godotengine
/
godot
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
godot/servers/rendering/renderer_rd/renderer_scene_render_rd.cpp

8947 lines
342 KiB
C++
Raw Normal View History

Fix code formatting issues and VS compilation Also temporarily disable multicheck build so that we get a full build even when there are style issues on Vulkan. Fixes #33356.
2019-11-05 11:01:00 +00:00
/*************************************************************************/
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
/* renderer_scene_render_rd.cpp */
Fix code formatting issues and VS compilation Also temporarily disable multicheck build so that we get a full build even when there are style issues on Vulkan. Fixes #33356.
2019-11-05 11:01:00 +00:00
/*************************************************************************/
/* This file is part of: */
/* GODOT ENGINE */
/* https://godotengine.org */
/*************************************************************************/
Update copyright statements to 2021 Happy new year to the wonderful Godot community! 2020 has been a tough year for most of us personally, but a good year for Godot development nonetheless with a huge amount of work done towards Godot 4.0 and great improvements backported to the long-lived 3.2 branch. We've had close to 400 contributors to engine code this year, authoring near 7,000 commit! (And that's only for the `master` branch and for the engine code, there's a lot more when counting docs, demos and other first-party repos.) Here's to a great year 2021 for all Godot users 🎆
2021-01-01 19:13:46 +00:00
/* Copyright (c) 2007-2021 Juan Linietsky, Ariel Manzur. */
/* Copyright (c) 2014-2021 Godot Engine contributors (cf. AUTHORS.md). */
Fix code formatting issues and VS compilation Also temporarily disable multicheck build so that we get a full build even when there are style issues on Vulkan. Fixes #33356.
2019-11-05 11:01:00 +00:00
/* */
/* Permission is hereby granted, free of charge, to any person obtaining */
/* a copy of this software and associated documentation files (the */
/* "Software"), to deal in the Software without restriction, including */
/* without limitation the rights to use, copy, modify, merge, publish, */
/* distribute, sublicense, and/or sell copies of the Software, and to */
/* permit persons to whom the Software is furnished to do so, subject to */
/* the following conditions: */
/* */
/* The above copyright notice and this permission notice shall be */
/* included in all copies or substantial portions of the Software. */
/* */
/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */
/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */
/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/
/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */
/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */
/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */
/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */
/*************************************************************************/
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
#include "renderer_scene_render_rd.h"
doc: Sync classref with current source Lots of internal API changes and some docstrings were lost in the conversion. I manually salvaged many of them but for all the rendering-related ones, an additional pass is needed. Added missing enum bindings in BaseMaterial3D and VisualServer.
2020-02-12 08:59:06 +00:00
Reorganized core/ directory, it was too fatty already -Removed FuncRef, since Callable makes it obsolete -Removed int_types.h as its obsolete in c++11+ -Changed color names code
2020-11-07 22:33:38 +00:00
#include "core/config/project_settings.h"
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
#include "core/os/os.h"
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
#include "renderer_compositor_rd.h"
Reorganize rendering server. -Made RenderingServerScene abstract, allowing reimplementation -RenderingServerRaster -> RenderingServerDefault, but this class is going away soon.
2020-12-03 21:09:47 +00:00
#include "servers/rendering/rendering_server_default.h"
Environment: Refactor code for readability + more - Makes all boolean setters/getters consistent. - Fixes bug where `glow_hdr_bleed_scale` was not used. - Split CameraEffects to their own source file. - Reorder all Environment method and properties declarations, definitions and bindings to be consistent with each other and with the order of property bindings. - Bind missing enum values added with SDFGI. - Remove unused SDFGI enhance_ssr boolean. - Sync doc changes after SDFGI merge and other misc changes.
2020-07-01 12:18:13 +00:00
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
uint64_t RendererSceneRenderRD::auto_exposure_counter = 2;
Environment sky more or less working.
2019-08-26 20:43:58 +00:00
Add proper quality settings to soft shadows
2020-04-10 09:30:36 +00:00
void get_vogel_disk(float *r_kernel, int p_sample_count) {
const float golden_angle = 2.4;
for (int i = 0; i < p_sample_count; i++) {
float r = Math::sqrt(float(i) + 0.5) / Math::sqrt(float(p_sample_count));
float theta = float(i) * golden_angle;
r_kernel[i * 4] = Math::cos(theta) * r;
r_kernel[i * 4 + 1] = Math::sin(theta) * r;
}
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
void RendererSceneRenderRD::_clear_reflection_data(ReflectionData &rd) {
Reflection probes working
2019-09-09 20:50:51 +00:00
rd.layers.clear();
rd.radiance_base_cubemap = RID();
Basic fast filtering implementation
2020-02-20 23:27:34 +00:00
if (rd.downsampled_radiance_cubemap.is_valid()) {
RD::get_singleton()->free(rd.downsampled_radiance_cubemap);
}
rd.downsampled_radiance_cubemap = RID();
rd.downsampled_layer.mipmaps.clear();
rd.coefficient_buffer = RID();
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
void RendererSceneRenderRD::_update_reflection_data(ReflectionData &rd, int p_size, int p_mipmaps, bool p_use_array, RID p_base_cube, int p_base_layer, bool p_low_quality) {
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
//recreate radiance and all data
Reflection probes working
2019-09-09 20:50:51 +00:00
int mipmaps = p_mipmaps;
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
uint32_t w = p_size, h = p_size;
Reflection probes working
2019-09-09 20:50:51 +00:00
if (p_use_array) {
Working sky shader implementation
2020-03-20 00:32:19 +00:00
int layers = p_low_quality ? 8 : roughness_layers;
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
Basic fast filtering implementation
2020-02-20 23:27:34 +00:00
for (int i = 0; i < layers; i++) {
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
ReflectionData::Layer layer;
uint32_t mmw = w;
uint32_t mmh = h;
layer.mipmaps.resize(mipmaps);
Basic fast filtering implementation
2020-02-20 23:27:34 +00:00
layer.views.resize(mipmaps);
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
for (int j = 0; j < mipmaps; j++) {
ReflectionData::Layer::Mipmap &mm = layer.mipmaps.write[j];
mm.size.width = mmw;
mm.size.height = mmh;
for (int k = 0; k < 6; k++) {
Reflection probes working
2019-09-09 20:50:51 +00:00
mm.views[k] = RD::get_singleton()->texture_create_shared_from_slice(RD::TextureView(), p_base_cube, p_base_layer + i * 6 + k, j);
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
Vector<RID> fbtex;
fbtex.push_back(mm.views[k]);
mm.framebuffers[k] = RD::get_singleton()->framebuffer_create(fbtex);
}
Basic fast filtering implementation
2020-02-20 23:27:34 +00:00
layer.views.write[j] = RD::get_singleton()->texture_create_shared_from_slice(RD::TextureView(), p_base_cube, p_base_layer + i * 6, j, RD::TEXTURE_SLICE_CUBEMAP);
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
mmw = MAX(1, mmw >> 1);
mmh = MAX(1, mmh >> 1);
}
rd.layers.push_back(layer);
}
} else {
Working sky shader implementation
2020-03-20 00:32:19 +00:00
mipmaps = p_low_quality ? 8 : mipmaps;
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
//regular cubemap, lower quality (aliasing, less memory)
ReflectionData::Layer layer;
uint32_t mmw = w;
uint32_t mmh = h;
Basic fast filtering implementation
2020-02-20 23:27:34 +00:00
layer.mipmaps.resize(mipmaps);
layer.views.resize(mipmaps);
for (int j = 0; j < mipmaps; j++) {
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
ReflectionData::Layer::Mipmap &mm = layer.mipmaps.write[j];
mm.size.width = mmw;
mm.size.height = mmh;
for (int k = 0; k < 6; k++) {
Reflection probes working
2019-09-09 20:50:51 +00:00
mm.views[k] = RD::get_singleton()->texture_create_shared_from_slice(RD::TextureView(), p_base_cube, p_base_layer + k, j);
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
Vector<RID> fbtex;
fbtex.push_back(mm.views[k]);
mm.framebuffers[k] = RD::get_singleton()->framebuffer_create(fbtex);
}
Basic fast filtering implementation
2020-02-20 23:27:34 +00:00
layer.views.write[j] = RD::get_singleton()->texture_create_shared_from_slice(RD::TextureView(), p_base_cube, p_base_layer, j, RD::TEXTURE_SLICE_CUBEMAP);
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
mmw = MAX(1, mmw >> 1);
mmh = MAX(1, mmh >> 1);
}
rd.layers.push_back(layer);
}
Reflection probes working
2019-09-09 20:50:51 +00:00
rd.radiance_base_cubemap = RD::get_singleton()->texture_create_shared_from_slice(RD::TextureView(), p_base_cube, p_base_layer, 0, RD::TEXTURE_SLICE_CUBEMAP);
Basic fast filtering implementation
2020-02-20 23:27:34 +00:00
RD::TextureFormat tf;
tf.format = RD::DATA_FORMAT_R16G16B16A16_SFLOAT;
tf.width = 64; // Always 64x64
tf.height = 64;
Rename RD texture "type" to "texture_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-11-27 03:50:05 +00:00
tf.texture_type = RD::TEXTURE_TYPE_CUBE;
Basic fast filtering implementation
2020-02-20 23:27:34 +00:00
tf.array_layers = 6;
tf.mipmaps = 7;
tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_STORAGE_BIT | RD::TEXTURE_USAGE_COLOR_ATTACHMENT_BIT;
rd.downsampled_radiance_cubemap = RD::get_singleton()->texture_create(tf, RD::TextureView());
{
uint32_t mmw = 64;
uint32_t mmh = 64;
rd.downsampled_layer.mipmaps.resize(7);
for (int j = 0; j < rd.downsampled_layer.mipmaps.size(); j++) {
ReflectionData::DownsampleLayer::Mipmap &mm = rd.downsampled_layer.mipmaps.write[j];
mm.size.width = mmw;
mm.size.height = mmh;
mm.view = RD::get_singleton()->texture_create_shared_from_slice(RD::TextureView(), rd.downsampled_radiance_cubemap, 0, j, RD::TEXTURE_SLICE_CUBEMAP);
mmw = MAX(1, mmw >> 1);
mmh = MAX(1, mmh >> 1);
}
}
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
void RendererSceneRenderRD::_create_reflection_fast_filter(ReflectionData &rd, bool p_use_arrays) {
Working sky shader implementation
2020-03-20 00:32:19 +00:00
storage->get_effects()->cubemap_downsample(rd.radiance_base_cubemap, rd.downsampled_layer.mipmaps[0].view, rd.downsampled_layer.mipmaps[0].size);
Basic fast filtering implementation
2020-02-20 23:27:34 +00:00
Working sky shader implementation
2020-03-20 00:32:19 +00:00
for (int i = 1; i < rd.downsampled_layer.mipmaps.size(); i++) {
storage->get_effects()->cubemap_downsample(rd.downsampled_layer.mipmaps[i - 1].view, rd.downsampled_layer.mipmaps[i].view, rd.downsampled_layer.mipmaps[i].size);
}
Basic fast filtering implementation
2020-02-20 23:27:34 +00:00
Working sky shader implementation
2020-03-20 00:32:19 +00:00
Vector<RID> views;
if (p_use_arrays) {
for (int i = 1; i < rd.layers.size(); i++) {
views.push_back(rd.layers[i].views[0]);
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
}
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
} else {
for (int i = 1; i < rd.layers[0].views.size(); i++) {
views.push_back(rd.layers[0].views[i]);
}
}
storage->get_effects()->cubemap_filter(rd.downsampled_radiance_cubemap, views, p_use_arrays);
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
void RendererSceneRenderRD::_create_reflection_importance_sample(ReflectionData &rd, bool p_use_arrays, int p_cube_side, int p_base_layer) {
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
if (p_use_arrays) {
//render directly to the layers
storage->get_effects()->cubemap_roughness(rd.radiance_base_cubemap, rd.layers[p_base_layer].views[0], p_cube_side, sky_ggx_samples_quality, float(p_base_layer) / (rd.layers.size() - 1.0), rd.layers[p_base_layer].mipmaps[0].size.x);
} else {
storage->get_effects()->cubemap_roughness(rd.layers[0].views[p_base_layer - 1], rd.layers[0].views[p_base_layer], p_cube_side, sky_ggx_samples_quality, float(p_base_layer) / (rd.layers[0].mipmaps.size() - 1.0), rd.layers[0].mipmaps[p_base_layer].size.x);
}
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
void RendererSceneRenderRD::_update_reflection_mipmaps(ReflectionData &rd, int p_start, int p_end) {
Add incremental update mode to sky
2020-07-11 00:07:30 +00:00
for (int i = p_start; i < p_end; i++) {
Use cubemap downsampler for reflection mipmaps
2020-12-25 00:39:17 +00:00
for (int j = 0; j < rd.layers[i].views.size() - 1; j++) {
RID view = rd.layers[i].views[j];
RID texture = rd.layers[i].views[j + 1];
Size2i size = rd.layers[i].mipmaps[j + 1].size;
storage->get_effects()->cubemap_downsample(view, texture, size);
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
}
}
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
void RendererSceneRenderRD::_sdfgi_erase(RenderBuffers *rb) {
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
for (uint32_t i = 0; i < rb->sdfgi->cascades.size(); i++) {
const SDFGI::Cascade &c = rb->sdfgi->cascades[i];
RD::get_singleton()->free(c.light_data);
RD::get_singleton()->free(c.light_aniso_0_tex);
RD::get_singleton()->free(c.light_aniso_1_tex);
RD::get_singleton()->free(c.sdf_tex);
RD::get_singleton()->free(c.solid_cell_dispatch_buffer);
RD::get_singleton()->free(c.solid_cell_buffer);
RD::get_singleton()->free(c.lightprobe_history_tex);
RD::get_singleton()->free(c.lightprobe_average_tex);
RD::get_singleton()->free(c.lights_buffer);
}
RD::get_singleton()->free(rb->sdfgi->render_albedo);
RD::get_singleton()->free(rb->sdfgi->render_emission);
RD::get_singleton()->free(rb->sdfgi->render_emission_aniso);
RD::get_singleton()->free(rb->sdfgi->render_sdf[0]);
RD::get_singleton()->free(rb->sdfgi->render_sdf[1]);
RD::get_singleton()->free(rb->sdfgi->render_sdf_half[0]);
RD::get_singleton()->free(rb->sdfgi->render_sdf_half[1]);
for (int i = 0; i < 8; i++) {
RD::get_singleton()->free(rb->sdfgi->render_occlusion[i]);
}
RD::get_singleton()->free(rb->sdfgi->render_geom_facing);
RD::get_singleton()->free(rb->sdfgi->lightprobe_data);
RD::get_singleton()->free(rb->sdfgi->lightprobe_history_scroll);
RD::get_singleton()->free(rb->sdfgi->occlusion_data);
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
RD::get_singleton()->free(rb->sdfgi->ambient_texture);
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
RD::get_singleton()->free(rb->sdfgi->cascades_ubo);
memdelete(rb->sdfgi);
rb->sdfgi = nullptr;
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
const Vector3i RendererSceneRenderRD::SDFGI::Cascade::DIRTY_ALL = Vector3i(0x7FFFFFFF, 0x7FFFFFFF, 0x7FFFFFFF);
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
void RendererSceneRenderRD::sdfgi_update(RID p_render_buffers, RID p_environment, const Vector3 &p_world_position) {
Fix spelling of a var, a struct, and message output
2020-07-17 16:02:06 +00:00
Environment *env = environment_owner.getornull(p_environment);
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers);
bool needs_sdfgi = env && env->sdfgi_enabled;
if (!needs_sdfgi) {
if (rb->sdfgi != nullptr) {
//erase it
_sdfgi_erase(rb);
_render_buffers_uniform_set_changed(p_render_buffers);
}
return;
}
static const uint32_t history_frames_to_converge[RS::ENV_SDFGI_CONVERGE_MAX] = { 5, 10, 15, 20, 25, 30 };
uint32_t requested_history_size = history_frames_to_converge[sdfgi_frames_to_converge];
if (rb->sdfgi && (rb->sdfgi->cascade_mode != env->sdfgi_cascades || rb->sdfgi->min_cell_size != env->sdfgi_min_cell_size || requested_history_size != rb->sdfgi->history_size || rb->sdfgi->uses_occlusion != env->sdfgi_use_occlusion || rb->sdfgi->y_scale_mode != env->sdfgi_y_scale)) {
//configuration changed, erase
_sdfgi_erase(rb);
}
SDFGI *sdfgi = rb->sdfgi;
if (sdfgi == nullptr) {
//re-create
rb->sdfgi = memnew(SDFGI);
sdfgi = rb->sdfgi;
sdfgi->cascade_mode = env->sdfgi_cascades;
sdfgi->min_cell_size = env->sdfgi_min_cell_size;
sdfgi->uses_occlusion = env->sdfgi_use_occlusion;
sdfgi->y_scale_mode = env->sdfgi_y_scale;
static const float y_scale[3] = { 1.0, 1.5, 2.0 };
sdfgi->y_mult = y_scale[sdfgi->y_scale_mode];
static const int cascasde_size[3] = { 4, 6, 8 };
sdfgi->cascades.resize(cascasde_size[sdfgi->cascade_mode]);
sdfgi->probe_axis_count = SDFGI::PROBE_DIVISOR + 1;
sdfgi->solid_cell_ratio = sdfgi_solid_cell_ratio;
sdfgi->solid_cell_count = uint32_t(float(sdfgi->cascade_size * sdfgi->cascade_size * sdfgi->cascade_size) * sdfgi->solid_cell_ratio);
float base_cell_size = sdfgi->min_cell_size;
RD::TextureFormat tf_sdf;
tf_sdf.format = RD::DATA_FORMAT_R8_UNORM;
tf_sdf.width = sdfgi->cascade_size; // Always 64x64
tf_sdf.height = sdfgi->cascade_size;
tf_sdf.depth = sdfgi->cascade_size;
Rename RD texture "type" to "texture_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-11-27 03:50:05 +00:00
tf_sdf.texture_type = RD::TEXTURE_TYPE_3D;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
tf_sdf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_STORAGE_BIT | RD::TEXTURE_USAGE_CAN_COPY_TO_BIT | RD::TEXTURE_USAGE_CAN_COPY_FROM_BIT;
{
RD::TextureFormat tf_render = tf_sdf;
tf_render.format = RD::DATA_FORMAT_R16_UINT;
sdfgi->render_albedo = RD::get_singleton()->texture_create(tf_render, RD::TextureView());
tf_render.format = RD::DATA_FORMAT_R32_UINT;
sdfgi->render_emission = RD::get_singleton()->texture_create(tf_render, RD::TextureView());
sdfgi->render_emission_aniso = RD::get_singleton()->texture_create(tf_render, RD::TextureView());
tf_render.format = RD::DATA_FORMAT_R8_UNORM; //at least its easy to visualize
for (int i = 0; i < 8; i++) {
sdfgi->render_occlusion[i] = RD::get_singleton()->texture_create(tf_render, RD::TextureView());
}
tf_render.format = RD::DATA_FORMAT_R32_UINT;
sdfgi->render_geom_facing = RD::get_singleton()->texture_create(tf_render, RD::TextureView());
tf_render.format = RD::DATA_FORMAT_R8G8B8A8_UINT;
sdfgi->render_sdf[0] = RD::get_singleton()->texture_create(tf_render, RD::TextureView());
sdfgi->render_sdf[1] = RD::get_singleton()->texture_create(tf_render, RD::TextureView());
tf_render.width /= 2;
tf_render.height /= 2;
tf_render.depth /= 2;
sdfgi->render_sdf_half[0] = RD::get_singleton()->texture_create(tf_render, RD::TextureView());
sdfgi->render_sdf_half[1] = RD::get_singleton()->texture_create(tf_render, RD::TextureView());
}
RD::TextureFormat tf_occlusion = tf_sdf;
tf_occlusion.format = RD::DATA_FORMAT_R16_UINT;
tf_occlusion.shareable_formats.push_back(RD::DATA_FORMAT_R16_UINT);
tf_occlusion.shareable_formats.push_back(RD::DATA_FORMAT_R4G4B4A4_UNORM_PACK16);
tf_occlusion.depth *= sdfgi->cascades.size(); //use depth for occlusion slices
tf_occlusion.width *= 2; //use width for the other half
RD::TextureFormat tf_light = tf_sdf;
tf_light.format = RD::DATA_FORMAT_R32_UINT;
tf_light.shareable_formats.push_back(RD::DATA_FORMAT_R32_UINT);
tf_light.shareable_formats.push_back(RD::DATA_FORMAT_E5B9G9R9_UFLOAT_PACK32);
RD::TextureFormat tf_aniso0 = tf_sdf;
tf_aniso0.format = RD::DATA_FORMAT_R8G8B8A8_UNORM;
RD::TextureFormat tf_aniso1 = tf_sdf;
tf_aniso1.format = RD::DATA_FORMAT_R8G8_UNORM;
int passes = nearest_shift(sdfgi->cascade_size) - 1;
//store lightprobe SH
RD::TextureFormat tf_probes;
tf_probes.format = RD::DATA_FORMAT_R16G16B16A16_SFLOAT;
tf_probes.width = sdfgi->probe_axis_count * sdfgi->probe_axis_count;
tf_probes.height = sdfgi->probe_axis_count * SDFGI::SH_SIZE;
tf_probes.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_STORAGE_BIT | RD::TEXTURE_USAGE_CAN_COPY_TO_BIT | RD::TEXTURE_USAGE_CAN_COPY_FROM_BIT;
Rename RD texture "type" to "texture_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-11-27 03:50:05 +00:00
tf_probes.texture_type = RD::TEXTURE_TYPE_2D_ARRAY;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
sdfgi->history_size = requested_history_size;
RD::TextureFormat tf_probe_history = tf_probes;
tf_probe_history.format = RD::DATA_FORMAT_R16G16B16A16_SINT; //signed integer because SH are signed
tf_probe_history.array_layers = sdfgi->history_size;
RD::TextureFormat tf_probe_average = tf_probes;
tf_probe_average.format = RD::DATA_FORMAT_R32G32B32A32_SINT; //signed integer because SH are signed
Rename RD texture "type" to "texture_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-11-27 03:50:05 +00:00
tf_probe_average.texture_type = RD::TEXTURE_TYPE_2D;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
sdfgi->lightprobe_history_scroll = RD::get_singleton()->texture_create(tf_probe_history, RD::TextureView());
sdfgi->lightprobe_average_scroll = RD::get_singleton()->texture_create(tf_probe_average, RD::TextureView());
{
//octahedral lightprobes
RD::TextureFormat tf_octprobes = tf_probes;
tf_octprobes.array_layers = sdfgi->cascades.size() * 2;
tf_octprobes.format = RD::DATA_FORMAT_R32_UINT; //pack well with RGBE
tf_octprobes.width = sdfgi->probe_axis_count * sdfgi->probe_axis_count * (SDFGI::LIGHTPROBE_OCT_SIZE + 2);
tf_octprobes.height = sdfgi->probe_axis_count * (SDFGI::LIGHTPROBE_OCT_SIZE + 2);
tf_octprobes.shareable_formats.push_back(RD::DATA_FORMAT_R32_UINT);
tf_octprobes.shareable_formats.push_back(RD::DATA_FORMAT_E5B9G9R9_UFLOAT_PACK32);
//lightprobe texture is an octahedral texture
sdfgi->lightprobe_data = RD::get_singleton()->texture_create(tf_octprobes, RD::TextureView());
RD::TextureView tv;
tv.format_override = RD::DATA_FORMAT_E5B9G9R9_UFLOAT_PACK32;
sdfgi->lightprobe_texture = RD::get_singleton()->texture_create_shared(tv, sdfgi->lightprobe_data);
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
//texture handling ambient data, to integrate with volumetric foc
RD::TextureFormat tf_ambient = tf_probes;
tf_ambient.array_layers = sdfgi->cascades.size();
tf_ambient.format = RD::DATA_FORMAT_R16G16B16A16_SFLOAT; //pack well with RGBE
tf_ambient.width = sdfgi->probe_axis_count * sdfgi->probe_axis_count;
tf_ambient.height = sdfgi->probe_axis_count;
Rename RD texture "type" to "texture_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-11-27 03:50:05 +00:00
tf_ambient.texture_type = RD::TEXTURE_TYPE_2D_ARRAY;
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
//lightprobe texture is an octahedral texture
sdfgi->ambient_texture = RD::get_singleton()->texture_create(tf_ambient, RD::TextureView());
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
}
sdfgi->cascades_ubo = RD::get_singleton()->uniform_buffer_create(sizeof(SDFGI::Cascade::UBO) * SDFGI::MAX_CASCADES);
sdfgi->occlusion_data = RD::get_singleton()->texture_create(tf_occlusion, RD::TextureView());
{
RD::TextureView tv;
tv.format_override = RD::DATA_FORMAT_R4G4B4A4_UNORM_PACK16;
sdfgi->occlusion_texture = RD::get_singleton()->texture_create_shared(tv, sdfgi->occlusion_data);
}
for (uint32_t i = 0; i < sdfgi->cascades.size(); i++) {
SDFGI::Cascade &cascade = sdfgi->cascades[i];
/* 3D Textures */
cascade.sdf_tex = RD::get_singleton()->texture_create(tf_sdf, RD::TextureView());
cascade.light_data = RD::get_singleton()->texture_create(tf_light, RD::TextureView());
cascade.light_aniso_0_tex = RD::get_singleton()->texture_create(tf_aniso0, RD::TextureView());
cascade.light_aniso_1_tex = RD::get_singleton()->texture_create(tf_aniso1, RD::TextureView());
{
RD::TextureView tv;
tv.format_override = RD::DATA_FORMAT_E5B9G9R9_UFLOAT_PACK32;
cascade.light_tex = RD::get_singleton()->texture_create_shared(tv, cascade.light_data);
RD::get_singleton()->texture_clear(cascade.light_tex, Color(0, 0, 0, 0), 0, 1, 0, 1);
RD::get_singleton()->texture_clear(cascade.light_aniso_0_tex, Color(0, 0, 0, 0), 0, 1, 0, 1);
RD::get_singleton()->texture_clear(cascade.light_aniso_1_tex, Color(0, 0, 0, 0), 0, 1, 0, 1);
}
cascade.cell_size = base_cell_size;
Vector3 world_position = p_world_position;
world_position.y *= sdfgi->y_mult;
int32_t probe_cells = sdfgi->cascade_size / SDFGI::PROBE_DIVISOR;
Vector3 probe_size = Vector3(1, 1, 1) * cascade.cell_size * probe_cells;
Vector3i probe_pos = Vector3i((world_position / probe_size + Vector3(0.5, 0.5, 0.5)).floor());
cascade.position = probe_pos * probe_cells;
cascade.dirty_regions = SDFGI::Cascade::DIRTY_ALL;
base_cell_size *= 2.0;
/* Probe History */
cascade.lightprobe_history_tex = RD::get_singleton()->texture_create(tf_probe_history, RD::TextureView());
RD::get_singleton()->texture_clear(cascade.lightprobe_history_tex, Color(0, 0, 0, 0), 0, 1, 0, tf_probe_history.array_layers); //needs to be cleared for average to work
cascade.lightprobe_average_tex = RD::get_singleton()->texture_create(tf_probe_average, RD::TextureView());
RD::get_singleton()->texture_clear(cascade.lightprobe_average_tex, Color(0, 0, 0, 0), 0, 1, 0, 1); //needs to be cleared for average to work
/* Buffers */
cascade.solid_cell_buffer = RD::get_singleton()->storage_buffer_create(sizeof(SDFGI::Cascade::SolidCell) * sdfgi->solid_cell_count);
cascade.solid_cell_dispatch_buffer = RD::get_singleton()->storage_buffer_create(sizeof(uint32_t) * 4, Vector<uint8_t>(), RD::STORAGE_BUFFER_USAGE_DISPATCH_INDIRECT);
cascade.lights_buffer = RD::get_singleton()->storage_buffer_create(sizeof(SDGIShader::Light) * MAX(SDFGI::MAX_STATIC_LIGHTS, SDFGI::MAX_DYNAMIC_LIGHTS));
{
Vector<RD::Uniform> uniforms;
{
RD::Uniform u;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_IMAGE;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
u.binding = 1;
u.ids.push_back(sdfgi->render_sdf[(passes & 1) ? 1 : 0]); //if passes are even, we read from buffer 0, else we read from buffer 1
uniforms.push_back(u);
}
{
RD::Uniform u;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_IMAGE;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
u.binding = 2;
u.ids.push_back(sdfgi->render_albedo);
uniforms.push_back(u);
}
{
RD::Uniform u;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_IMAGE;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
u.binding = 3;
for (int j = 0; j < 8; j++) {
u.ids.push_back(sdfgi->render_occlusion[j]);
}
uniforms.push_back(u);
}
{
RD::Uniform u;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_IMAGE;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
u.binding = 4;
u.ids.push_back(sdfgi->render_emission);
uniforms.push_back(u);
}
{
RD::Uniform u;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_IMAGE;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
u.binding = 5;
u.ids.push_back(sdfgi->render_emission_aniso);
uniforms.push_back(u);
}
{
RD::Uniform u;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_IMAGE;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
u.binding = 6;
u.ids.push_back(sdfgi->render_geom_facing);
uniforms.push_back(u);
}
{
RD::Uniform u;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_IMAGE;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
u.binding = 7;
u.ids.push_back(cascade.sdf_tex);
uniforms.push_back(u);
}
{
RD::Uniform u;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_IMAGE;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
u.binding = 8;
u.ids.push_back(sdfgi->occlusion_data);
uniforms.push_back(u);
}
{
RD::Uniform u;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
u.binding = 10;
u.ids.push_back(cascade.solid_cell_dispatch_buffer);
uniforms.push_back(u);
}
{
RD::Uniform u;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
u.binding = 11;
u.ids.push_back(cascade.solid_cell_buffer);
uniforms.push_back(u);
}
cascade.sdf_store_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, sdfgi_shader.preprocess.version_get_shader(sdfgi_shader.preprocess_shader, SDGIShader::PRE_PROCESS_STORE), 0);
}
{
Vector<RD::Uniform> uniforms;
{
RD::Uniform u;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_IMAGE;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
u.binding = 1;
u.ids.push_back(sdfgi->render_albedo);
uniforms.push_back(u);
}
{
RD::Uniform u;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_IMAGE;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
u.binding = 2;
u.ids.push_back(sdfgi->render_geom_facing);
uniforms.push_back(u);
}
{
RD::Uniform u;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_IMAGE;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
u.binding = 3;
u.ids.push_back(sdfgi->render_emission);
uniforms.push_back(u);
}
{
RD::Uniform u;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_IMAGE;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
u.binding = 4;
u.ids.push_back(sdfgi->render_emission_aniso);
uniforms.push_back(u);
}
{
RD::Uniform u;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
u.binding = 5;
u.ids.push_back(cascade.solid_cell_dispatch_buffer);
uniforms.push_back(u);
}
{
RD::Uniform u;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
u.binding = 6;
u.ids.push_back(cascade.solid_cell_buffer);
uniforms.push_back(u);
}
cascade.scroll_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, sdfgi_shader.preprocess.version_get_shader(sdfgi_shader.preprocess_shader, SDGIShader::PRE_PROCESS_SCROLL), 0);
}
{
Vector<RD::Uniform> uniforms;
{
RD::Uniform u;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_IMAGE;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
u.binding = 1;
for (int j = 0; j < 8; j++) {
u.ids.push_back(sdfgi->render_occlusion[j]);
}
uniforms.push_back(u);
}
{
RD::Uniform u;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_IMAGE;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
u.binding = 2;
u.ids.push_back(sdfgi->occlusion_data);
uniforms.push_back(u);
}
cascade.scroll_occlusion_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, sdfgi_shader.preprocess.version_get_shader(sdfgi_shader.preprocess_shader, SDGIShader::PRE_PROCESS_SCROLL_OCCLUSION), 0);
}
}
//direct light
for (uint32_t i = 0; i < sdfgi->cascades.size(); i++) {
SDFGI::Cascade &cascade = sdfgi->cascades[i];
Vector<RD::Uniform> uniforms;
{
RD::Uniform u;
u.binding = 1;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_TEXTURE;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
for (uint32_t j = 0; j < SDFGI::MAX_CASCADES; j++) {
if (j < rb->sdfgi->cascades.size()) {
u.ids.push_back(rb->sdfgi->cascades[j].sdf_tex);
} else {
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
u.ids.push_back(storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_3D_WHITE));
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
}
}
uniforms.push_back(u);
}
{
RD::Uniform u;
u.binding = 2;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_SAMPLER;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
u.ids.push_back(storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED));
uniforms.push_back(u);
}
{
RD::Uniform u;
u.binding = 3;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
u.ids.push_back(cascade.solid_cell_dispatch_buffer);
uniforms.push_back(u);
}
{
RD::Uniform u;
u.binding = 4;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
u.ids.push_back(cascade.solid_cell_buffer);
uniforms.push_back(u);
}
{
RD::Uniform u;
u.binding = 5;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_IMAGE;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
u.ids.push_back(cascade.light_data);
uniforms.push_back(u);
}
{
RD::Uniform u;
u.binding = 6;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_IMAGE;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
u.ids.push_back(cascade.light_aniso_0_tex);
uniforms.push_back(u);
}
{
RD::Uniform u;
u.binding = 7;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_IMAGE;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
u.ids.push_back(cascade.light_aniso_1_tex);
uniforms.push_back(u);
}
{
RD::Uniform u;
u.binding = 8;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
u.ids.push_back(rb->sdfgi->cascades_ubo);
uniforms.push_back(u);
}
{
RD::Uniform u;
u.binding = 9;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
u.ids.push_back(cascade.lights_buffer);
uniforms.push_back(u);
}
{
RD::Uniform u;
u.binding = 10;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_TEXTURE;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
u.ids.push_back(rb->sdfgi->lightprobe_texture);
uniforms.push_back(u);
}
cascade.sdf_direct_light_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, sdfgi_shader.direct_light.version_get_shader(sdfgi_shader.direct_light_shader, 0), 0);
}
//preprocess initialize uniform set
{
Vector<RD::Uniform> uniforms;
{
RD::Uniform u;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_IMAGE;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
u.binding = 1;
u.ids.push_back(sdfgi->render_albedo);
uniforms.push_back(u);
}
{
RD::Uniform u;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_IMAGE;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
u.binding = 2;
u.ids.push_back(sdfgi->render_sdf[0]);
uniforms.push_back(u);
}
sdfgi->sdf_initialize_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, sdfgi_shader.preprocess.version_get_shader(sdfgi_shader.preprocess_shader, SDGIShader::PRE_PROCESS_JUMP_FLOOD_INITIALIZE), 0);
}
{
Vector<RD::Uniform> uniforms;
{
RD::Uniform u;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_IMAGE;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
u.binding = 1;
u.ids.push_back(sdfgi->render_albedo);
uniforms.push_back(u);
}
{
RD::Uniform u;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_IMAGE;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
u.binding = 2;
u.ids.push_back(sdfgi->render_sdf_half[0]);
uniforms.push_back(u);
}
sdfgi->sdf_initialize_half_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, sdfgi_shader.preprocess.version_get_shader(sdfgi_shader.preprocess_shader, SDGIShader::PRE_PROCESS_JUMP_FLOOD_INITIALIZE_HALF), 0);
}
//jump flood uniform set
{
Vector<RD::Uniform> uniforms;
{
RD::Uniform u;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_IMAGE;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
u.binding = 1;
u.ids.push_back(sdfgi->render_sdf[0]);
uniforms.push_back(u);
}
{
RD::Uniform u;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_IMAGE;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
u.binding = 2;
u.ids.push_back(sdfgi->render_sdf[1]);
uniforms.push_back(u);
}
sdfgi->jump_flood_uniform_set[0] = RD::get_singleton()->uniform_set_create(uniforms, sdfgi_shader.preprocess.version_get_shader(sdfgi_shader.preprocess_shader, SDGIShader::PRE_PROCESS_JUMP_FLOOD), 0);
SWAP(uniforms.write[0].ids.write[0], uniforms.write[1].ids.write[0]);
sdfgi->jump_flood_uniform_set[1] = RD::get_singleton()->uniform_set_create(uniforms, sdfgi_shader.preprocess.version_get_shader(sdfgi_shader.preprocess_shader, SDGIShader::PRE_PROCESS_JUMP_FLOOD), 0);
}
//jump flood half uniform set
{
Vector<RD::Uniform> uniforms;
{
RD::Uniform u;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_IMAGE;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
u.binding = 1;
u.ids.push_back(sdfgi->render_sdf_half[0]);
uniforms.push_back(u);
}
{
RD::Uniform u;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_IMAGE;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
u.binding = 2;
u.ids.push_back(sdfgi->render_sdf_half[1]);
uniforms.push_back(u);
}
sdfgi->jump_flood_half_uniform_set[0] = RD::get_singleton()->uniform_set_create(uniforms, sdfgi_shader.preprocess.version_get_shader(sdfgi_shader.preprocess_shader, SDGIShader::PRE_PROCESS_JUMP_FLOOD), 0);
SWAP(uniforms.write[0].ids.write[0], uniforms.write[1].ids.write[0]);
sdfgi->jump_flood_half_uniform_set[1] = RD::get_singleton()->uniform_set_create(uniforms, sdfgi_shader.preprocess.version_get_shader(sdfgi_shader.preprocess_shader, SDGIShader::PRE_PROCESS_JUMP_FLOOD), 0);
}
//upscale half size sdf
{
Vector<RD::Uniform> uniforms;
{
RD::Uniform u;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_IMAGE;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
u.binding = 1;
u.ids.push_back(sdfgi->render_albedo);
uniforms.push_back(u);
}
{
RD::Uniform u;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_IMAGE;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
u.binding = 2;
u.ids.push_back(sdfgi->render_sdf_half[(passes & 1) ? 0 : 1]); //reverse pass order because half size
uniforms.push_back(u);
}
{
RD::Uniform u;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_IMAGE;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
u.binding = 3;
u.ids.push_back(sdfgi->render_sdf[(passes & 1) ? 0 : 1]); //reverse pass order because it needs an extra JFA pass
uniforms.push_back(u);
}
sdfgi->upscale_jfa_uniform_set_index = (passes & 1) ? 0 : 1;
sdfgi->sdf_upscale_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, sdfgi_shader.preprocess.version_get_shader(sdfgi_shader.preprocess_shader, SDGIShader::PRE_PROCESS_JUMP_FLOOD_UPSCALE), 0);
}
//occlusion uniform set
{
Vector<RD::Uniform> uniforms;
{
RD::Uniform u;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_IMAGE;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
u.binding = 1;
u.ids.push_back(sdfgi->render_albedo);
uniforms.push_back(u);
}
{
RD::Uniform u;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_IMAGE;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
u.binding = 2;
for (int i = 0; i < 8; i++) {
u.ids.push_back(sdfgi->render_occlusion[i]);
}
uniforms.push_back(u);
}
{
RD::Uniform u;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_IMAGE;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
u.binding = 3;
u.ids.push_back(sdfgi->render_geom_facing);
uniforms.push_back(u);
}
sdfgi->occlusion_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, sdfgi_shader.preprocess.version_get_shader(sdfgi_shader.preprocess_shader, SDGIShader::PRE_PROCESS_OCCLUSION), 0);
}
for (uint32_t i = 0; i < sdfgi->cascades.size(); i++) {
//integrate uniform
Vector<RD::Uniform> uniforms;
{
RD::Uniform u;
u.binding = 1;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_TEXTURE;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
for (uint32_t j = 0; j < SDFGI::MAX_CASCADES; j++) {
if (j < sdfgi->cascades.size()) {
u.ids.push_back(sdfgi->cascades[j].sdf_tex);
} else {
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
u.ids.push_back(storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_3D_WHITE));
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
}
}
uniforms.push_back(u);
}
{
RD::Uniform u;
u.binding = 2;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_TEXTURE;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
for (uint32_t j = 0; j < SDFGI::MAX_CASCADES; j++) {
if (j < sdfgi->cascades.size()) {
u.ids.push_back(sdfgi->cascades[j].light_tex);
} else {
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
u.ids.push_back(storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_3D_WHITE));
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
}
}
uniforms.push_back(u);
}
{
RD::Uniform u;
u.binding = 3;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_TEXTURE;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
for (uint32_t j = 0; j < SDFGI::MAX_CASCADES; j++) {
if (j < sdfgi->cascades.size()) {
u.ids.push_back(sdfgi->cascades[j].light_aniso_0_tex);
} else {
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
u.ids.push_back(storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_3D_WHITE));
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
}
}
uniforms.push_back(u);
}
{
RD::Uniform u;
u.binding = 4;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_TEXTURE;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
for (uint32_t j = 0; j < SDFGI::MAX_CASCADES; j++) {
if (j < sdfgi->cascades.size()) {
u.ids.push_back(sdfgi->cascades[j].light_aniso_1_tex);
} else {
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
u.ids.push_back(storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_3D_WHITE));
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
}
}
uniforms.push_back(u);
}
{
RD::Uniform u;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_SAMPLER;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
u.binding = 6;
u.ids.push_back(storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED));
uniforms.push_back(u);
}
{
RD::Uniform u;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
u.binding = 7;
u.ids.push_back(sdfgi->cascades_ubo);
uniforms.push_back(u);
}
{
RD::Uniform u;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_IMAGE;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
u.binding = 8;
u.ids.push_back(sdfgi->lightprobe_data);
uniforms.push_back(u);
}
{
RD::Uniform u;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_IMAGE;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
u.binding = 9;
u.ids.push_back(sdfgi->cascades[i].lightprobe_history_tex);
uniforms.push_back(u);
}
{
RD::Uniform u;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_IMAGE;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
u.binding = 10;
u.ids.push_back(sdfgi->cascades[i].lightprobe_average_tex);
uniforms.push_back(u);
}
{
RD::Uniform u;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_IMAGE;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
u.binding = 11;
u.ids.push_back(sdfgi->lightprobe_history_scroll);
uniforms.push_back(u);
}
{
RD::Uniform u;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_IMAGE;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
u.binding = 12;
u.ids.push_back(sdfgi->lightprobe_average_scroll);
uniforms.push_back(u);
}
{
RD::Uniform u;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_IMAGE;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
u.binding = 13;
RID parent_average;
if (i < sdfgi->cascades.size() - 1) {
parent_average = sdfgi->cascades[i + 1].lightprobe_average_tex;
} else {
Fix spelling & grammar in comments, docs, and messages
2020-07-16 16:54:15 +00:00
parent_average = sdfgi->cascades[i - 1].lightprobe_average_tex; //to use something, but it won't be used
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
}
u.ids.push_back(parent_average);
uniforms.push_back(u);
}
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
{
RD::Uniform u;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_IMAGE;
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
u.binding = 14;
u.ids.push_back(sdfgi->ambient_texture);
uniforms.push_back(u);
}
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
sdfgi->cascades[i].integrate_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, sdfgi_shader.integrate.version_get_shader(sdfgi_shader.integrate_shader, 0), 0);
}
sdfgi->uses_multibounce = env->sdfgi_use_multibounce;
sdfgi->energy = env->sdfgi_energy;
sdfgi->normal_bias = env->sdfgi_normal_bias;
sdfgi->probe_bias = env->sdfgi_probe_bias;
sdfgi->reads_sky = env->sdfgi_read_sky_light;
_render_buffers_uniform_set_changed(p_render_buffers);
return; //done. all levels will need to be rendered which its going to take a bit
}
//check for updates
sdfgi->uses_multibounce = env->sdfgi_use_multibounce;
sdfgi->energy = env->sdfgi_energy;
sdfgi->normal_bias = env->sdfgi_normal_bias;
sdfgi->probe_bias = env->sdfgi_probe_bias;
sdfgi->reads_sky = env->sdfgi_read_sky_light;
int32_t drag_margin = (sdfgi->cascade_size / SDFGI::PROBE_DIVISOR) / 2;
for (uint32_t i = 0; i < sdfgi->cascades.size(); i++) {
SDFGI::Cascade &cascade = sdfgi->cascades[i];
cascade.dirty_regions = Vector3i();
Vector3 probe_half_size = Vector3(1, 1, 1) * cascade.cell_size * float(sdfgi->cascade_size / SDFGI::PROBE_DIVISOR) * 0.5;
probe_half_size = Vector3(0, 0, 0);
Vector3 world_position = p_world_position;
world_position.y *= sdfgi->y_mult;
Vector3i pos_in_cascade = Vector3i((world_position + probe_half_size) / cascade.cell_size);
for (int j = 0; j < 3; j++) {
if (pos_in_cascade[j] < cascade.position[j]) {
while (pos_in_cascade[j] < (cascade.position[j] - drag_margin)) {
cascade.position[j] -= drag_margin * 2;
cascade.dirty_regions[j] += drag_margin * 2;
}
} else if (pos_in_cascade[j] > cascade.position[j]) {
while (pos_in_cascade[j] > (cascade.position[j] + drag_margin)) {
cascade.position[j] += drag_margin * 2;
cascade.dirty_regions[j] -= drag_margin * 2;
}
}
if (cascade.dirty_regions[j] == 0) {
continue; // not dirty
} else if (uint32_t(ABS(cascade.dirty_regions[j])) >= sdfgi->cascade_size) {
//moved too much, just redraw everything (make all dirty)
cascade.dirty_regions = SDFGI::Cascade::DIRTY_ALL;
break;
}
}
if (cascade.dirty_regions != Vector3i() && cascade.dirty_regions != SDFGI::Cascade::DIRTY_ALL) {
//see how much the total dirty volume represents from the total volume
uint32_t total_volume = sdfgi->cascade_size * sdfgi->cascade_size * sdfgi->cascade_size;
uint32_t safe_volume = 1;
for (int j = 0; j < 3; j++) {
safe_volume *= sdfgi->cascade_size - ABS(cascade.dirty_regions[j]);
}
uint32_t dirty_volume = total_volume - safe_volume;
if (dirty_volume > (safe_volume / 2)) {
//more than half the volume is dirty, make all dirty so its only rendered once
cascade.dirty_regions = SDFGI::Cascade::DIRTY_ALL;
}
}
}
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
int RendererSceneRenderRD::sdfgi_get_pending_region_count(RID p_render_buffers) const {
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers);
ERR_FAIL_COND_V(rb == nullptr, 0);
if (rb->sdfgi == nullptr) {
return 0;
}
int dirty_count = 0;
for (uint32_t i = 0; i < rb->sdfgi->cascades.size(); i++) {
const SDFGI::Cascade &c = rb->sdfgi->cascades[i];
if (c.dirty_regions == SDFGI::Cascade::DIRTY_ALL) {
dirty_count++;
} else {
for (int j = 0; j < 3; j++) {
if (c.dirty_regions[j] != 0) {
dirty_count++;
}
}
}
}
return dirty_count;
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
int RendererSceneRenderRD::_sdfgi_get_pending_region_data(RID p_render_buffers, int p_region, Vector3i &r_local_offset, Vector3i &r_local_size, AABB &r_bounds) const {
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers);
ERR_FAIL_COND_V(rb == nullptr, -1);
ERR_FAIL_COND_V(rb->sdfgi == nullptr, -1);
int dirty_count = 0;
for (uint32_t i = 0; i < rb->sdfgi->cascades.size(); i++) {
const SDFGI::Cascade &c = rb->sdfgi->cascades[i];
if (c.dirty_regions == SDFGI::Cascade::DIRTY_ALL) {
if (dirty_count == p_region) {
r_local_offset = Vector3i();
r_local_size = Vector3i(1, 1, 1) * rb->sdfgi->cascade_size;
r_bounds.position = Vector3((Vector3i(1, 1, 1) * -int32_t(rb->sdfgi->cascade_size >> 1) + c.position)) * c.cell_size * Vector3(1, 1.0 / rb->sdfgi->y_mult, 1);
r_bounds.size = Vector3(r_local_size) * c.cell_size * Vector3(1, 1.0 / rb->sdfgi->y_mult, 1);
return i;
}
dirty_count++;
} else {
for (int j = 0; j < 3; j++) {
if (c.dirty_regions[j] != 0) {
if (dirty_count == p_region) {
Vector3i from = Vector3i(0, 0, 0);
Vector3i to = Vector3i(1, 1, 1) * rb->sdfgi->cascade_size;
if (c.dirty_regions[j] > 0) {
//fill from the beginning
to[j] = c.dirty_regions[j];
} else {
//fill from the end
from[j] = to[j] + c.dirty_regions[j];
}
for (int k = 0; k < j; k++) {
// "chip" away previous regions to avoid re-voxelizing the same thing
if (c.dirty_regions[k] > 0) {
from[k] += c.dirty_regions[k];
} else if (c.dirty_regions[k] < 0) {
to[k] += c.dirty_regions[k];
}
}
r_local_offset = from;
r_local_size = to - from;
r_bounds.position = Vector3(from + Vector3i(1, 1, 1) * -int32_t(rb->sdfgi->cascade_size >> 1) + c.position) * c.cell_size * Vector3(1, 1.0 / rb->sdfgi->y_mult, 1);
r_bounds.size = Vector3(r_local_size) * c.cell_size * Vector3(1, 1.0 / rb->sdfgi->y_mult, 1);
return i;
}
dirty_count++;
}
}
}
}
return -1;
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
AABB RendererSceneRenderRD::sdfgi_get_pending_region_bounds(RID p_render_buffers, int p_region) const {
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
AABB bounds;
Vector3i from;
Vector3i size;
int c = _sdfgi_get_pending_region_data(p_render_buffers, p_region, from, size, bounds);
ERR_FAIL_COND_V(c == -1, AABB());
return bounds;
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
uint32_t RendererSceneRenderRD::sdfgi_get_pending_region_cascade(RID p_render_buffers, int p_region) const {
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
AABB bounds;
Vector3i from;
Vector3i size;
return _sdfgi_get_pending_region_data(p_render_buffers, p_region, from, size, bounds);
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
void RendererSceneRenderRD::_sdfgi_update_cascades(RID p_render_buffers) {
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers);
ERR_FAIL_COND(rb == nullptr);
if (rb->sdfgi == nullptr) {
return;
}
//update cascades
SDFGI::Cascade::UBO cascade_data[SDFGI::MAX_CASCADES];
int32_t probe_divisor = rb->sdfgi->cascade_size / SDFGI::PROBE_DIVISOR;
for (uint32_t i = 0; i < rb->sdfgi->cascades.size(); i++) {
Vector3 pos = Vector3((Vector3i(1, 1, 1) * -int32_t(rb->sdfgi->cascade_size >> 1) + rb->sdfgi->cascades[i].position)) * rb->sdfgi->cascades[i].cell_size;
cascade_data[i].offset[0] = pos.x;
cascade_data[i].offset[1] = pos.y;
cascade_data[i].offset[2] = pos.z;
cascade_data[i].to_cell = 1.0 / rb->sdfgi->cascades[i].cell_size;
cascade_data[i].probe_offset[0] = rb->sdfgi->cascades[i].position.x / probe_divisor;
cascade_data[i].probe_offset[1] = rb->sdfgi->cascades[i].position.y / probe_divisor;
cascade_data[i].probe_offset[2] = rb->sdfgi->cascades[i].position.z / probe_divisor;
cascade_data[i].pad = 0;
}
Reorganize RenderingDevice barriers -Removed sync to draw, now everything syncs to draw by default. -Fixed many validation layer errors. -Added support for VkImageViewUsageCreateInfo to fix validation layer warnings. -Texture, buffer, raster and compute functions now all allow spcifying which barriers will be used.
2021-01-26 00:52:58 +00:00
RD::get_singleton()->buffer_update(rb->sdfgi->cascades_ubo, 0, sizeof(SDFGI::Cascade::UBO) * SDFGI::MAX_CASCADES, cascade_data);
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
}
Rewrite culling to be more cache/thread friendly. -Uses a single array with all data -Massive performance improvement -Does not support threads yet, but code is now thread friendly
2020-12-26 12:05:36 +00:00
void RendererSceneRenderRD::sdfgi_update_probes(RID p_render_buffers, RID p_environment, const Vector<RID> &p_directional_lights, const RID *p_positional_light_instances, uint32_t p_positional_light_count) {
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers);
ERR_FAIL_COND(rb == nullptr);
if (rb->sdfgi == nullptr) {
return;
}
Fix spelling of a var, a struct, and message output
2020-07-17 16:02:06 +00:00
Environment *env = environment_owner.getornull(p_environment);
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
RENDER_TIMESTAMP(">SDFGI Update Probes");
/* Update Cascades UBO */
_sdfgi_update_cascades(p_render_buffers);
/* Update Dynamic Lights Buffer */
RENDER_TIMESTAMP("Update Lights");
/* Update dynamic lights */
{
Fix invalid buffer updates in SDFGI
2020-10-19 20:01:53 +00:00
int32_t cascade_light_count[SDFGI::MAX_CASCADES];
for (uint32_t i = 0; i < rb->sdfgi->cascades.size(); i++) {
SDFGI::Cascade &cascade = rb->sdfgi->cascades[i];
SDGIShader::Light lights[SDFGI::MAX_DYNAMIC_LIGHTS];
uint32_t idx = 0;
Rewrite culling to be more cache/thread friendly. -Uses a single array with all data -Massive performance improvement -Does not support threads yet, but code is now thread friendly
2020-12-26 12:05:36 +00:00
for (uint32_t j = 0; j < (uint32_t)p_directional_lights.size(); j++) {
Fix invalid buffer updates in SDFGI
2020-10-19 20:01:53 +00:00
if (idx == SDFGI::MAX_DYNAMIC_LIGHTS) {
break;
}
Rewrite culling to be more cache/thread friendly. -Uses a single array with all data -Massive performance improvement -Does not support threads yet, but code is now thread friendly
2020-12-26 12:05:36 +00:00
LightInstance *li = light_instance_owner.getornull(p_directional_lights[j]);
Fix invalid buffer updates in SDFGI
2020-10-19 20:01:53 +00:00
ERR_CONTINUE(!li);
Add sky_only setting to DirectionalLight3Ds
2020-10-22 06:22:53 +00:00
if (storage->light_directional_is_sky_only(li->light)) {
continue;
}
Fix invalid buffer updates in SDFGI
2020-10-19 20:01:53 +00:00
Vector3 dir = -li->transform.basis.get_axis(Vector3::AXIS_Z);
dir.y *= rb->sdfgi->y_mult;
dir.normalize();
lights[idx].direction[0] = dir.x;
lights[idx].direction[1] = dir.y;
lights[idx].direction[2] = dir.z;
Color color = storage->light_get_color(li->light);
color = color.to_linear();
lights[idx].color[0] = color.r;
lights[idx].color[1] = color.g;
lights[idx].color[2] = color.b;
lights[idx].type = RS::LIGHT_DIRECTIONAL;
lights[idx].energy = storage->light_get_param(li->light, RS::LIGHT_PARAM_ENERGY);
lights[idx].has_shadow = storage->light_has_shadow(li->light);
idx++;
}
AABB cascade_aabb;
cascade_aabb.position = Vector3((Vector3i(1, 1, 1) * -int32_t(rb->sdfgi->cascade_size >> 1) + cascade.position)) * cascade.cell_size;
cascade_aabb.size = Vector3(1, 1, 1) * rb->sdfgi->cascade_size * cascade.cell_size;
for (uint32_t j = 0; j < p_positional_light_count; j++) {
if (idx == SDFGI::MAX_DYNAMIC_LIGHTS) {
break;
}
LightInstance *li = light_instance_owner.getornull(p_positional_light_instances[j]);
ERR_CONTINUE(!li);
uint32_t max_sdfgi_cascade = storage->light_get_max_sdfgi_cascade(li->light);
if (i > max_sdfgi_cascade) {
continue;
}
if (!cascade_aabb.intersects(li->aabb)) {
continue;
}
Vector3 dir = -li->transform.basis.get_axis(Vector3::AXIS_Z);
//faster to not do this here
//dir.y *= rb->sdfgi->y_mult;
//dir.normalize();
lights[idx].direction[0] = dir.x;
lights[idx].direction[1] = dir.y;
lights[idx].direction[2] = dir.z;
Vector3 pos = li->transform.origin;
pos.y *= rb->sdfgi->y_mult;
lights[idx].position[0] = pos.x;
lights[idx].position[1] = pos.y;
lights[idx].position[2] = pos.z;
Color color = storage->light_get_color(li->light);
color = color.to_linear();
lights[idx].color[0] = color.r;
lights[idx].color[1] = color.g;
lights[idx].color[2] = color.b;
lights[idx].type = storage->light_get_type(li->light);
lights[idx].energy = storage->light_get_param(li->light, RS::LIGHT_PARAM_ENERGY);
lights[idx].has_shadow = storage->light_has_shadow(li->light);
lights[idx].attenuation = storage->light_get_param(li->light, RS::LIGHT_PARAM_ATTENUATION);
lights[idx].radius = storage->light_get_param(li->light, RS::LIGHT_PARAM_RANGE);
lights[idx].spot_angle = Math::deg2rad(storage->light_get_param(li->light, RS::LIGHT_PARAM_SPOT_ANGLE));
lights[idx].spot_attenuation = storage->light_get_param(li->light, RS::LIGHT_PARAM_SPOT_ATTENUATION);
idx++;
}
if (idx > 0) {
Reorganize RenderingDevice barriers -Removed sync to draw, now everything syncs to draw by default. -Fixed many validation layer errors. -Added support for VkImageViewUsageCreateInfo to fix validation layer warnings. -Texture, buffer, raster and compute functions now all allow spcifying which barriers will be used.
2021-01-26 00:52:58 +00:00
RD::get_singleton()->buffer_update(cascade.lights_buffer, 0, idx * sizeof(SDGIShader::Light), lights);
Fix invalid buffer updates in SDFGI
2020-10-19 20:01:53 +00:00
}
cascade_light_count[i] = idx;
}
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin();
RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, sdfgi_shader.direct_light_pipeline[SDGIShader::DIRECT_LIGHT_MODE_DYNAMIC]);
SDGIShader::DirectLightPushConstant push_constant;
push_constant.grid_size[0] = rb->sdfgi->cascade_size;
push_constant.grid_size[1] = rb->sdfgi->cascade_size;
push_constant.grid_size[2] = rb->sdfgi->cascade_size;
push_constant.max_cascades = rb->sdfgi->cascades.size();
push_constant.probe_axis_size = rb->sdfgi->probe_axis_count;
push_constant.multibounce = rb->sdfgi->uses_multibounce;
push_constant.y_mult = rb->sdfgi->y_mult;
for (uint32_t i = 0; i < rb->sdfgi->cascades.size(); i++) {
SDFGI::Cascade &cascade = rb->sdfgi->cascades[i];
Fix invalid buffer updates in SDFGI
2020-10-19 20:01:53 +00:00
push_constant.light_count = cascade_light_count[i];
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
push_constant.cascade = i;
Several GI related optimizations and fixes -SDFGI direct light is done over many frames -SDFGI Changed settings for rays/frame -SDFGI Misc optimizations -SDFGI Bug fix on probe scroll -GIProbe was not working, got it to work again -GIProbe dynamic objects were not working, fixed -Added a half size GI option.
2021-01-22 23:50:24 +00:00
if (rb->sdfgi->cascades[i].all_dynamic_lights_dirty || sdfgi_frames_to_update_light == RS::ENV_SDFGI_UPDATE_LIGHT_IN_1_FRAME) {
push_constant.process_offset = 0;
push_constant.process_increment = 1;
} else {
static uint32_t frames_to_update_table[RS::ENV_SDFGI_UPDATE_LIGHT_MAX] = {
1, 2, 4, 8, 16
};
uint32_t frames_to_update = frames_to_update_table[sdfgi_frames_to_update_light];
push_constant.process_offset = RSG::rasterizer->get_frame_number() % frames_to_update;
push_constant.process_increment = frames_to_update;
}
rb->sdfgi->cascades[i].all_dynamic_lights_dirty = false;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
RD::get_singleton()->compute_list_bind_uniform_set(compute_list, cascade.sdf_direct_light_uniform_set, 0);
RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(SDGIShader::DirectLightPushConstant));
RD::get_singleton()->compute_list_dispatch_indirect(compute_list, cascade.solid_cell_dispatch_buffer, 0);
}
RD::get_singleton()->compute_list_end();
}
RENDER_TIMESTAMP("Raytrace");
SDGIShader::IntegratePushConstant push_constant;
push_constant.grid_size[1] = rb->sdfgi->cascade_size;
push_constant.grid_size[2] = rb->sdfgi->cascade_size;
push_constant.grid_size[0] = rb->sdfgi->cascade_size;
push_constant.max_cascades = rb->sdfgi->cascades.size();
push_constant.probe_axis_size = rb->sdfgi->probe_axis_count;
push_constant.history_index = rb->sdfgi->render_pass % rb->sdfgi->history_size;
push_constant.history_size = rb->sdfgi->history_size;
Several GI related optimizations and fixes -SDFGI direct light is done over many frames -SDFGI Changed settings for rays/frame -SDFGI Misc optimizations -SDFGI Bug fix on probe scroll -GIProbe was not working, got it to work again -GIProbe dynamic objects were not working, fixed -Added a half size GI option.
2021-01-22 23:50:24 +00:00
static const uint32_t ray_count[RS::ENV_SDFGI_RAY_COUNT_MAX] = { 4, 8, 16, 32, 64, 96, 128 };
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
push_constant.ray_count = ray_count[sdfgi_ray_count];
push_constant.ray_bias = rb->sdfgi->probe_bias;
push_constant.image_size[0] = rb->sdfgi->probe_axis_count * rb->sdfgi->probe_axis_count;
push_constant.image_size[1] = rb->sdfgi->probe_axis_count;
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
push_constant.store_ambient_texture = env->volumetric_fog_enabled;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
RID sky_uniform_set = sdfgi_shader.integrate_default_sky_uniform_set;
push_constant.sky_mode = SDGIShader::IntegratePushConstant::SKY_MODE_DISABLED;
push_constant.y_mult = rb->sdfgi->y_mult;
if (rb->sdfgi->reads_sky && env) {
push_constant.sky_energy = env->bg_energy;
if (env->background == RS::ENV_BG_CLEAR_COLOR) {
push_constant.sky_mode = SDGIShader::IntegratePushConstant::SKY_MODE_COLOR;
Color c = storage->get_default_clear_color().to_linear();
push_constant.sky_color[0] = c.r;
push_constant.sky_color[1] = c.g;
push_constant.sky_color[2] = c.b;
} else if (env->background == RS::ENV_BG_COLOR) {
push_constant.sky_mode = SDGIShader::IntegratePushConstant::SKY_MODE_COLOR;
Color c = env->bg_color;
push_constant.sky_color[0] = c.r;
push_constant.sky_color[1] = c.g;
push_constant.sky_color[2] = c.b;
} else if (env->background == RS::ENV_BG_SKY) {
Sky *sky = sky_owner.getornull(env->sky);
if (sky && sky->radiance.is_valid()) {
if (sky->sdfgi_integrate_sky_uniform_set.is_null() || !RD::get_singleton()->uniform_set_is_valid(sky->sdfgi_integrate_sky_uniform_set)) {
Vector<RD::Uniform> uniforms;
{
RD::Uniform u;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_TEXTURE;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
u.binding = 0;
u.ids.push_back(sky->radiance);
uniforms.push_back(u);
}
{
RD::Uniform u;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_SAMPLER;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
u.binding = 1;
u.ids.push_back(storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED));
uniforms.push_back(u);
}
sky->sdfgi_integrate_sky_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, sdfgi_shader.integrate.version_get_shader(sdfgi_shader.integrate_shader, 0), 1);
}
sky_uniform_set = sky->sdfgi_integrate_sky_uniform_set;
push_constant.sky_mode = SDGIShader::IntegratePushConstant::SKY_MODE_SKY;
}
}
}
rb->sdfgi->render_pass++;
RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin();
RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, sdfgi_shader.integrate_pipeline[SDGIShader::INTEGRATE_MODE_PROCESS]);
int32_t probe_divisor = rb->sdfgi->cascade_size / SDFGI::PROBE_DIVISOR;
for (uint32_t i = 0; i < rb->sdfgi->cascades.size(); i++) {
push_constant.cascade = i;
push_constant.world_offset[0] = rb->sdfgi->cascades[i].position.x / probe_divisor;
push_constant.world_offset[1] = rb->sdfgi->cascades[i].position.y / probe_divisor;
push_constant.world_offset[2] = rb->sdfgi->cascades[i].position.z / probe_divisor;
RD::get_singleton()->compute_list_bind_uniform_set(compute_list, rb->sdfgi->cascades[i].integrate_uniform_set, 0);
RD::get_singleton()->compute_list_bind_uniform_set(compute_list, sky_uniform_set, 1);
RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(SDGIShader::IntegratePushConstant));
RD::get_singleton()->compute_list_dispatch_threads(compute_list, rb->sdfgi->probe_axis_count * rb->sdfgi->probe_axis_count, rb->sdfgi->probe_axis_count, 1, 8, 8, 1);
}
RD::get_singleton()->compute_list_add_barrier(compute_list); //wait until done
// Then store values into the lightprobe texture. Separating these steps has a small performance hit, but it allows for multiple bounces
RENDER_TIMESTAMP("Average Probes");
RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, sdfgi_shader.integrate_pipeline[SDGIShader::INTEGRATE_MODE_STORE]);
//convert to octahedral to store
push_constant.image_size[0] *= SDFGI::LIGHTPROBE_OCT_SIZE;
push_constant.image_size[1] *= SDFGI::LIGHTPROBE_OCT_SIZE;
for (uint32_t i = 0; i < rb->sdfgi->cascades.size(); i++) {
push_constant.cascade = i;
RD::get_singleton()->compute_list_bind_uniform_set(compute_list, rb->sdfgi->cascades[i].integrate_uniform_set, 0);
RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(SDGIShader::IntegratePushConstant));
RD::get_singleton()->compute_list_dispatch_threads(compute_list, rb->sdfgi->probe_axis_count * rb->sdfgi->probe_axis_count * SDFGI::LIGHTPROBE_OCT_SIZE, rb->sdfgi->probe_axis_count * SDFGI::LIGHTPROBE_OCT_SIZE, 1, 8, 8, 1);
}
RD::get_singleton()->compute_list_end();
RENDER_TIMESTAMP("<SDFGI Update Probes");
}
Replace Octree by DynamicBVH in cull code -Much greater pairing/unpairing performance -For now, using it for culling too, but this will change in a couple of days. -Added a paged allocator, to efficiently alloc/free some types of objects.
2020-12-23 16:52:58 +00:00
void RendererSceneRenderRD::_setup_giprobes(RID p_render_buffers, const Transform &p_transform, const PagedArray<RID> &p_gi_probes, uint32_t &r_gi_probes_used) {
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
r_gi_probes_used = 0;
RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers);
ERR_FAIL_COND(rb == nullptr);
RID gi_probe_buffer = render_buffers_get_gi_probe_buffer(p_render_buffers);
GI::GIProbeData gi_probe_data[RenderBuffers::MAX_GIPROBES];
bool giprobes_changed = false;
Transform to_camera;
to_camera.origin = p_transform.origin; //only translation, make local
for (int i = 0; i < RenderBuffers::MAX_GIPROBES; i++) {
RID texture;
Replace Octree by DynamicBVH in cull code -Much greater pairing/unpairing performance -For now, using it for culling too, but this will change in a couple of days. -Added a paged allocator, to efficiently alloc/free some types of objects.
2020-12-23 16:52:58 +00:00
if (i < (int)p_gi_probes.size()) {
GIProbeInstance *gipi = gi_probe_instance_owner.getornull(p_gi_probes[i]);
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
if (gipi) {
texture = gipi->texture;
GI::GIProbeData &gipd = gi_probe_data[i];
RID base_probe = gipi->probe;
Transform to_cell = storage->gi_probe_get_to_cell_xform(gipi->probe) * gipi->transform.affine_inverse() * to_camera;
gipd.xform[0] = to_cell.basis.elements[0][0];
gipd.xform[1] = to_cell.basis.elements[1][0];
gipd.xform[2] = to_cell.basis.elements[2][0];
gipd.xform[3] = 0;
gipd.xform[4] = to_cell.basis.elements[0][1];
gipd.xform[5] = to_cell.basis.elements[1][1];
gipd.xform[6] = to_cell.basis.elements[2][1];
gipd.xform[7] = 0;
gipd.xform[8] = to_cell.basis.elements[0][2];
gipd.xform[9] = to_cell.basis.elements[1][2];
gipd.xform[10] = to_cell.basis.elements[2][2];
gipd.xform[11] = 0;
gipd.xform[12] = to_cell.origin.x;
gipd.xform[13] = to_cell.origin.y;
gipd.xform[14] = to_cell.origin.z;
gipd.xform[15] = 1;
Vector3 bounds = storage->gi_probe_get_octree_size(base_probe);
gipd.bounds[0] = bounds.x;
gipd.bounds[1] = bounds.y;
gipd.bounds[2] = bounds.z;
gipd.dynamic_range = storage->gi_probe_get_dynamic_range(base_probe) * storage->gi_probe_get_energy(base_probe);
gipd.bias = storage->gi_probe_get_bias(base_probe);
gipd.normal_bias = storage->gi_probe_get_normal_bias(base_probe);
gipd.blend_ambient = !storage->gi_probe_is_interior(base_probe);
gipd.anisotropy_strength = 0;
gipd.ao = storage->gi_probe_get_ao(base_probe);
gipd.ao_size = Math::pow(storage->gi_probe_get_ao_size(base_probe), 4.0f);
gipd.mipmaps = gipi->mipmaps.size();
}
r_gi_probes_used++;
}
if (texture == RID()) {
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
texture = storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_3D_WHITE);
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
}
if (texture != rb->giprobe_textures[i]) {
giprobes_changed = true;
rb->giprobe_textures[i] = texture;
}
}
if (giprobes_changed) {
Check uniform set validity before freeing
2020-11-19 00:49:56 +00:00
if (RD::get_singleton()->uniform_set_is_valid(rb->gi_uniform_set)) {
RD::get_singleton()->free(rb->gi_uniform_set);
}
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
rb->gi_uniform_set = RID();
if (rb->volumetric_fog) {
if (RD::get_singleton()->uniform_set_is_valid(rb->volumetric_fog->uniform_set)) {
RD::get_singleton()->free(rb->volumetric_fog->uniform_set);
RD::get_singleton()->free(rb->volumetric_fog->uniform_set2);
}
rb->volumetric_fog->uniform_set = RID();
rb->volumetric_fog->uniform_set2 = RID();
}
}
Replace Octree by DynamicBVH in cull code -Much greater pairing/unpairing performance -For now, using it for culling too, but this will change in a couple of days. -Added a paged allocator, to efficiently alloc/free some types of objects.
2020-12-23 16:52:58 +00:00
if (p_gi_probes.size() > 0) {
Reorganize RenderingDevice barriers -Removed sync to draw, now everything syncs to draw by default. -Fixed many validation layer errors. -Added support for VkImageViewUsageCreateInfo to fix validation layer warnings. -Texture, buffer, raster and compute functions now all allow spcifying which barriers will be used.
2021-01-26 00:52:58 +00:00
RD::get_singleton()->buffer_update(gi_probe_buffer, 0, sizeof(GI::GIProbeData) * MIN((uint64_t)RenderBuffers::MAX_GIPROBES, p_gi_probes.size()), gi_probe_data);
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
}
}
Several GI related optimizations and fixes -SDFGI direct light is done over many frames -SDFGI Changed settings for rays/frame -SDFGI Misc optimizations -SDFGI Bug fix on probe scroll -GIProbe was not working, got it to work again -GIProbe dynamic objects were not working, fixed -Added a half size GI option.
2021-01-22 23:50:24 +00:00
void RendererSceneRenderRD::_process_gi(RID p_render_buffers, RID p_normal_roughness_buffer, RID p_gi_probe_buffer, RID p_environment, const CameraMatrix &p_projection, const Transform &p_transform, const PagedArray<RID> &p_gi_probes) {
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
RENDER_TIMESTAMP("Render GI");
RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers);
ERR_FAIL_COND(rb == nullptr);
Fix spelling of a var, a struct, and message output
2020-07-17 16:02:06 +00:00
Environment *env = environment_owner.getornull(p_environment);
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
Several GI related optimizations and fixes -SDFGI direct light is done over many frames -SDFGI Changed settings for rays/frame -SDFGI Misc optimizations -SDFGI Bug fix on probe scroll -GIProbe was not working, got it to work again -GIProbe dynamic objects were not working, fixed -Added a half size GI option.
2021-01-22 23:50:24 +00:00
if (rb->ambient_buffer.is_null() || rb->using_half_size_gi != gi.half_resolution) {
if (rb->ambient_buffer.is_valid()) {
RD::get_singleton()->free(rb->ambient_buffer);
RD::get_singleton()->free(rb->reflection_buffer);
}
RD::TextureFormat tf;
tf.format = RD::DATA_FORMAT_R16G16B16A16_SFLOAT;
tf.width = rb->width;
tf.height = rb->height;
if (gi.half_resolution) {
tf.width >>= 1;
tf.height >>= 1;
}
tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_STORAGE_BIT;
rb->reflection_buffer = RD::get_singleton()->texture_create(tf, RD::TextureView());
rb->ambient_buffer = RD::get_singleton()->texture_create(tf, RD::TextureView());
rb->using_half_size_gi = gi.half_resolution;
_render_buffers_uniform_set_changed(p_render_buffers);
}
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
GI::PushConstant push_constant;
push_constant.screen_size[0] = rb->width;
push_constant.screen_size[1] = rb->height;
push_constant.z_near = p_projection.get_z_near();
push_constant.z_far = p_projection.get_z_far();
push_constant.orthogonal = p_projection.is_orthogonal();
push_constant.proj_info[0] = -2.0f / (rb->width * p_projection.matrix[0][0]);
push_constant.proj_info[1] = -2.0f / (rb->height * p_projection.matrix[1][1]);
push_constant.proj_info[2] = (1.0f - p_projection.matrix[0][2]) / p_projection.matrix[0][0];
push_constant.proj_info[3] = (1.0f + p_projection.matrix[1][2]) / p_projection.matrix[1][1];
Replace Octree by DynamicBVH in cull code -Much greater pairing/unpairing performance -For now, using it for culling too, but this will change in a couple of days. -Added a paged allocator, to efficiently alloc/free some types of objects.
2020-12-23 16:52:58 +00:00
push_constant.max_giprobes = MIN((uint64_t)RenderBuffers::MAX_GIPROBES, p_gi_probes.size());
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
push_constant.high_quality_vct = gi_probe_quality == RS::GI_PROBE_QUALITY_HIGH;
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
bool use_sdfgi = rb->sdfgi != nullptr;
bool use_giprobes = push_constant.max_giprobes > 0;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
if (env) {
push_constant.ao_color[0] = env->ao_color.r;
push_constant.ao_color[1] = env->ao_color.g;
push_constant.ao_color[2] = env->ao_color.b;
} else {
push_constant.ao_color[0] = 0;
push_constant.ao_color[1] = 0;
push_constant.ao_color[2] = 0;
}
push_constant.cam_rotation[0] = p_transform.basis[0][0];
push_constant.cam_rotation[1] = p_transform.basis[1][0];
push_constant.cam_rotation[2] = p_transform.basis[2][0];
push_constant.cam_rotation[3] = 0;
push_constant.cam_rotation[4] = p_transform.basis[0][1];
push_constant.cam_rotation[5] = p_transform.basis[1][1];
push_constant.cam_rotation[6] = p_transform.basis[2][1];
push_constant.cam_rotation[7] = 0;
push_constant.cam_rotation[8] = p_transform.basis[0][2];
push_constant.cam_rotation[9] = p_transform.basis[1][2];
push_constant.cam_rotation[10] = p_transform.basis[2][2];
push_constant.cam_rotation[11] = 0;
if (rb->sdfgi) {
GI::SDFGIData sdfgi_data;
sdfgi_data.grid_size[0] = rb->sdfgi->cascade_size;
sdfgi_data.grid_size[1] = rb->sdfgi->cascade_size;
sdfgi_data.grid_size[2] = rb->sdfgi->cascade_size;
sdfgi_data.max_cascades = rb->sdfgi->cascades.size();
sdfgi_data.probe_axis_size = rb->sdfgi->probe_axis_count;
sdfgi_data.cascade_probe_size[0] = sdfgi_data.probe_axis_size - 1; //float version for performance
sdfgi_data.cascade_probe_size[1] = sdfgi_data.probe_axis_size - 1;
sdfgi_data.cascade_probe_size[2] = sdfgi_data.probe_axis_size - 1;
float csize = rb->sdfgi->cascade_size;
sdfgi_data.probe_to_uvw = 1.0 / float(sdfgi_data.cascade_probe_size[0]);
sdfgi_data.use_occlusion = rb->sdfgi->uses_occlusion;
//sdfgi_data.energy = rb->sdfgi->energy;
sdfgi_data.y_mult = rb->sdfgi->y_mult;
float cascade_voxel_size = (csize / sdfgi_data.cascade_probe_size[0]);
float occlusion_clamp = (cascade_voxel_size - 0.5) / cascade_voxel_size;
sdfgi_data.occlusion_clamp[0] = occlusion_clamp;
sdfgi_data.occlusion_clamp[1] = occlusion_clamp;
sdfgi_data.occlusion_clamp[2] = occlusion_clamp;
sdfgi_data.normal_bias = (rb->sdfgi->normal_bias / csize) * sdfgi_data.cascade_probe_size[0];
//vec2 tex_pixel_size = 1.0 / vec2(ivec2( (OCT_SIZE+2) * params.probe_axis_size * params.probe_axis_size, (OCT_SIZE+2) * params.probe_axis_size ) );
//vec3 probe_uv_offset = (ivec3(OCT_SIZE+2,OCT_SIZE+2,(OCT_SIZE+2) * params.probe_axis_size)) * tex_pixel_size.xyx;
uint32_t oct_size = SDFGI::LIGHTPROBE_OCT_SIZE;
sdfgi_data.lightprobe_tex_pixel_size[0] = 1.0 / ((oct_size + 2) * sdfgi_data.probe_axis_size * sdfgi_data.probe_axis_size);
sdfgi_data.lightprobe_tex_pixel_size[1] = 1.0 / ((oct_size + 2) * sdfgi_data.probe_axis_size);
sdfgi_data.lightprobe_tex_pixel_size[2] = 1.0;
sdfgi_data.energy = rb->sdfgi->energy;
sdfgi_data.lightprobe_uv_offset[0] = float(oct_size + 2) * sdfgi_data.lightprobe_tex_pixel_size[0];
sdfgi_data.lightprobe_uv_offset[1] = float(oct_size + 2) * sdfgi_data.lightprobe_tex_pixel_size[1];
sdfgi_data.lightprobe_uv_offset[2] = float((oct_size + 2) * sdfgi_data.probe_axis_size) * sdfgi_data.lightprobe_tex_pixel_size[0];
sdfgi_data.occlusion_renormalize[0] = 0.5;
sdfgi_data.occlusion_renormalize[1] = 1.0;
sdfgi_data.occlusion_renormalize[2] = 1.0 / float(sdfgi_data.max_cascades);
int32_t probe_divisor = rb->sdfgi->cascade_size / SDFGI::PROBE_DIVISOR;
for (uint32_t i = 0; i < sdfgi_data.max_cascades; i++) {
GI::SDFGIData::ProbeCascadeData &c = sdfgi_data.cascades[i];
Vector3 pos = Vector3((Vector3i(1, 1, 1) * -int32_t(rb->sdfgi->cascade_size >> 1) + rb->sdfgi->cascades[i].position)) * rb->sdfgi->cascades[i].cell_size;
Vector3 cam_origin = p_transform.origin;
cam_origin.y *= rb->sdfgi->y_mult;
pos -= cam_origin; //make pos local to camera, to reduce numerical error
c.position[0] = pos.x;
c.position[1] = pos.y;
c.position[2] = pos.z;
c.to_probe = 1.0 / (float(rb->sdfgi->cascade_size) * rb->sdfgi->cascades[i].cell_size / float(rb->sdfgi->probe_axis_count - 1));
Vector3i probe_ofs = rb->sdfgi->cascades[i].position / probe_divisor;
c.probe_world_offset[0] = probe_ofs.x;
c.probe_world_offset[1] = probe_ofs.y;
c.probe_world_offset[2] = probe_ofs.z;
c.to_cell = 1.0 / rb->sdfgi->cascades[i].cell_size;
}
Reorganize RenderingDevice barriers -Removed sync to draw, now everything syncs to draw by default. -Fixed many validation layer errors. -Added support for VkImageViewUsageCreateInfo to fix validation layer warnings. -Texture, buffer, raster and compute functions now all allow spcifying which barriers will be used.
2021-01-26 00:52:58 +00:00
RD::get_singleton()->buffer_update(gi.sdfgi_ubo, 0, sizeof(GI::SDFGIData), &sdfgi_data);
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
}
if (rb->gi_uniform_set.is_null() || !RD::get_singleton()->uniform_set_is_valid(rb->gi_uniform_set)) {
Vector<RD::Uniform> uniforms;
{
RD::Uniform u;
u.binding = 1;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_TEXTURE;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
for (uint32_t j = 0; j < SDFGI::MAX_CASCADES; j++) {
if (rb->sdfgi && j < rb->sdfgi->cascades.size()) {
u.ids.push_back(rb->sdfgi->cascades[j].sdf_tex);
} else {
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
u.ids.push_back(storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_3D_WHITE));
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
}
}
uniforms.push_back(u);
}
{
RD::Uniform u;
u.binding = 2;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_TEXTURE;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
for (uint32_t j = 0; j < SDFGI::MAX_CASCADES; j++) {
if (rb->sdfgi && j < rb->sdfgi->cascades.size()) {
u.ids.push_back(rb->sdfgi->cascades[j].light_tex);
} else {
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
u.ids.push_back(storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_3D_WHITE));
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
}
}
uniforms.push_back(u);
}
{
RD::Uniform u;
u.binding = 3;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_TEXTURE;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
for (uint32_t j = 0; j < SDFGI::MAX_CASCADES; j++) {
if (rb->sdfgi && j < rb->sdfgi->cascades.size()) {
u.ids.push_back(rb->sdfgi->cascades[j].light_aniso_0_tex);
} else {
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
u.ids.push_back(storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_3D_WHITE));
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
}
}
uniforms.push_back(u);
}
{
RD::Uniform u;
u.binding = 4;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_TEXTURE;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
for (uint32_t j = 0; j < SDFGI::MAX_CASCADES; j++) {
if (rb->sdfgi && j < rb->sdfgi->cascades.size()) {
u.ids.push_back(rb->sdfgi->cascades[j].light_aniso_1_tex);
} else {
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
u.ids.push_back(storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_3D_WHITE));
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
}
}
uniforms.push_back(u);
}
{
RD::Uniform u;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_TEXTURE;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
u.binding = 5;
if (rb->sdfgi) {
u.ids.push_back(rb->sdfgi->occlusion_texture);
} else {
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
u.ids.push_back(storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_3D_WHITE));
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
}
uniforms.push_back(u);
}
{
RD::Uniform u;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_SAMPLER;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
u.binding = 6;
u.ids.push_back(storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED));
uniforms.push_back(u);
}
{
RD::Uniform u;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_SAMPLER;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
u.binding = 7;
u.ids.push_back(storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED));
uniforms.push_back(u);
}
{
RD::Uniform u;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_IMAGE;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
u.binding = 9;
Several GI related optimizations and fixes -SDFGI direct light is done over many frames -SDFGI Changed settings for rays/frame -SDFGI Misc optimizations -SDFGI Bug fix on probe scroll -GIProbe was not working, got it to work again -GIProbe dynamic objects were not working, fixed -Added a half size GI option.
2021-01-22 23:50:24 +00:00
u.ids.push_back(rb->ambient_buffer);
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
uniforms.push_back(u);
}
{
RD::Uniform u;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_IMAGE;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
u.binding = 10;
Several GI related optimizations and fixes -SDFGI direct light is done over many frames -SDFGI Changed settings for rays/frame -SDFGI Misc optimizations -SDFGI Bug fix on probe scroll -GIProbe was not working, got it to work again -GIProbe dynamic objects were not working, fixed -Added a half size GI option.
2021-01-22 23:50:24 +00:00
u.ids.push_back(rb->reflection_buffer);
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
uniforms.push_back(u);
}
{
RD::Uniform u;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_TEXTURE;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
u.binding = 11;
if (rb->sdfgi) {
u.ids.push_back(rb->sdfgi->lightprobe_texture);
} else {
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
u.ids.push_back(storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_2D_ARRAY_WHITE));
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
}
uniforms.push_back(u);
}
{
RD::Uniform u;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_TEXTURE;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
u.binding = 12;
u.ids.push_back(rb->depth_texture);
uniforms.push_back(u);
}
{
RD::Uniform u;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_TEXTURE;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
u.binding = 13;
u.ids.push_back(p_normal_roughness_buffer);
uniforms.push_back(u);
}
{
RD::Uniform u;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_TEXTURE;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
u.binding = 14;
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
RID buffer = p_gi_probe_buffer.is_valid() ? p_gi_probe_buffer : storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_BLACK);
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
u.ids.push_back(buffer);
uniforms.push_back(u);
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
}
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
{
RD::Uniform u;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
u.binding = 15;
u.ids.push_back(gi.sdfgi_ubo);
uniforms.push_back(u);
}
{
RD::Uniform u;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
u.binding = 16;
u.ids.push_back(rb->giprobe_buffer);
uniforms.push_back(u);
}
{
RD::Uniform u;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_TEXTURE;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
u.binding = 17;
for (int i = 0; i < RenderBuffers::MAX_GIPROBES; i++) {
u.ids.push_back(rb->giprobe_textures[i]);
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
}
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
uniforms.push_back(u);
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
}
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
rb->gi_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, gi.shader.version_get_shader(gi.shader_version, 0), 0);
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
}
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
Several GI related optimizations and fixes -SDFGI direct light is done over many frames -SDFGI Changed settings for rays/frame -SDFGI Misc optimizations -SDFGI Bug fix on probe scroll -GIProbe was not working, got it to work again -GIProbe dynamic objects were not working, fixed -Added a half size GI option.
2021-01-22 23:50:24 +00:00
GI::Mode mode;
if (rb->using_half_size_gi) {
mode = (use_sdfgi && use_giprobes) ? GI::MODE_HALF_RES_COMBINED : (use_sdfgi ? GI::MODE_HALF_RES_SDFGI : GI::MODE_HALF_RES_GIPROBE);
} else {
mode = (use_sdfgi && use_giprobes) ? GI::MODE_COMBINED : (use_sdfgi ? GI::MODE_SDFGI : GI::MODE_GIPROBE);
}
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin();
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, gi.pipelines[mode]);
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
RD::get_singleton()->compute_list_bind_uniform_set(compute_list, rb->gi_uniform_set, 0);
RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(GI::PushConstant));
Several GI related optimizations and fixes -SDFGI direct light is done over many frames -SDFGI Changed settings for rays/frame -SDFGI Misc optimizations -SDFGI Bug fix on probe scroll -GIProbe was not working, got it to work again -GIProbe dynamic objects were not working, fixed -Added a half size GI option.
2021-01-22 23:50:24 +00:00
if (rb->using_half_size_gi) {
RD::get_singleton()->compute_list_dispatch_threads(compute_list, rb->width >> 1, rb->height >> 1, 1, 8, 8, 1);
} else {
RD::get_singleton()->compute_list_dispatch_threads(compute_list, rb->width, rb->height, 1, 8, 8, 1);
}
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
RD::get_singleton()->compute_list_end();
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
RID RendererSceneRenderRD::sky_create() {
Environment sky more or less working.
2019-08-26 20:43:58 +00:00
return sky_owner.make_rid(Sky());
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
void RendererSceneRenderRD::_sky_invalidate(Sky *p_sky) {
Environment sky more or less working.
2019-08-26 20:43:58 +00:00
if (!p_sky->dirty) {
p_sky->dirty = true;
p_sky->dirty_list = dirty_sky_list;
dirty_sky_list = p_sky;
}
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
void RendererSceneRenderRD::sky_set_radiance_size(RID p_sky, int p_radiance_size) {
Environment sky more or less working.
2019-08-26 20:43:58 +00:00
Sky *sky = sky_owner.getornull(p_sky);
ERR_FAIL_COND(!sky);
ERR_FAIL_COND(p_radiance_size < 32 || p_radiance_size > 2048);
if (sky->radiance_size == p_radiance_size) {
return;
}
sky->radiance_size = p_radiance_size;
Basic fast filtering implementation
2020-02-20 23:27:34 +00:00
Renaming of servers for coherency. VisualServer -> RenderingServer PhysicsServer -> PhysicsServer3D Physics2DServer -> PhysicsServer2D NavigationServer -> NavigationServer3D Navigation2DServer -> NavigationServer2D Also renamed corresponding files.
2020-03-27 18:21:27 +00:00
if (sky->mode == RS::SKY_MODE_REALTIME && sky->radiance_size != 256) {
Working sky shader implementation
2020-03-20 00:32:19 +00:00
WARN_PRINT("Realtime Skies can only use a radiance size of 256. Radiance size will be set to 256 internally.");
sky->radiance_size = 256;
Basic fast filtering implementation
2020-02-20 23:27:34 +00:00
}
Environment sky more or less working.
2019-08-26 20:43:58 +00:00
_sky_invalidate(sky);
Reflection probes working
2019-09-09 20:50:51 +00:00
if (sky->radiance.is_valid()) {
RD::get_singleton()->free(sky->radiance);
sky->radiance = RID();
}
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
_clear_reflection_data(sky->reflection);
Environment sky more or less working.
2019-08-26 20:43:58 +00:00
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
void RendererSceneRenderRD::sky_set_mode(RID p_sky, RS::SkyMode p_mode) {
Environment sky more or less working.
2019-08-26 20:43:58 +00:00
Sky *sky = sky_owner.getornull(p_sky);
ERR_FAIL_COND(!sky);
if (sky->mode == p_mode) {
return;
}
sky->mode = p_mode;
Basic fast filtering implementation
2020-02-20 23:27:34 +00:00
Renaming of servers for coherency. VisualServer -> RenderingServer PhysicsServer -> PhysicsServer3D Physics2DServer -> PhysicsServer2D NavigationServer -> NavigationServer3D Navigation2DServer -> NavigationServer2D Also renamed corresponding files.
2020-03-27 18:21:27 +00:00
if (sky->mode == RS::SKY_MODE_REALTIME && sky->radiance_size != 256) {
Working sky shader implementation
2020-03-20 00:32:19 +00:00
WARN_PRINT("Realtime Skies can only use a radiance size of 256. Radiance size will be set to 256 internally.");
sky_set_radiance_size(p_sky, 256);
Basic fast filtering implementation
2020-02-20 23:27:34 +00:00
}
Environment sky more or less working.
2019-08-26 20:43:58 +00:00
_sky_invalidate(sky);
Basic fast filtering implementation
2020-02-20 23:27:34 +00:00
if (sky->radiance.is_valid()) {
RD::get_singleton()->free(sky->radiance);
sky->radiance = RID();
}
_clear_reflection_data(sky->reflection);
Environment sky more or less working.
2019-08-26 20:43:58 +00:00
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
void RendererSceneRenderRD::sky_set_material(RID p_sky, RID p_material) {
Environment sky more or less working.
2019-08-26 20:43:58 +00:00
Sky *sky = sky_owner.getornull(p_sky);
ERR_FAIL_COND(!sky);
Working sky shader implementation
2020-03-20 00:32:19 +00:00
sky->material = p_material;
New lightmapper -Added LocalVector (needed it) -Added stb_rect_pack (It's pretty cool, we could probably use it for other stuff too) -Fixes and changes all around the place -Added library for 128 bits fixed point (required for Delaunay3D)
2020-05-01 12:34:23 +00:00
_sky_invalidate(sky);
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
Ref<Image> RendererSceneRenderRD::sky_bake_panorama(RID p_sky, float p_energy, bool p_bake_irradiance, const Size2i &p_size) {
New lightmapper -Added LocalVector (needed it) -Added stb_rect_pack (It's pretty cool, we could probably use it for other stuff too) -Fixes and changes all around the place -Added library for 128 bits fixed point (required for Delaunay3D)
2020-05-01 12:34:23 +00:00
Sky *sky = sky_owner.getornull(p_sky);
ERR_FAIL_COND_V(!sky, Ref<Image>());
_update_dirty_skys();
if (sky->radiance.is_valid()) {
RD::TextureFormat tf;
tf.format = RD::DATA_FORMAT_R32G32B32A32_SFLOAT;
tf.width = p_size.width;
tf.height = p_size.height;
tf.usage_bits = RD::TEXTURE_USAGE_STORAGE_BIT | RD::TEXTURE_USAGE_CAN_COPY_FROM_BIT;
RID rad_tex = RD::get_singleton()->texture_create(tf, RD::TextureView());
storage->get_effects()->copy_cubemap_to_panorama(sky->radiance, rad_tex, p_size, p_bake_irradiance ? roughness_layers : 0, sky->reflection.layers.size() > 1);
Vector<uint8_t> data = RD::get_singleton()->texture_get_data(rad_tex, 0);
RD::get_singleton()->free(rad_tex);
Ref<Image> img;
img.instance();
img->create(p_size.width, p_size.height, false, Image::FORMAT_RGBAF, data);
for (int i = 0; i < p_size.width; i++) {
for (int j = 0; j < p_size.height; j++) {
Color c = img->get_pixel(i, j);
c.r *= p_energy;
c.g *= p_energy;
c.b *= p_energy;
img->set_pixel(i, j, c);
}
}
return img;
}
return Ref<Image>();
Environment sky more or less working.
2019-08-26 20:43:58 +00:00
}
New lightmapper -Added LocalVector (needed it) -Added stb_rect_pack (It's pretty cool, we could probably use it for other stuff too) -Fixes and changes all around the place -Added library for 128 bits fixed point (required for Delaunay3D)
2020-05-01 12:34:23 +00:00
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
void RendererSceneRenderRD::_update_dirty_skys() {
Environment sky more or less working.
2019-08-26 20:43:58 +00:00
Sky *sky = dirty_sky_list;
while (sky) {
Working sky shader implementation
2020-03-20 00:32:19 +00:00
bool texture_set_dirty = false;
Environment sky more or less working.
2019-08-26 20:43:58 +00:00
//update sky configuration if texture is missing
Reflection probes working
2019-09-09 20:50:51 +00:00
if (sky->radiance.is_null()) {
int mipmaps = Image::get_image_required_mipmaps(sky->radiance_size, sky->radiance_size, Image::FORMAT_RGBAH) + 1;
uint32_t w = sky->radiance_size, h = sky->radiance_size;
Working sky shader implementation
2020-03-20 00:32:19 +00:00
int layers = roughness_layers;
Renaming of servers for coherency. VisualServer -> RenderingServer PhysicsServer -> PhysicsServer3D Physics2DServer -> PhysicsServer2D NavigationServer -> NavigationServer3D Navigation2DServer -> NavigationServer2D Also renamed corresponding files.
2020-03-27 18:21:27 +00:00
if (sky->mode == RS::SKY_MODE_REALTIME) {
Working sky shader implementation
2020-03-20 00:32:19 +00:00
layers = 8;
if (roughness_layers != 8) {
WARN_PRINT("When using REALTIME skies, roughness_layers should be set to 8 in the project settings for best quality reflections");
}
}
Reflection probes working
2019-09-09 20:50:51 +00:00
if (sky_use_cubemap_array) {
//array (higher quality, 6 times more memory)
RD::TextureFormat tf;
Basic fast filtering implementation
2020-02-20 23:27:34 +00:00
tf.array_layers = layers * 6;
Reflection probes working
2019-09-09 20:50:51 +00:00
tf.format = RD::DATA_FORMAT_R16G16B16A16_SFLOAT;
Rename RD texture "type" to "texture_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-11-27 03:50:05 +00:00
tf.texture_type = RD::TEXTURE_TYPE_CUBE_ARRAY;
Reflection probes working
2019-09-09 20:50:51 +00:00
tf.mipmaps = mipmaps;
tf.width = w;
tf.height = h;
Basic fast filtering implementation
2020-02-20 23:27:34 +00:00
tf.usage_bits = RD::TEXTURE_USAGE_COLOR_ATTACHMENT_BIT | RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_STORAGE_BIT;
Reflection probes working
2019-09-09 20:50:51 +00:00
sky->radiance = RD::get_singleton()->texture_create(tf, RD::TextureView());
Renaming of servers for coherency. VisualServer -> RenderingServer PhysicsServer -> PhysicsServer3D Physics2DServer -> PhysicsServer2D NavigationServer -> NavigationServer3D Navigation2DServer -> NavigationServer2D Also renamed corresponding files.
2020-03-27 18:21:27 +00:00
_update_reflection_data(sky->reflection, sky->radiance_size, mipmaps, true, sky->radiance, 0, sky->mode == RS::SKY_MODE_REALTIME);
Reflection probes working
2019-09-09 20:50:51 +00:00
} else {
//regular cubemap, lower quality (aliasing, less memory)
RD::TextureFormat tf;
tf.array_layers = 6;
tf.format = RD::DATA_FORMAT_R16G16B16A16_SFLOAT;
Rename RD texture "type" to "texture_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-11-27 03:50:05 +00:00
tf.texture_type = RD::TEXTURE_TYPE_CUBE;
Basic fast filtering implementation
2020-02-20 23:27:34 +00:00
tf.mipmaps = MIN(mipmaps, layers);
Reflection probes working
2019-09-09 20:50:51 +00:00
tf.width = w;
tf.height = h;
Basic fast filtering implementation
2020-02-20 23:27:34 +00:00
tf.usage_bits = RD::TEXTURE_USAGE_COLOR_ATTACHMENT_BIT | RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_STORAGE_BIT;
Reflection probes working
2019-09-09 20:50:51 +00:00
sky->radiance = RD::get_singleton()->texture_create(tf, RD::TextureView());
Renaming of servers for coherency. VisualServer -> RenderingServer PhysicsServer -> PhysicsServer3D Physics2DServer -> PhysicsServer2D NavigationServer -> NavigationServer3D Navigation2DServer -> NavigationServer2D Also renamed corresponding files.
2020-03-27 18:21:27 +00:00
_update_reflection_data(sky->reflection, sky->radiance_size, MIN(mipmaps, layers), false, sky->radiance, 0, sky->mode == RS::SKY_MODE_REALTIME);
Reflection probes working
2019-09-09 20:50:51 +00:00
}
Working sky shader implementation
2020-03-20 00:32:19 +00:00
texture_set_dirty = true;
}
Fix spelling & grammar in comments, docs, and messages
2020-07-16 16:54:15 +00:00
// Create subpass buffers if they haven't been created already
Working sky shader implementation
2020-03-20 00:32:19 +00:00
if (sky->half_res_pass.is_null() && !RD::get_singleton()->texture_is_valid(sky->half_res_pass) && sky->screen_size.x >= 4 && sky->screen_size.y >= 4) {
RD::TextureFormat tformat;
tformat.format = RD::DATA_FORMAT_R16G16B16A16_SFLOAT;
tformat.width = sky->screen_size.x / 2;
tformat.height = sky->screen_size.y / 2;
tformat.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_COLOR_ATTACHMENT_BIT;
Rename RD texture "type" to "texture_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-11-27 03:50:05 +00:00
tformat.texture_type = RD::TEXTURE_TYPE_2D;
Working sky shader implementation
2020-03-20 00:32:19 +00:00
sky->half_res_pass = RD::get_singleton()->texture_create(tformat, RD::TextureView());
Vector<RID> texs;
texs.push_back(sky->half_res_pass);
sky->half_res_framebuffer = RD::get_singleton()->framebuffer_create(texs);
texture_set_dirty = true;
}
if (sky->quarter_res_pass.is_null() && !RD::get_singleton()->texture_is_valid(sky->quarter_res_pass) && sky->screen_size.x >= 4 && sky->screen_size.y >= 4) {
RD::TextureFormat tformat;
tformat.format = RD::DATA_FORMAT_R16G16B16A16_SFLOAT;
tformat.width = sky->screen_size.x / 4;
tformat.height = sky->screen_size.y / 4;
tformat.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_COLOR_ATTACHMENT_BIT;
Rename RD texture "type" to "texture_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-11-27 03:50:05 +00:00
tformat.texture_type = RD::TEXTURE_TYPE_2D;
Working sky shader implementation
2020-03-20 00:32:19 +00:00
sky->quarter_res_pass = RD::get_singleton()->texture_create(tformat, RD::TextureView());
Vector<RID> texs;
texs.push_back(sky->quarter_res_pass);
sky->quarter_res_framebuffer = RD::get_singleton()->framebuffer_create(texs);
texture_set_dirty = true;
}
if (texture_set_dirty) {
for (int i = 0; i < SKY_TEXTURE_SET_MAX; i++) {
if (sky->texture_uniform_sets[i].is_valid() && RD::get_singleton()->uniform_set_is_valid(sky->texture_uniform_sets[i])) {
RD::get_singleton()->free(sky->texture_uniform_sets[i]);
sky->texture_uniform_sets[i] = RID();
}
}
Environment sky more or less working.
2019-08-26 20:43:58 +00:00
}
Working sky shader implementation
2020-03-20 00:32:19 +00:00
sky->reflection.dirty = true;
Add incremental update mode to sky
2020-07-11 00:07:30 +00:00
sky->processing_layer = 0;
Environment sky more or less working.
2019-08-26 20:43:58 +00:00
Sky *next = sky->dirty_list;
sky->dirty_list = nullptr;
sky->dirty = false;
sky = next;
}
dirty_sky_list = nullptr;
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
RID RendererSceneRenderRD::sky_get_radiance_texture_rd(RID p_sky) const {
Environment sky more or less working.
2019-08-26 20:43:58 +00:00
Sky *sky = sky_owner.getornull(p_sky);
ERR_FAIL_COND_V(!sky, RID());
Reflection probes working
2019-09-09 20:50:51 +00:00
return sky->radiance;
Environment sky more or less working.
2019-08-26 20:43:58 +00:00
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
RID RendererSceneRenderRD::sky_get_radiance_uniform_set_rd(RID p_sky, RID p_shader, int p_set) const {
Re-implemented screen space ambient occlusion
2020-01-25 10:18:55 +00:00
Sky *sky = sky_owner.getornull(p_sky);
ERR_FAIL_COND_V(!sky, RID());
if (sky->uniform_set.is_null() || !RD::get_singleton()->uniform_set_is_valid(sky->uniform_set)) {
sky->uniform_set = RID();
if (sky->radiance.is_valid()) {
Vector<RD::Uniform> uniforms;
{
RD::Uniform u;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_TEXTURE;
Re-implemented screen space ambient occlusion
2020-01-25 10:18:55 +00:00
u.binding = 0;
u.ids.push_back(sky->radiance);
uniforms.push_back(u);
}
sky->uniform_set = RD::get_singleton()->uniform_set_create(uniforms, p_shader, p_set);
}
}
return sky->uniform_set;
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
RID RendererSceneRenderRD::_get_sky_textures(Sky *p_sky, SkyTextureSetVersion p_version) {
Working sky shader implementation
2020-03-20 00:32:19 +00:00
if (p_sky->texture_uniform_sets[p_version].is_valid() && RD::get_singleton()->uniform_set_is_valid(p_sky->texture_uniform_sets[p_version])) {
return p_sky->texture_uniform_sets[p_version];
}
Vector<RD::Uniform> uniforms;
{
RD::Uniform u;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_TEXTURE;
Working sky shader implementation
2020-03-20 00:32:19 +00:00
u.binding = 0;
if (p_sky->radiance.is_valid() && p_version <= SKY_TEXTURE_SET_QUARTER_RES) {
u.ids.push_back(p_sky->radiance);
} else {
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
u.ids.push_back(storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_CUBEMAP_BLACK));
Working sky shader implementation
2020-03-20 00:32:19 +00:00
}
uniforms.push_back(u);
}
{
RD::Uniform u;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_TEXTURE;
Working sky shader implementation
2020-03-20 00:32:19 +00:00
u.binding = 1; // half res
Replace subpass textures with color in sky shader
2020-03-24 05:55:11 +00:00
if (p_sky->half_res_pass.is_valid() && p_version != SKY_TEXTURE_SET_HALF_RES && p_version != SKY_TEXTURE_SET_CUBEMAP_HALF_RES) {
if (p_version >= SKY_TEXTURE_SET_CUBEMAP) {
u.ids.push_back(p_sky->reflection.layers[0].views[1]);
Working sky shader implementation
2020-03-20 00:32:19 +00:00
} else {
u.ids.push_back(p_sky->half_res_pass);
}
} else {
Replace subpass textures with color in sky shader
2020-03-24 05:55:11 +00:00
if (p_version < SKY_TEXTURE_SET_CUBEMAP) {
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
u.ids.push_back(storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_WHITE));
Replace subpass textures with color in sky shader
2020-03-24 05:55:11 +00:00
} else {
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
u.ids.push_back(storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_CUBEMAP_BLACK));
Replace subpass textures with color in sky shader
2020-03-24 05:55:11 +00:00
}
Working sky shader implementation
2020-03-20 00:32:19 +00:00
}
uniforms.push_back(u);
}
{
RD::Uniform u;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_TEXTURE;
Working sky shader implementation
2020-03-20 00:32:19 +00:00
u.binding = 2; // quarter res
Replace subpass textures with color in sky shader
2020-03-24 05:55:11 +00:00
if (p_sky->quarter_res_pass.is_valid() && p_version != SKY_TEXTURE_SET_QUARTER_RES && p_version != SKY_TEXTURE_SET_CUBEMAP_QUARTER_RES) {
if (p_version >= SKY_TEXTURE_SET_CUBEMAP) {
u.ids.push_back(p_sky->reflection.layers[0].views[2]);
Working sky shader implementation
2020-03-20 00:32:19 +00:00
} else {
u.ids.push_back(p_sky->quarter_res_pass);
}
} else {
Replace subpass textures with color in sky shader
2020-03-24 05:55:11 +00:00
if (p_version < SKY_TEXTURE_SET_CUBEMAP) {
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
u.ids.push_back(storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_WHITE));
Replace subpass textures with color in sky shader
2020-03-24 05:55:11 +00:00
} else {
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
u.ids.push_back(storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_CUBEMAP_BLACK));
Replace subpass textures with color in sky shader
2020-03-24 05:55:11 +00:00
}
Working sky shader implementation
2020-03-20 00:32:19 +00:00
}
uniforms.push_back(u);
}
p_sky->texture_uniform_sets[p_version] = RD::get_singleton()->uniform_set_create(uniforms, sky_shader.default_shader_rd, SKY_SET_TEXTURES);
return p_sky->texture_uniform_sets[p_version];
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
RID RendererSceneRenderRD::sky_get_material(RID p_sky) const {
Working sky shader implementation
2020-03-20 00:32:19 +00:00
Sky *sky = sky_owner.getornull(p_sky);
ERR_FAIL_COND_V(!sky, RID());
return sky->material;
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
void RendererSceneRenderRD::_draw_sky(bool p_can_continue_color, bool p_can_continue_depth, RID p_fb, RID p_environment, const CameraMatrix &p_projection, const Transform &p_transform) {
Working sky shader implementation
2020-03-20 00:32:19 +00:00
ERR_FAIL_COND(!is_environment(p_environment));
Add fog to sky shaders
2020-08-18 04:12:51 +00:00
SkyMaterialData *material = nullptr;
Working sky shader implementation
2020-03-20 00:32:19 +00:00
Sky *sky = sky_owner.getornull(environment_get_sky(p_environment));
Add fog to sky shaders
2020-08-18 04:12:51 +00:00
RID sky_material;
Working sky shader implementation
2020-03-20 00:32:19 +00:00
Add fog to sky shaders
2020-08-18 04:12:51 +00:00
RS::EnvironmentBG background = environment_get_background(p_environment);
Working sky shader implementation
2020-03-20 00:32:19 +00:00
Add fog to sky shaders
2020-08-18 04:12:51 +00:00
if (!(background == RS::ENV_BG_CLEAR_COLOR || background == RS::ENV_BG_COLOR) || sky) {
ERR_FAIL_COND(!sky);
sky_material = sky_get_material(environment_get_sky(p_environment));
if (sky_material.is_valid()) {
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
material = (SkyMaterialData *)storage->material_get_data(sky_material, RendererStorageRD::SHADER_TYPE_SKY);
Add fog to sky shaders
2020-08-18 04:12:51 +00:00
if (!material || !material->shader_data->valid) {
material = nullptr;
}
}
if (!material) {
sky_material = sky_shader.default_material;
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
material = (SkyMaterialData *)storage->material_get_data(sky_material, RendererStorageRD::SHADER_TYPE_SKY);
Working sky shader implementation
2020-03-20 00:32:19 +00:00
}
}
Add fog to sky shaders
2020-08-18 04:12:51 +00:00
if (background == RS::ENV_BG_CLEAR_COLOR || background == RS::ENV_BG_COLOR) {
sky_material = sky_scene_state.fog_material;
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
material = (SkyMaterialData *)storage->material_get_data(sky_material, RendererStorageRD::SHADER_TYPE_SKY);
Working sky shader implementation
2020-03-20 00:32:19 +00:00
}
ERR_FAIL_COND(!material);
SkyShaderData *shader_data = material->shader_data;
ERR_FAIL_COND(!shader_data);
Basis sky_transform = environment_get_sky_orientation(p_environment);
sky_transform.invert();
float multiplier = environment_get_bg_energy(p_environment);
float custom_fov = environment_get_sky_custom_fov(p_environment);
// Camera
CameraMatrix camera;
if (custom_fov) {
float near_plane = p_projection.get_z_near();
float far_plane = p_projection.get_z_far();
float aspect = p_projection.get_aspect();
camera.set_perspective(custom_fov, aspect, near_plane, far_plane);
} else {
camera = p_projection;
}
sky_transform = p_transform.basis * sky_transform;
if (shader_data->uses_quarter_res) {
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
PipelineCacheRD *pipeline = &shader_data->pipelines[SKY_VERSION_QUARTER_RES];
Working sky shader implementation
2020-03-20 00:32:19 +00:00
RID texture_uniform_set = _get_sky_textures(sky, SKY_TEXTURE_SET_QUARTER_RES);
Vector<Color> clear_colors;
clear_colors.push_back(Color(0.0, 0.0, 0.0));
RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(sky->quarter_res_framebuffer, RD::INITIAL_ACTION_CLEAR, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_CLEAR, RD::FINAL_ACTION_DISCARD, clear_colors);
Add fog to sky shaders
2020-08-18 04:12:51 +00:00
storage->get_effects()->render_sky(draw_list, time, sky->quarter_res_framebuffer, sky_scene_state.uniform_set, sky_scene_state.fog_uniform_set, pipeline, material->uniform_set, texture_uniform_set, camera, sky_transform, multiplier, p_transform.origin);
Working sky shader implementation
2020-03-20 00:32:19 +00:00
RD::get_singleton()->draw_list_end();
}
if (shader_data->uses_half_res) {
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
PipelineCacheRD *pipeline = &shader_data->pipelines[SKY_VERSION_HALF_RES];
Working sky shader implementation
2020-03-20 00:32:19 +00:00
RID texture_uniform_set = _get_sky_textures(sky, SKY_TEXTURE_SET_HALF_RES);
Vector<Color> clear_colors;
clear_colors.push_back(Color(0.0, 0.0, 0.0));
RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(sky->half_res_framebuffer, RD::INITIAL_ACTION_CLEAR, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_CLEAR, RD::FINAL_ACTION_DISCARD, clear_colors);
Add fog to sky shaders
2020-08-18 04:12:51 +00:00
storage->get_effects()->render_sky(draw_list, time, sky->half_res_framebuffer, sky_scene_state.uniform_set, sky_scene_state.fog_uniform_set, pipeline, material->uniform_set, texture_uniform_set, camera, sky_transform, multiplier, p_transform.origin);
Working sky shader implementation
2020-03-20 00:32:19 +00:00
RD::get_singleton()->draw_list_end();
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
PipelineCacheRD *pipeline = &shader_data->pipelines[SKY_VERSION_BACKGROUND];
Working sky shader implementation
2020-03-20 00:32:19 +00:00
Add fog to sky shaders
2020-08-18 04:12:51 +00:00
RID texture_uniform_set;
if (sky) {
texture_uniform_set = _get_sky_textures(sky, SKY_TEXTURE_SET_BACKGROUND);
} else {
texture_uniform_set = sky_scene_state.fog_only_texture_uniform_set;
}
Working sky shader implementation
2020-03-20 00:32:19 +00:00
Re-Added screen space reflection.
2020-04-02 02:24:52 +00:00
RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(p_fb, RD::INITIAL_ACTION_CONTINUE, p_can_continue_color ? RD::FINAL_ACTION_CONTINUE : RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_CONTINUE, p_can_continue_depth ? RD::FINAL_ACTION_CONTINUE : RD::FINAL_ACTION_READ);
Add fog to sky shaders
2020-08-18 04:12:51 +00:00
storage->get_effects()->render_sky(draw_list, time, p_fb, sky_scene_state.uniform_set, sky_scene_state.fog_uniform_set, pipeline, material->uniform_set, texture_uniform_set, camera, sky_transform, multiplier, p_transform.origin);
Working sky shader implementation
2020-03-20 00:32:19 +00:00
RD::get_singleton()->draw_list_end();
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
void RendererSceneRenderRD::_setup_sky(RID p_environment, RID p_render_buffers, const CameraMatrix &p_projection, const Transform &p_transform, const Size2i p_screen_size) {
Working sky shader implementation
2020-03-20 00:32:19 +00:00
ERR_FAIL_COND(!is_environment(p_environment));
Add fog to sky shaders
2020-08-18 04:12:51 +00:00
SkyMaterialData *material = nullptr;
Working sky shader implementation
2020-03-20 00:32:19 +00:00
Sky *sky = sky_owner.getornull(environment_get_sky(p_environment));
Add fog to sky shaders
2020-08-18 04:12:51 +00:00
RID sky_material;
Working sky shader implementation
2020-03-20 00:32:19 +00:00
Add fog to sky shaders
2020-08-18 04:12:51 +00:00
SkyShaderData *shader_data = nullptr;
Working sky shader implementation
2020-03-20 00:32:19 +00:00
Add fog to sky shaders
2020-08-18 04:12:51 +00:00
RS::EnvironmentBG background = environment_get_background(p_environment);
if (!(background == RS::ENV_BG_CLEAR_COLOR || background == RS::ENV_BG_COLOR) || sky) {
ERR_FAIL_COND(!sky);
sky_material = sky_get_material(environment_get_sky(p_environment));
if (sky_material.is_valid()) {
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
material = (SkyMaterialData *)storage->material_get_data(sky_material, RendererStorageRD::SHADER_TYPE_SKY);
Add fog to sky shaders
2020-08-18 04:12:51 +00:00
if (!material || !material->shader_data->valid) {
material = nullptr;
}
Working sky shader implementation
2020-03-20 00:32:19 +00:00
}
Add fog to sky shaders
2020-08-18 04:12:51 +00:00
if (!material) {
sky_material = sky_shader.default_material;
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
material = (SkyMaterialData *)storage->material_get_data(sky_material, RendererStorageRD::SHADER_TYPE_SKY);
Add fog to sky shaders
2020-08-18 04:12:51 +00:00
}
Working sky shader implementation
2020-03-20 00:32:19 +00:00
Add fog to sky shaders
2020-08-18 04:12:51 +00:00
ERR_FAIL_COND(!material);
Working sky shader implementation
2020-03-20 00:32:19 +00:00
Add fog to sky shaders
2020-08-18 04:12:51 +00:00
shader_data = material->shader_data;
Working sky shader implementation
2020-03-20 00:32:19 +00:00
Add fog to sky shaders
2020-08-18 04:12:51 +00:00
ERR_FAIL_COND(!shader_data);
}
Working sky shader implementation
2020-03-20 00:32:19 +00:00
Add fog to sky shaders
2020-08-18 04:12:51 +00:00
if (sky) {
// Invalidate supbass buffers if screen size changes
if (sky->screen_size != p_screen_size) {
sky->screen_size = p_screen_size;
sky->screen_size.x = sky->screen_size.x < 4 ? 4 : sky->screen_size.x;
sky->screen_size.y = sky->screen_size.y < 4 ? 4 : sky->screen_size.y;
if (shader_data->uses_half_res) {
if (sky->half_res_pass.is_valid()) {
RD::get_singleton()->free(sky->half_res_pass);
sky->half_res_pass = RID();
}
_sky_invalidate(sky);
Working sky shader implementation
2020-03-20 00:32:19 +00:00
}
Add fog to sky shaders
2020-08-18 04:12:51 +00:00
if (shader_data->uses_quarter_res) {
if (sky->quarter_res_pass.is_valid()) {
RD::get_singleton()->free(sky->quarter_res_pass);
sky->quarter_res_pass = RID();
}
_sky_invalidate(sky);
Working sky shader implementation
2020-03-20 00:32:19 +00:00
}
Add fog to sky shaders
2020-08-18 04:12:51 +00:00
}
// Create new subpass buffers if necessary
if ((shader_data->uses_half_res && sky->half_res_pass.is_null()) ||
(shader_data->uses_quarter_res && sky->quarter_res_pass.is_null()) ||
sky->radiance.is_null()) {
Working sky shader implementation
2020-03-20 00:32:19 +00:00
_sky_invalidate(sky);
Add fog to sky shaders
2020-08-18 04:12:51 +00:00
_update_dirty_skys();
Working sky shader implementation
2020-03-20 00:32:19 +00:00
}
Add fog to sky shaders
2020-08-18 04:12:51 +00:00
if (shader_data->uses_time && time - sky->prev_time > 0.00001) {
sky->prev_time = time;
sky->reflection.dirty = true;
Reorganize rendering server. -Made RenderingServerScene abstract, allowing reimplementation -RenderingServerRaster -> RenderingServerDefault, but this class is going away soon.
2020-12-03 21:09:47 +00:00
RenderingServerDefault::redraw_request();
Add fog to sky shaders
2020-08-18 04:12:51 +00:00
}
Working sky shader implementation
2020-03-20 00:32:19 +00:00
Add fog to sky shaders
2020-08-18 04:12:51 +00:00
if (material != sky->prev_material) {
sky->prev_material = material;
sky->reflection.dirty = true;
}
Working sky shader implementation
2020-03-20 00:32:19 +00:00
Add fog to sky shaders
2020-08-18 04:12:51 +00:00
if (material->uniform_set_updated) {
material->uniform_set_updated = false;
sky->reflection.dirty = true;
}
Working sky shader implementation
2020-03-20 00:32:19 +00:00
Add fog to sky shaders
2020-08-18 04:12:51 +00:00
if (!p_transform.origin.is_equal_approx(sky->prev_position) && shader_data->uses_position) {
sky->prev_position = p_transform.origin;
sky->reflection.dirty = true;
}
Working sky shader implementation
2020-03-20 00:32:19 +00:00
Add fog to sky shaders
2020-08-18 04:12:51 +00:00
if (shader_data->uses_light) {
// Check whether the directional_light_buffer changes
bool light_data_dirty = false;
Working sky shader implementation
2020-03-20 00:32:19 +00:00
Add fog to sky shaders
2020-08-18 04:12:51 +00:00
if (sky_scene_state.ubo.directional_light_count != sky_scene_state.last_frame_directional_light_count) {
light_data_dirty = true;
for (uint32_t i = sky_scene_state.ubo.directional_light_count; i < sky_scene_state.max_directional_lights; i++) {
sky_scene_state.directional_lights[i].enabled = false;
}
}
if (!light_data_dirty) {
for (uint32_t i = 0; i < sky_scene_state.ubo.directional_light_count; i++) {
if (sky_scene_state.directional_lights[i].direction[0] != sky_scene_state.last_frame_directional_lights[i].direction[0] ||
sky_scene_state.directional_lights[i].direction[1] != sky_scene_state.last_frame_directional_lights[i].direction[1] ||
sky_scene_state.directional_lights[i].direction[2] != sky_scene_state.last_frame_directional_lights[i].direction[2] ||
sky_scene_state.directional_lights[i].energy != sky_scene_state.last_frame_directional_lights[i].energy ||
sky_scene_state.directional_lights[i].color[0] != sky_scene_state.last_frame_directional_lights[i].color[0] ||
sky_scene_state.directional_lights[i].color[1] != sky_scene_state.last_frame_directional_lights[i].color[1] ||
sky_scene_state.directional_lights[i].color[2] != sky_scene_state.last_frame_directional_lights[i].color[2] ||
sky_scene_state.directional_lights[i].enabled != sky_scene_state.last_frame_directional_lights[i].enabled ||
sky_scene_state.directional_lights[i].size != sky_scene_state.last_frame_directional_lights[i].size) {
light_data_dirty = true;
break;
}
Working sky shader implementation
2020-03-20 00:32:19 +00:00
}
}
Add fog to sky shaders
2020-08-18 04:12:51 +00:00
if (light_data_dirty) {
Reorganize RenderingDevice barriers -Removed sync to draw, now everything syncs to draw by default. -Fixed many validation layer errors. -Added support for VkImageViewUsageCreateInfo to fix validation layer warnings. -Texture, buffer, raster and compute functions now all allow spcifying which barriers will be used.
2021-01-26 00:52:58 +00:00
RD::get_singleton()->buffer_update(sky_scene_state.directional_light_buffer, 0, sizeof(SkyDirectionalLightData) * sky_scene_state.max_directional_lights, sky_scene_state.directional_lights);
Add fog to sky shaders
2020-08-18 04:12:51 +00:00
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
RendererSceneRenderRD::SkyDirectionalLightData *temp = sky_scene_state.last_frame_directional_lights;
Add fog to sky shaders
2020-08-18 04:12:51 +00:00
sky_scene_state.last_frame_directional_lights = sky_scene_state.directional_lights;
sky_scene_state.directional_lights = temp;
sky_scene_state.last_frame_directional_light_count = sky_scene_state.ubo.directional_light_count;
sky->reflection.dirty = true;
}
Working sky shader implementation
2020-03-20 00:32:19 +00:00
}
Add fog to sky shaders
2020-08-18 04:12:51 +00:00
}
Working sky shader implementation
2020-03-20 00:32:19 +00:00
Add fog to sky shaders
2020-08-18 04:12:51 +00:00
//setup fog variables
sky_scene_state.ubo.volumetric_fog_enabled = false;
if (p_render_buffers.is_valid()) {
if (render_buffers_has_volumetric_fog(p_render_buffers)) {
sky_scene_state.ubo.volumetric_fog_enabled = true;
Working sky shader implementation
2020-03-20 00:32:19 +00:00
Add fog to sky shaders
2020-08-18 04:12:51 +00:00
float fog_end = render_buffers_get_volumetric_fog_end(p_render_buffers);
if (fog_end > 0.0) {
sky_scene_state.ubo.volumetric_fog_inv_length = 1.0 / fog_end;
} else {
sky_scene_state.ubo.volumetric_fog_inv_length = 1.0;
Working sky shader implementation
2020-03-20 00:32:19 +00:00
}
Add fog to sky shaders
2020-08-18 04:12:51 +00:00
float fog_detail_spread = render_buffers_get_volumetric_fog_detail_spread(p_render_buffers); //reverse lookup
if (fog_detail_spread > 0.0) {
sky_scene_state.ubo.volumetric_fog_detail_spread = 1.0 / fog_detail_spread;
} else {
sky_scene_state.ubo.volumetric_fog_detail_spread = 1.0;
Working sky shader implementation
2020-03-20 00:32:19 +00:00
}
Add fog to sky shaders
2020-08-18 04:12:51 +00:00
}
Working sky shader implementation
2020-03-20 00:32:19 +00:00
Add fog to sky shaders
2020-08-18 04:12:51 +00:00
RID fog_uniform_set = render_buffers_get_volumetric_fog_sky_uniform_set(p_render_buffers);
Working sky shader implementation
2020-03-20 00:32:19 +00:00
Add fog to sky shaders
2020-08-18 04:12:51 +00:00
if (fog_uniform_set != RID()) {
sky_scene_state.fog_uniform_set = fog_uniform_set;
} else {
sky_scene_state.fog_uniform_set = sky_scene_state.default_fog_uniform_set;
Working sky shader implementation
2020-03-20 00:32:19 +00:00
}
}
Add fog to sky shaders
2020-08-18 04:12:51 +00:00
sky_scene_state.ubo.z_far = p_projection.get_z_far();
sky_scene_state.ubo.fog_enabled = environment_is_fog_enabled(p_environment);
sky_scene_state.ubo.fog_density = environment_get_fog_density(p_environment);
Add aerial perspective to fixed fog
2020-08-21 05:48:04 +00:00
sky_scene_state.ubo.fog_aerial_perspective = environment_get_fog_aerial_perspective(p_environment);
Add fog to sky shaders
2020-08-18 04:12:51 +00:00
Color fog_color = environment_get_fog_light_color(p_environment).to_linear();
float fog_energy = environment_get_fog_light_energy(p_environment);
sky_scene_state.ubo.fog_light_color[0] = fog_color.r * fog_energy;
sky_scene_state.ubo.fog_light_color[1] = fog_color.g * fog_energy;
sky_scene_state.ubo.fog_light_color[2] = fog_color.b * fog_energy;
sky_scene_state.ubo.fog_sun_scatter = environment_get_fog_sun_scatter(p_environment);
Reorganize RenderingDevice barriers -Removed sync to draw, now everything syncs to draw by default. -Fixed many validation layer errors. -Added support for VkImageViewUsageCreateInfo to fix validation layer warnings. -Texture, buffer, raster and compute functions now all allow spcifying which barriers will be used.
2021-01-26 00:52:58 +00:00
RD::get_singleton()->buffer_update(sky_scene_state.uniform_buffer, 0, sizeof(SkySceneState::UBO), &sky_scene_state.ubo);
Working sky shader implementation
2020-03-20 00:32:19 +00:00
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
void RendererSceneRenderRD::_update_sky(RID p_environment, const CameraMatrix &p_projection, const Transform &p_transform) {
Working sky shader implementation
2020-03-20 00:32:19 +00:00
ERR_FAIL_COND(!is_environment(p_environment));
Sky *sky = sky_owner.getornull(environment_get_sky(p_environment));
ERR_FAIL_COND(!sky);
RID sky_material = sky_get_material(environment_get_sky(p_environment));
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
SkyMaterialData *material = nullptr;
Working sky shader implementation
2020-03-20 00:32:19 +00:00
if (sky_material.is_valid()) {
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
material = (SkyMaterialData *)storage->material_get_data(sky_material, RendererStorageRD::SHADER_TYPE_SKY);
Working sky shader implementation
2020-03-20 00:32:19 +00:00
if (!material || !material->shader_data->valid) {
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
material = nullptr;
Working sky shader implementation
2020-03-20 00:32:19 +00:00
}
}
if (!material) {
sky_material = sky_shader.default_material;
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
material = (SkyMaterialData *)storage->material_get_data(sky_material, RendererStorageRD::SHADER_TYPE_SKY);
Working sky shader implementation
2020-03-20 00:32:19 +00:00
}
ERR_FAIL_COND(!material);
SkyShaderData *shader_data = material->shader_data;
ERR_FAIL_COND(!shader_data);
float multiplier = environment_get_bg_energy(p_environment);
Add incremental update mode to sky
2020-07-11 00:07:30 +00:00
bool update_single_frame = sky->mode == RS::SKY_MODE_REALTIME || sky->mode == RS::SKY_MODE_QUALITY;
RS::SkyMode sky_mode = sky->mode;
if (sky_mode == RS::SKY_MODE_AUTOMATIC) {
if (shader_data->uses_time || shader_data->uses_position) {
update_single_frame = true;
sky_mode = RS::SKY_MODE_REALTIME;
} else if (shader_data->uses_light || shader_data->ubo_size > 0) {
update_single_frame = false;
sky_mode = RS::SKY_MODE_INCREMENTAL;
} else {
update_single_frame = true;
sky_mode = RS::SKY_MODE_QUALITY;
}
}
if (sky->processing_layer == 0 && sky_mode == RS::SKY_MODE_INCREMENTAL) {
// On the first frame after creating sky, rebuild in single frame
update_single_frame = true;
sky_mode = RS::SKY_MODE_QUALITY;
}
int max_processing_layer = sky_use_cubemap_array ? sky->reflection.layers.size() : sky->reflection.layers[0].mipmaps.size();
Working sky shader implementation
2020-03-20 00:32:19 +00:00
// Update radiance cubemap
Add incremental update mode to sky
2020-07-11 00:07:30 +00:00
if (sky->reflection.dirty && (sky->processing_layer >= max_processing_layer || update_single_frame)) {
Working sky shader implementation
2020-03-20 00:32:19 +00:00
static const Vector3 view_normals[6] = {
Vector3(+1, 0, 0),
Vector3(-1, 0, 0),
Vector3(0, +1, 0),
Vector3(0, -1, 0),
Vector3(0, 0, +1),
Vector3(0, 0, -1)
};
static const Vector3 view_up[6] = {
Vector3(0, -1, 0),
Vector3(0, -1, 0),
Vector3(0, 0, +1),
Vector3(0, 0, -1),
Vector3(0, -1, 0),
Vector3(0, -1, 0)
};
CameraMatrix cm;
cm.set_perspective(90, 1, 0.01, 10.0);
CameraMatrix correction;
correction.set_depth_correction(true);
cm = correction * cm;
if (shader_data->uses_quarter_res) {
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
PipelineCacheRD *pipeline = &shader_data->pipelines[SKY_VERSION_CUBEMAP_QUARTER_RES];
Working sky shader implementation
2020-03-20 00:32:19 +00:00
Vector<Color> clear_colors;
clear_colors.push_back(Color(0.0, 0.0, 0.0));
RD::DrawListID cubemap_draw_list;
for (int i = 0; i < 6; i++) {
Transform local_view;
local_view.set_look_at(Vector3(0, 0, 0), view_normals[i], view_up[i]);
Replace subpass textures with color in sky shader
2020-03-24 05:55:11 +00:00
RID texture_uniform_set = _get_sky_textures(sky, SKY_TEXTURE_SET_CUBEMAP_QUARTER_RES);
Working sky shader implementation
2020-03-20 00:32:19 +00:00
cubemap_draw_list = RD::get_singleton()->draw_list_begin(sky->reflection.layers[0].mipmaps[2].framebuffers[i], RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_DISCARD);
Add fog to sky shaders
2020-08-18 04:12:51 +00:00
storage->get_effects()->render_sky(cubemap_draw_list, time, sky->reflection.layers[0].mipmaps[2].framebuffers[i], sky_scene_state.uniform_set, sky_scene_state.fog_uniform_set, pipeline, material->uniform_set, texture_uniform_set, cm, local_view.basis, multiplier, p_transform.origin);
Working sky shader implementation
2020-03-20 00:32:19 +00:00
RD::get_singleton()->draw_list_end();
}
}
if (shader_data->uses_half_res) {
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
PipelineCacheRD *pipeline = &shader_data->pipelines[SKY_VERSION_CUBEMAP_HALF_RES];
Working sky shader implementation
2020-03-20 00:32:19 +00:00
Vector<Color> clear_colors;
clear_colors.push_back(Color(0.0, 0.0, 0.0));
RD::DrawListID cubemap_draw_list;
for (int i = 0; i < 6; i++) {
Transform local_view;
local_view.set_look_at(Vector3(0, 0, 0), view_normals[i], view_up[i]);
Replace subpass textures with color in sky shader
2020-03-24 05:55:11 +00:00
RID texture_uniform_set = _get_sky_textures(sky, SKY_TEXTURE_SET_CUBEMAP_HALF_RES);
Working sky shader implementation
2020-03-20 00:32:19 +00:00
cubemap_draw_list = RD::get_singleton()->draw_list_begin(sky->reflection.layers[0].mipmaps[1].framebuffers[i], RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_DISCARD);
Add fog to sky shaders
2020-08-18 04:12:51 +00:00
storage->get_effects()->render_sky(cubemap_draw_list, time, sky->reflection.layers[0].mipmaps[1].framebuffers[i], sky_scene_state.uniform_set, sky_scene_state.fog_uniform_set, pipeline, material->uniform_set, texture_uniform_set, cm, local_view.basis, multiplier, p_transform.origin);
Working sky shader implementation
2020-03-20 00:32:19 +00:00
RD::get_singleton()->draw_list_end();
}
}
RD::DrawListID cubemap_draw_list;
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
PipelineCacheRD *pipeline = &shader_data->pipelines[SKY_VERSION_CUBEMAP];
Working sky shader implementation
2020-03-20 00:32:19 +00:00
for (int i = 0; i < 6; i++) {
Transform local_view;
local_view.set_look_at(Vector3(0, 0, 0), view_normals[i], view_up[i]);
Replace subpass textures with color in sky shader
2020-03-24 05:55:11 +00:00
RID texture_uniform_set = _get_sky_textures(sky, SKY_TEXTURE_SET_CUBEMAP);
Working sky shader implementation
2020-03-20 00:32:19 +00:00
cubemap_draw_list = RD::get_singleton()->draw_list_begin(sky->reflection.layers[0].mipmaps[0].framebuffers[i], RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_DISCARD);
Add fog to sky shaders
2020-08-18 04:12:51 +00:00
storage->get_effects()->render_sky(cubemap_draw_list, time, sky->reflection.layers[0].mipmaps[0].framebuffers[i], sky_scene_state.uniform_set, sky_scene_state.fog_uniform_set, pipeline, material->uniform_set, texture_uniform_set, cm, local_view.basis, multiplier, p_transform.origin);
Working sky shader implementation
2020-03-20 00:32:19 +00:00
RD::get_singleton()->draw_list_end();
}
Add incremental update mode to sky
2020-07-11 00:07:30 +00:00
if (sky_mode == RS::SKY_MODE_REALTIME) {
_create_reflection_fast_filter(sky->reflection, sky_use_cubemap_array);
if (sky_use_cubemap_array) {
_update_reflection_mipmaps(sky->reflection, 0, sky->reflection.layers.size());
}
Working sky shader implementation
2020-03-20 00:32:19 +00:00
} else {
Add incremental update mode to sky
2020-07-11 00:07:30 +00:00
if (update_single_frame) {
for (int i = 1; i < max_processing_layer; i++) {
Working sky shader implementation
2020-03-20 00:32:19 +00:00
_create_reflection_importance_sample(sky->reflection, sky_use_cubemap_array, 10, i);
}
Add incremental update mode to sky
2020-07-11 00:07:30 +00:00
if (sky_use_cubemap_array) {
_update_reflection_mipmaps(sky->reflection, 0, sky->reflection.layers.size());
}
Working sky shader implementation
2020-03-20 00:32:19 +00:00
} else {
Add incremental update mode to sky
2020-07-11 00:07:30 +00:00
if (sky_use_cubemap_array) {
// Multi-Frame so just update the first array level
_update_reflection_mipmaps(sky->reflection, 0, 1);
}
Working sky shader implementation
2020-03-20 00:32:19 +00:00
}
Add incremental update mode to sky
2020-07-11 00:07:30 +00:00
sky->processing_layer = 1;
Working sky shader implementation
2020-03-20 00:32:19 +00:00
}
sky->reflection.dirty = false;
Add incremental update mode to sky
2020-07-11 00:07:30 +00:00
} else {
if (sky_mode == RS::SKY_MODE_INCREMENTAL && sky->processing_layer < max_processing_layer) {
_create_reflection_importance_sample(sky->reflection, sky_use_cubemap_array, 10, sky->processing_layer);
if (sky_use_cubemap_array) {
_update_reflection_mipmaps(sky->reflection, sky->processing_layer, sky->processing_layer + 1);
}
sky->processing_layer++;
}
Working sky shader implementation
2020-03-20 00:32:19 +00:00
}
}
/* SKY SHADER */
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
void RendererSceneRenderRD::SkyShaderData::set_code(const String &p_code) {
Working sky shader implementation
2020-03-20 00:32:19 +00:00
//compile
code = p_code;
valid = false;
ubo_size = 0;
uniforms.clear();
if (code == String()) {
return; //just invalid, but no error
}
ShaderCompilerRD::GeneratedCode gen_code;
ShaderCompilerRD::IdentifierActions actions;
uses_time = false;
uses_half_res = false;
uses_quarter_res = false;
uses_position = false;
uses_light = false;
actions.render_mode_flags["use_half_res_pass"] = &uses_half_res;
actions.render_mode_flags["use_quarter_res_pass"] = &uses_quarter_res;
actions.usage_flag_pointers["TIME"] = &uses_time;
actions.usage_flag_pointers["POSITION"] = &uses_position;
actions.usage_flag_pointers["LIGHT0_ENABLED"] = &uses_light;
actions.usage_flag_pointers["LIGHT0_ENERGY"] = &uses_light;
actions.usage_flag_pointers["LIGHT0_DIRECTION"] = &uses_light;
actions.usage_flag_pointers["LIGHT0_COLOR"] = &uses_light;
Add light size to Sky Shaders
2020-04-17 06:38:22 +00:00
actions.usage_flag_pointers["LIGHT0_SIZE"] = &uses_light;
Working sky shader implementation
2020-03-20 00:32:19 +00:00
actions.usage_flag_pointers["LIGHT1_ENABLED"] = &uses_light;
actions.usage_flag_pointers["LIGHT1_ENERGY"] = &uses_light;
actions.usage_flag_pointers["LIGHT1_DIRECTION"] = &uses_light;
actions.usage_flag_pointers["LIGHT1_COLOR"] = &uses_light;
Add light size to Sky Shaders
2020-04-17 06:38:22 +00:00
actions.usage_flag_pointers["LIGHT1_SIZE"] = &uses_light;
Working sky shader implementation
2020-03-20 00:32:19 +00:00
actions.usage_flag_pointers["LIGHT2_ENABLED"] = &uses_light;
actions.usage_flag_pointers["LIGHT2_ENERGY"] = &uses_light;
actions.usage_flag_pointers["LIGHT2_DIRECTION"] = &uses_light;
actions.usage_flag_pointers["LIGHT2_COLOR"] = &uses_light;
Add light size to Sky Shaders
2020-04-17 06:38:22 +00:00
actions.usage_flag_pointers["LIGHT2_SIZE"] = &uses_light;
Working sky shader implementation
2020-03-20 00:32:19 +00:00
actions.usage_flag_pointers["LIGHT3_ENABLED"] = &uses_light;
actions.usage_flag_pointers["LIGHT3_ENERGY"] = &uses_light;
actions.usage_flag_pointers["LIGHT3_DIRECTION"] = &uses_light;
actions.usage_flag_pointers["LIGHT3_COLOR"] = &uses_light;
Add light size to Sky Shaders
2020-04-17 06:38:22 +00:00
actions.usage_flag_pointers["LIGHT3_SIZE"] = &uses_light;
Working sky shader implementation
2020-03-20 00:32:19 +00:00
actions.uniforms = &uniforms;
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
RendererSceneRenderRD *scene_singleton = (RendererSceneRenderRD *)RendererSceneRenderRD::singleton;
Working sky shader implementation
2020-03-20 00:32:19 +00:00
Renaming of servers for coherency. VisualServer -> RenderingServer PhysicsServer -> PhysicsServer3D Physics2DServer -> PhysicsServer2D NavigationServer -> NavigationServer3D Navigation2DServer -> NavigationServer2D Also renamed corresponding files.
2020-03-27 18:21:27 +00:00
Error err = scene_singleton->sky_shader.compiler.compile(RS::SHADER_SKY, code, &actions, path, gen_code);
Working sky shader implementation
2020-03-20 00:32:19 +00:00
ERR_FAIL_COND(err != OK);
if (version.is_null()) {
version = scene_singleton->sky_shader.shader.version_create();
}
#if 0
print_line("**compiling shader:");
print_line("**defines:\n");
for (int i = 0; i < gen_code.defines.size(); i++) {
print_line(gen_code.defines[i]);
}
print_line("\n**uniforms:\n" + gen_code.uniforms);
// print_line("\n**vertex_globals:\n" + gen_code.vertex_global);
// print_line("\n**vertex_code:\n" + gen_code.vertex);
print_line("\n**fragment_globals:\n" + gen_code.fragment_global);
print_line("\n**fragment_code:\n" + gen_code.fragment);
print_line("\n**light_code:\n" + gen_code.light);
#endif
scene_singleton->sky_shader.shader.version_set_code(version, gen_code.uniforms, gen_code.vertex_global, gen_code.vertex, gen_code.fragment_global, gen_code.light, gen_code.fragment, gen_code.defines);
ERR_FAIL_COND(!scene_singleton->sky_shader.shader.version_is_valid(version));
ubo_size = gen_code.uniform_total_size;
ubo_offsets = gen_code.uniform_offsets;
texture_uniforms = gen_code.texture_uniforms;
//update pipelines
for (int i = 0; i < SKY_VERSION_MAX; i++) {
RD::PipelineDepthStencilState depth_stencil_state;
depth_stencil_state.enable_depth_test = true;
depth_stencil_state.depth_compare_operator = RD::COMPARE_OP_LESS_OR_EQUAL;
RID shader_variant = scene_singleton->sky_shader.shader.version_get_shader(version, i);
pipelines[i].setup(shader_variant, RD::RENDER_PRIMITIVE_TRIANGLES, RD::PipelineRasterizationState(), RD::PipelineMultisampleState(), depth_stencil_state, RD::PipelineColorBlendState::create_disabled(), 0);
}
valid = true;
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
void RendererSceneRenderRD::SkyShaderData::set_default_texture_param(const StringName &p_name, RID p_texture) {
Working sky shader implementation
2020-03-20 00:32:19 +00:00
if (!p_texture.is_valid()) {
default_texture_params.erase(p_name);
} else {
default_texture_params[p_name] = p_texture;
}
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
void RendererSceneRenderRD::SkyShaderData::get_param_list(List<PropertyInfo> *p_param_list) const {
Working sky shader implementation
2020-03-20 00:32:19 +00:00
Map<int, StringName> order;
for (Map<StringName, ShaderLanguage::ShaderNode::Uniform>::Element *E = uniforms.front(); E; E = E->next()) {
Implement global and per instance shader uniforms. Adds two keywords to shader language for uniforms: -'global' -'instance' This allows them to reference values outside the material.
2020-04-17 02:52:00 +00:00
if (E->get().scope == ShaderLanguage::ShaderNode::Uniform::SCOPE_GLOBAL || E->get().scope == ShaderLanguage::ShaderNode::Uniform::SCOPE_INSTANCE) {
continue;
}
Working sky shader implementation
2020-03-20 00:32:19 +00:00
if (E->get().texture_order >= 0) {
order[E->get().texture_order + 100000] = E->key();
} else {
order[E->get().order] = E->key();
}
}
for (Map<int, StringName>::Element *E = order.front(); E; E = E->next()) {
PropertyInfo pi = ShaderLanguage::uniform_to_property_info(uniforms[E->get()]);
pi.name = E->get();
p_param_list->push_back(pi);
}
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
void RendererSceneRenderRD::SkyShaderData::get_instance_param_list(List<RendererStorage::InstanceShaderParam> *p_param_list) const {
Implement global and per instance shader uniforms. Adds two keywords to shader language for uniforms: -'global' -'instance' This allows them to reference values outside the material.
2020-04-17 02:52:00 +00:00
for (Map<StringName, ShaderLanguage::ShaderNode::Uniform>::Element *E = uniforms.front(); E; E = E->next()) {
if (E->get().scope != ShaderLanguage::ShaderNode::Uniform::SCOPE_INSTANCE) {
continue;
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
RendererStorage::InstanceShaderParam p;
Implement global and per instance shader uniforms. Adds two keywords to shader language for uniforms: -'global' -'instance' This allows them to reference values outside the material.
2020-04-17 02:52:00 +00:00
p.info = ShaderLanguage::uniform_to_property_info(E->get());
p.info.name = E->key(); //supply name
p.index = E->get().instance_index;
p.default_value = ShaderLanguage::constant_value_to_variant(E->get().default_value, E->get().type, E->get().hint);
p_param_list->push_back(p);
}
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
bool RendererSceneRenderRD::SkyShaderData::is_param_texture(const StringName &p_param) const {
Working sky shader implementation
2020-03-20 00:32:19 +00:00
if (!uniforms.has(p_param)) {
return false;
}
return uniforms[p_param].texture_order >= 0;
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
bool RendererSceneRenderRD::SkyShaderData::is_animated() const {
Working sky shader implementation
2020-03-20 00:32:19 +00:00
return false;
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
bool RendererSceneRenderRD::SkyShaderData::casts_shadows() const {
Working sky shader implementation
2020-03-20 00:32:19 +00:00
return false;
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
Variant RendererSceneRenderRD::SkyShaderData::get_default_parameter(const StringName &p_parameter) const {
Working sky shader implementation
2020-03-20 00:32:19 +00:00
if (uniforms.has(p_parameter)) {
ShaderLanguage::ShaderNode::Uniform uniform = uniforms[p_parameter];
Vector<ShaderLanguage::ConstantNode::Value> default_value = uniform.default_value;
return ShaderLanguage::constant_value_to_variant(default_value, uniform.type, uniform.hint);
}
return Variant();
}
Added ability to visualize native shaders
2021-01-05 23:01:50 +00:00
RS::ShaderNativeSourceCode RendererSceneRenderRD::SkyShaderData::get_native_source_code() const {
RendererSceneRenderRD *scene_singleton = (RendererSceneRenderRD *)RendererSceneRenderRD::singleton;
return scene_singleton->sky_shader.shader.version_get_native_source_code(version);
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
RendererSceneRenderRD::SkyShaderData::SkyShaderData() {
Working sky shader implementation
2020-03-20 00:32:19 +00:00
valid = false;
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
RendererSceneRenderRD::SkyShaderData::~SkyShaderData() {
RendererSceneRenderRD *scene_singleton = (RendererSceneRenderRD *)RendererSceneRenderRD::singleton;
Working sky shader implementation
2020-03-20 00:32:19 +00:00
ERR_FAIL_COND(!scene_singleton);
//pipeline variants will clear themselves if shader is gone
if (version.is_valid()) {
scene_singleton->sky_shader.shader.version_free(version);
}
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
RendererStorageRD::ShaderData *RendererSceneRenderRD::_create_sky_shader_func() {
Working sky shader implementation
2020-03-20 00:32:19 +00:00
SkyShaderData *shader_data = memnew(SkyShaderData);
return shader_data;
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
void RendererSceneRenderRD::SkyMaterialData::update_parameters(const Map<StringName, Variant> &p_parameters, bool p_uniform_dirty, bool p_textures_dirty) {
RendererSceneRenderRD *scene_singleton = (RendererSceneRenderRD *)RendererSceneRenderRD::singleton;
Working sky shader implementation
2020-03-20 00:32:19 +00:00
uniform_set_updated = true;
if ((uint32_t)ubo_data.size() != shader_data->ubo_size) {
p_uniform_dirty = true;
if (uniform_buffer.is_valid()) {
RD::get_singleton()->free(uniform_buffer);
uniform_buffer = RID();
}
ubo_data.resize(shader_data->ubo_size);
if (ubo_data.size()) {
uniform_buffer = RD::get_singleton()->uniform_buffer_create(ubo_data.size());
memset(ubo_data.ptrw(), 0, ubo_data.size()); //clear
}
//clear previous uniform set
if (uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(uniform_set)) {
RD::get_singleton()->free(uniform_set);
uniform_set = RID();
}
}
//check whether buffer changed
if (p_uniform_dirty && ubo_data.size()) {
update_uniform_buffer(shader_data->uniforms, shader_data->ubo_offsets.ptr(), p_parameters, ubo_data.ptrw(), ubo_data.size(), false);
RD::get_singleton()->buffer_update(uniform_buffer, 0, ubo_data.size(), ubo_data.ptrw());
}
uint32_t tex_uniform_count = shader_data->texture_uniforms.size();
if ((uint32_t)texture_cache.size() != tex_uniform_count) {
texture_cache.resize(tex_uniform_count);
p_textures_dirty = true;
//clear previous uniform set
if (uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(uniform_set)) {
RD::get_singleton()->free(uniform_set);
uniform_set = RID();
}
}
if (p_textures_dirty && tex_uniform_count) {
update_textures(p_parameters, shader_data->default_texture_params, shader_data->texture_uniforms, texture_cache.ptrw(), true);
}
if (shader_data->ubo_size == 0 && shader_data->texture_uniforms.size() == 0) {
// This material does not require an uniform set, so don't create it.
return;
}
if (!p_textures_dirty && uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(uniform_set)) {
//no reason to update uniform set, only UBO (or nothing) was needed to update
return;
}
Vector<RD::Uniform> uniforms;
{
if (shader_data->ubo_size) {
RD::Uniform u;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER;
Working sky shader implementation
2020-03-20 00:32:19 +00:00
u.binding = 0;
u.ids.push_back(uniform_buffer);
uniforms.push_back(u);
}
const RID *textures = texture_cache.ptrw();
for (uint32_t i = 0; i < tex_uniform_count; i++) {
RD::Uniform u;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_TEXTURE;
Working sky shader implementation
2020-03-20 00:32:19 +00:00
u.binding = 1 + i;
u.ids.push_back(textures[i]);
uniforms.push_back(u);
}
}
uniform_set = RD::get_singleton()->uniform_set_create(uniforms, scene_singleton->sky_shader.shader.version_get_shader(shader_data->version, 0), SKY_SET_MATERIAL);
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
RendererSceneRenderRD::SkyMaterialData::~SkyMaterialData() {
Working sky shader implementation
2020-03-20 00:32:19 +00:00
if (uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(uniform_set)) {
RD::get_singleton()->free(uniform_set);
}
if (uniform_buffer.is_valid()) {
RD::get_singleton()->free(uniform_buffer);
}
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
RendererStorageRD::MaterialData *RendererSceneRenderRD::_create_sky_material_func(SkyShaderData *p_shader) {
Working sky shader implementation
2020-03-20 00:32:19 +00:00
SkyMaterialData *material_data = memnew(SkyMaterialData);
material_data->shader_data = p_shader;
material_data->last_frame = false;
//update will happen later anyway so do nothing.
return material_data;
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
RID RendererSceneRenderRD::environment_create() {
Fix spelling of a var, a struct, and message output
2020-07-17 16:02:06 +00:00
return environment_owner.make_rid(Environment());
Environment sky more or less working.
2019-08-26 20:43:58 +00:00
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
void RendererSceneRenderRD::environment_set_background(RID p_env, RS::EnvironmentBG p_bg) {
Fix spelling of a var, a struct, and message output
2020-07-17 16:02:06 +00:00
Environment *env = environment_owner.getornull(p_env);
Environment sky more or less working.
2019-08-26 20:43:58 +00:00
ERR_FAIL_COND(!env);
env->background = p_bg;
}
Style: Enforce separation line between function definitions I couldn't find a tool that enforces it, so I went the manual route: ``` find -name "thirdparty" -prune \ -o -name "*.cpp" -o -name "*.h" -o -name "*.m" -o -name "*.mm" \ -o -name "*.glsl" > files perl -0777 -pi -e 's/\n}\n([^#])/\n}\n\n\1/g' $(cat files) misc/scripts/fix_style.sh -c ``` This adds a newline after all `}` on the first column, unless they are followed by `#` (typically `#endif`). This leads to having lots of places with two lines between function/class definitions, but clang-format then fixes it as we enforce max one line of separation. This doesn't fix potential occurrences of function definitions which are indented (e.g. for a helper class defined in a .cpp), but it's better than nothing. Also can't be made to run easily on CI/hooks so we'll have to be careful with new code. Part of #33027.
2020-05-14 12:29:06 +00:00
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
void RendererSceneRenderRD::environment_set_sky(RID p_env, RID p_sky) {
Fix spelling of a var, a struct, and message output
2020-07-17 16:02:06 +00:00
Environment *env = environment_owner.getornull(p_env);
Environment sky more or less working.
2019-08-26 20:43:58 +00:00
ERR_FAIL_COND(!env);
env->sky = p_sky;
}
Style: Enforce separation line between function definitions I couldn't find a tool that enforces it, so I went the manual route: ``` find -name "thirdparty" -prune \ -o -name "*.cpp" -o -name "*.h" -o -name "*.m" -o -name "*.mm" \ -o -name "*.glsl" > files perl -0777 -pi -e 's/\n}\n([^#])/\n}\n\n\1/g' $(cat files) misc/scripts/fix_style.sh -c ``` This adds a newline after all `}` on the first column, unless they are followed by `#` (typically `#endif`). This leads to having lots of places with two lines between function/class definitions, but clang-format then fixes it as we enforce max one line of separation. This doesn't fix potential occurrences of function definitions which are indented (e.g. for a helper class defined in a .cpp), but it's better than nothing. Also can't be made to run easily on CI/hooks so we'll have to be careful with new code. Part of #33027.
2020-05-14 12:29:06 +00:00
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
void RendererSceneRenderRD::environment_set_sky_custom_fov(RID p_env, float p_scale) {
Fix spelling of a var, a struct, and message output
2020-07-17 16:02:06 +00:00
Environment *env = environment_owner.getornull(p_env);
Environment sky more or less working.
2019-08-26 20:43:58 +00:00
ERR_FAIL_COND(!env);
env->sky_custom_fov = p_scale;
}
Style: Enforce separation line between function definitions I couldn't find a tool that enforces it, so I went the manual route: ``` find -name "thirdparty" -prune \ -o -name "*.cpp" -o -name "*.h" -o -name "*.m" -o -name "*.mm" \ -o -name "*.glsl" > files perl -0777 -pi -e 's/\n}\n([^#])/\n}\n\n\1/g' $(cat files) misc/scripts/fix_style.sh -c ``` This adds a newline after all `}` on the first column, unless they are followed by `#` (typically `#endif`). This leads to having lots of places with two lines between function/class definitions, but clang-format then fixes it as we enforce max one line of separation. This doesn't fix potential occurrences of function definitions which are indented (e.g. for a helper class defined in a .cpp), but it's better than nothing. Also can't be made to run easily on CI/hooks so we'll have to be careful with new code. Part of #33027.
2020-05-14 12:29:06 +00:00
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
void RendererSceneRenderRD::environment_set_sky_orientation(RID p_env, const Basis &p_orientation) {
Fix spelling of a var, a struct, and message output
2020-07-17 16:02:06 +00:00
Environment *env = environment_owner.getornull(p_env);
Environment sky more or less working.
2019-08-26 20:43:58 +00:00
ERR_FAIL_COND(!env);
env->sky_orientation = p_orientation;
}
Style: Enforce separation line between function definitions I couldn't find a tool that enforces it, so I went the manual route: ``` find -name "thirdparty" -prune \ -o -name "*.cpp" -o -name "*.h" -o -name "*.m" -o -name "*.mm" \ -o -name "*.glsl" > files perl -0777 -pi -e 's/\n}\n([^#])/\n}\n\n\1/g' $(cat files) misc/scripts/fix_style.sh -c ``` This adds a newline after all `}` on the first column, unless they are followed by `#` (typically `#endif`). This leads to having lots of places with two lines between function/class definitions, but clang-format then fixes it as we enforce max one line of separation. This doesn't fix potential occurrences of function definitions which are indented (e.g. for a helper class defined in a .cpp), but it's better than nothing. Also can't be made to run easily on CI/hooks so we'll have to be careful with new code. Part of #33027.
2020-05-14 12:29:06 +00:00
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
void RendererSceneRenderRD::environment_set_bg_color(RID p_env, const Color &p_color) {
Fix spelling of a var, a struct, and message output
2020-07-17 16:02:06 +00:00
Environment *env = environment_owner.getornull(p_env);
Environment sky more or less working.
2019-08-26 20:43:58 +00:00
ERR_FAIL_COND(!env);
env->bg_color = p_color;
}
Style: Enforce separation line between function definitions I couldn't find a tool that enforces it, so I went the manual route: ``` find -name "thirdparty" -prune \ -o -name "*.cpp" -o -name "*.h" -o -name "*.m" -o -name "*.mm" \ -o -name "*.glsl" > files perl -0777 -pi -e 's/\n}\n([^#])/\n}\n\n\1/g' $(cat files) misc/scripts/fix_style.sh -c ``` This adds a newline after all `}` on the first column, unless they are followed by `#` (typically `#endif`). This leads to having lots of places with two lines between function/class definitions, but clang-format then fixes it as we enforce max one line of separation. This doesn't fix potential occurrences of function definitions which are indented (e.g. for a helper class defined in a .cpp), but it's better than nothing. Also can't be made to run easily on CI/hooks so we'll have to be careful with new code. Part of #33027.
2020-05-14 12:29:06 +00:00
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
void RendererSceneRenderRD::environment_set_bg_energy(RID p_env, float p_energy) {
Fix spelling of a var, a struct, and message output
2020-07-17 16:02:06 +00:00
Environment *env = environment_owner.getornull(p_env);
Environment sky more or less working.
2019-08-26 20:43:58 +00:00
ERR_FAIL_COND(!env);
env->bg_energy = p_energy;
}
Style: Enforce separation line between function definitions I couldn't find a tool that enforces it, so I went the manual route: ``` find -name "thirdparty" -prune \ -o -name "*.cpp" -o -name "*.h" -o -name "*.m" -o -name "*.mm" \ -o -name "*.glsl" > files perl -0777 -pi -e 's/\n}\n([^#])/\n}\n\n\1/g' $(cat files) misc/scripts/fix_style.sh -c ``` This adds a newline after all `}` on the first column, unless they are followed by `#` (typically `#endif`). This leads to having lots of places with two lines between function/class definitions, but clang-format then fixes it as we enforce max one line of separation. This doesn't fix potential occurrences of function definitions which are indented (e.g. for a helper class defined in a .cpp), but it's better than nothing. Also can't be made to run easily on CI/hooks so we'll have to be careful with new code. Part of #33027.
2020-05-14 12:29:06 +00:00
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
void RendererSceneRenderRD::environment_set_canvas_max_layer(RID p_env, int p_max_layer) {
Fix spelling of a var, a struct, and message output
2020-07-17 16:02:06 +00:00
Environment *env = environment_owner.getornull(p_env);
Environment sky more or less working.
2019-08-26 20:43:58 +00:00
ERR_FAIL_COND(!env);
env->canvas_max_layer = p_max_layer;
}
Style: Enforce separation line between function definitions I couldn't find a tool that enforces it, so I went the manual route: ``` find -name "thirdparty" -prune \ -o -name "*.cpp" -o -name "*.h" -o -name "*.m" -o -name "*.mm" \ -o -name "*.glsl" > files perl -0777 -pi -e 's/\n}\n([^#])/\n}\n\n\1/g' $(cat files) misc/scripts/fix_style.sh -c ``` This adds a newline after all `}` on the first column, unless they are followed by `#` (typically `#endif`). This leads to having lots of places with two lines between function/class definitions, but clang-format then fixes it as we enforce max one line of separation. This doesn't fix potential occurrences of function definitions which are indented (e.g. for a helper class defined in a .cpp), but it's better than nothing. Also can't be made to run easily on CI/hooks so we'll have to be careful with new code. Part of #33027.
2020-05-14 12:29:06 +00:00
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
void RendererSceneRenderRD::environment_set_ambient_light(RID p_env, const Color &p_color, RS::EnvironmentAmbientSource p_ambient, float p_energy, float p_sky_contribution, RS::EnvironmentReflectionSource p_reflection_source, const Color &p_ao_color) {
Fix spelling of a var, a struct, and message output
2020-07-17 16:02:06 +00:00
Environment *env = environment_owner.getornull(p_env);
Environment sky more or less working.
2019-08-26 20:43:58 +00:00
ERR_FAIL_COND(!env);
env->ambient_light = p_color;
env->ambient_source = p_ambient;
env->ambient_light_energy = p_energy;
env->ambient_sky_contribution = p_sky_contribution;
env->reflection_source = p_reflection_source;
Re-implemented screen space ambient occlusion
2020-01-25 10:18:55 +00:00
env->ao_color = p_ao_color;
Environment sky more or less working.
2019-08-26 20:43:58 +00:00
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
RS::EnvironmentBG RendererSceneRenderRD::environment_get_background(RID p_env) const {
Fix spelling of a var, a struct, and message output
2020-07-17 16:02:06 +00:00
Environment *env = environment_owner.getornull(p_env);
Renaming of servers for coherency. VisualServer -> RenderingServer PhysicsServer -> PhysicsServer3D Physics2DServer -> PhysicsServer2D NavigationServer -> NavigationServer3D Navigation2DServer -> NavigationServer2D Also renamed corresponding files.
2020-03-27 18:21:27 +00:00
ERR_FAIL_COND_V(!env, RS::ENV_BG_MAX);
Environment sky more or less working.
2019-08-26 20:43:58 +00:00
return env->background;
}
Style: Enforce separation line between function definitions I couldn't find a tool that enforces it, so I went the manual route: ``` find -name "thirdparty" -prune \ -o -name "*.cpp" -o -name "*.h" -o -name "*.m" -o -name "*.mm" \ -o -name "*.glsl" > files perl -0777 -pi -e 's/\n}\n([^#])/\n}\n\n\1/g' $(cat files) misc/scripts/fix_style.sh -c ``` This adds a newline after all `}` on the first column, unless they are followed by `#` (typically `#endif`). This leads to having lots of places with two lines between function/class definitions, but clang-format then fixes it as we enforce max one line of separation. This doesn't fix potential occurrences of function definitions which are indented (e.g. for a helper class defined in a .cpp), but it's better than nothing. Also can't be made to run easily on CI/hooks so we'll have to be careful with new code. Part of #33027.
2020-05-14 12:29:06 +00:00
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
RID RendererSceneRenderRD::environment_get_sky(RID p_env) const {
Fix spelling of a var, a struct, and message output
2020-07-17 16:02:06 +00:00
Environment *env = environment_owner.getornull(p_env);
Environment sky more or less working.
2019-08-26 20:43:58 +00:00
ERR_FAIL_COND_V(!env, RID());
return env->sky;
}
Style: Enforce separation line between function definitions I couldn't find a tool that enforces it, so I went the manual route: ``` find -name "thirdparty" -prune \ -o -name "*.cpp" -o -name "*.h" -o -name "*.m" -o -name "*.mm" \ -o -name "*.glsl" > files perl -0777 -pi -e 's/\n}\n([^#])/\n}\n\n\1/g' $(cat files) misc/scripts/fix_style.sh -c ``` This adds a newline after all `}` on the first column, unless they are followed by `#` (typically `#endif`). This leads to having lots of places with two lines between function/class definitions, but clang-format then fixes it as we enforce max one line of separation. This doesn't fix potential occurrences of function definitions which are indented (e.g. for a helper class defined in a .cpp), but it's better than nothing. Also can't be made to run easily on CI/hooks so we'll have to be careful with new code. Part of #33027.
2020-05-14 12:29:06 +00:00
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
float RendererSceneRenderRD::environment_get_sky_custom_fov(RID p_env) const {
Fix spelling of a var, a struct, and message output
2020-07-17 16:02:06 +00:00
Environment *env = environment_owner.getornull(p_env);
Environment sky more or less working.
2019-08-26 20:43:58 +00:00
ERR_FAIL_COND_V(!env, 0);
return env->sky_custom_fov;
}
Style: Enforce separation line between function definitions I couldn't find a tool that enforces it, so I went the manual route: ``` find -name "thirdparty" -prune \ -o -name "*.cpp" -o -name "*.h" -o -name "*.m" -o -name "*.mm" \ -o -name "*.glsl" > files perl -0777 -pi -e 's/\n}\n([^#])/\n}\n\n\1/g' $(cat files) misc/scripts/fix_style.sh -c ``` This adds a newline after all `}` on the first column, unless they are followed by `#` (typically `#endif`). This leads to having lots of places with two lines between function/class definitions, but clang-format then fixes it as we enforce max one line of separation. This doesn't fix potential occurrences of function definitions which are indented (e.g. for a helper class defined in a .cpp), but it's better than nothing. Also can't be made to run easily on CI/hooks so we'll have to be careful with new code. Part of #33027.
2020-05-14 12:29:06 +00:00
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
Basis RendererSceneRenderRD::environment_get_sky_orientation(RID p_env) const {
Fix spelling of a var, a struct, and message output
2020-07-17 16:02:06 +00:00
Environment *env = environment_owner.getornull(p_env);
Environment sky more or less working.
2019-08-26 20:43:58 +00:00
ERR_FAIL_COND_V(!env, Basis());
return env->sky_orientation;
}
Style: Enforce separation line between function definitions I couldn't find a tool that enforces it, so I went the manual route: ``` find -name "thirdparty" -prune \ -o -name "*.cpp" -o -name "*.h" -o -name "*.m" -o -name "*.mm" \ -o -name "*.glsl" > files perl -0777 -pi -e 's/\n}\n([^#])/\n}\n\n\1/g' $(cat files) misc/scripts/fix_style.sh -c ``` This adds a newline after all `}` on the first column, unless they are followed by `#` (typically `#endif`). This leads to having lots of places with two lines between function/class definitions, but clang-format then fixes it as we enforce max one line of separation. This doesn't fix potential occurrences of function definitions which are indented (e.g. for a helper class defined in a .cpp), but it's better than nothing. Also can't be made to run easily on CI/hooks so we'll have to be careful with new code. Part of #33027.
2020-05-14 12:29:06 +00:00
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
Color RendererSceneRenderRD::environment_get_bg_color(RID p_env) const {
Fix spelling of a var, a struct, and message output
2020-07-17 16:02:06 +00:00
Environment *env = environment_owner.getornull(p_env);
Environment sky more or less working.
2019-08-26 20:43:58 +00:00
ERR_FAIL_COND_V(!env, Color());
return env->bg_color;
}
Style: Enforce separation line between function definitions I couldn't find a tool that enforces it, so I went the manual route: ``` find -name "thirdparty" -prune \ -o -name "*.cpp" -o -name "*.h" -o -name "*.m" -o -name "*.mm" \ -o -name "*.glsl" > files perl -0777 -pi -e 's/\n}\n([^#])/\n}\n\n\1/g' $(cat files) misc/scripts/fix_style.sh -c ``` This adds a newline after all `}` on the first column, unless they are followed by `#` (typically `#endif`). This leads to having lots of places with two lines between function/class definitions, but clang-format then fixes it as we enforce max one line of separation. This doesn't fix potential occurrences of function definitions which are indented (e.g. for a helper class defined in a .cpp), but it's better than nothing. Also can't be made to run easily on CI/hooks so we'll have to be careful with new code. Part of #33027.
2020-05-14 12:29:06 +00:00
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
float RendererSceneRenderRD::environment_get_bg_energy(RID p_env) const {
Fix spelling of a var, a struct, and message output
2020-07-17 16:02:06 +00:00
Environment *env = environment_owner.getornull(p_env);
Environment sky more or less working.
2019-08-26 20:43:58 +00:00
ERR_FAIL_COND_V(!env, 0);
return env->bg_energy;
}
Style: Enforce separation line between function definitions I couldn't find a tool that enforces it, so I went the manual route: ``` find -name "thirdparty" -prune \ -o -name "*.cpp" -o -name "*.h" -o -name "*.m" -o -name "*.mm" \ -o -name "*.glsl" > files perl -0777 -pi -e 's/\n}\n([^#])/\n}\n\n\1/g' $(cat files) misc/scripts/fix_style.sh -c ``` This adds a newline after all `}` on the first column, unless they are followed by `#` (typically `#endif`). This leads to having lots of places with two lines between function/class definitions, but clang-format then fixes it as we enforce max one line of separation. This doesn't fix potential occurrences of function definitions which are indented (e.g. for a helper class defined in a .cpp), but it's better than nothing. Also can't be made to run easily on CI/hooks so we'll have to be careful with new code. Part of #33027.
2020-05-14 12:29:06 +00:00
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
int RendererSceneRenderRD::environment_get_canvas_max_layer(RID p_env) const {
Fix spelling of a var, a struct, and message output
2020-07-17 16:02:06 +00:00
Environment *env = environment_owner.getornull(p_env);
Environment sky more or less working.
2019-08-26 20:43:58 +00:00
ERR_FAIL_COND_V(!env, 0);
return env->canvas_max_layer;
}
Style: Enforce separation line between function definitions I couldn't find a tool that enforces it, so I went the manual route: ``` find -name "thirdparty" -prune \ -o -name "*.cpp" -o -name "*.h" -o -name "*.m" -o -name "*.mm" \ -o -name "*.glsl" > files perl -0777 -pi -e 's/\n}\n([^#])/\n}\n\n\1/g' $(cat files) misc/scripts/fix_style.sh -c ``` This adds a newline after all `}` on the first column, unless they are followed by `#` (typically `#endif`). This leads to having lots of places with two lines between function/class definitions, but clang-format then fixes it as we enforce max one line of separation. This doesn't fix potential occurrences of function definitions which are indented (e.g. for a helper class defined in a .cpp), but it's better than nothing. Also can't be made to run easily on CI/hooks so we'll have to be careful with new code. Part of #33027.
2020-05-14 12:29:06 +00:00
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
Color RendererSceneRenderRD::environment_get_ambient_light_color(RID p_env) const {
Fix spelling of a var, a struct, and message output
2020-07-17 16:02:06 +00:00
Environment *env = environment_owner.getornull(p_env);
Environment sky more or less working.
2019-08-26 20:43:58 +00:00
ERR_FAIL_COND_V(!env, Color());
return env->ambient_light;
}
Style: Enforce separation line between function definitions I couldn't find a tool that enforces it, so I went the manual route: ``` find -name "thirdparty" -prune \ -o -name "*.cpp" -o -name "*.h" -o -name "*.m" -o -name "*.mm" \ -o -name "*.glsl" > files perl -0777 -pi -e 's/\n}\n([^#])/\n}\n\n\1/g' $(cat files) misc/scripts/fix_style.sh -c ``` This adds a newline after all `}` on the first column, unless they are followed by `#` (typically `#endif`). This leads to having lots of places with two lines between function/class definitions, but clang-format then fixes it as we enforce max one line of separation. This doesn't fix potential occurrences of function definitions which are indented (e.g. for a helper class defined in a .cpp), but it's better than nothing. Also can't be made to run easily on CI/hooks so we'll have to be careful with new code. Part of #33027.
2020-05-14 12:29:06 +00:00
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
RS::EnvironmentAmbientSource RendererSceneRenderRD::environment_get_ambient_source(RID p_env) const {
Fix spelling of a var, a struct, and message output
2020-07-17 16:02:06 +00:00
Environment *env = environment_owner.getornull(p_env);
Renaming of servers for coherency. VisualServer -> RenderingServer PhysicsServer -> PhysicsServer3D Physics2DServer -> PhysicsServer2D NavigationServer -> NavigationServer3D Navigation2DServer -> NavigationServer2D Also renamed corresponding files.
2020-03-27 18:21:27 +00:00
ERR_FAIL_COND_V(!env, RS::ENV_AMBIENT_SOURCE_BG);
Environment sky more or less working.
2019-08-26 20:43:58 +00:00
return env->ambient_source;
}
Style: Enforce separation line between function definitions I couldn't find a tool that enforces it, so I went the manual route: ``` find -name "thirdparty" -prune \ -o -name "*.cpp" -o -name "*.h" -o -name "*.m" -o -name "*.mm" \ -o -name "*.glsl" > files perl -0777 -pi -e 's/\n}\n([^#])/\n}\n\n\1/g' $(cat files) misc/scripts/fix_style.sh -c ``` This adds a newline after all `}` on the first column, unless they are followed by `#` (typically `#endif`). This leads to having lots of places with two lines between function/class definitions, but clang-format then fixes it as we enforce max one line of separation. This doesn't fix potential occurrences of function definitions which are indented (e.g. for a helper class defined in a .cpp), but it's better than nothing. Also can't be made to run easily on CI/hooks so we'll have to be careful with new code. Part of #33027.
2020-05-14 12:29:06 +00:00
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
float RendererSceneRenderRD::environment_get_ambient_light_energy(RID p_env) const {
Fix spelling of a var, a struct, and message output
2020-07-17 16:02:06 +00:00
Environment *env = environment_owner.getornull(p_env);
Environment sky more or less working.
2019-08-26 20:43:58 +00:00
ERR_FAIL_COND_V(!env, 0);
return env->ambient_light_energy;
}
Style: Enforce separation line between function definitions I couldn't find a tool that enforces it, so I went the manual route: ``` find -name "thirdparty" -prune \ -o -name "*.cpp" -o -name "*.h" -o -name "*.m" -o -name "*.mm" \ -o -name "*.glsl" > files perl -0777 -pi -e 's/\n}\n([^#])/\n}\n\n\1/g' $(cat files) misc/scripts/fix_style.sh -c ``` This adds a newline after all `}` on the first column, unless they are followed by `#` (typically `#endif`). This leads to having lots of places with two lines between function/class definitions, but clang-format then fixes it as we enforce max one line of separation. This doesn't fix potential occurrences of function definitions which are indented (e.g. for a helper class defined in a .cpp), but it's better than nothing. Also can't be made to run easily on CI/hooks so we'll have to be careful with new code. Part of #33027.
2020-05-14 12:29:06 +00:00
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
float RendererSceneRenderRD::environment_get_ambient_sky_contribution(RID p_env) const {
Fix spelling of a var, a struct, and message output
2020-07-17 16:02:06 +00:00
Environment *env = environment_owner.getornull(p_env);
Environment sky more or less working.
2019-08-26 20:43:58 +00:00
ERR_FAIL_COND_V(!env, 0);
return env->ambient_sky_contribution;
}
Style: Enforce separation line between function definitions I couldn't find a tool that enforces it, so I went the manual route: ``` find -name "thirdparty" -prune \ -o -name "*.cpp" -o -name "*.h" -o -name "*.m" -o -name "*.mm" \ -o -name "*.glsl" > files perl -0777 -pi -e 's/\n}\n([^#])/\n}\n\n\1/g' $(cat files) misc/scripts/fix_style.sh -c ``` This adds a newline after all `}` on the first column, unless they are followed by `#` (typically `#endif`). This leads to having lots of places with two lines between function/class definitions, but clang-format then fixes it as we enforce max one line of separation. This doesn't fix potential occurrences of function definitions which are indented (e.g. for a helper class defined in a .cpp), but it's better than nothing. Also can't be made to run easily on CI/hooks so we'll have to be careful with new code. Part of #33027.
2020-05-14 12:29:06 +00:00
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
RS::EnvironmentReflectionSource RendererSceneRenderRD::environment_get_reflection_source(RID p_env) const {
Fix spelling of a var, a struct, and message output
2020-07-17 16:02:06 +00:00
Environment *env = environment_owner.getornull(p_env);
Renaming of servers for coherency. VisualServer -> RenderingServer PhysicsServer -> PhysicsServer3D Physics2DServer -> PhysicsServer2D NavigationServer -> NavigationServer3D Navigation2DServer -> NavigationServer2D Also renamed corresponding files.
2020-03-27 18:21:27 +00:00
ERR_FAIL_COND_V(!env, RS::ENV_REFLECTION_SOURCE_DISABLED);
Environment sky more or less working.
2019-08-26 20:43:58 +00:00
return env->reflection_source;
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
Color RendererSceneRenderRD::environment_get_ao_color(RID p_env) const {
Fix spelling of a var, a struct, and message output
2020-07-17 16:02:06 +00:00
Environment *env = environment_owner.getornull(p_env);
Re-implemented screen space ambient occlusion
2020-01-25 10:18:55 +00:00
ERR_FAIL_COND_V(!env, Color());
return env->ao_color;
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
void RendererSceneRenderRD::environment_set_tonemap(RID p_env, RS::EnvironmentToneMapper p_tone_mapper, float p_exposure, float p_white, bool p_auto_exposure, float p_min_luminance, float p_max_luminance, float p_auto_exp_speed, float p_auto_exp_scale) {
Fix spelling of a var, a struct, and message output
2020-07-17 16:02:06 +00:00
Environment *env = environment_owner.getornull(p_env);
Tonemapping ported (not all parameters supported yet, only enough to get correct color)
2019-08-27 18:27:35 +00:00
ERR_FAIL_COND(!env);
Several fixes to 3D rendering, and multimesh implementation.
2019-09-14 03:37:42 +00:00
env->exposure = p_exposure;
Tonemapping ported (not all parameters supported yet, only enough to get correct color)
2019-08-27 18:27:35 +00:00
env->tone_mapper = p_tone_mapper;
Auto exposure re-implemented in Vulkan
2020-01-12 01:26:52 +00:00
if (!env->auto_exposure && p_auto_exposure) {
env->auto_exposure_version = ++auto_exposure_counter;
}
Tonemapping ported (not all parameters supported yet, only enough to get correct color)
2019-08-27 18:27:35 +00:00
env->auto_exposure = p_auto_exposure;
env->white = p_white;
env->min_luminance = p_min_luminance;
env->max_luminance = p_max_luminance;
env->auto_exp_speed = p_auto_exp_speed;
env->auto_exp_scale = p_auto_exp_scale;
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
void RendererSceneRenderRD::environment_set_glow(RID p_env, bool p_enable, Vector<float> p_levels, float p_intensity, float p_strength, float p_mix, float p_bloom_threshold, RS::EnvironmentGlowBlendMode p_blend_mode, float p_hdr_bleed_threshold, float p_hdr_bleed_scale, float p_hdr_luminance_cap) {
Fix spelling of a var, a struct, and message output
2020-07-17 16:02:06 +00:00
Environment *env = environment_owner.getornull(p_env);
-Refactored post processing, re-added glow and added a mix blend mode.
2020-01-10 00:40:26 +00:00
ERR_FAIL_COND(!env);
Optimize Glow with local memory
2020-09-19 22:02:32 +00:00
ERR_FAIL_COND_MSG(p_levels.size() != 7, "Size of array of glow levels must be 7");
-Refactored post processing, re-added glow and added a mix blend mode.
2020-01-10 00:40:26 +00:00
env->glow_enabled = p_enable;
Optimize Glow with local memory
2020-09-19 22:02:32 +00:00
env->glow_levels = p_levels;
-Refactored post processing, re-added glow and added a mix blend mode.
2020-01-10 00:40:26 +00:00
env->glow_intensity = p_intensity;
env->glow_strength = p_strength;
env->glow_mix = p_mix;
env->glow_bloom = p_bloom_threshold;
env->glow_blend_mode = p_blend_mode;
env->glow_hdr_bleed_threshold = p_hdr_bleed_threshold;
env->glow_hdr_bleed_scale = p_hdr_bleed_scale;
env->glow_hdr_luminance_cap = p_hdr_luminance_cap;
Move glow upscale quality to a global setting, for consistency
2020-03-30 13:46:03 +00:00
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
void RendererSceneRenderRD::environment_glow_set_use_bicubic_upscale(bool p_enable) {
Move glow upscale quality to a global setting, for consistency
2020-03-30 13:46:03 +00:00
glow_bicubic_upscale = p_enable;
Tonemapping ported (not all parameters supported yet, only enough to get correct color)
2019-08-27 18:27:35 +00:00
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
void RendererSceneRenderRD::environment_glow_set_use_high_quality(bool p_enable) {
Add high quality glow mode
2020-09-01 06:04:45 +00:00
glow_high_quality = p_enable;
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
void RendererSceneRenderRD::environment_set_sdfgi(RID p_env, bool p_enable, RS::EnvironmentSDFGICascades p_cascades, float p_min_cell_size, RS::EnvironmentSDFGIYScale p_y_scale, bool p_use_occlusion, bool p_use_multibounce, bool p_read_sky, float p_energy, float p_normal_bias, float p_probe_bias) {
Fix spelling of a var, a struct, and message output
2020-07-17 16:02:06 +00:00
Environment *env = environment_owner.getornull(p_env);
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
ERR_FAIL_COND(!env);
Add support for low-end 3D rendering. -Reduce number of uniform sets from 6 to 4. -Remove features in low end mode, in order to reduce the number of texture units fit to 16.
2020-12-07 21:27:38 +00:00
if (low_end) {
return;
}
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
env->sdfgi_enabled = p_enable;
env->sdfgi_cascades = p_cascades;
env->sdfgi_min_cell_size = p_min_cell_size;
env->sdfgi_use_occlusion = p_use_occlusion;
env->sdfgi_use_multibounce = p_use_multibounce;
env->sdfgi_read_sky_light = p_read_sky;
env->sdfgi_energy = p_energy;
env->sdfgi_normal_bias = p_normal_bias;
env->sdfgi_probe_bias = p_probe_bias;
env->sdfgi_y_scale = p_y_scale;
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
void RendererSceneRenderRD::environment_set_fog(RID p_env, bool p_enable, const Color &p_light_color, float p_light_energy, float p_sun_scatter, float p_density, float p_height, float p_height_density, float p_fog_aerial_perspective) {
Restored fog (non volumetric). Uses a simpler and more intuitive implementation based on density. Its less flexible than before, but its easier to get nice results.
2020-08-14 01:07:49 +00:00
Environment *env = environment_owner.getornull(p_env);
ERR_FAIL_COND(!env);
env->fog_enabled = p_enable;
env->fog_light_color = p_light_color;
env->fog_light_energy = p_light_energy;
env->fog_sun_scatter = p_sun_scatter;
env->fog_density = p_density;
env->fog_height = p_height;
env->fog_height_density = p_height_density;
Add aerial perspective to fixed fog
2020-08-21 05:48:04 +00:00
env->fog_aerial_perspective = p_fog_aerial_perspective;
Restored fog (non volumetric). Uses a simpler and more intuitive implementation based on density. Its less flexible than before, but its easier to get nice results.
2020-08-14 01:07:49 +00:00
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
bool RendererSceneRenderRD::environment_is_fog_enabled(RID p_env) const {
Restored fog (non volumetric). Uses a simpler and more intuitive implementation based on density. Its less flexible than before, but its easier to get nice results.
2020-08-14 01:07:49 +00:00
const Environment *env = environment_owner.getornull(p_env);
ERR_FAIL_COND_V(!env, false);
return env->fog_enabled;
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
Color RendererSceneRenderRD::environment_get_fog_light_color(RID p_env) const {
Restored fog (non volumetric). Uses a simpler and more intuitive implementation based on density. Its less flexible than before, but its easier to get nice results.
2020-08-14 01:07:49 +00:00
const Environment *env = environment_owner.getornull(p_env);
ERR_FAIL_COND_V(!env, Color());
return env->fog_light_color;
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
float RendererSceneRenderRD::environment_get_fog_light_energy(RID p_env) const {
Restored fog (non volumetric). Uses a simpler and more intuitive implementation based on density. Its less flexible than before, but its easier to get nice results.
2020-08-14 01:07:49 +00:00
const Environment *env = environment_owner.getornull(p_env);
ERR_FAIL_COND_V(!env, 0);
return env->fog_light_energy;
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
float RendererSceneRenderRD::environment_get_fog_sun_scatter(RID p_env) const {
Restored fog (non volumetric). Uses a simpler and more intuitive implementation based on density. Its less flexible than before, but its easier to get nice results.
2020-08-14 01:07:49 +00:00
const Environment *env = environment_owner.getornull(p_env);
ERR_FAIL_COND_V(!env, 0);
return env->fog_sun_scatter;
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
float RendererSceneRenderRD::environment_get_fog_density(RID p_env) const {
Restored fog (non volumetric). Uses a simpler and more intuitive implementation based on density. Its less flexible than before, but its easier to get nice results.
2020-08-14 01:07:49 +00:00
const Environment *env = environment_owner.getornull(p_env);
ERR_FAIL_COND_V(!env, 0);
return env->fog_density;
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
float RendererSceneRenderRD::environment_get_fog_height(RID p_env) const {
Restored fog (non volumetric). Uses a simpler and more intuitive implementation based on density. Its less flexible than before, but its easier to get nice results.
2020-08-14 01:07:49 +00:00
const Environment *env = environment_owner.getornull(p_env);
ERR_FAIL_COND_V(!env, 0);
return env->fog_height;
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
float RendererSceneRenderRD::environment_get_fog_height_density(RID p_env) const {
Restored fog (non volumetric). Uses a simpler and more intuitive implementation based on density. Its less flexible than before, but its easier to get nice results.
2020-08-14 01:07:49 +00:00
const Environment *env = environment_owner.getornull(p_env);
ERR_FAIL_COND_V(!env, 0);
return env->fog_height_density;
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
float RendererSceneRenderRD::environment_get_fog_aerial_perspective(RID p_env) const {
Add aerial perspective to fixed fog
2020-08-21 05:48:04 +00:00
const Environment *env = environment_owner.getornull(p_env);
ERR_FAIL_COND_V(!env, 0);
return env->fog_aerial_perspective;
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
void RendererSceneRenderRD::environment_set_volumetric_fog(RID p_env, bool p_enable, float p_density, const Color &p_light, float p_light_energy, float p_length, float p_detail_spread, float p_gi_inject, RenderingServer::EnvVolumetricFogShadowFilter p_shadow_filter) {
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
Environment *env = environment_owner.getornull(p_env);
ERR_FAIL_COND(!env);
Add support for low-end 3D rendering. -Reduce number of uniform sets from 6 to 4. -Remove features in low end mode, in order to reduce the number of texture units fit to 16.
2020-12-07 21:27:38 +00:00
if (low_end) {
return;
}
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
env->volumetric_fog_enabled = p_enable;
env->volumetric_fog_density = p_density;
env->volumetric_fog_light = p_light;
env->volumetric_fog_light_energy = p_light_energy;
Add fog to sky shaders
2020-08-18 04:12:51 +00:00
env->volumetric_fog_length = p_length;
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
env->volumetric_fog_detail_spread = p_detail_spread;
env->volumetric_fog_shadow_filter = p_shadow_filter;
env->volumetric_fog_gi_inject = p_gi_inject;
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
void RendererSceneRenderRD::environment_set_volumetric_fog_volume_size(int p_size, int p_depth) {
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
volumetric_fog_size = p_size;
volumetric_fog_depth = p_depth;
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
void RendererSceneRenderRD::environment_set_volumetric_fog_filter_active(bool p_enable) {
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
volumetric_fog_filter_active = p_enable;
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
void RendererSceneRenderRD::environment_set_volumetric_fog_directional_shadow_shrink_size(int p_shrink_size) {
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
p_shrink_size = nearest_power_of_2_templated(p_shrink_size);
if (volumetric_fog_directional_shadow_shrink == (uint32_t)p_shrink_size) {
return;
}
_clear_shadow_shrink_stages(directional_shadow.shrink_stages);
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
void RendererSceneRenderRD::environment_set_volumetric_fog_positional_shadow_shrink_size(int p_shrink_size) {
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
p_shrink_size = nearest_power_of_2_templated(p_shrink_size);
if (volumetric_fog_positional_shadow_shrink == (uint32_t)p_shrink_size) {
return;
}
for (uint32_t i = 0; i < shadow_atlas_owner.get_rid_count(); i++) {
ShadowAtlas *sa = shadow_atlas_owner.get_ptr_by_index(i);
_clear_shadow_shrink_stages(sa->shrink_stages);
}
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
void RendererSceneRenderRD::environment_set_sdfgi_ray_count(RS::EnvironmentSDFGIRayCount p_ray_count) {
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
sdfgi_ray_count = p_ray_count;
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
void RendererSceneRenderRD::environment_set_sdfgi_frames_to_converge(RS::EnvironmentSDFGIFramesToConverge p_frames) {
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
sdfgi_frames_to_converge = p_frames;
}
Several GI related optimizations and fixes -SDFGI direct light is done over many frames -SDFGI Changed settings for rays/frame -SDFGI Misc optimizations -SDFGI Bug fix on probe scroll -GIProbe was not working, got it to work again -GIProbe dynamic objects were not working, fixed -Added a half size GI option.
2021-01-22 23:50:24 +00:00
void RendererSceneRenderRD::environment_set_sdfgi_frames_to_update_light(RS::EnvironmentSDFGIFramesToUpdateLight p_update) {
sdfgi_frames_to_update_light = p_update;
}
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
void RendererSceneRenderRD::environment_set_ssr(RID p_env, bool p_enable, int p_max_steps, float p_fade_int, float p_fade_out, float p_depth_tolerance) {
Fix spelling of a var, a struct, and message output
2020-07-17 16:02:06 +00:00
Environment *env = environment_owner.getornull(p_env);
Re-Added screen space reflection.
2020-04-02 02:24:52 +00:00
ERR_FAIL_COND(!env);
Add support for low-end 3D rendering. -Reduce number of uniform sets from 6 to 4. -Remove features in low end mode, in order to reduce the number of texture units fit to 16.
2020-12-07 21:27:38 +00:00
if (low_end) {
return;
}
Re-Added screen space reflection.
2020-04-02 02:24:52 +00:00
env->ssr_enabled = p_enable;
env->ssr_max_steps = p_max_steps;
env->ssr_fade_in = p_fade_int;
env->ssr_fade_out = p_fade_out;
env->ssr_depth_tolerance = p_depth_tolerance;
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
void RendererSceneRenderRD::environment_set_ssr_roughness_quality(RS::EnvironmentSSRRoughnessQuality p_quality) {
Re-Added screen space reflection.
2020-04-02 02:24:52 +00:00
ssr_roughness_quality = p_quality;
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
RS::EnvironmentSSRRoughnessQuality RendererSceneRenderRD::environment_get_ssr_roughness_quality() const {
Re-Added screen space reflection.
2020-04-02 02:24:52 +00:00
return ssr_roughness_quality;
}
Port ASSAO to Godot to replace SAO
2020-12-08 05:37:09 +00:00
void RendererSceneRenderRD::environment_set_ssao(RID p_env, bool p_enable, float p_radius, float p_intensity, float p_power, float p_detail, float p_horizon, float p_sharpness, float p_light_affect, float p_ao_channel_affect) {
Fix spelling of a var, a struct, and message output
2020-07-17 16:02:06 +00:00
Environment *env = environment_owner.getornull(p_env);
Re-implemented screen space ambient occlusion
2020-01-25 10:18:55 +00:00
ERR_FAIL_COND(!env);
Add support for low-end 3D rendering. -Reduce number of uniform sets from 6 to 4. -Remove features in low end mode, in order to reduce the number of texture units fit to 16.
2020-12-07 21:27:38 +00:00
if (low_end) {
return;
}
Re-implemented screen space ambient occlusion
2020-01-25 10:18:55 +00:00
env->ssao_enabled = p_enable;
Revert "Replace SAO implementation with MSSAO"
2020-10-18 22:27:51 +00:00
env->ssao_radius = p_radius;
Re-implemented screen space ambient occlusion
2020-01-25 10:18:55 +00:00
env->ssao_intensity = p_intensity;
Port ASSAO to Godot to replace SAO
2020-12-08 05:37:09 +00:00
env->ssao_power = p_power;
env->ssao_detail = p_detail;
env->ssao_horizon = p_horizon;
env->ssao_sharpness = p_sharpness;
Re-implemented screen space ambient occlusion
2020-01-25 10:18:55 +00:00
env->ssao_direct_light_affect = p_light_affect;
env->ssao_ao_channel_affect = p_ao_channel_affect;
}
Port ASSAO to Godot to replace SAO
2020-12-08 05:37:09 +00:00
void RendererSceneRenderRD::environment_set_ssao_quality(RS::EnvironmentSSAOQuality p_quality, bool p_half_size, float p_adaptive_target, int p_blur_passes, float p_fadeout_from, float p_fadeout_to) {
Re-implemented screen space ambient occlusion
2020-01-25 10:18:55 +00:00
ssao_quality = p_quality;
Revert "Replace SAO implementation with MSSAO"
2020-10-18 22:27:51 +00:00
ssao_half_size = p_half_size;
Port ASSAO to Godot to replace SAO
2020-12-08 05:37:09 +00:00
ssao_adaptive_target = p_adaptive_target;
ssao_blur_passes = p_blur_passes;
ssao_fadeout_from = p_fadeout_from;
ssao_fadeout_to = p_fadeout_to;
Re-implemented screen space ambient occlusion
2020-01-25 10:18:55 +00:00
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
bool RendererSceneRenderRD::environment_is_ssao_enabled(RID p_env) const {
Fix spelling of a var, a struct, and message output
2020-07-17 16:02:06 +00:00
Environment *env = environment_owner.getornull(p_env);
Re-implemented screen space ambient occlusion
2020-01-25 10:18:55 +00:00
ERR_FAIL_COND_V(!env, false);
return env->ssao_enabled;
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
float RendererSceneRenderRD::environment_get_ssao_ao_affect(RID p_env) const {
Fix spelling of a var, a struct, and message output
2020-07-17 16:02:06 +00:00
Environment *env = environment_owner.getornull(p_env);
Static analyzer fixes: Removes unused code in OS. Fixes return types. Fixes few typos.
2020-12-08 18:58:49 +00:00
ERR_FAIL_COND_V(!env, 0.0);
Re-implemented screen space ambient occlusion
2020-01-25 10:18:55 +00:00
return env->ssao_ao_channel_affect;
}
Style: Enforce separation line between function definitions I couldn't find a tool that enforces it, so I went the manual route: ``` find -name "thirdparty" -prune \ -o -name "*.cpp" -o -name "*.h" -o -name "*.m" -o -name "*.mm" \ -o -name "*.glsl" > files perl -0777 -pi -e 's/\n}\n([^#])/\n}\n\n\1/g' $(cat files) misc/scripts/fix_style.sh -c ``` This adds a newline after all `}` on the first column, unless they are followed by `#` (typically `#endif`). This leads to having lots of places with two lines between function/class definitions, but clang-format then fixes it as we enforce max one line of separation. This doesn't fix potential occurrences of function definitions which are indented (e.g. for a helper class defined in a .cpp), but it's better than nothing. Also can't be made to run easily on CI/hooks so we'll have to be careful with new code. Part of #33027.
2020-05-14 12:29:06 +00:00
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
float RendererSceneRenderRD::environment_get_ssao_light_affect(RID p_env) const {
Fix spelling of a var, a struct, and message output
2020-07-17 16:02:06 +00:00
Environment *env = environment_owner.getornull(p_env);
Static analyzer fixes: Removes unused code in OS. Fixes return types. Fixes few typos.
2020-12-08 18:58:49 +00:00
ERR_FAIL_COND_V(!env, 0.0);
Re-implemented screen space ambient occlusion
2020-01-25 10:18:55 +00:00
return env->ssao_direct_light_affect;
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
bool RendererSceneRenderRD::environment_is_ssr_enabled(RID p_env) const {
Fix spelling of a var, a struct, and message output
2020-07-17 16:02:06 +00:00
Environment *env = environment_owner.getornull(p_env);
Re-implemented screen space ambient occlusion
2020-01-25 10:18:55 +00:00
ERR_FAIL_COND_V(!env, false);
Re-Added screen space reflection.
2020-04-02 02:24:52 +00:00
return env->ssr_enabled;
Re-implemented screen space ambient occlusion
2020-01-25 10:18:55 +00:00
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
bool RendererSceneRenderRD::environment_is_sdfgi_enabled(RID p_env) const {
Fix spelling of a var, a struct, and message output
2020-07-17 16:02:06 +00:00
Environment *env = environment_owner.getornull(p_env);
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
ERR_FAIL_COND_V(!env, false);
return env->sdfgi_enabled;
}
Re-implemented screen space ambient occlusion
2020-01-25 10:18:55 +00:00
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
bool RendererSceneRenderRD::is_environment(RID p_env) const {
Environment sky more or less working.
2019-08-26 20:43:58 +00:00
return environment_owner.owns(p_env);
}
Base 3D engine done, still untested, though.
2019-08-18 22:40:52 +00:00
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
Ref<Image> RendererSceneRenderRD::environment_bake_panorama(RID p_env, bool p_bake_irradiance, const Size2i &p_size) {
Fix spelling of a var, a struct, and message output
2020-07-17 16:02:06 +00:00
Environment *env = environment_owner.getornull(p_env);
New lightmapper -Added LocalVector (needed it) -Added stb_rect_pack (It's pretty cool, we could probably use it for other stuff too) -Fixes and changes all around the place -Added library for 128 bits fixed point (required for Delaunay3D)
2020-05-01 12:34:23 +00:00
ERR_FAIL_COND_V(!env, Ref<Image>());
if (env->background == RS::ENV_BG_CAMERA_FEED || env->background == RS::ENV_BG_CANVAS || env->background == RS::ENV_BG_KEEP) {
return Ref<Image>(); //nothing to bake
}
if (env->background == RS::ENV_BG_CLEAR_COLOR || env->background == RS::ENV_BG_COLOR) {
Color color;
if (env->background == RS::ENV_BG_CLEAR_COLOR) {
color = storage->get_default_clear_color();
} else {
color = env->bg_color;
}
color.r *= env->bg_energy;
color.g *= env->bg_energy;
color.b *= env->bg_energy;
Ref<Image> ret;
ret.instance();
ret->create(p_size.width, p_size.height, false, Image::FORMAT_RGBAF);
for (int i = 0; i < p_size.width; i++) {
for (int j = 0; j < p_size.height; j++) {
ret->set_pixel(i, j, color);
}
}
return ret;
}
if (env->background == RS::ENV_BG_SKY && env->sky.is_valid()) {
return sky_bake_panorama(env->sky, env->bg_energy, p_bake_irradiance, p_size);
}
return Ref<Image>();
}
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
////////////////////////////////////////////////////////////
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
RID RendererSceneRenderRD::reflection_atlas_create() {
Reflection probes working
2019-09-09 20:50:51 +00:00
ReflectionAtlas ra;
ra.count = GLOBAL_GET("rendering/quality/reflection_atlas/reflection_count");
ra.size = GLOBAL_GET("rendering/quality/reflection_atlas/reflection_size");
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
ra.cluster_builder = memnew(ClusterBuilderRD);
ra.cluster_builder->set_shared(&cluster_builder_shared);
ra.cluster_builder->setup(Size2i(ra.size, ra.size), max_cluster_elements, RID(), RID(), RID());
Reflection probes working
2019-09-09 20:50:51 +00:00
return reflection_atlas_owner.make_rid(ra);
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
void RendererSceneRenderRD::reflection_atlas_set_size(RID p_ref_atlas, int p_reflection_size, int p_reflection_count) {
Reflection probes working
2019-09-09 20:50:51 +00:00
ReflectionAtlas *ra = reflection_atlas_owner.getornull(p_ref_atlas);
ERR_FAIL_COND(!ra);
if (ra->size == p_reflection_size && ra->count == p_reflection_count) {
return; //no changes
}
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
ra->cluster_builder->setup(Size2i(ra->size, ra->size), max_cluster_elements, RID(), RID(), RID());
Basic fast filtering implementation
2020-02-20 23:27:34 +00:00
ra->size = p_reflection_size;
ra->count = p_reflection_count;
Reflection probes working
2019-09-09 20:50:51 +00:00
if (ra->reflection.is_valid()) {
//clear and invalidate everything
RD::get_singleton()->free(ra->reflection);
ra->reflection = RID();
Properly free reflection probe atlas
2020-03-01 07:26:57 +00:00
RD::get_singleton()->free(ra->depth_buffer);
ra->depth_buffer = RID();
Reflection probes working
2019-09-09 20:50:51 +00:00
for (int i = 0; i < ra->reflections.size(); i++) {
Properly free reflection probe atlas
2020-03-01 07:26:57 +00:00
_clear_reflection_data(ra->reflections.write[i].data);
Reflection probes working
2019-09-09 20:50:51 +00:00
if (ra->reflections[i].owner.is_null()) {
continue;
}
reflection_probe_release_atlas_index(ra->reflections[i].owner);
//rp->atlasindex clear
}
ra->reflections.clear();
}
}
Implement automatic LOD (Level of Detail) -Happens on import by default for all models -Just works (tm) -Biasing can be later adjusted per node or per viewport (as well as globally) -Disabled AABB.get_support test because its broken
2020-12-17 18:56:59 +00:00
int RendererSceneRenderRD::reflection_atlas_get_size(RID p_ref_atlas) const {
ReflectionAtlas *ra = reflection_atlas_owner.getornull(p_ref_atlas);
ERR_FAIL_COND_V(!ra, 0);
return ra->size;
}
Reflection probes working
2019-09-09 20:50:51 +00:00
////////////////////////
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
RID RendererSceneRenderRD::reflection_probe_instance_create(RID p_probe) {
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
ReflectionProbeInstance rpi;
rpi.probe = p_probe;
return reflection_probe_instance_owner.make_rid(rpi);
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
void RendererSceneRenderRD::reflection_probe_instance_set_transform(RID p_instance, const Transform &p_transform) {
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
ReflectionProbeInstance *rpi = reflection_probe_instance_owner.getornull(p_instance);
ERR_FAIL_COND(!rpi);
rpi->transform = p_transform;
rpi->dirty = true;
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
void RendererSceneRenderRD::reflection_probe_release_atlas_index(RID p_instance) {
Reflection probes working
2019-09-09 20:50:51 +00:00
ReflectionProbeInstance *rpi = reflection_probe_instance_owner.getornull(p_instance);
ERR_FAIL_COND(!rpi);
if (rpi->atlas.is_null()) {
return; //nothing to release
}
ReflectionAtlas *atlas = reflection_atlas_owner.getornull(rpi->atlas);
ERR_FAIL_COND(!atlas);
ERR_FAIL_INDEX(rpi->atlas_index, atlas->reflections.size());
atlas->reflections.write[rpi->atlas_index].owner = RID();
rpi->atlas_index = -1;
rpi->atlas = RID();
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
bool RendererSceneRenderRD::reflection_probe_instance_needs_redraw(RID p_instance) {
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
ReflectionProbeInstance *rpi = reflection_probe_instance_owner.getornull(p_instance);
ERR_FAIL_COND_V(!rpi, false);
if (rpi->rendering) {
return false;
}
if (rpi->dirty) {
return true;
}
Renaming of servers for coherency. VisualServer -> RenderingServer PhysicsServer -> PhysicsServer3D Physics2DServer -> PhysicsServer2D NavigationServer -> NavigationServer3D Navigation2DServer -> NavigationServer2D Also renamed corresponding files.
2020-03-27 18:21:27 +00:00
if (storage->reflection_probe_get_update_mode(rpi->probe) == RS::REFLECTION_PROBE_UPDATE_ALWAYS) {
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
return true;
}
Reflection probes working
2019-09-09 20:50:51 +00:00
return rpi->atlas_index == -1;
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
bool RendererSceneRenderRD::reflection_probe_instance_has_reflection(RID p_instance) {
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
ReflectionProbeInstance *rpi = reflection_probe_instance_owner.getornull(p_instance);
Reflection probes working
2019-09-09 20:50:51 +00:00
ERR_FAIL_COND_V(!rpi, false);
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
Reflection probes working
2019-09-09 20:50:51 +00:00
return rpi->atlas.is_valid();
}
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
bool RendererSceneRenderRD::reflection_probe_instance_begin_render(RID p_instance, RID p_reflection_atlas) {
Reflection probes working
2019-09-09 20:50:51 +00:00
ReflectionAtlas *atlas = reflection_atlas_owner.getornull(p_reflection_atlas);
ERR_FAIL_COND_V(!atlas, false);
Basic fast filtering implementation
2020-02-20 23:27:34 +00:00
ReflectionProbeInstance *rpi = reflection_probe_instance_owner.getornull(p_instance);
ERR_FAIL_COND_V(!rpi, false);
Renaming of servers for coherency. VisualServer -> RenderingServer PhysicsServer -> PhysicsServer3D Physics2DServer -> PhysicsServer2D NavigationServer -> NavigationServer3D Navigation2DServer -> NavigationServer2D Also renamed corresponding files.
2020-03-27 18:21:27 +00:00
if (storage->reflection_probe_get_update_mode(rpi->probe) == RS::REFLECTION_PROBE_UPDATE_ALWAYS && atlas->reflection.is_valid() && atlas->size != 256) {
Working sky shader implementation
2020-03-20 00:32:19 +00:00
WARN_PRINT("ReflectionProbes set to UPDATE_ALWAYS must have an atlas size of 256. Please update the atlas size in the ProjectSettings.");
reflection_atlas_set_size(p_reflection_atlas, 256, atlas->count);
Basic fast filtering implementation
2020-02-20 23:27:34 +00:00
}
Renaming of servers for coherency. VisualServer -> RenderingServer PhysicsServer -> PhysicsServer3D Physics2DServer -> PhysicsServer2D NavigationServer -> NavigationServer3D Navigation2DServer -> NavigationServer2D Also renamed corresponding files.
2020-03-27 18:21:27 +00:00
if (storage->reflection_probe_get_update_mode(rpi->probe) == RS::REFLECTION_PROBE_UPDATE_ALWAYS && atlas->reflection.is_valid() && atlas->reflections[0].data.layers[0].mipmaps.size() != 8) {
Improve cubemap importance sampling
2020-03-01 01:16:50 +00:00
// Invalidate reflection atlas, need to regenerate
RD::get_singleton()->free(atlas->reflection);
atlas->reflection = RID();
for (int i = 0; i < atlas->reflections.size(); i++) {
if (atlas->reflections[i].owner.is_null()) {
continue;
}
reflection_probe_release_atlas_index(atlas->reflections[i].owner);
}
atlas->reflections.clear();
}
Reflection probes working
2019-09-09 20:50:51 +00:00
if (atlas->reflection.is_null()) {
Basic fast filtering implementation
2020-02-20 23:27:34 +00:00
int mipmaps = MIN(roughness_layers, Image::get_image_required_mipmaps(atlas->size, atlas->size, Image::FORMAT_RGBAH) + 1);
Renaming of servers for coherency. VisualServer -> RenderingServer PhysicsServer -> PhysicsServer3D Physics2DServer -> PhysicsServer2D NavigationServer -> NavigationServer3D Navigation2DServer -> NavigationServer2D Also renamed corresponding files.
2020-03-27 18:21:27 +00:00
mipmaps = storage->reflection_probe_get_update_mode(rpi->probe) == RS::REFLECTION_PROBE_UPDATE_ALWAYS ? 8 : mipmaps; // always use 8 mipmaps with real time filtering
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
{
Reflection probes working
2019-09-09 20:50:51 +00:00
//reflection atlas was unused, create:
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
RD::TextureFormat tf;
Reflection probes working
2019-09-09 20:50:51 +00:00
tf.array_layers = 6 * atlas->count;
tf.format = RD::DATA_FORMAT_R16G16B16A16_SFLOAT;
Rename RD texture "type" to "texture_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-11-27 03:50:05 +00:00
tf.texture_type = RD::TEXTURE_TYPE_CUBE_ARRAY;
Basic fast filtering implementation
2020-02-20 23:27:34 +00:00
tf.mipmaps = mipmaps;
Reflection probes working
2019-09-09 20:50:51 +00:00
tf.width = atlas->size;
tf.height = atlas->size;
Basic fast filtering implementation
2020-02-20 23:27:34 +00:00
tf.usage_bits = RD::TEXTURE_USAGE_COLOR_ATTACHMENT_BIT | RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_STORAGE_BIT;
Reflection probes working
2019-09-09 20:50:51 +00:00
atlas->reflection = RD::get_singleton()->texture_create(tf, RD::TextureView());
}
{
RD::TextureFormat tf;
tf.format = RD::get_singleton()->texture_is_format_supported_for_usage(RD::DATA_FORMAT_D32_SFLOAT, RD::TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT) ? RD::DATA_FORMAT_D32_SFLOAT : RD::DATA_FORMAT_X8_D24_UNORM_PACK32;
tf.width = atlas->size;
tf.height = atlas->size;
tf.usage_bits = RD::TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT | RD::TEXTURE_USAGE_SAMPLING_BIT;
atlas->depth_buffer = RD::get_singleton()->texture_create(tf, RD::TextureView());
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
}
Reflection probes working
2019-09-09 20:50:51 +00:00
atlas->reflections.resize(atlas->count);
for (int i = 0; i < atlas->count; i++) {
Renaming of servers for coherency. VisualServer -> RenderingServer PhysicsServer -> PhysicsServer3D Physics2DServer -> PhysicsServer2D NavigationServer -> NavigationServer3D Navigation2DServer -> NavigationServer2D Also renamed corresponding files.
2020-03-27 18:21:27 +00:00
_update_reflection_data(atlas->reflections.write[i].data, atlas->size, mipmaps, false, atlas->reflection, i * 6, storage->reflection_probe_get_update_mode(rpi->probe) == RS::REFLECTION_PROBE_UPDATE_ALWAYS);
Reflection probes working
2019-09-09 20:50:51 +00:00
for (int j = 0; j < 6; j++) {
Vector<RID> fb;
fb.push_back(atlas->reflections.write[i].data.layers[0].mipmaps[0].views[j]);
fb.push_back(atlas->depth_buffer);
atlas->reflections.write[i].fbs[j] = RD::get_singleton()->framebuffer_create(fb);
}
}
GIProbes working.
2019-10-03 20:39:08 +00:00
Vector<RID> fb;
fb.push_back(atlas->depth_buffer);
atlas->depth_fb = RD::get_singleton()->framebuffer_create(fb);
Reflection probes working
2019-09-09 20:50:51 +00:00
}
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
Reflection probes working
2019-09-09 20:50:51 +00:00
if (rpi->atlas_index == -1) {
for (int i = 0; i < atlas->reflections.size(); i++) {
if (atlas->reflections[i].owner.is_null()) {
rpi->atlas_index = i;
break;
}
}
//find the one used last
if (rpi->atlas_index == -1) {
//everything is in use, find the one least used via LRU
uint64_t pass_min = 0;
for (int i = 0; i < atlas->reflections.size(); i++) {
ReflectionProbeInstance *rpi2 = reflection_probe_instance_owner.getornull(atlas->reflections[i].owner);
if (rpi2->last_pass < pass_min) {
pass_min = rpi2->last_pass;
rpi->atlas_index = i;
}
}
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
}
}
Reflection probes working
2019-09-09 20:50:51 +00:00
rpi->atlas = p_reflection_atlas;
rpi->rendering = true;
rpi->dirty = false;
Improve cubemap importance sampling
2020-03-01 01:16:50 +00:00
rpi->processing_layer = 1;
Reflection probes working
2019-09-09 20:50:51 +00:00
rpi->processing_side = 0;
return true;
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
bool RendererSceneRenderRD::reflection_probe_instance_postprocess_step(RID p_instance) {
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
ReflectionProbeInstance *rpi = reflection_probe_instance_owner.getornull(p_instance);
ERR_FAIL_COND_V(!rpi, false);
ERR_FAIL_COND_V(!rpi->rendering, false);
Reflection probes working
2019-09-09 20:50:51 +00:00
ERR_FAIL_COND_V(rpi->atlas.is_null(), false);
ReflectionAtlas *atlas = reflection_atlas_owner.getornull(rpi->atlas);
if (!atlas || rpi->atlas_index == -1) {
//does not belong to an atlas anymore, cancel (was removed from atlas or atlas changed while rendering)
rpi->rendering = false;
return false;
}
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
Renaming of servers for coherency. VisualServer -> RenderingServer PhysicsServer -> PhysicsServer3D Physics2DServer -> PhysicsServer2D NavigationServer -> NavigationServer3D Navigation2DServer -> NavigationServer2D Also renamed corresponding files.
2020-03-27 18:21:27 +00:00
if (storage->reflection_probe_get_update_mode(rpi->probe) == RS::REFLECTION_PROBE_UPDATE_ALWAYS) {
Working sky shader implementation
2020-03-20 00:32:19 +00:00
// Using real time reflections, all roughness is done in one step
_create_reflection_fast_filter(atlas->reflections.write[rpi->atlas_index].data, false);
rpi->rendering = false;
rpi->processing_side = 0;
rpi->processing_layer = 1;
return true;
}
Improve cubemap importance sampling
2020-03-01 01:16:50 +00:00
if (rpi->processing_layer > 1) {
Working sky shader implementation
2020-03-20 00:32:19 +00:00
_create_reflection_importance_sample(atlas->reflections.write[rpi->atlas_index].data, false, 10, rpi->processing_layer);
Improve cubemap importance sampling
2020-03-01 01:16:50 +00:00
rpi->processing_layer++;
if (rpi->processing_layer == atlas->reflections[rpi->atlas_index].data.layers[0].mipmaps.size()) {
rpi->rendering = false;
rpi->processing_side = 0;
rpi->processing_layer = 1;
return true;
}
return false;
} else {
Working sky shader implementation
2020-03-20 00:32:19 +00:00
_create_reflection_importance_sample(atlas->reflections.write[rpi->atlas_index].data, false, rpi->processing_side, rpi->processing_layer);
Basic fast filtering implementation
2020-02-20 23:27:34 +00:00
}
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
rpi->processing_side++;
if (rpi->processing_side == 6) {
rpi->processing_side = 0;
Improve cubemap importance sampling
2020-03-01 01:16:50 +00:00
rpi->processing_layer++;
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
}
Improve cubemap importance sampling
2020-03-01 01:16:50 +00:00
return false;
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
uint32_t RendererSceneRenderRD::reflection_probe_instance_get_resolution(RID p_instance) {
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
ReflectionProbeInstance *rpi = reflection_probe_instance_owner.getornull(p_instance);
ERR_FAIL_COND_V(!rpi, 0);
Reflection probes working
2019-09-09 20:50:51 +00:00
ReflectionAtlas *atlas = reflection_atlas_owner.getornull(rpi->atlas);
ERR_FAIL_COND_V(!atlas, 0);
return atlas->size;
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
RID RendererSceneRenderRD::reflection_probe_instance_get_framebuffer(RID p_instance, int p_index) {
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
ReflectionProbeInstance *rpi = reflection_probe_instance_owner.getornull(p_instance);
ERR_FAIL_COND_V(!rpi, RID());
ERR_FAIL_INDEX_V(p_index, 6, RID());
Reflection probes working
2019-09-09 20:50:51 +00:00
ReflectionAtlas *atlas = reflection_atlas_owner.getornull(rpi->atlas);
ERR_FAIL_COND_V(!atlas, RID());
return atlas->reflections[rpi->atlas_index].fbs[p_index];
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
RID RendererSceneRenderRD::reflection_probe_instance_get_depth_framebuffer(RID p_instance, int p_index) {
GIProbes working.
2019-10-03 20:39:08 +00:00
ReflectionProbeInstance *rpi = reflection_probe_instance_owner.getornull(p_instance);
ERR_FAIL_COND_V(!rpi, RID());
ERR_FAIL_INDEX_V(p_index, 6, RID());
ReflectionAtlas *atlas = reflection_atlas_owner.getornull(rpi->atlas);
ERR_FAIL_COND_V(!atlas, RID());
return atlas->depth_fb;
}
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
///////////////////////////////////////////////////////////
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
RID RendererSceneRenderRD::shadow_atlas_create() {
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
return shadow_atlas_owner.make_rid(ShadowAtlas());
}
Shadow map rendering optimization -All shadow rendering is done with raster now (no compute) -All shadow rendering is done by rendering directly to the shadow atlas -Improved how buffer clearing is done to optimize the above. -Ability to set shadows as 16 bits.
2021-01-24 19:00:20 +00:00
void RendererSceneRenderRD::_update_shadow_atlas(ShadowAtlas *shadow_atlas) {
if (shadow_atlas->size > 0 && shadow_atlas->depth.is_null()) {
RD::TextureFormat tf;
tf.format = shadow_atlas->use_16_bits ? RD::DATA_FORMAT_D16_UNORM : RD::DATA_FORMAT_D32_SFLOAT;
tf.width = shadow_atlas->size;
tf.height = shadow_atlas->size;
tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT;
shadow_atlas->depth = RD::get_singleton()->texture_create(tf, RD::TextureView());
Vector<RID> fb_tex;
fb_tex.push_back(shadow_atlas->depth);
shadow_atlas->fb = RD::get_singleton()->framebuffer_create(fb_tex);
}
}
void RendererSceneRenderRD::shadow_atlas_set_size(RID p_atlas, int p_size, bool p_16_bits) {
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
ShadowAtlas *shadow_atlas = shadow_atlas_owner.getornull(p_atlas);
ERR_FAIL_COND(!shadow_atlas);
ERR_FAIL_COND(p_size < 0);
p_size = next_power_of_2(p_size);
Shadow map rendering optimization -All shadow rendering is done with raster now (no compute) -All shadow rendering is done by rendering directly to the shadow atlas -Improved how buffer clearing is done to optimize the above. -Ability to set shadows as 16 bits.
2021-01-24 19:00:20 +00:00
if (p_size == shadow_atlas->size && p_16_bits == shadow_atlas->use_16_bits) {
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
return;
Style: Enforce braces around if blocks and loops Using clang-tidy's `readability-braces-around-statements`. https://clang.llvm.org/extra/clang-tidy/checks/readability-braces-around-statements.html
2020-05-14 14:41:43 +00:00
}
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
// erasing atlas
if (shadow_atlas->depth.is_valid()) {
RD::get_singleton()->free(shadow_atlas->depth);
shadow_atlas->depth = RID();
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
_clear_shadow_shrink_stages(shadow_atlas->shrink_stages);
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
}
for (int i = 0; i < 4; i++) {
//clear subdivisions
shadow_atlas->quadrants[i].shadows.resize(0);
shadow_atlas->quadrants[i].shadows.resize(1 << shadow_atlas->quadrants[i].subdivision);
}
//erase shadow atlas reference from lights
for (Map<RID, uint32_t>::Element *E = shadow_atlas->shadow_owners.front(); E; E = E->next()) {
LightInstance *li = light_instance_owner.getornull(E->key());
ERR_CONTINUE(!li);
li->shadow_atlases.erase(p_atlas);
}
//clear owners
shadow_atlas->shadow_owners.clear();
shadow_atlas->size = p_size;
Shadow map rendering optimization -All shadow rendering is done with raster now (no compute) -All shadow rendering is done by rendering directly to the shadow atlas -Improved how buffer clearing is done to optimize the above. -Ability to set shadows as 16 bits.
2021-01-24 19:00:20 +00:00
shadow_atlas->use_16_bits = p_size;
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
void RendererSceneRenderRD::shadow_atlas_set_quadrant_subdivision(RID p_atlas, int p_quadrant, int p_subdivision) {
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
ShadowAtlas *shadow_atlas = shadow_atlas_owner.getornull(p_atlas);
ERR_FAIL_COND(!shadow_atlas);
ERR_FAIL_INDEX(p_quadrant, 4);
ERR_FAIL_INDEX(p_subdivision, 16384);
uint32_t subdiv = next_power_of_2(p_subdivision);
if (subdiv & 0xaaaaaaaa) { //sqrt(subdiv) must be integer
subdiv <<= 1;
}
subdiv = int(Math::sqrt((float)subdiv));
//obtain the number that will be x*x
Style: Enforce braces around if blocks and loops Using clang-tidy's `readability-braces-around-statements`. https://clang.llvm.org/extra/clang-tidy/checks/readability-braces-around-statements.html
2020-05-14 14:41:43 +00:00
if (shadow_atlas->quadrants[p_quadrant].subdivision == subdiv) {
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
return;
Style: Enforce braces around if blocks and loops Using clang-tidy's `readability-braces-around-statements`. https://clang.llvm.org/extra/clang-tidy/checks/readability-braces-around-statements.html
2020-05-14 14:41:43 +00:00
}
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
//erase all data from quadrant
for (int i = 0; i < shadow_atlas->quadrants[p_quadrant].shadows.size(); i++) {
if (shadow_atlas->quadrants[p_quadrant].shadows[i].owner.is_valid()) {
shadow_atlas->shadow_owners.erase(shadow_atlas->quadrants[p_quadrant].shadows[i].owner);
LightInstance *li = light_instance_owner.getornull(shadow_atlas->quadrants[p_quadrant].shadows[i].owner);
ERR_CONTINUE(!li);
li->shadow_atlases.erase(p_atlas);
}
}
shadow_atlas->quadrants[p_quadrant].shadows.resize(0);
shadow_atlas->quadrants[p_quadrant].shadows.resize(subdiv * subdiv);
shadow_atlas->quadrants[p_quadrant].subdivision = subdiv;
//cache the smallest subdiv (for faster allocation in light update)
shadow_atlas->smallest_subdiv = 1 << 30;
for (int i = 0; i < 4; i++) {
if (shadow_atlas->quadrants[i].subdivision) {
shadow_atlas->smallest_subdiv = MIN(shadow_atlas->smallest_subdiv, shadow_atlas->quadrants[i].subdivision);
}
}
if (shadow_atlas->smallest_subdiv == 1 << 30) {
shadow_atlas->smallest_subdiv = 0;
}
//resort the size orders, simple bublesort for 4 elements..
int swaps = 0;
do {
swaps = 0;
for (int i = 0; i < 3; i++) {
if (shadow_atlas->quadrants[shadow_atlas->size_order[i]].subdivision < shadow_atlas->quadrants[shadow_atlas->size_order[i + 1]].subdivision) {
SWAP(shadow_atlas->size_order[i], shadow_atlas->size_order[i + 1]);
swaps++;
}
}
} while (swaps > 0);
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
bool RendererSceneRenderRD::_shadow_atlas_find_shadow(ShadowAtlas *shadow_atlas, int *p_in_quadrants, int p_quadrant_count, int p_current_subdiv, uint64_t p_tick, int &r_quadrant, int &r_shadow) {
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
for (int i = p_quadrant_count - 1; i >= 0; i--) {
int qidx = p_in_quadrants[i];
if (shadow_atlas->quadrants[qidx].subdivision == (uint32_t)p_current_subdiv) {
return false;
}
//look for an empty space
int sc = shadow_atlas->quadrants[qidx].shadows.size();
ShadowAtlas::Quadrant::Shadow *sarr = shadow_atlas->quadrants[qidx].shadows.ptrw();
int found_free_idx = -1; //found a free one
int found_used_idx = -1; //found existing one, must steal it
uint64_t min_pass = 0; // pass of the existing one, try to use the least recently used one (LRU fashion)
for (int j = 0; j < sc; j++) {
if (!sarr[j].owner.is_valid()) {
found_free_idx = j;
break;
}
LightInstance *sli = light_instance_owner.getornull(sarr[j].owner);
ERR_CONTINUE(!sli);
if (sli->last_scene_pass != scene_pass) {
//was just allocated, don't kill it so soon, wait a bit..
Style: Enforce braces around if blocks and loops Using clang-tidy's `readability-braces-around-statements`. https://clang.llvm.org/extra/clang-tidy/checks/readability-braces-around-statements.html
2020-05-14 14:41:43 +00:00
if (p_tick - sarr[j].alloc_tick < shadow_atlas_realloc_tolerance_msec) {
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
continue;
Style: Enforce braces around if blocks and loops Using clang-tidy's `readability-braces-around-statements`. https://clang.llvm.org/extra/clang-tidy/checks/readability-braces-around-statements.html
2020-05-14 14:41:43 +00:00
}
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
if (found_used_idx == -1 || sli->last_scene_pass < min_pass) {
found_used_idx = j;
min_pass = sli->last_scene_pass;
}
}
}
Style: Enforce braces around if blocks and loops Using clang-tidy's `readability-braces-around-statements`. https://clang.llvm.org/extra/clang-tidy/checks/readability-braces-around-statements.html
2020-05-14 14:41:43 +00:00
if (found_free_idx == -1 && found_used_idx == -1) {
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
continue; //nothing found
Style: Enforce braces around if blocks and loops Using clang-tidy's `readability-braces-around-statements`. https://clang.llvm.org/extra/clang-tidy/checks/readability-braces-around-statements.html
2020-05-14 14:41:43 +00:00
}
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
if (found_free_idx == -1 && found_used_idx != -1) {
found_free_idx = found_used_idx;
}
r_quadrant = qidx;
r_shadow = found_free_idx;
return true;
}
return false;
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
bool RendererSceneRenderRD::shadow_atlas_update_light(RID p_atlas, RID p_light_intance, float p_coverage, uint64_t p_light_version) {
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
ShadowAtlas *shadow_atlas = shadow_atlas_owner.getornull(p_atlas);
ERR_FAIL_COND_V(!shadow_atlas, false);
LightInstance *li = light_instance_owner.getornull(p_light_intance);
ERR_FAIL_COND_V(!li, false);
if (shadow_atlas->size == 0 || shadow_atlas->smallest_subdiv == 0) {
return false;
}
uint32_t quad_size = shadow_atlas->size >> 1;
int desired_fit = MIN(quad_size / shadow_atlas->smallest_subdiv, next_power_of_2(quad_size * p_coverage));
int valid_quadrants[4];
int valid_quadrant_count = 0;
int best_size = -1; //best size found
int best_subdiv = -1; //subdiv for the best size
//find the quadrants this fits into, and the best possible size it can fit into
for (int i = 0; i < 4; i++) {
int q = shadow_atlas->size_order[i];
int sd = shadow_atlas->quadrants[q].subdivision;
Style: Enforce braces around if blocks and loops Using clang-tidy's `readability-braces-around-statements`. https://clang.llvm.org/extra/clang-tidy/checks/readability-braces-around-statements.html
2020-05-14 14:41:43 +00:00
if (sd == 0) {
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
continue; //unused
Style: Enforce braces around if blocks and loops Using clang-tidy's `readability-braces-around-statements`. https://clang.llvm.org/extra/clang-tidy/checks/readability-braces-around-statements.html
2020-05-14 14:41:43 +00:00
}
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
int max_fit = quad_size / sd;
Style: Enforce braces around if blocks and loops Using clang-tidy's `readability-braces-around-statements`. https://clang.llvm.org/extra/clang-tidy/checks/readability-braces-around-statements.html
2020-05-14 14:41:43 +00:00
if (best_size != -1 && max_fit > best_size) {
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
break; //too large
Style: Enforce braces around if blocks and loops Using clang-tidy's `readability-braces-around-statements`. https://clang.llvm.org/extra/clang-tidy/checks/readability-braces-around-statements.html
2020-05-14 14:41:43 +00:00
}
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
valid_quadrants[valid_quadrant_count++] = q;
best_subdiv = sd;
if (max_fit >= desired_fit) {
best_size = max_fit;
}
}
ERR_FAIL_COND_V(valid_quadrant_count == 0, false);
uint64_t tick = OS::get_singleton()->get_ticks_msec();
//see if it already exists
if (shadow_atlas->shadow_owners.has(p_light_intance)) {
//it does!
uint32_t key = shadow_atlas->shadow_owners[p_light_intance];
uint32_t q = (key >> ShadowAtlas::QUADRANT_SHIFT) & 0x3;
uint32_t s = key & ShadowAtlas::SHADOW_INDEX_MASK;
bool should_realloc = shadow_atlas->quadrants[q].subdivision != (uint32_t)best_subdiv && (shadow_atlas->quadrants[q].shadows[s].alloc_tick - tick > shadow_atlas_realloc_tolerance_msec);
bool should_redraw = shadow_atlas->quadrants[q].shadows[s].version != p_light_version;
if (!should_realloc) {
shadow_atlas->quadrants[q].shadows.write[s].version = p_light_version;
//already existing, see if it should redraw or it's just OK
return should_redraw;
}
int new_quadrant, new_shadow;
//find a better place
if (_shadow_atlas_find_shadow(shadow_atlas, valid_quadrants, valid_quadrant_count, shadow_atlas->quadrants[q].subdivision, tick, new_quadrant, new_shadow)) {
//found a better place!
ShadowAtlas::Quadrant::Shadow *sh = &shadow_atlas->quadrants[new_quadrant].shadows.write[new_shadow];
if (sh->owner.is_valid()) {
//is taken, but is invalid, erasing it
shadow_atlas->shadow_owners.erase(sh->owner);
LightInstance *sli = light_instance_owner.getornull(sh->owner);
sli->shadow_atlases.erase(p_atlas);
}
//erase previous
shadow_atlas->quadrants[q].shadows.write[s].version = 0;
shadow_atlas->quadrants[q].shadows.write[s].owner = RID();
sh->owner = p_light_intance;
sh->alloc_tick = tick;
sh->version = p_light_version;
li->shadow_atlases.insert(p_atlas);
//make new key
key = new_quadrant << ShadowAtlas::QUADRANT_SHIFT;
key |= new_shadow;
//update it in map
shadow_atlas->shadow_owners[p_light_intance] = key;
//make it dirty, as it should redraw anyway
return true;
}
//no better place for this shadow found, keep current
//already existing, see if it should redraw or it's just OK
shadow_atlas->quadrants[q].shadows.write[s].version = p_light_version;
return should_redraw;
}
int new_quadrant, new_shadow;
//find a better place
if (_shadow_atlas_find_shadow(shadow_atlas, valid_quadrants, valid_quadrant_count, -1, tick, new_quadrant, new_shadow)) {
//found a better place!
ShadowAtlas::Quadrant::Shadow *sh = &shadow_atlas->quadrants[new_quadrant].shadows.write[new_shadow];
if (sh->owner.is_valid()) {
//is taken, but is invalid, erasing it
shadow_atlas->shadow_owners.erase(sh->owner);
LightInstance *sli = light_instance_owner.getornull(sh->owner);
sli->shadow_atlases.erase(p_atlas);
}
sh->owner = p_light_intance;
sh->alloc_tick = tick;
sh->version = p_light_version;
li->shadow_atlases.insert(p_atlas);
//make new key
uint32_t key = new_quadrant << ShadowAtlas::QUADRANT_SHIFT;
key |= new_shadow;
//update it in map
shadow_atlas->shadow_owners[p_light_intance] = key;
//make it dirty, as it should redraw anyway
return true;
}
//no place to allocate this light, apologies
return false;
}
Shadow map rendering optimization -All shadow rendering is done with raster now (no compute) -All shadow rendering is done by rendering directly to the shadow atlas -Improved how buffer clearing is done to optimize the above. -Ability to set shadows as 16 bits.
2021-01-24 19:00:20 +00:00
void RendererSceneRenderRD::_update_directional_shadow_atlas() {
if (directional_shadow.depth.is_null() && directional_shadow.size > 0) {
RD::TextureFormat tf;
tf.format = directional_shadow.use_16_bits ? RD::DATA_FORMAT_D16_UNORM : RD::DATA_FORMAT_D32_SFLOAT;
tf.width = directional_shadow.size;
tf.height = directional_shadow.size;
tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT;
directional_shadow.depth = RD::get_singleton()->texture_create(tf, RD::TextureView());
Vector<RID> fb_tex;
fb_tex.push_back(directional_shadow.depth);
directional_shadow.fb = RD::get_singleton()->framebuffer_create(fb_tex);
}
}
void RendererSceneRenderRD::directional_shadow_atlas_set_size(int p_size, bool p_16_bits) {
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
p_size = nearest_power_of_2_templated(p_size);
Shadow map rendering optimization -All shadow rendering is done with raster now (no compute) -All shadow rendering is done by rendering directly to the shadow atlas -Improved how buffer clearing is done to optimize the above. -Ability to set shadows as 16 bits.
2021-01-24 19:00:20 +00:00
if (directional_shadow.size == p_size && directional_shadow.use_16_bits == p_16_bits) {
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
return;
}
Directional lights and shadow mapping are functional.
2019-09-07 17:38:17 +00:00
directional_shadow.size = p_size;
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
if (directional_shadow.depth.is_valid()) {
RD::get_singleton()->free(directional_shadow.depth);
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
_clear_shadow_shrink_stages(directional_shadow.shrink_stages);
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
directional_shadow.depth = RID();
Shadow map rendering optimization -All shadow rendering is done with raster now (no compute) -All shadow rendering is done by rendering directly to the shadow atlas -Improved how buffer clearing is done to optimize the above. -Ability to set shadows as 16 bits.
2021-01-24 19:00:20 +00:00
_base_uniforms_changed();
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
}
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
void RendererSceneRenderRD::set_directional_shadow_count(int p_count) {
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
directional_shadow.light_count = p_count;
directional_shadow.current_light = 0;
}
Directional lights and shadow mapping are functional.
2019-09-07 17:38:17 +00:00
static Rect2i _get_directional_shadow_rect(int p_size, int p_shadow_count, int p_shadow_index) {
int split_h = 1;
int split_v = 1;
while (split_h * split_v < p_shadow_count) {
if (split_h == split_v) {
split_h <<= 1;
} else {
split_v <<= 1;
}
}
Rect2i rect(0, 0, p_size, p_size);
rect.size.width /= split_h;
rect.size.height /= split_v;
rect.position.x = rect.size.width * (p_shadow_index % split_h);
rect.position.y = rect.size.height * (p_shadow_index / split_h);
return rect;
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
int RendererSceneRenderRD::get_directional_light_shadow_size(RID p_light_intance) {
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
ERR_FAIL_COND_V(directional_shadow.light_count == 0, 0);
Directional lights and shadow mapping are functional.
2019-09-07 17:38:17 +00:00
Rect2i r = _get_directional_shadow_rect(directional_shadow.size, directional_shadow.light_count, 0);
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
LightInstance *light_instance = light_instance_owner.getornull(p_light_intance);
ERR_FAIL_COND_V(!light_instance, 0);
switch (storage->light_directional_get_shadow_mode(light_instance->light)) {
Renaming of servers for coherency. VisualServer -> RenderingServer PhysicsServer -> PhysicsServer3D Physics2DServer -> PhysicsServer2D NavigationServer -> NavigationServer3D Navigation2DServer -> NavigationServer2D Also renamed corresponding files.
2020-03-27 18:21:27 +00:00
case RS::LIGHT_DIRECTIONAL_SHADOW_ORTHOGONAL:
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
break; //none
Style: clang-format: Disable AllowShortCaseLabelsOnASingleLine Part of #33027.
2020-05-10 11:00:47 +00:00
case RS::LIGHT_DIRECTIONAL_SHADOW_PARALLEL_2_SPLITS:
r.size.height /= 2;
break;
case RS::LIGHT_DIRECTIONAL_SHADOW_PARALLEL_4_SPLITS:
r.size /= 2;
break;
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
}
Directional lights and shadow mapping are functional.
2019-09-07 17:38:17 +00:00
return MAX(r.size.width, r.size.height);
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
}
//////////////////////////////////////////////////
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
RID RendererSceneRenderRD::camera_effects_create() {
WIP CameraEffects implementation (bokeh not working for now)
2020-01-13 18:37:24 +00:00
return camera_effects_owner.make_rid(CameraEffects());
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
void RendererSceneRenderRD::camera_effects_set_dof_blur_quality(RS::DOFBlurQuality p_quality, bool p_use_jitter) {
DOF fully implemented, can be edited on the fly.
2020-01-16 00:23:21 +00:00
dof_blur_quality = p_quality;
dof_blur_use_jitter = p_use_jitter;
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
void RendererSceneRenderRD::camera_effects_set_dof_blur_bokeh_shape(RS::DOFBokehShape p_shape) {
DOF fully implemented, can be edited on the fly.
2020-01-16 00:23:21 +00:00
dof_blur_bokeh_shape = p_shape;
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
void RendererSceneRenderRD::camera_effects_set_dof_blur(RID p_camera_effects, bool p_far_enable, float p_far_distance, float p_far_transition, bool p_near_enable, float p_near_distance, float p_near_transition, float p_amount) {
WIP CameraEffects implementation (bokeh not working for now)
2020-01-13 18:37:24 +00:00
CameraEffects *camfx = camera_effects_owner.getornull(p_camera_effects);
ERR_FAIL_COND(!camfx);
camfx->dof_blur_far_enabled = p_far_enable;
camfx->dof_blur_far_distance = p_far_distance;
camfx->dof_blur_far_transition = p_far_transition;
camfx->dof_blur_near_enabled = p_near_enable;
camfx->dof_blur_near_distance = p_near_distance;
camfx->dof_blur_near_transition = p_near_transition;
camfx->dof_blur_amount = p_amount;
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
void RendererSceneRenderRD::camera_effects_set_custom_exposure(RID p_camera_effects, bool p_enable, float p_exposure) {
WIP CameraEffects implementation (bokeh not working for now)
2020-01-13 18:37:24 +00:00
CameraEffects *camfx = camera_effects_owner.getornull(p_camera_effects);
ERR_FAIL_COND(!camfx);
camfx->override_exposure_enabled = p_enable;
camfx->override_exposure = p_exposure;
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
RID RendererSceneRenderRD::light_instance_create(RID p_light) {
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
RID li = light_instance_owner.make_rid(LightInstance());
LightInstance *light_instance = light_instance_owner.getornull(li);
light_instance->self = li;
light_instance->light = p_light;
light_instance->light_type = storage->light_get_type(p_light);
return li;
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
void RendererSceneRenderRD::light_instance_set_transform(RID p_light_instance, const Transform &p_transform) {
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
LightInstance *light_instance = light_instance_owner.getornull(p_light_instance);
ERR_FAIL_COND(!light_instance);
light_instance->transform = p_transform;
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
void RendererSceneRenderRD::light_instance_set_aabb(RID p_light_instance, const AABB &p_aabb) {
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
LightInstance *light_instance = light_instance_owner.getornull(p_light_instance);
ERR_FAIL_COND(!light_instance);
light_instance->aabb = p_aabb;
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
void RendererSceneRenderRD::light_instance_set_shadow_transform(RID p_light_instance, const CameraMatrix &p_projection, const Transform &p_transform, float p_far, float p_split, int p_pass, float p_shadow_texel_size, float p_bias_scale, float p_range_begin, const Vector2 &p_uv_scale) {
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
LightInstance *light_instance = light_instance_owner.getornull(p_light_instance);
ERR_FAIL_COND(!light_instance);
Renaming of servers for coherency. VisualServer -> RenderingServer PhysicsServer -> PhysicsServer3D Physics2DServer -> PhysicsServer2D NavigationServer -> NavigationServer3D Navigation2DServer -> NavigationServer2D Also renamed corresponding files.
2020-03-27 18:21:27 +00:00
if (storage->light_get_type(light_instance->light) != RS::LIGHT_DIRECTIONAL) {
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
p_pass = 0;
}
ERR_FAIL_INDEX(p_pass, 4);
light_instance->shadow_transform[p_pass].camera = p_projection;
light_instance->shadow_transform[p_pass].transform = p_transform;
light_instance->shadow_transform[p_pass].farplane = p_far;
light_instance->shadow_transform[p_pass].split = p_split;
light_instance->shadow_transform[p_pass].bias_scale = p_bias_scale;
Support light size and soft shadows
2020-04-09 18:11:15 +00:00
light_instance->shadow_transform[p_pass].range_begin = p_range_begin;
Refactored shadowmapping. - Made shadow bias size independent, so it will remain when changing light or camera size. - Implemented normal offset bias, which greatly enhances quality. - Added transmission to subsurface scattering - Reimplemented shadow filter modes Closes #17260
2020-04-08 01:51:52 +00:00
light_instance->shadow_transform[p_pass].shadow_texel_size = p_shadow_texel_size;
Support light size and soft shadows
2020-04-09 18:11:15 +00:00
light_instance->shadow_transform[p_pass].uv_scale = p_uv_scale;
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
void RendererSceneRenderRD::light_instance_mark_visible(RID p_light_instance) {
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
LightInstance *light_instance = light_instance_owner.getornull(p_light_instance);
ERR_FAIL_COND(!light_instance);
light_instance->last_scene_pass = scene_pass;
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
RendererSceneRenderRD::ShadowCubemap *RendererSceneRenderRD::_get_shadow_cubemap(int p_size) {
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
if (!shadow_cubemaps.has(p_size)) {
ShadowCubemap sc;
{
RD::TextureFormat tf;
tf.format = RD::get_singleton()->texture_is_format_supported_for_usage(RD::DATA_FORMAT_D32_SFLOAT, RD::TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT) ? RD::DATA_FORMAT_D32_SFLOAT : RD::DATA_FORMAT_X8_D24_UNORM_PACK32;
tf.width = p_size;
tf.height = p_size;
Rename RD texture "type" to "texture_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-11-27 03:50:05 +00:00
tf.texture_type = RD::TEXTURE_TYPE_CUBE;
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
tf.array_layers = 6;
tf.usage_bits = RD::TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT | RD::TEXTURE_USAGE_SAMPLING_BIT;
sc.cubemap = RD::get_singleton()->texture_create(tf, RD::TextureView());
}
for (int i = 0; i < 6; i++) {
RID side_texture = RD::get_singleton()->texture_create_shared_from_slice(RD::TextureView(), sc.cubemap, i, 0);
Vector<RID> fbtex;
fbtex.push_back(side_texture);
sc.side_fb[i] = RD::get_singleton()->framebuffer_create(fbtex);
}
shadow_cubemaps[p_size] = sc;
}
return &shadow_cubemaps[p_size];
}
Implement decals Also implemented decal atlas, so projectors and other stuff can be added. Sidenote: Had to make RID hashable, so some unrelated includes changed in order to include it in hashfuncs.h
2020-04-14 03:05:21 +00:00
//////////////////////////
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
RID RendererSceneRenderRD::decal_instance_create(RID p_decal) {
Implement decals Also implemented decal atlas, so projectors and other stuff can be added. Sidenote: Had to make RID hashable, so some unrelated includes changed in order to include it in hashfuncs.h
2020-04-14 03:05:21 +00:00
DecalInstance di;
di.decal = p_decal;
return decal_instance_owner.make_rid(di);
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
void RendererSceneRenderRD::decal_instance_set_transform(RID p_decal, const Transform &p_transform) {
Implement decals Also implemented decal atlas, so projectors and other stuff can be added. Sidenote: Had to make RID hashable, so some unrelated includes changed in order to include it in hashfuncs.h
2020-04-14 03:05:21 +00:00
DecalInstance *di = decal_instance_owner.getornull(p_decal);
ERR_FAIL_COND(!di);
di->transform = p_transform;
}
GIProbes working.
2019-10-03 20:39:08 +00:00
/////////////////////////////////
Reorganize renderer code. So it can hopefully be made more cache efficient afterwards.
2020-12-31 12:42:56 +00:00
RID RendererSceneRenderRD::lightmap_instance_create(RID p_lightmap) {
LightmapInstance li;
li.lightmap = p_lightmap;
return lightmap_instance_owner.make_rid(li);
}
void RendererSceneRenderRD::lightmap_instance_set_transform(RID p_lightmap, const Transform &p_transform) {
LightmapInstance *li = lightmap_instance_owner.getornull(p_lightmap);
ERR_FAIL_COND(!li);
li->transform = p_transform;
}
/////////////////////////////////
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
RID RendererSceneRenderRD::gi_probe_instance_create(RID p_base) {
GIProbes working.
2019-10-03 20:39:08 +00:00
GIProbeInstance gi_probe;
gi_probe.probe = p_base;
RID rid = gi_probe_instance_owner.make_rid(gi_probe);
return rid;
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
void RendererSceneRenderRD::gi_probe_instance_set_transform_to_data(RID p_probe, const Transform &p_xform) {
GIProbes working.
2019-10-03 20:39:08 +00:00
GIProbeInstance *gi_probe = gi_probe_instance_owner.getornull(p_probe);
ERR_FAIL_COND(!gi_probe);
gi_probe->transform = p_xform;
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
bool RendererSceneRenderRD::gi_probe_needs_update(RID p_probe) const {
GIProbes working.
2019-10-03 20:39:08 +00:00
GIProbeInstance *gi_probe = gi_probe_instance_owner.getornull(p_probe);
ERR_FAIL_COND_V(!gi_probe, false);
Add support for low-end 3D rendering. -Reduce number of uniform sets from 6 to 4. -Remove features in low end mode, in order to reduce the number of texture units fit to 16.
2020-12-07 21:27:38 +00:00
if (low_end) {
return false;
}
GIProbes working.
2019-10-03 20:39:08 +00:00
//return true;
return gi_probe->last_probe_version != storage->gi_probe_get_version(gi_probe->probe);
}
Reorganize renderer code. So it can hopefully be made more cache efficient afterwards.
2020-12-31 12:42:56 +00:00
void RendererSceneRenderRD::gi_probe_update(RID p_probe, bool p_update_light_instances, const Vector<RID> &p_light_instances, const PagedArray<GeometryInstance *> &p_dynamic_objects) {
GIProbes working.
2019-10-03 20:39:08 +00:00
GIProbeInstance *gi_probe = gi_probe_instance_owner.getornull(p_probe);
ERR_FAIL_COND(!gi_probe);
Add support for low-end 3D rendering. -Reduce number of uniform sets from 6 to 4. -Remove features in low end mode, in order to reduce the number of texture units fit to 16.
2020-12-07 21:27:38 +00:00
if (low_end) {
return;
}
GIProbes working.
2019-10-03 20:39:08 +00:00
uint32_t data_version = storage->gi_probe_get_data_version(gi_probe->probe);
// (RE)CREATE IF NEEDED
if (gi_probe->last_probe_data_version != data_version) {
//need to re-create everything
if (gi_probe->texture.is_valid()) {
RD::get_singleton()->free(gi_probe->texture);
RD::get_singleton()->free(gi_probe->write_buffer);
gi_probe->mipmaps.clear();
}
Dynamic object support for GI Probes (a bit buggy still)
2019-10-11 02:14:56 +00:00
for (int i = 0; i < gi_probe->dynamic_maps.size(); i++) {
RD::get_singleton()->free(gi_probe->dynamic_maps[i].texture);
RD::get_singleton()->free(gi_probe->dynamic_maps[i].depth);
}
gi_probe->dynamic_maps.clear();
GIProbes working.
2019-10-03 20:39:08 +00:00
Vector3i octree_size = storage->gi_probe_get_octree_size(gi_probe->probe);
if (octree_size != Vector3i()) {
//can create a 3D texture
PoolVector is gone, replaced by Vector Typed `PoolTypeArray` types are now renamed `PackedTypeArray` and are sugar for `Vector<Type>`.
2020-02-17 21:06:54 +00:00
Vector<int> levels = storage->gi_probe_get_level_counts(gi_probe->probe);
GIProbes working.
2019-10-03 20:39:08 +00:00
RD::TextureFormat tf;
tf.format = RD::DATA_FORMAT_R8G8B8A8_UNORM;
tf.width = octree_size.x;
tf.height = octree_size.y;
tf.depth = octree_size.z;
Rename RD texture "type" to "texture_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-11-27 03:50:05 +00:00
tf.texture_type = RD::TEXTURE_TYPE_3D;
GIProbes working.
2019-10-03 20:39:08 +00:00
tf.mipmaps = levels.size();
tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_STORAGE_BIT | RD::TEXTURE_USAGE_CAN_COPY_TO_BIT;
gi_probe->texture = RD::get_singleton()->texture_create(tf, RD::TextureView());
Reorganize RenderingDevice barriers -Removed sync to draw, now everything syncs to draw by default. -Fixed many validation layer errors. -Added support for VkImageViewUsageCreateInfo to fix validation layer warnings. -Texture, buffer, raster and compute functions now all allow spcifying which barriers will be used.
2021-01-26 00:52:58 +00:00
RD::get_singleton()->texture_clear(gi_probe->texture, Color(0, 0, 0, 0), 0, levels.size(), 0, 1);
GIProbes working.
2019-10-03 20:39:08 +00:00
{
int total_elements = 0;
for (int i = 0; i < levels.size(); i++) {
total_elements += levels[i];
}
gi_probe->write_buffer = RD::get_singleton()->storage_buffer_create(total_elements * 16);
}
for (int i = 0; i < levels.size(); i++) {
GIProbeInstance::Mipmap mipmap;
mipmap.texture = RD::get_singleton()->texture_create_shared_from_slice(RD::TextureView(), gi_probe->texture, 0, i, RD::TEXTURE_SLICE_3D);
mipmap.level = levels.size() - i - 1;
mipmap.cell_offset = 0;
for (uint32_t j = 0; j < mipmap.level; j++) {
mipmap.cell_offset += levels[j];
}
mipmap.cell_count = levels[mipmap.level];
Vector<RD::Uniform> uniforms;
{
RD::Uniform u;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER;
GIProbes working.
2019-10-03 20:39:08 +00:00
u.binding = 1;
u.ids.push_back(storage->gi_probe_get_octree_buffer(gi_probe->probe));
uniforms.push_back(u);
}
{
RD::Uniform u;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER;
GIProbes working.
2019-10-03 20:39:08 +00:00
u.binding = 2;
u.ids.push_back(storage->gi_probe_get_data_buffer(gi_probe->probe));
uniforms.push_back(u);
}
{
RD::Uniform u;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER;
GIProbes working.
2019-10-03 20:39:08 +00:00
u.binding = 4;
u.ids.push_back(gi_probe->write_buffer);
uniforms.push_back(u);
}
Dynamic object support for GI Probes (a bit buggy still)
2019-10-11 02:14:56 +00:00
{
RD::Uniform u;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_TEXTURE;
Dynamic object support for GI Probes (a bit buggy still)
2019-10-11 02:14:56 +00:00
u.binding = 9;
u.ids.push_back(storage->gi_probe_get_sdf_texture(gi_probe->probe));
uniforms.push_back(u);
}
{
RD::Uniform u;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_SAMPLER;
Dynamic object support for GI Probes (a bit buggy still)
2019-10-11 02:14:56 +00:00
u.binding = 10;
Renaming of servers for coherency. VisualServer -> RenderingServer PhysicsServer -> PhysicsServer3D Physics2DServer -> PhysicsServer2D NavigationServer -> NavigationServer3D Navigation2DServer -> NavigationServer2D Also renamed corresponding files.
2020-03-27 18:21:27 +00:00
u.ids.push_back(storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED));
Dynamic object support for GI Probes (a bit buggy still)
2019-10-11 02:14:56 +00:00
uniforms.push_back(u);
}
GIProbes working.
2019-10-03 20:39:08 +00:00
{
Vector<RD::Uniform> copy_uniforms = uniforms;
if (i == 0) {
{
RD::Uniform u;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER;
GIProbes working.
2019-10-03 20:39:08 +00:00
u.binding = 3;
u.ids.push_back(gi_probe_lights_uniform);
copy_uniforms.push_back(u);
}
mipmap.uniform_set = RD::get_singleton()->uniform_set_create(copy_uniforms, giprobe_lighting_shader_version_shaders[GI_PROBE_SHADER_VERSION_COMPUTE_LIGHT], 0);
copy_uniforms = uniforms; //restore
{
RD::Uniform u;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_TEXTURE;
GIProbes working.
2019-10-03 20:39:08 +00:00
u.binding = 5;
u.ids.push_back(gi_probe->texture);
copy_uniforms.push_back(u);
}
mipmap.second_bounce_uniform_set = RD::get_singleton()->uniform_set_create(copy_uniforms, giprobe_lighting_shader_version_shaders[GI_PROBE_SHADER_VERSION_COMPUTE_SECOND_BOUNCE], 0);
} else {
mipmap.uniform_set = RD::get_singleton()->uniform_set_create(copy_uniforms, giprobe_lighting_shader_version_shaders[GI_PROBE_SHADER_VERSION_COMPUTE_MIPMAP], 0);
}
}
{
RD::Uniform u;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_IMAGE;
GIProbes working.
2019-10-03 20:39:08 +00:00
u.binding = 5;
u.ids.push_back(mipmap.texture);
uniforms.push_back(u);
}
mipmap.write_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, giprobe_lighting_shader_version_shaders[GI_PROBE_SHADER_VERSION_WRITE_TEXTURE], 0);
gi_probe->mipmaps.push_back(mipmap);
}
Dynamic object support for GI Probes (a bit buggy still)
2019-10-11 02:14:56 +00:00
{
uint32_t dynamic_map_size = MAX(MAX(octree_size.x, octree_size.y), octree_size.z);
uint32_t oversample = nearest_power_of_2_templated(4);
int mipmap_index = 0;
while (mipmap_index < gi_probe->mipmaps.size()) {
GIProbeInstance::DynamicMap dmap;
if (oversample > 0) {
dmap.size = dynamic_map_size * (1 << oversample);
dmap.mipmap = -1;
oversample--;
} else {
dmap.size = dynamic_map_size >> mipmap_index;
dmap.mipmap = mipmap_index;
mipmap_index++;
}
RD::TextureFormat dtf;
dtf.width = dmap.size;
dtf.height = dmap.size;
dtf.format = RD::DATA_FORMAT_R16G16B16A16_SFLOAT;
dtf.usage_bits = RD::TEXTURE_USAGE_STORAGE_BIT;
if (gi_probe->dynamic_maps.size() == 0) {
dtf.usage_bits |= RD::TEXTURE_USAGE_COLOR_ATTACHMENT_BIT;
}
dmap.texture = RD::get_singleton()->texture_create(dtf, RD::TextureView());
if (gi_probe->dynamic_maps.size() == 0) {
//render depth for first one
dtf.format = RD::get_singleton()->texture_is_format_supported_for_usage(RD::DATA_FORMAT_D32_SFLOAT, RD::TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT) ? RD::DATA_FORMAT_D32_SFLOAT : RD::DATA_FORMAT_X8_D24_UNORM_PACK32;
dtf.usage_bits = RD::TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT;
dmap.fb_depth = RD::get_singleton()->texture_create(dtf, RD::TextureView());
}
//just use depth as-is
dtf.format = RD::DATA_FORMAT_R32_SFLOAT;
dtf.usage_bits = RD::TEXTURE_USAGE_STORAGE_BIT | RD::TEXTURE_USAGE_COLOR_ATTACHMENT_BIT;
dmap.depth = RD::get_singleton()->texture_create(dtf, RD::TextureView());
if (gi_probe->dynamic_maps.size() == 0) {
dtf.format = RD::DATA_FORMAT_R8G8B8A8_UNORM;
dtf.usage_bits = RD::TEXTURE_USAGE_STORAGE_BIT | RD::TEXTURE_USAGE_COLOR_ATTACHMENT_BIT;
dmap.albedo = RD::get_singleton()->texture_create(dtf, RD::TextureView());
dmap.normal = RD::get_singleton()->texture_create(dtf, RD::TextureView());
dmap.orm = RD::get_singleton()->texture_create(dtf, RD::TextureView());
Vector<RID> fb;
fb.push_back(dmap.albedo);
fb.push_back(dmap.normal);
fb.push_back(dmap.orm);
fb.push_back(dmap.texture); //emission
fb.push_back(dmap.depth);
fb.push_back(dmap.fb_depth);
dmap.fb = RD::get_singleton()->framebuffer_create(fb);
{
Vector<RD::Uniform> uniforms;
{
RD::Uniform u;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER;
Dynamic object support for GI Probes (a bit buggy still)
2019-10-11 02:14:56 +00:00
u.binding = 3;
u.ids.push_back(gi_probe_lights_uniform);
uniforms.push_back(u);
}
{
RD::Uniform u;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_IMAGE;
Dynamic object support for GI Probes (a bit buggy still)
2019-10-11 02:14:56 +00:00
u.binding = 5;
u.ids.push_back(dmap.albedo);
uniforms.push_back(u);
}
{
RD::Uniform u;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_IMAGE;
Dynamic object support for GI Probes (a bit buggy still)
2019-10-11 02:14:56 +00:00
u.binding = 6;
u.ids.push_back(dmap.normal);
uniforms.push_back(u);
}
{
RD::Uniform u;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_IMAGE;
Dynamic object support for GI Probes (a bit buggy still)
2019-10-11 02:14:56 +00:00
u.binding = 7;
u.ids.push_back(dmap.orm);
uniforms.push_back(u);
}
{
RD::Uniform u;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_TEXTURE;
Dynamic object support for GI Probes (a bit buggy still)
2019-10-11 02:14:56 +00:00
u.binding = 8;
u.ids.push_back(dmap.fb_depth);
uniforms.push_back(u);
}
{
RD::Uniform u;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_TEXTURE;
Dynamic object support for GI Probes (a bit buggy still)
2019-10-11 02:14:56 +00:00
u.binding = 9;
u.ids.push_back(storage->gi_probe_get_sdf_texture(gi_probe->probe));
uniforms.push_back(u);
}
{
RD::Uniform u;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_SAMPLER;
Dynamic object support for GI Probes (a bit buggy still)
2019-10-11 02:14:56 +00:00
u.binding = 10;
Renaming of servers for coherency. VisualServer -> RenderingServer PhysicsServer -> PhysicsServer3D Physics2DServer -> PhysicsServer2D NavigationServer -> NavigationServer3D Navigation2DServer -> NavigationServer2D Also renamed corresponding files.
2020-03-27 18:21:27 +00:00
u.ids.push_back(storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED));
Dynamic object support for GI Probes (a bit buggy still)
2019-10-11 02:14:56 +00:00
uniforms.push_back(u);
}
{
RD::Uniform u;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_IMAGE;
Dynamic object support for GI Probes (a bit buggy still)
2019-10-11 02:14:56 +00:00
u.binding = 11;
u.ids.push_back(dmap.texture);
uniforms.push_back(u);
}
{
RD::Uniform u;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_IMAGE;
Dynamic object support for GI Probes (a bit buggy still)
2019-10-11 02:14:56 +00:00
u.binding = 12;
u.ids.push_back(dmap.depth);
uniforms.push_back(u);
}
dmap.uniform_set = RD::get_singleton()->uniform_set_create(uniforms, giprobe_lighting_shader_version_shaders[GI_PROBE_SHADER_VERSION_DYNAMIC_OBJECT_LIGHTING], 0);
}
} else {
bool plot = dmap.mipmap >= 0;
bool write = dmap.mipmap < (gi_probe->mipmaps.size() - 1);
Vector<RD::Uniform> uniforms;
{
RD::Uniform u;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_IMAGE;
Dynamic object support for GI Probes (a bit buggy still)
2019-10-11 02:14:56 +00:00
u.binding = 5;
u.ids.push_back(gi_probe->dynamic_maps[gi_probe->dynamic_maps.size() - 1].texture);
uniforms.push_back(u);
}
{
RD::Uniform u;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_IMAGE;
Dynamic object support for GI Probes (a bit buggy still)
2019-10-11 02:14:56 +00:00
u.binding = 6;
u.ids.push_back(gi_probe->dynamic_maps[gi_probe->dynamic_maps.size() - 1].depth);
uniforms.push_back(u);
}
if (write) {
{
RD::Uniform u;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_IMAGE;
Dynamic object support for GI Probes (a bit buggy still)
2019-10-11 02:14:56 +00:00
u.binding = 7;
u.ids.push_back(dmap.texture);
uniforms.push_back(u);
}
{
RD::Uniform u;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_IMAGE;
Dynamic object support for GI Probes (a bit buggy still)
2019-10-11 02:14:56 +00:00
u.binding = 8;
u.ids.push_back(dmap.depth);
uniforms.push_back(u);
}
}
{
RD::Uniform u;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_TEXTURE;
Dynamic object support for GI Probes (a bit buggy still)
2019-10-11 02:14:56 +00:00
u.binding = 9;
u.ids.push_back(storage->gi_probe_get_sdf_texture(gi_probe->probe));
uniforms.push_back(u);
}
{
RD::Uniform u;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_SAMPLER;
Dynamic object support for GI Probes (a bit buggy still)
2019-10-11 02:14:56 +00:00
u.binding = 10;
Renaming of servers for coherency. VisualServer -> RenderingServer PhysicsServer -> PhysicsServer3D Physics2DServer -> PhysicsServer2D NavigationServer -> NavigationServer3D Navigation2DServer -> NavigationServer2D Also renamed corresponding files.
2020-03-27 18:21:27 +00:00
u.ids.push_back(storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED));
Dynamic object support for GI Probes (a bit buggy still)
2019-10-11 02:14:56 +00:00
uniforms.push_back(u);
}
if (plot) {
{
RD::Uniform u;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_IMAGE;
Dynamic object support for GI Probes (a bit buggy still)
2019-10-11 02:14:56 +00:00
u.binding = 11;
u.ids.push_back(gi_probe->mipmaps[dmap.mipmap].texture);
uniforms.push_back(u);
}
}
CI: Update to clang-format 11 and apply ternary operator changes
2021-01-12 15:57:55 +00:00
dmap.uniform_set = RD::get_singleton()->uniform_set_create(
uniforms,
giprobe_lighting_shader_version_shaders[(write && plot) ? GI_PROBE_SHADER_VERSION_DYNAMIC_SHRINK_WRITE_PLOT : (write ? GI_PROBE_SHADER_VERSION_DYNAMIC_SHRINK_WRITE : GI_PROBE_SHADER_VERSION_DYNAMIC_SHRINK_PLOT)],
0);
Dynamic object support for GI Probes (a bit buggy still)
2019-10-11 02:14:56 +00:00
}
gi_probe->dynamic_maps.push_back(dmap);
}
}
GIProbes working.
2019-10-03 20:39:08 +00:00
}
gi_probe->last_probe_data_version = data_version;
Dynamic object support for GI Probes (a bit buggy still)
2019-10-11 02:14:56 +00:00
p_update_light_instances = true; //just in case
Re-implemented screen space ambient occlusion
2020-01-25 10:18:55 +00:00
_base_uniforms_changed();
GIProbes working.
2019-10-03 20:39:08 +00:00
}
// UDPDATE TIME
Fix code formatting issues and VS compilation Also temporarily disable multicheck build so that we get a full build even when there are style issues on Vulkan. Fixes #33356.
2019-11-05 11:01:00 +00:00
if (gi_probe->has_dynamic_object_data) {
Dynamic object support for GI Probes (a bit buggy still)
2019-10-11 02:14:56 +00:00
//if it has dynamic object data, it needs to be cleared
Reorganize RenderingDevice barriers -Removed sync to draw, now everything syncs to draw by default. -Fixed many validation layer errors. -Added support for VkImageViewUsageCreateInfo to fix validation layer warnings. -Texture, buffer, raster and compute functions now all allow spcifying which barriers will be used.
2021-01-26 00:52:58 +00:00
RD::get_singleton()->texture_clear(gi_probe->texture, Color(0, 0, 0, 0), 0, gi_probe->mipmaps.size(), 0, 1);
Dynamic object support for GI Probes (a bit buggy still)
2019-10-11 02:14:56 +00:00
}
GIProbes working.
2019-10-03 20:39:08 +00:00
Dynamic object support for GI Probes (a bit buggy still)
2019-10-11 02:14:56 +00:00
uint32_t light_count = 0;
GIProbes working.
2019-10-03 20:39:08 +00:00
Replace Octree by DynamicBVH in cull code -Much greater pairing/unpairing performance -For now, using it for culling too, but this will change in a couple of days. -Added a paged allocator, to efficiently alloc/free some types of objects.
2020-12-23 16:52:58 +00:00
if (p_update_light_instances || p_dynamic_objects.size() > 0) {
Dynamic object support for GI Probes (a bit buggy still)
2019-10-11 02:14:56 +00:00
light_count = MIN(gi_probe_max_lights, (uint32_t)p_light_instances.size());
GIProbes working.
2019-10-03 20:39:08 +00:00
Dynamic object support for GI Probes (a bit buggy still)
2019-10-11 02:14:56 +00:00
{
Transform to_cell = storage->gi_probe_get_to_cell_xform(gi_probe->probe);
Transform to_probe_xform = (gi_probe->transform * to_cell.affine_inverse()).affine_inverse();
//update lights
GIProbes working.
2019-10-03 20:39:08 +00:00
Dynamic object support for GI Probes (a bit buggy still)
2019-10-11 02:14:56 +00:00
for (uint32_t i = 0; i < light_count; i++) {
GIProbeLight &l = gi_probe_lights[i];
RID light_instance = p_light_instances[i];
RID light = light_instance_get_base_light(light_instance);
GIProbes working.
2019-10-03 20:39:08 +00:00
Dynamic object support for GI Probes (a bit buggy still)
2019-10-11 02:14:56 +00:00
l.type = storage->light_get_type(light);
Add sky_only setting to DirectionalLight3Ds
2020-10-22 06:22:53 +00:00
if (l.type == RS::LIGHT_DIRECTIONAL && storage->light_directional_is_sky_only(light)) {
light_count--;
continue;
}
Renaming of servers for coherency. VisualServer -> RenderingServer PhysicsServer -> PhysicsServer3D Physics2DServer -> PhysicsServer2D NavigationServer -> NavigationServer3D Navigation2DServer -> NavigationServer2D Also renamed corresponding files.
2020-03-27 18:21:27 +00:00
l.attenuation = storage->light_get_param(light, RS::LIGHT_PARAM_ATTENUATION);
l.energy = storage->light_get_param(light, RS::LIGHT_PARAM_ENERGY) * storage->light_get_param(light, RS::LIGHT_PARAM_INDIRECT_ENERGY);
l.radius = to_cell.basis.xform(Vector3(storage->light_get_param(light, RS::LIGHT_PARAM_RANGE), 0, 0)).length();
Dynamic object support for GI Probes (a bit buggy still)
2019-10-11 02:14:56 +00:00
Color color = storage->light_get_color(light).to_linear();
l.color[0] = color.r;
l.color[1] = color.g;
l.color[2] = color.b;
GIProbes working.
2019-10-03 20:39:08 +00:00
Renaming of servers for coherency. VisualServer -> RenderingServer PhysicsServer -> PhysicsServer3D Physics2DServer -> PhysicsServer2D NavigationServer -> NavigationServer3D Navigation2DServer -> NavigationServer2D Also renamed corresponding files.
2020-03-27 18:21:27 +00:00
l.spot_angle_radians = Math::deg2rad(storage->light_get_param(light, RS::LIGHT_PARAM_SPOT_ANGLE));
l.spot_attenuation = storage->light_get_param(light, RS::LIGHT_PARAM_SPOT_ATTENUATION);
GIProbes working.
2019-10-03 20:39:08 +00:00
Dynamic object support for GI Probes (a bit buggy still)
2019-10-11 02:14:56 +00:00
Transform xform = light_instance_get_base_transform(light_instance);
Vector3 pos = to_probe_xform.xform(xform.origin);
Vector3 dir = to_probe_xform.basis.xform(-xform.basis.get_axis(2)).normalized();
GIProbes working.
2019-10-03 20:39:08 +00:00
Dynamic object support for GI Probes (a bit buggy still)
2019-10-11 02:14:56 +00:00
l.position[0] = pos.x;
l.position[1] = pos.y;
l.position[2] = pos.z;
l.direction[0] = dir.x;
l.direction[1] = dir.y;
l.direction[2] = dir.z;
l.has_shadow = storage->light_has_shadow(light);
}
Reorganize RenderingDevice barriers -Removed sync to draw, now everything syncs to draw by default. -Fixed many validation layer errors. -Added support for VkImageViewUsageCreateInfo to fix validation layer warnings. -Texture, buffer, raster and compute functions now all allow spcifying which barriers will be used.
2021-01-26 00:52:58 +00:00
RD::get_singleton()->buffer_update(gi_probe_lights_uniform, 0, sizeof(GIProbeLight) * light_count, gi_probe_lights);
Dynamic object support for GI Probes (a bit buggy still)
2019-10-11 02:14:56 +00:00
}
GIProbes working.
2019-10-03 20:39:08 +00:00
}
Replace Octree by DynamicBVH in cull code -Much greater pairing/unpairing performance -For now, using it for culling too, but this will change in a couple of days. -Added a paged allocator, to efficiently alloc/free some types of objects.
2020-12-23 16:52:58 +00:00
if (gi_probe->has_dynamic_object_data || p_update_light_instances || p_dynamic_objects.size()) {
Dynamic object support for GI Probes (a bit buggy still)
2019-10-11 02:14:56 +00:00
// PROCESS MIPMAPS
if (gi_probe->mipmaps.size()) {
//can update mipmaps
Vector3i probe_size = storage->gi_probe_get_octree_size(gi_probe->probe);
GIProbePushConstant push_constant;
push_constant.limits[0] = probe_size.x;
push_constant.limits[1] = probe_size.y;
push_constant.limits[2] = probe_size.z;
push_constant.stack_size = gi_probe->mipmaps.size();
push_constant.emission_scale = 1.0;
push_constant.propagation = storage->gi_probe_get_propagation(gi_probe->probe);
push_constant.dynamic_range = storage->gi_probe_get_dynamic_range(gi_probe->probe);
push_constant.light_count = light_count;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
push_constant.aniso_strength = 0;
Dynamic object support for GI Probes (a bit buggy still)
2019-10-11 02:14:56 +00:00
/* print_line("probe update to version " + itos(gi_probe->last_probe_version));
print_line("propagation " + rtos(push_constant.propagation));
print_line("dynrange " + rtos(push_constant.dynamic_range));
*/
RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin();
int passes;
if (p_update_light_instances) {
passes = storage->gi_probe_is_using_two_bounces(gi_probe->probe) ? 2 : 1;
} else {
passes = 1; //only re-blitting is necessary
}
int wg_size = 64;
int wg_limit_x = RD::get_singleton()->limit_get(RD::LIMIT_MAX_COMPUTE_WORKGROUP_COUNT_X);
for (int pass = 0; pass < passes; pass++) {
if (p_update_light_instances) {
for (int i = 0; i < gi_probe->mipmaps.size(); i++) {
if (i == 0) {
RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, giprobe_lighting_shader_version_pipelines[pass == 0 ? GI_PROBE_SHADER_VERSION_COMPUTE_LIGHT : GI_PROBE_SHADER_VERSION_COMPUTE_SECOND_BOUNCE]);
} else if (i == 1) {
RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, giprobe_lighting_shader_version_pipelines[GI_PROBE_SHADER_VERSION_COMPUTE_MIPMAP]);
}
if (pass == 1 || i > 0) {
RD::get_singleton()->compute_list_add_barrier(compute_list); //wait til previous step is done
}
if (pass == 0 || i > 0) {
RD::get_singleton()->compute_list_bind_uniform_set(compute_list, gi_probe->mipmaps[i].uniform_set, 0);
} else {
RD::get_singleton()->compute_list_bind_uniform_set(compute_list, gi_probe->mipmaps[i].second_bounce_uniform_set, 0);
}
push_constant.cell_offset = gi_probe->mipmaps[i].cell_offset;
push_constant.cell_count = gi_probe->mipmaps[i].cell_count;
int wg_todo = (gi_probe->mipmaps[i].cell_count - 1) / wg_size + 1;
while (wg_todo) {
int wg_count = MIN(wg_todo, wg_limit_x);
RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(GIProbePushConstant));
RD::get_singleton()->compute_list_dispatch(compute_list, wg_count, 1, 1);
wg_todo -= wg_count;
push_constant.cell_offset += wg_count * wg_size;
}
}
GIProbes working.
2019-10-03 20:39:08 +00:00
RD::get_singleton()->compute_list_add_barrier(compute_list); //wait til previous step is done
}
Dynamic object support for GI Probes (a bit buggy still)
2019-10-11 02:14:56 +00:00
RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, giprobe_lighting_shader_version_pipelines[GI_PROBE_SHADER_VERSION_WRITE_TEXTURE]);
for (int i = 0; i < gi_probe->mipmaps.size(); i++) {
RD::get_singleton()->compute_list_bind_uniform_set(compute_list, gi_probe->mipmaps[i].write_uniform_set, 0);
push_constant.cell_offset = gi_probe->mipmaps[i].cell_offset;
push_constant.cell_count = gi_probe->mipmaps[i].cell_count;
int wg_todo = (gi_probe->mipmaps[i].cell_count - 1) / wg_size + 1;
while (wg_todo) {
int wg_count = MIN(wg_todo, wg_limit_x);
RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(GIProbePushConstant));
RD::get_singleton()->compute_list_dispatch(compute_list, wg_count, 1, 1);
wg_todo -= wg_count;
push_constant.cell_offset += wg_count * wg_size;
}
GIProbes working.
2019-10-03 20:39:08 +00:00
}
Dynamic object support for GI Probes (a bit buggy still)
2019-10-11 02:14:56 +00:00
}
RD::get_singleton()->compute_list_end();
}
}
GIProbes working.
2019-10-03 20:39:08 +00:00
Dynamic object support for GI Probes (a bit buggy still)
2019-10-11 02:14:56 +00:00
gi_probe->has_dynamic_object_data = false; //clear until dynamic object data is used again
GIProbes working.
2019-10-03 20:39:08 +00:00
Replace Octree by DynamicBVH in cull code -Much greater pairing/unpairing performance -For now, using it for culling too, but this will change in a couple of days. -Added a paged allocator, to efficiently alloc/free some types of objects.
2020-12-23 16:52:58 +00:00
if (p_dynamic_objects.size() && gi_probe->dynamic_maps.size()) {
Dynamic object support for GI Probes (a bit buggy still)
2019-10-11 02:14:56 +00:00
Vector3i octree_size = storage->gi_probe_get_octree_size(gi_probe->probe);
int multiplier = gi_probe->dynamic_maps[0].size / MAX(MAX(octree_size.x, octree_size.y), octree_size.z);
Transform oversample_scale;
oversample_scale.basis.scale(Vector3(multiplier, multiplier, multiplier));
Transform to_cell = oversample_scale * storage->gi_probe_get_to_cell_xform(gi_probe->probe);
Transform to_world_xform = gi_probe->transform * to_cell.affine_inverse();
Transform to_probe_xform = to_world_xform.affine_inverse();
AABB probe_aabb(Vector3(), octree_size);
//this could probably be better parallelized in compute..
Replace Octree by DynamicBVH in cull code -Much greater pairing/unpairing performance -For now, using it for culling too, but this will change in a couple of days. -Added a paged allocator, to efficiently alloc/free some types of objects.
2020-12-23 16:52:58 +00:00
for (int i = 0; i < (int)p_dynamic_objects.size(); i++) {
Reorganize renderer code. So it can hopefully be made more cache efficient afterwards.
2020-12-31 12:42:56 +00:00
GeometryInstance *instance = p_dynamic_objects[i];
Dynamic object support for GI Probes (a bit buggy still)
2019-10-11 02:14:56 +00:00
//transform aabb to giprobe
Reorganize renderer code. So it can hopefully be made more cache efficient afterwards.
2020-12-31 12:42:56 +00:00
AABB aabb = (to_probe_xform * geometry_instance_get_transform(instance)).xform(geometry_instance_get_aabb(instance));
Dynamic object support for GI Probes (a bit buggy still)
2019-10-11 02:14:56 +00:00
//this needs to wrap to grid resolution to avoid jitter
//also extend margin a bit just in case
Vector3i begin = aabb.position - Vector3i(1, 1, 1);
Vector3i end = aabb.position + aabb.size + Vector3i(1, 1, 1);
for (int j = 0; j < 3; j++) {
if ((end[j] - begin[j]) & 1) {
end[j]++; //for half extents split, it needs to be even
More GIProbe work and fixes
2019-10-05 13:27:43 +00:00
}
Dynamic object support for GI Probes (a bit buggy still)
2019-10-11 02:14:56 +00:00
begin[j] = MAX(begin[j], 0);
end[j] = MIN(end[j], octree_size[j] * multiplier);
GIProbes working.
2019-10-03 20:39:08 +00:00
}
Dynamic object support for GI Probes (a bit buggy still)
2019-10-11 02:14:56 +00:00
//aabb = aabb.intersection(probe_aabb); //intersect
aabb.position = begin;
aabb.size = end - begin;
GIProbes working.
2019-10-03 20:39:08 +00:00
Dynamic object support for GI Probes (a bit buggy still)
2019-10-11 02:14:56 +00:00
//print_line("aabb: " + aabb);
GIProbes working.
2019-10-03 20:39:08 +00:00
Dynamic object support for GI Probes (a bit buggy still)
2019-10-11 02:14:56 +00:00
for (int j = 0; j < 6; j++) {
//if (j != 0 && j != 3) {
// continue;
//}
static const Vector3 render_z[6] = {
Vector3(1, 0, 0),
Vector3(0, 1, 0),
Vector3(0, 0, 1),
Vector3(-1, 0, 0),
Vector3(0, -1, 0),
Vector3(0, 0, -1),
};
static const Vector3 render_up[6] = {
Vector3(0, 1, 0),
Vector3(0, 0, 1),
Vector3(0, 1, 0),
Vector3(0, 1, 0),
Vector3(0, 0, 1),
Vector3(0, 1, 0),
};
Vector3 render_dir = render_z[j];
Vector3 up_dir = render_up[j];
Vector3 center = aabb.position + aabb.size * 0.5;
Transform xform;
xform.set_look_at(center - aabb.size * 0.5 * render_dir, center, up_dir);
Vector3 x_dir = xform.basis.get_axis(0).abs();
int x_axis = int(Vector3(0, 1, 2).dot(x_dir));
Vector3 y_dir = xform.basis.get_axis(1).abs();
int y_axis = int(Vector3(0, 1, 2).dot(y_dir));
Vector3 z_dir = -xform.basis.get_axis(2);
int z_axis = int(Vector3(0, 1, 2).dot(z_dir.abs()));
Rect2i rect(aabb.position[x_axis], aabb.position[y_axis], aabb.size[x_axis], aabb.size[y_axis]);
bool x_flip = bool(Vector3(1, 1, 1).dot(xform.basis.get_axis(0)) < 0);
bool y_flip = bool(Vector3(1, 1, 1).dot(xform.basis.get_axis(1)) < 0);
bool z_flip = bool(Vector3(1, 1, 1).dot(xform.basis.get_axis(2)) > 0);
CameraMatrix cm;
cm.set_orthogonal(-rect.size.width / 2, rect.size.width / 2, -rect.size.height / 2, rect.size.height / 2, 0.0001, aabb.size[z_axis]);
Replace Octree by DynamicBVH in cull code -Much greater pairing/unpairing performance -For now, using it for culling too, but this will change in a couple of days. -Added a paged allocator, to efficiently alloc/free some types of objects.
2020-12-23 16:52:58 +00:00
if (cull_argument.size() == 0) {
cull_argument.push_back(nullptr);
}
cull_argument[0] = instance;
_render_material(to_world_xform * xform, cm, true, cull_argument, gi_probe->dynamic_maps[0].fb, Rect2i(Vector2i(), rect.size));
Dynamic object support for GI Probes (a bit buggy still)
2019-10-11 02:14:56 +00:00
GIProbeDynamicPushConstant push_constant;
zeromem(&push_constant, sizeof(GIProbeDynamicPushConstant));
push_constant.limits[0] = octree_size.x;
push_constant.limits[1] = octree_size.y;
push_constant.limits[2] = octree_size.z;
push_constant.light_count = p_light_instances.size();
push_constant.x_dir[0] = x_dir[0];
push_constant.x_dir[1] = x_dir[1];
push_constant.x_dir[2] = x_dir[2];
push_constant.y_dir[0] = y_dir[0];
push_constant.y_dir[1] = y_dir[1];
push_constant.y_dir[2] = y_dir[2];
push_constant.z_dir[0] = z_dir[0];
push_constant.z_dir[1] = z_dir[1];
push_constant.z_dir[2] = z_dir[2];
push_constant.z_base = xform.origin[z_axis];
push_constant.z_sign = (z_flip ? -1.0 : 1.0);
push_constant.pos_multiplier = float(1.0) / multiplier;
push_constant.dynamic_range = storage->gi_probe_get_dynamic_range(gi_probe->probe);
push_constant.flip_x = x_flip;
push_constant.flip_y = y_flip;
push_constant.rect_pos[0] = rect.position[0];
push_constant.rect_pos[1] = rect.position[1];
push_constant.rect_size[0] = rect.size[0];
push_constant.rect_size[1] = rect.size[1];
push_constant.prev_rect_ofs[0] = 0;
push_constant.prev_rect_ofs[1] = 0;
push_constant.prev_rect_size[0] = 0;
push_constant.prev_rect_size[1] = 0;
More correctly divide alpha by 8 for mipmaps, fixes voxel overocclusion.
2019-11-02 01:08:56 +00:00
push_constant.on_mipmap = false;
Dynamic objects in GIProbes are now affected by propagation parameter.
2019-11-02 01:28:53 +00:00
push_constant.propagation = storage->gi_probe_get_propagation(gi_probe->probe);
Fix code formatting issues and VS compilation Also temporarily disable multicheck build so that we get a full build even when there are style issues on Vulkan. Fixes #33356.
2019-11-05 11:01:00 +00:00
push_constant.pad[0] = 0;
push_constant.pad[1] = 0;
push_constant.pad[2] = 0;
Dynamic object support for GI Probes (a bit buggy still)
2019-10-11 02:14:56 +00:00
//process lighting
RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin();
RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, giprobe_lighting_shader_version_pipelines[GI_PROBE_SHADER_VERSION_DYNAMIC_OBJECT_LIGHTING]);
RD::get_singleton()->compute_list_bind_uniform_set(compute_list, gi_probe->dynamic_maps[0].uniform_set, 0);
RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(GIProbeDynamicPushConstant));
RD::get_singleton()->compute_list_dispatch(compute_list, (rect.size.x - 1) / 8 + 1, (rect.size.y - 1) / 8 + 1, 1);
//print_line("rect: " + itos(i) + ": " + rect);
for (int k = 1; k < gi_probe->dynamic_maps.size(); k++) {
// enlarge the rect if needed so all pixels fit when downscaled,
// this ensures downsampling is smooth and optimal because no pixels are left behind
//x
if (rect.position.x & 1) {
rect.size.x++;
push_constant.prev_rect_ofs[0] = 1; //this is used to ensure reading is also optimal
} else {
push_constant.prev_rect_ofs[0] = 0;
}
if (rect.size.x & 1) {
rect.size.x++;
}
GIProbes working.
2019-10-03 20:39:08 +00:00
Dynamic object support for GI Probes (a bit buggy still)
2019-10-11 02:14:56 +00:00
rect.position.x >>= 1;
rect.size.x = MAX(1, rect.size.x >> 1);
//y
if (rect.position.y & 1) {
rect.size.y++;
push_constant.prev_rect_ofs[1] = 1;
} else {
push_constant.prev_rect_ofs[1] = 0;
}
if (rect.size.y & 1) {
rect.size.y++;
}
rect.position.y >>= 1;
rect.size.y = MAX(1, rect.size.y >> 1);
//shrink limits to ensure plot does not go outside map
Use a different strategy to get textures from GPU (via buffers), for more compatibility.
2019-10-25 14:22:19 +00:00
if (gi_probe->dynamic_maps[k].mipmap > 0) {
Dynamic object support for GI Probes (a bit buggy still)
2019-10-11 02:14:56 +00:00
for (int l = 0; l < 3; l++) {
push_constant.limits[l] = MAX(1, push_constant.limits[l] >> 1);
}
}
GIProbes working.
2019-10-03 20:39:08 +00:00
Dynamic object support for GI Probes (a bit buggy still)
2019-10-11 02:14:56 +00:00
//print_line("rect: " + itos(i) + ": " + rect);
push_constant.rect_pos[0] = rect.position[0];
push_constant.rect_pos[1] = rect.position[1];
push_constant.prev_rect_size[0] = push_constant.rect_size[0];
push_constant.prev_rect_size[1] = push_constant.rect_size[1];
push_constant.rect_size[0] = rect.size[0];
push_constant.rect_size[1] = rect.size[1];
More correctly divide alpha by 8 for mipmaps, fixes voxel overocclusion.
2019-11-02 01:08:56 +00:00
push_constant.on_mipmap = gi_probe->dynamic_maps[k].mipmap > 0;
Dynamic object support for GI Probes (a bit buggy still)
2019-10-11 02:14:56 +00:00
RD::get_singleton()->compute_list_add_barrier(compute_list);
if (gi_probe->dynamic_maps[k].mipmap < 0) {
RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, giprobe_lighting_shader_version_pipelines[GI_PROBE_SHADER_VERSION_DYNAMIC_SHRINK_WRITE]);
} else if (k < gi_probe->dynamic_maps.size() - 1) {
RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, giprobe_lighting_shader_version_pipelines[GI_PROBE_SHADER_VERSION_DYNAMIC_SHRINK_WRITE_PLOT]);
} else {
RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, giprobe_lighting_shader_version_pipelines[GI_PROBE_SHADER_VERSION_DYNAMIC_SHRINK_PLOT]);
}
RD::get_singleton()->compute_list_bind_uniform_set(compute_list, gi_probe->dynamic_maps[k].uniform_set, 0);
RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(GIProbeDynamicPushConstant));
RD::get_singleton()->compute_list_dispatch(compute_list, (rect.size.x - 1) / 8 + 1, (rect.size.y - 1) / 8 + 1, 1);
More GIProbe work and fixes
2019-10-05 13:27:43 +00:00
}
Dynamic object support for GI Probes (a bit buggy still)
2019-10-11 02:14:56 +00:00
RD::get_singleton()->compute_list_end();
GIProbes working.
2019-10-03 20:39:08 +00:00
}
}
Dynamic object support for GI Probes (a bit buggy still)
2019-10-11 02:14:56 +00:00
gi_probe->has_dynamic_object_data = true; //clear until dynamic object data is used again
GIProbes working.
2019-10-03 20:39:08 +00:00
}
gi_probe->last_probe_version = storage->gi_probe_get_version(gi_probe->probe);
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
void RendererSceneRenderRD::_debug_giprobe(RID p_gi_probe, RD::DrawListID p_draw_list, RID p_framebuffer, const CameraMatrix &p_camera_with_transform, bool p_lighting, bool p_emission, float p_alpha) {
GIProbes working.
2019-10-03 20:39:08 +00:00
GIProbeInstance *gi_probe = gi_probe_instance_owner.getornull(p_gi_probe);
ERR_FAIL_COND(!gi_probe);
if (gi_probe->mipmaps.size() == 0) {
return;
}
CameraMatrix transform = (p_camera_with_transform * CameraMatrix(gi_probe->transform)) * CameraMatrix(storage->gi_probe_get_to_cell_xform(gi_probe->probe).affine_inverse());
int level = 0;
Dynamic object support for GI Probes (a bit buggy still)
2019-10-11 02:14:56 +00:00
Vector3i octree_size = storage->gi_probe_get_octree_size(gi_probe->probe);
GIProbes working.
2019-10-03 20:39:08 +00:00
GIProbeDebugPushConstant push_constant;
push_constant.alpha = p_alpha;
push_constant.dynamic_range = storage->gi_probe_get_dynamic_range(gi_probe->probe);
push_constant.cell_offset = gi_probe->mipmaps[level].cell_offset;
push_constant.level = level;
Dynamic object support for GI Probes (a bit buggy still)
2019-10-11 02:14:56 +00:00
push_constant.bounds[0] = octree_size.x >> level;
push_constant.bounds[1] = octree_size.y >> level;
push_constant.bounds[2] = octree_size.z >> level;
push_constant.pad = 0;
GIProbes working.
2019-10-03 20:39:08 +00:00
for (int i = 0; i < 4; i++) {
for (int j = 0; j < 4; j++) {
push_constant.projection[i * 4 + j] = transform.matrix[i][j];
}
}
if (giprobe_debug_uniform_set.is_valid()) {
RD::get_singleton()->free(giprobe_debug_uniform_set);
}
Vector<RD::Uniform> uniforms;
{
RD::Uniform u;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER;
GIProbes working.
2019-10-03 20:39:08 +00:00
u.binding = 1;
u.ids.push_back(storage->gi_probe_get_data_buffer(gi_probe->probe));
uniforms.push_back(u);
}
{
RD::Uniform u;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_TEXTURE;
GIProbes working.
2019-10-03 20:39:08 +00:00
u.binding = 2;
u.ids.push_back(gi_probe->texture);
uniforms.push_back(u);
}
{
RD::Uniform u;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_SAMPLER;
GIProbes working.
2019-10-03 20:39:08 +00:00
u.binding = 3;
Renaming of servers for coherency. VisualServer -> RenderingServer PhysicsServer -> PhysicsServer3D Physics2DServer -> PhysicsServer2D NavigationServer -> NavigationServer3D Navigation2DServer -> NavigationServer2D Also renamed corresponding files.
2020-03-27 18:21:27 +00:00
u.ids.push_back(storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED));
GIProbes working.
2019-10-03 20:39:08 +00:00
uniforms.push_back(u);
}
Dynamic object support for GI Probes (a bit buggy still)
2019-10-11 02:14:56 +00:00
int cell_count;
if (!p_emission && p_lighting && gi_probe->has_dynamic_object_data) {
cell_count = push_constant.bounds[0] * push_constant.bounds[1] * push_constant.bounds[2];
} else {
cell_count = gi_probe->mipmaps[level].cell_count;
}
GIProbes working.
2019-10-03 20:39:08 +00:00
giprobe_debug_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, giprobe_debug_shader_version_shaders[0], 0);
CI: Update to clang-format 11 and apply ternary operator changes
2021-01-12 15:57:55 +00:00
int giprobe_debug_pipeline = GI_PROBE_DEBUG_COLOR;
if (p_emission) {
giprobe_debug_pipeline = GI_PROBE_DEBUG_EMISSION;
} else if (p_lighting) {
giprobe_debug_pipeline = gi_probe->has_dynamic_object_data ? GI_PROBE_DEBUG_LIGHT_FULL : GI_PROBE_DEBUG_LIGHT;
}
RD::get_singleton()->draw_list_bind_render_pipeline(
p_draw_list,
giprobe_debug_shader_version_pipelines[giprobe_debug_pipeline].get_render_pipeline(RD::INVALID_ID, RD::get_singleton()->framebuffer_get_format(p_framebuffer)));
GIProbes working.
2019-10-03 20:39:08 +00:00
RD::get_singleton()->draw_list_bind_uniform_set(p_draw_list, giprobe_debug_uniform_set, 0);
RD::get_singleton()->draw_list_set_push_constant(p_draw_list, &push_constant, sizeof(GIProbeDebugPushConstant));
RD::get_singleton()->draw_list_draw(p_draw_list, false, cell_count, 36);
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
void RendererSceneRenderRD::_debug_sdfgi_probes(RID p_render_buffers, RD::DrawListID p_draw_list, RID p_framebuffer, const CameraMatrix &p_camera_with_transform) {
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers);
ERR_FAIL_COND(!rb);
if (!rb->sdfgi) {
return; //nothing to debug
}
SDGIShader::DebugProbesPushConstant push_constant;
for (int i = 0; i < 4; i++) {
for (int j = 0; j < 4; j++) {
push_constant.projection[i * 4 + j] = p_camera_with_transform.matrix[i][j];
}
}
//gen spheres from strips
uint32_t band_points = 16;
push_constant.band_power = 4;
push_constant.sections_in_band = ((band_points / 2) - 1);
push_constant.band_mask = band_points - 2;
Use Math_TAU and deg2rad/rad2deg in more places and optimize code
2020-04-03 09:50:40 +00:00
push_constant.section_arc = Math_TAU / float(push_constant.sections_in_band);
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
push_constant.y_mult = rb->sdfgi->y_mult;
uint32_t total_points = push_constant.sections_in_band * band_points;
uint32_t total_probes = rb->sdfgi->probe_axis_count * rb->sdfgi->probe_axis_count * rb->sdfgi->probe_axis_count;
push_constant.grid_size[0] = rb->sdfgi->cascade_size;
push_constant.grid_size[1] = rb->sdfgi->cascade_size;
push_constant.grid_size[2] = rb->sdfgi->cascade_size;
push_constant.cascade = 0;
push_constant.probe_axis_size = rb->sdfgi->probe_axis_count;
if (!rb->sdfgi->debug_probes_uniform_set.is_valid() || !RD::get_singleton()->uniform_set_is_valid(rb->sdfgi->debug_probes_uniform_set)) {
Vector<RD::Uniform> uniforms;
{
RD::Uniform u;
u.binding = 1;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
u.ids.push_back(rb->sdfgi->cascades_ubo);
uniforms.push_back(u);
}
{
RD::Uniform u;
u.binding = 2;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_TEXTURE;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
u.ids.push_back(rb->sdfgi->lightprobe_texture);
uniforms.push_back(u);
}
{
RD::Uniform u;
u.binding = 3;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_SAMPLER;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
u.ids.push_back(storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED));
uniforms.push_back(u);
}
{
RD::Uniform u;
u.binding = 4;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_TEXTURE;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
u.ids.push_back(rb->sdfgi->occlusion_texture);
uniforms.push_back(u);
}
rb->sdfgi->debug_probes_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, sdfgi_shader.debug_probes.version_get_shader(sdfgi_shader.debug_probes_shader, 0), 0);
}
RD::get_singleton()->draw_list_bind_render_pipeline(p_draw_list, sdfgi_shader.debug_probes_pipeline[SDGIShader::PROBE_DEBUG_PROBES].get_render_pipeline(RD::INVALID_FORMAT_ID, RD::get_singleton()->framebuffer_get_format(p_framebuffer)));
RD::get_singleton()->draw_list_bind_uniform_set(p_draw_list, rb->sdfgi->debug_probes_uniform_set, 0);
RD::get_singleton()->draw_list_set_push_constant(p_draw_list, &push_constant, sizeof(SDGIShader::DebugProbesPushConstant));
RD::get_singleton()->draw_list_draw(p_draw_list, false, total_probes, total_points);
if (sdfgi_debug_probe_dir != Vector3()) {
print_line("CLICK DEBUG ME?");
uint32_t cascade = 0;
Vector3 offset = Vector3((Vector3i(1, 1, 1) * -int32_t(rb->sdfgi->cascade_size >> 1) + rb->sdfgi->cascades[cascade].position)) * rb->sdfgi->cascades[cascade].cell_size * Vector3(1.0, 1.0 / rb->sdfgi->y_mult, 1.0);
Vector3 probe_size = rb->sdfgi->cascades[cascade].cell_size * (rb->sdfgi->cascade_size / SDFGI::PROBE_DIVISOR) * Vector3(1.0, 1.0 / rb->sdfgi->y_mult, 1.0);
Vector3 ray_from = sdfgi_debug_probe_pos;
Vector3 ray_to = sdfgi_debug_probe_pos + sdfgi_debug_probe_dir * rb->sdfgi->cascades[cascade].cell_size * Math::sqrt(3.0) * rb->sdfgi->cascade_size;
float sphere_radius = 0.2;
float closest_dist = 1e20;
sdfgi_debug_probe_enabled = false;
Vector3i probe_from = rb->sdfgi->cascades[cascade].position / (rb->sdfgi->cascade_size / SDFGI::PROBE_DIVISOR);
for (int i = 0; i < (SDFGI::PROBE_DIVISOR + 1); i++) {
for (int j = 0; j < (SDFGI::PROBE_DIVISOR + 1); j++) {
for (int k = 0; k < (SDFGI::PROBE_DIVISOR + 1); k++) {
Vector3 pos = offset + probe_size * Vector3(i, j, k);
Vector3 res;
if (Geometry3D::segment_intersects_sphere(ray_from, ray_to, pos, sphere_radius, &res)) {
float d = ray_from.distance_to(res);
if (d < closest_dist) {
closest_dist = d;
sdfgi_debug_probe_enabled = true;
sdfgi_debug_probe_index = probe_from + Vector3i(i, j, k);
}
}
}
}
}
if (sdfgi_debug_probe_enabled) {
print_line("found: " + sdfgi_debug_probe_index);
} else {
print_line("no found");
}
sdfgi_debug_probe_dir = Vector3();
}
if (sdfgi_debug_probe_enabled) {
uint32_t cascade = 0;
uint32_t probe_cells = (rb->sdfgi->cascade_size / SDFGI::PROBE_DIVISOR);
Vector3i probe_from = rb->sdfgi->cascades[cascade].position / probe_cells;
Vector3i ofs = sdfgi_debug_probe_index - probe_from;
if (ofs.x < 0 || ofs.y < 0 || ofs.z < 0) {
return;
}
if (ofs.x > SDFGI::PROBE_DIVISOR || ofs.y > SDFGI::PROBE_DIVISOR || ofs.z > SDFGI::PROBE_DIVISOR) {
return;
}
GIProbes working.
2019-10-03 20:39:08 +00:00
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
uint32_t mult = (SDFGI::PROBE_DIVISOR + 1);
uint32_t index = ofs.z * mult * mult + ofs.y * mult + ofs.x;
push_constant.probe_debug_index = index;
uint32_t cell_count = probe_cells * 2 * probe_cells * 2 * probe_cells * 2;
RD::get_singleton()->draw_list_bind_render_pipeline(p_draw_list, sdfgi_shader.debug_probes_pipeline[SDGIShader::PROBE_DEBUG_VISIBILITY].get_render_pipeline(RD::INVALID_FORMAT_ID, RD::get_singleton()->framebuffer_get_format(p_framebuffer)));
RD::get_singleton()->draw_list_bind_uniform_set(p_draw_list, rb->sdfgi->debug_probes_uniform_set, 0);
RD::get_singleton()->draw_list_set_push_constant(p_draw_list, &push_constant, sizeof(SDGIShader::DebugProbesPushConstant));
RD::get_singleton()->draw_list_draw(p_draw_list, false, cell_count, total_points);
}
GIProbes working.
2019-10-03 20:39:08 +00:00
}
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
////////////////////////////////
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
RID RendererSceneRenderRD::render_buffers_create() {
Base 3D engine done, still untested, though.
2019-08-18 22:40:52 +00:00
RenderBuffers rb;
rb.data = _create_render_buffer_data();
return render_buffers_owner.make_rid(rb);
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
void RendererSceneRenderRD::_allocate_blur_textures(RenderBuffers *rb) {
-Refactored post processing, re-added glow and added a mix blend mode.
2020-01-10 00:40:26 +00:00
ERR_FAIL_COND(!rb->blur[0].texture.is_null());
uint32_t mipmaps_required = Image::get_image_required_mipmaps(rb->width, rb->height, Image::FORMAT_RGBAH);
RD::TextureFormat tf;
tf.format = RD::DATA_FORMAT_R16G16B16A16_SFLOAT;
tf.width = rb->width;
tf.height = rb->height;
Rename RD texture "type" to "texture_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-11-27 03:50:05 +00:00
tf.texture_type = RD::TEXTURE_TYPE_2D;
Moved most of the effect code to compute. Simplifies it and improves performance. Improves image barrier handling per mipmap on RenderingDeviceVulkan
2020-04-11 17:43:12 +00:00
tf.usage_bits = RD::TEXTURE_USAGE_STORAGE_BIT | RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_CAN_COPY_TO_BIT;
-Refactored post processing, re-added glow and added a mix blend mode.
2020-01-10 00:40:26 +00:00
tf.mipmaps = mipmaps_required;
rb->blur[0].texture = RD::get_singleton()->texture_create(tf, RD::TextureView());
//the second one is smaller (only used for separatable part of blur)
tf.width >>= 1;
tf.height >>= 1;
tf.mipmaps--;
rb->blur[1].texture = RD::get_singleton()->texture_create(tf, RD::TextureView());
int base_width = rb->width;
int base_height = rb->height;
for (uint32_t i = 0; i < mipmaps_required; i++) {
RenderBuffers::Blur::Mipmap mm;
mm.texture = RD::get_singleton()->texture_create_shared_from_slice(RD::TextureView(), rb->blur[0].texture, 0, i);
mm.width = base_width;
mm.height = base_height;
rb->blur[0].mipmaps.push_back(mm);
if (i > 0) {
mm.texture = RD::get_singleton()->texture_create_shared_from_slice(RD::TextureView(), rb->blur[1].texture, 0, i - 1);
rb->blur[1].mipmaps.push_back(mm);
}
base_width = MAX(1, base_width >> 1);
base_height = MAX(1, base_height >> 1);
}
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
void RendererSceneRenderRD::_allocate_luminance_textures(RenderBuffers *rb) {
Auto exposure re-implemented in Vulkan
2020-01-12 01:26:52 +00:00
ERR_FAIL_COND(!rb->luminance.current.is_null());
int w = rb->width;
int h = rb->height;
while (true) {
w = MAX(w / 8, 1);
h = MAX(h / 8, 1);
RD::TextureFormat tf;
tf.format = RD::DATA_FORMAT_R32_SFLOAT;
tf.width = w;
tf.height = h;
tf.usage_bits = RD::TEXTURE_USAGE_STORAGE_BIT;
bool final = w == 1 && h == 1;
if (final) {
tf.usage_bits |= RD::TEXTURE_USAGE_SAMPLING_BIT;
}
RID texture = RD::get_singleton()->texture_create(tf, RD::TextureView());
rb->luminance.reduce.push_back(texture);
if (final) {
rb->luminance.current = RD::get_singleton()->texture_create(tf, RD::TextureView());
break;
}
}
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
void RendererSceneRenderRD::_free_render_buffer_data(RenderBuffers *rb) {
-Refactored post processing, re-added glow and added a mix blend mode.
2020-01-10 00:40:26 +00:00
if (rb->texture.is_valid()) {
RD::get_singleton()->free(rb->texture);
WIP CameraEffects implementation (bokeh not working for now)
2020-01-13 18:37:24 +00:00
rb->texture = RID();
}
if (rb->depth_texture.is_valid()) {
RD::get_singleton()->free(rb->depth_texture);
rb->depth_texture = RID();
-Refactored post processing, re-added glow and added a mix blend mode.
2020-01-10 00:40:26 +00:00
}
for (int i = 0; i < 2; i++) {
if (rb->blur[i].texture.is_valid()) {
RD::get_singleton()->free(rb->blur[i].texture);
rb->blur[i].texture = RID();
rb->blur[i].mipmaps.clear();
}
}
for (int i = 0; i < rb->luminance.reduce.size(); i++) {
RD::get_singleton()->free(rb->luminance.reduce[i]);
}
Auto exposure re-implemented in Vulkan
2020-01-12 01:26:52 +00:00
for (int i = 0; i < rb->luminance.reduce.size(); i++) {
RD::get_singleton()->free(rb->luminance.reduce[i]);
}
-Refactored post processing, re-added glow and added a mix blend mode.
2020-01-10 00:40:26 +00:00
rb->luminance.reduce.clear();
if (rb->luminance.current.is_valid()) {
RD::get_singleton()->free(rb->luminance.current);
rb->luminance.current = RID();
}
Re-implemented screen space ambient occlusion
2020-01-25 10:18:55 +00:00
Port ASSAO to Godot to replace SAO
2020-12-08 05:37:09 +00:00
if (rb->ssao.depth.is_valid()) {
Revert "Replace SAO implementation with MSSAO"
2020-10-18 22:27:51 +00:00
RD::get_singleton()->free(rb->ssao.depth);
Port ASSAO to Godot to replace SAO
2020-12-08 05:37:09 +00:00
RD::get_singleton()->free(rb->ssao.ao_deinterleaved);
RD::get_singleton()->free(rb->ssao.ao_pong);
RD::get_singleton()->free(rb->ssao.ao_final);
RD::get_singleton()->free(rb->ssao.importance_map[0]);
RD::get_singleton()->free(rb->ssao.importance_map[1]);
Re-implemented screen space ambient occlusion
2020-01-25 10:18:55 +00:00
Revert "Replace SAO implementation with MSSAO"
2020-10-18 22:27:51 +00:00
rb->ssao.depth = RID();
Port ASSAO to Godot to replace SAO
2020-12-08 05:37:09 +00:00
rb->ssao.ao_deinterleaved = RID();
rb->ssao.ao_pong = RID();
rb->ssao.ao_final = RID();
rb->ssao.importance_map[0] = RID();
rb->ssao.importance_map[1] = RID();
Re-implemented screen space ambient occlusion
2020-01-25 10:18:55 +00:00
rb->ssao.depth_slices.clear();
Port ASSAO to Godot to replace SAO
2020-12-08 05:37:09 +00:00
rb->ssao.ao_deinterleaved_slices.clear();
rb->ssao.ao_pong_slices.clear();
Re-implemented screen space ambient occlusion
2020-01-25 10:18:55 +00:00
}
Re-Added screen space reflection.
2020-04-02 02:24:52 +00:00
if (rb->ssr.blur_radius[0].is_valid()) {
RD::get_singleton()->free(rb->ssr.blur_radius[0]);
RD::get_singleton()->free(rb->ssr.blur_radius[1]);
rb->ssr.blur_radius[0] = RID();
rb->ssr.blur_radius[1] = RID();
}
if (rb->ssr.depth_scaled.is_valid()) {
RD::get_singleton()->free(rb->ssr.depth_scaled);
rb->ssr.depth_scaled = RID();
RD::get_singleton()->free(rb->ssr.normal_scaled);
rb->ssr.normal_scaled = RID();
}
Several GI related optimizations and fixes -SDFGI direct light is done over many frames -SDFGI Changed settings for rays/frame -SDFGI Misc optimizations -SDFGI Bug fix on probe scroll -GIProbe was not working, got it to work again -GIProbe dynamic objects were not working, fixed -Added a half size GI option.
2021-01-22 23:50:24 +00:00
if (rb->ambient_buffer.is_valid()) {
RD::get_singleton()->free(rb->ambient_buffer);
RD::get_singleton()->free(rb->reflection_buffer);
rb->ambient_buffer = RID();
rb->reflection_buffer = RID();
}
Re-Added screen space reflection.
2020-04-02 02:24:52 +00:00
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
void RendererSceneRenderRD::_process_sss(RID p_render_buffers, const CameraMatrix &p_camera) {
Re-implement subsurface scattering. The size settings are more "just works", with default scale and depth scale values that don't need much tweaking. Additionally, a "skin" mode was added so skin looks better. EDIT: Cleaned up SSR filter shader a bit.
2020-04-04 02:42:26 +00:00
RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers);
ERR_FAIL_COND(!rb);
bool can_use_effects = rb->width >= 8 && rb->height >= 8;
if (!can_use_effects) {
//just copy
return;
}
if (rb->blur[0].texture.is_null()) {
_allocate_blur_textures(rb);
_render_buffers_uniform_set_changed(p_render_buffers);
}
storage->get_effects()->sub_surface_scattering(rb->texture, rb->blur[0].mipmaps[0].texture, rb->depth_texture, p_camera, Size2i(rb->width, rb->height), sss_scale, sss_depth_scale, sss_quality);
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
void RendererSceneRenderRD::_process_ssr(RID p_render_buffers, RID p_dest_framebuffer, RID p_normal_buffer, RID p_specular_buffer, RID p_metallic, const Color &p_metallic_mask, RID p_environment, const CameraMatrix &p_projection, bool p_use_additive) {
Re-Added screen space reflection.
2020-04-02 02:24:52 +00:00
RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers);
ERR_FAIL_COND(!rb);
bool can_use_effects = rb->width >= 8 && rb->height >= 8;
if (!can_use_effects) {
//just copy
storage->get_effects()->merge_specular(p_dest_framebuffer, p_specular_buffer, p_use_additive ? RID() : rb->texture, RID());
return;
}
Fix spelling of a var, a struct, and message output
2020-07-17 16:02:06 +00:00
Environment *env = environment_owner.getornull(p_environment);
Re-Added screen space reflection.
2020-04-02 02:24:52 +00:00
ERR_FAIL_COND(!env);
ERR_FAIL_COND(!env->ssr_enabled);
if (rb->ssr.depth_scaled.is_null()) {
RD::TextureFormat tf;
tf.format = RD::DATA_FORMAT_R32_SFLOAT;
tf.width = rb->width / 2;
tf.height = rb->height / 2;
Rename RD texture "type" to "texture_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-11-27 03:50:05 +00:00
tf.texture_type = RD::TEXTURE_TYPE_2D;
Re-Added screen space reflection.
2020-04-02 02:24:52 +00:00
tf.usage_bits = RD::TEXTURE_USAGE_STORAGE_BIT;
rb->ssr.depth_scaled = RD::get_singleton()->texture_create(tf, RD::TextureView());
tf.format = RD::DATA_FORMAT_R8G8B8A8_UNORM;
rb->ssr.normal_scaled = RD::get_singleton()->texture_create(tf, RD::TextureView());
}
if (ssr_roughness_quality != RS::ENV_SSR_ROUGNESS_QUALITY_DISABLED && !rb->ssr.blur_radius[0].is_valid()) {
RD::TextureFormat tf;
tf.format = RD::DATA_FORMAT_R8_UNORM;
tf.width = rb->width / 2;
tf.height = rb->height / 2;
Rename RD texture "type" to "texture_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-11-27 03:50:05 +00:00
tf.texture_type = RD::TEXTURE_TYPE_2D;
Re-Added screen space reflection.
2020-04-02 02:24:52 +00:00
tf.usage_bits = RD::TEXTURE_USAGE_STORAGE_BIT | RD::TEXTURE_USAGE_SAMPLING_BIT;
rb->ssr.blur_radius[0] = RD::get_singleton()->texture_create(tf, RD::TextureView());
rb->ssr.blur_radius[1] = RD::get_singleton()->texture_create(tf, RD::TextureView());
}
if (rb->blur[0].texture.is_null()) {
_allocate_blur_textures(rb);
_render_buffers_uniform_set_changed(p_render_buffers);
}
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
storage->get_effects()->screen_space_reflection(rb->texture, p_normal_buffer, ssr_roughness_quality, rb->ssr.blur_radius[0], rb->ssr.blur_radius[1], p_metallic, p_metallic_mask, rb->depth_texture, rb->ssr.depth_scaled, rb->ssr.normal_scaled, rb->blur[0].mipmaps[1].texture, rb->blur[1].mipmaps[0].texture, Size2i(rb->width / 2, rb->height / 2), env->ssr_max_steps, env->ssr_fade_in, env->ssr_fade_out, env->ssr_depth_tolerance, p_projection);
Re-Added screen space reflection.
2020-04-02 02:24:52 +00:00
storage->get_effects()->merge_specular(p_dest_framebuffer, p_specular_buffer, p_use_additive ? RID() : rb->texture, rb->blur[0].mipmaps[1].texture);
Re-implemented screen space ambient occlusion
2020-01-25 10:18:55 +00:00
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
void RendererSceneRenderRD::_process_ssao(RID p_render_buffers, RID p_environment, RID p_normal_buffer, const CameraMatrix &p_projection) {
Re-implemented screen space ambient occlusion
2020-01-25 10:18:55 +00:00
RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers);
ERR_FAIL_COND(!rb);
Fix spelling of a var, a struct, and message output
2020-07-17 16:02:06 +00:00
Environment *env = environment_owner.getornull(p_environment);
Re-implemented screen space ambient occlusion
2020-01-25 10:18:55 +00:00
ERR_FAIL_COND(!env);
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
RENDER_TIMESTAMP("Process SSAO");
Port ASSAO to Godot to replace SAO
2020-12-08 05:37:09 +00:00
if (rb->ssao.ao_final.is_valid() && ssao_using_half_size != ssao_half_size) {
Revert "Replace SAO implementation with MSSAO"
2020-10-18 22:27:51 +00:00
RD::get_singleton()->free(rb->ssao.depth);
Port ASSAO to Godot to replace SAO
2020-12-08 05:37:09 +00:00
RD::get_singleton()->free(rb->ssao.ao_deinterleaved);
RD::get_singleton()->free(rb->ssao.ao_pong);
RD::get_singleton()->free(rb->ssao.ao_final);
RD::get_singleton()->free(rb->ssao.importance_map[0]);
RD::get_singleton()->free(rb->ssao.importance_map[1]);
Re-implemented screen space ambient occlusion
2020-01-25 10:18:55 +00:00
Revert "Replace SAO implementation with MSSAO"
2020-10-18 22:27:51 +00:00
rb->ssao.depth = RID();
Port ASSAO to Godot to replace SAO
2020-12-08 05:37:09 +00:00
rb->ssao.ao_deinterleaved = RID();
rb->ssao.ao_pong = RID();
rb->ssao.ao_final = RID();
rb->ssao.importance_map[0] = RID();
rb->ssao.importance_map[1] = RID();
Revert "Replace SAO implementation with MSSAO"
2020-10-18 22:27:51 +00:00
rb->ssao.depth_slices.clear();
Port ASSAO to Godot to replace SAO
2020-12-08 05:37:09 +00:00
rb->ssao.ao_deinterleaved_slices.clear();
rb->ssao.ao_pong_slices.clear();
}
int buffer_width;
int buffer_height;
int half_width;
int half_height;
if (ssao_half_size) {
buffer_width = (rb->width + 3) / 4;
buffer_height = (rb->height + 3) / 4;
half_width = (rb->width + 7) / 8;
half_height = (rb->height + 7) / 8;
} else {
buffer_width = (rb->width + 1) / 2;
buffer_height = (rb->height + 1) / 2;
half_width = (rb->width + 3) / 4;
half_height = (rb->height + 3) / 4;
Revert "Replace SAO implementation with MSSAO"
2020-10-18 22:27:51 +00:00
}
Port ASSAO to Godot to replace SAO
2020-12-08 05:37:09 +00:00
bool uniform_sets_are_invalid = false;
if (rb->ssao.depth.is_null()) {
Revert "Replace SAO implementation with MSSAO"
2020-10-18 22:27:51 +00:00
//allocate depth slices
Re-implemented screen space ambient occlusion
2020-01-25 10:18:55 +00:00
{
Revert "Replace SAO implementation with MSSAO"
2020-10-18 22:27:51 +00:00
RD::TextureFormat tf;
Port ASSAO to Godot to replace SAO
2020-12-08 05:37:09 +00:00
tf.format = RD::DATA_FORMAT_R16_SFLOAT;
tf.texture_type = RD::TEXTURE_TYPE_2D_ARRAY;
tf.width = buffer_width;
tf.height = buffer_height;
tf.mipmaps = 4;
tf.array_layers = 4;
Revert "Replace SAO implementation with MSSAO"
2020-10-18 22:27:51 +00:00
tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_STORAGE_BIT;
rb->ssao.depth = RD::get_singleton()->texture_create(tf, RD::TextureView());
Add named resources and debug labels in RenderDoc
2021-01-24 06:21:54 +00:00
RD::get_singleton()->set_resource_name(rb->ssao.depth, "SSAO Depth");
Revert "Replace SAO implementation with MSSAO"
2020-10-18 22:27:51 +00:00
for (uint32_t i = 0; i < tf.mipmaps; i++) {
Port ASSAO to Godot to replace SAO
2020-12-08 05:37:09 +00:00
RID slice = RD::get_singleton()->texture_create_shared_from_slice(RD::TextureView(), rb->ssao.depth, 0, i, RD::TEXTURE_SLICE_2D_ARRAY);
Revert "Replace SAO implementation with MSSAO"
2020-10-18 22:27:51 +00:00
rb->ssao.depth_slices.push_back(slice);
Add named resources and debug labels in RenderDoc
2021-01-24 06:21:54 +00:00
RD::get_singleton()->set_resource_name(rb->ssao.depth_slices[i], "SSAO Depth Mip " + itos(i) + " ");
Re-implemented screen space ambient occlusion
2020-01-25 10:18:55 +00:00
}
}
{
RD::TextureFormat tf;
Port ASSAO to Godot to replace SAO
2020-12-08 05:37:09 +00:00
tf.format = RD::DATA_FORMAT_R8G8_UNORM;
tf.texture_type = RD::TEXTURE_TYPE_2D_ARRAY;
tf.width = buffer_width;
tf.height = buffer_height;
tf.array_layers = 4;
tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_STORAGE_BIT;
rb->ssao.ao_deinterleaved = RD::get_singleton()->texture_create(tf, RD::TextureView());
Add named resources and debug labels in RenderDoc
2021-01-24 06:21:54 +00:00
RD::get_singleton()->set_resource_name(rb->ssao.ao_deinterleaved, "SSAO De-interleaved Array");
Port ASSAO to Godot to replace SAO
2020-12-08 05:37:09 +00:00
for (uint32_t i = 0; i < 4; i++) {
RID slice = RD::get_singleton()->texture_create_shared_from_slice(RD::TextureView(), rb->ssao.ao_deinterleaved, i, 0);
rb->ssao.ao_deinterleaved_slices.push_back(slice);
Add named resources and debug labels in RenderDoc
2021-01-24 06:21:54 +00:00
RD::get_singleton()->set_resource_name(rb->ssao.ao_deinterleaved_slices[i], "SSAO De-interleaved Array Layer " + itos(i) + " ");
Port ASSAO to Godot to replace SAO
2020-12-08 05:37:09 +00:00
}
}
{
RD::TextureFormat tf;
tf.format = RD::DATA_FORMAT_R8G8_UNORM;
tf.texture_type = RD::TEXTURE_TYPE_2D_ARRAY;
tf.width = buffer_width;
tf.height = buffer_height;
tf.array_layers = 4;
Re-implemented screen space ambient occlusion
2020-01-25 10:18:55 +00:00
tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_STORAGE_BIT;
Port ASSAO to Godot to replace SAO
2020-12-08 05:37:09 +00:00
rb->ssao.ao_pong = RD::get_singleton()->texture_create(tf, RD::TextureView());
Add named resources and debug labels in RenderDoc
2021-01-24 06:21:54 +00:00
RD::get_singleton()->set_resource_name(rb->ssao.ao_pong, "SSAO De-interleaved Array Pong");
Port ASSAO to Godot to replace SAO
2020-12-08 05:37:09 +00:00
for (uint32_t i = 0; i < 4; i++) {
RID slice = RD::get_singleton()->texture_create_shared_from_slice(RD::TextureView(), rb->ssao.ao_pong, i, 0);
rb->ssao.ao_pong_slices.push_back(slice);
Add named resources and debug labels in RenderDoc
2021-01-24 06:21:54 +00:00
RD::get_singleton()->set_resource_name(rb->ssao.ao_deinterleaved_slices[i], "SSAO De-interleaved Array Layer " + itos(i) + " Pong");
Port ASSAO to Godot to replace SAO
2020-12-08 05:37:09 +00:00
}
Re-implemented screen space ambient occlusion
2020-01-25 10:18:55 +00:00
}
Port ASSAO to Godot to replace SAO
2020-12-08 05:37:09 +00:00
{
RD::TextureFormat tf;
tf.format = RD::DATA_FORMAT_R8_UNORM;
tf.width = half_width;
tf.height = half_height;
tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_STORAGE_BIT;
rb->ssao.importance_map[0] = RD::get_singleton()->texture_create(tf, RD::TextureView());
Add named resources and debug labels in RenderDoc
2021-01-24 06:21:54 +00:00
RD::get_singleton()->set_resource_name(rb->ssao.importance_map[0], "SSAO Importance Map");
Port ASSAO to Godot to replace SAO
2020-12-08 05:37:09 +00:00
rb->ssao.importance_map[1] = RD::get_singleton()->texture_create(tf, RD::TextureView());
Add named resources and debug labels in RenderDoc
2021-01-24 06:21:54 +00:00
RD::get_singleton()->set_resource_name(rb->ssao.importance_map[1], "SSAO Importance Map Pong");
Port ASSAO to Godot to replace SAO
2020-12-08 05:37:09 +00:00
}
{
Re-implemented screen space ambient occlusion
2020-01-25 10:18:55 +00:00
RD::TextureFormat tf;
Revert "Replace SAO implementation with MSSAO"
2020-10-18 22:27:51 +00:00
tf.format = RD::DATA_FORMAT_R8_UNORM;
tf.width = rb->width;
tf.height = rb->height;
Re-implemented screen space ambient occlusion
2020-01-25 10:18:55 +00:00
tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_STORAGE_BIT;
Port ASSAO to Godot to replace SAO
2020-12-08 05:37:09 +00:00
rb->ssao.ao_final = RD::get_singleton()->texture_create(tf, RD::TextureView());
Add named resources and debug labels in RenderDoc
2021-01-24 06:21:54 +00:00
RD::get_singleton()->set_resource_name(rb->ssao.ao_final, "SSAO Final");
Port ASSAO to Godot to replace SAO
2020-12-08 05:37:09 +00:00
_render_buffers_uniform_set_changed(p_render_buffers);
Re-implemented screen space ambient occlusion
2020-01-25 10:18:55 +00:00
}
Port ASSAO to Godot to replace SAO
2020-12-08 05:37:09 +00:00
ssao_using_half_size = ssao_half_size;
uniform_sets_are_invalid = true;
Re-implemented screen space ambient occlusion
2020-01-25 10:18:55 +00:00
}
Port ASSAO to Godot to replace SAO
2020-12-08 05:37:09 +00:00
EffectsRD::SSAOSettings settings;
settings.radius = env->ssao_radius;
settings.intensity = env->ssao_intensity;
settings.power = env->ssao_power;
settings.detail = env->ssao_detail;
settings.horizon = env->ssao_horizon;
settings.sharpness = env->ssao_sharpness;
settings.quality = ssao_quality;
settings.half_size = ssao_half_size;
settings.adaptive_target = ssao_adaptive_target;
settings.blur_passes = ssao_blur_passes;
settings.fadeout_from = ssao_fadeout_from;
settings.fadeout_to = ssao_fadeout_to;
SSAO cleanup and fixes
2020-12-31 18:33:38 +00:00
settings.full_screen_size = Size2i(rb->width, rb->height);
Port ASSAO to Godot to replace SAO
2020-12-08 05:37:09 +00:00
settings.half_screen_size = Size2i(buffer_width, buffer_height);
SSAO cleanup and fixes
2020-12-31 18:33:38 +00:00
settings.quarter_screen_size = Size2i(half_width, half_height);
Port ASSAO to Godot to replace SAO
2020-12-08 05:37:09 +00:00
storage->get_effects()->generate_ssao(rb->depth_texture, p_normal_buffer, rb->ssao.depth, rb->ssao.depth_slices, rb->ssao.ao_deinterleaved, rb->ssao.ao_deinterleaved_slices, rb->ssao.ao_pong, rb->ssao.ao_pong_slices, rb->ssao.ao_final, rb->ssao.importance_map[0], rb->ssao.importance_map[1], p_projection, settings, uniform_sets_are_invalid);
-Refactored post processing, re-added glow and added a mix blend mode.
2020-01-10 00:40:26 +00:00
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
void RendererSceneRenderRD::_render_buffers_post_process_and_tonemap(RID p_render_buffers, RID p_environment, RID p_camera_effects, const CameraMatrix &p_projection) {
-Refactored post processing, re-added glow and added a mix blend mode.
2020-01-10 00:40:26 +00:00
RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers);
ERR_FAIL_COND(!rb);
Fix spelling of a var, a struct, and message output
2020-07-17 16:02:06 +00:00
Environment *env = environment_owner.getornull(p_environment);
-Refactored post processing, re-added glow and added a mix blend mode.
2020-01-10 00:40:26 +00:00
//glow (if enabled)
WIP CameraEffects implementation (bokeh not working for now)
2020-01-13 18:37:24 +00:00
CameraEffects *camfx = camera_effects_owner.getornull(p_camera_effects);
-Refactored post processing, re-added glow and added a mix blend mode.
2020-01-10 00:40:26 +00:00
Auto exposure re-implemented in Vulkan
2020-01-12 01:26:52 +00:00
bool can_use_effects = rb->width >= 8 && rb->height >= 8;
WIP CameraEffects implementation (bokeh not working for now)
2020-01-13 18:37:24 +00:00
if (can_use_effects && camfx && (camfx->dof_blur_near_enabled || camfx->dof_blur_far_enabled) && camfx->dof_blur_amount > 0.0) {
if (rb->blur[0].texture.is_null()) {
_allocate_blur_textures(rb);
Re-implemented screen space ambient occlusion
2020-01-25 10:18:55 +00:00
_render_buffers_uniform_set_changed(p_render_buffers);
WIP CameraEffects implementation (bokeh not working for now)
2020-01-13 18:37:24 +00:00
}
DOF fully implemented, can be edited on the fly.
2020-01-16 00:23:21 +00:00
float bokeh_size = camfx->dof_blur_amount * 64.0;
storage->get_effects()->bokeh_dof(rb->texture, rb->depth_texture, Size2i(rb->width, rb->height), rb->blur[0].mipmaps[0].texture, rb->blur[1].mipmaps[0].texture, rb->blur[0].mipmaps[1].texture, camfx->dof_blur_far_enabled, camfx->dof_blur_far_distance, camfx->dof_blur_far_transition, camfx->dof_blur_near_enabled, camfx->dof_blur_near_distance, camfx->dof_blur_near_transition, bokeh_size, dof_blur_bokeh_shape, dof_blur_quality, dof_blur_use_jitter, p_projection.get_z_near(), p_projection.get_z_far(), p_projection.is_orthogonal());
WIP CameraEffects implementation (bokeh not working for now)
2020-01-13 18:37:24 +00:00
}
Auto exposure re-implemented in Vulkan
2020-01-12 01:26:52 +00:00
if (can_use_effects && env && env->auto_exposure) {
if (rb->luminance.current.is_null()) {
_allocate_luminance_textures(rb);
Re-implemented screen space ambient occlusion
2020-01-25 10:18:55 +00:00
_render_buffers_uniform_set_changed(p_render_buffers);
Auto exposure re-implemented in Vulkan
2020-01-12 01:26:52 +00:00
}
bool set_immediate = env->auto_exposure_version != rb->auto_exposure_version;
rb->auto_exposure_version = env->auto_exposure_version;
double step = env->auto_exp_speed * time_step;
storage->get_effects()->luminance_reduction(rb->texture, Size2i(rb->width, rb->height), rb->luminance.reduce, rb->luminance.current, env->min_luminance, env->max_luminance, step, set_immediate);
//swap final reduce with prev luminance
SWAP(rb->luminance.current, rb->luminance.reduce.write[rb->luminance.reduce.size() - 1]);
Reorganize rendering server. -Made RenderingServerScene abstract, allowing reimplementation -RenderingServerRaster -> RenderingServerDefault, but this class is going away soon.
2020-12-03 21:09:47 +00:00
RenderingServerDefault::redraw_request(); //redraw all the time if auto exposure rendering is on
Auto exposure re-implemented in Vulkan
2020-01-12 01:26:52 +00:00
}
-Refactored post processing, re-added glow and added a mix blend mode.
2020-01-10 00:40:26 +00:00
int max_glow_level = -1;
Auto exposure re-implemented in Vulkan
2020-01-12 01:26:52 +00:00
if (can_use_effects && env && env->glow_enabled) {
-Refactored post processing, re-added glow and added a mix blend mode.
2020-01-10 00:40:26 +00:00
/* see that blur textures are allocated */
Optimize Glow with local memory
2020-09-19 22:02:32 +00:00
if (rb->blur[1].texture.is_null()) {
-Refactored post processing, re-added glow and added a mix blend mode.
2020-01-10 00:40:26 +00:00
_allocate_blur_textures(rb);
Re-implemented screen space ambient occlusion
2020-01-25 10:18:55 +00:00
_render_buffers_uniform_set_changed(p_render_buffers);
-Refactored post processing, re-added glow and added a mix blend mode.
2020-01-10 00:40:26 +00:00
}
Renaming of servers for coherency. VisualServer -> RenderingServer PhysicsServer -> PhysicsServer3D Physics2DServer -> PhysicsServer2D NavigationServer -> NavigationServer3D Navigation2DServer -> NavigationServer2D Also renamed corresponding files.
2020-03-27 18:21:27 +00:00
for (int i = 0; i < RS::MAX_GLOW_LEVELS; i++) {
Optimize Glow with local memory
2020-09-19 22:02:32 +00:00
if (env->glow_levels[i] > 0.0) {
-Refactored post processing, re-added glow and added a mix blend mode.
2020-01-10 00:40:26 +00:00
if (i >= rb->blur[1].mipmaps.size()) {
max_glow_level = rb->blur[1].mipmaps.size() - 1;
} else {
max_glow_level = i;
}
}
}
for (int i = 0; i < (max_glow_level + 1); i++) {
int vp_w = rb->blur[1].mipmaps[i].width;
int vp_h = rb->blur[1].mipmaps[i].height;
if (i == 0) {
Auto exposure re-implemented in Vulkan
2020-01-12 01:26:52 +00:00
RID luminance_texture;
if (env->auto_exposure && rb->luminance.current.is_valid()) {
luminance_texture = rb->luminance.current;
}
Optimize Glow with local memory
2020-09-19 22:02:32 +00:00
storage->get_effects()->gaussian_glow(rb->texture, rb->blur[1].mipmaps[i].texture, Size2i(vp_w, vp_h), env->glow_strength, glow_high_quality, true, env->glow_hdr_luminance_cap, env->exposure, env->glow_bloom, env->glow_hdr_bleed_threshold, env->glow_hdr_bleed_scale, luminance_texture, env->auto_exp_scale);
-Refactored post processing, re-added glow and added a mix blend mode.
2020-01-10 00:40:26 +00:00
} else {
Optimize Glow with local memory
2020-09-19 22:02:32 +00:00
storage->get_effects()->gaussian_glow(rb->blur[1].mipmaps[i - 1].texture, rb->blur[1].mipmaps[i].texture, Size2i(vp_w, vp_h), env->glow_strength, glow_high_quality);
-Refactored post processing, re-added glow and added a mix blend mode.
2020-01-10 00:40:26 +00:00
}
}
}
{
//tonemap
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
EffectsRD::TonemapSettings tonemap;
-Refactored post processing, re-added glow and added a mix blend mode.
2020-01-10 00:40:26 +00:00
Auto exposure re-implemented in Vulkan
2020-01-12 01:26:52 +00:00
if (can_use_effects && env && env->auto_exposure && rb->luminance.current.is_valid()) {
tonemap.use_auto_exposure = true;
tonemap.exposure_texture = rb->luminance.current;
tonemap.auto_exposure_grey = env->auto_exp_scale;
} else {
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
tonemap.exposure_texture = storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_WHITE);
Auto exposure re-implemented in Vulkan
2020-01-12 01:26:52 +00:00
}
if (can_use_effects && env && env->glow_enabled) {
-Refactored post processing, re-added glow and added a mix blend mode.
2020-01-10 00:40:26 +00:00
tonemap.use_glow = true;
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
tonemap.glow_mode = EffectsRD::TonemapSettings::GlowMode(env->glow_blend_mode);
Renaming of servers for coherency. VisualServer -> RenderingServer PhysicsServer -> PhysicsServer3D Physics2DServer -> PhysicsServer2D NavigationServer -> NavigationServer3D Navigation2DServer -> NavigationServer2D Also renamed corresponding files.
2020-03-27 18:21:27 +00:00
tonemap.glow_intensity = env->glow_blend_mode == RS::ENV_GLOW_BLEND_MODE_MIX ? env->glow_mix : env->glow_intensity;
Optimize Glow with local memory
2020-09-19 22:02:32 +00:00
for (int i = 0; i < RS::MAX_GLOW_LEVELS; i++) {
tonemap.glow_levels[i] = env->glow_levels[i];
}
-Refactored post processing, re-added glow and added a mix blend mode.
2020-01-10 00:40:26 +00:00
tonemap.glow_texture_size.x = rb->blur[1].mipmaps[0].width;
tonemap.glow_texture_size.y = rb->blur[1].mipmaps[0].height;
Move glow upscale quality to a global setting, for consistency
2020-03-30 13:46:03 +00:00
tonemap.glow_use_bicubic_upscale = glow_bicubic_upscale;
-Refactored post processing, re-added glow and added a mix blend mode.
2020-01-10 00:40:26 +00:00
tonemap.glow_texture = rb->blur[1].texture;
} else {
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
tonemap.glow_texture = storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_BLACK);
-Refactored post processing, re-added glow and added a mix blend mode.
2020-01-10 00:40:26 +00:00
}
Add screen space AA option, with FXAA implementation.
2020-04-12 04:49:10 +00:00
if (rb->screen_space_aa == RS::VIEWPORT_SCREEN_SPACE_AA_FXAA) {
tonemap.use_fxaa = true;
}
Add a debanding property to Viewport It can be enabled in the Project Settings (`rendering/quality/screen_filters/use_debanding`). It's disabled by default as it has a small performance impact and can make PNG screenshots much larger (due to how dithering works). It will also slightly brighten the scene's dark areas. As a result, it should be enabled only when banding is noticeable enough. This closes #17006.
2020-04-20 21:34:47 +00:00
tonemap.use_debanding = rb->use_debanding;
Add screen space AA option, with FXAA implementation.
2020-04-12 04:49:10 +00:00
tonemap.texture_size = Vector2i(rb->width, rb->height);
-Refactored post processing, re-added glow and added a mix blend mode.
2020-01-10 00:40:26 +00:00
if (env) {
tonemap.tonemap_mode = env->tone_mapper;
tonemap.white = env->white;
tonemap.exposure = env->exposure;
}
Refactored Mesh internals and formats. -Changed how mesh data is organized, hoping to make it more efficient on Vulkan and GLES. -Removed compression, it now always uses the most efficient format. -Added support for custom arrays (up to 8 custom formats) -Added support for 8 weights in skeleton data. -Added a simple optional versioning system for imported assets, to reimport if binary is newer -Fixes #43979 (I needed to test) WARNING: -NOT backwards compatible with previous 4.x-devel, will most likely never be, but it will force reimport scenes due to version change. -NOT backwards compatible with 3.x scenes, this will be eventually re-added. -Skeletons not working any longer, will fix in next PR.
2020-12-02 01:40:47 +00:00
tonemap.use_color_correction = false;
tonemap.use_1d_color_correction = false;
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
tonemap.color_correction_texture = storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_3D_WHITE);
Refactored Mesh internals and formats. -Changed how mesh data is organized, hoping to make it more efficient on Vulkan and GLES. -Removed compression, it now always uses the most efficient format. -Added support for custom arrays (up to 8 custom formats) -Added support for 8 weights in skeleton data. -Added a simple optional versioning system for imported assets, to reimport if binary is newer -Fixes #43979 (I needed to test) WARNING: -NOT backwards compatible with previous 4.x-devel, will most likely never be, but it will force reimport scenes due to version change. -NOT backwards compatible with 3.x scenes, this will be eventually re-added. -Skeletons not working any longer, will fix in next PR.
2020-12-02 01:40:47 +00:00
Environment brightness, contrast, saturation restore with color correction. Allow gradients and 2d images. Use shader versions for LUT in tonemap Co-authored-by: alex-poe <3957610+CptPotato@users.noreply.github.com> Co-authored-by: QbieShay <cislaghi.ilaria@gmail.com> Co-authored-by: Clay John <claynjohn@gmail.com>
2020-11-23 01:51:31 +00:00
if (can_use_effects && env) {
tonemap.use_bcs = env->adjustments_enabled;
tonemap.brightness = env->adjustments_brightness;
tonemap.contrast = env->adjustments_contrast;
tonemap.saturation = env->adjustments_saturation;
if (env->adjustments_enabled && env->color_correction.is_valid()) {
tonemap.use_color_correction = true;
Refactored Mesh internals and formats. -Changed how mesh data is organized, hoping to make it more efficient on Vulkan and GLES. -Removed compression, it now always uses the most efficient format. -Added support for custom arrays (up to 8 custom formats) -Added support for 8 weights in skeleton data. -Added a simple optional versioning system for imported assets, to reimport if binary is newer -Fixes #43979 (I needed to test) WARNING: -NOT backwards compatible with previous 4.x-devel, will most likely never be, but it will force reimport scenes due to version change. -NOT backwards compatible with 3.x scenes, this will be eventually re-added. -Skeletons not working any longer, will fix in next PR.
2020-12-02 01:40:47 +00:00
tonemap.use_1d_color_correction = env->use_1d_color_correction;
Environment brightness, contrast, saturation restore with color correction. Allow gradients and 2d images. Use shader versions for LUT in tonemap Co-authored-by: alex-poe <3957610+CptPotato@users.noreply.github.com> Co-authored-by: QbieShay <cislaghi.ilaria@gmail.com> Co-authored-by: Clay John <claynjohn@gmail.com>
2020-11-23 01:51:31 +00:00
tonemap.color_correction_texture = storage->texture_get_rd_texture(env->color_correction);
}
}
-Refactored post processing, re-added glow and added a mix blend mode.
2020-01-10 00:40:26 +00:00
storage->get_effects()->tonemapper(rb->texture, storage->render_target_get_rd_framebuffer(rb->render_target), tonemap);
}
storage->render_target_disable_clear_request(rb->render_target);
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
void RendererSceneRenderRD::_render_buffers_debug_draw(RID p_render_buffers, RID p_shadow_atlas) {
EffectsRD *effects = storage->get_effects();
-Refactored post processing, re-added glow and added a mix blend mode.
2020-01-10 00:40:26 +00:00
RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers);
ERR_FAIL_COND(!rb);
Renaming of servers for coherency. VisualServer -> RenderingServer PhysicsServer -> PhysicsServer3D Physics2DServer -> PhysicsServer2D NavigationServer -> NavigationServer3D Navigation2DServer -> NavigationServer2D Also renamed corresponding files.
2020-03-27 18:21:27 +00:00
if (debug_draw == RS::VIEWPORT_DEBUG_DRAW_SHADOW_ATLAS) {
-Refactored post processing, re-added glow and added a mix blend mode.
2020-01-10 00:40:26 +00:00
if (p_shadow_atlas.is_valid()) {
RID shadow_atlas_texture = shadow_atlas_get_texture(p_shadow_atlas);
Size2 rtsize = storage->render_target_get_size(rb->render_target);
Moved most of the effect code to compute. Simplifies it and improves performance. Improves image barrier handling per mipmap on RenderingDeviceVulkan
2020-04-11 17:43:12 +00:00
effects->copy_to_fb_rect(shadow_atlas_texture, storage->render_target_get_rd_framebuffer(rb->render_target), Rect2i(Vector2(), rtsize / 2), false, true);
-Refactored post processing, re-added glow and added a mix blend mode.
2020-01-10 00:40:26 +00:00
}
}
Renaming of servers for coherency. VisualServer -> RenderingServer PhysicsServer -> PhysicsServer3D Physics2DServer -> PhysicsServer2D NavigationServer -> NavigationServer3D Navigation2DServer -> NavigationServer2D Also renamed corresponding files.
2020-03-27 18:21:27 +00:00
if (debug_draw == RS::VIEWPORT_DEBUG_DRAW_DIRECTIONAL_SHADOW_ATLAS) {
-Refactored post processing, re-added glow and added a mix blend mode.
2020-01-10 00:40:26 +00:00
if (directional_shadow_get_texture().is_valid()) {
RID shadow_atlas_texture = directional_shadow_get_texture();
Size2 rtsize = storage->render_target_get_size(rb->render_target);
Moved most of the effect code to compute. Simplifies it and improves performance. Improves image barrier handling per mipmap on RenderingDeviceVulkan
2020-04-11 17:43:12 +00:00
effects->copy_to_fb_rect(shadow_atlas_texture, storage->render_target_get_rd_framebuffer(rb->render_target), Rect2i(Vector2(), rtsize / 2), false, true);
-Refactored post processing, re-added glow and added a mix blend mode.
2020-01-10 00:40:26 +00:00
}
}
Auto exposure re-implemented in Vulkan
2020-01-12 01:26:52 +00:00
Implement decals Also implemented decal atlas, so projectors and other stuff can be added. Sidenote: Had to make RID hashable, so some unrelated includes changed in order to include it in hashfuncs.h
2020-04-14 03:05:21 +00:00
if (debug_draw == RS::VIEWPORT_DEBUG_DRAW_DECAL_ATLAS) {
RID decal_atlas = storage->decal_atlas_get_texture();
if (decal_atlas.is_valid()) {
Size2 rtsize = storage->render_target_get_size(rb->render_target);
effects->copy_to_fb_rect(decal_atlas, storage->render_target_get_rd_framebuffer(rb->render_target), Rect2i(Vector2(), rtsize / 2), false, false, true);
}
}
Renaming of servers for coherency. VisualServer -> RenderingServer PhysicsServer -> PhysicsServer3D Physics2DServer -> PhysicsServer2D NavigationServer -> NavigationServer3D Navigation2DServer -> NavigationServer2D Also renamed corresponding files.
2020-03-27 18:21:27 +00:00
if (debug_draw == RS::VIEWPORT_DEBUG_DRAW_SCENE_LUMINANCE) {
Auto exposure re-implemented in Vulkan
2020-01-12 01:26:52 +00:00
if (rb->luminance.current.is_valid()) {
Size2 rtsize = storage->render_target_get_size(rb->render_target);
Moved most of the effect code to compute. Simplifies it and improves performance. Improves image barrier handling per mipmap on RenderingDeviceVulkan
2020-04-11 17:43:12 +00:00
effects->copy_to_fb_rect(rb->luminance.current, storage->render_target_get_rd_framebuffer(rb->render_target), Rect2(Vector2(), rtsize / 8), false, true);
Auto exposure re-implemented in Vulkan
2020-01-12 01:26:52 +00:00
}
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
}
Port ASSAO to Godot to replace SAO
2020-12-08 05:37:09 +00:00
if (debug_draw == RS::VIEWPORT_DEBUG_DRAW_SSAO && rb->ssao.ao_final.is_valid()) {
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
Size2 rtsize = storage->render_target_get_size(rb->render_target);
Port ASSAO to Godot to replace SAO
2020-12-08 05:37:09 +00:00
RID ao_buf = rb->ssao.ao_final;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
effects->copy_to_fb_rect(ao_buf, storage->render_target_get_rd_framebuffer(rb->render_target), Rect2(Vector2(), rtsize), false, true);
}
if (debug_draw == RS::VIEWPORT_DEBUG_DRAW_NORMAL_BUFFER && _render_buffers_get_normal_texture(p_render_buffers).is_valid()) {
Size2 rtsize = storage->render_target_get_size(rb->render_target);
effects->copy_to_fb_rect(_render_buffers_get_normal_texture(p_render_buffers), storage->render_target_get_rd_framebuffer(rb->render_target), Rect2(Vector2(), rtsize), false, false);
}
Several GI related optimizations and fixes -SDFGI direct light is done over many frames -SDFGI Changed settings for rays/frame -SDFGI Misc optimizations -SDFGI Bug fix on probe scroll -GIProbe was not working, got it to work again -GIProbe dynamic objects were not working, fixed -Added a half size GI option.
2021-01-22 23:50:24 +00:00
if (debug_draw == RS::VIEWPORT_DEBUG_DRAW_GI_BUFFER && rb->ambient_buffer.is_valid()) {
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
Size2 rtsize = storage->render_target_get_size(rb->render_target);
Several GI related optimizations and fixes -SDFGI direct light is done over many frames -SDFGI Changed settings for rays/frame -SDFGI Misc optimizations -SDFGI Bug fix on probe scroll -GIProbe was not working, got it to work again -GIProbe dynamic objects were not working, fixed -Added a half size GI option.
2021-01-22 23:50:24 +00:00
RID ambient_texture = rb->ambient_buffer;
RID reflection_texture = rb->reflection_buffer;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
effects->copy_to_fb_rect(ambient_texture, storage->render_target_get_rd_framebuffer(rb->render_target), Rect2(Vector2(), rtsize), false, false, false, true, reflection_texture);
}
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
void RendererSceneRenderRD::environment_set_adjustment(RID p_env, bool p_enable, float p_brightness, float p_contrast, float p_saturation, bool p_use_1d_color_correction, RID p_color_correction) {
Environment brightness, contrast, saturation restore with color correction. Allow gradients and 2d images. Use shader versions for LUT in tonemap Co-authored-by: alex-poe <3957610+CptPotato@users.noreply.github.com> Co-authored-by: QbieShay <cislaghi.ilaria@gmail.com> Co-authored-by: Clay John <claynjohn@gmail.com>
2020-11-23 01:51:31 +00:00
Environment *env = environment_owner.getornull(p_env);
ERR_FAIL_COND(!env);
env->adjustments_enabled = p_enable;
env->adjustments_brightness = p_brightness;
env->adjustments_contrast = p_contrast;
env->adjustments_saturation = p_saturation;
env->use_1d_color_correction = p_use_1d_color_correction;
env->color_correction = p_color_correction;
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
void RendererSceneRenderRD::_sdfgi_debug_draw(RID p_render_buffers, const CameraMatrix &p_projection, const Transform &p_transform) {
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers);
ERR_FAIL_COND(!rb);
if (!rb->sdfgi) {
return; //eh
}
if (!rb->sdfgi->debug_uniform_set.is_valid() || !RD::get_singleton()->uniform_set_is_valid(rb->sdfgi->debug_uniform_set)) {
Vector<RD::Uniform> uniforms;
{
RD::Uniform u;
u.binding = 1;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_TEXTURE;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
for (uint32_t i = 0; i < SDFGI::MAX_CASCADES; i++) {
if (i < rb->sdfgi->cascades.size()) {
u.ids.push_back(rb->sdfgi->cascades[i].sdf_tex);
} else {
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
u.ids.push_back(storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_3D_WHITE));
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
}
}
uniforms.push_back(u);
}
{
RD::Uniform u;
u.binding = 2;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_TEXTURE;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
for (uint32_t i = 0; i < SDFGI::MAX_CASCADES; i++) {
if (i < rb->sdfgi->cascades.size()) {
u.ids.push_back(rb->sdfgi->cascades[i].light_tex);
} else {
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
u.ids.push_back(storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_3D_WHITE));
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
}
}
uniforms.push_back(u);
}
{
RD::Uniform u;
u.binding = 3;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_TEXTURE;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
for (uint32_t i = 0; i < SDFGI::MAX_CASCADES; i++) {
if (i < rb->sdfgi->cascades.size()) {
u.ids.push_back(rb->sdfgi->cascades[i].light_aniso_0_tex);
} else {
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
u.ids.push_back(storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_3D_WHITE));
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
}
}
uniforms.push_back(u);
}
{
RD::Uniform u;
u.binding = 4;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_TEXTURE;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
for (uint32_t i = 0; i < SDFGI::MAX_CASCADES; i++) {
if (i < rb->sdfgi->cascades.size()) {
u.ids.push_back(rb->sdfgi->cascades[i].light_aniso_1_tex);
} else {
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
u.ids.push_back(storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_3D_WHITE));
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
}
}
uniforms.push_back(u);
}
{
RD::Uniform u;
u.binding = 5;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_TEXTURE;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
u.ids.push_back(rb->sdfgi->occlusion_texture);
uniforms.push_back(u);
}
{
RD::Uniform u;
u.binding = 8;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_SAMPLER;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
u.ids.push_back(storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED));
uniforms.push_back(u);
}
{
RD::Uniform u;
u.binding = 9;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
u.ids.push_back(rb->sdfgi->cascades_ubo);
uniforms.push_back(u);
}
{
RD::Uniform u;
u.binding = 10;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_IMAGE;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
u.ids.push_back(rb->texture);
uniforms.push_back(u);
}
{
RD::Uniform u;
u.binding = 11;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_TEXTURE;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
u.ids.push_back(rb->sdfgi->lightprobe_texture);
uniforms.push_back(u);
}
rb->sdfgi->debug_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, sdfgi_shader.debug_shader_version, 0);
}
RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin();
RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, sdfgi_shader.debug_pipeline);
RD::get_singleton()->compute_list_bind_uniform_set(compute_list, rb->sdfgi->debug_uniform_set, 0);
SDGIShader::DebugPushConstant push_constant;
push_constant.grid_size[0] = rb->sdfgi->cascade_size;
push_constant.grid_size[1] = rb->sdfgi->cascade_size;
push_constant.grid_size[2] = rb->sdfgi->cascade_size;
push_constant.max_cascades = rb->sdfgi->cascades.size();
push_constant.screen_size[0] = rb->width;
push_constant.screen_size[1] = rb->height;
push_constant.probe_axis_size = rb->sdfgi->probe_axis_count;
push_constant.use_occlusion = rb->sdfgi->uses_occlusion;
push_constant.y_mult = rb->sdfgi->y_mult;
Vector2 vp_half = p_projection.get_viewport_half_extents();
push_constant.cam_extent[0] = vp_half.x;
push_constant.cam_extent[1] = vp_half.y;
push_constant.cam_extent[2] = -p_projection.get_z_near();
push_constant.cam_transform[0] = p_transform.basis.elements[0][0];
push_constant.cam_transform[1] = p_transform.basis.elements[1][0];
push_constant.cam_transform[2] = p_transform.basis.elements[2][0];
push_constant.cam_transform[3] = 0;
push_constant.cam_transform[4] = p_transform.basis.elements[0][1];
push_constant.cam_transform[5] = p_transform.basis.elements[1][1];
push_constant.cam_transform[6] = p_transform.basis.elements[2][1];
push_constant.cam_transform[7] = 0;
push_constant.cam_transform[8] = p_transform.basis.elements[0][2];
push_constant.cam_transform[9] = p_transform.basis.elements[1][2];
push_constant.cam_transform[10] = p_transform.basis.elements[2][2];
push_constant.cam_transform[11] = 0;
push_constant.cam_transform[12] = p_transform.origin.x;
push_constant.cam_transform[13] = p_transform.origin.y;
push_constant.cam_transform[14] = p_transform.origin.z;
push_constant.cam_transform[15] = 1;
RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(SDGIShader::DebugPushConstant));
RD::get_singleton()->compute_list_dispatch_threads(compute_list, rb->width, rb->height, 1, 8, 8, 1);
RD::get_singleton()->compute_list_end();
Size2 rtsize = storage->render_target_get_size(rb->render_target);
storage->get_effects()->copy_to_fb_rect(rb->texture, storage->render_target_get_rd_framebuffer(rb->render_target), Rect2(Vector2(), rtsize), true);
-Refactored post processing, re-added glow and added a mix blend mode.
2020-01-10 00:40:26 +00:00
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
RID RendererSceneRenderRD::render_buffers_get_back_buffer_texture(RID p_render_buffers) {
-Refactored post processing, re-added glow and added a mix blend mode.
2020-01-10 00:40:26 +00:00
RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers);
ERR_FAIL_COND_V(!rb, RID());
Re-implemented screen space ambient occlusion
2020-01-25 10:18:55 +00:00
if (!rb->blur[0].texture.is_valid()) {
return RID(); //not valid at the moment
}
-Refactored post processing, re-added glow and added a mix blend mode.
2020-01-10 00:40:26 +00:00
return rb->blur[0].texture;
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
RID RendererSceneRenderRD::render_buffers_get_ao_texture(RID p_render_buffers) {
Re-implemented screen space ambient occlusion
2020-01-25 10:18:55 +00:00
RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers);
ERR_FAIL_COND_V(!rb, RID());
Port ASSAO to Godot to replace SAO
2020-12-08 05:37:09 +00:00
return rb->ssao.ao_final;
Re-implemented screen space ambient occlusion
2020-01-25 10:18:55 +00:00
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
RID RendererSceneRenderRD::render_buffers_get_gi_probe_buffer(RID p_render_buffers) {
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers);
ERR_FAIL_COND_V(!rb, RID());
if (rb->giprobe_buffer.is_null()) {
rb->giprobe_buffer = RD::get_singleton()->uniform_buffer_create(sizeof(GI::GIProbeData) * RenderBuffers::MAX_GIPROBES);
}
return rb->giprobe_buffer;
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
RID RendererSceneRenderRD::render_buffers_get_default_gi_probe_buffer() {
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
return default_giprobe_buffer;
}
Several GI related optimizations and fixes -SDFGI direct light is done over many frames -SDFGI Changed settings for rays/frame -SDFGI Misc optimizations -SDFGI Bug fix on probe scroll -GIProbe was not working, got it to work again -GIProbe dynamic objects were not working, fixed -Added a half size GI option.
2021-01-22 23:50:24 +00:00
RID RendererSceneRenderRD::render_buffers_get_gi_ambient_texture(RID p_render_buffers) {
RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers);
ERR_FAIL_COND_V(!rb, RID());
return rb->ambient_buffer;
}
RID RendererSceneRenderRD::render_buffers_get_gi_reflection_texture(RID p_render_buffers) {
RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers);
ERR_FAIL_COND_V(!rb, RID());
return rb->reflection_buffer;
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
uint32_t RendererSceneRenderRD::render_buffers_get_sdfgi_cascade_count(RID p_render_buffers) const {
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
const RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers);
ERR_FAIL_COND_V(!rb, 0);
ERR_FAIL_COND_V(!rb->sdfgi, 0);
return rb->sdfgi->cascades.size();
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
bool RendererSceneRenderRD::render_buffers_is_sdfgi_enabled(RID p_render_buffers) const {
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
const RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers);
ERR_FAIL_COND_V(!rb, false);
return rb->sdfgi != nullptr;
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
RID RendererSceneRenderRD::render_buffers_get_sdfgi_irradiance_probes(RID p_render_buffers) const {
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
const RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers);
ERR_FAIL_COND_V(!rb, RID());
ERR_FAIL_COND_V(!rb->sdfgi, RID());
return rb->sdfgi->lightprobe_texture;
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
Vector3 RendererSceneRenderRD::render_buffers_get_sdfgi_cascade_offset(RID p_render_buffers, uint32_t p_cascade) const {
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
const RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers);
ERR_FAIL_COND_V(!rb, Vector3());
ERR_FAIL_COND_V(!rb->sdfgi, Vector3());
ERR_FAIL_UNSIGNED_INDEX_V(p_cascade, rb->sdfgi->cascades.size(), Vector3());
return Vector3((Vector3i(1, 1, 1) * -int32_t(rb->sdfgi->cascade_size >> 1) + rb->sdfgi->cascades[p_cascade].position)) * rb->sdfgi->cascades[p_cascade].cell_size;
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
Vector3i RendererSceneRenderRD::render_buffers_get_sdfgi_cascade_probe_offset(RID p_render_buffers, uint32_t p_cascade) const {
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
const RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers);
ERR_FAIL_COND_V(!rb, Vector3i());
ERR_FAIL_COND_V(!rb->sdfgi, Vector3i());
ERR_FAIL_UNSIGNED_INDEX_V(p_cascade, rb->sdfgi->cascades.size(), Vector3i());
int32_t probe_divisor = rb->sdfgi->cascade_size / SDFGI::PROBE_DIVISOR;
return rb->sdfgi->cascades[p_cascade].position / probe_divisor;
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
float RendererSceneRenderRD::render_buffers_get_sdfgi_normal_bias(RID p_render_buffers) const {
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
const RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers);
ERR_FAIL_COND_V(!rb, 0);
ERR_FAIL_COND_V(!rb->sdfgi, 0);
return rb->sdfgi->normal_bias;
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
float RendererSceneRenderRD::render_buffers_get_sdfgi_cascade_probe_size(RID p_render_buffers, uint32_t p_cascade) const {
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
const RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers);
ERR_FAIL_COND_V(!rb, 0);
ERR_FAIL_COND_V(!rb->sdfgi, 0);
ERR_FAIL_UNSIGNED_INDEX_V(p_cascade, rb->sdfgi->cascades.size(), 0);
return float(rb->sdfgi->cascade_size) * rb->sdfgi->cascades[p_cascade].cell_size / float(rb->sdfgi->probe_axis_count - 1);
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
uint32_t RendererSceneRenderRD::render_buffers_get_sdfgi_cascade_probe_count(RID p_render_buffers) const {
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
const RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers);
ERR_FAIL_COND_V(!rb, 0);
ERR_FAIL_COND_V(!rb->sdfgi, 0);
return rb->sdfgi->probe_axis_count;
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
uint32_t RendererSceneRenderRD::render_buffers_get_sdfgi_cascade_size(RID p_render_buffers) const {
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
const RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers);
ERR_FAIL_COND_V(!rb, 0);
ERR_FAIL_COND_V(!rb->sdfgi, 0);
return rb->sdfgi->cascade_size;
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
bool RendererSceneRenderRD::render_buffers_is_sdfgi_using_occlusion(RID p_render_buffers) const {
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
const RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers);
ERR_FAIL_COND_V(!rb, false);
ERR_FAIL_COND_V(!rb->sdfgi, false);
return rb->sdfgi->uses_occlusion;
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
float RendererSceneRenderRD::render_buffers_get_sdfgi_energy(RID p_render_buffers) const {
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
const RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers);
Static analyzer fixes: Removes unused code in OS. Fixes return types. Fixes few typos.
2020-12-08 18:58:49 +00:00
ERR_FAIL_COND_V(!rb, 0.0);
ERR_FAIL_COND_V(!rb->sdfgi, 0.0);
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
return rb->sdfgi->energy;
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
RID RendererSceneRenderRD::render_buffers_get_sdfgi_occlusion_texture(RID p_render_buffers) const {
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
const RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers);
ERR_FAIL_COND_V(!rb, RID());
ERR_FAIL_COND_V(!rb->sdfgi, RID());
return rb->sdfgi->occlusion_texture;
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
bool RendererSceneRenderRD::render_buffers_has_volumetric_fog(RID p_render_buffers) const {
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
const RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers);
ERR_FAIL_COND_V(!rb, false);
return rb->volumetric_fog != nullptr;
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
RID RendererSceneRenderRD::render_buffers_get_volumetric_fog_texture(RID p_render_buffers) {
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
const RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers);
ERR_FAIL_COND_V(!rb || !rb->volumetric_fog, RID());
return rb->volumetric_fog->fog_map;
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
RID RendererSceneRenderRD::render_buffers_get_volumetric_fog_sky_uniform_set(RID p_render_buffers) {
Add fog to sky shaders
2020-08-18 04:12:51 +00:00
const RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers);
ERR_FAIL_COND_V(!rb, RID());
if (!rb->volumetric_fog) {
return RID();
}
return rb->volumetric_fog->sky_uniform_set;
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
float RendererSceneRenderRD::render_buffers_get_volumetric_fog_end(RID p_render_buffers) {
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
const RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers);
ERR_FAIL_COND_V(!rb || !rb->volumetric_fog, 0);
return rb->volumetric_fog->length;
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
float RendererSceneRenderRD::render_buffers_get_volumetric_fog_detail_spread(RID p_render_buffers) {
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
const RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers);
ERR_FAIL_COND_V(!rb || !rb->volumetric_fog, 0);
return rb->volumetric_fog->spread;
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
void RendererSceneRenderRD::render_buffers_configure(RID p_render_buffers, RID p_render_target, int p_width, int p_height, RS::ViewportMSAA p_msaa, RenderingServer::ViewportScreenSpaceAA p_screen_space_aa, bool p_use_debanding) {
Base 3D engine done, still untested, though.
2019-08-18 22:40:52 +00:00
RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers);
rb->width = p_width;
rb->height = p_height;
rb->render_target = p_render_target;
rb->msaa = p_msaa;
Add screen space AA option, with FXAA implementation.
2020-04-12 04:49:10 +00:00
rb->screen_space_aa = p_screen_space_aa;
Add a debanding property to Viewport It can be enabled in the Project Settings (`rendering/quality/screen_filters/use_debanding`). It's disabled by default as it has a small performance impact and can make PNG screenshots much larger (due to how dithering works). It will also slightly brighten the scene's dark areas. As a result, it should be enabled only when banding is noticeable enough. This closes #17006.
2020-04-20 21:34:47 +00:00
rb->use_debanding = p_use_debanding;
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
if (rb->cluster_builder == nullptr) {
rb->cluster_builder = memnew(ClusterBuilderRD);
}
rb->cluster_builder->set_shared(&cluster_builder_shared);
-Refactored post processing, re-added glow and added a mix blend mode.
2020-01-10 00:40:26 +00:00
_free_render_buffer_data(rb);
{
RD::TextureFormat tf;
tf.format = RD::DATA_FORMAT_R16G16B16A16_SFLOAT;
tf.width = rb->width;
tf.height = rb->height;
Implement MSAA
2020-04-12 18:33:57 +00:00
tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_STORAGE_BIT;
if (rb->msaa != RS::VIEWPORT_MSAA_DISABLED) {
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
tf.usage_bits |= RD::TEXTURE_USAGE_CAN_COPY_TO_BIT | RD::TEXTURE_USAGE_STORAGE_BIT;
Implement MSAA
2020-04-12 18:33:57 +00:00
} else {
tf.usage_bits |= RD::TEXTURE_USAGE_COLOR_ATTACHMENT_BIT;
}
-Refactored post processing, re-added glow and added a mix blend mode.
2020-01-10 00:40:26 +00:00
rb->texture = RD::get_singleton()->texture_create(tf, RD::TextureView());
}
WIP CameraEffects implementation (bokeh not working for now)
2020-01-13 18:37:24 +00:00
{
RD::TextureFormat tf;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
if (rb->msaa == RS::VIEWPORT_MSAA_DISABLED) {
tf.format = RD::get_singleton()->texture_is_format_supported_for_usage(RD::DATA_FORMAT_D24_UNORM_S8_UINT, RD::TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT) ? RD::DATA_FORMAT_D24_UNORM_S8_UINT : RD::DATA_FORMAT_D32_SFLOAT_S8_UINT;
} else {
tf.format = RD::DATA_FORMAT_R32_SFLOAT;
}
WIP CameraEffects implementation (bokeh not working for now)
2020-01-13 18:37:24 +00:00
tf.width = p_width;
tf.height = p_height;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT;
Implement MSAA
2020-04-12 18:33:57 +00:00
if (rb->msaa != RS::VIEWPORT_MSAA_DISABLED) {
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
tf.usage_bits |= RD::TEXTURE_USAGE_CAN_COPY_TO_BIT | RD::TEXTURE_USAGE_STORAGE_BIT;
} else {
tf.usage_bits |= RD::TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT;
Implement MSAA
2020-04-12 18:33:57 +00:00
}
WIP CameraEffects implementation (bokeh not working for now)
2020-01-13 18:37:24 +00:00
rb->depth_texture = RD::get_singleton()->texture_create(tf, RD::TextureView());
}
rb->data->configure(rb->texture, rb->depth_texture, p_width, p_height, p_msaa);
Re-implemented screen space ambient occlusion
2020-01-25 10:18:55 +00:00
_render_buffers_uniform_set_changed(p_render_buffers);
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
rb->cluster_builder->setup(Size2i(p_width, p_height), max_cluster_elements, rb->depth_texture, storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED), rb->texture);
Base 3D engine done, still untested, though.
2019-08-18 22:40:52 +00:00
}
Several GI related optimizations and fixes -SDFGI direct light is done over many frames -SDFGI Changed settings for rays/frame -SDFGI Misc optimizations -SDFGI Bug fix on probe scroll -GIProbe was not working, got it to work again -GIProbe dynamic objects were not working, fixed -Added a half size GI option.
2021-01-22 23:50:24 +00:00
void RendererSceneRenderRD::gi_set_use_half_resolution(bool p_enable) {
gi.half_resolution = p_enable;
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
void RendererSceneRenderRD::sub_surface_scattering_set_quality(RS::SubSurfaceScatteringQuality p_quality) {
Re-implement subsurface scattering. The size settings are more "just works", with default scale and depth scale values that don't need much tweaking. Additionally, a "skin" mode was added so skin looks better. EDIT: Cleaned up SSR filter shader a bit.
2020-04-04 02:42:26 +00:00
sss_quality = p_quality;
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
RS::SubSurfaceScatteringQuality RendererSceneRenderRD::sub_surface_scattering_get_quality() const {
Re-implement subsurface scattering. The size settings are more "just works", with default scale and depth scale values that don't need much tweaking. Additionally, a "skin" mode was added so skin looks better. EDIT: Cleaned up SSR filter shader a bit.
2020-04-04 02:42:26 +00:00
return sss_quality;
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
void RendererSceneRenderRD::sub_surface_scattering_set_scale(float p_scale, float p_depth_scale) {
Re-implement subsurface scattering. The size settings are more "just works", with default scale and depth scale values that don't need much tweaking. Additionally, a "skin" mode was added so skin looks better. EDIT: Cleaned up SSR filter shader a bit.
2020-04-04 02:42:26 +00:00
sss_scale = p_scale;
sss_depth_scale = p_depth_scale;
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
void RendererSceneRenderRD::shadows_quality_set(RS::ShadowQuality p_quality) {
Add proper quality settings to soft shadows
2020-04-10 09:30:36 +00:00
ERR_FAIL_INDEX_MSG(p_quality, RS::SHADOW_QUALITY_MAX, "Shadow quality too high, please see RenderingServer's ShadowQuality enum");
if (shadows_quality != p_quality) {
shadows_quality = p_quality;
switch (shadows_quality) {
case RS::SHADOW_QUALITY_HARD: {
penumbra_shadow_samples = 4;
soft_shadow_samples = 1;
shadows_quality_radius = 1.0;
} break;
case RS::SHADOW_QUALITY_SOFT_LOW: {
penumbra_shadow_samples = 8;
soft_shadow_samples = 4;
shadows_quality_radius = 2.0;
} break;
case RS::SHADOW_QUALITY_SOFT_MEDIUM: {
penumbra_shadow_samples = 12;
soft_shadow_samples = 8;
shadows_quality_radius = 2.0;
} break;
case RS::SHADOW_QUALITY_SOFT_HIGH: {
penumbra_shadow_samples = 24;
soft_shadow_samples = 16;
shadows_quality_radius = 3.0;
} break;
case RS::SHADOW_QUALITY_SOFT_ULTRA: {
penumbra_shadow_samples = 32;
soft_shadow_samples = 32;
shadows_quality_radius = 4.0;
} break;
case RS::SHADOW_QUALITY_MAX:
break;
}
get_vogel_disk(penumbra_shadow_kernel, penumbra_shadow_samples);
get_vogel_disk(soft_shadow_kernel, soft_shadow_samples);
}
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
void RendererSceneRenderRD::directional_shadow_quality_set(RS::ShadowQuality p_quality) {
Add proper quality settings to soft shadows
2020-04-10 09:30:36 +00:00
ERR_FAIL_INDEX_MSG(p_quality, RS::SHADOW_QUALITY_MAX, "Shadow quality too high, please see RenderingServer's ShadowQuality enum");
if (directional_shadow_quality != p_quality) {
directional_shadow_quality = p_quality;
switch (directional_shadow_quality) {
case RS::SHADOW_QUALITY_HARD: {
directional_penumbra_shadow_samples = 4;
directional_soft_shadow_samples = 1;
directional_shadow_quality_radius = 1.0;
} break;
case RS::SHADOW_QUALITY_SOFT_LOW: {
directional_penumbra_shadow_samples = 8;
directional_soft_shadow_samples = 4;
directional_shadow_quality_radius = 2.0;
} break;
case RS::SHADOW_QUALITY_SOFT_MEDIUM: {
directional_penumbra_shadow_samples = 12;
directional_soft_shadow_samples = 8;
directional_shadow_quality_radius = 2.0;
} break;
case RS::SHADOW_QUALITY_SOFT_HIGH: {
directional_penumbra_shadow_samples = 24;
directional_soft_shadow_samples = 16;
directional_shadow_quality_radius = 3.0;
} break;
case RS::SHADOW_QUALITY_SOFT_ULTRA: {
directional_penumbra_shadow_samples = 32;
directional_soft_shadow_samples = 32;
directional_shadow_quality_radius = 4.0;
} break;
case RS::SHADOW_QUALITY_MAX:
break;
}
get_vogel_disk(directional_penumbra_shadow_kernel, directional_penumbra_shadow_samples);
get_vogel_disk(directional_soft_shadow_kernel, directional_soft_shadow_samples);
}
Refactored shadowmapping. - Made shadow bias size independent, so it will remain when changing light or camera size. - Implemented normal offset bias, which greatly enhances quality. - Added transmission to subsurface scattering - Reimplemented shadow filter modes Closes #17260
2020-04-08 01:51:52 +00:00
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
int RendererSceneRenderRD::get_roughness_layers() const {
Environment sky more or less working.
2019-08-26 20:43:58 +00:00
return roughness_layers;
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
bool RendererSceneRenderRD::is_using_radiance_cubemap_array() const {
Environment sky more or less working.
2019-08-26 20:43:58 +00:00
return sky_use_cubemap_array;
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
RendererSceneRenderRD::RenderBufferData *RendererSceneRenderRD::render_buffers_get_data(RID p_render_buffers) {
-Refactored post processing, re-added glow and added a mix blend mode.
2020-01-10 00:40:26 +00:00
RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers);
Replace NULL with nullptr
2020-04-01 23:20:12 +00:00
ERR_FAIL_COND_V(!rb, nullptr);
-Refactored post processing, re-added glow and added a mix blend mode.
2020-01-10 00:40:26 +00:00
return rb->data;
}
Replace Octree by DynamicBVH in cull code -Much greater pairing/unpairing performance -For now, using it for culling too, but this will change in a couple of days. -Added a paged allocator, to efficiently alloc/free some types of objects.
2020-12-23 16:52:58 +00:00
void RendererSceneRenderRD::_setup_reflections(const PagedArray<RID> &p_reflections, const Transform &p_camera_inverse_transform, RID p_environment) {
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
cluster.reflection_count = 0;
Replace Octree by DynamicBVH in cull code -Much greater pairing/unpairing performance -For now, using it for culling too, but this will change in a couple of days. -Added a paged allocator, to efficiently alloc/free some types of objects.
2020-12-23 16:52:58 +00:00
for (uint32_t i = 0; i < (uint32_t)p_reflections.size(); i++) {
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
if (cluster.reflection_count == cluster.max_reflections) {
break;
}
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
ReflectionProbeInstance *rpi = reflection_probe_instance_owner.getornull(p_reflections[i]);
if (!rpi) {
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
continue;
}
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
cluster.reflection_sort[cluster.reflection_count].instance = rpi;
cluster.reflection_sort[cluster.reflection_count].depth = -p_camera_inverse_transform.xform(rpi->transform.origin).z;
cluster.reflection_count++;
}
if (cluster.reflection_count > 0) {
SortArray<Cluster::InstanceSort<ReflectionProbeInstance>> sort_array;
sort_array.sort(cluster.reflection_sort, cluster.reflection_count);
}
for (uint32_t i = 0; i < cluster.reflection_count; i++) {
ReflectionProbeInstance *rpi = cluster.reflection_sort[i].instance;
rpi->render_index = i;
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
RID base_probe = rpi->probe;
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
Cluster::ReflectionData &reflection_ubo = cluster.reflections[i];
Vector3 extents = storage->reflection_probe_get_extents(base_probe);
reflection_ubo.box_extents[0] = extents.x;
reflection_ubo.box_extents[1] = extents.y;
reflection_ubo.box_extents[2] = extents.z;
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
reflection_ubo.index = rpi->atlas_index;
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
Vector3 origin_offset = storage->reflection_probe_get_origin_offset(base_probe);
reflection_ubo.box_offset[0] = origin_offset.x;
reflection_ubo.box_offset[1] = origin_offset.y;
reflection_ubo.box_offset[2] = origin_offset.z;
reflection_ubo.mask = storage->reflection_probe_get_cull_mask(base_probe);
Shader optimizations to reduce VGPR usage and increase occupancy
2021-01-08 16:03:09 +00:00
reflection_ubo.intensity = storage->reflection_probe_get_intensity(base_probe);
reflection_ubo.ambient_mode = storage->reflection_probe_get_ambient_mode(base_probe);
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
Shader optimizations to reduce VGPR usage and increase occupancy
2021-01-08 16:03:09 +00:00
reflection_ubo.exterior = !storage->reflection_probe_is_interior(base_probe);
reflection_ubo.box_project = storage->reflection_probe_is_box_projection(base_probe);
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
Color ambient_linear = storage->reflection_probe_get_ambient_color(base_probe).to_linear();
float interior_ambient_energy = storage->reflection_probe_get_ambient_color_energy(base_probe);
reflection_ubo.ambient[0] = ambient_linear.r * interior_ambient_energy;
reflection_ubo.ambient[1] = ambient_linear.g * interior_ambient_energy;
reflection_ubo.ambient[2] = ambient_linear.b * interior_ambient_energy;
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
Transform transform = rpi->transform;
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
Transform proj = (p_camera_inverse_transform * transform).inverse();
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
RendererStorageRD::store_transform(proj, reflection_ubo.local_matrix);
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
current_cluster_builder->add_box(ClusterBuilderRD::BOX_TYPE_REFLECTION_PROBE, transform, extents);
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
rpi->last_pass = RSG::rasterizer->get_frame_number();
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
}
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
if (cluster.reflection_count) {
Reorganize RenderingDevice barriers -Removed sync to draw, now everything syncs to draw by default. -Fixed many validation layer errors. -Added support for VkImageViewUsageCreateInfo to fix validation layer warnings. -Texture, buffer, raster and compute functions now all allow spcifying which barriers will be used.
2021-01-26 00:52:58 +00:00
RD::get_singleton()->buffer_update(cluster.reflection_buffer, 0, cluster.reflection_count * sizeof(ReflectionData), cluster.reflections);
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
}
}
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
void RendererSceneRenderRD::_setup_lights(const PagedArray<RID> &p_lights, const Transform &p_camera_transform, RID p_shadow_atlas, bool p_using_shadows, uint32_t &r_directional_light_count, uint32_t &r_positional_light_count) {
Transform inverse_transform = p_camera_transform.affine_inverse();
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
r_directional_light_count = 0;
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
r_positional_light_count = 0;
Add fog to sky shaders
2020-08-18 04:12:51 +00:00
sky_scene_state.ubo.directional_light_count = 0;
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
Plane camera_plane(p_camera_transform.origin, -p_camera_transform.basis.get_axis(Vector3::AXIS_Z).normalized());
cluster.omni_light_count = 0;
cluster.spot_light_count = 0;
Replace Octree by DynamicBVH in cull code -Much greater pairing/unpairing performance -For now, using it for culling too, but this will change in a couple of days. -Added a paged allocator, to efficiently alloc/free some types of objects.
2020-12-23 16:52:58 +00:00
for (int i = 0; i < (int)p_lights.size(); i++) {
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
LightInstance *li = light_instance_owner.getornull(p_lights[i]);
if (!li) {
continue;
}
RID base = li->light;
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
ERR_CONTINUE(base.is_null());
RS::LightType type = storage->light_get_type(base);
switch (type) {
case RS::LIGHT_DIRECTIONAL: {
Add sky_only setting to DirectionalLight3Ds
2020-10-22 06:22:53 +00:00
// Copy to SkyDirectionalLightData
if (r_directional_light_count < sky_scene_state.max_directional_lights) {
SkyDirectionalLightData &sky_light_data = sky_scene_state.directional_lights[r_directional_light_count];
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
Transform light_transform = li->transform;
Add sky_only setting to DirectionalLight3Ds
2020-10-22 06:22:53 +00:00
Vector3 world_direction = light_transform.basis.xform(Vector3(0, 0, 1)).normalized();
sky_light_data.direction[0] = world_direction.x;
sky_light_data.direction[1] = world_direction.y;
sky_light_data.direction[2] = -world_direction.z;
float sign = storage->light_is_negative(base) ? -1 : 1;
sky_light_data.energy = sign * storage->light_get_param(base, RS::LIGHT_PARAM_ENERGY);
Color linear_col = storage->light_get_color(base).to_linear();
sky_light_data.color[0] = linear_col.r;
sky_light_data.color[1] = linear_col.g;
sky_light_data.color[2] = linear_col.b;
sky_light_data.enabled = true;
float angular_diameter = storage->light_get_param(base, RS::LIGHT_PARAM_SIZE);
if (angular_diameter > 0.0) {
// I know tan(0) is 0, but let's not risk it with numerical precision.
// technically this will keep expanding until reaching the sun, but all we care
// is expand until we reach the radius of the near plane (there can't be more occluders than that)
angular_diameter = Math::tan(Math::deg2rad(angular_diameter));
} else {
angular_diameter = 0.0;
}
sky_light_data.size = angular_diameter;
sky_scene_state.ubo.directional_light_count++;
}
if (r_directional_light_count >= cluster.max_directional_lights || storage->light_directional_is_sky_only(base)) {
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
continue;
}
Cluster::DirectionalLightData &light_data = cluster.directional_lights[r_directional_light_count];
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
Transform light_transform = li->transform;
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
Vector3 direction = inverse_transform.basis.xform(light_transform.basis.xform(Vector3(0, 0, 1))).normalized();
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
light_data.direction[0] = direction.x;
light_data.direction[1] = direction.y;
light_data.direction[2] = direction.z;
float sign = storage->light_is_negative(base) ? -1 : 1;
light_data.energy = sign * storage->light_get_param(base, RS::LIGHT_PARAM_ENERGY) * Math_PI;
Color linear_col = storage->light_get_color(base).to_linear();
light_data.color[0] = linear_col.r;
light_data.color[1] = linear_col.g;
light_data.color[2] = linear_col.b;
light_data.specular = storage->light_get_param(base, RS::LIGHT_PARAM_SPECULAR);
light_data.mask = storage->light_get_cull_mask(base);
float size = storage->light_get_param(base, RS::LIGHT_PARAM_SIZE);
light_data.size = 1.0 - Math::cos(Math::deg2rad(size)); //angle to cosine offset
Color shadow_col = storage->light_get_shadow_color(base).to_linear();
if (get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_PSSM_SPLITS) {
light_data.shadow_color1[0] = 1.0;
light_data.shadow_color1[1] = 0.0;
light_data.shadow_color1[2] = 0.0;
light_data.shadow_color1[3] = 1.0;
light_data.shadow_color2[0] = 0.0;
light_data.shadow_color2[1] = 1.0;
light_data.shadow_color2[2] = 0.0;
light_data.shadow_color2[3] = 1.0;
light_data.shadow_color3[0] = 0.0;
light_data.shadow_color3[1] = 0.0;
light_data.shadow_color3[2] = 1.0;
light_data.shadow_color3[3] = 1.0;
light_data.shadow_color4[0] = 1.0;
light_data.shadow_color4[1] = 1.0;
light_data.shadow_color4[2] = 0.0;
light_data.shadow_color4[3] = 1.0;
} else {
light_data.shadow_color1[0] = shadow_col.r;
light_data.shadow_color1[1] = shadow_col.g;
light_data.shadow_color1[2] = shadow_col.b;
light_data.shadow_color1[3] = 1.0;
light_data.shadow_color2[0] = shadow_col.r;
light_data.shadow_color2[1] = shadow_col.g;
light_data.shadow_color2[2] = shadow_col.b;
light_data.shadow_color2[3] = 1.0;
light_data.shadow_color3[0] = shadow_col.r;
light_data.shadow_color3[1] = shadow_col.g;
light_data.shadow_color3[2] = shadow_col.b;
light_data.shadow_color3[3] = 1.0;
light_data.shadow_color4[0] = shadow_col.r;
light_data.shadow_color4[1] = shadow_col.g;
light_data.shadow_color4[2] = shadow_col.b;
light_data.shadow_color4[3] = 1.0;
}
light_data.shadow_enabled = p_using_shadows && storage->light_has_shadow(base);
float angular_diameter = storage->light_get_param(base, RS::LIGHT_PARAM_SIZE);
if (angular_diameter > 0.0) {
// I know tan(0) is 0, but let's not risk it with numerical precision.
// technically this will keep expanding until reaching the sun, but all we care
// is expand until we reach the radius of the near plane (there can't be more occluders than that)
angular_diameter = Math::tan(Math::deg2rad(angular_diameter));
} else {
angular_diameter = 0.0;
}
if (light_data.shadow_enabled) {
RS::LightDirectionalShadowMode smode = storage->light_directional_get_shadow_mode(base);
int limit = smode == RS::LIGHT_DIRECTIONAL_SHADOW_ORTHOGONAL ? 0 : (smode == RS::LIGHT_DIRECTIONAL_SHADOW_PARALLEL_2_SPLITS ? 1 : 3);
light_data.blend_splits = storage->light_directional_get_blend_splits(base);
for (int j = 0; j < 4; j++) {
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
Rect2 atlas_rect = li->shadow_transform[j].atlas_rect;
CameraMatrix matrix = li->shadow_transform[j].camera;
float split = li->shadow_transform[MIN(limit, j)].split;
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
CameraMatrix bias;
bias.set_light_bias();
CameraMatrix rectm;
rectm.set_light_atlas_rect(atlas_rect);
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
Transform modelview = (inverse_transform * li->shadow_transform[j].transform).inverse();
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
CameraMatrix shadow_mtx = rectm * bias * matrix * modelview;
light_data.shadow_split_offsets[j] = split;
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
float bias_scale = li->shadow_transform[j].bias_scale;
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
light_data.shadow_bias[j] = storage->light_get_param(base, RS::LIGHT_PARAM_SHADOW_BIAS) * bias_scale;
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
light_data.shadow_normal_bias[j] = storage->light_get_param(base, RS::LIGHT_PARAM_SHADOW_NORMAL_BIAS) * li->shadow_transform[j].shadow_texel_size;
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
light_data.shadow_transmittance_bias[j] = storage->light_get_transmittance_bias(base) * bias_scale;
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
light_data.shadow_z_range[j] = li->shadow_transform[j].farplane;
light_data.shadow_range_begin[j] = li->shadow_transform[j].range_begin;
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
RendererStorageRD::store_camera(shadow_mtx, light_data.shadow_matrices[j]);
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
Vector2 uv_scale = li->shadow_transform[j].uv_scale;
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
uv_scale *= atlas_rect.size; //adapt to atlas size
switch (j) {
case 0: {
light_data.uv_scale1[0] = uv_scale.x;
light_data.uv_scale1[1] = uv_scale.y;
} break;
case 1: {
light_data.uv_scale2[0] = uv_scale.x;
light_data.uv_scale2[1] = uv_scale.y;
} break;
case 2: {
light_data.uv_scale3[0] = uv_scale.x;
light_data.uv_scale3[1] = uv_scale.y;
} break;
case 3: {
light_data.uv_scale4[0] = uv_scale.x;
light_data.uv_scale4[1] = uv_scale.y;
} break;
}
}
float fade_start = storage->light_get_param(base, RS::LIGHT_PARAM_SHADOW_FADE_START);
light_data.fade_from = -light_data.shadow_split_offsets[3] * MIN(fade_start, 0.999); //using 1.0 would break smoothstep
light_data.fade_to = -light_data.shadow_split_offsets[3];
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
light_data.shadow_volumetric_fog_fade = 1.0 / storage->light_get_shadow_volumetric_fog_fade(base);
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
light_data.soft_shadow_scale = storage->light_get_param(base, RS::LIGHT_PARAM_SHADOW_BLUR);
light_data.softshadow_angle = angular_diameter;
if (angular_diameter <= 0.0) {
light_data.soft_shadow_scale *= directional_shadow_quality_radius_get(); // Only use quality radius for PCF
}
}
r_directional_light_count++;
} break;
case RS::LIGHT_OMNI: {
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
if (cluster.omni_light_count >= cluster.max_lights) {
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
continue;
}
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
cluster.omni_light_sort[cluster.omni_light_count].instance = li;
cluster.omni_light_sort[cluster.omni_light_count].depth = camera_plane.distance_to(li->transform.origin);
cluster.omni_light_count++;
} break;
case RS::LIGHT_SPOT: {
if (cluster.spot_light_count >= cluster.max_lights) {
continue;
}
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
cluster.spot_light_sort[cluster.spot_light_count].instance = li;
cluster.spot_light_sort[cluster.spot_light_count].depth = camera_plane.distance_to(li->transform.origin);
cluster.spot_light_count++;
} break;
}
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
li->last_pass = RSG::rasterizer->get_frame_number();
}
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
if (cluster.omni_light_count) {
SortArray<Cluster::InstanceSort<LightInstance>> sorter;
sorter.sort(cluster.omni_light_sort, cluster.omni_light_count);
}
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
if (cluster.spot_light_count) {
SortArray<Cluster::InstanceSort<LightInstance>> sorter;
sorter.sort(cluster.spot_light_sort, cluster.spot_light_count);
}
Shader optimizations to reduce VGPR usage and increase occupancy
2021-01-08 16:03:09 +00:00
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
ShadowAtlas *shadow_atlas = nullptr;
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
if (p_shadow_atlas.is_valid() && p_using_shadows) {
shadow_atlas = shadow_atlas_owner.getornull(p_shadow_atlas);
}
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
for (uint32_t i = 0; i < (cluster.omni_light_count + cluster.spot_light_count); i++) {
uint32_t index = (i < cluster.omni_light_count) ? i : i - (cluster.omni_light_count);
Cluster::LightData &light_data = (i < cluster.omni_light_count) ? cluster.omni_lights[index] : cluster.spot_lights[index];
RS::LightType type = (i < cluster.omni_light_count) ? RS::LIGHT_OMNI : RS::LIGHT_SPOT;
LightInstance *li = (i < cluster.omni_light_count) ? cluster.omni_light_sort[index].instance : cluster.spot_light_sort[index].instance;
RID base = li->light;
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
cluster.lights_instances[i] = li->self;
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
Transform light_transform = li->transform;
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
float sign = storage->light_is_negative(base) ? -1 : 1;
Color linear_col = storage->light_get_color(base).to_linear();
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
light_data.attenuation = storage->light_get_param(base, RS::LIGHT_PARAM_ATTENUATION);
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
float energy = sign * storage->light_get_param(base, RS::LIGHT_PARAM_ENERGY) * Math_PI;
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
light_data.color[0] = linear_col.r * energy;
light_data.color[1] = linear_col.g * energy;
light_data.color[2] = linear_col.b * energy;
light_data.specular_amount = storage->light_get_param(base, RS::LIGHT_PARAM_SPECULAR) * 2.0;
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
float radius = MAX(0.001, storage->light_get_param(base, RS::LIGHT_PARAM_RANGE));
light_data.inv_radius = 1.0 / radius;
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
Vector3 pos = inverse_transform.xform(light_transform.origin);
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
light_data.position[0] = pos.x;
light_data.position[1] = pos.y;
light_data.position[2] = pos.z;
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
Vector3 direction = inverse_transform.basis.xform(light_transform.basis.xform(Vector3(0, 0, -1))).normalized();
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
light_data.direction[0] = direction.x;
light_data.direction[1] = direction.y;
light_data.direction[2] = direction.z;
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
float size = storage->light_get_param(base, RS::LIGHT_PARAM_SIZE);
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
light_data.size = size;
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
light_data.cone_attenuation = storage->light_get_param(base, RS::LIGHT_PARAM_SPOT_ATTENUATION);
float spot_angle = storage->light_get_param(base, RS::LIGHT_PARAM_SPOT_ANGLE);
light_data.cone_angle = Math::cos(Math::deg2rad(spot_angle));
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
light_data.mask = storage->light_get_cull_mask(base);
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
light_data.atlas_rect[0] = 0;
light_data.atlas_rect[1] = 0;
light_data.atlas_rect[2] = 0;
light_data.atlas_rect[3] = 0;
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
RID projector = storage->light_get_projector(base);
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
if (projector.is_valid()) {
Rect2 rect = storage->decal_atlas_get_texture_rect(projector);
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
if (type == RS::LIGHT_SPOT) {
light_data.projector_rect[0] = rect.position.x;
light_data.projector_rect[1] = rect.position.y + rect.size.height; //flip because shadow is flipped
light_data.projector_rect[2] = rect.size.width;
light_data.projector_rect[3] = -rect.size.height;
} else {
light_data.projector_rect[0] = rect.position.x;
light_data.projector_rect[1] = rect.position.y;
light_data.projector_rect[2] = rect.size.width;
light_data.projector_rect[3] = rect.size.height * 0.5; //used by dp, so needs to be half
}
} else {
light_data.projector_rect[0] = 0;
light_data.projector_rect[1] = 0;
light_data.projector_rect[2] = 0;
light_data.projector_rect[3] = 0;
}
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
if (shadow_atlas && shadow_atlas->shadow_owners.has(li->self)) {
// fill in the shadow information
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
light_data.shadow_enabled = true;
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
if (type == RS::LIGHT_SPOT) {
light_data.shadow_bias = (storage->light_get_param(base, RS::LIGHT_PARAM_SHADOW_BIAS) * radius / 10.0);
float shadow_texel_size = Math::tan(Math::deg2rad(spot_angle)) * radius * 2.0;
shadow_texel_size *= light_instance_get_shadow_texel_size(li->self, p_shadow_atlas);
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
light_data.shadow_normal_bias = storage->light_get_param(base, RS::LIGHT_PARAM_SHADOW_NORMAL_BIAS) * shadow_texel_size;
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
} else { //omni
light_data.shadow_bias = storage->light_get_param(base, RS::LIGHT_PARAM_SHADOW_BIAS) * radius / 10.0;
float shadow_texel_size = light_instance_get_shadow_texel_size(li->self, p_shadow_atlas);
light_data.shadow_normal_bias = storage->light_get_param(base, RS::LIGHT_PARAM_SHADOW_NORMAL_BIAS) * shadow_texel_size * 2.0; // applied in -1 .. 1 space
}
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
light_data.transmittance_bias = storage->light_get_transmittance_bias(base);
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
Rect2 rect = light_instance_get_shadow_atlas_rect(li->self, p_shadow_atlas);
light_data.atlas_rect[0] = rect.position.x;
light_data.atlas_rect[1] = rect.position.y;
light_data.atlas_rect[2] = rect.size.width;
light_data.atlas_rect[3] = rect.size.height;
light_data.soft_shadow_scale = storage->light_get_param(base, RS::LIGHT_PARAM_SHADOW_BLUR);
light_data.shadow_volumetric_fog_fade = 1.0 / storage->light_get_shadow_volumetric_fog_fade(base);
if (type == RS::LIGHT_OMNI) {
light_data.atlas_rect[3] *= 0.5; //one paraboloid on top of another
Transform proj = (inverse_transform * light_transform).inverse();
RendererStorageRD::store_transform(proj, light_data.shadow_matrix);
if (size > 0.0) {
light_data.soft_shadow_size = size;
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
} else {
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
light_data.soft_shadow_size = 0.0;
light_data.soft_shadow_scale *= shadows_quality_radius_get(); // Only use quality radius for PCF
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
}
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
} else if (type == RS::LIGHT_SPOT) {
Transform modelview = (inverse_transform * light_transform).inverse();
CameraMatrix bias;
bias.set_light_bias();
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
CameraMatrix shadow_mtx = bias * li->shadow_transform[0].camera * modelview;
RendererStorageRD::store_camera(shadow_mtx, light_data.shadow_matrix);
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
if (size > 0.0) {
CameraMatrix cm = li->shadow_transform[0].camera;
float half_np = cm.get_z_near() * Math::tan(Math::deg2rad(spot_angle));
light_data.soft_shadow_size = (size * 0.5 / radius) / (half_np / cm.get_z_near()) * rect.size.width;
} else {
light_data.soft_shadow_size = 0.0;
light_data.soft_shadow_scale *= shadows_quality_radius_get(); // Only use quality radius for PCF
}
}
} else {
light_data.shadow_enabled = false;
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
}
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
li->light_index = index;
current_cluster_builder->add_light(type == RS::LIGHT_SPOT ? ClusterBuilderRD::LIGHT_TYPE_SPOT : ClusterBuilderRD::LIGHT_TYPE_OMNI, light_transform, radius, spot_angle);
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
r_positional_light_count++;
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
}
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
if (cluster.omni_light_count) {
Reorganize RenderingDevice barriers -Removed sync to draw, now everything syncs to draw by default. -Fixed many validation layer errors. -Added support for VkImageViewUsageCreateInfo to fix validation layer warnings. -Texture, buffer, raster and compute functions now all allow spcifying which barriers will be used.
2021-01-26 00:52:58 +00:00
RD::get_singleton()->buffer_update(cluster.omni_light_buffer, 0, sizeof(Cluster::LightData) * cluster.omni_light_count, cluster.omni_lights);
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
}
if (cluster.spot_light_count) {
Reorganize RenderingDevice barriers -Removed sync to draw, now everything syncs to draw by default. -Fixed many validation layer errors. -Added support for VkImageViewUsageCreateInfo to fix validation layer warnings. -Texture, buffer, raster and compute functions now all allow spcifying which barriers will be used.
2021-01-26 00:52:58 +00:00
RD::get_singleton()->buffer_update(cluster.spot_light_buffer, 0, sizeof(Cluster::LightData) * cluster.spot_light_count, cluster.spot_lights);
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
}
if (r_directional_light_count) {
Reorganize RenderingDevice barriers -Removed sync to draw, now everything syncs to draw by default. -Fixed many validation layer errors. -Added support for VkImageViewUsageCreateInfo to fix validation layer warnings. -Texture, buffer, raster and compute functions now all allow spcifying which barriers will be used.
2021-01-26 00:52:58 +00:00
RD::get_singleton()->buffer_update(cluster.directional_light_buffer, 0, sizeof(Cluster::DirectionalLightData) * r_directional_light_count, cluster.directional_lights);
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
}
}
Replace Octree by DynamicBVH in cull code -Much greater pairing/unpairing performance -For now, using it for culling too, but this will change in a couple of days. -Added a paged allocator, to efficiently alloc/free some types of objects.
2020-12-23 16:52:58 +00:00
void RendererSceneRenderRD::_setup_decals(const PagedArray<RID> &p_decals, const Transform &p_camera_inverse_xform) {
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
Transform uv_xform;
uv_xform.basis.scale(Vector3(2.0, 1.0, 2.0));
uv_xform.origin = Vector3(-1.0, 0.0, -1.0);
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
uint32_t decal_count = p_decals.size();
cluster.decal_count = 0;
Replace Octree by DynamicBVH in cull code -Much greater pairing/unpairing performance -For now, using it for culling too, but this will change in a couple of days. -Added a paged allocator, to efficiently alloc/free some types of objects.
2020-12-23 16:52:58 +00:00
for (uint32_t i = 0; i < decal_count; i++) {
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
if (cluster.decal_count == cluster.max_decals) {
break;
}
DecalInstance *di = decal_instance_owner.getornull(p_decals[i]);
if (!di) {
continue;
}
RID decal = di->decal;
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
Transform xform = di->transform;
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
real_t distance = -p_camera_inverse_xform.xform(xform.origin).z;
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
if (storage->decal_is_distance_fade_enabled(decal)) {
float fade_begin = storage->decal_get_distance_fade_begin(decal);
float fade_length = storage->decal_get_distance_fade_length(decal);
if (distance > fade_begin) {
if (distance > fade_begin + fade_length) {
continue; // do not use this decal, its invisible
}
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
}
}
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
cluster.decal_sort[cluster.decal_count].instance = di;
cluster.decal_sort[cluster.decal_count].depth = distance;
cluster.decal_count++;
}
if (cluster.decal_count > 0) {
SortArray<Cluster::InstanceSort<DecalInstance>> sort_array;
sort_array.sort(cluster.decal_sort, cluster.decal_count);
}
for (uint32_t i = 0; i < cluster.decal_count; i++) {
DecalInstance *di = cluster.decal_sort[i].instance;
RID decal = di->decal;
Transform xform = di->transform;
float fade = 1.0;
if (storage->decal_is_distance_fade_enabled(decal)) {
real_t distance = -p_camera_inverse_xform.xform(xform.origin).z;
float fade_begin = storage->decal_get_distance_fade_begin(decal);
float fade_length = storage->decal_get_distance_fade_length(decal);
if (distance > fade_begin) {
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
fade = 1.0 - (distance - fade_begin) / fade_length;
}
}
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
Cluster::DecalData &dd = cluster.decals[i];
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
Vector3 decal_extents = storage->decal_get_extents(decal);
Transform scale_xform;
scale_xform.basis.scale(Vector3(decal_extents.x, decal_extents.y, decal_extents.z));
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
Transform to_decal_xform = (p_camera_inverse_xform * di->transform * scale_xform * uv_xform).affine_inverse();
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
RendererStorageRD::store_transform(to_decal_xform, dd.xform);
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
Vector3 normal = xform.basis.get_axis(Vector3::AXIS_Y).normalized();
normal = p_camera_inverse_xform.basis.xform(normal); //camera is normalized, so fine
dd.normal[0] = normal.x;
dd.normal[1] = normal.y;
dd.normal[2] = normal.z;
dd.normal_fade = storage->decal_get_normal_fade(decal);
RID albedo_tex = storage->decal_get_texture(decal, RS::DECAL_TEXTURE_ALBEDO);
RID emission_tex = storage->decal_get_texture(decal, RS::DECAL_TEXTURE_EMISSION);
if (albedo_tex.is_valid()) {
Rect2 rect = storage->decal_atlas_get_texture_rect(albedo_tex);
dd.albedo_rect[0] = rect.position.x;
dd.albedo_rect[1] = rect.position.y;
dd.albedo_rect[2] = rect.size.x;
dd.albedo_rect[3] = rect.size.y;
} else {
if (!emission_tex.is_valid()) {
continue; //no albedo, no emission, no decal.
}
dd.albedo_rect[0] = 0;
dd.albedo_rect[1] = 0;
dd.albedo_rect[2] = 0;
dd.albedo_rect[3] = 0;
}
RID normal_tex = storage->decal_get_texture(decal, RS::DECAL_TEXTURE_NORMAL);
if (normal_tex.is_valid()) {
Rect2 rect = storage->decal_atlas_get_texture_rect(normal_tex);
dd.normal_rect[0] = rect.position.x;
dd.normal_rect[1] = rect.position.y;
dd.normal_rect[2] = rect.size.x;
dd.normal_rect[3] = rect.size.y;
Basis normal_xform = p_camera_inverse_xform.basis * xform.basis.orthonormalized();
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
RendererStorageRD::store_basis_3x4(normal_xform, dd.normal_xform);
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
} else {
dd.normal_rect[0] = 0;
dd.normal_rect[1] = 0;
dd.normal_rect[2] = 0;
dd.normal_rect[3] = 0;
}
RID orm_tex = storage->decal_get_texture(decal, RS::DECAL_TEXTURE_ORM);
if (orm_tex.is_valid()) {
Rect2 rect = storage->decal_atlas_get_texture_rect(orm_tex);
dd.orm_rect[0] = rect.position.x;
dd.orm_rect[1] = rect.position.y;
dd.orm_rect[2] = rect.size.x;
dd.orm_rect[3] = rect.size.y;
} else {
dd.orm_rect[0] = 0;
dd.orm_rect[1] = 0;
dd.orm_rect[2] = 0;
dd.orm_rect[3] = 0;
}
if (emission_tex.is_valid()) {
Rect2 rect = storage->decal_atlas_get_texture_rect(emission_tex);
dd.emission_rect[0] = rect.position.x;
dd.emission_rect[1] = rect.position.y;
dd.emission_rect[2] = rect.size.x;
dd.emission_rect[3] = rect.size.y;
} else {
dd.emission_rect[0] = 0;
dd.emission_rect[1] = 0;
dd.emission_rect[2] = 0;
dd.emission_rect[3] = 0;
}
Color modulate = storage->decal_get_modulate(decal);
dd.modulate[0] = modulate.r;
dd.modulate[1] = modulate.g;
dd.modulate[2] = modulate.b;
dd.modulate[3] = modulate.a * fade;
dd.emission_energy = storage->decal_get_emission_energy(decal) * fade;
dd.albedo_mix = storage->decal_get_albedo_mix(decal);
dd.mask = storage->decal_get_cull_mask(decal);
dd.upper_fade = storage->decal_get_upper_fade(decal);
dd.lower_fade = storage->decal_get_lower_fade(decal);
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
current_cluster_builder->add_box(ClusterBuilderRD::BOX_TYPE_DECAL, xform, decal_extents);
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
}
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
if (cluster.decal_count > 0) {
Reorganize RenderingDevice barriers -Removed sync to draw, now everything syncs to draw by default. -Fixed many validation layer errors. -Added support for VkImageViewUsageCreateInfo to fix validation layer warnings. -Texture, buffer, raster and compute functions now all allow spcifying which barriers will be used.
2021-01-26 00:52:58 +00:00
RD::get_singleton()->buffer_update(cluster.decal_buffer, 0, sizeof(Cluster::DecalData) * cluster.decal_count, cluster.decals);
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
}
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
void RendererSceneRenderRD::_volumetric_fog_erase(RenderBuffers *rb) {
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
ERR_FAIL_COND(!rb->volumetric_fog);
RD::get_singleton()->free(rb->volumetric_fog->light_density_map);
RD::get_singleton()->free(rb->volumetric_fog->fog_map);
Add fog to sky shaders
2020-08-18 04:12:51 +00:00
if (rb->volumetric_fog->uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(rb->volumetric_fog->uniform_set)) {
RD::get_singleton()->free(rb->volumetric_fog->uniform_set);
}
if (rb->volumetric_fog->uniform_set2.is_valid() && RD::get_singleton()->uniform_set_is_valid(rb->volumetric_fog->uniform_set2)) {
RD::get_singleton()->free(rb->volumetric_fog->uniform_set2);
}
if (rb->volumetric_fog->sdfgi_uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(rb->volumetric_fog->sdfgi_uniform_set)) {
RD::get_singleton()->free(rb->volumetric_fog->sdfgi_uniform_set);
}
if (rb->volumetric_fog->sky_uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(rb->volumetric_fog->sky_uniform_set)) {
RD::get_singleton()->free(rb->volumetric_fog->sky_uniform_set);
}
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
memdelete(rb->volumetric_fog);
Add fog to sky shaders
2020-08-18 04:12:51 +00:00
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
rb->volumetric_fog = nullptr;
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
void RendererSceneRenderRD::_allocate_shadow_shrink_stages(RID p_base, int p_base_size, Vector<ShadowShrinkStage> &shrink_stages, uint32_t p_target_size) {
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
//create fog mipmaps
uint32_t fog_texture_size = p_target_size;
uint32_t base_texture_size = p_base_size;
ShadowShrinkStage first;
first.size = base_texture_size;
first.texture = p_base;
shrink_stages.push_back(first); //put depth first in case we dont find smaller ones
while (fog_texture_size < base_texture_size) {
base_texture_size = MAX(base_texture_size / 8, fog_texture_size);
ShadowShrinkStage s;
s.size = base_texture_size;
RD::TextureFormat tf;
tf.format = RD::DATA_FORMAT_R32_SFLOAT;
tf.width = base_texture_size;
tf.height = base_texture_size;
tf.usage_bits = RD::TEXTURE_USAGE_STORAGE_BIT;
if (base_texture_size == fog_texture_size) {
s.filter_texture = RD::get_singleton()->texture_create(tf, RD::TextureView());
tf.usage_bits |= RD::TEXTURE_USAGE_SAMPLING_BIT;
}
s.texture = RD::get_singleton()->texture_create(tf, RD::TextureView());
shrink_stages.push_back(s);
}
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
void RendererSceneRenderRD::_clear_shadow_shrink_stages(Vector<ShadowShrinkStage> &shrink_stages) {
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
for (int i = 1; i < shrink_stages.size(); i++) {
RD::get_singleton()->free(shrink_stages[i].texture);
if (shrink_stages[i].filter_texture.is_valid()) {
RD::get_singleton()->free(shrink_stages[i].filter_texture);
}
}
shrink_stages.clear();
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
void RendererSceneRenderRD::_update_volumetric_fog(RID p_render_buffers, RID p_environment, const CameraMatrix &p_cam_projection, const Transform &p_cam_transform, RID p_shadow_atlas, int p_directional_light_count, bool p_use_directional_shadows, int p_positional_light_count, int p_gi_probe_count) {
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers);
ERR_FAIL_COND(!rb);
Environment *env = environment_owner.getornull(p_environment);
float ratio = float(rb->width) / float((rb->width + rb->height) / 2);
uint32_t target_width = uint32_t(float(volumetric_fog_size) * ratio);
uint32_t target_height = uint32_t(float(volumetric_fog_size) / ratio);
if (rb->volumetric_fog) {
//validate
if (!env || !env->volumetric_fog_enabled || rb->volumetric_fog->width != target_width || rb->volumetric_fog->height != target_height || rb->volumetric_fog->depth != volumetric_fog_depth) {
_volumetric_fog_erase(rb);
_render_buffers_uniform_set_changed(p_render_buffers);
}
}
if (!env || !env->volumetric_fog_enabled) {
//no reason to enable or update, bye
return;
}
if (env && env->volumetric_fog_enabled && !rb->volumetric_fog) {
//required volumetric fog but not existing, create
rb->volumetric_fog = memnew(VolumetricFog);
rb->volumetric_fog->width = target_width;
rb->volumetric_fog->height = target_height;
rb->volumetric_fog->depth = volumetric_fog_depth;
RD::TextureFormat tf;
tf.format = RD::DATA_FORMAT_R16G16B16A16_SFLOAT;
tf.width = target_width;
tf.height = target_height;
tf.depth = volumetric_fog_depth;
Rename RD texture "type" to "texture_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-11-27 03:50:05 +00:00
tf.texture_type = RD::TEXTURE_TYPE_3D;
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
tf.usage_bits = RD::TEXTURE_USAGE_STORAGE_BIT;
rb->volumetric_fog->light_density_map = RD::get_singleton()->texture_create(tf, RD::TextureView());
tf.usage_bits |= RD::TEXTURE_USAGE_SAMPLING_BIT;
rb->volumetric_fog->fog_map = RD::get_singleton()->texture_create(tf, RD::TextureView());
_render_buffers_uniform_set_changed(p_render_buffers);
Add fog to sky shaders
2020-08-18 04:12:51 +00:00
Vector<RD::Uniform> uniforms;
{
RD::Uniform u;
u.binding = 0;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_TEXTURE;
Add fog to sky shaders
2020-08-18 04:12:51 +00:00
u.ids.push_back(rb->volumetric_fog->fog_map);
uniforms.push_back(u);
}
rb->volumetric_fog->sky_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, sky_shader.default_shader_rd, SKY_SET_FOG);
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
}
//update directional shadow
if (p_use_directional_shadows) {
Rename empty() to is_empty()
2020-12-15 12:04:21 +00:00
if (directional_shadow.shrink_stages.is_empty()) {
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
if (rb->volumetric_fog->uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(rb->volumetric_fog->uniform_set)) {
//invalidate uniform set, we will need a new one
RD::get_singleton()->free(rb->volumetric_fog->uniform_set);
rb->volumetric_fog->uniform_set = RID();
}
_allocate_shadow_shrink_stages(directional_shadow.depth, directional_shadow.size, directional_shadow.shrink_stages, volumetric_fog_directional_shadow_shrink);
}
if (directional_shadow.shrink_stages.size() > 1) {
RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin();
for (int i = 1; i < directional_shadow.shrink_stages.size(); i++) {
int32_t src_size = directional_shadow.shrink_stages[i - 1].size;
int32_t dst_size = directional_shadow.shrink_stages[i].size;
Rect2i r(0, 0, src_size, src_size);
int32_t shrink_limit = 8 / (src_size / dst_size);
storage->get_effects()->reduce_shadow(directional_shadow.shrink_stages[i - 1].texture, directional_shadow.shrink_stages[i].texture, Size2i(src_size, src_size), r, shrink_limit, compute_list);
RD::get_singleton()->compute_list_add_barrier(compute_list);
if (env->volumetric_fog_shadow_filter != RS::ENV_VOLUMETRIC_FOG_SHADOW_FILTER_DISABLED && directional_shadow.shrink_stages[i].filter_texture.is_valid()) {
Rect2i rf(0, 0, dst_size, dst_size);
storage->get_effects()->filter_shadow(directional_shadow.shrink_stages[i].texture, directional_shadow.shrink_stages[i].filter_texture, Size2i(dst_size, dst_size), rf, env->volumetric_fog_shadow_filter, compute_list);
}
}
RD::get_singleton()->compute_list_end();
}
}
ShadowAtlas *shadow_atlas = shadow_atlas_owner.getornull(p_shadow_atlas);
if (shadow_atlas) {
//shrink shadows that need to be shrunk
bool force_shrink_shadows = false;
Rename empty() to is_empty()
2020-12-15 12:04:21 +00:00
if (shadow_atlas->shrink_stages.is_empty()) {
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
if (rb->volumetric_fog->uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(rb->volumetric_fog->uniform_set)) {
//invalidate uniform set, we will need a new one
RD::get_singleton()->free(rb->volumetric_fog->uniform_set);
rb->volumetric_fog->uniform_set = RID();
}
_allocate_shadow_shrink_stages(shadow_atlas->depth, shadow_atlas->size, shadow_atlas->shrink_stages, volumetric_fog_positional_shadow_shrink);
force_shrink_shadows = true;
}
if (rb->volumetric_fog->last_shadow_filter != env->volumetric_fog_shadow_filter) {
//if shadow filter changed, invalidate caches
rb->volumetric_fog->last_shadow_filter = env->volumetric_fog_shadow_filter;
force_shrink_shadows = true;
}
cluster.lights_shadow_rect_cache_count = 0;
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
for (uint32_t i = 0; i < cluster.omni_light_count + cluster.spot_light_count; i++) {
Cluster::LightData &ld = i < cluster.omni_light_count ? cluster.omni_lights[i] : cluster.spot_lights[i - cluster.omni_light_count];
if (ld.shadow_enabled != 0) {
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
RID li = cluster.lights_instances[i];
ERR_CONTINUE(!shadow_atlas->shadow_owners.has(li));
uint32_t key = shadow_atlas->shadow_owners[li];
uint32_t quadrant = (key >> ShadowAtlas::QUADRANT_SHIFT) & 0x3;
uint32_t shadow = key & ShadowAtlas::SHADOW_INDEX_MASK;
ERR_CONTINUE((int)shadow >= shadow_atlas->quadrants[quadrant].shadows.size());
ShadowAtlas::Quadrant::Shadow &s = shadow_atlas->quadrants[quadrant].shadows.write[shadow];
if (!force_shrink_shadows && s.fog_version == s.version) {
continue; //do not update, no need
}
s.fog_version = s.version;
uint32_t quadrant_size = shadow_atlas->size >> 1;
Rect2i atlas_rect;
atlas_rect.position.x = (quadrant & 1) * quadrant_size;
atlas_rect.position.y = (quadrant >> 1) * quadrant_size;
uint32_t shadow_size = (quadrant_size / shadow_atlas->quadrants[quadrant].subdivision);
atlas_rect.position.x += (shadow % shadow_atlas->quadrants[quadrant].subdivision) * shadow_size;
atlas_rect.position.y += (shadow / shadow_atlas->quadrants[quadrant].subdivision) * shadow_size;
atlas_rect.size.x = shadow_size;
atlas_rect.size.y = shadow_size;
cluster.lights_shadow_rect_cache[cluster.lights_shadow_rect_cache_count] = atlas_rect;
cluster.lights_shadow_rect_cache_count++;
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
if (cluster.lights_shadow_rect_cache_count == cluster.max_lights * 2) {
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
break; //light limit reached
}
}
}
if (cluster.lights_shadow_rect_cache_count > 0) {
//there are shadows to be shrunk, try to do them in parallel
RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin();
for (int i = 1; i < shadow_atlas->shrink_stages.size(); i++) {
int32_t base_size = shadow_atlas->shrink_stages[0].size;
int32_t src_size = shadow_atlas->shrink_stages[i - 1].size;
int32_t dst_size = shadow_atlas->shrink_stages[i].size;
uint32_t rect_divisor = base_size / src_size;
int32_t shrink_limit = 8 / (src_size / dst_size);
//shrink in parallel for more performance
for (uint32_t j = 0; j < cluster.lights_shadow_rect_cache_count; j++) {
Rect2i src_rect = cluster.lights_shadow_rect_cache[j];
src_rect.position /= rect_divisor;
src_rect.size /= rect_divisor;
storage->get_effects()->reduce_shadow(shadow_atlas->shrink_stages[i - 1].texture, shadow_atlas->shrink_stages[i].texture, Size2i(src_size, src_size), src_rect, shrink_limit, compute_list);
}
RD::get_singleton()->compute_list_add_barrier(compute_list);
if (env->volumetric_fog_shadow_filter != RS::ENV_VOLUMETRIC_FOG_SHADOW_FILTER_DISABLED && shadow_atlas->shrink_stages[i].filter_texture.is_valid()) {
uint32_t filter_divisor = base_size / dst_size;
//filter in parallel for more performance
for (uint32_t j = 0; j < cluster.lights_shadow_rect_cache_count; j++) {
Rect2i dst_rect = cluster.lights_shadow_rect_cache[j];
dst_rect.position /= filter_divisor;
dst_rect.size /= filter_divisor;
storage->get_effects()->filter_shadow(shadow_atlas->shrink_stages[i].texture, shadow_atlas->shrink_stages[i].filter_texture, Size2i(dst_size, dst_size), dst_rect, env->volumetric_fog_shadow_filter, compute_list, true, false);
}
RD::get_singleton()->compute_list_add_barrier(compute_list);
for (uint32_t j = 0; j < cluster.lights_shadow_rect_cache_count; j++) {
Rect2i dst_rect = cluster.lights_shadow_rect_cache[j];
dst_rect.position /= filter_divisor;
dst_rect.size /= filter_divisor;
storage->get_effects()->filter_shadow(shadow_atlas->shrink_stages[i].texture, shadow_atlas->shrink_stages[i].filter_texture, Size2i(dst_size, dst_size), dst_rect, env->volumetric_fog_shadow_filter, compute_list, false, true);
}
}
}
RD::get_singleton()->compute_list_end();
}
}
//update volumetric fog
if (rb->volumetric_fog->uniform_set.is_null() || !RD::get_singleton()->uniform_set_is_valid(rb->volumetric_fog->uniform_set)) {
//re create uniform set if needed
Vector<RD::Uniform> uniforms;
{
RD::Uniform u;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_TEXTURE;
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
u.binding = 1;
if (shadow_atlas == nullptr || shadow_atlas->shrink_stages.size() == 0) {
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
u.ids.push_back(storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_BLACK));
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
} else {
u.ids.push_back(shadow_atlas->shrink_stages[shadow_atlas->shrink_stages.size() - 1].texture);
}
uniforms.push_back(u);
}
{
RD::Uniform u;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_TEXTURE;
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
u.binding = 2;
if (directional_shadow.shrink_stages.size() == 0) {
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
u.ids.push_back(storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_BLACK));
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
} else {
u.ids.push_back(directional_shadow.shrink_stages[directional_shadow.shrink_stages.size() - 1].texture);
}
uniforms.push_back(u);
}
{
RD::Uniform u;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER;
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
u.binding = 3;
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
u.ids.push_back(get_omni_light_buffer());
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
uniforms.push_back(u);
}
{
RD::Uniform u;
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER;
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
u.binding = 4;
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
u.ids.push_back(get_spot_light_buffer());
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
uniforms.push_back(u);
}
{
RD::Uniform u;
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER;
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
u.binding = 5;
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
u.ids.push_back(get_directional_light_buffer());
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
uniforms.push_back(u);
}
{
RD::Uniform u;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER;
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
u.binding = 6;
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
u.ids.push_back(rb->cluster_builder->get_cluster_buffer());
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
uniforms.push_back(u);
}
{
RD::Uniform u;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_SAMPLER;
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
u.binding = 7;
u.ids.push_back(storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED));
uniforms.push_back(u);
}
{
RD::Uniform u;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_IMAGE;
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
u.binding = 8;
u.ids.push_back(rb->volumetric_fog->light_density_map);
uniforms.push_back(u);
}
{
RD::Uniform u;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_IMAGE;
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
u.binding = 9;
u.ids.push_back(rb->volumetric_fog->fog_map);
uniforms.push_back(u);
}
{
RD::Uniform u;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_SAMPLER;
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
u.binding = 10;
u.ids.push_back(shadow_sampler);
uniforms.push_back(u);
}
{
RD::Uniform u;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER;
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
u.binding = 11;
u.ids.push_back(render_buffers_get_gi_probe_buffer(p_render_buffers));
uniforms.push_back(u);
}
{
RD::Uniform u;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_TEXTURE;
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
u.binding = 12;
for (int i = 0; i < RenderBuffers::MAX_GIPROBES; i++) {
u.ids.push_back(rb->giprobe_textures[i]);
}
uniforms.push_back(u);
}
{
RD::Uniform u;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_SAMPLER;
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
u.binding = 13;
u.ids.push_back(storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED));
uniforms.push_back(u);
}
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
{
RD::Uniform u;
u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER;
u.binding = 14;
u.ids.push_back(volumetric_fog.params_ubo);
uniforms.push_back(u);
}
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
rb->volumetric_fog->uniform_set = RD::get_singleton()->uniform_set_create(uniforms, volumetric_fog.shader.version_get_shader(volumetric_fog.shader_version, 0), 0);
SWAP(uniforms.write[7].ids.write[0], uniforms.write[8].ids.write[0]);
rb->volumetric_fog->uniform_set2 = RD::get_singleton()->uniform_set_create(uniforms, volumetric_fog.shader.version_get_shader(volumetric_fog.shader_version, 0), 0);
}
bool using_sdfgi = env->volumetric_fog_gi_inject > 0.0001 && env->sdfgi_enabled && (rb->sdfgi != nullptr);
if (using_sdfgi) {
if (rb->volumetric_fog->sdfgi_uniform_set.is_null() || !RD::get_singleton()->uniform_set_is_valid(rb->volumetric_fog->sdfgi_uniform_set)) {
Vector<RD::Uniform> uniforms;
{
RD::Uniform u;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER;
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
u.binding = 0;
u.ids.push_back(gi.sdfgi_ubo);
uniforms.push_back(u);
}
{
RD::Uniform u;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_TEXTURE;
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
u.binding = 1;
u.ids.push_back(rb->sdfgi->ambient_texture);
uniforms.push_back(u);
}
{
RD::Uniform u;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_TEXTURE;
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
u.binding = 2;
u.ids.push_back(rb->sdfgi->occlusion_texture);
uniforms.push_back(u);
}
rb->volumetric_fog->sdfgi_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, volumetric_fog.shader.version_get_shader(volumetric_fog.shader_version, VOLUMETRIC_FOG_SHADER_DENSITY_WITH_SDFGI), 1);
}
}
rb->volumetric_fog->length = env->volumetric_fog_length;
rb->volumetric_fog->spread = env->volumetric_fog_detail_spread;
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
VolumetricFogShader::ParamsUBO params;
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
Vector2 frustum_near_size = p_cam_projection.get_viewport_half_extents();
Vector2 frustum_far_size = p_cam_projection.get_far_plane_half_extents();
float z_near = p_cam_projection.get_z_near();
float z_far = p_cam_projection.get_z_far();
float fog_end = env->volumetric_fog_length;
Vector2 fog_far_size = frustum_near_size.lerp(frustum_far_size, (fog_end - z_near) / (z_far - z_near));
Vector2 fog_near_size;
if (p_cam_projection.is_orthogonal()) {
fog_near_size = fog_far_size;
} else {
fog_near_size = Vector2();
}
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
params.fog_frustum_size_begin[0] = fog_near_size.x;
params.fog_frustum_size_begin[1] = fog_near_size.y;
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
params.fog_frustum_size_end[0] = fog_far_size.x;
params.fog_frustum_size_end[1] = fog_far_size.y;
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
params.z_near = z_near;
params.z_far = z_far;
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
params.fog_frustum_end = fog_end;
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
params.fog_volume_size[0] = rb->volumetric_fog->width;
params.fog_volume_size[1] = rb->volumetric_fog->height;
params.fog_volume_size[2] = rb->volumetric_fog->depth;
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
params.directional_light_count = p_directional_light_count;
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
Color light = env->volumetric_fog_light.to_linear();
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
params.light_energy[0] = light.r * env->volumetric_fog_light_energy;
params.light_energy[1] = light.g * env->volumetric_fog_light_energy;
params.light_energy[2] = light.b * env->volumetric_fog_light_energy;
params.base_density = env->volumetric_fog_density;
params.detail_spread = env->volumetric_fog_detail_spread;
params.gi_inject = env->volumetric_fog_gi_inject;
params.cam_rotation[0] = p_cam_transform.basis[0][0];
params.cam_rotation[1] = p_cam_transform.basis[1][0];
params.cam_rotation[2] = p_cam_transform.basis[2][0];
params.cam_rotation[3] = 0;
params.cam_rotation[4] = p_cam_transform.basis[0][1];
params.cam_rotation[5] = p_cam_transform.basis[1][1];
params.cam_rotation[6] = p_cam_transform.basis[2][1];
params.cam_rotation[7] = 0;
params.cam_rotation[8] = p_cam_transform.basis[0][2];
params.cam_rotation[9] = p_cam_transform.basis[1][2];
params.cam_rotation[10] = p_cam_transform.basis[2][2];
params.cam_rotation[11] = 0;
params.filter_axis = 0;
params.max_gi_probes = env->volumetric_fog_gi_inject > 0.001 ? p_gi_probe_count : 0;
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
{
uint32_t cluster_size = rb->cluster_builder->get_cluster_size();
params.cluster_shift = get_shift_from_power_of_2(cluster_size);
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
uint32_t cluster_screen_width = (rb->width - 1) / cluster_size + 1;
uint32_t cluster_screen_height = (rb->height - 1) / cluster_size + 1;
params.cluster_type_size = cluster_screen_width * cluster_screen_height * (32 + 32);
params.cluster_width = cluster_screen_width;
params.max_cluster_element_count_div_32 = max_cluster_elements / 32;
params.screen_size[0] = rb->width;
params.screen_size[1] = rb->height;
}
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
/* Vector2 dssize = directional_shadow_get_size();
push_constant.directional_shadow_pixel_size[0] = 1.0 / dssize.x;
push_constant.directional_shadow_pixel_size[1] = 1.0 / dssize.y;
*/
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
RENDER_TIMESTAMP(">Volumetric Fog");
RENDER_TIMESTAMP("Render Fog");
Reorganize RenderingDevice barriers -Removed sync to draw, now everything syncs to draw by default. -Fixed many validation layer errors. -Added support for VkImageViewUsageCreateInfo to fix validation layer warnings. -Texture, buffer, raster and compute functions now all allow spcifying which barriers will be used.
2021-01-26 00:52:58 +00:00
RD::get_singleton()->buffer_update(volumetric_fog.params_ubo, 0, sizeof(VolumetricFogShader::ParamsUBO), &params);
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin();
bool use_filter = volumetric_fog_filter_active;
RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, volumetric_fog.pipelines[using_sdfgi ? VOLUMETRIC_FOG_SHADER_DENSITY_WITH_SDFGI : VOLUMETRIC_FOG_SHADER_DENSITY]);
RD::get_singleton()->compute_list_bind_uniform_set(compute_list, rb->volumetric_fog->uniform_set, 0);
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
if (using_sdfgi) {
RD::get_singleton()->compute_list_bind_uniform_set(compute_list, rb->volumetric_fog->sdfgi_uniform_set, 1);
}
RD::get_singleton()->compute_list_dispatch_threads(compute_list, rb->volumetric_fog->width, rb->volumetric_fog->height, rb->volumetric_fog->depth, 4, 4, 4);
RD::get_singleton()->compute_list_add_barrier(compute_list);
if (use_filter) {
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
RENDER_TIMESTAMP("Filter Fog");
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, volumetric_fog.pipelines[VOLUMETRIC_FOG_SHADER_FILTER]);
RD::get_singleton()->compute_list_bind_uniform_set(compute_list, rb->volumetric_fog->uniform_set, 0);
RD::get_singleton()->compute_list_dispatch_threads(compute_list, rb->volumetric_fog->width, rb->volumetric_fog->height, rb->volumetric_fog->depth, 8, 8, 1);
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
RD::get_singleton()->compute_list_end();
//need restart for buffer update
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
params.filter_axis = 1;
Reorganize RenderingDevice barriers -Removed sync to draw, now everything syncs to draw by default. -Fixed many validation layer errors. -Added support for VkImageViewUsageCreateInfo to fix validation layer warnings. -Texture, buffer, raster and compute functions now all allow spcifying which barriers will be used.
2021-01-26 00:52:58 +00:00
RD::get_singleton()->buffer_update(volumetric_fog.params_ubo, 0, sizeof(VolumetricFogShader::ParamsUBO), &params);
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
compute_list = RD::get_singleton()->compute_list_begin();
RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, volumetric_fog.pipelines[VOLUMETRIC_FOG_SHADER_FILTER]);
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
RD::get_singleton()->compute_list_bind_uniform_set(compute_list, rb->volumetric_fog->uniform_set2, 0);
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
if (using_sdfgi) {
RD::get_singleton()->compute_list_bind_uniform_set(compute_list, rb->volumetric_fog->sdfgi_uniform_set, 1);
}
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
RD::get_singleton()->compute_list_dispatch_threads(compute_list, rb->volumetric_fog->width, rb->volumetric_fog->height, rb->volumetric_fog->depth, 8, 8, 1);
RD::get_singleton()->compute_list_add_barrier(compute_list);
}
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
RENDER_TIMESTAMP("Integrate Fog");
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, volumetric_fog.pipelines[VOLUMETRIC_FOG_SHADER_FOG]);
RD::get_singleton()->compute_list_bind_uniform_set(compute_list, rb->volumetric_fog->uniform_set, 0);
RD::get_singleton()->compute_list_dispatch_threads(compute_list, rb->volumetric_fog->width, rb->volumetric_fog->height, 1, 8, 8, 1);
RD::get_singleton()->compute_list_end();
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
RENDER_TIMESTAMP("<Volumetric Fog");
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
}
Reorganize renderer code. So it can hopefully be made more cache efficient afterwards.
2020-12-31 12:42:56 +00:00
void RendererSceneRenderRD::render_scene(RID p_render_buffers, const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, const PagedArray<GeometryInstance *> &p_instances, const PagedArray<RID> &p_lights, const PagedArray<RID> &p_reflection_probes, const PagedArray<RID> &p_gi_probes, const PagedArray<RID> &p_decals, const PagedArray<RID> &p_lightmaps, RID p_environment, RID p_camera_effects, RID p_shadow_atlas, RID p_reflection_atlas, RID p_reflection_probe, int p_reflection_probe_pass, float p_screen_lod_threshold) {
-Refactored post processing, re-added glow and added a mix blend mode.
2020-01-10 00:40:26 +00:00
Color clear_color;
if (p_render_buffers.is_valid()) {
RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers);
ERR_FAIL_COND(!rb);
clear_color = storage->render_target_get_clear_request_color(rb->render_target);
} else {
clear_color = storage->get_default_clear_color();
}
Base 3D engine done, still untested, though.
2019-08-18 22:40:52 +00:00
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
//assign render indices to giprobes
Replace Octree by DynamicBVH in cull code -Much greater pairing/unpairing performance -For now, using it for culling too, but this will change in a couple of days. -Added a paged allocator, to efficiently alloc/free some types of objects.
2020-12-23 16:52:58 +00:00
for (uint32_t i = 0; i < (uint32_t)p_gi_probes.size(); i++) {
GIProbeInstance *giprobe_inst = gi_probe_instance_owner.getornull(p_gi_probes[i]);
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
if (giprobe_inst) {
giprobe_inst->render_index = i;
}
}
Replace Octree by DynamicBVH in cull code -Much greater pairing/unpairing performance -For now, using it for culling too, but this will change in a couple of days. -Added a paged allocator, to efficiently alloc/free some types of objects.
2020-12-23 16:52:58 +00:00
const PagedArray<RID> *lights = &p_lights;
const PagedArray<RID> *reflections = &p_reflection_probes;
const PagedArray<RID> *gi_probes = &p_gi_probes;
PagedArray<RID> empty;
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
if (get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_UNSHADED) {
Replace Octree by DynamicBVH in cull code -Much greater pairing/unpairing performance -For now, using it for culling too, but this will change in a couple of days. -Added a paged allocator, to efficiently alloc/free some types of objects.
2020-12-23 16:52:58 +00:00
lights = &empty;
reflections = &empty;
gi_probes = &empty;
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
}
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
if (render_buffers_owner.owns(p_render_buffers)) {
RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers);
current_cluster_builder = rb->cluster_builder;
} else if (reflection_probe_instance_owner.owns(p_reflection_probe)) {
ReflectionProbeInstance *rpi = reflection_probe_instance_owner.getornull(p_reflection_probe);
ReflectionAtlas *ra = reflection_atlas_owner.getornull(rpi->atlas);
if (!ra) {
ERR_PRINT("reflection probe has no reflection atlas! Bug?");
current_cluster_builder = nullptr;
} else {
current_cluster_builder = ra->cluster_builder;
}
} else {
ERR_PRINT("No cluster builder, bug"); //should never happen, will crash
current_cluster_builder = nullptr;
}
current_cluster_builder->begin(p_cam_transform, p_cam_projection, !p_reflection_probe.is_valid());
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
bool using_shadows = true;
if (p_reflection_probe.is_valid()) {
if (!storage->reflection_probe_renders_shadows(reflection_probe_instance_get_probe(p_reflection_probe))) {
using_shadows = false;
}
} else {
//do not render reflections when rendering a reflection probe
Replace Octree by DynamicBVH in cull code -Much greater pairing/unpairing performance -For now, using it for culling too, but this will change in a couple of days. -Added a paged allocator, to efficiently alloc/free some types of objects.
2020-12-23 16:52:58 +00:00
_setup_reflections(*reflections, p_cam_transform.affine_inverse(), p_environment);
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
}
uint32_t directional_light_count = 0;
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
uint32_t positional_light_count = 0;
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
_setup_lights(*lights, p_cam_transform, p_shadow_atlas, using_shadows, directional_light_count, positional_light_count);
Replace Octree by DynamicBVH in cull code -Much greater pairing/unpairing performance -For now, using it for culling too, but this will change in a couple of days. -Added a paged allocator, to efficiently alloc/free some types of objects.
2020-12-23 16:52:58 +00:00
_setup_decals(p_decals, p_cam_transform.affine_inverse());
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
current_cluster_builder->bake_cluster();
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
uint32_t gi_probe_count = 0;
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
if (p_render_buffers.is_valid()) {
_setup_giprobes(p_render_buffers, p_cam_transform, *gi_probes, gi_probe_count);
}
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
if (p_render_buffers.is_valid()) {
bool directional_shadows = false;
for (uint32_t i = 0; i < directional_light_count; i++) {
if (cluster.directional_lights[i].shadow_enabled) {
directional_shadows = true;
break;
}
}
_update_volumetric_fog(p_render_buffers, p_environment, p_cam_projection, p_cam_transform, p_shadow_atlas, directional_light_count, directional_shadows, positional_light_count, gi_probe_count);
}
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
_render_scene(p_render_buffers, p_cam_transform, p_cam_projection, p_cam_ortogonal, p_instances, directional_light_count, *gi_probes, p_lightmaps, p_environment, current_cluster_builder->get_cluster_buffer(), current_cluster_builder->get_cluster_size(), current_cluster_builder->get_max_cluster_elements(), p_camera_effects, p_shadow_atlas, p_reflection_atlas, p_reflection_probe, p_reflection_probe_pass, clear_color, p_screen_lod_threshold);
WIP CameraEffects implementation (bokeh not working for now)
2020-01-13 18:37:24 +00:00
if (p_render_buffers.is_valid()) {
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
if (debug_draw == RS::VIEWPORT_DEBUG_DRAW_CLUSTER_OMNI_LIGHTS || debug_draw == RS::VIEWPORT_DEBUG_DRAW_CLUSTER_SPOT_LIGHTS || debug_draw == RS::VIEWPORT_DEBUG_DRAW_CLUSTER_DECALS || debug_draw == RS::VIEWPORT_DEBUG_DRAW_CLUSTER_REFLECTION_PROBES) {
ClusterBuilderRD::ElementType elem_type = ClusterBuilderRD::ELEMENT_TYPE_MAX;
switch (debug_draw) {
case RS::VIEWPORT_DEBUG_DRAW_CLUSTER_OMNI_LIGHTS:
elem_type = ClusterBuilderRD::ELEMENT_TYPE_OMNI_LIGHT;
break;
case RS::VIEWPORT_DEBUG_DRAW_CLUSTER_SPOT_LIGHTS:
elem_type = ClusterBuilderRD::ELEMENT_TYPE_SPOT_LIGHT;
break;
case RS::VIEWPORT_DEBUG_DRAW_CLUSTER_DECALS:
elem_type = ClusterBuilderRD::ELEMENT_TYPE_DECAL;
break;
case RS::VIEWPORT_DEBUG_DRAW_CLUSTER_REFLECTION_PROBES:
elem_type = ClusterBuilderRD::ELEMENT_TYPE_REFLECTION_PROBE;
break;
default: {
}
}
current_cluster_builder->debug(elem_type);
}
WIP CameraEffects implementation (bokeh not working for now)
2020-01-13 18:37:24 +00:00
RENDER_TIMESTAMP("Tonemap");
_render_buffers_post_process_and_tonemap(p_render_buffers, p_environment, p_camera_effects, p_cam_projection);
_render_buffers_debug_draw(p_render_buffers, p_shadow_atlas);
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
if (debug_draw == RS::VIEWPORT_DEBUG_DRAW_SDFGI) {
_sdfgi_debug_draw(p_render_buffers, p_cam_projection, p_cam_transform);
}
WIP CameraEffects implementation (bokeh not working for now)
2020-01-13 18:37:24 +00:00
}
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
}
Reorganize renderer code. So it can hopefully be made more cache efficient afterwards.
2020-12-31 12:42:56 +00:00
void RendererSceneRenderRD::render_shadow(RID p_light, RID p_shadow_atlas, int p_pass, const PagedArray<GeometryInstance *> &p_instances, const Plane &p_camera_plane, float p_lod_distance_multiplier, float p_screen_lod_threshold) {
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
LightInstance *light_instance = light_instance_owner.getornull(p_light);
ERR_FAIL_COND(!light_instance);
Rect2i atlas_rect;
Shadow map rendering optimization -All shadow rendering is done with raster now (no compute) -All shadow rendering is done by rendering directly to the shadow atlas -Improved how buffer clearing is done to optimize the above. -Ability to set shadows as 16 bits.
2021-01-24 19:00:20 +00:00
uint32_t atlas_size;
RID atlas_fb;
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
bool using_dual_paraboloid = false;
bool using_dual_paraboloid_flip = false;
RID render_fb;
RID render_texture;
Shadow map rendering optimization -All shadow rendering is done with raster now (no compute) -All shadow rendering is done by rendering directly to the shadow atlas -Improved how buffer clearing is done to optimize the above. -Ability to set shadows as 16 bits.
2021-01-24 19:00:20 +00:00
float zfar;
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
Refactored shadowmapping. - Made shadow bias size independent, so it will remain when changing light or camera size. - Implemented normal offset bias, which greatly enhances quality. - Added transmission to subsurface scattering - Reimplemented shadow filter modes Closes #17260
2020-04-08 01:51:52 +00:00
bool use_pancake = false;
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
bool render_cubemap = false;
bool finalize_cubemap = false;
Shadow map rendering optimization -All shadow rendering is done with raster now (no compute) -All shadow rendering is done by rendering directly to the shadow atlas -Improved how buffer clearing is done to optimize the above. -Ability to set shadows as 16 bits.
2021-01-24 19:00:20 +00:00
bool flip_y = false;
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
CameraMatrix light_projection;
Transform light_transform;
Shadow map rendering optimization -All shadow rendering is done with raster now (no compute) -All shadow rendering is done by rendering directly to the shadow atlas -Improved how buffer clearing is done to optimize the above. -Ability to set shadows as 16 bits.
2021-01-24 19:00:20 +00:00
bool clear_region = true;
bool begin_texture = true;
bool end_texture = true;
Renaming of servers for coherency. VisualServer -> RenderingServer PhysicsServer -> PhysicsServer3D Physics2DServer -> PhysicsServer2D NavigationServer -> NavigationServer3D Navigation2DServer -> NavigationServer2D Also renamed corresponding files.
2020-03-27 18:21:27 +00:00
if (storage->light_get_type(light_instance->light) == RS::LIGHT_DIRECTIONAL) {
Shadow map rendering optimization -All shadow rendering is done with raster now (no compute) -All shadow rendering is done by rendering directly to the shadow atlas -Improved how buffer clearing is done to optimize the above. -Ability to set shadows as 16 bits.
2021-01-24 19:00:20 +00:00
_update_directional_shadow_atlas();
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
//set pssm stuff
if (light_instance->last_scene_shadow_pass != scene_pass) {
Directional lights and shadow mapping are functional.
2019-09-07 17:38:17 +00:00
light_instance->directional_rect = _get_directional_shadow_rect(directional_shadow.size, directional_shadow.light_count, directional_shadow.current_light);
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
directional_shadow.current_light++;
Directional lights and shadow mapping are functional.
2019-09-07 17:38:17 +00:00
light_instance->last_scene_shadow_pass = scene_pass;
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
}
Refactored shadowmapping. - Made shadow bias size independent, so it will remain when changing light or camera size. - Implemented normal offset bias, which greatly enhances quality. - Added transmission to subsurface scattering - Reimplemented shadow filter modes Closes #17260
2020-04-08 01:51:52 +00:00
use_pancake = storage->light_get_param(light_instance->light, RS::LIGHT_PARAM_SHADOW_PANCAKE_SIZE) > 0;
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
light_projection = light_instance->shadow_transform[p_pass].camera;
light_transform = light_instance->shadow_transform[p_pass].transform;
atlas_rect.position.x = light_instance->directional_rect.position.x;
atlas_rect.position.y = light_instance->directional_rect.position.y;
atlas_rect.size.width = light_instance->directional_rect.size.x;
atlas_rect.size.height = light_instance->directional_rect.size.y;
Shadow map rendering optimization -All shadow rendering is done with raster now (no compute) -All shadow rendering is done by rendering directly to the shadow atlas -Improved how buffer clearing is done to optimize the above. -Ability to set shadows as 16 bits.
2021-01-24 19:00:20 +00:00
int pass_count = 1;
Renaming of servers for coherency. VisualServer -> RenderingServer PhysicsServer -> PhysicsServer3D Physics2DServer -> PhysicsServer2D NavigationServer -> NavigationServer3D Navigation2DServer -> NavigationServer2D Also renamed corresponding files.
2020-03-27 18:21:27 +00:00
if (storage->light_directional_get_shadow_mode(light_instance->light) == RS::LIGHT_DIRECTIONAL_SHADOW_PARALLEL_4_SPLITS) {
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
atlas_rect.size.width /= 2;
atlas_rect.size.height /= 2;
if (p_pass == 1) {
atlas_rect.position.x += atlas_rect.size.width;
} else if (p_pass == 2) {
atlas_rect.position.y += atlas_rect.size.height;
} else if (p_pass == 3) {
atlas_rect.position.x += atlas_rect.size.width;
atlas_rect.position.y += atlas_rect.size.height;
}
Shadow map rendering optimization -All shadow rendering is done with raster now (no compute) -All shadow rendering is done by rendering directly to the shadow atlas -Improved how buffer clearing is done to optimize the above. -Ability to set shadows as 16 bits.
2021-01-24 19:00:20 +00:00
pass_count = 4;
Renaming of servers for coherency. VisualServer -> RenderingServer PhysicsServer -> PhysicsServer3D Physics2DServer -> PhysicsServer2D NavigationServer -> NavigationServer3D Navigation2DServer -> NavigationServer2D Also renamed corresponding files.
2020-03-27 18:21:27 +00:00
} else if (storage->light_directional_get_shadow_mode(light_instance->light) == RS::LIGHT_DIRECTIONAL_SHADOW_PARALLEL_2_SPLITS) {
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
atlas_rect.size.height /= 2;
if (p_pass == 0) {
} else {
atlas_rect.position.y += atlas_rect.size.height;
}
Shadow map rendering optimization -All shadow rendering is done with raster now (no compute) -All shadow rendering is done by rendering directly to the shadow atlas -Improved how buffer clearing is done to optimize the above. -Ability to set shadows as 16 bits.
2021-01-24 19:00:20 +00:00
pass_count = 2;
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
}
Directional lights and shadow mapping are functional.
2019-09-07 17:38:17 +00:00
light_instance->shadow_transform[p_pass].atlas_rect = atlas_rect;
light_instance->shadow_transform[p_pass].atlas_rect.position /= directional_shadow.size;
light_instance->shadow_transform[p_pass].atlas_rect.size /= directional_shadow.size;
Renaming of servers for coherency. VisualServer -> RenderingServer PhysicsServer -> PhysicsServer3D Physics2DServer -> PhysicsServer2D NavigationServer -> NavigationServer3D Navigation2DServer -> NavigationServer2D Also renamed corresponding files.
2020-03-27 18:21:27 +00:00
zfar = storage->light_get_param(light_instance->light, RS::LIGHT_PARAM_RANGE);
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
Shadow map rendering optimization -All shadow rendering is done with raster now (no compute) -All shadow rendering is done by rendering directly to the shadow atlas -Improved how buffer clearing is done to optimize the above. -Ability to set shadows as 16 bits.
2021-01-24 19:00:20 +00:00
render_fb = directional_shadow.fb;
render_texture = RID();
flip_y = true;
clear_region = false;
begin_texture = (directional_shadow.current_light == 1) && (p_pass == 0); //light is 1-index because it was incremented above
end_texture = (directional_shadow.current_light == directional_shadow.light_count) && (p_pass == pass_count - 1);
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
} else {
//set from shadow atlas
ShadowAtlas *shadow_atlas = shadow_atlas_owner.getornull(p_shadow_atlas);
ERR_FAIL_COND(!shadow_atlas);
ERR_FAIL_COND(!shadow_atlas->shadow_owners.has(p_light));
Shadow map rendering optimization -All shadow rendering is done with raster now (no compute) -All shadow rendering is done by rendering directly to the shadow atlas -Improved how buffer clearing is done to optimize the above. -Ability to set shadows as 16 bits.
2021-01-24 19:00:20 +00:00
_update_shadow_atlas(shadow_atlas);
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
uint32_t key = shadow_atlas->shadow_owners[p_light];
uint32_t quadrant = (key >> ShadowAtlas::QUADRANT_SHIFT) & 0x3;
uint32_t shadow = key & ShadowAtlas::SHADOW_INDEX_MASK;
ERR_FAIL_INDEX((int)shadow, shadow_atlas->quadrants[quadrant].shadows.size());
uint32_t quadrant_size = shadow_atlas->size >> 1;
atlas_rect.position.x = (quadrant & 1) * quadrant_size;
atlas_rect.position.y = (quadrant >> 1) * quadrant_size;
uint32_t shadow_size = (quadrant_size / shadow_atlas->quadrants[quadrant].subdivision);
atlas_rect.position.x += (shadow % shadow_atlas->quadrants[quadrant].subdivision) * shadow_size;
atlas_rect.position.y += (shadow / shadow_atlas->quadrants[quadrant].subdivision) * shadow_size;
atlas_rect.size.width = shadow_size;
atlas_rect.size.height = shadow_size;
Renaming of servers for coherency. VisualServer -> RenderingServer PhysicsServer -> PhysicsServer3D Physics2DServer -> PhysicsServer2D NavigationServer -> NavigationServer3D Navigation2DServer -> NavigationServer2D Also renamed corresponding files.
2020-03-27 18:21:27 +00:00
zfar = storage->light_get_param(light_instance->light, RS::LIGHT_PARAM_RANGE);
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
Renaming of servers for coherency. VisualServer -> RenderingServer PhysicsServer -> PhysicsServer3D Physics2DServer -> PhysicsServer2D NavigationServer -> NavigationServer3D Navigation2DServer -> NavigationServer2D Also renamed corresponding files.
2020-03-27 18:21:27 +00:00
if (storage->light_get_type(light_instance->light) == RS::LIGHT_OMNI) {
if (storage->light_omni_get_shadow_mode(light_instance->light) == RS::LIGHT_OMNI_SHADOW_CUBE) {
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
ShadowCubemap *cubemap = _get_shadow_cubemap(shadow_size / 2);
render_fb = cubemap->side_fb[p_pass];
render_texture = cubemap->cubemap;
light_projection = light_instance->shadow_transform[0].camera;
light_transform = light_instance->shadow_transform[0].transform;
render_cubemap = true;
finalize_cubemap = p_pass == 5;
Shadow map rendering optimization -All shadow rendering is done with raster now (no compute) -All shadow rendering is done by rendering directly to the shadow atlas -Improved how buffer clearing is done to optimize the above. -Ability to set shadows as 16 bits.
2021-01-24 19:00:20 +00:00
atlas_fb = shadow_atlas->fb;
atlas_size = shadow_atlas->size;
clear_region = false;
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
} else {
light_projection = light_instance->shadow_transform[0].camera;
light_transform = light_instance->shadow_transform[0].transform;
atlas_rect.size.height /= 2;
atlas_rect.position.y += p_pass * atlas_rect.size.height;
using_dual_paraboloid = true;
using_dual_paraboloid_flip = p_pass == 1;
Shadow map rendering optimization -All shadow rendering is done with raster now (no compute) -All shadow rendering is done by rendering directly to the shadow atlas -Improved how buffer clearing is done to optimize the above. -Ability to set shadows as 16 bits.
2021-01-24 19:00:20 +00:00
render_fb = shadow_atlas->fb;
flip_y = true;
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
}
Renaming of servers for coherency. VisualServer -> RenderingServer PhysicsServer -> PhysicsServer3D Physics2DServer -> PhysicsServer2D NavigationServer -> NavigationServer3D Navigation2DServer -> NavigationServer2D Also renamed corresponding files.
2020-03-27 18:21:27 +00:00
} else if (storage->light_get_type(light_instance->light) == RS::LIGHT_SPOT) {
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
light_projection = light_instance->shadow_transform[0].camera;
light_transform = light_instance->shadow_transform[0].transform;
Shadow map rendering optimization -All shadow rendering is done with raster now (no compute) -All shadow rendering is done by rendering directly to the shadow atlas -Improved how buffer clearing is done to optimize the above. -Ability to set shadows as 16 bits.
2021-01-24 19:00:20 +00:00
render_fb = shadow_atlas->fb;
Refactored shadowmapping. - Made shadow bias size independent, so it will remain when changing light or camera size. - Implemented normal offset bias, which greatly enhances quality. - Added transmission to subsurface scattering - Reimplemented shadow filter modes Closes #17260
2020-04-08 01:51:52 +00:00
Shadow map rendering optimization -All shadow rendering is done with raster now (no compute) -All shadow rendering is done by rendering directly to the shadow atlas -Improved how buffer clearing is done to optimize the above. -Ability to set shadows as 16 bits.
2021-01-24 19:00:20 +00:00
flip_y = true;
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
}
}
if (render_cubemap) {
//rendering to cubemap
Replace Octree by DynamicBVH in cull code -Much greater pairing/unpairing performance -For now, using it for culling too, but this will change in a couple of days. -Added a paged allocator, to efficiently alloc/free some types of objects.
2020-12-23 16:52:58 +00:00
_render_shadow(render_fb, p_instances, light_projection, light_transform, zfar, 0, 0, false, false, use_pancake, p_camera_plane, p_lod_distance_multiplier, p_screen_lod_threshold);
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
if (finalize_cubemap) {
//reblit
Shadow map rendering optimization -All shadow rendering is done with raster now (no compute) -All shadow rendering is done by rendering directly to the shadow atlas -Improved how buffer clearing is done to optimize the above. -Ability to set shadows as 16 bits.
2021-01-24 19:00:20 +00:00
Rect2 atlas_rect_norm = atlas_rect;
atlas_rect_norm.position.x /= float(atlas_size);
atlas_rect_norm.position.y /= float(atlas_size);
atlas_rect_norm.size.x /= float(atlas_size);
atlas_rect_norm.size.y /= float(atlas_size);
atlas_rect_norm.size.height /= 2;
storage->get_effects()->copy_cubemap_to_dp(render_texture, atlas_fb, atlas_rect_norm, light_projection.get_z_near(), light_projection.get_z_far(), false);
atlas_rect_norm.position.y += atlas_rect_norm.size.height;
storage->get_effects()->copy_cubemap_to_dp(render_texture, atlas_fb, atlas_rect_norm, light_projection.get_z_near(), light_projection.get_z_far(), true);
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
}
} else {
//render shadow
Shadow map rendering optimization -All shadow rendering is done with raster now (no compute) -All shadow rendering is done by rendering directly to the shadow atlas -Improved how buffer clearing is done to optimize the above. -Ability to set shadows as 16 bits.
2021-01-24 19:00:20 +00:00
_render_shadow(render_fb, p_instances, light_projection, light_transform, zfar, 0, 0, using_dual_paraboloid, using_dual_paraboloid_flip, use_pancake, p_camera_plane, p_lod_distance_multiplier, p_screen_lod_threshold, atlas_rect, flip_y, clear_region, begin_texture, end_texture);
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
}
Base 3D engine done, still untested, though.
2019-08-18 22:40:52 +00:00
}
Reorganize renderer code. So it can hopefully be made more cache efficient afterwards.
2020-12-31 12:42:56 +00:00
void RendererSceneRenderRD::render_material(const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, const PagedArray<GeometryInstance *> &p_instances, RID p_framebuffer, const Rect2i &p_region) {
Replace Octree by DynamicBVH in cull code -Much greater pairing/unpairing performance -For now, using it for culling too, but this will change in a couple of days. -Added a paged allocator, to efficiently alloc/free some types of objects.
2020-12-23 16:52:58 +00:00
_render_material(p_cam_transform, p_cam_projection, p_cam_ortogonal, p_instances, p_framebuffer, p_region);
Dynamic object support for GI Probes (a bit buggy still)
2019-10-11 02:14:56 +00:00
}
Reorganize renderer code. So it can hopefully be made more cache efficient afterwards.
2020-12-31 12:42:56 +00:00
void RendererSceneRenderRD::render_sdfgi(RID p_render_buffers, int p_region, const PagedArray<GeometryInstance *> &p_instances) {
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
//print_line("rendering region " + itos(p_region));
RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers);
ERR_FAIL_COND(!rb);
ERR_FAIL_COND(!rb->sdfgi);
AABB bounds;
Vector3i from;
Vector3i size;
int cascade_prev = _sdfgi_get_pending_region_data(p_render_buffers, p_region - 1, from, size, bounds);
int cascade_next = _sdfgi_get_pending_region_data(p_render_buffers, p_region + 1, from, size, bounds);
int cascade = _sdfgi_get_pending_region_data(p_render_buffers, p_region, from, size, bounds);
ERR_FAIL_COND(cascade < 0);
if (cascade_prev != cascade) {
//initialize render
Reorganize RenderingDevice barriers -Removed sync to draw, now everything syncs to draw by default. -Fixed many validation layer errors. -Added support for VkImageViewUsageCreateInfo to fix validation layer warnings. -Texture, buffer, raster and compute functions now all allow spcifying which barriers will be used.
2021-01-26 00:52:58 +00:00
RD::get_singleton()->texture_clear(rb->sdfgi->render_albedo, Color(0, 0, 0, 0), 0, 1, 0, 1);
RD::get_singleton()->texture_clear(rb->sdfgi->render_emission, Color(0, 0, 0, 0), 0, 1, 0, 1);
RD::get_singleton()->texture_clear(rb->sdfgi->render_emission_aniso, Color(0, 0, 0, 0), 0, 1, 0, 1);
RD::get_singleton()->texture_clear(rb->sdfgi->render_geom_facing, Color(0, 0, 0, 0), 0, 1, 0, 1);
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
}
//print_line("rendering cascade " + itos(p_region) + " objects: " + itos(p_cull_count) + " bounds: " + bounds + " from: " + from + " size: " + size + " cell size: " + rtos(rb->sdfgi->cascades[cascade].cell_size));
Replace Octree by DynamicBVH in cull code -Much greater pairing/unpairing performance -For now, using it for culling too, but this will change in a couple of days. -Added a paged allocator, to efficiently alloc/free some types of objects.
2020-12-23 16:52:58 +00:00
_render_sdfgi(p_render_buffers, from, size, bounds, p_instances, rb->sdfgi->render_albedo, rb->sdfgi->render_emission, rb->sdfgi->render_emission_aniso, rb->sdfgi->render_geom_facing);
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
if (cascade_next != cascade) {
RENDER_TIMESTAMP(">SDFGI Update SDF");
//done rendering! must update SDF
//clear dispatch indirect data
SDGIShader::PreprocessPushConstant push_constant;
zeromem(&push_constant, sizeof(SDGIShader::PreprocessPushConstant));
RENDER_TIMESTAMP("Scroll SDF");
//scroll
if (rb->sdfgi->cascades[cascade].dirty_regions != SDFGI::Cascade::DIRTY_ALL) {
//for scroll
Vector3i dirty = rb->sdfgi->cascades[cascade].dirty_regions;
push_constant.scroll[0] = dirty.x;
push_constant.scroll[1] = dirty.y;
push_constant.scroll[2] = dirty.z;
} else {
//for no scroll
push_constant.scroll[0] = 0;
push_constant.scroll[1] = 0;
push_constant.scroll[2] = 0;
}
Several GI related optimizations and fixes -SDFGI direct light is done over many frames -SDFGI Changed settings for rays/frame -SDFGI Misc optimizations -SDFGI Bug fix on probe scroll -GIProbe was not working, got it to work again -GIProbe dynamic objects were not working, fixed -Added a half size GI option.
2021-01-22 23:50:24 +00:00
rb->sdfgi->cascades[cascade].all_dynamic_lights_dirty = true;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
push_constant.grid_size = rb->sdfgi->cascade_size;
push_constant.cascade = cascade;
if (rb->sdfgi->cascades[cascade].dirty_regions != SDFGI::Cascade::DIRTY_ALL) {
Fix invalid buffer updates in SDFGI
2020-10-19 20:01:53 +00:00
RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin();
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
//must pre scroll existing data because not all is dirty
RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, sdfgi_shader.preprocess_pipeline[SDGIShader::PRE_PROCESS_SCROLL]);
RD::get_singleton()->compute_list_bind_uniform_set(compute_list, rb->sdfgi->cascades[cascade].scroll_uniform_set, 0);
RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(SDGIShader::PreprocessPushConstant));
RD::get_singleton()->compute_list_dispatch_indirect(compute_list, rb->sdfgi->cascades[cascade].solid_cell_dispatch_buffer, 0);
// no barrier do all together
RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, sdfgi_shader.preprocess_pipeline[SDGIShader::PRE_PROCESS_SCROLL_OCCLUSION]);
RD::get_singleton()->compute_list_bind_uniform_set(compute_list, rb->sdfgi->cascades[cascade].scroll_occlusion_uniform_set, 0);
Vector3i dirty = rb->sdfgi->cascades[cascade].dirty_regions;
Vector3i groups;
groups.x = rb->sdfgi->cascade_size - ABS(dirty.x);
groups.y = rb->sdfgi->cascade_size - ABS(dirty.y);
groups.z = rb->sdfgi->cascade_size - ABS(dirty.z);
RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(SDGIShader::PreprocessPushConstant));
RD::get_singleton()->compute_list_dispatch_threads(compute_list, groups.x, groups.y, groups.z, 4, 4, 4);
//no barrier, continue together
{
//scroll probes and their history also
SDGIShader::IntegratePushConstant ipush_constant;
ipush_constant.grid_size[1] = rb->sdfgi->cascade_size;
ipush_constant.grid_size[2] = rb->sdfgi->cascade_size;
ipush_constant.grid_size[0] = rb->sdfgi->cascade_size;
ipush_constant.max_cascades = rb->sdfgi->cascades.size();
ipush_constant.probe_axis_size = rb->sdfgi->probe_axis_count;
ipush_constant.history_index = 0;
ipush_constant.history_size = rb->sdfgi->history_size;
ipush_constant.ray_count = 0;
ipush_constant.ray_bias = 0;
ipush_constant.sky_mode = 0;
ipush_constant.sky_energy = 0;
ipush_constant.sky_color[0] = 0;
ipush_constant.sky_color[1] = 0;
ipush_constant.sky_color[2] = 0;
ipush_constant.y_mult = rb->sdfgi->y_mult;
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
ipush_constant.store_ambient_texture = false;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
ipush_constant.image_size[0] = rb->sdfgi->probe_axis_count * rb->sdfgi->probe_axis_count;
ipush_constant.image_size[1] = rb->sdfgi->probe_axis_count;
int32_t probe_divisor = rb->sdfgi->cascade_size / SDFGI::PROBE_DIVISOR;
ipush_constant.cascade = cascade;
ipush_constant.world_offset[0] = rb->sdfgi->cascades[cascade].position.x / probe_divisor;
ipush_constant.world_offset[1] = rb->sdfgi->cascades[cascade].position.y / probe_divisor;
ipush_constant.world_offset[2] = rb->sdfgi->cascades[cascade].position.z / probe_divisor;
ipush_constant.scroll[0] = dirty.x / probe_divisor;
ipush_constant.scroll[1] = dirty.y / probe_divisor;
ipush_constant.scroll[2] = dirty.z / probe_divisor;
RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, sdfgi_shader.integrate_pipeline[SDGIShader::INTEGRATE_MODE_SCROLL]);
RD::get_singleton()->compute_list_bind_uniform_set(compute_list, rb->sdfgi->cascades[cascade].integrate_uniform_set, 0);
RD::get_singleton()->compute_list_bind_uniform_set(compute_list, sdfgi_shader.integrate_default_sky_uniform_set, 1);
RD::get_singleton()->compute_list_set_push_constant(compute_list, &ipush_constant, sizeof(SDGIShader::IntegratePushConstant));
RD::get_singleton()->compute_list_dispatch_threads(compute_list, rb->sdfgi->probe_axis_count * rb->sdfgi->probe_axis_count, rb->sdfgi->probe_axis_count, 1, 8, 8, 1);
RD::get_singleton()->compute_list_add_barrier(compute_list);
RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, sdfgi_shader.integrate_pipeline[SDGIShader::INTEGRATE_MODE_SCROLL_STORE]);
RD::get_singleton()->compute_list_bind_uniform_set(compute_list, rb->sdfgi->cascades[cascade].integrate_uniform_set, 0);
RD::get_singleton()->compute_list_bind_uniform_set(compute_list, sdfgi_shader.integrate_default_sky_uniform_set, 1);
RD::get_singleton()->compute_list_set_push_constant(compute_list, &ipush_constant, sizeof(SDGIShader::IntegratePushConstant));
RD::get_singleton()->compute_list_dispatch_threads(compute_list, rb->sdfgi->probe_axis_count * rb->sdfgi->probe_axis_count, rb->sdfgi->probe_axis_count, 1, 8, 8, 1);
Several GI related optimizations and fixes -SDFGI direct light is done over many frames -SDFGI Changed settings for rays/frame -SDFGI Misc optimizations -SDFGI Bug fix on probe scroll -GIProbe was not working, got it to work again -GIProbe dynamic objects were not working, fixed -Added a half size GI option.
2021-01-22 23:50:24 +00:00
RD::get_singleton()->compute_list_add_barrier(compute_list);
if (rb->sdfgi->uses_multibounce) {
//multibounce requires this to be stored so direct light can read from it
RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, sdfgi_shader.integrate_pipeline[SDGIShader::INTEGRATE_MODE_STORE]);
//convert to octahedral to store
ipush_constant.image_size[0] *= SDFGI::LIGHTPROBE_OCT_SIZE;
ipush_constant.image_size[1] *= SDFGI::LIGHTPROBE_OCT_SIZE;
RD::get_singleton()->compute_list_bind_uniform_set(compute_list, rb->sdfgi->cascades[cascade].integrate_uniform_set, 0);
RD::get_singleton()->compute_list_bind_uniform_set(compute_list, sdfgi_shader.integrate_default_sky_uniform_set, 1);
RD::get_singleton()->compute_list_set_push_constant(compute_list, &ipush_constant, sizeof(SDGIShader::IntegratePushConstant));
RD::get_singleton()->compute_list_dispatch_threads(compute_list, rb->sdfgi->probe_axis_count * rb->sdfgi->probe_axis_count * SDFGI::LIGHTPROBE_OCT_SIZE, rb->sdfgi->probe_axis_count * SDFGI::LIGHTPROBE_OCT_SIZE, 1, 8, 8, 1);
}
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
}
//ok finally barrier
Fix invalid buffer updates in SDFGI
2020-10-19 20:01:53 +00:00
RD::get_singleton()->compute_list_end();
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
}
//clear dispatch indirect data
uint32_t dispatch_indirct_data[4] = { 0, 0, 0, 0 };
Reorganize RenderingDevice barriers -Removed sync to draw, now everything syncs to draw by default. -Fixed many validation layer errors. -Added support for VkImageViewUsageCreateInfo to fix validation layer warnings. -Texture, buffer, raster and compute functions now all allow spcifying which barriers will be used.
2021-01-26 00:52:58 +00:00
RD::get_singleton()->buffer_update(rb->sdfgi->cascades[cascade].solid_cell_dispatch_buffer, 0, sizeof(uint32_t) * 4, dispatch_indirct_data);
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
Fix invalid buffer updates in SDFGI
2020-10-19 20:01:53 +00:00
RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin();
Fix spelling & grammar in comments, docs, and messages
2020-07-16 16:54:15 +00:00
bool half_size = true; //much faster, very little difference
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
static const int optimized_jf_group_size = 8;
if (half_size) {
push_constant.grid_size >>= 1;
uint32_t cascade_half_size = rb->sdfgi->cascade_size >> 1;
RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, sdfgi_shader.preprocess_pipeline[SDGIShader::PRE_PROCESS_JUMP_FLOOD_INITIALIZE_HALF]);
RD::get_singleton()->compute_list_bind_uniform_set(compute_list, rb->sdfgi->sdf_initialize_half_uniform_set, 0);
RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(SDGIShader::PreprocessPushConstant));
RD::get_singleton()->compute_list_dispatch_threads(compute_list, cascade_half_size, cascade_half_size, cascade_half_size, 4, 4, 4);
RD::get_singleton()->compute_list_add_barrier(compute_list);
//must start with regular jumpflood
push_constant.half_size = true;
{
RENDER_TIMESTAMP("SDFGI Jump Flood (Half Size)");
uint32_t s = cascade_half_size;
RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, sdfgi_shader.preprocess_pipeline[SDGIShader::PRE_PROCESS_JUMP_FLOOD]);
int jf_us = 0;
//start with regular jump flood for very coarse reads, as this is impossible to optimize
while (s > 1) {
s /= 2;
push_constant.step_size = s;
RD::get_singleton()->compute_list_bind_uniform_set(compute_list, rb->sdfgi->jump_flood_half_uniform_set[jf_us], 0);
RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(SDGIShader::PreprocessPushConstant));
RD::get_singleton()->compute_list_dispatch_threads(compute_list, cascade_half_size, cascade_half_size, cascade_half_size, 4, 4, 4);
RD::get_singleton()->compute_list_add_barrier(compute_list);
jf_us = jf_us == 0 ? 1 : 0;
if (cascade_half_size / (s / 2) >= optimized_jf_group_size) {
break;
}
}
RENDER_TIMESTAMP("SDFGI Jump Flood Optimized (Half Size)");
//continue with optimized jump flood for smaller reads
RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, sdfgi_shader.preprocess_pipeline[SDGIShader::PRE_PROCESS_JUMP_FLOOD_OPTIMIZED]);
while (s > 1) {
s /= 2;
push_constant.step_size = s;
RD::get_singleton()->compute_list_bind_uniform_set(compute_list, rb->sdfgi->jump_flood_half_uniform_set[jf_us], 0);
RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(SDGIShader::PreprocessPushConstant));
RD::get_singleton()->compute_list_dispatch_threads(compute_list, cascade_half_size, cascade_half_size, cascade_half_size, optimized_jf_group_size, optimized_jf_group_size, optimized_jf_group_size);
RD::get_singleton()->compute_list_add_barrier(compute_list);
jf_us = jf_us == 0 ? 1 : 0;
}
}
// restore grid size for last passes
push_constant.grid_size = rb->sdfgi->cascade_size;
RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, sdfgi_shader.preprocess_pipeline[SDGIShader::PRE_PROCESS_JUMP_FLOOD_UPSCALE]);
RD::get_singleton()->compute_list_bind_uniform_set(compute_list, rb->sdfgi->sdf_upscale_uniform_set, 0);
RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(SDGIShader::PreprocessPushConstant));
RD::get_singleton()->compute_list_dispatch_threads(compute_list, rb->sdfgi->cascade_size, rb->sdfgi->cascade_size, rb->sdfgi->cascade_size, 4, 4, 4);
RD::get_singleton()->compute_list_add_barrier(compute_list);
//run one pass of fullsize jumpflood to fix up half size arctifacts
push_constant.half_size = false;
push_constant.step_size = 1;
RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, sdfgi_shader.preprocess_pipeline[SDGIShader::PRE_PROCESS_JUMP_FLOOD_OPTIMIZED]);
RD::get_singleton()->compute_list_bind_uniform_set(compute_list, rb->sdfgi->jump_flood_uniform_set[rb->sdfgi->upscale_jfa_uniform_set_index], 0);
RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(SDGIShader::PreprocessPushConstant));
RD::get_singleton()->compute_list_dispatch_threads(compute_list, rb->sdfgi->cascade_size, rb->sdfgi->cascade_size, rb->sdfgi->cascade_size, optimized_jf_group_size, optimized_jf_group_size, optimized_jf_group_size);
RD::get_singleton()->compute_list_add_barrier(compute_list);
} else {
//full size jumpflood
RENDER_TIMESTAMP("SDFGI Jump Flood");
RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, sdfgi_shader.preprocess_pipeline[SDGIShader::PRE_PROCESS_JUMP_FLOOD_INITIALIZE]);
RD::get_singleton()->compute_list_bind_uniform_set(compute_list, rb->sdfgi->sdf_initialize_uniform_set, 0);
RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(SDGIShader::PreprocessPushConstant));
RD::get_singleton()->compute_list_dispatch_threads(compute_list, rb->sdfgi->cascade_size, rb->sdfgi->cascade_size, rb->sdfgi->cascade_size, 4, 4, 4);
RD::get_singleton()->compute_list_add_barrier(compute_list);
push_constant.half_size = false;
{
uint32_t s = rb->sdfgi->cascade_size;
RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, sdfgi_shader.preprocess_pipeline[SDGIShader::PRE_PROCESS_JUMP_FLOOD]);
int jf_us = 0;
//start with regular jump flood for very coarse reads, as this is impossible to optimize
while (s > 1) {
s /= 2;
push_constant.step_size = s;
RD::get_singleton()->compute_list_bind_uniform_set(compute_list, rb->sdfgi->jump_flood_uniform_set[jf_us], 0);
RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(SDGIShader::PreprocessPushConstant));
RD::get_singleton()->compute_list_dispatch_threads(compute_list, rb->sdfgi->cascade_size, rb->sdfgi->cascade_size, rb->sdfgi->cascade_size, 4, 4, 4);
RD::get_singleton()->compute_list_add_barrier(compute_list);
jf_us = jf_us == 0 ? 1 : 0;
if (rb->sdfgi->cascade_size / (s / 2) >= optimized_jf_group_size) {
break;
}
}
RENDER_TIMESTAMP("SDFGI Jump Flood Optimized");
//continue with optimized jump flood for smaller reads
RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, sdfgi_shader.preprocess_pipeline[SDGIShader::PRE_PROCESS_JUMP_FLOOD_OPTIMIZED]);
while (s > 1) {
s /= 2;
push_constant.step_size = s;
RD::get_singleton()->compute_list_bind_uniform_set(compute_list, rb->sdfgi->jump_flood_uniform_set[jf_us], 0);
RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(SDGIShader::PreprocessPushConstant));
RD::get_singleton()->compute_list_dispatch_threads(compute_list, rb->sdfgi->cascade_size, rb->sdfgi->cascade_size, rb->sdfgi->cascade_size, optimized_jf_group_size, optimized_jf_group_size, optimized_jf_group_size);
RD::get_singleton()->compute_list_add_barrier(compute_list);
jf_us = jf_us == 0 ? 1 : 0;
}
}
}
RENDER_TIMESTAMP("SDFGI Occlusion");
// occlusion
{
uint32_t probe_size = rb->sdfgi->cascade_size / SDFGI::PROBE_DIVISOR;
Vector3i probe_global_pos = rb->sdfgi->cascades[cascade].position / probe_size;
RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, sdfgi_shader.preprocess_pipeline[SDGIShader::PRE_PROCESS_OCCLUSION]);
RD::get_singleton()->compute_list_bind_uniform_set(compute_list, rb->sdfgi->occlusion_uniform_set, 0);
for (int i = 0; i < 8; i++) {
//dispatch all at once for performance
Vector3i offset(i & 1, (i >> 1) & 1, (i >> 2) & 1);
if ((probe_global_pos.x & 1) != 0) {
offset.x = (offset.x + 1) & 1;
}
if ((probe_global_pos.y & 1) != 0) {
offset.y = (offset.y + 1) & 1;
}
if ((probe_global_pos.z & 1) != 0) {
offset.z = (offset.z + 1) & 1;
}
push_constant.probe_offset[0] = offset.x;
push_constant.probe_offset[1] = offset.y;
push_constant.probe_offset[2] = offset.z;
push_constant.occlusion_index = i;
RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(SDGIShader::PreprocessPushConstant));
Fix spelling & grammar in comments, docs, and messages
2020-07-16 16:54:15 +00:00
Vector3i groups = Vector3i(probe_size + 1, probe_size + 1, probe_size + 1) - offset; //if offset, it's one less probe per axis to compute
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
RD::get_singleton()->compute_list_dispatch(compute_list, groups.x, groups.y, groups.z);
}
RD::get_singleton()->compute_list_add_barrier(compute_list);
}
RENDER_TIMESTAMP("SDFGI Store");
// store
RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, sdfgi_shader.preprocess_pipeline[SDGIShader::PRE_PROCESS_STORE]);
RD::get_singleton()->compute_list_bind_uniform_set(compute_list, rb->sdfgi->cascades[cascade].sdf_store_uniform_set, 0);
RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(SDGIShader::PreprocessPushConstant));
RD::get_singleton()->compute_list_dispatch_threads(compute_list, rb->sdfgi->cascade_size, rb->sdfgi->cascade_size, rb->sdfgi->cascade_size, 4, 4, 4);
RD::get_singleton()->compute_list_end();
//clear these textures, as they will have previous garbage on next draw
Reorganize RenderingDevice barriers -Removed sync to draw, now everything syncs to draw by default. -Fixed many validation layer errors. -Added support for VkImageViewUsageCreateInfo to fix validation layer warnings. -Texture, buffer, raster and compute functions now all allow spcifying which barriers will be used.
2021-01-26 00:52:58 +00:00
RD::get_singleton()->texture_clear(rb->sdfgi->cascades[cascade].light_tex, Color(0, 0, 0, 0), 0, 1, 0, 1);
RD::get_singleton()->texture_clear(rb->sdfgi->cascades[cascade].light_aniso_0_tex, Color(0, 0, 0, 0), 0, 1, 0, 1);
RD::get_singleton()->texture_clear(rb->sdfgi->cascades[cascade].light_aniso_1_tex, Color(0, 0, 0, 0), 0, 1, 0, 1);
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
#if 0
Vector<uint8_t> data = RD::get_singleton()->texture_get_data(rb->sdfgi->cascades[cascade].sdf, 0);
Ref<Image> img;
img.instance();
for (uint32_t i = 0; i < rb->sdfgi->cascade_size; i++) {
Vector<uint8_t> subarr = data.subarray(128 * 128 * i, 128 * 128 * (i + 1) - 1);
img->create(rb->sdfgi->cascade_size, rb->sdfgi->cascade_size, false, Image::FORMAT_L8, subarr);
img->save_png("res://cascade_sdf_" + itos(cascade) + "_" + itos(i) + ".png");
}
//finalize render and update sdf
#endif
#if 0
Vector<uint8_t> data = RD::get_singleton()->texture_get_data(rb->sdfgi->render_albedo, 0);
Ref<Image> img;
img.instance();
for (uint32_t i = 0; i < rb->sdfgi->cascade_size; i++) {
Vector<uint8_t> subarr = data.subarray(128 * 128 * i * 2, 128 * 128 * (i + 1) * 2 - 1);
img->create(rb->sdfgi->cascade_size, rb->sdfgi->cascade_size, false, Image::FORMAT_RGB565, subarr);
img->convert(Image::FORMAT_RGBA8);
img->save_png("res://cascade_" + itos(cascade) + "_" + itos(i) + ".png");
}
//finalize render and update sdf
#endif
RENDER_TIMESTAMP("<SDFGI Update SDF");
}
}
Reorganize renderer code. So it can hopefully be made more cache efficient afterwards.
2020-12-31 12:42:56 +00:00
void RendererSceneRenderRD::render_particle_collider_heightfield(RID p_collider, const Transform &p_transform, const PagedArray<GeometryInstance *> &p_instances) {
Implement GPU Particle Collisions -Sphere Attractor -Box Attractor -Vector Field -Sphere Collider -Box Collider -Baked SDF Collider -Heightmap Collider
2020-10-08 00:29:49 +00:00
ERR_FAIL_COND(!storage->particles_collision_is_heightfield(p_collider));
Vector3 extents = storage->particles_collision_get_extents(p_collider) * p_transform.basis.get_scale();
CameraMatrix cm;
cm.set_orthogonal(-extents.x, extents.x, -extents.z, extents.z, 0, extents.y * 2.0);
Vector3 cam_pos = p_transform.origin;
cam_pos.y += extents.y;
Transform cam_xform;
cam_xform.set_look_at(cam_pos, cam_pos - p_transform.basis.get_axis(Vector3::AXIS_Y), -p_transform.basis.get_axis(Vector3::AXIS_Z).normalized());
RID fb = storage->particles_collision_get_heightfield_framebuffer(p_collider);
Replace Octree by DynamicBVH in cull code -Much greater pairing/unpairing performance -For now, using it for culling too, but this will change in a couple of days. -Added a paged allocator, to efficiently alloc/free some types of objects.
2020-12-23 16:52:58 +00:00
_render_particle_collider_heightfield(fb, cam_xform, cm, p_instances);
Implement GPU Particle Collisions -Sphere Attractor -Box Attractor -Vector Field -Sphere Collider -Box Collider -Baked SDF Collider -Heightmap Collider
2020-10-08 00:29:49 +00:00
}
Replace Octree by DynamicBVH in cull code -Much greater pairing/unpairing performance -For now, using it for culling too, but this will change in a couple of days. -Added a paged allocator, to efficiently alloc/free some types of objects.
2020-12-23 16:52:58 +00:00
void RendererSceneRenderRD::render_sdfgi_static_lights(RID p_render_buffers, uint32_t p_cascade_count, const uint32_t *p_cascade_indices, const PagedArray<RID> *p_positional_light_cull_result) {
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers);
ERR_FAIL_COND(!rb);
ERR_FAIL_COND(!rb->sdfgi);
_sdfgi_update_cascades(p_render_buffers); //need cascades updated for this
SDGIShader::Light lights[SDFGI::MAX_STATIC_LIGHTS];
Several GI related optimizations and fixes -SDFGI direct light is done over many frames -SDFGI Changed settings for rays/frame -SDFGI Misc optimizations -SDFGI Bug fix on probe scroll -GIProbe was not working, got it to work again -GIProbe dynamic objects were not working, fixed -Added a half size GI option.
2021-01-22 23:50:24 +00:00
uint32_t light_count[SDFGI::MAX_STATIC_LIGHTS];
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
for (uint32_t i = 0; i < p_cascade_count; i++) {
ERR_CONTINUE(p_cascade_indices[i] >= rb->sdfgi->cascades.size());
SDFGI::Cascade &cc = rb->sdfgi->cascades[p_cascade_indices[i]];
{ //fill light buffer
AABB cascade_aabb;
cascade_aabb.position = Vector3((Vector3i(1, 1, 1) * -int32_t(rb->sdfgi->cascade_size >> 1) + cc.position)) * cc.cell_size;
cascade_aabb.size = Vector3(1, 1, 1) * rb->sdfgi->cascade_size * cc.cell_size;
int idx = 0;
Replace Octree by DynamicBVH in cull code -Much greater pairing/unpairing performance -For now, using it for culling too, but this will change in a couple of days. -Added a paged allocator, to efficiently alloc/free some types of objects.
2020-12-23 16:52:58 +00:00
for (uint32_t j = 0; j < (uint32_t)p_positional_light_cull_result[i].size(); j++) {
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
if (idx == SDFGI::MAX_STATIC_LIGHTS) {
break;
}
LightInstance *li = light_instance_owner.getornull(p_positional_light_cull_result[i][j]);
ERR_CONTINUE(!li);
uint32_t max_sdfgi_cascade = storage->light_get_max_sdfgi_cascade(li->light);
if (p_cascade_indices[i] > max_sdfgi_cascade) {
continue;
}
if (!cascade_aabb.intersects(li->aabb)) {
continue;
}
lights[idx].type = storage->light_get_type(li->light);
Vector3 dir = -li->transform.basis.get_axis(Vector3::AXIS_Z);
if (lights[idx].type == RS::LIGHT_DIRECTIONAL) {
dir.y *= rb->sdfgi->y_mult; //only makes sense for directional
dir.normalize();
}
lights[idx].direction[0] = dir.x;
lights[idx].direction[1] = dir.y;
lights[idx].direction[2] = dir.z;
Vector3 pos = li->transform.origin;
pos.y *= rb->sdfgi->y_mult;
lights[idx].position[0] = pos.x;
lights[idx].position[1] = pos.y;
lights[idx].position[2] = pos.z;
Color color = storage->light_get_color(li->light);
color = color.to_linear();
lights[idx].color[0] = color.r;
lights[idx].color[1] = color.g;
lights[idx].color[2] = color.b;
lights[idx].energy = storage->light_get_param(li->light, RS::LIGHT_PARAM_ENERGY);
lights[idx].has_shadow = storage->light_has_shadow(li->light);
lights[idx].attenuation = storage->light_get_param(li->light, RS::LIGHT_PARAM_ATTENUATION);
lights[idx].radius = storage->light_get_param(li->light, RS::LIGHT_PARAM_RANGE);
lights[idx].spot_angle = Math::deg2rad(storage->light_get_param(li->light, RS::LIGHT_PARAM_SPOT_ANGLE));
lights[idx].spot_attenuation = storage->light_get_param(li->light, RS::LIGHT_PARAM_SPOT_ATTENUATION);
idx++;
}
if (idx > 0) {
Reorganize RenderingDevice barriers -Removed sync to draw, now everything syncs to draw by default. -Fixed many validation layer errors. -Added support for VkImageViewUsageCreateInfo to fix validation layer warnings. -Texture, buffer, raster and compute functions now all allow spcifying which barriers will be used.
2021-01-26 00:52:58 +00:00
RD::get_singleton()->buffer_update(cc.lights_buffer, 0, idx * sizeof(SDGIShader::Light), lights);
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
}
Several GI related optimizations and fixes -SDFGI direct light is done over many frames -SDFGI Changed settings for rays/frame -SDFGI Misc optimizations -SDFGI Bug fix on probe scroll -GIProbe was not working, got it to work again -GIProbe dynamic objects were not working, fixed -Added a half size GI option.
2021-01-22 23:50:24 +00:00
light_count[i] = idx;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
}
Several GI related optimizations and fixes -SDFGI direct light is done over many frames -SDFGI Changed settings for rays/frame -SDFGI Misc optimizations -SDFGI Bug fix on probe scroll -GIProbe was not working, got it to work again -GIProbe dynamic objects were not working, fixed -Added a half size GI option.
2021-01-22 23:50:24 +00:00
}
/* Static Lights */
RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin();
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
Several GI related optimizations and fixes -SDFGI direct light is done over many frames -SDFGI Changed settings for rays/frame -SDFGI Misc optimizations -SDFGI Bug fix on probe scroll -GIProbe was not working, got it to work again -GIProbe dynamic objects were not working, fixed -Added a half size GI option.
2021-01-22 23:50:24 +00:00
RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, sdfgi_shader.direct_light_pipeline[SDGIShader::DIRECT_LIGHT_MODE_STATIC]);
SDGIShader::DirectLightPushConstant dl_push_constant;
dl_push_constant.grid_size[0] = rb->sdfgi->cascade_size;
dl_push_constant.grid_size[1] = rb->sdfgi->cascade_size;
dl_push_constant.grid_size[2] = rb->sdfgi->cascade_size;
dl_push_constant.max_cascades = rb->sdfgi->cascades.size();
dl_push_constant.probe_axis_size = rb->sdfgi->probe_axis_count;
dl_push_constant.multibounce = false; // this is static light, do not multibounce yet
dl_push_constant.y_mult = rb->sdfgi->y_mult;
//all must be processed
dl_push_constant.process_offset = 0;
dl_push_constant.process_increment = 1;
for (uint32_t i = 0; i < p_cascade_count; i++) {
ERR_CONTINUE(p_cascade_indices[i] >= rb->sdfgi->cascades.size());
SDFGI::Cascade &cc = rb->sdfgi->cascades[p_cascade_indices[i]];
dl_push_constant.light_count = light_count[i];
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
dl_push_constant.cascade = p_cascade_indices[i];
if (dl_push_constant.light_count > 0) {
RD::get_singleton()->compute_list_bind_uniform_set(compute_list, cc.sdf_direct_light_uniform_set, 0);
RD::get_singleton()->compute_list_set_push_constant(compute_list, &dl_push_constant, sizeof(SDGIShader::DirectLightPushConstant));
RD::get_singleton()->compute_list_dispatch_indirect(compute_list, cc.solid_cell_dispatch_buffer, 0);
}
}
RD::get_singleton()->compute_list_end();
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
bool RendererSceneRenderRD::free(RID p_rid) {
Base 3D engine done, still untested, though.
2019-08-18 22:40:52 +00:00
if (render_buffers_owner.owns(p_rid)) {
RenderBuffers *rb = render_buffers_owner.getornull(p_rid);
-Refactored post processing, re-added glow and added a mix blend mode.
2020-01-10 00:40:26 +00:00
_free_render_buffer_data(rb);
Base 3D engine done, still untested, though.
2019-08-18 22:40:52 +00:00
memdelete(rb->data);
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
if (rb->sdfgi) {
_sdfgi_erase(rb);
}
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
if (rb->volumetric_fog) {
_volumetric_fog_erase(rb);
}
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
if (rb->cluster_builder) {
memdelete(rb->cluster_builder);
}
Base 3D engine done, still untested, though.
2019-08-18 22:40:52 +00:00
render_buffers_owner.free(p_rid);
Environment sky more or less working.
2019-08-26 20:43:58 +00:00
} else if (environment_owner.owns(p_rid)) {
//not much to delete, just free it
environment_owner.free(p_rid);
WIP CameraEffects implementation (bokeh not working for now)
2020-01-13 18:37:24 +00:00
} else if (camera_effects_owner.owns(p_rid)) {
//not much to delete, just free it
camera_effects_owner.free(p_rid);
Reflection probes working
2019-09-09 20:50:51 +00:00
} else if (reflection_atlas_owner.owns(p_rid)) {
reflection_atlas_set_size(p_rid, 0, 0);
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
ReflectionAtlas *ra = reflection_atlas_owner.getornull(p_rid);
if (ra->cluster_builder) {
memdelete(ra->cluster_builder);
}
Reflection probes working
2019-09-09 20:50:51 +00:00
reflection_atlas_owner.free(p_rid);
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
} else if (reflection_probe_instance_owner.owns(p_rid)) {
//not much to delete, just free it
Several fixes to 3D rendering, and multimesh implementation.
2019-09-14 03:37:42 +00:00
//ReflectionProbeInstance *rpi = reflection_probe_instance_owner.getornull(p_rid);
Reflection probes working
2019-09-09 20:50:51 +00:00
reflection_probe_release_atlas_index(p_rid);
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
reflection_probe_instance_owner.free(p_rid);
Implement decals Also implemented decal atlas, so projectors and other stuff can be added. Sidenote: Had to make RID hashable, so some unrelated includes changed in order to include it in hashfuncs.h
2020-04-14 03:05:21 +00:00
} else if (decal_instance_owner.owns(p_rid)) {
decal_instance_owner.free(p_rid);
Reorganize renderer code. So it can hopefully be made more cache efficient afterwards.
2020-12-31 12:42:56 +00:00
} else if (lightmap_instance_owner.owns(p_rid)) {
lightmap_instance_owner.free(p_rid);
GIProbes working.
2019-10-03 20:39:08 +00:00
} else if (gi_probe_instance_owner.owns(p_rid)) {
GIProbeInstance *gi_probe = gi_probe_instance_owner.getornull(p_rid);
if (gi_probe->texture.is_valid()) {
RD::get_singleton()->free(gi_probe->texture);
RD::get_singleton()->free(gi_probe->write_buffer);
}
Dynamic object support for GI Probes (a bit buggy still)
2019-10-11 02:14:56 +00:00
for (int i = 0; i < gi_probe->dynamic_maps.size(); i++) {
RD::get_singleton()->free(gi_probe->dynamic_maps[i].texture);
RD::get_singleton()->free(gi_probe->dynamic_maps[i].depth);
}
GIProbes working.
2019-10-03 20:39:08 +00:00
gi_probe_instance_owner.free(p_rid);
Environment sky more or less working.
2019-08-26 20:43:58 +00:00
} else if (sky_owner.owns(p_rid)) {
_update_dirty_skys();
Sky *sky = sky_owner.getornull(p_rid);
Working sky shader implementation
2020-03-20 00:32:19 +00:00
Reflection probes working
2019-09-09 20:50:51 +00:00
if (sky->radiance.is_valid()) {
RD::get_singleton()->free(sky->radiance);
sky->radiance = RID();
}
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
_clear_reflection_data(sky->reflection);
Working sky shader implementation
2020-03-20 00:32:19 +00:00
if (sky->uniform_buffer.is_valid()) {
RD::get_singleton()->free(sky->uniform_buffer);
sky->uniform_buffer = RID();
}
if (sky->half_res_pass.is_valid()) {
RD::get_singleton()->free(sky->half_res_pass);
sky->half_res_pass = RID();
}
if (sky->quarter_res_pass.is_valid()) {
RD::get_singleton()->free(sky->quarter_res_pass);
sky->quarter_res_pass = RID();
}
if (sky->material.is_valid()) {
storage->free(sky->material);
}
Environment sky more or less working.
2019-08-26 20:43:58 +00:00
sky_owner.free(p_rid);
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
} else if (light_instance_owner.owns(p_rid)) {
LightInstance *light_instance = light_instance_owner.getornull(p_rid);
//remove from shadow atlases..
for (Set<RID>::Element *E = light_instance->shadow_atlases.front(); E; E = E->next()) {
ShadowAtlas *shadow_atlas = shadow_atlas_owner.getornull(E->get());
ERR_CONTINUE(!shadow_atlas->shadow_owners.has(p_rid));
uint32_t key = shadow_atlas->shadow_owners[p_rid];
uint32_t q = (key >> ShadowAtlas::QUADRANT_SHIFT) & 0x3;
uint32_t s = key & ShadowAtlas::SHADOW_INDEX_MASK;
shadow_atlas->quadrants[q].shadows.write[s].owner = RID();
shadow_atlas->shadow_owners.erase(p_rid);
}
light_instance_owner.free(p_rid);
} else if (shadow_atlas_owner.owns(p_rid)) {
shadow_atlas_set_size(p_rid, 0);
shadow_atlas_owner.free(p_rid);
Base 3D engine done, still untested, though.
2019-08-18 22:40:52 +00:00
} else {
return false;
}
return true;
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
void RendererSceneRenderRD::set_debug_draw_mode(RS::ViewportDebugDraw p_debug_draw) {
-Refactored post processing, re-added glow and added a mix blend mode.
2020-01-10 00:40:26 +00:00
debug_draw = p_debug_draw;
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
void RendererSceneRenderRD::update() {
Environment sky more or less working.
2019-08-26 20:43:58 +00:00
_update_dirty_skys();
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
void RendererSceneRenderRD::set_time(double p_time, double p_step) {
Working sky shader implementation
2020-03-20 00:32:19 +00:00
time = p_time;
Auto exposure re-implemented in Vulkan
2020-01-12 01:26:52 +00:00
time_step = p_step;
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
void RendererSceneRenderRD::screen_space_roughness_limiter_set_active(bool p_enable, float p_amount, float p_limit) {
Added normalmap guided roughness mipmap generator, and a global roughness limiter.
2020-01-26 23:09:40 +00:00
screen_space_roughness_limiter = p_enable;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
screen_space_roughness_limiter_amount = p_amount;
screen_space_roughness_limiter_limit = p_limit;
Added normalmap guided roughness mipmap generator, and a global roughness limiter.
2020-01-26 23:09:40 +00:00
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
bool RendererSceneRenderRD::screen_space_roughness_limiter_is_active() const {
Added normalmap guided roughness mipmap generator, and a global roughness limiter.
2020-01-26 23:09:40 +00:00
return screen_space_roughness_limiter;
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
float RendererSceneRenderRD::screen_space_roughness_limiter_get_amount() const {
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
return screen_space_roughness_limiter_amount;
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
float RendererSceneRenderRD::screen_space_roughness_limiter_get_limit() const {
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
return screen_space_roughness_limiter_limit;
Added normalmap guided roughness mipmap generator, and a global roughness limiter.
2020-01-26 23:09:40 +00:00
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
TypedArray<Image> RendererSceneRenderRD::bake_render_uv2(RID p_base, const Vector<RID> &p_material_overrides, const Size2i &p_image_size) {
New lightmapper -Added LocalVector (needed it) -Added stb_rect_pack (It's pretty cool, we could probably use it for other stuff too) -Fixes and changes all around the place -Added library for 128 bits fixed point (required for Delaunay3D)
2020-05-01 12:34:23 +00:00
RD::TextureFormat tf;
tf.format = RD::DATA_FORMAT_R8G8B8A8_UNORM;
tf.width = p_image_size.width; // Always 64x64
tf.height = p_image_size.height;
tf.usage_bits = RD::TEXTURE_USAGE_COLOR_ATTACHMENT_BIT | RD::TEXTURE_USAGE_CAN_COPY_FROM_BIT;
RID albedo_alpha_tex = RD::get_singleton()->texture_create(tf, RD::TextureView());
RID normal_tex = RD::get_singleton()->texture_create(tf, RD::TextureView());
RID orm_tex = RD::get_singleton()->texture_create(tf, RD::TextureView());
tf.format = RD::DATA_FORMAT_R16G16B16A16_SFLOAT;
RID emission_tex = RD::get_singleton()->texture_create(tf, RD::TextureView());
tf.format = RD::DATA_FORMAT_R32_SFLOAT;
RID depth_write_tex = RD::get_singleton()->texture_create(tf, RD::TextureView());
tf.usage_bits = RD::TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT | RD::TEXTURE_USAGE_CAN_COPY_FROM_BIT;
tf.format = RD::get_singleton()->texture_is_format_supported_for_usage(RD::DATA_FORMAT_D32_SFLOAT, RD::TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT) ? RD::DATA_FORMAT_D32_SFLOAT : RD::DATA_FORMAT_X8_D24_UNORM_PACK32;
RID depth_tex = RD::get_singleton()->texture_create(tf, RD::TextureView());
Vector<RID> fb_tex;
fb_tex.push_back(albedo_alpha_tex);
fb_tex.push_back(normal_tex);
fb_tex.push_back(orm_tex);
fb_tex.push_back(emission_tex);
fb_tex.push_back(depth_write_tex);
fb_tex.push_back(depth_tex);
RID fb = RD::get_singleton()->framebuffer_create(fb_tex);
//RID sampled_light;
Reorganize renderer code. So it can hopefully be made more cache efficient afterwards.
2020-12-31 12:42:56 +00:00
GeometryInstance *gi = geometry_instance_create(p_base);
uint32_t sc = RSG::storage->mesh_get_surface_count(p_base);
Vector<RID> materials;
materials.resize(sc);
New lightmapper -Added LocalVector (needed it) -Added stb_rect_pack (It's pretty cool, we could probably use it for other stuff too) -Fixes and changes all around the place -Added library for 128 bits fixed point (required for Delaunay3D)
2020-05-01 12:34:23 +00:00
Reorganize renderer code. So it can hopefully be made more cache efficient afterwards.
2020-12-31 12:42:56 +00:00
for (uint32_t i = 0; i < sc; i++) {
if (i < (uint32_t)p_material_overrides.size()) {
materials.write[i] = p_material_overrides[i];
New lightmapper -Added LocalVector (needed it) -Added stb_rect_pack (It's pretty cool, we could probably use it for other stuff too) -Fixes and changes all around the place -Added library for 128 bits fixed point (required for Delaunay3D)
2020-05-01 12:34:23 +00:00
}
}
Reorganize renderer code. So it can hopefully be made more cache efficient afterwards.
2020-12-31 12:42:56 +00:00
geometry_instance_set_surface_materials(gi, materials);
Replace Octree by DynamicBVH in cull code -Much greater pairing/unpairing performance -For now, using it for culling too, but this will change in a couple of days. -Added a paged allocator, to efficiently alloc/free some types of objects.
2020-12-23 16:52:58 +00:00
if (cull_argument.size() == 0) {
cull_argument.push_back(nullptr);
}
Reorganize renderer code. So it can hopefully be made more cache efficient afterwards.
2020-12-31 12:42:56 +00:00
cull_argument[0] = gi;
Replace Octree by DynamicBVH in cull code -Much greater pairing/unpairing performance -For now, using it for culling too, but this will change in a couple of days. -Added a paged allocator, to efficiently alloc/free some types of objects.
2020-12-23 16:52:58 +00:00
_render_uv2(cull_argument, fb, Rect2i(0, 0, p_image_size.width, p_image_size.height));
New lightmapper -Added LocalVector (needed it) -Added stb_rect_pack (It's pretty cool, we could probably use it for other stuff too) -Fixes and changes all around the place -Added library for 128 bits fixed point (required for Delaunay3D)
2020-05-01 12:34:23 +00:00
Reorganize renderer code. So it can hopefully be made more cache efficient afterwards.
2020-12-31 12:42:56 +00:00
geometry_instance_free(gi);
New lightmapper -Added LocalVector (needed it) -Added stb_rect_pack (It's pretty cool, we could probably use it for other stuff too) -Fixes and changes all around the place -Added library for 128 bits fixed point (required for Delaunay3D)
2020-05-01 12:34:23 +00:00
TypedArray<Image> ret;
{
PackedByteArray data = RD::get_singleton()->texture_get_data(albedo_alpha_tex, 0);
Ref<Image> img;
img.instance();
img->create(p_image_size.width, p_image_size.height, false, Image::FORMAT_RGBA8, data);
RD::get_singleton()->free(albedo_alpha_tex);
ret.push_back(img);
}
{
PackedByteArray data = RD::get_singleton()->texture_get_data(normal_tex, 0);
Ref<Image> img;
img.instance();
img->create(p_image_size.width, p_image_size.height, false, Image::FORMAT_RGBA8, data);
RD::get_singleton()->free(normal_tex);
ret.push_back(img);
}
{
PackedByteArray data = RD::get_singleton()->texture_get_data(orm_tex, 0);
Ref<Image> img;
img.instance();
img->create(p_image_size.width, p_image_size.height, false, Image::FORMAT_RGBA8, data);
RD::get_singleton()->free(orm_tex);
ret.push_back(img);
}
{
PackedByteArray data = RD::get_singleton()->texture_get_data(emission_tex, 0);
Ref<Image> img;
img.instance();
img->create(p_image_size.width, p_image_size.height, false, Image::FORMAT_RGBAH, data);
RD::get_singleton()->free(emission_tex);
ret.push_back(img);
}
RD::get_singleton()->free(depth_write_tex);
RD::get_singleton()->free(depth_tex);
return ret;
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
void RendererSceneRenderRD::sdfgi_set_debug_probe_select(const Vector3 &p_position, const Vector3 &p_dir) {
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
sdfgi_debug_probe_pos = p_position;
sdfgi_debug_probe_dir = p_dir;
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
RendererSceneRenderRD *RendererSceneRenderRD::singleton = nullptr;
Working sky shader implementation
2020-03-20 00:32:19 +00:00
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
RID RendererSceneRenderRD::get_reflection_probe_buffer() {
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
return cluster.reflection_buffer;
}
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
RID RendererSceneRenderRD::get_omni_light_buffer() {
return cluster.omni_light_buffer;
}
RID RendererSceneRenderRD::get_spot_light_buffer() {
return cluster.spot_light_buffer;
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
}
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
RID RendererSceneRenderRD::get_directional_light_buffer() {
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
return cluster.directional_light_buffer;
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
RID RendererSceneRenderRD::get_decal_buffer() {
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
return cluster.decal_buffer;
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
int RendererSceneRenderRD::get_max_directional_lights() const {
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
return cluster.max_directional_lights;
}
Add support for low-end 3D rendering. -Reduce number of uniform sets from 6 to 4. -Remove features in low end mode, in order to reduce the number of texture units fit to 16.
2020-12-07 21:27:38 +00:00
bool RendererSceneRenderRD::is_low_end() const {
return low_end;
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
RendererSceneRenderRD::RendererSceneRenderRD(RendererStorageRD *p_storage) {
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
max_cluster_elements = GLOBAL_GET("rendering/cluster_builder/max_clustered_elements");
Environment sky more or less working.
2019-08-26 20:43:58 +00:00
storage = p_storage;
Working sky shader implementation
2020-03-20 00:32:19 +00:00
singleton = this;
Environment sky more or less working.
2019-08-26 20:43:58 +00:00
roughness_layers = GLOBAL_GET("rendering/quality/reflections/roughness_layers");
sky_ggx_samples_quality = GLOBAL_GET("rendering/quality/reflections/ggx_samples");
sky_use_cubemap_array = GLOBAL_GET("rendering/quality/reflections/texture_array_reflections");
Several GI related optimizations and fixes -SDFGI direct light is done over many frames -SDFGI Changed settings for rays/frame -SDFGI Misc optimizations -SDFGI Bug fix on probe scroll -GIProbe was not working, got it to work again -GIProbe dynamic objects were not working, fixed -Added a half size GI option.
2021-01-22 23:50:24 +00:00
sdfgi_ray_count = RS::EnvironmentSDFGIRayCount(CLAMP(int32_t(GLOBAL_GET("rendering/sdfgi/probe_ray_count")), 0, int32_t(RS::ENV_SDFGI_RAY_COUNT_MAX - 1)));
sdfgi_frames_to_converge = RS::EnvironmentSDFGIFramesToConverge(CLAMP(int32_t(GLOBAL_GET("rendering/sdfgi/frames_to_converge")), 0, int32_t(RS::ENV_SDFGI_CONVERGE_MAX - 1)));
sdfgi_frames_to_update_light = RS::EnvironmentSDFGIFramesToUpdateLight(CLAMP(int32_t(GLOBAL_GET("rendering/sdfgi/frames_to_update_lights")), 0, int32_t(RS::ENV_SDFGI_UPDATE_LIGHT_MAX - 1)));
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
Shadow map rendering optimization -All shadow rendering is done with raster now (no compute) -All shadow rendering is done by rendering directly to the shadow atlas -Improved how buffer clearing is done to optimize the above. -Ability to set shadows as 16 bits.
2021-01-24 19:00:20 +00:00
directional_shadow.size = GLOBAL_GET("rendering/quality/directional_shadow/size");
directional_shadow.use_16_bits = GLOBAL_GET("rendering/quality/directional_shadow/16_bits");
Add support for low-end 3D rendering. -Reduce number of uniform sets from 6 to 4. -Remove features in low end mode, in order to reduce the number of texture units fit to 16.
2020-12-07 21:27:38 +00:00
uint32_t textures_per_stage = RD::get_singleton()->limit_get(RD::LIMIT_MAX_TEXTURES_PER_SHADER_STAGE);
GIProbes working.
2019-10-03 20:39:08 +00:00
Add support for low-end 3D rendering. -Reduce number of uniform sets from 6 to 4. -Remove features in low end mode, in order to reduce the number of texture units fit to 16.
2020-12-07 21:27:38 +00:00
low_end = GLOBAL_GET("rendering/quality/rd_renderer/use_low_end_renderer");
if (textures_per_stage < 48) {
low_end = true;
}
if (!low_end) {
GIProbes working.
2019-10-03 20:39:08 +00:00
//kinda complicated to compute the amount of slots, we try to use as many as we can
gi_probe_max_lights = 32;
gi_probe_lights = memnew_arr(GIProbeLight, gi_probe_max_lights);
gi_probe_lights_uniform = RD::get_singleton()->uniform_buffer_create(gi_probe_max_lights * sizeof(GIProbeLight));
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
gi_probe_quality = RS::GIProbeQuality(CLAMP(int(GLOBAL_GET("rendering/quality/gi_probes/quality")), 0, 1));
GIProbes working.
2019-10-03 20:39:08 +00:00
String defines = "\n#define MAX_LIGHTS " + itos(gi_probe_max_lights) + "\n";
Vector<String> versions;
versions.push_back("\n#define MODE_COMPUTE_LIGHT\n");
versions.push_back("\n#define MODE_SECOND_BOUNCE\n");
versions.push_back("\n#define MODE_UPDATE_MIPMAPS\n");
versions.push_back("\n#define MODE_WRITE_TEXTURE\n");
Dynamic object support for GI Probes (a bit buggy still)
2019-10-11 02:14:56 +00:00
versions.push_back("\n#define MODE_DYNAMIC\n#define MODE_DYNAMIC_LIGHTING\n");
versions.push_back("\n#define MODE_DYNAMIC\n#define MODE_DYNAMIC_SHRINK\n#define MODE_DYNAMIC_SHRINK_WRITE\n");
versions.push_back("\n#define MODE_DYNAMIC\n#define MODE_DYNAMIC_SHRINK\n#define MODE_DYNAMIC_SHRINK_PLOT\n");
versions.push_back("\n#define MODE_DYNAMIC\n#define MODE_DYNAMIC_SHRINK\n#define MODE_DYNAMIC_SHRINK_PLOT\n#define MODE_DYNAMIC_SHRINK_WRITE\n");
GIProbes working.
2019-10-03 20:39:08 +00:00
giprobe_shader.initialize(versions, defines);
giprobe_lighting_shader_version = giprobe_shader.version_create();
for (int i = 0; i < GI_PROBE_SHADER_VERSION_MAX; i++) {
giprobe_lighting_shader_version_shaders[i] = giprobe_shader.version_get_shader(giprobe_lighting_shader_version, i);
giprobe_lighting_shader_version_pipelines[i] = RD::get_singleton()->compute_pipeline_create(giprobe_lighting_shader_version_shaders[i]);
}
}
Add support for low-end 3D rendering. -Reduce number of uniform sets from 6 to 4. -Remove features in low end mode, in order to reduce the number of texture units fit to 16.
2020-12-07 21:27:38 +00:00
if (!low_end) {
GIProbes working.
2019-10-03 20:39:08 +00:00
String defines;
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
Vector<String> versions;
GIProbes working.
2019-10-03 20:39:08 +00:00
versions.push_back("\n#define MODE_DEBUG_COLOR\n");
versions.push_back("\n#define MODE_DEBUG_LIGHT\n");
More GIProbe work and fixes
2019-10-05 13:27:43 +00:00
versions.push_back("\n#define MODE_DEBUG_EMISSION\n");
Dynamic object support for GI Probes (a bit buggy still)
2019-10-11 02:14:56 +00:00
versions.push_back("\n#define MODE_DEBUG_LIGHT\n#define MODE_DEBUG_LIGHT_FULL\n");
GIProbes working.
2019-10-03 20:39:08 +00:00
giprobe_debug_shader.initialize(versions, defines);
giprobe_debug_shader_version = giprobe_debug_shader.version_create();
for (int i = 0; i < GI_PROBE_DEBUG_MAX; i++) {
giprobe_debug_shader_version_shaders[i] = giprobe_debug_shader.version_get_shader(giprobe_debug_shader_version, i);
RD::PipelineRasterizationState rs;
rs.cull_mode = RD::POLYGON_CULL_FRONT;
RD::PipelineDepthStencilState ds;
ds.enable_depth_test = true;
ds.enable_depth_write = true;
ds.depth_compare_operator = RD::COMPARE_OP_LESS_OR_EQUAL;
giprobe_debug_shader_version_pipelines[i].setup(giprobe_debug_shader_version_shaders[i], RD::RENDER_PRIMITIVE_TRIANGLES, rs, RD::PipelineMultisampleState(), ds, RD::PipelineColorBlendState::create_disabled(), 0);
}
Untested support for compute shaders
2019-09-25 19:44:44 +00:00
}
DOF fully implemented, can be edited on the fly.
2020-01-16 00:23:21 +00:00
Working sky shader implementation
2020-03-20 00:32:19 +00:00
/* SKY SHADER */
{
// Start with the directional lights for the sky
sky_scene_state.max_directional_lights = 4;
uint32_t directional_light_buffer_size = sky_scene_state.max_directional_lights * sizeof(SkyDirectionalLightData);
sky_scene_state.directional_lights = memnew_arr(SkyDirectionalLightData, sky_scene_state.max_directional_lights);
sky_scene_state.last_frame_directional_lights = memnew_arr(SkyDirectionalLightData, sky_scene_state.max_directional_lights);
sky_scene_state.last_frame_directional_light_count = sky_scene_state.max_directional_lights + 1;
sky_scene_state.directional_light_buffer = RD::get_singleton()->uniform_buffer_create(directional_light_buffer_size);
String defines = "\n#define MAX_DIRECTIONAL_LIGHT_DATA_STRUCTS " + itos(sky_scene_state.max_directional_lights) + "\n";
// Initialize sky
Vector<String> sky_modes;
Replace subpass textures with color in sky shader
2020-03-24 05:55:11 +00:00
sky_modes.push_back(""); // Full size
sky_modes.push_back("\n#define USE_HALF_RES_PASS\n"); // Half Res
sky_modes.push_back("\n#define USE_QUARTER_RES_PASS\n"); // Quarter res
sky_modes.push_back("\n#define USE_CUBEMAP_PASS\n"); // Cubemap
sky_modes.push_back("\n#define USE_CUBEMAP_PASS\n#define USE_HALF_RES_PASS\n"); // Half Res Cubemap
sky_modes.push_back("\n#define USE_CUBEMAP_PASS\n#define USE_QUARTER_RES_PASS\n"); // Quarter res Cubemap
Working sky shader implementation
2020-03-20 00:32:19 +00:00
sky_shader.shader.initialize(sky_modes, defines);
}
// register our shader funds
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
storage->shader_set_data_request_function(RendererStorageRD::SHADER_TYPE_SKY, _create_sky_shader_funcs);
storage->material_set_data_request_function(RendererStorageRD::SHADER_TYPE_SKY, _create_sky_material_funcs);
Working sky shader implementation
2020-03-20 00:32:19 +00:00
{
ShaderCompilerRD::DefaultIdentifierActions actions;
actions.renames["COLOR"] = "color";
actions.renames["ALPHA"] = "alpha";
actions.renames["EYEDIR"] = "cube_normal";
actions.renames["POSITION"] = "params.position_multiplier.xyz";
actions.renames["SKY_COORDS"] = "panorama_coords";
actions.renames["SCREEN_UV"] = "uv";
actions.renames["TIME"] = "params.time";
Replace subpass textures with color in sky shader
2020-03-24 05:55:11 +00:00
actions.renames["HALF_RES_COLOR"] = "half_res_color";
actions.renames["QUARTER_RES_COLOR"] = "quarter_res_color";
Working sky shader implementation
2020-03-20 00:32:19 +00:00
actions.renames["RADIANCE"] = "radiance";
Add FOG, RADIANCE, and IRRADIANCE shader overrides
2020-08-21 03:24:46 +00:00
actions.renames["FOG"] = "custom_fog";
Working sky shader implementation
2020-03-20 00:32:19 +00:00
actions.renames["LIGHT0_ENABLED"] = "directional_lights.data[0].enabled";
Add light size to Sky Shaders
2020-04-17 06:38:22 +00:00
actions.renames["LIGHT0_DIRECTION"] = "directional_lights.data[0].direction_energy.xyz";
actions.renames["LIGHT0_ENERGY"] = "directional_lights.data[0].direction_energy.w";
actions.renames["LIGHT0_COLOR"] = "directional_lights.data[0].color_size.xyz";
actions.renames["LIGHT0_SIZE"] = "directional_lights.data[0].color_size.w";
Working sky shader implementation
2020-03-20 00:32:19 +00:00
actions.renames["LIGHT1_ENABLED"] = "directional_lights.data[1].enabled";
Add light size to Sky Shaders
2020-04-17 06:38:22 +00:00
actions.renames["LIGHT1_DIRECTION"] = "directional_lights.data[1].direction_energy.xyz";
actions.renames["LIGHT1_ENERGY"] = "directional_lights.data[1].direction_energy.w";
actions.renames["LIGHT1_COLOR"] = "directional_lights.data[1].color_size.xyz";
actions.renames["LIGHT1_SIZE"] = "directional_lights.data[1].color_size.w";
Working sky shader implementation
2020-03-20 00:32:19 +00:00
actions.renames["LIGHT2_ENABLED"] = "directional_lights.data[2].enabled";
Add light size to Sky Shaders
2020-04-17 06:38:22 +00:00
actions.renames["LIGHT2_DIRECTION"] = "directional_lights.data[2].direction_energy.xyz";
actions.renames["LIGHT2_ENERGY"] = "directional_lights.data[2].direction_energy.w";
actions.renames["LIGHT2_COLOR"] = "directional_lights.data[2].color_size.xyz";
actions.renames["LIGHT2_SIZE"] = "directional_lights.data[2].color_size.w";
Working sky shader implementation
2020-03-20 00:32:19 +00:00
actions.renames["LIGHT3_ENABLED"] = "directional_lights.data[3].enabled";
Add light size to Sky Shaders
2020-04-17 06:38:22 +00:00
actions.renames["LIGHT3_DIRECTION"] = "directional_lights.data[3].direction_energy.xyz";
actions.renames["LIGHT3_ENERGY"] = "directional_lights.data[3].direction_energy.w";
actions.renames["LIGHT3_COLOR"] = "directional_lights.data[3].color_size.xyz";
actions.renames["LIGHT3_SIZE"] = "directional_lights.data[3].color_size.w";
Working sky shader implementation
2020-03-20 00:32:19 +00:00
actions.renames["AT_CUBEMAP_PASS"] = "AT_CUBEMAP_PASS";
actions.renames["AT_HALF_RES_PASS"] = "AT_HALF_RES_PASS";
actions.renames["AT_QUARTER_RES_PASS"] = "AT_QUARTER_RES_PASS";
actions.custom_samplers["RADIANCE"] = "material_samplers[3]";
Replace subpass textures with color in sky shader
2020-03-24 05:55:11 +00:00
actions.usage_defines["HALF_RES_COLOR"] = "\n#define USES_HALF_RES_COLOR\n";
actions.usage_defines["QUARTER_RES_COLOR"] = "\n#define USES_QUARTER_RES_COLOR\n";
Add fog to sky shaders
2020-08-18 04:12:51 +00:00
actions.render_mode_defines["disable_fog"] = "#define DISABLE_FOG\n";
Working sky shader implementation
2020-03-20 00:32:19 +00:00
actions.sampler_array_name = "material_samplers";
actions.base_texture_binding_index = 1;
actions.texture_layout_set = 1;
actions.base_uniform_string = "material.";
actions.base_varying_index = 10;
actions.default_filter = ShaderLanguage::FILTER_LINEAR_MIPMAP;
actions.default_repeat = ShaderLanguage::REPEAT_ENABLE;
Implement global and per instance shader uniforms. Adds two keywords to shader language for uniforms: -'global' -'instance' This allows them to reference values outside the material.
2020-04-17 02:52:00 +00:00
actions.global_buffer_array_variable = "global_variables.data";
Working sky shader implementation
2020-03-20 00:32:19 +00:00
sky_shader.compiler.initialize(actions);
}
{
// default material and shader for sky shader
sky_shader.default_shader = storage->shader_create();
Add light size to Sky Shaders
2020-04-17 06:38:22 +00:00
storage->shader_set_code(sky_shader.default_shader, "shader_type sky; void fragment() { COLOR = vec3(0.0); } \n");
Working sky shader implementation
2020-03-20 00:32:19 +00:00
sky_shader.default_material = storage->material_create();
storage->material_set_shader(sky_shader.default_material, sky_shader.default_shader);
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
SkyMaterialData *md = (SkyMaterialData *)storage->material_get_data(sky_shader.default_material, RendererStorageRD::SHADER_TYPE_SKY);
Working sky shader implementation
2020-03-20 00:32:19 +00:00
sky_shader.default_shader_rd = sky_shader.shader.version_get_shader(md->shader_data->version, SKY_VERSION_BACKGROUND);
Add fog to sky shaders
2020-08-18 04:12:51 +00:00
sky_scene_state.uniform_buffer = RD::get_singleton()->uniform_buffer_create(sizeof(SkySceneState::UBO));
Working sky shader implementation
2020-03-20 00:32:19 +00:00
Vector<RD::Uniform> uniforms;
{
RD::Uniform u;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_SAMPLER;
Working sky shader implementation
2020-03-20 00:32:19 +00:00
u.binding = 0;
u.ids.resize(12);
RID *ids_ptr = u.ids.ptrw();
Renaming of servers for coherency. VisualServer -> RenderingServer PhysicsServer -> PhysicsServer3D Physics2DServer -> PhysicsServer2D NavigationServer -> NavigationServer3D Navigation2DServer -> NavigationServer2D Also renamed corresponding files.
2020-03-27 18:21:27 +00:00
ids_ptr[0] = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED);
ids_ptr[1] = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED);
ids_ptr[2] = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST_WITH_MIPMAPS, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED);
ids_ptr[3] = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED);
ids_ptr[4] = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST_WITH_MIPMAPS_ANISOTROPIC, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED);
ids_ptr[5] = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS_ANISOTROPIC, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED);
ids_ptr[6] = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST, RS::CANVAS_ITEM_TEXTURE_REPEAT_ENABLED);
ids_ptr[7] = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR, RS::CANVAS_ITEM_TEXTURE_REPEAT_ENABLED);
ids_ptr[8] = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST_WITH_MIPMAPS, RS::CANVAS_ITEM_TEXTURE_REPEAT_ENABLED);
ids_ptr[9] = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS, RS::CANVAS_ITEM_TEXTURE_REPEAT_ENABLED);
ids_ptr[10] = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST_WITH_MIPMAPS_ANISOTROPIC, RS::CANVAS_ITEM_TEXTURE_REPEAT_ENABLED);
ids_ptr[11] = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS_ANISOTROPIC, RS::CANVAS_ITEM_TEXTURE_REPEAT_ENABLED);
Working sky shader implementation
2020-03-20 00:32:19 +00:00
uniforms.push_back(u);
}
Implement global and per instance shader uniforms. Adds two keywords to shader language for uniforms: -'global' -'instance' This allows them to reference values outside the material.
2020-04-17 02:52:00 +00:00
{
RD::Uniform u;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER;
Implement global and per instance shader uniforms. Adds two keywords to shader language for uniforms: -'global' -'instance' This allows them to reference values outside the material.
2020-04-17 02:52:00 +00:00
u.binding = 1;
u.ids.push_back(storage->global_variables_get_storage_buffer());
uniforms.push_back(u);
}
Add fog to sky shaders
2020-08-18 04:12:51 +00:00
{
RD::Uniform u;
u.binding = 2;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER;
Add fog to sky shaders
2020-08-18 04:12:51 +00:00
u.ids.push_back(sky_scene_state.uniform_buffer);
uniforms.push_back(u);
}
{
RD::Uniform u;
u.binding = 3;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER;
Add fog to sky shaders
2020-08-18 04:12:51 +00:00
u.ids.push_back(sky_scene_state.directional_light_buffer);
uniforms.push_back(u);
}
sky_scene_state.uniform_set = RD::get_singleton()->uniform_set_create(uniforms, sky_shader.default_shader_rd, SKY_SET_UNIFORMS);
}
{
Vector<RD::Uniform> uniforms;
{
RD::Uniform u;
u.binding = 0;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_TEXTURE;
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
RID vfog = storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_3D_WHITE);
Add fog to sky shaders
2020-08-18 04:12:51 +00:00
u.ids.push_back(vfog);
uniforms.push_back(u);
}
sky_scene_state.default_fog_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, sky_shader.default_shader_rd, SKY_SET_FOG);
}
{
// Need defaults for using fog with clear color
sky_scene_state.fog_shader = storage->shader_create();
storage->shader_set_code(sky_scene_state.fog_shader, "shader_type sky; uniform vec4 clear_color; void fragment() { COLOR = clear_color.rgb; } \n");
sky_scene_state.fog_material = storage->material_create();
storage->material_set_shader(sky_scene_state.fog_material, sky_scene_state.fog_shader);
Vector<RD::Uniform> uniforms;
{
RD::Uniform u;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_TEXTURE;
Add fog to sky shaders
2020-08-18 04:12:51 +00:00
u.binding = 0;
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
u.ids.push_back(storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_CUBEMAP_BLACK));
Add fog to sky shaders
2020-08-18 04:12:51 +00:00
uniforms.push_back(u);
}
{
RD::Uniform u;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_TEXTURE;
Add fog to sky shaders
2020-08-18 04:12:51 +00:00
u.binding = 1;
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
u.ids.push_back(storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_WHITE));
Add fog to sky shaders
2020-08-18 04:12:51 +00:00
uniforms.push_back(u);
}
{
RD::Uniform u;
Rename RD uniform "type" to "uniform_type" "type" isn't very descriptive for exposed code, and this variable is exposed via RD_SETGET.
2020-10-17 01:19:21 +00:00
u.uniform_type = RD::UNIFORM_TYPE_TEXTURE;
Add fog to sky shaders
2020-08-18 04:12:51 +00:00
u.binding = 2;
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
u.ids.push_back(storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_WHITE));
Add fog to sky shaders
2020-08-18 04:12:51 +00:00
uniforms.push_back(u);
}
sky_scene_state.fog_only_texture_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, sky_shader.default_shader_rd, SKY_SET_TEXTURES);
Working sky shader implementation
2020-03-20 00:32:19 +00:00
}
Add support for low-end 3D rendering. -Reduce number of uniform sets from 6 to 4. -Remove features in low end mode, in order to reduce the number of texture units fit to 16.
2020-12-07 21:27:38 +00:00
if (!low_end) {
//SDFGI
{
Vector<String> preprocess_modes;
preprocess_modes.push_back("\n#define MODE_SCROLL\n");
preprocess_modes.push_back("\n#define MODE_SCROLL_OCCLUSION\n");
preprocess_modes.push_back("\n#define MODE_INITIALIZE_JUMP_FLOOD\n");
preprocess_modes.push_back("\n#define MODE_INITIALIZE_JUMP_FLOOD_HALF\n");
preprocess_modes.push_back("\n#define MODE_JUMPFLOOD\n");
preprocess_modes.push_back("\n#define MODE_JUMPFLOOD_OPTIMIZED\n");
preprocess_modes.push_back("\n#define MODE_UPSCALE_JUMP_FLOOD\n");
preprocess_modes.push_back("\n#define MODE_OCCLUSION\n");
preprocess_modes.push_back("\n#define MODE_STORE\n");
String defines = "\n#define OCCLUSION_SIZE " + itos(SDFGI::CASCADE_SIZE / SDFGI::PROBE_DIVISOR) + "\n";
sdfgi_shader.preprocess.initialize(preprocess_modes, defines);
sdfgi_shader.preprocess_shader = sdfgi_shader.preprocess.version_create();
for (int i = 0; i < SDGIShader::PRE_PROCESS_MAX; i++) {
sdfgi_shader.preprocess_pipeline[i] = RD::get_singleton()->compute_pipeline_create(sdfgi_shader.preprocess.version_get_shader(sdfgi_shader.preprocess_shader, i));
}
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
}
Add support for low-end 3D rendering. -Reduce number of uniform sets from 6 to 4. -Remove features in low end mode, in order to reduce the number of texture units fit to 16.
2020-12-07 21:27:38 +00:00
{
//calculate tables
String defines = "\n#define OCT_SIZE " + itos(SDFGI::LIGHTPROBE_OCT_SIZE) + "\n";
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
Add support for low-end 3D rendering. -Reduce number of uniform sets from 6 to 4. -Remove features in low end mode, in order to reduce the number of texture units fit to 16.
2020-12-07 21:27:38 +00:00
Vector<String> direct_light_modes;
direct_light_modes.push_back("\n#define MODE_PROCESS_STATIC\n");
direct_light_modes.push_back("\n#define MODE_PROCESS_DYNAMIC\n");
sdfgi_shader.direct_light.initialize(direct_light_modes, defines);
sdfgi_shader.direct_light_shader = sdfgi_shader.direct_light.version_create();
for (int i = 0; i < SDGIShader::DIRECT_LIGHT_MODE_MAX; i++) {
sdfgi_shader.direct_light_pipeline[i] = RD::get_singleton()->compute_pipeline_create(sdfgi_shader.direct_light.version_get_shader(sdfgi_shader.direct_light_shader, i));
}
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
}
Add support for low-end 3D rendering. -Reduce number of uniform sets from 6 to 4. -Remove features in low end mode, in order to reduce the number of texture units fit to 16.
2020-12-07 21:27:38 +00:00
{
//calculate tables
String defines = "\n#define OCT_SIZE " + itos(SDFGI::LIGHTPROBE_OCT_SIZE) + "\n";
defines += "\n#define SH_SIZE " + itos(SDFGI::SH_SIZE) + "\n";
Reorganize RenderingDevice barriers -Removed sync to draw, now everything syncs to draw by default. -Fixed many validation layer errors. -Added support for VkImageViewUsageCreateInfo to fix validation layer warnings. -Texture, buffer, raster and compute functions now all allow spcifying which barriers will be used.
2021-01-26 00:52:58 +00:00
if (sky_use_cubemap_array) {
defines += "\n#define USE_CUBEMAP_ARRAY\n";
}
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
Add support for low-end 3D rendering. -Reduce number of uniform sets from 6 to 4. -Remove features in low end mode, in order to reduce the number of texture units fit to 16.
2020-12-07 21:27:38 +00:00
Vector<String> integrate_modes;
integrate_modes.push_back("\n#define MODE_PROCESS\n");
integrate_modes.push_back("\n#define MODE_STORE\n");
integrate_modes.push_back("\n#define MODE_SCROLL\n");
integrate_modes.push_back("\n#define MODE_SCROLL_STORE\n");
sdfgi_shader.integrate.initialize(integrate_modes, defines);
sdfgi_shader.integrate_shader = sdfgi_shader.integrate.version_create();
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
Add support for low-end 3D rendering. -Reduce number of uniform sets from 6 to 4. -Remove features in low end mode, in order to reduce the number of texture units fit to 16.
2020-12-07 21:27:38 +00:00
for (int i = 0; i < SDGIShader::INTEGRATE_MODE_MAX; i++) {
sdfgi_shader.integrate_pipeline[i] = RD::get_singleton()->compute_pipeline_create(sdfgi_shader.integrate.version_get_shader(sdfgi_shader.integrate_shader, i));
}
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
{
Add support for low-end 3D rendering. -Reduce number of uniform sets from 6 to 4. -Remove features in low end mode, in order to reduce the number of texture units fit to 16.
2020-12-07 21:27:38 +00:00
Vector<RD::Uniform> uniforms;
{
RD::Uniform u;
u.uniform_type = RD::UNIFORM_TYPE_TEXTURE;
u.binding = 0;
u.ids.push_back(storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_CUBEMAP_WHITE));
uniforms.push_back(u);
}
{
RD::Uniform u;
u.uniform_type = RD::UNIFORM_TYPE_SAMPLER;
u.binding = 1;
u.ids.push_back(storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED));
uniforms.push_back(u);
}
sdfgi_shader.integrate_default_sky_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, sdfgi_shader.integrate.version_get_shader(sdfgi_shader.integrate_shader, 0), 1);
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
}
Add support for low-end 3D rendering. -Reduce number of uniform sets from 6 to 4. -Remove features in low end mode, in order to reduce the number of texture units fit to 16.
2020-12-07 21:27:38 +00:00
}
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
//GK
Add support for low-end 3D rendering. -Reduce number of uniform sets from 6 to 4. -Remove features in low end mode, in order to reduce the number of texture units fit to 16.
2020-12-07 21:27:38 +00:00
{
//calculate tables
String defines = "\n#define SDFGI_OCT_SIZE " + itos(SDFGI::LIGHTPROBE_OCT_SIZE) + "\n";
Vector<String> gi_modes;
Several GI related optimizations and fixes -SDFGI direct light is done over many frames -SDFGI Changed settings for rays/frame -SDFGI Misc optimizations -SDFGI Bug fix on probe scroll -GIProbe was not working, got it to work again -GIProbe dynamic objects were not working, fixed -Added a half size GI option.
2021-01-22 23:50:24 +00:00
gi_modes.push_back("\n#define USE_GIPROBES\n");
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
gi_modes.push_back("\n#define USE_SDFGI\n");
Several GI related optimizations and fixes -SDFGI direct light is done over many frames -SDFGI Changed settings for rays/frame -SDFGI Misc optimizations -SDFGI Bug fix on probe scroll -GIProbe was not working, got it to work again -GIProbe dynamic objects were not working, fixed -Added a half size GI option.
2021-01-22 23:50:24 +00:00
gi_modes.push_back("\n#define USE_SDFGI\n\n#define USE_GIPROBES\n");
gi_modes.push_back("\n#define MODE_HALF_RES\n#define USE_GIPROBES\n");
gi_modes.push_back("\n#define MODE_HALF_RES\n#define USE_SDFGI\n");
gi_modes.push_back("\n#define MODE_HALF_RES\n#define USE_SDFGI\n\n#define USE_GIPROBES\n");
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
Add support for low-end 3D rendering. -Reduce number of uniform sets from 6 to 4. -Remove features in low end mode, in order to reduce the number of texture units fit to 16.
2020-12-07 21:27:38 +00:00
gi.shader.initialize(gi_modes, defines);
gi.shader_version = gi.shader.version_create();
for (int i = 0; i < GI::MODE_MAX; i++) {
gi.pipelines[i] = RD::get_singleton()->compute_pipeline_create(gi.shader.version_get_shader(gi.shader_version, i));
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
}
Add support for low-end 3D rendering. -Reduce number of uniform sets from 6 to 4. -Remove features in low end mode, in order to reduce the number of texture units fit to 16.
2020-12-07 21:27:38 +00:00
gi.sdfgi_ubo = RD::get_singleton()->uniform_buffer_create(sizeof(GI::SDFGIData));
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
}
Add support for low-end 3D rendering. -Reduce number of uniform sets from 6 to 4. -Remove features in low end mode, in order to reduce the number of texture units fit to 16.
2020-12-07 21:27:38 +00:00
{
String defines = "\n#define OCT_SIZE " + itos(SDFGI::LIGHTPROBE_OCT_SIZE) + "\n";
Vector<String> debug_modes;
debug_modes.push_back("");
sdfgi_shader.debug.initialize(debug_modes, defines);
sdfgi_shader.debug_shader = sdfgi_shader.debug.version_create();
sdfgi_shader.debug_shader_version = sdfgi_shader.debug.version_get_shader(sdfgi_shader.debug_shader, 0);
sdfgi_shader.debug_pipeline = RD::get_singleton()->compute_pipeline_create(sdfgi_shader.debug_shader_version);
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
}
Add support for low-end 3D rendering. -Reduce number of uniform sets from 6 to 4. -Remove features in low end mode, in order to reduce the number of texture units fit to 16.
2020-12-07 21:27:38 +00:00
{
String defines = "\n#define OCT_SIZE " + itos(SDFGI::LIGHTPROBE_OCT_SIZE) + "\n";
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
Add support for low-end 3D rendering. -Reduce number of uniform sets from 6 to 4. -Remove features in low end mode, in order to reduce the number of texture units fit to 16.
2020-12-07 21:27:38 +00:00
Vector<String> versions;
versions.push_back("\n#define MODE_PROBES\n");
versions.push_back("\n#define MODE_VISIBILITY\n");
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
Add support for low-end 3D rendering. -Reduce number of uniform sets from 6 to 4. -Remove features in low end mode, in order to reduce the number of texture units fit to 16.
2020-12-07 21:27:38 +00:00
sdfgi_shader.debug_probes.initialize(versions, defines);
sdfgi_shader.debug_probes_shader = sdfgi_shader.debug_probes.version_create();
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
Add support for low-end 3D rendering. -Reduce number of uniform sets from 6 to 4. -Remove features in low end mode, in order to reduce the number of texture units fit to 16.
2020-12-07 21:27:38 +00:00
{
RD::PipelineRasterizationState rs;
rs.cull_mode = RD::POLYGON_CULL_DISABLED;
RD::PipelineDepthStencilState ds;
ds.enable_depth_test = true;
ds.enable_depth_write = true;
ds.depth_compare_operator = RD::COMPARE_OP_LESS_OR_EQUAL;
for (int i = 0; i < SDGIShader::PROBE_DEBUG_MAX; i++) {
RID debug_probes_shader_version = sdfgi_shader.debug_probes.version_get_shader(sdfgi_shader.debug_probes_shader, i);
sdfgi_shader.debug_probes_pipeline[i].setup(debug_probes_shader_version, RD::RENDER_PRIMITIVE_TRIANGLE_STRIPS, rs, RD::PipelineMultisampleState(), ds, RD::PipelineColorBlendState::create_disabled(), 0);
}
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
}
}
Add support for low-end 3D rendering. -Reduce number of uniform sets from 6 to 4. -Remove features in low end mode, in order to reduce the number of texture units fit to 16.
2020-12-07 21:27:38 +00:00
default_giprobe_buffer = RD::get_singleton()->uniform_buffer_create(sizeof(GI::GIProbeData) * RenderBuffers::MAX_GIPROBES);
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
}
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
{ //reflections
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
cluster.max_reflections = max_cluster_elements;
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
cluster.reflections = memnew_arr(Cluster::ReflectionData, cluster.max_reflections);
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
cluster.reflection_sort = memnew_arr(Cluster::InstanceSort<ReflectionProbeInstance>, cluster.max_decals);
cluster.reflection_buffer = RD::get_singleton()->storage_buffer_create(sizeof(Cluster::ReflectionData) * cluster.max_reflections);
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
}
{ //lights
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
cluster.max_lights = max_cluster_elements;
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
uint32_t light_buffer_size = cluster.max_lights * sizeof(Cluster::LightData);
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
cluster.omni_lights = memnew_arr(Cluster::LightData, cluster.max_lights);
cluster.omni_light_buffer = RD::get_singleton()->storage_buffer_create(light_buffer_size);
cluster.omni_light_sort = memnew_arr(Cluster::InstanceSort<LightInstance>, cluster.max_lights);
cluster.spot_lights = memnew_arr(Cluster::LightData, cluster.max_lights);
cluster.spot_light_buffer = RD::get_singleton()->storage_buffer_create(light_buffer_size);
cluster.spot_light_sort = memnew_arr(Cluster::InstanceSort<LightInstance>, cluster.max_lights);
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
//defines += "\n#define MAX_LIGHT_DATA_STRUCTS " + itos(cluster.max_lights) + "\n";
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
//used for volumetric fog shrinking
cluster.lights_instances = memnew_arr(RID, cluster.max_lights * 2);
cluster.lights_shadow_rect_cache = memnew_arr(Rect2i, cluster.max_lights * 2);
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
Rewrite culling to be more cache/thread friendly. -Uses a single array with all data -Massive performance improvement -Does not support threads yet, but code is now thread friendly
2020-12-26 12:05:36 +00:00
cluster.max_directional_lights = MAX_DIRECTIONAL_LIGHTS;
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
uint32_t directional_light_buffer_size = cluster.max_directional_lights * sizeof(Cluster::DirectionalLightData);
cluster.directional_lights = memnew_arr(Cluster::DirectionalLightData, cluster.max_directional_lights);
cluster.directional_light_buffer = RD::get_singleton()->uniform_buffer_create(directional_light_buffer_size);
}
{ //decals
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
cluster.max_decals = max_cluster_elements;
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
uint32_t decal_buffer_size = cluster.max_decals * sizeof(Cluster::DecalData);
cluster.decals = memnew_arr(Cluster::DecalData, cluster.max_decals);
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
cluster.decal_sort = memnew_arr(Cluster::InstanceSort<DecalInstance>, cluster.max_decals);
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
cluster.decal_buffer = RD::get_singleton()->storage_buffer_create(decal_buffer_size);
}
Add support for low-end 3D rendering. -Reduce number of uniform sets from 6 to 4. -Remove features in low end mode, in order to reduce the number of texture units fit to 16.
2020-12-07 21:27:38 +00:00
if (!low_end) {
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
String defines = "\n#define MAX_DIRECTIONAL_LIGHT_DATA_STRUCTS " + itos(cluster.max_directional_lights) + "\n";
Vector<String> volumetric_fog_modes;
volumetric_fog_modes.push_back("\n#define MODE_DENSITY\n");
volumetric_fog_modes.push_back("\n#define MODE_DENSITY\n#define ENABLE_SDFGI\n");
volumetric_fog_modes.push_back("\n#define MODE_FILTER\n");
volumetric_fog_modes.push_back("\n#define MODE_FOG\n");
volumetric_fog.shader.initialize(volumetric_fog_modes, defines);
volumetric_fog.shader_version = volumetric_fog.shader.version_create();
for (int i = 0; i < VOLUMETRIC_FOG_SHADER_MAX; i++) {
volumetric_fog.pipelines[i] = RD::get_singleton()->compute_pipeline_create(volumetric_fog.shader.version_get_shader(volumetric_fog.shader_version, i));
}
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
volumetric_fog.params_ubo = RD::get_singleton()->uniform_buffer_create(sizeof(VolumetricFogShader::ParamsUBO));
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
}
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
{
RD::SamplerState sampler;
Refactored 2D shader and lighting system -Removed normal/specular properties from nodes -Create CanvasTexture, which can contain normal/specular channels -Refactored, optimized and simplified 2D shaders -Use atlas for light textures. -Use a shadow atlas for shadow textures. -Use both items aboves to make light rendering stateless (faster). -Reorganized uniform sets for more efficiency.
2020-10-24 15:15:43 +00:00
sampler.mag_filter = RD::SAMPLER_FILTER_NEAREST;
sampler.min_filter = RD::SAMPLER_FILTER_NEAREST;
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
sampler.enable_compare = true;
sampler.compare_op = RD::COMPARE_OP_LESS;
shadow_sampler = RD::get_singleton()->sampler_create(sampler);
}
Implement MSAA
2020-04-12 18:33:57 +00:00
camera_effects_set_dof_blur_bokeh_shape(RS::DOFBokehShape(int(GLOBAL_GET("rendering/quality/depth_of_field/depth_of_field_bokeh_shape"))));
camera_effects_set_dof_blur_quality(RS::DOFBlurQuality(int(GLOBAL_GET("rendering/quality/depth_of_field/depth_of_field_bokeh_quality"))), GLOBAL_GET("rendering/quality/depth_of_field/depth_of_field_use_jitter"));
Port ASSAO to Godot to replace SAO
2020-12-08 05:37:09 +00:00
environment_set_ssao_quality(RS::EnvironmentSSAOQuality(int(GLOBAL_GET("rendering/quality/ssao/quality"))), GLOBAL_GET("rendering/quality/ssao/half_size"), GLOBAL_GET("rendering/quality/ssao/adaptive_target"), GLOBAL_GET("rendering/quality/ssao/blur_passes"), GLOBAL_GET("rendering/quality/ssao/fadeout_from"), GLOBAL_GET("rendering/quality/ssao/fadeout_to"));
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
screen_space_roughness_limiter = GLOBAL_GET("rendering/quality/screen_filters/screen_space_roughness_limiter_enabled");
screen_space_roughness_limiter_amount = GLOBAL_GET("rendering/quality/screen_filters/screen_space_roughness_limiter_amount");
screen_space_roughness_limiter_limit = GLOBAL_GET("rendering/quality/screen_filters/screen_space_roughness_limiter_limit");
Move glow upscale quality to a global setting, for consistency
2020-03-30 13:46:03 +00:00
glow_bicubic_upscale = int(GLOBAL_GET("rendering/quality/glow/upscale_mode")) > 0;
Add high quality glow mode
2020-09-01 06:04:45 +00:00
glow_high_quality = GLOBAL_GET("rendering/quality/glow/use_high_quality");
Re-Added screen space reflection.
2020-04-02 02:24:52 +00:00
ssr_roughness_quality = RS::EnvironmentSSRRoughnessQuality(int(GLOBAL_GET("rendering/quality/screen_space_reflection/roughness_quality")));
Re-implement subsurface scattering. The size settings are more "just works", with default scale and depth scale values that don't need much tweaking. Additionally, a "skin" mode was added so skin looks better. EDIT: Cleaned up SSR filter shader a bit.
2020-04-04 02:42:26 +00:00
sss_quality = RS::SubSurfaceScatteringQuality(int(GLOBAL_GET("rendering/quality/subsurface_scattering/subsurface_scattering_quality")));
sss_scale = GLOBAL_GET("rendering/quality/subsurface_scattering/subsurface_scattering_scale");
sss_depth_scale = GLOBAL_GET("rendering/quality/subsurface_scattering/subsurface_scattering_depth_scale");
Add proper quality settings to soft shadows
2020-04-10 09:30:36 +00:00
directional_penumbra_shadow_kernel = memnew_arr(float, 128);
directional_soft_shadow_kernel = memnew_arr(float, 128);
penumbra_shadow_kernel = memnew_arr(float, 128);
soft_shadow_kernel = memnew_arr(float, 128);
shadows_quality_set(RS::ShadowQuality(int(GLOBAL_GET("rendering/quality/shadows/soft_shadow_quality"))));
directional_shadow_quality_set(RS::ShadowQuality(int(GLOBAL_GET("rendering/quality/directional_shadow/soft_shadow_quality"))));
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
environment_set_volumetric_fog_volume_size(GLOBAL_GET("rendering/volumetric_fog/volume_size"), GLOBAL_GET("rendering/volumetric_fog/volume_depth"));
environment_set_volumetric_fog_filter_active(GLOBAL_GET("rendering/volumetric_fog/use_filter"));
environment_set_volumetric_fog_directional_shadow_shrink_size(GLOBAL_GET("rendering/volumetric_fog/directional_shadow_shrink"));
environment_set_volumetric_fog_positional_shadow_shrink_size(GLOBAL_GET("rendering/volumetric_fog/positional_shadow_shrink"));
Replace Octree by DynamicBVH in cull code -Much greater pairing/unpairing performance -For now, using it for culling too, but this will change in a couple of days. -Added a paged allocator, to efficiently alloc/free some types of objects.
2020-12-23 16:52:58 +00:00
cull_argument.set_page_pool(&cull_argument_pool);
Several GI related optimizations and fixes -SDFGI direct light is done over many frames -SDFGI Changed settings for rays/frame -SDFGI Misc optimizations -SDFGI Bug fix on probe scroll -GIProbe was not working, got it to work again -GIProbe dynamic objects were not working, fixed -Added a half size GI option.
2021-01-22 23:50:24 +00:00
gi.half_resolution = GLOBAL_GET("rendering/quality/gi/use_half_resolution");
Base 3D engine done, still untested, though.
2019-08-18 22:40:52 +00:00
}
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
RendererSceneRenderRD::~RendererSceneRenderRD() {
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
for (Map<int, ShadowCubemap>::Element *E = shadow_cubemaps.front(); E; E = E->next()) {
RD::get_singleton()->free(E->get().cubemap);
}
GIProbes working.
2019-10-03 20:39:08 +00:00
Add fog to sky shaders
2020-08-18 04:12:51 +00:00
if (sky_scene_state.uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(sky_scene_state.uniform_set)) {
RD::get_singleton()->free(sky_scene_state.uniform_set);
Working sky shader implementation
2020-03-20 00:32:19 +00:00
}
Add support for low-end 3D rendering. -Reduce number of uniform sets from 6 to 4. -Remove features in low end mode, in order to reduce the number of texture units fit to 16.
2020-12-07 21:27:38 +00:00
if (!low_end) {
RD::get_singleton()->free(default_giprobe_buffer);
RD::get_singleton()->free(gi_probe_lights_uniform);
RD::get_singleton()->free(gi.sdfgi_ubo);
Fixes recently introduced shader leaks
2020-06-27 13:08:20 +00:00
Add support for low-end 3D rendering. -Reduce number of uniform sets from 6 to 4. -Remove features in low end mode, in order to reduce the number of texture units fit to 16.
2020-12-07 21:27:38 +00:00
giprobe_debug_shader.version_free(giprobe_debug_shader_version);
giprobe_shader.version_free(giprobe_lighting_shader_version);
gi.shader.version_free(gi.shader_version);
sdfgi_shader.debug_probes.version_free(sdfgi_shader.debug_probes_shader);
sdfgi_shader.debug.version_free(sdfgi_shader.debug_shader);
sdfgi_shader.direct_light.version_free(sdfgi_shader.direct_light_shader);
sdfgi_shader.integrate.version_free(sdfgi_shader.integrate_shader);
sdfgi_shader.preprocess.version_free(sdfgi_shader.preprocess_shader);
Fixes recently introduced shader leaks
2020-06-27 13:08:20 +00:00
Add support for low-end 3D rendering. -Reduce number of uniform sets from 6 to 4. -Remove features in low end mode, in order to reduce the number of texture units fit to 16.
2020-12-07 21:27:38 +00:00
volumetric_fog.shader.version_free(volumetric_fog.shader_version);
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
RD::get_singleton()->free(volumetric_fog.params_ubo);
Add support for low-end 3D rendering. -Reduce number of uniform sets from 6 to 4. -Remove features in low end mode, in order to reduce the number of texture units fit to 16.
2020-12-07 21:27:38 +00:00
memdelete_arr(gi_probe_lights);
}
Add fog to sky shaders
2020-08-18 04:12:51 +00:00
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
SkyMaterialData *md = (SkyMaterialData *)storage->material_get_data(sky_shader.default_material, RendererStorageRD::SHADER_TYPE_SKY);
Working sky shader implementation
2020-03-20 00:32:19 +00:00
sky_shader.shader.version_free(md->shader_data->version);
RD::get_singleton()->free(sky_scene_state.directional_light_buffer);
Add fog to sky shaders
2020-08-18 04:12:51 +00:00
RD::get_singleton()->free(sky_scene_state.uniform_buffer);
Working sky shader implementation
2020-03-20 00:32:19 +00:00
memdelete_arr(sky_scene_state.directional_lights);
memdelete_arr(sky_scene_state.last_frame_directional_lights);
storage->free(sky_shader.default_shader);
storage->free(sky_shader.default_material);
Add fog to sky shaders
2020-08-18 04:12:51 +00:00
storage->free(sky_scene_state.fog_shader);
storage->free(sky_scene_state.fog_material);
Add proper quality settings to soft shadows
2020-04-10 09:30:36 +00:00
memdelete_arr(directional_penumbra_shadow_kernel);
memdelete_arr(directional_soft_shadow_kernel);
memdelete_arr(penumbra_shadow_kernel);
memdelete_arr(soft_shadow_kernel);
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
{
RD::get_singleton()->free(cluster.directional_light_buffer);
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
RD::get_singleton()->free(cluster.omni_light_buffer);
RD::get_singleton()->free(cluster.spot_light_buffer);
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
RD::get_singleton()->free(cluster.reflection_buffer);
RD::get_singleton()->free(cluster.decal_buffer);
memdelete_arr(cluster.directional_lights);
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
memdelete_arr(cluster.omni_lights);
memdelete_arr(cluster.spot_lights);
memdelete_arr(cluster.omni_light_sort);
memdelete_arr(cluster.spot_light_sort);
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
memdelete_arr(cluster.lights_shadow_rect_cache);
memdelete_arr(cluster.lights_instances);
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
memdelete_arr(cluster.reflections);
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
memdelete_arr(cluster.reflection_sort);
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
memdelete_arr(cluster.decals);
Added GPU based cluster builder Clustering is now GPU based, uses an implementation based on the Activision algorithm.
2021-01-17 16:25:38 +00:00
memdelete_arr(cluster.decal_sort);
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
}
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
RD::get_singleton()->free(shadow_sampler);
directional_shadow_atlas_set_size(0);
Replace Octree by DynamicBVH in cull code -Much greater pairing/unpairing performance -For now, using it for culling too, but this will change in a couple of days. -Added a paged allocator, to efficiently alloc/free some types of objects.
2020-12-23 16:52:58 +00:00
cull_argument.reset(); //avoid exit error
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
}