Skip to main content

WgpuRenderer

Struct WgpuRenderer 

Source
pub struct WgpuRenderer { /* private fields */ }
Expand description

wgpu backend that renders Scene commands into a presentation surface.

The renderer owns the GPU device, queue, surface, pipeline, and transient vertex buffer. Window creation stays outside the library so each application can choose its own windowing framework.

Implementations§

Source§

impl WgpuRenderer

Source

pub fn create_screen_primitive_batch( &self, primitives: &[ScreenPrimitive2d], budget: ScreenPrimitiveBudget, ) -> Result<ScreenPrimitiveBatch2d, ScreenPrimitiveError>

Uploads ordered analytic screen primitives under explicit residency/work limits. No draw or presentation occurs. The input slice is snapshotted.

Source

pub fn update_screen_primitive_batch( &self, batch: &mut ScreenPrimitiveBatch2d, primitives: &[ScreenPrimitive2d], ) -> Result<ScreenPrimitiveUploadReport, ScreenPrimitiveError>

Replaces the entire snapshot transactionally, reusing CPU/GPU capacity. Rejected input or allocation failure leaves the previous snapshot intact. Changed instance data invalidates all camera/viewport proofs. Clip-only changes preserve numeric proofs because they cannot change shader positions. GPU queue order protects prior submitted draws; no synchronous readback or wait is added.

Source

pub fn restore_screen_primitive_batch( &self, source: &ScreenPrimitiveBatch2d, ) -> Result<ScreenPrimitiveBatch2d, ScreenPrimitiveError>

Recreates a device-local batch from its complete retained CPU snapshot. The old batch remains valid for its old renderer and is not mutated.

Source§

impl WgpuRenderer

Source

pub async fn recover_device_and_surface( &mut self, ) -> Result<(), RendererInitError>

Requests a replacement logical device for the existing surface and recreates all renderer-owned transient pipelines and buffers.

Every external retained GPU resource deliberately keeps the previous renderer identity and is rejected until restored after this succeeds. Restore prepared and prepared-screen scenes, dynamic meshes, particle fields, scalar fields, 2D targets, and trails with their matching restore_* methods. Restore images before dependent image batches; glyph atlases before dependent glyph runs; and retained 3D meshes/scenes before their empty 3D target. Both 2D/3D targets and trails restore without prior pixels and must be redrawn. See each resource’s WgpuRenderer::restore_* method for its exact retained-state contract.

Replaced logical devices enter a bounded quarantine until this renderer is dropped. Some native swapchain drivers crash if a healthy previous device is destroyed immediately after the surface migrates to its replacement. Device recovery should therefore remain exceptional rather than a normal quality switch. Query WgpuRenderer::remaining_device_recoveries before attempting it.

Source§

impl WgpuRenderer

Source

pub fn diagnostics(&self) -> RendererDiagnostics<'_>

Returns allocation-free adapter, driver, DPI, and presentation metadata.

Record this alongside CPU preparation/upload timings and separately correlated asynchronous GPU timestamp samples. Metadata does not turn surface acquisition or a CPU interval into a GPU execution measurement.

Source§

impl WgpuRenderer

Source

pub fn restore_dynamic_mesh( &self, source: &DynamicMesh2d, ) -> Result<DynamicMesh2d, DynamicMeshError>

Recreates a dynamic mesh on this renderer from its retained CPU vertices.

Use this after recreating a renderer following device loss. The returned mesh belongs to this renderer and preserves the source mesh’s capacity so subsequent updates retain the same allocation behavior.

Source

pub fn create_dynamic_mesh( &self, vertices: &[DynamicVertex2d], ) -> Result<DynamicMesh2d, DynamicMeshError>

Creates mutable triangle-list geometry for a frequently changing visual.

The vertex count must be divisible by three. The mesh is owned by this renderer and may only be updated or drawn through it.

Source

pub fn create_dynamic_mesh_with_budget( &self, vertices: &[DynamicVertex2d], budget: DynamicMeshBudget, ) -> Result<DynamicMesh2d, DynamicMeshError>

Creates bounded mutable colored triangles for composition in a frame.

Validation and CPU reservation complete before any GPU allocation. The retained budget is also enforced by every later full or partial update.

Source

pub fn update_dynamic_mesh( &self, mesh: &mut DynamicMesh2d, vertices: &[DynamicVertex2d], ) -> Result<(), DynamicMeshError>

Replaces all dynamic mesh vertices, growing GPU capacity only when needed.

Source

pub fn update_dynamic_mesh_with_metrics( &self, mesh: &mut DynamicMesh2d, vertices: &[DynamicVertex2d], ) -> Result<DynamicMeshUpdateReport, DynamicMeshError>

Replaces all dynamic mesh vertices and returns CPU-side update metrics.

Source

pub fn update_dynamic_mesh_range( &self, mesh: &mut DynamicMesh2d, first_vertex: usize, vertices: &[DynamicVertex2d], ) -> Result<(), DynamicMeshError>

Replaces a contiguous vertex range without reallocating the mesh buffer.

The range must stay inside the current triangle list and its length must preserve triangle alignment, so callers cannot create partial triangles.

Source

pub fn update_dynamic_mesh_range_with_metrics( &self, mesh: &mut DynamicMesh2d, first_vertex: usize, vertices: &[DynamicVertex2d], ) -> Result<DynamicMeshUpdateReport, DynamicMeshError>

Replaces a triangle-aligned range and returns CPU-side update metrics.

Source

pub fn render_dynamic_mesh( &mut self, mesh: &DynamicMesh2d, background: Color, camera: &Camera2d, ) -> Result<RenderStatus, DynamicMeshRenderError>

Draws dynamic triangle-list geometry with a normalized clear color.

Every potentially visible triangle must remain fully inside the full surface clip volume and have a portable projected orientation. A hardware-clipped or ambiguous triangle returns DynamicMeshRenderError::Frame(RendererFrameError::InvalidGeometryTransform).

Source

pub fn render_dynamic_mesh_with_metrics( &mut self, mesh: &DynamicMesh2d, background: Color, camera: &Camera2d, ) -> Result<RenderReport, DynamicMeshRenderError>

Draws dynamic triangle-list geometry and returns per-frame CPU metrics.

Full-surface clip-volume crossing and association-dependent projected topology are rejected as DynamicMeshRenderError::Frame(RendererFrameError::InvalidGeometryTransform).

Source§

impl WgpuRenderer

Source

pub fn frame_cache_budget(&self) -> FrameCacheBudget

Returns renderer-owned idle composition cache limits.

Source

pub fn set_frame_cache_budget(&mut self, budget: FrameCacheBudget)

Changes limits and releases the old cache. In-flight resource retirement is handled by wgpu; the next frame warms storage under the new limits.

Source

pub fn clear_frame_cache(&mut self)

Releases cached composition allocations and GPU resource references.

Source

pub fn frame_cache_statistics(&self) -> FrameCacheStatistics

Returns idle retained bytes and the previous composed frame’s cache work. This does not include queue/backend allocations, source geometry, or LUTs.

Source§

impl WgpuRenderer

Source

pub fn begin_frame( &mut self, background: Color, budget: FrameBudget, ) -> Result<FrameComposer<'_>, FrameComposerError>

Begins one bounded heterogeneous frame with a single clear color.

Source§

impl WgpuRenderer

Source

pub fn create_glyph_atlas( &self, width: u32, height: u32, pixels: Vec<u8>, entries: Vec<GlyphAtlasEntry>, budget: GlyphAtlasBudget, ) -> Result<GlyphAtlas2d, GlyphError>

Creates a bounded glyph atlas from exact RGBA pixels and initial entries.

Source

pub fn upload_glyph( &self, atlas: &mut GlyphAtlas2d, entry: GlyphAtlasEntry, pixels: &[u8], ) -> Result<GlyphUploadReport, GlyphError>

Uploads or replaces one glyph rectangle after complete metadata preflight.

Reusing an identity requires the same rectangle. Changing atlas packing is an explicit new-atlas operation so existing retained runs cannot silently sample unrelated texels.

Source

pub fn create_glyph_run( &self, atlas: &GlyphAtlas2d, glyphs: Vec<PositionedGlyph2d>, budget: GlyphRunBudget, ) -> Result<GlyphRun2d, GlyphError>

Creates one retained instanced run from host-positioned glyph quads.

Source

pub fn update_glyph_run( &self, atlas: &GlyphAtlas2d, run: &mut GlyphRun2d, glyphs: &[PositionedGlyph2d], ) -> Result<GlyphRunUploadReport, GlyphError>

Updates host-shaped layout while reusing CPU and GPU capacity when possible.

Exact unchanged input uploads nothing; an empty slice clears the drawable run but retains capacity. The atlas must be the run’s original active-device atlas. All glyph sources, bounds and budget checks complete before changing drawable state. Synchronous rejection preserves the previous revision; device failure cannot roll back work already submitted to the queue.

Source

pub fn restore_glyph_atlas( &self, source: &GlyphAtlas2d, ) -> Result<GlyphAtlas2d, GlyphError>

Restores exact atlas pixels and metadata for this renderer generation.

Source

pub fn restore_glyph_run( &self, atlas: &GlyphAtlas2d, source: &GlyphRun2d, ) -> Result<GlyphRun2d, GlyphError>

Restores a run against the restored copy of its original atlas.

Source§

impl WgpuRenderer

Source

pub fn create_glyph_batch( &self, atlas: &GlyphAtlas2d, placements: &[GlyphRunPlacement2d<'_>], budget: GlyphBatchBudget, ) -> Result<GlyphBatch2d, GlyphError>

Snapshots same-atlas prepared placements without shaping or font work.

Foreign atlases/runs, overflowing destinations and count/byte/device-limit excess are rejected before GPU allocation. Input order is painter order.

Source

pub fn update_glyph_batch( &self, atlas: &GlyphAtlas2d, batch: &mut GlyphBatch2d, placements: &[GlyphRunPlacement2d<'_>], ) -> Result<GlyphBatchUploadReport, GlyphError>

Atomically replaces the snapshot, reusing CPU arrays and GPU capacity.

All runs must belong to the same current atlas. Synchronous validation or allocation failure preserves the old drawable batch. Source run mutations are not observed until this call. Clip-only changes do not upload glyphs; translation/tint changes update instance data without redoing text layout. Shrinking retains reusable capacity. GPU/device failure is not a rollback of queue work already submitted.

Source

pub fn restore_glyph_batch( &self, atlas: &GlyphAtlas2d, source: &GlyphBatch2d, ) -> Result<GlyphBatch2d, GlyphError>

Restores a snapshot against the restored copy of its original atlas.

The original runs need not remain alive. Translation, tint, clips and painter order are preserved. Another atlas, even with identical pixels, is rejected. The restored batch has no viewport-dependent cached proof.

Source§

impl WgpuRenderer

Source

pub fn update_image_batch( &self, image: &Image2d, batch: &mut ImageBatch2d, sprites: &[ImageSprite2d], ) -> Result<ImageBatchUploadReport, ImageError>

Atomically updates retained sprite descriptions from borrowed input.

Reuses both CPU arrays and the GPU buffer while capacity permits. Equal input submits no upload. Shrinking, including to zero sprites, retains capacity. Growth reserves exact bounded replacement arrays before changing drawable state. Synchronous validation/allocation errors preserve the old batch. Device failure is not a rollback of already submitted GPU work.

The original image and current renderer generation must match. The queue orders uploads after previously submitted draws; the frame borrow prevents updating a batch still referenced by an unsubmitted composer.

Source§

impl WgpuRenderer

Source

pub fn create_image_rgba8( &self, width: u32, height: u32, pixels: Vec<u8>, budget: ImageBudget, ) -> Result<Image2d, ImageError>

Creates an image while taking ownership of its exact recovery pixels.

Source

pub fn create_image_rgba8_from_slice( &self, width: u32, height: u32, pixels: &[u8], budget: ImageBudget, ) -> Result<Image2d, ImageError>

Copies a bounded byte slice into a retained image recovery snapshot.

Source

pub fn replace_image_rgba8( &self, image: &mut Image2d, width: u32, height: u32, pixels: Vec<u8>, ) -> Result<ImageUploadReport, ImageError>

Replaces dimensions and pixels atomically after full preflight.

Source

pub fn update_image_region( &self, image: &mut Image2d, region: ImageTexelRect, pixels: &[u8], ) -> Result<ImageUploadReport, ImageError>

Uploads one exact row-major RGBA region and updates recovery pixels.

Source

pub fn restore_image(&self, source: &Image2d) -> Result<Image2d, ImageError>

Recreates an exact image for this renderer from retained source bytes.

Source

pub fn create_image_batch( &self, image: &Image2d, sprites: Vec<ImageSprite2d>, budget: ImageBatchBudget, ) -> Result<ImageBatch2d, ImageError>

Creates one retained instanced batch tied to an image or atlas.

Source

pub fn replace_image_batch( &self, image: &Image2d, batch: &mut ImageBatch2d, sprites: Vec<ImageSprite2d>, ) -> Result<ImageBatchUploadReport, ImageError>

Updates a batch from owned input, reusing capacity when possible. Prefer Self::update_image_batch when the host retains its input storage.

Source

pub fn restore_image_batch( &self, image: &Image2d, source: &ImageBatch2d, ) -> Result<ImageBatch2d, ImageError>

Restores a batch against the restored copy of its original atlas.

Source§

impl WgpuRenderer

Source

pub async fn new( surface_target: impl Into<SurfaceTarget<'static>>, width: u32, height: u32, ) -> Result<Self, RendererInitError>

Creates a renderer for a window or canvas surface target.

width and height are physical surface pixels. Zero sizes are clamped to one pixel because wgpu surfaces cannot be configured at zero size. This convenience constructor assumes a display scale factor of 1.0; HiDPI hosts should use WgpuRenderer::new_with_options.

Source

pub async fn new_with_options( surface_target: impl Into<SurfaceTarget<'static>>, width: u32, height: u32, options: WgpuRendererOptions, ) -> Result<Self, RendererInitError>

Creates a renderer with explicit options.

Use this when measuring renderer throughput or when the host application needs control over presentation behavior or display scale.

Source

pub fn size(&self) -> (u32, u32)

Returns the configured surface size in physical pixels.

Source

pub fn gpu_timing_statistics(&self) -> GpuTimingStatistics

Returns optional GPU-query availability, bounded resources and losses. These diagnostics are independent of frame CPU timings and scene uploads.

Source

pub fn collect_gpu_timings(&mut self) -> GpuTimingBatch

Collects at most eight completed GPU pass intervals without waiting.

Call regularly when timing is enabled; uncollected samples occupy the fixed ring and eventually cause new samples to be dropped. Match sample IDs to submitted reports when excluding warmup or unrelated work. A call performs one nonblocking device poll and reports its CPU overhead. Device recovery discards pending old-device samples and resets the counters.

Source

pub fn logical_size(&self) -> (f32, f32)

Returns the render viewport size in logical screen pixels.

Scene clipping, stroke widths, shadows, and camera zoom use this logical coordinate space so their visual size remains stable across display DPI.

Source

pub fn logical_viewport(&self) -> Result<LogicalViewport, LogicalViewportError>

Returns the current camera viewport in logical screen pixels.

Source

pub fn physical_to_logical_screen( &self, position: PhysicalScreenPosition, ) -> Result<LogicalScreenPosition, RendererCoordinateError>

Converts a physical surface position into logical screen pixels.

Source

pub fn logical_to_physical_screen( &self, position: LogicalScreenPosition, ) -> Result<PhysicalScreenPosition, RendererCoordinateError>

Converts a logical screen position into physical surface pixels.

Source

pub fn scale_factor(&self) -> f64

Returns physical surface pixels per logical screen pixel.

Source

pub fn surface_present_mode(&self) -> RendererSurfacePresentMode

Returns the concrete presentation mode selected for the active surface.

RendererPresentMode::NoVsync prefers Immediate, then Mailbox, then FIFO. This method reveals which supported fallback was configured. The desktop compositor may still pace host redraw callbacks independently.

Source

pub fn set_pre_present_notify( &mut self, notify: impl Fn() + Send + Sync + 'static, )

Installs a host callback invoked immediately before synchronized surface presentation.

A winit host should capture its Arc<Window> and call Window::pre_present_notify from this callback. The renderer invokes it after command submission and immediately before Queue::present, but only when the concrete surface mode is Mailbox, FIFO, or FIFO-relaxed. Immediate presentation deliberately remains uncapped.

Source

pub fn clear_pre_present_notify(&mut self)

Removes the synchronized-presentation callback, if one is installed.

Source

pub fn adapter_name(&self) -> &str

Returns the active graphics adapter’s human-readable name.

Source

pub fn adapter_backend(&self) -> &'static str

Returns the active graphics API backend as a stable lowercase name.

Source

pub fn adapter_vendor_id(&self) -> u32

Returns the PCI vendor identifier reported for the active adapter.

Source

pub fn adapter_device_id(&self) -> u32

Returns the device identifier reported for the active adapter.

Source

pub fn adapter_pci_bus_id(&self) -> &str

Returns the backend-reported PCI bus address for the active adapter.

Vulkan adapters expose this as domain:bus:device.function when the driver supports VK_EXT_pci_bus_info. An empty string means that the backend cannot identify a physical adapter instance; release evidence must not treat model identifiers alone as instance identity.

Source

pub fn surface_format(&self) -> TextureFormat

Returns the texture format selected for the active presentation surface.

Source

pub fn surface_sample_count(&self) -> u32

Returns the raster sample count selected for the active surface format.

Source

pub fn adapter_driver(&self) -> &str

Returns the driver name reported by the active graphics adapter.

Source

pub fn adapter_driver_info(&self) -> &str

Returns the driver version or implementation detail reported by the adapter.

Source

pub fn quarantined_device_count(&self) -> usize

Returns previous logical devices retained for safe native-driver teardown.

Source

pub fn max_quarantined_device_count(&self) -> usize

Returns the configured maximum previous-device quarantine size.

Source

pub fn remaining_device_recoveries(&self) -> usize

Returns successful recoveries available before the bounded quarantine is full.

Source

pub fn wait_for_gpu_idle(&self) -> Result<(), PollError>

Blocks until all previously submitted GPU work has completed.

This is intended for diagnostic throughput measurements, readback, and controlled resource teardown. Calling it in an interactive frame loop defeats normal CPU/GPU pipelining.

Source

pub fn set_scale_factor( &mut self, scale_factor: f64, ) -> Result<(), RendererConfigurationError>

Replaces display scale without changing physical surface dimensions.

Invalid values return an error and leave the current scale unchanged.

Source

pub fn resize( &mut self, width: u32, height: u32, ) -> Result<(), RendererConfigurationError>

Reconfigures the surface after a host window resize.

Zero width or height is ignored because minimized windows often report zero size and wgpu cannot configure a zero-sized surface.

Source

pub fn resize_with_scale_factor( &mut self, width: u32, height: u32, scale_factor: f64, ) -> Result<(), RendererConfigurationError>

Reconfigures physical surface size and logical-to-physical display scale.

Invalid scale returns an error without changing either value. Zero physical dimensions update the scale but leave the minimized surface configuration unchanged.

Source§

impl WgpuRenderer

Source

pub fn create_mesh3d( &self, source: Mesh3d, ) -> Result<RetainedMesh3d, Mesh3dResourceError>

Uploads validated immutable topology into retained GPU buffers.

Counts, byte arithmetic, draw-count representation, and active-device buffer limits are checked before conversion staging allocation. Host staging reservation failure is returned explicitly; GPU allocation itself follows wgpu’s device-error model.

Source

pub fn restore_mesh3d( &self, source: &RetainedMesh3d, ) -> Result<RetainedMesh3d, Mesh3dResourceError>

Restores retained topology and its complete optional texture material onto this renderer after device replacement. Attribute/reserved buffer capacities, committed mip bytes/options, filtering, tint, UV transform, addressing and opaque/alpha-capable rebinding policy are preserved. Source handles remain unchanged. Separate calls restore separate GPU resources; use restore_scene3d to deduplicate shared scene resources.

Source

pub fn restore_scene3d( &self, scene: &mut Scene3d, ) -> Result<Scene3dRestoreReport, Mesh3dResourceError>

Atomically restores every stale retained mesh referenced by a 3D scene.

Distinct shared mesh and texture resources are uploaded once. Per-handle texture/filter/tint choices remain independent when topology is shared. Object IDs, insertion order, transforms, styles, visibility, scene provenance, and the next ID remain unchanged. If any capacity or host-staging allocation fails, the original scene is not modified. Targets remain separate resources and must be restored with WgpuRenderer::restore_render_target3d.

Source

pub fn create_render_target3d( &self, width: u32, height: u32, logical_viewport: LogicalViewport, ) -> Result<RenderTarget3d, Mesh3dResourceError>

Creates color/depth attachments for an explicit logical viewport.

width and height are physical target texels. Their aspect must match logical_viewport; the resulting texels-per-logical-pixel ratio controls wireframe width independently of the window DPI scale.

Source

pub fn restore_render_target3d( &self, source: &RenderTarget3d, ) -> Result<RenderTarget3d, Mesh3dResourceError>

Restores empty color/depth attachments with the source dimensions.

Source

pub fn render_scene3d_to_target( &mut self, target: &RenderTarget3d, scene: &Scene3d, camera: Camera3d, ) -> Result<Mesh3dRenderReport, Mesh3dRenderError>

Draws retained opaque/masked surfaces, sorted blended surfaces, then edges into a reusable premultiplied color/depth target.

Object insertion order does not determine visibility. Every surface writes and tests hardware depth. Model transforms are uploaded through a reusable instance buffer. Wholly inside objects retain their indexed topology; crossing surfaces use bounded homogeneous CPU clipping against all six frustum planes. Ambiguous classification/orientation produces an object-attributed error. Explicit display edges retain their complete interval-validated homogeneous shader clipper.

Source

pub fn validate_scene3d_for_target( &self, target: &RenderTarget3d, scene: &Scene3d, camera: Camera3d, budget: Mesh3dRenderBudget, ) -> Result<Mesh3dPreflightReport, Mesh3dRenderError>

Validates every visible object and exact submitted/generated work without changing target pixels, retained resources or submitting GPU work.

This is the same authoritative preflight used by rendering. Private bounded scratch may be allocated; the report is not a reusable draw authorization after the scene, target or camera changes. Mesh3dRenderBudget::surface_policy controls filled surfaces only. Native mode reports no CPU clipped/discarded source counts.

Source

pub fn render_scene3d_to_target_with_budget( &mut self, target: &RenderTarget3d, scene: &Scene3d, camera: Camera3d, budget: Mesh3dRenderBudget, ) -> Result<Mesh3dRenderReport, Mesh3dRenderError>

Draws a scene with explicit surface policy and submitted/generated limits.

All validation, generated counts and host/GPU capacity checks complete before the first GPU write or target mutation. Numerical ambiguity is attributed to the exact visible object; aggregate capacity remains a scene-level error. Wholly inside objects retain their indexed GPU path. With SurfaceRasterization3d::Native, all filled surfaces keep their original indices and hardware clipping defines boundary coverage. Finite shader arithmetic and independent mathematical-edge proofs still apply.

Source§

impl WgpuRenderer

Source

pub fn create_texture3d_rgba8_with_options( &self, width: u32, height: u32, pixels: Vec<u8>, budget: ImageBudget, options: Texture3dOptions, ) -> Result<Texture3d, Texture3dError>

Creates an independently filtered image with explicit alpha/mipmap policy. ImageBudget covers the complete committed CPU/GPU chain, not only mip zero.

Source

pub fn crop_texture3d_tile( &self, source: &Texture3d, region: ImageTexelRect, budget: ImageBudget, options: Texture3dOptions, ) -> Result<Texture3d, Texture3dError>

Crops physical source mip-zero texels into an independent local texture. Mip generation sees only the selected region. Later parent updates never propagate. Use full-domain normalized UVs and material Repeat to repeat a tile without sampling neighboring atlas entries at any mip level.

Source§

impl WgpuRenderer

Source

pub fn update_texture3d_region( &self, texture: &mut Texture3d, region: ImageTexelRect, pixels: &[u8], bytes_per_row: usize, budget: Texture3dUpdateBudget, ) -> Result<Texture3dUpdateReport, Texture3dUpdateError>

Updates this handle only, preserving all existing material/scene aliases.

Input is straight sRGB RGBA8 starting at byte zero; exact required length is (height - 1) * bytes_per_row + width * 4, with stride at least one row. Unique handles reuse GPU storage. Shared handles detach by GPU-copying the old chain before uploading the base patch and regenerated lower levels. Old materials remain unchanged until explicitly rebound with with_texture. Only affected lower-mip rectangles are regenerated and uploaded, with byte-identical filtering to full regeneration, including odd dimensions. CPU preparation still clones the complete recovery chain; shared handles still copy the complete GPU chain before applying these smaller uploads. Synchronous errors preserve CPU/GPU drawable state; asynchronous device loss is separate and recovery restores the last committed complete chain.

Source

pub fn update_scene3d_texture_region( &self, scene: &mut Scene3d, object_id: Object3dId, region: ImageTexelRect, pixels: &[u8], bytes_per_row: usize, budget: Texture3dUpdateBudget, ) -> Result<Texture3dUpdateReport, Texture3dUpdateError>

Updates only the selected object’s texture, preserving its ID/material and every external/other-object snapshot. A uniquely scene-owned texture reuses GPU storage; aliases detach. Mesh topology is never copied or uploaded. Texture, material alpha policy and final scene budgets are checked before writes. Private bounded scene bookkeeping may be prepared before failure.

Source§

impl WgpuRenderer

Source

pub fn create_texture3d_rgba8( &self, width: u32, height: u32, pixels: Vec<u8>, budget: ImageBudget, ) -> Result<Texture3d, Texture3dError>

Creates immutable opaque sRGB RGBA8 texels with shared nearest/linear bindings. Image/device limits and every alpha byte are validated before GPU creation.

Source

pub fn create_texture3d_rgba8_with_alpha( &self, width: u32, height: u32, pixels: Vec<u8>, budget: ImageBudget, ) -> Result<Texture3d, Texture3dError>

Creates sRGB RGBA8 texels with explicitly preserved straight alpha. Mask/blend materials sample alpha; Opaque surfaces deliberately ignore it.

Source

pub fn restore_texture3d( &self, source: &Texture3d, ) -> Result<Texture3d, Texture3dError>

Restores exact retained pixels on the current device. Existing handles remain unchanged; atomic restore_scene3d restores shared textures once.

Source

pub fn with_mesh3d_material( &self, mesh: &RetainedMesh3d, material: &TextureMaterial3d, ) -> Result<RetainedMesh3d, Texture3dError>

Associates a shared material with a new retained mesh handle without allocating/uploading topology. Existing clones and scene users keep their previous material; rebind selected IDs using Scene3d::set_mesh.

Source§

impl WgpuRenderer

Source

pub fn mesh3d_update_scratch_bytes(&self) -> usize

Renderer-owned dynamic mesh conversion capacity, separate from scene resources and FrameCacheBudget. Backend upload staging is excluded.

Source

pub fn clear_mesh3d_update_scratch(&mut self) -> usize

Releases retained dynamic conversion scratch without changing scene/GPU resources, returning the nominal CPU capacity released. Clear before requesting a smaller retained-scratch limit than a previous update used.

Source

pub fn update_scene3d_mesh( &mut self, scene: &mut Scene3d, object_id: Object3dId, source: Mesh3d, budget: DynamicMesh3dBudget, ) -> Result<DynamicMesh3dUpdateReport, DynamicMesh3dError>

Updates only the selected scene object, reusing unique GPU buffer capacity.

Other objects and exported immutable mesh/instance clones remain unchanged: aliases cause one whole-bundle copy-on-write. Once only this scene object owns the new bundle, subsequent fitting updates allocate no GPU buffers. Growth or an optional-attribute layout change replaces the whole bundle; all live vertices, indices, UVs, colors, normals and edges are uploaded even for an identical source.

Limits include retained capacity and old/new overlap. Validation, scene accounting and every fallible host preparation complete before GPU writes. Returned errors preserve the object’s previous geometry and scene state; private bounded conversion scratch may warm before a later device failure. Asynchronous device loss is separate; recovery restores committed source and reserved capacities. Queue ordering protects earlier submitted frames.

Source topology is a validated nonempty Mesh3d. Hide/remove an object for an empty chunk. Its ID, transform, visibility and material are preserved. UV-bearing material cannot be retained after removing source UVs.

Source§

impl WgpuRenderer

Source

pub fn create_mesh3d_with_budget( &self, source: Mesh3d, budget: Mesh3dUploadBudget, ) -> Result<RetainedMesh3d, Mesh3dResourceError>

Creates an immutable revision with explicit pre-allocation byte limits.

Source

pub fn replace_mesh3d( &self, mesh: &mut RetainedMesh3d, source: Mesh3d, budget: Mesh3dUploadBudget, ) -> Result<Mesh3dUploadReport, Mesh3dResourceError>

Atomically replaces this retained handle with a newly uploaded revision.

Clones and scene objects referencing the previous revision are unchanged; use Scene3d::set_mesh to rebind selected live objects. This is immutable replacement, not a stable-capacity in-place upload. Validation/allocation errors preserve the old handle; asynchronous device failure is separate. In-flight work keeps its old references. Nominal transient GPU/source overlap is at most the old revision plus the incoming configured limits.

Source§

impl WgpuRenderer

Source

pub fn create_particle_field( &self, instances: &[ParticleInstance2d], ) -> Result<ParticleField2d, ParticleFieldError>

Creates an instanced particle field owned by this renderer.

Each instance draws as one screen-space circle quad through a single GPU draw call. The retained CPU snapshot enables later restoration on a recreated renderer.

Source

pub fn create_particle_field_with_budget( &self, instances: &[ParticleInstance2d], budget: ParticleRenderBudget, ) -> Result<ParticleField2d, ParticleFieldError>

Creates an instanced particle field with hard retained-CPU, GPU-memory, upload, visible-instance, and visibility-work limits.

Source

pub fn set_particle_field_budget( &self, field: &mut ParticleField2d, budget: ParticleRenderBudget, ) -> Result<(), ParticleFieldError>

Replaces a particle field’s hard resource budget, shrinking memory when required.

Source

pub fn update_particle_field( &self, field: &mut ParticleField2d, instances: &[ParticleInstance2d], ) -> Result<ParticleStatistics, ParticleFieldError>

Replaces all particle instances, reusing GPU capacity whenever possible.

Source

pub fn update_particle_field_with_metrics( &self, field: &mut ParticleField2d, instances: &[ParticleInstance2d], ) -> Result<ParticleFieldUpdateReport, ParticleFieldError>

Replaces all particle instances and returns CPU-side preparation metrics.

Source

pub fn update_particle_field_range( &self, field: &mut ParticleField2d, first_instance: usize, instances: &[ParticleInstance2d], ) -> Result<ParticleStatistics, ParticleFieldError>

Replaces a contiguous particle range without reallocating its GPU buffer.

Source

pub fn update_particle_field_range_with_metrics( &self, field: &mut ParticleField2d, first_instance: usize, instances: &[ParticleInstance2d], ) -> Result<ParticleFieldUpdateReport, ParticleFieldError>

Replaces a particle range and returns CPU-side preparation metrics.

Source

pub fn restore_particle_field( &self, source: &ParticleField2d, ) -> Result<ParticleField2d, ParticleFieldError>

Recreates a particle field on this renderer from retained CPU instances.

Source

pub fn render_particle_field( &mut self, field: &mut ParticleField2d, background: Color, camera: &Camera2d, ) -> Result<RenderStatus, ParticleFieldRenderError>

Draws a particle field with a normalized clear color.

Source

pub fn render_particle_field_with_metrics( &mut self, field: &mut ParticleField2d, background: Color, camera: &Camera2d, ) -> Result<RenderReport, ParticleFieldRenderError>

Draws a particle field and returns regular per-frame renderer metrics.

Source

pub fn render_particle_field_to_target( &mut self, target: &RenderTarget2d, field: &mut ParticleField2d, camera: &Camera2d, load: RenderTargetLoad, ) -> Result<RenderReport, ParticleFieldRenderError>

Draws a particle field into an offscreen target.

Target dimensions may be lower than the presentation surface to bound raster cost. Camera coordinates and particle radii remain expressed in logical presentation pixels, so composing the target back to the surface preserves layout while trading only raster resolution.

Source§

impl WgpuRenderer

Source

pub fn create_scalar_field_texture( &self, field: ScalarField, ) -> Result<ScalarFieldTexture, ScalarFieldTextureError>

Creates a single-channel floating-point texture from a validated scalar grid.

Source

pub fn update_scalar_field_texture( &self, texture: &mut ScalarFieldTexture, field: ScalarField, ) -> Result<ScalarFieldUploadReport, ScalarFieldTextureError>

Replaces scalar data, recreating the texture only when dimensions change.

Source

pub fn update_scalar_field_texture_region( &self, texture: &mut ScalarFieldTexture, x: usize, y: usize, width: usize, height: usize, values: &[f32], ) -> Result<ScalarFieldUploadReport, ScalarFieldTextureError>

Updates a rectangular scalar texture region without recreating it.

Source

pub fn restore_scalar_field_texture( &self, source: &ScalarFieldTexture, ) -> Result<ScalarFieldTexture, ScalarFieldTextureError>

Recreates a scalar texture on this renderer from its retained CPU grid.

Source

pub fn render_scalar_field_texture( &mut self, texture: &ScalarFieldTexture, color_map: &ColorMap, value_range: (f32, f32), background: Color, ) -> Result<RenderStatus, ScalarFieldRenderError>

Draws a scalar texture across the logical viewport through a color map.

value_range maps minimum to the first color-map entry and maximum to the last. Values outside the range clamp to those endpoints.

Source

pub fn render_scalar_field_texture_with_metrics( &mut self, texture: &ScalarFieldTexture, color_map: &ColorMap, (minimum, maximum): (f32, f32), background: Color, ) -> Result<RenderReport, ScalarFieldRenderError>

Draws a scalar texture and returns normal CPU-side renderer metrics.

Source

pub fn render_scalar_field_texture_with_sampling( &mut self, texture: &ScalarFieldTexture, color_map: &ColorMap, (minimum, maximum): (f32, f32), sampling: ScalarFieldSampling, background: Color, ) -> Result<RenderStatus, ScalarFieldRenderError>

Draws a scalar texture with an explicit texel sampling mode.

Source

pub fn render_scalar_field_texture_to_target( &mut self, target: &RenderTarget2d, texture: &ScalarFieldTexture, color_map: &ColorMap, value_range: (f32, f32), background: Color, ) -> Result<RenderReport, ScalarFieldRenderError>

Renders a scalar heatmap into an offscreen target instead of presenting it.

Source

pub fn render_scalar_field_texture_to_target_with_sampling( &mut self, target: &RenderTarget2d, texture: &ScalarFieldTexture, color_map: &ColorMap, (minimum, maximum): (f32, f32), sampling: ScalarFieldSampling, background: Color, ) -> Result<RenderReport, ScalarFieldRenderError>

Renders a scalar heatmap into a target with explicit source sampling.

Source§

impl WgpuRenderer

Source

pub fn render_screen_primitive_batch_to_target( &mut self, target: &RenderTarget2d, batch: &ScreenPrimitiveBatch2d, pixel_scale: PhysicalPerLogical, viewport: Option<LogicalViewportRegion>, load: RenderTargetLoad, ) -> Result<RenderReport, RenderTargetError>

Draws a retained analytic screen batch into a single-sample color target.

pixel_scale is the target’s physical texels per logical pixel, independent of window DPI. Primitive positions and captured clips are local to the optional viewport. The target is explicitly loaded or cleared, including outside the viewport. Colors use the target’s existing premultiplied-alpha storage contract. This submits offscreen work, never a surface present.

Restored resources must belong to the active renderer. All synchronous source/viewport/clear-color validation happens before a queue write.

Source

pub fn render( &mut self, scene: &Scene, camera: &Camera2d, ) -> Result<RenderStatus, RendererFrameError>

Draws a scene using the supplied camera.

Scene positions and sizes are in world units unless a style explicitly says logical screen pixels. The renderer converts world coordinates to logical screen coordinates through Camera2d, then to normalized device coordinates for wgpu. Scissor rectangles are converted to physical surface pixels at the final backend boundary.

Source

pub fn render_screen_scene( &mut self, scene: &ScreenScene, ) -> Result<RenderStatus, RendererFrameError>

Draws fixed logical-screen geometry independently of any world camera.

Source

pub fn render_screen_scene_with_metrics( &mut self, scene: &ScreenScene, ) -> Result<RenderReport, RendererFrameError>

Draws fixed logical-screen geometry and returns CPU-side stage timings.

Source

pub fn render_with_metrics( &mut self, scene: &Scene, camera: &Camera2d, ) -> Result<RenderReport, RendererFrameError>

Draws a scene and returns a CPU-side timing breakdown.

Metrics separate tessellation, buffer upload, surface acquisition, and encode/submit/present dispatch. They do not measure GPU completion or the monitor scanout timestamp. Use this for diagnostics; WgpuRenderer::render is the simpler equivalent when stage timings are not needed.

Source

pub fn render_scene_in_viewport( &mut self, scene: &Scene, camera: &Camera2d, viewport: LogicalViewportRegion, ) -> Result<RenderStatus, RendererFrameError>

Draws a world scene through a camera into one bounded logical viewport.

Scene-local clips are interpreted relative to the viewport and are intersected with its physical bounds. The surface outside the viewport is cleared but receives no scene geometry.

Source

pub fn render_scene_in_viewport_with_metrics( &mut self, scene: &Scene, camera: &Camera2d, viewport: LogicalViewportRegion, ) -> Result<RenderReport, RendererFrameError>

Draws one bounded scene viewport and returns CPU-side stage timings.

Source

pub fn render_scene_to_target( &mut self, target: &RenderTarget2d, scene: &Scene, camera: &Camera2d, pixel_scale: PhysicalPerLogical, viewport: Option<LogicalViewportRegion>, load: RenderTargetLoad, ) -> Result<RenderReport, RenderTargetError>

Renders an ordinary world scene into an offscreen target.

Target dimensions are physical texels. pixel_scale defines their logical viewport density independently of window DPI. Scene-local clips are relative to viewport, or to the complete logical target when no region is supplied. load explicitly controls preservation or clearing.

Source

pub fn prepare_scene( &self, scene: &Scene, ) -> Result<PreparedScene, PreparedSceneError>

Tessellates a scene once and uploads immutable geometry to a dedicated GPU buffer.

Preparing is appropriate for geometry that will be drawn repeatedly while only the camera or target dimensions change. Any shape, style, gradient, ordering, or clipping change requires preparing a replacement scene.

Source

pub fn prepare_screen_scene( &self, scene: &ScreenScene, ) -> Result<PreparedScreenScene, PreparedSceneError>

Tessellates fixed logical-screen geometry once for repeated composition.

Source

pub fn restore_prepared_screen_scene( &self, source: &PreparedScreenScene, ) -> Result<PreparedScreenScene, PreparedSceneError>

Restores prepared logical-screen geometry on a recreated renderer.

Source

pub fn restore_prepared_scene( &self, source: &PreparedScene, ) -> Result<PreparedScene, PreparedSceneError>

Recreates prepared GPU resources on this renderer from a retained CPU snapshot.

This supports renderer recreation after device loss without requiring the original high-level Scene. The returned snapshot belongs to this renderer.

Source

pub fn render_prepared( &mut self, scene: &PreparedScene, camera: &Camera2d, ) -> Result<RenderStatus, PreparedSceneRenderError>

Draws geometry previously uploaded by WgpuRenderer::prepare_scene.

Camera and viewport changes are applied by the vertex shader and do not rebuild or re-upload the prepared geometry.

Source

pub fn render_prepared_screen_scene( &mut self, scene: &PreparedScreenScene, ) -> Result<RenderStatus, PreparedSceneRenderError>

Draws prepared logical-screen geometry independently of a world camera.

The prepared resource must have been built from a ScreenScene. The renderer cannot infer the original coordinate space from raw prepared vertices, so passing world-space geometry is a caller contract error.

Source

pub fn render_prepared_with_metrics( &mut self, scene: &PreparedScene, camera: &Camera2d, ) -> Result<RenderReport, PreparedSceneRenderError>

Draws prepared geometry and reports per-frame CPU timing.

Tessellation and geometry upload durations are zero because both happened in WgpuRenderer::prepare_scene. The camera uniform is still updated once per frame.

Source§

impl WgpuRenderer

Source

pub fn create_render_target( &self, width: u32, height: u32, ) -> Result<RenderTarget2d, RenderTargetError>

Creates an offscreen target with physical texture dimensions.

Source

pub fn create_trail_buffer( &mut self, width: u32, height: u32, ) -> Result<TrailBuffer2d, RenderTargetError>

Creates an empty ping-pong target pair for temporal accumulation.

Source

pub fn restore_render_target( &self, source: &RenderTarget2d, ) -> Result<RenderTarget2d, RenderTargetError>

Recreates an empty target with the source dimensions after device recovery.

GPU pixels are deliberately not retained; callers redraw their source state into the restored target.

Source

pub fn restore_trail_buffer( &mut self, source: &TrailBuffer2d, ) -> Result<TrailBuffer2d, RenderTargetError>

Recreates an empty, deterministically cleared trail buffer after recovery.

Source

pub fn accumulate_trail_buffer( &mut self, trails: &mut TrailBuffer2d, source: &RenderTarget2d, retention: f32, source_opacity: f32, source_blend: BlendMode, ) -> Result<RenderReport, RenderTargetError>

Retains one bounded frame of history and composites a fresh source target.

retention and source_opacity must be finite values in 0.0..=1.0. The source must be a distinct target; use WgpuRenderer::clear_trail_buffer to discard accumulated history deterministically.

Source

pub fn clear_trail_buffer( &mut self, trails: &mut TrailBuffer2d, color: Color, ) -> Result<RenderReport, RenderTargetError>

Clears both ping-pong targets, resetting all retained temporal history.

Source

pub fn compose_trail_buffer( &mut self, trails: &TrailBuffer2d, blend_mode: BlendMode, opacity: f32, background: Color, ) -> Result<RenderReport, RenderTargetError>

Presents the most recently accumulated temporal target.

Source

pub fn compose_render_target( &mut self, target: &RenderTarget2d, blend_mode: BlendMode, opacity: f32, background: Color, ) -> Result<RenderReport, RenderTargetError>

Composes a previously rendered target over the presentation surface.

Source§

impl WgpuRenderer

Source

pub fn render_layered_visualization( &mut self, target: &RenderTarget2d, scalar: &ScalarFieldTexture, color_map: &ColorMap, particles: &mut ParticleField2d, camera: &Camera2d, options: LayeredVisualizationOptions, ) -> Result<LayeredVisualizationReport, LayeredVisualizationError>

Draws a heatmap, overlays a budgeted particle field, and presents the target with one command encoder and one queue submission.

This fused path avoids synchronously submitting each layer and leaves substantially more CPU time for an expensive host simulation. The target should have the presentation surface’s aspect ratio; lowering both dimensions proportionally bounds raster cost without changing the logical camera contract.

Auto Trait Implementations§

Blanket Implementations§

Source§

impl<T> Any for T
where T: 'static + ?Sized,

Source§

fn type_id(&self) -> TypeId

Gets the TypeId of self. Read more
Source§

impl<T> Borrow<T> for T
where T: ?Sized,

Source§

fn borrow(&self) -> &T

Immutably borrows from an owned value. Read more
Source§

impl<T> BorrowMut<T> for T
where T: ?Sized,

Source§

fn borrow_mut(&mut self) -> &mut T

Mutably borrows from an owned value. Read more
Source§

impl<ST, DT> CastableFrom<ST, Initialized, Initialized> for DT
where ST: ?Sized, DT: ?Sized,

Source§

impl<ST, DT> CastableFrom<ST, Uninit, Uninit> for DT
where ST: ?Sized, DT: ?Sized,

Source§

impl<T> Downcast<T> for T

Source§

fn downcast(&self) -> &T

Source§

impl<T> From<T> for T

Source§

fn from(t: T) -> T

Returns the argument unchanged.

Source§

impl<T, U> Into<U> for T
where U: From<T>,

Source§

fn into(self) -> U

Calls U::from(self).

That is, this conversion is whatever the implementation of From<T> for U chooses to do.

Source§

impl<T> Read<Exclusive, BecauseExclusive> for T
where T: ?Sized,

Source§

impl<T, U> TryFrom<U> for T
where U: Into<T>,

Source§

type Error = !

The type returned in the event of a conversion error.
Source§

fn try_from(value: U) -> Result<T, !>

Performs the conversion.
Source§

impl<T, U> TryInto<U> for T
where U: TryFrom<T>,

Source§

type Error = <U as TryFrom<T>>::Error

The type returned in the event of a conversion error.
Source§

fn try_into(self) -> Result<U, <U as TryFrom<T>>::Error>

Performs the conversion.
Source§

impl<T> Upcast<T> for T

Source§

fn upcast(&self) -> Option<&T>

Source§

impl<T> WasmNotSend for T
where T: Send,

Source§

impl<T> WasmNotSendSync for T

Source§

impl<T> WasmNotSync for T
where T: Sync,