diff options
Diffstat (limited to 'servers/rendering')
35 files changed, 429 insertions, 179 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..a13e6c8172 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); @@ -319,16 +332,16 @@ public: _FORCE_INLINE_ void add_box(BoxType p_box_type, const Transform3D &p_transform, const Vector3 &p_half_extents) { 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 + // Extract scale and scale the matrix by it, makes things simpler. Vector3 scale = p_half_extents; for (uint32_t i = 0; i < 3; i++) { float s = xform.basis.rows[i].length(); @@ -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/sky.cpp b/servers/rendering/renderer_rd/environment/sky.cpp index 7e02f98ce9..814d145b66 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"); @@ -1627,7 +1630,10 @@ 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 +1648,10 @@ 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 ad63d3fd4b..1bba95f511 100644 --- a/servers/rendering/renderer_rd/forward_clustered/render_forward_clustered.cpp +++ b/servers/rendering/renderer_rd/forward_clustered/render_forward_clustered.cpp @@ -3768,6 +3768,7 @@ 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); @@ -3782,6 +3783,7 @@ 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) { 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..2aaea4288b 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 @@ -717,6 +717,8 @@ 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 25204f1abf..4065d253bb 100644 --- a/servers/rendering/renderer_rd/forward_mobile/render_forward_mobile.cpp +++ b/servers/rendering/renderer_rd/forward_mobile/render_forward_mobile.cpp @@ -2237,6 +2237,9 @@ 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(); @@ -2673,6 +2676,7 @@ 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); @@ -2687,6 +2691,7 @@ 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 */ 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..f9eecf4dc7 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 @@ -609,6 +609,7 @@ 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_scene_render_rd.cpp b/servers/rendering/renderer_rd/renderer_scene_render_rd.cpp index 2a87c4fa8d..33f6ac0276 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(); @@ -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. 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/forward_clustered/scene_forward_clustered.glsl b/servers/rendering/renderer_rd/shaders/forward_clustered/scene_forward_clustered.glsl index d32e6d717f..c6a3c42e57 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 @@ -550,13 +550,13 @@ 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 @@ -1305,24 +1305,26 @@ 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 +1353,8 @@ 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 +1800,7 @@ 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 +2084,7 @@ 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..90e902ca33 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; @@ -523,9 +529,15 @@ 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 @@ -1515,6 +1527,8 @@ 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/storage_rd/light_storage.cpp b/servers/rendering/renderer_rd/storage_rd/light_storage.cpp index 33e35a7a64..cdecb3828b 100644 --- a/servers/rendering/renderer_rd/storage_rd/light_storage.cpp +++ b/servers/rendering/renderer_rd/storage_rd/light_storage.cpp @@ -1464,7 +1464,6 @@ 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)); } diff --git a/servers/rendering/renderer_rd/storage_rd/light_storage.h b/servers/rendering/renderer_rd/storage_rd/light_storage.h index f7c8b8833b..d359219128 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 extents = Vector3(10, 10, 10); Vector3 origin_offset; bool interior = false; bool box_projection = false; diff --git a/servers/rendering/renderer_rd/storage_rd/mesh_storage.cpp b/servers/rendering/renderer_rd/storage_rd/mesh_storage.cpp index 46a0d85ba8..2015a2fca0 100644 --- a/servers/rendering/renderer_rd/storage_rd/mesh_storage.cpp +++ b/servers/rendering/renderer_rd/storage_rd/mesh_storage.cpp @@ -842,7 +842,7 @@ void MeshStorage::mesh_instance_set_blend_shape_weight(RID p_mesh_instance, int } void MeshStorage::_mesh_instance_clear(MeshInstance *mi) { - for (const RendererRD::MeshStorage::MeshInstance::Surface surface : mi->surfaces) { + 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); diff --git a/servers/rendering/renderer_rd/storage_rd/particles_storage.cpp b/servers/rendering/renderer_rd/storage_rd/particles_storage.cpp index f4c3e769a8..00fb8acca8 100644 --- a/servers/rendering/renderer_rd/storage_rd/particles_storage.cpp +++ b/servers/rendering/renderer_rd/storage_rd/particles_storage.cpp @@ -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..255719183f 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 @@ -251,7 +251,7 @@ 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); @@ -490,6 +490,16 @@ 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..50fb06c0d9 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 @@ -185,6 +185,9 @@ 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 +195,7 @@ 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 e8d9f486bb..2eaa7824fb 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; diff --git a/servers/rendering/shader_compiler.cpp b/servers/rendering/shader_compiler.cpp index 68542f32af..2710724066 100644 --- a/servers/rendering/shader_compiler.cpp +++ b/servers/rendering/shader_compiler.cpp @@ -1173,9 +1173,10 @@ String ShaderCompiler::_dump_node_code(const SL::Node *p_node, int p_level, Gene code += "("; - // if normal roughness texture is used, we will add logic to automatically switch between + // if color backbuffer, depth backbuffer or normal roughness texture is used, + // we will add logic to automatically switch between // sampler2D and sampler2D array and vec2 UV and vec3 UV. - bool normal_roughness_texture_used = false; + bool multiview_uv_needed = false; for (int i = 1; i < onode->arguments.size(); i++) { if (i > 1) { @@ -1220,8 +1221,8 @@ String ShaderCompiler::_dump_node_code(const SL::Node *p_node, int p_level, Gene } String node_code = _dump_node_code(onode->arguments[i], p_level, r_gen_code, p_actions, p_default_actions, p_assigning); - if (!RS::get_singleton()->is_low_end() && is_texture_func && i == 1) { - //need to map from texture to sampler in order to sample when using Vulkan GLSL + if (is_texture_func && i == 1) { + // If we're doing a texture lookup we need to check our texture argument StringName texture_uniform; bool correct_texture_uniform = false; @@ -1240,8 +1241,10 @@ String ShaderCompiler::_dump_node_code(const SL::Node *p_node, int p_level, Gene break; } - if (correct_texture_uniform) { + if (correct_texture_uniform && !RS::get_singleton()->is_low_end()) { + // Need to map from texture to sampler in order to sample when using Vulkan GLSL. String sampler_name; + bool is_depth_texture = false; bool is_normal_roughness_texture = false; if (actions.custom_samplers.has(texture_uniform)) { @@ -1251,6 +1254,8 @@ String ShaderCompiler::_dump_node_code(const SL::Node *p_node, int p_level, Gene const ShaderLanguage::ShaderNode::Uniform &u = shader->uniforms[texture_uniform]; if (u.hint == ShaderLanguage::ShaderNode::Uniform::HINT_SCREEN_TEXTURE) { is_screen_texture = true; + } else if (u.hint == ShaderLanguage::ShaderNode::Uniform::HINT_DEPTH_TEXTURE) { + is_depth_texture = true; } else if (u.hint == ShaderLanguage::ShaderNode::Uniform::HINT_NORMAL_ROUGHNESS_TEXTURE) { is_normal_roughness_texture = true; } @@ -1281,24 +1286,40 @@ String ShaderCompiler::_dump_node_code(const SL::Node *p_node, int p_level, Gene } String data_type_name = ""; - if (is_normal_roughness_texture) { + if (actions.check_multiview_samplers && (is_screen_texture || is_depth_texture || is_normal_roughness_texture)) { data_type_name = "multiviewSampler"; - normal_roughness_texture_used = true; + multiview_uv_needed = true; } else { data_type_name = ShaderLanguage::get_datatype_name(onode->arguments[i]->get_datatype()); } code += data_type_name + "(" + node_code + ", " + sampler_name + ")"; + } else if (actions.check_multiview_samplers && correct_texture_uniform && RS::get_singleton()->is_low_end()) { + // Texture function on low end hardware (i.e. OpenGL). + // We just need to know if the texture supports multiview. + + if (shader->uniforms.has(texture_uniform)) { + const ShaderLanguage::ShaderNode::Uniform &u = shader->uniforms[texture_uniform]; + if (u.hint == ShaderLanguage::ShaderNode::Uniform::HINT_SCREEN_TEXTURE) { + multiview_uv_needed = true; + } else if (u.hint == ShaderLanguage::ShaderNode::Uniform::HINT_DEPTH_TEXTURE) { + multiview_uv_needed = true; + } else if (u.hint == ShaderLanguage::ShaderNode::Uniform::HINT_NORMAL_ROUGHNESS_TEXTURE) { + multiview_uv_needed = true; + } + } + + code += node_code; } else { code += node_code; } - } else { - if (normal_roughness_texture_used && i == 2) { - // UV coordinate after using normal roughness texture. - node_code = "normal_roughness_uv(" + node_code + ".xy)"; - } + } else if (multiview_uv_needed && i == 2) { + // UV coordinate after using color, depth or normal roughness texture. + node_code = "multiview_uv(" + node_code + ".xy)"; code += node_code; + } else { + code += node_code; } } code += ")"; diff --git a/servers/rendering/shader_compiler.h b/servers/rendering/shader_compiler.h index 43bea213da..3bb29a545b 100644 --- a/servers/rendering/shader_compiler.h +++ b/servers/rendering/shader_compiler.h @@ -101,6 +101,7 @@ public: String instance_uniform_index_variable; uint32_t base_varying_index = 0; bool apply_luminance_multiplier = false; + bool check_multiview_samplers = false; }; private: diff --git a/servers/rendering/shader_language.cpp b/servers/rendering/shader_language.cpp index ba7ffda00a..7a4ac709f4 100644 --- a/servers/rendering/shader_language.cpp +++ b/servers/rendering/shader_language.cpp @@ -90,8 +90,9 @@ const char *ShaderLanguage::token_names[TK_MAX] = { "IDENTIFIER", "TRUE", "FALSE", - "REAL_CONSTANT", + "FLOAT_CONSTANT", "INT_CONSTANT", + "UINT_CONSTANT", "TYPE_VOID", "TYPE_BOOL", "TYPE_BVEC2", @@ -126,6 +127,7 @@ const char *ShaderLanguage::token_names[TK_MAX] = { "INTERPOLATION_FLAT", "INTERPOLATION_SMOOTH", "CONST", + "STRUCT", "PRECISION_LOW", "PRECISION_MID", "PRECISION_HIGH", @@ -169,6 +171,7 @@ const char *ShaderLanguage::token_names[TK_MAX] = { "CF_DO", "CF_SWITCH", "CF_CASE", + "CF_DEFAULT", "CF_BREAK", "CF_CONTINUE", "CF_RETURN", @@ -185,19 +188,26 @@ const char *ShaderLanguage::token_names[TK_MAX] = { "SEMICOLON", "PERIOD", "UNIFORM", + "UNIFORM_GROUP", "INSTANCE", "GLOBAL", "VARYING", - "IN", - "OUT", - "INOUT", + "ARG_IN", + "ARG_OUT", + "ARG_INOUT", "RENDER_MODE", - "SOURCE_COLOR", "HINT_DEFAULT_WHITE_TEXTURE", "HINT_DEFAULT_BLACK_TEXTURE", "HINT_DEFAULT_TRANSPARENT_TEXTURE", "HINT_NORMAL_TEXTURE", + "HINT_ROUGHNESS_NORMAL_TEXTURE", + "HINT_ROUGHNESS_R", + "HINT_ROUGHNESS_G", + "HINT_ROUGHNESS_B", + "HINT_ROUGHNESS_A", + "HINT_ROUGHNESS_GRAY", "HINT_ANISOTROPY_TEXTURE", + "HINT_SOURCE_COLOR", "HINT_RANGE", "HINT_INSTANCE_INDEX", "HINT_SCREEN_TEXTURE", @@ -612,7 +622,7 @@ ShaderLanguage::Token ShaderLanguage::_get_token() { char_idx += 2; include_positions.resize(include_positions.size() - 1); // Pop back. - tk_line = include_positions[include_positions.size() - 1].line; // Restore line. + tk_line = include_positions[include_positions.size() - 1].line - 1; // Restore line. } else { return _make_token(TK_ERROR, "Invalid include enter/exit hint token (@@> and @@<)"); @@ -2827,6 +2837,20 @@ const ShaderLanguage::BuiltinFuncDef ShaderLanguage::builtin_func_defs[] = { { "dFdx", TYPE_VEC3, { TYPE_VEC3, TYPE_VOID }, { "p" }, TAG_GLOBAL, false }, { "dFdx", TYPE_VEC4, { TYPE_VEC4, TYPE_VOID }, { "p" }, TAG_GLOBAL, false }, + // dFdxCoarse + + { "dFdxCoarse", TYPE_FLOAT, { TYPE_FLOAT, TYPE_VOID }, { "p" }, TAG_GLOBAL, true }, + { "dFdxCoarse", TYPE_VEC2, { TYPE_VEC2, TYPE_VOID }, { "p" }, TAG_GLOBAL, true }, + { "dFdxCoarse", TYPE_VEC3, { TYPE_VEC3, TYPE_VOID }, { "p" }, TAG_GLOBAL, true }, + { "dFdxCoarse", TYPE_VEC4, { TYPE_VEC4, TYPE_VOID }, { "p" }, TAG_GLOBAL, true }, + + // dFdxFine + + { "dFdxFine", TYPE_FLOAT, { TYPE_FLOAT, TYPE_VOID }, { "p" }, TAG_GLOBAL, true }, + { "dFdxFine", TYPE_VEC2, { TYPE_VEC2, TYPE_VOID }, { "p" }, TAG_GLOBAL, true }, + { "dFdxFine", TYPE_VEC3, { TYPE_VEC3, TYPE_VOID }, { "p" }, TAG_GLOBAL, true }, + { "dFdxFine", TYPE_VEC4, { TYPE_VEC4, TYPE_VOID }, { "p" }, TAG_GLOBAL, true }, + // dFdy { "dFdy", TYPE_FLOAT, { TYPE_FLOAT, TYPE_VOID }, { "p" }, TAG_GLOBAL, false }, @@ -2834,6 +2858,20 @@ const ShaderLanguage::BuiltinFuncDef ShaderLanguage::builtin_func_defs[] = { { "dFdy", TYPE_VEC3, { TYPE_VEC3, TYPE_VOID }, { "p" }, TAG_GLOBAL, false }, { "dFdy", TYPE_VEC4, { TYPE_VEC4, TYPE_VOID }, { "p" }, TAG_GLOBAL, false }, + // dFdyCoarse + + { "dFdyCoarse", TYPE_FLOAT, { TYPE_FLOAT, TYPE_VOID }, { "p" }, TAG_GLOBAL, true }, + { "dFdyCoarse", TYPE_VEC2, { TYPE_VEC2, TYPE_VOID }, { "p" }, TAG_GLOBAL, true }, + { "dFdyCoarse", TYPE_VEC3, { TYPE_VEC3, TYPE_VOID }, { "p" }, TAG_GLOBAL, true }, + { "dFdyCoarse", TYPE_VEC4, { TYPE_VEC4, TYPE_VOID }, { "p" }, TAG_GLOBAL, true }, + + // dFdyFine + + { "dFdyFine", TYPE_FLOAT, { TYPE_FLOAT, TYPE_VOID }, { "p" }, TAG_GLOBAL, true }, + { "dFdyFine", TYPE_VEC2, { TYPE_VEC2, TYPE_VOID }, { "p" }, TAG_GLOBAL, true }, + { "dFdyFine", TYPE_VEC3, { TYPE_VEC3, TYPE_VOID }, { "p" }, TAG_GLOBAL, true }, + { "dFdyFine", TYPE_VEC4, { TYPE_VEC4, TYPE_VOID }, { "p" }, TAG_GLOBAL, true }, + // fwidth { "fwidth", TYPE_FLOAT, { TYPE_FLOAT, TYPE_VOID }, { "p" }, TAG_GLOBAL, false }, @@ -2841,6 +2879,20 @@ const ShaderLanguage::BuiltinFuncDef ShaderLanguage::builtin_func_defs[] = { { "fwidth", TYPE_VEC3, { TYPE_VEC3, TYPE_VOID }, { "p" }, TAG_GLOBAL, false }, { "fwidth", TYPE_VEC4, { TYPE_VEC4, TYPE_VOID }, { "p" }, TAG_GLOBAL, false }, + // fwidthCoarse + + { "fwidthCoarse", TYPE_FLOAT, { TYPE_FLOAT, TYPE_VOID }, { "p" }, TAG_GLOBAL, true }, + { "fwidthCoarse", TYPE_VEC2, { TYPE_VEC2, TYPE_VOID }, { "p" }, TAG_GLOBAL, true }, + { "fwidthCoarse", TYPE_VEC3, { TYPE_VEC3, TYPE_VOID }, { "p" }, TAG_GLOBAL, true }, + { "fwidthCoarse", TYPE_VEC4, { TYPE_VEC4, TYPE_VOID }, { "p" }, TAG_GLOBAL, true }, + + // fwidthFine + + { "fwidthFine", TYPE_FLOAT, { TYPE_FLOAT, TYPE_VOID }, { "p" }, TAG_GLOBAL, true }, + { "fwidthFine", TYPE_VEC2, { TYPE_VEC2, TYPE_VOID }, { "p" }, TAG_GLOBAL, true }, + { "fwidthFine", TYPE_VEC3, { TYPE_VEC3, TYPE_VOID }, { "p" }, TAG_GLOBAL, true }, + { "fwidthFine", TYPE_VEC4, { TYPE_VEC4, TYPE_VOID }, { "p" }, TAG_GLOBAL, true }, + // Sub-functions. // array @@ -8233,6 +8285,10 @@ Error ShaderLanguage::_parse_shader(const HashMap<StringName, FunctionInfo> &p_f _set_error(vformat(RTR("Uniform instances are not yet implemented for '%s' shaders."), shader_type_identifier)); return ERR_PARSE_ERROR; } + if (OS::get_singleton()->get_current_rendering_method() == "gl_compatibility") { + _set_error(RTR("Uniform instances are not supported in gl_compatibility shaders.")); + return ERR_PARSE_ERROR; + } if (uniform_scope == ShaderNode::Uniform::SCOPE_LOCAL) { tk = _get_token(); if (tk.type != TK_UNIFORM) { diff --git a/servers/rendering/shader_preprocessor.cpp b/servers/rendering/shader_preprocessor.cpp index 97efd2df94..816d937e9f 100644 --- a/servers/rendering/shader_preprocessor.cpp +++ b/servers/rendering/shader_preprocessor.cpp @@ -78,19 +78,46 @@ char32_t ShaderPreprocessor::Tokenizer::peek() { return 0; } +int ShaderPreprocessor::Tokenizer::consume_line_continuations(int p_offset) { + int skips = 0; + + for (int i = index + p_offset; i < size; i++) { + char32_t c = code[i]; + if (c == '\\') { + if (i + 1 < size && code[i + 1] == '\n') { + // This line ends with "\" and "\n" continuation. + add_generated(Token('\n', line)); + line++; + skips++; + + i = i + 2; + index = i; + } else { + break; + } + } else if (!is_whitespace(c)) { + break; + } + } + return skips; +} + LocalVector<ShaderPreprocessor::Token> ShaderPreprocessor::Tokenizer::advance(char32_t p_what) { LocalVector<ShaderPreprocessor::Token> tokens; while (index < size) { char32_t c = code[index++]; - - tokens.push_back(ShaderPreprocessor::Token(c, line)); + if (c == '\\' && consume_line_continuations(-1) > 0) { + continue; + } if (c == '\n') { add_generated(ShaderPreprocessor::Token('\n', line)); line++; } + tokens.push_back(ShaderPreprocessor::Token(c, line)); + if (c == p_what || c == 0) { return tokens; } @@ -104,6 +131,11 @@ void ShaderPreprocessor::Tokenizer::skip_whitespace() { } } +bool ShaderPreprocessor::Tokenizer::consume_empty_line() { + // Read until newline and return true if the content was all whitespace/empty. + return tokens_to_string(advance('\n')).strip_edges().size() == 0; +} + String ShaderPreprocessor::Tokenizer::get_identifier(bool *r_is_cursor, bool p_started) { if (r_is_cursor != nullptr) { *r_is_cursor = false; @@ -113,6 +145,10 @@ String ShaderPreprocessor::Tokenizer::get_identifier(bool *r_is_cursor, bool p_s while (true) { char32_t c = peek(); + if (c == '\\' && consume_line_continuations(0) > 0) { + continue; + } + if (is_char_end(c) || c == '(' || c == ')' || c == ',' || c == ';') { break; } @@ -146,8 +182,10 @@ String ShaderPreprocessor::Tokenizer::get_identifier(bool *r_is_cursor, bool p_s String ShaderPreprocessor::Tokenizer::peek_identifier() { const int original = index; + const int original_line = line; String id = get_identifier(); index = original; + line = original_line; return id; } @@ -485,7 +523,9 @@ void ShaderPreprocessor::process_else(Tokenizer *p_tokenizer) { state->previous_region->to_line = line - 1; } - p_tokenizer->advance('\n'); + if (!p_tokenizer->consume_empty_line()) { + set_error(RTR("Invalid else."), p_tokenizer->get_line()); + } bool skip = false; for (int i = 0; i < state->current_branch->conditions.size(); i++) { @@ -508,17 +548,21 @@ void ShaderPreprocessor::process_else(Tokenizer *p_tokenizer) { } void ShaderPreprocessor::process_endif(Tokenizer *p_tokenizer) { + const int line = p_tokenizer->get_line(); + state->condition_depth--; if (state->condition_depth < 0) { - set_error(RTR("Unmatched endif."), p_tokenizer->get_line()); + set_error(RTR("Unmatched endif."), line); return; } if (state->previous_region != nullptr) { - state->previous_region->to_line = p_tokenizer->get_line() - 1; + state->previous_region->to_line = line - 1; state->previous_region = state->previous_region->parent; } - p_tokenizer->advance('\n'); + if (!p_tokenizer->consume_empty_line()) { + set_error(RTR("Invalid endif."), line); + } state->current_branch = state->current_branch->parent; state->branches.pop_back(); @@ -574,12 +618,10 @@ void ShaderPreprocessor::process_ifdef(Tokenizer *p_tokenizer) { return; } - p_tokenizer->skip_whitespace(); - if (!is_char_end(p_tokenizer->peek())) { + if (!p_tokenizer->consume_empty_line()) { set_error(RTR("Invalid ifdef."), line); return; } - p_tokenizer->advance('\n'); bool success = state->defines.has(label); start_branch_condition(p_tokenizer, success); @@ -598,12 +640,10 @@ void ShaderPreprocessor::process_ifndef(Tokenizer *p_tokenizer) { return; } - p_tokenizer->skip_whitespace(); - if (!is_char_end(p_tokenizer->peek())) { + if (!p_tokenizer->consume_empty_line()) { set_error(RTR("Invalid ifndef."), line); return; } - p_tokenizer->advance('\n'); bool success = !state->defines.has(label); start_branch_condition(p_tokenizer, success); @@ -628,9 +668,8 @@ void ShaderPreprocessor::process_include(Tokenizer *p_tokenizer) { } } path = path.substr(0, path.length() - 1); - p_tokenizer->skip_whitespace(); - if (path.is_empty() || !is_char_end(p_tokenizer->peek())) { + if (path.is_empty() || !p_tokenizer->consume_empty_line()) { set_error(RTR("Invalid path."), line); return; } @@ -728,25 +767,24 @@ void ShaderPreprocessor::process_pragma(Tokenizer *p_tokenizer) { return; } - p_tokenizer->advance('\n'); + if (!p_tokenizer->consume_empty_line()) { + set_error(RTR("Invalid pragma directive."), line); + return; + } } void ShaderPreprocessor::process_undef(Tokenizer *p_tokenizer) { const int line = p_tokenizer->get_line(); const String label = p_tokenizer->get_identifier(); - if (label.is_empty() || !state->defines.has(label)) { - set_error(RTR("Invalid name."), line); - return; - } - - p_tokenizer->skip_whitespace(); - if (!is_char_end(p_tokenizer->peek())) { + if (label.is_empty() || !p_tokenizer->consume_empty_line()) { set_error(RTR("Invalid undef."), line); return; } - memdelete(state->defines[label]); - state->defines.erase(label); + if (state->defines.has(label)) { + memdelete(state->defines[label]); + state->defines.erase(label); + } } void ShaderPreprocessor::add_region(int p_line, bool p_enabled, Region *p_parent_region) { @@ -957,15 +995,57 @@ bool ShaderPreprocessor::expand_macros_once(const String &p_line, int p_line_num String body = define->body; if (define->arguments.size() > 0) { // Complex macro with arguments. - int args_start = index + key.length(); - int args_end = p_line.find(")", args_start); - if (args_start == -1 || args_end == -1) { - set_error(RTR("Missing macro argument parenthesis."), p_line_number); - return false; + + int args_start = -1; + int args_end = -1; + int brackets_open = 0; + Vector<String> args; + for (int i = index_start - 1; i < p_line.length(); i++) { + bool add_argument = false; + bool reached_end = false; + char32_t c = p_line[i]; + + if (c == '(') { + brackets_open++; + if (brackets_open == 1) { + args_start = i + 1; + args_end = -1; + } + } else if (c == ')') { + brackets_open--; + if (brackets_open == 0) { + args_end = i; + add_argument = true; + reached_end = true; + } + } else if (c == ',') { + if (brackets_open == 1) { + args_end = i; + add_argument = true; + } + } + + if (add_argument) { + if (args_start == -1 || args_end == -1) { + set_error(RTR("Invalid macro argument list."), p_line_number); + return false; + } + + String arg = p_line.substr(args_start, args_end - args_start).strip_edges(); + if (arg.is_empty()) { + set_error(RTR("Invalid macro argument."), p_line_number); + return false; + } + args.append(arg); + + args_start = args_end + 1; + } + + if (reached_end) { + break; + } } - String values = result.substr(args_start + 1, args_end - (args_start + 1)); - Vector<String> args = values.split(","); if (args.size() != define->arguments.size()) { set_error(RTR("Invalid macro argument count."), p_line_number); return false; @@ -987,9 +1067,6 @@ bool ShaderPreprocessor::expand_macros_once(const String &p_line, int p_line_num result = result.substr(0, index) + " " + body + " " + result.substr(args_end + 1, result.length()); } else { result = result.substr(0, index) + body + result.substr(index + key.length(), result.length() - (index + key.length())); - // Manually reset index_start to where the body value of the define finishes. - // This ensures we don't skip another instance of this macro in the string. - index_start = index + body.length() + 1; } r_expanded = result; diff --git a/servers/rendering/shader_preprocessor.h b/servers/rendering/shader_preprocessor.h index 6e5533c575..2b7d2c274e 100644 --- a/servers/rendering/shader_preprocessor.h +++ b/servers/rendering/shader_preprocessor.h @@ -93,11 +93,13 @@ private: int get_line() const; int get_index() const; char32_t peek(); + int consume_line_continuations(int p_offset); void get_and_clear_generated(Vector<Token> *r_out); void backtrack(char32_t p_what); LocalVector<Token> advance(char32_t p_what); void skip_whitespace(); + bool consume_empty_line(); String get_identifier(bool *r_is_cursor = nullptr, bool p_started = false); String peek_identifier(); Token get_token(); |