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This repository has been archived on 2023-10-09.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
c99e70ae6c7baa7b827514fb5c33462aa74728e3
blender-archive/source/blender/draw/intern/draw_cache_impl_mesh.c

3337 lines
101 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"
Weight painting with draw manager This implements weight rendering with the draw manager, with all drawing options (Shading, wire, face masking, vertex masking). This is part of T51208 Reviewers: campbellbarton Subscribers: dfelinto Differential Revision: https://developer.blender.org/D2654
2017-05-03 18:55:40 +02:00
#include "DNA_object_types.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 "BKE_customdata.h"
Weight painting with draw manager This implements weight rendering with the draw manager, with all drawing options (Shading, wire, face masking, vertex masking). This is part of T51208 Reviewers: campbellbarton Subscribers: dfelinto Differential Revision: https://developer.blender.org/D2654
2017-05-03 18:55:40 +02:00
#include "BKE_deform.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 "BKE_DerivedMesh.h"
#include "BKE_editmesh.h"
DWM: Tangent support for editmesh
2017-05-09 08:50:45 +10:00
#include "BKE_editmesh_tangent.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 "BKE_mesh.h"
Weight painting with draw manager This implements weight rendering with the draw manager, with all drawing options (Shading, wire, face masking, vertex masking). This is part of T51208 Reviewers: campbellbarton Subscribers: dfelinto Differential Revision: https://developer.blender.org/D2654
2017-05-03 18:55:40 +02:00
#include "BKE_texture.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
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"
Use mesh draw cache for back-buffer selection Vertex/weight paint now work with core profile, resolves T51380.
2017-05-05 05:17:31 +10:00
#include "GPU_draw.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
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
/* ---------------------------------------------------------------------- */
Cleanup: use doxy categories
2017-05-05 05:07:52 +10:00
/** \name 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
}
Cleanup: rename ct -> len Also use 'defgroup' instead of 'index' for passing weight arg.
2017-05-04 23:35:50 +10:00
static int mesh_render_mat_len_get(Mesh *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
{
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
}
Cleanup: use doxy categories
2017-05-05 05:07:52 +10:00
/** \} */
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
/* ---------------------------------------------------------------------- */
Cleanup: use doxy categories
2017-05-05 05:07:52 +10:00
/** \name Mesh/BMesh Interface (indirect, partially cached access to complex data).
* \{ */
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 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;
Cleanup: rename ct -> len Also use 'defgroup' instead of 'index' for passing weight arg.
2017-05-04 23:35:50 +10:00
int mat_len;
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];
Weight painting with draw manager This implements weight rendering with the draw manager, with all drawing options (Shading, wire, face masking, vertex masking). This is part of T51208 Reviewers: campbellbarton Subscribers: dfelinto Differential Revision: https://developer.blender.org/D2654
2017-05-03 18:55:40 +02:00
MDeformVert *dvert;
DwM: texture paint support & mask mode Uses workaround so material slots are used when neither blender-internal or cycles are enabled.
2017-05-22 23:31:46 +10:00
MLoopUV *mloopuv;
Cleanup: move custom-data layers into a struct Changed because the values co-exist with active layers (Mesh.mloopuv, Mesh.mloopcol).
2017-05-22 17:21:22 +10:00
MLoopCol *mloopcol;
/* CustomData 'cd' cache for efficient access. */
struct {
struct {
MLoopUV **uv;
int uv_len;
int uv_active;
MLoopCol **vcol;
int vcol_len;
int vcol_active;
float (**tangent)[4];
int tangent_active;
bool *auto_vcol;
} layers;
/* Custom-data offsets (only needed for BMesh access) */
struct {
int crease;
int bweight;
int *uv;
int *vcol;
} offset;
struct {
char (*auto_mix)[32];
char (*uv)[32];
char (*vcol)[32];
char (*tangent)[32];
} uuid;
/* for certain cases we need an output loop-data storage (bmesh tangents) */
struct {
CustomData ldata;
/* grr, special case variable (use in place of 'dm->tangent_mask') */
char tangent_mask;
} output;
} cd;
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;
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];
Vertex paint with draw manager Reviewers: fclem Subscribers: campbellbarton, dfelinto Differential Revision: https://developer.blender.org/D2658
2017-05-04 19:59:27 +02:00
float (*vert_weight_color)[3];
char (*vert_color)[3];
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
short (*poly_normals_short)[3];
short (*vert_normals_short)[3];
Cleanup: clarify select terminology In mesh cache we have select: color, id, bool. Current naming didn't make this clear.
2017-05-17 11:46:26 +10:00
bool *edge_select_bool;
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
} 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,
Weight painting with draw manager This implements weight rendering with the draw manager, with all drawing options (Shading, wire, face masking, vertex masking). This is part of T51208 Reviewers: campbellbarton Subscribers: dfelinto Differential Revision: https://developer.blender.org/D2654
2017-05-03 18:55:40 +02:00
MR_DATATYPE_DVERT = 1 << 7,
Vertex paint with draw manager Reviewers: fclem Subscribers: campbellbarton, dfelinto Differential Revision: https://developer.blender.org/D2658
2017-05-04 19:59:27 +02:00
MR_DATATYPE_LOOPCOL = 1 << 8,
DwM: texture paint support & mask mode Uses workaround so material slots are used when neither blender-internal or cycles are enabled.
2017-05-22 23:31:46 +10:00
MR_DATATYPE_LOOPUV = 1 << 9,
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;
Cleanup: rename ct -> len Also use 'defgroup' instead of 'index' for passing weight arg.
2017-05-04 23:35:50 +10:00
rdata->mat_len = mesh_render_mat_len_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
Cleanup: move custom-data layers into a struct Changed because the values co-exist with active layers (Mesh.mloopuv, Mesh.mloopcol).
2017-05-22 17:21:22 +10:00
CustomData_reset(&rdata->cd.output.ldata);
DWM: Tangent support for editmesh
2017-05-09 08:50:45 +10:00
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
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);
Cleanup: move custom-data layers into a struct Changed because the values co-exist with active layers (Mesh.mloopuv, Mesh.mloopcol).
2017-05-22 17:21:22 +10:00
rdata->cd.offset.crease = CustomData_get_offset(&bm->edata, CD_CREASE);
rdata->cd.offset.bweight = 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
}
Weight painting with draw manager This implements weight rendering with the draw manager, with all drawing options (Shading, wire, face masking, vertex masking). This is part of T51208 Reviewers: campbellbarton Subscribers: dfelinto Differential Revision: https://developer.blender.org/D2654
2017-05-03 18:55:40 +02:00
if (types & (MR_DATATYPE_DVERT)) {
bm_ensure_types |= BM_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
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
}
Weight painting with draw manager This implements weight rendering with the draw manager, with all drawing options (Shading, wire, face masking, vertex masking). This is part of T51208 Reviewers: campbellbarton Subscribers: dfelinto Differential Revision: https://developer.blender.org/D2654
2017-05-03 18:55:40 +02:00
if (types & MR_DATATYPE_DVERT) {
rdata->vert_len = me->totvert;
rdata->dvert = CustomData_get_layer(&me->vdata, CD_MDEFORMVERT);
}
Vertex paint with draw manager Reviewers: fclem Subscribers: campbellbarton, dfelinto Differential Revision: https://developer.blender.org/D2658
2017-05-04 19:59:27 +02:00
if (types & MR_DATATYPE_LOOPCOL) {
rdata->loop_len = me->totloop;
Cleanup: move custom-data layers into a struct Changed because the values co-exist with active layers (Mesh.mloopuv, Mesh.mloopcol).
2017-05-22 17:21:22 +10:00
rdata->mloopcol = CustomData_get_layer(&me->ldata, CD_MLOOPCOL);
Vertex paint with draw manager Reviewers: fclem Subscribers: campbellbarton, dfelinto Differential Revision: https://developer.blender.org/D2658
2017-05-04 19:59:27 +02:00
}
DwM: texture paint support & mask mode Uses workaround so material slots are used when neither blender-internal or cycles are enabled.
2017-05-22 23:31:46 +10:00
if (types & MR_DATATYPE_LOOPUV) {
rdata->loop_len = me->totloop;
rdata->mloopuv = CustomData_get_layer(&me->ldata, CD_MLOOPUV);
}
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) {
DWM: Tangent support for editmesh
2017-05-09 08:50:45 +10:00
CustomData *cd_vdata, *cd_ldata;
if (me->edit_btmesh) {
BMesh *bm = me->edit_btmesh->bm;
cd_vdata = &bm->vdata;
cd_ldata = &bm->ldata;
}
else {
cd_vdata = &me->vdata;
cd_ldata = &me->ldata;
}
rdata->orco = CustomData_get_layer(cd_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);
}
}
/* don't access mesh directly, instead use vars taken from BMesh or Mesh */
#define me DONT_USE_THIS
Cleanup: warnings
2017-05-21 09:39:31 +10:00
#ifdef me /* quiet warning */
#endif
DWM: Tangent support for editmesh
2017-05-09 08:50:45 +10:00
Cleanup: move custom-data layers into a struct Changed because the values co-exist with active layers (Mesh.mloopuv, Mesh.mloopcol).
2017-05-22 17:21:22 +10:00
rdata->cd.layers.uv_len = CustomData_number_of_layers(cd_ldata, CD_MLOOPUV);
rdata->cd.layers.vcol_len = CustomData_number_of_layers(cd_ldata, CD_MLOOPCOL);
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
Cleanup: move custom-data layers into a struct Changed because the values co-exist with active layers (Mesh.mloopuv, Mesh.mloopcol).
2017-05-22 17:21:22 +10:00
rdata->cd.layers.uv = MEM_mallocN(sizeof(*rdata->cd.layers.uv) * rdata->cd.layers.uv_len, __func__);
rdata->cd.layers.vcol = MEM_mallocN(sizeof(*rdata->cd.layers.vcol) * rdata->cd.layers.vcol_len, __func__);
rdata->cd.layers.tangent = MEM_mallocN(sizeof(*rdata->cd.layers.tangent) * rdata->cd.layers.uv_len, __func__);
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
Cleanup: move custom-data layers into a struct Changed because the values co-exist with active layers (Mesh.mloopuv, Mesh.mloopcol).
2017-05-22 17:21:22 +10:00
rdata->cd.uuid.uv = MEM_mallocN(sizeof(*rdata->cd.uuid.uv) * rdata->cd.layers.uv_len, __func__);
rdata->cd.uuid.vcol = MEM_mallocN(sizeof(*rdata->cd.uuid.vcol) * rdata->cd.layers.vcol_len, __func__);
rdata->cd.uuid.tangent = MEM_mallocN(sizeof(*rdata->cd.uuid.tangent) * rdata->cd.layers.uv_len, __func__);
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
Cleanup: move custom-data layers into a struct Changed because the values co-exist with active layers (Mesh.mloopuv, Mesh.mloopcol).
2017-05-22 17:21:22 +10:00
rdata->cd.offset.uv = MEM_mallocN(sizeof(*rdata->cd.offset.uv) * rdata->cd.layers.uv_len, __func__);
rdata->cd.offset.vcol = MEM_mallocN(sizeof(*rdata->cd.offset.vcol) * rdata->cd.layers.vcol_len, __func__);
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
/* Allocate max */
Cleanup: move custom-data layers into a struct Changed because the values co-exist with active layers (Mesh.mloopuv, Mesh.mloopcol).
2017-05-22 17:21:22 +10:00
rdata->cd.layers.auto_vcol = MEM_callocN(
sizeof(*rdata->cd.layers.auto_vcol) * rdata->cd.layers.vcol_len, __func__);
rdata->cd.uuid.auto_mix = MEM_mallocN(
sizeof(*rdata->cd.uuid.auto_mix) * (rdata->cd.layers.vcol_len + rdata->cd.layers.uv_len), __func__);
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
/* 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 */
Cleanup: move custom-data layers into a struct Changed because the values co-exist with active layers (Mesh.mloopuv, Mesh.mloopcol).
2017-05-22 17:21:22 +10:00
for (int i = 0; i < rdata->cd.layers.vcol_len; ++i) {
DWM: Tangent support for editmesh
2017-05-09 08:50:45 +10:00
const char *name = CustomData_get_layer_name(cd_ldata, CD_MLOOPCOL, i);
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
unsigned int hash = BLI_ghashutil_strhash_p(name);
Cleanup: move custom-data layers into a struct Changed because the values co-exist with active layers (Mesh.mloopuv, Mesh.mloopcol).
2017-05-22 17:21:22 +10:00
BLI_snprintf(rdata->cd.uuid.vcol[i], sizeof(*rdata->cd.uuid.vcol), "c%u", hash);
rdata->cd.layers.vcol[i] = CustomData_get_layer_n(cd_ldata, CD_MLOOPCOL, i);
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 (rdata->edit_bmesh) {
Cleanup: move custom-data layers into a struct Changed because the values co-exist with active layers (Mesh.mloopuv, Mesh.mloopcol).
2017-05-22 17:21:22 +10:00
rdata->cd.offset.vcol[i] = CustomData_get_n_offset(&rdata->edit_bmesh->bm->ldata, CD_MLOOPCOL, i);
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
}
/* Gather number of auto layers. */
/* We only do vcols that are not overridden by uvs */
DWM: Tangent support for editmesh
2017-05-09 08:50:45 +10:00
if (CustomData_get_named_layer_index(cd_ldata, CD_MLOOPUV, name) == -1) {
Cleanup: rename layer-names -> layer-uuid These names are created from a hash, don't confuse with actual layer names.
2017-05-09 10:38:52 +10:00
BLI_snprintf(
Cleanup: move custom-data layers into a struct Changed because the values co-exist with active layers (Mesh.mloopuv, Mesh.mloopcol).
2017-05-22 17:21:22 +10:00
rdata->cd.uuid.auto_mix[rdata->cd.layers.uv_len + i],
sizeof(*rdata->cd.uuid.auto_mix), "a%u", hash);
rdata->cd.layers.auto_vcol[i] = true;
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
}
}
/* Start Fresh */
DWM: Tangent support for editmesh
2017-05-09 08:50:45 +10:00
CustomData_free_layers(cd_ldata, CD_MLOOPTANGENT, rdata->loop_len);
Cleanup: move custom-data layers into a struct Changed because the values co-exist with active layers (Mesh.mloopuv, Mesh.mloopcol).
2017-05-22 17:21:22 +10:00
for (int i = 0; i < rdata->cd.layers.uv_len; ++i) {
DWM: Tangent support for editmesh
2017-05-09 08:50:45 +10:00
const char *name = CustomData_get_layer_name(cd_ldata, CD_MLOOPUV, i);
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
unsigned int hash = BLI_ghashutil_strhash_p(name);
{
/* UVs */
Cleanup: move custom-data layers into a struct Changed because the values co-exist with active layers (Mesh.mloopuv, Mesh.mloopcol).
2017-05-22 17:21:22 +10:00
BLI_snprintf(rdata->cd.uuid.uv[i], sizeof(*rdata->cd.uuid.uv), "u%u", hash);
rdata->cd.layers.uv[i] = CustomData_get_layer_n(cd_ldata, CD_MLOOPUV, i);
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 (rdata->edit_bmesh) {
Cleanup: move custom-data layers into a struct Changed because the values co-exist with active layers (Mesh.mloopuv, Mesh.mloopcol).
2017-05-22 17:21:22 +10:00
rdata->cd.offset.uv[i] = CustomData_get_n_offset(&rdata->edit_bmesh->bm->ldata, CD_MLOOPUV, i);
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
}
Cleanup: move custom-data layers into a struct Changed because the values co-exist with active layers (Mesh.mloopuv, Mesh.mloopcol).
2017-05-22 17:21:22 +10:00
BLI_snprintf(rdata->cd.uuid.auto_mix[i], sizeof(*rdata->cd.uuid.auto_mix), "a%u", hash);
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
}
{
/* Tangents*/
Cleanup: move custom-data layers into a struct Changed because the values co-exist with active layers (Mesh.mloopuv, Mesh.mloopcol).
2017-05-22 17:21:22 +10:00
BLI_snprintf(rdata->cd.uuid.tangent[i], sizeof(*rdata->cd.uuid.tangent), "t%u", hash);
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 (rdata->edit_bmesh) {
DWM: Tangent support for editmesh
2017-05-09 08:50:45 +10:00
BMEditMesh *em = rdata->edit_bmesh;
BMesh *bm = em->bm;
Cleanup: move custom-data layers into a struct Changed because the values co-exist with active layers (Mesh.mloopuv, Mesh.mloopcol).
2017-05-22 17:21:22 +10:00
if (!CustomData_has_layer(&rdata->cd.output.ldata, CD_MLOOPTANGENT)) {
DWM: Tangent support for editmesh
2017-05-09 08:50:45 +10:00
bool calc_active_tangent = false;
float (*poly_normals)[3] = rdata->poly_normals;
float (*loop_normals)[3] = CustomData_get_layer(cd_ldata, CD_NORMAL);
char tangent_names[MAX_MTFACE][MAX_NAME];
int tangent_names_len = 0;
Cleanup: move custom-data layers into a struct Changed because the values co-exist with active layers (Mesh.mloopuv, Mesh.mloopcol).
2017-05-22 17:21:22 +10:00
for (tangent_names_len = 0; tangent_names_len < rdata->cd.layers.uv_len; tangent_names_len++) {
DWM: Tangent support for editmesh
2017-05-09 08:50:45 +10:00
BLI_strncpy(
tangent_names[tangent_names_len],
CustomData_get_layer_name(cd_ldata, CD_MLOOPUV, tangent_names_len), MAX_NAME);
}
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
DWM: Tangent support for editmesh
2017-05-09 08:50:45 +10:00
BKE_editmesh_loop_tangent_calc(
em, calc_active_tangent,
tangent_names, tangent_names_len,
poly_normals, loop_normals,
rdata->orco,
Cleanup: move custom-data layers into a struct Changed because the values co-exist with active layers (Mesh.mloopuv, Mesh.mloopcol).
2017-05-22 17:21:22 +10:00
&rdata->cd.output.ldata, bm->totloop,
&rdata->cd.output.tangent_mask);
DWM: Tangent support for editmesh
2017-05-09 08:50:45 +10:00
}
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
DWM: Tangent support for editmesh
2017-05-09 08:50:45 +10:00
/* note: BKE_editmesh_loop_tangent_calc calculates 'CD_TANGENT',
* not 'CD_MLOOPTANGENT' (as done below). It's OK, they're compatible. */
Cleanup: move custom-data layers into a struct Changed because the values co-exist with active layers (Mesh.mloopuv, Mesh.mloopcol).
2017-05-22 17:21:22 +10:00
rdata->cd.layers.tangent[i] = CustomData_get_layer_n(&rdata->cd.output.ldata, CD_TANGENT, i);
BLI_assert(rdata->cd.layers.tangent[i] != NULL);
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
DWM: Tangent support for editmesh
2017-05-09 08:50:45 +10:00
/* special case, we don't use offsets here */
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
}
else {
DWM: Tangent support for editmesh
2017-05-09 08:50:45 +10:00
#undef me
if (!CustomData_has_layer(cd_ldata, CD_NORMAL)) {
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
BKE_mesh_calc_normals_split(me);
}
DWM: Tangent support for editmesh
2017-05-09 08:50:45 +10:00
float (*loopnors)[3] = CustomData_get_layer(cd_ldata, CD_NORMAL);
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
Cleanup: move custom-data layers into a struct Changed because the values co-exist with active layers (Mesh.mloopuv, Mesh.mloopcol).
2017-05-22 17:21:22 +10:00
rdata->cd.layers.tangent[i] = CustomData_add_layer(
DWM: Tangent support for editmesh
2017-05-09 08:50:45 +10:00
cd_ldata, CD_MLOOPTANGENT, CD_CALLOC, NULL, me->totloop);
CustomData_set_layer_flag(cd_ldata, CD_MLOOPTANGENT, CD_FLAG_TEMPORARY);
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
Cleanup: move custom-data layers into a struct Changed because the values co-exist with active layers (Mesh.mloopuv, Mesh.mloopcol).
2017-05-22 17:21:22 +10:00
BKE_mesh_loop_tangents_ex(me->mvert, me->totvert, me->mloop, rdata->cd.layers.tangent[i],
loopnors, rdata->cd.layers.uv[i], me->totloop, me->mpoly, me->totpoly, NULL);
DWM: Tangent support for editmesh
2017-05-09 08:50:45 +10:00
#define me DONT_USE_THIS
Cleanup: warnings
2017-05-21 09:39:31 +10:00
#ifdef me /* quiet warning */
#endif
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
}
}
}
Cleanup: move custom-data layers into a struct Changed because the values co-exist with active layers (Mesh.mloopuv, Mesh.mloopcol).
2017-05-22 17:21:22 +10:00
rdata->cd.layers.uv_active = CustomData_get_active_layer_index(
DWM: Tangent support for editmesh
2017-05-09 08:50:45 +10:00
cd_ldata, CD_MLOOPUV) - CustomData_get_layer_index(cd_ldata, CD_MLOOPUV);
Cleanup: move custom-data layers into a struct Changed because the values co-exist with active layers (Mesh.mloopuv, Mesh.mloopcol).
2017-05-22 17:21:22 +10:00
rdata->cd.layers.vcol_active = CustomData_get_active_layer_index(
DWM: Tangent support for editmesh
2017-05-09 08:50:45 +10:00
cd_ldata, CD_MLOOPCOL) - CustomData_get_layer_index(cd_ldata, CD_MLOOPCOL);
Cleanup: move custom-data layers into a struct Changed because the values co-exist with active layers (Mesh.mloopuv, Mesh.mloopcol).
2017-05-22 17:21:22 +10:00
rdata->cd.layers.tangent_active = CustomData_get_active_layer_index(
DWM: Tangent support for editmesh
2017-05-09 08:50:45 +10:00
cd_ldata, CD_MLOOPTANGENT) - CustomData_get_layer_index(cd_ldata, CD_MLOOPTANGENT);
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
DWM: Tangent support for editmesh
2017-05-09 08:50:45 +10:00
#undef me
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
}
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->orco);
Cleanup: move custom-data layers into a struct Changed because the values co-exist with active layers (Mesh.mloopuv, Mesh.mloopcol).
2017-05-22 17:21:22 +10:00
MEM_SAFE_FREE(rdata->cd.offset.uv);
MEM_SAFE_FREE(rdata->cd.offset.vcol);
MEM_SAFE_FREE(rdata->cd.uuid.auto_mix);
MEM_SAFE_FREE(rdata->cd.uuid.uv);
MEM_SAFE_FREE(rdata->cd.uuid.vcol);
MEM_SAFE_FREE(rdata->cd.uuid.tangent);
MEM_SAFE_FREE(rdata->cd.layers.uv);
MEM_SAFE_FREE(rdata->cd.layers.vcol);
MEM_SAFE_FREE(rdata->cd.layers.tangent);
MEM_SAFE_FREE(rdata->cd.layers.auto_vcol);
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->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);
Vertex paint with draw manager Reviewers: fclem Subscribers: campbellbarton, dfelinto Differential Revision: https://developer.blender.org/D2658
2017-05-04 19:59:27 +02:00
MEM_SAFE_FREE(rdata->vert_weight_color);
Cleanup: clarify select terminology In mesh cache we have select: color, id, bool. Current naming didn't make this clear.
2017-05-17 11:46:26 +10:00
MEM_SAFE_FREE(rdata->edge_select_bool);
Vertex paint with draw manager Reviewers: fclem Subscribers: campbellbarton, dfelinto Differential Revision: https://developer.blender.org/D2658
2017-05-04 19:59:27 +02:00
MEM_SAFE_FREE(rdata->vert_color);
DWM: Tangent support for editmesh
2017-05-09 08:50:45 +10:00
Cleanup: move custom-data layers into a struct Changed because the values co-exist with active layers (Mesh.mloopuv, Mesh.mloopcol).
2017-05-22 17:21:22 +10:00
CustomData_free(&rdata->cd.output.ldata, rdata->loop_len);
DWM: Tangent support for editmesh
2017-05-09 08:50:45 +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
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
}
Cleanup: use doxy categories
2017-05-05 05:07:52 +10:00
/** \} */
/* ---------------------------------------------------------------------- */
/** \name Accessor Functions
* \{ */
Cleanup: rename layer-names -> layer-uuid These names are created from a hash, don't confuse with actual layer names.
2017-05-09 10:38:52 +10:00
static const char *mesh_render_data_uv_auto_layer_uuid_get(const MeshRenderData *rdata, int layer)
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
{
BLI_assert(rdata->types & MR_DATATYPE_SHADING);
Cleanup: move custom-data layers into a struct Changed because the values co-exist with active layers (Mesh.mloopuv, Mesh.mloopcol).
2017-05-22 17:21:22 +10:00
return rdata->cd.uuid.auto_mix[layer];
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
}
Cleanup: rename layer-names -> layer-uuid These names are created from a hash, don't confuse with actual layer names.
2017-05-09 10:38:52 +10:00
static const char *mesh_render_data_vcol_auto_layer_uuid_get(const MeshRenderData *rdata, int layer)
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
{
BLI_assert(rdata->types & MR_DATATYPE_SHADING);
Cleanup: move custom-data layers into a struct Changed because the values co-exist with active layers (Mesh.mloopuv, Mesh.mloopcol).
2017-05-22 17:21:22 +10:00
return rdata->cd.uuid.auto_mix[rdata->cd.layers.uv_len + layer];
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
}
Cleanup: rename layer-names -> layer-uuid These names are created from a hash, don't confuse with actual layer names.
2017-05-09 10:38:52 +10:00
static const char *mesh_render_data_uv_layer_uuid_get(const MeshRenderData *rdata, int layer)
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
{
BLI_assert(rdata->types & MR_DATATYPE_SHADING);
Cleanup: move custom-data layers into a struct Changed because the values co-exist with active layers (Mesh.mloopuv, Mesh.mloopcol).
2017-05-22 17:21:22 +10:00
return rdata->cd.uuid.uv[layer];
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
}
Cleanup: rename layer-names -> layer-uuid These names are created from a hash, don't confuse with actual layer names.
2017-05-09 10:38:52 +10:00
static const char *mesh_render_data_vcol_layer_uuid_get(const MeshRenderData *rdata, int layer)
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
{
BLI_assert(rdata->types & MR_DATATYPE_SHADING);
Cleanup: move custom-data layers into a struct Changed because the values co-exist with active layers (Mesh.mloopuv, Mesh.mloopcol).
2017-05-22 17:21:22 +10:00
return rdata->cd.uuid.vcol[layer];
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
}
Cleanup: rename layer-names -> layer-uuid These names are created from a hash, don't confuse with actual layer names.
2017-05-09 10:38:52 +10:00
static const char *mesh_render_data_tangent_layer_uuid_get(const MeshRenderData *rdata, int layer)
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
{
BLI_assert(rdata->types & MR_DATATYPE_SHADING);
Cleanup: move custom-data layers into a struct Changed because the values co-exist with active layers (Mesh.mloopuv, Mesh.mloopcol).
2017-05-22 17:21:22 +10:00
return rdata->cd.uuid.tangent[layer];
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
}
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
}
Cleanup: rename ct -> len Also use 'defgroup' instead of 'index' for passing weight arg.
2017-05-04 23:35:50 +10:00
static int mesh_render_data_mat_len_get(const MeshRenderData *rdata)
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
{
BLI_assert(rdata->types & MR_DATATYPE_POLY);
Cleanup: rename ct -> len Also use 'defgroup' instead of 'index' for passing weight arg.
2017-05-04 23:35:50 +10:00
return rdata->mat_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
}
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
}
Cleanup: use doxy categories
2017-05-05 05:07:52 +10:00
/** \} */
Cleanup: de-duplicate mesh data calculation
2017-05-05 04:56:28 +10:00
/* ---------------------------------------------------------------------- */
/** \name Internal Cache (Lazy Initialization)
* \{ */
/** Ensure #MeshRenderData.poly_normals_short */
static void mesh_render_data_ensure_poly_normals_short(MeshRenderData *rdata)
{
short (*pnors_short)[3] = rdata->poly_normals_short;
if (pnors_short == NULL) {
if (rdata->edit_bmesh) {
BMesh *bm = rdata->edit_bmesh->bm;
BMIter fiter;
BMFace *face;
int i;
pnors_short = rdata->poly_normals_short = MEM_mallocN(sizeof(*pnors_short) * rdata->poly_len, __func__);
BM_ITER_MESH_INDEX(face, &fiter, bm, BM_FACES_OF_MESH, i) {
normal_float_to_short_v3(pnors_short[i], face->no);
}
}
else {
float (*pnors)[3] = rdata->poly_normals;
if (!pnors) {
pnors = rdata->poly_normals = MEM_mallocN(sizeof(*pnors) * rdata->poly_len, __func__);
BKE_mesh_calc_normals_poly(
rdata->mvert, NULL, rdata->vert_len,
rdata->mloop, rdata->mpoly, rdata->loop_len, rdata->poly_len, pnors, true);
}
pnors_short = rdata->poly_normals_short = MEM_mallocN(sizeof(*pnors_short) * rdata->poly_len, __func__);
for (int i = 0; i < rdata->poly_len; i++) {
normal_float_to_short_v3(pnors_short[i], pnors[i]);
}
}
}
}
/** Ensure #MeshRenderData.vert_normals_short */
static void mesh_render_data_ensure_vert_normals_short(MeshRenderData *rdata)
{
short (*vnors_short)[3] = rdata->vert_normals_short;
if (vnors_short == NULL) {
if (rdata->edit_bmesh) {
BMesh *bm = rdata->edit_bmesh->bm;
BMIter viter;
BMVert *vert;
int i;
vnors_short = rdata->vert_normals_short = MEM_mallocN(sizeof(*vnors_short) * rdata->vert_len, __func__);
BM_ITER_MESH_INDEX(vert, &viter, bm, BM_VERT, i) {
normal_float_to_short_v3(vnors_short[i], vert->no);
}
}
else {
/* data from mesh used directly */
BLI_assert(0);
}
}
}
/** Ensure #MeshRenderData.vert_color */
static void mesh_render_data_ensure_vert_color(MeshRenderData *rdata)
{
char (*vcol)[3] = rdata->vert_color;
if (vcol == NULL) {
if (rdata->edit_bmesh) {
Vertex color access for edit-mode Currently not used, adding since legacy drawing can display this.
2017-05-08 12:41:05 +10:00
BMesh *bm = rdata->edit_bmesh->bm;
const int cd_loop_color_offset = CustomData_get_offset(&bm->ldata, CD_MLOOPCOL);
if (cd_loop_color_offset == -1) {
goto fallback;
}
vcol = rdata->vert_color = MEM_mallocN(sizeof(*vcol) * rdata->loop_len, __func__);
BMIter fiter;
BMFace *face;
int i = 0;
BM_ITER_MESH(face, &fiter, bm, BM_FACES_OF_MESH) {
BMLoop *l_iter, *l_first;
l_iter = l_first = BM_FACE_FIRST_LOOP(face);
do {
const MLoopCol *lcol = BM_ELEM_CD_GET_VOID_P(l_iter, cd_loop_color_offset);
vcol[i][0] = lcol->r;
vcol[i][1] = lcol->g;
vcol[i][2] = lcol->b;
i += 1;
} while ((l_iter = l_iter->next) != l_first);
}
BLI_assert(i == rdata->loop_len);
Cleanup: de-duplicate mesh data calculation
2017-05-05 04:56:28 +10:00
}
else {
Cleanup: move custom-data layers into a struct Changed because the values co-exist with active layers (Mesh.mloopuv, Mesh.mloopcol).
2017-05-22 17:21:22 +10:00
if (rdata->mloopcol == NULL) {
Cleanup: de-duplicate mesh data calculation
2017-05-05 04:56:28 +10:00
goto fallback;
}
vcol = rdata->vert_color = MEM_mallocN(sizeof(*vcol) * rdata->loop_len, __func__);
for (int i = 0; i < rdata->loop_len; i++) {
Cleanup: move custom-data layers into a struct Changed because the values co-exist with active layers (Mesh.mloopuv, Mesh.mloopcol).
2017-05-22 17:21:22 +10:00
vcol[i][0] = rdata->mloopcol[i].r;
vcol[i][1] = rdata->mloopcol[i].g;
vcol[i][2] = rdata->mloopcol[i].b;
Cleanup: de-duplicate mesh data calculation
2017-05-05 04:56:28 +10:00
}
}
}
return;
fallback:
vcol = rdata->vert_color = MEM_mallocN(sizeof(*vcol) * rdata->loop_len, __func__);
for (int i = 0; i < rdata->loop_len; i++) {
vcol[i][0] = 255;
vcol[i][1] = 255;
vcol[i][2] = 255;
}
}
/* TODO, move into shader? */
static void rgb_from_weight(float r_rgb[3], const float weight)
{
const float blend = ((weight / 2.0f) + 0.5f);
if (weight <= 0.25f) { /* blue->cyan */
r_rgb[0] = 0.0f;
r_rgb[1] = blend * weight * 4.0f;
r_rgb[2] = blend;
}
else if (weight <= 0.50f) { /* cyan->green */
r_rgb[0] = 0.0f;
r_rgb[1] = blend;
r_rgb[2] = blend * (1.0f - ((weight - 0.25f) * 4.0f));
}
else if (weight <= 0.75f) { /* green->yellow */
r_rgb[0] = blend * ((weight - 0.50f) * 4.0f);
r_rgb[1] = blend;
r_rgb[2] = 0.0f;
}
else if (weight <= 1.0f) { /* yellow->red */
r_rgb[0] = blend;
r_rgb[1] = blend * (1.0f - ((weight - 0.75f) * 4.0f));
r_rgb[2] = 0.0f;
}
else {
/* exceptional value, unclamped or nan,
* avoid uninitialized memory use */
r_rgb[0] = 1.0f;
r_rgb[1] = 0.0f;
r_rgb[2] = 1.0f;
}
}
Correct own error in copying vertex color to VBO's
2017-05-17 17:14:10 +10:00
/** Ensure #MeshRenderData.vert_weight_color */
Cleanup: de-duplicate mesh data calculation
2017-05-05 04:56:28 +10:00
static void mesh_render_data_ensure_vert_weight_color(MeshRenderData *rdata, const int defgroup)
{
float (*vweight)[3] = rdata->vert_weight_color;
if (vweight == NULL) {
if (rdata->edit_bmesh) {
BMesh *bm = rdata->edit_bmesh->bm;
const int cd_dvert_offset = CustomData_get_offset(&bm->vdata, CD_MDEFORMVERT);
if (cd_dvert_offset == -1) {
goto fallback;
}
BMIter viter;
BMVert *vert;
int i;
vweight = rdata->vert_weight_color = MEM_mallocN(sizeof(*vweight) * rdata->vert_len, __func__);
BM_ITER_MESH_INDEX(vert, &viter, bm, BM_VERT, i) {
const MDeformVert *dvert = BM_ELEM_CD_GET_VOID_P(vert, cd_dvert_offset);
float weight = defvert_find_weight(dvert, defgroup);
if (U.flag & USER_CUSTOM_RANGE) {
do_colorband(&U.coba_weight, weight, vweight[i]);
}
else {
rgb_from_weight(vweight[i], weight);
}
}
}
else {
if (rdata->dvert == NULL) {
goto fallback;
}
vweight = rdata->vert_weight_color = MEM_mallocN(sizeof(*vweight) * rdata->vert_len, __func__);
for (int i = 0; i < rdata->vert_len; i++) {
float weight = defvert_find_weight(&rdata->dvert[i], defgroup);
if (U.flag & USER_CUSTOM_RANGE) {
do_colorband(&U.coba_weight, weight, vweight[i]);
}
else {
rgb_from_weight(vweight[i], weight);
}
}
}
}
return;
fallback:
vweight = rdata->vert_weight_color = MEM_callocN(sizeof(*vweight) * rdata->vert_len, __func__);
for (int i = 0; i < rdata->vert_len; i++) {
vweight[i][2] = 0.5f;
}
}
Cleanup: clarify select terminology In mesh cache we have select: color, id, bool. Current naming didn't make this clear.
2017-05-17 11:46:26 +10:00
/** Ensure #MeshRenderData.edge_select_bool */
static void mesh_render_data_ensure_edge_select_bool(MeshRenderData *rdata, bool use_wire)
DWM: write mesh data to the VBO directly Was using intermediate functions to store each data type. This is still done in some places (for more involved access). For basic data: coords, normals, colors..etc this is unnecessarily indirect.
2017-05-17 11:28:44 +10:00
{
Cleanup: clarify select terminology In mesh cache we have select: color, id, bool. Current naming didn't make this clear.
2017-05-17 11:46:26 +10:00
bool *edge_select_bool = rdata->edge_select_bool;
if (edge_select_bool == NULL) {
edge_select_bool = rdata->edge_select_bool =
MEM_callocN(sizeof(*edge_select_bool) * rdata->edge_len, __func__);
DWM: write mesh data to the VBO directly Was using intermediate functions to store each data type. This is still done in some places (for more involved access). For basic data: coords, normals, colors..etc this is unnecessarily indirect.
2017-05-17 11:28:44 +10:00
for (int i = 0; i < rdata->poly_len; i++) {
MPoly *poly = &rdata->mpoly[i];
if (poly->flag & ME_FACE_SEL) {
for (int j = 0; j < poly->totloop; j++) {
MLoop *loop = &rdata->mloop[poly->loopstart + j];
if (use_wire) {
Cleanup: clarify select terminology In mesh cache we have select: color, id, bool. Current naming didn't make this clear.
2017-05-17 11:46:26 +10:00
edge_select_bool[loop->e] = true;
DWM: write mesh data to the VBO directly Was using intermediate functions to store each data type. This is still done in some places (for more involved access). For basic data: coords, normals, colors..etc this is unnecessarily indirect.
2017-05-17 11:28:44 +10:00
}
else {
Cleanup: clarify select terminology In mesh cache we have select: color, id, bool. Current naming didn't make this clear.
2017-05-17 11:46:26 +10:00
/* Not totally correct, will cause problems for edges with 3x faces. */
edge_select_bool[loop->e] = !edge_select_bool[loop->e];
DWM: write mesh data to the VBO directly Was using intermediate functions to store each data type. This is still done in some places (for more involved access). For basic data: coords, normals, colors..etc this is unnecessarily indirect.
2017-05-17 11:28:44 +10:00
}
}
}
}
}
}
Cleanup: de-duplicate mesh data calculation
2017-05-05 04:56:28 +10:00
/** \} */
Cleanup: use doxy categories
2017-05-05 05:07:52 +10:00
/* ---------------------------------------------------------------------- */
/** \name Internal Cache Generation
* \{ */
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
}
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];
Cleanup: move custom-data layers into a struct Changed because the values co-exist with active layers (Mesh.mloopuv, Mesh.mloopcol).
2017-05-22 17:21:22 +10:00
(*r_vert_uvs)[0] = ((MLoopUV *)BM_ELEM_CD_GET_VOID_P(bm_looptri[0], rdata->cd.offset.uv[uv_layer]))->uv;
(*r_vert_uvs)[1] = ((MLoopUV *)BM_ELEM_CD_GET_VOID_P(bm_looptri[1], rdata->cd.offset.uv[uv_layer]))->uv;
(*r_vert_uvs)[2] = ((MLoopUV *)BM_ELEM_CD_GET_VOID_P(bm_looptri[2], rdata->cd.offset.uv[uv_layer]))->uv;
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
}
else {
const MLoopTri *mlt = &rdata->mlooptri[tri_idx];
Cleanup: move custom-data layers into a struct Changed because the values co-exist with active layers (Mesh.mloopuv, Mesh.mloopcol).
2017-05-22 17:21:22 +10:00
(*r_vert_uvs)[0] = rdata->cd.layers.uv[uv_layer][mlt->tri[0]].uv;
(*r_vert_uvs)[1] = rdata->cd.layers.uv[uv_layer][mlt->tri[1]].uv;
(*r_vert_uvs)[2] = rdata->cd.layers.uv[uv_layer][mlt->tri[2]].uv;
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_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];
Cleanup: move custom-data layers into a struct Changed because the values co-exist with active layers (Mesh.mloopuv, Mesh.mloopcol).
2017-05-22 17:21:22 +10:00
(*r_vert_cols)[0] = &((MLoopCol *)BM_ELEM_CD_GET_VOID_P(bm_looptri[0], rdata->cd.offset.vcol[vcol_layer]))->r;
(*r_vert_cols)[1] = &((MLoopCol *)BM_ELEM_CD_GET_VOID_P(bm_looptri[1], rdata->cd.offset.vcol[vcol_layer]))->r;
(*r_vert_cols)[2] = &((MLoopCol *)BM_ELEM_CD_GET_VOID_P(bm_looptri[2], rdata->cd.offset.vcol[vcol_layer]))->r;
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
}
else {
const MLoopTri *mlt = &rdata->mlooptri[tri_idx];
Cleanup: move custom-data layers into a struct Changed because the values co-exist with active layers (Mesh.mloopuv, Mesh.mloopcol).
2017-05-22 17:21:22 +10:00
(*r_vert_cols)[0] = &rdata->cd.layers.vcol[vcol_layer][mlt->tri[0]].r;
(*r_vert_cols)[1] = &rdata->cd.layers.vcol[vcol_layer][mlt->tri[1]].r;
(*r_vert_cols)[2] = &rdata->cd.layers.vcol[vcol_layer][mlt->tri[2]].r;
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_tans_get(
MeshRenderData *rdata, const int tri_idx, const int tangent_layer,
float *(*r_vert_tans)[3])
{
if (rdata->edit_bmesh) {
const BMLoop **bm_looptri = (const BMLoop **)rdata->edit_bmesh->looptris[tri_idx];
Cleanup: move custom-data layers into a struct Changed because the values co-exist with active layers (Mesh.mloopuv, Mesh.mloopcol).
2017-05-22 17:21:22 +10:00
(*r_vert_tans)[0] = rdata->cd.layers.tangent[tangent_layer][BM_elem_index_get(bm_looptri[0])];
(*r_vert_tans)[1] = rdata->cd.layers.tangent[tangent_layer][BM_elem_index_get(bm_looptri[1])];
(*r_vert_tans)[2] = rdata->cd.layers.tangent[tangent_layer][BM_elem_index_get(bm_looptri[2])];
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
}
else {
const MLoopTri *mlt = &rdata->mlooptri[tri_idx];
Cleanup: move custom-data layers into a struct Changed because the values co-exist with active layers (Mesh.mloopuv, Mesh.mloopcol).
2017-05-22 17:21:22 +10:00
(*r_vert_tans)[0] = rdata->cd.layers.tangent[tangent_layer][mlt->tri[0]];
(*r_vert_tans)[1] = rdata->cd.layers.tangent[tangent_layer][mlt->tri[1]];
(*r_vert_tans)[2] = rdata->cd.layers.tangent[tangent_layer][mlt->tri[2]];
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
}
}
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(
DWM: Initial support vert & weight hidden faces
2017-05-17 16:01:18 +10:00
MeshRenderData *rdata, const int tri_idx, const bool use_hide,
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
{
Mesh Batch Cache: Split Edit Mesh mode Vertex Buffers for faster update.
2017-05-13 16:45:06 +02:00
BLI_assert(rdata->types & MR_DATATYPE_VERT);
BLI_assert(rdata->types & MR_DATATYPE_LOOPTRI);
BLI_assert(rdata->types & MR_DATATYPE_LOOP);
BLI_assert(rdata->types & 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
DWM: Initial support vert & weight hidden faces
2017-05-17 16:01:18 +10:00
/* Assume 'use_hide' */
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;
}
Missed last commit
2017-05-05 05:06:57 +10:00
mesh_render_data_ensure_poly_normals_short(rdata);
mesh_render_data_ensure_vert_normals_short(rdata);
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
(*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];
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
DWM: Initial support vert & weight hidden faces
2017-05-17 16:01:18 +10:00
if (use_hide && (rdata->mpoly[mlt->poly].flag & ME_HIDE)) {
return false;
}
Missed last commit
2017-05-05 05:06:57 +10:00
mesh_render_data_ensure_poly_normals_short(rdata);
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
Missed last commit
2017-05-05 05:06:57 +10:00
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
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: move custom-data layers into a struct Changed because the values co-exist with active layers (Mesh.mloopuv, Mesh.mloopcol).
2017-05-22 17:21:22 +10:00
if (rdata->cd.offset.crease != -1) {
float crease = BM_ELEM_CD_GET_FLOAT(be, rdata->cd.offset.crease);
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: move custom-data layers into a struct Changed because the values co-exist with active layers (Mesh.mloopuv, Mesh.mloopcol).
2017-05-22 17:21:22 +10:00
if (rdata->cd.offset.bweight != -1) {
float bweight = BM_ELEM_CD_GET_FLOAT(be, rdata->cd.offset.bweight);
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(
Mesh Batch Cache: Split Edit Mesh mode Vertex Buffers for faster update.
2017-05-13 16:45:06 +02:00
MeshRenderData *rdata, VertexBuffer *vbo_pos, VertexBuffer *vbo_nor, VertexBuffer *vbo_data,
const unsigned int pos_id, const unsigned int vnor_id, const unsigned int lnor_id, const unsigned int data_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
{
Cleanup: replace char array with struct
2017-04-12 14:08:44 +10:00
EdgeDrawAttr *eattr;
Mesh Batch Cache: Split Edit Mesh mode Vertex Buffers for faster update.
2017-05-13 16:45:06 +02:00
unsigned char fflag;
unsigned char vflag;
if (vbo_pos) {
for (int i = 0; i < 3; ++i) {
const float *pos = mesh_render_data_vert_co(rdata, tri_vert_idx[i]);
VertexBuffer_set_attrib(vbo_pos, pos_id, base_vert_idx + i, pos);
}
}
if (vbo_nor) {
float *tri_vert_cos[3];
short *tri_nor, *tri_vert_nors[3];
bool is_smooth;
mesh_render_data_looptri_cos_nors_smooth_get(
DWM: Initial support vert & weight hidden faces
2017-05-17 16:01:18 +10:00
rdata, f, false, &tri_vert_cos, &tri_nor, &tri_vert_nors, &is_smooth);
Mesh Batch Cache: Split Edit Mesh mode Vertex Buffers for faster update.
2017-05-13 16:45:06 +02:00
for (int i = 0; i < 3; ++i) {
/* TODO real loop normal */
const short *svnor = mesh_render_data_vert_nor(rdata, tri_vert_idx[i]);
const short *slnor = tri_vert_nors[i];
fflag = mesh_render_data_looptri_flag(rdata, f);
Gawain: promote 10_10_10 to first-class vertex format This format is part of OpenGL 3.3, and one of the reasons for choosing 3.3 over 3.2. Instead of checking #if USE_10_10_10 just use it wherever needed.
2017-05-21 15:59:34 -04:00
Mesh Batch Cache: Split Edit Mesh mode Vertex Buffers for faster update.
2017-05-13 16:45:06 +02:00
PackedNormal vnor = convert_i10_s3(svnor);
PackedNormal lnor = convert_i10_s3(slnor);
VertexBuffer_set_attrib(vbo_nor, vnor_id, base_vert_idx + i, &vnor);
VertexBuffer_set_attrib(vbo_nor, lnor_id, base_vert_idx + i, &lnor);
}
}
if (vbo_data) {
fflag = mesh_render_data_looptri_flag(rdata, f);
for (int i = 0; i < 3; ++i) {
int iedge = (i == 2) ? 0 : i+1;
eattr = mesh_render_data_edge_flag(rdata, tri_edge_idx[iedge]);
vflag = mesh_render_data_vertex_flag(rdata, tri_vert_idx[i]);
eattr->v_flag = fflag | vflag;
VertexBuffer_set_attrib(vbo_data, data_id, base_vert_idx + i, 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(
Mesh Batch Cache: Split Edit Mesh mode Vertex Buffers for faster update.
2017-05-13 16:45:06 +02:00
MeshRenderData *rdata, VertexBuffer *vbo_pos, VertexBuffer *vbo_nor, VertexBuffer *vbo_data,
const unsigned int pos_id, const unsigned int vnor_id, const unsigned int data_id,
const int edge_vert_idx[2], const int e, const int base_vert_idx)
Edit Mode overlays: separate multiple shaders for loose edges and verts
2017-03-01 14:07:04 +01:00
{
Mesh Batch Cache: Split Edit Mesh mode Vertex Buffers for faster update.
2017-05-13 16:45:06 +02:00
if (vbo_pos) {
for (int i = 0; i < 2; ++i) {
const float *pos = mesh_render_data_vert_co(rdata, edge_vert_idx[i]);
VertexBuffer_set_attrib(vbo_pos, pos_id, base_vert_idx + i, pos);
}
}
Edit Mode overlays: separate multiple shaders for loose edges and verts
2017-03-01 14:07:04 +01:00
Mesh Batch Cache: Split Edit Mesh mode Vertex Buffers for faster update.
2017-05-13 16:45:06 +02:00
if (vbo_nor) {
for (int i = 0; i < 2; ++i) {
short *nor = mesh_render_data_vert_nor(rdata, edge_vert_idx[i]);
PackedNormal vnor = convert_i10_s3(nor);
VertexBuffer_set_attrib(vbo_nor, vnor_id, base_vert_idx + i, &vnor);
}
}
if (vbo_data) {
EdgeDrawAttr *eattr = mesh_render_data_edge_flag(rdata, e);
for (int i = 0; i < 2; ++i) {
eattr->v_flag = mesh_render_data_vertex_flag(rdata, edge_vert_idx[i]);
VertexBuffer_set_attrib(vbo_data, data_id, base_vert_idx + i, 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(
Mesh Batch Cache: Split Edit Mesh mode Vertex Buffers for faster update.
2017-05-13 16:45:06 +02:00
MeshRenderData *rdata, VertexBuffer *vbo_pos, VertexBuffer *vbo_nor, VertexBuffer *vbo_data,
const unsigned int pos_id, const unsigned int vnor_id, const unsigned int data_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)
{
Mesh Batch Cache: Split Edit Mesh mode Vertex Buffers for faster update.
2017-05-13 16:45:06 +02:00
if (vbo_pos) {
const float *pos = mesh_render_data_vert_co(rdata, v);
VertexBuffer_set_attrib(vbo_pos, pos_id, base_vert_idx, pos);
}
if (vbo_nor) {
short *nor = mesh_render_data_vert_nor(rdata, v);
Mesh Batch Cache: Use compressed format. We don't compress UV by default because that would mean restricting them to [0, 1] Leaving all other code for later settings switch.
2017-05-15 17:55:02 +02:00
PackedNormal vnor = convert_i10_s3(nor);
Mesh Batch Cache: Split Edit Mesh mode Vertex Buffers for faster update.
2017-05-13 16:45:06 +02:00
VertexBuffer_set_attrib(vbo_nor, vnor_id, base_vert_idx, &vnor);
}
if (vbo_data) {
unsigned char vflag[4] = {0, 0, 0, 0};
vflag[0] = mesh_render_data_vertex_flag(rdata, v);
VertexBuffer_set_attrib(vbo_data, data_id, base_vert_idx, 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
Cleanup: use doxy categories
2017-05-05 05:07:52 +10:00
/** \} */
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
/* ---------------------------------------------------------------------- */
Cleanup: use doxy categories
2017-05-05 05:07:52 +10:00
/** \name Mesh Batch 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
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;
DWM: reuse position VBO for vpaint select overlay
2017-05-17 19:21:16 +10:00
ElementList *overlay_triangles_vpaint;
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 *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;
DWM: Initial support vert & weight hidden faces
2017-05-17 16:01:18 +10:00
VertexBuffer *pos_with_normals_visible_only; /* for paint modes with vert/face hide support. */
DWM: Share VBO positions & normals for mesh cache Select, weight & color cache included redundant position and normals. Use multiple VBO's per batch for more efficient storage.
2017-05-08 16:25:29 +10:00
VertexBuffer *tri_aligned_weights;
VertexBuffer *tri_aligned_vert_colors;
VertexBuffer *tri_aligned_select_id;
DwM: texture paint support & mask mode Uses workaround so material slots are used when neither blender-internal or cycles are enabled.
2017-05-22 23:31:46 +10:00
VertexBuffer *tri_aligned_uv; /* Active UV layer (mloopuv) */
Cleanup: clarify select terminology In mesh cache we have select: color, id, bool. Current naming didn't make this clear.
2017-05-17 11:46:26 +10:00
VertexBuffer *edge_pos_with_select_bool;
VertexBuffer *pos_with_select_bool;
Edit Mesh overlay: Ported Display Normals option
2017-03-09 01:29:58 +01:00
Batch *triangles_with_normals;
DWM: Initial support vert & weight hidden faces
2017-05-17 16:01:18 +10:00
/* Skip hidden (depending on paint select mode),
* 'pos_with_normals' or 'pos_with_normals_visible_only'. */
Weight painting with draw manager This implements weight rendering with the draw manager, with all drawing options (Shading, wire, face masking, vertex masking). This is part of T51208 Reviewers: campbellbarton Subscribers: dfelinto Differential Revision: https://developer.blender.org/D2654
2017-05-03 18:55:40 +02:00
Batch *triangles_with_weights;
Vertex paint with draw manager Reviewers: fclem Subscribers: campbellbarton, dfelinto Differential Revision: https://developer.blender.org/D2658
2017-05-04 19:59:27 +02:00
Batch *triangles_with_vert_colors;
DWM: Initial support vert & weight hidden faces
2017-05-17 16:01:18 +10:00
/* Always skip hidden */
Use mesh draw cache for back-buffer selection Vertex/weight paint now work with core profile, resolves T51380.
2017-05-05 05:17:31 +10:00
Batch *triangles_with_select_id;
DWM: Initial support vert & weight hidden faces
2017-05-17 16:01:18 +10:00
Edit Mesh overlay: Ported Display Normals option
2017-03-09 01:29:58 +01:00
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;
DwM: texture paint support & mask mode Uses workaround so material slots are used when neither blender-internal or cycles are enabled.
2017-05-22 23:31:46 +10:00
/* Texture Paint.*/
/* per-texture batch */
Batch **texpaint_triangles;
Batch *texpaint_triangles_single;
Mesh Batch Cache: Split Edit Mesh mode Vertex Buffers for faster update.
2017-05-13 16:45:06 +02:00
/* Edit Cage Mesh buffers */
VertexBuffer *ed_tri_pos;
VertexBuffer *ed_tri_nor; /* LoopNor, VertNor */
VertexBuffer *ed_tri_data;
VertexBuffer *ed_ledge_pos;
VertexBuffer *ed_ledge_nor; /* VertNor */
VertexBuffer *ed_ledge_data;
VertexBuffer *ed_lvert_pos;
VertexBuffer *ed_lvert_nor; /* VertNor */
VertexBuffer *ed_lvert_data;
VertexBuffer *ed_fcenter_pos;
VertexBuffer *ed_fcenter_nor;
Edit Mode overlays: separate multiple shaders for loose edges and verts
2017-03-01 14:07:04 +01:00
Batch *overlay_triangles;
Mesh Batch Cache: Split Edit Mesh mode Vertex Buffers for faster update.
2017-05-13 16:45:06 +02:00
Batch *overlay_triangles_nor; /* PRIM_POINTS */
Edit Mode overlays: separate multiple shaders for loose edges and verts
2017-03-01 14:07:04 +01:00
Batch *overlay_loose_edges;
Mesh Batch Cache: Split Edit Mesh mode Vertex Buffers for faster update.
2017-05-13 16:45:06 +02:00
Batch *overlay_loose_edges_nor; /* PRIM_POINTS */
Batch *overlay_loose_verts;
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;
Mesh Batch Cache: Split Edit Mesh mode Vertex Buffers for faster update.
2017-05-13 16:45:06 +02:00
Weight painting with draw manager This implements weight rendering with the draw manager, with all drawing options (Shading, wire, face masking, vertex masking). This is part of T51208 Reviewers: campbellbarton Subscribers: dfelinto Differential Revision: https://developer.blender.org/D2654
2017-05-03 18:55:40 +02:00
Batch *overlay_weight_faces;
Batch *overlay_weight_verts;
DWM: reuse position VBO for vpaint select overlay
2017-05-17 19:21:16 +10:00
Batch *overlay_paint_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
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;
Vertex paint with draw manager Reviewers: fclem Subscribers: campbellbarton, dfelinto Differential Revision: https://developer.blender.org/D2658
2017-05-04 19:59:27 +02:00
bool is_paint_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;
Cleanup: rename ct -> len Also use 'defgroup' instead of 'index' for passing weight arg.
2017-05-04 23:35:50 +10:00
int mat_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
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 */
Cleanup: rename ct -> len Also use 'defgroup' instead of 'index' for passing weight arg.
2017-05-04 23:35:50 +10:00
if (cache->mat_len != mesh_render_mat_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->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;
}
Vertex paint with draw manager Reviewers: fclem Subscribers: campbellbarton, dfelinto Differential Revision: https://developer.blender.org/D2658
2017-05-04 19:59:27 +02:00
if (cache->is_paint_dirty) {
Weight painting with draw manager This implements weight rendering with the draw manager, with all drawing options (Shading, wire, face masking, vertex masking). This is part of T51208 Reviewers: campbellbarton Subscribers: dfelinto Differential Revision: https://developer.blender.org/D2654
2017-05-03 18:55:40 +02:00
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)) ||
Cleanup: rename ct -> len Also use 'defgroup' instead of 'index' for passing weight arg.
2017-05-04 23:35:50 +10:00
(cache->mat_len != mesh_render_mat_len_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
Cleanup: rename ct -> len Also use 'defgroup' instead of 'index' for passing weight arg.
2017-05-04 23:35:50 +10:00
cache->mat_len = mesh_render_mat_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
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;
Vertex paint with draw manager Reviewers: fclem Subscribers: campbellbarton, dfelinto Differential Revision: https://developer.blender.org/D2658
2017-05-04 19:59:27 +02:00
cache->is_paint_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:
Mesh Batch Cache: Split Edit Mesh mode Vertex Buffers for faster update.
2017-05-13 16:45:06 +02:00
VERTEXBUFFER_DISCARD_SAFE(cache->ed_tri_data);
VERTEXBUFFER_DISCARD_SAFE(cache->ed_ledge_data);
VERTEXBUFFER_DISCARD_SAFE(cache->ed_lvert_data);
VERTEXBUFFER_DISCARD_SAFE(cache->ed_fcenter_nor); /* Contains select flag */
BATCH_DISCARD_SAFE(cache->overlay_triangles);
BATCH_DISCARD_SAFE(cache->overlay_loose_verts);
BATCH_DISCARD_SAFE(cache->overlay_loose_edges);
Fix memory leak with mesh cache (w/ select invalidate)
2017-05-16 14:55:29 +10:00
BATCH_DISCARD_ALL_SAFE(cache->overlay_facedots);
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
break;
Vertex paint with draw manager Reviewers: fclem Subscribers: campbellbarton, dfelinto Differential Revision: https://developer.blender.org/D2658
2017-05-04 19:59:27 +02:00
case BKE_MESH_BATCH_DIRTY_PAINT:
cache->is_paint_dirty = true;
Weight painting with draw manager This implements weight rendering with the draw manager, with all drawing options (Shading, wire, face masking, vertex masking). This is part of T51208 Reviewers: campbellbarton Subscribers: dfelinto Differential Revision: https://developer.blender.org/D2654
2017-05-03 18:55:40 +02:00
break;
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
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);
Re-use vertex position VBO in weight-paint mode
2017-05-18 14:53:43 +10:00
VERTEXBUFFER_DISCARD_SAFE(cache->pos_with_select_bool);
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
ELEMENTLIST_DISCARD_SAFE(cache->edges_in_order);
ELEMENTLIST_DISCARD_SAFE(cache->triangles_in_order);
DWM: reuse position VBO for vpaint select overlay
2017-05-17 19:21:16 +10:00
ELEMENTLIST_DISCARD_SAFE(cache->overlay_triangles_vpaint);
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 Batch Cache: Split Edit Mesh mode Vertex Buffers for faster update.
2017-05-13 16:45:06 +02:00
VERTEXBUFFER_DISCARD_SAFE(cache->ed_tri_pos);
VERTEXBUFFER_DISCARD_SAFE(cache->ed_tri_nor);
VERTEXBUFFER_DISCARD_SAFE(cache->ed_tri_data);
VERTEXBUFFER_DISCARD_SAFE(cache->ed_ledge_pos);
VERTEXBUFFER_DISCARD_SAFE(cache->ed_ledge_nor);
VERTEXBUFFER_DISCARD_SAFE(cache->ed_ledge_data);
VERTEXBUFFER_DISCARD_SAFE(cache->ed_lvert_pos);
VERTEXBUFFER_DISCARD_SAFE(cache->ed_lvert_nor);
VERTEXBUFFER_DISCARD_SAFE(cache->ed_lvert_data);
VERTEXBUFFER_DISCARD_SAFE(cache->ed_fcenter_pos);
VERTEXBUFFER_DISCARD_SAFE(cache->ed_fcenter_nor);
BATCH_DISCARD_SAFE(cache->overlay_triangles);
Fix memory leak with mesh cache
2017-05-16 11:45:23 +10:00
BATCH_DISCARD_SAFE(cache->overlay_triangles_nor);
Mesh Batch Cache: Split Edit Mesh mode Vertex Buffers for faster update.
2017-05-13 16:45:06 +02:00
BATCH_DISCARD_SAFE(cache->overlay_loose_verts);
BATCH_DISCARD_SAFE(cache->overlay_loose_edges);
Fix memory leak with mesh cache
2017-05-16 11:45:23 +10:00
BATCH_DISCARD_SAFE(cache->overlay_loose_edges_nor);
Mesh Batch Cache: Split Edit Mesh mode Vertex Buffers for faster update.
2017-05-13 16:45:06 +02:00
DWM: reuse position VBO for vpaint select overlay
2017-05-17 19:21:16 +10:00
BATCH_DISCARD_SAFE(cache->overlay_weight_faces);
Re-use vertex position VBO in weight-paint mode
2017-05-18 14:53:43 +10:00
BATCH_DISCARD_SAFE(cache->overlay_weight_verts);
DWM: reuse position VBO for vpaint select overlay
2017-05-17 19:21:16 +10:00
BATCH_DISCARD_ALL_SAFE(cache->overlay_paint_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);
DWM: Initial support vert & weight hidden faces
2017-05-17 16:01:18 +10:00
VERTEXBUFFER_DISCARD_SAFE(cache->pos_with_normals_visible_only);
DWM: Share VBO positions & normals for mesh cache Select, weight & color cache included redundant position and normals. Use multiple VBO's per batch for more efficient storage.
2017-05-08 16:25:29 +10:00
VERTEXBUFFER_DISCARD_SAFE(cache->tri_aligned_vert_colors);
VERTEXBUFFER_DISCARD_SAFE(cache->tri_aligned_weights);
BATCH_DISCARD_SAFE(cache->triangles_with_weights);
BATCH_DISCARD_SAFE(cache->triangles_with_vert_colors);
VERTEXBUFFER_DISCARD_SAFE(cache->tri_aligned_select_id);
DwM: texture paint support & mask mode Uses workaround so material slots are used when neither blender-internal or cycles are enabled.
2017-05-22 23:31:46 +10:00
VERTEXBUFFER_DISCARD_SAFE(cache->tri_aligned_uv);
Use mesh draw cache for back-buffer selection Vertex/weight paint now work with core profile, resolves T51380.
2017-05-05 05:17:31 +10:00
BATCH_DISCARD_SAFE(cache->triangles_with_select_id);
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) {
Cleanup: rename ct -> len Also use 'defgroup' instead of 'index' for passing weight arg.
2017-05-04 23:35:50 +10:00
for (int i = 0; i < cache->mat_len; ++i) {
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
ELEMENTLIST_DISCARD_SAFE(cache->shaded_triangles_in_order[i]);
}
}
if (cache->shaded_triangles) {
Cleanup: rename ct -> len Also use 'defgroup' instead of 'index' for passing weight arg.
2017-05-04 23:35:50 +10:00
for (int i = 0; i < cache->mat_len; ++i) {
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_DISCARD_SAFE(cache->shaded_triangles[i]);
}
}
MEM_SAFE_FREE(cache->shaded_triangles_in_order);
MEM_SAFE_FREE(cache->shaded_triangles);
DwM: texture paint support & mask mode Uses workaround so material slots are used when neither blender-internal or cycles are enabled.
2017-05-22 23:31:46 +10:00
if (cache->texpaint_triangles) {
for (int i = 0; i < cache->mat_len; ++i) {
BATCH_DISCARD_SAFE(cache->texpaint_triangles[i]);
}
}
MEM_SAFE_FREE(cache->texpaint_triangles);
BATCH_DISCARD_SAFE(cache->texpaint_triangles_single);
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. */
DWM: share positions & normals in shading VBO.
2017-05-09 10:56:22 +10:00
static VertexBuffer *mesh_batch_cache_get_tri_shading_data(MeshRenderData *rdata, MeshBatchCache *cache)
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
{
BLI_assert(rdata->types & (MR_DATATYPE_VERT | MR_DATATYPE_LOOPTRI | MR_DATATYPE_LOOP | MR_DATATYPE_POLY));
Cleanup: minor edits
2017-05-22 16:20:54 +10:00
#define USE_COMP_MESH_DATA
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 (cache->shaded_triangles_data == NULL) {
DWM: share positions & normals in shading VBO.
2017-05-09 10:56:22 +10:00
unsigned int vidx = 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
Cleanup: move custom-data layers into a struct Changed because the values co-exist with active layers (Mesh.mloopuv, Mesh.mloopcol).
2017-05-22 17:21:22 +10:00
if (rdata->cd.layers.uv_len + rdata->cd.layers.vcol_len == 0) {
Mesh Batch Cache: get rid of the ORCO VBO data, and reconstruct it in shader. With only one MADD instruction we recover the orco data and reduce both the storage and the fetching cost of an attrib layer.
2017-05-15 16:12:51 +02:00
return NULL;
}
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 */
Cleanup: move custom-data layers into a struct Changed because the values co-exist with active layers (Mesh.mloopuv, Mesh.mloopcol).
2017-05-22 17:21:22 +10:00
unsigned int *uv_id = MEM_mallocN(sizeof(*uv_id) * rdata->cd.layers.uv_len, "UV attrib format");
unsigned int *vcol_id = MEM_mallocN(sizeof(*vcol_id) * rdata->cd.layers.vcol_len, "Vcol attrib format");
unsigned int *tangent_id = MEM_mallocN(sizeof(*tangent_id) * rdata->cd.layers.uv_len, "Tangent attrib format");
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
Cleanup: move custom-data layers into a struct Changed because the values co-exist with active layers (Mesh.mloopuv, Mesh.mloopcol).
2017-05-22 17:21:22 +10:00
for (int i = 0; i < rdata->cd.layers.uv_len; i++) {
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
/* UV */
Cleanup: rename layer-names -> layer-uuid These names are created from a hash, don't confuse with actual layer names.
2017-05-09 10:38:52 +10:00
attrib_name = mesh_render_data_uv_layer_uuid_get(rdata, i);
Mesh Batch Cache: Use compressed format. We don't compress UV by default because that would mean restricting them to [0, 1] Leaving all other code for later settings switch.
2017-05-15 17:55:02 +02:00
#if defined(USE_COMP_MESH_DATA) && 0 /* these are clamped. Maybe use them as an option in the future */
Mesh Batch Cache: Define Compressed format for shading data. Deactivated by default. All shading attribs can be packed to take less VRAM at the cost of precision (not noticable in this case). UVs can be packed into I16 but that limits their positions into the [-1, +1] range. This could be a setting option in the future.
2017-05-14 16:13:00 +02:00
uv_id[i] = VertexFormat_add_attrib(format, attrib_name, COMP_I16, 2, NORMALIZE_INT_TO_FLOAT);
#else
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
uv_id[i] = VertexFormat_add_attrib(format, attrib_name, COMP_F32, 2, KEEP_FLOAT);
Mesh Batch Cache: Define Compressed format for shading data. Deactivated by default. All shading attribs can be packed to take less VRAM at the cost of precision (not noticable in this case). UVs can be packed into I16 but that limits their positions into the [-1, +1] range. This could be a setting option in the future.
2017-05-14 16:13:00 +02:00
#endif
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
/* Auto Name */
Cleanup: rename layer-names -> layer-uuid These names are created from a hash, don't confuse with actual layer names.
2017-05-09 10:38:52 +10:00
attrib_name = mesh_render_data_uv_auto_layer_uuid_get(rdata, i);
Mesh Batch Cache: Optimization, make use of gawain attrib aliases.
2017-05-13 22:53:11 +02:00
VertexFormat_add_alias(format, attrib_name);
Cleanup: move custom-data layers into a struct Changed because the values co-exist with active layers (Mesh.mloopuv, Mesh.mloopcol).
2017-05-22 17:21:22 +10:00
if (i == rdata->cd.layers.uv_active) {
Mesh Batch Cache: Optimization, make use of gawain attrib aliases.
2017-05-13 22:53:11 +02:00
VertexFormat_add_alias(format, "u");
}
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
/* Tangent */
Cleanup: rename layer-names -> layer-uuid These names are created from a hash, don't confuse with actual layer names.
2017-05-09 10:38:52 +10:00
attrib_name = mesh_render_data_tangent_layer_uuid_get(rdata, i);
GPUMaterial: Add support for tangent node.
2017-05-18 14:29:57 +02:00
/* WATCH IT : only specifying 3 component instead of 4 (4th is sign).
* That may cause some problem but I could not make it to fail (fclem) */
Mesh Batch Cache: Define Compressed format for shading data. Deactivated by default. All shading attribs can be packed to take less VRAM at the cost of precision (not noticable in this case). UVs can be packed into I16 but that limits their positions into the [-1, +1] range. This could be a setting option in the future.
2017-05-14 16:13:00 +02:00
#ifdef USE_COMP_MESH_DATA
Gawain: promote 10_10_10 to first-class vertex format This format is part of OpenGL 3.3, and one of the reasons for choosing 3.3 over 3.2. Instead of checking #if USE_10_10_10 just use it wherever needed.
2017-05-21 15:59:34 -04:00
/* Tangents need more precision than 10_10_10 */
Mesh Batch Cache: Define Compressed format for shading data. Deactivated by default. All shading attribs can be packed to take less VRAM at the cost of precision (not noticable in this case). UVs can be packed into I16 but that limits their positions into the [-1, +1] range. This could be a setting option in the future.
2017-05-14 16:13:00 +02:00
tangent_id[i] = VertexFormat_add_attrib(format, attrib_name, COMP_I16, 3, NORMALIZE_INT_TO_FLOAT);
#else
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
tangent_id[i] = VertexFormat_add_attrib(format, attrib_name, COMP_F32, 3, KEEP_FLOAT);
Mesh Batch Cache: Define Compressed format for shading data. Deactivated by default. All shading attribs can be packed to take less VRAM at the cost of precision (not noticable in this case). UVs can be packed into I16 but that limits their positions into the [-1, +1] range. This could be a setting option in the future.
2017-05-14 16:13:00 +02:00
#endif
Mesh Batch Cache: Optimization, make use of gawain attrib aliases.
2017-05-13 22:53:11 +02:00
Cleanup: move custom-data layers into a struct Changed because the values co-exist with active layers (Mesh.mloopuv, Mesh.mloopcol).
2017-05-22 17:21:22 +10:00
if (i == rdata->cd.layers.uv_active) {
Mesh Batch Cache: Optimization, make use of gawain attrib aliases.
2017-05-13 22:53:11 +02:00
VertexFormat_add_alias(format, "t");
}
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
}
Cleanup: move custom-data layers into a struct Changed because the values co-exist with active layers (Mesh.mloopuv, Mesh.mloopcol).
2017-05-22 17:21:22 +10:00
for (int i = 0; i < rdata->cd.layers.vcol_len; i++) {
Cleanup: rename layer-names -> layer-uuid These names are created from a hash, don't confuse with actual layer names.
2017-05-09 10:38:52 +10:00
attrib_name = mesh_render_data_vcol_layer_uuid_get(rdata, i);
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
vcol_id[i] = VertexFormat_add_attrib(format, attrib_name, COMP_U8, 3, NORMALIZE_INT_TO_FLOAT);
/* Auto layer */
Cleanup: move custom-data layers into a struct Changed because the values co-exist with active layers (Mesh.mloopuv, Mesh.mloopcol).
2017-05-22 17:21:22 +10:00
if (rdata->cd.layers.auto_vcol[i]) {
Cleanup: rename layer-names -> layer-uuid These names are created from a hash, don't confuse with actual layer names.
2017-05-09 10:38:52 +10:00
attrib_name = mesh_render_data_vcol_auto_layer_uuid_get(rdata, i);
Mesh Batch Cache: Optimization, make use of gawain attrib aliases.
2017-05-13 22:53:11 +02:00
VertexFormat_add_alias(format, attrib_name);
}
Cleanup: move custom-data layers into a struct Changed because the values co-exist with active layers (Mesh.mloopuv, Mesh.mloopcol).
2017-05-22 17:21:22 +10:00
if (i == rdata->cd.layers.vcol_active) {
Mesh Batch Cache: Optimization, make use of gawain attrib aliases.
2017-05-13 22:53:11 +02:00
VertexFormat_add_alias(format, "c");
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
}
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);
Cleanup: use more generic naming Can be used by any user that needs faces split by materials.
2017-05-23 17:48:05 +10:00
/* TODO deduplicate all verts and make use of ElementList in
* mesh_batch_cache_get_triangles_in_order_split_by_material. */
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
for (int i = 0; i < tri_len; i++) {
Mesh Batch Cache: get rid of the ORCO VBO data, and reconstruct it in shader. With only one MADD instruction we recover the orco data and reduce both the storage and the fetching cost of an attrib layer.
2017-05-15 16:12:51 +02:00
float *tri_uvs[3], *tri_tans[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
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
DWM: share positions & normals in shading VBO.
2017-05-09 10:56:22 +10:00
if (rdata->edit_bmesh == NULL ||
BM_elem_flag_test((rdata->edit_bmesh->looptris[i])[0]->f, BM_ELEM_HIDDEN) == 0)
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
/* UVs & TANGENTs */
Cleanup: move custom-data layers into a struct Changed because the values co-exist with active layers (Mesh.mloopuv, Mesh.mloopcol).
2017-05-22 17:21:22 +10:00
for (int j = 0; j < rdata->cd.layers.uv_len; j++) {
Mesh Batch Cache: Define Compressed format for shading data. Deactivated by default. All shading attribs can be packed to take less VRAM at the cost of precision (not noticable in this case). UVs can be packed into I16 but that limits their positions into the [-1, +1] range. This could be a setting option in the future.
2017-05-14 16:13:00 +02:00
/* UVs */
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
mesh_render_data_looptri_uvs_get(rdata, i, j, &tri_uvs);
Mesh Batch Cache: Use compressed format. We don't compress UV by default because that would mean restricting them to [0, 1] Leaving all other code for later settings switch.
2017-05-15 17:55:02 +02:00
#if defined(USE_COMP_MESH_DATA) && 0 /* these are clamped. Maybe use them as an option in the future */
Mesh Batch Cache: Define Compressed format for shading data. Deactivated by default. All shading attribs can be packed to take less VRAM at the cost of precision (not noticable in this case). UVs can be packed into I16 but that limits their positions into the [-1, +1] range. This could be a setting option in the future.
2017-05-14 16:13:00 +02:00
short s_uvs[3][2];
normal_float_to_short_v2(s_uvs[0], tri_uvs[0]);
normal_float_to_short_v2(s_uvs[1], tri_uvs[1]);
normal_float_to_short_v2(s_uvs[2], tri_uvs[2]);
#else
float **s_uvs = tri_uvs;
#endif
VertexBuffer_set_attrib(vbo, uv_id[j], vidx + 0, s_uvs[0]);
VertexBuffer_set_attrib(vbo, uv_id[j], vidx + 1, s_uvs[1]);
VertexBuffer_set_attrib(vbo, uv_id[j], vidx + 2, s_uvs[2]);
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
Mesh Batch Cache: Define Compressed format for shading data. Deactivated by default. All shading attribs can be packed to take less VRAM at the cost of precision (not noticable in this case). UVs can be packed into I16 but that limits their positions into the [-1, +1] range. This could be a setting option in the future.
2017-05-14 16:13:00 +02:00
/* Tangent */
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
mesh_render_data_looptri_tans_get(rdata, i, j, &tri_tans);
Mesh Batch Cache: Define Compressed format for shading data. Deactivated by default. All shading attribs can be packed to take less VRAM at the cost of precision (not noticable in this case). UVs can be packed into I16 but that limits their positions into the [-1, +1] range. This could be a setting option in the future.
2017-05-14 16:13:00 +02:00
#ifdef USE_COMP_MESH_DATA
Gawain: promote 10_10_10 to first-class vertex format This format is part of OpenGL 3.3, and one of the reasons for choosing 3.3 over 3.2. Instead of checking #if USE_10_10_10 just use it wherever needed.
2017-05-21 15:59:34 -04:00
/* Tangents need more precision than 10_10_10 */
Mesh Batch Cache: Define Compressed format for shading data. Deactivated by default. All shading attribs can be packed to take less VRAM at the cost of precision (not noticable in this case). UVs can be packed into I16 but that limits their positions into the [-1, +1] range. This could be a setting option in the future.
2017-05-14 16:13:00 +02:00
short s_tan[3][3];
normal_float_to_short_v3(s_tan[0], tri_tans[0]);
normal_float_to_short_v3(s_tan[1], tri_tans[1]);
normal_float_to_short_v3(s_tan[2], tri_tans[2]);
#else
float **s_tan = tri_tans;
#endif
VertexBuffer_set_attrib(vbo, tangent_id[j], vidx + 0, s_tan[0]);
VertexBuffer_set_attrib(vbo, tangent_id[j], vidx + 1, s_tan[1]);
VertexBuffer_set_attrib(vbo, tangent_id[j], vidx + 2, s_tan[2]);
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
}
/* VCOLs */
Cleanup: move custom-data layers into a struct Changed because the values co-exist with active layers (Mesh.mloopuv, Mesh.mloopcol).
2017-05-22 17:21:22 +10:00
for (int j = 0; j < rdata->cd.layers.vcol_len; j++) {
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
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]);
}
Mesh Batch Cache: get rid of the ORCO VBO data, and reconstruct it in shader. With only one MADD instruction we recover the orco data and reduce both the storage and the fetching cost of an attrib layer.
2017-05-15 16:12:51 +02:00
vidx += 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
}
}
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(vcol_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
}
Cleanup: minor edits
2017-05-22 16:20:54 +10:00
#undef USE_COMP_MESH_DATA
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;
}
DwM: texture paint support & mask mode Uses workaround so material slots are used when neither blender-internal or cycles are enabled.
2017-05-22 23:31:46 +10:00
static VertexBuffer *mesh_batch_cache_get_tri_uv_active(
MeshRenderData *rdata, MeshBatchCache *cache)
{
BLI_assert(rdata->types & (MR_DATATYPE_VERT | MR_DATATYPE_LOOPTRI | MR_DATATYPE_LOOP | MR_DATATYPE_LOOPUV));
BLI_assert(rdata->edit_bmesh == NULL);
if (cache->tri_aligned_uv == NULL) {
unsigned int vidx = 0;
static VertexFormat format = { 0 };
static struct { uint uv; } attr_id;
if (format.attrib_ct == 0) {
attr_id.uv = VertexFormat_add_attrib(&format, "uv", COMP_F32, 2, KEEP_FLOAT);
}
const int tri_len = mesh_render_data_looptri_len_get(rdata);
VertexBuffer *vbo = cache->tri_aligned_uv = VertexBuffer_create_with_format(&format);
const int vbo_len_capacity = tri_len * 3;
int vbo_len_used = 0;
VertexBuffer_allocate_data(vbo, vbo_len_capacity);
const MLoopUV *mloopuv = rdata->mloopuv;
for (int i = 0; i < tri_len; i++) {
const MLoopTri *mlt = &rdata->mlooptri[i];
VertexBuffer_set_attrib(vbo, attr_id.uv, vidx++, mloopuv[mlt->tri[0]].uv);
VertexBuffer_set_attrib(vbo, attr_id.uv, vidx++, mloopuv[mlt->tri[1]].uv);
VertexBuffer_set_attrib(vbo, attr_id.uv, vidx++, mloopuv[mlt->tri[2]].uv);
}
vbo_len_used = vidx;
BLI_assert(vbo_len_capacity == vbo_len_used);
}
return cache->tri_aligned_uv;
}
DWM: Initial support vert & weight hidden faces
2017-05-17 16:01:18 +10:00
static VertexBuffer *mesh_batch_cache_get_tri_pos_and_normals_ex(
MeshRenderData *rdata, const bool use_hide,
VertexBuffer **r_vbo)
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
DWM: Initial support vert & weight hidden faces
2017-05-17 16:01:18 +10:00
if (*r_vbo == NULL) {
Edit Mesh overlay: Ported Display Normals option
2017-03-09 01:29:58 +01:00
unsigned int vidx = 0, nidx = 0;
static VertexFormat format = { 0 };
Cleanup: group VBO attributes in a struct Some names are a bit arbitrary, this makes it clear which names are VBO attributes.
2017-05-17 12:22:22 +10:00
static struct { uint pos, nor; } attr_id;
Edit Mesh overlay: Ported Display Normals option
2017-03-09 01:29:58 +01:00
if (format.attrib_ct == 0) {
Cleanup: group VBO attributes in a struct Some names are a bit arbitrary, this makes it clear which names are VBO attributes.
2017-05-17 12:22:22 +10:00
attr_id.pos = VertexFormat_add_attrib(&format, "pos", COMP_F32, 3, KEEP_FLOAT);
attr_id.nor = VertexFormat_add_attrib(&format, "nor", COMP_I10, 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
DWM: Initial support vert & weight hidden faces
2017-05-17 16:01:18 +10:00
VertexBuffer *vbo = *r_vbo = 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(
DWM: Initial support vert & weight hidden faces
2017-05-17 16:01:18 +10:00
rdata, i, use_hide, &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) {
Mesh Batch Cache: Use compressed format. We don't compress UV by default because that would mean restricting them to [0, 1] Leaving all other code for later settings switch.
2017-05-15 17:55:02 +02:00
PackedNormal snor_pack[3] = {
Mesh Batch Cache: Define Compressed format for shading data. Deactivated by default. All shading attribs can be packed to take less VRAM at the cost of precision (not noticable in this case). UVs can be packed into I16 but that limits their positions into the [-1, +1] range. This could be a setting option in the future.
2017-05-14 16:13:00 +02:00
convert_i10_s3(tri_vert_nors[0]),
convert_i10_s3(tri_vert_nors[1]),
convert_i10_s3(tri_vert_nors[2])
Mesh Batch Cache: Use compressed format. We don't compress UV by default because that would mean restricting them to [0, 1] Leaving all other code for later settings switch.
2017-05-15 17:55:02 +02:00
};
PackedNormal *snor[3] = { &snor_pack[0], &snor_pack[1], &snor_pack[2] };
Gawain: promote 10_10_10 to first-class vertex format This format is part of OpenGL 3.3, and one of the reasons for choosing 3.3 over 3.2. Instead of checking #if USE_10_10_10 just use it wherever needed.
2017-05-21 15:59:34 -04:00
Cleanup: group VBO attributes in a struct Some names are a bit arbitrary, this makes it clear which names are VBO attributes.
2017-05-17 12:22:22 +10:00
VertexBuffer_set_attrib(vbo, attr_id.nor, nidx++, snor[0]);
VertexBuffer_set_attrib(vbo, attr_id.nor, nidx++, snor[1]);
VertexBuffer_set_attrib(vbo, attr_id.nor, nidx++, snor[2]);
Draw Manager: T51218 Support edit-mesh hide
2017-04-18 23:29:55 +10:00
}
else {
Mesh Batch Cache: Use compressed format. We don't compress UV by default because that would mean restricting them to [0, 1] Leaving all other code for later settings switch.
2017-05-15 17:55:02 +02:00
PackedNormal snor_pack = convert_i10_s3(tri_nor);
PackedNormal *snor = &snor_pack;
Gawain: promote 10_10_10 to first-class vertex format This format is part of OpenGL 3.3, and one of the reasons for choosing 3.3 over 3.2. Instead of checking #if USE_10_10_10 just use it wherever needed.
2017-05-21 15:59:34 -04:00
Cleanup: group VBO attributes in a struct Some names are a bit arbitrary, this makes it clear which names are VBO attributes.
2017-05-17 12:22:22 +10:00
VertexBuffer_set_attrib(vbo, attr_id.nor, nidx++, snor);
VertexBuffer_set_attrib(vbo, attr_id.nor, nidx++, snor);
VertexBuffer_set_attrib(vbo, attr_id.nor, nidx++, snor);
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: group VBO attributes in a struct Some names are a bit arbitrary, this makes it clear which names are VBO attributes.
2017-05-17 12:22:22 +10:00
VertexBuffer_set_attrib(vbo, attr_id.pos, vidx++, tri_vert_cos[0]);
VertexBuffer_set_attrib(vbo, attr_id.pos, vidx++, tri_vert_cos[1]);
VertexBuffer_set_attrib(vbo, attr_id.pos, 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
}
}
DWM: Initial support vert & weight hidden faces
2017-05-17 16:01:18 +10:00
return *r_vbo;
}
static VertexBuffer *mesh_batch_cache_get_tri_pos_and_normals(
MeshRenderData *rdata, MeshBatchCache *cache)
{
return mesh_batch_cache_get_tri_pos_and_normals_ex(
rdata, false,
&cache->pos_with_normals);
}
static VertexBuffer *mesh_batch_cache_get_tri_pos_and_normals_visible_only(
MeshRenderData *rdata, MeshBatchCache *cache)
{
return mesh_batch_cache_get_tri_pos_and_normals_ex(
rdata, true,
&cache->pos_with_normals_visible_only);
Edit Mesh overlay: Ported Display Normals option
2017-03-09 01:29:58 +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
DWM: Share VBO positions & normals for mesh cache Select, weight & color cache included redundant position and normals. Use multiple VBO's per batch for more efficient storage.
2017-05-08 16:25:29 +10:00
static VertexBuffer *mesh_batch_cache_get_tri_weights(
DWM: Initial support vert & weight hidden faces
2017-05-17 16:01:18 +10:00
MeshRenderData *rdata, MeshBatchCache *cache, bool use_hide, int defgroup)
Weight painting with draw manager This implements weight rendering with the draw manager, with all drawing options (Shading, wire, face masking, vertex masking). This is part of T51208 Reviewers: campbellbarton Subscribers: dfelinto Differential Revision: https://developer.blender.org/D2654
2017-05-03 18:55:40 +02:00
{
Cleanup: line length
2017-05-04 03:35:01 +10:00
BLI_assert(
rdata->types &
(MR_DATATYPE_VERT | MR_DATATYPE_LOOPTRI | MR_DATATYPE_LOOP | MR_DATATYPE_POLY | MR_DATATYPE_DVERT));
Weight painting with draw manager This implements weight rendering with the draw manager, with all drawing options (Shading, wire, face masking, vertex masking). This is part of T51208 Reviewers: campbellbarton Subscribers: dfelinto Differential Revision: https://developer.blender.org/D2654
2017-05-03 18:55:40 +02:00
DWM: Share VBO positions & normals for mesh cache Select, weight & color cache included redundant position and normals. Use multiple VBO's per batch for more efficient storage.
2017-05-08 16:25:29 +10:00
if (cache->tri_aligned_weights == NULL) {
unsigned int cidx = 0;
Weight painting with draw manager This implements weight rendering with the draw manager, with all drawing options (Shading, wire, face masking, vertex masking). This is part of T51208 Reviewers: campbellbarton Subscribers: dfelinto Differential Revision: https://developer.blender.org/D2654
2017-05-03 18:55:40 +02:00
static VertexFormat format = { 0 };
Cleanup: group VBO attributes in a struct Some names are a bit arbitrary, this makes it clear which names are VBO attributes.
2017-05-17 12:22:22 +10:00
static struct { uint col; } attr_id;
Weight painting with draw manager This implements weight rendering with the draw manager, with all drawing options (Shading, wire, face masking, vertex masking). This is part of T51208 Reviewers: campbellbarton Subscribers: dfelinto Differential Revision: https://developer.blender.org/D2654
2017-05-03 18:55:40 +02:00
if (format.attrib_ct == 0) {
Cleanup: group VBO attributes in a struct Some names are a bit arbitrary, this makes it clear which names are VBO attributes.
2017-05-17 12:22:22 +10:00
attr_id.col = VertexFormat_add_attrib(&format, "color", COMP_F32, 3, KEEP_FLOAT);
Weight painting with draw manager This implements weight rendering with the draw manager, with all drawing options (Shading, wire, face masking, vertex masking). This is part of T51208 Reviewers: campbellbarton Subscribers: dfelinto Differential Revision: https://developer.blender.org/D2654
2017-05-03 18:55:40 +02:00
}
const int tri_len = mesh_render_data_looptri_len_get(rdata);
DWM: Share VBO positions & normals for mesh cache Select, weight & color cache included redundant position and normals. Use multiple VBO's per batch for more efficient storage.
2017-05-08 16:25:29 +10:00
VertexBuffer *vbo = cache->tri_aligned_weights = VertexBuffer_create_with_format(&format);
Weight painting with draw manager This implements weight rendering with the draw manager, with all drawing options (Shading, wire, face masking, vertex masking). This is part of T51208 Reviewers: campbellbarton Subscribers: dfelinto Differential Revision: https://developer.blender.org/D2654
2017-05-03 18:55:40 +02:00
const int vbo_len_capacity = tri_len * 3;
int vbo_len_used = 0;
VertexBuffer_allocate_data(vbo, vbo_len_capacity);
DWM: write mesh data to the VBO directly Was using intermediate functions to store each data type. This is still done in some places (for more involved access). For basic data: coords, normals, colors..etc this is unnecessarily indirect.
2017-05-17 11:28:44 +10:00
mesh_render_data_ensure_vert_weight_color(rdata, defgroup);
const float (*vert_weight_color)[3] = rdata->vert_weight_color;
Weight painting with draw manager This implements weight rendering with the draw manager, with all drawing options (Shading, wire, face masking, vertex masking). This is part of T51208 Reviewers: campbellbarton Subscribers: dfelinto Differential Revision: https://developer.blender.org/D2654
2017-05-03 18:55:40 +02:00
DWM: write mesh data to the VBO directly Was using intermediate functions to store each data type. This is still done in some places (for more involved access). For basic data: coords, normals, colors..etc this is unnecessarily indirect.
2017-05-17 11:28:44 +10:00
if (rdata->edit_bmesh) {
for (int i = 0; i < tri_len; i++) {
const BMLoop **ltri = (const BMLoop **)rdata->edit_bmesh->looptris[i];
DWM: Initial support vert & weight hidden faces
2017-05-17 16:01:18 +10:00
/* Assume 'use_hide' */
DWM: write mesh data to the VBO directly Was using intermediate functions to store each data type. This is still done in some places (for more involved access). For basic data: coords, normals, colors..etc this is unnecessarily indirect.
2017-05-17 11:28:44 +10:00
if (!BM_elem_flag_test(ltri[0]->f, BM_ELEM_HIDDEN)) {
for (uint tri_corner = 0; tri_corner < 3; tri_corner++) {
const int v_index = BM_elem_index_get(ltri[tri_corner]->v);
Cleanup: group VBO attributes in a struct Some names are a bit arbitrary, this makes it clear which names are VBO attributes.
2017-05-17 12:22:22 +10:00
VertexBuffer_set_attrib(vbo, attr_id.col, cidx++, vert_weight_color[v_index]);
DWM: write mesh data to the VBO directly Was using intermediate functions to store each data type. This is still done in some places (for more involved access). For basic data: coords, normals, colors..etc this is unnecessarily indirect.
2017-05-17 11:28:44 +10:00
}
}
}
}
else {
for (int i = 0; i < tri_len; i++) {
const MLoopTri *mlt = &rdata->mlooptri[i];
DWM: Initial support vert & weight hidden faces
2017-05-17 16:01:18 +10:00
if (!(use_hide && (rdata->mpoly[mlt->poly].flag & ME_HIDE))) {
for (uint tri_corner = 0; tri_corner < 3; tri_corner++) {
const uint v_index = rdata->mloop[mlt->tri[tri_corner]].v;
VertexBuffer_set_attrib(vbo, attr_id.col, cidx++, vert_weight_color[v_index]);
}
DWM: write mesh data to the VBO directly Was using intermediate functions to store each data type. This is still done in some places (for more involved access). For basic data: coords, normals, colors..etc this is unnecessarily indirect.
2017-05-17 11:28:44 +10:00
}
Weight painting with draw manager This implements weight rendering with the draw manager, with all drawing options (Shading, wire, face masking, vertex masking). This is part of T51208 Reviewers: campbellbarton Subscribers: dfelinto Differential Revision: https://developer.blender.org/D2654
2017-05-03 18:55:40 +02:00
}
}
DWM: Share VBO positions & normals for mesh cache Select, weight & color cache included redundant position and normals. Use multiple VBO's per batch for more efficient storage.
2017-05-08 16:25:29 +10:00
vbo_len_used = cidx;
Weight painting with draw manager This implements weight rendering with the draw manager, with all drawing options (Shading, wire, face masking, vertex masking). This is part of T51208 Reviewers: campbellbarton Subscribers: dfelinto Differential Revision: https://developer.blender.org/D2654
2017-05-03 18:55:40 +02:00
if (vbo_len_capacity != vbo_len_used) {
VertexBuffer_resize_data(vbo, vbo_len_used);
}
}
DWM: Share VBO positions & normals for mesh cache Select, weight & color cache included redundant position and normals. Use multiple VBO's per batch for more efficient storage.
2017-05-08 16:25:29 +10:00
return cache->tri_aligned_weights;
Weight painting with draw manager This implements weight rendering with the draw manager, with all drawing options (Shading, wire, face masking, vertex masking). This is part of T51208 Reviewers: campbellbarton Subscribers: dfelinto Differential Revision: https://developer.blender.org/D2654
2017-05-03 18:55:40 +02:00
}
DWM: Share VBO positions & normals for mesh cache Select, weight & color cache included redundant position and normals. Use multiple VBO's per batch for more efficient storage.
2017-05-08 16:25:29 +10:00
static VertexBuffer *mesh_batch_cache_get_tri_vert_colors(
DWM: Initial support vert & weight hidden faces
2017-05-17 16:01:18 +10:00
MeshRenderData *rdata, MeshBatchCache *cache, bool use_hide)
Vertex paint with draw manager Reviewers: fclem Subscribers: campbellbarton, dfelinto Differential Revision: https://developer.blender.org/D2658
2017-05-04 19:59:27 +02:00
{
BLI_assert(
rdata->types &
(MR_DATATYPE_VERT | MR_DATATYPE_LOOPTRI | MR_DATATYPE_LOOP | MR_DATATYPE_POLY | MR_DATATYPE_LOOPCOL));
DWM: Share VBO positions & normals for mesh cache Select, weight & color cache included redundant position and normals. Use multiple VBO's per batch for more efficient storage.
2017-05-08 16:25:29 +10:00
if (cache->tri_aligned_vert_colors == NULL) {
unsigned int cidx = 0;
Vertex paint with draw manager Reviewers: fclem Subscribers: campbellbarton, dfelinto Differential Revision: https://developer.blender.org/D2658
2017-05-04 19:59:27 +02:00
static VertexFormat format = { 0 };
Cleanup: group VBO attributes in a struct Some names are a bit arbitrary, this makes it clear which names are VBO attributes.
2017-05-17 12:22:22 +10:00
static struct { uint col; } attr_id;
Vertex paint with draw manager Reviewers: fclem Subscribers: campbellbarton, dfelinto Differential Revision: https://developer.blender.org/D2658
2017-05-04 19:59:27 +02:00
if (format.attrib_ct == 0) {
Cleanup: group VBO attributes in a struct Some names are a bit arbitrary, this makes it clear which names are VBO attributes.
2017-05-17 12:22:22 +10:00
attr_id.col = VertexFormat_add_attrib(&format, "color", COMP_U8, 3, NORMALIZE_INT_TO_FLOAT);
Vertex paint with draw manager Reviewers: fclem Subscribers: campbellbarton, dfelinto Differential Revision: https://developer.blender.org/D2658
2017-05-04 19:59:27 +02:00
}
const int tri_len = mesh_render_data_looptri_len_get(rdata);
DWM: Share VBO positions & normals for mesh cache Select, weight & color cache included redundant position and normals. Use multiple VBO's per batch for more efficient storage.
2017-05-08 16:25:29 +10:00
VertexBuffer *vbo = cache->tri_aligned_vert_colors = VertexBuffer_create_with_format(&format);
Vertex paint with draw manager Reviewers: fclem Subscribers: campbellbarton, dfelinto Differential Revision: https://developer.blender.org/D2658
2017-05-04 19:59:27 +02:00
DWM: Share VBO positions & normals for mesh cache Select, weight & color cache included redundant position and normals. Use multiple VBO's per batch for more efficient storage.
2017-05-08 16:25:29 +10:00
const uint vbo_len_capacity = tri_len * 3;
Vertex paint with draw manager Reviewers: fclem Subscribers: campbellbarton, dfelinto Differential Revision: https://developer.blender.org/D2658
2017-05-04 19:59:27 +02:00
VertexBuffer_allocate_data(vbo, vbo_len_capacity);
DWM: write mesh data to the VBO directly Was using intermediate functions to store each data type. This is still done in some places (for more involved access). For basic data: coords, normals, colors..etc this is unnecessarily indirect.
2017-05-17 11:28:44 +10:00
mesh_render_data_ensure_vert_color(rdata);
const char (*vert_color)[3] = rdata->vert_color;
Vertex paint with draw manager Reviewers: fclem Subscribers: campbellbarton, dfelinto Differential Revision: https://developer.blender.org/D2658
2017-05-04 19:59:27 +02:00
DWM: write mesh data to the VBO directly Was using intermediate functions to store each data type. This is still done in some places (for more involved access). For basic data: coords, normals, colors..etc this is unnecessarily indirect.
2017-05-17 11:28:44 +10:00
if (rdata->edit_bmesh) {
for (int i = 0; i < tri_len; i++) {
const BMLoop **ltri = (const BMLoop **)rdata->edit_bmesh->looptris[i];
DWM: Initial support vert & weight hidden faces
2017-05-17 16:01:18 +10:00
/* Assume 'use_hide' */
DWM: write mesh data to the VBO directly Was using intermediate functions to store each data type. This is still done in some places (for more involved access). For basic data: coords, normals, colors..etc this is unnecessarily indirect.
2017-05-17 11:28:44 +10:00
if (!BM_elem_flag_test(ltri[0]->f, BM_ELEM_HIDDEN)) {
for (uint tri_corner = 0; tri_corner < 3; tri_corner++) {
Correct own error in copying vertex color to VBO's
2017-05-17 17:14:10 +10:00
const int l_index = BM_elem_index_get(ltri[tri_corner]);
VertexBuffer_set_attrib(vbo, attr_id.col, cidx++, vert_color[l_index]);
DWM: write mesh data to the VBO directly Was using intermediate functions to store each data type. This is still done in some places (for more involved access). For basic data: coords, normals, colors..etc this is unnecessarily indirect.
2017-05-17 11:28:44 +10:00
}
}
}
}
else {
for (int i = 0; i < tri_len; i++) {
const MLoopTri *mlt = &rdata->mlooptri[i];
DWM: Initial support vert & weight hidden faces
2017-05-17 16:01:18 +10:00
if (!(use_hide && (rdata->mpoly[mlt->poly].flag & ME_HIDE))) {
for (uint tri_corner = 0; tri_corner < 3; tri_corner++) {
Correct own error in copying vertex color to VBO's
2017-05-17 17:14:10 +10:00
const uint l_index = mlt->tri[tri_corner];
VertexBuffer_set_attrib(vbo, attr_id.col, cidx++, vert_color[l_index]);
DWM: Initial support vert & weight hidden faces
2017-05-17 16:01:18 +10:00
}
DWM: write mesh data to the VBO directly Was using intermediate functions to store each data type. This is still done in some places (for more involved access). For basic data: coords, normals, colors..etc this is unnecessarily indirect.
2017-05-17 11:28:44 +10:00
}
Vertex paint with draw manager Reviewers: fclem Subscribers: campbellbarton, dfelinto Differential Revision: https://developer.blender.org/D2658
2017-05-04 19:59:27 +02:00
}
}
DWM: Share VBO positions & normals for mesh cache Select, weight & color cache included redundant position and normals. Use multiple VBO's per batch for more efficient storage.
2017-05-08 16:25:29 +10:00
const uint vbo_len_used = cidx;
Vertex paint with draw manager Reviewers: fclem Subscribers: campbellbarton, dfelinto Differential Revision: https://developer.blender.org/D2658
2017-05-04 19:59:27 +02:00
if (vbo_len_capacity != vbo_len_used) {
VertexBuffer_resize_data(vbo, vbo_len_used);
}
}
DWM: Share VBO positions & normals for mesh cache Select, weight & color cache included redundant position and normals. Use multiple VBO's per batch for more efficient storage.
2017-05-08 16:25:29 +10:00
return cache->tri_aligned_vert_colors;
Vertex paint with draw manager Reviewers: fclem Subscribers: campbellbarton, dfelinto Differential Revision: https://developer.blender.org/D2658
2017-05-04 19:59:27 +02:00
}
DWM: Share VBO positions & normals for mesh cache Select, weight & color cache included redundant position and normals. Use multiple VBO's per batch for more efficient storage.
2017-05-08 16:25:29 +10:00
static VertexBuffer *mesh_batch_cache_get_tri_select_id(
Use mesh draw cache for back-buffer selection Vertex/weight paint now work with core profile, resolves T51380.
2017-05-05 05:17:31 +10:00
MeshRenderData *rdata, MeshBatchCache *cache, bool use_hide)
{
BLI_assert(
rdata->types &
(MR_DATATYPE_VERT | MR_DATATYPE_LOOPTRI | MR_DATATYPE_LOOP | MR_DATATYPE_POLY));
DWM: Share VBO positions & normals for mesh cache Select, weight & color cache included redundant position and normals. Use multiple VBO's per batch for more efficient storage.
2017-05-08 16:25:29 +10:00
if (cache->tri_aligned_select_id == NULL) {
unsigned int cidx = 0;
Use mesh draw cache for back-buffer selection Vertex/weight paint now work with core profile, resolves T51380.
2017-05-05 05:17:31 +10:00
static VertexFormat format = { 0 };
Cleanup: group VBO attributes in a struct Some names are a bit arbitrary, this makes it clear which names are VBO attributes.
2017-05-17 12:22:22 +10:00
static struct { uint col; } attr_id;
Use mesh draw cache for back-buffer selection Vertex/weight paint now work with core profile, resolves T51380.
2017-05-05 05:17:31 +10:00
if (format.attrib_ct == 0) {
Cleanup: group VBO attributes in a struct Some names are a bit arbitrary, this makes it clear which names are VBO attributes.
2017-05-17 12:22:22 +10:00
attr_id.col = VertexFormat_add_attrib(&format, "color", COMP_I32, 1, KEEP_INT);
Use mesh draw cache for back-buffer selection Vertex/weight paint now work with core profile, resolves T51380.
2017-05-05 05:17:31 +10:00
}
const int tri_len = mesh_render_data_looptri_len_get(rdata);
DWM: Share VBO positions & normals for mesh cache Select, weight & color cache included redundant position and normals. Use multiple VBO's per batch for more efficient storage.
2017-05-08 16:25:29 +10:00
VertexBuffer *vbo = cache->tri_aligned_select_id = VertexBuffer_create_with_format(&format);
Use mesh draw cache for back-buffer selection Vertex/weight paint now work with core profile, resolves T51380.
2017-05-05 05:17:31 +10:00
const int vbo_len_capacity = tri_len * 3;
int vbo_len_used = 0;
VertexBuffer_allocate_data(vbo, vbo_len_capacity);
DWM: write mesh data to the VBO directly Was using intermediate functions to store each data type. This is still done in some places (for more involved access). For basic data: coords, normals, colors..etc this is unnecessarily indirect.
2017-05-17 11:28:44 +10:00
if (rdata->edit_bmesh) {
for (int i = 0; i < tri_len; i++) {
const BMLoop **ltri = (const BMLoop **)rdata->edit_bmesh->looptris[i];
DWM: Initial support vert & weight hidden faces
2017-05-17 16:01:18 +10:00
/* Assume 'use_hide' */
if (!BM_elem_flag_test(ltri[0]->f, BM_ELEM_HIDDEN)) {
DWM: write mesh data to the VBO directly Was using intermediate functions to store each data type. This is still done in some places (for more involved access). For basic data: coords, normals, colors..etc this is unnecessarily indirect.
2017-05-17 11:28:44 +10:00
const int poly_index = BM_elem_index_get(ltri[0]->f);
int select_id;
GPU_select_index_get(poly_index + 1, &select_id);
for (uint tri_corner = 0; tri_corner < 3; tri_corner++) {
Cleanup: group VBO attributes in a struct Some names are a bit arbitrary, this makes it clear which names are VBO attributes.
2017-05-17 12:22:22 +10:00
VertexBuffer_set_attrib(vbo, attr_id.col, cidx++, &select_id);
DWM: write mesh data to the VBO directly Was using intermediate functions to store each data type. This is still done in some places (for more involved access). For basic data: coords, normals, colors..etc this is unnecessarily indirect.
2017-05-17 11:28:44 +10:00
}
}
}
}
else {
for (int i = 0; i < tri_len; i++) {
const MLoopTri *mlt = &rdata->mlooptri[i];
const int poly_index = mlt->poly;
if (!(use_hide && (rdata->mpoly[poly_index].flag & ME_HIDE))) {
int select_id;
GPU_select_index_get(poly_index + 1, &select_id);
for (uint tri_corner = 0; tri_corner < 3; tri_corner++) {
Cleanup: group VBO attributes in a struct Some names are a bit arbitrary, this makes it clear which names are VBO attributes.
2017-05-17 12:22:22 +10:00
VertexBuffer_set_attrib(vbo, attr_id.col, cidx++, &select_id);
DWM: write mesh data to the VBO directly Was using intermediate functions to store each data type. This is still done in some places (for more involved access). For basic data: coords, normals, colors..etc this is unnecessarily indirect.
2017-05-17 11:28:44 +10:00
}
}
Use mesh draw cache for back-buffer selection Vertex/weight paint now work with core profile, resolves T51380.
2017-05-05 05:17:31 +10:00
}
}
DWM: Share VBO positions & normals for mesh cache Select, weight & color cache included redundant position and normals. Use multiple VBO's per batch for more efficient storage.
2017-05-08 16:25:29 +10:00
vbo_len_used = cidx;
Use mesh draw cache for back-buffer selection Vertex/weight paint now work with core profile, resolves T51380.
2017-05-05 05:17:31 +10:00
if (vbo_len_capacity != vbo_len_used) {
VertexBuffer_resize_data(vbo, vbo_len_used);
}
}
DWM: Share VBO positions & normals for mesh cache Select, weight & color cache included redundant position and normals. Use multiple VBO's per batch for more efficient storage.
2017-05-08 16:25:29 +10:00
return cache->tri_aligned_select_id;
Use mesh draw cache for back-buffer selection Vertex/weight paint now work with core profile, resolves T51380.
2017-05-05 05:17:31 +10:00
}
Cleanup: include the type of data in function name Was confusing, some functions return vert or triangle data but naming wasn't very clear.
2017-05-05 05:41:30 +10:00
static VertexBuffer *mesh_batch_cache_get_vert_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 };
Cleanup: group VBO attributes in a struct Some names are a bit arbitrary, this makes it clear which names are VBO attributes.
2017-05-17 12:22:22 +10:00
static struct { uint pos, nor; } attr_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) {
Cleanup: group VBO attributes in a struct Some names are a bit arbitrary, this makes it clear which names are VBO attributes.
2017-05-17 12:22:22 +10:00
attr_id.pos = VertexFormat_add_attrib(&format, "pos", COMP_F32, 3, KEEP_FLOAT);
attr_id.nor = 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);
DWM: write mesh data to the VBO directly Was using intermediate functions to store each data type. This is still done in some places (for more involved access). For basic data: coords, normals, colors..etc this is unnecessarily indirect.
2017-05-17 11:28:44 +10:00
if (rdata->edit_bmesh) {
BMesh *bm = rdata->edit_bmesh->bm;
BMIter iter;
BMVert *eve;
uint i;
BM_ITER_MESH_INDEX (eve, &iter, bm, BM_VERTS_OF_MESH, i) {
static short no_short[3];
normal_float_to_short_v3(no_short, eve->no);
Cleanup: group VBO attributes in a struct Some names are a bit arbitrary, this makes it clear which names are VBO attributes.
2017-05-17 12:22:22 +10:00
VertexBuffer_set_attrib(vbo, attr_id.pos, i, eve->co);
VertexBuffer_set_attrib(vbo, attr_id.nor, i, no_short);
DWM: write mesh data to the VBO directly Was using intermediate functions to store each data type. This is still done in some places (for more involved access). For basic data: coords, normals, colors..etc this is unnecessarily indirect.
2017-05-17 11:28:44 +10:00
}
BLI_assert(i == vbo_len_capacity);
}
else {
for (int i = 0; i < vbo_len_capacity; ++i) {
Cleanup: group VBO attributes in a struct Some names are a bit arbitrary, this makes it clear which names are VBO attributes.
2017-05-17 12:22:22 +10:00
VertexBuffer_set_attrib(vbo, attr_id.pos, i, rdata->mvert[i].co);
VertexBuffer_set_attrib(vbo, attr_id.nor, i, rdata->mvert[i].no);
DWM: write mesh data to the VBO directly Was using intermediate functions to store each data type. This is still done in some places (for more involved access). For basic data: coords, normals, colors..etc this is unnecessarily indirect.
2017-05-17 11:28:44 +10:00
}
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
}
}
return cache->pos_in_order;
}
Mesh Batch Cache: Split Edit Mesh mode Vertex Buffers for faster update.
2017-05-13 16:45:06 +02:00
static VertexFormat *edit_mesh_overlay_pos_format(unsigned int *r_pos_id)
{
static VertexFormat format_pos = { 0 };
static unsigned pos_id;
if (format_pos.attrib_ct == 0) {
pos_id = VertexFormat_add_attrib(&format_pos, "pos", COMP_F32, 3, KEEP_FLOAT);
}
*r_pos_id = pos_id;
return &format_pos;
}
static VertexFormat *edit_mesh_overlay_nor_format(unsigned int *r_vnor_id, unsigned int *r_lnor_id)
{
static VertexFormat format_nor = { 0 };
static VertexFormat format_nor_loop = { 0 };
static unsigned vnor_id, vnor_loop_id, lnor_id;
if (format_nor.attrib_ct == 0) {
vnor_id = VertexFormat_add_attrib(&format_nor, "vnor", COMP_I10, 3, NORMALIZE_INT_TO_FLOAT);
vnor_loop_id = VertexFormat_add_attrib(&format_nor_loop, "vnor", COMP_I10, 3, NORMALIZE_INT_TO_FLOAT);
lnor_id = VertexFormat_add_attrib(&format_nor_loop, "lnor", COMP_I10, 3, NORMALIZE_INT_TO_FLOAT);
}
if (r_lnor_id) {
*r_vnor_id = vnor_loop_id;
*r_lnor_id = lnor_id;
return &format_nor_loop;
}
else {
*r_vnor_id = vnor_id;
return &format_nor;
}
}
static VertexFormat *edit_mesh_overlay_data_format(unsigned int *r_data_id)
{
static VertexFormat format_flag = { 0 };
static unsigned data_id;
if (format_flag.attrib_ct == 0) {
data_id = VertexFormat_add_attrib(&format_flag, "data", COMP_U8, 4, KEEP_INT);
}
*r_data_id = data_id;
return &format_flag;
}
static void mesh_batch_cache_create_overlay_tri_buffers(
MeshRenderData *rdata, MeshBatchCache *cache)
{
BLI_assert(rdata->types & (MR_DATATYPE_VERT | MR_DATATYPE_LOOPTRI));
const int tri_len = mesh_render_data_looptri_len_get(rdata);
const int vbo_len_capacity = tri_len * 3;
int vbo_len_used = 0;
/* Positions */
VertexBuffer *vbo_pos = NULL;
Cleanup: group VBO attributes in a struct Some names are a bit arbitrary, this makes it clear which names are VBO attributes.
2017-05-17 12:22:22 +10:00
static struct { uint pos, vnor, lnor, data; } attr_id;
Mesh Batch Cache: Split Edit Mesh mode Vertex Buffers for faster update.
2017-05-13 16:45:06 +02:00
if (cache->ed_tri_pos == NULL) {
Cleanup: group VBO attributes in a struct Some names are a bit arbitrary, this makes it clear which names are VBO attributes.
2017-05-17 12:22:22 +10:00
vbo_pos = cache->ed_tri_pos =
VertexBuffer_create_with_format(edit_mesh_overlay_pos_format(&attr_id.pos));
Mesh Batch Cache: Split Edit Mesh mode Vertex Buffers for faster update.
2017-05-13 16:45:06 +02:00
VertexBuffer_allocate_data(vbo_pos, vbo_len_capacity);
}
/* Normals */
VertexBuffer *vbo_nor = NULL;
if (cache->ed_tri_nor == NULL) {
Cleanup: group VBO attributes in a struct Some names are a bit arbitrary, this makes it clear which names are VBO attributes.
2017-05-17 12:22:22 +10:00
vbo_nor = cache->ed_tri_nor =
VertexBuffer_create_with_format(edit_mesh_overlay_nor_format(&attr_id.vnor, &attr_id.lnor));
Mesh Batch Cache: Split Edit Mesh mode Vertex Buffers for faster update.
2017-05-13 16:45:06 +02:00
VertexBuffer_allocate_data(vbo_nor, vbo_len_capacity);
}
/* Data */
VertexBuffer *vbo_data = NULL;
if (cache->ed_tri_data == NULL) {
Cleanup: group VBO attributes in a struct Some names are a bit arbitrary, this makes it clear which names are VBO attributes.
2017-05-17 12:22:22 +10:00
vbo_data = cache->ed_tri_data =
VertexBuffer_create_with_format(edit_mesh_overlay_data_format(&attr_id.data));
Mesh Batch Cache: Split Edit Mesh mode Vertex Buffers for faster update.
2017-05-13 16:45:06 +02:00
VertexBuffer_allocate_data(vbo_data, vbo_len_capacity);
}
for (int i = 0; i < tri_len; ++i) {
int tri_vert_idx[3], tri_edge_idx[3];
if (mesh_render_data_looptri_vert_edge_indices_get(rdata, i, tri_vert_idx, tri_edge_idx)) {
add_overlay_tri(
rdata, vbo_pos, vbo_nor, vbo_data,
Cleanup: group VBO attributes in a struct Some names are a bit arbitrary, this makes it clear which names are VBO attributes.
2017-05-17 12:22:22 +10:00
attr_id.pos, attr_id.vnor, attr_id.lnor, attr_id.data,
Mesh Batch Cache: Split Edit Mesh mode Vertex Buffers for faster update.
2017-05-13 16:45:06 +02:00
tri_vert_idx, tri_edge_idx, i, vbo_len_used);
vbo_len_used += 3;
}
}
/* Finish */
if (vbo_len_used != vbo_len_capacity) {
if (vbo_pos != NULL) {
VertexBuffer_resize_data(vbo_pos, vbo_len_used);
}
if (vbo_nor != NULL) {
VertexBuffer_resize_data(vbo_nor, vbo_len_used);
}
if (vbo_data != NULL) {
VertexBuffer_resize_data(vbo_data, vbo_len_used);
}
}
}
static void mesh_batch_cache_create_overlay_ledge_buffers(
MeshRenderData *rdata, MeshBatchCache *cache)
{
BLI_assert(rdata->types & (MR_DATATYPE_VERT | MR_DATATYPE_LOOPTRI));
const int ledge_len = mesh_render_data_loose_edges_len_get(rdata);
const int vbo_len_capacity = ledge_len * 2;
int vbo_len_used = 0;
/* Positions */
VertexBuffer *vbo_pos = NULL;
Cleanup: group VBO attributes in a struct Some names are a bit arbitrary, this makes it clear which names are VBO attributes.
2017-05-17 12:22:22 +10:00
static struct { uint pos, vnor, data; } attr_id;
Mesh Batch Cache: Split Edit Mesh mode Vertex Buffers for faster update.
2017-05-13 16:45:06 +02:00
if (cache->ed_ledge_pos == NULL) {
Cleanup: group VBO attributes in a struct Some names are a bit arbitrary, this makes it clear which names are VBO attributes.
2017-05-17 12:22:22 +10:00
vbo_pos = cache->ed_ledge_pos =
VertexBuffer_create_with_format(edit_mesh_overlay_pos_format(&attr_id.pos));
Mesh Batch Cache: Split Edit Mesh mode Vertex Buffers for faster update.
2017-05-13 16:45:06 +02:00
VertexBuffer_allocate_data(vbo_pos, vbo_len_capacity);
}
/* Normals */
VertexBuffer *vbo_nor = NULL;
if (cache->ed_ledge_nor == NULL) {
Cleanup: group VBO attributes in a struct Some names are a bit arbitrary, this makes it clear which names are VBO attributes.
2017-05-17 12:22:22 +10:00
vbo_nor = cache->ed_ledge_nor =
VertexBuffer_create_with_format(edit_mesh_overlay_nor_format(&attr_id.vnor, NULL));
Mesh Batch Cache: Split Edit Mesh mode Vertex Buffers for faster update.
2017-05-13 16:45:06 +02:00
VertexBuffer_allocate_data(vbo_nor, vbo_len_capacity);
}
/* Data */
VertexBuffer *vbo_data = NULL;
if (cache->ed_ledge_data == NULL) {
Cleanup: group VBO attributes in a struct Some names are a bit arbitrary, this makes it clear which names are VBO attributes.
2017-05-17 12:22:22 +10:00
vbo_data = cache->ed_ledge_data =
VertexBuffer_create_with_format(edit_mesh_overlay_data_format(&attr_id.data));
Mesh Batch Cache: Split Edit Mesh mode Vertex Buffers for faster update.
2017-05-13 16:45:06 +02:00
VertexBuffer_allocate_data(vbo_data, vbo_len_capacity);
}
for (int i = 0; i < ledge_len; ++i) {
int vert_idx[2];
Silence warnings in Mac (based on buildbot build report)
2017-05-17 18:13:10 +02:00
const bool ok = mesh_render_data_edge_verts_indices_get(rdata, rdata->loose_edges[i], vert_idx);
Mesh Batch Cache: Split Edit Mesh mode Vertex Buffers for faster update.
2017-05-13 16:45:06 +02:00
BLI_assert(ok); /* we don't add */
Silence warnings in Mac (based on buildbot build report)
2017-05-17 18:13:10 +02:00
UNUSED_VARS_NDEBUG(ok);
Mesh Batch Cache: Split Edit Mesh mode Vertex Buffers for faster update.
2017-05-13 16:45:06 +02:00
add_overlay_loose_edge(
rdata, vbo_pos, vbo_nor, vbo_data,
Cleanup: group VBO attributes in a struct Some names are a bit arbitrary, this makes it clear which names are VBO attributes.
2017-05-17 12:22:22 +10:00
attr_id.pos, attr_id.vnor, attr_id.data,
Mesh Batch Cache: Split Edit Mesh mode Vertex Buffers for faster update.
2017-05-13 16:45:06 +02:00
vert_idx, i, vbo_len_used);
vbo_len_used += 2;
}
/* Finish */
if (vbo_len_used != vbo_len_capacity) {
if (vbo_pos != NULL) {
VertexBuffer_resize_data(vbo_pos, vbo_len_used);
}
if (vbo_nor != NULL) {
VertexBuffer_resize_data(vbo_nor, vbo_len_used);
}
if (vbo_data != NULL) {
VertexBuffer_resize_data(vbo_data, vbo_len_used);
}
}
}
static void mesh_batch_cache_create_overlay_lvert_buffers(
MeshRenderData *rdata, MeshBatchCache *cache)
{
BLI_assert(rdata->types & (MR_DATATYPE_VERT | MR_DATATYPE_LOOPTRI));
const int lvert_len = mesh_render_data_loose_verts_len_get(rdata);
const int vbo_len_capacity = lvert_len;
int vbo_len_used = 0;
Cleanup: group VBO attributes in a struct Some names are a bit arbitrary, this makes it clear which names are VBO attributes.
2017-05-17 12:22:22 +10:00
static struct { uint pos, vnor, data; } attr_id;
Mesh Batch Cache: Split Edit Mesh mode Vertex Buffers for faster update.
2017-05-13 16:45:06 +02:00
/* Positions */
VertexBuffer *vbo_pos = NULL;
if (cache->ed_lvert_pos == NULL) {
Cleanup: group VBO attributes in a struct Some names are a bit arbitrary, this makes it clear which names are VBO attributes.
2017-05-17 12:22:22 +10:00
vbo_pos = cache->ed_lvert_pos =
VertexBuffer_create_with_format(edit_mesh_overlay_pos_format(&attr_id.pos));
Mesh Batch Cache: Split Edit Mesh mode Vertex Buffers for faster update.
2017-05-13 16:45:06 +02:00
VertexBuffer_allocate_data(vbo_pos, vbo_len_capacity);
}
/* Normals */
VertexBuffer *vbo_nor = NULL;
if (cache->ed_lvert_nor == NULL) {
Cleanup: group VBO attributes in a struct Some names are a bit arbitrary, this makes it clear which names are VBO attributes.
2017-05-17 12:22:22 +10:00
vbo_nor = cache->ed_lvert_nor =
VertexBuffer_create_with_format(edit_mesh_overlay_nor_format(&attr_id.vnor, NULL));
Mesh Batch Cache: Split Edit Mesh mode Vertex Buffers for faster update.
2017-05-13 16:45:06 +02:00
VertexBuffer_allocate_data(vbo_nor, vbo_len_capacity);
}
/* Data */
VertexBuffer *vbo_data = NULL;
if (cache->ed_lvert_data == NULL) {
Cleanup: group VBO attributes in a struct Some names are a bit arbitrary, this makes it clear which names are VBO attributes.
2017-05-17 12:22:22 +10:00
vbo_data = cache->ed_lvert_data =
VertexBuffer_create_with_format(edit_mesh_overlay_data_format(&attr_id.data));
Mesh Batch Cache: Split Edit Mesh mode Vertex Buffers for faster update.
2017-05-13 16:45:06 +02:00
VertexBuffer_allocate_data(vbo_data, vbo_len_capacity);
}
for (int i = 0; i < lvert_len; ++i) {
add_overlay_loose_vert(
rdata, vbo_pos, vbo_nor, vbo_data,
Cleanup: group VBO attributes in a struct Some names are a bit arbitrary, this makes it clear which names are VBO attributes.
2017-05-17 12:22:22 +10:00
attr_id.pos, attr_id.vnor, attr_id.data,
Mesh Batch Cache: Split Edit Mesh mode Vertex Buffers for faster update.
2017-05-13 16:45:06 +02:00
rdata->loose_verts[i], vbo_len_used);
vbo_len_used += 1;
}
/* Finish */
if (vbo_len_used != vbo_len_capacity) {
if (vbo_pos != NULL) {
VertexBuffer_resize_data(vbo_pos, vbo_len_used);
}
if (vbo_nor != NULL) {
VertexBuffer_resize_data(vbo_nor, vbo_len_used);
}
if (vbo_data != NULL) {
VertexBuffer_resize_data(vbo_data, vbo_len_used);
}
}
}
/* Position */
static VertexBuffer *mesh_batch_cache_get_edit_tri_pos(
MeshRenderData *rdata, MeshBatchCache *cache)
{
BLI_assert(rdata->types & MR_DATATYPE_VERT);
if (cache->ed_tri_pos == NULL) {
mesh_batch_cache_create_overlay_tri_buffers(rdata, cache);
}
return cache->ed_tri_pos;
}
static VertexBuffer *mesh_batch_cache_get_edit_ledge_pos(
MeshRenderData *rdata, MeshBatchCache *cache)
{
BLI_assert(rdata->types & MR_DATATYPE_VERT);
if (cache->ed_ledge_pos == NULL) {
mesh_batch_cache_create_overlay_ledge_buffers(rdata, cache);
}
return cache->ed_ledge_pos;
}
static VertexBuffer *mesh_batch_cache_get_edit_lvert_pos(
MeshRenderData *rdata, MeshBatchCache *cache)
{
BLI_assert(rdata->types & MR_DATATYPE_VERT);
if (cache->ed_lvert_pos == NULL) {
mesh_batch_cache_create_overlay_lvert_buffers(rdata, cache);
}
return cache->ed_lvert_pos;
}
/* Normal */
static VertexBuffer *mesh_batch_cache_get_edit_tri_nor(
MeshRenderData *rdata, MeshBatchCache *cache)
{
BLI_assert(rdata->types & MR_DATATYPE_VERT);
if (cache->ed_tri_nor == NULL) {
mesh_batch_cache_create_overlay_tri_buffers(rdata, cache);
}
return cache->ed_tri_nor;
}
static VertexBuffer *mesh_batch_cache_get_edit_ledge_nor(
MeshRenderData *rdata, MeshBatchCache *cache)
{
BLI_assert(rdata->types & MR_DATATYPE_VERT);
if (cache->ed_ledge_nor == NULL) {
mesh_batch_cache_create_overlay_ledge_buffers(rdata, cache);
}
return cache->ed_ledge_nor;
}
static VertexBuffer *mesh_batch_cache_get_edit_lvert_nor(
MeshRenderData *rdata, MeshBatchCache *cache)
{
BLI_assert(rdata->types & MR_DATATYPE_VERT);
if (cache->ed_lvert_nor == NULL) {
mesh_batch_cache_create_overlay_lvert_buffers(rdata, cache);
}
return cache->ed_lvert_nor;
}
/* Data */
static VertexBuffer *mesh_batch_cache_get_edit_tri_data(
MeshRenderData *rdata, MeshBatchCache *cache)
{
BLI_assert(rdata->types & MR_DATATYPE_VERT);
if (cache->ed_tri_data == NULL) {
mesh_batch_cache_create_overlay_tri_buffers(rdata, cache);
}
return cache->ed_tri_data;
}
static VertexBuffer *mesh_batch_cache_get_edit_ledge_data(
MeshRenderData *rdata, MeshBatchCache *cache)
{
BLI_assert(rdata->types & MR_DATATYPE_VERT);
if (cache->ed_ledge_data == NULL) {
mesh_batch_cache_create_overlay_ledge_buffers(rdata, cache);
}
return cache->ed_ledge_data;
}
static VertexBuffer *mesh_batch_cache_get_edit_lvert_data(
MeshRenderData *rdata, MeshBatchCache *cache)
{
BLI_assert(rdata->types & MR_DATATYPE_VERT);
if (cache->ed_lvert_data == NULL) {
mesh_batch_cache_create_overlay_lvert_buffers(rdata, cache);
}
return cache->ed_lvert_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 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);
DWM: write mesh data to the VBO directly Was using intermediate functions to store each data type. This is still done in some places (for more involved access). For basic data: coords, normals, colors..etc this is unnecessarily indirect.
2017-05-17 11:28:44 +10:00
if (rdata->edit_bmesh) {
for (int i = 0; i < tri_len; ++i) {
const BMLoop **ltri = (const BMLoop **)rdata->edit_bmesh->looptris[i];
if (!BM_elem_flag_test(ltri[0]->f, BM_ELEM_HIDDEN)) {
for (uint tri_corner = 0; tri_corner < 3; tri_corner++) {
add_generic_vertex(&elb, BM_elem_index_get(ltri[tri_corner]->v));
}
}
}
}
else {
for (int i = 0; i < tri_len; ++i) {
const MLoopTri *mlt = &rdata->mlooptri[i];
for (uint tri_corner = 0; tri_corner < 3; tri_corner++) {
add_generic_vertex(&elb, mlt->tri[tri_corner]);
}
Draw Manager: T51218 Support edit-mesh hide
2017-04-18 23:29:55 +10:00
}
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
}
cache->triangles_in_order = ElementList_build(&elb);
}
return cache->triangles_in_order;
}
Cleanup: use more generic naming Can be used by any user that needs faces split by materials.
2017-05-23 17:48:05 +10:00
static ElementList **mesh_batch_cache_get_triangles_in_order_split_by_material(
MeshRenderData *rdata, MeshBatchCache *cache)
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
{
DwM: optimize material access - Split BMesh/Mesh loops. - Loop over faces instead of tris. - Add out-of-bounds check for material index (rare but can happen).
2017-05-23 18:23:18 +10:00
BLI_assert(rdata->types & (MR_DATATYPE_VERT | MR_DATATYPE_POLY));
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 (cache->shaded_triangles_in_order == NULL) {
DwM: optimize material access - Split BMesh/Mesh loops. - Loop over faces instead of tris. - Add out-of-bounds check for material index (rare but can happen).
2017-05-23 18:23:18 +10:00
const int poly_len = mesh_render_data_polys_len_get(rdata);
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);
Cleanup: rename ct -> len Also use 'defgroup' instead of 'index' for passing weight arg.
2017-05-04 23:35:50 +10:00
const int mat_len = mesh_render_data_mat_len_get(rdata);
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: rename ct -> len Also use 'defgroup' instead of 'index' for passing weight arg.
2017-05-04 23:35:50 +10:00
int *mat_tri_len = MEM_callocN(sizeof(*mat_tri_len) * mat_len, __func__);
cache->shaded_triangles_in_order = MEM_callocN(sizeof(*cache->shaded_triangles) * mat_len, __func__);
ElementListBuilder *elb = MEM_callocN(sizeof(*elb) * mat_len, __func__);
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
DwM: optimize material access - Split BMesh/Mesh loops. - Loop over faces instead of tris. - Add out-of-bounds check for material index (rare but can happen).
2017-05-23 18:23:18 +10:00
/* Note that polygons (not triangles) are used here.
* This OK because result is _guaranteed_ to be the same. */
if (rdata->edit_bmesh) {
BMesh *bm = rdata->edit_bmesh->bm;
BMIter fiter;
BMFace *f;
BM_ITER_MESH(f, &fiter, bm, BM_FACES_OF_MESH) {
if (!BM_elem_flag_test(f, BM_ELEM_HIDDEN)) {
const short ma_id = f->mat_nr < mat_len ? f->mat_nr : 0;
mat_tri_len[ma_id] += (f->len - 2);
}
}
}
else {
for (uint i = 0; i < poly_len; i++) {
const MPoly *mp = &rdata->mpoly[i]; ;
const short ma_id = mp->mat_nr < mat_len ? mp->mat_nr : 0;
mat_tri_len[ma_id] += (mp->totloop - 2);
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
}
}
/* Init ELBs. */
Cleanup: rename ct -> len Also use 'defgroup' instead of 'index' for passing weight arg.
2017-05-04 23:35:50 +10:00
for (int i = 0; i < mat_len; ++i) {
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
ElementListBuilder_init(&elb[i], PRIM_TRIANGLES, mat_tri_len[i], tri_len * 3);
}
/* Populate ELBs. */
DwM: optimize material access - Split BMesh/Mesh loops. - Loop over faces instead of tris. - Add out-of-bounds check for material index (rare but can happen).
2017-05-23 18:23:18 +10:00
uint nidx = 0;
if (rdata->edit_bmesh) {
BMesh *bm = rdata->edit_bmesh->bm;
BMIter fiter;
BMFace *f;
BM_ITER_MESH(f, &fiter, bm, BM_FACES_OF_MESH) {
if (!BM_elem_flag_test(f, BM_ELEM_HIDDEN)) {
const short ma_id = f->mat_nr < mat_len ? f->mat_nr : 0;
Stupid mistake in material access optimization
2017-05-23 21:52:18 +10:00
for (int j = 2; j < f->len; j++) {
add_triangle_vertices(&elb[ma_id], nidx + 0, nidx + 1, nidx + 2);
nidx += 3;
}
DwM: optimize material access - Split BMesh/Mesh loops. - Loop over faces instead of tris. - Add out-of-bounds check for material index (rare but can happen).
2017-05-23 18:23:18 +10:00
}
}
}
else {
for (uint i = 0; i < poly_len; i++) {
const MPoly *mp = &rdata->mpoly[i]; ;
const short ma_id = mp->mat_nr < mat_len ? mp->mat_nr : 0;
Stupid mistake in material access optimization
2017-05-23 21:52:18 +10:00
for (int j = 2; j < mp->totloop; j++) {
add_triangle_vertices(&elb[ma_id], nidx + 0, nidx + 1, nidx + 2);
nidx += 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
}
}
/* Build ELBs. */
Cleanup: rename ct -> len Also use 'defgroup' instead of 'index' for passing weight arg.
2017-05-04 23:35:50 +10:00
for (int i = 0; i < mat_len; ++i) {
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->shaded_triangles_in_order[i] = ElementList_build(&elb[i]);
}
MEM_freeN(mat_tri_len);
MEM_freeN(elb);
}
return cache->shaded_triangles_in_order;
}
Cleanup: line length
2017-05-04 03:35:01 +10:00
static VertexBuffer *mesh_batch_cache_get_edge_pos_with_sel(
Cleanup: clarify select terminology In mesh cache we have select: color, id, bool. Current naming didn't make this clear.
2017-05-17 11:46:26 +10:00
MeshRenderData *rdata, MeshBatchCache *cache, bool use_wire, bool use_select_bool)
Weight painting with draw manager This implements weight rendering with the draw manager, with all drawing options (Shading, wire, face masking, vertex masking). This is part of T51208 Reviewers: campbellbarton Subscribers: dfelinto Differential Revision: https://developer.blender.org/D2654
2017-05-03 18:55:40 +02:00
{
BLI_assert(rdata->types & (MR_DATATYPE_VERT | MR_DATATYPE_EDGE | MR_DATATYPE_POLY | MR_DATATYPE_LOOP));
DWM: write mesh data to the VBO directly Was using intermediate functions to store each data type. This is still done in some places (for more involved access). For basic data: coords, normals, colors..etc this is unnecessarily indirect.
2017-05-17 11:28:44 +10:00
BLI_assert(rdata->edit_bmesh == NULL);
Weight painting with draw manager This implements weight rendering with the draw manager, with all drawing options (Shading, wire, face masking, vertex masking). This is part of T51208 Reviewers: campbellbarton Subscribers: dfelinto Differential Revision: https://developer.blender.org/D2654
2017-05-03 18:55:40 +02:00
Cleanup: clarify select terminology In mesh cache we have select: color, id, bool. Current naming didn't make this clear.
2017-05-17 11:46:26 +10:00
if (!cache->edge_pos_with_select_bool) {
Weight painting with draw manager This implements weight rendering with the draw manager, with all drawing options (Shading, wire, face masking, vertex masking). This is part of T51208 Reviewers: campbellbarton Subscribers: dfelinto Differential Revision: https://developer.blender.org/D2654
2017-05-03 18:55:40 +02:00
unsigned int vidx = 0, cidx = 0;
static VertexFormat format = { 0 };
Cleanup: group VBO attributes in a struct Some names are a bit arbitrary, this makes it clear which names are VBO attributes.
2017-05-17 12:22:22 +10:00
static struct { uint pos, sel; } attr_id;
Weight painting with draw manager This implements weight rendering with the draw manager, with all drawing options (Shading, wire, face masking, vertex masking). This is part of T51208 Reviewers: campbellbarton Subscribers: dfelinto Differential Revision: https://developer.blender.org/D2654
2017-05-03 18:55:40 +02:00
if (format.attrib_ct == 0) {
Cleanup: group VBO attributes in a struct Some names are a bit arbitrary, this makes it clear which names are VBO attributes.
2017-05-17 12:22:22 +10:00
attr_id.pos = VertexFormat_add_attrib(&format, "pos", COMP_F32, 3, KEEP_FLOAT);
attr_id.sel = VertexFormat_add_attrib(&format, "select", COMP_U8, 1, KEEP_INT);
Weight painting with draw manager This implements weight rendering with the draw manager, with all drawing options (Shading, wire, face masking, vertex masking). This is part of T51208 Reviewers: campbellbarton Subscribers: dfelinto Differential Revision: https://developer.blender.org/D2654
2017-05-03 18:55:40 +02:00
}
const int edge_len = mesh_render_data_edges_len_get(rdata);
Cleanup: clarify select terminology In mesh cache we have select: color, id, bool. Current naming didn't make this clear.
2017-05-17 11:46:26 +10:00
VertexBuffer *vbo = cache->edge_pos_with_select_bool = VertexBuffer_create_with_format(&format);
Weight painting with draw manager This implements weight rendering with the draw manager, with all drawing options (Shading, wire, face masking, vertex masking). This is part of T51208 Reviewers: campbellbarton Subscribers: dfelinto Differential Revision: https://developer.blender.org/D2654
2017-05-03 18:55:40 +02:00
const int vbo_len_capacity = edge_len * 2;
int vbo_len_used = 0;
VertexBuffer_allocate_data(vbo, vbo_len_capacity);
Cleanup: clarify select terminology In mesh cache we have select: color, id, bool. Current naming didn't make this clear.
2017-05-17 11:46:26 +10:00
if (use_select_bool) {
mesh_render_data_ensure_edge_select_bool(rdata, use_wire);
DWM: write mesh data to the VBO directly Was using intermediate functions to store each data type. This is still done in some places (for more involved access). For basic data: coords, normals, colors..etc this is unnecessarily indirect.
2017-05-17 11:28:44 +10:00
}
Cleanup: clarify select terminology In mesh cache we have select: color, id, bool. Current naming didn't make this clear.
2017-05-17 11:46:26 +10:00
bool *edge_select_bool = use_select_bool ? rdata->edge_select_bool : NULL;
Weight painting with draw manager This implements weight rendering with the draw manager, with all drawing options (Shading, wire, face masking, vertex masking). This is part of T51208 Reviewers: campbellbarton Subscribers: dfelinto Differential Revision: https://developer.blender.org/D2654
2017-05-03 18:55:40 +02:00
DWM: write mesh data to the VBO directly Was using intermediate functions to store each data type. This is still done in some places (for more involved access). For basic data: coords, normals, colors..etc this is unnecessarily indirect.
2017-05-17 11:28:44 +10:00
for (int i = 0; i < edge_len; i++) {
const MEdge *ed = &rdata->medge[i];
Weight painting with draw manager This implements weight rendering with the draw manager, with all drawing options (Shading, wire, face masking, vertex masking). This is part of T51208 Reviewers: campbellbarton Subscribers: dfelinto Differential Revision: https://developer.blender.org/D2654
2017-05-03 18:55:40 +02:00
DWM: write mesh data to the VBO directly Was using intermediate functions to store each data type. This is still done in some places (for more involved access). For basic data: coords, normals, colors..etc this is unnecessarily indirect.
2017-05-17 11:28:44 +10:00
uchar edge_vert_sel;
Cleanup: clarify select terminology In mesh cache we have select: color, id, bool. Current naming didn't make this clear.
2017-05-17 11:46:26 +10:00
if (use_select_bool && edge_select_bool[i]) {
DWM: write mesh data to the VBO directly Was using intermediate functions to store each data type. This is still done in some places (for more involved access). For basic data: coords, normals, colors..etc this is unnecessarily indirect.
2017-05-17 11:28:44 +10:00
edge_vert_sel = true;
Weight painting with draw manager This implements weight rendering with the draw manager, with all drawing options (Shading, wire, face masking, vertex masking). This is part of T51208 Reviewers: campbellbarton Subscribers: dfelinto Differential Revision: https://developer.blender.org/D2654
2017-05-03 18:55:40 +02:00
}
DWM: write mesh data to the VBO directly Was using intermediate functions to store each data type. This is still done in some places (for more involved access). For basic data: coords, normals, colors..etc this is unnecessarily indirect.
2017-05-17 11:28:44 +10:00
else if (use_wire) {
edge_vert_sel = false;
}
else {
continue;
}
Cleanup: group VBO attributes in a struct Some names are a bit arbitrary, this makes it clear which names are VBO attributes.
2017-05-17 12:22:22 +10:00
VertexBuffer_set_attrib(vbo, attr_id.sel, cidx++, &edge_vert_sel);
VertexBuffer_set_attrib(vbo, attr_id.sel, cidx++, &edge_vert_sel);
Weight painting with draw manager This implements weight rendering with the draw manager, with all drawing options (Shading, wire, face masking, vertex masking). This is part of T51208 Reviewers: campbellbarton Subscribers: dfelinto Differential Revision: https://developer.blender.org/D2654
2017-05-03 18:55:40 +02:00
Cleanup: group VBO attributes in a struct Some names are a bit arbitrary, this makes it clear which names are VBO attributes.
2017-05-17 12:22:22 +10:00
VertexBuffer_set_attrib(vbo, attr_id.pos, vidx++, rdata->mvert[ed->v1].co);
VertexBuffer_set_attrib(vbo, attr_id.pos, vidx++, rdata->mvert[ed->v2].co);
DWM: write mesh data to the VBO directly Was using intermediate functions to store each data type. This is still done in some places (for more involved access). For basic data: coords, normals, colors..etc this is unnecessarily indirect.
2017-05-17 11:28:44 +10:00
}
Weight painting with draw manager This implements weight rendering with the draw manager, with all drawing options (Shading, wire, face masking, vertex masking). This is part of T51208 Reviewers: campbellbarton Subscribers: dfelinto Differential Revision: https://developer.blender.org/D2654
2017-05-03 18:55:40 +02:00
vbo_len_used = vidx;
if (vbo_len_capacity != vbo_len_used) {
VertexBuffer_resize_data(vbo, vbo_len_used);
}
}
Cleanup: clarify select terminology In mesh cache we have select: color, id, bool. Current naming didn't make this clear.
2017-05-17 11:46:26 +10:00
return cache->edge_pos_with_select_bool;
Weight painting with draw manager This implements weight rendering with the draw manager, with all drawing options (Shading, wire, face masking, vertex masking). This is part of T51208 Reviewers: campbellbarton Subscribers: dfelinto Differential Revision: https://developer.blender.org/D2654
2017-05-03 18:55:40 +02:00
}
DWM: reuse position VBO for vpaint select overlay
2017-05-17 19:21:16 +10:00
static ElementList *mesh_batch_cache_get_tri_overlay_weight_faces(
Cleanup: clarify select terminology In mesh cache we have select: color, id, bool. Current naming didn't make this clear.
2017-05-17 11:46:26 +10:00
MeshRenderData *rdata, MeshBatchCache *cache)
Weight painting with draw manager This implements weight rendering with the draw manager, with all drawing options (Shading, wire, face masking, vertex masking). This is part of T51208 Reviewers: campbellbarton Subscribers: dfelinto Differential Revision: https://developer.blender.org/D2654
2017-05-03 18:55:40 +02:00
{
DWM: reuse position VBO for vpaint select overlay
2017-05-17 19:21:16 +10:00
BLI_assert(rdata->types & (MR_DATATYPE_VERT | MR_DATATYPE_LOOPTRI));
Weight painting with draw manager This implements weight rendering with the draw manager, with all drawing options (Shading, wire, face masking, vertex masking). This is part of T51208 Reviewers: campbellbarton Subscribers: dfelinto Differential Revision: https://developer.blender.org/D2654
2017-05-03 18:55:40 +02:00
DWM: reuse position VBO for vpaint select overlay
2017-05-17 19:21:16 +10:00
if (cache->overlay_triangles_vpaint == NULL) {
const int vert_len = mesh_render_data_verts_len_get(rdata);
Weight painting with draw manager This implements weight rendering with the draw manager, with all drawing options (Shading, wire, face masking, vertex masking). This is part of T51208 Reviewers: campbellbarton Subscribers: dfelinto Differential Revision: https://developer.blender.org/D2654
2017-05-03 18:55:40 +02:00
const int tri_len = mesh_render_data_looptri_len_get(rdata);
DWM: reuse position VBO for vpaint select overlay
2017-05-17 19:21:16 +10:00
ElementListBuilder elb;
ElementListBuilder_init(&elb, PRIM_TRIANGLES, tri_len, vert_len);
Weight painting with draw manager This implements weight rendering with the draw manager, with all drawing options (Shading, wire, face masking, vertex masking). This is part of T51208 Reviewers: campbellbarton Subscribers: dfelinto Differential Revision: https://developer.blender.org/D2654
2017-05-03 18:55:40 +02:00
DWM: reuse position VBO for vpaint select overlay
2017-05-17 19:21:16 +10:00
for (int i = 0; i < tri_len; ++i) {
DWM: write mesh data to the VBO directly Was using intermediate functions to store each data type. This is still done in some places (for more involved access). For basic data: coords, normals, colors..etc this is unnecessarily indirect.
2017-05-17 11:28:44 +10:00
const MLoopTri *mlt = &rdata->mlooptri[i];
DWM: correct unselected face overlay
2017-05-17 16:41:21 +10:00
if (!(rdata->mpoly[mlt->poly].flag & (ME_FACE_SEL | ME_HIDE))) {
DWM: write mesh data to the VBO directly Was using intermediate functions to store each data type. This is still done in some places (for more involved access). For basic data: coords, normals, colors..etc this is unnecessarily indirect.
2017-05-17 11:28:44 +10:00
for (uint tri_corner = 0; tri_corner < 3; tri_corner++) {
DWM: reuse position VBO for vpaint select overlay
2017-05-17 19:21:16 +10:00
add_generic_vertex(&elb, rdata->mloop[mlt->tri[tri_corner]].v);
DWM: write mesh data to the VBO directly Was using intermediate functions to store each data type. This is still done in some places (for more involved access). For basic data: coords, normals, colors..etc this is unnecessarily indirect.
2017-05-17 11:28:44 +10:00
}
Weight painting with draw manager This implements weight rendering with the draw manager, with all drawing options (Shading, wire, face masking, vertex masking). This is part of T51208 Reviewers: campbellbarton Subscribers: dfelinto Differential Revision: https://developer.blender.org/D2654
2017-05-03 18:55:40 +02:00
}
}
DWM: reuse position VBO for vpaint select overlay
2017-05-17 19:21:16 +10:00
cache->overlay_triangles_vpaint = ElementList_build(&elb);
Weight painting with draw manager This implements weight rendering with the draw manager, with all drawing options (Shading, wire, face masking, vertex masking). This is part of T51208 Reviewers: campbellbarton Subscribers: dfelinto Differential Revision: https://developer.blender.org/D2654
2017-05-03 18:55:40 +02:00
}
DWM: reuse position VBO for vpaint select overlay
2017-05-17 19:21:16 +10:00
return cache->overlay_triangles_vpaint;
Weight painting with draw manager This implements weight rendering with the draw manager, with all drawing options (Shading, wire, face masking, vertex masking). This is part of T51208 Reviewers: campbellbarton Subscribers: dfelinto Differential Revision: https://developer.blender.org/D2654
2017-05-03 18:55:40 +02:00
}
DWM: hide hidden verts in weight-paint mode
2017-05-17 18:09:40 +10:00
/**
* Non-edit mode vertices (only used for weight-paint mode).
*/
static VertexBuffer *mesh_batch_cache_get_vert_pos_with_overlay_data(
MeshRenderData *rdata, MeshBatchCache *cache)
Weight painting with draw manager This implements weight rendering with the draw manager, with all drawing options (Shading, wire, face masking, vertex masking). This is part of T51208 Reviewers: campbellbarton Subscribers: dfelinto Differential Revision: https://developer.blender.org/D2654
2017-05-03 18:55:40 +02:00
{
BLI_assert(rdata->types & (MR_DATATYPE_VERT));
DWM: write mesh data to the VBO directly Was using intermediate functions to store each data type. This is still done in some places (for more involved access). For basic data: coords, normals, colors..etc this is unnecessarily indirect.
2017-05-17 11:28:44 +10:00
BLI_assert(rdata->edit_bmesh == NULL);
Weight painting with draw manager This implements weight rendering with the draw manager, with all drawing options (Shading, wire, face masking, vertex masking). This is part of T51208 Reviewers: campbellbarton Subscribers: dfelinto Differential Revision: https://developer.blender.org/D2654
2017-05-03 18:55:40 +02:00
Cleanup: clarify select terminology In mesh cache we have select: color, id, bool. Current naming didn't make this clear.
2017-05-17 11:46:26 +10:00
if (cache->pos_with_select_bool == NULL) {
Re-use vertex position VBO in weight-paint mode
2017-05-18 14:53:43 +10:00
unsigned int cidx = 0;
Weight painting with draw manager This implements weight rendering with the draw manager, with all drawing options (Shading, wire, face masking, vertex masking). This is part of T51208 Reviewers: campbellbarton Subscribers: dfelinto Differential Revision: https://developer.blender.org/D2654
2017-05-03 18:55:40 +02:00
static VertexFormat format = { 0 };
Re-use vertex position VBO in weight-paint mode
2017-05-18 14:53:43 +10:00
static struct { uint data; } attr_id;
Weight painting with draw manager This implements weight rendering with the draw manager, with all drawing options (Shading, wire, face masking, vertex masking). This is part of T51208 Reviewers: campbellbarton Subscribers: dfelinto Differential Revision: https://developer.blender.org/D2654
2017-05-03 18:55:40 +02:00
if (format.attrib_ct == 0) {
Re-use vertex position VBO in weight-paint mode
2017-05-18 14:53:43 +10:00
attr_id.data = VertexFormat_add_attrib(&format, "data", COMP_I8, 1, KEEP_INT);
Weight painting with draw manager This implements weight rendering with the draw manager, with all drawing options (Shading, wire, face masking, vertex masking). This is part of T51208 Reviewers: campbellbarton Subscribers: dfelinto Differential Revision: https://developer.blender.org/D2654
2017-05-03 18:55:40 +02:00
}
const int vert_len = mesh_render_data_verts_len_get(rdata);
Cleanup: clarify select terminology In mesh cache we have select: color, id, bool. Current naming didn't make this clear.
2017-05-17 11:46:26 +10:00
VertexBuffer *vbo = cache->pos_with_select_bool = VertexBuffer_create_with_format(&format);
Weight painting with draw manager This implements weight rendering with the draw manager, with all drawing options (Shading, wire, face masking, vertex masking). This is part of T51208 Reviewers: campbellbarton Subscribers: dfelinto Differential Revision: https://developer.blender.org/D2654
2017-05-03 18:55:40 +02:00
const int vbo_len_capacity = vert_len;
int vbo_len_used = 0;
VertexBuffer_allocate_data(vbo, vbo_len_capacity);
for (int i = 0; i < vert_len; i++) {
DWM: write mesh data to the VBO directly Was using intermediate functions to store each data type. This is still done in some places (for more involved access). For basic data: coords, normals, colors..etc this is unnecessarily indirect.
2017-05-17 11:28:44 +10:00
const MVert *mv = &rdata->mvert[i];
DWM: hide hidden verts in weight-paint mode
2017-05-17 18:09:40 +10:00
const char data = mv->flag & (SELECT | ME_HIDE);
Re-use vertex position VBO in weight-paint mode
2017-05-18 14:53:43 +10:00
VertexBuffer_set_attrib(vbo, attr_id.data, cidx++, &data);
Weight painting with draw manager This implements weight rendering with the draw manager, with all drawing options (Shading, wire, face masking, vertex masking). This is part of T51208 Reviewers: campbellbarton Subscribers: dfelinto Differential Revision: https://developer.blender.org/D2654
2017-05-03 18:55:40 +02:00
}
Re-use vertex position VBO in weight-paint mode
2017-05-18 14:53:43 +10:00
vbo_len_used = cidx;
Weight painting with draw manager This implements weight rendering with the draw manager, with all drawing options (Shading, wire, face masking, vertex masking). This is part of T51208 Reviewers: campbellbarton Subscribers: dfelinto Differential Revision: https://developer.blender.org/D2654
2017-05-03 18:55:40 +02:00
if (vbo_len_capacity != vbo_len_used) {
VertexBuffer_resize_data(vbo, vbo_len_used);
}
}
Cleanup: clarify select terminology In mesh cache we have select: color, id, bool. Current naming didn't make this clear.
2017-05-17 11:46:26 +10:00
return cache->pos_with_select_bool;
Weight painting with draw manager This implements weight rendering with the draw manager, with all drawing options (Shading, wire, face masking, vertex masking). This is part of T51208 Reviewers: campbellbarton Subscribers: dfelinto Differential Revision: https://developer.blender.org/D2654
2017-05-03 18:55:40 +02:00
}
Cleanup: use doxy categories
2017-05-05 05:07:52 +10:00
/** \} */
/* ---------------------------------------------------------------------- */
/** \name Public API
* \{ */
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: line length
2017-05-04 03:35:01 +10:00
const int datatype = MR_DATATYPE_VERT | MR_DATATYPE_EDGE;
MeshRenderData *rdata = mesh_render_data_create(me, datatype);
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: line length
2017-05-04 03:35:01 +10:00
cache->all_edges = Batch_create(
Cleanup: include the type of data in function name Was confusing, some functions return vert or triangle data but naming wasn't very clear.
2017-05-05 05:41:30 +10:00
PRIM_LINES, mesh_batch_cache_get_vert_pos_and_nor_in_order(rdata, cache),
Cleanup: line length
2017-05-04 03:35:01 +10:00
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: line length
2017-05-04 03:35:01 +10:00
const int datatype = MR_DATATYPE_VERT | MR_DATATYPE_LOOPTRI;
MeshRenderData *rdata = mesh_render_data_create(me, datatype);
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: line length
2017-05-04 03:35:01 +10:00
cache->all_triangles = Batch_create(
Cleanup: include the type of data in function name Was confusing, some functions return vert or triangle data but naming wasn't very clear.
2017-05-05 05:41:30 +10:00
PRIM_TRIANGLES, mesh_batch_cache_get_vert_pos_and_nor_in_order(rdata, cache),
Cleanup: line length
2017-05-04 03:35:01 +10:00
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: line length
2017-05-04 03:35:01 +10:00
const int datatype = MR_DATATYPE_VERT | MR_DATATYPE_LOOPTRI | MR_DATATYPE_LOOP | MR_DATATYPE_POLY;
MeshRenderData *rdata = mesh_render_data_create(me, datatype);
Clay Engine: Support vertex normals (smooth shading), no face normals yet.
2017-02-15 13:32:35 +01:00
Cleanup: line length
2017-05-04 03:35:01 +10:00
cache->triangles_with_normals = Batch_create(
Cleanup: include the type of data in function name Was confusing, some functions return vert or triangle data but naming wasn't very clear.
2017-05-05 05:41:30 +10:00
PRIM_TRIANGLES, mesh_batch_cache_get_tri_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: rename ct -> len Also use 'defgroup' instead of 'index' for passing weight arg.
2017-05-04 23:35:50 +10:00
Batch *DRW_mesh_batch_cache_get_triangles_with_normals_and_weights(Mesh *me, int defgroup)
Weight painting with draw manager This implements weight rendering with the draw manager, with all drawing options (Shading, wire, face masking, vertex masking). This is part of T51208 Reviewers: campbellbarton Subscribers: dfelinto Differential Revision: https://developer.blender.org/D2654
2017-05-03 18:55:40 +02:00
{
MeshBatchCache *cache = mesh_batch_cache_get(me);
if (cache->triangles_with_weights == NULL) {
DWM: Initial support vert & weight hidden faces
2017-05-17 16:01:18 +10:00
const bool use_hide = (me->editflag & (ME_EDIT_PAINT_VERT_SEL | ME_EDIT_PAINT_FACE_SEL)) != 0;
Cleanup: line length
2017-05-04 03:35:01 +10:00
const int datatype =
MR_DATATYPE_VERT | MR_DATATYPE_LOOPTRI | MR_DATATYPE_LOOP | MR_DATATYPE_POLY | MR_DATATYPE_DVERT;
MeshRenderData *rdata = mesh_render_data_create(me, datatype);
Weight painting with draw manager This implements weight rendering with the draw manager, with all drawing options (Shading, wire, face masking, vertex masking). This is part of T51208 Reviewers: campbellbarton Subscribers: dfelinto Differential Revision: https://developer.blender.org/D2654
2017-05-03 18:55:40 +02:00
Cleanup: line length
2017-05-04 03:35:01 +10:00
cache->triangles_with_weights = Batch_create(
DWM: Initial support vert & weight hidden faces
2017-05-17 16:01:18 +10:00
PRIM_TRIANGLES, mesh_batch_cache_get_tri_weights(rdata, cache, use_hide, defgroup), NULL);
VertexBuffer *vbo_tris = use_hide ?
mesh_batch_cache_get_tri_pos_and_normals_visible_only(rdata, cache) :
mesh_batch_cache_get_tri_pos_and_normals(rdata, cache);
DWM: Share VBO positions & normals for mesh cache Select, weight & color cache included redundant position and normals. Use multiple VBO's per batch for more efficient storage.
2017-05-08 16:25:29 +10:00
DWM: Initial support vert & weight hidden faces
2017-05-17 16:01:18 +10:00
Batch_add_VertexBuffer(cache->triangles_with_weights, vbo_tris);
Weight painting with draw manager This implements weight rendering with the draw manager, with all drawing options (Shading, wire, face masking, vertex masking). This is part of T51208 Reviewers: campbellbarton Subscribers: dfelinto Differential Revision: https://developer.blender.org/D2654
2017-05-03 18:55:40 +02:00
mesh_render_data_free(rdata);
}
return cache->triangles_with_weights;
}
Vertex paint with draw manager Reviewers: fclem Subscribers: campbellbarton, dfelinto Differential Revision: https://developer.blender.org/D2658
2017-05-04 19:59:27 +02:00
Batch *DRW_mesh_batch_cache_get_triangles_with_normals_and_vert_colors(Mesh *me)
{
MeshBatchCache *cache = mesh_batch_cache_get(me);
if (cache->triangles_with_vert_colors == NULL) {
DWM: Initial support vert & weight hidden faces
2017-05-17 16:01:18 +10:00
const bool use_hide = (me->editflag & (ME_EDIT_PAINT_VERT_SEL | ME_EDIT_PAINT_FACE_SEL)) != 0;
Vertex paint with draw manager Reviewers: fclem Subscribers: campbellbarton, dfelinto Differential Revision: https://developer.blender.org/D2658
2017-05-04 19:59:27 +02:00
const int datatype =
MR_DATATYPE_VERT | MR_DATATYPE_LOOPTRI | MR_DATATYPE_LOOP | MR_DATATYPE_POLY | MR_DATATYPE_LOOPCOL;
MeshRenderData *rdata = mesh_render_data_create(me, datatype);
cache->triangles_with_vert_colors = Batch_create(
DWM: Initial support vert & weight hidden faces
2017-05-17 16:01:18 +10:00
PRIM_TRIANGLES, mesh_batch_cache_get_tri_vert_colors(rdata, cache, use_hide), NULL);
DWM: Share VBO positions & normals for mesh cache Select, weight & color cache included redundant position and normals. Use multiple VBO's per batch for more efficient storage.
2017-05-08 16:25:29 +10:00
DWM: Initial support vert & weight hidden faces
2017-05-17 16:01:18 +10:00
VertexBuffer *vbo_tris = use_hide ?
mesh_batch_cache_get_tri_pos_and_normals_visible_only(rdata, cache) :
mesh_batch_cache_get_tri_pos_and_normals(rdata, cache);
Batch_add_VertexBuffer(cache->triangles_with_vert_colors, vbo_tris);
Vertex paint with draw manager Reviewers: fclem Subscribers: campbellbarton, dfelinto Differential Revision: https://developer.blender.org/D2658
2017-05-04 19:59:27 +02:00
mesh_render_data_free(rdata);
}
return cache->triangles_with_vert_colors;
}
Use mesh draw cache for back-buffer selection Vertex/weight paint now work with core profile, resolves T51380.
2017-05-05 05:17:31 +10:00
struct Batch *DRW_mesh_batch_cache_get_triangles_with_select_id(struct Mesh *me, bool use_hide)
{
MeshBatchCache *cache = mesh_batch_cache_get(me);
if (cache->triangles_with_select_id == NULL) {
const int datatype =
MR_DATATYPE_VERT | MR_DATATYPE_LOOPTRI | MR_DATATYPE_LOOP | MR_DATATYPE_POLY;
MeshRenderData *rdata = mesh_render_data_create(me, datatype);
cache->triangles_with_select_id = Batch_create(
DWM: Share VBO positions & normals for mesh cache Select, weight & color cache included redundant position and normals. Use multiple VBO's per batch for more efficient storage.
2017-05-08 16:25:29 +10:00
PRIM_TRIANGLES, mesh_batch_cache_get_tri_select_id(rdata, cache, use_hide), NULL);
Fix weight-painting out of facemask mode w/ hidden faces Own error in recent commit
2017-05-17 19:07:31 +10:00
VertexBuffer *vbo_tris = use_hide ?
mesh_batch_cache_get_tri_pos_and_normals_visible_only(rdata, cache) :
mesh_batch_cache_get_tri_pos_and_normals(rdata, cache);
Batch_add_VertexBuffer(cache->triangles_with_select_id, vbo_tris);
Use mesh draw cache for back-buffer selection Vertex/weight paint now work with core profile, resolves T51380.
2017-05-05 05:17:31 +10:00
mesh_render_data_free(rdata);
}
return cache->triangles_with_select_id;
}
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: line length
2017-05-04 03:35:01 +10:00
const int datatype = MR_DATATYPE_VERT | MR_DATATYPE_LOOPTRI | MR_DATATYPE_LOOP | MR_DATATYPE_POLY;
MeshRenderData *rdata = mesh_render_data_create(me, datatype);
Clay Engine: Support vertex normals (smooth shading), no face normals yet.
2017-02-15 13:32:35 +01:00
Cleanup: line length
2017-05-04 03:35:01 +10:00
cache->points_with_normals = Batch_create(
Cleanup: include the type of data in function name Was confusing, some functions return vert or triangle data but naming wasn't very clear.
2017-05-05 05:41:30 +10:00
PRIM_POINTS, mesh_batch_cache_get_tri_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: line length
2017-05-04 03:35:01 +10:00
cache->all_verts = Batch_create(
Cleanup: include the type of data in function name Was confusing, some functions return vert or triangle data but naming wasn't very clear.
2017-05-05 05:41:30 +10:00
PRIM_POINTS, mesh_batch_cache_get_vert_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 };
Cleanup: group VBO attributes in a struct Some names are a bit arbitrary, this makes it clear which names are VBO attributes.
2017-05-17 12:22:22 +10:00
static struct { uint pos, n1, n2; } attr_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) {
Cleanup: group VBO attributes in a struct Some names are a bit arbitrary, this makes it clear which names are VBO attributes.
2017-05-17 12:22:22 +10:00
attr_id.pos = 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
Cleanup: group VBO attributes in a struct Some names are a bit arbitrary, this makes it clear which names are VBO attributes.
2017-05-17 12:22:22 +10:00
attr_id.n1 = VertexFormat_add_attrib(&format, "N1", COMP_I10, 3, NORMALIZE_INT_TO_FLOAT);
attr_id.n2 = 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
}
VertexBuffer *vbo = VertexBuffer_create_with_format(&format);
Cleanup: line length
2017-05-04 03:35:01 +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
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
Cleanup: group VBO attributes in a struct Some names are a bit arbitrary, this makes it clear which names are VBO attributes.
2017-05-17 12:22:22 +10:00
VertexBuffer_set_attrib(vbo, attr_id.pos, 2 * i, vcos1);
VertexBuffer_set_attrib(vbo, attr_id.n1, 2 * i, n1);
VertexBuffer_set_attrib(vbo, attr_id.n2, 2 * i, n2);
Draw Manager: T51218 Support edit-mesh hide
2017-04-18 23:29:55 +10:00
Cleanup: group VBO attributes in a struct Some names are a bit arbitrary, this makes it clear which names are VBO attributes.
2017-05-17 12:22:22 +10:00
VertexBuffer_set_attrib(vbo, attr_id.pos, 2 * i + 1, vcos2);
VertexBuffer_set_attrib(vbo, attr_id.n1, 2 * i + 1, n1);
VertexBuffer_set_attrib(vbo, attr_id.n2, 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 */
Cleanup: line length
2017-05-22 16:23:25 +10:00
const int options =
MR_DATATYPE_VERT | MR_DATATYPE_EDGE | MR_DATATYPE_LOOP | MR_DATATYPE_POLY |
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
if (cache->overlay_triangles == NULL) {
Cleanup: line length
2017-05-22 16:23:25 +10:00
cache->overlay_triangles = Batch_create(
PRIM_TRIANGLES, mesh_batch_cache_get_edit_tri_pos(rdata, cache), NULL);
Mesh Batch Cache: Split Edit Mesh mode Vertex Buffers for faster update.
2017-05-13 16:45:06 +02:00
Batch_add_VertexBuffer(cache->overlay_triangles, mesh_batch_cache_get_edit_tri_nor(rdata, cache));
Batch_add_VertexBuffer(cache->overlay_triangles, mesh_batch_cache_get_edit_tri_data(rdata, cache));
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) {
Cleanup: line length
2017-05-22 16:23:25 +10:00
cache->overlay_loose_edges = Batch_create(
PRIM_LINES, mesh_batch_cache_get_edit_ledge_pos(rdata, cache), NULL);
Mesh Batch Cache: Split Edit Mesh mode Vertex Buffers for faster update.
2017-05-13 16:45:06 +02:00
Batch_add_VertexBuffer(cache->overlay_loose_edges, mesh_batch_cache_get_edit_ledge_nor(rdata, cache));
Batch_add_VertexBuffer(cache->overlay_loose_edges, mesh_batch_cache_get_edit_ledge_data(rdata, cache));
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) {
Cleanup: line length
2017-05-22 16:23:25 +10:00
cache->overlay_loose_verts = Batch_create(
PRIM_POINTS, mesh_batch_cache_get_edit_lvert_pos(rdata, cache), NULL);
Mesh Batch Cache: Split Edit Mesh mode Vertex Buffers for faster update.
2017-05-13 16:45:06 +02:00
Batch_add_VertexBuffer(cache->overlay_loose_verts, mesh_batch_cache_get_edit_lvert_nor(rdata, cache));
Batch_add_VertexBuffer(cache->overlay_loose_verts, mesh_batch_cache_get_edit_lvert_data(rdata, cache));
}
Edit Mode overlays: separate multiple shaders for loose edges and verts
2017-03-01 14:07:04 +01:00
Mesh Batch Cache: Split Edit Mesh mode Vertex Buffers for faster update.
2017-05-13 16:45:06 +02:00
if (cache->overlay_triangles_nor == NULL) {
Cleanup: line length
2017-05-22 16:23:25 +10:00
cache->overlay_triangles_nor = Batch_create(
PRIM_POINTS, mesh_batch_cache_get_edit_tri_pos(rdata, cache), NULL);
Mesh Batch Cache: Split Edit Mesh mode Vertex Buffers for faster update.
2017-05-13 16:45:06 +02:00
Batch_add_VertexBuffer(cache->overlay_triangles_nor, mesh_batch_cache_get_edit_tri_nor(rdata, cache));
}
if (cache->overlay_loose_edges_nor == NULL) {
Cleanup: line length
2017-05-22 16:23:25 +10:00
cache->overlay_loose_edges_nor = Batch_create(
PRIM_POINTS, mesh_batch_cache_get_edit_ledge_pos(rdata, cache), NULL);
Mesh Batch Cache: Split Edit Mesh mode Vertex Buffers for faster update.
2017-05-13 16:45:06 +02:00
Batch_add_VertexBuffer(cache->overlay_loose_edges_nor, mesh_batch_cache_get_edit_ledge_nor(rdata, cache));
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
}
Mesh Batch Cache: Split Edit Mesh mode Vertex Buffers for faster update.
2017-05-13 16:45:06 +02:00
Batch *DRW_mesh_batch_cache_get_overlay_triangles_nor(Mesh *me)
{
MeshBatchCache *cache = mesh_batch_cache_get(me);
if (cache->overlay_triangles_nor == NULL) {
mesh_batch_cache_create_overlay_batches(me);
}
return cache->overlay_triangles_nor;
}
Batch *DRW_mesh_batch_cache_get_overlay_loose_edges_nor(Mesh *me)
{
MeshBatchCache *cache = mesh_batch_cache_get(me);
if (cache->overlay_loose_edges_nor == NULL) {
mesh_batch_cache_create_overlay_batches(me);
}
return cache->overlay_loose_edges_nor;
}
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 };
Cleanup: group VBO attributes in a struct Some names are a bit arbitrary, this makes it clear which names are VBO attributes.
2017-05-17 12:22:22 +10:00
static struct { uint pos, data; } attr_id;
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
if (format.attrib_ct == 0) {
Cleanup: group VBO attributes in a struct Some names are a bit arbitrary, this makes it clear which names are VBO attributes.
2017-05-17 12:22:22 +10:00
attr_id.pos = VertexFormat_add_attrib(&format, "pos", COMP_F32, 3, KEEP_FLOAT);
attr_id.data = VertexFormat_add_attrib(&format, "norAndFlag", COMP_I10, 4, NORMALIZE_INT_TO_FLOAT);
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);
DWM: Fix face center's not working w/ hidden faces
2017-05-08 16:16:28 +10:00
int vidx = 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);
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
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;
Cleanup: group VBO attributes in a struct Some names are a bit arbitrary, this makes it clear which names are VBO attributes.
2017-05-17 12:22:22 +10:00
VertexBuffer_set_attrib(vbo, attr_id.data, vidx, &nor);
Edit Mesh overlay: pack normals with face centers. Needed for face normals.
2017-03-02 15:03:40 +01:00
Cleanup: group VBO attributes in a struct Some names are a bit arbitrary, this makes it clear which names are VBO attributes.
2017-05-17 12:22:22 +10:00
VertexBuffer_set_attrib(vbo, attr_id.pos, vidx, pcenter);
DWM: Fix face center's not working w/ hidden faces
2017-05-08 16:16:28 +10:00
vidx += 1;
Cleanup: replace num/ct/tot with 'len'
2017-04-19 00:10:32 +10:00
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
}
DWM: Fix face center's not working w/ hidden faces
2017-05-08 16:16:28 +10:00
const int vbo_len_used = vidx;
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 */
Cleanup: line length
2017-05-04 03:35:01 +10:00
const int datatype =
MR_DATATYPE_VERT | MR_DATATYPE_LOOP | MR_DATATYPE_LOOPTRI |
MR_DATATYPE_POLY | MR_DATATYPE_SHADING;
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
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
Cleanup: naming
2017-05-05 11:31:42 +10:00
const int mat_len = mesh_render_data_mat_len_get(rdata);
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: naming
2017-05-05 11:31:42 +10:00
cache->shaded_triangles = MEM_callocN(sizeof(*cache->shaded_triangles) * mat_len, __func__);
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: use more generic naming Can be used by any user that needs faces split by materials.
2017-05-23 17:48:05 +10:00
ElementList **el = mesh_batch_cache_get_triangles_in_order_split_by_material(rdata, cache);
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
DWM: share positions & normals in shading VBO.
2017-05-09 10:56:22 +10:00
VertexBuffer *vbo = mesh_batch_cache_get_tri_pos_and_normals(rdata, cache);
Cleanup: naming
2017-05-05 11:31:42 +10:00
for (int i = 0; i < mat_len; ++i) {
Cleanup: line length
2017-05-04 03:35:01 +10:00
cache->shaded_triangles[i] = Batch_create(
Mesh Batch Cache: get rid of the ORCO VBO data, and reconstruct it in shader. With only one MADD instruction we recover the orco data and reduce both the storage and the fetching cost of an attrib layer.
2017-05-15 16:12:51 +02:00
PRIM_TRIANGLES, vbo, el[i]);
VertexBuffer *vbo_shading = mesh_batch_cache_get_tri_shading_data(rdata, cache);
if (vbo_shading) {
Batch_add_VertexBuffer(cache->shaded_triangles[i], vbo_shading);
}
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
}
DWM: share positions & normals in shading VBO.
2017-05-09 10:56:22 +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
mesh_render_data_free(rdata);
}
return cache->shaded_triangles;
}
DwM: texture paint support & mask mode Uses workaround so material slots are used when neither blender-internal or cycles are enabled.
2017-05-22 23:31:46 +10:00
Batch **DRW_mesh_batch_cache_get_surface_texpaint(Mesh *me)
{
MeshBatchCache *cache = mesh_batch_cache_get(me);
if (cache->texpaint_triangles == NULL) {
/* create batch from DM */
const int datatype =
MR_DATATYPE_VERT | MR_DATATYPE_LOOP | MR_DATATYPE_POLY | MR_DATATYPE_LOOPTRI | MR_DATATYPE_LOOPUV;
MeshRenderData *rdata = mesh_render_data_create(me, datatype);
const int mat_len = mesh_render_data_mat_len_get(rdata);
cache->texpaint_triangles = MEM_callocN(sizeof(*cache->texpaint_triangles) * mat_len, __func__);
Cleanup: use more generic naming Can be used by any user that needs faces split by materials.
2017-05-23 17:48:05 +10:00
ElementList **el = mesh_batch_cache_get_triangles_in_order_split_by_material(rdata, cache);
DwM: texture paint support & mask mode Uses workaround so material slots are used when neither blender-internal or cycles are enabled.
2017-05-22 23:31:46 +10:00
VertexBuffer *vbo = mesh_batch_cache_get_tri_pos_and_normals(rdata, cache);
for (int i = 0; i < mat_len; ++i) {
cache->texpaint_triangles[i] = Batch_create(
PRIM_TRIANGLES, vbo, el[i]);
VertexBuffer *vbo_uv = mesh_batch_cache_get_tri_uv_active(rdata, cache);
if (vbo_uv) {
Batch_add_VertexBuffer(cache->texpaint_triangles[i], vbo_uv);
}
}
mesh_render_data_free(rdata);
}
return cache->texpaint_triangles;
}
Batch *DRW_mesh_batch_cache_get_surface_texpaint_single(Mesh *me)
{
MeshBatchCache *cache = mesh_batch_cache_get(me);
if (cache->texpaint_triangles_single == NULL) {
/* create batch from DM */
const int datatype =
MR_DATATYPE_VERT | MR_DATATYPE_LOOP | MR_DATATYPE_POLY | MR_DATATYPE_LOOPTRI | MR_DATATYPE_LOOPUV;
MeshRenderData *rdata = mesh_render_data_create(me, datatype);
VertexBuffer *vbo = mesh_batch_cache_get_tri_pos_and_normals(rdata, cache);
cache->texpaint_triangles_single = Batch_create(
PRIM_TRIANGLES, vbo, NULL);
VertexBuffer *vbo_uv = mesh_batch_cache_get_tri_uv_active(rdata, cache);
if (vbo_uv) {
Batch_add_VertexBuffer(cache->texpaint_triangles_single, vbo_uv);
}
mesh_render_data_free(rdata);
}
return cache->texpaint_triangles_single;
}
Clean up weight/vertex painting code Now passing selection state instead of colors for the wire/face mask overlay thing. Also added masking indication on the faces in vertex paint.
2017-05-05 14:25:43 +02:00
Batch *DRW_mesh_batch_cache_get_weight_overlay_edges(Mesh *me, bool use_wire, bool use_sel)
Weight painting with draw manager This implements weight rendering with the draw manager, with all drawing options (Shading, wire, face masking, vertex masking). This is part of T51208 Reviewers: campbellbarton Subscribers: dfelinto Differential Revision: https://developer.blender.org/D2654
2017-05-03 18:55:40 +02:00
{
MeshBatchCache *cache = mesh_batch_cache_get(me);
Clean up weight/vertex painting code Now passing selection state instead of colors for the wire/face mask overlay thing. Also added masking indication on the faces in vertex paint.
2017-05-05 14:25:43 +02:00
if (cache->overlay_paint_edges == NULL) {
Weight painting with draw manager This implements weight rendering with the draw manager, with all drawing options (Shading, wire, face masking, vertex masking). This is part of T51208 Reviewers: campbellbarton Subscribers: dfelinto Differential Revision: https://developer.blender.org/D2654
2017-05-03 18:55:40 +02:00
/* create batch from Mesh */
Cleanup: line length
2017-05-04 03:35:01 +10:00
const int datatype = MR_DATATYPE_VERT | MR_DATATYPE_EDGE | MR_DATATYPE_POLY | MR_DATATYPE_LOOP;
MeshRenderData *rdata = mesh_render_data_create(me, datatype);
Weight painting with draw manager This implements weight rendering with the draw manager, with all drawing options (Shading, wire, face masking, vertex masking). This is part of T51208 Reviewers: campbellbarton Subscribers: dfelinto Differential Revision: https://developer.blender.org/D2654
2017-05-03 18:55:40 +02:00
Clean up weight/vertex painting code Now passing selection state instead of colors for the wire/face mask overlay thing. Also added masking indication on the faces in vertex paint.
2017-05-05 14:25:43 +02:00
cache->overlay_paint_edges = Batch_create(
PRIM_LINES, mesh_batch_cache_get_edge_pos_with_sel(rdata, cache, use_wire, use_sel), NULL);
Weight painting with draw manager This implements weight rendering with the draw manager, with all drawing options (Shading, wire, face masking, vertex masking). This is part of T51208 Reviewers: campbellbarton Subscribers: dfelinto Differential Revision: https://developer.blender.org/D2654
2017-05-03 18:55:40 +02:00
mesh_render_data_free(rdata);
}
Clean up weight/vertex painting code Now passing selection state instead of colors for the wire/face mask overlay thing. Also added masking indication on the faces in vertex paint.
2017-05-05 14:25:43 +02:00
return cache->overlay_paint_edges;
Weight painting with draw manager This implements weight rendering with the draw manager, with all drawing options (Shading, wire, face masking, vertex masking). This is part of T51208 Reviewers: campbellbarton Subscribers: dfelinto Differential Revision: https://developer.blender.org/D2654
2017-05-03 18:55:40 +02:00
}
Batch *DRW_mesh_batch_cache_get_weight_overlay_faces(Mesh *me)
{
MeshBatchCache *cache = mesh_batch_cache_get(me);
if (cache->overlay_weight_faces == NULL) {
/* create batch from Mesh */
Cleanup: line length
2017-05-04 03:35:01 +10:00
const int datatype = MR_DATATYPE_VERT | MR_DATATYPE_POLY | MR_DATATYPE_LOOP | MR_DATATYPE_LOOPTRI;
MeshRenderData *rdata = mesh_render_data_create(me, datatype);
Weight painting with draw manager This implements weight rendering with the draw manager, with all drawing options (Shading, wire, face masking, vertex masking). This is part of T51208 Reviewers: campbellbarton Subscribers: dfelinto Differential Revision: https://developer.blender.org/D2654
2017-05-03 18:55:40 +02:00
Cleanup: line length
2017-05-04 03:35:01 +10:00
cache->overlay_weight_faces = Batch_create(
DWM: reuse position VBO for vpaint select overlay
2017-05-17 19:21:16 +10:00
PRIM_TRIANGLES, mesh_batch_cache_get_vert_pos_and_nor_in_order(rdata, cache),
mesh_batch_cache_get_tri_overlay_weight_faces(rdata, cache));
Weight painting with draw manager This implements weight rendering with the draw manager, with all drawing options (Shading, wire, face masking, vertex masking). This is part of T51208 Reviewers: campbellbarton Subscribers: dfelinto Differential Revision: https://developer.blender.org/D2654
2017-05-03 18:55:40 +02:00
mesh_render_data_free(rdata);
}
return cache->overlay_weight_faces;
}
Batch *DRW_mesh_batch_cache_get_weight_overlay_verts(Mesh *me)
{
MeshBatchCache *cache = mesh_batch_cache_get(me);
if (cache->overlay_weight_verts == NULL) {
/* create batch from Mesh */
MeshRenderData *rdata = mesh_render_data_create(me, MR_DATATYPE_VERT);
Cleanup: line length
2017-05-04 03:35:01 +10:00
cache->overlay_weight_verts = Batch_create(
Re-use vertex position VBO in weight-paint mode
2017-05-18 14:53:43 +10:00
PRIM_POINTS, mesh_batch_cache_get_vert_pos_and_nor_in_order(rdata, cache), NULL);
Weight painting with draw manager This implements weight rendering with the draw manager, with all drawing options (Shading, wire, face masking, vertex masking). This is part of T51208 Reviewers: campbellbarton Subscribers: dfelinto Differential Revision: https://developer.blender.org/D2654
2017-05-03 18:55:40 +02:00
Re-use vertex position VBO in weight-paint mode
2017-05-18 14:53:43 +10:00
Batch_add_VertexBuffer(
cache->overlay_weight_verts,
mesh_batch_cache_get_vert_pos_with_overlay_data(rdata, cache));
Weight painting with draw manager This implements weight rendering with the draw manager, with all drawing options (Shading, wire, face masking, vertex masking). This is part of T51208 Reviewers: campbellbarton Subscribers: dfelinto Differential Revision: https://developer.blender.org/D2654
2017-05-03 18:55:40 +02:00
mesh_render_data_free(rdata);
}
return cache->overlay_weight_verts;
}
Cleanup: use doxy categories
2017-05-05 05:07:52 +10:00
/** \} */
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
#undef MESH_RENDER_FUNCTION
Copy Permalink