Enum ExecutionNode 

pub enum ExecutionNode {
    Brick {
        id: BrickId,
        timing_ns: u64,
        elements: u64,
    },
    Kernel {
        name: String,
        ptx_hash: u64,
        grid: (u32, u32, u32),
        block: (u32, u32, u32),
        shared_mem: u32,
        timing_ns: Option<u64>,
        arithmetic_intensity: Option<f32>,
        achieved_tflops: Option<f32>,
    },
    Transfer {
        src: String,
        dst: String,
        bytes: u64,
        direction: TransferDirection,
        timing_ns: Option<u64>,
    },
    Function {
        name: String,
        file: Option<String>,
        line: Option<u32>,
    },
    Layer {
        index: u32,
    },
    AsyncTask {
        name: String,
        poll_count: u64,
        yield_count: u64,
        total_poll_ns: u64,
    },
}

Execution graph node types.

PAR-201: Represents different levels of the execution hierarchy.

Variants

Brick

High-level brick (BrickId from v2)

Fields

id: BrickId

timing_ns: u64

elements: u64

Kernel

GPU kernel launch

Fields

name: String

ptx_hash: u64

FNV-1a hash of PTX source for identity

grid: (u32, u32, u32)

Grid dimensions (blocks)

block: (u32, u32, u32)

Block dimensions (threads)

shared_mem: u32

Shared memory bytes

timing_ns: Option<u64>

Kernel execution time in nanoseconds (Phase 9: for CPA)

arithmetic_intensity: Option<f32>

Arithmetic intensity (FLOPs/byte) for roofline analysis (Phase 9)

achieved_tflops: Option<f32>

Achieved throughput in TFLOP/s (Phase 9)

Transfer

Memory transfer operation (Phase 9: data movement topology)

Fields

src: String

Source location description

dst: String

Destination location description

bytes: u64

Bytes transferred

direction: TransferDirection

Transfer direction

timing_ns: Option<u64>

Transfer time in nanoseconds

Function

Rust function (from DWARF or manual annotation)

Fields

name: String

file: Option<String>

line: Option<u32>

Layer

Transformer layer grouping

Fields

index: u32

AsyncTask

Phase 11 (E.9.4): Async task metrics for poll efficiency tracking

Fields

name: String

Task name for identification

poll_count: u64

Number of times poll() was called

yield_count: u64

Number of times poll() returned Pending

total_poll_ns: u64

Total time spent in poll() (nanoseconds)

Implementations

impl ExecutionNode

pub fn name(&self) -> String

Get the display name of this node.

pub fn is_kernel(&self) -> bool

Check if this is a kernel node.

pub fn is_brick(&self) -> bool

Check if this is a brick node.

pub fn is_transfer(&self) -> bool

Check if this is a transfer node.

pub fn timing_ns(&self) -> Option<u64>

Get timing if available (bricks, kernels, and transfers).

pub fn ptx_hash(&self) -> Option<u64>

Get PTX hash if available (kernels only).

pub fn arithmetic_intensity(&self) -> Option<f32>

Get arithmetic intensity if available (kernels only, Phase 9).

pub fn achieved_tflops(&self) -> Option<f32>

Get achieved TFLOP/s if available (kernels only, Phase 9).

pub fn transfer_bytes(&self) -> Option<u64>

Get transfer bytes if available (transfers only, Phase 9).

Trait Implementations

impl Clone for ExecutionNode

fn clone(&self) -> ExecutionNode

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for ExecutionNode

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

Formats the value using the given formatter.