Struct SchedulerWarpConfig 

pub struct SchedulerWarpConfig {
    pub scheduler_warp_id: u32,
    pub scheduler: SchedulerConfig,
    pub work_queue_capacity: usize,
    pub poll_interval_ns: u32,
}

Configuration for the scheduler warp pattern in CUDA codegen.

This controls how the scheduler warp (warp 0 by default) is generated within each persistent kernel thread block. When enabled, the codegen produces a split: warp 0 handles work distribution, remaining warps do message processing.

Fields

scheduler_warp_id: u32

Which warp handles scheduling (default: 0).

scheduler: SchedulerConfig

The scheduling parameters.

work_queue_capacity: usize

Size of the global work queue (number of WorkItem slots). Must be power of 2. Used for round-robin and priority strategies.

poll_interval_ns: u32

Polling interval in nanoseconds for the scheduler warp when no work is available (default: 1000ns).

Implementations

impl SchedulerWarpConfig

pub fn new(scheduler: SchedulerConfig) -> Self

Create a new scheduler warp config with the given strategy.

pub fn disabled() -> Self

Create a static (disabled) scheduler warp config. Codegen will produce the original non-split kernel.

pub fn with_scheduler_warp(self, warp_id: u32) -> Self

Set the scheduler warp ID.

pub fn with_work_queue_capacity(self, capacity: usize) -> Self

Set the global work queue capacity.

pub fn with_poll_interval_ns(self, ns: u32) -> Self

Set the poll interval for the scheduler warp.

pub fn is_enabled(&self) -> bool

Check if the scheduler warp pattern should be generated.

Trait Implementations

impl Clone for SchedulerWarpConfig

fn clone(&self) -> SchedulerWarpConfig

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for SchedulerWarpConfig

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

Formats the value using the given formatter.

impl Default for SchedulerWarpConfig

fn default() -> Self

Returns the “default value” for a type.