Struct GpuLoadBalancer 

pub struct GpuLoadBalancer { /* private fields */ }

Distributes GPU work across multiple devices using a least-loaded strategy.

All mutating operations are thread-safe via an internal Mutex.

Example

use oxirs_vec::gpu::{GpuLoadBalancer, SimpleGpuDevice};

let balancer = GpuLoadBalancer::new();
balancer.register_device(SimpleGpuDevice::new(0, "GPU-0", 8192, 128));
balancer.register_device(SimpleGpuDevice::new(1, "GPU-1", 16384, 256));

if let Some(id) = balancer.select_device(512) {
    balancer.record_workload(id, 512);
    // ... do GPU work ...
    balancer.release_workload(id, 512);
}

Implementations

impl GpuLoadBalancer

pub fn new() -> Self

Create an empty load balancer with no registered devices.

pub fn register_device(&self, device: SimpleGpuDevice)

Register a GPU device. If a device with the same id already exists it is replaced (workload is reset to zero).

pub fn unregister_device(&self, device_id: u32)

Remove a device from the balancer.

pub fn select_device(&self, workload_mb: u64) -> Option<u32>

Select the device best suited to handle workload_mb megabytes of new work.

Returns the id of the device with the lowest current utilisation that has enough free memory to accept the workload, or None if no suitable device exists or no devices are registered.

pub fn record_workload(&self, device_id: u32, mb: u64) -> Result<()>

Record mb megabytes of additional workload on device_id.

Returns an error if device_id is not registered.

pub fn release_workload(&self, device_id: u32, mb: u64) -> Result<()>

Release mb megabytes of workload from device_id.

Clamps to zero to prevent underflow. Returns an error if the device is not registered.

pub fn utilization(&self, device_id: u32) -> Option<f64>

Utilisation of device_id as a fraction in [0.0, 1.0].

Returns None if the device is not registered.

pub fn total_capacity_mb(&self) -> u64

Sum of memory across all registered devices in MB.

pub fn device_count(&self) -> usize

Number of registered devices.

pub fn utilization_snapshot(&self) -> Vec<(u32, f64)>

Returns a snapshot of device IDs and their current utilisation.

Trait Implementations

impl Clone for GpuLoadBalancer

fn clone(&self) -> GpuLoadBalancer

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for GpuLoadBalancer

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

Formats the value using the given formatter.

impl Default for GpuLoadBalancer

fn default() -> Self

Returns the “default value” for a type.