Struct WorkloadDistributor 

pub struct WorkloadDistributor;

Splits a large vector index job across multiple GPU devices proportionally to their memory capacity.

The distributor is stateless: call distribute as many times as needed without side effects.

Implementations

impl WorkloadDistributor

pub fn new() -> Self

Create a new distributor.

pub fn distribute( &self, total_vectors: usize, devices: &[SimpleGpuDevice], ) -> Result<Vec<WorkloadChunk>>

Distribute total_vectors vectors across devices proportionally to each device’s memory_mb.

Returns one WorkloadChunk per device (in device order). Devices with zero memory are skipped. Returns an error if devices is empty or all devices have zero memory.

The last chunk absorbs any rounding remainder so that every vector is covered exactly once.

pub fn distribute_even( &self, total_vectors: usize, devices: &[SimpleGpuDevice], ) -> Result<Vec<WorkloadChunk>>

Distribute evenly (round-robin, ignoring memory ratios).

Useful when all devices are homogeneous. Returns an error if devices is empty.

Trait Implementations

impl Clone for WorkloadDistributor

fn clone(&self) -> WorkloadDistributor

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for WorkloadDistributor

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

Formats the value using the given formatter.

impl Default for WorkloadDistributor

fn default() -> WorkloadDistributor

Returns the “default value” for a type.