Struct BatchedComputePass 

pub struct BatchedComputePass { /* private fields */ }

Accumulates compute dispatch commands and issues them in batches.

Batching reduces command-encoder overhead by coalescing small dispatches and sorting them by pipeline so that pipeline switches are minimised.

When the number of pending commands reaches max_batch_size, an automatic flush is triggered and the commands are sorted by pipeline_id before being returned.

Implementations

impl BatchedComputePass

pub fn new(max_batch_size: usize) -> Self

Create a BatchedComputePass with the given max_batch_size.

A max_batch_size of 0 is treated as 1 (each submit is auto-flushed).

pub fn submit(&mut self, cmd: DispatchCommand) -> bool

Submit a dispatch command.

Returns true if an automatic flush was triggered (i.e., the pending queue reached max_batch_size). The caller should retrieve the flushed batch via flush when this returns true.

pub fn flush(&mut self) -> Vec<DispatchCommand>

Drain all pending commands, sorted by pipeline_id (ascending) to minimise pipeline state switches on the GPU.

Returns the sorted batch. If the queue is empty, returns an empty Vec.

pub fn pending_count(&self) -> usize

Number of commands currently pending.

pub fn total_flushed(&self) -> u64

Total commands flushed across all batches.

pub fn max_batch_size(&self) -> usize

Maximum batch size before an auto-flush is triggered.

Trait Implementations

impl Debug for BatchedComputePass

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

Formats the value using the given formatter.