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This repository has been archived on 2023-10-09.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. You cannot open issues or pull requests or push a commit.
Files
d7982f71e95e5caecb31dca54eb3cf0670aabbb3
blender-archive/source/blender/draw/engines/eevee/eevee_lightprobes.c

1857 lines
66 KiB
C
Raw Normal View History

Eevee: Introduction of world preconvolved envmap. For now only compute GGX convolution. The GGX LUT used for the split sum approximation (UE4) is merged with the LTX mag LUT that uses the same parameters (theta and roughness)
2017-04-18 12:50:09 +02:00
/*
* Copyright 2016, Blender Foundation.
*
* 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.
*
* Contributor(s): Blender Institute
*
*/
Cleanup: style, duplicate includes
2017-10-07 15:57:14 +11:00
/** \file eevee_lightprobes.c
Eevee: Introduction of world preconvolved envmap. For now only compute GGX convolution. The GGX LUT used for the split sum approximation (UE4) is merged with the LTX mag LUT that uses the same parameters (theta and roughness)
2017-04-18 12:50:09 +02:00
* \ingroup DNA
*/
Cleanup: BLI headers first Also move descriptions into doxy header
2017-11-20 14:11:45 +11:00
#include "DRW_render.h"
#include "BLI_utildefines.h"
Cleanup: replace BLI_dynstr w/ BLI_string_joinN Use in place of recently reverted 'DRW_shader_create_lib'
2018-01-17 21:40:43 +11:00
#include "BLI_string_utils.h"
Cleanup: BLI headers first Also move descriptions into doxy header
2017-11-20 14:11:45 +11:00
#include "BLI_rand.h"
Eevee: Introduction of world preconvolved envmap. For now only compute GGX convolution. The GGX LUT used for the split sum approximation (UE4) is merged with the LTX mag LUT that uses the same parameters (theta and roughness)
2017-04-18 12:50:09 +02:00
#include "DNA_world_types.h"
#include "DNA_texture_types.h"
#include "DNA_image_types.h"
Rename probe to light-probe Probe is a real general term, the new name is used often in docs online.
2017-06-12 20:59:54 +10:00
#include "DNA_lightprobe_types.h"
Eevee: First commit of Probe support.
2017-06-08 21:48:50 +02:00
#include "DNA_view3d_types.h"
Eevee: Introduction of world preconvolved envmap. For now only compute GGX convolution. The GGX LUT used for the split sum approximation (UE4) is merged with the LTX mag LUT that uses the same parameters (theta and roughness)
2017-04-18 12:50:09 +02:00
Collections and groups unification OVERVIEW * In 2.7 terminology, all layers and groups are now collection datablocks. * These collections are nestable, linkable, instanceable, overrideable, .. which opens up new ways to set up scenes and link + override data. * Viewport/render visibility and selectability are now a part of the collection and shared across all view layers and linkable. * View layers define which subset of the scene collection hierarchy is excluded for each. For many workflows one view layer can be used, these are more of an advanced feature now. OUTLINER * The outliner now has a "View Layer" display mode instead of "Collections", which can display the collections and/or objects in the view layer. * In this display mode, collections can be excluded with the right click menu. These will then be greyed out and their objects will be excluded. * To view collections not linked to any scene, the "Blender File" display mode can be used, with the new filtering option to just see Colleciton datablocks. * The outliner right click menus for collections and objects were reorganized. * Drag and drop still needs to be improved. Like before, dragging the icon or text gives different results, we'll unify this later. LINKING AND OVERRIDES * Collections can now be linked into the scene without creating an instance, with the link/append operator or from the collections view in the outliner. * Collections can get static overrides with the right click menu in the outliner, but this is rather unreliable and not clearly communicated at the moment. * We still need to improve the make override operator to turn collection instances into collections with overrides directly in the scene. PERFORMANCE * We tried to make performance not worse than before and improve it in some cases. The main thing that's still a bit slower is multiple scenes, we have to change the layer syncing to only updated affected scenes. * Collections keep a list of their parent collections for faster incremental updates in syncing and caching. * View layer bases are now in a object -> base hash to avoid quadratic time lookups internally and in API functions like visible_get(). VERSIONING * Compatibility with 2.7 files should be improved due to the new visibility controls. Of course users may not want to set up their scenes differently now to avoid having separate layers and groups. * Compatibility with 2.8 is mostly there, and was tested on Eevee demo and Hero files. There's a few things which are know to be not quite compatible, like nested layer collections inside groups. * The versioning code for 2.8 files is quite complicated, and isolated behind #ifdef so it can be removed at the end of the release cycle. KNOWN ISSUES * The G-key group operators in the 3D viewport were left mostly as is, they need to be modified still to fit better. * Same for the groups panel in the object properties. This needs to be updated still, or perhaps replaced by something better. * Collections must all have a unique name. Less restrictive namespacing is to be done later, we'll have to see how important this is as all objects within the collections must also have a unique name anyway. * Full scene copy and delete scene are exactly doing the right thing yet. Differential Revision: https://developer.blender.org/D3383 https://code.blender.org/2018/05/collections-and-groups/
2018-04-30 15:57:22 +02:00
#include "BKE_collection.h"
Eevee: Add parallax correction to probe reflections
2017-06-09 22:30:49 +02:00
#include "BKE_object.h"
DRW: Refactor: Less feature duplication with Gwn. This removes the need of custom attribs for instancing. Instancing works fully with dynamic batches & Gwn_VertFormat now. This is in prevision of the VAO manager patch.
2018-02-14 18:59:15 +01:00
#include "MEM_guardedalloc.h"
Eevee: Add parallax correction to probe reflections
2017-06-09 22:30:49 +02:00
Eevee: Material code refactor. Separate material handling inside another file. Make use of enums to identify shader variations. Group all 64*64 LUTs into one array texture. Only update world probe if world changes.
2017-06-04 12:12:58 +02:00
#include "GPU_material.h"
Eevee: Introduction of world preconvolved envmap. For now only compute GGX convolution. The GGX LUT used for the split sum approximation (UE4) is merged with the LTX mag LUT that uses the same parameters (theta and roughness)
2017-04-18 12:50:09 +02:00
#include "GPU_texture.h"
#include "GPU_glew.h"
Move EEVEE properties into scene We handle doversion for the scene properties, but not for the view layer overrides. Overrides will be implemented in a different way via dynamic overrides. For now this data is completely lost.
2018-05-16 19:34:24 +02:00
#include "DEG_depsgraph_query.h"
Eevee: Material code refactor. Separate material handling inside another file. Make use of enums to identify shader variations. Group all 64*64 LUTs into one array texture. Only update world probe if world changes.
2017-06-04 12:12:58 +02:00
#include "eevee_engine.h"
#include "eevee_private.h"
Cleanup: BLI headers first Also move descriptions into doxy header
2017-11-20 14:11:45 +11:00
#include "ED_screen.h"
Eevee: Irradiance Grid: Allocate needed resources instead of a static chunck. This commit makes the irradiance pool and render target sizes depend on the number of irradiance sample in the whole ViewLayer.
2017-12-04 17:20:20 +01:00
/* Rounded to nearest PowerOfTwo */
#if defined(IRRADIANCE_SH_L2)
#define IRRADIANCE_SAMPLE_SIZE_X 4 /* 3 in reality */
#define IRRADIANCE_SAMPLE_SIZE_Y 4 /* 3 in reality */
#elif defined(IRRADIANCE_CUBEMAP)
#define IRRADIANCE_SAMPLE_SIZE_X 8
#define IRRADIANCE_SAMPLE_SIZE_Y 8
#elif defined(IRRADIANCE_HL2)
#define IRRADIANCE_SAMPLE_SIZE_X 4 /* 3 in reality */
#define IRRADIANCE_SAMPLE_SIZE_Y 2
#endif
#define IRRADIANCE_MAX_POOL_LAYER 256 /* OpenGL 3.3 core requirement, can be extended but it's already very big */
#define IRRADIANCE_MAX_POOL_SIZE 1024
#define MAX_IRRADIANCE_SAMPLES \
(IRRADIANCE_MAX_POOL_SIZE / IRRADIANCE_SAMPLE_SIZE_X) * \
(IRRADIANCE_MAX_POOL_SIZE / IRRADIANCE_SAMPLE_SIZE_Y)
Eevee: Irradiance Visibility: Initial Implementation This augment the existing irradiance grid with a new visibility precomputation. We store a small shadowmap for each grid sample so that light does not leak through walls and such. The visibility parameter are similar to the one used by the Variance Shadow Map for point lights. Technical details: We store the visibility in the same texture (array) as the irradiance itself (in order to reduce the number of sampler). But the irradiance and the visibility are not the same data so we must encode them in order to use the same texture format. We use RGBA8 normalized texture and encode irradiance as RGBE (shared exponent). Using RGBE encoding instead of R11_G11_B10 may lead to some lighting changes, but quality seems to be nearly the same in my test cases. Using full RGBA16/32F maybe a future option but that will require much more memory and reduce the perf significantly. Visibility moments (VSM) are encoded as 16bits fixed point precision using a special range. This seems to retain enough precision for the needs. Also interpolation does not seems to be big problem (even though it's incorrect).
2017-12-02 14:28:29 +01:00
#define HAMMERSLEY_SIZE 1024
Eevee: Replace Cubemaps by octahedron maps for env. probes. This enables us to use 2D texture arrays for multiple probes. There is a little artifact with very high roughness caused elongated pixel due to the projection (along every 90° meridian).
2017-05-29 22:03:57 +02:00
static struct {
Eevee: Material code refactor. Separate material handling inside another file. Make use of enums to identify shader variations. Group all 64*64 LUTs into one array texture. Only update world probe if world changes.
2017-06-04 12:12:58 +02:00
struct GPUShader *probe_default_sh;
EEvEE: LookDev
2018-05-25 08:06:36 +02:00
struct GPUShader *probe_default_studiolight_sh;
Eevee: Add Irradiance Grid support Early implementation. Slow and still has quality 3 ways of storing irradiance: - Spherical Harmonics: Have problem with directionnal lighting. - HL2 diffuse cube: Very low resolution but smooth transitions. - Diffuse cube: High storage requirement. Also include some name change.
2017-06-13 17:39:39 +02:00
struct GPUShader *probe_filter_glossy_sh;
struct GPUShader *probe_filter_diffuse_sh;
Eevee: Irradiance Visibility: Initial Implementation This augment the existing irradiance grid with a new visibility precomputation. We store a small shadowmap for each grid sample so that light does not leak through walls and such. The visibility parameter are similar to the one used by the Variance Shadow Map for point lights. Technical details: We store the visibility in the same texture (array) as the irradiance itself (in order to reduce the number of sampler). But the irradiance and the visibility are not the same data so we must encode them in order to use the same texture format. We use RGBA8 normalized texture and encode irradiance as RGBE (shared exponent). Using RGBE encoding instead of R11_G11_B10 may lead to some lighting changes, but quality seems to be nearly the same in my test cases. Using full RGBA16/32F maybe a future option but that will require much more memory and reduce the perf significantly. Visibility moments (VSM) are encoded as 16bits fixed point precision using a special range. This seems to retain enough precision for the needs. Also interpolation does not seems to be big problem (even though it's incorrect).
2017-12-02 14:28:29 +01:00
struct GPUShader *probe_filter_visibility_sh;
Eevee: Probe Grid: Clear Grid buffers with world diffuse coefs. This make sure the values displayed by the "show data" sphere are initialized. Also this make the bounce lighting progress more apparent.
2017-09-28 21:17:57 +02:00
struct GPUShader *probe_grid_fill_sh;
Eevee: Add Grid debug display.
2017-06-14 13:45:54 +02:00
struct GPUShader *probe_grid_display_sh;
Eevee: Add data display for planar reflection. Maybe not very useful but it's here for feature parity with the other probes. The fragment shader clipping is here because I'm lazy and don't want to creating a proper quad Batch.
2017-06-17 01:59:01 +02:00
struct GPUShader *probe_planar_display_sh;
Eevee: Add Planar reflection blurring. This method is very cheap and inaccurate. This will fill the gap untill better model is supported.
2017-06-26 20:37:41 +02:00
struct GPUShader *probe_planar_downsample_sh;
Eevee: Probes: Add data display for cubemaps.
2017-06-15 00:10:34 +02:00
struct GPUShader *probe_cube_display_sh;
Eevee: Replace Cubemaps by octahedron maps for env. probes. This enables us to use 2D texture arrays for multiple probes. There is a little artifact with very high roughness caused elongated pixel due to the projection (along every 90° meridian).
2017-05-29 22:03:57 +02:00
struct GPUTexture *hammersley;
Eevee: Planar reflection: Fix bad texture binding.
2017-06-23 19:02:40 +02:00
struct GPUTexture *planar_pool_placeholder;
Eevee: HiZ buffer: Split into two 24bit depth buffer This way we don't have float precision issue we had before and we save some bandwidth.
2017-07-21 14:27:16 +02:00
struct GPUTexture *depth_placeholder;
Eevee: Planar Reflection: Remove distance approximation. This commit separate the depth texture into another texture array. This remove the need to output radial depth into alpha. Unfortunatly it's difficult to recover position from the non linear depth buffer when applying reflection without adding a bunch of stuff. This is in preparation of SSR planar reflections.
2017-07-23 14:03:27 +02:00
struct GPUTexture *depth_array_placeholder;
Eevee: Add AO support in planar reflection. Technically this enables the use of MinmaxZ pyramid inside the probe captures. Also Disable AO for cubemap probes because it shows big discontinuities at cubeface limits.
2017-06-23 21:08:53 +02:00
struct GPUTexture *cube_face_minmaxz;
Eevee: Replace Cubemaps by octahedron maps for env. probes. This enables us to use 2D texture arrays for multiple probes. There is a little artifact with very high roughness caused elongated pixel due to the projection (along every 90° meridian).
2017-05-29 22:03:57 +02:00
DRW: Refactor: Less feature duplication with Gwn. This removes the need of custom attribs for instancing. Instancing works fully with dynamic batches & Gwn_VertFormat now. This is in prevision of the VAO manager patch.
2018-02-14 18:59:15 +01:00
struct Gwn_VertFormat *format_probe_display_cube;
struct Gwn_VertFormat *format_probe_display_planar;
Eevee: Replace Cubemaps by octahedron maps for env. probes. This enables us to use 2D texture arrays for multiple probes. There is a little artifact with very high roughness caused elongated pixel due to the projection (along every 90° meridian).
2017-05-29 22:03:57 +02:00
} e_data = {NULL}; /* Engine data */
Eevee: Fix world probe with world without nodetree.
2017-06-29 17:32:13 +02:00
extern char datatoc_background_vert_glsl[];
Eevee: Material code refactor. Separate material handling inside another file. Make use of enums to identify shader variations. Group all 64*64 LUTs into one array texture. Only update world probe if world changes.
2017-06-04 12:12:58 +02:00
extern char datatoc_default_world_frag_glsl[];
Eevee: Add Irradiance Grid support Early implementation. Slow and still has quality 3 ways of storing irradiance: - Spherical Harmonics: Have problem with directionnal lighting. - HL2 diffuse cube: Very low resolution but smooth transitions. - Diffuse cube: High storage requirement. Also include some name change.
2017-06-13 17:39:39 +02:00
extern char datatoc_lightprobe_filter_glossy_frag_glsl[];
extern char datatoc_lightprobe_filter_diffuse_frag_glsl[];
Eevee: Irradiance Visibility: Initial Implementation This augment the existing irradiance grid with a new visibility precomputation. We store a small shadowmap for each grid sample so that light does not leak through walls and such. The visibility parameter are similar to the one used by the Variance Shadow Map for point lights. Technical details: We store the visibility in the same texture (array) as the irradiance itself (in order to reduce the number of sampler). But the irradiance and the visibility are not the same data so we must encode them in order to use the same texture format. We use RGBA8 normalized texture and encode irradiance as RGBE (shared exponent). Using RGBE encoding instead of R11_G11_B10 may lead to some lighting changes, but quality seems to be nearly the same in my test cases. Using full RGBA16/32F maybe a future option but that will require much more memory and reduce the perf significantly. Visibility moments (VSM) are encoded as 16bits fixed point precision using a special range. This seems to retain enough precision for the needs. Also interpolation does not seems to be big problem (even though it's incorrect).
2017-12-02 14:28:29 +01:00
extern char datatoc_lightprobe_filter_visibility_frag_glsl[];
Rename probe to light-probe Probe is a real general term, the new name is used often in docs online.
2017-06-12 20:59:54 +10:00
extern char datatoc_lightprobe_geom_glsl[];
extern char datatoc_lightprobe_vert_glsl[];
Eevee: Add data display for planar reflection. Maybe not very useful but it's here for feature parity with the other probes. The fragment shader clipping is here because I'm lazy and don't want to creating a proper quad Batch.
2017-06-17 01:59:01 +02:00
extern char datatoc_lightprobe_planar_display_frag_glsl[];
extern char datatoc_lightprobe_planar_display_vert_glsl[];
Eevee: Add Planar reflection blurring. This method is very cheap and inaccurate. This will fill the gap untill better model is supported.
2017-06-26 20:37:41 +02:00
extern char datatoc_lightprobe_planar_downsample_frag_glsl[];
extern char datatoc_lightprobe_planar_downsample_geom_glsl[];
extern char datatoc_lightprobe_planar_downsample_vert_glsl[];
Eevee: Probes: Add data display for cubemaps.
2017-06-15 00:10:34 +02:00
extern char datatoc_lightprobe_cube_display_frag_glsl[];
extern char datatoc_lightprobe_cube_display_vert_glsl[];
Eevee: Add Grid debug display.
2017-06-14 13:45:54 +02:00
extern char datatoc_lightprobe_grid_display_frag_glsl[];
extern char datatoc_lightprobe_grid_display_vert_glsl[];
Eevee: Probe Grid: Clear Grid buffers with world diffuse coefs. This make sure the values displayed by the "show data" sphere are initialized. Also this make the bounce lighting progress more apparent.
2017-09-28 21:17:57 +02:00
extern char datatoc_lightprobe_grid_fill_frag_glsl[];
Eevee: Add Grid debug display.
2017-06-14 13:45:54 +02:00
extern char datatoc_irradiance_lib_glsl[];
Eevee: Refactor of shading code to be more modular. This will enable creating shading models more easily.
2017-06-28 16:31:25 +02:00
extern char datatoc_lightprobe_lib_glsl[];
Eevee: Group octahedron map functions into one file.
2017-06-15 00:09:49 +02:00
extern char datatoc_octahedron_lib_glsl[];
Eevee: Replace Cubemaps by octahedron maps for env. probes. This enables us to use 2D texture arrays for multiple probes. There is a little artifact with very high roughness caused elongated pixel due to the projection (along every 90° meridian).
2017-05-29 22:03:57 +02:00
extern char datatoc_bsdf_common_lib_glsl[];
Eevee: Put all constant uniforms in a global UBO. This is an optimization / cleanup commit. The use of a global ubo remove lots of uniform lookups and only transfert data when needed. Lots of renaming for more consistent codestyle.
2018-01-21 17:25:10 +01:00
extern char datatoc_common_uniforms_lib_glsl[];
Eevee: Make use of the new view matrix UBO.
2018-03-10 00:00:04 +01:00
extern char datatoc_common_view_lib_glsl[];
Eevee: Replace Cubemaps by octahedron maps for env. probes. This enables us to use 2D texture arrays for multiple probes. There is a little artifact with very high roughness caused elongated pixel due to the projection (along every 90° meridian).
2017-05-29 22:03:57 +02:00
extern char datatoc_bsdf_sampling_lib_glsl[];
Eevee: Introduction of world preconvolved envmap. For now only compute GGX convolution. The GGX LUT used for the split sum approximation (UE4) is merged with the LTX mag LUT that uses the same parameters (theta and roughness)
2017-04-18 12:50:09 +02:00
Eevee: Material code refactor. Separate material handling inside another file. Make use of enums to identify shader variations. Group all 64*64 LUTs into one array texture. Only update world probe if world changes.
2017-06-04 12:12:58 +02:00
extern GlobalsUboStorage ts;
Eevee: Introduction of world preconvolved envmap. For now only compute GGX convolution. The GGX LUT used for the split sum approximation (UE4) is merged with the LTX mag LUT that uses the same parameters (theta and roughness)
2017-04-18 12:50:09 +02:00
/* *********** FUNCTIONS *********** */
Eevee: Irradiance Grid: Allocate needed resources instead of a static chunck. This commit makes the irradiance pool and render target sizes depend on the number of irradiance sample in the whole ViewLayer.
2017-12-04 17:20:20 +01:00
static void irradiance_pool_size_get(int visibility_size, int total_samples, int r_size[3])
{
/* Compute how many irradiance samples we can store per visibility sample. */
int irr_per_vis = (visibility_size / IRRADIANCE_SAMPLE_SIZE_X) *
(visibility_size / IRRADIANCE_SAMPLE_SIZE_Y);
/* The irradiance itself take one layer, hence the +1 */
int layer_ct = MIN2(irr_per_vis + 1, IRRADIANCE_MAX_POOL_LAYER);
int texel_ct = (int)ceilf((float)total_samples / (float)(layer_ct - 1));
r_size[0] = visibility_size * max_ii(1, min_ii(texel_ct, (IRRADIANCE_MAX_POOL_SIZE / visibility_size)));
r_size[1] = visibility_size * max_ii(1, (texel_ct / (IRRADIANCE_MAX_POOL_SIZE / visibility_size)));
r_size[2] = layer_ct;
}
Eevee: Replace Cubemaps by octahedron maps for env. probes. This enables us to use 2D texture arrays for multiple probes. There is a little artifact with very high roughness caused elongated pixel due to the projection (along every 90° meridian).
2017-05-29 22:03:57 +02:00
static struct GPUTexture *create_hammersley_sample_texture(int samples)
{
struct GPUTexture *tex;
float (*texels)[2] = MEM_mallocN(sizeof(float[2]) * samples, "hammersley_tex");
int i;
for (i = 0; i < samples; i++) {
BLI_rand : Make use of BLI_halton and BLI_hammersley
2017-09-26 21:38:23 +02:00
double dphi;
BLI_hammersley_1D(i, &dphi);
float phi = (float)dphi * 2.0f * M_PI;
Eevee: Modify Blue Noise.
2017-06-04 21:45:41 +02:00
texels[i][0] = cosf(phi);
Eevee: Replace Cubemaps by octahedron maps for env. probes. This enables us to use 2D texture arrays for multiple probes. There is a little artifact with very high roughness caused elongated pixel due to the projection (along every 90° meridian).
2017-05-29 22:03:57 +02:00
texels[i][1] = sinf(phi);
}
DRW: Remove DRWTextureFormat in favor or GPUTextureFormat. Because: - Less redundancy. - Better suffixes. Also a few modification to GPU_texture_create_* to simplify the API: - make the format explicit to the texture creation process. - remove the component count as it's specified in the GPUTextureFormat.
2018-04-30 16:02:24 +02:00
tex = DRW_texture_create_1D(samples, GPU_RG16F, DRW_TEX_WRAP, (float *)texels);
Eevee: Replace Cubemaps by octahedron maps for env. probes. This enables us to use 2D texture arrays for multiple probes. There is a little artifact with very high roughness caused elongated pixel due to the projection (along every 90° meridian).
2017-05-29 22:03:57 +02:00
MEM_freeN(texels);
return tex;
}
Eevee: Initial implementation of planar reflections. Still pretty barebone: No roughness support, No normal distortion support.
2017-06-17 00:08:03 +02:00
static void planar_pool_ensure_alloc(EEVEE_Data *vedata, int num_planar_ref)
{
/* XXX TODO OPTIMISATION : This is a complete waist of texture memory.
* Instead of allocating each planar probe for each viewport,
* only alloc them once using the biggest viewport resolution. */
EEVEE_TextureList *txl = vedata->txl;
const float *viewport_size = DRW_viewport_size_get();
/* TODO get screen percentage from layer setting */
// const DRWContextState *draw_ctx = DRW_context_state_get();
Rename any instance of scene layer or render layer in code with view layer The RenderResult struct still has a listbase of RenderLayer, but that's ok since this is strictly for rendering. * Subversion bump (to 2.80.2) * DNA low level doversion (renames) - only for .blend created since 2.80 started Note: We can't use DNA_struct_elem_find or get file version in init_structDNA, so we are manually iterating over the array of the SDNA elements instead. Note 2: This doversion change with renames can be reverted in a few months. But so far it's required for 2.8 files created between October 2016 and now. Reviewers: campbellbarton, sergey Differential Revision: https://developer.blender.org/D2927
2017-11-22 10:52:39 -02:00
// ViewLayer *view_layer = draw_ctx->view_layer;
Eevee: Initial implementation of planar reflections. Still pretty barebone: No roughness support, No normal distortion support.
2017-06-17 00:08:03 +02:00
float screen_percentage = 1.0f;
int width = (int)(viewport_size[0] * screen_percentage);
int height = (int)(viewport_size[1] * screen_percentage);
/* We need an Array texture so allocate it ourself */
Eevee: Fix OpenGl errors. Also assert if texture does not exists in draw manager. Keeping it sane.
2017-06-23 01:33:28 +02:00
if (!txl->planar_pool) {
if (num_planar_ref > 0) {
txl->planar_pool = DRW_texture_create_2D_array(width, height, max_ff(1, num_planar_ref),
DRW: Remove DRWTextureFormat in favor or GPUTextureFormat. Because: - Less redundancy. - Better suffixes. Also a few modification to GPU_texture_create_* to simplify the API: - make the format explicit to the texture creation process. - remove the component count as it's specified in the GPUTextureFormat.
2018-04-30 16:02:24 +02:00
GPU_R11F_G11F_B10F, DRW_TEX_FILTER | DRW_TEX_MIPMAP, NULL);
Eevee: Planar Reflection: Remove distance approximation. This commit separate the depth texture into another texture array. This remove the need to output radial depth into alpha. Unfortunatly it's difficult to recover position from the non linear depth buffer when applying reflection without adding a bunch of stuff. This is in preparation of SSR planar reflections.
2017-07-23 14:03:27 +02:00
txl->planar_depth = DRW_texture_create_2D_array(width, height, max_ff(1, num_planar_ref),
DRW: Remove DRWTextureFormat in favor or GPUTextureFormat. Because: - Less redundancy. - Better suffixes. Also a few modification to GPU_texture_create_* to simplify the API: - make the format explicit to the texture creation process. - remove the component count as it's specified in the GPUTextureFormat.
2018-04-30 16:02:24 +02:00
GPU_DEPTH_COMPONENT24, 0, NULL);
Eevee: Fix OpenGl errors. Also assert if texture does not exists in draw manager. Keeping it sane.
2017-06-23 01:33:28 +02:00
}
else if (num_planar_ref == 0) {
/* Makes Opengl Happy : Create a placeholder texture that will never be sampled but still bound to shader. */
DRW: Remove DRWTextureFormat in favor or GPUTextureFormat. Because: - Less redundancy. - Better suffixes. Also a few modification to GPU_texture_create_* to simplify the API: - make the format explicit to the texture creation process. - remove the component count as it's specified in the GPUTextureFormat.
2018-04-30 16:02:24 +02:00
txl->planar_pool = DRW_texture_create_2D_array(1, 1, 1, GPU_RGBA8, DRW_TEX_FILTER | DRW_TEX_MIPMAP, NULL);
txl->planar_depth = DRW_texture_create_2D_array(1, 1, 1, GPU_DEPTH_COMPONENT24, 0, NULL);
Eevee: Fix OpenGl errors. Also assert if texture does not exists in draw manager. Keeping it sane.
2017-06-23 01:33:28 +02:00
}
Eevee: Initial implementation of planar reflections. Still pretty barebone: No roughness support, No normal distortion support.
2017-06-17 00:08:03 +02:00
}
}
Eevee: CodeStyle: Separate Shader initialization.
2017-12-04 17:21:17 +01:00
static void lightprobe_shaders_init(void)
Eevee: Introduction of world preconvolved envmap. For now only compute GGX convolution. The GGX LUT used for the split sum approximation (UE4) is merged with the LTX mag LUT that uses the same parameters (theta and roughness)
2017-04-18 12:50:09 +02:00
{
Eevee: CodeStyle: Separate Shader initialization.
2017-12-04 17:21:17 +01:00
const char *filter_defines = "#define HAMMERSLEY_SIZE " STRINGIFY(HAMMERSLEY_SIZE) "\n"
Eevee: Add Grid debug display.
2017-06-14 13:45:54 +02:00
#if defined(IRRADIANCE_SH_L2)
Eevee: CodeStyle: Separate Shader initialization.
2017-12-04 17:21:17 +01:00
"#define IRRADIANCE_SH_L2\n"
Eevee: Add Grid debug display.
2017-06-14 13:45:54 +02:00
#elif defined(IRRADIANCE_CUBEMAP)
Eevee: CodeStyle: Separate Shader initialization.
2017-12-04 17:21:17 +01:00
"#define IRRADIANCE_CUBEMAP\n"
Eevee: Add Grid debug display.
2017-06-14 13:45:54 +02:00
#elif defined(IRRADIANCE_HL2)
Eevee: CodeStyle: Separate Shader initialization.
2017-12-04 17:21:17 +01:00
"#define IRRADIANCE_HL2\n"
Eevee: Add Grid debug display.
2017-06-14 13:45:54 +02:00
#endif
Eevee: CodeStyle: Separate Shader initialization.
2017-12-04 17:21:17 +01:00
"#define NOISE_SIZE 64\n";
Eevee: Add Grid debug display.
2017-06-14 13:45:54 +02:00
Eevee: CodeStyle: Separate Shader initialization.
2017-12-04 17:21:17 +01:00
char *shader_str = NULL;
Eevee: Make use of the new view matrix UBO.
2018-03-10 00:00:04 +01:00
char *vert_str = NULL;
Eevee: Add Grid debug display.
2017-06-14 13:45:54 +02:00
Cleanup: replace BLI_dynstr w/ BLI_string_joinN Use in place of recently reverted 'DRW_shader_create_lib'
2018-01-17 21:40:43 +11:00
shader_str = BLI_string_joinN(
Eevee: Make use of the new view matrix UBO.
2018-03-10 00:00:04 +01:00
datatoc_common_view_lib_glsl,
Eevee: Put all constant uniforms in a global UBO. This is an optimization / cleanup commit. The use of a global ubo remove lots of uniform lookups and only transfert data when needed. Lots of renaming for more consistent codestyle.
2018-01-21 17:25:10 +01:00
datatoc_common_uniforms_lib_glsl,
Cleanup: replace BLI_dynstr w/ BLI_string_joinN Use in place of recently reverted 'DRW_shader_create_lib'
2018-01-17 21:40:43 +11:00
datatoc_bsdf_common_lib_glsl,
datatoc_bsdf_sampling_lib_glsl,
datatoc_lightprobe_filter_glossy_frag_glsl);
Revert "Fix use-after free in DRW_shader_create_with_lib" Caused an error entering edit-mode. d60f26f37a & 2659500835
2018-01-17 20:35:06 +11:00
e_data.probe_filter_glossy_sh = DRW_shader_create(
datatoc_lightprobe_vert_glsl, datatoc_lightprobe_geom_glsl, shader_str, filter_defines);
e_data.probe_default_sh = DRW_shader_create(
datatoc_background_vert_glsl, NULL, datatoc_default_world_frag_glsl, NULL);
EEvEE: LookDev
2018-05-25 08:06:36 +02:00
e_data.probe_default_studiolight_sh = DRW_shader_create(
datatoc_background_vert_glsl, NULL, datatoc_default_world_frag_glsl, "#define LOOKDEV\n");
Eevee: CodeStyle: Separate Shader initialization.
2017-12-04 17:21:17 +01:00
MEM_freeN(shader_str);
Cleanup: replace BLI_dynstr w/ BLI_string_joinN Use in place of recently reverted 'DRW_shader_create_lib'
2018-01-17 21:40:43 +11:00
shader_str = BLI_string_joinN(
Eevee: Make use of the new view matrix UBO.
2018-03-10 00:00:04 +01:00
datatoc_common_view_lib_glsl,
Eevee: Put all constant uniforms in a global UBO. This is an optimization / cleanup commit. The use of a global ubo remove lots of uniform lookups and only transfert data when needed. Lots of renaming for more consistent codestyle.
2018-01-21 17:25:10 +01:00
datatoc_common_uniforms_lib_glsl,
Cleanup: replace BLI_dynstr w/ BLI_string_joinN Use in place of recently reverted 'DRW_shader_create_lib'
2018-01-17 21:40:43 +11:00
datatoc_bsdf_common_lib_glsl,
datatoc_bsdf_sampling_lib_glsl,
datatoc_lightprobe_filter_diffuse_frag_glsl);
Eevee: Probe Grid: Clear Grid buffers with world diffuse coefs. This make sure the values displayed by the "show data" sphere are initialized. Also this make the bounce lighting progress more apparent.
2017-09-28 21:17:57 +02:00
Eevee: CodeStyle: Separate Shader initialization.
2017-12-04 17:21:17 +01:00
e_data.probe_filter_diffuse_sh = DRW_shader_create_fullscreen(shader_str, filter_defines);
Revert "Fix use-after free in DRW_shader_create_with_lib" Caused an error entering edit-mode. d60f26f37a & 2659500835
2018-01-17 20:35:06 +11:00
Eevee: CodeStyle: Separate Shader initialization.
2017-12-04 17:21:17 +01:00
MEM_freeN(shader_str);
Eevee: Probes: Add data display for cubemaps.
2017-06-15 00:10:34 +02:00
Cleanup: replace BLI_dynstr w/ BLI_string_joinN Use in place of recently reverted 'DRW_shader_create_lib'
2018-01-17 21:40:43 +11:00
shader_str = BLI_string_joinN(
Eevee: Make use of the new view matrix UBO.
2018-03-10 00:00:04 +01:00
datatoc_common_view_lib_glsl,
Eevee: Put all constant uniforms in a global UBO. This is an optimization / cleanup commit. The use of a global ubo remove lots of uniform lookups and only transfert data when needed. Lots of renaming for more consistent codestyle.
2018-01-21 17:25:10 +01:00
datatoc_common_uniforms_lib_glsl,
Cleanup: replace BLI_dynstr w/ BLI_string_joinN Use in place of recently reverted 'DRW_shader_create_lib'
2018-01-17 21:40:43 +11:00
datatoc_bsdf_common_lib_glsl,
datatoc_bsdf_sampling_lib_glsl,
datatoc_lightprobe_filter_visibility_frag_glsl);
Eevee: Probes: Add data display for cubemaps.
2017-06-15 00:10:34 +02:00
Eevee: CodeStyle: Separate Shader initialization.
2017-12-04 17:21:17 +01:00
e_data.probe_filter_visibility_sh = DRW_shader_create_fullscreen(shader_str, filter_defines);
Eevee: Add data display for planar reflection. Maybe not very useful but it's here for feature parity with the other probes. The fragment shader clipping is here because I'm lazy and don't want to creating a proper quad Batch.
2017-06-17 01:59:01 +02:00
Eevee: CodeStyle: Separate Shader initialization.
2017-12-04 17:21:17 +01:00
MEM_freeN(shader_str);
Eevee: Add Planar reflection blurring. This method is very cheap and inaccurate. This will fill the gap untill better model is supported.
2017-06-26 20:37:41 +02:00
Cleanup: replace BLI_dynstr w/ BLI_string_joinN Use in place of recently reverted 'DRW_shader_create_lib'
2018-01-17 21:40:43 +11:00
shader_str = BLI_string_joinN(
datatoc_octahedron_lib_glsl,
Eevee: Make use of the new view matrix UBO.
2018-03-10 00:00:04 +01:00
datatoc_common_view_lib_glsl,
Eevee: Put all constant uniforms in a global UBO. This is an optimization / cleanup commit. The use of a global ubo remove lots of uniform lookups and only transfert data when needed. Lots of renaming for more consistent codestyle.
2018-01-21 17:25:10 +01:00
datatoc_common_uniforms_lib_glsl,
Cleanup: replace BLI_dynstr w/ BLI_string_joinN Use in place of recently reverted 'DRW_shader_create_lib'
2018-01-17 21:40:43 +11:00
datatoc_bsdf_common_lib_glsl,
datatoc_irradiance_lib_glsl,
datatoc_lightprobe_lib_glsl,
datatoc_lightprobe_grid_display_frag_glsl);
Revert "Fix use-after free in DRW_shader_create_with_lib" Caused an error entering edit-mode. d60f26f37a & 2659500835
2018-01-17 20:35:06 +11:00
Eevee: Make use of the new view matrix UBO.
2018-03-10 00:00:04 +01:00
vert_str = BLI_string_joinN(
datatoc_common_view_lib_glsl,
datatoc_lightprobe_grid_display_vert_glsl);
Revert "Fix use-after free in DRW_shader_create_with_lib" Caused an error entering edit-mode. d60f26f37a & 2659500835
2018-01-17 20:35:06 +11:00
Eevee: Make use of the new view matrix UBO.
2018-03-10 00:00:04 +01:00
e_data.probe_grid_display_sh = DRW_shader_create(vert_str, NULL, shader_str, filter_defines);
MEM_freeN(vert_str);
Eevee: CodeStyle: Separate Shader initialization.
2017-12-04 17:21:17 +01:00
MEM_freeN(shader_str);
Revert "Fix use-after free in DRW_shader_create_with_lib" Caused an error entering edit-mode. d60f26f37a & 2659500835
2018-01-17 20:35:06 +11:00
e_data.probe_grid_fill_sh = DRW_shader_create_fullscreen(
datatoc_lightprobe_grid_fill_frag_glsl, filter_defines);
Cleanup: replace BLI_dynstr w/ BLI_string_joinN Use in place of recently reverted 'DRW_shader_create_lib'
2018-01-17 21:40:43 +11:00
shader_str = BLI_string_joinN(
datatoc_octahedron_lib_glsl,
Eevee: Make use of the new view matrix UBO.
2018-03-10 00:00:04 +01:00
datatoc_common_view_lib_glsl,
Eevee: Put all constant uniforms in a global UBO. This is an optimization / cleanup commit. The use of a global ubo remove lots of uniform lookups and only transfert data when needed. Lots of renaming for more consistent codestyle.
2018-01-21 17:25:10 +01:00
datatoc_common_uniforms_lib_glsl,
Cleanup: replace BLI_dynstr w/ BLI_string_joinN Use in place of recently reverted 'DRW_shader_create_lib'
2018-01-17 21:40:43 +11:00
datatoc_bsdf_common_lib_glsl,
datatoc_lightprobe_lib_glsl,
datatoc_lightprobe_cube_display_frag_glsl);
Eevee: CodeStyle: Separate Shader initialization.
2017-12-04 17:21:17 +01:00
Eevee: Make use of the new view matrix UBO.
2018-03-10 00:00:04 +01:00
vert_str = BLI_string_joinN(
datatoc_common_view_lib_glsl,
datatoc_lightprobe_cube_display_vert_glsl);
e_data.probe_cube_display_sh = DRW_shader_create(vert_str, NULL, shader_str, NULL);
Revert "Fix use-after free in DRW_shader_create_with_lib" Caused an error entering edit-mode. d60f26f37a & 2659500835
2018-01-17 20:35:06 +11:00
Eevee: Make use of the new view matrix UBO.
2018-03-10 00:00:04 +01:00
MEM_freeN(vert_str);
Eevee: CodeStyle: Separate Shader initialization.
2017-12-04 17:21:17 +01:00
MEM_freeN(shader_str);
Eevee: Make use of the new view matrix UBO.
2018-03-10 00:00:04 +01:00
vert_str = BLI_string_joinN(
datatoc_common_view_lib_glsl,
datatoc_lightprobe_planar_display_vert_glsl);
shader_str = BLI_string_joinN(
datatoc_common_view_lib_glsl,
datatoc_lightprobe_planar_display_frag_glsl);
e_data.probe_planar_display_sh = DRW_shader_create(vert_str, NULL, shader_str, NULL);
MEM_freeN(vert_str);
MEM_freeN(shader_str);
Eevee: CodeStyle: Separate Shader initialization.
2017-12-04 17:21:17 +01:00
e_data.probe_planar_downsample_sh = DRW_shader_create(
datatoc_lightprobe_planar_downsample_vert_glsl,
datatoc_lightprobe_planar_downsample_geom_glsl,
datatoc_lightprobe_planar_downsample_frag_glsl,
NULL);
e_data.hammersley = create_hammersley_sample_texture(HAMMERSLEY_SIZE);
}
void EEVEE_lightprobes_init(EEVEE_ViewLayerData *sldata, EEVEE_Data *UNUSED(vedata))
{
Eevee: Put all constant uniforms in a global UBO. This is an optimization / cleanup commit. The use of a global ubo remove lots of uniform lookups and only transfert data when needed. Lots of renaming for more consistent codestyle.
2018-01-21 17:25:10 +01:00
EEVEE_CommonUniformBuffer *common_data = &sldata->common_data;
Eevee: CodeStyle: Separate Shader initialization.
2017-12-04 17:21:17 +01:00
bool update_all = false;
const DRWContextState *draw_ctx = DRW_context_state_get();
Move EEVEE properties into scene We handle doversion for the scene properties, but not for the view layer overrides. Overrides will be implemented in a different way via dynamic overrides. For now this data is completely lost.
2018-05-16 19:34:24 +02:00
const Scene *scene_eval = DEG_get_evaluated_scene(draw_ctx->depsgraph);
Eevee: CodeStyle: Separate Shader initialization.
2017-12-04 17:21:17 +01:00
/* Shaders */
if (!e_data.probe_filter_glossy_sh) {
lightprobe_shaders_init();
Eevee: Replace Cubemaps by octahedron maps for env. probes. This enables us to use 2D texture arrays for multiple probes. There is a little artifact with very high roughness caused elongated pixel due to the projection (along every 90° meridian).
2017-05-29 22:03:57 +02:00
}
Eevee: Put shadows and probes inside SceneLayerEngineData This remove the duplication of data for each viewport improving memory usage.
2017-05-30 22:29:20 +02:00
if (!sldata->probes) {
Rename probe to light-probe Probe is a real general term, the new name is used often in docs online.
2017-06-12 20:59:54 +10:00
sldata->probes = MEM_callocN(sizeof(EEVEE_LightProbesInfo), "EEVEE_LightProbesInfo");
Eevee: Probe Grid: Clear Grid buffers with world diffuse coefs. This make sure the values displayed by the "show data" sphere are initialized. Also this make the bounce lighting progress more apparent.
2017-09-28 21:17:57 +02:00
sldata->probes->grid_initialized = false;
Rename probe to light-probe Probe is a real general term, the new name is used often in docs online.
2017-06-12 20:59:54 +10:00
sldata->probe_ubo = DRW_uniformbuffer_create(sizeof(EEVEE_LightProbe) * MAX_PROBE, NULL);
Eevee: Add Irradiance Grid support Early implementation. Slow and still has quality 3 ways of storing irradiance: - Spherical Harmonics: Have problem with directionnal lighting. - HL2 diffuse cube: Very low resolution but smooth transitions. - Diffuse cube: High storage requirement. Also include some name change.
2017-06-13 17:39:39 +02:00
sldata->grid_ubo = DRW_uniformbuffer_create(sizeof(EEVEE_LightGrid) * MAX_GRID, NULL);
Eevee: Initial implementation of planar reflections. Still pretty barebone: No roughness support, No normal distortion support.
2017-06-17 00:08:03 +02:00
sldata->planar_ubo = DRW_uniformbuffer_create(sizeof(EEVEE_PlanarReflection) * MAX_PLANAR, NULL);
Eevee: Introduction of world preconvolved envmap. For now only compute GGX convolution. The GGX LUT used for the split sum approximation (UE4) is merged with the LTX mag LUT that uses the same parameters (theta and roughness)
2017-04-18 12:50:09 +02:00
}
Eevee: Update to support shader deferred compilation. World probe is tagged to refresh when it it's shader is updated. Probes are recomputed only after all meshes shaders have been compiled.
2018-03-06 02:19:28 +01:00
/* Only start doing probes if all materials have finished compiling. */
sldata->probes->all_materials_updated = true;
Eevee: Put all constant uniforms in a global UBO. This is an optimization / cleanup commit. The use of a global ubo remove lots of uniform lookups and only transfert data when needed. Lots of renaming for more consistent codestyle.
2018-01-21 17:25:10 +01:00
common_data->spec_toggle = true;
common_data->ssr_toggle = true;
common_data->sss_toggle = true;
Move EEVEE properties into scene We handle doversion for the scene properties, but not for the view layer overrides. Overrides will be implemented in a different way via dynamic overrides. For now this data is completely lost.
2018-05-16 19:34:24 +02:00
int prop_bounce_num = scene_eval->eevee.gi_diffuse_bounces;
Eevee : Probe Grid : Fix Update tagging.
2017-10-02 18:40:51 +02:00
if (sldata->probes->num_bounce != prop_bounce_num) {
Eevee: Irradiance Visibility: Initial Implementation This augment the existing irradiance grid with a new visibility precomputation. We store a small shadowmap for each grid sample so that light does not leak through walls and such. The visibility parameter are similar to the one used by the Variance Shadow Map for point lights. Technical details: We store the visibility in the same texture (array) as the irradiance itself (in order to reduce the number of sampler). But the irradiance and the visibility are not the same data so we must encode them in order to use the same texture format. We use RGBA8 normalized texture and encode irradiance as RGBE (shared exponent). Using RGBE encoding instead of R11_G11_B10 may lead to some lighting changes, but quality seems to be nearly the same in my test cases. Using full RGBA16/32F maybe a future option but that will require much more memory and reduce the perf significantly. Visibility moments (VSM) are encoded as 16bits fixed point precision using a special range. This seems to retain enough precision for the needs. Also interpolation does not seems to be big problem (even though it's incorrect).
2017-12-02 14:28:29 +01:00
sldata->probes->num_bounce = prop_bounce_num;
update_all = true;
Eevee : Probe Grid : Fix Update tagging.
2017-10-02 18:40:51 +02:00
}
Eevee : Add a setting for the number of indirect light bounce. This is used to tweak the overall spread of the lighting. It is a per renderlayer setting.
2017-10-01 02:19:10 +02:00
Move EEVEE properties into scene We handle doversion for the scene properties, but not for the view layer overrides. Overrides will be implemented in a different way via dynamic overrides. For now this data is completely lost.
2018-05-16 19:34:24 +02:00
int prop_cubemap_res = scene_eval->eevee.gi_cubemap_resolution;
Eevee: Probe: Add Cubemap Resolution option.
2017-10-10 21:56:11 +02:00
if (sldata->probes->cubemap_res != prop_cubemap_res) {
sldata->probes->cubemap_res = prop_cubemap_res;
Eevee: Irradiance Visibility: Initial Implementation This augment the existing irradiance grid with a new visibility precomputation. We store a small shadowmap for each grid sample so that light does not leak through walls and such. The visibility parameter are similar to the one used by the Variance Shadow Map for point lights. Technical details: We store the visibility in the same texture (array) as the irradiance itself (in order to reduce the number of sampler). But the irradiance and the visibility are not the same data so we must encode them in order to use the same texture format. We use RGBA8 normalized texture and encode irradiance as RGBE (shared exponent). Using RGBE encoding instead of R11_G11_B10 may lead to some lighting changes, but quality seems to be nearly the same in my test cases. Using full RGBA16/32F maybe a future option but that will require much more memory and reduce the perf significantly. Visibility moments (VSM) are encoded as 16bits fixed point precision using a special range. This seems to retain enough precision for the needs. Also interpolation does not seems to be big problem (even though it's incorrect).
2017-12-02 14:28:29 +01:00
update_all = true;
Eevee: Probe: Add Cubemap Resolution option.
2017-10-10 21:56:11 +02:00
sldata->probes->target_size = prop_cubemap_res >> 1;
Eevee: Fix bad framebuffer when changing probe resolution.
2018-03-05 00:49:24 +01:00
DRW_TEXTURE_FREE_SAFE(sldata->probe_depth_rt);
Eevee: Probe: Add Cubemap Resolution option.
2017-10-10 21:56:11 +02:00
DRW_TEXTURE_FREE_SAFE(sldata->probe_rt);
DRW_TEXTURE_FREE_SAFE(sldata->probe_pool);
}
Move EEVEE properties into scene We handle doversion for the scene properties, but not for the view layer overrides. Overrides will be implemented in a different way via dynamic overrides. For now this data is completely lost.
2018-05-16 19:34:24 +02:00
const int visibility_res = scene_eval->eevee.gi_visibility_resolution;
Eevee: Put all constant uniforms in a global UBO. This is an optimization / cleanup commit. The use of a global ubo remove lots of uniform lookups and only transfert data when needed. Lots of renaming for more consistent codestyle.
2018-01-21 17:25:10 +01:00
if (common_data->prb_irradiance_vis_size != visibility_res) {
common_data->prb_irradiance_vis_size = visibility_res;
Eevee: Irradiance Visibility: Initial Implementation This augment the existing irradiance grid with a new visibility precomputation. We store a small shadowmap for each grid sample so that light does not leak through walls and such. The visibility parameter are similar to the one used by the Variance Shadow Map for point lights. Technical details: We store the visibility in the same texture (array) as the irradiance itself (in order to reduce the number of sampler). But the irradiance and the visibility are not the same data so we must encode them in order to use the same texture format. We use RGBA8 normalized texture and encode irradiance as RGBE (shared exponent). Using RGBE encoding instead of R11_G11_B10 may lead to some lighting changes, but quality seems to be nearly the same in my test cases. Using full RGBA16/32F maybe a future option but that will require much more memory and reduce the perf significantly. Visibility moments (VSM) are encoded as 16bits fixed point precision using a special range. This seems to retain enough precision for the needs. Also interpolation does not seems to be big problem (even though it's incorrect).
2017-12-02 14:28:29 +01:00
update_all = true;
Eevee: Fix opengl error cause by bad texture configuration. This is fragile and the whole Texture/Framebuffer should be extended to be more flexible.
2017-06-24 01:46:07 +02:00
}
Eevee: Irradiance Visibility: Initial Implementation This augment the existing irradiance grid with a new visibility precomputation. We store a small shadowmap for each grid sample so that light does not leak through walls and such. The visibility parameter are similar to the one used by the Variance Shadow Map for point lights. Technical details: We store the visibility in the same texture (array) as the irradiance itself (in order to reduce the number of sampler). But the irradiance and the visibility are not the same data so we must encode them in order to use the same texture format. We use RGBA8 normalized texture and encode irradiance as RGBE (shared exponent). Using RGBE encoding instead of R11_G11_B10 may lead to some lighting changes, but quality seems to be nearly the same in my test cases. Using full RGBA16/32F maybe a future option but that will require much more memory and reduce the perf significantly. Visibility moments (VSM) are encoded as 16bits fixed point precision using a special range. This seems to retain enough precision for the needs. Also interpolation does not seems to be big problem (even though it's incorrect).
2017-12-02 14:28:29 +01:00
if (update_all) {
Eevee: Update to support shader deferred compilation. World probe is tagged to refresh when it it's shader is updated. Probes are recomputed only after all meshes shaders have been compiled.
2018-03-06 02:19:28 +01:00
sldata->probes->update_world |= PROBE_UPDATE_ALL;
Eevee: Irradiance Visibility: Initial Implementation This augment the existing irradiance grid with a new visibility precomputation. We store a small shadowmap for each grid sample so that light does not leak through walls and such. The visibility parameter are similar to the one used by the Variance Shadow Map for point lights. Technical details: We store the visibility in the same texture (array) as the irradiance itself (in order to reduce the number of sampler). But the irradiance and the visibility are not the same data so we must encode them in order to use the same texture format. We use RGBA8 normalized texture and encode irradiance as RGBE (shared exponent). Using RGBE encoding instead of R11_G11_B10 may lead to some lighting changes, but quality seems to be nearly the same in my test cases. Using full RGBA16/32F maybe a future option but that will require much more memory and reduce the perf significantly. Visibility moments (VSM) are encoded as 16bits fixed point precision using a special range. This seems to retain enough precision for the needs. Also interpolation does not seems to be big problem (even though it's incorrect).
2017-12-02 14:28:29 +01:00
sldata->probes->updated_bounce = 0;
sldata->probes->grid_initialized = false;
}
Eevee: Fix opengl error cause by bad texture configuration. This is fragile and the whole Texture/Framebuffer should be extended to be more flexible.
2017-06-24 01:46:07 +02:00
Eevee: Irradiance Visibility: Initial Implementation This augment the existing irradiance grid with a new visibility precomputation. We store a small shadowmap for each grid sample so that light does not leak through walls and such. The visibility parameter are similar to the one used by the Variance Shadow Map for point lights. Technical details: We store the visibility in the same texture (array) as the irradiance itself (in order to reduce the number of sampler). But the irradiance and the visibility are not the same data so we must encode them in order to use the same texture format. We use RGBA8 normalized texture and encode irradiance as RGBE (shared exponent). Using RGBE encoding instead of R11_G11_B10 may lead to some lighting changes, but quality seems to be nearly the same in my test cases. Using full RGBA16/32F maybe a future option but that will require much more memory and reduce the perf significantly. Visibility moments (VSM) are encoded as 16bits fixed point precision using a special range. This seems to retain enough precision for the needs. Also interpolation does not seems to be big problem (even though it's incorrect).
2017-12-02 14:28:29 +01:00
/* Setup Render Target Cubemap */
Eevee: Put shadows and probes inside SceneLayerEngineData This remove the duplication of data for each viewport improving memory usage.
2017-05-30 22:29:20 +02:00
if (!sldata->probe_rt) {
DRW: Remove DRWTextureFormat in favor or GPUTextureFormat. Because: - Less redundancy. - Better suffixes. Also a few modification to GPU_texture_create_* to simplify the API: - make the format explicit to the texture creation process. - remove the component count as it's specified in the GPUTextureFormat.
2018-04-30 16:02:24 +02:00
sldata->probe_depth_rt = DRW_texture_create_cube(sldata->probes->target_size, GPU_DEPTH_COMPONENT24, 0, NULL);
sldata->probe_rt = DRW_texture_create_cube(sldata->probes->target_size, GPU_RGBA16F, DRW_TEX_FILTER | DRW_TEX_MIPMAP, NULL);
Eevee: Introduction of world preconvolved envmap. For now only compute GGX convolution. The GGX LUT used for the split sum approximation (UE4) is merged with the LTX mag LUT that uses the same parameters (theta and roughness)
2017-04-18 12:50:09 +02:00
}
GPUFramebuffer: Refactor (Part 2) This refactor modernise the use of framebuffers. It also touches a lot of files so breaking down changes we have: - GPUTexture: Allow textures to be attached to more than one GPUFrameBuffer. This allows to create and configure more FBO without the need to attach and detach texture at drawing time. - GPUFrameBuffer: The wrapper starts to mimic opengl a bit closer. This allows to configure the framebuffer inside a context other than the one that will be rendering the framebuffer. We do the actual configuration when binding the FBO. We also Keep track of config validity and save drawbuffers state in the FBO. We remove the different bind/unbind functions. These make little sense now that we have separate contexts. - DRWFrameBuffer: We replace DRW_framebuffer functions by GPU_framebuffer ones to avoid another layer of abstraction. We move the DRW convenience functions to GPUFramebuffer instead and even add new ones. The MACRO GPU_framebuffer_ensure_config is pretty much all you need to create and config a GPUFramebuffer. - DRWTexture: Due to the removal of DRWFrameBuffer, we needed to create functions to create textures for thoses framebuffers. Pool textures are now using default texture parameters for the texture type asked. - DRWManager: Make sure no framebuffer object is bound when doing cache filling. - GPUViewport: Add new color_only_fb and depth_only_fb along with FB API usage update. This let draw engines render to color/depth only target and without the need to attach/detach textures. - WM_window: Assert when a framebuffer is bound when changing context. This balance the fact we are not track ogl context inside GPUFramebuffer. - Eevee, Clay, Mode engines: Update to new API. This comes with a lot of code simplification. This also come with some cleanups in some engine codes.
2018-03-25 17:46:48 +02:00
for (int i = 0; i < 6; ++i) {
GPU_framebuffer_ensure_config(&sldata->probe_face_fb[i], {
GPU_ATTACHMENT_TEXTURE_CUBEFACE(sldata->probe_depth_rt, i),
GPU_ATTACHMENT_TEXTURE_CUBEFACE(sldata->probe_rt, i)
});
}
Eevee: Planar reflection: Fix bad texture binding.
2017-06-23 19:02:40 +02:00
/* Placeholder planar pool: used when rendering planar reflections (avoid dependency loop). */
if (!e_data.planar_pool_placeholder) {
DRW: Remove DRWTextureFormat in favor or GPUTextureFormat. Because: - Less redundancy. - Better suffixes. Also a few modification to GPU_texture_create_* to simplify the API: - make the format explicit to the texture creation process. - remove the component count as it's specified in the GPUTextureFormat.
2018-04-30 16:02:24 +02:00
e_data.planar_pool_placeholder = DRW_texture_create_2D_array(1, 1, 1, GPU_RGBA8, DRW_TEX_FILTER, NULL);
Eevee: Planar reflection: Fix bad texture binding.
2017-06-23 19:02:40 +02:00
}
Eevee: HiZ buffer: Split into two 24bit depth buffer This way we don't have float precision issue we had before and we save some bandwidth.
2017-07-21 14:27:16 +02:00
if (!e_data.depth_placeholder) {
DRW: Remove DRWTextureFormat in favor or GPUTextureFormat. Because: - Less redundancy. - Better suffixes. Also a few modification to GPU_texture_create_* to simplify the API: - make the format explicit to the texture creation process. - remove the component count as it's specified in the GPUTextureFormat.
2018-04-30 16:02:24 +02:00
e_data.depth_placeholder = DRW_texture_create_2D(1, 1, GPU_DEPTH_COMPONENT24, 0, NULL);
Eevee: HiZ buffer: Split into two 24bit depth buffer This way we don't have float precision issue we had before and we save some bandwidth.
2017-07-21 14:27:16 +02:00
}
Eevee: Planar Reflection: Remove distance approximation. This commit separate the depth texture into another texture array. This remove the need to output radial depth into alpha. Unfortunatly it's difficult to recover position from the non linear depth buffer when applying reflection without adding a bunch of stuff. This is in preparation of SSR planar reflections.
2017-07-23 14:03:27 +02:00
if (!e_data.depth_array_placeholder) {
DRW: Remove DRWTextureFormat in favor or GPUTextureFormat. Because: - Less redundancy. - Better suffixes. Also a few modification to GPU_texture_create_* to simplify the API: - make the format explicit to the texture creation process. - remove the component count as it's specified in the GPUTextureFormat.
2018-04-30 16:02:24 +02:00
e_data.depth_array_placeholder = DRW_texture_create_2D_array(1, 1, 1, GPU_DEPTH_COMPONENT24, 0, NULL);
Eevee: Planar Reflection: Remove distance approximation. This commit separate the depth texture into another texture array. This remove the need to output radial depth into alpha. Unfortunatly it's difficult to recover position from the non linear depth buffer when applying reflection without adding a bunch of stuff. This is in preparation of SSR planar reflections.
2017-07-23 14:03:27 +02:00
}
Eevee: Introduction of world preconvolved envmap. For now only compute GGX convolution. The GGX LUT used for the split sum approximation (UE4) is merged with the LTX mag LUT that uses the same parameters (theta and roughness)
2017-04-18 12:50:09 +02:00
}
Rename any instance of scene layer or render layer in code with view layer The RenderResult struct still has a listbase of RenderLayer, but that's ok since this is strictly for rendering. * Subversion bump (to 2.80.2) * DNA low level doversion (renames) - only for .blend created since 2.80 started Note: We can't use DNA_struct_elem_find or get file version in init_structDNA, so we are manually iterating over the array of the SDNA elements instead. Note 2: This doversion change with renames can be reverted in a few months. But so far it's required for 2.8 files created between October 2016 and now. Reviewers: campbellbarton, sergey Differential Revision: https://developer.blender.org/D2927
2017-11-22 10:52:39 -02:00
void EEVEE_lightprobes_cache_init(EEVEE_ViewLayerData *sldata, EEVEE_Data *vedata)
Eevee: Introduction of world preconvolved envmap. For now only compute GGX convolution. The GGX LUT used for the split sum approximation (UE4) is merged with the LTX mag LUT that uses the same parameters (theta and roughness)
2017-04-18 12:50:09 +02:00
{
Eevee: Add Planar reflection blurring. This method is very cheap and inaccurate. This will fill the gap untill better model is supported.
2017-06-26 20:37:41 +02:00
EEVEE_TextureList *txl = vedata->txl;
EEVEE_PassList *psl = vedata->psl;
EEVEE_StorageList *stl = vedata->stl;
Rename probe to light-probe Probe is a real general term, the new name is used often in docs online.
2017-06-12 20:59:54 +10:00
EEVEE_LightProbesInfo *pinfo = sldata->probes;
Eevee: Material code refactor. Separate material handling inside another file. Make use of enums to identify shader variations. Group all 64*64 LUTs into one array texture. Only update world probe if world changes.
2017-06-04 12:12:58 +02:00
EEVEE: LightProbe: Implement Visibility Group. Works on every probe type. The function to see is EEVEE_lightprobes_obj_visibility_cb. Set pinfo->vis_data.cached to true makes the computation faster for multiple views using the same group. We could even sort the probes by group for that mater to speed things up even more (only applies to dynamic probes like the planar reflections because other probes are only rendered one at a time).
2018-04-24 12:45:59 +02:00
/* Make sure no aditionnal visibility check runs by default. */
Collections and groups unification OVERVIEW * In 2.7 terminology, all layers and groups are now collection datablocks. * These collections are nestable, linkable, instanceable, overrideable, .. which opens up new ways to set up scenes and link + override data. * Viewport/render visibility and selectability are now a part of the collection and shared across all view layers and linkable. * View layers define which subset of the scene collection hierarchy is excluded for each. For many workflows one view layer can be used, these are more of an advanced feature now. OUTLINER * The outliner now has a "View Layer" display mode instead of "Collections", which can display the collections and/or objects in the view layer. * In this display mode, collections can be excluded with the right click menu. These will then be greyed out and their objects will be excluded. * To view collections not linked to any scene, the "Blender File" display mode can be used, with the new filtering option to just see Colleciton datablocks. * The outliner right click menus for collections and objects were reorganized. * Drag and drop still needs to be improved. Like before, dragging the icon or text gives different results, we'll unify this later. LINKING AND OVERRIDES * Collections can now be linked into the scene without creating an instance, with the link/append operator or from the collections view in the outliner. * Collections can get static overrides with the right click menu in the outliner, but this is rather unreliable and not clearly communicated at the moment. * We still need to improve the make override operator to turn collection instances into collections with overrides directly in the scene. PERFORMANCE * We tried to make performance not worse than before and improve it in some cases. The main thing that's still a bit slower is multiple scenes, we have to change the layer syncing to only updated affected scenes. * Collections keep a list of their parent collections for faster incremental updates in syncing and caching. * View layer bases are now in a object -> base hash to avoid quadratic time lookups internally and in API functions like visible_get(). VERSIONING * Compatibility with 2.7 files should be improved due to the new visibility controls. Of course users may not want to set up their scenes differently now to avoid having separate layers and groups. * Compatibility with 2.8 is mostly there, and was tested on Eevee demo and Hero files. There's a few things which are know to be not quite compatible, like nested layer collections inside groups. * The versioning code for 2.8 files is quite complicated, and isolated behind #ifdef so it can be removed at the end of the release cycle. KNOWN ISSUES * The G-key group operators in the 3D viewport were left mostly as is, they need to be modified still to fit better. * Same for the groups panel in the object properties. This needs to be updated still, or perhaps replaced by something better. * Collections must all have a unique name. Less restrictive namespacing is to be done later, we'll have to see how important this is as all objects within the collections must also have a unique name anyway. * Full scene copy and delete scene are exactly doing the right thing yet. Differential Revision: https://developer.blender.org/D3383 https://code.blender.org/2018/05/collections-and-groups/
2018-04-30 15:57:22 +02:00
pinfo->vis_data.collection = NULL;
EEVEE: LightProbe: Implement Visibility Group. Works on every probe type. The function to see is EEVEE_lightprobes_obj_visibility_cb. Set pinfo->vis_data.cached to true makes the computation faster for multiple views using the same group. We could even sort the probes by group for that mater to speed things up even more (only applies to dynamic probes like the planar reflections because other probes are only rendered one at a time).
2018-04-24 12:45:59 +02:00
Eevee: Put all constant uniforms in a global UBO. This is an optimization / cleanup commit. The use of a global ubo remove lots of uniform lookups and only transfert data when needed. Lots of renaming for more consistent codestyle.
2018-01-21 17:25:10 +01:00
pinfo->do_cube_update = false;
Eevee: Move Spherical Harmonics to a new Probe UBO. Keep data packing tight to prevent use of padding floats
2017-06-04 16:50:22 +02:00
pinfo->num_cube = 1; /* at least one for the world */
Eevee: Irradiance grid: support for non-blocking update and multiple bounces.
2017-06-14 18:52:53 +02:00
pinfo->num_grid = 1;
Eevee: Initial implementation of planar reflections. Still pretty barebone: No roughness support, No normal distortion support.
2017-06-17 00:08:03 +02:00
pinfo->num_planar = 0;
Eevee: Irradiance Grid: Allocate needed resources instead of a static chunck. This commit makes the irradiance pool and render target sizes depend on the number of irradiance sample in the whole ViewLayer.
2017-12-04 17:20:20 +01:00
pinfo->total_irradiance_samples = 1;
Eevee: Add Irradiance Grid support Early implementation. Slow and still has quality 3 ways of storing irradiance: - Spherical Harmonics: Have problem with directionnal lighting. - HL2 diffuse cube: Very low resolution but smooth transitions. - Diffuse cube: High storage requirement. Also include some name change.
2017-06-13 17:39:39 +02:00
memset(pinfo->probes_cube_ref, 0, sizeof(pinfo->probes_cube_ref));
memset(pinfo->probes_grid_ref, 0, sizeof(pinfo->probes_grid_ref));
Eevee: Initial implementation of planar reflections. Still pretty barebone: No roughness support, No normal distortion support.
2017-06-17 00:08:03 +02:00
memset(pinfo->probes_planar_ref, 0, sizeof(pinfo->probes_planar_ref));
Eevee: Material code refactor. Separate material handling inside another file. Make use of enums to identify shader variations. Group all 64*64 LUTs into one array texture. Only update world probe if world changes.
2017-06-04 12:12:58 +02:00
{
Eevee: Draw background after depth prepass. Reduce overdraw. 0.4ms goes to 0.01ms if background is completely occluded.
2017-07-31 15:15:42 +02:00
psl->probe_background = DRW_pass_create("World Probe Background Pass", DRW_STATE_WRITE_COLOR | DRW_STATE_DEPTH_EQUAL);
Eevee: Material code refactor. Separate material handling inside another file. Make use of enums to identify shader variations. Group all 64*64 LUTs into one array texture. Only update world probe if world changes.
2017-06-04 12:12:58 +02:00
Gawain API naming refactor Use consistent prefix for gawain API names as well as some abbreviations to avoid over-long names, see: D2678
2017-06-19 20:18:04 +10:00
struct Gwn_Batch *geom = DRW_cache_fullscreen_quad_get();
Eevee: Material code refactor. Separate material handling inside another file. Make use of enums to identify shader variations. Group all 64*64 LUTs into one array texture. Only update world probe if world changes.
2017-06-04 12:12:58 +02:00
DRWShadingGroup *grp = NULL;
const DRWContextState *draw_ctx = DRW_context_state_get();
Scene *scene = draw_ctx->scene;
World *wo = scene->world;
float *col = ts.colorBackground;
EEvEE: LookDev
2018-05-25 08:06:36 +02:00
/* LookDev */
EEVEE: LookDev fadeout to the world horizon color
2018-05-29 12:38:19 +02:00
EEVEE_lookdev_cache_init(vedata, &grp, e_data.probe_default_studiolight_sh, psl->probe_background, wo, pinfo);
EEvEE: LookDev
2018-05-25 08:06:36 +02:00
/* END */
if (!grp && wo) {
Eevee: Material code refactor. Separate material handling inside another file. Make use of enums to identify shader variations. Group all 64*64 LUTs into one array texture. Only update world probe if world changes.
2017-06-04 12:12:58 +02:00
col = &wo->horr;
Eevee: Update to support shader deferred compilation. World probe is tagged to refresh when it it's shader is updated. Probes are recomputed only after all meshes shaders have been compiled.
2018-03-06 02:19:28 +01:00
bool wo_sh_compiled = true;
Eevee: Material code refactor. Separate material handling inside another file. Make use of enums to identify shader variations. Group all 64*64 LUTs into one array texture. Only update world probe if world changes.
2017-06-04 12:12:58 +02:00
if (wo->use_nodes && wo->nodetree) {
Eevee: Update to support shader deferred compilation. World probe is tagged to refresh when it it's shader is updated. Probes are recomputed only after all meshes shaders have been compiled.
2018-03-06 02:19:28 +01:00
static float error_col[3] = {1.0f, 0.0f, 1.0f};
static float compile_col[3] = {0.5f, 0.5f, 0.5f};
Rename probe to light-probe Probe is a real general term, the new name is used often in docs online.
2017-06-12 20:59:54 +10:00
struct GPUMaterial *gpumat = EEVEE_material_world_lightprobe_get(scene, wo);
Eevee: Material code refactor. Separate material handling inside another file. Make use of enums to identify shader variations. Group all 64*64 LUTs into one array texture. Only update world probe if world changes.
2017-06-04 12:12:58 +02:00
Eevee: Update to support shader deferred compilation. World probe is tagged to refresh when it it's shader is updated. Probes are recomputed only after all meshes shaders have been compiled.
2018-03-06 02:19:28 +01:00
GPUMaterialStatus status = GPU_material_status(gpumat);
switch (status) {
case GPU_MAT_SUCCESS:
grp = DRW_shgroup_material_create(gpumat, psl->probe_background);
DRW_shgroup_uniform_float(grp, "backgroundAlpha", &stl->g_data->background_alpha, 1);
DRW_shgroup_call_add(grp, geom, NULL);
wo_sh_compiled = true;
break;
case GPU_MAT_QUEUED:
pinfo->all_materials_updated = false;
wo_sh_compiled = false;
/* TODO Bypass probe compilation. */
col = compile_col;
break;
case GPU_MAT_FAILED:
default:
wo_sh_compiled = true;
col = error_col;
break;
Eevee: Material code refactor. Separate material handling inside another file. Make use of enums to identify shader variations. Group all 64*64 LUTs into one array texture. Only update world probe if world changes.
2017-06-04 12:12:58 +02:00
}
}
Eevee: Update lightprobes when deleting or linking a new world datablock
2018-01-17 00:23:31 +01:00
Eevee: Update to support shader deferred compilation. World probe is tagged to refresh when it it's shader is updated. Probes are recomputed only after all meshes shaders have been compiled.
2018-03-06 02:19:28 +01:00
if (wo->update_flag != 0 || pinfo->prev_world != wo || pinfo->prev_wo_sh_compiled != wo_sh_compiled) {
pinfo->update_world |= PROBE_UPDATE_ALL;
EEvEE: LookDev
2018-05-25 08:06:36 +02:00
pinfo->studiolight_index = 0;
Eevee: Update to support shader deferred compilation. World probe is tagged to refresh when it it's shader is updated. Probes are recomputed only after all meshes shaders have been compiled.
2018-03-06 02:19:28 +01:00
pinfo->prev_wo_sh_compiled = wo_sh_compiled;
pinfo->prev_world = wo;
}
wo->update_flag = 0;
Eevee: Update lightprobes when deleting or linking a new world datablock
2018-01-17 00:23:31 +01:00
}
else if (pinfo->prev_world) {
Eevee: Fix probes not working after a world update.
2018-03-06 02:36:00 +01:00
pinfo->update_world |= PROBE_UPDATE_ALL;
EEvEE: LookDev
2018-05-25 08:06:36 +02:00
pinfo->studiolight_index = 0;
Eevee: Update to support shader deferred compilation. World probe is tagged to refresh when it it's shader is updated. Probes are recomputed only after all meshes shaders have been compiled.
2018-03-06 02:19:28 +01:00
pinfo->prev_wo_sh_compiled = false;
Eevee: Update lightprobes when deleting or linking a new world datablock
2018-01-17 00:23:31 +01:00
pinfo->prev_world = NULL;
Eevee: Material code refactor. Separate material handling inside another file. Make use of enums to identify shader variations. Group all 64*64 LUTs into one array texture. Only update world probe if world changes.
2017-06-04 12:12:58 +02:00
}
/* Fallback if shader fails or if not using nodetree. */
if (grp == NULL) {
Eevee: Fix world probe with world without nodetree.
2017-06-29 17:32:13 +02:00
grp = DRW_shgroup_create(e_data.probe_default_sh, psl->probe_background);
Eevee: Material code refactor. Separate material handling inside another file. Make use of enums to identify shader variations. Group all 64*64 LUTs into one array texture. Only update world probe if world changes.
2017-06-04 12:12:58 +02:00
DRW_shgroup_uniform_vec3(grp, "color", col, 1);
Eevee: Fix Planar Reflection bug / background artifact. This was cause by a missing uniform.
2017-07-06 18:28:25 +02:00
DRW_shgroup_uniform_float(grp, "backgroundAlpha", &stl->g_data->background_alpha, 1);
Eevee: Remove Geometry shader usage for background. This fix the behaviour of the light path node that separates the probes background from the viewport background.
2017-06-29 17:07:41 +02:00
DRW_shgroup_call_add(grp, geom, NULL);
Eevee: Material code refactor. Separate material handling inside another file. Make use of enums to identify shader variations. Group all 64*64 LUTs into one array texture. Only update world probe if world changes.
2017-06-04 12:12:58 +02:00
}
}
Eevee: Replace Cubemaps by octahedron maps for env. probes. This enables us to use 2D texture arrays for multiple probes. There is a little artifact with very high roughness caused elongated pixel due to the projection (along every 90° meridian).
2017-05-29 22:03:57 +02:00
{
Eevee: Add Irradiance Grid support Early implementation. Slow and still has quality 3 ways of storing irradiance: - Spherical Harmonics: Have problem with directionnal lighting. - HL2 diffuse cube: Very low resolution but smooth transitions. - Diffuse cube: High storage requirement. Also include some name change.
2017-06-13 17:39:39 +02:00
psl->probe_glossy_compute = DRW_pass_create("LightProbe Glossy Compute", DRW_STATE_WRITE_COLOR);
Eevee: Replace Cubemaps by octahedron maps for env. probes. This enables us to use 2D texture arrays for multiple probes. There is a little artifact with very high roughness caused elongated pixel due to the projection (along every 90° meridian).
2017-05-29 22:03:57 +02:00
DRW: Refactor: Less feature duplication with Gwn. This removes the need of custom attribs for instancing. Instancing works fully with dynamic batches & Gwn_VertFormat now. This is in prevision of the VAO manager patch.
2018-02-14 18:59:15 +01:00
DRWShadingGroup *grp = DRW_shgroup_create(e_data.probe_filter_glossy_sh, psl->probe_glossy_compute);
Eevee: Probes: Add support for intensity tweak. This works for grid and cubemaps. The intensity is baked into the map itself. Thus you need to refresh/rebake the probe to see the changes.
2018-01-21 23:15:57 +01:00
DRW_shgroup_uniform_float(grp, "intensityFac", &pinfo->intensity_fac, 1);
DRW_shgroup_uniform_float(grp, "sampleCount", &pinfo->samples_ct, 1);
DRW_shgroup_uniform_float(grp, "invSampleCount", &pinfo->invsamples_ct, 1);
DRW_shgroup_uniform_float(grp, "roughnessSquared", &pinfo->roughness, 1);
DRW_shgroup_uniform_float(grp, "lodFactor", &pinfo->lodfactor, 1);
DRW_shgroup_uniform_float(grp, "lodMax", &pinfo->lod_rt_max, 1);
DRW_shgroup_uniform_float(grp, "texelSize", &pinfo->texel_size, 1);
DRW_shgroup_uniform_float(grp, "paddingSize", &pinfo->padding_size, 1);
DRW_shgroup_uniform_int(grp, "Layer", &pinfo->layer, 1);
Eevee: Replace Cubemaps by octahedron maps for env. probes. This enables us to use 2D texture arrays for multiple probes. There is a little artifact with very high roughness caused elongated pixel due to the projection (along every 90° meridian).
2017-05-29 22:03:57 +02:00
DRW_shgroup_uniform_texture(grp, "texHammersley", e_data.hammersley);
// DRW_shgroup_uniform_texture(grp, "texJitter", e_data.jitter);
Eevee: Put shadows and probes inside SceneLayerEngineData This remove the duplication of data for each viewport improving memory usage.
2017-05-30 22:29:20 +02:00
DRW_shgroup_uniform_texture(grp, "probeHdr", sldata->probe_rt);
DRW: Refactor: Less feature duplication with Gwn. This removes the need of custom attribs for instancing. Instancing works fully with dynamic batches & Gwn_VertFormat now. This is in prevision of the VAO manager patch.
2018-02-14 18:59:15 +01:00
DRW_shgroup_call_add(grp, DRW_cache_fullscreen_quad_get(), NULL);
Eevee: Replace Cubemaps by octahedron maps for env. probes. This enables us to use 2D texture arrays for multiple probes. There is a little artifact with very high roughness caused elongated pixel due to the projection (along every 90° meridian).
2017-05-29 22:03:57 +02:00
}
{
Eevee: Add Irradiance Grid support Early implementation. Slow and still has quality 3 ways of storing irradiance: - Spherical Harmonics: Have problem with directionnal lighting. - HL2 diffuse cube: Very low resolution but smooth transitions. - Diffuse cube: High storage requirement. Also include some name change.
2017-06-13 17:39:39 +02:00
psl->probe_diffuse_compute = DRW_pass_create("LightProbe Diffuse Compute", DRW_STATE_WRITE_COLOR);
Eevee: Replace Cubemaps by octahedron maps for env. probes. This enables us to use 2D texture arrays for multiple probes. There is a little artifact with very high roughness caused elongated pixel due to the projection (along every 90° meridian).
2017-05-29 22:03:57 +02:00
Eevee: Add Irradiance Grid support Early implementation. Slow and still has quality 3 ways of storing irradiance: - Spherical Harmonics: Have problem with directionnal lighting. - HL2 diffuse cube: Very low resolution but smooth transitions. - Diffuse cube: High storage requirement. Also include some name change.
2017-06-13 17:39:39 +02:00
DRWShadingGroup *grp = DRW_shgroup_create(e_data.probe_filter_diffuse_sh, psl->probe_diffuse_compute);
#ifdef IRRADIANCE_SH_L2
Eevee: Probes: Add support for intensity tweak. This works for grid and cubemaps. The intensity is baked into the map itself. Thus you need to refresh/rebake the probe to see the changes.
2018-01-21 23:15:57 +01:00
DRW_shgroup_uniform_int(grp, "probeSize", &pinfo->shres, 1);
Eevee: Add Irradiance Grid support Early implementation. Slow and still has quality 3 ways of storing irradiance: - Spherical Harmonics: Have problem with directionnal lighting. - HL2 diffuse cube: Very low resolution but smooth transitions. - Diffuse cube: High storage requirement. Also include some name change.
2017-06-13 17:39:39 +02:00
#else
Eevee: Probes: Add support for intensity tweak. This works for grid and cubemaps. The intensity is baked into the map itself. Thus you need to refresh/rebake the probe to see the changes.
2018-01-21 23:15:57 +01:00
DRW_shgroup_uniform_float(grp, "sampleCount", &pinfo->samples_ct, 1);
DRW_shgroup_uniform_float(grp, "invSampleCount", &pinfo->invsamples_ct, 1);
DRW_shgroup_uniform_float(grp, "lodFactor", &pinfo->lodfactor, 1);
DRW_shgroup_uniform_float(grp, "lodMax", &pinfo->lod_rt_max, 1);
Eevee: Add Grid debug display.
2017-06-14 13:45:54 +02:00
DRW_shgroup_uniform_texture(grp, "texHammersley", e_data.hammersley);
Eevee: Add Irradiance Grid support Early implementation. Slow and still has quality 3 ways of storing irradiance: - Spherical Harmonics: Have problem with directionnal lighting. - HL2 diffuse cube: Very low resolution but smooth transitions. - Diffuse cube: High storage requirement. Also include some name change.
2017-06-13 17:39:39 +02:00
#endif
Eevee: Probes: Add support for intensity tweak. This works for grid and cubemaps. The intensity is baked into the map itself. Thus you need to refresh/rebake the probe to see the changes.
2018-01-21 23:15:57 +01:00
DRW_shgroup_uniform_float(grp, "intensityFac", &pinfo->intensity_fac, 1);
Eevee: Put shadows and probes inside SceneLayerEngineData This remove the duplication of data for each viewport improving memory usage.
2017-05-30 22:29:20 +02:00
DRW_shgroup_uniform_texture(grp, "probeHdr", sldata->probe_rt);
Eevee: Replace Cubemaps by octahedron maps for env. probes. This enables us to use 2D texture arrays for multiple probes. There is a little artifact with very high roughness caused elongated pixel due to the projection (along every 90° meridian).
2017-05-29 22:03:57 +02:00
Gawain API naming refactor Use consistent prefix for gawain API names as well as some abbreviations to avoid over-long names, see: D2678
2017-06-19 20:18:04 +10:00
struct Gwn_Batch *geom = DRW_cache_fullscreen_quad_get();
Eevee: Replace Cubemaps by octahedron maps for env. probes. This enables us to use 2D texture arrays for multiple probes. There is a little artifact with very high roughness caused elongated pixel due to the projection (along every 90° meridian).
2017-05-29 22:03:57 +02:00
DRW_shgroup_call_add(grp, geom, NULL);
}
Eevee: Add Grid debug display.
2017-06-14 13:45:54 +02:00
Eevee: Irradiance Visibility: Initial Implementation This augment the existing irradiance grid with a new visibility precomputation. We store a small shadowmap for each grid sample so that light does not leak through walls and such. The visibility parameter are similar to the one used by the Variance Shadow Map for point lights. Technical details: We store the visibility in the same texture (array) as the irradiance itself (in order to reduce the number of sampler). But the irradiance and the visibility are not the same data so we must encode them in order to use the same texture format. We use RGBA8 normalized texture and encode irradiance as RGBE (shared exponent). Using RGBE encoding instead of R11_G11_B10 may lead to some lighting changes, but quality seems to be nearly the same in my test cases. Using full RGBA16/32F maybe a future option but that will require much more memory and reduce the perf significantly. Visibility moments (VSM) are encoded as 16bits fixed point precision using a special range. This seems to retain enough precision for the needs. Also interpolation does not seems to be big problem (even though it's incorrect).
2017-12-02 14:28:29 +01:00
{
psl->probe_visibility_compute = DRW_pass_create("LightProbe Visibility Compute", DRW_STATE_WRITE_COLOR);
DRWShadingGroup *grp = DRW_shgroup_create(e_data.probe_filter_visibility_sh, psl->probe_visibility_compute);
Eevee: Probes: Add support for intensity tweak. This works for grid and cubemaps. The intensity is baked into the map itself. Thus you need to refresh/rebake the probe to see the changes.
2018-01-21 23:15:57 +01:00
DRW_shgroup_uniform_int(grp, "outputSize", &pinfo->shres, 1);
DRW_shgroup_uniform_float(grp, "visibilityRange", &pinfo->visibility_range, 1);
DRW_shgroup_uniform_float(grp, "visibilityBlur", &pinfo->visibility_blur, 1);
DRW_shgroup_uniform_float(grp, "sampleCount", &pinfo->samples_ct, 1);
DRW_shgroup_uniform_float(grp, "invSampleCount", &pinfo->invsamples_ct, 1);
DRW_shgroup_uniform_float(grp, "storedTexelSize", &pinfo->texel_size, 1);
DRW_shgroup_uniform_float(grp, "nearClip", &pinfo->near_clip, 1);
DRW_shgroup_uniform_float(grp, "farClip", &pinfo->far_clip, 1);
Eevee: Irradiance Visibility: Initial Implementation This augment the existing irradiance grid with a new visibility precomputation. We store a small shadowmap for each grid sample so that light does not leak through walls and such. The visibility parameter are similar to the one used by the Variance Shadow Map for point lights. Technical details: We store the visibility in the same texture (array) as the irradiance itself (in order to reduce the number of sampler). But the irradiance and the visibility are not the same data so we must encode them in order to use the same texture format. We use RGBA8 normalized texture and encode irradiance as RGBE (shared exponent). Using RGBE encoding instead of R11_G11_B10 may lead to some lighting changes, but quality seems to be nearly the same in my test cases. Using full RGBA16/32F maybe a future option but that will require much more memory and reduce the perf significantly. Visibility moments (VSM) are encoded as 16bits fixed point precision using a special range. This seems to retain enough precision for the needs. Also interpolation does not seems to be big problem (even though it's incorrect).
2017-12-02 14:28:29 +01:00
DRW_shgroup_uniform_texture(grp, "texHammersley", e_data.hammersley);
DRW_shgroup_uniform_texture(grp, "probeDepth", sldata->probe_depth_rt);
struct Gwn_Batch *geom = DRW_cache_fullscreen_quad_get();
DRW_shgroup_call_add(grp, geom, NULL);
}
Eevee: Probe Grid: Clear Grid buffers with world diffuse coefs. This make sure the values displayed by the "show data" sphere are initialized. Also this make the bounce lighting progress more apparent.
2017-09-28 21:17:57 +02:00
{
psl->probe_grid_fill = DRW_pass_create("LightProbe Grid Floodfill", DRW_STATE_WRITE_COLOR);
DRWShadingGroup *grp = DRW_shgroup_create(e_data.probe_grid_fill_sh, psl->probe_grid_fill);
DRW: Rename DRW_shgroup_uniform_buffer into DRW_shgroup_uniform_texture_ref This is in order to not mix it with the incomming buffer textures.
2018-03-25 19:24:19 +02:00
DRW_shgroup_uniform_texture_ref(grp, "irradianceGrid", &sldata->irradiance_pool);
Eevee: Probe Grid: Clear Grid buffers with world diffuse coefs. This make sure the values displayed by the "show data" sphere are initialized. Also this make the bounce lighting progress more apparent.
2017-09-28 21:17:57 +02:00
struct Gwn_Batch *geom = DRW_cache_fullscreen_quad_get();
DRW_shgroup_call_add(grp, geom, NULL);
}
Eevee: Add Grid debug display.
2017-06-14 13:45:54 +02:00
{
DRW: Rename some DRW_STATE_* for more consistency.
2018-05-20 19:05:13 +02:00
DRWState state = DRW_STATE_WRITE_COLOR | DRW_STATE_WRITE_DEPTH | DRW_STATE_DEPTH_LESS_EQUAL | DRW_STATE_CULL_BACK;
Eevee: Add data display for planar reflection. Maybe not very useful but it's here for feature parity with the other probes. The fragment shader clipping is here because I'm lazy and don't want to creating a proper quad Batch.
2017-06-17 01:59:01 +02:00
psl->probe_display = DRW_pass_create("LightProbe Display", state);
Eevee: Probes: Add data display for cubemaps.
2017-06-15 00:10:34 +02:00
DRW: Refactor: Less feature duplication with Gwn. This removes the need of custom attribs for instancing. Instancing works fully with dynamic batches & Gwn_VertFormat now. This is in prevision of the VAO manager patch.
2018-02-14 18:59:15 +01:00
DRW_shgroup_instance_format(e_data.format_probe_display_cube, {
Cleanup: style, warning
2018-02-15 18:03:55 +11:00
{"probe_id", DRW_ATTRIB_INT, 1},
DRW: Refactor: Less feature duplication with Gwn. This removes the need of custom attribs for instancing. Instancing works fully with dynamic batches & Gwn_VertFormat now. This is in prevision of the VAO manager patch.
2018-02-14 18:59:15 +01:00
{"probe_location", DRW_ATTRIB_FLOAT, 3},
Cleanup: style, warning
2018-02-15 18:03:55 +11:00
{"sphere_size", DRW_ATTRIB_FLOAT, 1},
DRW: Refactor: Less feature duplication with Gwn. This removes the need of custom attribs for instancing. Instancing works fully with dynamic batches & Gwn_VertFormat now. This is in prevision of the VAO manager patch.
2018-02-14 18:59:15 +01:00
});
Cleanup: style, warning
2018-02-15 18:03:55 +11:00
DRWShadingGroup *grp = DRW_shgroup_instance_create(
e_data.probe_cube_display_sh,
psl->probe_display,
DRW_cache_sphere_get(),
e_data.format_probe_display_cube);
DRW: Refactor: Less feature duplication with Gwn. This removes the need of custom attribs for instancing. Instancing works fully with dynamic batches & Gwn_VertFormat now. This is in prevision of the VAO manager patch.
2018-02-14 18:59:15 +01:00
stl->g_data->cube_display_shgrp = grp;
DRW: Rename DRW_shgroup_uniform_buffer into DRW_shgroup_uniform_texture_ref This is in order to not mix it with the incomming buffer textures.
2018-03-25 19:24:19 +02:00
DRW_shgroup_uniform_texture_ref(grp, "probeCubes", &sldata->probe_pool);
Eevee: Put all constant uniforms in a global UBO. This is an optimization / cleanup commit. The use of a global ubo remove lots of uniform lookups and only transfert data when needed. Lots of renaming for more consistent codestyle.
2018-01-21 17:25:10 +01:00
DRW_shgroup_uniform_block(grp, "common_block", sldata->common_ubo);
Object Mode : Add probes data outlines and selectability This required some small changes to the data display shaders so that they match the way the object mode renders them. Strangely enough, I had to remove the normal attribute from the display code because it was being not bound as soon as I created another rendering call in object mode. The problem may be deeper but I did not have time for this so I derive the normal from the sphere pos.
2017-09-30 19:34:23 +02:00
DRW: Refactor: Less feature duplication with Gwn. This removes the need of custom attribs for instancing. Instancing works fully with dynamic batches & Gwn_VertFormat now. This is in prevision of the VAO manager patch.
2018-02-14 18:59:15 +01:00
DRW_shgroup_instance_format(e_data.format_probe_display_planar, {
Cleanup: style, warning
2018-02-15 18:03:55 +11:00
{"probe_id", DRW_ATTRIB_INT, 1},
DRW: Refactor: Less feature duplication with Gwn. This removes the need of custom attribs for instancing. Instancing works fully with dynamic batches & Gwn_VertFormat now. This is in prevision of the VAO manager patch.
2018-02-14 18:59:15 +01:00
{"probe_mat", DRW_ATTRIB_FLOAT, 16},
});
Cleanup: style, warning
2018-02-15 18:03:55 +11:00
grp = DRW_shgroup_instance_create(
e_data.probe_planar_display_sh,
psl->probe_display,
DRW_cache_quad_get(),
e_data.format_probe_display_planar);
DRW: Refactor: Less feature duplication with Gwn. This removes the need of custom attribs for instancing. Instancing works fully with dynamic batches & Gwn_VertFormat now. This is in prevision of the VAO manager patch.
2018-02-14 18:59:15 +01:00
stl->g_data->planar_display_shgrp = grp;
DRW: Rename DRW_shgroup_uniform_buffer into DRW_shgroup_uniform_texture_ref This is in order to not mix it with the incomming buffer textures.
2018-03-25 19:24:19 +02:00
DRW_shgroup_uniform_texture_ref(grp, "probePlanars", &txl->planar_pool);
Eevee: Add Grid debug display.
2017-06-14 13:45:54 +02:00
}
Eevee: Add Planar reflection blurring. This method is very cheap and inaccurate. This will fill the gap untill better model is supported.
2017-06-26 20:37:41 +02:00
{
psl->probe_planar_downsample_ps = DRW_pass_create("LightProbe Planar Downsample", DRW_STATE_WRITE_COLOR);
DRW: Refactor simple instancing. Instead of creating a new instancing shading group without attrib, we now have instancing calls. The benefits is that they can be culled. They can be used in conjuction with the standard and generate calls but shader must support it (which is generally not the case). We store a pointer to the actual count so that the number can be tweaked between redraw. This will makes multi layer rendering more efficient.
2018-03-01 19:27:38 +01:00
DRWShadingGroup *grp = DRW_shgroup_create(e_data.probe_planar_downsample_sh, psl->probe_planar_downsample_ps);
DRW: Rename DRW_shgroup_uniform_buffer into DRW_shgroup_uniform_texture_ref This is in order to not mix it with the incomming buffer textures.
2018-03-25 19:24:19 +02:00
DRW_shgroup_uniform_texture_ref(grp, "source", &txl->planar_pool);
Eevee: Put all constant uniforms in a global UBO. This is an optimization / cleanup commit. The use of a global ubo remove lots of uniform lookups and only transfert data when needed. Lots of renaming for more consistent codestyle.
2018-01-21 17:25:10 +01:00
DRW_shgroup_uniform_float(grp, "fireflyFactor", &sldata->common_data.ssr_firefly_fac, 1);
Cleanup: use 'uint' in draw manager
2018-05-11 07:48:52 +02:00
DRW_shgroup_call_instances_add(grp, DRW_cache_fullscreen_quad_get(), NULL, (uint *)&pinfo->num_planar);
Eevee: Add Planar reflection blurring. This method is very cheap and inaccurate. This will fill the gap untill better model is supported.
2017-06-26 20:37:41 +02:00
}
Eevee: Introduction of world preconvolved envmap. For now only compute GGX convolution. The GGX LUT used for the split sum approximation (UE4) is merged with the LTX mag LUT that uses the same parameters (theta and roughness)
2017-04-18 12:50:09 +02:00
}
EEVEE: LightProbe: Implement Visibility Group. Works on every probe type. The function to see is EEVEE_lightprobes_obj_visibility_cb. Set pinfo->vis_data.cached to true makes the computation faster for multiple views using the same group. We could even sort the probes by group for that mater to speed things up even more (only applies to dynamic probes like the planar reflections because other probes are only rendered one at a time).
2018-04-24 12:45:59 +02:00
bool EEVEE_lightprobes_obj_visibility_cb(bool vis_in, void *user_data)
{
EEVEE_ObjectEngineData *oed = (EEVEE_ObjectEngineData *)user_data;
/* test disabled if group is NULL */
Collections and groups unification OVERVIEW * In 2.7 terminology, all layers and groups are now collection datablocks. * These collections are nestable, linkable, instanceable, overrideable, .. which opens up new ways to set up scenes and link + override data. * Viewport/render visibility and selectability are now a part of the collection and shared across all view layers and linkable. * View layers define which subset of the scene collection hierarchy is excluded for each. For many workflows one view layer can be used, these are more of an advanced feature now. OUTLINER * The outliner now has a "View Layer" display mode instead of "Collections", which can display the collections and/or objects in the view layer. * In this display mode, collections can be excluded with the right click menu. These will then be greyed out and their objects will be excluded. * To view collections not linked to any scene, the "Blender File" display mode can be used, with the new filtering option to just see Colleciton datablocks. * The outliner right click menus for collections and objects were reorganized. * Drag and drop still needs to be improved. Like before, dragging the icon or text gives different results, we'll unify this later. LINKING AND OVERRIDES * Collections can now be linked into the scene without creating an instance, with the link/append operator or from the collections view in the outliner. * Collections can get static overrides with the right click menu in the outliner, but this is rather unreliable and not clearly communicated at the moment. * We still need to improve the make override operator to turn collection instances into collections with overrides directly in the scene. PERFORMANCE * We tried to make performance not worse than before and improve it in some cases. The main thing that's still a bit slower is multiple scenes, we have to change the layer syncing to only updated affected scenes. * Collections keep a list of their parent collections for faster incremental updates in syncing and caching. * View layer bases are now in a object -> base hash to avoid quadratic time lookups internally and in API functions like visible_get(). VERSIONING * Compatibility with 2.7 files should be improved due to the new visibility controls. Of course users may not want to set up their scenes differently now to avoid having separate layers and groups. * Compatibility with 2.8 is mostly there, and was tested on Eevee demo and Hero files. There's a few things which are know to be not quite compatible, like nested layer collections inside groups. * The versioning code for 2.8 files is quite complicated, and isolated behind #ifdef so it can be removed at the end of the release cycle. KNOWN ISSUES * The G-key group operators in the 3D viewport were left mostly as is, they need to be modified still to fit better. * Same for the groups panel in the object properties. This needs to be updated still, or perhaps replaced by something better. * Collections must all have a unique name. Less restrictive namespacing is to be done later, we'll have to see how important this is as all objects within the collections must also have a unique name anyway. * Full scene copy and delete scene are exactly doing the right thing yet. Differential Revision: https://developer.blender.org/D3383 https://code.blender.org/2018/05/collections-and-groups/
2018-04-30 15:57:22 +02:00
if (oed->test_data->collection == NULL)
EEVEE: LightProbe: Implement Visibility Group. Works on every probe type. The function to see is EEVEE_lightprobes_obj_visibility_cb. Set pinfo->vis_data.cached to true makes the computation faster for multiple views using the same group. We could even sort the probes by group for that mater to speed things up even more (only applies to dynamic probes like the planar reflections because other probes are only rendered one at a time).
2018-04-24 12:45:59 +02:00
return vis_in;
if (oed->test_data->cached == false)
oed->ob_vis_dirty = true;
/* early out, don't need to compute ob_vis yet. */
if (vis_in == false)
return vis_in;
if (oed->ob_vis_dirty) {
oed->ob_vis_dirty = false;
Collections and groups unification OVERVIEW * In 2.7 terminology, all layers and groups are now collection datablocks. * These collections are nestable, linkable, instanceable, overrideable, .. which opens up new ways to set up scenes and link + override data. * Viewport/render visibility and selectability are now a part of the collection and shared across all view layers and linkable. * View layers define which subset of the scene collection hierarchy is excluded for each. For many workflows one view layer can be used, these are more of an advanced feature now. OUTLINER * The outliner now has a "View Layer" display mode instead of "Collections", which can display the collections and/or objects in the view layer. * In this display mode, collections can be excluded with the right click menu. These will then be greyed out and their objects will be excluded. * To view collections not linked to any scene, the "Blender File" display mode can be used, with the new filtering option to just see Colleciton datablocks. * The outliner right click menus for collections and objects were reorganized. * Drag and drop still needs to be improved. Like before, dragging the icon or text gives different results, we'll unify this later. LINKING AND OVERRIDES * Collections can now be linked into the scene without creating an instance, with the link/append operator or from the collections view in the outliner. * Collections can get static overrides with the right click menu in the outliner, but this is rather unreliable and not clearly communicated at the moment. * We still need to improve the make override operator to turn collection instances into collections with overrides directly in the scene. PERFORMANCE * We tried to make performance not worse than before and improve it in some cases. The main thing that's still a bit slower is multiple scenes, we have to change the layer syncing to only updated affected scenes. * Collections keep a list of their parent collections for faster incremental updates in syncing and caching. * View layer bases are now in a object -> base hash to avoid quadratic time lookups internally and in API functions like visible_get(). VERSIONING * Compatibility with 2.7 files should be improved due to the new visibility controls. Of course users may not want to set up their scenes differently now to avoid having separate layers and groups. * Compatibility with 2.8 is mostly there, and was tested on Eevee demo and Hero files. There's a few things which are know to be not quite compatible, like nested layer collections inside groups. * The versioning code for 2.8 files is quite complicated, and isolated behind #ifdef so it can be removed at the end of the release cycle. KNOWN ISSUES * The G-key group operators in the 3D viewport were left mostly as is, they need to be modified still to fit better. * Same for the groups panel in the object properties. This needs to be updated still, or perhaps replaced by something better. * Collections must all have a unique name. Less restrictive namespacing is to be done later, we'll have to see how important this is as all objects within the collections must also have a unique name anyway. * Full scene copy and delete scene are exactly doing the right thing yet. Differential Revision: https://developer.blender.org/D3383 https://code.blender.org/2018/05/collections-and-groups/
2018-04-30 15:57:22 +02:00
oed->ob_vis = BKE_collection_has_object_recursive(oed->test_data->collection, oed->ob);
EEVEE: LightProbe: Implement Visibility Group. Works on every probe type. The function to see is EEVEE_lightprobes_obj_visibility_cb. Set pinfo->vis_data.cached to true makes the computation faster for multiple views using the same group. We could even sort the probes by group for that mater to speed things up even more (only applies to dynamic probes like the planar reflections because other probes are only rendered one at a time).
2018-04-24 12:45:59 +02:00
oed->ob_vis = (oed->test_data->invert) ? !oed->ob_vis : oed->ob_vis;
}
return vis_in && oed->ob_vis;
}
Rename any instance of scene layer or render layer in code with view layer The RenderResult struct still has a listbase of RenderLayer, but that's ok since this is strictly for rendering. * Subversion bump (to 2.80.2) * DNA low level doversion (renames) - only for .blend created since 2.80 started Note: We can't use DNA_struct_elem_find or get file version in init_structDNA, so we are manually iterating over the array of the SDNA elements instead. Note 2: This doversion change with renames can be reverted in a few months. But so far it's required for 2.8 files created between October 2016 and now. Reviewers: campbellbarton, sergey Differential Revision: https://developer.blender.org/D2927
2017-11-22 10:52:39 -02:00
void EEVEE_lightprobes_cache_add(EEVEE_ViewLayerData *sldata, Object *ob)
Eevee: Introduction of world preconvolved envmap. For now only compute GGX convolution. The GGX LUT used for the split sum approximation (UE4) is merged with the LTX mag LUT that uses the same parameters (theta and roughness)
2017-04-18 12:50:09 +02:00
{
Rename probe to light-probe Probe is a real general term, the new name is used often in docs online.
2017-06-12 20:59:54 +10:00
EEVEE_LightProbesInfo *pinfo = sldata->probes;
Eevee: Add Irradiance Grid support Early implementation. Slow and still has quality 3 ways of storing irradiance: - Spherical Harmonics: Have problem with directionnal lighting. - HL2 diffuse cube: Very low resolution but smooth transitions. - Diffuse cube: High storage requirement. Also include some name change.
2017-06-13 17:39:39 +02:00
LightProbe *probe = (LightProbe *)ob->data;
Eevee: First commit of Probe support.
2017-06-08 21:48:50 +02:00
Eevee: Add Irradiance Grid support Early implementation. Slow and still has quality 3 ways of storing irradiance: - Spherical Harmonics: Have problem with directionnal lighting. - HL2 diffuse cube: Very low resolution but smooth transitions. - Diffuse cube: High storage requirement. Also include some name change.
2017-06-13 17:39:39 +02:00
if ((probe->type == LIGHTPROBE_TYPE_CUBE && pinfo->num_cube >= MAX_PROBE) ||
Eevee: Probes: Add culling for planar probes. Planar probes that have no influence on pixels in the view are culled. This greatly improve performance when theses probes are offscreen.
2018-03-08 17:56:30 +01:00
(probe->type == LIGHTPROBE_TYPE_GRID && pinfo->num_grid >= MAX_PROBE) ||
(probe->type == LIGHTPROBE_TYPE_PLANAR && pinfo->num_grid >= MAX_PLANAR))
Eevee: Add Irradiance Grid support Early implementation. Slow and still has quality 3 ways of storing irradiance: - Spherical Harmonics: Have problem with directionnal lighting. - HL2 diffuse cube: Very low resolution but smooth transitions. - Diffuse cube: High storage requirement. Also include some name change.
2017-06-13 17:39:39 +02:00
{
Eevee: First commit of Probe support.
2017-06-08 21:48:50 +02:00
printf("Too much probes in the scene !!!\n");
Eevee: Add Irradiance Grid support Early implementation. Slow and still has quality 3 ways of storing irradiance: - Spherical Harmonics: Have problem with directionnal lighting. - HL2 diffuse cube: Very low resolution but smooth transitions. - Diffuse cube: High storage requirement. Also include some name change.
2017-06-13 17:39:39 +02:00
return;
Eevee: First commit of Probe support.
2017-06-08 21:48:50 +02:00
}
Eevee: Probes: Add culling for planar probes. Planar probes that have no influence on pixels in the view are culled. This greatly improve performance when theses probes are offscreen.
2018-03-08 17:56:30 +01:00
if (probe->type == LIGHTPROBE_TYPE_PLANAR) {
/* See if this planar probe is inside the view frustum. If not, no need to update it. */
/* NOTE: this could be bypassed if we want feedback loop mirrors for rendering. */
BoundBox bbox; float tmp[4][4];
const float min[3] = {-1.0f, -1.0f, -1.0f};
const float max[3] = { 1.0f, 1.0f, 1.0f};
BKE_boundbox_init_from_minmax(&bbox, min, max);
copy_m4_m4(tmp, ob->obmat);
normalize_v3(tmp[2]);
mul_v3_fl(tmp[2], probe->distinf);
for (int v = 0; v < 8; ++v) {
mul_m4_v3(tmp, bbox.vec[v]);
}
if (!DRW_culling_box_test(&bbox)) {
return; /* Culled */
}
}
Eevee: Cleanup, naming For functions which will allocate requested data if it does not exist yet "_ensure" is to be used instead of "_get". "_get" functions should return NULL in cases when requested data does not exist yet.
2017-11-29 10:36:58 +01:00
EEVEE_LightProbeEngineData *ped = EEVEE_lightprobe_data_ensure(ob);
Eevee: First commit of Probe support.
2017-06-08 21:48:50 +02:00
Eevee: Irradiance grid: support for non-blocking update and multiple bounces.
2017-06-14 18:52:53 +02:00
ped->num_cell = probe->grid_resolution_x * probe->grid_resolution_y * probe->grid_resolution_z;
Eevee: Irradiance Grid: Allocate needed resources instead of a static chunck. This commit makes the irradiance pool and render target sizes depend on the number of irradiance sample in the whole ViewLayer.
2017-12-04 17:20:20 +01:00
if ((probe->type == LIGHTPROBE_TYPE_GRID) &&
Cleanup: style, warning
2018-02-15 18:03:55 +11:00
((pinfo->total_irradiance_samples + ped->num_cell) >= MAX_IRRADIANCE_SAMPLES))
Eevee: Irradiance Grid: Allocate needed resources instead of a static chunck. This commit makes the irradiance pool and render target sizes depend on the number of irradiance sample in the whole ViewLayer.
2017-12-04 17:20:20 +01:00
{
printf("Too much grid samples !!!\n");
return;
}
Draw manager: Listen to depsgraph's ID update callbacks This replaces dedicated flag which wasn't clean who sets it and who clears it, and which was also trying to re-implement existing functionality in a way. Flushing is not currently very efficient but there are ways to speed this up a lot, but needs more investigation.
2017-11-28 17:05:52 +01:00
if (ped->need_full_update) {
ped->need_full_update = false;
Eevee: Add Irradiance Grid support Early implementation. Slow and still has quality 3 ways of storing irradiance: - Spherical Harmonics: Have problem with directionnal lighting. - HL2 diffuse cube: Very low resolution but smooth transitions. - Diffuse cube: High storage requirement. Also include some name change.
2017-06-13 17:39:39 +02:00
ped->need_update = true;
Eevee: Probes: Add data display for cubemaps.
2017-06-15 00:10:34 +02:00
ped->probe_id = 0;
Eevee: Fix grid probe updating when updating cube probe.
2017-06-15 00:23:00 +02:00
if (probe->type == LIGHTPROBE_TYPE_GRID) {
ped->updated_cells = 0;
Eevee: Probe Grid: Add "progressive rendering" This gives the user a more rapid preview of what the final result will be. The grid cells are renderer in a series of powers of two offsets.
2017-09-28 21:16:02 +02:00
ped->updated_lvl = 0;
Eevee: Fix grid probe updating when updating cube probe.
2017-06-15 00:23:00 +02:00
pinfo->updated_bounce = 0;
Eevee: Probe Grid: Clear Grid buffers with world diffuse coefs. This make sure the values displayed by the "show data" sphere are initialized. Also this make the bounce lighting progress more apparent.
2017-09-28 21:17:57 +02:00
pinfo->grid_initialized = false;
Eevee: Fix grid probe updating when updating cube probe.
2017-06-15 00:23:00 +02:00
}
Eevee: Add Irradiance Grid support Early implementation. Slow and still has quality 3 ways of storing irradiance: - Spherical Harmonics: Have problem with directionnal lighting. - HL2 diffuse cube: Very low resolution but smooth transitions. - Diffuse cube: High storage requirement. Also include some name change.
2017-06-13 17:39:39 +02:00
}
Eevee: First commit of Probe support.
2017-06-08 21:48:50 +02:00
Eevee: Update to support shader deferred compilation. World probe is tagged to refresh when it it's shader is updated. Probes are recomputed only after all meshes shaders have been compiled.
2018-03-06 02:19:28 +01:00
if (pinfo->update_world) {
Eevee: Add Irradiance Grid support Early implementation. Slow and still has quality 3 ways of storing irradiance: - Spherical Harmonics: Have problem with directionnal lighting. - HL2 diffuse cube: Very low resolution but smooth transitions. - Diffuse cube: High storage requirement. Also include some name change.
2017-06-13 17:39:39 +02:00
ped->need_update = true;
Eevee: Irradiance grid: support for non-blocking update and multiple bounces.
2017-06-14 18:52:53 +02:00
ped->updated_cells = 0;
Eevee: Probe Grid: Add "progressive rendering" This gives the user a more rapid preview of what the final result will be. The grid cells are renderer in a series of powers of two offsets.
2017-09-28 21:16:02 +02:00
ped->updated_lvl = 0;
Eevee: Probes: Add data display for cubemaps.
2017-06-15 00:10:34 +02:00
ped->probe_id = 0;
Eevee: Add Irradiance Grid support Early implementation. Slow and still has quality 3 ways of storing irradiance: - Spherical Harmonics: Have problem with directionnal lighting. - HL2 diffuse cube: Very low resolution but smooth transitions. - Diffuse cube: High storage requirement. Also include some name change.
2017-06-13 17:39:39 +02:00
}
Eevee: Put all constant uniforms in a global UBO. This is an optimization / cleanup commit. The use of a global ubo remove lots of uniform lookups and only transfert data when needed. Lots of renaming for more consistent codestyle.
2018-01-21 17:25:10 +01:00
pinfo->do_cube_update |= ped->need_update;
Eevee: Probes: Add culling for planar probes. Planar probes that have no influence on pixels in the view are culled. This greatly improve performance when theses probes are offscreen.
2018-03-08 17:56:30 +01:00
switch (probe->type) {
case LIGHTPROBE_TYPE_CUBE:
pinfo->probes_cube_ref[pinfo->num_cube] = ob;
pinfo->num_cube++;
break;
case LIGHTPROBE_TYPE_PLANAR:
pinfo->probes_planar_ref[pinfo->num_planar] = ob;
pinfo->num_planar++;
break;
case LIGHTPROBE_TYPE_GRID:
pinfo->probes_grid_ref[pinfo->num_grid] = ob;
pinfo->num_grid++;
pinfo->total_irradiance_samples += ped->num_cell;
break;
Eevee: Add Irradiance Grid support Early implementation. Slow and still has quality 3 ways of storing irradiance: - Spherical Harmonics: Have problem with directionnal lighting. - HL2 diffuse cube: Very low resolution but smooth transitions. - Diffuse cube: High storage requirement. Also include some name change.
2017-06-13 17:39:39 +02:00
}
Eevee: First commit of Probe support.
2017-06-08 21:48:50 +02:00
}
Eevee: Initial implementation of planar reflections. Still pretty barebone: No roughness support, No normal distortion support.
2017-06-17 00:08:03 +02:00
/* TODO find a nice name to push it to math_matrix.c */
static void scale_m4_v3(float R[4][4], float v[3])
{
for (int i = 0; i < 4; ++i)
mul_v3_v3(R[i], v);
}
Eevee: Planar Probe: Add supersampling jitter. This also fix a bug with the probe debug display when there was more than 2 probes. ped->probe_id was equal to 0 for all planar probes until the next frame. Resulting in all planar data debug to show probe 0.
2018-03-08 01:35:16 +01:00
static void EEVEE_planar_reflections_cache_finish(EEVEE_ViewLayerData *sldata, EEVEE_StorageList *stl)
{
EEVEE_LightProbesInfo *pinfo = sldata->probes;
Object *ob;
for (int i = 0; (ob = pinfo->probes_planar_ref[i]) && (i < MAX_PLANAR); i++) {
LightProbe *probe = (LightProbe *)ob->data;
EEVEE_LightProbeEngineData *ped = EEVEE_lightprobe_data_ensure(ob);
ped->probe_id = i;
/* Debug Display */
if (DRW_state_draw_support() &&
(probe->flag & LIGHTPROBE_FLAG_SHOW_DATA))
{
DRW_shgroup_call_dynamic_add(stl->g_data->planar_display_shgrp, &ped->probe_id, ob->obmat);
}
}
}
static void EEVEE_planar_reflections_updates(EEVEE_ViewLayerData *sldata)
Eevee: Initial implementation of planar reflections. Still pretty barebone: No roughness support, No normal distortion support.
2017-06-17 00:08:03 +02:00
{
EEVEE_LightProbesInfo *pinfo = sldata->probes;
Object *ob;
float mtx[4][4], normat[4][4], imat[4][4], rangemat[4][4];
Eevee: Planar Probe: Add supersampling jitter. This also fix a bug with the probe debug display when there was more than 2 probes. ped->probe_id was equal to 0 for all planar probes until the next frame. Resulting in all planar data debug to show probe 0.
2018-03-08 01:35:16 +01:00
float viewmat[4][4];
Eevee: Initial implementation of planar reflections. Still pretty barebone: No roughness support, No normal distortion support.
2017-06-17 00:08:03 +02:00
DRW_viewport_matrix_get(viewmat, DRW_MAT_VIEW);
zero_m4(rangemat);
rangemat[0][0] = rangemat[1][1] = rangemat[2][2] = 0.5f;
rangemat[3][0] = rangemat[3][1] = rangemat[3][2] = 0.5f;
rangemat[3][3] = 1.0f;
/* PLANAR REFLECTION */
for (int i = 0; (ob = pinfo->probes_planar_ref[i]) && (i < MAX_PLANAR); i++) {
LightProbe *probe = (LightProbe *)ob->data;
EEVEE_PlanarReflection *eplanar = &pinfo->planar_data[i];
Eevee: Cleanup, naming For functions which will allocate requested data if it does not exist yet "_ensure" is to be used instead of "_get". "_get" functions should return NULL in cases when requested data does not exist yet.
2017-11-29 10:36:58 +01:00
EEVEE_LightProbeEngineData *ped = EEVEE_lightprobe_data_ensure(ob);
Eevee: Initial implementation of planar reflections. Still pretty barebone: No roughness support, No normal distortion support.
2017-06-17 00:08:03 +02:00
/* Computing mtx : matrix that mirror position around object's XY plane. */
normalize_m4_m4(normat, ob->obmat); /* object > world */
invert_m4_m4(imat, normat); /* world > object */
float reflect[3] = {1.0f, 1.0f, -1.0f}; /* XY reflection plane */
scale_m4_v3(imat, reflect); /* world > object > mirrored obj */
mul_m4_m4m4(mtx, normat, imat); /* world > object > mirrored obj > world */
/* Reflect Camera Matrix. */
Eevee: Save and reset matrixstate for probe rendering.
2018-03-07 23:25:49 +01:00
mul_m4_m4m4(ped->mats.mat[DRW_MAT_VIEW], viewmat, mtx);
Eevee: Initial implementation of planar reflections. Still pretty barebone: No roughness support, No normal distortion support.
2017-06-17 00:08:03 +02:00
/* TODO FOV margin */
Eevee: Planar Probe: Add supersampling jitter. This also fix a bug with the probe debug display when there was more than 2 probes. ped->probe_id was equal to 0 for all planar probes until the next frame. Resulting in all planar data debug to show probe 0.
2018-03-08 01:35:16 +01:00
/* Temporal sampling jitter should be already applied to the DRW_MAT_WIN. */
DRW_viewport_matrix_get(ped->mats.mat[DRW_MAT_WIN], DRW_MAT_WIN);
/* Apply Projection Matrix. */
Eevee: Save and reset matrixstate for probe rendering.
2018-03-07 23:25:49 +01:00
mul_m4_m4m4(ped->mats.mat[DRW_MAT_PERS], ped->mats.mat[DRW_MAT_WIN], ped->mats.mat[DRW_MAT_VIEW]);
Eevee: Initial implementation of planar reflections. Still pretty barebone: No roughness support, No normal distortion support.
2017-06-17 00:08:03 +02:00
/* This is the matrix used to reconstruct texture coordinates.
* We use the original view matrix because it does not create
* visual artifacts if receiver is not perfectly aligned with
* the planar reflection probe. */
Eevee: Save and reset matrixstate for probe rendering.
2018-03-07 23:25:49 +01:00
mul_m4_m4m4(eplanar->reflectionmat, ped->mats.mat[DRW_MAT_WIN], viewmat); /* TODO FOV margin */
Eevee: Initial implementation of planar reflections. Still pretty barebone: No roughness support, No normal distortion support.
2017-06-17 00:08:03 +02:00
/* Convert from [-1, 1] to [0, 1] (NDC to Texture coord). */
mul_m4_m4m4(eplanar->reflectionmat, rangemat, eplanar->reflectionmat);
/* Compute clip plane equation / normal. */
float refpoint[3];
copy_v3_v3(eplanar->plane_equation, ob->obmat[2]);
normalize_v3(eplanar->plane_equation); /* plane normal */
Eevee: Fix Planar Probe offset changing influence calculation.
2017-06-17 01:56:45 +02:00
eplanar->plane_equation[3] = -dot_v3v3(eplanar->plane_equation, ob->obmat[3]);
/* Compute offset plane equation (fix missing texels near reflection plane). */
copy_v3_v3(ped->planer_eq_offset, eplanar->plane_equation);
Eevee: Initial implementation of planar reflections. Still pretty barebone: No roughness support, No normal distortion support.
2017-06-17 00:08:03 +02:00
mul_v3_v3fl(refpoint, eplanar->plane_equation, -probe->clipsta);
add_v3_v3(refpoint, ob->obmat[3]);
Eevee: Fix Planar Probe offset changing influence calculation.
2017-06-17 01:56:45 +02:00
ped->planer_eq_offset[3] = -dot_v3v3(eplanar->plane_equation, refpoint);
Eevee: Initial implementation of planar reflections. Still pretty barebone: No roughness support, No normal distortion support.
2017-06-17 00:08:03 +02:00
/* Compute XY clip planes. */
normalize_v3_v3(eplanar->clip_vec_x, ob->obmat[0]);
normalize_v3_v3(eplanar->clip_vec_y, ob->obmat[1]);
float vec[3] = {0.0f, 0.0f, 0.0f};
vec[0] = 1.0f; vec[1] = 0.0f; vec[2] = 0.0f;
mul_m4_v3(ob->obmat, vec); /* Point on the edge */
eplanar->clip_edge_x_pos = dot_v3v3(eplanar->clip_vec_x, vec);
vec[0] = 0.0f; vec[1] = 1.0f; vec[2] = 0.0f;
mul_m4_v3(ob->obmat, vec); /* Point on the edge */
eplanar->clip_edge_y_pos = dot_v3v3(eplanar->clip_vec_y, vec);
vec[0] = -1.0f; vec[1] = 0.0f; vec[2] = 0.0f;
mul_m4_v3(ob->obmat, vec); /* Point on the edge */
eplanar->clip_edge_x_neg = dot_v3v3(eplanar->clip_vec_x, vec);
vec[0] = 0.0f; vec[1] = -1.0f; vec[2] = 0.0f;
mul_m4_v3(ob->obmat, vec); /* Point on the edge */
eplanar->clip_edge_y_neg = dot_v3v3(eplanar->clip_vec_y, vec);
/* Facing factors */
float max_angle = max_ff(1e-2f, probe->falloff) * M_PI * 0.5f;
float min_angle = 0.0f;
eplanar->facing_scale = 1.0f / max_ff(1e-8f, cosf(min_angle) - cosf(max_angle));
eplanar->facing_bias = -min_ff(1.0f - 1e-8f, cosf(max_angle)) * eplanar->facing_scale;
/* Distance factors */
float max_dist = probe->distinf;
float min_dist = min_ff(1.0f - 1e-8f, 1.0f - probe->falloff) * probe->distinf;
eplanar->attenuation_scale = -1.0f / max_ff(1e-8f, max_dist - min_dist);
eplanar->attenuation_bias = max_dist * -eplanar->attenuation_scale;
}
}
Rename any instance of scene layer or render layer in code with view layer The RenderResult struct still has a listbase of RenderLayer, but that's ok since this is strictly for rendering. * Subversion bump (to 2.80.2) * DNA low level doversion (renames) - only for .blend created since 2.80 started Note: We can't use DNA_struct_elem_find or get file version in init_structDNA, so we are manually iterating over the array of the SDNA elements instead. Note 2: This doversion change with renames can be reverted in a few months. But so far it's required for 2.8 files created between October 2016 and now. Reviewers: campbellbarton, sergey Differential Revision: https://developer.blender.org/D2927
2017-11-22 10:52:39 -02:00
static void EEVEE_lightprobes_updates(EEVEE_ViewLayerData *sldata, EEVEE_PassList *psl, EEVEE_StorageList *stl)
Eevee: First commit of Probe support.
2017-06-08 21:48:50 +02:00
{
Rename probe to light-probe Probe is a real general term, the new name is used often in docs online.
2017-06-12 20:59:54 +10:00
EEVEE_LightProbesInfo *pinfo = sldata->probes;
Eevee: First commit of Probe support.
2017-06-08 21:48:50 +02:00
Object *ob;
Eevee: Initial implementation of planar reflections. Still pretty barebone: No roughness support, No normal distortion support.
2017-06-17 00:08:03 +02:00
/* CUBE REFLECTION */
Eevee: Add Irradiance Grid support Early implementation. Slow and still has quality 3 ways of storing irradiance: - Spherical Harmonics: Have problem with directionnal lighting. - HL2 diffuse cube: Very low resolution but smooth transitions. - Diffuse cube: High storage requirement. Also include some name change.
2017-06-13 17:39:39 +02:00
for (int i = 1; (ob = pinfo->probes_cube_ref[i]) && (i < MAX_PROBE); i++) {
Rename probe to light-probe Probe is a real general term, the new name is used often in docs online.
2017-06-12 20:59:54 +10:00
LightProbe *probe = (LightProbe *)ob->data;
EEVEE_LightProbe *eprobe = &pinfo->probe_data[i];
Eevee: Cleanup, naming For functions which will allocate requested data if it does not exist yet "_ensure" is to be used instead of "_get". "_get" functions should return NULL in cases when requested data does not exist yet.
2017-11-29 10:36:58 +01:00
EEVEE_LightProbeEngineData *ped = EEVEE_lightprobe_data_ensure(ob);
Eevee: First commit of Probe support.
2017-06-08 21:48:50 +02:00
Eevee: Add Irradiance Grid support Early implementation. Slow and still has quality 3 ways of storing irradiance: - Spherical Harmonics: Have problem with directionnal lighting. - HL2 diffuse cube: Very low resolution but smooth transitions. - Diffuse cube: High storage requirement. Also include some name change.
2017-06-13 17:39:39 +02:00
/* Update transforms */
copy_v3_v3(eprobe->position, ob->obmat[3]);
Eevee: Add parallax correction to probe reflections
2017-06-09 22:30:49 +02:00
/* Attenuation */
eprobe->attenuation_type = probe->attenuation_type;
eprobe->attenuation_fac = 1.0f / max_ff(1e-8f, probe->falloff);
unit_m4(eprobe->attenuationmat);
Probe: Remove Bounding Box parameter. After using it for like 30 sec, the min max bound box is absolutely not practical. Reverting into using a unit cube with object transform. This also simplify the code. In the future of center probes will be implemented using another object matrix (via an object pointer).
2017-06-09 23:21:23 +02:00
scale_m4_fl(eprobe->attenuationmat, probe->distinf);
mul_m4_m4m4(eprobe->attenuationmat, ob->obmat, eprobe->attenuationmat);
Eevee: Add parallax correction to probe reflections
2017-06-09 22:30:49 +02:00
invert_m4(eprobe->attenuationmat);
/* Parallax */
float dist;
Rename probe to light-probe Probe is a real general term, the new name is used often in docs online.
2017-06-12 20:59:54 +10:00
if ((probe->flag & LIGHTPROBE_FLAG_CUSTOM_PARALLAX) != 0) {
Eevee: Add parallax correction to probe reflections
2017-06-09 22:30:49 +02:00
eprobe->parallax_type = probe->parallax_type;
dist = probe->distpar;
}
else {
eprobe->parallax_type = probe->attenuation_type;
dist = probe->distinf;
}
unit_m4(eprobe->parallaxmat);
Probe: Remove Bounding Box parameter. After using it for like 30 sec, the min max bound box is absolutely not practical. Reverting into using a unit cube with object transform. This also simplify the code. In the future of center probes will be implemented using another object matrix (via an object pointer).
2017-06-09 23:21:23 +02:00
scale_m4_fl(eprobe->parallaxmat, dist);
mul_m4_m4m4(eprobe->parallaxmat, ob->obmat, eprobe->parallaxmat);
Eevee: Add parallax correction to probe reflections
2017-06-09 22:30:49 +02:00
invert_m4(eprobe->parallaxmat);
Eevee: Probes: Add data display for cubemaps.
2017-06-15 00:10:34 +02:00
/* Debug Display */
Group collection viewport/render options and remove collection visibility Users can change the group collection visibility in the outliner when looking at groups. Regular collections on the other hand don't have any special visibility control, if you need a collection to be invisible during render, either don't link it into the view layer used for F12, or disable it. This includes: * Updated unittests - update your lib/tests/layers folder. * Subversion bump - branches be aware of that. Note: Although we are using eval_ctx to determine the visibility of a group collection when rendering, the depsgraph is still using the same depsgraph for the viewport and the render engine, so at the moment the render visibility is ignored. Following next is a workaround for this separately to tag the groups before and after rendering to tackle that.
2017-12-14 11:46:49 -02:00
if (DRW_state_draw_support() &&
Depsgraph and collection enable/visibility Iterate over invisible objects too, so lamps can still lit the scene. Also, now you can use a collection to set an object to invisible, not only to visible. For example: Scene > Master collection > bedroom > furniture Scene > View Layer > bedroom (visible) > furniture (invisible) The View Layer has two linked collections, bedroom and furniture. This setup will make the furniture collection invisible. Note: Unlike what was suggested on D2849, this does not make collection visibility influence camera visibility. I will keep this as a separate patch. Reviewers: sergey Subscribers: sergey, brecht, fclem Differential Revision: https://developer.blender.org/D2849
2017-09-21 12:55:14 +02:00
(probe->flag & LIGHTPROBE_FLAG_SHOW_DATA))
{
Object Mode : Add probes data outlines and selectability This required some small changes to the data display shaders so that they match the way the object mode renders them. Strangely enough, I had to remove the normal attribute from the display code because it was being not bound as soon as I created another rendering call in object mode. The problem may be deeper but I did not have time for this so I derive the normal from the sphere pos.
2017-09-30 19:34:23 +02:00
ped->probe_size = probe->data_draw_size * 0.1f;
Depsgraph and collection enable/visibility Iterate over invisible objects too, so lamps can still lit the scene. Also, now you can use a collection to set an object to invisible, not only to visible. For example: Scene > Master collection > bedroom > furniture Scene > View Layer > bedroom (visible) > furniture (invisible) The View Layer has two linked collections, bedroom and furniture. This setup will make the furniture collection invisible. Note: Unlike what was suggested on D2849, this does not make collection visibility influence camera visibility. I will keep this as a separate patch. Reviewers: sergey Subscribers: sergey, brecht, fclem Differential Revision: https://developer.blender.org/D2849
2017-09-21 12:55:14 +02:00
DRW_shgroup_call_dynamic_add(
Object Mode : Add probes data outlines and selectability This required some small changes to the data display shaders so that they match the way the object mode renders them. Strangely enough, I had to remove the normal attribute from the display code because it was being not bound as soon as I created another rendering call in object mode. The problem may be deeper but I did not have time for this so I derive the normal from the sphere pos.
2017-09-30 19:34:23 +02:00
stl->g_data->cube_display_shgrp, &ped->probe_id, ob->obmat[3], &ped->probe_size);
Eevee: Probes: Add data display for cubemaps.
2017-06-15 00:10:34 +02:00
}
Eevee: First commit of Probe support.
2017-06-08 21:48:50 +02:00
}
Eevee: Add Irradiance Grid support Early implementation. Slow and still has quality 3 ways of storing irradiance: - Spherical Harmonics: Have problem with directionnal lighting. - HL2 diffuse cube: Very low resolution but smooth transitions. - Diffuse cube: High storage requirement. Also include some name change.
2017-06-13 17:39:39 +02:00
Eevee: Initial implementation of planar reflections. Still pretty barebone: No roughness support, No normal distortion support.
2017-06-17 00:08:03 +02:00
/* IRRADIANCE GRID */
Eevee: Add Irradiance Grid support Early implementation. Slow and still has quality 3 ways of storing irradiance: - Spherical Harmonics: Have problem with directionnal lighting. - HL2 diffuse cube: Very low resolution but smooth transitions. - Diffuse cube: High storage requirement. Also include some name change.
2017-06-13 17:39:39 +02:00
int offset = 1; /* to account for the world probe */
Eevee: Irradiance grid: support for non-blocking update and multiple bounces.
2017-06-14 18:52:53 +02:00
for (int i = 1; (ob = pinfo->probes_grid_ref[i]) && (i < MAX_GRID); i++) {
Eevee: Add Irradiance Grid support Early implementation. Slow and still has quality 3 ways of storing irradiance: - Spherical Harmonics: Have problem with directionnal lighting. - HL2 diffuse cube: Very low resolution but smooth transitions. - Diffuse cube: High storage requirement. Also include some name change.
2017-06-13 17:39:39 +02:00
LightProbe *probe = (LightProbe *)ob->data;
EEVEE_LightGrid *egrid = &pinfo->grid_data[i];
Eevee: Cleanup, naming For functions which will allocate requested data if it does not exist yet "_ensure" is to be used instead of "_get". "_get" functions should return NULL in cases when requested data does not exist yet.
2017-11-29 10:36:58 +01:00
EEVEE_LightProbeEngineData *ped = EEVEE_lightprobe_data_ensure(ob);
Eevee: Add Irradiance Grid support Early implementation. Slow and still has quality 3 ways of storing irradiance: - Spherical Harmonics: Have problem with directionnal lighting. - HL2 diffuse cube: Very low resolution but smooth transitions. - Diffuse cube: High storage requirement. Also include some name change.
2017-06-13 17:39:39 +02:00
Eevee: Irradiance Grid: Fix problem with bounce update. Before this patch, if one of the grid was updated (moved) only the subsequents evaluated grids had their level reset and had all their bounces recomputed.
2017-12-02 00:59:58 +01:00
/* If one grid has move we need to recompute all the lighting. */
if (!pinfo->grid_initialized) {
ped->updated_cells = 0;
ped->updated_lvl = 0;
ped->need_update = true;
}
Eevee: Probe Grid: Add "progressive rendering" This gives the user a more rapid preview of what the final result will be. The grid cells are renderer in a series of powers of two offsets.
2017-09-28 21:16:02 +02:00
/* Add one for level 0 */
ped->max_lvl = 1.0f + floorf(log2f((float)MAX3(probe->grid_resolution_x,
probe->grid_resolution_y,
probe->grid_resolution_z)));
Eevee: Add Irradiance Grid support Early implementation. Slow and still has quality 3 ways of storing irradiance: - Spherical Harmonics: Have problem with directionnal lighting. - HL2 diffuse cube: Very low resolution but smooth transitions. - Diffuse cube: High storage requirement. Also include some name change.
2017-06-13 17:39:39 +02:00
egrid->offset = offset;
Eevee: Probes: Change falloff of irradiance grid.
2017-06-14 22:45:20 +02:00
float fac = 1.0f / max_ff(1e-8f, probe->falloff);
egrid->attenuation_scale = fac / max_ff(1e-8f, probe->distinf);
egrid->attenuation_bias = fac;
Eevee: Add Irradiance Grid support Early implementation. Slow and still has quality 3 ways of storing irradiance: - Spherical Harmonics: Have problem with directionnal lighting. - HL2 diffuse cube: Very low resolution but smooth transitions. - Diffuse cube: High storage requirement. Also include some name change.
2017-06-13 17:39:39 +02:00
/* Set offset for the next grid */
Eevee: Irradiance grid: support for non-blocking update and multiple bounces.
2017-06-14 18:52:53 +02:00
offset += ped->num_cell;
Eevee: Add Irradiance Grid support Early implementation. Slow and still has quality 3 ways of storing irradiance: - Spherical Harmonics: Have problem with directionnal lighting. - HL2 diffuse cube: Very low resolution but smooth transitions. - Diffuse cube: High storage requirement. Also include some name change.
2017-06-13 17:39:39 +02:00
/* Update transforms */
Eevee: Irradiance grid: support for non-blocking update and multiple bounces.
2017-06-14 18:52:53 +02:00
float cell_dim[3], half_cell_dim[3];
cell_dim[0] = 2.0f / (float)(probe->grid_resolution_x);
cell_dim[1] = 2.0f / (float)(probe->grid_resolution_y);
cell_dim[2] = 2.0f / (float)(probe->grid_resolution_z);
mul_v3_v3fl(half_cell_dim, cell_dim, 0.5f);
Eevee: Add Irradiance Grid support Early implementation. Slow and still has quality 3 ways of storing irradiance: - Spherical Harmonics: Have problem with directionnal lighting. - HL2 diffuse cube: Very low resolution but smooth transitions. - Diffuse cube: High storage requirement. Also include some name change.
2017-06-13 17:39:39 +02:00
Eevee: Probes: Change falloff of irradiance grid.
2017-06-14 22:45:20 +02:00
/* Matrix converting world space to cell ranges. */
invert_m4_m4(egrid->mat, ob->obmat);
Eevee: Add Irradiance Grid support Early implementation. Slow and still has quality 3 ways of storing irradiance: - Spherical Harmonics: Have problem with directionnal lighting. - HL2 diffuse cube: Very low resolution but smooth transitions. - Diffuse cube: High storage requirement. Also include some name change.
2017-06-13 17:39:39 +02:00
/* First cell. */
Eevee: Irradiance grid: support for non-blocking update and multiple bounces.
2017-06-14 18:52:53 +02:00
copy_v3_fl(egrid->corner, -1.0f);
add_v3_v3(egrid->corner, half_cell_dim);
Eevee: Add Irradiance Grid support Early implementation. Slow and still has quality 3 ways of storing irradiance: - Spherical Harmonics: Have problem with directionnal lighting. - HL2 diffuse cube: Very low resolution but smooth transitions. - Diffuse cube: High storage requirement. Also include some name change.
2017-06-13 17:39:39 +02:00
mul_m4_v3(ob->obmat, egrid->corner);
/* Opposite neighbor cell. */
Eevee: Irradiance grid: support for non-blocking update and multiple bounces.
2017-06-14 18:52:53 +02:00
copy_v3_fl3(egrid->increment_x, cell_dim[0], 0.0f, 0.0f);
add_v3_v3(egrid->increment_x, half_cell_dim);
Eevee: Add Irradiance Grid support Early implementation. Slow and still has quality 3 ways of storing irradiance: - Spherical Harmonics: Have problem with directionnal lighting. - HL2 diffuse cube: Very low resolution but smooth transitions. - Diffuse cube: High storage requirement. Also include some name change.
2017-06-13 17:39:39 +02:00
add_v3_fl(egrid->increment_x, -1.0f);
mul_m4_v3(ob->obmat, egrid->increment_x);
sub_v3_v3(egrid->increment_x, egrid->corner);
Eevee: Irradiance grid: support for non-blocking update and multiple bounces.
2017-06-14 18:52:53 +02:00
copy_v3_fl3(egrid->increment_y, 0.0f, cell_dim[1], 0.0f);
add_v3_v3(egrid->increment_y, half_cell_dim);
Eevee: Add Irradiance Grid support Early implementation. Slow and still has quality 3 ways of storing irradiance: - Spherical Harmonics: Have problem with directionnal lighting. - HL2 diffuse cube: Very low resolution but smooth transitions. - Diffuse cube: High storage requirement. Also include some name change.
2017-06-13 17:39:39 +02:00
add_v3_fl(egrid->increment_y, -1.0f);
mul_m4_v3(ob->obmat, egrid->increment_y);
sub_v3_v3(egrid->increment_y, egrid->corner);
Eevee: Irradiance grid: support for non-blocking update and multiple bounces.
2017-06-14 18:52:53 +02:00
copy_v3_fl3(egrid->increment_z, 0.0f, 0.0f, cell_dim[2]);
add_v3_v3(egrid->increment_z, half_cell_dim);
Eevee: Add Irradiance Grid support Early implementation. Slow and still has quality 3 ways of storing irradiance: - Spherical Harmonics: Have problem with directionnal lighting. - HL2 diffuse cube: Very low resolution but smooth transitions. - Diffuse cube: High storage requirement. Also include some name change.
2017-06-13 17:39:39 +02:00
add_v3_fl(egrid->increment_z, -1.0f);
mul_m4_v3(ob->obmat, egrid->increment_z);
sub_v3_v3(egrid->increment_z, egrid->corner);
copy_v3_v3_int(egrid->resolution, &probe->grid_resolution_x);
Eevee: Add Grid debug display.
2017-06-14 13:45:54 +02:00
Eevee: Irradiance Visibility: Initial Implementation This augment the existing irradiance grid with a new visibility precomputation. We store a small shadowmap for each grid sample so that light does not leak through walls and such. The visibility parameter are similar to the one used by the Variance Shadow Map for point lights. Technical details: We store the visibility in the same texture (array) as the irradiance itself (in order to reduce the number of sampler). But the irradiance and the visibility are not the same data so we must encode them in order to use the same texture format. We use RGBA8 normalized texture and encode irradiance as RGBE (shared exponent). Using RGBE encoding instead of R11_G11_B10 may lead to some lighting changes, but quality seems to be nearly the same in my test cases. Using full RGBA16/32F maybe a future option but that will require much more memory and reduce the perf significantly. Visibility moments (VSM) are encoded as 16bits fixed point precision using a special range. This seems to retain enough precision for the needs. Also interpolation does not seems to be big problem (even though it's incorrect).
2017-12-02 14:28:29 +01:00
/* Visibility bias */
egrid->visibility_bias = 0.05f * probe->vis_bias;
egrid->visibility_bleed = probe->vis_bleedbias;
Avoid unnecessary sqrt calls
2018-02-15 21:38:24 +11:00
egrid->visibility_range = (
sqrtf(max_fff(len_squared_v3(egrid->increment_x),
len_squared_v3(egrid->increment_y),
len_squared_v3(egrid->increment_z))) + 1.0f);
Eevee: Irradiance Visibility: Initial Implementation This augment the existing irradiance grid with a new visibility precomputation. We store a small shadowmap for each grid sample so that light does not leak through walls and such. The visibility parameter are similar to the one used by the Variance Shadow Map for point lights. Technical details: We store the visibility in the same texture (array) as the irradiance itself (in order to reduce the number of sampler). But the irradiance and the visibility are not the same data so we must encode them in order to use the same texture format. We use RGBA8 normalized texture and encode irradiance as RGBE (shared exponent). Using RGBE encoding instead of R11_G11_B10 may lead to some lighting changes, but quality seems to be nearly the same in my test cases. Using full RGBA16/32F maybe a future option but that will require much more memory and reduce the perf significantly. Visibility moments (VSM) are encoded as 16bits fixed point precision using a special range. This seems to retain enough precision for the needs. Also interpolation does not seems to be big problem (even though it's incorrect).
2017-12-02 14:28:29 +01:00
Eevee: Add Grid debug display.
2017-06-14 13:45:54 +02:00
/* Debug Display */
Group collection viewport/render options and remove collection visibility Users can change the group collection visibility in the outliner when looking at groups. Regular collections on the other hand don't have any special visibility control, if you need a collection to be invisible during render, either don't link it into the view layer used for F12, or disable it. This includes: * Updated unittests - update your lib/tests/layers folder. * Subversion bump - branches be aware of that. Note: Although we are using eval_ctx to determine the visibility of a group collection when rendering, the depsgraph is still using the same depsgraph for the viewport and the render engine, so at the moment the render visibility is ignored. Following next is a workaround for this separately to tag the groups before and after rendering to tackle that.
2017-12-14 11:46:49 -02:00
if (DRW_state_draw_support() &&
Depsgraph and collection enable/visibility Iterate over invisible objects too, so lamps can still lit the scene. Also, now you can use a collection to set an object to invisible, not only to visible. For example: Scene > Master collection > bedroom > furniture Scene > View Layer > bedroom (visible) > furniture (invisible) The View Layer has two linked collections, bedroom and furniture. This setup will make the furniture collection invisible. Note: Unlike what was suggested on D2849, this does not make collection visibility influence camera visibility. I will keep this as a separate patch. Reviewers: sergey Subscribers: sergey, brecht, fclem Differential Revision: https://developer.blender.org/D2849
2017-09-21 12:55:14 +02:00
(probe->flag & LIGHTPROBE_FLAG_SHOW_DATA))
{
DRW: Refactor simple instancing. Instead of creating a new instancing shading group without attrib, we now have instancing calls. The benefits is that they can be culled. They can be used in conjuction with the standard and generate calls but shader must support it (which is generally not the case). We store a pointer to the actual count so that the number can be tweaked between redraw. This will makes multi layer rendering more efficient.
2018-03-01 19:27:38 +01:00
DRWShadingGroup *grp = DRW_shgroup_create(e_data.probe_grid_display_sh, psl->probe_display);
Eevee: Add Grid debug display.
2017-06-14 13:45:54 +02:00
DRW_shgroup_uniform_int(grp, "offset", &egrid->offset, 1);
DRW_shgroup_uniform_ivec3(grp, "grid_resolution", egrid->resolution, 1);
DRW_shgroup_uniform_vec3(grp, "corner", egrid->corner, 1);
DRW_shgroup_uniform_vec3(grp, "increment_x", egrid->increment_x, 1);
DRW_shgroup_uniform_vec3(grp, "increment_y", egrid->increment_y, 1);
DRW_shgroup_uniform_vec3(grp, "increment_z", egrid->increment_z, 1);
DRW: Rename DRW_shgroup_uniform_buffer into DRW_shgroup_uniform_texture_ref This is in order to not mix it with the incomming buffer textures.
2018-03-25 19:24:19 +02:00
DRW_shgroup_uniform_texture_ref(grp, "irradianceGrid", &sldata->irradiance_pool);
Eevee: Probes: Add data display for cubemaps.
2017-06-15 00:10:34 +02:00
DRW_shgroup_uniform_float(grp, "sphere_size", &probe->data_draw_size, 1);
Cleanup: use 'uint' in draw manager
2018-05-11 07:48:52 +02:00
DRW_shgroup_call_instances_add(grp, DRW_cache_sphere_get(), NULL, (uint *)&ped->num_cell);
Eevee: Add Grid debug display.
2017-06-14 13:45:54 +02:00
}
Eevee: Add Irradiance Grid support Early implementation. Slow and still has quality 3 ways of storing irradiance: - Spherical Harmonics: Have problem with directionnal lighting. - HL2 diffuse cube: Very low resolution but smooth transitions. - Diffuse cube: High storage requirement. Also include some name change.
2017-06-13 17:39:39 +02:00
}
Eevee: Introduction of world preconvolved envmap. For now only compute GGX convolution. The GGX LUT used for the split sum approximation (UE4) is merged with the LTX mag LUT that uses the same parameters (theta and roughness)
2017-04-18 12:50:09 +02:00
}
Rename any instance of scene layer or render layer in code with view layer The RenderResult struct still has a listbase of RenderLayer, but that's ok since this is strictly for rendering. * Subversion bump (to 2.80.2) * DNA low level doversion (renames) - only for .blend created since 2.80 started Note: We can't use DNA_struct_elem_find or get file version in init_structDNA, so we are manually iterating over the array of the SDNA elements instead. Note 2: This doversion change with renames can be reverted in a few months. But so far it's required for 2.8 files created between October 2016 and now. Reviewers: campbellbarton, sergey Differential Revision: https://developer.blender.org/D2927
2017-11-22 10:52:39 -02:00
void EEVEE_lightprobes_cache_finish(EEVEE_ViewLayerData *sldata, EEVEE_Data *vedata)
Eevee: Introduction of world preconvolved envmap. For now only compute GGX convolution. The GGX LUT used for the split sum approximation (UE4) is merged with the LTX mag LUT that uses the same parameters (theta and roughness)
2017-04-18 12:50:09 +02:00
{
Eevee: Put all constant uniforms in a global UBO. This is an optimization / cleanup commit. The use of a global ubo remove lots of uniform lookups and only transfert data when needed. Lots of renaming for more consistent codestyle.
2018-01-21 17:25:10 +01:00
EEVEE_CommonUniformBuffer *common_data = &sldata->common_data;
Rename probe to light-probe Probe is a real general term, the new name is used often in docs online.
2017-06-12 20:59:54 +10:00
EEVEE_LightProbesInfo *pinfo = sldata->probes;
Eevee: First commit of Probe support.
2017-06-08 21:48:50 +02:00
Object *ob;
Eevee: Introduction of world preconvolved envmap. For now only compute GGX convolution. The GGX LUT used for the split sum approximation (UE4) is merged with the LTX mag LUT that uses the same parameters (theta and roughness)
2017-04-18 12:50:09 +02:00
Eevee: Material code refactor. Separate material handling inside another file. Make use of enums to identify shader variations. Group all 64*64 LUTs into one array texture. Only update world probe if world changes.
2017-06-04 12:12:58 +02:00
/* Setup enough layers. */
/* Free textures if number mismatch. */
if (pinfo->num_cube != pinfo->cache_num_cube) {
DRW_TEXTURE_FREE_SAFE(sldata->probe_pool);
Eevee: Fix probes not working after a world update.
2018-03-06 02:36:00 +01:00
pinfo->cache_num_cube = pinfo->num_cube;
Eevee: Material code refactor. Separate material handling inside another file. Make use of enums to identify shader variations. Group all 64*64 LUTs into one array texture. Only update world probe if world changes.
2017-06-04 12:12:58 +02:00
}
Eevee: Initial implementation of planar reflections. Still pretty barebone: No roughness support, No normal distortion support.
2017-06-17 00:08:03 +02:00
if (pinfo->num_planar != pinfo->cache_num_planar) {
DRW_TEXTURE_FREE_SAFE(vedata->txl->planar_pool);
Eevee: Planar Reflection: Remove distance approximation. This commit separate the depth texture into another texture array. This remove the need to output radial depth into alpha. Unfortunatly it's difficult to recover position from the non linear depth buffer when applying reflection without adding a bunch of stuff. This is in preparation of SSR planar reflections.
2017-07-23 14:03:27 +02:00
DRW_TEXTURE_FREE_SAFE(vedata->txl->planar_depth);
Eevee: Initial implementation of planar reflections. Still pretty barebone: No roughness support, No normal distortion support.
2017-06-17 00:08:03 +02:00
pinfo->cache_num_planar = pinfo->num_planar;
}
Eevee: Irradiance Grid: Allocate needed resources instead of a static chunck. This commit makes the irradiance pool and render target sizes depend on the number of irradiance sample in the whole ViewLayer.
2017-12-04 17:20:20 +01:00
int irr_size[3];
Eevee: Put all constant uniforms in a global UBO. This is an optimization / cleanup commit. The use of a global ubo remove lots of uniform lookups and only transfert data when needed. Lots of renaming for more consistent codestyle.
2018-01-21 17:25:10 +01:00
irradiance_pool_size_get(common_data->prb_irradiance_vis_size, pinfo->total_irradiance_samples, irr_size);
Eevee: Irradiance Grid: Allocate needed resources instead of a static chunck. This commit makes the irradiance pool and render target sizes depend on the number of irradiance sample in the whole ViewLayer.
2017-12-04 17:20:20 +01:00
if ((irr_size[0] != pinfo->cache_irradiance_size[0]) ||
(irr_size[1] != pinfo->cache_irradiance_size[1]) ||
(irr_size[2] != pinfo->cache_irradiance_size[2]))
{
DRW_TEXTURE_FREE_SAFE(sldata->irradiance_pool);
DRW_TEXTURE_FREE_SAFE(sldata->irradiance_rt);
copy_v3_v3_int(pinfo->cache_irradiance_size, irr_size);
}
Eevee: Initial implementation of planar reflections. Still pretty barebone: No roughness support, No normal distortion support.
2017-06-17 00:08:03 +02:00
/* XXX this should be run each frame as it ensure planar_depth is set */
planar_pool_ensure_alloc(vedata, pinfo->num_planar);
Eevee: Material code refactor. Separate material handling inside another file. Make use of enums to identify shader variations. Group all 64*64 LUTs into one array texture. Only update world probe if world changes.
2017-06-04 12:12:58 +02:00
if (!sldata->probe_pool) {
Eevee: Probe: Add Cubemap Resolution option.
2017-10-10 21:56:11 +02:00
sldata->probe_pool = DRW_texture_create_2D_array(pinfo->cubemap_res, pinfo->cubemap_res, max_ff(1, pinfo->num_cube),
DRW: Remove DRWTextureFormat in favor or GPUTextureFormat. Because: - Less redundancy. - Better suffixes. Also a few modification to GPU_texture_create_* to simplify the API: - make the format explicit to the texture creation process. - remove the component count as it's specified in the GPUTextureFormat.
2018-04-30 16:02:24 +02:00
GPU_R11F_G11F_B10F, DRW_TEX_FILTER | DRW_TEX_MIPMAP, NULL);
Eevee: First commit of Probe support.
2017-06-08 21:48:50 +02:00
if (sldata->probe_filter_fb) {
GPUFramebuffer: Refactor (Part 2) This refactor modernise the use of framebuffers. It also touches a lot of files so breaking down changes we have: - GPUTexture: Allow textures to be attached to more than one GPUFrameBuffer. This allows to create and configure more FBO without the need to attach and detach texture at drawing time. - GPUFrameBuffer: The wrapper starts to mimic opengl a bit closer. This allows to configure the framebuffer inside a context other than the one that will be rendering the framebuffer. We do the actual configuration when binding the FBO. We also Keep track of config validity and save drawbuffers state in the FBO. We remove the different bind/unbind functions. These make little sense now that we have separate contexts. - DRWFrameBuffer: We replace DRW_framebuffer functions by GPU_framebuffer ones to avoid another layer of abstraction. We move the DRW convenience functions to GPUFramebuffer instead and even add new ones. The MACRO GPU_framebuffer_ensure_config is pretty much all you need to create and config a GPUFramebuffer. - DRWTexture: Due to the removal of DRWFrameBuffer, we needed to create functions to create textures for thoses framebuffers. Pool textures are now using default texture parameters for the texture type asked. - DRWManager: Make sure no framebuffer object is bound when doing cache filling. - GPUViewport: Add new color_only_fb and depth_only_fb along with FB API usage update. This let draw engines render to color/depth only target and without the need to attach/detach textures. - WM_window: Assert when a framebuffer is bound when changing context. This balance the fact we are not track ogl context inside GPUFramebuffer. - Eevee, Clay, Mode engines: Update to new API. This comes with a lot of code simplification. This also come with some cleanups in some engine codes.
2018-03-25 17:46:48 +02:00
GPU_framebuffer_texture_attach(sldata->probe_filter_fb, sldata->probe_pool, 0, 0);
Eevee: First commit of Probe support.
2017-06-08 21:48:50 +02:00
}
/* Tag probes to refresh */
Eevee: Update to support shader deferred compilation. World probe is tagged to refresh when it it's shader is updated. Probes are recomputed only after all meshes shaders have been compiled.
2018-03-06 02:19:28 +01:00
pinfo->update_world |= PROBE_UPDATE_CUBE;
Eevee: Fix probes not working after a world update.
2018-03-06 02:36:00 +01:00
}
Eevee: First commit of Probe support.
2017-06-08 21:48:50 +02:00
Eevee: Fix probes not working after a world update.
2018-03-06 02:36:00 +01:00
if ((pinfo->update_world & PROBE_UPDATE_CUBE) != 0) {
common_data->prb_num_render_cube = 0;
Eevee: Add Irradiance Grid support Early implementation. Slow and still has quality 3 ways of storing irradiance: - Spherical Harmonics: Have problem with directionnal lighting. - HL2 diffuse cube: Very low resolution but smooth transitions. - Diffuse cube: High storage requirement. Also include some name change.
2017-06-13 17:39:39 +02:00
for (int i = 1; (ob = pinfo->probes_cube_ref[i]) && (i < MAX_PROBE); i++) {
Eevee: Cleanup, naming For functions which will allocate requested data if it does not exist yet "_ensure" is to be used instead of "_get". "_get" functions should return NULL in cases when requested data does not exist yet.
2017-11-29 10:36:58 +01:00
EEVEE_LightProbeEngineData *ped = EEVEE_lightprobe_data_ensure(ob);
Eevee: First commit of Probe support.
2017-06-08 21:48:50 +02:00
ped->need_update = true;
ped->ready_to_shade = false;
Eevee: Probes: Add data display for cubemaps.
2017-06-15 00:10:34 +02:00
ped->probe_id = 0;
Eevee: First commit of Probe support.
2017-06-08 21:48:50 +02:00
}
Eevee: Material code refactor. Separate material handling inside another file. Make use of enums to identify shader variations. Group all 64*64 LUTs into one array texture. Only update world probe if world changes.
2017-06-04 12:12:58 +02:00
}
Eevee: Irradiance grid: support for non-blocking update and multiple bounces.
2017-06-14 18:52:53 +02:00
#ifdef IRRADIANCE_SH_L2
/* we need a signed format for Spherical Harmonics */
DRW: Remove DRWTextureFormat in favor or GPUTextureFormat. Because: - Less redundancy. - Better suffixes. Also a few modification to GPU_texture_create_* to simplify the API: - make the format explicit to the texture creation process. - remove the component count as it's specified in the GPUTextureFormat.
2018-04-30 16:02:24 +02:00
int irradiance_format = GPU_RGBA16F;
Eevee: Irradiance grid: support for non-blocking update and multiple bounces.
2017-06-14 18:52:53 +02:00
#else
DRW: Remove DRWTextureFormat in favor or GPUTextureFormat. Because: - Less redundancy. - Better suffixes. Also a few modification to GPU_texture_create_* to simplify the API: - make the format explicit to the texture creation process. - remove the component count as it's specified in the GPUTextureFormat.
2018-04-30 16:02:24 +02:00
int irradiance_format = GPU_RGBA8;
Eevee: Irradiance grid: support for non-blocking update and multiple bounces.
2017-06-14 18:52:53 +02:00
#endif
Eevee : Probe Grid : Fix Update tagging.
2017-10-02 18:40:51 +02:00
if (!sldata->irradiance_pool || !sldata->irradiance_rt) {
if (!sldata->irradiance_pool) {
Eevee: Irradiance Grid: Allocate needed resources instead of a static chunck. This commit makes the irradiance pool and render target sizes depend on the number of irradiance sample in the whole ViewLayer.
2017-12-04 17:20:20 +01:00
sldata->irradiance_pool = DRW_texture_create_2D_array(irr_size[0], irr_size[1], irr_size[2],
Eevee: Irradiance Visibility: Initial Implementation This augment the existing irradiance grid with a new visibility precomputation. We store a small shadowmap for each grid sample so that light does not leak through walls and such. The visibility parameter are similar to the one used by the Variance Shadow Map for point lights. Technical details: We store the visibility in the same texture (array) as the irradiance itself (in order to reduce the number of sampler). But the irradiance and the visibility are not the same data so we must encode them in order to use the same texture format. We use RGBA8 normalized texture and encode irradiance as RGBE (shared exponent). Using RGBE encoding instead of R11_G11_B10 may lead to some lighting changes, but quality seems to be nearly the same in my test cases. Using full RGBA16/32F maybe a future option but that will require much more memory and reduce the perf significantly. Visibility moments (VSM) are encoded as 16bits fixed point precision using a special range. This seems to retain enough precision for the needs. Also interpolation does not seems to be big problem (even though it's incorrect).
2017-12-02 14:28:29 +01:00
irradiance_format, DRW_TEX_FILTER, NULL);
Eevee : Probe Grid : Fix Update tagging.
2017-10-02 18:40:51 +02:00
}
if (!sldata->irradiance_rt) {
Eevee: Irradiance Grid: Allocate needed resources instead of a static chunck. This commit makes the irradiance pool and render target sizes depend on the number of irradiance sample in the whole ViewLayer.
2017-12-04 17:20:20 +01:00
sldata->irradiance_rt = DRW_texture_create_2D_array(irr_size[0], irr_size[1], irr_size[2],
Eevee: Irradiance Visibility: Initial Implementation This augment the existing irradiance grid with a new visibility precomputation. We store a small shadowmap for each grid sample so that light does not leak through walls and such. The visibility parameter are similar to the one used by the Variance Shadow Map for point lights. Technical details: We store the visibility in the same texture (array) as the irradiance itself (in order to reduce the number of sampler). But the irradiance and the visibility are not the same data so we must encode them in order to use the same texture format. We use RGBA8 normalized texture and encode irradiance as RGBE (shared exponent). Using RGBE encoding instead of R11_G11_B10 may lead to some lighting changes, but quality seems to be nearly the same in my test cases. Using full RGBA16/32F maybe a future option but that will require much more memory and reduce the perf significantly. Visibility moments (VSM) are encoded as 16bits fixed point precision using a special range. This seems to retain enough precision for the needs. Also interpolation does not seems to be big problem (even though it's incorrect).
2017-12-02 14:28:29 +01:00
irradiance_format, DRW_TEX_FILTER, NULL);
Eevee: Some probe update changes. Dont render grids if cubemap count changes (because it does not depends on). Recalc lighting if probe type change.
2017-06-14 22:46:48 +02:00
}
Eevee: Fix probes not working after a world update.
2018-03-06 02:36:00 +01:00
/* Tag probes to refresh */
Eevee: Update to support shader deferred compilation. World probe is tagged to refresh when it it's shader is updated. Probes are recomputed only after all meshes shaders have been compiled.
2018-03-06 02:19:28 +01:00
pinfo->update_world |= PROBE_UPDATE_GRID;
Eevee: Fix probes not working after a world update.
2018-03-06 02:36:00 +01:00
pinfo->grid_initialized = false;
}
Eevee: Some probe update changes. Dont render grids if cubemap count changes (because it does not depends on). Recalc lighting if probe type change.
2017-06-14 22:46:48 +02:00
Eevee: Fix probes not working after a world update.
2018-03-06 02:36:00 +01:00
if ((pinfo->update_world & PROBE_UPDATE_GRID) != 0) {
common_data->prb_num_render_grid = 0;
pinfo->updated_bounce = 0;
Eevee: Some probe update changes. Dont render grids if cubemap count changes (because it does not depends on). Recalc lighting if probe type change.
2017-06-14 22:46:48 +02:00
for (int i = 1; (ob = pinfo->probes_grid_ref[i]) && (i < MAX_PROBE); i++) {
Eevee: Cleanup, naming For functions which will allocate requested data if it does not exist yet "_ensure" is to be used instead of "_get". "_get" functions should return NULL in cases when requested data does not exist yet.
2017-11-29 10:36:58 +01:00
EEVEE_LightProbeEngineData *ped = EEVEE_lightprobe_data_ensure(ob);
Eevee: Some probe update changes. Dont render grids if cubemap count changes (because it does not depends on). Recalc lighting if probe type change.
2017-06-14 22:46:48 +02:00
ped->need_update = true;
ped->updated_cells = 0;
}
Eevee: Irradiance grid: support for non-blocking update and multiple bounces.
2017-06-14 18:52:53 +02:00
}
Eevee: Put all constant uniforms in a global UBO. This is an optimization / cleanup commit. The use of a global ubo remove lots of uniform lookups and only transfert data when needed. Lots of renaming for more consistent codestyle.
2018-01-21 17:25:10 +01:00
if (common_data->prb_num_render_grid > pinfo->num_grid) {
Eevee: Fix light grid being rendered after delete.
2017-10-10 19:21:32 +02:00
/* This can happen when deleting a probe. */
Eevee: Put all constant uniforms in a global UBO. This is an optimization / cleanup commit. The use of a global ubo remove lots of uniform lookups and only transfert data when needed. Lots of renaming for more consistent codestyle.
2018-01-21 17:25:10 +01:00
common_data->prb_num_render_grid = pinfo->num_grid;
Eevee: Fix light grid being rendered after delete.
2017-10-10 19:21:32 +02:00
}
Eevee: Planar Probe: Add supersampling jitter. This also fix a bug with the probe debug display when there was more than 2 probes. ped->probe_id was equal to 0 for all planar probes until the next frame. Resulting in all planar data debug to show probe 0.
2018-03-08 01:35:16 +01:00
EEVEE_planar_reflections_cache_finish(sldata, vedata->stl);
Eevee: Probes: Add data display for cubemaps.
2017-06-15 00:10:34 +02:00
EEVEE_lightprobes_updates(sldata, vedata->psl, vedata->stl);
Eevee: First commit of Probe support.
2017-06-08 21:48:50 +02:00
Eevee: Move Spherical Harmonics to a new Probe UBO. Keep data packing tight to prevent use of padding floats
2017-06-04 16:50:22 +02:00
DRW_uniformbuffer_update(sldata->probe_ubo, &sldata->probes->probe_data);
Eevee: Add Irradiance Grid support Early implementation. Slow and still has quality 3 ways of storing irradiance: - Spherical Harmonics: Have problem with directionnal lighting. - HL2 diffuse cube: Very low resolution but smooth transitions. - Diffuse cube: High storage requirement. Also include some name change.
2017-06-13 17:39:39 +02:00
DRW_uniformbuffer_update(sldata->grid_ubo, &sldata->probes->grid_data);
Eevee: Introduction of world preconvolved envmap. For now only compute GGX convolution. The GGX LUT used for the split sum approximation (UE4) is merged with the LTX mag LUT that uses the same parameters (theta and roughness)
2017-04-18 12:50:09 +02:00
}
Eevee: Add Planar reflection blurring. This method is very cheap and inaccurate. This will fill the gap untill better model is supported.
2017-06-26 20:37:41 +02:00
static void downsample_planar(void *vedata, int level)
{
EEVEE_PassList *psl = ((EEVEE_Data *)vedata)->psl;
EEVEE_StorageList *stl = ((EEVEE_Data *)vedata)->stl;
const float *size = DRW_viewport_size_get();
Eevee: Put all constant uniforms in a global UBO. This is an optimization / cleanup commit. The use of a global ubo remove lots of uniform lookups and only transfert data when needed. Lots of renaming for more consistent codestyle.
2018-01-21 17:25:10 +01:00
copy_v2_v2(stl->g_data->planar_texel_size, size);
Eevee: Add Planar reflection blurring. This method is very cheap and inaccurate. This will fill the gap untill better model is supported.
2017-06-26 20:37:41 +02:00
for (int i = 0; i < level - 1; ++i) {
Eevee: Put all constant uniforms in a global UBO. This is an optimization / cleanup commit. The use of a global ubo remove lots of uniform lookups and only transfert data when needed. Lots of renaming for more consistent codestyle.
2018-01-21 17:25:10 +01:00
stl->g_data->planar_texel_size[0] /= 2.0f;
stl->g_data->planar_texel_size[1] /= 2.0f;
min_ff(floorf(stl->g_data->planar_texel_size[0]), 1.0f);
min_ff(floorf(stl->g_data->planar_texel_size[1]), 1.0f);
Eevee: Add Planar reflection blurring. This method is very cheap and inaccurate. This will fill the gap untill better model is supported.
2017-06-26 20:37:41 +02:00
}
Eevee: Put all constant uniforms in a global UBO. This is an optimization / cleanup commit. The use of a global ubo remove lots of uniform lookups and only transfert data when needed. Lots of renaming for more consistent codestyle.
2018-01-21 17:25:10 +01:00
invert_v2(stl->g_data->planar_texel_size);
Eevee: Add Planar reflection blurring. This method is very cheap and inaccurate. This will fill the gap untill better model is supported.
2017-06-26 20:37:41 +02:00
DRW_draw_pass(psl->probe_planar_downsample_ps);
}
Eevee: Add Irradiance Grid support Early implementation. Slow and still has quality 3 ways of storing irradiance: - Spherical Harmonics: Have problem with directionnal lighting. - HL2 diffuse cube: Very low resolution but smooth transitions. - Diffuse cube: High storage requirement. Also include some name change.
2017-06-13 17:39:39 +02:00
/* Glossy filter probe_rt to probe_pool at index probe_idx */
Eevee: Probes: Add support for intensity tweak. This works for grid and cubemaps. The intensity is baked into the map itself. Thus you need to refresh/rebake the probe to see the changes.
2018-01-21 23:15:57 +01:00
static void glossy_filter_probe(
EEVEE_ViewLayerData *sldata, EEVEE_Data *vedata, EEVEE_PassList *psl, int probe_idx, float intensity)
Eevee: Introduction of world preconvolved envmap. For now only compute GGX convolution. The GGX LUT used for the split sum approximation (UE4) is merged with the LTX mag LUT that uses the same parameters (theta and roughness)
2017-04-18 12:50:09 +02:00
{
Rename probe to light-probe Probe is a real general term, the new name is used often in docs online.
2017-06-12 20:59:54 +10:00
EEVEE_LightProbesInfo *pinfo = sldata->probes;
Eevee: Material code refactor. Separate material handling inside another file. Make use of enums to identify shader variations. Group all 64*64 LUTs into one array texture. Only update world probe if world changes.
2017-06-04 12:12:58 +02:00
Eevee: Probes: Add support for intensity tweak. This works for grid and cubemaps. The intensity is baked into the map itself. Thus you need to refresh/rebake the probe to see the changes.
2018-01-21 23:15:57 +01:00
pinfo->intensity_fac = intensity;
Eevee: Fix problem with GPU_texture_generate_mipmap This function was called to recreate the lower mip level of the probe texture. But this is not it's usage and it introduced a stall. This patch add cubemap mipmap level regeneration in eevee_effects.c
2017-08-18 22:31:06 +02:00
/* Max lod used from the render target probe */
Eevee: Probe: Add Cubemap Resolution option.
2017-10-10 21:56:11 +02:00
pinfo->lod_rt_max = floorf(log2f(pinfo->target_size)) - 2.0f;
Eevee: Fix problem with GPU_texture_generate_mipmap This function was called to recreate the lower mip level of the probe texture. But this is not it's usage and it introduced a stall. This patch add cubemap mipmap level regeneration in eevee_effects.c
2017-08-18 22:31:06 +02:00
GPUFramebuffer: Refactor (Part 2) This refactor modernise the use of framebuffers. It also touches a lot of files so breaking down changes we have: - GPUTexture: Allow textures to be attached to more than one GPUFrameBuffer. This allows to create and configure more FBO without the need to attach and detach texture at drawing time. - GPUFrameBuffer: The wrapper starts to mimic opengl a bit closer. This allows to configure the framebuffer inside a context other than the one that will be rendering the framebuffer. We do the actual configuration when binding the FBO. We also Keep track of config validity and save drawbuffers state in the FBO. We remove the different bind/unbind functions. These make little sense now that we have separate contexts. - DRWFrameBuffer: We replace DRW_framebuffer functions by GPU_framebuffer ones to avoid another layer of abstraction. We move the DRW convenience functions to GPUFramebuffer instead and even add new ones. The MACRO GPU_framebuffer_ensure_config is pretty much all you need to create and config a GPUFramebuffer. - DRWTexture: Due to the removal of DRWFrameBuffer, we needed to create functions to create textures for thoses framebuffers. Pool textures are now using default texture parameters for the texture type asked. - DRWManager: Make sure no framebuffer object is bound when doing cache filling. - GPUViewport: Add new color_only_fb and depth_only_fb along with FB API usage update. This let draw engines render to color/depth only target and without the need to attach/detach textures. - WM_window: Assert when a framebuffer is bound when changing context. This balance the fact we are not track ogl context inside GPUFramebuffer. - Eevee, Clay, Mode engines: Update to new API. This comes with a lot of code simplification. This also come with some cleanups in some engine codes.
2018-03-25 17:46:48 +02:00
/* Start fresh */
GPU_framebuffer_ensure_config(&sldata->probe_filter_fb, {
GPU_ATTACHMENT_NONE,
GPU_ATTACHMENT_NONE
});
Eevee: Introduction of world preconvolved envmap. For now only compute GGX convolution. The GGX LUT used for the split sum approximation (UE4) is merged with the LTX mag LUT that uses the same parameters (theta and roughness)
2017-04-18 12:50:09 +02:00
/* 2 - Let gpu create Mipmaps for Filtered Importance Sampling. */
Eevee: Replace Cubemaps by octahedron maps for env. probes. This enables us to use 2D texture arrays for multiple probes. There is a little artifact with very high roughness caused elongated pixel due to the projection (along every 90° meridian).
2017-05-29 22:03:57 +02:00
/* Bind next framebuffer to be able to gen. mips for probe_rt. */
GPUFramebuffer: Refactor (Part 2) This refactor modernise the use of framebuffers. It also touches a lot of files so breaking down changes we have: - GPUTexture: Allow textures to be attached to more than one GPUFrameBuffer. This allows to create and configure more FBO without the need to attach and detach texture at drawing time. - GPUFrameBuffer: The wrapper starts to mimic opengl a bit closer. This allows to configure the framebuffer inside a context other than the one that will be rendering the framebuffer. We do the actual configuration when binding the FBO. We also Keep track of config validity and save drawbuffers state in the FBO. We remove the different bind/unbind functions. These make little sense now that we have separate contexts. - DRWFrameBuffer: We replace DRW_framebuffer functions by GPU_framebuffer ones to avoid another layer of abstraction. We move the DRW convenience functions to GPUFramebuffer instead and even add new ones. The MACRO GPU_framebuffer_ensure_config is pretty much all you need to create and config a GPUFramebuffer. - DRWTexture: Due to the removal of DRWFrameBuffer, we needed to create functions to create textures for thoses framebuffers. Pool textures are now using default texture parameters for the texture type asked. - DRWManager: Make sure no framebuffer object is bound when doing cache filling. - GPUViewport: Add new color_only_fb and depth_only_fb along with FB API usage update. This let draw engines render to color/depth only target and without the need to attach/detach textures. - WM_window: Assert when a framebuffer is bound when changing context. This balance the fact we are not track ogl context inside GPUFramebuffer. - Eevee, Clay, Mode engines: Update to new API. This comes with a lot of code simplification. This also come with some cleanups in some engine codes.
2018-03-25 17:46:48 +02:00
EEVEE_downsample_cube_buffer(vedata, sldata->probe_rt, (int)(pinfo->lod_rt_max));
Eevee: Introduction of world preconvolved envmap. For now only compute GGX convolution. The GGX LUT used for the split sum approximation (UE4) is merged with the LTX mag LUT that uses the same parameters (theta and roughness)
2017-04-18 12:50:09 +02:00
/* 3 - Render to probe array to the specified layer, do prefiltering. */
Eevee: Probe: Add Cubemap Resolution option.
2017-10-10 21:56:11 +02:00
float mipsize = pinfo->cubemap_res;
const int maxlevel = (int)floorf(log2f(pinfo->cubemap_res));
Eevee: Replace Cubemaps by octahedron maps for env. probes. This enables us to use 2D texture arrays for multiple probes. There is a little artifact with very high roughness caused elongated pixel due to the projection (along every 90° meridian).
2017-05-29 22:03:57 +02:00
const int min_lod_level = 3;
for (int i = 0; i < maxlevel - min_lod_level; i++) {
Eevee: Fix problem with GPU_texture_generate_mipmap This function was called to recreate the lower mip level of the probe texture. But this is not it's usage and it introduced a stall. This patch add cubemap mipmap level regeneration in eevee_effects.c
2017-08-18 22:31:06 +02:00
float bias = (i == 0) ? -1.0f : 1.0f;
Eevee: Replace Cubemaps by octahedron maps for env. probes. This enables us to use 2D texture arrays for multiple probes. There is a little artifact with very high roughness caused elongated pixel due to the projection (along every 90° meridian).
2017-05-29 22:03:57 +02:00
pinfo->texel_size = 1.0f / mipsize;
pinfo->padding_size = powf(2.0f, (float)(maxlevel - min_lod_level - 1 - i));
/* XXX : WHY THE HECK DO WE NEED THIS ??? */
/* padding is incorrect without this! float precision issue? */
if (pinfo->padding_size > 32) {
pinfo->padding_size += 5;
}
if (pinfo->padding_size > 16) {
pinfo->padding_size += 4;
}
else if (pinfo->padding_size > 8) {
pinfo->padding_size += 2;
}
else if (pinfo->padding_size > 4) {
pinfo->padding_size += 1;
}
Eevee: Move Spherical Harmonics to a new Probe UBO. Keep data packing tight to prevent use of padding floats
2017-06-04 16:50:22 +02:00
pinfo->layer = probe_idx;
Eevee: Replace Cubemaps by octahedron maps for env. probes. This enables us to use 2D texture arrays for multiple probes. There is a little artifact with very high roughness caused elongated pixel due to the projection (along every 90° meridian).
2017-05-29 22:03:57 +02:00
pinfo->roughness = (float)i / ((float)maxlevel - 4.0f);
Eevee: Introduction of world preconvolved envmap. For now only compute GGX convolution. The GGX LUT used for the split sum approximation (UE4) is merged with the LTX mag LUT that uses the same parameters (theta and roughness)
2017-04-18 12:50:09 +02:00
pinfo->roughness *= pinfo->roughness; /* Disney Roughness */
pinfo->roughness *= pinfo->roughness; /* Distribute Roughness accros lod more evenly */
CLAMP(pinfo->roughness, 1e-8f, 0.99999f); /* Avoid artifacts */
#if 1 /* Variable Sample count (fast) */
switch (i) {
case 0: pinfo->samples_ct = 1.0f; break;
case 1: pinfo->samples_ct = 16.0f; break;
case 2: pinfo->samples_ct = 32.0f; break;
case 3: pinfo->samples_ct = 64.0f; break;
default: pinfo->samples_ct = 128.0f; break;
}
#else /* Constant Sample count (slow) */
pinfo->samples_ct = 1024.0f;
#endif
pinfo->invsamples_ct = 1.0f / pinfo->samples_ct;
Eevee: Probe: Add Cubemap Resolution option.
2017-10-10 21:56:11 +02:00
pinfo->lodfactor = bias + 0.5f * log((float)(pinfo->target_size * pinfo->target_size) * pinfo->invsamples_ct) / log(2);
Eevee: Introduction of world preconvolved envmap. For now only compute GGX convolution. The GGX LUT used for the split sum approximation (UE4) is merged with the LTX mag LUT that uses the same parameters (theta and roughness)
2017-04-18 12:50:09 +02:00
GPUFramebuffer: Refactor (Part 2) This refactor modernise the use of framebuffers. It also touches a lot of files so breaking down changes we have: - GPUTexture: Allow textures to be attached to more than one GPUFrameBuffer. This allows to create and configure more FBO without the need to attach and detach texture at drawing time. - GPUFrameBuffer: The wrapper starts to mimic opengl a bit closer. This allows to configure the framebuffer inside a context other than the one that will be rendering the framebuffer. We do the actual configuration when binding the FBO. We also Keep track of config validity and save drawbuffers state in the FBO. We remove the different bind/unbind functions. These make little sense now that we have separate contexts. - DRWFrameBuffer: We replace DRW_framebuffer functions by GPU_framebuffer ones to avoid another layer of abstraction. We move the DRW convenience functions to GPUFramebuffer instead and even add new ones. The MACRO GPU_framebuffer_ensure_config is pretty much all you need to create and config a GPUFramebuffer. - DRWTexture: Due to the removal of DRWFrameBuffer, we needed to create functions to create textures for thoses framebuffers. Pool textures are now using default texture parameters for the texture type asked. - DRWManager: Make sure no framebuffer object is bound when doing cache filling. - GPUViewport: Add new color_only_fb and depth_only_fb along with FB API usage update. This let draw engines render to color/depth only target and without the need to attach/detach textures. - WM_window: Assert when a framebuffer is bound when changing context. This balance the fact we are not track ogl context inside GPUFramebuffer. - Eevee, Clay, Mode engines: Update to new API. This comes with a lot of code simplification. This also come with some cleanups in some engine codes.
2018-03-25 17:46:48 +02:00
GPU_framebuffer_ensure_config(&sldata->probe_filter_fb, {
GPU_ATTACHMENT_NONE,
GPU_ATTACHMENT_TEXTURE_MIP(sldata->probe_pool, i)
});
GPU_framebuffer_bind(sldata->probe_filter_fb);
GPU_framebuffer_viewport_set(sldata->probe_filter_fb, 0, 0, mipsize, mipsize);
Eevee: Add Irradiance Grid support Early implementation. Slow and still has quality 3 ways of storing irradiance: - Spherical Harmonics: Have problem with directionnal lighting. - HL2 diffuse cube: Very low resolution but smooth transitions. - Diffuse cube: High storage requirement. Also include some name change.
2017-06-13 17:39:39 +02:00
DRW_draw_pass(psl->probe_glossy_compute);
Eevee: Replace Cubemaps by octahedron maps for env. probes. This enables us to use 2D texture arrays for multiple probes. There is a little artifact with very high roughness caused elongated pixel due to the projection (along every 90° meridian).
2017-05-29 22:03:57 +02:00
mipsize /= 2;
CLAMP_MIN(mipsize, 1);
Eevee: Introduction of world preconvolved envmap. For now only compute GGX convolution. The GGX LUT used for the split sum approximation (UE4) is merged with the LTX mag LUT that uses the same parameters (theta and roughness)
2017-04-18 12:50:09 +02:00
}
Eevee: Replace Cubemaps by octahedron maps for env. probes. This enables us to use 2D texture arrays for multiple probes. There is a little artifact with very high roughness caused elongated pixel due to the projection (along every 90° meridian).
2017-05-29 22:03:57 +02:00
/* For shading, save max level of the octahedron map */
Eevee: Put all constant uniforms in a global UBO. This is an optimization / cleanup commit. The use of a global ubo remove lots of uniform lookups and only transfert data when needed. Lots of renaming for more consistent codestyle.
2018-01-21 17:25:10 +01:00
sldata->common_data.prb_lod_cube_max = (float)(maxlevel - min_lod_level) - 1.0f;
Eevee: Add Irradiance Grid support Early implementation. Slow and still has quality 3 ways of storing irradiance: - Spherical Harmonics: Have problem with directionnal lighting. - HL2 diffuse cube: Very low resolution but smooth transitions. - Diffuse cube: High storage requirement. Also include some name change.
2017-06-13 17:39:39 +02:00
}
/* Diffuse filter probe_rt to irradiance_pool at index probe_idx */
Eevee: Irradiance Visibility: Initial Implementation This augment the existing irradiance grid with a new visibility precomputation. We store a small shadowmap for each grid sample so that light does not leak through walls and such. The visibility parameter are similar to the one used by the Variance Shadow Map for point lights. Technical details: We store the visibility in the same texture (array) as the irradiance itself (in order to reduce the number of sampler). But the irradiance and the visibility are not the same data so we must encode them in order to use the same texture format. We use RGBA8 normalized texture and encode irradiance as RGBE (shared exponent). Using RGBE encoding instead of R11_G11_B10 may lead to some lighting changes, but quality seems to be nearly the same in my test cases. Using full RGBA16/32F maybe a future option but that will require much more memory and reduce the perf significantly. Visibility moments (VSM) are encoded as 16bits fixed point precision using a special range. This seems to retain enough precision for the needs. Also interpolation does not seems to be big problem (even though it's incorrect).
2017-12-02 14:28:29 +01:00
static void diffuse_filter_probe(
EEVEE_ViewLayerData *sldata, EEVEE_Data *vedata, EEVEE_PassList *psl, int offset,
Eevee: Probes: Add support for intensity tweak. This works for grid and cubemaps. The intensity is baked into the map itself. Thus you need to refresh/rebake the probe to see the changes.
2018-01-21 23:15:57 +01:00
float clipsta, float clipend, float vis_range, float vis_blur, float intensity)
Eevee: Add Irradiance Grid support Early implementation. Slow and still has quality 3 ways of storing irradiance: - Spherical Harmonics: Have problem with directionnal lighting. - HL2 diffuse cube: Very low resolution but smooth transitions. - Diffuse cube: High storage requirement. Also include some name change.
2017-06-13 17:39:39 +02:00
{
Eevee: Put all constant uniforms in a global UBO. This is an optimization / cleanup commit. The use of a global ubo remove lots of uniform lookups and only transfert data when needed. Lots of renaming for more consistent codestyle.
2018-01-21 17:25:10 +01:00
EEVEE_CommonUniformBuffer *common_data = &sldata->common_data;
Eevee: Add Irradiance Grid support Early implementation. Slow and still has quality 3 ways of storing irradiance: - Spherical Harmonics: Have problem with directionnal lighting. - HL2 diffuse cube: Very low resolution but smooth transitions. - Diffuse cube: High storage requirement. Also include some name change.
2017-06-13 17:39:39 +02:00
EEVEE_LightProbesInfo *pinfo = sldata->probes;
Eevee: Probes: Add support for intensity tweak. This works for grid and cubemaps. The intensity is baked into the map itself. Thus you need to refresh/rebake the probe to see the changes.
2018-01-21 23:15:57 +01:00
pinfo->intensity_fac = intensity;
Eevee: Irradiance Grid: Allocate needed resources instead of a static chunck. This commit makes the irradiance pool and render target sizes depend on the number of irradiance sample in the whole ViewLayer.
2017-12-04 17:20:20 +01:00
int pool_size[3];
Eevee: Put all constant uniforms in a global UBO. This is an optimization / cleanup commit. The use of a global ubo remove lots of uniform lookups and only transfert data when needed. Lots of renaming for more consistent codestyle.
2018-01-21 17:25:10 +01:00
irradiance_pool_size_get(common_data->prb_irradiance_vis_size, pinfo->total_irradiance_samples, pool_size);
Eevee: Irradiance Grid: Allocate needed resources instead of a static chunck. This commit makes the irradiance pool and render target sizes depend on the number of irradiance sample in the whole ViewLayer.
2017-12-04 17:20:20 +01:00
Eevee: Add Irradiance Grid support Early implementation. Slow and still has quality 3 ways of storing irradiance: - Spherical Harmonics: Have problem with directionnal lighting. - HL2 diffuse cube: Very low resolution but smooth transitions. - Diffuse cube: High storage requirement. Also include some name change.
2017-06-13 17:39:39 +02:00
/* find cell position on the virtual 3D texture */
/* NOTE : Keep in sync with load_irradiance_cell() */
#if defined(IRRADIANCE_SH_L2)
int size[2] = {3, 3};
#elif defined(IRRADIANCE_CUBEMAP)
int size[2] = {8, 8};
pinfo->samples_ct = 1024.0f;
#elif defined(IRRADIANCE_HL2)
int size[2] = {3, 2};
pinfo->samples_ct = 1024.0f;
#endif
Eevee: Irradiance Grid: Allocate needed resources instead of a static chunck. This commit makes the irradiance pool and render target sizes depend on the number of irradiance sample in the whole ViewLayer.
2017-12-04 17:20:20 +01:00
int cell_per_row = pool_size[0] / size[0];
Eevee: Irradiance grid: support for non-blocking update and multiple bounces.
2017-06-14 18:52:53 +02:00
int x = size[0] * (offset % cell_per_row);
int y = size[1] * (offset / cell_per_row);
Eevee: Add Irradiance Grid support Early implementation. Slow and still has quality 3 ways of storing irradiance: - Spherical Harmonics: Have problem with directionnal lighting. - HL2 diffuse cube: Very low resolution but smooth transitions. - Diffuse cube: High storage requirement. Also include some name change.
2017-06-13 17:39:39 +02:00
#ifndef IRRADIANCE_SH_L2
Eevee: Fix problem with GPU_texture_generate_mipmap This function was called to recreate the lower mip level of the probe texture. But this is not it's usage and it introduced a stall. This patch add cubemap mipmap level regeneration in eevee_effects.c
2017-08-18 22:31:06 +02:00
/* Tweaking parameters to balance perf. vs precision */
Eevee: Add Irradiance Grid support Early implementation. Slow and still has quality 3 ways of storing irradiance: - Spherical Harmonics: Have problem with directionnal lighting. - HL2 diffuse cube: Very low resolution but smooth transitions. - Diffuse cube: High storage requirement. Also include some name change.
2017-06-13 17:39:39 +02:00
const float bias = 0.0f;
pinfo->invsamples_ct = 1.0f / pinfo->samples_ct;
Eevee: Probe: Add Cubemap Resolution option.
2017-10-10 21:56:11 +02:00
pinfo->lodfactor = bias + 0.5f * log((float)(pinfo->target_size * pinfo->target_size) * pinfo->invsamples_ct) / log(2);
pinfo->lod_rt_max = floorf(log2f(pinfo->target_size)) - 2.0f;
Eevee: Add Irradiance Grid support Early implementation. Slow and still has quality 3 ways of storing irradiance: - Spherical Harmonics: Have problem with directionnal lighting. - HL2 diffuse cube: Very low resolution but smooth transitions. - Diffuse cube: High storage requirement. Also include some name change.
2017-06-13 17:39:39 +02:00
#else
pinfo->shres = 32; /* Less texture fetches & reduce branches */
Eevee: Refactor of shading code to be more modular. This will enable creating shading models more easily.
2017-06-28 16:31:25 +02:00
pinfo->lod_rt_max = 2.0f; /* Improve cache reuse */
Eevee: Add Irradiance Grid support Early implementation. Slow and still has quality 3 ways of storing irradiance: - Spherical Harmonics: Have problem with directionnal lighting. - HL2 diffuse cube: Very low resolution but smooth transitions. - Diffuse cube: High storage requirement. Also include some name change.
2017-06-13 17:39:39 +02:00
#endif
GPUFramebuffer: Refactor (Part 2) This refactor modernise the use of framebuffers. It also touches a lot of files so breaking down changes we have: - GPUTexture: Allow textures to be attached to more than one GPUFrameBuffer. This allows to create and configure more FBO without the need to attach and detach texture at drawing time. - GPUFrameBuffer: The wrapper starts to mimic opengl a bit closer. This allows to configure the framebuffer inside a context other than the one that will be rendering the framebuffer. We do the actual configuration when binding the FBO. We also Keep track of config validity and save drawbuffers state in the FBO. We remove the different bind/unbind functions. These make little sense now that we have separate contexts. - DRWFrameBuffer: We replace DRW_framebuffer functions by GPU_framebuffer ones to avoid another layer of abstraction. We move the DRW convenience functions to GPUFramebuffer instead and even add new ones. The MACRO GPU_framebuffer_ensure_config is pretty much all you need to create and config a GPUFramebuffer. - DRWTexture: Due to the removal of DRWFrameBuffer, we needed to create functions to create textures for thoses framebuffers. Pool textures are now using default texture parameters for the texture type asked. - DRWManager: Make sure no framebuffer object is bound when doing cache filling. - GPUViewport: Add new color_only_fb and depth_only_fb along with FB API usage update. This let draw engines render to color/depth only target and without the need to attach/detach textures. - WM_window: Assert when a framebuffer is bound when changing context. This balance the fact we are not track ogl context inside GPUFramebuffer. - Eevee, Clay, Mode engines: Update to new API. This comes with a lot of code simplification. This also come with some cleanups in some engine codes.
2018-03-25 17:46:48 +02:00
/* Start fresh */
GPU_framebuffer_ensure_config(&sldata->probe_filter_fb, {
GPU_ATTACHMENT_NONE,
GPU_ATTACHMENT_NONE
});
Eevee: Fix problem with GPU_texture_generate_mipmap This function was called to recreate the lower mip level of the probe texture. But this is not it's usage and it introduced a stall. This patch add cubemap mipmap level regeneration in eevee_effects.c
2017-08-18 22:31:06 +02:00
GPUFramebuffer: Refactor (Part 2) This refactor modernise the use of framebuffers. It also touches a lot of files so breaking down changes we have: - GPUTexture: Allow textures to be attached to more than one GPUFrameBuffer. This allows to create and configure more FBO without the need to attach and detach texture at drawing time. - GPUFrameBuffer: The wrapper starts to mimic opengl a bit closer. This allows to configure the framebuffer inside a context other than the one that will be rendering the framebuffer. We do the actual configuration when binding the FBO. We also Keep track of config validity and save drawbuffers state in the FBO. We remove the different bind/unbind functions. These make little sense now that we have separate contexts. - DRWFrameBuffer: We replace DRW_framebuffer functions by GPU_framebuffer ones to avoid another layer of abstraction. We move the DRW convenience functions to GPUFramebuffer instead and even add new ones. The MACRO GPU_framebuffer_ensure_config is pretty much all you need to create and config a GPUFramebuffer. - DRWTexture: Due to the removal of DRWFrameBuffer, we needed to create functions to create textures for thoses framebuffers. Pool textures are now using default texture parameters for the texture type asked. - DRWManager: Make sure no framebuffer object is bound when doing cache filling. - GPUViewport: Add new color_only_fb and depth_only_fb along with FB API usage update. This let draw engines render to color/depth only target and without the need to attach/detach textures. - WM_window: Assert when a framebuffer is bound when changing context. This balance the fact we are not track ogl context inside GPUFramebuffer. - Eevee, Clay, Mode engines: Update to new API. This comes with a lot of code simplification. This also come with some cleanups in some engine codes.
2018-03-25 17:46:48 +02:00
/* 4 - Compute spherical harmonics */
EEVEE_downsample_cube_buffer(vedata, sldata->probe_rt, (int)(pinfo->lod_rt_max));
GPU_framebuffer_ensure_config(&sldata->probe_filter_fb, {
GPU_ATTACHMENT_NONE,
GPU_ATTACHMENT_TEXTURE_LAYER(sldata->irradiance_rt, 0)
});
GPU_framebuffer_bind(sldata->probe_filter_fb);
GPU_framebuffer_viewport_set(sldata->probe_filter_fb, x, y, size[0], size[1]);
Eevee: Add Irradiance Grid support Early implementation. Slow and still has quality 3 ways of storing irradiance: - Spherical Harmonics: Have problem with directionnal lighting. - HL2 diffuse cube: Very low resolution but smooth transitions. - Diffuse cube: High storage requirement. Also include some name change.
2017-06-13 17:39:39 +02:00
DRW_draw_pass(psl->probe_diffuse_compute);
Eevee: Move Spherical Harmonics to a new Probe UBO. Keep data packing tight to prevent use of padding floats
2017-06-04 16:50:22 +02:00
Eevee: Irradiance Grid: Allocate needed resources instead of a static chunck. This commit makes the irradiance pool and render target sizes depend on the number of irradiance sample in the whole ViewLayer.
2017-12-04 17:20:20 +01:00
/* World irradiance have no visibility */
if (offset > 0) {
/* Compute visibility */
pinfo->samples_ct = 512.0f; /* TODO refine */
pinfo->invsamples_ct = 1.0f / pinfo->samples_ct;
Eevee: Put all constant uniforms in a global UBO. This is an optimization / cleanup commit. The use of a global ubo remove lots of uniform lookups and only transfert data when needed. Lots of renaming for more consistent codestyle.
2018-01-21 17:25:10 +01:00
pinfo->shres = common_data->prb_irradiance_vis_size;
Eevee: Irradiance Grid: Allocate needed resources instead of a static chunck. This commit makes the irradiance pool and render target sizes depend on the number of irradiance sample in the whole ViewLayer.
2017-12-04 17:20:20 +01:00
pinfo->visibility_range = vis_range;
pinfo->visibility_blur = vis_blur;
pinfo->near_clip = -clipsta;
pinfo->far_clip = -clipend;
Eevee: Put all constant uniforms in a global UBO. This is an optimization / cleanup commit. The use of a global ubo remove lots of uniform lookups and only transfert data when needed. Lots of renaming for more consistent codestyle.
2018-01-21 17:25:10 +01:00
pinfo->texel_size = 1.0f / (float)common_data->prb_irradiance_vis_size;
Eevee: Irradiance Grid: Allocate needed resources instead of a static chunck. This commit makes the irradiance pool and render target sizes depend on the number of irradiance sample in the whole ViewLayer.
2017-12-04 17:20:20 +01:00
Eevee: Put all constant uniforms in a global UBO. This is an optimization / cleanup commit. The use of a global ubo remove lots of uniform lookups and only transfert data when needed. Lots of renaming for more consistent codestyle.
2018-01-21 17:25:10 +01:00
int cell_per_col = pool_size[1] / common_data->prb_irradiance_vis_size;
cell_per_row = pool_size[0] / common_data->prb_irradiance_vis_size;
x = common_data->prb_irradiance_vis_size * (offset % cell_per_row);
y = common_data->prb_irradiance_vis_size * ((offset / cell_per_row) % cell_per_col);
Eevee: Irradiance Grid: Allocate needed resources instead of a static chunck. This commit makes the irradiance pool and render target sizes depend on the number of irradiance sample in the whole ViewLayer.
2017-12-04 17:20:20 +01:00
int layer = 1 + ((offset / cell_per_row) / cell_per_col);
GPUFramebuffer: Refactor (Part 2) This refactor modernise the use of framebuffers. It also touches a lot of files so breaking down changes we have: - GPUTexture: Allow textures to be attached to more than one GPUFrameBuffer. This allows to create and configure more FBO without the need to attach and detach texture at drawing time. - GPUFrameBuffer: The wrapper starts to mimic opengl a bit closer. This allows to configure the framebuffer inside a context other than the one that will be rendering the framebuffer. We do the actual configuration when binding the FBO. We also Keep track of config validity and save drawbuffers state in the FBO. We remove the different bind/unbind functions. These make little sense now that we have separate contexts. - DRWFrameBuffer: We replace DRW_framebuffer functions by GPU_framebuffer ones to avoid another layer of abstraction. We move the DRW convenience functions to GPUFramebuffer instead and even add new ones. The MACRO GPU_framebuffer_ensure_config is pretty much all you need to create and config a GPUFramebuffer. - DRWTexture: Due to the removal of DRWFrameBuffer, we needed to create functions to create textures for thoses framebuffers. Pool textures are now using default texture parameters for the texture type asked. - DRWManager: Make sure no framebuffer object is bound when doing cache filling. - GPUViewport: Add new color_only_fb and depth_only_fb along with FB API usage update. This let draw engines render to color/depth only target and without the need to attach/detach textures. - WM_window: Assert when a framebuffer is bound when changing context. This balance the fact we are not track ogl context inside GPUFramebuffer. - Eevee, Clay, Mode engines: Update to new API. This comes with a lot of code simplification. This also come with some cleanups in some engine codes.
2018-03-25 17:46:48 +02:00
const int vis_size = common_data->prb_irradiance_vis_size;
Eevee: Irradiance Grid: Allocate needed resources instead of a static chunck. This commit makes the irradiance pool and render target sizes depend on the number of irradiance sample in the whole ViewLayer.
2017-12-04 17:20:20 +01:00
GPUFramebuffer: Refactor (Part 2) This refactor modernise the use of framebuffers. It also touches a lot of files so breaking down changes we have: - GPUTexture: Allow textures to be attached to more than one GPUFrameBuffer. This allows to create and configure more FBO without the need to attach and detach texture at drawing time. - GPUFrameBuffer: The wrapper starts to mimic opengl a bit closer. This allows to configure the framebuffer inside a context other than the one that will be rendering the framebuffer. We do the actual configuration when binding the FBO. We also Keep track of config validity and save drawbuffers state in the FBO. We remove the different bind/unbind functions. These make little sense now that we have separate contexts. - DRWFrameBuffer: We replace DRW_framebuffer functions by GPU_framebuffer ones to avoid another layer of abstraction. We move the DRW convenience functions to GPUFramebuffer instead and even add new ones. The MACRO GPU_framebuffer_ensure_config is pretty much all you need to create and config a GPUFramebuffer. - DRWTexture: Due to the removal of DRWFrameBuffer, we needed to create functions to create textures for thoses framebuffers. Pool textures are now using default texture parameters for the texture type asked. - DRWManager: Make sure no framebuffer object is bound when doing cache filling. - GPUViewport: Add new color_only_fb and depth_only_fb along with FB API usage update. This let draw engines render to color/depth only target and without the need to attach/detach textures. - WM_window: Assert when a framebuffer is bound when changing context. This balance the fact we are not track ogl context inside GPUFramebuffer. - Eevee, Clay, Mode engines: Update to new API. This comes with a lot of code simplification. This also come with some cleanups in some engine codes.
2018-03-25 17:46:48 +02:00
GPU_framebuffer_ensure_config(&sldata->probe_filter_fb, {
GPU_ATTACHMENT_NONE,
GPU_ATTACHMENT_TEXTURE_LAYER(sldata->irradiance_rt, layer)
});
GPU_framebuffer_bind(sldata->probe_filter_fb);
GPU_framebuffer_viewport_set(sldata->probe_filter_fb, x, y, vis_size, vis_size);
Eevee: Irradiance Grid: Allocate needed resources instead of a static chunck. This commit makes the irradiance pool and render target sizes depend on the number of irradiance sample in the whole ViewLayer.
2017-12-04 17:20:20 +01:00
DRW_draw_pass(psl->probe_visibility_compute);
}
Eevee: Introduction of world preconvolved envmap. For now only compute GGX convolution. The GGX LUT used for the split sum approximation (UE4) is merged with the LTX mag LUT that uses the same parameters (theta and roughness)
2017-04-18 12:50:09 +02:00
}
Eevee: Replace Cubemaps by octahedron maps for env. probes. This enables us to use 2D texture arrays for multiple probes. There is a little artifact with very high roughness caused elongated pixel due to the projection (along every 90° meridian).
2017-05-29 22:03:57 +02:00
Eevee: Add Irradiance Grid support Early implementation. Slow and still has quality 3 ways of storing irradiance: - Spherical Harmonics: Have problem with directionnal lighting. - HL2 diffuse cube: Very low resolution but smooth transitions. - Diffuse cube: High storage requirement. Also include some name change.
2017-06-13 17:39:39 +02:00
/* Render the scene to the probe_rt texture. */
static void render_scene_to_probe(
Rename any instance of scene layer or render layer in code with view layer The RenderResult struct still has a listbase of RenderLayer, but that's ok since this is strictly for rendering. * Subversion bump (to 2.80.2) * DNA low level doversion (renames) - only for .blend created since 2.80 started Note: We can't use DNA_struct_elem_find or get file version in init_structDNA, so we are manually iterating over the array of the SDNA elements instead. Note 2: This doversion change with renames can be reverted in a few months. But so far it's required for 2.8 files created between October 2016 and now. Reviewers: campbellbarton, sergey Differential Revision: https://developer.blender.org/D2927
2017-11-22 10:52:39 -02:00
EEVEE_ViewLayerData *sldata, EEVEE_Data *vedata,
Eevee: Add Irradiance Grid support Early implementation. Slow and still has quality 3 ways of storing irradiance: - Spherical Harmonics: Have problem with directionnal lighting. - HL2 diffuse cube: Very low resolution but smooth transitions. - Diffuse cube: High storage requirement. Also include some name change.
2017-06-13 17:39:39 +02:00
const float pos[3], float clipsta, float clipend)
Eevee: First commit of Probe support.
2017-06-08 21:48:50 +02:00
{
Eevee: Add AO support in planar reflection. Technically this enables the use of MinmaxZ pyramid inside the probe captures. Also Disable AO for cubemap probes because it shows big discontinuities at cubeface limits.
2017-06-23 21:08:53 +02:00
EEVEE_TextureList *txl = vedata->txl;
EEVEE_PassList *psl = vedata->psl;
Rename probe to light-probe Probe is a real general term, the new name is used often in docs online.
2017-06-12 20:59:54 +10:00
EEVEE_LightProbesInfo *pinfo = sldata->probes;
Eevee: First commit of Probe support.
2017-06-08 21:48:50 +02:00
Eevee: More use of DRW_viewport_matrix_override_set_all
2018-03-03 01:48:16 +01:00
DRWMatrixState matstate;
float (*viewmat)[4] = matstate.mat[DRW_MAT_VIEW];
float (*viewinv)[4] = matstate.mat[DRW_MAT_VIEWINV];
float (*persmat)[4] = matstate.mat[DRW_MAT_PERS];
float (*persinv)[4] = matstate.mat[DRW_MAT_PERSINV];
float (*winmat)[4] = matstate.mat[DRW_MAT_WIN];
float (*wininv)[4] = matstate.mat[DRW_MAT_WININV];
float posmat[4][4];
Eevee: First commit of Probe support.
2017-06-08 21:48:50 +02:00
unit_m4(posmat);
/* Move to capture position */
Eevee: Add Irradiance Grid support Early implementation. Slow and still has quality 3 ways of storing irradiance: - Spherical Harmonics: Have problem with directionnal lighting. - HL2 diffuse cube: Very low resolution but smooth transitions. - Diffuse cube: High storage requirement. Also include some name change.
2017-06-13 17:39:39 +02:00
negate_v3_v3(posmat[3], pos);
Eevee: First commit of Probe support.
2017-06-08 21:48:50 +02:00
/* 1 - Render to each cubeface individually.
* We do this instead of using geometry shader because a) it's faster,
* b) it's easier than fixing the nodetree shaders (for view dependant effects). */
pinfo->layer = 0;
Eevee: Add Irradiance Grid support Early implementation. Slow and still has quality 3 ways of storing irradiance: - Spherical Harmonics: Have problem with directionnal lighting. - HL2 diffuse cube: Very low resolution but smooth transitions. - Diffuse cube: High storage requirement. Also include some name change.
2017-06-13 17:39:39 +02:00
perspective_m4(winmat, -clipsta, clipsta, -clipsta, clipsta, clipsta, clipend);
Eevee: First commit of Probe support.
2017-06-08 21:48:50 +02:00
Eevee: Add AO support in planar reflection. Technically this enables the use of MinmaxZ pyramid inside the probe captures. Also Disable AO for cubemap probes because it shows big discontinuities at cubeface limits.
2017-06-23 21:08:53 +02:00
/* Avoid using the texture attached to framebuffer when rendering. */
/* XXX */
GPUTexture *tmp_planar_pool = txl->planar_pool;
Eevee: HiZ buffer: Split into two 24bit depth buffer This way we don't have float precision issue we had before and we save some bandwidth.
2017-07-21 14:27:16 +02:00
GPUTexture *tmp_maxz = txl->maxzbuffer;
Eevee: Add AO support in planar reflection. Technically this enables the use of MinmaxZ pyramid inside the probe captures. Also Disable AO for cubemap probes because it shows big discontinuities at cubeface limits.
2017-06-23 21:08:53 +02:00
txl->planar_pool = e_data.planar_pool_placeholder;
Eevee: HiZ buffer: Split into two 24bit depth buffer This way we don't have float precision issue we had before and we save some bandwidth.
2017-07-21 14:27:16 +02:00
txl->maxzbuffer = e_data.depth_placeholder;
Eevee: Add AO support in planar reflection. Technically this enables the use of MinmaxZ pyramid inside the probe captures. Also Disable AO for cubemap probes because it shows big discontinuities at cubeface limits.
2017-06-23 21:08:53 +02:00
EEVEE: Group Timing stats for probe rendering.
2018-04-24 13:01:04 +02:00
DRW_stats_group_start("Cubemap Render");
Eevee: Put all constant uniforms in a global UBO. This is an optimization / cleanup commit. The use of a global ubo remove lots of uniform lookups and only transfert data when needed. Lots of renaming for more consistent codestyle.
2018-01-21 17:25:10 +01:00
/* Update common uniforms */
DRW_uniformbuffer_update(sldata->common_ubo, &sldata->common_data);
Eevee: First commit of Probe support.
2017-06-08 21:48:50 +02:00
for (int i = 0; i < 6; ++i) {
EEVEE: LightProbe: Implement Visibility Group. Works on every probe type. The function to see is EEVEE_lightprobes_obj_visibility_cb. Set pinfo->vis_data.cached to true makes the computation faster for multiple views using the same group. We could even sort the probes by group for that mater to speed things up even more (only applies to dynamic probes like the planar reflections because other probes are only rendered one at a time).
2018-04-24 12:45:59 +02:00
/* Recompute only on 1st drawloop. */
pinfo->vis_data.cached = (i != 0);
EEVEE: Group Timing stats for probe rendering.
2018-04-24 13:01:04 +02:00
DRW_stats_group_start("Cubemap Face");
Eevee: First commit of Probe support.
2017-06-08 21:48:50 +02:00
/* Setup custom matrices */
mul_m4_m4m4(viewmat, cubefacemat[i], posmat);
mul_m4_m4m4(persmat, winmat, viewmat);
invert_m4_m4(persinv, persmat);
invert_m4_m4(viewinv, viewmat);
Eevee: Implement Temporal Anti Aliasing / Super Sampling This adds TAA to eevee. The only thing important to note is that we need to keep the unjittered depth buffer so that the other engines are composited correctly.
2017-09-25 20:14:07 +02:00
invert_m4_m4(wininv, winmat);
Eevee: First commit of Probe support.
2017-06-08 21:48:50 +02:00
Eevee: More use of DRW_viewport_matrix_override_set_all
2018-03-03 01:48:16 +01:00
DRW_viewport_matrix_override_set_all(&matstate);
Eevee: First commit of Probe support.
2017-06-08 21:48:50 +02:00
Eevee: Shadows: Update cascaded shadowmaps when rendering probes. This is really resource intensive but there is no other way to correctly handle it.
2017-09-10 03:07:55 +02:00
/* Be sure that cascaded shadow maps are updated. */
EEVEE_draw_shadows(sldata, psl);
GPUFramebuffer: Refactor (Part 2) This refactor modernise the use of framebuffers. It also touches a lot of files so breaking down changes we have: - GPUTexture: Allow textures to be attached to more than one GPUFrameBuffer. This allows to create and configure more FBO without the need to attach and detach texture at drawing time. - GPUFrameBuffer: The wrapper starts to mimic opengl a bit closer. This allows to configure the framebuffer inside a context other than the one that will be rendering the framebuffer. We do the actual configuration when binding the FBO. We also Keep track of config validity and save drawbuffers state in the FBO. We remove the different bind/unbind functions. These make little sense now that we have separate contexts. - DRWFrameBuffer: We replace DRW_framebuffer functions by GPU_framebuffer ones to avoid another layer of abstraction. We move the DRW convenience functions to GPUFramebuffer instead and even add new ones. The MACRO GPU_framebuffer_ensure_config is pretty much all you need to create and config a GPUFramebuffer. - DRWTexture: Due to the removal of DRWFrameBuffer, we needed to create functions to create textures for thoses framebuffers. Pool textures are now using default texture parameters for the texture type asked. - DRWManager: Make sure no framebuffer object is bound when doing cache filling. - GPUViewport: Add new color_only_fb and depth_only_fb along with FB API usage update. This let draw engines render to color/depth only target and without the need to attach/detach textures. - WM_window: Assert when a framebuffer is bound when changing context. This balance the fact we are not track ogl context inside GPUFramebuffer. - Eevee, Clay, Mode engines: Update to new API. This comes with a lot of code simplification. This also come with some cleanups in some engine codes.
2018-03-25 17:46:48 +02:00
GPU_framebuffer_bind(sldata->probe_face_fb[i]);
GPU_framebuffer_clear_depth(sldata->probe_face_fb[i], 1.0);
Eevee: Shadows: Update cascaded shadowmaps when rendering probes. This is really resource intensive but there is no other way to correctly handle it.
2017-09-10 03:07:55 +02:00
Eevee: First commit of Probe support.
2017-06-08 21:48:50 +02:00
/* Depth prepass */
DRW_draw_pass(psl->depth_pass);
DRW_draw_pass(psl->depth_pass_cull);
Eevee: Draw background after depth prepass. Reduce overdraw. 0.4ms goes to 0.01ms if background is completely occluded.
2017-07-31 15:15:42 +02:00
DRW_draw_pass(psl->probe_background);
Eevee: Irradiance Visibility: Initial Implementation This augment the existing irradiance grid with a new visibility precomputation. We store a small shadowmap for each grid sample so that light does not leak through walls and such. The visibility parameter are similar to the one used by the Variance Shadow Map for point lights. Technical details: We store the visibility in the same texture (array) as the irradiance itself (in order to reduce the number of sampler). But the irradiance and the visibility are not the same data so we must encode them in order to use the same texture format. We use RGBA8 normalized texture and encode irradiance as RGBE (shared exponent). Using RGBE encoding instead of R11_G11_B10 may lead to some lighting changes, but quality seems to be nearly the same in my test cases. Using full RGBA16/32F maybe a future option but that will require much more memory and reduce the perf significantly. Visibility moments (VSM) are encoded as 16bits fixed point precision using a special range. This seems to retain enough precision for the needs. Also interpolation does not seems to be big problem (even though it's incorrect).
2017-12-02 14:28:29 +01:00
// EEVEE_create_minmax_buffer(vedata, sldata->probe_depth_rt);
Eevee: Add AO support in planar reflection. Technically this enables the use of MinmaxZ pyramid inside the probe captures. Also Disable AO for cubemap probes because it shows big discontinuities at cubeface limits.
2017-06-23 21:08:53 +02:00
GPUFramebuffer: Refactor (Part 2) This refactor modernise the use of framebuffers. It also touches a lot of files so breaking down changes we have: - GPUTexture: Allow textures to be attached to more than one GPUFrameBuffer. This allows to create and configure more FBO without the need to attach and detach texture at drawing time. - GPUFrameBuffer: The wrapper starts to mimic opengl a bit closer. This allows to configure the framebuffer inside a context other than the one that will be rendering the framebuffer. We do the actual configuration when binding the FBO. We also Keep track of config validity and save drawbuffers state in the FBO. We remove the different bind/unbind functions. These make little sense now that we have separate contexts. - DRWFrameBuffer: We replace DRW_framebuffer functions by GPU_framebuffer ones to avoid another layer of abstraction. We move the DRW convenience functions to GPUFramebuffer instead and even add new ones. The MACRO GPU_framebuffer_ensure_config is pretty much all you need to create and config a GPUFramebuffer. - DRWTexture: Due to the removal of DRWFrameBuffer, we needed to create functions to create textures for thoses framebuffers. Pool textures are now using default texture parameters for the texture type asked. - DRWManager: Make sure no framebuffer object is bound when doing cache filling. - GPUViewport: Add new color_only_fb and depth_only_fb along with FB API usage update. This let draw engines render to color/depth only target and without the need to attach/detach textures. - WM_window: Assert when a framebuffer is bound when changing context. This balance the fact we are not track ogl context inside GPUFramebuffer. - Eevee, Clay, Mode engines: Update to new API. This comes with a lot of code simplification. This also come with some cleanups in some engine codes.
2018-03-25 17:46:48 +02:00
/* Rebind Target FB */
GPU_framebuffer_bind(sldata->probe_face_fb[i]);
Eevee: Add AO support in planar reflection. Technically this enables the use of MinmaxZ pyramid inside the probe captures. Also Disable AO for cubemap probes because it shows big discontinuities at cubeface limits.
2017-06-23 21:08:53 +02:00
Eevee: First commit of Probe support.
2017-06-08 21:48:50 +02:00
/* Shading pass */
Eevee: Improve material variation managment. - Unify variations between default shaders and material shader. - Only create default shader passes if needed. Downside is that we have a big array of passes and shading grp in the vedata ... And it will double in size each time a new variation flag is added.
2017-06-22 02:41:17 +02:00
EEVEE_draw_default_passes(psl);
Eevee: First commit of Probe support.
2017-06-08 21:48:50 +02:00
DRW_draw_pass(psl->material_pass);
Eevee: Add SSS objects to probe rendering. This does NOT do the SSS diffusion when rendering probe but at least you can see your object without SSS in reflections / bounce light.
2017-11-14 21:13:38 +01:00
DRW_draw_pass(psl->sss_pass); /* Only output standard pass */
EEVEE: Group Timing stats for probe rendering.
2018-04-24 13:01:04 +02:00
DRW_stats_group_end();
Eevee: First commit of Probe support.
2017-06-08 21:48:50 +02:00
}
EEVEE: Group Timing stats for probe rendering.
2018-04-24 13:01:04 +02:00
DRW_stats_group_end();
EEVEE: LightProbe: Implement Visibility Group. Works on every probe type. The function to see is EEVEE_lightprobes_obj_visibility_cb. Set pinfo->vis_data.cached to true makes the computation faster for multiple views using the same group. We could even sort the probes by group for that mater to speed things up even more (only applies to dynamic probes like the planar reflections because other probes are only rendered one at a time).
2018-04-24 12:45:59 +02:00
/* Make sure no aditionnal visibility check runs after this. */
Collections and groups unification OVERVIEW * In 2.7 terminology, all layers and groups are now collection datablocks. * These collections are nestable, linkable, instanceable, overrideable, .. which opens up new ways to set up scenes and link + override data. * Viewport/render visibility and selectability are now a part of the collection and shared across all view layers and linkable. * View layers define which subset of the scene collection hierarchy is excluded for each. For many workflows one view layer can be used, these are more of an advanced feature now. OUTLINER * The outliner now has a "View Layer" display mode instead of "Collections", which can display the collections and/or objects in the view layer. * In this display mode, collections can be excluded with the right click menu. These will then be greyed out and their objects will be excluded. * To view collections not linked to any scene, the "Blender File" display mode can be used, with the new filtering option to just see Colleciton datablocks. * The outliner right click menus for collections and objects were reorganized. * Drag and drop still needs to be improved. Like before, dragging the icon or text gives different results, we'll unify this later. LINKING AND OVERRIDES * Collections can now be linked into the scene without creating an instance, with the link/append operator or from the collections view in the outliner. * Collections can get static overrides with the right click menu in the outliner, but this is rather unreliable and not clearly communicated at the moment. * We still need to improve the make override operator to turn collection instances into collections with overrides directly in the scene. PERFORMANCE * We tried to make performance not worse than before and improve it in some cases. The main thing that's still a bit slower is multiple scenes, we have to change the layer syncing to only updated affected scenes. * Collections keep a list of their parent collections for faster incremental updates in syncing and caching. * View layer bases are now in a object -> base hash to avoid quadratic time lookups internally and in API functions like visible_get(). VERSIONING * Compatibility with 2.7 files should be improved due to the new visibility controls. Of course users may not want to set up their scenes differently now to avoid having separate layers and groups. * Compatibility with 2.8 is mostly there, and was tested on Eevee demo and Hero files. There's a few things which are know to be not quite compatible, like nested layer collections inside groups. * The versioning code for 2.8 files is quite complicated, and isolated behind #ifdef so it can be removed at the end of the release cycle. KNOWN ISSUES * The G-key group operators in the 3D viewport were left mostly as is, they need to be modified still to fit better. * Same for the groups panel in the object properties. This needs to be updated still, or perhaps replaced by something better. * Collections must all have a unique name. Less restrictive namespacing is to be done later, we'll have to see how important this is as all objects within the collections must also have a unique name anyway. * Full scene copy and delete scene are exactly doing the right thing yet. Differential Revision: https://developer.blender.org/D3383 https://code.blender.org/2018/05/collections-and-groups/
2018-04-30 15:57:22 +02:00
pinfo->vis_data.collection = NULL;
EEVEE: LightProbe: Implement Visibility Group. Works on every probe type. The function to see is EEVEE_lightprobes_obj_visibility_cb. Set pinfo->vis_data.cached to true makes the computation faster for multiple views using the same group. We could even sort the probes by group for that mater to speed things up even more (only applies to dynamic probes like the planar reflections because other probes are only rendered one at a time).
2018-04-24 12:45:59 +02:00
Eevee: Disable specular when rendering probes. This prevents weird reflections and really strong indirect lighting.
2017-06-14 10:35:28 +02:00
/* Restore */
Eevee: Add AO support in planar reflection. Technically this enables the use of MinmaxZ pyramid inside the probe captures. Also Disable AO for cubemap probes because it shows big discontinuities at cubeface limits.
2017-06-23 21:08:53 +02:00
txl->planar_pool = tmp_planar_pool;
Eevee: HiZ buffer: Split into two 24bit depth buffer This way we don't have float precision issue we had before and we save some bandwidth.
2017-07-21 14:27:16 +02:00
txl->maxzbuffer = tmp_maxz;
Eevee: First commit of Probe support.
2017-06-08 21:48:50 +02:00
}
Eevee: Initial implementation of planar reflections. Still pretty barebone: No roughness support, No normal distortion support.
2017-06-17 00:08:03 +02:00
static void render_scene_to_planar(
Rename any instance of scene layer or render layer in code with view layer The RenderResult struct still has a listbase of RenderLayer, but that's ok since this is strictly for rendering. * Subversion bump (to 2.80.2) * DNA low level doversion (renames) - only for .blend created since 2.80 started Note: We can't use DNA_struct_elem_find or get file version in init_structDNA, so we are manually iterating over the array of the SDNA elements instead. Note 2: This doversion change with renames can be reverted in a few months. But so far it's required for 2.8 files created between October 2016 and now. Reviewers: campbellbarton, sergey Differential Revision: https://developer.blender.org/D2927
2017-11-22 10:52:39 -02:00
EEVEE_ViewLayerData *sldata, EEVEE_Data *vedata, int layer,
Eevee: Save and reset matrixstate for probe rendering.
2018-03-07 23:25:49 +01:00
EEVEE_LightProbeEngineData *ped)
Eevee: Initial implementation of planar reflections. Still pretty barebone: No roughness support, No normal distortion support.
2017-06-17 00:08:03 +02:00
{
EEVEE: LightProbe: Implement Visibility Group. Works on every probe type. The function to see is EEVEE_lightprobes_obj_visibility_cb. Set pinfo->vis_data.cached to true makes the computation faster for multiple views using the same group. We could even sort the probes by group for that mater to speed things up even more (only applies to dynamic probes like the planar reflections because other probes are only rendered one at a time).
2018-04-24 12:45:59 +02:00
EEVEE_LightProbesInfo *pinfo = sldata->probes;
Eevee: Initial implementation of planar reflections. Still pretty barebone: No roughness support, No normal distortion support.
2017-06-17 00:08:03 +02:00
EEVEE_FramebufferList *fbl = vedata->fbl;
EEVEE_TextureList *txl = vedata->txl;
EEVEE_PassList *psl = vedata->psl;
Eevee: Save and reset matrixstate for probe rendering.
2018-03-07 23:25:49 +01:00
float (*viewmat)[4] = ped->mats.mat[DRW_MAT_VIEW];
float (*viewinv)[4] = ped->mats.mat[DRW_MAT_VIEWINV];
float (*persmat)[4] = ped->mats.mat[DRW_MAT_PERS];
float (*persinv)[4] = ped->mats.mat[DRW_MAT_PERSINV];
float (*winmat)[4] = ped->mats.mat[DRW_MAT_WIN];
float (*wininv)[4] = ped->mats.mat[DRW_MAT_WININV];
Eevee: Initial implementation of planar reflections. Still pretty barebone: No roughness support, No normal distortion support.
2017-06-17 00:08:03 +02:00
invert_m4_m4(viewinv, viewmat);
invert_m4_m4(persinv, persmat);
Eevee: Save and reset matrixstate for probe rendering.
2018-03-07 23:25:49 +01:00
invert_m4_m4(wininv, winmat);
Eevee: Initial implementation of planar reflections. Still pretty barebone: No roughness support, No normal distortion support.
2017-06-17 00:08:03 +02:00
EEVEE: Group Timing stats for probe rendering.
2018-04-24 13:01:04 +02:00
DRW_stats_group_start("Planar Reflection");
Eevee: Save and reset matrixstate for probe rendering.
2018-03-07 23:25:49 +01:00
DRW_viewport_matrix_override_set_all(&ped->mats);
Eevee: Shadows: Update cascaded shadowmaps when rendering probes. This is really resource intensive but there is no other way to correctly handle it.
2017-09-10 03:07:55 +02:00
EEVEE: LightProbe: Implement Visibility Group. Works on every probe type. The function to see is EEVEE_lightprobes_obj_visibility_cb. Set pinfo->vis_data.cached to true makes the computation faster for multiple views using the same group. We could even sort the probes by group for that mater to speed things up even more (only applies to dynamic probes like the planar reflections because other probes are only rendered one at a time).
2018-04-24 12:45:59 +02:00
/* Don't reuse previous visibility. */
pinfo->vis_data.cached = false;
Eevee: Make use of culling when rendering the shadowmaps.
2018-03-02 18:35:25 +01:00
/* Be sure that cascaded shadow maps are updated. */
EEVEE_draw_shadows(sldata, psl);
Eevee: Shadows: Update cascaded shadowmaps when rendering probes. This is really resource intensive but there is no other way to correctly handle it.
2017-09-10 03:07:55 +02:00
/* Since we are rendering with an inverted view matrix, we need
* to invert the facing for backface culling to be the same. */
DRW_state_invert_facing();
Eevee: Add new clipping UBO. This fixes problems with the planar reflections.
2018-03-10 00:02:01 +01:00
/* Set clipping plan */
copy_v4_v4(sldata->clip_data.clip_planes[0], ped->planer_eq_offset);
DRW_uniformbuffer_update(sldata->clip_ubo, &sldata->clip_data);
DRW: Change clip planes API. The draw manager now just set the number of active clip planes. It's now up to the engine to specify the plane equations as uniform/ubo/constant.
2018-03-09 23:49:08 +01:00
DRW_state_clip_planes_count_set(1);
Eevee: Shadows: Update cascaded shadowmaps when rendering probes. This is really resource intensive but there is no other way to correctly handle it.
2017-09-10 03:07:55 +02:00
GPUFramebuffer: Refactor (Part 2) This refactor modernise the use of framebuffers. It also touches a lot of files so breaking down changes we have: - GPUTexture: Allow textures to be attached to more than one GPUFrameBuffer. This allows to create and configure more FBO without the need to attach and detach texture at drawing time. - GPUFrameBuffer: The wrapper starts to mimic opengl a bit closer. This allows to configure the framebuffer inside a context other than the one that will be rendering the framebuffer. We do the actual configuration when binding the FBO. We also Keep track of config validity and save drawbuffers state in the FBO. We remove the different bind/unbind functions. These make little sense now that we have separate contexts. - DRWFrameBuffer: We replace DRW_framebuffer functions by GPU_framebuffer ones to avoid another layer of abstraction. We move the DRW convenience functions to GPUFramebuffer instead and even add new ones. The MACRO GPU_framebuffer_ensure_config is pretty much all you need to create and config a GPUFramebuffer. - DRWTexture: Due to the removal of DRWFrameBuffer, we needed to create functions to create textures for thoses framebuffers. Pool textures are now using default texture parameters for the texture type asked. - DRWManager: Make sure no framebuffer object is bound when doing cache filling. - GPUViewport: Add new color_only_fb and depth_only_fb along with FB API usage update. This let draw engines render to color/depth only target and without the need to attach/detach textures. - WM_window: Assert when a framebuffer is bound when changing context. This balance the fact we are not track ogl context inside GPUFramebuffer. - Eevee, Clay, Mode engines: Update to new API. This comes with a lot of code simplification. This also come with some cleanups in some engine codes.
2018-03-25 17:46:48 +02:00
GPU_framebuffer_ensure_config(&fbl->planarref_fb, {
GPU_ATTACHMENT_TEXTURE_LAYER(txl->planar_depth, layer),
GPU_ATTACHMENT_TEXTURE_LAYER(txl->planar_pool, layer)
});
Eevee: Initial implementation of planar reflections. Still pretty barebone: No roughness support, No normal distortion support.
2017-06-17 00:08:03 +02:00
GPUFramebuffer: Refactor (Part 2) This refactor modernise the use of framebuffers. It also touches a lot of files so breaking down changes we have: - GPUTexture: Allow textures to be attached to more than one GPUFrameBuffer. This allows to create and configure more FBO without the need to attach and detach texture at drawing time. - GPUFrameBuffer: The wrapper starts to mimic opengl a bit closer. This allows to configure the framebuffer inside a context other than the one that will be rendering the framebuffer. We do the actual configuration when binding the FBO. We also Keep track of config validity and save drawbuffers state in the FBO. We remove the different bind/unbind functions. These make little sense now that we have separate contexts. - DRWFrameBuffer: We replace DRW_framebuffer functions by GPU_framebuffer ones to avoid another layer of abstraction. We move the DRW convenience functions to GPUFramebuffer instead and even add new ones. The MACRO GPU_framebuffer_ensure_config is pretty much all you need to create and config a GPUFramebuffer. - DRWTexture: Due to the removal of DRWFrameBuffer, we needed to create functions to create textures for thoses framebuffers. Pool textures are now using default texture parameters for the texture type asked. - DRWManager: Make sure no framebuffer object is bound when doing cache filling. - GPUViewport: Add new color_only_fb and depth_only_fb along with FB API usage update. This let draw engines render to color/depth only target and without the need to attach/detach textures. - WM_window: Assert when a framebuffer is bound when changing context. This balance the fact we are not track ogl context inside GPUFramebuffer. - Eevee, Clay, Mode engines: Update to new API. This comes with a lot of code simplification. This also come with some cleanups in some engine codes.
2018-03-25 17:46:48 +02:00
GPU_framebuffer_bind(fbl->planarref_fb);
GPU_framebuffer_clear_depth(fbl->planarref_fb, 1.0);
Eevee: Initial implementation of planar reflections. Still pretty barebone: No roughness support, No normal distortion support.
2017-06-17 00:08:03 +02:00
/* Avoid using the texture attached to framebuffer when rendering. */
Eevee: Add AO support in planar reflection. Technically this enables the use of MinmaxZ pyramid inside the probe captures. Also Disable AO for cubemap probes because it shows big discontinuities at cubeface limits.
2017-06-23 21:08:53 +02:00
/* XXX */
Eevee: Initial implementation of planar reflections. Still pretty barebone: No roughness support, No normal distortion support.
2017-06-17 00:08:03 +02:00
GPUTexture *tmp_planar_pool = txl->planar_pool;
Eevee: Planar Reflection: Remove distance approximation. This commit separate the depth texture into another texture array. This remove the need to output radial depth into alpha. Unfortunatly it's difficult to recover position from the non linear depth buffer when applying reflection without adding a bunch of stuff. This is in preparation of SSR planar reflections.
2017-07-23 14:03:27 +02:00
GPUTexture *tmp_planar_depth = txl->planar_depth;
Eevee: Planar reflection: Fix bad texture binding.
2017-06-23 19:02:40 +02:00
txl->planar_pool = e_data.planar_pool_placeholder;
Eevee: Planar Reflection: Remove distance approximation. This commit separate the depth texture into another texture array. This remove the need to output radial depth into alpha. Unfortunatly it's difficult to recover position from the non linear depth buffer when applying reflection without adding a bunch of stuff. This is in preparation of SSR planar reflections.
2017-07-23 14:03:27 +02:00
txl->planar_depth = e_data.depth_array_placeholder;
Eevee: Planar reflection: Fix bad texture binding.
2017-06-23 19:02:40 +02:00
Eevee: Planar Reflection: Add refraction support for reflected objects.
2018-03-10 00:03:25 +01:00
/* Slight modification: we handle refraction as normal
* shading and don't do SSRefraction. */
Eevee: Initial implementation of planar reflections. Still pretty barebone: No roughness support, No normal distortion support.
2017-06-17 00:08:03 +02:00
/* Depth prepass */
DRW_draw_pass(psl->depth_pass_clip);
DRW_draw_pass(psl->depth_pass_clip_cull);
Eevee: Planar Reflection: Add refraction support for reflected objects.
2018-03-10 00:03:25 +01:00
DRW_draw_pass(psl->refract_depth_pass);
DRW_draw_pass(psl->refract_depth_pass_cull);
Eevee: Initial implementation of planar reflections. Still pretty barebone: No roughness support, No normal distortion support.
2017-06-17 00:08:03 +02:00
Eevee: Draw background after depth prepass. Reduce overdraw. 0.4ms goes to 0.01ms if background is completely occluded.
2017-07-31 15:15:42 +02:00
/* Background */
DRW_draw_pass(psl->probe_background);
Eevee: Make MinmaxZ compatible with textureArray
2017-07-23 20:33:29 +02:00
EEVEE_create_minmax_buffer(vedata, tmp_planar_depth, layer);
Eevee: Add AO support in planar reflection. Technically this enables the use of MinmaxZ pyramid inside the probe captures. Also Disable AO for cubemap probes because it shows big discontinuities at cubeface limits.
2017-06-23 21:08:53 +02:00
Eevee: Rework GTAO This includes big improvement: - The horizon search is decoupled from the BSDF evaluation. This means using multiple BSDF nodes have a much lower impact when enbaling AO. - The horizon search is optimized by splitting the search into 4 corners searching similar directions to help which GPU cache coherence. - The AO options are now uniforms and do not trigger shader recompilation (aka. freeze UI). - Include a quality slider similar to the SSR one. - Add a switch for disabling bounce light approximation. - Fix problem with Bent Normals when occlusion get very dark. - Add a denoise option to that takes the neighbors pixel values via glsl derivatives. This reduces noise but exhibit 2x2 blocky artifacts. The downside : Separating the horizon search uses more memory (~3MB for each samples on HD viewport). We could lower the bit depth to 4bit per horizon but it produce noticeable banding (might be fixed with some dithering).
2017-08-18 15:06:51 +02:00
/* Compute GTAO Horizons */
Eevee: Fix AO in planar reflections.
2018-01-09 19:01:24 +01:00
EEVEE_occlusion_compute(sldata, vedata, tmp_planar_depth, layer);
Eevee: Rework GTAO This includes big improvement: - The horizon search is decoupled from the BSDF evaluation. This means using multiple BSDF nodes have a much lower impact when enbaling AO. - The horizon search is optimized by splitting the search into 4 corners searching similar directions to help which GPU cache coherence. - The AO options are now uniforms and do not trigger shader recompilation (aka. freeze UI). - Include a quality slider similar to the SSR one. - Add a switch for disabling bounce light approximation. - Fix problem with Bent Normals when occlusion get very dark. - Add a denoise option to that takes the neighbors pixel values via glsl derivatives. This reduces noise but exhibit 2x2 blocky artifacts. The downside : Separating the horizon search uses more memory (~3MB for each samples on HD viewport). We could lower the bit depth to 4bit per horizon but it produce noticeable banding (might be fixed with some dithering).
2017-08-18 15:06:51 +02:00
Eevee: Add AO support in planar reflection. Technically this enables the use of MinmaxZ pyramid inside the probe captures. Also Disable AO for cubemap probes because it shows big discontinuities at cubeface limits.
2017-06-23 21:08:53 +02:00
/* Rebind Planar FB */
GPUFramebuffer: Refactor (Part 2) This refactor modernise the use of framebuffers. It also touches a lot of files so breaking down changes we have: - GPUTexture: Allow textures to be attached to more than one GPUFrameBuffer. This allows to create and configure more FBO without the need to attach and detach texture at drawing time. - GPUFrameBuffer: The wrapper starts to mimic opengl a bit closer. This allows to configure the framebuffer inside a context other than the one that will be rendering the framebuffer. We do the actual configuration when binding the FBO. We also Keep track of config validity and save drawbuffers state in the FBO. We remove the different bind/unbind functions. These make little sense now that we have separate contexts. - DRWFrameBuffer: We replace DRW_framebuffer functions by GPU_framebuffer ones to avoid another layer of abstraction. We move the DRW convenience functions to GPUFramebuffer instead and even add new ones. The MACRO GPU_framebuffer_ensure_config is pretty much all you need to create and config a GPUFramebuffer. - DRWTexture: Due to the removal of DRWFrameBuffer, we needed to create functions to create textures for thoses framebuffers. Pool textures are now using default texture parameters for the texture type asked. - DRWManager: Make sure no framebuffer object is bound when doing cache filling. - GPUViewport: Add new color_only_fb and depth_only_fb along with FB API usage update. This let draw engines render to color/depth only target and without the need to attach/detach textures. - WM_window: Assert when a framebuffer is bound when changing context. This balance the fact we are not track ogl context inside GPUFramebuffer. - Eevee, Clay, Mode engines: Update to new API. This comes with a lot of code simplification. This also come with some cleanups in some engine codes.
2018-03-25 17:46:48 +02:00
GPU_framebuffer_bind(fbl->planarref_fb);
Eevee: Add AO support in planar reflection. Technically this enables the use of MinmaxZ pyramid inside the probe captures. Also Disable AO for cubemap probes because it shows big discontinuities at cubeface limits.
2017-06-23 21:08:53 +02:00
Eevee: Initial implementation of planar reflections. Still pretty barebone: No roughness support, No normal distortion support.
2017-06-17 00:08:03 +02:00
/* Shading pass */
Eevee: Improve material variation managment. - Unify variations between default shaders and material shader. - Only create default shader passes if needed. Downside is that we have a big array of passes and shading grp in the vedata ... And it will double in size each time a new variation flag is added.
2017-06-22 02:41:17 +02:00
EEVEE_draw_default_passes(psl);
Eevee: Initial implementation of planar reflections. Still pretty barebone: No roughness support, No normal distortion support.
2017-06-17 00:08:03 +02:00
DRW_draw_pass(psl->material_pass);
Eevee: Add SSS objects to probe rendering. This does NOT do the SSS diffusion when rendering probe but at least you can see your object without SSS in reflections / bounce light.
2017-11-14 21:13:38 +01:00
DRW_draw_pass(psl->sss_pass); /* Only output standard pass */
Eevee: Planar Reflection: Add refraction support for reflected objects.
2018-03-10 00:03:25 +01:00
DRW_draw_pass(psl->refract_pass);
Eevee: Initial implementation of planar reflections. Still pretty barebone: No roughness support, No normal distortion support.
2017-06-17 00:08:03 +02:00
Eevee: Planar: Add transparent objects. The ordering is left broken for the viewport as this can be quite heavy. Only do it if doing a render.
2018-03-08 01:36:49 +01:00
/* Transparent */
if (DRW_state_is_image_render()) {
/* Do the reordering only for offline because it can be costly. */
DRW_pass_sort_shgroup_z(psl->transparent_pass);
}
DRW_draw_pass(psl->transparent_pass);
Eevee: Initial implementation of planar reflections. Still pretty barebone: No roughness support, No normal distortion support.
2017-06-17 00:08:03 +02:00
DRW_state_invert_facing();
DRW_state_clip_planes_reset();
EEVEE: Group Timing stats for probe rendering.
2018-04-24 13:01:04 +02:00
DRW_stats_group_end();
EEVEE: LightProbe: Implement Visibility Group. Works on every probe type. The function to see is EEVEE_lightprobes_obj_visibility_cb. Set pinfo->vis_data.cached to true makes the computation faster for multiple views using the same group. We could even sort the probes by group for that mater to speed things up even more (only applies to dynamic probes like the planar reflections because other probes are only rendered one at a time).
2018-04-24 12:45:59 +02:00
/* Make sure no aditionnal visibility check runs after this. */
Collections and groups unification OVERVIEW * In 2.7 terminology, all layers and groups are now collection datablocks. * These collections are nestable, linkable, instanceable, overrideable, .. which opens up new ways to set up scenes and link + override data. * Viewport/render visibility and selectability are now a part of the collection and shared across all view layers and linkable. * View layers define which subset of the scene collection hierarchy is excluded for each. For many workflows one view layer can be used, these are more of an advanced feature now. OUTLINER * The outliner now has a "View Layer" display mode instead of "Collections", which can display the collections and/or objects in the view layer. * In this display mode, collections can be excluded with the right click menu. These will then be greyed out and their objects will be excluded. * To view collections not linked to any scene, the "Blender File" display mode can be used, with the new filtering option to just see Colleciton datablocks. * The outliner right click menus for collections and objects were reorganized. * Drag and drop still needs to be improved. Like before, dragging the icon or text gives different results, we'll unify this later. LINKING AND OVERRIDES * Collections can now be linked into the scene without creating an instance, with the link/append operator or from the collections view in the outliner. * Collections can get static overrides with the right click menu in the outliner, but this is rather unreliable and not clearly communicated at the moment. * We still need to improve the make override operator to turn collection instances into collections with overrides directly in the scene. PERFORMANCE * We tried to make performance not worse than before and improve it in some cases. The main thing that's still a bit slower is multiple scenes, we have to change the layer syncing to only updated affected scenes. * Collections keep a list of their parent collections for faster incremental updates in syncing and caching. * View layer bases are now in a object -> base hash to avoid quadratic time lookups internally and in API functions like visible_get(). VERSIONING * Compatibility with 2.7 files should be improved due to the new visibility controls. Of course users may not want to set up their scenes differently now to avoid having separate layers and groups. * Compatibility with 2.8 is mostly there, and was tested on Eevee demo and Hero files. There's a few things which are know to be not quite compatible, like nested layer collections inside groups. * The versioning code for 2.8 files is quite complicated, and isolated behind #ifdef so it can be removed at the end of the release cycle. KNOWN ISSUES * The G-key group operators in the 3D viewport were left mostly as is, they need to be modified still to fit better. * Same for the groups panel in the object properties. This needs to be updated still, or perhaps replaced by something better. * Collections must all have a unique name. Less restrictive namespacing is to be done later, we'll have to see how important this is as all objects within the collections must also have a unique name anyway. * Full scene copy and delete scene are exactly doing the right thing yet. Differential Revision: https://developer.blender.org/D3383 https://code.blender.org/2018/05/collections-and-groups/
2018-04-30 15:57:22 +02:00
pinfo->vis_data.collection = NULL;
EEVEE: LightProbe: Implement Visibility Group. Works on every probe type. The function to see is EEVEE_lightprobes_obj_visibility_cb. Set pinfo->vis_data.cached to true makes the computation faster for multiple views using the same group. We could even sort the probes by group for that mater to speed things up even more (only applies to dynamic probes like the planar reflections because other probes are only rendered one at a time).
2018-04-24 12:45:59 +02:00
Eevee: Initial implementation of planar reflections. Still pretty barebone: No roughness support, No normal distortion support.
2017-06-17 00:08:03 +02:00
/* Restore */
txl->planar_pool = tmp_planar_pool;
Eevee: Planar Reflection: Remove distance approximation. This commit separate the depth texture into another texture array. This remove the need to output radial depth into alpha. Unfortunatly it's difficult to recover position from the non linear depth buffer when applying reflection without adding a bunch of stuff. This is in preparation of SSR planar reflections.
2017-07-23 14:03:27 +02:00
txl->planar_depth = tmp_planar_depth;
Eevee: Initial implementation of planar reflections. Still pretty barebone: No roughness support, No normal distortion support.
2017-06-17 00:08:03 +02:00
}
Rename any instance of scene layer or render layer in code with view layer The RenderResult struct still has a listbase of RenderLayer, but that's ok since this is strictly for rendering. * Subversion bump (to 2.80.2) * DNA low level doversion (renames) - only for .blend created since 2.80 started Note: We can't use DNA_struct_elem_find or get file version in init_structDNA, so we are manually iterating over the array of the SDNA elements instead. Note 2: This doversion change with renames can be reverted in a few months. But so far it's required for 2.8 files created between October 2016 and now. Reviewers: campbellbarton, sergey Differential Revision: https://developer.blender.org/D2927
2017-11-22 10:52:39 -02:00
static void render_world_to_probe(EEVEE_ViewLayerData *sldata, EEVEE_PassList *psl)
Eevee: First commit of Probe support.
2017-06-08 21:48:50 +02:00
{
Rename probe to light-probe Probe is a real general term, the new name is used often in docs online.
2017-06-12 20:59:54 +10:00
EEVEE_LightProbesInfo *pinfo = sldata->probes;
Eevee: Save and reset matrixstate for probe rendering.
2018-03-07 23:25:49 +01:00
DRWMatrixState matstate;
float (*viewmat)[4] = matstate.mat[DRW_MAT_VIEW];
float (*viewinv)[4] = matstate.mat[DRW_MAT_VIEWINV];
float (*persmat)[4] = matstate.mat[DRW_MAT_PERS];
float (*persinv)[4] = matstate.mat[DRW_MAT_PERSINV];
float (*winmat)[4] = matstate.mat[DRW_MAT_WIN];
float (*wininv)[4] = matstate.mat[DRW_MAT_WININV];
Eevee: First commit of Probe support.
2017-06-08 21:48:50 +02:00
/* For world probe, we don't need to clear since we render the background directly. */
pinfo->layer = 0;
Eevee: Remove Geometry shader usage for background. This fix the behaviour of the light path node that separates the probes background from the viewport background.
2017-06-29 17:07:41 +02:00
perspective_m4(winmat, -0.1f, 0.1f, -0.1f, 0.1f, 0.1f, 1.0f);
Eevee: Fix broken World probe.
2017-09-26 21:43:14 +02:00
invert_m4_m4(wininv, winmat);
Eevee: Remove Geometry shader usage for background. This fix the behaviour of the light path node that separates the probes background from the viewport background.
2017-06-29 17:07:41 +02:00
for (int i = 0; i < 6; ++i) {
/* Setup custom matrices */
copy_m4_m4(viewmat, cubefacemat[i]);
mul_m4_m4m4(persmat, winmat, viewmat);
invert_m4_m4(persinv, persmat);
invert_m4_m4(viewinv, viewmat);
Eevee: Save and reset matrixstate for probe rendering.
2018-03-07 23:25:49 +01:00
DRW_viewport_matrix_override_set_all(&matstate);
Eevee: Remove Geometry shader usage for background. This fix the behaviour of the light path node that separates the probes background from the viewport background.
2017-06-29 17:07:41 +02:00
GPUFramebuffer: Refactor (Part 2) This refactor modernise the use of framebuffers. It also touches a lot of files so breaking down changes we have: - GPUTexture: Allow textures to be attached to more than one GPUFrameBuffer. This allows to create and configure more FBO without the need to attach and detach texture at drawing time. - GPUFrameBuffer: The wrapper starts to mimic opengl a bit closer. This allows to configure the framebuffer inside a context other than the one that will be rendering the framebuffer. We do the actual configuration when binding the FBO. We also Keep track of config validity and save drawbuffers state in the FBO. We remove the different bind/unbind functions. These make little sense now that we have separate contexts. - DRWFrameBuffer: We replace DRW_framebuffer functions by GPU_framebuffer ones to avoid another layer of abstraction. We move the DRW convenience functions to GPUFramebuffer instead and even add new ones. The MACRO GPU_framebuffer_ensure_config is pretty much all you need to create and config a GPUFramebuffer. - DRWTexture: Due to the removal of DRWFrameBuffer, we needed to create functions to create textures for thoses framebuffers. Pool textures are now using default texture parameters for the texture type asked. - DRWManager: Make sure no framebuffer object is bound when doing cache filling. - GPUViewport: Add new color_only_fb and depth_only_fb along with FB API usage update. This let draw engines render to color/depth only target and without the need to attach/detach textures. - WM_window: Assert when a framebuffer is bound when changing context. This balance the fact we are not track ogl context inside GPUFramebuffer. - Eevee, Clay, Mode engines: Update to new API. This comes with a lot of code simplification. This also come with some cleanups in some engine codes.
2018-03-25 17:46:48 +02:00
GPU_framebuffer_bind(sldata->probe_face_fb[i]);
GPU_framebuffer_clear_depth(sldata->probe_face_fb[i], 1.0f);
Eevee: Remove Geometry shader usage for background. This fix the behaviour of the light path node that separates the probes background from the viewport background.
2017-06-29 17:07:41 +02:00
DRW_draw_pass(psl->probe_background);
}
Eevee: Add Irradiance Grid support Early implementation. Slow and still has quality 3 ways of storing irradiance: - Spherical Harmonics: Have problem with directionnal lighting. - HL2 diffuse cube: Very low resolution but smooth transitions. - Diffuse cube: High storage requirement. Also include some name change.
2017-06-13 17:39:39 +02:00
}
Eevee: Probe Grid: Add "progressive rendering" This gives the user a more rapid preview of what the final result will be. The grid cells are renderer in a series of powers of two offsets.
2017-09-28 21:16:02 +02:00
static void lightprobe_cell_grid_location_get(EEVEE_LightGrid *egrid, int cell_idx, float r_local_cell[3])
Eevee: Add Irradiance Grid support Early implementation. Slow and still has quality 3 ways of storing irradiance: - Spherical Harmonics: Have problem with directionnal lighting. - HL2 diffuse cube: Very low resolution but smooth transitions. - Diffuse cube: High storage requirement. Also include some name change.
2017-06-13 17:39:39 +02:00
{
Eevee: Irradiance grid: support for non-blocking update and multiple bounces.
2017-06-14 18:52:53 +02:00
/* Keep in sync with lightprobe_grid_display_vert */
Eevee: Probe Grid: Add "progressive rendering" This gives the user a more rapid preview of what the final result will be. The grid cells are renderer in a series of powers of two offsets.
2017-09-28 21:16:02 +02:00
r_local_cell[2] = (float)(cell_idx % egrid->resolution[2]);
r_local_cell[1] = (float)((cell_idx / egrid->resolution[2]) % egrid->resolution[1]);
r_local_cell[0] = (float)(cell_idx / (egrid->resolution[2] * egrid->resolution[1]));
}
static void lightprobe_cell_world_location_get(EEVEE_LightGrid *egrid, float local_cell[3], float r_pos[3])
{
float tmp[3];
Eevee: Irradiance grid: support for non-blocking update and multiple bounces.
2017-06-14 18:52:53 +02:00
copy_v3_v3(r_pos, egrid->corner);
mul_v3_v3fl(tmp, egrid->increment_x, local_cell[0]);
add_v3_v3(r_pos, tmp);
mul_v3_v3fl(tmp, egrid->increment_y, local_cell[1]);
add_v3_v3(r_pos, tmp);
mul_v3_v3fl(tmp, egrid->increment_z, local_cell[2]);
add_v3_v3(r_pos, tmp);
Eevee: First commit of Probe support.
2017-06-08 21:48:50 +02:00
}
Eevee: Cleanup, split long probe refresh function There is no reason to have such a long function, it is really easy to break it down into a smaller ones, and call them from where needed. Makes them smaller and easier to follow. Also avoids use of confusing goto's.
2017-11-29 12:15:22 +01:00
static void lightprobes_refresh_world(EEVEE_ViewLayerData *sldata, EEVEE_Data *vedata)
Eevee: Material code refactor. Separate material handling inside another file. Make use of enums to identify shader variations. Group all 64*64 LUTs into one array texture. Only update world probe if world changes.
2017-06-04 12:12:58 +02:00
{
Eevee: Put all constant uniforms in a global UBO. This is an optimization / cleanup commit. The use of a global ubo remove lots of uniform lookups and only transfert data when needed. Lots of renaming for more consistent codestyle.
2018-01-21 17:25:10 +01:00
EEVEE_CommonUniformBuffer *common_data = &sldata->common_data;
Eevee: Update to support shader deferred compilation. World probe is tagged to refresh when it it's shader is updated. Probes are recomputed only after all meshes shaders have been compiled.
2018-03-06 02:19:28 +01:00
EEVEE_LightProbesInfo *pinfo = sldata->probes;
Eevee: Initial implementation of planar reflections. Still pretty barebone: No roughness support, No normal distortion support.
2017-06-17 00:08:03 +02:00
EEVEE_PassList *psl = vedata->psl;
Eevee: Fix temporal sampling not resetting. World Probe update was not resetting the TAA. This also reset the volume sampling in cases that needs it (world/grid probe diffuse coefs update).
2018-03-13 01:35:16 +01:00
EEVEE_StorageList *stl = vedata->stl;
Eevee: Save and reset matrixstate for probe rendering.
2018-03-07 23:25:49 +01:00
DRWMatrixState saved_mats;
/* We need to save the Matrices before overidding them */
DRW_viewport_matrix_get_all(&saved_mats);
Eevee: Put all constant uniforms in a global UBO. This is an optimization / cleanup commit. The use of a global ubo remove lots of uniform lookups and only transfert data when needed. Lots of renaming for more consistent codestyle.
2018-01-21 17:25:10 +01:00
Eevee: Cleanup, split long probe refresh function There is no reason to have such a long function, it is really easy to break it down into a smaller ones, and call them from where needed. Makes them smaller and easier to follow. Also avoids use of confusing goto's.
2017-11-29 12:15:22 +01:00
render_world_to_probe(sldata, psl);
Eevee: Update to support shader deferred compilation. World probe is tagged to refresh when it it's shader is updated. Probes are recomputed only after all meshes shaders have been compiled.
2018-03-06 02:19:28 +01:00
if (pinfo->update_world & PROBE_UPDATE_CUBE) {
Eevee: Probes: Add support for intensity tweak. This works for grid and cubemaps. The intensity is baked into the map itself. Thus you need to refresh/rebake the probe to see the changes.
2018-01-21 23:15:57 +01:00
glossy_filter_probe(sldata, vedata, psl, 0, 1.0);
Eevee: Fix grid lighting disappearing when adding a new cubemap probe.
2018-01-21 22:57:12 +01:00
common_data->prb_num_render_cube = 1;
Eevee: Cleanup, split long probe refresh function There is no reason to have such a long function, it is really easy to break it down into a smaller ones, and call them from where needed. Makes them smaller and easier to follow. Also avoids use of confusing goto's.
2017-11-29 12:15:22 +01:00
}
Eevee: Update to support shader deferred compilation. World probe is tagged to refresh when it it's shader is updated. Probes are recomputed only after all meshes shaders have been compiled.
2018-03-06 02:19:28 +01:00
if (pinfo->update_world & PROBE_UPDATE_GRID) {
Eevee: Probes: Add support for intensity tweak. This works for grid and cubemaps. The intensity is baked into the map itself. Thus you need to refresh/rebake the probe to see the changes.
2018-01-21 23:15:57 +01:00
diffuse_filter_probe(sldata, vedata, psl, 0, 0.0, 0.0, 0.0, 0.0, 1.0);
GPUFramebuffer: Refactor (Part 2) This refactor modernise the use of framebuffers. It also touches a lot of files so breaking down changes we have: - GPUTexture: Allow textures to be attached to more than one GPUFrameBuffer. This allows to create and configure more FBO without the need to attach and detach texture at drawing time. - GPUFrameBuffer: The wrapper starts to mimic opengl a bit closer. This allows to configure the framebuffer inside a context other than the one that will be rendering the framebuffer. We do the actual configuration when binding the FBO. We also Keep track of config validity and save drawbuffers state in the FBO. We remove the different bind/unbind functions. These make little sense now that we have separate contexts. - DRWFrameBuffer: We replace DRW_framebuffer functions by GPU_framebuffer ones to avoid another layer of abstraction. We move the DRW convenience functions to GPUFramebuffer instead and even add new ones. The MACRO GPU_framebuffer_ensure_config is pretty much all you need to create and config a GPUFramebuffer. - DRWTexture: Due to the removal of DRWFrameBuffer, we needed to create functions to create textures for thoses framebuffers. Pool textures are now using default texture parameters for the texture type asked. - DRWManager: Make sure no framebuffer object is bound when doing cache filling. - GPUViewport: Add new color_only_fb and depth_only_fb along with FB API usage update. This let draw engines render to color/depth only target and without the need to attach/detach textures. - WM_window: Assert when a framebuffer is bound when changing context. This balance the fact we are not track ogl context inside GPUFramebuffer. - Eevee, Clay, Mode engines: Update to new API. This comes with a lot of code simplification. This also come with some cleanups in some engine codes.
2018-03-25 17:46:48 +02:00
Eevee: Cleanup, split long probe refresh function There is no reason to have such a long function, it is really easy to break it down into a smaller ones, and call them from where needed. Makes them smaller and easier to follow. Also avoids use of confusing goto's.
2017-11-29 12:15:22 +01:00
SWAP(GPUTexture *, sldata->irradiance_pool, sldata->irradiance_rt);
GPUFramebuffer: Refactor (Part 2) This refactor modernise the use of framebuffers. It also touches a lot of files so breaking down changes we have: - GPUTexture: Allow textures to be attached to more than one GPUFrameBuffer. This allows to create and configure more FBO without the need to attach and detach texture at drawing time. - GPUFrameBuffer: The wrapper starts to mimic opengl a bit closer. This allows to configure the framebuffer inside a context other than the one that will be rendering the framebuffer. We do the actual configuration when binding the FBO. We also Keep track of config validity and save drawbuffers state in the FBO. We remove the different bind/unbind functions. These make little sense now that we have separate contexts. - DRWFrameBuffer: We replace DRW_framebuffer functions by GPU_framebuffer ones to avoid another layer of abstraction. We move the DRW convenience functions to GPUFramebuffer instead and even add new ones. The MACRO GPU_framebuffer_ensure_config is pretty much all you need to create and config a GPUFramebuffer. - DRWTexture: Due to the removal of DRWFrameBuffer, we needed to create functions to create textures for thoses framebuffers. Pool textures are now using default texture parameters for the texture type asked. - DRWManager: Make sure no framebuffer object is bound when doing cache filling. - GPUViewport: Add new color_only_fb and depth_only_fb along with FB API usage update. This let draw engines render to color/depth only target and without the need to attach/detach textures. - WM_window: Assert when a framebuffer is bound when changing context. This balance the fact we are not track ogl context inside GPUFramebuffer. - Eevee, Clay, Mode engines: Update to new API. This comes with a lot of code simplification. This also come with some cleanups in some engine codes.
2018-03-25 17:46:48 +02:00
GPU_framebuffer_texture_attach(sldata->probe_filter_fb, sldata->irradiance_rt, 0, 0);
GPU_framebuffer_bind(sldata->probe_filter_fb);
Eevee: Cleanup, split long probe refresh function There is no reason to have such a long function, it is really easy to break it down into a smaller ones, and call them from where needed. Makes them smaller and easier to follow. Also avoids use of confusing goto's.
2017-11-29 12:15:22 +01:00
DRW_draw_pass(psl->probe_grid_fill);
GPUFramebuffer: Refactor (Part 2) This refactor modernise the use of framebuffers. It also touches a lot of files so breaking down changes we have: - GPUTexture: Allow textures to be attached to more than one GPUFrameBuffer. This allows to create and configure more FBO without the need to attach and detach texture at drawing time. - GPUFrameBuffer: The wrapper starts to mimic opengl a bit closer. This allows to configure the framebuffer inside a context other than the one that will be rendering the framebuffer. We do the actual configuration when binding the FBO. We also Keep track of config validity and save drawbuffers state in the FBO. We remove the different bind/unbind functions. These make little sense now that we have separate contexts. - DRWFrameBuffer: We replace DRW_framebuffer functions by GPU_framebuffer ones to avoid another layer of abstraction. We move the DRW convenience functions to GPUFramebuffer instead and even add new ones. The MACRO GPU_framebuffer_ensure_config is pretty much all you need to create and config a GPUFramebuffer. - DRWTexture: Due to the removal of DRWFrameBuffer, we needed to create functions to create textures for thoses framebuffers. Pool textures are now using default texture parameters for the texture type asked. - DRWManager: Make sure no framebuffer object is bound when doing cache filling. - GPUViewport: Add new color_only_fb and depth_only_fb along with FB API usage update. This let draw engines render to color/depth only target and without the need to attach/detach textures. - WM_window: Assert when a framebuffer is bound when changing context. This balance the fact we are not track ogl context inside GPUFramebuffer. - Eevee, Clay, Mode engines: Update to new API. This comes with a lot of code simplification. This also come with some cleanups in some engine codes.
2018-03-25 17:46:48 +02:00
Eevee: Put all constant uniforms in a global UBO. This is an optimization / cleanup commit. The use of a global ubo remove lots of uniform lookups and only transfert data when needed. Lots of renaming for more consistent codestyle.
2018-01-21 17:25:10 +01:00
common_data->prb_num_render_grid = 1;
Eevee: Fix temporal sampling not resetting. World Probe update was not resetting the TAA. This also reset the volume sampling in cases that needs it (world/grid probe diffuse coefs update).
2018-03-13 01:35:16 +01:00
/* Reset volume history. */
stl->effects->volume_current_sample = -1;
common_data->vol_history_alpha = 0.0f;
Eevee: Cleanup, split long probe refresh function There is no reason to have such a long function, it is really easy to break it down into a smaller ones, and call them from where needed. Makes them smaller and easier to follow. Also avoids use of confusing goto's.
2017-11-29 12:15:22 +01:00
}
Eevee: Update to support shader deferred compilation. World probe is tagged to refresh when it it's shader is updated. Probes are recomputed only after all meshes shaders have been compiled.
2018-03-06 02:19:28 +01:00
pinfo->update_world = 0;
Eevee: Cleanup, split long probe refresh function There is no reason to have such a long function, it is really easy to break it down into a smaller ones, and call them from where needed. Makes them smaller and easier to follow. Also avoids use of confusing goto's.
2017-11-29 12:15:22 +01:00
DRW_viewport_request_redraw();
Eevee: Fix temporal sampling not resetting. World Probe update was not resetting the TAA. This also reset the volume sampling in cases that needs it (world/grid probe diffuse coefs update).
2018-03-13 01:35:16 +01:00
/* Do not let this frame accumulate. */
stl->effects->taa_current_sample = 1;
Eevee: Save and reset matrixstate for probe rendering.
2018-03-07 23:25:49 +01:00
DRW_viewport_matrix_override_set_all(&saved_mats);
Eevee: Cleanup, split long probe refresh function There is no reason to have such a long function, it is really easy to break it down into a smaller ones, and call them from where needed. Makes them smaller and easier to follow. Also avoids use of confusing goto's.
2017-11-29 12:15:22 +01:00
}
Eevee: First commit of Probe support.
2017-06-08 21:48:50 +02:00
Eevee: Cleanup, split long probe refresh function There is no reason to have such a long function, it is really easy to break it down into a smaller ones, and call them from where needed. Makes them smaller and easier to follow. Also avoids use of confusing goto's.
2017-11-29 12:15:22 +01:00
static void lightprobes_refresh_initialize_grid(EEVEE_ViewLayerData *sldata, EEVEE_Data *vedata)
{
EEVEE_LightProbesInfo *pinfo = sldata->probes;
EEVEE_PassList *psl = vedata->psl;
if (pinfo->grid_initialized) {
Eevee: Irradiance Grid: Fix Multibounce baking. The baking was only recording one bounce because the double buffer setup was buggy.
2017-12-02 00:47:42 +01:00
/* Grid is already initialized, nothing to do. */
Eevee: Cleanup, split long probe refresh function There is no reason to have such a long function, it is really easy to break it down into a smaller ones, and call them from where needed. Makes them smaller and easier to follow. Also avoids use of confusing goto's.
2017-11-29 12:15:22 +01:00
return;
}
/* Flood fill with world irradiance. */
GPUFramebuffer: Refactor (Part 2) This refactor modernise the use of framebuffers. It also touches a lot of files so breaking down changes we have: - GPUTexture: Allow textures to be attached to more than one GPUFrameBuffer. This allows to create and configure more FBO without the need to attach and detach texture at drawing time. - GPUFrameBuffer: The wrapper starts to mimic opengl a bit closer. This allows to configure the framebuffer inside a context other than the one that will be rendering the framebuffer. We do the actual configuration when binding the FBO. We also Keep track of config validity and save drawbuffers state in the FBO. We remove the different bind/unbind functions. These make little sense now that we have separate contexts. - DRWFrameBuffer: We replace DRW_framebuffer functions by GPU_framebuffer ones to avoid another layer of abstraction. We move the DRW convenience functions to GPUFramebuffer instead and even add new ones. The MACRO GPU_framebuffer_ensure_config is pretty much all you need to create and config a GPUFramebuffer. - DRWTexture: Due to the removal of DRWFrameBuffer, we needed to create functions to create textures for thoses framebuffers. Pool textures are now using default texture parameters for the texture type asked. - DRWManager: Make sure no framebuffer object is bound when doing cache filling. - GPUViewport: Add new color_only_fb and depth_only_fb along with FB API usage update. This let draw engines render to color/depth only target and without the need to attach/detach textures. - WM_window: Assert when a framebuffer is bound when changing context. This balance the fact we are not track ogl context inside GPUFramebuffer. - Eevee, Clay, Mode engines: Update to new API. This comes with a lot of code simplification. This also come with some cleanups in some engine codes.
2018-03-25 17:46:48 +02:00
GPU_framebuffer_texture_attach(sldata->probe_filter_fb, sldata->irradiance_rt, 0, 0);
GPU_framebuffer_bind(sldata->probe_filter_fb);
Eevee: Cleanup, split long probe refresh function There is no reason to have such a long function, it is really easy to break it down into a smaller ones, and call them from where needed. Makes them smaller and easier to follow. Also avoids use of confusing goto's.
2017-11-29 12:15:22 +01:00
DRW_draw_pass(psl->probe_grid_fill);
Eevee: Irradiance Grid: Fix Multibounce baking. The baking was only recording one bounce because the double buffer setup was buggy.
2017-12-02 00:47:42 +01:00
Eevee: Cleanup, split long probe refresh function There is no reason to have such a long function, it is really easy to break it down into a smaller ones, and call them from where needed. Makes them smaller and easier to follow. Also avoids use of confusing goto's.
2017-11-29 12:15:22 +01:00
SWAP(GPUTexture *, sldata->irradiance_pool, sldata->irradiance_rt);
GPUFramebuffer: Refactor (Part 2) This refactor modernise the use of framebuffers. It also touches a lot of files so breaking down changes we have: - GPUTexture: Allow textures to be attached to more than one GPUFrameBuffer. This allows to create and configure more FBO without the need to attach and detach texture at drawing time. - GPUFrameBuffer: The wrapper starts to mimic opengl a bit closer. This allows to configure the framebuffer inside a context other than the one that will be rendering the framebuffer. We do the actual configuration when binding the FBO. We also Keep track of config validity and save drawbuffers state in the FBO. We remove the different bind/unbind functions. These make little sense now that we have separate contexts. - DRWFrameBuffer: We replace DRW_framebuffer functions by GPU_framebuffer ones to avoid another layer of abstraction. We move the DRW convenience functions to GPUFramebuffer instead and even add new ones. The MACRO GPU_framebuffer_ensure_config is pretty much all you need to create and config a GPUFramebuffer. - DRWTexture: Due to the removal of DRWFrameBuffer, we needed to create functions to create textures for thoses framebuffers. Pool textures are now using default texture parameters for the texture type asked. - DRWManager: Make sure no framebuffer object is bound when doing cache filling. - GPUViewport: Add new color_only_fb and depth_only_fb along with FB API usage update. This let draw engines render to color/depth only target and without the need to attach/detach textures. - WM_window: Assert when a framebuffer is bound when changing context. This balance the fact we are not track ogl context inside GPUFramebuffer. - Eevee, Clay, Mode engines: Update to new API. This comes with a lot of code simplification. This also come with some cleanups in some engine codes.
2018-03-25 17:46:48 +02:00
GPU_framebuffer_texture_attach(sldata->probe_filter_fb, sldata->irradiance_rt, 0, 0);
GPU_framebuffer_bind(sldata->probe_filter_fb);
Eevee: Cleanup, split long probe refresh function There is no reason to have such a long function, it is really easy to break it down into a smaller ones, and call them from where needed. Makes them smaller and easier to follow. Also avoids use of confusing goto's.
2017-11-29 12:15:22 +01:00
DRW_draw_pass(psl->probe_grid_fill);
Eevee: Irradiance Grid: Fix Multibounce baking. The baking was only recording one bounce because the double buffer setup was buggy.
2017-12-02 00:47:42 +01:00
Eevee: Cleanup, split long probe refresh function There is no reason to have such a long function, it is really easy to break it down into a smaller ones, and call them from where needed. Makes them smaller and easier to follow. Also avoids use of confusing goto's.
2017-11-29 12:15:22 +01:00
SWAP(GPUTexture *, sldata->irradiance_pool, sldata->irradiance_rt);
GPUFramebuffer: Refactor (Part 2) This refactor modernise the use of framebuffers. It also touches a lot of files so breaking down changes we have: - GPUTexture: Allow textures to be attached to more than one GPUFrameBuffer. This allows to create and configure more FBO without the need to attach and detach texture at drawing time. - GPUFrameBuffer: The wrapper starts to mimic opengl a bit closer. This allows to configure the framebuffer inside a context other than the one that will be rendering the framebuffer. We do the actual configuration when binding the FBO. We also Keep track of config validity and save drawbuffers state in the FBO. We remove the different bind/unbind functions. These make little sense now that we have separate contexts. - DRWFrameBuffer: We replace DRW_framebuffer functions by GPU_framebuffer ones to avoid another layer of abstraction. We move the DRW convenience functions to GPUFramebuffer instead and even add new ones. The MACRO GPU_framebuffer_ensure_config is pretty much all you need to create and config a GPUFramebuffer. - DRWTexture: Due to the removal of DRWFrameBuffer, we needed to create functions to create textures for thoses framebuffers. Pool textures are now using default texture parameters for the texture type asked. - DRWManager: Make sure no framebuffer object is bound when doing cache filling. - GPUViewport: Add new color_only_fb and depth_only_fb along with FB API usage update. This let draw engines render to color/depth only target and without the need to attach/detach textures. - WM_window: Assert when a framebuffer is bound when changing context. This balance the fact we are not track ogl context inside GPUFramebuffer. - Eevee, Clay, Mode engines: Update to new API. This comes with a lot of code simplification. This also come with some cleanups in some engine codes.
2018-03-25 17:46:48 +02:00
Eevee: Cleanup, split long probe refresh function There is no reason to have such a long function, it is really easy to break it down into a smaller ones, and call them from where needed. Makes them smaller and easier to follow. Also avoids use of confusing goto's.
2017-11-29 12:15:22 +01:00
pinfo->grid_initialized = true;
}
Eevee: Add Irradiance Grid support Early implementation. Slow and still has quality 3 ways of storing irradiance: - Spherical Harmonics: Have problem with directionnal lighting. - HL2 diffuse cube: Very low resolution but smooth transitions. - Diffuse cube: High storage requirement. Also include some name change.
2017-06-13 17:39:39 +02:00
Eevee: Render: Make sure all probes are refreshed before rendering.
2018-02-03 01:50:38 +01:00
void EEVEE_lightprobes_refresh_planar(EEVEE_ViewLayerData *sldata, EEVEE_Data *vedata)
Eevee: Cleanup, split long probe refresh function There is no reason to have such a long function, it is really easy to break it down into a smaller ones, and call them from where needed. Makes them smaller and easier to follow. Also avoids use of confusing goto's.
2017-11-29 12:15:22 +01:00
{
Eevee: Put all constant uniforms in a global UBO. This is an optimization / cleanup commit. The use of a global ubo remove lots of uniform lookups and only transfert data when needed. Lots of renaming for more consistent codestyle.
2018-01-21 17:25:10 +01:00
EEVEE_CommonUniformBuffer *common_data = &sldata->common_data;
GPUFramebuffer: Refactor (Part 2) This refactor modernise the use of framebuffers. It also touches a lot of files so breaking down changes we have: - GPUTexture: Allow textures to be attached to more than one GPUFrameBuffer. This allows to create and configure more FBO without the need to attach and detach texture at drawing time. - GPUFrameBuffer: The wrapper starts to mimic opengl a bit closer. This allows to configure the framebuffer inside a context other than the one that will be rendering the framebuffer. We do the actual configuration when binding the FBO. We also Keep track of config validity and save drawbuffers state in the FBO. We remove the different bind/unbind functions. These make little sense now that we have separate contexts. - DRWFrameBuffer: We replace DRW_framebuffer functions by GPU_framebuffer ones to avoid another layer of abstraction. We move the DRW convenience functions to GPUFramebuffer instead and even add new ones. The MACRO GPU_framebuffer_ensure_config is pretty much all you need to create and config a GPUFramebuffer. - DRWTexture: Due to the removal of DRWFrameBuffer, we needed to create functions to create textures for thoses framebuffers. Pool textures are now using default texture parameters for the texture type asked. - DRWManager: Make sure no framebuffer object is bound when doing cache filling. - GPUViewport: Add new color_only_fb and depth_only_fb along with FB API usage update. This let draw engines render to color/depth only target and without the need to attach/detach textures. - WM_window: Assert when a framebuffer is bound when changing context. This balance the fact we are not track ogl context inside GPUFramebuffer. - Eevee, Clay, Mode engines: Update to new API. This comes with a lot of code simplification. This also come with some cleanups in some engine codes.
2018-03-25 17:46:48 +02:00
EEVEE_FramebufferList *fbl = vedata->fbl;
Eevee: Cleanup, split long probe refresh function There is no reason to have such a long function, it is really easy to break it down into a smaller ones, and call them from where needed. Makes them smaller and easier to follow. Also avoids use of confusing goto's.
2017-11-29 12:15:22 +01:00
EEVEE_TextureList *txl = vedata->txl;
Object *ob;
EEVEE_LightProbesInfo *pinfo = sldata->probes;
Eevee: Save and reset matrixstate for probe rendering.
2018-03-07 23:25:49 +01:00
DRWMatrixState saved_mats;
Eevee: Put all constant uniforms in a global UBO. This is an optimization / cleanup commit. The use of a global ubo remove lots of uniform lookups and only transfert data when needed. Lots of renaming for more consistent codestyle.
2018-01-21 17:25:10 +01:00
if (pinfo->num_planar == 0) {
Eevee: Render: Fix black reflections in 1st sample.
2018-02-03 02:33:35 +01:00
/* Disable SSR if we cannot read previous frame */
common_data->ssr_toggle = vedata->stl->g_data->valid_double_buffer;
Eevee: Probes: Fix last planar reflections remaining after deletion.
2018-03-08 18:02:42 +01:00
common_data->prb_num_planar = 0;
Eevee: Put all constant uniforms in a global UBO. This is an optimization / cleanup commit. The use of a global ubo remove lots of uniform lookups and only transfert data when needed. Lots of renaming for more consistent codestyle.
2018-01-21 17:25:10 +01:00
return;
}
Eevee: Planar Probe: Add supersampling jitter. This also fix a bug with the probe debug display when there was more than 2 probes. ped->probe_id was equal to 0 for all planar probes until the next frame. Resulting in all planar data debug to show probe 0.
2018-03-08 01:35:16 +01:00
EEVEE_planar_reflections_updates(sldata);
DRW_uniformbuffer_update(sldata->planar_ubo, &sldata->probes->planar_data);
Eevee: Save and reset matrixstate for probe rendering.
2018-03-07 23:25:49 +01:00
/* We need to save the Matrices before overidding them */
DRW_viewport_matrix_get_all(&saved_mats);
Eevee: Put all constant uniforms in a global UBO. This is an optimization / cleanup commit. The use of a global ubo remove lots of uniform lookups and only transfert data when needed. Lots of renaming for more consistent codestyle.
2018-01-21 17:25:10 +01:00
/* Temporary Remove all planar reflections (avoid lag effect). */
common_data->prb_num_planar = 0;
/* Turn off ssr to avoid black specular */
common_data->ssr_toggle = false;
common_data->sss_toggle = false;
DRW_uniformbuffer_update(sldata->common_ubo, &sldata->common_data);
Eevee: Cleanup, split long probe refresh function There is no reason to have such a long function, it is really easy to break it down into a smaller ones, and call them from where needed. Makes them smaller and easier to follow. Also avoids use of confusing goto's.
2017-11-29 12:15:22 +01:00
for (int i = 0; (ob = pinfo->probes_planar_ref[i]) && (i < MAX_PLANAR); i++) {
EEVEE_LightProbeEngineData *ped = EEVEE_lightprobe_data_ensure(ob);
EEVEE: LightProbe: Implement Visibility Group. Works on every probe type. The function to see is EEVEE_lightprobes_obj_visibility_cb. Set pinfo->vis_data.cached to true makes the computation faster for multiple views using the same group. We could even sort the probes by group for that mater to speed things up even more (only applies to dynamic probes like the planar reflections because other probes are only rendered one at a time).
2018-04-24 12:45:59 +02:00
LightProbe *prb = (LightProbe *)ob->data;
Collections and groups unification OVERVIEW * In 2.7 terminology, all layers and groups are now collection datablocks. * These collections are nestable, linkable, instanceable, overrideable, .. which opens up new ways to set up scenes and link + override data. * Viewport/render visibility and selectability are now a part of the collection and shared across all view layers and linkable. * View layers define which subset of the scene collection hierarchy is excluded for each. For many workflows one view layer can be used, these are more of an advanced feature now. OUTLINER * The outliner now has a "View Layer" display mode instead of "Collections", which can display the collections and/or objects in the view layer. * In this display mode, collections can be excluded with the right click menu. These will then be greyed out and their objects will be excluded. * To view collections not linked to any scene, the "Blender File" display mode can be used, with the new filtering option to just see Colleciton datablocks. * The outliner right click menus for collections and objects were reorganized. * Drag and drop still needs to be improved. Like before, dragging the icon or text gives different results, we'll unify this later. LINKING AND OVERRIDES * Collections can now be linked into the scene without creating an instance, with the link/append operator or from the collections view in the outliner. * Collections can get static overrides with the right click menu in the outliner, but this is rather unreliable and not clearly communicated at the moment. * We still need to improve the make override operator to turn collection instances into collections with overrides directly in the scene. PERFORMANCE * We tried to make performance not worse than before and improve it in some cases. The main thing that's still a bit slower is multiple scenes, we have to change the layer syncing to only updated affected scenes. * Collections keep a list of their parent collections for faster incremental updates in syncing and caching. * View layer bases are now in a object -> base hash to avoid quadratic time lookups internally and in API functions like visible_get(). VERSIONING * Compatibility with 2.7 files should be improved due to the new visibility controls. Of course users may not want to set up their scenes differently now to avoid having separate layers and groups. * Compatibility with 2.8 is mostly there, and was tested on Eevee demo and Hero files. There's a few things which are know to be not quite compatible, like nested layer collections inside groups. * The versioning code for 2.8 files is quite complicated, and isolated behind #ifdef so it can be removed at the end of the release cycle. KNOWN ISSUES * The G-key group operators in the 3D viewport were left mostly as is, they need to be modified still to fit better. * Same for the groups panel in the object properties. This needs to be updated still, or perhaps replaced by something better. * Collections must all have a unique name. Less restrictive namespacing is to be done later, we'll have to see how important this is as all objects within the collections must also have a unique name anyway. * Full scene copy and delete scene are exactly doing the right thing yet. Differential Revision: https://developer.blender.org/D3383 https://code.blender.org/2018/05/collections-and-groups/
2018-04-30 15:57:22 +02:00
pinfo->vis_data.collection = prb->visibility_grp;
EEVEE: LightProbe: Implement Visibility Group. Works on every probe type. The function to see is EEVEE_lightprobes_obj_visibility_cb. Set pinfo->vis_data.cached to true makes the computation faster for multiple views using the same group. We could even sort the probes by group for that mater to speed things up even more (only applies to dynamic probes like the planar reflections because other probes are only rendered one at a time).
2018-04-24 12:45:59 +02:00
pinfo->vis_data.invert = prb->flag & LIGHTPROBE_FLAG_INVERT_GROUP;
Eevee: Save and reset matrixstate for probe rendering.
2018-03-07 23:25:49 +01:00
render_scene_to_planar(sldata, vedata, i, ped);
Eevee: Cleanup, split long probe refresh function There is no reason to have such a long function, it is really easy to break it down into a smaller ones, and call them from where needed. Makes them smaller and easier to follow. Also avoids use of confusing goto's.
2017-11-29 12:15:22 +01:00
}
Eevee: Put all constant uniforms in a global UBO. This is an optimization / cleanup commit. The use of a global ubo remove lots of uniform lookups and only transfert data when needed. Lots of renaming for more consistent codestyle.
2018-01-21 17:25:10 +01:00
/* Restore */
common_data->prb_num_planar = pinfo->num_planar;
common_data->ssr_toggle = true;
common_data->sss_toggle = true;
Eevee: Planar: Add transparent objects. The ordering is left broken for the viewport as this can be quite heavy. Only do it if doing a render.
2018-03-08 01:36:49 +01:00
/* Prefilter for SSR */
if ((vedata->stl->effects->enabled_effects & EFFECT_SSR) != 0) {
Eevee: Cleanup, split long probe refresh function There is no reason to have such a long function, it is really easy to break it down into a smaller ones, and call them from where needed. Makes them smaller and easier to follow. Also avoids use of confusing goto's.
2017-11-29 12:15:22 +01:00
const int max_lod = 9;
DRW_stats_group_start("Planar Probe Downsample");
GPUFramebuffer: Refactor (Part 2) This refactor modernise the use of framebuffers. It also touches a lot of files so breaking down changes we have: - GPUTexture: Allow textures to be attached to more than one GPUFrameBuffer. This allows to create and configure more FBO without the need to attach and detach texture at drawing time. - GPUFrameBuffer: The wrapper starts to mimic opengl a bit closer. This allows to configure the framebuffer inside a context other than the one that will be rendering the framebuffer. We do the actual configuration when binding the FBO. We also Keep track of config validity and save drawbuffers state in the FBO. We remove the different bind/unbind functions. These make little sense now that we have separate contexts. - DRWFrameBuffer: We replace DRW_framebuffer functions by GPU_framebuffer ones to avoid another layer of abstraction. We move the DRW convenience functions to GPUFramebuffer instead and even add new ones. The MACRO GPU_framebuffer_ensure_config is pretty much all you need to create and config a GPUFramebuffer. - DRWTexture: Due to the removal of DRWFrameBuffer, we needed to create functions to create textures for thoses framebuffers. Pool textures are now using default texture parameters for the texture type asked. - DRWManager: Make sure no framebuffer object is bound when doing cache filling. - GPUViewport: Add new color_only_fb and depth_only_fb along with FB API usage update. This let draw engines render to color/depth only target and without the need to attach/detach textures. - WM_window: Assert when a framebuffer is bound when changing context. This balance the fact we are not track ogl context inside GPUFramebuffer. - Eevee, Clay, Mode engines: Update to new API. This comes with a lot of code simplification. This also come with some cleanups in some engine codes.
2018-03-25 17:46:48 +02:00
GPU_framebuffer_ensure_config(&fbl->planar_downsample_fb, {
GPU_ATTACHMENT_NONE,
GPU_ATTACHMENT_TEXTURE(txl->planar_pool)
});
GPU_framebuffer_recursive_downsample(fbl->planar_downsample_fb, max_lod, &downsample_planar, vedata);
Eevee: Cleanup, split long probe refresh function There is no reason to have such a long function, it is really easy to break it down into a smaller ones, and call them from where needed. Makes them smaller and easier to follow. Also avoids use of confusing goto's.
2017-11-29 12:15:22 +01:00
/* For shading, save max level of the planar map */
Eevee: Put all constant uniforms in a global UBO. This is an optimization / cleanup commit. The use of a global ubo remove lots of uniform lookups and only transfert data when needed. Lots of renaming for more consistent codestyle.
2018-01-21 17:25:10 +01:00
common_data->prb_lod_planar_max = (float)(max_lod);
Eevee: Cleanup, split long probe refresh function There is no reason to have such a long function, it is really easy to break it down into a smaller ones, and call them from where needed. Makes them smaller and easier to follow. Also avoids use of confusing goto's.
2017-11-29 12:15:22 +01:00
DRW_stats_group_end();
}
Eevee: Render: Make sure all probes are refreshed before rendering.
2018-02-03 01:50:38 +01:00
Eevee: Save and reset matrixstate for probe rendering.
2018-03-07 23:25:49 +01:00
DRW_viewport_matrix_override_set_all(&saved_mats);
Eevee: Planar: Add transparent objects. The ordering is left broken for the viewport as this can be quite heavy. Only do it if doing a render.
2018-03-08 01:36:49 +01:00
if (DRW_state_is_image_render()) {
/* Sort transparents because planar reflections could have re-sorted them. */
DRW_pass_sort_shgroup_z(vedata->psl->transparent_pass);
}
Eevee: Render: Make sure all probes are refreshed before rendering.
2018-02-03 01:50:38 +01:00
/* Disable SSR if we cannot read previous frame */
common_data->ssr_toggle = vedata->stl->g_data->valid_double_buffer;
Eevee: Cleanup, split long probe refresh function There is no reason to have such a long function, it is really easy to break it down into a smaller ones, and call them from where needed. Makes them smaller and easier to follow. Also avoids use of confusing goto's.
2017-11-29 12:15:22 +01:00
}
Eevee: Probe Grid: Clear Grid buffers with world diffuse coefs. This make sure the values displayed by the "show data" sphere are initialized. Also this make the bounce lighting progress more apparent.
2017-09-28 21:17:57 +02:00
Eevee: Cleanup, split long probe refresh function There is no reason to have such a long function, it is really easy to break it down into a smaller ones, and call them from where needed. Makes them smaller and easier to follow. Also avoids use of confusing goto's.
2017-11-29 12:15:22 +01:00
static void lightprobes_refresh_cube(EEVEE_ViewLayerData *sldata, EEVEE_Data *vedata)
{
Eevee: Put all constant uniforms in a global UBO. This is an optimization / cleanup commit. The use of a global ubo remove lots of uniform lookups and only transfert data when needed. Lots of renaming for more consistent codestyle.
2018-01-21 17:25:10 +01:00
EEVEE_CommonUniformBuffer *common_data = &sldata->common_data;
Eevee: Cleanup, split long probe refresh function There is no reason to have such a long function, it is really easy to break it down into a smaller ones, and call them from where needed. Makes them smaller and easier to follow. Also avoids use of confusing goto's.
2017-11-29 12:15:22 +01:00
EEVEE_PassList *psl = vedata->psl;
EEVEE_StorageList *stl = vedata->stl;
EEVEE_LightProbesInfo *pinfo = sldata->probes;
Object *ob;
for (int i = 1; (ob = pinfo->probes_cube_ref[i]) && (i < MAX_PROBE); i++) {
EEVEE_LightProbeEngineData *ped = EEVEE_lightprobe_data_ensure(ob);
if (!ped->need_update) {
continue;
Eevee : Probe Grid : Fix Update tagging.
2017-10-02 18:40:51 +02:00
}
Eevee: Cleanup, split long probe refresh function There is no reason to have such a long function, it is really easy to break it down into a smaller ones, and call them from where needed. Makes them smaller and easier to follow. Also avoids use of confusing goto's.
2017-11-29 12:15:22 +01:00
LightProbe *prb = (LightProbe *)ob->data;
Collections and groups unification OVERVIEW * In 2.7 terminology, all layers and groups are now collection datablocks. * These collections are nestable, linkable, instanceable, overrideable, .. which opens up new ways to set up scenes and link + override data. * Viewport/render visibility and selectability are now a part of the collection and shared across all view layers and linkable. * View layers define which subset of the scene collection hierarchy is excluded for each. For many workflows one view layer can be used, these are more of an advanced feature now. OUTLINER * The outliner now has a "View Layer" display mode instead of "Collections", which can display the collections and/or objects in the view layer. * In this display mode, collections can be excluded with the right click menu. These will then be greyed out and their objects will be excluded. * To view collections not linked to any scene, the "Blender File" display mode can be used, with the new filtering option to just see Colleciton datablocks. * The outliner right click menus for collections and objects were reorganized. * Drag and drop still needs to be improved. Like before, dragging the icon or text gives different results, we'll unify this later. LINKING AND OVERRIDES * Collections can now be linked into the scene without creating an instance, with the link/append operator or from the collections view in the outliner. * Collections can get static overrides with the right click menu in the outliner, but this is rather unreliable and not clearly communicated at the moment. * We still need to improve the make override operator to turn collection instances into collections with overrides directly in the scene. PERFORMANCE * We tried to make performance not worse than before and improve it in some cases. The main thing that's still a bit slower is multiple scenes, we have to change the layer syncing to only updated affected scenes. * Collections keep a list of their parent collections for faster incremental updates in syncing and caching. * View layer bases are now in a object -> base hash to avoid quadratic time lookups internally and in API functions like visible_get(). VERSIONING * Compatibility with 2.7 files should be improved due to the new visibility controls. Of course users may not want to set up their scenes differently now to avoid having separate layers and groups. * Compatibility with 2.8 is mostly there, and was tested on Eevee demo and Hero files. There's a few things which are know to be not quite compatible, like nested layer collections inside groups. * The versioning code for 2.8 files is quite complicated, and isolated behind #ifdef so it can be removed at the end of the release cycle. KNOWN ISSUES * The G-key group operators in the 3D viewport were left mostly as is, they need to be modified still to fit better. * Same for the groups panel in the object properties. This needs to be updated still, or perhaps replaced by something better. * Collections must all have a unique name. Less restrictive namespacing is to be done later, we'll have to see how important this is as all objects within the collections must also have a unique name anyway. * Full scene copy and delete scene are exactly doing the right thing yet. Differential Revision: https://developer.blender.org/D3383 https://code.blender.org/2018/05/collections-and-groups/
2018-04-30 15:57:22 +02:00
pinfo->vis_data.collection = prb->visibility_grp;
EEVEE: LightProbe: Implement Visibility Group. Works on every probe type. The function to see is EEVEE_lightprobes_obj_visibility_cb. Set pinfo->vis_data.cached to true makes the computation faster for multiple views using the same group. We could even sort the probes by group for that mater to speed things up even more (only applies to dynamic probes like the planar reflections because other probes are only rendered one at a time).
2018-04-24 12:45:59 +02:00
pinfo->vis_data.invert = prb->flag & LIGHTPROBE_FLAG_INVERT_GROUP;
Eevee: Cleanup, split long probe refresh function There is no reason to have such a long function, it is really easy to break it down into a smaller ones, and call them from where needed. Makes them smaller and easier to follow. Also avoids use of confusing goto's.
2017-11-29 12:15:22 +01:00
render_scene_to_probe(sldata, vedata, ob->obmat[3], prb->clipsta, prb->clipend);
Eevee: Probes: Add support for intensity tweak. This works for grid and cubemaps. The intensity is baked into the map itself. Thus you need to refresh/rebake the probe to see the changes.
2018-01-21 23:15:57 +01:00
glossy_filter_probe(sldata, vedata, psl, i, prb->intensity);
Eevee: Cleanup, split long probe refresh function There is no reason to have such a long function, it is really easy to break it down into a smaller ones, and call them from where needed. Makes them smaller and easier to follow. Also avoids use of confusing goto's.
2017-11-29 12:15:22 +01:00
ped->need_update = false;
ped->probe_id = i;
if (!ped->ready_to_shade) {
Eevee: Put all constant uniforms in a global UBO. This is an optimization / cleanup commit. The use of a global ubo remove lots of uniform lookups and only transfert data when needed. Lots of renaming for more consistent codestyle.
2018-01-21 17:25:10 +01:00
common_data->prb_num_render_cube++;
Eevee: Cleanup, split long probe refresh function There is no reason to have such a long function, it is really easy to break it down into a smaller ones, and call them from where needed. Makes them smaller and easier to follow. Also avoids use of confusing goto's.
2017-11-29 12:15:22 +01:00
ped->ready_to_shade = true;
Eevee: First commit of Probe support.
2017-06-08 21:48:50 +02:00
}
Eevee: Cleanup, split long probe refresh function There is no reason to have such a long function, it is really easy to break it down into a smaller ones, and call them from where needed. Makes them smaller and easier to follow. Also avoids use of confusing goto's.
2017-11-29 12:15:22 +01:00
#if 0
printf("Update Cubemap %d\n", i);
#endif
Eevee: First commit of Probe support.
2017-06-08 21:48:50 +02:00
DRW_viewport_request_redraw();
Eevee: Cleanup, split long probe refresh function There is no reason to have such a long function, it is really easy to break it down into a smaller ones, and call them from where needed. Makes them smaller and easier to follow. Also avoids use of confusing goto's.
2017-11-29 12:15:22 +01:00
/* Do not let this frame accumulate. */
stl->effects->taa_current_sample = 1;
/* Only do one probe per frame */
return;
}
Eevee: Render: Make sure all probes are refreshed before rendering.
2018-02-03 01:50:38 +01:00
pinfo->do_cube_update = false;
Eevee: Cleanup, split long probe refresh function There is no reason to have such a long function, it is really easy to break it down into a smaller ones, and call them from where needed. Makes them smaller and easier to follow. Also avoids use of confusing goto's.
2017-11-29 12:15:22 +01:00
}
Eevee: Probe Grid: Clear Grid buffers with world diffuse coefs. This make sure the values displayed by the "show data" sphere are initialized. Also this make the bounce lighting progress more apparent.
2017-09-28 21:17:57 +02:00
Eevee: Cleanup, split long probe refresh function There is no reason to have such a long function, it is really easy to break it down into a smaller ones, and call them from where needed. Makes them smaller and easier to follow. Also avoids use of confusing goto's.
2017-11-29 12:15:22 +01:00
static void lightprobes_refresh_all_no_world(EEVEE_ViewLayerData *sldata, EEVEE_Data *vedata)
{
Eevee: Put all constant uniforms in a global UBO. This is an optimization / cleanup commit. The use of a global ubo remove lots of uniform lookups and only transfert data when needed. Lots of renaming for more consistent codestyle.
2018-01-21 17:25:10 +01:00
EEVEE_CommonUniformBuffer *common_data = &sldata->common_data;
Eevee: Cleanup, split long probe refresh function There is no reason to have such a long function, it is really easy to break it down into a smaller ones, and call them from where needed. Makes them smaller and easier to follow. Also avoids use of confusing goto's.
2017-11-29 12:15:22 +01:00
EEVEE_PassList *psl = vedata->psl;
EEVEE_StorageList *stl = vedata->stl;
EEVEE_LightProbesInfo *pinfo = sldata->probes;
Object *ob;
const DRWContextState *draw_ctx = DRW_context_state_get();
RegionView3D *rv3d = draw_ctx->rv3d;
Eevee: Probe Grid: Clear Grid buffers with world diffuse coefs. This make sure the values displayed by the "show data" sphere are initialized. Also this make the bounce lighting progress more apparent.
2017-09-28 21:17:57 +02:00
Eevee: Cleanup, split long probe refresh function There is no reason to have such a long function, it is really easy to break it down into a smaller ones, and call them from where needed. Makes them smaller and easier to follow. Also avoids use of confusing goto's.
2017-11-29 12:15:22 +01:00
if (draw_ctx->evil_C != NULL) {
/* Only compute probes if not navigating or in playback */
struct wmWindowManager *wm = CTX_wm_manager(draw_ctx->evil_C);
if (((rv3d->rflag & RV3D_NAVIGATING) != 0) || ED_screen_animation_no_scrub(wm) != NULL) {
return;
}
}
Eevee: Save and reset matrixstate for probe rendering.
2018-03-07 23:25:49 +01:00
/* We need to save the Matrices before overidding them */
DRWMatrixState saved_mats;
DRW_viewport_matrix_get_all(&saved_mats);
Eevee: Cleanup, split long probe refresh function There is no reason to have such a long function, it is really easy to break it down into a smaller ones, and call them from where needed. Makes them smaller and easier to follow. Also avoids use of confusing goto's.
2017-11-29 12:15:22 +01:00
/* Make sure grid is initialized. */
lightprobes_refresh_initialize_grid(sldata, vedata);
/* Reflection probes depend on diffuse lighting thus on irradiance grid,
* so update them first. */
while (pinfo->updated_bounce < pinfo->num_bounce) {
Eevee: Put all constant uniforms in a global UBO. This is an optimization / cleanup commit. The use of a global ubo remove lots of uniform lookups and only transfert data when needed. Lots of renaming for more consistent codestyle.
2018-01-21 17:25:10 +01:00
common_data->prb_num_render_grid = pinfo->num_grid;
Eevee: Cleanup, avoid unneeded goto Ideally this also should be split into smaller functions, but will leave that for later a bit.
2017-11-29 12:19:57 +01:00
/* TODO(sergey): This logic can be split into smaller functions. */
Eevee: Cleanup, split long probe refresh function There is no reason to have such a long function, it is really easy to break it down into a smaller ones, and call them from where needed. Makes them smaller and easier to follow. Also avoids use of confusing goto's.
2017-11-29 12:15:22 +01:00
for (int i = 1; (ob = pinfo->probes_grid_ref[i]) && (i < MAX_GRID); i++) {
EEVEE_LightProbeEngineData *ped = EEVEE_lightprobe_data_ensure(ob);
if (!ped->need_update) {
continue;
}
EEVEE_LightGrid *egrid = &pinfo->grid_data[i];
LightProbe *prb = (LightProbe *)ob->data;
/* Find the next cell corresponding to the current level. */
bool valid_cell = false;
int cell_id = ped->updated_cells;
float pos[3], grid_loc[3];
/* Other levels */
int current_stride = 1 << max_ii(0, ped->max_lvl - ped->updated_lvl);
int prev_stride = current_stride << 1;
Eevee: Cleanup, avoid unneeded goto Ideally this also should be split into smaller functions, but will leave that for later a bit.
2017-11-29 12:19:57 +01:00
bool do_rendering = true;
Eevee: Cleanup, split long probe refresh function There is no reason to have such a long function, it is really easy to break it down into a smaller ones, and call them from where needed. Makes them smaller and easier to follow. Also avoids use of confusing goto's.
2017-11-29 12:15:22 +01:00
while (!valid_cell) {
cell_id = ped->updated_cells;
lightprobe_cell_grid_location_get(egrid, cell_id, grid_loc);
if (ped->updated_lvl == 0 && cell_id == 0) {
valid_cell = true;
ped->updated_cells = ped->num_cell;
Eevee: Cleanup, avoid too much nested indentation Just do early `continue`, makes code easier to follow than fewzillion of nested checks and loops and everything.
2017-11-29 11:59:15 +01:00
continue;
}
Eevee: Cleanup, split long probe refresh function There is no reason to have such a long function, it is really easy to break it down into a smaller ones, and call them from where needed. Makes them smaller and easier to follow. Also avoids use of confusing goto's.
2017-11-29 12:15:22 +01:00
else if (((((int)grid_loc[0] % current_stride) == 0) &&
(((int)grid_loc[1] % current_stride) == 0) &&
(((int)grid_loc[2] % current_stride) == 0)) &&
!((((int)grid_loc[0] % prev_stride) == 0) &&
(((int)grid_loc[1] % prev_stride) == 0) &&
(((int)grid_loc[2] % prev_stride) == 0)))
{
valid_cell = true;
Eevee: Cleanup, avoid too much nested indentation Just do early `continue`, makes code easier to follow than fewzillion of nested checks and loops and everything.
2017-11-29 11:59:15 +01:00
}
Eevee: Cleanup, split long probe refresh function There is no reason to have such a long function, it is really easy to break it down into a smaller ones, and call them from where needed. Makes them smaller and easier to follow. Also avoids use of confusing goto's.
2017-11-29 12:15:22 +01:00
ped->updated_cells++;
if (ped->updated_cells > ped->num_cell) {
Eevee: Cleanup, avoid unneeded goto Ideally this also should be split into smaller functions, but will leave that for later a bit.
2017-11-29 12:19:57 +01:00
do_rendering = false;
break;
Eevee: Cleanup, avoid too much nested indentation Just do early `continue`, makes code easier to follow than fewzillion of nested checks and loops and everything.
2017-11-29 11:59:15 +01:00
}
Eevee: Irradiance grid: support for non-blocking update and multiple bounces.
2017-06-14 18:52:53 +02:00
}
Eevee: Cleanup, avoid unneeded goto Ideally this also should be split into smaller functions, but will leave that for later a bit.
2017-11-29 12:19:57 +01:00
if (do_rendering) {
lightprobe_cell_world_location_get(egrid, grid_loc, pos);
SWAP(GPUTexture *, sldata->irradiance_pool, sldata->irradiance_rt);
/* Temporary Remove all probes. */
Eevee: Put all constant uniforms in a global UBO. This is an optimization / cleanup commit. The use of a global ubo remove lots of uniform lookups and only transfert data when needed. Lots of renaming for more consistent codestyle.
2018-01-21 17:25:10 +01:00
int tmp_num_render_grid = common_data->prb_num_render_grid;
int tmp_num_render_cube = common_data->prb_num_render_cube;
int tmp_num_planar = common_data->prb_num_planar;
Eevee: Irradiance Grid: Fix Multibounce baking. The baking was only recording one bounce because the double buffer setup was buggy.
2017-12-02 00:47:42 +01:00
float tmp_level_bias = egrid->level_bias;
Eevee: Put all constant uniforms in a global UBO. This is an optimization / cleanup commit. The use of a global ubo remove lots of uniform lookups and only transfert data when needed. Lots of renaming for more consistent codestyle.
2018-01-21 17:25:10 +01:00
common_data->prb_num_render_cube = 0;
common_data->prb_num_planar = 0;
Eevee: Cleanup, avoid unneeded goto Ideally this also should be split into smaller functions, but will leave that for later a bit.
2017-11-29 12:19:57 +01:00
/* Use light from previous bounce when capturing radiance. */
if (pinfo->updated_bounce == 0) {
Eevee: Irradiance Grid: Fix Multibounce baking. The baking was only recording one bounce because the double buffer setup was buggy.
2017-12-02 00:47:42 +01:00
/* But not on first bounce. */
Eevee: Put all constant uniforms in a global UBO. This is an optimization / cleanup commit. The use of a global ubo remove lots of uniform lookups and only transfert data when needed. Lots of renaming for more consistent codestyle.
2018-01-21 17:25:10 +01:00
common_data->prb_num_render_grid = 0;
Eevee: Cleanup, avoid unneeded goto Ideally this also should be split into smaller functions, but will leave that for later a bit.
2017-11-29 12:19:57 +01:00
}
Eevee: Irradiance Grid: Fix Multibounce baking. The baking was only recording one bounce because the double buffer setup was buggy.
2017-12-02 00:47:42 +01:00
else {
/* Remove bias */
egrid->level_bias = (float)(1 << 0);
DRW_uniformbuffer_update(sldata->grid_ubo, &sldata->probes->grid_data);
}
Collections and groups unification OVERVIEW * In 2.7 terminology, all layers and groups are now collection datablocks. * These collections are nestable, linkable, instanceable, overrideable, .. which opens up new ways to set up scenes and link + override data. * Viewport/render visibility and selectability are now a part of the collection and shared across all view layers and linkable. * View layers define which subset of the scene collection hierarchy is excluded for each. For many workflows one view layer can be used, these are more of an advanced feature now. OUTLINER * The outliner now has a "View Layer" display mode instead of "Collections", which can display the collections and/or objects in the view layer. * In this display mode, collections can be excluded with the right click menu. These will then be greyed out and their objects will be excluded. * To view collections not linked to any scene, the "Blender File" display mode can be used, with the new filtering option to just see Colleciton datablocks. * The outliner right click menus for collections and objects were reorganized. * Drag and drop still needs to be improved. Like before, dragging the icon or text gives different results, we'll unify this later. LINKING AND OVERRIDES * Collections can now be linked into the scene without creating an instance, with the link/append operator or from the collections view in the outliner. * Collections can get static overrides with the right click menu in the outliner, but this is rather unreliable and not clearly communicated at the moment. * We still need to improve the make override operator to turn collection instances into collections with overrides directly in the scene. PERFORMANCE * We tried to make performance not worse than before and improve it in some cases. The main thing that's still a bit slower is multiple scenes, we have to change the layer syncing to only updated affected scenes. * Collections keep a list of their parent collections for faster incremental updates in syncing and caching. * View layer bases are now in a object -> base hash to avoid quadratic time lookups internally and in API functions like visible_get(). VERSIONING * Compatibility with 2.7 files should be improved due to the new visibility controls. Of course users may not want to set up their scenes differently now to avoid having separate layers and groups. * Compatibility with 2.8 is mostly there, and was tested on Eevee demo and Hero files. There's a few things which are know to be not quite compatible, like nested layer collections inside groups. * The versioning code for 2.8 files is quite complicated, and isolated behind #ifdef so it can be removed at the end of the release cycle. KNOWN ISSUES * The G-key group operators in the 3D viewport were left mostly as is, they need to be modified still to fit better. * Same for the groups panel in the object properties. This needs to be updated still, or perhaps replaced by something better. * Collections must all have a unique name. Less restrictive namespacing is to be done later, we'll have to see how important this is as all objects within the collections must also have a unique name anyway. * Full scene copy and delete scene are exactly doing the right thing yet. Differential Revision: https://developer.blender.org/D3383 https://code.blender.org/2018/05/collections-and-groups/
2018-04-30 15:57:22 +02:00
pinfo->vis_data.collection = prb->visibility_grp;
EEVEE: LightProbe: Implement Visibility Group. Works on every probe type. The function to see is EEVEE_lightprobes_obj_visibility_cb. Set pinfo->vis_data.cached to true makes the computation faster for multiple views using the same group. We could even sort the probes by group for that mater to speed things up even more (only applies to dynamic probes like the planar reflections because other probes are only rendered one at a time).
2018-04-24 12:45:59 +02:00
pinfo->vis_data.invert = prb->flag & LIGHTPROBE_FLAG_INVERT_GROUP;
Eevee: Cleanup, avoid unneeded goto Ideally this also should be split into smaller functions, but will leave that for later a bit.
2017-11-29 12:19:57 +01:00
render_scene_to_probe(sldata, vedata, pos, prb->clipsta, prb->clipend);
Eevee: Irradiance Visibility: Initial Implementation This augment the existing irradiance grid with a new visibility precomputation. We store a small shadowmap for each grid sample so that light does not leak through walls and such. The visibility parameter are similar to the one used by the Variance Shadow Map for point lights. Technical details: We store the visibility in the same texture (array) as the irradiance itself (in order to reduce the number of sampler). But the irradiance and the visibility are not the same data so we must encode them in order to use the same texture format. We use RGBA8 normalized texture and encode irradiance as RGBE (shared exponent). Using RGBE encoding instead of R11_G11_B10 may lead to some lighting changes, but quality seems to be nearly the same in my test cases. Using full RGBA16/32F maybe a future option but that will require much more memory and reduce the perf significantly. Visibility moments (VSM) are encoded as 16bits fixed point precision using a special range. This seems to retain enough precision for the needs. Also interpolation does not seems to be big problem (even though it's incorrect).
2017-12-02 14:28:29 +01:00
diffuse_filter_probe(sldata, vedata, psl, egrid->offset + cell_id,
Eevee: Probes: Add support for intensity tweak. This works for grid and cubemaps. The intensity is baked into the map itself. Thus you need to refresh/rebake the probe to see the changes.
2018-01-21 23:15:57 +01:00
prb->clipsta, prb->clipend, egrid->visibility_range, prb->vis_blur,
prb->intensity);
Eevee: Cleanup, avoid unneeded goto Ideally this also should be split into smaller functions, but will leave that for later a bit.
2017-11-29 12:19:57 +01:00
/* To see what is going on. */
SWAP(GPUTexture *, sldata->irradiance_pool, sldata->irradiance_rt);
/* Restore */
Eevee: Put all constant uniforms in a global UBO. This is an optimization / cleanup commit. The use of a global ubo remove lots of uniform lookups and only transfert data when needed. Lots of renaming for more consistent codestyle.
2018-01-21 17:25:10 +01:00
common_data->prb_num_render_cube = tmp_num_render_cube;
Eevee: Cleanup, avoid unneeded goto Ideally this also should be split into smaller functions, but will leave that for later a bit.
2017-11-29 12:19:57 +01:00
pinfo->num_planar = tmp_num_planar;
Eevee: Irradiance Grid: Fix Multibounce baking. The baking was only recording one bounce because the double buffer setup was buggy.
2017-12-02 00:47:42 +01:00
if (pinfo->updated_bounce == 0) {
Eevee: Put all constant uniforms in a global UBO. This is an optimization / cleanup commit. The use of a global ubo remove lots of uniform lookups and only transfert data when needed. Lots of renaming for more consistent codestyle.
2018-01-21 17:25:10 +01:00
common_data->prb_num_render_grid = tmp_num_render_grid;
Eevee: Irradiance Grid: Fix Multibounce baking. The baking was only recording one bounce because the double buffer setup was buggy.
2017-12-02 00:47:42 +01:00
}
else {
egrid->level_bias = tmp_level_bias;
DRW_uniformbuffer_update(sldata->grid_ubo, &sldata->probes->grid_data);
}
#if 0
printf("Updated Grid %d : cell %d / %d, bounce %d / %d\n",
i, cell_id + 1, ped->num_cell, pinfo->updated_bounce + 1, pinfo->num_bounce);
#endif
Eevee: Add Irradiance Grid support Early implementation. Slow and still has quality 3 ways of storing irradiance: - Spherical Harmonics: Have problem with directionnal lighting. - HL2 diffuse cube: Very low resolution but smooth transitions. - Diffuse cube: High storage requirement. Also include some name change.
2017-06-13 17:39:39 +02:00
}
Eevee: Cleanup, split long probe refresh function There is no reason to have such a long function, it is really easy to break it down into a smaller ones, and call them from where needed. Makes them smaller and easier to follow. Also avoids use of confusing goto's.
2017-11-29 12:15:22 +01:00
if (ped->updated_cells >= ped->num_cell) {
ped->updated_lvl++;
ped->updated_cells = 0;
if (ped->updated_lvl > ped->max_lvl) {
ped->need_update = false;
}
egrid->level_bias = (float)(1 << max_ii(0, ped->max_lvl - ped->updated_lvl + 1));
DRW_uniformbuffer_update(sldata->grid_ubo, &sldata->probes->grid_data);
Eevee: Cleanup, avoid too much nested indentation Just do early `continue`, makes code easier to follow than fewzillion of nested checks and loops and everything.
2017-11-29 11:59:15 +01:00
}
Eevee: Cleanup, split long probe refresh function There is no reason to have such a long function, it is really easy to break it down into a smaller ones, and call them from where needed. Makes them smaller and easier to follow. Also avoids use of confusing goto's.
2017-11-29 12:15:22 +01:00
/* Only do one probe per frame */
Eevee: Cleanup, avoid too much nested indentation Just do early `continue`, makes code easier to follow than fewzillion of nested checks and loops and everything.
2017-11-29 11:59:15 +01:00
DRW_viewport_request_redraw();
/* Do not let this frame accumulate. */
stl->effects->taa_current_sample = 1;
Eevee: Fix temporal sampling not resetting. World Probe update was not resetting the TAA. This also reset the volume sampling in cases that needs it (world/grid probe diffuse coefs update).
2018-03-13 01:35:16 +01:00
/* Reset volume history. */
stl->effects->volume_current_sample = -1;
common_data->vol_history_alpha = 0.0f;
Eevee: Save and reset matrixstate for probe rendering.
2018-03-07 23:25:49 +01:00
/* Restore matrices */
DRW_viewport_matrix_override_set_all(&saved_mats);
Eevee: Cleanup, split long probe refresh function There is no reason to have such a long function, it is really easy to break it down into a smaller ones, and call them from where needed. Makes them smaller and easier to follow. Also avoids use of confusing goto's.
2017-11-29 12:15:22 +01:00
return;
Eevee: First commit of Probe support.
2017-06-08 21:48:50 +02:00
}
Eevee: Initial implementation of planar reflections. Still pretty barebone: No roughness support, No normal distortion support.
2017-06-17 00:08:03 +02:00
Eevee: Cleanup, split long probe refresh function There is no reason to have such a long function, it is really easy to break it down into a smaller ones, and call them from where needed. Makes them smaller and easier to follow. Also avoids use of confusing goto's.
2017-11-29 12:15:22 +01:00
pinfo->updated_bounce++;
Eevee: Put all constant uniforms in a global UBO. This is an optimization / cleanup commit. The use of a global ubo remove lots of uniform lookups and only transfert data when needed. Lots of renaming for more consistent codestyle.
2018-01-21 17:25:10 +01:00
common_data->prb_num_render_grid = pinfo->num_grid;
Eevee: Initial implementation of planar reflections. Still pretty barebone: No roughness support, No normal distortion support.
2017-06-17 00:08:03 +02:00
Eevee: Cleanup, split long probe refresh function There is no reason to have such a long function, it is really easy to break it down into a smaller ones, and call them from where needed. Makes them smaller and easier to follow. Also avoids use of confusing goto's.
2017-11-29 12:15:22 +01:00
if (pinfo->updated_bounce < pinfo->num_bounce) {
/* Retag all grids to update for next bounce */
for (int i = 1; (ob = pinfo->probes_grid_ref[i]) && (i < MAX_GRID); i++) {
EEVEE_LightProbeEngineData *ped = EEVEE_lightprobe_data_ensure(ob);
ped->need_update = true;
ped->updated_cells = 0;
ped->updated_lvl = 0;
}
/* Reset the next buffer so we can see the progress. */
Eevee: Irradiance Grid: Fix Multibounce baking. The baking was only recording one bounce because the double buffer setup was buggy.
2017-12-02 00:47:42 +01:00
/* irradiance_rt is already the next rt because of the previous SWAP */
GPUFramebuffer: Refactor (Part 2) This refactor modernise the use of framebuffers. It also touches a lot of files so breaking down changes we have: - GPUTexture: Allow textures to be attached to more than one GPUFrameBuffer. This allows to create and configure more FBO without the need to attach and detach texture at drawing time. - GPUFrameBuffer: The wrapper starts to mimic opengl a bit closer. This allows to configure the framebuffer inside a context other than the one that will be rendering the framebuffer. We do the actual configuration when binding the FBO. We also Keep track of config validity and save drawbuffers state in the FBO. We remove the different bind/unbind functions. These make little sense now that we have separate contexts. - DRWFrameBuffer: We replace DRW_framebuffer functions by GPU_framebuffer ones to avoid another layer of abstraction. We move the DRW convenience functions to GPUFramebuffer instead and even add new ones. The MACRO GPU_framebuffer_ensure_config is pretty much all you need to create and config a GPUFramebuffer. - DRWTexture: Due to the removal of DRWFrameBuffer, we needed to create functions to create textures for thoses framebuffers. Pool textures are now using default texture parameters for the texture type asked. - DRWManager: Make sure no framebuffer object is bound when doing cache filling. - GPUViewport: Add new color_only_fb and depth_only_fb along with FB API usage update. This let draw engines render to color/depth only target and without the need to attach/detach textures. - WM_window: Assert when a framebuffer is bound when changing context. This balance the fact we are not track ogl context inside GPUFramebuffer. - Eevee, Clay, Mode engines: Update to new API. This comes with a lot of code simplification. This also come with some cleanups in some engine codes.
2018-03-25 17:46:48 +02:00
GPU_framebuffer_texture_attach(sldata->probe_filter_fb, sldata->irradiance_rt, 0, 0);
GPU_framebuffer_bind(sldata->probe_filter_fb);
Eevee: Cleanup, split long probe refresh function There is no reason to have such a long function, it is really easy to break it down into a smaller ones, and call them from where needed. Makes them smaller and easier to follow. Also avoids use of confusing goto's.
2017-11-29 12:15:22 +01:00
DRW_draw_pass(psl->probe_grid_fill);
GPUFramebuffer: Refactor (Part 2) This refactor modernise the use of framebuffers. It also touches a lot of files so breaking down changes we have: - GPUTexture: Allow textures to be attached to more than one GPUFrameBuffer. This allows to create and configure more FBO without the need to attach and detach texture at drawing time. - GPUFrameBuffer: The wrapper starts to mimic opengl a bit closer. This allows to configure the framebuffer inside a context other than the one that will be rendering the framebuffer. We do the actual configuration when binding the FBO. We also Keep track of config validity and save drawbuffers state in the FBO. We remove the different bind/unbind functions. These make little sense now that we have separate contexts. - DRWFrameBuffer: We replace DRW_framebuffer functions by GPU_framebuffer ones to avoid another layer of abstraction. We move the DRW convenience functions to GPUFramebuffer instead and even add new ones. The MACRO GPU_framebuffer_ensure_config is pretty much all you need to create and config a GPUFramebuffer. - DRWTexture: Due to the removal of DRWFrameBuffer, we needed to create functions to create textures for thoses framebuffers. Pool textures are now using default texture parameters for the texture type asked. - DRWManager: Make sure no framebuffer object is bound when doing cache filling. - GPUViewport: Add new color_only_fb and depth_only_fb along with FB API usage update. This let draw engines render to color/depth only target and without the need to attach/detach textures. - WM_window: Assert when a framebuffer is bound when changing context. This balance the fact we are not track ogl context inside GPUFramebuffer. - Eevee, Clay, Mode engines: Update to new API. This comes with a lot of code simplification. This also come with some cleanups in some engine codes.
2018-03-25 17:46:48 +02:00
GPU_framebuffer_texture_attach(sldata->probe_filter_fb, sldata->probe_pool, 0, 0);
Eevee: Irradiance Grid: Fix Multibounce baking. The baking was only recording one bounce because the double buffer setup was buggy.
2017-12-02 00:47:42 +01:00
/* Swap AFTER */
SWAP(GPUTexture *, sldata->irradiance_pool, sldata->irradiance_rt);
Eevee: Cleanup, split long probe refresh function There is no reason to have such a long function, it is really easy to break it down into a smaller ones, and call them from where needed. Makes them smaller and easier to follow. Also avoids use of confusing goto's.
2017-11-29 12:15:22 +01:00
}
Eevee: Initial implementation of planar reflections. Still pretty barebone: No roughness support, No normal distortion support.
2017-06-17 00:08:03 +02:00
}
Eevee: Cleanup, split long probe refresh function There is no reason to have such a long function, it is really easy to break it down into a smaller ones, and call them from where needed. Makes them smaller and easier to follow. Also avoids use of confusing goto's.
2017-11-29 12:15:22 +01:00
/* Refresh cube probe when needed. */
lightprobes_refresh_cube(sldata, vedata);
Eevee: Save and reset matrixstate for probe rendering.
2018-03-07 23:25:49 +01:00
/* Restore matrices */
DRW_viewport_matrix_override_set_all(&saved_mats);
Eevee: Cleanup, split long probe refresh function There is no reason to have such a long function, it is really easy to break it down into a smaller ones, and call them from where needed. Makes them smaller and easier to follow. Also avoids use of confusing goto's.
2017-11-29 12:15:22 +01:00
}
Eevee: Add Planar reflection blurring. This method is very cheap and inaccurate. This will fill the gap untill better model is supported.
2017-06-26 20:37:41 +02:00
Eevee: Render: Make sure all probes are refreshed before rendering.
2018-02-03 01:50:38 +01:00
bool EEVEE_lightprobes_all_probes_ready(EEVEE_ViewLayerData *sldata, EEVEE_Data *UNUSED(vedata))
Eevee: Cleanup, split long probe refresh function There is no reason to have such a long function, it is really easy to break it down into a smaller ones, and call them from where needed. Makes them smaller and easier to follow. Also avoids use of confusing goto's.
2017-11-29 12:15:22 +01:00
{
Eevee: Put all constant uniforms in a global UBO. This is an optimization / cleanup commit. The use of a global ubo remove lots of uniform lookups and only transfert data when needed. Lots of renaming for more consistent codestyle.
2018-01-21 17:25:10 +01:00
EEVEE_LightProbesInfo *pinfo = sldata->probes;
EEVEE_CommonUniformBuffer *common_data = &sldata->common_data;
Eevee: Render: Make sure all probes are refreshed before rendering.
2018-02-03 01:50:38 +01:00
return ((pinfo->do_cube_update == false) &&
(pinfo->updated_bounce == pinfo->num_bounce) &&
(common_data->prb_num_render_cube == pinfo->num_cube));
}
void EEVEE_lightprobes_refresh(EEVEE_ViewLayerData *sldata, EEVEE_Data *vedata)
{
EEVEE_CommonUniformBuffer *common_data = &sldata->common_data;
Eevee: Update to support shader deferred compilation. World probe is tagged to refresh when it it's shader is updated. Probes are recomputed only after all meshes shaders have been compiled.
2018-03-06 02:19:28 +01:00
EEVEE_LightProbesInfo *pinfo = sldata->probes;
Eevee: Render: Make sure all probes are refreshed before rendering.
2018-02-03 01:50:38 +01:00
Eevee: Put all constant uniforms in a global UBO. This is an optimization / cleanup commit. The use of a global ubo remove lots of uniform lookups and only transfert data when needed. Lots of renaming for more consistent codestyle.
2018-01-21 17:25:10 +01:00
/* Disable specular lighting when rendering probes to avoid feedback loops (looks bad). */
common_data->spec_toggle = false;
common_data->ssr_toggle = false;
common_data->sss_toggle = false;
/* Disable AO until we find a way to hide really bad discontinuities between cubefaces. */
float tmp_ao_dist = common_data->ao_dist;
float tmp_ao_settings = common_data->ao_settings;
common_data->ao_settings = 0.0f;
common_data->ao_dist = 0.0f;
Eevee: Cleanup, split long probe refresh function There is no reason to have such a long function, it is really easy to break it down into a smaller ones, and call them from where needed. Makes them smaller and easier to follow. Also avoids use of confusing goto's.
2017-11-29 12:15:22 +01:00
/* Render world in priority */
Eevee: Update to support shader deferred compilation. World probe is tagged to refresh when it it's shader is updated. Probes are recomputed only after all meshes shaders have been compiled.
2018-03-06 02:19:28 +01:00
if (pinfo->update_world) {
Eevee: Cleanup, split long probe refresh function There is no reason to have such a long function, it is really easy to break it down into a smaller ones, and call them from where needed. Makes them smaller and easier to follow. Also avoids use of confusing goto's.
2017-11-29 12:15:22 +01:00
lightprobes_refresh_world(sldata, vedata);
}
Eevee: Update to support shader deferred compilation. World probe is tagged to refresh when it it's shader is updated. Probes are recomputed only after all meshes shaders have been compiled.
2018-03-06 02:19:28 +01:00
else if (EEVEE_lightprobes_all_probes_ready(sldata, vedata) == false && pinfo->all_materials_updated) {
Eevee: Cleanup, split long probe refresh function There is no reason to have such a long function, it is really easy to break it down into a smaller ones, and call them from where needed. Makes them smaller and easier to follow. Also avoids use of confusing goto's.
2017-11-29 12:15:22 +01:00
lightprobes_refresh_all_no_world(sldata, vedata);
Eevee: Add Planar reflection blurring. This method is very cheap and inaccurate. This will fill the gap untill better model is supported.
2017-06-26 20:37:41 +02:00
}
Eevee: SSR: Fix ssr nor working in ogl render.
2018-01-03 20:41:14 +01:00
Eevee: Put all constant uniforms in a global UBO. This is an optimization / cleanup commit. The use of a global ubo remove lots of uniform lookups and only transfert data when needed. Lots of renaming for more consistent codestyle.
2018-01-21 17:25:10 +01:00
/* Restore */
common_data->spec_toggle = true;
common_data->ssr_toggle = true;
common_data->sss_toggle = true;
common_data->ao_dist = tmp_ao_dist;
common_data->ao_settings = tmp_ao_settings;
Eevee: Material code refactor. Separate material handling inside another file. Make use of enums to identify shader variations. Group all 64*64 LUTs into one array texture. Only update world probe if world changes.
2017-06-04 12:12:58 +02:00
}
Rename probe to light-probe Probe is a real general term, the new name is used often in docs online.
2017-06-12 20:59:54 +10:00
void EEVEE_lightprobes_free(void)
Eevee: Replace Cubemaps by octahedron maps for env. probes. This enables us to use 2D texture arrays for multiple probes. There is a little artifact with very high roughness caused elongated pixel due to the projection (along every 90° meridian).
2017-05-29 22:03:57 +02:00
{
DRW: Refactor: Less feature duplication with Gwn. This removes the need of custom attribs for instancing. Instancing works fully with dynamic batches & Gwn_VertFormat now. This is in prevision of the VAO manager patch.
2018-02-14 18:59:15 +01:00
MEM_SAFE_FREE(e_data.format_probe_display_cube);
MEM_SAFE_FREE(e_data.format_probe_display_planar);
Eevee: Material code refactor. Separate material handling inside another file. Make use of enums to identify shader variations. Group all 64*64 LUTs into one array texture. Only update world probe if world changes.
2017-06-04 12:12:58 +02:00
DRW_SHADER_FREE_SAFE(e_data.probe_default_sh);
EEvEE: LookDev
2018-05-25 08:06:36 +02:00
DRW_SHADER_FREE_SAFE(e_data.probe_default_studiolight_sh);
Eevee: Add Irradiance Grid support Early implementation. Slow and still has quality 3 ways of storing irradiance: - Spherical Harmonics: Have problem with directionnal lighting. - HL2 diffuse cube: Very low resolution but smooth transitions. - Diffuse cube: High storage requirement. Also include some name change.
2017-06-13 17:39:39 +02:00
DRW_SHADER_FREE_SAFE(e_data.probe_filter_glossy_sh);
DRW_SHADER_FREE_SAFE(e_data.probe_filter_diffuse_sh);
Eevee: Irradiance Visibility: Initial Implementation This augment the existing irradiance grid with a new visibility precomputation. We store a small shadowmap for each grid sample so that light does not leak through walls and such. The visibility parameter are similar to the one used by the Variance Shadow Map for point lights. Technical details: We store the visibility in the same texture (array) as the irradiance itself (in order to reduce the number of sampler). But the irradiance and the visibility are not the same data so we must encode them in order to use the same texture format. We use RGBA8 normalized texture and encode irradiance as RGBE (shared exponent). Using RGBE encoding instead of R11_G11_B10 may lead to some lighting changes, but quality seems to be nearly the same in my test cases. Using full RGBA16/32F maybe a future option but that will require much more memory and reduce the perf significantly. Visibility moments (VSM) are encoded as 16bits fixed point precision using a special range. This seems to retain enough precision for the needs. Also interpolation does not seems to be big problem (even though it's incorrect).
2017-12-02 14:28:29 +01:00
DRW_SHADER_FREE_SAFE(e_data.probe_filter_visibility_sh);
Eevee: Probe Grid: Clear Grid buffers with world diffuse coefs. This make sure the values displayed by the "show data" sphere are initialized. Also this make the bounce lighting progress more apparent.
2017-09-28 21:17:57 +02:00
DRW_SHADER_FREE_SAFE(e_data.probe_grid_fill_sh);
Eevee: Add Grid debug display.
2017-06-14 13:45:54 +02:00
DRW_SHADER_FREE_SAFE(e_data.probe_grid_display_sh);
Eevee: Add data display for planar reflection. Maybe not very useful but it's here for feature parity with the other probes. The fragment shader clipping is here because I'm lazy and don't want to creating a proper quad Batch.
2017-06-17 01:59:01 +02:00
DRW_SHADER_FREE_SAFE(e_data.probe_planar_display_sh);
Eevee: Add Planar reflection blurring. This method is very cheap and inaccurate. This will fill the gap untill better model is supported.
2017-06-26 20:37:41 +02:00
DRW_SHADER_FREE_SAFE(e_data.probe_planar_downsample_sh);
Eevee: Probes: Add data display for cubemaps.
2017-06-15 00:10:34 +02:00
DRW_SHADER_FREE_SAFE(e_data.probe_cube_display_sh);
Eevee: Replace Cubemaps by octahedron maps for env. probes. This enables us to use 2D texture arrays for multiple probes. There is a little artifact with very high roughness caused elongated pixel due to the projection (along every 90° meridian).
2017-05-29 22:03:57 +02:00
DRW_TEXTURE_FREE_SAFE(e_data.hammersley);
Eevee: Planar reflection: Fix bad texture binding.
2017-06-23 19:02:40 +02:00
DRW_TEXTURE_FREE_SAFE(e_data.planar_pool_placeholder);
Eevee: HiZ buffer: Split into two 24bit depth buffer This way we don't have float precision issue we had before and we save some bandwidth.
2017-07-21 14:27:16 +02:00
DRW_TEXTURE_FREE_SAFE(e_data.depth_placeholder);
Eevee: Planar Reflection: Remove distance approximation. This commit separate the depth texture into another texture array. This remove the need to output radial depth into alpha. Unfortunatly it's difficult to recover position from the non linear depth buffer when applying reflection without adding a bunch of stuff. This is in preparation of SSR planar reflections.
2017-07-23 14:03:27 +02:00
DRW_TEXTURE_FREE_SAFE(e_data.depth_array_placeholder);
Eevee: Put shadows and probes inside SceneLayerEngineData This remove the duplication of data for each viewport improving memory usage.
2017-05-30 22:29:20 +02:00
}
Copy Permalink