Enum ExecMode 

pub enum ExecMode {
    Stream,
    Graph,
    Eager,
    MultiStream(usize),
}

How the schedule executes.

Stream (default) launches each Step on the default stream every run(). Graph captures the full schedule into a CUDA Graph on first run and replays the captured graph on subsequent runs — eliminates per-launch dispatch overhead (~10-20% on small-batch inference). Eager is a one-shot helper that compiles + runs + drops the executable in one call; useful for interactive debugging. MultiStream(n) allocates a pool of n streams and assigns each Step to a stream based on data dependencies — independent ops (e.g. unfused Q/K/V projections, FFN gate/up) run in parallel. Cross-stream synchronization uses CUDA events at producer-consumer boundaries.

Variants

Stream

Graph

Eager

MultiStream(usize)

Trait Implementations

impl Clone for ExecMode

fn clone(&self) -> ExecMode

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Copy for ExecMode

impl Debug for ExecMode

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

Formats the value using the given formatter.

impl Default for ExecMode

fn default() -> ExecMode

Returns the “default value” for a type.

impl Eq for ExecMode

impl PartialEq for ExecMode

fn eq(&self, other: &ExecMode) -> bool

Tests for self and other values to be equal, and is used by ==.

fn ne(&self, other: &Rhs) -> bool

Tests for !=. The default implementation is almost always sufficient, and should not be overridden without very good reason.

impl StructuralPartialEq for ExecMode