diff options
Diffstat (limited to 'servers/rendering/renderer_rd')
43 files changed, 587 insertions, 372 deletions
diff --git a/servers/rendering/renderer_rd/cluster_builder_rd.cpp b/servers/rendering/renderer_rd/cluster_builder_rd.cpp index 73a0c652a4..959a752fba 100644 --- a/servers/rendering/renderer_rd/cluster_builder_rd.cpp +++ b/servers/rendering/renderer_rd/cluster_builder_rd.cpp @@ -74,7 +74,7 @@ ClusterBuilderSharedDataRD::ClusterBuilderSharedDataRD() { cluster_debug.shader_pipeline = RD::get_singleton()->compute_pipeline_create(cluster_debug.shader); } - { // SPHERE + { // Sphere mesh data. static const uint32_t icosphere_vertex_count = 42; static const float icosphere_vertices[icosphere_vertex_count * 3] = { 0, 0, -1, 0.7236073, -0.5257253, -0.4472195, -0.276388, -0.8506492, -0.4472199, -0.8944262, 0, -0.4472156, -0.276388, 0.8506492, -0.4472199, 0.7236073, 0.5257253, -0.4472195, 0.276388, -0.8506492, 0.4472199, -0.7236073, -0.5257253, 0.4472195, -0.7236073, 0.5257253, 0.4472195, 0.276388, 0.8506492, 0.4472199, 0.8944262, 0, 0.4472156, 0, 0, 1, -0.1624555, -0.4999952, -0.8506544, 0.4253227, -0.3090114, -0.8506542, 0.2628688, -0.8090116, -0.5257377, 0.8506479, 0, -0.5257359, 0.4253227, 0.3090114, -0.8506542, -0.5257298, 0, -0.8506517, -0.6881894, -0.4999969, -0.5257362, -0.1624555, 0.4999952, -0.8506544, -0.6881894, 0.4999969, -0.5257362, 0.2628688, 0.8090116, -0.5257377, 0.9510579, -0.3090126, 0, 0.9510579, 0.3090126, 0, 0, -1, 0, 0.5877856, -0.8090167, 0, -0.9510579, -0.3090126, 0, -0.5877856, -0.8090167, 0, -0.5877856, 0.8090167, 0, -0.9510579, 0.3090126, 0, 0.5877856, 0.8090167, 0, 0, 1, 0, 0.6881894, -0.4999969, 0.5257362, -0.2628688, -0.8090116, 0.5257377, -0.8506479, 0, 0.5257359, -0.2628688, 0.8090116, 0.5257377, 0.6881894, 0.4999969, 0.5257362, 0.1624555, -0.4999952, 0.8506544, 0.5257298, 0, 0.8506517, -0.4253227, -0.3090114, 0.8506542, -0.4253227, 0.3090114, 0.8506542, 0.1624555, 0.4999952, 0.8506544 @@ -118,7 +118,7 @@ ClusterBuilderSharedDataRD::ClusterBuilderSharedDataRD() { sphere_overfit = 1.0 / min_d; } - { // CONE + { // Cone mesh data. static const uint32_t cone_vertex_count = 99; static const float cone_vertices[cone_vertex_count * 3] = { 0, 1, -1, 0.1950903, 0.9807853, -1, 0.3826835, 0.9238795, -1, 0.5555703, 0.8314696, -1, 0.7071068, 0.7071068, -1, 0.8314697, 0.5555702, -1, 0.9238795, 0.3826834, -1, 0.9807853, 0.1950903, -1, 1, 0, -1, 0.9807853, -0.1950902, -1, 0.9238796, -0.3826833, -1, 0.8314697, -0.5555702, -1, 0.7071068, -0.7071068, -1, 0.5555702, -0.8314697, -1, 0.3826833, -0.9238796, -1, 0.1950901, -0.9807853, -1, -3.25841e-7, -1, -1, -0.1950907, -0.9807852, -1, -0.3826839, -0.9238793, -1, -0.5555707, -0.8314693, -1, -0.7071073, -0.7071063, -1, -0.83147, -0.5555697, -1, -0.9238799, -0.3826827, -1, 0, 0, 0, -0.9807854, -0.1950894, -1, -1, 9.65599e-7, -1, -0.9807851, 0.1950913, -1, -0.9238791, 0.3826845, -1, -0.8314689, 0.5555713, -1, -0.7071059, 0.7071077, -1, -0.5555691, 0.8314704, -1, -0.3826821, 0.9238801, -1, -0.1950888, 0.9807856, -1 @@ -172,7 +172,7 @@ ClusterBuilderSharedDataRD::ClusterBuilderSharedDataRD() { cone_overfit = 1.0 / min_d; } - { // BOX + { // Box mesh data. static const uint32_t box_vertex_count = 8; static const float box_vertices[box_vertex_count * 3] = { -1, -1, -1, -1, -1, 1, -1, 1, -1, -1, 1, 1, 1, -1, -1, 1, -1, 1, 1, 1, -1, 1, 1, 1 @@ -219,8 +219,9 @@ ClusterBuilderSharedDataRD::~ClusterBuilderSharedDataRD() { void ClusterBuilderRD::_clear() { if (cluster_buffer.is_null()) { - return; //nothing to clear + return; } + RD::get_singleton()->free(cluster_buffer); RD::get_singleton()->free(cluster_render_buffer); RD::get_singleton()->free(element_buffer); @@ -254,7 +255,7 @@ void ClusterBuilderRD::setup(Size2i p_screen_size, uint32_t p_max_elements, RID cluster_screen_size.height = (p_screen_size.height - 1) / cluster_size + 1; max_elements_by_type = p_max_elements; - if (max_elements_by_type % 32) { //need to be 32 aligned + if (max_elements_by_type % 32) { // Needs to be aligned to 32. max_elements_by_type += 32 - (max_elements_by_type % 32); } @@ -264,7 +265,8 @@ void ClusterBuilderRD::setup(Size2i p_screen_size, uint32_t p_max_elements, RID uint32_t element_tag_bits_size = render_element_max / 32; uint32_t element_tag_depth_bits_size = render_element_max; - cluster_render_buffer_size = cluster_screen_size.x * cluster_screen_size.y * (element_tag_bits_size + element_tag_depth_bits_size) * 4; // tag bits (element was used) and tag depth (depth range in which it was used) + + cluster_render_buffer_size = cluster_screen_size.x * cluster_screen_size.y * (element_tag_bits_size + element_tag_depth_bits_size) * 4; // Tag bits (element was used) and tag depth (depth range in which it was used). cluster_render_buffer = RD::get_singleton()->storage_buffer_create(cluster_render_buffer_size); cluster_buffer = RD::get_singleton()->storage_buffer_create(cluster_buffer_size); @@ -379,9 +381,9 @@ void ClusterBuilderRD::begin(const Transform3D &p_view_transform, const Projecti projection = p_cam_projection; z_near = projection.get_z_near(); z_far = projection.get_z_far(); - orthogonal = p_cam_projection.is_orthogonal(); + camera_orthogonal = p_cam_projection.is_orthogonal(); adjusted_projection = projection; - if (!orthogonal) { + if (!camera_orthogonal) { adjusted_projection.adjust_perspective_znear(0.0001); } @@ -390,7 +392,7 @@ void ClusterBuilderRD::begin(const Transform3D &p_view_transform, const Projecti projection = correction * projection; adjusted_projection = correction * adjusted_projection; - //reset counts + // Reset counts. render_element_count = 0; for (uint32_t i = 0; i < ELEMENT_TYPE_MAX; i++) { cluster_count_by_type[i] = 0; @@ -402,14 +404,14 @@ void ClusterBuilderRD::bake_cluster() { RD::get_singleton()->draw_command_begin_label("Bake Light Cluster"); - //clear cluster buffer + // Clear cluster buffer. RD::get_singleton()->buffer_clear(cluster_buffer, 0, cluster_buffer_size, RD::BARRIER_MASK_RASTER | RD::BARRIER_MASK_COMPUTE); if (render_element_count > 0) { - //clear render buffer + // Clear render buffer. RD::get_singleton()->buffer_clear(cluster_render_buffer, 0, cluster_render_buffer_size, RD::BARRIER_MASK_RASTER); - { //fill state uniform + { // Fill state uniform. StateUniform state; @@ -425,13 +427,13 @@ void ClusterBuilderRD::bake_cluster() { RD::get_singleton()->buffer_update(state_uniform, 0, sizeof(StateUniform), &state, RD::BARRIER_MASK_RASTER | RD::BARRIER_MASK_COMPUTE); } - //update instances + // Update instances. RD::get_singleton()->buffer_update(element_buffer, 0, sizeof(RenderElementData) * render_element_count, render_elements, RD::BARRIER_MASK_RASTER | RD::BARRIER_MASK_COMPUTE); RENDER_TIMESTAMP("Render 3D Cluster Elements"); - //render elements + // Render elements. { RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(framebuffer, RD::INITIAL_ACTION_DROP, RD::FINAL_ACTION_DISCARD, RD::INITIAL_ACTION_DROP, RD::FINAL_ACTION_DISCARD); ClusterBuilderSharedDataRD::ClusterRender::PushConstant push_constant = {}; @@ -447,8 +449,16 @@ void ClusterBuilderRD::bake_cluster() { RD::get_singleton()->draw_list_bind_index_array(draw_list, shared->sphere_index_array); } break; case ELEMENT_TYPE_SPOT_LIGHT: { - RD::get_singleton()->draw_list_bind_vertex_array(draw_list, shared->cone_vertex_array); - RD::get_singleton()->draw_list_bind_index_array(draw_list, shared->cone_index_array); + // If the spot angle is above a certain threshold, use a sphere instead of a cone for building the clusters + // since the cone gets too flat/large (spot angle close to 90 degrees) or + // can't even cover the affected area of the light (spot angle above 90 degrees). + if (render_elements[i].has_wide_spot_angle) { + RD::get_singleton()->draw_list_bind_vertex_array(draw_list, shared->sphere_vertex_array); + RD::get_singleton()->draw_list_bind_index_array(draw_list, shared->sphere_index_array); + } else { + RD::get_singleton()->draw_list_bind_vertex_array(draw_list, shared->cone_vertex_array); + RD::get_singleton()->draw_list_bind_index_array(draw_list, shared->cone_index_array); + } } break; case ELEMENT_TYPE_DECAL: case ELEMENT_TYPE_REFLECTION_PROBE: { @@ -465,7 +475,7 @@ void ClusterBuilderRD::bake_cluster() { } RD::get_singleton()->draw_list_end(RD::BARRIER_MASK_COMPUTE); } - //store elements + // Store elements. RENDER_TIMESTAMP("Pack 3D Cluster Elements"); { @@ -509,7 +519,7 @@ void ClusterBuilderRD::debug(ElementType p_element) { push_constant.cluster_screen_size[1] = cluster_screen_size.y; push_constant.cluster_shift = get_shift_from_power_of_2(cluster_size); push_constant.cluster_type = p_element; - push_constant.orthogonal = orthogonal; + push_constant.orthogonal = camera_orthogonal; push_constant.z_far = z_far; push_constant.z_near = z_near; push_constant.max_cluster_element_count_div_32 = max_elements_by_type / 32; diff --git a/servers/rendering/renderer_rd/cluster_builder_rd.h b/servers/rendering/renderer_rd/cluster_builder_rd.h index 0b20a5d7ee..3ca7af70ca 100644 --- a/servers/rendering/renderer_rd/cluster_builder_rd.h +++ b/servers/rendering/renderer_rd/cluster_builder_rd.h @@ -43,13 +43,13 @@ class ClusterBuilderSharedDataRD { RID sphere_vertex_array; RID sphere_index_buffer; RID sphere_index_array; - float sphere_overfit = 0.0; //because an icosphere is not a perfect sphere, we need to enlarge it to cover the sphere area + float sphere_overfit = 0.0; // Because an icosphere is not a perfect sphere, we need to enlarge it to cover the sphere area. RID cone_vertex_buffer; RID cone_vertex_array; RID cone_index_buffer; RID cone_index_array; - float cone_overfit = 0.0; //because an cone mesh is not a perfect sphere, we need to enlarge it to cover the actual cone area + float cone_overfit = 0.0; // Because an cone mesh is not a perfect cone, we need to enlarge it to cover the actual cone area. RID box_vertex_buffer; RID box_vertex_array; @@ -73,6 +73,7 @@ class ClusterBuilderSharedDataRD { ClusterRenderShaderRD cluster_render_shader; RID shader_version; RID shader; + enum PipelineVersion { PIPELINE_NORMAL, PIPELINE_MSAA, @@ -85,10 +86,11 @@ class ClusterBuilderSharedDataRD { struct ClusterStore { struct PushConstant { uint32_t cluster_render_data_size; // how much data for a single cluster takes - uint32_t max_render_element_count_div_32; //divided by 32 + uint32_t max_render_element_count_div_32; // divided by 32 uint32_t cluster_screen_size[2]; - uint32_t render_element_count_div_32; //divided by 32 - uint32_t max_cluster_element_count_div_32; //divided by 32 + uint32_t render_element_count_div_32; // divided by 32 + uint32_t max_cluster_element_count_div_32; // divided by 32 + uint32_t pad1; uint32_t pad2; }; @@ -111,6 +113,7 @@ class ClusterBuilderSharedDataRD { uint32_t orthogonal; uint32_t max_cluster_element_count_div_32; + uint32_t pad1; uint32_t pad2; }; @@ -128,6 +131,8 @@ public: class ClusterBuilderRD { public: + static constexpr float WIDE_SPOT_ANGLE_THRESHOLD_DEG = 60.0f; + enum LightType { LIGHT_TYPE_OMNI, LIGHT_TYPE_SPOT @@ -144,21 +149,20 @@ public: ELEMENT_TYPE_DECAL, ELEMENT_TYPE_REFLECTION_PROBE, ELEMENT_TYPE_MAX, - }; private: ClusterBuilderSharedDataRD *shared = nullptr; struct RenderElementData { - uint32_t type; //0-4 + uint32_t type; // 0-4 uint32_t touches_near; uint32_t touches_far; uint32_t original_index; - float transform_inv[12]; //transposed transform for less space + float transform_inv[12]; // Transposed transform for less space. float scale[3]; - uint32_t pad; - }; + uint32_t has_wide_spot_angle; + }; // Keep aligned to 32 bytes. uint32_t cluster_count_by_type[ELEMENT_TYPE_MAX] = {}; uint32_t max_elements_by_type = 0; @@ -172,7 +176,7 @@ private: Projection projection; float z_far = 0; float z_near = 0; - bool orthogonal = false; + bool camera_orthogonal = false; enum Divisor { DIVISOR_1, @@ -188,26 +192,27 @@ private: Size2i cluster_screen_size; RID framebuffer; - RID cluster_render_buffer; //used for creating - RID cluster_buffer; //used for rendering - RID element_buffer; //used for storing, to hint element touches far plane or near plane + RID cluster_render_buffer; // Used for creating. + RID cluster_buffer; // Used for rendering. + RID element_buffer; // Used for storing, to hint element touches far plane or near plane. uint32_t cluster_render_buffer_size = 0; uint32_t cluster_buffer_size = 0; RID cluster_render_uniform_set; RID cluster_store_uniform_set; - //persistent data + // Persistent data. void _clear(); struct StateUniform { float projection[16]; float inv_z_far; - uint32_t screen_to_clusters_shift; // shift to obtain coordinates in block indices - uint32_t cluster_screen_width; // - uint32_t cluster_data_size; // how much data for a single cluster takes + uint32_t screen_to_clusters_shift; // Shift to obtain coordinates in block indices. + uint32_t cluster_screen_width; + uint32_t cluster_data_size; // How much data is needed for a single cluster. uint32_t cluster_depth_offset; + uint32_t pad0; uint32_t pad1; uint32_t pad2; @@ -224,10 +229,10 @@ public: _FORCE_INLINE_ void add_light(LightType p_type, const Transform3D &p_transform, float p_radius, float p_spot_aperture) { if (p_type == LIGHT_TYPE_OMNI && cluster_count_by_type[ELEMENT_TYPE_OMNI_LIGHT] == max_elements_by_type) { - return; //max number elements reached + return; // Max number elements reached. } if (p_type == LIGHT_TYPE_SPOT && cluster_count_by_type[ELEMENT_TYPE_SPOT_LIGHT] == max_elements_by_type) { - return; //max number elements reached + return; // Max number elements reached. } RenderElementData &e = render_elements[render_element_count]; @@ -242,15 +247,14 @@ public: radius *= p_radius; if (p_type == LIGHT_TYPE_OMNI) { - radius *= shared->sphere_overfit; // overfit icosphere + radius *= shared->sphere_overfit; // Overfit icosphere. - //omni float depth = -xform.origin.z; - if (orthogonal) { + if (camera_orthogonal) { e.touches_near = (depth - radius) < z_near; } else { - //contains camera inside light - float radius2 = radius * shared->sphere_overfit; // overfit again for outer size (camera may be outside actual sphere but behind an icosphere vertex) + // Contains camera inside light. + float radius2 = radius * shared->sphere_overfit; // Overfit again for outer size (camera may be outside actual sphere but behind an icosphere vertex) e.touches_near = xform.origin.length_squared() < radius2 * radius2; } @@ -265,12 +269,11 @@ public: cluster_count_by_type[ELEMENT_TYPE_OMNI_LIGHT]++; - } else { - //spot - radius *= shared->cone_overfit; // overfit icosphere + } else /*LIGHT_TYPE_SPOT */ { + radius *= shared->cone_overfit; // Overfit icosphere real_t len = Math::tan(Math::deg_to_rad(p_spot_aperture)) * radius; - //approximate, probably better to use a cone support function + // Approximate, probably better to use a cone support function. float max_d = -1e20; float min_d = 1e20; #define CONE_MINMAX(m_x, m_y) \ @@ -285,14 +288,13 @@ public: CONE_MINMAX(-1, -1); CONE_MINMAX(1, -1); - if (orthogonal) { + if (camera_orthogonal) { e.touches_near = min_d < z_near; } else { - //contains camera inside light Plane base_plane(-xform.basis.get_column(Vector3::AXIS_Z), xform.origin); float dist = base_plane.distance_to(Vector3()); if (dist >= 0 && dist < radius) { - //inside, check angle + // Contains camera inside light, check angle. float angle = Math::rad_to_deg(Math::acos((-xform.origin.normalized()).dot(-xform.basis.get_column(Vector3::AXIS_Z)))); e.touches_near = angle < p_spot_aperture * 1.05; //overfit aperture a little due to cone overfit } else { @@ -302,12 +304,23 @@ public: e.touches_far = max_d > z_far; - e.scale[0] = len * shared->cone_overfit; - e.scale[1] = len * shared->cone_overfit; - e.scale[2] = radius; + // If the spot angle is above the threshold, use a sphere instead of a cone for building the clusters + // since the cone gets too flat/large (spot angle close to 90 degrees) or + // can't even cover the affected area of the light (spot angle above 90 degrees). + if (p_spot_aperture > WIDE_SPOT_ANGLE_THRESHOLD_DEG) { + e.scale[0] = radius; + e.scale[1] = radius; + e.scale[2] = radius; + e.has_wide_spot_angle = true; + } else { + e.scale[0] = len * shared->cone_overfit; + e.scale[1] = len * shared->cone_overfit; + e.scale[2] = radius; + e.has_wide_spot_angle = false; + } e.type = ELEMENT_TYPE_SPOT_LIGHT; - e.original_index = cluster_count_by_type[ELEMENT_TYPE_SPOT_LIGHT]; //use omni since they share index + e.original_index = cluster_count_by_type[ELEMENT_TYPE_SPOT_LIGHT]; // Use omni light since they share index. RendererRD::MaterialStorage::store_transform_transposed_3x4(xform, e.transform_inv); @@ -317,19 +330,19 @@ public: render_element_count++; } - _FORCE_INLINE_ void add_box(BoxType p_box_type, const Transform3D &p_transform, const Vector3 &p_half_extents) { + _FORCE_INLINE_ void add_box(BoxType p_box_type, const Transform3D &p_transform, const Vector3 &p_half_size) { if (p_box_type == BOX_TYPE_DECAL && cluster_count_by_type[ELEMENT_TYPE_DECAL] == max_elements_by_type) { - return; //max number elements reached + return; // Max number elements reached. } if (p_box_type == BOX_TYPE_REFLECTION_PROBE && cluster_count_by_type[ELEMENT_TYPE_REFLECTION_PROBE] == max_elements_by_type) { - return; //max number elements reached + return; // Max number elements reached. } RenderElementData &e = render_elements[render_element_count]; Transform3D xform = view_xform * p_transform; - //extract scale and scale the matrix by it, makes things simpler - Vector3 scale = p_half_extents; + // Extract scale and scale the matrix by it, makes things simpler. + Vector3 scale = p_half_size; for (uint32_t i = 0; i < 3; i++) { float s = xform.basis.rows[i].length(); scale[i] *= s; @@ -339,10 +352,10 @@ public: float box_depth = Math::abs(xform.basis.xform_inv(Vector3(0, 0, -1)).dot(scale)); float depth = -xform.origin.z; - if (orthogonal) { + if (camera_orthogonal) { e.touches_near = depth - box_depth < z_near; } else { - //contains camera inside box + // Contains camera inside box. Vector3 inside = xform.xform_inv(Vector3(0, 0, 0)).abs(); e.touches_near = inside.x < scale.x && inside.y < scale.y && inside.z < scale.z; } diff --git a/servers/rendering/renderer_rd/environment/fog.cpp b/servers/rendering/renderer_rd/environment/fog.cpp index 2787693aeb..4253ea8610 100644 --- a/servers/rendering/renderer_rd/environment/fog.cpp +++ b/servers/rendering/renderer_rd/environment/fog.cpp @@ -82,11 +82,11 @@ void Fog::fog_volume_set_shape(RID p_fog_volume, RS::FogVolumeShape p_shape) { fog_volume->dependency.changed_notify(Dependency::DEPENDENCY_CHANGED_AABB); } -void Fog::fog_volume_set_extents(RID p_fog_volume, const Vector3 &p_extents) { +void Fog::fog_volume_set_size(RID p_fog_volume, const Vector3 &p_size) { FogVolume *fog_volume = fog_volume_owner.get_or_null(p_fog_volume); ERR_FAIL_COND(!fog_volume); - fog_volume->extents = p_extents; + fog_volume->size = p_size; fog_volume->dependency.changed_notify(Dependency::DEPENDENCY_CHANGED_AABB); } @@ -120,8 +120,8 @@ AABB Fog::fog_volume_get_aabb(RID p_fog_volume) const { case RS::FOG_VOLUME_SHAPE_CYLINDER: case RS::FOG_VOLUME_SHAPE_BOX: { AABB aabb; - aabb.position = -fog_volume->extents; - aabb.size = fog_volume->extents * 2; + aabb.position = -fog_volume->size / 2; + aabb.size = fog_volume->size; return aabb; } default: { @@ -131,10 +131,10 @@ AABB Fog::fog_volume_get_aabb(RID p_fog_volume) const { } } -Vector3 Fog::fog_volume_get_extents(RID p_fog_volume) const { +Vector3 Fog::fog_volume_get_size(RID p_fog_volume) const { const FogVolume *fog_volume = fog_volume_owner.get_or_null(p_fog_volume); ERR_FAIL_COND_V(!fog_volume, Vector3()); - return fog_volume->extents; + return fog_volume->size; } //////////////////////////////////////////////////////////////////////////////// @@ -210,7 +210,7 @@ void Fog::init_fog_shader(uint32_t p_max_directional_lights, int p_roughness_lay actions.renames["WORLD_POSITION"] = "world.xyz"; actions.renames["OBJECT_POSITION"] = "params.position"; actions.renames["UVW"] = "uvw"; - actions.renames["EXTENTS"] = "params.extents"; + actions.renames["SIZE"] = "params.size"; actions.renames["ALBEDO"] = "albedo"; actions.renames["DENSITY"] = "density"; actions.renames["EMISSION"] = "emission"; @@ -643,7 +643,7 @@ void Fog::volumetric_fog_update(const VolumetricFogSettings &p_settings, const P Vector3 position = fog_volume_instance->transform.get_origin(); RS::FogVolumeShape volume_type = RendererRD::Fog::get_singleton()->fog_volume_get_shape(fog_volume); - Vector3 extents = RendererRD::Fog::get_singleton()->fog_volume_get_extents(fog_volume); + Vector3 extents = RendererRD::Fog::get_singleton()->fog_volume_get_size(fog_volume) / 2; if (volume_type != RS::FOG_VOLUME_SHAPE_WORLD) { // Local fog volume. @@ -683,9 +683,9 @@ void Fog::volumetric_fog_update(const VolumetricFogSettings &p_settings, const P push_constant.position[0] = position.x; push_constant.position[1] = position.y; push_constant.position[2] = position.z; - push_constant.extents[0] = extents.x; - push_constant.extents[1] = extents.y; - push_constant.extents[2] = extents.z; + push_constant.size[0] = extents.x * 2; + push_constant.size[1] = extents.y * 2; + push_constant.size[2] = extents.z * 2; push_constant.corner[0] = min.x; push_constant.corner[1] = min.y; push_constant.corner[2] = min.z; diff --git a/servers/rendering/renderer_rd/environment/fog.h b/servers/rendering/renderer_rd/environment/fog.h index eb0a2fc7b5..0b6bcc29fb 100644 --- a/servers/rendering/renderer_rd/environment/fog.h +++ b/servers/rendering/renderer_rd/environment/fog.h @@ -53,7 +53,7 @@ private: struct FogVolume { RID material; - Vector3 extents = Vector3(1, 1, 1); + Vector3 size = Vector3(2, 2, 2); RS::FogVolumeShape shape = RS::FOG_VOLUME_SHAPE_BOX; @@ -83,7 +83,7 @@ private: float position[3]; float pad; - float extents[3]; + float size[3]; float pad2; int32_t corner[3]; @@ -239,12 +239,12 @@ public: Dependency *fog_volume_get_dependency(RID p_fog_volume) const; virtual void fog_volume_set_shape(RID p_fog_volume, RS::FogVolumeShape p_shape) override; - virtual void fog_volume_set_extents(RID p_fog_volume, const Vector3 &p_extents) override; + virtual void fog_volume_set_size(RID p_fog_volume, const Vector3 &p_size) override; virtual void fog_volume_set_material(RID p_fog_volume, RID p_material) override; virtual RS::FogVolumeShape fog_volume_get_shape(RID p_fog_volume) const override; RID fog_volume_get_material(RID p_fog_volume) const; virtual AABB fog_volume_get_aabb(RID p_fog_volume) const override; - Vector3 fog_volume_get_extents(RID p_fog_volume) const; + Vector3 fog_volume_get_size(RID p_fog_volume) const; /* FOG VOLUMES INSTANCE */ diff --git a/servers/rendering/renderer_rd/environment/gi.cpp b/servers/rendering/renderer_rd/environment/gi.cpp index a52716cd78..08133bf8d6 100644 --- a/servers/rendering/renderer_rd/environment/gi.cpp +++ b/servers/rendering/renderer_rd/environment/gi.cpp @@ -540,9 +540,7 @@ void GI::SDFGI::create(RID p_env, const Vector3 &p_world_position, uint32_t p_re occlusion_texture = RD::get_singleton()->texture_create_shared(tv, occlusion_data); } - for (uint32_t i = 0; i < cascades.size(); i++) { - SDFGI::Cascade &cascade = cascades[i]; - + for (SDFGI::Cascade &cascade : cascades) { /* 3D Textures */ cascade.sdf_tex = RD::get_singleton()->texture_create(tf_sdf, RD::TextureView()); @@ -743,9 +741,7 @@ void GI::SDFGI::create(RID p_env, const Vector3 &p_world_position, uint32_t p_re } //direct light - for (uint32_t i = 0; i < cascades.size(); i++) { - SDFGI::Cascade &cascade = cascades[i]; - + for (SDFGI::Cascade &cascade : cascades) { Vector<RD::Uniform> uniforms; { RD::Uniform u; @@ -1134,8 +1130,7 @@ void GI::SDFGI::free_data() { } GI::SDFGI::~SDFGI() { - for (uint32_t i = 0; i < cascades.size(); i++) { - const SDFGI::Cascade &c = cascades[i]; + for (const SDFGI::Cascade &c : cascades) { RD::get_singleton()->free(c.light_data); RD::get_singleton()->free(c.light_aniso_0_tex); RD::get_singleton()->free(c.light_aniso_1_tex); @@ -1198,8 +1193,7 @@ void GI::SDFGI::update(RID p_env, const Vector3 &p_world_position) { int32_t drag_margin = (cascade_size / SDFGI::PROBE_DIVISOR) / 2; - for (uint32_t i = 0; i < cascades.size(); i++) { - SDFGI::Cascade &cascade = cascades[i]; + for (SDFGI::Cascade &cascade : cascades) { cascade.dirty_regions = Vector3i(); Vector3 probe_half_size = Vector3(1, 1, 1) * cascade.cell_size * float(cascade_size / SDFGI::PROBE_DIVISOR) * 0.5; diff --git a/servers/rendering/renderer_rd/environment/sky.cpp b/servers/rendering/renderer_rd/environment/sky.cpp index 7e02f98ce9..7fff349b3c 100644 --- a/servers/rendering/renderer_rd/environment/sky.cpp +++ b/servers/rendering/renderer_rd/environment/sky.cpp @@ -347,7 +347,10 @@ void SkyRD::ReflectionData::update_reflection_data(int p_size, int p_mipmaps, bo tf.texture_type = RD::TEXTURE_TYPE_CUBE; tf.array_layers = 6; tf.mipmaps = p_low_quality ? 7 : mipmaps - 1; - tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_STORAGE_BIT | RD::TEXTURE_USAGE_COLOR_ATTACHMENT_BIT; + tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_COLOR_ATTACHMENT_BIT; + if (RendererSceneRenderRD::get_singleton()->_render_buffers_can_be_storage()) { + tf.usage_bits |= RD::TEXTURE_USAGE_STORAGE_BIT; + } downsampled_radiance_cubemap = RD::get_singleton()->texture_create(tf, RD::TextureView()); RD::get_singleton()->set_resource_name(downsampled_radiance_cubemap, "downsampled radiance cubemap"); @@ -571,7 +574,7 @@ RID SkyRD::Sky::get_textures(SkyTextureSetVersion p_version, RID p_default_shade u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; u.binding = 1; // half res if (p_version >= SKY_TEXTURE_SET_CUBEMAP) { - if (reflection.layers[0].views[1].is_valid() && p_version != SKY_TEXTURE_SET_CUBEMAP_HALF_RES) { + if (reflection.layers.size() && reflection.layers[0].views.size() >= 2 && reflection.layers[0].views[1].is_valid() && p_version != SKY_TEXTURE_SET_CUBEMAP_HALF_RES) { u.append_id(reflection.layers[0].views[1]); } else { u.append_id(texture_storage->texture_rd_get_default(RendererRD::TextureStorage::DEFAULT_RD_TEXTURE_CUBEMAP_BLACK)); @@ -591,7 +594,7 @@ RID SkyRD::Sky::get_textures(SkyTextureSetVersion p_version, RID p_default_shade u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; u.binding = 2; // quarter res if (p_version >= SKY_TEXTURE_SET_CUBEMAP) { - if (reflection.layers[0].views[2].is_valid() && p_version != SKY_TEXTURE_SET_CUBEMAP_QUARTER_RES) { + if (reflection.layers.size() && reflection.layers[0].views.size() >= 3 && reflection.layers[0].views[2].is_valid() && p_version != SKY_TEXTURE_SET_CUBEMAP_QUARTER_RES) { u.append_id(reflection.layers[0].views[2]); } else { u.append_id(texture_storage->texture_rd_get_default(RendererRD::TextureStorage::DEFAULT_RD_TEXTURE_CUBEMAP_BLACK)); @@ -1320,7 +1323,7 @@ void SkyRD::update_radiance_buffers(Ref<RenderSceneBuffersRD> p_render_buffers, // Note, we ignore environment_get_sky_orientation here as this is applied when we do our lookup in our scene shader. - if (shader_data->uses_quarter_res) { + if (shader_data->uses_quarter_res && roughness_layers >= 3) { RD::get_singleton()->draw_command_begin_label("Render Sky to Quarter Res Cubemap"); PipelineCacheRD *pipeline = &shader_data->pipelines[SKY_VERSION_CUBEMAP_QUARTER_RES]; @@ -1337,9 +1340,11 @@ void SkyRD::update_radiance_buffers(Ref<RenderSceneBuffersRD> p_render_buffers, RD::get_singleton()->draw_list_end(); } RD::get_singleton()->draw_command_end_label(); + } else if (shader_data->uses_quarter_res && roughness_layers < 3) { + ERR_PRINT_ED("Cannot use quarter res buffer in sky shader when roughness layers is less than 3. Please increase rendering/reflections/sky_reflections/roughness_layers."); } - if (shader_data->uses_half_res) { + if (shader_data->uses_half_res && roughness_layers >= 2) { RD::get_singleton()->draw_command_begin_label("Render Sky to Half Res Cubemap"); PipelineCacheRD *pipeline = &shader_data->pipelines[SKY_VERSION_CUBEMAP_HALF_RES]; @@ -1356,6 +1361,8 @@ void SkyRD::update_radiance_buffers(Ref<RenderSceneBuffersRD> p_render_buffers, RD::get_singleton()->draw_list_end(); } RD::get_singleton()->draw_command_end_label(); + } else if (shader_data->uses_half_res && roughness_layers < 2) { + ERR_PRINT_ED("Cannot use half res buffer in sky shader when roughness layers is less than 2. Please increase rendering/reflections/sky_reflections/roughness_layers."); } RD::DrawListID cubemap_draw_list; @@ -1627,7 +1634,10 @@ void SkyRD::update_dirty_skys() { tf.mipmaps = mipmaps; tf.width = w; tf.height = h; - tf.usage_bits = RD::TEXTURE_USAGE_COLOR_ATTACHMENT_BIT | RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_STORAGE_BIT; + tf.usage_bits = RD::TEXTURE_USAGE_COLOR_ATTACHMENT_BIT | RD::TEXTURE_USAGE_SAMPLING_BIT; + if (RendererSceneRenderRD::get_singleton()->_render_buffers_can_be_storage()) { + tf.usage_bits |= RD::TEXTURE_USAGE_STORAGE_BIT; + } sky->radiance = RD::get_singleton()->texture_create(tf, RD::TextureView()); @@ -1642,7 +1652,10 @@ void SkyRD::update_dirty_skys() { tf.mipmaps = MIN(mipmaps, layers); tf.width = w; tf.height = h; - tf.usage_bits = RD::TEXTURE_USAGE_COLOR_ATTACHMENT_BIT | RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_STORAGE_BIT; + tf.usage_bits = RD::TEXTURE_USAGE_COLOR_ATTACHMENT_BIT | RD::TEXTURE_USAGE_SAMPLING_BIT; + if (RendererSceneRenderRD::get_singleton()->_render_buffers_can_be_storage()) { + tf.usage_bits |= RD::TEXTURE_USAGE_STORAGE_BIT; + } sky->radiance = RD::get_singleton()->texture_create(tf, RD::TextureView()); diff --git a/servers/rendering/renderer_rd/forward_clustered/render_forward_clustered.cpp b/servers/rendering/renderer_rd/forward_clustered/render_forward_clustered.cpp index 93c741fd34..59e1f559c7 100644 --- a/servers/rendering/renderer_rd/forward_clustered/render_forward_clustered.cpp +++ b/servers/rendering/renderer_rd/forward_clustered/render_forward_clustered.cpp @@ -667,7 +667,7 @@ void RenderForwardClustered::_setup_environment(const RenderDataRD *p_render_dat if (get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_UNSHADED) { scene_state.ubo.ss_effects_flags = 0; - } else if (is_environment(p_render_data->environment)) { + } else if (p_render_data->reflection_probe.is_null() && is_environment(p_render_data->environment)) { scene_state.ubo.ssao_ao_affect = environment_get_ssao_ao_channel_affect(p_render_data->environment); scene_state.ubo.ssao_light_affect = environment_get_ssao_direct_light_affect(p_render_data->environment); uint32_t ss_flags = 0; @@ -1232,9 +1232,9 @@ void RenderForwardClustered::_update_volumetric_fog(Ref<RenderSceneBuffersRD> p_ /* Lighting */ -void RenderForwardClustered::setup_added_reflection_probe(const Transform3D &p_transform, const Vector3 &p_half_extents) { +void RenderForwardClustered::setup_added_reflection_probe(const Transform3D &p_transform, const Vector3 &p_half_size) { if (current_cluster_builder != nullptr) { - current_cluster_builder->add_box(ClusterBuilderRD::BOX_TYPE_REFLECTION_PROBE, p_transform, p_half_extents); + current_cluster_builder->add_box(ClusterBuilderRD::BOX_TYPE_REFLECTION_PROBE, p_transform, p_half_size); } } @@ -1244,9 +1244,9 @@ void RenderForwardClustered::setup_added_light(const RS::LightType p_type, const } } -void RenderForwardClustered::setup_added_decal(const Transform3D &p_transform, const Vector3 &p_half_extents) { +void RenderForwardClustered::setup_added_decal(const Transform3D &p_transform, const Vector3 &p_half_size) { if (current_cluster_builder != nullptr) { - current_cluster_builder->add_box(ClusterBuilderRD::BOX_TYPE_DECAL, p_transform, p_half_extents); + current_cluster_builder->add_box(ClusterBuilderRD::BOX_TYPE_DECAL, p_transform, p_half_size); } } @@ -1365,8 +1365,8 @@ void RenderForwardClustered::_pre_opaque_render(RenderDataRD *p_render_data, boo } //cube shadows are rendered in their own way - for (uint32_t i = 0; i < p_render_data->cube_shadows.size(); i++) { - _render_shadow_pass(p_render_data->render_shadows[p_render_data->cube_shadows[i]].light, p_render_data->shadow_atlas, p_render_data->render_shadows[p_render_data->cube_shadows[i]].pass, p_render_data->render_shadows[p_render_data->cube_shadows[i]].instances, camera_plane, lod_distance_multiplier, p_render_data->scene_data->screen_mesh_lod_threshold, true, true, true, p_render_data->render_info); + for (const int &index : p_render_data->cube_shadows) { + _render_shadow_pass(p_render_data->render_shadows[index].light, p_render_data->shadow_atlas, p_render_data->render_shadows[index].pass, p_render_data->render_shadows[index].instances, camera_plane, lod_distance_multiplier, p_render_data->scene_data->screen_mesh_lod_threshold, true, true, true, p_render_data->render_info); } if (p_render_data->directional_shadows.size()) { @@ -1480,7 +1480,7 @@ void RenderForwardClustered::_pre_opaque_render(RenderDataRD *p_render_data, boo uint32_t directional_light_count = 0; uint32_t positional_light_count = 0; light_storage->update_light_buffers(p_render_data, *p_render_data->lights, p_render_data->scene_data->cam_transform, p_render_data->shadow_atlas, using_shadows, directional_light_count, positional_light_count, p_render_data->directional_light_soft_shadows); - texture_storage->update_decal_buffer(*p_render_data->decals, p_render_data->scene_data->cam_transform.affine_inverse()); + texture_storage->update_decal_buffer(*p_render_data->decals, p_render_data->scene_data->cam_transform); p_render_data->directional_light_count = directional_light_count; @@ -1550,16 +1550,18 @@ void RenderForwardClustered::_process_sss(Ref<RenderSceneBuffersRD> p_render_buf void RenderForwardClustered::_render_scene(RenderDataRD *p_render_data, const Color &p_default_bg_color) { RendererRD::LightStorage *light_storage = RendererRD::LightStorage::get_singleton(); - Ref<RenderSceneBuffersRD> rb; + ERR_FAIL_NULL(p_render_data); + + Ref<RenderSceneBuffersRD> rb = p_render_data->render_buffers; + ERR_FAIL_COND(rb.is_null()); Ref<RenderBufferDataForwardClustered> rb_data; - if (p_render_data && p_render_data->render_buffers.is_valid()) { - rb = p_render_data->render_buffers; - if (rb->has_custom_data(RB_SCOPE_FORWARD_CLUSTERED)) { - // Our forward clustered custom data buffer will only be available when we're rendering our normal view. - // This will not be available when rendering reflection probes. - rb_data = rb->get_custom_data(RB_SCOPE_FORWARD_CLUSTERED); - } + if (rb->has_custom_data(RB_SCOPE_FORWARD_CLUSTERED)) { + // Our forward clustered custom data buffer will only be available when we're rendering our normal view. + // This will not be available when rendering reflection probes. + rb_data = rb->get_custom_data(RB_SCOPE_FORWARD_CLUSTERED); } + bool is_reflection_probe = p_render_data->reflection_probe.is_valid(); + static const int texture_multisamples[RS::VIEWPORT_MSAA_MAX] = { 1, 2, 4, 8 }; //first of all, make a new render pass @@ -1588,18 +1590,16 @@ void RenderForwardClustered::_render_scene(RenderDataRD *p_render_data, const Co p_render_data->voxel_gi_count = 0; - if (rb.is_valid()) { - if (rb->has_custom_data(RB_SCOPE_SDFGI)) { - Ref<RendererRD::GI::SDFGI> sdfgi = rb->get_custom_data(RB_SCOPE_SDFGI); - if (sdfgi.is_valid()) { - sdfgi->update_cascades(); - sdfgi->pre_process_gi(p_render_data->scene_data->cam_transform, p_render_data); - sdfgi->update_light(); - } + if (rb->has_custom_data(RB_SCOPE_SDFGI)) { + Ref<RendererRD::GI::SDFGI> sdfgi = rb->get_custom_data(RB_SCOPE_SDFGI); + if (sdfgi.is_valid()) { + sdfgi->update_cascades(); + sdfgi->pre_process_gi(p_render_data->scene_data->cam_transform, p_render_data); + sdfgi->update_light(); } - - gi.setup_voxel_gi_instances(p_render_data, p_render_data->render_buffers, p_render_data->scene_data->cam_transform, *p_render_data->voxel_gi_instances, p_render_data->voxel_gi_count); } + + gi.setup_voxel_gi_instances(p_render_data, p_render_data->render_buffers, p_render_data->scene_data->cam_transform, *p_render_data->voxel_gi_instances, p_render_data->voxel_gi_count); } else { ERR_PRINT("No render buffer nor reflection atlas, bug"); //should never happen, will crash current_cluster_builder = nullptr; @@ -1618,7 +1618,7 @@ void RenderForwardClustered::_render_scene(RenderDataRD *p_render_data, const Co // check if we need motion vectors if (get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_MOTION_VECTORS) { p_render_data->scene_data->calculate_motion_vectors = true; - } else if (rb.is_valid() && rb->get_use_taa()) { + } else if (!is_reflection_probe && rb->get_use_taa()) { p_render_data->scene_data->calculate_motion_vectors = true; } else { p_render_data->scene_data->calculate_motion_vectors = false; @@ -1641,9 +1641,9 @@ void RenderForwardClustered::_render_scene(RenderDataRD *p_render_data, const Co bool using_sdfgi = false; bool using_voxelgi = false; bool reverse_cull = p_render_data->scene_data->cam_transform.basis.determinant() < 0; - bool using_ssil = p_render_data->environment.is_valid() && environment_get_ssil_enabled(p_render_data->environment); + bool using_ssil = !is_reflection_probe && p_render_data->environment.is_valid() && environment_get_ssil_enabled(p_render_data->environment); - if (p_render_data->reflection_probe.is_valid()) { + if (is_reflection_probe) { uint32_t resolution = light_storage->reflection_probe_instance_get_resolution(p_render_data->reflection_probe); screen_size.x = resolution; screen_size.y = resolution; @@ -1657,7 +1657,7 @@ void RenderForwardClustered::_render_scene(RenderDataRD *p_render_data, const Co } reverse_cull = true; // for some reason our views are inverted - } else if (rb.is_valid()) { + } else { screen_size = rb->get_internal_size(); if (rb->get_use_taa() || get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_MOTION_VECTORS) { @@ -1685,8 +1685,6 @@ void RenderForwardClustered::_render_scene(RenderDataRD *p_render_data, const Co color_framebuffer = rb_data->get_color_pass_fb(color_pass_flags); color_only_framebuffer = rb_data->get_color_only_fb(); - } else { - ERR_FAIL(); //bug? } p_render_data->scene_data->emissive_exposure_normalization = -1.0; @@ -1695,7 +1693,7 @@ void RenderForwardClustered::_render_scene(RenderDataRD *p_render_data, const Co _setup_lightmaps(p_render_data, *p_render_data->lightmaps, p_render_data->scene_data->cam_transform); _setup_voxelgis(*p_render_data->voxel_gi_instances); - _setup_environment(p_render_data, p_render_data->reflection_probe.is_valid(), screen_size, !p_render_data->reflection_probe.is_valid(), p_default_bg_color, false); + _setup_environment(p_render_data, is_reflection_probe, screen_size, !is_reflection_probe, p_default_bg_color, false); _update_render_base_uniform_set(); //may have changed due to the above (light buffer enlarged, as an example) @@ -1707,7 +1705,7 @@ void RenderForwardClustered::_render_scene(RenderDataRD *p_render_data, const Co RD::get_singleton()->draw_command_end_label(); - if (rb.is_valid() && !p_render_data->reflection_probe.is_valid()) { + if (!is_reflection_probe) { if (using_voxelgi) { depth_pass_mode = PASS_MODE_DEPTH_NORMAL_ROUGHNESS_VOXEL_GI; } else if (p_render_data->environment.is_valid()) { @@ -1739,7 +1737,7 @@ void RenderForwardClustered::_render_scene(RenderDataRD *p_render_data, const Co } } - bool using_sss = rb_data.is_valid() && scene_state.used_sss && ss_effects->sss_get_quality() != RS::SUB_SURFACE_SCATTERING_QUALITY_DISABLED; + bool using_sss = rb_data.is_valid() && !is_reflection_probe && scene_state.used_sss && ss_effects->sss_get_quality() != RS::SUB_SURFACE_SCATTERING_QUALITY_DISABLED; if (using_sss && !using_separate_specular) { using_separate_specular = true; @@ -1772,7 +1770,7 @@ void RenderForwardClustered::_render_scene(RenderDataRD *p_render_data, const Co clear_color.r *= bg_energy_multiplier; clear_color.g *= bg_energy_multiplier; clear_color.b *= bg_energy_multiplier; - if ((rb.is_valid() && rb->has_custom_data(RB_SCOPE_FOG)) || environment_get_fog_enabled(p_render_data->environment)) { + if ((rb->has_custom_data(RB_SCOPE_FOG)) || environment_get_fog_enabled(p_render_data->environment)) { draw_sky_fog_only = true; RendererRD::MaterialStorage::get_singleton()->material_set_param(sky.sky_scene_state.fog_material, "clear_color", Variant(clear_color.srgb_to_linear())); } @@ -1782,7 +1780,7 @@ void RenderForwardClustered::_render_scene(RenderDataRD *p_render_data, const Co clear_color.r *= bg_energy_multiplier; clear_color.g *= bg_energy_multiplier; clear_color.b *= bg_energy_multiplier; - if ((rb.is_valid() && rb->has_custom_data(RB_SCOPE_FOG)) || environment_get_fog_enabled(p_render_data->environment)) { + if ((rb->has_custom_data(RB_SCOPE_FOG)) || environment_get_fog_enabled(p_render_data->environment)) { draw_sky_fog_only = true; RendererRD::MaterialStorage::get_singleton()->material_set_param(sky.sky_scene_state.fog_material, "clear_color", Variant(clear_color.srgb_to_linear())); } @@ -1791,7 +1789,7 @@ void RenderForwardClustered::_render_scene(RenderDataRD *p_render_data, const Co draw_sky = true; } break; case RS::ENV_BG_CANVAS: { - if (rb.is_valid()) { + if (!is_reflection_probe) { RID texture = RendererRD::TextureStorage::get_singleton()->render_target_get_rd_texture(rb->get_render_target()); copy_effects->copy_to_fb_rect(texture, color_only_framebuffer, Rect2i(), false, false, false, false, RID(), false, false, true); } @@ -1812,7 +1810,7 @@ void RenderForwardClustered::_render_scene(RenderDataRD *p_render_data, const Co RD::get_singleton()->draw_command_begin_label("Setup Sky"); // Setup our sky render information for this frame/viewport - if (p_render_data->reflection_probe.is_valid()) { + if (is_reflection_probe) { Vector3 eye_offset; Projection correction; correction.set_depth_correction(true); @@ -1849,7 +1847,7 @@ void RenderForwardClustered::_render_scene(RenderDataRD *p_render_data, const Co bool debug_sdfgi_probes = get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_SDFGI_PROBES; bool depth_pre_pass = bool(GLOBAL_GET("rendering/driver/depth_prepass/enable")) && depth_framebuffer.is_valid(); - bool using_ssao = depth_pre_pass && rb.is_valid() && p_render_data->environment.is_valid() && environment_get_ssao_enabled(p_render_data->environment); + bool using_ssao = depth_pre_pass && !is_reflection_probe && p_render_data->environment.is_valid() && environment_get_ssao_enabled(p_render_data->environment); bool continue_depth = false; if (depth_pre_pass) { //depth pre pass @@ -1881,7 +1879,7 @@ void RenderForwardClustered::_render_scene(RenderDataRD *p_render_data, const Co _pre_resolve_render(p_render_data, using_sdfgi || using_voxelgi); } - if (rb.is_valid() && rb->get_msaa_3d() != RS::VIEWPORT_MSAA_DISABLED) { + if (rb->get_msaa_3d() != RS::VIEWPORT_MSAA_DISABLED) { RENDER_TIMESTAMP("Resolve Depth Pre-Pass (MSAA)"); RD::get_singleton()->draw_command_begin_label("Resolve Depth Pre-Pass (MSAA)"); if (depth_pass_mode == PASS_MODE_DEPTH_NORMAL_ROUGHNESS || depth_pass_mode == PASS_MODE_DEPTH_NORMAL_ROUGHNESS_VOXEL_GI) { @@ -1915,7 +1913,7 @@ void RenderForwardClustered::_render_scene(RenderDataRD *p_render_data, const Co p_render_data->scene_data->directional_light_count = p_render_data->directional_light_count; p_render_data->scene_data->opaque_prepass_threshold = 0.0f; - _setup_environment(p_render_data, p_render_data->reflection_probe.is_valid(), screen_size, !p_render_data->reflection_probe.is_valid(), p_default_bg_color, rb.is_valid()); + _setup_environment(p_render_data, is_reflection_probe, screen_size, !is_reflection_probe, p_default_bg_color, true); RENDER_TIMESTAMP("Render Opaque Pass"); @@ -1996,7 +1994,7 @@ void RenderForwardClustered::_render_scene(RenderDataRD *p_render_data, const Co RD::get_singleton()->draw_command_end_label(); } - if (rb.is_valid() && !can_continue_color && rb->get_msaa_3d() != RS::VIEWPORT_MSAA_DISABLED) { + if (!can_continue_color && rb->get_msaa_3d() != RS::VIEWPORT_MSAA_DISABLED) { // Handle views individual, might want to look at rewriting our resolve to do both layers in one pass. for (uint32_t v = 0; v < rb->get_view_count(); v++) { RD::get_singleton()->texture_resolve_multisample(rb_data->get_color_msaa(v), rb->get_internal_texture(v)); @@ -2008,7 +2006,7 @@ void RenderForwardClustered::_render_scene(RenderDataRD *p_render_data, const Co } } - if (rb.is_valid() && !can_continue_depth && rb->get_msaa_3d() != RS::VIEWPORT_MSAA_DISABLED) { + if (!can_continue_depth && rb->get_msaa_3d() != RS::VIEWPORT_MSAA_DISABLED) { for (uint32_t v = 0; v < rb->get_view_count(); v++) { resolve_effects->resolve_depth(rb_data->get_depth_msaa(v), rb->get_depth_texture(v), rb->get_internal_size(), texture_multisamples[rb->get_msaa_3d()]); } @@ -2054,7 +2052,7 @@ void RenderForwardClustered::_render_scene(RenderDataRD *p_render_data, const Co rp_uniform_set = _setup_render_pass_uniform_set(RENDER_LIST_ALPHA, p_render_data, radiance_texture, true); - _setup_environment(p_render_data, p_render_data->reflection_probe.is_valid(), screen_size, !p_render_data->reflection_probe.is_valid(), p_default_bg_color, false); + _setup_environment(p_render_data, is_reflection_probe, screen_size, !is_reflection_probe, p_default_bg_color, false); { uint32_t transparent_color_pass_flags = (color_pass_flags | COLOR_PASS_FLAG_TRANSPARENT) & ~(COLOR_PASS_FLAG_SEPARATE_SPECULAR); @@ -2415,8 +2413,7 @@ void RenderForwardClustered::_render_shadow_process() { void RenderForwardClustered::_render_shadow_end(uint32_t p_barrier) { RD::get_singleton()->draw_command_begin_label("Shadow Render"); - for (uint32_t i = 0; i < scene_state.shadow_passes.size(); i++) { - SceneState::ShadowPass &shadow_pass = scene_state.shadow_passes[i]; + for (SceneState::ShadowPass &shadow_pass : scene_state.shadow_passes) { RenderListParameters render_list_parameters(render_list[RENDER_LIST_SECONDARY].elements.ptr() + shadow_pass.element_from, render_list[RENDER_LIST_SECONDARY].element_info.ptr() + shadow_pass.element_from, shadow_pass.element_count, shadow_pass.flip_cull, shadow_pass.pass_mode, 0, true, false, shadow_pass.rp_uniform_set, false, Vector2(), shadow_pass.lod_distance_multiplier, shadow_pass.screen_mesh_lod_threshold, 1, shadow_pass.element_from, RD::BARRIER_MASK_NO_BARRIER); _render_list_with_threads(&render_list_parameters, shadow_pass.framebuffer, RD::INITIAL_ACTION_DROP, RD::FINAL_ACTION_DISCARD, shadow_pass.initial_depth_action, shadow_pass.final_depth_action, Vector<Color>(), 1.0, 0, shadow_pass.rect); } @@ -2618,8 +2615,8 @@ void RenderForwardClustered::_render_sdfgi(Ref<RenderSceneBuffersRD> p_render_bu render_list[RENDER_LIST_SECONDARY].sort_by_key(); _fill_instance_data(RENDER_LIST_SECONDARY); - Vector3 half_extents = p_bounds.size * 0.5; - Vector3 center = p_bounds.position + half_extents; + Vector3 half_size = p_bounds.size * 0.5; + Vector3 center = p_bounds.position + half_size; Vector<RID> sbs = { p_albedo_texture, @@ -2647,16 +2644,16 @@ void RenderForwardClustered::_render_sdfgi(Ref<RenderSceneBuffersRD> p_render_bu fb_size.x = p_size[right_axis]; fb_size.y = p_size[up_axis]; - scene_data.cam_transform.origin = center + axis * half_extents; + scene_data.cam_transform.origin = center + axis * half_size; scene_data.cam_transform.basis.set_column(0, right); scene_data.cam_transform.basis.set_column(1, up); scene_data.cam_transform.basis.set_column(2, axis); //print_line("pass: " + itos(i) + " xform " + scene_data.cam_transform); - float h_size = half_extents[right_axis]; - float v_size = half_extents[up_axis]; - float d_size = half_extents[i] * 2.0; + float h_size = half_size[right_axis]; + float v_size = half_size[up_axis]; + float d_size = half_size[i] * 2.0; scene_data.cam_projection.set_orthogonal(-h_size, h_size, -v_size, v_size, 0, d_size); //print_line("pass: " + itos(i) + " cam hsize: " + rtos(h_size) + " vsize: " + rtos(v_size) + " dsize " + rtos(d_size)); @@ -3376,9 +3373,7 @@ int RenderForwardClustered::sdfgi_get_pending_region_count(const Ref<RenderScene Ref<RendererRD::GI::SDFGI> sdfgi = rb->get_custom_data(RB_SCOPE_SDFGI); int dirty_count = 0; - for (uint32_t i = 0; i < sdfgi->cascades.size(); i++) { - const RendererRD::GI::SDFGI::Cascade &c = sdfgi->cascades[i]; - + for (const RendererRD::GI::SDFGI::Cascade &c : sdfgi->cascades) { if (c.dirty_regions == RendererRD::GI::SDFGI::Cascade::DIRTY_ALL) { dirty_count++; } else { @@ -3771,6 +3766,7 @@ void RenderForwardClustered::_geometry_instance_dependency_changed(Dependency::D case Dependency::DEPENDENCY_CHANGED_MULTIMESH: case Dependency::DEPENDENCY_CHANGED_SKELETON_DATA: { static_cast<RenderGeometryInstance *>(p_tracker->userdata)->_mark_dirty(); + static_cast<GeometryInstanceForwardClustered *>(p_tracker->userdata)->data->dirty_dependencies = true; } break; case Dependency::DEPENDENCY_CHANGED_MULTIMESH_VISIBLE_INSTANCES: { GeometryInstanceForwardClustered *ginstance = static_cast<GeometryInstanceForwardClustered *>(p_tracker->userdata); @@ -3785,6 +3781,7 @@ void RenderForwardClustered::_geometry_instance_dependency_changed(Dependency::D } void RenderForwardClustered::_geometry_instance_dependency_deleted(const RID &p_dependency, DependencyTracker *p_tracker) { static_cast<RenderGeometryInstance *>(p_tracker->userdata)->_mark_dirty(); + static_cast<GeometryInstanceForwardClustered *>(p_tracker->userdata)->data->dirty_dependencies = true; } RenderGeometryInstance *RenderForwardClustered::geometry_instance_create(RID p_base) { @@ -4007,11 +4004,11 @@ RenderForwardClustered::~RenderForwardClustered() { RSG::light_storage->directional_shadow_atlas_set_size(0); { - for (uint32_t i = 0; i < scene_state.uniform_buffers.size(); i++) { - RD::get_singleton()->free(scene_state.uniform_buffers[i]); + for (const RID &rid : scene_state.uniform_buffers) { + RD::get_singleton()->free(rid); } - for (uint32_t i = 0; i < scene_state.implementation_uniform_buffers.size(); i++) { - RD::get_singleton()->free(scene_state.implementation_uniform_buffers[i]); + for (const RID &rid : scene_state.implementation_uniform_buffers) { + RD::get_singleton()->free(rid); } RD::get_singleton()->free(scene_state.lightmap_buffer); RD::get_singleton()->free(scene_state.lightmap_capture_buffer); diff --git a/servers/rendering/renderer_rd/forward_clustered/render_forward_clustered.h b/servers/rendering/renderer_rd/forward_clustered/render_forward_clustered.h index 9245f1b13a..8eb17ba6f4 100644 --- a/servers/rendering/renderer_rd/forward_clustered/render_forward_clustered.h +++ b/servers/rendering/renderer_rd/forward_clustered/render_forward_clustered.h @@ -640,9 +640,9 @@ public: RendererRD::SSEffects *get_ss_effects() { return ss_effects; } /* callback from updating our lighting UBOs, used to populate cluster builder */ - virtual void setup_added_reflection_probe(const Transform3D &p_transform, const Vector3 &p_half_extents) override; + virtual void setup_added_reflection_probe(const Transform3D &p_transform, const Vector3 &p_half_size) override; virtual void setup_added_light(const RS::LightType p_type, const Transform3D &p_transform, float p_radius, float p_spot_aperture) override; - virtual void setup_added_decal(const Transform3D &p_transform, const Vector3 &p_half_extents) override; + virtual void setup_added_decal(const Transform3D &p_transform, const Vector3 &p_half_size) override; virtual void base_uniforms_changed() override; _FORCE_INLINE_ virtual void update_uniform_sets() override { diff --git a/servers/rendering/renderer_rd/forward_clustered/scene_shader_forward_clustered.cpp b/servers/rendering/renderer_rd/forward_clustered/scene_shader_forward_clustered.cpp index 412406c0b4..3b3979b198 100644 --- a/servers/rendering/renderer_rd/forward_clustered/scene_shader_forward_clustered.cpp +++ b/servers/rendering/renderer_rd/forward_clustered/scene_shader_forward_clustered.cpp @@ -353,6 +353,11 @@ void SceneShaderForwardClustered::ShaderData::set_code(const String &p_code) { } int variant = shader_version + shader_flags; + + if (!static_cast<SceneShaderForwardClustered *>(singleton)->shader.is_variant_enabled(variant)) { + continue; + } + RID shader_variant = shader_singleton->shader.version_get_shader(version, variant); color_pipelines[i][j][l].setup(shader_variant, primitive_rd, raster_state, multisample_state, depth_stencil, blend_state, 0, singleton->default_specialization_constants); } @@ -503,7 +508,14 @@ void SceneShaderForwardClustered::init(const String p_defines) { shader.set_variant_enabled(SHADER_VERSION_DEPTH_PASS_MULTIVIEW, false); shader.set_variant_enabled(SHADER_VERSION_DEPTH_PASS_WITH_NORMAL_AND_ROUGHNESS_MULTIVIEW, false); shader.set_variant_enabled(SHADER_VERSION_DEPTH_PASS_WITH_NORMAL_AND_ROUGHNESS_AND_VOXEL_GI_MULTIVIEW, false); - // TODO Add a way to enable/disable color pass flags + + // Disable Color Passes + for (int i = 0; i < SHADER_COLOR_PASS_FLAG_COUNT; i++) { + // Selectively disable any shader pass that includes Multiview. + if ((i & SHADER_COLOR_PASS_FLAG_MULTIVIEW)) { + shader.set_variant_enabled(i + SHADER_VERSION_COLOR_PASS, false); + } + } } } @@ -625,6 +637,7 @@ void SceneShaderForwardClustered::init(const String p_defines) { actions.renames["VIEW_INDEX"] = "ViewIndex"; actions.renames["VIEW_MONO_LEFT"] = "0"; actions.renames["VIEW_RIGHT"] = "1"; + actions.renames["EYE_OFFSET"] = "eye_offset"; //for light actions.renames["VIEW"] = "view"; @@ -717,6 +730,8 @@ void SceneShaderForwardClustered::init(const String p_defines) { actions.global_buffer_array_variable = "global_shader_uniforms.data"; actions.instance_uniform_index_variable = "instances.data[instance_index_interp].instance_uniforms_ofs"; + actions.check_multiview_samplers = true; // make sure we check sampling multiview textures + compiler.initialize(actions); } diff --git a/servers/rendering/renderer_rd/forward_mobile/render_forward_mobile.cpp b/servers/rendering/renderer_rd/forward_mobile/render_forward_mobile.cpp index 816248567b..f9529de6dd 100644 --- a/servers/rendering/renderer_rd/forward_mobile/render_forward_mobile.cpp +++ b/servers/rendering/renderer_rd/forward_mobile/render_forward_mobile.cpp @@ -579,8 +579,8 @@ void RenderForwardMobile::_pre_opaque_render(RenderDataRD *p_render_data) { } //cube shadows are rendered in their own way - for (uint32_t i = 0; i < p_render_data->cube_shadows.size(); i++) { - _render_shadow_pass(p_render_data->render_shadows[p_render_data->cube_shadows[i]].light, p_render_data->shadow_atlas, p_render_data->render_shadows[p_render_data->cube_shadows[i]].pass, p_render_data->render_shadows[p_render_data->cube_shadows[i]].instances, camera_plane, lod_distance_multiplier, p_render_data->scene_data->screen_mesh_lod_threshold, true, true, true, p_render_data->render_info); + for (const int &index : p_render_data->cube_shadows) { + _render_shadow_pass(p_render_data->render_shadows[index].light, p_render_data->shadow_atlas, p_render_data->render_shadows[index].pass, p_render_data->render_shadows[index].instances, camera_plane, lod_distance_multiplier, p_render_data->scene_data->screen_mesh_lod_threshold, true, true, true, p_render_data->render_info); } if (p_render_data->directional_shadows.size()) { @@ -593,12 +593,10 @@ void RenderForwardMobile::_pre_opaque_render(RenderDataRD *p_render_data) { bool render_shadows = p_render_data->directional_shadows.size() || p_render_data->shadows.size(); + //prepare shadow rendering if (render_shadows) { RENDER_TIMESTAMP("Render Shadows"); - } - //prepare shadow rendering - if (render_shadows) { _render_shadow_begin(); //render directional shadows @@ -632,7 +630,7 @@ void RenderForwardMobile::_pre_opaque_render(RenderDataRD *p_render_data) { uint32_t directional_light_count = 0; uint32_t positional_light_count = 0; light_storage->update_light_buffers(p_render_data, *p_render_data->lights, p_render_data->scene_data->cam_transform, p_render_data->shadow_atlas, using_shadows, directional_light_count, positional_light_count, p_render_data->directional_light_soft_shadows); - texture_storage->update_decal_buffer(*p_render_data->decals, p_render_data->scene_data->cam_transform.affine_inverse()); + texture_storage->update_decal_buffer(*p_render_data->decals, p_render_data->scene_data->cam_transform); p_render_data->directional_light_count = directional_light_count; } @@ -640,16 +638,18 @@ void RenderForwardMobile::_pre_opaque_render(RenderDataRD *p_render_data) { void RenderForwardMobile::_render_scene(RenderDataRD *p_render_data, const Color &p_default_bg_color) { RendererRD::LightStorage *light_storage = RendererRD::LightStorage::get_singleton(); - Ref<RenderSceneBuffersRD> rb; + ERR_FAIL_NULL(p_render_data); + + Ref<RenderSceneBuffersRD> rb = p_render_data->render_buffers; + ERR_FAIL_COND(rb.is_null()); + Ref<RenderBufferDataForwardMobile> rb_data; - if (p_render_data->render_buffers.is_valid()) { - rb = p_render_data->render_buffers; - if (rb->has_custom_data(RB_SCOPE_MOBILE)) { - // Our forward mobile custom data buffer will only be available when we're rendering our normal view. - // This will not be available when rendering reflection probes. - rb_data = rb->get_custom_data(RB_SCOPE_MOBILE); - } + if (rb->has_custom_data(RB_SCOPE_MOBILE)) { + // Our forward mobile custom data buffer will only be available when we're rendering our normal view. + // This will not be available when rendering reflection probes. + rb_data = rb->get_custom_data(RB_SCOPE_MOBILE); } + bool is_reflection_probe = p_render_data->reflection_probe.is_valid(); RENDER_TIMESTAMP("Prepare 3D Scene"); @@ -697,7 +697,7 @@ void RenderForwardMobile::_render_scene(RenderDataRD *p_render_data, const Color p_render_data->render_info->info[RS::VIEWPORT_RENDER_INFO_TYPE_VISIBLE][RS::VIEWPORT_RENDER_INFO_OBJECTS_IN_FRAME] = p_render_data->instances->size(); } - if (p_render_data->reflection_probe.is_valid()) { + if (is_reflection_probe) { uint32_t resolution = light_storage->reflection_probe_instance_get_resolution(p_render_data->reflection_probe); screen_size.x = resolution; screen_size.y = resolution; @@ -754,7 +754,7 @@ void RenderForwardMobile::_render_scene(RenderDataRD *p_render_data, const Color RD::get_singleton()->draw_command_begin_label("Render Setup"); _setup_lightmaps(p_render_data, *p_render_data->lightmaps, p_render_data->scene_data->cam_transform); - _setup_environment(p_render_data, p_render_data->reflection_probe.is_valid(), screen_size, !p_render_data->reflection_probe.is_valid(), p_default_bg_color, false); + _setup_environment(p_render_data, is_reflection_probe, screen_size, !is_reflection_probe, p_default_bg_color, false); _update_render_base_uniform_set(); //may have changed due to the above (light buffer enlarged, as an example) @@ -832,7 +832,7 @@ void RenderForwardMobile::_render_scene(RenderDataRD *p_render_data, const Color RD::get_singleton()->draw_command_begin_label("Setup Sky"); // Setup our sky render information for this frame/viewport - if (p_render_data->reflection_probe.is_valid()) { + if (is_reflection_probe) { Vector3 eye_offset; Projection correction; correction.set_depth_correction(true); @@ -896,7 +896,7 @@ void RenderForwardMobile::_render_scene(RenderDataRD *p_render_data, const Color p_render_data->scene_data->directional_light_count = p_render_data->directional_light_count; - _setup_environment(p_render_data, p_render_data->reflection_probe.is_valid(), screen_size, !p_render_data->reflection_probe.is_valid(), p_default_bg_color, p_render_data->render_buffers.is_valid()); + _setup_environment(p_render_data, is_reflection_probe, screen_size, !is_reflection_probe, p_default_bg_color, p_render_data->render_buffers.is_valid()); if (using_subpass_transparent && using_subpass_post_process) { RENDER_TIMESTAMP("Render Opaque + Transparent + Tonemap"); @@ -1025,7 +1025,7 @@ void RenderForwardMobile::_render_scene(RenderDataRD *p_render_data, const Color } // this may be needed if we re-introduced steps that change info, not sure which do so in the previous implementation - // _setup_environment(p_render_data, p_render_data->reflection_probe.is_valid(), screen_size, !p_render_data->reflection_probe.is_valid(), p_default_bg_color, false); + // _setup_environment(p_render_data, is_reflection_probe, screen_size, !is_reflection_probe, p_default_bg_color, false); RD::FramebufferFormatID fb_format = RD::get_singleton()->framebuffer_get_format(framebuffer); RenderListParameters render_list_params(render_list[RENDER_LIST_ALPHA].elements.ptr(), render_list[RENDER_LIST_ALPHA].element_info.ptr(), render_list[RENDER_LIST_ALPHA].elements.size(), reverse_cull, PASS_MODE_COLOR, rp_uniform_set, spec_constant_base_flags, get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_WIREFRAME, Vector2(), p_render_data->scene_data->lod_distance_multiplier, p_render_data->scene_data->screen_mesh_lod_threshold, p_render_data->scene_data->view_count); @@ -1065,9 +1065,7 @@ void RenderForwardMobile::_render_scene(RenderDataRD *p_render_data, const Color _disable_clear_request(p_render_data); } - if (rb.is_valid()) { - _render_buffers_debug_draw(rb, p_render_data->shadow_atlas, p_render_data->occluder_debug_tex); - } + _render_buffers_debug_draw(rb, p_render_data->shadow_atlas, p_render_data->occluder_debug_tex); } /* these are being called from RendererSceneRenderRD::_pre_opaque_render */ @@ -1340,8 +1338,7 @@ void RenderForwardMobile::_render_shadow_process() { void RenderForwardMobile::_render_shadow_end(uint32_t p_barrier) { RD::get_singleton()->draw_command_begin_label("Shadow Render"); - for (uint32_t i = 0; i < scene_state.shadow_passes.size(); i++) { - SceneState::ShadowPass &shadow_pass = scene_state.shadow_passes[i]; + for (SceneState::ShadowPass &shadow_pass : scene_state.shadow_passes) { RenderListParameters render_list_parameters(render_list[RENDER_LIST_SECONDARY].elements.ptr() + shadow_pass.element_from, render_list[RENDER_LIST_SECONDARY].element_info.ptr() + shadow_pass.element_from, shadow_pass.element_count, shadow_pass.flip_cull, shadow_pass.pass_mode, shadow_pass.rp_uniform_set, 0, false, Vector2(), shadow_pass.lod_distance_multiplier, shadow_pass.screen_mesh_lod_threshold, 1, shadow_pass.element_from, RD::BARRIER_MASK_NO_BARRIER); _render_list_with_threads(&render_list_parameters, shadow_pass.framebuffer, RD::INITIAL_ACTION_DROP, RD::FINAL_ACTION_DISCARD, shadow_pass.initial_depth_action, shadow_pass.final_depth_action, Vector<Color>(), 1.0, 0, shadow_pass.rect); } @@ -1925,7 +1922,6 @@ void RenderForwardMobile::_fill_render_list(RenderListType p_render_list, const } void RenderForwardMobile::_setup_environment(const RenderDataRD *p_render_data, bool p_no_fog, const Size2i &p_screen_size, bool p_flip_y, const Color &p_default_bg_color, bool p_opaque_render_buffers, bool p_pancake_shadows, int p_index) { - Ref<RenderSceneBuffersRD> rd = p_render_data->render_buffers; RID env = is_environment(p_render_data->environment) ? p_render_data->environment : RID(); RID reflection_probe_instance = p_render_data->reflection_probe.is_valid() ? RendererRD::LightStorage::get_singleton()->reflection_probe_instance_get_probe(p_render_data->reflection_probe) : RID(); @@ -2238,6 +2234,9 @@ RenderGeometryInstance *RenderForwardMobile::geometry_instance_create(RID p_base ginstance->data->base = p_base; ginstance->data->base_type = type; + ginstance->data->dependency_tracker.userdata = ginstance; + ginstance->data->dependency_tracker.changed_callback = _geometry_instance_dependency_changed; + ginstance->data->dependency_tracker.deleted_callback = _geometry_instance_dependency_deleted; ginstance->_mark_dirty(); @@ -2674,6 +2673,7 @@ void RenderForwardMobile::_geometry_instance_dependency_changed(Dependency::Depe case Dependency::DEPENDENCY_CHANGED_MULTIMESH: case Dependency::DEPENDENCY_CHANGED_SKELETON_DATA: { static_cast<RenderGeometryInstance *>(p_tracker->userdata)->_mark_dirty(); + static_cast<GeometryInstanceForwardMobile *>(p_tracker->userdata)->data->dirty_dependencies = true; } break; case Dependency::DEPENDENCY_CHANGED_MULTIMESH_VISIBLE_INSTANCES: { GeometryInstanceForwardMobile *ginstance = static_cast<GeometryInstanceForwardMobile *>(p_tracker->userdata); @@ -2688,6 +2688,7 @@ void RenderForwardMobile::_geometry_instance_dependency_changed(Dependency::Depe } void RenderForwardMobile::_geometry_instance_dependency_deleted(const RID &p_dependency, DependencyTracker *p_tracker) { static_cast<RenderGeometryInstance *>(p_tracker->userdata)->_mark_dirty(); + static_cast<GeometryInstanceForwardMobile *>(p_tracker->userdata)->data->dirty_dependencies = true; } /* misc */ @@ -2810,8 +2811,8 @@ RenderForwardMobile::~RenderForwardMobile() { } { - for (uint32_t i = 0; i < scene_state.uniform_buffers.size(); i++) { - RD::get_singleton()->free(scene_state.uniform_buffers[i]); + for (const RID &rid : scene_state.uniform_buffers) { + RD::get_singleton()->free(rid); } RD::get_singleton()->free(scene_state.lightmap_buffer); RD::get_singleton()->free(scene_state.lightmap_capture_buffer); diff --git a/servers/rendering/renderer_rd/forward_mobile/scene_shader_forward_mobile.cpp b/servers/rendering/renderer_rd/forward_mobile/scene_shader_forward_mobile.cpp index 3d1d78c63d..cc4a7dfa47 100644 --- a/servers/rendering/renderer_rd/forward_mobile/scene_shader_forward_mobile.cpp +++ b/servers/rendering/renderer_rd/forward_mobile/scene_shader_forward_mobile.cpp @@ -517,6 +517,7 @@ void SceneShaderForwardMobile::init(const String p_defines) { actions.renames["VIEW_INDEX"] = "ViewIndex"; actions.renames["VIEW_MONO_LEFT"] = "0"; actions.renames["VIEW_RIGHT"] = "1"; + actions.renames["EYE_OFFSET"] = "eye_offset"; //for light actions.renames["VIEW"] = "view"; @@ -609,6 +610,7 @@ void SceneShaderForwardMobile::init(const String p_defines) { actions.instance_uniform_index_variable = "draw_call.instance_uniforms_ofs"; actions.apply_luminance_multiplier = true; // apply luminance multiplier to screen texture + actions.check_multiview_samplers = true; // make sure we check sampling multiview textures compiler.initialize(actions); } diff --git a/servers/rendering/renderer_rd/renderer_canvas_render_rd.cpp b/servers/rendering/renderer_rd/renderer_canvas_render_rd.cpp index 462b925134..f102bc0650 100644 --- a/servers/rendering/renderer_rd/renderer_canvas_render_rd.cpp +++ b/servers/rendering/renderer_rd/renderer_canvas_render_rd.cpp @@ -455,7 +455,7 @@ void RendererCanvasRenderRD::_render_item(RD::DrawListID p_draw_list, RID p_rend light_count++; - if (light_count == MAX_LIGHTS_PER_ITEM) { + if (light_count == MAX_LIGHTS_PER_ITEM - 1) { break; } } @@ -1431,6 +1431,7 @@ void RendererCanvasRenderRD::canvas_render_items(RID p_to_render_target, Item *p const Item::CommandMesh *cm = static_cast<const Item::CommandMesh *>(c); if (cm->mesh_instance.is_valid()) { mesh_storage->mesh_instance_check_for_update(cm->mesh_instance); + mesh_storage->mesh_instance_set_canvas_item_transform(cm->mesh_instance, canvas_transform_inverse * ci->final_transform); update_skeletons = true; } } @@ -1549,6 +1550,9 @@ void RendererCanvasRenderRD::light_set_texture(RID p_rid, RID p_texture) { if (cl->texture == p_texture) { return; } + + ERR_FAIL_COND(p_texture.is_valid() && !texture_storage->owns_texture(p_texture)); + if (cl->texture.is_valid()) { texture_storage->texture_remove_from_decal_atlas(cl->texture); } @@ -2630,6 +2634,7 @@ RendererCanvasRenderRD::RendererCanvasRenderRD() { // Default CanvasGroup shader. shader_type canvas_item; +render_mode unshaded; uniform sampler2D screen_texture : hint_screen_texture, repeat_disable, filter_nearest; @@ -2657,6 +2662,7 @@ void fragment() { // Default clip children shader. shader_type canvas_item; +render_mode unshaded; uniform sampler2D screen_texture : hint_screen_texture, repeat_disable, filter_nearest; diff --git a/servers/rendering/renderer_rd/renderer_scene_render_rd.cpp b/servers/rendering/renderer_rd/renderer_scene_render_rd.cpp index 2a87c4fa8d..5776414b14 100644 --- a/servers/rendering/renderer_rd/renderer_scene_render_rd.cpp +++ b/servers/rendering/renderer_rd/renderer_scene_render_rd.cpp @@ -253,6 +253,11 @@ void RendererSceneRenderRD::_render_buffers_copy_screen_texture(const RenderData Ref<RenderSceneBuffersRD> rb = p_render_data->render_buffers; ERR_FAIL_COND(rb.is_null()); + if (!rb->has_internal_texture()) { + // We're likely rendering reflection probes where we can't use our backbuffers. + return; + } + RD::get_singleton()->draw_command_begin_label("Copy screen texture"); rb->allocate_blur_textures(); @@ -275,7 +280,7 @@ void RendererSceneRenderRD::_render_buffers_copy_screen_texture(const RenderData for (int i = 1; i < mipmaps; i++) { RID source = dest; dest = rb->get_texture_slice(RB_SCOPE_BUFFERS, RB_TEX_BLUR_0, v, i); - Size2i msize = rb->get_texture_slice_size(RB_SCOPE_BUFFERS, RB_TEX_BLUR_0, v, i); + Size2i msize = rb->get_texture_slice_size(RB_SCOPE_BUFFERS, RB_TEX_BLUR_0, i); if (can_use_storage) { copy_effects->make_mipmap(source, dest, msize); @@ -292,6 +297,11 @@ void RendererSceneRenderRD::_render_buffers_copy_depth_texture(const RenderDataR Ref<RenderSceneBuffersRD> rb = p_render_data->render_buffers; ERR_FAIL_COND(rb.is_null()); + if (!rb->has_depth_texture()) { + // We're likely rendering reflection probes where we can't use our backbuffers. + return; + } + RD::get_singleton()->draw_command_begin_label("Copy depth texture"); // note, this only creates our back depth texture if we haven't already created it. @@ -321,9 +331,13 @@ void RendererSceneRenderRD::_render_buffers_copy_depth_texture(const RenderDataR void RendererSceneRenderRD::_render_buffers_post_process_and_tonemap(const RenderDataRD *p_render_data) { RendererRD::TextureStorage *texture_storage = RendererRD::TextureStorage::get_singleton(); + ERR_FAIL_NULL(p_render_data); + Ref<RenderSceneBuffersRD> rb = p_render_data->render_buffers; ERR_FAIL_COND(rb.is_null()); + ERR_FAIL_COND_MSG(p_render_data->reflection_probe.is_valid(), "Post processes should not be applied on reflection probes."); + // Glow, auto exposure and DoF (if enabled). Size2i internal_size = rb->get_internal_size(); @@ -434,7 +448,7 @@ void RendererSceneRenderRD::_render_buffers_post_process_and_tonemap(const Rende float luminance_multiplier = _render_buffers_get_luminance_multiplier(); for (uint32_t l = 0; l < rb->get_view_count(); l++) { for (int i = 0; i < (max_glow_level + 1); i++) { - Size2i vp_size = rb->get_texture_slice_size(RB_SCOPE_BUFFERS, RB_TEX_BLUR_1, l, i); + Size2i vp_size = rb->get_texture_slice_size(RB_SCOPE_BUFFERS, RB_TEX_BLUR_1, i); if (i == 0) { RID luminance_texture; @@ -488,7 +502,7 @@ void RendererSceneRenderRD::_render_buffers_post_process_and_tonemap(const Rende tonemap.glow_levels[i] = environment_get_glow_levels(p_render_data->environment)[i]; } - Size2i msize = rb->get_texture_slice_size(RB_SCOPE_BUFFERS, RB_TEX_BLUR_1, 0, 0); + Size2i msize = rb->get_texture_slice_size(RB_SCOPE_BUFFERS, RB_TEX_BLUR_1, 0); tonemap.glow_texture_size.x = msize.width; tonemap.glow_texture_size.y = msize.height; tonemap.glow_use_bicubic_upscale = glow_bicubic_upscale; @@ -918,11 +932,9 @@ void RendererSceneRenderRD::render_scene(const Ref<RenderSceneBuffers> &p_render RendererRD::TextureStorage *texture_storage = RendererRD::TextureStorage::get_singleton(); // getting this here now so we can direct call a bunch of things more easily - Ref<RenderSceneBuffersRD> rb; - if (p_render_buffers.is_valid()) { - rb = p_render_buffers; // cast it... - ERR_FAIL_COND(rb.is_null()); - } + ERR_FAIL_COND(p_render_buffers.is_null()); + Ref<RenderSceneBuffersRD> rb = p_render_buffers; + ERR_FAIL_COND(rb.is_null()); // setup scene data RenderSceneDataRD scene_data; diff --git a/servers/rendering/renderer_rd/renderer_scene_render_rd.h b/servers/rendering/renderer_rd/renderer_scene_render_rd.h index 6fa2f7a570..7c43021eb0 100644 --- a/servers/rendering/renderer_rd/renderer_scene_render_rd.h +++ b/servers/rendering/renderer_rd/renderer_scene_render_rd.h @@ -204,9 +204,9 @@ public: /* LIGHTING */ - virtual void setup_added_reflection_probe(const Transform3D &p_transform, const Vector3 &p_half_extents){}; + virtual void setup_added_reflection_probe(const Transform3D &p_transform, const Vector3 &p_half_size){}; virtual void setup_added_light(const RS::LightType p_type, const Transform3D &p_transform, float p_radius, float p_spot_aperture){}; - virtual void setup_added_decal(const Transform3D &p_transform, const Vector3 &p_half_extents){}; + virtual void setup_added_decal(const Transform3D &p_transform, const Vector3 &p_half_size){}; /* GI */ diff --git a/servers/rendering/renderer_rd/shader_rd.cpp b/servers/rendering/renderer_rd/shader_rd.cpp index 2d5263a3e2..533a912a34 100644 --- a/servers/rendering/renderer_rd/shader_rd.cpp +++ b/servers/rendering/renderer_rd/shader_rd.cpp @@ -165,8 +165,7 @@ void ShaderRD::_clear_version(Version *p_version) { } void ShaderRD::_build_variant_code(StringBuilder &builder, uint32_t p_variant, const Version *p_version, const StageTemplate &p_template) { - for (uint32_t i = 0; i < p_template.chunks.size(); i++) { - const StageTemplate::Chunk &chunk = p_template.chunks[i]; + for (const StageTemplate::Chunk &chunk : p_template.chunks) { switch (chunk.type) { case StageTemplate::Chunk::TYPE_VERSION_DEFINES: { builder.append("\n"); //make sure defines begin at newline diff --git a/servers/rendering/renderer_rd/shaders/effects/ss_effects_downsample.glsl b/servers/rendering/renderer_rd/shaders/effects/ss_effects_downsample.glsl index 134aae5ce7..b1ff46dd3b 100644 --- a/servers/rendering/renderer_rd/shaders/effects/ss_effects_downsample.glsl +++ b/servers/rendering/renderer_rd/shaders/effects/ss_effects_downsample.glsl @@ -56,7 +56,7 @@ vec4 screen_space_to_view_space_depth(vec4 p_depth) { float depth_linearize_mul = params.z_near; float depth_linearize_add = params.z_far; - // Optimised version of "-cameraClipNear / (cameraClipFar - projDepth * (cameraClipFar - cameraClipNear)) * cameraClipFar" + // Optimized 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 ); diff --git a/servers/rendering/renderer_rd/shaders/effects/ssao.glsl b/servers/rendering/renderer_rd/shaders/effects/ssao.glsl index 2a87e273bc..ffaa6872c9 100644 --- a/servers/rendering/renderer_rd/shaders/effects/ssao.glsl +++ b/servers/rendering/renderer_rd/shaders/effects/ssao.glsl @@ -221,7 +221,7 @@ void SSAOTap(const int p_quality_level, inout float r_obscurance_sum, inout floa // 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 + // calculate MIP based on the sample distance from the center, 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); @@ -259,7 +259,7 @@ void generate_SSAO_shadows_internal(out float r_shadow_term, out vec4 r_edges, o // 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) + // get left right top bottom neighboring 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; @@ -304,7 +304,7 @@ void generate_SSAO_shadows_internal(out float r_shadow_term, out vec4 r_edges, o float obscurance_sum = 0.0; float weight_sum = 0.0; - // 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) + // edge mask for between this and left/right/top/bottom neighbor 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); // 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 @@ -318,7 +318,7 @@ void generate_SSAO_shadows_internal(out float r_shadow_term, out vec4 r_edges, o 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" ) + //approximate neighboring 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); diff --git a/servers/rendering/renderer_rd/shaders/effects/ssao_importance_map.glsl b/servers/rendering/renderer_rd/shaders/effects/ssao_importance_map.glsl index 04f98964e8..d234ab4417 100644 --- a/servers/rendering/renderer_rd/shaders/effects/ssao_importance_map.glsl +++ b/servers/rendering/renderer_rd/shaders/effects/ssao_importance_map.glsl @@ -80,7 +80,7 @@ void main() { #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; + float center = textureLod(source_importance, uv, 0.0).x; vec2 half_pixel = params.half_screen_pixel_size; @@ -98,7 +98,7 @@ void main() { #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; + float center = textureLod(source_importance, uv, 0.0).x; vec2 half_pixel = params.half_screen_pixel_size; diff --git a/servers/rendering/renderer_rd/shaders/effects/ssao_interleave.glsl b/servers/rendering/renderer_rd/shaders/effects/ssao_interleave.glsl index f6a9a92fac..45cc62d361 100644 --- a/servers/rendering/renderer_rd/shaders/effects/ssao_interleave.glsl +++ b/servers/rendering/renderer_rd/shaders/effects/ssao_interleave.glsl @@ -60,8 +60,8 @@ void main() { 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_horizontal = (1 - mx) + my * 2; // neighboring, horizontal + int index_vertical = mx + (1 - my) * 2; // neighboring, 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; diff --git a/servers/rendering/renderer_rd/shaders/effects/ssil.glsl b/servers/rendering/renderer_rd/shaders/effects/ssil.glsl index 513791dfbf..de7b97953f 100644 --- a/servers/rendering/renderer_rd/shaders/effects/ssil.glsl +++ b/servers/rendering/renderer_rd/shaders/effects/ssil.glsl @@ -234,7 +234,7 @@ void SSILTap(const int p_quality_level, inout vec3 r_color_sum, inout float r_ob // 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 + // calculate MIP based on the sample distance from the center, similar to as described // in http://graphics.cs.williams.edu/papers/SAOHPG12/. float mip_level = (p_quality_level < SSIL_DEPTH_MIPS_ENABLE_AT_QUALITY_PRESET) ? (0) : (sample_pow_2_len + p_mip_offset); @@ -272,7 +272,7 @@ void generate_SSIL(out vec3 r_color, out vec4 r_edges, out float r_obscurance, o // 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) + // get left right top bottom neighboring 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; @@ -318,7 +318,7 @@ void generate_SSIL(out vec3 r_color, out vec4 r_edges, out float r_obscurance, o float obscurance_sum = 0.0; float weight_sum = 0.0; - // 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) + // edge mask for between this and left/right/top/bottom neighbor 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); // 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 diff --git a/servers/rendering/renderer_rd/shaders/effects/ssil_blur.glsl b/servers/rendering/renderer_rd/shaders/effects/ssil_blur.glsl index 47c56571f6..f48e6c4341 100644 --- a/servers/rendering/renderer_rd/shaders/effects/ssil_blur.glsl +++ b/servers/rendering/renderer_rd/shaders/effects/ssil_blur.glsl @@ -124,14 +124,14 @@ void main() { vec2 uv = (vec2(gl_GlobalInvocationID.xy) + vec2(0.5, 0.5)) * params.half_screen_pixel_size; - vec4 centre = textureLod(source_ssil, uv, 0.0); + vec4 center = textureLod(source_ssil, uv, 0.0); vec4 value = textureLod(source_ssil, vec2(uv + vec2(-half_pixel.x * 3, -half_pixel.y)), 0.0) * 0.2; value += textureLod(source_ssil, vec2(uv + vec2(+half_pixel.x, -half_pixel.y * 3)), 0.0) * 0.2; value += textureLod(source_ssil, vec2(uv + vec2(-half_pixel.x, +half_pixel.y * 3)), 0.0) * 0.2; value += textureLod(source_ssil, vec2(uv + vec2(+half_pixel.x * 3, +half_pixel.y)), 0.0) * 0.2; - vec4 sampled = value + centre * 0.2; + vec4 sampled = value + center * 0.2; #else #ifdef MODE_SMART diff --git a/servers/rendering/renderer_rd/shaders/effects/ssil_importance_map.glsl b/servers/rendering/renderer_rd/shaders/effects/ssil_importance_map.glsl index 6b6b02739d..193e3458ab 100644 --- a/servers/rendering/renderer_rd/shaders/effects/ssil_importance_map.glsl +++ b/servers/rendering/renderer_rd/shaders/effects/ssil_importance_map.glsl @@ -82,7 +82,7 @@ void main() { #ifdef PROCESS_MAPA vec2 uv = (vec2(ssC) + 0.5) * params.half_screen_pixel_size * 2.0; - float centre = textureLod(source_importance, uv, 0.0).x; + float center = textureLod(source_importance, uv, 0.0).x; vec2 half_pixel = params.half_screen_pixel_size; @@ -100,7 +100,7 @@ void main() { #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; + float center = textureLod(source_importance, uv, 0.0).x; vec2 half_pixel = params.half_screen_pixel_size; diff --git a/servers/rendering/renderer_rd/shaders/effects/ssil_interleave.glsl b/servers/rendering/renderer_rd/shaders/effects/ssil_interleave.glsl index 9e86ac0cf0..ed85b8ee4c 100644 --- a/servers/rendering/renderer_rd/shaders/effects/ssil_interleave.glsl +++ b/servers/rendering/renderer_rd/shaders/effects/ssil_interleave.glsl @@ -62,8 +62,8 @@ void main() { 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_horizontal = (1 - mx) + my * 2; // neighboring, horizontal + int index_vertical = mx + (1 - my) * 2; // neighboring, vertical int index_diagonal = (1 - mx) + (1 - my) * 2; // diagonal vec4 color = texelFetch(source_texture, ivec3(pix_pos / uvec2(params.size_modifier), index_center), 0); diff --git a/servers/rendering/renderer_rd/shaders/environment/sdfgi_direct_light.glsl b/servers/rendering/renderer_rd/shaders/environment/sdfgi_direct_light.glsl index 9f7449b8aa..06709f65d3 100644 --- a/servers/rendering/renderer_rd/shaders/environment/sdfgi_direct_light.glsl +++ b/servers/rendering/renderer_rd/shaders/environment/sdfgi_direct_light.glsl @@ -24,7 +24,7 @@ struct ProcessVoxel { uint albedo; // rgb bits 0-15 albedo, bits 16-21 are normal bits (set if geometry exists toward that side), extra 11 bits for neighbors. uint light; // rgbe8985 encoded total saved light, extra 2 bits for neighbors. uint light_aniso; // 55555 light anisotropy, extra 2 bits for neighbors. - //total neighbours: 26 + //total neighbors: 26 }; #ifdef MODE_PROCESS_STATIC @@ -443,10 +443,10 @@ void main() { imageStore(dst_aniso1, positioni, vec4(aniso1, 0.0, 0.0)); imageStore(dst_light, positioni, uvec4(light_total_rgbe)); - //also fill neighbours, so light interpolation during the indirect pass works + //also fill neighbors, so light interpolation during the indirect pass works - //recover the neighbour list from the leftover bits - uint neighbours = (voxel_albedo >> 21) | ((voxel_position >> 21) << 11) | ((process_voxels.data[voxel_index].light >> 30) << 22) | ((process_voxels.data[voxel_index].light_aniso >> 30) << 24); + //recover the neighbor list from the leftover bits + uint neighbors = (voxel_albedo >> 21) | ((voxel_position >> 21) << 11) | ((process_voxels.data[voxel_index].light >> 30) << 22) | ((process_voxels.data[voxel_index].light_aniso >> 30) << 24); const uint max_neighbours = 26; const ivec3 neighbour_positions[max_neighbours] = ivec3[]( @@ -478,7 +478,7 @@ void main() { ivec3(1, 1, 1)); for (uint i = 0; i < max_neighbours; i++) { - if (bool(neighbours & (1 << i))) { + if (bool(neighbors & (1 << i))) { ivec3 neighbour_pos = positioni + neighbour_positions[i]; imageStore(dst_light, neighbour_pos, uvec4(light_total_rgbe)); imageStore(dst_aniso0, neighbour_pos, aniso0); diff --git a/servers/rendering/renderer_rd/shaders/environment/sdfgi_preprocess.glsl b/servers/rendering/renderer_rd/shaders/environment/sdfgi_preprocess.glsl index bce98f4054..dd35ae3b73 100644 --- a/servers/rendering/renderer_rd/shaders/environment/sdfgi_preprocess.glsl +++ b/servers/rendering/renderer_rd/shaders/environment/sdfgi_preprocess.glsl @@ -102,10 +102,10 @@ dispatch_data; struct ProcessVoxel { uint position; // xyz 7 bit packed, extra 11 bits for neighbors. - uint albedo; //rgb bits 0-15 albedo, bits 16-21 are normal bits (set if geometry exists toward that side), extra 11 bits for neighbours - uint light; //rgbe8985 encoded total saved light, extra 2 bits for neighbours - uint light_aniso; //55555 light anisotropy, extra 2 bits for neighbours - //total neighbours: 26 + uint albedo; //rgb bits 0-15 albedo, bits 16-21 are normal bits (set if geometry exists toward that side), extra 11 bits for neighbors + uint light; //rgbe8985 encoded total saved light, extra 2 bits for neighbors + uint light_aniso; //55555 light anisotropy, extra 2 bits for neighbors + //total neighbors: 26 }; layout(set = 0, binding = 11, std430) restrict buffer writeonly ProcessVoxels { @@ -135,10 +135,10 @@ dispatch_data; struct ProcessVoxel { uint position; // xyz 7 bit packed, extra 11 bits for neighbors. - uint albedo; //rgb bits 0-15 albedo, bits 16-21 are normal bits (set if geometry exists toward that side), extra 11 bits for neighbours - uint light; //rgbe8985 encoded total saved light, extra 2 bits for neighbours - uint light_aniso; //55555 light anisotropy, extra 2 bits for neighbours - //total neighbours: 26 + uint albedo; //rgb bits 0-15 albedo, bits 16-21 are normal bits (set if geometry exists toward that side), extra 11 bits for neighbors + uint light; //rgbe8985 encoded total saved light, extra 2 bits for neighbors + uint light_aniso; //55555 light anisotropy, extra 2 bits for neighbors + //total neighbors: 26 }; layout(set = 0, binding = 6, std430) restrict buffer readonly ProcessVoxels { @@ -1016,14 +1016,14 @@ void main() { store_positions[index].albedo = rgb >> 1; //store as it comes (555) to avoid precision loss (and move away the alpha bit) store_positions[index].albedo |= (facing & 0x3F) << 15; // store facing in bits 15-21 - store_positions[index].albedo |= neighbour_bits << 21; //store lower 11 bits of neighbours with remaining albedo - store_positions[index].position |= (neighbour_bits >> 11) << 21; //store 11 bits more of neighbours with position + store_positions[index].albedo |= neighbour_bits << 21; //store lower 11 bits of neighbors with remaining albedo + store_positions[index].position |= (neighbour_bits >> 11) << 21; //store 11 bits more of neighbors with position store_positions[index].light = imageLoad(src_light, pos).r; store_positions[index].light_aniso = imageLoad(src_light_aniso, pos).r; - //add neighbours - store_positions[index].light |= (neighbour_bits >> 22) << 30; //store 2 bits more of neighbours with light - store_positions[index].light_aniso |= (neighbour_bits >> 24) << 30; //store 2 bits more of neighbours with aniso + //add neighbors + store_positions[index].light |= (neighbour_bits >> 22) << 30; //store 2 bits more of neighbors with light + store_positions[index].light_aniso |= (neighbour_bits >> 24) << 30; //store 2 bits more of neighbors with aniso } groupMemoryBarrier(); diff --git a/servers/rendering/renderer_rd/shaders/environment/volumetric_fog.glsl b/servers/rendering/renderer_rd/shaders/environment/volumetric_fog.glsl index 4658afd02d..8e4f5762fd 100644 --- a/servers/rendering/renderer_rd/shaders/environment/volumetric_fog.glsl +++ b/servers/rendering/renderer_rd/shaders/environment/volumetric_fog.glsl @@ -37,7 +37,7 @@ layout(push_constant, std430) uniform Params { vec3 position; float pad; - vec3 extents; + vec3 size; float pad2; ivec3 corner; @@ -184,36 +184,37 @@ void main() { vec4 local_pos = params.transform * world; local_pos.xyz /= local_pos.w; + vec3 half_size = params.size / 2.0; float sdf = -1.0; if (params.shape == 0) { // Ellipsoid // https://www.shadertoy.com/view/tdS3DG - float k0 = length(local_pos.xyz / params.extents); - float k1 = length(local_pos.xyz / (params.extents * params.extents)); + float k0 = length(local_pos.xyz / half_size); + float k1 = length(local_pos.xyz / (half_size * half_size)); sdf = k0 * (k0 - 1.0) / k1; } else if (params.shape == 1) { // Cone // https://iquilezles.org/www/articles/distfunctions/distfunctions.htm - // Compute the cone angle automatically to fit within the volume's extents. - float inv_height = 1.0 / max(0.001, params.extents.y); - float radius = 1.0 / max(0.001, (min(params.extents.x, params.extents.z) * 0.5)); + // Compute the cone angle automatically to fit within the volume's size. + float inv_height = 1.0 / max(0.001, half_size.y); + float radius = 1.0 / max(0.001, (min(half_size.x, half_size.z) * 0.5)); float hypotenuse = sqrt(radius * radius + inv_height * inv_height); float rsin = radius / hypotenuse; float rcos = inv_height / hypotenuse; vec2 c = vec2(rsin, rcos); float q = length(local_pos.xz); - sdf = max(dot(c, vec2(q, local_pos.y - params.extents.y)), -params.extents.y - local_pos.y); + sdf = max(dot(c, vec2(q, local_pos.y - half_size.y)), -half_size.y - local_pos.y); } else if (params.shape == 2) { // Cylinder // https://iquilezles.org/www/articles/distfunctions/distfunctions.htm - vec2 d = abs(vec2(length(local_pos.xz), local_pos.y)) - vec2(min(params.extents.x, params.extents.z), params.extents.y); + vec2 d = abs(vec2(length(local_pos.xz), local_pos.y)) - vec2(min(half_size.x, half_size.z), half_size.y); sdf = min(max(d.x, d.y), 0.0) + length(max(d, 0.0)); } else if (params.shape == 3) { // Box // https://iquilezles.org/www/articles/distfunctions/distfunctions.htm - vec3 q = abs(local_pos.xyz) - params.extents; + vec3 q = abs(local_pos.xyz) - half_size; sdf = length(max(q, 0.0)) + min(max(q.x, max(q.y, q.z)), 0.0); } @@ -222,7 +223,7 @@ void main() { #ifndef SDF_USED cull_mask = 1.0 - smoothstep(-0.1, 0.0, sdf); #endif - uvw = clamp((local_pos.xyz + params.extents) / (2.0 * params.extents), 0.0, 1.0); + uvw = clamp((local_pos.xyz + half_size) / params.size, 0.0, 1.0); } if (cull_mask > 0.0) { diff --git a/servers/rendering/renderer_rd/shaders/forward_clustered/scene_forward_clustered.glsl b/servers/rendering/renderer_rd/shaders/forward_clustered/scene_forward_clustered.glsl index d32e6d717f..21fa7fa148 100644 --- a/servers/rendering/renderer_rd/shaders/forward_clustered/scene_forward_clustered.glsl +++ b/servers/rendering/renderer_rd/shaders/forward_clustered/scene_forward_clustered.glsl @@ -118,13 +118,13 @@ layout(location = 10) out flat uint instance_index_interp; // !BAS! This needs to become an input once we implement our fallback! #define ViewIndex 0 #endif // has_VK_KHR_multiview -vec3 normal_roughness_uv(vec2 uv) { +vec3 multiview_uv(vec2 uv) { return vec3(uv, ViewIndex); } #else // USE_MULTIVIEW // Set to zero, not supported in non stereo #define ViewIndex 0 -vec2 normal_roughness_uv(vec2 uv) { +vec2 multiview_uv(vec2 uv) { return uv; } #endif //USE_MULTIVIEW @@ -315,9 +315,11 @@ void vertex_shader(in uint instance_index, in bool is_multimesh, in uint multime #ifdef USE_MULTIVIEW mat4 projection_matrix = scene_data.projection_matrix_view[ViewIndex]; mat4 inv_projection_matrix = scene_data.inv_projection_matrix_view[ViewIndex]; + vec3 eye_offset = scene_data.eye_offset[ViewIndex].xyz; #else mat4 projection_matrix = scene_data.projection_matrix; mat4 inv_projection_matrix = scene_data.inv_projection_matrix; + vec3 eye_offset = vec3(0.0, 0.0, 0.0); #endif //USE_MULTIVIEW //using world coordinates @@ -550,13 +552,13 @@ layout(location = 10) in flat uint instance_index_interp; // !BAS! This needs to become an input once we implement our fallback! #define ViewIndex 0 #endif // has_VK_KHR_multiview -vec3 normal_roughness_uv(vec2 uv) { +vec3 multiview_uv(vec2 uv) { return vec3(uv, ViewIndex); } #else // USE_MULTIVIEW // Set to zero, not supported in non stereo #define ViewIndex 0 -vec2 normal_roughness_uv(vec2 uv) { +vec2 multiview_uv(vec2 uv) { return uv; } #endif //USE_MULTIVIEW @@ -722,8 +724,10 @@ void fragment_shader(in SceneData scene_data) { //lay out everything, whatever is unused is optimized away anyway vec3 vertex = vertex_interp; #ifdef USE_MULTIVIEW - vec3 view = -normalize(vertex_interp - scene_data.eye_offset[ViewIndex].xyz); + vec3 eye_offset = scene_data.eye_offset[ViewIndex].xyz; + vec3 view = -normalize(vertex_interp - eye_offset); #else + vec3 eye_offset = vec3(0.0, 0.0, 0.0); vec3 view = -normalize(vertex_interp); #endif vec3 albedo = vec3(1.0); @@ -1305,24 +1309,26 @@ void fragment_shader(in SceneData scene_data) { } if (sc_use_forward_gi && bool(instances.data[instance_index].flags & INSTANCE_FLAGS_USE_VOXEL_GI)) { // process voxel_gi_instances - uint index1 = instances.data[instance_index].gi_offset & 0xFFFF; - vec3 ref_vec = normalize(reflect(-view, normal)); - ref_vec = mix(ref_vec, normal, roughness * roughness); + // Make vertex orientation the world one, but still align to camera. + vec3 cam_pos = mat3(scene_data.inv_view_matrix) * vertex; + vec3 cam_normal = mat3(scene_data.inv_view_matrix) * normal; + vec3 ref_vec = mat3(scene_data.inv_view_matrix) * normalize(reflect(-view, normal)); + //find arbitrary tangent and bitangent, then build a matrix - vec3 v0 = abs(normal.z) < 0.999 ? vec3(0.0, 0.0, 1.0) : vec3(0.0, 1.0, 0.0); - vec3 tangent = normalize(cross(v0, normal)); - vec3 bitangent = normalize(cross(tangent, normal)); - mat3 normal_mat = mat3(tangent, bitangent, normal); + vec3 v0 = abs(cam_normal.z) < 0.999 ? vec3(0.0, 0.0, 1.0) : vec3(0.0, 1.0, 0.0); + vec3 tangent = normalize(cross(v0, cam_normal)); + vec3 bitangent = normalize(cross(tangent, cam_normal)); + mat3 normal_mat = mat3(tangent, bitangent, cam_normal); vec4 amb_accum = vec4(0.0); vec4 spec_accum = vec4(0.0); - voxel_gi_compute(index1, vertex, normal, ref_vec, normal_mat, roughness * roughness, ambient_light, specular_light, spec_accum, amb_accum); + voxel_gi_compute(index1, cam_pos, cam_normal, ref_vec, normal_mat, roughness * roughness, ambient_light, specular_light, spec_accum, amb_accum); uint index2 = instances.data[instance_index].gi_offset >> 16; if (index2 != 0xFFFF) { - voxel_gi_compute(index2, vertex, normal, ref_vec, normal_mat, roughness * roughness, ambient_light, specular_light, spec_accum, amb_accum); + voxel_gi_compute(index2, cam_pos, cam_normal, ref_vec, normal_mat, roughness * roughness, ambient_light, specular_light, spec_accum, amb_accum); } if (amb_accum.a > 0.0) { @@ -1351,8 +1357,8 @@ void fragment_shader(in SceneData scene_data) { #endif // USE_MULTIVIEW for (int i = 0; i < 4; i++) { - const vec2 neighbours[4] = vec2[](vec2(-1, 0), vec2(1, 0), vec2(0, -1), vec2(0, 1)); - vec2 neighbour_coord = base_coord + neighbours[i] * scene_data.screen_pixel_size; + const vec2 neighbors[4] = vec2[](vec2(-1, 0), vec2(1, 0), vec2(0, -1), vec2(0, 1)); + vec2 neighbour_coord = base_coord + neighbors[i] * scene_data.screen_pixel_size; #ifdef USE_MULTIVIEW float neighbour_ang = dot(normal, textureLod(sampler2DArray(normal_roughness_buffer, material_samplers[SAMPLER_LINEAR_CLAMP]), vec3(neighbour_coord, ViewIndex), 0.0).xyz * 2.0 - 1.0); #else // USE_MULTIVIEW @@ -1798,7 +1804,7 @@ void fragment_shader(in SceneData scene_data) { shadow = float(shadow1 >> ((i - 4u) * 8u) & 0xFFu) / 255.0; } - shadow = shadow * directional_lights.data[i].shadow_opacity + 1.0 - directional_lights.data[i].shadow_opacity; + shadow = mix(1.0, shadow, directional_lights.data[i].shadow_opacity); #endif blur_shadow(shadow); @@ -2082,7 +2088,7 @@ void fragment_shader(in SceneData scene_data) { float sRed = floor((cRed / pow(2.0f, exps - B - N)) + 0.5f); float sGreen = floor((cGreen / pow(2.0f, exps - B - N)) + 0.5f); float sBlue = floor((cBlue / pow(2.0f, exps - B - N)) + 0.5f); - //store as 8985 to have 2 extra neighbour bits + //store as 8985 to have 2 extra neighbor bits uint light_rgbe = ((uint(sRed) & 0x1FFu) >> 1) | ((uint(sGreen) & 0x1FFu) << 8) | (((uint(sBlue) & 0x1FFu) >> 1) << 17) | ((uint(exps) & 0x1Fu) << 25); imageStore(emission_grid, grid_pos, uvec4(light_rgbe)); diff --git a/servers/rendering/renderer_rd/shaders/forward_clustered/scene_forward_clustered_inc.glsl b/servers/rendering/renderer_rd/shaders/forward_clustered/scene_forward_clustered_inc.glsl index 1f524313f2..8ff7a784dc 100644 --- a/servers/rendering/renderer_rd/shaders/forward_clustered/scene_forward_clustered_inc.glsl +++ b/servers/rendering/renderer_rd/shaders/forward_clustered/scene_forward_clustered_inc.glsl @@ -271,15 +271,16 @@ layout(r32ui, set = 1, binding = 13) uniform restrict uimage3D geom_facing_grid; #define multiviewSampler sampler2D #else -layout(set = 1, binding = 10) uniform texture2D depth_buffer; -layout(set = 1, binding = 11) uniform texture2D color_buffer; - #ifdef USE_MULTIVIEW +layout(set = 1, binding = 10) uniform texture2DArray depth_buffer; +layout(set = 1, binding = 11) uniform texture2DArray color_buffer; layout(set = 1, binding = 12) uniform texture2DArray normal_roughness_buffer; layout(set = 1, binding = 14) uniform texture2DArray ambient_buffer; layout(set = 1, binding = 15) uniform texture2DArray reflection_buffer; #define multiviewSampler sampler2DArray #else // USE_MULTIVIEW +layout(set = 1, binding = 10) uniform texture2D depth_buffer; +layout(set = 1, binding = 11) uniform texture2D color_buffer; layout(set = 1, binding = 12) uniform texture2D normal_roughness_buffer; layout(set = 1, binding = 14) uniform texture2D ambient_buffer; layout(set = 1, binding = 15) uniform texture2D reflection_buffer; diff --git a/servers/rendering/renderer_rd/shaders/forward_mobile/scene_forward_mobile.glsl b/servers/rendering/renderer_rd/shaders/forward_mobile/scene_forward_mobile.glsl index 5e64d4e651..a8b28bbd4f 100644 --- a/servers/rendering/renderer_rd/shaders/forward_mobile/scene_forward_mobile.glsl +++ b/servers/rendering/renderer_rd/shaders/forward_mobile/scene_forward_mobile.glsl @@ -112,9 +112,15 @@ layout(location = 9) out highp float dp_clip; // !BAS! This needs to become an input once we implement our fallback! #define ViewIndex 0 #endif +vec3 multiview_uv(vec2 uv) { + return vec3(uv, ViewIndex); +} #else // Set to zero, not supported in non stereo #define ViewIndex 0 +vec2 multiview_uv(vec2 uv) { + return uv; +} #endif //USE_MULTIVIEW invariant gl_Position; @@ -308,9 +314,11 @@ void main() { #ifdef USE_MULTIVIEW mat4 projection_matrix = scene_data.projection_matrix_view[ViewIndex]; mat4 inv_projection_matrix = scene_data.inv_projection_matrix_view[ViewIndex]; + vec3 eye_offset = scene_data.eye_offset[ViewIndex].xyz; #else mat4 projection_matrix = scene_data.projection_matrix; mat4 inv_projection_matrix = scene_data.inv_projection_matrix; + vec3 eye_offset = vec3(0.0, 0.0, 0.0); #endif //USE_MULTIVIEW //using world coordinates @@ -523,9 +531,15 @@ layout(location = 9) highp in float dp_clip; // !BAS! This needs to become an input once we implement our fallback! #define ViewIndex 0 #endif +vec3 multiview_uv(vec2 uv) { + return vec3(uv, ViewIndex); +} #else // Set to zero, not supported in non stereo #define ViewIndex 0 +vec2 multiview_uv(vec2 uv) { + return uv; +} #endif //USE_MULTIVIEW //defines to keep compatibility with vertex @@ -659,8 +673,10 @@ void main() { //lay out everything, whatever is unused is optimized away anyway vec3 vertex = vertex_interp; #ifdef USE_MULTIVIEW - vec3 view = -normalize(vertex_interp - scene_data.eye_offset[ViewIndex].xyz); + vec3 eye_offset = scene_data.eye_offset[ViewIndex].xyz; + vec3 view = -normalize(vertex_interp - eye_offset); #else + vec3 eye_offset = vec3(0.0, 0.0, 0.0); vec3 view = -normalize(vertex_interp); #endif vec3 albedo = vec3(1.0); @@ -1515,6 +1531,8 @@ void main() { } else { shadow = float(shadow1 >> ((i - 4) * 8) & 0xFF) / 255.0; } + + shadow = mix(1.0, shadow, directional_lights.data[i].shadow_opacity); #endif blur_shadow(shadow); diff --git a/servers/rendering/renderer_rd/shaders/forward_mobile/scene_forward_mobile_inc.glsl b/servers/rendering/renderer_rd/shaders/forward_mobile/scene_forward_mobile_inc.glsl index 631ff0575b..78b39a356d 100644 --- a/servers/rendering/renderer_rd/shaders/forward_mobile/scene_forward_mobile_inc.glsl +++ b/servers/rendering/renderer_rd/shaders/forward_mobile/scene_forward_mobile_inc.glsl @@ -154,8 +154,15 @@ layout(set = 1, binding = 5) uniform highp texture2D directional_shadow_atlas; // this needs to change to providing just the lightmap we're using.. layout(set = 1, binding = 6) uniform texture2DArray lightmap_textures[MAX_LIGHTMAP_TEXTURES]; +#ifdef USE_MULTIVIEW +layout(set = 1, binding = 9) uniform highp texture2DArray depth_buffer; +layout(set = 1, binding = 10) uniform mediump texture2DArray color_buffer; +#define multiviewSampler sampler2DArray +#else layout(set = 1, binding = 9) uniform highp texture2D depth_buffer; layout(set = 1, binding = 10) uniform mediump texture2D color_buffer; +#define multiviewSampler sampler2D +#endif // USE_MULTIVIEW /* Set 2 Skeleton & Instancing (can change per item) */ diff --git a/servers/rendering/renderer_rd/shaders/scene_forward_lights_inc.glsl b/servers/rendering/renderer_rd/shaders/scene_forward_lights_inc.glsl index b30b0c8169..9dda62c28d 100644 --- a/servers/rendering/renderer_rd/shaders/scene_forward_lights_inc.glsl +++ b/servers/rendering/renderer_rd/shaders/scene_forward_lights_inc.glsl @@ -794,8 +794,13 @@ void light_process_spot(uint idx, vec3 vertex, vec3 eye_vec, vec3 normal, vec3 v float light_length = length(light_rel_vec); float spot_attenuation = get_omni_attenuation(light_length, spot_lights.data[idx].inv_radius, spot_lights.data[idx].attenuation); vec3 spot_dir = spot_lights.data[idx].direction; - float scos = max(dot(-normalize(light_rel_vec), spot_dir), spot_lights.data[idx].cone_angle); - float spot_rim = max(0.0001, (1.0 - scos) / (1.0 - spot_lights.data[idx].cone_angle)); + + // This conversion to a highp float is crucial to prevent light leaking + // due to precision errors in the following calculations (cone angle is mediump). + highp float cone_angle = spot_lights.data[idx].cone_angle; + float scos = max(dot(-normalize(light_rel_vec), spot_dir), cone_angle); + float spot_rim = max(0.0001, (1.0 - scos) / (1.0 - cone_angle)); + spot_attenuation *= 1.0 - pow(spot_rim, spot_lights.data[idx].cone_attenuation); float light_attenuation = spot_attenuation; vec3 color = spot_lights.data[idx].color; diff --git a/servers/rendering/renderer_rd/shaders/skeleton.glsl b/servers/rendering/renderer_rd/shaders/skeleton.glsl index f5b233cca0..59c161548c 100644 --- a/servers/rendering/renderer_rd/shaders/skeleton.glsl +++ b/servers/rendering/renderer_rd/shaders/skeleton.glsl @@ -51,6 +51,15 @@ layout(push_constant, std430) uniform Params { bool normalized_blend_shapes; uint pad0; uint pad1; + + vec2 skeleton_transform_x; + vec2 skeleton_transform_y; + + vec2 skeleton_transform_offset; + vec2 inverse_transform_x; + + vec2 inverse_transform_y; + vec2 inverse_transform_offset; } params; @@ -158,8 +167,12 @@ void main() { m += mat4(bone_transforms.data[bones_23.x], bone_transforms.data[bones_23.x + 1], vec4(0.0, 0.0, 1.0, 0.0), vec4(0.0, 0.0, 0.0, 1.0)) * weights_23.x; m += mat4(bone_transforms.data[bones_23.y], bone_transforms.data[bones_23.y + 1], vec4(0.0, 0.0, 1.0, 0.0), vec4(0.0, 0.0, 0.0, 1.0)) * weights_23.y; - //reverse order because its transposed - vertex = (vec4(vertex, 0.0, 1.0) * m).xy; + mat4 skeleton_matrix = mat4(vec4(params.skeleton_transform_x, 0.0, 0.0), vec4(params.skeleton_transform_y, 0.0, 0.0), vec4(0.0, 0.0, 1.0, 0.0), vec4(params.skeleton_transform_offset, 0.0, 1.0)); + mat4 inverse_matrix = mat4(vec4(params.inverse_transform_x, 0.0, 0.0), vec4(params.inverse_transform_y, 0.0, 0.0), vec4(0.0, 0.0, 1.0, 0.0), vec4(params.inverse_transform_offset, 0.0, 1.0)); + + m = skeleton_matrix * transpose(m) * inverse_matrix; + + vertex = (m * vec4(vertex, 0.0, 1.0)).xy; } uint dst_offset = index * params.vertex_stride; diff --git a/servers/rendering/renderer_rd/storage_rd/light_storage.cpp b/servers/rendering/renderer_rd/storage_rd/light_storage.cpp index 673fc25595..e65d842a67 100644 --- a/servers/rendering/renderer_rd/storage_rd/light_storage.cpp +++ b/servers/rendering/renderer_rd/storage_rd/light_storage.cpp @@ -232,6 +232,8 @@ void LightStorage::light_set_projector(RID p_light, RID p_texture) { return; } + ERR_FAIL_COND(p_texture.is_valid() && !texture_storage->owns_texture(p_texture)); + if (light->type != RS::LIGHT_DIRECTIONAL && light->projector.is_valid()) { texture_storage->texture_remove_from_decal_atlas(light->projector, light->type == RS::LIGHT_OMNI); } @@ -387,6 +389,13 @@ uint64_t LightStorage::light_get_version(RID p_light) const { return light->version; } +uint32_t LightStorage::light_get_cull_mask(RID p_light) const { + const Light *light = light_owner.get_or_null(p_light); + ERR_FAIL_COND_V(!light, 0); + + return light->cull_mask; +} + AABB LightStorage::light_get_aabb(RID p_light) const { const Light *light = light_owner.get_or_null(p_light); ERR_FAIL_COND_V(!light, AABB()); @@ -568,8 +577,6 @@ void LightStorage::update_light_buffers(RenderDataRD *p_render_data, const Paged r_directional_light_count = 0; r_positional_light_count = 0; - Plane camera_plane(-p_camera_transform.basis.get_column(Vector3::AXIS_Z).normalized(), p_camera_transform.origin); - omni_light_count = 0; spot_light_count = 0; @@ -720,7 +727,7 @@ void LightStorage::update_light_buffers(RenderDataRD *p_render_data, const Paged } Transform3D light_transform = light_instance->transform; - const real_t distance = camera_plane.distance_to(light_transform.origin); + const real_t distance = p_camera_transform.origin.distance_to(light_transform.origin); if (light->distance_fade) { const float fade_begin = light->distance_fade_begin; @@ -745,7 +752,7 @@ void LightStorage::update_light_buffers(RenderDataRD *p_render_data, const Paged } Transform3D light_transform = light_instance->transform; - const real_t distance = camera_plane.distance_to(light_transform.origin); + const real_t distance = p_camera_transform.origin.distance_to(light_transform.origin); if (light->distance_fade) { const float fade_begin = light->distance_fade_begin; @@ -787,6 +794,7 @@ void LightStorage::update_light_buffers(RenderDataRD *p_render_data, const Paged RS::LightType type = (i < omni_light_count) ? RS::LIGHT_OMNI : RS::LIGHT_SPOT; LightInstance *light_instance = (i < omni_light_count) ? omni_light_sort[index].light_instance : spot_light_sort[index].light_instance; Light *light = (i < omni_light_count) ? omni_light_sort[index].light : spot_light_sort[index].light; + real_t distance = (i < omni_light_count) ? omni_light_sort[index].depth : spot_light_sort[index].depth; if (using_forward_ids) { forward_id_storage->map_forward_id(type == RS::LIGHT_OMNI ? RendererRD::FORWARD_ID_TYPE_OMNI_LIGHT : RendererRD::FORWARD_ID_TYPE_SPOT_LIGHT, light_instance->forward_id, index); @@ -803,7 +811,6 @@ void LightStorage::update_light_buffers(RenderDataRD *p_render_data, const Paged float fade_begin = 0.0; float fade_shadow = 0.0; float fade_length = 0.0; - real_t distance = 0.0; float fade = 1.0; float shadow_opacity_fade = 1.0; @@ -811,7 +818,6 @@ void LightStorage::update_light_buffers(RenderDataRD *p_render_data, const Paged fade_begin = light->distance_fade_begin; fade_shadow = light->distance_fade_shadow; fade_length = light->distance_fade_length; - distance = camera_plane.distance_to(light_transform.origin); // Use `smoothstep()` to make opacity changes more gradual and less noticeable to the player. if (distance > fade_begin) { @@ -1067,14 +1073,14 @@ void LightStorage::reflection_probe_set_max_distance(RID p_probe, float p_distan reflection_probe->dependency.changed_notify(Dependency::DEPENDENCY_CHANGED_REFLECTION_PROBE); } -void LightStorage::reflection_probe_set_extents(RID p_probe, const Vector3 &p_extents) { +void LightStorage::reflection_probe_set_size(RID p_probe, const Vector3 &p_size) { ReflectionProbe *reflection_probe = reflection_probe_owner.get_or_null(p_probe); ERR_FAIL_COND(!reflection_probe); - if (reflection_probe->extents == p_extents) { + if (reflection_probe->size == p_size) { return; } - reflection_probe->extents = p_extents; + reflection_probe->size = p_size; reflection_probe->dependency.changed_notify(Dependency::DEPENDENCY_CHANGED_REFLECTION_PROBE); } @@ -1146,8 +1152,8 @@ AABB LightStorage::reflection_probe_get_aabb(RID p_probe) const { ERR_FAIL_COND_V(!reflection_probe, AABB()); AABB aabb; - aabb.position = -reflection_probe->extents; - aabb.size = reflection_probe->extents * 2.0; + aabb.position = -reflection_probe->size / 2; + aabb.size = reflection_probe->size; return aabb; } @@ -1166,11 +1172,11 @@ uint32_t LightStorage::reflection_probe_get_cull_mask(RID p_probe) const { return reflection_probe->cull_mask; } -Vector3 LightStorage::reflection_probe_get_extents(RID p_probe) const { +Vector3 LightStorage::reflection_probe_get_size(RID p_probe) const { const ReflectionProbe *reflection_probe = reflection_probe_owner.get_or_null(p_probe); ERR_FAIL_COND_V(!reflection_probe, Vector3()); - return reflection_probe->extents; + return reflection_probe->size; } Vector3 LightStorage::reflection_probe_get_origin_offset(RID p_probe) const { @@ -1467,7 +1473,6 @@ bool LightStorage::reflection_probe_instance_begin_render(RID p_instance, RID p_ fb.push_back(atlas->depth_buffer); atlas->depth_fb = RD::get_singleton()->framebuffer_create(fb); - atlas->render_buffers->cleanup(); atlas->render_buffers->configure_for_reflections(Size2i(atlas->size, atlas->size)); } @@ -1556,6 +1561,11 @@ bool LightStorage::reflection_probe_instance_postprocess_step(RID p_instance) { if (rpi->processing_side == 6) { rpi->processing_side = 0; rpi->processing_layer++; + if (rpi->processing_layer == atlas->reflections[rpi->atlas_index].data.layers[0].mipmaps.size()) { + rpi->rendering = false; + rpi->processing_layer = 1; + return true; + } } return false; @@ -1671,7 +1681,7 @@ void LightStorage::update_reflection_probe_buffer(RenderDataRD *p_render_data, c ReflectionData &reflection_ubo = reflections[i]; - Vector3 extents = probe->extents; + Vector3 extents = probe->size / 2; rpi->cull_mask = probe->cull_mask; diff --git a/servers/rendering/renderer_rd/storage_rd/light_storage.h b/servers/rendering/renderer_rd/storage_rd/light_storage.h index f7c8b8833b..c36d1ef503 100644 --- a/servers/rendering/renderer_rd/storage_rd/light_storage.h +++ b/servers/rendering/renderer_rd/storage_rd/light_storage.h @@ -226,7 +226,7 @@ private: Color ambient_color; float ambient_color_energy = 1.0; float max_distance = 0; - Vector3 extents = Vector3(1, 1, 1); + Vector3 size = Vector3(20, 20, 20); Vector3 origin_offset; bool interior = false; bool box_projection = false; @@ -517,13 +517,6 @@ public: return light->color; } - _FORCE_INLINE_ uint32_t light_get_cull_mask(RID p_light) { - const Light *light = light_owner.get_or_null(p_light); - ERR_FAIL_COND_V(!light, 0); - - return light->cull_mask; - } - _FORCE_INLINE_ bool light_is_distance_fade_enabled(RID p_light) { const Light *light = light_owner.get_or_null(p_light); return light->distance_fade; @@ -575,6 +568,7 @@ public: virtual RS::LightBakeMode light_get_bake_mode(RID p_light) override; virtual uint32_t light_get_max_sdfgi_cascade(RID p_light) override; virtual uint64_t light_get_version(RID p_light) const override; + virtual uint32_t light_get_cull_mask(RID p_light) const override; Dependency *light_get_dependency(RID p_light) const; @@ -791,7 +785,7 @@ public: virtual void reflection_probe_set_ambient_color(RID p_probe, const Color &p_color) override; virtual void reflection_probe_set_ambient_energy(RID p_probe, float p_energy) override; virtual void reflection_probe_set_max_distance(RID p_probe, float p_distance) override; - virtual void reflection_probe_set_extents(RID p_probe, const Vector3 &p_extents) override; + virtual void reflection_probe_set_size(RID p_probe, const Vector3 &p_size) override; virtual void reflection_probe_set_origin_offset(RID p_probe, const Vector3 &p_offset) override; virtual void reflection_probe_set_as_interior(RID p_probe, bool p_enable) override; virtual void reflection_probe_set_enable_box_projection(RID p_probe, bool p_enable) override; @@ -805,7 +799,7 @@ public: virtual AABB reflection_probe_get_aabb(RID p_probe) const override; virtual RS::ReflectionProbeUpdateMode reflection_probe_get_update_mode(RID p_probe) const override; virtual uint32_t reflection_probe_get_cull_mask(RID p_probe) const override; - virtual Vector3 reflection_probe_get_extents(RID p_probe) const override; + virtual Vector3 reflection_probe_get_size(RID p_probe) const override; virtual Vector3 reflection_probe_get_origin_offset(RID p_probe) const override; virtual float reflection_probe_get_origin_max_distance(RID p_probe) const override; virtual float reflection_probe_get_mesh_lod_threshold(RID p_probe) const override; diff --git a/servers/rendering/renderer_rd/storage_rd/material_storage.h b/servers/rendering/renderer_rd/storage_rd/material_storage.h index 0ac5557659..ac217d9a49 100644 --- a/servers/rendering/renderer_rd/storage_rd/material_storage.h +++ b/servers/rendering/renderer_rd/storage_rd/material_storage.h @@ -323,13 +323,13 @@ public: // http://andrewthall.org/papers/df64_qf128.pdf #ifdef REAL_T_IS_DOUBLE - static _FORCE_INLINE_ void split_double(double a, float *ahi, float *alo) { + static _FORCE_INLINE_ void split_double(double a, float *a_hi, float *a_lo) { const double SPLITTER = (1 << 29) + 1; double t = a * SPLITTER; - double thi = t - (t - a); - double tlo = a - thi; - *ahi = (float)thi; - *alo = (float)tlo; + double t_hi = t - (t - a); + double t_lo = a - t_hi; + *a_hi = (float)t_hi; + *a_lo = (float)t_lo; } #endif diff --git a/servers/rendering/renderer_rd/storage_rd/mesh_storage.cpp b/servers/rendering/renderer_rd/storage_rd/mesh_storage.cpp index 62da62403f..96618c3352 100644 --- a/servers/rendering/renderer_rd/storage_rd/mesh_storage.cpp +++ b/servers/rendering/renderer_rd/storage_rd/mesh_storage.cpp @@ -586,6 +586,8 @@ void MeshStorage::mesh_set_custom_aabb(RID p_mesh, const AABB &p_aabb) { Mesh *mesh = mesh_owner.get_or_null(p_mesh); ERR_FAIL_COND(!mesh); mesh->custom_aabb = p_aabb; + + mesh->dependency.changed_notify(Dependency::DEPENDENCY_CHANGED_AABB); } AABB MeshStorage::mesh_get_custom_aabb(RID p_mesh) const { @@ -842,15 +844,15 @@ void MeshStorage::mesh_instance_set_blend_shape_weight(RID p_mesh_instance, int } void MeshStorage::_mesh_instance_clear(MeshInstance *mi) { - for (uint32_t i = 0; i < mi->surfaces.size(); i++) { - if (mi->surfaces[i].versions) { - for (uint32_t j = 0; j < mi->surfaces[i].version_count; j++) { - RD::get_singleton()->free(mi->surfaces[i].versions[j].vertex_array); + for (const RendererRD::MeshStorage::MeshInstance::Surface &surface : mi->surfaces) { + if (surface.versions) { + for (uint32_t j = 0; j < surface.version_count; j++) { + RD::get_singleton()->free(surface.versions[j].vertex_array); } - memfree(mi->surfaces[i].versions); + memfree(surface.versions); } - if (mi->surfaces[i].vertex_buffer.is_valid()) { - RD::get_singleton()->free(mi->surfaces[i].vertex_buffer); + if (surface.vertex_buffer.is_valid()) { + RD::get_singleton()->free(surface.vertex_buffer); } } mi->surfaces.clear(); @@ -866,8 +868,8 @@ void MeshStorage::_mesh_instance_clear(MeshInstance *mi) { void MeshStorage::_mesh_instance_add_surface(MeshInstance *mi, Mesh *mesh, uint32_t p_surface) { if (mesh->blend_shape_count > 0 && mi->blend_weights_buffer.is_null()) { mi->blend_weights.resize(mesh->blend_shape_count); - for (uint32_t i = 0; i < mi->blend_weights.size(); i++) { - mi->blend_weights[i] = 0; + for (float &weight : mi->blend_weights) { + weight = 0; } mi->blend_weights_buffer = RD::get_singleton()->storage_buffer_create(sizeof(float) * mi->blend_weights.size(), mi->blend_weights.to_byte_array()); mi->weights_dirty = true; @@ -930,6 +932,11 @@ void MeshStorage::mesh_instance_check_for_update(RID p_mesh_instance) { } } +void MeshStorage::mesh_instance_set_canvas_item_transform(RID p_mesh_instance, const Transform2D &p_transform) { + MeshInstance *mi = mesh_instance_owner.get_or_null(p_mesh_instance); + mi->canvas_item_transform_2d = p_transform; +} + void MeshStorage::update_mesh_instances() { while (dirty_mesh_instance_weights.first()) { MeshInstance *mi = dirty_mesh_instance_weights.first()->self(); @@ -981,6 +988,25 @@ void MeshStorage::update_mesh_instances() { push_constant.skin_stride = (mi->mesh->surfaces[i]->skin_buffer_size / mi->mesh->surfaces[i]->vertex_count) / 4; push_constant.skin_weight_offset = (mi->mesh->surfaces[i]->format & RS::ARRAY_FLAG_USE_8_BONE_WEIGHTS) ? 4 : 2; + Transform2D transform = Transform2D(); + if (sk && sk->use_2d) { + transform = mi->canvas_item_transform_2d.affine_inverse() * sk->base_transform_2d; + } + push_constant.skeleton_transform_x[0] = transform.columns[0][0]; + push_constant.skeleton_transform_x[1] = transform.columns[0][1]; + push_constant.skeleton_transform_y[0] = transform.columns[1][0]; + push_constant.skeleton_transform_y[1] = transform.columns[1][1]; + push_constant.skeleton_transform_offset[0] = transform.columns[2][0]; + push_constant.skeleton_transform_offset[1] = transform.columns[2][1]; + + Transform2D inverse_transform = transform.affine_inverse(); + push_constant.inverse_transform_x[0] = inverse_transform.columns[0][0]; + push_constant.inverse_transform_x[1] = inverse_transform.columns[0][1]; + push_constant.inverse_transform_y[0] = inverse_transform.columns[1][0]; + push_constant.inverse_transform_y[1] = inverse_transform.columns[1][1]; + push_constant.inverse_transform_offset[0] = inverse_transform.columns[2][0]; + push_constant.inverse_transform_offset[1] = inverse_transform.columns[2][1]; + push_constant.blend_shape_count = mi->mesh->blend_shape_count; push_constant.normalized_blend_shapes = mi->mesh->blend_shape_mode == RS::BLEND_SHAPE_MODE_NORMALIZED; push_constant.pad0 = 0; @@ -1779,8 +1805,12 @@ void MeshStorage::multimesh_set_visible_instances(RID p_multimesh, int p_visible } if (multimesh->data_cache.size()) { - //there is a data cache.. + // There is a data cache, but we may need to update some sections. _multimesh_mark_all_dirty(multimesh, false, true); + int start = multimesh->visible_instances >= 0 ? multimesh->visible_instances : multimesh->instances; + for (int i = start; i < p_visible; i++) { + _multimesh_mark_dirty(multimesh, i, true); + } } multimesh->visible_instances = p_visible; diff --git a/servers/rendering/renderer_rd/storage_rd/mesh_storage.h b/servers/rendering/renderer_rd/storage_rd/mesh_storage.h index b62da5fd7b..c921523941 100644 --- a/servers/rendering/renderer_rd/storage_rd/mesh_storage.h +++ b/servers/rendering/renderer_rd/storage_rd/mesh_storage.h @@ -178,6 +178,7 @@ private: bool weights_dirty = false; SelfList<MeshInstance> weight_update_list; SelfList<MeshInstance> array_update_list; + Transform2D canvas_item_transform_2d; MeshInstance() : weight_update_list(this), array_update_list(this) {} }; @@ -256,6 +257,14 @@ private: uint32_t normalized_blend_shapes; uint32_t pad0; uint32_t pad1; + float skeleton_transform_x[2]; + float skeleton_transform_y[2]; + + float skeleton_transform_offset[2]; + float inverse_transform_x[2]; + + float inverse_transform_y[2]; + float inverse_transform_offset[2]; }; enum { @@ -548,6 +557,7 @@ public: virtual void mesh_instance_set_skeleton(RID p_mesh_instance, RID p_skeleton) override; virtual void mesh_instance_set_blend_shape_weight(RID p_mesh_instance, int p_shape, float p_weight) override; virtual void mesh_instance_check_for_update(RID p_mesh_instance) override; + virtual void mesh_instance_set_canvas_item_transform(RID p_mesh_instance, const Transform2D &p_transform) override; virtual void update_mesh_instances() override; /* MULTIMESH API */ diff --git a/servers/rendering/renderer_rd/storage_rd/particles_storage.cpp b/servers/rendering/renderer_rd/storage_rd/particles_storage.cpp index e4149f6bbd..00fb8acca8 100644 --- a/servers/rendering/renderer_rd/storage_rd/particles_storage.cpp +++ b/servers/rendering/renderer_rd/storage_rd/particles_storage.cpp @@ -851,9 +851,9 @@ void ParticlesStorage::_particles_process(Particles *p_particles, double p_delta collision_heightmap_texture = p_particles->sdf_collision_texture; //replace in all other history frames where used because parameters are no longer valid if screen moves - for (uint32_t i = 1; i < p_particles->frame_history.size(); i++) { - if (p_particles->frame_history[i].collider_count > 0 && p_particles->frame_history[i].colliders[0].type == ParticlesFrameParams::COLLISION_TYPE_2D_SDF) { - p_particles->frame_history[i].colliders[0] = frame_params.colliders[0]; + for (ParticlesFrameParams ¶ms : p_particles->frame_history) { + if (params.collider_count > 0 && params.colliders[0].type == ParticlesFrameParams::COLLISION_TYPE_2D_SDF) { + params.colliders[0] = frame_params.colliders[0]; } } } @@ -1145,7 +1145,7 @@ void ParticlesStorage::particles_set_view_axis(RID p_particles, const Vector3 &p return; } - if (particles->particle_buffer.is_null()) { + if (particles->particle_buffer.is_null() || particles->trail_bind_pose_uniform_set.is_null()) { return; //particles have not processed yet } diff --git a/servers/rendering/renderer_rd/storage_rd/render_scene_buffers_rd.cpp b/servers/rendering/renderer_rd/storage_rd/render_scene_buffers_rd.cpp index 31377a10a0..f5d6404f01 100644 --- a/servers/rendering/renderer_rd/storage_rd/render_scene_buffers_rd.cpp +++ b/servers/rendering/renderer_rd/storage_rd/render_scene_buffers_rd.cpp @@ -50,25 +50,19 @@ void RenderSceneBuffersRD::_bind_methods() { // ClassDB::bind_method(D_METHOD("create_texture_view", "context", "name", "view_name", "view"), &RenderSceneBuffersRD::has_texture); ClassDB::bind_method(D_METHOD("get_texture", "context", "name"), &RenderSceneBuffersRD::get_texture); // ClassDB::bind_method(D_METHOD("get_texture_format", "context", "name"), &RenderSceneBuffersRD::get_texture_format); - ClassDB::bind_method(D_METHOD("get_texture_slice", "context", "name", "layer", "mipmap"), &RenderSceneBuffersRD::get_texture_slice); - ClassDB::bind_method(D_METHOD("get_texture_slice_size", "context", "name", "layer", "mipmap"), &RenderSceneBuffersRD::get_texture_slice_size); + ClassDB::bind_method(D_METHOD("get_texture_slice", "context", "name", "layer", "mipmap", "layers", "mipmaps"), &RenderSceneBuffersRD::get_texture_slice); + ClassDB::bind_method(D_METHOD("get_texture_slice_size", "context", "name", "mipmap"), &RenderSceneBuffersRD::get_texture_slice_size); ClassDB::bind_method(D_METHOD("clear_context", "context"), &RenderSceneBuffersRD::clear_context); } void RenderSceneBuffersRD::update_sizes(NamedTexture &p_named_texture) { ERR_FAIL_COND(p_named_texture.texture.is_null()); - uint32_t size = p_named_texture.format.array_layers * p_named_texture.format.mipmaps; - p_named_texture.sizes.resize(size); + p_named_texture.sizes.resize(p_named_texture.format.mipmaps); Size2i mipmap_size = Size2i(p_named_texture.format.width, p_named_texture.format.height); - for (uint32_t mipmap = 0; mipmap < p_named_texture.format.mipmaps; mipmap++) { - for (uint32_t layer = 0; layer < p_named_texture.format.array_layers; layer++) { - uint32_t index = layer * p_named_texture.format.mipmaps + mipmap; - - p_named_texture.sizes.ptrw()[index] = mipmap_size; - } + p_named_texture.sizes.ptrw()[mipmap] = mipmap_size; mipmap_size.width = MAX(1, mipmap_size.width >> 1); mipmap_size.height = MAX(1, mipmap_size.height >> 1); @@ -251,7 +245,7 @@ RID RenderSceneBuffersRD::create_texture(const StringName &p_context, const Stri } RID RenderSceneBuffersRD::create_texture_from_format(const StringName &p_context, const StringName &p_texture_name, const RD::TextureFormat &p_texture_format, RD::TextureView p_view, bool p_unique) { - // TODO p_unique, if p_unique is true, this is a texture that can be shared. This will be implemented later as an optimisation. + // TODO p_unique, if p_unique is true, this is a texture that can be shared. This will be implemented later as an optimization. NTKey key(p_context, p_texture_name); @@ -324,7 +318,7 @@ const RD::TextureFormat RenderSceneBuffersRD::get_texture_format(const StringNam return named_textures[key].format; } -RID RenderSceneBuffersRD::get_texture_slice(const StringName &p_context, const StringName &p_texture_name, const uint32_t p_layer, const uint32_t p_mipmap) { +RID RenderSceneBuffersRD::get_texture_slice(const StringName &p_context, const StringName &p_texture_name, const uint32_t p_layer, const uint32_t p_mipmap, const uint32_t p_layers, const uint32_t p_mipmaps) { NTKey key(p_context, p_texture_name); // check if this is a known texture @@ -334,36 +328,41 @@ RID RenderSceneBuffersRD::get_texture_slice(const StringName &p_context, const S // check if we're in bounds ERR_FAIL_UNSIGNED_INDEX_V(p_layer, named_texture.format.array_layers, RID()); + ERR_FAIL_COND_V(p_layers == 0, RID()); + ERR_FAIL_COND_V(p_layer + p_layers > named_texture.format.array_layers, RID()); ERR_FAIL_UNSIGNED_INDEX_V(p_mipmap, named_texture.format.mipmaps, RID()); + ERR_FAIL_COND_V(p_mipmaps == 0, RID()); + ERR_FAIL_COND_V(p_mipmap + p_mipmaps > named_texture.format.mipmaps, RID()); - // if we don't have multiple layers or mipmaps, we can just return our texture as is - if (named_texture.format.array_layers == 1 && named_texture.format.mipmaps == 1) { + // asking the whole thing? just return the original + if (p_layer == 0 && p_mipmap == 0 && named_texture.format.array_layers == p_layers && named_texture.format.mipmaps == p_mipmaps) { return named_texture.texture; } - // get our index and make sure we have enough entries in our slices vector - uint32_t index = p_layer * named_texture.format.mipmaps + p_mipmap; - while (named_texture.slices.size() <= int(index)) { - named_texture.slices.push_back(RID()); + // see if we have this + NTSliceKey slice_key(p_layer, p_layers, p_mipmap, p_mipmaps); + if (named_texture.slices.has(slice_key)) { + return named_texture.slices[slice_key]; } - // create our slice if we don't have it already - if (named_texture.slices[index].is_null()) { - named_texture.slices.ptrw()[index] = RD::get_singleton()->texture_create_shared_from_slice(RD::TextureView(), named_texture.texture, p_layer, p_mipmap); + // create our slice + RID &slice = named_texture.slices[slice_key]; + slice = RD::get_singleton()->texture_create_shared_from_slice(RD::TextureView(), named_texture.texture, p_layer, p_mipmap, p_mipmaps, p_layers > 1 ? RD::TEXTURE_SLICE_2D_ARRAY : RD::TEXTURE_SLICE_2D, p_layers); - Array arr; - arr.push_back(p_context); - arr.push_back(p_texture_name); - arr.push_back(itos(p_layer)); - arr.push_back(itos(p_mipmap)); - RD::get_singleton()->set_resource_name(named_texture.slices[index], String("RenderBuffer {0}/{1} slice {2}/{3}").format(arr)); - } + Array arr; + arr.push_back(p_context); + arr.push_back(p_texture_name); + arr.push_back(itos(p_layer)); + arr.push_back(itos(p_layers)); + arr.push_back(itos(p_mipmap)); + arr.push_back(itos(p_mipmaps)); + RD::get_singleton()->set_resource_name(slice, String("RenderBuffer {0}/{1}, layer {2}/{3}, mipmap {4}/{5}").format(arr)); // and return our slice - return named_texture.slices[index]; + return slice; } -Size2i RenderSceneBuffersRD::get_texture_slice_size(const StringName &p_context, const StringName &p_texture_name, const uint32_t p_layer, const uint32_t p_mipmap) { +Size2i RenderSceneBuffersRD::get_texture_slice_size(const StringName &p_context, const StringName &p_texture_name, const uint32_t p_mipmap) { NTKey key(p_context, p_texture_name); // check if this is a known texture @@ -372,14 +371,10 @@ Size2i RenderSceneBuffersRD::get_texture_slice_size(const StringName &p_context, ERR_FAIL_COND_V(named_texture.texture.is_null(), Size2i()); // check if we're in bounds - ERR_FAIL_UNSIGNED_INDEX_V(p_layer, named_texture.format.array_layers, Size2i()); ERR_FAIL_UNSIGNED_INDEX_V(p_mipmap, named_texture.format.mipmaps, Size2i()); - // get our index - uint32_t index = p_layer * named_texture.format.mipmaps + p_mipmap; - - // and return our size - return named_texture.sizes[index]; + // return our size + return named_texture.sizes[p_mipmap]; } void RenderSceneBuffersRD::clear_context(const StringName &p_context) { @@ -490,6 +485,16 @@ Ref<RenderBufferCustomDataRD> RenderSceneBuffersRD::get_custom_data(const String // Depth texture +bool RenderSceneBuffersRD::has_depth_texture() { + RendererRD::TextureStorage *texture_storage = RendererRD::TextureStorage::get_singleton(); + RID depth = texture_storage->render_target_get_override_depth(render_target); + if (depth.is_valid()) { + return true; + } else { + return has_texture(RB_SCOPE_BUFFERS, RB_TEX_DEPTH); + } +} + RID RenderSceneBuffersRD::get_depth_texture() { RendererRD::TextureStorage *texture_storage = RendererRD::TextureStorage::get_singleton(); RID depth = texture_storage->render_target_get_override_depth(render_target); diff --git a/servers/rendering/renderer_rd/storage_rd/render_scene_buffers_rd.h b/servers/rendering/renderer_rd/storage_rd/render_scene_buffers_rd.h index dc849fd56a..9a299a3415 100644 --- a/servers/rendering/renderer_rd/storage_rd/render_scene_buffers_rd.h +++ b/servers/rendering/renderer_rd/storage_rd/render_scene_buffers_rd.h @@ -93,7 +93,6 @@ private: } static uint32_t hash(const NTKey &p_val) { - // FIXME, properly hash two stringnames together uint32_t h = p_val.context.hash(); h = hash_murmur3_one_32(p_val.buffer_name.hash(), h); return hash_fmix32(h); @@ -106,6 +105,33 @@ private: } }; + struct NTSliceKey { + uint32_t layer; + uint32_t layers; + uint32_t mipmap; + uint32_t mipmaps; + + bool operator==(const NTSliceKey &p_val) const { + return (layer == p_val.layer) && (layers == p_val.layers) && (mipmap == p_val.mipmap) && (mipmaps == p_val.mipmaps); + } + + static uint32_t hash(const NTSliceKey &p_val) { + uint32_t h = hash_murmur3_one_32(p_val.layer); + h = hash_murmur3_one_32(p_val.layers, h); + h = hash_murmur3_one_32(p_val.mipmap, h); + h = hash_murmur3_one_32(p_val.mipmaps, h); + return hash_fmix32(h); + } + + NTSliceKey() {} + NTSliceKey(uint32_t p_layer, uint32_t p_layers, uint32_t p_mipmap, uint32_t p_mipmaps) { + layer = p_layer; + layers = p_layers; + mipmap = p_mipmap; + mipmaps = p_mipmaps; + } + }; + struct NamedTexture { // Cache the data used to create our texture RD::TextureFormat format; @@ -113,7 +139,7 @@ private: // Our texture objects, slices are lazy (i.e. only created when requested). RID texture; - Vector<RID> slices; + mutable HashMap<NTSliceKey, RID, NTSliceKey> slices; Vector<Size2i> sizes; }; @@ -154,8 +180,8 @@ public: RID create_texture_view(const StringName &p_context, const StringName &p_texture_name, const StringName p_view_name, RD::TextureView p_view = RD::TextureView()); RID get_texture(const StringName &p_context, const StringName &p_texture_name) const; const RD::TextureFormat get_texture_format(const StringName &p_context, const StringName &p_texture_name) const; - RID get_texture_slice(const StringName &p_context, const StringName &p_texture_name, const uint32_t p_layer, const uint32_t p_mipmap); - Size2i get_texture_slice_size(const StringName &p_context, const StringName &p_texture_name, const uint32_t p_layer, const uint32_t p_mipmap); + RID get_texture_slice(const StringName &p_context, const StringName &p_texture_name, const uint32_t p_layer, const uint32_t p_mipmap, const uint32_t p_layers = 1, const uint32_t p_mipmaps = 1); + Size2i get_texture_slice_size(const StringName &p_context, const StringName &p_texture_name, const uint32_t p_mipmap); void clear_context(const StringName &p_context); @@ -185,6 +211,9 @@ public: // For our internal textures we provide some easy access methods. + _FORCE_INLINE_ bool has_internal_texture() const { + return has_texture(RB_SCOPE_BUFFERS, RB_TEX_COLOR); + } _FORCE_INLINE_ RID get_internal_texture() const { return get_texture(RB_SCOPE_BUFFERS, RB_TEX_COLOR); } @@ -192,6 +221,7 @@ public: return get_texture_slice(RB_SCOPE_BUFFERS, RB_TEX_COLOR, p_layer, 0); } + bool has_depth_texture(); RID get_depth_texture(); RID get_depth_texture(const uint32_t p_layer); diff --git a/servers/rendering/renderer_rd/storage_rd/texture_storage.cpp b/servers/rendering/renderer_rd/storage_rd/texture_storage.cpp index 06fda8fa9e..0ee9b28826 100644 --- a/servers/rendering/renderer_rd/storage_rd/texture_storage.cpp +++ b/servers/rendering/renderer_rd/storage_rd/texture_storage.cpp @@ -422,8 +422,7 @@ TextureStorage::TextureStorage() { tformat.usage_bits = RD::TEXTURE_USAGE_COLOR_ATTACHMENT_BIT | RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_STORAGE_BIT | RD::TEXTURE_USAGE_CAN_UPDATE_BIT | RD::TEXTURE_USAGE_VRS_ATTACHMENT_BIT; tformat.texture_type = RD::TEXTURE_TYPE_2D; if (!RD::get_singleton()->has_feature(RD::SUPPORTS_ATTACHMENT_VRS)) { - tformat.usage_bits = RD::TEXTURE_USAGE_COLOR_ATTACHMENT_BIT | RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_STORAGE_BIT | RD::TEXTURE_USAGE_CAN_UPDATE_BIT; - tformat.format = RD::DATA_FORMAT_R8_UNORM; + tformat.usage_bits = RD::TEXTURE_USAGE_COLOR_ATTACHMENT_BIT | RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_CAN_UPDATE_BIT; } Vector<uint8_t> pv; @@ -1854,10 +1853,10 @@ void TextureStorage::decal_free(RID p_rid) { decal_owner.free(p_rid); } -void TextureStorage::decal_set_extents(RID p_decal, const Vector3 &p_extents) { +void TextureStorage::decal_set_size(RID p_decal, const Vector3 &p_size) { Decal *decal = decal_owner.get_or_null(p_decal); ERR_FAIL_COND(!decal); - decal->extents = p_extents; + decal->size = p_size; decal->dependency.changed_notify(Dependency::DEPENDENCY_CHANGED_AABB); } @@ -1907,7 +1906,7 @@ void TextureStorage::decal_set_cull_mask(RID p_decal, uint32_t p_layers) { Decal *decal = decal_owner.get_or_null(p_decal); ERR_FAIL_COND(!decal); decal->cull_mask = p_layers; - decal->dependency.changed_notify(Dependency::DEPENDENCY_CHANGED_AABB); + decal->dependency.changed_notify(Dependency::DEPENDENCY_CHANGED_DECAL); } void TextureStorage::decal_set_distance_fade(RID p_decal, bool p_enabled, float p_begin, float p_length) { @@ -1950,7 +1949,14 @@ AABB TextureStorage::decal_get_aabb(RID p_decal) const { Decal *decal = decal_owner.get_or_null(p_decal); ERR_FAIL_COND_V(!decal, AABB()); - return AABB(-decal->extents, decal->extents * 2.0); + return AABB(-decal->size / 2, decal->size); +} + +uint32_t TextureStorage::decal_get_cull_mask(RID p_decal) const { + Decal *decal = decal_owner.get_or_null(p_decal); + ERR_FAIL_COND_V(!decal, 0); + + return decal->cull_mask; } Dependency *TextureStorage::decal_get_dependency(RID p_decal) { @@ -2206,6 +2212,12 @@ void TextureStorage::decal_instance_set_transform(RID p_decal_instance, const Tr di->transform = p_transform; } +void TextureStorage::decal_instance_set_sorting_offset(RID p_decal_instance, float p_sorting_offset) { + DecalInstance *di = decal_instance_owner.get_or_null(p_decal_instance); + ERR_FAIL_COND(!di); + di->sorting_offset = p_sorting_offset; +} + /* DECAL DATA API */ void TextureStorage::free_decal_data() { @@ -2233,7 +2245,7 @@ void TextureStorage::set_max_decals(const uint32_t p_max_decals) { decal_buffer = RD::get_singleton()->storage_buffer_create(decal_buffer_size); } -void TextureStorage::update_decal_buffer(const PagedArray<RID> &p_decals, const Transform3D &p_camera_inverse_xform) { +void TextureStorage::update_decal_buffer(const PagedArray<RID> &p_decals, const Transform3D &p_camera_xform) { ForwardIDStorage *forward_id_storage = ForwardIDStorage::get_singleton(); Transform3D uv_xform; @@ -2257,7 +2269,7 @@ void TextureStorage::update_decal_buffer(const PagedArray<RID> &p_decals, const Transform3D xform = decal_instance->transform; - real_t distance = -p_camera_inverse_xform.xform(xform.origin).z; + real_t distance = p_camera_xform.origin.distance_to(xform.origin); if (decal->distance_fade) { float fade_begin = decal->distance_fade_begin; @@ -2272,7 +2284,7 @@ void TextureStorage::update_decal_buffer(const PagedArray<RID> &p_decals, const decal_sort[decal_count].decal_instance = decal_instance; decal_sort[decal_count].decal = decal; - decal_sort[decal_count].depth = distance; + decal_sort[decal_count].depth = distance - decal_instance->sorting_offset; decal_count++; } @@ -2292,11 +2304,10 @@ void TextureStorage::update_decal_buffer(const PagedArray<RID> &p_decals, const decal_instance->cull_mask = decal->cull_mask; - Transform3D xform = decal_instance->transform; float fade = 1.0; if (decal->distance_fade) { - const real_t distance = -p_camera_inverse_xform.xform(xform.origin).z; + const real_t distance = decal_sort[i].depth + decal_instance->sorting_offset; const float fade_begin = decal->distance_fade_begin; const float fade_length = decal->distance_fade_length; @@ -2308,15 +2319,20 @@ void TextureStorage::update_decal_buffer(const PagedArray<RID> &p_decals, const DecalData &dd = decals[i]; - Vector3 decal_extents = decal->extents; + Vector3 decal_extents = decal->size / 2; Transform3D scale_xform; scale_xform.basis.scale(decal_extents); - Transform3D to_decal_xform = (p_camera_inverse_xform * xform * scale_xform * uv_xform).affine_inverse(); + + Transform3D xform = decal_instance->transform; + + Transform3D camera_inverse_xform = p_camera_xform.affine_inverse(); + + Transform3D to_decal_xform = (camera_inverse_xform * xform * scale_xform * uv_xform).affine_inverse(); MaterialStorage::store_transform(to_decal_xform, dd.xform); Vector3 normal = xform.basis.get_column(Vector3::AXIS_Y).normalized(); - normal = p_camera_inverse_xform.basis.xform(normal); //camera is normalized, so fine + normal = camera_inverse_xform.basis.xform(normal); //camera is normalized, so fine dd.normal[0] = normal.x; dd.normal[1] = normal.y; @@ -2350,7 +2366,7 @@ void TextureStorage::update_decal_buffer(const PagedArray<RID> &p_decals, const dd.normal_rect[2] = rect.size.x; dd.normal_rect[3] = rect.size.y; - Basis normal_xform = p_camera_inverse_xform.basis * xform.basis.orthonormalized(); + Basis normal_xform = camera_inverse_xform.basis * xform.basis.orthonormalized(); MaterialStorage::store_basis_3x4(normal_xform, dd.normal_xform); } else { dd.normal_rect[0] = 0; diff --git a/servers/rendering/renderer_rd/storage_rd/texture_storage.h b/servers/rendering/renderer_rd/storage_rd/texture_storage.h index 1558342c3b..c16f5274ad 100644 --- a/servers/rendering/renderer_rd/storage_rd/texture_storage.h +++ b/servers/rendering/renderer_rd/storage_rd/texture_storage.h @@ -235,7 +235,7 @@ private: } decal_atlas; struct Decal { - Vector3 extents = Vector3(1, 1, 1); + Vector3 size = Vector3(2, 2, 2); RID textures[RS::DECAL_TEXTURE_MAX]; float emission_energy = 1.0; float albedo_mix = 1.0; @@ -258,6 +258,7 @@ private: struct DecalInstance { RID decal; Transform3D transform; + float sorting_offset = 0.0; uint32_t cull_mask = 0; RendererRD::ForwardID forward_id = -1; }; @@ -560,7 +561,7 @@ public: virtual void decal_initialize(RID p_decal) override; virtual void decal_free(RID p_rid) override; - virtual void decal_set_extents(RID p_decal, const Vector3 &p_extents) override; + virtual void decal_set_size(RID p_decal, const Vector3 &p_size) override; virtual void decal_set_texture(RID p_decal, RS::DecalTexture p_type, RID p_texture) override; virtual void decal_set_emission_energy(RID p_decal, float p_energy) override; virtual void decal_set_albedo_mix(RID p_decal, float p_mix) override; @@ -576,9 +577,9 @@ public: virtual void texture_add_to_decal_atlas(RID p_texture, bool p_panorama_to_dp = false) override; virtual void texture_remove_from_decal_atlas(RID p_texture, bool p_panorama_to_dp = false) override; - _FORCE_INLINE_ Vector3 decal_get_extents(RID p_decal) { + _FORCE_INLINE_ Vector3 decal_get_size(RID p_decal) { const Decal *decal = decal_owner.get_or_null(p_decal); - return decal->extents; + return decal->size; } _FORCE_INLINE_ RID decal_get_texture(RID p_decal, RS::DecalTexture p_texture) { @@ -637,6 +638,7 @@ public: } virtual AABB decal_get_aabb(RID p_decal) const override; + virtual uint32_t decal_get_cull_mask(RID p_decal) const override; Dependency *decal_get_dependency(RID p_decal); /* DECAL INSTANCE API */ @@ -646,6 +648,7 @@ public: virtual RID decal_instance_create(RID p_decal) override; virtual void decal_instance_free(RID p_decal_instance) override; virtual void decal_instance_set_transform(RID p_decal_instance, const Transform3D &p_transform) override; + virtual void decal_instance_set_sorting_offset(RID p_decal_instance, float p_sorting_offset) override; _FORCE_INLINE_ RID decal_instance_get_base(RID p_decal_instance) const { DecalInstance *di = decal_instance_owner.get_or_null(p_decal_instance); @@ -677,7 +680,7 @@ public: void free_decal_data(); void set_max_decals(const uint32_t p_max_decals); RID get_decal_buffer() { return decal_buffer; } - void update_decal_buffer(const PagedArray<RID> &p_decals, const Transform3D &p_camera_inverse_xform); + void update_decal_buffer(const PagedArray<RID> &p_decals, const Transform3D &p_camera_xform); /* RENDER TARGET API */ diff --git a/servers/rendering/renderer_rd/storage_rd/utilities.cpp b/servers/rendering/renderer_rd/storage_rd/utilities.cpp index d2f5e6f224..cabac4e9ee 100644 --- a/servers/rendering/renderer_rd/storage_rd/utilities.cpp +++ b/servers/rendering/renderer_rd/storage_rd/utilities.cpp @@ -286,6 +286,10 @@ bool Utilities::has_os_feature(const String &p_feature) const { return true; } + if (p_feature == "astc" && RD::get_singleton()->texture_is_format_supported_for_usage(RD::DATA_FORMAT_ASTC_4x4_UNORM_BLOCK, RD::TEXTURE_USAGE_SAMPLING_BIT)) { + return true; + } + return false; } |