Skip to content
Project referenceDevelopmentEngine ccf7d03f
Authored
  • Output requests, render-model generations and owned GPU resource caches.

Output requests, render-model generations and owned GPU resource caches.

Start with rendering tasks and frame ownership.

RenderRuntime combines output requests, runtime assets and a neutral device into executable frames. It owns render-model and Pipeline/Feature generations, asset-to-GPU caches and retirement queues. It does not depend on asset authoring, AssetPipeline or shader compilation tools and has no built-in 2D/3D world model.

The host supplies TypeCatalog, a borrowed device, diagnostics and required artifact providers. Runtime owns its created models, resource services, targets and generations. Construction composes services; the containing host/subsystem supplies the frame lifecycle. Device destruction remains the host’s responsibility.

ContentRenderTargetArtifactProvider reads already compiled artifacts through logical content locations. Runtime does not compile authoring sources or search physical asset directories. Keep the content owner and converter generation alive for their borrowed operations.

Implement IRenderModel to interpret a content model through its stable extension identity. Pipeline/Feature extensions describe drawing and required device capabilities. IRenderFrameGraphContributor adds neutral operations to the final graph; IRenderTargetArtifactProvider supplies deployed shader/texture data. These are distinct responsibilities.

Contributor collection changes create a new immutable snapshot. A frame captures one snapshot; registration changes affect the next frame. Registration does not transfer lifecycle ownership, and unregistering does not permit immediate disposal while the current frame can still call the instance.

Each contributor builds inside its own graph mutation scope. An ordinary extension failure reports diagnostics and rolls back that contribution while preserving accepted graph state. Retirement Pending still propagates to the owner. Final Compile constructs the complete graph once; graphCompileCount exposes the observed count.

Without a usable primary output, skip its graph construction and input-coordinate conversion while independent offscreen outputs continue. Missing artifacts, insufficient capabilities and corrupt content have distinct diagnostics. Candidate failure retains last-good state.

Exit first cancels candidates and retires generations, then drops request/contributor references, releases targets, compositors, uploads and resource services, and finally closes the device frame. Pending or failed retirement is not success. Private scratch may retain capacity but clears extension references after a frame, failure and shutdown; it is not a caller-owned persistent snapshot.

Supply live, explicitly owned dependencies as described above. These examples compile against the host contract; they do not initialize a complete product or verify a device.

using Inno.Core.Diagnostics;
using Inno.Extensibility.Types;
using Inno.Rendering;
using Inno.Rendering.Runtime;
static RenderRuntime CreateRuntime(
TypeCatalog types,
IRenderDevice device,
IDiagnosticReporter diagnostics
) {
return new RenderRuntime(types, device, diagnostics);
}

The source project defines the host declarations. Source API extraction does not establish script visibility, native behavior or completed generation retirement.

Browse types and members.

The compiler-selected script declarations are separate from this project’s host CLR surface. Read the script calling guide and the associated module contracts before using a projected member.

Project: Inno.Rendering.Runtime
Source: src/services/rendering/Inno.Rendering.Runtime/Inno.Rendering.Runtime.csproj

Type API Description
ContentRenderTargetArtifactProvider engine Reads immutable render target artifacts from one verified logical content store.
GraphicsSettings engine Exposes current rendering configuration and immutable device state.
IPreparedViewDrawable engine Encodes a resource-ready draw into the owning model’s scene pass.
IRenderModel engine Builds view requests from host output sessions without owning a host window.
IRenderRequestSink engine Accepts rendering-model-neutral requests without exposing runtime or backend ownership.
IRenderResourceService engine Resolves neutral assets and provider-owned uploads into opaque resources for the active device generation.
IRenderRuntimeReloadTransaction engine Controls one isolated rendering-extension candidate from preparation through atomic completion.
IRenderTargetArtifactProvider engine Resolves immutable, source-free target artifacts for the active rendering device.
IViewContentCollector engine Collects all active world-content sources for one exact view.
IViewContentFrameSource engine Completes frame-local input after every output has routed its views.
IViewContentSink engine Receives world items without imposing a rendering-model sort key.
IViewContentSource engine Contributes model-independent world items to selected views.
IViewDrawable engine Prepares one model-independent drawable for a specific render pass.
IViewPointerTarget engine Receives pointer input after the rendering model resolves visible draw order.
RenderExtensionStateContext engine Restores neutral render settings through the canonical asset’s actual owner and pinned references.
RenderFeatureContext engine Supplies one configured feature with frame-scoped graph services.
RenderFeatureExtensionAttribute engine Marks a reloadable pipeline feature implementation with a stable extension identifier.
RenderFrameData engine Carries generation-scoped, frame-only payloads between a request producer and its pipeline.
RenderFrameStatistics engine Reports read-only statistics for the most recently completed render frame.
RenderGeometry engine Provides generation-scoped GPU buffers for an imported geometry helper asset.
RenderGeometrySection engine Describes one indexed range in resolved backend-neutral geometry.
RenderMaterialPass engine Represents one frame-resolved material pass and its material-owned bindings.
RenderModelExtensionAttribute engine Marks a reloadable rendering model with a stable identity.
RenderModelOutput engine Gets one model’s prepared pipeline data for a host-owned target.
RenderOutputInput engine Frame-local input located in the output viewport’s physical pixels.
RenderOutputLayer engine Assigns exactly one rendering model and its world-content sources to an output layer.
RenderOutputRoute engine States the exact model order and exclusive world-content assignment for one output.
RenderOutputSession engine Declares a host output without naming a camera or scene model.
RenderPersistentResourceId engine Identifies one provider-owned persistent GPU resource without exposing a native handle.
RenderPipeline engine Builds frame-local passes without prescribing a rendering model.
RenderPipelineContext engine Supplies one request and frame-scoped services to a render pipeline.
RenderPipelineExtensionAttribute engine Marks a reloadable render pipeline implementation with a stable extension identifier.
RenderPipelineFeature engine Adds capability-aware passes without owning frame graph state.
RenderRequest engine Requests one pipeline-defined rendering operation without prescribing world semantics.
RenderRequestProvider engine Produces arbitrary render requests without prescribing a scene or rendering model.
RenderRequestProviderContext engine Supplies frame timing, capabilities and the request sink to one provider invocation.
RenderRequestProviderExtensionAttribute engine Marks a reloadable provider that produces model-neutral render requests each frame.
RenderResourceLimits engine Sets finite native resource and asynchronous readback admission limits for one rendering owner.
RenderResourceMap engine Stores open semantic graph resources for one pipeline request.
RenderResourceProvider engine Provides protected construction operations for replaceable render resource-service implementations.
RenderResourceProvider.MaterialBinding engine Represents one material binding staged by a derived resource service.
RenderResourceStatistics engine Reports control-thread resource occupancy and admission pressure without exposing backend objects.
RenderRuntime engine Owns the sole graphics frame boundary and executes model-neutral render requests.
RenderRuntimeFactory engine Creates one rendering lifecycle feature for every isolated runtime session.
RenderTargetArtifactStatus engine Describes the current availability of one target-specific rendering artifact.
RenderTargetStore engine Owns persistent offscreen targets without exposing backend-native handles.
RenderTextureSubresourceData engine Stores one complete texture mip and addressable layer, slice, or cubemap-face upload.
RenderView engine Describes one backend-neutral view produced by a rendering model.
SelectedViewContentCollector engine Restricts a neutral world-content collector to an output layer’s source IDs.
ViewContentContext engine Supplies one content source with the host-selected scene roots and an exact view.
ViewContentItem engine One world item supplied to a rendering model without choosing its sort policy.
ViewContentSourceExtensionAttribute engine Discovers an independently reloadable world-content source.
GraphicsSettings Runtime Exposes current rendering configuration and immutable device state.
IPreparedViewDrawable Runtime Encodes a resource-ready draw into the owning model’s scene pass.
IRenderModel Runtime Builds view requests from host output sessions without owning a host window.
IRenderRequestSink Runtime Accepts rendering-model-neutral requests without exposing runtime or backend ownership.
IRenderResourceService Runtime Resolves neutral assets and provider-owned uploads into opaque resources for the active device generation.
IViewContentCollector Runtime Collects all active world-content sources for one exact view.
IViewContentFrameSource Runtime Completes frame-local input after every output has routed its views.
IViewContentSink Runtime Receives world items without imposing a rendering-model sort key.
IViewContentSource Runtime Contributes model-independent world items to selected views.
IViewDrawable Runtime Prepares one model-independent drawable for a specific render pass.
IViewPointerTarget Runtime Receives pointer input after the rendering model resolves visible draw order.
RenderExtensionStateContext Runtime Restores neutral render settings through the canonical asset’s actual owner and pinned references.
RenderFeatureContext Runtime Supplies one configured feature with frame-scoped graph services.
RenderFeatureExtensionAttribute Runtime Marks a reloadable pipeline feature implementation with a stable extension identifier.
RenderFrameData Runtime Carries generation-scoped, frame-only payloads between a request producer and its pipeline.
RenderFrameStatistics Runtime Reports read-only statistics for the most recently completed render frame.
RenderGeometry Runtime Provides generation-scoped GPU buffers for an imported geometry helper asset.
RenderGeometrySection Runtime Describes one indexed range in resolved backend-neutral geometry.
RenderMaterialPass Runtime Represents one frame-resolved material pass and its material-owned bindings.
RenderModelExtensionAttribute Runtime Marks a reloadable rendering model with a stable identity.
RenderModelOutput Runtime Gets one model’s prepared pipeline data for a host-owned target.
RenderOutputInput Runtime Frame-local input located in the output viewport’s physical pixels.
RenderOutputLayer Runtime Assigns exactly one rendering model and its world-content sources to an output layer.
RenderOutputRoute Runtime States the exact model order and exclusive world-content assignment for one output.
RenderOutputSession Runtime Declares a host output without naming a camera or scene model.
RenderPersistentResourceId Runtime Identifies one provider-owned persistent GPU resource without exposing a native handle.
RenderPipeline Runtime Builds frame-local passes without prescribing a rendering model.
RenderPipelineContext Runtime Supplies one request and frame-scoped services to a render pipeline.
RenderPipelineExtensionAttribute Runtime Marks a reloadable render pipeline implementation with a stable extension identifier.
RenderPipelineFeature Runtime Adds capability-aware passes without owning frame graph state.
RenderRequest Runtime Requests one pipeline-defined rendering operation without prescribing world semantics.
RenderRequestProvider Runtime Produces arbitrary render requests without prescribing a scene or rendering model.
RenderRequestProviderContext Runtime Supplies frame timing, capabilities and the request sink to one provider invocation.
RenderRequestProviderExtensionAttribute Runtime Marks a reloadable provider that produces model-neutral render requests each frame.
RenderResourceMap Runtime Stores open semantic graph resources for one pipeline request.
RenderTargetArtifactStatus Runtime Describes the current availability of one target-specific rendering artifact.
RenderTextureSubresourceData Runtime Stores one complete texture mip and addressable layer, slice, or cubemap-face upload.
RenderView Runtime Describes one backend-neutral view produced by a rendering model.
SelectedViewContentCollector Runtime Restricts a neutral world-content collector to an output layer’s source IDs.
ViewContentContext Runtime Supplies one content source with the host-selected scene roots and an exact view.
ViewContentItem Runtime One world item supplied to a rendering model without choosing its sort policy.
ViewContentSourceExtensionAttribute Runtime Discovers an independently reloadable world-content source.