summaryrefslogtreecommitdiff
path: root/servers
diff options
context:
space:
mode:
Diffstat (limited to 'servers')
-rw-r--r--servers/physics_2d/area_2d_sw.cpp4
-rw-r--r--servers/physics_2d/body_2d_sw.h2
-rw-r--r--servers/physics_2d/broad_phase_2d_hash_grid.cpp2
-rw-r--r--servers/physics_2d/space_2d_sw.h2
-rw-r--r--servers/physics_3d/area_3d_sw.cpp4
-rw-r--r--servers/physics_3d/body_3d_sw.h2
-rw-r--r--servers/physics_3d/space_3d_sw.h2
-rw-r--r--servers/rendering/renderer_canvas_cull.cpp18
-rw-r--r--servers/rendering/renderer_rd/effects_rd.cpp577
-rw-r--r--servers/rendering/renderer_rd/effects_rd.h159
-rw-r--r--servers/rendering/renderer_rd/renderer_canvas_render_rd.cpp8
-rw-r--r--servers/rendering/renderer_rd/renderer_scene_render_forward.cpp79
-rw-r--r--servers/rendering/renderer_rd/renderer_scene_render_forward.h18
-rw-r--r--servers/rendering/renderer_rd/renderer_scene_render_rd.cpp304
-rw-r--r--servers/rendering/renderer_rd/renderer_scene_render_rd.h72
-rw-r--r--servers/rendering/renderer_rd/renderer_storage_rd.cpp20
-rw-r--r--servers/rendering/renderer_rd/renderer_storage_rd.h2
-rw-r--r--servers/rendering/renderer_rd/shader_compiler_rd.cpp19
-rw-r--r--servers/rendering/renderer_rd/shaders/SCsub4
-rw-r--r--servers/rendering/renderer_rd/shaders/ssao.glsl601
-rw-r--r--servers/rendering/renderer_rd/shaders/ssao_blur.glsl245
-rw-r--r--servers/rendering/renderer_rd/shaders/ssao_downsample.glsl206
-rw-r--r--servers/rendering/renderer_rd/shaders/ssao_importance_map.glsl126
-rw-r--r--servers/rendering/renderer_rd/shaders/ssao_interleave.glsl119
-rw-r--r--servers/rendering/renderer_rd/shaders/ssao_minify.glsl45
-rw-r--r--servers/rendering/renderer_scene.h4
-rw-r--r--servers/rendering/renderer_scene_cull.cpp1843
-rw-r--r--servers/rendering/renderer_scene_cull.h421
-rw-r--r--servers/rendering/renderer_scene_render.h29
-rw-r--r--servers/rendering/rendering_device.cpp4
-rw-r--r--servers/rendering/rendering_device.h1
-rw-r--r--servers/rendering/rendering_device_binds.cpp2
-rw-r--r--servers/rendering/rendering_server_default.cpp25
-rw-r--r--servers/rendering/rendering_server_default.h8
-rw-r--r--servers/rendering/rendering_server_wrap_mt.h8
-rw-r--r--servers/rendering/shader_language.cpp445
-rw-r--r--servers/rendering/shader_language.h8
-rw-r--r--servers/rendering_server.cpp31
-rw-r--r--servers/rendering_server.h15
-rw-r--r--servers/text_server.cpp53
-rw-r--r--servers/xr/xr_positional_tracker.cpp36
-rw-r--r--servers/xr/xr_positional_tracker.h16
42 files changed, 3709 insertions, 1880 deletions
diff --git a/servers/physics_2d/area_2d_sw.cpp b/servers/physics_2d/area_2d_sw.cpp
index 7485f31afc..31704c9f66 100644
--- a/servers/physics_2d/area_2d_sw.cpp
+++ b/servers/physics_2d/area_2d_sw.cpp
@@ -199,7 +199,7 @@ void Area2DSW::set_monitorable(bool p_monitorable) {
}
void Area2DSW::call_queries() {
- if (monitor_callback_id.is_valid() && !monitored_bodies.empty()) {
+ if (monitor_callback_id.is_valid() && !monitored_bodies.is_empty()) {
Variant res[5];
Variant *resptr[5];
for (int i = 0; i < 5; i++) {
@@ -234,7 +234,7 @@ void Area2DSW::call_queries() {
}
}
- if (area_monitor_callback_id.is_valid() && !monitored_areas.empty()) {
+ if (area_monitor_callback_id.is_valid() && !monitored_areas.is_empty()) {
Variant res[5];
Variant *resptr[5];
for (int i = 0; i < 5; i++) {
diff --git a/servers/physics_2d/body_2d_sw.h b/servers/physics_2d/body_2d_sw.h
index 19e4b92a99..1b940a5b54 100644
--- a/servers/physics_2d/body_2d_sw.h
+++ b/servers/physics_2d/body_2d_sw.h
@@ -164,7 +164,7 @@ public:
_FORCE_INLINE_ int get_max_contacts_reported() const { return contacts.size(); }
- _FORCE_INLINE_ bool can_report_contacts() const { return !contacts.empty(); }
+ _FORCE_INLINE_ bool can_report_contacts() const { return !contacts.is_empty(); }
_FORCE_INLINE_ void add_contact(const Vector2 &p_local_pos, const Vector2 &p_local_normal, real_t p_depth, int p_local_shape, const Vector2 &p_collider_pos, int p_collider_shape, ObjectID p_collider_instance_id, const RID &p_collider, const Vector2 &p_collider_velocity_at_pos);
_FORCE_INLINE_ void add_exception(const RID &p_exception) { exceptions.insert(p_exception); }
diff --git a/servers/physics_2d/broad_phase_2d_hash_grid.cpp b/servers/physics_2d/broad_phase_2d_hash_grid.cpp
index c8b3d193c9..4fca48e771 100644
--- a/servers/physics_2d/broad_phase_2d_hash_grid.cpp
+++ b/servers/physics_2d/broad_phase_2d_hash_grid.cpp
@@ -261,7 +261,7 @@ void BroadPhase2DHashGrid::_exit_grid(Element *p_elem, const Rect2 &p_rect, bool
}
}
- if (pb->object_set.empty() && pb->static_object_set.empty()) {
+ if (pb->object_set.is_empty() && pb->static_object_set.is_empty()) {
if (hash_table[idx] == pb) {
hash_table[idx] = pb->next;
} else {
diff --git a/servers/physics_2d/space_2d_sw.h b/servers/physics_2d/space_2d_sw.h
index 93b62e0ba4..173157d307 100644
--- a/servers/physics_2d/space_2d_sw.h
+++ b/servers/physics_2d/space_2d_sw.h
@@ -187,7 +187,7 @@ public:
int test_body_ray_separation(Body2DSW *p_body, const Transform2D &p_transform, bool p_infinite_inertia, Vector2 &r_recover_motion, PhysicsServer2D::SeparationResult *r_results, int p_result_max, real_t p_margin);
void set_debug_contacts(int p_amount) { contact_debug.resize(p_amount); }
- _FORCE_INLINE_ bool is_debugging_contacts() const { return !contact_debug.empty(); }
+ _FORCE_INLINE_ bool is_debugging_contacts() const { return !contact_debug.is_empty(); }
_FORCE_INLINE_ void add_debug_contact(const Vector2 &p_contact) {
if (contact_debug_count < contact_debug.size()) {
contact_debug.write[contact_debug_count++] = p_contact;
diff --git a/servers/physics_3d/area_3d_sw.cpp b/servers/physics_3d/area_3d_sw.cpp
index 571f1435de..e67112c634 100644
--- a/servers/physics_3d/area_3d_sw.cpp
+++ b/servers/physics_3d/area_3d_sw.cpp
@@ -199,7 +199,7 @@ void Area3DSW::set_monitorable(bool p_monitorable) {
}
void Area3DSW::call_queries() {
- if (monitor_callback_id.is_valid() && !monitored_bodies.empty()) {
+ if (monitor_callback_id.is_valid() && !monitored_bodies.is_empty()) {
Variant res[5];
Variant *resptr[5];
for (int i = 0; i < 5; i++) {
@@ -234,7 +234,7 @@ void Area3DSW::call_queries() {
}
}
- if (area_monitor_callback_id.is_valid() && !monitored_areas.empty()) {
+ if (area_monitor_callback_id.is_valid() && !monitored_areas.is_empty()) {
Variant res[5];
Variant *resptr[5];
for (int i = 0; i < 5; i++) {
diff --git a/servers/physics_3d/body_3d_sw.h b/servers/physics_3d/body_3d_sw.h
index 6dbda8670a..6c6ffc996a 100644
--- a/servers/physics_3d/body_3d_sw.h
+++ b/servers/physics_3d/body_3d_sw.h
@@ -178,7 +178,7 @@ public:
}
_FORCE_INLINE_ int get_max_contacts_reported() const { return contacts.size(); }
- _FORCE_INLINE_ bool can_report_contacts() const { return !contacts.empty(); }
+ _FORCE_INLINE_ bool can_report_contacts() const { return !contacts.is_empty(); }
_FORCE_INLINE_ void add_contact(const Vector3 &p_local_pos, const Vector3 &p_local_normal, real_t p_depth, int p_local_shape, const Vector3 &p_collider_pos, int p_collider_shape, ObjectID p_collider_instance_id, const RID &p_collider, const Vector3 &p_collider_velocity_at_pos);
_FORCE_INLINE_ void add_exception(const RID &p_exception) { exceptions.insert(p_exception); }
diff --git a/servers/physics_3d/space_3d_sw.h b/servers/physics_3d/space_3d_sw.h
index 22535a6adb..668afa832c 100644
--- a/servers/physics_3d/space_3d_sw.h
+++ b/servers/physics_3d/space_3d_sw.h
@@ -182,7 +182,7 @@ public:
PhysicsDirectSpaceState3DSW *get_direct_state();
void set_debug_contacts(int p_amount) { contact_debug.resize(p_amount); }
- _FORCE_INLINE_ bool is_debugging_contacts() const { return !contact_debug.empty(); }
+ _FORCE_INLINE_ bool is_debugging_contacts() const { return !contact_debug.is_empty(); }
_FORCE_INLINE_ void add_debug_contact(const Vector3 &p_contact) {
if (contact_debug_count < contact_debug.size()) {
contact_debug.write[contact_debug_count++] = p_contact;
diff --git a/servers/rendering/renderer_canvas_cull.cpp b/servers/rendering/renderer_canvas_cull.cpp
index 53deba512e..72ae719c93 100644
--- a/servers/rendering/renderer_canvas_cull.cpp
+++ b/servers/rendering/renderer_canvas_cull.cpp
@@ -825,10 +825,10 @@ void RendererCanvasCull::canvas_item_add_nine_patch(RID p_item, const Rect2 &p_r
style->source = p_source;
style->draw_center = p_draw_center;
style->color = p_modulate;
- style->margin[MARGIN_LEFT] = p_topleft.x;
- style->margin[MARGIN_TOP] = p_topleft.y;
- style->margin[MARGIN_RIGHT] = p_bottomright.x;
- style->margin[MARGIN_BOTTOM] = p_bottomright.y;
+ style->margin[SIDE_LEFT] = p_topleft.x;
+ style->margin[SIDE_TOP] = p_topleft.y;
+ style->margin[SIDE_RIGHT] = p_bottomright.x;
+ style->margin[SIDE_BOTTOM] = p_bottomright.y;
style->axis_x = p_x_axis_mode;
style->axis_y = p_y_axis_mode;
}
@@ -874,7 +874,7 @@ void RendererCanvasCull::canvas_item_add_polygon(RID p_item, const Vector<Point2
ERR_FAIL_COND(uv_size != 0 && (uv_size != pointcount));
#endif
Vector<int> indices = Geometry2D::triangulate_polygon(p_points);
- ERR_FAIL_COND_MSG(indices.empty(), "Invalid polygon data, triangulation failed.");
+ ERR_FAIL_COND_MSG(indices.is_empty(), "Invalid polygon data, triangulation failed.");
Item::CommandPolygon *polygon = canvas_item->alloc_command<Item::CommandPolygon>();
ERR_FAIL_COND(!polygon);
@@ -889,10 +889,10 @@ void RendererCanvasCull::canvas_item_add_triangle_array(RID p_item, const Vector
int vertex_count = p_points.size();
ERR_FAIL_COND(vertex_count == 0);
- ERR_FAIL_COND(!p_colors.empty() && p_colors.size() != vertex_count && p_colors.size() != 1);
- ERR_FAIL_COND(!p_uvs.empty() && p_uvs.size() != vertex_count);
- ERR_FAIL_COND(!p_bones.empty() && p_bones.size() != vertex_count * 4);
- ERR_FAIL_COND(!p_weights.empty() && p_weights.size() != vertex_count * 4);
+ ERR_FAIL_COND(!p_colors.is_empty() && p_colors.size() != vertex_count && p_colors.size() != 1);
+ ERR_FAIL_COND(!p_uvs.is_empty() && p_uvs.size() != vertex_count);
+ ERR_FAIL_COND(!p_bones.is_empty() && p_bones.size() != vertex_count * 4);
+ ERR_FAIL_COND(!p_weights.is_empty() && p_weights.size() != vertex_count * 4);
Vector<int> indices = p_indices;
diff --git a/servers/rendering/renderer_rd/effects_rd.cpp b/servers/rendering/renderer_rd/effects_rd.cpp
index cbb000cb8c..b33255b54b 100644
--- a/servers/rendering/renderer_rd/effects_rd.cpp
+++ b/servers/rendering/renderer_rd/effects_rd.cpp
@@ -31,6 +31,7 @@
#include "effects_rd.h"
#include "core/config/project_settings.h"
+#include "core/math/math_defs.h"
#include "core/os/os.h"
#include "thirdparty/misc/cubemap_coeffs.h"
@@ -125,6 +126,33 @@ RID EffectsRD::_get_compute_uniform_set_from_texture(RID p_texture, bool p_use_m
return uniform_set;
}
+RID EffectsRD::_get_compute_uniform_set_from_texture_and_sampler(RID p_texture, RID p_sampler) {
+ TextureSamplerPair tsp;
+ tsp.texture = p_texture;
+ tsp.sampler = p_sampler;
+
+ if (texture_sampler_to_compute_uniform_set_cache.has(tsp)) {
+ RID uniform_set = texture_sampler_to_compute_uniform_set_cache[tsp];
+ if (RD::get_singleton()->uniform_set_is_valid(uniform_set)) {
+ return uniform_set;
+ }
+ }
+
+ Vector<RD::Uniform> uniforms;
+ RD::Uniform u;
+ u.uniform_type = RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE;
+ u.binding = 0;
+ u.ids.push_back(p_sampler);
+ u.ids.push_back(p_texture);
+ uniforms.push_back(u);
+ //any thing with the same configuration (one texture in binding 0 for set 0), is good
+ RID uniform_set = RD::get_singleton()->uniform_set_create(uniforms, ssao.blur_shader.version_get_shader(ssao.blur_shader_version, 0), 0);
+
+ texture_sampler_to_compute_uniform_set_cache[tsp] = uniform_set;
+
+ return uniform_set;
+}
+
RID EffectsRD::_get_compute_uniform_set_from_texture_pair(RID p_texture1, RID p_texture2, bool p_use_mipmaps) {
TexturePair tp;
tp.texture1 = p_texture1;
@@ -951,157 +979,345 @@ void EffectsRD::bokeh_dof(RID p_base_texture, RID p_depth_texture, const Size2i
RD::get_singleton()->compute_list_end();
}
-void EffectsRD::generate_ssao(RID p_depth_buffer, RID p_normal_buffer, const Size2i &p_depth_buffer_size, RID p_depth_mipmaps_texture, const Vector<RID> &depth_mipmaps, RID p_ao1, bool p_half_size, RID p_ao2, RID p_upscale_buffer, float p_intensity, float p_radius, float p_bias, const CameraMatrix &p_projection, RS::EnvironmentSSAOQuality p_quality, RS::EnvironmentSSAOBlur p_blur, float p_edge_sharpness) {
- //minify first
- ssao.minify_push_constant.orthogonal = p_projection.is_orthogonal();
- ssao.minify_push_constant.z_near = p_projection.get_z_near();
- ssao.minify_push_constant.z_far = p_projection.get_z_far();
- ssao.minify_push_constant.pixel_size[0] = 1.0 / p_depth_buffer_size.x;
- ssao.minify_push_constant.pixel_size[1] = 1.0 / p_depth_buffer_size.y;
- ssao.minify_push_constant.source_size[0] = p_depth_buffer_size.x;
- ssao.minify_push_constant.source_size[1] = p_depth_buffer_size.y;
+void EffectsRD::gather_ssao(RD::ComputeListID p_compute_list, const Vector<RID> p_ao_slices, const SSAOSettings &p_settings, bool p_adaptive_base_pass) {
+ RD::get_singleton()->compute_list_bind_uniform_set(p_compute_list, ssao.gather_uniform_set, 0);
+ if ((p_settings.quality == RS::ENV_SSAO_QUALITY_ULTRA) && !p_adaptive_base_pass) {
+ RD::get_singleton()->compute_list_bind_uniform_set(p_compute_list, ssao.importance_map_uniform_set, 1);
+ }
+
+ for (int i = 0; i < 4; i++) {
+ if ((p_settings.quality == RS::ENV_SSAO_QUALITY_VERY_LOW) && ((i == 1) || (i == 2))) {
+ continue;
+ }
+
+ ssao.gather_push_constant.pass_coord_offset[0] = i % 2;
+ ssao.gather_push_constant.pass_coord_offset[1] = i / 2;
+ ssao.gather_push_constant.pass_uv_offset[0] = ((i % 2) - 0.0) / p_settings.screen_size.x;
+ ssao.gather_push_constant.pass_uv_offset[1] = ((i / 2) - 0.0) / p_settings.screen_size.y;
+ ssao.gather_push_constant.pass = i;
+ RD::get_singleton()->compute_list_bind_uniform_set(p_compute_list, _get_uniform_set_from_image(p_ao_slices[i]), 2);
+ RD::get_singleton()->compute_list_set_push_constant(p_compute_list, &ssao.gather_push_constant, sizeof(SSAOGatherPushConstant));
+
+ int x_groups = ((p_settings.screen_size.x >> (p_settings.half_size ? 2 : 1)) - 1) / 8 + 1;
+ int y_groups = ((p_settings.screen_size.y >> (p_settings.half_size ? 2 : 1)) - 1) / 8 + 1;
+ RD::get_singleton()->compute_list_dispatch(p_compute_list, x_groups, y_groups, 1);
+ }
+ RD::get_singleton()->compute_list_add_barrier(p_compute_list);
+}
+
+void EffectsRD::generate_ssao(RID p_depth_buffer, RID p_normal_buffer, RID p_depth_mipmaps_texture, const Vector<RID> &depth_mipmaps, RID p_ao, const Vector<RID> p_ao_slices, RID p_ao_pong, const Vector<RID> p_ao_pong_slices, RID p_upscale_buffer, RID p_importance_map, RID p_importance_map_pong, const CameraMatrix &p_projection, const SSAOSettings &p_settings, bool p_invalidate_uniform_sets) {
RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin();
/* FIRST PASS */
- // Minify the depth buffer.
-
- for (int i = 0; i < depth_mipmaps.size(); i++) {
- if (i == 0) {
- RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, ssao.pipelines[SSAO_MINIFY_FIRST]);
- RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_depth_buffer), 0);
- } else {
- if (i == 1) {
- RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, ssao.pipelines[SSAO_MINIFY_MIPMAP]);
+ // Downsample and deinterleave the depth buffer.
+ {
+ if (p_invalidate_uniform_sets) {
+ Vector<RD::Uniform> uniforms;
+ {
+ RD::Uniform u;
+ u.uniform_type = RD::UNIFORM_TYPE_IMAGE;
+ u.binding = 0;
+ u.ids.push_back(depth_mipmaps[1]);
+ uniforms.push_back(u);
}
+ {
+ RD::Uniform u;
+ u.uniform_type = RD::UNIFORM_TYPE_IMAGE;
+ u.binding = 1;
+ u.ids.push_back(depth_mipmaps[2]);
+ uniforms.push_back(u);
+ }
+ {
+ RD::Uniform u;
+ u.uniform_type = RD::UNIFORM_TYPE_IMAGE;
+ u.binding = 2;
+ u.ids.push_back(depth_mipmaps[3]);
+ uniforms.push_back(u);
+ }
+ ssao.downsample_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, ssao.downsample_shader.version_get_shader(ssao.downsample_shader_version, 2), 2);
+ }
+
+ float depth_linearize_mul = -p_projection.matrix[3][2];
+ float depth_linearize_add = p_projection.matrix[2][2];
+ if (depth_linearize_mul * depth_linearize_add < 0) {
+ depth_linearize_add = -depth_linearize_add;
+ }
+
+ ssao.downsample_push_constant.orthogonal = p_projection.is_orthogonal();
+ ssao.downsample_push_constant.z_near = depth_linearize_mul;
+ ssao.downsample_push_constant.z_far = depth_linearize_add;
+ if (ssao.downsample_push_constant.orthogonal) {
+ ssao.downsample_push_constant.z_near = p_projection.get_z_near();
+ ssao.downsample_push_constant.z_far = p_projection.get_z_far();
+ }
+ ssao.downsample_push_constant.pixel_size[0] = 1.0 / p_settings.screen_size.x;
+ ssao.downsample_push_constant.pixel_size[1] = 1.0 / p_settings.screen_size.y;
+ ssao.downsample_push_constant.radius_sq = p_settings.radius * p_settings.radius;
+
+ int downsample_pipeline = SSAO_DOWNSAMPLE;
+ if (p_settings.quality == RS::ENV_SSAO_QUALITY_VERY_LOW) {
+ downsample_pipeline = SSAO_DOWNSAMPLE_HALF;
+ } else if (p_settings.quality > RS::ENV_SSAO_QUALITY_MEDIUM) {
+ downsample_pipeline = SSAO_DOWNSAMPLE_MIPMAP;
+ }
- RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(depth_mipmaps[i - 1]), 0);
+ if (p_settings.half_size) {
+ downsample_pipeline++;
}
- RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(depth_mipmaps[i]), 1);
- RD::get_singleton()->compute_list_set_push_constant(compute_list, &ssao.minify_push_constant, sizeof(SSAOMinifyPushConstant));
- // shrink after set
- ssao.minify_push_constant.source_size[0] = MAX(1, ssao.minify_push_constant.source_size[0] >> 1);
- ssao.minify_push_constant.source_size[1] = MAX(1, ssao.minify_push_constant.source_size[1] >> 1);
+ RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, ssao.pipelines[downsample_pipeline]);
+ RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_depth_buffer), 0);
+ RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(depth_mipmaps[0]), 1);
+ if (p_settings.quality > RS::ENV_SSAO_QUALITY_MEDIUM) {
+ RD::get_singleton()->compute_list_bind_uniform_set(compute_list, ssao.downsample_uniform_set, 2);
+ }
+ RD::get_singleton()->compute_list_set_push_constant(compute_list, &ssao.downsample_push_constant, sizeof(SSAODownsamplePushConstant));
- int x_groups = (ssao.minify_push_constant.source_size[0] - 1) / 8 + 1;
- int y_groups = (ssao.minify_push_constant.source_size[1] - 1) / 8 + 1;
+ int x_groups = (MAX(1, p_settings.screen_size.x >> (p_settings.half_size ? 2 : 1)) - 1) / 8 + 1;
+ int y_groups = (MAX(1, p_settings.screen_size.y >> (p_settings.half_size ? 2 : 1)) - 1) / 8 + 1;
RD::get_singleton()->compute_list_dispatch(compute_list, x_groups, y_groups, 1);
RD::get_singleton()->compute_list_add_barrier(compute_list);
}
/* SECOND PASS */
- // Gather samples
-
- RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, ssao.pipelines[(SSAO_GATHER_LOW + p_quality) + (p_half_size ? 4 : 0)]);
- RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_depth_mipmaps_texture), 0);
- RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_ao1), 1);
- if (!p_half_size) {
- RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_depth_buffer), 2);
- }
- RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_normal_buffer), 3);
-
- ssao.gather_push_constant.screen_size[0] = p_depth_buffer_size.x;
- ssao.gather_push_constant.screen_size[1] = p_depth_buffer_size.y;
- if (p_half_size) {
- ssao.gather_push_constant.screen_size[0] >>= 1;
- ssao.gather_push_constant.screen_size[1] >>= 1;
- }
- ssao.gather_push_constant.z_far = p_projection.get_z_far();
- ssao.gather_push_constant.z_near = p_projection.get_z_near();
- ssao.gather_push_constant.orthogonal = p_projection.is_orthogonal();
-
- ssao.gather_push_constant.proj_info[0] = -2.0f / (ssao.gather_push_constant.screen_size[0] * p_projection.matrix[0][0]);
- ssao.gather_push_constant.proj_info[1] = -2.0f / (ssao.gather_push_constant.screen_size[1] * p_projection.matrix[1][1]);
- ssao.gather_push_constant.proj_info[2] = (1.0f - p_projection.matrix[0][2]) / p_projection.matrix[0][0];
- ssao.gather_push_constant.proj_info[3] = (1.0f + p_projection.matrix[1][2]) / p_projection.matrix[1][1];
- //ssao.gather_push_constant.proj_info[2] = (1.0f - p_projection.matrix[0][2]) / p_projection.matrix[0][0];
- //ssao.gather_push_constant.proj_info[3] = -(1.0f + p_projection.matrix[1][2]) / p_projection.matrix[1][1];
-
- ssao.gather_push_constant.radius = p_radius;
-
- ssao.gather_push_constant.proj_scale = float(p_projection.get_pixels_per_meter(ssao.gather_push_constant.screen_size[0]));
- ssao.gather_push_constant.bias = p_bias;
- ssao.gather_push_constant.intensity_div_r6 = p_intensity / pow(p_radius, 6.0f);
-
- ssao.gather_push_constant.pixel_size[0] = 1.0 / p_depth_buffer_size.x;
- ssao.gather_push_constant.pixel_size[1] = 1.0 / p_depth_buffer_size.y;
-
- RD::get_singleton()->compute_list_set_push_constant(compute_list, &ssao.gather_push_constant, sizeof(SSAOGatherPushConstant));
+ // Sample SSAO
+ {
+ ssao.gather_push_constant.screen_size[0] = p_settings.screen_size.x;
+ ssao.gather_push_constant.screen_size[1] = p_settings.screen_size.y;
+
+ ssao.gather_push_constant.half_screen_pixel_size[0] = 1.0 / p_settings.half_screen_size.x;
+ ssao.gather_push_constant.half_screen_pixel_size[1] = 1.0 / p_settings.half_screen_size.y;
+ float tan_half_fov_x = 1.0 / p_projection.matrix[0][0];
+ float tan_half_fov_y = 1.0 / p_projection.matrix[1][1];
+ ssao.gather_push_constant.NDC_to_view_mul[0] = tan_half_fov_x * 2.0;
+ ssao.gather_push_constant.NDC_to_view_mul[1] = tan_half_fov_y * -2.0;
+ ssao.gather_push_constant.NDC_to_view_add[0] = tan_half_fov_x * -1.0;
+ ssao.gather_push_constant.NDC_to_view_add[1] = tan_half_fov_y;
+ ssao.gather_push_constant.is_orthogonal = p_projection.is_orthogonal();
+
+ ssao.gather_push_constant.half_screen_pixel_size_x025[0] = ssao.gather_push_constant.half_screen_pixel_size[0] * 0.25;
+ ssao.gather_push_constant.half_screen_pixel_size_x025[1] = ssao.gather_push_constant.half_screen_pixel_size[1] * 0.25;
+
+ float radius_near_limit = (p_settings.radius * 1.2f);
+ if (p_settings.quality <= RS::ENV_SSAO_QUALITY_LOW) {
+ radius_near_limit *= 1.50f;
+
+ if (p_settings.quality == RS::ENV_SSAO_QUALITY_VERY_LOW) {
+ ssao.gather_push_constant.radius *= 0.8f;
+ }
+ if (p_settings.half_size) {
+ ssao.gather_push_constant.radius *= 0.5f;
+ }
+ }
+ radius_near_limit /= tan_half_fov_y;
+ ssao.gather_push_constant.radius = p_settings.radius;
+ ssao.gather_push_constant.intensity = p_settings.intensity;
+ ssao.gather_push_constant.shadow_power = p_settings.power;
+ ssao.gather_push_constant.shadow_clamp = 0.98;
+ ssao.gather_push_constant.fade_out_mul = -1.0 / (p_settings.fadeout_to - p_settings.fadeout_from);
+ ssao.gather_push_constant.fade_out_add = p_settings.fadeout_from / (p_settings.fadeout_to - p_settings.fadeout_from) + 1.0;
+ ssao.gather_push_constant.horizon_angle_threshold = p_settings.horizon;
+ ssao.gather_push_constant.inv_radius_near_limit = 1.0f / radius_near_limit;
+ ssao.gather_push_constant.neg_inv_radius = -1.0 / ssao.gather_push_constant.radius;
+
+ ssao.gather_push_constant.load_counter_avg_div = 9.0 / float((p_settings.quarter_size.x) * (p_settings.quarter_size.y) * 255);
+ ssao.gather_push_constant.adaptive_sample_limit = p_settings.adaptive_target;
+
+ ssao.gather_push_constant.detail_intensity = p_settings.detail;
+ ssao.gather_push_constant.quality = MAX(0, p_settings.quality - 1);
+ ssao.gather_push_constant.size_multiplier = p_settings.half_size ? 2 : 1;
+
+ if (p_invalidate_uniform_sets) {
+ Vector<RD::Uniform> uniforms;
+ {
+ RD::Uniform u;
+ u.uniform_type = RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE;
+ u.binding = 0;
+ u.ids.push_back(ssao.mirror_sampler);
+ u.ids.push_back(p_depth_mipmaps_texture);
+ uniforms.push_back(u);
+ }
+ {
+ RD::Uniform u;
+ u.uniform_type = RD::UNIFORM_TYPE_IMAGE;
+ u.binding = 1;
+ u.ids.push_back(p_normal_buffer);
+ uniforms.push_back(u);
+ }
+ {
+ RD::Uniform u;
+ u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER;
+ u.binding = 2;
+ u.ids.push_back(ssao.gather_constants_buffer);
+ uniforms.push_back(u);
+ }
+ ssao.gather_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, ssao.gather_shader.version_get_shader(ssao.gather_shader_version, 0), 0);
+ }
- int x_groups = (ssao.gather_push_constant.screen_size[0] - 1) / 8 + 1;
- int y_groups = (ssao.gather_push_constant.screen_size[1] - 1) / 8 + 1;
+ if (p_invalidate_uniform_sets) {
+ Vector<RD::Uniform> uniforms;
+ {
+ RD::Uniform u;
+ u.uniform_type = RD::UNIFORM_TYPE_IMAGE;
+ u.binding = 0;
+ u.ids.push_back(p_ao_pong);
+ uniforms.push_back(u);
+ }
+ {
+ RD::Uniform u;
+ u.uniform_type = RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE;
+ u.binding = 1;
+ u.ids.push_back(default_sampler);
+ u.ids.push_back(p_importance_map);
+ uniforms.push_back(u);
+ }
+ {
+ RD::Uniform u;
+ u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER;
+ u.binding = 2;
+ u.ids.push_back(ssao.importance_map_load_counter);
+ uniforms.push_back(u);
+ }
+ ssao.importance_map_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, ssao.gather_shader.version_get_shader(ssao.gather_shader_version, 2), 1);
+ }
- RD::get_singleton()->compute_list_dispatch(compute_list, x_groups, y_groups, 1);
- RD::get_singleton()->compute_list_add_barrier(compute_list);
+ if (p_settings.quality == RS::ENV_SSAO_QUALITY_ULTRA) {
+ ssao.importance_map_push_constant.half_screen_pixel_size[0] = 1.0 / p_settings.half_screen_size.x;
+ ssao.importance_map_push_constant.half_screen_pixel_size[1] = 1.0 / p_settings.half_screen_size.y;
+ ssao.importance_map_push_constant.intensity = p_settings.intensity;
+ ssao.importance_map_push_constant.power = p_settings.power;
+ //base pass
+ RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, ssao.pipelines[SSAO_GATHER_BASE]);
+ gather_ssao(compute_list, p_ao_pong_slices, p_settings, true);
+ //generate importance map
+ int x_groups = (p_settings.quarter_size.x - 1) / 8 + 1;
+ int y_groups = (p_settings.quarter_size.y - 1) / 8 + 1;
+
+ RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, ssao.pipelines[SSAO_GENERATE_IMPORTANCE_MAP]);
+ RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_ao_pong), 0);
+ RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_importance_map), 1);
+ RD::get_singleton()->compute_list_set_push_constant(compute_list, &ssao.importance_map_push_constant, sizeof(SSAOImportanceMapPushConstant));
+ RD::get_singleton()->compute_list_dispatch(compute_list, x_groups, y_groups, 1);
+ RD::get_singleton()->compute_list_add_barrier(compute_list);
+ //process importance map A
+ RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, ssao.pipelines[SSAO_PROCESS_IMPORTANCE_MAPA]);
+ RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_importance_map), 0);
+ RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_importance_map_pong), 1);
+ RD::get_singleton()->compute_list_set_push_constant(compute_list, &ssao.importance_map_push_constant, sizeof(SSAOImportanceMapPushConstant));
+ RD::get_singleton()->compute_list_dispatch(compute_list, x_groups, y_groups, 1);
+ RD::get_singleton()->compute_list_add_barrier(compute_list);
+ //process Importance Map B
+ RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, ssao.pipelines[SSAO_PROCESS_IMPORTANCE_MAPB]);
+ RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_importance_map_pong), 0);
+ RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_importance_map), 1);
+ RD::get_singleton()->compute_list_bind_uniform_set(compute_list, ssao.counter_uniform_set, 2);
+ RD::get_singleton()->compute_list_set_push_constant(compute_list, &ssao.importance_map_push_constant, sizeof(SSAOImportanceMapPushConstant));
+ RD::get_singleton()->compute_list_dispatch(compute_list, x_groups, y_groups, 1);
+ RD::get_singleton()->compute_list_add_barrier(compute_list);
- /* THIRD PASS */
- // Blur horizontal
-
- ssao.blur_push_constant.edge_sharpness = p_edge_sharpness;
- ssao.blur_push_constant.filter_scale = p_blur;
- ssao.blur_push_constant.screen_size[0] = ssao.gather_push_constant.screen_size[0];
- ssao.blur_push_constant.screen_size[1] = ssao.gather_push_constant.screen_size[1];
- ssao.blur_push_constant.z_far = p_projection.get_z_far();
- ssao.blur_push_constant.z_near = p_projection.get_z_near();
- ssao.blur_push_constant.orthogonal = p_projection.is_orthogonal();
- ssao.blur_push_constant.axis[0] = 1;
- ssao.blur_push_constant.axis[1] = 0;
-
- if (p_blur != RS::ENV_SSAO_BLUR_DISABLED) {
- RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, ssao.pipelines[p_half_size ? SSAO_BLUR_PASS_HALF : SSAO_BLUR_PASS]);
- RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_ao1), 0);
- if (p_half_size) {
- RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_depth_mipmaps_texture), 1);
+ RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, ssao.pipelines[SSAO_GATHER_ADAPTIVE]);
} else {
- RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_depth_buffer), 1);
+ RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, ssao.pipelines[SSAO_GATHER]);
}
- RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_ao2), 3);
-
- RD::get_singleton()->compute_list_set_push_constant(compute_list, &ssao.blur_push_constant, sizeof(SSAOBlurPushConstant));
- RD::get_singleton()->compute_list_dispatch(compute_list, x_groups, y_groups, 1);
- RD::get_singleton()->compute_list_add_barrier(compute_list);
-
- /* THIRD PASS */
- // Blur vertical
-
- ssao.blur_push_constant.axis[0] = 0;
- ssao.blur_push_constant.axis[1] = 1;
+ gather_ssao(compute_list, p_ao_slices, p_settings, false);
+ }
- RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_ao2), 0);
- RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_ao1), 3);
+ // /* THIRD PASS */
+ // // Blur
+ //
+ {
+ ssao.blur_push_constant.edge_sharpness = 1.0 - p_settings.sharpness;
+ ssao.blur_push_constant.half_screen_pixel_size[0] = 1.0 / p_settings.half_screen_size.x;
+ ssao.blur_push_constant.half_screen_pixel_size[1] = 1.0 / p_settings.half_screen_size.y;
+
+ int blur_passes = p_settings.quality > RS::ENV_SSAO_QUALITY_VERY_LOW ? p_settings.blur_passes : 1;
+
+ for (int pass = 0; pass < blur_passes; pass++) {
+ int blur_pipeline = SSAO_BLUR_PASS;
+ if (p_settings.quality > RS::ENV_SSAO_QUALITY_VERY_LOW) {
+ if (pass < blur_passes - 2) {
+ blur_pipeline = SSAO_BLUR_PASS_WIDE;
+ }
+ blur_pipeline = SSAO_BLUR_PASS_SMART;
+ }
- RD::get_singleton()->compute_list_set_push_constant(compute_list, &ssao.blur_push_constant, sizeof(SSAOBlurPushConstant));
+ for (int i = 0; i < 4; i++) {
+ if ((p_settings.quality == RS::ENV_SSAO_QUALITY_VERY_LOW) && ((i == 1) || (i == 2))) {
+ continue;
+ }
+
+ RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, ssao.pipelines[blur_pipeline]);
+ if (pass % 2 == 0) {
+ if (p_settings.quality == RS::ENV_SSAO_QUALITY_VERY_LOW) {
+ RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_ao_slices[i]), 0);
+ } else {
+ RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture_and_sampler(p_ao_slices[i], ssao.mirror_sampler), 0);
+ }
+
+ RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_ao_pong_slices[i]), 1);
+ } else {
+ if (p_settings.quality == RS::ENV_SSAO_QUALITY_VERY_LOW) {
+ RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_ao_pong_slices[i]), 0);
+ } else {
+ RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture_and_sampler(p_ao_pong_slices[i], ssao.mirror_sampler), 0);
+ }
+ RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_ao_slices[i]), 1);
+ }
+ RD::get_singleton()->compute_list_set_push_constant(compute_list, &ssao.blur_push_constant, sizeof(SSAOBlurPushConstant));
+
+ int x_groups = ((p_settings.screen_size.x >> (p_settings.half_size ? 2 : 1)) - 1) / 8 + 1;
+ int y_groups = ((p_settings.screen_size.y >> (p_settings.half_size ? 2 : 1)) - 1) / 8 + 1;
+
+ RD::get_singleton()->compute_list_dispatch(compute_list, x_groups, y_groups, 1);
+ }
- RD::get_singleton()->compute_list_dispatch(compute_list, x_groups, y_groups, 1);
+ if (p_settings.quality > RS::ENV_SSAO_QUALITY_VERY_LOW) {
+ RD::get_singleton()->compute_list_add_barrier(compute_list);
+ }
+ }
}
- if (p_half_size) { //must upscale
- /* FOURTH PASS */
- // upscale if half size
- //back to full size
- ssao.blur_push_constant.screen_size[0] = p_depth_buffer_size.x;
- ssao.blur_push_constant.screen_size[1] = p_depth_buffer_size.y;
+ /* FOURTH PASS */
+ // Interleave buffers
+ // back to full size
+ {
+ ssao.interleave_push_constant.inv_sharpness = 1.0 - p_settings.sharpness;
+ ssao.interleave_push_constant.pixel_size[0] = 1.0 / p_settings.screen_size.x;
+ ssao.interleave_push_constant.pixel_size[1] = 1.0 / p_settings.screen_size.y;
+ ssao.interleave_push_constant.size_modifier = uint32_t(p_settings.half_size ? 4 : 2);
+
+ int interleave_pipeline = SSAO_INTERLEAVE_HALF;
+ if (p_settings.quality == RS::ENV_SSAO_QUALITY_LOW) {
+ interleave_pipeline = SSAO_INTERLEAVE;
+ } else if (p_settings.quality >= RS::ENV_SSAO_QUALITY_MEDIUM) {
+ interleave_pipeline = SSAO_INTERLEAVE_SMART;
+ }
- RD::get_singleton()->compute_list_add_barrier(compute_list);
- RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, ssao.pipelines[SSAO_BLUR_UPSCALE]);
+ RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, ssao.pipelines[interleave_pipeline]);
- RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_ao1), 0);
- RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_upscale_buffer), 3);
- RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_depth_buffer), 1);
- RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_depth_mipmaps_texture), 2);
+ RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_upscale_buffer), 0);
+ if (p_settings.quality > RS::ENV_SSAO_QUALITY_VERY_LOW && p_settings.blur_passes % 2 == 0) {
+ RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_ao), 1);
+ } else {
+ RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_ao_pong), 1);
+ }
- RD::get_singleton()->compute_list_set_push_constant(compute_list, &ssao.blur_push_constant, sizeof(SSAOBlurPushConstant)); //not used but set anyway
+ RD::get_singleton()->compute_list_set_push_constant(compute_list, &ssao.interleave_push_constant, sizeof(SSAOInterleavePushConstant));
- x_groups = (p_depth_buffer_size.x - 1) / 8 + 1;
- y_groups = (p_depth_buffer_size.y - 1) / 8 + 1;
+ int x_groups = (p_settings.screen_size.x - 1) / 8 + 1;
+ int y_groups = (p_settings.screen_size.y - 1) / 8 + 1;
RD::get_singleton()->compute_list_dispatch(compute_list, x_groups, y_groups, 1);
+ RD::get_singleton()->compute_list_add_barrier(compute_list);
}
RD::get_singleton()->compute_list_end();
+
+ int zero[1] = { 0 };
+ RD::get_singleton()->buffer_update(ssao.importance_map_load_counter, 0, sizeof(uint32_t), &zero, false);
}
void EffectsRD::roughness_limit(RID p_source_normal, RID p_roughness, const Size2i &p_size, float p_curve) {
@@ -1486,57 +1702,142 @@ EffectsRD::EffectsRD() {
{
// Initialize ssao
+
+ RD::SamplerState sampler;
+ sampler.mag_filter = RD::SAMPLER_FILTER_NEAREST;
+ sampler.min_filter = RD::SAMPLER_FILTER_NEAREST;
+ sampler.mip_filter = RD::SAMPLER_FILTER_NEAREST;
+ sampler.repeat_u = RD::SAMPLER_REPEAT_MODE_MIRRORED_REPEAT;
+ sampler.repeat_v = RD::SAMPLER_REPEAT_MODE_MIRRORED_REPEAT;
+ sampler.repeat_w = RD::SAMPLER_REPEAT_MODE_MIRRORED_REPEAT;
+ sampler.max_lod = 4;
+
+ ssao.mirror_sampler = RD::get_singleton()->sampler_create(sampler);
+
uint32_t pipeline = 0;
{
Vector<String> ssao_modes;
- ssao_modes.push_back("\n#define MINIFY_START\n");
ssao_modes.push_back("\n");
+ ssao_modes.push_back("\n#define USE_HALF_SIZE\n");
+ ssao_modes.push_back("\n#define GENERATE_MIPS\n");
+ ssao_modes.push_back("\n#define GENERATE_MIPS\n#define USE_HALF_SIZE");
+ ssao_modes.push_back("\n#define USE_HALF_BUFFERS\n");
+ ssao_modes.push_back("\n#define USE_HALF_BUFFERS\n#define USE_HALF_SIZE");
- ssao.minify_shader.initialize(ssao_modes);
+ ssao.downsample_shader.initialize(ssao_modes);
- ssao.minify_shader_version = ssao.minify_shader.version_create();
+ ssao.downsample_shader_version = ssao.downsample_shader.version_create();
- for (int i = 0; i <= SSAO_MINIFY_MIPMAP; i++) {
- ssao.pipelines[pipeline] = RD::get_singleton()->compute_pipeline_create(ssao.minify_shader.version_get_shader(ssao.minify_shader_version, i));
+ for (int i = 0; i <= SSAO_DOWNSAMPLE_HALF_RES_HALF; i++) {
+ ssao.pipelines[pipeline] = RD::get_singleton()->compute_pipeline_create(ssao.downsample_shader.version_get_shader(ssao.downsample_shader_version, i));
pipeline++;
}
}
{
Vector<String> ssao_modes;
- ssao_modes.push_back("\n#define SSAO_QUALITY_LOW\n");
+
ssao_modes.push_back("\n");
- ssao_modes.push_back("\n#define SSAO_QUALITY_HIGH\n");
- ssao_modes.push_back("\n#define SSAO_QUALITY_ULTRA\n");
- ssao_modes.push_back("\n#define SSAO_QUALITY_LOW\n#define USE_HALF_SIZE\n");
- ssao_modes.push_back("\n#define USE_HALF_SIZE\n");
- ssao_modes.push_back("\n#define SSAO_QUALITY_HIGH\n#define USE_HALF_SIZE\n");
- ssao_modes.push_back("\n#define SSAO_QUALITY_ULTRA\n#define USE_HALF_SIZE\n");
+ ssao_modes.push_back("\n#define SSAO_BASE\n");
+ ssao_modes.push_back("\n#define ADAPTIVE\n");
ssao.gather_shader.initialize(ssao_modes);
ssao.gather_shader_version = ssao.gather_shader.version_create();
- for (int i = SSAO_GATHER_LOW; i <= SSAO_GATHER_ULTRA_HALF; i++) {
- ssao.pipelines[pipeline] = RD::get_singleton()->compute_pipeline_create(ssao.gather_shader.version_get_shader(ssao.gather_shader_version, i - SSAO_GATHER_LOW));
+ for (int i = SSAO_GATHER; i <= SSAO_GATHER_ADAPTIVE; i++) {
+ ssao.pipelines[pipeline] = RD::get_singleton()->compute_pipeline_create(ssao.gather_shader.version_get_shader(ssao.gather_shader_version, i - SSAO_GATHER));
pipeline++;
}
+
+ ssao.gather_constants_buffer = RD::get_singleton()->uniform_buffer_create(sizeof(SSAOGatherConstants));
+ SSAOGatherConstants gather_constants;
+
+ const int sub_pass_count = 5;
+ for (int pass = 0; pass < 4; pass++) {
+ for (int subPass = 0; subPass < sub_pass_count; subPass++) {
+ int a = pass;
+ int b = subPass;
+
+ int spmap[5]{ 0, 1, 4, 3, 2 };
+ b = spmap[subPass];
+
+ float ca, sa;
+ float angle0 = (float(a) + float(b) / float(sub_pass_count)) * Math_PI * 0.5f;
+
+ ca = Math::cos(angle0);
+ sa = Math::sin(angle0);
+
+ float scale = 1.0f + (a - 1.5f + (b - (sub_pass_count - 1.0f) * 0.5f) / float(sub_pass_count)) * 0.07f;
+
+ gather_constants.rotation_matrices[pass * 20 + subPass * 4 + 0] = scale * ca;
+ gather_constants.rotation_matrices[pass * 20 + subPass * 4 + 1] = scale * -sa;
+ gather_constants.rotation_matrices[pass * 20 + subPass * 4 + 2] = -scale * sa;
+ gather_constants.rotation_matrices[pass * 20 + subPass * 4 + 3] = -scale * ca;
+ }
+ }
+
+ RD::get_singleton()->buffer_update(ssao.gather_constants_buffer, 0, sizeof(SSAOGatherConstants), &gather_constants, false);
}
{
Vector<String> ssao_modes;
- ssao_modes.push_back("\n#define MODE_FULL_SIZE\n");
- ssao_modes.push_back("\n");
- ssao_modes.push_back("\n#define MODE_UPSCALE\n");
+ ssao_modes.push_back("\n#define GENERATE_MAP\n");
+ ssao_modes.push_back("\n#define PROCESS_MAPA\n");
+ ssao_modes.push_back("\n#define PROCESS_MAPB\n");
+
+ ssao.importance_map_shader.initialize(ssao_modes);
+
+ ssao.importance_map_shader_version = ssao.importance_map_shader.version_create();
+
+ for (int i = SSAO_GENERATE_IMPORTANCE_MAP; i <= SSAO_PROCESS_IMPORTANCE_MAPB; i++) {
+ ssao.pipelines[pipeline] = RD::get_singleton()->compute_pipeline_create(ssao.importance_map_shader.version_get_shader(ssao.importance_map_shader_version, i - SSAO_GENERATE_IMPORTANCE_MAP));
+
+ pipeline++;
+ }
+ ssao.importance_map_load_counter = RD::get_singleton()->storage_buffer_create(sizeof(uint32_t));
+ int zero[1] = { 0 };
+ RD::get_singleton()->buffer_update(ssao.importance_map_load_counter, 0, sizeof(uint32_t), &zero, false);
+
+ Vector<RD::Uniform> uniforms;
+ {
+ RD::Uniform u;
+ u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER;
+ u.binding = 0;
+ u.ids.push_back(ssao.importance_map_load_counter);
+ uniforms.push_back(u);
+ }
+ ssao.counter_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, ssao.importance_map_shader.version_get_shader(ssao.importance_map_shader_version, 2), 2);
+ }
+ {
+ Vector<String> ssao_modes;
+ ssao_modes.push_back("\n#define MODE_NON_SMART\n");
+ ssao_modes.push_back("\n#define MODE_SMART\n");
+ ssao_modes.push_back("\n#define MODE_WIDE\n");
ssao.blur_shader.initialize(ssao_modes);
ssao.blur_shader_version = ssao.blur_shader.version_create();
- for (int i = SSAO_BLUR_PASS; i <= SSAO_BLUR_UPSCALE; i++) {
+ for (int i = SSAO_BLUR_PASS; i <= SSAO_BLUR_PASS_WIDE; i++) {
ssao.pipelines[pipeline] = RD::get_singleton()->compute_pipeline_create(ssao.blur_shader.version_get_shader(ssao.blur_shader_version, i - SSAO_BLUR_PASS));
pipeline++;
}
}
+ {
+ Vector<String> ssao_modes;
+ ssao_modes.push_back("\n#define MODE_NON_SMART\n");
+ ssao_modes.push_back("\n#define MODE_SMART\n");
+ ssao_modes.push_back("\n#define MODE_HALF\n");
+
+ ssao.interleave_shader.initialize(ssao_modes);
+
+ ssao.interleave_shader_version = ssao.interleave_shader.version_create();
+ for (int i = SSAO_INTERLEAVE; i <= SSAO_INTERLEAVE_HALF; i++) {
+ ssao.pipelines[pipeline] = RD::get_singleton()->compute_pipeline_create(ssao.interleave_shader.version_get_shader(ssao.interleave_shader_version, i - SSAO_INTERLEAVE));
+
+ pipeline++;
+ }
+ }
ERR_FAIL_COND(pipeline != SSAO_MAX);
}
@@ -1777,6 +2078,10 @@ EffectsRD::~EffectsRD() {
RD::get_singleton()->free(index_buffer); //array gets freed as dependency
RD::get_singleton()->free(filter.coefficient_buffer);
+ RD::get_singleton()->free(ssao.mirror_sampler);
+ RD::get_singleton()->free(ssao.gather_constants_buffer);
+ RD::get_singleton()->free(ssao.importance_map_load_counter);
+
bokeh.shader.version_free(bokeh.shader_version);
copy.shader.version_free(copy.shader_version);
copy_to_fb.shader.version_free(copy_to_fb.shader_version);
@@ -1791,7 +2096,9 @@ EffectsRD::~EffectsRD() {
specular_merge.shader.version_free(specular_merge.shader_version);
ssao.blur_shader.version_free(ssao.blur_shader_version);
ssao.gather_shader.version_free(ssao.gather_shader_version);
- ssao.minify_shader.version_free(ssao.minify_shader_version);
+ ssao.downsample_shader.version_free(ssao.downsample_shader_version);
+ ssao.interleave_shader.version_free(ssao.interleave_shader_version);
+ ssao.importance_map_shader.version_free(ssao.importance_map_shader_version);
ssr.shader.version_free(ssr.shader_version);
ssr_filter.shader.version_free(ssr_filter.shader_version);
ssr_scale.shader.version_free(ssr_scale.shader_version);
diff --git a/servers/rendering/renderer_rd/effects_rd.h b/servers/rendering/renderer_rd/effects_rd.h
index 3afc111b9d..8731466dea 100644
--- a/servers/rendering/renderer_rd/effects_rd.h
+++ b/servers/rendering/renderer_rd/effects_rd.h
@@ -51,7 +51,9 @@
#include "servers/rendering/renderer_rd/shaders/specular_merge.glsl.gen.h"
#include "servers/rendering/renderer_rd/shaders/ssao.glsl.gen.h"
#include "servers/rendering/renderer_rd/shaders/ssao_blur.glsl.gen.h"
-#include "servers/rendering/renderer_rd/shaders/ssao_minify.glsl.gen.h"
+#include "servers/rendering/renderer_rd/shaders/ssao_downsample.glsl.gen.h"
+#include "servers/rendering/renderer_rd/shaders/ssao_importance_map.glsl.gen.h"
+#include "servers/rendering/renderer_rd/shaders/ssao_interleave.glsl.gen.h"
#include "servers/rendering/renderer_rd/shaders/subsurface_scattering.glsl.gen.h"
#include "servers/rendering/renderer_rd/shaders/tonemap.glsl.gen.h"
@@ -288,71 +290,121 @@ class EffectsRD {
} bokeh;
enum SSAOMode {
- SSAO_MINIFY_FIRST,
- SSAO_MINIFY_MIPMAP,
- SSAO_GATHER_LOW,
- SSAO_GATHER_MEDIUM,
- SSAO_GATHER_HIGH,
- SSAO_GATHER_ULTRA,
- SSAO_GATHER_LOW_HALF,
- SSAO_GATHER_MEDIUM_HALF,
- SSAO_GATHER_HIGH_HALF,
- SSAO_GATHER_ULTRA_HALF,
+ SSAO_DOWNSAMPLE,
+ SSAO_DOWNSAMPLE_HALF_RES,
+ SSAO_DOWNSAMPLE_MIPMAP,
+ SSAO_DOWNSAMPLE_MIPMAP_HALF_RES,
+ SSAO_DOWNSAMPLE_HALF,
+ SSAO_DOWNSAMPLE_HALF_RES_HALF,
+ SSAO_GATHER,
+ SSAO_GATHER_BASE,
+ SSAO_GATHER_ADAPTIVE,
+ SSAO_GENERATE_IMPORTANCE_MAP,
+ SSAO_PROCESS_IMPORTANCE_MAPA,
+ SSAO_PROCESS_IMPORTANCE_MAPB,
SSAO_BLUR_PASS,
- SSAO_BLUR_PASS_HALF,
- SSAO_BLUR_UPSCALE,
+ SSAO_BLUR_PASS_SMART,
+ SSAO_BLUR_PASS_WIDE,
+ SSAO_INTERLEAVE,
+ SSAO_INTERLEAVE_SMART,
+ SSAO_INTERLEAVE_HALF,
SSAO_MAX
};
- struct SSAOMinifyPushConstant {
+ struct SSAODownsamplePushConstant {
float pixel_size[2];
float z_far;
float z_near;
- int32_t source_size[2];
uint32_t orthogonal;
- uint32_t pad;
+ float radius_sq;
+ uint32_t pad[2];
};
struct SSAOGatherPushConstant {
int32_t screen_size[2];
- float z_far;
- float z_near;
+ int pass;
+ int quality;
+
+ float half_screen_pixel_size[2];
+ int size_multiplier;
+ float detail_intensity;
+
+ float NDC_to_view_mul[2];
+ float NDC_to_view_add[2];
+
+ float pad[2];
+ float half_screen_pixel_size_x025[2];
- uint32_t orthogonal;
- float intensity_div_r6;
float radius;
- float bias;
+ float intensity;
+ float shadow_power;
+ float shadow_clamp;
- float proj_info[4];
- float pixel_size[2];
- float proj_scale;
- uint32_t pad;
+ float fade_out_mul;
+ float fade_out_add;
+ float horizon_angle_threshold;
+ float inv_radius_near_limit;
+
+ bool is_orthogonal;
+ float neg_inv_radius;
+ float load_counter_avg_div;
+ float adaptive_sample_limit;
+
+ int32_t pass_coord_offset[2];
+ float pass_uv_offset[2];
+ };
+
+ struct SSAOGatherConstants {
+ float rotation_matrices[80]; //5 vec4s * 4
+ };
+
+ struct SSAOImportanceMapPushConstant {
+ float half_screen_pixel_size[2];
+ float intensity;
+ float power;
};
struct SSAOBlurPushConstant {
float edge_sharpness;
- int32_t filter_scale;
- float z_far;
- float z_near;
- uint32_t orthogonal;
- uint32_t pad[3];
- int32_t axis[2];
- int32_t screen_size[2];
+ float pad;
+ float half_screen_pixel_size[2];
+ };
+
+ struct SSAOInterleavePushConstant {
+ float inv_sharpness;
+ uint32_t size_modifier;
+ float pixel_size[2];
};
struct SSAO {
- SSAOMinifyPushConstant minify_push_constant;
- SsaoMinifyShaderRD minify_shader;
- RID minify_shader_version;
+ SSAODownsamplePushConstant downsample_push_constant;
+ SsaoDownsampleShaderRD downsample_shader;
+ RID downsample_shader_version;
+ RID downsample_uniform_set;
SSAOGatherPushConstant gather_push_constant;
SsaoShaderRD gather_shader;
RID gather_shader_version;
+ RID gather_uniform_set;
+ RID gather_constants_buffer;
+ bool gather_initialized = false;
+
+ SSAOImportanceMapPushConstant importance_map_push_constant;
+ SsaoImportanceMapShaderRD importance_map_shader;
+ RID importance_map_shader_version;
+ RID importance_map_load_counter;
+ RID importance_map_uniform_set;
+ RID counter_uniform_set;
SSAOBlurPushConstant blur_push_constant;
SsaoBlurShaderRD blur_shader;
RID blur_shader_version;
+ SSAOInterleavePushConstant interleave_push_constant;
+ SsaoInterleaveShaderRD interleave_shader;
+ RID interleave_shader_version;
+
+ RID mirror_sampler;
RID pipelines[SSAO_MAX];
} ssao;
@@ -598,13 +650,27 @@ class EffectsRD {
}
};
+ struct TextureSamplerPair {
+ RID texture;
+ RID sampler;
+ _FORCE_INLINE_ bool operator<(const TextureSamplerPair &p_pair) const {
+ if (texture == p_pair.texture) {
+ return sampler < p_pair.sampler;
+ } else {
+ return texture < p_pair.texture;
+ }
+ }
+ };
+
Map<RID, RID> texture_to_compute_uniform_set_cache;
Map<TexturePair, RID> texture_pair_to_compute_uniform_set_cache;
Map<TexturePair, RID> image_pair_to_compute_uniform_set_cache;
+ Map<TextureSamplerPair, RID> texture_sampler_to_compute_uniform_set_cache;
RID _get_uniform_set_from_image(RID p_texture);
RID _get_uniform_set_from_texture(RID p_texture, bool p_use_mipmaps = false);
RID _get_compute_uniform_set_from_texture(RID p_texture, bool p_use_mipmaps = false);
+ RID _get_compute_uniform_set_from_texture_and_sampler(RID p_texture, RID p_sampler);
RID _get_compute_uniform_set_from_texture_pair(RID p_texture, RID p_texture2, bool p_use_mipmaps = false);
RID _get_compute_uniform_set_from_image_pair(RID p_texture, RID p_texture2);
@@ -664,9 +730,30 @@ public:
Vector2i texture_size;
};
+ struct SSAOSettings {
+ float radius = 1.0;
+ float intensity = 2.0;
+ float power = 1.5;
+ float detail = 0.5;
+ float horizon = 0.06;
+ float sharpness = 0.98;
+
+ RS::EnvironmentSSAOQuality quality = RS::ENV_SSAO_QUALITY_MEDIUM;
+ bool half_size = false;
+ float adaptive_target = 0.5;
+ int blur_passes = 2;
+ float fadeout_from = 50.0;
+ float fadeout_to = 300.0;
+
+ Size2i screen_size = Size2i();
+ Size2i half_screen_size = Size2i();
+ Size2i quarter_size = Size2i();
+ };
+
void tonemapper(RID p_source_color, RID p_dst_framebuffer, const TonemapSettings &p_settings);
- void generate_ssao(RID p_depth_buffer, RID p_normal_buffer, const Size2i &p_depth_buffer_size, RID p_depth_mipmaps_texture, const Vector<RID> &depth_mipmaps, RID p_ao1, bool p_half_size, RID p_ao2, RID p_upscale_buffer, float p_intensity, float p_radius, float p_bias, const CameraMatrix &p_projection, RS::EnvironmentSSAOQuality p_quality, RS::EnvironmentSSAOBlur p_blur, float p_edge_sharpness);
+ void gather_ssao(RD::ComputeListID p_compute_list, const Vector<RID> p_ao_slices, const SSAOSettings &p_settings, bool p_adaptive_base_pass);
+ void generate_ssao(RID p_depth_buffer, RID p_normal_buffer, RID p_depth_mipmaps_texture, const Vector<RID> &depth_mipmaps, RID p_ao, const Vector<RID> p_ao_slices, RID p_ao_pong, const Vector<RID> p_ao_pong_slices, RID p_upscale_buffer, RID p_importance_map, RID p_importance_map_pong, const CameraMatrix &p_projection, const SSAOSettings &p_settings, bool p_invalidate_uniform_sets);
void roughness_limit(RID p_source_normal, RID p_roughness, const Size2i &p_size, float p_curve);
void cubemap_downsample(RID p_source_cubemap, RID p_dest_cubemap, const Size2i &p_size);
diff --git a/servers/rendering/renderer_rd/renderer_canvas_render_rd.cpp b/servers/rendering/renderer_rd/renderer_canvas_render_rd.cpp
index 8fa56b182c..591018346a 100644
--- a/servers/rendering/renderer_rd/renderer_canvas_render_rd.cpp
+++ b/servers/rendering/renderer_rd/renderer_canvas_render_rd.cpp
@@ -601,10 +601,10 @@ void RendererCanvasRenderRD::_render_item(RD::DrawListID p_draw_list, const Item
push_constant.flags |= FLAGS_NINEPACH_DRAW_CENTER;
}
- push_constant.ninepatch_margins[0] = np->margin[MARGIN_LEFT];
- push_constant.ninepatch_margins[1] = np->margin[MARGIN_TOP];
- push_constant.ninepatch_margins[2] = np->margin[MARGIN_RIGHT];
- push_constant.ninepatch_margins[3] = np->margin[MARGIN_BOTTOM];
+ push_constant.ninepatch_margins[0] = np->margin[SIDE_LEFT];
+ push_constant.ninepatch_margins[1] = np->margin[SIDE_TOP];
+ push_constant.ninepatch_margins[2] = np->margin[SIDE_RIGHT];
+ push_constant.ninepatch_margins[3] = np->margin[SIDE_BOTTOM];
RD::get_singleton()->draw_list_set_push_constant(p_draw_list, &push_constant, sizeof(PushConstant));
RD::get_singleton()->draw_list_bind_index_array(p_draw_list, shader.quad_index_array);
diff --git a/servers/rendering/renderer_rd/renderer_scene_render_forward.cpp b/servers/rendering/renderer_rd/renderer_scene_render_forward.cpp
index 93f4b7bc8e..80bd429f2f 100644
--- a/servers/rendering/renderer_rd/renderer_scene_render_forward.cpp
+++ b/servers/rendering/renderer_rd/renderer_scene_render_forward.cpp
@@ -766,7 +766,7 @@ void RendererSceneRenderForward::_allocate_normal_roughness_texture(RenderBuffer
tf.format = RD::DATA_FORMAT_R8G8B8A8_UNORM;
tf.width = rb->width;
tf.height = rb->height;
- tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT;
+ tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_STORAGE_BIT;
if (rb->msaa != RS::VIEWPORT_MSAA_DISABLED) {
tf.usage_bits |= RD::TEXTURE_USAGE_CAN_COPY_TO_BIT | RD::TEXTURE_USAGE_STORAGE_BIT;
@@ -782,7 +782,7 @@ void RendererSceneRenderForward::_allocate_normal_roughness_texture(RenderBuffer
fb.push_back(rb->normal_roughness_buffer);
rb->depth_normal_roughness_fb = RD::get_singleton()->framebuffer_create(fb);
} else {
- tf.usage_bits = RD::TEXTURE_USAGE_COLOR_ATTACHMENT_BIT | RD::TEXTURE_USAGE_CAN_COPY_FROM_BIT | RD::TEXTURE_USAGE_SAMPLING_BIT;
+ tf.usage_bits = RD::TEXTURE_USAGE_COLOR_ATTACHMENT_BIT | RD::TEXTURE_USAGE_CAN_COPY_FROM_BIT | RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_STORAGE_BIT;
tf.samples = rb->texture_samples;
rb->normal_roughness_buffer_msaa = RD::get_singleton()->texture_create(tf, RD::TextureView());
@@ -894,7 +894,7 @@ void RendererSceneRenderForward::_fill_instances(RenderList::Element **p_element
} else {
id.gi_offset = 0xFFFFFFFF;
}
- } else if (!e->instance->lightmap_sh.empty()) {
+ } else if (!e->instance->lightmap_sh.is_empty()) {
if (lightmap_captures_used < scene_state.max_lightmap_captures) {
const Color *src_capture = e->instance->lightmap_sh.ptr();
LightmapCaptureData &lcd = scene_state.lightmap_captures[lightmap_captures_used];
@@ -914,7 +914,7 @@ void RendererSceneRenderForward::_fill_instances(RenderList::Element **p_element
id.flags |= INSTANCE_DATA_FLAG_USE_GI_BUFFERS;
}
- if (!low_end && !e->instance->gi_probe_instances.empty()) {
+ if (!low_end && !e->instance->gi_probe_instances.is_empty()) {
uint32_t written = 0;
for (int j = 0; j < e->instance->gi_probe_instances.size(); j++) {
RID probe = e->instance->gi_probe_instances[j];
@@ -1521,7 +1521,7 @@ void RendererSceneRenderForward::_add_geometry_with_material(InstanceBase *p_ins
e->geometry_index = p_geometry_index;
e->material_index = e->material->index;
e->uses_instancing = e->instance->base_type == RS::INSTANCE_MULTIMESH;
- e->uses_lightmap = e->instance->lightmap != nullptr || !e->instance->lightmap_sh.empty();
+ e->uses_lightmap = e->instance->lightmap != nullptr || !e->instance->lightmap_sh.is_empty();
e->uses_forward_gi = has_alpha && (e->instance->gi_probe_instances.size() || p_using_sdfgi);
e->shader_index = e->shader_index;
e->depth_layer = e->instance->depth_layer;
@@ -1532,7 +1532,7 @@ void RendererSceneRenderForward::_add_geometry_with_material(InstanceBase *p_ins
}
}
-void RendererSceneRenderForward::_fill_render_list(InstanceBase **p_cull_result, int p_cull_count, PassMode p_pass_mode, bool p_using_sdfgi) {
+void RendererSceneRenderForward::_fill_render_list(const PagedArray<InstanceBase *> &p_instances, PassMode p_pass_mode, const CameraMatrix &p_cam_projection, const Transform &p_cam_transform, bool p_using_sdfgi) {
scene_state.current_shader_index = 0;
scene_state.current_material_index = 0;
scene_state.used_sss = false;
@@ -1540,12 +1540,19 @@ void RendererSceneRenderForward::_fill_render_list(InstanceBase **p_cull_result,
scene_state.used_normal_texture = false;
scene_state.used_depth_texture = false;
+ Plane near_plane(p_cam_transform.origin, -p_cam_transform.basis.get_axis(Vector3::AXIS_Z));
+ near_plane.d += p_cam_projection.get_z_near();
+ float z_max = p_cam_projection.get_z_far() - p_cam_projection.get_z_near();
+
uint32_t geometry_index = 0;
//fill list
- for (int i = 0; i < p_cull_count; i++) {
- InstanceBase *inst = p_cull_result[i];
+ for (int i = 0; i < (int)p_instances.size(); i++) {
+ InstanceBase *inst = p_instances[i];
+
+ inst->depth = near_plane.distance_to(inst->transform.origin);
+ inst->depth_layer = CLAMP(int(inst->depth * 16 / z_max), 0, 15);
//add geometry for drawing
switch (inst->base_type) {
@@ -1635,14 +1642,14 @@ void RendererSceneRenderForward::_fill_render_list(InstanceBase **p_cull_result,
}
}
-void RendererSceneRenderForward::_setup_lightmaps(InstanceBase **p_lightmap_cull_result, int p_lightmap_cull_count, const Transform &p_cam_transform) {
+void RendererSceneRenderForward::_setup_lightmaps(const PagedArray<InstanceBase *> &p_lightmaps, const Transform &p_cam_transform) {
uint32_t lightmaps_used = 0;
- for (int i = 0; i < p_lightmap_cull_count; i++) {
+ for (int i = 0; i < (int)p_lightmaps.size(); i++) {
if (i >= (int)scene_state.max_lightmaps) {
break;
}
- InstanceBase *lm = p_lightmap_cull_result[i];
+ InstanceBase *lm = p_lightmaps[i];
Basis to_lm = lm->transform.basis.inverse() * p_cam_transform.basis;
to_lm = to_lm.inverse().transposed(); //will transform normals
RendererStorageRD::store_transform_3x3(to_lm, scene_state.lightmaps[i].normal_xform);
@@ -1654,7 +1661,7 @@ void RendererSceneRenderForward::_setup_lightmaps(InstanceBase **p_lightmap_cull
}
}
-void RendererSceneRenderForward::_render_scene(RID p_render_buffer, const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, InstanceBase **p_cull_result, int p_cull_count, int p_directional_light_count, RID *p_gi_probe_cull_result, int p_gi_probe_cull_count, InstanceBase **p_lightmap_cull_result, int p_lightmap_cull_count, RID p_environment, RID p_camera_effects, RID p_shadow_atlas, RID p_reflection_atlas, RID p_reflection_probe, int p_reflection_probe_pass, const Color &p_default_bg_color, float p_screen_lod_threshold) {
+void RendererSceneRenderForward::_render_scene(RID p_render_buffer, const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, const PagedArray<InstanceBase *> &p_instances, int p_directional_light_count, const PagedArray<RID> &p_gi_probes, const PagedArray<InstanceBase *> &p_lightmaps, RID p_environment, RID p_camera_effects, RID p_shadow_atlas, RID p_reflection_atlas, RID p_reflection_probe, int p_reflection_probe_pass, const Color &p_default_bg_color, float p_screen_lod_threshold) {
RenderBufferDataForward *render_buffer = nullptr;
if (p_render_buffer.is_valid()) {
render_buffer = (RenderBufferDataForward *)render_buffers_get_data(p_render_buffer);
@@ -1709,7 +1716,7 @@ void RendererSceneRenderForward::_render_scene(RID p_render_buffer, const Transf
opaque_framebuffer = render_buffer->color_fb;
- if (!low_end && p_gi_probe_cull_count > 0) {
+ if (!low_end && p_gi_probes.size() > 0) {
using_giprobe = true;
render_buffer->ensure_gi();
}
@@ -1776,13 +1783,13 @@ void RendererSceneRenderForward::_render_scene(RID p_render_buffer, const Transf
ERR_FAIL(); //bug?
}
- _setup_lightmaps(p_lightmap_cull_result, p_lightmap_cull_count, p_cam_transform);
+ _setup_lightmaps(p_lightmaps, p_cam_transform);
_setup_environment(p_environment, p_render_buffer, p_cam_projection, p_cam_transform, p_reflection_probe, p_reflection_probe.is_valid(), screen_pixel_size, p_shadow_atlas, !p_reflection_probe.is_valid(), p_default_bg_color, p_cam_projection.get_z_near(), p_cam_projection.get_z_far(), false);
_update_render_base_uniform_set(); //may have changed due to the above (light buffer enlarged, as an example)
render_list.clear();
- _fill_render_list(p_cull_result, p_cull_count, PASS_MODE_COLOR, using_sdfgi);
+ _fill_render_list(p_instances, PASS_MODE_COLOR, p_cam_projection, p_cam_transform, using_sdfgi);
bool using_sss = !low_end && render_buffer && scene_state.used_sss && sub_surface_scattering_get_quality() != RS::SUB_SURFACE_SCATTERING_QUALITY_DISABLED;
@@ -1864,7 +1871,7 @@ void RendererSceneRenderForward::_render_scene(RID p_render_buffer, const Transf
clear_color = p_default_bg_color;
}
- RID rp_uniform_set = _setup_render_pass_uniform_set(p_render_buffer, radiance_texture, p_shadow_atlas, p_reflection_atlas, p_gi_probe_cull_result, p_gi_probe_cull_count);
+ RID rp_uniform_set = _setup_render_pass_uniform_set(p_render_buffer, radiance_texture, p_shadow_atlas, p_reflection_atlas, p_gi_probes);
render_list.sort_by_key(false);
@@ -1903,7 +1910,7 @@ void RendererSceneRenderForward::_render_scene(RID p_render_buffer, const Transf
}
if (using_sdfgi || using_giprobe) {
- _process_gi(p_render_buffer, render_buffer->normal_roughness_buffer, render_buffer->ambient_buffer, render_buffer->reflection_buffer, render_buffer->giprobe_buffer, p_environment, p_cam_projection, p_cam_transform, p_gi_probe_cull_result, p_gi_probe_cull_count);
+ _process_gi(p_render_buffer, render_buffer->normal_roughness_buffer, render_buffer->ambient_buffer, render_buffer->reflection_buffer, render_buffer->giprobe_buffer, p_environment, p_cam_projection, p_cam_transform, p_gi_probes);
}
_setup_environment(p_environment, p_render_buffer, p_cam_projection, p_cam_transform, p_reflection_probe, p_reflection_probe.is_valid(), screen_pixel_size, p_shadow_atlas, !p_reflection_probe.is_valid(), p_default_bg_color, p_cam_projection.get_z_near(), p_cam_projection.get_z_far(), p_render_buffer.is_valid());
@@ -1949,8 +1956,8 @@ void RendererSceneRenderForward::_render_scene(RID p_render_buffer, const Transf
dc.set_depth_correction(true);
CameraMatrix cm = (dc * p_cam_projection) * CameraMatrix(p_cam_transform.affine_inverse());
RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(opaque_framebuffer, RD::INITIAL_ACTION_CONTINUE, will_continue_color ? RD::FINAL_ACTION_CONTINUE : RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_CONTINUE, will_continue_depth ? RD::FINAL_ACTION_CONTINUE : RD::FINAL_ACTION_READ);
- for (int i = 0; i < p_gi_probe_cull_count; i++) {
- _debug_giprobe(p_gi_probe_cull_result[i], draw_list, opaque_framebuffer, cm, get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_GI_PROBE_LIGHTING, get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_GI_PROBE_EMISSION, 1.0);
+ for (int i = 0; i < (int)p_gi_probes.size(); i++) {
+ _debug_giprobe(p_gi_probes[i], draw_list, opaque_framebuffer, cm, get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_GI_PROBE_LIGHTING, get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_GI_PROBE_EMISSION, 1.0);
}
RD::get_singleton()->draw_list_end();
}
@@ -2027,7 +2034,7 @@ void RendererSceneRenderForward::_render_scene(RID p_render_buffer, const Transf
}
}
-void RendererSceneRenderForward::_render_shadow(RID p_framebuffer, InstanceBase **p_cull_result, int p_cull_count, const CameraMatrix &p_projection, const Transform &p_transform, float p_zfar, float p_bias, float p_normal_bias, bool p_use_dp, bool p_use_dp_flip, bool p_use_pancake, const Plane &p_camera_plane, float p_lod_distance_multiplier, float p_screen_lod_threshold) {
+void RendererSceneRenderForward::_render_shadow(RID p_framebuffer, const PagedArray<InstanceBase *> &p_instances, const CameraMatrix &p_projection, const Transform &p_transform, float p_zfar, float p_bias, float p_normal_bias, bool p_use_dp, bool p_use_dp_flip, bool p_use_pancake, const Plane &p_camera_plane, float p_lod_distance_multiplier, float p_screen_lod_threshold) {
RENDER_TIMESTAMP("Setup Rendering Shadow");
_update_render_base_uniform_set();
@@ -2046,9 +2053,9 @@ void RendererSceneRenderForward::_render_shadow(RID p_framebuffer, InstanceBase
PassMode pass_mode = p_use_dp ? PASS_MODE_SHADOW_DP : PASS_MODE_SHADOW;
- _fill_render_list(p_cull_result, p_cull_count, pass_mode);
+ _fill_render_list(p_instances, pass_mode, p_projection, p_transform);
- RID rp_uniform_set = _setup_render_pass_uniform_set(RID(), RID(), RID(), RID(), nullptr, 0);
+ RID rp_uniform_set = _setup_render_pass_uniform_set(RID(), RID(), RID(), RID(), PagedArray<RID>());
RENDER_TIMESTAMP("Render Shadow");
@@ -2064,7 +2071,7 @@ void RendererSceneRenderForward::_render_shadow(RID p_framebuffer, InstanceBase
}
}
-void RendererSceneRenderForward::_render_particle_collider_heightfield(RID p_fb, const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, InstanceBase **p_cull_result, int p_cull_count) {
+void RendererSceneRenderForward::_render_particle_collider_heightfield(RID p_fb, const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, const PagedArray<InstanceBase *> &p_instances) {
RENDER_TIMESTAMP("Setup Render Collider Heightfield");
_update_render_base_uniform_set();
@@ -2079,9 +2086,9 @@ void RendererSceneRenderForward::_render_particle_collider_heightfield(RID p_fb,
PassMode pass_mode = PASS_MODE_SHADOW;
- _fill_render_list(p_cull_result, p_cull_count, pass_mode);
+ _fill_render_list(p_instances, pass_mode, p_cam_projection, p_cam_transform);
- RID rp_uniform_set = _setup_render_pass_uniform_set(RID(), RID(), RID(), RID(), nullptr, 0);
+ RID rp_uniform_set = _setup_render_pass_uniform_set(RID(), RID(), RID(), RID(), PagedArray<RID>());
RENDER_TIMESTAMP("Render Collider Heightield");
@@ -2097,7 +2104,7 @@ void RendererSceneRenderForward::_render_particle_collider_heightfield(RID p_fb,
}
}
-void RendererSceneRenderForward::_render_material(const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, InstanceBase **p_cull_result, int p_cull_count, RID p_framebuffer, const Rect2i &p_region) {
+void RendererSceneRenderForward::_render_material(const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, const PagedArray<InstanceBase *> &p_instances, RID p_framebuffer, const Rect2i &p_region) {
RENDER_TIMESTAMP("Setup Rendering Material");
_update_render_base_uniform_set();
@@ -2112,9 +2119,9 @@ void RendererSceneRenderForward::_render_material(const Transform &p_cam_transfo
render_list.clear();
PassMode pass_mode = PASS_MODE_DEPTH_MATERIAL;
- _fill_render_list(p_cull_result, p_cull_count, pass_mode);
+ _fill_render_list(p_instances, pass_mode, p_cam_projection, p_cam_transform);
- RID rp_uniform_set = _setup_render_pass_uniform_set(RID(), RID(), RID(), RID(), nullptr, 0);
+ RID rp_uniform_set = _setup_render_pass_uniform_set(RID(), RID(), RID(), RID(), PagedArray<RID>());
RENDER_TIMESTAMP("Render Material");
@@ -2136,7 +2143,7 @@ void RendererSceneRenderForward::_render_material(const Transform &p_cam_transfo
}
}
-void RendererSceneRenderForward::_render_uv2(InstanceBase **p_cull_result, int p_cull_count, RID p_framebuffer, const Rect2i &p_region) {
+void RendererSceneRenderForward::_render_uv2(const PagedArray<InstanceBase *> &p_instances, RID p_framebuffer, const Rect2i &p_region) {
RENDER_TIMESTAMP("Setup Rendering UV2");
_update_render_base_uniform_set();
@@ -2151,9 +2158,9 @@ void RendererSceneRenderForward::_render_uv2(InstanceBase **p_cull_result, int p
render_list.clear();
PassMode pass_mode = PASS_MODE_DEPTH_MATERIAL;
- _fill_render_list(p_cull_result, p_cull_count, pass_mode);
+ _fill_render_list(p_instances, pass_mode, CameraMatrix(), Transform());
- RID rp_uniform_set = _setup_render_pass_uniform_set(RID(), RID(), RID(), RID(), nullptr, 0);
+ RID rp_uniform_set = _setup_render_pass_uniform_set(RID(), RID(), RID(), RID(), PagedArray<RID>());
RENDER_TIMESTAMP("Render Material");
@@ -2197,7 +2204,7 @@ void RendererSceneRenderForward::_render_uv2(InstanceBase **p_cull_result, int p
}
}
-void RendererSceneRenderForward::_render_sdfgi(RID p_render_buffers, const Vector3i &p_from, const Vector3i &p_size, const AABB &p_bounds, InstanceBase **p_cull_result, int p_cull_count, const RID &p_albedo_texture, const RID &p_emission_texture, const RID &p_emission_aniso_texture, const RID &p_geom_facing_texture) {
+void RendererSceneRenderForward::_render_sdfgi(RID p_render_buffers, const Vector3i &p_from, const Vector3i &p_size, const AABB &p_bounds, const PagedArray<InstanceBase *> &p_instances, const RID &p_albedo_texture, const RID &p_emission_texture, const RID &p_emission_aniso_texture, const RID &p_geom_facing_texture) {
RENDER_TIMESTAMP("Render SDFGI");
_update_render_base_uniform_set();
@@ -2209,7 +2216,7 @@ void RendererSceneRenderForward::_render_sdfgi(RID p_render_buffers, const Vecto
render_list.clear();
PassMode pass_mode = PASS_MODE_SDF;
- _fill_render_list(p_cull_result, p_cull_count, pass_mode);
+ _fill_render_list(p_instances, pass_mode, CameraMatrix(), Transform());
render_list.sort_by_key(false);
_fill_instances(render_list.elements, render_list.element_count, true);
@@ -2453,7 +2460,7 @@ void RendererSceneRenderForward::_update_render_base_uniform_set() {
}
}
-RID RendererSceneRenderForward::_setup_render_pass_uniform_set(RID p_render_buffers, RID p_radiance_texture, RID p_shadow_atlas, RID p_reflection_atlas, RID *p_gi_probe_cull_result, int p_gi_probe_cull_count) {
+RID RendererSceneRenderForward::_setup_render_pass_uniform_set(RID p_render_buffers, RID p_radiance_texture, RID p_shadow_atlas, RID p_reflection_atlas, const PagedArray<RID> &p_gi_probes) {
if (render_pass_uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(render_pass_uniform_set)) {
RD::get_singleton()->free(render_pass_uniform_set);
}
@@ -2516,8 +2523,8 @@ RID RendererSceneRenderForward::_setup_render_pass_uniform_set(RID p_render_buff
u.ids.resize(MAX_GI_PROBES);
RID default_tex = storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_3D_WHITE);
for (int i = 0; i < MAX_GI_PROBES; i++) {
- if (i < p_gi_probe_cull_count) {
- RID tex = gi_probe_instance_get_texture(p_gi_probe_cull_result[i]);
+ if (i < (int)p_gi_probes.size()) {
+ RID tex = gi_probe_instance_get_texture(p_gi_probes[i]);
if (!tex.is_valid()) {
tex = default_tex;
}
diff --git a/servers/rendering/renderer_rd/renderer_scene_render_forward.h b/servers/rendering/renderer_rd/renderer_scene_render_forward.h
index 94284e509c..eca83893c3 100644
--- a/servers/rendering/renderer_rd/renderer_scene_render_forward.h
+++ b/servers/rendering/renderer_rd/renderer_scene_render_forward.h
@@ -266,7 +266,7 @@ class RendererSceneRenderForward : public RendererSceneRenderRD {
void _update_render_base_uniform_set();
RID _setup_sdfgi_render_pass_uniform_set(RID p_albedo_texture, RID p_emission_texture, RID p_emission_aniso_texture, RID p_geom_facing_texture);
- RID _setup_render_pass_uniform_set(RID p_render_buffers, RID p_radiance_texture, RID p_shadow_atlas, RID p_reflection_atlas, RID *p_gi_probe_cull_result, int p_gi_probe_cull_count);
+ RID _setup_render_pass_uniform_set(RID p_render_buffers, RID p_radiance_texture, RID p_shadow_atlas, RID p_reflection_atlas, const PagedArray<RID> &p_gi_probes);
struct LightmapData {
float normal_xform[12];
@@ -567,26 +567,26 @@ class RendererSceneRenderForward : public RendererSceneRenderRD {
};
void _setup_environment(RID p_environment, RID p_render_buffers, const CameraMatrix &p_cam_projection, const Transform &p_cam_transform, RID p_reflection_probe, bool p_no_fog, const Size2 &p_screen_pixel_size, RID p_shadow_atlas, bool p_flip_y, const Color &p_default_bg_color, float p_znear, float p_zfar, bool p_opaque_render_buffers = false, bool p_pancake_shadows = false);
- void _setup_lightmaps(InstanceBase **p_lightmap_cull_result, int p_lightmap_cull_count, const Transform &p_cam_transform);
+ void _setup_lightmaps(const PagedArray<InstanceBase *> &p_lightmaps, const Transform &p_cam_transform);
void _fill_instances(RenderList::Element **p_elements, int p_element_count, bool p_for_depth, bool p_has_sdfgi = false, bool p_has_opaque_gi = false);
void _render_list(RenderingDevice::DrawListID p_draw_list, RenderingDevice::FramebufferFormatID p_framebuffer_Format, RenderList::Element **p_elements, int p_element_count, bool p_reverse_cull, PassMode p_pass_mode, bool p_no_gi, RID p_render_pass_uniform_set, bool p_force_wireframe = false, const Vector2 &p_uv_offset = Vector2(), const Plane &p_lod_plane = Plane(), float p_lod_distance_multiplier = 0.0, float p_screen_lod_threshold = 0.0);
_FORCE_INLINE_ void _add_geometry(InstanceBase *p_instance, uint32_t p_surface, RID p_material, PassMode p_pass_mode, uint32_t p_geometry_index, bool p_using_sdfgi = false);
_FORCE_INLINE_ void _add_geometry_with_material(InstanceBase *p_instance, uint32_t p_surface, MaterialData *p_material, RID p_material_rid, PassMode p_pass_mode, uint32_t p_geometry_index, bool p_using_sdfgi = false);
- void _fill_render_list(InstanceBase **p_cull_result, int p_cull_count, PassMode p_pass_mode, bool p_using_sdfgi = false);
+ void _fill_render_list(const PagedArray<InstanceBase *> &p_instances, PassMode p_pass_mode, const CameraMatrix &p_cam_projection, const Transform &p_cam_transform, bool p_using_sdfgi = false);
Map<Size2i, RID> sdfgi_framebuffer_size_cache;
bool low_end = false;
protected:
- virtual void _render_scene(RID p_render_buffer, const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, InstanceBase **p_cull_result, int p_cull_count, int p_directional_light_count, RID *p_gi_probe_cull_result, int p_gi_probe_cull_count, InstanceBase **p_lightmap_cull_result, int p_lightmap_cull_count, RID p_environment, RID p_camera_effects, RID p_shadow_atlas, RID p_reflection_atlas, RID p_reflection_probe, int p_reflection_probe_pass, const Color &p_default_bg_color, float p_lod_threshold);
- virtual void _render_shadow(RID p_framebuffer, InstanceBase **p_cull_result, int p_cull_count, const CameraMatrix &p_projection, const Transform &p_transform, float p_zfar, float p_bias, float p_normal_bias, bool p_use_dp, bool p_use_dp_flip, bool p_use_pancake, const Plane &p_camera_plane = Plane(), float p_lod_distance_multiplier = 0, float p_screen_lod_threshold = 0.0);
- virtual void _render_material(const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, InstanceBase **p_cull_result, int p_cull_count, RID p_framebuffer, const Rect2i &p_region);
- virtual void _render_uv2(InstanceBase **p_cull_result, int p_cull_count, RID p_framebuffer, const Rect2i &p_region);
- virtual void _render_sdfgi(RID p_render_buffers, const Vector3i &p_from, const Vector3i &p_size, const AABB &p_bounds, InstanceBase **p_cull_result, int p_cull_count, const RID &p_albedo_texture, const RID &p_emission_texture, const RID &p_emission_aniso_texture, const RID &p_geom_facing_texture);
- virtual void _render_particle_collider_heightfield(RID p_fb, const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, InstanceBase **p_cull_result, int p_cull_count);
+ virtual void _render_scene(RID p_render_buffer, const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, const PagedArray<InstanceBase *> &p_instances, int p_directional_light_count, const PagedArray<RID> &p_gi_probes, const PagedArray<InstanceBase *> &p_lightmaps, RID p_environment, RID p_camera_effects, RID p_shadow_atlas, RID p_reflection_atlas, RID p_reflection_probe, int p_reflection_probe_pass, const Color &p_default_bg_color, float p_lod_threshold);
+ virtual void _render_shadow(RID p_framebuffer, const PagedArray<InstanceBase *> &p_instances, const CameraMatrix &p_projection, const Transform &p_transform, float p_zfar, float p_bias, float p_normal_bias, bool p_use_dp, bool p_use_dp_flip, bool p_use_pancake, const Plane &p_camera_plane = Plane(), float p_lod_distance_multiplier = 0.0, float p_screen_lod_threshold = 0.0);
+ virtual void _render_material(const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, const PagedArray<InstanceBase *> &p_instances, RID p_framebuffer, const Rect2i &p_region);
+ virtual void _render_uv2(const PagedArray<InstanceBase *> &p_instances, RID p_framebuffer, const Rect2i &p_region);
+ virtual void _render_sdfgi(RID p_render_buffers, const Vector3i &p_from, const Vector3i &p_size, const AABB &p_bounds, const PagedArray<InstanceBase *> &p_instances, const RID &p_albedo_texture, const RID &p_emission_texture, const RID &p_emission_aniso_texture, const RID &p_geom_facing_texture);
+ virtual void _render_particle_collider_heightfield(RID p_fb, const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, const PagedArray<InstanceBase *> &p_instances);
public:
virtual void set_time(double p_time, double p_step);
diff --git a/servers/rendering/renderer_rd/renderer_scene_render_rd.cpp b/servers/rendering/renderer_rd/renderer_scene_render_rd.cpp
index 4543ced8dc..dd4762fec8 100644
--- a/servers/rendering/renderer_rd/renderer_scene_render_rd.cpp
+++ b/servers/rendering/renderer_rd/renderer_scene_render_rd.cpp
@@ -1153,7 +1153,7 @@ void RendererSceneRenderRD::_sdfgi_update_cascades(RID p_render_buffers) {
RD::get_singleton()->buffer_update(rb->sdfgi->cascades_ubo, 0, sizeof(SDFGI::Cascade::UBO) * SDFGI::MAX_CASCADES, cascade_data, true);
}
-void RendererSceneRenderRD::sdfgi_update_probes(RID p_render_buffers, RID p_environment, const RID *p_directional_light_instances, uint32_t p_directional_light_count, const RID *p_positional_light_instances, uint32_t p_positional_light_count) {
+void RendererSceneRenderRD::sdfgi_update_probes(RID p_render_buffers, RID p_environment, const Vector<RID> &p_directional_lights, const RID *p_positional_light_instances, uint32_t p_positional_light_count) {
RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers);
ERR_FAIL_COND(rb == nullptr);
if (rb->sdfgi == nullptr) {
@@ -1179,12 +1179,12 @@ void RendererSceneRenderRD::sdfgi_update_probes(RID p_render_buffers, RID p_envi
SDGIShader::Light lights[SDFGI::MAX_DYNAMIC_LIGHTS];
uint32_t idx = 0;
- for (uint32_t j = 0; j < p_directional_light_count; j++) {
+ for (uint32_t j = 0; j < (uint32_t)p_directional_lights.size(); j++) {
if (idx == SDFGI::MAX_DYNAMIC_LIGHTS) {
break;
}
- LightInstance *li = light_instance_owner.getornull(p_directional_light_instances[j]);
+ LightInstance *li = light_instance_owner.getornull(p_directional_lights[j]);
ERR_CONTINUE(!li);
if (storage->light_directional_is_sky_only(li->light)) {
@@ -1402,7 +1402,7 @@ void RendererSceneRenderRD::sdfgi_update_probes(RID p_render_buffers, RID p_envi
RENDER_TIMESTAMP("<SDFGI Update Probes");
}
-void RendererSceneRenderRD::_setup_giprobes(RID p_render_buffers, const Transform &p_transform, RID *p_gi_probe_cull_result, int p_gi_probe_cull_count, uint32_t &r_gi_probes_used) {
+void RendererSceneRenderRD::_setup_giprobes(RID p_render_buffers, const Transform &p_transform, const PagedArray<RID> &p_gi_probes, uint32_t &r_gi_probes_used) {
r_gi_probes_used = 0;
RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers);
ERR_FAIL_COND(rb == nullptr);
@@ -1417,8 +1417,8 @@ void RendererSceneRenderRD::_setup_giprobes(RID p_render_buffers, const Transfor
for (int i = 0; i < RenderBuffers::MAX_GIPROBES; i++) {
RID texture;
- if (i < p_gi_probe_cull_count) {
- GIProbeInstance *gipi = gi_probe_instance_owner.getornull(p_gi_probe_cull_result[i]);
+ if (i < (int)p_gi_probes.size()) {
+ GIProbeInstance *gipi = gi_probe_instance_owner.getornull(p_gi_probes[i]);
if (gipi) {
texture = gipi->texture;
@@ -1489,12 +1489,12 @@ void RendererSceneRenderRD::_setup_giprobes(RID p_render_buffers, const Transfor
}
}
- if (p_gi_probe_cull_count > 0) {
- RD::get_singleton()->buffer_update(gi_probe_buffer, 0, sizeof(GI::GIProbeData) * MIN(RenderBuffers::MAX_GIPROBES, p_gi_probe_cull_count), gi_probe_data, true);
+ if (p_gi_probes.size() > 0) {
+ RD::get_singleton()->buffer_update(gi_probe_buffer, 0, sizeof(GI::GIProbeData) * MIN((uint64_t)RenderBuffers::MAX_GIPROBES, p_gi_probes.size()), gi_probe_data, true);
}
}
-void RendererSceneRenderRD::_process_gi(RID p_render_buffers, RID p_normal_roughness_buffer, RID p_ambient_buffer, RID p_reflection_buffer, RID p_gi_probe_buffer, RID p_environment, const CameraMatrix &p_projection, const Transform &p_transform, RID *p_gi_probe_cull_result, int p_gi_probe_cull_count) {
+void RendererSceneRenderRD::_process_gi(RID p_render_buffers, RID p_normal_roughness_buffer, RID p_ambient_buffer, RID p_reflection_buffer, RID p_gi_probe_buffer, RID p_environment, const CameraMatrix &p_projection, const Transform &p_transform, const PagedArray<RID> &p_gi_probes) {
RENDER_TIMESTAMP("Render GI");
RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers);
@@ -1512,7 +1512,7 @@ void RendererSceneRenderRD::_process_gi(RID p_render_buffers, RID p_normal_rough
push_constant.proj_info[1] = -2.0f / (rb->height * p_projection.matrix[1][1]);
push_constant.proj_info[2] = (1.0f - p_projection.matrix[0][2]) / p_projection.matrix[0][0];
push_constant.proj_info[3] = (1.0f + p_projection.matrix[1][2]) / p_projection.matrix[1][1];
- push_constant.max_giprobes = MIN(RenderBuffers::MAX_GIPROBES, p_gi_probe_cull_count);
+ push_constant.max_giprobes = MIN((uint64_t)RenderBuffers::MAX_GIPROBES, p_gi_probes.size());
push_constant.high_quality_vct = gi_probe_quality == RS::GI_PROBE_QUALITY_HIGH;
push_constant.use_sdfgi = rb->sdfgi != nullptr;
@@ -3122,7 +3122,7 @@ RS::EnvironmentSSRRoughnessQuality RendererSceneRenderRD::environment_get_ssr_ro
return ssr_roughness_quality;
}
-void RendererSceneRenderRD::environment_set_ssao(RID p_env, bool p_enable, float p_radius, float p_intensity, float p_bias, float p_light_affect, float p_ao_channel_affect, RS::EnvironmentSSAOBlur p_blur, float p_bilateral_sharpness) {
+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) {
Environment *env = environment_owner.getornull(p_env);
ERR_FAIL_COND(!env);
@@ -3133,15 +3133,21 @@ void RendererSceneRenderRD::environment_set_ssao(RID p_env, bool p_enable, float
env->ssao_enabled = p_enable;
env->ssao_radius = p_radius;
env->ssao_intensity = p_intensity;
- env->ssao_bias = p_bias;
+ env->ssao_power = p_power;
+ env->ssao_detail = p_detail;
+ env->ssao_horizon = p_horizon;
+ env->ssao_sharpness = p_sharpness;
env->ssao_direct_light_affect = p_light_affect;
env->ssao_ao_channel_affect = p_ao_channel_affect;
- env->ssao_blur = p_blur;
}
-void RendererSceneRenderRD::environment_set_ssao_quality(RS::EnvironmentSSAOQuality p_quality, bool p_half_size) {
+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) {
ssao_quality = p_quality;
ssao_half_size = p_half_size;
+ ssao_adaptive_target = p_adaptive_target;
+ ssao_blur_passes = p_blur_passes;
+ ssao_fadeout_from = p_fadeout_from;
+ ssao_fadeout_to = p_fadeout_to;
}
bool RendererSceneRenderRD::environment_is_ssao_enabled(RID p_env) const {
@@ -4055,7 +4061,7 @@ bool RendererSceneRenderRD::gi_probe_needs_update(RID p_probe) const {
return gi_probe->last_probe_version != storage->gi_probe_get_version(gi_probe->probe);
}
-void RendererSceneRenderRD::gi_probe_update(RID p_probe, bool p_update_light_instances, const Vector<RID> &p_light_instances, int p_dynamic_object_count, InstanceBase **p_dynamic_objects) {
+void RendererSceneRenderRD::gi_probe_update(RID p_probe, bool p_update_light_instances, const Vector<RID> &p_light_instances, const PagedArray<InstanceBase *> &p_dynamic_objects) {
GIProbeInstance *gi_probe = gi_probe_instance_owner.getornull(p_probe);
ERR_FAIL_COND(!gi_probe);
@@ -4414,7 +4420,7 @@ void RendererSceneRenderRD::gi_probe_update(RID p_probe, bool p_update_light_ins
uint32_t light_count = 0;
- if (p_update_light_instances || p_dynamic_object_count > 0) {
+ if (p_update_light_instances || p_dynamic_objects.size() > 0) {
light_count = MIN(gi_probe_max_lights, (uint32_t)p_light_instances.size());
{
@@ -4464,7 +4470,7 @@ void RendererSceneRenderRD::gi_probe_update(RID p_probe, bool p_update_light_ins
}
}
- if (gi_probe->has_dynamic_object_data || p_update_light_instances || p_dynamic_object_count) {
+ if (gi_probe->has_dynamic_object_data || p_update_light_instances || p_dynamic_objects.size()) {
// PROCESS MIPMAPS
if (gi_probe->mipmaps.size()) {
//can update mipmaps
@@ -4557,7 +4563,7 @@ void RendererSceneRenderRD::gi_probe_update(RID p_probe, bool p_update_light_ins
gi_probe->has_dynamic_object_data = false; //clear until dynamic object data is used again
- if (p_dynamic_object_count && gi_probe->dynamic_maps.size()) {
+ if (p_dynamic_objects.size() && gi_probe->dynamic_maps.size()) {
Vector3i octree_size = storage->gi_probe_get_octree_size(gi_probe->probe);
int multiplier = gi_probe->dynamic_maps[0].size / MAX(MAX(octree_size.x, octree_size.y), octree_size.z);
@@ -4571,7 +4577,7 @@ void RendererSceneRenderRD::gi_probe_update(RID p_probe, bool p_update_light_ins
AABB probe_aabb(Vector3(), octree_size);
//this could probably be better parallelized in compute..
- for (int i = 0; i < p_dynamic_object_count; i++) {
+ for (int i = 0; i < (int)p_dynamic_objects.size(); i++) {
InstanceBase *instance = p_dynamic_objects[i];
//not used, so clear
instance->depth_layer = 0;
@@ -4642,7 +4648,12 @@ void RendererSceneRenderRD::gi_probe_update(RID p_probe, bool p_update_light_ins
CameraMatrix cm;
cm.set_orthogonal(-rect.size.width / 2, rect.size.width / 2, -rect.size.height / 2, rect.size.height / 2, 0.0001, aabb.size[z_axis]);
- _render_material(to_world_xform * xform, cm, true, &instance, 1, gi_probe->dynamic_maps[0].fb, Rect2i(Vector2i(), rect.size));
+ if (cull_argument.size() == 0) {
+ cull_argument.push_back(nullptr);
+ }
+ cull_argument[0] = instance;
+
+ _render_material(to_world_xform * xform, cm, true, cull_argument, gi_probe->dynamic_maps[0].fb, Rect2i(Vector2i(), rect.size));
GIProbeDynamicPushConstant push_constant;
zeromem(&push_constant, sizeof(GIProbeDynamicPushConstant));
@@ -5081,21 +5092,24 @@ void RendererSceneRenderRD::_free_render_buffer_data(RenderBuffers *rb) {
rb->luminance.current = RID();
}
- if (rb->ssao.ao[0].is_valid()) {
+ if (rb->ssao.depth.is_valid()) {
RD::get_singleton()->free(rb->ssao.depth);
- RD::get_singleton()->free(rb->ssao.ao[0]);
- if (rb->ssao.ao[1].is_valid()) {
- RD::get_singleton()->free(rb->ssao.ao[1]);
- }
- if (rb->ssao.ao_full.is_valid()) {
- RD::get_singleton()->free(rb->ssao.ao_full);
- }
+ 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]);
rb->ssao.depth = RID();
- rb->ssao.ao[0] = RID();
- rb->ssao.ao[1] = RID();
- rb->ssao.ao_full = RID();
+ 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();
rb->ssao.depth_slices.clear();
+ rb->ssao.ao_deinterleaved_slices.clear();
+ rb->ssao.ao_pong_slices.clear();
}
if (rb->ssr.blur_radius[0].is_valid()) {
@@ -5194,64 +5208,132 @@ void RendererSceneRenderRD::_process_ssao(RID p_render_buffers, RID p_environmen
RENDER_TIMESTAMP("Process SSAO");
- if (rb->ssao.ao[0].is_valid() && rb->ssao.ao_full.is_valid() != ssao_half_size) {
+ if (rb->ssao.ao_final.is_valid() && ssao_using_half_size != ssao_half_size) {
RD::get_singleton()->free(rb->ssao.depth);
- RD::get_singleton()->free(rb->ssao.ao[0]);
- if (rb->ssao.ao[1].is_valid()) {
- RD::get_singleton()->free(rb->ssao.ao[1]);
- }
- if (rb->ssao.ao_full.is_valid()) {
- RD::get_singleton()->free(rb->ssao.ao_full);
- }
+ 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]);
rb->ssao.depth = RID();
- rb->ssao.ao[0] = RID();
- rb->ssao.ao[1] = RID();
- rb->ssao.ao_full = RID();
+ 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();
rb->ssao.depth_slices.clear();
+ 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;
}
-
- if (!rb->ssao.ao[0].is_valid()) {
+ bool uniform_sets_are_invalid = false;
+ if (rb->ssao.depth.is_null()) {
//allocate depth slices
{
RD::TextureFormat tf;
- tf.format = RD::DATA_FORMAT_R32_SFLOAT;
- tf.width = rb->width / 2;
- tf.height = rb->height / 2;
- tf.mipmaps = Image::get_image_required_mipmaps(tf.width, tf.height, Image::FORMAT_RF) + 1;
+ 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;
tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_STORAGE_BIT;
rb->ssao.depth = RD::get_singleton()->texture_create(tf, RD::TextureView());
for (uint32_t i = 0; i < tf.mipmaps; i++) {
- RID slice = RD::get_singleton()->texture_create_shared_from_slice(RD::TextureView(), rb->ssao.depth, 0, i);
+ RID slice = RD::get_singleton()->texture_create_shared_from_slice(RD::TextureView(), rb->ssao.depth, 0, i, RD::TEXTURE_SLICE_2D_ARRAY);
rb->ssao.depth_slices.push_back(slice);
}
}
{
RD::TextureFormat tf;
- tf.format = RD::DATA_FORMAT_R8_UNORM;
- tf.width = ssao_half_size ? rb->width / 2 : rb->width;
- tf.height = ssao_half_size ? rb->height / 2 : rb->height;
+ 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[0] = RD::get_singleton()->texture_create(tf, RD::TextureView());
- rb->ssao.ao[1] = RD::get_singleton()->texture_create(tf, RD::TextureView());
+ rb->ssao.ao_deinterleaved = RD::get_singleton()->texture_create(tf, RD::TextureView());
+ 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);
+ }
}
- if (ssao_half_size) {
- //upsample texture
+ {
+ 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;
+ tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_STORAGE_BIT;
+ rb->ssao.ao_pong = RD::get_singleton()->texture_create(tf, RD::TextureView());
+ 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);
+ }
+ }
+
+ {
+ 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());
+ rb->ssao.importance_map[1] = RD::get_singleton()->texture_create(tf, RD::TextureView());
+ }
+ {
RD::TextureFormat tf;
tf.format = RD::DATA_FORMAT_R8_UNORM;
tf.width = rb->width;
tf.height = rb->height;
tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_STORAGE_BIT;
- rb->ssao.ao_full = RD::get_singleton()->texture_create(tf, RD::TextureView());
+ rb->ssao.ao_final = RD::get_singleton()->texture_create(tf, RD::TextureView());
+ _render_buffers_uniform_set_changed(p_render_buffers);
}
-
- _render_buffers_uniform_set_changed(p_render_buffers);
+ ssao_using_half_size = ssao_half_size;
+ uniform_sets_are_invalid = true;
}
- storage->get_effects()->generate_ssao(rb->depth_texture, p_normal_buffer, Size2i(rb->width, rb->height), rb->ssao.depth, rb->ssao.depth_slices, rb->ssao.ao[0], rb->ssao.ao_full.is_valid(), rb->ssao.ao[1], rb->ssao.ao_full, env->ssao_intensity, env->ssao_radius, env->ssao_bias, p_projection, ssao_quality, env->ssao_blur, env->ssao_blur_edge_sharpness);
+ 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;
+ settings.screen_size = Size2i(rb->width, rb->height);
+ settings.half_screen_size = Size2i(buffer_width, buffer_height);
+ settings.quarter_size = Size2i(half_width, half_height);
+
+ 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);
}
void RendererSceneRenderRD::_render_buffers_post_process_and_tonemap(RID p_render_buffers, RID p_environment, RID p_camera_effects, const CameraMatrix &p_projection) {
@@ -5431,9 +5513,9 @@ void RendererSceneRenderRD::_render_buffers_debug_draw(RID p_render_buffers, RID
}
}
- if (debug_draw == RS::VIEWPORT_DEBUG_DRAW_SSAO && rb->ssao.ao[0].is_valid()) {
+ 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_full.is_valid() ? rb->ssao.ao_full : rb->ssao.ao[0];
+ 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);
}
@@ -5621,7 +5703,7 @@ RID RendererSceneRenderRD::render_buffers_get_ao_texture(RID p_render_buffers) {
RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers);
ERR_FAIL_COND_V(!rb, RID());
- return rb->ssao.ao_full.is_valid() ? rb->ssao.ao_full : rb->ssao.ao[0];
+ return rb->ssao.ao_final;
}
RID RendererSceneRenderRD::render_buffers_get_gi_probe_buffer(RID p_render_buffers) {
@@ -5923,11 +6005,11 @@ RendererSceneRenderRD::RenderBufferData *RendererSceneRenderRD::render_buffers_g
return rb->data;
}
-void RendererSceneRenderRD::_setup_reflections(RID *p_reflection_probe_cull_result, int p_reflection_probe_cull_count, const Transform &p_camera_inverse_transform, RID p_environment) {
- for (int i = 0; i < p_reflection_probe_cull_count; i++) {
- RID rpi = p_reflection_probe_cull_result[i];
+void RendererSceneRenderRD::_setup_reflections(const PagedArray<RID> &p_reflections, const Transform &p_camera_inverse_transform, RID p_environment) {
+ for (uint32_t i = 0; i < (uint32_t)p_reflections.size(); i++) {
+ RID rpi = p_reflections[i];
- if (i >= (int)cluster.max_reflections) {
+ if (i >= cluster.max_reflections) {
reflection_probe_instance_set_render_index(rpi, 0); //invalid, but something needs to be set
continue;
}
@@ -5978,19 +6060,19 @@ void RendererSceneRenderRD::_setup_reflections(RID *p_reflection_probe_cull_resu
reflection_probe_instance_set_render_pass(rpi, RSG::rasterizer->get_frame_number());
}
- if (p_reflection_probe_cull_count) {
- RD::get_singleton()->buffer_update(cluster.reflection_buffer, 0, MIN(cluster.max_reflections, (unsigned int)p_reflection_probe_cull_count) * sizeof(ReflectionData), cluster.reflections, true);
+ if (p_reflections.size()) {
+ RD::get_singleton()->buffer_update(cluster.reflection_buffer, 0, MIN(cluster.max_reflections, (unsigned int)p_reflections.size()) * sizeof(ReflectionData), cluster.reflections, true);
}
}
-void RendererSceneRenderRD::_setup_lights(RID *p_light_cull_result, int p_light_cull_count, const Transform &p_camera_inverse_transform, RID p_shadow_atlas, bool p_using_shadows, uint32_t &r_directional_light_count, uint32_t &r_positional_light_count) {
+void RendererSceneRenderRD::_setup_lights(const PagedArray<RID> &p_lights, const Transform &p_camera_inverse_transform, RID p_shadow_atlas, bool p_using_shadows, uint32_t &r_directional_light_count, uint32_t &r_positional_light_count) {
uint32_t light_count = 0;
r_directional_light_count = 0;
r_positional_light_count = 0;
sky_scene_state.ubo.directional_light_count = 0;
- for (int i = 0; i < p_light_cull_count; i++) {
- RID li = p_light_cull_result[i];
+ for (int i = 0; i < (int)p_lights.size(); i++) {
+ RID li = p_lights[i];
RID base = light_instance_get_base_light(li);
ERR_CONTINUE(base.is_null());
@@ -6343,15 +6425,15 @@ void RendererSceneRenderRD::_setup_lights(RID *p_light_cull_result, int p_light_
}
}
-void RendererSceneRenderRD::_setup_decals(const RID *p_decal_instances, int p_decal_count, const Transform &p_camera_inverse_xform) {
+void RendererSceneRenderRD::_setup_decals(const PagedArray<RID> &p_decals, const Transform &p_camera_inverse_xform) {
Transform uv_xform;
uv_xform.basis.scale(Vector3(2.0, 1.0, 2.0));
uv_xform.origin = Vector3(-1.0, 0.0, -1.0);
- p_decal_count = MIN((uint32_t)p_decal_count, cluster.max_decals);
+ uint32_t decal_count = MIN((uint32_t)p_decals.size(), cluster.max_decals);
int idx = 0;
- for (int i = 0; i < p_decal_count; i++) {
- RID di = p_decal_instances[i];
+ for (uint32_t i = 0; i < decal_count; i++) {
+ RID di = p_decals[i];
RID decal = decal_instance_get_base(di);
Transform xform = decal_instance_get_transform(di);
@@ -6599,7 +6681,7 @@ void RendererSceneRenderRD::_update_volumetric_fog(RID p_render_buffers, RID p_e
//update directional shadow
if (p_use_directional_shadows) {
- if (directional_shadow.shrink_stages.empty()) {
+ if (directional_shadow.shrink_stages.is_empty()) {
if (rb->volumetric_fog->uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(rb->volumetric_fog->uniform_set)) {
//invalidate uniform set, we will need a new one
RD::get_singleton()->free(rb->volumetric_fog->uniform_set);
@@ -6634,7 +6716,7 @@ void RendererSceneRenderRD::_update_volumetric_fog(RID p_render_buffers, RID p_e
bool force_shrink_shadows = false;
- if (shadow_atlas->shrink_stages.empty()) {
+ if (shadow_atlas->shrink_stages.is_empty()) {
if (rb->volumetric_fog->uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(rb->volumetric_fog->uniform_set)) {
//invalidate uniform set, we will need a new one
RD::get_singleton()->free(rb->volumetric_fog->uniform_set);
@@ -7019,7 +7101,7 @@ void RendererSceneRenderRD::_update_volumetric_fog(RID p_render_buffers, RID p_e
RD::get_singleton()->compute_list_end();
}
-void RendererSceneRenderRD::render_scene(RID p_render_buffers, const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, InstanceBase **p_cull_result, int p_cull_count, RID *p_light_cull_result, int p_light_cull_count, RID *p_reflection_probe_cull_result, int p_reflection_probe_cull_count, RID *p_gi_probe_cull_result, int p_gi_probe_cull_count, RID *p_decal_cull_result, int p_decal_cull_count, InstanceBase **p_lightmap_cull_result, int p_lightmap_cull_count, RID p_environment, RID p_camera_effects, RID p_shadow_atlas, RID p_reflection_atlas, RID p_reflection_probe, int p_reflection_probe_pass, float p_screen_lod_threshold) {
+void RendererSceneRenderRD::render_scene(RID p_render_buffers, const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, const PagedArray<InstanceBase *> &p_instances, const PagedArray<RID> &p_lights, const PagedArray<RID> &p_reflection_probes, const PagedArray<RID> &p_gi_probes, const PagedArray<RID> &p_decals, const PagedArray<InstanceBase *> &p_lightmaps, RID p_environment, RID p_camera_effects, RID p_shadow_atlas, RID p_reflection_atlas, RID p_reflection_probe, int p_reflection_probe_pass, float p_screen_lod_threshold) {
Color clear_color;
if (p_render_buffers.is_valid()) {
RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers);
@@ -7030,17 +7112,23 @@ void RendererSceneRenderRD::render_scene(RID p_render_buffers, const Transform &
}
//assign render indices to giprobes
- for (int i = 0; i < p_gi_probe_cull_count; i++) {
- GIProbeInstance *giprobe_inst = gi_probe_instance_owner.getornull(p_gi_probe_cull_result[i]);
+ for (uint32_t i = 0; i < (uint32_t)p_gi_probes.size(); i++) {
+ GIProbeInstance *giprobe_inst = gi_probe_instance_owner.getornull(p_gi_probes[i]);
if (giprobe_inst) {
giprobe_inst->render_index = i;
}
}
+ const PagedArray<RID> *lights = &p_lights;
+ const PagedArray<RID> *reflections = &p_reflection_probes;
+ const PagedArray<RID> *gi_probes = &p_gi_probes;
+
+ PagedArray<RID> empty;
+
if (get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_UNSHADED) {
- p_light_cull_count = 0;
- p_reflection_probe_cull_count = 0;
- p_gi_probe_cull_count = 0;
+ lights = &empty;
+ reflections = &empty;
+ gi_probes = &empty;
}
cluster.builder.begin(p_cam_transform.affine_inverse(), p_cam_projection); //prepare cluster
@@ -7053,17 +7141,17 @@ void RendererSceneRenderRD::render_scene(RID p_render_buffers, const Transform &
}
} else {
//do not render reflections when rendering a reflection probe
- _setup_reflections(p_reflection_probe_cull_result, p_reflection_probe_cull_count, p_cam_transform.affine_inverse(), p_environment);
+ _setup_reflections(*reflections, p_cam_transform.affine_inverse(), p_environment);
}
uint32_t directional_light_count = 0;
uint32_t positional_light_count = 0;
- _setup_lights(p_light_cull_result, p_light_cull_count, p_cam_transform.affine_inverse(), p_shadow_atlas, using_shadows, directional_light_count, positional_light_count);
- _setup_decals(p_decal_cull_result, p_decal_cull_count, p_cam_transform.affine_inverse());
+ _setup_lights(*lights, p_cam_transform.affine_inverse(), p_shadow_atlas, using_shadows, directional_light_count, positional_light_count);
+ _setup_decals(p_decals, p_cam_transform.affine_inverse());
cluster.builder.bake_cluster(); //bake to cluster
uint32_t gi_probe_count = 0;
- _setup_giprobes(p_render_buffers, p_cam_transform, p_gi_probe_cull_result, p_gi_probe_cull_count, gi_probe_count);
+ _setup_giprobes(p_render_buffers, p_cam_transform, *gi_probes, gi_probe_count);
if (p_render_buffers.is_valid()) {
bool directional_shadows = false;
@@ -7076,7 +7164,7 @@ void RendererSceneRenderRD::render_scene(RID p_render_buffers, const Transform &
_update_volumetric_fog(p_render_buffers, p_environment, p_cam_projection, p_cam_transform, p_shadow_atlas, directional_light_count, directional_shadows, positional_light_count, gi_probe_count);
}
- _render_scene(p_render_buffers, p_cam_transform, p_cam_projection, p_cam_ortogonal, p_cull_result, p_cull_count, directional_light_count, p_gi_probe_cull_result, p_gi_probe_cull_count, p_lightmap_cull_result, p_lightmap_cull_count, p_environment, p_camera_effects, p_shadow_atlas, p_reflection_atlas, p_reflection_probe, p_reflection_probe_pass, clear_color, p_screen_lod_threshold);
+ _render_scene(p_render_buffers, p_cam_transform, p_cam_projection, p_cam_ortogonal, p_instances, directional_light_count, *gi_probes, p_lightmaps, p_environment, p_camera_effects, p_shadow_atlas, p_reflection_atlas, p_reflection_probe, p_reflection_probe_pass, clear_color, p_screen_lod_threshold);
if (p_render_buffers.is_valid()) {
RENDER_TIMESTAMP("Tonemap");
@@ -7089,7 +7177,7 @@ void RendererSceneRenderRD::render_scene(RID p_render_buffers, const Transform &
}
}
-void RendererSceneRenderRD::render_shadow(RID p_light, RID p_shadow_atlas, int p_pass, InstanceBase **p_cull_result, int p_cull_count, const Plane &p_camera_plane, float p_lod_distance_multiplier, float p_screen_lod_threshold) {
+void RendererSceneRenderRD::render_shadow(RID p_light, RID p_shadow_atlas, int p_pass, const PagedArray<InstanceBase *> &p_instances, const Plane &p_camera_plane, float p_lod_distance_multiplier, float p_screen_lod_threshold) {
LightInstance *light_instance = light_instance_owner.getornull(p_light);
ERR_FAIL_COND(!light_instance);
@@ -7240,7 +7328,7 @@ void RendererSceneRenderRD::render_shadow(RID p_light, RID p_shadow_atlas, int p
if (render_cubemap) {
//rendering to cubemap
- _render_shadow(render_fb, p_cull_result, p_cull_count, light_projection, light_transform, zfar, 0, 0, false, false, use_pancake, p_camera_plane, p_lod_distance_multiplier, p_screen_lod_threshold);
+ _render_shadow(render_fb, p_instances, light_projection, light_transform, zfar, 0, 0, false, false, use_pancake, p_camera_plane, p_lod_distance_multiplier, p_screen_lod_threshold);
if (finalize_cubemap) {
//reblit
atlas_rect.size.height /= 2;
@@ -7251,7 +7339,7 @@ void RendererSceneRenderRD::render_shadow(RID p_light, RID p_shadow_atlas, int p
} else {
//render shadow
- _render_shadow(render_fb, p_cull_result, p_cull_count, light_projection, light_transform, zfar, bias, normal_bias, using_dual_paraboloid, using_dual_paraboloid_flip, use_pancake, p_camera_plane, p_lod_distance_multiplier, p_screen_lod_threshold);
+ _render_shadow(render_fb, p_instances, light_projection, light_transform, zfar, bias, normal_bias, using_dual_paraboloid, using_dual_paraboloid_flip, use_pancake, p_camera_plane, p_lod_distance_multiplier, p_screen_lod_threshold);
//copy to atlas
if (use_linear_depth) {
@@ -7265,11 +7353,11 @@ void RendererSceneRenderRD::render_shadow(RID p_light, RID p_shadow_atlas, int p
}
}
-void RendererSceneRenderRD::render_material(const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, InstanceBase **p_cull_result, int p_cull_count, RID p_framebuffer, const Rect2i &p_region) {
- _render_material(p_cam_transform, p_cam_projection, p_cam_ortogonal, p_cull_result, p_cull_count, p_framebuffer, p_region);
+void RendererSceneRenderRD::render_material(const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, const PagedArray<InstanceBase *> &p_instances, RID p_framebuffer, const Rect2i &p_region) {
+ _render_material(p_cam_transform, p_cam_projection, p_cam_ortogonal, p_instances, p_framebuffer, p_region);
}
-void RendererSceneRenderRD::render_sdfgi(RID p_render_buffers, int p_region, InstanceBase **p_cull_result, int p_cull_count) {
+void RendererSceneRenderRD::render_sdfgi(RID p_render_buffers, int p_region, const PagedArray<InstanceBase *> &p_instances) {
//print_line("rendering region " + itos(p_region));
RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers);
ERR_FAIL_COND(!rb);
@@ -7292,7 +7380,7 @@ void RendererSceneRenderRD::render_sdfgi(RID p_render_buffers, int p_region, Ins
}
//print_line("rendering cascade " + itos(p_region) + " objects: " + itos(p_cull_count) + " bounds: " + bounds + " from: " + from + " size: " + size + " cell size: " + rtos(rb->sdfgi->cascades[cascade].cell_size));
- _render_sdfgi(p_render_buffers, from, size, bounds, p_cull_result, p_cull_count, rb->sdfgi->render_albedo, rb->sdfgi->render_emission, rb->sdfgi->render_emission_aniso, rb->sdfgi->render_geom_facing);
+ _render_sdfgi(p_render_buffers, from, size, bounds, p_instances, rb->sdfgi->render_albedo, rb->sdfgi->render_emission, rb->sdfgi->render_emission_aniso, rb->sdfgi->render_geom_facing);
if (cascade_next != cascade) {
RENDER_TIMESTAMP(">SDFGI Update SDF");
@@ -7606,7 +7694,7 @@ void RendererSceneRenderRD::render_sdfgi(RID p_render_buffers, int p_region, Ins
}
}
-void RendererSceneRenderRD::render_particle_collider_heightfield(RID p_collider, const Transform &p_transform, InstanceBase **p_cull_result, int p_cull_count) {
+void RendererSceneRenderRD::render_particle_collider_heightfield(RID p_collider, const Transform &p_transform, const PagedArray<InstanceBase *> &p_instances) {
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;
@@ -7620,16 +7708,14 @@ void RendererSceneRenderRD::render_particle_collider_heightfield(RID p_collider,
RID fb = storage->particles_collision_get_heightfield_framebuffer(p_collider);
- _render_particle_collider_heightfield(fb, cam_xform, cm, p_cull_result, p_cull_count);
+ _render_particle_collider_heightfield(fb, cam_xform, cm, p_instances);
}
-void RendererSceneRenderRD::render_sdfgi_static_lights(RID p_render_buffers, uint32_t p_cascade_count, const uint32_t *p_cascade_indices, const RID **p_positional_light_cull_result, const uint32_t *p_positional_light_cull_count) {
+void RendererSceneRenderRD::render_sdfgi_static_lights(RID p_render_buffers, uint32_t p_cascade_count, const uint32_t *p_cascade_indices, const PagedArray<RID> *p_positional_light_cull_result) {
RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers);
ERR_FAIL_COND(!rb);
ERR_FAIL_COND(!rb->sdfgi);
- ERR_FAIL_COND(p_positional_light_cull_count == 0);
-
_sdfgi_update_cascades(p_render_buffers); //need cascades updated for this
RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin();
@@ -7665,7 +7751,7 @@ void RendererSceneRenderRD::render_sdfgi_static_lights(RID p_render_buffers, uin
int idx = 0;
- for (uint32_t j = 0; j < p_positional_light_cull_count[i]; j++) {
+ for (uint32_t j = 0; j < (uint32_t)p_positional_light_cull_result[i].size(); j++) {
if (idx == SDFGI::MAX_STATIC_LIGHTS) {
break;
}
@@ -7904,8 +7990,11 @@ TypedArray<Image> RendererSceneRenderRD::bake_render_uv2(RID p_base, const Vecto
}
}
- InstanceBase *cull = &ins;
- _render_uv2(&cull, 1, fb, Rect2i(0, 0, p_image_size.width, p_image_size.height));
+ if (cull_argument.size() == 0) {
+ cull_argument.push_back(nullptr);
+ }
+ cull_argument[0] = &ins;
+ _render_uv2(cull_argument, fb, Rect2i(0, 0, p_image_size.width, p_image_size.height));
TypedArray<Image> ret;
@@ -8395,7 +8484,7 @@ RendererSceneRenderRD::RendererSceneRenderRD(RendererStorageRD *p_storage) {
cluster.lights_instances = memnew_arr(RID, cluster.max_lights);
cluster.lights_shadow_rect_cache = memnew_arr(Rect2i, cluster.max_lights);
- cluster.max_directional_lights = 8;
+ cluster.max_directional_lights = MAX_DIRECTIONAL_LIGHTS;
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);
@@ -8435,7 +8524,7 @@ RendererSceneRenderRD::RendererSceneRenderRD(RendererStorageRD *p_storage) {
camera_effects_set_dof_blur_bokeh_shape(RS::DOFBokehShape(int(GLOBAL_GET("rendering/quality/depth_of_field/depth_of_field_bokeh_shape"))));
camera_effects_set_dof_blur_quality(RS::DOFBlurQuality(int(GLOBAL_GET("rendering/quality/depth_of_field/depth_of_field_bokeh_quality"))), GLOBAL_GET("rendering/quality/depth_of_field/depth_of_field_use_jitter"));
- environment_set_ssao_quality(RS::EnvironmentSSAOQuality(int(GLOBAL_GET("rendering/quality/ssao/quality"))), GLOBAL_GET("rendering/quality/ssao/half_size"));
+ environment_set_ssao_quality(RS::EnvironmentSSAOQuality(int(GLOBAL_GET("rendering/quality/ssao/quality"))), GLOBAL_GET("rendering/quality/ssao/half_size"), GLOBAL_GET("rendering/quality/ssao/adaptive_target"), GLOBAL_GET("rendering/quality/ssao/blur_passes"), GLOBAL_GET("rendering/quality/ssao/fadeout_from"), GLOBAL_GET("rendering/quality/ssao/fadeout_to"));
screen_space_roughness_limiter = GLOBAL_GET("rendering/quality/screen_filters/screen_space_roughness_limiter_enabled");
screen_space_roughness_limiter_amount = GLOBAL_GET("rendering/quality/screen_filters/screen_space_roughness_limiter_amount");
screen_space_roughness_limiter_limit = GLOBAL_GET("rendering/quality/screen_filters/screen_space_roughness_limiter_limit");
@@ -8456,6 +8545,8 @@ RendererSceneRenderRD::RendererSceneRenderRD(RendererStorageRD *p_storage) {
environment_set_volumetric_fog_filter_active(GLOBAL_GET("rendering/volumetric_fog/use_filter"));
environment_set_volumetric_fog_directional_shadow_shrink_size(GLOBAL_GET("rendering/volumetric_fog/directional_shadow_shrink"));
environment_set_volumetric_fog_positional_shadow_shrink_size(GLOBAL_GET("rendering/volumetric_fog/positional_shadow_shrink"));
+
+ cull_argument.set_page_pool(&cull_argument_pool);
}
RendererSceneRenderRD::~RendererSceneRenderRD() {
@@ -8520,4 +8611,5 @@ RendererSceneRenderRD::~RendererSceneRenderRD() {
RD::get_singleton()->free(shadow_sampler);
directional_shadow_atlas_set_size(0);
+ cull_argument.reset(); //avoid exit error
}
diff --git a/servers/rendering/renderer_rd/renderer_scene_render_rd.h b/servers/rendering/renderer_rd/renderer_scene_render_rd.h
index 3afe6e3f4a..e4dc98571e 100644
--- a/servers/rendering/renderer_rd/renderer_scene_render_rd.h
+++ b/servers/rendering/renderer_rd/renderer_scene_render_rd.h
@@ -104,17 +104,17 @@ protected:
};
virtual RenderBufferData *_create_render_buffer_data() = 0;
- void _setup_lights(RID *p_light_cull_result, int p_light_cull_count, const Transform &p_camera_inverse_transform, RID p_shadow_atlas, bool p_using_shadows, uint32_t &r_directional_light_count, uint32_t &r_positional_light_count);
- void _setup_decals(const RID *p_decal_instances, int p_decal_count, const Transform &p_camera_inverse_xform);
- void _setup_reflections(RID *p_reflection_probe_cull_result, int p_reflection_probe_cull_count, const Transform &p_camera_inverse_transform, RID p_environment);
- void _setup_giprobes(RID p_render_buffers, const Transform &p_transform, RID *p_gi_probe_cull_result, int p_gi_probe_cull_count, uint32_t &r_gi_probes_used);
-
- virtual void _render_scene(RID p_render_buffer, const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, InstanceBase **p_cull_result, int p_cull_count, int p_directional_light_count, RID *p_gi_probe_cull_result, int p_gi_probe_cull_count, InstanceBase **p_lightmap_cull_result, int p_lightmap_cull_count, RID p_environment, RID p_camera_effects, RID p_shadow_atlas, RID p_reflection_atlas, RID p_reflection_probe, int p_reflection_probe_pass, const Color &p_default_color, float p_screen_lod_threshold) = 0;
- virtual void _render_shadow(RID p_framebuffer, InstanceBase **p_cull_result, int p_cull_count, const CameraMatrix &p_projection, const Transform &p_transform, float p_zfar, float p_bias, float p_normal_bias, bool p_use_dp, bool use_dp_flip, bool p_use_pancake, const Plane &p_camera_plane = Plane(), float p_lod_distance_multiplier = 0.0, float p_screen_lod_threshold = 0.0) = 0;
- virtual void _render_material(const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, InstanceBase **p_cull_result, int p_cull_count, RID p_framebuffer, const Rect2i &p_region) = 0;
- virtual void _render_uv2(InstanceBase **p_cull_result, int p_cull_count, RID p_framebuffer, const Rect2i &p_region) = 0;
- virtual void _render_sdfgi(RID p_render_buffers, const Vector3i &p_from, const Vector3i &p_size, const AABB &p_bounds, InstanceBase **p_cull_result, int p_cull_count, const RID &p_albedo_texture, const RID &p_emission_texture, const RID &p_emission_aniso_texture, const RID &p_geom_facing_texture) = 0;
- virtual void _render_particle_collider_heightfield(RID p_fb, const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, InstanceBase **p_cull_result, int p_cull_count) = 0;
+ void _setup_lights(const PagedArray<RID> &p_lights, const Transform &p_camera_inverse_transform, RID p_shadow_atlas, bool p_using_shadows, uint32_t &r_directional_light_count, uint32_t &r_positional_light_count);
+ void _setup_decals(const PagedArray<RID> &p_decals, const Transform &p_camera_inverse_xform);
+ void _setup_reflections(const PagedArray<RID> &p_reflections, const Transform &p_camera_inverse_transform, RID p_environment);
+ void _setup_giprobes(RID p_render_buffers, const Transform &p_transform, const PagedArray<RID> &p_gi_probes, uint32_t &r_gi_probes_used);
+
+ virtual void _render_scene(RID p_render_buffer, const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, const PagedArray<InstanceBase *> &p_instances, int p_directional_light_count, const PagedArray<RID> &p_gi_probes, const PagedArray<InstanceBase *> &p_lightmaps, RID p_environment, RID p_camera_effects, RID p_shadow_atlas, RID p_reflection_atlas, RID p_reflection_probe, int p_reflection_probe_pass, const Color &p_default_color, float p_screen_lod_threshold) = 0;
+ virtual void _render_shadow(RID p_framebuffer, const PagedArray<InstanceBase *> &p_instances, const CameraMatrix &p_projection, const Transform &p_transform, float p_zfar, float p_bias, float p_normal_bias, bool p_use_dp, bool use_dp_flip, bool p_use_pancake, const Plane &p_camera_plane = Plane(), float p_lod_distance_multiplier = 0.0, float p_screen_lod_threshold = 0.0) = 0;
+ virtual void _render_material(const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, const PagedArray<InstanceBase *> &p_instances, RID p_framebuffer, const Rect2i &p_region) = 0;
+ virtual void _render_uv2(const PagedArray<InstanceBase *> &p_instances, RID p_framebuffer, const Rect2i &p_region) = 0;
+ virtual void _render_sdfgi(RID p_render_buffers, const Vector3i &p_from, const Vector3i &p_size, const AABB &p_bounds, const PagedArray<InstanceBase *> &p_instances, const RID &p_albedo_texture, const RID &p_emission_texture, const RID &p_emission_aniso_texture, const RID &p_geom_facing_texture) = 0;
+ virtual void _render_particle_collider_heightfield(RID p_fb, const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, const PagedArray<InstanceBase *> &p_instances) = 0;
virtual void _debug_giprobe(RID p_gi_probe, RenderingDevice::DrawListID p_draw_list, RID p_framebuffer, const CameraMatrix &p_camera_with_transform, bool p_lighting, bool p_emission, float p_alpha);
void _debug_sdfgi_probes(RID p_render_buffers, RD::DrawListID p_draw_list, RID p_framebuffer, const CameraMatrix &p_camera_with_transform);
@@ -134,8 +134,11 @@ protected:
void _setup_sky(RID p_environment, RID p_render_buffers, const CameraMatrix &p_projection, const Transform &p_transform, const Size2i p_screen_size);
void _update_sky(RID p_environment, const CameraMatrix &p_projection, const Transform &p_transform);
void _draw_sky(bool p_can_continue_color, bool p_can_continue_depth, RID p_fb, RID p_environment, const CameraMatrix &p_projection, const Transform &p_transform);
- void _process_gi(RID p_render_buffers, RID p_normal_roughness_buffer, RID p_ambient_buffer, RID p_reflection_buffer, RID p_gi_probe_buffer, RID p_environment, const CameraMatrix &p_projection, const Transform &p_transform, RID *p_gi_probe_cull_result, int p_gi_probe_cull_count);
+ void _process_gi(RID p_render_buffers, RID p_normal_roughness_buffer, RID p_ambient_buffer, RID p_reflection_buffer, RID p_gi_probe_buffer, RID p_environment, const CameraMatrix &p_projection, const Transform &p_transform, const PagedArray<RID> &p_gi_probes);
+ // needed for a single argument calls (material and uv2)
+ PagedArrayPool<InstanceBase *> cull_argument_pool;
+ PagedArray<InstanceBase *> cull_argument; //need this to exist
private:
RS::ViewportDebugDraw debug_draw = RS::VIEWPORT_DEBUG_DRAW_DISABLED;
double time_step = 0;
@@ -737,13 +740,14 @@ private:
/// SSAO
bool ssao_enabled = false;
- float ssao_radius = 1;
- float ssao_intensity = 1;
- float ssao_bias = 0.01;
+ float ssao_radius = 1.0;
+ float ssao_intensity = 2.0;
+ float ssao_power = 1.5;
+ float ssao_detail = 0.5;
+ float ssao_horizon = 0.06;
+ float ssao_sharpness = 0.98;
float ssao_direct_light_affect = 0.0;
float ssao_ao_channel_affect = 0.0;
- float ssao_blur_edge_sharpness = 4.0;
- RS::EnvironmentSSAOBlur ssao_blur = RS::ENV_SSAO_BLUR_3x3;
/// SSR
///
@@ -777,6 +781,12 @@ private:
RS::EnvironmentSSAOQuality ssao_quality = RS::ENV_SSAO_QUALITY_MEDIUM;
bool ssao_half_size = false;
+ bool ssao_using_half_size = false;
+ float ssao_adaptive_target = 0.5;
+ int ssao_blur_passes = 2;
+ float ssao_fadeout_from = 50.0;
+ float ssao_fadeout_to = 300.0;
+
bool glow_bicubic_upscale = false;
bool glow_high_quality = false;
RS::EnvironmentSSRRoughnessQuality ssr_roughness_quality = RS::ENV_SSR_ROUGNESS_QUALITY_LOW;
@@ -861,8 +871,12 @@ private:
struct SSAO {
RID depth;
Vector<RID> depth_slices;
- RID ao[2];
- RID ao_full; //when using half-size
+ RID ao_deinterleaved;
+ Vector<RID> ao_deinterleaved_slices;
+ RID ao_pong;
+ Vector<RID> ao_pong_slices;
+ RID ao_final;
+ RID importance_map[2];
} ssao;
struct SSR {
@@ -1502,7 +1516,7 @@ public:
virtual int sdfgi_get_pending_region_count(RID p_render_buffers) const;
virtual AABB sdfgi_get_pending_region_bounds(RID p_render_buffers, int p_region) const;
virtual uint32_t sdfgi_get_pending_region_cascade(RID p_render_buffers, int p_region) const;
- virtual void sdfgi_update_probes(RID p_render_buffers, RID p_environment, const RID *p_directional_light_instances, uint32_t p_directional_light_count, const RID *p_positional_light_instances, uint32_t p_positional_light_count);
+ virtual void sdfgi_update_probes(RID p_render_buffers, RID p_environment, const Vector<RID> &p_directional_lights, const RID *p_positional_light_instances, uint32_t p_positional_light_count);
RID sdfgi_get_ubo() const { return gi.sdfgi_ubo; }
/* SKY API */
@@ -1567,8 +1581,8 @@ public:
virtual void environment_set_volumetric_fog_positional_shadow_shrink_size(int p_shrink_size);
void 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);
- void environment_set_ssao(RID p_env, bool p_enable, float p_radius, float p_intensity, float p_bias, float p_light_affect, float p_ao_channel_affect, RS::EnvironmentSSAOBlur p_blur, float p_bilateral_sharpness);
- void environment_set_ssao_quality(RS::EnvironmentSSAOQuality p_quality, bool p_half_size);
+ void 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);
+ void 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);
bool environment_is_ssao_enabled(RID p_env) const;
float environment_get_ssao_ao_affect(RID p_env) const;
float environment_get_ssao_light_affect(RID p_env) const;
@@ -1811,7 +1825,7 @@ public:
RID gi_probe_instance_create(RID p_base);
void gi_probe_instance_set_transform_to_data(RID p_probe, const Transform &p_xform);
bool gi_probe_needs_update(RID p_probe) const;
- void gi_probe_update(RID p_probe, bool p_update_light_instances, const Vector<RID> &p_light_instances, int p_dynamic_object_count, InstanceBase **p_dynamic_objects);
+ void gi_probe_update(RID p_probe, bool p_update_light_instances, const Vector<RID> &p_light_instances, const PagedArray<RendererSceneRender::InstanceBase *> &p_dynamic_objects);
void gi_probe_set_quality(RS::GIProbeQuality p_quality) { gi_probe_quality = p_quality; }
@@ -1886,16 +1900,16 @@ public:
float render_buffers_get_volumetric_fog_end(RID p_render_buffers);
float render_buffers_get_volumetric_fog_detail_spread(RID p_render_buffers);
- void render_scene(RID p_render_buffers, const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, InstanceBase **p_cull_result, int p_cull_count, RID *p_light_cull_result, int p_light_cull_count, RID *p_reflection_probe_cull_result, int p_reflection_probe_cull_count, RID *p_gi_probe_cull_result, int p_gi_probe_cull_count, RID *p_decal_cull_result, int p_decal_cull_count, InstanceBase **p_lightmap_cull_result, int p_lightmap_cull_count, RID p_environment, RID p_shadow_atlas, RID p_camera_effects, RID p_reflection_atlas, RID p_reflection_probe, int p_reflection_probe_pass, float p_screen_lod_threshold);
+ void render_scene(RID p_render_buffers, const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, const PagedArray<InstanceBase *> &p_instances, const PagedArray<RID> &p_lights, const PagedArray<RID> &p_reflection_probes, const PagedArray<RID> &p_gi_probes, const PagedArray<RID> &p_decals, const PagedArray<InstanceBase *> &p_lightmaps, RID p_environment, RID p_camera_effects, RID p_shadow_atlas, RID p_reflection_atlas, RID p_reflection_probe, int p_reflection_probe_pass, float p_screen_lod_threshold);
- void render_shadow(RID p_light, RID p_shadow_atlas, int p_pass, InstanceBase **p_cull_result, int p_cull_count, const Plane &p_camera_plane = Plane(), float p_lod_distance_multiplier = 0, float p_screen_lod_threshold = 0.0);
+ void render_shadow(RID p_light, RID p_shadow_atlas, int p_pass, const PagedArray<InstanceBase *> &p_instances, const Plane &p_camera_plane = Plane(), float p_lod_distance_multiplier = 0, float p_screen_lod_threshold = 0.0);
- void render_material(const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, InstanceBase **p_cull_result, int p_cull_count, RID p_framebuffer, const Rect2i &p_region);
+ void render_material(const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, const PagedArray<InstanceBase *> &p_instances, RID p_framebuffer, const Rect2i &p_region);
- void render_sdfgi(RID p_render_buffers, int p_region, InstanceBase **p_cull_result, int p_cull_count);
- void render_sdfgi_static_lights(RID p_render_buffers, uint32_t p_cascade_count, const uint32_t *p_cascade_indices, const RID **p_positional_light_cull_result, const uint32_t *p_positional_light_cull_count);
+ void render_sdfgi(RID p_render_buffers, int p_region, const PagedArray<InstanceBase *> &p_instances);
+ void render_sdfgi_static_lights(RID p_render_buffers, uint32_t p_cascade_count, const uint32_t *p_cascade_indices, const PagedArray<RID> *p_positional_light_cull_result);
- void render_particle_collider_heightfield(RID p_collider, const Transform &p_transform, InstanceBase **p_cull_result, int p_cull_count);
+ void render_particle_collider_heightfield(RID p_collider, const Transform &p_transform, const PagedArray<InstanceBase *> &p_instances);
virtual void set_scene_pass(uint64_t p_pass) {
scene_pass = p_pass;
diff --git a/servers/rendering/renderer_rd/renderer_storage_rd.cpp b/servers/rendering/renderer_rd/renderer_storage_rd.cpp
index 60c0bd1603..9da4fab75d 100644
--- a/servers/rendering/renderer_rd/renderer_storage_rd.cpp
+++ b/servers/rendering/renderer_rd/renderer_storage_rd.cpp
@@ -537,7 +537,7 @@ Ref<Image> RendererStorageRD::_validate_texture_format(const Ref<Image> &p_image
RID RendererStorageRD::texture_2d_create(const Ref<Image> &p_image) {
ERR_FAIL_COND_V(p_image.is_null(), RID());
- ERR_FAIL_COND_V(p_image->empty(), RID());
+ ERR_FAIL_COND_V(p_image->is_empty(), RID());
TextureToRDFormat ret_format;
Ref<Image> image = _validate_texture_format(p_image, ret_format);
@@ -620,7 +620,7 @@ RID RendererStorageRD::texture_2d_layered_create(const Vector<Ref<Image>> &p_lay
Image::Format valid_format = Image::FORMAT_MAX;
for (int i = 0; i < p_layers.size(); i++) {
- ERR_FAIL_COND_V(p_layers[i]->empty(), RID());
+ ERR_FAIL_COND_V(p_layers[i]->is_empty(), RID());
if (i == 0) {
valid_width = p_layers[i]->get_width();
@@ -855,7 +855,7 @@ RID RendererStorageRD::texture_proxy_create(RID p_base) {
}
void RendererStorageRD::_texture_2d_update(RID p_texture, const Ref<Image> &p_image, int p_layer, bool p_immediate) {
- ERR_FAIL_COND(p_image.is_null() || p_image->empty());
+ ERR_FAIL_COND(p_image.is_null() || p_image->is_empty());
Texture *tex = texture_owner.getornull(p_texture);
ERR_FAIL_COND(!tex);
@@ -1039,7 +1039,7 @@ Ref<Image> RendererStorageRD::texture_2d_get(RID p_texture) const {
Ref<Image> image;
image.instance();
image->create(tex->width, tex->height, tex->mipmaps > 1, tex->validated_format, data);
- ERR_FAIL_COND_V(image->empty(), Ref<Image>());
+ ERR_FAIL_COND_V(image->is_empty(), Ref<Image>());
if (tex->format != tex->validated_format) {
image->convert(tex->format);
}
@@ -1062,7 +1062,7 @@ Ref<Image> RendererStorageRD::texture_2d_layer_get(RID p_texture, int p_layer) c
Ref<Image> image;
image.instance();
image->create(tex->width, tex->height, tex->mipmaps > 1, tex->validated_format, data);
- ERR_FAIL_COND_V(image->empty(), Ref<Image>());
+ ERR_FAIL_COND_V(image->is_empty(), Ref<Image>());
if (tex->format != tex->validated_format) {
image->convert(tex->format);
}
@@ -1090,7 +1090,7 @@ Vector<Ref<Image>> RendererStorageRD::texture_3d_get(RID p_texture) const {
Ref<Image> img;
img.instance();
img->create(bs.size.width, bs.size.height, false, tex->validated_format, sub_region);
- ERR_FAIL_COND_V(img->empty(), Vector<Ref<Image>>());
+ ERR_FAIL_COND_V(img->is_empty(), Vector<Ref<Image>>());
if (tex->format != tex->validated_format) {
img->convert(tex->format);
}
@@ -2408,9 +2408,6 @@ void RendererStorageRD::mesh_add_surface(RID p_mesh, const RS::SurfaceData &p_su
Mesh *mesh = mesh_owner.getornull(p_mesh);
ERR_FAIL_COND(!mesh);
- //ensure blend shape consistency
- ERR_FAIL_COND(mesh->blend_shape_count && p_surface.bone_aabbs.size() != mesh->bone_aabbs.size());
-
#ifdef DEBUG_ENABLED
//do a validation, to catch errors first
{
@@ -2576,6 +2573,11 @@ void RendererStorageRD::mesh_add_surface(RID p_mesh, const RS::SurfaceData &p_su
mesh->bone_aabbs = p_surface.bone_aabbs;
mesh->aabb = p_surface.aabb;
} else {
+ if (mesh->bone_aabbs.size() < p_surface.bone_aabbs.size()) {
+ // ArrayMesh::_surface_set_data only allocates bone_aabbs up to max_bone
+ // Each surface may affect different numbers of bones.
+ mesh->bone_aabbs.resize(p_surface.bone_aabbs.size());
+ }
for (int i = 0; i < p_surface.bone_aabbs.size(); i++) {
mesh->bone_aabbs.write[i].merge_with(p_surface.bone_aabbs[i]);
}
diff --git a/servers/rendering/renderer_rd/renderer_storage_rd.h b/servers/rendering/renderer_rd/renderer_storage_rd.h
index e4199ffd12..328e4d4999 100644
--- a/servers/rendering/renderer_rd/renderer_storage_rd.h
+++ b/servers/rendering/renderer_rd/renderer_storage_rd.h
@@ -1408,7 +1408,7 @@ public:
if (r_surface_count == 0) {
return nullptr;
}
- if (mesh->material_cache.empty()) {
+ if (mesh->material_cache.is_empty()) {
mesh->material_cache.resize(mesh->surface_count);
for (uint32_t i = 0; i < r_surface_count; i++) {
mesh->material_cache.write[i] = mesh->surfaces[i]->material;
diff --git a/servers/rendering/renderer_rd/shader_compiler_rd.cpp b/servers/rendering/renderer_rd/shader_compiler_rd.cpp
index 2c1d2a84fd..ae392b7586 100644
--- a/servers/rendering/renderer_rd/shader_compiler_rd.cpp
+++ b/servers/rendering/renderer_rd/shader_compiler_rd.cpp
@@ -920,7 +920,7 @@ String ShaderCompilerRD::_dump_node_code(const SL::Node *p_node, int p_level, Ge
if (adnode->datatype == SL::TYPE_STRUCT) {
declaration += _mkid(adnode->struct_name);
} else {
- declaration = _prestr(adnode->precision) + _typestr(adnode->datatype);
+ declaration += _prestr(adnode->precision) + _typestr(adnode->datatype);
}
for (int i = 0; i < adnode->declarations.size(); i++) {
if (i > 0) {
@@ -930,7 +930,11 @@ String ShaderCompilerRD::_dump_node_code(const SL::Node *p_node, int p_level, Ge
}
declaration += _mkid(adnode->declarations[i].name);
declaration += "[";
- declaration += itos(adnode->declarations[i].size);
+ if (adnode->size_expression != nullptr) {
+ declaration += _dump_node_code(adnode->size_expression, p_level, r_gen_code, p_actions, p_default_actions, p_assigning);
+ } else {
+ declaration += itos(adnode->declarations[i].size);
+ }
declaration += "]";
int sz = adnode->declarations[i].initializer.size();
if (sz > 0) {
@@ -986,12 +990,13 @@ String ShaderCompilerRD::_dump_node_code(const SL::Node *p_node, int p_level, Ge
if (anode->call_expression != nullptr) {
code += ".";
code += _dump_node_code(anode->call_expression, p_level, r_gen_code, p_actions, p_default_actions, p_assigning, false);
- }
-
- if (anode->index_expression != nullptr) {
+ } else if (anode->index_expression != nullptr) {
code += "[";
code += _dump_node_code(anode->index_expression, p_level, r_gen_code, p_actions, p_default_actions, p_assigning);
code += "]";
+ } else if (anode->assign_expression != nullptr) {
+ code += "=";
+ code += _dump_node_code(anode->assign_expression, p_level, r_gen_code, p_actions, p_default_actions, true, false);
}
if (anode->name == time_name) {
@@ -1229,8 +1234,10 @@ String ShaderCompilerRD::_dump_node_code(const SL::Node *p_node, int p_level, Ge
code += "[";
code += _dump_node_code(mnode->index_expression, p_level, r_gen_code, p_actions, p_default_actions, p_assigning);
code += "]";
+ } else if (mnode->assign_expression != nullptr) {
+ code += "=";
+ code += _dump_node_code(mnode->assign_expression, p_level, r_gen_code, p_actions, p_default_actions, true, false);
}
-
} break;
}
diff --git a/servers/rendering/renderer_rd/shaders/SCsub b/servers/rendering/renderer_rd/shaders/SCsub
index cb62882deb..deaa9668df 100644
--- a/servers/rendering/renderer_rd/shaders/SCsub
+++ b/servers/rendering/renderer_rd/shaders/SCsub
@@ -21,8 +21,10 @@ if "RD_GLSL" in env["BUILDERS"]:
env.RD_GLSL("luminance_reduce.glsl")
env.RD_GLSL("bokeh_dof.glsl")
env.RD_GLSL("ssao.glsl")
- env.RD_GLSL("ssao_minify.glsl")
+ env.RD_GLSL("ssao_downsample.glsl")
+ env.RD_GLSL("ssao_importance_map.glsl")
env.RD_GLSL("ssao_blur.glsl")
+ env.RD_GLSL("ssao_interleave.glsl")
env.RD_GLSL("roughness_limiter.glsl")
env.RD_GLSL("screen_space_reflection.glsl")
env.RD_GLSL("screen_space_reflection_filter.glsl")
diff --git a/servers/rendering/renderer_rd/shaders/ssao.glsl b/servers/rendering/renderer_rd/shaders/ssao.glsl
index 346338181a..315ef8fa13 100644
--- a/servers/rendering/renderer_rd/shaders/ssao.glsl
+++ b/servers/rendering/renderer_rd/shaders/ssao.glsl
@@ -1,249 +1,486 @@
+///////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+// Copyright (c) 2016, Intel Corporation
+// 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.
+///////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+// File changes (yyyy-mm-dd)
+// 2016-09-07: filip.strugar@intel.com: first commit
+// 2020-12-05: clayjohn: convert to Vulkan and Godot
+///////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+
#[compute]
#version 450
VERSION_DEFINES
-layout(local_size_x = 8, local_size_y = 8, local_size_z = 1) in;
+#define SSAO_ADAPTIVE_TAP_BASE_COUNT 5
+
+#define INTELSSAO_MAIN_DISK_SAMPLE_COUNT (32)
+const vec4 sample_pattern[INTELSSAO_MAIN_DISK_SAMPLE_COUNT] = {
+ vec4(0.78488064, 0.56661671, 1.500000, -0.126083), vec4(0.26022232, -0.29575172, 1.500000, -1.064030), vec4(0.10459357, 0.08372527, 1.110000, -2.730563), vec4(-0.68286800, 0.04963045, 1.090000, -0.498827),
+ vec4(-0.13570161, -0.64190155, 1.250000, -0.532765), vec4(-0.26193795, -0.08205118, 0.670000, -1.783245), vec4(-0.61177456, 0.66664219, 0.710000, -0.044234), vec4(0.43675563, 0.25119025, 0.610000, -1.167283),
+ vec4(0.07884444, 0.86618668, 0.640000, -0.459002), vec4(-0.12790935, -0.29869005, 0.600000, -1.729424), vec4(-0.04031125, 0.02413622, 0.600000, -4.792042), vec4(0.16201244, -0.52851415, 0.790000, -1.067055),
+ vec4(-0.70991218, 0.47301072, 0.640000, -0.335236), vec4(0.03277707, -0.22349690, 0.600000, -1.982384), vec4(0.68921727, 0.36800742, 0.630000, -0.266718), vec4(0.29251814, 0.37775412, 0.610000, -1.422520),
+ vec4(-0.12224089, 0.96582592, 0.600000, -0.426142), vec4(0.11071457, -0.16131058, 0.600000, -2.165947), vec4(0.46562141, -0.59747696, 0.600000, -0.189760), vec4(-0.51548797, 0.11804193, 0.600000, -1.246800),
+ vec4(0.89141309, -0.42090443, 0.600000, 0.028192), vec4(-0.32402530, -0.01591529, 0.600000, -1.543018), vec4(0.60771245, 0.41635221, 0.600000, -0.605411), vec4(0.02379565, -0.08239821, 0.600000, -3.809046),
+ vec4(0.48951152, -0.23657045, 0.600000, -1.189011), vec4(-0.17611565, -0.81696892, 0.600000, -0.513724), vec4(-0.33930185, -0.20732205, 0.600000, -1.698047), vec4(-0.91974425, 0.05403209, 0.600000, 0.062246),
+ vec4(-0.15064627, -0.14949332, 0.600000, -1.896062), vec4(0.53180975, -0.35210401, 0.600000, -0.758838), vec4(0.41487166, 0.81442589, 0.600000, -0.505648), vec4(-0.24106961, -0.32721516, 0.600000, -1.665244)
+};
+
+// these values can be changed (up to SSAO_MAX_TAPS) with no changes required elsewhere; values for 4th and 5th preset are ignored but array needed to avoid compilation errors
+// the actual number of texture samples is two times this value (each "tap" has two symmetrical depth texture samples)
+const int num_taps[5] = { 3, 5, 12, 0, 0 };
+
+#define SSAO_TILT_SAMPLES_ENABLE_AT_QUALITY_PRESET (99) // to disable simply set to 99 or similar
+#define SSAO_TILT_SAMPLES_AMOUNT (0.4)
+//
+#define SSAO_HALOING_REDUCTION_ENABLE_AT_QUALITY_PRESET (1) // to disable simply set to 99 or similar
+#define SSAO_HALOING_REDUCTION_AMOUNT (0.6) // values from 0.0 - 1.0, 1.0 means max weighting (will cause artifacts, 0.8 is more reasonable)
+//
+#define SSAO_NORMAL_BASED_EDGES_ENABLE_AT_QUALITY_PRESET (2) // to disable simply set to 99 or similar
+#define SSAO_NORMAL_BASED_EDGES_DOT_THRESHOLD (0.5) // use 0-0.1 for super-sharp normal-based edges
+//
+#define SSAO_DETAIL_AO_ENABLE_AT_QUALITY_PRESET (1) // whether to use detail; to disable simply set to 99 or similar
+//
+#define SSAO_DEPTH_MIPS_ENABLE_AT_QUALITY_PRESET (2) // !!warning!! the MIP generation on the C++ side will be enabled on quality preset 2 regardless of this value, so if changing here, change the C++ side too
+#define SSAO_DEPTH_MIPS_GLOBAL_OFFSET (-4.3) // best noise/quality/performance tradeoff, found empirically
+//
+// !!warning!! the edge handling is hard-coded to 'disabled' on quality level 0, and enabled above, on the C++ side; while toggling it here will work for
+// testing purposes, it will not yield performance gains (or correct results)
+#define SSAO_DEPTH_BASED_EDGES_ENABLE_AT_QUALITY_PRESET (1)
+//
+#define SSAO_REDUCE_RADIUS_NEAR_SCREEN_BORDER_ENABLE_AT_QUALITY_PRESET (1)
+
+#define SSAO_MAX_TAPS 32
+#define SSAO_MAX_REF_TAPS 512
+#define SSAO_ADAPTIVE_TAP_BASE_COUNT 5
+#define SSAO_ADAPTIVE_TAP_FLEXIBLE_COUNT (SSAO_MAX_TAPS - SSAO_ADAPTIVE_TAP_BASE_COUNT)
+#define SSAO_DEPTH_MIP_LEVELS 4
-#define TWO_PI 6.283185307179586476925286766559
+layout(local_size_x = 8, local_size_y = 8, local_size_z = 1) in;
-#ifdef SSAO_QUALITY_HIGH
-#define NUM_SAMPLES (20)
-#endif
+layout(set = 0, binding = 0) uniform sampler2DArray source_depth_mipmaps;
+layout(rgba8, set = 0, binding = 1) uniform restrict readonly image2D source_normal;
+layout(set = 0, binding = 2) uniform Constants { //get into a lower set
+ vec4 rotation_matrices[20];
+}
+constants;
-#ifdef SSAO_QUALITY_ULTRA
-#define NUM_SAMPLES (48)
+#ifdef ADAPTIVE
+layout(rg8, set = 1, binding = 0) uniform restrict readonly image2DArray source_ssao;
+layout(set = 1, binding = 1) uniform sampler2D source_importance;
+layout(set = 1, binding = 2, std430) buffer Counter {
+ uint sum;
+}
+counter;
#endif
-#ifdef SSAO_QUALITY_LOW
-#define NUM_SAMPLES (8)
-#endif
+layout(rg8, set = 2, binding = 0) uniform restrict writeonly image2D dest_image;
-#if !defined(SSAO_QUALITY_LOW) && !defined(SSAO_QUALITY_HIGH) && !defined(SSAO_QUALITY_ULTRA)
-#define NUM_SAMPLES (12)
-#endif
+// This push_constant is full - 128 bytes - if you need to add more data, consider adding to the uniform buffer instead
+layout(push_constant, binding = 1, std430) uniform Params {
+ ivec2 screen_size;
+ int pass;
+ int quality;
-// If using depth mip levels, the log of the maximum pixel offset before we need to switch to a lower
-// miplevel to maintain reasonable spatial locality in the cache
-// If this number is too small (< 3), too many taps will land in the same pixel, and we'll get bad variance that manifests as flashing.
-// If it is too high (> 5), we'll get bad performance because we're not using the MIP levels effectively
-#define LOG_MAX_OFFSET (3)
-
-// This must be less than or equal to the MAX_MIP_LEVEL defined in SSAO.cpp
-#define MAX_MIP_LEVEL (4)
-
-// This is the number of turns around the circle that the spiral pattern makes. This should be prime to prevent
-// taps from lining up. This particular choice was tuned for NUM_SAMPLES == 9
-
-const int ROTATIONS[] = int[](
- 1, 1, 2, 3, 2, 5, 2, 3, 2,
- 3, 3, 5, 5, 3, 4, 7, 5, 5, 7,
- 9, 8, 5, 5, 7, 7, 7, 8, 5, 8,
- 11, 12, 7, 10, 13, 8, 11, 8, 7, 14,
- 11, 11, 13, 12, 13, 19, 17, 13, 11, 18,
- 19, 11, 11, 14, 17, 21, 15, 16, 17, 18,
- 13, 17, 11, 17, 19, 18, 25, 18, 19, 19,
- 29, 21, 19, 27, 31, 29, 21, 18, 17, 29,
- 31, 31, 23, 18, 25, 26, 25, 23, 19, 34,
- 19, 27, 21, 25, 39, 29, 17, 21, 27);
-
-//#define NUM_SPIRAL_TURNS (7)
-const int NUM_SPIRAL_TURNS = ROTATIONS[NUM_SAMPLES - 1];
-
-layout(set = 0, binding = 0) uniform sampler2D source_depth_mipmaps;
-layout(r8, set = 1, binding = 0) uniform restrict writeonly image2D dest_image;
-
-#ifndef USE_HALF_SIZE
-layout(set = 2, binding = 0) uniform sampler2D source_depth;
-#endif
+ vec2 half_screen_pixel_size;
+ int size_multiplier;
+ float detail_intensity;
-layout(set = 3, binding = 0) uniform sampler2D source_normal;
+ vec2 NDC_to_view_mul;
+ vec2 NDC_to_view_add;
-layout(push_constant, binding = 1, std430) uniform Params {
- ivec2 screen_size;
- float z_far;
- float z_near;
+ vec2 pad2;
+ vec2 half_screen_pixel_size_x025;
- bool orthogonal;
- float intensity_div_r6;
float radius;
- float bias;
-
- vec4 proj_info;
- vec2 pixel_size;
- float proj_scale;
- uint pad;
+ float intensity;
+ float shadow_power;
+ float shadow_clamp;
+
+ float fade_out_mul;
+ float fade_out_add;
+ float horizon_angle_threshold;
+ float inv_radius_near_limit;
+
+ bool is_orthogonal;
+ float neg_inv_radius;
+ float load_counter_avg_div;
+ float adaptive_sample_limit;
+
+ ivec2 pass_coord_offset;
+ vec2 pass_uv_offset;
}
params;
-vec3 reconstructCSPosition(vec2 S, float z) {
- if (params.orthogonal) {
- return vec3((S.xy * params.proj_info.xy + params.proj_info.zw), z);
+// packing/unpacking for edges; 2 bits per edge mean 4 gradient values (0, 0.33, 0.66, 1) for smoother transitions!
+float pack_edges(vec4 p_edgesLRTB) {
+ p_edgesLRTB = round(clamp(p_edgesLRTB, 0.0, 1.0) * 3.05);
+ return dot(p_edgesLRTB, vec4(64.0 / 255.0, 16.0 / 255.0, 4.0 / 255.0, 1.0 / 255.0));
+}
+
+vec3 NDC_to_view_space(vec2 p_pos, float p_viewspace_depth) {
+ if (params.is_orthogonal) {
+ return vec3((params.NDC_to_view_mul * p_pos.xy + params.NDC_to_view_add), p_viewspace_depth);
} else {
- return vec3((S.xy * params.proj_info.xy + params.proj_info.zw) * z, z);
+ return vec3((params.NDC_to_view_mul * p_pos.xy + params.NDC_to_view_add) * p_viewspace_depth, p_viewspace_depth);
}
}
-vec3 getPosition(ivec2 ssP) {
- vec3 P;
-#ifdef USE_HALF_SIZE
- P.z = texelFetch(source_depth_mipmaps, ssP, 0).r;
- P.z = -P.z;
-#else
- P.z = texelFetch(source_depth, ssP, 0).r;
+// calculate effect radius and fit our screen sampling pattern inside it
+void calculate_radius_parameters(const float p_pix_center_length, const vec2 p_pixel_size_at_center, out float r_lookup_radius, out float r_radius, out float r_fallof_sq) {
+ r_radius = params.radius;
- P.z = P.z * 2.0 - 1.0;
- if (params.orthogonal) {
- P.z = ((P.z + (params.z_far + params.z_near) / (params.z_far - params.z_near)) * (params.z_far - params.z_near)) / 2.0;
- } else {
- P.z = 2.0 * params.z_near * params.z_far / (params.z_far + params.z_near - P.z * (params.z_far - params.z_near));
- }
- P.z = -P.z;
-#endif
- // Offset to pixel center
- P = reconstructCSPosition(vec2(ssP) + vec2(0.5), P.z);
- return P;
+ // when too close, on-screen sampling disk will grow beyond screen size; limit this to avoid closeup temporal artifacts
+ const float too_close_limit = clamp(p_pix_center_length * params.inv_radius_near_limit, 0.0, 1.0) * 0.8 + 0.2;
+
+ r_radius *= too_close_limit;
+
+ // 0.85 is to reduce the radius to allow for more samples on a slope to still stay within influence
+ r_lookup_radius = (0.85 * r_radius) / p_pixel_size_at_center.x;
+
+ // used to calculate falloff (both for AO samples and per-sample weights)
+ r_fallof_sq = -1.0 / (r_radius * r_radius);
}
-/** Returns a unit vector and a screen-space radius for the tap on a unit disk (the caller should scale by the actual disk radius) */
-vec2 tapLocation(int sampleNumber, float spinAngle, out float ssR) {
- // Radius relative to ssR
- float alpha = (float(sampleNumber) + 0.5) * (1.0 / float(NUM_SAMPLES));
- float angle = alpha * (float(NUM_SPIRAL_TURNS) * 6.28) + spinAngle;
+vec4 calculate_edges(const float p_center_z, const float p_left_z, const float p_right_z, const float p_top_z, const float p_bottom_z) {
+ // slope-sensitive depth-based edge detection
+ vec4 edgesLRTB = vec4(p_left_z, p_right_z, p_top_z, p_bottom_z) - p_center_z;
+ vec4 edgesLRTB_slope_adjusted = edgesLRTB + edgesLRTB.yxwz;
+ edgesLRTB = min(abs(edgesLRTB), abs(edgesLRTB_slope_adjusted));
+ return clamp((1.3 - edgesLRTB / (p_center_z * 0.040)), 0.0, 1.0);
+}
- ssR = alpha;
- return vec2(cos(angle), sin(angle));
+vec3 decode_normal(vec3 p_encoded_normal) {
+ vec3 normal = p_encoded_normal * 2.0 - 1.0;
+ return normal;
}
-/** Read the camera-space position of the point at screen-space pixel ssP + unitOffset * ssR. Assumes length(unitOffset) == 1 */
-vec3 getOffsetPosition(ivec2 ssP, float ssR) {
- // Derivation:
- // mipLevel = floor(log(ssR / MAX_OFFSET));
+vec3 load_normal(ivec2 p_pos) {
+ vec3 encoded_normal = imageLoad(source_normal, p_pos).xyz;
+ encoded_normal.z = 1.0 - encoded_normal.z;
+ return decode_normal(encoded_normal);
+}
- int mipLevel = clamp(int(floor(log2(ssR))) - LOG_MAX_OFFSET, 0, MAX_MIP_LEVEL);
+vec3 load_normal(ivec2 p_pos, ivec2 p_offset) {
+ vec3 encoded_normal = imageLoad(source_normal, p_pos + p_offset).xyz;
+ encoded_normal.z = 1.0 - encoded_normal.z;
+ return decode_normal(encoded_normal);
+}
- vec3 P;
+// all vectors in viewspace
+float calculate_pixel_obscurance(vec3 p_pixel_normal, vec3 p_hit_delta, float p_fallof_sq) {
+ float length_sq = dot(p_hit_delta, p_hit_delta);
+ float NdotD = dot(p_pixel_normal, p_hit_delta) / sqrt(length_sq);
- // We need to divide by 2^mipLevel to read the appropriately scaled coordinate from a MIP-map.
- // Manually clamp to the texture size because texelFetch bypasses the texture unit
- ivec2 mipP = clamp(ssP >> mipLevel, ivec2(0), (params.screen_size >> mipLevel) - ivec2(1));
+ float falloff_mult = max(0.0, length_sq * p_fallof_sq + 1.0);
-#ifdef USE_HALF_SIZE
- P.z = texelFetch(source_depth_mipmaps, mipP, mipLevel).r;
- P.z = -P.z;
-#else
- if (mipLevel < 1) {
- //read from depth buffer
- P.z = texelFetch(source_depth, mipP, 0).r;
- P.z = P.z * 2.0 - 1.0;
- if (params.orthogonal) {
- P.z = ((P.z + (params.z_far + params.z_near) / (params.z_far - params.z_near)) * (params.z_far - params.z_near)) / 2.0;
- } else {
- P.z = 2.0 * params.z_near * params.z_far / (params.z_far + params.z_near - P.z * (params.z_far - params.z_near));
- }
- P.z = -P.z;
+ return max(0, NdotD - params.horizon_angle_threshold) * falloff_mult;
+}
- } else {
- //read from mipmaps
- P.z = texelFetch(source_depth_mipmaps, mipP, mipLevel - 1).r;
- P.z = -P.z;
+void SSAO_tap_inner(const int p_quality_level, inout float r_obscurance_sum, inout float r_weight_sum, const vec2 p_sampling_uv, const float p_mip_level, const vec3 p_pix_center_pos, vec3 p_pixel_normal, const float p_fallof_sq, const float p_weight_mod) {
+ // get depth at sample
+ float viewspace_sample_z = textureLod(source_depth_mipmaps, vec3(p_sampling_uv, params.pass), p_mip_level).x;
+
+ // convert to viewspace
+ vec3 hit_pos = NDC_to_view_space(p_sampling_uv.xy, viewspace_sample_z).xyz;
+ vec3 hit_delta = hit_pos - p_pix_center_pos;
+
+ float obscurance = calculate_pixel_obscurance(p_pixel_normal, hit_delta, p_fallof_sq);
+ float weight = 1.0;
+
+ if (p_quality_level >= SSAO_HALOING_REDUCTION_ENABLE_AT_QUALITY_PRESET) {
+ float reduct = max(0, -hit_delta.z);
+ reduct = clamp(reduct * params.neg_inv_radius + 2.0, 0.0, 1.0);
+ weight = SSAO_HALOING_REDUCTION_AMOUNT * reduct + (1.0 - SSAO_HALOING_REDUCTION_AMOUNT);
}
-#endif
+ weight *= p_weight_mod;
+ r_obscurance_sum += obscurance * weight;
+ r_weight_sum += weight;
+}
+
+void SSAOTap(const int p_quality_level, inout float r_obscurance_sum, inout float r_weight_sum, const int p_tap_index, const mat2 p_rot_scale, const vec3 p_pix_center_pos, vec3 p_pixel_normal, const vec2 p_normalized_screen_pos, const float p_mip_offset, const float p_fallof_sq, float p_weight_mod, vec2 p_norm_xy, float p_norm_xy_length) {
+ vec2 sample_offset;
+ float sample_pow_2_len;
+
+ // patterns
+ {
+ vec4 new_sample = sample_pattern[p_tap_index];
+ sample_offset = new_sample.xy * p_rot_scale;
+ sample_pow_2_len = new_sample.w; // precalculated, same as: sample_pow_2_len = log2( length( new_sample.xy ) );
+ p_weight_mod *= new_sample.z;
+ }
+
+ // snap to pixel center (more correct obscurance math, avoids artifacts)
+ sample_offset = round(sample_offset);
+
+ // calculate MIP based on the sample distance from the centre, similar to as described
+ // in http://graphics.cs.williams.edu/papers/SAOHPG12/.
+ float mip_level = (p_quality_level < SSAO_DEPTH_MIPS_ENABLE_AT_QUALITY_PRESET) ? (0) : (sample_pow_2_len + p_mip_offset);
+
+ vec2 sampling_uv = sample_offset * params.half_screen_pixel_size + p_normalized_screen_pos;
+
+ SSAO_tap_inner(p_quality_level, r_obscurance_sum, r_weight_sum, sampling_uv, mip_level, p_pix_center_pos, p_pixel_normal, p_fallof_sq, p_weight_mod);
+
+ // for the second tap, just use the mirrored offset
+ vec2 sample_offset_mirrored_uv = -sample_offset;
- // Offset to pixel center
- P = reconstructCSPosition(vec2(ssP) + vec2(0.5), P.z);
+ // tilt the second set of samples so that the disk is effectively rotated by the normal
+ // effective at removing one set of artifacts, but too expensive for lower quality settings
+ if (p_quality_level >= SSAO_TILT_SAMPLES_ENABLE_AT_QUALITY_PRESET) {
+ float dot_norm = dot(sample_offset_mirrored_uv, p_norm_xy);
+ sample_offset_mirrored_uv -= dot_norm * p_norm_xy_length * p_norm_xy;
+ sample_offset_mirrored_uv = round(sample_offset_mirrored_uv);
+ }
+
+ // snap to pixel center (more correct obscurance math, avoids artifacts)
+ vec2 sampling_mirrored_uv = sample_offset_mirrored_uv * params.half_screen_pixel_size + p_normalized_screen_pos;
- return P;
+ SSAO_tap_inner(p_quality_level, r_obscurance_sum, r_weight_sum, sampling_mirrored_uv, mip_level, p_pix_center_pos, p_pixel_normal, p_fallof_sq, p_weight_mod);
}
-/** Compute the occlusion due to sample with index \a i about the pixel at \a ssC that corresponds
- to camera-space point \a C with unit normal \a n_C, using maximum screen-space sampling radius \a ssDiskRadius
+void generate_SSAO_shadows_internal(out float r_shadow_term, out vec4 r_edges, out float r_weight, const vec2 p_pos, int p_quality_level, bool p_adaptive_base) {
+ vec2 pos_rounded = trunc(p_pos);
+ uvec2 upos = uvec2(pos_rounded);
+
+ const int number_of_taps = (p_adaptive_base) ? (SSAO_ADAPTIVE_TAP_BASE_COUNT) : (num_taps[p_quality_level]);
+ float pix_z, pix_left_z, pix_top_z, pix_right_z, pix_bottom_z;
+
+ vec4 valuesUL = textureGather(source_depth_mipmaps, vec3(pos_rounded * params.half_screen_pixel_size, params.pass));
+ vec4 valuesBR = textureGather(source_depth_mipmaps, vec3((pos_rounded + vec2(1.0)) * params.half_screen_pixel_size, params.pass));
+
+ // get this pixel's viewspace depth
+ pix_z = valuesUL.y;
+
+ // get left right top bottom neighbouring pixels for edge detection (gets compiled out on quality_level == 0)
+ pix_left_z = valuesUL.x;
+ pix_top_z = valuesUL.z;
+ pix_right_z = valuesBR.z;
+ pix_bottom_z = valuesBR.x;
- Note that units of H() in the HPG12 paper are meters, not
- unitless. The whole falloff/sampling function is therefore
- unitless. In this implementation, we factor out (9 / radius).
+ vec2 normalized_screen_pos = pos_rounded * params.half_screen_pixel_size + params.half_screen_pixel_size_x025;
+ vec3 pix_center_pos = NDC_to_view_space(normalized_screen_pos, pix_z);
- Four versions of the falloff function are implemented below
-*/
-float sampleAO(in ivec2 ssC, in vec3 C, in vec3 n_C, in float ssDiskRadius, in float p_radius, in int tapIndex, in float randomPatternRotationAngle) {
- // Offset on the unit disk, spun for this pixel
- float ssR;
- vec2 unitOffset = tapLocation(tapIndex, randomPatternRotationAngle, ssR);
- ssR *= ssDiskRadius;
+ // Load this pixel's viewspace normal
+ uvec2 full_res_coord = upos * 2 * params.size_multiplier + params.pass_coord_offset.xy;
+ vec3 pixel_normal = load_normal(ivec2(full_res_coord));
- ivec2 ssP = ivec2(ssR * unitOffset) + ssC;
+ const vec2 pixel_size_at_center = NDC_to_view_space(normalized_screen_pos.xy + params.half_screen_pixel_size, pix_center_pos.z).xy - pix_center_pos.xy;
- if (any(lessThan(ssP, ivec2(0))) || any(greaterThanEqual(ssP, params.screen_size))) {
- return 0.0;
+ float pixel_lookup_radius;
+ float fallof_sq;
+
+ // calculate effect radius and fit our screen sampling pattern inside it
+ float viewspace_radius;
+ calculate_radius_parameters(length(pix_center_pos), pixel_size_at_center, pixel_lookup_radius, viewspace_radius, fallof_sq);
+
+ // calculate samples rotation/scaling
+ mat2 rot_scale_matrix;
+ uint pseudo_random_index;
+
+ {
+ vec4 rotation_scale;
+ // reduce effect radius near the screen edges slightly; ideally, one would render a larger depth buffer (5% on each side) instead
+ if (!p_adaptive_base && (p_quality_level >= SSAO_REDUCE_RADIUS_NEAR_SCREEN_BORDER_ENABLE_AT_QUALITY_PRESET)) {
+ float near_screen_border = min(min(normalized_screen_pos.x, 1.0 - normalized_screen_pos.x), min(normalized_screen_pos.y, 1.0 - normalized_screen_pos.y));
+ near_screen_border = clamp(10.0 * near_screen_border + 0.6, 0.0, 1.0);
+ pixel_lookup_radius *= near_screen_border;
+ }
+
+ // load & update pseudo-random rotation matrix
+ pseudo_random_index = uint(pos_rounded.y * 2 + pos_rounded.x) % 5;
+ rotation_scale = constants.rotation_matrices[params.pass * 5 + pseudo_random_index];
+ rot_scale_matrix = mat2(rotation_scale.x * pixel_lookup_radius, rotation_scale.y * pixel_lookup_radius, rotation_scale.z * pixel_lookup_radius, rotation_scale.w * pixel_lookup_radius);
}
- // The occluding point in camera space
- vec3 Q = getOffsetPosition(ssP, ssR);
+ // the main obscurance & sample weight storage
+ float obscurance_sum = 0.0;
+ float weight_sum = 0.0;
- vec3 v = Q - C;
+ // edge mask for between this and left/right/top/bottom neighbour pixels - not used in quality level 0 so initialize to "no edge" (1 is no edge, 0 is edge)
+ vec4 edgesLRTB = vec4(1.0, 1.0, 1.0, 1.0);
- float vv = dot(v, v);
- float vn = dot(v, n_C);
+ // Move center pixel slightly towards camera to avoid imprecision artifacts due to using of 16bit depth buffer; a lot smaller offsets needed when using 32bit floats
+ pix_center_pos *= 0.9992;
- const float epsilon = 0.01;
- float radius2 = p_radius * p_radius;
+ if (!p_adaptive_base && (p_quality_level >= SSAO_DEPTH_BASED_EDGES_ENABLE_AT_QUALITY_PRESET)) {
+ edgesLRTB = calculate_edges(pix_z, pix_left_z, pix_right_z, pix_top_z, pix_bottom_z);
+ }
- // A: From the HPG12 paper
- // Note large epsilon to avoid overdarkening within cracks
- //return float(vv < radius2) * max((vn - bias) / (epsilon + vv), 0.0) * radius2 * 0.6;
+ // adds a more high definition sharp effect, which gets blurred out (reuses left/right/top/bottom samples that we used for edge detection)
+ if (!p_adaptive_base && (p_quality_level >= SSAO_DETAIL_AO_ENABLE_AT_QUALITY_PRESET)) {
+ // disable in case of quality level 4 (reference)
+ if (p_quality_level != 4) {
+ //approximate neighbouring pixels positions (actually just deltas or "positions - pix_center_pos" )
+ vec3 normalized_viewspace_dir = vec3(pix_center_pos.xy / pix_center_pos.zz, 1.0);
+ vec3 pixel_left_delta = vec3(-pixel_size_at_center.x, 0.0, 0.0) + normalized_viewspace_dir * (pix_left_z - pix_center_pos.z);
+ vec3 pixel_right_delta = vec3(+pixel_size_at_center.x, 0.0, 0.0) + normalized_viewspace_dir * (pix_right_z - pix_center_pos.z);
+ vec3 pixel_top_delta = vec3(0.0, -pixel_size_at_center.y, 0.0) + normalized_viewspace_dir * (pix_top_z - pix_center_pos.z);
+ vec3 pixel_bottom_delta = vec3(0.0, +pixel_size_at_center.y, 0.0) + normalized_viewspace_dir * (pix_bottom_z - pix_center_pos.z);
+
+ const float range_reduction = 4.0f; // this is to avoid various artifacts
+ const float modified_fallof_sq = range_reduction * fallof_sq;
+
+ vec4 additional_obscurance;
+ additional_obscurance.x = calculate_pixel_obscurance(pixel_normal, pixel_left_delta, modified_fallof_sq);
+ additional_obscurance.y = calculate_pixel_obscurance(pixel_normal, pixel_right_delta, modified_fallof_sq);
+ additional_obscurance.z = calculate_pixel_obscurance(pixel_normal, pixel_top_delta, modified_fallof_sq);
+ additional_obscurance.w = calculate_pixel_obscurance(pixel_normal, pixel_bottom_delta, modified_fallof_sq);
+
+ obscurance_sum += params.detail_intensity * dot(additional_obscurance, edgesLRTB);
+ }
+ }
- // B: Smoother transition to zero (lowers contrast, smoothing out corners). [Recommended]
- float f = max(radius2 - vv, 0.0);
- return f * f * f * max((vn - params.bias) / (epsilon + vv), 0.0);
+ // Sharp normals also create edges - but this adds to the cost as well
+ if (!p_adaptive_base && (p_quality_level >= SSAO_NORMAL_BASED_EDGES_ENABLE_AT_QUALITY_PRESET)) {
+ vec3 neighbour_normal_left = load_normal(ivec2(full_res_coord), ivec2(-2, 0));
+ vec3 neighbour_normal_right = load_normal(ivec2(full_res_coord), ivec2(2, 0));
+ vec3 neighbour_normal_top = load_normal(ivec2(full_res_coord), ivec2(0, -2));
+ vec3 neighbour_normal_bottom = load_normal(ivec2(full_res_coord), ivec2(0, 2));
- // C: Medium contrast (which looks better at high radii), no division. Note that the
- // contribution still falls off with radius^2, but we've adjusted the rate in a way that is
- // more computationally efficient and happens to be aesthetically pleasing.
- // return 4.0 * max(1.0 - vv * invRadius2, 0.0) * max(vn - bias, 0.0);
+ const float dot_threshold = SSAO_NORMAL_BASED_EDGES_DOT_THRESHOLD;
- // D: Low contrast, no division operation
- // return 2.0 * float(vv < radius * radius) * max(vn - bias, 0.0);
-}
+ vec4 normal_edgesLRTB;
+ normal_edgesLRTB.x = clamp((dot(pixel_normal, neighbour_normal_left) + dot_threshold), 0.0, 1.0);
+ normal_edgesLRTB.y = clamp((dot(pixel_normal, neighbour_normal_right) + dot_threshold), 0.0, 1.0);
+ normal_edgesLRTB.z = clamp((dot(pixel_normal, neighbour_normal_top) + dot_threshold), 0.0, 1.0);
+ normal_edgesLRTB.w = clamp((dot(pixel_normal, neighbour_normal_bottom) + dot_threshold), 0.0, 1.0);
-void main() {
- // Pixel being shaded
- ivec2 ssC = ivec2(gl_GlobalInvocationID.xy);
- if (any(greaterThanEqual(ssC, params.screen_size))) { //too large, do nothing
- return;
+ edgesLRTB *= normal_edgesLRTB;
}
- // World space point being shaded
- vec3 C = getPosition(ssC);
+ const float global_mip_offset = SSAO_DEPTH_MIPS_GLOBAL_OFFSET;
+ float mip_offset = (p_quality_level < SSAO_DEPTH_MIPS_ENABLE_AT_QUALITY_PRESET) ? (0) : (log2(pixel_lookup_radius) + global_mip_offset);
-#ifdef USE_HALF_SIZE
- vec3 n_C = texelFetch(source_normal, ssC << 1, 0).xyz * 2.0 - 1.0;
-#else
- vec3 n_C = texelFetch(source_normal, ssC, 0).xyz * 2.0 - 1.0;
+ // Used to tilt the second set of samples so that the disk is effectively rotated by the normal
+ // effective at removing one set of artifacts, but too expensive for lower quality settings
+ vec2 norm_xy = vec2(pixel_normal.x, pixel_normal.y);
+ float norm_xy_length = length(norm_xy);
+ norm_xy /= vec2(norm_xy_length, -norm_xy_length);
+ norm_xy_length *= SSAO_TILT_SAMPLES_AMOUNT;
+
+ // standard, non-adaptive approach
+ if ((p_quality_level != 3) || p_adaptive_base) {
+ for (int i = 0; i < number_of_taps; i++) {
+ SSAOTap(p_quality_level, obscurance_sum, weight_sum, i, rot_scale_matrix, pix_center_pos, pixel_normal, normalized_screen_pos, mip_offset, fallof_sq, 1.0, norm_xy, norm_xy_length);
+ }
+ }
+#ifdef ADAPTIVE
+ else {
+ // add new ones if needed
+ vec2 full_res_uv = normalized_screen_pos + params.pass_uv_offset.xy;
+ float importance = textureLod(source_importance, full_res_uv, 0.0).x;
+
+ // this is to normalize SSAO_DETAIL_AO_AMOUNT across all pixel regardless of importance
+ obscurance_sum *= (SSAO_ADAPTIVE_TAP_BASE_COUNT / float(SSAO_MAX_TAPS)) + (importance * SSAO_ADAPTIVE_TAP_FLEXIBLE_COUNT / float(SSAO_MAX_TAPS));
+
+ // load existing base values
+ vec2 base_values = imageLoad(source_ssao, ivec3(upos, params.pass)).xy;
+ weight_sum += base_values.y * float(SSAO_ADAPTIVE_TAP_BASE_COUNT * 4.0);
+ obscurance_sum += (base_values.x) * weight_sum;
+
+ // increase importance around edges
+ float edge_count = dot(1.0 - edgesLRTB, vec4(1.0, 1.0, 1.0, 1.0));
+
+ float avg_total_importance = float(counter.sum) * params.load_counter_avg_div;
+
+ float importance_limiter = clamp(params.adaptive_sample_limit / avg_total_importance, 0.0, 1.0);
+ importance *= importance_limiter;
+
+ float additional_sample_count = SSAO_ADAPTIVE_TAP_FLEXIBLE_COUNT * importance;
+
+ const float blend_range = 3.0;
+ const float blend_range_inv = 1.0 / blend_range;
+
+ additional_sample_count += 0.5;
+ uint additional_samples = uint(additional_sample_count);
+ uint additional_samples_to = min(SSAO_MAX_TAPS, additional_samples + SSAO_ADAPTIVE_TAP_BASE_COUNT);
+
+ for (uint i = SSAO_ADAPTIVE_TAP_BASE_COUNT; i < additional_samples_to; i++) {
+ additional_sample_count -= 1.0f;
+ float weight_mod = clamp(additional_sample_count * blend_range_inv, 0.0, 1.0);
+ SSAOTap(p_quality_level, obscurance_sum, weight_sum, int(i), rot_scale_matrix, pix_center_pos, pixel_normal, normalized_screen_pos, mip_offset, fallof_sq, weight_mod, norm_xy, norm_xy_length);
+ }
+ }
#endif
- n_C = normalize(n_C);
- n_C.y = -n_C.y; //because this code reads flipped
- // Hash function used in the HPG12 AlchemyAO paper
- float randomPatternRotationAngle = mod(float((3 * ssC.x ^ ssC.y + ssC.x * ssC.y) * 10), TWO_PI);
+ // early out for adaptive base - just output weight (used for the next pass)
+ if (p_adaptive_base) {
+ float obscurance = obscurance_sum / weight_sum;
+
+ r_shadow_term = obscurance;
+ r_edges = vec4(0.0);
+ r_weight = weight_sum;
+ return;
+ }
+
+ // calculate weighted average
+ float obscurance = obscurance_sum / weight_sum;
- // Reconstruct normals from positions. These will lead to 1-pixel black lines
- // at depth discontinuities, however the blur will wipe those out so they are not visible
- // in the final image.
+ // calculate fadeout (1 close, gradient, 0 far)
+ float fade_out = clamp(pix_center_pos.z * params.fade_out_mul + params.fade_out_add, 0.0, 1.0);
- // Choose the screen-space sample radius
- // proportional to the projected area of the sphere
+ // Reduce the SSAO shadowing if we're on the edge to remove artifacts on edges (we don't care for the lower quality one)
+ if (!p_adaptive_base && (p_quality_level >= SSAO_DEPTH_BASED_EDGES_ENABLE_AT_QUALITY_PRESET)) {
+ // when there's more than 2 opposite edges, start fading out the occlusion to reduce aliasing artifacts
+ float edge_fadeout_factor = clamp((1.0 - edgesLRTB.x - edgesLRTB.y) * 0.35, 0.0, 1.0) + clamp((1.0 - edgesLRTB.z - edgesLRTB.w) * 0.35, 0.0, 1.0);
- float ssDiskRadius = -params.proj_scale * params.radius;
- if (!params.orthogonal) {
- ssDiskRadius = -params.proj_scale * params.radius / C.z;
+ fade_out *= clamp(1.0 - edge_fadeout_factor, 0.0, 1.0);
}
- float sum = 0.0;
- for (int i = 0; i < NUM_SAMPLES; ++i) {
- sum += sampleAO(ssC, C, n_C, ssDiskRadius, params.radius, i, randomPatternRotationAngle);
+
+ // strength
+ obscurance = params.intensity * obscurance;
+
+ // clamp
+ obscurance = min(obscurance, params.shadow_clamp);
+
+ // fadeout
+ obscurance *= fade_out;
+
+ // conceptually switch to occlusion with the meaning being visibility (grows with visibility, occlusion == 1 implies full visibility),
+ // to be in line with what is more commonly used.
+ float occlusion = 1.0 - obscurance;
+
+ // modify the gradient
+ // note: this cannot be moved to a later pass because of loss of precision after storing in the render target
+ occlusion = pow(clamp(occlusion, 0.0, 1.0), params.shadow_power);
+
+ // outputs!
+ r_shadow_term = occlusion; // Our final 'occlusion' term (0 means fully occluded, 1 means fully lit)
+ r_edges = edgesLRTB; // These are used to prevent blurring across edges, 1 means no edge, 0 means edge, 0.5 means half way there, etc.
+ r_weight = weight_sum;
+}
+
+void main() {
+ float out_shadow_term;
+ float out_weight;
+ vec4 out_edges;
+ ivec2 ssC = ivec2(gl_GlobalInvocationID.xy);
+ if (any(greaterThanEqual(ssC, params.screen_size))) { //too large, do nothing
+ return;
}
- float A = max(0.0, 1.0 - sum * params.intensity_div_r6 * (5.0 / float(NUM_SAMPLES)));
+ vec2 uv = vec2(gl_GlobalInvocationID) + vec2(0.5);
+#ifdef SSAO_BASE
+ generate_SSAO_shadows_internal(out_shadow_term, out_edges, out_weight, uv, params.quality, true);
+
+ imageStore(dest_image, ivec2(gl_GlobalInvocationID.xy), vec4(out_shadow_term, out_weight / (float(SSAO_ADAPTIVE_TAP_BASE_COUNT) * 4.0), 0.0, 0.0));
+#else
+ generate_SSAO_shadows_internal(out_shadow_term, out_edges, out_weight, uv, params.quality, false); // pass in quality levels
+ if (params.quality == 0) {
+ out_edges = vec4(1.0);
+ }
- imageStore(dest_image, ssC, vec4(A));
+ imageStore(dest_image, ivec2(gl_GlobalInvocationID.xy), vec4(out_shadow_term, pack_edges(out_edges), 0.0, 0.0));
+#endif
}
diff --git a/servers/rendering/renderer_rd/shaders/ssao_blur.glsl b/servers/rendering/renderer_rd/shaders/ssao_blur.glsl
index 3e63e3cb59..510a777048 100644
--- a/servers/rendering/renderer_rd/shaders/ssao_blur.glsl
+++ b/servers/rendering/renderer_rd/shaders/ssao_blur.glsl
@@ -1,3 +1,22 @@
+///////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+// Copyright (c) 2016, Intel Corporation
+// 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.
+///////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+// File changes (yyyy-mm-dd)
+// 2016-09-07: filip.strugar@intel.com: first commit
+// 2020-12-05: clayjohn: convert to Vulkan and Godot
+///////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+
#[compute]
#version 450
@@ -7,147 +26,129 @@ VERSION_DEFINES
layout(local_size_x = 8, local_size_y = 8, local_size_z = 1) in;
layout(set = 0, binding = 0) uniform sampler2D source_ssao;
-layout(set = 1, binding = 0) uniform sampler2D source_depth;
-#ifdef MODE_UPSCALE
-layout(set = 2, binding = 0) uniform sampler2D source_depth_mipmaps;
-#endif
-layout(r8, set = 3, binding = 0) uniform restrict writeonly image2D dest_image;
-
-//////////////////////////////////////////////////////////////////////////////////////////////
-// Tunable Parameters:
+layout(rg8, set = 1, binding = 0) uniform restrict writeonly image2D dest_image;
layout(push_constant, binding = 1, std430) uniform Params {
- float edge_sharpness; /** Increase to make depth edges crisper. Decrease to reduce flicker. */
- int filter_scale;
- float z_far;
- float z_near;
- bool orthogonal;
- uint pad0;
- uint pad1;
- uint pad2;
- ivec2 axis; /** (1, 0) or (0, 1) */
- ivec2 screen_size;
+ float edge_sharpness;
+ float pad;
+ vec2 half_screen_pixel_size;
}
params;
-/** Filter radius in pixels. This will be multiplied by SCALE. */
-#define R (4)
+vec4 unpack_edges(float p_packed_val) {
+ uint packed_val = uint(p_packed_val * 255.5);
+ vec4 edgesLRTB;
+ edgesLRTB.x = float((packed_val >> 6) & 0x03) / 3.0;
+ edgesLRTB.y = float((packed_val >> 4) & 0x03) / 3.0;
+ edgesLRTB.z = float((packed_val >> 2) & 0x03) / 3.0;
+ edgesLRTB.w = float((packed_val >> 0) & 0x03) / 3.0;
+
+ return clamp(edgesLRTB + params.edge_sharpness, 0.0, 1.0);
+}
+
+void add_sample(float p_ssao_value, float p_edge_value, inout float r_sum, inout float r_sum_weight) {
+ float weight = p_edge_value;
+
+ r_sum += (weight * p_ssao_value);
+ r_sum_weight += weight;
+}
+
+#ifdef MODE_WIDE
+vec2 sample_blurred_wide(vec2 p_coord) {
+ vec2 vC = textureLodOffset(source_ssao, vec2(p_coord), 0.0, ivec2(0, 0)).xy;
+ vec2 vL = textureLodOffset(source_ssao, vec2(p_coord), 0.0, ivec2(-2, 0)).xy;
+ vec2 vT = textureLodOffset(source_ssao, vec2(p_coord), 0.0, ivec2(0, -2)).xy;
+ vec2 vR = textureLodOffset(source_ssao, vec2(p_coord), 0.0, ivec2(2, 0)).xy;
+ vec2 vB = textureLodOffset(source_ssao, vec2(p_coord), 0.0, ivec2(0, 2)).xy;
+
+ float packed_edges = vC.y;
+ vec4 edgesLRTB = unpack_edges(packed_edges);
+ edgesLRTB.x *= unpack_edges(vL.y).y;
+ edgesLRTB.z *= unpack_edges(vT.y).w;
+ edgesLRTB.y *= unpack_edges(vR.y).x;
+ edgesLRTB.w *= unpack_edges(vB.y).z;
+
+ float ssao_value = vC.x;
+ float ssao_valueL = vL.x;
+ float ssao_valueT = vT.x;
+ float ssao_valueR = vR.x;
+ float ssao_valueB = vB.x;
+
+ float sum_weight = 0.8f;
+ float sum = ssao_value * sum_weight;
+
+ add_sample(ssao_valueL, edgesLRTB.x, sum, sum_weight);
+ add_sample(ssao_valueR, edgesLRTB.y, sum, sum_weight);
+ add_sample(ssao_valueT, edgesLRTB.z, sum, sum_weight);
+ add_sample(ssao_valueB, edgesLRTB.w, sum, sum_weight);
+
+ float ssao_avg = sum / sum_weight;
+
+ ssao_value = ssao_avg;
+
+ return vec2(ssao_value, packed_edges);
+}
+#endif
+
+#ifdef MODE_SMART
+vec2 sample_blurred(vec3 p_pos, vec2 p_coord) {
+ float packed_edges = texelFetch(source_ssao, ivec2(p_pos.xy), 0).y;
+ vec4 edgesLRTB = unpack_edges(packed_edges);
+
+ vec4 valuesUL = textureGather(source_ssao, vec2(p_coord - params.half_screen_pixel_size * 0.5));
+ vec4 valuesBR = textureGather(source_ssao, vec2(p_coord + params.half_screen_pixel_size * 0.5));
+
+ float ssao_value = valuesUL.y;
+ float ssao_valueL = valuesUL.x;
+ float ssao_valueT = valuesUL.z;
+ float ssao_valueR = valuesBR.z;
+ float ssao_valueB = valuesBR.x;
+
+ float sum_weight = 0.5;
+ float sum = ssao_value * sum_weight;
+
+ add_sample(ssao_valueL, edgesLRTB.x, sum, sum_weight);
+ add_sample(ssao_valueR, edgesLRTB.y, sum, sum_weight);
+
+ add_sample(ssao_valueT, edgesLRTB.z, sum, sum_weight);
+ add_sample(ssao_valueB, edgesLRTB.w, sum, sum_weight);
-//////////////////////////////////////////////////////////////////////////////////////////////
+ float ssao_avg = sum / sum_weight;
-// Gaussian coefficients
-const float gaussian[R + 1] =
- //float[](0.356642, 0.239400, 0.072410, 0.009869);
- //float[](0.398943, 0.241971, 0.053991, 0.004432, 0.000134); // stddev = 1.0
- float[](0.153170, 0.144893, 0.122649, 0.092902, 0.062970); // stddev = 2.0
-//float[](0.111220, 0.107798, 0.098151, 0.083953, 0.067458, 0.050920, 0.036108); // stddev = 3.0
+ ssao_value = ssao_avg;
+
+ return vec2(ssao_value, packed_edges);
+}
+#endif
void main() {
// Pixel being shaded
ivec2 ssC = ivec2(gl_GlobalInvocationID.xy);
- if (any(greaterThanEqual(ssC, params.screen_size))) { //too large, do nothing
- return;
- }
-
-#ifdef MODE_UPSCALE
-
- //closest one should be the same pixel, but check nearby just in case
- float depth = texelFetch(source_depth, ssC, 0).r;
-
- depth = depth * 2.0 - 1.0;
- if (params.orthogonal) {
- depth = ((depth + (params.z_far + params.z_near) / (params.z_far - params.z_near)) * (params.z_far - params.z_near)) / 2.0;
- } else {
- depth = 2.0 * params.z_near * params.z_far / (params.z_far + params.z_near - depth * (params.z_far - params.z_near));
- }
-
- vec2 pixel_size = 1.0 / vec2(params.screen_size);
- vec2 closest_uv = vec2(ssC) * pixel_size + pixel_size * 0.5;
- vec2 from_uv = closest_uv;
- vec2 ps2 = pixel_size; // * 2.0;
-
- float closest_depth = abs(textureLod(source_depth_mipmaps, closest_uv, 0.0).r - depth);
-
- vec2 offsets[4] = vec2[](vec2(ps2.x, 0), vec2(-ps2.x, 0), vec2(0, ps2.y), vec2(0, -ps2.y));
- for (int i = 0; i < 4; i++) {
- vec2 neighbour = from_uv + offsets[i];
- float neighbour_depth = abs(textureLod(source_depth_mipmaps, neighbour, 0.0).r - depth);
- if (neighbour_depth < closest_depth) {
- closest_uv = neighbour;
- closest_depth = neighbour_depth;
- }
- }
-
- float visibility = textureLod(source_ssao, closest_uv, 0.0).r;
- imageStore(dest_image, ssC, vec4(visibility));
-#else
- float depth = texelFetch(source_depth, ssC, 0).r;
+#ifdef MODE_NON_SMART
-#ifdef MODE_FULL_SIZE
- depth = depth * 2.0 - 1.0;
+ vec2 halfPixel = params.half_screen_pixel_size * 0.5f;
- if (params.orthogonal) {
- depth = ((depth + (params.z_far + params.z_near) / (params.z_far - params.z_near)) * (params.z_far - params.z_near)) / 2.0;
- } else {
- depth = 2.0 * params.z_near * params.z_far / (params.z_far + params.z_near - depth * (params.z_far - params.z_near));
- }
+ vec2 uv = (vec2(gl_GlobalInvocationID.xy) + vec2(0.5, 0.5)) * params.half_screen_pixel_size;
-#endif
- float depth_divide = 1.0 / params.z_far;
-
- //depth *= depth_divide;
-
- /*
- if (depth > params.z_far * 0.999) {
- discard; //skybox
- }
- */
-
- float sum = texelFetch(source_ssao, ssC, 0).r;
-
- // Base weight for depth falloff. Increase this for more blurriness,
- // decrease it for better edge discrimination
- float BASE = gaussian[0];
- float totalWeight = BASE;
- sum *= totalWeight;
-
- ivec2 clamp_limit = params.screen_size - ivec2(1);
-
- for (int r = -R; r <= R; ++r) {
- // We already handled the zero case above. This loop should be unrolled and the static branch optimized out,
- // so the IF statement has no runtime cost
- if (r != 0) {
- ivec2 ppos = ssC + params.axis * (r * params.filter_scale);
- float value = texelFetch(source_ssao, clamp(ppos, ivec2(0), clamp_limit), 0).r;
- ivec2 rpos = clamp(ppos, ivec2(0), clamp_limit);
-
- float temp_depth = texelFetch(source_depth, rpos, 0).r;
-#ifdef MODE_FULL_SIZE
- temp_depth = temp_depth * 2.0 - 1.0;
- if (params.orthogonal) {
- temp_depth = ((temp_depth + (params.z_far + params.z_near) / (params.z_far - params.z_near)) * (params.z_far - params.z_near)) / 2.0;
- } else {
- temp_depth = 2.0 * params.z_near * params.z_far / (params.z_far + params.z_near - temp_depth * (params.z_far - params.z_near));
- }
- //temp_depth *= depth_divide;
-#endif
- // spatial domain: offset gaussian tap
- float weight = 0.3 + gaussian[abs(r)];
- //weight *= max(0.0, dot(temp_normal, normal));
+ vec2 centre = textureLod(source_ssao, vec2(uv), 0.0).xy;
- // range domain (the "bilateral" weight). As depth difference increases, decrease weight.
- weight *= max(0.0, 1.0 - params.edge_sharpness * abs(temp_depth - depth));
+ vec4 vals;
+ vals.x = textureLod(source_ssao, vec2(uv + vec2(-halfPixel.x * 3, -halfPixel.y)), 0.0).x;
+ vals.y = textureLod(source_ssao, vec2(uv + vec2(+halfPixel.x, -halfPixel.y * 3)), 0.0).x;
+ vals.z = textureLod(source_ssao, vec2(uv + vec2(-halfPixel.x, +halfPixel.y * 3)), 0.0).x;
+ vals.w = textureLod(source_ssao, vec2(uv + vec2(+halfPixel.x * 3, +halfPixel.y)), 0.0).x;
- sum += value * weight;
- totalWeight += weight;
- }
- }
+ vec2 sampled = vec2(dot(vals, vec4(0.2)) + centre.x * 0.2, centre.y);
- const float epsilon = 0.0001;
- float visibility = sum / (totalWeight + epsilon);
+#else
+#ifdef MODE_SMART
+ vec2 sampled = sample_blurred(vec3(gl_GlobalInvocationID), (vec2(gl_GlobalInvocationID.xy) + vec2(0.5, 0.5)) * params.half_screen_pixel_size);
+#else // MODE_WIDE
+ vec2 sampled = sample_blurred_wide((vec2(gl_GlobalInvocationID.xy) + vec2(0.5, 0.5)) * params.half_screen_pixel_size);
+#endif
- imageStore(dest_image, ssC, vec4(visibility));
#endif
+ imageStore(dest_image, ivec2(ssC), vec4(sampled, 0.0, 0.0));
}
diff --git a/servers/rendering/renderer_rd/shaders/ssao_downsample.glsl b/servers/rendering/renderer_rd/shaders/ssao_downsample.glsl
new file mode 100644
index 0000000000..cb2d31f70d
--- /dev/null
+++ b/servers/rendering/renderer_rd/shaders/ssao_downsample.glsl
@@ -0,0 +1,206 @@
+///////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+// Copyright (c) 2016, Intel Corporation
+// 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.
+///////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+// File changes (yyyy-mm-dd)
+// 2016-09-07: filip.strugar@intel.com: first commit
+// 2020-12-05: clayjohn: convert to Vulkan and Godot
+///////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+
+#[compute]
+
+#version 450
+
+VERSION_DEFINES
+
+layout(local_size_x = 8, local_size_y = 8, local_size_z = 1) in;
+
+layout(push_constant, binding = 1, std430) uniform Params {
+ vec2 pixel_size;
+ float z_far;
+ float z_near;
+ bool orthogonal;
+ float radius_sq;
+ uvec2 pad;
+}
+params;
+
+layout(set = 0, binding = 0) uniform sampler2D source_depth;
+
+layout(r16f, set = 1, binding = 0) uniform restrict writeonly image2DArray dest_image0; //rename
+#ifdef GENERATE_MIPS
+layout(r16f, set = 2, binding = 0) uniform restrict writeonly image2DArray dest_image1;
+layout(r16f, set = 2, binding = 1) uniform restrict writeonly image2DArray dest_image2;
+layout(r16f, set = 2, binding = 2) uniform restrict writeonly image2DArray dest_image3;
+#endif
+
+vec4 screen_space_to_view_space_depth(vec4 p_depth) {
+ if (params.orthogonal) {
+ vec4 depth = p_depth * 2.0 - 1.0;
+ return ((depth + (params.z_far + params.z_near) / (params.z_far - params.z_near)) * (params.z_far - params.z_near)) / 2.0;
+ }
+
+ float depth_linearize_mul = params.z_near;
+ float depth_linearize_add = params.z_far;
+
+ // Optimised version of "-cameraClipNear / (cameraClipFar - projDepth * (cameraClipFar - cameraClipNear)) * cameraClipFar"
+
+ // Set your depth_linearize_mul and depth_linearize_add to:
+ // depth_linearize_mul = ( cameraClipFar * cameraClipNear) / ( cameraClipFar - cameraClipNear );
+ // depth_linearize_add = cameraClipFar / ( cameraClipFar - cameraClipNear );
+
+ return depth_linearize_mul / (depth_linearize_add - p_depth);
+}
+
+float screen_space_to_view_space_depth(float p_depth) {
+ if (params.orthogonal) {
+ float depth = p_depth * 2.0 - 1.0;
+ return ((depth + (params.z_far + params.z_near) / (params.z_far - params.z_near)) * (params.z_far - params.z_near)) / (2.0 * params.z_far);
+ }
+
+ float depth_linearize_mul = params.z_near;
+ float depth_linearize_add = params.z_far;
+
+ return depth_linearize_mul / (depth_linearize_add - p_depth);
+}
+
+#ifdef GENERATE_MIPS
+
+shared float depth_buffer[4][8][8];
+
+float mip_smart_average(vec4 p_depths) {
+ float closest = min(min(p_depths.x, p_depths.y), min(p_depths.z, p_depths.w));
+ float fallof_sq = -1.0f / params.radius_sq;
+ vec4 dists = p_depths - closest.xxxx;
+ vec4 weights = clamp(dists * dists * fallof_sq + 1.0, 0.0, 1.0);
+ return dot(weights, p_depths) / dot(weights, vec4(1.0, 1.0, 1.0, 1.0));
+}
+
+void prepare_depths_and_mips(vec4 p_samples, uvec2 p_output_coord, uvec2 p_gtid) {
+ p_samples = screen_space_to_view_space_depth(p_samples);
+
+ depth_buffer[0][p_gtid.x][p_gtid.y] = p_samples.w;
+ depth_buffer[1][p_gtid.x][p_gtid.y] = p_samples.z;
+ depth_buffer[2][p_gtid.x][p_gtid.y] = p_samples.x;
+ depth_buffer[3][p_gtid.x][p_gtid.y] = p_samples.y;
+
+ imageStore(dest_image0, ivec3(p_output_coord.x, p_output_coord.y, 0), vec4(p_samples.w));
+ imageStore(dest_image0, ivec3(p_output_coord.x, p_output_coord.y, 1), vec4(p_samples.z));
+ imageStore(dest_image0, ivec3(p_output_coord.x, p_output_coord.y, 2), vec4(p_samples.x));
+ imageStore(dest_image0, ivec3(p_output_coord.x, p_output_coord.y, 3), vec4(p_samples.y));
+
+ uint depth_array_index = 2 * (p_gtid.y % 2) + (p_gtid.x % 2);
+ uvec2 depth_array_offset = ivec2(p_gtid.x % 2, p_gtid.y % 2);
+ ivec2 buffer_coord = ivec2(p_gtid) - ivec2(depth_array_offset);
+
+ p_output_coord /= 2;
+ groupMemoryBarrier();
+ barrier();
+
+ // if (still_alive) <-- all threads alive here
+ {
+ float sample_00 = depth_buffer[depth_array_index][buffer_coord.x + 0][buffer_coord.y + 0];
+ float sample_01 = depth_buffer[depth_array_index][buffer_coord.x + 0][buffer_coord.y + 1];
+ float sample_10 = depth_buffer[depth_array_index][buffer_coord.x + 1][buffer_coord.y + 0];
+ float sample_11 = depth_buffer[depth_array_index][buffer_coord.x + 1][buffer_coord.y + 1];
+
+ float avg = mip_smart_average(vec4(sample_00, sample_01, sample_10, sample_11));
+ imageStore(dest_image1, ivec3(p_output_coord.x, p_output_coord.y, depth_array_index), vec4(avg));
+ depth_buffer[depth_array_index][buffer_coord.x][buffer_coord.y] = avg;
+ }
+
+ bool still_alive = p_gtid.x % 4 == depth_array_offset.x && p_gtid.y % 4 == depth_array_offset.y;
+
+ p_output_coord /= 2;
+ groupMemoryBarrier();
+ barrier();
+
+ if (still_alive) {
+ float sample_00 = depth_buffer[depth_array_index][buffer_coord.x + 0][buffer_coord.y + 0];
+ float sample_01 = depth_buffer[depth_array_index][buffer_coord.x + 0][buffer_coord.y + 2];
+ float sample_10 = depth_buffer[depth_array_index][buffer_coord.x + 2][buffer_coord.y + 0];
+ float sample_11 = depth_buffer[depth_array_index][buffer_coord.x + 2][buffer_coord.y + 2];
+
+ float avg = mip_smart_average(vec4(sample_00, sample_01, sample_10, sample_11));
+ imageStore(dest_image2, ivec3(p_output_coord.x, p_output_coord.y, depth_array_index), vec4(avg));
+ depth_buffer[depth_array_index][buffer_coord.x][buffer_coord.y] = avg;
+ }
+
+ still_alive = p_gtid.x % 8 == depth_array_offset.x && depth_array_offset.y % 8 == depth_array_offset.y;
+
+ p_output_coord /= 2;
+ groupMemoryBarrier();
+ barrier();
+
+ if (still_alive) {
+ float sample_00 = depth_buffer[depth_array_index][buffer_coord.x + 0][buffer_coord.y + 0];
+ float sample_01 = depth_buffer[depth_array_index][buffer_coord.x + 0][buffer_coord.y + 4];
+ float sample_10 = depth_buffer[depth_array_index][buffer_coord.x + 4][buffer_coord.y + 0];
+ float sample_11 = depth_buffer[depth_array_index][buffer_coord.x + 4][buffer_coord.y + 4];
+
+ float avg = mip_smart_average(vec4(sample_00, sample_01, sample_10, sample_11));
+ imageStore(dest_image3, ivec3(p_output_coord.x, p_output_coord.y, depth_array_index), vec4(avg));
+ }
+}
+#else
+#ifndef USE_HALF_BUFFERS
+void prepare_depths(vec4 p_samples, uvec2 p_tid) {
+ p_samples = screen_space_to_view_space_depth(p_samples);
+
+ imageStore(dest_image0, ivec3(p_tid, 0), vec4(p_samples.w));
+ imageStore(dest_image0, ivec3(p_tid, 1), vec4(p_samples.z));
+ imageStore(dest_image0, ivec3(p_tid, 2), vec4(p_samples.x));
+ imageStore(dest_image0, ivec3(p_tid, 3), vec4(p_samples.y));
+}
+#endif
+#endif
+
+void main() {
+#ifdef USE_HALF_BUFFERS
+#ifdef USE_HALF_SIZE
+ float sample_00 = texelFetch(source_depth, ivec2(4 * gl_GlobalInvocationID.x + 0, 4 * gl_GlobalInvocationID.y + 0), 0).x;
+ float sample_11 = texelFetch(source_depth, ivec2(4 * gl_GlobalInvocationID.x + 2, 4 * gl_GlobalInvocationID.y + 2), 0).x;
+#else
+ float sample_00 = texelFetch(source_depth, ivec2(2 * gl_GlobalInvocationID.x + 0, 2 * gl_GlobalInvocationID.y + 0), 0).x;
+ float sample_11 = texelFetch(source_depth, ivec2(2 * gl_GlobalInvocationID.x + 1, 2 * gl_GlobalInvocationID.y + 1), 0).x;
+#endif
+ sample_00 = screen_space_to_view_space_depth(sample_00);
+ sample_11 = screen_space_to_view_space_depth(sample_11);
+
+ imageStore(dest_image0, ivec3(gl_GlobalInvocationID.xy, 0), vec4(sample_00));
+ imageStore(dest_image0, ivec3(gl_GlobalInvocationID.xy, 3), vec4(sample_11));
+#else //!USE_HALF_BUFFERS
+#ifdef USE_HALF_SIZE
+ ivec2 depth_buffer_coord = 4 * ivec2(gl_GlobalInvocationID.xy);
+ ivec2 output_coord = ivec2(gl_GlobalInvocationID);
+
+ vec2 uv = (vec2(depth_buffer_coord) + 0.5f) * params.pixel_size;
+ vec4 samples;
+ samples.x = textureLodOffset(source_depth, uv, 0, ivec2(0, 2)).x;
+ samples.y = textureLodOffset(source_depth, uv, 0, ivec2(2, 2)).x;
+ samples.z = textureLodOffset(source_depth, uv, 0, ivec2(2, 0)).x;
+ samples.w = textureLodOffset(source_depth, uv, 0, ivec2(0, 0)).x;
+#else
+ ivec2 depth_buffer_coord = 2 * ivec2(gl_GlobalInvocationID.xy);
+ ivec2 output_coord = ivec2(gl_GlobalInvocationID);
+
+ vec2 uv = (vec2(depth_buffer_coord) + 0.5f) * params.pixel_size;
+ vec4 samples = textureGather(source_depth, uv);
+#endif
+#ifdef GENERATE_MIPS
+ prepare_depths_and_mips(samples, output_coord, gl_LocalInvocationID.xy);
+#else
+ prepare_depths(samples, gl_GlobalInvocationID.xy);
+#endif
+#endif
+}
diff --git a/servers/rendering/renderer_rd/shaders/ssao_importance_map.glsl b/servers/rendering/renderer_rd/shaders/ssao_importance_map.glsl
new file mode 100644
index 0000000000..6aa7624261
--- /dev/null
+++ b/servers/rendering/renderer_rd/shaders/ssao_importance_map.glsl
@@ -0,0 +1,126 @@
+///////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+// Copyright (c) 2016, Intel Corporation
+// 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.
+///////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+// File changes (yyyy-mm-dd)
+// 2016-09-07: filip.strugar@intel.com: first commit
+// 2020-12-05: clayjohn: convert to Vulkan and Godot
+///////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+
+#[compute]
+
+#version 450
+
+VERSION_DEFINES
+
+layout(local_size_x = 8, local_size_y = 8, local_size_z = 1) in;
+
+#ifdef GENERATE_MAP
+layout(set = 0, binding = 0) uniform sampler2DArray source_ssao;
+#else
+layout(set = 0, binding = 0) uniform sampler2D source_importance;
+#endif
+layout(r8, set = 1, binding = 0) uniform restrict writeonly image2D dest_image;
+
+#ifdef PROCESS_MAPB
+layout(set = 2, binding = 0, std430) buffer Counter {
+ uint sum;
+}
+counter;
+#endif
+
+layout(push_constant, binding = 1, std430) uniform Params {
+ vec2 half_screen_pixel_size;
+ float intensity;
+ float power;
+}
+params;
+
+void main() {
+ // Pixel being shaded
+ ivec2 ssC = ivec2(gl_GlobalInvocationID.xy);
+
+#ifdef GENERATE_MAP
+ // importance map stuff
+ uvec2 base_position = ssC * 2;
+
+ vec2 base_uv = (vec2(base_position) + vec2(0.5f, 0.5f)) * params.half_screen_pixel_size;
+
+ float avg = 0.0;
+ float minV = 1.0;
+ float maxV = 0.0;
+ for (int i = 0; i < 4; i++) {
+ vec4 vals = textureGather(source_ssao, vec3(base_uv, i));
+
+ // apply the same modifications that would have been applied in the main shader
+ vals = params.intensity * vals;
+
+ vals = 1 - vals;
+
+ vals = pow(clamp(vals, 0.0, 1.0), vec4(params.power));
+
+ avg += dot(vec4(vals.x, vals.y, vals.z, vals.w), vec4(1.0 / 16.0, 1.0 / 16.0, 1.0 / 16.0, 1.0 / 16.0));
+
+ maxV = max(maxV, max(max(vals.x, vals.y), max(vals.z, vals.w)));
+ minV = min(minV, min(min(vals.x, vals.y), min(vals.z, vals.w)));
+ }
+
+ float min_max_diff = maxV - minV;
+
+ imageStore(dest_image, ssC, vec4(pow(clamp(min_max_diff * 2.0, 0.0, 1.0), 0.8)));
+#endif
+
+#ifdef PROCESS_MAPA
+ vec2 uv = (vec2(ssC) + 0.5f) * params.half_screen_pixel_size * 2.0;
+
+ float centre = textureLod(source_importance, uv, 0.0).x;
+
+ vec2 half_pixel = params.half_screen_pixel_size;
+
+ vec4 vals;
+ vals.x = textureLod(source_importance, uv + vec2(-half_pixel.x * 3, -half_pixel.y), 0.0).x;
+ vals.y = textureLod(source_importance, uv + vec2(+half_pixel.x, -half_pixel.y * 3), 0.0).x;
+ vals.z = textureLod(source_importance, uv + vec2(+half_pixel.x * 3, +half_pixel.y), 0.0).x;
+ vals.w = textureLod(source_importance, uv + vec2(-half_pixel.x, +half_pixel.y * 3), 0.0).x;
+
+ float avg = dot(vals, vec4(0.25, 0.25, 0.25, 0.25));
+
+ imageStore(dest_image, ssC, vec4(avg));
+#endif
+
+#ifdef PROCESS_MAPB
+ vec2 uv = (vec2(ssC) + 0.5f) * params.half_screen_pixel_size * 2.0;
+
+ float centre = textureLod(source_importance, uv, 0.0).x;
+
+ vec2 half_pixel = params.half_screen_pixel_size;
+
+ vec4 vals;
+ vals.x = textureLod(source_importance, uv + vec2(-half_pixel.x, -half_pixel.y * 3), 0.0).x;
+ vals.y = textureLod(source_importance, uv + vec2(+half_pixel.x * 3, -half_pixel.y), 0.0).x;
+ vals.z = textureLod(source_importance, uv + vec2(+half_pixel.x, +half_pixel.y * 3), 0.0).x;
+ vals.w = textureLod(source_importance, uv + vec2(-half_pixel.x * 3, +half_pixel.y), 0.0).x;
+
+ float avg = dot(vals, vec4(0.25, 0.25, 0.25, 0.25));
+
+ imageStore(dest_image, ssC, vec4(avg));
+
+ // sum the average; to avoid overflowing we assume max AO resolution is not bigger than 16384x16384; so quarter res (used here) will be 4096x4096, which leaves us with 8 bits per pixel
+ uint sum = uint(clamp(avg, 0.0, 1.0) * 255.0 + 0.5);
+
+ // save every 9th to avoid InterlockedAdd congestion - since we're blurring, this is good enough; compensated by multiplying load_counter_avg_div by 9
+ if (((ssC.x % 3) + (ssC.y % 3)) == 0) {
+ atomicAdd(counter.sum, sum);
+ }
+#endif
+}
diff --git a/servers/rendering/renderer_rd/shaders/ssao_interleave.glsl b/servers/rendering/renderer_rd/shaders/ssao_interleave.glsl
new file mode 100644
index 0000000000..4fdf334aa5
--- /dev/null
+++ b/servers/rendering/renderer_rd/shaders/ssao_interleave.glsl
@@ -0,0 +1,119 @@
+///////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+// Copyright (c) 2016, Intel Corporation
+// 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.
+///////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+// File changes (yyyy-mm-dd)
+// 2016-09-07: filip.strugar@intel.com: first commit
+// 2020-12-05: clayjohn: convert to Vulkan and Godot
+///////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+#[compute]
+
+#version 450
+
+VERSION_DEFINES
+
+layout(local_size_x = 8, local_size_y = 8, local_size_z = 1) in;
+
+layout(rgba8, set = 0, binding = 0) uniform restrict writeonly image2D dest_image;
+layout(set = 1, binding = 0) uniform sampler2DArray source_texture;
+
+layout(push_constant, binding = 1, std430) uniform Params {
+ float inv_sharpness;
+ uint size_modifier;
+ vec2 pixel_size;
+}
+params;
+
+vec4 unpack_edges(float p_packed_val) {
+ uint packed_val = uint(p_packed_val * 255.5);
+ vec4 edgesLRTB;
+ edgesLRTB.x = float((packed_val >> 6) & 0x03) / 3.0;
+ edgesLRTB.y = float((packed_val >> 4) & 0x03) / 3.0;
+ edgesLRTB.z = float((packed_val >> 2) & 0x03) / 3.0;
+ edgesLRTB.w = float((packed_val >> 0) & 0x03) / 3.0;
+
+ return clamp(edgesLRTB + params.inv_sharpness, 0.0, 1.0);
+}
+
+void main() {
+ ivec2 ssC = ivec2(gl_GlobalInvocationID.xy);
+ if (any(greaterThanEqual(ssC, ivec2(1.0 / params.pixel_size)))) { //too large, do nothing
+ return;
+ }
+
+#ifdef MODE_SMART
+ float ao;
+ uvec2 pix_pos = uvec2(gl_GlobalInvocationID.xy);
+ vec2 uv = (gl_GlobalInvocationID.xy + vec2(0.5)) * params.pixel_size;
+
+ // calculate index in the four deinterleaved source array texture
+ int mx = int(pix_pos.x % 2);
+ int my = int(pix_pos.y % 2);
+ int index_center = mx + my * 2; // center index
+ int index_horizontal = (1 - mx) + my * 2; // neighbouring, horizontal
+ int index_vertical = mx + (1 - my) * 2; // neighbouring, vertical
+ int index_diagonal = (1 - mx) + (1 - my) * 2; // diagonal
+
+ vec2 center_val = texelFetch(source_texture, ivec3(pix_pos / uvec2(params.size_modifier), index_center), 0).xy;
+
+ ao = center_val.x;
+
+ vec4 edgesLRTB = unpack_edges(center_val.y);
+
+ // convert index shifts to sampling offsets
+ float fmx = float(mx);
+ float fmy = float(my);
+
+ // in case of an edge, push sampling offsets away from the edge (towards pixel center)
+ float fmxe = (edgesLRTB.y - edgesLRTB.x);
+ float fmye = (edgesLRTB.w - edgesLRTB.z);
+
+ // calculate final sampling offsets and sample using bilinear filter
+ vec2 uv_horizontal = (gl_GlobalInvocationID.xy + vec2(0.5) + vec2(fmx + fmxe - 0.5, 0.5 - fmy)) * params.pixel_size;
+ float ao_horizontal = textureLod(source_texture, vec3(uv_horizontal, index_horizontal), 0.0).x;
+ vec2 uv_vertical = (gl_GlobalInvocationID.xy + vec2(0.5) + vec2(0.5 - fmx, fmy - 0.5 + fmye)) * params.pixel_size;
+ float ao_vertical = textureLod(source_texture, vec3(uv_vertical, index_vertical), 0.0).x;
+ vec2 uv_diagonal = (gl_GlobalInvocationID.xy + vec2(0.5) + vec2(fmx - 0.5 + fmxe, fmy - 0.5 + fmye)) * params.pixel_size;
+ float ao_diagonal = textureLod(source_texture, vec3(uv_diagonal, index_diagonal), 0.0).x;
+
+ // reduce weight for samples near edge - if the edge is on both sides, weight goes to 0
+ vec4 blendWeights;
+ blendWeights.x = 1.0;
+ blendWeights.y = (edgesLRTB.x + edgesLRTB.y) * 0.5;
+ blendWeights.z = (edgesLRTB.z + edgesLRTB.w) * 0.5;
+ blendWeights.w = (blendWeights.y + blendWeights.z) * 0.5;
+
+ // calculate weighted average
+ float blendWeightsSum = dot(blendWeights, vec4(1.0, 1.0, 1.0, 1.0));
+ ao = dot(vec4(ao, ao_horizontal, ao_vertical, ao_diagonal), blendWeights) / blendWeightsSum;
+
+ imageStore(dest_image, ivec2(gl_GlobalInvocationID.xy), vec4(ao));
+#else // !MODE_SMART
+
+ vec2 uv = (gl_GlobalInvocationID.xy + vec2(0.5)) * params.pixel_size;
+#ifdef MODE_HALF
+ float a = textureLod(source_texture, vec3(uv, 0), 0.0).x;
+ float d = textureLod(source_texture, vec3(uv, 3), 0.0).x;
+ float avg = (a + d) * 0.5;
+
+#else
+ float a = textureLod(source_texture, vec3(uv, 0), 0.0).x;
+ float b = textureLod(source_texture, vec3(uv, 1), 0.0).x;
+ float c = textureLod(source_texture, vec3(uv, 2), 0.0).x;
+ float d = textureLod(source_texture, vec3(uv, 3), 0.0).x;
+ float avg = (a + b + c + d) * 0.25;
+
+#endif
+ imageStore(dest_image, ivec2(gl_GlobalInvocationID.xy), vec4(avg));
+#endif
+}
diff --git a/servers/rendering/renderer_rd/shaders/ssao_minify.glsl b/servers/rendering/renderer_rd/shaders/ssao_minify.glsl
deleted file mode 100644
index 263fca386f..0000000000
--- a/servers/rendering/renderer_rd/shaders/ssao_minify.glsl
+++ /dev/null
@@ -1,45 +0,0 @@
-#[compute]
-
-#version 450
-
-VERSION_DEFINES
-
-layout(local_size_x = 8, local_size_y = 8, local_size_z = 1) in;
-
-layout(push_constant, binding = 1, std430) uniform Params {
- vec2 pixel_size;
- float z_far;
- float z_near;
- ivec2 source_size;
- bool orthogonal;
- uint pad;
-}
-params;
-
-#ifdef MINIFY_START
-layout(set = 0, binding = 0) uniform sampler2D source_texture;
-#else
-layout(r32f, set = 0, binding = 0) uniform restrict readonly image2D source_image;
-#endif
-layout(r32f, set = 1, binding = 0) uniform restrict writeonly image2D dest_image;
-
-void main() {
- ivec2 pos = ivec2(gl_GlobalInvocationID.xy);
-
- if (any(greaterThan(pos, params.source_size >> 1))) { //too large, do nothing
- return;
- }
-
-#ifdef MINIFY_START
- float depth = texelFetch(source_texture, pos << 1, 0).r * 2.0 - 1.0;
- if (params.orthogonal) {
- depth = ((depth + (params.z_far + params.z_near) / (params.z_far - params.z_near)) * (params.z_far - params.z_near)) / 2.0;
- } else {
- depth = 2.0 * params.z_near * params.z_far / (params.z_far + params.z_near - depth * (params.z_far - params.z_near));
- }
-#else
- float depth = imageLoad(source_image, pos << 1).r;
-#endif
-
- imageStore(dest_image, pos, vec4(depth));
-}
diff --git a/servers/rendering/renderer_scene.h b/servers/rendering/renderer_scene.h
index 56c38beaa3..22af720ae7 100644
--- a/servers/rendering/renderer_scene.h
+++ b/servers/rendering/renderer_scene.h
@@ -132,9 +132,9 @@ public:
virtual void 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) = 0;
virtual void environment_set_ssr_roughness_quality(RS::EnvironmentSSRRoughnessQuality p_quality) = 0;
- virtual void environment_set_ssao(RID p_env, bool p_enable, float p_radius, float p_intensity, float p_bias, float p_light_affect, float p_ao_channel_affect, RS::EnvironmentSSAOBlur p_blur, float p_bilateral_sharpness) = 0;
+ virtual void 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) = 0;
- virtual void environment_set_ssao_quality(RS::EnvironmentSSAOQuality p_quality, bool p_half_size) = 0;
+ virtual void 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) = 0;
virtual void environment_set_sdfgi(RID p_env, bool p_enable, RS::EnvironmentSDFGICascades p_cascades, float p_min_cell_size, RS::EnvironmentSDFGIYScale p_y_scale, bool p_use_occlusion, bool p_use_multibounce, bool p_read_sky, float p_energy, float p_normal_bias, float p_probe_bias) = 0;
diff --git a/servers/rendering/renderer_scene_cull.cpp b/servers/rendering/renderer_scene_cull.cpp
index 88a0859a28..3ec2a4c9c8 100644
--- a/servers/rendering/renderer_scene_cull.cpp
+++ b/servers/rendering/renderer_scene_cull.cpp
@@ -30,6 +30,7 @@
#include "renderer_scene_cull.h"
+#include "core/config/project_settings.h"
#include "core/os/os.h"
#include "rendering_server_default.h"
#include "rendering_server_globals.h"
@@ -108,8 +109,8 @@ bool RendererSceneCull::is_camera(RID p_camera) const {
/* SCENARIO API */
-void *RendererSceneCull::_instance_pair(void *p_self, OctreeElementID, Instance *p_A, int, OctreeElementID, Instance *p_B, int) {
- //RendererSceneCull *self = (RendererSceneCull*)p_self;
+void RendererSceneCull::_instance_pair(Instance *p_A, Instance *p_B) {
+ RendererSceneCull *self = (RendererSceneCull *)singleton;
Instance *A = p_A;
Instance *B = p_B;
@@ -122,90 +123,84 @@ void *RendererSceneCull::_instance_pair(void *p_self, OctreeElementID, Instance
InstanceLightData *light = static_cast<InstanceLightData *>(B->base_data);
InstanceGeometryData *geom = static_cast<InstanceGeometryData *>(A->base_data);
- InstanceLightData::PairInfo pinfo;
- pinfo.geometry = A;
- pinfo.L = geom->lighting.push_back(B);
-
- List<InstanceLightData::PairInfo>::Element *E = light->geometries.push_back(pinfo);
+ geom->lights.insert(B);
+ light->geometries.insert(A);
if (geom->can_cast_shadows) {
light->shadow_dirty = true;
}
- geom->lighting_dirty = true;
- return E; //this element should make freeing faster
- } else if (B->base_type == RS::INSTANCE_REFLECTION_PROBE && ((1 << A->base_type) & RS::INSTANCE_GEOMETRY_MASK)) {
+ if (A->scenario && A->array_index >= 0) {
+ InstanceData &idata = A->scenario->instance_data[A->array_index];
+ idata.flags |= InstanceData::FLAG_GEOM_LIGHTING_DIRTY;
+ }
+
+ } else if (self->pair_volumes_to_mesh && B->base_type == RS::INSTANCE_REFLECTION_PROBE && ((1 << A->base_type) & RS::INSTANCE_GEOMETRY_MASK)) {
InstanceReflectionProbeData *reflection_probe = static_cast<InstanceReflectionProbeData *>(B->base_data);
InstanceGeometryData *geom = static_cast<InstanceGeometryData *>(A->base_data);
- InstanceReflectionProbeData::PairInfo pinfo;
- pinfo.geometry = A;
- pinfo.L = geom->reflection_probes.push_back(B);
-
- List<InstanceReflectionProbeData::PairInfo>::Element *E = reflection_probe->geometries.push_back(pinfo);
+ geom->reflection_probes.insert(B);
+ reflection_probe->geometries.insert(A);
- geom->reflection_dirty = true;
+ if (A->scenario && A->array_index >= 0) {
+ InstanceData &idata = A->scenario->instance_data[A->array_index];
+ idata.flags |= InstanceData::FLAG_GEOM_REFLECTION_DIRTY;
+ }
- return E; //this element should make freeing faster
- } else if (B->base_type == RS::INSTANCE_DECAL && ((1 << A->base_type) & RS::INSTANCE_GEOMETRY_MASK)) {
+ } else if (self->pair_volumes_to_mesh && B->base_type == RS::INSTANCE_DECAL && ((1 << A->base_type) & RS::INSTANCE_GEOMETRY_MASK)) {
InstanceDecalData *decal = static_cast<InstanceDecalData *>(B->base_data);
InstanceGeometryData *geom = static_cast<InstanceGeometryData *>(A->base_data);
- InstanceDecalData::PairInfo pinfo;
- pinfo.geometry = A;
- pinfo.L = geom->decals.push_back(B);
-
- List<InstanceDecalData::PairInfo>::Element *E = decal->geometries.push_back(pinfo);
+ geom->decals.insert(B);
+ decal->geometries.insert(A);
- geom->decal_dirty = true;
+ if (A->scenario && A->array_index >= 0) {
+ InstanceData &idata = A->scenario->instance_data[A->array_index];
+ idata.flags |= InstanceData::FLAG_GEOM_DECAL_DIRTY;
+ }
- return E; //this element should make freeing faster
} else if (B->base_type == RS::INSTANCE_LIGHTMAP && ((1 << A->base_type) & RS::INSTANCE_GEOMETRY_MASK)) {
InstanceLightmapData *lightmap_data = static_cast<InstanceLightmapData *>(B->base_data);
InstanceGeometryData *geom = static_cast<InstanceGeometryData *>(A->base_data);
if (A->dynamic_gi) {
- InstanceLightmapData::PairInfo pinfo;
- pinfo.geometry = A;
- pinfo.L = geom->lightmap_captures.push_back(B);
- List<InstanceLightmapData::PairInfo>::Element *E = lightmap_data->geometries.push_back(pinfo);
- ((RendererSceneCull *)p_self)->_instance_queue_update(A, false, false); //need to update capture
- return E; //this element should make freeing faster
- } else {
- return nullptr;
+ geom->lightmap_captures.insert(A);
+ lightmap_data->geometries.insert(B);
+
+ if (A->scenario && A->array_index >= 0) {
+ InstanceData &idata = A->scenario->instance_data[A->array_index];
+ idata.flags |= InstanceData::FLAG_LIGHTMAP_CAPTURE;
+ }
+ ((RendererSceneCull *)self)->_instance_queue_update(A, false, false); //need to update capture
}
} else if (B->base_type == RS::INSTANCE_GI_PROBE && ((1 << A->base_type) & RS::INSTANCE_GEOMETRY_MASK)) {
InstanceGIProbeData *gi_probe = static_cast<InstanceGIProbeData *>(B->base_data);
InstanceGeometryData *geom = static_cast<InstanceGeometryData *>(A->base_data);
- InstanceGIProbeData::PairInfo pinfo;
- pinfo.geometry = A;
- pinfo.L = geom->gi_probes.push_back(B);
+ geom->gi_probes.insert(B);
- List<InstanceGIProbeData::PairInfo>::Element *E;
if (A->dynamic_gi) {
- E = gi_probe->dynamic_geometries.push_back(pinfo);
+ gi_probe->dynamic_geometries.insert(A);
} else {
- E = gi_probe->geometries.push_back(pinfo);
+ gi_probe->geometries.insert(A);
}
- geom->gi_probes_dirty = true;
-
- return E; //this element should make freeing faster
+ if (A->scenario && A->array_index >= 0) {
+ InstanceData &idata = A->scenario->instance_data[A->array_index];
+ idata.flags |= InstanceData::FLAG_GEOM_GI_PROBE_DIRTY;
+ }
} else if (B->base_type == RS::INSTANCE_GI_PROBE && A->base_type == RS::INSTANCE_LIGHT) {
InstanceGIProbeData *gi_probe = static_cast<InstanceGIProbeData *>(B->base_data);
- return gi_probe->lights.insert(A);
+ gi_probe->lights.insert(A);
} else if (B->base_type == RS::INSTANCE_PARTICLES_COLLISION && A->base_type == RS::INSTANCE_PARTICLES) {
RSG::storage->particles_add_collision(A->base, B);
}
-
- return nullptr;
}
-void RendererSceneCull::_instance_unpair(void *p_self, OctreeElementID, Instance *p_A, int, OctreeElementID, Instance *p_B, int, void *udata) {
- //RendererSceneCull *self = (RendererSceneCull*)p_self;
+void RendererSceneCull::_instance_unpair(Instance *p_A, Instance *p_B) {
+ RendererSceneCull *self = (RendererSceneCull *)singleton;
Instance *A = p_A;
Instance *B = p_B;
@@ -218,68 +213,76 @@ void RendererSceneCull::_instance_unpair(void *p_self, OctreeElementID, Instance
InstanceLightData *light = static_cast<InstanceLightData *>(B->base_data);
InstanceGeometryData *geom = static_cast<InstanceGeometryData *>(A->base_data);
- List<InstanceLightData::PairInfo>::Element *E = reinterpret_cast<List<InstanceLightData::PairInfo>::Element *>(udata);
-
- geom->lighting.erase(E->get().L);
- light->geometries.erase(E);
+ geom->lights.erase(B);
+ light->geometries.erase(A);
if (geom->can_cast_shadows) {
light->shadow_dirty = true;
}
- geom->lighting_dirty = true;
- } else if (B->base_type == RS::INSTANCE_REFLECTION_PROBE && ((1 << A->base_type) & RS::INSTANCE_GEOMETRY_MASK)) {
+ if (A->scenario && A->array_index >= 0) {
+ InstanceData &idata = A->scenario->instance_data[A->array_index];
+ idata.flags |= InstanceData::FLAG_GEOM_LIGHTING_DIRTY;
+ }
+
+ } else if (self->pair_volumes_to_mesh && B->base_type == RS::INSTANCE_REFLECTION_PROBE && ((1 << A->base_type) & RS::INSTANCE_GEOMETRY_MASK)) {
InstanceReflectionProbeData *reflection_probe = static_cast<InstanceReflectionProbeData *>(B->base_data);
InstanceGeometryData *geom = static_cast<InstanceGeometryData *>(A->base_data);
- List<InstanceReflectionProbeData::PairInfo>::Element *E = reinterpret_cast<List<InstanceReflectionProbeData::PairInfo>::Element *>(udata);
+ geom->reflection_probes.erase(B);
+ reflection_probe->geometries.erase(A);
- geom->reflection_probes.erase(E->get().L);
- reflection_probe->geometries.erase(E);
+ if (A->scenario && A->array_index >= 0) {
+ InstanceData &idata = A->scenario->instance_data[A->array_index];
+ idata.flags |= InstanceData::FLAG_GEOM_REFLECTION_DIRTY;
+ }
- geom->reflection_dirty = true;
- } else if (B->base_type == RS::INSTANCE_DECAL && ((1 << A->base_type) & RS::INSTANCE_GEOMETRY_MASK)) {
+ } else if (self->pair_volumes_to_mesh && B->base_type == RS::INSTANCE_DECAL && ((1 << A->base_type) & RS::INSTANCE_GEOMETRY_MASK)) {
InstanceDecalData *decal = static_cast<InstanceDecalData *>(B->base_data);
InstanceGeometryData *geom = static_cast<InstanceGeometryData *>(A->base_data);
- List<InstanceDecalData::PairInfo>::Element *E = reinterpret_cast<List<InstanceDecalData::PairInfo>::Element *>(udata);
+ geom->decals.erase(B);
+ decal->geometries.erase(A);
- geom->decals.erase(E->get().L);
- decal->geometries.erase(E);
+ if (A->scenario && A->array_index >= 0) {
+ InstanceData &idata = A->scenario->instance_data[A->array_index];
+ idata.flags |= InstanceData::FLAG_GEOM_DECAL_DIRTY;
+ }
- geom->decal_dirty = true;
} else if (B->base_type == RS::INSTANCE_LIGHTMAP && ((1 << A->base_type) & RS::INSTANCE_GEOMETRY_MASK)) {
- if (udata) { //only for dynamic geometries
- InstanceLightmapData *lightmap_data = static_cast<InstanceLightmapData *>(B->base_data);
- InstanceGeometryData *geom = static_cast<InstanceGeometryData *>(A->base_data);
+ InstanceLightmapData *lightmap_data = static_cast<InstanceLightmapData *>(B->base_data);
+ InstanceGeometryData *geom = static_cast<InstanceGeometryData *>(A->base_data);
+ if (A->dynamic_gi) {
+ geom->lightmap_captures.erase(B);
- List<InstanceLightmapData::PairInfo>::Element *E = reinterpret_cast<List<InstanceLightmapData::PairInfo>::Element *>(udata);
+ if (geom->lightmap_captures.is_empty() && A->scenario && A->array_index >= 0) {
+ InstanceData &idata = A->scenario->instance_data[A->array_index];
+ idata.flags &= ~uint32_t(InstanceData::FLAG_LIGHTMAP_CAPTURE);
+ }
- geom->lightmap_captures.erase(E->get().L);
- lightmap_data->geometries.erase(E);
- ((RendererSceneCull *)p_self)->_instance_queue_update(A, false, false); //need to update capture
+ lightmap_data->geometries.erase(A);
+ ((RendererSceneCull *)self)->_instance_queue_update(A, false, false); //need to update capture
}
} else if (B->base_type == RS::INSTANCE_GI_PROBE && ((1 << A->base_type) & RS::INSTANCE_GEOMETRY_MASK)) {
InstanceGIProbeData *gi_probe = static_cast<InstanceGIProbeData *>(B->base_data);
InstanceGeometryData *geom = static_cast<InstanceGeometryData *>(A->base_data);
- List<InstanceGIProbeData::PairInfo>::Element *E = reinterpret_cast<List<InstanceGIProbeData::PairInfo>::Element *>(udata);
-
- geom->gi_probes.erase(E->get().L);
+ geom->gi_probes.erase(B);
if (A->dynamic_gi) {
- gi_probe->dynamic_geometries.erase(E);
+ gi_probe->dynamic_geometries.erase(A);
} else {
- gi_probe->geometries.erase(E);
+ gi_probe->geometries.erase(A);
}
- geom->gi_probes_dirty = true;
+ if (A->scenario && A->array_index >= 0) {
+ InstanceData &idata = A->scenario->instance_data[A->array_index];
+ idata.flags |= InstanceData::FLAG_GEOM_GI_PROBE_DIRTY;
+ }
} else if (B->base_type == RS::INSTANCE_GI_PROBE && A->base_type == RS::INSTANCE_LIGHT) {
InstanceGIProbeData *gi_probe = static_cast<InstanceGIProbeData *>(B->base_data);
- Set<Instance *>::Element *E = reinterpret_cast<Set<Instance *>::Element *>(udata);
-
- gi_probe->lights.erase(E);
+ gi_probe->lights.erase(A);
} else if (B->base_type == RS::INSTANCE_PARTICLES_COLLISION && A->base_type == RS::INSTANCE_PARTICLES) {
RSG::storage->particles_remove_collision(A->base, B);
}
@@ -291,8 +294,6 @@ RID RendererSceneCull::scenario_create() {
RID scenario_rid = scenario_owner.make_rid(scenario);
scenario->self = scenario_rid;
- scenario->octree.set_pair_callback(_instance_pair, this);
- scenario->octree.set_unpair_callback(_instance_unpair, this);
scenario->reflection_probe_shadow_atlas = scene_render->shadow_atlas_create();
scene_render->shadow_atlas_set_size(scenario->reflection_probe_shadow_atlas, 1024); //make enough shadows for close distance, don't bother with rest
scene_render->shadow_atlas_set_quadrant_subdivision(scenario->reflection_probe_shadow_atlas, 0, 4);
@@ -300,6 +301,10 @@ RID RendererSceneCull::scenario_create() {
scene_render->shadow_atlas_set_quadrant_subdivision(scenario->reflection_probe_shadow_atlas, 2, 4);
scene_render->shadow_atlas_set_quadrant_subdivision(scenario->reflection_probe_shadow_atlas, 3, 8);
scenario->reflection_atlas = scene_render->reflection_atlas_create();
+
+ scenario->instance_aabbs.set_page_pool(&instance_aabb_page_pool);
+ scenario->instance_data.set_page_pool(&instance_data_page_pool);
+
return scenario_rid;
}
@@ -375,10 +380,20 @@ void RendererSceneCull::_instance_update_mesh_instance(Instance *p_instance) {
if (needs_instance != p_instance->mesh_instance.is_valid()) {
if (needs_instance) {
p_instance->mesh_instance = RSG::storage->mesh_instance_create(p_instance->base);
+
} else {
RSG::storage->free(p_instance->mesh_instance);
p_instance->mesh_instance = RID();
}
+
+ if (p_instance->scenario && p_instance->array_index >= 0) {
+ InstanceData &idata = p_instance->scenario->instance_data[p_instance->array_index];
+ if (p_instance->mesh_instance.is_valid()) {
+ idata.flags |= InstanceData::FLAG_USES_MESH_INSTANCE;
+ } else {
+ idata.flags &= ~uint32_t(InstanceData::FLAG_USES_MESH_INSTANCE);
+ }
+ }
}
if (p_instance->mesh_instance.is_valid()) {
@@ -395,14 +410,14 @@ void RendererSceneCull::instance_set_base(RID p_instance, RID p_base) {
if (instance->base_type != RS::INSTANCE_NONE) {
//free anything related to that base
- if (scenario && instance->octree_id) {
- scenario->octree.erase(instance->octree_id); //make dependencies generated by the octree go away
- instance->octree_id = 0;
+ if (scenario && instance->indexer_id.is_valid()) {
+ _unpair_instance(instance);
}
if (instance->mesh_instance.is_valid()) {
RSG::storage->free(instance->mesh_instance);
instance->mesh_instance = RID();
+ // no need to set instance data flag here, as it was freed above
}
switch (instance->base_type) {
@@ -556,9 +571,8 @@ void RendererSceneCull::instance_set_scenario(RID p_instance, RID p_scenario) {
if (instance->scenario) {
instance->scenario->instances.remove(&instance->scenario_item);
- if (instance->octree_id) {
- instance->scenario->octree.erase(instance->octree_id); //make dependencies generated by the octree go away
- instance->octree_id = 0;
+ if (instance->indexer_id.is_valid()) {
+ _unpair_instance(instance);
}
switch (instance->base_type) {
@@ -642,6 +656,9 @@ void RendererSceneCull::instance_set_layer_mask(RID p_instance, uint32_t p_mask)
ERR_FAIL_COND(!instance);
instance->layer_mask = p_mask;
+ if (instance->scenario && instance->array_index >= 0) {
+ instance->scenario->instance_data[instance->array_index].layer_mask = p_mask;
+ }
}
void RendererSceneCull::instance_set_transform(RID p_instance, const Transform &p_transform) {
@@ -715,45 +732,12 @@ void RendererSceneCull::instance_set_visible(RID p_instance, bool p_visible) {
instance->visible = p_visible;
- switch (instance->base_type) {
- case RS::INSTANCE_LIGHT: {
- if (RSG::storage->light_get_type(instance->base) != RS::LIGHT_DIRECTIONAL && instance->octree_id && instance->scenario) {
- instance->scenario->octree.set_pairable(instance->octree_id, p_visible, 1 << RS::INSTANCE_LIGHT, p_visible ? RS::INSTANCE_GEOMETRY_MASK : 0);
- }
-
- } break;
- case RS::INSTANCE_REFLECTION_PROBE: {
- if (instance->octree_id && instance->scenario) {
- instance->scenario->octree.set_pairable(instance->octree_id, p_visible, 1 << RS::INSTANCE_REFLECTION_PROBE, p_visible ? RS::INSTANCE_GEOMETRY_MASK : 0);
- }
-
- } break;
- case RS::INSTANCE_DECAL: {
- if (instance->octree_id && instance->scenario) {
- instance->scenario->octree.set_pairable(instance->octree_id, p_visible, 1 << RS::INSTANCE_DECAL, p_visible ? RS::INSTANCE_GEOMETRY_MASK : 0);
- }
-
- } break;
- case RS::INSTANCE_LIGHTMAP: {
- if (instance->octree_id && instance->scenario) {
- instance->scenario->octree.set_pairable(instance->octree_id, p_visible, 1 << RS::INSTANCE_LIGHTMAP, p_visible ? RS::INSTANCE_GEOMETRY_MASK : 0);
- }
-
- } break;
- case RS::INSTANCE_GI_PROBE: {
- if (instance->octree_id && instance->scenario) {
- instance->scenario->octree.set_pairable(instance->octree_id, p_visible, 1 << RS::INSTANCE_GI_PROBE, p_visible ? (RS::INSTANCE_GEOMETRY_MASK | (1 << RS::INSTANCE_LIGHT)) : 0);
- }
-
- } break;
- case RS::INSTANCE_PARTICLES_COLLISION: {
- if (instance->octree_id && instance->scenario) {
- instance->scenario->octree.set_pairable(instance->octree_id, p_visible, 1 << RS::INSTANCE_PARTICLES_COLLISION, p_visible ? (1 << RS::INSTANCE_PARTICLES) : 0);
- }
-
- } break;
- default: {
+ if (p_visible) {
+ if (instance->scenario != nullptr) {
+ _instance_queue_update(instance, true, false);
}
+ } else if (instance->indexer_id.is_valid()) {
+ _unpair_instance(instance);
}
}
@@ -824,21 +808,21 @@ Vector<ObjectID> RendererSceneCull::instances_cull_aabb(const AABB &p_aabb, RID
const_cast<RendererSceneCull *>(this)->update_dirty_instances(); // check dirty instances before culling
- int culled = 0;
- Instance *cull[1024];
- culled = scenario->octree.cull_aabb(p_aabb, cull, 1024);
-
- for (int i = 0; i < culled; i++) {
- Instance *instance = cull[i];
- ERR_CONTINUE(!instance);
- if (instance->object_id.is_null()) {
- continue;
+ struct CullAABB {
+ Vector<ObjectID> instances;
+ _FORCE_INLINE_ bool operator()(void *p_data) {
+ Instance *p_instance = (Instance *)p_data;
+ if (!p_instance->object_id.is_null()) {
+ instances.push_back(p_instance->object_id);
+ }
+ return false;
}
+ };
- instances.push_back(instance->object_id);
- }
-
- return instances;
+ CullAABB cull_aabb;
+ scenario->indexers[Scenario::INDEXER_GEOMETRY].aabb_query(p_aabb, cull_aabb);
+ scenario->indexers[Scenario::INDEXER_VOLUMES].aabb_query(p_aabb, cull_aabb);
+ return cull_aabb.instances;
}
Vector<ObjectID> RendererSceneCull::instances_cull_ray(const Vector3 &p_from, const Vector3 &p_to, RID p_scenario) const {
@@ -847,21 +831,21 @@ Vector<ObjectID> RendererSceneCull::instances_cull_ray(const Vector3 &p_from, co
ERR_FAIL_COND_V(!scenario, instances);
const_cast<RendererSceneCull *>(this)->update_dirty_instances(); // check dirty instances before culling
- int culled = 0;
- Instance *cull[1024];
- culled = scenario->octree.cull_segment(p_from, p_from + p_to * 10000, cull, 1024);
-
- for (int i = 0; i < culled; i++) {
- Instance *instance = cull[i];
- ERR_CONTINUE(!instance);
- if (instance->object_id.is_null()) {
- continue;
+ struct CullRay {
+ Vector<ObjectID> instances;
+ _FORCE_INLINE_ bool operator()(void *p_data) {
+ Instance *p_instance = (Instance *)p_data;
+ if (!p_instance->object_id.is_null()) {
+ instances.push_back(p_instance->object_id);
+ }
+ return false;
}
+ };
- instances.push_back(instance->object_id);
- }
-
- return instances;
+ CullRay cull_ray;
+ scenario->indexers[Scenario::INDEXER_GEOMETRY].ray_query(p_from, p_to, cull_ray);
+ scenario->indexers[Scenario::INDEXER_VOLUMES].ray_query(p_from, p_to, cull_ray);
+ return cull_ray.instances;
}
Vector<ObjectID> RendererSceneCull::instances_cull_convex(const Vector<Plane> &p_convex, RID p_scenario) const {
@@ -870,22 +854,23 @@ Vector<ObjectID> RendererSceneCull::instances_cull_convex(const Vector<Plane> &p
ERR_FAIL_COND_V(!scenario, instances);
const_cast<RendererSceneCull *>(this)->update_dirty_instances(); // check dirty instances before culling
- int culled = 0;
- Instance *cull[1024];
-
- culled = scenario->octree.cull_convex(p_convex, cull, 1024);
+ Vector<Vector3> points = Geometry3D::compute_convex_mesh_points(&p_convex[0], p_convex.size());
- for (int i = 0; i < culled; i++) {
- Instance *instance = cull[i];
- ERR_CONTINUE(!instance);
- if (instance->object_id.is_null()) {
- continue;
+ struct CullConvex {
+ Vector<ObjectID> instances;
+ _FORCE_INLINE_ bool operator()(void *p_data) {
+ Instance *p_instance = (Instance *)p_data;
+ if (!p_instance->object_id.is_null()) {
+ instances.push_back(p_instance->object_id);
+ }
+ return false;
}
+ };
- instances.push_back(instance->object_id);
- }
-
- return instances;
+ CullConvex cull_convex;
+ scenario->indexers[Scenario::INDEXER_GEOMETRY].convex_query(p_convex.ptr(), p_convex.size(), points.ptr(), points.size(), cull_convex);
+ scenario->indexers[Scenario::INDEXER_VOLUMES].convex_query(p_convex.ptr(), p_convex.size(), points.ptr(), points.size(), cull_convex);
+ return cull_convex.instances;
}
void RendererSceneCull::instance_geometry_set_flag(RID p_instance, RS::InstanceFlags p_flags, bool p_enabled) {
@@ -898,6 +883,15 @@ void RendererSceneCull::instance_geometry_set_flag(RID p_instance, RS::InstanceF
case RS::INSTANCE_FLAG_USE_BAKED_LIGHT: {
instance->baked_light = p_enabled;
+ if (instance->scenario && instance->array_index >= 0) {
+ InstanceData &idata = instance->scenario->instance_data[instance->array_index];
+ if (instance->baked_light) {
+ idata.flags |= InstanceData::FLAG_USES_BAKED_LIGHT;
+ } else {
+ idata.flags &= ~uint32_t(InstanceData::FLAG_USES_BAKED_LIGHT);
+ }
+ }
+
} break;
case RS::INSTANCE_FLAG_USE_DYNAMIC_GI: {
if (p_enabled == instance->dynamic_gi) {
@@ -905,10 +899,8 @@ void RendererSceneCull::instance_geometry_set_flag(RID p_instance, RS::InstanceF
return;
}
- if (instance->octree_id != 0) {
- //remove from octree, it needs to be re-paired
- instance->scenario->octree.erase(instance->octree_id);
- instance->octree_id = 0;
+ if (instance->indexer_id.is_valid()) {
+ _unpair_instance(instance);
_instance_queue_update(instance, true, true);
}
@@ -919,6 +911,15 @@ void RendererSceneCull::instance_geometry_set_flag(RID p_instance, RS::InstanceF
case RS::INSTANCE_FLAG_DRAW_NEXT_FRAME_IF_VISIBLE: {
instance->redraw_if_visible = p_enabled;
+ if (instance->scenario && instance->array_index >= 0) {
+ InstanceData &idata = instance->scenario->instance_data[instance->array_index];
+ if (instance->redraw_if_visible) {
+ idata.flags |= InstanceData::FLAG_REDRAW_IF_VISIBLE;
+ } else {
+ idata.flags &= ~uint32_t(InstanceData::FLAG_REDRAW_IF_VISIBLE);
+ }
+ }
+
} break;
default: {
}
@@ -930,6 +931,23 @@ void RendererSceneCull::instance_geometry_set_cast_shadows_setting(RID p_instanc
ERR_FAIL_COND(!instance);
instance->cast_shadows = p_shadow_casting_setting;
+
+ if (instance->scenario && instance->array_index >= 0) {
+ InstanceData &idata = instance->scenario->instance_data[instance->array_index];
+
+ if (instance->cast_shadows != RS::SHADOW_CASTING_SETTING_SHADOWS_ONLY) {
+ idata.flags |= InstanceData::FLAG_CAST_SHADOWS;
+ } else {
+ idata.flags &= ~uint32_t(InstanceData::FLAG_CAST_SHADOWS);
+ }
+
+ if (instance->cast_shadows == RS::SHADOW_CASTING_SETTING_SHADOWS_ONLY) {
+ idata.flags |= InstanceData::FLAG_CAST_SHADOWS_ONLY;
+ } else {
+ idata.flags &= ~uint32_t(InstanceData::FLAG_CAST_SHADOWS_ONLY);
+ }
+ }
+
_instance_queue_update(instance, false, true);
}
@@ -1067,7 +1085,11 @@ void RendererSceneCull::_update_instance(Instance *p_instance) {
InstanceReflectionProbeData *reflection_probe = static_cast<InstanceReflectionProbeData *>(p_instance->base_data);
scene_render->reflection_probe_instance_set_transform(reflection_probe->instance, p_instance->transform);
- reflection_probe->reflection_dirty = true;
+
+ if (p_instance->scenario && p_instance->array_index >= 0) {
+ InstanceData &idata = p_instance->scenario->instance_data[p_instance->array_index];
+ idata.flags |= InstanceData::FLAG_REFLECTION_PROBE_DIRTY;
+ }
}
if (p_instance->base_type == RS::INSTANCE_DECAL) {
@@ -1102,7 +1124,7 @@ void RendererSceneCull::_update_instance(Instance *p_instance) {
//make sure lights are updated if it casts shadow
if (geom->can_cast_shadows) {
- for (List<Instance *>::Element *E = geom->lighting.front(); E; E = E->next()) {
+ for (Set<Instance *>::Element *E = geom->lights.front(); E; E = E->next()) {
InstanceLightData *light = static_cast<InstanceLightData *>(E->get()->base_data);
light->shadow_dirty = true;
}
@@ -1112,7 +1134,7 @@ void RendererSceneCull::_update_instance(Instance *p_instance) {
//affected by lightmap captures, must update capture info!
_update_instance_lightmap_captures(p_instance);
} else {
- if (!p_instance->lightmap_sh.empty()) {
+ if (!p_instance->lightmap_sh.is_empty()) {
p_instance->lightmap_sh.clear(); //don't need SH
p_instance->lightmap_target_sh.clear(); //don't need SH
}
@@ -1124,8 +1146,8 @@ void RendererSceneCull::_update_instance(Instance *p_instance) {
InstanceLightmapData *lightmap_data = static_cast<InstanceLightmapData *>(p_instance->base_data);
//erase dependencies, since no longer a lightmap
- for (List<InstanceLightmapData::PairInfo>::Element *E = lightmap_data->geometries.front(); E; E = E->next()) {
- Instance *geom = E->get().geometry;
+ for (Set<Instance *>::Element *E = lightmap_data->geometries.front(); E; E = E->next()) {
+ Instance *geom = E->get();
_instance_queue_update(geom, true, false);
}
}
@@ -1138,41 +1160,175 @@ void RendererSceneCull::_update_instance(Instance *p_instance) {
p_instance->transformed_aabb = new_aabb;
- if (!p_instance->scenario) {
+ if (p_instance->scenario == nullptr || !p_instance->visible || Math::is_zero_approx(p_instance->transform.basis.determinant())) {
+ p_instance->prev_transformed_aabb = p_instance->transformed_aabb;
return;
}
- if (p_instance->octree_id == 0) {
- uint32_t base_type = 1 << p_instance->base_type;
- uint32_t pairable_mask = 0;
- bool pairable = false;
+ //quantize to improve moving object performance
+ AABB bvh_aabb = p_instance->transformed_aabb;
+
+ if (p_instance->indexer_id.is_valid() && bvh_aabb != p_instance->prev_transformed_aabb) {
+ //assume motion, see if bounds need to be quantized
+ AABB motion_aabb = bvh_aabb.merge(p_instance->prev_transformed_aabb);
+ float motion_longest_axis = motion_aabb.get_longest_axis_size();
+ float longest_axis = p_instance->transformed_aabb.get_longest_axis_size();
- if (p_instance->base_type == RS::INSTANCE_LIGHT || p_instance->base_type == RS::INSTANCE_REFLECTION_PROBE || p_instance->base_type == RS::INSTANCE_DECAL || p_instance->base_type == RS::INSTANCE_LIGHTMAP) {
- pairable_mask = p_instance->visible ? RS::INSTANCE_GEOMETRY_MASK : 0;
- pairable = true;
+ if (motion_longest_axis < longest_axis * 2) {
+ //moved but not a lot, use motion aabb quantizing
+ float quantize_size = Math::pow(2.0, Math::ceil(Math::log(motion_longest_axis) / Math::log(2.0))) * 0.5; //one fifth
+ bvh_aabb.quantize(quantize_size);
}
+ }
- if (p_instance->base_type == RS::INSTANCE_PARTICLES_COLLISION) {
- pairable_mask = p_instance->visible ? (1 << RS::INSTANCE_PARTICLES) : 0;
- pairable = true;
+ if (!p_instance->indexer_id.is_valid()) {
+ if ((1 << p_instance->base_type) & RS::INSTANCE_GEOMETRY_MASK) {
+ p_instance->indexer_id = p_instance->scenario->indexers[Scenario::INDEXER_GEOMETRY].insert(bvh_aabb, p_instance);
+ } else {
+ p_instance->indexer_id = p_instance->scenario->indexers[Scenario::INDEXER_VOLUMES].insert(bvh_aabb, p_instance);
}
- if (p_instance->base_type == RS::INSTANCE_GI_PROBE) {
- //lights and geometries
- pairable_mask = p_instance->visible ? RS::INSTANCE_GEOMETRY_MASK | (1 << RS::INSTANCE_LIGHT) : 0;
- pairable = true;
+ p_instance->array_index = p_instance->scenario->instance_data.size();
+ InstanceData idata;
+ idata.instance = p_instance;
+ idata.layer_mask = p_instance->layer_mask;
+ idata.flags = p_instance->base_type; //changing it means de-indexing, so this never needs to be changed later
+ idata.base_rid = p_instance->base;
+ switch (p_instance->base_type) {
+ case RS::INSTANCE_LIGHT: {
+ idata.instance_data_rid = static_cast<InstanceLightData *>(p_instance->base_data)->instance;
+ } break;
+ case RS::INSTANCE_REFLECTION_PROBE: {
+ idata.instance_data_rid = static_cast<InstanceReflectionProbeData *>(p_instance->base_data)->instance;
+ } break;
+ case RS::INSTANCE_DECAL: {
+ idata.instance_data_rid = static_cast<InstanceDecalData *>(p_instance->base_data)->instance;
+ } break;
+ case RS::INSTANCE_GI_PROBE: {
+ idata.instance_data_rid = static_cast<InstanceGIProbeData *>(p_instance->base_data)->probe_instance;
+ } break;
+ default: {
+ }
}
- // not inside octree
- p_instance->octree_id = p_instance->scenario->octree.create(p_instance, new_aabb, 0, pairable, base_type, pairable_mask);
+ if (p_instance->base_type == RS::INSTANCE_REFLECTION_PROBE) {
+ //always dirty when added
+ idata.flags |= InstanceData::FLAG_REFLECTION_PROBE_DIRTY;
+ }
+ if (p_instance->cast_shadows != RS::SHADOW_CASTING_SETTING_SHADOWS_ONLY) {
+ idata.flags |= InstanceData::FLAG_CAST_SHADOWS;
+ }
+ if (p_instance->cast_shadows == RS::SHADOW_CASTING_SETTING_SHADOWS_ONLY) {
+ idata.flags |= InstanceData::FLAG_CAST_SHADOWS_ONLY;
+ }
+ if (p_instance->redraw_if_visible) {
+ idata.flags |= InstanceData::FLAG_REDRAW_IF_VISIBLE;
+ }
+ // dirty flags should not be set here, since no pairing has happened
+ if (p_instance->baked_light) {
+ idata.flags |= InstanceData::FLAG_USES_BAKED_LIGHT;
+ }
+ if (p_instance->mesh_instance.is_valid()) {
+ idata.flags |= InstanceData::FLAG_USES_MESH_INSTANCE;
+ }
+ p_instance->scenario->instance_data.push_back(idata);
+ p_instance->scenario->instance_aabbs.push_back(InstanceBounds(p_instance->transformed_aabb));
} else {
- /*
- if (new_aabb==p_instance->data.transformed_aabb)
- return;
- */
+ if ((1 << p_instance->base_type) & RS::INSTANCE_GEOMETRY_MASK) {
+ p_instance->scenario->indexers[Scenario::INDEXER_GEOMETRY].update(p_instance->indexer_id, bvh_aabb);
+ } else {
+ p_instance->scenario->indexers[Scenario::INDEXER_VOLUMES].update(p_instance->indexer_id, bvh_aabb);
+ }
+ p_instance->scenario->instance_aabbs[p_instance->array_index] = InstanceBounds(p_instance->transformed_aabb);
+ }
+
+ //move instance and repair
+ pair_pass++;
+
+ PairInstances pair;
+
+ pair.instance = p_instance;
+ pair.pair_allocator = &pair_allocator;
+ pair.pair_pass = pair_pass;
+ pair.pair_mask = 0;
+
+ if ((1 << p_instance->base_type) & RS::INSTANCE_GEOMETRY_MASK) {
+ pair.pair_mask |= 1 << RS::INSTANCE_LIGHT;
+ pair.pair_mask |= 1 << RS::INSTANCE_GI_PROBE;
+ pair.pair_mask |= 1 << RS::INSTANCE_LIGHTMAP;
- p_instance->scenario->octree.move(p_instance->octree_id, new_aabb);
+ if (pair_volumes_to_mesh) {
+ pair.pair_mask |= 1 << RS::INSTANCE_DECAL;
+ pair.pair_mask |= 1 << RS::INSTANCE_REFLECTION_PROBE;
+ }
+ pair.bvh2 = &p_instance->scenario->indexers[Scenario::INDEXER_VOLUMES];
+ } else if (p_instance->base_type == RS::INSTANCE_LIGHT) {
+ pair.pair_mask |= RS::INSTANCE_GEOMETRY_MASK;
+ pair.bvh = &p_instance->scenario->indexers[Scenario::INDEXER_GEOMETRY];
+
+ if (RSG::storage->light_get_bake_mode(p_instance->base) == RS::LIGHT_BAKE_DYNAMIC) {
+ pair.pair_mask |= (1 << RS::INSTANCE_GI_PROBE);
+ pair.bvh2 = &p_instance->scenario->indexers[Scenario::INDEXER_VOLUMES];
+ }
+ } else if (pair_volumes_to_mesh && (p_instance->base_type == RS::INSTANCE_REFLECTION_PROBE || p_instance->base_type == RS::INSTANCE_DECAL)) {
+ pair.pair_mask = RS::INSTANCE_GEOMETRY_MASK;
+ pair.bvh = &p_instance->scenario->indexers[Scenario::INDEXER_GEOMETRY];
+ } else if (p_instance->base_type == RS::INSTANCE_PARTICLES_COLLISION) {
+ pair.pair_mask = (1 << RS::INSTANCE_PARTICLES);
+ pair.bvh = &p_instance->scenario->indexers[Scenario::INDEXER_GEOMETRY];
+ } else if (p_instance->base_type == RS::INSTANCE_GI_PROBE) {
+ //lights and geometries
+ pair.pair_mask = RS::INSTANCE_GEOMETRY_MASK | (1 << RS::INSTANCE_LIGHT);
+ pair.bvh = &p_instance->scenario->indexers[Scenario::INDEXER_GEOMETRY];
+ pair.bvh2 = &p_instance->scenario->indexers[Scenario::INDEXER_VOLUMES];
+ }
+
+ pair.pair();
+
+ p_instance->prev_transformed_aabb = p_instance->transformed_aabb;
+}
+
+void RendererSceneCull::_unpair_instance(Instance *p_instance) {
+ if (!p_instance->indexer_id.is_valid()) {
+ return; //nothing to do
+ }
+
+ while (p_instance->pairs.first()) {
+ InstancePair *pair = p_instance->pairs.first()->self();
+ Instance *other_instance = p_instance == pair->a ? pair->b : pair->a;
+ _instance_unpair(p_instance, other_instance);
+ pair_allocator.free(pair);
+ }
+
+ if ((1 << p_instance->base_type) & RS::INSTANCE_GEOMETRY_MASK) {
+ p_instance->scenario->indexers[Scenario::INDEXER_GEOMETRY].remove(p_instance->indexer_id);
+ } else {
+ p_instance->scenario->indexers[Scenario::INDEXER_VOLUMES].remove(p_instance->indexer_id);
+ }
+
+ p_instance->indexer_id = DynamicBVH::ID();
+
+ //replace this by last
+ int32_t swap_with_index = p_instance->scenario->instance_data.size() - 1;
+ if (swap_with_index != p_instance->array_index) {
+ p_instance->scenario->instance_data[swap_with_index].instance->array_index = p_instance->array_index; //swap
+ p_instance->scenario->instance_data[p_instance->array_index] = p_instance->scenario->instance_data[swap_with_index];
+ p_instance->scenario->instance_aabbs[p_instance->array_index] = p_instance->scenario->instance_aabbs[swap_with_index];
+ }
+
+ // pop last
+ p_instance->scenario->instance_data.pop_back();
+ p_instance->scenario->instance_aabbs.pop_back();
+
+ //uninitialize
+ p_instance->array_index = -1;
+ if ((1 << p_instance->base_type) & RS::INSTANCE_GEOMETRY_MASK) {
+ // Clear these now because the InstanceData containing the dirty flags is gone
+ p_instance->light_instances.clear();
+ p_instance->reflection_probe_instances.clear();
+ //p_instance->decal_instances.clear(); will implement later
+ p_instance->gi_probe_instances.clear();
}
}
@@ -1264,7 +1420,7 @@ void RendererSceneCull::_update_instance_lightmap_captures(Instance *p_instance)
float accum_blend = 0.0;
InstanceGeometryData *geom = static_cast<InstanceGeometryData *>(p_instance->base_data);
- for (List<Instance *>::Element *E = geom->lightmap_captures.front(); E; E = E->next()) {
+ for (Set<Instance *>::Element *E = geom->lightmap_captures.front(); E; E = E->next()) {
Instance *lightmap = E->get();
bool interior = RSG::storage->lightmap_is_interior(lightmap->base);
@@ -1332,389 +1488,311 @@ void RendererSceneCull::_update_instance_lightmap_captures(Instance *p_instance)
}
}
-bool RendererSceneCull::_light_instance_update_shadow(Instance *p_instance, const Transform p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_orthogonal, bool p_cam_vaspect, RID p_shadow_atlas, Scenario *p_scenario, float p_screen_lod_threshold) {
+void RendererSceneCull::_light_instance_setup_directional_shadow(int p_shadow_index, Instance *p_instance, const Transform p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_orthogonal, bool p_cam_vaspect) {
InstanceLightData *light = static_cast<InstanceLightData *>(p_instance->base_data);
Transform light_transform = p_instance->transform;
light_transform.orthonormalize(); //scale does not count on lights
- bool animated_material_found = false;
-
- switch (RSG::storage->light_get_type(p_instance->base)) {
- case RS::LIGHT_DIRECTIONAL: {
- Plane camera_plane(p_cam_transform.get_origin(), -p_cam_transform.basis.get_axis(Vector3::AXIS_Z));
-
- real_t max_distance = p_cam_projection.get_z_far();
- real_t shadow_max = RSG::storage->light_get_param(p_instance->base, RS::LIGHT_PARAM_SHADOW_MAX_DISTANCE);
- if (shadow_max > 0 && !p_cam_orthogonal) { //its impractical (and leads to unwanted behaviors) to set max distance in orthogonal camera
- max_distance = MIN(shadow_max, max_distance);
- }
- max_distance = MAX(max_distance, p_cam_projection.get_z_near() + 0.001);
- real_t min_distance = MIN(p_cam_projection.get_z_near(), max_distance);
-
- RS::LightDirectionalShadowDepthRangeMode depth_range_mode = RSG::storage->light_directional_get_shadow_depth_range_mode(p_instance->base);
-
- real_t pancake_size = RSG::storage->light_get_param(p_instance->base, RS::LIGHT_PARAM_SHADOW_PANCAKE_SIZE);
-
- if (depth_range_mode == RS::LIGHT_DIRECTIONAL_SHADOW_DEPTH_RANGE_OPTIMIZED) {
- //optimize min/max
- Vector<Plane> planes = p_cam_projection.get_projection_planes(p_cam_transform);
- int cull_count = p_scenario->octree.cull_convex(planes, instance_shadow_cull_result, MAX_INSTANCE_CULL, RS::INSTANCE_GEOMETRY_MASK);
- Plane base(p_cam_transform.origin, -p_cam_transform.basis.get_axis(2));
- //check distance max and min
+ real_t max_distance = p_cam_projection.get_z_far();
+ real_t shadow_max = RSG::storage->light_get_param(p_instance->base, RS::LIGHT_PARAM_SHADOW_MAX_DISTANCE);
+ if (shadow_max > 0 && !p_cam_orthogonal) { //its impractical (and leads to unwanted behaviors) to set max distance in orthogonal camera
+ max_distance = MIN(shadow_max, max_distance);
+ }
+ max_distance = MAX(max_distance, p_cam_projection.get_z_near() + 0.001);
+ real_t min_distance = MIN(p_cam_projection.get_z_near(), max_distance);
- bool found_items = false;
- real_t z_max = -1e20;
- real_t z_min = 1e20;
+ RS::LightDirectionalShadowDepthRangeMode depth_range_mode = RSG::storage->light_directional_get_shadow_depth_range_mode(p_instance->base);
- for (int i = 0; i < cull_count; i++) {
- Instance *instance = instance_shadow_cull_result[i];
- if (!instance->visible || !((1 << instance->base_type) & RS::INSTANCE_GEOMETRY_MASK) || !static_cast<InstanceGeometryData *>(instance->base_data)->can_cast_shadows) {
- continue;
- }
+ real_t pancake_size = RSG::storage->light_get_param(p_instance->base, RS::LIGHT_PARAM_SHADOW_PANCAKE_SIZE);
- if (static_cast<InstanceGeometryData *>(instance->base_data)->material_is_animated) {
- animated_material_found = true;
- }
+ real_t range = max_distance - min_distance;
- real_t max, min;
- instance->transformed_aabb.project_range_in_plane(base, min, max);
+ int splits = 0;
+ switch (RSG::storage->light_directional_get_shadow_mode(p_instance->base)) {
+ case RS::LIGHT_DIRECTIONAL_SHADOW_ORTHOGONAL:
+ splits = 1;
+ break;
+ case RS::LIGHT_DIRECTIONAL_SHADOW_PARALLEL_2_SPLITS:
+ splits = 2;
+ break;
+ case RS::LIGHT_DIRECTIONAL_SHADOW_PARALLEL_4_SPLITS:
+ splits = 4;
+ break;
+ }
- if (max > z_max) {
- z_max = max;
- }
+ real_t distances[5];
- if (min < z_min) {
- z_min = min;
- }
+ distances[0] = min_distance;
+ for (int i = 0; i < splits; i++) {
+ distances[i + 1] = min_distance + RSG::storage->light_get_param(p_instance->base, RS::LightParam(RS::LIGHT_PARAM_SHADOW_SPLIT_1_OFFSET + i)) * range;
+ };
- found_items = true;
- }
+ distances[splits] = max_distance;
- if (found_items) {
- min_distance = MAX(min_distance, z_min);
- max_distance = MIN(max_distance, z_max);
- }
- }
+ real_t texture_size = scene_render->get_directional_light_shadow_size(light->instance);
- real_t range = max_distance - min_distance;
+ bool overlap = RSG::storage->light_directional_get_blend_splits(p_instance->base);
- int splits = 0;
- switch (RSG::storage->light_directional_get_shadow_mode(p_instance->base)) {
- case RS::LIGHT_DIRECTIONAL_SHADOW_ORTHOGONAL:
- splits = 1;
- break;
- case RS::LIGHT_DIRECTIONAL_SHADOW_PARALLEL_2_SPLITS:
- splits = 2;
- break;
- case RS::LIGHT_DIRECTIONAL_SHADOW_PARALLEL_4_SPLITS:
- splits = 4;
- break;
- }
+ real_t first_radius = 0.0;
- real_t distances[5];
+ real_t min_distance_bias_scale = distances[1];
- distances[0] = min_distance;
- for (int i = 0; i < splits; i++) {
- distances[i + 1] = min_distance + RSG::storage->light_get_param(p_instance->base, RS::LightParam(RS::LIGHT_PARAM_SHADOW_SPLIT_1_OFFSET + i)) * range;
- };
+ cull.shadow_count = p_shadow_index + 1;
+ cull.shadows[p_shadow_index].cascade_count = splits;
+ cull.shadows[p_shadow_index].light_instance = light->instance;
- distances[splits] = max_distance;
+ for (int i = 0; i < splits; i++) {
+ RENDER_TIMESTAMP("Culling Directional Light split" + itos(i));
- real_t texture_size = scene_render->get_directional_light_shadow_size(light->instance);
+ // setup a camera matrix for that range!
+ CameraMatrix camera_matrix;
- bool overlap = RSG::storage->light_directional_get_blend_splits(p_instance->base);
+ real_t aspect = p_cam_projection.get_aspect();
- real_t first_radius = 0.0;
+ if (p_cam_orthogonal) {
+ Vector2 vp_he = p_cam_projection.get_viewport_half_extents();
- real_t min_distance_bias_scale = pancake_size > 0 ? distances[1] / 10.0 : 0;
+ camera_matrix.set_orthogonal(vp_he.y * 2.0, aspect, distances[(i == 0 || !overlap) ? i : i - 1], distances[i + 1], false);
+ } else {
+ real_t fov = p_cam_projection.get_fov(); //this is actually yfov, because set aspect tries to keep it
+ camera_matrix.set_perspective(fov, aspect, distances[(i == 0 || !overlap) ? i : i - 1], distances[i + 1], true);
+ }
- for (int i = 0; i < splits; i++) {
- RENDER_TIMESTAMP("Culling Directional Light split" + itos(i));
+ //obtain the frustum endpoints
- // setup a camera matrix for that range!
- CameraMatrix camera_matrix;
+ Vector3 endpoints[8]; // frustum plane endpoints
+ bool res = camera_matrix.get_endpoints(p_cam_transform, endpoints);
+ ERR_CONTINUE(!res);
- real_t aspect = p_cam_projection.get_aspect();
+ // obtain the light frustum ranges (given endpoints)
- if (p_cam_orthogonal) {
- Vector2 vp_he = p_cam_projection.get_viewport_half_extents();
+ Transform transform = light_transform; //discard scale and stabilize light
- camera_matrix.set_orthogonal(vp_he.y * 2.0, aspect, distances[(i == 0 || !overlap) ? i : i - 1], distances[i + 1], false);
- } else {
- real_t fov = p_cam_projection.get_fov(); //this is actually yfov, because set aspect tries to keep it
- camera_matrix.set_perspective(fov, aspect, distances[(i == 0 || !overlap) ? i : i - 1], distances[i + 1], true);
- }
+ Vector3 x_vec = transform.basis.get_axis(Vector3::AXIS_X).normalized();
+ Vector3 y_vec = transform.basis.get_axis(Vector3::AXIS_Y).normalized();
+ Vector3 z_vec = transform.basis.get_axis(Vector3::AXIS_Z).normalized();
+ //z_vec points against the camera, like in default opengl
- //obtain the frustum endpoints
+ real_t x_min = 0.f, x_max = 0.f;
+ real_t y_min = 0.f, y_max = 0.f;
+ real_t z_min = 0.f, z_max = 0.f;
- Vector3 endpoints[8]; // frustum plane endpoints
- bool res = camera_matrix.get_endpoints(p_cam_transform, endpoints);
- ERR_CONTINUE(!res);
+ // FIXME: z_max_cam is defined, computed, but not used below when setting up
+ // ortho_camera. Commented out for now to fix warnings but should be investigated.
+ real_t x_min_cam = 0.f, x_max_cam = 0.f;
+ real_t y_min_cam = 0.f, y_max_cam = 0.f;
+ real_t z_min_cam = 0.f;
+ //real_t z_max_cam = 0.f;
- // obtain the light frustm ranges (given endpoints)
+ real_t bias_scale = 1.0;
+ real_t aspect_bias_scale = 1.0;
- Transform transform = light_transform; //discard scale and stabilize light
+ //used for culling
- Vector3 x_vec = transform.basis.get_axis(Vector3::AXIS_X).normalized();
- Vector3 y_vec = transform.basis.get_axis(Vector3::AXIS_Y).normalized();
- Vector3 z_vec = transform.basis.get_axis(Vector3::AXIS_Z).normalized();
- //z_vec points agsint the camera, like in default opengl
+ for (int j = 0; j < 8; j++) {
+ real_t d_x = x_vec.dot(endpoints[j]);
+ real_t d_y = y_vec.dot(endpoints[j]);
+ real_t d_z = z_vec.dot(endpoints[j]);
- real_t x_min = 0.f, x_max = 0.f;
- real_t y_min = 0.f, y_max = 0.f;
- real_t z_min = 0.f, z_max = 0.f;
+ if (j == 0 || d_x < x_min) {
+ x_min = d_x;
+ }
+ if (j == 0 || d_x > x_max) {
+ x_max = d_x;
+ }
- // FIXME: z_max_cam is defined, computed, but not used below when setting up
- // ortho_camera. Commented out for now to fix warnings but should be investigated.
- real_t x_min_cam = 0.f, x_max_cam = 0.f;
- real_t y_min_cam = 0.f, y_max_cam = 0.f;
- real_t z_min_cam = 0.f;
- //real_t z_max_cam = 0.f;
+ if (j == 0 || d_y < y_min) {
+ y_min = d_y;
+ }
+ if (j == 0 || d_y > y_max) {
+ y_max = d_y;
+ }
- real_t bias_scale = 1.0;
- real_t aspect_bias_scale = 1.0;
+ if (j == 0 || d_z < z_min) {
+ z_min = d_z;
+ }
+ if (j == 0 || d_z > z_max) {
+ z_max = d_z;
+ }
+ }
- //used for culling
+ real_t radius = 0;
+ real_t soft_shadow_expand = 0;
+ Vector3 center;
- for (int j = 0; j < 8; j++) {
- real_t d_x = x_vec.dot(endpoints[j]);
- real_t d_y = y_vec.dot(endpoints[j]);
- real_t d_z = z_vec.dot(endpoints[j]);
+ {
+ //camera viewport stuff
- if (j == 0 || d_x < x_min) {
- x_min = d_x;
- }
- if (j == 0 || d_x > x_max) {
- x_max = d_x;
- }
+ for (int j = 0; j < 8; j++) {
+ center += endpoints[j];
+ }
+ center /= 8.0;
- if (j == 0 || d_y < y_min) {
- y_min = d_y;
- }
- if (j == 0 || d_y > y_max) {
- y_max = d_y;
- }
+ //center=x_vec*(x_max-x_min)*0.5 + y_vec*(y_max-y_min)*0.5 + z_vec*(z_max-z_min)*0.5;
- if (j == 0 || d_z < z_min) {
- z_min = d_z;
- }
- if (j == 0 || d_z > z_max) {
- z_max = d_z;
- }
+ for (int j = 0; j < 8; j++) {
+ real_t d = center.distance_to(endpoints[j]);
+ if (d > radius) {
+ radius = d;
}
+ }
- real_t radius = 0;
- real_t soft_shadow_expand = 0;
- Vector3 center;
-
- {
- //camera viewport stuff
-
- for (int j = 0; j < 8; j++) {
- center += endpoints[j];
- }
- center /= 8.0;
-
- //center=x_vec*(x_max-x_min)*0.5 + y_vec*(y_max-y_min)*0.5 + z_vec*(z_max-z_min)*0.5;
-
- for (int j = 0; j < 8; j++) {
- real_t d = center.distance_to(endpoints[j]);
- if (d > radius) {
- radius = d;
- }
- }
-
- radius *= texture_size / (texture_size - 2.0); //add a texel by each side
-
- if (i == 0) {
- first_radius = radius;
- } else {
- bias_scale = radius / first_radius;
- }
-
- z_min_cam = z_vec.dot(center) - radius;
-
- {
- float soft_shadow_angle = RSG::storage->light_get_param(p_instance->base, RS::LIGHT_PARAM_SIZE);
-
- if (soft_shadow_angle > 0.0 && pancake_size > 0.0) {
- float z_range = (z_vec.dot(center) + radius + pancake_size) - z_min_cam;
- soft_shadow_expand = Math::tan(Math::deg2rad(soft_shadow_angle)) * z_range;
+ radius *= texture_size / (texture_size - 2.0); //add a texel by each side
- x_max += soft_shadow_expand;
- y_max += soft_shadow_expand;
+ if (i == 0) {
+ first_radius = radius;
+ } else {
+ bias_scale = radius / first_radius;
+ }
- x_min -= soft_shadow_expand;
- y_min -= soft_shadow_expand;
- }
- }
+ z_min_cam = z_vec.dot(center) - radius;
- x_max_cam = x_vec.dot(center) + radius + soft_shadow_expand;
- x_min_cam = x_vec.dot(center) - radius - soft_shadow_expand;
- y_max_cam = y_vec.dot(center) + radius + soft_shadow_expand;
- y_min_cam = y_vec.dot(center) - radius - soft_shadow_expand;
+ {
+ float soft_shadow_angle = RSG::storage->light_get_param(p_instance->base, RS::LIGHT_PARAM_SIZE);
- if (depth_range_mode == RS::LIGHT_DIRECTIONAL_SHADOW_DEPTH_RANGE_STABLE) {
- //this trick here is what stabilizes the shadow (make potential jaggies to not move)
- //at the cost of some wasted resolution. Still the quality increase is very well worth it
+ if (soft_shadow_angle > 0.0) {
+ float z_range = (z_vec.dot(center) + radius + pancake_size) - z_min_cam;
+ soft_shadow_expand = Math::tan(Math::deg2rad(soft_shadow_angle)) * z_range;
- real_t unit = radius * 2.0 / texture_size;
+ x_max += soft_shadow_expand;
+ y_max += soft_shadow_expand;
- x_max_cam = Math::stepify(x_max_cam, unit);
- x_min_cam = Math::stepify(x_min_cam, unit);
- y_max_cam = Math::stepify(y_max_cam, unit);
- y_min_cam = Math::stepify(y_min_cam, unit);
- }
+ x_min -= soft_shadow_expand;
+ y_min -= soft_shadow_expand;
}
+ }
- //now that we now all ranges, we can proceed to make the light frustum planes, for culling octree
+ x_max_cam = x_vec.dot(center) + radius + soft_shadow_expand;
+ x_min_cam = x_vec.dot(center) - radius - soft_shadow_expand;
+ y_max_cam = y_vec.dot(center) + radius + soft_shadow_expand;
+ y_min_cam = y_vec.dot(center) - radius - soft_shadow_expand;
- Vector<Plane> light_frustum_planes;
- light_frustum_planes.resize(6);
+ if (depth_range_mode == RS::LIGHT_DIRECTIONAL_SHADOW_DEPTH_RANGE_STABLE) {
+ //this trick here is what stabilizes the shadow (make potential jaggies to not move)
+ //at the cost of some wasted resolution. Still the quality increase is very well worth it
- //right/left
- light_frustum_planes.write[0] = Plane(x_vec, x_max);
- light_frustum_planes.write[1] = Plane(-x_vec, -x_min);
- //top/bottom
- light_frustum_planes.write[2] = Plane(y_vec, y_max);
- light_frustum_planes.write[3] = Plane(-y_vec, -y_min);
- //near/far
- light_frustum_planes.write[4] = Plane(z_vec, z_max + 1e6);
- light_frustum_planes.write[5] = Plane(-z_vec, -z_min); // z_min is ok, since casters further than far-light plane are not needed
+ real_t unit = radius * 2.0 / texture_size;
- int cull_count = p_scenario->octree.cull_convex(light_frustum_planes, instance_shadow_cull_result, MAX_INSTANCE_CULL, RS::INSTANCE_GEOMETRY_MASK);
+ x_max_cam = Math::stepify(x_max_cam, unit);
+ x_min_cam = Math::stepify(x_min_cam, unit);
+ y_max_cam = Math::stepify(y_max_cam, unit);
+ y_min_cam = Math::stepify(y_min_cam, unit);
+ }
+ }
- // a pre pass will need to be needed to determine the actual z-near to be used
+ //now that we know all ranges, we can proceed to make the light frustum planes, for culling octree
- Plane near_plane(light_transform.origin, -light_transform.basis.get_axis(2));
+ Vector<Plane> light_frustum_planes;
+ light_frustum_planes.resize(6);
- real_t cull_max = 0;
- for (int j = 0; j < cull_count; j++) {
- real_t min, max;
- Instance *instance = instance_shadow_cull_result[j];
- if (!instance->visible || !((1 << instance->base_type) & RS::INSTANCE_GEOMETRY_MASK) || !static_cast<InstanceGeometryData *>(instance->base_data)->can_cast_shadows) {
- cull_count--;
- SWAP(instance_shadow_cull_result[j], instance_shadow_cull_result[cull_count]);
- j--;
- continue;
- }
+ //right/left
+ light_frustum_planes.write[0] = Plane(x_vec, x_max);
+ light_frustum_planes.write[1] = Plane(-x_vec, -x_min);
+ //top/bottom
+ light_frustum_planes.write[2] = Plane(y_vec, y_max);
+ light_frustum_planes.write[3] = Plane(-y_vec, -y_min);
+ //near/far
+ light_frustum_planes.write[4] = Plane(z_vec, z_max + 1e6);
+ light_frustum_planes.write[5] = Plane(-z_vec, -z_min); // z_min is ok, since casters further than far-light plane are not needed
- instance->transformed_aabb.project_range_in_plane(Plane(z_vec, 0), min, max);
- instance->depth = near_plane.distance_to(instance->transform.origin);
- instance->depth_layer = 0;
- if (j == 0 || max > cull_max) {
- cull_max = max;
- }
+ // a pre pass will need to be needed to determine the actual z-near to be used
- if (instance->mesh_instance.is_valid()) {
- RSG::storage->mesh_instance_check_for_update(instance->mesh_instance);
- }
- }
+ if (pancake_size > 0) {
+ z_max = z_vec.dot(center) + radius + pancake_size;
+ }
- if (cull_max > z_max) {
- z_max = cull_max;
- }
+ if (aspect != 1.0) {
+ // if the aspect is different, then the radius will become larger.
+ // if this happens, then bias needs to be adjusted too, as depth will increase
+ // to do this, compare the depth of one that would have resulted from a square frustum
- if (pancake_size > 0) {
- z_max = z_vec.dot(center) + radius + pancake_size;
+ CameraMatrix camera_matrix_square;
+ if (p_cam_orthogonal) {
+ Vector2 vp_he = camera_matrix.get_viewport_half_extents();
+ if (p_cam_vaspect) {
+ camera_matrix_square.set_orthogonal(vp_he.x * 2.0, 1.0, distances[(i == 0 || !overlap) ? i : i - 1], distances[i + 1], true);
+ } else {
+ camera_matrix_square.set_orthogonal(vp_he.y * 2.0, 1.0, distances[(i == 0 || !overlap) ? i : i - 1], distances[i + 1], false);
}
+ } else {
+ Vector2 vp_he = camera_matrix.get_viewport_half_extents();
+ if (p_cam_vaspect) {
+ camera_matrix_square.set_frustum(vp_he.x * 2.0, 1.0, Vector2(), distances[(i == 0 || !overlap) ? i : i - 1], distances[i + 1], true);
+ } else {
+ camera_matrix_square.set_frustum(vp_he.y * 2.0, 1.0, Vector2(), distances[(i == 0 || !overlap) ? i : i - 1], distances[i + 1], false);
+ }
+ }
- if (aspect != 1.0) {
- // if the aspect is different, then the radius will become larger.
- // if this happens, then bias needs to be adjusted too, as depth will increase
- // to do this, compare the depth of one that would have resulted from a square frustum
-
- CameraMatrix camera_matrix_square;
- if (p_cam_orthogonal) {
- Vector2 vp_he = camera_matrix.get_viewport_half_extents();
- if (p_cam_vaspect) {
- camera_matrix_square.set_orthogonal(vp_he.x * 2.0, 1.0, distances[(i == 0 || !overlap) ? i : i - 1], distances[i + 1], true);
- } else {
- camera_matrix_square.set_orthogonal(vp_he.y * 2.0, 1.0, distances[(i == 0 || !overlap) ? i : i - 1], distances[i + 1], false);
- }
- } else {
- Vector2 vp_he = camera_matrix.get_viewport_half_extents();
- if (p_cam_vaspect) {
- camera_matrix_square.set_frustum(vp_he.x * 2.0, 1.0, Vector2(), distances[(i == 0 || !overlap) ? i : i - 1], distances[i + 1], true);
- } else {
- camera_matrix_square.set_frustum(vp_he.y * 2.0, 1.0, Vector2(), distances[(i == 0 || !overlap) ? i : i - 1], distances[i + 1], false);
- }
- }
-
- Vector3 endpoints_square[8]; // frustum plane endpoints
- res = camera_matrix_square.get_endpoints(p_cam_transform, endpoints_square);
- ERR_CONTINUE(!res);
- Vector3 center_square;
- real_t z_max_square = 0;
-
- for (int j = 0; j < 8; j++) {
- center_square += endpoints_square[j];
-
- real_t d_z = z_vec.dot(endpoints_square[j]);
-
- if (j == 0 || d_z > z_max_square) {
- z_max_square = d_z;
- }
- }
-
- if (cull_max > z_max_square) {
- z_max_square = cull_max;
- }
-
- center_square /= 8.0;
-
- real_t radius_square = 0;
-
- for (int j = 0; j < 8; j++) {
- real_t d = center_square.distance_to(endpoints_square[j]);
- if (d > radius_square) {
- radius_square = d;
- }
- }
-
- radius_square *= texture_size / (texture_size - 2.0); //add a texel by each side
+ Vector3 endpoints_square[8]; // frustum plane endpoints
+ res = camera_matrix_square.get_endpoints(p_cam_transform, endpoints_square);
+ ERR_CONTINUE(!res);
+ Vector3 center_square;
- if (pancake_size > 0) {
- z_max_square = z_vec.dot(center_square) + radius_square + pancake_size;
- }
+ for (int j = 0; j < 8; j++) {
+ center_square += endpoints_square[j];
+ }
- real_t z_min_cam_square = z_vec.dot(center_square) - radius_square;
+ center_square /= 8.0;
- aspect_bias_scale = (z_max - z_min_cam) / (z_max_square - z_min_cam_square);
+ real_t radius_square = 0;
- // this is not entirely perfect, because the cull-adjusted z-max may be different
- // but at least it's warranted that it results in a greater bias, so no acne should be present either way.
- // pancaking also helps with this.
+ for (int j = 0; j < 8; j++) {
+ real_t d = center_square.distance_to(endpoints_square[j]);
+ if (d > radius_square) {
+ radius_square = d;
}
+ }
- {
- CameraMatrix ortho_camera;
- real_t half_x = (x_max_cam - x_min_cam) * 0.5;
- real_t half_y = (y_max_cam - y_min_cam) * 0.5;
+ radius_square *= texture_size / (texture_size - 2.0); //add a texel by each side
- ortho_camera.set_orthogonal(-half_x, half_x, -half_y, half_y, 0, (z_max - z_min_cam));
+ float z_max_square = z_vec.dot(center_square) + radius_square + pancake_size;
- Vector2 uv_scale(1.0 / (x_max_cam - x_min_cam), 1.0 / (y_max_cam - y_min_cam));
+ real_t z_min_cam_square = z_vec.dot(center_square) - radius_square;
- Transform ortho_transform;
- ortho_transform.basis = transform.basis;
- ortho_transform.origin = x_vec * (x_min_cam + half_x) + y_vec * (y_min_cam + half_y) + z_vec * z_max;
+ aspect_bias_scale = (z_max - z_min_cam) / (z_max_square - z_min_cam_square);
- {
- Vector3 max_in_view = p_cam_transform.affine_inverse().xform(z_vec * cull_max);
- Vector3 dir_in_view = p_cam_transform.xform_inv(z_vec).normalized();
- cull_max = dir_in_view.dot(max_in_view);
- }
+ // this is not entirely perfect, because the cull-adjusted z-max may be different
+ // but at least it's warranted that it results in a greater bias, so no acne should be present either way.
+ // pancaking also helps with this.
+ }
- scene_render->light_instance_set_shadow_transform(light->instance, ortho_camera, ortho_transform, z_max - z_min_cam, distances[i + 1], i, radius * 2.0 / texture_size, bias_scale * aspect_bias_scale * min_distance_bias_scale, z_max, uv_scale);
- }
+ {
+ CameraMatrix ortho_camera;
+ real_t half_x = (x_max_cam - x_min_cam) * 0.5;
+ real_t half_y = (y_max_cam - y_min_cam) * 0.5;
+
+ ortho_camera.set_orthogonal(-half_x, half_x, -half_y, half_y, 0, (z_max - z_min_cam));
+
+ Vector2 uv_scale(1.0 / (x_max_cam - x_min_cam), 1.0 / (y_max_cam - y_min_cam));
+
+ Transform ortho_transform;
+ ortho_transform.basis = transform.basis;
+ ortho_transform.origin = x_vec * (x_min_cam + half_x) + y_vec * (y_min_cam + half_y) + z_vec * z_max;
+
+ cull.shadows[p_shadow_index].cascades[i].frustum = Frustum(light_frustum_planes);
+ cull.shadows[p_shadow_index].cascades[i].projection = ortho_camera;
+ cull.shadows[p_shadow_index].cascades[i].transform = ortho_transform;
+ cull.shadows[p_shadow_index].cascades[i].zfar = z_max - z_min_cam;
+ cull.shadows[p_shadow_index].cascades[i].split = distances[i + 1];
+ cull.shadows[p_shadow_index].cascades[i].shadow_texel_size = radius * 2.0 / texture_size;
+ cull.shadows[p_shadow_index].cascades[i].bias_scale = bias_scale * aspect_bias_scale * min_distance_bias_scale;
+ cull.shadows[p_shadow_index].cascades[i].range_begin = z_max;
+ cull.shadows[p_shadow_index].cascades[i].uv_scale = uv_scale;
+ }
+ }
+}
- RSG::storage->update_mesh_instances();
+bool RendererSceneCull::_light_instance_update_shadow(Instance *p_instance, const Transform p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_orthogonal, bool p_cam_vaspect, RID p_shadow_atlas, Scenario *p_scenario, float p_screen_lod_threshold) {
+ InstanceLightData *light = static_cast<InstanceLightData *>(p_instance->base_data);
- scene_render->render_shadow(light->instance, p_shadow_atlas, i, (RendererSceneRender::InstanceBase **)instance_shadow_cull_result, cull_count, camera_plane, p_cam_projection.get_lod_multiplier(), p_screen_lod_threshold);
- }
+ Transform light_transform = p_instance->transform;
+ light_transform.orthonormalize(); //scale does not count on lights
+
+ bool animated_material_found = false;
+ switch (RSG::storage->light_get_type(p_instance->base)) {
+ case RS::LIGHT_DIRECTIONAL: {
} break;
case RS::LIGHT_OMNI: {
RS::LightOmniShadowMode shadow_mode = RSG::storage->light_omni_get_shadow_mode(p_instance->base);
@@ -1736,33 +1814,48 @@ bool RendererSceneCull::_light_instance_update_shadow(Instance *p_instance, cons
planes.write[4] = light_transform.xform(Plane(Vector3(0, -1, z).normalized(), radius));
planes.write[5] = light_transform.xform(Plane(Vector3(0, 0, -z), 0));
- int cull_count = p_scenario->octree.cull_convex(planes, instance_shadow_cull_result, MAX_INSTANCE_CULL, RS::INSTANCE_GEOMETRY_MASK);
+ geometry_instances_to_shadow_render.clear();
+ instance_shadow_cull_result.clear();
+
+ Vector<Vector3> points = Geometry3D::compute_convex_mesh_points(&planes[0], planes.size());
+
+ struct CullConvex {
+ PagedArray<Instance *> *result;
+ _FORCE_INLINE_ bool operator()(void *p_data) {
+ Instance *p_instance = (Instance *)p_data;
+ result->push_back(p_instance);
+ return false;
+ }
+ };
+
+ CullConvex cull_convex;
+ cull_convex.result = &instance_shadow_cull_result;
+
+ p_scenario->indexers[Scenario::INDEXER_GEOMETRY].convex_query(planes.ptr(), planes.size(), points.ptr(), points.size(), cull_convex);
+
Plane near_plane(light_transform.origin, light_transform.basis.get_axis(2) * z);
- for (int j = 0; j < cull_count; j++) {
+ for (int j = 0; j < (int)instance_shadow_cull_result.size(); j++) {
Instance *instance = instance_shadow_cull_result[j];
if (!instance->visible || !((1 << instance->base_type) & RS::INSTANCE_GEOMETRY_MASK) || !static_cast<InstanceGeometryData *>(instance->base_data)->can_cast_shadows) {
- cull_count--;
- SWAP(instance_shadow_cull_result[j], instance_shadow_cull_result[cull_count]);
- j--;
+ continue;
} else {
if (static_cast<InstanceGeometryData *>(instance->base_data)->material_is_animated) {
animated_material_found = true;
}
- instance->depth = near_plane.distance_to(instance->transform.origin);
- instance->depth_layer = 0;
-
if (instance->mesh_instance.is_valid()) {
RSG::storage->mesh_instance_check_for_update(instance->mesh_instance);
}
}
+
+ geometry_instances_to_shadow_render.push_back(instance);
}
RSG::storage->update_mesh_instances();
scene_render->light_instance_set_shadow_transform(light->instance, CameraMatrix(), light_transform, radius, 0, i, 0);
- scene_render->render_shadow(light->instance, p_shadow_atlas, i, (RendererSceneRender::InstanceBase **)instance_shadow_cull_result, cull_count);
+ scene_render->render_shadow(light->instance, p_shadow_atlas, i, geometry_instances_to_shadow_render);
}
} else { //shadow cube
@@ -1795,30 +1888,46 @@ bool RendererSceneCull::_light_instance_update_shadow(Instance *p_instance, cons
Vector<Plane> planes = cm.get_projection_planes(xform);
- int cull_count = p_scenario->octree.cull_convex(planes, instance_shadow_cull_result, MAX_INSTANCE_CULL, RS::INSTANCE_GEOMETRY_MASK);
+ geometry_instances_to_shadow_render.clear();
+ instance_shadow_cull_result.clear();
+
+ Vector<Vector3> points = Geometry3D::compute_convex_mesh_points(&planes[0], planes.size());
+
+ struct CullConvex {
+ PagedArray<Instance *> *result;
+ _FORCE_INLINE_ bool operator()(void *p_data) {
+ Instance *p_instance = (Instance *)p_data;
+ result->push_back(p_instance);
+ return false;
+ }
+ };
+
+ CullConvex cull_convex;
+ cull_convex.result = &instance_shadow_cull_result;
+
+ p_scenario->indexers[Scenario::INDEXER_GEOMETRY].convex_query(planes.ptr(), planes.size(), points.ptr(), points.size(), cull_convex);
Plane near_plane(xform.origin, -xform.basis.get_axis(2));
- for (int j = 0; j < cull_count; j++) {
+
+ for (int j = 0; j < (int)instance_shadow_cull_result.size(); j++) {
Instance *instance = instance_shadow_cull_result[j];
if (!instance->visible || !((1 << instance->base_type) & RS::INSTANCE_GEOMETRY_MASK) || !static_cast<InstanceGeometryData *>(instance->base_data)->can_cast_shadows) {
- cull_count--;
- SWAP(instance_shadow_cull_result[j], instance_shadow_cull_result[cull_count]);
- j--;
+ continue;
} else {
if (static_cast<InstanceGeometryData *>(instance->base_data)->material_is_animated) {
animated_material_found = true;
}
- instance->depth = near_plane.distance_to(instance->transform.origin);
- instance->depth_layer = 0;
if (instance->mesh_instance.is_valid()) {
RSG::storage->mesh_instance_check_for_update(instance->mesh_instance);
}
}
+
+ geometry_instances_to_shadow_render.push_back(instance);
}
RSG::storage->update_mesh_instances();
scene_render->light_instance_set_shadow_transform(light->instance, cm, xform, radius, 0, i, 0);
- scene_render->render_shadow(light->instance, p_shadow_atlas, i, (RendererSceneRender::InstanceBase **)instance_shadow_cull_result, cull_count);
+ scene_render->render_shadow(light->instance, p_shadow_atlas, i, geometry_instances_to_shadow_render);
}
//restore the regular DP matrix
@@ -1836,32 +1945,48 @@ bool RendererSceneCull::_light_instance_update_shadow(Instance *p_instance, cons
cm.set_perspective(angle * 2.0, 1.0, 0.01, radius);
Vector<Plane> planes = cm.get_projection_planes(light_transform);
- int cull_count = p_scenario->octree.cull_convex(planes, instance_shadow_cull_result, MAX_INSTANCE_CULL, RS::INSTANCE_GEOMETRY_MASK);
+
+ geometry_instances_to_shadow_render.clear();
+ instance_shadow_cull_result.clear();
+
+ Vector<Vector3> points = Geometry3D::compute_convex_mesh_points(&planes[0], planes.size());
+
+ struct CullConvex {
+ PagedArray<Instance *> *result;
+ _FORCE_INLINE_ bool operator()(void *p_data) {
+ Instance *p_instance = (Instance *)p_data;
+ result->push_back(p_instance);
+ return false;
+ }
+ };
+
+ CullConvex cull_convex;
+ cull_convex.result = &instance_shadow_cull_result;
+
+ p_scenario->indexers[Scenario::INDEXER_GEOMETRY].convex_query(planes.ptr(), planes.size(), points.ptr(), points.size(), cull_convex);
Plane near_plane(light_transform.origin, -light_transform.basis.get_axis(2));
- for (int j = 0; j < cull_count; j++) {
+
+ for (int j = 0; j < (int)instance_shadow_cull_result.size(); j++) {
Instance *instance = instance_shadow_cull_result[j];
if (!instance->visible || !((1 << instance->base_type) & RS::INSTANCE_GEOMETRY_MASK) || !static_cast<InstanceGeometryData *>(instance->base_data)->can_cast_shadows) {
- cull_count--;
- SWAP(instance_shadow_cull_result[j], instance_shadow_cull_result[cull_count]);
- j--;
+ continue;
} else {
if (static_cast<InstanceGeometryData *>(instance->base_data)->material_is_animated) {
animated_material_found = true;
}
- instance->depth = near_plane.distance_to(instance->transform.origin);
- instance->depth_layer = 0;
if (instance->mesh_instance.is_valid()) {
RSG::storage->mesh_instance_check_for_update(instance->mesh_instance);
}
}
+ geometry_instances_to_shadow_render.push_back(instance);
}
RSG::storage->update_mesh_instances();
scene_render->light_instance_set_shadow_transform(light->instance, cm, light_transform, radius, 0, 0, 0);
- scene_render->render_shadow(light->instance, p_shadow_atlas, 0, (RendererSceneRender::InstanceBase **)instance_shadow_cull_result, cull_count);
+ scene_render->render_shadow(light->instance, p_shadow_atlas, 0, geometry_instances_to_shadow_render);
} break;
}
@@ -2008,10 +2133,11 @@ void RendererSceneCull::_prepare_scene(const Transform p_cam_transform, const Ca
// - p_cam_transform will be a transform in the middle of our two eyes
// - p_cam_projection is a wider frustrum that encompasses both eyes
+ Instance *render_reflection_probe = instance_owner.getornull(p_reflection_probe); //if null, not rendering to it
+
Scenario *scenario = scenario_owner.getornull(p_scenario);
render_pass++;
- uint32_t camera_layer_mask = p_visible_layers;
scene_render->set_scene_pass(render_pass);
@@ -2026,250 +2152,360 @@ void RendererSceneCull::_prepare_scene(const Transform p_cam_transform, const Ca
Vector<Plane> planes = p_cam_projection.get_projection_planes(p_cam_transform);
Plane near_plane(p_cam_transform.origin, -p_cam_transform.basis.get_axis(2).normalized());
- float z_far = p_cam_projection.get_z_far();
+
+ uint64_t frame_number = RSG::rasterizer->get_frame_number();
+ float lightmap_probe_update_speed = RSG::storage->lightmap_get_probe_capture_update_speed() * RSG::rasterizer->get_frame_delta_time();
/* STEP 2 - CULL */
- instance_cull_count = scenario->octree.cull_convex(planes, instance_cull_result, MAX_INSTANCE_CULL);
- light_cull_count = 0;
- reflection_probe_cull_count = 0;
- decal_cull_count = 0;
- gi_probe_cull_count = 0;
- lightmap_cull_count = 0;
+ cull.frustum = Frustum(planes);
- //light_samplers_culled=0;
+ Vector<RID> directional_lights;
+ // directional lights
+ {
+ //reset shadows
+ for (int i = 0; i < RendererSceneRender::MAX_DIRECTIONAL_LIGHTS; i++) {
+ for (int j = 0; j < RendererSceneRender::MAX_DIRECTIONAL_LIGHT_CASCADES; j++) {
+ cull.shadows[i].cascades[j].cull_result.clear();
+ }
+ }
- /*
- print_line("OT: "+rtos( (OS::get_singleton()->get_ticks_usec()-t)/1000.0));
- print_line("OTO: "+itos(p_scenario->octree.get_octant_count()));
- print_line("OTE: "+itos(p_scenario->octree.get_elem_count()));
- print_line("OTP: "+itos(p_scenario->octree.get_pair_count()));
- */
+ cull.shadow_count = 0;
- /* STEP 3 - PROCESS PORTALS, VALIDATE ROOMS */
- //removed, will replace with culling
+ Vector<Instance *> lights_with_shadow;
- /* STEP 4 - REMOVE FURTHER CULLED OBJECTS, ADD LIGHTS */
- uint64_t frame_number = RSG::rasterizer->get_frame_number();
- float lightmap_probe_update_speed = RSG::storage->lightmap_get_probe_capture_update_speed() * RSG::rasterizer->get_frame_delta_time();
+ for (List<Instance *>::Element *E = scenario->directional_lights.front(); E; E = E->next()) {
+ if (!E->get()->visible) {
+ continue;
+ }
+
+ if (directional_lights.size() > RendererSceneRender::MAX_DIRECTIONAL_LIGHTS) {
+ break;
+ }
+
+ InstanceLightData *light = static_cast<InstanceLightData *>(E->get()->base_data);
- for (int i = 0; i < instance_cull_count; i++) {
- Instance *ins = instance_cull_result[i];
+ //check shadow..
- bool keep = false;
+ if (light) {
+ if (p_using_shadows && p_shadow_atlas.is_valid() && RSG::storage->light_has_shadow(E->get()->base) && !(RSG::storage->light_get_type(E->get()->base) == RS::LIGHT_DIRECTIONAL && RSG::storage->light_directional_is_sky_only(E->get()->base))) {
+ lights_with_shadow.push_back(E->get());
+ }
+ //add to list
+ directional_lights.push_back(light->instance);
+ }
+ }
- if ((camera_layer_mask & ins->layer_mask) == 0) {
- //failure
- } else if (ins->base_type == RS::INSTANCE_LIGHT && ins->visible) {
- if (light_cull_count < MAX_LIGHTS_CULLED) {
- InstanceLightData *light = static_cast<InstanceLightData *>(ins->base_data);
+ scene_render->set_directional_shadow_count(lights_with_shadow.size());
- if (!light->geometries.empty()) {
- //do not add this light if no geometry is affected by it..
- light_cull_result[light_cull_count] = ins;
- light_instance_cull_result[light_cull_count] = light->instance;
- if (p_shadow_atlas.is_valid() && RSG::storage->light_has_shadow(ins->base)) {
- scene_render->light_instance_mark_visible(light->instance); //mark it visible for shadow allocation later
- }
+ for (int i = 0; i < lights_with_shadow.size(); i++) {
+ _light_instance_setup_directional_shadow(i, lights_with_shadow[i], p_cam_transform, p_cam_projection, p_cam_orthogonal, p_cam_vaspect);
+ }
+ }
+
+ { //sdfgi
+ cull.sdfgi.region_count = 0;
- light_cull_count++;
+ for (int i = 0; i < SDFGI_MAX_CASCADES * SDFGI_MAX_REGIONS_PER_CASCADE; i++) {
+ cull.sdfgi.region_cull_result[i].clear();
+ }
+
+ for (int i = 0; i < SDFGI_MAX_CASCADES; i++) {
+ cull.sdfgi.cascade_lights[i].clear();
+ }
+
+ if (p_render_buffers.is_valid()) {
+ for (int i = 0; i < SDFGI_MAX_CASCADES; i++) {
+ cull.sdfgi.cascade_lights[i].clear();
+ }
+ cull.sdfgi.cascade_light_count = 0;
+
+ uint32_t prev_cascade = 0xFFFFFFFF;
+ uint32_t pending_region_count = scene_render->sdfgi_get_pending_region_count(p_render_buffers);
+
+ for (uint32_t i = 0; i < pending_region_count; i++) {
+ cull.sdfgi.region_aabb[i] = scene_render->sdfgi_get_pending_region_bounds(p_render_buffers, i);
+ uint32_t region_cascade = scene_render->sdfgi_get_pending_region_cascade(p_render_buffers, i);
+ cull.sdfgi.region_cascade[i] = region_cascade;
+
+ if (region_cascade != prev_cascade) {
+ cull.sdfgi.cascade_light_index[cull.sdfgi.cascade_light_count] = region_cascade;
+ cull.sdfgi.cascade_light_count++;
+ prev_cascade = region_cascade;
}
}
- } else if (ins->base_type == RS::INSTANCE_REFLECTION_PROBE && ins->visible) {
- if (reflection_probe_cull_count < MAX_REFLECTION_PROBES_CULLED) {
- InstanceReflectionProbeData *reflection_probe = static_cast<InstanceReflectionProbeData *>(ins->base_data);
- if (p_reflection_probe != reflection_probe->instance) {
- //avoid entering The Matrix
+ cull.sdfgi.region_count = pending_region_count;
+ }
+ }
- if (!reflection_probe->geometries.empty()) {
- //do not add this light if no geometry is affected by it..
+ {
+ //pre-clear results
+ geometry_instances_to_render.clear();
+ light_cull_result.clear();
+ lightmap_cull_result.clear();
+ reflection_probe_instance_cull_result.clear();
+ light_instance_cull_result.clear();
+ gi_probe_instance_cull_result.clear();
+ lightmap_cull_result.clear();
+ decal_instance_cull_result.clear();
+ mesh_instance_cull_result.clear();
+ }
+
+ {
+ uint64_t cull_count = scenario->instance_data.size();
+ uint32_t sdfgi_last_light_index = 0xFFFFFFFF;
+ uint32_t sdfgi_last_light_cascade = 0xFFFFFFFF;
+
+ for (uint64_t i = 0; i < cull_count; i++) {
+ bool mesh_visible = false;
+
+ if (scenario->instance_aabbs[i].in_frustum(cull.frustum)) {
+ InstanceData &idata = scenario->instance_data[i];
+ uint32_t base_type = idata.flags & InstanceData::FLAG_BASE_TYPE_MASK;
+
+ if ((p_visible_layers & idata.layer_mask) == 0) {
+ //failure
+ } else if (base_type == RS::INSTANCE_LIGHT) {
+ light_cull_result.push_back(idata.instance);
+ light_instance_cull_result.push_back(idata.instance_data_rid);
+ if (p_shadow_atlas.is_valid() && RSG::storage->light_has_shadow(idata.base_rid)) {
+ scene_render->light_instance_mark_visible(idata.instance_data_rid); //mark it visible for shadow allocation later
+ }
- if (reflection_probe->reflection_dirty || scene_render->reflection_probe_instance_needs_redraw(reflection_probe->instance)) {
+ } else if (base_type == RS::INSTANCE_REFLECTION_PROBE) {
+ if (render_reflection_probe != idata.instance) {
+ //avoid entering The Matrix
+
+ if ((idata.flags & InstanceData::FLAG_REFLECTION_PROBE_DIRTY) || scene_render->reflection_probe_instance_needs_redraw(idata.instance_data_rid)) {
+ InstanceReflectionProbeData *reflection_probe = static_cast<InstanceReflectionProbeData *>(idata.instance->base_data);
+ cull.lock.lock();
if (!reflection_probe->update_list.in_list()) {
reflection_probe->render_step = 0;
reflection_probe_render_list.add_last(&reflection_probe->update_list);
}
+ cull.lock.unlock();
- reflection_probe->reflection_dirty = false;
+ idata.flags &= ~uint32_t(InstanceData::FLAG_REFLECTION_PROBE_DIRTY);
}
- if (scene_render->reflection_probe_instance_has_reflection(reflection_probe->instance)) {
- reflection_probe_instance_cull_result[reflection_probe_cull_count] = reflection_probe->instance;
- reflection_probe_cull_count++;
+ if (scene_render->reflection_probe_instance_has_reflection(idata.instance_data_rid)) {
+ reflection_probe_instance_cull_result.push_back(idata.instance_data_rid);
}
}
- }
- }
- } else if (ins->base_type == RS::INSTANCE_DECAL && ins->visible) {
- if (decal_cull_count < MAX_DECALS_CULLED) {
- InstanceDecalData *decal = static_cast<InstanceDecalData *>(ins->base_data);
+ } else if (base_type == RS::INSTANCE_DECAL) {
+ decal_instance_cull_result.push_back(idata.instance_data_rid);
+
+ } else if (base_type == RS::INSTANCE_GI_PROBE) {
+ InstanceGIProbeData *gi_probe = static_cast<InstanceGIProbeData *>(idata.instance->base_data);
+ cull.lock.lock();
+ if (!gi_probe->update_element.in_list()) {
+ gi_probe_update_list.add(&gi_probe->update_element);
+ }
+ cull.lock.unlock();
+ gi_probe_instance_cull_result.push_back(idata.instance_data_rid);
- if (!decal->geometries.empty()) {
- //do not add this decal if no geometry is affected by it..
- decal_instance_cull_result[decal_cull_count] = decal->instance;
- decal_cull_count++;
- }
- }
+ } else if (base_type == RS::INSTANCE_LIGHTMAP) {
+ lightmap_cull_result.push_back(idata.instance);
+ } else if (((1 << base_type) & RS::INSTANCE_GEOMETRY_MASK) && !(idata.flags & InstanceData::FLAG_CAST_SHADOWS_ONLY)) {
+ bool keep = true;
- } else if (ins->base_type == RS::INSTANCE_GI_PROBE && ins->visible) {
- InstanceGIProbeData *gi_probe = static_cast<InstanceGIProbeData *>(ins->base_data);
- if (!gi_probe->update_element.in_list()) {
- gi_probe_update_list.add(&gi_probe->update_element);
- }
+ if (idata.flags & InstanceData::FLAG_REDRAW_IF_VISIBLE) {
+ RenderingServerDefault::redraw_request();
+ }
- if (gi_probe_cull_count < MAX_GI_PROBES_CULLED) {
- gi_probe_instance_cull_result[gi_probe_cull_count] = gi_probe->probe_instance;
- gi_probe_cull_count++;
- }
- } else if (ins->base_type == RS::INSTANCE_LIGHTMAP && ins->visible) {
- if (lightmap_cull_count < MAX_LIGHTMAPS_CULLED) {
- lightmap_cull_result[lightmap_cull_count] = ins;
- lightmap_cull_count++;
- }
+ if (base_type == RS::INSTANCE_MESH) {
+ mesh_visible = true;
+ } else if (base_type == RS::INSTANCE_PARTICLES) {
+ //particles visible? process them
+ if (RSG::storage->particles_is_inactive(idata.base_rid)) {
+ //but if nothing is going on, don't do it.
+ keep = false;
+ } else {
+ cull.lock.lock();
+ RSG::storage->particles_request_process(idata.base_rid);
+ cull.lock.unlock();
+ RSG::storage->particles_set_view_axis(idata.base_rid, -p_cam_transform.basis.get_axis(2).normalized());
+ //particles visible? request redraw
+ RenderingServerDefault::redraw_request();
+ }
+ }
- } else if (((1 << ins->base_type) & RS::INSTANCE_GEOMETRY_MASK) && ins->visible && ins->cast_shadows != RS::SHADOW_CASTING_SETTING_SHADOWS_ONLY) {
- keep = true;
+ if (pair_volumes_to_mesh && (idata.flags & InstanceData::FLAG_GEOM_LIGHTING_DIRTY)) {
+ InstanceGeometryData *geom = static_cast<InstanceGeometryData *>(idata.instance->base_data);
+ int l = 0;
+ //only called when lights AABB enter/exit this geometry
+ idata.instance->light_instances.resize(geom->lights.size());
- InstanceGeometryData *geom = static_cast<InstanceGeometryData *>(ins->base_data);
+ for (Set<Instance *>::Element *E = geom->lights.front(); E; E = E->next()) {
+ InstanceLightData *light = static_cast<InstanceLightData *>(E->get()->base_data);
- if (ins->redraw_if_visible) {
- RenderingServerDefault::redraw_request();
- }
+ idata.instance->light_instances.write[l++] = light->instance;
+ }
- if (ins->base_type == RS::INSTANCE_PARTICLES) {
- //particles visible? process them
- if (RSG::storage->particles_is_inactive(ins->base)) {
- //but if nothing is going on, don't do it.
- keep = false;
- } else {
- RSG::storage->particles_request_process(ins->base);
- RSG::storage->particles_set_view_axis(ins->base, -p_cam_transform.basis.get_axis(2).normalized());
- //particles visible? request redraw
- RenderingServerDefault::redraw_request();
- }
- }
+ idata.flags &= ~uint32_t(InstanceData::FLAG_GEOM_LIGHTING_DIRTY);
+ }
- if (geom->lighting_dirty) {
- int l = 0;
- //only called when lights AABB enter/exit this geometry
- ins->light_instances.resize(geom->lighting.size());
+ if (pair_volumes_to_mesh && (idata.flags & InstanceData::FLAG_GEOM_REFLECTION_DIRTY)) {
+ InstanceGeometryData *geom = static_cast<InstanceGeometryData *>(idata.instance->base_data);
+ int l = 0;
+ //only called when reflection probe AABB enter/exit this geometry
+ idata.instance->reflection_probe_instances.resize(geom->reflection_probes.size());
- for (List<Instance *>::Element *E = geom->lighting.front(); E; E = E->next()) {
- InstanceLightData *light = static_cast<InstanceLightData *>(E->get()->base_data);
+ for (Set<Instance *>::Element *E = geom->reflection_probes.front(); E; E = E->next()) {
+ InstanceReflectionProbeData *reflection_probe = static_cast<InstanceReflectionProbeData *>(E->get()->base_data);
- ins->light_instances.write[l++] = light->instance;
- }
+ idata.instance->reflection_probe_instances.write[l++] = reflection_probe->instance;
+ }
- geom->lighting_dirty = false;
- }
+ idata.flags &= ~uint32_t(InstanceData::FLAG_GEOM_REFLECTION_DIRTY);
+ }
- if (geom->reflection_dirty) {
- int l = 0;
- //only called when reflection probe AABB enter/exit this geometry
- ins->reflection_probe_instances.resize(geom->reflection_probes.size());
+ if (pair_volumes_to_mesh && (idata.flags & InstanceData::FLAG_GEOM_DECAL_DIRTY)) {
+ //InstanceGeometryData *geom = static_cast<InstanceGeometryData *>(idata.instance->base_data);
+ //todo for GLES3
+ idata.flags &= ~uint32_t(InstanceData::FLAG_GEOM_DECAL_DIRTY);
+ }
- for (List<Instance *>::Element *E = geom->reflection_probes.front(); E; E = E->next()) {
- InstanceReflectionProbeData *reflection_probe = static_cast<InstanceReflectionProbeData *>(E->get()->base_data);
+ if (idata.flags & InstanceData::FLAG_GEOM_GI_PROBE_DIRTY) {
+ InstanceGeometryData *geom = static_cast<InstanceGeometryData *>(idata.instance->base_data);
+ int l = 0;
+ //only called when reflection probe AABB enter/exit this geometry
+ idata.instance->gi_probe_instances.resize(geom->gi_probes.size());
- ins->reflection_probe_instances.write[l++] = reflection_probe->instance;
- }
+ for (Set<Instance *>::Element *E = geom->gi_probes.front(); E; E = E->next()) {
+ InstanceGIProbeData *gi_probe = static_cast<InstanceGIProbeData *>(E->get()->base_data);
- geom->reflection_dirty = false;
- }
+ idata.instance->gi_probe_instances.write[l++] = gi_probe->probe_instance;
+ }
- if (geom->gi_probes_dirty) {
- int l = 0;
- //only called when reflection probe AABB enter/exit this geometry
- ins->gi_probe_instances.resize(geom->gi_probes.size());
+ idata.flags &= ~uint32_t(InstanceData::FLAG_GEOM_GI_PROBE_DIRTY);
+ }
- for (List<Instance *>::Element *E = geom->gi_probes.front(); E; E = E->next()) {
- InstanceGIProbeData *gi_probe = static_cast<InstanceGIProbeData *>(E->get()->base_data);
+ if ((idata.flags & InstanceData::FLAG_LIGHTMAP_CAPTURE) && idata.instance->last_frame_pass != frame_number && !idata.instance->lightmap_target_sh.is_empty() && !idata.instance->lightmap_sh.is_empty()) {
+ Color *sh = idata.instance->lightmap_sh.ptrw();
+ const Color *target_sh = idata.instance->lightmap_target_sh.ptr();
+ for (uint32_t j = 0; j < 9; j++) {
+ sh[j] = sh[j].lerp(target_sh[j], MIN(1.0, lightmap_probe_update_speed));
+ }
+ idata.instance->last_frame_pass = frame_number;
+ }
- ins->gi_probe_instances.write[l++] = gi_probe->probe_instance;
+ if (keep) {
+ geometry_instances_to_render.push_back(idata.instance);
+ }
}
-
- geom->gi_probes_dirty = false;
}
- if (ins->last_frame_pass != frame_number && !ins->lightmap_target_sh.empty() && !ins->lightmap_sh.empty()) {
- Color *sh = ins->lightmap_sh.ptrw();
- const Color *target_sh = ins->lightmap_target_sh.ptr();
- for (uint32_t j = 0; j < 9; j++) {
- sh[j] = sh[j].lerp(target_sh[j], MIN(1.0, lightmap_probe_update_speed));
+ for (uint32_t j = 0; j < cull.shadow_count; j++) {
+ for (uint32_t k = 0; k < cull.shadows[j].cascade_count; k++) {
+ if (scenario->instance_aabbs[i].in_frustum(cull.shadows[j].cascades[k].frustum)) {
+ InstanceData &idata = scenario->instance_data[i];
+ uint32_t base_type = idata.flags & InstanceData::FLAG_BASE_TYPE_MASK;
+
+ if (((1 << base_type) & RS::INSTANCE_GEOMETRY_MASK) && idata.flags & InstanceData::FLAG_CAST_SHADOWS) {
+ cull.shadows[j].cascades[k].cull_result.push_back(idata.instance);
+ mesh_visible = true;
+ }
+ }
}
}
- if (ins->mesh_instance.is_valid()) {
- RSG::storage->mesh_instance_check_for_update(ins->mesh_instance);
+ for (uint32_t j = 0; j < cull.sdfgi.region_count; j++) {
+ if (scenario->instance_aabbs[i].in_aabb(cull.sdfgi.region_aabb[j])) {
+ InstanceData &idata = scenario->instance_data[i];
+ uint32_t base_type = idata.flags & InstanceData::FLAG_BASE_TYPE_MASK;
+
+ if (base_type == RS::INSTANCE_LIGHT) {
+ InstanceLightData *instance_light = (InstanceLightData *)idata.instance->base_data;
+ if (instance_light->bake_mode == RS::LIGHT_BAKE_STATIC && cull.sdfgi.region_cascade[j] <= instance_light->max_sdfgi_cascade) {
+ if (sdfgi_last_light_index != i || sdfgi_last_light_cascade != cull.sdfgi.region_cascade[j]) {
+ sdfgi_last_light_index = i;
+ sdfgi_last_light_cascade = cull.sdfgi.region_cascade[j];
+ cull.sdfgi.cascade_lights[sdfgi_last_light_cascade].push_back(instance_light->instance);
+ }
+ }
+ } else if ((1 << base_type) & RS::INSTANCE_GEOMETRY_MASK) {
+ if (idata.flags & InstanceData::FLAG_USES_BAKED_LIGHT) {
+ cull.sdfgi.region_cull_result[j].push_back(idata.instance);
+ mesh_visible = true;
+ }
+ }
+ }
}
- ins->depth = near_plane.distance_to(ins->transform.origin);
- ins->depth_layer = CLAMP(int(ins->depth * 16 / z_far), 0, 15);
+ if (mesh_visible && scenario->instance_data[i].flags & InstanceData::FLAG_USES_MESH_INSTANCE) {
+ mesh_instance_cull_result.push_back(scenario->instance_data[i].instance->mesh_instance);
+ }
}
- if (!keep) {
- // remove, no reason to keep
- instance_cull_count--;
- SWAP(instance_cull_result[i], instance_cull_result[instance_cull_count]);
- i--;
- ins->last_render_pass = 0; // make invalid
- } else {
- ins->last_render_pass = render_pass;
+ if (mesh_instance_cull_result.size()) {
+ for (uint64_t i = 0; i < mesh_instance_cull_result.size(); i++) {
+ RSG::storage->mesh_instance_check_for_update(mesh_instance_cull_result[i]);
+ }
+ RSG::storage->update_mesh_instances();
}
- ins->last_frame_pass = frame_number;
}
- RSG::storage->update_mesh_instances();
+ //render shadows
- /* STEP 5 - PROCESS LIGHTS */
+ for (uint32_t i = 0; i < cull.shadow_count; i++) {
+ for (uint32_t j = 0; j < cull.shadows[i].cascade_count; j++) {
+ const Cull::Shadow::Cascade &c = cull.shadows[i].cascades[j];
+ // print_line("shadow " + itos(i) + " cascade " + itos(j) + " elements: " + itos(c.cull_result.size()));
+ scene_render->light_instance_set_shadow_transform(cull.shadows[i].light_instance, c.projection, c.transform, c.zfar, c.split, j, c.shadow_texel_size, c.bias_scale, c.range_begin, c.uv_scale);
+ scene_render->render_shadow(cull.shadows[i].light_instance, p_shadow_atlas, j, c.cull_result, near_plane, p_cam_projection.get_lod_multiplier(), p_screen_lod_threshold);
+ }
+ }
- RID *directional_light_ptr = &light_instance_cull_result[light_cull_count];
- directional_light_count = 0;
+ //render SDFGI
- // directional lights
{
- Instance **lights_with_shadow = (Instance **)alloca(sizeof(Instance *) * scenario->directional_lights.size());
- int directional_shadow_count = 0;
-
- for (List<Instance *>::Element *E = scenario->directional_lights.front(); E; E = E->next()) {
- if (light_cull_count + directional_light_count >= MAX_LIGHTS_CULLED) {
- break;
+ if (cull.sdfgi.region_count > 0) {
+ //update regions
+ for (uint32_t i = 0; i < cull.sdfgi.region_count; i++) {
+ scene_render->render_sdfgi(p_render_buffers, i, cull.sdfgi.region_cull_result[i]);
+ }
+ //check if static lights were culled
+ bool static_lights_culled = false;
+ for (uint32_t i = 0; i < cull.sdfgi.cascade_light_count; i++) {
+ if (cull.sdfgi.cascade_lights[i].size()) {
+ static_lights_culled = true;
+ break;
+ }
}
- if (!E->get()->visible) {
- continue;
+ if (static_lights_culled) {
+ scene_render->render_sdfgi_static_lights(p_render_buffers, cull.sdfgi.cascade_light_count, cull.sdfgi.cascade_light_index, cull.sdfgi.cascade_lights);
}
+ }
- InstanceLightData *light = static_cast<InstanceLightData *>(E->get()->base_data);
-
- //check shadow..
-
- if (light) {
- if (p_using_shadows && p_shadow_atlas.is_valid() && RSG::storage->light_has_shadow(E->get()->base) && !(RSG::storage->light_get_type(E->get()->base) == RS::LIGHT_DIRECTIONAL && RSG::storage->light_directional_is_sky_only(E->get()->base))) {
- lights_with_shadow[directional_shadow_count++] = E->get();
- }
- //add to list
- directional_light_ptr[directional_light_count++] = light->instance;
- }
+ if (p_render_buffers.is_valid()) {
+ scene_render->sdfgi_update_probes(p_render_buffers, p_environment, directional_lights, scenario->dynamic_lights.ptr(), scenario->dynamic_lights.size());
}
+ }
- scene_render->set_directional_shadow_count(directional_shadow_count);
+ //light_samplers_culled=0;
- for (int i = 0; i < directional_shadow_count; i++) {
- RENDER_TIMESTAMP(">Rendering Directional Light " + itos(i));
+ /*
+ print_line("OT: "+rtos( (OS::get_singleton()->get_ticks_usec()-t)/1000.0));
+ print_line("OTO: "+itos(p_scenario->octree.get_octant_count()));
+ print_line("OTE: "+itos(p_scenario->octree.get_elem_count()));
+ print_line("OTP: "+itos(p_scenario->octree.get_pair_count()));
+ */
- _light_instance_update_shadow(lights_with_shadow[i], p_cam_transform, p_cam_projection, p_cam_orthogonal, p_cam_vaspect, p_shadow_atlas, scenario, p_screen_lod_threshold);
+ /* STEP 3 - PROCESS PORTALS, VALIDATE ROOMS */
+ //removed, will replace with culling
- RENDER_TIMESTAMP("<Rendering Directional Light " + itos(i));
- }
- }
+ /* STEP 4 - REMOVE FURTHER CULLED OBJECTS, ADD LIGHTS */
+
+ /* STEP 5 - PROCESS POSITIONAL LIGHTS */
if (p_using_shadows) { //setup shadow maps
//SortArray<Instance*,_InstanceLightsort> sorter;
//sorter.sort(light_cull_result,light_cull_count);
- for (int i = 0; i < light_cull_count; i++) {
+ for (uint32_t i = 0; i < (uint32_t)light_cull_result.size(); i++) {
Instance *ins = light_cull_result[i];
if (!p_shadow_atlas.is_valid() || !RSG::storage->light_has_shadow(ins->base)) {
@@ -2364,74 +2600,9 @@ void RendererSceneCull::_prepare_scene(const Transform p_cam_transform, const Ca
}
}
- /* UPDATE SDFGI */
-
- if (p_render_buffers.is_valid()) {
- uint32_t cascade_index[8];
- uint32_t cascade_sizes[8];
- const RID *cascade_ptrs[8];
- uint32_t cascade_count = 0;
- uint32_t sdfgi_light_cull_count = 0;
-
- uint32_t prev_cascade = 0xFFFFFFFF;
- for (int i = 0; i < scene_render->sdfgi_get_pending_region_count(p_render_buffers); i++) {
- AABB region = scene_render->sdfgi_get_pending_region_bounds(p_render_buffers, i);
- uint32_t region_cascade = scene_render->sdfgi_get_pending_region_cascade(p_render_buffers, i);
-
- if (region_cascade != prev_cascade) {
- cascade_sizes[cascade_count] = 0;
- cascade_index[cascade_count] = region_cascade;
- cascade_ptrs[cascade_count] = &sdfgi_light_cull_result[sdfgi_light_cull_count];
- cascade_count++;
- sdfgi_light_cull_pass++;
- prev_cascade = region_cascade;
- }
- uint32_t sdfgi_cull_count = scenario->octree.cull_aabb(region, instance_shadow_cull_result, MAX_INSTANCE_CULL);
-
- for (uint32_t j = 0; j < sdfgi_cull_count; j++) {
- Instance *ins = instance_shadow_cull_result[j];
-
- bool keep = false;
-
- if (ins->base_type == RS::INSTANCE_LIGHT && ins->visible) {
- InstanceLightData *instance_light = (InstanceLightData *)ins->base_data;
- if (instance_light->bake_mode != RS::LIGHT_BAKE_STATIC || region_cascade > instance_light->max_sdfgi_cascade) {
- continue;
- }
-
- if (sdfgi_light_cull_pass != instance_light->sdfgi_cascade_light_pass && sdfgi_light_cull_count < MAX_LIGHTS_CULLED) {
- instance_light->sdfgi_cascade_light_pass = sdfgi_light_cull_pass;
- sdfgi_light_cull_result[sdfgi_light_cull_count++] = instance_light->instance;
- cascade_sizes[cascade_count - 1]++;
- }
- } else if ((1 << ins->base_type) & RS::INSTANCE_GEOMETRY_MASK) {
- if (ins->baked_light) {
- keep = true;
- if (ins->mesh_instance.is_valid()) {
- RSG::storage->mesh_instance_check_for_update(ins->mesh_instance);
- }
- }
- }
-
- if (!keep) {
- // remove, no reason to keep
- sdfgi_cull_count--;
- SWAP(instance_shadow_cull_result[j], instance_shadow_cull_result[sdfgi_cull_count]);
- j--;
- }
- }
-
- RSG::storage->update_mesh_instances();
-
- scene_render->render_sdfgi(p_render_buffers, i, (RendererSceneRender::InstanceBase **)instance_shadow_cull_result, sdfgi_cull_count);
- //have to save updated cascades, then update static lights.
- }
-
- if (sdfgi_light_cull_count) {
- scene_render->render_sdfgi_static_lights(p_render_buffers, cascade_count, cascade_index, cascade_ptrs, cascade_sizes);
- }
-
- scene_render->sdfgi_update_probes(p_render_buffers, p_environment, directional_light_ptr, directional_light_count, scenario->dynamic_lights.ptr(), scenario->dynamic_lights.size());
+ //append the directional lights to the lights culled
+ for (int i = 0; i < directional_lights.size(); i++) {
+ light_instance_cull_result.push_back(directional_lights[i]);
}
}
@@ -2468,7 +2639,7 @@ void RendererSceneCull::_render_scene(RID p_render_buffers, const Transform p_ca
/* PROCESS GEOMETRY AND DRAW SCENE */
RENDER_TIMESTAMP("Render Scene ");
- scene_render->render_scene(p_render_buffers, p_cam_transform, p_cam_projection, p_cam_orthogonal, (RendererSceneRender::InstanceBase **)instance_cull_result, instance_cull_count, light_instance_cull_result, light_cull_count + directional_light_count, reflection_probe_instance_cull_result, reflection_probe_cull_count, gi_probe_instance_cull_result, gi_probe_cull_count, decal_instance_cull_result, decal_cull_count, (RendererSceneRender::InstanceBase **)lightmap_cull_result, lightmap_cull_count, p_environment, camera_effects, p_shadow_atlas, p_reflection_probe.is_valid() ? RID() : scenario->reflection_atlas, p_reflection_probe, p_reflection_probe_pass, p_screen_lod_threshold);
+ scene_render->render_scene(p_render_buffers, p_cam_transform, p_cam_projection, p_cam_orthogonal, geometry_instances_to_render, light_instance_cull_result, reflection_probe_instance_cull_result, gi_probe_instance_cull_result, decal_instance_cull_result, lightmap_cull_result, p_environment, camera_effects, p_shadow_atlas, p_reflection_probe.is_valid() ? RID() : scenario->reflection_atlas, p_reflection_probe, p_reflection_probe_pass, p_screen_lod_threshold);
}
void RendererSceneCull::render_empty_scene(RID p_render_buffers, RID p_scenario, RID p_shadow_atlas) {
@@ -2483,7 +2654,7 @@ void RendererSceneCull::render_empty_scene(RID p_render_buffers, RID p_scenario,
environment = scenario->fallback_environment;
}
RENDER_TIMESTAMP("Render Empty Scene ");
- scene_render->render_scene(p_render_buffers, Transform(), CameraMatrix(), true, nullptr, 0, nullptr, 0, nullptr, 0, nullptr, 0, nullptr, 0, nullptr, 0, environment, RID(), p_shadow_atlas, scenario->reflection_atlas, RID(), 0, 0);
+ scene_render->render_scene(p_render_buffers, Transform(), CameraMatrix(), true, PagedArray<RendererSceneRender::InstanceBase *>(), PagedArray<RID>(), PagedArray<RID>(), PagedArray<RID>(), PagedArray<RID>(), PagedArray<RendererSceneRender::InstanceBase *>(), RID(), RID(), p_shadow_atlas, scenario->reflection_atlas, RID(), 0, 0);
#endif
}
@@ -2758,35 +2929,34 @@ void RendererSceneCull::render_probes() {
update_lights = true;
}
- instance_cull_count = 0;
- for (List<InstanceGIProbeData::PairInfo>::Element *E = probe->dynamic_geometries.front(); E; E = E->next()) {
- if (instance_cull_count < MAX_INSTANCE_CULL) {
- Instance *ins = E->get().geometry;
- if (!ins->visible) {
- continue;
- }
- InstanceGeometryData *geom = (InstanceGeometryData *)ins->base_data;
+ geometry_instances_to_render.clear();
- if (geom->gi_probes_dirty) {
- //giprobes may be dirty, so update
- int l = 0;
- //only called when reflection probe AABB enter/exit this geometry
- ins->gi_probe_instances.resize(geom->gi_probes.size());
+ for (Set<Instance *>::Element *E = probe->dynamic_geometries.front(); E; E = E->next()) {
+ Instance *ins = E->get();
+ if (!ins->visible) {
+ continue;
+ }
+ InstanceGeometryData *geom = (InstanceGeometryData *)ins->base_data;
- for (List<Instance *>::Element *F = geom->gi_probes.front(); F; F = F->next()) {
- InstanceGIProbeData *gi_probe2 = static_cast<InstanceGIProbeData *>(F->get()->base_data);
+ if (ins->scenario && ins->array_index >= 0 && (ins->scenario->instance_data[ins->array_index].flags & InstanceData::FLAG_GEOM_GI_PROBE_DIRTY)) {
+ //giprobes may be dirty, so update
+ int l = 0;
+ //only called when reflection probe AABB enter/exit this geometry
+ ins->gi_probe_instances.resize(geom->gi_probes.size());
- ins->gi_probe_instances.write[l++] = gi_probe2->probe_instance;
- }
+ for (Set<Instance *>::Element *F = geom->gi_probes.front(); F; F = F->next()) {
+ InstanceGIProbeData *gi_probe2 = static_cast<InstanceGIProbeData *>(F->get()->base_data);
- geom->gi_probes_dirty = false;
+ ins->gi_probe_instances.write[l++] = gi_probe2->probe_instance;
}
- instance_cull_result[instance_cull_count++] = E->get().geometry;
+ ins->scenario->instance_data[ins->array_index].flags &= ~uint32_t(InstanceData::FLAG_GEOM_GI_PROBE_DIRTY);
}
+
+ geometry_instances_to_render.push_back(E->get());
}
- scene_render->gi_probe_update(probe->probe_instance, update_lights, probe->light_instances, instance_cull_count, (RendererSceneRender::InstanceBase **)instance_cull_result);
+ scene_render->gi_probe_update(probe->probe_instance, update_lights, probe->light_instances, geometry_instances_to_render);
gi_probe_update_list.remove(gi_probe);
@@ -2800,16 +2970,32 @@ void RendererSceneCull::render_particle_colliders() {
if (hfpc->scenario && hfpc->base_type == RS::INSTANCE_PARTICLES_COLLISION && RSG::storage->particles_collision_is_heightfield(hfpc->base)) {
//update heightfield
- int cull_count = hfpc->scenario->octree.cull_aabb(hfpc->transformed_aabb, instance_cull_result, MAX_INSTANCE_CULL); //@TODO: cull mask missing
- for (int i = 0; i < cull_count; i++) {
+ instance_cull_result.clear();
+ geometry_instances_to_render.clear();
+
+ struct CullAABB {
+ PagedArray<Instance *> *result;
+ _FORCE_INLINE_ bool operator()(void *p_data) {
+ Instance *p_instance = (Instance *)p_data;
+ result->push_back(p_instance);
+ return false;
+ }
+ };
+
+ CullAABB cull_aabb;
+ cull_aabb.result = &instance_cull_result;
+ hfpc->scenario->indexers[Scenario::INDEXER_GEOMETRY].aabb_query(hfpc->transformed_aabb, cull_aabb);
+ hfpc->scenario->indexers[Scenario::INDEXER_VOLUMES].aabb_query(hfpc->transformed_aabb, cull_aabb);
+
+ for (int i = 0; i < (int)instance_cull_result.size(); i++) {
Instance *instance = instance_cull_result[i];
- if (!instance->visible || !((1 << instance->base_type) & (RS::INSTANCE_GEOMETRY_MASK & (~(1 << RS::INSTANCE_PARTICLES))))) { //all but particles to avoid self collision
- cull_count--;
- SWAP(instance_cull_result[i], instance_cull_result[cull_count]);
+ if (!instance || !((1 << instance->base_type) & (RS::INSTANCE_GEOMETRY_MASK & (~(1 << RS::INSTANCE_PARTICLES))))) { //all but particles to avoid self collision
+ continue;
}
+ geometry_instances_to_render.push_back(instance);
}
- scene_render->render_particle_collider_heightfield(hfpc->base, hfpc->transform, (RendererSceneRender::InstanceBase **)instance_cull_result, cull_count);
+ scene_render->render_particle_collider_heightfield(hfpc->base, hfpc->transform, geometry_instances_to_render);
}
heightfield_particle_colliders_update_list.erase(heightfield_particle_colliders_update_list.front());
}
@@ -3009,7 +3195,7 @@ void RendererSceneCull::_update_dirty_instance(Instance *p_instance) {
if (can_cast_shadows != geom->can_cast_shadows) {
//ability to cast shadows change, let lights now
- for (List<Instance *>::Element *E = geom->lighting.front(); E; E = E->next()) {
+ for (Set<Instance *>::Element *E = geom->lights.front(); E; E = E->next()) {
InstanceLightData *light = static_cast<InstanceLightData *>(E->get()->base_data);
light->shadow_dirty = true;
}
@@ -3060,6 +3246,12 @@ void RendererSceneCull::update_dirty_instances() {
}
void RendererSceneCull::update() {
+ //optimize bvhs
+ for (uint32_t i = 0; i < scenario_owner.get_rid_count(); i++) {
+ Scenario *s = scenario_owner.get_ptr_by_index(i);
+ s->indexers[Scenario::INDEXER_GEOMETRY].optimize_incremental(indexer_update_iterations);
+ s->indexers[Scenario::INDEXER_VOLUMES].optimize_incremental(indexer_update_iterations);
+ }
scene_render->update();
update_dirty_instances();
render_particle_colliders();
@@ -3082,6 +3274,9 @@ bool RendererSceneCull::free(RID p_rid) {
while (scenario->instances.first()) {
instance_set_scenario(scenario->instances.first()->self()->self, RID());
}
+ scenario->instance_aabbs.reset();
+ scenario->instance_data.reset();
+
scene_render->free(scenario->reflection_probe_shadow_atlas);
scene_render->free(scenario->reflection_atlas);
scenario_owner.free(p_rid);
@@ -3130,7 +3325,67 @@ RendererSceneCull *RendererSceneCull::singleton = nullptr;
RendererSceneCull::RendererSceneCull() {
render_pass = 1;
singleton = this;
+ pair_volumes_to_mesh = false;
+
+ instance_cull_result.set_page_pool(&instance_cull_page_pool);
+ mesh_instance_cull_result.set_page_pool(&rid_cull_page_pool);
+ instance_shadow_cull_result.set_page_pool(&instance_cull_page_pool);
+ instance_sdfgi_cull_result.set_page_pool(&instance_cull_page_pool);
+ light_cull_result.set_page_pool(&instance_cull_page_pool);
+
+ geometry_instances_to_render.set_page_pool(&base_instance_cull_page_pool);
+ geometry_instances_to_shadow_render.set_page_pool(&base_instance_cull_page_pool);
+ lightmap_cull_result.set_page_pool(&base_instance_cull_page_pool);
+
+ reflection_probe_instance_cull_result.set_page_pool(&rid_cull_page_pool);
+ light_instance_cull_result.set_page_pool(&rid_cull_page_pool);
+ gi_probe_instance_cull_result.set_page_pool(&rid_cull_page_pool);
+ decal_instance_cull_result.set_page_pool(&rid_cull_page_pool);
+
+ for (int i = 0; i < RendererSceneRender::MAX_DIRECTIONAL_LIGHTS; i++) {
+ for (int j = 0; j < RendererSceneRender::MAX_DIRECTIONAL_LIGHT_CASCADES; j++) {
+ cull.shadows[i].cascades[j].cull_result.set_page_pool(&base_instance_cull_page_pool);
+ }
+ }
+
+ for (int i = 0; i < SDFGI_MAX_CASCADES * SDFGI_MAX_REGIONS_PER_CASCADE; i++) {
+ cull.sdfgi.region_cull_result[i].set_page_pool(&base_instance_cull_page_pool);
+ }
+
+ for (int i = 0; i < SDFGI_MAX_CASCADES; i++) {
+ cull.sdfgi.cascade_lights[i].set_page_pool(&rid_cull_page_pool);
+ }
+
+ indexer_update_iterations = GLOBAL_GET("rendering/spatial_indexer/update_iterations_per_frame");
}
RendererSceneCull::~RendererSceneCull() {
+ instance_cull_result.reset();
+ mesh_instance_cull_result.reset();
+ instance_shadow_cull_result.reset();
+ instance_sdfgi_cull_result.reset();
+ light_cull_result.reset();
+
+ geometry_instances_to_render.reset();
+ geometry_instances_to_shadow_render.reset();
+ lightmap_cull_result.reset();
+
+ reflection_probe_instance_cull_result.reset();
+ light_instance_cull_result.reset();
+ gi_probe_instance_cull_result.reset();
+ decal_instance_cull_result.reset();
+
+ for (int i = 0; i < RendererSceneRender::MAX_DIRECTIONAL_LIGHTS; i++) {
+ for (int j = 0; j < RendererSceneRender::MAX_DIRECTIONAL_LIGHT_CASCADES; j++) {
+ cull.shadows[i].cascades[j].cull_result.reset();
+ }
+ }
+
+ for (int i = 0; i < SDFGI_MAX_CASCADES * SDFGI_MAX_REGIONS_PER_CASCADE; i++) {
+ cull.sdfgi.region_cull_result[i].reset();
+ }
+
+ for (int i = 0; i < SDFGI_MAX_CASCADES; i++) {
+ cull.sdfgi.cascade_lights[i].reset();
+ }
}
diff --git a/servers/rendering/renderer_scene_cull.h b/servers/rendering/renderer_scene_cull.h
index 4d82d873cc..b755efd877 100644
--- a/servers/rendering/renderer_scene_cull.h
+++ b/servers/rendering/renderer_scene_cull.h
@@ -34,30 +34,26 @@
#include "core/templates/pass_func.h"
#include "servers/rendering/renderer_compositor.h"
+#include "core/math/dynamic_bvh.h"
#include "core/math/geometry_3d.h"
#include "core/math/octree.h"
#include "core/os/semaphore.h"
#include "core/os/thread.h"
#include "core/templates/local_vector.h"
+#include "core/templates/paged_allocator.h"
+#include "core/templates/paged_array.h"
#include "core/templates/rid_owner.h"
#include "core/templates/self_list.h"
#include "servers/rendering/renderer_scene.h"
#include "servers/rendering/renderer_scene_render.h"
#include "servers/xr/xr_interface.h"
-
class RendererSceneCull : public RendererScene {
public:
RendererSceneRender *scene_render;
enum {
- MAX_INSTANCE_CULL = 65536,
- MAX_LIGHTS_CULLED = 4096,
- MAX_REFLECTION_PROBES_CULLED = 4096,
- MAX_DECALS_CULLED = 4096,
- MAX_GI_PROBES_CULLED = 4096,
- MAX_ROOM_CULL = 32,
- MAX_LIGHTMAPS_CULLED = 4096,
- MAX_EXTERIOR_PORTALS = 128,
+ SDFGI_MAX_CASCADES = 8,
+ SDFGI_MAX_REGIONS_PER_CASCADE = 3
};
uint64_t render_pass;
@@ -113,12 +109,165 @@ public:
struct Instance;
+ struct PlaneSign {
+ _ALWAYS_INLINE_ PlaneSign() {}
+ _ALWAYS_INLINE_ PlaneSign(const Plane &p_plane) {
+ if (p_plane.normal.x > 0) {
+ signs[0] = 0;
+ } else {
+ signs[0] = 3;
+ }
+ if (p_plane.normal.y > 0) {
+ signs[1] = 1;
+ } else {
+ signs[1] = 4;
+ }
+ if (p_plane.normal.z > 0) {
+ signs[2] = 2;
+ } else {
+ signs[2] = 5;
+ }
+ }
+
+ uint32_t signs[3];
+ };
+
+ struct Frustum {
+ Vector<Plane> planes;
+ Vector<PlaneSign> plane_signs;
+ const Plane *planes_ptr;
+ const PlaneSign *plane_signs_ptr;
+ uint32_t plane_count;
+
+ _ALWAYS_INLINE_ Frustum() {}
+ _ALWAYS_INLINE_ Frustum(const Frustum &p_frustum) {
+ planes = p_frustum.planes;
+ plane_signs = p_frustum.plane_signs;
+
+ planes_ptr = planes.ptr();
+ plane_signs_ptr = plane_signs.ptr();
+ plane_count = p_frustum.plane_count;
+ }
+ _ALWAYS_INLINE_ void operator=(const Frustum &p_frustum) {
+ planes = p_frustum.planes;
+ plane_signs = p_frustum.plane_signs;
+
+ planes_ptr = planes.ptr();
+ plane_signs_ptr = plane_signs.ptr();
+ plane_count = p_frustum.plane_count;
+ }
+ _ALWAYS_INLINE_ Frustum(const Vector<Plane> &p_planes) {
+ planes = p_planes;
+ planes_ptr = planes.ptrw();
+ plane_count = planes.size();
+ for (int i = 0; i < planes.size(); i++) {
+ PlaneSign ps(p_planes[i]);
+ plane_signs.push_back(ps);
+ }
+
+ plane_signs_ptr = plane_signs.ptr();
+ }
+ };
+
+ struct InstanceBounds {
+ // Efficiently store instance bounds.
+ // Because bounds checking is performed first,
+ // keep it separated from data.
+
+ real_t bounds[6];
+ _ALWAYS_INLINE_ InstanceBounds() {}
+
+ _ALWAYS_INLINE_ InstanceBounds(const AABB &p_aabb) {
+ bounds[0] = p_aabb.position.x;
+ bounds[1] = p_aabb.position.y;
+ bounds[2] = p_aabb.position.z;
+ bounds[3] = p_aabb.position.x + p_aabb.size.x;
+ bounds[4] = p_aabb.position.y + p_aabb.size.y;
+ bounds[5] = p_aabb.position.z + p_aabb.size.z;
+ }
+ _ALWAYS_INLINE_ bool in_frustum(const Frustum &p_frustum) const {
+ // This is not a full SAT check and the possibility of false positives exist,
+ // but the tradeoff vs performance is still very good.
+
+ for (uint32_t i = 0; i < p_frustum.plane_count; i++) {
+ Vector3 min(
+ bounds[p_frustum.plane_signs_ptr[i].signs[0]],
+ bounds[p_frustum.plane_signs_ptr[i].signs[1]],
+ bounds[p_frustum.plane_signs_ptr[i].signs[2]]);
+
+ if (p_frustum.planes_ptr[i].distance_to(min) >= 0.0) {
+ return false;
+ }
+ }
+
+ return true;
+ }
+ _ALWAYS_INLINE_ bool in_aabb(const AABB &p_aabb) const {
+ Vector3 end = p_aabb.position + p_aabb.size;
+
+ if (bounds[0] >= end.x) {
+ return false;
+ }
+ if (bounds[3] <= p_aabb.position.x) {
+ return false;
+ }
+ if (bounds[1] >= end.y) {
+ return false;
+ }
+ if (bounds[4] <= p_aabb.position.y) {
+ return false;
+ }
+ if (bounds[2] >= end.z) {
+ return false;
+ }
+ if (bounds[5] <= p_aabb.position.z) {
+ return false;
+ }
+
+ return true;
+ }
+ };
+
+ struct InstanceData {
+ // Store instance pointer as well as common instance processing information,
+ // to make processing more cache friendly.
+ enum Flags {
+ FLAG_BASE_TYPE_MASK = 0xFF,
+ FLAG_CAST_SHADOWS = (1 << 8),
+ FLAG_CAST_SHADOWS_ONLY = (1 << 9),
+ FLAG_REDRAW_IF_VISIBLE = (1 << 10),
+ FLAG_GEOM_LIGHTING_DIRTY = (1 << 11),
+ FLAG_GEOM_REFLECTION_DIRTY = (1 << 12),
+ FLAG_GEOM_DECAL_DIRTY = (1 << 13),
+ FLAG_GEOM_GI_PROBE_DIRTY = (1 << 14),
+ FLAG_LIGHTMAP_CAPTURE = (1 << 15),
+ FLAG_USES_BAKED_LIGHT = (1 << 16),
+ FLAG_USES_MESH_INSTANCE = (1 << 17),
+ FLAG_REFLECTION_PROBE_DIRTY = (1 << 18),
+ };
+
+ uint32_t flags = 0;
+ uint32_t layer_mask = 0; //for fast layer-mask discard
+ RID base_rid;
+ RID instance_data_rid;
+ Instance *instance = nullptr;
+ };
+
+ PagedArrayPool<InstanceBounds> instance_aabb_page_pool;
+ PagedArrayPool<InstanceData> instance_data_page_pool;
+
struct Scenario {
+ enum IndexerType {
+ INDEXER_GEOMETRY, //for geometry
+ INDEXER_VOLUMES, //for everything else
+ INDEXER_MAX
+ };
+
+ DynamicBVH indexers[INDEXER_MAX];
+
RS::ScenarioDebugMode debug;
RID self;
- Octree<Instance, true> octree;
-
List<Instance *> directional_lights;
RID environment;
RID fallback_environment;
@@ -130,13 +279,22 @@ public:
LocalVector<RID> dynamic_lights;
- Scenario() { debug = RS::SCENARIO_DEBUG_DISABLED; }
+ PagedArray<InstanceBounds> instance_aabbs;
+ PagedArray<InstanceData> instance_data;
+
+ Scenario() {
+ indexers[INDEXER_GEOMETRY].set_index(INDEXER_GEOMETRY);
+ indexers[INDEXER_VOLUMES].set_index(INDEXER_VOLUMES);
+ debug = RS::SCENARIO_DEBUG_DISABLED;
+ }
};
+ int indexer_update_iterations = 0;
+
mutable RID_PtrOwner<Scenario> scenario_owner;
- static void *_instance_pair(void *p_self, OctreeElementID, Instance *p_A, int, OctreeElementID, Instance *p_B, int);
- static void _instance_unpair(void *p_self, OctreeElementID, Instance *p_A, int, OctreeElementID, Instance *p_B, int, void *);
+ static void _instance_pair(Instance *p_A, Instance *p_B);
+ static void _instance_unpair(Instance *p_A, Instance *p_B);
static void _instance_update_mesh_instance(Instance *p_instance);
@@ -152,6 +310,17 @@ public:
/* INSTANCING API */
+ struct InstancePair {
+ Instance *a;
+ Instance *b;
+ SelfList<InstancePair> list_a;
+ SelfList<InstancePair> list_b;
+ InstancePair() :
+ list_a(this), list_b(this) {}
+ };
+
+ PagedAllocator<InstancePair> pair_allocator;
+
struct InstanceBaseData {
virtual ~InstanceBaseData() {}
};
@@ -159,7 +328,8 @@ public:
struct Instance : RendererSceneRender::InstanceBase {
RID self;
//scenario stuff
- OctreeElementID octree_id;
+ DynamicBVH::ID indexer_id;
+ int32_t array_index;
Scenario *scenario;
SelfList<Instance> scenario_item;
@@ -181,13 +351,15 @@ public:
Vector<Color> lightmap_target_sh; //target is used for incrementally changing the SH over time, this avoids pops in some corner cases and when going interior <-> exterior
- uint64_t last_render_pass;
uint64_t last_frame_pass;
uint64_t version; // changes to this, and changes to base increase version
InstanceBaseData *base_data;
+ SelfList<InstancePair>::List pairs;
+ uint64_t pair_check;
+
virtual void dependency_deleted(RID p_dependency) {
if (p_dependency == base) {
singleton->instance_set_base(self, RID());
@@ -205,7 +377,6 @@ public:
Instance() :
scenario_item(this),
update_item(this) {
- octree_id = 0;
scenario = nullptr;
update_aabb = false;
@@ -220,12 +391,14 @@ public:
lod_begin_hysteresis = 0;
lod_end_hysteresis = 0;
- last_render_pass = 0;
last_frame_pass = 0;
version = 1;
base_data = nullptr;
custom_aabb = nullptr;
+
+ pair_check = 0;
+ array_index = -1;
}
~Instance() {
@@ -242,50 +415,33 @@ public:
void _instance_queue_update(Instance *p_instance, bool p_update_aabb, bool p_update_dependencies = false);
struct InstanceGeometryData : public InstanceBaseData {
- List<Instance *> lighting;
- bool lighting_dirty;
+ Set<Instance *> lights;
bool can_cast_shadows;
bool material_is_animated;
- List<Instance *> decals;
- bool decal_dirty;
-
- List<Instance *> reflection_probes;
- bool reflection_dirty;
-
- List<Instance *> gi_probes;
- bool gi_probes_dirty;
-
- List<Instance *> lightmap_captures;
+ Set<Instance *> decals;
+ Set<Instance *> reflection_probes;
+ Set<Instance *> gi_probes;
+ Set<Instance *> lightmap_captures;
InstanceGeometryData() {
- lighting_dirty = false;
- reflection_dirty = true;
can_cast_shadows = true;
material_is_animated = true;
- gi_probes_dirty = true;
- decal_dirty = true;
}
};
struct InstanceReflectionProbeData : public InstanceBaseData {
Instance *owner;
- struct PairInfo {
- List<Instance *>::Element *L; //reflection iterator in geometry
- Instance *geometry;
- };
- List<PairInfo> geometries;
+ Set<Instance *> geometries;
RID instance;
- bool reflection_dirty;
SelfList<InstanceReflectionProbeData> update_list;
int render_step;
InstanceReflectionProbeData() :
update_list(this) {
- reflection_dirty = true;
render_step = -1;
}
};
@@ -294,11 +450,7 @@ public:
Instance *owner;
RID instance;
- struct PairInfo {
- List<Instance *>::Element *L; //reflection iterator in geometry
- Instance *geometry;
- };
- List<PairInfo> geometries;
+ Set<Instance *> geometries;
InstanceDecalData() {
}
@@ -307,26 +459,19 @@ public:
SelfList<InstanceReflectionProbeData>::List reflection_probe_render_list;
struct InstanceLightData : public InstanceBaseData {
- struct PairInfo {
- List<Instance *>::Element *L; //light iterator in geometry
- Instance *geometry;
- };
-
RID instance;
uint64_t last_version;
List<Instance *>::Element *D; // directional light in scenario
bool shadow_dirty;
- List<PairInfo> geometries;
+ Set<Instance *> geometries;
Instance *baked_light;
RS::LightBakeMode bake_mode;
uint32_t max_sdfgi_cascade = 2;
- uint64_t sdfgi_cascade_light_pass = 0;
-
InstanceLightData() {
bake_mode = RS::LIGHT_BAKE_DISABLED;
shadow_dirty = true;
@@ -339,13 +484,8 @@ public:
struct InstanceGIProbeData : public InstanceBaseData {
Instance *owner;
- struct PairInfo {
- List<Instance *>::Element *L; //gi probe iterator in geometry
- Instance *geometry;
- };
-
- List<PairInfo> geometries;
- List<PairInfo> dynamic_geometries;
+ Set<Instance *> geometries;
+ Set<Instance *> dynamic_geometries;
Set<Instance *> lights;
@@ -383,40 +523,104 @@ public:
SelfList<InstanceGIProbeData>::List gi_probe_update_list;
struct InstanceLightmapData : public InstanceBaseData {
- struct PairInfo {
- List<Instance *>::Element *L; //iterator in geometry
- Instance *geometry;
- };
- List<PairInfo> geometries;
-
+ Set<Instance *> geometries;
Set<Instance *> users;
InstanceLightmapData() {
}
};
+ uint64_t pair_pass = 1;
+
+ struct PairInstances {
+ Instance *instance = nullptr;
+ PagedAllocator<InstancePair> *pair_allocator = nullptr;
+ SelfList<InstancePair>::List pairs_found;
+ DynamicBVH *bvh = nullptr;
+ DynamicBVH *bvh2 = nullptr; //some may need to cull in two
+ uint32_t pair_mask;
+ uint64_t pair_pass;
+
+ _FORCE_INLINE_ bool operator()(void *p_data) {
+ Instance *p_instance = (Instance *)p_data;
+ if (instance != p_instance && instance->transformed_aabb.intersects(p_instance->transformed_aabb) && (pair_mask & (1 << p_instance->base_type))) {
+ //test is more coarse in indexer
+ p_instance->pair_check = pair_pass;
+ InstancePair *pair = pair_allocator->alloc();
+ pair->a = instance;
+ pair->b = p_instance;
+ pairs_found.add(&pair->list_a);
+ }
+ return false;
+ }
+
+ void pair() {
+ if (bvh) {
+ bvh->aabb_query(instance->transformed_aabb, *this);
+ }
+ if (bvh2) {
+ bvh2->aabb_query(instance->transformed_aabb, *this);
+ }
+ while (instance->pairs.first()) {
+ InstancePair *pair = instance->pairs.first()->self();
+ Instance *other_instance = instance == pair->a ? pair->b : pair->a;
+ if (other_instance->pair_check != pair_pass) {
+ //unpaired
+ _instance_unpair(instance, other_instance);
+ } else {
+ //kept
+ other_instance->pair_check = 0; // if kept, then put pair check to zero, so we can distinguish with the newly added ones
+ }
+
+ pair_allocator->free(pair);
+ }
+ while (pairs_found.first()) {
+ InstancePair *pair = pairs_found.first()->self();
+ pairs_found.remove(pairs_found.first());
+
+ if (pair->b->pair_check == pair_pass) {
+ //paired
+ _instance_pair(instance, pair->b);
+ }
+ pair->a->pairs.add(&pair->list_a);
+ pair->b->pairs.add(&pair->list_b);
+ }
+ }
+ };
+
+ struct CullResult {
+ PagedArray<Instance *> *result;
+ _FORCE_INLINE_ bool operator()(void *p_data) {
+ Instance *p_instance = (Instance *)p_data;
+ result->push_back(p_instance);
+ return false;
+ }
+ };
+
Set<Instance *> heightfield_particle_colliders_update_list;
- int instance_cull_count;
- Instance *instance_cull_result[MAX_INSTANCE_CULL];
- Instance *instance_shadow_cull_result[MAX_INSTANCE_CULL]; //used for generating shadowmaps
- Instance *light_cull_result[MAX_LIGHTS_CULLED];
- RID sdfgi_light_cull_result[MAX_LIGHTS_CULLED];
- RID light_instance_cull_result[MAX_LIGHTS_CULLED];
- uint64_t sdfgi_light_cull_pass = 0;
- int light_cull_count;
- int directional_light_count;
- RID reflection_probe_instance_cull_result[MAX_REFLECTION_PROBES_CULLED];
- RID decal_instance_cull_result[MAX_DECALS_CULLED];
- int reflection_probe_cull_count;
- int decal_cull_count;
- RID gi_probe_instance_cull_result[MAX_GI_PROBES_CULLED];
- int gi_probe_cull_count;
- Instance *lightmap_cull_result[MAX_LIGHTS_CULLED];
- int lightmap_cull_count;
+ PagedArrayPool<Instance *> instance_cull_page_pool;
+ PagedArrayPool<RendererSceneRender::InstanceBase *> base_instance_cull_page_pool;
+ PagedArrayPool<RID> rid_cull_page_pool;
+
+ PagedArray<Instance *> instance_cull_result;
+ PagedArray<RID> mesh_instance_cull_result;
+ PagedArray<RendererSceneRender::InstanceBase *> geometry_instances_to_render;
+ PagedArray<Instance *> instance_shadow_cull_result;
+ PagedArray<RendererSceneRender::InstanceBase *> geometry_instances_to_shadow_render;
+ PagedArray<Instance *> instance_sdfgi_cull_result;
+ PagedArray<Instance *> light_cull_result;
+ PagedArray<RendererSceneRender::InstanceBase *> lightmap_cull_result;
+ PagedArray<RID> reflection_probe_instance_cull_result;
+ PagedArray<RID> light_instance_cull_result;
+
+ PagedArray<RID> gi_probe_instance_cull_result;
+ PagedArray<RID> decal_instance_cull_result;
RID_PtrOwner<Instance> instance_owner;
+ bool pair_volumes_to_mesh; // used in traditional forward, unnecesary on clustered
+
virtual RID instance_create();
virtual void instance_set_base(RID p_instance, RID p_base);
@@ -460,11 +664,56 @@ public:
_FORCE_INLINE_ void _update_instance_aabb(Instance *p_instance);
_FORCE_INLINE_ void _update_dirty_instance(Instance *p_instance);
_FORCE_INLINE_ void _update_instance_lightmap_captures(Instance *p_instance);
+ void _unpair_instance(Instance *p_instance);
+
+ void _light_instance_setup_directional_shadow(int p_shadow_index, Instance *p_instance, const Transform p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_orthogonal, bool p_cam_vaspect);
_FORCE_INLINE_ bool _light_instance_update_shadow(Instance *p_instance, const Transform p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_orthogonal, bool p_cam_vaspect, RID p_shadow_atlas, Scenario *p_scenario, float p_scren_lod_threshold);
RID _render_get_environment(RID p_camera, RID p_scenario);
+ struct Cull {
+ struct Shadow {
+ RID light_instance;
+ struct Cascade {
+ Frustum frustum;
+
+ CameraMatrix projection;
+ Transform transform;
+ real_t zfar;
+ real_t split;
+ real_t shadow_texel_size;
+ real_t bias_scale;
+ real_t range_begin;
+ Vector2 uv_scale;
+
+ PagedArray<RendererSceneRender::InstanceBase *> cull_result;
+
+ } cascades[RendererSceneRender::MAX_DIRECTIONAL_LIGHT_CASCADES]; //max 4 cascades
+ uint32_t cascade_count;
+
+ } shadows[RendererSceneRender::MAX_DIRECTIONAL_LIGHTS];
+
+ uint32_t shadow_count;
+
+ struct SDFGI {
+ //have arrays here because SDFGI functions expects this, plus regions can have areas
+ PagedArray<RendererSceneRender::InstanceBase *> region_cull_result[SDFGI_MAX_CASCADES * SDFGI_MAX_REGIONS_PER_CASCADE];
+ AABB region_aabb[SDFGI_MAX_CASCADES * SDFGI_MAX_REGIONS_PER_CASCADE]; //max 3 regions per cascade
+ uint32_t region_cascade[SDFGI_MAX_CASCADES * SDFGI_MAX_REGIONS_PER_CASCADE]; //max 3 regions per cascade
+ uint32_t region_count = 0;
+
+ PagedArray<RID> cascade_lights[SDFGI_MAX_CASCADES];
+ uint32_t cascade_light_index[SDFGI_MAX_CASCADES];
+ uint32_t cascade_light_count = 0;
+
+ } sdfgi;
+
+ SpinLock lock;
+
+ Frustum frustum;
+ } cull;
+
bool _render_reflection_probe_step(Instance *p_instance, int p_step);
void _prepare_scene(const Transform p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_orthogonal, bool p_cam_vaspect, RID p_render_buffers, RID p_environment, uint32_t p_visible_layers, RID p_scenario, RID p_shadow_atlas, RID p_reflection_probe, float p_screen_lod_threshold, bool p_using_shadows = true);
void _render_scene(RID p_render_buffers, const Transform p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_orthogonal, RID p_environment, RID p_force_camera_effects, RID p_scenario, RID p_shadow_atlas, RID p_reflection_probe, int p_reflection_probe_pass, float p_screen_lod_threshold);
@@ -516,8 +765,8 @@ public:
PASS6(environment_set_ssr, RID, bool, int, float, float, float)
PASS1(environment_set_ssr_roughness_quality, RS::EnvironmentSSRRoughnessQuality)
- PASS9(environment_set_ssao, RID, bool, float, float, float, float, float, RS::EnvironmentSSAOBlur, float)
- PASS2(environment_set_ssao_quality, RS::EnvironmentSSAOQuality, bool)
+ PASS10(environment_set_ssao, RID, bool, float, float, float, float, float, float, float, float)
+ PASS6(environment_set_ssao_quality, RS::EnvironmentSSAOQuality, bool, float, int, float, float)
PASS11(environment_set_glow, RID, bool, Vector<float>, float, float, float, float, RS::EnvironmentGlowBlendMode, float, float, float)
PASS1(environment_glow_set_use_bicubic_upscale, bool)
diff --git a/servers/rendering/renderer_scene_render.h b/servers/rendering/renderer_scene_render.h
index e1f3ea9b6b..d5e8f391cd 100644
--- a/servers/rendering/renderer_scene_render.h
+++ b/servers/rendering/renderer_scene_render.h
@@ -32,13 +32,19 @@
#define RENDERINGSERVERSCENERENDER_H
#include "core/math/camera_matrix.h"
+#include "core/templates/paged_array.h"
#include "servers/rendering/renderer_storage.h"
class RendererSceneRender {
public:
+ enum {
+ MAX_DIRECTIONAL_LIGHTS = 8,
+ MAX_DIRECTIONAL_LIGHT_CASCADES = 4
+ };
/* SHADOW ATLAS API */
- virtual RID shadow_atlas_create() = 0;
+ virtual RID
+ shadow_atlas_create() = 0;
virtual void shadow_atlas_set_size(RID p_atlas, int p_size) = 0;
virtual void shadow_atlas_set_quadrant_subdivision(RID p_atlas, int p_quadrant, int p_subdivision) = 0;
virtual bool shadow_atlas_update_light(RID p_atlas, RID p_light_intance, float p_coverage, uint64_t p_light_version) = 0;
@@ -55,7 +61,7 @@ public:
virtual int sdfgi_get_pending_region_count(RID p_render_buffers) const = 0;
virtual AABB sdfgi_get_pending_region_bounds(RID p_render_buffers, int p_region) const = 0;
virtual uint32_t sdfgi_get_pending_region_cascade(RID p_render_buffers, int p_region) const = 0;
- virtual void sdfgi_update_probes(RID p_render_buffers, RID p_environment, const RID *p_directional_light_instances, uint32_t p_directional_light_count, const RID *p_positional_light_instances, uint32_t p_positional_light_count) = 0;
+ virtual void sdfgi_update_probes(RID p_render_buffers, RID p_environment, const Vector<RID> &p_directional_lights, const RID *p_positional_light_instances, uint32_t p_positional_light_count) = 0;
/* SKY API */
@@ -96,9 +102,9 @@ public:
virtual void 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) = 0;
virtual void environment_set_ssr_roughness_quality(RS::EnvironmentSSRRoughnessQuality p_quality) = 0;
- virtual void environment_set_ssao(RID p_env, bool p_enable, float p_radius, float p_intensity, float p_bias, float p_light_affect, float p_ao_channel_affect, RS::EnvironmentSSAOBlur p_blur, float p_bilateral_sharpness) = 0;
+ virtual void 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) = 0;
- virtual void environment_set_ssao_quality(RS::EnvironmentSSAOQuality p_quality, bool p_half_size) = 0;
+ virtual void 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) = 0;
virtual void environment_set_sdfgi(RID p_env, bool p_enable, RS::EnvironmentSDFGICascades p_cascades, float p_min_cell_size, RS::EnvironmentSDFGIYScale p_y_scale, bool p_use_occlusion, bool p_use_multibounce, bool p_read_sky, float p_energy, float p_normal_bias, float p_probe_bias) = 0;
@@ -171,6 +177,7 @@ public:
AABB aabb;
AABB transformed_aabb;
+ AABB prev_transformed_aabb;
struct InstanceShaderParameter {
int32_t index = -1;
@@ -231,17 +238,17 @@ public:
virtual RID gi_probe_instance_create(RID p_gi_probe) = 0;
virtual void gi_probe_instance_set_transform_to_data(RID p_probe, const Transform &p_xform) = 0;
virtual bool gi_probe_needs_update(RID p_probe) const = 0;
- virtual void gi_probe_update(RID p_probe, bool p_update_light_instances, const Vector<RID> &p_light_instances, int p_dynamic_object_count, InstanceBase **p_dynamic_objects) = 0;
+ virtual void gi_probe_update(RID p_probe, bool p_update_light_instances, const Vector<RID> &p_light_instances, const PagedArray<RendererSceneRender::InstanceBase *> &p_dynamic_objects) = 0;
virtual void gi_probe_set_quality(RS::GIProbeQuality) = 0;
- virtual void render_scene(RID p_render_buffers, const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, InstanceBase **p_cull_result, int p_cull_count, RID *p_light_cull_result, int p_light_cull_count, RID *p_reflection_probe_cull_result, int p_reflection_probe_cull_count, RID *p_gi_probe_cull_result, int p_gi_probe_cull_count, RID *p_decal_cull_result, int p_decal_cull_count, InstanceBase **p_lightmap_cull_result, int p_lightmap_cull_count, RID p_environment, RID p_camera_effects, RID p_shadow_atlas, RID p_reflection_atlas, RID p_reflection_probe, int p_reflection_probe_pass, float p_screen_lod_threshold) = 0;
+ virtual void render_scene(RID p_render_buffers, const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, const PagedArray<InstanceBase *> &p_instances, const PagedArray<RID> &p_lights, const PagedArray<RID> &p_reflection_probes, const PagedArray<RID> &p_gi_probes, const PagedArray<RID> &p_decals, const PagedArray<InstanceBase *> &p_lightmaps, RID p_environment, RID p_camera_effects, RID p_shadow_atlas, RID p_reflection_atlas, RID p_reflection_probe, int p_reflection_probe_pass, float p_screen_lod_threshold) = 0;
- virtual void render_shadow(RID p_light, RID p_shadow_atlas, int p_pass, InstanceBase **p_cull_result, int p_cull_count, const Plane &p_camera_plane = Plane(), float p_lod_distance_multiplier = 0, float p_screen_lod_threshold = 0.0) = 0;
- virtual void render_material(const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, InstanceBase **p_cull_result, int p_cull_count, RID p_framebuffer, const Rect2i &p_region) = 0;
- virtual void render_sdfgi(RID p_render_buffers, int p_region, InstanceBase **p_cull_result, int p_cull_count) = 0;
- virtual void render_sdfgi_static_lights(RID p_render_buffers, uint32_t p_cascade_count, const uint32_t *p_cascade_indices, const RID **p_positional_light_cull_result, const uint32_t *p_positional_light_cull_count) = 0;
- virtual void render_particle_collider_heightfield(RID p_collider, const Transform &p_transform, InstanceBase **p_cull_result, int p_cull_count) = 0;
+ virtual void render_shadow(RID p_light, RID p_shadow_atlas, int p_pass, const PagedArray<InstanceBase *> &p_instances, const Plane &p_camera_plane = Plane(), float p_lod_distance_multiplier = 0, float p_screen_lod_threshold = 0.0) = 0;
+ virtual void render_material(const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, const PagedArray<InstanceBase *> &p_instances, RID p_framebuffer, const Rect2i &p_region) = 0;
+ virtual void render_sdfgi(RID p_render_buffers, int p_region, const PagedArray<InstanceBase *> &p_instances) = 0;
+ virtual void render_sdfgi_static_lights(RID p_render_buffers, uint32_t p_cascade_count, const uint32_t *p_cascade_indices, const PagedArray<RID> *p_positional_lights) = 0;
+ virtual void render_particle_collider_heightfield(RID p_collider, const Transform &p_transform, const PagedArray<InstanceBase *> &p_instances) = 0;
virtual void set_scene_pass(uint64_t p_pass) = 0;
virtual void set_time(double p_time, double p_step) = 0;
diff --git a/servers/rendering/rendering_device.cpp b/servers/rendering/rendering_device.cpp
index ba30670082..e78140bcec 100644
--- a/servers/rendering/rendering_device.cpp
+++ b/servers/rendering/rendering_device.cpp
@@ -68,7 +68,7 @@ RID RenderingDevice::_texture_create(const Ref<RDTextureFormat> &p_format, const
Vector<Vector<uint8_t>> data;
for (int i = 0; i < p_data.size(); i++) {
Vector<uint8_t> byte_slice = p_data[i];
- ERR_FAIL_COND_V(byte_slice.empty(), RID());
+ ERR_FAIL_COND_V(byte_slice.is_empty(), RID());
data.push_back(byte_slice);
}
return texture_create(p_format->base, p_view->base, data);
@@ -154,7 +154,7 @@ RID RenderingDevice::shader_create_from_bytecode(const Ref<RDShaderBytecode> &p_
String error = p_bytecode->get_stage_compile_error(stage);
ERR_FAIL_COND_V_MSG(error != String(), RID(), "Can't create a shader from an errored bytecode. Check errors in source bytecode.");
sd.spir_v = p_bytecode->get_stage_bytecode(stage);
- if (sd.spir_v.empty()) {
+ if (sd.spir_v.is_empty()) {
continue;
}
stage_data.push_back(sd);
diff --git a/servers/rendering/rendering_device.h b/servers/rendering/rendering_device.h
index 524d7749c9..b3d4e66f6c 100644
--- a/servers/rendering/rendering_device.h
+++ b/servers/rendering/rendering_device.h
@@ -433,6 +433,7 @@ public:
TEXTURE_SLICE_2D,
TEXTURE_SLICE_CUBEMAP,
TEXTURE_SLICE_3D,
+ TEXTURE_SLICE_2D_ARRAY,
};
virtual RID texture_create_shared_from_slice(const TextureView &p_view, RID p_with_texture, uint32_t p_layer, uint32_t p_mipmap, TextureSliceType p_slice_type = TEXTURE_SLICE_2D) = 0;
diff --git a/servers/rendering/rendering_device_binds.cpp b/servers/rendering/rendering_device_binds.cpp
index af9ecef0dd..90df85f961 100644
--- a/servers/rendering/rendering_device_binds.cpp
+++ b/servers/rendering/rendering_device_binds.cpp
@@ -163,7 +163,7 @@ Error RDShaderFile::parse_versions_from_text(const String &p_text, const String
ERR_FAIL_V_MSG(ERR_PARSE_ERROR, "When writing compute shaders, [compute] mustbe the only stage present.");
}
- if (version_texts.empty()) {
+ if (version_texts.is_empty()) {
version_texts[""] = ""; //make sure a default version exists
}
diff --git a/servers/rendering/rendering_server_default.cpp b/servers/rendering/rendering_server_default.cpp
index 47f7fa07d5..da6c3ef6f4 100644
--- a/servers/rendering/rendering_server_default.cpp
+++ b/servers/rendering/rendering_server_default.cpp
@@ -45,17 +45,17 @@ int RenderingServerDefault::changes = 0;
/* BLACK BARS */
void RenderingServerDefault::black_bars_set_margins(int p_left, int p_top, int p_right, int p_bottom) {
- black_margin[MARGIN_LEFT] = p_left;
- black_margin[MARGIN_TOP] = p_top;
- black_margin[MARGIN_RIGHT] = p_right;
- black_margin[MARGIN_BOTTOM] = p_bottom;
+ black_margin[SIDE_LEFT] = p_left;
+ black_margin[SIDE_TOP] = p_top;
+ black_margin[SIDE_RIGHT] = p_right;
+ black_margin[SIDE_BOTTOM] = p_bottom;
}
void RenderingServerDefault::black_bars_set_images(RID p_left, RID p_top, RID p_right, RID p_bottom) {
- black_image[MARGIN_LEFT] = p_left;
- black_image[MARGIN_TOP] = p_top;
- black_image[MARGIN_RIGHT] = p_right;
- black_image[MARGIN_BOTTOM] = p_bottom;
+ black_image[SIDE_LEFT] = p_left;
+ black_image[SIDE_TOP] = p_top;
+ black_image[SIDE_RIGHT] = p_right;
+ black_image[SIDE_BOTTOM] = p_bottom;
}
void RenderingServerDefault::_draw_margins() {
@@ -101,11 +101,16 @@ void RenderingServerDefault::draw(bool p_swap_buffers, double frame_step) {
TIMESTAMP_BEGIN()
+ uint64_t time_usec = OS::get_singleton()->get_ticks_usec();
+
RSG::scene->update(); //update scenes stuff before updating instances
+ frame_setup_time = double(OS::get_singleton()->get_ticks_usec() - time_usec) / 1000.0;
+
RSG::storage->update_particles(); //need to be done after instances are updated (colliders and particle transforms), and colliders are rendered
RSG::scene->render_probes();
+
RSG::viewport->draw_viewports();
RSG::canvas_render->update();
@@ -159,6 +164,10 @@ void RenderingServerDefault::draw(bool p_swap_buffers, double frame_step) {
frame_profile_frame = RSG::storage->get_captured_timestamps_frame();
}
+float RenderingServerDefault::get_frame_setup_time_cpu() const {
+ return frame_setup_time;
+}
+
void RenderingServerDefault::sync() {
}
diff --git a/servers/rendering/rendering_server_default.h b/servers/rendering/rendering_server_default.h
index dd1a41a4f7..922cf08f3b 100644
--- a/servers/rendering/rendering_server_default.h
+++ b/servers/rendering/rendering_server_default.h
@@ -72,6 +72,8 @@ class RenderingServerDefault : public RenderingServer {
uint64_t frame_profile_frame;
Vector<FrameProfileArea> frame_profile;
+ float frame_setup_time = 0;
+
public:
//if editor is redrawing when it shouldn't, enable this and put a breakpoint in _changes_changed()
//#define DEBUG_CHANGES
@@ -590,8 +592,8 @@ public:
BIND6(environment_set_ssr, RID, bool, int, float, float, float)
BIND1(environment_set_ssr_roughness_quality, EnvironmentSSRRoughnessQuality)
- BIND9(environment_set_ssao, RID, bool, float, float, float, float, float, EnvironmentSSAOBlur, float)
- BIND2(environment_set_ssao_quality, EnvironmentSSAOQuality, bool)
+ BIND10(environment_set_ssao, RID, bool, float, float, float, float, float, float, float, float)
+ BIND6(environment_set_ssao_quality, EnvironmentSSAOQuality, bool, float, int, float, float)
BIND11(environment_set_glow, RID, bool, Vector<float>, float, float, float, float, EnvironmentGlowBlendMode, float, float, float)
BIND1(environment_glow_set_use_bicubic_upscale, bool)
@@ -845,6 +847,8 @@ public:
/* TESTING */
+ virtual float get_frame_setup_time_cpu() const;
+
virtual void set_boot_image(const Ref<Image> &p_image, const Color &p_color, bool p_scale, bool p_use_filter = true);
virtual void set_default_clear_color(const Color &p_color);
diff --git a/servers/rendering/rendering_server_wrap_mt.h b/servers/rendering/rendering_server_wrap_mt.h
index 20556779fe..1e6c3b8f71 100644
--- a/servers/rendering/rendering_server_wrap_mt.h
+++ b/servers/rendering/rendering_server_wrap_mt.h
@@ -498,9 +498,9 @@ public:
FUNC6(environment_set_ssr, RID, bool, int, float, float, float)
FUNC1(environment_set_ssr_roughness_quality, EnvironmentSSRRoughnessQuality)
- FUNC9(environment_set_ssao, RID, bool, float, float, float, float, float, EnvironmentSSAOBlur, float)
+ FUNC10(environment_set_ssao, RID, bool, float, float, float, float, float, float, float, float)
- FUNC2(environment_set_ssao_quality, EnvironmentSSAOQuality, bool)
+ FUNC6(environment_set_ssao_quality, EnvironmentSSAOQuality, bool, float, int, float, float)
FUNC11(environment_set_sdfgi, RID, bool, EnvironmentSDFGICascades, float, EnvironmentSDFGIYScale, bool, bool, bool, float, float, float)
FUNC1(environment_set_sdfgi_ray_count, EnvironmentSDFGIRayCount)
@@ -776,6 +776,10 @@ public:
return rendering_server->get_frame_profile();
}
+ virtual float get_frame_setup_time_cpu() const {
+ return rendering_server->get_frame_setup_time_cpu();
+ }
+
virtual void sdfgi_set_debug_probe_select(const Vector3 &p_position, const Vector3 &p_dir) {
rendering_server->sdfgi_set_debug_probe_select(p_position, p_dir);
}
diff --git a/servers/rendering/shader_language.cpp b/servers/rendering/shader_language.cpp
index 742ad8a7bf..ba343cf851 100644
--- a/servers/rendering/shader_language.cpp
+++ b/servers/rendering/shader_language.cpp
@@ -913,6 +913,7 @@ void ShaderLanguage::clear() {
char_idx = 0;
error_set = false;
error_str = "";
+ last_const = false;
while (nodes) {
Node *n = nodes;
nodes = nodes->next;
@@ -920,7 +921,7 @@ void ShaderLanguage::clear() {
}
}
-bool ShaderLanguage::_find_identifier(const BlockNode *p_block, bool p_allow_reassign, const FunctionInfo &p_function_info, const StringName &p_identifier, DataType *r_data_type, IdentifierType *r_type, bool *r_is_const, int *r_array_size, StringName *r_struct_name) {
+bool ShaderLanguage::_find_identifier(const BlockNode *p_block, bool p_allow_reassign, const FunctionInfo &p_function_info, const StringName &p_identifier, DataType *r_data_type, IdentifierType *r_type, bool *r_is_const, int *r_array_size, StringName *r_struct_name, ConstantNode::Value *r_constant_value) {
if (p_function_info.built_ins.has(p_identifier)) {
if (r_data_type) {
*r_data_type = p_function_info.built_ins[p_identifier].type;
@@ -968,6 +969,9 @@ bool ShaderLanguage::_find_identifier(const BlockNode *p_block, bool p_allow_rea
if (r_struct_name) {
*r_struct_name = p_block->variables[p_identifier].struct_name;
}
+ if (r_constant_value) {
+ *r_constant_value = p_block->variables[p_identifier].value;
+ }
return true;
}
@@ -1028,6 +1032,9 @@ bool ShaderLanguage::_find_identifier(const BlockNode *p_block, bool p_allow_rea
}
if (shader->constants.has(p_identifier)) {
+ if (r_is_const) {
+ *r_is_const = true;
+ }
if (r_data_type) {
*r_data_type = shader->constants[p_identifier].type;
}
@@ -1040,6 +1047,11 @@ bool ShaderLanguage::_find_identifier(const BlockNode *p_block, bool p_allow_rea
if (r_struct_name) {
*r_struct_name = shader->constants[p_identifier].type_str;
}
+ if (r_constant_value) {
+ if (shader->constants[p_identifier].initializer && shader->constants[p_identifier].initializer->values.size() == 1) {
+ *r_constant_value = shader->constants[p_identifier].initializer->values[0];
+ }
+ }
return true;
}
@@ -3241,6 +3253,137 @@ bool ShaderLanguage::_propagate_function_call_sampler_builtin_reference(StringNa
ERR_FAIL_V(false); //bug? function not found
}
+ShaderLanguage::Node *ShaderLanguage::_parse_array_constructor(BlockNode *p_block, const FunctionInfo &p_function_info, DataType p_type, const StringName &p_struct_name, int p_array_size) {
+ DataType type = TYPE_VOID;
+ String struct_name = "";
+ int array_size = 0;
+ bool auto_size = false;
+ Token tk = _get_token();
+
+ if (tk.type == TK_CURLY_BRACKET_OPEN) {
+ auto_size = true;
+ } else {
+ if (shader->structs.has(tk.text)) {
+ type = TYPE_STRUCT;
+ struct_name = tk.text;
+ } else {
+ if (!is_token_variable_datatype(tk.type)) {
+ _set_error("Invalid data type for array");
+ return nullptr;
+ }
+ type = get_token_datatype(tk.type);
+ }
+ tk = _get_token();
+ if (tk.type == TK_BRACKET_OPEN) {
+ TkPos pos = _get_tkpos();
+ tk = _get_token();
+ if (tk.type == TK_BRACKET_CLOSE) {
+ array_size = p_array_size;
+ tk = _get_token();
+ } else {
+ _set_tkpos(pos);
+
+ Node *n = _parse_and_reduce_expression(p_block, p_function_info);
+ if (!n || n->type != Node::TYPE_CONSTANT || n->get_datatype() != TYPE_INT) {
+ _set_error("Expected single integer constant > 0");
+ return nullptr;
+ }
+
+ ConstantNode *cnode = (ConstantNode *)n;
+ if (cnode->values.size() == 1) {
+ array_size = cnode->values[0].sint;
+ if (array_size <= 0) {
+ _set_error("Expected single integer constant > 0");
+ return nullptr;
+ }
+ } else {
+ _set_error("Expected single integer constant > 0");
+ return nullptr;
+ }
+
+ tk = _get_token();
+ if (tk.type != TK_BRACKET_CLOSE) {
+ _set_error("Expected ']'");
+ return nullptr;
+ } else {
+ tk = _get_token();
+ }
+ }
+ } else {
+ _set_error("Expected '['");
+ return nullptr;
+ }
+
+ if (type != p_type || struct_name != p_struct_name || array_size != p_array_size) {
+ String error_str = "Cannot convert from '";
+ if (type == TYPE_STRUCT) {
+ error_str += struct_name;
+ } else {
+ error_str += get_datatype_name(type);
+ }
+ error_str += "[";
+ error_str += itos(array_size);
+ error_str += "]'";
+ error_str += " to '";
+ if (type == TYPE_STRUCT) {
+ error_str += p_struct_name;
+ } else {
+ error_str += get_datatype_name(p_type);
+ }
+ error_str += "[";
+ error_str += itos(p_array_size);
+ error_str += "]'";
+ _set_error(error_str);
+ return nullptr;
+ }
+ }
+
+ ArrayConstructNode *an = alloc_node<ArrayConstructNode>();
+ an->datatype = p_type;
+ an->struct_name = p_struct_name;
+
+ if (tk.type == TK_PARENTHESIS_OPEN || auto_size) { // initialization
+ while (true) {
+ Node *n = _parse_and_reduce_expression(p_block, p_function_info);
+ if (!n) {
+ return nullptr;
+ }
+
+ if (p_type != n->get_datatype() || p_struct_name != n->get_datatype_name()) {
+ _set_error("Invalid assignment of '" + (n->get_datatype() == TYPE_STRUCT ? n->get_datatype_name() : get_datatype_name(n->get_datatype())) + "' to '" + (type == TYPE_STRUCT ? struct_name : get_datatype_name(type)) + "'");
+ return nullptr;
+ }
+
+ tk = _get_token();
+ if (tk.type == TK_COMMA) {
+ an->initializer.push_back(n);
+ } else if (!auto_size && tk.type == TK_PARENTHESIS_CLOSE) {
+ an->initializer.push_back(n);
+ break;
+ } else if (auto_size && tk.type == TK_CURLY_BRACKET_CLOSE) {
+ an->initializer.push_back(n);
+ break;
+ } else {
+ if (auto_size) {
+ _set_error("Expected '}' or ','");
+ } else {
+ _set_error("Expected ')' or ','");
+ }
+ return nullptr;
+ }
+ }
+ if (an->initializer.size() != p_array_size) {
+ _set_error("Array size mismatch");
+ return nullptr;
+ }
+ } else {
+ _set_error("Expected array initialization!");
+ return nullptr;
+ }
+
+ return an;
+}
+
ShaderLanguage::Node *ShaderLanguage::_parse_expression(BlockNode *p_block, const FunctionInfo &p_function_info) {
Vector<Expression> expression;
@@ -3384,142 +3527,10 @@ ShaderLanguage::Node *ShaderLanguage::_parse_expression(BlockNode *p_block, cons
Node *nexpr;
if (pstruct->members[i]->array_size != 0) {
- DataType type = pstruct->members[i]->get_datatype();
- String struct_name = pstruct->members[i]->struct_name;
- int array_size = pstruct->members[i]->array_size;
-
- DataType type2;
- String struct_name2 = "";
- int array_size2 = 0;
-
- bool auto_size = false;
-
- tk = _get_token();
-
- if (tk.type == TK_CURLY_BRACKET_OPEN) {
- auto_size = true;
- } else {
- if (shader->structs.has(tk.text)) {
- type2 = TYPE_STRUCT;
- struct_name2 = tk.text;
- } else {
- if (!is_token_variable_datatype(tk.type)) {
- _set_error("Invalid data type for array");
- return nullptr;
- }
- type2 = get_token_datatype(tk.type);
- }
-
- tk = _get_token();
- if (tk.type == TK_BRACKET_OPEN) {
- TkPos pos2 = _get_tkpos();
- tk = _get_token();
- if (tk.type == TK_BRACKET_CLOSE) {
- array_size2 = array_size;
- tk = _get_token();
- } else {
- _set_tkpos(pos2);
-
- Node *n = _parse_and_reduce_expression(p_block, p_function_info);
- if (!n || n->type != Node::TYPE_CONSTANT || n->get_datatype() != TYPE_INT) {
- _set_error("Expected single integer constant > 0");
- return nullptr;
- }
-
- ConstantNode *cnode = (ConstantNode *)n;
- if (cnode->values.size() == 1) {
- array_size2 = cnode->values[0].sint;
- if (array_size2 <= 0) {
- _set_error("Expected single integer constant > 0");
- return nullptr;
- }
- } else {
- _set_error("Expected single integer constant > 0");
- return nullptr;
- }
-
- tk = _get_token();
- if (tk.type != TK_BRACKET_CLOSE) {
- _set_error("Expected ']'");
- return nullptr;
- } else {
- tk = _get_token();
- }
- }
- } else {
- _set_error("Expected '['");
- return nullptr;
- }
-
- if (type != type2 || struct_name != struct_name2 || array_size != array_size2) {
- String error_str = "Cannot convert from '";
- if (type2 == TYPE_STRUCT) {
- error_str += struct_name2;
- } else {
- error_str += get_datatype_name(type2);
- }
- error_str += "[";
- error_str += itos(array_size2);
- error_str += "]'";
- error_str += " to '";
- if (type == TYPE_STRUCT) {
- error_str += struct_name;
- } else {
- error_str += get_datatype_name(type);
- }
- error_str += "[";
- error_str += itos(array_size);
- error_str += "]'";
- _set_error(error_str);
- return nullptr;
- }
- }
-
- ArrayConstructNode *an = alloc_node<ArrayConstructNode>();
- an->datatype = type;
- an->struct_name = struct_name;
-
- if (tk.type == TK_PARENTHESIS_OPEN || auto_size) { // initialization
- while (true) {
- Node *n = _parse_and_reduce_expression(p_block, p_function_info);
- if (!n) {
- return nullptr;
- }
-
- if (type != n->get_datatype() || struct_name != n->get_datatype_name()) {
- _set_error("Invalid assignment of '" + (n->get_datatype() == TYPE_STRUCT ? n->get_datatype_name() : get_datatype_name(n->get_datatype())) + "' to '" + (type == TYPE_STRUCT ? struct_name : get_datatype_name(type)) + "'");
- return nullptr;
- }
-
- tk = _get_token();
- if (tk.type == TK_COMMA) {
- an->initializer.push_back(n);
- continue;
- } else if (!auto_size && tk.type == TK_PARENTHESIS_CLOSE) {
- an->initializer.push_back(n);
- break;
- } else if (auto_size && tk.type == TK_CURLY_BRACKET_CLOSE) {
- an->initializer.push_back(n);
- break;
- } else {
- if (auto_size) {
- _set_error("Expected '}' or ','");
- } else {
- _set_error("Expected ')' or ','");
- }
- return nullptr;
- }
- }
- if (an->initializer.size() != array_size) {
- _set_error("Array size mismatch");
- return nullptr;
- }
- } else {
- _set_error("Expected array initialization!");
+ nexpr = _parse_array_constructor(p_block, p_function_info, pstruct->members[i]->get_datatype(), pstruct->members[i]->struct_name, pstruct->members[i]->array_size);
+ if (!nexpr) {
return nullptr;
}
-
- nexpr = an;
} else {
nexpr = _parse_and_reduce_expression(p_block, p_function_info);
if (!nexpr) {
@@ -3722,6 +3733,7 @@ ShaderLanguage::Node *ShaderLanguage::_parse_expression(BlockNode *p_block, cons
} else {
//an identifier
+ last_const = false;
_set_tkpos(pos);
DataType data_type;
@@ -3749,6 +3761,7 @@ ShaderLanguage::Node *ShaderLanguage::_parse_expression(BlockNode *p_block, cons
_set_error("Unknown identifier in expression: " + String(identifier));
return nullptr;
}
+ last_const = is_const;
if (ident_type == IDENTIFIER_FUNCTION) {
_set_error("Can't use function as identifier: " + String(identifier));
@@ -3758,16 +3771,30 @@ ShaderLanguage::Node *ShaderLanguage::_parse_expression(BlockNode *p_block, cons
Node *index_expression = nullptr;
Node *call_expression = nullptr;
+ Node *assign_expression = nullptr;
if (array_size > 0) {
tk = _get_token();
- if (tk.type != TK_BRACKET_OPEN && tk.type != TK_PERIOD) {
- _set_error("Expected '[' or '.'");
+ if (tk.type != TK_BRACKET_OPEN && tk.type != TK_PERIOD && tk.type != TK_OP_ASSIGN) {
+ _set_error("Expected '[','.' or '='");
return nullptr;
}
- if (tk.type == TK_PERIOD) {
+ if (tk.type == TK_OP_ASSIGN) {
+ if (is_const) {
+ _set_error("Constants cannot be modified.");
+ return nullptr;
+ }
+ if (shader->varyings.has(identifier) && current_function != String("vertex")) {
+ _set_error("Varyings can only be assigned in vertex function.");
+ return nullptr;
+ }
+ assign_expression = _parse_array_constructor(p_block, p_function_info, data_type, struct_name, array_size);
+ if (!assign_expression) {
+ return nullptr;
+ }
+ } else if (tk.type == TK_PERIOD) {
completion_class = TAG_ARRAY;
p_block->block_tag = SubClassTag::TAG_ARRAY;
call_expression = _parse_and_reduce_expression(p_block, p_function_info);
@@ -3791,7 +3818,7 @@ ShaderLanguage::Node *ShaderLanguage::_parse_expression(BlockNode *p_block, cons
if (index_expression->type == Node::TYPE_CONSTANT) {
ConstantNode *cnode = (ConstantNode *)index_expression;
if (cnode) {
- if (!cnode->values.empty()) {
+ if (!cnode->values.is_empty()) {
int value = cnode->values[0].sint;
if (value < 0 || value >= array_size) {
_set_error(vformat("Index [%s] out of range [%s..%s]", value, 0, array_size - 1));
@@ -3814,6 +3841,7 @@ ShaderLanguage::Node *ShaderLanguage::_parse_expression(BlockNode *p_block, cons
arrname->struct_name = struct_name;
arrname->index_expression = index_expression;
arrname->call_expression = call_expression;
+ arrname->assign_expression = assign_expression;
arrname->is_const = is_const;
expr = arrname;
@@ -4154,7 +4182,18 @@ ShaderLanguage::Node *ShaderLanguage::_parse_expression(BlockNode *p_block, cons
if (array_size > 0) {
tk = _get_token();
- if (tk.type == TK_PERIOD) {
+ if (tk.type == TK_OP_ASSIGN) {
+ if (last_const) {
+ last_const = false;
+ _set_error("Constants cannot be modified.");
+ return nullptr;
+ }
+ Node *assign_expression = _parse_array_constructor(p_block, p_function_info, member_type, member_struct_name, array_size);
+ if (!assign_expression) {
+ return nullptr;
+ }
+ mn->assign_expression = assign_expression;
+ } else if (tk.type == TK_PERIOD) {
_set_error("Nested array length() is not yet implemented");
return nullptr;
} else if (tk.type == TK_BRACKET_OPEN) {
@@ -4171,7 +4210,7 @@ ShaderLanguage::Node *ShaderLanguage::_parse_expression(BlockNode *p_block, cons
if (index_expression->type == Node::TYPE_CONSTANT) {
ConstantNode *cnode = (ConstantNode *)index_expression;
if (cnode) {
- if (!cnode->values.empty()) {
+ if (!cnode->values.is_empty()) {
int value = cnode->values[0].sint;
if (value < 0 || value >= array_size) {
_set_error(vformat("Index [%s] out of range [%s..%s]", value, 0, array_size - 1));
@@ -4189,7 +4228,7 @@ ShaderLanguage::Node *ShaderLanguage::_parse_expression(BlockNode *p_block, cons
mn->index_expression = index_expression;
} else {
- _set_error("Expected '[' or '.'");
+ _set_error("Expected '[','.' or '='");
return nullptr;
}
}
@@ -5010,17 +5049,53 @@ Error ShaderLanguage::_parse_block(BlockNode *p_block, const FunctionInfo &p_fun
decl.name = name;
decl.size = 0U;
+ pos = _get_tkpos();
tk = _get_token();
if (tk.type == TK_BRACKET_CLOSE) {
unknown_size = true;
} else {
if (tk.type != TK_INT_CONSTANT || ((int)tk.constant) <= 0) {
+ _set_tkpos(pos);
+ Node *n = _parse_and_reduce_expression(p_block, p_function_info);
+ if (n) {
+ if (n->type == Node::TYPE_VARIABLE) {
+ VariableNode *vn = static_cast<VariableNode *>(n);
+ if (vn) {
+ ConstantNode::Value v;
+ DataType data_type;
+
+ _find_identifier(p_block, false, p_function_info, vn->name, &data_type, nullptr, &is_const, nullptr, nullptr, &v);
+
+ if (is_const) {
+ if (data_type == TYPE_INT) {
+ int32_t value = v.sint;
+ if (value > 0) {
+ node->size_expression = n;
+ decl.size = (uint32_t)value;
+ }
+ } else if (data_type == TYPE_UINT) {
+ uint32_t value = v.uint;
+ if (value > 0U) {
+ node->size_expression = n;
+ decl.size = value;
+ }
+ }
+ }
+ }
+ } else if (n->type == Node::TYPE_OPERATOR) {
+ _set_error("Array size expressions are not yet implemented.");
+ return ERR_PARSE_ERROR;
+ }
+ }
+ } else if (((int)tk.constant) > 0) {
+ decl.size = (uint32_t)tk.constant;
+ }
+
+ if (decl.size == 0U) {
_set_error("Expected integer constant > 0 or ']'");
return ERR_PARSE_ERROR;
}
-
- decl.size = ((uint32_t)tk.constant);
tk = _get_token();
if (tk.type != TK_BRACKET_CLOSE) {
@@ -5218,7 +5293,7 @@ Error ShaderLanguage::_parse_block(BlockNode *p_block, const FunctionInfo &p_fun
_set_error("Expected array initialization");
return ERR_PARSE_ERROR;
}
- if (is_const) {
+ if (node->is_const) {
_set_error("Expected initialization of constant");
return ERR_PARSE_ERROR;
}
@@ -5252,6 +5327,13 @@ Error ShaderLanguage::_parse_block(BlockNode *p_block, const FunctionInfo &p_fun
}
decl.initializer = n;
+ if (n->type == Node::TYPE_CONSTANT) {
+ ConstantNode *const_node = static_cast<ConstantNode *>(n);
+ if (const_node && const_node->values.size() == 1) {
+ var.value = const_node->values[0];
+ }
+ }
+
if (var.type == TYPE_STRUCT ? (var.struct_name != n->get_datatype_name()) : (var.type != n->get_datatype())) {
_set_error("Invalid assignment of '" + (n->get_datatype() == TYPE_STRUCT ? n->get_datatype_name() : get_datatype_name(n->get_datatype())) + "' to '" + (var.type == TYPE_STRUCT ? String(var.struct_name) : get_datatype_name(var.type)) + "'");
return ERR_PARSE_ERROR;
@@ -5420,18 +5502,29 @@ Error ShaderLanguage::_parse_block(BlockNode *p_block, const FunctionInfo &p_fun
ControlFlowNode *flow = (ControlFlowNode *)switch_block->statements[i];
if (flow) {
if (flow->flow_op == FLOW_OP_CASE) {
- ConstantNode *n2 = static_cast<ConstantNode *>(flow->expressions[0]);
- if (!n2) {
- return ERR_PARSE_ERROR;
- }
- if (n2->values.empty()) {
- return ERR_PARSE_ERROR;
- }
- if (constants.has(n2->values[0].sint)) {
- _set_error("Duplicated case label: '" + itos(n2->values[0].sint) + "'");
- return ERR_PARSE_ERROR;
+ if (flow->expressions[0]->type == Node::TYPE_CONSTANT) {
+ ConstantNode *cn = static_cast<ConstantNode *>(flow->expressions[0]);
+ if (!cn || cn->values.is_empty()) {
+ return ERR_PARSE_ERROR;
+ }
+ if (constants.has(cn->values[0].sint)) {
+ _set_error("Duplicated case label: '" + itos(cn->values[0].sint) + "'");
+ return ERR_PARSE_ERROR;
+ }
+ constants.insert(cn->values[0].sint);
+ } else if (flow->expressions[0]->type == Node::TYPE_VARIABLE) {
+ VariableNode *vn = static_cast<VariableNode *>(flow->expressions[0]);
+ if (!vn) {
+ return ERR_PARSE_ERROR;
+ }
+ ConstantNode::Value v;
+ _find_identifier(p_block, false, p_function_info, vn->name, nullptr, nullptr, nullptr, nullptr, nullptr, &v);
+ if (constants.has(v.sint)) {
+ _set_error("Duplicated case label: '" + itos(v.sint) + "'");
+ return ERR_PARSE_ERROR;
+ }
+ constants.insert(v.sint);
}
- constants.insert(n2->values[0].sint);
} else if (flow->flow_op == FLOW_OP_DEFAULT) {
continue;
} else {
@@ -5467,12 +5560,38 @@ Error ShaderLanguage::_parse_block(BlockNode *p_block, const FunctionInfo &p_fun
tk = _get_token();
}
+ Node *n = nullptr;
+
if (tk.type != TK_INT_CONSTANT) {
- _set_error("Expected integer constant");
- return ERR_PARSE_ERROR;
- }
+ bool correct_constant_expression = false;
+ DataType data_type;
+
+ if (tk.type == TK_IDENTIFIER) {
+ bool is_const;
+ _find_identifier(p_block, false, p_function_info, tk.text, &data_type, nullptr, &is_const);
+ if (is_const) {
+ if (data_type == TYPE_INT) {
+ correct_constant_expression = true;
+ }
+ }
+ }
+ if (!correct_constant_expression) {
+ _set_error("Expected integer constant");
+ return ERR_PARSE_ERROR;
+ }
+
+ VariableNode *vn = alloc_node<VariableNode>();
+ vn->name = tk.text;
+ n = vn;
+ } else {
+ ConstantNode::Value v;
+ v.sint = (int)tk.constant * sign;
- int constant = (int)tk.constant * sign;
+ ConstantNode *cn = alloc_node<ConstantNode>();
+ cn->values.push_back(v);
+ cn->datatype = TYPE_INT;
+ n = cn;
+ }
tk = _get_token();
@@ -5484,12 +5603,6 @@ Error ShaderLanguage::_parse_block(BlockNode *p_block, const FunctionInfo &p_fun
ControlFlowNode *cf = alloc_node<ControlFlowNode>();
cf->flow_op = FLOW_OP_CASE;
- ConstantNode *n = alloc_node<ConstantNode>();
- ConstantNode::Value v;
- v.sint = constant;
- n->values.push_back(v);
- n->datatype = TYPE_INT;
-
BlockNode *case_block = alloc_node<BlockNode>();
case_block->block_type = BlockNode::BLOCK_TYPE_CASE;
case_block->parent_block = p_block;
diff --git a/servers/rendering/shader_language.h b/servers/rendering/shader_language.h
index 9d2d591542..d8fc316f47 100644
--- a/servers/rendering/shader_language.h
+++ b/servers/rendering/shader_language.h
@@ -414,6 +414,7 @@ public:
StringName name;
Node *index_expression = nullptr;
Node *call_expression = nullptr;
+ Node *assign_expression = nullptr;
bool is_const = false;
virtual DataType get_datatype() const { return datatype_cache; }
@@ -437,6 +438,7 @@ public:
DataType datatype = TYPE_VOID;
String struct_name;
bool is_const = false;
+ Node *size_expression = nullptr;
struct Declaration {
StringName name;
@@ -496,6 +498,7 @@ public:
int line; //for completion
int array_size;
bool is_const;
+ ConstantNode::Value value;
};
Map<StringName, Variable> variables;
@@ -526,6 +529,7 @@ public:
StringName name;
Node *owner = nullptr;
Node *index_expression = nullptr;
+ Node *assign_expression = nullptr;
bool has_swizzling_duplicates = false;
virtual DataType get_datatype() const { return datatype; }
@@ -774,6 +778,7 @@ private:
int tk_line;
StringName current_function;
+ bool last_const = false;
struct TkPos {
int char_idx;
@@ -819,7 +824,7 @@ private:
IDENTIFIER_CONSTANT,
};
- bool _find_identifier(const BlockNode *p_block, bool p_allow_reassign, const FunctionInfo &p_function_info, const StringName &p_identifier, DataType *r_data_type = nullptr, IdentifierType *r_type = nullptr, bool *r_is_const = nullptr, int *r_array_size = nullptr, StringName *r_struct_name = nullptr);
+ bool _find_identifier(const BlockNode *p_block, bool p_allow_reassign, const FunctionInfo &p_function_info, const StringName &p_identifier, DataType *r_data_type = nullptr, IdentifierType *r_type = nullptr, bool *r_is_const = nullptr, int *r_array_size = nullptr, StringName *r_struct_name = nullptr, ConstantNode::Value *r_constant_value = nullptr);
bool _is_operator_assign(Operator p_op) const;
bool _validate_assign(Node *p_node, const FunctionInfo &p_function_info, String *r_message = nullptr);
bool _validate_operator(OperatorNode *p_op, DataType *r_ret_type = nullptr);
@@ -861,6 +866,7 @@ private:
bool _propagate_function_call_sampler_builtin_reference(StringName p_name, int p_argument, const StringName &p_builtin);
Node *_parse_expression(BlockNode *p_block, const FunctionInfo &p_function_info);
+ Node *_parse_array_constructor(BlockNode *p_block, const FunctionInfo &p_function_info, DataType p_type, const StringName &p_struct_name, int p_array_size);
ShaderLanguage::Node *_reduce_expression(BlockNode *p_block, ShaderLanguage::Node *p_node);
Node *_parse_and_reduce_expression(BlockNode *p_block, const FunctionInfo &p_function_info);
diff --git a/servers/rendering_server.cpp b/servers/rendering_server.cpp
index b643b14040..e9bfb7ecf5 100644
--- a/servers/rendering_server.cpp
+++ b/servers/rendering_server.cpp
@@ -1655,6 +1655,10 @@ void RenderingServer::_bind_methods() {
ClassDB::bind_method(D_METHOD("viewport_get_render_info", "viewport", "info"), &RenderingServer::viewport_get_render_info);
ClassDB::bind_method(D_METHOD("viewport_set_debug_draw", "viewport", "draw"), &RenderingServer::viewport_set_debug_draw);
+ ClassDB::bind_method(D_METHOD("viewport_set_measure_render_time", "viewport", "enable"), &RenderingServer::viewport_set_measure_render_time);
+ ClassDB::bind_method(D_METHOD("viewport_get_measured_render_time_cpu", "viewport"), &RenderingServer::viewport_get_measured_render_time_cpu);
+ ClassDB::bind_method(D_METHOD("viewport_get_measured_render_time_gpu", "viewport"), &RenderingServer::viewport_get_measured_render_time_gpu);
+
ClassDB::bind_method(D_METHOD("environment_create"), &RenderingServer::environment_create);
ClassDB::bind_method(D_METHOD("environment_set_background", "env", "bg"), &RenderingServer::environment_set_background);
ClassDB::bind_method(D_METHOD("environment_set_sky", "env", "sky"), &RenderingServer::environment_set_sky);
@@ -1668,7 +1672,7 @@ void RenderingServer::_bind_methods() {
ClassDB::bind_method(D_METHOD("environment_set_tonemap", "env", "tone_mapper", "exposure", "white", "auto_exposure", "min_luminance", "max_luminance", "auto_exp_speed", "auto_exp_grey"), &RenderingServer::environment_set_tonemap);
ClassDB::bind_method(D_METHOD("environment_set_adjustment", "env", "enable", "brightness", "contrast", "saturation", "use_1d_color_correction", "color_correction"), &RenderingServer::environment_set_adjustment);
ClassDB::bind_method(D_METHOD("environment_set_ssr", "env", "enable", "max_steps", "fade_in", "fade_out", "depth_tolerance"), &RenderingServer::environment_set_ssr);
- ClassDB::bind_method(D_METHOD("environment_set_ssao", "env", "enable", "radius", "intensity", "bias", "light_affect", "ao_channel_affect", "blur", "bilateral_sharpness"), &RenderingServer::environment_set_ssao);
+ ClassDB::bind_method(D_METHOD("environment_set_ssao", "env", "enable", "radius", "intensity", "power", "detail", "horizon", "sharpness", "light_affect", "ao_channel_affect"), &RenderingServer::environment_set_ssao);
ClassDB::bind_method(D_METHOD("environment_set_fog", "env", "enable", "light_color", "light_energy", "sun_scatter", "density", "height", "height_density", "aerial_perspective"), &RenderingServer::environment_set_fog);
ClassDB::bind_method(D_METHOD("scenario_create"), &RenderingServer::scenario_create);
@@ -1813,6 +1817,9 @@ void RenderingServer::_bind_methods() {
ClassDB::bind_method(D_METHOD("is_render_loop_enabled"), &RenderingServer::is_render_loop_enabled);
ClassDB::bind_method(D_METHOD("set_render_loop_enabled", "enabled"), &RenderingServer::set_render_loop_enabled);
+
+ ClassDB::bind_method(D_METHOD("get_frame_setup_time_cpu"), &RenderingServer::get_frame_setup_time_cpu);
+
ADD_PROPERTY(PropertyInfo(Variant::BOOL, "render_loop_enabled"), "set_render_loop_enabled", "is_render_loop_enabled");
BIND_CONSTANT(NO_INDEX_ARRAY);
@@ -2038,11 +2045,7 @@ void RenderingServer::_bind_methods() {
BIND_ENUM_CONSTANT(ENV_SSR_ROUGNESS_QUALITY_MEDIUM);
BIND_ENUM_CONSTANT(ENV_SSR_ROUGNESS_QUALITY_HIGH);
- BIND_ENUM_CONSTANT(ENV_SSAO_BLUR_DISABLED);
- BIND_ENUM_CONSTANT(ENV_SSAO_BLUR_1x1);
- BIND_ENUM_CONSTANT(ENV_SSAO_BLUR_2x2);
- BIND_ENUM_CONSTANT(ENV_SSAO_BLUR_3x3);
-
+ BIND_ENUM_CONSTANT(ENV_SSAO_QUALITY_VERY_LOW);
BIND_ENUM_CONSTANT(ENV_SSAO_QUALITY_LOW);
BIND_ENUM_CONSTANT(ENV_SSAO_QUALITY_MEDIUM);
BIND_ENUM_CONSTANT(ENV_SSAO_QUALITY_HIGH);
@@ -2309,9 +2312,18 @@ RenderingServer::RenderingServer() {
ProjectSettings::get_singleton()->set_custom_property_info("rendering/quality/depth_of_field/depth_of_field_bokeh_quality", PropertyInfo(Variant::INT, "rendering/quality/depth_of_field/depth_of_field_bokeh_quality", PROPERTY_HINT_ENUM, "Very Low (Fastest),Low (Fast),Medium (Average),High (Slow)"));
GLOBAL_DEF("rendering/quality/depth_of_field/depth_of_field_use_jitter", false);
- GLOBAL_DEF("rendering/quality/ssao/quality", 1);
- ProjectSettings::get_singleton()->set_custom_property_info("rendering/quality/ssao/quality", PropertyInfo(Variant::INT, "rendering/quality/ssao/quality", PROPERTY_HINT_ENUM, "Low (Fast),Medium (Average),High (Slow),Ultra (Slower)"));
+ GLOBAL_DEF("rendering/quality/ssao/quality", 2);
+ ProjectSettings::get_singleton()->set_custom_property_info("rendering/quality/ssao/quality", PropertyInfo(Variant::INT, "rendering/quality/ssao/quality", PROPERTY_HINT_ENUM, "Very Low (Fast),Low (Fast),Medium (Average),High (Slow),Ultra (Custom)"));
GLOBAL_DEF("rendering/quality/ssao/half_size", false);
+ GLOBAL_DEF("rendering/quality/ssao/half_size.mobile", true);
+ GLOBAL_DEF("rendering/quality/ssao/adaptive_target", 0.5);
+ ProjectSettings::get_singleton()->set_custom_property_info("rendering/quality/ssao/adaptive_target", PropertyInfo(Variant::FLOAT, "rendering/quality/ssao/adaptive_target", PROPERTY_HINT_RANGE, "0.0,1.0,0.01"));
+ GLOBAL_DEF("rendering/quality/ssao/blur_passes", 2);
+ ProjectSettings::get_singleton()->set_custom_property_info("rendering/quality/ssao/blur_passes", PropertyInfo(Variant::INT, "rendering/quality/ssao/blur_passes", PROPERTY_HINT_RANGE, "0,6"));
+ GLOBAL_DEF("rendering/quality/ssao/fadeout_from", 50.0);
+ ProjectSettings::get_singleton()->set_custom_property_info("rendering/quality/ssao/fadeout_from", PropertyInfo(Variant::FLOAT, "rendering/quality/ssao/fadeout_from", PROPERTY_HINT_RANGE, "0.0,512,0.1,or_greater"));
+ GLOBAL_DEF("rendering/quality/ssao/fadeout_to", 300.0);
+ ProjectSettings::get_singleton()->set_custom_property_info("rendering/quality/ssao/fadeout_to", PropertyInfo(Variant::FLOAT, "rendering/quality/ssao/fadeout_to", PROPERTY_HINT_RANGE, "64,65536,0.1,or_greater"));
GLOBAL_DEF("rendering/quality/screen_filters/screen_space_roughness_limiter_enabled", true);
GLOBAL_DEF("rendering/quality/screen_filters/screen_space_roughness_limiter_amount", 0.25);
@@ -2354,6 +2366,9 @@ RenderingServer::RenderingServer() {
ProjectSettings::get_singleton()->set_custom_property_info("rendering/volumetric_fog/directional_shadow_shrink", PropertyInfo(Variant::INT, "rendering/volumetric_fog/directional_shadow_shrink", PROPERTY_HINT_RANGE, "32,2048,1"));
GLOBAL_DEF("rendering/volumetric_fog/positional_shadow_shrink", 512);
ProjectSettings::get_singleton()->set_custom_property_info("rendering/volumetric_fog/positional_shadow_shrink", PropertyInfo(Variant::INT, "rendering/volumetric_fog/positional_shadow_shrink", PROPERTY_HINT_RANGE, "32,2048,1"));
+
+ GLOBAL_DEF("rendering/spatial_indexer/update_iterations_per_frame", 10);
+ ProjectSettings::get_singleton()->set_custom_property_info("rendering/spatial_indexer/update_iterations_per_frame", PropertyInfo(Variant::INT, "rendering/spatial_indexer/update_iterations_per_frame", PROPERTY_HINT_RANGE, "0,1024,1"));
}
RenderingServer::~RenderingServer() {
diff --git a/servers/rendering_server.h b/servers/rendering_server.h
index 6f7359b276..f403264a96 100644
--- a/servers/rendering_server.h
+++ b/servers/rendering_server.h
@@ -940,23 +940,17 @@ public:
virtual void environment_set_ssr_roughness_quality(EnvironmentSSRRoughnessQuality p_quality) = 0;
- enum EnvironmentSSAOBlur {
- ENV_SSAO_BLUR_DISABLED,
- ENV_SSAO_BLUR_1x1,
- ENV_SSAO_BLUR_2x2,
- ENV_SSAO_BLUR_3x3,
- };
-
- virtual void environment_set_ssao(RID p_env, bool p_enable, float p_radius, float p_intensity, float p_bias, float p_light_affect, float p_ao_channel_affect, EnvironmentSSAOBlur p_blur, float p_bilateral_sharpness) = 0;
+ virtual void 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) = 0;
enum EnvironmentSSAOQuality {
+ ENV_SSAO_QUALITY_VERY_LOW,
ENV_SSAO_QUALITY_LOW,
ENV_SSAO_QUALITY_MEDIUM,
ENV_SSAO_QUALITY_HIGH,
ENV_SSAO_QUALITY_ULTRA,
};
- virtual void environment_set_ssao_quality(EnvironmentSSAOQuality p_quality, bool p_half_size) = 0;
+ virtual void environment_set_ssao_quality(EnvironmentSSAOQuality p_quality, bool p_half_size, float p_adaptive_target, int p_blur_passes, float p_fadeout_from, float p_fadeout_to) = 0;
enum EnvironmentSDFGICascades {
ENV_SDFGI_CASCADES_4,
@@ -1415,6 +1409,8 @@ public:
virtual Vector<FrameProfileArea> get_frame_profile() = 0;
virtual uint64_t get_frame_profile_frame() = 0;
+ virtual float get_frame_setup_time_cpu() const = 0;
+
/* TESTING */
virtual RID get_test_cube() = 0;
@@ -1488,7 +1484,6 @@ VARIANT_ENUM_CAST(RenderingServer::EnvironmentReflectionSource);
VARIANT_ENUM_CAST(RenderingServer::EnvironmentGlowBlendMode);
VARIANT_ENUM_CAST(RenderingServer::EnvironmentToneMapper);
VARIANT_ENUM_CAST(RenderingServer::EnvironmentSSRRoughnessQuality);
-VARIANT_ENUM_CAST(RenderingServer::EnvironmentSSAOBlur);
VARIANT_ENUM_CAST(RenderingServer::EnvironmentSSAOQuality);
VARIANT_ENUM_CAST(RenderingServer::SubSurfaceScatteringQuality);
VARIANT_ENUM_CAST(RenderingServer::DOFBlurQuality);
diff --git a/servers/text_server.cpp b/servers/text_server.cpp
index 30dfa60ba3..bdd0dbcf4e 100644
--- a/servers/text_server.cpp
+++ b/servers/text_server.cpp
@@ -550,7 +550,7 @@ void TextServer::draw_hex_code_box(RID p_canvas, int p_size, const Vector2 &p_po
Vector<Vector2i> TextServer::shaped_text_get_line_breaks_adv(RID p_shaped, const Vector<float> &p_width, int p_start, bool p_once, uint8_t /*TextBreakFlag*/ p_break_flags) const {
Vector<Vector2i> lines;
- ERR_FAIL_COND_V(p_width.empty(), lines);
+ ERR_FAIL_COND_V(p_width.is_empty(), lines);
const_cast<TextServer *>(this)->shaped_text_update_breaks(p_shaped);
const Vector<Glyph> &logical = const_cast<TextServer *>(this)->shaped_text_sort_logical(p_shaped);
@@ -905,7 +905,7 @@ Vector<Vector2> TextServer::shaped_text_get_selection(RID p_shaped, int p_start,
float off = 0.0f;
for (int i = 0; i < v_size; i++) {
for (int k = 0; k < glyphs[i].repeat; k++) {
- if (glyphs[i].count > 0 && glyphs[i].index != 0) {
+ if ((glyphs[i].count > 0) && ((glyphs[i].index != 0) || ((glyphs[i].flags & GRAPHEME_IS_SPACE) == GRAPHEME_IS_SPACE))) {
if (glyphs[i].start < end && glyphs[i].end > start) {
// Grapheme fully in selection range.
if (glyphs[i].start >= start && glyphs[i].end <= end) {
@@ -963,6 +963,10 @@ Vector<Vector2> TextServer::shaped_text_get_selection(RID p_shaped, int p_start,
// Merge intersecting ranges.
int i = 0;
while (i < ranges.size()) {
+ i++;
+ }
+ i = 0;
+ while (i < ranges.size()) {
int j = i + 1;
while (j < ranges.size()) {
if (Math::is_equal_approx(ranges[i].y, ranges[j].x, UNIT_EPSILON)) {
@@ -1034,31 +1038,36 @@ int TextServer::shaped_text_hit_test_position(RID p_shaped, float p_coords) cons
float off = 0.0f;
for (int i = 0; i < v_size; i++) {
- for (int k = 0; k < glyphs[i].repeat; k++) {
- if (glyphs[i].count > 0) {
- float advance = 0.f;
- for (int j = 0; j < glyphs[i].count; j++) {
- advance += glyphs[i + j].advance;
+ if (glyphs[i].count > 0) {
+ float advance = 0.f;
+ for (int j = 0; j < glyphs[i].count; j++) {
+ advance += glyphs[i + j].advance * glyphs[i + j].repeat;
+ }
+ if (((glyphs[i].flags & GRAPHEME_IS_VIRTUAL) == GRAPHEME_IS_VIRTUAL) && (p_coords >= off && p_coords < off + advance)) {
+ if ((glyphs[i].flags & GRAPHEME_IS_RTL) == GRAPHEME_IS_RTL) {
+ return glyphs[i].end;
+ } else {
+ return glyphs[i].start;
}
- // Place caret to the left of clicked grapheme.
- if (p_coords >= off && p_coords < off + advance / 2) {
- if ((glyphs[i].flags & GRAPHEME_IS_RTL) == GRAPHEME_IS_RTL) {
- return glyphs[i].end;
- } else {
- return glyphs[i].start;
- }
+ }
+ // Place caret to the left of clicked grapheme.
+ if (p_coords >= off && p_coords < off + advance / 2) {
+ if ((glyphs[i].flags & GRAPHEME_IS_RTL) == GRAPHEME_IS_RTL) {
+ return glyphs[i].end;
+ } else {
+ return glyphs[i].start;
}
- // Place caret to the right of clicked grapheme.
- if (p_coords >= off + advance / 2 && p_coords < off + advance) {
- if ((glyphs[i].flags & GRAPHEME_IS_RTL) == GRAPHEME_IS_RTL) {
- return glyphs[i].start;
- } else {
- return glyphs[i].end;
- }
+ }
+ // Place caret to the right of clicked grapheme.
+ if (p_coords >= off + advance / 2 && p_coords < off + advance) {
+ if ((glyphs[i].flags & GRAPHEME_IS_RTL) == GRAPHEME_IS_RTL) {
+ return glyphs[i].start;
+ } else {
+ return glyphs[i].end;
}
}
- off += glyphs[i].advance;
}
+ off += glyphs[i].advance * glyphs[i].repeat;
}
return 0;
}
diff --git a/servers/xr/xr_positional_tracker.cpp b/servers/xr/xr_positional_tracker.cpp
index 4455ca3db1..91469b1aec 100644
--- a/servers/xr/xr_positional_tracker.cpp
+++ b/servers/xr/xr_positional_tracker.cpp
@@ -42,17 +42,17 @@ void XRPositionalTracker::_bind_methods() {
ClassDB::bind_method(D_METHOD("get_tracker_id"), &XRPositionalTracker::get_tracker_id);
ClassDB::bind_method(D_METHOD("get_tracker_name"), &XRPositionalTracker::get_tracker_name);
ClassDB::bind_method(D_METHOD("get_joy_id"), &XRPositionalTracker::get_joy_id);
- ClassDB::bind_method(D_METHOD("get_tracks_orientation"), &XRPositionalTracker::get_tracks_orientation);
+ ClassDB::bind_method(D_METHOD("is_tracking_orientation"), &XRPositionalTracker::is_tracking_orientation);
ClassDB::bind_method(D_METHOD("get_orientation"), &XRPositionalTracker::get_orientation);
- ClassDB::bind_method(D_METHOD("get_tracks_position"), &XRPositionalTracker::get_tracks_position);
+ ClassDB::bind_method(D_METHOD("is_tracking_position"), &XRPositionalTracker::is_tracking_position);
ClassDB::bind_method(D_METHOD("get_position"), &XRPositionalTracker::get_position);
- ClassDB::bind_method(D_METHOD("get_hand"), &XRPositionalTracker::get_hand);
+ ClassDB::bind_method(D_METHOD("get_tracker_hand"), &XRPositionalTracker::get_tracker_hand);
ClassDB::bind_method(D_METHOD("get_transform", "adjust_by_reference_frame"), &XRPositionalTracker::get_transform);
ClassDB::bind_method(D_METHOD("get_mesh"), &XRPositionalTracker::get_mesh);
// these functions we don't want to expose to normal users but do need to be callable from GDNative
- ClassDB::bind_method(D_METHOD("_set_type", "type"), &XRPositionalTracker::set_type);
- ClassDB::bind_method(D_METHOD("_set_name", "name"), &XRPositionalTracker::set_name);
+ ClassDB::bind_method(D_METHOD("_set_tracker_type", "type"), &XRPositionalTracker::set_tracker_type);
+ ClassDB::bind_method(D_METHOD("_set_tracker_name", "name"), &XRPositionalTracker::set_tracker_name);
ClassDB::bind_method(D_METHOD("_set_joy_id", "joy_id"), &XRPositionalTracker::set_joy_id);
ClassDB::bind_method(D_METHOD("_set_orientation", "orientation"), &XRPositionalTracker::set_orientation);
ClassDB::bind_method(D_METHOD("_set_rw_position", "rw_position"), &XRPositionalTracker::set_rw_position);
@@ -63,7 +63,7 @@ void XRPositionalTracker::_bind_methods() {
ADD_PROPERTY(PropertyInfo(Variant::FLOAT, "rumble"), "set_rumble", "get_rumble");
};
-void XRPositionalTracker::set_type(XRServer::TrackerType p_type) {
+void XRPositionalTracker::set_tracker_type(XRServer::TrackerType p_type) {
if (type != p_type) {
type = p_type;
hand = XRPositionalTracker::TRACKER_HAND_UNKNOWN;
@@ -81,7 +81,7 @@ XRServer::TrackerType XRPositionalTracker::get_tracker_type() const {
return type;
};
-void XRPositionalTracker::set_name(const String &p_name) {
+void XRPositionalTracker::set_tracker_name(const String &p_name) {
name = p_name;
};
@@ -101,14 +101,14 @@ int XRPositionalTracker::get_joy_id() const {
return joy_id;
};
-bool XRPositionalTracker::get_tracks_orientation() const {
- return tracks_orientation;
+bool XRPositionalTracker::is_tracking_orientation() const {
+ return tracking_orientation;
};
void XRPositionalTracker::set_orientation(const Basis &p_orientation) {
_THREAD_SAFE_METHOD_
- tracks_orientation = true; // obviously we have this
+ tracking_orientation = true; // obviously we have this
orientation = p_orientation;
};
@@ -118,8 +118,8 @@ Basis XRPositionalTracker::get_orientation() const {
return orientation;
};
-bool XRPositionalTracker::get_tracks_position() const {
- return tracks_position;
+bool XRPositionalTracker::is_tracking_position() const {
+ return tracking_position;
};
void XRPositionalTracker::set_position(const Vector3 &p_position) {
@@ -130,7 +130,7 @@ void XRPositionalTracker::set_position(const Vector3 &p_position) {
real_t world_scale = xr_server->get_world_scale();
ERR_FAIL_COND(world_scale == 0);
- tracks_position = true; // obviously we have this
+ tracking_position = true; // obviously we have this
rw_position = p_position / world_scale;
};
@@ -147,7 +147,7 @@ Vector3 XRPositionalTracker::get_position() const {
void XRPositionalTracker::set_rw_position(const Vector3 &p_rw_position) {
_THREAD_SAFE_METHOD_
- tracks_position = true; // obviously we have this
+ tracking_position = true; // obviously we have this
rw_position = p_rw_position;
};
@@ -169,11 +169,11 @@ Ref<Mesh> XRPositionalTracker::get_mesh() const {
return mesh;
};
-XRPositionalTracker::TrackerHand XRPositionalTracker::get_hand() const {
+XRPositionalTracker::TrackerHand XRPositionalTracker::get_tracker_hand() const {
return hand;
};
-void XRPositionalTracker::set_hand(const XRPositionalTracker::TrackerHand p_hand) {
+void XRPositionalTracker::set_tracker_hand(const XRPositionalTracker::TrackerHand p_hand) {
XRServer *xr_server = XRServer::get_singleton();
ERR_FAIL_NULL(xr_server);
@@ -227,8 +227,8 @@ XRPositionalTracker::XRPositionalTracker() {
name = "Unknown";
joy_id = -1;
tracker_id = 0;
- tracks_orientation = false;
- tracks_position = false;
+ tracking_orientation = false;
+ tracking_position = false;
hand = TRACKER_HAND_UNKNOWN;
rumble = 0.0;
};
diff --git a/servers/xr/xr_positional_tracker.h b/servers/xr/xr_positional_tracker.h
index 4bf2017224..c0976d3c13 100644
--- a/servers/xr/xr_positional_tracker.h
+++ b/servers/xr/xr_positional_tracker.h
@@ -59,9 +59,9 @@ private:
StringName name; // (unique) name of the tracker
int tracker_id; // tracker index id that is unique per type
int joy_id; // if we also have a related joystick entity, the id of the joystick
- bool tracks_orientation; // do we track orientation?
+ bool tracking_orientation; // do we track orientation?
Basis orientation; // our orientation
- bool tracks_position; // do we track position?
+ bool tracking_position; // do we track position?
Vector3 rw_position; // our position "in the real world, so without world_scale applied"
Ref<Mesh> mesh; // when available, a mesh that can be used to render this tracker
TrackerHand hand; // if known, the hand this tracker is held in
@@ -71,23 +71,23 @@ protected:
static void _bind_methods();
public:
- void set_type(XRServer::TrackerType p_type);
+ void set_tracker_type(XRServer::TrackerType p_type);
XRServer::TrackerType get_tracker_type() const;
- void set_name(const String &p_name);
+ void set_tracker_name(const String &p_name);
StringName get_tracker_name() const;
int get_tracker_id() const;
void set_joy_id(int p_joy_id);
int get_joy_id() const;
- bool get_tracks_orientation() const;
+ bool is_tracking_orientation() const;
void set_orientation(const Basis &p_orientation);
Basis get_orientation() const;
- bool get_tracks_position() const;
+ bool is_tracking_position() const;
void set_position(const Vector3 &p_position); // set position with world_scale applied
Vector3 get_position() const; // get position with world_scale applied
void set_rw_position(const Vector3 &p_rw_position);
Vector3 get_rw_position() const;
- XRPositionalTracker::TrackerHand get_hand() const;
- void set_hand(const XRPositionalTracker::TrackerHand p_hand);
+ XRPositionalTracker::TrackerHand get_tracker_hand() const;
+ void set_tracker_hand(const XRPositionalTracker::TrackerHand p_hand);
real_t get_rumble() const;
void set_rumble(real_t p_rumble);
void set_mesh(const Ref<Mesh> &p_mesh);