Module compute_graph 

GPU compute graph — a typed, topologically-ordered execution planner.

Models the task graph used internally by modern GPU driver stacks (e.g. DirectX 12 command lists, Vulkan render graphs, WebGPU compute passes).

Node types

Variant	Meaning
NodeKind::Kernel	Dispatch a compute shader / kernel.
NodeKind::Copy	Transfer data between buffers (DMA-style).
NodeKind::Barrier	Memory / execution barrier between stages.

Workflow

1. Create a ComputeGraph.
2. Add nodes via ComputeGraph::add_node.
3. Add resource bindings via ComputeGraph::bind_resource.
4. Add edges via ComputeGraph::add_edge.
5. Call ComputeGraph::execution_order to obtain a valid ordering.
6. Optionally call ComputeGraph::validate to check for binding issues.

Structs

ComputeGraph

Directed acyclic graph of compute nodes with resource binding management.

ExecutionPlan

The result of topological ordering: an ordered list of node IDs.

GraphNode

A single node in the compute graph.

PipelineStageFlags

Bit flags representing pipeline stages for barrier nodes.

ResourceBinding

A named resource bound to a specific node.

Enums

GraphError

Errors produced by the compute graph.

NodeKind

The functional category of a compute graph node.

ResourceAccess

Read / write access mode for a resource binding.