Struct Scheduler 

pub struct Scheduler { /* private fields */ }

A demand-driven, parallel tile evaluator.

One scheduler can run many pipelines; the pool and cache are reused.

Implementations

impl Scheduler

pub fn new(options: SchedulerOptions) -> Result<Scheduler, PixelsError>

Build a scheduler with options.

Errors

Returns PixelsError::Io if worker threads cannot be spawned.

pub fn with_defaults() -> Result<Scheduler, PixelsError>

A scheduler with default tuning.

Errors

As Scheduler::new.

pub fn global() -> Result<Arc<Scheduler>, PixelsError>

The process-wide scheduler, built on first use with default options: one worker per core and the default cache budget.

This is what an output runs on unless told otherwise, so every pipeline in a process shares one pool of threads and one tile cache however many run at once. Concurrent runs queue their tiles on the same workers rather than each spawning a pool of their own, which is what keeps a server handling forty requests at a time from running forty threads per core. Its workers sleep while there is no work and live as long as the process.

Errors

Returns PixelsError::Io if the first call cannot spawn the worker threads. A later call tries again.

pub const fn threads(&self) -> usize

The worker thread count.

pub fn cache(&self) -> &Arc<TileCache>

The shared tile cache.

pub fn run( &self, image: &Image, consume: impl FnMut(Region, &Tile<'_>) -> Result<(), PixelsError>, ) -> Result<RunStats, PixelsError>

Evaluate image, handing each output tile to consume in order.

consume is called on the calling thread, in tile order, so a sink may hold non-Send state and an encoder still sees rows top to bottom.

Errors

Propagates any producer, op or consumer error. A failure anywhere fails the whole run; the sink is simply not given the remaining tiles (ARCHITECTURE §Failure model).

Trait Implementations

impl Debug for Scheduler

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

Formats the value using the given formatter.