Struct ExecutionPlan 

pub struct ExecutionPlan {
    pub steps: Vec<PlanStep>,
    pub num_streams: usize,
    pub pool_bytes: usize,
    pub kernel_count_original: usize,
    pub kernel_count_fused: usize,
    pub event_count: usize,
}

The compiled, optimised execution plan for a ComputeGraph.

Contains an ordered list of PlanSteps that can be submitted to an executor backend (sequential simulator or real CUDA graph capture).

Fields

steps: Vec<PlanStep>

Steps in submission order.

num_streams: usize

Number of distinct CUDA streams used.

pool_bytes: usize

Total device memory pool size in bytes.

kernel_count_original: usize

Number of kernel launches (before fusion).

kernel_count_fused: usize

Number of kernel launches (after fusion).

event_count: usize

Number of cross-stream event sync pairs.

Implementations

impl ExecutionPlan

pub fn build(graph: &ComputeGraph, max_streams: usize) -> GraphResult<Self>

Compiles a ComputeGraph into an ExecutionPlan.

Runs all analysis and optimisation passes in order: topo → liveness → fusion → memory → stream → linearise.

Parameters

• graph — the computation graph to compile.
• max_streams — upper bound on concurrent CUDA streams.

Errors

Propagates errors from any analysis or optimisation pass.

pub fn steps_on(&self, stream: StreamId) -> Vec<&PlanStep>

Returns all steps assigned to a specific stream, in order.

pub fn total_steps(&self) -> usize

Returns the total number of steps.

pub fn compute_steps(&self) -> usize

Returns the number of kernel-launch steps (after fusion).

Trait Implementations

impl Clone for ExecutionPlan

fn clone(&self) -> ExecutionPlan

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for ExecutionPlan

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl Display for ExecutionPlan

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.