT_TO_FLOAT);
n2_id = VertexFormat_add_attrib(&format, "N2", COMP_I10, 3, NORMALIZE_INT_TO_FLOAT);
Convert MBC_ API to Mesh (instead of derived mesh) and move it to mesh_render This includes a few fixes in the MBC_ api. The idea here is for this to be the only interface the render engines will deal with for the meshes. If we need to expose special options for sculpting engine we refactor this accordingly. But for now we are shaping this in a per-case base. Note: * We still need to hook up to the depsgraph to force clear/update of batch_cache when mesh changes (I'm waiting for Sergey Sharybin's depsgraph update for this though) * Also ideally we could/should use BMesh directly instead of DerivedMesh, but this will do for now. Note 2: In the end I renamed the `BKE_mesh_render` functions to `static mesh_render`. We can re-expose them as BKE_* later once we need it. Reviewers: merwin Subscribers: fclem Differential Revision: https://developer.blender.org/D2476
2017-01-25 09:16:29 +01:00
#else
Gawain: VertexFormat_add_attrib (function name change) See intern/gawain for the API change. Other files are updated to use the new name. Also updated every call site to the recommended style: unsigned int foo = VertexFormat_add_attrib(format, "foo", COMP_ ... )
2017-04-06 00:59:04 -04:00
n1_id = VertexFormat_add_attrib(&format, "N1", COMP_F32, 3, KEEP_FLOAT);
n2_id = VertexFormat_add_attrib(&format, "N2", COMP_F32, 3, KEEP_FLOAT);
Convert MBC_ API to Mesh (instead of derived mesh) and move it to mesh_render This includes a few fixes in the MBC_ api. The idea here is for this to be the only interface the render engines will deal with for the meshes. If we need to expose special options for sculpting engine we refactor this accordingly. But for now we are shaping this in a per-case base. Note: * We still need to hook up to the depsgraph to force clear/update of batch_cache when mesh changes (I'm waiting for Sergey Sharybin's depsgraph update for this though) * Also ideally we could/should use BMesh directly instead of DerivedMesh, but this will do for now. Note 2: In the end I renamed the `BKE_mesh_render` functions to `static mesh_render`. We can re-expose them as BKE_* later once we need it. Reviewers: merwin Subscribers: fclem Differential Revision: https://developer.blender.org/D2476
2017-01-25 09:16:29 +01:00
#endif
}
VertexBuffer *vbo = VertexBuffer_create_with_format(&format);
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
MeshRenderData *rdata = mesh_render_data_create(me, MR_DATATYPE_VERT | MR_DATATYPE_EDGE | MR_DATATYPE_LOOP | MR_DATATYPE_POLY);
Convert MBC_ API to Mesh (instead of derived mesh) and move it to mesh_render This includes a few fixes in the MBC_ api. The idea here is for this to be the only interface the render engines will deal with for the meshes. If we need to expose special options for sculpting engine we refactor this accordingly. But for now we are shaping this in a per-case base. Note: * We still need to hook up to the depsgraph to force clear/update of batch_cache when mesh changes (I'm waiting for Sergey Sharybin's depsgraph update for this though) * Also ideally we could/should use BMesh directly instead of DerivedMesh, but this will do for now. Note 2: In the end I renamed the `BKE_mesh_render` functions to `static mesh_render`. We can re-expose them as BKE_* later once we need it. Reviewers: merwin Subscribers: fclem Differential Revision: https://developer.blender.org/D2476
2017-01-25 09:16:29 +01:00
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
const int edge_len = mesh_render_data_edges_len_get(rdata);
Convert MBC_ API to Mesh (instead of derived mesh) and move it to mesh_render This includes a few fixes in the MBC_ api. The idea here is for this to be the only interface the render engines will deal with for the meshes. If we need to expose special options for sculpting engine we refactor this accordingly. But for now we are shaping this in a per-case base. Note: * We still need to hook up to the depsgraph to force clear/update of batch_cache when mesh changes (I'm waiting for Sergey Sharybin's depsgraph update for this though) * Also ideally we could/should use BMesh directly instead of DerivedMesh, but this will do for now. Note 2: In the end I renamed the `BKE_mesh_render` functions to `static mesh_render`. We can re-expose them as BKE_* later once we need it. Reviewers: merwin Subscribers: fclem Differential Revision: https://developer.blender.org/D2476
2017-01-25 09:16:29 +01:00
Cleanup: replace num/ct/tot with 'len'
2017-04-19 00:10:32 +10:00
const int vbo_len_capacity = edge_len * 2; /* these are PRIM_LINE verts, not mesh verts */
int vbo_len_used = 0;
VertexBuffer_allocate_data(vbo, vbo_len_capacity);
for (int i = 0; i < edge_len; ++i) {
Rework of BKE's mesh_render to support BMesh directly. Note that since there is no (efficient) ways to get arrays of MVert/MEdge/etc. out of a BMesh, I refactored quite heavily internals of BKE_mesh_render. Now, when you do need acess to mesh data to generate cached batches, you create an abstract struct from mesh (either Mesh or BMesh if available), and then use advanced helpers to extract needed data, on a per-item basis (no more handling of arrays of verts/edges/... in batches code). This allows to: * Avoid having to create arrays of BMesh elements. * Take advantage of existing advanced BMesh topology and connectivity data. Reviewers: dfelinto, fclem, merwin Differential Revision: https://developer.blender.org/D2521
2017-02-23 11:54:40 +01:00
float *vcos1, *vcos2;
float *pnor1 = NULL, *pnor2 = NULL;
Draw Manager: T51218 Support edit-mesh hide
2017-04-18 23:29:55 +10:00
bool is_manifold;
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
if (mesh_render_data_edge_vcos_manifold_pnors(rdata, i, &vcos1, &vcos2, &pnor1, &pnor2, &is_manifold)) {
Convert MBC_ API to Mesh (instead of derived mesh) and move it to mesh_render This includes a few fixes in the MBC_ api. The idea here is for this to be the only interface the render engines will deal with for the meshes. If we need to expose special options for sculpting engine we refactor this accordingly. But for now we are shaping this in a per-case base. Note: * We still need to hook up to the depsgraph to force clear/update of batch_cache when mesh changes (I'm waiting for Sergey Sharybin's depsgraph update for this though) * Also ideally we could/should use BMesh directly instead of DerivedMesh, but this will do for now. Note 2: In the end I renamed the `BKE_mesh_render` functions to `static mesh_render`. We can re-expose them as BKE_* later once we need it. Reviewers: merwin Subscribers: fclem Differential Revision: https://developer.blender.org/D2476
2017-01-25 09:16:29 +01:00
#if USE_10_10_10
Draw Manager: T51218 Support edit-mesh hide
2017-04-18 23:29:55 +10:00
PackedNormal n1value = { .x = 0, .y = 0, .z = +511 };
PackedNormal n2value = { .x = 0, .y = 0, .z = -511 };
Convert MBC_ API to Mesh (instead of derived mesh) and move it to mesh_render This includes a few fixes in the MBC_ api. The idea here is for this to be the only interface the render engines will deal with for the meshes. If we need to expose special options for sculpting engine we refactor this accordingly. But for now we are shaping this in a per-case base. Note: * We still need to hook up to the depsgraph to force clear/update of batch_cache when mesh changes (I'm waiting for Sergey Sharybin's depsgraph update for this though) * Also ideally we could/should use BMesh directly instead of DerivedMesh, but this will do for now. Note 2: In the end I renamed the `BKE_mesh_render` functions to `static mesh_render`. We can re-expose them as BKE_* later once we need it. Reviewers: merwin Subscribers: fclem Differential Revision: https://developer.blender.org/D2476
2017-01-25 09:16:29 +01:00
Draw Manager: T51218 Support edit-mesh hide
2017-04-18 23:29:55 +10:00
if (is_manifold) {
n1value = convert_i10_v3(pnor1);
n2value = convert_i10_v3(pnor2);
}
Convert MBC_ API to Mesh (instead of derived mesh) and move it to mesh_render This includes a few fixes in the MBC_ api. The idea here is for this to be the only interface the render engines will deal with for the meshes. If we need to expose special options for sculpting engine we refactor this accordingly. But for now we are shaping this in a per-case base. Note: * We still need to hook up to the depsgraph to force clear/update of batch_cache when mesh changes (I'm waiting for Sergey Sharybin's depsgraph update for this though) * Also ideally we could/should use BMesh directly instead of DerivedMesh, but this will do for now. Note 2: In the end I renamed the `BKE_mesh_render` functions to `static mesh_render`. We can re-expose them as BKE_* later once we need it. Reviewers: merwin Subscribers: fclem Differential Revision: https://developer.blender.org/D2476
2017-01-25 09:16:29 +01:00
Draw Manager: T51218 Support edit-mesh hide
2017-04-18 23:29:55 +10:00
const PackedNormal *n1 = &n1value;
const PackedNormal *n2 = &n2value;
Convert MBC_ API to Mesh (instead of derived mesh) and move it to mesh_render This includes a few fixes in the MBC_ api. The idea here is for this to be the only interface the render engines will deal with for the meshes. If we need to expose special options for sculpting engine we refactor this accordingly. But for now we are shaping this in a per-case base. Note: * We still need to hook up to the depsgraph to force clear/update of batch_cache when mesh changes (I'm waiting for Sergey Sharybin's depsgraph update for this though) * Also ideally we could/should use BMesh directly instead of DerivedMesh, but this will do for now. Note 2: In the end I renamed the `BKE_mesh_render` functions to `static mesh_render`. We can re-expose them as BKE_* later once we need it. Reviewers: merwin Subscribers: fclem Differential Revision: https://developer.blender.org/D2476
2017-01-25 09:16:29 +01:00
#else
Draw Manager: T51218 Support edit-mesh hide
2017-04-18 23:29:55 +10:00
const float dummy1[3] = { 0.0f, 0.0f, +1.0f };
const float dummy2[3] = { 0.0f, 0.0f, -1.0f };
Convert MBC_ API to Mesh (instead of derived mesh) and move it to mesh_render This includes a few fixes in the MBC_ api. The idea here is for this to be the only interface the render engines will deal with for the meshes. If we need to expose special options for sculpting engine we refactor this accordingly. But for now we are shaping this in a per-case base. Note: * We still need to hook up to the depsgraph to force clear/update of batch_cache when mesh changes (I'm waiting for Sergey Sharybin's depsgraph update for this though) * Also ideally we could/should use BMesh directly instead of DerivedMesh, but this will do for now. Note 2: In the end I renamed the `BKE_mesh_render` functions to `static mesh_render`. We can re-expose them as BKE_* later once we need it. Reviewers: merwin Subscribers: fclem Differential Revision: https://developer.blender.org/D2476
2017-01-25 09:16:29 +01:00
Draw Manager: T51218 Support edit-mesh hide
2017-04-18 23:29:55 +10:00
const float *n1 = (is_manifold) ? pnor1 : dummy1;
const float *n2 = (is_manifold) ? pnor2 : dummy2;
Convert MBC_ API to Mesh (instead of derived mesh) and move it to mesh_render This includes a few fixes in the MBC_ api. The idea here is for this to be the only interface the render engines will deal with for the meshes. If we need to expose special options for sculpting engine we refactor this accordingly. But for now we are shaping this in a per-case base. Note: * We still need to hook up to the depsgraph to force clear/update of batch_cache when mesh changes (I'm waiting for Sergey Sharybin's depsgraph update for this though) * Also ideally we could/should use BMesh directly instead of DerivedMesh, but this will do for now. Note 2: In the end I renamed the `BKE_mesh_render` functions to `static mesh_render`. We can re-expose them as BKE_* later once we need it. Reviewers: merwin Subscribers: fclem Differential Revision: https://developer.blender.org/D2476
2017-01-25 09:16:29 +01:00
#endif
Draw Manager: T51218 Support edit-mesh hide
2017-04-18 23:29:55 +10:00
VertexBuffer_set_attrib(vbo, pos_id, 2 * i, vcos1);
VertexBuffer_set_attrib(vbo, n1_id, 2 * i, n1);
VertexBuffer_set_attrib(vbo, n2_id, 2 * i, n2);
VertexBuffer_set_attrib(vbo, pos_id, 2 * i + 1, vcos2);
VertexBuffer_set_attrib(vbo, n1_id, 2 * i + 1, n1);
VertexBuffer_set_attrib(vbo, n2_id, 2 * i + 1, n2);
Convert MBC_ API to Mesh (instead of derived mesh) and move it to mesh_render This includes a few fixes in the MBC_ api. The idea here is for this to be the only interface the render engines will deal with for the meshes. If we need to expose special options for sculpting engine we refactor this accordingly. But for now we are shaping this in a per-case base. Note: * We still need to hook up to the depsgraph to force clear/update of batch_cache when mesh changes (I'm waiting for Sergey Sharybin's depsgraph update for this though) * Also ideally we could/should use BMesh directly instead of DerivedMesh, but this will do for now. Note 2: In the end I renamed the `BKE_mesh_render` functions to `static mesh_render`. We can re-expose them as BKE_* later once we need it. Reviewers: merwin Subscribers: fclem Differential Revision: https://developer.blender.org/D2476
2017-01-25 09:16:29 +01:00
Cleanup: replace num/ct/tot with 'len'
2017-04-19 00:10:32 +10:00
vbo_len_used += 2;
Draw Manager: T51218 Support edit-mesh hide
2017-04-18 23:29:55 +10:00
}
}
Cleanup: replace num/ct/tot with 'len'
2017-04-19 00:10:32 +10:00
if (vbo_len_used != vbo_len_capacity) {
VertexBuffer_resize_data(vbo, vbo_len_used);
Convert MBC_ API to Mesh (instead of derived mesh) and move it to mesh_render This includes a few fixes in the MBC_ api. The idea here is for this to be the only interface the render engines will deal with for the meshes. If we need to expose special options for sculpting engine we refactor this accordingly. But for now we are shaping this in a per-case base. Note: * We still need to hook up to the depsgraph to force clear/update of batch_cache when mesh changes (I'm waiting for Sergey Sharybin's depsgraph update for this though) * Also ideally we could/should use BMesh directly instead of DerivedMesh, but this will do for now. Note 2: In the end I renamed the `BKE_mesh_render` functions to `static mesh_render`. We can re-expose them as BKE_* later once we need it. Reviewers: merwin Subscribers: fclem Differential Revision: https://developer.blender.org/D2476
2017-01-25 09:16:29 +01:00
}
OpenGL: use PRIM instead of GL enum everywhere else Well, everywhere that uses Gawain for drawing. Places that call OpenGL directly still use GL enums. Part of T49043
2017-04-08 00:50:36 -04:00
cache->fancy_edges = Batch_create(PRIM_LINES, vbo, NULL);
Convert MBC_ API to Mesh (instead of derived mesh) and move it to mesh_render This includes a few fixes in the MBC_ api. The idea here is for this to be the only interface the render engines will deal with for the meshes. If we need to expose special options for sculpting engine we refactor this accordingly. But for now we are shaping this in a per-case base. Note: * We still need to hook up to the depsgraph to force clear/update of batch_cache when mesh changes (I'm waiting for Sergey Sharybin's depsgraph update for this though) * Also ideally we could/should use BMesh directly instead of DerivedMesh, but this will do for now. Note 2: In the end I renamed the `BKE_mesh_render` functions to `static mesh_render`. We can re-expose them as BKE_* later once we need it. Reviewers: merwin Subscribers: fclem Differential Revision: https://developer.blender.org/D2476
2017-01-25 09:16:29 +01:00
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
mesh_render_data_free(rdata);
Convert MBC_ API to Mesh (instead of derived mesh) and move it to mesh_render This includes a few fixes in the MBC_ api. The idea here is for this to be the only interface the render engines will deal with for the meshes. If we need to expose special options for sculpting engine we refactor this accordingly. But for now we are shaping this in a per-case base. Note: * We still need to hook up to the depsgraph to force clear/update of batch_cache when mesh changes (I'm waiting for Sergey Sharybin's depsgraph update for this though) * Also ideally we could/should use BMesh directly instead of DerivedMesh, but this will do for now. Note 2: In the end I renamed the `BKE_mesh_render` functions to `static mesh_render`. We can re-expose them as BKE_* later once we need it. Reviewers: merwin Subscribers: fclem Differential Revision: https://developer.blender.org/D2476
2017-01-25 09:16:29 +01:00
}
Rework of BKE's mesh_render to support BMesh directly. Note that since there is no (efficient) ways to get arrays of MVert/MEdge/etc. out of a BMesh, I refactored quite heavily internals of BKE_mesh_render. Now, when you do need acess to mesh data to generate cached batches, you create an abstract struct from mesh (either Mesh or BMesh if available), and then use advanced helpers to extract needed data, on a per-item basis (no more handling of arrays of verts/edges/... in batches code). This allows to: * Avoid having to create arrays of BMesh elements. * Take advantage of existing advanced BMesh topology and connectivity data. Reviewers: dfelinto, fclem, merwin Differential Revision: https://developer.blender.org/D2521
2017-02-23 11:54:40 +01:00
return cache->fancy_edges;
Convert MBC_ API to Mesh (instead of derived mesh) and move it to mesh_render This includes a few fixes in the MBC_ api. The idea here is for this to be the only interface the render engines will deal with for the meshes. If we need to expose special options for sculpting engine we refactor this accordingly. But for now we are shaping this in a per-case base. Note: * We still need to hook up to the depsgraph to force clear/update of batch_cache when mesh changes (I'm waiting for Sergey Sharybin's depsgraph update for this though) * Also ideally we could/should use BMesh directly instead of DerivedMesh, but this will do for now. Note 2: In the end I renamed the `BKE_mesh_render` functions to `static mesh_render`. We can re-expose them as BKE_* later once we need it. Reviewers: merwin Subscribers: fclem Differential Revision: https://developer.blender.org/D2476
2017-01-25 09:16:29 +01:00
}
Edit Mode overlays: separate multiple shaders for loose edges and verts
2017-03-01 14:07:04 +01:00
static void mesh_batch_cache_create_overlay_batches(Mesh *me)
Clay Engine: Edit mesh overlays Based on the previous overlay shader from merwin. This shader takes care of clipped vertex cases and do all edit mode face info in one pass (except face centers). As the shading is done one the triangle itself the visual can't go beyond the surface of the mesh. That leads to half displayed edges on the outline of the mesh. This problem can be fixed by a second pass. This is work in progress.
2017-02-26 21:07:37 +01:00
{
Edit Mode overlays: separate multiple shaders for loose edges and verts
2017-03-01 14:07:04 +01:00
/* Since MR_DATATYPE_OVERLAY is slow to generate, generate them all at once */
int options = MR_DATATYPE_VERT | MR_DATATYPE_EDGE | MR_DATATYPE_LOOPTRI | MR_DATATYPE_OVERLAY;
Clay Engine: Edit mesh overlays Based on the previous overlay shader from merwin. This shader takes care of clipped vertex cases and do all edit mode face info in one pass (except face centers). As the shading is done one the triangle itself the visual can't go beyond the surface of the mesh. That leads to half displayed edges on the outline of the mesh. This problem can be fixed by a second pass. This is work in progress.
2017-02-26 21:07:37 +01:00
Edit Mode overlays: separate multiple shaders for loose edges and verts
2017-03-01 14:07:04 +01:00
MeshBatchCache *cache = mesh_batch_cache_get(me);
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
MeshRenderData *rdata = mesh_render_data_create(me, options);
Edit Mode overlays: separate multiple shaders for loose edges and verts
2017-03-01 14:07:04 +01:00
static VertexFormat format = { 0 };
static unsigned pos_id, data_id;
if (format.attrib_ct == 0) {
/* initialize vertex format */
Gawain: VertexFormat_add_attrib (function name change) See intern/gawain for the API change. Other files are updated to use the new name. Also updated every call site to the recommended style: unsigned int foo = VertexFormat_add_attrib(format, "foo", COMP_ ... )
2017-04-06 00:59:04 -04:00
pos_id = VertexFormat_add_attrib(&format, "pos", COMP_F32, 3, KEEP_FLOAT);
data_id = VertexFormat_add_attrib(&format, "data", COMP_U8, 4, KEEP_INT);
Edit Mode overlays: separate multiple shaders for loose edges and verts
2017-03-01 14:07:04 +01:00
}
Clay Engine: Edit mesh overlays Based on the previous overlay shader from merwin. This shader takes care of clipped vertex cases and do all edit mode face info in one pass (except face centers). As the shading is done one the triangle itself the visual can't go beyond the surface of the mesh. That leads to half displayed edges on the outline of the mesh. This problem can be fixed by a second pass. This is work in progress.
2017-02-26 21:07:37 +01:00
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
const int tri_len = mesh_render_data_looptri_len_get(rdata);
const int ledge_len = mesh_render_data_loose_edges_len_get(rdata);
const int lvert_len = mesh_render_data_loose_verts_len_get(rdata);
Edit Mesh overlays: added looses edges and verts as triangle. We loose a bit of memory for (more?) performance.
2017-02-28 14:30:34 +01:00
Edit Mode overlays: separate multiple shaders for loose edges and verts
2017-03-01 14:07:04 +01:00
if (cache->overlay_triangles == NULL) {
VertexBuffer *vbo = VertexBuffer_create_with_format(&format);
Cleanup: replace num/ct/tot with 'len'
2017-04-19 00:10:32 +10:00
const int vbo_len_capacity = tri_len * 3;
int vbo_len_used = 0;
VertexBuffer_allocate_data(vbo, vbo_len_capacity);
Edit Mesh overlays: added looses edges and verts as triangle. We loose a bit of memory for (more?) performance.
2017-02-28 14:30:34 +01:00
Cleanup: replace num/ct/tot with 'len'
2017-04-19 00:10:32 +10:00
for (int i = 0; i < tri_len; ++i) {
EditMode: remove edge-hash from MeshRenderData LoopTri's can access this data.
2017-04-12 13:42:27 +10:00
int tri_vert_idx[3], tri_edge_idx[3];
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
if (mesh_render_data_looptri_vert_edge_indices_get(rdata, i, tri_vert_idx, tri_edge_idx)) {
Draw Manager: T51218 Support edit-mesh hide
2017-04-18 23:29:55 +10:00
add_overlay_tri(
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
rdata, vbo, pos_id, data_id,
Cleanup: replace num/ct/tot with 'len'
2017-04-19 00:10:32 +10:00
tri_vert_idx, tri_edge_idx, i, vbo_len_used);
vbo_len_used += 3;
Draw Manager: T51218 Support edit-mesh hide
2017-04-18 23:29:55 +10:00
}
}
Cleanup: replace num/ct/tot with 'len'
2017-04-19 00:10:32 +10:00
if (vbo_len_used != vbo_len_capacity) {
VertexBuffer_resize_data(vbo, vbo_len_used);
Edit Mesh overlays: added looses edges and verts as triangle. We loose a bit of memory for (more?) performance.
2017-02-28 14:30:34 +01:00
}
OpenGL: use PRIM instead of GL enum everywhere else Well, everywhere that uses Gawain for drawing. Places that call OpenGL directly still use GL enums. Part of T49043
2017-04-08 00:50:36 -04:00
cache->overlay_triangles = Batch_create(PRIM_TRIANGLES, vbo, NULL);
Edit Mode overlays: separate multiple shaders for loose edges and verts
2017-03-01 14:07:04 +01:00
}
Clay Engine: Edit mesh overlays Based on the previous overlay shader from merwin. This shader takes care of clipped vertex cases and do all edit mode face info in one pass (except face centers). As the shading is done one the triangle itself the visual can't go beyond the surface of the mesh. That leads to half displayed edges on the outline of the mesh. This problem can be fixed by a second pass. This is work in progress.
2017-02-26 21:07:37 +01:00
Edit Mode overlays: separate multiple shaders for loose edges and verts
2017-03-01 14:07:04 +01:00
if (cache->overlay_loose_edges == NULL) {
VertexBuffer *vbo = VertexBuffer_create_with_format(&format);
Cleanup: replace num/ct/tot with 'len'
2017-04-19 00:10:32 +10:00
const int vbo_len_capacity = ledge_len * 2;
int vbo_len_used = 0;
VertexBuffer_allocate_data(vbo, vbo_len_capacity);
Clay Engine: Edit mesh overlays Based on the previous overlay shader from merwin. This shader takes care of clipped vertex cases and do all edit mode face info in one pass (except face centers). As the shading is done one the triangle itself the visual can't go beyond the surface of the mesh. That leads to half displayed edges on the outline of the mesh. This problem can be fixed by a second pass. This is work in progress.
2017-02-26 21:07:37 +01:00
Cleanup: replace num/ct/tot with 'len'
2017-04-19 00:10:32 +10:00
for (int i = 0; i < ledge_len; ++i) {
Edit Mode overlays: separate multiple shaders for loose edges and verts
2017-03-01 14:07:04 +01:00
int vert_idx[2];
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
bool ok = mesh_render_data_edge_verts_indices_get(rdata, rdata->loose_edges[i], vert_idx);
Cleanup: replace num/ct/tot with 'len'
2017-04-19 00:10:32 +10:00
assert(ok); /* we don't add */
add_overlay_loose_edge(
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
rdata, vbo, pos_id, data_id,
vert_idx[0], vert_idx[1], rdata->loose_edges[i], vbo_len_used);
Cleanup: replace num/ct/tot with 'len'
2017-04-19 00:10:32 +10:00
vbo_len_used += 2;
Clay Engine: Edit mesh overlays Based on the previous overlay shader from merwin. This shader takes care of clipped vertex cases and do all edit mode face info in one pass (except face centers). As the shading is done one the triangle itself the visual can't go beyond the surface of the mesh. That leads to half displayed edges on the outline of the mesh. This problem can be fixed by a second pass. This is work in progress.
2017-02-26 21:07:37 +01:00
}
Cleanup: replace num/ct/tot with 'len'
2017-04-19 00:10:32 +10:00
BLI_assert(vbo_len_used == vbo_len_capacity);
OpenGL: use PRIM instead of GL enum everywhere else Well, everywhere that uses Gawain for drawing. Places that call OpenGL directly still use GL enums. Part of T49043
2017-04-08 00:50:36 -04:00
cache->overlay_loose_edges = Batch_create(PRIM_LINES, vbo, NULL);
Clay Engine: Edit mesh overlays Based on the previous overlay shader from merwin. This shader takes care of clipped vertex cases and do all edit mode face info in one pass (except face centers). As the shading is done one the triangle itself the visual can't go beyond the surface of the mesh. That leads to half displayed edges on the outline of the mesh. This problem can be fixed by a second pass. This is work in progress.
2017-02-26 21:07:37 +01:00
}
Edit Mode overlays: separate multiple shaders for loose edges and verts
2017-03-01 14:07:04 +01:00
if (cache->overlay_loose_verts == NULL) {
VertexBuffer *vbo = VertexBuffer_create_with_format(&format);
Cleanup: replace num/ct/tot with 'len'
2017-04-19 00:10:32 +10:00
const int vbo_len_capacity = lvert_len;
int vbo_len_used = 0;
VertexBuffer_allocate_data(vbo, vbo_len_capacity);
Edit Mode overlays: separate multiple shaders for loose edges and verts
2017-03-01 14:07:04 +01:00
Cleanup: replace num/ct/tot with 'len'
2017-04-19 00:10:32 +10:00
for (int i = 0; i < lvert_len; ++i) {
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
add_overlay_loose_vert(rdata, vbo, pos_id, data_id,
rdata->loose_verts[i], vbo_len_used);
Cleanup: replace num/ct/tot with 'len'
2017-04-19 00:10:32 +10:00
vbo_len_used += 1;
Clay Engine: Edit mesh overlays Based on the previous overlay shader from merwin. This shader takes care of clipped vertex cases and do all edit mode face info in one pass (except face centers). As the shading is done one the triangle itself the visual can't go beyond the surface of the mesh. That leads to half displayed edges on the outline of the mesh. This problem can be fixed by a second pass. This is work in progress.
2017-02-26 21:07:37 +01:00
}
Cleanup: replace num/ct/tot with 'len'
2017-04-19 00:10:32 +10:00
BLI_assert(vbo_len_used == vbo_len_capacity);
OpenGL: use PRIM instead of GL enum everywhere else Well, everywhere that uses Gawain for drawing. Places that call OpenGL directly still use GL enums. Part of T49043
2017-04-08 00:50:36 -04:00
cache->overlay_loose_verts = Batch_create(PRIM_POINTS, vbo, NULL);
Clay Engine: Edit mesh overlays Based on the previous overlay shader from merwin. This shader takes care of clipped vertex cases and do all edit mode face info in one pass (except face centers). As the shading is done one the triangle itself the visual can't go beyond the surface of the mesh. That leads to half displayed edges on the outline of the mesh. This problem can be fixed by a second pass. This is work in progress.
2017-02-26 21:07:37 +01:00
}
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
mesh_render_data_free(rdata);
Clay Engine: Edit mesh overlays Based on the previous overlay shader from merwin. This shader takes care of clipped vertex cases and do all edit mode face info in one pass (except face centers). As the shading is done one the triangle itself the visual can't go beyond the surface of the mesh. That leads to half displayed edges on the outline of the mesh. This problem can be fixed by a second pass. This is work in progress.
2017-02-26 21:07:37 +01:00
}
Cleanup: move draw-cache creation from BKE to DRW Creating draw-cache should only ever be used by the draw-manager.
2017-04-21 21:14:11 +10:00
Batch *DRW_mesh_batch_cache_get_overlay_triangles(Mesh *me)
Convert MBC_ API to Mesh (instead of derived mesh) and move it to mesh_render This includes a few fixes in the MBC_ api. The idea here is for this to be the only interface the render engines will deal with for the meshes. If we need to expose special options for sculpting engine we refactor this accordingly. But for now we are shaping this in a per-case base. Note: * We still need to hook up to the depsgraph to force clear/update of batch_cache when mesh changes (I'm waiting for Sergey Sharybin's depsgraph update for this though) * Also ideally we could/should use BMesh directly instead of DerivedMesh, but this will do for now. Note 2: In the end I renamed the `BKE_mesh_render` functions to `static mesh_render`. We can re-expose them as BKE_* later once we need it. Reviewers: merwin Subscribers: fclem Differential Revision: https://developer.blender.org/D2476
2017-01-25 09:16:29 +01:00
{
Edit Mode overlays: separate multiple shaders for loose edges and verts
2017-03-01 14:07:04 +01:00
MeshBatchCache *cache = mesh_batch_cache_get(me);
Convert MBC_ API to Mesh (instead of derived mesh) and move it to mesh_render This includes a few fixes in the MBC_ api. The idea here is for this to be the only interface the render engines will deal with for the meshes. If we need to expose special options for sculpting engine we refactor this accordingly. But for now we are shaping this in a per-case base. Note: * We still need to hook up to the depsgraph to force clear/update of batch_cache when mesh changes (I'm waiting for Sergey Sharybin's depsgraph update for this though) * Also ideally we could/should use BMesh directly instead of DerivedMesh, but this will do for now. Note 2: In the end I renamed the `BKE_mesh_render` functions to `static mesh_render`. We can re-expose them as BKE_* later once we need it. Reviewers: merwin Subscribers: fclem Differential Revision: https://developer.blender.org/D2476
2017-01-25 09:16:29 +01:00
Edit Mode overlays: separate multiple shaders for loose edges and verts
2017-03-01 14:07:04 +01:00
if (cache->overlay_triangles == NULL) {
mesh_batch_cache_create_overlay_batches(me);
}
Convert MBC_ API to Mesh (instead of derived mesh) and move it to mesh_render This includes a few fixes in the MBC_ api. The idea here is for this to be the only interface the render engines will deal with for the meshes. If we need to expose special options for sculpting engine we refactor this accordingly. But for now we are shaping this in a per-case base. Note: * We still need to hook up to the depsgraph to force clear/update of batch_cache when mesh changes (I'm waiting for Sergey Sharybin's depsgraph update for this though) * Also ideally we could/should use BMesh directly instead of DerivedMesh, but this will do for now. Note 2: In the end I renamed the `BKE_mesh_render` functions to `static mesh_render`. We can re-expose them as BKE_* later once we need it. Reviewers: merwin Subscribers: fclem Differential Revision: https://developer.blender.org/D2476
2017-01-25 09:16:29 +01:00
Edit Mode overlays: separate multiple shaders for loose edges and verts
2017-03-01 14:07:04 +01:00
return cache->overlay_triangles;
Edit Mesh overlays: added looses edges and verts as triangle. We loose a bit of memory for (more?) performance.
2017-02-28 14:30:34 +01:00
}
Cleanup: move draw-cache creation from BKE to DRW Creating draw-cache should only ever be used by the draw-manager.
2017-04-21 21:14:11 +10:00
Batch *DRW_mesh_batch_cache_get_overlay_loose_edges(Mesh *me)
Edit Mesh overlays: added looses edges and verts as triangle. We loose a bit of memory for (more?) performance.
2017-02-28 14:30:34 +01:00
{
Edit Mode overlays: separate multiple shaders for loose edges and verts
2017-03-01 14:07:04 +01:00
MeshBatchCache *cache = mesh_batch_cache_get(me);
Edit Mesh overlays: added looses edges and verts as triangle. We loose a bit of memory for (more?) performance.
2017-02-28 14:30:34 +01:00
Edit Mode overlays: separate multiple shaders for loose edges and verts
2017-03-01 14:07:04 +01:00
if (cache->overlay_loose_edges == NULL) {
mesh_batch_cache_create_overlay_batches(me);
}
Edit Mesh overlays: added looses edges and verts as triangle. We loose a bit of memory for (more?) performance.
2017-02-28 14:30:34 +01:00
Edit Mode overlays: separate multiple shaders for loose edges and verts
2017-03-01 14:07:04 +01:00
return cache->overlay_loose_edges;
Convert MBC_ API to Mesh (instead of derived mesh) and move it to mesh_render This includes a few fixes in the MBC_ api. The idea here is for this to be the only interface the render engines will deal with for the meshes. If we need to expose special options for sculpting engine we refactor this accordingly. But for now we are shaping this in a per-case base. Note: * We still need to hook up to the depsgraph to force clear/update of batch_cache when mesh changes (I'm waiting for Sergey Sharybin's depsgraph update for this though) * Also ideally we could/should use BMesh directly instead of DerivedMesh, but this will do for now. Note 2: In the end I renamed the `BKE_mesh_render` functions to `static mesh_render`. We can re-expose them as BKE_* later once we need it. Reviewers: merwin Subscribers: fclem Differential Revision: https://developer.blender.org/D2476
2017-01-25 09:16:29 +01:00
}
Cleanup: move draw-cache creation from BKE to DRW Creating draw-cache should only ever be used by the draw-manager.
2017-04-21 21:14:11 +10:00
Batch *DRW_mesh_batch_cache_get_overlay_loose_verts(Mesh *me)
Convert MBC_ API to Mesh (instead of derived mesh) and move it to mesh_render This includes a few fixes in the MBC_ api. The idea here is for this to be the only interface the render engines will deal with for the meshes. If we need to expose special options for sculpting engine we refactor this accordingly. But for now we are shaping this in a per-case base. Note: * We still need to hook up to the depsgraph to force clear/update of batch_cache when mesh changes (I'm waiting for Sergey Sharybin's depsgraph update for this though) * Also ideally we could/should use BMesh directly instead of DerivedMesh, but this will do for now. Note 2: In the end I renamed the `BKE_mesh_render` functions to `static mesh_render`. We can re-expose them as BKE_* later once we need it. Reviewers: merwin Subscribers: fclem Differential Revision: https://developer.blender.org/D2476
2017-01-25 09:16:29 +01:00
{
MeshBatchCache *cache = mesh_batch_cache_get(me);
Edit Mode overlays: separate multiple shaders for loose edges and verts
2017-03-01 14:07:04 +01:00
if (cache->overlay_loose_verts == NULL) {
mesh_batch_cache_create_overlay_batches(me);
Convert MBC_ API to Mesh (instead of derived mesh) and move it to mesh_render This includes a few fixes in the MBC_ api. The idea here is for this to be the only interface the render engines will deal with for the meshes. If we need to expose special options for sculpting engine we refactor this accordingly. But for now we are shaping this in a per-case base. Note: * We still need to hook up to the depsgraph to force clear/update of batch_cache when mesh changes (I'm waiting for Sergey Sharybin's depsgraph update for this though) * Also ideally we could/should use BMesh directly instead of DerivedMesh, but this will do for now. Note 2: In the end I renamed the `BKE_mesh_render` functions to `static mesh_render`. We can re-expose them as BKE_* later once we need it. Reviewers: merwin Subscribers: fclem Differential Revision: https://developer.blender.org/D2476
2017-01-25 09:16:29 +01:00
}
Edit Mode overlays: separate multiple shaders for loose edges and verts
2017-03-01 14:07:04 +01:00
return cache->overlay_loose_verts;
Convert MBC_ API to Mesh (instead of derived mesh) and move it to mesh_render This includes a few fixes in the MBC_ api. The idea here is for this to be the only interface the render engines will deal with for the meshes. If we need to expose special options for sculpting engine we refactor this accordingly. But for now we are shaping this in a per-case base. Note: * We still need to hook up to the depsgraph to force clear/update of batch_cache when mesh changes (I'm waiting for Sergey Sharybin's depsgraph update for this though) * Also ideally we could/should use BMesh directly instead of DerivedMesh, but this will do for now. Note 2: In the end I renamed the `BKE_mesh_render` functions to `static mesh_render`. We can re-expose them as BKE_* later once we need it. Reviewers: merwin Subscribers: fclem Differential Revision: https://developer.blender.org/D2476
2017-01-25 09:16:29 +01:00
}
Cleanup: move draw-cache creation from BKE to DRW Creating draw-cache should only ever be used by the draw-manager.
2017-04-21 21:14:11 +10:00
Batch *DRW_mesh_batch_cache_get_overlay_facedots(Mesh *me)
Edit Mode overlay: Moved Shaders to draw modules and resolved some draw issue. We don't want to clutter gpu_shader.c with engine specific code Added face's center dot Simplified loose vert shader
2017-03-02 01:07:03 +01:00
{
MeshBatchCache *cache = mesh_batch_cache_get(me);
if (cache->overlay_facedots == NULL) {
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
MeshRenderData *rdata = mesh_render_data_create(me, MR_DATATYPE_VERT | MR_DATATYPE_LOOP | MR_DATATYPE_POLY);
Edit Mode overlay: Moved Shaders to draw modules and resolved some draw issue. We don't want to clutter gpu_shader.c with engine specific code Added face's center dot Simplified loose vert shader
2017-03-02 01:07:03 +01:00
static VertexFormat format = { 0 };
static unsigned pos_id, data_id;
if (format.attrib_ct == 0) {
/* initialize vertex format */
Gawain: VertexFormat_add_attrib (function name change) See intern/gawain for the API change. Other files are updated to use the new name. Also updated every call site to the recommended style: unsigned int foo = VertexFormat_add_attrib(format, "foo", COMP_ ... )
2017-04-06 00:59:04 -04:00
pos_id = VertexFormat_add_attrib(&format, "pos", COMP_F32, 3, KEEP_FLOAT);
Edit Mesh overlay: pack normals with face centers. Needed for face normals.
2017-03-02 15:03:40 +01:00
#if USE_10_10_10
Gawain: VertexFormat_add_attrib (function name change) See intern/gawain for the API change. Other files are updated to use the new name. Also updated every call site to the recommended style: unsigned int foo = VertexFormat_add_attrib(format, "foo", COMP_ ... )
2017-04-06 00:59:04 -04:00
data_id = VertexFormat_add_attrib(&format, "norAndFlag", COMP_I10, 4, NORMALIZE_INT_TO_FLOAT);
Edit Mesh overlay: pack normals with face centers. Needed for face normals.
2017-03-02 15:03:40 +01:00
#else
Gawain: VertexFormat_add_attrib (function name change) See intern/gawain for the API change. Other files are updated to use the new name. Also updated every call site to the recommended style: unsigned int foo = VertexFormat_add_attrib(format, "foo", COMP_ ... )
2017-04-06 00:59:04 -04:00
data_id = VertexFormat_add_attrib(&format, "norAndFlag", COMP_F32, 4, KEEP_FLOAT);
Edit Mesh overlay: pack normals with face centers. Needed for face normals.
2017-03-02 15:03:40 +01:00
#endif
Edit Mode overlay: Moved Shaders to draw modules and resolved some draw issue. We don't want to clutter gpu_shader.c with engine specific code Added face's center dot Simplified loose vert shader
2017-03-02 01:07:03 +01:00
}
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
const int vbo_len_capacity = mesh_render_data_polys_len_get(rdata);
Cleanup: replace num/ct/tot with 'len'
2017-04-19 00:10:32 +10:00
int vbo_len_used = 0;
Edit Mode overlay: Moved Shaders to draw modules and resolved some draw issue. We don't want to clutter gpu_shader.c with engine specific code Added face's center dot Simplified loose vert shader
2017-03-02 01:07:03 +01:00
VertexBuffer *vbo = VertexBuffer_create_with_format(&format);
Cleanup: replace num/ct/tot with 'len'
2017-04-19 00:10:32 +10:00
VertexBuffer_allocate_data(vbo, vbo_len_capacity);
Edit Mode overlay: Moved Shaders to draw modules and resolved some draw issue. We don't want to clutter gpu_shader.c with engine specific code Added face's center dot Simplified loose vert shader
2017-03-02 01:07:03 +01:00
Cleanup: replace num/ct/tot with 'len'
2017-04-19 00:10:32 +10:00
for (int i = 0; i < vbo_len_capacity; ++i) {
Edit Mesh overlay: pack normals with face centers. Needed for face normals.
2017-03-02 15:03:40 +01:00
float pcenter[3], pnor[3];
Remove bool -> int cast Wont work as expected on big endian systems
2017-04-12 18:23:40 +10:00
bool selected = false;
Edit Mesh overlay: pack normals with face centers. Needed for face normals.
2017-03-02 15:03:40 +01:00
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
if (mesh_render_data_pnors_pcenter_select_get(rdata, i, pnor, pcenter, &selected)) {
Edit Mesh overlay: pack normals with face centers. Needed for face normals.
2017-03-02 15:03:40 +01:00
#if USE_10_10_10
Draw Manager: T51218 Support edit-mesh hide
2017-04-18 23:29:55 +10:00
PackedNormal nor = { .x = 0, .y = 0, .z = -511 };
nor = convert_i10_v3(pnor);
nor.w = selected ? 1 : 0;
VertexBuffer_set_attrib(vbo, data_id, i, &nor);
Edit Mesh overlay: pack normals with face centers. Needed for face normals.
2017-03-02 15:03:40 +01:00
#else
Draw Manager: T51218 Support edit-mesh hide
2017-04-18 23:29:55 +10:00
float nor[4] = {pnor[0], pnor[1], pnor[2], selected ? 1 : 0};
VertexBuffer_set_attrib(vbo, data_id, i, nor);
Edit Mesh overlay: pack normals with face centers. Needed for face normals.
2017-03-02 15:03:40 +01:00
#endif
Draw Manager: T51218 Support edit-mesh hide
2017-04-18 23:29:55 +10:00
VertexBuffer_set_attrib(vbo, pos_id, i, pcenter);
Cleanup: replace num/ct/tot with 'len'
2017-04-19 00:10:32 +10:00
vbo_len_used += 1;
Draw Manager: T51218 Support edit-mesh hide
2017-04-18 23:29:55 +10:00
}
Edit Mode overlay: Moved Shaders to draw modules and resolved some draw issue. We don't want to clutter gpu_shader.c with engine specific code Added face's center dot Simplified loose vert shader
2017-03-02 01:07:03 +01:00
}
Cleanup: replace num/ct/tot with 'len'
2017-04-19 00:10:32 +10:00
if (vbo_len_used != vbo_len_capacity) {
VertexBuffer_resize_data(vbo, vbo_len_used);
}
Edit Mode overlay: Moved Shaders to draw modules and resolved some draw issue. We don't want to clutter gpu_shader.c with engine specific code Added face's center dot Simplified loose vert shader
2017-03-02 01:07:03 +01:00
OpenGL: use PRIM instead of GL enum everywhere else Well, everywhere that uses Gawain for drawing. Places that call OpenGL directly still use GL enums. Part of T49043
2017-04-08 00:50:36 -04:00
cache->overlay_facedots = Batch_create(PRIM_POINTS, vbo, NULL);
Edit Mode overlay: Moved Shaders to draw modules and resolved some draw issue. We don't want to clutter gpu_shader.c with engine specific code Added face's center dot Simplified loose vert shader
2017-03-02 01:07:03 +01:00
Cleanup: remove type prefix from render-data vars Render data is never mixed with other object types, no need to make a distinction.
2017-04-20 16:24:35 +10:00
mesh_render_data_free(rdata);
Edit Mode overlay: Moved Shaders to draw modules and resolved some draw issue. We don't want to clutter gpu_shader.c with engine specific code Added face's center dot Simplified loose vert shader
2017-03-02 01:07:03 +01:00
}
return cache->overlay_facedots;
}
Eevee: Make default shaders works. - Added New Batch cache routine to get the split mesh for each material. (optimization needed) - Did some shader modifications to create default_frag and support a somwhat correct fresnel for lamps (to finish)
2017-04-25 18:46:59 +02:00
Batch **DRW_mesh_batch_cache_get_surface_shaded(Mesh *me)
{
MeshBatchCache *cache = mesh_batch_cache_get(me);
if (cache->shaded_triangles == NULL) {
/* create batch from DM */
Mesh Batch Cache: Add support for uv/vcol/tangent data layers. This is a bad implementation waiting for some improvement that are : - Gawain support for multiple names. -> will remove duplication of attribute alloc - Glsl safe name string escaping. -> will remove any risk of hash colision. Old glsl drawing was relying on attribute being choosen by DerivedMesh at drawing time. For this reason, we declare all possible attrib "name" for each data layer inside the batches. In the glsl code, we declare required data by type and name. Then Gawain only bind vbos if names correspond. This is way cleaner as we do not need to access the CD itself when drawing. One other problem is that the hash maybe 11 caracters long and rapidly overload gawain's attrib name buffer.
2017-05-01 17:58:15 +02:00
const int datatype = MR_DATATYPE_VERT | MR_DATATYPE_LOOP | MR_DATATYPE_LOOPTRI | MR_DATATYPE_POLY | MR_DATATYPE_SHADING;
MeshRenderData *rdata = mesh_render_data_create(me, datatype);
Eevee: Make default shaders works. - Added New Batch cache routine to get the split mesh for each material. (optimization needed) - Did some shader modifications to create default_frag and support a somwhat correct fresnel for lamps (to finish)
2017-04-25 18:46:59 +02:00
const int mat_ct = mesh_render_data_mat_ct_get(rdata);
cache->shaded_triangles = MEM_callocN(sizeof(*cache->shaded_triangles) * mat_ct, "shaded triangles batches");
ElementList **el = mesh_batch_cache_get_shaded_triangles_in_order(rdata, cache);
for (int i = 0; i < mat_ct; ++i) {
cache->shaded_triangles[i] = Batch_create(PRIM_TRIANGLES, mesh_batch_cache_get_shading_data(rdata, cache), el[i]);
}
mesh_render_data_free(rdata);
}
return cache->shaded_triangles;
}
Convert MBC_ API to Mesh (instead of derived mesh) and move it to mesh_render This includes a few fixes in the MBC_ api. The idea here is for this to be the only interface the render engines will deal with for the meshes. If we need to expose special options for sculpting engine we refactor this accordingly. But for now we are shaping this in a per-case base. Note: * We still need to hook up to the depsgraph to force clear/update of batch_cache when mesh changes (I'm waiting for Sergey Sharybin's depsgraph update for this though) * Also ideally we could/should use BMesh directly instead of DerivedMesh, but this will do for now. Note 2: In the end I renamed the `BKE_mesh_render` functions to `static mesh_render`. We can re-expose them as BKE_* later once we need it. Reviewers: merwin Subscribers: fclem Differential Revision: https://developer.blender.org/D2476
2017-01-25 09:16:29 +01:00
#undef MESH_RENDER_FUNCTION
Copy Permalink