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

4747 lines
181 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
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::sdfgi_update(RID p_render_buffers, RID p_environment, const Vector3 &p_world_position) {
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
RendererSceneEnvironmentRD *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
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
rb->sdfgi->erase();
memdelete(rb->sdfgi);
rb->sdfgi = nullptr;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
}
return;
}
static const uint32_t history_frames_to_converge[RS::ENV_SDFGI_CONVERGE_MAX] = { 5, 10, 15, 20, 25, 30 };
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
uint32_t requested_history_size = history_frames_to_converge[gi.sdfgi_frames_to_converge];
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 && (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
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
rb->sdfgi->erase();
memdelete(rb->sdfgi);
rb->sdfgi = nullptr;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
}
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
RendererSceneGIRD::SDFGI *sdfgi = rb->sdfgi;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
if (sdfgi == nullptr) {
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
// re-create
rb->sdfgi = gi.create_sdfgi(env, p_world_position, requested_history_size);
} else {
//check for updates
rb->sdfgi->update(env, p_world_position);
}
}
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
int RendererSceneRenderRD::sdfgi_get_pending_region_count(RID p_render_buffers) const {
RenderBuffers *rb = render_buffers_owner.getornull(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
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
ERR_FAIL_COND_V(rb == nullptr, 0);
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
if (rb->sdfgi == nullptr) {
return 0;
}
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
int dirty_count = 0;
for (uint32_t i = 0; i < rb->sdfgi->cascades.size(); i++) {
const RendererSceneGIRD::SDFGI::Cascade &c = rb->sdfgi->cascades[i];
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
if (c.dirty_regions == RendererSceneGIRD::SDFGI::Cascade::DIRTY_ALL) {
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
dirty_count++;
} else {
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
for (int j = 0; j < 3; j++) {
if (c.dirty_regions[j] != 0) {
dirty_count++;
}
}
Working sky shader implementation
2020-03-20 00:32:19 +00:00
}
}
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
return dirty_count;
}
Working sky shader implementation
2020-03-20 00:32:19 +00:00
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
AABB RendererSceneRenderRD::sdfgi_get_pending_region_bounds(RID p_render_buffers, int p_region) const {
AABB bounds;
Vector3i from;
Vector3i size;
RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers);
ERR_FAIL_COND_V(rb == nullptr, AABB());
ERR_FAIL_COND_V(rb->sdfgi == nullptr, AABB());
Working sky shader implementation
2020-03-20 00:32:19 +00:00
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
int c = rb->sdfgi->get_pending_region_data(p_region, from, size, bounds);
ERR_FAIL_COND_V(c == -1, AABB());
return bounds;
}
Working sky shader implementation
2020-03-20 00:32:19 +00:00
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
uint32_t RendererSceneRenderRD::sdfgi_get_pending_region_cascade(RID p_render_buffers, int p_region) const {
AABB bounds;
Vector3i from;
Vector3i size;
RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers);
ERR_FAIL_COND_V(rb == nullptr, -1);
ERR_FAIL_COND_V(rb->sdfgi == nullptr, -1);
Working sky shader implementation
2020-03-20 00:32:19 +00:00
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
return rb->sdfgi->get_pending_region_data(p_region, from, size, bounds);
}
Working sky shader implementation
2020-03-20 00:32:19 +00:00
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
RID RendererSceneRenderRD::sky_allocate() {
return sky.allocate_sky_rid();
}
void RendererSceneRenderRD::sky_initialize(RID p_rid) {
sky.initialize_sky_rid(p_rid);
}
Working sky shader implementation
2020-03-20 00:32:19 +00:00
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
void RendererSceneRenderRD::sky_set_radiance_size(RID p_sky, int p_radiance_size) {
sky.sky_set_radiance_size(p_sky, p_radiance_size);
Working sky shader implementation
2020-03-20 00:32:19 +00:00
}
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
void RendererSceneRenderRD::sky_set_mode(RID p_sky, RS::SkyMode p_mode) {
sky.sky_set_mode(p_sky, p_mode);
}
Working sky shader implementation
2020-03-20 00:32:19 +00:00
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
void RendererSceneRenderRD::sky_set_material(RID p_sky, RID p_material) {
sky.sky_set_material(p_sky, p_material);
Working sky shader implementation
2020-03-20 00:32:19 +00:00
}
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
Ref<Image> RendererSceneRenderRD::sky_bake_panorama(RID p_sky, float p_energy, bool p_bake_irradiance, const Size2i &p_size) {
return sky.sky_bake_panorama(p_sky, p_energy, p_bake_irradiance, p_size);
Working sky shader implementation
2020-03-20 00:32:19 +00:00
}
Make Servers truly Thread Safe -Rendering server now uses a split RID allocate/initialize internally, this allows generating RIDs immediately but initialization to happen later on the proper thread (as rendering APIs generally requiere to call on the right thread). -RenderingServerWrapMT is no more, multithreading is done in RenderingServerDefault. -Some functions like texture or mesh creation, when renderer supports it, can register and return immediately (so no waiting for server API to flush, and saving staging and command buffer memory). -3D physics server changed to be made multithread friendly. -Added PhysicsServer3DWrapMT to use 3D physics server from multiple threads. -Disablet Bullet (too much effort to make multithread friendly, this needs to be fixed eventually).
2021-02-09 16:19:03 +00:00
RID RendererSceneRenderRD::environment_allocate() {
return environment_owner.allocate_rid();
}
void RendererSceneRenderRD::environment_initialize(RID p_rid) {
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
environment_owner.initialize_rid(p_rid, RendererSceneEnvironmentRD());
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) {
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
RendererSceneEnvironmentRD *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) {
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
RendererSceneEnvironmentRD *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) {
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
RendererSceneEnvironmentRD *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) {
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
RendererSceneEnvironmentRD *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) {
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
RendererSceneEnvironmentRD *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) {
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
RendererSceneEnvironmentRD *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) {
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
RendererSceneEnvironmentRD *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) {
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
RendererSceneEnvironmentRD *env = environment_owner.getornull(p_env);
Environment sky more or less working.
2019-08-26 20:43:58 +00:00
ERR_FAIL_COND(!env);
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
env->set_ambient_light(p_color, p_ambient, p_energy, p_sky_contribution, p_reflection_source, 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 {
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
RendererSceneEnvironmentRD *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 {
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
RendererSceneEnvironmentRD *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 {
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
RendererSceneEnvironmentRD *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 {
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
RendererSceneEnvironmentRD *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 {
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
RendererSceneEnvironmentRD *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 {
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
RendererSceneEnvironmentRD *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 {
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
RendererSceneEnvironmentRD *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 {
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
RendererSceneEnvironmentRD *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 {
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
RendererSceneEnvironmentRD *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 {
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
RendererSceneEnvironmentRD *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 {
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
RendererSceneEnvironmentRD *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 {
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
RendererSceneEnvironmentRD *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 {
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
RendererSceneEnvironmentRD *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) {
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
RendererSceneEnvironmentRD *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);
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
env->set_tonemap(p_tone_mapper, p_exposure, p_white, p_auto_exposure, p_min_luminance, p_max_luminance, p_auto_exp_speed, p_auto_exp_scale);
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_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) {
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
RendererSceneEnvironmentRD *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);
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
env->set_glow(p_enable, p_levels, p_intensity, p_strength, p_mix, p_bloom_threshold, p_blend_mode, p_hdr_bleed_threshold, p_hdr_bleed_scale, 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;
}
Improve SDFGI indirect light feedback loop -Use occlusion for feedback, further reduces light leaking. -More control on feedback, now its a slider.
2021-02-08 00:08:59 +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, float p_bounce_feedback, bool p_read_sky, float p_energy, float p_normal_bias, float p_probe_bias) {
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
RendererSceneEnvironmentRD *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);
Remove low_end option from renderer, being replaced by separate implementation
2021-04-07 11:06:48 +00:00
if (!is_dynamic_gi_supported()) {
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
return;
}
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
env->set_sdfgi(p_enable, p_cascades, p_min_cell_size, p_y_scale, p_use_occlusion, p_bounce_feedback, p_read_sky, p_energy, p_normal_bias, p_probe_bias);
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_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) {
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
RendererSceneEnvironmentRD *env = environment_owner.getornull(p_env);
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
ERR_FAIL_COND(!env);
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
env->set_fog(p_enable, p_light_color, p_light_energy, p_sun_scatter, p_density, p_height, p_height_density, 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 {
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
const RendererSceneEnvironmentRD *env = environment_owner.getornull(p_env);
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
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 {
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
const RendererSceneEnvironmentRD *env = environment_owner.getornull(p_env);
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
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 {
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
const RendererSceneEnvironmentRD *env = environment_owner.getornull(p_env);
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
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 {
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
const RendererSceneEnvironmentRD *env = environment_owner.getornull(p_env);
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
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 {
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
const RendererSceneEnvironmentRD *env = environment_owner.getornull(p_env);
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
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 {
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
const RendererSceneEnvironmentRD *env = environment_owner.getornull(p_env);
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
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 {
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
const RendererSceneEnvironmentRD *env = environment_owner.getornull(p_env);
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
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 {
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
const RendererSceneEnvironmentRD *env = environment_owner.getornull(p_env);
Add aerial perspective to fixed fog
2020-08-21 05:48:04 +00:00
ERR_FAIL_COND_V(!env, 0);
return env->fog_aerial_perspective;
}
Simplify Volumetric Fog -Always use temporal reproject, it just loos way better than any other filter. -By always using termporal reproject, the shadowmap reduction can be done away with, massively improving performance. -Disadvantage of temporal reproject is update latency so.. -Made sure a gaussian filter runs in XY after fog, this allows to keep stability and lower latency.
2021-02-06 14:51:56 +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, bool p_temporal_reprojection, float p_temporal_reprojection_amount) {
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
RendererSceneEnvironmentRD *env = environment_owner.getornull(p_env);
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
ERR_FAIL_COND(!env);
Remove low_end option from renderer, being replaced by separate implementation
2021-04-07 11:06:48 +00:00
if (!is_volumetric_supported()) {
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
return;
}
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
env->set_volumetric_fog(p_enable, p_density, p_light, p_light_energy, p_length, p_detail_spread, p_gi_inject, p_temporal_reprojection, p_temporal_reprojection_amount);
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
}
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_sdfgi_ray_count(RS::EnvironmentSDFGIRayCount p_ray_count) {
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
gi.sdfgi_ray_count = 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
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
void RendererSceneRenderRD::environment_set_sdfgi_frames_to_converge(RS::EnvironmentSDFGIFramesToConverge p_frames) {
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
gi.sdfgi_frames_to_converge = p_frames;
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
void RendererSceneRenderRD::environment_set_sdfgi_frames_to_update_light(RS::EnvironmentSDFGIFramesToUpdateLight p_update) {
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
gi.sdfgi_frames_to_update_light = p_update;
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
}
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) {
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
RendererSceneEnvironmentRD *env = environment_owner.getornull(p_env);
Re-Added screen space reflection.
2020-04-02 02:24:52 +00:00
ERR_FAIL_COND(!env);
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
env->set_ssr(p_enable, p_max_steps, p_fade_int, p_fade_out, p_depth_tolerance);
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::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) {
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
RendererSceneEnvironmentRD *env = environment_owner.getornull(p_env);
Re-implemented screen space ambient occlusion
2020-01-25 10:18:55 +00:00
ERR_FAIL_COND(!env);
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
env->set_ssao(p_enable, p_radius, p_intensity, p_power, p_detail, p_horizon, p_sharpness, p_light_affect, p_ao_channel_affect);
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
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 {
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
RendererSceneEnvironmentRD *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 {
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
RendererSceneEnvironmentRD *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 {
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
RendererSceneEnvironmentRD *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 {
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
RendererSceneEnvironmentRD *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 {
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
RendererSceneEnvironmentRD *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) {
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
RendererSceneEnvironmentRD *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;
Rename `instance()`->`instantiate()` when it's a verb
2021-06-17 22:03:09 +00:00
ret.instantiate();
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
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;
Reorganize Project Settings -Advanced Settings toggle also hides advanced properties when disabled -Simplified Advanced Bar (errors were just plain redundant) -Reorganized rendering quality settings. -Reorganized miscelaneous settings for clean up.
2021-02-17 16:44:49 +00:00
ra.count = GLOBAL_GET("rendering/reflections/reflection_atlas/reflection_count");
ra.size = GLOBAL_GET("rendering/reflections/reflection_atlas/reflection_size");
Reflection probes working
2019-09-09 20:50:51 +00:00
Move clustered renderer functionality
2021-03-23 10:46:31 +00:00
if (is_clustered_enabled()) {
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());
} else {
ra.cluster_builder = nullptr;
}
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 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
}
Move clustered renderer functionality
2021-03-23 10:46:31 +00:00
if (ra->cluster_builder) {
// only if we're using our cluster
ra->cluster_builder->setup(Size2i(ra->size, ra->size), max_cluster_elements, RID(), RID(), RID());
}
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
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++) {
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
ra->reflections.write[i].data.clear_reflection_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;
Fixes to forward mobile * use valid format for framebuffer: VK_FORMAT_A2B10G10R10_UNORM_PACK32 * Unfortunately cant be used for compute. * Mobile will need to do refprobe, sky, mipmapblurring using raster.
2021-07-15 20:06:33 +00:00
rpi.forward_id = _allocate_forward_id(FORWARD_ID_TYPE_REFLECTION_PROBE);
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
return reflection_probe_instance_owner.make_rid(rpi);
}
Rename Transform to Transform3D in core
2020-10-17 05:08:21 +00:00
void RendererSceneRenderRD::reflection_probe_instance_set_transform(RID p_instance, const Transform3D &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);
Porting cubemap compute shaders to raster for the mobile renderer
2021-07-27 13:33:47 +00:00
RD::get_singleton()->draw_command_begin_label("Reflection probe render");
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()) {
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
int mipmaps = MIN(sky.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;
Use 32bit instead of 64bit 3D render buffer on mobile
2021-06-17 10:37:08 +00:00
tf.format = _render_buffers_get_color_format();
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;
Fixes to forward mobile * use valid format for framebuffer: VK_FORMAT_A2B10G10R10_UNORM_PACK32 * Unfortunately cant be used for compute. * Mobile will need to do refprobe, sky, mipmapblurring using raster.
2021-07-15 20:06:33 +00:00
tf.usage_bits = RD::TEXTURE_USAGE_COLOR_ATTACHMENT_BIT | RD::TEXTURE_USAGE_SAMPLING_BIT | (_render_buffers_can_be_storage() ? RD::TEXTURE_USAGE_STORAGE_BIT : 0);
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++) {
Porting cubemap compute shaders to raster for the mobile renderer
2021-07-27 13:33:47 +00:00
atlas->reflections.write[i].data.update_reflection_data(storage, atlas->size, mipmaps, false, atlas->reflection, i * 6, storage->reflection_probe_get_update_mode(rpi->probe) == RS::REFLECTION_PROBE_UPDATE_ALWAYS, sky.roughness_layers, _render_buffers_get_color_format());
Reflection probes working
2019-09-09 20:50:51 +00:00
for (int j = 0; j < 6; j++) {
Use subpasses to do 3D rendering and resolve in mobile renderer
2021-06-26 10:49:25 +00:00
atlas->reflections.write[i].fbs[j] = reflection_probe_create_framebuffer(atlas->reflections.write[i].data.layers[0].mipmaps[0].views[j], atlas->depth_buffer);
Reflection probes working
2019-09-09 20:50:51 +00:00
}
}
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
Fix indexing of multiple reflection probes
2021-07-06 03:00:33 +00:00
if (rpi->atlas_index != -1) { // should we fail if this is still -1 ?
atlas->reflections.write[rpi->atlas_index].owner = p_instance;
}
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;
Porting cubemap compute shaders to raster for the mobile renderer
2021-07-27 13:33:47 +00:00
RD::get_singleton()->draw_command_end_label();
Reflection probes working
2019-09-09 20:50:51 +00:00
return true;
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
}
Use subpasses to do 3D rendering and resolve in mobile renderer
2021-06-26 10:49:25 +00:00
RID RendererSceneRenderRD::reflection_probe_create_framebuffer(RID p_color, RID p_depth) {
Vector<RID> fb;
fb.push_back(p_color);
fb.push_back(p_depth);
return RD::get_singleton()->framebuffer_create(fb);
}
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
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
atlas->reflections.write[rpi->atlas_index].data.create_reflection_fast_filter(storage, false);
Working sky shader implementation
2020-03-20 00:32:19 +00:00
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) {
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
atlas->reflections.write[rpi->atlas_index].data.create_reflection_importance_sample(storage, false, 10, rpi->processing_layer, sky.sky_ggx_samples_quality);
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 {
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
atlas->reflections.write[rpi->atlas_index].data.create_reflection_importance_sample(storage, false, rpi->processing_side, rpi->processing_layer, sky.sky_ggx_samples_quality);
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();
}
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);
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
}
//////////////////////////////////////////////////
Make Servers truly Thread Safe -Rendering server now uses a split RID allocate/initialize internally, this allows generating RIDs immediately but initialization to happen later on the proper thread (as rendering APIs generally requiere to call on the right thread). -RenderingServerWrapMT is no more, multithreading is done in RenderingServerDefault. -Some functions like texture or mesh creation, when renderer supports it, can register and return immediately (so no waiting for server API to flush, and saving staging and command buffer memory). -3D physics server changed to be made multithread friendly. -Added PhysicsServer3DWrapMT to use 3D physics server from multiple threads. -Disablet Bullet (too much effort to make multithread friendly, this needs to be fixed eventually).
2021-02-09 16:19:03 +00:00
RID RendererSceneRenderRD::camera_effects_allocate() {
return camera_effects_owner.allocate_rid();
}
void RendererSceneRenderRD::camera_effects_initialize(RID p_rid) {
camera_effects_owner.initialize_rid(p_rid, CameraEffects());
WIP CameraEffects implementation (bokeh not working for now)
2020-01-13 18:37:24 +00:00
}
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);
Fixes to forward mobile * use valid format for framebuffer: VK_FORMAT_A2B10G10R10_UNORM_PACK32 * Unfortunately cant be used for compute. * Mobile will need to do refprobe, sky, mipmapblurring using raster.
2021-07-15 20:06:33 +00:00
if (light_instance->light_type != RS::LIGHT_DIRECTIONAL) {
light_instance->forward_id = _allocate_forward_id(light_instance->light_type == RS::LIGHT_OMNI ? FORWARD_ID_TYPE_OMNI_LIGHT : FORWARD_ID_TYPE_SPOT_LIGHT);
}
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
return li;
}
Rename Transform to Transform3D in core
2020-10-17 05:08:21 +00:00
void RendererSceneRenderRD::light_instance_set_transform(RID p_light_instance, const Transform3D &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;
}
Rename Transform to Transform3D in core
2020-10-17 05:08:21 +00:00
void RendererSceneRenderRD::light_instance_set_shadow_transform(RID p_light_instance, const CameraMatrix &p_projection, const Transform3D &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);
Rewrote how barriers work for faster rendering -Added more finegrained control in RenderingDevice API -Optimized barriers (use less ones for thee same) -General optimizations -Shadows render all together unbarriered -GI can render together with shadows. -SDFGI can render together with depth-preoass. -General fixes -Added GPU detection
2021-02-02 19:51:36 +00:00
ERR_FAIL_INDEX(p_pass, 6);
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
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());
}
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
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);
}
Dynamic object support for GI Probes (a bit buggy still)
2019-10-11 02:14:56 +00:00
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
shadow_cubemaps[p_size] = sc;
}
Dynamic object support for GI Probes (a bit buggy still)
2019-10-11 02:14:56 +00:00
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
return &shadow_cubemaps[p_size];
}
Dynamic object support for GI Probes (a bit buggy still)
2019-10-11 02:14:56 +00:00
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
//////////////////////////
GIProbes working.
2019-10-03 20:39:08 +00:00
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
RID RendererSceneRenderRD::decal_instance_create(RID p_decal) {
DecalInstance di;
di.decal = p_decal;
Fixes to forward mobile * use valid format for framebuffer: VK_FORMAT_A2B10G10R10_UNORM_PACK32 * Unfortunately cant be used for compute. * Mobile will need to do refprobe, sky, mipmapblurring using raster.
2021-07-15 20:06:33 +00:00
di.forward_id = _allocate_forward_id(FORWARD_ID_TYPE_DECAL);
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
return decal_instance_owner.make_rid(di);
}
GIProbes working.
2019-10-03 20:39:08 +00:00
Rename Transform to Transform3D in core
2020-10-17 05:08:21 +00:00
void RendererSceneRenderRD::decal_instance_set_transform(RID p_decal, const Transform3D &p_transform) {
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +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
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
/////////////////////////////////
GIProbes working.
2019-10-03 20:39:08 +00:00
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
RID RendererSceneRenderRD::lightmap_instance_create(RID p_lightmap) {
LightmapInstance li;
li.lightmap = p_lightmap;
return lightmap_instance_owner.make_rid(li);
}
Rename Transform to Transform3D in core
2020-10-17 05:08:21 +00:00
void RendererSceneRenderRD::lightmap_instance_set_transform(RID p_lightmap, const Transform3D &p_transform) {
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
LightmapInstance *li = lightmap_instance_owner.getornull(p_lightmap);
ERR_FAIL_COND(!li);
li->transform = p_transform;
}
Dynamic object support for GI Probes (a bit buggy still)
2019-10-11 02:14:56 +00:00
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
/////////////////////////////////
Dynamic object support for GI Probes (a bit buggy still)
2019-10-11 02:14:56 +00:00
Rename GI Classes * GIProbe is now VoxelGI * BakedLightmap is now LightmapGI As godot adds more ways to provide GI (as an example, SDFGI in 4.0), the different techniques (which have different pros/cons) need to be properly named to avoid confusion.
2021-06-04 22:47:26 +00:00
RID RendererSceneRenderRD::voxel_gi_instance_create(RID p_base) {
return gi.voxel_gi_instance_create(p_base);
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
}
Dynamic object support for GI Probes (a bit buggy still)
2019-10-11 02:14:56 +00:00
Rename GI Classes * GIProbe is now VoxelGI * BakedLightmap is now LightmapGI As godot adds more ways to provide GI (as an example, SDFGI in 4.0), the different techniques (which have different pros/cons) need to be properly named to avoid confusion.
2021-06-04 22:47:26 +00:00
void RendererSceneRenderRD::voxel_gi_instance_set_transform_to_data(RID p_probe, const Transform3D &p_xform) {
gi.voxel_gi_instance_set_transform_to_data(p_probe, p_xform);
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
}
Dynamic object support for GI Probes (a bit buggy still)
2019-10-11 02:14:56 +00:00
Rename GI Classes * GIProbe is now VoxelGI * BakedLightmap is now LightmapGI As godot adds more ways to provide GI (as an example, SDFGI in 4.0), the different techniques (which have different pros/cons) need to be properly named to avoid confusion.
2021-06-04 22:47:26 +00:00
bool RendererSceneRenderRD::voxel_gi_needs_update(RID p_probe) const {
Remove low_end option from renderer, being replaced by separate implementation
2021-04-07 11:06:48 +00:00
if (!is_dynamic_gi_supported()) {
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
return false;
GIProbes working.
2019-10-03 20:39:08 +00:00
}
Rename GI Classes * GIProbe is now VoxelGI * BakedLightmap is now LightmapGI As godot adds more ways to provide GI (as an example, SDFGI in 4.0), the different techniques (which have different pros/cons) need to be properly named to avoid confusion.
2021-06-04 22:47:26 +00:00
return gi.voxel_gi_needs_update(p_probe);
GIProbes working.
2019-10-03 20:39:08 +00:00
}
Rename GI Classes * GIProbe is now VoxelGI * BakedLightmap is now LightmapGI As godot adds more ways to provide GI (as an example, SDFGI in 4.0), the different techniques (which have different pros/cons) need to be properly named to avoid confusion.
2021-06-04 22:47:26 +00:00
void RendererSceneRenderRD::voxel_gi_update(RID p_probe, bool p_update_light_instances, const Vector<RID> &p_light_instances, const PagedArray<GeometryInstance *> &p_dynamic_objects) {
Remove low_end option from renderer, being replaced by separate implementation
2021-04-07 11:06:48 +00:00
if (!is_dynamic_gi_supported()) {
GIProbes working.
2019-10-03 20:39:08 +00:00
return;
}
Rename GI Classes * GIProbe is now VoxelGI * BakedLightmap is now LightmapGI As godot adds more ways to provide GI (as an example, SDFGI in 4.0), the different techniques (which have different pros/cons) need to be properly named to avoid confusion.
2021-06-04 22:47:26 +00:00
gi.voxel_gi_update(p_probe, p_update_light_instances, p_light_instances, p_dynamic_objects, this);
GIProbes working.
2019-10-03 20:39:08 +00:00
}
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
}
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
rb->sdfgi->debug_probes(p_draw_list, p_framebuffer, p_camera_with_transform);
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);
Use fragment shader instead of compute shader for effects for mobile renderer
2021-07-20 11:40:16 +00:00
// TODO make sure texture_create_shared_from_slice works for multiview
-Refactored post processing, re-added glow and added a mix blend mode.
2020-01-10 00:40:26 +00:00
RD::TextureFormat tf;
Use fragment shader instead of compute shader for effects for mobile renderer
2021-07-20 11:40:16 +00:00
tf.format = _render_buffers_get_color_format(); // RD::DATA_FORMAT_R16G16B16A16_SFLOAT;
-Refactored post processing, re-added glow and added a mix blend mode.
2020-01-10 00:40:26 +00:00
tf.width = rb->width;
tf.height = rb->height;
Use fragment shader instead of compute shader for effects for mobile renderer
2021-07-20 11:40:16 +00:00
tf.texture_type = rb->view_count > 1 ? RD::TEXTURE_TYPE_2D_ARRAY : RD::TEXTURE_TYPE_2D;
tf.array_layers = rb->view_count;
tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_CAN_COPY_TO_BIT;
if (_render_buffers_can_be_storage()) {
tf.usage_bits += RD::TEXTURE_USAGE_STORAGE_BIT;
} 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
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;
Use fragment shader instead of compute shader for effects for mobile renderer
2021-07-20 11:40:16 +00:00
if (!_render_buffers_can_be_storage()) {
Vector<RID> fb;
fb.push_back(mm.texture);
mm.fb = RD::get_singleton()->framebuffer_create(fb);
}
if (!_render_buffers_can_be_storage()) {
// and half texture, this is an intermediate result so just allocate a texture, is this good enough?
tf.width = MAX(1, base_width >> 1);
tf.height = base_height;
tf.mipmaps = 1; // 1 or 0?
mm.half_texture = RD::get_singleton()->texture_create(tf, RD::TextureView());
Vector<RID> half_fb;
half_fb.push_back(mm.half_texture);
mm.half_fb = RD::get_singleton()->framebuffer_create(half_fb);
}
-Refactored post processing, re-added glow and added a mix blend mode.
2020-01-10 00:40:26 +00:00
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);
Use fragment shader instead of compute shader for effects for mobile renderer
2021-07-20 11:40:16 +00:00
if (!_render_buffers_can_be_storage()) {
Vector<RID> fb;
fb.push_back(mm.texture);
mm.fb = RD::get_singleton()->framebuffer_create(fb);
// We can re-use the half texture here as it is an intermediate result
}
-Refactored post processing, re-added glow and added a mix blend mode.
2020-01-10 00:40:26 +00:00
rb->blur[1].mipmaps.push_back(mm);
}
base_width = MAX(1, base_width >> 1);
base_height = MAX(1, base_height >> 1);
}
Implemented raster versions of bokeh shaders to replace broken gaussian implementation
2021-07-29 07:43:17 +00:00
if (!_render_buffers_can_be_storage()) {
// create 4 weight textures, 2 full size, 2 half size
tf.format = RD::DATA_FORMAT_R16_SFLOAT; // We could probably use DATA_FORMAT_R8_SNORM if we don't pre-multiply by blur_size but that depends on whether we can remove DEPTH_GAP
tf.width = rb->width;
tf.height = rb->height;
tf.texture_type = rb->view_count > 1 ? RD::TEXTURE_TYPE_2D_ARRAY : RD::TEXTURE_TYPE_2D;
tf.array_layers = rb->view_count;
tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_CAN_COPY_TO_BIT | RD::TEXTURE_USAGE_COLOR_ATTACHMENT_BIT;
tf.mipmaps = 1;
for (uint32_t i = 0; i < 4; i++) {
// associated blur texture
RID texture;
if (i == 0) {
texture = rb->texture;
} else if (i == 1) {
texture = rb->blur[0].mipmaps[0].texture;
} else if (i == 2) {
texture = rb->blur[1].mipmaps[0].texture;
} else if (i == 3) {
texture = rb->blur[0].mipmaps[1].texture;
}
// create weight texture
rb->weight_buffers[i].weight = RD::get_singleton()->texture_create(tf, RD::TextureView());
// create frame buffer
Vector<RID> fb;
fb.push_back(texture);
fb.push_back(rb->weight_buffers[i].weight);
rb->weight_buffers[i].fb = RD::get_singleton()->framebuffer_create(fb);
if (i == 1) {
// next 2 are half size
tf.width = MAX(1, tf.width >> 1);
tf.height = MAX(1, tf.height >> 1);
}
}
{
// and finally an FB for just our base weights
Vector<RID> fb;
fb.push_back(rb->weight_buffers[0].weight);
rb->base_weight_fb = RD::get_singleton()->framebuffer_create(fb);
}
}
-Refactored post processing, re-added glow and added a mix blend mode.
2020-01-10 00:40:26 +00:00
}
Fix read from screen and depth texture
2021-08-11 11:00:12 +00:00
void RendererSceneRenderRD::_allocate_depth_backbuffer_textures(RenderBuffers *rb) {
ERR_FAIL_COND(!rb->depth_back_texture.is_null());
{
RD::TextureFormat tf;
if (rb->view_count > 1) {
tf.texture_type = RD::TEXTURE_TYPE_2D_ARRAY;
}
// We're not using this as a depth stencil, just copying our data into this. May need to look into using a different format on mobile, maybe R16?
tf.format = RD::DATA_FORMAT_R32_SFLOAT;
tf.width = rb->width;
tf.height = rb->height;
tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT;
tf.array_layers = rb->view_count; // create a layer for every view
tf.usage_bits |= RD::TEXTURE_USAGE_CAN_COPY_TO_BIT | RD::TEXTURE_USAGE_STORAGE_BIT;
tf.usage_bits |= RD::TEXTURE_USAGE_COLOR_ATTACHMENT_BIT; // set this as color attachment because we're copying data into it, it's not actually used as a depth buffer
rb->depth_back_texture = RD::get_singleton()->texture_create(tf, RD::TextureView());
}
if (!_render_buffers_can_be_storage()) {
// create framebuffer so we can write into this...
Vector<RID> fb;
fb.push_back(rb->depth_back_texture);
rb->depth_back_fb = RD::get_singleton()->framebuffer_create(fb, RD::INVALID_ID, rb->view_count);
}
}
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;
bool final = w == 1 && h == 1;
Use fragment shader instead of compute shader for effects for mobile renderer
2021-07-20 11:40:16 +00:00
if (_render_buffers_can_be_storage()) {
tf.usage_bits = RD::TEXTURE_USAGE_STORAGE_BIT;
if (final) {
tf.usage_bits |= RD::TEXTURE_USAGE_SAMPLING_BIT;
}
} else {
tf.usage_bits = RD::TEXTURE_USAGE_COLOR_ATTACHMENT_BIT | RD::TEXTURE_USAGE_SAMPLING_BIT;
Auto exposure re-implemented in Vulkan
2020-01-12 01:26:52 +00:00
}
RID texture = RD::get_singleton()->texture_create(tf, RD::TextureView());
rb->luminance.reduce.push_back(texture);
Use fragment shader instead of compute shader for effects for mobile renderer
2021-07-20 11:40:16 +00:00
if (!_render_buffers_can_be_storage()) {
Vector<RID> fb;
fb.push_back(texture);
rb->luminance.fb.push_back(RD::get_singleton()->framebuffer_create(fb));
}
Auto exposure re-implemented in Vulkan
2020-01-12 01:26:52 +00:00
if (final) {
rb->luminance.current = RD::get_singleton()->texture_create(tf, RD::TextureView());
Use fragment shader instead of compute shader for effects for mobile renderer
2021-07-20 11:40:16 +00:00
if (!_render_buffers_can_be_storage()) {
Vector<RID> fb;
fb.push_back(rb->luminance.current);
rb->luminance.current_fb = RD::get_singleton()->framebuffer_create(fb);
}
Auto exposure re-implemented in Vulkan
2020-01-12 01:26:52 +00:00
break;
}
}
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
void RendererSceneRenderRD::_free_render_buffer_data(RenderBuffers *rb) {
Use fragment shader instead of compute shader for effects for mobile renderer
2021-07-20 11:40:16 +00:00
if (rb->texture_fb.is_valid()) {
RD::get_singleton()->free(rb->texture_fb);
rb->texture_fb = RID();
}
-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
}
Fix read from screen and depth texture
2021-08-11 11:00:12 +00:00
if (rb->depth_back_fb.is_valid()) {
RD::get_singleton()->free(rb->depth_back_fb);
rb->depth_back_fb = RID();
}
if (rb->depth_back_texture.is_valid()) {
RD::get_singleton()->free(rb->depth_back_texture);
rb->depth_back_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++) {
Use fragment shader instead of compute shader for effects for mobile renderer
2021-07-20 11:40:16 +00:00
for (int m = 0; m < rb->blur[i].mipmaps.size(); m++) {
// do we free the texture slice here? or is it enough to free the main texture?
// do free the mobile extra stuff
if (rb->blur[i].mipmaps[m].fb.is_valid()) {
RD::get_singleton()->free(rb->blur[i].mipmaps[m].fb);
}
if (rb->blur[i].mipmaps[m].half_fb.is_valid()) {
RD::get_singleton()->free(rb->blur[i].mipmaps[m].half_fb);
}
if (rb->blur[i].mipmaps[m].half_texture.is_valid()) {
RD::get_singleton()->free(rb->blur[i].mipmaps[m].half_texture);
}
}
rb->blur[i].mipmaps.clear();
-Refactored post processing, re-added glow and added a mix blend mode.
2020-01-10 00:40:26 +00:00
if (rb->blur[i].texture.is_valid()) {
RD::get_singleton()->free(rb->blur[i].texture);
rb->blur[i].texture = RID();
}
}
Use fragment shader instead of compute shader for effects for mobile renderer
2021-07-20 11:40:16 +00:00
for (int i = 0; i < rb->luminance.fb.size(); i++) {
RD::get_singleton()->free(rb->luminance.fb[i]);
}
rb->luminance.fb.clear();
-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 < rb->luminance.reduce.size(); i++) {
RD::get_singleton()->free(rb->luminance.reduce[i]);
}
rb->luminance.reduce.clear();
Use fragment shader instead of compute shader for effects for mobile renderer
2021-07-20 11:40:16 +00:00
if (rb->luminance.current_fb.is_valid()) {
RD::get_singleton()->free(rb->luminance.current_fb);
rb->luminance.current_fb = RID();
}
-Refactored post processing, re-added glow and added a mix blend mode.
2020-01-10 00:40:26 +00:00
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);
}
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;
}
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
RendererSceneEnvironmentRD *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);
}
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);
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
RendererSceneEnvironmentRD *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");
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
Store SSAO uniform sets per viewport
2021-06-13 17:42:10 +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, rb->ssao.downsample_uniform_set, rb->ssao.gather_uniform_set, rb->ssao.importance_map_uniform_set);
-Refactored post processing, re-added glow and added a mix blend mode.
2020-01-10 00:40:26 +00:00
}
Fix read from screen and depth texture
2021-08-11 11:00:12 +00:00
void RendererSceneRenderRD::_render_buffers_copy_screen_texture(const RenderDataRD *p_render_data) {
RenderBuffers *rb = render_buffers_owner.getornull(p_render_data->render_buffers);
ERR_FAIL_COND(!rb);
RD::get_singleton()->draw_command_begin_label("Copy screen texture");
if (rb->blur[0].texture.is_null()) {
_allocate_blur_textures(rb);
}
// @TODO IMPLEMENT MULTIVIEW, all effects need to support stereo buffers or effects are only applied to the left eye
bool can_use_storage = _render_buffers_can_be_storage();
if (can_use_storage) {
storage->get_effects()->copy_to_rect(rb->texture, rb->blur[0].mipmaps[0].texture, Rect2i(0, 0, rb->width, rb->height));
for (int i = 1; i < rb->blur[0].mipmaps.size(); i++) {
storage->get_effects()->make_mipmap(rb->blur[0].mipmaps[i - 1].texture, rb->blur[0].mipmaps[i].texture, Size2i(rb->blur[0].mipmaps[i].width, rb->blur[0].mipmaps[i].height));
}
} else {
storage->get_effects()->copy_to_fb_rect(rb->texture, rb->blur[0].mipmaps[0].fb, Rect2i(0, 0, rb->width, rb->height));
for (int i = 1; i < rb->blur[0].mipmaps.size(); i++) {
storage->get_effects()->make_mipmap_raster(rb->blur[0].mipmaps[i - 1].texture, rb->blur[0].mipmaps[i].fb, Size2i(rb->blur[0].mipmaps[i].width, rb->blur[0].mipmaps[i].height));
}
}
RD::get_singleton()->draw_command_end_label();
}
void RendererSceneRenderRD::_render_buffers_copy_depth_texture(const RenderDataRD *p_render_data) {
RenderBuffers *rb = render_buffers_owner.getornull(p_render_data->render_buffers);
ERR_FAIL_COND(!rb);
RD::get_singleton()->draw_command_begin_label("Copy depth texture");
if (rb->depth_back_texture.is_null()) {
_allocate_depth_backbuffer_textures(rb);
}
// @TODO IMPLEMENT MULTIVIEW, all effects need to support stereo buffers or effects are only applied to the left eye
bool can_use_storage = _render_buffers_can_be_storage();
if (can_use_storage) {
storage->get_effects()->copy_to_rect(rb->depth_texture, rb->depth_back_texture, Rect2i(0, 0, rb->width, rb->height));
} else {
storage->get_effects()->copy_to_fb_rect(rb->depth_texture, rb->depth_back_fb, Rect2i(0, 0, rb->width, rb->height));
}
RD::get_singleton()->draw_command_end_label();
}
Split RenderDataRD struct from RenderState struct to simplify passing our render data around the renderer.
2021-05-05 05:41:12 +00:00
void RendererSceneRenderRD::_render_buffers_post_process_and_tonemap(const RenderDataRD *p_render_data) {
RenderBuffers *rb = render_buffers_owner.getornull(p_render_data->render_buffers);
-Refactored post processing, re-added glow and added a mix blend mode.
2020-01-10 00:40:26 +00:00
ERR_FAIL_COND(!rb);
Split RenderDataRD struct from RenderState struct to simplify passing our render data around the renderer.
2021-05-05 05:41:12 +00:00
RendererSceneEnvironmentRD *env = environment_owner.getornull(p_render_data->environment);
-Refactored post processing, re-added glow and added a mix blend mode.
2020-01-10 00:40:26 +00:00
//glow (if enabled)
Split RenderDataRD struct from RenderState struct to simplify passing our render data around the renderer.
2021-05-05 05:41:12 +00:00
CameraEffects *camfx = camera_effects_owner.getornull(p_render_data->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;
Use fragment shader instead of compute shader for effects for mobile renderer
2021-07-20 11:40:16 +00:00
bool can_use_storage = _render_buffers_can_be_storage();
// @TODO IMPLEMENT MULTIVIEW, all effects need to support stereo buffers or effects are only applied to the left eye
Auto exposure re-implemented in Vulkan
2020-01-12 01:26:52 +00:00
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) {
Use fragment shader instead of compute shader for effects for mobile renderer
2021-07-20 11:40:16 +00:00
RD::get_singleton()->draw_command_begin_label("DOF");
WIP CameraEffects implementation (bokeh not working for now)
2020-01-13 18:37:24 +00:00
if (rb->blur[0].texture.is_null()) {
_allocate_blur_textures(rb);
}
Implemented raster versions of bokeh shaders to replace broken gaussian implementation
2021-07-29 07:43:17 +00:00
EffectsRD::BokehBuffers buffers;
// textures we use
buffers.base_texture_size = Size2i(rb->width, rb->height);
buffers.base_texture = rb->texture;
buffers.depth_texture = rb->depth_texture;
buffers.secondary_texture = rb->blur[0].mipmaps[0].texture;
buffers.half_texture[0] = rb->blur[1].mipmaps[0].texture;
buffers.half_texture[1] = rb->blur[0].mipmaps[1].texture;
float bokeh_size = camfx->dof_blur_amount * 64.0;
Use fragment shader instead of compute shader for effects for mobile renderer
2021-07-20 11:40:16 +00:00
if (can_use_storage) {
Implemented raster versions of bokeh shaders to replace broken gaussian implementation
2021-07-29 07:43:17 +00:00
storage->get_effects()->bokeh_dof(buffers, 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_render_data->z_near, p_render_data->z_far, p_render_data->cam_ortogonal);
Use fragment shader instead of compute shader for effects for mobile renderer
2021-07-20 11:40:16 +00:00
} else {
Implemented raster versions of bokeh shaders to replace broken gaussian implementation
2021-07-29 07:43:17 +00:00
// set framebuffers
buffers.base_fb = rb->texture_fb;
buffers.secondary_fb = rb->weight_buffers[1].fb;
buffers.half_fb[0] = rb->weight_buffers[2].fb;
buffers.half_fb[1] = rb->weight_buffers[3].fb;
buffers.weight_texture[0] = rb->weight_buffers[0].weight;
buffers.weight_texture[1] = rb->weight_buffers[1].weight;
buffers.weight_texture[2] = rb->weight_buffers[2].weight;
buffers.weight_texture[3] = rb->weight_buffers[3].weight;
// set weight buffers
buffers.base_weight_fb = rb->base_weight_fb;
storage->get_effects()->bokeh_dof_raster(buffers, 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, p_render_data->z_near, p_render_data->z_far, p_render_data->cam_ortogonal);
Use fragment shader instead of compute shader for effects for mobile renderer
2021-07-20 11:40:16 +00:00
}
RD::get_singleton()->draw_command_end_label();
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) {
Use fragment shader instead of compute shader for effects for mobile renderer
2021-07-20 11:40:16 +00:00
RD::get_singleton()->draw_command_begin_label("Auto exposure");
Auto exposure re-implemented in Vulkan
2020-01-12 01:26:52 +00:00
if (rb->luminance.current.is_null()) {
_allocate_luminance_textures(rb);
}
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;
Use fragment shader instead of compute shader for effects for mobile renderer
2021-07-20 11:40:16 +00:00
if (can_use_storage) {
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);
} else {
storage->get_effects()->luminance_reduction_raster(rb->texture, Size2i(rb->width, rb->height), rb->luminance.reduce, rb->luminance.fb, rb->luminance.current, env->min_luminance, env->max_luminance, step, set_immediate);
}
Auto exposure re-implemented in Vulkan
2020-01-12 01:26:52 +00:00
//swap final reduce with prev luminance
SWAP(rb->luminance.current, rb->luminance.reduce.write[rb->luminance.reduce.size() - 1]);
Use fragment shader instead of compute shader for effects for mobile renderer
2021-07-20 11:40:16 +00:00
if (!can_use_storage) {
SWAP(rb->luminance.current_fb, rb->luminance.fb.write[rb->luminance.fb.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
Use fragment shader instead of compute shader for effects for mobile renderer
2021-07-20 11:40:16 +00:00
RD::get_singleton()->draw_command_end_label();
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) {
Use fragment shader instead of compute shader for effects for mobile renderer
2021-07-20 11:40:16 +00:00
RD::get_singleton()->draw_command_begin_label("Gaussian Glow");
-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);
}
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;
}
Use fragment shader instead of compute shader for effects for mobile renderer
2021-07-20 11:40:16 +00:00
if (can_use_storage) {
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);
} else {
storage->get_effects()->gaussian_glow_raster(rb->texture, rb->blur[1].mipmaps[i].half_fb, rb->blur[1].mipmaps[i].half_texture, rb->blur[1].mipmaps[i].fb, 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 {
Use fragment shader instead of compute shader for effects for mobile renderer
2021-07-20 11:40:16 +00:00
if (can_use_storage) {
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);
} else {
storage->get_effects()->gaussian_glow_raster(rb->blur[1].mipmaps[i - 1].texture, rb->blur[1].mipmaps[i].half_fb, rb->blur[1].mipmaps[i].half_texture, rb->blur[1].mipmaps[i].fb, Vector2(1.0 / vp_w, 1.0 / 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
}
}
Use fragment shader instead of compute shader for effects for mobile renderer
2021-07-20 11:40:16 +00:00
RD::get_singleton()->draw_command_end_label();
-Refactored post processing, re-added glow and added a mix blend mode.
2020-01-10 00:40:26 +00:00
}
{
Use fragment shader instead of compute shader for effects for mobile renderer
2021-07-20 11:40:16 +00:00
RD::get_singleton()->draw_command_begin_label("Tonemap");
-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);
}
}
Add stereoscopic rendering through multiview
2021-05-07 13:19:04 +00:00
tonemap.view_count = p_render_data->view_count;
-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);
Use fragment shader instead of compute shader for effects for mobile renderer
2021-07-20 11:40:16 +00:00
RD::get_singleton()->draw_command_end_label();
-Refactored post processing, re-added glow and added a mix blend mode.
2020-01-10 00:40:26 +00:00
}
storage->render_target_disable_clear_request(rb->render_target);
}
Use subpasses to do 3D rendering and resolve in mobile renderer
2021-06-26 10:49:25 +00:00
void RendererSceneRenderRD::_post_process_subpass(RID p_source_texture, RID p_framebuffer, const RenderDataRD *p_render_data) {
RD::get_singleton()->draw_command_begin_label("Post Process Subpass");
RenderBuffers *rb = render_buffers_owner.getornull(p_render_data->render_buffers);
ERR_FAIL_COND(!rb);
RendererSceneEnvironmentRD *env = environment_owner.getornull(p_render_data->environment);
bool can_use_effects = rb->width >= 8 && rb->height >= 8;
RD::DrawListID draw_list = RD::get_singleton()->draw_list_switch_to_next_pass();
EffectsRD::TonemapSettings tonemap;
if (env) {
tonemap.tonemap_mode = env->tone_mapper;
tonemap.exposure = env->exposure;
tonemap.white = env->white;
}
// We don't support glow or auto exposure here, if they are needed, don't use subpasses!
// The problem is that we need to use the result so far and process them before we can
// apply this to our results.
if (can_use_effects && env && env->glow_enabled) {
ERR_FAIL_MSG("Glow is not supported when using subpasses.");
}
if (can_use_effects && env && env->auto_exposure) {
ERR_FAIL_MSG("Glow is not supported when using subpasses.");
}
tonemap.use_glow = false;
tonemap.glow_texture = storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_BLACK);
tonemap.use_auto_exposure = false;
tonemap.exposure_texture = storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_WHITE);
tonemap.use_color_correction = false;
tonemap.use_1d_color_correction = false;
tonemap.color_correction_texture = storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_3D_WHITE);
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;
tonemap.use_1d_color_correction = env->use_1d_color_correction;
tonemap.color_correction_texture = storage->texture_get_rd_texture(env->color_correction);
}
}
tonemap.use_debanding = rb->use_debanding;
tonemap.texture_size = Vector2i(rb->width, rb->height);
tonemap.view_count = p_render_data->view_count;
storage->get_effects()->tonemapper(draw_list, p_source_texture, RD::get_singleton()->framebuffer_get_format(p_framebuffer), tonemap);
RD::get_singleton()->draw_command_end_label();
}
void RendererSceneRenderRD::_disable_clear_request(const RenderDataRD *p_render_data) {
RenderBuffers *rb = render_buffers_owner.getornull(p_render_data->render_buffers);
ERR_FAIL_COND(!rb);
storage->render_target_disable_clear_request(rb->render_target);
}
Implement occlusion culling Added an occlusion culling system with support for static occluder meshes. It can be enabled via `Project Settings > Rendering > Occlusion Culling > Use Occlusion Culling`. Occluders are defined via the new `Occluder3D` resource and instanced using the new `OccluderInstance3D` node. The occluders can also be automatically baked from a scene using the built-in editor plugin.
2021-04-20 16:40:24 +00:00
void RendererSceneRenderRD::_render_buffers_debug_draw(RID p_render_buffers, RID p_shadow_atlas, RID p_occlusion_buffer) {
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
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);
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
effects->copy_to_fb_rect(decal_atlas, storage->render_target_get_rd_framebuffer(rb->render_target), Rect2i(Vector2(), rtsize / 2), false, false, true);
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
}
}
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
if (debug_draw == RS::VIEWPORT_DEBUG_DRAW_SCENE_LUMINANCE) {
if (rb->luminance.current.is_valid()) {
Size2 rtsize = storage->render_target_get_size(rb->render_target);
effects->copy_to_fb_rect(rb->luminance.current, storage->render_target_get_rd_framebuffer(rb->render_target), Rect2(Vector2(), rtsize / 8), false, true);
}
}
if (debug_draw == RS::VIEWPORT_DEBUG_DRAW_SSAO && rb->ssao.ao_final.is_valid()) {
Size2 rtsize = storage->render_target_get_size(rb->render_target);
RID ao_buf = rb->ssao.ao_final;
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);
}
if (debug_draw == RS::VIEWPORT_DEBUG_DRAW_GI_BUFFER && rb->ambient_buffer.is_valid()) {
Size2 rtsize = storage->render_target_get_size(rb->render_target);
RID ambient_texture = rb->ambient_buffer;
RID reflection_texture = rb->reflection_buffer;
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);
}
Implement occlusion culling Added an occlusion culling system with support for static occluder meshes. It can be enabled via `Project Settings > Rendering > Occlusion Culling > Use Occlusion Culling`. Occluders are defined via the new `Occluder3D` resource and instanced using the new `OccluderInstance3D` node. The occluders can also be automatically baked from a scene using the built-in editor plugin.
2021-04-20 16:40:24 +00:00
if (debug_draw == RS::VIEWPORT_DEBUG_DRAW_OCCLUDERS) {
if (p_occlusion_buffer.is_valid()) {
Size2 rtsize = storage->render_target_get_size(rb->render_target);
effects->copy_to_fb_rect(storage->texture_get_rd_texture(p_occlusion_buffer), storage->render_target_get_rd_framebuffer(rb->render_target), Rect2i(Vector2(), rtsize), true, false);
}
}
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +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) {
RendererSceneEnvironmentRD *env = environment_owner.getornull(p_env);
ERR_FAIL_COND(!env);
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
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;
-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;
}
Fix read from screen and depth texture
2021-08-11 11:00:12 +00:00
RID RendererSceneRenderRD::render_buffers_get_back_depth_texture(RID p_render_buffers) {
RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers);
ERR_FAIL_COND_V(!rb, RID());
if (!rb->depth_back_texture.is_valid()) {
return RID(); //not valid at the moment
}
return rb->depth_back_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
}
Rename GI Classes * GIProbe is now VoxelGI * BakedLightmap is now LightmapGI As godot adds more ways to provide GI (as an example, SDFGI in 4.0), the different techniques (which have different pros/cons) need to be properly named to avoid confusion.
2021-06-04 22:47:26 +00:00
RID RendererSceneRenderRD::render_buffers_get_voxel_gi_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());
Rename GI Classes * GIProbe is now VoxelGI * BakedLightmap is now LightmapGI As godot adds more ways to provide GI (as an example, SDFGI in 4.0), the different techniques (which have different pros/cons) need to be properly named to avoid confusion.
2021-06-04 22:47:26 +00:00
if (rb->gi.voxel_gi_buffer.is_null()) {
rb->gi.voxel_gi_buffer = RD::get_singleton()->uniform_buffer_create(sizeof(RendererSceneGIRD::VoxelGIData) * RendererSceneGIRD::MAX_VOXEL_GI_INSTANCES);
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
}
Rename GI Classes * GIProbe is now VoxelGI * BakedLightmap is now LightmapGI As godot adds more ways to provide GI (as an example, SDFGI in 4.0), the different techniques (which have different pros/cons) need to be properly named to avoid confusion.
2021-06-04 22:47:26 +00:00
return rb->gi.voxel_gi_buffer;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
}
Rename GI Classes * GIProbe is now VoxelGI * BakedLightmap is now LightmapGI As godot adds more ways to provide GI (as an example, SDFGI in 4.0), the different techniques (which have different pros/cons) need to be properly named to avoid confusion.
2021-06-04 22:47:26 +00:00
RID RendererSceneRenderRD::render_buffers_get_default_voxel_gi_buffer() {
return gi.default_voxel_gi_buffer;
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
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());
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
int32_t probe_divisor = rb->sdfgi->cascade_size / RendererSceneGIRD::SDFGI::PROBE_DIVISOR;
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->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;
}
Use 32bit instead of 64bit 3D render buffer on mobile
2021-06-17 10:37:08 +00:00
RD::DataFormat RendererSceneRenderRD::_render_buffers_get_color_format() {
return RD::DATA_FORMAT_R16G16B16A16_SFLOAT;
}
Fixes to forward mobile * use valid format for framebuffer: VK_FORMAT_A2B10G10R10_UNORM_PACK32 * Unfortunately cant be used for compute. * Mobile will need to do refprobe, sky, mipmapblurring using raster.
2021-07-15 20:06:33 +00:00
bool RendererSceneRenderRD::_render_buffers_can_be_storage() {
return true;
}
Add stereoscopic rendering through multiview
2021-05-07 13:19:04 +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, uint32_t p_view_count) {
Fix various typos with codespell Found via `codespell -q 3 -S ./thirdparty,*.po,./DONORS.md -L ackward,ang,ans,ba,beng,cas,childs,childrens,dof,doubleclick,fave,findn,hist,inout,leapyear,lod,nd,numer,ois,ony,paket,seeked,sinc,switchs,te,uint`
2021-07-07 15:17:32 +00:00
ERR_FAIL_COND_MSG(p_view_count == 0, "Must have at least 1 view");
Add stereoscopic rendering through multiview
2021-05-07 13:19:04 +00:00
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;
Add stereoscopic rendering through multiview
2021-05-07 13:19:04 +00:00
rb->view_count = p_view_count;
Move clustered renderer functionality
2021-03-23 10:46:31 +00:00
if (is_clustered_enabled()) {
if (rb->cluster_builder == nullptr) {
rb->cluster_builder = memnew(ClusterBuilderRD);
}
rb->cluster_builder->set_shared(&cluster_builder_shared);
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
}
-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;
Add stereoscopic rendering through multiview
2021-05-07 13:19:04 +00:00
if (rb->view_count > 1) {
tf.texture_type = RD::TEXTURE_TYPE_2D_ARRAY;
}
Use 32bit instead of 64bit 3D render buffer on mobile
2021-06-17 10:37:08 +00:00
tf.format = _render_buffers_get_color_format();
-Refactored post processing, re-added glow and added a mix blend mode.
2020-01-10 00:40:26 +00:00
tf.width = rb->width;
tf.height = rb->height;
Add stereoscopic rendering through multiview
2021-05-07 13:19:04 +00:00
tf.array_layers = rb->view_count; // create a layer for every view
Use subpasses to do 3D rendering and resolve in mobile renderer
2021-06-26 10:49:25 +00:00
tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | (_render_buffers_can_be_storage() ? RD::TEXTURE_USAGE_STORAGE_BIT : 0) | RD::TEXTURE_USAGE_COLOR_ATTACHMENT_BIT;
Implement MSAA
2020-04-12 18:33:57 +00:00
if (rb->msaa != RS::VIEWPORT_MSAA_DISABLED) {
Use subpasses to do 3D rendering and resolve in mobile renderer
2021-06-26 10:49:25 +00:00
tf.usage_bits |= RD::TEXTURE_USAGE_CAN_COPY_TO_BIT;
Implement MSAA
2020-04-12 18:33:57 +00:00
}
Use subpasses to do 3D rendering and resolve in mobile renderer
2021-06-26 10:49:25 +00:00
tf.usage_bits |= RD::TEXTURE_USAGE_INPUT_ATTACHMENT_BIT; // only needed when using subpasses in the mobile renderer
-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;
Add stereoscopic rendering through multiview
2021-05-07 13:19:04 +00:00
if (rb->view_count > 1) {
tf.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
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;
Add stereoscopic rendering through multiview
2021-05-07 13:19:04 +00:00
tf.array_layers = rb->view_count; // create a layer for every view
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
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());
}
Use fragment shader instead of compute shader for effects for mobile renderer
2021-07-20 11:40:16 +00:00
if (!_render_buffers_can_be_storage()) {
// ONLY USED ON MOBILE RENDERER, ONLY USED FOR POST EFFECTS!
Vector<RID> fb;
fb.push_back(rb->texture);
rb->texture_fb = RD::get_singleton()->framebuffer_create(fb, RenderingDevice::INVALID_ID, rb->view_count);
}
Use subpasses to do 3D rendering and resolve in mobile renderer
2021-06-26 10:49:25 +00:00
RID target_texture = storage->render_target_get_rd_texture(rb->render_target);
rb->data->configure(rb->texture, rb->depth_texture, target_texture, p_width, p_height, p_msaa, p_view_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
Move clustered renderer functionality
2021-03-23 10:46:31 +00:00
if (is_clustered_enabled()) {
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);
}
Implement more rendering options as specialization constants * Shadow quality settings now specialization constant. * Decal and light projector filters can be set. * Changing those settings forces re-creation of the pipelines. These changes should help improve performance related to shadow mapping, and allows improving performance by sacrificing decal and light projector quality.
2021-07-19 19:41:55 +00:00
_update_shader_quality_settings();
Add proper quality settings to soft shadows
2020-04-10 09:30:36 +00:00
}
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);
}
Implement more rendering options as specialization constants * Shadow quality settings now specialization constant. * Decal and light projector filters can be set. * Changing those settings forces re-creation of the pipelines. These changes should help improve performance related to shadow mapping, and allows improving performance by sacrificing decal and light projector quality.
2021-07-19 19:41:55 +00:00
_update_shader_quality_settings();
}
void RendererSceneRenderRD::decals_set_filter(RenderingServer::DecalFilter p_filter) {
if (decals_filter == p_filter) {
return;
}
decals_filter = p_filter;
_update_shader_quality_settings();
}
void RendererSceneRenderRD::light_projectors_set_filter(RenderingServer::LightProjectorFilter p_filter) {
if (light_projectors_filter == p_filter) {
return;
}
light_projectors_filter = p_filter;
_update_shader_quality_settings();
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 {
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
return sky.roughness_layers;
Environment sky more or less working.
2019-08-26 20:43:58 +00:00
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
bool RendererSceneRenderRD::is_using_radiance_cubemap_array() const {
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
return sky.sky_use_cubemap_array;
Environment sky more or less working.
2019-08-26 20:43:58 +00:00
}
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;
}
Rename Transform to Transform3D in core
2020-10-17 05:08:21 +00:00
void RendererSceneRenderRD::_setup_reflections(const PagedArray<RID> &p_reflections, const Transform3D &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);
}
Fixes to forward mobile * use valid format for framebuffer: VK_FORMAT_A2B10G10R10_UNORM_PACK32 * Unfortunately cant be used for compute. * Mobile will need to do refprobe, sky, mipmapblurring using raster.
2021-07-15 20:06:33 +00:00
bool using_forward_ids = _uses_forward_ids();
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.reflection_count; i++) {
ReflectionProbeInstance *rpi = cluster.reflection_sort[i].instance;
Fixes to forward mobile * use valid format for framebuffer: VK_FORMAT_A2B10G10R10_UNORM_PACK32 * Unfortunately cant be used for compute. * Mobile will need to do refprobe, sky, mipmapblurring using raster.
2021-07-15 20:06:33 +00:00
if (using_forward_ids) {
_map_forward_id(FORWARD_ID_TYPE_REFLECTION_PROBE, rpi->forward_id, 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);
Create mobile renderer
2021-03-10 11:23:55 +00:00
rpi->cull_mask = storage->reflection_probe_get_cull_mask(base_probe);
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
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;
Rename Transform to Transform3D in core
2020-10-17 05:08:21 +00:00
Transform3D transform = rpi->transform;
Transform3D 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
Create mobile renderer
2021-03-10 11:23:55 +00:00
if (current_cluster_builder != nullptr) {
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) {
Fix reflection probe
2021-04-20 13:59:46 +00:00
RD::get_singleton()->buffer_update(cluster.reflection_buffer, 0, cluster.reflection_count * sizeof(Cluster::ReflectionData), cluster.reflections, RD::BARRIER_MASK_RASTER | RD::BARRIER_MASK_COMPUTE);
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
}
}
Use specialization constants in clustered renderer * Keep track of when projector, softshadow or directional sofshadow were enabled. * Enable them via specializaton constant where it makes sense. * Re-implements soft shadows. * Re-implements light projectors.
2021-07-12 23:32:05 +00:00
void RendererSceneRenderRD::_setup_lights(const PagedArray<RID> &p_lights, const Transform3D &p_camera_transform, RID p_shadow_atlas, bool p_using_shadows, uint32_t &r_directional_light_count, uint32_t &r_positional_light_count, bool &r_directional_light_soft_shadows) {
Rename Transform to Transform3D in core
2020-10-17 05:08:21 +00:00
Transform3D inverse_transform = p_camera_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
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;
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
sky.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;
Use specialization constants in clustered renderer * Keep track of when projector, softshadow or directional sofshadow were enabled. * Enable them via specializaton constant where it makes sense. * Re-implements soft shadows. * Re-implements light projectors.
2021-07-12 23:32:05 +00:00
r_directional_light_soft_shadows = false;
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
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
if (r_directional_light_count < sky.sky_scene_state.max_directional_lights) {
RendererSceneSkyRD::SkyDirectionalLightData &sky_light_data = sky.sky_scene_state.directional_lights[r_directional_light_count];
Rename Transform to Transform3D in core
2020-10-17 05:08:21 +00:00
Transform3D 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));
Use specialization constants in clustered renderer * Keep track of when projector, softshadow or directional sofshadow were enabled. * Enable them via specializaton constant where it makes sense. * Re-implements soft shadows. * Re-implements light projectors.
2021-07-12 23:32:05 +00:00
if (storage->light_has_shadow(base)) {
r_directional_light_soft_shadows = true;
}
Add sky_only setting to DirectionalLight3Ds
2020-10-22 06:22:53 +00:00
} else {
angular_diameter = 0.0;
}
sky_light_data.size = angular_diameter;
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
sky.sky_scene_state.ubo.directional_light_count++;
Add sky_only setting to DirectionalLight3Ds
2020-10-22 06:22:53 +00:00
}
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];
Rename Transform to Transform3D in core
2020-10-17 05:08:21 +00:00
Transform3D 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);
Rename Transform to Transform3D in core
2020-10-17 05:08:21 +00:00
Transform3D 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;
Fix directional shadow bias * Simplified code a lot, bias based on normalized cascade size. * Lets scale cascades, max distance, etc. without creating acne. * Fixed normal biasing in directional shadows. I removed normal biasing in both omni and spot shadows, since the technique can't be easily implemented there. Will need to be replaced by something else.
2021-07-29 16:30:39 +00:00
light_data.shadow_bias[j] = storage->light_get_param(base, RS::LIGHT_PARAM_SHADOW_BIAS) / 100.0 * 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;
Assorted fixes to UV unwrapping and GPU lightmapper Various fixes to UV2 unwrapping and the GPU lightmapper. Listed here for context in case of git blame/bisect: * Fix UV2 unwrapping on import, also cleaned up the unwrap cache code. * Fix saving of RGBA images in EXR format. * Fixes to the GPU lightmapper: - Added padding between atlas elements, avoids bleeding. - Remove old SDF generation code. - Fix baked attenuation for Omni/Spot lights. - Fix baking of material properties onto UV2 (wireframe was wrongly used before). - Disable statically baked lights for objects that have a lightmap texture to avoid applying the same light twice. - Fix lightmap pairing in RendererSceneCull. - Fix UV2 array generated from `RenderingServer::mesh_surface_get_arrays()`. - Port autoexposure fix for OIDN from 3.x. - Save debug textures as EXR when using floating point format.
2021-04-25 21:36:39 +00:00
light_data.bake_mode = storage->light_get_bake_mode(base);
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
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
Fixes to forward mobile * use valid format for framebuffer: VK_FORMAT_A2B10G10R10_UNORM_PACK32 * Unfortunately cant be used for compute. * Mobile will need to do refprobe, sky, mipmapblurring using raster.
2021-07-15 20:06:33 +00:00
bool using_forward_ids = _uses_forward_ids();
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
Fixes to forward mobile * use valid format for framebuffer: VK_FORMAT_A2B10G10R10_UNORM_PACK32 * Unfortunately cant be used for compute. * Mobile will need to do refprobe, sky, mipmapblurring using raster.
2021-07-15 20:06:33 +00:00
if (using_forward_ids) {
_map_forward_id(type == RS::LIGHT_OMNI ? FORWARD_ID_TYPE_OMNI_LIGHT : FORWARD_ID_TYPE_SPOT_LIGHT, li->forward_id, index);
}
Rename Transform to Transform3D in core
2020-10-17 05:08:21 +00:00
Transform3D 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;
Assorted fixes to UV unwrapping and GPU lightmapper Various fixes to UV2 unwrapping and the GPU lightmapper. Listed here for context in case of git blame/bisect: * Fix UV2 unwrapping on import, also cleaned up the unwrap cache code. * Fix saving of RGBA images in EXR format. * Fixes to the GPU lightmapper: - Added padding between atlas elements, avoids bleeding. - Remove old SDF generation code. - Fix baked attenuation for Omni/Spot lights. - Fix baking of material properties onto UV2 (wireframe was wrongly used before). - Disable statically baked lights for objects that have a lightmap texture to avoid applying the same light twice. - Fix lightmap pairing in RendererSceneCull. - Fix UV2 array generated from `RenderingServer::mesh_surface_get_arrays()`. - Port autoexposure fix for OIDN from 3.x. - Save debug textures as EXR when using floating point format.
2021-04-25 21:36:39 +00:00
light_data.bake_mode = storage->light_get_bake_mode(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
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
Invert spotlight angle attenuation Inverted the spotlight angle attenuation so a higher value results in a dimmer light, this makes it more consistent with the distance attenuation. Also changed the way spotlighs are computed in SDFGI and GIPorbes and GPU lightmapper, now it matches the falloff used in the scene rendering code.
2021-02-06 20:39:08 +00:00
light_data.inv_spot_attenuation = 1.0f / storage->light_get_param(base, RS::LIGHT_PARAM_SPOT_ATTENUATION);
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 spot_angle = storage->light_get_param(base, RS::LIGHT_PARAM_SPOT_ANGLE);
Invert spotlight angle attenuation Inverted the spotlight angle attenuation so a higher value results in a dimmer light, this makes it more consistent with the distance attenuation. Also changed the way spotlighs are computed in SDFGI and GIPorbes and GPU lightmapper, now it matches the falloff used in the scene rendering code.
2021-02-06 20:39:08 +00:00
light_data.cos_spot_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) {
Fix directional shadow bias * Simplified code a lot, bias based on normalized cascade size. * Lets scale cascades, max distance, etc. without creating acne. * Fixed normal biasing in directional shadows. I removed normal biasing in both omni and spot shadows, since the technique can't be easily implemented there. Will need to be replaced by something else.
2021-07-29 16:30:39 +00:00
light_data.shadow_bias = (storage->light_get_param(base, RS::LIGHT_PARAM_SHADOW_BIAS) / 100.0);
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
} else { //omni
Fix directional shadow bias * Simplified code a lot, bias based on normalized cascade size. * Lets scale cascades, max distance, etc. without creating acne. * Fixed normal biasing in directional shadows. I removed normal biasing in both omni and spot shadows, since the technique can't be easily implemented there. Will need to be replaced by something else.
2021-07-29 16:30:39 +00:00
light_data.shadow_bias = storage->light_get_param(base, RS::LIGHT_PARAM_SHADOW_BIAS) / 100.0;
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
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
Rename Transform to Transform3D in core
2020-10-17 05:08:21 +00:00
Transform3D proj = (inverse_transform * light_transform).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
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) {
Rename Transform to Transform3D in core
2020-10-17 05:08:21 +00:00
Transform3D modelview = (inverse_transform * light_transform).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
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
}
Create mobile renderer
2021-03-10 11:23:55 +00:00
li->cull_mask = storage->light_get_cull_mask(base);
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
Create mobile renderer
2021-03-10 11:23:55 +00:00
if (current_cluster_builder != nullptr) {
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
}
Rewrote how barriers work for faster rendering -Added more finegrained control in RenderingDevice API -Optimized barriers (use less ones for thee same) -General optimizations -Shadows render all together unbarriered -GI can render together with shadows. -SDFGI can render together with depth-preoass. -General fixes -Added GPU detection
2021-02-02 19:51:36 +00:00
//update without barriers
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) {
Fixes to get Godot running again on Intel IGP -Fixed strange bug with shadowed instance_param (this should not have worked anywhere, odd..) -Cleaned up barrier usage further.
2021-02-05 03:02:06 +00:00
RD::get_singleton()->buffer_update(cluster.omni_light_buffer, 0, sizeof(Cluster::LightData) * cluster.omni_light_count, cluster.omni_lights, RD::BARRIER_MASK_RASTER | RD::BARRIER_MASK_COMPUTE);
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) {
Fixes to get Godot running again on Intel IGP -Fixed strange bug with shadowed instance_param (this should not have worked anywhere, odd..) -Cleaned up barrier usage further.
2021-02-05 03:02:06 +00:00
RD::get_singleton()->buffer_update(cluster.spot_light_buffer, 0, sizeof(Cluster::LightData) * cluster.spot_light_count, cluster.spot_lights, RD::BARRIER_MASK_RASTER | RD::BARRIER_MASK_COMPUTE);
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) {
Fixes to get Godot running again on Intel IGP -Fixed strange bug with shadowed instance_param (this should not have worked anywhere, odd..) -Cleaned up barrier usage further.
2021-02-05 03:02:06 +00:00
RD::get_singleton()->buffer_update(cluster.directional_light_buffer, 0, sizeof(Cluster::DirectionalLightData) * r_directional_light_count, cluster.directional_lights, RD::BARRIER_MASK_RASTER | RD::BARRIER_MASK_COMPUTE);
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
}
}
Rename Transform to Transform3D in core
2020-10-17 05:08:21 +00:00
void RendererSceneRenderRD::_setup_decals(const PagedArray<RID> &p_decals, const Transform3D &p_camera_inverse_xform) {
Transform3D uv_xform;
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
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
Rename Transform to Transform3D in core
2020-10-17 05:08:21 +00:00
Transform3D 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);
}
Fixes to forward mobile * use valid format for framebuffer: VK_FORMAT_A2B10G10R10_UNORM_PACK32 * Unfortunately cant be used for compute. * Mobile will need to do refprobe, sky, mipmapblurring using raster.
2021-07-15 20:06:33 +00:00
bool using_forward_ids = _uses_forward_ids();
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.decal_count; i++) {
DecalInstance *di = cluster.decal_sort[i].instance;
RID decal = di->decal;
Fixes to forward mobile * use valid format for framebuffer: VK_FORMAT_A2B10G10R10_UNORM_PACK32 * Unfortunately cant be used for compute. * Mobile will need to do refprobe, sky, mipmapblurring using raster.
2021-07-15 20:06:33 +00:00
if (using_forward_ids) {
_map_forward_id(FORWARD_ID_TYPE_DECAL, di->forward_id, i);
}
Create mobile renderer
2021-03-10 11:23:55 +00:00
di->cull_mask = storage->decal_get_cull_mask(decal);
Rename Transform to Transform3D in core
2020-10-17 05:08:21 +00:00
Transform3D xform = di->transform;
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 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);
Rename Transform to Transform3D in core
2020-10-17 05:08:21 +00:00
Transform3D scale_xform;
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
scale_xform.basis.scale(Vector3(decal_extents.x, decal_extents.y, decal_extents.z));
Rename Transform to Transform3D in core
2020-10-17 05:08:21 +00:00
Transform3D 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);
Create mobile renderer
2021-03-10 11:23:55 +00:00
if (current_cluster_builder != nullptr) {
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) {
Fixes to get Godot running again on Intel IGP -Fixed strange bug with shadowed instance_param (this should not have worked anywhere, odd..) -Cleaned up barrier usage further.
2021-02-05 03:02:06 +00:00
RD::get_singleton()->buffer_update(cluster.decal_buffer, 0, sizeof(Cluster::DecalData) * cluster.decal_count, cluster.decals, RD::BARRIER_MASK_RASTER | RD::BARRIER_MASK_COMPUTE);
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);
Added temporal reprojection to Volumetric Fog -It's an option, just enable it -Just works, don't have to do anything else.
2021-02-05 13:51:49 +00:00
RD::get_singleton()->free(rb->volumetric_fog->prev_light_density_map);
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
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;
}
Rename GI Classes * GIProbe is now VoxelGI * BakedLightmap is now LightmapGI As godot adds more ways to provide GI (as an example, SDFGI in 4.0), the different techniques (which have different pros/cons) need to be properly named to avoid confusion.
2021-06-04 22:47:26 +00:00
void RendererSceneRenderRD::_update_volumetric_fog(RID p_render_buffers, RID p_environment, const CameraMatrix &p_cam_projection, const Transform3D &p_cam_transform, RID p_shadow_atlas, int p_directional_light_count, bool p_use_directional_shadows, int p_positional_light_count, int p_voxel_gi_count) {
Move clustered renderer functionality
2021-03-23 10:46:31 +00:00
ERR_FAIL_COND(!is_clustered_enabled()); // can't use volumetric fog without clustered
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);
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
RendererSceneEnvironmentRD *env = environment_owner.getornull(p_environment);
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
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);
}
}
if (!env || !env->volumetric_fog_enabled) {
//no reason to enable or update, bye
return;
}
Rewrote how barriers work for faster rendering -Added more finegrained control in RenderingDevice API -Optimized barriers (use less ones for thee same) -General optimizations -Shadows render all together unbarriered -GI can render together with shadows. -SDFGI can render together with depth-preoass. -General fixes -Added GPU detection
2021-02-02 19:51:36 +00:00
RENDER_TIMESTAMP(">Volumetric Fog");
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
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 temporal reprojection to Volumetric Fog -It's an option, just enable it -Just works, don't have to do anything else.
2021-02-05 13:51:49 +00:00
tf.usage_bits = RD::TEXTURE_USAGE_STORAGE_BIT | RD::TEXTURE_USAGE_CAN_COPY_FROM_BIT;
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
rb->volumetric_fog->light_density_map = RD::get_singleton()->texture_create(tf, RD::TextureView());
Added temporal reprojection to Volumetric Fog -It's an option, just enable it -Just works, don't have to do anything else.
2021-02-05 13:51:49 +00:00
tf.usage_bits = RD::TEXTURE_USAGE_STORAGE_BIT | RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_CAN_COPY_TO_BIT;
rb->volumetric_fog->prev_light_density_map = RD::get_singleton()->texture_create(tf, RD::TextureView());
RD::get_singleton()->texture_clear(rb->volumetric_fog->prev_light_density_map, Color(0, 0, 0, 0), 0, 1, 0, 1);
tf.usage_bits = RD::TEXTURE_USAGE_STORAGE_BIT | RD::TEXTURE_USAGE_SAMPLING_BIT;
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
rb->volumetric_fog->fog_map = RD::get_singleton()->texture_create(tf, RD::TextureView());
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);
}
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
rb->volumetric_fog->sky_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, sky.sky_shader.default_shader_rd, RendererSceneSkyRD::SKY_SET_FOG);
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
}
//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;
Simplify Volumetric Fog -Always use temporal reproject, it just loos way better than any other filter. -By always using termporal reproject, the shadowmap reduction can be done away with, massively improving performance. -Disadvantage of temporal reproject is update latency so.. -Made sure a gaussian filter runs in XY after fog, this allows to keep stability and lower latency.
2021-02-06 14:51:56 +00:00
ShadowAtlas *shadow_atlas = shadow_atlas_owner.getornull(p_shadow_atlas);
if (shadow_atlas == nullptr || shadow_atlas->depth.is_null()) {
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 {
Simplify Volumetric Fog -Always use temporal reproject, it just loos way better than any other filter. -By always using termporal reproject, the shadowmap reduction can be done away with, massively improving performance. -Disadvantage of temporal reproject is update latency so.. -Made sure a gaussian filter runs in XY after fog, this allows to keep stability and lower latency.
2021-02-06 14:51:56 +00:00
u.ids.push_back(shadow_atlas->depth);
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_TEXTURE;
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
u.binding = 2;
Simplify Volumetric Fog -Always use temporal reproject, it just loos way better than any other filter. -By always using termporal reproject, the shadowmap reduction can be done away with, massively improving performance. -Disadvantage of temporal reproject is update latency so.. -Made sure a gaussian filter runs in XY after fog, this allows to keep stability and lower latency.
2021-02-06 14:51:56 +00:00
if (directional_shadow.depth.is_valid()) {
u.ids.push_back(directional_shadow.depth);
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
} else {
Simplify Volumetric Fog -Always use temporal reproject, it just loos way better than any other filter. -By always using termporal reproject, the shadowmap reduction can be done away with, massively improving performance. -Disadvantage of temporal reproject is update latency so.. -Made sure a gaussian filter runs in XY after fog, this allows to keep stability and lower latency.
2021-02-06 14:51:56 +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
}
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;
Rename GI Classes * GIProbe is now VoxelGI * BakedLightmap is now LightmapGI As godot adds more ways to provide GI (as an example, SDFGI in 4.0), the different techniques (which have different pros/cons) need to be properly named to avoid confusion.
2021-06-04 22:47:26 +00:00
u.ids.push_back(render_buffers_get_voxel_gi_buffer(p_render_buffers));
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_TEXTURE;
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
u.binding = 12;
Rename GI Classes * GIProbe is now VoxelGI * BakedLightmap is now LightmapGI As godot adds more ways to provide GI (as an example, SDFGI in 4.0), the different techniques (which have different pros/cons) need to be properly named to avoid confusion.
2021-06-04 22:47:26 +00:00
for (int i = 0; i < RendererSceneGIRD::MAX_VOXEL_GI_INSTANCES; i++) {
u.ids.push_back(rb->gi.voxel_gi_textures[i]);
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 = 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 temporal reprojection to Volumetric Fog -It's an option, just enable it -Just works, don't have to do anything else.
2021-02-05 13:51:49 +00:00
{
RD::Uniform u;
u.uniform_type = RD::UNIFORM_TYPE_TEXTURE;
u.binding = 15;
u.ids.push_back(rb->volumetric_fog->prev_light_density_map);
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;
Rename GI Classes * GIProbe is now VoxelGI * BakedLightmap is now LightmapGI As godot adds more ways to provide GI (as an example, SDFGI in 4.0), the different techniques (which have different pros/cons) need to be properly named to avoid confusion.
2021-06-04 22:47:26 +00:00
params.max_voxel_gi_instances = env->volumetric_fog_gi_inject > 0.001 ? p_voxel_gi_count : 0;
Added temporal reprojection to Volumetric Fog -It's an option, just enable it -Just works, don't have to do anything else.
2021-02-05 13:51:49 +00:00
params.temporal_frame = RSG::rasterizer->get_frame_number() % VolumetricFog::MAX_TEMPORAL_FRAMES;
Rename Transform to Transform3D in core
2020-10-17 05:08:21 +00:00
Transform3D to_prev_cam_view = rb->volumetric_fog->prev_cam_transform.affine_inverse() * p_cam_transform;
Added temporal reprojection to Volumetric Fog -It's an option, just enable it -Just works, don't have to do anything else.
2021-02-05 13:51:49 +00:00
storage->store_transform(to_prev_cam_view, params.to_prev_view);
params.use_temporal_reprojection = env->volumetric_fog_temporal_reprojection;
params.temporal_blend = env->volumetric_fog_temporal_reprojection_amount;
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
Rewrote how barriers work for faster rendering -Added more finegrained control in RenderingDevice API -Optimized barriers (use less ones for thee same) -General optimizations -Shadows render all together unbarriered -GI can render together with shadows. -SDFGI can render together with depth-preoass. -General fixes -Added GPU detection
2021-02-02 19:51:36 +00:00
RD::get_singleton()->draw_command_begin_label("Render Volumetric Fog");
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("Render Fog");
Rewrote how barriers work for faster rendering -Added more finegrained control in RenderingDevice API -Optimized barriers (use less ones for thee same) -General optimizations -Shadows render all together unbarriered -GI can render together with shadows. -SDFGI can render together with depth-preoass. -General fixes -Added GPU detection
2021-02-02 19:51:36 +00:00
RD::get_singleton()->buffer_update(volumetric_fog.params_ubo, 0, sizeof(VolumetricFogShader::ParamsUBO), &params, RD::BARRIER_MASK_COMPUTE);
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);
}
Rewrote how barriers work for faster rendering -Added more finegrained control in RenderingDevice API -Optimized barriers (use less ones for thee same) -General optimizations -Shadows render all together unbarriered -GI can render together with shadows. -SDFGI can render together with depth-preoass. -General fixes -Added GPU detection
2021-02-02 19:51:36 +00:00
RD::get_singleton()->compute_list_dispatch_threads(compute_list, rb->volumetric_fog->width, rb->volumetric_fog->height, rb->volumetric_fog->depth);
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
Rewrote how barriers work for faster rendering -Added more finegrained control in RenderingDevice API -Optimized barriers (use less ones for thee same) -General optimizations -Shadows render all together unbarriered -GI can render together with shadows. -SDFGI can render together with depth-preoass. -General fixes -Added GPU detection
2021-02-02 19:51:36 +00:00
RD::get_singleton()->draw_command_end_label();
Added temporal reprojection to Volumetric Fog -It's an option, just enable it -Just works, don't have to do anything else.
2021-02-05 13:51:49 +00:00
RD::get_singleton()->compute_list_end();
RD::get_singleton()->texture_copy(rb->volumetric_fog->light_density_map, rb->volumetric_fog->prev_light_density_map, Vector3(0, 0, 0), Vector3(0, 0, 0), Vector3(rb->volumetric_fog->width, rb->volumetric_fog->height, rb->volumetric_fog->depth), 0, 0, 0, 0);
compute_list = RD::get_singleton()->compute_list_begin();
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
if (use_filter) {
Rewrote how barriers work for faster rendering -Added more finegrained control in RenderingDevice API -Optimized barriers (use less ones for thee same) -General optimizations -Shadows render all together unbarriered -GI can render together with shadows. -SDFGI can render together with depth-preoass. -General fixes -Added GPU detection
2021-02-02 19:51:36 +00:00
RD::get_singleton()->draw_command_begin_label("Filter Fog");
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);
Rewrote how barriers work for faster rendering -Added more finegrained control in RenderingDevice API -Optimized barriers (use less ones for thee same) -General optimizations -Shadows render all together unbarriered -GI can render together with shadows. -SDFGI can render together with depth-preoass. -General fixes -Added GPU detection
2021-02-02 19:51:36 +00:00
RD::get_singleton()->compute_list_dispatch_threads(compute_list, rb->volumetric_fog->width, rb->volumetric_fog->height, 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
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);
}
Rewrote how barriers work for faster rendering -Added more finegrained control in RenderingDevice API -Optimized barriers (use less ones for thee same) -General optimizations -Shadows render all together unbarriered -GI can render together with shadows. -SDFGI can render together with depth-preoass. -General fixes -Added GPU detection
2021-02-02 19:51:36 +00:00
RD::get_singleton()->compute_list_dispatch_threads(compute_list, rb->volumetric_fog->width, rb->volumetric_fog->height, rb->volumetric_fog->depth);
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
RD::get_singleton()->compute_list_add_barrier(compute_list);
Rewrote how barriers work for faster rendering -Added more finegrained control in RenderingDevice API -Optimized barriers (use less ones for thee same) -General optimizations -Shadows render all together unbarriered -GI can render together with shadows. -SDFGI can render together with depth-preoass. -General fixes -Added GPU detection
2021-02-02 19:51:36 +00:00
RD::get_singleton()->draw_command_end_label();
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
RENDER_TIMESTAMP("Integrate Fog");
Rewrote how barriers work for faster rendering -Added more finegrained control in RenderingDevice API -Optimized barriers (use less ones for thee same) -General optimizations -Shadows render all together unbarriered -GI can render together with shadows. -SDFGI can render together with depth-preoass. -General fixes -Added GPU detection
2021-02-02 19:51:36 +00:00
RD::get_singleton()->draw_command_begin_label("Integrate Fog");
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::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);
Rewrote how barriers work for faster rendering -Added more finegrained control in RenderingDevice API -Optimized barriers (use less ones for thee same) -General optimizations -Shadows render all together unbarriered -GI can render together with shadows. -SDFGI can render together with depth-preoass. -General fixes -Added GPU detection
2021-02-02 19:51:36 +00:00
RD::get_singleton()->compute_list_dispatch_threads(compute_list, rb->volumetric_fog->width, rb->volumetric_fog->height, 1);
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
Rewrote how barriers work for faster rendering -Added more finegrained control in RenderingDevice API -Optimized barriers (use less ones for thee same) -General optimizations -Shadows render all together unbarriered -GI can render together with shadows. -SDFGI can render together with depth-preoass. -General fixes -Added GPU detection
2021-02-02 19:51:36 +00:00
RD::get_singleton()->compute_list_end(RD::BARRIER_MASK_RASTER);
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");
Rewrote how barriers work for faster rendering -Added more finegrained control in RenderingDevice API -Optimized barriers (use less ones for thee same) -General optimizations -Shadows render all together unbarriered -GI can render together with shadows. -SDFGI can render together with depth-preoass. -General fixes -Added GPU detection
2021-02-02 19:51:36 +00:00
RD::get_singleton()->draw_command_end_label();
Added temporal reprojection to Volumetric Fog -It's an option, just enable it -Just works, don't have to do anything else.
2021-02-05 13:51:49 +00:00
rb->volumetric_fog->prev_cam_transform = p_cam_transform;
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
}
Split RenderDataRD struct from RenderState struct to simplify passing our render data around the renderer.
2021-05-05 05:41:12 +00:00
bool RendererSceneRenderRD::_needs_post_prepass_render(RenderDataRD *p_render_data, bool p_use_gi) {
if (p_render_data->render_buffers.is_valid()) {
RenderBuffers *rb = render_buffers_owner.getornull(p_render_data->render_buffers);
Rewrote how barriers work for faster rendering -Added more finegrained control in RenderingDevice API -Optimized barriers (use less ones for thee same) -General optimizations -Shadows render all together unbarriered -GI can render together with shadows. -SDFGI can render together with depth-preoass. -General fixes -Added GPU detection
2021-02-02 19:51:36 +00:00
if (rb->sdfgi != nullptr) {
return true;
}
-Refactored post processing, re-added glow and added a mix blend mode.
2020-01-10 00:40:26 +00:00
}
Rewrote how barriers work for faster rendering -Added more finegrained control in RenderingDevice API -Optimized barriers (use less ones for thee same) -General optimizations -Shadows render all together unbarriered -GI can render together with shadows. -SDFGI can render together with depth-preoass. -General fixes -Added GPU detection
2021-02-02 19:51:36 +00:00
return false;
}
Base 3D engine done, still untested, though.
2019-08-18 22:40:52 +00:00
Split RenderDataRD struct from RenderState struct to simplify passing our render data around the renderer.
2021-05-05 05:41:12 +00:00
void RendererSceneRenderRD::_post_prepass_render(RenderDataRD *p_render_data, bool p_use_gi) {
if (p_render_data->render_buffers.is_valid()) {
Rewrote how barriers work for faster rendering -Added more finegrained control in RenderingDevice API -Optimized barriers (use less ones for thee same) -General optimizations -Shadows render all together unbarriered -GI can render together with shadows. -SDFGI can render together with depth-preoass. -General fixes -Added GPU detection
2021-02-02 19:51:36 +00:00
if (p_use_gi) {
Split RenderDataRD struct from RenderState struct to simplify passing our render data around the renderer.
2021-05-05 05:41:12 +00:00
RenderBuffers *rb = render_buffers_owner.getornull(p_render_data->render_buffers);
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
ERR_FAIL_COND(rb == nullptr);
if (rb->sdfgi == nullptr) {
return;
}
Split RenderDataRD struct from RenderState struct to simplify passing our render data around the renderer.
2021-05-05 05:41:12 +00:00
RendererSceneEnvironmentRD *env = environment_owner.getornull(p_render_data->environment);
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
rb->sdfgi->update_probes(env, sky.sky_owner.getornull(env->sky));
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
}
}
Rewrote how barriers work for faster rendering -Added more finegrained control in RenderingDevice API -Optimized barriers (use less ones for thee same) -General optimizations -Shadows render all together unbarriered -GI can render together with shadows. -SDFGI can render together with depth-preoass. -General fixes -Added GPU detection
2021-02-02 19:51:36 +00:00
}
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
Split RenderDataRD struct from RenderState struct to simplify passing our render data around the renderer.
2021-05-05 05:41:12 +00:00
void RendererSceneRenderRD::_pre_resolve_render(RenderDataRD *p_render_data, bool p_use_gi) {
if (p_render_data->render_buffers.is_valid()) {
Rewrote how barriers work for faster rendering -Added more finegrained control in RenderingDevice API -Optimized barriers (use less ones for thee same) -General optimizations -Shadows render all together unbarriered -GI can render together with shadows. -SDFGI can render together with depth-preoass. -General fixes -Added GPU detection
2021-02-02 19:51:36 +00:00
if (p_use_gi) {
RD::get_singleton()->compute_list_end();
}
}
}
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
Rename GI Classes * GIProbe is now VoxelGI * BakedLightmap is now LightmapGI As godot adds more ways to provide GI (as an example, SDFGI in 4.0), the different techniques (which have different pros/cons) need to be properly named to avoid confusion.
2021-06-04 22:47:26 +00:00
void RendererSceneRenderRD::_pre_opaque_render(RenderDataRD *p_render_data, bool p_use_ssao, bool p_use_gi, RID p_normal_roughness_buffer, RID p_voxel_gi_buffer) {
Rewrote how barriers work for faster rendering -Added more finegrained control in RenderingDevice API -Optimized barriers (use less ones for thee same) -General optimizations -Shadows render all together unbarriered -GI can render together with shadows. -SDFGI can render together with depth-preoass. -General fixes -Added GPU detection
2021-02-02 19:51:36 +00:00
// Render shadows while GI is rendering, due to how barriers are handled, this should happen at the same time
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
Split RenderDataRD struct from RenderState struct to simplify passing our render data around the renderer.
2021-05-05 05:41:12 +00:00
if (p_render_data->render_buffers.is_valid() && p_use_gi) {
RenderBuffers *rb = render_buffers_owner.getornull(p_render_data->render_buffers);
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
ERR_FAIL_COND(rb == nullptr);
Fix VoxelGI (works again). It was broken due to #46046, this makes it work again.
2021-06-12 23:37:56 +00:00
if (rb->sdfgi != nullptr) {
rb->sdfgi->store_probes();
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
}
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
}
Rewrote how barriers work for faster rendering -Added more finegrained control in RenderingDevice API -Optimized barriers (use less ones for thee same) -General optimizations -Shadows render all together unbarriered -GI can render together with shadows. -SDFGI can render together with depth-preoass. -General fixes -Added GPU detection
2021-02-02 19:51:36 +00:00
render_state.cube_shadows.clear();
render_state.shadows.clear();
render_state.directional_shadows.clear();
Split RenderDataRD struct from RenderState struct to simplify passing our render data around the renderer.
2021-05-05 05:41:12 +00:00
Plane camera_plane(p_render_data->cam_transform.origin, -p_render_data->cam_transform.basis.get_axis(Vector3::AXIS_Z));
float lod_distance_multiplier = p_render_data->cam_projection.get_lod_multiplier();
Rewrote how barriers work for faster rendering -Added more finegrained control in RenderingDevice API -Optimized barriers (use less ones for thee same) -General optimizations -Shadows render all together unbarriered -GI can render together with shadows. -SDFGI can render together with depth-preoass. -General fixes -Added GPU detection
2021-02-02 19:51:36 +00:00
{
for (int i = 0; i < render_state.render_shadow_count; i++) {
LightInstance *li = light_instance_owner.getornull(render_state.render_shadows[i].light);
if (storage->light_get_type(li->light) == RS::LIGHT_DIRECTIONAL) {
render_state.directional_shadows.push_back(i);
} else if (storage->light_get_type(li->light) == RS::LIGHT_OMNI && storage->light_omni_get_shadow_mode(li->light) == RS::LIGHT_OMNI_SHADOW_CUBE) {
render_state.cube_shadows.push_back(i);
} else {
render_state.shadows.push_back(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
}
Rewrote how barriers work for faster rendering -Added more finegrained control in RenderingDevice API -Optimized barriers (use less ones for thee same) -General optimizations -Shadows render all together unbarriered -GI can render together with shadows. -SDFGI can render together with depth-preoass. -General fixes -Added GPU detection
2021-02-02 19:51:36 +00:00
//cube shadows are rendered in their own way
for (uint32_t i = 0; i < render_state.cube_shadows.size(); i++) {
Fix Render Info * Fixed and redone the process to obtain render information from a viewport * Some stats, such as material changes are too difficult to guess on Vulkan, were removed. * Separated visible and shadow stats, which causes confusion. * Texture, buffer and general video memory can be queried now. * Fixed the performance metrics too.
2021-07-02 23:14:19 +00:00
_render_shadow_pass(render_state.render_shadows[render_state.cube_shadows[i]].light, p_render_data->shadow_atlas, render_state.render_shadows[render_state.cube_shadows[i]].pass, render_state.render_shadows[render_state.cube_shadows[i]].instances, camera_plane, lod_distance_multiplier, p_render_data->screen_lod_threshold, true, true, true, p_render_data->render_info);
Rewrote how barriers work for faster rendering -Added more finegrained control in RenderingDevice API -Optimized barriers (use less ones for thee same) -General optimizations -Shadows render all together unbarriered -GI can render together with shadows. -SDFGI can render together with depth-preoass. -General fixes -Added GPU detection
2021-02-02 19:51:36 +00:00
}
if (render_state.directional_shadows.size()) {
//open the pass for directional shadows
_update_directional_shadow_atlas();
RD::get_singleton()->draw_list_begin(directional_shadow.fb, RD::INITIAL_ACTION_DROP, RD::FINAL_ACTION_DISCARD, RD::INITIAL_ACTION_CLEAR, RD::FINAL_ACTION_CONTINUE);
RD::get_singleton()->draw_list_end();
}
}
// Render GI
bool render_shadows = render_state.directional_shadows.size() || render_state.shadows.size();
Split RenderDataRD struct from RenderState struct to simplify passing our render data around the renderer.
2021-05-05 05:41:12 +00:00
bool render_gi = p_render_data->render_buffers.is_valid() && p_use_gi;
Rewrote how barriers work for faster rendering -Added more finegrained control in RenderingDevice API -Optimized barriers (use less ones for thee same) -General optimizations -Shadows render all together unbarriered -GI can render together with shadows. -SDFGI can render together with depth-preoass. -General fixes -Added GPU detection
2021-02-02 19:51:36 +00:00
if (render_shadows && render_gi) {
RENDER_TIMESTAMP("Render GI + Render Shadows (parallel)");
} else if (render_shadows) {
RENDER_TIMESTAMP("Render Shadows");
} else if (render_gi) {
RENDER_TIMESTAMP("Render GI");
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
}
Rewrote how barriers work for faster rendering -Added more finegrained control in RenderingDevice API -Optimized barriers (use less ones for thee same) -General optimizations -Shadows render all together unbarriered -GI can render together with shadows. -SDFGI can render together with depth-preoass. -General fixes -Added GPU detection
2021-02-02 19:51:36 +00:00
//prepare shadow rendering
if (render_shadows) {
_render_shadow_begin();
//render directional shadows
for (uint32_t i = 0; i < render_state.directional_shadows.size(); i++) {
Fix Render Info * Fixed and redone the process to obtain render information from a viewport * Some stats, such as material changes are too difficult to guess on Vulkan, were removed. * Separated visible and shadow stats, which causes confusion. * Texture, buffer and general video memory can be queried now. * Fixed the performance metrics too.
2021-07-02 23:14:19 +00:00
_render_shadow_pass(render_state.render_shadows[render_state.directional_shadows[i]].light, p_render_data->shadow_atlas, render_state.render_shadows[render_state.directional_shadows[i]].pass, render_state.render_shadows[render_state.directional_shadows[i]].instances, camera_plane, lod_distance_multiplier, p_render_data->screen_lod_threshold, false, i == render_state.directional_shadows.size() - 1, false, p_render_data->render_info);
Rewrote how barriers work for faster rendering -Added more finegrained control in RenderingDevice API -Optimized barriers (use less ones for thee same) -General optimizations -Shadows render all together unbarriered -GI can render together with shadows. -SDFGI can render together with depth-preoass. -General fixes -Added GPU detection
2021-02-02 19:51:36 +00:00
}
//render positional shadows
for (uint32_t i = 0; i < render_state.shadows.size(); i++) {
Fix Render Info * Fixed and redone the process to obtain render information from a viewport * Some stats, such as material changes are too difficult to guess on Vulkan, were removed. * Separated visible and shadow stats, which causes confusion. * Texture, buffer and general video memory can be queried now. * Fixed the performance metrics too.
2021-07-02 23:14:19 +00:00
_render_shadow_pass(render_state.render_shadows[render_state.shadows[i]].light, p_render_data->shadow_atlas, render_state.render_shadows[render_state.shadows[i]].pass, render_state.render_shadows[render_state.shadows[i]].instances, camera_plane, lod_distance_multiplier, p_render_data->screen_lod_threshold, i == 0, i == render_state.shadows.size() - 1, true, p_render_data->render_info);
Rewrote how barriers work for faster rendering -Added more finegrained control in RenderingDevice API -Optimized barriers (use less ones for thee same) -General optimizations -Shadows render all together unbarriered -GI can render together with shadows. -SDFGI can render together with depth-preoass. -General fixes -Added GPU detection
2021-02-02 19:51:36 +00:00
}
_render_shadow_process();
}
//start GI
if (render_gi) {
Rename GI Classes * GIProbe is now VoxelGI * BakedLightmap is now LightmapGI As godot adds more ways to provide GI (as an example, SDFGI in 4.0), the different techniques (which have different pros/cons) need to be properly named to avoid confusion.
2021-06-04 22:47:26 +00:00
gi.process_gi(p_render_data->render_buffers, p_normal_roughness_buffer, p_voxel_gi_buffer, p_render_data->environment, p_render_data->cam_projection, p_render_data->cam_transform, *p_render_data->voxel_gi_instances, this);
Rewrote how barriers work for faster rendering -Added more finegrained control in RenderingDevice API -Optimized barriers (use less ones for thee same) -General optimizations -Shadows render all together unbarriered -GI can render together with shadows. -SDFGI can render together with depth-preoass. -General fixes -Added GPU detection
2021-02-02 19:51:36 +00:00
}
//Do shadow rendering (in parallel with GI)
if (render_shadows) {
_render_shadow_end(RD::BARRIER_MASK_NO_BARRIER);
}
if (render_gi) {
RD::get_singleton()->compute_list_end(RD::BARRIER_MASK_NO_BARRIER); //use a later barrier
}
Split RenderDataRD struct from RenderState struct to simplify passing our render data around the renderer.
2021-05-05 05:41:12 +00:00
if (p_render_data->render_buffers.is_valid()) {
Rewrote how barriers work for faster rendering -Added more finegrained control in RenderingDevice API -Optimized barriers (use less ones for thee same) -General optimizations -Shadows render all together unbarriered -GI can render together with shadows. -SDFGI can render together with depth-preoass. -General fixes -Added GPU detection
2021-02-02 19:51:36 +00:00
if (p_use_ssao) {
Split RenderDataRD struct from RenderState struct to simplify passing our render data around the renderer.
2021-05-05 05:41:12 +00:00
_process_ssao(p_render_data->render_buffers, p_render_data->environment, p_normal_roughness_buffer, p_render_data->cam_projection);
Rewrote how barriers work for faster rendering -Added more finegrained control in RenderingDevice API -Optimized barriers (use less ones for thee same) -General optimizations -Shadows render all together unbarriered -GI can render together with shadows. -SDFGI can render together with depth-preoass. -General fixes -Added GPU detection
2021-02-02 19:51:36 +00:00
}
}
Fixes to get Godot running again on Intel IGP -Fixed strange bug with shadowed instance_param (this should not have worked anywhere, odd..) -Cleaned up barrier usage further.
2021-02-05 03:02:06 +00:00
//full barrier here, we need raster, transfer and compute and it depends from the previous work
RD::get_singleton()->barrier(RD::BARRIER_MASK_ALL, RD::BARRIER_MASK_ALL);
Rewrote how barriers work for faster rendering -Added more finegrained control in RenderingDevice API -Optimized barriers (use less ones for thee same) -General optimizations -Shadows render all together unbarriered -GI can render together with shadows. -SDFGI can render together with depth-preoass. -General fixes -Added GPU detection
2021-02-02 19:51:36 +00:00
if (current_cluster_builder) {
Split RenderDataRD struct from RenderState struct to simplify passing our render data around the renderer.
2021-05-05 05:41:12 +00:00
current_cluster_builder->begin(p_render_data->cam_transform, p_render_data->cam_projection, !p_render_data->reflection_probe.is_valid());
Rewrote how barriers work for faster rendering -Added more finegrained control in RenderingDevice API -Optimized barriers (use less ones for thee same) -General optimizations -Shadows render all together unbarriered -GI can render together with shadows. -SDFGI can render together with depth-preoass. -General fixes -Added GPU detection
2021-02-02 19:51:36 +00:00
}
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;
Split RenderDataRD struct from RenderState struct to simplify passing our render data around the renderer.
2021-05-05 05:41:12 +00:00
if (p_render_data->reflection_probe.is_valid()) {
if (!storage->reflection_probe_renders_shadows(reflection_probe_instance_get_probe(p_render_data->reflection_probe))) {
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
using_shadows = false;
}
} else {
//do not render reflections when rendering a reflection probe
Split RenderDataRD struct from RenderState struct to simplify passing our render data around the renderer.
2021-05-05 05:41:12 +00:00
_setup_reflections(*p_render_data->reflection_probes, p_render_data->cam_transform.affine_inverse(), p_render_data->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;
Use specialization constants in clustered renderer * Keep track of when projector, softshadow or directional sofshadow were enabled. * Enable them via specializaton constant where it makes sense. * Re-implements soft shadows. * Re-implements light projectors.
2021-07-12 23:32:05 +00:00
_setup_lights(*p_render_data->lights, p_render_data->cam_transform, p_render_data->shadow_atlas, using_shadows, directional_light_count, positional_light_count, p_render_data->directional_light_soft_shadows);
Split RenderDataRD struct from RenderState struct to simplify passing our render data around the renderer.
2021-05-05 05:41:12 +00:00
_setup_decals(*p_render_data->decals, p_render_data->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
Split RenderDataRD struct from RenderState struct to simplify passing our render data around the renderer.
2021-05-05 05:41:12 +00:00
p_render_data->directional_light_count = directional_light_count;
Refactor light clustering Move from high end to shared, so all renderers use clustering.
2020-07-22 23:39:09 +00:00
Rewrote how barriers work for faster rendering -Added more finegrained control in RenderingDevice API -Optimized barriers (use less ones for thee same) -General optimizations -Shadows render all together unbarriered -GI can render together with shadows. -SDFGI can render together with depth-preoass. -General fixes -Added GPU detection
2021-02-02 19:51:36 +00:00
if (current_cluster_builder) {
current_cluster_builder->bake_cluster();
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
Split RenderDataRD struct from RenderState struct to simplify passing our render data around the renderer.
2021-05-05 05:41:12 +00:00
if (p_render_data->render_buffers.is_valid()) {
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
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;
}
}
Move clustered renderer functionality
2021-03-23 10:46:31 +00:00
if (is_volumetric_supported()) {
Rename GI Classes * GIProbe is now VoxelGI * BakedLightmap is now LightmapGI As godot adds more ways to provide GI (as an example, SDFGI in 4.0), the different techniques (which have different pros/cons) need to be properly named to avoid confusion.
2021-06-04 22:47:26 +00:00
_update_volumetric_fog(p_render_data->render_buffers, p_render_data->environment, p_render_data->cam_projection, p_render_data->cam_transform, p_render_data->shadow_atlas, directional_light_count, directional_shadows, positional_light_count, render_state.voxel_gi_count);
Move clustered renderer functionality
2021-03-23 10:46:31 +00:00
}
Rewrote how barriers work for faster rendering -Added more finegrained control in RenderingDevice API -Optimized barriers (use less ones for thee same) -General optimizations -Shadows render all together unbarriered -GI can render together with shadows. -SDFGI can render together with depth-preoass. -General fixes -Added GPU detection
2021-02-02 19:51:36 +00:00
}
}
Fix Render Info * Fixed and redone the process to obtain render information from a viewport * Some stats, such as material changes are too difficult to guess on Vulkan, were removed. * Separated visible and shadow stats, which causes confusion. * Texture, buffer and general video memory can be queried now. * Fixed the performance metrics too.
2021-07-02 23:14:19 +00:00
void RendererSceneRenderRD::render_scene(RID p_render_buffers, const CameraData *p_camera_data, const PagedArray<GeometryInstance *> &p_instances, const PagedArray<RID> &p_lights, const PagedArray<RID> &p_reflection_probes, const PagedArray<RID> &p_voxel_gi_instances, const PagedArray<RID> &p_decals, const PagedArray<RID> &p_lightmaps, RID p_environment, RID p_camera_effects, RID p_shadow_atlas, RID p_occluder_debug_tex, RID p_reflection_atlas, RID p_reflection_probe, int p_reflection_probe_pass, float p_screen_lod_threshold, const RenderShadowData *p_render_shadows, int p_render_shadow_count, const RenderSDFGIData *p_render_sdfgi_regions, int p_render_sdfgi_region_count, const RenderSDFGIUpdateData *p_sdfgi_update_data, RendererScene::RenderInfo *r_render_info) {
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
// getting this here now so we can direct call a bunch of things more easily
RenderBuffers *rb = nullptr;
if (p_render_buffers.is_valid()) {
rb = render_buffers_owner.getornull(p_render_buffers);
Create mobile renderer
2021-03-10 11:23:55 +00:00
ERR_FAIL_COND(!rb);
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
}
Rewrote how barriers work for faster rendering -Added more finegrained control in RenderingDevice API -Optimized barriers (use less ones for thee same) -General optimizations -Shadows render all together unbarriered -GI can render together with shadows. -SDFGI can render together with depth-preoass. -General fixes -Added GPU detection
2021-02-02 19:51:36 +00:00
//assign render data
Split RenderDataRD struct from RenderState struct to simplify passing our render data around the renderer.
2021-05-05 05:41:12 +00:00
RenderDataRD render_data;
Rewrote how barriers work for faster rendering -Added more finegrained control in RenderingDevice API -Optimized barriers (use less ones for thee same) -General optimizations -Shadows render all together unbarriered -GI can render together with shadows. -SDFGI can render together with depth-preoass. -General fixes -Added GPU detection
2021-02-02 19:51:36 +00:00
{
Split RenderDataRD struct from RenderState struct to simplify passing our render data around the renderer.
2021-05-05 05:41:12 +00:00
render_data.render_buffers = p_render_buffers;
Add stereoscopic rendering through multiview
2021-05-07 13:19:04 +00:00
// Our first camera is used by default
render_data.cam_transform = p_camera_data->main_transform;
render_data.cam_projection = p_camera_data->main_projection;
render_data.view_projection[0] = p_camera_data->main_projection;
render_data.cam_ortogonal = p_camera_data->is_ortogonal;
render_data.view_count = p_camera_data->view_count;
for (uint32_t v = 0; v < p_camera_data->view_count; v++) {
render_data.view_projection[v] = p_camera_data->view_projection[v];
}
render_data.z_near = p_camera_data->main_projection.get_z_near();
render_data.z_far = p_camera_data->main_projection.get_z_far();
Split RenderDataRD struct from RenderState struct to simplify passing our render data around the renderer.
2021-05-05 05:41:12 +00:00
render_data.instances = &p_instances;
render_data.lights = &p_lights;
render_data.reflection_probes = &p_reflection_probes;
Rename GI Classes * GIProbe is now VoxelGI * BakedLightmap is now LightmapGI As godot adds more ways to provide GI (as an example, SDFGI in 4.0), the different techniques (which have different pros/cons) need to be properly named to avoid confusion.
2021-06-04 22:47:26 +00:00
render_data.voxel_gi_instances = &p_voxel_gi_instances;
Split RenderDataRD struct from RenderState struct to simplify passing our render data around the renderer.
2021-05-05 05:41:12 +00:00
render_data.decals = &p_decals;
render_data.lightmaps = &p_lightmaps;
render_data.environment = p_environment;
render_data.camera_effects = p_camera_effects;
render_data.shadow_atlas = p_shadow_atlas;
render_data.reflection_atlas = p_reflection_atlas;
render_data.reflection_probe = p_reflection_probe;
render_data.reflection_probe_pass = p_reflection_probe_pass;
Add stereoscopic rendering through multiview
2021-05-07 13:19:04 +00:00
// this should be the same for all cameras..
render_data.lod_distance_multiplier = p_camera_data->main_projection.get_lod_multiplier();
render_data.lod_camera_plane = Plane(p_camera_data->main_transform.get_origin(), -p_camera_data->main_transform.basis.get_axis(Vector3::AXIS_Z));
Split RenderDataRD struct from RenderState struct to simplify passing our render data around the renderer.
2021-05-05 05:41:12 +00:00
if (get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_DISABLE_LOD) {
render_data.screen_lod_threshold = 0.0;
} else {
render_data.screen_lod_threshold = p_screen_lod_threshold;
}
Rewrote how barriers work for faster rendering -Added more finegrained control in RenderingDevice API -Optimized barriers (use less ones for thee same) -General optimizations -Shadows render all together unbarriered -GI can render together with shadows. -SDFGI can render together with depth-preoass. -General fixes -Added GPU detection
2021-02-02 19:51:36 +00:00
render_state.render_shadows = p_render_shadows;
render_state.render_shadow_count = p_render_shadow_count;
render_state.render_sdfgi_regions = p_render_sdfgi_regions;
render_state.render_sdfgi_region_count = p_render_sdfgi_region_count;
render_state.sdfgi_update_data = p_sdfgi_update_data;
Fix Render Info * Fixed and redone the process to obtain render information from a viewport * Some stats, such as material changes are too difficult to guess on Vulkan, were removed. * Separated visible and shadow stats, which causes confusion. * Texture, buffer and general video memory can be queried now. * Fixed the performance metrics too.
2021-07-02 23:14:19 +00:00
render_data.render_info = r_render_info;
Rewrote how barriers work for faster rendering -Added more finegrained control in RenderingDevice API -Optimized barriers (use less ones for thee same) -General optimizations -Shadows render all together unbarriered -GI can render together with shadows. -SDFGI can render together with depth-preoass. -General fixes -Added GPU detection
2021-02-02 19:51:36 +00:00
}
PagedArray<RID> empty;
if (get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_UNSHADED) {
Split RenderDataRD struct from RenderState struct to simplify passing our render data around the renderer.
2021-05-05 05:41:12 +00:00
render_data.lights = &empty;
render_data.reflection_probes = &empty;
Rename GI Classes * GIProbe is now VoxelGI * BakedLightmap is now LightmapGI As godot adds more ways to provide GI (as an example, SDFGI in 4.0), the different techniques (which have different pros/cons) need to be properly named to avoid confusion.
2021-06-04 22:47:26 +00:00
render_data.voxel_gi_instances = &empty;
Rewrote how barriers work for faster rendering -Added more finegrained control in RenderingDevice API -Optimized barriers (use less ones for thee same) -General optimizations -Shadows render all together unbarriered -GI can render together with shadows. -SDFGI can render together with depth-preoass. -General fixes -Added GPU detection
2021-02-02 19:51:36 +00:00
}
//sdfgi first
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
if (rb != nullptr && rb->sdfgi != nullptr) {
Rewrote how barriers work for faster rendering -Added more finegrained control in RenderingDevice API -Optimized barriers (use less ones for thee same) -General optimizations -Shadows render all together unbarriered -GI can render together with shadows. -SDFGI can render together with depth-preoass. -General fixes -Added GPU detection
2021-02-02 19:51:36 +00:00
for (int i = 0; i < render_state.render_sdfgi_region_count; i++) {
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
rb->sdfgi->render_region(p_render_buffers, render_state.render_sdfgi_regions[i].region, render_state.render_sdfgi_regions[i].instances, this);
Rewrote how barriers work for faster rendering -Added more finegrained control in RenderingDevice API -Optimized barriers (use less ones for thee same) -General optimizations -Shadows render all together unbarriered -GI can render together with shadows. -SDFGI can render together with depth-preoass. -General fixes -Added GPU detection
2021-02-02 19:51:36 +00:00
}
if (render_state.sdfgi_update_data->update_static) {
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
rb->sdfgi->render_static_lights(p_render_buffers, render_state.sdfgi_update_data->static_cascade_count, p_sdfgi_update_data->static_cascade_indices, render_state.sdfgi_update_data->static_positional_lights, this);
Rewrote how barriers work for faster rendering -Added more finegrained control in RenderingDevice API -Optimized barriers (use less ones for thee same) -General optimizations -Shadows render all together unbarriered -GI can render together with shadows. -SDFGI can render together with depth-preoass. -General fixes -Added GPU detection
2021-02-02 19:51:36 +00:00
}
}
Color clear_color;
if (p_render_buffers.is_valid()) {
clear_color = storage->render_target_get_clear_request_color(rb->render_target);
} else {
clear_color = storage->get_default_clear_color();
}
Rename GI Classes * GIProbe is now VoxelGI * BakedLightmap is now LightmapGI As godot adds more ways to provide GI (as an example, SDFGI in 4.0), the different techniques (which have different pros/cons) need to be properly named to avoid confusion.
2021-06-04 22:47:26 +00:00
//assign render indices to voxel_gi_instances
Create mobile renderer
2021-03-10 11:23:55 +00:00
if (is_dynamic_gi_supported()) {
Rename GI Classes * GIProbe is now VoxelGI * BakedLightmap is now LightmapGI As godot adds more ways to provide GI (as an example, SDFGI in 4.0), the different techniques (which have different pros/cons) need to be properly named to avoid confusion.
2021-06-04 22:47:26 +00:00
for (uint32_t i = 0; i < (uint32_t)p_voxel_gi_instances.size(); i++) {
RendererSceneGIRD::VoxelGIInstance *voxel_gi_inst = gi.voxel_gi_instance_owner.getornull(p_voxel_gi_instances[i]);
if (voxel_gi_inst) {
voxel_gi_inst->render_index = i;
Create mobile renderer
2021-03-10 11:23:55 +00:00
}
Rewrote how barriers work for faster rendering -Added more finegrained control in RenderingDevice API -Optimized barriers (use less ones for thee same) -General optimizations -Shadows render all together unbarriered -GI can render together with shadows. -SDFGI can render together with depth-preoass. -General fixes -Added GPU detection
2021-02-02 19:51:36 +00:00
}
}
Split RenderDataRD struct from RenderState struct to simplify passing our render data around the renderer.
2021-05-05 05:41:12 +00:00
if (render_buffers_owner.owns(render_data.render_buffers)) {
// render_data.render_buffers == p_render_buffers so we can use our already retrieved rb
Move clustered renderer functionality
2021-03-23 10:46:31 +00:00
current_cluster_builder = rb->cluster_builder;
Split RenderDataRD struct from RenderState struct to simplify passing our render data around the renderer.
2021-05-05 05:41:12 +00:00
} else if (reflection_probe_instance_owner.owns(render_data.reflection_probe)) {
ReflectionProbeInstance *rpi = reflection_probe_instance_owner.getornull(render_data.reflection_probe);
Rewrote how barriers work for faster rendering -Added more finegrained control in RenderingDevice API -Optimized barriers (use less ones for thee same) -General optimizations -Shadows render all together unbarriered -GI can render together with shadows. -SDFGI can render together with depth-preoass. -General fixes -Added GPU detection
2021-02-02 19:51:36 +00:00
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 {
Move clustered renderer functionality
2021-03-23 10:46:31 +00:00
ERR_PRINT("No render buffer nor reflection atlas, bug"); //should never happen, will crash
Rewrote how barriers work for faster rendering -Added more finegrained control in RenderingDevice API -Optimized barriers (use less ones for thee same) -General optimizations -Shadows render all together unbarriered -GI can render together with shadows. -SDFGI can render together with depth-preoass. -General fixes -Added GPU detection
2021-02-02 19:51:36 +00:00
current_cluster_builder = nullptr;
}
Rename GI Classes * GIProbe is now VoxelGI * BakedLightmap is now LightmapGI As godot adds more ways to provide GI (as an example, SDFGI in 4.0), the different techniques (which have different pros/cons) need to be properly named to avoid confusion.
2021-06-04 22:47:26 +00:00
render_state.voxel_gi_count = 0;
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
Fix VoxelGI (works again). It was broken due to #46046, this makes it work again.
2021-06-12 23:37:56 +00:00
if (rb != nullptr) {
if (rb->sdfgi) {
rb->sdfgi->update_cascades();
Add stereoscopic rendering through multiview
2021-05-07 13:19:04 +00:00
rb->sdfgi->pre_process_gi(render_data.cam_transform, &render_data, this);
Fix VoxelGI (works again). It was broken due to #46046, this makes it work again.
2021-06-12 23:37:56 +00:00
rb->sdfgi->update_light();
}
Fix SDFGI * Broken by 7513b73902f3ed7ac45aa9aefc2f6f333ae731ff, fixes #49631
2021-07-04 00:24:37 +00:00
gi.setup_voxel_gi_instances(render_data.render_buffers, render_data.cam_transform, *render_data.voxel_gi_instances, render_state.voxel_gi_count, this);
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
}
Rewrote how barriers work for faster rendering -Added more finegrained control in RenderingDevice API -Optimized barriers (use less ones for thee same) -General optimizations -Shadows render all together unbarriered -GI can render together with shadows. -SDFGI can render together with depth-preoass. -General fixes -Added GPU detection
2021-02-02 19:51:36 +00:00
render_state.depth_prepass_used = false;
//calls _pre_opaque_render between depth pre-pass and opaque pass
Create mobile renderer
2021-03-10 11:23:55 +00:00
if (current_cluster_builder != nullptr) {
Split RenderDataRD struct from RenderState struct to simplify passing our render data around the renderer.
2021-05-05 05:41:12 +00:00
render_data.cluster_buffer = current_cluster_builder->get_cluster_buffer();
render_data.cluster_size = current_cluster_builder->get_cluster_size();
render_data.cluster_max_elements = current_cluster_builder->get_max_cluster_elements();
Create mobile renderer
2021-03-10 11:23:55 +00:00
}
WIP CameraEffects implementation (bokeh not working for now)
2020-01-13 18:37:24 +00:00
Split RenderDataRD struct from RenderState struct to simplify passing our render data around the renderer.
2021-05-05 05:41:12 +00:00
_render_scene(&render_data, clear_color);
WIP CameraEffects implementation (bokeh not working for now)
2020-01-13 18:37:24 +00:00
if (p_render_buffers.is_valid()) {
Use subpasses to do 3D rendering and resolve in mobile renderer
2021-06-26 10:49:25 +00:00
/*
_debug_draw_cluster(p_render_buffers);
RENDER_TIMESTAMP("Tonemap");
_render_buffers_post_process_and_tonemap(&render_data);
*/
_render_buffers_debug_draw(p_render_buffers, p_shadow_atlas, p_occluder_debug_tex);
if (debug_draw == RS::VIEWPORT_DEBUG_DRAW_SDFGI && rb != nullptr && rb->sdfgi != nullptr) {
rb->sdfgi->debug_draw(render_data.cam_projection, render_data.cam_transform, rb->width, rb->height, rb->render_target, rb->texture);
}
}
}
void RendererSceneRenderRD::_debug_draw_cluster(RID p_render_buffers) {
if (p_render_buffers.is_valid() && current_cluster_builder != nullptr) {
RS::ViewportDebugDraw dd = get_debug_draw_mode();
if (dd == RS::VIEWPORT_DEBUG_DRAW_CLUSTER_OMNI_LIGHTS || dd == RS::VIEWPORT_DEBUG_DRAW_CLUSTER_SPOT_LIGHTS || dd == RS::VIEWPORT_DEBUG_DRAW_CLUSTER_DECALS || dd == RS::VIEWPORT_DEBUG_DRAW_CLUSTER_REFLECTION_PROBES) {
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
ClusterBuilderRD::ElementType elem_type = ClusterBuilderRD::ELEMENT_TYPE_MAX;
Use subpasses to do 3D rendering and resolve in mobile renderer
2021-06-26 10:49:25 +00:00
switch (dd) {
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
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: {
}
}
Use subpasses to do 3D rendering and resolve in mobile renderer
2021-06-26 10:49:25 +00:00
current_cluster_builder->debug(elem_type);
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
}
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
}
Fix Render Info * Fixed and redone the process to obtain render information from a viewport * Some stats, such as material changes are too difficult to guess on Vulkan, were removed. * Separated visible and shadow stats, which causes confusion. * Texture, buffer and general video memory can be queried now. * Fixed the performance metrics too.
2021-07-02 23:14:19 +00:00
void RendererSceneRenderRD::_render_shadow_pass(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, bool p_open_pass, bool p_close_pass, bool p_clear_region, RendererScene::RenderInfo *p_render_info) {
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;
Rename Transform to Transform3D in core
2020-10-17 05:08:21 +00:00
Transform3D light_transform;
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_DIRECTIONAL) {
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;
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;
}
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;
}
}
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
light_instance->shadow_transform[p_pass].atlas_rect = atlas_rect;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
light_instance->shadow_transform[p_pass].atlas_rect.position /= directional_shadow.size;
light_instance->shadow_transform[p_pass].atlas_rect.size /= directional_shadow.size;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
zfar = storage->light_get_param(light_instance->light, RS::LIGHT_PARAM_RANGE);
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
render_fb = directional_shadow.fb;
render_texture = RID();
flip_y = true;
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
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
} else {
//set from shadow atlas
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
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
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));
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
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
_update_shadow_atlas(shadow_atlas);
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
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
uint32_t key = shadow_atlas->shadow_owners[p_light];
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
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
uint32_t quadrant = (key >> ShadowAtlas::QUADRANT_SHIFT) & 0x3;
uint32_t shadow = key & ShadowAtlas::SHADOW_INDEX_MASK;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
ERR_FAIL_INDEX((int)shadow, shadow_atlas->quadrants[quadrant].shadows.size());
Rewrote how barriers work for faster rendering -Added more finegrained control in RenderingDevice API -Optimized barriers (use less ones for thee same) -General optimizations -Shadows render all together unbarriered -GI can render together with shadows. -SDFGI can render together with depth-preoass. -General fixes -Added GPU detection
2021-02-02 19:51:36 +00:00
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
uint32_t quadrant_size = shadow_atlas->size >> 1;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
atlas_rect.position.x = (quadrant & 1) * quadrant_size;
atlas_rect.position.y = (quadrant >> 1) * quadrant_size;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
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;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
atlas_rect.size.width = shadow_size;
atlas_rect.size.height = shadow_size;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
zfar = storage->light_get_param(light_instance->light, RS::LIGHT_PARAM_RANGE);
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +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) {
ShadowCubemap *cubemap = _get_shadow_cubemap(shadow_size / 2);
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
render_fb = cubemap->side_fb[p_pass];
render_texture = cubemap->cubemap;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
light_projection = light_instance->shadow_transform[p_pass].camera;
light_transform = light_instance->shadow_transform[p_pass].transform;
render_cubemap = true;
finalize_cubemap = p_pass == 5;
atlas_fb = shadow_atlas->fb;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
atlas_size = shadow_atlas->size;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
if (p_pass == 0) {
_render_shadow_begin();
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
}
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
} else {
light_projection = light_instance->shadow_transform[0].camera;
light_transform = light_instance->shadow_transform[0].transform;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
atlas_rect.size.height /= 2;
atlas_rect.position.y += p_pass * atlas_rect.size.height;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
using_dual_paraboloid = true;
using_dual_paraboloid_flip = p_pass == 1;
render_fb = shadow_atlas->fb;
flip_y = true;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
}
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
} else if (storage->light_get_type(light_instance->light) == RS::LIGHT_SPOT) {
light_projection = light_instance->shadow_transform[0].camera;
light_transform = light_instance->shadow_transform[0].transform;
render_fb = shadow_atlas->fb;
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
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
flip_y = true;
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
}
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
if (render_cubemap) {
//rendering to cubemap
Fix Render Info * Fixed and redone the process to obtain render information from a viewport * Some stats, such as material changes are too difficult to guess on Vulkan, were removed. * Separated visible and shadow stats, which causes confusion. * Texture, buffer and general video memory can be queried now. * Fixed the performance metrics too.
2021-07-02 23:14:19 +00:00
_render_shadow_append(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, Rect2(), false, true, true, true, p_render_info);
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
if (finalize_cubemap) {
_render_shadow_process();
_render_shadow_end();
//reblit
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);
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
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
//restore transform so it can be properly used
light_instance_set_shadow_transform(p_light, CameraMatrix(), light_instance->transform, zfar, 0, 0, 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
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
} else {
//render shadow
Fix Render Info * Fixed and redone the process to obtain render information from a viewport * Some stats, such as material changes are too difficult to guess on Vulkan, were removed. * Separated visible and shadow stats, which causes confusion. * Texture, buffer and general video memory can be queried now. * Fixed the performance metrics too.
2021-07-02 23:14:19 +00:00
_render_shadow_append(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, p_clear_region, p_open_pass, p_close_pass, p_render_info);
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +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
Rename Transform to Transform3D in core
2020-10-17 05:08:21 +00:00
void RendererSceneRenderRD::render_material(const Transform3D &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, const PagedArray<GeometryInstance *> &p_instances, RID p_framebuffer, const Rect2i &p_region) {
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
_render_material(p_cam_transform, p_cam_projection, p_cam_ortogonal, p_instances, p_framebuffer, p_region);
}
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
Rename Transform to Transform3D in core
2020-10-17 05:08:21 +00:00
void RendererSceneRenderRD::render_particle_collider_heightfield(RID p_collider, const Transform3D &p_transform, const PagedArray<GeometryInstance *> &p_instances) {
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +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);
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
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
Vector3 cam_pos = p_transform.origin;
cam_pos.y += extents.y;
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
Rename Transform to Transform3D in core
2020-10-17 05:08:21 +00:00
Transform3D cam_xform;
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
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());
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
RID fb = storage->particles_collision_get_heightfield_framebuffer(p_collider);
Rewrote how barriers work for faster rendering -Added more finegrained control in RenderingDevice API -Optimized barriers (use less ones for thee same) -General optimizations -Shadows render all together unbarriered -GI can render together with shadows. -SDFGI can render together with depth-preoass. -General fixes -Added GPU detection
2021-02-02 19:51:36 +00:00
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
_render_particle_collider_heightfield(fb, cam_xform, cm, p_instances);
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
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) {
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
rb->sdfgi->erase();
memdelete(rb->sdfgi);
rb->sdfgi = nullptr;
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
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)) {
Fixes to forward mobile * use valid format for framebuffer: VK_FORMAT_A2B10G10R10_UNORM_PACK32 * Unfortunately cant be used for compute. * Mobile will need to do refprobe, sky, mipmapblurring using raster.
2021-07-15 20:06:33 +00:00
ReflectionProbeInstance *rpi = reflection_probe_instance_owner.getornull(p_rid);
_free_forward_id(FORWARD_ID_TYPE_REFLECTION_PROBE, rpi->forward_id);
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)) {
Fixes to forward mobile * use valid format for framebuffer: VK_FORMAT_A2B10G10R10_UNORM_PACK32 * Unfortunately cant be used for compute. * Mobile will need to do refprobe, sky, mipmapblurring using raster.
2021-07-15 20:06:33 +00:00
DecalInstance *di = decal_instance_owner.getornull(p_rid);
_free_forward_id(FORWARD_ID_TYPE_DECAL, di->forward_id);
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
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);
Rename GI Classes * GIProbe is now VoxelGI * BakedLightmap is now LightmapGI As godot adds more ways to provide GI (as an example, SDFGI in 4.0), the different techniques (which have different pros/cons) need to be properly named to avoid confusion.
2021-06-04 22:47:26 +00:00
} else if (gi.voxel_gi_instance_owner.owns(p_rid)) {
RendererSceneGIRD::VoxelGIInstance *voxel_gi = gi.voxel_gi_instance_owner.getornull(p_rid);
if (voxel_gi->texture.is_valid()) {
RD::get_singleton()->free(voxel_gi->texture);
RD::get_singleton()->free(voxel_gi->write_buffer);
GIProbes working.
2019-10-03 20:39:08 +00:00
}
Rename GI Classes * GIProbe is now VoxelGI * BakedLightmap is now LightmapGI As godot adds more ways to provide GI (as an example, SDFGI in 4.0), the different techniques (which have different pros/cons) need to be properly named to avoid confusion.
2021-06-04 22:47:26 +00:00
for (int i = 0; i < voxel_gi->dynamic_maps.size(); i++) {
RD::get_singleton()->free(voxel_gi->dynamic_maps[i].texture);
RD::get_singleton()->free(voxel_gi->dynamic_maps[i].depth);
Dynamic object support for GI Probes (a bit buggy still)
2019-10-11 02:14:56 +00:00
}
Rename GI Classes * GIProbe is now VoxelGI * BakedLightmap is now LightmapGI As godot adds more ways to provide GI (as an example, SDFGI in 4.0), the different techniques (which have different pros/cons) need to be properly named to avoid confusion.
2021-06-04 22:47:26 +00:00
gi.voxel_gi_instance_owner.free(p_rid);
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
} else if (sky.sky_owner.owns(p_rid)) {
sky.update_dirty_skys();
sky.free_sky(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);
}
Fixes to forward mobile * use valid format for framebuffer: VK_FORMAT_A2B10G10R10_UNORM_PACK32 * Unfortunately cant be used for compute. * Mobile will need to do refprobe, sky, mipmapblurring using raster.
2021-07-15 20:06:33 +00:00
if (light_instance->light_type != RS::LIGHT_DIRECTIONAL) {
_free_forward_id(light_instance->light_type == RS::LIGHT_OMNI ? FORWARD_ID_TYPE_OMNI_LIGHT : FORWARD_ID_TYPE_SPOT_LIGHT, light_instance->forward_id);
}
Rewrote large part of rendering, omni and spot shadows now work.
2019-09-07 01:51:27 +00:00
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() {
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
sky.update_dirty_skys();
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::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;
Rename `instance()`->`instantiate()` when it's a verb
2021-06-17 22:03:09 +00:00
img.instantiate();
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
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;
Rename `instance()`->`instantiate()` when it's a verb
2021-06-17 22:03:09 +00:00
img.instantiate();
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
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;
Rename `instance()`->`instantiate()` when it's a verb
2021-06-17 22:03:09 +00:00
img.instantiate();
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
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;
Rename `instance()`->`instantiate()` when it's a verb
2021-06-17 22:03:09 +00:00
img.instantiate();
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
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) {
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
gi.sdfgi_debug_probe_pos = p_position;
gi.sdfgi_debug_probe_dir = p_dir;
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
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;
}
Move clustered renderer functionality
2021-03-23 10:46:31 +00:00
bool RendererSceneRenderRD::is_dynamic_gi_supported() const {
// usable by default (unless low end = true)
return true;
}
bool RendererSceneRenderRD::is_clustered_enabled() const {
// used by default.
return true;
}
bool RendererSceneRenderRD::is_volumetric_supported() const {
// usable by default (unless low end = true)
return true;
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
}
Create mobile renderer
2021-03-10 11:23:55 +00:00
uint32_t RendererSceneRenderRD::get_max_elements() const {
return GLOBAL_GET("rendering/limits/cluster_builder/max_clustered_elements");
}
RenderingServer reorganization
2020-12-04 18:26:24 +00:00
RendererSceneRenderRD::RendererSceneRenderRD(RendererStorageRD *p_storage) {
Create mobile renderer
2021-03-10 11:23:55 +00:00
max_cluster_elements = get_max_elements();
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
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
Reorganize Project Settings -Advanced Settings toggle also hides advanced properties when disabled -Simplified Advanced Bar (errors were just plain redundant) -Reorganized rendering quality settings. -Reorganized miscelaneous settings for clean up.
2021-02-17 16:44:49 +00:00
directional_shadow.size = GLOBAL_GET("rendering/shadows/directional_shadow/size");
directional_shadow.use_16_bits = GLOBAL_GET("rendering/shadows/directional_shadow/16_bits");
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
Working sky shader implementation
2020-03-20 00:32:19 +00:00
/* SKY SHADER */
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
sky.init(storage);
Working sky shader implementation
2020-03-20 00:32:19 +00:00
Combine init_gi and init_sdfgi into a single init
2021-03-11 03:35:05 +00:00
/* GI */
Remove low_end option from renderer, being replaced by separate implementation
2021-04-07 11:06:48 +00:00
if (is_dynamic_gi_supported()) {
Combine init_gi and init_sdfgi into a single init
2021-03-11 03:35:05 +00:00
gi.init(storage, &sky);
Addition of SDFGI for open world global illumination Move GI to a deferred pass
2020-06-25 13:33:28 +00:00
}
Don't use trashy max_decals value to prevent crashes
2021-02-01 20:43:08 +00:00
{ //decals
cluster.max_decals = max_cluster_elements;
uint32_t decal_buffer_size = cluster.max_decals * sizeof(Cluster::DecalData);
cluster.decals = memnew_arr(Cluster::DecalData, cluster.max_decals);
cluster.decal_sort = memnew_arr(Cluster::InstanceSort<DecalInstance>, cluster.max_decals);
cluster.decal_buffer = RD::get_singleton()->storage_buffer_create(decal_buffer_size);
}
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);
Rewrote how barriers work for faster rendering -Added more finegrained control in RenderingDevice API -Optimized barriers (use less ones for thee same) -General optimizations -Shadows render all together unbarriered -GI can render together with shadows. -SDFGI can render together with depth-preoass. -General fixes -Added GPU detection
2021-02-02 19:51:36 +00:00
cluster.reflection_sort = memnew_arr(Cluster::InstanceSort<ReflectionProbeInstance>, 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_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
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);
}
Remove low_end option from renderer, being replaced by separate implementation
2021-04-07 11:06:48 +00:00
if (is_volumetric_supported()) {
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);
}
Reorganize Project Settings -Advanced Settings toggle also hides advanced properties when disabled -Simplified Advanced Bar (errors were just plain redundant) -Reorganized rendering quality settings. -Reorganized miscelaneous settings for clean up.
2021-02-17 16:44:49 +00:00
camera_effects_set_dof_blur_bokeh_shape(RS::DOFBokehShape(int(GLOBAL_GET("rendering/camera/depth_of_field/depth_of_field_bokeh_shape"))));
camera_effects_set_dof_blur_quality(RS::DOFBlurQuality(int(GLOBAL_GET("rendering/camera/depth_of_field/depth_of_field_bokeh_quality"))), GLOBAL_GET("rendering/camera/depth_of_field/depth_of_field_use_jitter"));
environment_set_ssao_quality(RS::EnvironmentSSAOQuality(int(GLOBAL_GET("rendering/environment/ssao/quality"))), GLOBAL_GET("rendering/environment/ssao/half_size"), GLOBAL_GET("rendering/environment/ssao/adaptive_target"), GLOBAL_GET("rendering/environment/ssao/blur_passes"), GLOBAL_GET("rendering/environment/ssao/fadeout_from"), GLOBAL_GET("rendering/environment/ssao/fadeout_to"));
screen_space_roughness_limiter = GLOBAL_GET("rendering/anti_aliasing/screen_space_roughness_limiter/enabled");
screen_space_roughness_limiter_amount = GLOBAL_GET("rendering/anti_aliasing/screen_space_roughness_limiter/amount");
screen_space_roughness_limiter_limit = GLOBAL_GET("rendering/anti_aliasing/screen_space_roughness_limiter/limit");
glow_bicubic_upscale = int(GLOBAL_GET("rendering/environment/glow/upscale_mode")) > 0;
glow_high_quality = GLOBAL_GET("rendering/environment/glow/use_high_quality");
ssr_roughness_quality = RS::EnvironmentSSRRoughnessQuality(int(GLOBAL_GET("rendering/environment/screen_space_reflection/roughness_quality")));
sss_quality = RS::SubSurfaceScatteringQuality(int(GLOBAL_GET("rendering/environment/subsurface_scattering/subsurface_scattering_quality")));
sss_scale = GLOBAL_GET("rendering/environment/subsurface_scattering/subsurface_scattering_scale");
sss_depth_scale = GLOBAL_GET("rendering/environment/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);
Reorganize Project Settings -Advanced Settings toggle also hides advanced properties when disabled -Simplified Advanced Bar (errors were just plain redundant) -Reorganized rendering quality settings. -Reorganized miscelaneous settings for clean up.
2021-02-17 16:44:49 +00:00
shadows_quality_set(RS::ShadowQuality(int(GLOBAL_GET("rendering/shadows/shadows/soft_shadow_quality"))));
directional_shadow_quality_set(RS::ShadowQuality(int(GLOBAL_GET("rendering/shadows/directional_shadow/soft_shadow_quality"))));
Added volumetric fog effect.
2020-08-13 01:21:01 +00:00
Reorganize Project Settings -Advanced Settings toggle also hides advanced properties when disabled -Simplified Advanced Bar (errors were just plain redundant) -Reorganized rendering quality settings. -Reorganized miscelaneous settings for clean up.
2021-02-17 16:44:49 +00:00
environment_set_volumetric_fog_volume_size(GLOBAL_GET("rendering/environment/volumetric_fog/volume_size"), GLOBAL_GET("rendering/environment/volumetric_fog/volume_depth"));
environment_set_volumetric_fog_filter_active(GLOBAL_GET("rendering/environment/volumetric_fog/use_filter"));
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
Implement more rendering options as specialization constants * Shadow quality settings now specialization constant. * Decal and light projector filters can be set. * Changing those settings forces re-creation of the pipelines. These changes should help improve performance related to shadow mapping, and allows improving performance by sacrificing decal and light projector quality.
2021-07-19 19:41:55 +00:00
decals_set_filter(RS::DecalFilter(int(GLOBAL_GET("rendering/textures/decals/filter"))));
light_projectors_set_filter(RS::LightProjectorFilter(int(GLOBAL_GET("rendering/textures/light_projectors/filter"))));
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);
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
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
if (sky.sky_scene_state.uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(sky.sky_scene_state.uniform_set)) {
RD::get_singleton()->free(sky.sky_scene_state.uniform_set);
Working sky shader implementation
2020-03-20 00:32:19 +00:00
}
Remove low_end option from renderer, being replaced by separate implementation
2021-04-07 11:06:48 +00:00
if (is_dynamic_gi_supported()) {
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
gi.free();
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
}
Add fog to sky shaders
2020-08-18 04:12:51 +00:00
Moving GI code into RendererServerGIRD Moving Skyshader code into RendererServerSkyRD
2021-02-13 12:08:08 +00:00
RendererSceneSkyRD::SkyMaterialData *md = (RendererSceneSkyRD::SkyMaterialData *)storage->material_get_data(sky.sky_shader.default_material, RendererStorageRD::SHADER_TYPE_SKY);
sky.sky_shader.shader.version_free(md->shader_data->version);
RD::get_singleton()->free(sky.sky_scene_state.directional_light_buffer);
RD::get_singleton()->free(sky.sky_scene_state.uniform_buffer);
memdelete_arr(sky.sky_scene_state.directional_lights);
memdelete_arr(sky.sky_scene_state.last_frame_directional_lights);
storage->free(sky.sky_shader.default_shader);
storage->free(sky.sky_shader.default_material);
storage->free(sky.sky_scene_state.fog_shader);
storage->free(sky.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);
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
}